A question about basis set

Top Page > Browsing

A question about basis set

Date: 2006/07/07 04:21
Name: Peter Winey <peterwiney@gmail.com>: Dear Dr.Ozaki,

Recently I was playing with openmx and encountered couple of problems.
One of them is related to the basis set, I think.

Here are the details:
When I run the C60_DC.dat calculation that you provided as an example,
everything appears to be ok. To check how the number of basis sets
affects the result, I modified the basis set line in the input file
from:

C C4.0-s1p1 C_LDA
to:
C C4.0-s2p2 C_LDA

and run the calculation again. Now the calculation does not seem to
converge. Below are the lines showing NormRD.

<DFT> NormRD = 1.000000000000 Criterion = 0.000000001000
<DFT> NormRD = 505.330342597744 Criterion = 0.000000001000
<DFT> NormRD = 506.619535017050 Criterion = 0.000000001000
<DFT> NormRD = 502.973423692965 Criterion = 0.000000001000
<DFT> NormRD = 381.117023865255 Criterion = 0.000000001000
<DFT> NormRD = 351.352164182782 Criterion = 0.000000001000
<DFT> NormRD = 265.587854074567 Criterion = 0.000000001000
<DFT> NormRD = 391.731830360534 Criterion = 0.000000001000
<DFT> NormRD = 137.325800547367 Criterion = 0.000000001000
<DFT> NormRD = 536.362748116627 Criterion = 0.000000001000
<DFT> NormRD = 223.268837201871 Criterion = 0.000000001000
<DFT> NormRD = 553.799971195841 Criterion = 0.000000001000
<DFT> NormRD = 124.356101869723 Criterion = 0.000000001000
<DFT> NormRD = 181.220527701639 Criterion = 0.000000001000
<DFT> NormRD = 115.141249838955 Criterion = 0.000000001000
<DFT> NormRD = 189.427878836131 Criterion = 0.000000001000
<DFT> NormRD = 141.454696615966 Criterion = 0.000000001000
<DFT> NormRD = 79.861710470534 Criterion = 0.000000001000
<DFT> NormRD = 112.029666714555 Criterion = 0.000000001000
<DFT> NormRD = 148.208874534117 Criterion = 0.000000001000
...

Could you help me to identify what was wrong?
Thanks in advance for your help!

-Peter

Page: [1]

Re: A question about basis set ( No.1 )

Date: 2006/07/10 22:13
Name: T.Ozaki

Hi,

I performed the same calculation, and obtained the convergent
result as follows:

SCF= 1 NormRD= 1.000000000000 Uele= -104.224142291171
SCF= 2 NormRD= 0.282692921550 Uele= -92.004281331209
SCF= 3 NormRD= 0.130594203923 Uele= -94.558703455327
SCF= 4 NormRD= 0.055468788877 Uele= -95.905971772074
SCF= 5 NormRD= 0.026672871887 Uele= -96.358015140301
SCF= 6 NormRD= 0.013443866376 Uele= -96.527102830466
SCF= 7 NormRD= 0.005525912309 Uele= -96.625285482559
SCF= 8 NormRD= 0.001127771215 Uele= -96.656488848760
SCF= 9 NormRD= 0.000531279310 Uele= -96.649760172546
SCF= 10 NormRD= 0.000252692788 Uele= -96.649063151801
SCF= 11 NormRD= 0.000064988974 Uele= -96.648804910039
SCF= 12 NormRD= 0.000039272247 Uele= -96.648816162146
SCF= 13 NormRD= 0.000044521895 Uele= -96.648867538333
SCF= 14 NormRD= 0.000004069321 Uele= -96.648782584094
SCF= 15 NormRD= 0.000002722798 Uele= -96.648781972128
SCF= 16 NormRD= 0.000006680434 Uele= -96.648780149096
SCF= 17 NormRD= 0.000001891980 Uele= -96.648781785242
SCF= 18 NormRD= 0.000001222400 Uele= -96.648782462235
SCF= 19 NormRD= 0.000001077218 Uele= -96.648783398319
SCF= 20 NormRD= 0.000000746105 Uele= -96.648783142529
SCF= 21 NormRD= 0.000000574732 Uele= -96.648783003786
SCF= 22 NormRD= 0.000000621744 Uele= -96.648783231491
SCF= 23 NormRD= 0.000000463928 Uele= -96.648783055105
SCF= 24 NormRD= 0.000005100631 Uele= -96.648785232430
SCF= 25 NormRD= 0.000000376489 Uele= -96.648782799019
SCF= 26 NormRD= 0.000000417988 Uele= -96.648782935596
SCF= 27 NormRD= 0.000000386418 Uele= -96.648782908274
SCF= 28 NormRD= 0.000000377505 Uele= -96.648782921428
SCF= 29 NormRD= 0.000000358025 Uele= -96.648782943094
SCF= 30 NormRD= 0.000000351320 Uele= -96.648782939424
SCF= 31 NormRD= 0.000000000000 Uele= -96.648782939351

Note that I changed only from C4.0-s1p1 to C4.0-s2p2 in this
input file, and did not change any other parameters.

I wonder that the lapack routine failed to converge on your machine.
To confirm this, please try a different routine by
scf.lapack.dste dstedc # dstegr|dstedc|dstevx, default=dstevx

Regards,

TO

Re: A question about basis set ( No.2 )

Date: 2006/07/10 23:49
Name: Peter Winey <peterwiney@gmail.com>

Dr.Ozaki,

It is very helpful to know that the problem I had is specific to my machine.

I tried all three lapack routines and did not solve the problem. Then I thought
that my compilation of the openmx code might have some problem.

I used intel icc compiler Version 9.0 and I knew the "-O3" flag does not work well.
So in my current compilation I used the "-O2" flag, as suggested by you in an old
post on this forum.

Anyway, I tried re-compiling the code just now using the "-O1" flag and found the
problem got solved! I am happy again. :-)

Thanks,
-Peter

Page: [1]