Synthesis and microstructure of bulk nanocrystalline ... - Springer Link

9 downloads 0 Views 669KB Size Report
C. LANGLOIS, M.J. H ¨YTCH, P. LANGLOIS, S. LARTIGUE-KORINEK, and Y. CHAMPION. The synthesis of bulk nanocrystalline copper (NC-Cu) by powder ...
Synthesis and Microstructure of Bulk Nanocrystalline Copper ¨ C. LANGLOIS, M.J. HYTCH, P. LANGLOIS, S. LARTIGUE-KORINEK, and Y. CHAMPION The synthesis of bulk nanocrystalline copper (NC-Cu) by powder metallurgy is presented, from compaction of nanocrystalline powders to sintering and differential extrusion. At each step of the process, the microstructure is characterized using X-ray diffraction (XRD) analysis, field-emission gun scanning electron microscopy (FEG-SEM), and transmission electron microscopy (TEM). Particular attention is given to the concept of grain size in nanostructured materials and the comparison of results from the different characterization techniques. The fully dense material has a grain size of 100 nm with a microstructure best described in terms of the distribution of high-angle grain boundaries (GBs), twin boundaries, and low-angle GBs. Dislocations occur in half the grains and at most of the twin boundaries. The GBs are shown to be crystalline, and no evidence is found for amorphous interfacial regions. It is proposed that the grain size be defined only in terms of high-angle GBs, excluding low-angle GBs, for the discussion of mechanical properties. In this respect, the microstructure is compared with the NC-Cu material produced by other synthesis techniques. Powder metallurgy (P/M) processing is revealed as an alternative for the production of large-size submicrocrystalline and NC materials.

I. INTRODUCTION

NANOCRYSTALLINE copper (NC-Cu) has been shown to have exceptional mechanical properties, from high strength and ductility[1] to near-perfect elastoplasticity,[2] high toughness and ductility,[3] and “super extensibility.”[4] Superplasticity may even be achievable for small-enough grain sizes.[5] Both the study of mechanical properties and the development of potential applications require the preparation of sufficiently large amounts of bulk nanostructured material. So far, results have mainly been reported by two specimen-preparation techniques that produce very different microstructures. Equal-channel angular pressing[6] is able to produce extremely large specimens. The grain size of such specimens is quoted to be in the range of 150 nm, limited by the underlying dislocation-based mechanism. Other severe plastic deformation (SPD) techniques produce variants in the microstructure; high-pressure torsion (HPT)[7] allows further grain refinement (up to 20 nm), and severe cold rolling at liquid-nitrogen temperature[3] produces an interesting bimodal nanomicrostructure (with grain sizes of 200 nm and 1 to 3 m). Electrodeposition is another technique that produces reasonably large specimens. For example, the lowest normalized superplastic temperature was reported for Ni,[8] and super extensibility was obtained for copper. The grain size was quoted to be as low as 20 nm, although an extremely large proportion will be subgrains separated by low angle GBs. C. LANGLOIS, formerly Ph.D. Student, Centre d’Etudes de Chimie Métallurgique CECM-CNRS, is Associate Professor, Laboratoire Matériaux Et Phénom`enes Quantiques, LPS, ESPCI 10 Rue Vauquelin 75252 Paris Cedex ¨ 05 France. M.J. HYTCH, S. LARTIGUE-KORINEK, and Y. CHAMPION, Senior Researchers, are with the Centre d’Etudes de Chimie Métallurgique CECM-CNRS, 94407 Vitry-sur-Seine, France. Contact e-mail: champion@ glvt-cnrs.fr P. LANGLOIS, Senior Researcher, is with the Laboratoire d’Ingénierie des Matériaux et des Hautes Pressions LIMHP-CNRS, Université Paris XIII, 93430 Villetaneuse, France. Manuscript submitted June 28, 2004. METALLURGICAL AND MATERIALS TRANSACTIONS A

The SPD and electrodeposition are seen as the most relevant preparation techniques, because they produce fully dense bulk materials. The SPD technique, as an in-situ refinement process, has the additional advantage of producing noncontaminated samples. In contrast, powder metallurgy (P/M) processing, either of nanopowders or ball-milled powders, is often criticized regarding bulk materials fabrication. The main reasons are that sample purity and ultimate densification are difficult to obtain. Contamination can be avoided by compacting nanopowders in situ at high pressure, but the resulting specimens are small.[9] Larger specimens can be made by keeping particles oxidized[10] and with in-situ processing, but specimens then suffer from residual porosity.[11] If chemistry and densification can be controlled, P/M technology offers new perspectives for nanostructures. The grain size and the nature of grain boundaries (GBs) can, in principle, be modified during the grain-assembly process, thus providing additional control over properties. Furthermore, the P/M process allows the fabrication of specific macro and micro objects with shapes inaccessible to standard forming and machining techniques. Here, we present the synthesis of fully dense NC-Cu by P/M, allowing the production of large specimens, notably for tensile testing, with a 15-mm gage length and 3-mm diameter.[2] A major aspect will concern the characterization of the materials microstructure during the processing, as well as the final material. A wide range of characterization techniques will also be employed in order to arrive at a coherent model for the microstructure. Some of these techniques have been specially adapted to the characterization of materials structured at the nanoscale. The notion of grain size will be particularly important. Plasticity of metals is governed by the generation, movement, and interaction of dislocations. Materials can be considered as being nanocrystalline and, therefore, having significantly different mechanical properties to their coarser-grained counterparts, for grain sizes significantly larger than those for, e.g., optical properties, due to quantum confinement. Relevant VOLUME 36A, DECEMBER 2005—3451

length scales for dislocation-based phenomena are in the 50to 100-nm range. However, not all boundaries between grains affect mechanisms involving dislocations to the same degree. This will lead us to propose a definition of the grain size in terms of the surrounding GBs that appear the most relevant to the discussion of mechanical properties. Microstructures of NC-Cu material obtained from different synthesis techniques will be discussed in consequence. II. SAMPLE PROCESSING The synthesis of a fully dense NC-Cu sample can be divided into five stages: (1) NC-Cu powder production by evaporation condensation, (2) powder precompaction under cold isostatic pressing, (3) sintering under a controlled reductive atmosphere, (4) ultimate-densification hydrostatic differential extrusion at room temperature, and (5) annealing. Details on powder production can be found elsewhere.[12] The nanocrystals of Cu are covered by a thin (2 to 3 nm) layer of copper oxide Cu2O due to the diluted presence of water in the organic solvent used to store the powders. This passivation acts as a protection against pyrophoric effects that allows the powders to be handled in air. In order to allow the transfer of pressure homogeneously to the sample during further compaction, a crushable latex tube is filled with NC-Cu powder, using a piezo-electrical vibrator. The tube allows the transfer of pressure homogeneously to the sample during compaction. The density after filling is about 8 pct of the density for pure copper. The relative density of the NC-Cu material (dr) is given by > Ve

me

dr  100 #

d0

(pct)

with me being the sample weight, Ve the sample volume, and d0 the density of pure copper (8.96). The volume was measured geometrically with an accuracy of about 1 pct. Compaction is carried out at under an isostatic pressure of 400 MPa for 5 minutes at room temperature (Figure 1(a)). During compaction, the dimensions (diameter and length) of the sample are reduced by a factor of 2 and the relative density reaches approximately 70 pct. Further compaction at this stage is not desirable, since an open porosity is necessary for reducing the oxide layer. Sintering conditions were determined from dilatometry experiments under H2 flow and by monitoring sintering reactions with X-ray diffraction (XRD) analysis, in order to achieve the highest densification.[13] Owing to the oxide layer, the powder processing occurs in two separate steps: (1) the reduction of the oxide under H2 flow and (2) sintering of the powder with shrinkage. A slow temperature ramp was chosen to carry out the two successive stages. Sintering was performed between 20 °C and 240 °C at a rate of 0.5 °C/min, followed by a 10-minute plateau and then slow cooling at an estimated rate of 0.3 °C/min. Sintered samples have typical dimensions of 25 mm in length and 8 mm in diameter and have a relative density of around 90 pct. Final densification is obtained by means of hydrostatic differential extrusion at room temperature. The sintered sample is machined to fit in a cylindrical copper billet that is extruded under fluid pressure through a tungsten carbide die 3452—VOLUME 36A, DECEMBER 2005

Fig. 1—Description of compaction and extrusion processing steps: (a) cold isostatic pressing and (b) hydrostatic differential extrusion.

(Figure 1(b)). Frictional and redundant losses are minimized by using 45-deg conical dies. Hydrostatic differential extrusion is of prime interest for limiting grain growth, since it allows temperature to be substituted by pressure when deforming materials in their plastic domain. Direct extrusion, i.e., with only atmospheric pressure at the exit of the die, results in multiple fractures of the sample inside the billet. Plastic flow is improved by applying a backpressure to the lower chamber (cf. Figure 1(b)). This is known to avoid the dramatic release of potentially high METALLURGICAL AND MATERIALS TRANSACTIONS A

Table I. Chemical Purity of Sintered NC-Cu Material: ICP-OES Analysis; Impurities Comparable to Raw Material Element Al Ca Cr Fe K Mg Na Ni S Si Zn

Fig. 2—Thermal stability of the NC-Cu: microhardness measurements (Vickers) after different annealing times (hours) and temperatures (°C).

stresses immediately after the die and prevents the emergence of tensile deformation zones near the axis by limiting the stress gradient along the axis. Furthermore, the compressivestress state is globally higher, leading to better densification. The effective stress causing extrusion depends on the difference of pressure between the upper and lower chambers and can be estimated by the following semiempirical formula: P  c(6.28 HV  430) ln Q D> d R d (MPa) where HV is the mean Vickers hardness for the composite billet, and D and d are the diameters of the billet and of the extrudate, respectively. In order to limit the overall deformation due to the extrusion step, the differential pressure was limited to 400 MPa, which corresponds to the minimum required for a reduction of the sections of 3:2, i.e., for providing extruded samples with final dimensions of 50 mm in length and 5 mm in diameter. As anticipated, the density after extrusion reaches 99  1 pct. After the extrusion, the residual stresses inside the material are reduced by annealing at low temperatures to avoid activating grain growth. In order to estimate the grain-growth kinetics, experiments of annealing at four different temperatures were carried out. The grain size was monitored by microhardness testing, with the assumption that the hardness decreases with grain coarsening. Measurements were averaged over a series of ten indentations performed with a load of 100 g for 30 seconds using a standard Vickers square pyramidal indenter installed on a Zeiss optical microscope (Zeiss, Oberkochen, Germany). Hardness values for each time/temperature couple are reported in Figure 2. Optimum conditions lie between 140 °C and 165 °C in order to avoid grain growth (visible by a decrease in hardness) and to minimize the annealing time. At 200 °C, the grain size shows a dramatic coarsening after a few minutes and is, therefore, not reported on the figure. Accordingly, annealing conditions were fixed to 30 minutes at 155 °C. The XRD measurements confirm that strain release occurs without grain growth. Chemical analyses of the Cu rod used for powder synthesis as well as that of the as-extruded NC-Cu were performed via inductively coupled plasma–optical emission spectroscopy (ICP-OES). Table I shows the results for the two materials. METALLURGICAL AND MATERIALS TRANSACTIONS A

Sample: Sintered NC-Cu (wt ppm) 35 337 1.5 38 74 2.4 44 19 48 112 23

The main impurities in the extruded NC-Cu are the same as in the original copper rod, showing that no pollution occurs during the sample processing. III. CHARACTERIZATION TECHNIQUES The microstructure of complex materials can only be fully understood by using a range of characterization techniques and subsequent comparison of results. We apply this strategy to the study of the evolution of the microstructure through the successive fabrication stages described in the previous section. The most significant structural parameters are grain size, residual porosity, and GB distribution in terms of low-angle, high-angle, and twin boundaries. The presence of dislocations will also be investigated. The variations of these features for each step of the process are studied by the mean of scanning electron microscopy (SEM), transmission electron microscopy (TEM), and XRD. The SEM observations are performed with a LEO 1530 fieldemission gun–scanning electron microscope (Zeiss, Oberkochen, Germany) operating at 5 kV using an in-lens and classical secondary electron detector. For the compacted and sintered NC-Cu material, surfaces after fracture are observed without any particular preparation. Electrochemical etching was performed with an orthophosphoric-based solution at a voltage of 2 V and 0 °C to reveal grain morphology and porosity on the extruded and annealed samples. The XRD measurements were carried out with a PHILIPS* *PHILIPS is a trademark of Philips Electronic Instruments Corp., Mahwah, NJ.

PW 1762 detector with Co K radiation. The X-ray wavelengths of K1  1.78896 and K2  1.79285 were selected using a graphite monochromator. Calibration of the instrumental line broadening was achieved using a LaB6 powder standard. Peak profiles were fitted by a split pseudo-Voigt function using the EVA software from Bruker (Bruker AXS, Karlsruhe, Germany) and SOCABIM*. *SOCABIM is a trademark of Bruker, AXS, Karlsruhe, Germany.

Thin foils for TEM observations were obtained by electrochemical thinning with a STRUERS TENUPOL 3* apparatus *STRUERS and TENUPOL 3 are trademarks of Struers, Ballerup, Denmark.

VOLUME 36A, DECEMBER 2005—3453

using a commercial D2 solution from Struers at 6 V and 6 °C. Conventional TEM investigations were conducted on a JEOL* 2000 EX *JEOL is a trademark of Japan Electron Optics Ltd., Tokyo.

transmission electron microscope operating at 200 kV. A. The XRD Line-Broadening Analysis In this work, a specific analysis is used to obtain the true diffraction-coherent domain size (crystallite size) by removing the contribution from twins and stacking faults as well as dislocation-strain fields. The classical Williamson–Hall (classical W-H) plot can be modified to take into account the twinning and stacking faults as well as the dislocation density.[14] The starting point is the following classical W-H relationship:[15] K 

1  2hK D

where 2sinq , l (2q)cosq , K  l K

(2 )  the integral breadth of a reflection, h

K is a microstrain ratio, and K

D  the crystallite size. The contribution of stacking faults and twins in the classical W-H equation has to be removed from the broadening of the peaks:[16] 1 K (1.5a  b)/a # W(K)   2hK D where a is the lattice parameter, and  and are the probabilities for stacking faults and twin boundaries, respectively. Warren defines  and as the proportion of atomic planes that are either faulted or forming twin boundaries.[16] Values of W(K), different for each reflection of an fcc crystal, are listed in Reference 14. For simplicity, we use a global parameter (T), defined as T  (1.5a  b)>a and, so 1  2hK D This equation gives a Warren-modified W-H (Warren W-H) plot. A dislocation-broadening correction can be introduced by changing the variables for the strain term:[17] K T # W (K) 

K T # W (K) 

1  f (KC 1/2 ) D

with f being a quadratic function of KC 1/2. When dislocations are the primary source of strain in a crystal, KC 1/2 is the 3454—VOLUME 36A, DECEMBER 2005

relevant parameter for the W-H plots. The parameters for KC 1/2 are reported in Reference 14 for copper. This equation is represented as the Warren–Ungár W-H plot. In practice, the broadening of each peak, after correction for instrumental broadening, is initially plotted on a classical W-H plot. If the data points do not form a straight line, the T parameter is adjusted in order to minimize the scatter. This approach is presented in Figures 3(a) and (b) for the NC-Cu powder. On the corresponding classical W-H plot, the data points are not aligned (Figure 3(a)), whereas on the Warren W-H plot, little or no scatter remains after adjustment of the T parameter (Figure 3(b)). It can be concluded that the broadening from stacking faults and twin boundaries is fully accounted for, and the crystallite size can be determined with precision. If significant scattering of data points persists on the Warren W-H plot (Figures 3(c) and (d)), the data are represented on a Warren–Ungár W-H plot, for which the dislocation contribution to broadening is included. The T parameter is readjusted until the data points can be fitted by a quadratic curve (Figure 4(a)). The crystallite size is then given by the intercept. In principle, the dislocation density could be obtained from the quadratic equation, but would require a full Warren–Averbach (W-A) analysis. Table II is a summary of the grain size evolution during processing. IV. MICROSTRUCTURAL EVOLUTION DURING PROCESSING A. NC-Cu Powder For a pure metal powder, the grain size will be defined as the diameter of the particles. This corresponds to the crystallite size obtained by XRD once the twin boundaries’ contribution has been removed. The value determined from the intercept on the Warren W-H plot (Figure 3(b)) is 40  3 nm. Due to the spherical shape of the particles, this value needs to be adjusted by the form factor of 4/3, giving 53  4 nm as the average size. The TEM investigations give a smaller average value of 40  10 nm, the larger error due to the more limited statistics. The cuprite oxide on the surface of the grains is 2 to 3 nm in thickness.[18] Concerning defects, the X-ray line-broadening analysis gives the best fit for a T parameter of 0.0187 and no dislocation contribution. The absence of dislocations is supported by TEM observations.[18] The slope of the W-H fit curve is negligible, as for a perfectly relaxed spherical powder. The value of , the twin probability, can be set equal to T/a because , the stacking-fault density, is zero. This assumption is justified by the fact that stacking faults are usually generated by deformation and dislocation motion in fcc metals and is confirmed by measuring  from W-A analysis on 111-222 and 200-400 sets of reflections for a NC-Cu powder obtained by inert-gas condensation.[19] Moreover, the stacking-fault energy of copper is around 50 mJ mol 1, which is sufficiently high to prevent the formation of such defects. The twin probability for our NC-Cu powder is then equal to 0.0052. In a monocrystalline sample, this would correspond to a twin every 40 nm, on average. However, in our case, due to the small grain size, the meaning of cannot be so easily interpreted. METALLURGICAL AND MATERIALS TRANSACTIONS A

Fig. 3—XRD line-broadening analysis including twin contribution: (a) and (c) classical W-H plots and (b) and (d) Warren W-H plots for (a) and (b) NC-Cu powder and (c) and (d) compacted NC-Cu powder. The scatter is significantly reduced for Warren W-H plots; refer to the text for parameters.

Fig. 4—XRD line-broadening analysis including dislocation contribution: Warren–Ungár W-H plots for (a) compacted NC-Cu powder, (b) sintered material, and (c) extruded material. For (c), the (220) reflection has been removed due to the crystallographic fiber texture. METALLURGICAL AND MATERIALS TRANSACTIONS A

VOLUME 36A, DECEMBER 2005—3455

Table II. Evolution of the Microstructure during the Processing Steps Determined by FEG-SEM, TEM, and XRD; Grain Size Given for Domains Surrounded by Different Types of GBs (cf. Figures 9 and 10)* Size of Domains Surrounded by Fabrication Step

Particle Size

Twin Density

General GBs

General GBs and Twins

General GBs, Twins, and Low-Angle GBs

Powder

0.68 pct







0.75 pct







Sintered

50  5 nm XRD-TEM 60  5 nm XRD-SEMTEM —

0.45 pct



0 pct*

100  10 nm TEM 100  10 nm TEM



Extruded

150  15 nm TEM 140  15 nm SEM-TEM

Compact

90  5 nm TEM-XRD

*Twins no longer in exact orientation relation due to extrinsic dislocations.

B. Compacted Powder A SEM micrograph representative of the structure of the compact is shown in Figure 5(a). The copper particles are loosely agglomerated, covered by small particles identified as cuprite, the relative density of the green body being around 70 pct. For the compacted powder, the grain size is defined as the crystallite size without the twin contribution. The SEM observations give a mean agglomerate size of 80  10 nm, estimated by cross-counting. The TEM observations are more accurate for the grain-size determination, even if the overlapping of the grains complicates counting. It was necessary to acquire micrographs of the same area with several different tilt orientations. Cross-counting then gives a mean grain size of 65  5 nm. Except for twinning, the grains are free of any substructure (Figure 5(b)). The XRD analysis results in a value of 0.020  0.005 for the T parameter, corresponding to a twin probability of  0.0055, which is close to the value for the powder. The Warren–Ungár W-H plot (Figure 4(a)) shows almost no scatter, in contrast to the best-fitting Warren W-H plot, meaning that dislocations are present in the material. Dislocations were not, however, observed by TEM. This might be explained by relaxing of the thin films, with dislocations emerging at the many free surfaces of the porous material, which is not the case for the XRD bulk samples. The fact that grains appear faceted on the TEM micrographs agrees with the idea that grains are deformed during compaction. Experimentally, the compaction pressure is 400 MPa, which is higher than the rearrangement threshold stress of the grains and, consequently, grains are subjected to cold deformation.[20] The grain size from XRD analysis after removing the twinning and stacking-fault contribution is 60  5 nm, slightly lower than the value from the TEM observations. This could be explained by an additional residual stress (diffraction-order independent) in the material that keeps the grain under elastic compression, slightly enlarging the width of the diffraction peaks. C. Sintered Material The transformations engendered during sintering are more significant than those produced during compaction. The material after sintering is now a polycrystalline metal of a density around 90 pct, with well defined GB. The size of domains surrounded by general GBs (i.e., high-angle GBs 3456—VOLUME 36A, DECEMBER 2005

Fig. 5—Evolution of microstructure during the processing steps: (a) FEGSEM micrograph of compacted NC-Cu powder, (b) the corresponding bright-field TEM micrograph, (c) FEG-SEM micrograph of sintered material (fracture surface), (d) the corresponding bright-field TEM micrograph, (e) FEG-SEM micrograph of extruded material, and ( f ) the corresponding bright-field TEM micrograph.

excluding twins), as well as grains surrounded by general boundaries and twins, will be monitored. The general GBs correspond, in our study, to high-angle GBs, excepting twins. The contrast between the compacted microstructure and the sintered microstructure is particularly striking from SEM observations. The micrographs of the fracture surface shown METALLURGICAL AND MATERIALS TRANSACTIONS A

in Figures 5(c) and (d) clearly show an equiaxed structure resulting from intergranular fracture. Observations of several areas reveal the residual porosity as “missing grains” in the microstructure, homogeneously present in the sample. The grain size obtained by delineating the GBs is an overestimation, because the fracture path was certainly along the largest grains and small grains are hardly visible at this observation scale. The mean value is 200 nm, with sizes between 60 and 700 nm. The TEM observations reveal the presence of smaller grains not visible on the SEM micrographs (Figure 5(d)). The mean size of grains surrounded by general boundaries is 150  15 nm. Including twins gives a grain size of 100  10 nm. The XRD analysis gives a value of 85  5 nm after removing the twin contribution. The difference between the two values is large, but might be explained by the following argument. The diffractometer used has an instrumental broadening that is equivalent to a grain size of 120 nm. Because of this, the broadening due to the largest grains of the distribution is partly masked in the XRD line profiles. Further work is, however, necessary to clarify this apparent inconsistency. The TEM micrographs reveal the presence of numerous twins with perfectly parallel boundaries due to grain growth. The XRD analysis results in a value of 0.0035  0.0005 for the twin probability . Similar values are obtained in NCCu or nickel with an equivalent grain size and containing a large number of twins.[21]

Fig. 6—FEG-SEM micrograph showing residual porosity of the extruded material (extrusion axis vertical). Pores enlarged by electrochemical etching.

D. Fully Densified Bulk NC-Cu The deformation of the NC-Cu during the final processing step of extrusion is relatively low and corresponds to an effective strain of about 0.6. The objective here is to reduce the porosity. Depending on the density of the sintered material, which is around 90 pct, the extrusion process can increase the density by 10 pct, reaching more than 99 pct of the theoretical density of copper. The level of porosity and its distribution in the matrix can be better estimated from SEM observations. (Refer, for example, to the micrograph of an axial cut of the NC-Cu rod shown in Figure 6). The grains can be seen to be slightly elongated along the extrusion axis. The pores are clearly visible, elongated along the extrusion axis similar to the grains. It should be noted that the pores are enlarged due to the chemical etching used to prepare the surface. When the density of the sintered material is high (90 pct), there is a radial distribution of pores after extrusion, the outer zone of the rod showing no pores at all. On the other hand, a less dense sintered material results in a residual porosity homogeneously distributed in the matrix after differential extrusion. This suggests that the pressure should be increased (both in the higher and lower part of the extrusion container) to make the extrusion stage more efficient in densifying the material. The grain-size distribution of domains delimited by general GBs can be provided by the polished and etched–surface SEM observations (Figure 7), the twin boundaries being invisible on the micrographs. This distribution measured from the images is lognormal-centered on 140  10 nm. The TEM dark-field micrographs reveal that the larger grains are often subdivided by low-angle GBs (Figure 8(a)) or twins (Figure 8(c)). The low-angle disorientations are typically less than 4 deg and agree with the number of METALLURGICAL AND MATERIALS TRANSACTIONS A

Fig. 7—Size distribution of the domains surrounded by general GBs in extruded NC-Cu material: histogram determined from FEG-SEM micrographs.

dislocations per unit length along the boundaries. The low-angle boundaries can either be curved (Figure 8(a)) or completely straight (Figure 8(b)). Figure 8(c) shows a darkfield micrograph where a twin relationship between the dark and bright parts of the grain has been identified by grain orientation using TEM diffraction. The twin boundaries contain, however, many extrinsic intergranular dislocations arranged in a nonperiodic network (Figure 8(d)). This results VOLUME 36A, DECEMBER 2005—3457

Fig. 8—Microstructure of extruded NC-Cu: dark-field TEM micrographs showing the following arrowed features: (1) curved low-angle GB, micrograph (a); (2) straight low-angle GB, micrograph (a); (3) twins, micrographs (a) through (c); (4) intragranular dislocation array perpendicular to twins, micrographs (a) and (c); and (5) pseudoperiodic array of intergranular dislocations in a general GB, micrograph (d).

Fig. 9—Sketch of a zone of the microstructure after extrusion. Gray solid lines are the GBs containing intergranular dislocations. Dotted lines are the low-angle GBs. Starred lines are the twin boundaries. Dark lines are the high-angle GBs.

in a misorientation from a perfect twin relationship of 2 to 4 deg, also in agreement with the dislocation-line density. We shall call these boundaries perturbed twins. This explains the very low value of 0.0005 found for the twin probability in the XRD analysis. Domains separated by perturbed twin boundaries appear as differently oriented crystallites, and the anisotropic contribution of the twins is no longer significant. We estimate the size of the domains surrounded by general GBs or perturbed twins to be 100  10 nm, as for the sintered material. The XRD analysis, which includes the contribution of low-angle GBs, provides a mean grain size of 90  5 nm, after dislocation correction. The TEM examination reveals that about half of the grains are completely free of dislocations (excepting those in lowangle GBs). When present, however, they are often observed perpendicular to the twin boundaries (Figures 8(a) and (c)). On the other hand, intergranular dislocations are frequently observed, both in general and twin boundaries. The occurrence of dislocations has been analyzed in part of the microstructure (Figure 9). Nearly periodic dislocation arrays appear in about 12 pct of the GBs examined. Preliminary analysis of the GB geometrical parameters shows that intergranular dislocations occur mainly in GBs close to coincidence orientations. Observed proportions of the respective types of GBs are reported in Table III. A study is in progress to increase the statistical accuracy of these parameters, using automated orientation measurements by convergent-beam electron diffraction in tandem with TEM. In twin boundaries, extrinsic dislocations are arranged as nonstrictly periodic networks, although they are periodically spaced in other GBs. It is recalled that extrinsic dislocations are intergranular defects that do not participate to the equilibrium GB structure and give rise to long-range stress fields. They result generally from the interaction of lattice dislocations with GBs. The return to equilibrium implies a rearrangement as a periodic network of intrinsic dislocations in near-coincidence GBs and their disappearance in general GBs. The accommodation of extrinsic GB dislocations involve several thermally activated processes, most of them being controlled by intergranular diffusion. In all cases, the presence of periodically spaced dislocations in numerous GBs implies that GBs are in an intermediate state of equilibrium.[22,23,24] An additional important point is that the mere presence of intergranular dislocations in general GBs proves that they are crystalline in nature. If an amorphous region existed at GBs, dislocations could not exist. From these different observations, we propose a representation of the microstructure in Figure 10: (a) domains of 140 nm subdivided by perturbed twin boundaries or low-angle GBs, with a resulting grain size of around 100 nm;

Table III. Characteristics of a Set of GBs Examined in the Extruded Material; Results are Presented in Terms of the Presence of Intergranular Dislocations and Respective Proportions of Large-Angle GBs, Low-Angle GBs, and Twin Boundaries Proportion of the Different GB Types Number of GBs Examined Proportion of GBs containing intergranular dislocations 3458—VOLUME 36A, DECEMBER 2005

118

General GBs

Twins

Low-Angle GBs

12 pct

91.5 pct

5 pct

3.5 pct

METALLURGICAL AND MATERIALS TRANSACTIONS A

Fig. 10—Schematic model of extruded NC-Cu microstructure: (a) population of grains free of internal structure; and (b) larger grains containing low-angle GBs (point 1), perturbed twins (point 2), and boarded by general GBs (point 3).

(b) small grains of 90 nm free of dislocations; (c) few dislocations in the matrix, often aligned in sequence perpendicular to boundaries; (d) high-angle GBs containing many intergranular dislocations; (e) crystalline GBs with no evidence of amorphous regions; and (f) residual porosity of 1 pct. This is a refinement of the structural model presented previously.[2] V. DISCUSSION The microstructural evolution in terms of grain size and occurrence of lattice and intergranular dislocations now needs to be discussed and compared to NC-Cu material obtained by other synthesis routes. The microstructure itself has been determined using different characterization techniques that will be comparatively discussed as a preamble. The characterization techniques used have differing sensibilities to structural features. For example, XRD is the most accurate for determining the domain size surrounded by any kind of GB (general, twin, or low-angle), except when a significant proportion of grains are too large compared with the experimental broadening. It is possible to eliminate the contribution due to perfect twins if necessary, but not that due to low-angle boundaries. The XRD, therefore, always leads to the smallest grain-size measurements. The FEG-SEM analysis of polished surfaces reveals the distribution of grains surrounded by general GBs, but not twins, perturbed twins, or low-angle boundaries. However, the distribution of grain sizes is available to this technique. The nature of GBs is best analyzed by TEM observations, but these lack the statistical degree of either XRD or SEM. From the TEM analysis, it is possible to estimate the grain size including any or all types of boundaries, thus revealing, for example, the main difference between the sintered and extruded material. The TEM technique is, of course, the only technique capable of analyzing the localization and distribution of dislocations. However, XRD analysis has proved valuable in confirming their presence in samples, notably for the compacted material. Twin density is best determined by XRD analysis, but METALLURGICAL AND MATERIALS TRANSACTIONS A

only for perfect twin configurations. Finally, TEM provides information on perturbed twins, while SEM is unusable in both cases. From microstructural investigations, it appears that the most significant grain growth occurs during the sintering step of the processing route, the grain size increasing from 50 nm to over 100 nm. While some growth is inevitable, as the density increases from 70 to 90 pct, the main part is likely to have occurred during the slow cooling of the sintered material. Recent experiments have shown that grain growth can be limited by accelerating the cooling of the sample. Between the sintered and extruded material, the most significant difference is not in the grain size per se, but in the nature of the GBs and grain morphology. The grains become elongated; however, for the moment, this phenomenon has not been fully quantified. During extrusion, the twin boundaries remain present but become loaded with extrinsic dislocations, thus leaving the perfect twin orientation. Evidence for dislocation activity during the extrusion stage is provided by both XRD analysis and direct TEM observations. Many more intergranular dislocations are observed, for example, between the sintered and extruded material. The main microstructural feature concerns the type and distribution of GBs. General high-angle boundaries, twins, and low-angle boundaries are all barriers to dislocation movement, but with differing relative strengths (References 25 through 27, respectively). Low-angle boundaries present the least resistance to dislocation motion. Twin boundaries present an efficient barrier for movement of lattice dislocations. They constitute hardening elements of the microstructure. More generally, the contribution of GBs to a macroscopic property roughly depends on the GB “character” distribution and also from their repartition in the polycrystal space. Twin GBs are less prone to accommodation of lattice dislocations, and they are expected to increase the yield stress at low temperatures. In the case of a predominance of GB sliding, high-angle GBs are expected to behave as preferential paths for sliding, allowing a high level of deformation if they dominate the microstructure. Thus, an interesting property should be expected in our microstructure that exhibits the highest density of high-angle GBs obtained to date. It is instructive to compare the microstructure of the material described here with that of material produced by other synthesis routes. The NC-Cu produced by electrodeposition, for example, has a microstructure composed of large domains surrounded by high-angle GBs of micron size. These domains are subdivided into much smaller grains of around 20 to 30 nm by low-angle GBs with less than 10 deg of misorientation.[7] The microstructure of NC-Cu produced by SPD is typically in the form of submicronic domains between 150 and 350 nm in size, surrounded by both large-angle and low-angle GBs, the later proportion ranging from 20 to 80 pct, depending on the synthesis route.[28] Taking low-angle GBs into account for all NC-Cu materials, our grain size is, therefore, larger than those of electrodeposited material and lower than those in a SPD deformed material. In so far as the ability to undergo high deformation is related to the presence of high-angle GBs, our NC-Cu displays a 100-nm grain size, which is significantly smaller than that for other NC-Cu materials. It is not surprising that the microstructure of SPD deformed material is different from that of to the NC-Cu produced by VOLUME 36A, DECEMBER 2005—3459

P/M—the only similar processing step is the final extrusion. This is to be compared with the very large accumulated strain needed to refine the grains during SPD techniques (for example, 10 for equal-channel angular extrusion[1]). Indeed, only one pass is applied in the processing, compared to the several passes needed in SPD, as the aim is to densify the material and not to refine the already nanometric grain size. The objective of the study was to characterize thoroughly the microstructure of bulk NC-Cu produced by P/M, so as to provide a firm basis for exploring the material’s interesting mechanical properties: near-perfect elastoplasticity, the absence of apparent necking, high yield stress, and enhanced ductility.[2] These properties show significant differences from materials produced by electrodeposition, SPD, or HPT, notably in the absence of sustained work hardening or necking. We believe these differences can be explained in terms of the smaller grain size and the lack of large specimens, particularly for tensile testing. In this respect, the most significant microstructural differences concern the grain size surrounded by general GBs, excluding low-angle GBs. It would seem that the grain size of domains surrounded by general GBs is the most significant parameter for mechanical properties, followed by the grain size including low-angle boundaries. However, additional work, especially postdeformation observations, has to be carried out before developing these arguments further. VI. CONCLUSIONS Bulk NC-Cu can be synthesized using P/M techniques with a mean grain size of less than 100 nm. The microstructure consists of high-angle GBs, perturbed twin-boundaries, and low-angle GBs. It is important in this respect to use a range of characterization techniques and to adapt XRD line-broadening analysis to include twin and dislocation contributions. Mechanical properties should be interpreted in consequence, notably when comparing with materials produced from other processing routes. The proportion of high-angle to low-angle boundaries is of prime importance, as they form very different barriers to the motion of dislocations. We suggest that the most relevant definition of grain size is with respect to general GBs. The XRD analysis always gives the very smallest grain size, which includes both general and low-angle GBs, and, therefore, may not be the most relevant parameter for understanding the mechanical properties of the material. As a final remark, each processing step gives a new material with its own microstructure and properties, and not only the final material can be of interest for industrial applications. Controlled porosity and nanostructured grains in a sample of large dimensions could find functional applications, for instance, in the medical field.

3460—VOLUME 36A, DECEMBER 2005

ACKNOWLEDGMENTS The authors thank Jean-Louis Pastol and Audrey Valette for help with the FEG-SEM observations and Jean-Claude Rouchaud for the ICP-OES analysis. Louisette Priester is acknowledged for fruitful discussions on the description of grain boundaries and their interactions with dislocations.

REFERENCES 1. R.Z. Valiev. I.V. Alexandrov, Y.T. Zhu, and T.C. Lowe: J. Mater. Res., 2002, vol. 17, pp. 5-8. 2. Y. Champion, C. Langlois, S. Guérin-Mailly, P. Langlois, J.-L. Bonnentien, and M.J. H ¨ytch: Science, 2003, vol. 300, pp. 310-11. 3. Y. Wang, M. Chen, F. Zhou, and E. Ma: Nature, 2002, vol. 419, pp. 912-15. 4. L. Lu, M.L. Sui, and K. Lu: Science, 2000, vol. 287, pp. 1463-66. 5. K.V. Ivanov, I.V. Ratochka, and Y.R. Kolobov: Nanostruct. Mater., 1999, vol. 12, pp. 947-50. 6. L. Lu, L.B. Wang, B.Z. Ding, and K. Lu: J. Mater. Res., 2000, vol. 15, pp. 270-73. 7. H. Jiang, Y.T. Zhu, D.P. Butt, I.V. Alexandrov, and T.C. Lowe: Mater. Sci. Eng. A, 2000, vol. 290, pp. 128-38. 8. S.X. McFadden, R.S. Mishra, R.Z. Valiev, A.P. Zhilyaev, and A.K. Mukherjee: Nature, 1999, vol. 398, pp. 684-86. 9. G.W. Nieman, J.R. Weertman, and R.W. Siegal: J. Mater. Res., 1991, vol. 6, pp. 1012-027. 10. P. Apte, B.H. Suits, and R.W. Siegel: Nanostruct. Mater., 1997, vol. 9, pp. 501-04. 11. X.J. Wu, L.G. Du, H.F. Zhang, J.F. Liu, Y.S. Zhou, Z.S. Li, L.Y. Xiong, and Y.L. Bai: Nanostruct. Mater., 1999, vol. 12, pp. 221-24. 12. Y. Champion and J. Bigot: J. Nanostruct. Mater., 1998, vol. 10, pp. 1097-110. 13. Y. Champion, F. Bernard, N. Guigue-Millot, and P. Perriat: Mater. Sci. Eng., 2003, vol. 260, pp. 258-66. 14. T. Ungár, S. Ott, P.G. Sanders, A. Borbély, and J.R. Weertman: Acta Mater., 1998, vol. 46, pp. 3693-99. 15. G.K. Williamson and W.H. Hall: Acta Metall., 1953, vol. 1, pp. 22-31. 16. B.E. Warren: in Progress in Metal Physics, Pergamon Publishing Corp., New York, NY, 1959, vol. 8, pp. 147-202. 17. T. Ungár: Mater. Sci. Eng. A, 2001, vols. 309–310, pp. 14-22. 18. Y. Champion and J. Bigot: Mater. Sci. Eng., 1996, vols. 217–218, pp. 58-63. 19. P.G. Sanders, A.B. Witney, J.R. Weertman, R.Z. Valiev, and R.W. Siegel: Mater. Sci. Eng., 1995, vol. 204, pp. 7-11. 20. O. Dominguez, Y. Champion, and J. Bigot: Proc. Annual Meeting of TMS, Austin, TX, 1996, D.L. Bourell, ed., TMS, Warrendale, PA, 1996. 21. G.W. Nieman, J.R. Weertman, and R.W. Siegel: Materials Research Society Symposia Proceedings, Materials Research Society, Pittsburgh, PA, 1991. 22. A.A. Nazarov, A.E. Romanov, and R.Z. Valiev: Scripta Metall. Mater., 1990, vol. 24, pp. 1984-94. 23. L. Priester: Mater. Sci. Eng. A, 2001, vol. 309–310, pp. 430-39. 24. S. Lartigue-Korinek and L. Priester: J. Am. Ceram. Soc., 1988, vol. 71, pp. 430-37. 25. A.W. Thompson: Acta Metall., 1975, vol. 23, pp. 1337-42. 26. S. Mahajan and G.Y. Chin: Acta Metall., 1973, vol. 21, pp. 173-79. 27. J.P. Hirth and J. Lothe: Theory of Dislocations, Wiley, New York, NY, 1982. 28. O.V. Mishin, D. Juul Jensen, and N. Hansen: Mater. Sci. Eng. A, 2003, vol. 342, pp. 320-28.

METALLURGICAL AND MATERIALS TRANSACTIONS A