Nuclear Ensemble Approach with Importance Sampling - Aix-Marseille Université Accéder directement au contenu
Article Dans Une Revue Journal of Chemical Theory and Computation Année : 2018

Nuclear Ensemble Approach with Importance Sampling

Fábris Kossoski
Mario Barbatti

Résumé

We show that the importance sampling technique can effectively augment the range of problems where the nuclear ensemble approach can be applied. A sampling probability distribution function initially determines the collection of initial conditions for which calculations are performed, as usual. Then, results for a distinct target distribution are computed by introducing compensating importance sampling weights for each sampled point. This mapping between the two probability distributions can be performed whenever they are both explicitly constructed. Perhaps most notably, this procedure allows for the computation of temperature dependent observables. As a test case, we investigated the UV absorption spectra of phenol, which has been shown to have a marked temperature dependence. Application of the proposed technique to a range that covers 500 K provides results that converge to those obtained with conventional sampling. We further show that an overall improved rate of convergence is obtained when sampling is performed at intermediate temperatures. The comparison between calculated and the available measured cross sections is very satisfactory, as the main features of the spectra are correctly reproduced. As a second test case, one of Tully's classical models 1 was revisited, and we show that the computation of dynamical observables also profits from the importance sampling technique. In summary, the strategy developed here can be employed to assess the role of temperature for any property calculated within the nuclear ensemble method, with the same computational cost as doing so for a single temperature.

Mots clés

Fichier principal
Vignette du fichier
p125_kossoski_importance-sampling_jctc_2018-preprint.pdf (660.63 Ko) Télécharger le fichier
Origine : Publication financée par une institution
Loading...

Dates et versions

hal-01965459 , version 1 (26-12-2018)

Licence

Paternité

Identifiants

  • HAL Id : hal-01965459 , version 1

Citer

Fábris Kossoski, Mario Barbatti. Nuclear Ensemble Approach with Importance Sampling. Journal of Chemical Theory and Computation, 2018, 14 (6), pp.3173-3183. ⟨hal-01965459⟩
30 Consultations
266 Téléchargements

Partager

Gmail Facebook X LinkedIn More