Thanks for the help everyone! Unfortunately adjusting the Mixing weight and tolerance did not help. Switching to ACML did not help the issue either, although it did speed up the calculation noticably (I never could get ATLAS to work for me, but this system is a bit of a mess so oh well).
I have tried switching my XC authors to RPBE, I was using RevPBEP; I noticed SIESTA prints a warning about in the atandard output when RevPBE is used. Don't worry, I have switched the Psuedo files as well. I'll let you know how it works. Its a large system with dense k-point sampling so it takes about half a day before I know whether the SCF calculations are converging or blowing up. Cheers! Rob ________________________________________ From: Juan Beltran [beltr...@averell.umh.ac.be] Sent: Thursday, July 01, 2010 9:45 AM To: siesta-l@uam.es Subject: Re: [SIESTA-L] SCF Convergence issues Hi Rob, some convergency problems I had for binary oxides were solved in most cases by increasing the MaxSCFIterations and DM.Tolerance while reducing the DM.MixingWeight parameter. I was using these settings which can be again set to normal values once the system is close to the ground state minimum: MaxSCFIterations 400 DM.MixingWeight 0.005 DM.Tolerance 2.d-3 I hope it helps, Juan On 07/01/2010 03:33 PM, Koch, Robert J wrote: > Hello all! > > Thanks for all the help so far! > > I'm now having an issue getting dDmax in my SCF calculations to move anywhere > below 1.0: > > siesta: iscf Eharris(eV) E_KS(eV) FreeEng(eV) dDmax Ef(eV) > siesta: 1 -55279.4019 -54866.8984 -54866.8984 1.8078 -4.1119 > timer: Routine,Calls,Time,% = IterSCF 1 78511.227 96.73 > elaps: Routine,Calls,Wall,% = IterSCF 1 3972.463 97.47 > siesta: 2 -56569.5706 -54935.4817 -54998.0432 1.4103 -0.6808 > siesta: 3 -56458.1370 -54943.8818 -55040.7807 1.2056 -1.5842 > siesta: 4 -56653.2502 -54947.8429 -55036.0493 1.8025 -1.2048 > siesta: 5 -56850.7918 -54948.5110 -55021.8909 1.6664 -1.9393 > siesta: 6 -56925.8454 -54947.7056 -55013.4623 1.9448 -2.1386 > siesta: 7 -56988.6476 -54945.1035 -55005.7856 1.9900 -2.1200 > siesta: 8 -56903.7554 -54942.1942 -54995.7603 2.0173 -2.4747 > siesta: 9 -56889.2759 -54938.0499 -54997.5861 2.0881 -2.2071 > siesta: 10 -56777.1928 -54934.0747 -54989.4649 2.1059 -2.4257 > siesta: 11 -56776.7970 -54928.9014 -54988.8325 2.1666 -2.2701 > > I have tried raising the electronic temperature (quite high now actually), > I've systematically adjusted the mixing parameters, my energy cutoff is at > 400 Ry, I'm using 18 k-points. > > The system I'm studying is a bulk Aurivillius phase (n=3) with Ti on the B > site and La on the A site. It is non magnetic (600 electrons included in the > calculation). Specifically Aurivillius phases are a perovskite structure > layered between (BiO2)2+ layers. > > Can anyone point me in a direction which would allow me to figure out what is > going on? > > Cheers, > Rob