(Jul 12, 2010) Errata Fixed

Post date: Sep 5, 2014 3:11:03 AM

The problem has been fixed, or rather explained, as of October 7, 2010. The detailed information is given below.

Issue:

======

Siam Quantum version 1.2.1 (and prior) calculated incorrectly the total energy of nitrogen molecule. At bond length 1.5 Angstroms, Siam Quantum predicted the total energy to be -107.769 Hartrees (HF/3-21G) whereas Gaussian software package gave -108.070 Hartrees, which was the correct value.

However, at bond length 1.3 Angstroms, both programs were in agreement.

Explaination:

=============

In the case of 1.5 Angstroms bond length, unfortunately there is a low-lying local minima when solving Hartree-Fock equation. The final answer seems to be sensitive to the initial guess of the density matrix. If Huckel guess is used, SCF cycle converges to correct value. However, if the core Hamiltonian is used, the SCF cycle arrives at the wrong minima. This is true for both Gaussian program and Siam Quantum.

For example, try the following calculations on Gaussian program and you will get two different answers.

case 1) #P HF/3-21G Guess=Core

case 2) #P HF/3-21G Guess=Huckel

In Siam Quantum, this problem is solved as of version 1.2.2 by changing the way in which we construct initial guess of the density matrix. In this case, the initial density matrix is set to simply identity matrix. This works surprisingly well, and finally solve the problem.

Teepanis Chachiyo. October 7, 2010.

=============== Original issue regarding the energy of nitrogen molecule ============

=============== Affected version 1.2.1 and prior ====================================

Kaniknun Srijiwangsa has found that when using Siam Quantum to compute the total energy of nitrogen molecule (N2) at large bond length (more than 1.3 angstroms), the energy is much higher than the results from Gaussian and Gamess.

In short, Siam Quantum gives the wrong! result for N2 molecule.

We are very serious about trying to get the right answer, and so are working to solve the problem. Preliminary test indicated that the error might due to poor convergence algorithm rather than mathematical or programming error.

This error affects all version of Siam Quantum, including the new release 1.2.1. And, we would like to thank Kaniknun Srijiwangsa for pointing out the bug.

Best wishes,

Teepanis Chachiyo.