Ceilidh ... Re^2: linear dependence and gradient out of range errors

Firefly and PC GAMESS-related discussion club

Re^2: linear dependence and gradient out of range errors

Dear Fabian

~ 10 atoms : No, this is not a reason for getting "GRADIENT OUT OF RANGE". It seems that the HW pseudo-potential was not built for Pd internally in FF but I have examined the Job using SBKJC and the structure was optimized without any problem. So I think that there is another problem here.

~ 200 atoms: Why must I replace the 6-31+G* basis set which is widely used in the literature by another basis set ? One may say some reasons about the efficiency of such basis sets but my main problem is the error.

Thanks for your attention.

Sincerely Yours
David G.
JCU

On Tue Jan 11 '11 5:29pm, Fabian Weber wrote
--------------------------------------------
>Dear David,

>I have some suggestions about your errors:

>>OPTIMIZATION ABORTED.
>> -- GRADIENT OUT OF RANGE
>> -- MAXIMUM ALLOWED FORCE (FMAXT) = 10.000

>This normally means, that your geometry is too far from an optimized geometry.

>If this occures right at the beginning of your calculation it will probably help if you "preoptimize" your input-geometry with a forcefield-method. Forcefield-methods like MM2 are available in ChemDraw 3D or Molden.

>If this Error occures after Firefly started "optimizing" your Geometry this may mean, that Firefly changed your geometry too much. You can tell firefly to take smaller steps by adding

> $statpt DXMAX=0.1 TRUPD=.f. $end

>to your input-file. DXMAX controls the stepsize (default is 0.3). The TRUPD-flag (default=.t.) lets Firefly increase the stepsize. (For further information see the manual)

>>*** FATAL ERROR ***
>> THE INPUT BASIS SET CONTAINS A LINEAR DEPENDENCE.

>This error already gives some clues why it occures. One of them is:

>> IF THE LINEAR DEPENDENCE IS A RESULT OF USING MANY DIFFUSE
>> BASIS FUNCTIONS, YOU ARE OUT OF LUCK, THE SCF NEVER WILL
>> CONVERGE WITH THIS DEGREE OF LINEAR DEPENDENCE

>This might be caused by your 6-31+G* basis set. Do you really need this basis set? You could try Ahlrichs VDZ as a more modern, similar complex basis set with less diffuse basis functions. aug-cc-pCVDZ is a more modern basis set with diffuse basis functions ("aug" is for "augmented" which means that diffuse functions have been added)

>>It seems that the errors are related to the external basis sets

>Why do you think that your errors are related to your external basis sets? Have you tried a different basis set?

>best regards,

>Fabian
>
>
>
>
>
>On Tue Jan 11 '11 0:15am, David G. wrote
>----------------------------------------
>>Dear Firefly Users

>>When I use external basis sets and ECPs for my system which includes Pd, C, and H; FF shows me two kind of errors that depends on the size of the system :

>>1. ~ 10 atoms :

>>OPTIMIZATION ABORTED.
>> -- GRADIENT OUT OF RANGE
>> -- MAXIMUM ALLOWED FORCE (FMAXT) = 10.000

>>2. ~ 200 atoms :

>>*** FATAL ERROR ***
>> THE INPUT BASIS SET CONTAINS A LINEAR DEPENDENCE.
>> INPUT BASIS DIMENSION= 3577 SALC DIMENSION= 3577
>> THE SMALLEST EIGENVALUE OF THE OVERLAP MATRIX IS 8.610577E-15
>> THERE ARE 68 EIGENVALUES LESS THAN 1.00E-10
>> THE NUMBER OF LINEARLY INDEPENDENT ORBITALS KEPT IS 3509

>> IF THERE ARE ANY SMALL EIGENVALUES:
>> CHECK YOUR OUTPUT FILE FOR DUPLICATE ATOMIC COORDINATES, OR
>> FOR DUPLICATE BASIS FUNCTIONS, OR FOR ZERO CONTRACTIONS
>> COEFFICIENTS, IF FOUND, THEN RESUBMIT THIS JOB. OTHERWISE,
>> IF THE LINEAR DEPENDENCE IS A RESULT OF USING MANY DIFFUSE
>> BASIS FUNCTIONS, YOU ARE OUT OF LUCK, THE SCF NEVER WILL
>> CONVERGE WITH THIS DEGREE OF LINEAR DEPENDENCE

>> IF THE SALC SPACE DOES NOT MATCH THE INPUT BASIS DIMENSION:
>> CHECK TO BE SURE YOU SCRUPULOUSLY FOLLOWED THE CONVENTION
>> FOR AXIS ORIENTATION IN $DATA (ESP. PERPENDICULAR C2 AXES)
>> AND THEN RESUBMIT THIS JOB.

>>the basis sets are standard and has been downloaded from EMSL website.

>>Hydrogen : 6-31G EMSL Basis Set Exchange Library

>>Carbon: 6-31+G* EMSL Basis Set Exchange Library

>>Palladium: LANL2DZ ECP EMSL Basis Set Exchange Library

>>It seems that the errors are related to the external basis sets and I cannot solve the problem. Also, there are not different internal basis sets which have ECPs in Firefly; and I have found only SBKJC and HW. I appreciate any suggestions in advance to find which type of basis sets and ECPs should be used for Firefly and how can we recognize them?
>>
>>
>>--
>>Sincerely Yours
>>David G.
>>JCU
>>
>>
>>
>>
>>
>>

Tue Jan 11 '11 10:15pm

This message read 956 times