light weight metal corrosion and modeling for corrosion prevention, life prediction and assessment

Light Weight Metal Corrosion and Modeling for Corrosion Prevention, Life Prediction

and Assessment

Selected peer reviewed papers from the 2nd Workshop on Corrosion Modeling for Life Prediction (CMLP 2010),

Rome, Italy, 18 to 20 April 2010, held under the auspices of the Office of Naval Research Global and the Università degli Studi di Milano.

Edited by

Stefano P. Trasatti

Juliet Ippolito

TRANS TECH PUBLICATIONS LTD Switzerland • UK • USA

Copyright 2010 Trans Tech Publications Ltd, Switzerland

All rights reserved. No part of the contents of this publication may be reproduced or transmitted in any form or by any means without the written permission of the publisher.

Trans Tech Publications Ltd Laubisrutistr. 24 CH-8712 Stafa-Zurich Switzerland http://www.ttp.net

Volume 138 of Advanced Materials Research ISSN 1022-6680

Full text available online at http://www.scientific.net

Distributed worldwide by and in the Americas by

Trans Tech Publications Ltd Trans Tech Publications Inc. Laubisrutistr. 24 PO Box 699, May Street CH-8712 Stafa-Zurich Enfield, NH 03748 Switzerland USA

Phone: +1 (603) 632-7377 Fax: +41 (44) 922 10 33 Fax: +1 (603) 632-5611 e-mail: [email protected] e-mail: [email protected]

Preface This volume contains some of the contributions presented at the 2nd Workshop on

Corrosion Modeling for Life Prediction (CMLP 2010), Rome, Italy, 18 to 20 April

2010, held under the auspices of the Office of Naval Research Global and the

Università degli Studi di Milano.

In organizing the workshop every effort was made to invite corrosionists working in

the field of lightweight alloys and dealing with modelling. Their expertise provided a

base to discuss corrosion problems and solutions for Military and Aerospace Systems

and Facilities, thus laying the foundations for tackling still unsolved issues.

The use of lightweight metals and composites to replace heavy structural materials for

military hardware and weapon systems (ships, aircraft, ground vehicles, etc.) is a new

strategic requirement for defence forces, falling under Naval S&T Strategic Plans.

Objectives of the workshop were to seek the state of the art outside the continental

United States in the field of low density metallic materials and composites for

structural applications, as well as in modeling and simulation software tools capable

of generating and identifying damage evolution data for health monitoring, corrosion

control, life prediction and assessment of civil and military hardware systems.

We would like to use this opportunity to gratefully acknowledge the invaluable

contribution, to make the Workshop a success, of all those participating in this

venture, from the organizers to supporting institutions and companies, to speakers and

attendees, as well as to the hotel staff, despite the volcanic eruption occurred in Island

in the very days of this event.

August 2010

Stefano P. Trasatti, Juliet Ippolito

Table of Contents

Sponsors, Acknowledgments, Organizing Committee

Preface

Enhancing the Localized Corrosion Resistance of High Strength 7010 Al-AlloyM.B. Kannan and V.S. Raja 1

Electrochemical Behavior of Nickel-Aluminum Alloys in Sodium Chloride SolutionsK.V. Rybalka, L.A. Beketaeva, V.S. Shaldaev, N.G. Bukhan’ko and A.D. Davydov 7

Characterization of Bronze Corrosion Products on Exposition to Sulphur DioxideB. De Filippo, L. Campanella, A. Brotzu, S. Natali and D. Ferro 21

Electrochemical Methods to Assist Corrosion Control of Lightweight AlloysM. Curioni and G.E. Thompson 29

Surface Protection for Aircraft Maintenanceby Means of Zinc Rich PrimersG. Bockmair and K. Kranzeder 41

Thin, Nanoparticulate Coatings for the Improvement of the Corrosion and PassivationBehavior of AZ Magnesium AlloysF. Feil and W. Fürbeth 47

Electrochemical Characteristics of PEO Treated Electric Arc Coatings on LightweightAlloysH.M. Nykyforchyn, V.I. Pokhmurskii, M.D. Klapkiv, M.M. Student and J. Ippolito 55

Hybrid Coatings Based on Conducting Polymers and Polysiloxane Chains for CorrosionProtection of Al AlloysM. Trueba, S.P. Trasatti and D.O. Flamini 63

A Composite Coating for Corrosion and Wear Protection of AM60B Magnesium AlloyA. Da Forno and M. Bestetti 79

Continuum Damage Model for Biodegradable Magnesium Alloy StentD. Gastaldi, V. Sassi, L. Petrini, M. Vedani, S.P. Trasatti and F. Migliavacca 85

Prediction of Morphological Properties of Smart-Coatings for Cr Replacement, Based onMathematical ModellingB. Bozzini, I. Sgura, D. Lacitignola, C. Mele, M. Marchitto and A. Ciliberto 93

Understanding Nanoscale Wetting Using Dynamic Local Contact Angle MethodM. Losada, K. Mackie, J.H. Osborne and S. Chaudhuri 107

Two-Dimensional Numerical Modelling of Hydrogen Diffusion in Metals Assisted by BothStress and StrainJ. Toribio, V. Kharin, D. Vergara and M. Lorenzo 117

Approach to Iron Corrosion via the Numerical Simulation of a Galvanic CellG. Colicchio, D. Mansutti and M.L. Santarelli 127

Prognostic Tools for Lifetime Prediction of Aircraft Coatings: Paint DegradationJ.M. Colwell, J.H. Khan, G. Will, K.E. Fairfull-Smith, S.E. Bottle, G.A. George and A. Trueman 137

Enhancing the Localized Corrosion Resistance of High Strength

7010 Al-alloy

M. Bobby Kannan 1, a and V.S. Raja 2, b 1Discipline of Chemical Engineering, James Cook University, Townsville 4811, Australia

2 Department of Metallurgical Engineering and Materials Science, Indian Institute of Technology

Bombay, Mumbai, India [email protected],

[email protected]

Keywords: Aluminium alloys; Stress corrosion cracking; Intergranular corrosion; Heat-treatment;

Scandium alloying

Abstract. This paper brings out the developments on heat-treatment and alloying to improve the

stress corrosion cracking (SCC) behavior of 7010 Al-alloy. The role of microstructures including

the grain boundary precipitates and recystallized grains and the relation of intergranular corrosion

(IGC) on the SCC behavior of 7010 Al-alloy have been discussed.

Introduction

Development of high strength Al-alloys continues to be an important area of research due to the

demand for alloys possesing high strength-to-weight ratio for aerospace applications. 7010 Al-alloy

has been recently developed to substitute the conventional 7075 Al-alloy. The low level of

impurities and presence of zirconium in 7010 Al-alloy contribute to its improvement in the

mechanical properties [1]. However, successful applications of this alloy will highly depend on its

localized corrosion behavior. Unfortunately, susceptibility of Al-alloys to stress corrosion cracking

(SCC), a form of localized corrosion, is directly proportional to their strength levels. Hence, SCC

evaluatation of 7010 Al-alloy becomes vital. Puiggali et al. [2] and Robinson [3] have studied the

SCC behavior of 7010 Al-alloys. The first published work by Puiggali et al. [2] reports that an

improvement in the SCC resistance of 7010 Al-alloy can achieved by over aging. However, the

measured ductility of the over aged alloy even in the air tested sample is not appreciable. On other

side, Robinson [3] work, testing carried out under constant load test method on smooth bar samples,

on 7010 forgings subjected to two-step over aging and RRA (RRA retrogression and re-aging)

tempered treatments show that this alloy was susceptible to SCC. Hence, a systematic work was

undertaken in our laboratory to understand the SCC behavior of 7010 Al-alloy and also to enhance

its SCC resistance through novel heat-treatment and alloying [4-11].

Multi-step Heat-treatment

A detailed study on the effect of heat-treatment (multi-step aging) on various tensile properties of

7010 Al-alloy was carried out by the authors [4,5]. The heat-treatment steps involved in the multi-

step aging treatment are shown in Fig. 1. The multi-step aging treatment is advantageous over RRA

treatment by the fact that the former can be applied to even thick plates, while the latter is restricted

to only thin sheets. The data from this work are presented in Table 1. In air, the peak aged alloy

exhibited 10 % elongation (E) and reduction in area (RA) and 561 MPa ultimate tensile strength

(UTS). However, in 3.5 % NaCl solution, the peak aged alloy suffered a significant loss in ductility

and strength. Thus, %E and %RA of the alloy decreased to 3 and the UTS to 515 MPa. On the

contrary, the over aged alloy showed high ductility in air as well as in 3.5% NaCl solution, with

only a 10 % loss of its peak strength. Thus, the over aged alloy exhibited 10%E, 28 %RA and 504

MPa UTS when tested in air. In 3.5% NaCl, the over aged alloy exhibited 10%E, 24 %RA and 491

MPa UTS. Our study notably shows that through multi-step aging a higher percentage elongation of

peak aged and over aged alloys could be achieved than that was reported by Puiggali et al. [1].

Enhancing the Localized Corrosion Resistance of High Strength

7010 Al-alloy

M. Bobby Kannan 1, a and V.S. Raja 2, b 1Discipline of Chemical Engineering, James Cook University, Townsville 4811, Australia

2 Department of Metallurgical Engineering and Materials Science, Indian Institute of Technology

Bombay, Mumbai, India [email protected],

[email protected]

Keywords: Aluminium alloys; Stress corrosion cracking; Intergranular corrosion; Heat-treatment;

Scandium alloying

Abstract. This paper brings out the developments on heat-treatment and alloying to improve the

stress corrosion cracking (SCC) behavior of 7010 Al-alloy. The role of microstructures including

the grain boundary precipitates and recystallized grains and the relation of intergranular corrosion

(IGC) on the SCC behavior of 7010 Al-alloy have been discussed.

Introduction

Development of high strength Al-alloys continues to be an important area of research due to the

demand for alloys possesing high strength-to-weight ratio for aerospace applications. 7010 Al-alloy

has been recently developed to substitute the conventional 7075 Al-alloy. The low level of

impurities and presence of zirconium in 7010 Al-alloy contribute to its improvement in the

mechanical properties [1]. However, successful applications of this alloy will highly depend on its

localized corrosion behavior. Unfortunately, susceptibility of Al-alloys to stress corrosion cracking

(SCC), a form of localized corrosion, is directly proportional to their strength levels. Hence, SCC

evaluatation of 7010 Al-alloy becomes vital. Puiggali et al. [2] and Robinson [3] have studied the

SCC behavior of 7010 Al-alloys. The first published work by Puiggali et al. [2] reports that an

improvement in the SCC resistance of 7010 Al-alloy can achieved by over aging. However, the

measured ductility of the over aged alloy even in the air tested sample is not appreciable. On other

side, Robinson [3] work, testing carried out under constant load test method on smooth bar samples,

on 7010 forgings subjected to two-step over aging and RRA (RRA retrogression and re-aging)

tempered treatments show that this alloy was susceptible to SCC. Hence, a systematic work was

undertaken in our laboratory to understand the SCC behavior of 7010 Al-alloy and also to enhance

its SCC resistance through novel heat-treatment and alloying [4-11].

Multi-step Heat-treatment

A detailed study on the effect of heat-treatment (multi-step aging) on various tensile properties of

7010 Al-alloy was carried out by the authors [4,5]. The heat-treatment steps involved in the multi-

step aging treatment are shown in Fig. 1. The multi-step aging treatment is advantageous over RRA

treatment by the fact that the former can be applied to even thick plates, while the latter is restricted

to only thin sheets. The data from this work are presented in Table 1. In air, the peak aged alloy

exhibited 10 % elongation (E) and reduction in area (RA) and 561 MPa ultimate tensile strength

(UTS). However, in 3.5 % NaCl solution, the peak aged alloy suffered a significant loss in ductility

and strength. Thus, %E and %RA of the alloy decreased to 3 and the UTS to 515 MPa. On the

contrary, the over aged alloy showed high ductility in air as well as in 3.5% NaCl solution, with

only a 10 % loss of its peak strength. Thus, the over aged alloy exhibited 10%E, 28 %RA and 504

MPa UTS when tested in air. In 3.5% NaCl, the over aged alloy exhibited 10%E, 24 %RA and 491

MPa UTS. Our study notably shows that through multi-step aging a higher percentage elongation of

peak aged and over aged alloys could be achieved than that was reported by Puiggali et al. [1].

Fig. 1 Heat-treatment steps of the multi-step aging treatment on 7010 Al-alloy.

Table 1. SSRT data of 7010 Al-alloys in peak aged and over aged conditions tested in air and in

3.5% NaCl solution at 10-6/s strain rate [4].

Temper

% Elongation % Reduction in Area Ultimate Tensile Strength

(MPa)

Air 3.5% NaCl Air 3.5% NaCl Air 3.5% NaCl

Peak aged 10 3 10 3 561 515

Overaged 10 10 28 24 504 491

The reasons for the dependence of SCC susceptibility on the heat treatment were analysed

through fractography. Fractographs of peak aged 7010 Al-alloy showed that the recrystallized grains

were predominantly attacked along the grain boundary (Fig.2 (a)) leading to intergranular cleavage

failure, while the over aged alloy exhibited predominantly ductile failure (Fig.2 (b)) [4,5]. Only the

peak aged alloy was found to be sensitive to cracking along recrystallized grains, although the over

aged alloy also contained recrystallized grains (Fig. 3(a) and (b)). Since the grain boundary area in

peak aged alloy is most susceptible for cracking, it is suggested that the difference in the

morphology and chemistry of grain boundary precipitates (GBPs) in both heat treated conditions are

likely to affect the SCC. To follow this, transmission electron microscope (TEM) studies were

carried out [4]. TEM photographs revealed that in the peak aged alloy, the GBPs were continuously

decorated along the grain boundaries (Fig.4(a)), whereas in the over aged alloy the GBPs were

found to be coarse and disconnected (Fig.4(b)). The GBPs were mainly η particles having the

chemistry of MgZn2 [1]. The precipitates are anodic (based on the TEM-EDX analysis) to the Al

matrix and hence suggested to undergo selective dissolution in corrosive environment [4]. For this

reason, the grain boundaries of peak aged alloy, where η precipitates lie in a continuous manner,

suffered severe cracking in 3.5% NaCl medium. As the grain precipitates were separated in the over

aged condition, the over aged alloy offered more resistance to cracking. The higher Cu of GBPs of

over aged alloy than that of peak aged alloy made MgZn2 precipitates noble and thereby minimized

the dissolution of these precipitates.

Arresting Recrystallization

Since recrystallized grains were found to be the weakest zones of intergranular stress corrosion

cracking (IGSCC), the SCC resistance of Sc containing 7010 Al-alloy was examined [4,7,8], as Sc

is known to inhibit recrystallization in Al-alloys [12-13]. Fig. 5a confirms that addition of

0.25 wt.% Sc to 7010-Al alloy inhibited recrystallization. Interestingly, the alloy containing Sc

showed higher SCC resistance even in the peak aged condition (Table 3). Thus, the alloy exhibited

12.5 %E, 16.4 %RA and 560 MPa UTS when tested in 3.5 % NaCl solution at 10-6/s strain rate.

Fig. 1 Heat-treatment steps of the multi-step aging treatment on 7010 Al-alloy.

Table 1. SSRT data of 7010 Al-alloys in peak aged and over aged conditions tested in air and in

3.5% NaCl solution at 10-6/s strain rate [4].

Temper


(MPa)


Peak aged 10 3 10 3 561 515

Overaged 10 10 28 24 504 491

The reasons for the dependence of SCC susceptibility on the heat treatment were analysed

through fractography. Fractographs of peak aged 7010 Al-alloy showed that the recrystallized grains

were predominantly attacked along the grain boundary (Fig.2 (a)) leading to intergranular cleavage

failure, while the over aged alloy exhibited predominantly ductile failure (Fig.2 (b)) [4,5]. Only the

peak aged alloy was found to be sensitive to cracking along recrystallized grains, although the over

aged alloy also contained recrystallized grains (Fig. 3(a) and (b)). Since the grain boundary area in

peak aged alloy is most susceptible for cracking, it is suggested that the difference in the

morphology and chemistry of grain boundary precipitates (GBPs) in both heat treated conditions are

likely to affect the SCC. To follow this, transmission electron microscope (TEM) studies were

carried out [4]. TEM photographs revealed that in the peak aged alloy, the GBPs were continuously

decorated along the grain boundaries (Fig.4(a)), whereas in the over aged alloy the GBPs were

found to be coarse and disconnected (Fig.4(b)). The GBPs were mainly η particles having the

chemistry of MgZn2 [1]. The precipitates are anodic (based on the TEM-EDX analysis) to the Al

matrix and hence suggested to undergo selective dissolution in corrosive environment [4]. For this

reason, the grain boundaries of peak aged alloy, where η precipitates lie in a continuous manner,

suffered severe cracking in 3.5% NaCl medium. As the grain precipitates were separated in the over

aged condition, the over aged alloy offered more resistance to cracking. The higher Cu of GBPs of

over aged alloy than that of peak aged alloy made MgZn2 precipitates noble and thereby minimized

the dissolution of these precipitates.

Arresting Recrystallization

Since recrystallized grains were found to be the weakest zones of intergranular stress corrosion

cracking (IGSCC), the SCC resistance of Sc containing 7010 Al-alloy was examined [4,7,8], as Sc

is known to inhibit recrystallization in Al-alloys [12-13]. Fig. 5a confirms that addition of

0.25 wt.% Sc to 7010-Al alloy inhibited recrystallization. Interestingly, the alloy containing Sc

showed higher SCC resistance even in the peak aged condition (Table 3). Thus, the alloy exhibited

12.5 %E, 16.4 %RA and 560 MPa UTS when tested in 3.5 % NaCl solution at 10-6/s strain rate.

2 Light Weight Metal Corrosion and Modeling

Comparing with the base alloy, Sc containing alloy exhibited about 4 fold increase in %E and 5 fold

increase in % RA, in spite of the latter exhibiting a 10 % higher UTS than the former. Due to a

sharp reduction in recrystallization, the fracture surface of Sc containing alloy revealed predominant

ductile features (Fig. 5b).

Intergranular Corrosion

In order to understand the relationship between the intergranular corrosion (IGC) suceptibility of

7010 Al-alloys to its SCC suceptibility, the alloys were examined for IGC using ASTM G110-92

standard. Figs.6 (a-c) represent the photographs of the peak aged, over aged and Sc containing peak

aged 7010 Al-alloys after exposure to IGC test solution (4M NaCl, 0.5M KNO3 and 0.1M HNO3 in

distilled water) for 48h. Pits were observed in all the alloys. However, the cross-section analysis of

the alloys revealed various modes of attack (Figs. 7( a-c)). Peak aged alloy underwent high IGC

attack, whereas the overaged alloy showed only marginal suceptibility to IGC and some evidence of

pitting corrosion. As expected, Sc containing peak aged alloy showed significant improvment in

the IGC resistance as compared to the peak aged base alloy. Comparing the SCC and IGC behavior

of 7010-Al-alloys there is a clear indication that when the alloy is susceptibilite to IGC it is pron to

SCC.

Fig. 3 Three-dimensional optical microstructures of (a) peak aged alloy and (b) overaged 7010

Al-alloy shows equiaxed recrystallized grains. L, T and S indicate longitudinal (rolling),

transverse and short-transverse directions respectively [4].

(a) (b)

Fig. 2 SEM fractographs of 7010 Al-alloy in: (a) peak aged condition, shows typical

intergranular cracking of recrystallized grains; and (b) over aged condition, shows

predominant ductile fracture [4].

(a) (b)

Comparing with the base alloy, Sc containing alloy exhibited about 4 fold increase in %E and 5 fold

increase in % RA, in spite of the latter exhibiting a 10 % higher UTS than the former. Due to a

sharp reduction in recrystallization, the fracture surface of Sc containing alloy revealed predominant

ductile features (Fig. 5b).

Intergranular Corrosion

In order to understand the relationship between the intergranular corrosion (IGC) suceptibility of

7010 Al-alloys to its SCC suceptibility, the alloys were examined for IGC using ASTM G110-92

standard. Figs.6 (a-c) represent the photographs of the peak aged, over aged and Sc containing peak

aged 7010 Al-alloys after exposure to IGC test solution (4M NaCl, 0.5M KNO3 and 0.1M HNO3 in

distilled water) for 48h. Pits were observed in all the alloys. However, the cross-section analysis of

the alloys revealed various modes of attack (Figs. 7( a-c)). Peak aged alloy underwent high IGC

attack, whereas the overaged alloy showed only marginal suceptibility to IGC and some evidence of

pitting corrosion. As expected, Sc containing peak aged alloy showed significant improvment in

the IGC resistance as compared to the peak aged base alloy. Comparing the SCC and IGC behavior

of 7010-Al-alloys there is a clear indication that when the alloy is susceptibilite to IGC it is pron to

SCC.

Fig. 3 Three-dimensional optical microstructures of (a) peak aged alloy and (b) overaged 7010

Al-alloy shows equiaxed recrystallized grains. L, T and S indicate longitudinal (rolling),

transverse and short-transverse directions respectively [4].

(a) (b)

Fig. 2 SEM fractographs of 7010 Al-alloy in: (a) peak aged condition, shows typical

intergranular cracking of recrystallized grains; and (b) over aged condition, shows

predominant ductile fracture [4].

(a) (b)

Advanced Materials Research Vol. 138 3

Table 3 SSRT data of base 7010 Al-alloy and 0.25 wt.% Sc containing alloy in peak aged conditions

tested in air and 3.5% NaCl solution at 10-6/s strain rate [4].

Conclusions

Recrystallized grains in 7010 Al-alloy were found to be more susceptible towards intergranular

SCC. Preferential dissolution of MgZn2 precipitates assisted by its continuous decoration along the

grain boundary makes the grain boundary an easy path for crack growth. Both the above conditions

were modified by over aging treatment, but with about 10% loss in UTS. As the recrystallized

grains were the weakest areas of cracking in peak aged base alloy, through inhibiting

Alloy


(MPa)


Base 10 3 10 3 561 515

0.25 wt.%

Sc added

13.4

12.5

15.8

16.4

560

560

(b)

Fig. 4 TEM micrographs of 7010 Al-alloy in: (a) peak aged condition, shows fine

precipitates in the matrix and continuous grain boundary precipitates ; and (b) over aged

condition, shows coarse precipitates in the matrix and broken network of coarse grain

boundary precipitates [4].

(a) (b)

Fig. 5 Micrographs of 0.25 wt.% Sc containing 7010 Al-alloy: (a) Optical micrograph

shows fibrous non-recrystallized grains and (b) SEM fractograph reveals predominant

ducitle failure [4].

(a) (b)

Table 3 SSRT data of base 7010 Al-alloy and 0.25 wt.% Sc containing alloy in peak aged conditions

tested in air and 3.5% NaCl solution at 10-6/s strain rate [4].

Conclusions

Recrystallized grains in 7010 Al-alloy were found to be more susceptible towards intergranular

SCC. Preferential dissolution of MgZn2 precipitates assisted by its continuous decoration along the

grain boundary makes the grain boundary an easy path for crack growth. Both the above conditions

were modified by over aging treatment, but with about 10% loss in UTS. As the recrystallized

grains were the weakest areas of cracking in peak aged base alloy, through inhibiting

Alloy


(MPa)


Base 10 3 10 3 561 515

0.25 wt.%

Sc added

13.4

12.5

15.8

16.4

560

560

(b)

Fig. 4 TEM micrographs of 7010 Al-alloy in: (a) peak aged condition, shows fine

precipitates in the matrix and continuous grain boundary precipitates ; and (b) over aged

condition, shows coarse precipitates in the matrix and broken network of coarse grain

boundary precipitates [4].

(a) (b)

Fig. 5 Micrographs of 0.25 wt.% Sc containing 7010 Al-alloy: (a) Optical micrograph

shows fibrous non-recrystallized grains and (b) SEM fractograph reveals predominant

ducitle failure [4].

(a) (b)


recrystallization by Sc addition the SCC resistance was improved significantly. Further, it was

found that the alloy suceptibile to IGC is also susceptible to SCC.

Fig. 7 Cross-section views of IGC tested 7010 Al-alloy in (a) peak aged, (b)

over aged, and (c) 0.25 wt.% Sc containing peak aged shows various degree of

intergranular corrosion.

(a) (b)

(c)

Fig. 6 Photographs of IGC tested 7010 Al-alloy in (a) peak aged, (b) over aged

and (c) 0.25 wt.% Sc containing peak aged shows pitting corrosion.

(a) (b) (c)

recrystallization by Sc addition the SCC resistance was improved significantly. Further, it was

found that the alloy suceptibile to IGC is also susceptible to SCC.

Fig. 7 Cross-section views of IGC tested 7010 Al-alloy in (a) peak aged, (b)

over aged, and (c) 0.25 wt.% Sc containing peak aged shows various degree of

intergranular corrosion.

(a) (b)

(c)

Fig. 6 Photographs of IGC tested 7010 Al-alloy in (a) peak aged, (b) over aged

and (c) 0.25 wt.% Sc containing peak aged shows pitting corrosion.

(a) (b) (c)


References

[1] A.K. Mukhopadhyay, G.M. Reddy, K.S. Prasad, S.V. Kamat, A. Dutta, C.Mondal: J.T.Staley

Honorary Symposium on Al Alloys, Advances in the Metallurgy of Al Alloys, ASM

International (Indianapolis, USA, November 5-8, 2001), p.63.

[2] M. Puiggali, A. Zielinski, J.M. Olive, E. Renauld, D. Desjardins, M. Cid: Corro. Sci. 40 (1998),

p.805.

[3] J.S.Robinson: Mater. Sci. Forum 331-337 (2000), p.1653.

[4] M. Bobby Kannan: Ph.D. thesis, Indian Institute of Technology Bombay, India, May 2005.

[5] M. Bobby Kannan, V.S. Raja, R. Raman, A .K. Mukhopadhyay: Corrosion 59 (2003), p. 881.

[6] M. Bobby Kannan, V.S. Raja, A .K. Mukhopadhyay: Scripta Mater. 51 (2004), p.1075.

[7] M. Bobby Kannan, V.S. Raja, A .K. Mukhopadhyay, P. Schmuki: Metall. Mater. Trans. A 36

(2005), p.3257.

[8] M. Bobby Kannan, V.S. Raja: Engineering Fracture Mechanics 77 (2010), p.249.

[9] M. Bobby Kannan, V.S. Raja: J. Mater. Sci. 42 (2007), p.5458.

[10] M. Bobby Kannan, V.S. Raja: Advances in Mater. Sci. 7 (2007), p. 21.


[12] L.A. Willey: US Patent 3619181, 1971.

[13] Y.W. Riddle, T.H. Sanders Jr.: Mater. Sci. Forum 331-337 (2000), p.799.

References

[1] A.K. Mukhopadhyay, G.M. Reddy, K.S. Prasad, S.V. Kamat, A. Dutta, C.Mondal: J.T.Staley

Honorary Symposium on Al Alloys, Advances in the Metallurgy of Al Alloys, ASM

International (Indianapolis, USA, November 5-8, 2001), p.63.

[2] M. Puiggali, A. Zielinski, J.M. Olive, E. Renauld, D. Desjardins, M. Cid: Corro. Sci. 40 (1998),

p.805.

[3] J.S.Robinson: Mater. Sci. Forum 331-337 (2000), p.1653.

[4] M. Bobby Kannan: Ph.D. thesis, Indian Institute of Technology Bombay, India, May 2005.

[5] M. Bobby Kannan, V.S. Raja, R. Raman, A .K. Mukhopadhyay: Corrosion 59 (2003), p. 881.

[6] M. Bobby Kannan, V.S. Raja, A .K. Mukhopadhyay: Scripta Mater. 51 (2004), p.1075.

[7] M. Bobby Kannan, V.S. Raja, A .K. Mukhopadhyay, P. Schmuki: Metall. Mater. Trans. A 36

(2005), p.3257.

[8] M. Bobby Kannan, V.S. Raja: Engineering Fracture Mechanics 77 (2010), p.249.


[10] M. Bobby Kannan, V.S. Raja: Advances in Mater. Sci. 7 (2007), p. 21.


[12] L.A. Willey: US Patent 3619181, 1971.

[13] Y.W. Riddle, T.H. Sanders Jr.: Mater. Sci. Forum 331-337 (2000), p.799.


Electrochemical Behavior of Nickel-Aluminum Alloys in Sodium Chloride Solutions

Konstantin V. Rybalka1,a, Luiza A. Beketaeva1,a , Vyacheslav S. Shaldaev1,a, Nataliya G. Bukhan’ko2,b and Alexey D. Davydov1,c

1Frumkin Institute of Physical Chemistry and Electrochemistry, Russian Academy of Sciences, Leninskii pr. 31, Moscow, 119991 Russia

2Department of Chemistry, Moscow State University, Moscow, Russia

aemail: [email protected], bemail: [email protected], cemail: [email protected]

Keywords: nickel-aluminum alloys, sodium chloride solutions, general corrosion, pitting corrosion

Abstract. The anodic and cathodic reactions involved in the corrosion process on several nickel-

aluminum alloys including two intermetallic compounds NiAl and Ni3Al in the NaCl solutions are

studied. A procedure of pretreatment of test specimens and measuring the anodic and cathodic

voltammograms is developed. It enabled us to obtain reproducible results including Tafel portions

of voltammograms. The corrosion potentials Ecorr and corrosion currents icorr are determined by the

coordinates of the intersection of anodic and cathodic Tafel plots. The dependences of Ecorr and icorr

on the alloy composition (the content of nickel in the binary nickel-aluminum alloys), on the

concentration of NaCl, and рН of unbuffered NaCl solutions with the additions of HCl or NaOH are

determined.

The anodic behavior of the alloys in a wide potential range is studied using the potentiodynamic

method and the method of stepwise raising anodic potential with an exposure of electrode at each

potential for a certain time. The dependences of pitting potential on the concentration of solution are

determined for two intermetallic compounds.

Introduction

Corrosion modeling is one of rapidly developing branches of corrosion science. There are various

approaches to simulating corrosion processes (see [1-6] and literature cited therein).

In some cases, the computational programs are developed, which enable one to calculate the

corrosion potentials Ecorr and corrosion current densities icorr using known kinetic parameters of

anodic and cathodic reactions (the coefficients in the Tafel equation, the exchange current, etc.)

involved in the corrosion process. It is better, if the formalized experimental dependences of Ecorr

and icorr on the solution composition and corrosion conditions are available. The computational

programs should be supplemented with the corresponding database. The larger is the database and

more precise are the data, the more valuable are the programs for calculating the corrosion rate.

Alternative approaches determine the electrochemical measurements, which should be

performed (for example, to measure the variation in the current caused by the variation of potential

imposed), and relate the experimental results to icorr by using a certain model (for example, using a

certain equivalent circuit of corrosion system). The fitting process of the input and the output is

frequently used. The application of these models is commonly limited.

In all cases, the estimation of icorr requires a large number of experiments. The corrosion

experiment is the most time-consuming and expensive part of corrosion study, even if relatively

rapid electrochemical methods are used to determine the corrosion rate.

Moreover, there are some problems associated with determining the corrosion rate using relatively

rapid electrochemical methods.

In the course of electrochemical experiments, a potential is imposed or a current is passed

through the model electrochemical system using an external power source, commonly, a

potentiostat/galvanostat. This can lead to irreversible changes in the system; hence, the results of

electrochemical measurements will not correspond to the natural corrosion conditions of test metal.

Electrochemical Behavior of Nickel-Aluminum Alloys in Sodium Chloride Solutions

Konstantin V. Rybalka1,a, Luiza A. Beketaeva1,a , Vyacheslav S. Shaldaev1,a, Nataliya G. Bukhan’ko2,b and Alexey D. Davydov1,c

1Frumkin Institute of Physical Chemistry and Electrochemistry, Russian Academy of Sciences, Leninskii pr. 31, Moscow, 119991 Russia

2Department of Chemistry, Moscow State University, Moscow, Russia

aemail: [email protected], bemail: [email protected], cemail: [email protected]

Keywords: nickel-aluminum alloys, sodium chloride solutions, general corrosion, pitting corrosion

Abstract. The anodic and cathodic reactions involved in the corrosion process on several nickel-

aluminum alloys including two intermetallic compounds NiAl and Ni3Al in the NaCl solutions are

studied. A procedure of pretreatment of test specimens and measuring the anodic and cathodic

voltammograms is developed. It enabled us to obtain reproducible results including Tafel portions

of voltammograms. The corrosion potentials Ecorr and corrosion currents icorr are determined by the

coordinates of the intersection of anodic and cathodic Tafel plots. The dependences of Ecorr and icorr

on the alloy composition (the content of nickel in the binary nickel-aluminum alloys), on the

concentration of NaCl, and рН of unbuffered NaCl solutions with the additions of HCl or NaOH are

determined.

The anodic behavior of the alloys in a wide potential range is studied using the potentiodynamic

method and the method of stepwise raising anodic potential with an exposure of electrode at each

potential for a certain time. The dependences of pitting potential on the concentration of solution are

determined for two intermetallic compounds.

Introduction

Corrosion modeling is one of rapidly developing branches of corrosion science. There are various

approaches to simulating corrosion processes (see [1-6] and literature cited therein).

In some cases, the computational programs are developed, which enable one to calculate the

corrosion potentials Ecorr and corrosion current densities icorr using known kinetic parameters of

anodic and cathodic reactions (the coefficients in the Tafel equation, the exchange current, etc.)

involved in the corrosion process. It is better, if the formalized experimental dependences of Ecorr

and icorr on the solution composition and corrosion conditions are available. The computational

programs should be supplemented with the corresponding database. The larger is the database and

more precise are the data, the more valuable are the programs for calculating the corrosion rate.

Alternative approaches determine the electrochemical measurements, which should be

performed (for example, to measure the variation in the current caused by the variation of potential

imposed), and relate the experimental results to icorr by using a certain model (for example, using a

certain equivalent circuit of corrosion system). The fitting process of the input and the output is

frequently used. The application of these models is commonly limited.

In all cases, the estimation of icorr requires a large number of experiments. The corrosion

experiment is the most time-consuming and expensive part of corrosion study, even if relatively

rapid electrochemical methods are used to determine the corrosion rate.

Moreover, there are some problems associated with determining the corrosion rate using relatively

rapid electrochemical methods.

In the course of electrochemical experiments, a potential is imposed or a current is passed

through the model electrochemical system using an external power source, commonly, a

potentiostat/galvanostat. This can lead to irreversible changes in the system; hence, the results of

electrochemical measurements will not correspond to the natural corrosion conditions of test metal.

To reduce this drawback of electrochemical measurements, the researchers try to conduct

them in the vicinity of the corrosion potential, to apply various relaxation methods, and to perform

the measurement for a short time, which also lead to the difficulties in interpreting the experimental

results concerning their applicability to the free-corrosion conditions. The difficulties, in particular,

can be associated with the fact that the steady-state corrosion process is not reached.

To check the applicability of a model, the results, which are obtained using a model, are

compared with the corrosion rate data from solution analysis and from weight loss of the specimens.

These nonelectrochemical experiments are also commonly rather short (they take several hours). In

the literature, usually, the examples are presented, when the electrochemical methods proposed give

the results, which agree well with the results of other methods. However, this cannot assure well

agreement in other cases.

The most frequently used electrochemical method for determining the corrosion current and

corrosion potential in the case of uniform corrosion is as follows: the corrosion current and

corrosion potential are determined by the coordinates of intersection point of extended anodic and

cathodic Tafel portions of voltammogram. However, even in this rather simple method, not

everything is quite clear.

To obtain reproducible results, it is necessary to prepare properly the test specimen surface

prior to the electrochemical experiment. Various methods of surface pretreatment are used.

The mechanical treatment is used to remove the surface layers, which differ from the bulk

metal in their composition. As a result of this treatment, some mechanical defects are formed in the

surface layer.

As a result of electrochemical polishing, a film containing oxygen and species of polishing

electrolyte is formed.

The cathodic reduction in order to remove spontaneously formed oxide films from the

surface is a good method that provides reproducible results of electrochemical measurements.

However, the corrosion behavior of these pretreated specimens can differ significantly from that of

commercial parts.

The initial conditions of electrochemical experiments are different for different methods of

pretreatment leading to a discrepancy between the measured results.

The corrosion rate depends on the surface state of test specimen. We are interested in the

corrosion rate of specimen, when its surface state is similar to that at the free-corrosion potential in

the real corrosion conditions. In the electrochemical studies, the metal potential is varied by using a

potentiostat (or similar equipment) in order to obtain anodic and cathodic voltammograms.

Possibly, this changes the surface state of test specimen and, hence, its corrosion rate as compared

with real corrosion conditions, when the corrosion potential is reached spontaneously as a result of

the interaction of the specimen surface with the corrosion medium, with various components of

medium, for example, water, which is a passivator, and chloride ion, which is well-known activator.

The voltammograms not necessarily contain the Tafel portions, i.e. the linear curve portions

on the overpotential vs. the logarithm of current density coordinates. Then, the problem is analyzed

by using alternative methods (see, for example, [1, 2]).

Great difficulties emerge in numerous and practically important cases, when a certain period

of general, uniform corrosion, so-called induction period tind, is followed by the local, pitting

corrosion. Here, addition problems arise. It is difficult to determine the induction period for the

corrosion conditions by using the electrochemical methods, because the measurements of induction

period as a function of electrode potential, concentration and temperature of aggressive solution,

and passive film thickness, were performed, at best, for hours [7-13] (in exceptional cases, for

hundreds hours [14], which is too short time as compared with real corrosion conditions. The

extrapolation of these results to longer period of time can lead to large errors in the determined tind.

Then, how can the possibility of pitting corrosion be determined using the electrochemical

measurements? There are many works, where the values of pitting potential Epit of various metals

and alloys in the solutions of various compositions and concentrations measured under various

experimental conditions are reported. In most cases, Epit is determined by the anodic

To reduce this drawback of electrochemical measurements, the researchers try to conduct

them in the vicinity of the corrosion potential, to apply various relaxation methods, and to perform

the measurement for a short time, which also lead to the difficulties in interpreting the experimental

results concerning their applicability to the free-corrosion conditions. The difficulties, in particular,

can be associated with the fact that the steady-state corrosion process is not reached.

To check the applicability of a model, the results, which are obtained using a model, are

compared with the corrosion rate data from solution analysis and from weight loss of the specimens.

These nonelectrochemical experiments are also commonly rather short (they take several hours). In

the literature, usually, the examples are presented, when the electrochemical methods proposed give

the results, which agree well with the results of other methods. However, this cannot assure well

agreement in other cases.

The most frequently used electrochemical method for determining the corrosion current and

corrosion potential in the case of uniform corrosion is as follows: the corrosion current and

corrosion potential are determined by the coordinates of intersection point of extended anodic and

cathodic Tafel portions of voltammogram. However, even in this rather simple method, not

everything is quite clear.

To obtain reproducible results, it is necessary to prepare properly the test specimen surface

prior to the electrochemical experiment. Various methods of surface pretreatment are used.

The mechanical treatment is used to remove the surface layers, which differ from the bulk

metal in their composition. As a result of this treatment, some mechanical defects are formed in the

surface layer.

As a result of electrochemical polishing, a film containing oxygen and species of polishing

electrolyte is formed.

The cathodic reduction in order to remove spontaneously formed oxide films from the

surface is a good method that provides reproducible results of electrochemical measurements.

However, the corrosion behavior of these pretreated specimens can differ significantly from that of

commercial parts.

The initial conditions of electrochemical experiments are different for different methods of

pretreatment leading to a discrepancy between the measured results.

The corrosion rate depends on the surface state of test specimen. We are interested in the

corrosion rate of specimen, when its surface state is similar to that at the free-corrosion potential in

the real corrosion conditions. In the electrochemical studies, the metal potential is varied by using a

potentiostat (or similar equipment) in order to obtain anodic and cathodic voltammograms.

Possibly, this changes the surface state of test specimen and, hence, its corrosion rate as compared

with real corrosion conditions, when the corrosion potential is reached spontaneously as a result of

the interaction of the specimen surface with the corrosion medium, with various components of

medium, for example, water, which is a passivator, and chloride ion, which is well-known activator.

The voltammograms not necessarily contain the Tafel portions, i.e. the linear curve portions

on the overpotential vs. the logarithm of current density coordinates. Then, the problem is analyzed

by using alternative methods (see, for example, [1, 2]).

Great difficulties emerge in numerous and practically important cases, when a certain period

of general, uniform corrosion, so-called induction period tind, is followed by the local, pitting

corrosion. Here, addition problems arise. It is difficult to determine the induction period for the

corrosion conditions by using the electrochemical methods, because the measurements of induction

period as a function of electrode potential, concentration and temperature of aggressive solution,

and passive film thickness, were performed, at best, for hours [7-13] (in exceptional cases, for

hundreds hours [14], which is too short time as compared with real corrosion conditions. The

extrapolation of these results to longer period of time can lead to large errors in the determined tind.

Then, how can the possibility of pitting corrosion be determined using the electrochemical

measurements? There are many works, where the values of pitting potential Epit of various metals

and alloys in the solutions of various compositions and concentrations measured under various

experimental conditions are reported. In most cases, Epit is determined by the anodic


potentiodynamic curves or, using the step-by-step raising of anodic potential, before the onset of

steep increase of the anodic current, which corresponds to the breakdown of metal passivity. An

important problem in determining Epit as applied to the corrosion conditions is that, when the

potential is changed from Eoc to Epit in the potentiodynamic experiments and using step-by step

raising of potential, as a rule, the surface state of test specimen is significantly changed under the

action of external power source. Certainly, there is no assurance that similar changes take place

under real corrosion conditions. Therefore, the electrochemical measurements can give inadequate

values of Epit, and the conclusion about the possibility of pitting corrosion can be invalid.

The danger of pitting corrosion is difficult to predict also due to specific features of the stage

of pitting development.

The pitting corrosion can develop by various ways. A large number of pits can form

immediately and, then, they can grow rapidly in width and slowly in depth. In another limiting case,

it is important to determine the rate of pit deepening. This is complicated by the fact that the rate of

pit deepening varies in the time [15]; in addition, some pits deepen, initially, rapidly, and, after a

certain time, they virtually stop to grow, and some of later formed pits become deeper than the

earlier formed pits (see, for example, [16]). The heterogeneity of alloy, the presence of excessive

phases (sulfides, carbides, etc.), which promote the local dissolution, is also of significance. This

can be important not only at the stage of breakdown of metal passivity, at the initial stage of pitting

corrosion, but also at the stage of pitting development, deepening of individual pits.

In the majority of works, the pitting development is studied under the potentiostatic

conditions, at a potential higher than the free-corrosion potential. This method takes a shorted time

and the results are more reproducible; however, its relation to real corrosion is frequently unknown.

In the experimental part of this work, certainly, we cannot overcome all aforementioned

difficulties. However, we performed the experimental study with regard for these difficulties and

took into consideration the relativity of measured results.

The electrochemical behavior of several binary Ni-Al alloys of various compositions in the

NaCl solutions is studied.

In the literature, there are the works devoted to the anodic behavior of binary alloys of nickel

with other metals at various ratios between the components, for example, Ni-Cr [17, 18] and

Ni-Ti [19] alloys. We have found no systematic studies of electrochemical behavior of Ni-Al alloys.

Only fragmentary information on this subject is available [20, 21].

Experimental Procedure

The specimens of 50 at. % Ni – 50 at. % Al alloy were produced from aluminum (99.999%) and electrolytic

nickel in an electric arc furnace in the argon atmosphere. Zirconium was used as a getter in the melting. The

ingots were subjected to the homogenizing annealing for 1 month at a temperature 550ºС with quenching in

the ice water. This regime of annealing was chosen in accordance with the phase diagram of the Al-Ni

system [22].

The phase composition of the alloys was determined by the XRD method using a DRON-4

diffractometer (CuKα- radiation). The data identification was carried out using the STOE program. Figure 1

gives the phase composition of the alloy.

A horizontal disk 1.32 cm2 in area was used as the test electrode. Immediately prior to the

experiment, the electrode surface was polished with fine emery paper, degreased with alcohol, and

rinsed with twice-distilled water. All measurements were performed with an IPC 2000 compact

potentiostat.

In the work, unbuffered NaCl solutions are used, because they are most similar to the real

media. Therefore, the results of all experiments in the solutions with various рН values refer to the

pH values in the bulk solution.

potentiodynamic curves or, using the step-by-step raising of anodic potential, before the onset of

steep increase of the anodic current, which corresponds to the breakdown of metal passivity. An

important problem in determining Epit as applied to the corrosion conditions is that, when the

potential is changed from Eoc to Epit in the potentiodynamic experiments and using step-by step

raising of potential, as a rule, the surface state of test specimen is significantly changed under the

action of external power source. Certainly, there is no assurance that similar changes take place

under real corrosion conditions. Therefore, the electrochemical measurements can give inadequate

values of Epit, and the conclusion about the possibility of pitting corrosion can be invalid.

The danger of pitting corrosion is difficult to predict also due to specific features of the stage

of pitting development.

The pitting corrosion can develop by various ways. A large number of pits can form

immediately and, then, they can grow rapidly in width and slowly in depth. In another limiting case,

it is important to determine the rate of pit deepening. This is complicated by the fact that the rate of

pit deepening varies in the time [15]; in addition, some pits deepen, initially, rapidly, and, after a

certain time, they virtually stop to grow, and some of later formed pits become deeper than the

earlier formed pits (see, for example, [16]). The heterogeneity of alloy, the presence of excessive

phases (sulfides, carbides, etc.), which promote the local dissolution, is also of significance. This

can be important not only at the stage of breakdown of metal passivity, at the initial stage of pitting

corrosion, but also at the stage of pitting development, deepening of individual pits.

In the majority of works, the pitting development is studied under the potentiostatic

conditions, at a potential higher than the free-corrosion potential. This method takes a shorted time

and the results are more reproducible; however, its relation to real corrosion is frequently unknown.

In the experimental part of this work, certainly, we cannot overcome all aforementioned

difficulties. However, we performed the experimental study with regard for these difficulties and

took into consideration the relativity of measured results.

The electrochemical behavior of several binary Ni-Al alloys of various compositions in the

NaCl solutions is studied.

In the literature, there are the works devoted to the anodic behavior of binary alloys of nickel

with other metals at various ratios between the components, for example, Ni-Cr [17, 18] and

Ni-Ti [19] alloys. We have found no systematic studies of electrochemical behavior of Ni-Al alloys.

Only fragmentary information on this subject is available [20, 21].

Experimental Procedure

The specimens of 50 at. % Ni – 50 at. % Al alloy were produced from aluminum (99.999%) and electrolytic

nickel in an electric arc furnace in the argon atmosphere. Zirconium was used as a getter in the melting. The

ingots were subjected to the homogenizing annealing for 1 month at a temperature 550ºС with quenching in

the ice water. This regime of annealing was chosen in accordance with the phase diagram of the Al-Ni

system [22].

The phase composition of the alloys was determined by the XRD method using a DRON-4

diffractometer (CuKα- radiation). The data identification was carried out using the STOE program. Figure 1

gives the phase composition of the alloy.

A horizontal disk 1.32 cm2 in area was used as the test electrode. Immediately prior to the

experiment, the electrode surface was polished with fine emery paper, degreased with alcohol, and

rinsed with twice-distilled water. All measurements were performed with an IPC 2000 compact

potentiostat.

In the work, unbuffered NaCl solutions are used, because they are most similar to the real

media. Therefore, the results of all experiments in the solutions with various рН values refer to the

pH values in the bulk solution.


Fig. 1. The phase composition of the alloys by the results of XRD method: (a) Ni50Al50;

(b) Ni66Al34; (c) Ni75Al25; (d) Ni95Al5 (solid solution of Al in Ni).

To estimate the corrosion rate of alloys (the corrosion current densities icorr) in the NaCl

solutions with various concentrations and pH values, the anodic and cathodic voltammograms were

measured, the Tafel portions in these curves were revealed, and the coordinates of the intersection

point between the extended Tafel portions were determined.

Prior to measuring the voltammograms, the test specimens were held in the NaCl solutions

up to reaching a constant value of open-circuit potential Eoc.. Then, the anodic voltammogram was

measured from Eoc. in the direction of higher anodic potentials up to the value, which enabled us to

obtain an anodic Tafel curve portion. The measurements were performed in the potentiodynamic

mode with a potential scan rate of 10-3

V/s. Then, the potential was switched-off, and the test

specimen was held in the solution, until the value of Eoc. became equal to that, from which the

a)

b)

c)

d)

Fig. 1. The phase composition of the alloys by the results of XRD method: (a) Ni50Al50;

(b) Ni66Al34; (c) Ni75Al25; (d) Ni95Al5 (solid solution of Al in Ni).

To estimate the corrosion rate of alloys (the corrosion current densities icorr) in the NaCl

solutions with various concentrations and pH values, the anodic and cathodic voltammograms were

measured, the Tafel portions in these curves were revealed, and the coordinates of the intersection

point between the extended Tafel portions were determined.

Prior to measuring the voltammograms, the test specimens were held in the NaCl solutions

up to reaching a constant value of open-circuit potential Eoc.. Then, the anodic voltammogram was

measured from Eoc. in the direction of higher anodic potentials up to the value, which enabled us to

obtain an anodic Tafel curve portion. The measurements were performed in the potentiodynamic

mode with a potential scan rate of 10-3

V/s. Then, the potential was switched-off, and the test

specimen was held in the solution, until the value of Eoc. became equal to that, from which the

a)

b)

c)

d)


measurement of the anodic voltammogram was started. Thereafter, the cathodic voltammogram

was measured from the same Eoc. value under similar potentiodynamic conditions. This procedure

enabled us to obtain well reproducible anodic and cathodic Tafel plots and to use them for

determining the corrosion potential Ecorr and the corrosion current density icorr. The results of

experiments (Ecorr and icorr), in which, first, the cathodic Tafel portion and, then, the anodic one

were obtained, virtually coincided with the results of above experiments.

The experiments were performed in the NaCl solutions at various pH values. The required

pH values were obtained by adding the corresponding amounts of HCl or NaOH.

Experimental Results and Discussion

The open-circuit potential. By the example of Ni50Al50 alloy, Fig. 2 shows the variation of

the open-circuit potential Eoc. with the time after the mechanical polishing of specimen surface and

immersing it into 0.01 M NaCl solution. In the neutral and weakly alkaline (pH 4 – 10) solutions,

Eoc. shifts to less negative values. This is commonly related to the self-passivation as a result of

prevailing interaction of metal surface with water. In the solutions with pH 2 and 2.5, the potential

shifts to more negative values. This can be associated with specimen surface activation under the

joint action of Cl- and H

+ ions.

Fig. 2. The variation of the open-circuit potential Eoc. with the time after the mechanical polishing

of specimen surface and immersing it into 0.01 M NaCl solution with various pH values: (1) 11,

(2) 10, (3) 8, (4) 7, (5) 4, (6) 3, (7) 2, and (8) 2.

Figure 3 gives the dependences of Eoc. on the pH value for several Ni-Al alloys with various

ratios between the components and for nickel and aluminum. The Eoc. values were measured after

an hour exposure of specimens in 0.01 M NaCl solution. For all alloys, the open-circuit potentials

are close to that of nickel. The aluminum potentials, as would be expected, are much more negative.

measurement of the anodic voltammogram was started. Thereafter, the cathodic voltammogram

was measured from the same Eoc. value under similar potentiodynamic conditions. This procedure

enabled us to obtain well reproducible anodic and cathodic Tafel plots and to use them for

determining the corrosion potential Ecorr and the corrosion current density icorr. The results of

experiments (Ecorr and icorr), in which, first, the cathodic Tafel portion and, then, the anodic one

were obtained, virtually coincided with the results of above experiments.

The experiments were performed in the NaCl solutions at various pH values. The required

pH values were obtained by adding the corresponding amounts of HCl or NaOH.

Experimental Results and Discussion

The open-circuit potential. By the example of Ni50Al50 alloy, Fig. 2 shows the variation of

the open-circuit potential Eoc. with the time after the mechanical polishing of specimen surface and

immersing it into 0.01 M NaCl solution. In the neutral and weakly alkaline (pH 4 – 10) solutions,

Eoc. shifts to less negative values. This is commonly related to the self-passivation as a result of

prevailing interaction of metal surface with water. In the solutions with pH 2 and 2.5, the potential

shifts to more negative values. This can be associated with specimen surface activation under the

joint action of Cl- and H

+ ions.

Fig. 2. The variation of the open-circuit potential Eoc. with the time after the mechanical polishing

of specimen surface and immersing it into 0.01 M NaCl solution with various pH values: (1) 11,

(2) 10, (3) 8, (4) 7, (5) 4, (6) 3, (7) 2, and (8) 2.

Figure 3 gives the dependences of Eoc. on the pH value for several Ni-Al alloys with various

ratios between the components and for nickel and aluminum. The Eoc. values were measured after

an hour exposure of specimens in 0.01 M NaCl solution. For all alloys, the open-circuit potentials

are close to that of nickel. The aluminum potentials, as would be expected, are much more negative.


Different characters of time dependences of Eoc. at different pH values indicates that the

corrosion rates of an alloy of a given composition can be different in the solutions with different

acidity.

Fig. 3. The pH dependences of open-circuit potential Eoc. for (1) nickel and (2) Ni95Al5,

(3) Ni75Al25, (4) Ni66Al34, (5) Ni50Al50 alloys, and (6) Al after an hour exposure of specimens in

0.01 M NaCl solution.

Anodic and cathodic behavior of alloys near the corrosion potential. Determination of

corrosion potential and corrosion current density. Figure 4 gives the anodic and cathodic

voltammograms on the Tafel coordinates. The voltammograms were measured on nickel and three

alloys with various contents of nickel in 0.5 M NaCl solution. By extrapolating the Tafel curve

portions to their intersection point, the corrosion potentials and corrosion currents were determined.

It should be noted that the used procedure of obtaining anodic and cathodic Tafel plots

provides the conditions, which allow one to determine Ecorr and icorr by extrapolating only one –

anodic or cathodic – Tafel curve portion.

In the solutions with low pH values (pH 2 and 3), we failed to obtain well-defined linear

curve portions on the Tafel coordinates in the anodic voltammogram. We always managed to obtain

the Tafel plot by subtracting the cathodic current, which is determined by the cathodic Tafel curve

portion extended in the anodic direction from the total current in the initial (adjacent to Ecorr)

segment of anodic voltammogram (with regard for opposite signs of anodic and cathodic currents).

This procedure is based on the fact that at the potentials around Ecorr, the cathodic and anodic

voltammograms are distorted by the anodic and cathodic reactions proceeding simultaneously with

close rates.

Different characters of time dependences of Eoc. at different pH values indicates that the

corrosion rates of an alloy of a given composition can be different in the solutions with different

acidity.

Fig. 3. The pH dependences of open-circuit potential Eoc. for (1) nickel and (2) Ni95Al5,

(3) Ni75Al25, (4) Ni66Al34, (5) Ni50Al50 alloys, and (6) Al after an hour exposure of specimens in

0.01 M NaCl solution.

Anodic and cathodic behavior of alloys near the corrosion potential. Determination of

corrosion potential and corrosion current density. Figure 4 gives the anodic and cathodic

voltammograms on the Tafel coordinates. The voltammograms were measured on nickel and three

alloys with various contents of nickel in 0.5 M NaCl solution. By extrapolating the Tafel curve

portions to their intersection point, the corrosion potentials and corrosion currents were determined.

It should be noted that the used procedure of obtaining anodic and cathodic Tafel plots

provides the conditions, which allow one to determine Ecorr and icorr by extrapolating only one –

anodic or cathodic – Tafel curve portion.

In the solutions with low pH values (pH 2 and 3), we failed to obtain well-defined linear

curve portions on the Tafel coordinates in the anodic voltammogram. We always managed to obtain

the Tafel plot by subtracting the cathodic current, which is determined by the cathodic Tafel curve

portion extended in the anodic direction from the total current in the initial (adjacent to Ecorr)

segment of anodic voltammogram (with regard for opposite signs of anodic and cathodic currents).

This procedure is based on the fact that at the potentials around Ecorr, the cathodic and anodic

voltammograms are distorted by the anodic and cathodic reactions proceeding simultaneously with

close rates.


Fig. 4. Tafel portions of anodic and cathodic voltammograms measured on (a) nickel and

(b) Ni95Al5, (c) Ni66Al34, (d) Ni50Al50 alloys in 0.5 M NaCl solution.

Table 1. Corrosion potentials and currents of Ni50Al50 alloy in 0.5 M NaCl solution after a five-hour

exposure of electrode in the solutions with various pH values.

рН icorr [µA/cm2] Ecorr [V]

2.5 10.5 -0.368

3 14.8 -0.359

4 4.5 -0.237

6 0.44 -0.068

8 0.48 -0.105

11 0.33 -0.085

12 0.35 -0.075

Fig. 4. Tafel portions of anodic and cathodic voltammograms measured on (a) nickel and

(b) Ni95Al5, (c) Ni66Al34, (d) Ni50Al50 alloys in 0.5 M NaCl solution.

Table 1. Corrosion potentials and currents of Ni50Al50 alloy in 0.5 M NaCl solution after a five-hour

exposure of electrode in the solutions with various pH values.

рН icorr [µA/cm2] Ecorr [V]

2.5 10.5 -0.368

3 14.8 -0.359

4 4.5 -0.237

6 0.44 -0.068

8 0.48 -0.105

11 0.33 -0.085

12 0.35 -0.075


Table 1 lists the data for Ni50Al50 alloy in 0.5 M NaCl solution with various pH values.

Figure 5 gives the data for several alloys, nickel and aluminum in 0.01 M NaCl solution.

The character of dependence of corrosion rate (icorr) of aluminum on the pH value is well

known [23]: the corrosion rate is almost identical in the solutions with pH of 4 to 10 and steeply

increases in the acidic and alkaline solutions (Fig. 5). Nickel and all alloys studied exhibit low

corrosion rates in the neutral and alkaline solutions and significantly higher corrosion rates in the

acidic solutions (Fig. 5 and Table 1). For example, from Table 1 it is seen that the corrosion current

density of Ni50Al50 alloy in the neutral and alkaline 0.5 M NaCl solutions is about 0.5 µA/cm2, and

in the solutions with pH 2.5 and 3, it is about 10 µA/cm2.

The corrosion currents of nickel-aluminum alloys are intermediate between those of

individual components: nickel and aluminum. The values of icorr monotonically decrease with

increasing content of nickel in the alloy (Fig. 6). The corrosion potentials of all alloys studied are

approximately equal to Ecorr of nickel, i.e. they are much more positive than Ecorr of aluminum.

Fig. 5. The dependences of corrosion current density icorr on the pH value of 0.01 M NaCl solutions

for (1) Ni50Al50, (2) Ni66Al34, (3) Ni75Al25, and (4) Ni95Al5, alloys, (5) nickel, and (6) aluminum.

Table 1 lists the data for Ni50Al50 alloy in 0.5 M NaCl solution with various pH values.

Figure 5 gives the data for several alloys, nickel and aluminum in 0.01 M NaCl solution.

The character of dependence of corrosion rate (icorr) of aluminum on the pH value is well

known [23]: the corrosion rate is almost identical in the solutions with pH of 4 to 10 and steeply

increases in the acidic and alkaline solutions (Fig. 5). Nickel and all alloys studied exhibit low

corrosion rates in the neutral and alkaline solutions and significantly higher corrosion rates in the

acidic solutions (Fig. 5 and Table 1). For example, from Table 1 it is seen that the corrosion current

density of Ni50Al50 alloy in the neutral and alkaline 0.5 M NaCl solutions is about 0.5 µA/cm2, and

in the solutions with pH 2.5 and 3, it is about 10 µA/cm2.

The corrosion currents of nickel-aluminum alloys are intermediate between those of

individual components: nickel and aluminum. The values of icorr monotonically decrease with

increasing content of nickel in the alloy (Fig. 6). The corrosion potentials of all alloys studied are

approximately equal to Ecorr of nickel, i.e. they are much more positive than Ecorr of aluminum.

Fig. 5. The dependences of corrosion current density icorr on the pH value of 0.01 M NaCl solutions

for (1) Ni50Al50, (2) Ni66Al34, (3) Ni75Al25, and (4) Ni95Al5, alloys, (5) nickel, and (6) aluminum.


Fig. 6. The dependences of (a) corrosion potential and (b) current density on the alloy

composition in 0.5 M NaCl solution.

Fig. 7. The dependence of cathodic current on Ni50Al50 alloy on the pH of 0.5 M NaCl

solution at a potential, which is by 0.08 V more negative than Eoc.

Fig. 6. The dependences of (a) corrosion potential and (b) current density on the alloy

composition in 0.5 M NaCl solution.

Fig. 7. The dependence of cathodic current on Ni50Al50 alloy on the pH of 0.5 M NaCl

solution at a potential, which is by 0.08 V more negative than Eoc.


Figure 7 gives the dependence of cathodic current, which was measured at a potential by

0.08 V more negative than Eoc, on the pH of 0.5 M NaCl solution. An abrupt increase in the cathodic

current is observed approximately in the same range of pH values, where the corrosion current

steeply increases (Fig. 5). In the acidified solutions, the reduction of dissolved oxygen is

supplemented by the reduction of H+ cations. The lower is the pH value of solution, the higher is the

contribution of the latter reaction to the total cathodic current. The change of the cathodic process in

the acidified solutions can be one of the reasons for an increase of the corrosion current.

Anodic behavior, pitting potential. The anodic voltammograms were measured using two

methods: the potentiodynamic method and stepwise increase of anodic potential with an exposure

of specimen at each potential for a certain time. Both methods are frequently used in the studies of

pitting corrosion. In different series of experiments, the voltammograms were measured with no

preliminary exposure of test specimen at Eoc in the same solution or after a preliminary exposure up

to reaching virtually constant value of Eoc. By way of example, Fig. 8 gives an anodic

potentiodynamic (a potential scan rate of 5 10-4

V/s) voltammogram, which was measured on

Ni50Al50 alloy in 0.1 M NaCl solution (the direct and reverse branches) with no preliminary

exposure. In the voltammogram, a passivity region is observed, which is limited by an abrupt

increase of the anodic current due to the breakdown of alloy passivity with aggressive Cl- ions at the

pitting potential Epit. The reverse branch of voltammogram enables us to estimate the alloy

repassivation potential Erp.

The anodic voltammograms measured in the acidified solutions (Fig. 9) differ significantly

from those obtained in the neutral solutions: they contain no passivity region, the current starts to

increase steeply immediately from Eoc. Thus, an increase in the corrosion current density in the

acidified solutions (Fig. 5) is associated with an increase in the rates of both cathodic (Fig. 7) and

anodic (Fig. 9) reactions.

Fig. 8. The anodic potentiodynamic curve, which was measured on Ni50Al50 alloy in

0.1 M NaCl solution: (1) and (2) are the direct branch (2 is a continuation of 1) and (3) is the reverse

branches.

Figure 7 gives the dependence of cathodic current, which was measured at a potential by

0.08 V more negative than Eoc, on the pH of 0.5 M NaCl solution. An abrupt increase in the cathodic

current is observed approximately in the same range of pH values, where the corrosion current

steeply increases (Fig. 5). In the acidified solutions, the reduction of dissolved oxygen is

supplemented by the reduction of H+ cations. The lower is the pH value of solution, the higher is the

contribution of the latter reaction to the total cathodic current. The change of the cathodic process in

the acidified solutions can be one of the reasons for an increase of the corrosion current.

Anodic behavior, pitting potential. The anodic voltammograms were measured using two

methods: the potentiodynamic method and stepwise increase of anodic potential with an exposure

of specimen at each potential for a certain time. Both methods are frequently used in the studies of

pitting corrosion. In different series of experiments, the voltammograms were measured with no

preliminary exposure of test specimen at Eoc in the same solution or after a preliminary exposure up

to reaching virtually constant value of Eoc. By way of example, Fig. 8 gives an anodic

potentiodynamic (a potential scan rate of 5 10-4

V/s) voltammogram, which was measured on

Ni50Al50 alloy in 0.1 M NaCl solution (the direct and reverse branches) with no preliminary

exposure. In the voltammogram, a passivity region is observed, which is limited by an abrupt

increase of the anodic current due to the breakdown of alloy passivity with aggressive Cl- ions at the

pitting potential Epit. The reverse branch of voltammogram enables us to estimate the alloy

repassivation potential Erp.

The anodic voltammograms measured in the acidified solutions (Fig. 9) differ significantly

from those obtained in the neutral solutions: they contain no passivity region, the current starts to

increase steeply immediately from Eoc. Thus, an increase in the corrosion current density in the

acidified solutions (Fig. 5) is associated with an increase in the rates of both cathodic (Fig. 7) and

anodic (Fig. 9) reactions.


0.1 M NaCl solution: (1) and (2) are the direct branch (2 is a continuation of 1) and (3) is the reverse

branches.



0.1 M NaCl solution with pH 3 due to an addition of HCl (the direct and reverse branches).

Figure 10 gives the dependences of potentials Epit and Erp on the logarithm of concentration

of NaCl solution (the experiments with no preliminary exposure of test specimen). Both potentials

linearly decrease with increasing log C, which is typical for the plots of Epit vs. log C for other

metals and alloys (see, for example, [24-26]).

Fig. 10. Dependences of potentials (1) Epit and (2) Erp of Ni50Al50 alloy on the logarithm of

concentration of NaCl solution.


0.1 M NaCl solution with pH 3 due to an addition of HCl (the direct and reverse branches).

Figure 10 gives the dependences of potentials Epit and Erp on the logarithm of concentration

of NaCl solution (the experiments with no preliminary exposure of test specimen). Both potentials

linearly decrease with increasing log C, which is typical for the plots of Epit vs. log C for other

metals and alloys (see, for example, [24-26]).

Fig. 10. Dependences of potentials (1) Epit and (2) Erp of Ni50Al50 alloy on the logarithm of

concentration of NaCl solution.


The potential Epit, which was determined by the same procedure, in 0.01 M NaCl solution

was virtually (within the limits of experimental error) independent of the alloy composition: in the

alloys with nickel content from 50 to 95 at. %, it was 0.07 ±0.02 V, which corresponds to Epit for

nickel under similar conditions.

Figure 11 gives the anodic voltammogram, which was measured by increasing stepwise the

potential on Ni75Al25 alloy in 0.5 M NaCl solution. The dependences of Epit on the log C, which were

obtained by two methods: (1) the stepwise increase of potential and (2) the potentiodynamic method

are compared on Fig. 12. It is seen that the values of Epit are significantly higher in the first case.

In the potentiodynamic method of determining Epit, its value can variously depend on the

potential scan rate: with increasing potential scan rate Epit can increase [24, 27-29], decrease [30,

31] or be independent of scan rate [30, 32]. The dependence of Epit on the potential scan rate can

have a complex character [33]. In the rather simple case that Epit linearly depends on the potential

scan rate, it can be extrapolated to zero scan rate. However, even in this case, the questions of

pretreatment of specimens and other aforementioned peculiarities of electrochemical methods of

investigating pitting corrosion remain open.

Frequently, the researchers attempt to measure “close to the steady-state values” by using

the method of stepwise increase of potential with an exposure of specimen at each potential for a

certain time. However, in this case, before reaching the potential of passivity breakdown, an anodic

oxide film, which possesses protective properties, grows on the electrode, and high experimental

values of Epit are obtained. The presence of similar protective anodic oxide film on a part under

natural corrosion conditions is not assured.

Fig. 11. The anodic characteristic, which was measured by increasing stepwise the potential of

Ni75Al25 alloy in 0.5 M NaCl solution. The values of potential, V: (1) -0.10; (2) -0.08; (3) -0.06;

(4) -0.04; (5) -0.02; (6) -0.01; (7) 0.0; (8) 0.01. The specimen was preliminary held in the same

solution for 2.5 h.

The potential Epit, which was determined by the same procedure, in 0.01 M NaCl solution

was virtually (within the limits of experimental error) independent of the alloy composition: in the

alloys with nickel content from 50 to 95 at. %, it was 0.07 ±0.02 V, which corresponds to Epit for

nickel under similar conditions.

Figure 11 gives the anodic voltammogram, which was measured by increasing stepwise the

potential on Ni75Al25 alloy in 0.5 M NaCl solution. The dependences of Epit on the log C, which were

obtained by two methods: (1) the stepwise increase of potential and (2) the potentiodynamic method

are compared on Fig. 12. It is seen that the values of Epit are significantly higher in the first case.

In the potentiodynamic method of determining Epit, its value can variously depend on the

potential scan rate: with increasing potential scan rate Epit can increase [24, 27-29], decrease [30,

31] or be independent of scan rate [30, 32]. The dependence of Epit on the potential scan rate can

have a complex character [33]. In the rather simple case that Epit linearly depends on the potential

scan rate, it can be extrapolated to zero scan rate. However, even in this case, the questions of

pretreatment of specimens and other aforementioned peculiarities of electrochemical methods of

investigating pitting corrosion remain open.

Frequently, the researchers attempt to measure “close to the steady-state values” by using

the method of stepwise increase of potential with an exposure of specimen at each potential for a

certain time. However, in this case, before reaching the potential of passivity breakdown, an anodic

oxide film, which possesses protective properties, grows on the electrode, and high experimental

values of Epit are obtained. The presence of similar protective anodic oxide film on a part under

natural corrosion conditions is not assured.

Fig. 11. The anodic characteristic, which was measured by increasing stepwise the potential of

Ni75Al25 alloy in 0.5 M NaCl solution. The values of potential, V: (1) -0.10; (2) -0.08; (3) -0.06;

(4) -0.04; (5) -0.02; (6) -0.01; (7) 0.0; (8) 0.01. The specimen was preliminary held in the same

solution for 2.5 h.


Fig. 12. The dependences of Epit on the log C, which were obtained on Ni75Al25 alloy in 0.5 M NaCl

solution by two methods: the stepwise increase of potential (squares) and the potentiodynamic

method (circles).

In all cases, it is better to orient to the lowest experimental values of Epit. For example, from

the above results it is seen that on Ni50Al50 alloy, the potentials Eoc, which are reached in several

(5-10) hours in the 0.5 and 1.0 M NaCl solutions, fall into a potential range between Epit and Erp,

which were determined potentiodynamically with no preliminary exposure of electrode at Eoc. This

indicates that the pitting corrosion of alloy is possible in these solutions.

Conclusions

Based on the results of electrochemical measurements, the corrosion potentials and

corrosion current densities are estimated for several nickel-aluminum alloys in the NaCl solutions

as functions of the ratio between the components in the binary alloys, concentration and pH value of

unbuffered solutions. The dependences of potential of alloy passivity breakdown (pitting potential)

on the concentration of NaCl are obtained. The possibility of pitting corrosion of alloys is

considered.

ACKNOWLEDGMENTS

This work was supported by the fundamental research program “New Approaches to Enhancing

Corrosion Resistance, Radiation Hardness of Materials, and Radiation Environment Safety”,

Division of Chemistry and Material Science, Russian Academy of Sciences.

Fig. 12. The dependences of Epit on the log C, which were obtained on Ni75Al25 alloy in 0.5 M NaCl

solution by two methods: the stepwise increase of potential (squares) and the potentiodynamic

method (circles).

In all cases, it is better to orient to the lowest experimental values of Epit. For example, from

the above results it is seen that on Ni50Al50 alloy, the potentials Eoc, which are reached in several

(5-10) hours in the 0.5 and 1.0 M NaCl solutions, fall into a potential range between Epit and Erp,

which were determined potentiodynamically with no preliminary exposure of electrode at Eoc. This

indicates that the pitting corrosion of alloy is possible in these solutions.

Conclusions

Based on the results of electrochemical measurements, the corrosion potentials and

corrosion current densities are estimated for several nickel-aluminum alloys in the NaCl solutions

as functions of the ratio between the components in the binary alloys, concentration and pH value of

unbuffered solutions. The dependences of potential of alloy passivity breakdown (pitting potential)

on the concentration of NaCl are obtained. The possibility of pitting corrosion of alloys is

considered.

ACKNOWLEDGMENTS

This work was supported by the fundamental research program “New Approaches to Enhancing

Corrosion Resistance, Radiation Hardness of Materials, and Radiation Environment Safety”,

Division of Chemistry and Material Science, Russian Academy of Sciences.


References

[1] V. Feliu, J.A. Gonzales and S. Feliu: Corros. Sci. Vol. 49 (2007), p. 3241.

[2] H.J. Flitt and D.P. Schweinsberg: Corros. Sci. Vol. 47 (2005), p. 3034.

[3] V.S. Beleevskii, K.A. Konev, V.V. Novosadov and V.Yu. Vasil’ev: Protection of Metals Vol. 40 (2004), p. 566.

[4] Y.-T. Zhao, X.-P. Guo, H.-H. Li and Z.-H. Dong: Corros. Sci. Vol. 48 (2006), p. 2913.

[5] E. McCafferty: Corros. Sci. Vol. 47 (2005), p. 3202.

[6] F. Mansfeld: Corros. Sci. Vol. 47 (2005), p. 3178.

[7] Aa Broli, H. Holtan and T.B. Andreassen: Werkst. und Korros. Vol. 27 (1976), p. 497.

[8] M. Janik-Chachor: Werkst. und Korros. Vol. 30 (1979), p. 255.

[9] H.-J. Engell and N.D. Stolica: Z. Phys. Chem. N.F. Vol. 20 (1959), p.113.

[10] T.P. Hoar and W.R. Jacobs: Nature Vol. 216 (1967), p. 1299.

[11] K.E. Heusler and L. Fischer: Werkst. und Korros. Vol. 27 (1976), p. 551.

[12] H.J. Engell and N.D. Stolica: Arch. Eisenhuttenwes. Vol. 30 (1959), p. 239.

[13] A.D. Davydov, V.S. Shaldaev and G.R. Engel'gardt: Russian J. Electrochem. Vol. 42 (2006), p. 121.

[14] J.-H. Wang, C.C. Su and Z. Szklarska-Smialowska: Corrosion Vol. 44 (1988), p.732.

[15] A.D. Davydov: Russian J Electrochem. Vol. 44 (2008), p. 835.

[16] K.V. Rybalka, V.S. Shaldaev, L.A. Beketaeva, A.N. Malofeeva and A.D. Davydov: Russian J Electrochem. Vol. 46 (2010), p. 196.

[17] A.D. Davydov, E.N. Kiriyak and V.D. Kashcheev: Soviet Electrochem. Vol. 14 (1978), p. 352.

[18] T. Jabs, P. Borthen and H.-H. Strehblow: J. Electrochem. Soc. Vol. 144 (1997), p. 1231.

[19] L.M. Glukhov, N.G. Bukhan’ko and A.D. Davydov: Russian J Electrochem. Vol. 44 (2008), p. 332.

[20] Z. Zhang, E. Akiyama, Y. Watanabe, Y. Katada and K. Tsuzaki: Corros. Sci. Vol. 49 (2007), p. 2962.

[21] B.B. Rodrigues and A.W. Hassel: J. Electrochem. Soc. Vol. 155 (2008), p. K31-K37.

[22] M.F. Singleton, J.L. Murray and P. Nash, in: Binary Alloy Phase Diagrams, Second Edition, edited by T.B. Massalski, p. 181-184, ASM International, Materials Park, Ohio 1 (1990).

[23] H. Kaesche: Die Korrosion der Metalle (Springer-Verlag, Berlin 1979).

[24] K.-S. Lei, D.D. Macdonald, B.G. Pound and B.E.Wilde: J. Electrochem. Soc. Vol. 135 (1988), p.1625.

[25] L. Organ, J.R. Scully, A.S. Mikhailov and J. Hudson: Electrochim. Acta Vol. 51 (2005), p. 225.


[27] T. Szauer and J. Jakobs: Corros. Sci. Vol. 16 (1976), p.945.

[28] Aa Broli and H. Holtan: Corros. Sci. Vol. 13 (1973), p.237.

[29] P. Leckie and H.H. Uhlig: J. Electrochem. Soc. Vol. 113 (1966), p.1261.

[30] K. Sugimoto, S. Matsuda, Y. Ogiwara. And K. Kitamura: J. Electrochem. Soc. Vol. 132 (1985), p. 1791.

[31] A.P. Bond and E.A. Lizlovs: J. Electrochem. Soc. Vol. 115 (1968), p. 1130.

[32] W. Schwenk: Corros. Sci. Vol. 5 (1965), p. 245.

[33] B. Baroux: Corros. Sci. Vol. 28 (1988), p. 969.

References

[1] V. Feliu, J.A. Gonzales and S. Feliu: Corros. Sci. Vol. 49 (2007), p. 3241.

[2] H.J. Flitt and D.P. Schweinsberg: Corros. Sci. Vol. 47 (2005), p. 3034.

[3] V.S. Beleevskii, K.A. Konev, V.V. Novosadov and V.Yu. Vasil’ev: Protection of Metals Vol. 40 (2004), p. 566.

[4] Y.-T. Zhao, X.-P. Guo, H.-H. Li and Z.-H. Dong: Corros. Sci. Vol. 48 (2006), p. 2913.

[5] E. McCafferty: Corros. Sci. Vol. 47 (2005), p. 3202.

[6] F. Mansfeld: Corros. Sci. Vol. 47 (2005), p. 3178.

[7] Aa Broli, H. Holtan and T.B. Andreassen: Werkst. und Korros. Vol. 27 (1976), p. 497.


[9] H.-J. Engell and N.D. Stolica: Z. Phys. Chem. N.F. Vol. 20 (1959), p.113.

[10] T.P. Hoar and W.R. Jacobs: Nature Vol. 216 (1967), p. 1299.

[11] K.E. Heusler and L. Fischer: Werkst. und Korros. Vol. 27 (1976), p. 551.

[12] H.J. Engell and N.D. Stolica: Arch. Eisenhuttenwes. Vol. 30 (1959), p. 239.

[13] A.D. Davydov, V.S. Shaldaev and G.R. Engel'gardt: Russian J. Electrochem. Vol. 42 (2006), p. 121.

[14] J.-H. Wang, C.C. Su and Z. Szklarska-Smialowska: Corrosion Vol. 44 (1988), p.732.

[15] A.D. Davydov: Russian J Electrochem. Vol. 44 (2008), p. 835.

[16] K.V. Rybalka, V.S. Shaldaev, L.A. Beketaeva, A.N. Malofeeva and A.D. Davydov: Russian J Electrochem. Vol. 46 (2010), p. 196.

[17] A.D. Davydov, E.N. Kiriyak and V.D. Kashcheev: Soviet Electrochem. Vol. 14 (1978), p. 352.

[18] T. Jabs, P. Borthen and H.-H. Strehblow: J. Electrochem. Soc. Vol. 144 (1997), p. 1231.

[19] L.M. Glukhov, N.G. Bukhan’ko and A.D. Davydov: Russian J Electrochem. Vol. 44 (2008), p. 332.

[20] Z. Zhang, E. Akiyama, Y. Watanabe, Y. Katada and K. Tsuzaki: Corros. Sci. Vol. 49 (2007), p. 2962.

[21] B.B. Rodrigues and A.W. Hassel: J. Electrochem. Soc. Vol. 155 (2008), p. K31-K37.

[22] M.F. Singleton, J.L. Murray and P. Nash, in: Binary Alloy Phase Diagrams, Second Edition, edited by T.B. Massalski, p. 181-184, ASM International, Materials Park, Ohio 1 (1990).

[23] H. Kaesche: Die Korrosion der Metalle (Springer-Verlag, Berlin 1979).

[24] K.-S. Lei, D.D. Macdonald, B.G. Pound and B.E.Wilde: J. Electrochem. Soc. Vol. 135 (1988), p.1625.

[25] L. Organ, J.R. Scully, A.S. Mikhailov and J. Hudson: Electrochim. Acta Vol. 51 (2005), p. 225.


[27] T. Szauer and J. Jakobs: Corros. Sci. Vol. 16 (1976), p.945.

[28] Aa Broli and H. Holtan: Corros. Sci. Vol. 13 (1973), p.237.

[29] P. Leckie and H.H. Uhlig: J. Electrochem. Soc. Vol. 113 (1966), p.1261.

[30] K. Sugimoto, S. Matsuda, Y. Ogiwara. And K. Kitamura: J. Electrochem. Soc. Vol. 132 (1985), p. 1791.

[31] A.P. Bond and E.A. Lizlovs: J. Electrochem. Soc. Vol. 115 (1968), p. 1130.

[32] W. Schwenk: Corros. Sci. Vol. 5 (1965), p. 245.

[33] B. Baroux: Corros. Sci. Vol. 28 (1988), p. 969.


Characterization of bronze corrosion products on exposition to sulphur dioxide

B. De Filippo*, L. Campanella*, A. Brotzu**, S. Natali**, D. Ferro***

*“La Sapienza” University of Rome, dep . of Chemistry , p.le Aldo Moro 5, Rome ** La Sapienza” University of Rome, dep . ICMA, via Eudossiana 18, Rome *** CNR-ISMN, Co dep of Chemistry, P.le A. Moro 5, 00185 Roma (Italia)

Abstract

In the main frame of the research aimed to model the corrosion growth on bronze surface, the

objective of the work here reported has been to characterize the corrosion products formed on

laboratory samples of bronze alloy (Cu Sn12), during the early stage of exposure to moist air with

sulfur dioxide. A cycling corrosion cabinet was used to control 200 ppm gas concentration, relative

humidity (RH) and temperature, according to the DIN 50018 (Kesternich test).The method is

designed to evaluate how well the surface resists to sulfur dioxide corrosion; the test cycle consists

of 8 hours exposure to sulfur dioxide at 40°C temperature and 100% relative humidity, followed by

12 hours drying at room condition.

Weight variation, Spectrophotometer, Scanning Electron Microscopy with X-ray microanalysis

(SEM-EDS), X-ray Diffraction (XRD) analysis were carried out for the tarnish products

characterization. Some of the compound identified were brochantite (Cu4(OH)6SO4), chalcanthite

(CuSO4·5H2O) cuprite (Cu2O), cassiterite (SnO2) and ottemannite (Sn2O3).

Keywords: bronze, X-ray diffraction, colorimetric measurements, atmospheric corrosion,

Introduction

In nature most of the oxidation process of bronze metal works exposed in environmental condition

are electrochemical and involve interaction between the metal surface, the adsorbed moisture and

various atmospheric gases (SO2, CO2, NOx, hydrocarbons and so on).

Focusing the attention on the interaction between copper and sulfur dioxide, it is well known that

the presence of small amounts of SO2 emissions can accelerate corrosion factor of bronze exposed

in urban and industrial areas [1-3]

The reaction behavior of bronze corrosion is often assumed to be similar as that of pure copper,

however the alloying elements strongly modify the bronze corrosion behavior [4-6]. Tarnish

products formed on bronze during the exposition have been investigated and characterized, as

reported in the recent literature [7-9], even if the electrochemical corrosion process are still not well

understood. Robbiola and other [10-12] have deeply studied the mechanism of patinas formation

and the role of cycling action of acid rain on the patina composition. In particular they have pointed

out the different patinas formed on outdoor bronze metalwork as a function of the exposure

geometry [6].

Considering the role of the relative humidity, a significant corrosion rate increase was investigated

at 70-75% relative humidity (RH) in the atmosphere containing SO2. According to the studies of

Tomashov and others [13-17] the critical humidity was associated with the formation of a

continuum condensated moisture layer on the metal surface.

The aim of the present works, which is the first step of a more complex research, is to investigate

the tarnish copper compounds developed on bronze alloys exposed to sulfur dioxide and submitted

to cycling wet and dry tests at 200 ppm, 40°C and 100% RH. The evolution of each corrosion

process have been monitored using Weight variation, Spectrophotometer, Scanning Electron

Microscopy with X-ray microanalysis (SEM-EDS), X-ray Diffraction (XRD).

Characterization of bronze corrosion products on exposition to sulphur dioxide

B. De Filippo*, L. Campanella*, A. Brotzu**, S. Natali**, D. Ferro***

*“La Sapienza” University of Rome, dep . of Chemistry , p.le Aldo Moro 5, Rome ** La Sapienza” University of Rome, dep . ICMA, via Eudossiana 18, Rome *** CNR-ISMN, Co dep of Chemistry, P.le A. Moro 5, 00185 Roma (Italia)

Abstract

In the main frame of the research aimed to model the corrosion growth on bronze surface, the

objective of the work here reported has been to characterize the corrosion products formed on

laboratory samples of bronze alloy (Cu Sn12), during the early stage of exposure to moist air with

sulfur dioxide. A cycling corrosion cabinet was used to control 200 ppm gas concentration, relative

humidity (RH) and temperature, according to the DIN 50018 (Kesternich test).The method is

designed to evaluate how well the surface resists to sulfur dioxide corrosion; the test cycle consists

of 8 hours exposure to sulfur dioxide at 40°C temperature and 100% relative humidity, followed by

12 hours drying at room condition.

Weight variation, Spectrophotometer, Scanning Electron Microscopy with X-ray microanalysis

(SEM-EDS), X-ray Diffraction (XRD) analysis were carried out for the tarnish products

characterization. Some of the compound identified were brochantite (Cu4(OH)6SO4), chalcanthite

(CuSO4·5H2O) cuprite (Cu2O), cassiterite (SnO2) and ottemannite (Sn2O3).

Keywords: bronze, X-ray diffraction, colorimetric measurements, atmospheric corrosion,

Introduction

In nature most of the oxidation process of bronze metal works exposed in environmental condition

are electrochemical and involve interaction between the metal surface, the adsorbed moisture and

various atmospheric gases (SO2, CO2, NOx, hydrocarbons and so on).

Focusing the attention on the interaction between copper and sulfur dioxide, it is well known that

the presence of small amounts of SO2 emissions can accelerate corrosion factor of bronze exposed

in urban and industrial areas [1-3]

The reaction behavior of bronze corrosion is often assumed to be similar as that of pure copper,

however the alloying elements strongly modify the bronze corrosion behavior [4-6]. Tarnish

products formed on bronze during the exposition have been investigated and characterized, as

reported in the recent literature [7-9], even if the electrochemical corrosion process are still not well

understood. Robbiola and other [10-12] have deeply studied the mechanism of patinas formation

and the role of cycling action of acid rain on the patina composition. In particular they have pointed

out the different patinas formed on outdoor bronze metalwork as a function of the exposure

geometry [6].

Considering the role of the relative humidity, a significant corrosion rate increase was investigated

at 70-75% relative humidity (RH) in the atmosphere containing SO2. According to the studies of

Tomashov and others [13-17] the critical humidity was associated with the formation of a

continuum condensated moisture layer on the metal surface.

The aim of the present works, which is the first step of a more complex research, is to investigate

the tarnish copper compounds developed on bronze alloys exposed to sulfur dioxide and submitted

to cycling wet and dry tests at 200 ppm, 40°C and 100% RH. The evolution of each corrosion

process have been monitored using Weight variation, Spectrophotometer, Scanning Electron

Microscopy with X-ray microanalysis (SEM-EDS), X-ray Diffraction (XRD).

The results will be used as input data for the validation of a mathematical model, that simulate

corrosion process on bronze surfaces as a function of specific environmental condition.

Materials and Methods

The cast bronze specimens have been realized starting from pure elements. Copper and tin has been

melted in an electric furnace a 1150 °C with borax protective slag. The molten metal has been cast

in a graphite die, analyzed with Energy Dispersion Scansion analysis in order to verify its chemical

composition and then homogenized in an electric furnace at 600 °C for 72 hours (air cooling) in

order to obtain monophase alpha bronze microstructure.

The cast bronze chemical composition is: Copper 87.9 %w, Tin 12.1 %w; it has been chosen as a

representative alloy used in the past (ancient Greek type).

Corrosion test has been carried out in a Erichsen Mod. 519/AUTO cyclic corrosion cabinet

following the indication of DIN 50018 standard. This standard describes a wet and dry corrosion

test designed in order to evaluate the resistance of metallic surface to sulphur dioxide corrosion.

Bronze specimens have been exposed to an atmosphere containing about 200 ppm of SO2 at 40°C

and 100% RH for 8 hours (wet cycle), subsequently they have been exposed to room condition for

16 hours (dry cycle). Each wet and dry cycle has been repeated 20 times. Specimens have been

weighted at the beginning of each cycle.

Bronze corroded specimens have been characterized employing several analytical techniques.

Particularly the morphology and the microstructure of the bronze corroded surface has been

investigated with Scanning Electron Microscopy and XRD analysis. SEM observations and X-ray

Energy Dispersion Scansion analysis have been carried out both on the surface and on the section of

the specimens. In this way we were able to identify the elements characterizing the corrosion patina,

to highlight the morphology of corrosion products and to measure patina thickness. XRD analysis

have been carried out only on the specimens surface in order to identify the corrosion products.

Corroded surfaces have been also characterized by colorimetric CIELab non destructive

measurements. CIELab test analyses the spectrum of the light reflected by surfaces, through an

instrument constituted by a spectrophotometer. The detected signals may be represented as a graph

with the wavelength (nm) on the horizontal axis and the reflected light percentage (%) on the

vertical axis. The graph can be elaborated in order to obtained the brightness of the reflected light

brightens (L*) and two different parameters (a* and b*) which quantitatively define the reflected

light putting it in the tridimensional CIELab color space. L* is always positive and it represents the

light brightness (a black surface give L=0 while a white surface give L*=100); a* is the

redness/greenness ratio and a* define the position on the redness-greenness axis (a* is positive in

the red region, negative in the green region) and b* is the ratio blueness/yellowness and define the

position on the blueness – yellowness axis (b* is positive in the yellow region and negative in the

blue region). If reflected light is characterized by a*=0 and b*= 0, it means that all light wavelength

light components are present and the light will be white if L*=100 and lack if L*=0. Patinas with

comparable colorimetric measurements are probably formed by the same copper compounds.

Results and discussion

Corrosion products characterization

Figure 1 reports the weight variation measured during sulphur dioxide corrosion test. It can be

clearly seen that during the first 5 cycles all specimens exposed to a sulphur dioxide-moist ambient

show a continuous weight increment. From the fifth up to the tenth cycle a strong weight decrease

as been observed. After the tenth cycle the weight decrease is more slow.

The results will be used as input data for the validation of a mathematical model, that simulate

corrosion process on bronze surfaces as a function of specific environmental condition.


The cast bronze specimens have been realized starting from pure elements. Copper and tin has been

melted in an electric furnace a 1150 °C with borax protective slag. The molten metal has been cast

in a graphite die, analyzed with Energy Dispersion Scansion analysis in order to verify its chemical

composition and then homogenized in an electric furnace at 600 °C for 72 hours (air cooling) in

order to obtain monophase alpha bronze microstructure.

The cast bronze chemical composition is: Copper 87.9 %w, Tin 12.1 %w; it has been chosen as a

representative alloy used in the past (ancient Greek type).

Corrosion test has been carried out in a Erichsen Mod. 519/AUTO cyclic corrosion cabinet

following the indication of DIN 50018 standard. This standard describes a wet and dry corrosion

test designed in order to evaluate the resistance of metallic surface to sulphur dioxide corrosion.

Bronze specimens have been exposed to an atmosphere containing about 200 ppm of SO2 at 40°C

and 100% RH for 8 hours (wet cycle), subsequently they have been exposed to room condition for

16 hours (dry cycle). Each wet and dry cycle has been repeated 20 times. Specimens have been

weighted at the beginning of each cycle.

Bronze corroded specimens have been characterized employing several analytical techniques.

Particularly the morphology and the microstructure of the bronze corroded surface has been

investigated with Scanning Electron Microscopy and XRD analysis. SEM observations and X-ray

Energy Dispersion Scansion analysis have been carried out both on the surface and on the section of

the specimens. In this way we were able to identify the elements characterizing the corrosion patina,

to highlight the morphology of corrosion products and to measure patina thickness. XRD analysis

have been carried out only on the specimens surface in order to identify the corrosion products.

Corroded surfaces have been also characterized by colorimetric CIELab non destructive

measurements. CIELab test analyses the spectrum of the light reflected by surfaces, through an

instrument constituted by a spectrophotometer. The detected signals may be represented as a graph

with the wavelength (nm) on the horizontal axis and the reflected light percentage (%) on the

vertical axis. The graph can be elaborated in order to obtained the brightness of the reflected light

brightens (L*) and two different parameters (a* and b*) which quantitatively define the reflected

light putting it in the tridimensional CIELab color space. L* is always positive and it represents the

light brightness (a black surface give L=0 while a white surface give L*=100); a* is the

redness/greenness ratio and a* define the position on the redness-greenness axis (a* is positive in

the red region, negative in the green region) and b* is the ratio blueness/yellowness and define the

position on the blueness – yellowness axis (b* is positive in the yellow region and negative in the

blue region). If reflected light is characterized by a*=0 and b*= 0, it means that all light wavelength

light components are present and the light will be white if L*=100 and lack if L*=0. Patinas with

comparable colorimetric measurements are probably formed by the same copper compounds.

Results and discussion

Corrosion products characterization

Figure 1 reports the weight variation measured during sulphur dioxide corrosion test. It can be

clearly seen that during the first 5 cycles all specimens exposed to a sulphur dioxide-moist ambient

show a continuous weight increment. From the fifth up to the tenth cycle a strong weight decrease

as been observed. After the tenth cycle the weight decrease is more slow.


Figure 1 Weight variation vs. n. of exposition cycles

The macroscopic observation of corroded surfaces shows that green corrosion products appear in

the central zone ever since the first cycle. In few cycles these corrosion products cover the whole

specimen surface and their appearance becomes more and more compact.

Characterization test have been carried out after 1, 3, 5, 10, 15 and 20 cycles.

Fig. 2 and 3 show the SEM observations and the EDS spectrum of surfaces exposed to SO2 and

100% of relative humidity for 1 and 3 cycles.

Figure 2 SEM observation and EDS spectrum of the surface of a specimen exposed for 1 cycle to SO2

Bronze surfaces appear covered by small crystals composed by copper, sulfur and oxygen.

On corroded surfaces it is possible to localize areas without these crystals characterized by

numerous microcracks which grow inside the materials. EDS analysis carried out in these zones

show that they are rich of tin, sulfur and oxygen.

0

0.5

1

1.5

2

2.5

0 2 4 6 8 10 12 14 16 18 20

n. cycles

Weight variation (mg/cm2)

Figure 1 Weight variation vs. n. of exposition cycles

The macroscopic observation of corroded surfaces shows that green corrosion products appear in

the central zone ever since the first cycle. In few cycles these corrosion products cover the whole

specimen surface and their appearance becomes more and more compact.

Characterization test have been carried out after 1, 3, 5, 10, 15 and 20 cycles.

Fig. 2 and 3 show the SEM observations and the EDS spectrum of surfaces exposed to SO2 and

100% of relative humidity for 1 and 3 cycles.


Bronze surfaces appear covered by small crystals composed by copper, sulfur and oxygen.

On corroded surfaces it is possible to localize areas without these crystals characterized by

numerous microcracks which grow inside the materials. EDS analysis carried out in these zones

show that they are rich of tin, sulfur and oxygen.

0

0.5

1

1.5

2

2.5

0 2 4 6 8 10 12 14 16 18 20

n. cycles

Weight variation (mg/cm2)



The XRD analysis (Fig. 4) carried out on the same specimens allows to identify the minerals

compounds which have been formed during the first step of the corrosion process. The principal

identified compound are Brochantite (Cu4SO4(OH)6), Cuprite (Cu2O) and Ottemannite (Sn2S3).

Some trace of Cassiterite (SnO2) has also been detected.

Figure 4 XRD spectrum of the surface of a specimen exposed for 3 cycle to SO2 (C= Cuprite, B=Brocanthite and

O= Ottemannite)

The small crystals shown in figure 2 and 3 are Brochantite crystals . This hydrated copper sulphate

develops over the cuprite layer which covers the bronze surface. The localized tin rich areas are

made of oxide and sulphide tin compounds (Ottemannite and Cassiterite).

Brochantite crystals grow rapidly in dimension and coalesce together forming brochantite

palquettes (area 3 of Fig. 3) which cover the surface hiding eventual corrosion products like the

ottemannite pits. After 5 cycle the surface is completely covered by copper corrosion products.

After 15 cycle the XRD analysis (Fig. 5) show the presence on the surface of Brochantite and

Chalcanthite (CuSO4*5H2O) with some rare traces of Ottemannite [3].

0

200

400

600

800

1000

1200

10.00 20.00 30.00 40.00 50.00 60.00 70.00

2Θ2Θ2Θ2Θ

Intensity

BB

B

B

BB B

B

B

C

C

C

O

O

O

O

O

OO

O

C B

B


The XRD analysis (Fig. 4) carried out on the same specimens allows to identify the minerals

compounds which have been formed during the first step of the corrosion process. The principal

identified compound are Brochantite (Cu4SO4(OH)6), Cuprite (Cu2O) and Ottemannite (Sn2S3).

Some trace of Cassiterite (SnO2) has also been detected.

Figure 4 XRD spectrum of the surface of a specimen exposed for 3 cycle to SO2 (C= Cuprite, B=Brocanthite and

O= Ottemannite)

The small crystals shown in figure 2 and 3 are Brochantite crystals . This hydrated copper sulphate

develops over the cuprite layer which covers the bronze surface. The localized tin rich areas are

made of oxide and sulphide tin compounds (Ottemannite and Cassiterite).

Brochantite crystals grow rapidly in dimension and coalesce together forming brochantite

palquettes (area 3 of Fig. 3) which cover the surface hiding eventual corrosion products like the

ottemannite pits. After 5 cycle the surface is completely covered by copper corrosion products.

After 15 cycle the XRD analysis (Fig. 5) show the presence on the surface of Brochantite and

Chalcanthite (CuSO4*5H2O) with some rare traces of Ottemannite [3].

0

200

400

600

800

1000

1200

10.00 20.00 30.00 40.00 50.00 60.00 70.00

2Θ2Θ2Θ2Θ

Intensity

BB

B

B

BB B

B

B

C

C

C

O

O

O

O

O

OO

O

C B

B


Figure 5 XRD spectrum of the surface of a specimen exposed for 15 cycle to SO2 (Ca= Chalcanthite, B=Brochantite

and O= Ottemannite)

As before said after 5 cycle, the specimens start to loose weight. This decreasing is initially strong,

but after the tenth cycle becomes more soft. SEM observations of the cross sections show that the

alterated layer thickness is always rising with a linear trend. After 5 exposition cycle (Fig. 6) it

appears as a compact layer with an average thickness of about 15 µm The EDS analysis carried out

in several zone of this compact patina highlight that it is composed principally by copper-sulphur-

oxygen compounds confirming that the XRD data which indicate Brochantite (copper

hydroxysulphates) as principal compound forming the corrosion patina. At the end of the

experimentation (15/20 Cycle) the average patina thickness is about 20-30 µm (Fig. 7).

Our results are in agreement with the brochantite formation mechanism proposed by Odnevall et al.

[18] They indicated that, in the initial oxidation process, cuprite formation is followed by posnjakite

(Cu4SO4(OH)6*H2O), as a precursors phase to brochantite. They also revealed the existence of

intermediate amorphous compounds including copper and sulfate before the formation of posnjakite

and brochantite. In our case both posnjakite and amorphous compounds weren’t detected during the

corrosion tests, probably due to the high levels of SO2.

Figure 6 SEM observation the cross section of a

specimen exposed for 5cycle to SO2

Figure 7 SEM observation of the cross section of a

specimen exposed for 15 cycle to SO2

0

200

400

600

800

1000

1200

10 20 30 40 50 60 70

2Θ2Θ2Θ2Θ

intensity

CaCa

Ca

Ca

Ca

Ca

CaCaCa

B

B

B

BB

B

BB

B

Ca

B

Ca

Ca

B

OOOO

O

Figure 5 XRD spectrum of the surface of a specimen exposed for 15 cycle to SO2 (Ca= Chalcanthite, B=Brochantite

and O= Ottemannite)

As before said after 5 cycle, the specimens start to loose weight. This decreasing is initially strong,

but after the tenth cycle becomes more soft. SEM observations of the cross sections show that the

alterated layer thickness is always rising with a linear trend. After 5 exposition cycle (Fig. 6) it

appears as a compact layer with an average thickness of about 15 µm The EDS analysis carried out

in several zone of this compact patina highlight that it is composed principally by copper-sulphur-

oxygen compounds confirming that the XRD data which indicate Brochantite (copper

hydroxysulphates) as principal compound forming the corrosion patina. At the end of the

experimentation (15/20 Cycle) the average patina thickness is about 20-30 µm (Fig. 7).

Our results are in agreement with the brochantite formation mechanism proposed by Odnevall et al.

[18] They indicated that, in the initial oxidation process, cuprite formation is followed by posnjakite

(Cu4SO4(OH)6*H2O), as a precursors phase to brochantite. They also revealed the existence of

intermediate amorphous compounds including copper and sulfate before the formation of posnjakite

and brochantite. In our case both posnjakite and amorphous compounds weren’t detected during the

corrosion tests, probably due to the high levels of SO2.

Figure 6 SEM observation the cross section of a

specimen exposed for 5cycle to SO2

Figure 7 SEM observation of the cross section of a

specimen exposed for 15 cycle to SO2

0

200

400

600

800

1000

1200

10 20 30 40 50 60 70

2Θ2Θ2Θ2Θ

intensity

CaCa

Ca

Ca

Ca

Ca

CaCaCa

B

B

B

BB

B

BB

B

Ca

B

Ca

Ca

B

OOOO

O


Cross sections observation are useful also for the characterization of the pit corrosion morphology

and chemistry. Figure 8 shows the cross section of specimen exposed to SO2-humidity atmosphere

for 10 cycles after a chemical etch with FeCl2 carried out in order to reveal the metallurgical

structure of the alloy and the interaction between structure and corrosion pit. This chemical etch has

unfortunately melt all the corrosion products, so in this micro-photos pits appear empty. Pits are

usually full of a corrosion products whose structure is characterized by a wide web of microcracks.

EDS analysis carried out on pit show spectra similar to those reported in Fig. 2 (zone 2) and in

figure 3 (zone 4).They grow under the specimens surface following the bronze grain structure. Their

formation is probably due to the recombination cycles of pure elements and their oxides. The

average depth of these pit is about 25 µm.

Figure 8 SEM observation of the cross section of a specimen exposed for 20 cycle to SO2 (etched with FeCl2)

Corrosion products in the patina are always composed by copper-sulfur-oxygen compounds

confirming the XRD analysis that indicate the presence of two different copper sulphate

compounds. However it has been observed that even if the global patina thickness grow with the

number of exposition cycles, it becomes more and more less compact. It appears in many areas

disgregated showing several zone without corrosion products. This is probably due to a progressive

fragmentation arising form the strong chemical etch of SO2 which brings to the loss of pieces of

patina and this also justifies the weight decrease recorded after the fifth exposition cycle.

CIELab Characterization

The Colorimetric method is a non destructive investigation technique based on the surface color

quantification. It can give useful information on modifications occurred on a metallic surface during

a corrosion process. The reflective coefficient, which corresponds to the intensity of the reflected

light, is reported in the diagram of Figure 9 as a function of the wavelength. The 0 cycle points are

those of the original bronze, while the others represent the behavior after different number of

cycles. The non corroded reflective curve is characterized by a continuous increase of the reflective

coefficient with the wavelength and by a inflection point at a wavelength of about 550 nm.. As

expected the curve after each SO2 exposition are characterized by lower reflective coefficient value

at every wavelength. The reflective curves of the corroded surfaces show indeed a slowly rising

Cross sections observation are useful also for the characterization of the pit corrosion morphology

and chemistry. Figure 8 shows the cross section of specimen exposed to SO2-humidity atmosphere

for 10 cycles after a chemical etch with FeCl2 carried out in order to reveal the metallurgical

structure of the alloy and the interaction between structure and corrosion pit. This chemical etch has

unfortunately melt all the corrosion products, so in this micro-photos pits appear empty. Pits are

usually full of a corrosion products whose structure is characterized by a wide web of microcracks.

EDS analysis carried out on pit show spectra similar to those reported in Fig. 2 (zone 2) and in

figure 3 (zone 4).They grow under the specimens surface following the bronze grain structure. Their

formation is probably due to the recombination cycles of pure elements and their oxides. The

average depth of these pit is about 25 µm.

Figure 8 SEM observation of the cross section of a specimen exposed for 20 cycle to SO2 (etched with FeCl2)

Corrosion products in the patina are always composed by copper-sulfur-oxygen compounds

confirming the XRD analysis that indicate the presence of two different copper sulphate

compounds. However it has been observed that even if the global patina thickness grow with the

number of exposition cycles, it becomes more and more less compact. It appears in many areas

disgregated showing several zone without corrosion products. This is probably due to a progressive

fragmentation arising form the strong chemical etch of SO2 which brings to the loss of pieces of

patina and this also justifies the weight decrease recorded after the fifth exposition cycle.

CIELab Characterization

The Colorimetric method is a non destructive investigation technique based on the surface color

quantification. It can give useful information on modifications occurred on a metallic surface during

a corrosion process. The reflective coefficient, which corresponds to the intensity of the reflected

light, is reported in the diagram of Figure 9 as a function of the wavelength. The 0 cycle points are

those of the original bronze, while the others represent the behavior after different number of

cycles. The non corroded reflective curve is characterized by a continuous increase of the reflective

coefficient with the wavelength and by a inflection point at a wavelength of about 550 nm.. As

expected the curve after each SO2 exposition are characterized by lower reflective coefficient value

at every wavelength. The reflective curves of the corroded surfaces show indeed a slowly rising


trend up to a wavelength of about 580 nm followed by a slow decrease. This trend can be quantify

calculating the values of the parameters L*, a* and b* reported in Table 1. The L*, a* and b*

parameters rapidly decrease after the first cycles, after this point the reduction trend begins lower.

Figure 9 Bronze specimens reflective curves (before and after the sulphur dioxide treatments)

Table 1 CIELab parameters related to the reflective curves of Fig. 9

0 cycle 5 cycles 10 cycles 15 cycles 20 cycles

L* 78.01 52.34 50.91 50.28 48.92

a* 7.2 -3.68 -3.73 -6.21 -6.27

b* 20.91 17.94 14.33 12.52 11.71

The L* parameter reduction is due to the nature and the morphology of minerals which form the

corrosion patina. They reflect the light less than a pure metal , and furthermore the corroded surface

is rough and this yields to a further light reflection reduction. The decreasing of a* and b*

highlights the modification of the surface color from the yellow which characterize the non

corroded bronze to a green-blue which indeed is characteristic of many copper-sulfur minerals.

The utility of this measurements is linked to the non destructive nature of this kind of test. Surfaces

with same color parameters probably are made of the same compounds. Obtain a deep correlation

between color quantified parameters and chemical and morphological nature of corrosion products,

can be an useful instruments for evaluating a corrosion phenomena which happen on bronze

artifacts exposed to the ambient. In this way useful information on the origin of the corrosion and

on the better restoring ,actions can be obtained with simple non destructive measurements.

Conclusion

The employed studying methodology has been suitable to obtain the information necessary for the

future mathematical model. This model needs the knowledge of the composition of the corrosion

layer after each exposition cycle. During the experimentation the modifications of the corrosion

layer composition have been identified. In particular the development of copper corrosion products

(brochantite, cuprite and chalcanthite) has been monitored characterizing their morphology, their

thickness etc. Moreover it has been studied also the development of tin corrosion products

(ottemannite and cassiterite) which forms pit which grow under the metal surface. All the data

obtained are also interesting to better understand all the corrosion mechanisms which brings to

bronze surface degradation.

Surface chemical modifications have been also monitored analyzing its color. The corrosion

phenomena brings to color variation which can be quantified analyzing the spectrum of the reflected

light. The studied corrosion phenomena cause the decrease of all the CIELab parameters which

trend up to a wavelength of about 580 nm followed by a slow decrease. This trend can be quantify

calculating the values of the parameters L*, a* and b* reported in Table 1. The L*, a* and b*

parameters rapidly decrease after the first cycles, after this point the reduction trend begins lower.

Figure 9 Bronze specimens reflective curves (before and after the sulphur dioxide treatments)

Table 1 CIELab parameters related to the reflective curves of Fig. 9

0 cycle 5 cycles 10 cycles 15 cycles 20 cycles

L* 78.01 52.34 50.91 50.28 48.92

a* 7.2 -3.68 -3.73 -6.21 -6.27

b* 20.91 17.94 14.33 12.52 11.71

The L* parameter reduction is due to the nature and the morphology of minerals which form the

corrosion patina. They reflect the light less than a pure metal , and furthermore the corroded surface

is rough and this yields to a further light reflection reduction. The decreasing of a* and b*

highlights the modification of the surface color from the yellow which characterize the non

corroded bronze to a green-blue which indeed is characteristic of many copper-sulfur minerals.

The utility of this measurements is linked to the non destructive nature of this kind of test. Surfaces

with same color parameters probably are made of the same compounds. Obtain a deep correlation

between color quantified parameters and chemical and morphological nature of corrosion products,

can be an useful instruments for evaluating a corrosion phenomena which happen on bronze

artifacts exposed to the ambient. In this way useful information on the origin of the corrosion and

on the better restoring ,actions can be obtained with simple non destructive measurements.

Conclusion

The employed studying methodology has been suitable to obtain the information necessary for the

future mathematical model. This model needs the knowledge of the composition of the corrosion

layer after each exposition cycle. During the experimentation the modifications of the corrosion

layer composition have been identified. In particular the development of copper corrosion products

(brochantite, cuprite and chalcanthite) has been monitored characterizing their morphology, their

thickness etc. Moreover it has been studied also the development of tin corrosion products

(ottemannite and cassiterite) which forms pit which grow under the metal surface. All the data

obtained are also interesting to better understand all the corrosion mechanisms which brings to

bronze surface degradation.

Surface chemical modifications have been also monitored analyzing its color. The corrosion

phenomena brings to color variation which can be quantified analyzing the spectrum of the reflected

light. The studied corrosion phenomena cause the decrease of all the CIELab parameters which


indicate a strong reduction of the reflective coefficient and a shift of the color from yellow (typical

of bronze metal) to green-blue (typical of many copper minerals).

The study of color modification and their correlation to corrosion development is necessary to the

develop this new non destructive analyzing methodology. Even if at this step of research the study

have to be considered only preliminary, from the obtained results it is possible to forecast an

interesting application of the methodology to observe the degradation state of a bronze surface from

colorimetric measurements as absolutely non destructive, not invasive methodology.

References

[1] T. E. Graedel, K Nassau, J. P: Franey, Copper patinas formed in the atmosphere-I. Introduction,

Corros. Sci 27 (1987) 639-657.

[2] A. Kratschmer, I. Odnevall Wallander, C. Leygraf, The evolution of outdoor copper patina,

Corrosion Science 44 (2002) 425-450.

[3] J.M. Bastidas, A. Lopez-Delgado, F. A. Lopez, Characterization of artificially patinated layer

on artistic bronze exposed to laboratory SO2 contamination, Journal of Materials Science 32

(1997) p.129-133.

[4] K. Nassau, A.E. Miller, T.E. Graedel, The reaction of simulated rain with copper, copper

patina, and some copper compounds, Corrosion Science 27 (1987) 703-709.

[5 ] T. E. Graedel, Corros.Sci. 45 (2003) 2851

[6] C. Chiavari, K. Rahmouni, H. Takenout, S. Joiret, P. Vermaut, L. Robbiola, Composition and

electrochemical properties of natural patinas of outdoor bronze monuments, Electrochimica

Acta, 52 (2007) 7760-7769.

[7] J. H. Payer, Corrosion Processes in the Development of Thin Tarnish Films, Electrical Contacts-

1990 Proceedings of the Thirty Sixth IEEE Holm Conference on Electrical Contacts meeting

jointly

with the Fifteenth International Conference on Electrical Contacts, pp. 203–211, Piscataway, NJ:

IEEE, 1990.

[8] B. I. Rickett and J.H. Payer Composition of Copper tarnish products formed in moist air with

trace levels of pollutant gas: sulfur dioxide and sulfur dioxide/nitrogen dioxide

J. Electrochemical Soc. Vol 142, N.11, 1995, 3713-3722].

[9] L. Morselli, E. Bernardi, C. Chiavari, G. Brunoro, Corrosion of 85-5-5-5 bronze in natural and

synthetic acid rain, Appl. Phys. A 79 (2004) 363-367.

[10] L. Robbiola, C.Fiaud, Apport de l’analyse statistique des produits de corrosion a la

comprehension des processus de dégradation des bronzes archéologiques, Revue

d’Archeometrie, 16 (1992), 109-119

[11] E. Bernardi, C. Chiavari, B. Lenza, C. Martini, L. Morselli, F. Ospitali, L. Robbiola The

atmospheric corrosion of quaternary bronzes, the leaching action of acid rain, Corrosion

Science 51 (2009), pp 159-170

[12] I. Ondevall Wallinder, C. Leygraf, A study of copper runoff in an urban atmosphere, Corros.

Sci. 39 (1997) 12, 2039-2052.

[13] W. H. J. Vernon, Trans. Faraday Soc. 27 (1931), 255.

[14] N. D. Tomashov Theory of Corrosion and protection of metals, B.H.Tytell, Translator,

E.C.Greco, Editor, p. 125 NACE, Houston, Texas (1972)

[15] P.B. P. Phipps and D.W. Rice, in corrosion Chemistry American Chemical Society,

Washington DC (1979).

[16] S. K. Chawla, J.H. Payer, The Early Stage of Atmospheric Corrosion of Copper by Sulfur

Dioxide, J. Electrochem. Soc. 137 (1990) 1 60-64.

[17] D.A. Scott. Copper and Bronze in Art: Corrosion, Colorants, Conservation, The Getty

Conservation Institute Los Angeles 2002, 45-46.

[18] I. Odnevall and C. Leygraf, J. Electrochem. Soc., 142, 3682, 1995

indicate a strong reduction of the reflective coefficient and a shift of the color from yellow (typical

of bronze metal) to green-blue (typical of many copper minerals).

The study of color modification and their correlation to corrosion development is necessary to the

develop this new non destructive analyzing methodology. Even if at this step of research the study

have to be considered only preliminary, from the obtained results it is possible to forecast an

interesting application of the methodology to observe the degradation state of a bronze surface from

colorimetric measurements as absolutely non destructive, not invasive methodology.

References

[1] T. E. Graedel, K Nassau, J. P: Franey, Copper patinas formed in the atmosphere-I. Introduction,

Corros. Sci 27 (1987) 639-657.

[2] A. Kratschmer, I. Odnevall Wallander, C. Leygraf, The evolution of outdoor copper patina,

Corrosion Science 44 (2002) 425-450.

[3] J.M. Bastidas, A. Lopez-Delgado, F. A. Lopez, Characterization of artificially patinated layer

on artistic bronze exposed to laboratory SO2 contamination, Journal of Materials Science 32

(1997) p.129-133.

[4] K. Nassau, A.E. Miller, T.E. Graedel, The reaction of simulated rain with copper, copper

patina, and some copper compounds, Corrosion Science 27 (1987) 703-709.

[5 ] T. E. Graedel, Corros.Sci. 45 (2003) 2851

[6] C. Chiavari, K. Rahmouni, H. Takenout, S. Joiret, P. Vermaut, L. Robbiola, Composition and

electrochemical properties of natural patinas of outdoor bronze monuments, Electrochimica

Acta, 52 (2007) 7760-7769.

[7] J. H. Payer, Corrosion Processes in the Development of Thin Tarnish Films, Electrical Contacts-

1990 Proceedings of the Thirty Sixth IEEE Holm Conference on Electrical Contacts meeting

jointly

with the Fifteenth International Conference on Electrical Contacts, pp. 203–211, Piscataway, NJ:

IEEE, 1990.

[8] B. I. Rickett and J.H. Payer Composition of Copper tarnish products formed in moist air with

trace levels of pollutant gas: sulfur dioxide and sulfur dioxide/nitrogen dioxide

J. Electrochemical Soc. Vol 142, N.11, 1995, 3713-3722].

[9] L. Morselli, E. Bernardi, C. Chiavari, G. Brunoro, Corrosion of 85-5-5-5 bronze in natural and

synthetic acid rain, Appl. Phys. A 79 (2004) 363-367.

[10] L. Robbiola, C.Fiaud, Apport de l’analyse statistique des produits de corrosion a la

comprehension des processus de dégradation des bronzes archéologiques, Revue

d’Archeometrie, 16 (1992), 109-119

[11] E. Bernardi, C. Chiavari, B. Lenza, C. Martini, L. Morselli, F. Ospitali, L. Robbiola The

atmospheric corrosion of quaternary bronzes, the leaching action of acid rain, Corrosion

Science 51 (2009), pp 159-170

[12] I. Ondevall Wallinder, C. Leygraf, A study of copper runoff in an urban atmosphere, Corros.

Sci. 39 (1997) 12, 2039-2052.

[13] W. H. J. Vernon, Trans. Faraday Soc. 27 (1931), 255.

[14] N. D. Tomashov Theory of Corrosion and protection of metals, B.H.Tytell, Translator,

E.C.Greco, Editor, p. 125 NACE, Houston, Texas (1972)

[15] P.B. P. Phipps and D.W. Rice, in corrosion Chemistry American Chemical Society,

Washington DC (1979).

[16] S. K. Chawla, J.H. Payer, The Early Stage of Atmospheric Corrosion of Copper by Sulfur

Dioxide, J. Electrochem. Soc. 137 (1990) 1 60-64.

[17] D.A. Scott. Copper and Bronze in Art: Corrosion, Colorants, Conservation, The Getty

Conservation Institute Los Angeles 2002, 45-46.

[18] I. Odnevall and C. Leygraf, J. Electrochem. Soc., 142, 3682, 1995


Electrochemical Methods to Assist Corrosion Control of

Lightweight Alloys

M. Curioni1, G. E. Thompson1

Corrosion and Protection Centre, School of Materials, The University of Manchester, Manchester M13 9PL, UK

[email protected]

Abstract

In this work, the use of a differential aeration technique (split-cell) to assist understanding of

contributions to the corrosion of an aerospace aluminium alloy is demonstrated. The setup

comprised two similar specimens immersed in differentially aerated test solutions and coupled by a

zero resistance ammeter. The individual electrochemical responses of the coupled specimens during

linear polarization were interpreted in relation to the aeration condition, alloy composition and

surface preparation. Further, the same setup was employed to investigate corrosion inhibition by

observing the current and potential transients after inhibitor addition to the aerated or deareated

compartment. It was found that the split-cell technique provides detailed understanding of the

corrosion process in multiphase alloys and provides important information on inhibitor

performance.

Introduction

In order to control corrosion of light alloys, a variety of approaches including anodic treatment [1],

conversion coating, sol-gel coating[2],and application of organic coatings are employed, with

validation of performance provided, for example, by exposure to a salt spray test (SST). SST is

generally accepted by the aerospace industry as one of the reliable tools for performance

assessment, but it is time-consuming and provides little information on the corrosion process and

how the candidate protection system interacts with the aggressive environment and with the

substrate. For this reason, electrochemical techniques are generally preferred for the development of

a protection system, since they provide valuable information on corrosion rates, inhibition and

barrier effects [3]. However, selection of the most appropriate electrochemical techniques is

necessary, since conflicting requirements simultaneously coexist. Direct current (DC) techniques,

such as linear polarization, apply a relatively large perturbation to the corroding system, producing

an electrical response which can be easily recorded and, generally, interpreted readily. However, the

relatively large polarization of the surface may produce results which do not necessarily correlate

Electrochemical Methods to Assist Corrosion Control of

Lightweight Alloys

M. Curioni1, G. E. Thompson1

Corrosion and Protection Centre, School of Materials, The University of Manchester, Manchester M13 9PL, UK

[email protected]

Abstract

In this work, the use of a differential aeration technique (split-cell) to assist understanding of

contributions to the corrosion of an aerospace aluminium alloy is demonstrated. The setup

comprised two similar specimens immersed in differentially aerated test solutions and coupled by a

zero resistance ammeter. The individual electrochemical responses of the coupled specimens during

linear polarization were interpreted in relation to the aeration condition, alloy composition and

surface preparation. Further, the same setup was employed to investigate corrosion inhibition by

observing the current and potential transients after inhibitor addition to the aerated or deareated

compartment. It was found that the split-cell technique provides detailed understanding of the

corrosion process in multiphase alloys and provides important information on inhibitor

performance.

Introduction

In order to control corrosion of light alloys, a variety of approaches including anodic treatment [1],

conversion coating, sol-gel coating[2],and application of organic coatings are employed, with

validation of performance provided, for example, by exposure to a salt spray test (SST). SST is

generally accepted by the aerospace industry as one of the reliable tools for performance

assessment, but it is time-consuming and provides little information on the corrosion process and

how the candidate protection system interacts with the aggressive environment and with the

substrate. For this reason, electrochemical techniques are generally preferred for the development of

a protection system, since they provide valuable information on corrosion rates, inhibition and

barrier effects [3]. However, selection of the most appropriate electrochemical techniques is

necessary, since conflicting requirements simultaneously coexist. Direct current (DC) techniques,

such as linear polarization, apply a relatively large perturbation to the corroding system, producing

an electrical response which can be easily recorded and, generally, interpreted readily. However, the

relatively large polarization of the surface may produce results which do not necessarily correlate

with in-service behaviour, since the corrosion process proceeds at the corrosion potential [3].

Further, traditional DC techniques do not provide information on the relative behaviour of the

different metallurgical phases present on the alloy surface or the local differences due to the

development of microenvironments.

The purpose of the present work is to illustrate the use of split-cell techniques that overcome some

of the previously described issues. The split-cell technique [4-6] is based on a cell with two

identical cylindrical compartments connected by a porous frit, enabling current flow but preventing

significant mass exchange. In each of the two compartments, filled with the test electrolyte,

individual specimens of the alloy under investigation are immersed and the two specimens are

connected through a zero resistance ammeter (ZRA). After a sufficient settling time, oxygen-free

nitrogen is passed into one compartment and air is passed into the other compartment. If a potential

sweep (linear polarization) is applied to the two coupled electrodes using a third auxiliary electrode,

the electrical response of each electrode can be recorded, providing information on the effects of

different microstructure, composition or surface preparation on the cathodic and anodic activities.

By performing a single cyclic polarization scan, it is possible to identify changes in the pitting

potential induced by microstructure or surface condition, cathodic activity in relation to oxygen

reduction or hydrogen evolution and the compositional/metallurgical conditions that influence the

corrosion potential.

Alternatively, the current generated by differential aeration can be measured by the ZRA, without

application of external potential [4, 5]. In this case, as a result of the deareation, the cathodic

reaction of the specimen in the nitrogen purged compartment is restricted and this specimen

becomes a net anode with respect to the specimen immersed in the aerated compartment, with a

relatively small change in the corrosion potential. Once this situation is established, the effect of a

candidate compound as an anodic, cathodic or mixed inhibitor can be evaluated by adding the

compound to one of the two compartments and observing the effect on the current and potential

transients.

Experimental setup.

The split cell experiments where performed on acetone degreased and on acetone degreased, etched

and desmutted specimens of AA 2024 T3 and Al 1000 ppm wt. Cu alloys. The etching and

desmutting treatment involved immersion in 10 % wt. NaOH at 60o C for 30 seconds and

desmutting in 30% vol. HNO3 solution. All the specimens used for inhibition studies were etched

and desmutted. After pre-treatment, the specimens were assembled to form spade electrodes and,

subsequently, masked. The split cell comprised two 800 ml cylindrical compartments connected by

a porous gas frit. One specimen was immersed in each compartment and the two specimens were

with in-service behaviour, since the corrosion process proceeds at the corrosion potential [3].

Further, traditional DC techniques do not provide information on the relative behaviour of the

different metallurgical phases present on the alloy surface or the local differences due to the

development of microenvironments.

The purpose of the present work is to illustrate the use of split-cell techniques that overcome some

of the previously described issues. The split-cell technique [4-6] is based on a cell with two

identical cylindrical compartments connected by a porous frit, enabling current flow but preventing

significant mass exchange. In each of the two compartments, filled with the test electrolyte,

individual specimens of the alloy under investigation are immersed and the two specimens are

connected through a zero resistance ammeter (ZRA). After a sufficient settling time, oxygen-free

nitrogen is passed into one compartment and air is passed into the other compartment. If a potential

sweep (linear polarization) is applied to the two coupled electrodes using a third auxiliary electrode,

the electrical response of each electrode can be recorded, providing information on the effects of

different microstructure, composition or surface preparation on the cathodic and anodic activities.

By performing a single cyclic polarization scan, it is possible to identify changes in the pitting

potential induced by microstructure or surface condition, cathodic activity in relation to oxygen

reduction or hydrogen evolution and the compositional/metallurgical conditions that influence the

corrosion potential.

Alternatively, the current generated by differential aeration can be measured by the ZRA, without

application of external potential [4, 5]. In this case, as a result of the deareation, the cathodic

reaction of the specimen in the nitrogen purged compartment is restricted and this specimen

becomes a net anode with respect to the specimen immersed in the aerated compartment, with a

relatively small change in the corrosion potential. Once this situation is established, the effect of a

candidate compound as an anodic, cathodic or mixed inhibitor can be evaluated by adding the

compound to one of the two compartments and observing the effect on the current and potential

transients.

Experimental setup.

The split cell experiments where performed on acetone degreased and on acetone degreased, etched

and desmutted specimens of AA 2024 T3 and Al 1000 ppm wt. Cu alloys. The etching and

desmutting treatment involved immersion in 10 % wt. NaOH at 60o C for 30 seconds and

desmutting in 30% vol. HNO3 solution. All the specimens used for inhibition studies were etched

and desmutted. After pre-treatment, the specimens were assembled to form spade electrodes and,

subsequently, masked. The split cell comprised two 800 ml cylindrical compartments connected by

a porous gas frit. One specimen was immersed in each compartment and the two specimens were


connected through a ZRA (Solartron 1280 for inhibitor studies and Concerto Multichannel

Potentiostat for polarization experiments at 10 mV min-1). The reference electrode was placed in the

areated compartment. Initially, the electrolyte was naturally aerated 3.5 % wt. NaCl in both sides of

the cell. The specimens were immersed simultaneously in the two compartments and left under

naturally aerated conditions for 15 minutes. Subsequently, nitrogen was passed into one

compartment and air into the other. For inhibition studies, after 30 minutes from commencement of

the experiment, 100 ppm. wt. sodium tartrate or chromic acid were added to one of the two

compartments and the effects on current and potential were recorded. In the Figures presented later,

the current flowing in the external circuit from the net cathode (aerated compartment) to the net

anode (deareated compartment) is taken as positive.

Understanding alloy behaviour by split-cell polarization. When a multiphase aluminum alloy is

immersed in an aggressive electrolyte, a variety of metallurgical phases are exposed simultaneously

at the surface [7, 8] and, as the corrosion process proceeds, different microenvironments can

develop locally. The combined generation of microenvironments and the presence of alloying

elements, both in second phase material and in solid solution, results in a relatively complicated

system. Specifically, whilst appropriate second phase material provides sites for the cathodic

reaction, in the presence of alloying elements in solid solution in the matrix that are nobler than

aluminum, as corrosion proceeds enrichment of such alloying elements takes place at the

metal/oxide interface. This is due to the difference in the Gibbs free energy per equivalent for

alumina formation and the oxide of the respective alloying element. Thus, for alloying elements

with Gibbs free energies per equivalent less negative than alumina, aluminium preferentially

oxidizes and the very near-surface concentration of the nobler element increases. As the reaction

proceeds, a critical concentration (enrichment), corresponding to equivalent surface activities of

alloying elements and aluminium, is attained and co-oxidation proceeds in the alloy proportions

[9-14]. Therefore, the chemical nature of the anodic phase (the matrix) changes locally and with

time. In order to investigate this condition, split cell polarization is particularly useful. Using this

method and the selection of the appropriate representative specimens, it is possible to gain insight

into the relative contributions to the cathodic and anodic activities of the alloy matrix and second

phase materials at different stages of the corrosion process. In Figure 1, linear polarization curves,

obtained with the split cell arrangement, are presented for unetched Al-1000 ppm wt. Cu,

etched/desmutted Al-1000 ppm wt. Cu and unetched AA2024 T3 aluminum alloy. The as received

Al-Cu 1000 ppm alloy is representative of a region of matrix of the practical alloy where the

corrosion process did not start and no material has been anodically removed; therefore, a well

established copper enriched layer is not expected. The etched/desmutted Al-Cu 1000 ppm wt. alloy

connected through a ZRA (Solartron 1280 for inhibitor studies and Concerto Multichannel

Potentiostat for polarization experiments at 10 mV min-1). The reference electrode was placed in the

areated compartment. Initially, the electrolyte was naturally aerated 3.5 % wt. NaCl in both sides of

the cell. The specimens were immersed simultaneously in the two compartments and left under

naturally aerated conditions for 15 minutes. Subsequently, nitrogen was passed into one

compartment and air into the other. For inhibition studies, after 30 minutes from commencement of

the experiment, 100 ppm. wt. sodium tartrate or chromic acid were added to one of the two

compartments and the effects on current and potential were recorded. In the Figures presented later,

the current flowing in the external circuit from the net cathode (aerated compartment) to the net

anode (deareated compartment) is taken as positive.

Understanding alloy behaviour by split-cell polarization. When a multiphase aluminum alloy is

immersed in an aggressive electrolyte, a variety of metallurgical phases are exposed simultaneously

at the surface [7, 8] and, as the corrosion process proceeds, different microenvironments can

develop locally. The combined generation of microenvironments and the presence of alloying

elements, both in second phase material and in solid solution, results in a relatively complicated

system. Specifically, whilst appropriate second phase material provides sites for the cathodic

reaction, in the presence of alloying elements in solid solution in the matrix that are nobler than

aluminum, as corrosion proceeds enrichment of such alloying elements takes place at the

metal/oxide interface. This is due to the difference in the Gibbs free energy per equivalent for

alumina formation and the oxide of the respective alloying element. Thus, for alloying elements

with Gibbs free energies per equivalent less negative than alumina, aluminium preferentially

oxidizes and the very near-surface concentration of the nobler element increases. As the reaction

proceeds, a critical concentration (enrichment), corresponding to equivalent surface activities of

alloying elements and aluminium, is attained and co-oxidation proceeds in the alloy proportions

[9-14]. Therefore, the chemical nature of the anodic phase (the matrix) changes locally and with

time. In order to investigate this condition, split cell polarization is particularly useful. Using this

method and the selection of the appropriate representative specimens, it is possible to gain insight

into the relative contributions to the cathodic and anodic activities of the alloy matrix and second

phase materials at different stages of the corrosion process. In Figure 1, linear polarization curves,

obtained with the split cell arrangement, are presented for unetched Al-1000 ppm wt. Cu,

etched/desmutted Al-1000 ppm wt. Cu and unetched AA2024 T3 aluminum alloy. The as received

Al-Cu 1000 ppm alloy is representative of a region of matrix of the practical alloy where the

corrosion process did not start and no material has been anodically removed; therefore, a well

established copper enriched layer is not expected. The etched/desmutted Al-Cu 1000 ppm wt. alloy


represents regions on the practical alloy where corrosion has removed significant amounts of

material, producing a well-developed copper-rich layer [13]. The curves recoded for the practical

alloys account for the cumulative behavior of the matrix and the second phase material.

For the unetched Al-Cu 1000 ppm alloy, at the commencement of the polarization, at -780 mV

SCE, the current is very low for both the aerated and deareated specimen. As the potential is

reduced, the low current is maintained for both specimens until about -1200 mV SCE. Here, the

current flowing across the specimen immersed in the oxygen containing compartment increases

significantly to reach approximately 10 microamps cm-2

when the polarization is reversed. The

specimen immersed in the absence of oxygen does not show such an increase, but displays a current

of about 1 microamp cm-2 at the inversion potential. After the reverse of potential, a significant

hysteresis is observed for the specimen immersed in the presence of oxygen, while this effect is less

significant for the deareated specimen. At about -1100 mV and -1000 mV SCE for the deareated

and aerated specimens respectively, the current recorded during the descending branch becomes

superimposible on the curves recorded during the ascending branch, until about -740 mV SCE. At

the previous potential, a pitting potential is evident and the current sharply increases to high values.

The maximum anodic current is significantly higher at the reversal potential for the specimen

immersed in the presence of oxygen. From these data, information representative of an alloy matrix

unattached by corrosion, can be extracted. The low cathodic currents during the initial descending

branch indicate that the air-formed film covering the alloy is highly protective and does not allow

cathodic reactions, either oxygen reduction or hydrogen evolution, to be efficiently supported.

However, if the specimen is polarized to sufficiently negative values, local cathodic sites may

become activated, generating a local increase in pH. This pH increase produces thinning the air-

formed oxide in a relatively large surrounding region. At other locations within the high pH region,

the air formed film may become sufficiently thin that an additional cathodic site may be activated

and the process propagates. Once this mechanism is onset, the cathodic reactions can be sustained at

more positive potential, during the ascending cathodic branch. This consideration applies both to

the oxygen reduction (in the oxygen containing compartment) and to the hydrogen evolution (in the

oxygen free compartment) reactions, although it is more important in the oxygen-containing

compartment. At about -1000 mV SCE, the rate of oxygen reduction decreases, reducing the

localized pH differences and promoting the re-establishment of a strong air-formed film, thereby

hindering the cathodic reaction. When this occurs, the value of the cathodic currents drops and it

does not depend on the presence of oxygen in the electrolyte. The presence of increased cathodic

activity in the oxygen compartment does not affect the pitting potential, i.e. the pitting potential

does not depend on the oxygen concentration. However, once the potential is sufficient to promote

represents regions on the practical alloy where corrosion has removed significant amounts of

material, producing a well-developed copper-rich layer [13]. The curves recoded for the practical

alloys account for the cumulative behavior of the matrix and the second phase material.

For the unetched Al-Cu 1000 ppm alloy, at the commencement of the polarization, at -780 mV

SCE, the current is very low for both the aerated and deareated specimen. As the potential is

reduced, the low current is maintained for both specimens until about -1200 mV SCE. Here, the

current flowing across the specimen immersed in the oxygen containing compartment increases

significantly to reach approximately 10 microamps cm-2

when the polarization is reversed. The

specimen immersed in the absence of oxygen does not show such an increase, but displays a current

of about 1 microamp cm-2 at the inversion potential. After the reverse of potential, a significant

hysteresis is observed for the specimen immersed in the presence of oxygen, while this effect is less

significant for the deareated specimen. At about -1100 mV and -1000 mV SCE for the deareated

and aerated specimens respectively, the current recorded during the descending branch becomes

superimposible on the curves recorded during the ascending branch, until about -740 mV SCE. At

the previous potential, a pitting potential is evident and the current sharply increases to high values.

The maximum anodic current is significantly higher at the reversal potential for the specimen

immersed in the presence of oxygen. From these data, information representative of an alloy matrix

unattached by corrosion, can be extracted. The low cathodic currents during the initial descending

branch indicate that the air-formed film covering the alloy is highly protective and does not allow

cathodic reactions, either oxygen reduction or hydrogen evolution, to be efficiently supported.

However, if the specimen is polarized to sufficiently negative values, local cathodic sites may

become activated, generating a local increase in pH. This pH increase produces thinning the air-

formed oxide in a relatively large surrounding region. At other locations within the high pH region,

the air formed film may become sufficiently thin that an additional cathodic site may be activated

and the process propagates. Once this mechanism is onset, the cathodic reactions can be sustained at

more positive potential, during the ascending cathodic branch. This consideration applies both to

the oxygen reduction (in the oxygen containing compartment) and to the hydrogen evolution (in the

oxygen free compartment) reactions, although it is more important in the oxygen-containing

compartment. At about -1000 mV SCE, the rate of oxygen reduction decreases, reducing the

localized pH differences and promoting the re-establishment of a strong air-formed film, thereby

hindering the cathodic reaction. When this occurs, the value of the cathodic currents drops and it

does not depend on the presence of oxygen in the electrolyte. The presence of increased cathodic

activity in the oxygen compartment does not affect the pitting potential, i.e. the pitting potential

does not depend on the oxygen concentration. However, once the potential is sufficient to promote


pitting, this is more severe in the oxygen-containing compartment, due to the modifications on the

alloy surface associated with the low pH due to high cathodic activity during the cathodic branch.

Similar considerations apply to the etched and desmutted alloy. However, here, the etching

treatment has generated a copper-enriched layer in the alloy very near-surface, immediately beneath

the air-formed film; the copper-enriched layer may comprise copper-rich nanoparticles that act as

effective cathodic sites[13-15]. As a consequence, during the descending cathodic branch in the

oxygen-containing compartment, the values of current start to increase at about -1000 mV SCE, and

a much reduced hysteresis effect is observed after the potential is reversed. This indicates that, for

the etched alloy, no activation of the cathodic sites for oxygen reduction is necessary, since the

etching treatment has already provided available sites. Conversely, in the oxygen-free compartment,

the cathodic reaction of hydrogen evolution increases significantly below -1300 mV SCE and, once

the cathodic areas are activated, they remain effective after the potential is reversed to about -

1100 mV SCE. The pitting potential is not affected by the oxygen concentration but, as in the

previous case, the maximum pitting current is higher in the oxygen containing compartment.

Etching does not increase the pitting potential with respect to the as-received alloy. With reference

to a real corrosion condition, the information gathered here suggests that once an anodic event

consumes aluminium aluminum , new cathodic sites are also generated on the alloy matrix by

localized copper enrichment.

A similar experiment performed on the AA2024T3 alloy shows that, in the oxygen-containing

compartment, the cathodic reaction of oxygen reduction can proceed under diffusion control below

-600 mV SCE, since the value of current is independent of the applied potential over a relatively

wide range. This suggests that the alloy surface always presents a very high number of active

cathodic sites. Only at about -1100 mV SCE, an increase in current is observed, indicating that the

contribution to the total current from hydrogen evolution becomes significant. In the deareated

compartment, a progressive increase of the current due to the hydrogen evolution is evident with

decreasing potential, with current values similar to those observed for the etched alloy during the

ascending cathodic branch. This indicates that the alloy provides a significant number of effective

cathodic sites for hydrogen evolution, which require little activation. The pitting potential for the

alloy is higher (about 100 mV) than the model Al-Cu alloy. This can be related to the increased

presence of copper with respect to the model alloy.

Summarizing the results and translating the linear polarization behavior to a real corrosion

condition at the (freely variable) open circuit potential, it can be concluded that, during the early

stages of immersion, all the cathodic activity is located on the second phase material, as indicated

by the high cathodic current of the AA 2024 T3 alloy at all potentials. However, as corrosion

progress, an enriched layer is generated on the alloy matrix and the behavior changes from the

pitting, this is more severe in the oxygen-containing compartment, due to the modifications on the

alloy surface associated with the low pH due to high cathodic activity during the cathodic branch.

Similar considerations apply to the etched and desmutted alloy. However, here, the etching

treatment has generated a copper-enriched layer in the alloy very near-surface, immediately beneath

the air-formed film; the copper-enriched layer may comprise copper-rich nanoparticles that act as

effective cathodic sites[13-15]. As a consequence, during the descending cathodic branch in the

oxygen-containing compartment, the values of current start to increase at about -1000 mV SCE, and

a much reduced hysteresis effect is observed after the potential is reversed. This indicates that, for

the etched alloy, no activation of the cathodic sites for oxygen reduction is necessary, since the

etching treatment has already provided available sites. Conversely, in the oxygen-free compartment,

the cathodic reaction of hydrogen evolution increases significantly below -1300 mV SCE and, once

the cathodic areas are activated, they remain effective after the potential is reversed to about -

1100 mV SCE. The pitting potential is not affected by the oxygen concentration but, as in the

previous case, the maximum pitting current is higher in the oxygen containing compartment.

Etching does not increase the pitting potential with respect to the as-received alloy. With reference

to a real corrosion condition, the information gathered here suggests that once an anodic event

consumes aluminium aluminum , new cathodic sites are also generated on the alloy matrix by

localized copper enrichment.

A similar experiment performed on the AA2024T3 alloy shows that, in the oxygen-containing

compartment, the cathodic reaction of oxygen reduction can proceed under diffusion control below

-600 mV SCE, since the value of current is independent of the applied potential over a relatively

wide range. This suggests that the alloy surface always presents a very high number of active

cathodic sites. Only at about -1100 mV SCE, an increase in current is observed, indicating that the

contribution to the total current from hydrogen evolution becomes significant. In the deareated

compartment, a progressive increase of the current due to the hydrogen evolution is evident with

decreasing potential, with current values similar to those observed for the etched alloy during the

ascending cathodic branch. This indicates that the alloy provides a significant number of effective

cathodic sites for hydrogen evolution, which require little activation. The pitting potential for the

alloy is higher (about 100 mV) than the model Al-Cu alloy. This can be related to the increased

presence of copper with respect to the model alloy.

Summarizing the results and translating the linear polarization behavior to a real corrosion

condition at the (freely variable) open circuit potential, it can be concluded that, during the early

stages of immersion, all the cathodic activity is located on the second phase material, as indicated

by the high cathodic current of the AA 2024 T3 alloy at all potentials. However, as corrosion

progress, an enriched layer is generated on the alloy matrix and the behavior changes from the


behavior of the unetched model alloy to that recorded for the etched Al-Cu 1000 ppm alloy.

Therefore, if the corrosion potential drops due to localized corrosion events, significant

contributions to the cathodic reaction of oxygen reduction can arise from local sites on the matrix

where corrosion events have produced a copper-enriched layer. Conversely, little contribution from

hydrogen evolution is expected to arise from enrichment, even at low potentials, due to the low

values of cathodic current measured for the etched specimen in the deareated compartment.

0.01

0.1

1

10

100

1000

-1400 -1200 -1000 -800 -600

0.1

1

10

100

1000

-1400 -1200 -1000 -800 -600

0.1

1

10

100

1000

a

Deaerated Compartment

Aerated Compartment

E

Current Density / µA cm

-2

S

b


-2


Aerated Compartment


Aerated Compartment

c


-2

Potential / V

Figure 1. Polarization curves obtained in aerated and deareated conditions with the split cell setup:

a) degreased Al-Cu 1000 ppm. wt. alloy, b) degreased, etched and desmutted Al-Cu 1000 ppm. wt.

alloy and c) degreased AA2024 T3 aluminium alloy. Sweep rate 10 mV min-1. Arrows in a) indicate

direction of polarization; S: polarization start, E: polarization end

Split cell applied to inhibition studies. Having achieved a fundamental understanding of the local

and general phenomena associated with the multiphase alloy and to the dynamic enrichment of

copper at anodic locations, the split cell technique has been employed to investigate the effect of

behavior of the unetched model alloy to that recorded for the etched Al-Cu 1000 ppm alloy.

Therefore, if the corrosion potential drops due to localized corrosion events, significant

contributions to the cathodic reaction of oxygen reduction can arise from local sites on the matrix

where corrosion events have produced a copper-enriched layer. Conversely, little contribution from

hydrogen evolution is expected to arise from enrichment, even at low potentials, due to the low

values of cathodic current measured for the etched specimen in the deareated compartment.

0.01

0.1

1

10

100

1000

-1400 -1200 -1000 -800 -600

0.1

1

10

100

1000

-1400 -1200 -1000 -800 -600

0.1

1

10

100

1000

a


Aerated Compartment

E


-2

S

b


-2


Aerated Compartment


Aerated Compartment

c


-2

Potential / V

Figure 1. Polarization curves obtained in aerated and deareated conditions with the split cell setup:

a) degreased Al-Cu 1000 ppm. wt. alloy, b) degreased, etched and desmutted Al-Cu 1000 ppm. wt.

alloy and c) degreased AA2024 T3 aluminium alloy. Sweep rate 10 mV min-1. Arrows in a) indicate

direction of polarization; S: polarization start, E: polarization end

Split cell applied to inhibition studies. Having achieved a fundamental understanding of the local

and general phenomena associated with the multiphase alloy and to the dynamic enrichment of

copper at anodic locations, the split cell technique has been employed to investigate the effect of


corrosion inhibitors. This is achieved by connecting the two specimens through a ZRA, without

applying any external polarization but utilizing the differential aeration effect. In Figures 2 and 3,

the results from split cell experiments on chromic acid and sodium tartrate respectively added in to

the 3.5% NaCl solution in the anodic and cathodic compartment are presented. During the initial

stage at the commencement of the experiments, the net currents are generally relatively low and are

not reproducible due to preferential corrosion of one of the two specimens. When the gas flow is

established, an increase in the net current is observed due to the generation of the differentially

aerated conditions. Correspondingly, as the gas flow is started, a small decrease in potential is

generally observed. This relates to the deactivation of the cathodic sites on the surface of the

specimen in the anodic compartment. Specifically, during the split cell experiment, the corrosion

potential depends upon the reaction kinetics, and the number and size of the anodic and cathodic

sites on the two specimens. Therefore, as the gas flows, oxygen is removed from the anodic

compartment by the nitrogen gas and, consequently, the cathodic sites on the anodic specimen

became deactivated. Since the two specimens are short circuited by the ZRA, they have the same

potential. Therefore, a reduction of cathodic activity on one specimen results in a decrease in

potential, since the anodic activity is not affected by the de-aeration. After this initial potential drop,

a slow potential recovery is generally observed. This is due to the increasing availability of cathodic

sites on the specimen in the cathodic compartment due to the progression of the corrosion process,

as indicated by the linear polarization results. As the alloy matrix is locally attacked, a copper-rich

layer is developed and increased cathodic activity is possible. As this proceeds on the specimen in

the aerated compartment, the polarization induced by the de-activation of the cathodic site on the

anodic specimen becomes less important.

Having considered the effects of differential aeration on current and potential, the influence of

addition of inhibitors can be examined. In the case of chromic acid, addition to the anodic side

results in a negative current spike, followed by a positive current surge and a slow current decay.

The potential, conversely, shows a positive spike and a negative overshoot, followed by a slow

decay. The negative current spike can be attributed to a burst of charge due to the injection of Cr6+

ions in the oxygen free compartment which can readily be reduced to Cr3+ species on cathodic sites

on the specimen. Therefore, the anodic specimen transiently becomes cathodic due to the

availability of electron acceptors in the solution. In agreement, the potential shows a positive spike

indicating an increase in overall cathodic activity. As all the cathodic sites on the anodic specimen

are covered with reaction products, the direction of the current is restored. Subsequently, a

temporary current increase is observed, due to the anodic activation promoted by the pH reduction

following chromic acid addition. However, this activation is only temporary, since chromate ions

corrosion inhibitors. This is achieved by connecting the two specimens through a ZRA, without

applying any external polarization but utilizing the differential aeration effect. In Figures 2 and 3,

the results from split cell experiments on chromic acid and sodium tartrate respectively added in to

the 3.5% NaCl solution in the anodic and cathodic compartment are presented. During the initial

stage at the commencement of the experiments, the net currents are generally relatively low and are

not reproducible due to preferential corrosion of one of the two specimens. When the gas flow is

established, an increase in the net current is observed due to the generation of the differentially

aerated conditions. Correspondingly, as the gas flow is started, a small decrease in potential is

generally observed. This relates to the deactivation of the cathodic sites on the surface of the

specimen in the anodic compartment. Specifically, during the split cell experiment, the corrosion

potential depends upon the reaction kinetics, and the number and size of the anodic and cathodic

sites on the two specimens. Therefore, as the gas flows, oxygen is removed from the anodic

compartment by the nitrogen gas and, consequently, the cathodic sites on the anodic specimen

became deactivated. Since the two specimens are short circuited by the ZRA, they have the same

potential. Therefore, a reduction of cathodic activity on one specimen results in a decrease in

potential, since the anodic activity is not affected by the de-aeration. After this initial potential drop,

a slow potential recovery is generally observed. This is due to the increasing availability of cathodic

sites on the specimen in the cathodic compartment due to the progression of the corrosion process,

as indicated by the linear polarization results. As the alloy matrix is locally attacked, a copper-rich

layer is developed and increased cathodic activity is possible. As this proceeds on the specimen in

the aerated compartment, the polarization induced by the de-activation of the cathodic site on the

anodic specimen becomes less important.

Having considered the effects of differential aeration on current and potential, the influence of

addition of inhibitors can be examined. In the case of chromic acid, addition to the anodic side

results in a negative current spike, followed by a positive current surge and a slow current decay.

The potential, conversely, shows a positive spike and a negative overshoot, followed by a slow

decay. The negative current spike can be attributed to a burst of charge due to the injection of Cr6+

ions in the oxygen free compartment which can readily be reduced to Cr3+ species on cathodic sites

on the specimen. Therefore, the anodic specimen transiently becomes cathodic due to the

availability of electron acceptors in the solution. In agreement, the potential shows a positive spike

indicating an increase in overall cathodic activity. As all the cathodic sites on the anodic specimen

are covered with reaction products, the direction of the current is restored. Subsequently, a

temporary current increase is observed, due to the anodic activation promoted by the pH reduction

following chromic acid addition. However, this activation is only temporary, since chromate ions


act as a mild anodic inhibitor, as suggested by the progressive decrease of current and by previous

results on the effect of addition of sodium chromate [6].

However, the more evident effect of sodium chromate is the cathodic inhibition, observed when the

addition is performed in the aerated compartment. As sodium chromate is added to the cathodic

compartment, a positive current spike is observed as a result of the increased availability of

electrons acceptors on the surface of the cathodic specimens. However, after a relatively short time,

complete coverage of the cathodic sites by the reaction products take place, and the current drops to

very low values since further reduction reactions are inhibited. Accordingly, a positive potential

spike, associated with increased overall cathodic activity is observed after inhibitor addition,

followed by a potential drop due to de-activation of the cathodic sites on the cathodic specimen due

to coverage by reaction products. From these results, it can be concluded that chromic acid acts both

as an anodic and cathodic inhibitor and that a fast reduction reaction takes place preferentially on

the cathodic sites, blocking oxygen reduction.

0 500 1000 1500 2000 2500 3000 3500 4000

-20

-15

-10

-5

0

5

10

15

20

25 Inhibitor IN

Current density / µA cm

-2

a

0 500 1000 1500 2000 2500 3000 3500 4000

-20

-15

-10

-5

0

5

10

15

20

25

b

No Gas Gas ON Inhibitor INNo Gas

0 500 1000 1500 2000 2500 3000 3500 4000

-0.74

-0.72

-0.70

-0.68

-0.66

-0.64

-0.62

-0.60

c

Gas ON

Potential / V (SCE)

Time / s

0 500 1000 1500 2000 2500 3000 3500 4000

-0.74

-0.72

-0.70

-0.68

-0.66

-0.64

-0.62

-0.60

d

Time / s

Figure 2. Current density vs. time plots (a,b) and potential vs. time plots (c,d) for 100 ppm. wt.

chromic acid added in the anodic (a,c) and cathodic (b,d) compartments.

act as a mild anodic inhibitor, as suggested by the progressive decrease of current and by previous

results on the effect of addition of sodium chromate [6].

However, the more evident effect of sodium chromate is the cathodic inhibition, observed when the

addition is performed in the aerated compartment. As sodium chromate is added to the cathodic

compartment, a positive current spike is observed as a result of the increased availability of

electrons acceptors on the surface of the cathodic specimens. However, after a relatively short time,

complete coverage of the cathodic sites by the reaction products take place, and the current drops to

very low values since further reduction reactions are inhibited. Accordingly, a positive potential

spike, associated with increased overall cathodic activity is observed after inhibitor addition,

followed by a potential drop due to de-activation of the cathodic sites on the cathodic specimen due

to coverage by reaction products. From these results, it can be concluded that chromic acid acts both

as an anodic and cathodic inhibitor and that a fast reduction reaction takes place preferentially on

the cathodic sites, blocking oxygen reduction.

0 500 1000 1500 2000 2500 3000 3500 4000

-20

-15

-10

-5

0

5

10

15

20

25 Inhibitor IN


-2

a

0 500 1000 1500 2000 2500 3000 3500 4000

-20

-15

-10

-5

0

5

10

15

20

25

b

No Gas Gas ON Inhibitor INNo Gas

0 500 1000 1500 2000 2500 3000 3500 4000

-0.74

-0.72

-0.70

-0.68

-0.66

-0.64

-0.62

-0.60

c

Gas ON

Potential / V (SCE)

Time / s

0 500 1000 1500 2000 2500 3000 3500 4000

-0.74

-0.72

-0.70

-0.68

-0.66

-0.64

-0.62

-0.60

d

Time / s


chromic acid added in the anodic (a,c) and cathodic (b,d) compartments.


Similar experiments performed by adding sodium tartrate to each compartment show a contrasting

behavior. Specifically, a significant decrease in the measured current is revealed when the addition

is made to the anodic compartment, with little effect is associated with the addition in the cathodic

compartment, in agreement with previous results on tartaric acid [5]. Interestingly, when the sodium

tartrate is added either to the anodic or cathodic compartment, no significant effects on the potential

are revealed, even if the current is significantly reduced in the case of the anodic addition as

expected for a purely anodic inhibitor. For a cathodic inhibitor added to the cathodic compartment,

all the cathodic sites on both specimens (deareation on one side and inhibition on the other prevent

cathodic reactions) are virtually deactivated, producing a consequent potential drop. Conversely, for

an anodic inhibitor added to the anodic compartment, the anodic reactions on the cathodic specimen

are still possible since aeration does not restrict the anodic reactions, and the effect on potential is

reduced.

This is supported by the linear polarization experiments, indicating that corrosion on the practical

alloy is cathodically, not anodically, limited; therefore, a progressive reduction in the anodic

activity on one specimen is readily balanced and does not change dramatically the corrosion

potential.

-500 0 500 1000 1500 2000 2500 3000 3500 4000

-10

-5

0

5

10

15

20

25 Inhibitor INGas ONNo Gas


-2

A (s)

a b

dc

-500 0 500 1000 1500 2000 2500 3000 3500 4000

-10

-5

0

5

10

15

20


D (s)

-500 0 500 1000 1500 2000 2500 3000 3500 4000

-0.70

-0.68

-0.66

-0.64

-0.62

-0.60

Potential / V (SCE)

Time / s

-500 0 500 1000 1500 2000 2500 3000 3500 4000

-0.70

-0.69

-0.68

-0.67

-0.66

-0.65

-0.64

-0.63

-0.62

-0.61

-0.60

Time / s


sodium tartrate added in the anodic (a,c) and cathodic (b,d) compartments.

Similar experiments performed by adding sodium tartrate to each compartment show a contrasting

behavior. Specifically, a significant decrease in the measured current is revealed when the addition

is made to the anodic compartment, with little effect is associated with the addition in the cathodic

compartment, in agreement with previous results on tartaric acid [5]. Interestingly, when the sodium

tartrate is added either to the anodic or cathodic compartment, no significant effects on the potential

are revealed, even if the current is significantly reduced in the case of the anodic addition as

expected for a purely anodic inhibitor. For a cathodic inhibitor added to the cathodic compartment,

all the cathodic sites on both specimens (deareation on one side and inhibition on the other prevent

cathodic reactions) are virtually deactivated, producing a consequent potential drop. Conversely, for

an anodic inhibitor added to the anodic compartment, the anodic reactions on the cathodic specimen

are still possible since aeration does not restrict the anodic reactions, and the effect on potential is

reduced.

This is supported by the linear polarization experiments, indicating that corrosion on the practical

alloy is cathodically, not anodically, limited; therefore, a progressive reduction in the anodic

activity on one specimen is readily balanced and does not change dramatically the corrosion

potential.

-500 0 500 1000 1500 2000 2500 3000 3500 4000

-10

-5

0

5

10

15

20



-2

A (s)

a b

dc

-500 0 500 1000 1500 2000 2500 3000 3500 4000

-10

-5

0

5

10

15

20


D (s)

-500 0 500 1000 1500 2000 2500 3000 3500 4000

-0.70

-0.68

-0.66

-0.64

-0.62

-0.60

Potential / V (SCE)

Time / s

-500 0 500 1000 1500 2000 2500 3000 3500 4000

-0.70

-0.69

-0.68

-0.67

-0.66

-0.65

-0.64

-0.63

-0.62

-0.61

-0.60

Time / s


sodium tartrate added in the anodic (a,c) and cathodic (b,d) compartments.


Conclusions

In this work, the application of split cell techniques to corrosion studies has been demonstrated.

Further, by performing linear polarization on identical specimens immersed in the two separate

compartment of the split cell, the behavior under aerated and deareated conditions, representative of

the different microenvironments that can exist on the alloy surface during corrosion, it is possible to

identify the cathodic activity of matrix and second phase material with respect to oxygen reduction

and hydrogen evolution. It was revealed that the copper-containing aluminum matrix is not an

effective cathode for oxygen reduction or hydrogen evolution, unless an enriched layer is produced

by etching or corrosion. Etched model alloys are representative of locations where anodic events on

the corroding surface have produced a copper-rich layer. This layer provides a relatively efficient

cathode for the reaction of oxygen reduction, but is not particularly effective for hydrogen

evolution. The linear polarization results provide the foundation for the interpretation of the split

cell experiments performed without the need for external polarization to investigate the inhibition

process. It has been demonstrated that by examination of the current and potential transients of two

freely-corroding, differentially aerated specimens following inhibitor addition, it is possible to

identify the anodic or cathodic inhibition effect and to investigate the timescale for inhibition in a

relatively unperturbed condition. It was shown that chromic acid is effective as a cathodic and

anodic inhibitor, while sodium tartrate is only effective as an anodic inhibitor. Since the corrosion

of the alloy is cathodically controlled, when cathodic inhibition is operative the potential decreases

significantly.

Acknowledgements

EPSRC is acknowledged for provision of financial support through the LATEST Portolio

Partnership.

Bibliography

[1] M. Curioni, P. Skeleton, G. E. Thompson, J. Ferguson, Proc. Advanced Materials Research,

2008.

[2] M. Schem, T. Schmidt, J. Gerwann, M. Wittmar, M. Veith, G. E. Thompson, I.S. Molchan,

T.Hashimoto, P. Skeldon, A. R. Phani, S. Santucci, M. L. Zheludkevich, Corrosion Science, 51

(2009) 2304-2315.

[3] M. Pourbaix, Corrosion Science, 14 (1974) 25-82.

[4] W. J. Clark, J. D. Ramsey, R. L. McCreery, G. S. Frankel, Journal of the Electrochemical

Society, 149 (2002) B179-B185.

Conclusions

In this work, the application of split cell techniques to corrosion studies has been demonstrated.

Further, by performing linear polarization on identical specimens immersed in the two separate

compartment of the split cell, the behavior under aerated and deareated conditions, representative of

the different microenvironments that can exist on the alloy surface during corrosion, it is possible to

identify the cathodic activity of matrix and second phase material with respect to oxygen reduction

and hydrogen evolution. It was revealed that the copper-containing aluminum matrix is not an

effective cathode for oxygen reduction or hydrogen evolution, unless an enriched layer is produced

by etching or corrosion. Etched model alloys are representative of locations where anodic events on

the corroding surface have produced a copper-rich layer. This layer provides a relatively efficient

cathode for the reaction of oxygen reduction, but is not particularly effective for hydrogen

evolution. The linear polarization results provide the foundation for the interpretation of the split

cell experiments performed without the need for external polarization to investigate the inhibition

process. It has been demonstrated that by examination of the current and potential transients of two

freely-corroding, differentially aerated specimens following inhibitor addition, it is possible to

identify the anodic or cathodic inhibition effect and to investigate the timescale for inhibition in a

relatively unperturbed condition. It was shown that chromic acid is effective as a cathodic and

anodic inhibitor, while sodium tartrate is only effective as an anodic inhibitor. Since the corrosion

of the alloy is cathodically controlled, when cathodic inhibition is operative the potential decreases

significantly.

Acknowledgements

EPSRC is acknowledged for provision of financial support through the LATEST Portolio

Partnership.

Bibliography

[1] M. Curioni, P. Skeleton, G. E. Thompson, J. Ferguson, Proc. Advanced Materials Research,

2008.

[2] M. Schem, T. Schmidt, J. Gerwann, M. Wittmar, M. Veith, G. E. Thompson, I.S. Molchan,

T.Hashimoto, P. Skeldon, A. R. Phani, S. Santucci, M. L. Zheludkevich, Corrosion Science, 51

(2009) 2304-2315.

[3] M. Pourbaix, Corrosion Science, 14 (1974) 25-82.

[4] W. J. Clark, J. D. Ramsey, R. L. McCreery, G. S. Frankel, Journal of the Electrochemical

Society, 149 (2002) B179-B185.


[5] M. Curioni, P. Skeldon, E. Koroleva, G. E. Thompson, J. Ferguson, Journal of the

Electrochemical Society, 156 (2009) C147-C153.

[6] N. C. Rosero-Navarro, M. Curioni, R. Bingham, A. Durán, M. Aparicio, R. A. Cottis,

G. E. Thompson, Corrosion Science, in press (2010).

[7] Totten, MacKenzie, Handbook of Aluminium, Marcel Dekker, 2003.

[8] M. Curioni, M. Saenz De Miera, P. Skeldon, G. E. Thompson, J. Ferguson, Journal of the


[9] H. Habazaki, M. A. Paez, K. Shimizu, P. Skeldon, G. E. Thompson, G. C. Wood, X. Zhou,

Corrosion Science, 38 (1996) 1033-1042.

[10] P. Skeldon, X. Zhou, G. E. Thompson, G. C. Wood, H. Habazaki, K. Shimizu, Thin Solid

Films, 293 (1997) 327-332.

[11] I. Pires, L. Quintino, C. M. Rangel, G. E. Thompson, P. Skeldon, X. Zhou, Transactions of the

Institute of Metal Finishing (UK). Vol. 78, no. 5, pp. 179-185. Sept. 2000, (2000).

[12] S. Garcia-Vergara, P. Skeldon, G. E. Thompson, P. Bailey, T. C. Q. Noakes, H. Habazaki,

K. Shimizu, Applied Surface Science, 205 (2003) 121-127.

[13] Y. Liu, F. Colin, P. Skeldon, G. E. Thompson, X. Zhou, H. Habazaki, K. Shimizu, Corrosion

Science, 45 (2003) 1539-1544.

[14] M. Curioni, F. Roeth, S. J. Garcia-Vergara, T. Hashimoto, P. Skeldon, G. E. Thompson,

J. Ferguson, Surface and Interface Analysis, 42 (2010) 234-240.

[15] Y. Liu, M. A. Arenas, S.J. Garcia-Vergara, T. Hashimoto, P. Skeldon, G. E. Thompson,

H. Habazaki, P. Bailey, T. C. Q. Noakes, Corrosion Science, 50 (2008) 1475-1480.

[5] M. Curioni, P. Skeldon, E. Koroleva, G. E. Thompson, J. Ferguson, Journal of the


[6] N. C. Rosero-Navarro, M. Curioni, R. Bingham, A. Durán, M. Aparicio, R. A. Cottis,

G. E. Thompson, Corrosion Science, in press (2010).

[7] Totten, MacKenzie, Handbook of Aluminium, Marcel Dekker, 2003.

[8] M. Curioni, M. Saenz De Miera, P. Skeldon, G. E. Thompson, J. Ferguson, Journal of the


[9] H. Habazaki, M. A. Paez, K. Shimizu, P. Skeldon, G. E. Thompson, G. C. Wood, X. Zhou,

Corrosion Science, 38 (1996) 1033-1042.

[10] P. Skeldon, X. Zhou, G. E. Thompson, G. C. Wood, H. Habazaki, K. Shimizu, Thin Solid

Films, 293 (1997) 327-332.

[11] I. Pires, L. Quintino, C. M. Rangel, G. E. Thompson, P. Skeldon, X. Zhou, Transactions of the

Institute of Metal Finishing (UK). Vol. 78, no. 5, pp. 179-185. Sept. 2000, (2000).

[12] S. Garcia-Vergara, P. Skeldon, G. E. Thompson, P. Bailey, T. C. Q. Noakes, H. Habazaki,

K. Shimizu, Applied Surface Science, 205 (2003) 121-127.

[13] Y. Liu, F. Colin, P. Skeldon, G. E. Thompson, X. Zhou, H. Habazaki, K. Shimizu, Corrosion

Science, 45 (2003) 1539-1544.

[14] M. Curioni, F. Roeth, S. J. Garcia-Vergara, T. Hashimoto, P. Skeldon, G. E. Thompson,

J. Ferguson, Surface and Interface Analysis, 42 (2010) 234-240.

[15] Y. Liu, M. A. Arenas, S.J. Garcia-Vergara, T. Hashimoto, P. Skeldon, G. E. Thompson,

H. Habazaki, P. Bailey, T. C. Q. Noakes, Corrosion Science, 50 (2008) 1475-1480.


Surface Protection for Aircraft Maintenance by means of Zinc Rich Primers

Georg Bockmair, Katharina Kranzeder

Wehrwissenschaftliches Institut für Werk- und Betriebsstoffe (WIWeB),

Institutsweg 1, 85435 Erding, Germany

[email protected] [email protected]

Keywords: Surface protection, zinc rich primers, corrosion inhibitors, aircraft maintenance

Abstract

The aerospace industry urgently needs environmentally friendly materials and processes for

corrosion protection of aluminium alloys in aircraft structures. Until now this has been achieved by

hexavalent chromium based compounds in either surface pre-treatments or primers. Due to its

carcinogenic properties the use of chromates is restricted and a ban is expected soon. Up to now an

all over recognized replacement of chromates is not available for aircraft maintenance, although a

lot of research has been done and promising results also exist for some chromium-free conversion

coatings and for magnesium rich primers.

WIWeB found out in laboratory scale and by flight trials that thin layers of zinc rich primers, if

applied with dry film thickness of 10 – 20 µm, can be used successfully to prevent corrosion on

aluminium for aircraft. Solvent based as well as water based zinc rich primers have been tested. The

major part of the work presented is from further investigations which show, that another great

improvement can be achieved, when thin layers of organic adhesion inhibitors like 2-

aminopropyltriethoxysilane and 4-t-butylbenzoic acid are applied on the unclad Al 2024 panels,

which had been scrubbed with abrasive pads before. This process is followed by the application of

the thin film of zinc rich epoxy primer before the usual 2-pack epoxy primer is applied.

Introduction



highly efficient hexavalent chromium based compounds in either surface pre-treatments or primers.

This also includes the application of the chromate containing wash primer for aluminium pre-

treatment in aircraft maintenance. Because of its carcinogenic properties the use of chromates will

be restricted by the European REACH legislation in the near future. Due to the existing European

legislation for hazardous materials chromates up to now may only be used, when no alternatives

exist. Therefore intensive efforts to substitute chromates in surface protection have begun.

Chromates have always been the best choice for high specific strength aircraft aluminium, because

of their extremely good corrosion behaviour, including their efficiency to resist intergranular and

exfoliation corrosion. When looking for alternatives there is the possibility to put the focus either

more on the pre-treatment of the surface or on the primer. Substitutes of the first category comprise

passivating solutions with compounds of Ti, Cr, Zr, Mn or other metals, which build stable oxide

layers [1, 2, 3], which are a very good adhesive base for the following organic coating, similar to

anodizing layers, which are well known for new built aircraft.

Surface Protection for Aircraft Maintenance by means of Zinc Rich Primers

Georg Bockmair, Katharina Kranzeder

Wehrwissenschaftliches Institut für Werk- und Betriebsstoffe (WIWeB),

Institutsweg 1, 85435 Erding, Germany

[email protected] [email protected]

Keywords: Surface protection, zinc rich primers, corrosion inhibitors, aircraft maintenance

Abstract



hexavalent chromium based compounds in either surface pre-treatments or primers. Due to its

carcinogenic properties the use of chromates is restricted and a ban is expected soon. Up to now an

all over recognized replacement of chromates is not available for aircraft maintenance, although a

lot of research has been done and promising results also exist for some chromium-free conversion

coatings and for magnesium rich primers.

WIWeB found out in laboratory scale and by flight trials that thin layers of zinc rich primers, if

applied with dry film thickness of 10 – 20 µm, can be used successfully to prevent corrosion on

aluminium for aircraft. Solvent based as well as water based zinc rich primers have been tested. The

major part of the work presented is from further investigations which show, that another great

improvement can be achieved, when thin layers of organic adhesion inhibitors like 2-

aminopropyltriethoxysilane and 4-t-butylbenzoic acid are applied on the unclad Al 2024 panels,

which had been scrubbed with abrasive pads before. This process is followed by the application of

the thin film of zinc rich epoxy primer before the usual 2-pack epoxy primer is applied.

Introduction



highly efficient hexavalent chromium based compounds in either surface pre-treatments or primers.

This also includes the application of the chromate containing wash primer for aluminium pre-

treatment in aircraft maintenance. Because of its carcinogenic properties the use of chromates will

be restricted by the European REACH legislation in the near future. Due to the existing European

legislation for hazardous materials chromates up to now may only be used, when no alternatives

exist. Therefore intensive efforts to substitute chromates in surface protection have begun.

Chromates have always been the best choice for high specific strength aircraft aluminium, because

of their extremely good corrosion behaviour, including their efficiency to resist intergranular and

exfoliation corrosion. When looking for alternatives there is the possibility to put the focus either

more on the pre-treatment of the surface or on the primer. Substitutes of the first category comprise

passivating solutions with compounds of Ti, Cr, Zr, Mn or other metals, which build stable oxide

layers [1, 2, 3], which are a very good adhesive base for the following organic coating, similar to

anodizing layers, which are well known for new built aircraft.

Instead of solid oxide layers on the Al surface after a conversion process extremely thin layers of

organic adhesion promoters such as silanes should also be very promising [4].

Such interface active compounds slow down the corrosion process by suppressing the cathodic

reaction by preventing the diffusion of ions to the metallic surface.

The other possibility of improving the corrosion behaviour of a metal part relies on the formulation

of paints with appropriate binders and highly efficient anti-corrosion pigments. Normally 2-pack

epoxy primers are the first choice followed by the slightly less corrosion resistant 2-pack

polyurethane primers. 2-Pack polyurethane binders are used for topcoats. The anti-corrosion

pigments in the primer must not dissolve too easily in water, otherwise blistering of the coating may

occure. Magnesium (Mg) rich primers have been reported to be very effective as anti-corrosion

paints due to the great electrochemical potential difference of Mg to the Al substrate, which is to be

protected [5, 6].

Another procedure to protect Al alloys after removal of the old coating is based on 2-pack zinc rich

epoxy primers. Zinc rich primers predominantly are used for protection of steel constructions. Here

they show excellent corrosion protection. The mechanism of corrosion protection on aluminium

comprises two steps. In the beginning after damaging the coating and exposing the Al substrate a

cathodic protection is assumed, when there is a conductive connection between the Al substrate and

the Zn rich primer [7]. When corrosion goes on, the barrier protection prevails, because corrosion

products become more compact and seal the Al alloy underneath. In contrast to Zn rich primers for

steel constructions due to the lack of a real rough surface beforehand and the low cohesive strength

of Zn rich coatings a lower dry film thickness (10 – 20 µm) is recommended.

Other concepts like self-healing primers [8, 9] or the sol-gel technology are also reported to be very

useful. The aim of this work, however, is to show that Zn epoxy primers are not only a very

powerful means for the protection of steel, but also for maintenance of aircraft aluminium.

Furthermore our research shall give evidence that remarkable improvements can be achieved, when

certain corrosion inhibitors of the adsorption type are used between surface roughening and Zn rich

priming.

Experimental

Whereas test panels for laboratory testing have been degreased and roughened by using abrasive

pads, the conditions of the surface for the flight trial the wings of a Tornado fighter have been

achieved by plastic media blasting and a subsequent detergent cleaning.

The evaluation of the appearance of the Tornado wings after almost one year of flying including

four inspections every 80 flight hours with cross cut adhesion testing [9] showed excellent

behaviour (grade 0 – 1). Corrosion testing with test panels in the laboratory in accordance with ISO

4623-2 / ISO 4628-10 [10], 1000 h (filiform corrosion test) respectively ISO 11997-1 (cycle B,

6 weeks) / ISO 4628-8 [10] resulted in ≤ 4 mm for the majority length of filiform threads and

≤ 1 mm for the degree of undercutting and corrosion. Solvent based as well as water based Zn rich

primers have been tested by WIWeB. Both kinds of paint gave evidence for very good behaviour in

the flight trial. For depot overhaul in work shops of the German Air Force, however, only solvent

based primers are recommended currently due to the difficulties of having a real clean surface, as

little oil leakages have been observed, which deteriorated the wetting of the surface in the time

between cleaning and painting. This is a clear disadvantage when priming old aircraft with water

based paints, as these have absolutely no tolerance to oily contaminants.

Further research in WIWeB laboratory showed that even more progress could be achieved, when

the Al 2024 panels, which had been treated by an abrasive pad before, were dipped after another

Instead of solid oxide layers on the Al surface after a conversion process extremely thin layers of

organic adhesion promoters such as silanes should also be very promising [4].

Such interface active compounds slow down the corrosion process by suppressing the cathodic

reaction by preventing the diffusion of ions to the metallic surface.

The other possibility of improving the corrosion behaviour of a metal part relies on the formulation

of paints with appropriate binders and highly efficient anti-corrosion pigments. Normally 2-pack

epoxy primers are the first choice followed by the slightly less corrosion resistant 2-pack

polyurethane primers. 2-Pack polyurethane binders are used for topcoats. The anti-corrosion

pigments in the primer must not dissolve too easily in water, otherwise blistering of the coating may

occure. Magnesium (Mg) rich primers have been reported to be very effective as anti-corrosion

paints due to the great electrochemical potential difference of Mg to the Al substrate, which is to be

protected [5, 6].

Another procedure to protect Al alloys after removal of the old coating is based on 2-pack zinc rich

epoxy primers. Zinc rich primers predominantly are used for protection of steel constructions. Here

they show excellent corrosion protection. The mechanism of corrosion protection on aluminium

comprises two steps. In the beginning after damaging the coating and exposing the Al substrate a

cathodic protection is assumed, when there is a conductive connection between the Al substrate and

the Zn rich primer [7]. When corrosion goes on, the barrier protection prevails, because corrosion

products become more compact and seal the Al alloy underneath. In contrast to Zn rich primers for

steel constructions due to the lack of a real rough surface beforehand and the low cohesive strength

of Zn rich coatings a lower dry film thickness (10 – 20 µm) is recommended.

Other concepts like self-healing primers [8, 9] or the sol-gel technology are also reported to be very

useful. The aim of this work, however, is to show that Zn epoxy primers are not only a very

powerful means for the protection of steel, but also for maintenance of aircraft aluminium.

Furthermore our research shall give evidence that remarkable improvements can be achieved, when

certain corrosion inhibitors of the adsorption type are used between surface roughening and Zn rich

priming.

Experimental

Whereas test panels for laboratory testing have been degreased and roughened by using abrasive

pads, the conditions of the surface for the flight trial the wings of a Tornado fighter have been

achieved by plastic media blasting and a subsequent detergent cleaning.

The evaluation of the appearance of the Tornado wings after almost one year of flying including

four inspections every 80 flight hours with cross cut adhesion testing [9] showed excellent

behaviour (grade 0 – 1). Corrosion testing with test panels in the laboratory in accordance with ISO

4623-2 / ISO 4628-10 [10], 1000 h (filiform corrosion test) respectively ISO 11997-1 (cycle B,

6 weeks) / ISO 4628-8 [10] resulted in ≤ 4 mm for the majority length of filiform threads and

≤ 1 mm for the degree of undercutting and corrosion. Solvent based as well as water based Zn rich

primers have been tested by WIWeB. Both kinds of paint gave evidence for very good behaviour in

the flight trial. For depot overhaul in work shops of the German Air Force, however, only solvent

based primers are recommended currently due to the difficulties of having a real clean surface, as

little oil leakages have been observed, which deteriorated the wetting of the surface in the time

between cleaning and painting. This is a clear disadvantage when priming old aircraft with water

based paints, as these have absolutely no tolerance to oily contaminants.

Further research in WIWeB laboratory showed that even more progress could be achieved, when

the Al 2024 panels, which had been treated by an abrasive pad before, were dipped after another


4 hours into a 0.1 % solution (acetone) of corrosion inhibitor Asconium-144*, which consists of

amines of fatty acids and 3-Aminopropyltriethoxysilane and 4-t-Butylbenzoic acid.

Table 1 shows the results of cross cut adhesion testing and the majority length of threads after

filiform corrosion (ISO 4623-2, 1000 h, evaluation ISO 4628-10) as well as the degree of

delamination after the cyclic corrosion test (ISO 11997-1, cycle B, 6 weeks, evaluation

ISO 4628-8). A comparison of the results from Al 2024 laboratory panels shows that both corrosion

tests give by far the best results, when the triple pre-treatment (scrubbing, surface inhibiting, Zn

priming) is chosen.

Summary

The application of Zn epoxy primers with a dry film thickness of 10 – 20 µm has been reported to

be a powerful means for aircraft maintenance after roughening of the surface by abrasive pads or

plastic media blasting. In this work we could show in laboratory scale with unclad Al 2024 panels

that a further remarkable improvement of the corrosion protection could be achieved when a

subsequent application of a thin film containing corrosion inhibitor takes place before the Zn epoxy

primer is sprayed onto the panels.

Further work is necessary to find out, if the chemical composition of the corrosion inhibitor used, is

already optimized. In particular the influence of the individual components should be investigated.

Table 1 Adhesion and corrosion behaviour of 2pack epoxy primer after pre-treatment of

unclad Al 2024 panels by abrasive pad, diluted corrosion inhibitor Asconium 144

and Zn epoxy primer (dry film 20 µm)

* Ascotec Company, Saint-Etienne, France

Pre-treatment

Results

Cross cut

ISO 2409,

grade

Filiform corrosion,

majority length of

threads M [mm]

ISO 4623-2 /

ISO 4628-10, 1000 h

Cyclic corrosion,

ISO 11997-1,

cycle B, 6 weeks /

ISO 4628-8, degree of

delamination d [mm]

1 Abrasive pad +

2-pack Zn epoxy

primer

0 4.2

(fig. 1)

2.2

(fig. 3)

2 Abrasive pad +

application of

corrosion

inhibitor +

2-pack Zn epoxy

primer

0 2.1 (fig. 2)

0.1 (fig. 4)

no use of abrasive pad

3 2-Pack Zn epoxy

primer 1 6.0 2.3

no use of Zn epoxy primer

4 Abrasive pad +

application of

corrosion

inhibitor

0 11.7 2.4

4 hours into a 0.1 % solution (acetone) of corrosion inhibitor Asconium-144*, which consists of

amines of fatty acids and 3-Aminopropyltriethoxysilane and 4-t-Butylbenzoic acid.

Table 1 shows the results of cross cut adhesion testing and the majority length of threads after

filiform corrosion (ISO 4623-2, 1000 h, evaluation ISO 4628-10) as well as the degree of

delamination after the cyclic corrosion test (ISO 11997-1, cycle B, 6 weeks, evaluation

ISO 4628-8). A comparison of the results from Al 2024 laboratory panels shows that both corrosion

tests give by far the best results, when the triple pre-treatment (scrubbing, surface inhibiting, Zn

priming) is chosen.

Summary

The application of Zn epoxy primers with a dry film thickness of 10 – 20 µm has been reported to

be a powerful means for aircraft maintenance after roughening of the surface by abrasive pads or

plastic media blasting. In this work we could show in laboratory scale with unclad Al 2024 panels

that a further remarkable improvement of the corrosion protection could be achieved when a

subsequent application of a thin film containing corrosion inhibitor takes place before the Zn epoxy

primer is sprayed onto the panels.

Further work is necessary to find out, if the chemical composition of the corrosion inhibitor used, is

already optimized. In particular the influence of the individual components should be investigated.

Table 1 Adhesion and corrosion behaviour of 2pack epoxy primer after pre-treatment of

unclad Al 2024 panels by abrasive pad, diluted corrosion inhibitor Asconium 144

and Zn epoxy primer (dry film 20 µm)

* Ascotec Company, Saint-Etienne, France

Pre-treatment

Results

Cross cut

ISO 2409,

grade

Filiform corrosion,

majority length of

threads M [mm]

ISO 4623-2 /

ISO 4628-10, 1000 h

Cyclic corrosion,

ISO 11997-1,

cycle B, 6 weeks /

ISO 4628-8, degree of

delamination d [mm]

1 Abrasive pad +

2-pack Zn epoxy

primer

0 4.2

(fig. 1)

2.2

(fig. 3)

2 Abrasive pad +

application of

corrosion

inhibitor +

2-pack Zn epoxy

primer

0 2.1 (fig. 2)

0.1 (fig. 4)

no use of abrasive pad

3 2-Pack Zn epoxy

primer 1 6.0 2.3

no use of Zn epoxy primer

4 Abrasive pad +

application of

corrosion

inhibitor

0 11.7 2.4


Fig. 1

Corrosion behaviour, resistance to filiform corrosion (ISO 4623-2, 1000 h / ISO 4628-10),

Al 2024 unclad, after local paint stripping

pre-treatment: abrasive pad + 2-pack Zn epoxy primer (DFT 20 µm)

primer: 2-pack epoxy

Total DFT: 60 µm

Fig. 2


Al 2024 unclad, after local paint stripping,

pre-treatment: abrasive pad + application of corrosion inhibitor + 2-pack Zn epoxy primer

(DFT 20 µm)


Total DFT: 69 µm

Fig. 1


Al 2024 unclad, after local paint stripping



Total DFT: 60 µm

Fig. 2


Al 2024 unclad, after local paint stripping,


(DFT 20 µm)


Total DFT: 69 µm


Fig. 3

Corrosion behaviour, resistance to cyclic corrosion (ISO 11997-1, cycle B, 6 weeks /

ISO 4628-10), Al 2024 unclad



Total DFT: 60 µm

Fig. 4

Corrosion behaviour, resistance to cyclic corrosion (EN ISO 11997-1, cycle B, 6 weeks /

ISO 4628-10), Al 2024 unclad


(DFT 20 µm)


Total DFT: 67 µm

Fig. 3

Corrosion behaviour, resistance to cyclic corrosion (ISO 11997-1, cycle B, 6 weeks /

ISO 4628-10), Al 2024 unclad



Total DFT: 60 µm

Fig. 4

Corrosion behaviour, resistance to cyclic corrosion (EN ISO 11997-1, cycle B, 6 weeks /

ISO 4628-10), Al 2024 unclad


(DFT 20 µm)


Total DFT: 67 µm


References

[1] U. Jüptner: Tagung “Leichtmetall-Anwendungen – Neue Entwicklungen in der

Oberflächentechnik”, DFO, Düsseldorf, 16. – 17.03.2004, 50 - 56

[2] Ch. Ruhland: J. Oberflächentechn. 8, 56 – 59 (2003)

[3] T. Wendel: Tagung „Leichtmetall-Anwendungen – Neue Entwicklungen in der

Oberflächentechnik“, DFO, Düsseldorf, 16. – 17.03.2004, 32 – 48

[4] J. Mulder: „Vorbehandlung von Aluminium mit Silanen“, DFO/DGO-Tagung

„Leichtmetall-Anwendungen, Neue Entwicklungen in der Oberflächentechnik“, Münster,

2001

[5] M. Nanna, G. Bierwagen: Journ. of Coat. Techn., Research, Vol. 1, No. 2, April 2004, 69 –

80

[6] G. Bockmair, K. Kranzeder : Aluminium, 81 (2005), 777 – 779

[7] Rudolf: Tagung „Leichtmetall-Anwendungen – Neue Entwicklungen in der

Oberflächentechnik“, DFO, Düsseldorf, 16. – 17.03.2004

[8] N. Voevodin, D. Buhrmaster, V. Balbyshev, A. Khramov, J. Johnson, R. Mantz: “Non-

Chromated Coating Systems for Corrosion Protection of Aircraft Aluminium”, Tri-Service

Corrosion Conference, Orlando, Florida, 2005

[9] D. Raps: Ph D thesis, “Development of a Self-Healing Corrosion Protection Coating System

for High Strength Aluminium Alloys”, Technical University Munich, 2008

[10] ISO-Standards

ISO 4623-2

Paints and varnishes – determination of resistance to filiform corrosion – part 2: aluminium

substrates

ISO 4628-8

Paints and varnishes – evaluation of degradation of coatings – designation of quantity and

size of defects, and of intensity of uniform changes in appearance – part 8: assessment of

degree of delamination and corrosion around scribe

ISO 4628-10

Paints and varnishes - evaluation of degradation of coatings – designation of quantity and


degree of filiform corrosion

ISO 11997-1

Paints and varnishes – determination of resistance to cyclic corrosion conditions – part 1:

wet (salt fog) / dry / humidity

References

[1] U. Jüptner: Tagung “Leichtmetall-Anwendungen – Neue Entwicklungen in der

Oberflächentechnik”, DFO, Düsseldorf, 16. – 17.03.2004, 50 - 56

[2] Ch. Ruhland: J. Oberflächentechn. 8, 56 – 59 (2003)

[3] T. Wendel: Tagung „Leichtmetall-Anwendungen – Neue Entwicklungen in der

Oberflächentechnik“, DFO, Düsseldorf, 16. – 17.03.2004, 32 – 48

[4] J. Mulder: „Vorbehandlung von Aluminium mit Silanen“, DFO/DGO-Tagung

„Leichtmetall-Anwendungen, Neue Entwicklungen in der Oberflächentechnik“, Münster,

2001

[5] M. Nanna, G. Bierwagen: Journ. of Coat. Techn., Research, Vol. 1, No. 2, April 2004, 69 –

80

[6] G. Bockmair, K. Kranzeder : Aluminium, 81 (2005), 777 – 779

[7] Rudolf: Tagung „Leichtmetall-Anwendungen – Neue Entwicklungen in der

Oberflächentechnik“, DFO, Düsseldorf, 16. – 17.03.2004

[8] N. Voevodin, D. Buhrmaster, V. Balbyshev, A. Khramov, J. Johnson, R. Mantz: “Non-

Chromated Coating Systems for Corrosion Protection of Aircraft Aluminium”, Tri-Service

Corrosion Conference, Orlando, Florida, 2005

[9] D. Raps: Ph D thesis, “Development of a Self-Healing Corrosion Protection Coating System

for High Strength Aluminium Alloys”, Technical University Munich, 2008

[10] ISO-Standards

ISO 4623-2

Paints and varnishes – determination of resistance to filiform corrosion – part 2: aluminium

substrates

ISO 4628-8

Paints and varnishes – evaluation of degradation of coatings – designation of quantity and


degree of delamination and corrosion around scribe

ISO 4628-10

Paints and varnishes - evaluation of degradation of coatings – designation of quantity and


degree of filiform corrosion

ISO 11997-1

Paints and varnishes – determination of resistance to cyclic corrosion conditions – part 1:

wet (salt fog) / dry / humidity


Thin, Nanoparticulate Coatings for the Improvement of the Corrosion and Passivation Behavior of AZ Magnesium Alloys

Florian Feil, Wolfram Fijrbeth DECHEMA e.V., Karl-Winnacker-lnstitut,

60486 Frankfurt am Main, Germany

[email protected], [email protected]

Keywords: magnesium, passivation, sol-gel, corrosion inhibitors, electrochemical impedance spectroscopy, EIS.

Abstract. We developed multilayered, purely inorganic coatings for the corrosion protection of AZ magnesium alloys. Polymeric acid-catalyzed sols form relatively dense coatings, but any direct contact to the reactive magnesium substrate has to be avoided. However polymeric sols based on SO2, B2O3, A1203, Zr02 and up to 5% of lanthanide salts can be used to seal samples with prime coat based on aqueous nanoparticle dispersions. Without organic network modification, these sealings have to be kept thin to avoid cracks. However if the coating process with aqueous dispersions and polymeric sols is alternated, a kind of lamellar sandwich structure can be formed which stays crack-free up to several layers. The performance and the protective properties of these coatings were studied with different methods (EIS, salt spray tests and electron microscopy).

Introduction

Improving the corrosion performance of magnesium alloys is a crucial challenge for their increased application as construction materials. In contrast to aluminum based alloys, these reactive magnesium alloys do not form a stable, self-protective oxide layer at pH values lower than 12. Here we present a procedure to apply protective multilayered coatings onto magnesium alloys based on a combination of various approaches. Even inorganic corrosion inhibitors, i.e. lanthanide salts, could be included for improvement of the passivation behavior or even self-healing abilities. In contrast to most coating systems, which contain organic components in at least one part, these inorganic coatings could offer higher mechanical and thermal stability.

Coating Composition

Si02-nanoparticles have a high surface activity [I]. Thus, coatings based on Si02-nanoparticles can be densified at moderate temperatures, suitable even for heat sensitive magnesium alloys. Aqueous SO2-nanoparticle dispersions are inexpensive and obtainable from different vendors. For coating application aqueous solutions of borax, sodium phosphate, magnesium nitrate and lithium or potassium hydroxide for instance were added as binders and to further decrease the sintering temperature. Dip-coating application and properties of coatings based on ~evasil' 200were described elsewhere [2,3]. These coatings (typical composition: 80-85% Si02 8-1 1% B2O3, 2-3 % P2O5, 1-2% Na20, 1-2% K20, 1-2% Li20 and 1-2% MgO) are quite thin (200-500 nm) and stay crack-free even after heat treatment between 200-400 "C. However, complete densification is not possible under these mild conditions. An increase of the coating thickness can be achieved by multiple applications of the dispersions each time followed by a drying and sintering step. So far, up to 5 layers could be applied onto AZ91 by dip coating, resulting in crack-free coatings of 1.5 pm thickness.

Coatings based on polymeric sols are usually less porous than coatings based on particulate sols or dispersions [4,5,6], but usually are synthesized under acid catalysis. Because these acid stabilized


sols would react with the sensitive magnesium substrate, they cannot be applied directly. However coatings based on aqueous nanoparticle dispersion can act as primer for an additional sealing with a polymeric sol.

Usually organic substituents on silicon are introduced to reduce the degree of cross-linking and to get a higher flexibility. However, to retain the thermal and mechanical coating stability we abstained from organic network modifications. Instead we developed purely inorganic sealings based on Tetraethoxysilane (TEOS), Triethoxyborane (TEB), Zirconium(1V)propoxide (Zr(OPr)4) and Rare Earth Nitrates (RE = Sm(III), Ce(III), Eu(II1) or La(II1)) in iso-Propanol at pH 4 and a solid content of ca. 8%. A particular Zirconium content in the coating usually offers higher chemical stability, especially in alkaline environments. Rare Earth salts can act as cathodic corrosion inhibitors. These coatings were applied onto AZ31 and AZ91 plates, pretreated with an aqueous prime coat by dip coating (withdrawal speed 40 mmlmin). After sealing the surface becomes much smoother (Fig.la, here the dipping edge of the sealing on top of the prime coat is shown). Because the resulting, highly cross-linked coatings are very rigid, the maximum coating thickness is strongly limited. However thin, about 100 nm thick sealings on top of a particulate prime coat stay crack free even after sintering (Fig. 1 b), while the prime coat retains its porous, particulate character.

Fig. 1: SEM images of a sol-gel sealing on an aqueous prime coat on AZ91; sealing composition: 36.7% Si02, 52.6% Zr02, 5.8% B2O3, 4.8% Eu2O3; sintered at 250 "C12 h; a) surface at the dipping edge between primer and sealing; b) cross section prepared by focused argon ion beam.

To increase the coating thickness and to reduce the degree of cross-linking Triethoxysilane (TREOS) can be used as a silica source. The Si-H bond is relatively stable at moderate acidic conditions, while it gets hydrolyzed in mineral acids or in alkaline environments [7]. We developed a stable sol consisting of TEOS, TREOS, TEB, triisopropylaluminium (Al(OPr)3) and Rare Earth nitrates in acidic ethanol solution (solid content: 20 Wt%, pH 2 with HN03). Because the Si-H bond decomposes easily at higher pH, any contact to the reactive substrate has to be avoided. However crack-free sol-gel sealings can be applied by dip coating onto magnesium samples, pre-coated with a nanoparticulate coating. After sintering, these sealings are up to 500 nm thick, resulting in almost 800 nm thick coatings (Fig. 2). Nevertheless the Si-H bond decomposes during drying and sintering under the release of hydrogen gas forming bubbles in the coating layer, which stay separated and do not lead to open channels or pores. Beside these bubbles the surface is smooth and crack-free.


Fig. 2: SEM images of a sol-gel sealing on an aqueous prime coat on AZ91; sealing composition: 90.7% Si02 (TRE0S:TEOS = 5:2), 5.3% B2O3, 1.7% A1203, 2.2% Sm203, sintered at 250 "CI2 h; a) surface; b) focused argon ion beam prepared cross section.

The direct application of a second polymeric sol-gel layer is not possible without the formation of cracks, because the possibilities for network relaxation are limited in these inorganic coatings. However if the coating process with aqueous dispersions and polymeric sols is alternated, a kind of lamellar sandwich structure can be formed which stays crack-free up to several layers. The wettability of the Si02/Zr02 layers with aqueous dispersions is good. However, particulate coatings on top of TREOS based layers tend to delaminate, so that only Si02/Zr02 based sealings could be used for this process. So far up to six alternating layers, which equal almost 2 ym, could be applied without significant crack formation. Especially in BSE pictures the Zr02 containing sealings can nicely be distinguished from the mainly Si02 containing dispersion based layers (Fig. 3a).

Fig. 3: SEM images of cross-sections of a 4-fold alternating layer on AZ91 composed of an aqueous dispersion (84.3% Si02 (Levasil 200, 0 20 nm) 9.0% B203, 1.2% P205, 1.7% Na20, 1.5% G O , 1.5% Li20, 1.1% MgO) and a polymeric sol (35.4% SiOz, 55.3% Zr02, 5.6% B203, 3.7% Ce203, lighter layers, sintered at 250 "C); before (a) and after (b) 72 h of impedance measurement (OCP, 10 mV amplitude, 5 mM NaCl); deposition of the corrosion products (dark) under the intact coating.

Electrochemical characterization Electrochemical investigations of the corrosion performance were mainly performed with electrochemical impedance spectroscopy (EIS, Zahner M 6 , 10 KHz-5 mHz ,OCP, amplitude: 10 mV, three-electrode setup vs. SCE in 5 mM NaCl or 0,l M Na2S04). All experiments showed here use AZ91 as substrate. Results for AZ3 1 are concordant.


The impedance spectra of sealed samples consist of two distinct time constants: a RC element at medium frequencies and a RC Element at low frequencies (Fig. 4, Fig. 5). The absolute impedance is increasing with immersion time until eventually the absolute impedance starts to decrease and some additional inductive time constants appear.

For uncoated magnesium samples these low-frequency signals which appear like inductivities, are present in the spectra from the beginning and are generally correlated to local corrosion [8,9]. This shows that these coatings inhibit local corrosion for a certain time. The increase of the absolute impedance can also be observed for samples, just treated with an aqueous prime coat [3]. However for sealed samples the starting and the maximum impedance is much higher and the period until local corrosion appears is much longer. This is especially true for sealings which contain Rare Earth oxides (Fig. 5).

Fig. 4: EIS spectra (bode plot, OCP, amplitude 10 mV, 5 mM NaCl) of AZ91 coated with an aqueous prime coat and a sol-gel sealing (92.3% Si02 from TEOS and TRES, 5.4% B2O3, 2.3% A1203, sintered at 250 "C) after 1-59 h.

. . . .

Fig. 5: EIS spectra (bode plot, OCP, amplitude 10 mV, 5 mM NaC1) of AZ91 coated with an aqueous prime coat and a sol-gel sealing (36.8% Si02 Erom TEOS, 5.8% B203, 52.8% ZrO2, 4.6% Ce203, sintered at 250 "C) after 1-301 h.

Advanced Materials Research Vol. 138 5 1

The RC element at medium frequencies is generally attributed to the film or pore resistance Rp, and the coating resistance Cf, and the RC element at low frequencies to the polarization or charge transfer resistance RCt and the double layer capacity Cdl. At high frequencies the impedance equals the electrolyte resistance Rel. The typical equivalent circuit for non-isolating coatings (figure 6) was used to calculate the impedance elements for these spectra. Due to surface inhomogenities, constant phase elements (CPE) are used for simulation instead of real capacities (The loss factor ranges from 0.85-0.95 indicating almost ideal behavior).

This equivalent circuit fits excellent for spectra after 6 or more hours of immersion, with deviations from 0.4% to 1%. The average error for measurements under 6 h is usually larger. Hilbert transformation shows deviations especially in the LF range for these measurements, indicating a change of the system parameters during the measurement [lo]. With proceeding time the system becomes more stable and Hilbert transformation gives consistent data.

Fig. 6: Equivalent circuit used to model the spectra of coated magnesium alloys.

In Fig. 7 calculated Rpo values for various coatings are shown in dependence of the immersion time in 5 mM NaC1. Starting from a coating resistance of 10 JSC2.cm2 after 1 hour of immersion, an aqueous prime coatings can passivate up to 20 KC2.cm2 and is stable for up to two days. A sealed sample without corrosion inhibitor passivates up to 60 JSC2.cm2 and shows a better long-term stability (up to 5 days). Sealings containing lanthanides show improved corrosion performance. Samarium, Europium and Cerium have a minor effect on the coating resistance and behave quite similar, but they have a remarkable effect on the long term stability (Fig. 7). This is probably due to their applicability as cathodic corrosion inhibitor. Cerium shows the biggest influence in this series, maybe due to its better solubility.

A fourfold, alternating coating can reach coating resistances of 150 KC2.cm2. It is noteworthy that even if coating breakthrough appears, these coatings can recover completely in electrochemical investigations. This is probably due to the inhibitor effect of the Rare Earth salts, which may become dissolved during corrosion and then precipitate on cathodic areas, due to their low solubility under alkaline conditions.

The coating capacity Cf can be calculated from the CPE and be correlated to the coating composition and thickness:

c = Yo . (coma)"-l (1) (Yo = CPE, om, = frequency of maximum phase shift, a = loss factor)

(GO= dielectric permittivity of vacuum, E, = permittivity of the oxide layer, A = surface area, d = coating thickness)

With immersion time Cf is increasing for all studied coatings (shifting to lower frequencies). According to Eq. 2 an increase of the coating thickness would have a contrary effect. Thus, the explanation for the increase of Cf must be an increase of E, by water uptake of the coating [ l 11 or by change of the coating composition by deposition of corrosion products.


Fig. 7: Calculated Rpo from EIS spectra (OCP, amplitude 10 mV, 5 mM NaCl) of A291 coated with an aqueous prime coat and different sealings in dependence of time.

Electron microscopy of cross sections of sealed samples also verifies the deposition of corrosion products under the intact coating explaining the passivation process. Even if the coating density could be increased compared to the aqueous prime coat there is still diffusion of electrolyte and uniform corrosion with the deposition of a passivating layer. However the performance of sealed samples is much better (Fig. 3).

Summary

Compared to the limited thickness, multilayered, alternating coatings based on aqueous dispersions and polymeric sols offer decent protective properties, but do not isolate the substrate completely from environmental influences. During corrosion tests a passivation process can be observed depending on coating thickness and composition, indicating coating porosity. Depending on the electrolyte used during electrochemical investigation a more or less stable layer composed of corrosion products is formed under the intact coating, increasing the coating resistance. This process could be used, similar to chemical conversion, to build up protective coating systems with similar properties like natural protective oxide skins on other metals without significantly changing the device appearance and dimension.

Acknowledgements

This project has been funded by the German Ministry for Economics and Technology (BMWi) via the Arbeitsgemeinschaft industrieller Forschungsvereinigungen ,,Otto von Guericke" e.V. (AiF). We also thank Dr. Peter Thissen of the University of Paderborn for the FIB prepared SEM pictures.


References

[ l] D.-M. Liu: J. Mater. Sci. Lett. 17 (1998) 467

[2] F. Feil, W. Furbeth, M. Schutze: Surface Engineering 24 (2008) 198

[3] F. Feil, W. Furbeth, M. Schutze: Electrochimica Acta 54 (2009) 2478

[4] C.J. Brinker, G.W. Scherer: Sol-gel science: The physics and chemistry of sol-gel processing, Academic Press Inc., Boston (1990)

[5] C.J. Brinker, G.C. Frye, A.J. Hurd, C. S. Ashley: Thin solid films 201 (1991) 97

[6] M.M. Collinson, N. Moore, P.N. Deepa, M. Kanungo: Langmuir 19 (2003) 7669

[7] E. Cordoncillo, F. J. Guaita, P. Escribano, C. Philippe, B. Viana, C. Sanchez: Optical Materials 18 (2001) 309

[8] Chen, J. Wang, E. Han, J. Dong, W. Ke: Electrochim. Acta 52 (2007) 3299

[9] N. Pebere, C. Riera, F. Dabosi: Electrochim. Acta 35 (1990) 555

[lo] W. Ehm, H. Gohr, R. Kaus, B. Roseler, C. A. Schiller: Acta Chim. Hung. 137 (2000) 145

[ l l ] M.L. Zheludkevich, R. Serra, M.F. Montemor, K.A. Yasakau, I.M. Miranda Salvado, M.G.S. Ferreira: Electrochimica Acta 5 1 (2006) 208

Electrochemical characteristics of PEO treated electric arc coatings on lightweight alloys

Nykyforchyn H.M. a, Pokhmurskii V. I.b, Klapkiv M.D. c, Student M.M. d,

Ippolito J. e

Karpenko Physico-Mechanical Institute, 5 Naukova St., 79601 Lviv, Ukraine

a [email protected], b [email protected], c [email protected],

d [email protected]

Universitа degli Studi di Napoli Federico II; Via Pansini 5, 80131 Naples, Italy

e [email protected]

Keywords: steel, aluminum titanium, magnesium alloys, plasma electrolyte oxidation, ceramic coatings, corrosion properties

Abstract. The complex technology of the surface treatment of Al, Mg, Ti alloys for size

reconstruction and strengthening is presented herein. This consists of electric arc spraying of

aluminum alloys or powder wire in an aluminum shell and then treatment with plasma electrolytic

oxidation (PEO). Once treated, oxide-ceramic coatings maintain extreme hardness, durability and

resistance to wear. At the same time their corrosion-resistant properties are also significant.

Dynamic potential dependences were studied for electric arc Al coatings and PEO treatments on Al,

Mg, Ti alloys and corrosion currents were analyzed for exposure to a corrosive environment for a

period of from 1 hour to 30 days. It was established that PEO treated coatings on Al alloys have a

higher corrosion resistance than untreated sprayed coatings.

In Mg alloys, an intermediate layer of aluminum electric arc coating between the substrate

and PEO-treated coating is necessary in order to ensure high corrosion resistance. This is due to the

specifics of the formation of the MgO and Al2O3 oxide phases in the plasma discharge channels. At

the same time Al coatings on Ti alloys, including those of post-PEO treatment, were characterized

as having lower corrosion resistance within the range of electrode potential from corrosion potential

up to repassivation potential, than were untreated Ti-alloys. Yet it was found that the corrosion

resistance of PEO treated coatings increases at higher anode potentials. Under cathode polarization

the hydrogen discharge is less likely to occur on PEO-coatings than on untreated Ti alloys which

more effectively prevents hydrogenation.

Introduction.

The PEO method has been applied to a generation of protective coatings on light alloys and

consists of synthesizing oxide-ceramic coatings in the spark discharge channels of the metal-

electrolyte system. Environmental friendliness and relatively low costs are additional advantages of

PEO technology which are used primarily for surface treatment of aluminium alloys. These coatings

show improved corrosion and wear resistance, increased voltage breakdown and higher thermal

stability. Numerous experimental results and discussion of the PEO treatment tendencies in surface

engineering are summarized in reviews [1, 2].

These days aluminium, magnesium, titanium and zirconium alloys have become the object

of the PEO method [3-10]. However, a narrow range of electric and physical parameters of alloy vs.

electrolytic reaction, and a narrow concentration range of electrolyte content still exists. PEO

coatings on magnesium alloys have improved properties however their corrosion resistance is not

sufficient due to their thorough permeating layers of porosity [11] in spite of attempts at porosity

reduction. [12, 13]. Compared to Al and Mg alloys, PEO coatings on titanium alloys are

characterized by considerably reduced thickness and their corrosion resistance depends upon the

Electrochemical characteristics of PEO treated electric arc coatings on lightweight alloys

Nykyforchyn H.M. a, Pokhmurskii V. I.b, Klapkiv M.D. c, Student M.M. d,

Ippolito J. e

Karpenko Physico-Mechanical Institute, 5 Naukova St., 79601 Lviv, Ukraine

a [email protected], b [email protected], c [email protected],

d [email protected]

Universitа degli Studi di Napoli Federico II; Via Pansini 5, 80131 Naples, Italy

e [email protected]

Keywords: steel, aluminum titanium, magnesium alloys, plasma electrolyte oxidation, ceramic coatings, corrosion properties

Abstract. The complex technology of the surface treatment of Al, Mg, Ti alloys for size

reconstruction and strengthening is presented herein. This consists of electric arc spraying of

aluminum alloys or powder wire in an aluminum shell and then treatment with plasma electrolytic

oxidation (PEO). Once treated, oxide-ceramic coatings maintain extreme hardness, durability and

resistance to wear. At the same time their corrosion-resistant properties are also significant.

Dynamic potential dependences were studied for electric arc Al coatings and PEO treatments on Al,

Mg, Ti alloys and corrosion currents were analyzed for exposure to a corrosive environment for a

period of from 1 hour to 30 days. It was established that PEO treated coatings on Al alloys have a

higher corrosion resistance than untreated sprayed coatings.

In Mg alloys, an intermediate layer of aluminum electric arc coating between the substrate

and PEO-treated coating is necessary in order to ensure high corrosion resistance. This is due to the

specifics of the formation of the MgO and Al2O3 oxide phases in the plasma discharge channels. At

the same time Al coatings on Ti alloys, including those of post-PEO treatment, were characterized

as having lower corrosion resistance within the range of electrode potential from corrosion potential

up to repassivation potential, than were untreated Ti-alloys. Yet it was found that the corrosion

resistance of PEO treated coatings increases at higher anode potentials. Under cathode polarization

the hydrogen discharge is less likely to occur on PEO-coatings than on untreated Ti alloys which

more effectively prevents hydrogenation.

Introduction.

The PEO method has been applied to a generation of protective coatings on light alloys and

consists of synthesizing oxide-ceramic coatings in the spark discharge channels of the metal-

electrolyte system. Environmental friendliness and relatively low costs are additional advantages of

PEO technology which are used primarily for surface treatment of aluminium alloys. These coatings

show improved corrosion and wear resistance, increased voltage breakdown and higher thermal

stability. Numerous experimental results and discussion of the PEO treatment tendencies in surface

engineering are summarized in reviews [1, 2].

These days aluminium, magnesium, titanium and zirconium alloys have become the object

of the PEO method [3-10]. However, a narrow range of electric and physical parameters of alloy vs.

electrolytic reaction, and a narrow concentration range of electrolyte content still exists. PEO

coatings on magnesium alloys have improved properties however their corrosion resistance is not

sufficient due to their thorough permeating layers of porosity [11] in spite of attempts at porosity

reduction. [12, 13]. Compared to Al and Mg alloys, PEO coatings on titanium alloys are

characterized by considerably reduced thickness and their corrosion resistance depends upon the

aggressiveness of the corrosion environment [7]. Therefore, oxide ceramic PEO coatings on Al

alloys have an advantage over other light metals due to the complexity of their properties. Thus

oxide ceramic PEO coatings on Al alloys are considerably advantageous because of their superior

complex properties compared to similar coatings on other light metals.

Recently a new approach has been developed [14] which consists of a combination of Al

coating on light alloys, including Al alloys and even Fe alloys, with additional PEO treatment of a

preliminarily obtained Al coating. One of the variants of a preliminary layer of Al coating consists

of electric arc coating with the use of powder wire. This approach has some advantages: a) better

coating properties obtained with PEO treatment of aluminum alloys compared to PEO treatment of

other light metals; b) the application of PEO coating on metals that are unable to be treated by a

PEO method (for example, Fe-base alloys); c) altering of powder content to reach a flexible change

in properties (for example, to decrease the embrittlement of ceramic coatings, which is a serious

problem for practical applications under cyclic loading); d) possibility for re-dimensioning (by PEO

treatment) of components after intensive wear.

The purpose of this study is to conduct an electrochemical evaluation of corrosion properties

of combined PEO coatings on Al, Mg and Ti alloys.

Methodology. Aluminium alloy D16 (3.8-3.9 Cu; 1.2-1.8 Mg; 0.8 Zn; 0.4-0.5 Si; 0.3-0.9 Mn; 0.4-0.5 Fe;

the rest Al), magnesium alloy МА-5 (0.15-0.5 Mn; 7.8-9.2 Al; 0.2-0.8 Zn; the rest Mg) and titanium

alloy VT8 (6.5 Al; 3.5 Mn; 0.2 Cr) were all used as substrates. Electric arc layers from solid mass

wires D16, AMg-6 (5.8-6.8 Mg; 0.4 Si; 0.4 Fe, the rest Al), combined D16+АMg-6, as well as

powder wire in an aluminum shell with mixture (55%)В4С+(45%)NiCrBSi or

(55%)SiC+45%)NiCrBSi. Thickness of layers was from 50 µm to 300 µm. The following regimens

of arc spraying were used: voltage 32 V; current 100 A, air pressure 0.6 MPa, spraying distance

100 mm.

PEO of the sprayed layers was carried out in an anode-cathode regimen at 20 A/dm2 current

density for 120 min at the cathodic-to-anodic-current-density ratio of 1:1. A 3 g/l KOH + 2 g/l

water glass electrolyte was used. The method is described in detail in ref. [9]. Electrochemical tests

were conducted in 3% NaCl solution using the IPC-Pro potentiostat (scanning rate 2 mV/s).

A reference (Cl–, AgCl | Ag) counter electrode was used.

The microstructure of the coatings’ cross-section was investigated using electron

microscope LEO 1455 VP with EDXS analyzer.

Experimental results and discussion.

A cross-section of sprayed layers is shown on Fig. 1 and that were of a lamellar structure

(zone 2 on Fig. 1a, b). The lamellas were subdivided into thin oxide films of Al2O3. Inclusions of

carbide phases and segregation of Ni were present in the layers that had been sprayed by powder

wires. A clear interface with the substrate was observed throughout all the layers and consisted of a

porosity of 2.5%-2.8% for AMg-6 coatings, 5.8% to 6.1% for D16 and about 4.8% for a mixture of

D16+AMg-6, respectively.

In general PEO essentially alters the structure of the sprayed layers, specifically: lamellas

disappear and porosity diminishes (see Fig. 1). In cases of spraying the content of В4С+NiCrBSi by

powder wire, coarse parts of carbide and nickel alloy were detected in nanosizes during the PEO

process (Fig. 1f) A partial decomposition of carbides was also found. The section of the PEO

coating closer to the surface was partially porous while, at the same time, the inner section was

practically nonporous which had been caused, evidently, by the positive effect on nickel.

Where free carbon was revealed, longitudinal cracks in the PEO coatings from the

SiС+NiCrBSi powder mixture were observed on the sprayed layers closer to external surface

(Fig. 1g). In the internal part of the PEO, coating carbon in the embodiment of submicron sizes was

aggressiveness of the corrosion environment [7]. Therefore, oxide ceramic PEO coatings on Al

alloys have an advantage over other light metals due to the complexity of their properties. Thus

oxide ceramic PEO coatings on Al alloys are considerably advantageous because of their superior

complex properties compared to similar coatings on other light metals.

Recently a new approach has been developed [14] which consists of a combination of Al

coating on light alloys, including Al alloys and even Fe alloys, with additional PEO treatment of a

preliminarily obtained Al coating. One of the variants of a preliminary layer of Al coating consists

of electric arc coating with the use of powder wire. This approach has some advantages: a) better

coating properties obtained with PEO treatment of aluminum alloys compared to PEO treatment of

other light metals; b) the application of PEO coating on metals that are unable to be treated by a

PEO method (for example, Fe-base alloys); c) altering of powder content to reach a flexible change

in properties (for example, to decrease the embrittlement of ceramic coatings, which is a serious

problem for practical applications under cyclic loading); d) possibility for re-dimensioning (by PEO

treatment) of components after intensive wear.

The purpose of this study is to conduct an electrochemical evaluation of corrosion properties

of combined PEO coatings on Al, Mg and Ti alloys.

Methodology. Aluminium alloy D16 (3.8-3.9 Cu; 1.2-1.8 Mg; 0.8 Zn; 0.4-0.5 Si; 0.3-0.9 Mn; 0.4-0.5 Fe;

the rest Al), magnesium alloy МА-5 (0.15-0.5 Mn; 7.8-9.2 Al; 0.2-0.8 Zn; the rest Mg) and titanium

alloy VT8 (6.5 Al; 3.5 Mn; 0.2 Cr) were all used as substrates. Electric arc layers from solid mass

wires D16, AMg-6 (5.8-6.8 Mg; 0.4 Si; 0.4 Fe, the rest Al), combined D16+АMg-6, as well as

powder wire in an aluminum shell with mixture (55%)В4С+(45%)NiCrBSi or

(55%)SiC+45%)NiCrBSi. Thickness of layers was from 50 µm to 300 µm. The following regimens

of arc spraying were used: voltage 32 V; current 100 A, air pressure 0.6 MPa, spraying distance

100 mm.

PEO of the sprayed layers was carried out in an anode-cathode regimen at 20 A/dm2 current

density for 120 min at the cathodic-to-anodic-current-density ratio of 1:1. A 3 g/l KOH + 2 g/l

water glass electrolyte was used. The method is described in detail in ref. [9]. Electrochemical tests

were conducted in 3% NaCl solution using the IPC-Pro potentiostat (scanning rate 2 mV/s).

A reference (Cl–, AgCl | Ag) counter electrode was used.

The microstructure of the coatings’ cross-section was investigated using electron

microscope LEO 1455 VP with EDXS analyzer.

Experimental results and discussion.

A cross-section of sprayed layers is shown on Fig. 1 and that were of a lamellar structure

(zone 2 on Fig. 1a, b). The lamellas were subdivided into thin oxide films of Al2O3. Inclusions of

carbide phases and segregation of Ni were present in the layers that had been sprayed by powder

wires. A clear interface with the substrate was observed throughout all the layers and consisted of a

porosity of 2.5%-2.8% for AMg-6 coatings, 5.8% to 6.1% for D16 and about 4.8% for a mixture of

D16+AMg-6, respectively.

In general PEO essentially alters the structure of the sprayed layers, specifically: lamellas

disappear and porosity diminishes (see Fig. 1). In cases of spraying the content of В4С+NiCrBSi by

powder wire, coarse parts of carbide and nickel alloy were detected in nanosizes during the PEO

process (Fig. 1f) A partial decomposition of carbides was also found. The section of the PEO

coating closer to the surface was partially porous while, at the same time, the inner section was

practically nonporous which had been caused, evidently, by the positive effect on nickel.

Where free carbon was revealed, longitudinal cracks in the PEO coatings from the

SiС+NiCrBSi powder mixture were observed on the sprayed layers closer to external surface

(Fig. 1g). In the internal part of the PEO, coating carbon in the embodiment of submicron sizes was



Figure 1. PEO coatings on sprayed layers from D 16 alloy on aluminum (a), magnesium (b, c), and titanium (d) alloys, as well as from powder wires B4C+NiCrBSi (e, fl, and Sic+-NiCrBSi

(g, h) on aluminum alloy: 1 - substrate, 2 - sprayed arc layer, 3 and 3'- PEO coating.

present (Fig. 1h) while silicon carbide was not absent which indicates that it decomposed in the

PEO process.

The PEO coating on the sprayed layer of 50 µm in thickness on magnesium alloy (Fig. 1c)

consists of two subzones: non-mixed phases MgO (zone 3') and Al2O3 (zone 3). Transverse cracks

were observed in subzone 3' what led to an exfoliation of the coating during corrosion tests. This

leads us to conclude that specimens of a 50 µm intermediate aluminum layer, as shown in Fig. 1b,

can be recommended for application in corrosion environments.

Aluminum layers on titanium alloy were sprayed with D16 mixture and AMg-6 wires. The

structure of PEO coatings (zones 3 and 3') of a 50µm and 100µm thickness is presented on Fig. 1d

and extends to the titanium substrate (zone 1). Zone 3' of the PEO coating, which consists of phase

ТіО2, is adjacent to the titanium substrate. The tongues of these phases (ТіО2) grew in zone 3 with

.Al2O3 and, in some cases, even reached the surface.

Electrochemical properties.

Polarization curves for sprayed Al coatings are presented on Fig. 2. Their corrosion

potentials Ec shift in a negative direction compared to the D16 substrate potential

(Fig. 2, curves 1-4), the maximum difference being of 370 mV.

-1750 -1500 -1250 -1000 -750 -500 -250 0

1E-5

1E-4

1E-3

0.01

0.1

1

10

100

67

5

43

21

E, mV

i, m

A/c

m2

Figure 2. Polarization curves after 1 hour exposure in 3% NaCl solution for: 1 –

substrate D16; 2 – sprayed D16 on substrate D16; 3 –– sprayed AMg-6 on substrate D16; 4 –

sprayed layer (D16+AMg6) on D16; 5– sprayed AMg-6 on substrate D16 with PEO coating;

6 – sprayed D16 on substrate D16 with PEO coating; 7 –sprayed (D16+AMg-6) on substrate

D16 with PEO coating.

This effect can be explained by the porosity of the sprayed layer and, evidently, its

thermodynamic instability. Besides these two factors a sharp shift in the corrosion potential Ec for

sprayed AMg-6 to -1300 mV and mixture D16 + AMg-6 to -1020 mV is defined by a concentration

of Mg in the sprayed layer (more in the AMg-6 layer than in a mixture with D16). In this way the

corrosion potential is more negative for the AMg-6 layer only. At anodic polarization the wide

ranges of passivation potentials can be observed. The passivation current ip for sprayed layers D16

and AMg-6 reaches 0,3 mA/cm2 for mixture D16 + AMG-6 – 0,15 mA/cm

2, and is more in order

present (Fig. 1h) while silicon carbide was not absent which indicates that it decomposed in the

PEO process.

The PEO coating on the sprayed layer of 50 µm in thickness on magnesium alloy (Fig. 1c)

consists of two subzones: non-mixed phases MgO (zone 3') and Al2O3 (zone 3). Transverse cracks

were observed in subzone 3' what led to an exfoliation of the coating during corrosion tests. This

leads us to conclude that specimens of a 50 µm intermediate aluminum layer, as shown in Fig. 1b,

can be recommended for application in corrosion environments.

Aluminum layers on titanium alloy were sprayed with D16 mixture and AMg-6 wires. The

structure of PEO coatings (zones 3 and 3') of a 50µm and 100µm thickness is presented on Fig. 1d

and extends to the titanium substrate (zone 1). Zone 3' of the PEO coating, which consists of phase

ТіО2, is adjacent to the titanium substrate. The tongues of these phases (ТіО2) grew in zone 3 with

.Al2O3 and, in some cases, even reached the surface.

Electrochemical properties.

Polarization curves for sprayed Al coatings are presented on Fig. 2. Their corrosion

potentials Ec shift in a negative direction compared to the D16 substrate potential

(Fig. 2, curves 1-4), the maximum difference being of 370 mV.

-1750 -1500 -1250 -1000 -750 -500 -250 0

1E-5

1E-4

1E-3

0.01

0.1

1

10

100

67

5

43

21

E, mV

i, m

A/c

m2

Figure 2. Polarization curves after 1 hour exposure in 3% NaCl solution for: 1 –

substrate D16; 2 – sprayed D16 on substrate D16; 3 –– sprayed AMg-6 on substrate D16; 4 –

sprayed layer (D16+AMg6) on D16; 5– sprayed AMg-6 on substrate D16 with PEO coating;

6 – sprayed D16 on substrate D16 with PEO coating; 7 –sprayed (D16+AMg-6) on substrate

D16 with PEO coating.

This effect can be explained by the porosity of the sprayed layer and, evidently, its

thermodynamic instability. Besides these two factors a sharp shift in the corrosion potential Ec for

sprayed AMg-6 to -1300 mV and mixture D16 + AMg-6 to -1020 mV is defined by a concentration

of Mg in the sprayed layer (more in the AMg-6 layer than in a mixture with D16). In this way the

corrosion potential is more negative for the AMg-6 layer only. At anodic polarization the wide

ranges of passivation potentials can be observed. The passivation current ip for sprayed layers D16

and AMg-6 reaches 0,3 mA/cm2 for mixture D16 + AMG-6 – 0,15 mA/cm

2, and is more in order


than substrate D16, indicating that the corrosion properties of the sprayed layers are less resistant

than the aluminium substrate. Polarization curves for PEO treated sprayed Al coatings are presented

on Fig. 2, curves 5-7. As can be seen, corrosion potentials shifted in a positive direction for the

treated AMg-6 layers to -790 mV, D16 - 510 mV and (D16+AMG-6) to -500 mV. The

corresponding passivation current is 0.04 mA/cm2, 0. 016 mA/cm

2 and 0.004 mA/cm

2. This is

evidence of the strong protective role that the PEO treatment played in improving the corrosion

resistant sprayed layers.

Despite the electric arc coatings on aluminum alloy D16, these coatings on magnesium alloy

МА5 increase their corrosion resistance as verified by the shift in the corrosion potential. (Fig. 3,

curves 1-3). Additionally, the PEO treatment of these layers increases the corrosion properties of

their surface (Fig. 3, curve 4-5). Correspondingly, the corrosion potential shifted from -1540 mV for

-3000 -2500 -2000 -1500 -1000 -500 0 500

1E-4

1E-3

0,01

0,1

1

10

100

5

4

3

2

i, m

A/c

m2

E, mV

1

Figure 3. Polarization curves after 1 hour exposure to 3% NaCl solution for: 1 – substrate

MA5; 2 – sprayed D16 on substrate MA5; 3 –– sprayed (D16+AMg-6) on substrate MA5; 4 -

sprayed D16 on substrate MA5 with PEO coating; 5 – sprayed layer (D16+AMg-6) on substrate

MA5; 5– sprayed (D16+AMg-6) on substrate MA5 with PEO coating.

-4000 -3000 -2000 -1000 0 1000 2000

1E-4

1E-3

0,01

0,1

1

10

100

1

4

2

E, mV

i, m

A/c

m2

3

Figure 4. Polarization curves after 1 hour exposure in 3% NaCl solution for: 1 – substrate

VT8; 2 - sprayed (D16+AMg-6) on substrate VT8 (t = 100 µm) with PEO coating; 3- sprayed

(D16+AMg-6) on substrate VT8 (t = 50 µm) with PEO coating; 4 – sprayed (D16+AMg-6) on

substrate VT8 (t = 300 µm) with PEO coating.

than substrate D16, indicating that the corrosion properties of the sprayed layers are less resistant

than the aluminium substrate. Polarization curves for PEO treated sprayed Al coatings are presented

on Fig. 2, curves 5-7. As can be seen, corrosion potentials shifted in a positive direction for the

treated AMg-6 layers to -790 mV, D16 - 510 mV and (D16+AMG-6) to -500 mV. The

corresponding passivation current is 0.04 mA/cm2, 0. 016 mA/cm

2 and 0.004 mA/cm

2. This is

evidence of the strong protective role that the PEO treatment played in improving the corrosion

resistant sprayed layers.

Despite the electric arc coatings on aluminum alloy D16, these coatings on magnesium alloy

МА5 increase their corrosion resistance as verified by the shift in the corrosion potential. (Fig. 3,

curves 1-3). Additionally, the PEO treatment of these layers increases the corrosion properties of

their surface (Fig. 3, curve 4-5). Correspondingly, the corrosion potential shifted from -1540 mV for

-3000 -2500 -2000 -1500 -1000 -500 0 500

1E-4

1E-3

0,01

0,1

1

10

100

5

4

3

2

i, m

A/c

m2

E, mV

1

Figure 3. Polarization curves after 1 hour exposure to 3% NaCl solution for: 1 – substrate

MA5; 2 – sprayed D16 on substrate MA5; 3 –– sprayed (D16+AMg-6) on substrate MA5; 4 -

sprayed D16 on substrate MA5 with PEO coating; 5 – sprayed layer (D16+AMg-6) on substrate

MA5; 5– sprayed (D16+AMg-6) on substrate MA5 with PEO coating.

-4000 -3000 -2000 -1000 0 1000 2000

1E-4

1E-3

0,01

0,1

1

10

100

1

4

2

E, mV

i, m

A/c

m2

3

Figure 4. Polarization curves after 1 hour exposure in 3% NaCl solution for: 1 – substrate

VT8; 2 - sprayed (D16+AMg-6) on substrate VT8 (t = 100 µm) with PEO coating; 3- sprayed

(D16+AMg-6) on substrate VT8 (t = 50 µm) with PEO coating; 4 – sprayed (D16+AMg-6) on

substrate VT8 (t = 300 µm) with PEO coating.


substrate to – 1170 і -750 mV for PEO coatings on sprayed layers from wires D16 and

(D16+AMg-6). The wide expanse of passivity on these PEO coatings polarization curves can be

observed, such as in the D16 coating and its corresponding passivation currents of 2·10-2

and 6·10-

3 mA/cm

2. This comparison demonstrates the advantage of the wire mixture (Д16+АМг-6) for, in

this example, the corrosion resistance of the PEO treated coating is consistent with the corrosion

resistant alloy D16.

Other characteristics of the polarization curves for coatings on titanium alloy VT-8 (Fig. 4)

are as follows: the corrosion potentials of PEO treated coatings shift to a negative direction

compared to the substrate;

The corrosion current also increases in relation to the thickness of the sprayed layer. For

thickness t = 50 and 100 µm (at a depth of a PEO treatment of 100 – 120 µm, this means that the

oxide has penetrated into the substrate), and the anode branch for t = 50 µm overlaps on another

curve (for t = 100 µm). For t = 300 µm the anode branch has shifted up and to the left, indicating

-4000 -2000 0 2000 4000 60001E-4

1E-3

0,01

0,1

1

10

100

7

65

9

88

8

4

3

21

E, mV

i, m

A/c

m2

a

-4000 -2000 0 20001E-4

1E-3

0,01

0,1

1

10

100

98

76

54

2

1

i, m

A/c

m2

E, mV

3

b

Figure 5. Polarization curves in 3% NaCl solution for PEO treated sprayed layers from

SiС+NiCrBSi (а) and В4С+NiCrBSi (b) after exposure times of: 1 – 1 ½ hours, 2 – 1 days, 3 – 2

days, 4 - 4 days, 5 – 7 days, 6 – 10 days, 7 – 15 days, 8 – 21 days, 9 – 30 days.

substrate to – 1170 і -750 mV for PEO coatings on sprayed layers from wires D16 and

(D16+AMg-6). The wide expanse of passivity on these PEO coatings polarization curves can be

observed, such as in the D16 coating and its corresponding passivation currents of 2·10-2

and 6·10-

3 mA/cm

2. This comparison demonstrates the advantage of the wire mixture (Д16+АМг-6) for, in

this example, the corrosion resistance of the PEO treated coating is consistent with the corrosion

resistant alloy D16.

Other characteristics of the polarization curves for coatings on titanium alloy VT-8 (Fig. 4)

are as follows: the corrosion potentials of PEO treated coatings shift to a negative direction

compared to the substrate;

The corrosion current also increases in relation to the thickness of the sprayed layer. For

thickness t = 50 and 100 µm (at a depth of a PEO treatment of 100 – 120 µm, this means that the

oxide has penetrated into the substrate), and the anode branch for t = 50 µm overlaps on another

curve (for t = 100 µm). For t = 300 µm the anode branch has shifted up and to the left, indicating

-4000 -2000 0 2000 4000 60001E-4

1E-3

0,01

0,1

1

10

100

7

65

9

88

8

4

3

21

E, mV

i, m

A/c

m2

a

-4000 -2000 0 20001E-4

1E-3

0,01

0,1

1

10

100

98

76

54

2

1

i, m

A/c

m2

E, mV

3

b

Figure 5. Polarization curves in 3% NaCl solution for PEO treated sprayed layers from

SiС+NiCrBSi (а) and В4С+NiCrBSi (b) after exposure times of: 1 – 1 ½ hours, 2 – 1 days, 3 – 2

days, 4 - 4 days, 5 – 7 days, 6 – 10 days, 7 – 15 days, 8 – 21 days, 9 – 30 days.


lower corrosion resistance. Therefore, for t = 50 µm coatings, the electrochemical parameters are

Ec = -1160 mV, ip = 0,015 mA/сm2, but for t = 300 µm Ec = - 1380 mV, ip = 0,040 mA/cm

2. The

passivity sections for PEO treated coatings are essentially more exstensive than for the titanium

substrate and the passivation current practically coincides with the repassivation current of the

titanium alloy (intersection point of curves 1 and 2 on Fig. 4).

Table 1. Electrochemical parameters of PEO coatings on the layers sprayed by the powder

wires n depending upon time of preliminary exposure to 3% NaCl solution

Parameters Powder

content

Time of exposition, hours

1 24 48 56 168 240 360 504 720

E, mV SiС+

NiCrBSi

-470 -630 -664 -728 -740 -730 -752 -773 -776

ip, mA/сm2 7·10

-2 0.65 1.0 1.5 1.0 1.1 1.4 2.5 1.3

E,mV В4С+

NiCrBSi

-700 -610 -615 -770 -746 -740 -670 -622 -620

ip, mA/cm2 2 10

-2 0.25 0.2 0.27 0.24 0.1 1.0 1.3 0.9

The polarization curves for PEO coatings on layers sprayed by power wires were obtained

for the different times of the preliminary exposure to the corrosion environment (Fig. 5). For the

exposure time of 1-30 days the currents of passivity are from 1 to 1 ½ order higher than after a

1 hour exposure however the corrosion potentials had changed slightly (Table 1) and were in the

range of potentials for PEO coatings on sprayed layers from solid mass wires (see Fig. 1, curves 5-

7). However the PEO coatings on the layers from powder wires were less resistant than PEO

coatings on layers from solid mass wires.

Conclusions

1. Plasma electrolytic oxidation preliminary obtained by arc spraying of aluminium

alloy on aluminium, magnesium and titanium substrates has good future prospects as a sound

method of surface treatment for improving not only mechanical, but also corrosion properties. The

use of powder wires gives additional advantages because of their possibility to alter the content of

PEO coating within a wide range.

2. For PEO treatment of corrosion sensitive magnesium alloys it is recommended to

leave the PEO untreated layer to about a 50 µm thickness of sprayed corrosion resistant aluminium

alloy which serves as a barrier for penetration of the corrosion environment to the magnesium

substrate.

References

1. R.C. Barik, J.A. Wharton, R.J.K. Wood, K.R. Stokes and R.L. Jones: Surface and Coatings

Technology Vol. 199 (2005), p.158

2. P. Gupta, G. Tenhundfeld, E.O. Daigle, D. Riabkov: Surface and Coating Technology Vol. 201

(2007), p. 8746

3. A.L. Yerokhin, X. Nie, A. Leyland, A. Matthews: Surface and Coatings Technology Vol. 130

(2000), p. 195

4. W. Dietzel , M.D. Klapkiv

, H.M. Nykyforchyn

, V.M. Posuvailo, C. Blawert: Mater. Sci. N 5

(2004), p. 585

5. C. Blawert, T.V. Heitmann, W. Dietzel, H.M. Nykyforchyn, M.D. Klapkiv: Surface and Coatings

Technology Vol. 200 (2005), p. 68

lower corrosion resistance. Therefore, for t = 50 µm coatings, the electrochemical parameters are

Ec = -1160 mV, ip = 0,015 mA/сm2, but for t = 300 µm Ec = - 1380 mV, ip = 0,040 mA/cm

2. The

passivity sections for PEO treated coatings are essentially more exstensive than for the titanium

substrate and the passivation current practically coincides with the repassivation current of the

titanium alloy (intersection point of curves 1 and 2 on Fig. 4).

Table 1. Electrochemical parameters of PEO coatings on the layers sprayed by the powder

wires n depending upon time of preliminary exposure to 3% NaCl solution

Parameters Powder

content

Time of exposition, hours

1 24 48 56 168 240 360 504 720

E, mV SiС+

NiCrBSi

-470 -630 -664 -728 -740 -730 -752 -773 -776

ip, mA/сm2 7·10

-2 0.65 1.0 1.5 1.0 1.1 1.4 2.5 1.3

E,mV В4С+

NiCrBSi

-700 -610 -615 -770 -746 -740 -670 -622 -620

ip, mA/cm2 2 10

-2 0.25 0.2 0.27 0.24 0.1 1.0 1.3 0.9

The polarization curves for PEO coatings on layers sprayed by power wires were obtained

for the different times of the preliminary exposure to the corrosion environment (Fig. 5). For the

exposure time of 1-30 days the currents of passivity are from 1 to 1 ½ order higher than after a

1 hour exposure however the corrosion potentials had changed slightly (Table 1) and were in the

range of potentials for PEO coatings on sprayed layers from solid mass wires (see Fig. 1, curves 5-

7). However the PEO coatings on the layers from powder wires were less resistant than PEO

coatings on layers from solid mass wires.

Conclusions

1. Plasma electrolytic oxidation preliminary obtained by arc spraying of aluminium

alloy on aluminium, magnesium and titanium substrates has good future prospects as a sound

method of surface treatment for improving not only mechanical, but also corrosion properties. The

use of powder wires gives additional advantages because of their possibility to alter the content of

PEO coating within a wide range.

2. For PEO treatment of corrosion sensitive magnesium alloys it is recommended to

leave the PEO untreated layer to about a 50 µm thickness of sprayed corrosion resistant aluminium

alloy which serves as a barrier for penetration of the corrosion environment to the magnesium

substrate.

References

1. R.C. Barik, J.A. Wharton, R.J.K. Wood, K.R. Stokes and R.L. Jones: Surface and Coatings

Technology Vol. 199 (2005), p.158

2. P. Gupta, G. Tenhundfeld, E.O. Daigle, D. Riabkov: Surface and Coating Technology Vol. 201

(2007), p. 8746

3. A.L. Yerokhin, X. Nie, A. Leyland, A. Matthews: Surface and Coatings Technology Vol. 130

(2000), p. 195

4. W. Dietzel , M.D. Klapkiv

, H.M. Nykyforchyn

, V.M. Posuvailo, C. Blawert: Mater. Sci. N 5

(2004), p. 585

5. C. Blawert, T.V. Heitmann, W. Dietzel, H.M. Nykyforchyn, M.D. Klapkiv: Surface and Coatings



6. C. Blawert, V. Heitmann, W. Dietzel, H.M. Nykyforchyn, M.D. Klapkiv: Surface and Coatings


7. M.D. Klapkiv , N. Y. Povstiana, H.M. Nykyforchyn: Mater. Sci. N 2 (2006), p. 277

8. H.M. Nykyforchyn, V.S. Agarwala, M.D. Klapkiv, V.M. Posuvailo: Advanced Materials

Research Vol. 38 (2008), p. 27

9. H.M. Nykyforchyn, M. D. Klapkiv, V. M. Posuvailo: Surface and Coatings Technology,

Vol. 100-101 (1998), p 219

10. R. Arrabal, E. Matykina, T. Hashimoto, P. Skeldon, G.E. Thompson: Surface and Coatings


11. W. Dietzel, H. M. Nykyforchyn, M. D. Klapkiv, V. M. Posuvailo, C. Blawert: High Temp. Mat.

Proc., Vol. 8, N 4 (2004) p.635

12. V.T.Yavors'kyi, I.P. Mertsalo, M.D. Klapkiv, L.V, Savchuk, V.T. Olynets': Mater. Sci. Vol. 39,

N 5 (2003), p. 745

13. Bong Young Yoo, Duck Young Hwang, Jin Young Cho, Hyun Hee Cho, Dong Hyuk Shin.

www.electrochem.org/meetings/scheduler/abstracts/214/1566.pdf

14. Pokhmurskii V., Nykyforchyn G., Student M., Klapkiv M., Pokhmurska H., Wielage B., Grund

T., A. Wank: Journal of Thermal Spray Technology, Vol 16 N (5-6) (2007), p. 998

6. C. Blawert, V. Heitmann, W. Dietzel, H.M. Nykyforchyn, M.D. Klapkiv: Surface and Coatings


7. M.D. Klapkiv , N. Y. Povstiana, H.M. Nykyforchyn: Mater. Sci. N 2 (2006), p. 277

8. H.M. Nykyforchyn, V.S. Agarwala, M.D. Klapkiv, V.M. Posuvailo: Advanced Materials

Research Vol. 38 (2008), p. 27

9. H.M. Nykyforchyn, M. D. Klapkiv, V. M. Posuvailo: Surface and Coatings Technology,

Vol. 100-101 (1998), p 219

10. R. Arrabal, E. Matykina, T. Hashimoto, P. Skeldon, G.E. Thompson: Surface and Coatings


11. W. Dietzel, H. M. Nykyforchyn, M. D. Klapkiv, V. M. Posuvailo, C. Blawert: High Temp. Mat.

Proc., Vol. 8, N 4 (2004) p.635

12. V.T.Yavors'kyi, I.P. Mertsalo, M.D. Klapkiv, L.V, Savchuk, V.T. Olynets': Mater. Sci. Vol. 39,

N 5 (2003), p. 745

13. Bong Young Yoo, Duck Young Hwang, Jin Young Cho, Hyun Hee Cho, Dong Hyuk Shin.

www.electrochem.org/meetings/scheduler/abstracts/214/1566.pdf

14. Pokhmurskii V., Nykyforchyn G., Student M., Klapkiv M., Pokhmurska H., Wielage B., Grund

T., A. Wank: Journal of Thermal Spray Technology, Vol 16 N (5-6) (2007), p. 998


Hybrid coatings based on conducting polymers and polysiloxane chains for corrosion protection of Al alloys

TRUEBA Monica1, a, TRASATTI Stefano P.1,b, FLAMINI Daniel O.2,c 1Department of Physical Chemistry and Electrochemistry,

Università degli Studi di Milano, Via Golgi 19, 20133 Milan, Italy

2Department of Chemical Engineering, Universidad Nacional del Sur,

Av. Alem 1253, Bahia Blanca, Argentina

[email protected], [email protected], [email protected]

Keywords: Al alloys, corrosion protection, silane-based treatments, conducting polymers.

Abstract. It was previously demonstrated that the use of a pyrrole-based silane (PySi) for surface

treatment of Al alloys provides both active and barrier protection due to the deposition of a hybrid

coating, containing polypyrrole and polysiloxane chains. To further explore these features, a wider

range of Al substrates and different silane-based formulations in terms of silane molecule, solvent

nature, water amount and pH, were investigated. Also, some tests were carried out by using aniline-

based silane (AniSi). Structural/morphological characterization of the coatings, as well as the

investigation of PySi solutions by diverse spectroscopic techniques, in addition to corrosion tests in

NaCl, strongly support the very promising protection performance of the hybrid film. This is

indicated as well from the preliminary results obtained with the AniSi-based approach. Thus,

typical silane-based treatments with principally barrier action can gain in active properties if the

silane compound contains monomers of conducting polymers as a funtional group.

Introduction

The demand of lightweight aluminium and its alloys is increasing year by year, becoming this metal

the most used after steel. Severe federal emission regulations as well as fuel economy standards

make Al alloys the materials of choice especially in aerospace and automotive industries. Reduced

vehicle weight along with powertrain efficiency, are directly related to improved fuel consumption

and CO2 emission reduction. Besides the wide range of forms that Al can take (bar, tube, sheet, etc.)

and the variety of surface finishing available (e.g. anodizing), this metal is 100% recyclable [1].

Heavy restrictions of Cr (VI)-based treatments for corrosion inhibition of Al alloys have

conditioned the development of several alternatives among which conducting polymers are

intensively investigated as active coatings [2-8]. One of the challenges in developing these coatings

is to overcome processability difficulties and to improve the adhesion. We have developed a new

promising approach, which allows one to obtain a composite film containing polypyrrole units and

polysiloxane linkages in a simple way by using a pyrrolil-silicon compound as a primer on as-

received Al wrought alloys [9]. Besides the high degree of compactness, an improved adhesion due

to silanol group preferential adsorption and condensation at the metal/film interface is obtained.

Also, mixed protection in terms of passive (barrier) and active (anodic) actions provided by

polysiloxane and polypyrrole chains within the composite network, is revealed. Last but not least,

the deposition of this hybrid coating is carried out by dipping the metal in the hydrolyzed solution,

similar to the procedure employed for silane-based treatments.

The present work summarizes our recent studies that further confirm the promising features of the

pyrrole-based silane when compared with organofunctional silanes like mono- and bis- amino

silanes, imidazole-based silane and octylsilane, as well as mixture of silanes. Preliminary results of

aniline-based silane are also presented.

Hybrid coatings based on conducting polymers and polysiloxane chains for corrosion protection of Al alloys

TRUEBA Monica1, a, TRASATTI Stefano P.1,b, FLAMINI Daniel O.2,c 1Department of Physical Chemistry and Electrochemistry,

Università degli Studi di Milano, Via Golgi 19, 20133 Milan, Italy

2Department of Chemical Engineering, Universidad Nacional del Sur,

Av. Alem 1253, Bahia Blanca, Argentina


Keywords: Al alloys, corrosion protection, silane-based treatments, conducting polymers.

Abstract. It was previously demonstrated that the use of a pyrrole-based silane (PySi) for surface

treatment of Al alloys provides both active and barrier protection due to the deposition of a hybrid

coating, containing polypyrrole and polysiloxane chains. To further explore these features, a wider

range of Al substrates and different silane-based formulations in terms of silane molecule, solvent

nature, water amount and pH, were investigated. Also, some tests were carried out by using aniline-

based silane (AniSi). Structural/morphological characterization of the coatings, as well as the

investigation of PySi solutions by diverse spectroscopic techniques, in addition to corrosion tests in

NaCl, strongly support the very promising protection performance of the hybrid film. This is

indicated as well from the preliminary results obtained with the AniSi-based approach. Thus,

typical silane-based treatments with principally barrier action can gain in active properties if the

silane compound contains monomers of conducting polymers as a funtional group.

Introduction

The demand of lightweight aluminium and its alloys is increasing year by year, becoming this metal

the most used after steel. Severe federal emission regulations as well as fuel economy standards

make Al alloys the materials of choice especially in aerospace and automotive industries. Reduced

vehicle weight along with powertrain efficiency, are directly related to improved fuel consumption

and CO2 emission reduction. Besides the wide range of forms that Al can take (bar, tube, sheet, etc.)

and the variety of surface finishing available (e.g. anodizing), this metal is 100% recyclable [1].

Heavy restrictions of Cr (VI)-based treatments for corrosion inhibition of Al alloys have

conditioned the development of several alternatives among which conducting polymers are

intensively investigated as active coatings [2-8]. One of the challenges in developing these coatings

is to overcome processability difficulties and to improve the adhesion. We have developed a new

promising approach, which allows one to obtain a composite film containing polypyrrole units and

polysiloxane linkages in a simple way by using a pyrrolil-silicon compound as a primer on as-

received Al wrought alloys [9]. Besides the high degree of compactness, an improved adhesion due

to silanol group preferential adsorption and condensation at the metal/film interface is obtained.

Also, mixed protection in terms of passive (barrier) and active (anodic) actions provided by

polysiloxane and polypyrrole chains within the composite network, is revealed. Last but not least,

the deposition of this hybrid coating is carried out by dipping the metal in the hydrolyzed solution,

similar to the procedure employed for silane-based treatments.

The present work summarizes our recent studies that further confirm the promising features of the

pyrrole-based silane when compared with organofunctional silanes like mono- and bis- amino

silanes, imidazole-based silane and octylsilane, as well as mixture of silanes. Preliminary results of

aniline-based silane are also presented.

Experimental part

Materials. High purity chemicals were used, except pyrrole (Py) that was freshly distilled under

reduced pressure in a nitrogen atmosphere prior to use. Millipore MiliQ water was used where

needed. Commercial wrought Al alloys (AA) of series 1xxx, 2xxx, 5xxx, 6xxx, and 7xxx were

purchased from AVIOMETAL S.p.a. (Table 1). Plates with thickness 1 to 1.5 mm (depending on

the alloy) were cut into 20 x 30 mm coupons that were used after ultrasonic cleaning in n-hexane,

acetone and methanol for 15 min. each. Commercial alloys AA6061 (extruded) and AA5052 were

also studied after pre-treatment with mild alkaline solution.

Table 1. Chemical composition (wt%) of commercial wrought Al alloys (AA).

Al alloys Si Fe Cu Mn Mg Zn Ti Cr

AA1050 O 0.25 0.40 0.05 0.05 0.05 0.07 0.05 <0.03

AA1050 H24 0.14 0.25 < 0.01 0.01 < 0.01 < 0.01 0,01 < 0.01

AA6082 T6 0.90 0.36 0.04 0.56 1.00 0.02 0.02 0.04

AA5754 H111 0.08 0.26 0.03 0.18 2.73 0.01 <0.01 0.05

AA5083 H111 0.17 0.32 0.04 0.62 4.32 0.03 0.02 0.07

AA2024 T3 0.15 0.25 4.67 0.63 1.34 0.02 0.06 0.01

AA7075 T6 0.08 0.13 1.60 0.02 2.52 5.90 0.04 0.19

Surface treatment. Surface treatments were performed by immersion of the specimens in

hydrolyzed solutions of silane-based compounds as listed in Table 2. Solutions were prepared at

4%v/v in methanol/water (95:5) at pH 4 and then left to hydrolyze under stagnant conditions for

three days. Other experimental conditions, such as solvent/water proportion, low VOC ter-butilic

alcohol (tBuOH) in place of methanol and solution pH, were also investigated. After solvent

cleaning, specimens were pre-heated at 120 °C for 20 min. and then immersed in the testing

solutions. After 1-3 minutes immersion, the so-modified substrates were dried in hot-air stream and

cured in an open-to-air sand oven from 1 to 2 hours, depending on the immersion time. In some

cases, a multiple immersion was also carried out with intermediate 20 min. thermal treatment.

Characterization techniques. Silane-based hydrolyzed solutions were characterized by recording

UV-Visible, Attenuated Total Reflection IR and 29

Si NMR spectra, as a function of time. UV-

Visible spectra were taken using a JASCO V530 spectrometer in the wavelength scan mode in the

region 200-800 nm. Transmission spectra (ATR) were obtained in the range 4000 to 400 cm-1

with

a Perkin-Elmer Spotlight 300/Spectrum 100 spectrophotometer equipped with a diamond crystal. 29

Si NMR spectra were recorded at room temperature using a Bruker 500 spectrometer equipped

with a 5mm BB probe and operating at 99.36 MHz. As-adsorbed layers and cured films were

characterized by reflection-absorption IR (RAIR) and diffusion-reflection (DRIFT) spectroscopy.

The RAIR spectra were recorded with a Bio-Rad FTS-40 spectrophotometer in the scanning range

4000 to 400 cm-1

with a spectral resolution of 4 cm-1

and scan number of 64. DRIFT spectra were

obtained with a Spectrum One (Perkin-Elmer) spectrophotometer in the range 4000 to 500 cm-1

(32 scans). For X-ray photoelectron spectroscopy an ESCA system (XI ASCII Surface Science

Instruments) operating at 10-8

-10-9

Torr with Al anode (1486.6 eV) and 1 eV of energy resolution

was used. Morphology was examined on a scanning electron microscope (SEM), using a LEO 1430

microscope equipped with an EDX spectrometer at a chamber pressure of 8x10-6

Torr and 20 keV

accelerating voltage. For cross-section examination, coated specimens were mounted in a cold-

working resin, followed by polishing up to 1 µm with diamond paste and non-aqueous solvent.

Experimental part

Materials. High purity chemicals were used, except pyrrole (Py) that was freshly distilled under

reduced pressure in a nitrogen atmosphere prior to use. Millipore MiliQ water was used where

needed. Commercial wrought Al alloys (AA) of series 1xxx, 2xxx, 5xxx, 6xxx, and 7xxx were

purchased from AVIOMETAL S.p.a. (Table 1). Plates with thickness 1 to 1.5 mm (depending on

the alloy) were cut into 20 x 30 mm coupons that were used after ultrasonic cleaning in n-hexane,

acetone and methanol for 15 min. each. Commercial alloys AA6061 (extruded) and AA5052 were

also studied after pre-treatment with mild alkaline solution.

Table 1. Chemical composition (wt%) of commercial wrought Al alloys (AA).

Al alloys Si Fe Cu Mn Mg Zn Ti Cr

AA1050 O 0.25 0.40 0.05 0.05 0.05 0.07 0.05 <0.03

AA1050 H24 0.14 0.25 < 0.01 0.01 < 0.01 < 0.01 0,01 < 0.01

AA6082 T6 0.90 0.36 0.04 0.56 1.00 0.02 0.02 0.04

AA5754 H111 0.08 0.26 0.03 0.18 2.73 0.01 <0.01 0.05

AA5083 H111 0.17 0.32 0.04 0.62 4.32 0.03 0.02 0.07

AA2024 T3 0.15 0.25 4.67 0.63 1.34 0.02 0.06 0.01

AA7075 T6 0.08 0.13 1.60 0.02 2.52 5.90 0.04 0.19

Surface treatment. Surface treatments were performed by immersion of the specimens in

hydrolyzed solutions of silane-based compounds as listed in Table 2. Solutions were prepared at

4%v/v in methanol/water (95:5) at pH 4 and then left to hydrolyze under stagnant conditions for

three days. Other experimental conditions, such as solvent/water proportion, low VOC ter-butilic

alcohol (tBuOH) in place of methanol and solution pH, were also investigated. After solvent

cleaning, specimens were pre-heated at 120 °C for 20 min. and then immersed in the testing

solutions. After 1-3 minutes immersion, the so-modified substrates were dried in hot-air stream and

cured in an open-to-air sand oven from 1 to 2 hours, depending on the immersion time. In some

cases, a multiple immersion was also carried out with intermediate 20 min. thermal treatment.

Characterization techniques. Silane-based hydrolyzed solutions were characterized by recording

UV-Visible, Attenuated Total Reflection IR and 29

Si NMR spectra, as a function of time. UV-

Visible spectra were taken using a JASCO V530 spectrometer in the wavelength scan mode in the

region 200-800 nm. Transmission spectra (ATR) were obtained in the range 4000 to 400 cm-1

with

a Perkin-Elmer Spotlight 300/Spectrum 100 spectrophotometer equipped with a diamond crystal. 29

Si NMR spectra were recorded at room temperature using a Bruker 500 spectrometer equipped

with a 5mm BB probe and operating at 99.36 MHz. As-adsorbed layers and cured films were

characterized by reflection-absorption IR (RAIR) and diffusion-reflection (DRIFT) spectroscopy.

The RAIR spectra were recorded with a Bio-Rad FTS-40 spectrophotometer in the scanning range

4000 to 400 cm-1

with a spectral resolution of 4 cm-1

and scan number of 64. DRIFT spectra were

obtained with a Spectrum One (Perkin-Elmer) spectrophotometer in the range 4000 to 500 cm-1

(32 scans). For X-ray photoelectron spectroscopy an ESCA system (XI ASCII Surface Science

Instruments) operating at 10-8

-10-9

Torr with Al anode (1486.6 eV) and 1 eV of energy resolution

was used. Morphology was examined on a scanning electron microscope (SEM), using a LEO 1430

microscope equipped with an EDX spectrometer at a chamber pressure of 8x10-6

Torr and 20 keV

accelerating voltage. For cross-section examination, coated specimens were mounted in a cold-

working resin, followed by polishing up to 1 µm with diamond paste and non-aqueous solvent.


Table 2. Silane-based molecules used for surface treatment of Al alloys.

N-(3-trimethoxysililpropyl)pyrrole (PySi)

N-(3-trimethoxysililpropyl)aniline (AniSi)

N-(3-triethoxysililpropyl)dihydroimidazole (HImSi)

Octylsilane (OSi)

Bis-(trimethoxysilylpropyl)amine (BA)

Vinyltriacetoxysilane (V)

Methyltrimethoxysilane (PMeSi)

Corrosion tests. Protection performance of the coatings was evaluated at room temperature in

quiescent, naturally aerated near neutral (pH 6.5 ± 0.2) 0.6 M NaCl solution, prepared with reagent

grade NaCl (98%, Aldrich). Bare substrates were also tested. Single-cycle anodic polarization was

recorded at a scan rate of 10 mVmin-1

after an open circuit equilibration for 10 min. The direction

of the scan was reversed as current density reached about 5 x 10-3

A/cm2 up to complete

repassivation (cathodic current). The open circuit potential (Eoc) monitoring was carried out for at

least 15 hour. Long-term immersion tests were carried out in open-air test solutions for 7 days,

according to the ASTM procedure G31 [10]. Electrochemical measurements were performed using

a single-compartment O-ring cell with a working (active) surface of 1 cm2. A Pt sheet was used as a

counterelectrode, and an external SCE as a reference electrode, connected to the working

compartment via a salt bridge containing the test solution and a Luggin capillary. Data were

recorded by means of a PC driven Solartron 1286 potentiostat. At the end of each test, the surfaces

and cross-sections were examined by scanning electron microscopy.

Table 2. Silane-based molecules used for surface treatment of Al alloys.

N-(3-trimethoxysililpropyl)pyrrole (PySi)

N-(3-trimethoxysililpropyl)aniline (AniSi)

N-(3-triethoxysililpropyl)dihydroimidazole (HImSi)

Octylsilane (OSi)

Bis-(trimethoxysilylpropyl)amine (BA)

Vinyltriacetoxysilane (V)

Methyltrimethoxysilane (PMeSi)

Corrosion tests. Protection performance of the coatings was evaluated at room temperature in

quiescent, naturally aerated near neutral (pH 6.5 ± 0.2) 0.6 M NaCl solution, prepared with reagent

grade NaCl (98%, Aldrich). Bare substrates were also tested. Single-cycle anodic polarization was

recorded at a scan rate of 10 mVmin-1

after an open circuit equilibration for 10 min. The direction

of the scan was reversed as current density reached about 5 x 10-3

A/cm2 up to complete

repassivation (cathodic current). The open circuit potential (Eoc) monitoring was carried out for at

least 15 hour. Long-term immersion tests were carried out in open-air test solutions for 7 days,

according to the ASTM procedure G31 [10]. Electrochemical measurements were performed using

a single-compartment O-ring cell with a working (active) surface of 1 cm2. A Pt sheet was used as a

counterelectrode, and an external SCE as a reference electrode, connected to the working

compartment via a salt bridge containing the test solution and a Luggin capillary. Data were

recorded by means of a PC driven Solartron 1286 potentiostat. At the end of each test, the surfaces

and cross-sections were examined by scanning electron microscopy.


Results and Discussion

Pyrrole-based silane

Morphology and Structure. Figure 1 summarizes the cross-section morphology of the films

obtained after surface treatment with different silane-based solutions at 4% v/v in MetOH/H2O

(95:5) and pH 4. The morphology of eletrogenerated polypyrrole (PPy) film is also included [9].

PMeSi film is characterized by a non-continuous layer of ca. 10 µm, constituted of solid

microparticles with rounded-like shape (Fig. 1a). On the contrary, when the octyl chain or the

propylpyrrole group substitute the methyl group (Fig. 1b,c, respectively), a compact amorphous-like

morphology is obtained and the estimated thickness for the latter is higher (at least twice). In

addition, PPySi film is well adhered to the metal substrate (Fig. 1d) due to silanol preferential

adsorption on the metal surface in contrast, for example, to PPy. The morphological features of

PPySi and POSi films, deposited from the silanes hydrolysed solutions in a mixture of terbutilic

alcohol and water, were closely similar to those shown in Figure 1b,c.

Figure 1. Cross-section SEM images of as-prepared films on AA6082: (a) PMeSi, (b) POSi,

(c) PPySi, (d) PPy (white arrow points the Ppy-filled pit promoted during elecrpolymerization)

The main structural features of PPySi, as deduced by RAIR spectroscopy [9], are summarized as

follows: intense pyrrole ring vibrations between 1600-1200 cm-1

, doping-induced bands (D) as well

as vibration modes characteristic of bulk (B) and tail (T) pyrrole groups, which indicate the

presence of pyrrole oligomers (at least tetramers) in the hybrid coating. The wide and intense band

in the Si-O-Si region (1200-1000cm-1

) with a sharp peak at 1086 cm-1

resulting from siloxane cyclic

segments indicates polysiloxane chains. These characteristics were independent of the solvent used

in PySi solution, namely methanol and terbutilic alcohol, as shown in Figure 2a,b respectively.

Despite the PPySi and POSi similar morphology (Fig. 1), the RAIR spectrum for the latter was

almost featureless as illustrated in Figure 2c. An important difference is reflected by the progression

of coupled CH2 wag modes between 1325-1260 cm-1

, as revealed for PPySi film with a spacing of

17 cm-1

(inset of Fig. 2b) that suggest an almost perfect all-trans conformation [9,11]. Conversely,

these progression bands are not detected at all for the octyl chain in the spectrum of POSi, even

though the corresponding stretching vibrations (between 3000-2700 cm-1

) are quite intense, which

points to a disordered structure. This is also suggested by the broad band in the range of 3500-

3000 cm-1

which is attributed to hydrogen-bonded silicon. Another feature repeatedly observed was

the self-assembling of PySi layer on the metallic surface in the immersion step, which is illustrated

for extruded AA6061 in Figure 3. The adsorbed layer obtained in the immersion step, readily shows

the main structural features of PPySi film pointed out above with prevailing sharpening of the bands

after the subsequent thermal treatment (curing). This behaviour was similar for all studied Al alloys,

as well as for other metallic specimens like low carbon steel and galvanized steel (not shown).

(a) (b)

(c) (d)


Pyrrole-based silane

Morphology and Structure. Figure 1 summarizes the cross-section morphology of the films

obtained after surface treatment with different silane-based solutions at 4% v/v in MetOH/H2O

(95:5) and pH 4. The morphology of eletrogenerated polypyrrole (PPy) film is also included [9].

PMeSi film is characterized by a non-continuous layer of ca. 10 µm, constituted of solid

microparticles with rounded-like shape (Fig. 1a). On the contrary, when the octyl chain or the

propylpyrrole group substitute the methyl group (Fig. 1b,c, respectively), a compact amorphous-like

morphology is obtained and the estimated thickness for the latter is higher (at least twice). In

addition, PPySi film is well adhered to the metal substrate (Fig. 1d) due to silanol preferential

adsorption on the metal surface in contrast, for example, to PPy. The morphological features of

PPySi and POSi films, deposited from the silanes hydrolysed solutions in a mixture of terbutilic

alcohol and water, were closely similar to those shown in Figure 1b,c.

Figure 1. Cross-section SEM images of as-prepared films on AA6082: (a) PMeSi, (b) POSi,

(c) PPySi, (d) PPy (white arrow points the Ppy-filled pit promoted during elecrpolymerization)

The main structural features of PPySi, as deduced by RAIR spectroscopy [9], are summarized as

follows: intense pyrrole ring vibrations between 1600-1200 cm-1

, doping-induced bands (D) as well

as vibration modes characteristic of bulk (B) and tail (T) pyrrole groups, which indicate the

presence of pyrrole oligomers (at least tetramers) in the hybrid coating. The wide and intense band

in the Si-O-Si region (1200-1000cm-1

) with a sharp peak at 1086 cm-1

resulting from siloxane cyclic

segments indicates polysiloxane chains. These characteristics were independent of the solvent used

in PySi solution, namely methanol and terbutilic alcohol, as shown in Figure 2a,b respectively.

Despite the PPySi and POSi similar morphology (Fig. 1), the RAIR spectrum for the latter was

almost featureless as illustrated in Figure 2c. An important difference is reflected by the progression

of coupled CH2 wag modes between 1325-1260 cm-1

, as revealed for PPySi film with a spacing of

17 cm-1

(inset of Fig. 2b) that suggest an almost perfect all-trans conformation [9,11]. Conversely,

these progression bands are not detected at all for the octyl chain in the spectrum of POSi, even

though the corresponding stretching vibrations (between 3000-2700 cm-1

) are quite intense, which

points to a disordered structure. This is also suggested by the broad band in the range of 3500-

3000 cm-1

which is attributed to hydrogen-bonded silicon. Another feature repeatedly observed was

the self-assembling of PySi layer on the metallic surface in the immersion step, which is illustrated

for extruded AA6061 in Figure 3. The adsorbed layer obtained in the immersion step, readily shows

the main structural features of PPySi film pointed out above with prevailing sharpening of the bands

after the subsequent thermal treatment (curing). This behaviour was similar for all studied Al alloys,

as well as for other metallic specimens like low carbon steel and galvanized steel (not shown).

(a) (b)

(c) (d)


Figure 2. RAIR spectra of AA2024 after treatment with: (a) PySi in MetOH; (b) PySi in tBuOH;

(c) OSi in tBuOH.

Figure 3 . RAIR spectra of extruded AA6061 after treatment with PySi: (a) adsorbed layer,

(b) PPySi film (after curing).

Studies of PySi hydrolyzed solution by ATR, UV-Visible and 29

Si NMR spectroscopy as a function

of time suggest that pyrrole oligomerization is readily promoted in solution, in contrast with silanol

groups condensation, which could justify the self-assembling of PySi macro-oligomers in the

adsorption step and film thickness in the order of microns after curing. Figure 4 shows the ATR

spectra of PySi and MeSi solutions after 10 days and 2 months. The B and T modes of pyrrole

oligomers increase with time, while the corresponding features of silanol groups remain almost

unchanged contrary to MeSi with significant decrease in intensity of SiOH stretching vibrations.

Figure 2. RAIR spectra of AA2024 after treatment with: (a) PySi in MetOH; (b) PySi in tBuOH;

(c) OSi in tBuOH.

Figure 3 . RAIR spectra of extruded AA6061 after treatment with PySi: (a) adsorbed layer,

(b) PPySi film (after curing).

Studies of PySi hydrolyzed solution by ATR, UV-Visible and 29

Si NMR spectroscopy as a function

of time suggest that pyrrole oligomerization is readily promoted in solution, in contrast with silanol

groups condensation, which could justify the self-assembling of PySi macro-oligomers in the

adsorption step and film thickness in the order of microns after curing. Figure 4 shows the ATR

spectra of PySi and MeSi solutions after 10 days and 2 months. The B and T modes of pyrrole

oligomers increase with time, while the corresponding features of silanol groups remain almost

unchanged contrary to MeSi with significant decrease in intensity of SiOH stretching vibrations.


Figure 4. ATR spectra after 10 days (----) and two months (―) two months for hydrolysed

solutions of: (a) PySi; (b) MeSi.

29

Si NMR after 1 month gives similar indications when comparing PySi with HImSi spectra, as

shown in Figure 5. The T structures, prevailing for the former, indicate mainly dimerization via

silanol groups (T1) with some siloxane linear linking (T

2), remaining in solution an appreciated

amount of unreacted silanol groups (T0

H) [12]. In the case of HimSi, hydrolized groups are not

revealed whereas linear and tridimensional crosslinking prevail (T2 and T

3 structures, respectively),

making this solution less stable over time. The delayed silanol condensation, which is probably

associated to pyrrole rings oligomerization, constitutes an important feature for practical

applications.

Oligomerization of PySi in solution via pyrrole ring is further indicated by the evolution of UV/Vis.

spectra with time, as compared to those obtained for pyrrole (Py), both illustrated in Figure 6. There

is a considerable variation in the experimental UV/Visible data of PPy and Py oligomers. Therefore

the experimental results must be viewed with some caution. On combining polymer and oligomers

reported data, both experimental and theoretical [13], the spectral features of Py (Fig. 6a) suggest

oligomerization up to at least six rings in 15 days, indicated by the broad band at about 590 nm.

These relatively small size oligomers result likely from intermediate species like dimmers and

tetramers as revealed by the peak at 320 nm and the broad feature between 400-500 nm with two

maxima (at ca. 430 and 460 nm), respectively, showing up at early stages of the test. Absorption

characteristics of PySi (Fig. 6b) resemble closely those of Py but the peaks appear red shifted in

about 15 nm as a result of N-substitution. The main differences are given by the much more intense

absorption overall the spectral range as well as by the important broadening between 300-400 nm.

This progressive red shift could be attributed to σ-π mixing effect due to silicon-based substituent

and to changes in torsional angles providing conformational ordering, while the significant intensity

suggests an increase of solution viscosity [14,15]. The latter is likely supported by the splitting in

the range of 400-500 nm that vanishes after 20 days to produce two maxima of higher energy (blue-

shifted) with several sub-bands peaks. This feature becomes dominant, as the oligomers get longer

(more than nine Py units) [13].

Figure 4. ATR spectra after 10 days (----) and two months (―) two months for hydrolysed

solutions of: (a) PySi; (b) MeSi.

29

Si NMR after 1 month gives similar indications when comparing PySi with HImSi spectra, as

shown in Figure 5. The T structures, prevailing for the former, indicate mainly dimerization via

silanol groups (T1) with some siloxane linear linking (T

2), remaining in solution an appreciated

amount of unreacted silanol groups (T0

H) [12]. In the case of HimSi, hydrolized groups are not

revealed whereas linear and tridimensional crosslinking prevail (T2 and T

3 structures, respectively),

making this solution less stable over time. The delayed silanol condensation, which is probably

associated to pyrrole rings oligomerization, constitutes an important feature for practical

applications.

Oligomerization of PySi in solution via pyrrole ring is further indicated by the evolution of UV/Vis.

spectra with time, as compared to those obtained for pyrrole (Py), both illustrated in Figure 6. There

is a considerable variation in the experimental UV/Visible data of PPy and Py oligomers. Therefore

the experimental results must be viewed with some caution. On combining polymer and oligomers

reported data, both experimental and theoretical [13], the spectral features of Py (Fig. 6a) suggest

oligomerization up to at least six rings in 15 days, indicated by the broad band at about 590 nm.

These relatively small size oligomers result likely from intermediate species like dimmers and

tetramers as revealed by the peak at 320 nm and the broad feature between 400-500 nm with two

maxima (at ca. 430 and 460 nm), respectively, showing up at early stages of the test. Absorption

characteristics of PySi (Fig. 6b) resemble closely those of Py but the peaks appear red shifted in

about 15 nm as a result of N-substitution. The main differences are given by the much more intense

absorption overall the spectral range as well as by the important broadening between 300-400 nm.

This progressive red shift could be attributed to σ-π mixing effect due to silicon-based substituent

and to changes in torsional angles providing conformational ordering, while the significant intensity

suggests an increase of solution viscosity [14,15]. The latter is likely supported by the splitting in

the range of 400-500 nm that vanishes after 20 days to produce two maxima of higher energy (blue-

shifted) with several sub-bands peaks. This feature becomes dominant, as the oligomers get longer

(more than nine Py units) [13].


Figure 5. 29

Si NMR spectra of hydrolyzed solutions after 1 month: (a) PySi; (b) HimSi

Figure 6. UV-Visible absorption spectra as a function of time of (a) Py and (b) PySi methanol-

based solutions.

Protection performance. The general trend of corrosion protection for PMeSi, POSi and PPySi

films, according to the potential-current responses during anodic polarization scans, is summarized

in Figure 7 for AA6082. The behaviour obtained for the alloy modified with electrodeposited PPy

film is also shown. As expected, POSi shows improved barrier protection with respect to PMeSi,

but no repassivation regions are detected in both cases. These are clearly reflected for the coatings

(a)

(b)

Figure 5. 29

Si NMR spectra of hydrolyzed solutions after 1 month: (a) PySi; (b) HimSi

Figure 6. UV-Visible absorption spectra as a function of time of (a) Py and (b) PySi methanol-

based solutions.

Protection performance. The general trend of corrosion protection for PMeSi, POSi and PPySi

films, according to the potential-current responses during anodic polarization scans, is summarized

in Figure 7 for AA6082. The behaviour obtained for the alloy modified with electrodeposited PPy

film is also shown. As expected, POSi shows improved barrier protection with respect to PMeSi,

but no repassivation regions are detected in both cases. These are clearly reflected for the coatings

(a)

(b)


containing PPy chains, i.e. PPy and PPySi at potentials higher than about -200 mV vs SCE. The

curves show a “stairs-up-like” shape where almost consecutive current rise and its stabilization with

the direction of the scan prevail, which has been attributed to PPy chains anodic activity against

substrate corrosion [9]. Nevertheless, PPySi gives a wide passive region at the beginning of the

polarization (between ca. -700 and -300 mV), contrary to PPy, due to silanol groups condensation at

the metal/film interface thus favouring the coating adhesion. It is well known that conducting

polymers adhesion on reactive substrates is highly limited. Thus, the use of a pyrrole-based silane

allows this problem to be significantly diminished.

Figure 7. Anodic polarizations in 0.6M NaCl of AA6082 with different coatings. Left: (―) bare

alloy, (--) PMeSi, (--) POSi, (-◊-) electrodeposited PPy. Right: bare alloy, (-⊗-) PPySi.

The passive/active behaviour in the anodic polarization scans was also exhibited by PPySi film

when solvent, water amount and pH were changed in PySi formulation and applied on several

metallic substrates pre-treated with solvent or mild-alkaline cleaning. Figures 8 and 9 summarize

these results, which include the alloys response after treatment with other mono-silanes, bis-silanes

and/or mixtures of silanes. Figure 8 illustrates for AA6061 (mild-alkaline cleaning) the improved

protection offered by PPySi in terms of initial passivity when silane solution at pH 6 is used. Yet,

both PPySi-based films show the breakdown at closely similar potentials (about -500 mV vs SCE)

that is shifted more than 200 mV with respect to that of bare substrate. This indicates that PySi

adhesion is more favoured under neutral conditions and is determined by silanol adsorption and

condensation at the metal/PySi interface. Figure 8b shows the very modest protection offered by

hydrophobic mono-silane like OSi or a mixture of bis-amino and vinyltriacetxy silanes (BAV)

compared to PySi (at pH 4). In addition, although the film composed by a mixture of bis-amino and

OSi (BAPO) shows the highest breakdown and reflects the same features of PPySi (Fig. 8a), the

protection of the latter can be considered much better if silane composition is taken into account,

i.e., mono-silane versus a mixture of bis- and mono-silanes.

containing PPy chains, i.e. PPy and PPySi at potentials higher than about -200 mV vs SCE. The

curves show a “stairs-up-like” shape where almost consecutive current rise and its stabilization with

the direction of the scan prevail, which has been attributed to PPy chains anodic activity against

substrate corrosion [9]. Nevertheless, PPySi gives a wide passive region at the beginning of the

polarization (between ca. -700 and -300 mV), contrary to PPy, due to silanol groups condensation at

the metal/film interface thus favouring the coating adhesion. It is well known that conducting

polymers adhesion on reactive substrates is highly limited. Thus, the use of a pyrrole-based silane

allows this problem to be significantly diminished.

Figure 7. Anodic polarizations in 0.6M NaCl of AA6082 with different coatings. Left: (―) bare

alloy, (--) PMeSi, (--) POSi, (-◊-) electrodeposited PPy. Right: bare alloy, (-⊗-) PPySi.

The passive/active behaviour in the anodic polarization scans was also exhibited by PPySi film

when solvent, water amount and pH were changed in PySi formulation and applied on several

metallic substrates pre-treated with solvent or mild-alkaline cleaning. Figures 8 and 9 summarize

these results, which include the alloys response after treatment with other mono-silanes, bis-silanes

and/or mixtures of silanes. Figure 8 illustrates for AA6061 (mild-alkaline cleaning) the improved

protection offered by PPySi in terms of initial passivity when silane solution at pH 6 is used. Yet,

both PPySi-based films show the breakdown at closely similar potentials (about -500 mV vs SCE)

that is shifted more than 200 mV with respect to that of bare substrate. This indicates that PySi

adhesion is more favoured under neutral conditions and is determined by silanol adsorption and

condensation at the metal/PySi interface. Figure 8b shows the very modest protection offered by

hydrophobic mono-silane like OSi or a mixture of bis-amino and vinyltriacetxy silanes (BAV)

compared to PySi (at pH 4). In addition, although the film composed by a mixture of bis-amino and

OSi (BAPO) shows the highest breakdown and reflects the same features of PPySi (Fig. 8a), the

protection of the latter can be considered much better if silane composition is taken into account,

i.e., mono-silane versus a mixture of bis- and mono-silanes.


Figure 8. Anodic polarization scans of AA6061 (mild-alkaline cleaning) as a function of PySi-

based solution pH and silane nature: (a) bare Al alloy (―) AA6061 modified with PySi 4%v/v in

MetOH/H2O (95:5) at (--) pH 4 and (--) pH 6; (b) bare Al alloy (―) AA6061 modified with

similar solutions at pH 4 of (--) BAV mixture (3/1) 5%v/v, (--) OSi 5%v/v, (-◊-) BAOSi

mixture (2/1) 6%v/v.

Figure 9 shows the effect of the solvent nature and water content in silane formulation on film

performance for Mg-rich alloys AA5083 and AA5052, respectively. The amount of water seems to

be a more critical parameter according to the instable initial passivity of PPySi-coated alloys when

tBuOH and MetOH are used in a proportion solvent/water of 75:25. The negative effect of water in

silane-based treatments is typically related to incomplete condensation and induced structural defect

in the coating like hydrogen-bonded silicon. Notwithstanding the differences in the barrier action, at

currents between 10-8

-10-6

A/cm2 up to about 10

-3 A/cm

2, the curves show the “stairs-up-like” trend

attributed to the active action of PPy chains in the hybrid PPySi network. In this respect, it is

important to note that POSi shows only barrier-like protection, which is also importantly affected

by increasing water amount in silane formulation (Fig. 9a,b). Moreover, the use of HImSi does not

offer protection at all in contrast to PySi (Fig. 9b).

Other corrosion tests employed for PPySi protection performance evaluation, such as potential-time

monitoring and long-term immersion in NaCl solution, correlate well with the results discussed

above and support previous findings [9]. Some results are given in Figures 10 and 11. Open circuit

potential monitoring of AA6061 treated with PySi in MetOH/H2O (95:5) at pH 6 (Fig. 10), reveals

an average open circuit potential more positive than the corresponding of the bare substrate for

more than 7 hours with negative transients, suggesting the regeneration of the substrate passivity by

the anodic action of PPy in the hybrid PPySi coating.

(a) (b)

Figure 8. Anodic polarization scans of AA6061 (mild-alkaline cleaning) as a function of PySi-

based solution pH and silane nature: (a) bare Al alloy (―) AA6061 modified with PySi 4%v/v in

MetOH/H2O (95:5) at (--) pH 4 and (--) pH 6; (b) bare Al alloy (―) AA6061 modified with

similar solutions at pH 4 of (--) BAV mixture (3/1) 5%v/v, (--) OSi 5%v/v, (-◊-) BAOSi

mixture (2/1) 6%v/v.

Figure 9 shows the effect of the solvent nature and water content in silane formulation on film

performance for Mg-rich alloys AA5083 and AA5052, respectively. The amount of water seems to

be a more critical parameter according to the instable initial passivity of PPySi-coated alloys when

tBuOH and MetOH are used in a proportion solvent/water of 75:25. The negative effect of water in

silane-based treatments is typically related to incomplete condensation and induced structural defect

in the coating like hydrogen-bonded silicon. Notwithstanding the differences in the barrier action, at

currents between 10-8

-10-6

A/cm2 up to about 10

-3 A/cm

2, the curves show the “stairs-up-like” trend

attributed to the active action of PPy chains in the hybrid PPySi network. In this respect, it is

important to note that POSi shows only barrier-like protection, which is also importantly affected

by increasing water amount in silane formulation (Fig. 9a,b). Moreover, the use of HImSi does not

offer protection at all in contrast to PySi (Fig. 9b).

Other corrosion tests employed for PPySi protection performance evaluation, such as potential-time

monitoring and long-term immersion in NaCl solution, correlate well with the results discussed

above and support previous findings [9]. Some results are given in Figures 10 and 11. Open circuit

potential monitoring of AA6061 treated with PySi in MetOH/H2O (95:5) at pH 6 (Fig. 10), reveals

an average open circuit potential more positive than the corresponding of the bare substrate for

more than 7 hours with negative transients, suggesting the regeneration of the substrate passivity by

the anodic action of PPy in the hybrid PPySi coating.

(a) (b)


Figure 9. Anodic polarization scans of Mg-rich alloys as a function of water content and solvent

nature in silane formulation: (a) bare Al alloy (―) AA5083 modified with (--) PySi 4%v/v at pH

4 in tBuOH/H2O (75:25), (--) the same in tBuOH/H2O (90:10), (-◊-) OSi 4%v/v at pH 4 in

tBuOH/H2O (90:10); (b) bare Al alloy (―) AA5052 modified with solutions at 4%v/v pH 4 in

MetOH/H2O (75:25) of (--) HImSi, (-◊-) OSi, (--) PySi.

Figure 10. Open circuit potential monitoring in 0.6 M NaCl for AA6061: (····) bare alloy,

(―) treated with PySi in MetOH/H2O (95:5) at pH 6.

(a) (b)

Figure 9. Anodic polarization scans of Mg-rich alloys as a function of water content and solvent

nature in silane formulation: (a) bare Al alloy (―) AA5083 modified with (--) PySi 4%v/v at pH

4 in tBuOH/H2O (75:25), (--) the same in tBuOH/H2O (90:10), (-◊-) OSi 4%v/v at pH 4 in

tBuOH/H2O (90:10); (b) bare Al alloy (―) AA5052 modified with solutions at 4%v/v pH 4 in

MetOH/H2O (75:25) of (--) HImSi, (-◊-) OSi, (--) PySi.

Figure 10. Open circuit potential monitoring in 0.6 M NaCl for AA6061: (····) bare alloy,

(―) treated with PySi in MetOH/H2O (95:5) at pH 6.

(a) (b)


Figure 11 shows the surface SEM images of bare Al alloy AA2024 and the same treated with

methanol-based solutions of PySi and HImSi at pH 4 after immersion in NaCl solution for 3 days.

HImSi provides some protection (Fig. 11b) if compared to the bare specimen, even though

undercoating corrosion can be appreciated. No film disbonding but likely-passive defects are

revealed for PySi-treated alloy (Fig. 11c). This result further supports the anodic galvanic action of

PPy chains in the hybrid PPySi coating, which is in addition highly crosslinked due to pyrrole rings

linking and silanol condensation.

Figure 11. Surface SEM images of modified AA2024 after immersion for 3 days in 0.6 M NaCl

solution: (a) bare alloy, (b) treated with HImSi in MetOH/H2O (95:5) at pH 4, (c) treated with

PySi in MetOH/H2O (95:5) at pH 4.

XPS analysis of HImSi- and PySi-treated AA2024 was carried out before and after the immersion

test. Figure 12a,b show N1s and Cu 2p3/2 profiles of as-prepared PPySi film, which are composed

by a peak at 399.4 eV indicating aza-type nitrogen species (-N=) and at 932.8 eV attributed to Cu(I)

species, respectively. This result agrees with previously proposed Cu interaction with pyrrole ring

that promotes the formation of imine-like nitrogen [9,16]. After immersion, copper species were not

detected in the XPS spectra and N1s line is constitutes of two peaks (Fig.12c), one major peak at

399.9 eV assigned to neutral pyrrolylium nitrogen, and a second peak of lower intensity and higher

energy at 401.0 eV indicative of positively charged nitrogen species (N+). This electron-deficient

structure may be due either to a covalently bonded hydroxyl group or a delocalized positive charge

associated with the charge carrier [17]. Since the surface morphology does not reveal significant

attack (Fig. 11c), the second consideration appears to be the most probable, which is promoted by

copper leaching and dissolved oxygen-assisted reoxidation of PPy chains. The proportion of

positively charged nitrogen to neutral nitrogen (N+/N) equals 0.22, providing a measure of the

intrinsic oxidation state of PPy, which is closely similar to that reported typically for the doped

polymer.

Opposite behaviour was found for HImSi-treated AA2024. No copper whereas two nitrogen

species, namely imine-like and positively charged nitrogens, are revealed in the N1s profile before

and after immersion in NaCl solution, as illustrated in Figure 13. In addition, no shift in peaks

energy has occurred, positioned at 399.6 and 400.7 eV, respectively. The main difference is given

by the increase of –N= intensity coupled with the decrease of the N+ peak, which is a consequence

of increased basicity associated to the corrosion of the Al substrate. This correlates well to the

surface morphology in Figure 11b, revealing coating disbonding as a result of local pH increase.

(a) (b) (c)

Figure 11 shows the surface SEM images of bare Al alloy AA2024 and the same treated with

methanol-based solutions of PySi and HImSi at pH 4 after immersion in NaCl solution for 3 days.

HImSi provides some protection (Fig. 11b) if compared to the bare specimen, even though

undercoating corrosion can be appreciated. No film disbonding but likely-passive defects are

revealed for PySi-treated alloy (Fig. 11c). This result further supports the anodic galvanic action of

PPy chains in the hybrid PPySi coating, which is in addition highly crosslinked due to pyrrole rings

linking and silanol condensation.

Figure 11. Surface SEM images of modified AA2024 after immersion for 3 days in 0.6 M NaCl

solution: (a) bare alloy, (b) treated with HImSi in MetOH/H2O (95:5) at pH 4, (c) treated with

PySi in MetOH/H2O (95:5) at pH 4.

XPS analysis of HImSi- and PySi-treated AA2024 was carried out before and after the immersion

test. Figure 12a,b show N1s and Cu 2p3/2 profiles of as-prepared PPySi film, which are composed

by a peak at 399.4 eV indicating aza-type nitrogen species (-N=) and at 932.8 eV attributed to Cu(I)

species, respectively. This result agrees with previously proposed Cu interaction with pyrrole ring

that promotes the formation of imine-like nitrogen [9,16]. After immersion, copper species were not

detected in the XPS spectra and N1s line is constitutes of two peaks (Fig.12c), one major peak at

399.9 eV assigned to neutral pyrrolylium nitrogen, and a second peak of lower intensity and higher

energy at 401.0 eV indicative of positively charged nitrogen species (N+). This electron-deficient

structure may be due either to a covalently bonded hydroxyl group or a delocalized positive charge

associated with the charge carrier [17]. Since the surface morphology does not reveal significant

attack (Fig. 11c), the second consideration appears to be the most probable, which is promoted by

copper leaching and dissolved oxygen-assisted reoxidation of PPy chains. The proportion of

positively charged nitrogen to neutral nitrogen (N+/N) equals 0.22, providing a measure of the

intrinsic oxidation state of PPy, which is closely similar to that reported typically for the doped

polymer.

Opposite behaviour was found for HImSi-treated AA2024. No copper whereas two nitrogen

species, namely imine-like and positively charged nitrogens, are revealed in the N1s profile before

and after immersion in NaCl solution, as illustrated in Figure 13. In addition, no shift in peaks

energy has occurred, positioned at 399.6 and 400.7 eV, respectively. The main difference is given

by the increase of –N= intensity coupled with the decrease of the N+ peak, which is a consequence

of increased basicity associated to the corrosion of the Al substrate. This correlates well to the

surface morphology in Figure 11b, revealing coating disbonding as a result of local pH increase.

(a) (b) (c)


Figure 12. XPS analysis of AA2024 treated with PySi: (a) N1s line deconvolution of as-prepared

PPySi film, (b) Cu??? line deconvolution of as-prepared PPySi film, (c) N1s line deconvolution

after immersion in 0.6M NaCl solution for 3 days.

Figure 13. XPS N1s line deconvolution of AA2024 treated with HImSi: (a) as-prepared coated

specimen, (b) after immersion in 0.6M NaCl solution for 3 days.

(a) (b)

(c)

(a) (b)

Figure 12. XPS analysis of AA2024 treated with PySi: (a) N1s line deconvolution of as-prepared

PPySi film, (b) Cu??? line deconvolution of as-prepared PPySi film, (c) N1s line deconvolution

after immersion in 0.6M NaCl solution for 3 days.

Figure 13. XPS N1s line deconvolution of AA2024 treated with HImSi: (a) as-prepared coated

specimen, (b) after immersion in 0.6M NaCl solution for 3 days.

(a) (b)

(c)

(a) (b)


Aniline-based silane

Preliminary studies of aniline-based silane (AniSi) for surface treatment of wrought Al alloys

(Table 1) indicate as well that barrier/active protection can be obtained, as shown in Figure 14. The

AniSi-R treatment in the figure refers to AniSi solution aged for 10 days with an appreciable pink

coloration.

Figure 14 Anodic polarizations in 0.6M NaCl of wrought Al alloys (solid line) after treatment with

(--) AniSi and (--) AniSi-R solutions: (a) AA1050, (b) AA6082, (c) AA5083, (d) AA2024,

(e) AA7075.

(a) (b)

(c) (d)

(e)

Aniline-based silane

Preliminary studies of aniline-based silane (AniSi) for surface treatment of wrought Al alloys

(Table 1) indicate as well that barrier/active protection can be obtained, as shown in Figure 14. The

AniSi-R treatment in the figure refers to AniSi solution aged for 10 days with an appreciable pink

coloration.

Figure 14 Anodic polarizations in 0.6M NaCl of wrought Al alloys (solid line) after treatment with

(--) AniSi and (--) AniSi-R solutions: (a) AA1050, (b) AA6082, (c) AA5083, (d) AA2024,

(e) AA7075.

(a) (b)

(c) (d)

(e)


While in all samples a “stairs-up-like” breakdown is produced, the barrier action at the beginning of

the scan is mainly influenced by the alloy nature. For the most reactive alloys like AA2024 and

AA7075 this can be attributed to these metallic surfaces lower affinity for silanol adsorption and

condensation, limiting the coating adhesion, as it was pointed out for PySi [9]. The barrier action

appears also to be less effective when AniSi-R treatment is used, as revealed by the behaviour of

AA1050 and AA6082 in Figure 14a,b, respectively. The UV/Visible spectra of AniSi solution

reported as a function of time in Figure 15 show an emerging peak at 500 nm in concomitance with

the observed pink coloration. This has been attributed to radical cation of aniline as intermediate

[18]. The presence of charged species in AniSi-R solution would originate a more porous structure

due to counterions needed for charge compensation, thus limiting the barrier properties of the

coating. It is to be noted that the absorption spectrum intensity importantly decrease after about a

month and a red-brownish precipitate was observed at the test tube walls. This indicates that AniSi

oligomeric species in solution are less stable than those derived from PySi (Fig. 6b), showing any

precipitation for more than two month.

Figure 15. UV-Visible absorption spectra as a function of time of AniSi methanol-based solution.

Figure 16 compares the behaviour obtained for AA2024 and AA6082 after treatment with acidic

solutions of AniSi, AniSi-R and OSi, as well as with AniSi solution at its natural pH (8.5). All

AniSi-based coatings perform better than the corresponding based on the alkylsilane, even though

acidic conditions are preferable for the former.

While in all samples a “stairs-up-like” breakdown is produced, the barrier action at the beginning of

the scan is mainly influenced by the alloy nature. For the most reactive alloys like AA2024 and

AA7075 this can be attributed to these metallic surfaces lower affinity for silanol adsorption and

condensation, limiting the coating adhesion, as it was pointed out for PySi [9]. The barrier action

appears also to be less effective when AniSi-R treatment is used, as revealed by the behaviour of

AA1050 and AA6082 in Figure 14a,b, respectively. The UV/Visible spectra of AniSi solution

reported as a function of time in Figure 15 show an emerging peak at 500 nm in concomitance with

the observed pink coloration. This has been attributed to radical cation of aniline as intermediate

[18]. The presence of charged species in AniSi-R solution would originate a more porous structure

due to counterions needed for charge compensation, thus limiting the barrier properties of the

coating. It is to be noted that the absorption spectrum intensity importantly decrease after about a

month and a red-brownish precipitate was observed at the test tube walls. This indicates that AniSi

oligomeric species in solution are less stable than those derived from PySi (Fig. 6b), showing any

precipitation for more than two month.

Figure 15. UV-Visible absorption spectra as a function of time of AniSi methanol-based solution.

Figure 16 compares the behaviour obtained for AA2024 and AA6082 after treatment with acidic

solutions of AniSi, AniSi-R and OSi, as well as with AniSi solution at its natural pH (8.5). All

AniSi-based coatings perform better than the corresponding based on the alkylsilane, even though

acidic conditions are preferable for the former.


Figure 16. Anodic polarizations in 0.6M NaCl of: (a) AA2024 and (b) AA6082, after treatment

with (--) AniSi, (--) AniSi-R, (-◊-) AniSi at pH 8.5, (----) OSi.

Conclusions

Generally, silane-based coatings act essentially as a physical barrier, while doping with chemicals

having corrosion inhibiting properties is needed for these films to acquire active protection. The

present work has supported our previous investigations [9], showing that active behaviour can be

obtained if silane molecule contains a monomer of a conducting polymer as a functional group.

Thus, a hybrid structure with barrier/active actions, as well as improved morphology and adhesion,

can be obtained following a simplified procedure.

References

[1] Information on http://www.world-aluminium.org

[2] (a) D.E. Tallman, G. Spinks, A. Dominis, G.G. Wallace, J. Solid State Electrochem. 6 (2002),

p. 73; (b) G. Spinks, A. Dominis, G.G. Wallace, D.E. Tallman, J. Solid State Electrochem. 6

(2002), p. 85

[3] T.D. Nguyen, M. Keddam, H. Takenouti, Electrochem Solid State Lett. 6 (2003), p. B25

[4] J. He, D.E. Tallman, G.P. Bierwagen, J. Electrochem. Soc. 151 (2004), p. B644

[5] G. Paliwoda-Porebska, M. Rohwerder, M. Stratmann, U. Rammelt, L.M. Duc, W. Plieth,

J. Solid State Electrochem. 10 (2006), p. 730

[6] I.L. Lehr, S.B. Saidman, Corr. Sci. 49 (2007), p. 2210

[7] N.C.T. Martins, T. Moura e Silva, M.F. Montemor, J.C.S. Fernandes, M.G.S. Ferreira,

Electrochim. Acta 53 (2008), p. 4754

[8] K.R.L. Castagno, D.S. Azambuja, V. Dalmoro, J. Appl. Electrochem. 39 (2009), p. 93

[9] (a) M. Trueba, S.P. Trasatti, Prog. Org. Coat. 66 (2009), p. 254; (b) ibid, p. 265

[10] ASTM G31 – 72 (1999) Standard Practice for Laboratory Immersion Corrosion Testing of

Metals, US

[11] D.L. Elmore, D.B. Chase, Y. Liu, J.F. Rabolt, Vib. Spectrosc. 34 (2004), p. 37

(a) (b)

Figure 16. Anodic polarizations in 0.6M NaCl of: (a) AA2024 and (b) AA6082, after treatment

with (--) AniSi, (--) AniSi-R, (-◊-) AniSi at pH 8.5, (----) OSi.

Conclusions

Generally, silane-based coatings act essentially as a physical barrier, while doping with chemicals

having corrosion inhibiting properties is needed for these films to acquire active protection. The

present work has supported our previous investigations [9], showing that active behaviour can be

obtained if silane molecule contains a monomer of a conducting polymer as a functional group.

Thus, a hybrid structure with barrier/active actions, as well as improved morphology and adhesion,

can be obtained following a simplified procedure.

References

[1] Information on http://www.world-aluminium.org

[2] (a) D.E. Tallman, G. Spinks, A. Dominis, G.G. Wallace, J. Solid State Electrochem. 6 (2002),

p. 73; (b) G. Spinks, A. Dominis, G.G. Wallace, D.E. Tallman, J. Solid State Electrochem. 6

(2002), p. 85

[3] T.D. Nguyen, M. Keddam, H. Takenouti, Electrochem Solid State Lett. 6 (2003), p. B25

[4] J. He, D.E. Tallman, G.P. Bierwagen, J. Electrochem. Soc. 151 (2004), p. B644

[5] G. Paliwoda-Porebska, M. Rohwerder, M. Stratmann, U. Rammelt, L.M. Duc, W. Plieth,

J. Solid State Electrochem. 10 (2006), p. 730

[6] I.L. Lehr, S.B. Saidman, Corr. Sci. 49 (2007), p. 2210

[7] N.C.T. Martins, T. Moura e Silva, M.F. Montemor, J.C.S. Fernandes, M.G.S. Ferreira,

Electrochim. Acta 53 (2008), p. 4754

[8] K.R.L. Castagno, D.S. Azambuja, V. Dalmoro, J. Appl. Electrochem. 39 (2009), p. 93

[9] (a) M. Trueba, S.P. Trasatti, Prog. Org. Coat. 66 (2009), p. 254; (b) ibid, p. 265

[10] ASTM G31 – 72 (1999) Standard Practice for Laboratory Immersion Corrosion Testing of

Metals, US

[11] D.L. Elmore, D.B. Chase, Y. Liu, J.F. Rabolt, Vib. Spectrosc. 34 (2004), p. 37

(a) (b)


[12] M.-C. Brochier Salon, P.-A. Bayle, M. Abdelmouleh, S. Boufi, M. Naceur Belgacem, Colloids

Surf. A: Physicochem. Eng. Aspects 312 (2008), p. 83

[13] S. Okur, U. Salzner, J. Phys. Chem. A 112 (2008), p. 11842

[14] M. Fujiki, J. Am. Chem. Soc. 118 (1996), p. 7424

[15] G.P. van der Laan, M.P. de Haas, A. Hummel, H. Frey, M. Moller, J. Phys. Chem. 100 (1996),

p. 5470

[16] V.W.L. Lim, E.T. Kang, K.G. Neoh, Macromol. Chem. Phys. 202 (2001), p. 2824

[17] M.V. Zeller, S.J. Hahn, Surf. Iterface Anal. 11 (1988), p. 327

[18] K.G. Neoh, E.T. Kang, K.L. Tan, Polymer 34 (1993), p. 3921

[12] M.-C. Brochier Salon, P.-A. Bayle, M. Abdelmouleh, S. Boufi, M. Naceur Belgacem, Colloids

Surf. A: Physicochem. Eng. Aspects 312 (2008), p. 83

[13] S. Okur, U. Salzner, J. Phys. Chem. A 112 (2008), p. 11842

[14] M. Fujiki, J. Am. Chem. Soc. 118 (1996), p. 7424

[15] G.P. van der Laan, M.P. de Haas, A. Hummel, H. Frey, M. Moller, J. Phys. Chem. 100 (1996),

p. 5470

[16] V.W.L. Lim, E.T. Kang, K.G. Neoh, Macromol. Chem. Phys. 202 (2001), p. 2824

[17] M.V. Zeller, S.J. Hahn, Surf. Iterface Anal. 11 (1988), p. 327

[18] K.G. Neoh, E.T. Kang, K.L. Tan, Polymer 34 (1993), p. 3921


A composite coating for corrosion and wear protection of AM60B

magnesium alloy

Anna Da Forno1, a and Massimiliano Bestetti1, b 1 Politecnico di Milano, Via Mancinelli 7, 20131 Milano ITALY

[email protected],

b [email protected]

Keywords: Magnesium alloys, micro-arc anodizing, corrosion resistance, wear resistance.

Abstract. In this paper a protection process against corrosion and wear for AM60B magnesium

alloys, by multilayer approach, is related. The coating consists of a porous oxide layer, obtained in

micro-arc anodizing regime, and two or three layers deposited by sol-gel technique. The anodic

oxidation pre-treatment improves the adhesion of the sol-gel layers, which are responsible for the

sealing of the anodic oxide pores and for the corrosion protection effect. In addition the multilayer

system significantly improves AM60B alloy wear resistance. Scanning Electron Microscopy (SEM)

and X-Ray Diffraction (XRD) were employed to assess morphology and crystallographic structure.

Electrochemical polarization and wear tests were performed in order to evaluate the corrosion

resistance behaviour and the wear resistance of the coated magnesium alloys.

Introduction

Magnesium alloys are extremely attractive engineering materials and are employed in a variety of

applications. Because of their low density and high strength to weight ratio they are of interest to

transport industries and where a weight reduction is required. However, magnesium is too reactive

and generally exhibits a poor corrosion resistance particularly in chloride environments and by

galvanic coupling with more noble metals. Moreover impurities and second phases act as active

cathodic sites causing local galvanic acceleration of corrosion [1]. Many techniques have been

developed in order to increase corrosion resistance of magnesium alloys such as electroplating,

conversion coatings, anodic oxidation [2]. Micro arc anodic oxidation is a relatively new surface

treatment technique and emphasis is drawn to its potential for surface modification. By means of

this process mixed oxides can be synthesised [3], containing the elements coming both from the

substrate and the electrolytic solution, which significantly affect the surface properties and favour

the adhesion of post-treatment coatings, for example painting [4] or hybrid organic-inorganic sol gel

coatings [5]. Usually to improve the adhesion between a metal and a non metallic coating, an

interlayer compatible with both is necessary [6]. Hybrid organic-inorganic coatings have been

suggested as top coats on anodic oxides; the inorganic component improves the mechanical

properties while the inorganic one increases flexibility and compatibility with organic paint systems.

In this paper an anodic oxide produced by micro-arc anodizing is used as interlayer to improve the

adhesion of a hybrid organic-inorganic top coat layer produced by sol-gel technique and deposited

by spraying.

Experimental

Anodic oxidation experiments were performed on a die cast AM60B alloy (Q-panel) provided by

Norsk Hydro (Table 1).

A composite coating for corrosion and wear protection of AM60B

magnesium alloy

Anna Da Forno1, a and Massimiliano Bestetti1, b 1 Politecnico di Milano, Via Mancinelli 7, 20131 Milano ITALY

[email protected],

b [email protected]

Keywords: Magnesium alloys, micro-arc anodizing, corrosion resistance, wear resistance.

Abstract. In this paper a protection process against corrosion and wear for AM60B magnesium

alloys, by multilayer approach, is related. The coating consists of a porous oxide layer, obtained in

micro-arc anodizing regime, and two or three layers deposited by sol-gel technique. The anodic

oxidation pre-treatment improves the adhesion of the sol-gel layers, which are responsible for the

sealing of the anodic oxide pores and for the corrosion protection effect. In addition the multilayer

system significantly improves AM60B alloy wear resistance. Scanning Electron Microscopy (SEM)

and X-Ray Diffraction (XRD) were employed to assess morphology and crystallographic structure.

Electrochemical polarization and wear tests were performed in order to evaluate the corrosion

resistance behaviour and the wear resistance of the coated magnesium alloys.

Introduction

Magnesium alloys are extremely attractive engineering materials and are employed in a variety of

applications. Because of their low density and high strength to weight ratio they are of interest to

transport industries and where a weight reduction is required. However, magnesium is too reactive

and generally exhibits a poor corrosion resistance particularly in chloride environments and by

galvanic coupling with more noble metals. Moreover impurities and second phases act as active

cathodic sites causing local galvanic acceleration of corrosion [1]. Many techniques have been

developed in order to increase corrosion resistance of magnesium alloys such as electroplating,

conversion coatings, anodic oxidation [2]. Micro arc anodic oxidation is a relatively new surface

treatment technique and emphasis is drawn to its potential for surface modification. By means of

this process mixed oxides can be synthesised [3], containing the elements coming both from the

substrate and the electrolytic solution, which significantly affect the surface properties and favour

the adhesion of post-treatment coatings, for example painting [4] or hybrid organic-inorganic sol gel

coatings [5]. Usually to improve the adhesion between a metal and a non metallic coating, an

interlayer compatible with both is necessary [6]. Hybrid organic-inorganic coatings have been

suggested as top coats on anodic oxides; the inorganic component improves the mechanical

properties while the inorganic one increases flexibility and compatibility with organic paint systems.

In this paper an anodic oxide produced by micro-arc anodizing is used as interlayer to improve the

adhesion of a hybrid organic-inorganic top coat layer produced by sol-gel technique and deposited

by spraying.

Experimental

Anodic oxidation experiments were performed on a die cast AM60B alloy (Q-panel) provided by

Norsk Hydro (Table 1).

Table 1. Chemical composition of AM60B alloy (wt. %).

Al Zn Mn Si Fe Cu Ni Mg

5.9 0.05 0.27 0.025 0.0017 0.0017 0.0004 balance

Specimens of 10×3x0.3 cm were cut from Q-panels, polished with P1200 silicon carbide abrasive

paper, degreased with acetone, rinsed with distilled water and dried in air. All the solutions were

prepared by using analytical grade reagents and distilled water. The samples were anodized in

alkaline electrolyte which consists of 3.0 M potassium hydroxide, 0.21 M sodium phosphate,

0.15 M aluminium nitrate with or without addition of 0.6 M potassium fluoride. Two AISI 316

panels were used as cathodes. The anodization processes were carried out under voltage control and

the initial temperature of the electrolytic solution was 30°C. During anodization the cell voltage

raised linearly from 0 to the maintenance voltage (70-80V) in a fixed time ramp of 255 s, and then it

was kept at maintenance voltage, with a total time of 10 min. Beyond a threshold cell voltage

(around 55 V) sparks occur on the anode surface, at higher voltages they become more and more

intense and in the last minutes they become smaller and less intense until they disappear.

A hybrid sol-gel derived coating was prepared by mixing together with TEOS (4.7 g Aldrich), 3-

metacryloxypropyl trimethoxysilane (MEMO 10.4 g Aldrich, 99.7 %), ethyl alcohol (15.8 g Fluka,

96%), distilled water (4.9 g), tert-butylhydroperoxide (1.9 g Aldrich, 99.7%). Sol-gel coatings were

applied to the anodized samples by spraying them with 2 ml cm-2

of fresh solution. One to three

layers were applied to the samples. Every layer was cured at 70°C for 24 hours for intermediate

passes and 48 hours for the final ones.

Surface morphology and cross-section of anodized coatings were observed by scanning electron

microscopy (SEM) and the phase composition was investigated by X-ray diffraction (XRD).

Hardness tests were performed on a micro-hardness tester Fisher Fisherscope H 100CP, with a

loading rate of 10 mN per 10 s on the surface. The polarization measurements were conducted in

aqueous solution of 3.5 wt% sodium chloride at 25oC by using a conventional three electrode cell:

an activated titanium mesh was used as counter electrode and a saturated calomel electrode SCE as

the reference electrode. The specimens were covered by Teflon except for an area of 1×1cm2 and

acted as working electrode. All the measurements were performed with a computer controlled

EG&G Princeton Applied Research PAR 273A potentiostat. The scan was conducted from -2.1V to

0.5V with a scan rate of 5 mV s-1

. Flat-on-cylinder wear tests (load: 3 kg, ω: 200 turns min-1

,

t: 5 min, ∅ 4 cm) were carried out on anodized and sol-gel coated samples, on only anodized

samples and on as cast AM60B.

Results

Figure 1 shows the SEM surface morphology of AM60B samples anodized at cell voltage 70 V at

different anodization times (2 min; 5 min; 9 min). After 2 minutes the porosity is not yet developed,

while after 5 min and especially after 9 min pores are quite evident. Because of the high porosity of

the anodic oxides produced by micro-arc anodizing the contact between the substrate and the

environment aggressive compounds is not prevented. To provide a sufficient protection against

corrosion a thin anodic oxide is grown in order to favour the adhesion of the top coat layers. As the

micrograph of Figure 2 shows, the anodic oxide thickness is around 1.5 µm after10 minutes.

Table 1. Chemical composition of AM60B alloy (wt. %).

Al Zn Mn Si Fe Cu Ni Mg

5.9 0.05 0.27 0.025 0.0017 0.0017 0.0004 balance

Specimens of 10×3x0.3 cm were cut from Q-panels, polished with P1200 silicon carbide abrasive

paper, degreased with acetone, rinsed with distilled water and dried in air. All the solutions were

prepared by using analytical grade reagents and distilled water. The samples were anodized in

alkaline electrolyte which consists of 3.0 M potassium hydroxide, 0.21 M sodium phosphate,

0.15 M aluminium nitrate with or without addition of 0.6 M potassium fluoride. Two AISI 316

panels were used as cathodes. The anodization processes were carried out under voltage control and

the initial temperature of the electrolytic solution was 30°C. During anodization the cell voltage

raised linearly from 0 to the maintenance voltage (70-80V) in a fixed time ramp of 255 s, and then it

was kept at maintenance voltage, with a total time of 10 min. Beyond a threshold cell voltage

(around 55 V) sparks occur on the anode surface, at higher voltages they become more and more

intense and in the last minutes they become smaller and less intense until they disappear.

A hybrid sol-gel derived coating was prepared by mixing together with TEOS (4.7 g Aldrich), 3-

metacryloxypropyl trimethoxysilane (MEMO 10.4 g Aldrich, 99.7 %), ethyl alcohol (15.8 g Fluka,

96%), distilled water (4.9 g), tert-butylhydroperoxide (1.9 g Aldrich, 99.7%). Sol-gel coatings were

applied to the anodized samples by spraying them with 2 ml cm-2

of fresh solution. One to three

layers were applied to the samples. Every layer was cured at 70°C for 24 hours for intermediate

passes and 48 hours for the final ones.

Surface morphology and cross-section of anodized coatings were observed by scanning electron

microscopy (SEM) and the phase composition was investigated by X-ray diffraction (XRD).

Hardness tests were performed on a micro-hardness tester Fisher Fisherscope H 100CP, with a

loading rate of 10 mN per 10 s on the surface. The polarization measurements were conducted in

aqueous solution of 3.5 wt% sodium chloride at 25oC by using a conventional three electrode cell:

an activated titanium mesh was used as counter electrode and a saturated calomel electrode SCE as

the reference electrode. The specimens were covered by Teflon except for an area of 1×1cm2 and

acted as working electrode. All the measurements were performed with a computer controlled

EG&G Princeton Applied Research PAR 273A potentiostat. The scan was conducted from -2.1V to

0.5V with a scan rate of 5 mV s-1

. Flat-on-cylinder wear tests (load: 3 kg, ω: 200 turns min-1

,

t: 5 min, ∅ 4 cm) were carried out on anodized and sol-gel coated samples, on only anodized

samples and on as cast AM60B.

Results

Figure 1 shows the SEM surface morphology of AM60B samples anodized at cell voltage 70 V at

different anodization times (2 min; 5 min; 9 min). After 2 minutes the porosity is not yet developed,

while after 5 min and especially after 9 min pores are quite evident. Because of the high porosity of

the anodic oxides produced by micro-arc anodizing the contact between the substrate and the

environment aggressive compounds is not prevented. To provide a sufficient protection against

corrosion a thin anodic oxide is grown in order to favour the adhesion of the top coat layers. As the

micrograph of Figure 2 shows, the anodic oxide thickness is around 1.5 µm after10 minutes.


Figure 1- SEM surface micrographs of AM60B after anodizing (70 V, 2 min - 5 min and 9 min).

Figure 2- Thickness of anodic oxide produced at 70V (10 min).

X-ray diffraction has been employed to assess the crystallographic structure of the anodic oxides.

The X-ray diffraction (XRD) analysis (Figure 3) shows that the layer produced by micro-arc

anodizing mainly consists of magnesium oxide MgO.

Figure 3- XRD pattern of anodic oxide

Figure 1- SEM surface micrographs of AM60B after anodizing (70 V, 2 min - 5 min and 9 min).

Figure 2- Thickness of anodic oxide produced at 70V (10 min).

X-ray diffraction has been employed to assess the crystallographic structure of the anodic oxides.

The X-ray diffraction (XRD) analysis (Figure 3) shows that the layer produced by micro-arc

anodizing mainly consists of magnesium oxide MgO.

Figure 3- XRD pattern of anodic oxide


Table 2-Vickers microhardness (10mN, 10s) of anodized and

sol gel treated samples.

Sample

Vickers microhardness, HV

Anodic oxide 181±18

Anodic oxide +2 sol-gel

layers 405±32


layers 351±97

AM60B 120±27

AM60B+2 sol-gel layers 142±34


Table 2 reports Vickers microhardness values of samples anodized and sol-gel coated by spraying.

Because the sol-gel coating thickness is around 1.5µm, hardness values are affected by the

underlying oxide layer. By comparing the hardness values it’s possible to conclude that there are no

significant differences in hardness values between the bare alloy and the alloy coated with two or

three sol-gel layers probably because the hybrid film is absent or very thin, while the specimens

with magnesium anodic oxide as inter layers have higher hardness particularly when two sol-gel

layers are applied. The hardness of the multilayer systems anodic oxide + 2 sol-gel layers is almost

twice of those of anodic oxide. These results prove that magnesium oxides play an important role

for the sol-gel coatings adhesion.

Flat on cylinder wear tests were carried out onto the bare alloy, the alloy coated with two sol-gel

layers, the anodic oxide and the anodic oxide coated with two sol-gel layers. The removed volumes

(V/L) are reported in Table 3.

Table 3- Removed volume (V/L) in wear tests

Sample

Removed volume (V/L),

mm2

Anodic oxide 0.016

Anodic oxide +2 sol-gel layers 0.006

AM60B 0.070

AM60B+2 sol-gel layers 0.115

Anodised and sol-gel coated samples have significantly lower value of removed volumes than the

specimens untreated or directly sol-gel coated. Therefore sol-gel films, albeit organic-inorganic, are

able to improve significantly the wear behaviour compared with the uncoated samples.

Figure 4 and Figure 5 show that sol-gel treatments (two or three sol-gel layers) performed on the

bare alloy do not increase the corrosion resistance. The sample produced in fluoride free solution

has a corrosion current density of around 3 x 10-8

A cm-2

and a corrosion potential of -0.5V, while

the corrosion current density of the sample coated with two sol-gel layers without the anodic oxide

as interlayer is 10-5

A cm-2

and the corrosion potential is -1.5V.

Table 2-Vickers microhardness (10mN, 10s) of anodized and

sol gel treated samples.

Sample

Vickers microhardness, HV

Anodic oxide 181±18


layers 405±32


layers 351±97

AM60B 120±27



Table 2 reports Vickers microhardness values of samples anodized and sol-gel coated by spraying.

Because the sol-gel coating thickness is around 1.5µm, hardness values are affected by the

underlying oxide layer. By comparing the hardness values it’s possible to conclude that there are no

significant differences in hardness values between the bare alloy and the alloy coated with two or

three sol-gel layers probably because the hybrid film is absent or very thin, while the specimens

with magnesium anodic oxide as inter layers have higher hardness particularly when two sol-gel

layers are applied. The hardness of the multilayer systems anodic oxide + 2 sol-gel layers is almost

twice of those of anodic oxide. These results prove that magnesium oxides play an important role

for the sol-gel coatings adhesion.

Flat on cylinder wear tests were carried out onto the bare alloy, the alloy coated with two sol-gel

layers, the anodic oxide and the anodic oxide coated with two sol-gel layers. The removed volumes

(V/L) are reported in Table 3.

Table 3- Removed volume (V/L) in wear tests

Sample

Removed volume (V/L),

mm2

Anodic oxide 0.016

Anodic oxide +2 sol-gel layers 0.006

AM60B 0.070

AM60B+2 sol-gel layers 0.115

Anodised and sol-gel coated samples have significantly lower value of removed volumes than the

specimens untreated or directly sol-gel coated. Therefore sol-gel films, albeit organic-inorganic, are

able to improve significantly the wear behaviour compared with the uncoated samples.

Figure 4 and Figure 5 show that sol-gel treatments (two or three sol-gel layers) performed on the

bare alloy do not increase the corrosion resistance. The sample produced in fluoride free solution

has a corrosion current density of around 3 x 10-8

A cm-2

and a corrosion potential of -0.5V, while

the corrosion current density of the sample coated with two sol-gel layers without the anodic oxide

as interlayer is 10-5

A cm-2

and the corrosion potential is -1.5V.


Figure 4- Electrochemical polarisation curves of the alloy, the anodic oxide produced in fluoride

solution, the oxide produced in fluoride free solution, coated with 2 sol-gel layers.

Figure 5 - Electrochemical polarisation curves of the alloy, the anodic oxide produced in fluoride


Conclusions

In this study a way to protect AM60B magnesium alloy specimens by means of a composite coating

was evaluated. The samples were anodized by micro-arc technique and the resulting MgO oxide

was sealed with a hybrid organic-inorganic layer. The electrochemical polarization tests showed that

the composite coatings composed by anodic oxide and two sol-gel layers produced an increase of

the corrosion potential [-0.5V(SCE)] and three orders of magnitude decrease of the corrosion

current densities (10-8

A cm-2

). A significant enhancement of wear resistance in comparison with

anodized or untreated samples was also achieved.

Figure 4- Electrochemical polarisation curves of the alloy, the anodic oxide produced in fluoride


Figure 5 - Electrochemical polarisation curves of the alloy, the anodic oxide produced in fluoride


Conclusions

In this study a way to protect AM60B magnesium alloy specimens by means of a composite coating

was evaluated. The samples were anodized by micro-arc technique and the resulting MgO oxide

was sealed with a hybrid organic-inorganic layer. The electrochemical polarization tests showed that

the composite coatings composed by anodic oxide and two sol-gel layers produced an increase of

the corrosion potential [-0.5V(SCE)] and three orders of magnitude decrease of the corrosion

current densities (10-8

A cm-2

). A significant enhancement of wear resistance in comparison with

anodized or untreated samples was also achieved.


References

[1] G. Song and A. Atrens: Adv. Eng. Mat. Vol. 5 (2003), p. 837.

[2] J.E. Gray and B. Luan: J. Alloys Compd. Vol. 336 (2002), p. 88.

[3] A.L. Yerokhin, X. Nie, A. Leyland, A. Matthews and S.J. Dowey, Surf. Coat. Technol. Vol.122

(1999), p. 73.

[4] M. Bestetti, P.L. Cavallotti, A. Da Forno and S. Pozzi. Trans. IMF Vol. 85(6) (2007), p. 316.

[5] A.N. Kramov, V.N. Balbyshev, L.S. Kasten and R.A.Mantz, Thin Solid Films Vol 514 (2006),

p. 174.

[6] W. Boysen, A. Frattini, N. Pellegri and O.de Sanctis, Surf. Coat. Technol.Vol. 122 (1999), p. 14

References

[1] G. Song and A. Atrens: Adv. Eng. Mat. Vol. 5 (2003), p. 837.

[2] J.E. Gray and B. Luan: J. Alloys Compd. Vol. 336 (2002), p. 88.

[3] A.L. Yerokhin, X. Nie, A. Leyland, A. Matthews and S.J. Dowey, Surf. Coat. Technol. Vol.122

(1999), p. 73.

[4] M. Bestetti, P.L. Cavallotti, A. Da Forno and S. Pozzi. Trans. IMF Vol. 85(6) (2007), p. 316.

[5] A.N. Kramov, V.N. Balbyshev, L.S. Kasten and R.A.Mantz, Thin Solid Films Vol 514 (2006),

p. 174.

[6] W. Boysen, A. Frattini, N. Pellegri and O.de Sanctis, Surf. Coat. Technol.Vol. 122 (1999), p. 14


Continuum damage model for biodegradable Magnesium alloy stent

Dario Gastaldi1, a, Valentina Sassi1, Lorenza Petrini1, Maurizio Vedani2, Stefano Trasatti3, Francesco Migliavacca1

1Laboratory of Biological Structure Mechanics (LaBS) Department of Structural Engineering,

Politecnico di Milano piazza Leonardo da Vinci 32, 20133, Milano, Italy.

2Dept. of Mechanical Engineering, Politecnico di Milano, Italy

3Dept. of Physical Chemistry and Electrochemistry, University of Milan, Italy

[email protected]

Keywords: Bioresorbable stent, damage model, Magnesium alloy, Metallic biomaterials.

Abstract. The main drawback of conventional stenting procedure is the high risk of restenosis. The

idea of a stent that "disappears" after having fulfilled its mission is very intriguing and fascinating.

The stent mass should diminishing in time to allow the gradual transmission of the mechanical load

to the surrounding tissues. Magnesium and its alloys seem to be among the most appealing materials

to design biodegradable stents. The objective of this work is to develop, in a finite element (FE)

framework, a model of magnesium degradation able to predict the corrosion rate and thus providing

a valuable tool to design biodegradable stents. Continuum damage approach is suitable for

modelling different damage mechanisms, including several types of corrosion. Corrosion is

modelled by a scalar damage field which accounts for the material strength loss due to geometrical

discontinuities. As damage progresses, the material stiffness decreases. Corrosion damage results as

the superposition of stress corrosion process and uniform corrosion. The former describes the stress-

mediated localization of the corrosion attack through a stress-dependent evolution law similar to the

one used in analytical models, while the latter affects the free surface of the material exposed to an

aggressive environment. The effects of both phenomena described are modelled through a linear

composition of the two specific damage evolution laws. The model, developed in a FE framework,

manages the mesh dependency, typical of strain-softening behaviour, including the FE characteristic

length in the damage evolution law definition. The developed model is able to reproduce the

behaviour of different magnesium alloys subjected to static and slow-strain-rate corrosion tests.

Moreover, 3D stenting procedures accounting for the interaction with the arterial vessel are

simulated.

Introduction

From the beginning of stenting clinical routine (1990s) millions of atherosclerotic arteries have been

treated avoiding very invasive surgical operations with great advantages for patients and National

Health Systems. However, if stent implantation reduces the stenosis either in the immediate and in

the long term period, the device presence in the blood vessel may induce a new tissue thickening

nullifying the effect of the procedure. Among the solutions proposed to overcome this drawback, the

adoption of degradable materials coupled with new shapes of the device seems to be very

promising. Recently, particular attention has been devoted to the use of magnesium (Mg) alloys for

designing innovative stents [1, 2]. Indeed, the dissolution of the magnesium inside the human body

is well tolerated, being magnesium one of the main component of the human metabolism.

Moreover, its mechanical properties allow to sustain the arterial wall and to prevent the elastic

recoil of the vessel after stent expansion. On the other hand, the high rate of magnesium corrosion

in a physiological environment has limited up to now its employment for endoprostheses

production. In this context, the development of a numerical procedure that allows to study the

efficacy of different stent designs made of different magnesium alloys might be very useful for

Continuum damage model for biodegradable Magnesium alloy stent

Dario Gastaldi1, a, Valentina Sassi1, Lorenza Petrini1, Maurizio Vedani2, Stefano Trasatti3, Francesco Migliavacca1

1Laboratory of Biological Structure Mechanics (LaBS) Department of Structural Engineering,

Politecnico di Milano piazza Leonardo da Vinci 32, 20133, Milano, Italy.

2Dept. of Mechanical Engineering, Politecnico di Milano, Italy

3Dept. of Physical Chemistry and Electrochemistry, University of Milan, Italy

[email protected]

Keywords: Bioresorbable stent, damage model, Magnesium alloy, Metallic biomaterials.

Abstract. The main drawback of conventional stenting procedure is the high risk of restenosis. The

idea of a stent that "disappears" after having fulfilled its mission is very intriguing and fascinating.

The stent mass should diminishing in time to allow the gradual transmission of the mechanical load

to the surrounding tissues. Magnesium and its alloys seem to be among the most appealing materials

to design biodegradable stents. The objective of this work is to develop, in a finite element (FE)

framework, a model of magnesium degradation able to predict the corrosion rate and thus providing

a valuable tool to design biodegradable stents. Continuum damage approach is suitable for

modelling different damage mechanisms, including several types of corrosion. Corrosion is

modelled by a scalar damage field which accounts for the material strength loss due to geometrical

discontinuities. As damage progresses, the material stiffness decreases. Corrosion damage results as

the superposition of stress corrosion process and uniform corrosion. The former describes the stress-

mediated localization of the corrosion attack through a stress-dependent evolution law similar to the

one used in analytical models, while the latter affects the free surface of the material exposed to an

aggressive environment. The effects of both phenomena described are modelled through a linear

composition of the two specific damage evolution laws. The model, developed in a FE framework,

manages the mesh dependency, typical of strain-softening behaviour, including the FE characteristic

length in the damage evolution law definition. The developed model is able to reproduce the

behaviour of different magnesium alloys subjected to static and slow-strain-rate corrosion tests.

Moreover, 3D stenting procedures accounting for the interaction with the arterial vessel are

simulated.

Introduction

From the beginning of stenting clinical routine (1990s) millions of atherosclerotic arteries have been

treated avoiding very invasive surgical operations with great advantages for patients and National

Health Systems. However, if stent implantation reduces the stenosis either in the immediate and in

the long term period, the device presence in the blood vessel may induce a new tissue thickening

nullifying the effect of the procedure. Among the solutions proposed to overcome this drawback, the

adoption of degradable materials coupled with new shapes of the device seems to be very

promising. Recently, particular attention has been devoted to the use of magnesium (Mg) alloys for

designing innovative stents [1, 2]. Indeed, the dissolution of the magnesium inside the human body

is well tolerated, being magnesium one of the main component of the human metabolism.

Moreover, its mechanical properties allow to sustain the arterial wall and to prevent the elastic

recoil of the vessel after stent expansion. On the other hand, the high rate of magnesium corrosion

in a physiological environment has limited up to now its employment for endoprostheses

production. In this context, the development of a numerical procedure that allows to study the

efficacy of different stent designs made of different magnesium alloys might be very useful for

bioengineering company involved in stent production and might give a contribution in the

improvement of stenting routine success.


Finite element (FE) models accounting for degradation may provide a valuable tool to obtain

information about the performance of biodegradable devices over time.

Micro-scale models of corrosion based on diffusion-controlled processes are present in literature

[3]. Such kind of models, even if consistent with the multiphysics degradation process, are not

easily applicable to an actual medical device due to the difficulty to operate experimental-based

calibration procedures for all model parameters and the high computational costs of tridimensional

geometries. Therefore it is necessary to describe the corrosion behavior at a higher dimensional

scale adopting a phenomenological approach. Continuum damage mechanics (CDM) allows to

model the loss of mechanical strength of a material due to the presence of geometrical

discontinuities through the definition of a scalar field which quantifies the distribution of the

damage [4,5].

In this work two corrosion mechanisms are considered as different degradation processes.

Assuming a linear superposition of the different scalar fields the corrosion damage variable D

responsible for the global degradation is:

D=DU+DSC

where the uniform corrosion damage DU accounts for the mass loss occurring to the material

when exposed to aggressive environment, even if it is unstressed; DSC describes the damage related

to stress corrosion (SC) process, namely the localization of the corrosion attack in the areas of the

material where the stress is more concentrated and a stress dependent evolution of the corrosive

phenomena occurs.

The model is implemented into a FE framework using the commercial code ABAQUS/Explicit

(Dassault Systemes Simulia Corp., RI, USA) by means of a user subroutine (VUSDFLD) devoted to

the calculation of the damage increment through the evolution law affecting the updated stress state

in the explicit time integration scheme. The intrinsic FE discretization procedure is obtained

through the call of the subroutine in each material point of the grid.

The definition of different characteristic dimensions δ for the corrosion processes, allows to

implement into the FE framework a well established strategy to manage and reduce the mesh

dependency effect, intrinsic to the CDM approach, on the numerical results.

The strain localization phenomena, commonly encountered in the FE implementation of non linear

damage laws characterized by a strain-softening behavior, requires a proper reduction of the mesh

sensitivity phenomena, namely by a direct dependency of damage evolution on the characteristic

FE length Le. This strategy is adopted in the evolution law through the ratio Le/δ in order to scale the numerical grid characteristic length over a relevant characteristic dimension of the corrosion

process.

The damage evolution law for the uniform corrosion process is:

where the notation dotted DU represents the time derivative, kU is a parameter related to the

kinetics of the uniform corrosion process and δU is a characteristic dimension of the uniform

corrosion process (e.g. the critical thickness of the corrosion film).

In order to guarantee that finite elements characterized by different Le (i.e. different volume and

associated mass) will degrade consistently with the corrosion phenomenon, an inverse dependency

on the Le is required in the damage evolution law. In particular, the higher the volume element, the

longer time will be required for the complete erosion of its mass.

bioengineering company involved in stent production and might give a contribution in the

improvement of stenting routine success.


Finite element (FE) models accounting for degradation may provide a valuable tool to obtain

information about the performance of biodegradable devices over time.

Micro-scale models of corrosion based on diffusion-controlled processes are present in literature

[3]. Such kind of models, even if consistent with the multiphysics degradation process, are not

easily applicable to an actual medical device due to the difficulty to operate experimental-based

calibration procedures for all model parameters and the high computational costs of tridimensional

geometries. Therefore it is necessary to describe the corrosion behavior at a higher dimensional

scale adopting a phenomenological approach. Continuum damage mechanics (CDM) allows to

model the loss of mechanical strength of a material due to the presence of geometrical

discontinuities through the definition of a scalar field which quantifies the distribution of the

damage [4,5].

In this work two corrosion mechanisms are considered as different degradation processes.

Assuming a linear superposition of the different scalar fields the corrosion damage variable D

responsible for the global degradation is:

D=DU+DSC

where the uniform corrosion damage DU accounts for the mass loss occurring to the material

when exposed to aggressive environment, even if it is unstressed; DSC describes the damage related

to stress corrosion (SC) process, namely the localization of the corrosion attack in the areas of the

material where the stress is more concentrated and a stress dependent evolution of the corrosive

phenomena occurs.

The model is implemented into a FE framework using the commercial code ABAQUS/Explicit

(Dassault Systemes Simulia Corp., RI, USA) by means of a user subroutine (VUSDFLD) devoted to

the calculation of the damage increment through the evolution law affecting the updated stress state

in the explicit time integration scheme. The intrinsic FE discretization procedure is obtained

through the call of the subroutine in each material point of the grid.

The definition of different characteristic dimensions δ for the corrosion processes, allows to

implement into the FE framework a well established strategy to manage and reduce the mesh

dependency effect, intrinsic to the CDM approach, on the numerical results.

The strain localization phenomena, commonly encountered in the FE implementation of non linear

damage laws characterized by a strain-softening behavior, requires a proper reduction of the mesh

sensitivity phenomena, namely by a direct dependency of damage evolution on the characteristic

FE length Le. This strategy is adopted in the evolution law through the ratio Le/δ in order to scale the numerical grid characteristic length over a relevant characteristic dimension of the corrosion

process.

The damage evolution law for the uniform corrosion process is:

where the notation dotted DU represents the time derivative, kU is a parameter related to the

kinetics of the uniform corrosion process and δU is a characteristic dimension of the uniform

corrosion process (e.g. the critical thickness of the corrosion film).

In order to guarantee that finite elements characterized by different Le (i.e. different volume and

associated mass) will degrade consistently with the corrosion phenomenon, an inverse dependency

on the Le is required in the damage evolution law. In particular, the higher the volume element, the

longer time will be required for the complete erosion of its mass.


The damage evolution law assumed for the SC process, used in [6] to model the same

phenomenon on stainless steel, is:

σ∗eq is an equivalent stress governing the stress corrosion mechanism. Different measures for the

equivalent stress can be proposed. Under the assumption that the corrosion rate is higher at tensile

stressed regions, in this model the maximum principal stress is adopted, but other choices could be

the equivalent Von Mises stress or other stress components compositions.

σth corresponds to the equivalent stress value below which the stress corrosion process does not

occur. This stress threshold is highly related to the following concurrent factors: the material

composition, the metallurgical conditions and the corrosive environment; in literature a wide range

for the stress threshold has been reported (from 30% of the yield stress to 90% of the ultimate

tensile stress). Due to the lack of knowledge on the Mg-alloys behavior in physiological conditions,

in this model σth is set to 50% of the yield stress [7].

S and R are constants related to the kinetics of the stress corrosion process and can be a function of the corrosive environment; in [6] the pH dependency is considered, but the context of this

research (i.e. physiologic environment), suggest to adopt a constant pH factor, hence S and R are kept constant. δSC is a characteristic dimension of the stress corrosion process (e.g. the characteristic

length of the local oxidation process zone or a characteristic dimension related to the pitting

process).

Preliminary in vitro weight loss tests were conducted on five commercially available alloys

(AZ31, AZ61, AZ80, ZK60 and ZM21). Cylindrical specimens (height 5 mm, diameter 10 mm for

ZK60 and 14 mm for the other alloys) were previously polished down to 600 grit, rinsed in distilled

water, degreased ultrasonically with anhydrous ethanol for 5 minutes and dried with warm air. The

samples were immersed into a solution simulating plasma (Kokubo-c-SBF at 37°C) which was

changed every 24 hours to prevent saturation and to maintain pH to acceptable values (pH=7.4).

Surface area/solution volume ratio was 0.2 cm-1.

In order to identify the uniform corrosion parameters an axisymmetrical model (4-node

axisymmetric elements with Le /δU =0.4) was used to replicate the immersion test. In these analyses

δU was set to 100 µm consistently with experimental observations on Mg alloys in literature [8].

The SC damage model was applied to a simple testing geometry. A uniaxial tensile test on a flat

sample was simulated applying a displacement boundary condition.

The stresses produced were kept within the elastic limit of the material and thus Mg was

modeled with an isotropic linear elastic behavior. Several simulations were conducted varying R

and S.

A biodegradable device structure is simulated using ABAQUS/Explicit with the aim of showing

the capabilities of the developed corrosion model. The geometry of the first Mg stent currently used

in clinical trials (Biotronik, Berlin, Germany) was reconstructed from literature images assuming

0.15 mm of thickness. A periodic unit of the stent of 4.54 mm in length, corresponding to 4 rings

with links, was used in this study with 84476 8-node linear reduced brick (C3D8R) elements. An

average element characteristic length of 20 µm was obtained.

Uniaxial tensile tests were conducted on the five alloys in order to select the most suitable one. In

particular the limited plastic strain range of Mg alloys is the main pitfall for their use in balloon

expandable stent applications, therefore the alloy with the maximum strain limit was selected.

The effect of a hypothetical surface treatment able to reduce the uniform corrosion rate to a third

of the experimental value was also studied. Four cases were simulated, in the following referred as

case SCUexp: superposition of SC and uniform corrosion as estimated from experimental immersion

The damage evolution law assumed for the SC process, used in [6] to model the same

phenomenon on stainless steel, is:

σ∗eq is an equivalent stress governing the stress corrosion mechanism. Different measures for the

equivalent stress can be proposed. Under the assumption that the corrosion rate is higher at tensile

stressed regions, in this model the maximum principal stress is adopted, but other choices could be

the equivalent Von Mises stress or other stress components compositions.

σth corresponds to the equivalent stress value below which the stress corrosion process does not

occur. This stress threshold is highly related to the following concurrent factors: the material

composition, the metallurgical conditions and the corrosive environment; in literature a wide range

for the stress threshold has been reported (from 30% of the yield stress to 90% of the ultimate

tensile stress). Due to the lack of knowledge on the Mg-alloys behavior in physiological conditions,

in this model σth is set to 50% of the yield stress [7].

S and R are constants related to the kinetics of the stress corrosion process and can be a function of the corrosive environment; in [6] the pH dependency is considered, but the context of this

research (i.e. physiologic environment), suggest to adopt a constant pH factor, hence S and R are kept constant. δSC is a characteristic dimension of the stress corrosion process (e.g. the characteristic

length of the local oxidation process zone or a characteristic dimension related to the pitting

process).

Preliminary in vitro weight loss tests were conducted on five commercially available alloys

(AZ31, AZ61, AZ80, ZK60 and ZM21). Cylindrical specimens (height 5 mm, diameter 10 mm for

ZK60 and 14 mm for the other alloys) were previously polished down to 600 grit, rinsed in distilled

water, degreased ultrasonically with anhydrous ethanol for 5 minutes and dried with warm air. The

samples were immersed into a solution simulating plasma (Kokubo-c-SBF at 37°C) which was

changed every 24 hours to prevent saturation and to maintain pH to acceptable values (pH=7.4).

Surface area/solution volume ratio was 0.2 cm-1.

In order to identify the uniform corrosion parameters an axisymmetrical model (4-node

axisymmetric elements with Le /δU =0.4) was used to replicate the immersion test. In these analyses

δU was set to 100 µm consistently with experimental observations on Mg alloys in literature [8].

The SC damage model was applied to a simple testing geometry. A uniaxial tensile test on a flat

sample was simulated applying a displacement boundary condition.

The stresses produced were kept within the elastic limit of the material and thus Mg was

modeled with an isotropic linear elastic behavior. Several simulations were conducted varying R

and S.

A biodegradable device structure is simulated using ABAQUS/Explicit with the aim of showing

the capabilities of the developed corrosion model. The geometry of the first Mg stent currently used

in clinical trials (Biotronik, Berlin, Germany) was reconstructed from literature images assuming

0.15 mm of thickness. A periodic unit of the stent of 4.54 mm in length, corresponding to 4 rings

with links, was used in this study with 84476 8-node linear reduced brick (C3D8R) elements. An

average element characteristic length of 20 µm was obtained.

Uniaxial tensile tests were conducted on the five alloys in order to select the most suitable one. In

particular the limited plastic strain range of Mg alloys is the main pitfall for their use in balloon

expandable stent applications, therefore the alloy with the maximum strain limit was selected.

The effect of a hypothetical surface treatment able to reduce the uniform corrosion rate to a third

of the experimental value was also studied. Four cases were simulated, in the following referred as

case SCUexp: superposition of SC and uniform corrosion as estimated from experimental immersion


tests; case SC: SC alone; case Ũ: hypothetical reduced uniform corrosion only; case SCŨ : superposition of SC and Ũ.

Results

From the results of these experimental campaign, summarized in Table 1, ZM21 turned out to have

the highest strain at break; moreover, it has comparable mechanical properties to Mg alloy WE43

used for the stent under clinical trial. For these reasons the ZM21 alloy was selected for the

simulations. The material was modeled as a homogeneous, isotropic, elasto-plastic material through

a Von Mises plasticity model with isotropic hardening. Poisson modulus and density were

respectively set to 0.35 and 1.83 g/cm3.

Table 1: Mechanical properties of the available alloys obtained from uniaxial tensile testing.

In Fig. 1 the numerical results are compared with those from the experiments in term of mass loss.

Results showed how the uniform corrosion damage model was able to provide corrosion kinetics

consistent with experimental measurements when changing kU in the range 10-2-10

-1 [h

-1].

Figure 1: Mass loss (ML) versus time (t) curves obtained from immersion tests (dashed lines)

and with the uniform corrosion model (dotted lines). Experimental results and optical images of the

specimens observed after 90 h (on the right).

tests; case SC: SC alone; case Ũ: hypothetical reduced uniform corrosion only; case SCŨ : superposition of SC and Ũ.

Results

From the results of these experimental campaign, summarized in Table 1, ZM21 turned out to have

the highest strain at break; moreover, it has comparable mechanical properties to Mg alloy WE43

used for the stent under clinical trial. For these reasons the ZM21 alloy was selected for the

simulations. The material was modeled as a homogeneous, isotropic, elasto-plastic material through

a Von Mises plasticity model with isotropic hardening. Poisson modulus and density were

respectively set to 0.35 and 1.83 g/cm3.

Table 1: Mechanical properties of the available alloys obtained from uniaxial tensile testing.

In Fig. 1 the numerical results are compared with those from the experiments in term of mass loss.

Results showed how the uniform corrosion damage model was able to provide corrosion kinetics

consistent with experimental measurements when changing kU in the range 10-2-10

-1 [h

-1].

Figure 1: Mass loss (ML) versus time (t) curves obtained from immersion tests (dashed lines)

and with the uniform corrosion model (dotted lines). Experimental results and optical images of the

specimens observed after 90 h (on the right).


Figure 2

Fig. 2 shows the global damage and mass loss per unit area curves obtained for the whole tensile

test sample changing S for three values of R.

For R>1 it is possible to identify a critical value of global damage, around 0.2-0.3, above which

failure occurred almost instantaneously. This critical value for global damage is a characteristic

value dependent on the sample geometry. Moreover the critical global damage value seems

independent on the material-environment interface parameters (R and S) adopted. The same

behavior was observed experimentally and correctly reproduced with a CDM model for stainless

steel in [6]. The results show that, when increasing the R value, the complete damage for the

structure is reached for lower specific mass loss (i.e. a more localized corrosive attack).

About the results of stenting procedure modeling the effect of the different damage contributions

is clarified in Fig. 3: uniform corrosion alone produced an even loss of material from surface

(case Ũ) while SC concentrated where maximum principal stress (top) exceeded σth (case SC) forming pitting that was further amplified when both mechanisms were present (case SCŨ ).

Figure 2

Fig. 2 shows the global damage and mass loss per unit area curves obtained for the whole tensile

test sample changing S for three values of R.

For R>1 it is possible to identify a critical value of global damage, around 0.2-0.3, above which

failure occurred almost instantaneously. This critical value for global damage is a characteristic

value dependent on the sample geometry. Moreover the critical global damage value seems

independent on the material-environment interface parameters (R and S) adopted. The same

behavior was observed experimentally and correctly reproduced with a CDM model for stainless

steel in [6]. The results show that, when increasing the R value, the complete damage for the

structure is reached for lower specific mass loss (i.e. a more localized corrosive attack).

About the results of stenting procedure modeling the effect of the different damage contributions

is clarified in Fig. 3: uniform corrosion alone produced an even loss of material from surface

(case Ũ) while SC concentrated where maximum principal stress (top) exceeded σth (case SC) forming pitting that was further amplified when both mechanisms were present (case SCŨ ).


Figure 3: Maximum principal stress is mapped on a portion of the stent (top). At bottom the

damage after 50 time units is plotted with SC only (case SC), with reduced uniform corrosion alone (case Ũ) and with both mechanisms activated (case SCŨ ).

Conclusions

This study presents a novel approach to describe the corrosion of metallic biomaterials through a

phenomenological model that can be applied to the complex geometry of a medical device such as

an endovascular stent. A key feature of CDM approach, implemented into the FE framework, is that

it is not restricted to only one damage mechanisms but it is possible to account for different

corrosion processes (e.g. fatigue corrosion may play an important role in the resistance of a stent

device subject to pulsatile flow) or other damage mechanisms. For example, observing the

mechanical behavior of Mg-alloys in general, it is reasonable that a limited amount of plastic

damage may occur even at low temperature (i.e. 37°C), in particular at the high strain rate at which

a stent is deployed. Considering that the study of plastic damage was the original application of the

CDM approach, the presented CDM based models for corrosion damage could contribute to explore

the interaction of plastic damage with corrosion damage which is one of the current challenges in

degradable materials modelling. The response of the models adopted for uniform corrosion and SC

damage is phenomenologically consistent with experimental observations of these phenomena. A

preliminary experimental campaign for uniform corrosion measurements for different Mg alloys has

been done in order to perform a quantitative validation process about uniform corrosion. About SC

model a broad sensitivity analysis has been done simulating the SC experimental tests.

Given the complex and irreproducible conditions in the human body the exact lifetime of an

implanted device is clearly not predictable a priori. Nonetheless, the modelling approach described in this work accounting for the combined effects of aggressive environment and mechanical loading,

could be highly useful when comparing different stent designs in terms of durability and scaffolding

ability over time.

Figure 3: Maximum principal stress is mapped on a portion of the stent (top). At bottom the

damage after 50 time units is plotted with SC only (case SC), with reduced uniform corrosion alone (case Ũ) and with both mechanisms activated (case SCŨ ).

Conclusions

This study presents a novel approach to describe the corrosion of metallic biomaterials through a

phenomenological model that can be applied to the complex geometry of a medical device such as

an endovascular stent. A key feature of CDM approach, implemented into the FE framework, is that

it is not restricted to only one damage mechanisms but it is possible to account for different

corrosion processes (e.g. fatigue corrosion may play an important role in the resistance of a stent

device subject to pulsatile flow) or other damage mechanisms. For example, observing the

mechanical behavior of Mg-alloys in general, it is reasonable that a limited amount of plastic

damage may occur even at low temperature (i.e. 37°C), in particular at the high strain rate at which

a stent is deployed. Considering that the study of plastic damage was the original application of the

CDM approach, the presented CDM based models for corrosion damage could contribute to explore

the interaction of plastic damage with corrosion damage which is one of the current challenges in

degradable materials modelling. The response of the models adopted for uniform corrosion and SC

damage is phenomenologically consistent with experimental observations of these phenomena. A

preliminary experimental campaign for uniform corrosion measurements for different Mg alloys has

been done in order to perform a quantitative validation process about uniform corrosion. About SC

model a broad sensitivity analysis has been done simulating the SC experimental tests.

Given the complex and irreproducible conditions in the human body the exact lifetime of an

implanted device is clearly not predictable a priori. Nonetheless, the modelling approach described in this work accounting for the combined effects of aggressive environment and mechanical loading,

could be highly useful when comparing different stent designs in terms of durability and scaffolding

ability over time.


Acknowledgements

The authors gratefully acknowledge the support from IIT (Italian Institute of Technology, Genoa),

Fondazione Cassa di Risparmio di Trento e Rovereto and Politecnico di Milano Progetto 5xmille.

References

[1] Zartner P., Cesnjevar R., Singer H., Weyand M., 2005, Absorbable coronary stents. New

promising technology, Catheterization and Cardiovascular Interventions, 66, 590-594.

[2] Hanawa T., 2009, Materials for metallic stents, Journal of Artificial Organs, 12(2), 73-79.

[3] Scheiner, S., Hellmich, C., 2007. Stable pitting corrosion of stainless steel as diffusion-

controlled dissolution process with a sharp moving electrode boundary. Corrosion Science 49,

319-346.

[4] Garud Y.S. 1991, Quantitative evaluation of environmentally assisted cracking: a survey of

developments and application of modelling concepts, Journal of Pressure Vessel Technology,

113 (1), 1-9.

[5] Bolotin V.V., Shipkov A.A. 2001, Mechanical aspects of corrosion fatigue and stress corrosion

cracking, International Journal of Solids and Structures, 38, 7297-7318.

[6] da Costa-Mattos H.S., I.N. Bastos I.N., Gomes J.A.C.P. 2008, A simple model for slow strain

rate and constant load corrosion tests of austenitic stainless steel in acid aqueous solution

containing sodium chloride, Corrosion Science, 50, 2858-2866.

[7] Winzer, N., Atrens, A., Dietzel, W., Song, G.L., Kainer, KU., 2007. Stress corrosion cracking

in magnesium alloys: Characterization and prevention. Journal of the Minerals, Metals and

Materials Society 59, 49-53.

[8] Levésque, J., Hermawan, H., Dub´e, D., Mantovani, D., 2008. Design of a pseudo-

physiological test bench specific to the development of biodegradable metallic biomaterials.

Acta Biomater 4, 284-295.

Acknowledgements

The authors gratefully acknowledge the support from IIT (Italian Institute of Technology, Genoa),

Fondazione Cassa di Risparmio di Trento e Rovereto and Politecnico di Milano Progetto 5xmille.

References

[1] Zartner P., Cesnjevar R., Singer H., Weyand M., 2005, Absorbable coronary stents. New

promising technology, Catheterization and Cardiovascular Interventions, 66, 590-594.

[2] Hanawa T., 2009, Materials for metallic stents, Journal of Artificial Organs, 12(2), 73-79.

[3] Scheiner, S., Hellmich, C., 2007. Stable pitting corrosion of stainless steel as diffusion-

controlled dissolution process with a sharp moving electrode boundary. Corrosion Science 49,

319-346.

[4] Garud Y.S. 1991, Quantitative evaluation of environmentally assisted cracking: a survey of

developments and application of modelling concepts, Journal of Pressure Vessel Technology,

113 (1), 1-9.

[5] Bolotin V.V., Shipkov A.A. 2001, Mechanical aspects of corrosion fatigue and stress corrosion

cracking, International Journal of Solids and Structures, 38, 7297-7318.

[6] da Costa-Mattos H.S., I.N. Bastos I.N., Gomes J.A.C.P. 2008, A simple model for slow strain

rate and constant load corrosion tests of austenitic stainless steel in acid aqueous solution

containing sodium chloride, Corrosion Science, 50, 2858-2866.

[7] Winzer, N., Atrens, A., Dietzel, W., Song, G.L., Kainer, KU., 2007. Stress corrosion cracking

in magnesium alloys: Characterization and prevention. Journal of the Minerals, Metals and

Materials Society 59, 49-53.

[8] Levésque, J., Hermawan, H., Dub´e, D., Mantovani, D., 2008. Design of a pseudo-

physiological test bench specific to the development of biodegradable metallic biomaterials.

Acta Biomater 4, 284-295.


Prediction of Morphological Properties of Smart-Coatings

for Cr Replacement, Based on Mathematical Modelling

Benedetto Bozzini1, a, Ivonne Sgura2, Deborah Lacitignola3,

Claudio Mele1, Mariapia Marchitto4, Antonio Ciliberto4 1Dipartimento di Ingegneria dell’Innovazione, Università del Salento, via Monteroni, 73100

Lecce - Italy

2Dipartimento di Matematica, Università del Salento, via Arnesano, 73100 Lecce – Italy

3Dipartimento di Scienze Motorie e della Salute, Università di Cassino, Campus Folcara,

Loc. S.Angelo, I-03043 Cassino, Italy

4Alenia Aeronautica S.p.A. Viale dell'Aeronautica, s/n, 80038 Pomigliano d'Arco (Na) - Italy

acorresponding author: [email protected]

Keywords: Mn, electrodeposition, ionic liquids, reaction diffusion, pattern formation, Turing instability, numerical simulation.

Abstract. In this paper we present an extension of a mathematical model for the morphological evolution of metal electrodeposits – recently developed by some of the authors – accounting for mass-transport of electroactive species from the bulk of the bath to the cathode surface. The implementation of mass-transport effects is specially necessary for the quantitative rationalisation of electrodeposition processes from ionic liquids, since these electrolytes exhibit a viscosity that is notably higher than that of cognate aqueous solutions and consequently mass-transport control is active at all practically relevant plating rates. In this work we show that, if mass-transport is coupled to cathodic adsorption of ionic liquid species and surface diffusion of adatoms, it can lead to electrodeposit smoothing. This seemingly paradoxical theoretical result has been validated by a series of Mn electrodeposition experiments from aqueous baths and eutectic ionic liquids. The latter solutions have been shown to be able to form remarkably smoother coatings than the former ones. Mn electroplates have been proposed for Cd replacement and their corrosion protection performance seems comparable, but so far the required surface finish quality has not been achieved with aqueous electrolytes. Ionic liquids thus seem to provide a viable approach to aeronautic-grade Mn electroplating.

Introduction

Within the framework of Green-aircraft issues, a key point is the replacement of Cd coatings. A prospective system of notable interest is the electrodeposition of Mn and Mn-based alloys. Though feasible [1-3], coating processes of this type from aqueous solutions are difficult to manage and - to the best of the current technology - exhibit excessive roughness for practical applications, essentially due to the high rate of concurrent hydrogen evolution. For this reason, the use of room temperature ionic liquids (RTIL) as solvents for electrodeposition is highly appealing [4,5]. In particular, it is expected that the peculiar combination of electrochemistry and mass-transport prevailing in RTIL solutions will favour the growth of smooth Mn-based coatings. Of course, in the field of electrocrystallisation, a combination of experimental and mathematical modelling work is the key to the development of successful solutions: for this reason in our group, we are performing research on Mn electrodeposition from eutectic RTILs as well as mathematical modelling of these processes.

Prediction of Morphological Properties of Smart-Coatings

for Cr Replacement, Based on Mathematical Modelling

Benedetto Bozzini1, a, Ivonne Sgura2, Deborah Lacitignola3,

Claudio Mele1, Mariapia Marchitto4, Antonio Ciliberto4 1Dipartimento di Ingegneria dell’Innovazione, Università del Salento, via Monteroni, 73100

Lecce - Italy

2Dipartimento di Matematica, Università del Salento, via Arnesano, 73100 Lecce – Italy

3Dipartimento di Scienze Motorie e della Salute, Università di Cassino, Campus Folcara,

Loc. S.Angelo, I-03043 Cassino, Italy

4Alenia Aeronautica S.p.A. Viale dell'Aeronautica, s/n, 80038 Pomigliano d'Arco (Na) - Italy

acorresponding author: [email protected]

Keywords: Mn, electrodeposition, ionic liquids, reaction diffusion, pattern formation, Turing instability, numerical simulation.

Abstract. In this paper we present an extension of a mathematical model for the morphological evolution of metal electrodeposits – recently developed by some of the authors – accounting for mass-transport of electroactive species from the bulk of the bath to the cathode surface. The implementation of mass-transport effects is specially necessary for the quantitative rationalisation of electrodeposition processes from ionic liquids, since these electrolytes exhibit a viscosity that is notably higher than that of cognate aqueous solutions and consequently mass-transport control is active at all practically relevant plating rates. In this work we show that, if mass-transport is coupled to cathodic adsorption of ionic liquid species and surface diffusion of adatoms, it can lead to electrodeposit smoothing. This seemingly paradoxical theoretical result has been validated by a series of Mn electrodeposition experiments from aqueous baths and eutectic ionic liquids. The latter solutions have been shown to be able to form remarkably smoother coatings than the former ones. Mn electroplates have been proposed for Cd replacement and their corrosion protection performance seems comparable, but so far the required surface finish quality has not been achieved with aqueous electrolytes. Ionic liquids thus seem to provide a viable approach to aeronautic-grade Mn electroplating.

Introduction

Within the framework of Green-aircraft issues, a key point is the replacement of Cd coatings. A prospective system of notable interest is the electrodeposition of Mn and Mn-based alloys. Though feasible [1-3], coating processes of this type from aqueous solutions are difficult to manage and - to the best of the current technology - exhibit excessive roughness for practical applications, essentially due to the high rate of concurrent hydrogen evolution. For this reason, the use of room temperature ionic liquids (RTIL) as solvents for electrodeposition is highly appealing [4,5]. In particular, it is expected that the peculiar combination of electrochemistry and mass-transport prevailing in RTIL solutions will favour the growth of smooth Mn-based coatings. Of course, in the field of electrocrystallisation, a combination of experimental and mathematical modelling work is the key to the development of successful solutions: for this reason in our group, we are performing research on Mn electrodeposition from eutectic RTILs as well as mathematical modelling of these processes.

This paper is chiefly focused on mathematical modelling, but also reports some original data regarding electrodeposition of Mn from innovative aqueous electrolytes containing chelating agents as well as from eutectic RTILs. The mathematical model is based on a reaction-diffusion system, devised to the model coupling between surface morphology and surface composition, as a means of understanding the formation of morphological patterns found in electrodeposition and leading to the roughening of growing metals. In this particular case, the discussion is restricted to the case of one chemical species adsorbed at the surface of the growing cathode and source terms for both the chemical and the morphological equations of a simple, but electrochemically realistic form. As far as adsorption is concerned, from in situ spectroelectrochemical investigations both in the absence [5] and in the presence of electrodeposition reactions [6, 7], it can be concluded that electrodic adsorption of RTILs seems a general feature of the electrochemical processes carried out in these electrolytes. Furthermore, metal electrodeposition has been proved to profoundly affect the RTIL adsorption behaviour [6]. Recently, we have developed a series of approaches to the analytical and numerical solution of these electrodeposition modelling issues, ultimately based on pure Butler-Volmer kinetics [8-10]. In this paper we describe the extension of our approach to systems, such as RTIL-based baths, in which mass-transport contributions are vital.

As far as Mn-based coatings are concerned, a small group of papers concerning electrodeposition of Mn and Mn alloys from RTILs has appeared in literature, essentially reporting electrochemical (chronopotentiometry, chronoamperometry, voltammetry) [11-17], morphological (SEM, TEM, AFM) [11-16] and compositional studies (XRD, AAS, EDS) [11-15]. Effects of applied potential and bath temperature on material characteristics of Mn films and Mn alloys electrodeposited in the following systems have been studied: Mn in butylmethylpyrrolidinium bis(trifluoromethylsulfony)imide (BMP–TFSI) [11, 12], Mn in RTILs based on dicyanamide (DCA) anion [17], Zn-Mn alloys in tri-1-butylmethylammonium bis((trifluoromethane)sulfonyl)imide (Bu3MeN-TFSI) [16], Zn-Mn in an eutectic RTIL obtained combining urea-choline chloride in a 2:1 molar ratio [13], Al-Mn alloys in 1-ethyl-3-methylimidazolium chloride, [EMI]Cl [14]. BMP-TFSI and DCA-based RTILs has been proved to be more beneficial then Bu3MeN-TFSI for electroplate quality, due to their lower viscosity and higher conductivity that result in more efficient mass-transport and lower voltage drop [12, 17]. SEM micrographs revealed a morphology highly influenced by deposition potential and temperature. At -1.8 V vs Fc/Fc+ and 50° C a fine nano-structure of the Mn deposited in BMP-TFSI was observed, that has been reported to change into a flat morphology at more cathodic potentials. At higher deposition temperatures, the Mn deposits become less uniform, with a fibrous structure [11]. The crystallite size of Mn deposits seems to be reduced by increasing the cathodic electrodeposition potential [12]. The surface morphologies of Zn–Mn alloys strongly depend on the alloy composition. As the Mn contents in the Zn-Mn alloys increases, the grain size of the surface coatings increases and the colour of the Mn coatings changes from a bright and mirror-like silver shine to dark black, but reflective, and consisting mainly of spherical grains [13, 16]. The following types of deposit structures were observed, in order of increasing cathodic overpotential: bright silvery layers, black and smooth adherent films, rougher and powdery deposits [13]. Boric acid has been used as an additive for obtaining more adherent deposits, under otherwise identical plating conditions [13]. If a critical amount of this additive is exceeded, the Mn content increases, though at the expense of deposit quality, since dendrites tend to develop. In the deposition of single-phase fcc Al-Mn alloys, increasing Mn content up to 7.5 % leads to a decrease in grain size; Mn contents higher than ca. 10% give rise to the formation of amorphous deposits [14]. In supercapacitor technology, manganese oxide is a promising alternative to highly performant, but notably expensive ruthenium oxide [11, 12, 18].

This paper is chiefly focused on mathematical modelling, but also reports some original data regarding electrodeposition of Mn from innovative aqueous electrolytes containing chelating agents as well as from eutectic RTILs. The mathematical model is based on a reaction-diffusion system, devised to the model coupling between surface morphology and surface composition, as a means of understanding the formation of morphological patterns found in electrodeposition and leading to the roughening of growing metals. In this particular case, the discussion is restricted to the case of one chemical species adsorbed at the surface of the growing cathode and source terms for both the chemical and the morphological equations of a simple, but electrochemically realistic form. As far as adsorption is concerned, from in situ spectroelectrochemical investigations both in the absence [5] and in the presence of electrodeposition reactions [6, 7], it can be concluded that electrodic adsorption of RTILs seems a general feature of the electrochemical processes carried out in these electrolytes. Furthermore, metal electrodeposition has been proved to profoundly affect the RTIL adsorption behaviour [6]. Recently, we have developed a series of approaches to the analytical and numerical solution of these electrodeposition modelling issues, ultimately based on pure Butler-Volmer kinetics [8-10]. In this paper we describe the extension of our approach to systems, such as RTIL-based baths, in which mass-transport contributions are vital.

As far as Mn-based coatings are concerned, a small group of papers concerning electrodeposition of Mn and Mn alloys from RTILs has appeared in literature, essentially reporting electrochemical (chronopotentiometry, chronoamperometry, voltammetry) [11-17], morphological (SEM, TEM, AFM) [11-16] and compositional studies (XRD, AAS, EDS) [11-15]. Effects of applied potential and bath temperature on material characteristics of Mn films and Mn alloys electrodeposited in the following systems have been studied: Mn in butylmethylpyrrolidinium bis(trifluoromethylsulfony)imide (BMP–TFSI) [11, 12], Mn in RTILs based on dicyanamide (DCA) anion [17], Zn-Mn alloys in tri-1-butylmethylammonium bis((trifluoromethane)sulfonyl)imide (Bu3MeN-TFSI) [16], Zn-Mn in an eutectic RTIL obtained combining urea-choline chloride in a 2:1 molar ratio [13], Al-Mn alloys in 1-ethyl-3-methylimidazolium chloride, [EMI]Cl [14]. BMP-TFSI and DCA-based RTILs has been proved to be more beneficial then Bu3MeN-TFSI for electroplate quality, due to their lower viscosity and higher conductivity that result in more efficient mass-transport and lower voltage drop [12, 17]. SEM micrographs revealed a morphology highly influenced by deposition potential and temperature. At -1.8 V vs Fc/Fc+ and 50° C a fine nano-structure of the Mn deposited in BMP-TFSI was observed, that has been reported to change into a flat morphology at more cathodic potentials. At higher deposition temperatures, the Mn deposits become less uniform, with a fibrous structure [11]. The crystallite size of Mn deposits seems to be reduced by increasing the cathodic electrodeposition potential [12]. The surface morphologies of Zn–Mn alloys strongly depend on the alloy composition. As the Mn contents in the Zn-Mn alloys increases, the grain size of the surface coatings increases and the colour of the Mn coatings changes from a bright and mirror-like silver shine to dark black, but reflective, and consisting mainly of spherical grains [13, 16]. The following types of deposit structures were observed, in order of increasing cathodic overpotential: bright silvery layers, black and smooth adherent films, rougher and powdery deposits [13]. Boric acid has been used as an additive for obtaining more adherent deposits, under otherwise identical plating conditions [13]. If a critical amount of this additive is exceeded, the Mn content increases, though at the expense of deposit quality, since dendrites tend to develop. In the deposition of single-phase fcc Al-Mn alloys, increasing Mn content up to 7.5 % leads to a decrease in grain size; Mn contents higher than ca. 10% give rise to the formation of amorphous deposits [14]. In supercapacitor technology, manganese oxide is a promising alternative to highly performant, but notably expensive ruthenium oxide [11, 12, 18].


Experimental

The eutectic RTIL electrolyte was obtained by mixing urea with choline chloride both in powder forms (ChCl) in a 2:1 molar ratio and heating to a temperature of 80 °C under continuous magnetic

stirring, until a clear, colourless liquid formed. Subsequently, 0.5 M MnCl2⋅4H2O was added to the liquid and stirring was continued for 30 min to reach a homogeneous solution. For comparison, we also tested the following chelate-based aqueous baths, optimised for deposit adherence and compactness. The inorganic components of the baths were: (NH4)2SO4 1M, MnSO4 0.59M, pH 2.5 (adjusted with H2SO4). This solution was then made 20 mM in each of the following chelating compounds: (i) Titriplex-I: nitrilotriacetic acid (#3 –COOH groups); (ii) EDTA: ethylenediaminetetra-acetic acid: (#4 –COOH groups); (iii) DTPA: diethylenetriamine penta-acetate (#5 –COOH groups). Mn coatings were deposited onto Fe360 carbon steel, polished with grit papers and diamond paste doen to mirror finish, degreased in 1M NaOH in a ultrasound bath, rinsed with ultrapure water and dried in an N2 flow. The active electrodic area was 3 cm2, defined by a PTFE window. The counter electrodes were: for the RTIL-based solution an Au wire loop of ca. 4 cm2 area; for the aqueous bath a Pb slab of 10 cm2 area. The RTIL bath was operated in a prismatic cell with undivided electrolyte. With the aqueous bath, electrodeposition was performed in a two-compartment cell, with the anolyte separed from the catholyte by a glass frit and a 15 cm long tube, in order to minimise the precipitation of MnO2 at the anode. The catholyte was stirred with a magnetic bar and its pH was continuosly monitored and controlled to ±0.2 units by dropwise additions of 10 vol% H2SO4. The anolyte had the same composition as the bath, apart from the Mn(II) salt.

Galvanostatic electrodeposition of films of thicknesses in the range ca. 10÷15 µm was carried out at room temperature with an AMEL 5000 system. With the RTIL bath, electrodeposition could be run

at low current densities (-2÷-20 mA cm-2) thanks to the absence of concurrent electrolyte decomposition reactions. In the aqueous systems, high current densities are instead required in order to achieve a catholyte chemistry adequate for the electrodeposition of Mn (see, e.g. [19]): the optimal conditions were empirically found to be -400 mA cm-2. The cathodic efficiencies were

determined gravimetrically and found to be ca. 44÷56% and ca. 93% in the aqueous and RTIL electrolytes, respectively. The morphology of the electrodeposits was studied by SEM (Cambridge Stereoscan 360) and the phase structure of the coatings by powder XRD (Philips PW 1830 diffractometer, Philips PW 1820 goniometer, unmonochromated Cu Kα, scan rate 1 deg s-1).

Experimental

The eutectic RTIL electrolyte was obtained by mixing urea with choline chloride both in powder forms (ChCl) in a 2:1 molar ratio and heating to a temperature of 80 °C under continuous magnetic

stirring, until a clear, colourless liquid formed. Subsequently, 0.5 M MnCl2⋅4H2O was added to the liquid and stirring was continued for 30 min to reach a homogeneous solution. For comparison, we also tested the following chelate-based aqueous baths, optimised for deposit adherence and compactness. The inorganic components of the baths were: (NH4)2SO4 1M, MnSO4 0.59M, pH 2.5 (adjusted with H2SO4). This solution was then made 20 mM in each of the following chelating compounds: (i) Titriplex-I: nitrilotriacetic acid (#3 –COOH groups); (ii) EDTA: ethylenediaminetetra-acetic acid: (#4 –COOH groups); (iii) DTPA: diethylenetriamine penta-acetate (#5 –COOH groups). Mn coatings were deposited onto Fe360 carbon steel, polished with grit papers and diamond paste doen to mirror finish, degreased in 1M NaOH in a ultrasound bath, rinsed with ultrapure water and dried in an N2 flow. The active electrodic area was 3 cm2, defined by a PTFE window. The counter electrodes were: for the RTIL-based solution an Au wire loop of ca. 4 cm2 area; for the aqueous bath a Pb slab of 10 cm2 area. The RTIL bath was operated in a prismatic cell with undivided electrolyte. With the aqueous bath, electrodeposition was performed in a two-compartment cell, with the anolyte separed from the catholyte by a glass frit and a 15 cm long tube, in order to minimise the precipitation of MnO2 at the anode. The catholyte was stirred with a magnetic bar and its pH was continuosly monitored and controlled to ±0.2 units by dropwise additions of 10 vol% H2SO4. The anolyte had the same composition as the bath, apart from the Mn(II) salt.

Galvanostatic electrodeposition of films of thicknesses in the range ca. 10÷15 µm was carried out at room temperature with an AMEL 5000 system. With the RTIL bath, electrodeposition could be run

at low current densities (-2÷-20 mA cm-2) thanks to the absence of concurrent electrolyte decomposition reactions. In the aqueous systems, high current densities are instead required in order to achieve a catholyte chemistry adequate for the electrodeposition of Mn (see, e.g. [19]): the optimal conditions were empirically found to be -400 mA cm-2. The cathodic efficiencies were

determined gravimetrically and found to be ca. 44÷56% and ca. 93% in the aqueous and RTIL electrolytes, respectively. The morphology of the electrodeposits was studied by SEM (Cambridge Stereoscan 360) and the phase structure of the coatings by powder XRD (Philips PW 1830 diffractometer, Philips PW 1820 goniometer, unmonochromated Cu Kα, scan rate 1 deg s-1).


Morphology and Structure of Electrodeposited Mn based Coatings

Fig. 1 – SEM micrographs of Mn coatings electrodeposited galvanostatically at -400 mA cm-2 from aqueous

solutions containing 20 mM: (A) Titriplex-I (thickness 13.5±4.7 µm); (B) EDTA (thickness 11.2±0.8 µm); (C)

DTPA (thickness 14.3±2.8 µm).

Fig. 2 – X-ray diffractograms of Mn coatings electrodeposited galvanostatically at -400 mA cm-2 from aqueous solutions containing 20 mM: (A) Titriplex-I; (B) EDTA; (C) DTPA and from the eutectic RTIL electrolyte at: (D) - mA cm-2; (E) -5 mA cm-2; (F) -20 mA cm-2.

In Fig. 1 we report SEM micrographs of galvanostatic deposits, obtained from the three investigated aqueous solutions at -400 mA cm-2. The phase structure – as revealed by XRD (Fig. 2) - of the

deposits is γ-Mn with a certain degree of (002) preferred orientation (powder diffractograms yield 100% and 14% relative intensities for the (101) and (002) reflections, respectively) and tiny

amounts of β-Mn, that has been found in the literature to be correlated with the precipitation of

Morphology and Structure of Electrodeposited Mn based Coatings

Fig. 1 – SEM micrographs of Mn coatings electrodeposited galvanostatically at -400 mA cm-2 from aqueous

solutions containing 20 mM: (A) Titriplex-I (thickness 13.5±4.7 µm); (B) EDTA (thickness 11.2±0.8 µm); (C)

DTPA (thickness 14.3±2.8 µm).

Fig. 2 – X-ray diffractograms of Mn coatings electrodeposited galvanostatically at -400 mA cm-2 from aqueous solutions containing 20 mM: (A) Titriplex-I; (B) EDTA; (C) DTPA and from the eutectic RTIL electrolyte at: (D) - mA cm-2; (E) -5 mA cm-2; (F) -20 mA cm-2.

In Fig. 1 we report SEM micrographs of galvanostatic deposits, obtained from the three investigated aqueous solutions at -400 mA cm-2. The phase structure – as revealed by XRD (Fig. 2) - of the

deposits is γ-Mn with a certain degree of (002) preferred orientation (powder diffractograms yield 100% and 14% relative intensities for the (101) and (002) reflections, respectively) and tiny

amounts of β-Mn, that has been found in the literature to be correlated with the precipitation of


Mn(OH)2 [20, 21]. The high-temperature γ-Mn allotropic form, often accompanied by the

equilibrium α modification – not detected in our electrodeposits – is the typical phase of electrodeposits obtained at high cathodic polarisations, in the presence of adsorbable additives such as SCN¯ [20, 22]. The use of chelating agents in aqueous solutions allows the formation of compact and adherent deposit, nevertheless approaches to a radical improvement of roughness do not seem to be available at the moment of this writing. The morphological quality of the deposits seems to be correlated with the number of chelating moieties present in the additive: globular deposits with hydrogen-related holes are found with Titriplex-I, closer-packed globuli with EDTA and separate, tinier, pyramidal crystallites tend to form in the presence of DTPA.

Fig. 3 – SEM micrographs of Mn coatings electrodeposited galvanostatically from the eutectic RTIL electrolyte at

different current densities: (A) -2 mA cm-2 (low magnification, thickness 12.9±1.4 µm); (B) -5 mA cm-2 (low

magnification, thickness 11.8±0.9 µm); (C) -10 mA cm-2 (low magnification, thickness 13.6±1.1 µm); (D) -2 mA cm-2 (high magnification); (E) -5 mA cm-2 (high magnification); (F) -10 mA cm-2 (high magnification); (G) -

15 mA cm-2 (low magnification thickness 10.7±0.8 µm); (H) 15 mA cm-2 (high magnification); (I) -20 mA cm-2 (low

magnification thickness 14.2±2.3 µm).

In general, the deposits obtained from the eutectic RTIL exhibit a notably smoother finish with respect to those grown from aqueous solutions. A selection of representative SEM micrographs is shown in Fig. 3. The Mn coatings grown at current densities of -10 mA cm-2 exhibit a remarkably regular, nanocrystalline texture. The micrographs reported in panels (A)-(C), (G) and (I) of Fig. 3 were recorded at essentially the same magnifications as the images of Fig. 1. At higher current

Mn(OH)2 [20, 21]. The high-temperature γ-Mn allotropic form, often accompanied by the

equilibrium α modification – not detected in our electrodeposits – is the typical phase of electrodeposits obtained at high cathodic polarisations, in the presence of adsorbable additives such as SCN¯ [20, 22]. The use of chelating agents in aqueous solutions allows the formation of compact and adherent deposit, nevertheless approaches to a radical improvement of roughness do not seem to be available at the moment of this writing. The morphological quality of the deposits seems to be correlated with the number of chelating moieties present in the additive: globular deposits with hydrogen-related holes are found with Titriplex-I, closer-packed globuli with EDTA and separate, tinier, pyramidal crystallites tend to form in the presence of DTPA.

Fig. 3 – SEM micrographs of Mn coatings electrodeposited galvanostatically from the eutectic RTIL electrolyte at

different current densities: (A) -2 mA cm-2 (low magnification, thickness 12.9±1.4 µm); (B) -5 mA cm-2 (low

magnification, thickness 11.8±0.9 µm); (C) -10 mA cm-2 (low magnification, thickness 13.6±1.1 µm); (D) -2 mA cm-2 (high magnification); (E) -5 mA cm-2 (high magnification); (F) -10 mA cm-2 (high magnification); (G) -

15 mA cm-2 (low magnification thickness 10.7±0.8 µm); (H) 15 mA cm-2 (high magnification); (I) -20 mA cm-2 (low

magnification thickness 14.2±2.3 µm).

In general, the deposits obtained from the eutectic RTIL exhibit a notably smoother finish with respect to those grown from aqueous solutions. A selection of representative SEM micrographs is shown in Fig. 3. The Mn coatings grown at current densities of -10 mA cm-2 exhibit a remarkably regular, nanocrystalline texture. The micrographs reported in panels (A)-(C), (G) and (I) of Fig. 3 were recorded at essentially the same magnifications as the images of Fig. 1. At higher current


densities the nanocrystals tend to aggregate first into ridges and successively give rise to flat islands,

exhibiting typical dimensions of some tens of µm. Also in the case of Mn films electrodeposited

from the RTIL, the crystalline structure is γ-Mn, with a considerable (002) preferred orientation. It is worth noting that – to the best of the authors’ knowledge – the only structural description of Mn electrodeposits from RTILs (in this case: butyl methyl pyrrolidinium bis((tri fluoromethyl)sulfonyl) imide) reports the formation of amorphous Mn onto W cathodes, that - as expected - crystallises into the equilibrium a form after heat treatment at 450°C for 24 h [12].

The model

In this paper we consider a modification of the 2D reaction-diffusion system studied in [8], coupling one equation for the morphological dynamics with one for the surface chemical dynamics. With the same notations and assumptions used in [8] for the meaning of the source terms, the system in adimensional form is given by:

[ ]

−+∆=∂∂

−

++∆=

∂∂

,),(),(

,1

21

2

θθηθησθθ

ηθη

εηρη

η

KKdt

t (1)

where η(x,y.,t) is the dimensionless electrode shape and θ(x,y,t) is the surface coverage with the

adsorbed chemical species defined for (x,y,t) ∈ Ω × [0, T], with Ω= [0, L1] × [0, L2], L1, L2 characteristic lengths of the electrode and T a characteristic time of the electrodeposition process. The system (1) is equipped with: zero-flux boundary conditions, that is the normal derivatives along

the electrode boundary are ,0,0 =∂∂

=∂∂

nn

θη and with initial conditions for t=0 given by

.),(),,()0,,(),,()0,,( 00 Ω∈== yxyxyxyxyx θθηη (2)

In both equations 2

2

2

2

yx ∂

∂+

∂

∂=∆ is the Laplace operator and the adimensional parameters are:

.,1

,,**

*

*

*

ss

sc

s

sDD

Dd

DDt

βρσ

βα

ετ ===== (3)

τ is the dimensionless time and *sD is the surface diffusion coefficient of adatoms, α >0, β >0

account for localisation of the ECD process and the presence of an adsorbate at the growing cathode in the first equation for the morphology.

In the second equation for the chemistry, *scD is the surface diffusion coefficient of RTIL and d is

the ratio between the diffusion coefficients of species. The surface coverage dynamics is described, as customary in chemical kinetics, in terms of a material balance with a source term containing positive and negative contribution related to adsorption and desorption. In fact, we have

),,(),(),(),,(),( **2

*1 θηθηθηθηθη DESADSADS KKKKK +== (4)

where *

ADSK and *

DESK represent the adsorption and desorption rate constants, respectively.

As in [8], here we assume that the exchange current densities and Tafel slopes of metal discharge

and electrosorption of ligands are coupled quantum charge-transfer effects [23].

densities the nanocrystals tend to aggregate first into ridges and successively give rise to flat islands,

exhibiting typical dimensions of some tens of µm. Also in the case of Mn films electrodeposited

from the RTIL, the crystalline structure is γ-Mn, with a considerable (002) preferred orientation. It is worth noting that – to the best of the authors’ knowledge – the only structural description of Mn electrodeposits from RTILs (in this case: butyl methyl pyrrolidinium bis((tri fluoromethyl)sulfonyl) imide) reports the formation of amorphous Mn onto W cathodes, that - as expected - crystallises into the equilibrium a form after heat treatment at 450°C for 24 h [12].

The model

In this paper we consider a modification of the 2D reaction-diffusion system studied in [8], coupling one equation for the morphological dynamics with one for the surface chemical dynamics. With the same notations and assumptions used in [8] for the meaning of the source terms, the system in adimensional form is given by:

[ ]

−+∆=∂∂

−

++∆=

∂∂

,),(),(

,1

21

2

θθηθησθθ

ηθη

εηρη

η

KKdt

t (1)

where η(x,y.,t) is the dimensionless electrode shape and θ(x,y,t) is the surface coverage with the

adsorbed chemical species defined for (x,y,t) ∈ Ω × [0, T], with Ω= [0, L1] × [0, L2], L1, L2 characteristic lengths of the electrode and T a characteristic time of the electrodeposition process. The system (1) is equipped with: zero-flux boundary conditions, that is the normal derivatives along

the electrode boundary are ,0,0 =∂∂

=∂∂

nn

θη and with initial conditions for t=0 given by

.),(),,()0,,(),,()0,,( 00 Ω∈== yxyxyxyxyx θθηη (2)

In both equations 2

2

2

2

yx ∂

∂+

∂

∂=∆ is the Laplace operator and the adimensional parameters are:

.,1

,,**

*

*

*

ss

sc

s

sDD

Dd

DDt

βρσ

βα

ετ ===== (3)

τ is the dimensionless time and *sD is the surface diffusion coefficient of adatoms, α >0, β >0

account for localisation of the ECD process and the presence of an adsorbate at the growing cathode in the first equation for the morphology.

In the second equation for the chemistry, *scD is the surface diffusion coefficient of RTIL and d is

the ratio between the diffusion coefficients of species. The surface coverage dynamics is described, as customary in chemical kinetics, in terms of a material balance with a source term containing positive and negative contribution related to adsorption and desorption. In fact, we have

),,(),(),(),,(),( **2

*1 θηθηθηθηθη DESADSADS KKKKK +== (4)

where *

ADSK and *

DESK represent the adsorption and desorption rate constants, respectively.

As in [8], here we assume that the exchange current densities and Tafel slopes of metal discharge

and electrosorption of ligands are coupled quantum charge-transfer effects [23].


In this study we introduce the dependence on mass-transport via the limiting current density

approach (for electrochemical details, see, e.g. [24]). Assuming that mass-transport is stationary and

that the concentration gradient at the electrode surface can be linearised, the limiting current density

iL is customarily written as: bL

Ci zF

δ= ⋅ , where Cb is the bulk concentration of the electroactive

species and δ is the thickness of the concentration boundary layer, the other symbols have the usual

meaning. In this work we shall define 1

Liν = .

These hypotheses allow a simplification in the parameter space such that we can have the following forms for the above rates:

)]exp(2

1[),exp()exp(1

)exp(121 εαη

εθ

ηνη

+=+

= KKba

aAK (5)

where α =ln(1/2) < 0 and ν ≥ 0 .

It is worth noting that for ν=0, the system reduces to the one studied in [8], for which linear stability analysis is provided and presence of spatial patterns induced by diffusion is shown to occur in a suitable region of the parameter space. In particular, in [8] the role of the diffusion parameter d and of the Tafel slope a on stability and pattern selection has been highlighted. In fact, by varying the parameter a, two different types of patterns, in fact encounterd in electroplating experiments, have been obtained by numerical simulations as stationary Turing patterns, that have been denominated spots and spots and worms. In this context, our aim is to investigate - from the mathematical and numerical points of view - the

role of the new parameter ν ≥ 0, in order to understand the effects of mass-transport, related to the well-known high viscosity of RTILs in terms of stabilising effects on a spotty pattern configuration. In the next sections, we follow the same mathematical steps performed in [8], that is the linear and the Turing stability analyses of the equilibria of system (1) with choice (5) for the adsorption and desorption laws.

Stability Analysis

In this section we analyze the stability properties of the non-trivial homogeneous equilibrium of

model (5) both in the space-independent case and in the presence of diffusion (d ≠ 0). The homogeneous steady-states are the solutions of the following set of equations

=−

=

−

+.0),(),(

,01

21 eeeee

e

e

e

e

KK θθηθη

θη

εηη (6)

It is easy to see, that the same spatially homogeneous equilibria found in [8] are obtained also in the present case, that is:

+

==

+

=ε

εε

θηε

ε1

,1

),(,2

,0 21 eeEE (7)

In this study we introduce the dependence on mass-transport via the limiting current density

approach (for electrochemical details, see, e.g. [24]). Assuming that mass-transport is stationary and

that the concentration gradient at the electrode surface can be linearised, the limiting current density

iL is customarily written as: bL

Ci zF

δ= ⋅ , where Cb is the bulk concentration of the electroactive

species and δ is the thickness of the concentration boundary layer, the other symbols have the usual

meaning. In this work we shall define 1

Liν = .

These hypotheses allow a simplification in the parameter space such that we can have the following forms for the above rates:

)]exp(2

1[),exp()exp(1

)exp(121 εαη

εθ

ηνη

+=+

= KKba

aAK (5)

where α =ln(1/2) < 0 and ν ≥ 0 .

It is worth noting that for ν=0, the system reduces to the one studied in [8], for which linear stability analysis is provided and presence of spatial patterns induced by diffusion is shown to occur in a suitable region of the parameter space. In particular, in [8] the role of the diffusion parameter d and of the Tafel slope a on stability and pattern selection has been highlighted. In fact, by varying the parameter a, two different types of patterns, in fact encounterd in electroplating experiments, have been obtained by numerical simulations as stationary Turing patterns, that have been denominated spots and spots and worms. In this context, our aim is to investigate - from the mathematical and numerical points of view - the

role of the new parameter ν ≥ 0, in order to understand the effects of mass-transport, related to the well-known high viscosity of RTILs in terms of stabilising effects on a spotty pattern configuration. In the next sections, we follow the same mathematical steps performed in [8], that is the linear and the Turing stability analyses of the equilibria of system (1) with choice (5) for the adsorption and desorption laws.

Stability Analysis

In this section we analyze the stability properties of the non-trivial homogeneous equilibrium of

model (5) both in the space-independent case and in the presence of diffusion (d ≠ 0). The homogeneous steady-states are the solutions of the following set of equations

=−

=

−

+.0),(),(

,01

21 eeeee

e

e

e

e

KK θθηθη

θη

εηη (6)

It is easy to see, that the same spatially homogeneous equilibria found in [8] are obtained also in the present case, that is:

+

==

+

=ε

εε

θηε

ε1

,1

),(,2

,0 21 eeEE (7)


To investigate the stability properties of these steady states, we have to consider the Jacobian matrix of the system (1) , that is

( )

−−−

−

−

+

+=

)],([)(

1

)2(2

),(

22121

2

eeee

ee

e

ee

ee

KKKKK

J

θηθσθσ

ρηθη

ηηρ

θη

θθηη

(8)

where

),(),(),,(),,( 21

22

11

11 eeeeeeee

KK

KK

KK

KK θη

θθη

ηθη

θθη

η θηθη ∂

∂=

∂

∂=

∂

∂=

∂

∂=

Hence, for the trivial equilibrium E1, we have

+

++

−

+

++

+−

=

2exp

2

12exp

212ln2

02)( 1

εε

εε

νσ

εε

εε

νσ

ερε

bAbAEJ (9)

For the non trivial equilibrium E2, the Jacobian matrix is given by

( )

+−

+

−+=

εεα

σε

εασ

ερ

ερε

)1)(exp(

1

2ln)exp(1)(

0000

2

2

2AzAz

EJ (10)

where

)1(

2

0 +

++=

εεεε

αbaa

and )/exp(1

10 εν a

z+

= . (11)

Linear stability analysis assures that (ηe, θe) is linearly stable if and only if

0)),(det( >eeJ θη (12)

.0)),(( <eeJtr θη (13)

Moreover, we search conditions assuring that an equilibrium (ηe, θe) can undergo Turing instability. We recall that a reaction-diffusion system exhibits Turing or diffusion-driven instability, if a homogeneous steady state is stable to small perturbations in the absence of diffusion, but it is unstable to small spatial perturbations when diffusion is present [9]. By using standard linear theory

(see e.g. [9]), it can be shown that (ηe, θe) can undergo Turing instability if the relationships in (12) and (13) are satisfied together with

02211 >+ ee JdJ (14)

)),,(det(4

)( 22211

ee

ee

Jd

JdJθη>

+ (15)

where Jije are the entries of the Jacobian matrix in Eq. 9 or Eq. 10.

As far as the trivial equilibrium E1 is concerned, the two eigenvalues of the Jacobian matrix J(E1) are both real and negative, i.e. E1 is always a stable node and, even when diffusion is present

To investigate the stability properties of these steady states, we have to consider the Jacobian matrix of the system (1) , that is

( )

−−−

−

−

+

+=

)],([)(

1

)2(2

),(

22121

2

eeee

ee

e

ee

ee

KKKKK

J

θηθσθσ

ρηθη

ηηρ

θη

θθηη

(8)

where

),(),(),,(),,( 21

22

11

11 eeeeeeee

KK

KK

KK

KK θη

θθη

ηθη

θθη

η θηθη ∂

∂=

∂

∂=

∂

∂=

∂

∂=

Hence, for the trivial equilibrium E1, we have

+

++

−

+

++

+−

=

2exp

2

12exp

212ln2

02)( 1

εε

εε

νσ

εε

εε

νσ

ερε

bAbAEJ (9)

For the non trivial equilibrium E2, the Jacobian matrix is given by

( )

+−

+

−+=

εεα

σε

εασ

ερ

ερε

)1)(exp(

1

2ln)exp(1)(

0000

2

2

2AzAz

EJ (10)

where

)1(

2

0 +

++=

εεεε

αbaa

and )/exp(1

10 εν a

z+

= . (11)

Linear stability analysis assures that (ηe, θe) is linearly stable if and only if

0)),(det( >eeJ θη (12)

.0)),(( <eeJtr θη (13)

Moreover, we search conditions assuring that an equilibrium (ηe, θe) can undergo Turing instability. We recall that a reaction-diffusion system exhibits Turing or diffusion-driven instability, if a homogeneous steady state is stable to small perturbations in the absence of diffusion, but it is unstable to small spatial perturbations when diffusion is present [9]. By using standard linear theory

(see e.g. [9]), it can be shown that (ηe, θe) can undergo Turing instability if the relationships in (12) and (13) are satisfied together with

02211 >+ ee JdJ (14)

)),,(det(4

)( 22211

ee

ee

Jd

JdJθη>

+ (15)

where Jije are the entries of the Jacobian matrix in Eq. 9 or Eq. 10.

As far as the trivial equilibrium E1 is concerned, the two eigenvalues of the Jacobian matrix J(E1) are both real and negative, i.e. E1 is always a stable node and, even when diffusion is present


(d ≠ 0), the necessary conditions in Eq. 14 for Turing instability are never satisfied, being d J11(E1)+

J22(E1) <0 for all values of d>0 and ν ≥0. On the other hand, we show that the non trivial equilibrium E2 , which is more interesting from the physical point of view, can undergo diffusion driven instability. The above set of general conditions in Eqs. 12-15 for diffusion-driven instability of the equilibrium E2 can thus be specialised as

>+

−−

+

+−

>+−

>−

<+−

01

)2)(lnexp()(4

)1(

)1)(exp()(

0)1)(exp()(

0)2)(lnexp()(

0)1)(exp()(

00

2

2

300

3

300

3

00

300

3

εεανρσ

εεεανσρε

εανσρεεανρσ

εανσρε

AzdAzd

Azd

Az

Az

(16)

Where )(0 νz given in Eq. 11 introduces the dependence on the new parameter.

Since these inequalities involve the model parameters, they allow us to locate a region in the parameter space such that E2 is stable to small perturbations in the absence of diffusion, but it can be unstable to small spatial perturbations when the diffusion parameter d is non-zero and greater than a critical value d*.

In the case ν=0, an extensive analysis has been performed in [8], here we wish to focus on the role

of the mass-transport kinetic control, that is for ν>0. In particular, we study the behavior of E2 in

dependence of the parameters (d, ν) when – in order to fix ideas and to continue the cases without and with mass-transport - all the other parameters are fixed to the values set in [8] to generate

spatial patterns in absence of mass-transport control (ν=0).

Hence, let us fix ρ = 40, σ = 2, ε = 0.5, b = 1, A = 1. Moreover, let be a = 0.9 to consider the case of a spotty pattern. With this choice, we have E2 = (2, 1/3) and the conditions in Eq. 16 become

>

>

>+

<

)(0 6.0496+1

1/2)-(ln(2) d 1801.2-

6.0496+1

56.990-5d

81

64

)(0 6.0496+1

56.990-5d

)(0 6.04961

1/2)-(ln(2) 1350.9

)(0 6.0496+1

56.990-5

4

2

3

2

1

C

C

C

C

νν

ν

ν

ν

It is easy to see that:

(C1) is satisfied for ν < ν*= 1.7188;

(C2) is always satisfied for all ν ≥0;

(C3) identifies one curve d3(ν)=) 6.0496+5(1

56.990

ν in the plane (d,ν), such that for all values of these

parameters above this curve, the non trivial equilibrium E2 undergoes Turing instability;

(C4) identifies two curves d+(ν) = ν 0.75621+ 0.125

0.46107

and d-(ν)=ν 0.75621+ 0.125

0.44026 such that for all

values above the curve d+(ν) and below the curve d-(ν) Turing instability arises.

(d ≠ 0), the necessary conditions in Eq. 14 for Turing instability are never satisfied, being d J11(E1)+

J22(E1) <0 for all values of d>0 and ν ≥0. On the other hand, we show that the non trivial equilibrium E2 , which is more interesting from the physical point of view, can undergo diffusion driven instability. The above set of general conditions in Eqs. 12-15 for diffusion-driven instability of the equilibrium E2 can thus be specialised as

>+

−−

+

+−

>+−

>−

<+−

01

)2)(lnexp()(4

)1(

)1)(exp()(

0)1)(exp()(

0)2)(lnexp()(

0)1)(exp()(

00

2

2

300

3

300

3

00

300

3

εεανρσ

εεεανσρε

εανσρεεανρσ

εανσρε

AzdAzd

Azd

Az

Az

(16)

Where )(0 νz given in Eq. 11 introduces the dependence on the new parameter.

Since these inequalities involve the model parameters, they allow us to locate a region in the parameter space such that E2 is stable to small perturbations in the absence of diffusion, but it can be unstable to small spatial perturbations when the diffusion parameter d is non-zero and greater than a critical value d*.

In the case ν=0, an extensive analysis has been performed in [8], here we wish to focus on the role

of the mass-transport kinetic control, that is for ν>0. In particular, we study the behavior of E2 in

dependence of the parameters (d, ν) when – in order to fix ideas and to continue the cases without and with mass-transport - all the other parameters are fixed to the values set in [8] to generate

spatial patterns in absence of mass-transport control (ν=0).

Hence, let us fix ρ = 40, σ = 2, ε = 0.5, b = 1, A = 1. Moreover, let be a = 0.9 to consider the case of a spotty pattern. With this choice, we have E2 = (2, 1/3) and the conditions in Eq. 16 become

>

>

>+

<

)(0 6.0496+1

1/2)-(ln(2) d 1801.2-

6.0496+1

56.990-5d

81

64

)(0 6.0496+1

56.990-5d

)(0 6.04961

1/2)-(ln(2) 1350.9

)(0 6.0496+1

56.990-5

4

2

3

2

1

C

C

C

C

νν

ν

ν

ν

It is easy to see that:

(C1) is satisfied for ν < ν*= 1.7188;

(C2) is always satisfied for all ν ≥0;

(C3) identifies one curve d3(ν)=) 6.0496+5(1

56.990

ν in the plane (d,ν), such that for all values of these

parameters above this curve, the non trivial equilibrium E2 undergoes Turing instability;

(C4) identifies two curves d+(ν) = ν 0.75621+ 0.125

0.46107

and d-(ν)=ν 0.75621+ 0.125

0.44026 such that for all

values above the curve d+(ν) and below the curve d-(ν) Turing instability arises.


These results are summarized in Fig. 4.

Fig. 4 – Dynamic scenario in the (d,ν) parameter space.

We found that there exists an interval I =[0, ν*] where, in absence of diffusion, the homogeneous

equilibrium E2 is stable and such that for ν > ν* it becomes unstable. This is shown by the blue line

describing the condition (C1), that is the trace tr(ν) of the Jacobian matrix J(E2).

In presence of diffusion, for all ν ∈ I and for d=d(ν) ≤ d+(ν), (that is below the red continuous

curve) , E2 continues to be a stable equilibrium, while for d=d(ν) > d+(ν) it undergoes diffusion driven instability.

For example, in Fig. 4, we highlight the cases (i) d=40 > d+(0) :=d*=36.886 and (ii) d=10. In

case (i), for ν=0 we have a situation similar to that studied in [8] and a spotty pattern configuration

arises, while for 0 < ν ≤ ν*, that is ν ∈ I (red dashed line), the Turing instability condition holds and the corresponding patterns will be investigated in the next section by numerical simulations.

In case (ii) d=10, the following result holds: the stability interval is I = [0, ν0] U ]ν0, ν∗] = I0 U I1, such that in I0 the homogeneous equilibrium E2 remains stable, while in I1 (red dashed line) the

Turing instability holds. Of course, the interval I0 tends to zero for d ≥ d*=d+(0). Also in this case,

the patterns arising for ν ∈ I1 will be investigated by numerical simulations. Moreover, we consider

also the case (iii), ν > ν* for any value of the diffusion parameter d >0, when the non trivial equilibrium is unstable. Numerical investigations will elucidate also the dynamic details of the fate of the steady state of the reaction diffusion system in this situation.

These results are summarized in Fig. 4.

Fig. 4 – Dynamic scenario in the (d,ν) parameter space.

We found that there exists an interval I =[0, ν*] where, in absence of diffusion, the homogeneous

equilibrium E2 is stable and such that for ν > ν* it becomes unstable. This is shown by the blue line

describing the condition (C1), that is the trace tr(ν) of the Jacobian matrix J(E2).

In presence of diffusion, for all ν ∈ I and for d=d(ν) ≤ d+(ν), (that is below the red continuous

curve) , E2 continues to be a stable equilibrium, while for d=d(ν) > d+(ν) it undergoes diffusion driven instability.

For example, in Fig. 4, we highlight the cases (i) d=40 > d+(0) :=d*=36.886 and (ii) d=10. In

case (i), for ν=0 we have a situation similar to that studied in [8] and a spotty pattern configuration

arises, while for 0 < ν ≤ ν*, that is ν ∈ I (red dashed line), the Turing instability condition holds and the corresponding patterns will be investigated in the next section by numerical simulations.

In case (ii) d=10, the following result holds: the stability interval is I = [0, ν0] U ]ν0, ν∗] = I0 U I1, such that in I0 the homogeneous equilibrium E2 remains stable, while in I1 (red dashed line) the

Turing instability holds. Of course, the interval I0 tends to zero for d ≥ d*=d+(0). Also in this case,

the patterns arising for ν ∈ I1 will be investigated by numerical simulations. Moreover, we consider

also the case (iii), ν > ν* for any value of the diffusion parameter d >0, when the non trivial equilibrium is unstable. Numerical investigations will elucidate also the dynamic details of the fate of the steady state of the reaction diffusion system in this situation.


Numerical Simulations and Comparison with Experiments

In order to analyse the three cases (i)—(iii) described in the previous section, we solve numerically the coupled reaction-diffusion equations (1) by the software COMSOL Multiphysics [25] based on the finite-element method. All simulations are performed by fixing the starting approximations in

Eq. 2 given by η0(x,y)=2ηe + cη cos(x,y) and θ0(x,y)=2θe + cθ cos(x,y), representing a spatial perturbation of the homogeneous non-trivial equilibrium E2. The initial profile solutions are those

defined above with amplitudes: cη = 0.05, cθ = 0.001. The spatial and temporal intervals considered

are Ω =[0, 100] × [0,70] and t ∈ [0, 200], respectively.

In order to outline the stabilising role of the mass-transport, ν has been varied, whereas all the other parameters have been kept fixed as before, that is such that the equilibrium E2 is stable when d = 0.

The points of the parameter subspace (d,ν) chosen for the COMSOL simulations are depicted in Fig. 5.

Fig. 5 – A subset of the dynamic scenario in the (d,ν) parameter space with indication of the point chosen for the numerical simulations.

COMSOL Simulations:

Case i) for d ≥ d* , for example d=40, and ν ∈ I, spotty patterns still arise, but for increasing values

of ν the number of spots increases and they become larger and higher, see Fig. 6 and 7. This type of pattern change is coherent with the overall morphology variation obtained by changing from aqueous (Fig. 1) to ionic liquid (Fig. 3) solvents Case ii) for d=10, in the interval I0, the non trivial homogeneous equilibrium E2 is still stable, but in

the interval I1, several kinds of patterns arise. For increasing values of ν, labyrintine, stripes and patterns with larger spots are found, as shown in Fig. 8. Moreover, increasing the diffusion

parameter d and fixing ν=1 such that we are still within the Turing zone, the labyrinthine patterns

Numerical Simulations and Comparison with Experiments

In order to analyse the three cases (i)—(iii) described in the previous section, we solve numerically the coupled reaction-diffusion equations (1) by the software COMSOL Multiphysics [25] based on the finite-element method. All simulations are performed by fixing the starting approximations in

Eq. 2 given by η0(x,y)=2ηe + cη cos(x,y) and θ0(x,y)=2θe + cθ cos(x,y), representing a spatial perturbation of the homogeneous non-trivial equilibrium E2. The initial profile solutions are those

defined above with amplitudes: cη = 0.05, cθ = 0.001. The spatial and temporal intervals considered

are Ω =[0, 100] × [0,70] and t ∈ [0, 200], respectively.

In order to outline the stabilising role of the mass-transport, ν has been varied, whereas all the other parameters have been kept fixed as before, that is such that the equilibrium E2 is stable when d = 0.

The points of the parameter subspace (d,ν) chosen for the COMSOL simulations are depicted in Fig. 5.

Fig. 5 – A subset of the dynamic scenario in the (d,ν) parameter space with indication of the point chosen for the numerical simulations.

COMSOL Simulations:

Case i) for d ≥ d* , for example d=40, and ν ∈ I, spotty patterns still arise, but for increasing values

of ν the number of spots increases and they become larger and higher, see Fig. 6 and 7. This type of pattern change is coherent with the overall morphology variation obtained by changing from aqueous (Fig. 1) to ionic liquid (Fig. 3) solvents Case ii) for d=10, in the interval I0, the non trivial homogeneous equilibrium E2 is still stable, but in

the interval I1, several kinds of patterns arise. For increasing values of ν, labyrintine, stripes and patterns with larger spots are found, as shown in Fig. 8. Moreover, increasing the diffusion

parameter d and fixing ν=1 such that we are still within the Turing zone, the labyrinthine patterns


become stripes and then large spots (as in case (i)). This mathematical scenario corresponds to the experiments carried out with the ionic liquid bath at different plating current densities (Fig. 3): the relationship between the plating rate and the surface diffusion coefficient d is extensively discussed in [8]. One can notice that the changes in morphologies corresponding to low- (Fig. 3, panels D-F) and high-current-density conditions (Fig. 3, panels G-I) (under identical mass-transport conditions)

closely resemble the changes in patters observed by changing d at fixed ν (compare Fig. 6, panels B-C and Fig. 8, panels E-F).

Case iii) for any d, if ν > ν *, but for ν ≅ ν * (Fig. 6, panel C and Fig. 8, panel H), a transition zone

is present where E2 destabilizes, but does not attain another steady solution. Instead for ν >> ν* the trivial equilibrium E1 is attained.

(Α) ν=0 (Β) ν=1 (C) ν=2 ν=10, equilibrium E1=0 Fig. 6 – Numerical simulations corresponding to Case (i), d=40

Fig. 7 - Numerical simulations corresponding to Case (i): d=40, Spotty patterns for ν=0,0.5, 1

become stripes and then large spots (as in case (i)). This mathematical scenario corresponds to the experiments carried out with the ionic liquid bath at different plating current densities (Fig. 3): the relationship between the plating rate and the surface diffusion coefficient d is extensively discussed in [8]. One can notice that the changes in morphologies corresponding to low- (Fig. 3, panels D-F) and high-current-density conditions (Fig. 3, panels G-I) (under identical mass-transport conditions)

closely resemble the changes in patters observed by changing d at fixed ν (compare Fig. 6, panels B-C and Fig. 8, panels E-F).

Case iii) for any d, if ν > ν *, but for ν ≅ ν * (Fig. 6, panel C and Fig. 8, panel H), a transition zone

is present where E2 destabilizes, but does not attain another steady solution. Instead for ν >> ν* the trivial equilibrium E1 is attained.

(Α) ν=0 (Β) ν=1 (C) ν=2 ν=10, equilibrium E1=0 Fig. 6 – Numerical simulations corresponding to Case (i), d=40

Fig. 7 - Numerical simulations corresponding to Case (i): d=40, Spotty patterns for ν=0,0.5, 1


(D) ν=0, equilibrium E2=2 (E) ν=0.5 (F) ν=1

(G) ν=1.5 (H) ν=2 ν=10, equilibrium E1=0

Fig. 8 - Numerical simulations corresponding to Case (ii), d=10

Conclusions

In this paper we discuss and validate experimentally an extension - incorporating mass-transport effects - of a reaction-diffusion model coupling morphology development and cathodic adsorption chemistry for metal electroplating, recently developed by some of the authors. The source term for the morphology equation has been modified in order to account for mass-transport of electroactive species from the bath bulk to the cathodic surface, by suitably combining the limiting current density (iL) expression derived by the boundary-layer theory, with Tafel equations for charge-transfer kinetics. Two major dynamic effects are thus obtained: (i) a modification of the Turing instability scenario, leading from the formation of patters corresponding to outgrowth of individual crystallites, to a general electrode smoothing; (ii) instabilisation of the otherwise stable equilibrium undergoing Turing instability, if a critical value of iL

-1 is exceeded: this effect is a manifestation of (boundary-layer) diffusion-induced instability that appears right from the linear stability analysis, owing to a modification of the source term accounting for mass transport, before the introduction of (explicitly space-dependent) diffusion instability. This way, we can differentiate between: (i) instability induced by 2D mass-transport (surface diffusion), i.e. Turing instability and (ii) instability induced by 3D mass-transport (limiting current density). A smoothing effect of mass-transport might sound paradoxical in view of the typical behaviour of aqueous solutions, but it is worth recalling that: (i) such smoothing effect is the result of a subtle synergy of: mass-transport from the bulk of the solution, surface diffusion of adatoms, cathodic adsoprtion; (ii) in many aqueous plating systems, cathodic adsorption is not an intrinsic aspect of the metal plating process as it is instead for an ionic liquid electrolyte. The Mn electrodeposition experiments reported in this paper – apart from validating our extended mathematical model - have achieved two major metal plating results in their own right: (i) we have shown that chelating agents can play a vital role in the electrodeposition of compact Mn films – though not smooth enough for aeronautic applications - from aqueous solutions with a relatively high cathodic efficiency, their effectiveness seems to correlate positively with the number of

chelating moieties present in the molecule; (ii) notably smooth, compact, γ-phase Mn layers can be

(D) ν=0, equilibrium E2=2 (E) ν=0.5 (F) ν=1

(G) ν=1.5 (H) ν=2 ν=10, equilibrium E1=0

Fig. 8 - Numerical simulations corresponding to Case (ii), d=10

Conclusions

In this paper we discuss and validate experimentally an extension - incorporating mass-transport effects - of a reaction-diffusion model coupling morphology development and cathodic adsorption chemistry for metal electroplating, recently developed by some of the authors. The source term for the morphology equation has been modified in order to account for mass-transport of electroactive species from the bath bulk to the cathodic surface, by suitably combining the limiting current density (iL) expression derived by the boundary-layer theory, with Tafel equations for charge-transfer kinetics. Two major dynamic effects are thus obtained: (i) a modification of the Turing instability scenario, leading from the formation of patters corresponding to outgrowth of individual crystallites, to a general electrode smoothing; (ii) instabilisation of the otherwise stable equilibrium undergoing Turing instability, if a critical value of iL

-1 is exceeded: this effect is a manifestation of (boundary-layer) diffusion-induced instability that appears right from the linear stability analysis, owing to a modification of the source term accounting for mass transport, before the introduction of (explicitly space-dependent) diffusion instability. This way, we can differentiate between: (i) instability induced by 2D mass-transport (surface diffusion), i.e. Turing instability and (ii) instability induced by 3D mass-transport (limiting current density). A smoothing effect of mass-transport might sound paradoxical in view of the typical behaviour of aqueous solutions, but it is worth recalling that: (i) such smoothing effect is the result of a subtle synergy of: mass-transport from the bulk of the solution, surface diffusion of adatoms, cathodic adsoprtion; (ii) in many aqueous plating systems, cathodic adsorption is not an intrinsic aspect of the metal plating process as it is instead for an ionic liquid electrolyte. The Mn electrodeposition experiments reported in this paper – apart from validating our extended mathematical model - have achieved two major metal plating results in their own right: (i) we have shown that chelating agents can play a vital role in the electrodeposition of compact Mn films – though not smooth enough for aeronautic applications - from aqueous solutions with a relatively high cathodic efficiency, their effectiveness seems to correlate positively with the number of

chelating moieties present in the molecule; (ii) notably smooth, compact, γ-phase Mn layers can be


plated with cathodic efficiencies in excess of 90% from eutectic ionic liquid electrolytes; these layer exhibit both a corrosion performance and a surface finish that can be regarded as adequate for the aeronautic industry.

References

[1] B. Bozzini, E. Griskonis, A. Fanigliulo, A. Sulcius: Surf. & Coat. Technol. Vol. 154 (2002) p. 294

[2] B. Bozzini, E. Griskonis, A. Sulcius, P.L. Cavallotti: Plat. Surf. Fin. Vol. 88 (2001), p. 64 [3] B. Bozzini: Trans. Inst. Met. Finish. Vol. 78 (2000), p. 93 [4] B. Bozzini, A. Bund, B. Busson, Ch. Humbert, A. Ispas, C. Mele, A. Tadjeddine: Electrochem.

Commun. Vol. 12 (2010), p. 56 [5] G. Giovannelli, L. D'Urzo, G. Maggiulli, S. Natali, C. Pagliara, I. Sgura and B. Bozzini: J. Solid State Electrochem. Vol. 14 (2010), p. 479 [6] B. Bozzini, B. Busson, Ch. Humbert, C. Mele, P. Raffa and A. Tadjeddine. "An in situ SFG

Investigation of Au electrodeposition from the room temperature ionic liquid BMP-TFSA, containing Au(I) cyanocomplex" In preparation.

[7] B. Bozzini, E. Tondo, A. Bund, A. Ispas, C. Mele. "Electrodeposition of Au from [EMIm][TFSI] room temperature ionic liquid: a study based on Surface-Enhanced Raman Spectroscopy" In preparation.

[8] B. Bozzini, D. Lacitignola, I. Sgura: Mathematical Biosciences and Engineering. Vol. 7, N: 2 (2010), p.237

[9] B. Bozzini, D. Lacitignola, I. Sgura. "Travelling Waves in a Reaction-Diffusion Model for Electrodeposition": Mathematics and Computers in Simulation. In press.

[10] B. Bozzini, D. Lacitignola, I. Sgura: Journal of Physics: Conference Series Vol. 96 (2008), p. 012051

[11] J.-K. Chang, C.-H Huang, W.-T. Tsai, M.-J. Deng, I.-W. Sun, P.-Y. Chen: Electrochim. Acta Vol. 53 (2008) p. 4447

[12] M.-J. Deng, P.-Y. Chen, I.-W. Sun: Electrochim. Acta Vol. 53 (2007), p. 1931 [13] P.P Chung, P.A. Cantwell, G.D. Wilcox, G.W. Critchlow: Trans. Inst. Met. Finish. Vol. 86(4)

(2008), p. 211 [14] S. Ruan, C.A. Schuh: Acta Mater. Vol. 57 (2009), p. 3810 [15] D.-X. Zhuang, M.-J. Deng, P.-Y. Chen, I.-W. Sun: J. Electrochem. Soc. Vol. 155 (2009),

p. D575 [16] P.-Y. Chen, C.L. Hussey: Electrochim. Acta Vol. 52 (2007), p. 1857 [17] M.-J. Deng, P.-Y. Chen, T.-I. Leong, I.-W. Sun, J.-K. Chang, W.-T. Tsai: Electrochem.

Commun. Vol. 10 (2008), p. 213 [18] J.-K. Chang, M.-T Lee, C.-W. Cheng, W.-T. Tsai, M.-J. Deng, I.-W. Sun: J. Mater. Chem.

Vol. 19 (2009), p. 3732 [19] Q. Wei, X. Ren, J. Du, S. Wie, S.R. Hu: Minerals Engineering Vol. 23 (2010), p. 578. [20] P. Díaz-Arista, R. Antaño-López, Y. Meas, R. Ortega, E. Chainet, P. Ozil, G. Trejo:

Electrochim. Acta Vol. 51 (2006), p. 4393 [21] P. Díaz-Arista, G. Trejo: Surf. & Coat. Technol. Vol. 201 (2006), p. 3359 [22] T. Agladze, in: New Materials and Technologies in Surface Finishing for Better Corrosion and

Tribology Properties, A. Choms, Ed., E.G. Leuze Vlg., Saulgau (D) (1993) p. 109 [23] S.U.M. Kahn: J. Phys. Chem, Vol. 92 (1988), p. 2541 [24] B. Bozzini. J. Chem. Edu. Vol. 77 (2000), p. 100 [25] COMSOL MULTIPHYSICS v.3.5a User’s guide (2009).

plated with cathodic efficiencies in excess of 90% from eutectic ionic liquid electrolytes; these layer exhibit both a corrosion performance and a surface finish that can be regarded as adequate for the aeronautic industry.

References

[1] B. Bozzini, E. Griskonis, A. Fanigliulo, A. Sulcius: Surf. & Coat. Technol. Vol. 154 (2002) p. 294

[2] B. Bozzini, E. Griskonis, A. Sulcius, P.L. Cavallotti: Plat. Surf. Fin. Vol. 88 (2001), p. 64 [3] B. Bozzini: Trans. Inst. Met. Finish. Vol. 78 (2000), p. 93 [4] B. Bozzini, A. Bund, B. Busson, Ch. Humbert, A. Ispas, C. Mele, A. Tadjeddine: Electrochem.

Commun. Vol. 12 (2010), p. 56 [5] G. Giovannelli, L. D'Urzo, G. Maggiulli, S. Natali, C. Pagliara, I. Sgura and B. Bozzini: J. Solid State Electrochem. Vol. 14 (2010), p. 479 [6] B. Bozzini, B. Busson, Ch. Humbert, C. Mele, P. Raffa and A. Tadjeddine. "An in situ SFG

Investigation of Au electrodeposition from the room temperature ionic liquid BMP-TFSA, containing Au(I) cyanocomplex" In preparation.

[7] B. Bozzini, E. Tondo, A. Bund, A. Ispas, C. Mele. "Electrodeposition of Au from [EMIm][TFSI] room temperature ionic liquid: a study based on Surface-Enhanced Raman Spectroscopy" In preparation.

[8] B. Bozzini, D. Lacitignola, I. Sgura: Mathematical Biosciences and Engineering. Vol. 7, N: 2 (2010), p.237

[9] B. Bozzini, D. Lacitignola, I. Sgura. "Travelling Waves in a Reaction-Diffusion Model for Electrodeposition": Mathematics and Computers in Simulation. In press.

[10] B. Bozzini, D. Lacitignola, I. Sgura: Journal of Physics: Conference Series Vol. 96 (2008), p. 012051

[11] J.-K. Chang, C.-H Huang, W.-T. Tsai, M.-J. Deng, I.-W. Sun, P.-Y. Chen: Electrochim. Acta Vol. 53 (2008) p. 4447

[12] M.-J. Deng, P.-Y. Chen, I.-W. Sun: Electrochim. Acta Vol. 53 (2007), p. 1931 [13] P.P Chung, P.A. Cantwell, G.D. Wilcox, G.W. Critchlow: Trans. Inst. Met. Finish. Vol. 86(4)

(2008), p. 211 [14] S. Ruan, C.A. Schuh: Acta Mater. Vol. 57 (2009), p. 3810 [15] D.-X. Zhuang, M.-J. Deng, P.-Y. Chen, I.-W. Sun: J. Electrochem. Soc. Vol. 155 (2009),

p. D575 [16] P.-Y. Chen, C.L. Hussey: Electrochim. Acta Vol. 52 (2007), p. 1857 [17] M.-J. Deng, P.-Y. Chen, T.-I. Leong, I.-W. Sun, J.-K. Chang, W.-T. Tsai: Electrochem.

Commun. Vol. 10 (2008), p. 213 [18] J.-K. Chang, M.-T Lee, C.-W. Cheng, W.-T. Tsai, M.-J. Deng, I.-W. Sun: J. Mater. Chem.

Vol. 19 (2009), p. 3732 [19] Q. Wei, X. Ren, J. Du, S. Wie, S.R. Hu: Minerals Engineering Vol. 23 (2010), p. 578. [20] P. Díaz-Arista, R. Antaño-López, Y. Meas, R. Ortega, E. Chainet, P. Ozil, G. Trejo:

Electrochim. Acta Vol. 51 (2006), p. 4393 [21] P. Díaz-Arista, G. Trejo: Surf. & Coat. Technol. Vol. 201 (2006), p. 3359 [22] T. Agladze, in: New Materials and Technologies in Surface Finishing for Better Corrosion and

Tribology Properties, A. Choms, Ed., E.G. Leuze Vlg., Saulgau (D) (1993) p. 109 [23] S.U.M. Kahn: J. Phys. Chem, Vol. 92 (1988), p. 2541 [24] B. Bozzini. J. Chem. Edu. Vol. 77 (2000), p. 100 [25] COMSOL MULTIPHYSICS v.3.5a User’s guide (2009).


Understanding Nanoscale Wetting using Dynamic Local Contact Angle Method

Martin Losada1,a, KatherineMackie1,a, Joseph H. Osborne2,b,

Santanu Chaudhuri1,a

1ISP/Applied Sciences Laboratory, WashingtonStateUniversity, Spokane, WA99210

2The Boeing Company, 9725 E Marginal Way South, Seattle, WA98108


Keywords: Molecular dynamics, dynamic local contact angle, coatings, hydrophobicity

Abstract: A multiscale quantum/classical-framework for hydrophobicity and UV absorption in heterogeneous coatings is presented. Atomistic water droplet simulations on coated oxide surface are used to define nanoscale contact-angles using a new numerical technique called the dynamic local contact angle (DLCA) method. The DLCA method is well suited to calculate macroscopic contact angles for polymeric and composite coatings. The accuracy of the method is tested for a series of common polymers and composites. In addition, the sensitivity of the contact angles towards functional groups and nanoscale roughness are tested using varying molecular structures. Fluorinated polyhedral oligomericsilsesquioxanes (F-POSS) molecular frameworks are used as a model system. Changes in contact angle and UV absorption spectrum as a function of hydrophobic chain length are calculated to test the feasibility of developing a virtual framework for new coating design connecting atomistic calculations to continuum level material properties.

Introduction

Hydrophobicity and durability of coatings are closely related. In general, hydrophobic coatings are defined as coatings with net macroscopic contact-angles above 90o for water droplet. Intrinsic surface energy is a large contributor to the contact angle. However nanoscale roughness, such as might result from surface degradation and/or incorporation of nanoparticles to a polymer, can alter wetting and change macroscopic contact angles and these correlations are not clearly understood. This is primarily because nanoparticles are not inert additives in a composite. The nanophase alters the host polymer in myriad ways. As a result, composition of multicomponent composite coatings manifests into local transitions in wetting (Wenzel state) and the contact angles further gets modified by local domains. So a net contact angle in multiphasic system becomes an increasingly inaccurate measure of average hydrophobic behavior of coatings and interaction of fluid at the material interfaces. A natural consequence of this complexity leads to an important question: What is the length scale at which material inhomogeneity needs to be treated explicitly in order to connect local fluctuations to their macroscopic behavior? As such, there is an important need to identify changes in macroscopic coating properties from altered atomistic scale environment in composites and nanoscale functional materials. A new multiscale framework is therefore proposed to address these challenging problems using atomistic simulations. It is based on performing interfacial dynamics calculation of nanoscale water droplet on coating surfaces. To connect such a dynamic environment to macroscale properties of coating, a localized definition of contact-angle is developed. Changes in this contact angle as a function of material inhomogeneity, nano-to-macroscale surface roughness, and thermal conditions can be valuable tool for material development and understanding. The atomistic scale definition of contact-angle is based on contact line segments. A contact-line is defined as the line around a water droplet on a surface divided into small linear facets where the water droplet creates the tri-phase boundary between air, water and the solid surface. The length of

Understanding Nanoscale Wetting using Dynamic Local Contact Angle Method

Martin Losada1,a, KatherineMackie1,a, Joseph H. Osborne2,b,

Santanu Chaudhuri1,a

1ISP/Applied Sciences Laboratory, WashingtonStateUniversity, Spokane, WA99210

2The Boeing Company, 9725 E Marginal Way South, Seattle, WA98108


Keywords: Molecular dynamics, dynamic local contact angle, coatings, hydrophobicity

Abstract: A multiscale quantum/classical-framework for hydrophobicity and UV absorption in heterogeneous coatings is presented. Atomistic water droplet simulations on coated oxide surface are used to define nanoscale contact-angles using a new numerical technique called the dynamic local contact angle (DLCA) method. The DLCA method is well suited to calculate macroscopic contact angles for polymeric and composite coatings. The accuracy of the method is tested for a series of common polymers and composites. In addition, the sensitivity of the contact angles towards functional groups and nanoscale roughness are tested using varying molecular structures. Fluorinated polyhedral oligomericsilsesquioxanes (F-POSS) molecular frameworks are used as a model system. Changes in contact angle and UV absorption spectrum as a function of hydrophobic chain length are calculated to test the feasibility of developing a virtual framework for new coating design connecting atomistic calculations to continuum level material properties.

Introduction

Hydrophobicity and durability of coatings are closely related. In general, hydrophobic coatings are defined as coatings with net macroscopic contact-angles above 90o for water droplet. Intrinsic surface energy is a large contributor to the contact angle. However nanoscale roughness, such as might result from surface degradation and/or incorporation of nanoparticles to a polymer, can alter wetting and change macroscopic contact angles and these correlations are not clearly understood. This is primarily because nanoparticles are not inert additives in a composite. The nanophase alters the host polymer in myriad ways. As a result, composition of multicomponent composite coatings manifests into local transitions in wetting (Wenzel state) and the contact angles further gets modified by local domains. So a net contact angle in multiphasic system becomes an increasingly inaccurate measure of average hydrophobic behavior of coatings and interaction of fluid at the material interfaces. A natural consequence of this complexity leads to an important question: What is the length scale at which material inhomogeneity needs to be treated explicitly in order to connect local fluctuations to their macroscopic behavior? As such, there is an important need to identify changes in macroscopic coating properties from altered atomistic scale environment in composites and nanoscale functional materials. A new multiscale framework is therefore proposed to address these challenging problems using atomistic simulations. It is based on performing interfacial dynamics calculation of nanoscale water droplet on coating surfaces. To connect such a dynamic environment to macroscale properties of coating, a localized definition of contact-angle is developed. Changes in this contact angle as a function of material inhomogeneity, nano-to-macroscale surface roughness, and thermal conditions can be valuable tool for material development and understanding. The atomistic scale definition of contact-angle is based on contact line segments. A contact-line is defined as the line around a water droplet on a surface divided into small linear facets where the water droplet creates the tri-phase boundary between air, water and the solid surface. The length of

each contact line segment is determined by the material inhomogeneity and the size of the water droplet. In addition to contact line segments, atomistic molecular dynamic trajectories are analyzed to track time-dependent changes in the droplet shape and the coating constituents. The information obtained from nanoscale simulations can be extended to microscale using finite-element methods to account for macroscale surface roughness. In finite elements length-scale, surfaces roughness in µm range can be considered to be the only factor that can change the nanoscale calculated contact-angles. This assumption is reasonable for our multiscale framework where surface roughness, flow and thermal equilibrium are only considered in macroscale. Polyhedral oligomericsilsesquioxanes (POSS) are one type of hybrid inorganic/organic material of the form (RSiO3/2)n, where organic substituents (Rn) are attached to an oxygen-silicon cage. The most common POSS cage is the R8-Si8O12, a molecule with a cubic array of silicon atoms and bridging oxygen atoms with eight R groups at the vertexes of the cube. When these Si-O cage structures are incorporated into organic polymers, exciting possibilities for the development of new materials are often realized [1], with properties superior to the original organic polymer. For example, the low surface energy properties of fluorinated POSS compounds have been used to augment hydrophobic properties of both fluorinated and nonfluorinated polymers [2-4].

Material and Methods

Multiple common polymers and nanoscale molecular framework consisting of fluorinated polyhedral oligomericsilsesquioxanes (F-POSS) are considered in this work. The F-POSS molecules are shown in Fig. 1 (A-C) with increasing length of hydrophobic chain attached to the Si-O-Si framework. Fig. 1D has a hydrolyzed Si-O-Si linkage to simulated damaged or degraded F-POSS framework. Different ways in which POSS can be used as a functionalizable framework is explained elsewhere [1].

A

B

C

D

Fig. 1: POSS molecules with variable chain-lengths to create different nano-roughness There is an accumulating series of proof that Wenzel/Cassie regimes are not as easily definable in nanoscale. We demonstrate that a dynamic view of wetting-dewetting dynamics is possible using a dynamics local contact angle (DLCA) method. In addition, the sensitivity of the contact angles towards molecular level changes is tested using varying molecular framework consisting of F-POSS molecules shown in Fig. 1. The results demonstrate that DLCA method can allow for predicting macroscale contact angles.

each contact line segment is determined by the material inhomogeneity and the size of the water droplet. In addition to contact line segments, atomistic molecular dynamic trajectories are analyzed to track time-dependent changes in the droplet shape and the coating constituents. The information obtained from nanoscale simulations can be extended to microscale using finite-element methods to account for macroscale surface roughness. In finite elements length-scale, surfaces roughness in µm range can be considered to be the only factor that can change the nanoscale calculated contact-angles. This assumption is reasonable for our multiscale framework where surface roughness, flow and thermal equilibrium are only considered in macroscale. Polyhedral oligomericsilsesquioxanes (POSS) are one type of hybrid inorganic/organic material of the form (RSiO3/2)n, where organic substituents (Rn) are attached to an oxygen-silicon cage. The most common POSS cage is the R8-Si8O12, a molecule with a cubic array of silicon atoms and bridging oxygen atoms with eight R groups at the vertexes of the cube. When these Si-O cage structures are incorporated into organic polymers, exciting possibilities for the development of new materials are often realized [1], with properties superior to the original organic polymer. For example, the low surface energy properties of fluorinated POSS compounds have been used to augment hydrophobic properties of both fluorinated and nonfluorinated polymers [2-4].

Material and Methods

Multiple common polymers and nanoscale molecular framework consisting of fluorinated polyhedral oligomericsilsesquioxanes (F-POSS) are considered in this work. The F-POSS molecules are shown in Fig. 1 (A-C) with increasing length of hydrophobic chain attached to the Si-O-Si framework. Fig. 1D has a hydrolyzed Si-O-Si linkage to simulated damaged or degraded F-POSS framework. Different ways in which POSS can be used as a functionalizable framework is explained elsewhere [1].

A

B

C

D

Fig. 1: POSS molecules with variable chain-lengths to create different nano-roughness There is an accumulating series of proof that Wenzel/Cassie regimes are not as easily definable in nanoscale. We demonstrate that a dynamic view of wetting-dewetting dynamics is possible using a dynamics local contact angle (DLCA) method. In addition, the sensitivity of the contact angles towards molecular level changes is tested using varying molecular framework consisting of F-POSS molecules shown in Fig. 1. The results demonstrate that DLCA method can allow for predicting macroscale contact angles.


General MD simulations:Classical simulations of water droplets on aluminum oxide and polymer-coated oxide surfaces were carried out using the Discover MD simulation program and the COMPASS ab initio parameterized force-field; these are well-suited to condensed matter simulations. Both the bulk and surfaces of all the systems presented here have been considered. The oxide surface is kept fixed and the primary effect from the oxide surface is Coulombic long range forces. All of the MD simulations used periodic boundary conditions in all three directions. As a result, the vacuum layer on top of the oxide-polymer layers wasexpanded (120 Å) to avoid interaction between periodic images in vertical directions. Descriptions of the COMPASS force-field and the accuracy of its water model are published elsewhere[5,6]. Overall, the potential for water-polymer interaction and consistent treatment across the materials landscape allowed us to use the same force-field across various polymeric and oxide systems, something that is very difficult to achieve using competing models. Also, the COMPASS model for water, although a three-site model similar to SPC/E, is quite accurate for predicting ice and water densities [5,7]. The COMPASS water force-field, however, faces difficulty at high pressures and temperatures. The water droplet of 4 nm radius was used in all simulations. For the Fluorinated Polyhedral OligomericSilsesquioxanes molecules, three different structures were considered: 8C6F13-POSS molecules, named as F-POSS-3, 8C6H4F9-POSS molecules, named as F-POSS-4, and 8C10H4F17-POSS molecules, named as F-POSS-5, structures A, B, and C in Figure 1, respectively.The initial F-POSS-3,4,5 configurations are obtained by building an amorphous layer composed of forty molecules using the Amorphous Cell Tools of the Materials Studio (Accelrys Inc.) suite of programs. The F-POSS-3,4,5-water system, are then built as a two layer systems, where the first layer corresponds to the amorphous F-POSS-and the second to the water droplet. All the simulations were performed at 298 K in the NVT (constant number of particles, volume, and temperature) canonical ensemble. The COMPASS force field, which has been used for the simulation of POSS and polymer/POSS composites [6,8,9], is employed for the simulations. The initial configurations of the F-POSS-3,4,5 -water droplet systems were subjected to energy minimization for 8000 iterations. Then, NVT MD production simulations that last 300 psfor equilibration and statistics were collected from last 200 ps of a total 500 ps simulation. All simulations were carried out 300 K unless otherwise noted for low temperature tests. The resulting trajectorieswere then used to study the hydrophobic properties by computing the contact angle between the amorphous polymer layer(s) and the water droplet. In MD simulations, the van der Waals interactions were calculated using the atom-based method with a cut-off distance of 9.50 Å, the Coulombic interactions were taken into account using the Ewald method. To integrate the equations of motion, we use the Discover simulation program (Accelrys Inc.) and canonical (NVT) ensemble. The temperature is adjusted using the Nose method. The DLCA algorithm to collect contact angles from MD trajectories are described below. The DLCA Algorithm:First, we will describe how DLCA methods determine contact angle. An atomistic model of a water droplet is a collection of points in 3D space. We identify the atoms on the outer surface of a water droplet by calculating a mathematical quantity, called the convex hull; the convex hull is calculated by expanding these atomic coordinates into a spherical distribution of points representing the atoms’vdW radii. The vdW radii of the outermost atoms from the water droplet and the coating’s surface touch along a three point contact line. The polyhedral boundary of the droplet on the coating surface is a collection of all the contiguous contact lines, each corresponding to a line segment with a local contact angle with respect to the surface. The local contact angle is thus the dihedral angle between two intersecting triangles—one from the convex hull of water and the other from the convex hull of the polymeric surface as shown in Fig.2.

General MD simulations:Classical simulations of water droplets on aluminum oxide and polymer-coated oxide surfaces were carried out using the Discover MD simulation program and the COMPASS ab initio parameterized force-field; these are well-suited to condensed matter simulations. Both the bulk and surfaces of all the systems presented here have been considered. The oxide surface is kept fixed and the primary effect from the oxide surface is Coulombic long range forces. All of the MD simulations used periodic boundary conditions in all three directions. As a result, the vacuum layer on top of the oxide-polymer layers wasexpanded (120 Å) to avoid interaction between periodic images in vertical directions. Descriptions of the COMPASS force-field and the accuracy of its water model are published elsewhere[5,6]. Overall, the potential for water-polymer interaction and consistent treatment across the materials landscape allowed us to use the same force-field across various polymeric and oxide systems, something that is very difficult to achieve using competing models. Also, the COMPASS model for water, although a three-site model similar to SPC/E, is quite accurate for predicting ice and water densities [5,7]. The COMPASS water force-field, however, faces difficulty at high pressures and temperatures. The water droplet of 4 nm radius was used in all simulations. For the Fluorinated Polyhedral OligomericSilsesquioxanes molecules, three different structures were considered: 8C6F13-POSS molecules, named as F-POSS-3, 8C6H4F9-POSS molecules, named as F-POSS-4, and 8C10H4F17-POSS molecules, named as F-POSS-5, structures A, B, and C in Figure 1, respectively.The initial F-POSS-3,4,5 configurations are obtained by building an amorphous layer composed of forty molecules using the Amorphous Cell Tools of the Materials Studio (Accelrys Inc.) suite of programs. The F-POSS-3,4,5-water system, are then built as a two layer systems, where the first layer corresponds to the amorphous F-POSS-and the second to the water droplet. All the simulations were performed at 298 K in the NVT (constant number of particles, volume, and temperature) canonical ensemble. The COMPASS force field, which has been used for the simulation of POSS and polymer/POSS composites [6,8,9], is employed for the simulations. The initial configurations of the F-POSS-3,4,5 -water droplet systems were subjected to energy minimization for 8000 iterations. Then, NVT MD production simulations that last 300 psfor equilibration and statistics were collected from last 200 ps of a total 500 ps simulation. All simulations were carried out 300 K unless otherwise noted for low temperature tests. The resulting trajectorieswere then used to study the hydrophobic properties by computing the contact angle between the amorphous polymer layer(s) and the water droplet. In MD simulations, the van der Waals interactions were calculated using the atom-based method with a cut-off distance of 9.50 Å, the Coulombic interactions were taken into account using the Ewald method. To integrate the equations of motion, we use the Discover simulation program (Accelrys Inc.) and canonical (NVT) ensemble. The temperature is adjusted using the Nose method. The DLCA algorithm to collect contact angles from MD trajectories are described below. The DLCA Algorithm:First, we will describe how DLCA methods determine contact angle. An atomistic model of a water droplet is a collection of points in 3D space. We identify the atoms on the outer surface of a water droplet by calculating a mathematical quantity, called the convex hull; the convex hull is calculated by expanding these atomic coordinates into a spherical distribution of points representing the atoms’vdW radii. The vdW radii of the outermost atoms from the water droplet and the coating’s surface touch along a three point contact line. The polyhedral boundary of the droplet on the coating surface is a collection of all the contiguous contact lines, each corresponding to a line segment with a local contact angle with respect to the surface. The local contact angle is thus the dihedral angle between two intersecting triangles—one from the convex hull of water and the other from the convex hull of the polymeric surface as shown in Fig.2.


Fig. 2: (a) Faceting on the coating surface and the nanoscale water droplet, (b) DLCA algorithm defining a series of contact angles along the contact line and (c) contact angles color coded along the contact line of a droplet obtained from data from a frame in a 200 ps simulation trajectory for PDMS.

The DLCA algorithm developed uses surface triangulation methods for the water droplet and the surface atoms separately. During the dynamics, each step has a set of triangulated surfaces (as shown in Figure 2a) formed of atomic coordinates in the simulation cell. Furthermore, the definition of contact between coating and droplet cannot use only the atomic coordinates. In a more atomistic definition, contact is made when the vdW surface (dependent on vdWradii of constituent atoms) of atoms from the coatings surface touches the average vdW surface of the water molecules. As a result, each of the atomic coordinates was expandedinto a spherical distribution of points around the atomic position to represent the vdW sphere. Points on this sphere are subsequently used for calculation of surface and water convex hull sets. A macroscopic contact angle is calculated by taking a weighted average of the time-dependent distribution of the DLCA values. The macroscopic contact angle is calculated as an average of local contact angles from each frame and can be expressed as:

,and . (1)

Where, n number of frames are used for averaging, N is the number of contact line segments in a droplet for a frame t, and |ri| is the length of the contact line segment with the same local contact angle θi. A distribution in the value of θican be color-coded with respect to its position on the contact line projections as shown in Fig. 1c. UV absorption calculations:UV absorption calculations: For calculation of UV spectrum of F-POSS containing coating, each individual F-POSS molecules and the polymeric dimmers are used. In this case, an F-POSS/polycarbonate coating was simulated. The geometries of all (F-POSS) molecules displayed in Fig. 1 and that of the polycarbonate dimer structure (see Figure 3) have been fully optimized using the HF/6-31+G(d) [10] level of theory with no symmetry constraints. Becke’s[11] exchange functional in combination with Becke’s[12] three-parameter hybrid functional using the LYP [13] correlation functional of Lee, yang, and Parr (B3LYP) was employed in the computations of the electronic absorption spectra using Time-dependent Density Functional Theory (TD-DFT) [14] at the same level of calculation. To evaluate the basis set dependence of the predicted UV spectra of the polycarbonate dimer, two additional calculations were carried out using the triply split 6-311G* and the 6-311G** basis sets, augmented with one and two polarization functions, respectively. Thus, TD-DFT excited-state calculations were determined, utilizing the previously optimized ground-state geometries, at the B3LYP/6-31+G(d)//HF/6-31+G(d) level for F-POSS molecules and at the B3LYP/6-31+G*, 6-311G* and 6-311G**//HF/6-31+G* level for the polycarbonate dimer. For the formal foundation of TD-DFT, we refer the reader to the paper by Gross and Runge[15]. The lowest three single excited states were investigated to simulate the absorption spectra. The absorption spectra were obtained using the visualization tools in the

Fig. 2: (a) Faceting on the coating surface and the nanoscale water droplet, (b) DLCA algorithm defining a series of contact angles along the contact line and (c) contact angles color coded along the contact line of a droplet obtained from data from a frame in a 200 ps simulation trajectory for PDMS.

The DLCA algorithm developed uses surface triangulation methods for the water droplet and the surface atoms separately. During the dynamics, each step has a set of triangulated surfaces (as shown in Figure 2a) formed of atomic coordinates in the simulation cell. Furthermore, the definition of contact between coating and droplet cannot use only the atomic coordinates. In a more atomistic definition, contact is made when the vdW surface (dependent on vdWradii of constituent atoms) of atoms from the coatings surface touches the average vdW surface of the water molecules. As a result, each of the atomic coordinates was expandedinto a spherical distribution of points around the atomic position to represent the vdW sphere. Points on this sphere are subsequently used for calculation of surface and water convex hull sets. A macroscopic contact angle is calculated by taking a weighted average of the time-dependent distribution of the DLCA values. The macroscopic contact angle is calculated as an average of local contact angles from each frame and can be expressed as:

,and . (1)

Where, n number of frames are used for averaging, N is the number of contact line segments in a droplet for a frame t, and |ri| is the length of the contact line segment with the same local contact angle θi. A distribution in the value of θican be color-coded with respect to its position on the contact line projections as shown in Fig. 1c. UV absorption calculations:UV absorption calculations: For calculation of UV spectrum of F-POSS containing coating, each individual F-POSS molecules and the polymeric dimmers are used. In this case, an F-POSS/polycarbonate coating was simulated. The geometries of all (F-POSS) molecules displayed in Fig. 1 and that of the polycarbonate dimer structure (see Figure 3) have been fully optimized using the HF/6-31+G(d) [10] level of theory with no symmetry constraints. Becke’s[11] exchange functional in combination with Becke’s[12] three-parameter hybrid functional using the LYP [13] correlation functional of Lee, yang, and Parr (B3LYP) was employed in the computations of the electronic absorption spectra using Time-dependent Density Functional Theory (TD-DFT) [14] at the same level of calculation. To evaluate the basis set dependence of the predicted UV spectra of the polycarbonate dimer, two additional calculations were carried out using the triply split 6-311G* and the 6-311G** basis sets, augmented with one and two polarization functions, respectively. Thus, TD-DFT excited-state calculations were determined, utilizing the previously optimized ground-state geometries, at the B3LYP/6-31+G(d)//HF/6-31+G(d) level for F-POSS molecules and at the B3LYP/6-31+G*, 6-311G* and 6-311G**//HF/6-31+G* level for the polycarbonate dimer. For the formal foundation of TD-DFT, we refer the reader to the paper by Gross and Runge[15]. The lowest three single excited states were investigated to simulate the absorption spectra. The absorption spectra were obtained using the visualization tools in the


GaussView4.1 program [16]. All calculations were carried out with the Gaussian 03 [17] software package. Results

The study of the complex surface dynamics of water droplets on an alumina surface starts with the

investigation of water on an alumina surface. The common forms of alumina, such as α-alumina,

crystallize in an 3R c lattice. The oxygen-terminated (001) surface is used as the oxide face on

which a thin layer of coatings (30-50) nm is applied. A series of polymers were used to perform

NVT simulations under ambient and supercooled conditions (300 and 150 K). Minimal evaporation

from the nanoscale water droplet was observed and the droplet deformed according to the interfacial

tension and showed different wetting behavior. A summary of DLCA calculated angles are

tabulated in Table 1.

During the MD simulations ofpolymer coated oxide surfaces, the water droplet is in a dynamic state with the three phase contact line displaying varying local contact angles; the droplet moves around the unconstrained polymeric surface and changes shapes along the contact line. Most importantly, the use of an advanced force field (COMPASS) with explicit terms for water interaction of different functional groups at the interface allowed the surface to respond to the solvation effects accurately at a local level needed for DLCA method and understanding roles of F-POSS molecules embedded on the surface. The temperature dependence of contact angle is also tested. In general contact angle reduced with lowering of temperature in most case however PPMA, Polyurethane and PDMS/silica composite showed slight increase (Table 1).

Table 1: Contact angles from DLCA analysis at 300 and 150 [K] calculated from an averaging of local contact angles.

Polymer 300 [K] 150 [K] Experiment

PDMS 103.86 100.63 101

PMPS 88.73 85.79 90

PMMA 67.8 63.25 68-70

PPMA 86.99 88.96 73

Polystyrene 85.19 77.85 86 +/- 2

Polycarbonate 80.57 65.78 81

Polyurethane 77.76 86.27 73

PDMS-Silica 86.76 89.25 93

The comparison of DLCA calculated and experimental contact-angles are quite good as can be seen in Table 1. It needs to be noted that experimental contact angles are from surfaces of arbitrary roughness and vary almost 5-10% depending on synthesis methods, curing temperature, and solvents used. Considering such variability in coatings, the general similarity is striking. Only PDMS showed hydrophobic contact angles in nanoscale. Changes with roughness can be calculated numerically using the roughness modified Young’s equation if we consider nanoscale rough surface is similar to a smooth surface as assumed in the derivation of Young’s equation. As indicated in the introduction, the incorporation of POSS molecules into a polymeric system enhances its properties. For example, the incorporation of fluorinated-POSS molecules into the

GaussView4.1 program [16]. All calculations were carried out with the Gaussian 03 [17] software package. Results

The study of the complex surface dynamics of water droplets on an alumina surface starts with the

investigation of water on an alumina surface. The common forms of alumina, such as α-alumina,

crystallize in an 3R c lattice. The oxygen-terminated (001) surface is used as the oxide face on

which a thin layer of coatings (30-50) nm is applied. A series of polymers were used to perform

NVT simulations under ambient and supercooled conditions (300 and 150 K). Minimal evaporation

from the nanoscale water droplet was observed and the droplet deformed according to the interfacial

tension and showed different wetting behavior. A summary of DLCA calculated angles are

tabulated in Table 1.

During the MD simulations ofpolymer coated oxide surfaces, the water droplet is in a dynamic state with the three phase contact line displaying varying local contact angles; the droplet moves around the unconstrained polymeric surface and changes shapes along the contact line. Most importantly, the use of an advanced force field (COMPASS) with explicit terms for water interaction of different functional groups at the interface allowed the surface to respond to the solvation effects accurately at a local level needed for DLCA method and understanding roles of F-POSS molecules embedded on the surface. The temperature dependence of contact angle is also tested. In general contact angle reduced with lowering of temperature in most case however PPMA, Polyurethane and PDMS/silica composite showed slight increase (Table 1).

Table 1: Contact angles from DLCA analysis at 300 and 150 [K] calculated from an averaging of local contact angles.

Polymer 300 [K] 150 [K] Experiment

PDMS 103.86 100.63 101

PMPS 88.73 85.79 90

PMMA 67.8 63.25 68-70

PPMA 86.99 88.96 73

Polystyrene 85.19 77.85 86 +/- 2

Polycarbonate 80.57 65.78 81

Polyurethane 77.76 86.27 73

PDMS-Silica 86.76 89.25 93

The comparison of DLCA calculated and experimental contact-angles are quite good as can be seen in Table 1. It needs to be noted that experimental contact angles are from surfaces of arbitrary roughness and vary almost 5-10% depending on synthesis methods, curing temperature, and solvents used. Considering such variability in coatings, the general similarity is striking. Only PDMS showed hydrophobic contact angles in nanoscale. Changes with roughness can be calculated numerically using the roughness modified Young’s equation if we consider nanoscale rough surface is similar to a smooth surface as assumed in the derivation of Young’s equation. As indicated in the introduction, the incorporation of POSS molecules into a polymeric system enhances its properties. For example, the incorporation of fluorinated-POSS molecules into the


polycarbonate (PC) matrix improved the hydroponic and self-cleaning properties of PC as reported by Dodiuk et al. [3]. They also reported that the PC surface roughness was augmented by means of silica and POSS layers. In addition, PC exhibits high light transmission properties (92 %). Thus, transparent PC substrates and PC/F-POSS polymer composites are an attractive system to benchmark the DLCA methods. To investigate the hydrophobic properties of polymer composites, MD simulations of PC coated with F-POSS-3 and F-POSS-5 units were also carried out under the same conditions. For these simulations, the initial configurations are obtained by building a three layer system composed of amorphous PC, first layer; F-POSS molecules, second layer and finally the water droplet for the third layer. Thus, the systems under study are PC/F-POSS-3 and PC/F-POSS-5 with nanoscale water droplet. Table 2 shows the calculated DLCA contact angle for the five different molecular systems considered here along with the tabulated value for PC.

Table 2: Molecular roughness in nanoscale using POSS molecules and DLCA calculated contact angles in nanoscale surfaces

Molecular System Contact Angle

F-POSS-3 103.5 F-POSS-4 86.2 F-POSS-5 62.6 F-POSS-3-hydrolyzed 103.2 Polycarbonate 80.6 Polycarbonate/F-POSS-3 88.2 Polycarbonate/F-POSS-5 97.1

The molecular systems can be divided into two groups, the one-layer systems, composed of F-POSS-3, F-POSS-4, F-POSS-5, and PC and the two-layer systems, composed of PC/F-POSS-3 and PC/F-POSS-5 molecules. Among the one-layer systems, F-POSS-5 is calculated to have the smallest contact angle, followed by polycarbonate, F-POSS-4, and F-POSS-3 with contact angles of 60.3, 80.6, 82.2 and 103.5 degree, respectively. The trend displayed by the DLCA contact angles for F-POSS coatings can be analyzed by using the

degree of fluorination and chain-length of the R group present in the (RSi-O3/2)n framework. For a

completely fluorinated group in F-POSS-3 (R= C6F13 as in Fig. 1A), the contact angle is highest and

hydrophobic (103.5 degree) for wetting by nanoscale water droplet. Somewhat less fluorinated

F-POSS-4 (R=C6F9H4) has the same chain-length as F-POSS-3 but a contact angle of 86.2 degree.

The transition from a hydrophobic to hydrophilic range (< 90 degree) can be attributed to the four

H-atoms as in Si-CH2-CH2- linkages in place of Si-CF2-CF2- linkages present in F-POSS-3. The

DLCA nanoscale-sensitive measurements thus can identify small changes in local atomistic

composition as reflected in the average DLCA contact angles. The F-POSS-5 simulations test if

increasing the length of the fluorinated domain in the R group from C6F9H4 to C10F17H4 will

increase surface. The DLCA contact angle from F-POSS-5 simulations shows (Table 2) the effects

of increasing surface roughness. According to Wenzel’s formula, increasing surface roughness for

hydrophilic surfaces should lead to increased wetting. F-POSS-5 shows an increased wetting as the

DLCA contact angle is calculated to be 62.6 degree, a decrease from F-POSS-4 contact angle of

86.2 degree. The trend is thus consistent with expected macroscopic behavior in contact angle for

the interface in a Wenzel domain. In Wenzel domain, the changing contact angle is governed by

cosθw= r cosθ, where the Wenzel angle (θw) is the surface roughness (r) modified Young’s equation

contact angle for ideally smooth surface.

polycarbonate (PC) matrix improved the hydroponic and self-cleaning properties of PC as reported by Dodiuk et al. [3]. They also reported that the PC surface roughness was augmented by means of silica and POSS layers. In addition, PC exhibits high light transmission properties (92 %). Thus, transparent PC substrates and PC/F-POSS polymer composites are an attractive system to benchmark the DLCA methods. To investigate the hydrophobic properties of polymer composites, MD simulations of PC coated with F-POSS-3 and F-POSS-5 units were also carried out under the same conditions. For these simulations, the initial configurations are obtained by building a three layer system composed of amorphous PC, first layer; F-POSS molecules, second layer and finally the water droplet for the third layer. Thus, the systems under study are PC/F-POSS-3 and PC/F-POSS-5 with nanoscale water droplet. Table 2 shows the calculated DLCA contact angle for the five different molecular systems considered here along with the tabulated value for PC.

Table 2: Molecular roughness in nanoscale using POSS molecules and DLCA calculated contact angles in nanoscale surfaces

Molecular System Contact Angle

F-POSS-3 103.5 F-POSS-4 86.2 F-POSS-5 62.6 F-POSS-3-hydrolyzed 103.2 Polycarbonate 80.6 Polycarbonate/F-POSS-3 88.2 Polycarbonate/F-POSS-5 97.1

The molecular systems can be divided into two groups, the one-layer systems, composed of F-POSS-3, F-POSS-4, F-POSS-5, and PC and the two-layer systems, composed of PC/F-POSS-3 and PC/F-POSS-5 molecules. Among the one-layer systems, F-POSS-5 is calculated to have the smallest contact angle, followed by polycarbonate, F-POSS-4, and F-POSS-3 with contact angles of 60.3, 80.6, 82.2 and 103.5 degree, respectively. The trend displayed by the DLCA contact angles for F-POSS coatings can be analyzed by using the

degree of fluorination and chain-length of the R group present in the (RSi-O3/2)n framework. For a

completely fluorinated group in F-POSS-3 (R= C6F13 as in Fig. 1A), the contact angle is highest and

hydrophobic (103.5 degree) for wetting by nanoscale water droplet. Somewhat less fluorinated

F-POSS-4 (R=C6F9H4) has the same chain-length as F-POSS-3 but a contact angle of 86.2 degree.

The transition from a hydrophobic to hydrophilic range (< 90 degree) can be attributed to the four

H-atoms as in Si-CH2-CH2- linkages in place of Si-CF2-CF2- linkages present in F-POSS-3. The

DLCA nanoscale-sensitive measurements thus can identify small changes in local atomistic

composition as reflected in the average DLCA contact angles. The F-POSS-5 simulations test if

increasing the length of the fluorinated domain in the R group from C6F9H4 to C10F17H4 will

increase surface. The DLCA contact angle from F-POSS-5 simulations shows (Table 2) the effects

of increasing surface roughness. According to Wenzel’s formula, increasing surface roughness for

hydrophilic surfaces should lead to increased wetting. F-POSS-5 shows an increased wetting as the

DLCA contact angle is calculated to be 62.6 degree, a decrease from F-POSS-4 contact angle of

86.2 degree. The trend is thus consistent with expected macroscopic behavior in contact angle for

the interface in a Wenzel domain. In Wenzel domain, the changing contact angle is governed by

cosθw= r cosθ, where the Wenzel angle (θw) is the surface roughness (r) modified Young’s equation

contact angle for ideally smooth surface.


The Wenzel type behavior described above changes when these nanoscale F-POSS domains are used as surface functionalization group in polycarbonate (PC) matrix. The predicted contact angle for PC, 80.6, is very close to the experimental value of 81.3 reported by Dodiuk et al.[3]. The PC/F-POSS-3 has a contact angle of 88.2 degree.For PC matrix with F-POSS-5, the average contact angles increasesfurther to a hydrophobic value (97.1). Thus,after coating the PC surface with F-POSS-3 and F-POSS-5, the contact angles increase by 7.6 and 16.4 degree, respectively.This trend is counterintuitive if we consider surface roughness (Wenzel type behavior) is the only controlling factor. If a Cassie-Baxter type behavior is followed for heterogeneous surfaces, there isa net increase in net hydrophobic fraction between F-POSS-3 and F-POSS-5. Thisis due to increase of hydrophobic chain lengths from (-CF2-)6in F-POSS-3 to (-CF2-)8in F-POSS-5. It is therefore reasonable to assume that a Cassie-Baxter type behavior is displayed by PC/F-POSS compositions. It is however hard to quantify the exact nature of the surface sampled by the moving boundary of the nanoscale water droplet.As the DLCA method samples all such hydrophobic and hydrophilic domains, an average measure of interfacial tension is provided. This can combine both roughness and inhomogeneity along as many as 500 contact line segments each with a local contact angle calculated during the dynamic simulations.Furthermore, the F-POSS nanoscale domains do not act in isolation. These domains interact with PC and should show a different behavior compared to F-POSS by themselves. As a result, although the trend of contact angles suggests a Cassie-Baxter type behavior, the role of surface roughness and changes in the PC interfacial tension cannot be completely ruled out. Further investigations are currently ongoing to identify the Cassie-Wenzel transition and the possibility of coexistence of Cassie and Wenzel type effects together in parts of the droplet. However, the strength of DLCA technique allows us to quantify the net effect on nanoscale wetting of rough, inhomogeneous coatings with variations in functional groups. To gain further insights into the optical properties of hydrophobic coatings composed of PC and (F-POSS) molecules, geometry optimization and electronic spectra calculations were performed at the same level of theory used for the (F-POSS) molecules, for the PC prototype dimer structure shown in Fig. 3.

Fig. 3: Optimized geometry of a PC dimer structure.

Fig. 4 shows the computed UV spectra for the PC dimer along with the computed spectra for the (A), (B), and (C) (F-POSS) molecules in Fig. 1.

The Wenzel type behavior described above changes when these nanoscale F-POSS domains are used as surface functionalization group in polycarbonate (PC) matrix. The predicted contact angle for PC, 80.6, is very close to the experimental value of 81.3 reported by Dodiuk et al.[3]. The PC/F-POSS-3 has a contact angle of 88.2 degree.For PC matrix with F-POSS-5, the average contact angles increasesfurther to a hydrophobic value (97.1). Thus,after coating the PC surface with F-POSS-3 and F-POSS-5, the contact angles increase by 7.6 and 16.4 degree, respectively.This trend is counterintuitive if we consider surface roughness (Wenzel type behavior) is the only controlling factor. If a Cassie-Baxter type behavior is followed for heterogeneous surfaces, there isa net increase in net hydrophobic fraction between F-POSS-3 and F-POSS-5. Thisis due to increase of hydrophobic chain lengths from (-CF2-)6in F-POSS-3 to (-CF2-)8in F-POSS-5. It is therefore reasonable to assume that a Cassie-Baxter type behavior is displayed by PC/F-POSS compositions. It is however hard to quantify the exact nature of the surface sampled by the moving boundary of the nanoscale water droplet.As the DLCA method samples all such hydrophobic and hydrophilic domains, an average measure of interfacial tension is provided. This can combine both roughness and inhomogeneity along as many as 500 contact line segments each with a local contact angle calculated during the dynamic simulations.Furthermore, the F-POSS nanoscale domains do not act in isolation. These domains interact with PC and should show a different behavior compared to F-POSS by themselves. As a result, although the trend of contact angles suggests a Cassie-Baxter type behavior, the role of surface roughness and changes in the PC interfacial tension cannot be completely ruled out. Further investigations are currently ongoing to identify the Cassie-Wenzel transition and the possibility of coexistence of Cassie and Wenzel type effects together in parts of the droplet. However, the strength of DLCA technique allows us to quantify the net effect on nanoscale wetting of rough, inhomogeneous coatings with variations in functional groups. To gain further insights into the optical properties of hydrophobic coatings composed of PC and (F-POSS) molecules, geometry optimization and electronic spectra calculations were performed at the same level of theory used for the (F-POSS) molecules, for the PC prototype dimer structure shown in Fig. 3.

Fig. 3: Optimized geometry of a PC dimer structure.

Fig. 4 shows the computed UV spectra for the PC dimer along with the computed spectra for the (A), (B), and (C) (F-POSS) molecules in Fig. 1.


Fig.4:Computed UV spectra of PC dimer using ab initio TDDFTmethods.

It is worth noting that the computed UV spectrum for PC is based on a dimer prototype structure and it is taken as a model for qualitative discussions. However, the PC UV experimental spectrum reported by Larosa et al. [18] exhibits a broader absorption feature than that of the calculated spectrum. This is mainly due to the approximation made by the current calculation, in which the prototype molecule is too small to capture all the contributions of the aromatic rings of the larger chains in the experimental sample. Nonetheless, the calculated spectrum captures the expected trends in UV absorption as more F-POSS domains are present in the coating. Therefore the calculated spectra shown in Fig. 4 can serve as a guide to predict the optical properties of coated samples of PC with (F-POSS) species for practical applications. Accordingly, a coated sample made of PC and 8C6F13-POSS (red trace in Fig. 4) species would exhibit the strongest UV absorption in the 150-280 nm range, whereas a sample made of PC and 8C6H4F9-POSS (blue trace in Fig. 4) will show the lowest combined absorption in the same range, where PC absorbance is the dominant. We can also generate UV absorption maps using these data for inhomogeneous coatings if surface compositions are known from experiments or simulations. Consequently, for transparent coatings with hydrophobic properties, the PC phase is vulnerable to upper atmosphere UV (so-called UVC region between 100-280 nm range in upper atmosphere solar spectrum) and thus fillers that can protect the coating in the same absorption range can be developed for stable formulations usable in aerospace applications. The predicted optical properties can be combined with estimations of contact angle to evaluate hydrophobic stability and design possible mechanism for dissipation of the absorbed energy. For the PC/ F-POSS model coatings, the DLCA method provided a consistentestimate of contact angles and predicted changes in the nanoscale wetting. The changes outlined in this work are not completely linear with the F-POSS nanoscale domains. The Si-O- cage and the functional group with hydrophobic tails interact with PC chainsand the net changes can be quantified using an atomistic simulations based method as presented in this work.The force-fields used can capture the time-averaged interactions between polymeric matrix, the nanophase (F-POSS) and the water nanodroplet. In addition, some well-known mechanisms of degradation can be incorporated in the nanoscale domains as we find that F-POSS-3 is a stronger upper atmosphere UV absorber (Fig. 4). The DLCA methodaddresses the concerns regarding identifying Cassie-Baxter and Wenzel domains implicitly, by providing the time-averaged contact angles based on a localized and faceted description of water contact lines on polymeric surfaces. In addition to nanoscale roughness, coatings prepared with similar material (e.g. PC/F-POSS-5) can be fabricated into superhydrophobic coatings if microscale roughness can be controlled during spin-coating or curing.

Fig.4:Computed UV spectra of PC dimer using ab initio TDDFTmethods.

It is worth noting that the computed UV spectrum for PC is based on a dimer prototype structure and it is taken as a model for qualitative discussions. However, the PC UV experimental spectrum reported by Larosa et al. [18] exhibits a broader absorption feature than that of the calculated spectrum. This is mainly due to the approximation made by the current calculation, in which the prototype molecule is too small to capture all the contributions of the aromatic rings of the larger chains in the experimental sample. Nonetheless, the calculated spectrum captures the expected trends in UV absorption as more F-POSS domains are present in the coating. Therefore the calculated spectra shown in Fig. 4 can serve as a guide to predict the optical properties of coated samples of PC with (F-POSS) species for practical applications. Accordingly, a coated sample made of PC and 8C6F13-POSS (red trace in Fig. 4) species would exhibit the strongest UV absorption in the 150-280 nm range, whereas a sample made of PC and 8C6H4F9-POSS (blue trace in Fig. 4) will show the lowest combined absorption in the same range, where PC absorbance is the dominant. We can also generate UV absorption maps using these data for inhomogeneous coatings if surface compositions are known from experiments or simulations. Consequently, for transparent coatings with hydrophobic properties, the PC phase is vulnerable to upper atmosphere UV (so-called UVC region between 100-280 nm range in upper atmosphere solar spectrum) and thus fillers that can protect the coating in the same absorption range can be developed for stable formulations usable in aerospace applications. The predicted optical properties can be combined with estimations of contact angle to evaluate hydrophobic stability and design possible mechanism for dissipation of the absorbed energy. For the PC/ F-POSS model coatings, the DLCA method provided a consistentestimate of contact angles and predicted changes in the nanoscale wetting. The changes outlined in this work are not completely linear with the F-POSS nanoscale domains. The Si-O- cage and the functional group with hydrophobic tails interact with PC chainsand the net changes can be quantified using an atomistic simulations based method as presented in this work.The force-fields used can capture the time-averaged interactions between polymeric matrix, the nanophase (F-POSS) and the water nanodroplet. In addition, some well-known mechanisms of degradation can be incorporated in the nanoscale domains as we find that F-POSS-3 is a stronger upper atmosphere UV absorber (Fig. 4). The DLCA methodaddresses the concerns regarding identifying Cassie-Baxter and Wenzel domains implicitly, by providing the time-averaged contact angles based on a localized and faceted description of water contact lines on polymeric surfaces. In addition to nanoscale roughness, coatings prepared with similar material (e.g. PC/F-POSS-5) can be fabricated into superhydrophobic coatings if microscale roughness can be controlled during spin-coating or curing.


Conclusion

In conclusion, the DLCA technique is well suited for nanoscale wetting problems and produces consistent values of contact angles in comparison to experimental values. These angles can be subsequently modified by surface roughness effects in microscale if nanoscale surface is considered smooth surface or Young’s equation contact angle values. Currently there is no predictive tool for addressing nano/microscale roughness in systematic way. The current simulations based method is a promising direction for coatings research. The success can be attributed to the COMPASS force field being successful in representing interfacial forces. In additionthe length scale is adequate anddynamics islong enough to produce accurate thermodynamic states even in nanoscale that leads to reasonable interfacial tension of solid-liquid interfaces reflected in the contact angles calculated using DLCA method. More importantly, the average contact angles can trace localized inhomogeneity as in the case of PDMS/silica and F-POSS/polycarbonate composites. The sensitivity of the method to molecular level changes is also excellent as we show that changes that are only two carbon chain long can alter the wetting behavior of the entire coating. The changes in local contact angle sampling frequency from the current 2 to1 ps per frame of MD simulation trajectory only varied contact angle by ±2 degrees. Therefore, the accuracy of DLCA technique is well suited to present the effect of molecular level roughness and heterogeneity at a local level as demonstrated for functionalize polycarbonate coatings. The simulations presented here also provide ways of designing new coating in a simulations-base environment as long as reliable force-fields for different polymeric and inorganic components are available or developed to represent interfacial forces (van der Waals and Coulombic forces) correctly. First-principles parameterized force-field such as COMPASS has the advantage in using values well represented in higher level theories for a consistent treatment of water interactions. It is thus expected that detailed approach with local contact angle effects and dynamic equilibrium effects can both be used using DLCA and similar techniques to address some important questions in coating design to protect metals from corrosion and UV degradation. Acknowledgements: SC acknowledges funding from Boeing Company, Office of Naval Research (Grant #N00014-04-1-0688and N00014-06-1-0315). Researchers at Boeing, Gerould Young, and Y. M. Gupta are thanked for encouragement and discussions.

References

[1] D. B. L. Cordes, P. D.; Rataboul, F., Chem. Rev. 110 (2010) p. 2081. [2] S. T. Iacono, S. M. Budy, J. M. Mabry, J. Smith, D. W., Macromolecules 40 (2007) p. 9517. [3] H. R. Dodiuk, P. F.; Kenig, S., Polym. Adv. Technol. 18 (2007) p. 746. [4] A. Tuteja, W. Choi, M. Ma, J. M. Mabry, S. A. Mazzella, G. C. Rutledge, G. H. McKinkley,

R. E. Cohen, Science 318 (2007) p. 1618. [5] D. Rigby, Fluid Phase Equilibria 217 (2004) p. 77. [6] H. Sun, J. Phys. Chem. B 102 (1998) p. 7338. [7] M. J. McQuaid, H. Sun, D. Rigby, J. Comput. Chem. 25 (2004) p. 61. [8] T. C. Ionescu, F. Qi, C. McCabe, A. Striolo, J. kieffer, P. T. Cummings, J. Phys. Chem. B

110 (2006) p. 2502. [9] S. Bizet, S. Galy, J.-F. Gerard, Polymer 47 (2006) p. 8219. [10] J. C. Cramer: Essentials of Computational Chemistry: Theories and Models, West Sussex,

England, 2002. [11] A. D. Becke, Phys. Rev. A 38 (1988) p. 3098. [12] A. D. Becke, J. Chem. Phys. 98 (1993) p. 5648. [13] C. Y. Lee, W.; Parr, R. G., Phys. Rev. B 37 (1988) p. 785. [14] R. E. S. Stratmann, G. E.; Frisch, M. L., J. Chem. Phys. 109 (1998) p. 8218. [15] E. G. Runge, E. K. U., Phys. Rev. Lett. 52 (1984) p. 997.

Conclusion

In conclusion, the DLCA technique is well suited for nanoscale wetting problems and produces consistent values of contact angles in comparison to experimental values. These angles can be subsequently modified by surface roughness effects in microscale if nanoscale surface is considered smooth surface or Young’s equation contact angle values. Currently there is no predictive tool for addressing nano/microscale roughness in systematic way. The current simulations based method is a promising direction for coatings research. The success can be attributed to the COMPASS force field being successful in representing interfacial forces. In additionthe length scale is adequate anddynamics islong enough to produce accurate thermodynamic states even in nanoscale that leads to reasonable interfacial tension of solid-liquid interfaces reflected in the contact angles calculated using DLCA method. More importantly, the average contact angles can trace localized inhomogeneity as in the case of PDMS/silica and F-POSS/polycarbonate composites. The sensitivity of the method to molecular level changes is also excellent as we show that changes that are only two carbon chain long can alter the wetting behavior of the entire coating. The changes in local contact angle sampling frequency from the current 2 to1 ps per frame of MD simulation trajectory only varied contact angle by ±2 degrees. Therefore, the accuracy of DLCA technique is well suited to present the effect of molecular level roughness and heterogeneity at a local level as demonstrated for functionalize polycarbonate coatings. The simulations presented here also provide ways of designing new coating in a simulations-base environment as long as reliable force-fields for different polymeric and inorganic components are available or developed to represent interfacial forces (van der Waals and Coulombic forces) correctly. First-principles parameterized force-field such as COMPASS has the advantage in using values well represented in higher level theories for a consistent treatment of water interactions. It is thus expected that detailed approach with local contact angle effects and dynamic equilibrium effects can both be used using DLCA and similar techniques to address some important questions in coating design to protect metals from corrosion and UV degradation. Acknowledgements: SC acknowledges funding from Boeing Company, Office of Naval Research (Grant #N00014-04-1-0688and N00014-06-1-0315). Researchers at Boeing, Gerould Young, and Y. M. Gupta are thanked for encouragement and discussions.

References

[1] D. B. L. Cordes, P. D.; Rataboul, F., Chem. Rev. 110 (2010) p. 2081. [2] S. T. Iacono, S. M. Budy, J. M. Mabry, J. Smith, D. W., Macromolecules 40 (2007) p. 9517. [3] H. R. Dodiuk, P. F.; Kenig, S., Polym. Adv. Technol. 18 (2007) p. 746. [4] A. Tuteja, W. Choi, M. Ma, J. M. Mabry, S. A. Mazzella, G. C. Rutledge, G. H. McKinkley,

R. E. Cohen, Science 318 (2007) p. 1618. [5] D. Rigby, Fluid Phase Equilibria 217 (2004) p. 77. [6] H. Sun, J. Phys. Chem. B 102 (1998) p. 7338. [7] M. J. McQuaid, H. Sun, D. Rigby, J. Comput. Chem. 25 (2004) p. 61. [8] T. C. Ionescu, F. Qi, C. McCabe, A. Striolo, J. kieffer, P. T. Cummings, J. Phys. Chem. B

110 (2006) p. 2502. [9] S. Bizet, S. Galy, J.-F. Gerard, Polymer 47 (2006) p. 8219. [10] J. C. Cramer: Essentials of Computational Chemistry: Theories and Models, West Sussex,

England, 2002. [11] A. D. Becke, Phys. Rev. A 38 (1988) p. 3098. [12] A. D. Becke, J. Chem. Phys. 98 (1993) p. 5648. [13] C. Y. Lee, W.; Parr, R. G., Phys. Rev. B 37 (1988) p. 785. [14] R. E. S. Stratmann, G. E.; Frisch, M. L., J. Chem. Phys. 109 (1998) p. 8218. [15] E. G. Runge, E. K. U., Phys. Rev. Lett. 52 (1984) p. 997.


[16] A. Frisch, R. D. Denington, II, T. A. Keith, A. B. Nielsen, A. J. Holder, GaussView 4.1. Gaussian, Inc., Wallingford, 2003.

[17] A. F. Frisch, M. J.; Trucks, G. W., Gaussian 03, Revision E.01. Gaussian, Inc., Wallingford, 2004.

[18] C. Larosa, E. Stura, R. Eggenhoeffner, C. Nicolini, Materials 2 (2009) p. 1193.

[16] A. Frisch, R. D. Denington, II, T. A. Keith, A. B. Nielsen, A. J. Holder, GaussView 4.1. Gaussian, Inc., Wallingford, 2003.

[17] A. F. Frisch, M. J.; Trucks, G. W., Gaussian 03, Revision E.01. Gaussian, Inc., Wallingford, 2004.

[18] C. Larosa, E. Stura, R. Eggenhoeffner, C. Nicolini, Materials 2 (2009) p. 1193.


TWO-DIMENSIONAL NUMERICAL MODELLING OF HYDROGEN

DIFFUSION IN METALS ASSISTED BY BOTH STRESS AND STRAIN

Jesús Toribio1, a, V. Kharin1,b, D. Vergara1,c and M. Lorenzo1,d 1Department of Materials Engineering, University of Salamanca

E.P.S., Campus Viriato, Avda. Requejo 33, 49022 Zamora, Spain Tel: (34-980) 54 50 00; Fax: (34-980) 54 50 02

[email protected],

[email protected],

[email protected],

[email protected]

Keywords: numerical modelling, weighted residual method, hydrogen diffusion, axisymmetric notch.

Abstract. The present work is based on previous research on the one-dimensional (1D) analysis of

the hydrogen diffusion process, and proposes a numerical approach of the same phenomenon in

two-dimensional (2D) situations, e.g. notches. The weighted residual method was used to solve

numerically the differential equations set out when the geometry was discretized through the

application of the finite element method. Three-node triangular elements were used in the

discretization, due to its simplicity, and a numerical algorithm was numerically implemented to

obtain the hydrogen concentration distribution in the material at different time increments. The

model is a powerful tool to analyze hydrogen embrittlement phenomena in structural materials.

Introduction

The influence of hydrogen on fracture depends on hydrogen concentration, C, in the sites where

localized material damage might occur. The accumulation of hydrogen in these zones proceeds by

diffusion from external or internal sources, i.e., local fracture event takes place when and where

hydrogen concentration reaches some critical value, Ccr, which is conditioned by the stress-strain

state in the material [1]. This is expressed by the following equation

Ccr = Ccr (σi, εi), (1)

which reflects the influence, in a general case, of the stress-strain state through its invariants, represented by the principal components of stresses and strains, σi and εi (i = 1, 2, 3) respectively.

Hydrogen diffusion within metals is also governed by the stress-strain state therein. Roughly, it

may be considered that hydrogen diffuses in metals obeying a Fick type diffusion law including

additional terms to account for the effect of the stress-strain state. Concerning the role of stress-

strain, this is commonly associated with its hydrostatic component σ and its equivalent plastic strain

εP, respectively. According to this, hydrogen diffuses not only to the points of minimum

concentration (driven by its gradient), but also to the sites of maximum hydrostatic stress (driven by

its gradient). The diffusion process may be also conditioned by the plastic strain: according to

previous studies [2,3] the role of this variable must be emphasized, since it controls hydrogen

accumulation in the material through its influence on both hydrogen solubility and diffusivity.

Following this way, i.e., considering the diffusion assisted by stress and strain as a responsible

transport mode, it is possible to evaluate the amount of hydrogen accumulated in metal, thereby

determining the locations where fracture initiation process might take place in the case of hydrogen

embrittlement phenomena. In order to improve an understanding of the diffusion phenomena inside

materials, and particularly the effects that stresses and strains exert on diffusion, it is useful to reveal

the time evolution of hydrogen concentration in the relevant sites concerning particular geometries

of the studied test-pieces or components, especially in the more representative locations.

TWO-DIMENSIONAL NUMERICAL MODELLING OF HYDROGEN

DIFFUSION IN METALS ASSISTED BY BOTH STRESS AND STRAIN

Jesús Toribio1, a, V. Kharin1,b, D. Vergara1,c and M. Lorenzo1,d 1Department of Materials Engineering, University of Salamanca

E.P.S., Campus Viriato, Avda. Requejo 33, 49022 Zamora, Spain Tel: (34-980) 54 50 00; Fax: (34-980) 54 50 02

[email protected],

[email protected],

[email protected],

[email protected]

Keywords: numerical modelling, weighted residual method, hydrogen diffusion, axisymmetric notch.

Abstract. The present work is based on previous research on the one-dimensional (1D) analysis of

the hydrogen diffusion process, and proposes a numerical approach of the same phenomenon in

two-dimensional (2D) situations, e.g. notches. The weighted residual method was used to solve

numerically the differential equations set out when the geometry was discretized through the

application of the finite element method. Three-node triangular elements were used in the

discretization, due to its simplicity, and a numerical algorithm was numerically implemented to

obtain the hydrogen concentration distribution in the material at different time increments. The

model is a powerful tool to analyze hydrogen embrittlement phenomena in structural materials.

Introduction

The influence of hydrogen on fracture depends on hydrogen concentration, C, in the sites where

localized material damage might occur. The accumulation of hydrogen in these zones proceeds by

diffusion from external or internal sources, i.e., local fracture event takes place when and where

hydrogen concentration reaches some critical value, Ccr, which is conditioned by the stress-strain

state in the material [1]. This is expressed by the following equation

Ccr = Ccr (σi, εi), (1)

which reflects the influence, in a general case, of the stress-strain state through its invariants, represented by the principal components of stresses and strains, σi and εi (i = 1, 2, 3) respectively.

Hydrogen diffusion within metals is also governed by the stress-strain state therein. Roughly, it

may be considered that hydrogen diffuses in metals obeying a Fick type diffusion law including

additional terms to account for the effect of the stress-strain state. Concerning the role of stress-

strain, this is commonly associated with its hydrostatic component σ and its equivalent plastic strain

εP, respectively. According to this, hydrogen diffuses not only to the points of minimum

concentration (driven by its gradient), but also to the sites of maximum hydrostatic stress (driven by

its gradient). The diffusion process may be also conditioned by the plastic strain: according to

previous studies [2,3] the role of this variable must be emphasized, since it controls hydrogen

accumulation in the material through its influence on both hydrogen solubility and diffusivity.

Following this way, i.e., considering the diffusion assisted by stress and strain as a responsible

transport mode, it is possible to evaluate the amount of hydrogen accumulated in metal, thereby

determining the locations where fracture initiation process might take place in the case of hydrogen

embrittlement phenomena. In order to improve an understanding of the diffusion phenomena inside

materials, and particularly the effects that stresses and strains exert on diffusion, it is useful to reveal

the time evolution of hydrogen concentration in the relevant sites concerning particular geometries

of the studied test-pieces or components, especially in the more representative locations.

To this end, numerous analyses [4-7] have focused various aspects of hydrogen diffusion in

metals affected by mechanical factors, such as stress or strain. Some analytical closed-form

solutions as well as numerical approaches have been developed under certain more or less

restraining assumptions or simplifications.

Concerning these latter, several previous studies were limited to considerations with one spatial

dimension of the sole diffusion depth and straight-line diffusion flux (1D situations), but with an

advantage of taking explicitly into account a lot of complicating factors which could arise during

nonsteady-state elastoplastic loading histories.

Several analyses dealt with two-dimensional (2D) situations, when diffusion was disturbed by

geometrical or stress-state inhomegeneities and proceeded along curvilinear trajectories, e.g. near

notches or cracks. However, these studies have been notably less extensive so far, in particular with

regard to the nonsteady-state stress-strain fields, e.g., slow strain-rate test conditions or cyclic

loading.

Therefore, considering the afore-said state of art with regard to hydrogen diffusion models, the

present paper aims to give a preliminary depiction of some advances towards the modeling of 2D

stress-strain assisted hydrogen diffusion under transient loading conditions.

According to [1] the driving force of the hydrogen diffusion can be expressed by means of the

gradient of hydrogen chemical potential as follows

HD µ−∇=X . (2)

And the latter can be expressed via the coefficient of solubility of hydrogen in the metal, Ks, that

represents the density of available sites for hydrogen,

=

S

H lnK

CRTµ , (3)

where R is the universal gas constant and T is the absolute temperature. The coefficient of solubility Ks depends on temperature, the level of hydrostatic stresses, σ, the microstructure of an alloy, its chemical and phase composition, and the density of hydrogen traps (traps for hydrogen in metals are formed by lattice imperfections: dislocations are, as a rule, the strongest type of traps but not the only one). The overall density of traps depends on the level of plastic strains, which may be represented via the second invariant of the plastic strain tensor: the effective or equivalent plastic strain εP. In addition, plastic strains may affect the phase composition of the alloy, thereby causing variations in the solubility of hydrogen, as in austenitic steels, via the strain-induced γ→α transformation. Thus, parallel with hydrostatic stresses and temperature, the level of plastic strains is also a variable affecting the solubility of hydrogen in metals, namely,

),,( PSεS TKK εσ= . (4)

This quantity can be represented as a product of the factor depending on the level of plastic strains εP by the function of hydrostatic stresses as follows (cf. [1]):

( )σεεσ Ω= exp),(),,( PSεPS TKTK , with RT

VH=Ω , (5)

where VH is the partial molar volume of hydrogen in metal.

The gradient of any solubility-affecting factor can induce a diffusion flux [8]. Under the assumption that the distribution of temperature in the solid is uniform (a hypothesis accepted throughout the present work), the diffusion flux J can be expressed as follows:

To this end, numerous analyses [4-7] have focused various aspects of hydrogen diffusion in

metals affected by mechanical factors, such as stress or strain. Some analytical closed-form

solutions as well as numerical approaches have been developed under certain more or less

restraining assumptions or simplifications.

Concerning these latter, several previous studies were limited to considerations with one spatial

dimension of the sole diffusion depth and straight-line diffusion flux (1D situations), but with an

advantage of taking explicitly into account a lot of complicating factors which could arise during

nonsteady-state elastoplastic loading histories.

Several analyses dealt with two-dimensional (2D) situations, when diffusion was disturbed by

geometrical or stress-state inhomegeneities and proceeded along curvilinear trajectories, e.g. near

notches or cracks. However, these studies have been notably less extensive so far, in particular with

regard to the nonsteady-state stress-strain fields, e.g., slow strain-rate test conditions or cyclic

loading.

Therefore, considering the afore-said state of art with regard to hydrogen diffusion models, the

present paper aims to give a preliminary depiction of some advances towards the modeling of 2D

stress-strain assisted hydrogen diffusion under transient loading conditions.

According to [1] the driving force of the hydrogen diffusion can be expressed by means of the

gradient of hydrogen chemical potential as follows

HD µ−∇=X . (2)

And the latter can be expressed via the coefficient of solubility of hydrogen in the metal, Ks, that

represents the density of available sites for hydrogen,

=

S

H lnK

CRTµ , (3)

where R is the universal gas constant and T is the absolute temperature. The coefficient of solubility Ks depends on temperature, the level of hydrostatic stresses, σ, the microstructure of an alloy, its chemical and phase composition, and the density of hydrogen traps (traps for hydrogen in metals are formed by lattice imperfections: dislocations are, as a rule, the strongest type of traps but not the only one). The overall density of traps depends on the level of plastic strains, which may be represented via the second invariant of the plastic strain tensor: the effective or equivalent plastic strain εP. In addition, plastic strains may affect the phase composition of the alloy, thereby causing variations in the solubility of hydrogen, as in austenitic steels, via the strain-induced γ→α transformation. Thus, parallel with hydrostatic stresses and temperature, the level of plastic strains is also a variable affecting the solubility of hydrogen in metals, namely,

),,( PSεS TKK εσ= . (4)

This quantity can be represented as a product of the factor depending on the level of plastic strains εP by the function of hydrostatic stresses as follows (cf. [1]):

( )σεεσ Ω= exp),(),,( PSεPS TKTK , with RT

VH=Ω , (5)

where VH is the partial molar volume of hydrogen in metal.

The gradient of any solubility-affecting factor can induce a diffusion flux [8]. Under the assumption that the distribution of temperature in the solid is uniform (a hypothesis accepted throughout the present work), the diffusion flux J can be expressed as follows:


∇−==

S

D

K

CDC

RT

XDCJ ln , (6)

where D is the diffusion coefficient of hydrogen in the metal. Actually, the diffusivity should be regarded not as a constant but as a function of the level of

plastic strains, i.e., εP, to reflect the influence of the variable phase composition or the density of traps on the averaged (microscopic) mobility of diffusion species in addition to their effect on the solubility of hydrogen reflected by eq. (4) and (5).

Substituting relation (5) in (6), the equation for the stress-strain assisted diffusion flux of hydrogen is obtained, according to which it is assumed that hydrogen diffusion through material proceeds toward the sites where occur the lowest concentration or the higher hydrostatic stress and the highest solubility.

The combination of these factors results in an equation for the stress-strain assisted diffusion flux of hydrogen which is:

( ) ( )( )

PSε

PSεP

∇+∇Ω−∇−=

εε

σεK

KCCDJ

(7)

Following the standard way, using the matter conservation law [9]:

Jdt

dCdiv−= , (8)

together with the expression (7) for the flux, the equation of diffusion in terms of the sole concentration can be obtained as

∇+∇Ω−∇⋅∇=

∂∂

ε

εσs

s

K

KDCCD

t

C (9)

In this formulation, neither stress state nor temperature (and thus, temperature dependent material

characteristics, such as D or Ω) are required to be stationary, but can be time dependent. To simplify

this preliminary study, diffusion coefficient, D, as well as temperature, are considered to be spatially

uniform, i.e., their gradients are zero, although this is not an essential restriction. This leads to the

known equation of hydrogen diffusion assisted by both stress and strain in terms of concentration:

)ln ln ( S

2

S

2 KCKCCDdt

dC∇−∇∇−∇= , (10)

where the coefficients may be time dependent.

The geometry of interest, selected here as an example to develop the analysis methodology, is

sketched in Fig. 1, which shows how the three-dimensional (3D) testpiece geometry can be analyzed

as a two-dimensional (2D) region (shaded figure) due to its axial symmetry.

The boundary conditions, both the mechanical and those for hydrogen entry into metal, are

depicted in Fig. 2. There an arbitrary axial load is applied over the boundary SL, a definite

environmentally controlled equilibrium concentration of hydrogen, Ceq, is imposed over the

boundary Seq exposed to the hydrogenating environment, and on both symmetry axes denoted Sf, the

null values of the hydrogen flux, J , and of the normal component of displacement are imposed.

∇−==

S

D

K

CDC

RT

XDCJ ln , (6)

where D is the diffusion coefficient of hydrogen in the metal. Actually, the diffusivity should be regarded not as a constant but as a function of the level of

plastic strains, i.e., εP, to reflect the influence of the variable phase composition or the density of traps on the averaged (microscopic) mobility of diffusion species in addition to their effect on the solubility of hydrogen reflected by eq. (4) and (5).

Substituting relation (5) in (6), the equation for the stress-strain assisted diffusion flux of hydrogen is obtained, according to which it is assumed that hydrogen diffusion through material proceeds toward the sites where occur the lowest concentration or the higher hydrostatic stress and the highest solubility.

The combination of these factors results in an equation for the stress-strain assisted diffusion flux of hydrogen which is:

( ) ( )( )

PSε

PSεP

∇+∇Ω−∇−=

εε

σεK

KCCDJ

(7)

Following the standard way, using the matter conservation law [9]:

Jdt

dCdiv−= , (8)

together with the expression (7) for the flux, the equation of diffusion in terms of the sole concentration can be obtained as

∇+∇Ω−∇⋅∇=

∂∂

ε

εσs

s

K

KDCCD

t

C (9)

In this formulation, neither stress state nor temperature (and thus, temperature dependent material

characteristics, such as D or Ω) are required to be stationary, but can be time dependent. To simplify

this preliminary study, diffusion coefficient, D, as well as temperature, are considered to be spatially

uniform, i.e., their gradients are zero, although this is not an essential restriction. This leads to the

known equation of hydrogen diffusion assisted by both stress and strain in terms of concentration:

)ln ln ( S

2

S

2 KCKCCDdt

dC∇−∇∇−∇= , (10)

where the coefficients may be time dependent.

The geometry of interest, selected here as an example to develop the analysis methodology, is

sketched in Fig. 1, which shows how the three-dimensional (3D) testpiece geometry can be analyzed

as a two-dimensional (2D) region (shaded figure) due to its axial symmetry.

The boundary conditions, both the mechanical and those for hydrogen entry into metal, are

depicted in Fig. 2. There an arbitrary axial load is applied over the boundary SL, a definite

environmentally controlled equilibrium concentration of hydrogen, Ceq, is imposed over the

boundary Seq exposed to the hydrogenating environment, and on both symmetry axes denoted Sf, the

null values of the hydrogen flux, J , and of the normal component of displacement are imposed.


Fig. 1. Analyzed geometry.

Fig. 2. Boundary conditions in an axisymmetric geometry.

The following expressions represent explicitly the mentioned boundary conditions for diffusion:

0 ff

SS==⋅ ΓJnJ

(11)

eqSeq CC = (12)

where n

is the external unit normal vector to the respective surface. For convenience in further numerical implementation of the diffusion boundary-value problem (10)-(12), the equilibrium equation (12) at the side of hydrogen entry from the environment is substituted by the next mass-exchange condition

)( eqSS eqeqCCJnJ −==⋅ Γ α

, (13)

where α is the mass-exchange rate coefficient which controls the velocity of approaching the equilibrium of hydrogen between environment and the entry surface layer Seq of a testpiece. To represent adequately the equilibrium entry condition (12) by means of relation (13), this rate coefficient may be chosen arbitrarily, but large enough to ensure practically instantaneous (with respect to the characteristic time scale for diffusion) attainment of the equilibrium given by equality (12). The adequacy of a choice can be easily confirmed a posteriori.

Finally, to finish the diffusion problem statement, hydrogen accumulation in the initially

hydrogen-free sample may be considered, so that the zero initial condition

00t

==

C (14)

will be placed throughout a whole testpiece of interest.

Fig. 1. Analyzed geometry.

Fig. 2. Boundary conditions in an axisymmetric geometry.

The following expressions represent explicitly the mentioned boundary conditions for diffusion:

0 ff

SS==⋅ ΓJnJ

(11)

eqSeq CC = (12)

where n

is the external unit normal vector to the respective surface. For convenience in further numerical implementation of the diffusion boundary-value problem (10)-(12), the equilibrium equation (12) at the side of hydrogen entry from the environment is substituted by the next mass-exchange condition

)( eqSS eqeqCCJnJ −==⋅ Γ α

, (13)

where α is the mass-exchange rate coefficient which controls the velocity of approaching the equilibrium of hydrogen between environment and the entry surface layer Seq of a testpiece. To represent adequately the equilibrium entry condition (12) by means of relation (13), this rate coefficient may be chosen arbitrarily, but large enough to ensure practically instantaneous (with respect to the characteristic time scale for diffusion) attainment of the equilibrium given by equality (12). The adequacy of a choice can be easily confirmed a posteriori.

Finally, to finish the diffusion problem statement, hydrogen accumulation in the initially

hydrogen-free sample may be considered, so that the zero initial condition

00t

==

C (14)

will be placed throughout a whole testpiece of interest.


Numerical approach

Obviously, quantitative modeling of stress-strain assisted hydrogen diffusion requires the stress-

strain field in a testpiece of interest to be known. Even for rather simple cases, such as a notched bar

being considered here, neither the exact solutions nor the closed form ones are usually available.

Thus, one must count on some sort of the numerical solution of the mechanical portion of the

coupled problem of the stress-strain assisted diffusion. The finite element method (FEM) approach,

well-developed for both linear and nonlinear analyses of deformable solid mechanics, is a right

choice to perform the stress analysis as a prerequisite for diffusion calculations.

In due course, simulation of diffusion accompanying mechanical loading of a sample also

requires numerical treatment. To this end, expansion of the FEM approach based on the same

framework of the spatial finite elements used in the stress and strain analysis is a natural choice for

numerical modelling of diffusion.

Some advanced general-purpose finite-element codes, well adapted for stress analysis, in

particular, such as ABAQUS or MSC.MARC, have certain capabilities to simulate the stress-strain

assisted diffusion, too. Unfortunately, they still are limited in some rather important aspects. With

regard to ABAQUS, this allows one to perform simulations of the stress-strain assisted diffusion

governed by equation (10) "over" the data of an accomplished solution of a geometrically and

physically nonlinear stress-strain analysis, i.e., for the stationary stress-strain field at the end of

some preliminary loading trajectory, but not for the case of simultaneous transient loading and

hydrogenation.

Another one, the MSC.MARC code with certain user subroutines may be employed to simulate

the transient stress-assisted diffusion as far as corresponding transport equation (10) has

mathematically the same form as the equation of convective heat transfer implemented in this

software, although, not for the geometrically nonlinear (large deformation) cases. In addition, it has

another rather strong shortcoming that requires the definition of the values of stress gradient at the

finite element nodes, which is accompanied by the accuracy loss in the displacement-based FEM

procedures. Indeed, using MSC.MARC the stresses per se are derived with an inevitable loss of

accuracy from the displacement gradients, so that the best approximation of stresses is achieved at

the element integration points, and calculation of the second derivatives of displacements must

substantially worsen the analysis accuracy.

With this in mind, it seems to be a reasonable compromise to consider a FEM implementation of

the modelling of stress-strain assisted diffusion over the previously (or simultaneously) performed

stress-strain analysis taking the nodal values of stresses and strains, obtained with a post-processing

technique, as the entry data for diffusion, i.e., constructing a finite-element approximation of the

stress-strain field with the aid of the same finite-element shape functions used in the mechanical

analysis to approximate the displacement fields.

Following this way, the standard weighted residuals procedure together with finite element

approximation of both fields of the hydrostatic stress σ and the hydrogen concentration C [10] may

be adopted to develop corresponding procedure for diffusion modeling coupled with the stress-

strain analysis.

Proceeding with the standard weighted residuals approach [10] to find an approximated solution

of the diffusion boundary-value problem (7)-(11) and (13), the approximation of concentration is

represented in terms of a linear combination of the spatial trial functions generated over a certain

sort of finite elements which discretise the solid under consideration, Ni(x), where x stands for the

instantaneous coordinates of material points over the volume of considered region V occupied by a

testpiece, so that

C(x,t) = Σ Ci(t)Ni(x), i = 1, 2, …, n (15)

where Ci are the nodal values of concentration, and the sum is taken over all the nodes of the finite

element discretisation. Then, the best approximation of the considered boundary-value problem will

Numerical approach

Obviously, quantitative modeling of stress-strain assisted hydrogen diffusion requires the stress-

strain field in a testpiece of interest to be known. Even for rather simple cases, such as a notched bar

being considered here, neither the exact solutions nor the closed form ones are usually available.

Thus, one must count on some sort of the numerical solution of the mechanical portion of the

coupled problem of the stress-strain assisted diffusion. The finite element method (FEM) approach,

well-developed for both linear and nonlinear analyses of deformable solid mechanics, is a right

choice to perform the stress analysis as a prerequisite for diffusion calculations.

In due course, simulation of diffusion accompanying mechanical loading of a sample also

requires numerical treatment. To this end, expansion of the FEM approach based on the same

framework of the spatial finite elements used in the stress and strain analysis is a natural choice for

numerical modelling of diffusion.

Some advanced general-purpose finite-element codes, well adapted for stress analysis, in

particular, such as ABAQUS or MSC.MARC, have certain capabilities to simulate the stress-strain

assisted diffusion, too. Unfortunately, they still are limited in some rather important aspects. With

regard to ABAQUS, this allows one to perform simulations of the stress-strain assisted diffusion

governed by equation (10) "over" the data of an accomplished solution of a geometrically and

physically nonlinear stress-strain analysis, i.e., for the stationary stress-strain field at the end of

some preliminary loading trajectory, but not for the case of simultaneous transient loading and

hydrogenation.

Another one, the MSC.MARC code with certain user subroutines may be employed to simulate

the transient stress-assisted diffusion as far as corresponding transport equation (10) has

mathematically the same form as the equation of convective heat transfer implemented in this

software, although, not for the geometrically nonlinear (large deformation) cases. In addition, it has

another rather strong shortcoming that requires the definition of the values of stress gradient at the

finite element nodes, which is accompanied by the accuracy loss in the displacement-based FEM

procedures. Indeed, using MSC.MARC the stresses per se are derived with an inevitable loss of

accuracy from the displacement gradients, so that the best approximation of stresses is achieved at

the element integration points, and calculation of the second derivatives of displacements must

substantially worsen the analysis accuracy.

With this in mind, it seems to be a reasonable compromise to consider a FEM implementation of

the modelling of stress-strain assisted diffusion over the previously (or simultaneously) performed

stress-strain analysis taking the nodal values of stresses and strains, obtained with a post-processing

technique, as the entry data for diffusion, i.e., constructing a finite-element approximation of the

stress-strain field with the aid of the same finite-element shape functions used in the mechanical

analysis to approximate the displacement fields.

Following this way, the standard weighted residuals procedure together with finite element

approximation of both fields of the hydrostatic stress σ and the hydrogen concentration C [10] may

be adopted to develop corresponding procedure for diffusion modeling coupled with the stress-

strain analysis.

Proceeding with the standard weighted residuals approach [10] to find an approximated solution

of the diffusion boundary-value problem (7)-(11) and (13), the approximation of concentration is

represented in terms of a linear combination of the spatial trial functions generated over a certain

sort of finite elements which discretise the solid under consideration, Ni(x), where x stands for the

instantaneous coordinates of material points over the volume of considered region V occupied by a

testpiece, so that

C(x,t) = Σ Ci(t)Ni(x), i = 1, 2, …, n (15)

where Ci are the nodal values of concentration, and the sum is taken over all the nodes of the finite

element discretisation. Then, the best approximation of the considered boundary-value problem will


be obtained with the nodal concentrations Ci which nullify all the residuals, which correspond to

both the differential equation (10) and the boundary conditions (11) and (13), with the weights

Wi(x), so that this yields the system of equations as follows

[ ] 0dV DC C DC D dV C

eqSfSSε

Sε =−−

∇−∇Ω−∇∇−− Γ

+∫∫∫ • JJW

K

KWW iii

VV

σ ∀i = 1, 2,…, n (16)

where • represents the inner product.

Adopting the Galerkin method as a particular form of weighted residuals, i.e., considering the

weights Wi to be the same as the trial functions Ni, after standard transformations of integrals in the

relation (16), the next system of the ordinary differential equations with respect to nodal

concentrations Ci(t) may be derived:

( )

0dS C dS

dV D dV

eqSfS eqS

eq

Sε

Sε

=

−++

∇∇−∇∇Ω−∇∇+

∫∫ ∫

∑ ∫∫

Γ

•••

ijiji

ijijijjjij

NdSNNCJN

K

KNNNNDNNDCNNC

j VV

αα

σ

(17)

or in the matrix form,

F C K C M =+ (18)

where the dot represents time derivative, and the boldface layout is used to denote matrices

(vectors).

Within the standard framework of development of the finite element procedures, considering the

region V subdivided with a set of finite elements Ve, that is V = ΣVe, corresponding global matrices

which appear in equation (18) are the result of the assembling of respective element matrices

defined as follows

∫=eV

ji dVNN em (19)

∫∫ +

∇∇∇∇Ω−∇∇= •••

eqS

ji

Sε

Sεjiji dS N N D-N N N N ασ dV

K

KNNDD

eV

ijek (20)

∫ ∫+−= Γ

fS

iN

eqS

ieq dS N C dS J αef (21)

where trial functions Ni acquire now the meaning of the corresponding element shape functions.

In these equations the stress-field is supposed to be known as a certain finite element

approximation with the use of the same trial functions (or element shape functions) of the form

similar to the one employed for the concentration (15), i.e.,

σ(x,t) = Σ σi(t)Ni(x), i= 1, 2, …, n (22)

εP(x,t) = Σ εPi(t)Ni(x), i= 1, 2, …, n (23)

be obtained with the nodal concentrations Ci which nullify all the residuals, which correspond to

both the differential equation (10) and the boundary conditions (11) and (13), with the weights

Wi(x), so that this yields the system of equations as follows

[ ] 0dV DC C DC D dV C

eqSfSSε

Sε =−−

∇−∇Ω−∇∇−− Γ

+∫∫∫ • JJW

K

KWW iii

VV

σ ∀i = 1, 2,…, n (16)

where • represents the inner product.

Adopting the Galerkin method as a particular form of weighted residuals, i.e., considering the

weights Wi to be the same as the trial functions Ni, after standard transformations of integrals in the

relation (16), the next system of the ordinary differential equations with respect to nodal

concentrations Ci(t) may be derived:

( )

0dS C dS

dV D dV

eqSfS eqS

eq

Sε

Sε

=

−++

∇∇−∇∇Ω−∇∇+

∫∫ ∫

∑ ∫∫

Γ

•••

ijiji

ijijijjjij

NdSNNCJN

K

KNNNNDNNDCNNC

j VV

αα

σ

(17)

or in the matrix form,

F C K C M =+ (18)

where the dot represents time derivative, and the boldface layout is used to denote matrices

(vectors).

Within the standard framework of development of the finite element procedures, considering the

region V subdivided with a set of finite elements Ve, that is V = ΣVe, corresponding global matrices

which appear in equation (18) are the result of the assembling of respective element matrices

defined as follows

∫=eV

ji dVNN em (19)

∫∫ +

∇∇∇∇Ω−∇∇= •••

eqS

ji

Sε

Sεjiji dS N N D-N N N N ασ dV

K

KNNDD

eV

ijek (20)

∫ ∫+−= Γ

fS

iN

eqS

ieq dS N C dS J αef (21)

where trial functions Ni acquire now the meaning of the corresponding element shape functions.

In these equations the stress-field is supposed to be known as a certain finite element

approximation with the use of the same trial functions (or element shape functions) of the form

similar to the one employed for the concentration (15), i.e.,

σ(x,t) = Σ σi(t)Ni(x), i= 1, 2, …, n (22)

εP(x,t) = Σ εPi(t)Ni(x), i= 1, 2, …, n (23)


where σi(t) and εPi(t) are, respectively, the known nodal stress and plastic strain values over a

prescribed loading history, which must be obtained on the phase of purely mechanical analysis. This

latter may be performed either simultaneously with or previously to the diffusion calculations, as far

as hydrogen diffusion is not supposed to affect the stress-strain state evolution in a solid.

Now, having reduced the diffusion boundary-problem to the system of first-order differential

equations (17) or (18), the solution of this latter along the t-axis may be undertaken with the aid of

the general easily programmable time-marching numerical scheme proposed for diffusion-type

equations elsewhere [11]. Limiting to the first-order approximation of the unknowns within every

single time interval, the nodal concentration values Cm–1 and Cm at the start and the end,

respectively, of the m-th time interval [tm–1,tm] are related as follows

( )( ) FC KKMCC =+∆∆+− −− 11 / mmm ttθ , (24)

where ∆t = tm – tm–1, and the constant θ must be chosen in a manner assuring the stability of this

time-marching scheme. Obviously, for the first time interval (at m = 1) the array C0 must be

determined according to the prescribed initial conditions of the analyzed problem. Then subsequent

values of Cm are to be found from (24) solving corresponding linear algebraic system by any

suitable means. In particular, in a symbolic form the matrix equation (24) renders the following

solution, also in matrix form

( ) ( ) tt mmm ∆−∆++= −−

− 1

1

1 C KFKMCC θ , (25)

which invites one to employ suitable algorithms of matrix inversion on this route. The described

procedure of time integration was proven to be unconditionally stable when the parameter θ is

between 0.5 and 1, i.e., θ ∈[0.5, 1].

This way, the numerical approach to the modeling of the stress-strain assisted diffusion is

determined in general terms. Its further transformation into a working practical code follows

established FEM procedures of element formulation (i.e., the choice of appropriate element

geometry, its shape functions, derivation of respective element matrices, which are involved in

equations (18), the use of numerical integration, etc). Since diffusion modeling under consideration

is planned as a supplementary one to a stress-strain analysis to be performed with the use of a

certain general purpose FEM code, it seems naturally to use the same spatial elements formulation

for both mechanical and diffusion phases whenever there appeared no specific reasons to change the

element type.

Concerning the described numerical approach, some final comments are worthy to be made.

Firstly, it is known [12] that strong accuracy deterioration problems may occur when Galerkin

method is applied to the transport equation (10), which is a kind of a convection-dominated

problem, and a mesh-related parameter called the Peclet number increases too much. In such cases

Petrov-Galerkin methods are considered to be a better choice. Fortunately, this complication has

never been approached in performed simulations with the magnitudes of governing material

parameters associated with common metal-hydrogen systems, considered geometries and loadings,

as well as reasonable finite element meshes.

Next, whenever diffusion is considered to proceed simultaneously along with a non-steady state

loading history, such as if slow strain rate tests or constant extension rate tests were simulated, the

stress-field is obviously time dependent, and so, the stress dependent element matrices do, too. In

addition to this fact, when large geometry changes occur, the deformed distances become the

diffusion paths of interest, so that coordinates x must be continuously updated with deformation

displacements, and thus, they also become time dependent. As a result, all the element matrices in

equations (18) must be updated throughout the simulation histories, i.e., they must be recalculated

on every time step of diffusion modeling. This makes the full-scale calculations extremely time

where σi(t) and εPi(t) are, respectively, the known nodal stress and plastic strain values over a

prescribed loading history, which must be obtained on the phase of purely mechanical analysis. This

latter may be performed either simultaneously with or previously to the diffusion calculations, as far

as hydrogen diffusion is not supposed to affect the stress-strain state evolution in a solid.

Now, having reduced the diffusion boundary-problem to the system of first-order differential

equations (17) or (18), the solution of this latter along the t-axis may be undertaken with the aid of

the general easily programmable time-marching numerical scheme proposed for diffusion-type

equations elsewhere [11]. Limiting to the first-order approximation of the unknowns within every

single time interval, the nodal concentration values Cm–1 and Cm at the start and the end,

respectively, of the m-th time interval [tm–1,tm] are related as follows

( )( ) FC KKMCC =+∆∆+− −− 11 / mmm ttθ , (24)

where ∆t = tm – tm–1, and the constant θ must be chosen in a manner assuring the stability of this

time-marching scheme. Obviously, for the first time interval (at m = 1) the array C0 must be

determined according to the prescribed initial conditions of the analyzed problem. Then subsequent

values of Cm are to be found from (24) solving corresponding linear algebraic system by any

suitable means. In particular, in a symbolic form the matrix equation (24) renders the following

solution, also in matrix form

( ) ( ) tt mmm ∆−∆++= −−

− 1

1

1 C KFKMCC θ , (25)

which invites one to employ suitable algorithms of matrix inversion on this route. The described

procedure of time integration was proven to be unconditionally stable when the parameter θ is

between 0.5 and 1, i.e., θ ∈[0.5, 1].

This way, the numerical approach to the modeling of the stress-strain assisted diffusion is

determined in general terms. Its further transformation into a working practical code follows

established FEM procedures of element formulation (i.e., the choice of appropriate element

geometry, its shape functions, derivation of respective element matrices, which are involved in

equations (18), the use of numerical integration, etc). Since diffusion modeling under consideration

is planned as a supplementary one to a stress-strain analysis to be performed with the use of a

certain general purpose FEM code, it seems naturally to use the same spatial elements formulation

for both mechanical and diffusion phases whenever there appeared no specific reasons to change the

element type.

Concerning the described numerical approach, some final comments are worthy to be made.

Firstly, it is known [12] that strong accuracy deterioration problems may occur when Galerkin

method is applied to the transport equation (10), which is a kind of a convection-dominated

problem, and a mesh-related parameter called the Peclet number increases too much. In such cases

Petrov-Galerkin methods are considered to be a better choice. Fortunately, this complication has

never been approached in performed simulations with the magnitudes of governing material

parameters associated with common metal-hydrogen systems, considered geometries and loadings,

as well as reasonable finite element meshes.

Next, whenever diffusion is considered to proceed simultaneously along with a non-steady state

loading history, such as if slow strain rate tests or constant extension rate tests were simulated, the

stress-field is obviously time dependent, and so, the stress dependent element matrices do, too. In

addition to this fact, when large geometry changes occur, the deformed distances become the

diffusion paths of interest, so that coordinates x must be continuously updated with deformation

displacements, and thus, they also become time dependent. As a result, all the element matrices in

equations (18) must be updated throughout the simulation histories, i.e., they must be recalculated

on every time step of diffusion modeling. This makes the full-scale calculations extremely time


consuming. To diminish this consumption of time, very expensive from the computational point of

view, some reasonable model reductions might be not only advisable, but also necessary.

Improvement of the workability of the developed modelling scheme

To proceed with simulations of stress-strain assisted diffusion with rather modest computational

facilities, it turned out to be indeed necessary to reduce the FEM-problem size. Among two possible

approaches, i.e., coarsening of the mesh of the modeled "full-scale" specimen or shrinking the

domain of diffusion simulation focusing on the locations of prospective hydrogen assisted fracture

initiation near the notch, the second one seems to be preferable. The relevant data about stress-strain

fields may be transferred to this domain from the full scale mechanical analyses, performing their

interpolation for the finite element mesh for diffusion, if convenient.

To succeed on this way, in the particular application case of the notched tensile specimens

analyzed in this paper, one may take an advantage of the fact that the notch can act as a localized

disturbance of the uniform stress-strain field in a smooth tensile specimen, depending on notch

parameters of depths and width, as displayed by the data of the hydrostatic stress distribution in

Fig. 3 and the equivalent plastic strain in Fig. 4.

Fig. 3. Hydrostatic stress distribution in an axisymmetric notched bar subjected to tension loading

indicating homogenization of the stress state away from the notch area (results obtained by means of

the commercial finite element code MSC.MARC).

Fig. 4. Equivalent plastic strain distribution in an axisymmetric notched bar subjected to tension

loading indicating homogenization of the stress state away from the notch area (results obtained by

means of the commercial finite element code MSC.MARC).

In addition, a notch may do the same with regard to the diffusion phenomenon from both the

points of view of the geometry and the stress-strain effects on the transport phenomenon, if

compared with the stress-strain unassisted diffusion in a smooth cylinder. In particular, the range of

consuming. To diminish this consumption of time, very expensive from the computational point of

view, some reasonable model reductions might be not only advisable, but also necessary.

Improvement of the workability of the developed modelling scheme

To proceed with simulations of stress-strain assisted diffusion with rather modest computational

facilities, it turned out to be indeed necessary to reduce the FEM-problem size. Among two possible

approaches, i.e., coarsening of the mesh of the modeled "full-scale" specimen or shrinking the

domain of diffusion simulation focusing on the locations of prospective hydrogen assisted fracture

initiation near the notch, the second one seems to be preferable. The relevant data about stress-strain

fields may be transferred to this domain from the full scale mechanical analyses, performing their

interpolation for the finite element mesh for diffusion, if convenient.

To succeed on this way, in the particular application case of the notched tensile specimens

analyzed in this paper, one may take an advantage of the fact that the notch can act as a localized

disturbance of the uniform stress-strain field in a smooth tensile specimen, depending on notch

parameters of depths and width, as displayed by the data of the hydrostatic stress distribution in

Fig. 3 and the equivalent plastic strain in Fig. 4.

Fig. 3. Hydrostatic stress distribution in an axisymmetric notched bar subjected to tension loading

indicating homogenization of the stress state away from the notch area (results obtained by means of

the commercial finite element code MSC.MARC).

Fig. 4. Equivalent plastic strain distribution in an axisymmetric notched bar subjected to tension

loading indicating homogenization of the stress state away from the notch area (results obtained by

means of the commercial finite element code MSC.MARC).

In addition, a notch may do the same with regard to the diffusion phenomenon from both the

points of view of the geometry and the stress-strain effects on the transport phenomenon, if

compared with the stress-strain unassisted diffusion in a smooth cylinder. In particular, the range of


the disturbing effect of a notch on stress-unassisted transport phenomena in solids can be estimated

from Fig. 5, where vanishing of the notch effect corresponds to fairly radial flow trajectories, or

concentration contour bands parallel to the cylinder surface, the same as it occurs in smooth bars.

Fig. 5. Diffusing specie distribution in a bar in the course of stress-strain-unassisted diffusion

(obtained with the finite element code MSC.MARC).

Thus, it follows that at some reasonable distance these both effects on hydrogen diffusion (of

notch geometry and non-uniform stresses and strains) vanish and diffusion becomes stress-strain-

notch unaffected. According to this, a reduced geometry can be considered, which includes the

region of stress-strain assisted and notch-affected transport. This zone of interest may be defined

from analyses of the FEM-solutions of the problems of mechanics about the stress-strain state in

notched bars and stress-strain unassisted transport, which may be obtained, e.g., with the aid of

whichever available FEM-code, such as examples displayed in Figs. 3 and 4. To this end it is only

necessary to substitute the rest of the specimen (the "remote" portion) by corresponding boundary

conditions derived, e.g., from the available closed-form solution of the transport problem for

smooth homogeneous cylinder, which may be found elsewhere [13]. If pertinent, the easily

generable notch- (but not stress and strain-) affected solution of the transport problem may be used

for this purpose in certain cases.

As an example, the reduced size domain to calculate the stress-strain affected distribution of

diffusible hydrogen in a particular notched specimen under tension loading, the same as considered

above, is verified as shown in Fig. 6, where corresponding boundary conditions are indicated. There

a mesh of linear triangular elements is employed, although this is not a matter of particular essence

with regard to the presented approach.

Fig. 6. Reduced size domain and boundary conditions for simulations of stress-strain assisted

diffusion.

the disturbing effect of a notch on stress-unassisted transport phenomena in solids can be estimated

from Fig. 5, where vanishing of the notch effect corresponds to fairly radial flow trajectories, or

concentration contour bands parallel to the cylinder surface, the same as it occurs in smooth bars.

Fig. 5. Diffusing specie distribution in a bar in the course of stress-strain-unassisted diffusion

(obtained with the finite element code MSC.MARC).

Thus, it follows that at some reasonable distance these both effects on hydrogen diffusion (of

notch geometry and non-uniform stresses and strains) vanish and diffusion becomes stress-strain-

notch unaffected. According to this, a reduced geometry can be considered, which includes the

region of stress-strain assisted and notch-affected transport. This zone of interest may be defined

from analyses of the FEM-solutions of the problems of mechanics about the stress-strain state in

notched bars and stress-strain unassisted transport, which may be obtained, e.g., with the aid of

whichever available FEM-code, such as examples displayed in Figs. 3 and 4. To this end it is only

necessary to substitute the rest of the specimen (the "remote" portion) by corresponding boundary

conditions derived, e.g., from the available closed-form solution of the transport problem for

smooth homogeneous cylinder, which may be found elsewhere [13]. If pertinent, the easily

generable notch- (but not stress and strain-) affected solution of the transport problem may be used

for this purpose in certain cases.

As an example, the reduced size domain to calculate the stress-strain affected distribution of

diffusible hydrogen in a particular notched specimen under tension loading, the same as considered

above, is verified as shown in Fig. 6, where corresponding boundary conditions are indicated. There

a mesh of linear triangular elements is employed, although this is not a matter of particular essence

with regard to the presented approach.

Fig. 6. Reduced size domain and boundary conditions for simulations of stress-strain assisted

diffusion.


CONCLUSIONS

• The numerical approach presented in this paper allows one to calculate the distribution of

hydrogen in stressed and strained solids with limited expenditure of computer resources.

• This model is considered to be useful to improve the knowledge of the role played by the factor

of hydrogen accumulation in prospective rupture sites by stress-strain assisted diffusion, which is

one of the key items in hydrogen embrittlement, a very dangerous phenomenon that frequently

accompanies metals and alloys in service.

• Proposed computational model seems to be a promising tool as an aid to develop the life-

prediction analyses for metallic components and structure subjected to any kind of hydrogen

embrittlement in service.

Acknowledgments

The authors wish to thank the financial support of their research at the University of Salamanca

provided by the following Spanish Institutions: Ministry for Science and Technology (MCYT;

Grant MAT2002-01831), Ministry for Education and Science (MEC; Grant BIA2005-08965),

Ministry for Science and Innovation (MCINN; Grant BIA2008-06810), Junta de Castilla y León

(JCyL; Grants SA067A05, SA111A07 and SA039A08).

References

[1] J. Toribio and V. Kharin: Nucl. Eng. Design Vol. 182 (1998), p. 149-163.

[2] J. Toribio, V. Kharin, D. Vergara, J.A. Blanco and J.G. Ballesteros: Corros. Sci. Vol. 49

(2007), p. 3557-3569.

[3] J. Toribio, M. Lorenzo, D. Vergara and V. Kharin, in: CORROSION 2010 Conference and

Expo, edited by NACE International, paper 10295 (2010), p. 1-13.

[4] P. Sofronis and R.M. McMeecking: J. Mech. Phys. Solids Vol. 37 (1989), p. 317-350.

[5] A.H.M Krom, R.W.J. Koers and A. Bakker: J. Mech. Phys. Solids Vol. 47 (1999), p. 971-992.

[6] A.H.M Krom, H.J. Maier, R.W.J. Koers and A. Bakker: Mater. Sci. Eng. Vol. A271 (1999),

p. 22-30.

[7] M. Wang, E. Akiyama and K. Tsuzaki: Scripta Mater. Vol. 53 (2005), p. 713-718.

[8] J. Toribio and V. Kharin: Fusion Eng. Design. Vol. 51-52 (2000), p. 213-218.

[9] P. Shewmon, in: Diffusion in Solids, TMS (1989).

[10] O.C. Zienkiewicz and K. Morgan: Finite Elements and Approximation, Wiley-Interscience

Publication (1983).

[11] O.C. Zienkiewicz, W.L. Wood, N.W. Hine and R.L. Taylor: Int. J. Num. Meth. Eng. Vol 20

(1984), p. 1529-1552.

[12] O.C. Zienkiewicz and R.L. Taylor, in: The Finite Element Method: Solids and Fluids

Mechanics, Dynamics and Non-Linearity, McGraw-Hill Book Company (1991).

[13] H.S. Carslaw and J.C. Jaegger, Conduction of Heat in Solids, Oxford Clarendon Press (1959).

CONCLUSIONS

• The numerical approach presented in this paper allows one to calculate the distribution of

hydrogen in stressed and strained solids with limited expenditure of computer resources.

• This model is considered to be useful to improve the knowledge of the role played by the factor

of hydrogen accumulation in prospective rupture sites by stress-strain assisted diffusion, which is

one of the key items in hydrogen embrittlement, a very dangerous phenomenon that frequently

accompanies metals and alloys in service.

• Proposed computational model seems to be a promising tool as an aid to develop the life-

prediction analyses for metallic components and structure subjected to any kind of hydrogen

embrittlement in service.

Acknowledgments

The authors wish to thank the financial support of their research at the University of Salamanca

provided by the following Spanish Institutions: Ministry for Science and Technology (MCYT;

Grant MAT2002-01831), Ministry for Education and Science (MEC; Grant BIA2005-08965),

Ministry for Science and Innovation (MCINN; Grant BIA2008-06810), Junta de Castilla y León

(JCyL; Grants SA067A05, SA111A07 and SA039A08).

References

[1] J. Toribio and V. Kharin: Nucl. Eng. Design Vol. 182 (1998), p. 149-163.

[2] J. Toribio, V. Kharin, D. Vergara, J.A. Blanco and J.G. Ballesteros: Corros. Sci. Vol. 49

(2007), p. 3557-3569.

[3] J. Toribio, M. Lorenzo, D. Vergara and V. Kharin, in: CORROSION 2010 Conference and

Expo, edited by NACE International, paper 10295 (2010), p. 1-13.

[4] P. Sofronis and R.M. McMeecking: J. Mech. Phys. Solids Vol. 37 (1989), p. 317-350.

[5] A.H.M Krom, R.W.J. Koers and A. Bakker: J. Mech. Phys. Solids Vol. 47 (1999), p. 971-992.

[6] A.H.M Krom, H.J. Maier, R.W.J. Koers and A. Bakker: Mater. Sci. Eng. Vol. A271 (1999),

p. 22-30.

[7] M. Wang, E. Akiyama and K. Tsuzaki: Scripta Mater. Vol. 53 (2005), p. 713-718.

[8] J. Toribio and V. Kharin: Fusion Eng. Design. Vol. 51-52 (2000), p. 213-218.

[9] P. Shewmon, in: Diffusion in Solids, TMS (1989).

[10] O.C. Zienkiewicz and K. Morgan: Finite Elements and Approximation, Wiley-Interscience

Publication (1983).

[11] O.C. Zienkiewicz, W.L. Wood, N.W. Hine and R.L. Taylor: Int. J. Num. Meth. Eng. Vol 20

(1984), p. 1529-1552.

[12] O.C. Zienkiewicz and R.L. Taylor, in: The Finite Element Method: Solids and Fluids

Mechanics, Dynamics and Non-Linearity, McGraw-Hill Book Company (1991).

[13] H.S. Carslaw and J.C. Jaegger, Conduction of Heat in Solids, Oxford Clarendon Press (1959).


Approach to Iron Corrosion via

the Numerical Simulation of a Galvanic Cell

COLICCHIO Giuseppina1, a, MANSUTTI Daniela2,b

and SANTARELLI Maria Laura3,c 1I.N.S.E.A.N., Via di Vallerano, 139 – 00128 Rome, Italy

2I.A.C. (C.N.R.), Via dei Taurini, 19 – 00186 Rome, Italy

3Dip. Ing. Chimica Materiali Ambiente, Sapienza Università di Roma,

Via Eudossiana, 18 – 00185 Rome, Italy


Keywords: Iron, redox reaction, kinetics, PDE, numerical simulation

Abstract. A mathematical model of the galvanic iron corrosion is, here, presented. The iron(III)-

hydroxide formation is considered together with the redox reaction. The PDE system, assembled on

the basis of the fundamental holding electro-chemistry laws, is numerically solved by a locally

refined FD method. For verification purpose we have assembled an experimental galvanic cell; in

the present work, we report two tests cases, with acidic and neutral electrolitical solution, where the

computed electric potential compares well with the measured experimental one

Introduction

Corrosion processes have been object of numerical simulations built upon 'ab initio' molecular

dynamics (for example [1], [2]); these ones have the limit of focussing on microscopic details while

missing the description of global macroscopic aspects and are extremely time consuming (105 hours

cpu time for 10-6sec. physical time).

Interesting descriptions have been developed also with inverse modelling approach, quantitative

descriptions of the corroded area elaborated on the basis of in situ electrostatic measurements (see,

for example, [3] and [4]).

The model, here, described provides Direct Numerical Simulation (DNS) of the iron corrosion

process as it can be represented via a galvanic cell with anode made of iron. In particular we bring to

reader’s attention an extension to the model presented by one author in [5], where iron(III)-oxyde

formation comes into play. In order to support the numerical results we have also built an

experimental galvanic cell and obtained a good match of the experimental measurements and of the

numerical results.

This paper is organized as follows: next section details the experimental cell, in the third section

the mathematical and numerical models are presented, the fourth section is devoted to a preliminary

numerical test and, in the fifth section, conclusions are drawn on the basis of the comparison of

numerical and experimental results.

Experimental Device

We focus on the galvanic corrosion of an iron sample (see Fig. 1) that may be represented as the

oxidation process occurring in a galvanic cell where the iron sample plays the role of the anode.

The experimental galvanic cell that we have monitored in laboratory is drawn at Fig.2: it consists of

an empty cylinder made of glass, closed at the two extrema by cylindric electrodes, the anode, made

of iron, and the cathode, made of platinum, typically inert material. The electrodes result 70 cm

Approach to Iron Corrosion via

the Numerical Simulation of a Galvanic Cell

COLICCHIO Giuseppina1, a, MANSUTTI Daniela2,b

and SANTARELLI Maria Laura3,c 1I.N.S.E.A.N., Via di Vallerano, 139 – 00128 Rome, Italy

2I.A.C. (C.N.R.), Via dei Taurini, 19 – 00186 Rome, Italy

3Dip. Ing. Chimica Materiali Ambiente, Sapienza Università di Roma,

Via Eudossiana, 18 – 00185 Rome, Italy


Keywords: Iron, redox reaction, kinetics, PDE, numerical simulation

Abstract. A mathematical model of the galvanic iron corrosion is, here, presented. The iron(III)-

hydroxide formation is considered together with the redox reaction. The PDE system, assembled on

the basis of the fundamental holding electro-chemistry laws, is numerically solved by a locally

refined FD method. For verification purpose we have assembled an experimental galvanic cell; in

the present work, we report two tests cases, with acidic and neutral electrolitical solution, where the

computed electric potential compares well with the measured experimental one

Introduction

Corrosion processes have been object of numerical simulations built upon 'ab initio' molecular

dynamics (for example [1], [2]); these ones have the limit of focussing on microscopic details while

missing the description of global macroscopic aspects and are extremely time consuming (105 hours

cpu time for 10-6sec. physical time).

Interesting descriptions have been developed also with inverse modelling approach, quantitative

descriptions of the corroded area elaborated on the basis of in situ electrostatic measurements (see,

for example, [3] and [4]).

The model, here, described provides Direct Numerical Simulation (DNS) of the iron corrosion

process as it can be represented via a galvanic cell with anode made of iron. In particular we bring to

reader’s attention an extension to the model presented by one author in [5], where iron(III)-oxyde

formation comes into play. In order to support the numerical results we have also built an

experimental galvanic cell and obtained a good match of the experimental measurements and of the

numerical results.

This paper is organized as follows: next section details the experimental cell, in the third section

the mathematical and numerical models are presented, the fourth section is devoted to a preliminary

numerical test and, in the fifth section, conclusions are drawn on the basis of the comparison of

numerical and experimental results.

Experimental Device

We focus on the galvanic corrosion of an iron sample (see Fig. 1) that may be represented as the

oxidation process occurring in a galvanic cell where the iron sample plays the role of the anode.

The experimental galvanic cell that we have monitored in laboratory is drawn at Fig.2: it consists of

an empty cylinder made of glass, closed at the two extrema by cylindric electrodes, the anode, made

of iron, and the cathode, made of platinum, typically inert material. The electrodes result 70 cm

apart and their diameter is 5 cm long; the ratio between these quantities allows to assume the

migration

Figure 1 – Sketch of an iron corrosion occurrence

of the produced ions from one to the other electrode being negligible. Also the gravity effect is

reduced by placing the cell on a horizontal plane. The electrodes are connected via a conducting

wire outside of the cell where a potenziometer is placed in order to measure the electric potential

gradient developed during the redox reactions.

Figure 2 – Sketch of the experimental galvanic cell

It is known that main chemical reactions occurring in the galvanic cell are the following [7]:

at the anode,

(Ra1) Fe Fe2+

+ 2e- Fe

3+ +3e

- (two- step oxidation)

V

Anode (-)

Iron Cathode (+)

Platinum

- +

Potentiometer

Water

apart and their diameter is 5 cm long; the ratio between these quantities allows to assume the

migration

Figure 1 – Sketch of an iron corrosion occurrence

of the produced ions from one to the other electrode being negligible. Also the gravity effect is

reduced by placing the cell on a horizontal plane. The electrodes are connected via a conducting

wire outside of the cell where a potenziometer is placed in order to measure the electric potential

gradient developed during the redox reactions.

Figure 2 – Sketch of the experimental galvanic cell

It is known that main chemical reactions occurring in the galvanic cell are the following [7]:

at the anode,

(Ra1) Fe Fe2+

+ 2e- Fe

3+ +3e

- (two- step oxidation)

V

Anode (-)

Iron Cathode (+)

Platinum

- +

Potentiometer

Water


at the cathode,

(Rc1) 2H+

+ 2e- H2 (reduction in acidic environment)

(Rc2) O2 + H2O + 4e- 4OH

- (reduction in natural environment)

in the electrolyte solution,

(Rb1 )Fe3+

+ 3OH- Fe(OH)3 (towards rust formation)

(Rb2) H2O H+ + OH

- (water molecule dissociation law)

(Rb3) HA H+ + A

- (if acidic electrolyte solution)

being Fe2+

,the ferrous ion, and Fe

3+, the ferric ion, OH

- , the hydroxyl ion, Fe(OH)3, the iron (III)

hydroxide (rust precursor).

In the following, two tests cases in the above laboratory cell will be presented, one with acidic

electrolyte solution and one with neutral electrolyte solution.

Mathematical Model

In this paragraph we describe the revised version of a model that we have presented at the

conference of Società Italiana per la Matematica Applicata e Industriale, 2008 [9].

The geometrical shape of the described experimental cell is such that the unidimensional

approximation appears appropriate. Let us fix the origin of the reference frame at the anode and

choose the (x) axis parallel to the symmetry axis of the cylindrical structure.

The chemical species involved in the process are: Fe3+

, H+, OH

-, Fe(OH)3, A

-. Let Ck = Ck(t,x),

k=1,.., 5, be respectively their concentration (moles/litre) at time t at the abscissa x.

We have represented the reactions listed at the previous section by imposing, in the bulk, the mass

conservation law, the condition of electro-neutrality of the system and the rate law of

absorption/production of chemical species during a chemical reaction and by imposing, at the

electrodes, the Butler-Volmer equation or the Hurd equation, where it applies, for the current

density produced by redox reactions [6, 7].

Neglecting the convective motion of the electrolyte, and under the assumption that concentrations

are small, the mass flux kJ of the k-th species is given, with a good approximation, by the Planck-

Nernst law [7]:

φφφφ∇∇∇∇−−−−∇∇∇∇−−−−====k

kk

kkkC

RT

FDzCDJ (1)

where Dk is the diffusion constant, zk the charge number, F Faraday’s constant, φ the electric potential in the electrolyte, R the gas constant and T the absolute temperature. The law (1)

represents the fact that the ions are transported by migration within the electric field and by

molecular diffusion.

For each species, by mass conservation, the following transport equation can be written in the

electrolyte solution space:

kk

k SJt

C=⋅∇+

∂

∂ (2)

where Sk is a source term that takes into account the production (or absorption) of ions of the k-th

species due to chemical reactions (rate law), as, for example, in the case of the ions Fe3+ that

contributes to the formation of rust. For a binary reaction occurring between species i and j, on

phenomenological basis, the rate law states:

0, if k-th species is neither a produced nor a reacting species

Sk = kij (Ci)νi

( Cj)νj

, if k-th is a produced species

νκ kik (Ci)νi

( Ck)νk

, if k-th is a reacting species

at the cathode,

(Rc1) 2H+

+ 2e- H2 (reduction in acidic environment)

(Rc2) O2 + H2O + 4e- 4OH

- (reduction in natural environment)

in the electrolyte solution,

(Rb1 )Fe3+

+ 3OH- Fe(OH)3 (towards rust formation)

(Rb2) H2O H+ + OH

- (water molecule dissociation law)

(Rb3) HA H+ + A

- (if acidic electrolyte solution)

being Fe2+

,the ferrous ion, and Fe

3+, the ferric ion, OH

- , the hydroxyl ion, Fe(OH)3, the iron (III)

hydroxide (rust precursor).

In the following, two tests cases in the above laboratory cell will be presented, one with acidic

electrolyte solution and one with neutral electrolyte solution.

Mathematical Model

In this paragraph we describe the revised version of a model that we have presented at the

conference of Società Italiana per la Matematica Applicata e Industriale, 2008 [9].

The geometrical shape of the described experimental cell is such that the unidimensional

approximation appears appropriate. Let us fix the origin of the reference frame at the anode and

choose the (x) axis parallel to the symmetry axis of the cylindrical structure.

The chemical species involved in the process are: Fe3+

, H+, OH

-, Fe(OH)3, A

-. Let Ck = Ck(t,x),

k=1,.., 5, be respectively their concentration (moles/litre) at time t at the abscissa x.

We have represented the reactions listed at the previous section by imposing, in the bulk, the mass

conservation law, the condition of electro-neutrality of the system and the rate law of

absorption/production of chemical species during a chemical reaction and by imposing, at the

electrodes, the Butler-Volmer equation or the Hurd equation, where it applies, for the current

density produced by redox reactions [6, 7].

Neglecting the convective motion of the electrolyte, and under the assumption that concentrations

are small, the mass flux kJ of the k-th species is given, with a good approximation, by the Planck-

Nernst law [7]:

φφφφ∇∇∇∇−−−−∇∇∇∇−−−−====k

kk

kkkC

RT

FDzCDJ (1)

where Dk is the diffusion constant, zk the charge number, F Faraday’s constant, φ the electric potential in the electrolyte, R the gas constant and T the absolute temperature. The law (1)

represents the fact that the ions are transported by migration within the electric field and by

molecular diffusion.

For each species, by mass conservation, the following transport equation can be written in the

electrolyte solution space:

kk

k SJt

C=⋅∇+

∂

∂ (2)

where Sk is a source term that takes into account the production (or absorption) of ions of the k-th

species due to chemical reactions (rate law), as, for example, in the case of the ions Fe3+ that

contributes to the formation of rust. For a binary reaction occurring between species i and j, on

phenomenological basis, the rate law states:

0, if k-th species is neither a produced nor a reacting species

Sk = kij (Ci)νi

( Cj)νj

, if k-th is a produced species

νκ kik (Ci)νi

( Ck)νk

, if k-th is a reacting species


where kij, νi are respectively the rate constant of the reaction between species i-th and species j-th

(experimentally measured) and the stoichiometric coefficient of the i-th species within the

considered reaction.

However we have to bear in mind also the law of dissociation of the water molecule that states the

permanent well known balance between the concentrations of the hydrogen ions and the hydroxyl

ions; actually, this balance allows to compute the concentration of one of them once the other is

known:

14ClogClog32

====−−−−−−−− .

We choose to compute C3 by solving the (simple!) nonlinear equation (3) in place of the related

equation in system (2).

It should be noticed that, once the expression (1) is substituted in equation (2), this one takes the

form of a convection-diffusion transport equation, where the transport velocity is proportional to

φφφφ∇∇∇∇ . In addition an equation for the electric potential φ can be obtained considering the total electric current and imposing the electrical neutrality of the electrolyte. In fact the electrolyte is electrically

neutral apart from the very thin double layers adjacent to the electrodes. The electric current density

i in the electrolyte is given in terms of the mass fluxes of ions by the Faraday’s law:

k

k

k JzFi ∑= . (4)

A statement of electro-neutrality is that the electric current density is non-divergent, that is:

0=⋅∇ i . (5)

Inserting the expression (4) into equation (5) and taking into account the Planck-Nernst law (1),

we obtain the equation for the electric potential:

∑∑∑∑∑∑∑∑ ∇∇∇∇−−−−====

∇∇∇∇

⋅⋅⋅⋅∇∇∇∇

k

k

2

kk

k

k

k

2

k CDzCRT

FDzφφφφ (6)

Let us notice that when the concentration gradients are negligible, equation (6) reduces to the

Laplace’s equation, 02 =∇ φ .

Boundary conditions. At the anode and at the cathode, the redox reactions release respectively

the following current densities:

∑∑∑∑ ==== −−−− ++++−−−−

/(/(/(/(−−−−−−−−====

2

1k kk1k

k0

k

a

)RT

F)nz(exp(

i

n

)RTF3exp(1zi

ηηηηαααα

ηηηη

(7)

(3)

where kij, νi are respectively the rate constant of the reaction between species i-th and species j-th

(experimentally measured) and the stoichiometric coefficient of the i-th species within the

considered reaction.

However we have to bear in mind also the law of dissociation of the water molecule that states the

permanent well known balance between the concentrations of the hydrogen ions and the hydroxyl

ions; actually, this balance allows to compute the concentration of one of them once the other is

known:

14ClogClog32

====−−−−−−−− .

We choose to compute C3 by solving the (simple!) nonlinear equation (3) in place of the related

equation in system (2).

It should be noticed that, once the expression (1) is substituted in equation (2), this one takes the

form of a convection-diffusion transport equation, where the transport velocity is proportional to

φφφφ∇∇∇∇ . In addition an equation for the electric potential φ can be obtained considering the total electric current and imposing the electrical neutrality of the electrolyte. In fact the electrolyte is electrically

neutral apart from the very thin double layers adjacent to the electrodes. The electric current density

i in the electrolyte is given in terms of the mass fluxes of ions by the Faraday’s law:

k

k

k JzFi ∑= . (4)

A statement of electro-neutrality is that the electric current density is non-divergent, that is:

0=⋅∇ i . (5)

Inserting the expression (4) into equation (5) and taking into account the Planck-Nernst law (1),

we obtain the equation for the electric potential:

∑∑∑∑∑∑∑∑ ∇∇∇∇−−−−====

∇∇∇∇

⋅⋅⋅⋅∇∇∇∇

k

k

2

kk

k

k

k

2

k CDzCRT

FDzφφφφ (6)

Let us notice that when the concentration gradients are negligible, equation (6) reduces to the

Laplace’s equation, 02 =∇ φ .

Boundary conditions. At the anode and at the cathode, the redox reactions release respectively

the following current densities:

∑∑∑∑ ==== −−−− ++++−−−−

/(/(/(/(−−−−−−−−====

2

1k kk1k

k0

k

a

)RT

F)nz(exp(

i

n

)RTF3exp(1zi

ηηηηαααα

ηηηη

(7)

(3)


where i0k is the exchange current density of the k-th reaction (experimental), nk is the electron

stoichiometric coefficient at the k-th step, α (and ακ) is the transfer coefficient (experimental), zk

is the charge number of the k-th ion species and η is the overvoltage produced by the anodic reactions, being ηηηη = E−−−−δφδφδφδφ , with δφ , the variation of potential between the electrode and the

electrolyte (soon after the electronic double layer) and E , the electrode potential at zero current.

The last quantity is given as a function of the standard potential E0 using the following Nernst

equation:

j

j

j aFz

RTEE ∑+= ln0 ν ,

where the summation is extended to all the species participating at the electrode reactions, jν are

the stoichiometric coefficients and aj are the activities of the ions. In the hypothesis of a diluted

solution with a large excess of solvent, as we suppose in the present case, the activities of the ions

can be approximated with the molar concentration Cj.

The overvoltages ηk (k=2,3), attaining to the cathodic reactions, are obtained by ηηηη k= E−−−−δφδφδφδφ k, with

where we can derive the meaning of the symbols from the above expression of E.

The expression (7) is known as the Hurd equation for the two-step oxidation of iron and the

expressions in (8) are known as Butler-Volmer equations (ic2, the current density released by H+

reduction (Rc1) and, ic3, the current density released by OH- formation (Rc2)). Then, boundary

conditions are expressed in terms of the current density associated to each chemical species:

at the anode (x = 0), F z1 J1 = ia,, J2 = J4 = J5 = 0

at the cathode (x = xmax), F z2 J2 = ic2 , J1 = J4 =J5 = 0.

Moreover the electro-neutrality of the system (we are considering a galvanic cell) has still to be

considered; we express it by forcing the current densities at the electrodes to be equal:

ia = ic . (9)

We observe that via this condition the presence of O2 within the cathodic reactions enhances the

anodic iron dissolution as it is expected to be.

Initial conditions. The initial conditions are dependent only on the pH of the electrolitical

solution. According to the dissociation law of the water molecule and to the definition of pH, we

have:

,ClogFz

RTEE kk

0

kk νννν++++====

(((( ))))

(((( ))))

−−−−====

−−−−====

++++====

3

3

033c

2

2

022c

3c2cc

RT

Fz1expii

RT

Fz1expii

with,iii

ηηηηαααα

ηηηηαααα (8)

where i0k is the exchange current density of the k-th reaction (experimental), nk is the electron

stoichiometric coefficient at the k-th step, α (and ακ) is the transfer coefficient (experimental), zk

is the charge number of the k-th ion species and η is the overvoltage produced by the anodic reactions, being ηηηη = E−−−−δφδφδφδφ , with δφ , the variation of potential between the electrode and the

electrolyte (soon after the electronic double layer) and E , the electrode potential at zero current.

The last quantity is given as a function of the standard potential E0 using the following Nernst

equation:

j

j

j aFz

RTEE ∑+= ln0 ν ,

where the summation is extended to all the species participating at the electrode reactions, jν are

the stoichiometric coefficients and aj are the activities of the ions. In the hypothesis of a diluted

solution with a large excess of solvent, as we suppose in the present case, the activities of the ions

can be approximated with the molar concentration Cj.

The overvoltages ηk (k=2,3), attaining to the cathodic reactions, are obtained by ηηηη k= E−−−−δφδφδφδφ k, with

where we can derive the meaning of the symbols from the above expression of E.

The expression (7) is known as the Hurd equation for the two-step oxidation of iron and the

expressions in (8) are known as Butler-Volmer equations (ic2, the current density released by H+

reduction (Rc1) and, ic3, the current density released by OH- formation (Rc2)). Then, boundary

conditions are expressed in terms of the current density associated to each chemical species:

at the anode (x = 0), F z1 J1 = ia,, J2 = J4 = J5 = 0

at the cathode (x = xmax), F z2 J2 = ic2 , J1 = J4 =J5 = 0.

Moreover the electro-neutrality of the system (we are considering a galvanic cell) has still to be

considered; we express it by forcing the current densities at the electrodes to be equal:

ia = ic . (9)

We observe that via this condition the presence of O2 within the cathodic reactions enhances the

anodic iron dissolution as it is expected to be.

Initial conditions. The initial conditions are dependent only on the pH of the electrolitical

solution. According to the dissociation law of the water molecule and to the definition of pH, we

have:

,ClogFz

RTEE kk

0

kk νννν++++====

(((( ))))

(((( ))))

−−−−====

−−−−====

++++====

3

3

033c

2

2

022c

3c2cc

RT

Fz1expii

RT

Fz1expii

with,iii

ηηηηαααα

ηηηηαααα (8)


14loglog 1010 =+=−− −+ pOHpHCCOHH

;

and for pure water it is pH=7, that is both ion concentrations equal 10–7

moles/litre.

When an acid is dissolved in water, it dissociates in ions according to the following formulas:

−+ +→ AHHA and ,litre/moles10CC

pH

AH

−−−−======== −−−−++++

so the concentrations of +H ions increases and the pH of the solution decreases. Concerning the

initial concentration COH -=10

–(14-pH) , it is considered negligible versus the electro-neutrality of the

solution. The initial concentrations of the other species (k=1,4) are assumed to be null.

Numerical procedure. The system of equations (2) and equation (6), completed by condition

(9) and initial and boundary conditions, is numerically solved by a finite difference discretization.

We have adopted a stretched distribution of nodes towards the electrodes (according to function

sech(.)) in order to follow better the rapid increase (at the anode) and decrease (at the cathode) of

reacting ionic species. The assignment of the discrete unknowns on the mesh is staggered, with

concentrations and electric potential in the middle of a cell and the fluxes Jk at the extrema. We

have chosen second order time and space approximation schemes. In particular time integration has

been accomplished by a predictor-corrector methods based on the mid-point Euler scheme. The

splitting of main operators (for mass conservation and for electro-neutrality) allows straightforward

numerical solution. For the sake of completeness, it has to be noticed that, as the electric potential

appears in the model just through its gradient, the value of φ at one point x had to be arbitrarily

chosen.

The numerical solutions presented in the following sections are obtained with space and time steps

respectively ∆x=xmax /50 and ∆t=10-3sec. For the values of the electrochemical constants

appearing in the model we refer the reader to [8].

Preliminary Numerical Test.

In this section, just to get a qualitative insight to the mathematical numerical model performance,

we show the numerical results for a preliminary test with weakly acidic electrolitical solution

(ph = 6). Let us observe, first, that (Fig. 3) the concentration of Fe3+ is almost negligible but, yet,

Figure 3 – Ferric ion concentration spacial distribution at t = 4.7 h (left) and

its time evolution at the neighborhood of the anode (right)

14loglog 1010 =+=−− −+ pOHpHCCOHH

;

and for pure water it is pH=7, that is both ion concentrations equal 10–7

moles/litre.

When an acid is dissolved in water, it dissociates in ions according to the following formulas:

−+ +→ AHHA and ,litre/moles10CC

pH

AH

−−−−======== −−−−++++

so the concentrations of +H ions increases and the pH of the solution decreases. Concerning the

initial concentration COH -=10

–(14-pH) , it is considered negligible versus the electro-neutrality of the

solution. The initial concentrations of the other species (k=1,4) are assumed to be null.

Numerical procedure. The system of equations (2) and equation (6), completed by condition

(9) and initial and boundary conditions, is numerically solved by a finite difference discretization.

We have adopted a stretched distribution of nodes towards the electrodes (according to function

sech(.)) in order to follow better the rapid increase (at the anode) and decrease (at the cathode) of

reacting ionic species. The assignment of the discrete unknowns on the mesh is staggered, with

concentrations and electric potential in the middle of a cell and the fluxes Jk at the extrema. We

have chosen second order time and space approximation schemes. In particular time integration has

been accomplished by a predictor-corrector methods based on the mid-point Euler scheme. The

splitting of main operators (for mass conservation and for electro-neutrality) allows straightforward

numerical solution. For the sake of completeness, it has to be noticed that, as the electric potential

appears in the model just through its gradient, the value of φ at one point x had to be arbitrarily

chosen.

The numerical solutions presented in the following sections are obtained with space and time steps

respectively ∆x=xmax /50 and ∆t=10-3sec. For the values of the electrochemical constants

appearing in the model we refer the reader to [8].

Preliminary Numerical Test.

In this section, just to get a qualitative insight to the mathematical numerical model performance,

we show the numerical results for a preliminary test with weakly acidic electrolitical solution

(ph = 6). Let us observe, first, that (Fig. 3) the concentration of Fe3+ is almost negligible but, yet,

Figure 3 – Ferric ion concentration spacial distribution at t = 4.7 h (left) and



Figure 4 – Iron(III) hydroxide concentration spacial distribution at t = 4.7 h (left) and


Figure 5 – Hydrogen ion concentration spacial distribution at t = 4.7 h (left) and

current density evolution (right)

reasonably behaving: in the vicinity of the anode (right plot), it increases rapidly at initial time,

while, after, it approaches a sort of passivation regime; moreover, it concentrates around the anode,

where rust, eventually, would grow, whereas it is completely null elsewhere in the cell (left plot).

The same distribution is, coherently, observed for Iron(III) hydroxide (Fig. 4, left plot), which, in

addition, grows indefinitely in the vicinity of the anode (right plot) due to the lack of the mechanism

of formation of rust in the mathematical model. In Fig. 5 (right) the graph of current density versus

time exhibits the same shape of Fe3+ concentration evolution curve at the anode, as it has to be in

order to describe the flow of the electrons released by Fe atoms. The plot of H+ concentration

spacial distribution in Fig 5 (left) describes the fact that within the bulk it remains unchanged,

whereas it contributes to one of the reduction reactions at the cathode and, also, moves to the anode

in order to counteract the OH – decrease for iron(III) hydroxide production (water dissociation law).

Numerical versus Experimental Test. Conclusions

We have tested the mathematical model by reproducing the experimental galvanic cell filled either

with an acidic solution – the case that has most impact on industrial applications - and, as a further

reliability check-up, with a neutral solution.

Figure 4 – Iron(III) hydroxide concentration spacial distribution at t = 4.7 h (left) and


Figure 5 – Hydrogen ion concentration spacial distribution at t = 4.7 h (left) and

current density evolution (right)

reasonably behaving: in the vicinity of the anode (right plot), it increases rapidly at initial time,

while, after, it approaches a sort of passivation regime; moreover, it concentrates around the anode,

where rust, eventually, would grow, whereas it is completely null elsewhere in the cell (left plot).

The same distribution is, coherently, observed for Iron(III) hydroxide (Fig. 4, left plot), which, in

addition, grows indefinitely in the vicinity of the anode (right plot) due to the lack of the mechanism

of formation of rust in the mathematical model. In Fig. 5 (right) the graph of current density versus

time exhibits the same shape of Fe3+ concentration evolution curve at the anode, as it has to be in

order to describe the flow of the electrons released by Fe atoms. The plot of H+ concentration

spacial distribution in Fig 5 (left) describes the fact that within the bulk it remains unchanged,

whereas it contributes to one of the reduction reactions at the cathode and, also, moves to the anode

in order to counteract the OH – decrease for iron(III) hydroxide production (water dissociation law).

Numerical versus Experimental Test. Conclusions

We have tested the mathematical model by reproducing the experimental galvanic cell filled either

with an acidic solution – the case that has most impact on industrial applications - and, as a further

reliability check-up, with a neutral solution.


In the acidic case, ph = 4, the computed solution exhibits, for the concentrations of the Fe3+

ions

and Fe(OH)3 oxyde at the first internal point aside the anode, the time evolution plotted in Fig.6:

it is

Figure 6 – Test-acidic: time evolution of the numerical concentrations

at the neighborhood of the anode.

apparent that the ferric ion concentration increases asymptotically towards a value (O(10-7

)

mole/m3), whereas, after the first hour, the iron(III)-oxyde concentration grows at constant velocity:

the anode loses continuously iron atoms and releases ferric ions and electrons and, after one hour

adjustement, a constant portion of ferric ions contributes to the oxyde formation. The time

evolution of the current density (see plot in Fig. 7) results reasonably similar to that of the Fe3+

concentration, so the more the concentration of Fe(OH)3 oxyde increases and the more the current

density growth (its derivative) decreases. This behaviour may be referred as a sort of passivation

[6], although this term is mainly used to indicate the effect of the layer of rust that forms on the

anode and substantially prevents from new corrosion events. We stress that the trend of the

numerical Fe(OH)3 oxyde to be indefinitely increasing is indeed unnatural and is due to the

absence, in the mathematical model, of a mechanism describing the saturation of the electrolytic

solution to iron(III)- hydroxide and, also, of rust formation.

Figure 7 - Test_acidic: time evolution of the numerical current density

released in the neighbourhood of the anode

The electric potential gradient developed during the simulated operation time of the galvanic cell

(about 4.5 h)) is drawn in Figure 8. We observe a very good qualitative matching with the

experimental sketch in Figure 9, with respect to either the value and the passivation time (let us just

notice that the experimental observation time interval is shorter and lasts 2 h).

In the acidic case, ph = 4, the computed solution exhibits, for the concentrations of the Fe3+

ions

and Fe(OH)3 oxyde at the first internal point aside the anode, the time evolution plotted in Fig.6:

it is

Figure 6 – Test-acidic: time evolution of the numerical concentrations

at the neighborhood of the anode.

apparent that the ferric ion concentration increases asymptotically towards a value (O(10-7

)

mole/m3), whereas, after the first hour, the iron(III)-oxyde concentration grows at constant velocity:

the anode loses continuously iron atoms and releases ferric ions and electrons and, after one hour

adjustement, a constant portion of ferric ions contributes to the oxyde formation. The time

evolution of the current density (see plot in Fig. 7) results reasonably similar to that of the Fe3+

concentration, so the more the concentration of Fe(OH)3 oxyde increases and the more the current

density growth (its derivative) decreases. This behaviour may be referred as a sort of passivation

[6], although this term is mainly used to indicate the effect of the layer of rust that forms on the

anode and substantially prevents from new corrosion events. We stress that the trend of the

numerical Fe(OH)3 oxyde to be indefinitely increasing is indeed unnatural and is due to the

absence, in the mathematical model, of a mechanism describing the saturation of the electrolytic

solution to iron(III)- hydroxide and, also, of rust formation.

Figure 7 - Test_acidic: time evolution of the numerical current density

released in the neighbourhood of the anode

The electric potential gradient developed during the simulated operation time of the galvanic cell

(about 4.5 h)) is drawn in Figure 8. We observe a very good qualitative matching with the

experimental sketch in Figure 9, with respect to either the value and the passivation time (let us just

notice that the experimental observation time interval is shorter and lasts 2 h).


Figure 8 - Test_acidic: time evolution of the numerical electric

potential gradient of the cell

As further support to the mathematical and numerical model proposed, we display the curves of the

numerical and experimental electric potential gradient for the case of neutral electrolytical solution,

ph = 7, respectively in Figure 10 and Figure 11. Also in this case we observe a substantial

agreement over the initial two operation hours, when most changes occur. We notice that the value

of φ is considerably smaller than in the acidic case, actually, corrosion is less active in presence of a neutral solution due to the lower amount of hydrogen ions, so that also the evolution towards

passivation is slower.

Figure 10 - Test_neutral: time evolution of the

numerical electric potential gradient of the cell

Next improvement of the mathematical model is the inclusion of rust formation from the iron(III)

oxyde and of the solutal convection with gravity effects. These aspects will make our approach to

the simulation of a galvanic cell quite closer to the real process and will allow to recover

exhaustively the passivation phenomenon, an important aspect for the implications on prevention

treatments.

Figure 9 - Test_acidic:

time evolution of the experimental

electric potential gradient of the cell


experimental electric potential gradient of the cell

Figure 8 - Test_acidic: time evolution of the numerical electric

potential gradient of the cell

As further support to the mathematical and numerical model proposed, we display the curves of the

numerical and experimental electric potential gradient for the case of neutral electrolytical solution,

ph = 7, respectively in Figure 10 and Figure 11. Also in this case we observe a substantial

agreement over the initial two operation hours, when most changes occur. We notice that the value

of φ is considerably smaller than in the acidic case, actually, corrosion is less active in presence of a neutral solution due to the lower amount of hydrogen ions, so that also the evolution towards

passivation is slower.


numerical electric potential gradient of the cell

Next improvement of the mathematical model is the inclusion of rust formation from the iron(III)

oxyde and of the solutal convection with gravity effects. These aspects will make our approach to

the simulation of a galvanic cell quite closer to the real process and will allow to recover

exhaustively the passivation phenomenon, an important aspect for the implications on prevention

treatments.

Figure 9 - Test_acidic:

time evolution of the experimental

electric potential gradient of the cell


experimental electric potential gradient of the cell


Acknowledgements. Besides the institutional funding from C.N.R. in benefit of one author, we

acknowledge the financial support of Soprintendenza ai Beni Culturali Regione Autonoma Valle

d'Aosta (project "Analisi di processi di corrosione di materiali ferrosi", 2003-2005) that allowed this

research to be.

References

[1] T. A. Arias, J. Cline, A. A. Rigosand: Physics and Materials Science, Gather/Scatter Vol. 13

(1997), p. 2.

[2] M. R. Radeke, E. A. Carter: Ann. Rev. Phys. Chem Vol. 48 (1997), p. 243

[3] G. Inglese: Inverse problems Vol. 13 (1997), p. 977.

[4] G. Alessandrini, L. Del Piero, L. Rondi: Inverse problems Vol. 19 (2003), p. 973.

[5] V. Botte, D. Mansutti and A. Pascarelli: Applied Numerical Mathematics Vol. 55 (2005),

p. 253.

[6] L. Kiss: Kinetics of electrochemical metal dissolution (Elsevier, Hungary 1988).

[7] J. S. Newman: Electrochemical systems (Prentice-Hall, Englewood Cliffs, NJ 1991).

[8] R. Parson: Handbook of electrochemical constants (Academic Press, New York 1959).

[9] D. Mansutti, G. Colicchio, M.L. Santarelli: Communications to SIMAI Congress Vol. 3 (2009),

p. 331.

Acknowledgements. Besides the institutional funding from C.N.R. in benefit of one author, we

acknowledge the financial support of Soprintendenza ai Beni Culturali Regione Autonoma Valle

d'Aosta (project "Analisi di processi di corrosione di materiali ferrosi", 2003-2005) that allowed this

research to be.

References

[1] T. A. Arias, J. Cline, A. A. Rigosand: Physics and Materials Science, Gather/Scatter Vol. 13

(1997), p. 2.

[2] M. R. Radeke, E. A. Carter: Ann. Rev. Phys. Chem Vol. 48 (1997), p. 243

[3] G. Inglese: Inverse problems Vol. 13 (1997), p. 977.

[4] G. Alessandrini, L. Del Piero, L. Rondi: Inverse problems Vol. 19 (2003), p. 973.

[5] V. Botte, D. Mansutti and A. Pascarelli: Applied Numerical Mathematics Vol. 55 (2005),

p. 253.

[6] L. Kiss: Kinetics of electrochemical metal dissolution (Elsevier, Hungary 1988).

[7] J. S. Newman: Electrochemical systems (Prentice-Hall, Englewood Cliffs, NJ 1991).

[8] R. Parson: Handbook of electrochemical constants (Academic Press, New York 1959).

[9] D. Mansutti, G. Colicchio, M.L. Santarelli: Communications to SIMAI Congress Vol. 3 (2009),

p. 331.


Prognostic Tools for Lifetime Prediction of Aircraft Coatings:

Paint Degradation

John M Colwell1,a, Javaid H Khan2,b, Geoffrey Will2,c,

Kathryn E Fairfull-Smith3,d, Steven E Bottle3,e, Graeme A George1,f and Antony Trueman4,g

1Defence Materials Technology Centre, School of Mechanical and Mining Engineering, University

of Queensland, St Lucia, QLD, 4072, Australia

2Chemistry Discipline, Faculty of Science and Technology, Queensland University of Technology,

Brisbane, QLD, 4001, Australia

3ARC Centre of Excellence for Free Radical Chemistry and Biotechnology, Queensland University

of Technology, Brisbane, QLD, 4001, Australia

4Maritime Platforms Division, Defence Science and Technology Organisation, Fishermans Bend,

VIC, 3207, Australia

[email protected],

[email protected],

[email protected],

[email protected],

[email protected],

[email protected],

[email protected]

Keywords: Aircraft coating, Paint degradation, Fourier transform infrared, Profluorescent nitroxide

Abstract. A direct interrogation, portable analysis technique (portable FT-IR) and a novel

environment-monitoring profluorescent sensor for studying aircraft coating degradation have been

developed. For the direct interrogation approach, a standard military aircraft paint: 459-line

Anzothane flexible polyurethane (lead free) has been used to illustrate a new potential field

technique to evaluate coating service lifetime, portable FT-IR. This technique allows direct analysis

of chemical changes within the degrading coatings and has the potential to evaluate service lifetime

when coupled with advanced statistical analysis methods (chemometrics). The degradation

environment monitoring sensors are embodied in a profluorescent environment-sensitive witness

patch that may be analysed in-service by a field-deployable fluorescence spectrometer. Accelerated

ageing for both the paint and the witness patches has been undertaken and their capabilities as

aircraft paint degradation monitors assessed.

Introduction

Aircraft coatings are designed to primarily offer a corrosion resistant barrier to the

underlying metallic structure. Typical corrosion-resistant coatings protect the metallic substrate by

two mechanisms: by acting as a physical barrier to isolate the substrate, and by containing reactive

material such as pigments or inhibitors that interact with metallic components of the vehicle and

inhibit corrosion. As a coating degrades, due to environmental exposure, the barrier properties

change as a result of chemical changes within the coating’s structure that ultimately lead to

mechanical breakdown.

The loss of properties of a paint coating when exposed to the environment may arise by the

scission of the polymer chains and by further cross-linking reactions that alter the Glass Transition

Temperature (Tg) of the polymer coating [1]. These processes result in localised shrinkage of the

coating at the surface that can result in micro-crack formation [2]. The environmental stresses that

lead to loss of properties include: photo-oxidation, thermo-oxidation and/or hydrolytic degradation.

Prognostic Tools for Lifetime Prediction of Aircraft Coatings:

Paint Degradation

John M Colwell1,a, Javaid H Khan2,b, Geoffrey Will2,c,

Kathryn E Fairfull-Smith3,d, Steven E Bottle3,e, Graeme A George1,f and Antony Trueman4,g

1Defence Materials Technology Centre, School of Mechanical and Mining Engineering, University

of Queensland, St Lucia, QLD, 4072, Australia

2Chemistry Discipline, Faculty of Science and Technology, Queensland University of Technology,

Brisbane, QLD, 4001, Australia

3ARC Centre of Excellence for Free Radical Chemistry and Biotechnology, Queensland University

of Technology, Brisbane, QLD, 4001, Australia

4Maritime Platforms Division, Defence Science and Technology Organisation, Fishermans Bend,

VIC, 3207, Australia

[email protected],

[email protected],

[email protected],

[email protected],

[email protected],

[email protected],

[email protected]

Keywords: Aircraft coating, Paint degradation, Fourier transform infrared, Profluorescent nitroxide

Abstract. A direct interrogation, portable analysis technique (portable FT-IR) and a novel

environment-monitoring profluorescent sensor for studying aircraft coating degradation have been

developed. For the direct interrogation approach, a standard military aircraft paint: 459-line

Anzothane flexible polyurethane (lead free) has been used to illustrate a new potential field

technique to evaluate coating service lifetime, portable FT-IR. This technique allows direct analysis

of chemical changes within the degrading coatings and has the potential to evaluate service lifetime

when coupled with advanced statistical analysis methods (chemometrics). The degradation

environment monitoring sensors are embodied in a profluorescent environment-sensitive witness

patch that may be analysed in-service by a field-deployable fluorescence spectrometer. Accelerated

ageing for both the paint and the witness patches has been undertaken and their capabilities as

aircraft paint degradation monitors assessed.

Introduction

Aircraft coatings are designed to primarily offer a corrosion resistant barrier to the

underlying metallic structure. Typical corrosion-resistant coatings protect the metallic substrate by

two mechanisms: by acting as a physical barrier to isolate the substrate, and by containing reactive

material such as pigments or inhibitors that interact with metallic components of the vehicle and

inhibit corrosion. As a coating degrades, due to environmental exposure, the barrier properties

change as a result of chemical changes within the coating’s structure that ultimately lead to

mechanical breakdown.

The loss of properties of a paint coating when exposed to the environment may arise by the

scission of the polymer chains and by further cross-linking reactions that alter the Glass Transition

Temperature (Tg) of the polymer coating [1]. These processes result in localised shrinkage of the

coating at the surface that can result in micro-crack formation [2]. The environmental stresses that

lead to loss of properties include: photo-oxidation, thermo-oxidation and/or hydrolytic degradation.

Figure 1: Development of crack depth as a function of ageing time in a paint (protective

scheme) coated metal substrate. Diagram courtesy of Graham Clark.

Protective coating breakdown has a long time-frame when compared to corrosion of the

metallic substrate once exposed to the atmosphere (Figure 1). Coating breakdown is a complex

process that is not as well understood as the corrosion process that occurs once the coating fails. As

a result, there are only few groups researching service lifetime prediction for corrosion protective

coatings [3]. The initial stages of coating breakdown, due to photo- or thermo-oxidation, usually

involve free-radical processes that lead to small changes in the chemical composition of the coating

that are typically not detectable by traditional analysis methods. At later stages in the coating

breakdown, the chemical changes within the coating are more obvious and can be detected by

analysis methods such as Fourier transform infrared spectroscopy (FT-IR) [4]. However, methods

for analysis of coating degradation are almost exclusively laboratory based, which limits their

practical use. There is, therefore, a need for suitable analysis tools that can be used to assess coating

degradation in the field.

This research is concerned with the development of tools that may be appropriate to non-

destructively determine the ageing characteristics of a coating and provide a basis for the

determination of the residual life by inspection of the cumulative degradation using techniques that

are adaptable to field inspection. Here we present complementary methods that we are researching

for coating breakdown analysis: a profluorescent environment-sensitive witness patch that detects

oxidative conditions during the early stages of degradation by the production of fluorescence and a

portable FT-IR spectrometer that can be used to show the chemical changes associated with coating

breakdown during its mid-late stages.

Experimental

Paint. The paint under study was a two-part polyurethane, comprised of a pigmented

polyester polyol base resin (459-Line Anzothane flexible polyurethane (lead free)) and an isocyanate

curing solution (455-9007 Super Anzothane curing solution). The base resin and curing solution

were sourced from the Valspar (Australia) Corporation Pty Limited and prepared according to the

manufacturer’s guidelines. The mixed paint was sprayed onto aluminium slides. Some samples

were prepared by the Defence Science and Technology Organisation (DSTO), Melbourne, Australia.

For these samples, a chromate pretreatment was applied to an aluminium panel, followed by an

epoxy primer (MIL-P-23377 from the Valspar [Australia] Corporation Pty Limited), then the

polyurethane topcoat.

Figure 1: Development of crack depth as a function of ageing time in a paint (protective

scheme) coated metal substrate. Diagram courtesy of Graham Clark.

Protective coating breakdown has a long time-frame when compared to corrosion of the

metallic substrate once exposed to the atmosphere (Figure 1). Coating breakdown is a complex

process that is not as well understood as the corrosion process that occurs once the coating fails. As

a result, there are only few groups researching service lifetime prediction for corrosion protective

coatings [3]. The initial stages of coating breakdown, due to photo- or thermo-oxidation, usually

involve free-radical processes that lead to small changes in the chemical composition of the coating

that are typically not detectable by traditional analysis methods. At later stages in the coating

breakdown, the chemical changes within the coating are more obvious and can be detected by

analysis methods such as Fourier transform infrared spectroscopy (FT-IR) [4]. However, methods

for analysis of coating degradation are almost exclusively laboratory based, which limits their

practical use. There is, therefore, a need for suitable analysis tools that can be used to assess coating

degradation in the field.

This research is concerned with the development of tools that may be appropriate to non-

destructively determine the ageing characteristics of a coating and provide a basis for the

determination of the residual life by inspection of the cumulative degradation using techniques that

are adaptable to field inspection. Here we present complementary methods that we are researching

for coating breakdown analysis: a profluorescent environment-sensitive witness patch that detects

oxidative conditions during the early stages of degradation by the production of fluorescence and a

portable FT-IR spectrometer that can be used to show the chemical changes associated with coating

breakdown during its mid-late stages.

Experimental

Paint. The paint under study was a two-part polyurethane, comprised of a pigmented

polyester polyol base resin (459-Line Anzothane flexible polyurethane (lead free)) and an isocyanate

curing solution (455-9007 Super Anzothane curing solution). The base resin and curing solution

were sourced from the Valspar (Australia) Corporation Pty Limited and prepared according to the

manufacturer’s guidelines. The mixed paint was sprayed onto aluminium slides. Some samples

were prepared by the Defence Science and Technology Organisation (DSTO), Melbourne, Australia.

For these samples, a chromate pretreatment was applied to an aluminium panel, followed by an

epoxy primer (MIL-P-23377 from the Valspar [Australia] Corporation Pty Limited), then the

polyurethane topcoat.


Paint Ageing.

Q-Sun. Polyurethane coating samples were exposed in a Q-Sun Xe-3 system

equipped with a xenon arc light source with daylight filters manufactured by Q-panel. The exposure

conditions were cycles of 18 hours light (0.68 W/m2 at 340 nm) with 60% humidity and 47 °C air

temperature followed by 6 hours dark cycle with 60 % humidity and 50 °C air temperature.

Paint Analysis.

Portable FT-IR. Portable FT-IR analysis was conducted using an A2 Exoscan

handheld FT-IR spectrometer manufactured by A2 Technologies (Figure 2). Spectra were measured

using a specular reflectance head and 64 scans at 8 cm-1

resolution. Spectra were corrected using a

Kramers-Kronig (KK) transformation in Omnic Version 7.2 (Thermo Electron Corporation).

Figure 2: A2 Exoscan portable FT-IR spectrometer [5]

FT-IR ATR. FT-IR ATR spectra of the paint coating samples were measured with a

Nicolet Nexus 870 FT-IR spectrometer equipped with Nicolet Endurance horizontal diamond ATR

accessory. Spectra were collected using 64 scans at 4 cm-1

resolution.

Chemometrics. Chemometrics analysis was performed using Solo (Eigenvector

Research Incorporated). FT-IR ATR spectra were preprocessed by first applying an automatic

background correction using Omnic Version 7.2 (Thermo Electron Corporation), then normalising

by setting the total integrated spectral absorbance to 1. Principle Component Analysis (PCA) was

applied over the spectral range: 850 – 1800 cm-1

.

Witness Patch. Profluorescent environment-sensitive witness patches were prepared by

mixing, the profluorescent nitroxide (PFN): 10-(Phenylethynyl)-9-(1,1,3,3-tetramethylisoindolin-2-

yloxyl-5-ethynyl)anthracene (BPETMIOA) [6], with a solvent-borne oxidatively sensitive polymer

mixture and casting into thin films (~3 µm). Control patches, without BPETMIOA or with

BPETMIOA’s fluorescent parent compound - bis(phenylethynyl)anthracene (BPEA) - were

prepared by casting the polymer mixture alone or first mixing with BPEA.

Witness Patch Ageing.

Thermo-oxidation. Thermo-oxidation of the witness patch, and witness patch

control samples, was undertaken at 120 °C in an air-circulating oven. Samples were placed on

aluminium slides during ageing.

Photo-oxidation. Photo-oxidation of the witness patch, and witness patch control

samples, was performed using an Heraeus Suntest CPS+ at an irradiance of 765 W/m2 and a

temperature of 35 °C. The Suntest was fitted with a coated quartz filter, alone, simulating severe

(unnatural) UV stress. Witness patch, and witness patch control samples were mounted on quartz

Paint Ageing.

Q-Sun. Polyurethane coating samples were exposed in a Q-Sun Xe-3 system

equipped with a xenon arc light source with daylight filters manufactured by Q-panel. The exposure

conditions were cycles of 18 hours light (0.68 W/m2 at 340 nm) with 60% humidity and 47 °C air

temperature followed by 6 hours dark cycle with 60 % humidity and 50 °C air temperature.

Paint Analysis.

Portable FT-IR. Portable FT-IR analysis was conducted using an A2 Exoscan

handheld FT-IR spectrometer manufactured by A2 Technologies (Figure 2). Spectra were measured

using a specular reflectance head and 64 scans at 8 cm-1

resolution. Spectra were corrected using a

Kramers-Kronig (KK) transformation in Omnic Version 7.2 (Thermo Electron Corporation).

Figure 2: A2 Exoscan portable FT-IR spectrometer [5]

FT-IR ATR. FT-IR ATR spectra of the paint coating samples were measured with a

Nicolet Nexus 870 FT-IR spectrometer equipped with Nicolet Endurance horizontal diamond ATR

accessory. Spectra were collected using 64 scans at 4 cm-1

resolution.

Chemometrics. Chemometrics analysis was performed using Solo (Eigenvector

Research Incorporated). FT-IR ATR spectra were preprocessed by first applying an automatic

background correction using Omnic Version 7.2 (Thermo Electron Corporation), then normalising

by setting the total integrated spectral absorbance to 1. Principle Component Analysis (PCA) was

applied over the spectral range: 850 – 1800 cm-1

.

Witness Patch. Profluorescent environment-sensitive witness patches were prepared by

mixing, the profluorescent nitroxide (PFN): 10-(Phenylethynyl)-9-(1,1,3,3-tetramethylisoindolin-2-

yloxyl-5-ethynyl)anthracene (BPETMIOA) [6], with a solvent-borne oxidatively sensitive polymer

mixture and casting into thin films (~3 µm). Control patches, without BPETMIOA or with

BPETMIOA’s fluorescent parent compound - bis(phenylethynyl)anthracene (BPEA) - were

prepared by casting the polymer mixture alone or first mixing with BPEA.

Witness Patch Ageing.

Thermo-oxidation. Thermo-oxidation of the witness patch, and witness patch

control samples, was undertaken at 120 °C in an air-circulating oven. Samples were placed on

aluminium slides during ageing.

Photo-oxidation. Photo-oxidation of the witness patch, and witness patch control

samples, was performed using an Heraeus Suntest CPS+ at an irradiance of 765 W/m2 and a

temperature of 35 °C. The Suntest was fitted with a coated quartz filter, alone, simulating severe

(unnatural) UV stress. Witness patch, and witness patch control samples were mounted on quartz


slides then each covered with an aluminium mask that left a 5 mm diameter circular spot of the

sample exposed to the light. The back-face of the quartz slides were covered to prevent reflected

light from striking the sample from beneath the mask.

Witness Patch Analysis.

Fluorescence Spectroscopy. Fluorescence spectroscopy was performed using a

Varian Cary Eclipse fluorescence spectrophotometer. Spectra were measured using a fibre optic

probe with a 45° read-head. Three measurements were made at each analysis time and averaged.

UV-Vis Spectroscopy. UV-Vis spectroscopy was performed using a Varian Cary

5000 UV-Vis-NIR spectrophotometer operating in reflectance mode.

Fluorescence Imaging. Images were captured using a Sony DSC-S75 Cyber-shot

3.3 mega pixel digital camera attached to a Carl Zeiss Stemi 2000-C stereoscopic microscope via a

Carl Zeiss Vario-Sonnar adaptor. Fluorescence excitation was provided by a UVGL-58 Mineralight

handheld UV lamp operating in Long Wave mode (peak emission ~365 nm). The lamp was held at

an angle above the sample. The photomask was removed during imaging. 3D representations of

fluorescence images were prepared using ImageJ, image analysis software.


Paint Degradation. Current military aircraft coating systems are based mainly on a

chromate pretreatment to an aerospace aluminium alloy surface, a chromate-containing epoxy-

polyamide primer, and a polyurethane topcoat. Being the topcoat, the polyurethane is exposed first

to environmental stresses and is the main barrier to corrosive elements for the underlying aluminium

alloy. One of the major environmental stresses for exposed surfaces of the aircraft is photo-

oxidation. Due to the durability of the coatings used for aircraft applications, testing of resistance to

photo-oxidation, and other environmental stresses, can be a long process. To accelerate this

process, testing of paints’ resistance to photo-oxidation can be undertaken by using photo-ageing

systems such as a Q-Sun.

Q-Sun Ageing of Polyurethane. Polyurethane topcoats were exposed to accelerated

photo-oxidation conditions in a Q-Sun system to assess the chemical changes that occurred under

these conditions as a function of ageing time. Two approaches were taken for analysis: bench top

FT-IR ATR, and portable FT-IR (reflectance). As the portable FT-IR was previously untested for

this application, the bench top system provided a suitable control for observing instrumental

differences. One significant difference noted between the two techniques (after correction) is due to

the nature of sampling (ATR versus reflectance) and is evident in the 850 – 1200 cm-1

region of the

spectra, where silica (filler) absorbance bands occur (Figures 3 and 4). This difference can be

attributed to artefacts produced in the reflectance spectra, measured using the portable FT-IR, as a

result of restrahlen band dispersions, which are known for highly absorbing species such as silica

[7]. The other main difference between FT-IR ATR and corrected portable FT-IR reflectance

spectra was the mixed-mode spectra obtained from the portable FT-IR. Uncorrected reflectance

spectra from the portable FT-IR showed a combination of diffuse and specular characteristics.

Diffuse and specular characteristics each require specific correction algorithms. In this case, the

spectral window in which we are mainly interested is below 1800 cm-1

, where the spectra from the

portable FT-IR showed dominant specular characteristics. This allowed for correction of specular

artefacts using a Kramers-Kronig (KK) transformation.

FT-IR ATR analysis during Q-Sun ageing showed significant changes within the amide II

region (1500 – 1600 cm-1

) of the IR spectra as a function of ageing time (Figure 3). Similar changes

were observed in spectra collected with the portable FT-IR (Figure 4). These spectral changes

correspond to changes in the structure of the crosslinks (urea and urethane groups) within the

polyurethane network. Changes in these chemical groups have been noted during UV-ageing of

other polyurethane coatings [4]. We have also noticed similar spectral changes in an earlier study

that involved the thermo-oxidative degradation of the same polyurethane coating studied here [8].

slides then each covered with an aluminium mask that left a 5 mm diameter circular spot of the

sample exposed to the light. The back-face of the quartz slides were covered to prevent reflected

light from striking the sample from beneath the mask.

Witness Patch Analysis.

Fluorescence Spectroscopy. Fluorescence spectroscopy was performed using a

Varian Cary Eclipse fluorescence spectrophotometer. Spectra were measured using a fibre optic

probe with a 45° read-head. Three measurements were made at each analysis time and averaged.

UV-Vis Spectroscopy. UV-Vis spectroscopy was performed using a Varian Cary

5000 UV-Vis-NIR spectrophotometer operating in reflectance mode.

Fluorescence Imaging. Images were captured using a Sony DSC-S75 Cyber-shot

3.3 mega pixel digital camera attached to a Carl Zeiss Stemi 2000-C stereoscopic microscope via a

Carl Zeiss Vario-Sonnar adaptor. Fluorescence excitation was provided by a UVGL-58 Mineralight

handheld UV lamp operating in Long Wave mode (peak emission ~365 nm). The lamp was held at

an angle above the sample. The photomask was removed during imaging. 3D representations of

fluorescence images were prepared using ImageJ, image analysis software.


Paint Degradation. Current military aircraft coating systems are based mainly on a

chromate pretreatment to an aerospace aluminium alloy surface, a chromate-containing epoxy-

polyamide primer, and a polyurethane topcoat. Being the topcoat, the polyurethane is exposed first

to environmental stresses and is the main barrier to corrosive elements for the underlying aluminium

alloy. One of the major environmental stresses for exposed surfaces of the aircraft is photo-

oxidation. Due to the durability of the coatings used for aircraft applications, testing of resistance to

photo-oxidation, and other environmental stresses, can be a long process. To accelerate this

process, testing of paints’ resistance to photo-oxidation can be undertaken by using photo-ageing

systems such as a Q-Sun.

Q-Sun Ageing of Polyurethane. Polyurethane topcoats were exposed to accelerated

photo-oxidation conditions in a Q-Sun system to assess the chemical changes that occurred under

these conditions as a function of ageing time. Two approaches were taken for analysis: bench top

FT-IR ATR, and portable FT-IR (reflectance). As the portable FT-IR was previously untested for

this application, the bench top system provided a suitable control for observing instrumental

differences. One significant difference noted between the two techniques (after correction) is due to

the nature of sampling (ATR versus reflectance) and is evident in the 850 – 1200 cm-1

region of the

spectra, where silica (filler) absorbance bands occur (Figures 3 and 4). This difference can be

attributed to artefacts produced in the reflectance spectra, measured using the portable FT-IR, as a

result of restrahlen band dispersions, which are known for highly absorbing species such as silica

[7]. The other main difference between FT-IR ATR and corrected portable FT-IR reflectance

spectra was the mixed-mode spectra obtained from the portable FT-IR. Uncorrected reflectance

spectra from the portable FT-IR showed a combination of diffuse and specular characteristics.

Diffuse and specular characteristics each require specific correction algorithms. In this case, the

spectral window in which we are mainly interested is below 1800 cm-1

, where the spectra from the

portable FT-IR showed dominant specular characteristics. This allowed for correction of specular

artefacts using a Kramers-Kronig (KK) transformation.

FT-IR ATR analysis during Q-Sun ageing showed significant changes within the amide II

region (1500 – 1600 cm-1

) of the IR spectra as a function of ageing time (Figure 3). Similar changes

were observed in spectra collected with the portable FT-IR (Figure 4). These spectral changes

correspond to changes in the structure of the crosslinks (urea and urethane groups) within the

polyurethane network. Changes in these chemical groups have been noted during UV-ageing of

other polyurethane coatings [4]. We have also noticed similar spectral changes in an earlier study

that involved the thermo-oxidative degradation of the same polyurethane coating studied here [8].


The spectral changes observed by FT-IR ATR during photo-oxidative ageing were analysed by an

advanced statistical method (chemometrics) and a correlation between the observed spectral

changes and ageing time was generated (Figure 5).

Principal-component analysis (PCA) was the chemometrics analysis method that was

applied to our data. PCA is the statistical analysis of spectral data in order to reduce the data set to

the minimum number of principal components which, with an appropriate precision can explain all

of the variations in the spectra. The total spectral output from the data set may be expressed as a

series of linear combinations of these components, so the outputs from a PCA analysis of a data set

are the score vector, which indicates how closely each sample relates to the principal component

while the loading vector contains the spectral information in the principal component. The first

principal component (PC1) should account for the greatest variance in the data set, here 63 %. Thus

the spectrum at each ageing time will be a linear combination of principal components given by the

product score multiplied by the loading for that particular sample. Generally, the plot of PC1

against time captures important spectral changes within a data set [9].

PCA analysis showed that there was a correlation between spectral changes, represented as a

‘Score on PC1’, during accelerated photo-oxidation and ageing time over long ageing times

(129 days). The spectral changes represented by PC1 can be summarised as a decrease in polymer

matrix absorbance bands, as a result of oxidative degradation, and an increase in pigment

absorbance bands due to exposure of the pigments after degradation of the polymer matrix as the

sample was aged in the Q-Sun.

One of the benefits of using chemometrics techniques as methods for correlating spectral

changes with ageing time is that data that represent a failure criterion can be used as inputs into the

analysis. In this way, it may be possible to relate the observed spectral changes with how close the

coating is to failing. The coatings in this study did not fail within the analysis time and so this was

not able to be tested. The coating failure criterion was set to mechanical breakdown (micro-

cracking), which was determined by microscopic examination. Microscopic examination showed

that degradation of the coatings had not proceeded far enough to show micro-cracking over the

studied ageing time. Even so, FT-IR was able to show significant changes within the chemical

structure of the degraded coatings before mechanical breakdown and, therefore, has high potential

to be used as a prognostic tool for monitoring the lifetime of aircraft, and other, paint coatings.

The spectral changes observed by FT-IR ATR during photo-oxidative ageing were analysed by an

advanced statistical method (chemometrics) and a correlation between the observed spectral

changes and ageing time was generated (Figure 5).

Principal-component analysis (PCA) was the chemometrics analysis method that was

applied to our data. PCA is the statistical analysis of spectral data in order to reduce the data set to

the minimum number of principal components which, with an appropriate precision can explain all

of the variations in the spectra. The total spectral output from the data set may be expressed as a

series of linear combinations of these components, so the outputs from a PCA analysis of a data set

are the score vector, which indicates how closely each sample relates to the principal component

while the loading vector contains the spectral information in the principal component. The first

principal component (PC1) should account for the greatest variance in the data set, here 63 %. Thus

the spectrum at each ageing time will be a linear combination of principal components given by the

product score multiplied by the loading for that particular sample. Generally, the plot of PC1

against time captures important spectral changes within a data set [9].

PCA analysis showed that there was a correlation between spectral changes, represented as a

‘Score on PC1’, during accelerated photo-oxidation and ageing time over long ageing times

(129 days). The spectral changes represented by PC1 can be summarised as a decrease in polymer

matrix absorbance bands, as a result of oxidative degradation, and an increase in pigment

absorbance bands due to exposure of the pigments after degradation of the polymer matrix as the

sample was aged in the Q-Sun.

One of the benefits of using chemometrics techniques as methods for correlating spectral

changes with ageing time is that data that represent a failure criterion can be used as inputs into the

analysis. In this way, it may be possible to relate the observed spectral changes with how close the

coating is to failing. The coatings in this study did not fail within the analysis time and so this was

not able to be tested. The coating failure criterion was set to mechanical breakdown (micro-

cracking), which was determined by microscopic examination. Microscopic examination showed

that degradation of the coatings had not proceeded far enough to show micro-cracking over the

studied ageing time. Even so, FT-IR was able to show significant changes within the chemical

structure of the degraded coatings before mechanical breakdown and, therefore, has high potential

to be used as a prognostic tool for monitoring the lifetime of aircraft, and other, paint coatings.


wavenumber / cm-1

700800900100011001200130014001500160017001800

arbitrary units

0 days

20 days

52 days

68 days

Unaged

Aged for 68 days

Figure 3: FT-IR ATR spectra from a polyurethane sample aged in the Q-Sun over 68 days.

Spectra were normalised using the area of the peaks due to C-H stretching between 2700 –

3020 cm-1, then zeroed at 1815 cm

-1 to minimise sampling differences due to scattering and

carbon pigment composition.

wavenumber / cm-1

700800900100011001200130014001500160017001800

arbitrary units

0 days

20 days

52 days

68 days

Unaged

Aged for 68 days

Figure 3: FT-IR ATR spectra from a polyurethane sample aged in the Q-Sun over 68 days.

Spectra were normalised using the area of the peaks due to C-H stretching between 2700 –

3020 cm-1, then zeroed at 1815 cm

-1 to minimise sampling differences due to scattering and

carbon pigment composition.


wavenumber / cm-1

700800900100011001200130014001500160017001800

effective absorbance

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0 days

20 days

52 days

68 days

Unaged

Aged for 68 days

Figure 4: A2 Exoscan spectra from of a polyurethane sample aged in the Q-Sun over 68 days.

Spectra were corrected for specular reflectance artefacts using a Kramers-Kronig (KK)

transformation.

wavenumber / cm-1

700800900100011001200130014001500160017001800

effective absorbance

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0 days

20 days

52 days

68 days

Unaged

Aged for 68 days

Figure 4: A2 Exoscan spectra from of a polyurethane sample aged in the Q-Sun over 68 days.

Spectra were corrected for specular reflectance artefacts using a Kramers-Kronig (KK)

transformation.


time / days

0 20 40 60 80 100 120 140

Score on PC1

-0.008

-0.006

-0.004

-0.002

0.000

0.002

0.004

0.006

Figure 5: Score on PC1 versus ageing time in the Q-Sun for a DSTO-prepared polyurethane

paint coating sample. Spectra used for chemometrics analysis were measured using FT-IR

ATR.

Witness Patch. The witness patch contains a profluorescent nitroxide (PFN), BPETMIOA,

which can be used as a sensor of carbon-centred free-radicals [6]. During oxidative degradation of

polymer materials such as paint coatings, carbon-centred free-radicals are formed [9]. Just like a

paint coating, the polymer-based witness patch is sensitive to oxidative degradation and will form

carbon-centred free-radicals during exposure. Trapping of carbon-centred free-radicals, formed as

the patch is exposed to an oxidative environment, can be monitored by observing changes in the

intensity of fluorescence from the BPETMIOA (Figure 6). In this way, the witness patch can be

used as a monitor for degradative thermo- or photo-oxidative environments near a paint coating.

Like the constituents of the paint coatings, BPETMIOA is an organic compound and is, therefore,

potentially susceptible to oxidative degradation. The stability of BPETMIOA can be monitored by

observing changes in its UV-Vis spectrum.

time / days

0 20 40 60 80 100 120 140

Score on PC1

-0.008

-0.006

-0.004

-0.002

0.000

0.002

0.004

0.006

Figure 5: Score on PC1 versus ageing time in the Q-Sun for a DSTO-prepared polyurethane

paint coating sample. Spectra used for chemometrics analysis were measured using FT-IR

ATR.

Witness Patch. The witness patch contains a profluorescent nitroxide (PFN), BPETMIOA,

which can be used as a sensor of carbon-centred free-radicals [6]. During oxidative degradation of

polymer materials such as paint coatings, carbon-centred free-radicals are formed [9]. Just like a

paint coating, the polymer-based witness patch is sensitive to oxidative degradation and will form

carbon-centred free-radicals during exposure. Trapping of carbon-centred free-radicals, formed as

the patch is exposed to an oxidative environment, can be monitored by observing changes in the

intensity of fluorescence from the BPETMIOA (Figure 6). In this way, the witness patch can be

used as a monitor for degradative thermo- or photo-oxidative environments near a paint coating.

Like the constituents of the paint coatings, BPETMIOA is an organic compound and is, therefore,

potentially susceptible to oxidative degradation. The stability of BPETMIOA can be monitored by

observing changes in its UV-Vis spectrum.


Figure 6: Sensing of carbon-centred free-radicals (R.) by BPETMIOA.

Thermo-oxidative ageing. Accelerated thermo-oxidative ageing of a witness patch

at 120 °C in air led to an increase in the observed fluorescence intensity up to a peak at 16 hours

ageing time (Figure 7). This indicates that the witness patch can be used as a thermo-oxidative-

environment monitor. Even so, concurrent degradation of the BPETMIOA chromophore was

observed by monitoring UV-Vis spectra over this time (Figure 7). The observed degradation of the

BPETMIOA chromophore is presumably due to secondary oxidation in the already oxidized patch,

which has been reported for other PFN in polypropylene [10], and limits the sensitivity of

BPETMIOA under these conditions. Further development of PFN that are stable under these

conditions is necessary to enhance the sensitivity of this technique.

Figure 6: Sensing of carbon-centred free-radicals (R.) by BPETMIOA.

Thermo-oxidative ageing. Accelerated thermo-oxidative ageing of a witness patch

at 120 °C in air led to an increase in the observed fluorescence intensity up to a peak at 16 hours

ageing time (Figure 7). This indicates that the witness patch can be used as a thermo-oxidative-

environment monitor. Even so, concurrent degradation of the BPETMIOA chromophore was

observed by monitoring UV-Vis spectra over this time (Figure 7). The observed degradation of the

BPETMIOA chromophore is presumably due to secondary oxidation in the already oxidized patch,

which has been reported for other PFN in polypropylene [10], and limits the sensitivity of

BPETMIOA under these conditions. Further development of PFN that are stable under these

conditions is necessary to enhance the sensitivity of this technique.


time / h

0 2 4 6 8 10 12 14 16 18 20

integrated UV-Vis absorbance (365 - 497 nm) / a.u.

0

2

4

6

8

10

12

14

I/I 0 for integrated fluorescence signal

0

1

2

3

4

5

6

7

8

9

10

Figure 7: Fluorescence and UV-Vis changes in a witness patch as a function of ageing time at

120 °C in air.

Photo-oxidative ageing. Accelerated photo-oxidative ageing of photo-masked

samples was undertaken in an environment of severe unnatural UV stress, which represents a strong

test for the field durability of the witness patch. Fluorescence imaging showed that there was a

clear, visible fluorescence increase of the exposed patch followed by a substantial decrease with

increasing ageing time (Figure 8). The fluorescence signal peaked between 75 - 105 minutes, and

then started to decay as the BPETMIOA, itself, degraded. The susceptibility of BPETMIOA (the

PFN contained in the witness patch) to photo-degradation can be ascribed to the instability of the

parent fluorophore on which BPETMIOA is based (Figure 9). The integrated fluorescence signal

from the exposed circle as a function of ageing time is shown in Figure 10 for the witness patch and

for both the fluorescent parent (BPEA) and the BPETMIOA-free control samples. Figure 10 shows

a clear decay of the fluorescence from the BPEA from the start of ageing, indicating that it was not

stable under the ageing conditions. Also, the fluorescence from the BPETMIOA-free control

samples remained low and constant throughout ageing, indicating that any fluorescence changes

observed for the other samples were only due to changes in the fluorescent compounds. As

concluded for the thermo-oxidative case, new, more stable PFN are required to enhance the

sensitivity and field applicability of this technique.

time / h

0 2 4 6 8 10 12 14 16 18 20

integrated UV-Vis absorbance (365 - 497 nm) / a.u.

0

2

4

6

8

10

12

14

I/I 0 for integrated fluorescence signal

0

1

2

3

4

5

6

7

8

9

10

Figure 7: Fluorescence and UV-Vis changes in a witness patch as a function of ageing time at

120 °C in air.

Photo-oxidative ageing. Accelerated photo-oxidative ageing of photo-masked

samples was undertaken in an environment of severe unnatural UV stress, which represents a strong

test for the field durability of the witness patch. Fluorescence imaging showed that there was a

clear, visible fluorescence increase of the exposed patch followed by a substantial decrease with

increasing ageing time (Figure 8). The fluorescence signal peaked between 75 - 105 minutes, and

then started to decay as the BPETMIOA, itself, degraded. The susceptibility of BPETMIOA (the

PFN contained in the witness patch) to photo-degradation can be ascribed to the instability of the

parent fluorophore on which BPETMIOA is based (Figure 9). The integrated fluorescence signal

from the exposed circle as a function of ageing time is shown in Figure 10 for the witness patch and

for both the fluorescent parent (BPEA) and the BPETMIOA-free control samples. Figure 10 shows

a clear decay of the fluorescence from the BPEA from the start of ageing, indicating that it was not

stable under the ageing conditions. Also, the fluorescence from the BPETMIOA-free control

samples remained low and constant throughout ageing, indicating that any fluorescence changes

observed for the other samples were only due to changes in the fluorescent compounds. As

concluded for the thermo-oxidative case, new, more stable PFN are required to enhance the

sensitivity and field applicability of this technique.


Figure 8: 3D representations of fluorescence images from a masked (inner circle exposed)

witness patch aged under conditions of severe unnatural UV stress. Left to right: 0, 75, and

190 min ageing time. The effect around the edge of the exposed circle is due to roughening of

the surface at the edge of the photomask. This roughening produces an artefact: optical

enhancement of the fluorescent signal at the roughened edges.

Figure 9: Structures of BPETMIOA and the fluorescent compound on which it is based,

BPEA.

Figure 8: 3D representations of fluorescence images from a masked (inner circle exposed)

witness patch aged under conditions of severe unnatural UV stress. Left to right: 0, 75, and

190 min ageing time. The effect around the edge of the exposed circle is due to roughening of

the surface at the edge of the photomask. This roughening produces an artefact: optical

enhancement of the fluorescent signal at the roughened edges.

Figure 9: Structures of BPETMIOA and the fluorescent compound on which it is based,

BPEA.


time / min

0 20 40 60 80 100 120 140 160 180 200

average fluorescence intensity per pixel

0

10

20

30

40

50

60

70

80

fluorescent parent (BPEA) control

witness patch

BPETMIOA-free control

Figure 10: Average fluorescence intensity per pixel (from exposed circle in fluorescence

images) as a function of ageing time for the witness patch and control samples aged under

conditions of severe unnatural UV stress.

Summary

Two potential in-field techniques (portable FT-IR and an environment-sensitive witness

patch) for the analysis of aircraft coating degradation have been evaluated.

Portable FT-IR (reflectance) and benchtop FT-IR ATR were used for the direct chemical analysis

of a polyurethane aircraft coating’s breakdown due to photo-oxidation in a Q-Sun and showed

similar spectral changes. A correlation between Q-Sun ageing time and FT-IR ATR spectral

changes was observed by advanced statistical analysis (chemometrics).

The witness patch has shown to be a powerful visible oxidative environment indicator under both

thermo- and photo-oxidative conditions. Further development of the witness patch is required to

overcome issues with stability of the fluorescent signalling compounds.

The use of techniques such as these may allow the field evaluation of paint coatings and estimates

of damage and service life. It is envisaged that these techniques will be able to detect the early signs

of coating breakdown before the underlying metallic structure is exposed to the environment. This

will allow a preventative maintenance schedules to be implemented and will be applicable across

the greater spectrum of coating, composites and sealing agents used in the aircraft area.

Acknowledgements

The authors acknowledge the financial support of DSTO, Melbourne for part of this work.

This research was partly conducted within the Defence Materials Technology Centre, which was

established and is supported by the Australian Government’s Defence Future Capability Technology

Centre (DFCTC) initiative. The support of the Australian Research Council Centre of Excellence

Scheme under the ARC Centres of Excellence Program, CE0561607, is also acknowledged. The

authors also gratefully acknowledge Prof. Graham Clark, RMIT, for production of Figure 1.

time / min

0 20 40 60 80 100 120 140 160 180 200

average fluorescence intensity per pixel

0

10

20

30

40

50

60

70

80

fluorescent parent (BPEA) control

witness patch

BPETMIOA-free control

Figure 10: Average fluorescence intensity per pixel (from exposed circle in fluorescence

images) as a function of ageing time for the witness patch and control samples aged under

conditions of severe unnatural UV stress.

Summary

Two potential in-field techniques (portable FT-IR and an environment-sensitive witness

patch) for the analysis of aircraft coating degradation have been evaluated.

Portable FT-IR (reflectance) and benchtop FT-IR ATR were used for the direct chemical analysis

of a polyurethane aircraft coating’s breakdown due to photo-oxidation in a Q-Sun and showed

similar spectral changes. A correlation between Q-Sun ageing time and FT-IR ATR spectral

changes was observed by advanced statistical analysis (chemometrics).

The witness patch has shown to be a powerful visible oxidative environment indicator under both

thermo- and photo-oxidative conditions. Further development of the witness patch is required to

overcome issues with stability of the fluorescent signalling compounds.

The use of techniques such as these may allow the field evaluation of paint coatings and estimates

of damage and service life. It is envisaged that these techniques will be able to detect the early signs

of coating breakdown before the underlying metallic structure is exposed to the environment. This

will allow a preventative maintenance schedules to be implemented and will be applicable across

the greater spectrum of coating, composites and sealing agents used in the aircraft area.

Acknowledgements

The authors acknowledge the financial support of DSTO, Melbourne for part of this work.

This research was partly conducted within the Defence Materials Technology Centre, which was

established and is supported by the Australian Government’s Defence Future Capability Technology

Centre (DFCTC) initiative. The support of the Australian Research Council Centre of Excellence

Scheme under the ARC Centres of Excellence Program, CE0561607, is also acknowledged. The

authors also gratefully acknowledge Prof. Graham Clark, RMIT, for production of Figure 1.


References

[1] B.M.D. Fernando, X. Shi, S.G. Croll: J. Coat. Technol. Res. Vol. 5 (2008), p. 1.

[2] G.P. Bierwagen: Prog. Org. Coat. Vol. 15 (1987), p.179.

[3] G. Bierwagen: J. Coat. Technol. Res. Vol. 5 (2008), p. 133.

[4] X.F. Yang et al.: Polym. Degrad. Stab. Vol. 74 (2001), p. 341.

[5] Sourced from http://www.a2technologies.com/exoscan_battery.html

[6] K.E. Fairfull-Smith, S.E. Bottle: Eur. J. Org. Chem. Vol. 2008 (2008), p. 5391.

[7] P.G. Appleyard, N. Davies: Opt. Eng. Vol. 43 (2004), p. 376.

[8] J.H. Khan, G. Will, G.A. George, J.M. Colwell, A. Trueman: Corrosion and Prevention

2009: Conference Proceedings, paper 094.

[9] P.J. Halley, G.A. George: Chemorheology of Polymers (Cambridge University Press, United

Kingdom 2009).

[10] A.S. Micallef et al.: Polym. Degrad. Stab. Vol. 89 (2005), p. 427.

References

[1] B.M.D. Fernando, X. Shi, S.G. Croll: J. Coat. Technol. Res. Vol. 5 (2008), p. 1.

[2] G.P. Bierwagen: Prog. Org. Coat. Vol. 15 (1987), p.179.

[3] G. Bierwagen: J. Coat. Technol. Res. Vol. 5 (2008), p. 133.

[4] X.F. Yang et al.: Polym. Degrad. Stab. Vol. 74 (2001), p. 341.

[5] Sourced from http://www.a2technologies.com/exoscan_battery.html

[6] K.E. Fairfull-Smith, S.E. Bottle: Eur. J. Org. Chem. Vol. 2008 (2008), p. 5391.

[7] P.G. Appleyard, N. Davies: Opt. Eng. Vol. 43 (2004), p. 376.

[8] J.H. Khan, G. Will, G.A. George, J.M. Colwell, A. Trueman: Corrosion and Prevention

2009: Conference Proceedings, paper 094.

[9] P.J. Halley, G.A. George: Chemorheology of Polymers (Cambridge University Press, United

Kingdom 2009).

[10] A.S. Micallef et al.: Polym. Degrad. Stab. Vol. 89 (2005), p. 427.


This page has been reformatted by Knovel to provide easier navigation.

AUTHORS INDEX

Index Terms Links

B

Beketaeva, L.A. 7

Bestetti, M. 79

Bockmair, G. 41

Bottle, S.E. 137

Bozzini, B. 93

Brotzu, A. 21

Bukhan’ko, N.G. 7

C

Campanella, L. 21

Chaudhuri, S. 107

Ciliberto, A. 93

Colicchio, G. 127

Colwell, J.M. 137

Curioni, M. 29

D

Da Forno, A. 79

Davydov, A.D. 7

De Filippo, B. 21

F

Fairfull-Smith, K.E. 137

Feil, F. 47

Ferro, D. 21

Flamini, D.O. 63

Furbeth, W. 47

G

Gastaldi, D. 85

George, G.A. 137

Index Terms Links


I

Ippolito, J. 55

K

Kannan, M.B. 1

Khan, J.H. 137

Kharin, V. 117

Klapkiv, M.D. 55

Kranzeder, K. 41

L

Lacitignola, D. 93

Lorenzo, M. 117

Losada, M. 107

M

Mackie, K. 107

Mansutti, D. 127

Marchitto, M. 93

Mele, C. 93

Migliavacca, F. 85

N

Natali, S. 21

Nykyforchyn, H.M. 55

O

Osborne, J.H. 107

P

Petrini, L. 85

Pokhmurskii, V.I. 55

R

Raja, V.S. 1

Rybalka, K.V. 7

Index Terms Links


S

Santarelli, M.L. 127

Sassi, V. 85

Sgura, I. 93

Shaldaev, V.S. 7

Student, M.M. 55

T

Thompson, G.E. 29

Toribio, J. 117

Trasatti, S.P. 63 85

Trueba, M. 63

Trueman, A. 137

V

Vedani, M. 85

Vergara, D. 117

W

Will, G. 137


KEYWORDS INDEX

Index Terms Links

A

Aircraft Coating 137

Aircraft Maintenance 41

Aluminium Alloy 1 63

Aluminum Titanium 55

Atmospheric Corrosion 21

Axisymmetric Notch 117

B

Bioresorbable Stent 85

Bronze 21

C

Ceramic Coating 55

Coating 107

Colorimetric Measurement 21

Conducting Polymer 63

Corrosion Inhibitor 41 47

Corrosion Property 55

Corrosion Protection 63

Corrosion-Resistance 79

D

Damage Model 85

Dynamic Local Contact Angle 107

E

Electrochemical Impedance 47

Spectroscopy (EIS)

Electrodeposition 93

F

Fourier Transform Infrared 137

Index Terms Links


G

General Corrosion 7

H

Heat Treatment 1

Hydrogen Diffusion 117

Hydrophobicity 107

I

Intergranular Corrosion 1

Ionic Liquid 93

Iron 127

K

Kinetics 127

M

Magnesium 47

Magnesium Alloy 55 79 85

Metallic Biomaterial 85

Micro-Arc Anodizing 79

Mn 93

Molecular Dynamics (MD) 107

N

Nickel-Aluminum Alloy 7

Numerical Modeling 117

Numerical Simulation 93 123

P

Paint Degradation 137

Passivation 47

Pattern Formation 93

PDE 127

Pitting Corrosion 7

Plasma Electrolyte Oxidation 55

Profluorescent Nitroxide 137

Index Terms Links


R

Reaction Diffusion 93

Redox Reaction 127

S

Scandium Alloying 1

Silane-Based Treatment 63

Sodium Chloride Solution 7

Sol-Gel 47

Steel 55

Stress Corrosion Cracking 1

Surface Protection 41

T

Turing Instability 93

W

Wear Resistance 79

Weighted Residual Method 117

X

X-Ray Diffraction (XRD) 21

Z

Zinc Rich Primers 41

light weight metal corrosion and modeling for corrosion prevention, life prediction and assessment

Documents