Re: About changing charge/multiplicity during QM-MM

From: Gerard Rowe (
Date: Mon Mar 25 2019 - 13:34:09 CDT


It sounds like you may have a fortunate outcome. I don't know the nature of your redox partner, but it's not an unreasonable assumption that the spin state will be the same between reactant and product if your partner becomes a ligand. If the redox partner started as a radical and became a closed-shell molecule, your final wavefunction is probably going to behave, and the trajectory could be well-suited for analysis through molecular orbital plots or spin-density plots. I think the most useful information will come out of spin density for a redox reaction. Things get more complicated (but not impossible) if the redox partner has unpaired electrons in it.

As for carrying out stand-alone Orca calculations from the temp folder files of QM/MM, you should copy the inp, gbw, and point charge files into a new folder. Unless you have a lot of experience with QM programs, I'd stick to single point calculations here because you would have to play a tricky game of atomic restraints for QM geometry minimization.

Your choice of level of theory is pretty good for QM/MM, but if I were trying to get a more reliable wavefunction, I'd probably change the basis set to def2-svp, and possibly specify def2-tzvp for the metal ion. If you don't have restrictions on how much CPU time you use for QM/MM, you may also play around with other local DFT functionals like PBE or M06L to see if they lead to different equilibrium structures. M06L, in particular, is parameterized to handle non-covalent interactions.


From: Francesco Pietra <>
Sent: Monday, March 25, 2019 12:09 PM
To: NAMD; Gerard Rowe
Subject: Re: namd-l: About changing charge/multiplicity during QM-MM

The redox partner is included but, regrettably, I have not yet carried out an analysis of what orca could tell. I simply observed that now the iron ions and the redox partner have reached a bonding distance, which remains stable on long QM-MM simulations, and the type of complex that is formed is experimentally known to have FeIII, never FeII. So, I came here much too early with my question, but, to go on, I'll take into account what you said. At any event the type of ligands of the Fe ions (in a protein) are for high spin, which over-complicates the matter. In addition, the simulation, because of the type and size of the system, was started at SlowConv but was continued at SOSCF in order to get convergence . All over it was at what might well be a too low level of DFT ("! UKS BP86 RI SV def2/J enGrad KDIIS SOSCF").

On Mon, Mar 25, 2019 at 3:53 PM Gerard Rowe <<>> wrote:

If your metal ion is definitely undergoing redox change, that's going to be very difficult to treat with QM/MM unless you have also included the redox partner in the QM region. Even then, it's going to be problematic because the spin state of your resulting system is unlikely to be reasonable. This is especially true if your final system is supposed to have unpaired electrons in different regions that aren't in communication anymore. What kind of system are you studying?

I can imagine two scenarios where this will work out:

  1. Your Fe(II) state is low-spin d6 (singlet) and the redox partner is a doublet. In the product state, The Fe(III) is a low-spin doublet and the redox partner is now a singlet.
  2. Your Fe(II) state is high-spin d6 (sextet) and the redox partner is a singlet. In the product state, The Fe(III) is a high-spin quintet and the redox partner is now a doublet.

If scenario 1 applies, you're in luck, and the spin system will be reasonable. If it's scenario 2, you're going to run into serious issues unless the redox partner happens to be bound to the metal. Long-range separation of unpaired spin in a DFT calculation often leads to problems and non-physical solutions. You can check the spin density of your resulting wavefunction in a similar way that you can visualize molecular orbitals.


From:<> <<>> on behalf of Francesco Pietra <<>>
Sent: Monday, March 25, 2019 3:16 AM
Subject: namd-l: About changing charge/multiplicity during QM-MM

I dare posing a general question on QM-MM, surely a naive question to those experienced with QM-MM (I am doing such simulations for the first time).

I am observing a change in my metalloprotein system where apparently the iron ions become bounded to other ligands, changing from FeII to FeIII. The overall charge/multiplicity is not maintained, as far as I can understand. Nonetheless, ORCA is repeating its SCF cycles, converging after a number of cycles. I have already restarted the simulation six times for 24hr on 36 cores with the same settings, in particular as to the charge and multiplicity, and the structure of the complex is no more changing, as far as it can be judged from geometry. I can't figure out how the system readjusts itself so that the QM calculation can go on.
Thanks for any comment
francesco pietra

This archive was generated by hypermail 2.1.6 : Tue Dec 31 2019 - 23:20:39 CST