The CRF at Sandia has long been at the forefront of efforts to describe reacting flow processes in terms of elementary, as opposed to globalized, chemical reactions. This focus led to the development of the Chemkin software and associated application codes for various reactor configurations. Although these capabilities were initially applied to homogeneous gas reactions such as combustion, they are equally useful in simulating surface chemistry, in particular the catalytic reactions that are of enormous importance in fields such as chemical synthesis and exhaust-gas aftertreatment. We are currently using these tools to develop elementary reaction mechanisms for emerging technologies that remove nitrogen oxides (NOx) from the exhaust of lean-burn engines such as diesels. This longstanding problem cannot be addressed with conventional catalytic converters and is central to meeting increasingly stringent air quality standards. We believe that understanding the underlying chemistry at a fundamental level will allow these new technologies to be used in the most effective way possible.
Because the first-principles prediction of rate constants for catalytic reactions is very difficult, these parameters are often obtained by matching the results of reactor simulations to the results of carefully planned experiments. This can be a very challenging computational problem, due to both the time involved in carrying out a single simulation and the large number of parameters whose values must be extracted. At Sandia, we have taken advantage of a unique set of capabilities in applying this approach to the development of reaction mechanisms for an aftertreatment technology known as the lean NOx trap (LNT). These capabilities include our longstanding experience with the Chemkin software, our recent development of a Chemkin-based transient plug flow reactor code, the availability of efficient large-scale optimization software such as APPSPACK (developed at Sandia) to handle the adjustment of parameters, and access to state-of-the-art massively parallel computers such as Red Sky. These have allowed us to construct a 45-step reaction mechanism, complete with values for all of the kinetic parameters, that accurately describes the normal functioning of an LNT under a wide variety of conditions. In addition, we have developed an auxiliary 11-step mechanism that describes the sulfur poisoning of an LNT as well as its high-temperature regeneration. We are continuing to use this approach to address other facets of LNT operation and plan to apply it to additional aftertreatment technologies as well