Uhf ccsd(t) optimization


I am trying to optimize a radical using uhf ccsd(t) (since the gradients are available) but my calculation stops after three cycles or less. It usually gets stuck at ‘Computing (T) contributions to CC density…’ I tried using cachelevel 0 and print 2 but it is not enough.
I would appreciate any suggestions on to run these types of jobs.


Could you provide the input file that you are using?

Yes, it looks like this:

memory 55000 mb

molecule {
0 2
C 1.313939694789 -0.701279287504 0.000410981520
C 1.451970099761 0.671261841198 -0.000047101955
C 0.188096572370 1.195752463848 0.000303879890
C -0.767219489460 0.097046462496 -0.000044051768
O 0.040597331215 -1.093507006625 -0.000309993224
H 2.058744754948 -1.477374315903 0.000355010857
H 2.390051501460 1.196555898414 -0.000214138689
H -0.120613208728 2.225373962714 0.000463825978
O -1.953923815055 0.023595105110 -0.000196037747

set globals {
reference uhf
basis cc-pvtz
dertype first
max_disp_g_convergence 1e-6
max_force_g_convergence 1.0e-6
max_energy_g_convergence 7
e_convergence 10
r_convergence 10
cachelevel 0
print 2


Thanks for the input. It seems your scratch directory is getting full because of which you are getting a psio write error. If you have not set your PSI_SCRATCH variable, psi4 writes all the scratch files to the /tmp folder which might not be big enough for these kind of calculations. I suggest setting your PSI_SCRATCH variable appropriately.