Skip to content

Ethanol

On hold

Alchemical transformation is on hold as of 2026-10-13 and is not part of the current release. This page was last run against 0.6.4. See On hold.

Tested against md-tools 0.6.4. Every command and every number on these pages comes from a run executed as written, at that version, on NVIDIA RTX 3080 GPUs with CUDA and mixed precision.

Nine heavy-atom-and-hydrogen atoms in a box of TIP3P water. Ethanol is here for one reason: it is the smallest system on which an alchemical free energy can be computed, checked against an experimental number, and checked again against a completely different alchemical route — all in about an hour of GPU time.

Why this system

Hydration free energy is the simplest free energy a simulation can be asked for. There is no binding site, no pocket water, no conformational selection: a molecule is removed from water, and the work of removing it is the answer. That makes it the right first alchemical calculation, and the right place to learn what the machinery does before pointing it at a protein.

Ethanol specifically, rather than ethane or a noble gas:

  • Its hydration is dominated by ELECTROSTATICS. Ethanol hydrogen-bonds; ethane does not. Of the two contributions FreeSolv separates for these molecules, ethane's is almost all cavity work and ethanol's is almost all charging. A ladder that is correct for ethane may barely exercise the part where electrostatic decoupling can go wrong, and that part is where the difficulty is.
  • It is in FreeSolv v0.52 as mobley_2310185, with an experimental hydration free energy of −5.00 kcal/mol. So there is an outside number.
  • It is small. 1,869 particles. Twelve windows of 1 ns each finish in under an hour spread over four cards.

One warning about the FreeSolv comparison, because it is the easiest way to draw a wrong conclusion from this tutorial. FreeSolv also publishes a calculated value for each molecule, and that column is GAFF/AM1-BCC. These pages use openff-2.2.1 (Sage)/AM1-BCC. It is a different force field, so that column is not the reference for anything here — agreeing with it would be reassuring, not validating, and disagreeing with it would not be a defect.

Build it

Exactly as for any other small molecule, and in two steps: a parameter package first, then a box that reuses it. A small molecule has no library of residue templates, so the parameters have to be produced before anything can be solvated.

ethanol.smi:

CCO ethanol

mkdir -p EOH && cd EOH
md-openmm build-top --parameterize -i ethanol.smi --config param.config \
    -op build/parameter/EOH.pdb -os build/parameter/EOH.xml \
    -log build/parameterize.log --resname EOH

A .smi states the chemistry and no coordinates, so a conformer is embedded. Coordinates are not part of a package's identity, so the same SMILES gives the same package whichever conformer comes out of the embedding.

Then the box, reusing that package rather than parameterising again. build.config:

# build.config -- ethanol in a cube of TIP3P, no salt
solute:
  kind: ligand
  residue_name: EOH
  parameters: ./build/parameter        # a PATH, relative to this file

solvent:
  model: TIP3P
  padding_nm: 1.35
  box_shape: cube
  ionic_strength_molar: 0.0
  cutoff_nm: 0.9

constraints:
  type: HBonds
  rigid_water: true

md-openmm build-top --config build.config -odir build

param.config is the parameterisation half — force field and charge method — and it is the same file the paracetamol page walks through key by key; nothing about it is specific to this molecule.

No salt, and that is deliberate. Decoupling changes what is in the box, and a neutral molecule in a neutral box keeps the net charge fixed at zero throughout — which is what lets this calculation avoid a finite-size correction entirely. (A charged ligand does not, and combine-topology refuses to build a decoupling plan for one, by name.)

Where to go next

method what it is for
Hydration free energy ΔG of taking ethanol out of water, by alchemical decoupling published
cMD ordinary dynamics on this box planned
REST2 nothing to enhance here — one rotatable bond not planned

A row with no link is a page that does not exist yet. It is listed so the method is visible, not to suggest it is written.