EVOLUTION OF r-VALUE DURING THE TENSILE DEFORMATION OF ALUMINIUM

The elastic strain ratios associated with specific ideal orientations as well as with particular experimental textures were derived from the second order Hill approximation. The single crystal elastic constants for pure aluminum and copper were used. Plastic strain ratios were then calculated for the experimental textures using the FC (full constraint), RC (relaxed constraint) and LW (least work) crystal plasticity codes. Tests were also carried out to determine the r-values along various directions in the sheet during Liiders band propagation, as well as the dependence on strain of the plastic r-value. It is shown that the observed strain dependence of the r-value is related to Liiders band propagation and to the evolution of the texture and not to the elastic-plastic transition.


I INTRODUCTION
It has been commonly observed that the conventional or cumulative plastic strain ratio rpi changes fairly rapidly during the initial stages of straining, finally adopting a near constant value at tensile strains of 15% or 20% (Welch et al., 1983;Daniel and Jonas, 1992). Such changes in r-value could have their origin in the difference between rel and rpt, where rel is the elastic strain ratio. However, it has recently been shown (Daniel and Jonas, 1992) that the evolution of r-value during the tensile deformation of steels is not primarily associated with the difference between rel and rnl although it may be affected by the propagation of Liiders bands. Instead, it appears to be a direct consequence of the evolution of the texture during the early stages of tensile deformation.
Depending on the actual texture present in the sheet, as well as on the inclination O of the axis of the tensile specimen with respect to the initial rolling direction, the texture changes can lead to increases or decreases in r-value, as well as to the relative absence of change.
The conclusions described above were based on crystal plasticity calculations and texture measurements carried out on bcc materials and on the occurrence of combined 110 <111> and 112 <111> slip. Of interest in the present investigation was the extent to which the general phenomenology outlined above applies to fcc metals such as aluminum, in which 111 <110> glide is taking place instead. It will be shown below that, while rel and rnl generally differ in fcc as well as bcc materials, the dependence of rnt on strain observed in samples of aluminum sheets is again related to Liiders band propagation and to texture evolution and not to the elastic-plastic transition. 150 J. SAVOIE ETAL.

II MEASUREMENT OF EXPERIMENTAL r-VALUES
Changes in width and length were measured continuously during the tensile testing of commercial 1145 grade aluminum and of both continuously and batch annealed versions of a 5XXX series experimental alloy. These measurements were carded out at the Alcan KRDC laboratories in Kingston on 190ram x 19mm specimens cut from sheets at several angles with respect to the rolling direction. The samples were then reduced to a constant width of 13mm over a length of 100ram. A double extensometer technique was employed, which allowed the total elongations and width variations to be measured simultaneously . Data acquisition of the load applied to the specimen was also performed; this was used to calculate the stress-strain curves. As no information about the thickness changes could be collected during straining (due to the specimen geometry), the elastic r-value (rd) could not be calculated directly from the experimental data. Furthermore, it was not possible to derive the thickness decrease indirectly from the length increase, as this requires knowledge of the volume change during extension. For this reason, in what follows, calculated ra-values are used, as they are known accurately (Daniel and Jonas, 1992). A further problem arose from the fact that Ltiders band propagation takes place rather inhomogeneously as waves of localized deformation pass through the material. Nevertheless, it was possible to measure the average width and thickness strains associated with the Liiders stage of deformation. Their ratio defines an average plastic r-value, referred to here as rLdrs. Experimental measurements of this ratio were carded out on each specimen of the batch and continuous annealed experimental material by using a single width extensometer. (The thickness strain was determined from the width and length strains in the manner described below.) The experimental plastic r-values (rpl) were evaluated on the basis of the following def'mition: rp % (1) where e and e are the plastic strains along the width and thickness directions of the tensile specimen, respectively. The latter were deduced from the relations: % In wl and t --(el "[" 'w) (2) where w and wi represent the initial (at the beginning of plastic straining) and final (at a def'med plastic strain) widths. For this purpose, it is assumed that the specimen volume remains unchanged during plastic straining and that: %= ln (3) where and 1 represent the initial and final gauge lengths. (Here a gauge length of 50.8mm was used.) In what follows, two different defmitions were used for the plastic r-values: the cumulative and the instantaneous. The former is defined by eq. (1), whereas the latter is given by: Instron extensometers of the LVDT type were used. The width extensometer was calibrated to work over the range 10mm-20mm. EVOLUTION OF R-VALUE 151 A (4) where Ae and As, represent small simultaneous strain increases. In terms of the ew vs e, plots presented below, the latter corresponds to the slope at any point, while the former is equivalent to the tangent defined by the coordinate ratio ew:e ,.
Stress-strain curves for the 1145 grade and the continuously annealed experimental alloy, both tested along the rolling direction, are illustrated in Figures l a and lb, respectively. The experimental material displays some Liiders elongation followed by the occurrence of dynamic strain aging. The corresponding experimental e vs t, curves are presented in Figure 2. The dependence of the r-value on elongation strain in the two materials (Figure 3) was calculated from the experimental curves of Figure 2. However, because of the fluctuations in the measured curves of Figure 2a and the discontinuous yielding associated with the material of Figure 2b, the r-values determined in this way display oscillations and irregularities. For this reason, a second order polynomial was fitted to the experimental curves of Figure 2, leading to the smoothed t; vs e, curves also shown in Figure 2. These smoothed curves were used to derive the cumulative and instantaneous r-value curves presented in Figures 4 and 5, respectively, for the two classes of material (grade 1145 and the continuous annealed experimental alloy).
The r-values pertaining to the 1145 material, shown in Figures 3a, 4a and 5a, change relatively slowly with strain. As will be shown below, these changes are associated with texture evolution. By contrast, the r-values associated with the experimental material evolve fairly rapidly with strain, particularly in the first 1.5% of deformation (see Figures 3b, 4b and 5b). These changes are associated with the transition from Liiders band propagation to generalized plastic flow, as will also be seen in more detail below.

III EXPERIMENTAL R-VALUES FOR THE PRESENT MATERIALS
The angular variation in instantaneous r-value at the beginning of generalized plastic deformation, determined as described above, was measured in the present three materials. The results are displayed in Figures 6a to 6c, respectively. All these materials were annealed (i.e. recrystallized). For comparison purposes, the r-values corresponding to a cold rolled material, taken from the work of MacEwen et al. (1992), are shown in  In the case of the cold rolled texture, the r-value increases more or less progressively from RD to TD (Figure 6d). This behaviour is explained by the presence (see below) A polynomial function of higher order would have reproduced the strong fluctuations of the curve.
These are due to the inherent characteristics of the extensometers used. The present calculations were carded out using the rate sensitive plasticity approach of T6th et al. (1988). This involves simulating the tensile deformation of a specimen inclined at an angle O to the rolling direction. In the case of very long specimens and when rotating grips are used, the macroscopic boundary conditions are the following: where /12 characterizes the macroscopic shear rate in tension (with El2 0 for O 0 and 90) and q the contraction ratio-/22//11, which varies from 0 to 1. The subscripts 1, 2, and 3 refer to the tensile, width and thickness directions, respectively. The r-value is related to q by the equation: r-q (6) 1-q In order to calculate the angular variation of r from the texture, several different microscopic conditions can be employed. In the FC (full constraint) case, the shear rates t 0. in each crystallite are prescribed as fully constrained. Some of the shear components are fully relaxed for the RC (relaxed constraint) model, whereas all three are partially relaxed by a factor :j for the present LW (least work) model (Engler et al., 1993). (Here gq :ij-te, with 0 < : < 1, and go as the fully relaxed value of the shear rate component under consideration). Accordingly, the following microscopic strain rate tensor t was used (expressed here in the rolling coordinate system, where the subscripts 1, 2 and 3 now refer to the rolling, transverse and normal directions, respectively). The above coordinate system coincides with the tensile reference frame for 0 0.
For an imposed value of q and a given grain orientation g, the local plastic work rate lq (g) (per unit volume) was calculated using the FC, one RC (pancake,/;13 and 23 fully relaxed) and LW deformation modes. Here: Smax (q(g) js. ,t.s (8) S=I where ),, and % are the shear rates and shear stresses, respectively, that apply to slip system s. A peculiarity of the rate sensitive model is that the z' values (which differ from z0) vary from one slip system to another, especially for high values of m. It is not therefore possible to calculate the plastic work rate per unit CRSS (or 0) using solely the sum of the crystallographic shears, as in rate insensitive models. The local plastic work rate must thus be defined as the sum of the crystallographic shears multiplied by the resolved shear stress .s pertaining to each slip system. The LW model used here is similar to that proposed in a previous paper (Engler et al., 1993), where a range of values is employed for , the degree of relaxation of each shear rate g/. In this model, the : ij are determined from the differences A I in deformation energy between the full constraint case and the special type of RC case in which all three shear rate components are relaxed. For this purpose, a shear capacity txij is defined as the ratio of the energy saved by relaxing the constraints to the amount of the fully relaxed shear: FC A._.__ Wq _ff, qc (9) An orientation associated with a shear component gij that leads to a large energy reduction will be associated with a higher value of s i than an orientation with a small reduction of energy. i2 is then calculated from: ct0 (10) where ct0" is defined, for simplicity, as the maximum shear capacity for all possible orientations in orientation space.

IV.2 Plastic Strain Ratio
For a polycrystalline sample, the macroscopic rate of plastic work corresponding to a given q value is given by the relations in Bunge (1970;1982). However, such calculations are based on the Taylor factor Mq(g) instead of on the rate of plastic work l.q(g) defined above. For this the following relations, where M is replaced by reason, W, were used. Various q-responses of the material were studied. As in the Taylor factor approach, the value of q that minimizes the macroscopic deformation energy was selected. Using the series expansion formalism, the rate of plastic work for a given texture and value of q is then def'med as: where wU" are the series expansion coefficients of the plastic work rates integrated throughout Euler space (Bunge, 1982). The CU coefficients were determined from the initial sheet textures by means of X-ray measurements, using standard techniques (see below).
Bunge (1982) has shown that the mean Taylor factor calculated along the lines of equation (11) can be used to derive the energy dissipated when the strain rate tensor is expressed in the specimen coordinate system as opposed to the one in which the ODF is described (i.e. the rolling reference frame). As the normal direction in the rolling reference frame coincides with the thickness axis of the sample coordinate system, the above relation becomes: , Z , , , . c;" To predict r-values from the present experimental ODF's, a data bank of w coefficients was prepared corresponding to several deformation modes. This was carried out in the rolling reference frame, for which O 0. For non-zero values of 19, rt was found by minimizing ff'q(O) with the aid of a polynomial regression. The r-value obtained in this way corresponds to the instantaneous value associated with a given texture rather than the average or cumulative r-value conventionally measured at 15 or 20% strain.
IV.3 Texture measurements For the texture measurements, four incomplete pole figures: 111}, 002 }, {022 and 113 (with 5 < ct < 80) were measured using a Siemens texture goniometer. After correction of the pole figure data, the ODF's were calculated using the series expansion method, leading to the determination of the C/ coefficients up to =22 (Bunge, 1982).
For the reduction of ghost effects, the exponential method was used in order to determine the odd terms of the series expansions (Van Houtte, 1991).

IV.4 Plastic r-values of single texture components
The angular variation of r-value was calculated using the pancake model for five single texture components. (The latter led to the best agreement with the experimental data, see section V). The ODF coefficients associated with each of these ideal orientations were calculated using a gaussian scatter width of to o 8 and a volume fraction M 100%. This type of gaussian distribution was used to account for the misorientation of individual grains with respect to the ideal orientation. The results are presented in Figure 7.
It is clear from Figure 7a, that the cube (recrystallization) component is responsible for the angular variation in r-value displayed in Figure 6a. The secondary components, such as the Goss and R (Figures 7b and 7c, respectively), disturb the symmetry associated with the cube texture: the presence of the Goss increases the value of r at 90 , while that of the R increases r at approximately 45.
The calculated r-values corresponding to the main rolling texture components are displayed in Figures 7b to 7e for the Goss, S3, Bs and Cu, respectively. Here it can be seen that there is a marked maximum at 45. However, the maximum r-value in the rolled material of Figure 6d is reached at 90. This can be attributed to the presence of the Goss component in the rolling texture (see Figure 7b).

V COMPARISON BETWEEN EXPERIMENTAL DATA AND MODELLING PREDICTIONS
The experimental r-values displayed in Figure 6 (for e 0.025) are compared with the modelling data obtained from the FC and pancake models in Figure 8. The initial texture coefficients of the sheet were employed in these calculations. For these materials, the FC mode does not reproduce the experimental data as well as the predictions of the pancake model. In the as-rolled material, the FC model leads to large deviations around 45; these discrepancies are consistent with the elongated grain shapes observed in the rolled samples. The use of the least work model leads to a compromise between the predictions of the pancake and FC models. Some calculations carded out in this way are presented in Figure 9. Here it can be seen that the predictions for the three annealed materials are in reasonable agreement with the observations and are about as accurate as those of the pancake model in Figure 8. For the as-rolled material, however, only the pancake predictions fit the experimental data in a reasonable way.
Note that the R 113 <575> and S 123 <634> components are not considered here as distinct as they lie close to each other and display similar calculated angular variations in r-value.

VI EFFECT OF TEXTURE EVOLUTION ON R-VALUE
The textures of selected specimens were measured after tensile deformation: along RD alone for the 1145 material and continuously annealed experimental alloy, and along directions taken at intervals of 15 between RD and TD for the batch annealed experimental alloy. In the latter case, the sample symmetry is monoclinic. The ODF's measured before and after tensile deformation are shown in Figure 10 for the batch annealed experimental alloy. Here the tensile specimen was aligned along RD. The results obtained for specimens aligned at an angle of 0 with respect to RD are presented in Figure 11. It can be seen from these ODF's that the textures changed significantly during plastic deformation (to e 0.15 0.20). These texture changes can be expected to contribute to r-value evolution during tensile testing. In order to test the magnitude of this effect, r-values were calculated using the texture coefficients determined before and after tensile straining. Then these were compared with the corresponding instantaneous experimental r's. The results are summarized in Table 1 in the form of the changes in r, t r, taking place during testing or predicted from the texture data using the pancake model. For seven of the eight samples, the change in r-value during straining can be related directly to texture evolution, as the predicted increases or decreases correspond approximately to the experimental ones (within a dir difference of 0.05). However, in one case, the changes are of opposite sign and differ by 0.  The angular variation of the elastic and plastic strain ratios pertaining to the main ideal orientations observed experimentally in fcc sheets was presented in Figure 7. For this purpose, ra(O) was calculated using the second order Hill approximation (Bunge, 1974).
The dependence of Young's modulus on O was also calculated using the elasticity model ( Figure 12). As aluminum displays little elastic anisotropy due to its unique elastic properties, similar calculations were also carded out using the elastic constants of pure copper ( Table 2). Given that, in the model calculations, the plastic anisotropy depends only on the texture, the rpl predictions for aluminum and copper are identical. By contrast, the elastic data for the copper display much stronger elastic anisotropy than those for the aluminum, although the two materials follow the same tendencies.
As expected, E(O) and r a (O) display opposite dependences on O. The copper r a curves follow the general tendencies of the rpt (O) curves, even though it was shown above that rpt and r a are not strictly proportional along a given O direction. The aluminum rd-values remain close to 1; thus re is greater or less than rt depending on whether the latter is less than or greater than 1, respectively. The elastic-plastic transition in r-value is thus expected to involve a rapid change from a value close to 1 to r as soon as plastic deformation is initiated (see Figure 13). VII.2 Transition from Liiders band propagation to generalized plastic flow Experimental measurements of the r/ers-values were carded out on each specimen of the batch and continuous annealed experimental material, as explained in section II. In the batch annealed material, the average value of rLaders determined in the O range 0 to 45 , inclusive, was 0.5, whereas it was 1.3 over the O range 60 to 90. For the continuous annealed material, the average over the O range 0 to 60 , inclusive, was 0.4, whereas it was 1.0 over the 75 to 90 interval (Table 3). It thus appears that the measured r.de,-values for both experimental materials differ distinctly from  those determined after conventional flow has begun (see Figures 6a and 6b). Let ct represent the inclination of the Liiders front to the tensile axis. Then the value a 54.7 applies to isotropic materials (r 1), while ct is greater than cti for the first O range (e.g. ct 64.3 for r 0.3) and less than afor the second O range (e.g. ct 52.2 for r 1.5) (see Appendix). Thus, tensile testing over the O range 0 to 45 or 0 to 60 , which led to rtd,r values below one (more thickness decrease than width decrease) seems to involve Liiders fronts with higher inclinations with respect to the tensile axis; by contrast, testing over the O range 60 to 90 , which led to rLrs values above one (i.e. more width decrease than thickness decrease) appears to be associated with Ltiders front inclinations that are closer to 45. It thus appears that, in the experimental alloy, the r-value changes suddenly during the transition from Liiders band propagation to generalized plastic flow. By contrast, during tensile testing of the 1145 alloy, in which no Liiders bands are observed, the transition between r a and rpi is much smoother, although both materials display the same kind of angular dependence of r-value on angle O. Schematic illustrations of the expected strain dependences of r in the cases of homogeneous and inhomogeneous yielding are presented in Figure 13. VIII CONCLUSIONS 1. The angular variations in r-value (with respect to the rolling direction) were determined from experimental data using the instantaneous definition of r. The variations in the annealed materials were consistent with the cube texture being dominant. The r-value profile ha the as-rolled material was consistent with the presence of cold rolling texture components. 2. The angular variation in r-value of the annealed materials can be reproduced very accurately using both the pancake and least work deformation modes. However, the as-rolled material data could only be reproduced with the pancake mode. This behaviour is consistent with the presence of pancaked grains after rolling. 3. The plastic r-value changes with strain when both the instantaneous and cumulative definitions of r are used. These changes can be attributed to the evolution of the texture during deformation. 4. The elastic-plastic transition in r-value displays a rapid change from a value close to 1 (for aluminum) to rpt as soon as plastic deformation is initiated. When Liiders band propagation occurs, the r-value changes suddenly during the transition from Liiders band generalized plastic, flow. These changes can involve rapid increases or decreases in r-value. By contrast, for materials in which no Liidres bands are observed, the transition between ret and rt is much smoother.