Influence of Pre-Deformation of 5052H112 Alloy on Tensile Propertiesand Fracture Resistance under Vibration

Kuo-Tsung Huang1,+, Shih-Hsien Chang2 and Truan-Sheng Lui3

1Department of Auto-Mechanics, National Kangshan Agricultural Industrial Senior High School,Kaohsiung 82049, Taiwan, R. O. China2Department of Materials and Mineral Resources Engineering, National Taipei University of Technology,Taipei, 10608, Taiwan, R. O. China3Department of Materials Science and Engineering, National Cheng Kung University,Tainan, 70101, Taiwan, R. O. China

This study prepared a 5052H112 alloy under different cold-rolled reductions to explore its microstructure and mechanical properties. Theexperimental results indicated that the cold-rolled specimens had good tensile properties and better vibration fracture resistance, due to the highdensities of dislocation and small aspect ratio of crystal grain. The stress-elongation curves of all specimens showed the serrated yielding. Thehigh densities of dislocation and small aspect ratio of crystal grain introduced by cold-rolling could hold the mobile dislocations long enough tolet Mg atoms form atmospheres around them. In addition, the crack propagation behavior of all specimens showed that slip bands can beobserved in the vicinity of the main crack and be suppressed by increasing the cold-rolled reduction. Crack propagation showed that rolledspecimens exhibit decreasing crack propagation rates with increased matrix strengthening (cold-rolling). Therefore, it can be concluded that alarge number of dislocation tangles introduced by cold-rolling are effective in improving mechanical properties.[doi:10.2320/matertrans.M2014305]

(Received August 22, 2014; Accepted October 27, 2014; Published January 25, 2015)

Keywords: cold-rolling, serrated yielding, dislocation tangles, fracture resistance

1. Introduction

In the recent years, weight reduction has become a keyissue for automotive manufacturers. For this reason, Al-Mg(aluminum magnesium) alloys (5XXX series) have greatattention in automotive industry due to their excellent highstrength and weight ratio; corrosion resistance, weld abilityand recycling potential. Therefore, aluminum alloys couldreplace heavier materials in the automobiles to reduce theautomobile weight. Meanwhile, requirements for fuel con-sumption, environmental laws, and global warming issueshave significant influence on the choice of the materials.13)

Al-Mg alloys have been used in many transportationsystems that could possibly experience abnormal vibrationprocesses. In particular, severe fracture problems could occuras an applied vibration frequency meets with resonance.46)

Studies of this failure are necessary in order to ensure thereliability of the applications. From our previous reports,6)

the resonant vibration cracks often initiate at different stressconcentration sites on the surface and grow simultaneouslyinto the body of the specimen. This study examined acommercial 5052 Al-Mg alloy to explore its crack prop-agation behaviors under resonant vibration. Tests wereperformed on 5052 Al-Mg alloys, which are classified asnon-heat treated alloys, prior to cold-rolling in order tocontrol normal variations of mechanical properties.

Furthermore, the advantages of this alloy include goodcorrosion resistance, high mechanical properties, and ex-cellent formability. Thus, it can be used in car bodyapplications. Since 5052 Al-Mg alloys have great potentialfor microstructural evolution through deformation processes,and possess enhanced mechanical properties,5) which could

also affect crack propagation behaviors under resonantvibration, it is necessary to study the effects of priordeformation on the 5052 Al-Mg alloys under differentcold-rolled reductions.

Previous reports have pointed out that7,8) high densities ofdislocations and grain boundaries introduced by a pre-straincan enhance the strength and the ability for absorbing thevibration energy, thus improving the crack propagationresistance of Al-Mg alloys. However, the effects of thedifferent pre-strains on vibration behavior of 5052 Al-Mgalloys are still not clear. In this investigation, we aim toclarify the effect of prior deformation of a 5052 Al-Mg alloysbase metal on microstructural features, tensile properties andvibration fracture resistance under different amounts of pre-strain.

2. Experimental Procedures

The material used in this study was hot rolled aluminumalloy 5052H112 (H112 temper is strain hardened as thefinal operation). The chemical composition (mass%) of the5052H112 alloy is listed in Table 1, which is prepared withdifferent cold-rolled reductions (12%, 25% and 50%) alongthe rolling direction at room temperature (25°C). In thisstudy, the base metal was designated as “BM”. The rolledsheets of 12%, 25%, and 50%, were designated as “R12”,“R25”, and “R50”, respectively.

Microstructural features of the specimens were examinedby optical microscopy and transmission electron microscopy(TEM). The thin-foil disk specimens of HF2000 FE-TEMwere cut 3mm in diameter from the matrix of the differentcold rolled specimens and were then jet electropolished byan electrical discharge machine in a 50ml HNO3 + 200mlCH3OH solution at 248K. To study the Vickers hardness+Corresponding author, E-mail: morganhuang@pchome.com.tw

Materials Transactions, Vol. 56, No. 2 (2015) pp. 236 to 241©2015 The Japan Institute of Metals and Materials EXPRESS REGULAR ARTICLE

http://dx.doi.org/10.2320/matertrans.M2014305

profile in the vicinity of the main crack, the Vickers indenterwas set to be a 0.98N for 15 s. In tensile tests, the rolledsheets were cut along the rolling direction, and had a strainrate fixed at 8.3 © 10¹4 s¹1. To explore the effects ofmicrostructural features, tensile test specimens, with 15mmgauge in length and 3mm in thickness, were used.

Results of vibration tests on the specimen’s shape anddimension are illustrated in Fig. 1(a). Figure 1(c) schemati-cally depicts the vibration set-up. To generate the vibration,one end of the test specimen was clamped on a vibrationshaker. The two V-notches adjacent to the clamp weredesigned to confine the resonance to a location between them.Deflection of the specimens at the other end, opposite thevibration shaker, was measured using an acceleration sensor.In this study, two kinds of vibration test were performed.First, acceleration of the vibration shaker was fixed at 1.1Gas the vibrational fracture test. The effect of deflectionamplitude on the cold-rolled specimens was investigated.Second, in order to obtain the close surface strain of eachspecimen, the initial deflection was fixed to 6.5mm as thevibrational fatigue test. The crack propagation rate andvibrational fracture behavior near the V notches wereinvestigated. G represents the acceleration of the gravityconstant, which is 9.8m·s¹2. In order to determine theresonance frequency, taken from the one leading to the largestdeflection, the vibration frequency was scanned continuously,as shown in Fig. 1(b). The resonant frequencies for allspecimens were detected as within a range of 42 « 1Hz.

The effects of microstructural evolution of all vibrationspecimens on the crack propagation paths were examined byan optical microscope equipped with an image analyzer. Allthe specimens’ surfaces were finished by polishing with0.3 µm Al2O3, which allowed direct observation of the crackpropagation path throughout the microstructure. Therefore,the crack propagation path could be observed directly on thespecimen surface. To measure the crack propagation rate, thevibration test was temporarily stopped after 60 s of vibrationand the sum of two projected cracks was calculated.Moreover, the quantitative data of crack tortuosity could bedetermined, as shown in Fig. 2. Then, the value of cracktortuosity was defined as the length of the main crack dividedby the projected length of the crack along its macroscopicpropagation direction. The macroscopic crack propagationdirection was parallel to the tip direction of the two V-notches(the notch is shown in Fig. 1(a)). After the vibration test, wecontinued to control the deflection of the specimens until thecrack propagation path was large enough, and then cut off thespecimens along the crack propagation path using a cuttingmachine. The fracture surface of the specimens near theV-notches was observed by SEM.

Unit: mm

(b)(a)

40 42 44 46Frequency, f /Hz

2

4

6

8

Def

lect

ion,

D/m

m

resonant frequency

1. vibration controller 5. specimen

2. acceleration sensor 6. deflection sensor

3. vibration shaker 7. recorder

4. specimen clamp

(c)

Fig. 1 (a) Shape and dimension of the resonant vibration test specimen, (b) resonant frequency, (c) the resonant vibration setup.

main crack length = Lm

branch crack

projected crack length along

notch direction = Lp crack path tortuosity = Lm / Lp

V-notch

Width direction

Fig. 2 Schematic drawing of the cracking features generated by resonant.

Table 1 Chemical composition of the 5052H112 alloy (mass%).

Element Mg Fe Cr Si Mn

(mass%) 2.58 0.26 0.21 0.15 0.1

Influence of Pre-Deformation of 5052H112 Alloy on Tensile Properties and Fracture Resistance under Vibration 237

3. Results and Discussion

3.1 Effect of cold-rolling on microstructure and tensileproperties

Figure 3(a) shows a typical metallographic image of an Al-Mg alloy as a rolled 5052H112(BM) sample. A compara-tively prolonged grain structure was observed in the cold-rolled specimens, as shown in Figs. 3(b), 3(c) and 3(d). Asseen, the prolonged grain of the structure increased withincreased cold-rolled reduction. The aspect ratio (width/length) of grain size in the cross-sections of BM, R12, R25and R50 specimens were significantly different (BM: 0.56,R12: 0.33, R25: 0.20 and R50: 0.14). Moreover, the TEMimages revealed that, in the matrix, there were few particlesbut many dislocations. The density of the dislocationincreases as the cold-rolled reduction was increased, asshown in Fig. 4.

Figure 5 shows the stress-elongation curves. All speci-mens exhibited a similar serrated yielding behavior, but aslight difference in serrated flow was observed. The intensityof the serrated flow will be discussed in the next paragraph.The tensile results indicated that the strength obviouslyincreased as the cold-rolled reduction increased (0.2% offsetyield stress: 106, 205, 237 and 258MPa, respectively, forBM, R12, R25 and R50 specimens). Conversely, the uniformelongation decreased as the cold-rolled reduction increased;however, the decrease was slight for the R25 and R50specimens (BM: 16.7%, R12: 4.7%, R25: 3.0%, R50: 1.9%).It is clear that the effects of rolled specimens were significantimprovement of strength and decrease in elongation. Thesefeatures are shown in the comparison of Figs. 3 and 5.In particular, cold-rolled reduction was very effective inenhancing strength.

The stress-elongation curves (Fig. 5) show the serratedphenomenon, also known as the PortevinLeChatelier effect(PLC effect). PLC effect is the phenomena that solute-dislocation occurs mainly at the obstacles commonlyobserved in all tensile test specimens held. The solute atomsdiffused to these dislocations, causing atmospheres to formaround them. When the deformation continued, the disloca-

tions broke away from these atmospheres and multiplied,which led to serrated yielding.911) According to previousstudies,12) a material with a rolled structure possessedserration anisotropy, whereas a fully re-crystallized materialdid not. This implies that the application of various cold-rolledreductions may result in variations in the serration magnitude.In this study, the high densities of dislocation and small grainsize (a greater area fraction of grain boundaries along the lineof the mobile dislocations), introduced by strain hardening;could hold the mobile dislocations long enough to allow Mgatoms to form atmospheres around them.13) As a result,before necking of the specimen occurred, the PLC effectincreased as strain-hardening increased. Figure 5 shows theserrated yielding behaviors for all specimens. The intensity ofthe serrated flow increased slightly as the cold-rolledreductions increased. This could be ascribed to the cold-rolled reduction. As a result, the increase of the dislocationtangles and the decrease of the grain size, which are the mainsources of obstacles, are as shown in Fig. 4.

3.2 Effect of cold-rolling on vibration fracture resistanceFigure 6 shows the data of deflection amplitude versus the

number of vibration cycles (D-N curves). Figure 6(a) showsthat the D-N curves of all rolled specimens were recordedunder identical accelerations (G = 1.1). This indicates thatthe initial deflection amplitude of the BM specimen was thelowest, and the initial deflection amplitude of the R50specimen was the highest. Previous study indicated that theinitial deflection amplitude is inversely proportional to the

(a)

100μm

100μm

(c) (d)

100μm

(b)

100μm

Fig. 3 Optical microscopic of (a) BM, (b) R12, (c) R25, (d) R50. (Thearrow indicates the rolling direction.)

(a)

0.5μm

(b)

0.5μm

Fig. 4 The TEM images of (a) BM, (b) R50. (The samples didn’t carry outvibration test, and an arrow indicates the rolling direction.)

0 10 20 30 40Elongation (%)

50

100

150

200

250

300

Nom

inal

str

ess,

σ/M

Pa

BM

R12

R25

R50

Fig. 5 Stress-strain curves.

K.-T. Huang, S.-H. Chang and T.-S. Lui238

damping capacity of the specimens under identical accel-erations.4) Comparison of the initial deflection amplitude ofFig. 6(a) with the cold-rolled reduction shows that the initialdeflection amplitude of the specimens was related to the cold-rolled reduction. As a result, the initial deflection amplitudeof the specimens tended to increase as the cold rolledreduction increased. To avoid the effect of damping capacityas far as possible, the vibration force (increase or decrease)of the specimens was controlled to obtain identical initialdeflection amplitudes, thus, allowing measurement of thevibration of the D-N curves under identical initial deflectionconditions (i.e. constant vibration strain conditions). Underthe initial deflection of 6.5mm, all the specimens exhibited acommon trend. With increased vibration cycles the deflectionamplitude increased in the initial ascending stage, thendecreased in the final descending stage. The D-N curve alsoreveals a distinct plateau stage of maximum deflectionamplitude in between the initial ascending stage and the finaldescending stage. However, the resonant vibration character-istics of the specimens were identified from the D-N curves,which can be divided into two stages, as indicated inFig. 6(b): Stage I (initial ascending stage or work hardeningstage) and Stage II (final descending stage). The increasedability in work hardening raised the effective elastic modulus,and the deflection amplitude then increased with a decreasingdamping capacity.14) The work hardening stage was desig-nated as Stage I, and the vibration fracture resistance wasconsidered being strongly dependent on the duration ofStage I. When the vibration cycles were prolonged, there wasa significant decrease in deflection amplitude, which could bedesignated as Stage II. This was resulted from the inwardpropagation of major cracks and the deviation of the actualvibration frequency from the resonant frequency.15) Ourprevious study5) confirmed that Al-2.5Mg specimens showeda deviation of the actual vibration frequency from theresonant frequency when the deflection amplitude wasdecreased from the maximum value due to the reducedeffective cross-sectional area of the test specimen. Therefore,the vibration life in this study can be defined as the vibrationcycle number when the deflection amplitude reached itsmaximum value (i.e. the abovementioned duration ofStage I).

The Al-Mg alloys, with a high amount of Mg content andsevere dislocation tangles, possessed greater tensile deforma-tion resistance, which could be ascribed to the interactionbetween the dislocations and dissolved Mg. Although thehigh amount of Mg content and severe dislocation tanglestended to decrease the damping capacity under an identicalacceleration, the initial ascending deflection was significantlysuppressed under identical initial deflection amplitudes.5)

Therefore, the high Mg content and severe dislocationtangles was advantageous to enhance the crack propagationresistance. Figure 6(b) shows that, during Stage I, thedeflections of all test specimens increased with increasingvibration cycles until reaching the maximum value. In thisstage, the ascending deflection decreased with increased cold-rolling reduction.

Meanwhile, the sum of two projected cracks versus thenumber of vibration cycles is shown in Fig. 7. The crackpropagation rate significantly decreased as the rolledreduction increased. Compared to the BM specimen, thecold-rolled specimens possessed a longer period of maximumdeflection and a greater fracture resistance to crack growth.In particular, the R50 specimen had the longest period ofmaximum deflection, the best crack growth resistance and thelongest vibration life.

Figure 7 also shows that the cracks had already beengenerated in Stage I, and the crack propagation resistance ofthe R50 specimen was the best among all specimens. In thisstudy, crack resistance can be divided into two stages: crackinitiation and propagation. High densities of dislocation andsmall aspect ratio of crystal grain introduced by cold-rollingcould enhance the strength and decrease the work hardeningrate. As a result, the appearance of the initial crack tended tobe slightly slow as the cold-rolling increased (Fig. 7). Inaddition, due to different crack propagation rates, it isconcluded that the change in matrix properties, rather thanthe crack propagation, is the major factor leading to theascending deflection. As supporting evidence of this state-ment, the microhardness data revealed increased workhardening in this stage. During Stage I of ascendingdeflection (Fig. 6 (b)), the microhardness in the vicinity ofthe main crack increased with increasing vibration cycles.Taking the BM specimen as an illustration, the original

0 10 20 30Number of vibration cycles (×1000)

4

5

6

7

8

9

10

Def

lect

ion

ampl

itude

, D/m

m

R50R25R12BM

(a)

0 10 20 30Number of vibration cycles (×1000)

4

5

6

7

8

9

10

Def

lect

ion

ampl

itude

, D/m

m

R50R25R12BM

(b)

Fig. 6 Deflection amplitude versus number of vibration cycles (a) G = 1.1, (b) D = 6.5mm.

Influence of Pre-Deformation of 5052H112 Alloy on Tensile Properties and Fracture Resistance under Vibration 239

microhardness before vibration was 72.3 VHN. It increasedto 88.5 VHN with vibration cycles of 5000 (see Fig. 6(b), itis in the stage of ascending deflection). The final hardnessreached about 95.2 VHN in the stage of maximum deflection.According to McGuire et al.,16) the increase in workhardening could raise the effective elastic modules, thusincreasing deflection. This result agrees with the variation ofn values, as listed in Table 2. It can be concluded thatthe work hardening behavior plays an important role inascending deflection.

The effective elastic modulus and resonance frequencyof a material can be reduced by inducing cracks;14,17) thefrequency can be regarded as the resonant frequency as longas the total crack length is not long enough to have anobvious effect on the reduction of the resonant frequency.Compared with the crack propagation rate (Fig. 7) anddeflection data (Fig. 6(b)), the cracks that continuedpropagating during vibration was induced quite early inStage I of the ascending deflection. Cracking had a greateffect on reducing the cross-sectional area of the test

specimen, thereby reducing the resonant frequency. In avibration test of constant vibrational frequency, such as in thecurrent study, reduction of the resonant frequency by crackssuppressed the deflection since the vibration condition wasno longer resonant. This implies that the phenomenon ofdescending deflection in Stage II (Fig. 6(b)) was due to anincreasing deviation in the test frequency from the resonantfrequency.17)

Cold-rolling induced a large number of dislocation tanglesand decreased aspect ratio of crystal grain (PLC effectincreased), which improved matrix strengthening and reducedthe crack propagation rate. Therefore, prior deformation of the5052 alloys effectively increased crack propagation resist-ance, and thus, increased the vibration period of maximumdeflection. Consequently, matrix strengthening by cold-roll-ing was regarded as one of the main factors affecting vibrationfracture resistance. In addition, the number of dislocationtangles introduced by cold-rolling was greater, which canimprove the tensile strength and decrease the work hardening.However, an ascending deflection of the D-N curve wasreduced significantly under identical initial deflection con-ditions; this is another factor in improving the vibrationfracture resistance.

The vibration fracture resistance was examined byobserving the fracture path. Figure 8 shows the vibrationfracture pattern, and all specimens commonly demonstrate aflat feature that is closely related to slip deformation behavior.This feature can further compare the results of Fig. 9 usingthe evidence of specimen BM. This agrees with findingsshown in Fig. 8(a) where more slip bands cracking in thefracture path can be directly identified.

Figure 9 shows the crack propagation behavior of theBM and R50 specimens. All specimens had similar crackpropagation features, which reveal the appearance of slipbands in the vicinity of the main crack. Cold-rolling induceda large number of dislocation tangles and decreased aspectratio of the crystal grain, which suppressed the mobiledislocations in Fig. 9 and reduced the crack tortuosity inTable 2. The slip bands in the vicinity of the main crackdecreased with increasing rolled reduction. Our previousstudy confirmed that18) a small grain size can increasevibration fracture resistance resulting from plastic deforma-tion in the vicinity of the main crack. According to above-mentioned observations, it is reasonable to suggest that thecold rolling plays an important role in affecting vibrationpropagation behavior.

Table 2 The crack path of tortuosity and n value of the test specimens.

Specimen BM R12 R25 R50

Crack path of tortuosity 1.26 1.23 1.17 1.10

n value 0.35 0.09 0.07 0.05

0 10 20 30 40Number of vibration cycles (x1000)

0

1

2

3

4

Proj

ecte

d cr

ack

leng

th (

L/m

m)

R50R25R12BM

Fig. 7 The sum of two projected cracks versus number of vibration cyclesunder resonant vibration.

(b)

5μm5μm

(a)

Fig. 8 SEM micrographics of the fracture surface for (a) BM, (b) R50. (The arrow indicates the crack propagation direction.)

K.-T. Huang, S.-H. Chang and T.-S. Lui240

Using an optical microscope, crack propagation paths ofall surface etched specimens were observed as intra-granularfractures on Fig. 10. A significant difference in vibrationfracture resistance was correlated with the changes of cracktortuosity resulting from matrix strengthening by cold-rolling. Thus, pre-deformation plays an important role inimproving the vibration fracture resistance of the 5052 H112Al-Mg alloy.

4. Conclusions

According to the experimental results discussed in theprevious sections, the following conclusions can be drawn:cold-rolling, which is effective in improving tensile strength,is also significant in reducing the elongation of non-heattreatment of a 5052H112 Al-Mg alloy. The vibrationaldeflection increased initially with increasing vibration cycles.During this stage of ascending deflection, a crack had alreadybeen generated. Meanwhile, the microhardness at this regionnear the main crack also showed ascending behavior. Afterthe initial ascending deflection (Stage I), the main crack grewto a critical length, the effective cross-section area of thespecimens reduced. All specimens deviated from the actualvibration frequency resulting in the deflection amplitudedecreasing from its maximum.

Cold-rolling has a great effect on increasing this period ofmaximum deflection under identical initial deflection con-ditions. A large number of dislocation tangles introduced bycold-rolling can reduce the work hardening rate and crackpropagation rate, thereby improving the vibration fractureresistance of a 5052H112 Al-Mg alloy.

Acknowledgements

The publication charge payment is supported by ASSABSTEELS TAIWAN CO., LTD. In addition, the authorswould like to express their appreciations for Professor Li-HuiChen.

REFERENCES

1) D. Li and A. Ghosh: Mater. Sci. Eng. A 352 (2003) 279286.2) W. S. Miller, L. Zhuang, J. Bottema, A. J. Wittebrood, P. De Smet, A.

Haszler and A. Vieregge: Mater. Sci. Eng. A 280 (2000) 3749.3) F. Ozturk, S. Toros and S. Kilic: Arch. Mater. Sci. Eng. 34 (2008) 95

98.4) D. S. Jiang, T. S. Lui and L. H. Chen: Scr. Mater. 36 (1997) 1520.5) K. T. Huang, T. S. Lui and L. H. Chen: Mater. Trans. 46 (2005) 2268

2275.6) K. T. Huang, T. S. Lui and L. H. Chen: Mater. Trans. 47 (2006) 2504

2511.7) H. Masumoto, M. Hinai and S. Sawaya: J. Japan Inst. Metals. 24 (1983)

681688.8) A. Turnbull and E. R. De Los Rios: Fatigue Fract. Eng. Mater. Struct.

18 (1995) 13551366.9) A. H. Cottrell: Philos. Mag. 44 (1953) 829832.10) R. K. Ham and D. Jaffrey: Philos. Mag. 15 (1967) 247256.11) M. C. Chen, L. H. Chen and T. S. Lui: Metall. Mater. Trans. A 27

(1996) 16911694.12) J. M. Reed and M. E. Walter: Mater. Sci. Eng. A 359 (2003) 110.13) X.-M. Cheng and J. G. Morris: Scr. Mater. 43 (2000) 651658.14) S. M. McGuire, M. E. Fine, O. Buck and J. D. Achenbach: J. Mater.

Res. 8 (1993) 22162223.15) J. H. Horning, T. S. Lui and L. H. Chen: Aluminum Trans. 2 (2000)

283302.16) S. M. McGuire, M. E. Fine and J. D. Achenbach: Metall. Trans. A 26

(1995) 11231127.17) D. S. Jiang, T. S. Lui and L. H. Chen: Mater. Trans. 41 (2000) 499506.18) K. T. Huang, T. S. Lui and L. H. Chen: Mater. Trans. 47 (2006) 2405

2412.

20μm

(b)

20μm

(a)

Fig. 9 Optical micrographs of the macroscopic crack propagation for (a) BM, (b) R50. (The arrow indicates the crack propagationdirection.)

50μm

macroscopic crack

Fig. 10 Optical micrograph of the crack propagation path for BM (Thearrow indicates the macroscopic crack propagation direction).

Influence of Pre-Deformation of 5052H112 Alloy on Tensile Properties and Fracture Resistance under Vibration 241

http://dx.doi.org/10.1016/S0921-5093(02)00915-2http://dx.doi.org/10.1016/S0921-5093(99)00653-Xhttp://dx.doi.org/10.1016/S1359-6462(96)00339-9http://dx.doi.org/10.2320/matertrans.46.2268http://dx.doi.org/10.2320/matertrans.46.2268http://dx.doi.org/10.2320/matertrans.47.2504http://dx.doi.org/10.2320/matertrans.47.2504http://dx.doi.org/10.2320/matertrans1960.24.681http://dx.doi.org/10.2320/matertrans1960.24.681http://dx.doi.org/10.1111/j.1460-2695.1995.tb00861.xhttp://dx.doi.org/10.1111/j.1460-2695.1995.tb00861.xhttp://dx.doi.org/10.1080/14786436708227697http://dx.doi.org/10.1007/BF02649826http://dx.doi.org/10.1007/BF02649826http://dx.doi.org/10.1016/S0921-5093(02)00729-3http://dx.doi.org/10.1016/S1359-6462(00)00474-7http://dx.doi.org/10.1557/JMR.1993.2216http://dx.doi.org/10.1557/JMR.1993.2216http://dx.doi.org/10.1007/BF02670609http://dx.doi.org/10.1007/BF02670609http://dx.doi.org/10.2320/matertrans1989.41.499http://dx.doi.org/10.2320/matertrans.47.2405http://dx.doi.org/10.2320/matertrans.47.2405