Ensemble Generation

The main feature of crest is the generation of an ensemble and the search for the global minimum of a structure. This can be done with the following command:

crest struc.xyz

This results in a file called crest_best.xyz containing the best conformer found, crest_conformers.xyz containing the conformer ensemble sorted by energy and without duplicates, and the crest_rotamers.xyz file containing an ensemble of rotamers sorted by energy.

Many features of xtb can be used also with crest, for example, implicit solvation and the different GFN methods. The command line flags are similar to xtb, e.g., --alpb <solvent> or --gfnff.
You can define the number of cores to use with the **-T ** argument. All available flags are listed in the [documentation](https://crest-lab.github.io/crest-docs/page/documentation/keywords.html).

crest comes with mostly robust defaults. However, setting robust defaults covering the complete chemical space is not possible. Therefore, it might be necessary to adjust these defaults like the MD length (--mdlen <REAL>). To get a feeling on whether the settings are reasonable, the --keepdir flag can be used when starting a crest run. This will leave a MDFILES folder with different trajectories of the crest runs. Inspecting them with, e.g., Molden can be used to check how the molecule behaves during the simulations. More on this is explained in the respective chapter.

Warning: Per default, crest generates an ensemble with structures in a certain energy window, e.g., for a conformer search 6 kcal/mol. Every conformer above this window will be discarded and is lost forever. Especially for electronically challenging systems, where the applied method, e.g., GFN2-xTB, is expected to have larger errors, important conformers are lost and cannot be recovered by following ensemble refinements. Thus, it is very important to increase the energy window for such cases with the --ewin <REAL> flag.

Exercise

As a simple example, compute the ensemble of histidine including implicit water solvation with the GFN-FF method.

20

O     -1.0689000000   -1.6919000000   -0.7660000000
O     -2.4946000000   -1.1223000000    0.9038000000
N      1.6797000000    0.1409000000    1.1624000000
N     -2.7494000000    1.3693000000   -0.2178000000
N      2.8201000000   -0.3365000000   -0.6825000000
C     -0.3579000000    1.3004000000    0.3217000000
C     -1.5062000000    0.6397000000   -0.4567000000
C      0.9671000000    0.6020000000    0.1195000000
C     -1.7498000000   -0.7917000000   -0.0102000000
C      1.6556000000    0.3150000000   -1.0031000000
C      2.8043000000   -0.4249000000    0.6290000000
H     -0.5942000000    1.3196000000    1.3946000000
H     -0.2433000000    2.3450000000    0.0039000000
H     -1.3124000000    0.6416000000   -1.5355000000
H      1.4441000000    0.1943000000    2.1450000000
H      1.3992000000    0.5299000000   -2.0301000000
H     -2.6419000000    2.3346000000   -0.5280000000
H     -2.9358000000    1.4198000000    0.7835000000
H      3.5655000000   -0.8810000000    1.2458000000
H     -1.2415000000   -2.6142000000   -0.4801000000

After the CREST calculation, you will find an ensemble and some additional files in your working directory. The ensemble will be dumped in XYZ format in a file called crest_conformers.xyz. You can have a look at the ensemble file by opening it with, e.g., molden.

Warning: CREST works stochastically and is thus non-deterministic. For molecules with higher flexibility, the results of CREST might differ and multiple runs should be used.