Born 1958, in Washington DC, USA.
Professor of Chemistry, Department of Chemistry, Queen's University, Canada.
Web: external link
NSERC of Canada University Research Fellow, 1988-1997; Alexander von Humboldt Stiftung Fellow, 1994-1995; Noranda Lecture Award of the Canadian Society for Chemistry, 1999; Fellow of the Chemical Institute of Canada, 1999; Canada Research Chair (Tier I), Aug. 2007-present; Fellow of the American Physical Society (chemical physics division), 2007; Gerhard Herzberg award of the Canadian Society for Analytical Sciences and Spectroscopy, 2013; John C. Polanyi Award of the Canadian Society for Chemistry, 6.2014; Alexander von Humboldt Research Award, 2017-2019
More than 205 scientific papers.
Carrington was an early proponent of time-independent iterative methods for calculating ro-vibrational spectra. He demonstrated that by exploiting the structure of commonly used basis sets, quadrature grids, and molecular Hamiltonians it is possible to efficiently use iterative eigensolvers to compute spectra. A key advantage of this general approach is that it obviates calculation of multi-dimensional integrals.
By using Smolyak's recipe for combining 1D quadratures with nested 1D quadratures that he optimises for problems of quantum dynamics, Carrington and his group have developed a practical strategy for avoiding direct product quadratures when solving the Schrödinger equation. Structure is exploited to evaluate matrix-vector products efficiently by doing sums sequentially.
Carrington and his group demonstrated that it is possible to obtain accurate potential energy surfaces from neural networks. They showed that a neural network can be designed so that the final potential energy surface is a sum of products, advantageous for several methods of quantum dynamics.
Carrington's group has developed contracted basis set methods that make it possible use contracted basis sets in conjunction with iterative methods. The crucial idea is to avoid transforming back to the original basis. This obviates the need to store a vector in a primitive product basis. There are two key ideas. 1) The contracted basis functions do not have a shared index. 2) Rather than storing values of the potential on a full-dimensional grid, one stores instead a potential matrix in an intermediate basis of products of contracted functions and discrete variable representation functions. These ideas have been applied to CH5+, methane, the water dimer, and other molecules.
Carrington and his group developed an iterative method for computing vibrational spectra using tensor rank reduction that requires only MBs of memory. The memory cost of the method scales linearly with the number of atoms.