Bug in running of IP/EA CCSD?
Clicked A Few Times
12:02:42 PM PDT - Mon, Sep 22nd 2014
Hello all,

I have been having difficulties with some errors with the newest snapshot of nwchem including the IPCCSD and EACCSD modules. I have tested my code and it works great for the ozone and F2 example files however. When I attempt to get the IP and EA of a 1-D chain of LiH atoms, I stumble upon strange results.

Here are my input files for the IPCCSD and EACCSD:

IP:

charge 0
memory stack 3500 mb heap 100 mb global 8000 mb
geometry nocenter noautosym noautoz
Li 0 0 0
H 0 0 1.5
Li 0 0 3
H 0 0 4.5
Li 0 0 6.0
H 0 0 7.5
end
basis
  • library 6-31g*
end
scf
rhf; singlet;
end
tce
ipccsd

nroots 2

freeze atomic

tilesize 20

thresh 1d-7
end
set tce:thresheom 1d-4
set tce:thresh1 1d-3
task tce energy

Results for the IP run:

No. of initial right vectors 12

For some reason it won't let me post this unless I delete most of the iterations so here is the error bombout message: 

EOM-CCSD right-hand side iterations

--------------------------------------------------------------

     Residuum       Omega / hartree  Omega / eV    Cpu    Wall

--------------------------------------------------------------

Iteration  18 using   37 trial vectors

  0.0000000876516   0.0000000000000    0.00000

  0.0001016081551   0.2140089616994    5.82348     0.2    31.1



Iteration  19 using   38 trial vectors

------------------------------------------------------------------------

tce_eom_xdiagon: failed to extend subspace                   0

------------------------------------------------------------------------

------------------------------------------------------------------------

 current input line :

   26: task tce energy

------------------------------------------------------------------------

------------------------------------------------------------------------

This type of error is most commonly associated with calculations not reaching convergence criteria

As you can see the first root, goes to 0.00 eV. What exactly does this mean????

EA:

charge 0
memory stack 3500 mb heap 100 mb global 8000 mb
geometry nocenter noautosym noautoz
Li 0 0 0
H 0 0 1.5
Li 0 0 3
H 0 0 4.5
Li 0 0 6.0
H 0 0 7.5
end
basis
  • library 6-31g*
end
scf
rhf; singlet;
end
tce
eaccsd

nroots 2

freeze atomic

tilesize 30

thresh 1d-6
end
task tce energy

No. of initial right vectors 608

EOM-CCSD right-hand side iterations

--------------------------------------------------------------

     Residuum       Omega / hartree  Omega / eV    Cpu    Wall

--------------------------------------------------------------



Iteration   1 using  608 trial vectors

------------------------------------------------------------------------

tce_eom_xdiagon: nxtrails too big                   4

------------------------------------------------------------------------

------------------------------------------------------------------------

 current input line :

   24: task tce energy

------------------------------------------------------------------------

------------------------------------------------------------------------

------------------------------------------------------------------------

For more information see the NWChem manual at http://nwchemgit.github.io/index.php/NWChem_Documentation

I am not sure why in this case there are so many trial vectors, or how I can go about fixing this.

Thanks in advance,

-Cole

Clicked A Few Times
4:02:54 PM PDT - Mon, Sep 22nd 2014
Hi Cole,
please use maxeorb to define initial starts for EA/IP-EOMCCSD calculations (as described in the manual).
It significantly reduces the number of initial guesses and results in a good convergence of eigensolver.

New inputs are attached:
####### IP ############
echo
start lih_5_ip
title lih_5_ip

permanent_dir /dtemp/kowalski

memory stack 1600 mb heap 100 mb global 1600 mb

charge 0

geometry  noautosym
Li 0 0 0
H 0 0 1.5
Li 0 0 3
H 0 0 4.5
Li 0 0 6.0
H 0 0 7.5
end

basis
 * library 6-31g*
end

scf
rhf
singlet
end

tce
ipccsd
nroots 2
freeze atomic
tilesize 20
thresh 1d-6
end

set tce:maxeorb 0.37

set tce:thresheom 1d-4
task tce energy
#################################



#### EA ##########################
echo
start lih_5_ea
title lih_5_ea

permanent_dir /dtemp/kowalski

memory stack 1600 mb heap 100 mb global 1600 mb

charge 0

geometry  noautosym
Li 0 0 0
H 0 0 1.5
Li 0 0 3
H 0 0 4.5
Li 0 0 6.0
H 0 0 7.5
end

basis
 * library 6-31g*
end

scf
rhf
singlet
end

tce
eaccsd
nroots 2
freeze atomic
tilesize 20
thresh 1d-6
end

set tce:maxeorb 0.0132

set tce:thresheom 1d-4
task tce energy
################################



Calculated IPs
Iteration  13 using   23 trial vectors

  0.0000005179033   0.2140087813386    5.82348

  0.0000004428821   0.3411417911087    9.28294     0.0     0.0

--------------------------------------------------------------
(5.82 & 9.28 eV)


Calculated EAs
Iteration  12 using   24 trial vectors

  0.0000002746568  -0.0398998406135   -1.08573

  0.0000004933803   0.0062559815381    0.17023     0.0     0.1

--------------------------------------------------------------
(-1.09 and 0.17 eV)

Best,
Karol

Clicked A Few Times
10:14:47 AM PDT - Mon, Sep 29th 2014
Hi Karol,

Thank you so much for your quick answer. I have only one question. How exactly do you specify the value of max eorb (since they have different values in both cases). Is there just a general number I can use for all molecules? such as .10 or .01?

Thanks,
-Cole


Forum >> NWChem's corner >> Running NWChem