click atoms here to define each coordinate
energy
0max
Guided tutorials · worked examples
Molecule
Scan coordinates (define exactly 2)
Mode
Force field
Energy units
—
—
Point—
Relative energy — kJ/mol
Span (barrier) — kJ/mol
Path ends (minima only) — hold 1 / 2, then click the map

Hold 1 / 2 and click a green pin — or leave both empty for the two deepest.

1. Load a molecule · 2. click atoms to define the two scan coordinates · 3. Run scan.

Nice work! 🎉

PES Explorer — how it works

A potential-energy surface is the molecule's energy as a function of a few chosen internal coordinates. You pick two, the scan maps the energy over their grid, and everything else here reads that map.

Molecule

Type a name or SMILES (butane, CCO) and press Load, pick a family and hit 🎲 for a random one, or search the catalog. Anything not in the catalog is embedded and relaxed on the fly before the scan starts.

Scan coordinates

Define exactly two coordinates — they become the X and Y axes of the surface. Click a card to make it active, then click atoms in the molecule:

  • 2 atoms → a bond length (Å)
  • 3 atoms → a bond angle (°)
  • 4 atoms → a dihedral / torsion (°)

A full card resets itself on your next atom click, and finishing one jumps straight to the other, so you can pick both in one run. The from / to / steps boxes set that axis's range and how finely it is sampled — the scan costs one relaxation per grid node, so 20 × 20 is four hundred of them. A bond inside a ring cannot be scanned (nothing can rotate about it without breaking the ring).

Relaxed vs. rigid

Relaxed: each grid node holds the two scanned coordinates at their target values and lets every other atom settle. This is the true surface — the energy of the best structure available at that pair of coordinates — and it is what you want for barriers.

Rigid: freeze the molecule and only move the scanned atoms, then take a single-point energy. Much faster, and it shows raw sterics — the clash a relaxation would have relieved.

Force field

Every energy on the surface comes from the chosen SO3LR model. A different model has a different energy zero and a slightly different shape, so switching reloads the molecule rather than mixing two surfaces in one plot.

The second model (UMA) is far too slow to drive a whole grid; it is used only to refine a traced reaction path — re-optimizing its R / TS / P landmarks — with the button under the path controls.

Energy units

Purely a display choice: everything is computed in eV and converted for printing, so switching the unit re-prints the readouts, the colourbar and the chart without recomputing or rescanning anything. Every energy shown is relative to the lowest point of the scan.

Reading the surface

  • Hover any point to fly the molecule viewer to that geometry; click to pin a small structure on the surface and select it. Clicking never deletes — select a pin and press Delete, right-click it, or use its ✕ in the list.
  • Height and colour are the energy, from the lowest point of the scan to the highest; the colourbar over the graph gives the scale.
  • ⤓ Roll downhill walks the selected point into the nearest basin and pins the minimum it lands on. Steepest hops between grid nodes; gradient and momentum step down −∇E over the smooth interpolant.
  • ∇ Compute forces & critical points takes the in-plane gradient of the scanned grid and classifies every node by the ring of eight neighbours around it. Each kind gets its own shape, so you can read it off the surface at any zoom and over any colour of terrain: a green pin is a minimum, an amber four-pointed star a saddle (up two ways, down the other two — a transition state), a small hollow red triangle a maximum. Neighbouring nodes of the same kind — a flat basin classifies all of itself — are merged into the one marker at the cluster's extreme node. The arrows switch shows the −∇E field, and the colour toggle recolours the surface by |∇E| — which is zero at every critical point, so the blue spots are exactly the stationary ones.
  • Hiding the markers again: once the points are computed, the same button is the show/hide switch — click it (or press p) to take every marker off the surface and click it again to bring them back. Each entry of the key beside it — minimum / saddle / maximum — hides just that kind, so you can keep the minima and drop a crowd of saddles. Nothing is recomputed and nothing is lost: the points are still there, so a reaction path and the ① / ② ends work exactly the same while they are hidden. Only a new molecule or a new scan actually clears them.
  • Path ends: only minima may be endpoints. Hold 1 (or click ① start), move over the surface — the nearest minimum previews in red — and click to drop it; 2 does the same for the end. Esc cancels, ✕ forgets both, and with neither set the trace runs between the two deepest minima.
  • ⛰ Trace reaction path finds the route whose highest point is as low as possible, so it necessarily crosses the lowest saddle: it reads reactant → transition state → product and reports the barrier both ways. It also pops up the energy projected onto each scan coordinate.
  • ✦ Refine path re-optimizes just R, TS and P with the second force field, holding the scanned coordinates fixed, and reports that model's barriers beside the scan's.

Sampling — dynamics on this surface

Drop a walker on the finished surface and run Langevin (thermal) dynamics on the very energy you just mapped — not a fresh simulation, the same interpolated surface you can see. It rolls into basins, rattles around, and if it is warm enough it hops the barriers.

The surface starts ghosted and lights up wherever the trajectory goes, so a run literally paints what it explored. Start from different points, with Keep previous runs on, to compare what each start reaches.

  • Temperature is the one control that changes the sampling: it sets the thermal energy the walker carries against the barriers on this surface (kT vs ΔE). You can slide it while the dynamics is running — the walker takes the new temperature immediately, so you can warm a trapped run until it escapes and watch it happen.
  • Steps sets how long the trajectory runs; the timestep only paces it — a larger one covers ground faster but samples more coarsely.
Energy along the reaction path

Projected onto each scan coordinate — the traced R → TS → P path over the whole surface's energies.

RTSPall scanned points