Constraints

For electronically more complex systems, such as transition metal complexes, or if you wish to preserve specific structural motifs (e.g., a transition state), crest offers the option to constrain certain bonds, dihedrals, and torsion angles.

To do this, you can provide a file with a format similar to the xtb input using the –cinp flag:

crest struc.xyz --cinp <FILE> --subrmsd

The --subrmsd flag causes crest to compare only those parts of the structure that were also included in the metadynamics bias potential.

All possible constraints are listed in the documentation.

Note: Exact atom fixing is not possible due to instabilities during MD and MTD simulations.

To simplify the manual definition of constraints, crest offers the option to automatically generate the required input file. For this, the --constrain <atomlist> keyword can be used, where <atomlist> defines the indices of the atoms to be constrained, e.g.:

crest struc.xyz --constrain 1-5,8,12

This results in a .xcontrol.sample file, which can be renamed and used as input for the crest run as well as a coord.ref file used as reference for the constraining.

Additionally, crest provides the option to automatically constrain certain interaction motifs. For example, –cmetal constrains the bonds between a metal and all ligands to prevent unwanted behavior during MD and MTD simulations. More information can be found in the documentation.

Exercise

As an example, consider the transition state of the oxidative addition of benzene to an Ir catalyst. Search for conformers of the transition state with GFN-FF in THF by constraining the relevant atoms. To do this, examine the structure using software like Molden, display the labels and atom numbers, and take note of the atom numbers of atoms relevant for the transition state (you can also identify by performing a xtb frequency calculation and checking the imaginary frequency). Then, call crest with the --constrain <atomlist> flag, replacing <atomlist> with the appropriate atom numbers. Rename the resulting .xcontrol.sample file to a name of your choice, such as xtb.inp, and start crest again, providing this input file. As this is a electronically more complex molecule, increase the energy window to 30 kcal/mol (--ewin 30). Finally, verify that crest_best.xyz represents a reasonable transition state, similar to what was shown before.

   60
FINAL HEAT OF FORMATION =   -10.652236
Ir     0.183562     0.007648    -0.034276
 B     1.105603    -1.318961     1.507833
 O     1.750576    -0.765030     2.609861
 O     1.056026    -2.707974     1.409728
 C     2.218956    -1.890502     3.322169
 C     1.760197    -3.147882     2.551868
 B     1.830913    -0.804148    -1.261756
 O     1.569192    -0.985666    -2.621398
 O     3.102286    -1.126315    -0.790834
 C     2.786325    -1.492861    -3.123711
 C     3.796277    -1.516005    -1.956777
 B     1.356214     1.835824     0.243472
 O     2.297869     2.181902    -0.725072
 O     1.160116     2.714197     1.313025
 C     2.831475     3.397478    -0.243161
 C     2.109970     3.731991     1.079859
 C    -2.187317    -0.733022     1.465202
 C    -2.128710    -1.676766     0.343637
 N    -1.207227     0.200038     1.549249
 N    -1.094669    -1.556386    -0.527617
 C    -3.178877    -0.732774     2.438232
 C    -3.143042     0.198364     3.452403
 C    -2.124571     1.130718     3.501850
 C    -1.158193     1.111240     2.528940
 C    -3.069231    -2.673918     0.114600
 C    -2.933610    -3.508373    -0.972445
 C    -1.862043    -3.358253    -1.831075
 C    -0.945976    -2.368038    -1.583761
 C    -1.159801     1.404740    -1.125045
 C    -1.105228     2.754179    -0.875819
 C    -2.068280     0.907573    -2.026386
 C    -1.949133     3.618338    -1.544122
 C    -2.915488     1.762955    -2.702242
 C    -2.853341     3.121896    -2.461411
 H     0.313687     0.668710    -1.526625
 H    -0.349040     1.820852     2.518451
 H    -2.089589     1.864410     4.290756
 H    -3.910826     0.200329     4.209757
 H    -3.971000    -1.459019     2.401558
 H    -3.903732    -2.792065     0.782309
 H    -3.663763    -4.281399    -1.153805
 H    -1.746239    -4.007743    -2.683587
 H    -0.098647    -2.209801    -2.227424
 H     1.086337    -3.781560     3.144643
 H     2.603808    -3.765459     2.213226
 H     1.806418    -1.846429     4.338517
 H     3.312104    -1.815563     3.390209
 H     1.581634     4.693743     1.040438
 H     2.796937     3.752779     1.936209
 H     2.672051     4.165503    -1.012870
 H     3.914153     3.265315    -0.109524
 H     4.621871    -0.807232    -2.106982
 H     4.225315    -2.512403    -1.787757
 H     3.110396    -0.849172    -3.952446
 H     2.592694    -2.496512    -3.524796
 H    -0.400619     3.123541    -0.151170
 H    -1.901896     4.677345    -1.349469
 H    -3.513811     3.794581    -2.985626
 H    -3.625714     1.370496    -3.411901
 H    -2.119903    -0.155991    -2.183099

Please note that GFN-FF and GFN2-xTB are not reliable enough to predict accurate conformational ranking for such a system. For production runs, the ensemble has to be refined with DFT.