You are on page 1of 11

Effect of Loading Condition and Stress State on Damage Evolution of Silicon Particles in an Al-Si-Mg-Base Cast Alloy

M.D. DIGHE, A.M. GOKHALE, and M.F. HORSTEMEYER Damage evolution of Si particles in a Sr modified cast A356(T6) Al alloy is quantitatively characterized as a function of strain under tension, compression, and torsion. The fraction of damaged Si particles, their size distributions, and orientation distribution of particle cracks are measured by image analysis and stereological techniques. Silicon particle cracking and debonding are the predominant damage modes. Particle debonding is observed only under externally applied tensile loads, whereas particle cracking is observed under all loading conditions. The relative contributions of Si particle debonding and fracture to the total damage strongly depend on stress state and temperature. For all loading conditions and stress states studied, the average size of damaged Si particles is considerably larger than the bulk average size. The rate of damage accumulation is different for different loading conditions. At a given strain level, Si particle damage is lowest under compression and highest under torsion. The anisotropy of the damage is highly dependent on the deformation path and stress state. Under uniaxial tension, the cracks in the broken Si particles are mostly perpendicular to the loading direction, whereas in the compression test specimens they are parallel to the loading direction. The Si particle cracks in the torsion and notch-tension test specimens do not exhibit preferred orientations. The quantitative microstructural data are used to test damage evolution models.

I. INTRODUCTION

THE Al-Si-Mg-base cast alloys are widely used for automotive and aerospace structural applications. The cast microstructure of these alloys typically contains aluminum rich dendrite cells, silicon particles in the interdendritic regions, microporosity, and numerous minor phases. Mechanical properties of cast Al-Si-Mg-base alloys depend on the amount and size of macrodefects such as internal oxide layers and shrinkage macroporosity,[1,2,3] microporosity,[4,5] dendrite cell size,[6,7] and size and shape of Si particles[8] present in the interdendritic regions. Deformation and fracture of these alloys is usually associated with gradual microstructural damage accumulation.[812] In good quality castings having low porosity and macrodefects, the dominant damage micromechanisms under overload or LCF conditions are fracture[812] and debonding[12] of the Si particles. As the Si particles are clustered in the interdendritic regions, their fracture and debonding facilitates crack initiation in the interdendritic regions. The interlinkage of such microcracks usually causes the final failure (Figure 1). Therefore, to understand and model the deformation and fracture behavior of cast Al-Si-Mg alloys, it is necessary to quantify the evolution of microstructural damage processes such as fracture and debonding of silicon particles as a function of strain under various loading conditions. Such data are useful to

M.D. DIGHE, Advanced Technology Manager, is with the Hi TecMetal Group, Cleveland, OH 44092. A.M. GOKHALE, FASM, Professor, is with the Department of Materials Science and Engineering, Georgia Institute of Technology, Atlanta, GA 30332-0245. M.F. HORSTEMEYER, Manager, is with the Center for Materials and Engineering Science, Sandia National Laboratories, Livermore, CA 94551-0969. Manuscript submitted February 28, 2001.
METALLURGICAL AND MATERIALS TRANSACTIONS A

verify the existing theoretical models for the damage evolution and fracture,[8,10,12,15,16] and they are required for development and validation of internal state variable (ISV)based damage mechanics models for damage evolution and fracture[17,18] of cast Al-alloy components. Earlier studies on the damage evolution of silicon particles in Al-Si-Mg alloys only considered damage evolution under uniaxial tensile stress, and they are mostly of qualitative nature. Few investigations report quantitative data on the evolution of silicon particle damage as a function of tensile strain in Al-Si alloys. Yeh and Liu[8] reported damage evolution in A357 alloy, Gurland and Plateau[9] studied damage evolution as a function of tensile strain in Al-13 pct Si alloy, and Doglione et al.[11] investigated damage evolution of Si particles in A356 alloy. Recently, Caceres and Griffiths[10] studied the cracking of silicon particles in a cast Al-7Si0.4Mg alloy during plastic deformation under uniaxial tension. Effects of temperature[20] and strain rate[21] on the damage of Si particles in A356 alloy have been reported by the authors, but to the best of the authors knowledge, damage evolution related to Si particles in Al-Si-Mg base (or any other Al alloys) has not been studied as a function of compression, torsion, or various triaxial stress states to understand the effect of loading condition and stress state on the damage evolution processes. The objective of this contribution is to report detailed quantitative microstructural analysis of the damage evolution of Si particles in a cast A356 alloy (which is a typical Al-Si-Mg-base alloy) as a function of strain under tension, compression, and torsion. The Si particle damage is also quantified in notch tension test specimens to understand the effect of triaxial stress state on the damage development. It is shown that the damage evolution strongly depends on the loading condition and stress state. The experimental data have been utilized to compute the size dependence of
VOLUME 33A, MARCH 2002555

Fig. 1Fracture caused by interlinkage of microcracks through the aluminum dendrites.

of the dislocation pileup. The model predicts that the effective shear stress required for particle cracking may increase with an increase in the particle size (D) if the length of pileup (L) is much larger than D. Yeh and Liu[8] reported damage evolution of Si particles as a function of strain in a cast A357 alloy under uniaxial tensile stress. They observed that silicon particle fracture initiates at the onset of plastic deformation, and the extent of this damage increases with an increase in plastic strain. Further, the fraction of fractured particles increases more rapidly with plastic strain in the alloys having higher yield stress. At all plastic strain levels, most of the cracks in the particles tend to be perpendicular to the tensile axis. Yeh and Liu developed a model for particle cracking that is based on a dislocation pileup mechanism. Their model gives the following relation between the fraction of cracked particles, f, plastic strain , and the applied tensile stress, . f [ADm2L2M 2 5( 0)]/[5n 1] [3] In this equation, n is the exponent of the assumed power law stress-strain relationship, 0 is the total strain at which microyielding occurs, A is a constant, L is the dislocation pileup length, M is the two-dimensional Schmidt factor, D is the particle size, and m is the average number of pileups in a slip band. If one assumes that 0, and the power law hardening equation given by K n, then Eq. [3] takes the following form. f {[AK 5 Dm2L2M 2]/[5n 1]} (5n1) [4] Equations [3] and [4] predict that at a given value of plastic strain, the fraction of broken particles, f, is directly proportional to the particle size D. Doglione et al.[11] have reported an experimental study on the Si particle damage in A356-T6 alloy. They performed experiments on a chill cast alloy having average Si particle size of 2.6 m (and aspect ratio 1.3) and on a sand cast alloy having average size of 3.4 m (and aspect ratio 1.4). They measured the fraction of broken particles as a function of tensile strain for both the alloys. Interestingly, they did not observe any significant difference in the variation of the fraction of broken Si particles with plastic strain in the sand cast and chill cast alloy specimens, although the average Si particle size in the two groups differed by about 30 pct. Recently, Caceres and Griffiths[10] studied the cracking of silicon particles in a cast Al-7Si-0.4Mg alloy during plastic deformation. Different microstructures were produced by varying the solidification rate and time for solution treatment, in order to study the effect of microstructural variables on damage evolution. It was observed that the eutectic silicon particles progressively crack with applied plastic strain in tension. In coarser structures, the particle cracking occurs very rapidly, while in finer structures, it is more gradual. They used dispersion hardening models to calculate the stresses in the silicon particles. Assuming that the particle fracture follows Weibull statistics, they suggested the following equation for the probability of particle cracking. p 1 exp

the probability of Si particle fracture under tension, compression, and torsion, to test the damage models, and to compute parameters of an ISV-based damage evolution model that incorporates damage nucleation, growth, and coalescence.[17,18] II. BACKGROUND Numerous particle-cracking models have been proposed for damage evolution of particulate phases in metals and alloys.[815] These models mainly differ in the failure criterion, and on this basis, they can be classified into two groups: (1) models that utilize stress or strain based failure criterion[12,13,24,25] and (2) energy based models.[9] Further, some models are based on continuum mechanics,[1,5,9,12,18,19,23] whereas others are based on dislocation theory.[8,13,24] Particle cracking theories[9,26,27] predict that (1) the stress required for particle cracking decreases with an increase in particle size, and therefore, larger particles are expected to crack before the smaller ones; and (2) higher aspect ratio particles crack more easily than equiaxed or spherical particles. All the models pertain to damage evolution under tension. Very few models have been experimentally verified, perhaps due to lack of detailed quantitative microstructural data on the damage evolution. Gurland and Plateau[9] have given the following equation for critical tensile stress to develop a crack in a particle of size D.

[E /(Dq2]1/2

[1]

In Eq. [1], q is the stress concentration factor at the particle, E is the weighted average of the elastic moduli of the particle and matrix, and is the interfacial energy of the crack. Thus, the stress required to crack a particle is inversely proportional to the square root of the particle size, and therefore, according to this model, larger particles crack at lower stresses. Smith and Barnby[22] developed a model for particle cracking on the basis of a dislocation pileup mechanism. They showed that the effective shear stress, , for cracking a particle of size D is given as follows:

[D/(L D)]1/2 [2 ( (L(1 ))]1/2

[2]

V p V0 0

[5]

In Eq. [2], is the Poissons ratio, is the crack surface energy, is the matrix shear modulus, and L is the length
556VOLUME 33A, MARCH 2002

In Eq. [5], p represents the probability of cracking of a particle of volume V (1/6)D3; p is the stress in the
METALLURGICAL AND MATERIALS TRANSACTIONS A

particle; V0 (1/6)D3 0 and m are constants; and D is the size of the silicon particles. These authors concluded that the Weibull statistics could represent the fracture behavior of the silicon particles in uniaxial tension. III. EXPERIMENTAL A. Materials and Heat Treatment The experiments were performed on a commercial A356 alloy of nominal composition Al-7 wt pct Si-0.4 wt pct Mg. Rectangular plate castings of dimensions 203.2 139.7 25.4 mm were made by using iron chills on the top, bottom, and side opposite to the riser end of the plate to simulate a permanent mold casting. A no-bake silica sand was used for the riser and the down sprue. A ceramic filter was placed between the riser and the down sprue. The melt was grain refined, strontium modified, and degassed using a rotary degasser. The pouring temperature was in the range of 950 to 977 K. The castings were cooled for a period of about 16 hours and then removed from the molds. The cast plates were solution treated at 810 K for 16 hours, hot water quenched at 344 K, and then artificially aged at 492 K for 4 hours to achieve a T6 temper. Mechanical test specimens were machined from the low porosity (0.1 pct) regions of the heat-treated cast plates near the chill end. B. Mechanical Tests A series of interrupted tests were carried out at different strain levels under uniaxial tension, compression, and torsion. All test specimens were drawn from similar locations in different plate castings, and their microstructures were statistically similar. Tension, compression, and torsion tests were performed at room temperature and under displacement control at the strain rate of 104. Tensile tests were carried out according to the ASTM E18 standard. The interrupted tensile tests were conducted at different strain levels ranging from 0.3 to 7.5 pct (failure strain). The uniaxial compression tests were performed on specimens of 25.4 mm diameter and 50.8 mm length. The compression test specimens were designed to produce homogenous deformation throughout the specimen by machining grooves into the specimen ends.[28] Lubrication was placed into the grooves to alleviate frictional forces to prevent barreling of the specimen.[29] These quasi-static tests were interrupted at different compression strain levels ranging from 0.1 to 25 pct. Torsional testing involved holding one end of the specimen and rotating or twisting the other end. The Lindholm specimen geometry was used for the torsion test specimens. The specimens were notched to a very thin section in the middle portion to minimize the gradient in the strain from one end to the other. The interrupted torsion tests were performed at torsional strain levels ranging from 0.1 to 45 pct. The notch Bridgman tension test specimens had a notch radius of 2.97 mm and specimen diameter of 9.525 mm. Notch tensile tests were performed at 222, 294, and 394 K, at a strain rate of 0.01 per second. These tests were continued to final fracture of the specimens. C. Quantitative Metallography The specimens were cut in the center along vertical planes containing the applied load direction (torsional axis in case
METALLURGICAL AND MATERIALS TRANSACTIONS A

of torsion test specimens) and were polished using standard metallographic procedures. Details are given elsewhere.[12] The microstructures were observed and quantified in the unetched condition. In the case of compression samples, the region near the center of the specimen was selected for measurements. For the torsion samples, the thinnest region showed the maximum damage. The damage was uniform within this thin section and hence the thin section was selected for the measurements. In the case of notched tensile test specimens, the measurements were performed only in regions within 3 mm of the fracture surface, where the damaged particles were concentrated. To quantify the Si particle damage, contiguous fields of view (at magnification 500 times) were grabbed and stored in the computer memory to create large-area tiled microstructural regions at high resolution.[30,31,32] Particle counting on such a large area high-resolution image eliminates the bias in particle counting due to edge effects for all practical purposes.[30,31,32] Broken and debonded silicon particles were measured interactively over 200 to 300 such contiguous fields. The number density, size distribution, volume fraction, and nearest neighbor distribution of damaged (fractured/debonded) Si particles were measured in each specimen. This was done by classifying the damaged Si particles into two groups, namely, broken (fractured) and debonded particles. Any particle that showed a separation at the interface with the Al rich matrix was classified as a debonded particle. On the other hand, if a particle showed cracking, then it was counted as a broken particle. The geometric attributes of these damaged particles such as the maximum dimension, minimum dimension, shape factor, perimeter, area, and equivalent circle diameter were measured by means of extensive image analysis algorithms. In addition, data were gathered on the orientations and the centroid coordinates of the damaged particles. To quantify the anisotropy of the silicon particle damage with respect to the loading axis, the orientation of the cracks with respect to the loading axis and the orientation of the Si particles with respect to the loading axis were measured. The overall bulk microstructure was also quantified in a similar manner. IV. RESULTS AND DISCUSSION A. Qualitative Microstructural Observations and Results In the present context, fracture and/or debonding of Si particles constitutes damage. Figures 2 through 6 illustrate the damaged Si particles in tension, compression, torsion, and notch-tension specimens. Careful examination of microstructure leads to the following observations. (1) In tensile test specimens, the cracks in the Si particles are mostly perpendicular to the loading direction (Figure 2), whereas in the compression test specimens, the cracks are parallel to the loading direction (Figure 3). On the other hand, Si particle cracks in the torsion test specimens (Figure 4) and notch-tension test specimens (Figure 5) do not exhibit any preferred orientations. Therefore, the anisotropy of cracks and damaged Si particles strongly depends on the loading condition and stress state. (2) For all loading conditions and stress states studied, the average size of damaged Si particles appears to be considerably larger than the bulk average size of the Si
VOLUME 33A, MARCH 2002557

Fig. 2Optical micrograph depicting damaged Si particles in a tensile test specimen at 7.5 pct strain level.

Fig. 5Optical micrograph depicting damaged Si particle in a notch-tension test specimen tested at 394 K.

Fig. 3Optical micrograph depicting the damaged Si particle in a compression test specimen at 22.5 pct strain level.

Fig. 6Debonded Si particles of the tensile test specimen strained to 7.5 pct strain level.

cracking of the silicon particles appears to be a phenomenon that is strongly influenced by its local microstructural environment, which is stochastic in nature. (3) Both fractured (broken/cracked) and debonded Si particles are present in uniaxial tension (Figure 6) and notchtension test specimens, whereas compression and torsion test specimens exhibit only fractured particles. B. Evolution of Silicon Particle Damage with Strain
Fig. 4Optical micrograph depicting damaged Si particles in a torsion test specimen at 5 pct strain level.

particles. However, all large Si particles are not damaged, and some small Si particles are cracked. Small fractured Si particles are mostly found in clusters of many small particles. Also, there are some isolated large particles that have not cracked at all. Therefore, the
558VOLUME 33A, MARCH 2002

Figures 7 through 9 show the number fraction of damaged (broken or debonded) Si particles as a function of strain under uniaxial tension, compression, and torsion. Figure 10 shows a plot of number fraction of damaged Si particles as a function of temperature for notch-tension test specimens. The notch-tension test data have been reported in an earlier contribution.[20] Inspection of these plots leads to the following observations and conclusions. (1) For all loading conditions studied, damaged Si particles
METALLURGICAL AND MATERIALS TRANSACTIONS A

Fig. 7Number fraction of damaged (broken/debonded) Si particles vs global remote tensile strain under uniaxial tension.

Fig. 9Number fraction of damaged Si particles vs strain for torsion test specimens.

Fig. 10Number fraction of debonded and broken Si particles vs temperature for notch-tension test specimens.

Fig. 8Number fraction of damaged Si particles vs strain under uniaxial compression.

are observed at very low strain levels (0.3 pct). It can be concluded that under all loading conditions, the particle damage initiates at the onset of local plastic deformation. (2) For all loading conditions, the number fraction of the damaged Si particles increases monotonically with strain. However, the rate of damage accumulation varies for different loading conditions. (3) At a given strain level, the Si particle damage is lowest under compression and highest under torsion. C. Relative Contributions of Particle Debonding and Cracking Damage Mechanisms All the earlier experimental studies on damage evolution of Si particles in Al-Si-Mg base alloys only considered particle cracking. In the earlier studies, Si particle debonding is either combined with particle cracking or is completely ignored. In the present investigation, both fractured and debonded Si particles are observed in the uniaxial tension test specimens and notch-tension test specimens, whereas only particle cracking is observed under compressive and torsional loading conditions. This implies that a critical
METALLURGICAL AND MATERIALS TRANSACTIONS A

amount of hydrostatic tensile stress is required to initiate Si particle debonding. In the compression and torsion test specimens, the local hydrostatic tensile stresses around Si particles are not sufficiently large to initiate particle debonding. On the other hand, as the Si particle cracking is observed under all the loading conditions, initiation of cracks in Si particles is likely to be governed by principal tensile stress for fracture. This is consistent with the observations that, in compression, the cracks are perpendicular to the local tensile axis (parallel to applied compressive loading axis), and in tension the cracks are perpendicular to the loading axis, indicating mode-I type fracture. On the other hand, in torsion, the principal stress axis rotates as the Si particles and Al rich grains reorient themselves, and therefore, there is no preferred direction for crack orientations. Figure 11 shows relative contributions of debonding and particle fracture to the total Si particle damage as a function of tensile strain. It is interesting to observe that the rate of increase of fraction of debonded Si particles with tensile strain is significantly lower than the rate of increase of fractured Si particles with strain. There is no significant increase in the fraction of debonded particles at high strain levels. At the 7.5 pct strain level (failed tensile test specimen), the ratio of debonded Si particles to fractured Si particles in the bulk microstructure is about 0.3. However, interestingly enough, on the fracture profile of the same specimen, the ratio of the number of debonded Si particles to fractured Si particles was observed to be about two. This may be due to preference of the fracture path for debonded
VOLUME 33A, MARCH 2002559

Fig. 12Relative contributions of debonding and particle fracture to the total Si particle damage as a function of temperature in notch-tension test specimens.

Fig. 11Relative contributions of debonding and particle fracture to the total Si particle damage as a function of uniaxial tensile strain.

particles rather than fractured particles, and/or profuse debonding (rather than cracking) of the particles ahead of the crack tip in the final stages of the fracture process due to the presence of a highly triaxial stress state. Figure 12 shows a plot of the fraction of debonded and broken Si particles as a function of temperature for notch tension test specimens. These data bring out an important effect of state of stress on the Si particle damage. In the uniaxial tension specimens, at any given strain level, the fraction of debonded Si particles is lower than the fraction of broken Si particles (Figure 11). However, in the notch tension specimens, the fraction of debonded Si particles is higher than the fraction of broken Si particles, except at the highest temperature. Because the stress triaxiality is higher for the notch tensile test than for the uniaxial tensile test, debonding would be expected to be higher for the notch tensile test. This verifies that debonding is mainly driven by the hydrostatic tensile stress but the Si particle fracture is driven by the local maximum principal stress. The data in Figure 12 reveal that the fraction of debonded Si particles varies significantly with the temperature, whereas the fraction of fractured Si particles is not as sensitive to the change in the temperature. Fracture of silicon is mainly a function of the maximum principal stress, which changes minimally with temperature. However, debonding is a function of the thermoelastic constitutive response of the silicon particles and the interface between the particles and the matrix, and the thermoplastic response of the matrix. Since the thermoplastic response of the aluminum matrix is highly dependent on temperature, the interface response is also expected to depend on temperature, which gives rise to the temperature dependence of the particle debonding. As the rate of silicon particle debonding increases significantly with decrease in temperature, the contribution of this fracture micromechanism to crack initiation is expected to be higher at lower temperatures. One final note about Figure 12 is that there appears to be a transition temperature at which dominant damage mechanism changes. At cold and ambient temperatures, the fraction of debonded Si particles is larger than the fraction of broken particles. However, at the highest temperature studied, the
560VOLUME 33A, MARCH 2002

damage due to particle fracture is larger than that due to debonding. This transition is related to the relative influence of temperature on the elastic moduli of silicon, which then changes the stress state, for fracture vs the plastic response of the aluminum matrix in conjunction with the changing elastic moduli of the silicon for debonding. Clearly, the temperature influence on the aluminum is greater since we observe a greater difference in the debonding than silicon fracture over the temperature range valuated in this study. The present work leads to the following new observations and conclusions regarding Si particle debonding and cracking mechanisms in the Sr modified A356 (T6) alloy. (1) As Si particle debonding is observed only in tensile test (smooth as well as notch tension) specimens, it may be said that Si particle debonding requires an externally applied tensile stress that gives rise to a critical local hydrostatic tensile stress. On the other hand, Si particle cracking is observed under all loading conditions (tension, compression, and torsion). (2) The extent of particle damage due to debonding depends on the local hydrostatic stress level. A larger triaxial stress state (predominant in notch tensile test specimens) increases the damage due to particle debonding. (3) Under an externally applied tensile load, damage due to particle cracking increases with tensile strain at a significantly higher rate than damage due to particle debonding. (4) The extent of damage due to particle debonding is more sensitive to temperature than particle cracking. D. Size Dependence of Silicon Particle Damage In the present work, the largest dimension of a particle (major axis) is regarded as the measure of Si particle size. In the bulk microstructure of the present specimens, the average Si particle size is 5.7 m. Figure 13 shows the overall Si particle size distribution in the bulk microstructure, whereas Figure 14 shows the size distribution of damaged (broken and/or debonded) Si particles in the tensile test specimen with 7.5 pct strain. Observe that the sizes of the damaged Si particles are significantly larger than the overall bulk average size of 5.7 m. This trend is observed for all loading conditions and stress states studied in this investigation. To quantify the size dependence of probability of particle damage, it is essential to analyze a fraction of particles in the overall population having a given size that are damaged at a given strain. This fraction P is essentially the
METALLURGICAL AND MATERIALS TRANSACTIONS A

Fig. 15A histogram plot of the probability of particle damage P vs Si particle size D for failed tensile test specimen (7.5 pct strain). Fig. 13Overall Si particle size distribution in the bulk microstructure.

Fig. 14Size distribution of damaged Si particles in the tensile test specimen with 7.5 pct strain.

Fig. 16A histogram plot of the probability of particle damage P vs Si particle size D for a compression test specimen (22.5 pct strain).

probability that a particle of given size will be damaged at a given strain level for a given loading condition, and it is formally defined as follows. P [Number of damaged Si particles of given size]/[number of Si particles of same size in overall bulk population]. Figure 15 shows a histogram plot of the fraction P vs Si particle size D in a failed tensile test specimen (7.5 pct strain). Observe that P increases with an increase in the particle size in a nonlinear manner. Figures 16 and 17 show similar plots for compression and torsion test specimens. These data show that a significant fraction of largest particles (28 m size range) is damaged, but only a small fraction of particles in the average size range (4 to 6 m) or below average size are damaged. This clearly shows that larger particles have significantly higher probability of fracture,
METALLURGICAL AND MATERIALS TRANSACTIONS A

[6]

Fig. 17A histogram plot of the probability of particle damage P vs Si particle size D for a torsion test specimen (45 pct strain).
VOLUME 33A, MARCH 2002561

Fig. 18Average size of damaged Si particles vs strain for compression test specimens.

Fig. 20Orientation distribution of cracks in Si particles in a tensile test specimen (7.5 pct strain).

Fig. 19Average size of damaged Si particles vs strain for torsion test specimens.

Fig. 21Orientation distribution of cracks in Si particles in a compression test specimen (22.5 pct strain).

which should be accounted for in the models for damage evolution. It is important to point out that in this analysis, fractured and debonded Si particles are combined together in a single group. The data in Figures 15 through 17 show that for all three loading conditions, at a given strain level, the probability of particle damage increases with the size. In the uniaxial tensile test specimens, the average size of damaged Si particles does not vary significantly with strain. However, in the compression and torsion test specimens, the average size of damaged Si particles decreases with the increase in the strain (Figures 18 and 19). These observations support the hypothesis that, at least for compression and torsion loading conditions, larger particles are the first to fracture, and the extent of damage in the smaller particle size range gradually increases with strain. E. Anisotropy of Si Particle Damage As mentioned earlier, the morphological anisotropy of the damage is highly dependent on the deformation path and
562VOLUME 33A, MARCH 2002

stress triaxiality. Figures 20 through 22 show plots of orientation distribution of cracks in broken Si particles for specimens strained under tension, compression, and torsion. In these data, the crack orientation is specified by the angle between the loading direction (torsion axis in the case of torsion specimens) and the crack. Observe that in the tensile test specimen (Figure 20) most of the cracks have an orientation angle in the range of 80 to 100 deg; i.e., they are approximately perpendicular to the tensile axis. However, in the compression test specimen strained up to 22.5 pct compressive strain (Figure 21), most of the cracks are in the orientation range of 0 and 30 deg (or 150 to 180 deg) with respect to the loading axis, i.e., they are mostly parallel to the loading axis (i.e., direction of compressive load). These cracks are oriented perpendicular to the induced tensile stress component resulting from the Poisson effect. Figure 22 shows the distribution of the orientations of the cracks with respect to torsion axis for a torsion test specimen. These cracks do not show strong preferred orientations. These data
METALLURGICAL AND MATERIALS TRANSACTIONS A

Fig. 23Plots of ln ( f ) vs ln ( ) for damaged (broken or debonded) Si particles and only broken Si particles in uniaxial tension test. Fig. 22Orientation distribution of cracks in Si particles in a torsion test specimen (45 pct strain).

clearly demonstrate that the anisotropy of the damage is highly dependent on the loading condition and stress state.
F. Damage Evolution Models The experimental observations on the size dependence of damage are consistent with the model proposed by Gurland and Plateau.[9] According to their model (Eq. [1]), the stress required to crack a particle is inversely proportional to the square root of the particle size, and therefore, larger particles should crack at lower stresses. Consequently, the model predicts that the probability of particle damage should increase with the particle size, as experimentally observed under all loading conditions in the present work. Yeh and Liu[8] proposed a model for the fracture of silicon particles under uniaxial tension. Their model assumes a strain controlled particle fracture phenomenon, where the fracture occurs due to the pile-up stresses developed due to dislocation movement during the deformation process (Eqs. [3] and [4]). A power law hardening equation given by K n is assumed in Yeh and Lius model, and therefore, one can write
f {[AK 5Dm2L2M 2]/[5n 1]} (5n1) [7] Therefore, if Yeh and Lius model is applicable, a plot of ln ( f ) vs ln ( ) should be linear and the slope of the linear plot should yield the value of the strain hardening coefficient n. Figures 23 shows these plots for the total fraction of damaged Si particles (i.e., fractured and/or broken), as well as for the fraction of broken Si particles. Although both plots appear to be linear, the values of the strain hardening coefficient n calculated from their slopes are negative. Therefore, it can be concluded that Yeh and Lius model is not applicable to the present data. Recently, Horstemeyer and Gokhale[17] have proposed a continuum model that incorporates void nucleation (i.e., particle cracking and debonding) as well as void growth and void coalescence. This model has been implemented into a finite element code.[18] In this model, the total damage evolution is divided into three equations, one each for nucleation, growth, and coalescence. In this model, the Si particle damage by fracture or debonding implies void nucleation. Therefore, data discussed thus far in this article relate to only void nucleation. The void nucleation equation of Horstemeyer
METALLURGICAL AND MATERIALS TRANSACTIONS A

and Gokhale[17] is used to model the results from the cast A356 (T6) aluminum data under compression, tension, and torsion. The integrated form of the void nucleation rate equation is given by

(t) Ccoeff exp

J32 (t)d 1/2 4 KIC f 1/3 27 J23

J3 I1 b 3/2 c J2 J2

[8]

where (t) is the void nucleation density (which is the same as the number density of damaged Si particles), (t) is the strain at time t, and Ccoeff is a material constant. The material parameters a, b, and c relate to the volume fraction of damaged Si particles arising from local microstresses in the material. These constants are determined from experimentally measured data on the Si particle damage evolution under tension, compression, and torsion at different strain levels. The stress state dependence on damage evolution is captured in Eq. [8] by using the stress invariants denoted by I1, J2, and J3, respectively. The term I1 is the first invariant of stress (I1 kk). The term J2 is the second invariant of deviatoric stress (J2 1/2 SijSij), where Sij ij 1/3ijkk . The term J3 is the third invariant of deviatoric stress (J3 SijSjkSiki). The parameter f is the volume fraction of silicon particles, d is the average silicon particle size, and KIC is the bulk fracture toughness of A356 alloy. For the cast A356 aluminum alloy in our study, KIC 17.3 MPam, d 6 m, and f 0.07. The stress state parameters were determined to be a 615.46 GPa, b 58.64 GPa, c 30.011 GPa, and Ccoeff 90. Figure 24 shows the comparison of this damage model with the experimental results for compression, tension, and torsion, respectively. The agreement between the predictions of the model and the experimental data is very much expected, as the same data have been used for calculation of the model parameters. An interesting feature of Horstemeyer and Gokhales model[17] is that once the parameters of the model are known, it is possible to predict the fraction of damaged Si particles for any stress state. Therefore, Horstemeyer and Gokhales model can be tested by using the damage data on notch tension specimens, which were not used to compute the model parameters. At room temperature, the experimentally measured value of the fraction of damaged Si particles is equal to 1.7 pct for the notch tension specimen. For the same
VOLUME 33A, MARCH 2002563

5.

6. 7.

8.
Fig. 24Comparison of damage nucleation data with Horstemeyer and Gokhale model illustrating the differences in Si particle damage evolution under different deformation paths.

loading condition, temperature, and specimen geometry, the fraction of damaged Si particles calculated (Reference 18 provides details of such calculations) from the model is 1.8 pct. Therefore, there is a good agreement between the prediction of the model and the data. However, comparison with more experimental data involving different stress states is clearly necessary to confirm the models predictive capability. V. SUMMARY AND CONCLUSIONS A detailed microstructural analysis has been carried out to quantify damage evolution of Si particles in a cast A356 (T6) alloy, as a function of strain, under tension, compression, and torsion. Analysis of the experimental data leads to the following conclusions. 1. For all loading conditions studied, particle damage initiates at the onset of local plastic deformation. 2. Silicon particle debonding requires an externally applied tensile stress that gives rise to a critical local hydrostatic tensile stress, since debonding is observed only in tensile test specimens (smooth as well as notch tension). A larger triaxial stress state (predominant in notch-tensile test specimens) increases the damage due to particle debonding. On the other hand, Si particle cracking is governed by the local maximum tensile principal stress for fracture, and therefore, it is observed under all loading conditions (tension, compression, and torsion). 3. Under an externally applied tensile load, the damage due to particle cracking increases with strain at a significantly higher rate than the damage due to debonding. On the other hand, the damage due to debonding is significantly more sensitive to temperature than the damage due to particle fracture. 4. For all loading conditions and stress states studied, the average size of damaged Si particles is considerably larger than the bulk average size. However, all large Si particles are not damaged, and some small Si particles are cracked. The silicon particle damage is a phenomenon that is strongly influenced by the local microstructural environment, which is stochastic in nature. The average probability of Si particle damage (calculated from experimental data) strongly depends on particle size: larger
564VOLUME 33A, MARCH 2002

particles have a significantly higher probability of damage. At least for compression and torsion loading conditions, larger particles are the first to fracture, and the extent of damage in the smaller particle size range gradually increases with strain. For all loading conditions, the number fraction (and volume fraction) of the damaged Si particles increases monotonically with strain. The rate of damage accumulation is different for different loading conditions. Further, for a given strain level, the Si particle damage is lowest under compression and highest under torsion. The morphological anisotropy of damage is highly dependent on the deformation path and stress triaxiality. Under uniaxial tension, cracks in the broken Si particles are mostly perpendicular to the loading direction, whereas in the compression test specimens, they are parallel to the loading direction. On the other hand, the Si particle cracks in the torsion test specimens and notchtension test specimens do not exhibit any preferred orientations.

ACKNOWLEDGMENTS The authors thank Richard Osborne (General Motors Corp.) and Srinath Viswanathan (Oak Ridge National Laboratory) for numerous useful discussions, G.A. Schulke (Chrysler Corp.) for casting and heat treatment of A356 alloy plates, and Westmoreland Mechanical Testing and Research Laboratories for conducting the mechanical tests. The research conducted at the Georgia Institute of Technology (AMG and MDD) was supported by research grants from the United States National Science Foundation (Grant No. DMR-9816618) and American Foundry Society. The financial support is gratefully acknowledged. REFERENCES
1. M.J. Couper, A.E. Neeson, and J.R. Griffiths: Fat. Fract. Eng. Mater., 1990, vol. 13 (3), pp. 213-27. 2. M.K. Surappa, E. Blank, and J.C. Jaquet: Scripta Metall., 1986, vol. 20, pp. 1281-86. 3. K. Radhakrishnan and S. Seshan: Trans. Ind. Inst. Met., 1984, vol. 34, pp. 169-71. 4. A.M. Samuel and F.H. Samuel: Metall. Mater. Trans. A, 1995, vol. 26A, pp. 2359-72. 5. E.N. Pan, C.S. Lin, and C.R. Loper: Am. Foundrymen Soc. Trans., 1990, vol. 98, pp. 735-46. 6. R.E. Spear and G.R. Gardener: Am. Foundrymen Soc. Trans., 1963, vol. 71, pp. 209-15. 7. J.F. Major, A. Makinde, P.D. Lee, B. Chamberlain, T. Scappaticci, and D. Richman: Proc. Int. Congr. Expos. on Vehicle Suspension System Advancement, ASAE Publication, Detroit, MI, 1994, pp. 117-28. 8. Jien-Wei Yeh and Wen-Pin Liu: Metall. Mater. Trans. A, 1996, vol. 27A, pp. 3558-68. 9. J. Gurland and J. Plateau: Trans. ASM, 1963, vol. 56, pp. 442-52. 10. C.H. Caceres and J.R. Griffiths: Acta Mater., 1996, vol. 44, pp. 25-33. 11. R. Doglione, J.L. Douziech, C. Berdin, and D. Francois: Mater. Sci. Forum, 1996, pp. 130-39. 12. M.D. Dighe: Masters Thesis, Georgia Institute of Technology, Atlanta, GA, 1999. 13. M.F. Ashby: Phil. Mag., 1966, vol. 14, pp. 1157-78. 14. J.P. Hirth and W.D. Nix: Acta Metall., 1985, vol. 33, pp. 359-68. 15. A.S. Argon, J. Im, and R. Safoglu: Metall. Trans. A, 1975, vol. 6A, pp. 825-37.
METALLURGICAL AND MATERIALS TRANSACTIONS A

16. F.T. Lee, J.F. Major, and F.H. Samuel: Metall Mater. Trans. A, 1995, vol. 26A, pp. 1553-70. 17. M.F. Horstmeyer and A.M. Gokhale: Int. J. Solids Struct., 1999, vol. 36, pp. 5029-55. 18. M.F. Horstemeyer, J. Lathrop, A.M. Gokhale, and M. Dighe: Theor. App. Fract. Mech., 2000, vol. 33, pp. 31-47. 19. C.H. Caceres, C.J. Davidson, and J.R. Griffiths: Mater. Sci. Eng., 1995, vol. A197, pp. 268-78. 20. M.D. Dighe, A.M. Gokhale, and M.F. Horstemeyer: Metall. Mater. Trans. A, 1998, vol. 29A, pp. 905-07. 21. M.D. Dighe, A.M. Gokhale, and M.F. Horstemeyer: Metall. Mater. Trans. A, 2000, vol. 31A, pp. 1725-31. 22. E. Smith and J.J. Barnby: Met. Sci. J., 1967, vol. 1, p. 56. 23. F.A. McClintock: Ductility, ASM, Metals Park, OH, 1968. 24. C. Zener: Fract. Met., ASM, Metals Park, OH, 1948.

25. G.T. Hahn and A.R. Rosenfield: Metall. Trans. A, 1975, vol. 6A, pp. 653-68. 26. T.B. Cox and J.R. Low: Metall. Trans., 1974, vol. 5, pp. 1457-70. 27. T.C. Lindley, G. Oates, and C.E. Richards: Acta Metall., 1970, vol. 18, pp. 1127-36. 28. W.A. Kawahara: Exp. Tech., 1990, MarchApril, pp. 58-60. 29. S.S. Hecker, M.G. Stout, and D.T. Eash: Proc. Workshop on Plasticity of Metals at Finite Strains: Theory, Experiment, and Computation, E.H. Lee and R.L. Mallet, eds., Stanford University, Stanford, CA, 1982, pp. 162-201. 30. A. Tewari, M.D. Dighe, and A.M. Gokhale: Mater. Characterization, 1998, vol. 40, pp. 119-32. 31. P. Louis and A.M. Gokhale: Acta Mater., 1996, vol. 44, pp. 1519-28. 32. P. Louis and A.M. Gokhale: Metall. Mater. Trans. A, 1995, vol. 26A, pp. 1449-56.

METALLURGICAL AND MATERIALS TRANSACTIONS A

VOLUME 33A, MARCH 2002565

You might also like