MRCCSD optimisation stops without error

Hello,

I'm trying to use MRCC to calculate a MRCCSD calculation. To make sure this worked, I copied (and adapted) the CH+ test that you have in the MTEST directory (I removed the check of the energy, and added gopt=full and freq=on). After 14 MCSCF steps, the energies swapped between either -37.9443749851044956, -37.9443749851044885 and, a few times, -37.9443749851045027 au. It continued to do this until it ran out of steps. The calculation, however, continued to the coupled cluster gradient calculation. After the LR-CC iterations had completed, the calculation stopped without warning or any error. As far as Slurm is concerned, the calculation completed successfully, so the calculation was not killed.

I'm not quite sure what the problem is here. I'll attach the output below. I have run a few calculations with MRCC so far (simple CCSDT/cc-pVDZ optimisations/frequencies on H2O and CH3NH2) and had they completed without issues, as far as I could tell.

I compiled MRCC using this command:./build.mrcc Intel -pOMP -pMPI=IntelMPI -llibxc -f/home/granem94/mrcc/build

I am running MRCC 25.1.2 on a HPC (MRCC is installed in my home directory, though, as the HCP admins use EasyBuild to install programs officially and MRCC is not on EasyBuild). I wrote a script that just basically loads the compiler and starts MRCC using a command similar to: dmrcc MINP > output.out. The version of ifort I used was 2021.4.0. I did use a newer version of MKL, but I was told by the HPC admins that it shouldn't make too much difference.

Unfortunately, analytical gradients are not available with MCSCF orbitals.

Oh... but then how do you run an analytic MRCCSD calculation?

Also, how would that affect the behaviour of the MCSCF calculation?

And why didn't the program end with an error or warning?