Ceilidh ... Re^4: ROUNDOFF ERROR IN BEND message when reading a Z matrix

Firefly and PC GAMESS-related discussion club

Re^4: ROUNDOFF ERROR IN BEND message when reading a Z matrix

Reinaldo Pis Diez
reinaldo.pisdiez@gmail.com

That's exactly what I did. The error message appears when I added ifzmat to the input after a previous run with dlc/auto only.

The strange thing here (to me at least) is that the manual states that ifzmat is used to freeze some internal coords but the dlc generates delocalized coords, which are usually much more that 3N-6. Then, how should I connect the two coord types?

Let me make things clearer: for a trial system X(OH)4, there are 21 internal coords. However, dlc/auto generates 30 coords. After that, the program prints

INDEP./SYM. INTERNAL COORDINATES

and the expected 21 internal coords are expressed as combination of the 30 delocalized ones. I could easily identify bonds and angles by looking at the table of automatically generated coords but the identification of the internal coords generated as above is not so easy at all. I understand that ifzmat acts on the internal coords then, how could I "see" what coords to freeze?

On Mon Aug 2 '10 9:04pm, Jim Kress wrote
----------------------------------------
>Try running your input without the IFZMAT but with EXETYP=CHECK. That will list all the internals. Then select the internals from that list and rebuild your IFZMAT from the auto generated internals you got from the EXETYP=CHECK run.

>On Mon Aug 2 '10 0:38am, Reinaldo Pis Diez wrote
>------------------------------------------------
>>Thanks to Alex and Jim for their help/hints. I must say that I finally rearranged the original z matrix and freeze the internal coords using ifreez within $statpt.

>>I'd like to mention that I couldn't use ifrzat together with dlc=.t. and auto=.t. because it is limited to 16 atoms in the current version and I have much more atoms to freeze.

>>Also, I was unable to combine dlc=.t. and auto=.t. with ifzmat as Jim suggested. According to the INPUT section of the manual, ifzmat must have the same format as izmat. Thus, I try something like

>> $zmat dlc=.t. auto=.t. ifzmat(1)=1,1,2, 1,1,3 $end

>>to see what happens and got the message

>>ERROR: UNABLE TO PROJECT DLC!

>>after the z matrix was automatically generated. I don't know what happens here, but it is clear that the problem is with ifzmat as if it is removed, the example runs smoothly. I guess that I'm missing something here. Can anyone provide a running example using dlc/auto/ifzmat?

>>Thanks in advance.

>>Regards,

>>Reinaldo

>>On Sat Jul 31 '10 2:02am, Jim Kress wrote
>>-----------------------------------------
>>>Why don't you just use

>>>$ZMAT DLC=.true. AUTO=.true. $END

>>>to automatically generate your internal coordinates. It's much more efficient and effective. Then, if you want to freeze any of these internal coordinates, you can use the IFZMAT/FVALUE keywords in $ZMAT.

>>>See page 32 (and following) in this documentation:

>>>for more info.

>>>Jim

>>>On Mon Jul 26 '10 0:24am, Reinaldo Pis Diez wrote
>>>-------------------------------------------------
>>>>Dear folks,

>>>>I'm using Firefly 7.1.G, build number 5618, to study a system containing about a hundred atoms. I'm using cartesian coordinates due to construction issues. Scf is ok, slow but convergent. Now, I have to optimize the neighborhood of a given atom, so I decide to use nzvar=3M-6 in $contrl, and "build" the z matrix in $zmat using zmat(1)=1,1,2, etc, etc. I've checked the z matrix and is ok with respect to connectivity issues.
>>>>Despite the ifreez option I'm using to freeze almost all the internal coordinates, Firefly returns an error message during the reading of the z matrix in check mode:

>>>> --- ENCODED Z MATRIX ---
>>>> COORD TYPE I J K L M N
>>>> 1 1 2 1
>>>> 2 1 3 1
>>>>...
>>>> 309 3 105 69 33 21
>>>> ROUNDOFF ERROR IN BEND - STOP

>>>>where the 309th coord is the last one. What is the source of that error? (Almost) linear bond or torsion angles perhaps?
>>>>I could provide the entire input if needed.
>>>>Thanks in advance.
>>>>Regards,

>>>>Reinaldo
>>>>

Tue Aug 3 '10 6:08pm

This message read 930 times