Carbon Reactions

4 downloads 0 Views 1MB Size Report
reaction is the graphitization of the amorphous carbon, at temperatures well below ... transmission electron microscopy, metal–carbon reactions, graphitization of ...
Microscopy Microanalysis

Microsc. Microanal. 8, 288–304, 2002 DOI: 10.1017.S1431927602020226

AND

© MICROSCOPY SOCIETY OF AMERICA 2002

In Situ TEM Studies of Metal–Carbon Reactions Robert Sinclair,1, * Toshio Itoh,1,2 and Richard Chin 1 1

Department of Materials Science and Engineering, Stanford University, 416 Escondido Mall, Building 550, Stanford, CA 94305-2205, USA 2 Applied Materials Inc., Santa Clara, CA 95054, USA

Abstract: The reactions which occur between amorphous carbon and a number of first transition metals ~Ti, Cr, Fe, Co, Ni, and Cu! have been studied by transmission electron microscopy ~TEM!. The materials are in thin-film form with the metal layer sandwiched between thicker carbon layers. In four cases, the predominant reaction is the graphitization of the amorphous carbon, at temperatures well below 8008C. This is brought about by the elements themselves in the case of Co and Ni, and by metastable carbides in the case of Fe ~Fe3C! and Cr ~Cr3C22x !. The Ti–C and Cu–C systems do not exhibit graphitization. For the former, only TiC is produced up to 10008C, while the carbon does not react at all with copper. In situ TEM studies show the mechanism to be of the dissolution-precipitation type, which is equivalent to the metal-mediated crystallization process for amorphous silicon and germanium. The heat of graphitization is found to be 18–19 kcal/mol-C by differential scanning calorimetry. Key words: In situ transmission electron microscopy, metal–carbon reactions, graphitization of carbon, nickel–carbon system, iron–carbon system, chromium–carbon system

I NTR ODUCTION Metal–carbon systems are important both practically and scientifically. For instance, metal carbides have extensive engineering applications as hard coatings or as strengthening precipitates in steels. On the other hand, metal–carbon multilayers are employed for X-ray reflectors and focusing devices ~Barbee, 1987!. Accordingly, study of the reactions which can take place between the elements has fundamental significance. In this article, we consider carbon-transition metal layered thin films fabricated by physical vapor deposition. Carbon prepared this way is amorphous while the metal Received March 25, 2002; accepted April 8, 2002. *Corresponding author

usually adopts its equilibrium crystal structure. Thus, not only is there the possibility of carbide formation, as either stable or metastable phases, but also the graphitization of the carbon or the solid-state amorphization of the mixture. A variety of metals is reported, ranging from those with strong carbide forming tendency ~e.g., Ti! to those with no equilibrium carbide ~e.g., Ni!. It will be seen that carbon graphitization is a prominent reaction when carbon is in excess, with increasing difficulty as the metal carbide itself becomes more stable. Transmission electron microscopy ~TEM; Thomas, 1962! is ideally suited to characterize the reaction behavior. Microstructural evolution is readily determined by bright and dark field imaging, interface reactions are elucidated by high resolution microscopy ~HREM!, and phase identification can be carried out by selected area diffraction analysis

In Situ TEM Studies of Metal–Carbon Reactions

~SADP!. Furthermore, in situ heating studies provide an elegant and fairly rapid means of surveying the types of reactions which take place.

B ACKGR OUND Carbon is a group IV element which can be produced in many forms ~Pierson, 1993!. Of direct relevance here are graphite or graphitic carbon ~with predominantly sp 2 bonding!, and diamond or diamond-like amorphous carbon ~with predominantly sp 3 bonding!. The former represent the more stable phases at ambient temperature and pressure. In contrast, the group IV elements below carbon in the periodic table, Si and Ge, exist only with tetrahedral sp 3 bonding: The amorphous state is normally created by vapor deposition at room temperature, while the diamond cubic structure represents the stable phase. Transformation of an amorphous phase to its inherently more stable crystalline counterpart generally requires a thermally assisted kinetic process involving nucleation and growth. For Si and Ge, this can take place at about 550–6008C. However, it is known that such crystallization temperatures can be significantly reduced in the presence of some metals, particularly those which have a eutectic phase diagram with Si and Ge ~e.g., see Konno and Sinclair, 1994a, for a review!. From in situ HREM observations ~Sinclair and Konno, 1994; Konno and Sinclair, 1995a, 1995b!, it was shown that the dominant mechanism first involves dissolution of the amorphous element into the metal phase, thereby supersaturating it with respect to the crystalline diamond cubic phase. Subsequent precipitation of the latter inside the metal, followed by atomic diffusion from the amorphous phase to the crystalline nuclei, brings about the reaction at lower temperatures than can occur in the elemental matrix itself, a mechanism we termed metal-mediated crystallization ~MMC!. Amorphous carbon is even more difficult to “crystallize,” or rather “graphitize,” often involving temperatures in the range 1700–30008C ~Franklin, 1954; Pierson, 1993!. ~Diamond formation of course requires high temperatures and excessive pressures.! However, the catalytic nature of metals in assisting the transformation is also well known ~Fitzer and Kegel, 1968; Gillot et al., 1968; Marsh and Warburton, 1970; Fischbach, 1971; Oya and Marsh, 1982! although temperatures more than 15008C are still typical. The most suggested mechanisms are a carbon dissolution-

289

precipitation process, essentially equivalent to our MMC, and a carbide formation-disintegration sequence whereby a metal carbide is formed as an intermediary. It is also interesting to note that diamond ~Strong, 1963!, filamentary carbon ~Baker and Harris, 1978!, and carbon nanotube ~Iijima and Ichihashi, 1993! formation are also significantly promoted in the presence of metals. However, direct interpretation of the mechanism is difficult to achieve because of the high temperatures or complex atmospheric conditions involved. In an extension of his work on MMC of Si and Ge, Konno also examined a simple metal–carbon system ~Co–C! to see whether or not graphitization was amenable to the same type of investigation ~Konno and Sinclair, 1995c!. Indeed, employing an a-C/Co/a-C trilayer film, he found that there was a strong exothermic calorimetric peak at about 5508C, decreasing in temperature as the metal layer thickness was increased. In situ TEM observations and ex situ microstructural determination showed a striking resemblance to the behavior of the MMC of Si and Ge. It was concluded that an equivalent mechanism was taking place, involving metallic cobalt only and not its metastable carbide, Co2C ~Hofer and Peebles, 1947; Konno and Sinclair, 1994b!. Furthermore, subsequent work on the influence of the cobalt alloy magnetic layer in computer hard discs on its protective diamond-like carbon coating demonstrated that graphitization takes place at about 6008C, which has severe consequences for the stability of such important technological structures ~Ramirez et al., 1999!. The present study was therefore undertaken to establish whether the MMC mechanism is more widely applicable to metal–carbon reactions, utilizing TEM and in situ studies as the primary investigative technique. A range of first-transition series metals was chosen, which now includes Ti, Cr, Fe, Co, Ni, and Cu.

E XPERIMENTAL P R OCEDUR ES The materials in this study were prepared in thin-film form by sputter deposition at room temperature. The targets were the element themselves, of at least 99.9% purity, with 3 mtorr argon sputtering gas in a chamber with 5 31028 torr background pressure. A 3-in. DC magnetron gun was employed for the carbon deposition, at a rate of about 0.03 nm/s, while 2-in. guns were used for the metals ~typically 0.07 nm/s, as determined by standard rate runs!. A specially

290

Robert Sinclair et al.

Figure 1. Schematic structure of the trilayer thin films used in this study; free-standing films for through-foil TEM samples ~a! and films deposited on oxidized silicon substrate for cross-sectional TEM samples ~b!.

designed magnetic gun was used for the ferromagnetic elements Fe, Co, and Ni. Two configurations were fabricated, as shown schematically in Figure 1. For through-foil TEM examination, 20 nm carbon/10 nm metal/20 nm carbon trilayers were prepared on glass substrates coated with photoresist. For crosssectional viewing, both in situ and ex situ, 100 nm carbon/ 20 nm metal/100 nm carbon trilayers were deposited onto oxidized silicon wafers. Substrate sizes were usually 25 mm square. The through-foil samples were floated off from their substrates by simply dissolving in acetone and were then supported by 75-mesh “oyster-shell” type copper grids. Crosssectioning was achieved by the Bravman–Sinclair method ~Bravman and Sinclair, 1984!, or modifications thereof, ensuring that the Gatan G1 bond epoxy was as thin as possible. In addition, some multilayer samples were also made, consisting of 10 periods of 5 nm carbon/5 nm metal. The through-foil specimens are useful for overall viewing of the microstructural changes and phase identification by diffraction analysis, while the cross sections are important for observing the interface reactions and the mechanisms of graphitization. TEM was carried out using a Philips EM430ST microscope, operating at 300 kV with a resolution of 0.19 nm or better. Heating was achieved in situ with a Philips single tilt heating holder and ex situ in a radiant heating vacuum furnace. The microstructures produced by both types of annealing experiments were remarkably similar for purport-

edly identical conditions of time and temperature ~unless noted!, fulfilling, therefore, one of our major criteria for checking the veracity of the in situ experiment ~Sinclair et al., 1988; Ko and Sinclair, 1994!. Furthermore, as the reactions of interest must take place primarily at the metal– carbon interface, the through-foil configuration essentially represents the bulk behavior, as there is no interference from the surfaces. Accordingly, structures and phase development observed from the cross section samples were also compared with the through-foil material to ensure reproducibility of the observations. Finally, the reaction temperatures were also found to be consistent with those indicated by exothermic peaks in differential scanning calorimetry ~DSC! scans on the multilayer samples. The procedure for the in situ experiment occasionally employed continuous video-recording, but in most cases followed a step-wise annealing sequence. Samples were annealed, under observation, for 30 min at temperature intervals of 508C starting between 2508C and 3008C. Images and diffraction patterns were obtained at temperature over the next 10 min, whereupon the temperature was raised at about 208C/min to the next level, taking care not to overshoot. All observations described here involving annealing above 7008C were taken from samples treated ex situ. Because of the good thermal conductivity of silicon, the crosssectional observations were uniform across the specimen. However, the free-standing through-foil films displayed noticeable temperature gradients and so TEM was always carried out within 10 mm of the copper grid bars. Differential scanning calorimetry was performed in a Dupont 910 DSC in an argon atmosphere, utilizing freestanding multilayer samples. Although the mass was relatively small ~e.g., 0.6 mg!, reasonable results were obtained, with the scan from a second run subtracted from the first to remove the background.

R ESULTS Cobalt–Carbon The results for this system ~which has a eutectic form! have been published in full elsewhere ~Konno and Sinclair, 1994b, 1995c!, but a brief summary is relevant here. An amorphous carbon/cobalt trilayer exhibits all the main features of MMC. Thus, an exothermic DSC peak occurs on heating between 500–6008C, with decreasing peak temperature as the metal

In Situ TEM Studies of Metal–Carbon Reactions

Figure 2. Differential scanning calorimetry scan of a nickelamorphous carbon multilayer film taken at a heating rate of 308C/minute. The exothermic peak at about 5408C is quite clear.

layer thickness increases. TEM examination after annealing under these conditions shows that the cobalt appears to “move through” the carbon, leaving a graphitic structure behind. The calorimetric peak is measured to be 19 6 2 kcal/mol-C, compared to that for amorphous silicon crystallization of 12 kcal/mol-C. This is believed to be the heat of graphitization of carbon. It is thought to be its first known measurement as, without the metal, the transformation temperatures are too high for most DSC apparatus. Meanwhile, for cobalt-rich cosputtered ~i.e., homogeneous! Co–C thin films, a metastable carbide with composition close to Co2C is formed at about 2508C ~Hofer and Peebles, 1947; Clarke and Jack, 1951; Nagakura, 1961; Konno and Sinclair, 1994b!, decomposing to cobalt and graphite above about 4008C.

Nickel–Carbon The equilibrium nickel–carbon phase diagram is also eutectic, and there have been reports of nickel inducing graphitization either in molten form ~Fitzer and Kegel, 1968! or at 9508C in thin foils ~Derbyshire et al., 1972!. Our preliminary study showed that a reaction occurs at much lower temperatures ~Itoh and Sinclair, 1994!. For instance, differential scanning calorimetry on the multilayer ~Fig. 2! exhibits an exothermic peak between 5008C and 6008C, with a heat of reaction of 18 kJ/mol-C.

291

In situ TEM shows that the reaction is somewhat more complex than simple graphitization. For instance, heating to 2758C for 30 min produces some rather large singlecrystal grains within the matrix of small nickel crystallites ~e.g., Fig. 3!. The selected area diffraction pattern is consistent with that of the known metastable carbide, Ni3C ~Nagakura, 1957!. These crystals eventually comprise up to about 10% of the observed areas, but rather than continuing to grow upon heating, they disappear and decompose by 3258C back to face-centered cubic nickel and amorphous carbon ~e.g., Fig. 4!. By DSC and ex situ observations on a Ni-10%C alloy, Nishitani et al. ~1985! had concluded that the metastable carbide decomposed into nickel and graphite above 5008C, but the in situ experiment clearly shows that this is not the case here. A further interesting observation is that the high resolution image of the carbon itself, well out of contact with the nickel, shows subtle but definite changes in appearance ~Fig. 5!. The normal HREM amorphous image is still random, but shows many regions in which the image is layered over distances of about 1 nm, rather like a nano-crystalline graphite. This was also observed in our study of diamond-like carbon on cobalt alloy magnetic media ~Ramirez et al., 1999!. The carbon itself shows no further changes up to 9008C annealing, and our interpretation is that it represents only the initial stages of graphitization, or ordering, of the carbon material. Presumably, both the Ni3C formation and the short-range ordering of the carbon do not have enthalpies sufficiently high to be noticed during the DSC scans. Annealing the free-standing film up to 5008C brings about some nickel grain growth and the appearance of large areas of lighter shading, which become even more distinct by 5508C ~e.g., Fig. 6!. At this stage, graphite diffraction rings are also observed. However, the cross-sectional views, especially in situ, reveal more clearly the nature of the reaction ~e.g., Fig. 7!. The nickel film becomes broken up into individual particles which gradually migrate through the carbon film. The carbon is amorphous ahead of the nickel grains, and is graphitized behind them. Highresolution imaging shows that the graphite basal planes are “contoured” on the nickel particles from which they emerge ~e.g., Fig. 8! and so the phase which is produced is technically not crystalline graphite, which would have regular translational symmetry of the carbon atoms within the basal plane. Accordingly, the product is best described as graphitized carbon, rather than graphite itself. Continued annealing brings about further transformation of the amorphous carbon, which is virtually complete by the end of the

292

Robert Sinclair et al.

Figure 3. Through-foil bright-field image ~a!, selected area diffraction pattern ~b!, and dark-field image ~c!, using a diffraction spot of Ni3C and part of the polycrystalline Ni ring, of a free-standing film annealed in situ at 2758C for 30 min. The larger grain is Ni3C, while the smaller ones are nickel.

5508C annealing treatment, consistent with the DSC exothermic peak. Thus, the dissolution-precipitation mechanism is substantiated by our observations. In summary therefore, nickel also achieves graphitization of amorphous carbon at a similar temperature to, and with an equivalent enthalpy of reaction as, that induced by cobalt. However, part of the nickel film is converted to its metastable carbide Ni3C, which decomposes before graphitization occurs and so plays no role in the latter reaction.

Iron–Carbon In the field of metallurgy, the iron–carbon system is perhaps the most important technologically. The iron carbide phase, cementite ~Fe3C!, is also metastable, but it is so strongly formed in steels that its presence is always recog-

nized in the phase diagram ~Massalski et al., 1986!. The behavior of Fe/C layered thin films is then clearly of particular interest ~Itoh and Sinclair, 1995!. Increasing the temperature of the free-standing film brings about a gradual transformation of the alpha-iron ~ferrite! layer to cementite. Thus at 3008C, many new diffraction spots are present in the SADP, and some grains are larger than in the as-deposited condition. Further increases in temperature produce more diffraction spots and decreasing intensity of the BCC iron rings ~e.g., Fig. 9!, so that by 5008C, only the former are present. As the polycrystalline rings become more complete, it is confirmed that they match those expected for cementite. Plan view images at 5008C show large white “islands,” whose only diffraction effect is that of amorphous carbon. Cross-sectional views confirm that these are voids in the metal film, adjacent to

In Situ TEM Studies of Metal–Carbon Reactions

293

Figure 4. Through-foil dark-field image and selected area diffraction pattern of the free-standing film annealed in situ for 30 min at 2758C ~a!, at 3008C ~b!, and at 3258C ~c!. The dark-field images were formed using a diffraction spot of Ni3C and part of the nickel ~200! ring. Gradual dissolution of the Ni3C is revealed.

which some graphite is formed ~e.g., Fig. 10!. The crosssectional view shows that the metal film is otherwise intact at this temperature, but thicker by about 10% compared to the elemental iron. At this stage, however, the majority of the carbon is still amorphous. Raising the temperature further to 5508C brings about rapid changes, which were recorded by more frequent photographs and by continuous video-recordings. The metal layer splits into two, and eventually into individual particles which then migrate through the amorphous carbon ~Fig. 11!. Behind these migrating particles remains graphite, the basal planes of which are parallel to the contours of the metal. Microdiffraction and high resolution images ~e.g., Fig. 12! show that the metal phase participating in this behavior is cementite itself. Thus, unlike the

Co–C and Ni–C systems, the amorphous carbon is graphitized by the iron carbide and not by the elemental phase. This is remarkably similar to the Pd–Si and Ni–Si systems, whereby amorphous silicon is crystallized by a metalmediated process utilizing a silicide phase @Pd2Si and NiSi2 , respectively ~Lau and van der Weg, 1978; Hayzelden and Batstone, 1993!#. This might be then termed carbidemediated graphitization. By the time the temperature is raised to 6008C, the graphitization process is complete. Microdiffraction shows that occasionally a few alpha-iron crystals are present. Whether these result from the decomposition of cementite or remain from the initially unconverted iron layer is not completely clear, but as no a-Fe was found at 5008C or 5508C, the former seems the more likely possibility.

294

Robert Sinclair et al.

Figure 5. Cross-sectional high resolution image of an a-C/Ni/a-C trilayer film as deposited ~a! and annealed in situ at 3008C for 30 min ~b!. The changing character of the a-C structure is revealed, as is a small amount of graphitization at the Ni–C interface.

Figure 6. Through-foil bright-field image ~a!, selected area diffraction pattern ~b!, and dark-field image ~c!, using the nickel ~111! diffraction ring, of the a-C/Ni/a-C free-standing film annealed in situ at 5508C for 30 min. Bright areas correspond to gaps in the previously continuous metal layer.

Chromium–Carbon The Cr–C system is also interesting metallurgically, as chromium carbides act as strengthening precipitates in steels. There are three stable carbide phases @Cr23C6 , Cr7C3 , and

Cr3C2 ~Massalski et al., 1986!# and at least two known metastable ones @Cr2C ~Lux and Eberle, 1961; Zaslavskays et al., 1972! and Cr3C ~Inoue and Masumoto, 1979!#. There is no complete account of the reactions between Cr and C, and so the present work is of basic interest also. As will be

In Situ TEM Studies of Metal–Carbon Reactions

295

Figure 7. Cross-sectional bright-field image ~a!, selected area diffraction pattern ~b!, dark-field image ~c!, using the graphite ~0002! diffraction ring, and dark-field image ~d!, using the nickel ~111! diffraction ring, of the a-C/Ni/a-C trilayer film annealed in situ at 5008C for 30 min. The migration of the nickel grains into the amorphous carbon is captured, leaving graphitic carbon behind.

seen, the sequence of events is even more complex than those of the other systems, and so a more complete description will be given elsewhere. First, the as-deposited trilayer is noticeably different from the other systems, particularly with regard to the polycrystalline metal layer, the grain boundaries of which are much more distinct. The cross-sectional view shows that this arises from incursion of the carbon extensively down them. Furthermore, a thin amorphous layer is present at both Cr–C interfaces ~e.g., Fig. 13!, presumably due to intermixing as in solid-state amorphization processes. @Note that Cr–C fulfills the major requirements of the latter, having a negative heat of mixing and one fast diffusing element ~Johnson, 1986!.# As-deposited metal–silicon interfaces ~e.g., Holloway et al., 1989a, 1989b! show similar behavior. Upon annealing in the temperature range 300–5008C, it appeared at first sight that these amorphous layers grew, as

in the “classical” solid-state amorphization process ~e.g., Fig. 14!. However, high-resolution imaging in through-foil and cross section, and selected area diffraction from the former, show that in fact a “nano-crystalline” phase is produced, with grain sizes of approximately 1–2 nm in diameter. The diffraction rings match those of the metastable hexagonal phase Cr2C, which is also produced by tempering high chromium content steels around 5008C ~Irvine et al., 1960; Zaslavskays et al., 1972!. As can be seen, the metal film remains intact at this stage. This situation prevails up to 6508C, apart from the sharpening of the Cr2C diffraction rings due to grain growth. However, after about 20 min annealing at 6508C, large acicular grains emerge, the diffraction patterns of which show they are largely single crystal ~e.g., Fig. 15!. These consume the whole metal film by 80 min annealing. Through the use of over 20 single crystal diffraction patterns, it was

296

Robert Sinclair et al.

Figure 16 are so similar to those found in the other systems that it is reasonable to conclude that an equivalent mechanism is taking place. The only carbide present is still the Cr3C22x phase, with no evidence of other phases. Likewise, high-resolution imaging ~Fig. 17! also reveals the relationship of the graphite basal planes with the contours of the particle from which they emerged. Thus the Cr–C system also displays the features of carbide-mediated carbon crystallization, employing metastable Cr3C22x .

Titanium–Carbon

Figure 8. Cross-sectional high-resolution image of the a-C/Ni/a-C trilayer film annealed in situ at 5008C for 30 min, showing the graphite layer behind a migrating nickel grain.

found that there was no uniquely good match to any of the stable or metastable phases mentioned earlier. However, all patterns could be indexed satisfactorily according to the structural data recently reported for a further orthorhombic metastable phase designated as Cr3C22x ~Bouzy et al., 1991, 1993!. This phase was discovered after annealing cosputtered carbon-rich amorphous Cr–C alloys ~Bouzy et al., 1991!. It is interesting that its precursor in our study is a nano-crystalline metastable phase, which must also be highly disordered due to the large grain boundary area. Annealing the cross section samples in situ did not reproduce this behavior, although ex situ annealing did. Instead, Cr7C3 crystals ~one of the equilibrium carbides! are produced throughout, or on, the carbon layer between 5008C and 6008C. It appears that a surface reaction is both kinetically and thermodynamically favorable here. Because of this disparity, the in situ experiment is regarded as unreliable in this system for annealing higher than 5008C and so those results were rejected. Both though-foil and cross section ex situ samples annealed at 8008C and 9008C displayed evidence of graphite formation. Indeed, the cross section micrographs shown in

TiC is one of the most stable carbides, having a melting point over 30008C. Accordingly, it is relevant to see if the trend of increasing difficulty ~i.e., temperature! of graphitization is followed in this system. Trilayer films were made as before. It is of some interest that the as-deposited Ti–C interfaces did not show any vestiges of solid-state amorphization mixing, because the Ti–Si system is very prominent in this regard ~Holloway and Sinclair, 1987!. The rock salt structure titanium carbide replaced the titanium layer upon annealing at 4008C and no further changes were detected up to 10008C, which was the maximum temperature tested in this study ~e.g., Fig. 18!. Presumably, the diffusion of carbon is too slow even at these temperatures to bring about the graphitization process by a carbide-mediated mechanism. It should be noted that the amorphous carbon layer showed no transformation within itself even at 10008C.

Copper–Carbon Copper–carbon has no known carbide phases and the solubility of C in solid Cu is extremely low. DSC measurements carried out on multilayers comprising 10 layers of 5 nm Cu/5 nm C showed no exothermic peaks up to 6008C, even though copper is a lower-melting-point ~i.e., higher diffusivity! metal than cobalt or nickel. In situ heating studies of the multilayer also showed no graphitization reaction, but rather the copper tended to agglomerate as heating progressed. Accordingly, it seems that copper does not fit into the behavior of the other metals in that it does not bring about graphitization in the solid state, even at temperatures whereby carbon atomic diffusion in the metal lattice is expected to be fairly rapid. Whether this is due to the low solubility of C in Cu or arises from an interface effect is not clear at present, as this system was not studied further.

In Situ TEM Studies of Metal–Carbon Reactions

297

Figure 9. Through-foil bright-field images and associated selected area diffraction patterns of an a-C/Fe/a-C free-standing film annealed in situ at 3508C ~a!, 4008C ~b!, and 4508C ~c!. The emergence of a new phase is revealed, the diffraction pattern of which is consistent with that of cementite ~Fe3C!.

D ISCUSSION It has been found in this research that the amorphous carbon produced by sputter deposition can be converted to a graphitic form at abnormally low temperatures by a number of transition metals. No such behavior occurs until at least 10008C for the carbon by itself. The microstructures

developed during the transformation are remarkably similar to those occurring during metal-mediated crystallization of amorphous group IV elemental semiconductors, and reactions followed in situ also show striking similarity to that process ~Sinclair and Konno, 1994!. It is believed, therefore, that they are equivalent: Thus carbon dissolves into the metal by atomic diffusion, supersaturating it with respect to the more stable “crystalline” phase due to the

298

Robert Sinclair et al.

Figure 10. High-resolution image of an a-C/Fe/a-C trilayer film annealed in situ at 5008C. Graphite is seen to be present occasionally at the boundary between the cementite and voids in the metal layer.

higher free energy of the amorphous state. “Crystalline” nucleation can subsequently take place, with growth of the crystals brought about by diffusion through the metal lattice, which is more rapid than the direct rearrangement of the atoms in the amorphous network. In the case of cobalt and nickel, the graphitization takes place between 5008C and 6008C, with the face-centered cubic elemental metals being the catalytic phases. Metastable carbides can be formed, but do not take part in the reaction. Thus, Ni3C is produced at lower temperatures ~e.g., 3508C!, but decomposes back into nickel and carbon before the transformation. Co2C is only formed from amorphous Co–C alloys. The heat of reaction was found, by differential scanning calorimetry, to be about 18–19 kcal/ mol-C. It is thought that this is the first direct determination of the heat of graphitization, since the direct transformation takes place at too high a temperature for calorimetric analysis. @It should be noted that the crystallization enthalpies determined by our MMC experiments on Si and Ge match those for the “pure” elements themselves ~Konno and Sinclair, 1995a, 1995b!.# The basal planes of the graphite tend to follow the contours of the metal particles from which they emerge, and so there is no long-range translational symmetry within these graphite planes. The phase produced is therefore not technically “crystalline graphite,” but the graphitic nature is most definitely present. As the carbide-forming tendency of the metals becomes more dominant, the graphitization mechanism be-

Figure 11. Cross-sectional bright field images of the a-C/Fe/a-C trilayer film annealed in situ at 5508C for 5 mins ~a!, 10 mins ~b!, and 30 mins ~c!, revealing the progression of cementite particles through the carbon, creating graphite in their wake.

comes more complex. Thus, in the Fe–C system, alpha-iron is converted to cementite ~Fe3C! at temperatures up to 5008C, with little change in the layered structure. However, as the temperature is raised further, the cementite particles appear to move through the amorphous carbon, as carbon atoms are diffusing through this phase, from the amorphous layer to the growing graphitic phase. Thus, cementite is the relevant medium for the process, rather than iron itself. This is similar to the role that silicides play in the Pd–Si and Ni–Si systems ~Lau and van der Weg, 1978; Hayzelden and Batstone, 1993!. It should be noted that although cementite is formally a metastable phase in the Fe–C system, its tendency to form ~as in steels! must be sufficiently strong that it is produced in the present thinfilm reactions.

In Situ TEM Studies of Metal–Carbon Reactions

299

Figure 12. High-resolution image of an a-C/Fe/a-C trilayer film annealed in situ at 5508C showing the relationship between the graphitized carbon basal planes and the cementite particle from which they emerged.

Figure 14. Cross-sectional high-resolution images of an a-C/Cr/a-C trilayer film annealed in situ at 3008C ~a!, 4008C ~b!, and 5008C ~c! for 30 min, showing the Cr layer transformation to extremely fine-grained Cr2C.

Figure 13. Cross-sectional high-resolution image of an as-deposited a-C/Cr/a-C trilayer film, showing carbon incursion down the Cr grain boundaries and solid-state amorphization at the Cr–C interfaces.

Chromium has an even stronger carbide-forming tendency, with three equilibrium carbide phases and several metastable ones. The initial reaction is to produce an extremely fine-grained ~e.g., 1–2 nm or so! metastable carbide

~Cr2C! by about 5008C. This then converts to a second metastable carbide at about 6508C ~Cr3C22x ! with still no graphitization. The metal-mediated reaction then takes place above 7508C, with the latter metastable phase bringing it about. Neither decomposition nor transformation of Cr3C22x to a stable carbide takes place up to 9008C. Finally in two systems, no graphitization was encountered. In the Ti–C system, the extremely stable phase TiC was produced at about 4008C, but it achieved no graphitization up to 10008C, the highest temperature studied here. Presumably this arises from the very slow diffusion rates in this high melting temperature compound. Indeed, using known diffusion data for TiC, it is estimated that a temperature of approximately 15008C would be required to bring about equivalent changes to those, say, in the Ni–C system. On the other hand, copper ~which like cobalt and nickel has no stable carbides! also does not bring about graphitization

300

Robert Sinclair et al.

Figure 15. Through-foil bright-field image ~a!, selected area diffraction pattern ~b!, and dark-field image ~c!, using the Cr3C22x ~040! diffraction spot, for an a-C/Cr/a-C free-standing film annealed at 6508C for 30 min. The selected area diffraction pattern reveals the newly formed phase to be metastable Cr3C22x .

up to 6008C, even in Cu–C multilayers. We interpret this as probably arising from low solubility of carbon in copper. It may be noticed that alpha-iron, which also has low carbon solubility, must be converted to cementite before graphitization takes place: As there is no copper carbide, this is not a possibility in the Cu–C system. Although we did not make formal measurements of the diffusion kinetics, it is clear that the graphitization process, if it occurs as we describe, should be consistent with known diffusional behavior. In other words, the rate of growth of the graphite particles is directly proportional to the flux of carbon atoms across the thin metal medium. Konno analyzed this for the Al–Si, Ag–Si, and Ag–Ge systems ~Konno and Sinclair, 1992, 1995a, 1995b!. He found good agreement with a simplified one-dimensional diffusion model for the first two cases, but discrepancies for the third. Preliminary calculations for the present systems ~Itoh, 1996! are similar. Thus, diffusional estimates for Ni–C and Cr–C are consistent with the observed growth rates and temperatures. However for Fe–C, the calculated growth rates are about three orders of magnitude too slow. Since this system should be the

most well established, it would be somewhat surprising, but not completely unlikely, that the diffusion coefficients for carbon in cementite were unreliable. Also, there is the possibility of an interface-controlled reaction contributing to the kinetic process. While the solution to this difficulty is clearly of interest, it was beyond the scope of the present work. Because TEM was used as the primary investigative tool, it is also worthwhile considering its role in these experiments. First, it is notable that the results obtained by in situ observations compare favorably with those from ex situ annealed bulk samples in almost all cases, and are consistent with the calorimetric data. There appear, for instance, to be no electron-beam effects in these materials. Second, the through-foil specimens provide direct observation of the “bulk” behavior because the reaction always initiates at the metal–carbon interfaces. Their strength lies in the wide areas for imaging observation and the ability of selected area diffraction analysis for phase identification. Third, the cross-sectional views yield significant information about the role of interfaces, the exact location of graphite formation, and indeed about the mechanism of the

In Situ TEM Studies of Metal–Carbon Reactions

301

Figure 16. Cross-sectional bright-field images of a-C/Cr/a-C trilayer films annealed ex situ at 8008C ~a! and 9008C ~b! showing the graphitization of the amorphous carbon by migration of Cr3C22x particles.

reaction itself. Thus, the combination of through-foil and cross section TEM is especially powerful. Finally, the in situ experiment, as has been pointed out before ~Sinclair et al., 1988!, yields unique insight into the behavior, and because one is continually observing the same area, does not miss any stage in the reaction. It also provides a rapid survey of the phenomena, by gradually ramping up the temperature. Equivalent information from ex situ annealing alone, followed by individual sample preparation, would involve significantly larger investments in time and might miss key aspects of the process. Therefore, in situ TEM has provided fascinating new data on an important and interesting topic, namely the graphitization of amorphous carbon, as brought about at abnormally low temperatures, by transition metals.

C ONCLUSIONS 1. Elemental nickel, like elemental cobalt, brings about the graphitization of amorphous carbon thin films at temperatures between 5008C and 6008C, with an enthalpy of

Figure 17. Cross-sectional high-resolution image of an a-C/Cr/ a-C trilayer film annealed ex situ at 8008C showing the emergence of graphite from the metastable Cr3C22x phase.

reaction about 18 kcal/mol-C. A metastable phase Ni3C is partially formed at about 3008C, decomposing prior to the graphitization process. 2. Iron reacts with carbon at temperatures between 3008C and 5008C to produce cementite ~Fe3C!, its metastable carbide. The cementite layer is largely continuous but shows some agglomeration. Raising the temperature further, up to 5508C, brings about graphitization of the remaining amorphous carbon, employing cementite itself. 3. Chromium reacts with amorphous carbon at 3508– 5008C to produce metastable Cr2C, which transforms to a second metastable phase Cr3C22x at about 6508C. Above 7508C, the latter phase brings about the graphitization of amorphous carbon. 4. Titanium reacts with amorphous carbon to form TiC at about 4008C, with no further changes up to 10008C, the maximum temperature studied here.

302

Robert Sinclair et al.

Figure 18. Through-foil bright-field image ~a!, selected area diffraction pattern ~b!, and dark field image ~c!, using the TiC ~111! ring, of an a-C ~40 nm!/Ti ~10 nm!/a-C ~40 nm! free-standing film annealed ex situ at 9008C for 30 min. No transformation of the amorphous carbon is detected.

5. There is no reaction observed between elemental copper and amorphous carbon up to 6008C. 6. In situ transmission electron microscopy shows that the graphitization behavior is consistent with a dissolutionprecipitation process, similar to the metal-mediated crystallization behavior of amorphous silicon and germanium.

scientific community have been profoundly influenced by his leadership in the application of transmission electron microscopy. This research is supported by the Department of Energy, Basic Energy Sciences Division under Grant No. DE-FG03-99ER45786.

A CKNOWLEDGMENTS

R EFER ENCES

It is a distinct pleasure and privilege to dedicate this work in honor of Professor Gareth Thomas’ 70th birthday. One of us ~R.S.! has benefited directly from his guidance, inspiration, and friendship. Countless others in the international

Baker, R.T.K. & Harris, P.S. ~1978!. The formation of filamentous carbon. In Chemistry and Physics of Carbon, Walker, P.L. & Thrower, P.A. ~Eds.!, Vol. 14, pp. 83–165. New York: Marcel Dekker.

In Situ TEM Studies of Metal–Carbon Reactions

Barbee, T.W., Jr. ~1987!. Multilayer structures: Atomic engineering in its infancy. In Physics, Fabrication, and Applications of Multilayered Structures, Dhez, P. & Weisbuch, C. ~Eds.!, pp. 17– 32. New York, London: Plenum Press. Bouzy, E., Bauer-Grosse, E. & Le Caer, G. ~1993!. NaCl and filled Re 3B-type structures for two metastable chromium carbides. Philos Mag B 68, 619–638. Bouzy, E., Le Caer, G. & Bauer-Grosse, E. ~1991!. Pseudotenfold twins in a new metastable chromium carbide and associated growth morphologies. Philos Mag Lett 64, 1–6. Bravman, J.C. & Sinclair, R. ~1984!. The preparation of crosssection specimens for transmission electron microscopy. J Electron Microsc Tech 1, 53–61. Clarke, J. & Jack, K.H. ~1951!. The preparation and the crystal structures of cobalt nitride, Co2N, of cobalt carbonitrides, Co2 ~C,N!, and of cobalt carbide. Chem Ind 17, 1004–1005. Derbyshire, F.J., Presland, A.E.B. & Trimm, D.L. ~1972!. The formation of graphite films by precipitation of carbon from nickel foils. Carbon 10, 114–115. Fischbach, D.B. ~1971!. The kinetics and mechanism of graphitization. In Chemistry and Physics of Carbon, Walker, P.L. ~Ed.!, Vol. 7. pp. 1–105. New York: Marcel Dekker. Fitzer, E. & Kegel, B. ~1968!. Reactions of carbon saturated vanadium carbide with disordered carbon ~Effects on catalytic graphitization!. Carbon 6, 433–446. Franklin, R.E. ~1954!. Crystallite growth in graphitizing and non-graphitizing carbons. Proc R Soc London A 209, 196–218. Gillot, J., Lux, B., Cornuault, P. & du Chaffaut, F. ~1968!. Catalysis of the graphitization of a furfuryl alcohol coke by titanium and vanadium. Ber Deut Keram Ges 45, 224–230. Hayzelden, C. & Batstone, J.L. ~1993!. Silicide formation and silicide-mediated crystallization of nickel-implanted amorphous silicon thin films. J Appl Phys 73, 8279–8289. Hofer, L.J.E. & Peebles, W.C. ~1947!. Preparation and X-ray diffraction studies of a new cobalt carbide. J Am Chem Soc 69, 893–899. Holloway, K. & Sinclair, R. ~1987!. Amorphous Ti–Si alloy formed by interdiffusion of amorphous Si and crystalline Ti multilayers. J Appl Phys 61, 1359–1364. Holloway, K., Do, K.B. & Sinclair, R. ~1989!. Interfacial reactions on annealing molybdenum-silicon multilayers. J Appl Phys 65, 474–480. Holloway, K., Sinclair, R. & Nathan, M. ~1989!. Amorphous silicide formation by thermal reaction: A comparison of several metal-silicon systems. J Vac Sci Technol A 7, 1479–1483. Iijima, S. & Ichihashi, T. ~1993!. Single-shell carbon nanotubes of 1-nm diameter. Nature 363, 603–615. Inoue, A. & Masumoto, T. ~1979!. Formation of nonequilibrium Cr3C carbide in Cr-C binary alloys quenched rapidly from the melt. Scripta Met 13, 711–715. Irvine, K.J., Crowe, D.J. & Pickering, F.B. ~1960!. The physical metallurgy of 12% chromium steels. J Iron Steel Inst 195, 386.

303

Itoh, T. & Sinclair, R. ~1994!. Nickel mediated transformation of amorphous carbon to graphite. Mater Res Soc Symp Proc 349, 31–36. Itoh, T. & Sinclair, R. ~1995!. In situ TEM study of reactions in iron/amorphous carbon layered thin films. Mater Res Soc Symp Proc 382, 45–50. Itoh, T. ~1996!. Study of Reactions in Transition Metal/Amorphous Carbon Layered Thin Films Using Transmission Electron Microscopy. Ph.D. Dissertation, Stanford University, pp. 123–134. Johnson, W.L. ~1986!. Thermodynamic and kinetic aspects of the crystal to glass transformation in metallic materials. Prog Mater Sci 30, 81–134. Ko, D.H. & Sinclair, R. ~1994!. In situ, dynamic high resolution transmission electron microscopy: Application to Pt/GaAs interfacial reactions. Ultramicroscopy 54, 166–178. Konno, T.J. & Sinclair, R. ~1992!. Crystallization of silicon in aluminum/amorphous-silicon multilayers. Philos Mag B 66, 729–765. Konno, T.J. & Sinclair, R. ~1994a!. Metal-contact induced crystallization of semiconductors. Mater Sci Eng A A179/A180, 426–432. Konno, T.J. & Sinclair R. ~1994b!. Crystallization of co-sputtered amorphous cobalt-carbon alloys. Acta Metall Mater 42, 1231–1247. Konno, T.J. & Sinclair, R. ~1995a!. Metal-mediated crystallization of amorphous silicon in silicon-silver layered systems. Philos Mag B 71, 163–178. Konno, T.J. & Sinclair, R. ~1995b!. Metal-mediated crystallization of amorphous germanium in germanium-silver layered systems. Philos Mag B 71, 179–199. Konno, T.J. & Sinclair, R. ~1995c! Crystallization of amorphous carbon in carbon-cobalt layered thin films. Acta Metall Mater 42, 471–484. Lau, S.S. & van der Weg, W.F. ~1978!. Solid phase epitaxy. In Thin Films—Interdiffusion and Reactions, Poate, J.M., Tu, K.N. & Mayer, J.W. ~Eds.!, pp. 433–480. Chichester, UK: Wiley. Lux, H. & Eberle, L. ~1961!. Zur Kenntnis der Chrom~II!-Salze und des Chrom~II!-oxids, III. Chem Ber 94, 1562–1571. Marsh, H. & Warburton, A.P. ~1970!. Catalysis of graphitisation. J Appl Chem 20, 133–142. Massalski, T.B., Murray, J.L, Bennett, L.H. & Baker, H., Eds. ~1986!. Binary Alloy Phase Diagrams. Metals Park, OH: American Society for Metals. Nagakura, S. ~1957!. Study of metallic carbides by electron diffraction. Part I. Formation and decomposition of nickel carbide. J Phys Soc Jpn 12, 482–494. Nagakura, S. ~1961!. Study of metallic carbides by electron diffraction. Part IV. Cobalt carbides. J Phys Soc Jpn 16, 1213– 1219. Nishitani, S.R., Ishihara, K.N., Suzuki, R.O. & Shingu, P.H. ~1985!. Metastable solid solubility limit of carbon in the Ni-C system. J Mater Sci Lett 4, 872–875.

304

Robert Sinclair et al.

Oya, A. & Marsh, H. ~1982!. Phenomena of catalytic graphitization. J Mater Sci 17, 309–322. Pierson, H.O. ~1993!. Handbook of Carbon, Graphite, Diamond and Fullerenes. Park Ridge, NJ: Noyes Publications. Ramirez, A.G., Itoh, T. & Sinclair, R. ~1999!. Crystallization of amorphous carbon thin films in the presence of magnetic media. J Appl Phys 85, 1508–1513. Sinclair, R. & Konno, T.J. ~1994!. In situ HREM: Application to metal-mediated crystallization. Ultramicroscopy 56, 225–232. Sinclair, R., Yamashita, T., Parker, M.A., Kim, K.B., Holloway, K. & Schwartzman, A.F. ~1988!. The development of in situ

high resolution electron microscopy. Acta Crystallogr A 44, 965–975. Strong, H.M. ~1963!. Catalytic effects in transformation of graphite to diamond. J Chem Phys 39, 2057–2062. Thomas, G. ~1962!. Transmission Electron Microscopy of Metals. New York: Wiley. Zaslavskays, L.V., Ivanova, N.V. & Lashko, N.F. ~1972!. Separation of alloyed chromium carbides of the Cr2C ~metastable! and Cr23C6 or Cr7C3 types isolated from steels and alloys. In Chemical Properties and Analysis of Refractory Compounds, Samsonov, G.V. ~Ed.!, pp. 68–71. New York: Consultants Bureau.