Journal of Applied Physics 93, 4207 (2003); https://doi.org/10.1063/1.1558227 93, 4207

© 2003 American Institute of Physics.

Nucleation, growth, and graphitization ofdiamond nanocrystals during chlorination ofcarbidesCite as: Journal of Applied Physics 93, 4207 (2003); https://doi.org/10.1063/1.1558227Submitted: 04 September 2002 . Accepted: 14 January 2003 . Published Online: 21 March 2003

Sascha Welz, Yury Gogotsi, and Michael J. McNallan

ARTICLES YOU MAY BE INTERESTED IN

Induced nucleation of diamond powderApplied Physics Letters 59, 546 (1991); https://doi.org/10.1063/1.105434

Transformation of nanodiamond into carbon onions: A comparative study by high-resolutiontransmission electron microscopy, electron energy-loss spectroscopy, x-ray diffraction, small-angle x-ray scattering, and ultraviolet Raman spectroscopyJournal of Applied Physics 97, 074302 (2005); https://doi.org/10.1063/1.1868054

Diamond-Graphite Equilibrium Line from Growth and Graphitization of DiamondThe Journal of Chemical Physics 35, 383 (1961); https://doi.org/10.1063/1.1731938

Nucleation, growth, and graphitization of diamond nanocrystalsduring chlorination of carbides

Sascha WelzDepartment of Civil and Material Engineering, University of Illinois at Chicago, 842 West Taylor Street,Chicago, Illinois 60607

Yury Gogotsia)

Department of Materials Engineering, Drexel University, 3141 Chestnut Street, Philadelphia,Pennsylvania 19104 and Department of Mechanical and Industrial Engineering,University of Illinois at Chicago, 842 West Taylor Street, Chicago, Illinois 60607

Michael J. McNallanDepartment of Mechanical and Industrial Engineering, University of Illinois at Chicago,842 West Taylor Street, Chicago, Illinois 60607

~Received 4 September 2002; accepted 14 January 2003!

Synthesis of nano- and microcrystallinesp3-bonded carbon~diamond! with cubic and hexagonalstructure by extraction of silicon from silicon carbide in chlorine-containing gases has been reportedrecently. This process is attractive because it can produce diamond at ambient pressure andtemperatures below 1000 °C. No plasma or other high-energy activation is required, thus providingan opportunity for large-scale synthesis. However, the mechanism of diamond formation has notbeen previously analyzed. This work reports on the formation mechanisms of diamond as well as thetransformation of diamond to graphite and onionlike carbon upon heating. Study of SiC/carboninterfaces showed that direct epitaxial growth of diamond on SiC is possible, in agreement withprevious molecular-dynamics simulation. However, random nucleation of diamond from amorphoussp3-bonded carbon produced as the result of extraction of Si from SiC has also been demonstrated.It has been shown that the presence of hydrogen in the environment is not required for diamondsynthesis. However, hydrogen can stabilize the nanocrystals and lead to the growth of thick diamondlayers. If no hydrogen is added, diamond nanocrystals transform to graphite, forming carbon onionsand other curved graphitic nanostructures. ©2003 American Institute of Physics.@DOI: 10.1063/1.1558227#

I. INTRODUCTION

Diamond is one of the most interesting carbon poly-morphs due to its extreme hardness and thermal conductivityas well as other useful properties. Synthetic diamond is com-mercially produced using chemical vapor deposition~CVD!,1,2 shock-wave,3 or high-pressure processes.4 Meth-ods for diamond synthesis which include use of explosivemixtures,5 plasma treatment,6 fullerene precursors,7,8 or highpressures continue to appear. Nanocrystalline diamond inparticular has received attention recently,8–13 partially due toa general interest to nanotechnology and the exploration ofnanoscale structures. It can be produced by CVD, shock-wave, and other techniques and its structure and propertieshave been studied fairly well.

What is commonly referred to as ‘‘diamond’’ is appli-cable to cubic 3C diamond having the Fd3m structure. Nano-crystalline diamond samples can contain a variety of othersp3 bonded structures along with 3C diamond. Hexagonal~2H, 4H, 6H, 8H, and 10H! and rhombohedral~15R and21R! diamond phases have been theoretically predicted14–16

as well as found in nature~‘‘lonsdaleite’’!17 and produced inthe laboratory.18,19 4H, 6H, and 8H structures have been ob-served experimentally,19–22 although they are energetically

less favorable with respect to their parent structures 2C and2H. Rhombohedral polytypes such as 9R and 15R have alsobeen observed.23,24 The so-calledn-diamond org-carbon,25

belonging to the F4̄3m space group26 completes the group ofknown diamond phases, although some othersp3-bondedcarbon structures have been theoretically predicted.

A different approach to the production of carbon coat-ings, including diamond films on silicon carbide~SiC!, wasreported recently.27,28 It has been previously shown that se-lective etching of carbides is an attractive technique for syn-thesis of carbide derived carbon~CDC! coatings.29 Super-critical water30 or halogens can be used to remove metalsfrom carbides producing such carbon coatings. The CDCcoatings formed by selective etching in halogens include avariety of carbon phases depending on experimentalconditions.27,28,31Since SiCl4 is much more thermodynami-cally stable than CCl4 , chlorine or HCl react selectively withthe silicon at SiC surfaces by the reaction

SiC12Cl25SiCl41C, ~1!

SiC14HCl5SiCl41C12H2, ~2!

leaving carbon behind on the SiC substrate.27 The process ofCDC formation is schematically shown in Fig. 1. Chlorina-tion of SiC in the presence of hydrogen in the gas mixtureleads to a stable conversion of SiC to diamond with theaverage crystallite size of 5–10 nm.28 Thick and thin coat-

a!Author to whom all correspondence should be addressed; electronic mail:gogotsi@drexel.edu

JOURNAL OF APPLIED PHYSICS VOLUME 93, NUMBER 7 1 APRIL 2003

42070021-8979/2003/93(7)/4207/8/$20.00 © 2003 American Institute of Physics

ings or polycrystalline powders can be prepared using thismethod. The linear reaction kinetics32 allow transformationto any depth, until the complete SiC particle or component isconverted to carbon. In order for the chlorination reaction toproceed, two molecules of Cl2(g) or four molecules ofHCl~g! must be transported to the SiC/C interface and onemolecule of SiCl4(g) and two molecules of hydrogen mustbe transported away from the interface for each atom of sili-con extracted and carbon produced@Eqs.~1! and~2!#. Linearkinetics imply that the carbon film is nanoporous and allowseasy permeation of Cl2 , HCl, H2 , and SiCl4 molecules.Nanocrystalline diamond-structured CDC coatings demon-strate high-hardness values and Young’s modulus28 as well asexcellent tribological properties.33

Initially, we reported diamond formation from SiC in thepresence of hydrogen during chlorination.27,28 However,large nanocrystalline areas producing characteristic diamondfringing or diamond microcrystals were found in some re-gions of SiC samples treated in pure chlorine with no hydro-gen. This is similar to the diamond synthesis fromfullerene,7,8,34 which showed that hydrogen is not essentialfor the growth of diamond outside its range of thermody-namic stability.

This article describes the mechanism of formation andevolution of the nanocrystalline diamond~nanodiamond! inCDC produced by extraction of metals from carbides.

II. EXPERIMENT

A. Synthesis

The apparatus used to investigate the interactions of SiCwith halogens at atmospheric pressure has been describedelsewhere.27 Synthesis experiments were performed usingcommercially available sintereda-SiC ~Hexoloy!. Thesesamples were sectioned into disks 16 mm in diameter and 1mm thick. The disks were cleaned ultrasonically, rinsed inacetone, and placed in a quartz sample holder. The SiC

samples were exposed to flowing gas mixtures of 1%–3.5%Cl2 , 0%–2% H2, and balance Ar, which was used as a car-rier gas. Experiments were conducted at 1000 °C with thetypical exposure time of 24 h. At the end of each experimen-tal run, the furnace and reaction gas mixture were securedand an argon purge was initiated through the reaction cham-ber during the cool down period, as described in Ref. 32.Titanium carbide~TiC! powder was used in some experi-ments to compare the structure of CDC produced from dif-ferent carbides. TiC samples were chlorinated for 2 h at800 °C in Ar-3.5%Cl2 .

B. Analysis

TEM samples were cut from CDC-coated sintered SiCsamples to the size that covers a 2-mm hole in a standard3-mm diameter Cu washer. They were mechanically groundon lapping films using a Tri-Pod Polisher to the final speci-men thickness of;4 mm. After that, Ar ion milling with anangle of;8° relative to the surface and a rotation angle of60° perpendicular to the interface was applied until a holeappeared in the center of the sample and the edges of thehole were electron transparent. The prepared specimens wereanalyzed using JEOL JEM-3010~300 kV! and JEOL JEM-2010F ~200 kV! transmission electron microscopes~TEM!.Energy dispersive x-ray spectroscopy~EDS! was used toidentify the carbon areas that were free from impurities andshowed only traces of silicon and chlorine, and electron en-ergy loss spectroscopy was used to identify areas of diamondfrom the carbon edge shape. Subsequently, selected areaelectron diffraction~SAD! was performed on about one-micron-sized nanocrystalline areas and microcrystals, andconvergent-beam electron diffraction~CBED! was per-formed with a 5 and 10-nm electron probe.

III. RESULTS AND DISCUSSION

A. Mechanism of conversion of SiC to diamond

CDC coatings produced with a chlorine to hydrogen ra-tio of about 2:1 had high hardness and Young’s modulus.28

TEM shows that these coatings were built of nanocrystalswith an average size of 5 nm~Fig. 2!. Lattice fringing, SAD,CBED, and EELS confirm the formation of diamond-structured carbon in this layer. Lattice fringes and diffractionspots at;0.193 and;0.218 nm, as well as characteristicCBED images, suggest the formation of 2H hexagonal dia-mond ~lonsdaleite! along with cubic diamond. Lonsdaleitehas been often observed in nanocrystalline diamond filmsproduced by CVD,19,35and accompanied SiC in some naturalsamples.36

Recent studies have shown that it is not difficult to formnanometer-sized diamonds.34 Moreover, several groups re-ported nucleation ofsp3-bonded carbon and nanocrystallinediamond after surface treatment of SiC by fluorocarbonplasma37 and bombardment with hydrogen38 or carbonions.39 Thus, there is consistent evidence of conversion ofcarbides into diamond after the removal of metal atoms fromthe carbide lattice under various experimental conditions.According to the original concept for metastable growth ofdiamond suggested by Spitsyn,40 it is necessary to conserve

FIG. 1. Schematic showing the difference between CDC coating~a! andconventional deposition techniques~b!. The starting material is SiC or an-other metal carbide. The process takes place in a fused silica tube furnaceunder ambient pressure at temperatures starting from 200 °C and typicallynot exceeding 1200 °C. The surface of SiC converts into CDC by etching ina halogen-containing gas mixture. The shape of the sample stays unchangedand CDC coatings show excellent adhesion to the substrate.

4208 J. Appl. Phys., Vol. 93, No. 7, 1 April 2003 Welz, Gogotsi, and McNallan

the orientational effect of the surface carbon atoms and touse carbon-containing molecules withsp3 bonding that canbe attached to the diamond surface in a complementary man-ner. Both conditions can be satisfied when Si is extractedfrom SiC, forming carbon atoms in thesp3 hybridization. Itis worth of mentioning that the very first diamond growthexperiments of Derjagin and Spitsyn were conducted in thecarbon-halogen systems using CBr4 and CI4 .40

Several groups have predicted that diamond may bemore stable at the nanoscale than graphite. A simple expla-nation for this is the high-surface energy of graphite edgescompared to diamond. While most predictions were done forroom-temperature conditions, some researchers calculatedthe critical size of diamond crystals for high temperatures aswell. We plotted data from Gamarnik41 in Fig. 3~a! as a func-tion of temperature. According to this prediction, diamondnanocrystals formed in the temperature range of interest forCDC synthesis~below 1200 °C) should be less than 5 nm insize and the maximum size of stable diamond crystals shouldnot exceed 10 nm at room temperature. Statistical analysisdone using TEM micrographs of the diamonds in CDC coat-ings produced on SiC samples chlorinated at 1000 °C showsthree populations of crystals@Fig. 3~b!#. The first one has themaximum at 5 nm@the theoretically predicted critical crystalsize for this temperature is 4.8 nm~Ref. 41!# with the num-ber of crystals larger than 5 nm being very small and goingto zero above 10 nm. These statistics were based on analysisof 173 crystals and is in excellent agreement with the pre-dicted crystal size@Fig. 3~a!#. It clearly shows that nanodia-mond growth in CDC synthesis occurs under equilibriumconditions, as expected from a slow process when noquenching of metastable structures is expected, and with noplasma or other high-energy activation that could cause for-mation of high-energy phases. However, it is necessary toremember that the free energy difference between nanocrys-

talline graphite and nanodiamond is only 3–4 kJ/mol,41

while the free energy change in the reaction~1! is2495 kJ/mol at 950 °C. Thus, additional energy advantagecan be obtained if the reactions of carbon formation fromSiC and diamond formations are thermodynamicallycoupled. The model of thermodynamically coupledreactions42 can be used to explain the formation of diamondin reactions~1! and ~2!. In any case, appropriate conditionsfacilitating diamond formation must be created to producenanodiamond instead of graphite or amorphous carbon. Wehave identified the following factors that favor diamond for-mation upon treatment of carbides in HCl or H2 /Cl2 mix-tures:

• Tetrahedrally-bonded carbide source

• Low temperature–moderate carbon mobility

• Atomic hydrogen formed at the SiC/C interface by re-action ~2!

• Matching substrate that allows epitaxial growth

However, we found in the same samples about 30 largercrystals about 15–20 nm in size and 13 crystals between 200and 800 nm in size. The number of larger crystals was notsufficient to plot a probability histogram; therefore, thosegroups are shown as single bars in the right part of Fig. 3~b!.These crystals had rounded or irregular shapes@Fig. 4~a!#,because they were constrained by solid carbon. EELS spectra@Fig. 4~b!# confirm sp3 bonding of carbon and diamondstructure in the crystals. Diamond has a single loss featurewith an onset at about 289 eV due to itss* electronic stateswhile graphite has an additional absorption starting at about285 eV owing to its lower-lying antibondingp* states.Amorphous or disordered carbon has a peak at about 285 eV,similar to graphite. SAD identified many of these crystals ashexagonal andn-diamond~also known asg-carbon!. We be-

FIG. 2. Low ~a! and high-resolution~b!, ~c! TEM images of nanocrystal-line 2H diamond~b! and cubic 3C dia-mond ~c! surrounded by amorphouscarbon. White frame in~a! shows anapproximate location of the area mag-nified in ~b!.

FIG. 3. Theoretical dependence of theequilibrium diamond crystal size onthe temperature41 ~a! and experimen-tally measured size distribution of dia-mond crystals in CDC~b!. 173 nanoc-rystals smaller than 10 nm, 30nanocrystals from 15 to 20 nm, and 13microcrystals ~up to 800 nm! werecounted in the CDC sample understudy.

4209J. Appl. Phys., Vol. 93, No. 7, 1 April 2003 Welz, Gogotsi, and McNallan

lieve that forbidden reflexes observed in the SAD pattern ofn-diamond, in addition to cubic diamond reflexes, originatefrom the incorporation of impurities and a lattice distortionresulting in a lower symmetry.

The presence of micrometer-size crystals in the samplesshows that consumption of amorphous carbon surroundingthem or gas-phase transport may occur under appropriateconditions and lead to microcrystalline diamond growth.Graphite is energetically preferred over diamond under theseconditions. However, it is necessary to take into account thefact that the nucleation of graphite on diamond at these fairlylow temperatures may be kinetically hindered because of alow mobility of carbon atoms. The experimentally measuredgraphitization activation energy is about 188 kJ/mol.43 Thus,growth of separate diamond crystals may continue by con-sumption of amorphous carbon surrounding them~Fig. 4! bysurface or gas-phase transport in the nanoporous CDC layer.The most probable explanation is the CVD-like mechanisminvolving the gas transfer of carbon. This process should besimilar to CVD diamond synthesis, which has been thor-oughly studied.40 CVD diamond growth was observed withthe use of CCl4 or CBr4 . The C–Cl bonding energy is;335 kJ/mol and the formation of such bonds and the pres-

ence of volatile or mobile CClx fragments is expected fromthermodynamic analysis. CH4 and other hydrocarbons mayform when hydrogen is added to the environment. The theo-retical aspects of hydrogen-assisted diamond growth at lowpressures have been described in many publications40 and donot need discussion here.

An alternative mechanism of diamond crystal growth bycoalescence of nanocrystals seems to be less probable. First,it would lead to a continuous distribution of grain sizes,which is not observed@Fig. 3~b!#. Second, this mechanismmust provide a very high concentration of defects as well assintered microcrystalline regions, which were not found. Fur-thermore, compaction of nanodiamond requires high pres-sures and temperatures,10 so that nanocrystals would graphi-tize rather then coalesce under our experimental conditions.

B. Nucleation of diamond

A knowledge of the nucleation mechanism is very im-portant for understanding the diamond formation. It can beassumed that the tetrahedrally coordinated SiC lattice, whichis preserved during the chlorination, acts as a template forgrowth of diamond,30,27 and that diamond grows by direct

FIG. 4. TEM image of diamond microcrystals surrounded by disorderedsp2-bonded carbon~a! and EELS spectrum of the central crystal in comparison withthe spectrum of graphite taken from a different area in the same sample~b!.

FIG. 5. High-resolution TEM imagesof SiC/carbon interfaces showing lossof Si from the SiC lattice. Formationof amorphous carbon and nucleationof diamond at the SiC surface duringhalogenation treatment constitute thefirst stage of CDC formation.

4210 J. Appl. Phys., Vol. 93, No. 7, 1 April 2003 Welz, Gogotsi, and McNallan

transformation of the SiC lattice because ofsp3 bonding ofcarbon in SiC and the similar structure ofb-SiC, which has adiamond lattice where 50% of carbon atoms are replacedwith Si. However, our work done usinga-SiC showed thatany SiC polytype can be converted to diamond, although thequestion remains as to how the carbide lattice structure~cu-bic or hexagonal! affects the probability of cubic or hexago-nal diamond formation. Molecular-dynamics simulation us-ing empirical interatomic Tersoff potentials shows that for anSi-terminated ~1000! 6H–SiC surface very high-latticestrains do not allow direct continuous growth of diamond onSiC and fragmentation, leading to nanocrystalline material,must occur.44 Small diamond clusters on SiC are predicted tohave a good adhesion to the substrate and maintainsp3 co-ordination of carbon atoms.44

TEM study of SiC/carbon interfaces~Fig. 5! suggestedthat chlorine removes Si selectively from the SiC lattice,leading to lattice disorder and formation of carbon. Signifi-cant amounts of silicon~about 5%! were often observed inthe amorphous carbon at the interface. This shows that theSiC lattice may collapse before the removal of Si is com-pleted. Diamond nanocrystals were observed either directlyon SiC or close to the SiC/carbon interface. Separation of thediamond crystals from SiC can be explained either by nucle-ation of crystals from amorphous carbon formed from SiC orby break-off due to propagation of the reaction front, causingdetachment of the crystal from the substrate. The possibilityof the latter mechanism is confirmed by the observation ofdiamond crystals growing epitaxially on SiC~Figs. 6 and 7!.Only small~less then 5 nm! crystals were observed in directcontact with SiC. This result is in agreement with our mod-eling work44 showing that only very small nanocrystals canbe accommodated on SiC surfaces. If the size of the crystalincreases, the stress due to lattice mismatch becomes toolarge and an amorphous/disordered interlayer will be re-quired to accommodate diamond on SiC. The presence ofsuch a layer can be seen in Fig. 7. Figure 7 also shows thatthe structural relations between the diamond nanocrystalsand SiC grains are mainly random. Random orientation ofdiamond nanocrystals~Fig. 2! in CDC supports the nonepi-taxial growth mechanism of diamond. A very similar randomnucleation of nanodiamond was also observed in CVDsynthesis.45 Since random orientation of diamond nanocrys-

tals is typical for CDC@Fig. 2~b!#, we assume that the growthof nanocrystalline diamond mainly occurred from highly dis-orderedsp3 carbon produced by selective etching of SiC. Inmost cases, SiC was first converted to amorphoussp3 carbonand then the formation of diamond occurred within nanom-eters off the SiC/carbon interface. Growth of larger diamondcrystals@Figs. 3~a! and~b!# might be the result of gas trans-port reactions in the nanocrystalline regions@Fig. 2~b!#.However, if no hydrogen was added to the gas, nanocrystal-line diamond was slowly transformed to the thermodynami-cally stable graphitic carbon during the long-term treatmentat 1000 °C, and we consequently observed only amorphousand graphitic carbon at the distance of more than 3mm fromthe SiC/carbon interface. Thus, the role of hydrogen is pri-marily in stabilization of dangling bonds of carbon on thesurface of diamond nanocrystals. The presence of hydrogenalong with chlorine helps to maintain thesp3 hybridizationof carbon and retards the formation ofsp2 bonded carbonwhich can act as a seed for further graphite growth. There-fore, the addition of hydrogen stabilized the diamond phaseand allowed the continuous growth of a diamond-structuredfilm on the surface of SiC.

It is important to note that most of the diamond nanoc-rystals, especially the ones located in amorphous carbon

FIG. 6. Epitaxial growth of 2H diamond on SiC. No amorphous materialand a good lattice matching have been observed. Lattice spacing of 2Hdiamond is 0.219 nm in@100# direction whereas SiC spacing is 0.255 nm.~b! shows the magnified region of the area framed in~a!.

FIG. 7. Formation of diamond and amorphous carbon at the SiC/carboninterface. Amorphous carbon functions as a stress relieving interlayer.Sample was treated in Cl2-H2 for 24 h at 1000 °C.

FIG. 8. TEM image~a!, SAD pattern~b!, and high-resolution TEM image~c! of diamond nanocrystals embedded in amorphous carbon in CDC pro-duced by chlorination of TiC. Diffuse rings in the SAD pattern~b! originatefrom amorphous carbon. Bright dots come from diamond nanocrystals.Sample was treated in Ar13.5%Cl2 for 2 h at 800 °C.

4211J. Appl. Phys., Vol. 93, No. 7, 1 April 2003 Welz, Gogotsi, and McNallan

~Fig. 8!, did not have a faceted shape. Spherical nanodia-mond particles were also produced in shock compressionsynthesis. Modeling shows that spherical nanodiamonds areenergetically favored compared to cubo-octahedra at thenanoscale.46 Figure 8 shows that not only SiC but other car-bides, such as TiC, can be used to synthesize nanodiamond.

C. Transformation from diamond to graphite

Since diamond is not the stable phase of carbon at am-bient pressure, it is of interest to know in which way thecarbon atoms of the diamond structure rearrange to the morestable graphite structure and under what conditions such atransformation is likely to occur. By comparing samples afterdifferent treatments, we found that shorter treatment times ofabout 2 h or less favor the formation of diamond structure atthe interface in the Cl2 treated SiC samples. The addition ofhydrogen stabilizes the formation of diamond resulting in thefurther growth of nanocrystalline diamond films and evenmicrocrystalline diamond. Further annealing during CDCgrowth at the temperature of 1000 °C leads to graphitizationof diamond. This process occurs at a distance from the SiC/diamond interface, generating a diamond/graphite interface.

CDC coatings produced by the treatment of sintered SiCin Ar-3.5% Cl2 were black in color and graphitic, accordingto x-ray diffraction and Raman spectroscopy. However,cross-sectional hardness measurements using nanoindenterdemonstrated the existence of an intermediate layer severalmicrometers in thickness between the SiC and graphitic car-bon. The material in this layer had an average hardness ofabout 20–30 GPa and Young’s modulus of 200–300 GPa,compared to the 1.8 GPa hardness and 17–18 GPa modulusof the surface CDC layer. TEM studies of this intermediatelayer @Fig. 9~a!# have shown that it consists of a mixture ofgraphitic carbon~onions, ribbons, and disordered carbon!and nanocrystalline diamond.

High-resolution transmission electron microscopy~HR-TEM! revealed the graphitization process at the interface be-tween diamond and graphite. The size of the transformingnanocrystalline diamonds varies from 2 to 5 nm. Figure 9~a!shows a HRTEM image of the interface between a nanocrys-talline diamond area and an area where the graphite lattice isdeveloping and represents the transformation of a hexagonaldiamond nanocrystal into a graphitic onion-type structure@Fig. 9~b!#. A growing carbon onion can be seen along withcurved graphite planes and EELS verifies the presence ofboth diamond and graphite in the area of Fig 9~a!. The pref-erential crystal plane for graphitization seems to be the@100#plane of hexagonal diamond. The lattice spacing for thenewly formed graphitic planes is 0.344 nm, which is close tothat of planar graphite~0.335 nm!. In general, in the graphi-tization process three diamond planes are expected to trans-form into two graphene sheets, because there is a 2/3 corre-spondence between the density of graphite and diamond.Moreover, it has been shown that when three~111! diamondplanes match up with two~0001! planes of graphite, the in-terfacial strain is minimized.47 The formation of the graphenesheets evades dangling bond formation since the newlyformed graphite planes merge at the interface with the dia-mond structure. The graphitization proceeds towards the@100# direction as well as the@010# direction in hexagonaldiamond. Simultaneously, the diamond/graphite interfacemoves towards the core of the diamond crystal.

The formation of onionlike carbon during the annealingof nanodiamond~2–5 nm! above 1000 °C has been reportedin the literature.9,48 It was suggested that the transformationresembles a zipperlike opening of three cubic~111! diamondplanes to form two graphene sheets.9 The formation of othercurved graphitic structures such as nanotubes and nanofoldswas observed upon graphitization of larger diamondcrystals.9

Following the work of Kutznetsovet al.9 for cubic dia-mond, the schematic in Fig. 9~c! illustrates how the graphi-tization occurs for hexagonal diamond. These considerationsare unaided by molecular dynamics or quantum-mechanicalcalculations and are limited to geometrical aspects. Three~100! hexagonal diamond planes transform into two graphiticsheets by a zipperlike opening process. The marked areashows two groups of sixfold rings plus one additional atomcapable of facilitating the growth of a sixfold graphite ring.Following the exfoliation process indicated by the arrows,the atoms of the inner diamond layer may distribute in aninterchanging manner forming two graphitic layers. If a dia-mond crystal starts transforming to graphite on more thanone side, the formation of a graphitic onion structure islikely. After the initial transformation takes place, the nextstep may include the arrangement of spherical shells aroundthe nanocrystalline diamond core. At this stage, the shells arenot perfectly formed because of the required rearrangementof atoms and spheridization, which may occur during furtherannealing. However, defects in growing graphene sheets andthe spherical shape of many diamond nanocrystals~Fig. 8!probably facilitate the formation of onions. The process con-tinues and the residual diamond core finally disappears re-sulting in a completely formed carbon onion structure, such

FIG. 9. Transformation of diamond nanocrystals to carbon onions.~a! High-resolution TEM micrograph showing the beginning of transformation.~b!TEM micrograph showing carbon onions in the outer CDC layer.~c! Sche-matic showing how three@100# hexagonal diamond planes transform intotwo @1100# graphite planes.

4212 J. Appl. Phys., Vol. 93, No. 7, 1 April 2003 Welz, Gogotsi, and McNallan

as shown in Fig. 9~b!. The planes of the final spherical onionstructure exhibit a smoother curvature than the transitionalstates. The cores of most carbon onions were hollow@Fig.9~b!#, with a few being completely filled@Fig. 10~a!# enclos-ing a fullerene sphere and/or a planar cluster of a few carbonatoms. Whiled-spacing values for hollow onion structuresdo not change in the inner of the onion compared to the outershells, the lattice spacing close to the core of the filled onion@Fig. 10~a!# decreases down to;75% of the graphite inter-planar spacing. Specifically, the lattice spacing of the outeronion shells was 0.36 nm whereas values in the inner shellsclose to the core are compressed by about 25% to approxi-mately 0.27 nm. Earlier publications49 proposed onionlikestructures as nanoscopic pressure cells and estimated thepressure in the onion core up to 36 GPa. Interplanar com-pression may be due to the shift fromsp2- to sp3-type bond-ing caused by interaction between the onion shells or due toheating by the electron beam during the study. EELS spectrarecorded from nonhollow onion structures correspond to car-bon atoms in thesp2-state. Nosp3-bonding was found.

The histogram in Fig. 10~b! illustrates the size distribu-tion of carbon onions, with both hollow and filled onionstructures included. The total number of analyzed onions is40. The mean of the distribution is 12.95 nm, and carbononions with the size of 12 nm were frequently observed.Most of the nanocrystalline diamond crystals observed were5 nm in size@Fig. 3~b!#. As a result of annealing, the dia-mond nanocrystals transform to carbon onions and their sizeincreases. Since the density of diamond is higher than that ofgraphite~3.515 and 2.267 g/cm3, respectively50! and manyonions are hollow inside, an increase in volume is expected.However, a more uniform distribution of onion sizes@Fig.10~b!# compared to the nanodiamonds@Fig. 3~b!# and thepresence of onions 20–40 nm in size suggests that theymight continue growing by consuming the amorphous car-bon surrounding them.

IV. CONCLUSIONS

The structure of a typical CDC coating produced bychlorination of SiC at temperatures of about 1000 °C is sum-

marized in Fig. 11. Nanocrystalline diamond is formed afterextraction of silicon from the carbide and can be surroundedby amorphous carbon. Continued heat treatment of the nano-diamond at this temperature results in graphitization and theformation of carbon onions within a few micrometers of theSiC/carbon interface. Structures subsequent to the diamond/graphite interface contain carbon onions, as well as curvedgraphite sheets, some planar graphite, and porous and disor-dered amorphous carbon. This onion-containing layer cangrow to the thickness of more than 100mm with uniformstructure and properties. The diamond-containing interlayerand the sawlike interface towards the metal carbide explainthe excellent adhesion of CDC coatings and the onionlikesurface layer is responsible for their excellent tribologicalproperties.

This study of SiC/carbon interfaces showed that directgrowth of diamond on SiC is possible, in agreement withprevious molecular dynamics simulations. Nucleation of dia-mond from amorphoussp3-bonded carbon produced as theresult of extraction of Si from SiC has also been demon-strated.

It has been shown that the presence of hydrogen in theenvironment is not required for diamond synthesis. However,hydrogen can stabilize the nanocrystals and lead to the

FIG. 10. HRTEM image of a carbon onion in the outer CDC layer~a! and size distribution of carbon onions~b!.

FIG. 11. Structural variations through the thickness of the CDC conversionlayer on SiC: schematic diagram and TEM images.

4213J. Appl. Phys., Vol. 93, No. 7, 1 April 2003 Welz, Gogotsi, and McNallan

growth of thick diamond-structured layers. If no hydrogen isadded, diamond nanocrystals transform to graphite, formingcarbon onions and other curved graphitic nanostructures.

ACKNOWLEDGMENTS

We thank Dr. A. Nicholls for help with TEM analysisand helpful discussions, D.A. Ersoy and A. Lee for thesamples~all UIC!, and Dr. V.L. Kuznetsov~Russian Acad.Sci., Novosibirsk! and Beth Carroll~Drexel University! forcritical comments on the manuscript. This work was sup-ported by DARPA via ONR contract and NSF through Award9813400. The electron microscopes used in this work areoperated by the Research Resources Center at UIC. TheJEM-2010F purchase was supported by the NSF~DMR-9601792!.

1W. A. Yarbrough and R. Messier, Science247, 688 ~1990!.2J. C. Angus and C. C. Hayman, Science241, 913 ~1988!.3G. Burkhard, K. Dan, Y. Tanabe, A. B. Sawaoka, and K. Yamada, Jpn. J.Appl. Phys., Part 233, L876 ~1994!.

4F. P. Bundy, H. T. Hall, and R. H. Wentorf, Nature~London! 176, 51~1955!.

5Y. Li, Y. Qian, H. Liao, Y. Ding, L. Yang, C. Xu, F. Li, and G. Zhou,Science281, 246 ~1998!.

6R. Roy, K. A. Cherian, J. P. Cheng, A. Badzian, C. Langlade, H. Dewan,and W. Drawl, Innovations Mater. Res.1, 65-87~1996!.

7M. N. Regueiro, P. Monceau, and J. L. Hodeau, Nature~London! 355, 237~1992!.

8D. M. Gruen, S. Liu, A. R. Krauss, and X. Pan, J. Appl. Phys.75, 1758~1994!.

9V. L. Kutznetsov, I. L. Zilberberg, Y. V. Butenko, and A. L. Chuvilin, J.Appl. Phys.86, 863 ~1999!.

10A. L. Vereschagin,Detonation Nanodiamonds~Altai State Technical Uni-versity, Barnaul, Russian Federation, 2001!.

11A. I. Laymkin, E. A. Petrov, A. P. Ershov, G. V. Sakovitch, A. M. Staver,and V. M. Titov, Doklady Academii Nauk~USSR! 302, 440 ~1988!.

12V. Y. Dolmatov, Usp. Khim.70, 687 ~2001!.13N. R. Greiner, D. S. Phillips, and F. J. D. Johnson, Nature~London! 20,

100 ~1988!.14A. K. Sharma, H. G. Salunke, G. P. Das, and P. Ayyub, J. Phys.: Condens.

Matter 8, 5801~1996!.15K. E. Spear, A. W. Phelps, and W. B. White, J. Mater. Res.5, 2277~1990!.16A. W. Phelps, W. Howard, and D. K. Smith, J. Mater. Res.8, 2835~1993!.17R. E. Hannemann, H. M. Strong, and F. P. Bundy, Science155, 995

~1967!.18F. P. Bundy and J. S. Kasper, J. Chem. Phys.46, 3437~1967!.19S. R. Silva, G. A. Amaratunga, E. K. Salje, and K. M. Knowles, J. Mater.

Sci. 29, 4962~1994!.20K. Maruyama, M. Makino, N. Kikukawa, and M. Shiraishi, J. Mater. Sci.

Lett. 11, 116 ~1992!.21M. Frenklach, R. Kematick, D. Huang, W. Howard, K. E. Spear, A. W.

Phelps, and R. Koba, J. Appl. Phys.66, 395 ~1989!.

22R. Kapil, B. R. Mehta, and V. D. Vankar, Thin Solid Films312, 106~1998!.

23Y. Lifshitz, X. F. Duan, N. G. Shang, Q. Li, L. Wan, I. Bello, and S. T.Lee, Nature~London! 412, 131 ~2001!.

24R. Kapil, B. R. Mehta, and V. D. Vankar, Appl. Phys. Lett.68, 2520~1996!.

25H. Hirai and K. I. Kondo, Science253, 772 ~1991!.26M. Rossi, G. Vitali, M. L. Terranova, and V. Sessa, Appl. Phys. Lett.63,

2765 ~1993!.27Y. G. Gogotsi, J. D. Jeon, and M. J. McNallan, J. Mater. Chem.7, 1841-

1848 ~1997!.28Y. Gogotsi, S. Welz, D. A. Ersoy, and M. J. McNallan, Nature~London!

411, 283 ~2001!.29Y. G. Gogotsi and M. Yoshimura, Nature~London! 367, 628 ~1994!.30Y. G. Gogotsi, P. Kofstad, K. G. Nickel, and M. Yoshimura, Diamond

Relat. Mater.5, 151 ~1996!.31Y. G. Gogotsi,Formation of Carbon Coatings on Carbide Fibers and

Particles by Disproportionation Reactions, Kluwer, Dordrecht, 1997, pp.217-230.

32D. Ersoy, M. J. McNallan, and Y. G. Gogotsi, Mater. Res. Innovations5,55 ~2001!.

33D. A. Ersoy, M. J. McNallan, Y. Gogotsi, and A. Erdemir, STLE Tribol.Trans.43, 809 ~2000!.

34D. M. Gruen, Annu. Rev. Mater. Sci.29, 211 ~1999!.35M. Zarrabian, N. Fourches-Coulon, G. Turban, C. Marhic, and M. Lancin,

Appl. Phys. Lett.70, 2535~1997!.36R. M. Hough, I. Gilmour, C. T. Pillinger, W. Arden, K. W. R. Gilkes, J.

Yuan, and H. J. Milledge, Nature~London! 378, 41 ~1995!.37K. J. Grannen and R. P. H. Chang, J. Mater. Res.9, 2154~1994!.38J. M. J. Lannon, J. S. Gold, and C. D. Stinespring, J. Appl. Phys.77, 3823

~1995!.39V. Heera, W. Skorupa, B. Pecz, and L. Dobos, Appl. Phys. Lett.76, 2847

~2000!.40B. V. Spitsyn, inHandbook of Crystal Growth; Vol. 3, edited by D. T. J.

Hurle ~Elsevier, New York, 1994!, pp. 401–456.41M. Y. Gamarnik, J. Appl. Phys.86, 863 ~1999!.42J.-T. Wang, C.-B. Cao, and P.-J. Zheng, J. Electrochem. Soc.141, 278-281

~1994!.43Y. V. Butenko, V. L. Kuznetsov, A. L. Chuvilin, V. N. Kolomiichuk, S. V.

Stankus, R. A. Khairulin, and B. Segall, J. Appl. Phys.88, 4380~2000!.44Y. Gogotsi, V. Kamyshenko, V. Shevchenko, S. Welz, D. A. Ersoy, and M.

J. McNallan, inProceedings NATO ASI on Functional Gradient Materialsand Surface Layers Prepared by Fine Particles Technology, edited by M.I. Baraton and I. V. Uvarova~Kluwer, Dordrecht, 2000!.

45S. Jiao, A. Sumant, M. A. Kirk, D. M. Gruen, A. R. Krauss, and O.Auciello, J. Appl. Phys.90, 118 ~2001!.

46O. Shenderova, V. V. Zhirnov, and D. W. Brenner, Carbon Nanostructures,Critical Reviews in Solid State and Materials Science~to be published!.

47W. R. L. Lambrecht, C. H. Lee, B. Segall, J. C. Angus, Z. Li, and M.Sunkara, Nature~London! 364, 607 ~1993!.

48V. L. Kuznetsov, A. L. Chuvilin, Yu. V. Butenko, I. Yu. Malkov, and V. M.Titov, Chem. Phys. Lett.222, 343 ~1994!.

49F. Banhart and P. M. Ajayan, Nature~London! 382, 433 ~1996!.50T. Evans, inThe Properties of Diamond, edited by J. E. Field~London,

1979!, p. 403.

4214 J. Appl. Phys., Vol. 93, No. 7, 1 April 2003 Welz, Gogotsi, and McNallan