Skip to content

Chemistry Seminar Series 291 (9/21/18)

September 21, 2018 - 3:00pm

Anthony Dutoi, University of the Pacific

Abstract:

Many chemical problems contain electrons that are doing something interesting. This includes bond breaking, which is at the heart of chemistry, as well as electrons being excited or transferred in photophysical processes. Sometimes the need for high energetic accuracy alone motivates detailed descriptions of electronic wavefunctions. For all of these reasons, and in spite of the apparent obstacles to doing so, the electronic structure community still seeks methods that can be systematically improved to arbitrary accuracy, and which are applicable to large systems. This talk will present work on a fresh approach to this problem, which is showing a lot of promise in initial tests.The well-worn motif of dividing a super-system into chemically meaningful fragments is adopted. However, a shortcoming of existing fragment-based methods is that electron correlation (if included) is modeled at the level of individual electrons, effectively recomputing local many-body fluctuations for each separate inter-fragment interaction. We show that, without approximation, the ab initio Hamiltonian may be rewritten in terms of sub-system fluctuations between locally correlated electronic states (the excitonic basis). This not only reduces the computational cost of conventional highly correlated electronic structure methods, but also the scaling.Using this foundation, we developed a working excitonic coupled-cluster (X-CC) code. We have acquired promising results on test systems of chains of Be atoms (the dimer is already a notoriously difficult multiconfigurational, semicovalent problem). In a toy 6-31G basis, our pilot code can handle 100 Be atoms (200 active electrons) in about 11 minutes, achieving a total dissociation energy comparable to conventional CCSD 
(coupled-cluster with single and double substitutions). A calculation of this size is challenging to even complete with existing production-level CCSD codes. Upcoming real-world applications and extensions of this work will be discussed alongside the conceptual aspects of the theory. 

Flyer PDF: dutoi_anthony_chemistry_flyer.pdf