ALPB and GBSA
xtb has two implicit solvation models. ALPB (analytical linearized Poisson–Boltzmann) improves on GBSA and is now the default choice. Both are parametrised for GFN1, GFN2 and GFN-FF.
Usage
xtb ala.xyz --opt tight --alpb water
xtb ala.xyz --opt tight --gbsa water
crest ala.xyz --gfn2 --alpb water
Solvents are named water (or h2o), methanol, acetonitrile,
dmso, thf, toluene, chcl3 and so on. The full list is in the solvent
table of the xtb documentation.
Example: geometry in the gas phase and in water
Alanine dipeptide was optimised from the same starting geometry in the gas phase and with ALPB(water). In the gas phase the ring-like C7eq structure with an intramolecular hydrogen bond is favoured; in water that hydrogen bond gains less and other regions compete.
| Gas phase | ALPB(water) | |
|---|---|---|
| Total energy (Eh) | −32.975847892266 | −32.991360406923 |
| φ / ψ (°) | −81.4 / 74.0 | −82.5 / 73.1 |
| HOMO–LUMO gap (eV) | 4.480 | 4.588 |
| Wall time | 0.52 s | 0.39 s |
Both optimisations ended in the C7eq region, about 1° apart in φ and ψ. Adding the solvent did not move the minimum nearest the starting geometry. That does not make C7eq the most stable structure in water: an optimisation only goes downhill, so it cannot tell whether another region lies lower. The ALPB total energy is 9.7 kcal/mol (0.0155 Eh) lower; since the two geometries are almost the same, the difference is mostly the solvation term. The CREST search for the same molecule in water is in the CREST chapter.
Caveats
- A single optimisation only reaches the minimum nearest the start. To see whether the most stable conformer changes with solvent, look at the whole ensemble with CREST.
- The difference between ALPB and gas-phase energies can serve as a solvation free energy, but for quantitative values DFT-level SMD or COSMO-RS is better.
- Continuum models cannot represent specific hydrogen bonds to solvent molecules. If they matter, add a few explicit solvent molecules.