Dear Xiao,
Emilio's answer is reasonable as always. MD in a coarse grained grid shows often discontinuities in total energy. What I don't understand on the contrary is why the symmetry isn't recovered explicitly after each MD step or, better after each self-consistent step. This would avoid any symmetry drifts along time, and the raise of eventual unphysical complex eigenvalues. This also raises another question related to the efficiency of H matrix element calculation. Since H should be Hermitian by definition why aren't just the upper triangular elements calculated and then copied by symmetry into the lower triangular part. Naturally this would reduce the construction time by two. If however all the elements are calculated I think a simple symmetry recovery transformation would be simple and wise to implement. Best Regards, Cosmin :*~*:._.:*~*:._.:*~*:._.:*~*:._.:*~*:._.:*~*: : Cosmin ROMAN, Ph.D. Student. : TIMA-INPG, MNS Group, : 46 Avenue F¨lix Viallet, 38031 Grenoble Cedex, FRANCE. : Tel : 00 33 4 76 57 48 34 : Fax : 00 33 4 76 47 38 14 e-mail : [EMAIL PROTECTED] : :*~*:._.:*~*:._.:*~*:._.:*~*:._.:*~*:._.:*~*: ----- Original Message ----- From: "Emilio Artacho" <[EMAIL PROTECTED]> To: <SIESTA-L@LISTSERV.UAM.ES> Sent: Saturday, June 19, 2004 11:35 AM Subject: Re: [SIESTA-L] about the Hamiltonian matrix and orbital overlap matrix > Dear Xiao > > One possible source of symmetry breakings of > different kinds is the real-space mesh (or grid) used > for calculating many of the integrals. > Clearly, the higher the mesh cutoff the smaller the > effect. But,I don't think you need to recover your symmetry > completely by a very high mesh cutoff in order to do > well converged calculations: it will depend on what > you want to calculate. > > Emilio > > Xiao Changyong wrote: > > >Dear Roman, > > > >Thank you for your advices. > > > >I have calculated Fe crystal, the symmetry of S and H are all acceptable > >from the constructioin of H and S till the exit of calculation. > > > >But for molecule, such as H2O, it is much different. After the construction > >of H and S, H and S exhibit symmetry. However, in > >MD process, the unsymmetry problem appeared with H, usually happened at the > >9th digit after decimal point. > >I am not sure if this is acceptable. > > > > > >Best regards, > >Xiao, David. > > > > > >-----Original Message----- > >From: Siesta, Self-Consistent DFT LCAO program, http://www.uam.es/siesta > >[mailto:[EMAIL PROTECTED] Behalf Of Cosmin ROMAN > >Sent:
