Hi,
I'm performing ROCCSD for my small model (23 atoms) and calculations don't give me final energy after my server's time limit which i cannot extend.
Initial checkpoint file comes from ROHF calculations.
My input:
%oldchk=**.rohf.chk
%chk=**.ccsd.chk
%mem=24gb
%Nproc=24
# ROCCSD=(saveAmplitudes,maxcyc=250,FreezeNobleGasCore,T1Diag)
# NoSym guess=read gen 5d
then charge, multiplicity, geometry and basis set of course (ccpVDZ(C,H), ccpVTZ(Fe,N,O))
Tail of my log file looks like this:
" Iteration Nr. 138
**********************
DD1Dir will call FoFJK 46 times, MxPair= 84
NAB= 1292 NAA= 703 NBB= 561 NumPrc= 24.
DE(Corr)= -2.2764181 E(CORR)= -1868.8815128 Delta= 8.29D-07
NORM(A)= 0.13878515D+01
Iteration Nr. 139
**********************
DD1Dir will call FoFJK 46 times, MxPair= 84
NAB= 1292 NAA= 703 NBB= 561 NumPrc= 24."
Gaussian error file tells me that:
"JOB XXX ON XXX CANCELLED AT XXX DUE TO TIME LIMIT"
Is there anything I can do or the rohf geometry is not good enough?
The calculation simply takes too long for your cluster.
23 atoms are not a very small system for CCSD but is also not very large.
Normally, quantum chemical calculations do not require such a large number of iterations (138 in your case) to converge. There indeed may be some problems with your system. You can try to optimize the configuration with MP2 or some DFT methods before applying CCSD.
I have another question that whether CCSD is really needed for your system. Such high-level method is too slow for most systems and many technical problems could arise. For instance, if you are using QM/MM, the size of QM region is also a probelm. Checking the convergence on the QM size requires the calculation of larger systems, which is really prohibited for high-level methods.
Thank you for the answer.
The system was optimized with DFT (using triple-dzeta basis set) before applying CCSD.
I need CCSD method in order to obtain the most accurate energy of my system in order to use it in parametrization later on.
- Badges
- Science topic
- Similar topics
- Filing