Skip to main content

Graham D. Fletcher

Computational Scientist

Graham Fletcher has a background in quantum chemistry and supercomputing. His research interests focus on the development of highly scalable methods and algorithms for the accurate and reliable prediction of chemical phenomena.

Biography

Graham Fletcher has a background in quantum chemistry and supercomputing. His research interests focus on the development of highly scalable methods and algorithms for the accurate and reliable prediction of chemical phenomena. He gained a Ph.D. (physics) from the University of York, UK, under Graham Doggett, who worked with Coulson on molecular orbital theory. Graham has extensive experience in high performance computing, and has worked at Iowa State University, NASA Ames Research Center, and Daresbury Laboratory in the UK. More recently, he is the originator of the Variational Subspace Valence Bond (VSVB) method.[1]

VSVB is a high-accuracy ab initio electronic structure method, and the only general, scalable alternative to molecular orbital theory. VSVB obviates the famous N! problem’ allowing wave functions to be constructed from objects familiar to chemists such as bonds and lone pairs’. Indeed, the VSVB wave function can be built lego-style’ from pre-defined orbitals,[2] obviating the costly optimization step required in mainstream QM methods. VSVB can model many classes of chemical problem in a single method (ground and excited states; open- and closed-shell systems; bond-breaking/formation; resonance; etc) while achieving near-linear concurrency to 1 million ranks on leadership computers.[2]

[1] Graham D. Fletcher, The variational subspace valence bond method”, J. Chem. Phys. 142, 134112 (2015). http://​dx​.doi​.org/​1​0​.​1​0​6​3​/​1​.​4​9​16743 

[2] Graham D. Fletcher, Colleen Bertoni, Murat Keçeli, Michael D’Mello, Valence: A Massively Parallel Implementation of the Variational Subspace Valence Bond Method”, J. Comput. Chem., Software News and Updates, 40, 1664 (2019). https://​doi​.org/​1​0​.​1​0​0​2​/​j​c​c​.​25818