Projecting existing def2-SVP into def2-TZVP without redoing def2-SVP

Questions
#1

I have a fully converged DFT calculation (using Q-Chem 6.0) that used a def2-SVP basis set. I am trying to use the converged wave function result from the def2-SVP calculation as a starting point for a def2-TZVP calculation, without the def2-SVP calculation having to go through another complete SCF calculation to get an (already existing) converged wave function. I cannot find a way to do this in the Q-Chem manual. The only thing I find is the “dual-basis approximation” in Section 4.7.1 of the manual that requires the def2-SVP calculation to be re-run before the def2-TZVP guess is generated.

Is there any way to have Q-Chem project the existing def2-SVP orbitals into a starting set for the def2-TZVP calculation and not require the def2-SVP calculation to be redone?

#2

What you need is basis-set projection, keyword BASIS2 in $rem.

#3

Thanks for your reply.

That is the section of the manual I referenced. The example given requires the smaller basis set SCF be run to completion before the larger basis set run can be performed.

I don’t see information provided on how to use the files from an already converged smaller basis set run, from a separate job, to be used in the projection without a complete SCF job being run in the smaller basis set.

Is there an example of using BASIS2 using files from a previously run SCF of the smaller basis set as the input for the larger basis set run without having to rerun the smaller basis set job?

#4

Just be careful because “dual basis” (a perturbative correction amounting to one Roothaan step in the larger basis set, following convergence in the smaller one) is different from “basis-set projection” (using small basis solution as a guess but fully converging the large basis). Both methods use the BASIS2 keyword. For what you want to do, you would have needed to save the small (def2-SVP) scratch files:

qchem -save small.in small.out small.scr

which would put the scratch files in $QCSCRATCH/small.scr. The “small.in” is a normal SCF with def2-SVP. To make sure you don’t accidentally overwrite the converged small-basis scratch files, I suggest

cp -r $QCSCRATCH/small.scr $QCSCRATCH/large.scr

Then run

qchem -save large.in large.out large.scr

where “large.in” has the following:

$rem
scf_guess read
basis def2-TZVP ! "target" (large) basis
basis2 def2-SVP ! small basis from previous calculation
... (etc) ...
$end

You should notice 2 SCF cycles in the small basis, which is the minimum number to guarantee that F and P are consistent. Then there’s a basis-projection step, may be a bit slow as I’m not sure how well parallelized that is as compared to the SCF cycles. Then the SCF cycles in the large basis should begin.

#5

Thanks John. I’ll give it a try tomorrow.