- Posts: 4
- Thank you received: 0
restart geometry optimization
- jlievin
- Topic Author
- Offline
- New Member
Less
More
1 year 4 months ago #1375
by jlievin
restart geometry optimization was created by jlievin
Hi,
I'm new to MRCC and have some basic questions about the geometry optimization.
1) I'm trying to restart a geometry optimization after a first run in which optmaxit has been set to 2 to avoid reaching the time limit of the cluster I'm calculating on.
I then run MRCC for 2 more iterations in the directory of the first calculation, using the MINP resulting from the first 2 iterations, to which I've added rest=1. Repeating this several times leads to the requested convergence. However, in my simple test case (H2O df-HF/VDZ) convergence is achieved after 13 iterations, whereas a one-step calculation only requires 5. I get the same result without rest=1, and I suppose that that the restart command has been ignored, maybe because after 2 iterations MRCC has stopped on an error (convergence fails)?
Is it a solution to overcome this problem?
2) The final geometry is provided in Cartesian coordinates and I wonder if it would be possible to obtain the Z matrix coordinates as well?
Many thanks in advance for any comments or help.
Jacky
I'm new to MRCC and have some basic questions about the geometry optimization.
1) I'm trying to restart a geometry optimization after a first run in which optmaxit has been set to 2 to avoid reaching the time limit of the cluster I'm calculating on.
I then run MRCC for 2 more iterations in the directory of the first calculation, using the MINP resulting from the first 2 iterations, to which I've added rest=1. Repeating this several times leads to the requested convergence. However, in my simple test case (H2O df-HF/VDZ) convergence is achieved after 13 iterations, whereas a one-step calculation only requires 5. I get the same result without rest=1, and I suppose that that the restart command has been ignored, maybe because after 2 iterations MRCC has stopped on an error (convergence fails)?
Is it a solution to overcome this problem?
2) The final geometry is provided in Cartesian coordinates and I wonder if it would be possible to obtain the Z matrix coordinates as well?
Many thanks in advance for any comments or help.
Jacky
Please Log in or Create an account to join the conversation.
- Boris
- Offline
- New Member
Less
More
- Posts: 2
- Thank you received: 0
1 year 4 months ago #1376
by Boris
Replied by Boris on topic restart geometry optimization
Dear Jacky,
1) The rest keyword controls the restart of the CI and CC calculations and has no effect on geometry optimization. The difference in the number of the steps is because the Hessian matrix is not saved and the optimizer always starts with a diagonal Hessian. Currently there is no option for restarting geometry optimization with the Hessian of a previous calculation. If wall time limit is an issue, please let me know and I can create a patch for you.
2) As far as I know, MRCC cannot transform Cartesian coordinates to Z matrix. You can use MOLDEN for this. If you have a large number of molecules you should check this python code: github.com/robashaw/geomConvert
Hopefully I could answer your questions.
József
1) The rest keyword controls the restart of the CI and CC calculations and has no effect on geometry optimization. The difference in the number of the steps is because the Hessian matrix is not saved and the optimizer always starts with a diagonal Hessian. Currently there is no option for restarting geometry optimization with the Hessian of a previous calculation. If wall time limit is an issue, please let me know and I can create a patch for you.
2) As far as I know, MRCC cannot transform Cartesian coordinates to Z matrix. You can use MOLDEN for this. If you have a large number of molecules you should check this python code: github.com/robashaw/geomConvert
Hopefully I could answer your questions.
József
Please Log in or Create an account to join the conversation.
- jlievin
- Topic Author
- Offline
- New Member
Less
More
- Posts: 4
- Thank you received: 0
1 year 4 months ago #1377
by jlievin
Replied by jlievin on topic restart geometry optimization
Dear József,
Thanks for your answer!
Yes, the wall time limit is a real issue for optimizing geometries with high level CC approaches as CCSDT or CCSDTQ. If it is not too complicated, it would be great to have the possibility of restarting the hessian from the previous step.
Using the python script is a convenient solution for obtaining the Zmatrix coordinates.
Best regards,
Jacky
Thanks for your answer!
Yes, the wall time limit is a real issue for optimizing geometries with high level CC approaches as CCSDT or CCSDTQ. If it is not too complicated, it would be great to have the possibility of restarting the hessian from the previous step.
Using the python script is a convenient solution for obtaining the Zmatrix coordinates.
Best regards,
Jacky
Please Log in or Create an account to join the conversation.
Time to create page: 0.041 seconds