Invalid shell type specified in basis

General discussion of the Cambridge quantum Monte Carlo code CASINO; how to install and setup; how to use it; what it does; applications.
Post Reply
Posts: 16
Joined: Sun Jun 02, 2013 11:41 am

Invalid shell type specified in basis

Post by s.nasiri »

Dear All
I have computed the anionic form of the Lithium atom by MCSCF with 4 electron and 9 orbitals by the Gamess package. However, there is an error about the basis set. I could not find similar problem in the previous topics. I pasted the input file of the gamess and the last line of the CASINO output as the follows. I used aug-cc-pvtz basis set.
any help would be appreciated.

#################### input###################

$QMCWFN dettol=0.00001 $END
Li 3.0 0.00000 0.00000 0.00000


143 -1.0000000000E-05 2 Free
144 -1.0000000000E-05 2 Free
145 -1.0000000000E-05 2 Free
146 -1.0000000000E-05 2 Free
147 1.0000000000E-05 2 Free
148 1.0000000000E-05 2 Free
149 -1.0000000000E-05 2 Free
150 -1.0000000000E-05 2 Free
151 -1.0000000000E-05 2 Free
152 -1.0000000000E-05 2 Free

Excitation specifications (see manual):
No excitations specifed.

Reading Gaussian wave function and associated data

Title : TITLE

Generating code : GAMESS
Method : MCSCF
DFT functional : NONE
Periodicity : 0 (molecule)
Spin restricted? : Yes

Total number of electrons : 4

Basis set information

Number of atoms : 1
Number of shells : 14
Number of basis fns : 46
Number of primitives : 32
Highest ang momentum : 4

Invalid shell type specified in basis (s,sp,p,d,2s,3p only in GAMESS).

CASINO internal traceback:
Problem detected at SETUP
Called from MAIN


--Job's stderr--
Katharina Doblhoff
Posts: 84
Joined: Tue Jun 17, 2014 6:50 am

Re: Invalid shell type specified in basis

Post by Katharina Doblhoff »

The solution is actually in the error message. aug-ccp-vTZ for Li contains d and f shells. These are not supported by the converter.
There are 4 solutions to this:
a.) Kick out the d and f function of your basis set and redo the MCSCF calculation (but this may lower the accuracy of your MCSCF calculation, so you may want to check)
b.) Implement d and f orbitals in the converter (I would not advise this. It is nasty to get the transformation matrices right and other people have done this)
c.) Use gaussian instead of gamess and the converter I wrote. There should be some forum entries on this somewhere
d.) Contact Albert DeFusco and ask him whether he may want to give you his gamess code with internal converter. His code can by now treat d and f functions.

I guess a and d are the fastest solutions since you need to change least.

However, you may still additionally need to change a line of code in CASINO itself. At least the version of CASINO that I am using still forbids the use of d and f orbitals since it assumes that none of the converters can handle those orbitals. If using a correct converter, these lines can safely be removed (and they may have been removed in a newer version of CASINO - I did not check).
Posts: 16
Joined: Sun Jun 02, 2013 11:41 am

Re: Invalid shell type specified in basis

Post by s.nasiri »

Dear Katharina Doblhoff

Thanks so much for the comments. In fact this version of the Gamess has been given by Albert DeFusco.
Also, the Gaussian does not print more than 50 determinants.
Katharina Doblhoff
Posts: 84
Joined: Tue Jun 17, 2014 6:50 am

Re: Invalid shell type specified in basis

Post by Katharina Doblhoff »

Ah, sorry. You are in casino already.
As I said before, you may need to change the casino code...
The problem lies in the gaussian.f90 file. Do you have a version of CASINO where you can access the source code? If Albert is sure that his f terms are correct, then it should be safe to comment the lines in the code that rise this error and stop the code. If you do not have a source code, you may want to discuss with Mike whether he wants to change this. (Which may be a bit dangerous since the converter for the "normal" GAMESS code, cannot handle f functions).
Mike Towler
Posts: 237
Joined: Thu May 30, 2013 11:03 pm
Location: Florence

Re: Invalid shell type specified in basis

Post by Mike Towler »

Yes go ahead and remove the error trap in your personal copy of CASINO if you're happy that f/g work.. (you can do this even in the obfuscated version of the source code by just searching the *.f90 files in the CASINO/src directory for a suitable substring from the error message).

To make sure that the GAMESS interface does in fact work for higher angular momentum functions, the wave function files generated by GAMESS need to pass the tests in the CASINO/examples/generic/gauss_dfg/gamess directory. According to the data in the file CASINO/examples/generic/gauss_dfg/RESULTS both the Defusco interface and the ancient gamess2qmc converter supplied with CASINO are currently certified as correct only up to d functions. If someone can supply me evidence to the contrary I will update the RESULTS file and remove the error trap.. (or better, make it more sophisticated so it understands the difference between the two GAMESS interfaces).

Post Reply