Stabilized quasiNewton optimization of noisy potential energy surfaces
Abstract
Optimizations of atomic positions belong to the most commonly performed tasks in electronic structure calculations. Many simulations like global minimum searches or characterizations of chemical reactions require performing hundreds or thousands of minimizations or saddle computations. To automatize these tasks, optimization algorithms must not only be efficient but also very reliable. Unfortunately, computational noise in forces and energies is inherent to electronic structure codes. This computational noise poses a severe problem to the stability of efficient optimization methods like the limitedmemory Broyden–Fletcher–Goldfarb–Shanno algorithm. We here present a technique that allows obtaining significant curvature information of noisy potential energy surfaces. We use this technique to construct both, a stabilized quasiNewton minimization method and a stabilized quasiNewton saddle finding approach. We demonstrate with the help of benchmarks that both the minimizer and the saddle finding approach are superior to comparable existing methods.
 Authors:
 Department of Physics, University of Basel, Klingelbergstrasse 82, CH4056 Basel (Switzerland)
 Institute for Advanced Studies in Basic Sciences, P.O. Box 451951159, IRZanjan (Iran, Islamic Republic of)
 Computational and Systems Biology, Biozentrum, University of Basel, CH4056 Basel (Switzerland)
 Publication Date:
 OSTI Identifier:
 22416002
 Resource Type:
 Journal Article
 Resource Relation:
 Journal Name: Journal of Chemical Physics; Journal Volume: 142; Journal Issue: 3; Other Information: (c) 2015 AIP Publishing LLC; Country of input: International Atomic Energy Agency (IAEA)
 Country of Publication:
 United States
 Language:
 English
 Subject:
 37 INORGANIC, ORGANIC, PHYSICAL AND ANALYTICAL CHEMISTRY; ALGORITHMS; BENCHMARKS; CALCULATION METHODS; CHEMICAL REACTIONS; COMPARATIVE EVALUATIONS; COMPUTERIZED SIMULATION; ELECTRONIC STRUCTURE; MINIMIZATION; POTENTIAL ENERGY; SURFACES
Citation Formats
Schaefer, Bastian, Goedecker, Stefan, Email: stefan.goedecker@unibas.ch, Alireza Ghasemi, S., and Roy, Shantanu. Stabilized quasiNewton optimization of noisy potential energy surfaces. United States: N. p., 2015.
Web. doi:10.1063/1.4905665.
Schaefer, Bastian, Goedecker, Stefan, Email: stefan.goedecker@unibas.ch, Alireza Ghasemi, S., & Roy, Shantanu. Stabilized quasiNewton optimization of noisy potential energy surfaces. United States. doi:10.1063/1.4905665.
Schaefer, Bastian, Goedecker, Stefan, Email: stefan.goedecker@unibas.ch, Alireza Ghasemi, S., and Roy, Shantanu. 2015.
"Stabilized quasiNewton optimization of noisy potential energy surfaces". United States.
doi:10.1063/1.4905665.
@article{osti_22416002,
title = {Stabilized quasiNewton optimization of noisy potential energy surfaces},
author = {Schaefer, Bastian and Goedecker, Stefan, Email: stefan.goedecker@unibas.ch and Alireza Ghasemi, S. and Roy, Shantanu},
abstractNote = {Optimizations of atomic positions belong to the most commonly performed tasks in electronic structure calculations. Many simulations like global minimum searches or characterizations of chemical reactions require performing hundreds or thousands of minimizations or saddle computations. To automatize these tasks, optimization algorithms must not only be efficient but also very reliable. Unfortunately, computational noise in forces and energies is inherent to electronic structure codes. This computational noise poses a severe problem to the stability of efficient optimization methods like the limitedmemory Broyden–Fletcher–Goldfarb–Shanno algorithm. We here present a technique that allows obtaining significant curvature information of noisy potential energy surfaces. We use this technique to construct both, a stabilized quasiNewton minimization method and a stabilized quasiNewton saddle finding approach. We demonstrate with the help of benchmarks that both the minimizer and the saddle finding approach are superior to comparable existing methods.},
doi = {10.1063/1.4905665},
journal = {Journal of Chemical Physics},
number = 3,
volume = 142,
place = {United States},
year = 2015,
month = 1
}

This paper shows how the application of augmented Lagrangian functions and quasiNewton techniques can be utilized for power system voltage optimization. The developed algorithm is attractive for three reasons: it can accommodate power system constraints in a straightforward manner, it is capable of reaching a solution even from infeasible startingpoints and it converges in a few iterations. The proposed algorithm offers substantial improvements in the computational efficiency. Computer runs have been performed and the results prove the efficiency of the algorithm.

On the characterization of qsuperlinear convergence of quasiNewton methods for constrained optimization
In this paper we present a short, straightforward and selfcontained derivation of the BoggsTolleWang characterization of those, quasiNewton methods for equalityconstrained optimization which produce iterates which are qsuperlinearly convergent. 
Quasiclassical scattering study with Fourier analysis of the H + H/sub 2/ and D + H/sub 2/ reactions at E/sub rel/ = 0. 65 eV. Comparison of results on two potential energy surfaces
A quasiclassical trajectory study of the H + H/sub 2/ and D + H/sub 2/ reactions has been carried out at E/sub rel/ = 0.65 eV. The calculations were carried out for both the PorterKarplus and SiegbahnLiuTruhlarHorowitz potential energy surfaces. Very accurate determinations have been made of the product vibrational, rotational, and translational energy distributions using Fourier analysis. The results for the different reactions and different surfaces are compared with each other and with other work on these systems. The results indicate that there is little coupling between vibrational and translational degrees of freedom except very near the transition state.more » 
Quasiclassical trajectory calculations of the thermal rate coefficient for the O + OH. >. O/sub 2/ + H reaction on realistic double manybody expansion potential energy surfaces for groundstate HO/sub 2/
Quasiclassical trajectory calculations of the thermal rate coefficient for the title reaction have been carried out over the temperature range 250 less than or equal to T less than or equal to 2500 K by using two recently reported DMBE potential energy surfaces for the ground state of the hydroperoxyl radical. The results are compared with each other and with experiment. The agreement is good. The authors results support previous theoretical calculations by Miller on the MeliusBlint potential energy surface in that nonstatistical recrossing effects are very important. For the DMBE II (DMBE III) potential energy surface, these nonstatistical correctionsmore » 
An improved quasidiabatic representation of the 1, 2, 3{sup 1}A coupled adiabatic potential energy surfaces of phenol in the full 33 internal coordinates
In a recent work we constructed a quasidiabatic representation, H{sup d}, of the 1, 2, 3{sup 1}A adiabatic states of phenol from high level multireference single and double excitation configuration interaction electronic structure data, energies, energy gradients, and derivative couplings. That H{sup d} accurately describes surface minima, saddle points, and also regions of strong nonadiabatic interactions, reproducing the locus of conical intersection seams and the coordinate dependence of the derivative couplings. The present work determines the accuracy of H{sup d} for describing phenol photodissociation. Additionally, we demonstrate that a modest energetic shift of two diabats yields a quantifiably more accuratemore »