Acta mater. Vol. 44, No. 4, pp.

141S1435,

1996

Pergamon

093%7151(!45)002!30-1

Elsevier ScienceLtd Copyright 0 1996Acta Metallurgica Inc. Printed in Great Britain. All rights reserved 1359-6454/96 $15.00+ 0.00

RECRYSTALLIZATION TEXTURE DEVELOPMENT COMMERCIAL Al-Mn-Mg ALLOYS

0. DAALAND and E. NES*

IN

Hydra Aluminium a.s., R &D Centre, Karmoy, N-4265 HBvik, Norway and INorwegian Institute of Technology, Department of Metallurgy, University of Trondheim, Alfred Getz vei 2b, 7034 Trondheim, Norway (Received 6 December 1994; in revised form 5 July 1995)

Abstract-A detailed investigation of the evolution of recrystallization textures during annealing after hot
rolling of an AllMnlMg alloy and after hot and cold rolling of an AllMnO.SMg alloy has been carried out, using the EBSP-technique for identification of the individual grain orientations. The growth of grains with a W( 111) orientation relationship to components in the rolling texture has been followed and compared to the growth of random oriented grains. No oriented growth effect due to an enhanced mobility of 40( 111) boundaries could be detected. However, during low temperature annealing of one of the alloys grains belonging to different texture components were observed to grow at different rates. This special case is associated with a precipitation reaction occurring simultaneously with recrystallization, and the observation is interpreted as a precipitation-induced-effect reflecting that particle stimulated nucleation (PSN) is more sensitive to precipitation than nucleation from the competing W( 111) sites. In order to illustrate the effect of grains growing at different rates on the structure and texture development, a computer model has been applied for simulating the recrystallization reaction. Reasonable predictions of the kinetics, grain size distribution and final texture (volume fractions) have been obtained.

1. INTRODUCTION The present paper is the result of an international collaborative research programme involving both industrial and university partners (see Ref. [l]), the objective being to obtain a better understanding of the evolution of texture and microstructure during hot and cold rolling and subsequent annealing of commercial aluminium alloys. In the first part of this work [l], the development of microstructure and texture during hot rolling was investigated and interpreted. The major part of that work concentrated on the origin of cube texture, which is a major cause of anisotropy after hot rolling. The present paper is focused on the recrystallization behaviour, with special emphasis on the mechanisms responsible for the development of the recrystallization texture on annealing. The origin of recrystallization texture formation has been a subject of debate for many years, and has been ascribed to both oriented nucleation and oriented growth interpretations. The objective of the present paper is to study in detail the nucleation and growth of strong 40(111) components during annealing. Special focus has been directed towards the influence of precipitates on the recrystallization texture development, which is often a major problem during processing of industrial/commercial alloys.

2. EXPERIMENTAL DETAILS

2.1. Alloy conditions and annealing treatment The investigated alloys are designated as AllMnlMg and AllMnOSMg, and details about chemical compositions and hot rolling are given in Ref. [l]. For in-depth studies of recrystallization texture development it is preferable that rather strong textures are produced during annealing. In the case of the AllMnlMg alloy a suitable condition for such a study was obtained by simply using the alloy in the final hot rolled condition (i.e. exit tandem mill). Annealing treatment of this material was subsequently performed in a salt bath at 330C. To obtain a suitable condition for recrystallization studies in the case of the AllMnO.SMg alloy, samples taken at the 20 mm hot rolling gauge were further processed by cold rolling in a laboratory rolling mill to a strain of E = 3.5. Annealing treatment was subsequently performed in a salt bath at temperatures 300, 350 and 450C. 2.2. Texture measurements Details about the measurement of figures are given in Ref. [l]. In this paper, ODFs were corrected with respect to the according to the method of Liicke and X-ray pole some of the ghost errors co-workers

1413

1414

DAALAND and NES:

RECRYSTALLIZATION

TEXTURE

[2,3] in which a model ODF consisting of a few Gauss scattering components is adapted to the experimental ODF (for details see Ref. [4]). Further evaluation of the ODF data has been obtained by calculation of the volume fraction of the main texture components. The growth of recrystallized grains was followed using the EBSP-technique (described in more detail in Ref. [ 11)on a series of partly recrystallized samples. By inspection of pole figures, obtained from the EBSP-examination, the grains were categorised into several ideal components {hkl)(uvw) corresponding to the main components found in the recrystallized texture. If the observed grain orientation was within a 15 scatter (in all directions) from a defined ideal component {hkl}(uvw) it was classified as belonging to that specific component, otherwise, it was designated as being part of the random component. In order to link individual orientations to the microstructure, electron channelling contrast micrographs of the investigated areas were recorded. For each annealing condition, on average, 200 grains were analysed and their size (grain area) measured from the channelling contrast micrographs using a computerised image analyser (Kontron Videoplan). 2.3. Optical microscopy Optical microscopy has been used for the following purposes: (i) measuring the fraction of recrystallized material in partly recrystallized samples. A point counting analysis has been used in order to measure the volume fraction recrystallized material (X,) versus annealing time. A grid with 121 grid points was positioned in the focusing eyepiece of a Leitz optical microscope. For each sample, more than 700 grid points were taken into account, i.e. more than six randomly selected areas were considered. The grid spacing relative to the scale of the structure is the

most critical item in determining the efficiency of the analysis [5]. For a recrystallized fraction of less than 0.5, this spacing was chosen in such a way that the average number of points falling in any one recrystallized colony did not exceed unity. If the fraction recrystallized was greater than 0.5, the same rule was applied with respect to the non-recrystallized colonies. To meet these requirements, the magnification was adjusted between 200x and 500x depending on the grain size and the degree of deformation prior to annealing; (ii) measuring the interfacial boundary area between untransformed and transformed material (to be used in the analysis of Gokhale and DeHoff [6]). The area per unit volume, S,, for the boundary between recrystallized and non-recrystallized material was determined using the following expression: S, = 2N, (1)

where NL is the average number of intersections (which separates transformed and untransformed material) per unit length of a random test line. The expression [equation (l)] is general, and involves no assumptions about the form of the structure. The only requirement when applying equation (l), is that the test lines are positioned randomly, i.e. in random directions and at random inspection planes [7]; (iii) grain size measurements. Average grain size in fully recrystallized structures were measured using a lineal intercept technique. A minimum of 200 grains were taken into account for each sample.
3. RESULTS

The AllMnlMg alloy after final hot rolling contains a fully developed and strong /I-fibre (centre section of the sheet) as illustrated in Figs l(a) and (c),
_(112) _( 123)
<631>
I

_( 011)
<I L>

Q, -cona.

<ill>

f Id T
15

b YJ _

AHMnlMg

alloy

\ 0,/o-o---o\ 2 /O 0 \ / \ Oio/

10

<,_o,o

All MnO.BMg alloy

lb) (a) Fig. 1. Rolling texture in the two alloys, shown by means of ODFs (a) AllMnlMg
gauge (centre section); (b) AllMnOSMg

alloy, final hot rolling alloy cold rolled to strain, 6 = 3.5 (surface section); (c) /!-fibres.

DAALAND

and NES:

RECRYSTALLIZATION

TEXTURE

1415

for more details on texture evolution see Ref. [l]. It is emphasised that the ODF also contain peaks located at or in the vicinity of the cube position {OOl}( 100). Furthermore, orientations rotated away from the ideal cube position around RD (towards the Goss orientation {110}(001)) and TD are found. A characteristic feature in the surface texture of the hot rolled AllMnO.SMg alloy slab, is a strong (001) (110) shear component (45 ND-rotated cube), penetrating to depths of l/3 of the sheet thickness (see Ref. [1], Fig. 17). The central region of the hot band displays a more normal rolling texture, showing a strong maxima at the S component { 123}(634) (16 x random). It is well known that the {OOl}(llO) orientation, when subjected to plane strain deformation, rotates preferentially towards the Cu orientation { 112}( 111). During cold rolling of the AllMnO.SMg alloy to a strain of c = 3.5 the surface region follows this pattern of rotation to a certain extent resulting in the development of a /?-fibre texture [Figs l(b) and (c)] with intensities of 17, 14 and 8 x random at the Cu, S and Bs {110}(112) positions, respectively. 3.1. Evolution of the recrystallization
texture

2o

fit 15 % 3 9" n -0 100 200 300 400 500 600 10

ANNEALING TIME [s] (4

-7 g 5 d

.ic
40 30

", 20 f -I 10 9 n 0 1000 2000 3000 4000 5000 6000

ANNEALING TIME [s] (W Fig. 2. The development of specific texture components during recrystallization: (a) AllMnlMg alloy annealed at 330C; (b) AllMnOSMg alloy annealed at 300C.

In order to get an overall picture of the texture transition in the two alloys during recrystallization, a detailed ODF analysis (computer-program developed by the Aachen-group [4]) has been performed on samples prepared to various stages of recrystallization. Figures 2(a) and (b) show the volume fractions of the main texture components derived from the ODFs and plotted as a function of the annealing time. In the AllMnlMg alloy case [Fig. 2(a)] the rolling texture is qualitatively retained in the early stages of annealing. It can be seen that the S component, in the unannealed state, occupies a much larger volume (N 32%) compared to the other main components of the rolling texture (16% Bs and 14% Cu). Even though the Bs orientation shows a higher peak intensity in the ODF, the S orientation possesses a larger volume fraction due to its lower symmetry (four variants instead of two which is the case for the Cu and Bs orientation). On annealing, the cube texture {OOl}( 100) evolves progressively with a simultaneous decrease in all the main rolling texture components, resulting in an approximately 23% cube orientation in the fully recrystallized condition. The texture in the fully recrystallized condition is shown in Fig. 3(a), by means of ODF and pole figure, [the positions of the ideal orientations are given in Fig. 3(d)]. It is to be noted that the cube texture is accompanied by small amounts of Goss { 1 lO}(OOl) (-3%) and R texture {123}(634) (-5%), but they are occupying only a few percent of the total volume. No other specific components are evident which indicates that a significant amount of the material (N 70%) consists of randomly spread orientations. As shown in Fig. 4(a), a rather homogeneous microstructure is obtained in the fully recrystallized condition.

In the AllMnO.SMg alloy case [Fig. 2(b)] the Cu and S orientation are found as the main components in the as-rolled structure occupying approximately 30 and 27%, respectively. During annealing at 300C the main recrystallization textures, ND rotated cube {001}(130) and P {llO}(lll), are seen to grow at the expense of all rolling texture components Cu, S and Bs. Together with the main recrystallization texture components ND rotated cube and P it can be seen that small amounts of the exact cube orientation { lOO}(OOl) evolve in the later stages of annealing. In the ODF this is observed as a smearing out of the ND rotated cube towards the exact cube. In the fully recrystallized condition, shown in Fig. 3(b), the recrystallization texture contains approximately 42% ND rotated cube, 18% P, 4% cube and 10% R texture { 123}(634). The amount of material associated with the random spread of orientations is therefore less than 30%. It is to be noted that the resulting grain structure is very inhomogeneous containing a wide spread of grain sizes. As shown by the micrograph in Fig. 4(b) the grain size ranges from less than 10 pm to approximately 70pm. The average grain size in this structure, using a lineal intercept method, was measured to be 57 pm. Note that this grain size effect is a surface region phenomenon typically observed during low temperature (or batch) annealing. Annealing the material at a higher temperature, i.e.

all 0.5 1 /y+-qy: 2

LEVELS :

0.5

(111)

ND mLCutm

(4 lb)

(c)

(4

Fig. 3. Annealing texture, fully recrystallized condition (a) AllMnlMg alloy annealed at 330C; (b) AllMnOSMg alloy annealed at 300C; (c) AllMnO.SMg alloy annealed at 450C; (d) positions of the ideal orientations refered to in this paper.

DAALAND

and NE%

RECRYSTALLIZATION

TEXTURE

1417

1418

DAALAND

and NES:

RECRYSTALLIZATION

TEXTURE

0.4 5 0 -

Y s-o,4 c g-03

-1,2 -I,6
10

. ,,,,I

100

I
TIME

1000 [secl

10000

ANNEALING (4

son-Mehl (time-dependent) case, respectively. By determination of the boundary area (S,) in addition to the volume fraction of recrystallized material (X,) per unit volume, it is possible to deduce the nature of the time dependence of the nucleation process i.e. whether it is of site saturation or Johnson-Mehl type kinetics. This analysis is based on a Laplace transform methodology proposed by Gokhale and DeHoff [6] and further applied to recrystallization by Vandermeer and Rath [8]. A stereological point counting procedure was used in order to evaluate S, (for details see Section 2). From these data the extended interfacial area per unit volume, S,,, can be calculated from the following definition (according to DeHoff

PI>:
lOOGil j__

1000

7y_;
?AHMnlMg ?
10

&x=(1.

(3)

w lo-

,,,,I

100

TIME

AIlMnOSMg

1000 [set]

., L

alloy alby 10000

ANNEALING

( 8 Fig. 5. Progress in recrystallization for the two alloys. (a) Fraction recrystallized material as a function of annealing time, plotted in order to determine the JMAK-exponent. (b) Plot of the extended interfacial area per unit volume as a function of annealing time. 350 and 450C results in a progressively much finer grain structure with an average grain size of 40 and 12 pm, respectively, see Fig. 4(c). In addition, a much weaker texture is found showing a relatively larger spread around the main component, Fig. 3(c). The P component is in this case very weak with an intensity of less than 1 x random. Note also that some intensity seems to remain from the deformation texture at the Bs position. 3.2. Nucleation of recrystallization

A plot of S,,, versus annealing time now gives us a straight line with slope m, as shown in Fig. 5(b). According to the above mentioned methodology the nucleation kinetics will be site saturated or time dependent (Johnson-Mehl) when the quantity 3m - 2n equals 0 and 1, respectively. In the AllMnlMg alloy investigated here, the quantity 3m - 2n was found to equal 0, while in the AllMnO.SMg alloy a value of 0.1 was found. Consequently, it is concluded that in both cases the available nucleation sites are activated in a time so short it may be considered to be zero relative to the total annealing time, i.e. a situation near site saturation kinetics is the most realistic. 3.2.2. Nucleation sites. The nature of the nucleation sites for development of a major cube texture component in the case of the AllMnl Mg alloy, was documented in detail in Ref. [l]. A similar detailed characterisation of nucleation sites corresponding to
RD

3.2.1. Nucleation kinetics. The progress of recrystallization is well described by the familiar equation, based on the Johnson-Mehl-Avrami-Kolmogorovtheory: X, = 1 - exp( - kt) (2) where X, is the fraction recrystallized material, k is a constant and the exponent n usually is referred to as the JMAK-exponent. By plotting log(ln(l/l - X,(t))) vs log t the JMAK-exponent can be obtained as shown in Fig. 5(a). The n value is found to equal 2.4 for the AllMnlMg alloy and 1.6 for the AllMnOSMg alloy. These values correspond to those commonly found in recrystallization but are much smaller than those expected from the classical analysis. With constant growth rate in three dimensions n should be 3 and 4 for the site saturation and John-

Fig. 6. (111) pole figure showing the orientation of subgrains within a cube or close-to-cube oriented band found in the cold rolled substructure of the AllMnO.SMg alloy. (EBSPstep scan, surface section of the sheet.)

DAALAND

and NES:

RECRYSTALLIZATION

TEXTURE

1419

the major recrystallization textures in the case of the AllMnO.SMg alloy has proven to be difficult, due to the small subgrain size obtained after cold rolling. The EBSP-technique has some limitations according to resolution when the subgrains become less than 0.5-l pm in diameter, giving rather diffuse EBSPs which are often difficult to interpret/index. An extremely complicated microstructure is obtained after cold rolling the AllMnO.SMg alloy, because of the high accumulated strain (6 N 6.5, after hot and cold rolling) felt by the material. Orientation measurements collected during a step scan in the normal direction revealed that each texture component on average was less than 1 pm thick. In the investigation of the as-cold deformed structure, no direct evidence was found of material having the ND rotated cube or P orientation. As only very small volumes of deformed material are needed for nucleation of recrystallization, this is perhaps not surprising. The only indication of a possible nucleation site detected is shown by means of a (111) pole figure in Fig. 6. A banded feature of subgrains with cube orientation (showing scattering towards rotated cube), was found during an EBSP-step scan in the rolling direction. It was possible to follow this cube oriented band for approximately 50 pm, along the rolling direction. Furthermore, it is interesting to note that the orientations in question are indeed present in the deformation texture, showing up as small peaks indicated by arrows in the ODF shown in Fig. l(b). This does, of course, demonstrate that the deformation actually produces sites with the ND rotated cube and P orientation. It is expected that these orientations are present in the matrix as banded features being of a similar type to the cube band sites observed in the Al 1Mn 1Mg alloy, see Ref. [ 11. Concerning nucleation characteristics it is furthermore assumed that the special dynamic recovery properties of the cube orientation do, to some extent, also apply to the ND rotated cube, giving these sites an advantage. 3.3. Growth of recrystallized grains

out that due to the rather simplified classification into only two categories (cube and random) the random category of grains in addition to truly random oriented grains also includes small amounts of Goss {110}(001) and R texture {123}(634). Looking at the pole figure for the random category in Fig. 7(a), a diffuse texture can be observed. However, as shown in Section 3.1, the Goss and R texture components are not a dominant part of the recrystallization texture in the fully transformed state and hence they are not expected to represent a major proportion of the grains in the random category. Note that the largest grain in the random category was, without success, checked for a possible 40( 11 I)-relationship to the main components in the /I-fibre. Based on

cube

(4

random

F30 3 f$ 25 iii 2 2o cl g 15 d a 10 w s B 2 5-

_ 0 0 100 200 ANNEALING 300 400 TIME [SEC] 500

(b)

The growth of recrystallized grains in the two alloys has been investigated in detail by applying the EBSP-technique. From a series of partly recrystallized samples the evolution was followed by simultaneous determination of size and crystallographic orientation of the individual recrystallized grains in the structure. The grains were classified into several groups or ideal orientations {hkl}(uuw), corresponding to the main recrystallization textures observed in each case, i.e. cube {100}(001> oriented grains in the case of the AllMnlMg alloy annealed at 33OC, and ND rotated cube { 100}(310), P { 1 10}<111) and cube {100}(001) grains in the case of the AllMnO.SMg alloy annealed at 300C. 3.3.1. AllMnIMg alloy. Examples of grain orientations, according to the orientation classification in the case of AllMnl Mg alloy, are shown by means of EBSP pole figures in Fig. 7(a). It should be pointed

$-z~ y-::
B t;; 10
I 4 0 0 0 100 200 300 ANNEALING TIME [set] (o) of grains belonging to different texture components for the AllMnlMg alloy during annealing at 330C. (a) (111) pole figure representation of the EBSP results showing the orientation classification. (b) Average grain diameter vs annealing time. (c) Largest grain diameter vs annealing time. 400 500 AIlMnlMg alloy

Fig. 7. Evolution

1420

DAALAND

and NES:

RECRYSTALLIZATION

TEXTURE

microstructural observations it is reasonable to assume that the random oriented grains are basically associated with the second phase particles, i.e. they have developed by particle stimulated nucleation (PSN). Nucleation from other sites, such as grain boundary regions, may however also be important sources for the random oriented grains. In order to gain insight into the growth reaction the grain size measurements (equivalent diameter calculated from measurement of grain area) are presented in two different ways: the average grain diameter (D) vs annealing time [Fig. 7(b)] and the diameter of the largest grain (D,,) vs annealing time [Fig. 7(c)]. Following the evolution of average size, the cube grains are for all annealing times seen to be somewhat larger than grains of random orientation. A small growth rate difference between the components might be observed. However, it is difficult to say whether this small difference really reflects a higher growth rate of the cube grains. In the fully recrystallized condition the average cube grain is approximately 15% larger than the average random grain. Looking at the evolution of the largest grain [Fig. 7(c)] no growth rate difference between the components is observed, although still cube grains are consistently larger than grains belonging to other texture components. Table 1 lists all the important transformation data obtained from the EBSP-investigation in the case of the AllMnlMg alloy annealed at 330C. Note that by assuming random distribution of sites and spherical grains the measured area fraction of the texture components is equal to the volume fraction. In addition to the area or volume fraction of cube and random oriented grains, the table also includes the approximate nuclei/grain-density, calculated assuming the growing grains to be spherical. The densities are seen to be approximately constant during the entire recrystallization period, i.e. site saturation seems to be fulfilled in this material. This is in accordance with the results obtained from the stereological procedure used in Section 3.2. It is important to be aware that the number of grains of cube orientation is significantly smaller (approximately 4 with the volume fraction times, in accordance measurements) than the number of grains of other orientations. In the fully recrystallized condition, the cube oriented grains were measured to account for 29% of the recrystallization texture, the remaining part consisting of randomly oriented grains. Note that the volume fraction of cube texture measured

using EBSP is in reasonable agreement with the volume fraction calculated from the ODF-data (Section 3.1). 3.3.2. AllMnO.SMg alloy. EBSP-pole figures of the various types of grains in the case of the AllMnO.SMg alloy, are shown in Fig. 8(a). The evolution of grain sizes during annealing are shown in Figs 8(b) and (c). From the average grain size vs annealing time-plot [Fig. 8(b)] it is obvious that in this alloy a strong relationship exists between the size of the grain and its orientation. Grains of different orientations grow at different rates, the ND rotated cube and P grains growing significantly faster than the random component, though with a steadily falling growth rate. Both these recrystallization texture components are according to the oriented growth theory, associated with a fast growing 40(111) orientation relationship with respect to the Cu component. Note that the evolution of grains with orientations close to the exact cube orientation { lOO}(OOl) has been followed throughout the entire transformation, although in the ODF investigation (Section 3.1) the cube component was found to evolve only in the later stages of recrystallization. In order to distinguish between the exact cube {lOO}(OOl) and the ND rotated cube {100}(013) component the magnitude of the allowed deviation from the ideal was reduced to 5 in the case of the exact cube. Even though the statistics for the cube oriented grains are rather poor, it seems that this grain orientation also is evolving at an enhanced rate compared with grains of random orientation. Due to insufficient data for complete characterisation, the cube component is given by a broken line in Fig. 8. Note that the grains belonging to the random category seem in fact to stop growing completely after reaching a size of approximately 7-8 pm in average diameter. As can be seen from Fig. 8(b) this corresponds to approximately 800 s of annealing or a total of 20% recrystallization. The small and equiaxed grains with a random orientation are clearly associated with the large particles in the structure, and it is reasonable to assume that they have been formed by particle stimulated nucleation (PSN). During the rest of the recrystallization period, the growth of these PSN-grains are obviously hindered and they are often engulfed by larger recrystallized grains becoming island grains. Further insight into the growth reaction is obtained by monitoring the largest grain as a function of annealing time, shown in Fig. 8(c). Also in this plot a significant

Table 1. Volume fraction and number of erains for the texture conmonents in the AllMnlMe allov (330C). determined by EBSP (values in parenthkes correspond to the &me fraction obtained from OGF:data) Annealing Volume W) time (s) fraction Cube Random Tntnl _-_-. Cube Random 70 2 6 R 4 x-lo 2 x 104 100 5 9 14 5 x- ;013 2 x 104 140 10 16 26 5 x;o 2 x 104 180 4 26 40 5 x ;013 2 x 104 240 20 42 62 3 x 10 1 x 1014 500 29 (23) 71 100 3 x lOI3 1 x 104

No. of grains per unit volume (II-)

DAALAND

and NES:

RECRYSTALLIZATION

TEXTURE

1421

ND rot. cube

a3
k,

.$

..,.,*....
.: .., ,..
ID

.. ..,,

*.

...@*,,

,..I

,..

...

cube

(a)

random

Random

* _

2 0
0 1000 ANNEALING (b)

2000
TIME [set]

3000

4000

.$70 2 60

5 =; 50 z 5 40 30

g 20 iz = 10 = 0 0 0 1000 ANNEALING 2000

All MnO.SMg 3000 TIME [set]

alloy

4000

(c)
Fig. 8. Evolution of grains belonging to different texture components for the AllMnOSMg alloy during annealing at 300C. (a) (111) pole figure representation of the EBSP results showing the orientation classification. (b) Average grain diameter vs annealing time. (c) Largest grain diameter vs annealing time.

growth rate difference between the components is observed in the early stages of recrystallization. However, a careful inspection of the transformation curves reveals that it is only in the initial stage (< 800 s of annealing) that the 40 (111) grains (ND rotated cube and P) have their big advantage. On

further annealing the largest grain for all components seems, in fact, to grow at the same rate. An example of an EBSP-analysis of the various types of grains in a partly recrystallized specimen is shown in Fig. 9. The two largest grains A and B in the micrograph exhibit the ND rotated cube and P orientation, respectively. Both these orientations grow rapidly and consume the slower growing PSNgrains, which are clearly associated with the particles. The neighbouring regions next to the large recrystallized grains have been analysed carefully in order to monitor the relationship to the deformation matrix. Based on an oriented growth mechanism it should be expected that a 40(111) orientation relationship is frequently found. However, as seen from the pole figures in Fig. 9, there is no apparent preferred orientation adjacent to the large recrystallized grains. Further investigation of the substructure-areas next to the large recrystallized grains revealed that each deformation texture component on average was less than 1 pm thick. The whole B-fibre was found to be present within a distance of less than -5 pm. The volume fraction and grain densities of the classified texture components obtained from the EBSP-measurements on the AllMnO.SMg alloy, are presented in Table 2. The most curious observation in this alloy is the drastic reduction in the number of nuclei in the early stages of recrystallization. As can be seen from the table, a significant drop by a factor of approximately three is observed for all the texture components between 400 and 800 s of annealing. On further annealing the grain/nucleation densities remain reasonably constant. Concerning the density numbers it is to be noted that the number of grains with a random orientation (i.e. PSN-grains) is 5-10 times larger than grains of ND rotated cube, P or cube orientation. However, since the PSN-grains are effectively outgrown by these latter rapidly growing components, the PSN-grains make a rather small contribution to the final texture. In the fully recrystallized condition the random category was found to account for 9% of the total volume of orientations. The rest of the material contained 48% ND rotated cube, 26% P and 17% cube orientation. Because of the large difference in growth rates between grains of different orientations a characteristic feature in this AllMnO.SMg alloy after complete recrystallization at 300C is a very inhomogeneous grain structure, containing a large range of grain sizes. The grain shapes are irregular, often with jagged boundaries as shown by the channelling contrast micrograph in Fig. 9 or in more detail in Fig. 10. The large grains (marked B in Fig. 10) were found to contain a large number of small island grains (marked A in Fig. lo), most of them surrounding large particles. This structure clearly emerged because some grains grew more rapidly than others, the slower ones subsequently being surrounded. As was shown by the ODF in Fig. 3(c), annealing of the AllMnO.SMg alloy at a higher temperature (450C) results in a much weaker texture

1422

DAALAND

and NES:

RECRYSTALLIZATION

TEXTURE

ND

<ill>-poles RD

<ill>-poles
RD

Fig. 9. Channelling contrast micrograph of partly recrystallized AllMnOSMg alloy (annealed 400 s at 300C). The orientation of both recrystallized grains and the surrounding subgrains is illustrated by means of EBSP (111) pole figures. The numbers in the micrograph correspond to the numbered (111) poles.

and a totally different grain structure [Fig. 4(c)]. The growth reaction at 450C has not been investigated, due to the very short annealing times required to produce partly recrystallized structures. However,

measurements of grain sizes have been performed on a fully recrystallized sample annealed at 450C. These grain size results are presented in Table 3, which also includes the grain sizes obtained for the

Table 2. Volume fraction and number of grains for the texture components in the AIlMnOSMg alloy (3OOC), determined by EBSP (values in parentheses correspond to volume fractions obtained from ODF-data) Annealing Volume W) time (s) fraction ND rotated P Cube Random Total ND rotated P Cube Random cube 2 cube 2 x310 9 5014 2 2 6 1 200 1 400 3 1 1 4 x9,014 x 104 x lOI3 x 105 6 6 3 3 800 10 5 1 6 x2:0 x 10 x 10 x 104 6 5 4 4 1400 21 11 4 10 x5:0 x 10 x 10 x 1014 6 4 3 3 1800 34 12 5 10 x6:O13 x 10 x 103 x 104 4000 48 (42) 26 (18) 17 (4) 9 5 4 3 3 ZOl~ x 10 x 10 x 104

No. of grains per unit volume (m-))

DAALAND

and NES:

RECRYSTALLIZATION

TEXTURE

1423

Fig. 10. Channelling contrast micrograph from a specimen of the AllMnOSMg alloy recrystallized at 3OOC,showing small island grains (A) associated with particles which are outgrown by the larger grains (B) having a ND-rotated cube, P or cube orientation. AllMnlMg and AllMnO.SMg alloy annealed at 330C and 3OOC, respectively. Note that during high temperature annealing of the AllMnO.SMg alloy, the 40( 11 l)-grains (ND rotated cube and P) are of the same size as the random grains, which indicates that in this case the PSN-grains are able to grow as rapidly as the others. Simple calculation shows that the nuclei/grain-density (or the number of activated sites) is the same during annealing at 300 and 450 (of the order 10 5mm3). Pole figures determined from the EBSP investigation on fully recrystallized samples are shown in Fig. 11. Each EBSP pole figure represents approximately 500 grains and, as can be seen, they reproduce the essential feature of the recrystallization texture measured using X-ray diffraction (see pole figures in Fig. 3).
Table 3. Average

3.3.3. Precipitation-eflects. The coarse-grain-structure observed in the AllMnO.SMg alloy after annealing at low temperature, naturally raise suspicion of a Zener-drag effect, i.e. precipitation occurring concurrently to recrystallization. Such reactions are sometimes unavoidable in industrial alloys when batch annealing treatments are performed using a slow heating rate (typically 50C h-l). In order to quantify the amount of precipitation taking place during annealing, the electrical conductivity has been measured and the results are shown in Fig. 12(a). To indicate the kinetics of the recrystallization reaction, hardness measurements have also been included in Fig. 12(a). In the as-cold rolled condition the AllMnO.SMg material is seen to have a conductivity of approximately 23.8 m 0-l mmm2, and after
for the different

grain diameter D(pm) and largest grain diameter &,(rm) texture components in fully recrystallized samples AllMnlMg alloy (330C) All Mn0.5Mg alloy (300C)

AlIMnOSMg alloy (450C)

D
ND rotated P Cube Random cube 21 23

D nlar
63 53

D
21 23 18 8

D max
61 54 41 40

D
9 8 8

DIM
18 14 19

1424

DAALAND and NE.9

RECRYSTALLIZATION TEXTURE

(cl (4 0 ) Fig. 11. (111) EBSP pole figures showing the texture in fully recrystallized samples of (a) AllMnlMg alloy annealed at 330C; (b) AllMnOSMg alloy annealed at 300C; and (c) AllMnOSMg alloy annealed at 450C. annealing for 1000 s at 300C an increase of 1.0 m R- mmT2 was observed. A large amount of precipitation is therefore expected to occur at the early stages of the transformation which presumably will affect the growth of recrystallization grains by allowing only a selected proportion of nuclei to grow. During further annealing the electrical conductivity continues to increase and seems to reach a saturation value of 25.6 ma- mmm2 when the material is fully recrystallized. It is interesting to note that a similar behaviour is observed during annealing at 35OC, i.e. a rapid increase in conductivity during recrystallization of the material, reaching a saturation when the material is in the fully recrystallized condition. However, the amount of precipitation is weaker at this temperature, the result being a somewhat finer average grain size. During annealing at 450C no increase in conductivity is observed during the recrystallization period. For comparison, conductivity measurements in the case of the AllMnlMg alloy [Fig. 12(b)] show a very small increase in the conductivity during the recrystallization period. This indicates that no significant precipitation is taking place in this alloy. 3.4. Grain size distributions From the results presented in the previous section, the size distribution of grain section areas of recrystallized grain has been derived, for the fully recrystallized samples. The results are shown in Fig. 13 where the frequency of occurrence is plotted against A /A on a logarithmic scale. Note that A is the grain section area and 2 the corresponding geometrical mean value. Looking at the AllMnlMg alloy [Fig. 13(a)] it is interesting to observe that the size distribution is close to being of a log normal type. A similar result is obtained in the case of the AllMnO.SMg alloy when annealing is performed at 450C [Fig. 13(c)]. However, annealing the AllMnO.SMg alloy at 300C results in a size distribution which is extremely broad [Fig. 13(b)], reflecting the different growth rates for grains of the different orientations. The grains are seen to be fairly evenly distributed over a wide range of sizes, and the maximum grain size is 334 times larger than the average size. 4. DISCUSSION Fig. 12. Electrical conductivity and hardness as a function of annealing time for (a) AllMnOSMg alloy annealed at 300, 350 and 450C; and (b) AllMnlMg alloy annealed at 330C. 4.1. The formation
of recrystallization textures

Taking a global look, i.e. by looking at the rolled material as a whole, the number of activated sites (in

DAALAND

and NES:

RECRYSTALLIZATION

TEXTURE

1425

-4

-3

-2

-1 I"(Pk,

(4

Ail Mn0.5Mg alloy

2 5 3

0.3

0.2

s lg 0.1 IJ. 0.0 .b -3 -2 -1 0 1 2 3 4

In(NAm) @)

ifi d
i Y =

All Mn0.5Mg
0.4

alloy

0.3

generation of new cubes. It follows that subgrains belonging to the cube orientation have one to two orders of magnitude more potential as nucleation sites for recrystallization compared to other matrix subgrains (PSN excluded). Still the probability for a given cube subgrain to develop into a new grain is very low. 4.1.1. Nucleation of the cube. After final hot rolling of the AllMnlMg alloy (as-deformed condition) the microstructure contains cube oriented subgrains distributed as continuous sheets parallel to the rolling plane and positioned in-between stable rolling texture components, for more details see Ref. [l]. It was documented in Ref. [l] that these cube bands serve as preferred sites for nucleation of cube texture on annealing. The crucial question now is: what makes these flattened old cube grains special compared to similar shaped grains of other orientations? In order to answer this question it seems appropriate to start out by reiterating the following important structural aspects (from Ref. [l]). In the as-hot rolled structure (final gauge) the mean subgrain size of cube and matrix subgrains were found to be 1.5 and 1.0 pm, respectively, with the cube subgrain distribution having a much larger tail towards larger subgrain sizes. The extraordinary potential for nucleation from cube bands (see sketch in Fig. 14) as compared to nucleation from other sites can be illustrated as follows. The critical diameter 1, for a subgrain in the cube band to bow out and start growing into the surrounding matrix, is obtained by combining the GibbsThompson nucleation criterion, I, = 4yo,/Pn with the relation for the stored energy in the matrix substructure, PO = c(( ysB/6, ): 1 C
uYSB

t; 5 0.2 8 w E 0.0 -4 -3 -2 -1 0 1 2 3 4

(4)

0.1

In(A/Am)

(4
Fig. 13. Experimental size distributions of recrystallized grains: (a) AllMnlMg alloy annealed at 330C; (b) AllMnOSMg alloy annealed at 300C; (c) AllMnOSMg alloy annealed at 450C.

where ycB is the grain boundary energy (of the nucleus), CIis a geometric constant, 6,, the average matrix subgrain size and ysa is the energy of a sub-boundary. The latter can be estimated from the Read-Schockley relation: ysB = (BGb/4a (1 - v)) In e0 /0,, giving: (5) where 13is the sub-boundary misorientation and Bc the maximum value of this, and v is Poisson s ratio. By substituting typical values for heavily rolled aluminium (a = 3, 8 = 3 and Bc = 15) into equation (5) this gives: 1, = 2.26,. It follows that the critical recrystallization nucleus in the present material has a size of 2-3 pm. Looking at the distribution data for cube subgrains [given in Fig. 14(b) in [l]], it can be seen that the largest subgrains inside the cube band are typically of this size or even larger. Thus, cube subgrains of critical size are present in the asdeformed condition, and there is no need for further subgrain growth or coalescence in order to produce

the range 10 3-1014m~3) is very small compared to available sites (total number of subgrains in the range 10 8-1019 m- ). That is, only one subgrain out of 104-lo5 succeeds in nucleating a new grain. In terms of the nucleation of cube grains it is possible to be a little more specific. If it is assumed that the cube grains are nucleated from banded features which are spaced approximately 30 pm apart, then the site density in cube bands is approximately lo9 m- which is generated from a subgrain density exposed to the surrounding matrix of 10 m-*. In other words, 1 out of 1000 subgrains in the band are potential sites for

1426

DAALAND

and NES:

RECRYSTALLIZATION

TEXTURE

ORIENTATION

HIGH-ANGLE

GRAIN -

Fig. 14. Schematic cube band configuration.

a viable cube nucleus. In other words, incubation time for the cubes. Orientation Aspects

there

is no

of Cube Grain Nucleation

As reported in Ref. [1], all components of the fi-fibre could be found as neighbour orientations to the cube bands. Accordingly some of the subgrains within most cube bands are likely to satisfy a 40( 111) orientation relationship with respect to the neighbour orientations. A recent investigation on cold rolled copper by Duggan et al. [lo] has demonstrated that the material adjacent to the cube nucleus frequently has an orientation which can be related by a -4O( 111) relationship. These workers concluded that the cube texture arises from the competition between nuclei in a variety of orientation environments, the successful grains having environments into which they initially grow characterised by a 40( 111) relationship. They called this process a microgrowth selection effect, in which the local texture determines the viability and competitive advantage of nuclei through a mobility advantage of certain boundaries. Looking at the pole figures in Fig. 12 in Ref. [1], which show examples of the lattice rotation associated with EBSP-scans across different cube bands, it can be seen that the orientation just above or below the band sometimes has a (111) pole more or less in common with the cube orientation inside the band. An obvious question now follows: is such a favourable cube band environment a frequent observation also in the present case, and further, how important is this effect concerning determination of the final texture after annealing? Preliminary investigations [ 111 seem to confirm the Duggan et al. results also for the present case. 4.1.2. Origin of preferential growth. The Classic Oriented Growth Mechanism The oriented growth hypothesis claims that grains with a special orientation relationship to the deformed matrix will have a higher growth rate compared with grains of random orientation. In fee metals, grains with a -4O( 111) orientation relationship seem to be of this special character. According

to this mechanism the cube orientation (in the case of the AllMnlMg alloy) should have a fast growing 40( 111) orientation relationship with respect to the S orientation. The ND rotated cube and the P orientation (in the case of the AllMnOSMg alloy) may be interpreted in a similar manner since they both exhibit a 40( 111) orientation relationship with respect to the Cu orientation. However, several arguments may be raised against such a growth mechanism. Usually it is found that only one out of four (111) poles are rotation axes, and only one rotation direction is found for each. As an example, looking at the rotations necessary to produce the cube orientation from the (123 }(412) or S component, it is seen that only one of the poles of each of the four {123}(412) components may act as a rotation axis (with one sense only) to form the cube orientation. If growth selection were made from randomly oriented sites, then eight variants would be produced. In response to this variant selection problem, the concept of compromise texture has been introduced. The cube orientation has, for instance, been claimed to have a good compromise character in the sense that it is highly symmetrical with respect to several rolling texture components and so can grow more readily. However, large variations are found in the rotational relationship characterised by -4O( 111). Assuming an oriented growth mechanism, this implies that grains of significantly different orientation can grow at rates comparable with those of grains having the exact orientation relationship. However, deviations from the preferred rotational relationship, 40(111), have been shown by Liebmann et al. [12] to cause substantial changes in mobility compared to the ideal case. The simple concept of an optimal over-all growth of the 40( 111) components fails to explain some important details of the transformation curves in Fig. 2. In the AllMnlMg alloy case [Fig. 2(a)] the main component which develops during recrystallization is the cube orientation. However, this component evolves at the expense of all the main rolling texture components Cu, S and Bs. Although the 40(111) relationship with respect to the cube is fulfilled best

DAALAND

and NES:

RECRYSTALLIZATION

TEXTURE

1427

for the S orientation, this component does not seem to be reduced preferentially. In the fully recrystallized state the cube is found to coexist with the retained rolling or R texture which is very close to the S orientation of the rolling texture. However, the S orientation is the one which is supposedly consumed by growing cube grains. These observations certainly cast doubt on the growth selection as the main cause for cube texture development. A similar failure of the oriented growth theory is observed in the case of the AllMnOSMg alloy [Fig. 2(b)]. During recrystallization of this alloy the development of the ND rotated cube and P orientation is accompanied by a decrease in both Cu and S rolling texture, although a 40( 111) orientation relationship with respect to these components is found only for the Cu component. In the as-rolled state the Cu and S orientations are about equal in volume, constituting 30 and 27%, respectively. Based on oriented growth it is strange that the cube orientation, which has a good 40(111) relationship to the S orientation, develops only in the later stages of recrystallization and constitutes only 5% of the total volume in the fully transformed structure. An interesting observation with regard to the ND rotated cube components is observed when inspecting the (111) pole figure presented in Fig. 3(b). During annealing, the two ND rotated cube components are seen not to evolve in a similar manner. The (001) [310] component develops much more strongly than the symmetrically equivalent (001)[3io] component. Such a preference is difficult to explain using oriented growth. On the Transformation
Potential of 40C 11 I) Grains

AllMnIMg Alloy. Regarding this alloy, the experimental results (Fig. 7) clearly indicate that there is no significant growth selection in this case. A plot of the initial stage in the growth of cube is shown in more detail in Fig. 15(a). In this figure the size of the largest recrystallized cube grain and largest random grain have been plotted as a function of annealing time. Additionally, in Fig. 15(a) the largest cube band subgrain (taken from the investigation of the same alloy in Ref. [1]) has been plotted. It is interesting to note that the curve of the largest recrystallized cube grain extrapolates down to the values of the largest cube band subgrain in the as-deformed condition. A certain discrepancy in the straight line behaviour is observed. However, it is to be noted that due to rather poor statistics (largest subgrain from a total distribution of 50 subgrains only) these subgrain (or new grain) sizes are probably underestimated. Based on the subgrain-grain size measurements shown in Fig. 15(a) it seems clear that there is no growth selection in this case. The largest cube grain and the largest random grain seem to have equal growth rates throughout the transformation. If

55 5. p 1 45 _ _

?? LARGEST CUBE BAND SUEGRAIN

0 LARGEST CUBE GRAIN * LARGEST RANDOM GRAIN

40 -

There have been few attempts in the past to directly measure and compare the growth rate of 40( 111) components/grains to that of grains belonging to other texture components, in order to test a possible oriented growth effect. To provide some further insight into the complicated texture transition during recrystallization, a systematic effort has been made in the two alloys, trying to measure the growth rate of different grain orientation. By applying the EBSPtechnique for identification of grain orientations, the size of recrystallized grains were measured at different stages during annealing. For each annealing condition the grains were assigned to different texture groups, viz. cube, ND rotated cube, P or random (see Figs 7 and 8). Before discussing the results of the grain size measurements in detail, it must be emphasised that it was shown (using a stereological procedure, Section 3.2) that the nucleation in both alloys is best described by a site saturation situation (all nuclei are created at t = 0). This means that nucleation is completed within a time period which is much shorter than that required for complete transformation. It may therefore be stated that no ND rotated cube, P or cube oriented grains are nucleated at an earlier time than other orientations.

5 -- . ... _I 0 0 ,; I 20 8 8 a 8 I # 8 f t f 0 t 0 0 40 60 80 100 120 140 160 180 200 ANNEALING TIME [set] (a]

LN

ANNEALING (W

TIME [set]

Fig. 15. (a) The growth of cube and random oriented grains as observed in the AllMnlMg alloy on annealing after final hot rolling. (b) Schematic illustration of the growth of grains with different orientation, during annealing after cold rolling of the AllMnOSMg alloy.

1428

DAALAND and NES:

RECRYSTALLIZATION

TEXTURE

a growth selection was present, this should have been revealed through a faster increase in the size of cubes compared to random grains. This is not observed, and neither does there seem to be any micro-growth selection. A possible micro-growth selection effect of the cube oriented grains (as described by Duggan et al. [lo]) in the present case should have been visible in Fig. 15(a). Thus growth selection is excluded, and the preference of cube must be explained at the nucleation stage. The same conclusion is found in two recent papers by Hjelen et al. [13] and Doherty et al. [14], who performed extensive investigations of the growth of cube grains during recrystallization of aluminium after cold rolling and hot extrusion, respectively. As shown in Section 4.1.1 there is no incubation time for cube since subgrains of critical size (or larger) are present in the as-deformed condition. This behaviour contrasts with the nucleation and growth of other orientations. The non-cube grains (basically nucleated at particles), which invariably develop from much smaller cells, require an incubation time, i.e. subgrain growth and recrystallization of the deformation zone before reaching the critical size of 2-3 pm in diameter. In order to explain the observed size difference of 5-10 pm between cube and random oriented grains, it now follows that this mainly reflects the size advantage the cube grains carry with them from the as-deformed condition, see Fig. 15(a). Additionally, the particle sites are also expected to be more clustered as compared to the cube sites leading to earlier impingement of the growing PSN-grains. Despite the lack of a growth selection, there is, however, an element of 40( 111) orientation relationship, as shown by Vatne et al. [ll]. Investigating the same material by means of detailed EBSP-investigations, it was shown that the probability of the nucleation and growth of cube oriented recrystallized grains from cube bands is higher when the band is surrounded by the S deformation texture component to which cube makes a 40( 111) boundary. For further discussions see Refs [ll, 15, 161. AIlMnO.5Mg Alloy. In the AllMnO.SMg alloy case an obvious interpretation, when considering only the transformation curves in Fig. 8, seems to be a strong oriented growth effect. However, taking into account all the experimental observations in the present study, such a boundary mobility interpretation does not appear convincing. A main deficiency is the fact that the higher mobility of the ND rotated cube and P grains seems to be operative only in the early stages of recrystallization. As shown in Fig. 8(c) for the evolution of the largest grain, the 40( 111) components are growing faster only in the first 20-30% of recrystallization. In a classical oriented growth situation the 40( 111) component should have continued to grow at a higher rate than the random component, during the entire transformation, as long as the preferred Cu orientation still remains as part of the deformation texture. However, this is not

observed. The second striking result in disfavour of a true oriented growth effect is connected to the microtexture observations reported in Section 3.3. As oriented growth depends on an orientation relationship between the deformed matrix and the recrystallized grain, it is to be expected that oriented growth will be most pronounced if the deformed matrix has a unique orientation. It will be least effective if the matrix orientations are diffuse. EBSP-investigation of the as-cold deformed AllMnO.SMg material revealed that the whole /I-fibre was present within a sheet of cold deformed metal which was less than 5 pm thick. The reason for this highly deformed structure of course is the rather brutal treatment experienced during hot rolling (no recrystallization), and subsequently heavy cold rolling. It follows that during subsequent annealing of this complicated deformation structure both the ND rotated cube and the P grains, almost from the nucleation stage on, were exposed to a deformation matrix which in orientation terms covered the total /I-fibre. Further, out of this /I-fibre the Cu component contributed only -30%. In such a situation the basic concept of oriented growth becomes difficult to apply, even when introducing the idea of compromise texture. A final point which also disfavours a classical oriented growth interpretation in the case of the AllMnO.SMg alloy is connected to the observation of a much weaker recrystallization texture when the material is annealed at a higher temperature. Annealing performed at 450C resulted in a microstructure containing much smaller grains, in which the average size of ND rotated cube and P grains were clearly no larger than grains of other orientations (see Table 3). Note that in a classical oriented growth situation the relative size difference between grains of different orientation found at low temperature annealing, should have been found also after high temperature annealing. Injuence of Concurrent Precipitation

From the above discussion, it follows that the strong growth selection effect found in the AllMnO.SMg alloy during low temperature annealing must be due to other reasons than a classical oriented growth effect. It is known that the AllMnO.SMg cast ingot was hot rolled without a prior homogenisation treatment, leading to a retainment of a large amount of elements in solid solution before the start of hot rolling. In addition, the hot rolling was performed with a relatively rapidly falling temperature. Subsequent cold rolling and annealing at a low temperature resulted in precipitation of Fe-, Mn-bearing phases occurring concomitantly with recrystallization, as reported in Section 3.3. It is well established that the recrystallization behaviour of a highly supersaturated alloy may be drastically altered if decomposition reactions occur simultaneously with recrystallization [ 17, 181. The precipitation reaction has a profound influence on both the recrystallization kinetics and the final grain size. The very low JMAK-

DAALAND

and NES:

RECRYSTALLIZATION

TEXTURE

1429

exponent (n = 1.6), which was obtained for the AllMnOSMg alloy, can therefore be accounted for by the concurrent precipitation reaction. The basic principles concerning the competition between recrystallization and precipitation have been described by Hornbogen [19] and Kiister [20]. In general, the higher the annealing temperature the sooner recrystallization will start. Precipitation reactions, however, will follow the typical C-curve type kinetics. Therefore, at high temperature, recrystallization can precede precipitation without interference from precipitating second phases. Given these conditions, recrystallized grain size can be small, given a large number of potential nucleation sites. It is expected that PSN will be the favourable nucleation mechanisms under these circumstances. Such an interpretation fits well with the observations in the case of the AllMnO.SMg alloy annealed at 450C. Annealing at low temperature, however, results in precipitation concurrent with recrystallization which can affect both the nucleation and growth stage of recrystallization. The conductivity measurements [Fig. 12(a)] confirm such a precipitation reaction when annealing at 300C (or 350C). The obvious question then is: why does this precipitation reaction only affect the random oriented grains (i.e. the PSN activity) and not the growth of the other components? All components are probably affected by the precipitates, but the PSN grains are the ones most severely affected. Based on the detailed characterisation of cube nucleation sites in the AllMnlMg alloy a reasonable hypothesis in the case of the AllMnOSMg alloy may be stated as follows:

illustrated as shown in Fig. 15(b). As indicated in this figure, the nuclei or grains of ND rotated cube, P or cube orientation, which have reached the critical Gibbs-Thompson diameter, starts to grow at time t = 0 simultaneously with the start of precipitation. Stated in other words, nucleation from the banded heterogeneities will be less affected by precipitates since the viable nuclei are present at the onset of precipitation. Assuming, furthermore, that these nuclei can easily expand during early stages of recrystallization (by a bulging mechanism similar to the cube band nucleation), the mobile boundary created will be able to grow although it will probably be hindered by the precipitates ahead of the boundary. This behaviour contrasts with the nucleation and growth of the PSN-grains. As shown in Fig. 15(b) the PSN-grains, which invariably develop from much smaller cells, require an incubation time rrsN, i.e. subgrain growth and recrystallization of the deformation zone, before reaching the critical GibbsThompson diameter of size 2.2 .&, . As shown by Humphreys [21] the critical stage of PSN is not the growth within the zone but the growth out of it. During the time required to consume the deformation zone (&.,) sufficient precipitation will take place. Hence, the grain will be exposed to a Zener-drag when it starts growing out of the deformation zone. In such a situation the critical diameter for nucleation at a particle will be given by:

a,=

=VT (;qc >

2Yo* Pz)

(9 Nucleation and growth of recrystallization in the

AllMnO.SMg alloy is comparable with the recrystallization reaction in the case of the AllMnlMg alloy, i.e. the texture evolution during recrystallization can be described as a competition between PSN and nucleation from banded features within the deformation microstructure. The occurrence of such a banded feature in the case of the AllMnO.SMg alloy (with a cube or slightly rotated cube orientation) was reported in Section 3.2 (Fig. 6). It must also be emphasised that small peaks at the ND rotated cube and P position were found in the ODF of the as-cold rolled AllMnO.SMg alloy [Fig. l(b), ND rotated cube position shown with arrows], an observation which proves that these orientations are indeed present in significant proportions in the as-deformed condition. (ii) Nuclei of size equal to (or larger than) the Gibbs-Thompson diameter are expected to be present inside these band -heterogeneities of ND-rotated cube, P and cube orientation, in the as-deformed condition. The growth of grains with different orientations in a situation as described above may be schematically

where yGB is the grain boundary energy, PE is the driving pressure for recrystallization due to the stored energy in the matrix, and Pz is the Zener pressure due to the drag from small dispersoid particles. From this relation it is seen that a reduction in the effective driving force, due to a Zener drag effect, will increase the critical nucleus size necessary for successful nucleation to occur. It follows that the precipitation reaction may severely affect the formation of the PSN-grains.
A Model for the Infiuence of Precipitates PSN-activity on the

In the following a first-approach model for the influence of precipitates on the PSN-activity will be introduced. In explaining the precipitation-effect, two contributions of the Zener drag need to be considered: (i) Zener-drag from dispersoids precipitated during homogenisation/preheating and hot rolling; and (ii) Zener-drag from small particles precipitated during (low temperature or batch) annealing after cold rolling. An estimate of the Zener drag from dispersoids is given by:

1430

DAALAND and NES: RECRYSTALLIZATION TEXTURE shown in Fig. 16 correspond to drag expressions of the form: Pg(Z) = 27my,,r2cr/6, and P$(II) = my,,r, where n is the area density of precipitates on the sub-boundaries. Inserting typical values, r = 0.01 pm and n = lOI m- [22], &, = 1 pm and LY = 3, leads to drags of the order P!(Z) - 2 x 10syos and P$(ZZ) - 3 x 106y,, The true value is expected to be somewhere between these limits, probably closest to Pg(II) due to the high flexibility of the boundary. This means that the Zener-drag due to precipitates is considerably higher than that caused by the dispersoids and thus the total retarding pressure is expected to be of the order of Pz - 106yoB.The driving pressure for recrystallization, i.e. the stored deformation energy, is estimated to be of the order of PE = 106yGe, which is then of the same size as Pz. This means that unless a new PSN-grain has consumed the deformation zone before impingement on the precipitated phase, growth beyond the deformation zone will be very difficult. A large amount of the PSN-nuclei is therefore expected to have great difficulty in growing out of the deformation zone.
Deformation / zone

wherefis the volume fraction and r the radius of the dispersoids. According to Strid [22], typical values are f= 0.001 and r = 0.05 pm which give a retarding pressure of the order Pi - 104yo,. It is to be noted that the dispersoids are expected to be randomly distributed. The small particles precipitated during annealing, on the other hand, are expected to nucleate preferentially on the subgrain structure. In order to give an estimate of the Zener-drag pressure from these precipitates it is necessary to estimate the number of particles in contact with a moving grain boundary. Two limiting cases, which are schematically illustrated in Fig. 16, will be discussed. Case I-the moving grain boundary is macroscopically flat and the precipitates interact with this boundary as if they were randomly distributed in space. Case II-the moving grain boundary is flexible and adapts a jagged shape due to the interaction with the precipitates on the substructure. The Zener-drag pressure due to the precipitates is given by P$ = nNy,,r, where N is the area density on the moving grain boundary and r the radius of the precipitates. The two limiting cases

t=O

Grain boundary-

Dispersoids

Fig. 16. Particle stimulated nucleation, schematic illustration of the grain boundary migration during precipitation. Case I-the grain boundary is macroscopically flat. Case II-the grain boundary is flexible.

DAALAND

and NES:

RECRYSTALLIZATION

TEXTURE

1431

Although the number of nuclei having a ND rotated cube, P or cube orientation is much less than the number of random oriented/PSN-nuclei (c.f. Table 2) they will become active sites earlier in the transformation. Under these circumstances the PSNgrains will be outgrown by the more rapidly growing grains of ND-rotated cube, P and cube orientation, a situation which is consistent with the experimental observation of island grains (Fig. 10). It is interesting to note that there are large random oriented grains in the structure [looking at the evolution of the largest grain, Fig. 8(c)], indicating that random oriented nuclei are able to grow if their initial size is large enough. However, although random oriented grains as large as 40 pm were found, they usually showed very little evidence of substantial growth, and their contribution to the final texture is therefore small. It is reasonable to expect that grains of all orientations will be hindered by the precipitates if they have not reached the critical size. This gives a satisfactory explanation to the reduction in the number of nucleation sites (or grains) which was observed for all components early in the transformation, see Table 2. Accordingly it is concluded that the more rapid growth of the ND rotated cube, P and cube grains at low temperature annealing of the AllMnOSMg alloy is due to a precipitation-induced effect. The higher growth rate of the 40(111) components reflects that these grains have a much shorter incubation time for nucleation than the random category, and accordingly become less affected by the concurrent precipitation occurring at low annealing temperatures. 4.2. Growth of recrystallization, lation approach
a computer simu-

To further the understanding of recrystallization, simulations of the growth reaction following nucleation have been performed using a Cray supercomputer. The main purpose of these simulations is to investigate the effect of grains growing at different rates, on the structure and texture development. The computer simulation procedure applied in this work was originally developed by Mahin et al. [23], but has later been refined to include more general growth rate and nucleation models by Sztre et al. [24], Marthinsen et al. [25] and Furu et al. [26]. It is a three-dimensional model where nuclei distributed freely in space, i.e. in a cube with periodic boundary conditions, start to grow according to a given nucleation model. The nucleated grains grow as given by a specified growth law and the transformation is complete when the grains impinge on one another. The resulting microstructure is analysed in a two-dimensional section through the cube. It is a powerful model which with a modem supercomputer may handle more than lo4 nucleation sites, allowing the kinetics as well as the structural aspect of the reaction, i.e. the size distribution of recrystallized grains, to be followed.

4.2.1. Growth rate modifications. In the present work the model has been modified by introducing a non-uniformity with respect to how the grains grow, allowing the textural aspect of recrystallization to be studied. The motivation for this modification is based on the experimental results obtained for the AllMnO.SMg alloy in Section 3.3, in which different categories of grains (i.e. different crystallographic orientations) were found to grow at significantly different rates. Based on these observations the nuclei for recrystallization in the simulation procedure, were distributed between four different classes according to the nuclei/grain-density numbers calculated for each classified texture component (given in Table 2 in Section 3.3). As a first approach (Case I), a specific growth rate was then assigned to each class, corresponding to the difference in initial growth rate observed experimentally, i.e. from studying the growth of the largest grain, Fig. 8(c). The grains were subsequently allowed to grow isotropically at constant rates until impingement. In addition to using constant but different growth rates for each class, simulations have also been performed in which each class was associated with a decreasing growth rate different from the others (Case IIA). Such a decrease in growth rate is highly realistic, and may for instance be caused by a concurrent precipitation reaction, as shown in the case of the AllMnOSMg alloy in the present work. A fall in growth rate G(t) was obtained by modifying the numerical code as described by Satre et al. [24], and using the following formula: G(t) = &/(a + t), where t is annealing time. By selecting different values for a and I,,, the shape of the decreasing growth rate curve can be varied, for more details see Refs [24,25]. Three components were given roughly the same growth rate variation, while one (the random component) was assumed to fall off more rapidly. A further refinement of the simulation procedure (Case IIB) has been included by taking into account the experimentally observed drop in the number of grains (or nuclei) during transformation of the AllMnO.SMg alloy. After approximately 10% total transformation a drastic reduction in the number of nuclei was observed for all the texture components. Using the same condition for growth as in Case IIA, this effect was achieved in the simulation by physically stopping 2/3 of the grains of each class after a total transformation of 10%. It is to be noted that all simulations were performed using site saturation nucleation kinetics (i.e. all nuclei activated at time, t = 0), and the nuclei were randomly distributed in space. 4.2.2. Simulation results. An example of the microstructure evolution during simulation (Case I) is shown in Fig. 17. The evolution of the average grain diameter as a function of the total area fraction transformed, as obtained by experiments and by computer simulation, is shown in Fig. 18. In order to compare the experimental and the simulated results, the fastest growing component in the simulation is

1432

DAALAND

and NES:

RECRYSTALLIZATION

TEXTURE

(b)

Fig. 17. Time-lapse sequence showing the development of the microstructure in simulation Case I, as seen on a planar section through the transforming body: (a) 24% transformed; (b) 48% transformed; (c) 80% transformed; (d) 100% transformed. Note the occurrence of island grains. adjusted in such a way as to result in 10% recrystallization, in accordance with the experimental observations. The simulated and experimental results in terms of texture (volume fractions) and average grain sizes in the fully recrystallized state are listed in Table 4. In simulation Case I, the grain size is seen to develop rapidly in the initial stage and the growth rate then falls off quickly due to impingement [Fig. 18(a)]. A general underestimation of the experimental results is obtained in this case. Somewhat larger grain sizes are predicted in the simulation Case IIA [Fig. 18(b)], however, the average grain size for the different classes are still too small compared with the experimental observations as shown in Table 4. A reasonably good fit with experimental results is observed in the last simulation approach, Case IIB [Figs 18(c) and (d)]. Except for the slowest growing random component, the experimental and simulated results for the other components are well in line. It is interesting to note that although quite large variations in the input parameters are used in the simulations (Cases I and II), the effect on the evolution of the average grain size is much less dramatic. The analysis used for determination of the nuclei/grain-density in Section 3.3 was based on the measured grain size and volume fraction recrystallized, and by assuming isotropical (i.e. spherical) growth of the grains. In order to check the validity of such a simple analysis, it was applied to the simulation Case IIA. By investigating selected two-dimensional planes or sections through the computer generated sample/cube, it was shown that the nuclei/ grain-density remained approximately constant throughout the entire transformation. The simulation therefore indicates that the procedure applied in the experimental investigation is a reasonable method for determination of the nuclei/grain-density. In addition, based on this observation it follows that the drastic drop in the density observed experimentally in the case of AllMnOSMg alloy is a realistic one, a phenomenon attributed to precipitation in Section 4.1.2. The grain size distributions determined by the computer simulations are shown in Fig. 19. Using constant but different growth rates (Case I) results in a significant broadening of the size distribution, Fig. 19(a), as compared with simulation using only one component. Note that when the size distribution of each separate class is plotted, site saturation distributions are obtained. When different classes are combined in a simulation their individual site saturation peaks will be superimposed resulting in several peaks in the size distribution. As can be seen in Fig. 19, distinct peaks in the distribution can be observed reflecting the different growth rates of the different classes, which accordingly should give a smoother

DAALAND
-g 30 3 ozp5 _ CASEI b_ w

and NES:

RECRYSTALLIZATION

TEXTURE

1433

CASE

IIA

Random

0.4 w (a)

0.6

0.6

1 .o

0.0

0.2

0.4 w (b)

0.6

0.6

1.0

0.4 XV) @)

0.6

0.6

1 .o

0.0

0.2

0.4 X0) (d)

0.6

0.6

1 .o

Fig. 18. Simulated transformation curves showing the evolution of the average grain size as a function of annealing time. (a) Case I; (b) Case IIA; (c) Case IIB; and (d) Case IIB+ experimental results, AllMnOSMg alloy annealed at 300C.

distribution curve. The use of decreasing growth rates further broadens the distribution, as shown in Figs 19(b) and (c). However, compared with the experimental results [Fig. 13(b)] all the simulated distributions are somewhat narrower. This indicates that the recrystallization process is more complex than assumed by the simulation model. It is noted, however, that Case IIB comes close to the experimental observations. It is expected that the experimental situation corresponds to a dispersion of growth rates within each class, instead of four distinct growth rates used in the present simulations. While such refinements will have consequences for the detailed character of the simulated size distribution they will not affect the general multipeak broadening which is well captured by the Case IIB simulation, compare Figs 13(b) and 19(c). For a detailed discussion of the
Table 4. Simulated and experimental

differences between simulated and experimental size distributions, see Refs [24,25]. The simulated results for the fraction of recrystallized material as a function of annealing time, are shown in Fig. 20. The results have been plotted in the classical way for a JMAK-interpretation, i.e. log ln[l/l - X(t)] versus log t. For the simulation Case I, a least square fit to a straight line yields a slope of 3 in accordance with the original site saturation case. This is not surprising since the simulation in Case I in kinetic terms is equivalent to simulations using only one component. However, when decreasing growth rates are included in the simulation the n-value is reduced to 2.5 and 1.8 for Cases IIA and IIB, respectively. The latter value, which takes into account the concurrent precipitation reaction in the early stages of recrystallization, is in reasonable
alloy, 3OOC), fully transformed state

results (AllMn0.5Mg

Experiments
X-rav

Computer simulation Case I

46 26 21 7 19 16 15 6

EBSP
48 26 17 9 21 23 18 9

Case IIA
49 29 15 I 22 18 15 6

Case IIB
49 21 19 6 21 22 17 6

Texture
W)

Grain (rem)

size

ND rot. cube P Cube Random ND rot. cube P Cube Random

42 18 4 -

1434

DAALAND

and NES:

RECRYSTALLIZATION

TEXTURE

agreement with the experimentally observed value of 1.6. It can be concluded that the computer model which has been applied in the present work is not fully capable of simulating the recrystallization process in the case of the AllMnOSMg alloy. However, computer simulation using four components improves the simulation prediction, compared to a more simple modelling approach in which all grains are considered to behave alike (i.e. using only one growth rate). A size distribution being much broader in combination with a JMAK-exponent (Case IIB) being close to the experimentally observed value, indicates that the model has important elements in it which are closer to a correct description of the recrystallization reaction.

-2.5

-3.51I
-4.0

/
I -3.8 *

-3.6

I log

I
-3.4

-3.2

I
-3.0

-2.8

time (arb. units)

Fig. 20. Transformation kinetics (JMAK-plot) for the cases investigated by means of computer simulation.

5. CONCLUSIONS

0.3

A detailed study of the development of recrystallization textures has been carried out, following in detail the evolution of grains with a 40( 111) orientation relationship to components within the rolling texture distribution.

6 $j 0.2
2 & 0.1 0.0 -4 -3 -2 -1 In(Gull) 1 2 3 4

(1) Annealing of an AllMnlMg

(a)

-4

-3

-2

-1

0 In(NAm) 0 )

5
p 6 5 s g u 0.0 -4 -3 -2 -1 In(ZAm) 0.3 0.2 0.1

(4
Fig. 19, Grain size distributions obtained from simulation (a) Case I; (b) Case IIA; (c) Case IIB.

alloy after final hot rolling provided clear and direct evidence for the central role of oriented nucleation and for the negligible role of oriented growth in the development of cube texture. No enhanced growth rate of the 40( 111) related cube ({OOl}( 100)) grains, nucleated systematically within cube band sites, was found as compared to the random oriented grains developed by particle stimulated nucleation (PSN). The small size difference observed between cube and random oriented grains is attributed to the fact that cube oriented subgrains (or grains) of critical size are present within the cube bands in the as-deformed condition. (2) Low temperature annealing of an AllMnO.SMg alloy after cold rolling resulted in a strong oriented growth effect in the sense that grains of the 40(111) components, ND rotated cube ({OOl}( 130)) and P ({ 1 lO}( 11 l)), grew significantly faster than the randomly oriented grains. Due to a concurrent precipitation reaction occurring at this temperature the growth selection observed is interpreted as a precipitation induced effect, reflecting that particle stimulated nucleation (responsible for the random oriented grains) is more sensitive to precipitation than nucleation from the competing 40(111> sites. (3) In order to illustrate the effect of grains growing at different rates on the structure and texture development, a computer simulation procedure has been applied for simulation of the recrystallization reaction. By modifying the simulation model according to the experimental observations in the case of the AllMnOSMg alloy, the kinetics of the reaction as well as the final grain size distribution and texture (volume fractions) can be predicted in a reasonable way.

DAALAND

and NES:

RECRYSTALLIZATION
10.

TEXTURE

1435

Acknowledgements-The authors would like to acknowledge the financial support provided by Hydro Aluminium, Pechiney and the French Norwegian Foundation (FNS). The valuable discussions with members of the project committee, Drs G. M. Raynaud, P. Lequeu, R. Shahani at Pechiney, Drs J. Strid and S. E. Nzss at Hydro Aluminium and Professor J. Driver and Dr C. Maurice at the Ecole des Mines, are greatly appreciated. Moreover, thanks are due to Dr Dorte Juul Jensen at Riss National Laboratory, Denmark, for performing the neutron texture measurements and to Dr Olaf Engler at the RWTH Aachen, Germany, for the X-ray texture measurements.

B. J. Duggan, K. Liicke, G. Kijhlhoff and C. S. Lee,

Acta. metall. mater. 42, 1921 (1993).

11. H. E. Vatne, 0. Daaland and E. Nes, Mater. Sci. Forum

157-162, 57 (1956). 1087 (1994).

12. B. Liebmann, K. Liicke and G. Masing, Z. Metallk. 47, 13. J. Hjelen, R. 0rsund and E. Nes, Acta. metall. 39, 1377
(1991).

14. R. D. Doherty, K. Kashyap and S. Panchanadeeswaran, Acia. metall. mater. 41, 3029 (1993). 15. I. Samajdar and R. D. Doherty, Scripta metall. mater.
32, 845 (1995).

REFERENCES 1. 0. Daaland and E. Nes, Acta. mater. 44, 1389 (1995). 2. K. Liicke, J. Pospiech, K,-H. Virnih and J. Jura, Acta. metall. 29, 167 (1981). 3. K. Liicke, J. Pospiech, J. Jura and J. Hirsch, Z. Metallk.
77, 312 (1986). 4. J. Hirsch and K. Liicke, Acta metall. 36, 2883 (1988). 5. J. E. Hilliard, in Recrystallization, Grain Growih aid Textures (Edited bv H. Maruolin). v. 267. ASM. Metal

16. H. E. Vatne, T. Furu and E. Nes, Mater. Sci. Technol., in press. 17. R. D. Doherty and J. W. Martin, J. Inst. Metals 91,332
(1962).

18. E. Nes and J. D. Embury,

(1975).

Z. Metallk.

66, 9,

589 349

19. E. Hombogen, Prakt. Metallographie, (1970). 20. U. Koster, in Recrystallization of Metallic

Materials

Park, OH. (1966)._ 6. A. M. Gokhale and R. T. DeHoff, Metall. Trans. 16A,

Quantitative Stereology. Addison Wesley, New York (1970). 8. R. A. Vandermeer and B. B. Rath, Metall. Trans. 20A. 391 (1989). 9. R. T. DeHoff, Proc. 7th Int. Symp. on Metallurgy and Materials Science (Edited by N. Hansen et al.), Riss National Laboratory p. 35. Denmark (1986). 559 (1985). 7. E. E. Underwood,

(Edited by F. Haessner), p. 215, 1st edition. Rieder-Verlag, Stuttgart (197 1). 21. F. J. Humphreys, Acta. metall. 25, 1323 (1977). 22. J. Strid, unpubhshed work, Hydro Alummium (1990). 23. K. W. Mahin. K. Hanson and J. W. Morris Jr. Acta.
metall. 28, 44j (1980).

24. T. 0. Saetre, 0. Hunderi and E. Nes, Acta. metall. 34,

981 (1986).

25. K. Marthinsen, 0. Lohne and E. Nes, Acta. metall. 37,

135 (1989).

26. T. Furu, K. Marthinsen and E. Nes, Mater. Sci. Technol. 6, 1093 (1990).