Preparation and characterization of nanocrystalline copper powders (original) (raw)

PREPARATION AND CHARACTERIZATION OF NANOCRYSTALLINE COPPER POWDERS

Y. Champion and J. Bigot
CECM-CNRS 15 rue Georges Urbain 94407 Vitry Cedex, France

(Received April 3, 1995)

Introduction

Owing to potentially interesting properties linked to the high volume fraction of grain boundary atoms with respect to those with a bulk environment, nanocrystalline materials have attracted substantial interest [1]. Metallic nanocrystals have been produced using a gas condensation technique [2, 3, 4]. We have developed a different process (cryogenic melting) to produce such materials and this has been used to produce nanometric powders of Fe,Ni,Co,Cr\mathrm{Fe}, \mathrm{Ni}, \mathrm{Co}, \mathrm{Cr} and alloys thereof [5, 6, 7]. The essential advantage is that moderate quantities (about 50 g per hour) are produced. This is about the minimum required [8] to produce macroscopic compacts (through pressing and subsequent sintering) which are essential for the study of the mechanical properties of these materials.

Here, we report on the production of nanocrystalline Cu powders using the cryogenic melting technique. Copper has been chosen in order to subsequently compare its mechanical properties with those of the other metals prepared previously ( Fe,Ni,Co,Cr\mathrm{Fe}, \mathrm{Ni}, \mathrm{Co}, \mathrm{Cr} ) and also for its potential intrinsic mechanical properties and applications [9]. The powders were characterized using X-ray diffraction, transmission electron microscopy and X-ray photoelectron spectroscopy (XPS).

Experimental

The preparation technique consists of overheating molten metal in contact with a cryogenic liquid (nitrogen in the case of Cu powders). To avoid contamination through reaction with a crucible, the metal is heated in a r.f. levitation furnace. A turbulent calefaction layer, consisting of gas vapours from the cryogenic liquid and supersaturated metal vapour [7], is formed around the molten metal. The nanocrystalline particles would then be formed here by rapid condensation of the metallic vapours. The production of nanocrystalline powders by this technique depends critically on the vapour pressure of the overheated metal. Moreover, since r.f. heating is used, the electrical conductivity of the metal is crucial [7]. Since copper has a particularly high electrical and thermal conductivity at low temperatures, the temperature of the overheated melt can be substantially lowered through an excessive rate of arrival of solid copper feed stock: a stationary regime of powder production can be difficult to obtain. This is overcome by premelting the copper sample before filling with liquid nitrogen, allowing a high initial temperature of the molten metal to be reached. Subsequent overheating is thus maintained after filling the vessel with liquid nitrogen and this enables a constant feed stock regime to be obtained. Maintaining as stationary conditions as possi-

ble during powder production is essential, given the correlation between resulting particle size, melt overheating and vapour pressure [7].

The experiments were carried out with an initial 30 g copper load and with a feed rate of about 1.5 g per minute. The production of nanocrystalline Cu powders is estimated from weight loss to be about 60 g per hour and the yield is of order 75%75 \%. The loss can be essentially accounted for by the rather primitive means of recovery used (linen filters washed with pure hexane). In order to reduce their subsequent oxidation, the powders are stored in pure hexane. Electron microscopy was carried out using a TOPCON 002B microscope (working at 200 KV with a point resolution of 0.18 nm ) on the powders as-deposited on a carbon grid.

Morphology of the Powders

Figure 1 shows that the nanocrystalline Cu particles are nearly spherical and are aggregated. Aggregation has also been observed on nanocrystalline Fe powders prepared previously by this method. It has been suggested that this is related to the magnetic interactions between particles [7] and not necessarily associated with the process (such aggregation has also been observed for Fe particles as well as for other magnetic particles obtained by the gas condensation process [2]). Other work [10], from the measurement of the fractal dimensions of the aggregates, suggested that these were formed by a ballistic mechanism in the calefaction layer. In the case of copper, the formation of aggregates could be explained by this latter mechanism. In addition, some of the particles are clearly joined or even sintered. Sintering was seldom observed in the case of nanocrystalline Fe powders prepared by the same process [7], probably due to the fact that their sintering temperature is likely to be much higher than that of nanocrystalline Cu .

The particle size distribution in the powders was evaluated from such micrographs by manual counting on populations of about 150 particles. The histogram (figure 2) shows a nearly log-normal distribution shape, often observed as a size-distribution function [11]. The distribution is peaked at around 25 nm , the

Figure 1. Bright field transmission electron micrograph of the nanocrystalline Cu powder.

Figure 2. Particle size histogram corresponding to figure 1. A nearly log-normal distribution function peaked at 25 nm is obtained. The calculated average particle size is 35 nm .
mean size of particles is 35 nm with a standard deviation of 16 nm . The minimum size is of about 10 nm and the maximum is of 100 nm . The majority of particles ( 23%23 \% ) have a size of around 25 nm . This distribution is comparable to the distribution of nanocrystalline Fe powders [8].

Structure and Composition of the Particles

X-rays diffraction (figure 3) shows that nanocrystalline Cu is fcc, with parameters and relative intensities consistent with those of a reference copper specimen. A slight broadening of the Cu diffraction peaks is observed and is related to particle size. This broadening, using the Scherrer formula, gives an average size of the order of 25 nm , in agreement with the electron microscopy. Additional diffraction peaks are iden-

Figure 3. X-ray diffraction scan of the nanocrystalline Cu powder. The broadening of the copper peaks observed, corresponds to an average particle size of 35 nm . Broader Cu2O\mathrm{Cu}_{2} \mathrm{O} peaks with significant intensity are also visible.

Figure 4. XPS spectra on the as prepared nanocrystalline Cu powder (A) and on the same powder after ( Ar∗\mathrm{Ar}^{*} ) ion sputtering (B). The increasing of the Cu LMM peak from (A) to (B) indicates that the copper oxide phase is located at the surface of the particles. (The Ag 3 p peaks are due to the silver paint used to maintain the powder on the specimen holder.)

tified as due to Cu2O\mathrm{Cu}_{2} \mathrm{O}. Their relative intensities indicate the presence of a significant amount of this oxide phase, corresponding here to 20wt%20 \mathrm{wt} \%. The broadening of Cu2O\mathrm{Cu}_{2} \mathrm{O} peaks might be associated to a combination of small crystallite size and of lattice strains in the oxide. Only Cu2O\mathrm{Cu}_{2} \mathrm{O} was observed ( CuO has not been detected).

XPS analyses were carried out on the as-prepared powders (figure 4). According to Wagner [12], the Auger parameter measured from the experimental spectra is close to the value for Cu2O\mathrm{Cu}_{2} \mathrm{O}. The slight deviation of this parameter might be related to non stoechiometry. In addition, the comparison between the shape of the 2 p Cu peak and a reference spectrum [13] indicates that CuO is not present in the sample. XPS experiments were also carried out on the same powders after ( Ar∗\mathrm{Ar}^{*} ) ion sputtering. The resulting spectrum and the one for the as-prepared powders are compared in figure 4. Sputtering of the particle surfaces leads to a significant increase of the metallic copper contribution to the peak at 570 eV . These first experiments are not able to provide quantitative data as the thickness of the layer around the particles, but suggest that Cu2O\mathrm{Cu}_{2} \mathrm{O} is located mostly at the particle surfaces, as could be expected.

The high resolution micrograph in figure 5 shows a typical ( 35 nm ) spherical particle. Examination of the image contrast shows that this is made up of distinct phases: the centre is metallic copper, oriented here along ⟨110⟩\langle 110\rangle. This metallic kernel is surrounded by the oxide phase (Cu2O)\left(\mathrm{Cu}_{2} \mathrm{O}\right) which appears to be strained due to the accommodation of the particle curvature. Finally, the whole particle is surrounded by an amorphous layer which could be either some kind of copper oxide or carbon compound. Some regions of the oxide layer are in epitaxial condition with the copper lattice with a cube-on-cube orientation relationship. The oxide lattice adapts a nearly spherical shape at the interface with the amorphous phase, by an arrangement of large and small facets. The large facets lie along the densest (111) plane of the structure

Figure 5. High resolution image of a 35 nm size particle. The centre of the particle is metallic copper oriented along ⟨110⟩\langle 110\rangle, and the lines crossing the particle are associated to {111}\{111\} twin boundaries. The copper kernel is surrounded by a Cu2O\mathrm{Cu}_{2} \mathrm{O} layer, relatively strained and partially in epitaxy with the copper lattice. The oxide phase is faceted following large (A) and small (B) facets. The whole particle is surrounded by an amorphous layer giving it an approximately spherical shape.

(arrow A in the figure 5). The small facets (arrow B in the figure 5) lie also along dense planes and are associated with ledges to accommodate the curvature between two adjacent large facets. Finally, the match between this faceted surface and the external spherical surface is achieved by the amorphous layer of variable thickness.

Thickness of oxide can be measured from the HREM micrograph figure 6 (here it corresponds to the lighter lattice fringes close to the edge; a weak Fresnel fringe is also visible), and varies as a function of curvature and/or crystalline orientation. It ranges from nearly 3 nm for the greatest curvature to below 1 nm on the flatter portions of the particle. On the other hand, the amorphous layer has a roughly constant thickness of about 2 nm . Recent results using high resolution electron microscopy show the same type structure for nanocrystalline Fe [14] and Ni [15] particles, both prepared by the evaporation technique. The case of Ni powders is very similar to copper, with a faceted 2 nm thick NiO layer around metallic nickel, surrounded by an amorphous carbonate hydroxide. These nanoparticles are thus readily oxidised, for both methods of preparation; this oxide layer may passivate the particles, hence avoiding total oxidation. Such layers will have to be removed for the properties of purely metallic single phase compacts to be assessed.

The high resolution images show that planar defects are present in both single (figure 5) and sintered particles (figure 6). These defects are believed to be twin boundaries since they are parallel to the (111) planes and the contrast level remain the same on both sides. Figure 5 and also the lower magnification image (figure 1) show that such boundaries start at the edge and cross the particle. In the particular case of figure 6 , five of these planar defects are simultaneously present and connected at a five-fold junction.

Figure 6. HREM micrograph of two sintered 55 nm size particles. The sharp contrast variation indicates that particles are surrounded by a 2.5 nm thick Cu2O\mathrm{Cu}_{2} \mathrm{O} layer and an outer 2.5 nm thick amorphous layer. Linear contrasts inside the particle are associated with {111}\{111\} twin boundaries meeting at a five fold junction (D).

The angle between two successive planes is about 72∘72^{\circ} which is consistent with an arrangement of five successive {111}\{111\} twin ( 70.53∘70.53^{\circ} misoriented) boundaries locked by a 7.5∘7.5^{\circ}-edge disclination. The disclination line is located at the five-fold junction ( D on the figure 6 ). Disclinations are believed to be more stable than dislocations within nanometric materials [16]. This may be due to their strain field which is parabolically dependent on the distance from the disclination line, whilst the strain field around a dislocations line follows a logarithmic law. As a result, such disclinations would have to transform into dislocations with grain coarsening.

Conclusion

We have shown that the cryogenic melting technique is suitable for the preparation of nanocrystalline Cu powders with a significant production rate of 60 g per hour. The as-prepared powders are characterized by a narrow size distribution centred on 25 nm . The particles are superficially oxidised with a Cu2O\mathrm{Cu}_{2} \mathrm{O} layer of average thickness of 2.5 nm around a metallic copper kernel. The oxide phase is strained to accommodate the curvature of the particle surface, faceted and partially in epitaxial relation with the copper lattice. Finally, the particles are covered with an approximately 2.5 nm thick amorphous phase, which has yet to be identified. Planar defects associated to twins are observed. Such twinning, seldom observed in “microcrystalline” copper and more generally in fcc metals, has been related to nanoscale structural effects.

The process must be improved in order to reduce oxidation (although subsequent powders now contain only 5wt.%5 \mathrm{wt} . \% of Cu2O\mathrm{Cu}_{2} \mathrm{O} ) and sintering during formation of the particles. This is crucial for the subsequent study of the compaction and the sintering of the powders to produce dense metallic specimens.

Acknowledgements

We would like to thank O. Dominguez for fruitful discussion, J-L. Bonnentien for assistance in the preparation of the powders. The XPS spectra were obtained with the help of M-G. Barthés-Labrousse and L. Minel. We would also like to thank J-P. Chevalier for critically reading the manuscript.

References

1. L.E. Brus and R.W. Siegel, J. Mater. Res. 4705 (1989).
2. K. Kimoto, Y. Kamiya, M. Nonoyama and R. Uyeda, Jpn J. Appl. Phys. 2702 (1963).
3. S. Yatsuya, S. Kasukabe and R. Uyeda, Jpn J. Appl. Phys. 121675 (1973).
4. S. Yatsuya, Y. Tsukasaki, K. Mihama and R. Uyeda, J. Cryst. Growth 45490 (1978).
5. J. Bigot, A.G. Goursat, G. Vernet, J.F. Rimbert, J. Foulard and T. Sarle, French patent n ∘{ }^{\circ} Z 8307414, (1983).
6. Y.H. Zhou, Thesis, University Paris 6, (1990).
7. J. Bigot, Ann. Chim. Fr. 18117 (1993).
8. O. Dominguez, M. Phillippot and J. Bigot, Scr. Metall. Mater. 3213 (1994).
9. R.Z. Valiev, E.V. Kozlov, Yu.F. Yvanov, J. Lian, A.A. Nazarov and B. Baudelet, Acta Metall. Mater. 422467 (1994).
10. M. Tence, J.P. Chevalier and R. Jullien, J. Physique 471989 (1986).
11. C.G. Granqvist and R.A. Buhrman, J. Appl. Phys. 472200 (1976).
12. C.D. Wagner, Appendix 5 in Vol. 1 of Practical Surface Analysis (2nd ed.), ed. D. Briggs and M.P. Seah, pub. John Wiley and Sons, Chichester (1990).
13. M.P. Seab, Surf. Interface Anal. 2222 (1980).
14. C.M. Hsu, H.M. Lin and K.R. Tsai, J. Appl. Phys. 764793 (1994).
15. T. Uchikoshi, Y. Sakka, M. Yoshitake and K. Yoshihara, Nanostructured Materials. 4199 (1994).
16. R.C. Pond, “Line defects in solids” in Dislocations in solids, ed. F.R.N. Nabarro, Vol. 8, pub. Elsevier Science, Amsterdam (1989).