McBabe IJP 2009
McBabe IJP 2009
McBabe IJP 2009
MST Division, Los Alamos National Laboratory, Los Alamos, NM 87545, USA School of Civil Engineering, University of Sydney, Sydney, NSW 2006, Australia c MPA Division, Los Alamos National Laboratory, Los Alamos, NM 87545, USA
b
a r t i c l e
i n f o
a b s t r a c t
We have used electron backscatter diffraction (EBSD) to quantify the contributions of rst generation and second generation twinning to the total plastic strain of zirconium compressed at 76 K. For compression parallel to a primary c-axis texture, prismatic slip and rst generation f1 1 2 2g compression twinning are the dominant deformation mechanisms with twinning accommodating roughly one third of the plastic strain. Second generation f1 0 1 2g and f1 1 2 1g tensile twins increase with the third power of the rst generation f1 1 2 2g twin fraction. For compression perpendicular to the primary c-axis texture, prismatic slip and rst generation f1 0 1 2g tensile twinning are the dominant deformation mecha nisms with a small contribution from rst generation f1 1 2 1g tensile twinning. Above approximately 17% strain, second generation f1 1 2 2g compression twins begin to make a contribution to the overall strain. These observations are used to explain the measured mechanical responses and texture evolution during deformation of zirconium. 2008 Elsevier Ltd. All rights reserved.
Article history: Received 24 September 2007 Received in nal revised form 1 February 2008 Available online 13 April 2008
1. Introduction Twinning is an important deformation mode in hcp metals such as zirconium (Reed-Hill, 1964; Reed-Hill et al., 1965; Akhtar, 1973; Garde et al., 1973; Song and Gray, 1995; Kaschner and Gray, 2000; Bingert et al., 2002; McCabe et al., 2006) and contributes to its good ductility at temperatures as low as 76 K. In our previous work, we qualitatively described the deformation mechanisms observed in zirconium as a function of temperature and initial texture (McCabe et al., 2006). The key deformation modes observed at 76 K are shown schematically in Fig. 1, and include prismatic slip,
* Corresponding author. Tel.: +1 505 606 1649; fax: +1 505 667 8021. E-mail address: rmccabe@lanl.gov (R.J. McCabe). 0749-6419/$ - see front matter 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.ijplas.2008.03.010
455
1 3
two tensile twins, f1 0 1 2g and f1 1 2 1g, and one compressive twin, f1 1 2 2g, where compressive and tensile twinning are named for their ability to accommodate strain during c-axis deformation. For compression parallel to a primary c-axis ber texture (through thickness, TT, direction), we observed that the dominant deformation mechanisms were rst generation compressive f1 1 2 2g twinning and prismatic slip. Second generation tensile twinning, twins within the rst generation twins, was observed and proposed to have a signicant effect on the evolving microstructure and stress strain response. For compression perpendicular to a primary c-axis ber texture (in-plane, IP, direction), we observed that the dominant deformation mechanisms were rst generation f1 0 1 2g tensile twinning and prismatic slip, with a small number of rst generationf1 1 2 1g tensile twins and second genera tion compressive f1 1 2 2g twins at high strains. The volume fraction of rst generation and second generation twinning, and hence, the overall contribution of twinning to deformation at a given applied strain was not determined quantitatively. Polycrystal plasticity models are routinely applied to describe constitutive plastic response and texture evolution. Historically, the active deformation modes and single crystal hardening parameters for zirconium have been inferred by tting to available experimental information such as stressstrain curves and deformation textures (Tome et al., 2001; Kaschner et al., 2006). The Visco-Plastic Self-Consistent (VPSC) model has been successful in capturing the texture and hardening evolution for deformation along a single strain path (Lebensohn and Tome, 1993; Rangaswamy et al., 2002). However, the VPSC model does not accurately predict the active deformation modes, and thus has been less successful in capturing the response for deformation conditions that include a change in strain path (e.g. change in strain direction or temperature). A recent improvement to the VPSC model known as the Composite Grain Model (CGM) (Proust et al., 2007) accounts for deformation within rst generation deformation twins and qualitatively agrees with the observed mechanisms (McCabe et al., 2006). However, there is a need to experimentally determine the quantitative contribution of different deformation modes to the total plastic strain in order to properly incorporate the correct physical mechanisms into the polycrystal plasticity models. Quantitative twin fraction measurements have been made previously in hcp metals using optical techniques (Reed-Hill et al., 1965; Zhang et al., 1998; Nemat-Nasser et al., 1999; Jiang et al., 2007; Lou et al., 2007), orientation distribution (diffraction/texture) techniques (Rangaswamy et al., 2002; Backx et al., 2007; Jain and Agnew, 2007), and EBSD techniques (Glavicic et al., 2004; Battaini et al., 2007). For our application, we strongly believe that data obtained by EBSD is more reliable than with the two other techniques. Reed-Hill et al. (1965) used optical microscopy to quantitatively measure twin fractions in zirconium to strains of about 9%. The material and straining conditions of that study were such that twinning deformation was due primarily to only two twinning modes, f1 1 2 1g and f1 0 1 2g. The two twinning modes were distinguished from one another optically based on differences in morphological features, a differentiation technique that can be ambiguous. The data collected during an EBSD scan allows for verication of the crystallographic relation between two adjacent grain fragments, and one can directly determine if, and which kind of twinning relation exists between two fragments. In our study, the material and deformation conditions result in up to three coexisting twinning modes occurring in multiple twinning generations. The EBSD orientation data can be used to unambiguously characterize twin boundaries. In addition, the EBSD orientation data lends itself to
456
semi-automatic twin identication and analysis (Mason et al., 2002; Henrie and Mason, 2004; Proust et al., in press), largely removing human subjectivity from the process of twin measurements. Battaini and his coworkers (Battaini et al., 2007) contend that neutron diffraction, although not easily available, would be the best technique for analyzing twinned microstructures. We agree that neutron diffraction is an excellent technique for quantifying twin contributions if the operative twin modes are known a priori and the operative twin modes result in a distinct change in texture. Mag nesium alloys may be optimal materials for diffraction techniques where f1 0 1 2g twinning is the dominant twinning mode and it results in a nearly 90 change in crystal orientation and can cause a signicant change in texture (Backx et al., 2007; Jain and Agnew, 2007). However, when different twinning modes are active simultaneously it becomes more difcult to determine the contribution of each twinning mode to the bulk texture. It will be shown in this study that diffraction techniques by themselves are not well suited for determining all twin contributions for some deformation condi tions in zirconium. Rangaswamy et al. (2002) previously used neutron diffraction to measure f1 1 2 2g twinning fractions in zirconium of similar pedigree to the present study. We will show here that it is likely that the previous neutron diffraction study underestimates the contribution of f1 1 2 2g twinning at higher strains, which resulted in the prediction of twin saturation, and totally neglected the contributions of second generation f1 0 1 2g and f1 1 2 1g twins. The purpose of the present work is to quantify the contributions the various twinning modes make to the deformation of zirconium using EBSD. The results will allow for further renement of polycrystal models by providing quantitative experimental measurements of the activity of the various deformation modes in zirconium. 2. Experimental We have used clock-rolled and annealed, high-purity Zr with a strong c-axis ber texture and an average grain size of 1520 lm. Clock-rolling is a multiple pass rolling sequence where the plate is rotated to the various multiples of 45 between passes. Due to this processing path, the plate is expected to be relatively isotropic in the in-plane directions. The plate is of similar pedigree to the plate used by Kaschner and Gray (2000) and the impurity chemistry is expected to be similar including an oxygen content of less than 50 ppm. Compression cubes with approximate dimensions of 6.8 mm in the plate through thickness (TT) direction and dimensions of 7.8 8.8 mm in the in-plane (IP) directions were cut from the clock-rolled plate and quasi-statically deformed to various strains at 76 K with the compression direction corresponding to either the TT or IP direction. Samples were deformed to a range of nal strains in uniaxial compression at a strain rate of 0.001/s in an equilibrium liquid nitrogen bath. In order to minimize frictional effects, samples were compressed between tungsten carbide platens coated with MoSi2 lubricant. Further details of the mechanical characterization are given elsewhere (Kaschner et al., 2006). Following deformation, the cubes were cut in half with the section plane containing the plate TT direction and an IP direction, the IP direction corresponding to the compression direction for IP compressed samples. The specimens were metallographically mounted in conductive epoxy and the surface was prepared by grinding to an 800 grit SiC nish followed by chemically polishing with a solution of the ratio 45H2O:45HNO3:10HF. Automated EBSD scans were performed at 25 kV in an FEI XL30 SEM equipped with TSL/EDAX data acquisition software. The orientation data was analyzed using TSL/EDAX OIMTM Analysis software. EBSD textures were obtained using a 10 lm step size in a hexagonal grid over half of the cube cross-section from one compression face to the center (hereafter referred to as large scale scans) The area scanned during the large scale scans was around 25 mm2 . A 0.2 lm step size scan of a 120 240 lm area was performed at the center of each sample and at other positions to examine the microstructure (hereafter referred to as small scale scans). Twin statistics for the small scale scans were analyzed using the OIMTM Analysis software and a twin program written at Los Alamos National Laboratory (Henrie and Mason, 2004; Proust et al., in press). The twin program is used to identify the parent orientation inside a grain and to assign twin modes and generations to the remaining grain fragments. The parent orientation is identied by systematically analyzing twin variants and/or comparing Schmid factors, relative sizes, and aspect ratios between grain fragments sharing twin relationships. Further details are given in Henrie and Mason (2004); Proust et al. (in press).
457
As with optical microscopy and X-ray diffraction, EBSD data is obtained from a two-dimensional slice of a three-dimensional specimen. Our twin fraction measurements based on the small and large scale scans are area fraction measurements. However, if the twins are small compared to the sampling area and randomly distributed within the samples, the area fraction is expected to be equivalent to the volume fraction. We are also reassured by the fact that our EBSD based pole gures compare very well with neutron diffraction derived pole gures taken from samples from the same plate having undergone similar levels of strain (Kaschner et al., 2006).
3. Results Representative stressstrain and hardening curves for 76 K compression in the TT and IP directions are shown in Fig. 2. Evident from the curves is the signicant anisotropy characteristic of textured polycrystalline zirconium. The mechanical behavior of the 76 K, TT sample is characterized by a sharp yield stress with, thereafter, a slowly increasing work hardening rate. Both the sharp yield and increasing work hardening rate are often characteristics of deformation twinning. IP compression exhibits a much lower yield than the TT case. The IP work hardening rate shortly after yield is higher than in the TT case, but eventually reaches a maximum and begins to decrease at a strain of about 17% and a stress above 500 MPa. An EBSD orientation map for the undeformed microstructure is shown in Fig. 3. The undeformed microstructure consists of twin free grains with an average grain size of 1520 lm, and a majority of grains have a c-axis orientation within 30 of the TT direction. The grain size across the thickness of the sample was found to be very uniform. EBSD orientation maps for samples compressed 9.3% TT and 14.3% IP are shown in Fig. 4a and b, respectively. Both deformation microstructures consist of heavily twinned grains. The regions of the EBSD scans corresponding to different zirconium twinning modes and twinning generations were determined using the OIMTM Analysis software and a twin program (Henrie and Mason, 2004; Proust et al., in press). The results are shown in twin maps in Fig. 5. f1 1 2 2g compressive twinning during TT compression ips the c-axis by 64.3, and tends to align it almost perpendicular to the TT direction. This is evident in Fig. 4a where the material that has been reoriented by compressive twinning is oriented away from (0 0 0 1) pole (red) and toward thef1 0 1 0g and f2 1 1 0g (blue to green) poles. Most of the tensile twins observed during TT compression are second generation tensile twins and are completely within rst generation compressive twins. First gen eration f1 0 1 2g tensile twinning during 76 K IP compression ips the c-axis by 85.2, and tends to align it with the compression axis near the equator. This is evident in the IPF map where the material that has been reoriented by tensile twinning is oriented near the f1 0 1 0gand f2 1 1 0g (blue to green) poles. A small number of compressive twins are seen in Fig. 4b, but signicant fractions are not observed until after a strain of approximately 17%. EBSD derived pole gures (PFs) showing the evolution of texture with strain are presented in Fig. 6. These pole gures compare very well with neutron diffraction derived pole gures taken from samples from the same plate having undergone similar levels of strain (Kaschner et al., 2006). The undeformed (initial) zirconium has a strong c-axis ber texture about the TT direction with the in-plane (IP) texture of the plate being relatively homogeneous. The texture across the thickness of the sample was measured and found to be very uniform. The reorientation due to compressive twinning during TT compression is evident in the (0 0 0 1) pole gures as a reduction in intensity near the pole (TT direction) and increase in intensity near the equator (Tome et al., 2001; Rangaswamy et al., 2002). The reorientation due to tensile twinning during IP compression is evident in the (0 0 0 1) pole gures as a reduction in intensity near the pole (TT direction) and an increase in intensity near the equator along the compression direction, labeled IP1 in the pole gures (Kaschner et al., 2001; Tome et al., 2001). Fig. 7a represents the distribution of c-axes orientations with respect to the original TT direction as a function of strain; the data was obtained from the large scale scans. The reorientation due to twinning even at low strains is evident from this gure in the form of c-axis orientations near 90. Fig. 8 was derived directly from the small scale scan data from the sample compressed to 14.3% in the TT direction. The raggedness of Fig. 8 compared to Fig. 7a is due to the much smaller sampling area used to construct Fig. 8. The distribution of c-axis orientations are broken up into contributions from the
458
Fig. 2. Representative (a) stressstrain and (b) hardening curves for 76 K compression in the through thickness (TT) and inplane (IP) orientations.
various twinning elements (un-twinned material, rst generation compression twins, etc.. . .) similar to as represented on Fig. 5a. First generation f1 1 2 2g compressive twins (C1-1) can be differentiated from the rest of the material based on c-axis orientations. Therefore, one can determine a cutoff angle in the c-axis distribution to measure the fraction of material that has been reoriented by rst generation compressive twins (C1-1). If we assume the cutoff angle is 60 for the sample deformed 9.3% by TT compression, we obtain the same volume fraction as measured using the twin recognition code on the small scale scan of the same sample. A second observation from Fig. 8 is the material that has second generation tensile twinned (T1-2 and T2-2) cannot be differentiated from the non-twinned material based on c-axis orientations. The peak for T1-2 lies directly below the peak for non-twinned material. Therefore, the material that has re-twinned, and should be counted both towards the rst
459
Fig. 3. EBSD orientation map showing the undeformed microstructure. Colors in all orientation maps designate crystal orientations parallel to the TT orientation.
generation compressive twin fraction and the second generation tensile twin fraction, cannot be measured base on the distribution of c-axes. One of the main objectives of this work is to quantitatively measure the contributions for all twin modes during compression of zirconium samples at 76 K. It is desirable to study zirconium in compression because of the anisotropic nature of its response; it behaves differently in tension and compression. However, there are compromises incurred by using compression testing making it more challenging to accurately quantify twinning contributions. A primary challenge is due to frictional effects at the compression platens resulting in stress and strain states that are not totally uniform across the height of the specimen. This results in non-uniform twinning across the height of the sample, the magnitude of which will become evident in the next section. It is also a challenge to accurately account for second generation twinning. Much of the next section addresses these challenges in order to obtain high-quality quantitative data. 4. Twin fraction determination The analysis of twinning fractions during 76 K compression is complicated due to second generation twinning and the non-uniform distribution of twinning within the specimens. However, it is possible to determine the contributions of all twinning modes during deformation by incorporating data from both the large scale and small scale scans. EBSD measurements are generally statistically limited by the step size and/or scan area. A compromise is usually made between step size and scan area to obtain data of interest in a timely manner. For the large scale scans the data of interest are orientation distributions such as those shown in texture plots (Fig. 6) or in c-axis distribution plots (Fig. 7). The primary benet of large scale scans is a more representative sampling. Textures based on our large scale scans are very similar to pole gures obtained using X-ray diffraction (XRD) and neutron diffraction (Kaschner et al., 2006). As discussed before, applying only a texture based method would result in a considerable underestimation of rst generation twinning and total neglect of second generation twinning due to the overlap in the texture peaks representing non-twinned material and second generation twins. With texture based methods, it is only possible to estimate the fraction of material that has rst generation twinned and not re-twinned. This is essentially the red areas of Fig. 5a. The fraction of re-twinned material, the material contributing to both the rst generation compressive twinning and the second generation tensile twinning, must be measured by alternate means and will be discussed later.
460
Fig. 4. EBSD orientation maps showing microstructures for samples deformed (a) 9.3% TT and (b) 14.3% IP in compression. The compression direction is the (a) TT and (b) IP1 direction. Colors designate crystal orientations parallel to the TT direction.
The uncorrected rst generation twin fractions (twinned but not re-twinned material) were measured by summing the volume fraction contributions above an appropriate cutoff angle within the caxis orientation distribution plots (Fig. 7) of the large scale scans. The appropriate cutoff angle was determined using the small scale scan data, and was found to vary with strain. The c-axis distribution was analyzed for each small scale scan similar to what was done for Fig. 8 to determine the appropriate cutoff angle for a given strain condition. In each case, a cutoff angle was determined that gave the same volume fraction of rst generation twins as was measured directly from the small scale scan data similar to Fig. 5a. Fig. 9a and b shows the areas on the small scale scan with angles greater than the cutoff angle for TT compression to 9.3% and IP compression to 14.3%, respectively. In Fig. 9a, the red regions are those where the angle between the c-axis and TT direction is greater than a cutoff of 60, and in Fig. 9b, the blue regions are those where the angle between the c-axis and TT direction is greater than a cutoff of 70. The red and blue regions in Fig. 9a and b, respectively, have the same volume fraction as the red and blue regions in Fig. 5a and b, respectively, and the colored regions correspond very well between the two images. Twinning is not uniform across the sample due to the non-uniform stress and strain state resulting from frictional effects at the compression platens. The question becomes, what is the appropriate vol-
461
Fig. 5. Twin maps showing the observed twin structure for (a) 9.3% TT and (b) 14.3% IP compression determined using the twin program.
ume to analyze in order to quantitatively determine the twinning fraction? Since the compression test measures the average axial stress and strain response of the entire compression sample including the region near the platens, one can measure the average twin fraction over the entire sample. By making the assumptions that the sample is axi-symmetric about the center of the compression axis and symmetric about the center plane perpendicular to the compression direction, we can use the orientation distribution of the entire large scale scan area to measure the average uncorrected rst generation twin fraction for the entire sample. Another measure of the microstructure is the twin fraction in the region of the specimen where frictional effects are minimal and the stress state is most nearly truly uniaxial, that is the center of the sample. We can use the orientation distribution for a portion of the large scale scan near the center of the sample to measure the twin fraction. In Fig. 10a, there are two curves representing the uncorrected rst generation f1 1 2 2g compression twin fraction (C1-1). The lower curve represents the average over the entire specimen and the upper curve represents the average over the center half of the specimen. The maximum difference between the two curves is around 10% of the measured volume fraction. For the IP compression samples, there is no systematic variation between the two curves and only the whole sample average uncorrected curve (T1-1) is represented in Fig. 10b.
462
Fig. 6. (0 0 0 1) and 1 0 1 0 pole gures showing the evolution of texture with strain. (a) Undeformed zirconium plate texture. Specimens compressed parallel to the plate through thickness (TT) direction to strains of (b) 6.2%, (c) 9.3%, (d) 14.3%, and (e) 18.2%. Specimens compressed parallel to the plate in-plane direction (IP) to (f) 4.2%, (g) 8.9%, (h) 14.3%, and (i) 18.7%. The loading direction is parallel to the ND for the TT experiments and parallel to the IP1 for IP experiments. Maximum intensities are marked on each PF.
The fraction of re-twinned material, the material contributing to both the rst generation compressive twinning and the second generation tensile twinning, is measured using the small scale scan data.
463
a
Number Fraction (bin size 2)
0.06 0.05 0.04 0.03 0.02 0.01 0 0 10 20 30 40 50 60 70 80 90 Angle Between c-axis and TT direction Strain 0.00% 2.22% 6.27% 9.25% 14.34% 18.25%
b
Number Fraction (bin size 2)
0.06 0.05 0.04 0.03 0.02 0.01 0 0 10 20 30 40 50 60 70 80 90 Angle Between c-axis and TT direction Strain 0.00% 4.28% 8.94% 14.27% 16.97% 18.77%
Fig. 7. Distribution of c-axis orientations with respect to the TT direction as a function of strain for (a) TT and (b) IP compression. The peaks observed near 90 are due to compressive twins in (a) and tensile twins in (b).
The primary benet of the small scale EBSD scans is they clearly show the active twinning modes including second generation twinning and beyond. In addition, twin statistics obtained from small scale scans can be very accurate for the small area measured. While these statistics may not be an accurate representation of twinning fractions for the entire sample because of twinning non-uniformities, they can be used to examine trends in the dependence of second generation twinning on the fraction of rst generation twinning. Figs. 11 and 12 show the dependence of the fraction of second generation f1 0 1 2g and f1 1 2 1g tensile twins, respectively, as a function of the fraction of rst generation compressive twins using small scale scans on TT compressed samples. There are two curves for each plot, one with the x-axis representing the non-re-twinned (uncorrected) fraction of rst generation compressive twins. The x-axis for the other curve in Fig. 11 represents the corrected (twinned and re-f1 0 1 2g-twinned) fraction of rst generation compressive twins. Likewise, the x-axis for the second curve in Fig. 12 represents the corrected (twinned and re-f1 1 2 1g-twinned) fraction of rst generation compressive twins. Interestingly, both corrected curves follow a y = Ax3dependence. These curves are based on small scale scans from the centers of TT samples compressed to 6.3%, 9.3%, 14.3% and from a small scale scan taken approximately midway between the center and compression face of the 6.3% sample along the midline. Using these curves, the fraction of second generation twinning and, thus, the corrected fraction of rst generation compressive twinning can be estimated.
464
0.06 0.05 Volume Fraction 0.04 0.03 0.02 0.01 0 0 10 20 30 40 50 60 70 80 90 Angle Between c-axis and TT direction
Fig. 8. Distribution of c-axis orientations with respect to the TT direction for parent and twinning components observed in the microstructure for TT compression to 14.3% strain. The plot is based on the 0.2 lm step size, 120 240 lm scan from the center of the specimen. In the legend C1, T1, and T2 refer to the twinning mode as per Fig. 1 and the number following refers to the twinning generation.
Fig. 10 shows the evolution of twin volume fraction during deformation at 76 K in the TT and IP directions. Measurements of the uncorrected rst generation twin fractions (C1-1 for TT and T1-1 for IP) were made using the large scale scan, c-axis distribution data shown in Fig. 7. The corrected curves in Fig. 10 add the fraction of second generation twins to the uncorrected rst generation twin fractions. For the TT case, the curves in Figs. 11 and 12 were used to estimate the contribution of second generation tensile twin fractions (T1-2 and T2-2) based on the measured uncorrected (C1-1) values. There are two corrected C1-1 curves corresponding to the two uncorrected C1-1 curves discussed above. Fractions of T2-1 and C1-2 for IP compression are from direct measurements from small scale scans at the centers of specimens. Fig. 13 shows how the active deformation mechanisms contribute to the compressive strain. The twinning contributions were estimated using the expression e V f Sm, where Vf is the twin fraction, S is the characteristic twinning shear and m is the volume average twin Schmid factor (Reed-Hill et al., 1965). Volume average Schmid factors were determined from the small scale EBSD data as a function of twin fraction. The m values did not vary signicantly with twin fraction and/or total strain. Table 1 shows the twinning shear values and volume average Schmid factors used in creating Fig. 13. In each case, prismatic slip was assumed to make up the difference in strain and was determined by subtracting the twinning contributions from the total compressive strain. Again, in the TT case there are curves based on the average twin fraction for the entire sample and curves based on the average twin fraction measured in the center half of the specimen. 5. Discussion 5.1. In-plane compression The prominent feature of the behavior of zirconium during IP compression at 76 K is the transition that occurs at a strain of around 17%. This transition is most evident in the hardening versus strain plot in Fig. 2b where the work hardening dramatically changes from an increasing rate to a decreasing rate. The behavior during IP compression will rst be discussed in terms of the active deformation mechanisms and microstructural changes occurring up to the transition, and then during and beyond the transition.
465
Fig. 9. (a) For 9.3% TT compression, all grain fragments in which the angle between the c-axis and original TT direction is greater than a cutoff angle of 60 are shown in red. The area fraction of dark red grains is the same as in Fig. 5a, and the red areas correspond well between the two gures. (b) For 14.3% IP compression, all grain fragments in which the angle between the caxis and original TT direction is greater than a cutoff angle of 70 are shown in blue. The area fraction of dark blue grains is the same as in Fig. 5b, and the blue areas correspond well between the two gures.
The observed mechanical properties and microstructural changes up to the transition are largely due to rst generation tensile twinning and prismatic slip. From Fig. 10b we see that the rate (slope of the volume fraction versus strain curve) of rst generation f1 0 1 2g tensile twinning increases with strain until the transition at 17% strain. Approximately 45% of the microstructure is reoriented by rst generation f1 0 1 2g tensile twinning by 17% strain. The fraction of rst generation f1 1 2 1g tensile twins initially increases with strain and then reaches a plateau at a volume fraction just under 5%. Compressive twinning does not contribute signicantly up to the transition. Fig. 13b shows that prismatic slip accommodates 8090% of the compressive strain. While rst generation f1 0 1 2g tensile twinning is relatively inefcient at accommodating strain, it has a major effect on the hardening behavior. The primary hardening effect is a reorientation effect; material is reoriented from a soft orientation to a hard orientation. Most of the grains in the unde formed material are oriented favorably for both f1 0 1 2g tensile twinning and prismatic slip, and both mechanisms are often active in the same grains (McCabe et al., 2006). The crystal orientation of the
466
a 0.7
0.6
Twin Volume Fraction
C1-1 Corrected
0.15
0.2
0.5
0.4
T1-1 Corrected
0.3
0.2
0.1
Fig. 10. Evolution of twinning with strain for (a) TT and (b) IP compression. C1-1 in (a) is the measure of the rst generation compressive twins that have not re-twinned, and C1-1 Corrected adds the contributions of the second generation tensile twins. The lower curve in each case is the average twin fraction for the entire sample and the upper curve is the fraction measured from the middle half of the sample. At 18% strain both curves halve the same value and the uncertainty is indicated by an error bar. T1-1 in (b) is the measure of the rst generation tensile twins that have not re-twinned. T1-1 Corrected adds the contributions of the second generation compressive twins.
rst generation f1 0 1 2g tensile twins, with the c-axis being nearly parallel to the compression direction (cf. Fig. 6), is favorable for neither prismatic slip nor tensile twinning. These orientations generally have near zero Schmid factors for prismatic slip and negative Schmid factors for both tensile twinning
467
0.16 2nd Gen Tensile Twin Fraction 0.14 0.12 0.1 0.08 0.06 0.04 0.02 0 0 0.2 0.4 Compressive Twin Fraction 0.6 Corrected
y = Ax3
Fig. 11. Dependence of 2nd generation f1 0 1 2g tensile twin fraction on rst generation compressive twin fraction. The open diamonds represent the uncorrected fraction of rst generation compressive twins. The diamonds on squares are corrected for the 2nd generation f1 0 1 2g tensile twins.
0.016 2nd Gen Tensile Twin Fraction 0.014 0.012 0.01 0.008 0.006 0.004 0.002 0 0 0.2 0.4 Compressive Twin Fraction 0.6
y = Bx3
Corrected
Fig. 12. Dependence of 2nd generation f1 1 2 1g tensile twin fraction on rst generation compressive twin fraction. The open triangles represent the uncorrected fraction of rst generation compressive twins. The triangles on squares are corrected for the 2nd generation f1 1 2 1g tensile twins.
modes. As the volume fraction of rst generation f1 0 1 2g tensile twins increases, there is less material available for easy prismatic slip or rst generation tensile twinning. The result is a decrease in the contributions to compressive strain of prismatic slip, a plateau in the volume fraction of rst genera tion f1 1 2 1g tensile twins, and an increase in the strain hardening rate. In addition to hardening associated with reorientation, it is also possible that rst generation f1 0 1 2g tensile twinning results in hardening due to HallPetch effects (Meyers et al., 2001). As rst
468
a
Fraction of Compressive Strain
Prismatic C1-1 T1 -2 T2 -2
0.05
0.1
TT-Compressive Strain
0.15
0.2
b
Fraction of Compressive Strain
1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.05 0.1 0.15 0.2
Prismatic T1 -1 T2 -1 C1-2
IP Compressive Strain
Fig. 13. Contribution of deformation mechanisms to (a) TT and (b) IP compressive strain. C1-1 and T1-1 represent corrected values. Prismatic slip was estimated as the balance of the total imposed strain not accommodated by twinning. For the TT case, there are two curves for each mechanism. The curve with the larger symbols represents the average for the entire compression sample and the curve with the smaller sample is the average for the center half of the sample.
generation f1 0 1 2g tensile twinning proceeds, twins divide grains and may act as barriers to dislocation glide (Yablinsky et al., 2006; Jiang et al., 2007). Meyers et al. have also shown that the effect of grain size on twinning is even more pronounced (Meyers et al., 2001). In the current study we found that, at low strains, rst generation f1 1 2 1g twins are generally observed to be long and narrow, usually spanning entire grains from boundary to boundary. They also seem to be more prevalent in larger
469
Table 1 Values of twinning shears and volume average Schmid factors used for calculating the contributions to compressive strain of the active twin components Twin mode C1-1 (TT) T1-2 (TT) T2-2 (TT) T1-1 (IP) T2-1 (IP) C1-2 (IP) S .23 .17 .63 .17 .63 .23 m .41 .41 .42 .38 .37 .45
grains, and they frequently occur in grains that also exhibit rst generation f1 0 1 2g tensile twinning. As the fraction of rst generation f1 0 1 2g tensile twins increases and the sizes of un-twinned volumes decrease, the volumes available for f1 1 2 1g twinning decrease. As it becomes more difcult for prismatic slip and f1 1 2 1g tensile twinning to accommodate strain, the rate of rst generation f1 0 1 2g tensile twinning increases. f1 0 1 2g tensile twin boundaries act as barriers to f1 1 2 1g twins and slip systems within the parent grains because thef1 0 1 2g twins are at angles to the f1 1 2 1g twin planes and prismatic slip planes. When considering only single f1 0 1 2g twin variants, there should be no self HallPetch effect because the twin boundaries are parallel to their own twinning shear direction. HallPetch effects on f1 0 1 2g twinning would only occur in grains with more than one f1 0 1 2g twin variant. At around 17% deformation, an obvious transition occurs in the behavior of the IP compressed zirconium. As seen in Fig. 10b, this transition is accompanied by the appearance of second generation f1 1 2 2g compressive twins. At the transition, the ow stress has become high enough to activate f1 1 2 2g compressive twinning in favorably oriented volumes. As seen in Fig. 2a, the ow stress at the transition is roughly equivalent to the yield stress for TT compression, that is the stress that is nec essary to activate rst generation f1 1 2 2g compressive twins during TT compression as will be discussed in the next section. Although only a small fraction of original grains were oriented well for compressive twinning during IP compression, nearly all rst generation f1 0 1 2g twins are well ori ented for compressive twinning. Moreover, the second generation f1 1 2 2g compressive twinned volumes are once again oriented favorably for prismatic slip and tensile twinning. The additional active deformation mechanisms result in a decrease in the strain hardening. 5.2. Through thickness compression For 76 K TT compression, the initially observed mechanical properties and microstructural changes are, again, largely due to the interplay of two mechanisms; in this case rst generation f1 1 2 2g com pressive twinning and prismatic slip. Fig. 10a shows the important role of rst generation f1 1 2 2g compressive twinning having reoriented around 60% of the microstructure at 18% strain. Fig. 13b shows that prismatic slip accommodates 6070% of the compressive strain. While the resolved shear stress necessary to activate prismatic slip is considerably less than that for f1 1 2 2g compressive twinning, deformation in many grains cannot be accommodated by prismatic slip. A large fraction of grains in the initial microstructure are oriented with their c-axis close to the compression direction resulting in nearly 50% of them having a Schmid factor of less than 0.1 for prismatic slip. Bulk plastic strain cannot occur until the stress is high enough to activate compressive twinning in the hardest grains. Once the stress is high enough to activate compressive twinning in the harder grains, prismatic slip is able to accommodate strain in some of the softer neighboring grains. Nearly all the grains are oriented for f1 1 2 2g compressive twinning with more than 28% of all ini tial grains having a Schmid factor between 0.48 and 0.5 for at least one of the f1 1 2 2g compressive twin variants. Moreover, we have observed that most twinned grains exhibit more than one f1 1 2 2g variant. For example at 6% strain, 80% of the grains at the center of the deformed specimen that have twinned present two or more compressive twin variants. The stress required for f1 1 2 2g compressive twinning is likely reached nearly simultaneously in a signicant fraction of grains. This
470
is the likely reason for the initial rapid rise in f1 1 2 2g compressive twinning fraction and the small work hardening rate immediately following yield. Another important difference with the IP case is that f1 1 2 2g compressive twinning during TT compression ips material from a hard orientation to a soft orientation. Compressive twinned material is ideally oriented for prismatic slip and/or second generation tensile twinning. This may partially explain the lower work hardening rate following yield than in the IP case. The prismatic activity shown in Fig. 13a for TT compression occurs both in parent grains and within rst generation f1 1 2 2g compressive twins. One nal, major difference between IP and TT compression is in the manifestation of second gen eration twinning. During IP compression, rst generation f1 0 1 2g tensile twinning ips material to orientations ideal for second generation f1 1 2 2g compressive twinning, but the stress is not high enough until approximately 17% strain to activate compressive twinning. At 17% strain, rst generation f1 0 1 2g tensile twins are generally relatively large and there is only a minimal grain (rst generation twin) size effect on second generation twinning. During TT compression, rst generation f1 1 2 2g com pressive twinning ips material to orientations ideal for second generation f1 0 1 2g tensile twinning. At yield, where the initial rise in f1 1 2 2g compressive twinning fraction is observed, the stress is al ready much larger than the stress necessary to initiate rst generation f1 0 1 2g tensile twinning in the 2g compressive twins are the limiting IP case. The volume fraction and size of the rst generation f1 1 2 factors to second generation tensile twinning. As the fraction and size of rst generation f1 1 2 2g compressive twins increase, second generation tensile twinning becomes more prevalent. The curves in Fig. 11 suggest two interesting phenomena with regard to second generation twinning during TT compression. For one, the uncorrected curve suggests that the observed uncorrected C1-1 fraction should saturate between 40 and 50 vol%. At higher strains, the rate of second generation f1 0 1 2g tensile twinning should be greater than the rate of rst generation f1 1 2 2g compressive twinning. This means that there are more rst generation compressive twins that are reoriented by tensile twinning than new rst generation compressive twins being created; giving the impression that compressive twinning has considerably slowed down. This phenomenon explains why neutron diffraction based twin fraction measurements, that neglect re-twinning, report a so called saturation of twinning during deformation (Rangaswamy et al., 2002). Secondly, if the dependence of second generation f1 0 1 2g tensile twinning on rst generation compressive twin fraction continues to follow the same 3 Ax dependence, the fraction of second generation f1 0 1 2g tensile twins will account for half of the rst generation f1 1 2 2g compressive twin volume when that twin fraction (corrected) is approximately 75%. There is an obvious physical limitation to the fraction of an individual twinning mode. Assuming all grains were initially oriented for that twinning mode, the largest volume fraction possible is one. Based on this premise, the contribution of (corrected) rst generation f1 1 2 2g compressive twinning should eventually begin to asymptote to one. It has been shown that second generation f1 0 1 2g tensile twinning reorients the material back to an orientation similar to the initial parent orientation. This orientation is again ideal for f1 1 2 2g compressive twinning, and we have started to see third gener ation f1 1 2 2g compressive twinning in our small scale scans at the center of our sample TT compressed to 14.3% strain. It is possible that the compressive and tensile twinning and re-twinning may occur for several generations, with the primary factor limiting twinning being the ever decreasing effective grain size. 6. Summary and conclusions The purpose of the present work is to quantify the contributions the various twinning modes make to the 76 K deformation of zirconium. Because of twinning non-uniformities and second generation twinning, quantitative measures of twin fractions were made by combining data from large scale and small scale EBSD scans. For IP compression, the rate of rst generation f1 0 1 2g tensile twinning increases slowly up to approximately 17% strain beyond which the rate of twinning decreases. The fraction of rst generation f1 1 2 1g tensile twins initially increases with strain and then reaches a plateau at a fraction just below
471
5%. Above approximately 17% strain, second generation f1 1 2 2g compression twins within rst gen eration f1 0 1 2g tensile twins begin to make a contribution to the overall strain. The increase in hard ening rate is largely due to rst generation f1 0 1 2g tensile twinning reorienting material from a soft orientation to a hard orientation. At approximately 17% strain, the stress is high enough to activate second generation f1 1 2 2g compression twinning. The added deformation mechanisms results in a decreased hardening rate. For TT compression, f1 1 2 2g compression twinning and prismatic slip are the dominant deforma tion mechanisms following yield. Second generation f1 0 1 2g and f1 1 2 1g tensile twins observed within the rst generation f1 1 2 2g compression twins increase with the third power of the rst gen eration twin fraction. First generation f1 1 2 2g compression twinning reorients material from a hard orientation to a soft orientation. The soft orientation is ideal for second generation tensile twinning and/or prismatic slip. The rate of second generation tensile twinning depends largely on the fraction and size of rst generation f1 1 2 2g compressive twins. Acknowledgements This research is supported by the Department of Energy, Ofce of Science, Ofce of Basic Energy Sciences. Authors acknowledge discussions with Carlos Tom, Mike Baskes, Irene Beyerlein, Bjrn Clausen, George Kaschner, Marek Niewczas, and S.G. Srinivasan. Manuel Lovato performed the mechanical tests. Ann Kelly helped prepare the EBSD samples. References
Akhtar, A., 1973. Compression of zirconium single crystals parallel to the c-axis. J. Nucl. Mater. 47, 7986. Backx, P., Petrov, R., Kestens, L., 2007. Inuence of annealing treatments on twins in magnesium alloy AZ31. Mater. Sci. Forum 550, 375380. Battaini, M., Pereloma, E.V., Davies, C.H.J., 2007. Orientation effect on mechanical properties of commercially pure titanium at room temperature. Metall. Mater. Trans. A 38A, 276285. Bingert, J.F., Mason, T.A., Kaschner, G.C., Maudlin, P.J., Gray, G.T., 2002. Deformation twinning in polycrystalline Zr: insights from electron backscattered diffraction characterization. Metall. Mater. Trans. A 33, 955963. Garde, A.M., Aigeltin, E., Reed-Hill, R.E., 1973. Relationship between deformation twinning and the stressstrain behavior of polycrystalline titanium and zirconium at 77 K. Metall. Trans. 4, 24612468. Glavicic, M.G., Salem, A.A., Semiatin, S.L., 2004. X-ray line-broadening analysis of deformation mechanisms during rolling of commercial-purity titanium. Acta Mater. 52, 647655. Henrie, B.L., Mason, T.A., 2004. A fully automated technique for twin identication through electron backscatter diffraction. JOM 56, 241242. Jain, A., Agnew, S.R., 2007. Modeling the temperature dependent effect of twinning on the behavior of magnesium alloy AZ31B sheet. Mater. Sci. Eng. A 462, 2936. Jiang, L., Jonas, J.J., Luo, A.A., Sachdev, A.K., Godet, S., 2007. Inuence of {1012} extension twinning on the ow stress behavior of AZ31 Mg alloy. Mater. Sci. Eng. A, 302446. Kaschner, G.C., Bingert, J.F., Liu, C., Lovato, M.L., Maudlin, P.J., Stout, M.G., Tome, C.N., 2001. Mechanical response of zirconium II. Experimental and nite element analysis of bent beams. Acta Mater. 49, 30973108. Kaschner, G.C., Gray, G.T., 2000. Inuence of crystallographic texture and interstitial impurities on the mechanical behavior of zirconium. Metall. Mater. Trans. A 31, 19972003. Kaschner, G.C., Tom, C.N., Beyerlein, I.J., Vogel, S.C., Brown, D.W., McCabe, R.J., 2006. Role of twinning in the hardening response of zirconium during temperature change reloads. Acta Mater. 54, 28872896. Lebensohn, R.A., Tome, C.N., 1993. A self-consistent anisotropic approach for the simulation of plastic deformation and texture development of polycrystals: application to zirconium alloys. Acta Metall. Mater. 41, 26112624. Lou, X.Y., Li, M., Boger, R.K., Agnew, S.R., Wagoner, R.H., 2007. Hardening evolution of AZ31B Mg sheet. Int. J. Plast. 23, 4486. Mason, T.A., Bingert, J.F., Kaschner, G.C., Wright, S.I., Larsen, R.J., 2002. Advances in deformation twin characterization using electron backscattered diffraction data. Metall. Mater. Trans. A 33, 949954. McCabe, R.J., Cerreta, E.K., Misra, A., Kaschner, G.C., Tome, C.N., 2006. Effects of texture, temperature, and strain on the deformation modes of zirconium. Phil. Mag. A 86, 35953611. Meyers, M.A., Vohringer, O., Lubarda, V.A., 2001. The onset of twinning in metals: a constitutive description. Acta Mater. 49, 40254039. Nemat-Nasser, S., Guo, W.G., Cheng, J.Y., 1999. Mechanical properties and deformation mechanisms of a commercially pure titanium. Acta Mater. 47, 37053720. Proust, G., McCabe, R.J., Rogers, J.T., in press. Improvements on the automated twin identication technique using electron backscattered diffraction data. Metall. Mater. Trans. A. Proust, G., Tome, C.N., Kaschner, G.C., 2007. Modeling texture, twinning and hardening evolution during deformation of hexagonal materials. Acta Mater. 55, 21372148. Rangaswamy, P., Bourke, M.A.M., Brown, D.W., Kaschner, G.C., Rogge, R.B., Stout, M.G., Tome, C.N., 2002. A study of twinning in zirconium using neutron diffraction and polycrystalline modeling. Metall. Mater. Trans. A 33, 757763.
472
Reed-Hill, R.E., 1964. In: Reed-Hill, R.E., Hirth, J.P., Rogers, H.C. (Eds.), Role of deformation twinning in the plastic deformation of a polycrystalline anisotropic material, Deformation Twinning. Gordon and Breach, New York, pp. 295320. Reed-Hill, R.E., Buchanan, E.R., Caldwell, J.F.W., 1965. Quantitative measurement of fraction of tensile strain due to twinning in polycrystalline zirconium at 77 K. Trans. Metall. Soc. AIME 233, 17161718. Song, S.G., Gray III, G.T., 1995. Inuence of temperature and strain rate on slip and twinning behavior of Zr. Metall. Mater. Trans. A 26A, 26652675. Tome, C.N., Maudlin, P.J., Lebensohn, R.A., Kaschner, G.C., 2001. Mechanical response of zirconium: I. Derivation of a polycrystal constitutive law and nite element analysis. Acta Mater. 49, 30853096. Yablinsky, C.A., Cerreta, E.K., Gray, G.T., Brown, D.W., Vogel, S.C., 2006. The effects of twinning on the work-hardening behavior and microstructural evolution fo hafnium. Metall. Mater. Trans. A 37, 19071915. Zhang, Z.F., Gu, H.C., Tan, X.L., 1998. Low-cycle fatigue behaviors of commercial-purity titanium. Mater. Sci. Eng. A (Struct. Mater.: Prop. Microstruct. Process.) A252, 8592.