The development of new molecules and materials capable of catalysis such as reduction of CO2 is a major challenge. In order to make progress towards rational design of catalysts and photocatalysts, we need an efficient simulation framework that includes quantum mechanical effects to describe bond rearrangement and charge transfer while simultaneously being able to address conformational flexibility and sampling issues.
We aim to develop such a framework that will be tailored to upcoming large-scale resources that make extensive use of graphical processing units (GPUs). In order to accomplish this, we have assembled a team representing applied mathematics, computer science, chemistry, biochemical simulation, and protein design. We will develop new algorithms that exploit both element and rank sparsity in high-order tensors representing electronic wavefunctions or interaction operators. These new algorithms will be implemented in the context of domain specific languages and parallel frameworks targeting next-generation architectures as exemplified by the Sierra, Summit and Aurora computer systems that will come online in the next few years. We will use these new algorithms and frameworks, as implemented on large-scale computational resources, in order to launch a program targeting design of enzymatic and/or organometallic systems for photocatalytic reduction of CO2.
The catalyst design process enabled will dovetail with experiments performed at BES user facilities. Ultrafast characterization of photocatalysts is a leading use case for LCLS and LCLS-II, and our team is well-positioned to translate the computational technologies developed in this proposed work directly into high impact experiments conducted at these and other facilities.