Ec

Ja

b

a

ARRA

KCDCCNPP

1

mTnmmatarltati

ctc

0d

Materials Chemistry and Physics 129 (2011) 569– 579

Contents lists available at ScienceDirect

Materials Chemistry and Physics

j ourna l ho me pag e: www.elsev ier .com/ locate /matchemphys

lectrical impedance analysis based quantification of microstructural changes inoncretes due to non-steady state chloride migration

itendra Jaina, Narayanan Neithalathb,∗

School of Civil Engineering, Purdue University, West Lafayette, IN 47907, USASchool of Sustainable Engineering and the Built Environment, Arizona State University, Box No. 875306, Tempe, AZ 85287, USA

r t i c l e i n f o

rticle history:eceived 28 July 2010eceived in revised form 11 April 2011ccepted 26 April 2011

eywords:. Electrical characterization (impedance). Microstructurehloride transport

a b s t r a c t

Non-steady state migration (NSSM) test is commonly employed to evaluate chloride ion ingress into con-cretes. This paper quantifies the microstructural changes in concretes subjected to NSSM tests, induced bythe chemical constitution of the cement and replacement materials. Electrical impedance spectroscopy(EIS) and equivalent electrical circuit modeling are used to obtain characteristic features of the concretepore structure that aid in this quantification. A methodology is provided in this paper to convert thebulk resistances of the specimens under the action of any applied voltages to the ones correspondingto a 30 V applied potential. The relationships between: (i) bulk resistance of the specimens before theNSSM test and the NSSM coefficient Dnssm, (ii) resistance of the connected pores and the microstructural

hloride bindingon-steady state migrationorosityore connectivity factor

parameter �ˇ (product of porosity and pore connectivity), and (iii) bulk resistance after the NSSM testcorresponding to a 30 V potential and the chloride penetration depths, demonstrate the effect of chloridebinding. The capacitance related to the pore–solid interface obtained from the equivalent circuit is usedto indicate that the chloride binding products are formed at the interfaces and influence the tortuosity ofthe pore system. The experiments and the model suggest an average reduction of about 10% in �ˇ valuesafter the NSSM test.

. Introduction

Chloride induced steel reinforcement corrosion is one of theost prominent deterioration mechanisms in reinforced concrete.

his has resulted in a large number of studies on the mecha-isms of chloride transport in concrete, the effects of the concreteicrostructure on transport, and means to modify the materialicrostructure so as to reduce chloride ion ingress. It is well

ccepted that, in the absence of electric fields, chloride ions areransported in concrete through diffusion (movement of ions under

chemical potential), absorption or wick action [1]. Diffusion isegarded as the primary means of transporting chlorides to theevel of reinforcing steel [2]; however, in the case of low coverhickness, capillary absorption might also be a dominant mech-nism. The microstructural changes induced in concretes throughhe use of supplementary cementing materials that inhibit chlorideon transport have also been well documented [3,4].

Several experimental techniques are available to determine the

hloride ion penetration resistance of concretes. In general, theseest methods can be classified into two types – one in which thehloride ion transport into the saturated concrete is slow because of

∗ Corresponding author. Tel.: +1 480 965 6023; fax: +1 480 965 0557.E-mail address: Narayanan.Neithalath@asu.edu (N. Neithalath).

254-0584/$ – see front matter © 2011 Elsevier B.V. All rights reserved.oi:10.1016/j.matchemphys.2011.04.057

© 2011 Elsevier B.V. All rights reserved.

the concentration gradients (diffusion tests), and the other in whichthe transport is accelerated through the application of an electricfield [5–7]. The chloride migration tests as well as the conductivitytests fall into the second category. Accelerated tests such as therapid chloride permeability test (ASTM C 1012), which basicallyis just a conductivity or resistivity test, and the non-steady statemigration test (NT Build 492 [8]) are commonly used to determinethe chloride transport parameters of concretes. Different means ofpredicting the diffusion or migration coefficients from acceleratedtests have also been proposed [9–11].

When chloride ions penetrate into concrete, either in slow or inaccelerated test conditions, they interact with the pore system ofconcrete, and a few ions “bind” to the hydration products. Consid-ering the contribution of chloride binding in transport is importantbecause of three reasons [12]: (i) reduction in free chloride ionsnear the reinforcing steel that reduces the chances of corrosion,(ii) reduction in overall chloride concentration in the penetratingmedia thus retarding the penetration rate, and (iii) formation ofFriedel’s salt, resulting in a less porous microstructure that furtherslows down transport. Chloride binding depends on a host of fac-tors including the chemical composition of the ingredients, cation

associated with Cl−, pore structure of concrete, and the appliedelectrical field. It is reported that the binding capacity increaseswith increase in the C3A content of cement (especially in the highchloride concentration range of 1–3 M) [12], or the increase in fly

570 J. Jain, N. Neithalath / Materials Chemistry and Physics 129 (2011) 569– 579

Table 1Chemical composition and physical characteristics of materials used in this study.

Composition (% by mass)/property Cement Fly ash Silica fume VCAS

Silica (SiO2) 20.2 50.24 93.4 54.2Alumina (Al2O3) 4.70 28.78 0.42 17.8Iron oxide (Fe2O3) 3.00 5.72 0.52 1.00Calcium oxide (CaO) 61.9 5.86 1.91 24.2Magnesium oxide (MgO) 2.60 1.74 0 0.80Sodium oxide (Na2O) 0.19

0.96b 0.25 0.75Potassium oxide (K2O) 0.82 0.79 0.20Sulfur trioxide (SO3) 3.90 0.51 0.34 <0.10Loss on ignition 1.90 2.80 2.30 <0.50Median particle size (�m) 13.0 20.0 <1a 8.00Density (kg m−3) 3150 2400 2200 2600

acm

rumitmmtirhpatsstsnifibtfama

2

2

eaaiabmpcaoVrcTs

NSSM cells, just before the NSSM test and after the completion of the test. The EIS

a Agglomerates can approach or exceed the size of cement grains.b Equivalent alkali content.

sh content in concrete [13] because of the increased aluminateontent. The use of silica fume as a partial cement replacementaterial is reported to reduce the chloride binding [14].The influence of electrical field on chloride binding has not

eceived significant attention, and there is a need to furthernderstand the combined effect of electrical field and supple-entary cementing materials on chloride binding [12]. This study

ntends to address this concern by investigating the changes inhe microstructure of concretes modified using different supple-

entary cementing materials, when subjected to non-steady stateigration (NSSM) tests. In order to use the non-steady state migra-

ion coefficient as a predictive parameter of the chloride transportn concrete, it is important to understand the influence of chlo-ide binding during the NSSM test vis-à-vis natural diffusion, andow different supplementary cementing materials influence thishenomenon. The influence of supplementary cementing materi-ls such as fly ash and a vitreous calcium aluminosilicate (VCAS)hat has higher alumina contents than the cement it replaces, andilica fume that has a lower alumina content, is investigated in thistudy. The reason for the use of non-steady state migration tests ishe following. In the NSSM test, the stronger electrical field and thehorter test duration result in reduced chloride binding than in aon-steady state diffusion test [15]. If the microstructural changes

nduced as a result of binding during NSSM test can be quanti-ed, it establishes the sensitivity of the method used to quantifyinding and hence can be used for conditions such as the diffusionest in which more amounts of binding products are likely to beormed. Electrical impedance spectroscopy (EIS) along with equiv-lent electrical circuit modeling is used in this paper to quantify theicrostructural changes in plain and modified concretes before and

fter subjecting them to NSSM tests.

. Experimental program

.1. Materials and mixture proportions

Type I Ordinary Portland cement conforming to ASTM C 150 was used for all thexperiments described in this study. A Class F fly ash (FA) conforming to ASTM C 618,

dry densified silica fume (SF) corresponding to ASTM C 1240, or vitreous calciumlumino-silicate (VCAS) were used as partial cement replacement materials. VCASs a relatively new high performance pozzolanic material manufactured by heating

blend of ground silica, lime, and alumina compounds to a molten state, followedy solidification by quenching, and final grinding. VCAS, like metakaolin, is an alu-inosilicate, with a very similar SiO2 content but a lower Al2O3 content [16,17]. The

hysical and chemical characteristics of these materials are shown in Table 1. Con-rete mixtures were proportioned by replacing either 10% or 20% of cement with flysh (indicated as FA 10 or FA 20 respectively in the tables and graphs), or 6% or 9%f cement with silica fume (indicated as SF 6 or SF 9), or 9% or 15% of cement withCAS (indicated as VCAS 9 or VCAS 15), by mass. The water-to-cementing materials

atio (w/cm) for all the concretes was maintained at 0.40. The cementing materialsontent (cement + replacement materials) of the mixtures was fixed at 430 kg m−3.able 2 provides the mixture proportions for all the concrete mixtures used in thistudy. Cylindrical specimens (100 mm diameter × 200 mm height) were cast and

Fig. 1. Schematic of a NSSM test set up.

demolded after a day of casting and cured in a walk-in chamber at 23 ± 2 ◦C and 98%RH until the desired age of testing.

2.2. Test methods

2.2.1. Non-steady state migration test as per NT Build 49250 mm thick slices were cut from the cylindrical concrete specimens for non-

steady state migration test (NSSM) as per NT Build 492 [8] after the concretes werecured for 28, 56, or 90 days in saturated conditions. Two such specimens from twodifferent cylinders corresponding to the same mixture were used for the trans-port tests, and the values reported are the average values from those two tests. Aschematic of the migration cell is shown in Fig. 1. The catholyte cell was filled with2 N NaCl solution and the anolyte cell with 0.3 N NaOH solution. External electricalpotential was applied axially across the specimen, which forces the chloride ions inthe cell to migrate into the specimen. The voltage to be applied and the total testduration were determined based on the initial current passing through the speci-men when a 30 V potential was applied. After the specified test duration (24 h forall the specimens used in this study), the specimen was axially split and 0.1 N silvernitrate solution sprayed on the freshly split sections. The chloride penetration depthwas then measured from the visible white silver chloride precipitation. At least 7measurements were made in each split side of the sample and the average of 28measurements (14 per specimen) denoted as the penetration depth (xd). The NSSMvalues, Dnssm (in m2 s−1), were calculated by employing the penetration depths inEq. (1):

Dnssm = RT

zFE

xd − ˛√

xd

t(1)

where R is the molar gas constant (8.314 J (K mol)−1), Z is the absolute value of ionvalence (1 for chloride ions), F is the Faraday constant (96,485 C mol−1), T is theaverage value of the initial and final temperatures in the anolyte solution in K, xd

is the average value of the penetration depths in m, and t is the test duration inseconds.

The terms E and ̨ in Eq. (1) are given by:

E = U − 2L

(2a)

˛ = 2

√RT

zFE× erf −1

(1 − 2Cd

Co

)(2b)

where U is the absolute value of the applied voltage in volts, L is the thickness ofthe specimen in m, Cd is the chloride concentration at which white silver chlorideprecipitates (Cd ≈ 0.07 N for plain concrete – same value used for concretes withcement replacement materials also because even appreciable changes in Cd do notinfluence the calculated Dnssm to any considerable degree), and Co is the chlorideconcentration in the catholyte solution (2 N).

2.2.2. Electrical impedance spectroscopyThe bulk electrical resistances (Rb) of the specimens before and after the rapid

chloride migration test were determined using electrical impedance spectroscopy(EIS). EIS measurements were performed on the concrete specimens enclosed in the

measurements after the test were carried out when the cell temperature reachednear ambient levels so that the measured conductivity is not influenced significantlyby the temperature change. EIS spectra were obtained using a SolartronTM 1260 gain-phase analyzer operating at a frequency range of 1 Hz to 10 MHz. A 250 mV AC signal

J. Jain, N. Neithalath / Materials Chemistry and Physics 129 (2011) 569– 579 571

Table 2Mixture proportions for 1 m3 of concrete used in this study.

Mixture ID Cement (kg) Fine agg. (kg) Coarse agg. (kg) Fly ash (kg) Silica fume (kg) VCAS (kg)

Plain 430 743 1058 0 0 0FA 10 387 735 1054 43 0 0FA 20 344 730 1050 86 0 0SF 6 404 738 1053 0 26 0SF 9 391 730 1053 0 39 0VCAS 9 391 735 1050 0 0 43

A of 0.4

wmio

�

wt

3N

tfpcr

3t

otwAcw

Fc

VCAS 15 365 735 1050

ll concretes were proportioned to an effective water-to-cementing materials ratio

as employed and 10 measurements per decade of frequency were recorded. Theeeting point of the bulk and electrode arcs in a Nyquist plot (plot of real vs. imag-

nary impedance) is denoted as the bulk resistance (Rb). The effective conductivityf the specimen (�eff) was calculated as:

eff = L

RbA(3)

here L is the length (50 mm), and A is the cross sectional area (7854 mm2) respec-ively of the specimen.

. Microstructural features before the NSSM test and theSSM coefficients of different concretes

This section deals with the quantification of microstructural fea-ures of plain and modified concretes using a parameter extractedrom electrical impedance test results before the chloride trans-ort test, the non-steady state migration coefficients (Dnssm) of theoncrete mixtures, and their relationship with the measured bulkesistance values.

.1. Microstructural parameter (�ˇ) extracted before the NSSMest

The effective conductivities (�eff) of the concrete specimensbtained from bulk resistances (Rb) are shown in Fig. 2 as a func-ion of their curing durations. The effective conductivity decreases

ith an increase in curing duration for all mixtures as expected.fter 28 days of curing, it can be seen that the fly ash modified con-retes have similar effective conductivities as that of plain concrete,hich can be attributed to the lack of pozzolanic reaction at early

ig. 2. Effective conductivities of the plain and modified concretes as a function ofuring duration.

0 0 65

0.

ages, but at later ages the �eff values are lower than those of theplain concrete. The concretes modified using the high-performancesupplementary cementing materials (silica fume and VCAS) couldbe expected to have lower porosities as compared to the plain con-crete because of the fine filler effect of particles of silica fume andVCAS, and also the consumption of calcium hydroxide (CH) to formadditional C–S–H gel. This reduces the effective conductivities ofthese specimens significantly as compared to the plain concrete.The modified concrete with 9% cement replacement by silica fumeshows the lowest values of �eff at all ages. The VCAS modified con-cretes also show much lower conductivities than the plain and flyash modified mixtures. The specimens containing higher dosagesof VCAS and silica fume are seen to behave similarly at later agesas far as �eff is considered. A comparison of the hydration behaviorand the porosities of cement pastes containing VCAS or silica fumecan be found in [16].

The effective conductivity of concrete can be related to the poresolution conductivity (�pore), pore volume fraction (�), and poreconnectivity factor (ˇ) according to the well-known relationship:

�eff = �pore� ̌ (4)

The pore solution conductivities (�pore) of all the concretes usedin this study were calculated using the equivalent ionic conductiv-ities of the highly conductive species in the pore solution (OH−,Na+, K+), the ionic concentrations, and the species valences [18].The ionic concentrations were obtained from the cement and/orsupplementary cementing materials content, the oxide composi-tions of these materials, w/cm, and the degrees of hydration. Thedegrees of hydration of companion cement pastes containing thesame proportions of replacement materials as the concretes weredetermined using a model based on measured non-evaporablewater contents, the details of which have been published elsewhere[16,19]. The values of �pore thus determined for all the mixturesare given in Table 3. The pore solution conductivities can be seen toincrease with an increase in curing duration. Plain concrete exhibitsthe highest pore solution conductivity at all ages whereas the �pore

values are the lowest for silica fume modified concretes. The reasonfor lower �pore values of silica fume modified concretes is elabo-rated elsewhere [20].

From the known values of �eff and �pore, a microstructuralparameter �ˇ (the product of porosity and pore connectivity fac-

tor) can be determined using Eq. (4). When the porosities of theconcretes are determined using methods such as vacuum satura-tion, the pore connectivity factor (ˇ) can be separated out. Suchan attempt is not made in this study because: (i) both the vol-

Table 3Pore solution conductivities for the plain and modified concretes at different ages.

Age (days) Pore solution conductivity (S m−1)

Plain FA 10 FA 20 SF 6 SF 9 VCAS 9 VCAS 15

28 14.94 13.33 12.53 11.18 9.81 14.74 14.5456 15.55 13.74 12.73 11.38 9.81 15.34 14.9490 15.93 14.34 12.93 11.38 10.01 15.55 15.34

572 J. Jain, N. Neithalath / Materials Chemistry and Physics 129 (2011) 569– 579

Fe

uis

abowAotisovtVacccabfiauva

acdadtmt

3

(t4aa

ig. 3. Evolution of microstructural parameter (�ˇ) for different mixtures at differ-nt curing durations.

me fraction and connectivity of the pore network are influentialn moisture and ionic transport, and (ii) the porosities of all thepecimens were not determined after the chloride transport tests.

Fig. 3 shows the microstructural parameter � ̌ of the plainnd modified concrete mixtures as a function of curing durationefore the ingress of chloride ions. Reduction in the values of � ̌ isbserved with an increase in curing duration for all the specimens,hich provides a quantification of microstructure densification.t early ages, the fly ash modified mixtures show higher valuesf � ̌ as compared to plain concrete, which can be attributed tohe dilution effect of fly ash at these ages. After 56 days of cur-ng, the fly ash modified concrete with 20% cement replacementhows a lower value of � ̌ than plain concrete, and after 90 daysf curing both 10% and 20% fly ash modified mixtures show loweralues of � ̌ indicating the effectiveness of the pozzolanic reac-ion in pore structure refinement. The � ̌ values of silica fume andCAS modified mixtures are considerably lower than those of plainnd fly ash modified concretes at all ages. The � ̌ values of con-rete containing 6% silica fume is found to be similar to that of theoncrete containing 9% VCAS, and that of 9% silica fume modifiedoncrete similar to that of 15% VCAS modified concrete, especiallyt later ages. The effectiveness of these replacement materials cane adequately captured from the observation that 15% VCAS modi-ed concrete and 9% silica fume modified concrete have �ˇ valuesbout 3 times lower than those of the plain concrete at all ages. These of �eff values along with the pore solution conductivities pro-ides a reliable means of quantification of microstructural changess function of curing duration.

It can also be seen from Fig. 3 that the � ̌ values of plain and flysh modified concretes, and those of silica fume and VCAS modifiedoncretes form separate groups. The differences in microstructureue to the addition of a moderately active pozzolanic material (flysh) and highly active pozzolans (VCAS and silica fume) are imme-iately obvious. Since � ̌ is obtained by normalizing �eff with �pore,he influence of lower pore solution conductivity of silica fume

odified mixtures is not reflected in �ˇ, and it is a true microstruc-ural parameter.

.2. Non-steady state migration coefficients

Table 4 shows the non-steady state migration coefficientsDnssm) for the plain and modified concretes cured for various dura-

ions, along with the applied electrical potentials based on NT Build92 (depending on the initial currents when a potential of 30 V ispplied). All the modified concretes show lower values of Dnssm atll ages relative to the plain concrete, indicating better resistance

Fig. 4. Relationship between bulk resistance before NSSM test and Dnssm for plainand modified concretes.

to chloride ion penetration. Fly ash modified mixtures show lowerDnssm values at later ages and at higher replacement levels. The sil-ica fume and VCAS modified concretes show much lower values ofDnssm as compared to those of plain and fly ash modified concretes.This is in line with the higher pozzolanic activity of these materi-als [16]. The Dnssm values of the modified concretes are lower thanthose of plain concrete by a factor similar to the ratio of their �ˇvalues. It can be noticed from Table 4 that, at later ages, higher elec-trical potentials (40 V or 60 V instead of 30 V) were used for silicafume and VCAS modified concretes in the NSSM test because of theirdenser microstructure, which could influence the measured bulkresistance of the specimens after the test, and the microstructure.This aspect is discussed in detail later.

Fig. 4 shows the relationship between bulk resistance (Rb) ofthe specimen before the start of the NSSM test and the Dnssm val-ues for the plain and modified concretes at all ages. Two distinctrelationships can be noticed in this figure: one for the plain andsilica fume modified concretes, and the other for concretes con-taining fly ash and VCAS. For a particular value of bulk resistance,the Dnssm values are found to be lower for the fly ash or VCAS modi-fied mixtures. As can be seen from Table 1, fly ash and VCAS containmuch higher amount of aluminates which could be responsible forbinding the chloride ions and forming chloroaluminate compoundsthat get deposited on the pore structure, thus reducing the migra-tion coefficients [12,21]. Chloride binding results in the alterationof the pore solution concentration and produces changes in thepore structure of concrete. These factors strongly influence the rateof chloride ion penetration into concrete. In a later section of thispaper, electrical impedance spectra along with equivalent circuitmodeling is used to quantify the microstructural changes occur-ring in different concrete mixtures as a result of their exposure tonon-steady state migration test conditions.

The relationship between the microstructural parameter �ˇbefore the start of the NSSM test and Dnssm is depicted in Fig. 5.A linear relationship is observed between these parameters, with

Dnssm increasing with an increase in �ˇ. The relationship is strongerat lower values of � ̌ (i.e., at later ages) for all the mixtures, andless strong at early ages especially for the plain and fly ash mod-ified mixtures. This is because at early ages, the pore structure is

J. Jain, N. Neithalath / Materials Chemistry and Physics 129 (2011) 569– 579 573

Table 4Applied voltages and Dnssm values at 28, 56 and 90 days of curing for all concretes.

Replacement material % of cement replacement Applied voltage in NSSM testa, V Dnssm (×10−12), m2 s−1

28 days 56 days 90 days 28 days 56 days 90 days

None 0 30 30 30 9.94 8.62 8.57Fly ash 10 30 30 30 9.15 5.84 5.39Fly ash 20 30 30 30 7.20 4.88 4.53Silica fume 6 40 60 60 5.14 3.50 3.47Silica fume 9 40 60 60 3.95 2.90 1.65VCAS 9 30 40 40 5.75 4.78 1.96

l is ap

lpii

3r

tmebowsemqiNotttsi

Ft

VCAS 15 30

a Applied voltages are based on the initial current recorded when a 30 V potentia

ess refined for the plain and fly ash modified concretes (whereozzolanic reaction has yet started), leading to larger chloride ion

ngress and consequently increased formation of binding productsn these mixtures.

.3. Electrical circuit model for material microstructure andelevance of model parameters

The Nyquist plots obtained from EIS contain several features ofhe material microstructure that are otherwise difficult to deter-

ine. Previous studies have used different types of equivalentlectrical circuit models to represent the Nyquist plots of cement-ased materials. Equivalent electrical circuits are a combinationf resistors, capacitors, and/or constant-phase elements, whichhen assembled in certain configurations, provide reliable repre-

entations of the pore structure of the material. Among severalquivalent electrical circuits used in the study of cement-basedaterials, the one depicted in Fig. 6(a) has been shown to ade-

uately capture the material structure [22–24], and hence is usedn this study. This model accounts for only the bulk arc in a typicalyquist plot. The resistance Re in the circuit denotes the resistancef electrolytes between the electrodes and the concrete sample inhe NSSM test set up (Fig. 1). The resistance Rc is associated with

he connected pores in the concrete (percolating pores), while Ruc ishe resistance of the unconnected or isolated pores in the materialtructure. C1 is the dielectric capacitance related to the solid phasen the concrete (paste and the aggregates), and C2 is the capacitance

ig. 5. Relationship between the microstructural parameter (�ˇ) before the NSSMest and Dnssm.

40 40 3.33 2.83 1.60

plied.

associated with the double layer present between the pore wallsand the pore solution. Ri and C3 denote the resistance and capac-itance of the specimen–electrode interface. The total frequencydependent impedance Z(ω) of the system can be represented as:

Z(ω) = Re + Z1Z2

Z1 + Z2+ Z3 (5)

Z1, Z2, and Z3 are the impedances of the element groups in the cir-cuit. The impedances Z1 and Z2, belonging to the Rc–C1 and Ruc–C2combinations respectively (bulk part of the system) are denoted as:

Z1 = Rc

1 + (jωRcC1)˛ (6a)

Z2 = Ruc[1 + (jωRucC2)−ˇ] (6b)

The terms ̨ and ̌ are the dispersion factors. The equivalent cir-cuit model parameters were extracted from the impedance spectrausing ZViewTM software. Fig. 6(b) shows the EIS spectra and repre-sentative fits obtained from the equivalent electrical circuit for thebulk arc for plain concrete, and two selected modified concretes. Itcan be seen that the fits are in good agreement with the measuredimpedance spectra.

Rc is the most important parameter that is used from the equiv-alent circuit model here because transport is dominated by theconnected or percolating pores. Since Rc represents the connectedpores in the system, and the microstructural parameter � ̌ repre-sents the combination of the porosity and pore connectivity, it isonly natural to expect good relationships between these parame-ters. Fig. 7 shows the variation of � ̌ determined before the NSSMtest with Rc (also determined before the NSSM test), in which twoseparate relationships are observed, one for the plain and silicafume modified concretes, and the other for fly ash and VCAS modi-fied concretes. For the same value of Rc, the � ̌ values of fly ash andVCAS modified mixtures are lower than those of plain and silicafume modified concretes. This is similar to the Rb–Dnssm relation-ships shown in Fig. 4. Based on these results, it is believed thattracking the changes in Rc with chloride transport through equiv-alent circuit models can provide reliable indicators of the changesin microstructure (�ˇ) brought about by non-steady state chloridemigration.

4. Microstructural changes as a result of NSSM testing

4.1. Changes in bulk resistance due to non-steady state chlorideion migration

Table 5 shows the bulk resistances (Rb) of all the specimensbefore and after the migration test determined using EIS. Asexpected, the bulk resistance of plain concrete is the lowest and

the silica fume concretes is the highest at all curing durations. Ingeneral, the bulk resistances of the samples are found to be higherafter the migration test, with the X-intercept of the Nyquist plotshifting to the right after the NSSM test. The increase in resistance

574 J. Jain, N. Neithalath / Materials Chemistry and Physics 129 (2011) 569– 579

Fig. 6. (a) Equivalent electrical circuit used for the analysis of EIS measurements, and (b) EIS spectra and fits using the equivalent electrical circuit.

Table 5Bulk resistance of plain and modified concretes before and after the migration test.

Replacement material Bulk resistance (Rb), �

28 days 56 days 90 days

Before NSSM test After NSSM test Before NSSM test After NSSM test Before NSSM test After NSSM test

Plain 429 462 485 520 606 674FA 10 431 460 489 557 777 717FA 20 448 522 654 719 982 1074SF 6 819 910 1228 1248 1723 1628a (1811)SF 9 1667 1674 2621 2893 3241 2748a (3533)VCAS 9 610 726 788 975 1336 1083a (1662)VCAS 15 922 1053 1419 1788 2360 2354a (2364)

sons dT

cbdnibf

Ft

a Unexpected behavior where Rb decreases after the NSSM of chloride ions – reahe corrected values are shown in brackets where necessary.

ould be attributed to the formation of new solid products aidedy the ingress of chloride ions or the replacement of highly con-uctive OH− ions by the less conductive Cl− ions. However, in the

on-steady state test time interval, it is less likely that the OH−

ons are removed from the specimen [6,25]. Hence the increase inulk resistances after the NSSM test could be an indication of theormation of new solid products either within the pores or as elec-

ig. 7. Relationship between � ̌ and the resistance of connected pores (Rc) beforehe NSSM test.

escribed in Section 4.1 and a methodology for correction described in Section 4.2.

trochemical double layers along the pore walls [26]. When highconcentrations of external chloride solutions (>1 M) were used, ithas been shown that the amount of bound chlorides was similar innon-steady state migration and natural diffusion experiments on aparticular concrete [27] even though the mechanism of interactionbetween the chlorides and the hydrated cement phases, and therates of reaction were different. Based on the high concentrationof external chloride solution used for the NSSM test in this study(2 M), and the fact that chloride binding is found to be similar tothat in diffusion test in such cases, it can be postulated that theincrease in bulk resistance is a result of the formation of chloridebinding products. It needs to be mentioned here that in anotherstudy that used forced ionic migration [22], the bulk resistance wasfound to decrease for the first 100 h, but in that case, the externalchloride concentration was only 1 M and the driving potential only120 V m−1. The influence of a higher external chloride concentra-tion (2 M) that will facilitate increased binding as shown in [27] isevident from these results.

From Table 5, it can be noticed that for the high-performanceconcrete mixtures (silica fume and VCAS modified concretes) curedfor 90 days, the bulk resistances after the test are found to belower, which is contrary to what is expected. These specimenshave a much denser microstructure and have undergone migrationtests under higher applied voltages. Increased applied voltages canresult in increased chloride ion penetration, thus causing a reduc-tion in measured bulk resistance. In order to facilitate comparisonbetween all the specimens and their microstructure when sub-jected to chloride ingress, it is necessary that the applied voltages

are the same. Since the NSSM tests were carried out in this studyusing the NT Build 492 protocol that necessitates different appliedvoltages based on the initial current, a methodology is developedin this paper that adjusts all the measured bulk resistances to the

J. Jain, N. Neithalath / Materials Chemistr

Fd

eT

4a

idaailTv

v

wn(a

dtdsmamdpw3tifd(

ig. 8. Relationship between the experimental and calculated chloride penetrationepths.

quivalent of those when a 30 V (600 V m−1) potential is applied.his procedure is explained below.

.2. New bulk resistances corresponding to a 30 V (600 V m−1)pplied potential

Increase in applied electrical potentials drives more chlorideons into the specimen, thereby increasing the ion penetrationepth. Hence the first step is to determine the penetration depthsnalytically for the cases where only a 600 V m−1 (30 V) potential ispplied as against the 800 V m−1 (40 V) or 1200 V m−1 (60 V) appliedn the experiments. This is accomplished by multiplying the calcu-ated drift velocities of the chloride ions with the testing time (24 h).he magnitude of the drift velocity is calculated from a modifiedersion of the Einstein equation as [28]:

d = zeED

kT(7)

here z is the ion valence, e is the electron charge (C), E is the mag-itude of the applied electric field (V m−1), D is the bulk diffusivitym2 s−1), k is the Boltzmann constant (1.38 × 10−23 m2 kg s−2 K−1),nd T is the absolute temperature (K).

Fig. 8 shows the relationship between the chloride penetrationepths calculated from the drift velocities using the same elec-rical potential that was applied in the experiments, and thoseetermined using colorimetric techniques (spraying 0.1 N AgNO3olution) from the specimens that underwent the migration experi-ents (see Section 2.2.1). Though the calculated penetration depths

re slightly lower than the experimental values, a reasonable agree-ent between these values suggests that the penetration depths

etermined from vd can provide reliable indications of the actualenetration depth. Based on this observation, the drift velocitiesere recalculated using an E value of 600 V m−1 (corresponding to

0 V) and the new penetration depths (xd)cal-30V determined forhose specimens subjected to a potential of greater than 600 V m−1

n the NSSM tests. Table 6 shows the penetration depths obtainedrom the NSSM experiments, (xd)exp, and those calculated usingrift velocities and an applied electrical potential of 600 V m−1,xd)cal-30V. A significant observation from Fig. 8 and Table 6 con-

y and Physics 129 (2011) 569– 579 575

cerns the (xd)exp values of silica fume and VCAS modified concretes.The chloride penetration depths (experimental) for VCAS modifiedconcretes are much lower than those of silica fume modified con-cretes even though the � ̌ values of these mixtures are similar oreven higher for the VCAS modified concretes (see Fig. 3). The influ-ence of aluminate phases in VCAS in binding the chloride ions, andthus reducing the penetration depths is obvious from these results.A similar trend is found when plain and fly ash modified concretesare compared. The recalculated penetration depths, (xd)cal-30V, arefound to be lower than the experimentally measured ones in thecase of specimens that were subjected to a higher potential in theNSSM tests. The bulk resistances can then be adjusted based on thenew penetration depths, as detailed below.

Using the actual experimentally applied voltages and the actualmeasured bulk resistances, the actual experimental currents (iexp)can be determined for the specimens before and after the NSSMtests. The difference between the iexp values before and after theNSSM test (�iexp) was multiplied by the ratio of the penetrationdepths [(xd)cal-30V/(xd)exp] to account for the corrected differencein current because of a 30 V (600 V m−1) potential (�ical-30V). Thenew current after the NSSM test when a new voltage (Vnew = 30 V)is applied, (inew-30V)after NSSM, can be expressed as:

(inew-30V)after NSSM = Vnew

(Rb)after NSSM− �ical-30V (8)

The new bulk resistance corresponding to the applied voltage of30 V (Rb-30V) is:

(Rb-30V)after NSSM = Vnew

inew-30V(9)

Substituting Vnew (30 V) from Eq. (8) into Eq. (9), and simplifying,

(Rb-30V)after NSSM = (Rb-V)after NSSM

[1 + �ical-30V

inew-30V

](10)

The values of (Rb-30V)after NSSM are the new bulk resistances afterthe NSSM test if only 600 V m−1 potentials were applied insteadof 800 V m−1 or 1200 V m−1 for certain specimens in the NSSMtest. The simplified set of calculations given above provides a rapidmeans of calculating the bulk resistances corresponding to a differ-ent applied voltage other than that from the experiment.

When the bulk resistances of the specimens after subject-ing to prescribed voltages (based on NT Build 492) in theNSSM test, (Rb-V)after NSSM, and those values recalculated basedon a 30 V applied potential, (Rb-30V)after NSSM, are compared, the(Rb-30V)after NSSM values are always higher than the resistancesobtained when these specimens were actually subjected to 40 Vor 60 V potential. The four values marked by ‘a’ in Table 5 also arehigher than the corresponding Rb values before the NSSM test whenthe above described corrections were incorporated. The adjustedvalues of bulk resistances (Rb-30V)after NSSM were used to calcu-late the final effective conductivities for the specimens after themigration test, which can be used in the determination of effectiveconductivity after the NSSM test, (�eff)after NSSM. The resistances ofconnected pores after the NSSM test, (Rc)after NSSM, obtained fromthe equivalent electrical circuit model were also modified in pro-portion to the new bulk resistance values.

Fig. 9 shows the relationship between the calculated bulkresistances (Rb-30V)after NSSM and the calculated penetration depth((xd)cal-30V) for all the specimens. As expected, lower penetrationdepths are observed for specimens with higher bulk resistance. The(Rb-30V)after NSSM–xd relationship also follows a trend noticed for

the Rb–Dnssm and Rc–� ̌ relationships (Figs. 4 and 7), showing theinfluence of specimen composition on the chloride ion transportbehavior as detected by an electrical property (Rb or Rc), and thematerial microstructural feature (�ˇ).

576 J. Jain, N. Neithalath / Materials Chemistry and Physics 129 (2011) 569– 579

Table 6Penetration depths at applied experimental voltages and at 30 V for all the specimens.

Mixture type 28 days 56 days 90 days

(xd)exp, mm (xd)cal-30V, mm (xd)exp, mm (xd)cal-30V, mm (xd)exp, mm (xd)cal-30V, mm

Plain 19.8 19.8 17.3 17.3 17.2 17.2FA 10 18.3 18.3 11.7 11.7 10.8 10.8FA 20 14.4 14.4 9.77 9.77 9.10 9.10SF 6 13.1 9.80a 13.9 6.96a 13.8 6.90a

SF 9 10.6 7.91a 11.7 5.83a 6.56 3.28a

VCAS 9 11.5 11.5 12.6 9.48a 5.22 3.91a

VCAS15 6.65 6.65 7.48 5.61a 3.99 2.99a

N tentiatial.

4a

twbbhftpat6dfiptgssm

Fp

o difference in the experimental and calculated depths when the experimental poa Changed penetration depths after correcting for 600 V m−1 (30 V) applied poten

.3. Equivalent electrical circuit model parameters before andfter the NSSM test

The variation in electrical circuit parameters (described in Sec-ion 3.3) can also provide indications of the microstructural changesith chloride transport. It was shown in the earlier section that the

ulk resistances after the NSSM test were higher than the valuesefore the test. The Rb values after the NSSM test were found to beigher, on an average, by about 9% when compared to the values

or the pristine specimens. Fig. 10(a) shows the values of the resis-ance of connected pores (Rc) and the resistance of unconnectedores (Ruc) obtained from the equivalent circuit model before andfter the electromigration test in the non-steady state. The resis-ance of connected pores is found to increase by an average of about% after the NSSM test and the resistance of unconnected poresecreases by about 15%. The increase in Rc can be attributed to theormation of solid products during the migration process. When Rc

ncreases, i.e., when the number and connectivity of the connectedores decrease, it is essential that the number and connectivity ofhe unconnected pores increase since some of the connected poreset converted into unconnected pores through the formation of

olid products. When the capacitance of the solid phase (C1) is con-idered, it can be seen from Fig. 10(b) that the increase after theigration test, attributable to new solid product formation, is of

ig. 9. Relationship between corrected bulk resistance after the test and calculatedenetration depths for an applied electrical potential of 30 V.

l is 30 V.

the order of 6%, similar to the increase in Rc values. This is expectedsince the changes in both Rc and C1 describe the changes in thebulk of the specimen. However, when the values of C2, the inter-facial capacitance associated with the double layer between thepore walls and the pore solution, before and after the NSSM testsare observed, it can be found that there is a much larger increase(∼47%) in its values after the test, as shown in Fig. 10(b). This is inline with previous studies that have reported that the increase in C2values point to the formation of new solid products predominantlyat the pore walls, resulting in an increase in the surface area of thepores [22,23], and thus influencing the transport lengths. In fact,when the effective resistances of the connected and unconnectedpores [(RcRuc)/(Rc + Ruc)] from the model shown in Fig. 6(a) are con-sidered, it was found that there is no difference at all betweenthe values before and after the NSSM test (Reff-after = 1.01 Reff-before)because an increase in Rc is accompanied by a decrease in Ruc. Sinceelectrical resistance is a bulk quantity, it can be stated that the over-all porosity, which also is a bulk quantity related to the resistance,does not change significantly during the non-steady state migra-tion process. Experimental evidence for this fact has been presented[29] where it has been shown that the change in total porosity isconsiderable only in the anolyte side during a migration test. Sincethe chloride ions would not have reached the anolyte in a non-steady state test, it can be safely stated that the overall porosity ofthe specimens does not change significantly. The pore connectivity(or tortuosity) could then be the microstructural parameter thatchanges considerably, as evidenced by the significant increase inthe interfacial capacitance C2 after the NSSM test. The changes in Rc,Ruc, C1, and C2 when subjected to forced ionic migration shown inFig. 10 confirm that the non-steady state migration test carried outin accordance with NT Build 492 results in the formation of chloridebinding products that changes the material microstructure.

4.4. Microstructural parameter � ̌ after the NSSM test

4.4.1. � ̌ from Rc derived from the equivalent circuit modelThe excellent relationships between the microstructural param-

eter � ̌ before the test and the resistance of connected pores (Rc)before the test are shown in Fig. 7. Since the changes in Rc influ-ence the resistance of the unconnected pores (Ruc) also as describedin the previous section, Rc is considered as the equivalent circuitparameter that can be easily related to the microstructural changes.The �ˇ values after the NSSM tests are obtained by using the sameRc–� ̌ relationships before the NSSM tests (Fig. 7) along with theRc values after the test. Fig. 11(a) shows the relationship betweenthe � ̌ values before and after the NSSM test for all the specimensused in this study. For the plain and silica fume modified mixtures,the � ̌ values after the test are about 96% of those before the test

whereas for the fly ash and VCAS modified mixtures (containinghigher amount of aluminates), the � ̌ values are further lower afterthe test (∼87%). The use of equivalent circuit models as described inthis paper provides a convenient means to quantify the microstruc-

J. Jain, N. Neithalath / Materials Chemistry and Physics 129 (2011) 569– 579 577

tances

tmtcbh

4c

tsf

tif

t

wtii

ttfc

Fc

Fig. 10. (a) Resistances of the connected and unconnected pores, and (b) capaci

ural changes that occur in concretes as a result of rapid chlorideigration. The microstructural densification, especially the forma-

ion of chloride binding products in the pore walls (as shown by thehanges in the interfacial capacitance C2), is adequately capturedy the reduced � ̌ values, particularly for the mixtures containingigher amounts of aluminates.

.4.2. � ̌ from Rb (specimen conductivity) and pore solutiononductivity after the NSSM test

� ̌ values after the NSSM test were calculated using Eq. (4), usinghe �eff values obtained from (Rb-30V)after NSSM. The changed poreolution conductivity (�pore) due to ionic ingress is calculated asollows.

In the electrical migration test, all the ions in the solution par-icipate in the current transport and their individual contributions given by their transference numbers (ti) determined using theollowing equation [30,31]:

i = �iziCi∑�iziCi

(11)

here �i is the equivalent ionic conductivity (cm2 S mol−1), Zi ishe valence, and Ci is the concentration (mol l−1) respectively of theon i. The equivalent ionic conductivities of the major ionic speciesn concrete pore solution can be found in [18].

Since this paper is concerned about the changes occurring in

he non-steady state, the upstream cell (cathode) reactions andhe ionic movement occurring along the specimen–catholyte inter-ace are considered to be more prominent than the downstreamell (anode) reactions where the chloride ions would not have

ig. 11. Microstructural parameter (�ˇ) before and after the NSSM test: (a) from resisonductivity.

of the solid phase and the pore–solid interface, before and after the NSSM test.

reached in the time duration (24 h) under study. The experimen-tal chloride penetration depths shown in Table 5 indicate that inno case has the chlorides penetrated through the entire specimen,i.e., the breakthrough time is much larger than the test duration of24 h.

Considering the cathodic cell of the test set up (Fig. 1), the ionicspecies present are Na+ and Cl−. In addition, the electrode reac-tion produces OH− ions also. Thus the transference number in thecathode can be written as:

tNa+ + tCl− + tOH− = 1 (12)

The concentrations of Na+ and Cl− ions in the catholyte (2 N NaClsolution) are known. If the total charge passed during the NSSM testis Q (given as

∫i(t)dt, where i(t) is the current passing through the

specimen at time t), then Q/F (where F is the Faraday’s number)moles of OH− ions are produced at the cathode. To maintain elec-troneutrality of the cathodic solution, [(tNa

+) · Q/F] moles of Na+ ionswill leave the specimen and move to the catholyte while (tCl

−) · Q/Fmoles of Cl− ions will migrate towards the concrete sample fromthe cathode. Though this equation for transference numbers forthe cathodic cell is generally valid for steady state conditions, it hasbeen shown that, if there is an unlimited source of chloride ions inthe catholyte (catholyte concentration typically greater than 1 M),and if a significant amount of charge has been passed, this equationcan be used for the non-steady state also [30].

The initial concentrations of all the ionic species present in theconcrete pore solution were obtained as described earlier. Theseconcentrations were then modified by accounting for the amountof Cl− ions that have penetrated the specimen and the amount of

tance of connected pores, and (b) from effective conductivity and pore solution

5 emistr

Ncocmuiriabwtip

5

ccmata

(

(

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

78 J. Jain, N. Neithalath / Materials Ch

a+ ions that were removed from the specimens. The changed con-entrations were used in an equation that uses the weighted sumf equivalent conductivities of each ionic species [18] to obtain thehanged pore solution conductivities after the migration test. Theicrostructural parameter � ̌ after the NSSM test can be calculated

sing this pore solution conductivity and the effective conductiv-ty determined from corrected bulk resistance (Rb-30V)after NSSM. Theelationship between the � ̌ values before and after the NSSM testss shown in Fig. 11(b). It is seen from this figure also that � ̌ valuesfter the NSSM test are only about 90% of the corresponding valuesefore the test, indicating changes in the material microstructurehen subjected to non-steady state chloride ion ingress. The dis-

inction between the mixtures based on their aluminate contentss less obvious in this case, possibly due to the uncertainties in theore solution conductivity estimation after the NSSM test.

. Conclusions

This paper has dealt with the quantification of microstructuralhanges that occur in concretes containing different supplementaryementing materials when subjected to non-steady state chlorideigration tests. Electrical impedance spectroscopy (EIS) and an

ssociated equivalent electrical circuit model were used to extracthe relevant parameters of interest. The following conclusions arerrived at from this study:

(i) The microstructural parameter (�ˇ) obtained from the elec-trical conductivity data before the start of the NSSM testclearly brought out the relative effectiveness of moderate andhigh-performance pozzolanic cement replacement materialsin microstructural densification. The fly ash and VCAS modifiedconcretes, because of their higher aluminate contents, showedlower Dnssm values at the same bulk resistances as comparedto plain and silica fume modified mixtures. When the Dnssm

values were related to the microstructural parameter (�ˇ), afairly linear relationship was noticed, with the Dnssm valuesincreasing with increasing �ˇ.

(ii) The resistance of the connected pores (Rc) from the equiva-lent electrical circuit model was considered as one of the mostimportant model parameters because the ionic transport is pri-marily governed by the connected pores in the system. Theuse of Rc to track the changes in � ̌ was found to be a reliablemethod to follow the microstructural changes induced becauseof accelerated chloride transport tests in concretes.

iii) A methodology employing the use of theoretical penetrationdepths along with the actual experimental currents, measuredbulk resistances, and the applied voltages was used to obtainthe Rb values corresponding to an applied potential of 30 V[(Rb-30V)after NSSM]. This was required because the applied volt-ages in the NSSM test were different based on the initial currentpassed when a 30 V potential was applied. The (Rb-30V)after NSSMvalues were used to determine the effective electrical con-ductivity of the specimens after the NSSM test. The values ofthe resistance of connected pores (Rc) and the capacitance ofthe solid phase (C1), which were higher after the NSSM test,indicate the formation of solid products as a result of chloridetransport. The large increase in the capacitance related to thepore–solid interface (C2) after the NSSM test suggested thatthe chloride binding products are formed in the interface andinfluence the tortuosity (or connectivity, ˇ) of the transportpaths.

iv) Microstructural parameter � ̌ of the specimens after the NSSMtest was determined by relating it to the Rc values after the testfrom EIS models, similar to the relationship between Rc and �ˇbefore the NSSM test. The specimen conductivities obtained

[

y and Physics 129 (2011) 569– 579

from Rb-30V values were also used along with pore solutionconductivities after the NSSM test estimated based on ionictransference numbers to obtain the values of � ̌ after the test.Both these methods predicted an average reduction of about10% in the � ̌ values as a result of non-steady state chloridemigration for all the specimens used in this study.

References

[1] N.R. Buenfled, G.K. Glass, A.M. Hassanein, J.-Z. Zhang, Chloride transport inconcrete subjected to electric field, J. Mater. Civil Eng. 10 (1998) 220–228.

[2] H.W. Song, C.H. Lee, K.Y. Ann, Factors influencing chloride transport in concretestructures exposed to marine environments, Cem. Concr. Compos. 30 (2008)113–121.

[3] M. Hisada, S. Nagataki, N. Otsuki, Evaluation of mineral admixtures on the view-point of chloride ion migration through mortar, Cem. Concr. Compos. 21 (1999)443–448.

[4] K.A. Gruber, T. Ramlochan, A. Boddy, R.D. Hooton, M.D.A. Thomas, Increasingconcrete durability with high reactivity metakaolin, Cem. Concr. Compos. 23(2001) 479–484.

[5] L. Tang, L.O. Nilsson, Rapid determination of the chloride diffusivity in concreteby applying an electric field, ACI Mater. J. 89 (1992) 49–53.

[6] T. Zhang, O.E. Gjørv, An electrochemical method for accelerated testing of chlo-ride diffusivity in concrete, Cem. Concr. Res. 24 (1994) 1534–1548.

[7] P.E. Streicher, M.G. Alexander, A chloride conduction test for concrete, Cem.Concr. Res. 25 (1995) 1284–1294.

[8] NT Build 492, Concrete, Mortar and Cement-based Repair Materials: ChlorideMigration Coefficient from Non-steady-state Migration Experiments, NordtestMethod 492, 1999.

[9] C. Andrade, Calculation of chloride diffusion coefficients in concrete from ionicmigration measurements, Cem. Concr. Res. 23 (1993) 724–742.

10] A. Delagrave, J. Marchand, E. Samson, Prediction of diffusion coefficients incement based materials on the basis of migration experiments, Cem. Concr.Res. 26 (1996) 1831–1842.

11] M. Castellote, C. Andrade, C. Alonso, Measurements of the steady and non-steady state chloride diffusion coefficients in a migration test by means ofmonitoring the conductivity in the anolyte chamber: comparison with naturaldiffusion tests, Cem. Concr. Res. 31 (2001) 1411–1420.

12] Q. Yuan, C. Shi, G. De Schutter, K. Audenaert, D. Deng, Chloride binding of cementbased materials subjected to external chloride environment – a review, Constr.Build. Mater. 23 (2009) 1–13.

13] R.K. Dhir, M.A.K. El-Mohr, T.D. Dyer, Developing chloride resisting concreteusing PFA, Cem. Concr. Res. 27 (1997) 1633–1639.

14] C. Arya, N.R. Buenfeld, J.B. Newman, Factors influencing chloride binding inconcrete, Cem. Concr. Res. 20 (1990) 291–300.

15] R. Loser, B. Lothenbach, A. Leemann, M. Tuchschmid, Chloride resistance ofconcrete and its binding capacity – comparison between experiments andthermodynamic modeling, Cem. Concr. Compos. 32 (2010) 31–42.

16] N. Neithalath, J. Persun, A. Hossain, Hydration in high-performance cementi-tious systems containing vitreous calcium aluminosilicate or silica fume, Cem.Concr. Res. 39 (2009) 473–481.

17] S. Shirazi, A. Hossain, J. Persun, N. Neithalath, Properties of concrete containingvitreous calcium aluminosilicate pozzolan, J. Transport. Res. Board 20 (2008)32–38.

18] K.A. Snyder, X. Feng, B.D. Keen, T.O. Mason, Estimating the electrical conductiv-ity of cement paste pore solutions from OH− , K+, and Na+ concentrations, Cem.Concr. Res. 33 (2003) 793–798.

19] N. Schwarz, N. Neithalath, Influence of a fine glass powder on cement hydration:comparison to fly ash and modeling the degree of hydration, Cem. Concr. Res.38 (2008) 429–436.

20] K. Tori, M. Kawamura, Pore structure and chloride ion permeability of mortarscontaining silica fume, Cem. Concr. Compos. 16 (1994) 279–286.

21] J. Jain, N. Neithalath, Chloride transport in fly ash and glass powder modifiedconcretes: influence of test methods on microstructure, Cem. Concr. Compos.32 (2010) 148–156.

22] I. Sánchez, X.R. Nóvoa, G. de Vera, M.A. Climent, Microstructural modifica-tions in Portland cement concrete due to forced ionic migration tests: studyby impedance spectroscopy, Cem. Concr. Res. 38 (2008) 1015–1025.

23] B. Diaz, X.R. Nóvoa, M.C. Perez, Study of the chloride diffusion in mortar: a newmethod of determining diffusion coefficients based on impedance measure-ments, Cem. Concr. Compos. 28 (2006) 237–245.

24] N. Neithalath, J. Jain, Relating rapid chloride transport parameters of concretesto microstructural features extracted from electrical impedance, Cem. Concr.Res. 40 (2010) 1041–1051.

25] C.C. Yang, S.W. Cho, J.M. Chi, R. Huang, An electrochemical method for acceler-ated chloride migration test in cement-based materials, Mater. Chem. Phys. 77(2002) 461–469.

26] M. Siegwart, J.F. Lyness, B.J. McFarland, Change of pore size in concrete due to

electrochemical chloride extraction and possible implications for the migrationof ions, Cem. Concr. Res. 33 (2003) 1211–1221.

27] M. Castellote, C. Andrade, C. Alonso, Chloride-binding isotherms in concretesubmitted to non-steady-state migration experiments, Cem. Concr. Res. 29(1999) 1799–1806.

emistr

[

[

[30] M. Castellote, C. Andrade, C. Alonso, Phenomenological mass-balance-based

J. Jain, N. Neithalath / Materials Ch

28] K.A. Snyder, C. Ferraris, N.S. Martys, E.J. Garboczi, Using impedance spec-

troscopy to assess the viability of the rapid chloride test for determiningconcrete conductivity, J. Res. Natl. Inst. Stand. Technol. 105 (2000) 497–509.

29] B. Diaz, L. Freire, P. Merion, X.R. Novoa, M.C. Perez, Impedance spectroscopystudy of saturated mortar samples, Electrochim. Acta 53 (2008) 7549–7555.

[

y and Physics 129 (2011) 569– 579 579

model of migration tests in stationary conditions – application to non-steadystate tests, Cem. Concr. Res. 30 (2000) 185–1893.

31] W. Prince, R. Gagne, The effects of types of solutions used in accelerated chloridemigration tests for concrete, Cem. Concr. Res. 31 (2001) 775–780.