Switching to another root during CCSD calculation of N2 triplet

Dear Forum,

We calculate the energy of the lowest triplet of N2 by MCSCF followed by the CCSD calculation. Using the patches posted on the MRCC website convergence of my calculations require much less iterations then in the un-patched version and we get convergence also at longer bondlength's. In the un-patched version the calculation does not converge at longer N-N bond length's. 

It seems however that the calculation switches to another root during its iterations. for example at bond length R1=1.7775 The calculation is almost converged in iteration 23 to an energy that seems right to me (comparing with the energy found at a little smaller bond length)  at iteration 24 however the solution makes a jump and converges to this root. (output is attached). The same effect is also visible in the un-patched version although there is no convergence in that case

Is there a setting that might prevent this jump?

Thanks in advance for any help.

Jos Suijker
Technical University of Eindhoven
The Netherlands

Dear Jos,
You should try root following, that is, set diag=follow in the input.

Dear Prof. Kallay

Thank you very much, this solution worked perfect. The calculation converged easily last night. We should have found this keyword ourselves in the manual it is there.

Regards,

Jos Suijker