skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: A finite difference Davidson procedure to sidestep full ab initio hessian calculation: Application to characterization of stationary points and transition state searches

Abstract

The cost of calculating nuclear hessians, either analytically or by finite difference methods, during the course of quantum chemical analyses can be prohibitive for systems containing hundreds of atoms. In many applications, though, only a few eigenvalues and eigenvectors, and not the full hessian, are required. For instance, the lowest one or two eigenvalues of the full hessian are sufficient to characterize a stationary point as a minimum or a transition state (TS), respectively. We describe here a method that can eliminate the need for hessian calculations for both the characterization of stationary points as well as searches for saddle points. A finite differences implementation of the Davidson method that uses only first derivatives of the energy to calculate the lowest eigenvalues and eigenvectors of the hessian is discussed. This method can be implemented in conjunction with geometry optimization methods such as partitioned-rational function optimization (P-RFO) to characterize stationary points on the potential energy surface. With equal ease, it can be combined with interpolation methods that determine TS guess structures, such as the freezing string method, to generate approximate hessian matrices in lieu of full hessians as input to P-RFO for TS optimization. This approach is shown to achieve significantmore » cost savings relative to exact hessian calculation when applied to both stationary point characterization as well as TS optimization. The basic reason is that the present approach scales one power of system size lower since the rate of convergence is approximately independent of the size of the system. Therefore, the finite-difference Davidson method is a viable alternative to full hessian calculation for stationary point characterization and TS search particularly when analytical hessians are not available or require substantial computational effort.« less

Authors:
;  [1];  [2]
  1. Department of Chemical and Biomolecular Engineering, University of California, Berkeley, California 94720 (United States)
  2. Department of Chemistry, University of California, Berkeley, California 94720 (United States)
Publication Date:
OSTI Identifier:
22252982
Resource Type:
Journal Article
Resource Relation:
Journal Name: Journal of Chemical Physics; Journal Volume: 140; Journal Issue: 16; Other Information: (c) 2014 AIP Publishing LLC; Country of input: International Atomic Energy Agency (IAEA)
Country of Publication:
United States
Language:
English
Subject:
71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSICS; CHEMICAL ANALYSIS; EIGENVALUES; EIGENVECTORS; FINITE DIFFERENCE METHOD; OPTIMIZATION; POTENTIAL ENERGY

Citation Formats

Sharada, Shaama Mallikarjun, Bell, Alexis T., E-mail: mhg@bastille.cchem.berkeley.edu, E-mail: bell@cchem.berkeley.edu, and Head-Gordon, Martin, E-mail: mhg@bastille.cchem.berkeley.edu, E-mail: bell@cchem.berkeley.edu. A finite difference Davidson procedure to sidestep full ab initio hessian calculation: Application to characterization of stationary points and transition state searches. United States: N. p., 2014. Web. doi:10.1063/1.4871660.
Sharada, Shaama Mallikarjun, Bell, Alexis T., E-mail: mhg@bastille.cchem.berkeley.edu, E-mail: bell@cchem.berkeley.edu, & Head-Gordon, Martin, E-mail: mhg@bastille.cchem.berkeley.edu, E-mail: bell@cchem.berkeley.edu. A finite difference Davidson procedure to sidestep full ab initio hessian calculation: Application to characterization of stationary points and transition state searches. United States. doi:10.1063/1.4871660.
Sharada, Shaama Mallikarjun, Bell, Alexis T., E-mail: mhg@bastille.cchem.berkeley.edu, E-mail: bell@cchem.berkeley.edu, and Head-Gordon, Martin, E-mail: mhg@bastille.cchem.berkeley.edu, E-mail: bell@cchem.berkeley.edu. Mon . "A finite difference Davidson procedure to sidestep full ab initio hessian calculation: Application to characterization of stationary points and transition state searches". United States. doi:10.1063/1.4871660.
@article{osti_22252982,
title = {A finite difference Davidson procedure to sidestep full ab initio hessian calculation: Application to characterization of stationary points and transition state searches},
author = {Sharada, Shaama Mallikarjun and Bell, Alexis T., E-mail: mhg@bastille.cchem.berkeley.edu, E-mail: bell@cchem.berkeley.edu and Head-Gordon, Martin, E-mail: mhg@bastille.cchem.berkeley.edu, E-mail: bell@cchem.berkeley.edu},
abstractNote = {The cost of calculating nuclear hessians, either analytically or by finite difference methods, during the course of quantum chemical analyses can be prohibitive for systems containing hundreds of atoms. In many applications, though, only a few eigenvalues and eigenvectors, and not the full hessian, are required. For instance, the lowest one or two eigenvalues of the full hessian are sufficient to characterize a stationary point as a minimum or a transition state (TS), respectively. We describe here a method that can eliminate the need for hessian calculations for both the characterization of stationary points as well as searches for saddle points. A finite differences implementation of the Davidson method that uses only first derivatives of the energy to calculate the lowest eigenvalues and eigenvectors of the hessian is discussed. This method can be implemented in conjunction with geometry optimization methods such as partitioned-rational function optimization (P-RFO) to characterize stationary points on the potential energy surface. With equal ease, it can be combined with interpolation methods that determine TS guess structures, such as the freezing string method, to generate approximate hessian matrices in lieu of full hessians as input to P-RFO for TS optimization. This approach is shown to achieve significant cost savings relative to exact hessian calculation when applied to both stationary point characterization as well as TS optimization. The basic reason is that the present approach scales one power of system size lower since the rate of convergence is approximately independent of the size of the system. Therefore, the finite-difference Davidson method is a viable alternative to full hessian calculation for stationary point characterization and TS search particularly when analytical hessians are not available or require substantial computational effort.},
doi = {10.1063/1.4871660},
journal = {Journal of Chemical Physics},
number = 16,
volume = 140,
place = {United States},
year = {Mon Apr 28 00:00:00 EDT 2014},
month = {Mon Apr 28 00:00:00 EDT 2014}
}
  • SCF and single-reference ACPF calculations were performed in order to determine the structure, stability, and harmonic vibrational frequencies of stationary points for the HCO radical in the ground ({sup 2}A{prime}) and first excited ({sup 2}A{double_prime}) states. Very large and flexible basis sets including two f functions on the heavy atoms and two d functions on hydrogen were used. The calculated geometries and vibrational frequencies are in good agreement with available experimental data. The relative stabilities are now also much better balanced compared to previous theoretical results. 41 refs., 2 figs., 8 tabs.
  • GVB-POL-CI ab initio calculations of the geometries, energetics, and normal mode frequencies of C/sub 2/H/sub 2/, C/sub 2/H/sub 3/, and the transition state for the addition reaction of H + C/sub 2/H/sub 2/ are presented. In addition, normal mode frequencies for the isotopic variants D + C/sub 2/D/sub 2/, D + C/sub 2/H/sub 2/, and H + C/sub 2/D/sub 2/ are preented. These results are compared to experimental values for C/sub 2/H/sub 2/ and to ab initio values of Hagase and Kern, and semiempirical values of Keil, Lynch, Cowfer, and Michael. The results are also used to calculate the apparentmore » bimolecular addition rate constant using conventional RRKM theory for chemical activation. The calculated rate constants and their isotopic variants are compared as a function of temperature and pressure to available experimental information. The agreement is little different from that obtained by Keil et al. with a similar calculation using semiempirical values for acetylene, transition-state, and vinyl radical properties. In particular, the calculated high-pressure limit of the rate constant appears to be at least 1 order of magnitude higher than the experimental limit. Several possible reasons for this discrepancy are discussed.« less
  • We present ab initio (GVB--POL--CI) calculations for enough of the region about the abstraction and exchange saddle points for H(T)+CH/sub 4/(CD/sub 4/) to perform a full normal mode analysis of the transition states. The resulting normal mode frequencies are compared to four other published surfaces: an ab initio UHF--SCF calculation by Carsky and Zahradnik, a semiempirical surface by Raff, and two semiempirical surfaces by Kurylo, Hollinden, and Timmons. Significant quantitative and qualitative differences exist between the POL--CI results and those of the other surfaces. Transition state theory rate constants and vibrationally adiabatic reaction threshold energies were computed for all surfacesmore » and compared to available experimental values. For abstraction, the POL--CI rates are in good agreement with experimental rates and in better agreement than are the rates of any of the other surfaces. For exchange, uncertainties in the experimental values and in the importance of vibrationally nonadiabatic effects cloud the comparison of theory to experiment. Tentative conclusions are that the POL--CI barrier is too low by several kcal. Unless vibrationaly nonadiabatic effects are severe, the POL--CI surface is still in better agreement with experiment than are the other surfaces. The rates for a simple 3-atom transition state theory model (where CH/sub 3/ is treated as an atom) are compared to the rates for the full 6-atom model. The kinetic energy coupling of reaction coordinate modes to methyl group modes is identified as being of primary importance in determining the accuracy of the 3-atom model for this system. Substantial coupling in abstraction, but not exchange, causes the model to fail for abstraction but succeed for exchange.« less
  • The finite-difference boundary value method is applied to the calculation of Born--Oppenheimer vibrational energies and expectation values of R$sup -2$ for an excited state of H$sub 2$. We estimate the accuracy attainable by this method, point out a systematic error in the previous calculations of Tobin and Hinze, and correct several unjustified statements in the literature. Finally we point out that there is a large uncertainty in the final results due to choice of interpolation scheme. (AIP)
  • Convergence properties of the augmented Hessian (AH) method when searching for stationary points of an arbitrary fixed index are investigated. It is shown that the displacement vector of this method is proportional to one of the Hessian eigenvectors if the current point is far from a stationary one of the required index. A simple and reliable criterion for nearness of the current point to a stationary one of the desired index is proposed. The efficiency of a new one-dimensional optimization scheme that uses this criterion is studied. The case of coincidence of Hessian eigenvalues, which is a bottleneck of themore » standard AH method, is analyzed. A relation of the AH method to those by Poppinger and Wales is outlined. The correctness of the results obtained is illustrated on an example of a model surface. 24 refs., 1 fig., 1 tab.« less