The annular sample geometry produced some interesting and unexpected oxide growth behavior that would not have been observed on planar coupons. Each topic associated with this study is discussed in turn, and some important conclusions are drawn from data obtained and observations made.
Mass Change Data and Growth Geometry
The annular sample geometry presents challenges for a detailed quantitative analysis of mass change. Our observations regarding OD surface oxidation of the annulus do, however, support an unreacted shrinking core model, at least up to 850 °C. This requires a non-porous solid reactant, which can be claimed from integrity of the high-magnification SEM microstructures. The oxide layer is porous and contains cracks and fissures, produced by the volume change in transforming WC into WO3, generating a factor of three difference in CTE in addition to formation of carbon dioxide. With the annular geometry, there are clearly additional stresses placed on the OD surface oxide, from radial growth on a convex surface, when compared to growth on the planar surface of a coupon.
Figures 3, 4, 5, 6, 7, 8, and 9 present changes in oxide growth geometry with temperature. One important feature of these changes is behavior of growth on the ID surface. Oxide formation is thought to be suppressed by generation of compressive stresses in the oxide layer, which tend to close fissures, isolate pores, and so restrict free airflow to the reacting metal surface. On the ID of the annulus, the oxide becomes more protective and the shrinking core oxidation process is shut down. The illustrations over the 700–1100 °C temperature range exhibit limited oxide formation on the ID of the annulus samples, as compared to that found on the OD surface.
Oxide growth geometry on the OD surfaces changes with increasing exposure temperature. After 1 h at 700 °C, a thin, rigid layer peels or spalls off exposed surfaces, including perfect “washer” shapes from the sides of the annulus. Even the chamfered edges generate their own thin rings of oxide. At this temperature, and at 850 °C, this geometrically structured growth, perpendicular from all OD surfaces, continues. Basu and Sarin  explain this behavior by noting the tendency for WO3 to grow in a columnar mode, thus restricting lateral expansion. However, there is a departure from this form at 1000 °C. Oxide formation is no longer unidirectional, and for the annulus, this resulted in growth of a conically shaped oxide formation from the planar ends and formation of a “bobbin” geometry from the curved OD surface of the annulus.
This change in oxidation mode may be related to a change in oxides present. Below 800 °C, WO3 and CoWO4 are the only oxides formed [1, 2]. At 850 °C, columnar-controlled growth is still evident (Figs. 3 and 5). At higher temperatures, oxidation reactions in air may produce cobalto-cobaltic oxide (Co3O4). However, at ~ 950 °C, this compound breaks down into CoO and O2. The complexities introduced by these potential changes in oxide compound formation are significant. Together with an increased probability of direct CoO formation, they may account for the loss of columnar growth control and concomitant shape conformance to the underlying metal surface. An increase in CO2 will exacerbate porosity and defects and so allow more open access of air to the oxidation reaction zone. In addition, freedom for lateral oxide growth may also explain the significant increase in oxidation rate.
Scanning Electron Microscopy
Selective SEM studies of the sample set were conducted. Figure 13, taken at 20kX, depicts the very small grain size and the presence of small, angular WC particles. EDS data (Fig. 12) support the composition as 94% W, 6% Co, but with an apparent addition of 0.5% Cr.
The oxide formed at 700 °C is porous. Although the ID oxide separated from the cermet substrate after a 3-h exposure to air at 700 °C (Fig. 12), linescan data presented in Fig. 14 indicate no evidence for oxygen diffusion into the base material. Oxide formation and growth take place on the metal surface by allowing air to flood the oxide/cermet interface via linked pores and other defects. It is only during cooling down that oxide breakaway occurs at the reaction interface, due to the large difference in CTE. At this temperature, there is not a great deal of difference in oxide growth rate on the ID and OD after a 3-h exposure, so that oxide layers formed are of a similar thickness. These results support the unreacted shrinking core model for oxidation at 700 °C.
All oxidized samples, to some degree, show a dark-blue outer band within the OD oxidized layer. This is exhibited in very clear form by Sample T507-C, exposed for 3 h @850 °C, (Fig. 5). A sector of this sample was removed for SEM evaluation. EDS data in Fig. 15 show that concentration of Co in this outer region of the oxide (i.e., zone of oxide that formed first) contains about twice the level of Co as later (inner) oxide growth. It was also noted that:
It was concluded that the samples picked up moisture when exposed to air after removal from the furnace and formed a dark-blue compound [Co(H2O)4(OH)2], predominantly on exposed surfaces in the area with higher Co content. Exposure in the SEM was sufficient to dehydrate this compound and remove the coloration. The reason why initial oxide formation is rich in Co requires further investigation.
The bushing sample oxidized at 1000 °C was examined in detail, (Figs. 16, 17, 18, 19, and 20). A sector of the bobbin-oxide structure was carefully broken from the substrate for evaluation in the SEM. Its removal revealed a spiral pattern on the cermet surface, at the growth front of the oxide layer. The carbon map (Fig. 17d) clearly indicates oxidation of WC is a dominant factor in spiral formation. Distribution densities appear to be identical for W and Co, so it is not clear whether Co is important in this growth mode, or whether the Co map simply reflects the intimate mixture of W and Co in the cermet.
Spiral formations occur in a few binary alloy eutectic structures (e.g., Al–Mg, Zn–Mg), but these develop from growth anisotropy in the liquid phase. In this case, reactions are chemical between a gas and a solid. A more rigorous investigation is needed to determine how stresses generated by contiguous W–O, Co–O, and W-Co–O compound formation lead to this spiral structure on the reaction interface. In addition, it should be noted that this apparent 2D “spiral” may be a 3D helix, which becomes an integral part of the basal oxide structure developed over time. Such helix formations may be limited by continued oxidation of WC on the helical bands, so that they simply become consumed as the oxide layer grows away from the reaction surface. An initial helical growth mode, however, would help to explain the departure from columnar propagation established at lower temperatures.
Linescan data were acquired from a mounted and polished sample exposed at 1000 °C/1 h. The etched microstructure shown in Fig. 18a exhibits a striking ruby-red band close to the ID reaction interface, but just inside the oxide layer. An investigation of color compounds of W and Co ruled out the possibility that this was an oxide. It is more likely that this is a color generated during metallographic preparation. Water reaction with cobalt can produce [Co(H2O)6]2+ ions. The reason why Co in this location, just within the oxide layer, is so sensitive to reaction with water is not known. This reaction was not observed elsewhere.
Light microscope and SEM images, depicting reactions at 1000 °C (Fig. 18), suggest that inward diffusion of oxygen has taken place on the ID surface. Linescan data (Fig. 19) produced a very different picture from the one obtained at 700 °C. Within the reaction zone, several bands, apparently containing different compounds (A, B, and C. D is substrate), are discernible on the backscattered electron image. Zone C, with a thickness of 120 µm, does not show clear indications of element diffusion, but there is definite substrate bonding, substantial enough to resist oxide layer separation on cooling.
Zone C is also notable for the flat cobalt profile. Neither binder element, Co and Cr, was detected in this zone, indicating concentrations of less than 1.0 wt. %. In contrast, oxygen is at its highest level of the four zones and has clearly reacted with tungsten to form WO3.
It would appear that the A/B Zone boundary may have been at the original ID surface. The following reactions are taking place at 1000 °C:
Zone A: Oxygen reacts with the surface to form WO3 and CoWO4.
Zone B: Oxygen diffuses into the bulk material, principally along the WC/binder interfaces. Outward diffusion of binder (Co and Cr) contributes to the formation of CoO and CoWO4 (and possibly CoCrO4).
Zone C: This zone is depleted of binder, but inward diffusion of oxygen generates a band of WO3. Near the interface with Zone D, a layer sufficiently rich in Co reacts with water to produce a red color on the metallographic mount.
Zone D: Chemistry abruptly changes to original matrix. There is very little evidence of any element diffusion (in or out) beyond the C/D boundary.
This complex oxide surface structure has formed in the constricting environment of the annulus ID. Development of compressive stress, produced by the natural tendency for expansion as tungsten oxide is formed, has closed cracks and other major oxide defects. Inward diffusion of oxygen creates three distinct layers as it meets zones with different chemistry. At 1000C, it appears that diffusion of the Co binder phase has been activated across distinct zones (Zones C → B, and A in Fig. 19). Throughout the zones, WC reacts “in situ” allowing oxygen to move inward along the major phase interfaces. After 1 h at 1,000 °C, the combined zones have developed to a thickness of 300 µm. Overall porosity, chiefly resulting from WO3 formation, allows CO2 to escape without creating major defects in the layered structure.
A linescan across the OD surface oxide (remnant), shown in Fig. 20, presents element reaction profiles consistent with the unreacted shrinking core mechanism. There is one clearly defined reaction interface, which separates oxidized substrate from the unoxidized core. These data are in sharp contrast to results obtained from the reaction zone on the ID surface of the annulus. In addition, as illustrated in Fig. 7, the OD oxide has grown to a thickness of about 5 mm in 1 h. A summary of physical and chemical reactions occurring over the 700–1100 °C temperature range is presented in Table 4.
The phenomenon of “oxide jacking” is well known for damage inflicted by stresses generated during oxide growth and the accompanying volume expansion, on ferritic steel. Damage in structures of all kinds has been documented, from historic buildings to bridges, and even to granite countertops. In a confined space, oxidation of iron proceeds with the interdiffusion of oxygen into the steel and iron diffusion into the oxide. A large volume expansion of the oxide accompanies this process, and according to Harris, an early protagonist of this process, “…stresses generated are of sufficient magnitude to deform or fracture all known materials” .
Although volume expansion accompanying oxide growth of tungsten carbide is substantial, it would appear that in a constricted space at elevated temperature, the oxidation process is arrested or slowed down, due to inability of oxygen to reach the unreacted WC surface. The magnitude of compressive stress required to achieve this is not known. However, the overriding advantage of the shrinking core oxidation mechanism is that it only requires ingress of oxygen to the unreacted surface to be stopped. This is achieved by the self-generated pressure of oxide growth, which seals off fissures and other defect diffusion pathways in the oxide layer. In contrast to the oxide jacking process at ambient temperatures, oxide formation on a WC cermet in an ID annular space, for example, can become protective at very high temperatures.
Although the bushing sample geometry is not ideally suited for making accurate mass change measurements as a function of oxidation time and exposure temperature, an attempt to determine activation energy for the reaction was made. Employing the parabolic rate law for oxidation of WC-6% Co over the temperature range of 600–850 °C to obtain a rate constant and then fitting this into the Arrhenius equation, an activation energy of 259 kJ/mol was obtained.
Within the limits of tools available for this study, our results support an unreacted shrinking core model for oxidation of WC-6% Co cermet. These experiments have expanded the oxidation reaction timescale and increased the temperature range. It is understood that in relation to oxidation experienced by tool bits, exposure periods in terms of minutes are more appropriate. However, observations made on the small bushing may lead to use of WC cermets in high-temperature bearings or other applications where annular geometry is utilized.