Restarting coupled-cluster

Thank you, that is helpful.

The memory information immediately preceding the warning is the same as in the original (interrupted) calculation. But until the first iteration is complete, I will not know if the amplitudes were read correctly from fort.16.

>> Memory requirements /Mbyte/:
>> Minimal Optimal
>> Real*8: 125253.5563 125253.5563
>> Integer: 136.3759
>> Total: 125389.9322 125389.9322
>>
>> The number of active orbitals has been changed!
>> It is dangerous to restart the program!

In my experience, the lines:

>> The number of active orbitals has been changed!
>> It is dangerous to restart the program!

are not a problem if the block of lines:

>> Memory requirements /Mbyte/:
>> Minimal Optimal
>> Real*8: 125253.5563 125253.5563
>> Integer: 136.3759
>> Total: 125389.9322 125389.9322

during the last execution of xmrcc, are the same as they were in the original (interrupted) calculation. Let's see how the first two iterations turn out. The first iteration might have a high residual vector norm, but the second one will be roughly the same size or smaller than the last norm value from the interrupted calculation (if everything is happening correctly).

With best wishes!
Nike

In the interrupted calculation (irest=2), the "norm of residual vector" reached 0.000039 (iteration 19). In the restart (irest=1), after "Initial cluster amplitudes are read from unit 16", the norm was 1.216 (iter 1), 0.380 (iter 2), ... 0.0077 (iter 9). This is worse than the original irest=2 (norm=0.139 at iter 1 and 0.0011 at iter 9). So the restart did not work as I expected.

In the interrupted job the energy at iter 1 was -193.97293885 and at iter 19 was -193.97634401. Upon restart, the energy at iter 1 was -193.97524292. So the restarted initial energy was much better than the original starting energy. Could this mean that low-order amplitudes were read correctly, but not high-order amplitudes?

There was one time when the restart did work as I expected. Maybe that job was
interrupted at a lucky moment during the iteration.

Thanks and best regards,
Karl