Multi‐level quantum monte Carlo wave functions for complex reactions: The decomposition of α‐hydroxy‐dimethylnitrosamine

Abstract

We present here several novel features of our recently proposed Jastrow linear generalized valence bond (J-LGVB) wave functions, which allow a consistently accurate description of complex potential energy surfaces (PES) of medium-large systems within quantum Monte Carlo (QMC). In particular, we develop a multilevel scheme to treat different regions of the molecule at different levels of the theory. As prototypical study case, we investigate the decomposition of α-hydroxy-dimethylnitrosamine, a carcinogenic metabolite of dimethylnitrosamine (NDMA), through a two-step mechanism of isomerization followed by a retro-ene reaction. We compute a reliable reaction path with the quadratic configuration interaction method and employ QMC for the calculation of the electronic energies. We show that the use of multideterminantal wave functions is very important to correctly describe the critical points of this PES within QMC, and that our multilevel J-LGVB approach is an effective tool to significantly reduce the cost of QMC calculations without loss of accuracy. As regards the complex PES of α-hydroxy-dimethylnitrosamine, the accurate energies computed with our approach allows us to confirm the validity of the two-step reaction mechanism of decomposition originally proposed within density functional theory, but with some important differences in the barrier heights of the individual steps.

Accurate barrier heights using diffusion Monte Carlo

“…There have been only a few tests of the performance of QMC methods for reaction barriers. DMC studies of reaction barriers have been calculated for H + H 2 [16][17][18], several organic molecules [19][20][21][22][23], surface reactions [24][25][26], and others [27,28]. These have generally found that DMC can get close to or within chemical accuracy for reaction barriers, often improving on DFT results.…”

Multiple-Resonance Local Wave Functions for Accurate Excited States in Quantum Monte Carlo

“…To illustrate the potential of a local approach in QMC, we briefly describe the linear-scaling scheme we recently introduced for ground states, which couples electron pairs locally to progressively construct wave functions of increasing quality. − In our approach, the occupied orbitals in the reference are localized over one or two centers along with the corresponding antibonding orbitals in the virtual space. The determinants are then generated by correlating two pairs of electrons sitting in adjacent bonding orbitals and the corresponding two antibonding orbitals in a so-called CAS­(4,4) expansion (i.e., by including the excitations in the active space obtained by correlating four electrons in the two bonding and two antibonding orbitals).…”

“…The availability of accurate Jastrow–Slater wave functions with a small and transferable determinantal component is therefore highly desirable. In previous work with Fracchia, we have shown that this is possible with the use of orbitals localized over one or two centers and introduced a local correlation scheme in QMC to progressively construct accurate and compact multideterminant wave functions of increasing quality, which were successfully applied to the ground state of closed and open shell systems. − By correlating two pairs of electrons sitting in adjacent bonding orbitals with the corresponding antibonding orbitals, we achieved linear scaling and, consequently, a significant computational gain with respect to other local expansions also employed in QMC as in the perfect-pair generalized-valence-bond or the valence-bond approach built on atomic-centered orbitals …”

Estudo de propriedade energéticas de dímeros via Monte Carlo Quântico

“…4 Por isso, há uma grande quantidade de trabalhos que visam o estudo de propriedades energéticas de átomos, moléculas e sistemas mais complexos. [5][6][7][8][9][10][11][12][13][14][15][16] A maior parte desses estudos, trabalham com sistemas no estado fundamental, mas há também muitos trabalhos com estados excitados. [17][18][19] Com o avanço computacional e o desenvolvimento de novas metodologias e técnicas relacionadas ao MCQ, ampliou-se a sua aplicação para situações mais complexas, como por exemplo, para casos em que se considera o efeito do solvente, para área de materiais em geral e supercondutores.…”

“…) ρ(q) (7) em que M é o número de coordenadas q obtidas ao longo da integração de Monte Carlo, sobre a densidade de probabilidade. O sub-índice ρ indica que os pontos utilizados para o cálculo da média foram selecionados de acordo com a densidade de probabilidade (equação 6).…”