Polyhedrm Vol.13, No. 8, pp. 131S1327, 1994 Pergamon Ekvier Science Ltd. Printed in Great Britain

REACTION BONDED REFRACTORY METAL CARBIDE/CARBON COMPOSITES

ANDREW N. MacINNES and ANDREW R. BARRON” Department of Chemistry, Harvard University, Cambridge, MA 02138, U.S.A.

and

JASON J. LI and THOMAS R. GILBERT Department of Chemistry and Barnett Institute of Chemical Analysis and Materials Science, Northeastern University, Boston, MA 02115, U.S.A.

Abstract-A simple chemical process has been developed that allows the rapid conversion of carbon substrates to the refractory metal carbides (MC) of titanium, zirconium, hafnium, vanadium, niobium, tantalum and tungsten. As a demonstration of the technique, graphite rods are infiltrated with an ethanolic solution of various metal oxide-halides, and resistively fired (1350-3000°C) under an argon atmosphere. The final composite produced is found to contain a residual carbon core within a fully dense metal carbide-carbon matrix outer layer. The effects of both firing temperatures, and the number of infiltration-fire treatments, have been investigated in order to determine optimum firing conditions (i.e. maximum carbide formation) for each individual metal carbide. To demonstrate the applicability of the technique for a range of substrate materials, niobium and titanium carbides were formed on carbon fibre tows and non-porous substrates, respectively. The formation of mixed Nb- Ti carbide coatings of varying composition were investigated in order to determine the factors controlling the formation of ternary phases. The identity of the transition metal precursors and the possible pathway of their conversion to the appropriate carbide is discussed. All the composites have been characterized by X-ray photoelectron spectroscopy (XPS), scanning electron microscopy (SEM) with associated energy-dispersive X-ray (EDX) analysis, and X-ray diffraction (XRD).

INTRODUCTION carbide to provide physical compatibility with the carbon substrate and prevent the outward diffusion Carbon fibre reinforced carbon matrix (carbon- of carbon. The outer layer in the multi-layered coat- carbon) composites are currently being proposed ing scheme would then be a suitable oxide (possibly as candidate structural materials for aerospace use the native oxide formed on the carbide inner layer) based on their high specific strength at elevated providing a barrier to the ingress of oxygen into the temperatures (> 2OOo”C).’However, the successful carbon substrate. application of such composites will be dependent Although inert barriers coatings formed by tra- on the development of a suitable high temperature ditional methods [e.g. physical vapour deposition coating system to alleviate the susceptibility of the (PVD), chemical vapour deposition (CVD) and sol- carbon based materials to oxidation above 400°C. gel techniques] do provide diffusion barriers: they Strife and Sheehan’ have reviewed coating require- are, however, generally not resistant to thermal ments and proposed a multi-layered coating system. cycle operations imposed on the composite The inner layer would consist of a refractory metal material. This failure during thermal cycling is due to the thermal expansion mismatch between the *Author to whom all correspondence should be coating (metal carbide or oxide) and the carbon addressed. matrix. Cracking and spallation of the coating 1315 1316 A. N. MacINNES et al.

Coating Coating

Substrate Substrate I I I

(I) (11)

ensues as a result of thermal stresses generated at We have previously communicated that niobium an essentially atomically abrupt interface (I). It carbide (NbC) coatings can be produced on gra- becomes clear, therefore, that before full exploi- phitic carbon by the infiltration of an ethanolic tation of carboncarbon composites can be solution of niobium pentachloride into the carbon realized, it is essential to alleviate the thermal stress substrate followed by resistive firing.Y In addition, generated in these materials during thermal cycling. we have shown, in preliminary studies, that this A common approach in the matching of two dis- process may be extended to the more dense carbon similar materials such as quartz and borosilicate fibre substrates.“,” In this present paper we extend glass is to produce a graded interface ; in terms of these communications to the formation of carbide chemical composition and mechanical properties coatings of the refractory metals titanium, (II). If a graded interface will overcome the thermal zirconium, hafnium, vanadium, niobium, tantalum expansion mismatch between carbon substrates and and tungsten. In addition, the formation of ternary ceramic coatings, the problem then becomes one of carbides, the effect of the carbon substrate identity developing suitable routes to the formation of such on carbide formation, and the possible mechanism a graded interface.4 In a related study we have for carbide formation are presented. reported the formation of graded interfaces between niobium metal and ceramic coatings via thermal, RESULTS AND DISCUSSION ion beam and laser treatment.5.6s7 However, these physical techniques do not lend themselves to appli- Dissolution, in wet (as received) ethanol, of the cation for carbon substrates. We have, therefore, early transition metal halides listed in Table 1 initiated a program of research to examine the results in the evolution of HCl and the exothermic possibility of facile chemical reactions leading to the formation of solvated oxide--chlorides (see after). production of a graded interface between graphite The immersion of carbon graphite rods in these carbon substrates and refractory metal carbide oxide-chloride solutions followed by drying and coatings.’ resistively firing in the apparatus shown in Fig. 1,

Table 1. Chemical characterization and optimum firing temperature for the metal carbide coated carbon rods

Optimum firing temperature (“C) XPS peak shift for carbide Max. carbide content Max. metal content Metal Precursor system XRD XPS XPS c 1s M 2~312

Titanium TiClJEtOH 2200 2000 2200-2500 281.5 454.63 Zirconium ZrCL,/EtOH 2200 2250 1500-1700 Hafnium HfCl,/EtOH 2200 Y 150&1750 Y L1 Vanadium VCl,/EtOH 2000 2200-2500 2200 282.24 513.10 Niobium NbC&/EtOH 1800 223@-2750 210&2750 282.5 203.6’ Tantatum TaCl,/EtOH 1800 2000 2000-2250 281.75 226.96 Tungsten WCl,/EtOH 2200 LI 2750 n 31-5

n No carbide detected on the surface by XPS. bNb 3d5/2. 1317

BNchimber electrode clamps

----)- power supply -----_

t Ar Fig. 1. Schematic diagram of apparatus used for resistively firing carbon rods.

Fi g. 2. SEM cross sectional micrographs of untreated (a), and Nb treated (b) rod. In the latter can be seen residual graphite particle (A) within a NbC/C matrix (B). 1318 A. N. MacINNES et al.

Fig. 2-continued. results in the formation of the appropriate metal The XRD spectrum of the pulverized fired rods carbide in a carbide-carbon compact. (Fig. 3) exhibit reflections due to the presence of The SEM cross sectional micrograph of an un- graphitic carbon and the appropriate refractory- treated rod is shown in Fig. 2a, while the equivalent metal carbide. Oxides are only observed by XRD view for a niobium treated rod is shown in Fig. as a minor constituent for some of the samples 2b. Whereas the untreated rods are porous, but of (Fig. 3). It should be noted that the minimum firing homogeneous composition, it can be clearly seen temperature required for the detection of the metal from Fig. 2b that under the reaction conditions carbide by XRD (see Table 1) is dependent on the of the experiment a fully impregnated composite identity of the metals. As is discussed below, sig- material is produced. The composite consists of nificant quantities of the carbides are formed at residual carbon particles (Fig. 2b. A) within a metal lesser firing temperatures, but are presumably carbide-carbon matrix (Fig. 2b. B). Similar bulk amorphous. microstructures are obtained for all the other metal The surface chemical analysis of the fired rods carbides. was enabled by X-ray photoelectron spectroscopy Reaction bonded refractory metal composites 1319

0.4

0.0 0.3 2.2 4.2 6.1 8.1 10.0

ANGLE ( degrees 2 theta ) [x 1O1] Fig. 3. XRD of pulverized niobium treated carbon rod (Cu-K, radiation NbC peaks are indexed, C refers to graphite peaks).

(XPS), the results of which are given for the opti- mum firing conditions in Table 1. As a typical result Carbide Fig. 4 shows the Nbxd and Cl, regions for a NbCl, treated rod, with the peak-fits performed for the Carbon chemical states of niobium and carbon present.‘* The niobium is present as both carbide and oxide (Nb205 and a suboxide) and the carbon as graphite, VW carbide and adventitious hydrocarbons. In all the samples the signal for the adventitious hydrocarbon that the carbide : carbon ratios of the bulk material and oxides are eliminated after argon ion beam as determined by XRD are lower than those found sputtering of the surface. The oxide present is there- by EDX (a near-surface technique) of the surface fore due to surface oxidation of the carbide rather of an uncleaved rod. These results are consistent than as a result of unreacted precursor. It should with the absorption of the precursor solution into be noted that no chlorine was detected for any of the carbon rod followed by carbide formation the samples, consistent with the complete decompo- throughout the rod during the subsequent firing sition of the precursors. step. In the cases where carbide is detected the metal The relative carbide content of the treated rods to carbon ratio is higher for EDX (a near surface will clearly be dependent on a number of possible technique) than the metal carbide to carbon graph- factors, including : the number of infiltration-firing ite ratio determined by XRD (a bulk technique). sequences, the firing temperature, and the identity Although this may be due, in part, to the formation and quantity of the transition metal. We have of amorphous metal-carbide, an EDX map of the employed the results from XRD and XPS to deter- cleaved face, (cross-section) of a whole rod (Fig. mine the experimental parameters for the maximum 5) shows a decrease of metal content (in this case carbon to carbide conversion. Since the detection titanium) with increasing penetration depth, i.e. the of nipbium carbide by XRD at firing temperatures surface (the left hand side of the micrograph) is-rich as low as 14OO”C,allows for a comparison over a in metal carbide while the core is primarily carbon wide temperature range, this system was chosen as graphite, as shown in the schematic cross section of the model system. It should be noted that while the a treated graphite rod, III. Given this gradient in concentration of the precursor solution will limit elemental composition between the surface and the the theoretical yield of carbide formation when very centre of the rods, it is not unsurprising, therefore, dilute solutions are used, the use of saturated solu- 1320 (a) 4000

203.63 35.3 3d” carbide

206.36 26.5 3d3 carbide 207.02 15.9 3d5 peatoxide r B 209.62 11.8 3d3 peatoxide

8

214.0 Binding Energy (eV)

(b) 2000

==W X Assignment

282.57 20.1 carbide 284.18 592 graphite 285.24 20.7 advmdtious hydrocarbon

Binding Energy (eV) Fig. 4. (a) Nbxd and (b) C,, XPS spectra (monochromatized Al-K, radiation) of a fired carbon rod surface. Peak positions are included (eV).

Fig. 5. Titanium EDX-map of a cross section of a single treated rod. The surface of the rod is at the far left of the picture with the centre at the far right. Reaction bonded refractory metal composites 1321

0.16 -

g 0.14- , i 0.12 - 5 4 0.10 -

b 0.08 -

0.06! . I . I . I I 1 o.oT 0 5 10 15 20 25 1400 1800 2200 2600 3OOcl infiltration firing sequences (n) Temperature ( “C) Fig. 6. A plot of relative carbide content in the bulk of a Fig. 7. A plot of relative carbide content in the bulk and niobium treated rod, as determined by XRD, vs number on the surface of a niobium treated rod, as determined of infiltration firing sequences (n). by XRD and XPS respectively, vs the firing temperature.

tions (l-3 mol dmm3) in our experiments ensures and 1495°C; the former being the temperature at that the precursor is not the limiting reagent. which chlorine is liberated from the oxychlorides, In a typical series of experiments the carbon rods while the latter is the decomposition point of the were subjected to a number (n) of infiltration-firing least stable oxide of niobium, Nb205. Thus, carbide sequences (n = 1,5, 10,23) and then pulverized for formation only occurs at temperatures above the XRD analysis. The results for the niobium treated decomposition point of the precursor materials. rods are shown in Fig. 6. It can be seen that the The constant bulk carbide formation between 1600 bulk carbide content relative to residual graphite and 2400°C indicates that once the precursor increases approximately two-fold between one and decomposes carbide formation is essentially tem- five cycles, and then remains essentially unchanged. perature independent. As such, the increase in the This effect may be readily explained in terms of amount of carbide formations in the niobium sys- the SEM and EDX data (cf. Figs 2 and 4) which tems at temperatures above 2400°C may seem mis- indicated that a dense composite is formed after placed. However, two factors may account for this, firing. Thus, once sufficient reaction has occurred firstly, it should be noted that the melting point of to produce a dense material within the surface layer niobium is 2470°C and the formation of carbide of the rod, no further uptake of the precursor may be enhanced by the formation of a fluid state solution will occur due to pore-blocking, and as a which allows for maximum diffusion and reactivity. consequence, no further bulk carbide formation Secondly, it has been previously observed that spon- results. taneous catalytic reaction occurs between carbon In order to investigate the effects of firing tem- and niobium oxide above 25OO”C, and this may peratures on carbide content a series of carbon rods enhance the formation of carbide in the present were subjected to four soak-fire sequences over a system.13 range of temperatures, lOOO-3OOO”C,and the rela- As with the bulk carbide content the fraction of tive carbide composition of the rods determined in niobium carbide on the surface generally increases the bulk and on the surface by XRD and XPS, with increasing temperature (Fig. 7). However, respectively. The results for the formation of NbC below 1800°C no niobium carbide is observed on as a convenient example, are presented in Fig. 7. In the surface ; all the niobium being present as oxide, the case of niobium no crystalline carbide is formed and above 2600°C a small decrease in niobium car- below 1400°C. Between 1600 and 2400°C the quali- bide content is observed. The decrease at high tem- tative extent of carbide formation in bulk (as deter- peratures is consistent with the thermal decompo- mined by comparison of the relative intensity of the sition (decarburization) and oxidation of the NbC and graphite peaks in the XRD) is significant carbide at the firing temperatures.’ The lack of car- and relatively constant; above 2400°C a further bide formation on the surface of the fired rod below increase in the relative amount of carbide results. 1800°C is not as easily explained but presumably it From the foregoing it is clear, therefore, that for the is due to the oxidation of the small amount of bulk material three distinct temperature regimes are niobium on the surface. Thus, Fig. 7 indicates that present. The lowest range, below 14OO”C,does not the temperature range for maximization of carbide result in crystalline carbide formation and may be vs oxide vs carbon graphite both on the surface and related to the decomposition point of the precursor the bulk of a niobium treated rod is 22OCL26OO”C. compounds (see below) which occurs between 1200 Since we are primarily concerned with the max- 1322 A. N. MacINNES et al. imization of carbide on the surface as opposed to a carbide but a contiguous coating of oxide and/or bulk infiltration, we have investigated the surface carbide to inhibit oxidation of the substrate, then carbide : carbon temperature profile for each of the based on the above data we propose that our metal precursor systems by XPS. The results for titanium, niobium and vanadium treatments are each element are briefly discussed below. The opti- most suitable for the oxidation protection of porous mum firing temperature, that at which the car- carbon substrates. In order to determine the appli- bide : carbon ratio is a maximum, for each system cability of this carbide formation methodology for was thus determined, and is given in Table 1. In dense carbon substrates we have investigated the addition, since the carbides all form a native oxide formation of NbC and TiC on carbon fibres and on their surface we have determined the tem- non-porous carbon graphite rods, respectively.” perature at which the elemental composition is max- These studies are described below. imized for the metal. It should be noted that in some cases there is a significant difference between the Non-porous carbon substrates optimum temperature determined from XPS and XRD. This disparity suggests that a balance exists SEM-EDX studies on the porous rod (cf. Fig. 5) between surface coverage and bulk penetration. indicated that infiltration/carbide formation The titanium treated rods show a marked occurred at depths of ca 500-600 pm. Since the increase, from ca 20 to 30%, in the titanium content depth of penetration will be dependent on the den- on the carbon surface at about 2000°C. This tem- sity and homogeneity of the carbon substrate we perature is also the point at which the chemical have treated a highly compacted (dense) bulk car- identity of the titanium alters from oxide to carbide. bon substrate (Union Carbide). The substrate was However, at higher temperatures, 275O”C, the car- treated in an analogous manner to that described bide content again decreases, to ca 15%, consistent in Table 1 for the optimum TiC formation. Shown with decarburization. By contrast, the treated rods in Fig. 8 is the Ti EDX map of the cross section of for the heavier group 4 metals, zirconium and a treated non-porous rod. It can be clearly seen that hafnium, show a slow decrease in metal content at the average penetration depth, as indicated by the higher temperatures, possibly consistent with the lighter area in the micrograph, is ca 15 pm. It should loss of precursor by volatilization. Whereas, the also be noted that deeper penetration has occurred XRD studies indicate carbide formation for both in a number of areas, but this is most probably due elements, the rods’ surface consists of entirely oxide, to cracks (defects) in the surface, or areas of the with no carbon being detected. The only exception rod that have been insufficiently compacted during is for zirconium at 2250°C where a small quantity processing. (4.5%) of surface carbide is detected. Carbon fibres represent perhaps the most dense The vanadium treated samples show that while form of graphitic materials, and as such should the vanadium content increases only moderately exhibit the lowest penetration during carbide for- from 1500°C to 22OO”C, the carbide content mation. Carbon fibre tows were treated with an increases, over the same temperature range, from ethanolic solution of niobium pentachloride and zero to being the exclusive chemical species on the subseq,uently fired in an analogous manner to that surface. However, at 2750°C there is no vanadium for the carbon rods described above. SEM micro- detected on the surface implying the complete vola- graphs of the fired tows show that although residual tilization of either the precursor and/or the resulting material (shown by EDX to be niobium rich) was carbide or oxide has occurred. formed between the fibres, the individual fibres are As would be expected from the similarity in their themselves coated (Fig. 9a) and may be readily sep- chemistry the trends observed for niobium and tan- arated from any excess material by gentle teasing. talum are related, both showing oxide at low tem- The presence of niobium in these coatings is con- peratures, increasing carbide above 1800 and firmed by Nb EDX-mapping (Fig. 9b). The cut 1750°C for niobium and tantalum, respectively. ends of the fibres are clearly distinguishable as dark Finally, in line with the XRD results a slight areas, indicating the absence of Nb, in contrast to decrease in carbide content at high temperatures. the light appearance of the fibre surface (Nb). The Although tungsten carbide (WC) was detected by available data can, therefore, be interpreted by pro- XRD no carbide was observed on the surface of the posing that the firing sequence has, indeed, pro- treated rods by XPS over the range of 175&275O”C. duced a niobium carbide (and perhaps oxide) coat- However, no carbon was detected indicating that ing on the fibres (as is evident from the increased the surface is entirely oxide.14 diameter), 7 pm when compared to the untreated Given that the oxidation protection of a carbon fibres 5 pm. The presence of a carbide coating on composite would require the formation, not only of the individual fibre was confirmed by XPS and AES, Reaction bonded refractory metal composites 1323

Fig. 8. Titanium EDX-mad of a cross section of a non-porous carbon rod. The surface of the rod (A) is indicated by the arrow.

as well as oxidation studies.” SEM of individual solutions. The resulting relative metal (carbide and fibres show that the carbon fibre core remaining oxide) content was the-n determined by XPS analy- after treatment is approximately 2 pm in diameter. sis. The results of these runs are shown in Fig. 10. Thus the carbide coating is ca 2.5 pm in depth, The formation of solid solutions, as opposed to significantly lower than for the other substrates dis- separate NbC and TiC formation, was confirmed cussed. by XRD. It can be seen that the relative niobium and Formation of Nb-Ti carbide solid solutions titanium content of rods fired at 1800°C is within experimental error of the theoretical values based While the reaction bonded process described on the complete conversion of the metal compounds above provides a convenient route to metal carbide in the precursor solution. Thus, the metal content of coatings on carbon substrates, its chemical sim- the resulting carbide/graphite compact is controlled plicity has the added potential that mixed carbide by the relative amounts of each precursor metal phases should be readily prepared by painting the halide. In contrast, the analysis of rods fired at substrate with solutions containing~more than one 2500°C are all titanium deficient. In fact the metal salt. Niobium carbide and titanium carbide niobium content varies little compared to that of the are known to form solid-‘solutions,‘5 and as such precursor solution. The reason for the low titanium represent a suitable system, for study of a mixed content cannot be due to thermal loss of the carbide metal carbide coating ot&bbn substrates. (decarburization) since for the pure titanium system Porous carbon substrates were painted with etha- the optimum carbide content as determined by XPS nolic solutions consisting of mixtures of NbCls and is obtained by firing at temperatures between 2200 Tic&. The metal content of five solutions was con- and 2500°C. Thus, the low titanium content must trolled such that the niobium content was equal to presumably be due to the preferential formation of 0, 25, 50, 75 and 100%. If the relative rates of NbC over Tic. If the rate of NbC formation at formation of NbC and TiC are similar then the 2500°C is sufficiently greater than that of TiC then relative proportion of niobium should be retained the available carbon substrate will be consumed by from solution to the carbide compact. The carbon the formation of NbC, and further heating will only rods were fired at either 1800 or 2500°C to deter- cause evaporation of the titanium precursor, lead- mine the effect of firing temperature on mixed metal ing to a niobium rich composite. ? ?

;

Fig. 9. SE1 micrograph of Nb treated carbon fibres (a) with associated Nb EDX-map (b). N.B. The cut ends of the fibres are clearly distinguishable as dark areas, indicating the absence of Nb, in contrast to the light appearance of the fibre surface (Nb). Reaction bonded refractory metal composites 1325 oxidation may occur, resulting in the formation of a variety of oxide and oxychloride salts, e.g. VOC153- which is commonly prepared by the interaction of ethanolic HCl with V205.‘6 The ethanolic hydrolysis of niobium and tan- talum pentahalides has been reported to result in the formation of the oxychlorides MOC13 and MOrCl as well as the oxides M20S.Z0’21FAB mass spectra of the niobium and tantalum precursor solu- 0.2 0.4 0.6 0.8 Solution Nb content (96) tions indicates the presence of both M-O and M- Cl fragments, while the 93Nb NMR spectra is con- Fig. 10. A plot of the relative niobium and titanium content (% Nb) of rods fired at 1800 and 2500°C as a sistent with the formation of solvated oxide-chlor- function of the precursor solution composition (% Nb). ide species. The dashed line represents the theoretical values based Since the WC16 and the oxychlorides, W02C12 on the precursor solution. and WOCL,, are all volatile under the conditions employed for carbide formation it is unlikely that these are the direct precursors to the metal carbide. Tungsten hexachloride readily dissolves in dry etha- Decomposition of the inorganic precursors nol to form a deep blue solution, however, in the All of the halide salts used as precursors are vol- presence of excess water hydrolysis occurs resulting atile at the firing temperatures employed for carbide in the formation of tungstic acid (eq. 5). formation. It is clear, therefore, that the metal hal- WCl,+5H,O- WOj*2Hz0+6HCl (5) ides cannot themselves be involved in the carbide forming reaction. In fact, none of the halides, MCI,, In the present case complete hydrolysis does not retain their chemical identity in moist alcoholic occur since FAB Mass Spectral data indicates that solutions ; each reacts with the water present in the tungsten-chloride bonds are retained since peaks solvent to liberate HCl gas and form solvated oxides due to lWClJ+ are observed as the major tungsten and oxide chlorides.‘6 containing fragments along with mod’. As with Titanium trichloride has been reported to react the group 5 metals it is probable that the tungsten with water, in the air, to give titanic acid which precursor solution consists of a mixture of oxide in the presence of the HCl liberated yields blue and oxide-chloride species, possibly including solutions of titanium oxydichloride (eq. 1).17 w2C1604]2- which has been recently isolated.16 Thus, based on literature precedent, and our TiCl,+H*O -H,Ti03-----+ TiOCl* (1) qualitative spectroscopic and analytical studies we While zirconium and hafnium tetrachloride have propose that the precursor solutions consist of sol- both been shown to form the analogous oxy- vated metal oxide