Oxidation resistant tungsten carbide hardmetals

Int. Journal of Refractory Metals and Hard Materials 66 (2017) 135–143

Contents lists available at ScienceDirect

Int. Journal of Refractory Metals and Hard Materials journal homepage: www.elsevier.com/locate/IJRMHM

Oxidation resistant tungsten carbide hardmetals Samuel A. Humphry-Baker a,⁎, Ke Peng a,b, William E. Lee a a b

Department of Materials, Imperial College London, Prince Consort Road, London SW7 2BP, UK State Key Laboratory for Powder Metallurgy, Central South University, Changsha, 410083, China

a r t i c l e

i n f o

Article history: Received 12 October 2016 Received in revised form 7 March 2017 Accepted 16 March 2017 Available online 18 March 2017 Keywords: Metal matrix composites Weight loss X-ray diffraction SEM Oxidation

a b s t r a c t We present a new method for retarding the oxidation rate of hardmetals. By diffusion impregnating a WC-FeCr hardmetal with silicon, we manufacture two-layered silicide coatings consisting of an FeSix outer crust and WSi2 beneath. The structure results from a preferential reaction between silicon and the metallic binder. The FeSix outer layer is crucial to providing oxidation resistance as when exposed to oxygen it passivates, forming a protective SiO2 surface film – while simultaneously preventing exposure of the underlying WSi2, which is known to oxidise in an active manner. Our analysis shows the coating method is applicable to various hardmetals structures. © 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http:// creativecommons.org/licenses/by/4.0/).

1. Introduction Tungsten carbide hardmetals have excellent wear and thermal properties, yet are susceptible to oxidation at relatively moderate temperatures. Their wear resistance results from combining hard WC particles in a matrix of a ductile metallic binder, commonly Co, Ni, Fe, Cr or some combination thereof. However, oxidation can degrade their mechanical properties [1] and thus limit their useful life. Furthermore, hardmetals are being considered as shielding materials for fusion reactors [2,3], due to their high attenuation of high-energy neutrons [2], impressive fracture toughness [4] and manufacturability [5]. However, their susceptibility to oxidation could be hazardous in a power plant accident scenario. For example, a combined air ingress and loss-of-coolant event could cause a temperature rise in the first-wall components to over 1000 °C [6], whereupon tungsten oxides would form and subsequently volatilize, releasing dangerous transmutation products such as Os. Since the oxide scale on hardmetals is highly cracked and porous, the scales are generally unprotective [7–9] and thus degradation would continue uninhibited. Attempts have been made to engineer hardmetals that are oxidation resistant. WC-Co is the most widely studied system, and we therefore begin discussing these materials (although we note that Co is excluded from application in fusion devices due its poor radiological properties [10]). Oxidation resistance in WC-Co is most commonly achieved by additions of cubic carbides, e.g. TiC, TaC and NbC [11–13]. The resulting oxides form with a smaller volume increase and thus transition from linear ⁎ Corresponding author. E-mail address: [email protected] (S.A. Humphry-Baker).

(i.e. unprotective) to parabolic (protective) kinetics [12]. However, the binder can play an important role too: Ni in the binder is reported to be detrimental [7,13], and Cr beneficial [14,15]. Increasing binder content is generally reported to increase oxidation rate [8,16], however some reports dispute this [17] – and our recent demonstration of anomalously fast oxidation kinetics in highly porous WC samples suggests that the role of binder metal is not well understood [18]. Since Fe is a relatively uncommon binder element, literature on the oxidation of properties of WC-Fe is sparse. Promisingly, however, Fe-based hardmetals are reported to have improved oxidation resistance over pure Co ones [19,20]. Coatings are typically more effective than these bulk modifications: boron is often impregnated into the surface, tending to form the ternary tungsten boride, e.g. WCoB in the case of WC-Co [21,22] — with the added advantage of improved surface hardness [22]. Such boride coatings can retard the rate of oxidation by about an order of magnitude [23], however their lifetime in oxidising environments is limited as boria volatilizes easily. To further enhance oxidation resistance, new coating methods are needed. Silicon impregnation has been successful for many structural alloys, due to their tendency to induce robust SiO2 scales upon high temperature oxidation [24,25]. Examples include iron- [26], tungsten[27] and vanadium-based systems [28]. However, such coatings have been little – if at all – studied in hardmetals. In what follows, a WCFeCr hardmetal is impregnated with Si by pack cementation, the structure characterised, and its properties compared to the untreated substrate. The coating suppresses the oxidation rate by 3–4 orders of magnitude and is stable up to 1200 °C. Its impressive performance is explained by preferential segregation of the iron binder to the surface during coating formation.

http://dx.doi.org/10.1016/j.ijrmhm.2017.03.009 0263-4368/© 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

136

S.A. Humphry-Baker et al. / Int. Journal of Refractory Metals and Hard Materials 66 (2017) 135–143

2. Material and methods

3. Results

2.1. Coating fabrication

3.1. Microstructure of substrate and coating

Tungsten carbide hardmetals were supplied by Sandvik Hyperion Ltd., in the form of a solid plate with dimensions 40 × 20 × 8 mm. The material had a nominal composition of 90 wt% WC particles and 10 wt% ferritic iron-chromium binder, which itself had a composition of 92 wt% Fe and 8 wt% Cr. Hereafter we refer to this material as WCFe or the substrate. Coating of WC-Fe samples was performed via the pack cementation method. The powder pack consisted of two components: Si and NaF powders (Alpha Aesar), of 99% and 99.5% purity, of mesh size 50 and 90, respectively. Powders were weighed in the weight ratio 80 Si:20 NaF, mixed in a mortar with a pestle and loaded into a lidtopped alumina crucible and packed around pellets of WC-Fe that had been cut to dimensions of 7 × 4 × 4 mm. The pack was heated to 1000 °C in a tube furnace in flowing Ar - 5% H2 gas and held isothermally for 4 h. The average pellet mass gain and coating thickness were determined with a mass balance and digital micrometer with accuracies of ±0.1 mg and ±2 μm respectively.

Fig. 1 shows the phases and microstructure of the substrate before coating application. Fig. 1(a) is an XRD pattern, which indicates there are three phases present: hexagonal WC (PDF 101-3982), ferritic α-Fe (1-1267) and a small quantity of cubic Fe3W3C phase (170-7470). Fig. 1(b) shows a representative SEM cross-sectional image, showing mainly WC grains with light grey contrast and Fe binder black. The phase in dark grey contrast is Fe3W3C, also known as eta phase, which is a reaction product from sintering under a sub-stoichiometric carbon content, due to a narrow and carbon-rich two-phase region in the WC-Fe phase diagram [29]. The volume averaged spherical grain size for WC is measured to be 0.8 ± 0.12 μm. Fig. 2 shows some typical SEM micrographs of the siliconised coating. Fig. 2(a) is a cross-sectional image through the thickness of the coating. The lower part of the figure shows the substrate, which contains some small pores of a few μm in diameter, which were typically observed in the region between 20 and 200 μm from the substrate-coating interface. Above the substrate is the siliconised coating, which had a typical thickness of 65 ± 9 μm, corresponding to an average mass gain of 15.4 ± 0.5 mg/cm2 per pellet — as measured on three nominally identical samples. The coating structure contains an outer crust, which appears in dark contrast. The inset of Fig. 2 is a higher magnification image of the crust, revealing it is composed of two layers of differing chemical composition. Fig. 2(b) is a micrograph of the pellet surface, showing typical surface roughness morphology. The separation of convex asperities on the surface is about 5–10 μm. Fig. 3 shows XRD taken from (a) the coating surface and (b) from the central section of the coating after the top 30 μm of material was removed by mechanical polishing, i.e. in the plane of the sample surface. The peaks in the top surface pattern are well matched with iron silicides, namely: FeSi, FeSi2, Fe3Si7, (PDFs 192-3808, 20-0532 and 35-0822

2.2. Characterisation The substrate material, as well as coated samples, were characterised by X-ray diffraction (XRD), using a PANalytical X'Pert powder diffractometer with a Cu radiation source operated at 40 kV and 40 mA. Patterns were collected at a scan rate of 2 °C/min over a scan range of 20–90 ° 2θ. The patterns were matched to ICDD Powder Diffraction Files (PDFs) and analyzed using the Rietveld method to determine the relative phase fractions, employing a pseudo-Voigt profile function. Scanning electron microscopy images were collected using a JSM 6010 SEM, operated in secondary electron imaging mode with an Energy Dispersive X-ray (EDX) detector that was used to determine the chemical composition of microstructural features. Grain size was determined stereographically using the linear intercept method.

2.3. Oxidation tests For oxidation tests, samples were loaded into an alumina crucible inside a STA 449 F5 Jupiter Thermogravimetric Analyser (TGA). In each experiment, the sample was heated to the set point at a rate of 20 °C/ min in high purity argon, and held isothermally. Once the temperature stabilised, synthetic air (80% N2; 20% O2) was flowed over the sample at 100 ml/min for a set time interval, after which the flow gas was switched back to Ar and cooled. Further details of this procedure are given in a previous study [18]. To calculate the oxidation rate, the mass gain signal was normalised by the instantaneous sample surface area. The initial area was measured using a micrometer of accuracy ± 0.002 mm – and for coated samples this was assumed constant, since the amount of oxide up-take was small. However, for uncoated samples the area reduction during oxidation was significant and calculated by assuming that the substrate (of density 14.1 g/cm3) recedes isotropically in all directions and that the mass gain upon formation of the oxide film is about 19.4%. This mass increase factor was calculated using the following equation:

f −WC þ ð1−f Þ−Fe þ ð2 f þ 1=2Þ O2 →ð2f −1Þ−WO3 þ ð1− f Þ−FeWO4 þ f −CO2 ;

ð1Þ

where f is the mole fraction of WC, which, is about f = 0.72 for our WC-Fe samples (based on a nominal mass fraction of 0.9).

Fig. 1. Structure of the substrate, (a) XRD pattern showing the phase composition: hexagonal WC, α-Fe, and Fe3W3C; (b) an SEM micrograph showing the typical phase morphology: WC appears light grey, α-Fe black, and Fe3W3C dark grey.


137

with a small amount of WC with volume percent of 71, 29 and b1% respectively. Thus the crust is greatly enriched in iron relative to the coating bulk. 3.2. Oxidation kinetics

Fig. 2. SEM images of the coating, (a) cross-section showing a 65 μm thick coating, with an FeSix-rich outer crust, which is itself composed of two layers (inset); (b) micrograph of the pellet surface, illustrating the surface roughness.

respectively). Rietveld analysis revealed that their volume percent are about 10, 31 and 58% respectively. A small amount of WSi2 phase (192–4516) is also detected, probably due to penetration of the X-ray beam through the crust (and we note that no NaF or other Na-containing compounds from the powder pack are detected). By contrast, the pattern taken from the coating bulk matches mainly WSi2 and FeSi,

Fig. 3. XRD scans of the coating showing separation in iron and tungsten within the coating. The outer crust is mainly FeSi2 and Fe3Si7, and the bulk of the coating is mainly WSi2 with some FeSi.

The oxidation mass gain kinetics of the siliconised and substrate material are compared in Fig. 4, between 800 and 1000 °C. Each thermogravimetry trace starts at the moment air is injected into the furnace, i.e. once the sample is already settled at the isothermal set-point. For the substrate, a large and monotonic mass gain begins at the instant that air is introduced. The oxidation rate increases with increasing temperature, reaching a total mass gain of 7, 10 and 15 mg/cm2 after 10 min at 800, 900, and 1000 °C respectively. By contrast, the siliconised specimens show mass gains of between about 0.01 and 0.02 mg/cm2 over the same temperature interval, and thus the curves appear flat and superimposed on one another. To allow more quantitative comparison, Fig. 5 compares the oxidation rates of the substrate and siliconised samples. Each datapoint represents the overall gradient of the mass gain signal, as determined using a least-squares linear fit, after any substrate surface area corrections are applied. The substrate shows rapid increase in oxidation rate over the temperature range 600–1000 °C, i.e. from about 0.04 to 90 mg/cm2 h. The data can be well fit with two straight lines corresponding to two regimes of temperature dependence: a low temperature regime with a high activation energy of 258 ± 59 kJ/mol, and a high temperature regime with an activation energy of 46 ± 2 kJ/mol. By comparison, the siliconised samples maintain very low oxidation rates of between 0.02 and 0.08 mg/cm2 h, even up to 1150 °C, with a very low activation energy of 15 ± 9 kJ/mol. The oxidation rates are N 3 orders of magnitude lower than the substrate at the highest temperatures. As indicated by the error bars in Fig. 5, the rates of mass gain are close to the accuracy of the experiment. The siliconised samples were stable up to 1150 °C, but failed at 1200 °C – i.e. the oxidation protection was lost, as shown by an increase in oxidation rate of 2 orders of magnitude. This temperature is close to the melting point of Fe3Si7 phase at 1209 °C. 3.3. Siliconised oxide layer structure Fig. 6 shows XRD patterns from the surface of the siliconised coatings. Each pattern is taken from an oxidised sample after a 30 min

Fig. 4. Siliconisation treatment dramatically reduces composite oxidation rate. From 800 to 1000 °C, the mass gain is about 40–80 mg/cm2 h for the substrate but only 0.02– 0.04 mg/cm2 h in the siliconised state. A reduced dataset is shown for clarity, with plotpoints every 1 min.

138


dramatically in Fig. 5, the XRD pattern is well matched to FeWO4 and WO3 phases (96-900-8125 and 96-101-0619 respectively). Fig. 7 shows cross-sectional SEM micrographs of the oxidised samples after treatments at 700, 900 and 1150 °C for 30 min, compared to the as-coated state (before). The upper-most part of the coating is shown, i.e. in the as-coated state only the region of transition from mainly WSi2 bulk to mainly FeSi2 + Fe3Si7 outer crust (as shown in Fig. 3). In all three oxidised samples a layer of SiO2, between about 2 and 10 μm in thickness, is visible above the outermost iron silicide layer. The SiO2 thickness is uneven, and appears to penetrate into asperities in the outer iron silicide crust, in some cases forming encapsulated regions of SiO2 deep into it. Overall, the amount of SiO2 in the outer layer, relative to iron silicide, increases with increasing temperature; at 1150 °C the SiO2 penetrates well over half of the crust depth. This is in agreement with the increasing intensity of the (101) SiO2 peak in Fig. 6, and the increasing rate of mass gain in Fig. 5. Furthermore, pores are visible in the oxidised coatings, and the porosity increases with increasing temperature. The porosity is concentrated within the outer iron silicide crust, however at the higher oxidation temperature of 1150 °C there is also significant porosity in the WSi2 + FeSi2 bulk. 4. Discussion Fig. 5. Siliconised coatings decrease oxidation rate by 3–4 orders of magnitude in the range 800–1000 °C. The coating failed at 1200 °C, close to the melting point of the Fe3Si7 phase of 1209 °C. Error bars represent the largest deviation in slope from fits to final or initial 10% of data.

TGA test at various temperatures between 700 and 1200 °C – such as those depicted in Fig. 4. The key feature to note is the emergence of SiO2, which was best-matched to the low cristobalite polymorph (PDF 1-76-937), as revealed by the appearance of the (101) peak at 22° 2 θ. Further inspection of Fig. 6 reveals further key transitions occurring with increasing temperature in the iron silicide crust: firstly, at about 700–800 °C there is a gradual conversion from predominantly Fe3Si7 in the as-coated condition to predominantly FeSi2. Next, above about 900 °C, the Fe3Si7 is replaced by FeSi, which is accompanied by a marked increase in the SiO2 peak intensity. In none of the XRD patterns below 1200 °C was any evidence of Fe or W oxides found. However, at 1200 °C i.e. the same temperature at which the oxidation rate increases

4.1. Iron silicide segregation to surface The most important aspect of this work are the low rates of oxygen uptake in coated samples, as shown in Figs. 4 and 5. To the authors knowledge such a level of oxidation resistance is yet to be reported in any coated hardmetal. However, the chemical segregation of these coatings, displayed in Figs. 2 and 3 was unexpected, which is crucial to explaining their corrosion properties — and thus we start by discussing this feature. From the formation of an iron silicide crust that is absent of W (which constitutes 85 wt% of the substrate) it can be concluded that Fe atoms migrated preferentially towards the surface during impregnation. Analogous processes have been reported previously: surface demixing has been observed during treatment of initially homogeneous mixed oxide phases, where the more thermodynamically stable oxide will segregate to the surface [30]. More generally, production of functionally graded materials is generally categorised into two main classes of processes: (i) constructive, i.e. coating or stacking of constituents in sequence; and (ii) transport-based, where diffusional phenomena create the compositional gradients [31]. The coating microstructures shown in Figs. 2 and 3 appear to display both constructive and transport processes simultaneously, i.e. silicon is coated constructively, and iron is transported to the surface diffusionally. Such preferential migration of the binder metal during impregnation of a composite structure is therefore unusual. To explain such chemical segregation, we consider the driving forces for silicon reacting with each phase. For the carbide phase silicon displaces carbon: WC þ 2Si→WSi2 þ C:

ð2Þ

While for the binder phase silicon simply alloys with iron to form the intermetallic: Fe þ xSi→FeSix :

Fig. 6. XRD linescans of siliconised samples after oxidation treatments at various temperatures between 700 and 1200 °C for 30 min. An SiO2 peak emerges at 22° 2 θ, which increases in intensity with temperature. In addition, the surface crust evolves from predominantly Fe3Si7 as-coated, to FeSi2 at moderate temperatures, and to FeSi above 900 °C. At 1200 °C, when the coating fails, oxides of W and Fe begin to form.

ð3Þ

Before calculating the driving forces for these reactions, we first comment on the production of carbon, as predicted by Eq. (2). We note that no carbon-rich phases were detected in XRD (e.g. Fig. 3), corresponding to where WSi2 forms. We therefore speculate that carbon is either accommodated in solution, or that it precipitates out and the particle size is too small to be detected. In calculating the free energy of reaction for Eq. (2) and (3), we report in Table 1 the standard reaction quantities for the relevant compounds at room temperature. These are the free energy, enthalpy, and entropy of compound formation, denoted


139

Fig. 7. Cross-sections of the siliconised samples after 30 min TGA treatments. An SiO2 layer is clearly visible above the FeSix crust, the depth of which increases with increasing temperature.

ΔGf°, ΔHf° and ΔSf° respectively. Table 1 also reports the free energy of formation at the siliconising treatment temperature, ΔGf T, which is calculated using the equation: ΔG f T ¼ ΔH f ° −TΔS f ° ;

ð4Þ

where T = 1273 K, i.e. the temperature at which the coatings form. With these free energies of formation, ΔGf T, in mind we now consider the energetics of siliconising the WC and Fe phases in turn: • For siliconising WC, the free energy of reaction, given by Eq. (2), is equal to the difference in free energy of formation, ΔGf T, of WSi2 and WC: about −33.6 + 15.2 = −18.4 kJ/mol, or −27.6 kJ/mol of Si. • For siliconising Fe, the free energy of reaction, given by Eq. (3), is simply ΔGf T for the FeSix phase: about −35.1, −24.0 and −21.7 kJ/mol in the case of FeSi, FeSi2 and Fe3Si7, respectively, or about −70.1, −36.0 and −31.1 kJ/mol of Si. Since the free energies in the latter case (Fe) are all more strongly negative, siliconising Fe is more thermodynamically favourable than siliconising WC, regardless of the FeSix stoichiometry that is formed. The thermodynamic preference for reaction (3) is aided by faster diffusion kinetics in iron silicides; for example the silicide growth constants of metal silicides on pure Si, which are strongly related to diffusion rates, are about 3 orders of magnitude larger in FeSi than in WSi2 at 1000 K [32], and thus the overall rate of reaction between Fe and the inward diffusing Si is likely to be faster. Overall, therefore, a strong kinetic and thermodynamic bias exists for net migration of the Fe-rich binder phase to the surface. 4.2. Formation of SiO2 scale We next consider how the silicon dioxide passivating layer forms. Although the oxidation resistance of iron silicides has been reported in the past, such as Fe3Si coatings on carbon steels [33], generally coatings of high Si content are believed to be unachievable, due to poor surface Table 1 Standard reaction quantities at 298.15 K and atmospheric pressure for the compounds per mole of atoms. The quantity ΔfGT is the reaction free energy at 1273 K, as calculated using Eq. (4). Compound

ΔGf°/(J/mol)

ΔHf°/(J/mol)

ΔSf°/(J/mol)

ΔGf T/(J/mol)

Reference

WC WSi2 FeSi FeSi2 Fe3Si7

−18,664 −31,921 −32,575 −24,248 −18,036

−19,716 −31,393 −31,806 −24,333 −16,904

−3.528 1.771 2.572 −0.285 3.795

−15,225 −33,647 −35,080 −23,970 −21,735

[49] [50] [37] [37] [37]

adhesion [34]. In addition, the lack of any literature reports on siliconised hardmetals means an explicit discussion of their oxidation mechanism is absent. The lack of any iron or tungsten oxides in the XRD of Fig. 6 below 1200 °C suggests that the iron silicides are selectively oxidised, i.e. they exclusively form SiO2, thus preventing the underlying Fe from being attacked. Similarly, a variety of transition metal silicide thin films have been observed to selectively oxidise leading to passivation [35,36], i.e. without oxidation of the parent metal. To examine whether the mixed FexSi crusts characterised in Figs. 2-3, are indeed predicted to passivate, we first consider the sequence of phase transformations in the crust surface during oxidation, as shown by Fig. 6. At 700–800 °C, the replacement of Fe3Si7 with FeSi2, suggests that at low temperatures oxidation occurs via the following reaction: Fe3 Si7 þ O2 →3FeSi2 þ SiO2

ð5Þ

At higher temperatures, i.e. at 900 °C and above, the disilicide is replaced by the monosilicide, which suggests a further reaction occurs: FeSi2 þ O2 →FeSi þ SiO2 :

ð6Þ

Since these silicides are all line compounds – with limited solid stoichiometry range – and the sequence of transformation does not correlate with the temperatures of oxidation (Fe3Si7 is more stable at higher temperatures than FeSi2 [37]), it is logical to conclude that the phase transitions in the silicide crust are chemically induced, i.e. by the consumption of Si in the growing SiO2 layer and thus decrease of Si in FeSix. This conclusion is supported by observations of porosity in the outer FeSix crust, as shown in Fig. 7; the net flow of Si towards the surface would yield a corresponding flow of vacancies into the coating, which will eventually coalesce into pores. Such pores cannot be accounted for by a pore opening stress due to changes in atomic packing density of the phases, since the newly formed phases have a lower atomic packing density than the pre-oxidation ones (packing densities of the Fe3Si7, FeSi2 and FeSi phases are about 83, 80 and 72 atoms/nm3 respectively). Passivation by selective oxidation is usually explained in terms of thermodynamic driving force: if the free energy of formation of the metal oxide, ΔGf°, is sufficiently outweighed by that of the SiO2 formation, then passivation will occur [35]. Fig. 8 plots ΔGf° per mole of O2 as a function of temperature, where data are from linear fits to the Ellingham diagram shown in [38]. The figure shows the formation energy for SiO2 lies well below that for FeO or Fe2O3 over the entire temperature range of interest in this study, indicating that in an iron silicide film, there is a thermodynamic driving force for preferential oxidation

140


Fig. 8. The free energy of formation for some oxides of interest in this study. The formation energy of SiO2 is the lowest of all phases, suggesting preferential oxidation of Si is thermodynamically preferable in iron and tungsten silicides.

of Si. The energetics for WO3, taken from data in [39], are shown for comparison, indicating that preferential oxidation of Si is also predicted for tungsten silicide films (which is relevant to our discussion in the following section). 4.3. Oxidation rate: Selective vs. passive oxidation We now turn our analysis to explaining the most unexpected result of this study: the very low oxidation rates after siliconisation. Importantly, by comparing to literature reported oxidation rates we find the kinetics are significantly slower than similarly treated metallic W, i.e. WSi2-coated W, which is known to oxidise in an active manner over the same temperature range of interest [27,40]. This comparison is shown in Fig. 9a, which plots the oxidation rates for the current study

alongside results from Si-rich coatings on W alloys [27,40]. Various mole fractions of Si, x, are shown, up to a maximum of 0.66, which is directly comparable to the starting composition of coatings in the present study (i.e. the compound FeSi2). Alongside the W-Si data, we also report the evaporation rates of WO3 to allow interpretation of the oxidation mechanism. The data we reproduce is taken from literature data on vapor pressure [41], and converted into evaporation rate using the Langmuir equation. Clearly, the coatings in this study outperform siliconised W by about an order of magnitude across a broad temperature range. The apparent activation energy for oxidation is also much lower for our coatings; about 15 ± 9 kJ/mol v.s. about 80 kJ/mol for W alloys [27], suggesting different processes are limiting the rates of oxidation in each case. We attribute the enhanced oxidation resistance of siliconised WC-Fe to a fundamental difference in oxidation mechanism between the compounds WSi2 and FeSi2. On the one hand, WSi2 oxidation is known to proceed (at least at temperatures below 1300 °C) via a mechanism known as active oxidation [40]. In this mode, both W and Si are oxidised simultaneously, leading to the production of a mixed SiO2-WO3 composite scale, whereby the WO3 forms a percolating network that acts as a short circuit for oxygen diffusion. The scale is therefore unprotective and oxidation kinetics are rapid. The scale can only become protective at temperatures above 1300 °C, as shown by the large drop in the oxidation rate shown in Fig. 9a between 1200 and 1300 °C. At this point [40], WO3 evaporates from the scale at a similar rate as it is formed, as shown by the solid line in Fig. 9a, thus preventing its accumulation and allowing a continuous SiO2 layer to form (vaporisation will be present at all temperatures, however below about 1000 °C the vaporisation rate is at least an order of magnitude slower than the oxidation rate, therefore its effect on the mass gain kinetics will be small). This type of transitional behaviour is also known to occur in other refractory metal silicides such as Mo and Ta [42], although the onset of passivation varies due to different vapor pressures of the respective oxides. Comparison of the two regimes – i.e. active vs. passive oxidation – is shown schematically in Fig. 9b. In the case of siliconised W, the WO3 grains form a continuous network to the underlying WSi2, and therefore

Fig. 9. Comparison of siliconised W and WC-Fe, (a) oxidation rate of siliconised W from the literature [27], within the active oxidation regime, is typically 1–2 orders of magnitude faster than the oxidation kinetics of siliconised WC-Fe from this study. The transition from active to passive oxidation in W0.34Si0.66 coincides with the temperature at which the weight gain from oxidation is equal to the weight gain from WO3 evaporation. This is explained by a fundamental difference in the oxidation mechanism, shown in (b): WSi2 will oxidise actively, forming an unprotective WO3 + SiO2 layer, while FeSi2 oxidise passively, forming a protective SiO2 surface.


it remains unprotected, while in the case of siliconised WC-Fe, the FeSix crust allows the formation of a continuous SiO2 layer. 4.4. Property mapping for protective oxides Why then does siliconised W not oxidise passively? A simple thermodynamic argument was used to explain passive oxidation for FeSix films in this study (as shown in Fig. 8), however in the case of WSi2, passivation was also predicted on the same thermodynamic basis (as is also the case for silicides of Ti, Nb, and Ta [35]) and thus the kinetics of reaction must also be considered too. That is, oxygen transport through the SiO2 layer (i.e. consumption of Si) must be slower than diffusion of Si in the metal silicide beneath (i.e. injection of Si). These thermodynamic and kinetic criteria can be summarised as follows: 1) The protective oxide (in this case SiO2) is more stable than the parent metal oxide, as discussed in Section 4.2. 2) Diffusion of Si in the sub-stoichiometric metal silicide must be faster than diffusion of O in the SiO2 [24]. Addressing the first point, i.e. the thermodynamics of passivation, we replace our relatively simple analysis of Fig. 8, which was based only on the free energy of oxide formation, with an enthalpy of reaction, ΔHR, which includes extra information about the consumption of metal silicide at the silicide/oxide interface, as proposed by Bartur [35,43]. The reaction enthalpy is based on the following: ðy=x þ ½Þ Si þ My O ¼ y=x Mx Si þ ½ SiO2 ;

ð7Þ

where MyO is the generalised metal oxide and MxSi is the generalised silicide. The equation therefore represents the balance between two extreme cases: on the right, where protective SiO2 coating forms; and on the left, where the silicide coating dissociates and an unproductive metal oxide is generated. The enthalpy for the reaction in Eq. (7) is therefore given as [35]: ΔH R ° ¼

1 y ΔH ° SiO2 −ΔH ° My O þ ΔH ° Mx Si ; 2 x

ð8Þ

141

κ is larger, diffusion in the silicide is faster, and thus more likely to passivate; if κ is smaller, silicide diffusion will be slower and thus a mixed oxide scale is more likely. To help understand the interplay of the kinetics and thermodynamics, Fig. 10 plots the reaction heats, given by Eq. (8) on the x-axis, and the silicide growth constant, given by Eq. (9) on the y-axis, for several transition metal silicides, where data was available [32,35]. The combination of kinetic and thermodynamic data in Fig. 10 has to our knowledge not been reported before — and we therefore make some observations here. What is immediately clear from Fig. 10 is that in most cases SiO2 scale formation is thermodynamically favourable. Only Hf lies in the upper (active) region and active oxidation is indeed generally observed experimentally [44] (NB: Zr behaves similarly to Hf [45], although diffusion data were not available). The next element below Hf, which is Ti, has only a mildly negative reaction heat for passivation (lying just below the dotted line), with a magnitude of 40– 80 kJ/mol, suggesting that at low temperatures, the kinetics of Si diffusion may be insufficient to supply a passivating scale. Indeed experimentally a transition in behaviour is observed: with passivizated SiO2 scales at high temperatures and mixed oxide scales below about 900 °C [46]. We turn now to the remaining elements, i.e. those that can be more clearly placed in the thermodynamically passive region (i.e. N200 kJ/mol). Here, elements with the slowest growth constants, e.g. W, Mo, Re, and V, are expected to exhibit a transitional behaviour depending on whether kinetics or thermodynamics dominate, i.e. similar to Ti. Although data is not available for Re, this is indeed the case for Mo and W [40], as discussed above. In the case of V, recent evidence suggests that continuous Si coatings are formed, even at temperatures as low as 650 °C, although there was clear evidence of V2O5 formation in the scale [47]. For the purposes of this analysis we ignore the platinum group metals, e.g. Ru, Rh, Pt and Ir, as they are prohibitively expensive for most structural applications. The remaining elements, Fe, Ni, Co and Cr, have all been suggested as candidate binder elements for use in WC-composites and we now consider these in detail. All show relatively moderate thermodynamic driving forces (250–350 kJ/mol) and rapid diffusion kinetics. We can

where a negative value will correspond to SiO2 formation, and viceversa. Although a reaction free energy is a more preferable quantity to use, use of reaction enthalpy, according to Eq. (8), can be justified because, in general, the entropy of parent metal oxide formation is typically (i) small in comparison to the enthalpy of oxide formation, and (ii) negative, and of a similar magnitude to that for SiO2. Taking for example the case of CrSi2: the formation entropies for SiO2 (cristobalite) and CrO2 are ΔSf° = − 0.1809 and − 0.1751 kJ/mol-K respectively, and thus the difference between ΔHR° and ΔGR° for these phases is about 7 kJ/mol, which is small compared to the value of ΔHR°, which varies between about 150 and 300 kJ/mol. Thus, the general trends seen for ΔHR° will likely hold for ΔGR° as well. Addressing the second point, i.e. the kinetics of passivation, we consider the diffusion rate of Si in the silicide: Although standard data for tracer diffusivity, D*, are generally unavailable, we can report the reactive diffusion rates, as measured by the growth constant, κ, during kinetic studies of growth of silicide thin films between Si and the metal [32]: κ ¼ 2ðD =kT Þ ðΔG=xÞ expð−Q=kT Þ;

ð9Þ

where ΔG is the Gibbs free energy of formation for the silicide compound MSix, where x is the stoichiometry of Si atoms, and Q is the activation energy for the diffusion of Si. Thus, the intuitive difference between these growth constants, κ, and D* is that they are biased by the heat of compound formation. The reported κ values correspond to much faster diffusion rates than D* therefore and can not be explicitly compared to such diffusivities in an absolute scale. However, they can give some indication as to the likely relative magnitudes between different silicides, and thus their propensity to form fully protective scales: If

Fig. 10. Thermodynamic and kinetic factors in forming protective SiO2 films on silicides. The y-axis represents the thermodynamic driving force for passivation (given by Eq. (8), and as reported by Bartur [35]): a negative value predicts passive SiO2 formation; positive predicts a mixed oxide scale (active). The x-axis represents the growth constants of metal silicide thin films, as given by Eq. (9), which are dominated by Si diffusion through them: a larger value tends towards likely total oxidation; smaller towards selective oxidation. All data points represent MSi2 except those in bold, which represent MSi. Error bars represent the range in values where several oxide phases are possible.

142


therefore reasonably predict that a WC-M hardmetal, where M = Ni, Co or Cr, could selectively oxidise after siliconisation, in a similar way to the present WC-Fe hardmetal, provided a binder-rich crust forms during impregnation. Binder-rich crust formation can be rationalised based on a similar argument as in Section 4.1 concerning the heat of MSi2 formation: The formation enthalpies for M = Ni, Co and Cr are all larger, about 29, 34 and 26 kJ/mol respectively [48], than that for Fe (24 kJ/mol) [37], suggesting binder-rich crusts would form for all three. Of those, two melt congruently at temperatures at least 100 °C higher than FeSi2: 1599 and 1606 K for CoSi2 and CrSi2 respectively, suggesting that coatings of such hardmetals could potentially remain protective at even higher temperatures than are achieved in this study. However, such predictions must be validated experimentally; further studies on hardmetals with alternative binders are required. These need not be limited to impregnation of Si, but indeed other preferentially oxidising species such as Cr and Al.

5. Conclusions We have shown that WC-Fe hardmetals can be rendered oxidation resistant by diffusion impregnation with silicon. Such oxidation resistance critically relies on the compositionally two-layered nature of the siliconised coatings: The coating surface is made up exclusively of iron silicides, which is not expected based on relatively minor iron content in the substrate. Under high temperature oxidation, behaviour is dominated by this iron silicide outer layer, which passivates to form a protective SiO2 scale that is stable to 1200 °C. Unexpectedly, the kinetics of mass gain are 1–2 orders of magnitude slower than similarly coated W alloys, which tend to oxidise in an active manner over the same temperature range. By considering both kinetic and thermodynamic requirements for passivation, the improved performance in WC-Fe is explained by the relatively rapid diffusion that can occur in the FeSi2 coating, vs. more refractory coatings such as WSi2 that form on W alloys. By extending the analysis to consider a variety of transition metals, we predict that similar coatings could be manufactured from WC hardmetals employing other common binder metals, and furthermore, such coatings could be amenable to develop protective SiO2 coatings under high temperature oxidation. The results presented here offer some insight for the development of armour materials for fusion reactors, where oxidation resistance is an important requirement for reactor safety. Importantly, the improved performance of siliconised hardmetals over similarly treated tungsten alloys holds promise for their employment in fusion devices. While our results are most applicable to hardmetals, our observations of preferential de-mixing of the binder to form a tungsten-free surface, could have implications for a number of other particulate cermet systems, which combine a highly refractory ceramic skeleton within a relatively mobile metallic matrix.

Acknowledgements The authors would like to thank the EPSRC and Tokamak Energy Ltd for financial support. Funding was provided through programme grant (EP/K008749/1) Materials Systems for Extreme Environments (XMat), and through Lambert 2 agreement (TE140409-1), respectively. The authors would also like to thank J.M. Marshall of Sandvik Hyperion for providing the hardmetal sample used in this study. References [1] B. Casas, X. Ramis, M. Anglada, J.M. Salla, L. Llanes, Oxidation-induced strength degradation of WC–Co hardmetals, Int. J. Refract. Met. Hard Mater. 19 (2001) 303–309. [2] C.G. Windsor, J.G. Morgan, P.F. Buxton, Heat deposition into the superconducting central column of a spherical tokamak fusion plant, Nucl. Fusion 55 (2015) 23014.

[3] B.G. Hong, Y.S. Hwang, J.S. Kang, D.W. Lee, H.G. Joo, M. Ono, Conceptual design study of a superconducting spherical tokamak reactor with a self-consistent system analysis code, Nucl. Fusion 51 (2011). [4] L.S. Sigl, H.F. Fischmeister, On the fracture toughness of cemented carbides, Acta Metall. 36 (1988) 887–897. [5] P.K. Mehrotra, 1.07 - powder processing and green shaping A2, in: Vinod K. Sarin (Ed.), Compr. Hard Mater, Elsevier, Oxford 2014, pp. 213–235. [6] D. Maisonnier, I. Cook, S. Pierre, B. Lorenzo, B. Edgar, B. Karin, D.P. Luigi, F. Robin, G. Luciano, H. Stephan, N. Claudio, N. Prachai, P. Aldo, T. Neill, W. David, The European power plant conceptual study, Fusion Eng. Des. 75–79 (2005) 1173–1179. [7] V.B. Voitovich, V.V. Sverdel, R.F. Voitovich, E.I. Golovko, Oxidation of WC-Co, WC-Ni and WC-Co-Ni hard metals in the temperature range 500–800 °C, Int. J. Refract. Met. Hard Mater. 14 (1996) 289–295. [8] S.N. Basu, V.K. Sarin, Oxidation behavior of WC-Co, Mater. Sci. Eng. A 209 (1996) 206–212. [9] L. Del Campo, R.B. Pérez-Sáez, L. González-Fernández, M.J. Tello, Kinetics inversion in isothermal oxidation of uncoated WC-based carbides between 450 and 800°C, Corros. Sci. 51 (2009) 707–712. [10] M. Victoria, N. Baluc, P. Spätig, Structural materials for fusion reactors, Nucl. Fusion 41 (2001) 1047. [11] S. Huang, J. Xiong, Z. Guo, W. Wan, L. Tang, H. Zhong, W. Zhou, B. Wang, Oxidation of WC-TiC-TaC-Co hard materials at relatively low temperature, Int. J. Refract. Met. Hard Mater. 48 (2015) 134–140. [12] C. Barbatti, J. Garcia, P. Brito, A.R. Pyzalla, Influence of WC replacement by TiC and (Ta, Nb) C on the oxidation resistance of Co-based cemented carbides, Int. J. Refract. Met. Hard Mater. 27 (2009) 768–776. [13] S.K. Bhaumik, R. Balasubramaniam, G.S. Upadhyaya, M.L. Vaidya, Oxidation behaviour of hard and binder phase modified WC-10Co cemented carbides, J. Mater. Sci. Lett. 11 (1992) 1457–1459. [14] M. Jafari, M.H. Enayati, M. Salehi, S.M. Nahvi, C.G. Park, Comparison between oxidation kinetics of HVOF sprayed WC–12Co and WC–10Co–4Cr coatings, Int. J. Refract. Met. Hard Mater. 41 (2013) 78–84. [15] V.A. Tracey, Nickel in hardmetals, Int. J. Refract. Met. Hard Mater. 11 (1992) 137–149. [16] F. Lofaj, Y.S. Kaganovskii, Kinetics of WC-Co oxidation accompanied by swelling, J. Mater. Sci. 30 (1995) 1811–1817. [17] M. Aristizabal, J.M. Sanchez, N. Rodriguez, F. Ibarreta, R. Martinez, Comparison of the oxidation behaviour of WC–Co and WC–Ni–Co–Cr cemented carbides, Corros. Sci. 53 (2011) 2754–2760. [18] S.A. Humphry-Baker, W.E. Lee, Tungsten carbide is more oxidation resistant than tungsten when processed to full density, Scr. Mater. 116 (2016) 67–70. [19] L. Prakash, 1.02 - fundamentals and general applications of hardmetals A2, in: Vinod K. Sarin (Ed.), Compr. Hard Mater, Elsevier, Oxford 2014, pp. 29–90. [20] B. Wittmann, W.D. Schubert, B. Lux, Hardmetals with iron-based binder, Eur. 2002 Lausanne Hard Mater. Proc 2002, pp. 303–308. [21] D. Mercurio-Lavaud, K. Petrov, P. Peshev, Z. Zakhariev, The elemental and phase composition of boride coatings deposited by diffusion on a WC-Co alloy, J. Common Met. 67 (1979) 59–63. [22] Z.T. Zahariev, M.I. Marinov, Superhard boride layer deposition on a carbide-cobalt hard alloy, J. Alloys Compd. 201 (1993) 1–3. [23] P. Peshev, Z. Zakhariev, K. Petrov, The composition and oxidation resistance of coatings deposited on hard carbide alloys using TiB2, J. Common Met. 67 (1979) 351–356. [24] H.W. Grünling, R. Bauer, The role of silicon in corrosion-resistant high temperature coatings, Thin Solid Films 95 (1982) 3–20. [25] G.H. Meier, Research on oxidation and embrittlement of intermetallic compounds in the US, Oxid. Intermet. (2008) 15–58. [26] S.N. Basu, G.J. Yurek, Effect of alloy grain size and silicon content on the oxidation of austenitic Fe-Cr-Ni-Mn-Si alloys in pure O2, Oxid. Met. 36 (1991) 281–315. [27] F. Koch, H. Bolt, Self passivating W-based alloys as plasma facing material for nuclear fusion, Phys. Scr. 2007 (2007) 100. [28] N. Chaia, S. Mathieu, T. Cozzika, F. Rouillard, C. Desgranges, J.L. Courouau, C. Petitjean, N. David, M. Vilasi, An overview of the oxidation performance of silicide diffusion coatings for vanadium-based alloys for generation IV reactors, Corros. Sci. 66 (2013) 285–291. [29] C.M. Fernandes, A.M.R. Senos, Cemented carbide phase diagrams: a review, Int. J. Refract. Met. Hard Mater. 29 (2011) 405–418. [30] H. Schmalzried, W. Laqua, Multicomponent oxides in oxygen potential gradients, Oxid. Met. 15 (1981) 339–353. [31] A. Mortensen, S. Suresh, Functionally graded metals and metal-ceramic composites: part 1 processing, Int. Mater. Rev. 40 (1995) 239–265. [32] P. Gas, F.M. d'Heurle, 4 diffusion in silicides, in: D.L. Beke (Ed.), Diffus. Semicond, Springer, Berlin Heidelberg 1998, pp. 1–38. [33] A.L. Cabrera, J.F. Kirner, Formation of silicon diffusion coatings on ferrous alloys from their reaction with silane, Surf. Coat. Technol. 39 (1989) 43–51. [34] M.A. Harper, R.A. Rapp, Codeposited chromium and silicon diffusion coatings for FeBase alloys via pack cementation, Oxid. Met. 42 (1994) 303–333. [35] M. Bartur, Thermal oxidation of transition metal silicides: the role of mass transport, Thin Solid Films 107 (1983) 55–65. [36] H. Jiang, C.S. Petersson, M.-A. Nicolet, Thermal oxidation of transition metal silicides, Thin Solid Films 140 (1986) 115–130. [37] P. Franke, D. Neuschütz, P. Franke, D. Neuschütz, Binary Systems. Part 3: Binary Systems From Cs-K to Mg-Zr · Fe-Si: Datasheet From Landolt-Börnstein - Group IV Physical Chemistry · Volume 19B3: “Binary systems. Part 3: Binary Systems from Cs-K to Mg-Zr” in SpringerMaterials ((http://dx.doi.org/10.1007/10757413_43), Springer-Verlag Berlin Heidelberg, n.d. http://materials.springer.com/lb/docs/sm_ lbs_978-3-540-31688-6_43).

S.A. Humphry-Baker et al. / Int. Journal of Refractory Metals and Hard Materials 66 (2017) 135–143 [38] S.R.J. Saunders, J.R. Nicholls, Chapter 14 - Oxidation, hot corrosion and protection of metallic materials A2, in: Robert W. Cahn, P. Haasen (Eds.), Phys. Metall. Fourth Revis. Enhanc. Ed, Oxford, North-Holland 1996, pp. 1291–1361. [39] Thermodynamic Properties of Compounds, RuO3 to Ti3O5: Datasheet from LandoltBörnstein - Group IV Physical Chemistry · Volume 19A4: “Pure Substances. Part 4 _ Compounds from HgH_g to ZnTe_g” in SpringerMaterials ((http://dx.doi.org/10. 1007/10688868_16), Springer-Verlag Berlin Heidelberg, n.d. http://materials. springer.com/lb/docs/sm_lbs_978-3-540-45274-4_16). [40] H.-S. Kim, J.-K. Yoon, G.-H. Kim, J.-M. Doh, S.-I. Kwun, K.-T. Hong, Growth behavior and microstructure of oxide scales grown on WSi2 coating, Intermetallics 16 (2008) 360–372. [41] P.E. Blackburn, M. Hoch, H.L. Johnston, The vaporization of molybdenum and tungsten oxides, J. Phys. Chem. 62 (1958) 769–773. [42] D.A. Berztiss, R.R. Cerchiara, E.A. Gulbransen, F.S. Pettit, G.H. Meier, Oxidation of MoSi2 and comparison with other silicide materials, Mater. Sci. Eng. A 155 (1992) 165–181. [43] M. Bartur, M.-A. Nicolet, Thermal oxidation of transition metal silicides on Si: summary, J. Electrochem. Soc. 131 (1984) 371–375. [44] S.P. Murarka, C.C. Chang, Thermal oxidation of hafnium silicide films on silicon, Appl. Phys. Lett. 37 (1980) 639–641.

143

[45] W.J. Strydom, J.C. Lombaard, R. Pretorius, Thermal oxidation of the silicides CoSi2, CrSi2, NiSi2, PtSi, TiSi2 and ZrSi2, Thin Solid Films 131 (1985) 215–231. [46] F. d'Heurle, E.A. Irene, C.Y. Ting, Oxidation of silicide thin films: TiSi2, Appl. Phys. Lett. 42 (1983) 361–363. [47] S. Mathieu, N. Chaia, M. Le Flem, M. Vilasi, Multi-layered silicides coating for vanadium alloys for generation IV reactors, Surf. Coat. Technol. 206 (2012) 4594–4600. [48] M.E. Schlesinger, Thermodynamics of solid transition-metal silicides, Chem. Rev. 90 (1990) 607–628. [49] P. Franke, D. Neuschütz, P. Franke, D. Neuschütz, Binary systems. Part 2: Elements and Binary Systems from B – C to Cr – Zr · C-W_: Datasheet from Landolt-Börnstein - Group IV Physical Chemistry · Volume 19B2: “Binary systems. Part 2: Elements and Binary Systems from B – C to Cr – Zr” in SpringerMaterials ((http://dx.doi. org/10.1007/10757405_49), Springer-Verlag Berlin Heidelberg, n.d. http://materials.springer.com/lb/docs/sm_lbs_978-3-540-45201-0_49). [50] B. Predel, O. Madelung, Si-W (Silicon-Tungsten): Datasheet from Landolt-Börnstein - Group IV Physical Chemistry · Volume 5J: “Pu-Re – Zn-Zr” in SpringerMaterials ((http://dx.doi.org/10.1007/10551312_2742), Springer-Verlag Berlin Heidelberg, n.d. http://materials.springer.com/lb/docs/sm_lbs_978-3-540-70705-9_2742).