[molpro-user] Internal and explicit initial guessing

[Taube, Andrew]

I am having a weird result when comparing results for basis sets from the internal basis set library and from explicitly including the same basis set in the input file. While the results converge to the same value of RHF energy, the initial guess is different.

For example, for F- in a cc-pVDZ basis including the basis set explicitly:
***,F- cc-pVDZ explicitly
memory,400,M
basis={
s, F , 14710.0000000, 2207.0000000, 502.8000000, 142.6000000, 46.4700000, 16.7000000, 6.3560000, 1.3160000, 0.3897000
c, 1.9, 0.0007210, 0.0055530, 0.0282670, 0.1064440, 0.2868140, 0.4486410, 0.2647610, 0.0153330, -0.002332
c, 1.9, -0.0001650, -0.0013080, -0.0064950, -0.0266910, -0.0736900, -0.1707760, -0.1123270, 0.5628140, 0.568778
c, 9.9, 1
p, F , 22.6700000, 4.9770000, 1.3470000, 0.3471000
c, 1.4, 0.0448780, 0.2357180, 0.5085210, 0.45812
c, 4.4, 1
d, F , 1.6400000
c, 1.1, 1
}
thrdscf=1e-12,thrdscf_max=1e-9
gthresh,throvl=1e-9,throrth=1e-7
gdirect
geometry={f;
}
charge=-1
hf

has the following set of iterations:
ITERATION   DDIFF            GRAD            ENERGY        2-EL.EN.            DIPOLE MOMENTS         DIIS   CPU(IT)  CPU(TOT)
    1      0.000D+00      0.000D+00       -99.29578519     95.175396   0.000000   0.000000   0.000000    0      5.7      9.2   diag
    2      0.000D+00      0.131D+00       -99.34548119     88.831609   0.000000   0.000000   0.000000    1      3.0     12.2   diag
    3      0.206D+00      0.728D-01       -99.36598299     91.014455   0.000000   0.000000   0.000000    2      3.1     15.2   diag
    4      0.688D-01      0.506D-03       -99.36598357     91.025429   0.000000   0.000000   0.000000    3      3.0     18.2   diag
    5      0.401D-03      0.113D-04       -99.36598357     91.025398   0.000000   0.000000   0.000000    3      3.0     21.2   diag
    6      0.184D-04      0.259D-05       -99.36598357     91.025417   0.000000   0.000000   0.000000    3      2.0     23.2   diag
    7      0.558D-05      0.141D-06       -99.36598357     91.025424   0.000000   0.000000   0.000000    0      2.0     25.1   orth

While calling the same basis set from the internal library,
***,F- cc-pVDZ implicitly
memory,400,M
basis={
f=cc-pVDZ
}
thrdscf=1e-12,thrdscf_max=1e-9
gthresh,throvl=1e-9,throrth=1e-7
gdirect
geometry={f;
}
charge=-1
hf

has the following iteration set:
ITERATION   DDIFF            GRAD            ENERGY        2-EL.EN.            DIPOLE MOMENTS         DIIS   CPU(IT)  CPU(TOT)
    1      0.000D+00      0.000D+00       -99.29785341     95.113039   0.000000   0.000000   0.000000    0      5.4      7.5   diag
    2      0.000D+00      0.129D+00       -99.34608743     88.864010   0.000000   0.000000   0.000000    1      2.9     10.3   diag
    3      0.202D+00      0.717D-01       -99.36598302     91.014701   0.000000   0.000000   0.000000    2      2.8     13.1   diag
    4      0.678D-01      0.494D-03       -99.36598357     91.025427   0.000000   0.000000   0.000000    3      3.1     16.2   diag
    5      0.392D-03      0.110D-04       -99.36598357     91.025398   0.000000   0.000000   0.000000    3      3.1     19.3   diag
    6      0.179D-04      0.252D-05       -99.36598357     91.025417   0.000000   0.000000   0.000000    3      3.1     22.4   diag
    7      0.542D-05      0.138D-06       -99.36598357     91.025424   0.000000   0.000000   0.000000    0      3.0     25.4   orth

Looking at the first iteration there is a difference of 2 mH. In this case, that isn't a big deal, however, I have added basis sets to the basis set library and for some larger basis sets, the resultant iterations converge to different states. I can send those examples as desired. Can someone explain why this behavior is occurring and how to control it?

These calculations were run with molpro version 2009.1, patch level 29.

Thanks,
Andrew

---
Andrew G. Taube
D. E. Shaw Research
email: Andrew.Taube at DEShawResearch.com
Tel: +1-212-478-0118
Fax: +1-212-845-1118



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.molpro.net/pipermail/molpro-user/attachments/20100629/44862750/attachment.html>