We optimize the lattice and test several distances
The system can be cut in several variations into a a QM and an MM part breaking a B-N bond
variation QM atoms
var1 B(1),H(2)
var2 B(5),H(6)
var3 N(3),H(4)
var4 N(7),H(8)
var5 B(1),H(2),N(3),H(4)
var6 B(5),H(6),N(7),H(8)
Variation one is equivalent to variation two, and variation 3 should be equivalent with variation 4
Variation five is equivalent to variation six
Here are the distances (Angstrom) as obtained with a QM and an MM method
distance qm mm
B(1)-H(2) 1.182 1.185
B(5)-H(6) 1.182 1.185
N(3)-H(4) 1.007 1.045
N(7)-H(8) 1.007 1.045
B(1)-N(3) 1.431 1.508
B(5)-N(7) 1.431 1.508
Now we try the hybrid engine with several variations for the QM region
Two capping methods are tried as well.
variation capping energy B(1)-H(2) B(5)-H(6) N(3)-H(4) N(7)-H(8) B(1)-N(3) B(5)-N(7)
var1 fixed -2.901499 1.184 1.185 1.045 1.045 1.508 1.508
var1 fractional -2.787165 1.198 1.182 1.044 1.044 1.673 1.505
var2 fixed -2.901499 1.185 1.184 1.045 1.045 1.508 1.508
var2 fractional -2.787165 1.182 1.198 1.044 1.044 1.505 1.673
var3 fixed -4.791110 1.184 1.184 0.993 1.044 1.508 1.507
var3 fractional -4.733046 1.184 1.184 0.997 1.045 1.657 1.506
var4 fixed -4.791110 1.184 1.184 1.044 0.993 1.507 1.508
var4 fractional -4.733046 1.184 1.184 1.045 0.997 1.506 1.657
var5 fixed -6.741093 1.189 1.187 1.003 1.045 1.390 1.489
var5 fractional -6.648313 1.198 1.183 1.003 1.045 1.405 1.505
var6 fixed -6.741093 1.187 1.189 1.045 1.003 1.489 1.390
var6 fractional -6.648313 1.183 1.198 1.045 1.003 1.505 1.405
Here are some observations
* generally the fixed capping seems a bit better
Here are some remarks
* Starting from the initial very bad structure the fixed capping fails completely for variant 5 and 6
(not if you use as qm engine band and as mm engine dftb)
* A reasonable starting geometry can avoid strange collapses
* The more the two engines disagree about the capped QM region, the stronger the capping forces