Download files
Abstract
Solid oxide fuel cells operate at elevated temperature, oxidising fuel gases to generate electricity. The fuel gas streams in the fuel cell systems are rich in carbon and have very low oxygen potential. Under these conditions, alloys can undergo metal dusting, which causes pitting or general thinning of the alloys. This process is not yet fully understood. It is, hence, not possible to accurately predict the susceptibility of a particular alloy in the atmospheres relevant to SOFC.
Model Fe-Cr and Fe-Ni-Cr alloys were exposed to test the hypothesis that cementite formation and its decomposition is necessary for metal dusting to occur. A series of ferritic and austenitic engineering alloys were also exposed to compare their dusting rates. Two specimens of each alloy were studied, one was etched in a H3PO4-15%H2SO4-21%H2O solution and the other was ground to a 600-grit finish. The alloys were exposed to a CO-26%H2-6%H2O gas mixture at 680oC under thermal cyclic conditions. The hot gas composition corresponded to ac = 2.9 and an oxygen potential high enough to oxidise chromium, but not iron or nickel.
All the alloys were shown to undergo internal carburisation, metal dusting and coking once the protective chromium oxide scale was damaged. Fe-25Cr was less resistant than Fe-60Cr because of its lower chromium content. However, ferritic Fe-25Cr-based steels are more resistant to dusting than austenitic Fe-25Cr-25Ni. The present findings are consistent with the earlier conclusions that cementite formation is essential for dusting on ferritic steels and that dusting of austenitic alloys does not involve the prior formation of cementite and its decomposition.
The onset of metal dusting was more accelerated for most austenitic engineering alloys (Alloy 800, Inconel 601, 690, 693 and Alloy 602CA) than for engineering ferritic steels (Fe-27Cr-0.001Y). However, the alloy with the best performance was austenitic Inconel 625, which was still protected by its Cr2O3 scale after 500 one-hour cycles. In both ferritic and austenitic chromia-formers alloys, the surface ground specimens were more resistant to metal dusting than the electropolished specimens. In contrast, ferritic alumina-formers with electropolished surfaces did not dust during the entire experimental periods of 1200 one-hour cycle, but the alloys with ground surfaces slowly underwent dusting attack.
The coke deposits formed consisted largely of graphite nanotubes, containing small particles at the tube tips. These particles were identified as single crystal cementite, in the case of ferritic steels, and austenite, for the austenitic alloys. This is not in agreement with the currently accepted dusting model for ferritic steels that cementite decomposition yields iron particles, which catalyse coke deposition. EDX analysis of the cementite particles, showed that the only metal detected was iron, thus differing in chemistry from the (Fe,Cr)3C surface layer. Similarly, the austenite particles contained only nickel and iron, differing in chemistry from the disintegrated alloy surfaces. These results suggested that the particles were formed in the coke in the carbon-supersaturated gas, rather than disintegration of the alloy surface layer. Strong orientation relationships were determined between the graphite and cementite particles; however, no clear crystallographic relationship was deduced between the graphite and austenite.
Relative alloy performance appears consistent using the present multiple one-hour cycle and the results of others using a smaller number of lengthy cycles. Hourly thermal cycling was shown to accelerate the dusting onset for both electropolished chromia-formers and surface ground alumina-formers.
Protective oxide scales spall at a critical thickness and carbon attack results when the alloy surfaces are depleted of scale-forming elements and healing becomes impossible. On this basis, analytical models were developed and used to predict the incubation periods for oxide failure and the subsequent carbon attack. Upon testing, these models were, however, found to be not qualitatively meaningful in predicting the onset of dusting observed in the present study. Gross oversimplifications involved in the model and the absence of reliable data for many parameters required for the computations prevented even an approximate quantitative prediction.