MICROSCOPY RESEARCH AND TECHNIQUE 40:101–121 (1998) Microstructure Investigations of Ball Milled Materials J.Y. HUANG, Y.K. WU, AND H.Q. YE* Laboratory of Atomic Imaging of Solids, Institute of Metal Research, Chinese Academy of Sciences, Shenyang 110015, People’s Republic of China KEY WORDS mechanical alloying; high-resolution electron microscopy; nanocrystalline; deformation twinning; solid solution ABSTRACT HREM and FEG TEM were emphasized and extensively used to follow the most subtle changes in the structure and composition of ball-milled Cu, Fe-Cu, and thermally decomposed Fe60Cu40. Some significant results are obtained and summarized as follows: HREM shows that the deformation of ball-milled copper proceeds mainly by twinning and shear bands (SBs) formation. The nano-grains formed during ball milling (BM) contain a high density of dislocations. The grain boundaries (GBs) of nanocrystalline (NC) Cu prepared by BM are ordered, curved, and strained, but disordering, lattice distortion, and nanovoids in local regions were frequently observed. Nanoscale composition analysis on mechanically alloyed Fe16Cu84 shows that the average Fe content in both the interior of grains and the GBs is close to the designed composition, which proves that a supersaturated solid solution has really formed. However, the Fe content is rather inhomogeneous between the larger and smaller grains, which infers the inhomogeneous mixing of Fe and Cu during mechanical alloying (MA). NC structure and the mechanical force-enhanced fast diffusion are the reasons of the formation of supersaturated solid solutions in immiscible systems with positive enthalpy of mixing. HREM observations carried out with the thermally decomposed Fe60Cu40 solid solution show that the Nishiyama (N-W) or Kurdyumov-Sachs (K-S) orientation relationships exist between a-Fe and Cu. Energy dispersive X-ray spectra (EDXS) results show that the Cu content in these a-Fe grains reaches as high as 9.5 at.% even after heating to 1,400°C, which is even higher than the maximum solubility of Cu in g-Fe at 1,094°C. Microsc. Res. Tech. 40:101–121, 1998. r 1998 Wiley-Liss, Inc. INTRODUCTION Mechanical alloying (MA)/ball milling (BM) (often refers to the milling of single component powders) has attracted considerable interest from the materials science community in recent years. This is mainly due to the wide range of materials exhibiting intriguing properties and structures that can be fabricated using this method. Many materials such as oxide dispersion strengthened (ODS) alloys (Benjamin, 1970), intermetallics (Morris and Morris, 1996; Nash et al., 1995), amorphous and nanocrystalline materials (Fecht, 1995; Koch et al., 1983; Schwarz and Koch, 1986; Shingu et al., 1988), and a number of nonequilibrium structures (Gente et al., 1993; Huang et al., 1994a; Ma and Atzmon, 1991; Schwarz, 1996; Shingu and Ishihara, 1995; Shingu et al., 1990) have been synthesized by MA. Although the underlying fundamentals of the MA process are increasingly well understood (Koch, 1995), many details during the mechanical treatment of solids remain unclear. For example, it is well known that repeated mechanical deformation and cold-welding often lead to the formation of nanocrystals, but the formation mechanisms and the structural characterization of the nanocrystals are poorly understood. This is mainly due to the difficulty in preparing specimens from mechanically alloyed powders for direct transmission electron microscopy (TEM) observations. The simplest method involves suspending the powder in a volatile liquid and pipetting the suspension on to a carbon support film. Usually, TEM images can only be r 1998 WILEY-LISS, INC. obtained from the edges of mechanically alloyed powders. However, the microstructure of the edges may differ from that in the interior of the particles, and for powders with a large size distribution the small fraction may not be representative of the whole. A complex method is that the powders are consolidated by hot compaction or by extrusion, but the original structure during such a process may be destroyed. Most commonly, powder and nickel are plated over a substrate and the resultant composite is electrochemically polished (Field and Fraser, 1978; Kang and Benn, 1987; Yang et al., 1987). Recently, two novel techniques [powder-fixing method (Huang et al., 1994b) and powder-dispersing method (Huang et al., 1994c)] for specimen preparation of mechanically alloyed powders were developed. Using the two new techniques some original microstructural information of ball milled materials was obtained in our laboratory in the past 6 years and is reviewed in this paper. This type of work has been less commonly reported due to difficulties in specimen preparation but is, in fact, very important since many results obtained by MA or BM are rather puzzling and in many cases direct observations seem to be very necessary. It is expected that TEM or high-resolution electron micros- Contract grant sponsor: National Natural Science Foundation of China. *Correspondence to: H.Q. Ye, Laboratory of Atomic Imaging of Solids, Institute of Metal Research, Chinese Academy of Sciences, Shenyang 110015, China. E-mail: email@example.com Received 20 February 1996; accepted in revised form 14 March 1996 102 J.Y. HUANG ET AL. copy (HREM) could shed some new light on the existing problems. MATERIALS AND METHODS An irregular copper powder (99.0% 1 pure) with a particle size smaller than 70 µm was milled in a vibratory ball mill in an argon atmosphere. Some Fig. 1. SEM micrograph of copper powders after 20 hours of milling, showing the formation of spheres. powders after 20 hours of ball milling have agglomerated to spheres or flat fragments with diameters of about 2–2.5 mm due to cold-welding, as shown in Figure 1. Some of them were mechanically ground, dimpled, and ion thinned for HREM observations. During MA of Fe-Cu system, two different methods were used to prepare the TEM specimens. For mechanically alloyed Fe16Cu84, the powder-dispersing procedure was used, which involved mechanical grinding and subsequent ion beam thinning of the small agglomerates formed during MA (Huang et al., 1994c). The specimens for Fe60Cu40 powders after 1, 3, and 60 hours of MA were prepared by the powder-fixing method (Huang et al., 1994b), which involved mixing the powders into a small amount of epoxy resin and then a piece of metal net was embedded into the mixture. The as-prepared samples were flattened and solidified for a few hours, and then were mechanically ground, dimpled, and finally thinned by ion beam milling. The MA experiment was performed in a planetary ball mill under the protection of an argon atmosphere. The mechanically alloyed Fe60Cu40 powders were heated in a differential thermal analyzer (DTA). After DTA heating to 1,400°C and then cooling to room temperature the powders have agglomerated to a small Fig. 2. TEM micrographs of copper powders after 20 hours of milling. a: Bright field image. b: Dark field image. c: Corresponding EDP. MICROSTRUCTURE OF BALL MILLED MATERIALS Fig. 3. HREM images showing complex configurations of higher order twins. 1–5 indicate the five twin variants. Arrowheads in b indicate the grain boundary. 103 104 J.Y. HUANG ET AL. Fig. 4. a: HREM image showing the lenticular morphology and the thickening of the multiple twins; the steps at one of the TB are denoted by arrowheads. b: Local magnification of shear bands in a, showing the dislocations piling up in it. c: Local magnification of a, showing a low-angle grain boundary produced in the shear bands. Notice the 51116 planes of region A deviating by about 5°. cylindric block with length 5 mm and radius 1.5 mm. The agglomerate was cut into thin slices, polished, and ion beam thinned for TEM observations. HREM observations were carried out with a JEM-2000EXII electron microscopy operated at 200 kV. The energy dispersive X-ray spectroscopy (EDXS) was performed in a Hitachi HF-2000 FEG TEM with a probe diameter of about 1 nm in the microanalysis mode. RESULTS AND DISCUSSION Ball Mill of Pure Copper (Huang et al., 1995, 1996a) The grain size obtained from X-ray diffraction (XRD) using the Scherrer formula for the unmilled copper is 340 nm. From TEM observations, after 20 hours of milling the grain size varies from region to region, but generally it falls in the range of 10–100 nm. Typical bright and dark field images with their corresponding electron diffraction pattern (EDP) are shown in Figure 2a–c, respectively. The different grain size obtained in one specimen infers that the microstructure is inhomogeneous through the specimen. HREM revealed that a number of deformation twins were generated in these Cu grains. Figure 3 shows higher order twins formed in the Cu grains. In Figure 3a the five segments do not arrange in the manner of fivefold symmetry. In order to fill the whole space, many internal distortions were produced, as evidenced by mismatch dislocations at the interface between segments 1 and 2, and the multiple twins in segment 2 as well as the edge dislocation in segment 5, etc. Figure 3b shows other high-order twins. The strained regions in segments 4 and 5 cross continuously through the twin boundary (TB). In some cases, twinning takes place on only one set of twinning planes. Figure 4a shows such a grain that was fully deformed by multiple twinning. Most of the twin lamellas, which MICROSTRUCTURE OF BALL MILLED MATERIALS Fig. 4. are thickened by passage of some steps, are curved and their tendency to be lenticular has become obvious. The TBs are not one atomic plane thick but extend over a few layers, and in many regions they have developed to bands with a width of 2–5 nm. Figure 4b is a higher magnification of a TB in Figure 4a, and it shows clearly that a number of dislocations are piling up in it. It is hereby suggested that the described structure is in fact the ‘‘shear bands’’ (SBs). Figure 5 shows SBs formation in another grain. There are four SBs in the interior of this grain with a width of 5–10 nm and an average spacing of 10 nm running parallel to the slip planes. In SB 1, microtwins can also be found. The lattice image in most of the SBs has heavily distorted, indicating that severe plastic deformation has occurred in these regions. It should be mentioned that the lower right part of this grain was not deformed, and a misorientation of about 10° has been produced between the deformed region and the undeformed one. Thus, a low-angle grain boundary (LAGB) was formed. 105 (Continued.) With the development of the SBs, subgrains tend to form in these bands, and this has been accepted as a typical mechanism for grain size reduction during plastic deformation (Fecht et al., 1990; Hansen, 1982) (hereafter referred to as the first way). We indeed observed some LAGBs in some of the SBs: a typical example is shown in Figure 4c, in which the lattice image of region A has rotated an angle of about 5° with respect to the twinning planes, and there is a strong tendency to form subgrains. In addition to the first way, some LAGBs were found to form at the tip of the SBs (Fig. 5) or the TBs (hereafter referred to as the second way). Figure 6a shows a HREM image of four grains. Deformation is rather inhomogeneous through these four grains. Grain 4 was deformed by multiple twinning, while grain 1, 2, and 3 were not deformed. Grain 1 has rotated an angle of about 4° with respect to grain 4. From a local magnification of the grain boundary (GB) as shown in Figure 6b, it can be seen that a number of dislocations are piling up in the GB region. It is 106 J.Y. HUANG ET AL. Fig. 5. HREM image showing the shear bands (denoted by 1–4), and the low-angle grain boundary formed at the tip of the shear bands (indicated by the black arrowheads). Disordered and lattice rotation regions are marked by black dots and a white arrow, respectively, the short and long black lines indicating the microtwins and the misorientation between the two neighboring grains, respectively. interesting that most of these dislocations are in pairs, and generally they belong to two types (denoted by ‘‘A’’ and ‘‘B’’). The Burgers vector bA 5 (1/2) or (1/2), while bB 5 (1/2) or (1/2), and they both have an angle of 60° with respect to the dislocation lines. With continuous refinement of the grain size, the number of GBs increased rapidly. Figure 7 shows a HREM image of a GB. Most of this GB is ordered, but lattice distortion in local regions can also be detected (see regions ‘‘P’’ and ‘‘O’’). Figure 8a is a typical HREM image containing a number of grains and GBs. All the GBs in Figure 8a are ordered, curved, and show a strained appearance. Figure 8b is a local magnification of grain 1 in Figure 8a; a number of dislocations were found in the interior of this grain. All the observed dislocations are also in pairs. Detailed analyses found that these dislocations belong to two different types (marked by ‘‘A’’ and ‘‘B’’) and the Burgers vector bA 5 (1/2) or (1/2), and both have an angle of QA 5 60° with dislocation lines. So dislocation A is a mobile one. However, bB 5 (1/2), QB 5 90°, and it is immobile. It was believed for many years that a face-centercubic (f.c.c.) metal does not deform by twinning, since they have enough slip systems to accommodate deformation. In fact, mechanical twins were never observed in copper when deformed at room temperature by cyclic fatigue (Liu et al., 1994; Winter et al., 1981) or tensile stress (Hansen and Ralph, 1982). However, Blewitt et al. (1957) showed that twins were formed during the tensile testing of certain copper crystals at low temperatures. Since this initial work, several workers have reported additional observations of this mode of deformation. Smith (1958) found that copper after shocks of above about 200 kbars shows numerous mechanical twins, which, however, make no significant contribution to hardening. Johari and Thomas (1964) showed that after a critical pressure the mode of deformation changed from slip to microtwinning in explosively deformed Cu. The critical pressure depends on the stacking fault energy. Matthews (1970) pointed out that the deformation of copper thin film under tensile stresses was accompanied by the generation of lenticu- MICROSTRUCTURE OF BALL MILLED MATERIALS Fig. 6. a: HREM image showing a low-angle grain boundary formed at the tip of multiple twins. Black dots indicate the grain boundaries, black lines show a misorientation of about 5° between grain 1 and 4. b: Local magnification of the grain boundary between grains 1 and 4 showing a number of dislocations piling up on the grain boundary. A and B indicate the two types of dislocations. 107 108 J.Y. HUANG ET AL. Fig. 7. HREM image showing the ordered state of a grain boundary. Arrows indicate the grain boundary. P and O show the lattice distortion regions. lar twins whose growth seemed to take place by the migration of twinning dislocations. It appears from the present state of research on mechanical twinning that the latter is favored by deformation at low temperatures or very high strain rates as well as by lowering stacking fault energy (by alloying) (Thornton and Mitchell, 1962). The following factors may account for the generation of mechanical twins in ball milled copper (Huang et al., 1995, 1996a). Firstly, the pressure induced by BM was larger than the critical shear stress for twinning. Secondly, the grain size reaches a critical value below which twinning is the preferred mode of deformation. Finally, the strain rate induced by BM is rather high. The resultant materials from MA are usually nanocrystals. MA produces the nanostructure not by cluster assembly but by the structural decomposition of coarser-grained structures as a result of severe plastic deformation. While the structural characteristic of the nanostructured materials (NC-materials) produced by other methods, such as gas-condensation/vacuum compaction (Gleiter, 1990), have been studied extensively, that produced by MA was rarely reported. We got some structural information of NC-copper produced by MA at an atomic level. Most of the investigated nanograins are heavily strained and contain a high density of dislocations. This is in contrast to the NC-Cu prepared by sliding wear (Ganapathi and Rigney, 1990), in which no structural defects were detected inside grains. This is possibly caused by the different grain size of the resultant materials. In our case the grain size of the resultant NC-Cu is 10–100 nm, while in the latter one, a grain size of 4–5 nm was reported. Generally, if the grain size is too small, the possibility of activating a dislocation is rare according to the Hall-Petch relationship. The GB structure in NC-materials is still controversial in terms of whether a ‘‘gas-like’’ structure can exist in the GBs of NC-materials. Most of the research accomplished to date shows that the GB structure of NC-materials is ordered and is similar to those in coarse-grained materials (Ganapathi and Rigney, 1990; Li et al., 1993; Siegel, 1994; Suryanarayana, 1995). In ball milled copper, the conclusions are similar to those drawn above. All the investigated GBs are ordered, curved, and strained. The lattice distortion and nanovoids were frequently encountered in the matrix nearby GBs, and GBs contain a high density of dislocations in several cases. It can be concluded that there is no characteristic difference between the nanograined and coarse-grained GBs in ball milled copper. Mechanical Alloying in an Immiscible Fe-Cu System A number of studies have been focused on the FeXCu(100-X) system since it was shown that the miscibility of Fe in Cu can be greatly enhanced, 0 , X , 60, through MA (Corespo et al., 1993; Eckert et al., 1993; Jiang et al., 1993; Ma and Atzmon, 1995; Ma et al., 1993; Marci et al., 1993; Uensh et al., 1992; Yavari et al., 1992). However, direct observation on the mechanically alloyed Fe-Cu alloys has not been reported. In the present paper, the mechanically alloyed Fe16Cu84 and Fe60Cu40 were directly observed by HREM. Both atomic MICROSTRUCTURE OF BALL MILLED MATERIALS Fig. 8. a: HREM image showing the nanocrystals formed in ball milled copper. Arrowhead indicates the strain and the lattice distortion. b: Local magnification of grain 1 in a showing two types of dislocations (60° and 90° dislocations marked by A and B, respectively). The Burgers circuits of each were drawn. 109 110 J.Y. HUANG ET AL. Fig. 9. TEM micrographs of MA Fe16Cu84. a: Bright field image. b: Dark field image. c: Corresponding EDP. Fig. 10. TEM micrographs of MA Fe60Cu40. a: Bright field image. b: Dark field image. c: Corresponding EDP. level structure and nano-scale composition information were obtained. Figures 9 and 10 are the TEM micrographs corresponding to specimen Fe16Cu84 and Fe60Cu40, respectively. Both the EDPs show a set of fcc phases, which agrees well with the XRD results (Huang et al., 1996c). From the bright and dark field images it can be seen that the grain size varies from 10 to 100 nm. Figure 11 shows a HREM image of 3 grains in MA Fe16Cu84: in the large grain, i.e., grain 2, small twin lamellas can be found. However, the structure of the smaller grains 1 and 3 is almost perfect. These GBs are curved and slightly strained; lattice distortion or nanovoids can also be detected in local regions, as marked by circles in Figure 11. In many cases, the size of the Fe grains can be as small as 2–5 nm, and these small Fe grains tend to insert into the interior of the Cu grains, as shown in Figure 12. Figure 13 is a HREM image containing a MICROSTRUCTURE OF BALL MILLED MATERIALS Fig. 11. HREM image of GB structures in MA Fe16Cu84. 1–3 denote the number of grains. Arrowheads indicate the GBs. Circles mark nanovoids or local disordering in the GBs. Fig. 12. HREM image of MA Fe16Cu84 showing an Fe grain inserting into the center of a Cu grain. 111 112 J.Y. HUANG ET AL. Fig. 13. HREM image of GB structures in MA Fe16Cu8. Circle indicates a small Fe grain. 1–5 denote the number of grains. number of grains and GBs in an Fe16Cu84 specimen. These grains are flattened and some nano-scale layered structures have developed. Even though the diffraction peak from the bcc phase has completely disappeared in the XRD of this specimen, small Fe domains can still be detected in the HREM image shown in Figure 13, as indicated by a circle. Figure 14 shows the HREM image of Fe60Cu40 after 1 hour MA. The grain size has reached nanometer scale. The Cu grains are coarse and their size is about 5–30 nm; while the Fe grains are very small (less than 5 nm), and some domains with only a few atomic planes were also detected, as indicated by circles in Figure 14. After 3 hours MA, some large Cu grains can still be found, as shown in Figure 15. It can be seen that this large Cu grain is heavily strained and the deformation occurred in a rather inhomogeneous way. The center of Figure 15 is a highly deformed region of about 10 nm in width that extends throughout the grain. The HREM image of this region reveals microstructures consisting of individual grains with a diameter of about 10 nm, which are slightly rotated with respect to each other at a rotation angle of less than 20°. After 60 hours MA, most of the grains are very small and their size is less than 5 nm. Although no bcc diffraction is observed in both the EDP and XRD, some ultra-fine Fe domains can still be detected, as marked by the circles in Figure 16. The composition of both the grain interior and the GBs of specimen Fe16Cu84 was analyzed by EDXS. Typical EDXS spectra from the grain interior and the MICROSTRUCTURE OF BALL MILLED MATERIALS Fig. 14. 113 HREM image of Fe60Cu40 after 1 hour of MA. Circles show some small Fe domains. GBs are shown in Figure 17. The result showed that the average Fe content (about 16%) in both the grain interior and the GBs is very close to the designed composition (16%), but the Fe content in both cases is rather inhomogeneous; generally, the smaller the grain size, the higher the Fe content. However, in most cases the Fe content falls in the range of about 5 to 20%, which infers that most Fe atoms have dissolved in the Cu lattice and true alloying has occurred. A standard method for identifying the phase that is formed under MA is XRD. In many investigations, the lattice parameters of phases such as solid solutions are determined from the XRD patterns and plotted as a function of composition. Continuous variation of the lattice parameters with composition, such as a Vegard law dependence, is commonly taken as an indication that the material is single phase. However, the question arises as to the homogeneity of such solid solution phases. While the structure of a solid solution is readily identified by XRD, data about its homogeneity are not easily obtained. This has been demonstrated by X-ray investigations on Co/Cu multi-layers and on decomposed Co-Cu solid solutions (Michaelsen, 1995). Although these materials consist of separate fcc Cu and fcc Co regions, they exhibit a diffraction pattern that can hardly be distinguished from the XRD of a solid solution having the same overall composition, if the Cu and Co regions are coherent and their size is smaller than about 5–8 nm. Michaelsen (1995) therefore concluded that it is not possible to determine the supersaturation of a solid solution using a conventional XRD. Complementary measurements have to be performed in order to prove the homogeneity of a supersaturated alloy. In the present experiment, EDXS result shows that the average Fe contents in the interior of most fcc grains in the FCB specimen are close to the designed composition, which is certainly direct evidence for the formation of supersaturated solid solutions. However, the mixing is not homogeneous since the data of the Fe contents in both the interior of grains and the GBs are rather scattering. The mechanism by which a solid solution with positive heat of mixing is formed upon milling is still a subject of controversy. Yavari et al. (1992) addressed the interfacial energy as a driving force for homogenization. They found that the chemical component of the interfacial energy is insufficient to dissolve nanograins of average size as determined by XRD, and only the tail of the grain size distribution with r , 1 nm will dissolve initially. They then suggested that codeformation re- Fig. 15. HREM image of Fe60Cu40 after 3 hours of MA, showing the generation of subgrains in the shear band (denoted by 1–3). Fig. 16. HREM image of Fe60Cu40 after 60 hours of MA. Circles indicate some small Fe domains. MICROSTRUCTURE OF BALL MILLED MATERIALS 115 Fig. 18. The chemical contribution schem of the Fe/Cu interfaces to the enthalpy of a Fe60Cu40 composite as a function of the size d of the Fe and Cu regions. Fig. 17. EDXS spectra of MA Fe16Cu84 in the interior grains (a) and in the GBs (b). sults in the formation of thin Fe layers, which, by undergoing necking to striction, generate smaller fragments. Their calculations show that when the fragments reach diameters as small as 2 nm, the Fe-Cu interfacial energy, which has a large chemical component, results in their dissolution. Further deformation during MA then generates more such fragments that also dissolve until a complete solution is obtained. This kind of tip radii or fragments were indeed frequently observed in our HREM images (Figs. 12–14 and 16), which are surely direct evidence to support Yavari et al.’s model. According to this model, in going from a NC-mixture of bcc-Fe and fcc-Cu crystals to a fcc solid solution of equal grain size, the interfacial energy is reduced by DE 5 6Vmschem/d, where Vm is the molar volume and schem is the chemical contribution to the interfacial energy. DHmix 5 DE 1 TDSmix, where DHmix and DSmix are the enthalpy and entropy of mixing, respectively. The refinement of the microstructure enhances the chemical contribution of the interface enthalpy as shown in Figure 18 for different sizes of the Fe and Cu regions. It can be seen that the grain size down to 1 nm raises the enthalpy of a Fe60Cu40 composite above 17 KJ/mol, which is sufficient to allow for a phase transformation of the composite into a supersaturated solid solution; there is still additional enthalpy from the heavy mechanical deformation, which is not considered. Thus, the enthalpy due to the contribution of alloying at the interface and mechanical deformation may be large enough to overcome the calculated positive enthalpy of mixing of 12 KJ/mol for Fe60Cu40. It is useful to compare the results of MA with that of plastic deformation under pressure in Fe-Cu. Teplov et al. (1995) recently reported that NC structure and nonequilibrium solid solutions of Fe80Cu20 and Fe20Cu80 were obtained after severe plastic deformation by shear under pressure. Based on grain rolling, a mechanism of superplastic deformation has been proposed. The main idea of such a mechanism is that plastic deformation of a fine-grain polycrystal is fully accommodated by the GBs. In the meantime, very fast volume diffusion may also exist (Gryaznov et al., 1992). Very fast GB diffusion and volume diffusion occur during the cold forced superplastic deformation by shear under pressure. Obviously, diminishing the grain sizes to minimal value, fast diffusion, and high internal stress in the fine-grain materials also take place during MA. During the early stage of MA, the grain size decreases rapidly to a steady value after only a few 116 J.Y. HUANG ET AL. Fig. 19. Susceptibility of Fe60Cu40 solid solution during heating and subsequent cooling at 10°C/min. sliding. This can increase the diffusion coefficient tremendously. The volume diffusion coefficient Dv may be calculated by the expression (Gryaznov et al., 1992): 4Dv · t 5 d2, where t is the time of plastic deformation, d is the grain size. For example, the typical milling time is about 1 hour from a NC-Cu and -Fe composite to a fcc solid solution, but at any instant of time during MA only a small fraction of the powder is undergoing impact or shear due to ball motion. Thus the deformation time t is also only a small fraction of the total milling time. Supposing t 5 600 seconds, and the grain size before alloying is about 5 nm; substituting t and d to the above equation, we obtain Dv 5 10216cm2/sec, which is several orders of magnitude higher than the volume coefficient of the diffusion of Cu in Fe at 300 K (Dv 5 10242cm2/sec) (Anand and Agarwala, 1966; Teplov et al., 1995). So NC structure and mechanical driven fast diffusion are the reasons for the formation of solid solution in immiscible systems with positive enthalpy of mixing. Fig. 20. XRD patterns of the as-milled Fe60Cu40 and after annealing at different temperatures. hours (generally less than 5 hours) (Eckert et al., 1993; Huang et al., 1994a) due to the mobility of dislocations. Further deformation can be accommodated only by GB Thermal Decomposition of Mechanically Alloyed NC-Fe60Cu40 (Huang et al., 1996b) Figure 19 shows the susceptibility of Fe60Cu40 during decomposition heating to 915°C and subsequently cooling. The susceptibility tends to vanish around the Curie temperature Tc 5 250°C of the supersaturated fcc-FeCu solid solution before increasing as decomposition sets in above 300°C. The decomposition finishes at about 450°C. Abrupt changes of the susceptibility are observed at about 640 = 760°C on heating and from 800 to 640°C on cooling. Figure 20 shows the XRD patterns of Fe60Cu40 after milling and subsequently annealing at different temperatures. It is seen that after heating to 300°C, the bcc MICROSTRUCTURE OF BALL MILLED MATERIALS phase appears. The Fe60Cu40 specimen after DTA heating to 1,400°C and cooling to room temperature was characterized by TEM. Figure 21 shows the general TEM morphology of some Fe grains. The four Fe grains shown in Figure 21a with a size of about 200–400 nm are nearly spherical in shape; in some cases they contain many dislocations (indicated by arrowhead). Those shown in Figure 21b and c with a size of about 500–600 nm take an approximately cubic form, and some structural defects can also be detected in the center of the Fe grain shown in Figure 21b. Figure 22a shows a typical EDP between the a- and g-phases. One can see that the N-W orientation relationship, i.e., (110)a//(111)g; [00 1]a//[01 1] g, exists between them, which can also be clearly identified from the HREM image shown in Figure 22b. In several cases, another orientation relationship, the K-S orientation relationship, was also encountered, and a typical example is shown in Figure 23. In this case, the K-S orientation relationship holds with: (110)a//(111)g; [1 1 1]a//[01 1]g. Even after DTA heating to 1,400°C and subsequent cooling to room temperature, some small Fe grains were still detected. Figure 24 shows a low magnification HREM image of an Fe grain with approximately spherical morphology and a size of about 50 nm. A number of edge dislocations are realized in the interior of this grain. The strained region at the interface extends to about 2 nm. The corresponding EDP in Figure 24b shows that there also exists the K-S orientation relationship between Fe and the matrix. The composition of both the a-Fe and fcc Cu phases was analyzed by EDXS in a HF-2000 FEG-TEM with a probe diameter of 1 nm. The result showed that the Fe content in the matrix is about 1 at.%, whereas the Cu content within the Fe grains reaches as high as 9.5%, which even exceeds the highest solubility of Cu in g-Fe (8.5% at 1,094°C). A typical EDXS spectrum of an Fe grain is shown in Figure 25. From the thermomagnetic measurement and heat treatment results, the decomposition of the fcc Fe60Cu40 supersaturated solid solution could take place in the following way, on heating: fcc Fe60Cu40 (I) =300=460°C aFe(Cu) 1 g-Fe(Cu) 1 Cu(Fe) (II) =640=760°C g-Fe(Cu) 1 Cu(Fe) (III) =1,081°C g-Fe(Cu) 1 Cu(melted) (IV) and upon the subsequent cooling: g-Fe(Cu) 1 Cu(Fe)(melted) (IV)=1,126°C g-Fe(Cu) 1 Cu(Fe) (V) =800=640°C a-Fe(Cu) 1 Cu(Fe) (VI). The decomposition of the fcc solid solution from stage I to II has been investigated by many researchers (Corespo et al., 1993; Drbohlav and Yavari, 1994; Jiang et al., 1993; Marci et al., 1993). At the initial stage of decomposition, precipitation begins with the formation of the g-Fe phase. With prolonged annealing or with the rise of temperature, g-Fe precipitates grow, losing their fully coherency. When the annealing temperature is further raised, stable a-Fe precipitates nucleate from large g-Fe precipitates or directly from the dislocation tangles of the fcc Cu phase, as well as at the grain boundaries. In the meantime, due to excellent coherency between the g-phase and the matrix fcc phase, some small g-precipitates will not transform to the a-phase. The sluggish transformation of g- to a-Fe in a Cu matrix, which only reaches completion after cold working, was already reported by Easterling and Miekk (1967). Fig. 21. 117 TEM morphology of Fe60Cu40 solid solution after DTA run. With the rising of temperature, there is a a- = g-Fe transformation at 640 = 760°C (II = III), as proved by the abrupt decrease of the susceptibility shown in Figure 19. The reverse transformation is martensitic and occurs from 800 to 640°C. Obviously, the a t g transformations occur in a wide temperature range. The different grain size may be responsible for this behavior. After DTA heating to 1,400°C, a different grain size from 50 (Fig. 24) to 600 nm (Fig. 21) was observed. It has been reported (Cech and Turnbull, 1956) 118 J.Y. HUANG ET AL. Fig. 22. TEM micrographs showing the N-W orientation relationship between the a- and g-phases. a: EDP image. b: HREM image. Arrowheads indicate the interface between the a- and g-phases. Fig. 23. TEM micrographs showing the K-S orientation relationship between the a- and g-phases. a: EDP image. b: HREM image. Arrowheads indicate the interface between the a- and g-phases. that the martensitic transformation temperature can be decreased remarkably with decreasing grain size. The orientation relationship between the austensite and martensite in as-cast Fe-Cu alloys was established by Easterling and Weatherly (1969), and they found that the K-S orientation relationship exists between the austenite and martensite. However, in thermally treated MA Fe60Cu40, both the K-S and N-W orientation relationships were found between a-Fe and Cu phases. Due to excellent coherency between the g and Cu phases, this orientation relationship should also represent the one between g- and a-Fe. Also, after DTA heating to 1,400°C, the solubility of Cu in a-Fe exceeds the highest solubility of Cu in g-Fe (8.5 at.% above MICROSTRUCTURE OF BALL MILLED MATERIALS 119 Fig. 24. TEM micrographs showing a nanometer-scale a-Fe(Cu) remained after being heated to 1,400°C (a), and the K-S orientation relationship between the a- and g-phases (b). The indexing of EDP is the same as that of Figure 23. 1,094°C). This may be caused by the small grain size, since the grain size in Fe60Cu40 after DTA run is still not very large (50–600 nm). The martensitic transformation in bulk Fe-Cu alloys has been investigated by many researchers (Easterling and Miekk, 1967), but due to limited solubility in equilibrium between Cu and Fe, all the investigated compositions are Fe-rich ones with the Cu content less than 2 at.%. From these investigations, it is suggested that diffusionless decomposition of austenite occurs both by massive and martensitic mechanisms (Easterling and Miekk, 1967). The transformation tempera- 120 J.Y. HUANG ET AL. quite different from that in as-cast alloys, which is possibly due to the different fabrication process and the non-equilibrium microstructure (especially the high content of Cu in Fe) of MA. Because of the lack of substantial experimental data in as-cast alloys, we could not satisfactorily compare our results with that of the bulk materials at the present time. ACKNOWLEDGMENTS This work was supported by the National Natural Science Foundation of China, which is gratefully acknowledged. REFERENCES Fig. 25. A typical EDXS spectrum of the a-Fe phase. Fig. 26. The massive and martensite transformation temperatures in Fe-Cu alloy after Easterling and Miekk (1967), the allotropic transformation start temperature after Hansen (1958), and the martensite transformation temperature obtained in the present work. tures for both these reactions have been determined in the bulk materials (Easterling and Miekk, 1967), and are given in Figure 26. It is seen that there is no experimental data corresponding to the Cu composition in our alloy, but by extrapolating the Mass or Ms curves to our composition (9.5% Cu), the transformation temperature should be less than 500°C, which is lower than 700 = 622°C in the Fe60Cu40 alloy. Even by extrapolating the a t g allotropic transformation line to 9.5%, a temperature of 500°C might be obtained, which is still lower than that in Fe60Cu40. It seems that the martensitic transformation in thermally treated MA Fe-Cu is Anand, D.A., and Agarwala, R.A. (1966) Diffusion of copper in iron. J. Appl. Physiol., 37:4248–4251. Benjamin, J.S. (1970) Mechanical alloying. Sci. Am., 234:40–48. Blewitt, T.H., Coltman, R., and Redman, J.K. (1957) Low temperature deformation of copper single crystal. J. Appl. Physiol., 28:651–660. Cech, R.E., and Turnbull, D. (1956) Heterogeneous nucleation of the martensite transformation. J. Met., 8:124–132. Corespo, P., Hernando, A., Yavari, A.R., Drbohlav, O., Garcia Escorial, A., Barandiaran, J.M., and Orue, I. (1993) Magnetic behavior of metastable fcc Fe-Cu after thermal treatments. Physiol. Rev., B48:7134–7139. Drbohlav, O., and Yavari, A.R. (1994) Mechanical alloying and thermal decomposition of ferromagnetic nanocrystalline fcc-Fe50Cu50. Acta Metall. Mater., 43:1799–1809. Easterling, K.E., and Miekk, H.M. (1967) The martensitic transformation of iron precipitates in a copper matrix. Acta Metall., 15:1133– 1141. Easterling, K.E., and Weatherly, G.C. (1969) On the nucleation of martensite in iron precipitates. Acta Metall., 17:845–852.35. Eckert, J., Holzer, J.C., Krill III, C.E., and Jonson, W.L. (1993) Mechanical driven alloying and grain size changes in nanocrystalline Fe-Cu powders. J. Appl. Physiol., 73:2794–2802. Fecht, H.J. (1995) Nano-phase materials by mechanical attrition: Synthesis and characterization. In: Nanophase Materials. G.C. Hadjipanayis, and R.W. Siegel, eds. Kluwer Academic Publishers, pp. 125–144. Fecht, H.J., Hellstern, E., and Johnson, W.L. (1990) Nanocrystalline metals prepared by high energy ball milling. Metall. Trans., A21: 2333–2337. Field, R.D., and Fraser, H.L. (1978) Microstructure observations of metal powders using analytical electron microscopy. Met. Trans., A9:131–134. Ganapathi, S.K., and Rigney, D.A. (1990) An HREM study of the nanocrystalline materials prepared by sliding wear. Script. Metall. Mater., 24:1675–1678. Gente, G., Oehring, M., and Bormann, R. (1993) Formation of the thermodynamically unstable solid solutions in the Cu-Co system by mechanical alloying. Physiol. Rev., B48:13244–13252. Gleiter, H. (1990) Nanocrystalline materials. Prog. Mater. Sci., 33:223– 315. Gryaznov, V.G., Yu, M., Tanakov, V., and Trusov, L.I. (1992) Plastic and mass-transfer in contacting nanoparticles. J. Mater. Sci., 27:4829– 4841. Hansen, N. (1958) Constitution of Binary Alloys. McGraw-Hill Book Company, Inc., New York, pp. 580. Hansen, N. (1982) Cold deformation microstructures. Mater. Sci. Tech., 6:1039–1047. Hansen, N., and Ralph, B. (1982) The strain and grain size dependence of the follow stress of copper. Acta Metall., 30:411–417. Huang, J.Y., He, A.Q., and Wu, Y.K. (1994a) Nanocrystalline Cu-Fe solid solutions prepared by mechanical alloying. Nanostruct. Mater., 4:1–10. Huang, J.Y., He, A.Q., Wu, Y.K., and Ye, H.Q. (1994b) A new technique for specimen preparation for TEM studies of mechanically alloyed powders. J. Mater. Sci. Lett., 13:1201–1203. Huang, J.Y., Wu, Y.K., and Ye, H.Q. (1994c) A novel technique for specimen preparation of metal of ceramic powders for TEM or HREM observations. Matt. Lett., 21:167–170. Huang, J.Y., Wu, Y.K., and Ye, H.Q. (1995) Ball milling of ductile metals. Mater. Sci. Eng., A199:165–172. Huang, J.Y., Wu, Y.K., and Ye, H.Q. (1996a) Deformation structures in ball milled copper. Acta Metall. Mater. MICROSTRUCTURE OF BALL MILLED MATERIALS Huang, J.Y., Wu, Y.K., and Ye, H.Q. (1996b) Thermal decomposition of nanocrystalline Fe60Cu40. J. Mater. Res. Huang, J.Y., Yu, Y.D., Wu, Y.K., and Ye, H.Q. (1996c) Microstructure and nano-scale composition analysis on the mechanically alloyed FeXCu(100-X) (X 5 16, 60). Jiang, J.Z., Gonser, U., Gente, C., and Bormann, R. (1993) Thermal stability of the unstable fcc Fe50Cu50 phase prepared by mechanical alloying. Appl. Phys. Lett., 63:1056–1058. Johari, O., and Thomas, G. (1964) Substructures in explosively deformed Cu and Cu-Al alloys. Acta Metall., 2:1153–1159. Kang, S.K., and Benn, R.C. (1987) Characterization INCONEL alloy MA 6000 powder. Met. Trans., A18:747–752. Koch, C.C. (1995) Non-equilibrium materials by mechanical alloying. Mater. Trans. JIM, 36:85–95. Koch, C.C., Cavin, O.B., Mckamey, C.G., and Scarbrough, J.O. (1983) Preparation of ‘‘amorphous’’ Ni40Nb60 by mechanical alloying. Appl. Phys. Lett., 43:1017–1019. Li, D.X., Ping, D.H., Ye, H.Q., Qin, X.Y., and Wu, X.J. (1993) HREM studies of the microstructure in nanocrystalline materials. Matt. Lett., 18:29–34. Liu, C.D., Bassim, M.N., and You, D.X. (1994) Dislocation structures in fatigued polycrystalline copper. Acta Metall. Mater., 42:3695–3704. Ma, E., and Atzmon, M. (1991) Calorimetric evidence for polymorphous constraints on metastable Zr-Al phase formation by mechanical alloying. Physiol. Rev. Lett., 67:1126–1129. Ma, E., and Atzmon, M. (1995) Phase transformation by mechanical alloying in binary systems. Mater. Chem. Phys., 39:249–267. Ma, E., Atzmon, M., and Pinkerton, F. (1993) Thermal and magnetic properties of metastable FeXCu100-X solid solutions formed by mechanical alloying. J. Appl. Physiol., 74:943–962. Marci, P.P., Enzo, S., Cowlam, N., Frattini, R., Principi, G., and Hu, W.H. (1993) Mechanical alloying of immiscible Cu70TM30 alloys (TM 5 Fe,Co). Philos. Mag., B71:249–259. Matthews, J.W. (1970) Role of deformation twins in the fracture of single crystal film. Acta Metall., 18:175–181. Michaelsen, C. (1995) On the structure and homogeneity of solid solution: The limits of conventional X-ray diffraction. Philos. Mag., A72:813–823. Morris, D.G., and Morris, M.A. (1996) Milling of intermetallics and attempts at their ductilization. Scripta Mot. 34:45–51. Nash, P., Wu, S-H., Ur, S.-C., and Dollar, M. (1995) Design of dispersion strengthened NiAl. Mater. Trans. JIM, 36:351–356. 121 Schwarz, R.B. (1996) Introduction to the view point set on: Mechanical alloying. Scripta Mater., 34:1–4. Schwarz, R.B., and Koch, C.C. (1986) Formation of amorphous alloys by the mechanical alloying of crystalline powders of pure metals and intermetallics. Appl. Physiol. Lett., 49:146–148. Shingu, P.H., and Ishihara, K.N. (1995) Non-equilibrium materials by mechanical alloying. Mater. Trans. JIM, 36:96–101. Shingu, P.H., Huang, B., Nishitani, S.R., and Nasu, S. (1988) Nanometer order crystalline structures of Al-Fe alloys produced by mechanical alloying. Trans. JIM, (Suppl) 29:3–10. Shingu, P.H., Ishihara, K.N., Uenishi, K., Kuyama, J., Huang, B., and Nasu, S. (1990) Metastable alloy phase formation by repeated rolling in some common metallic binary alloy systems. In: Solid State Powder Processing. A.H. Clauer and J.J. deBarbadillo, eds. The Minerals, Metals and Materials Society, pp. 21–34. Siegel, R.W. (1994) Mechanical properties of nano-phase materials. In: Nano-Phase Materials. G.C. Hadjipanayis, and R.W. Siegel, eds. Kluwer Academic Publishers, pp. 233–261. Smith, C.S. (1958) Metallographic studies of metals after explosive shock. Trans. Met. Soc. AIME., 212:574–589. Suryanarayana, C. (1995) Nanocrystalline materials. Inter. Mater. Rev., 40:41–64. Teplov, V.A., Pilugin, V.P., Gaviko, V.S., and Chernyshov, E.G. (1995) Nanocrystalline structure of non-equilibrium Fe-Cu alloy obtained by severe plastic deformation under pressure. Nano. Mater., 6:437– 440. Thornton, T.R., and Mitchell, T.E. (1962) Deformation twinning in alloys at low temperature. Philos. Mag., 7:361–375. Uenish, K., Kobayashi, K.F., Nasu, S., Hatano, H., Ishihara, K.N., and Shingu, P.H. (1992) Mechanical alloying in the Fe-Cu system. Z. Metallkd., 83:132–135. Winter, A.T., Pedersen, O.B., and Rasmussen, K.V. (1981) Dislocation microstructures in fatigued copper polycrystal. Acta Metall., 29:735– 748. Yang, J.C., Nash, P., and Bradley, S. (1987) A technique for producing thin foils of mechanically alloyed Ni-Ti amorphous powders. J. Mater. Sci. Lett. 6:982–984. Yavari, A.R., Desre, P.J., and Benameur, T. (1992) Mechanical alloying of immiscible elements. Physiol. Rev. Lett., 68:2235–2238.