# NWPW Optimization Fail to Converge for a Tutorial Job

Viewed 3008 times, With a total of 3 Posts

 Chiensh Member Profile Send PM
 Clicked A Few Times Threads 18 Posts 34
 12:21:31 AM PDT - Fri, May 6th 2011 Dear All, I have compiled a copy of NWChem 6.0 with on our SGI Altex server, and I tried to run a NWPW tutorial job obtained from the following page http://www.nwchem-sw.org/index.php/Plane-Wave_Density_Functional_Theory I chose "NWPW Tutorial 5: Optimizing the Unit Cell and Geometry of Diamond" to check the NWPW module in my copy, and this job was supposed to converge in the 6th optimization cycle as it is shown. The only different of my input from the the one shown in the tutorial is definition of "permanent_dir" and "scratch_dir" are removed. However, my calculation failed to converge within 40 optimization iterations. Part of the output are given below, can anyone help me to figure out what is the problem ? Thanks a lot. Regards, Dominic Chien I used the these compilers, MKL and MPI library to compile CC=gcc FC=ifort MPI=mpt-2.04 MKL 10.2.0.013 This is the first iteration of my calculation -------- Step 1 -------- Geometry "geometry" -> "geometry" --------------------------------- Output coordinates in angstroms (scale by 1.889725989 to convert to a.u.) No. Tag Charge X Y Z ---- ---------------- ---------- -------------- -------------- -------------- 1 C 6.0000 1.88357541 1.88401967 1.88308402 2 C 6.0000 -0.00008246 -0.00039948 1.88381307 3 C 6.0000 -0.00008905 1.88315026 0.00014969 4 C 6.0000 1.88325462 0.00001380 -0.00010021 5 C 6.0000 2.82514175 2.82520561 2.82555541 6 C 6.0000 0.94210694 0.94213554 2.82468173 7 C 6.0000 0.94154529 2.82542784 0.94150850 8 C 6.0000 2.82536182 0.94123718 0.94188317 Step Energy Delta E Gmax Grms Xrms Xmax Walltime ---- ---------------- -------- -------- -------- -------- -------- -------- 1 -45.06493468 -7.2D-03 0.04795 0.01449 0.11787 0.39107 284.7 This is the result of the 6th cycle in my calculation -------- Step 6 -------- Geometry "geometry" -> "geometry" --------------------------------- Output coordinates in angstroms (scale by 1.889725989 to convert to a.u.) No. Tag Charge X Y Z ---- ---------------- ---------- -------------- -------------- -------------- 1 C 6.0000 1.82723706 1.82729824 1.82705720 2 C 6.0000 0.00000990 -0.00005990 1.82729710 3 C 6.0000 -0.00000770 1.82706000 0.00002961 4 C 6.0000 1.82712035 0.00006237 -0.00002685 5 C 6.0000 2.74074475 2.74072927 2.74088332 6 C 6.0000 0.91366245 0.91369992 2.74065081 7 C 6.0000 0.91351325 2.74080620 0.91352781 8 C 6.0000 2.74078530 0.91348139 0.91365483 Step Energy Delta E Gmax Grms Xrms Xmax Walltime ---- ---------------- -------- -------- -------- -------- -------- -------- 6 -45.07688298 -5.9D-08 0.00072 0.00041 0.00001 0.00003 643.3 Following is the final result of my job -------- Step 40 -------- Geometry "geometry" -> "geometry" --------------------------------- Output coordinates in angstroms (scale by 1.889725989 to convert to a.u.) No. Tag Charge X Y Z ---- ---------------- ---------- -------------- -------------- -------------- 1 C 6.0000 1.82723137 1.82725492 1.82706722 2 C 6.0000 0.00002348 -0.00004480 1.82731800 3 C 6.0000 0.00000670 1.82703722 0.00004368 4 C 6.0000 1.82710444 0.00006880 0.00002764 5 C 6.0000 2.74071464 2.74073912 2.74084961 6 C 6.0000 0.91365030 0.91366721 2.74063091 7 C 6.0000 0.91350995 2.74079951 0.91347033 8 C 6.0000 2.74076657 0.91345935 0.91358173 Step Energy Delta E Gmax Grms Xrms Xmax Walltime ---- ---------------- -------- -------- -------- -------- -------- -------- 40 -45.07688299 -9.7D-10 0.00186 0.00071 0.00001 0.00001 3243.6 ok ok --------------------------------------------------------------- Failed to converge in maximum number of steps or available time --------------------------------------------------------------- Following is the converged result of the 6th cycle for the tutorial ---------------------- Optimization converged ---------------------- Step Energy Delta E Gmax Grms Xrms Xmax Walltime ---- ---------------- -------- -------- -------- -------- -------- -------- 6 -45.07688304 -1.1D-07 0.00037 0.00021 0.00002 0.00003 174.5 ok ok ok ok Geometry "geometry" -> "geometry" --------------------------------- Output coordinates in angstroms (scale by 1.889725989 to convert to a.u.) No. Tag Charge X Y Z ---------------- ---------- -------------- -------------- -------------- 1 C 6.0000 1.82723789 1.82729813 1.82705440 2 C 6.0000 0.00000857 -0.00006053 1.82730027 3 C 6.0000 -0.00000584 1.82706061 0.00002852 4 C 6.0000 1.82712018 0.00006354 -0.00002544 5 C 6.0000 2.74074195 2.74072805 2.74088522 6 C 6.0000 0.91366407 0.91370055 2.74064976 7 C 6.0000 0.91351181 2.74080771 0.91352917 8 C 6.0000 2.74078843 0.91348115 0.91365446 Edited On 12:23:34 AM PDT - Fri, May 6th 2011 by Chiensh

 Chiensh Member Profile Send PM
 Clicked A Few Times Threads 18 Posts 34
 7:05:11 PM PDT - Sun, May 8th 2011 I just confirm that this unconverged result only happens when more than 1 cpu is required. I got the converged result in the 6th optimization cycle as expect when only 1 cpu was required. Does any one able to reproduce this error in NWChem ver. 6.0?

 HUWagner Member Profile Send PM
 Just Got Here Threads 1 Posts 2
 4:04:08 AM PDT - Fri, Aug 19th 2011 NWPW Optimization Fail to Converge for a Tutorial Job - Question Hi Dominic Chien, could you please insert here the input file to this calculation example. I couldn't find it in the tutorial. Regards, HUWagner

 Guest - Send PM
 3:56:10 PM PDT - Fri, Aug 19th 2011 Dominic Chien, Basically your energy is converged. The problem your seeing is that the number of plane-waves in the calculation changes wrt to the unit cell. There's a fix for this problem in the development branch. Another way to fix this problem is the use a larger cutoff energy. Eric

 Who's here now Members 0 Guests 1 Bots/Crawler 0

AWC's: 2.5.10 MediaWiki - Stand Alone Forum Extension
Forum theme style by: AWC