Docs are under construction. Content may be incomplete or change.

Read and Compare Energies

Page type: Task guide
On this page

Tako displays energies from several operations, but the scientific quantity depends on the operation and result field. A Single Point energy belongs to the supplied geometry. An Optimization final energy belongs to the final optimizer frame. An MD final energy is one instantaneous endpoint. Molecular vibration ZPE is a separate nested correction. This page describes the interface and artifacts; the quantity hierarchy is developed in Comparable Energies and Thermodynamic Cycles.

Identify the operation and result state

Select the calculation in Explorer or the right Calc panel. Before reading a number, confirm:

  • operation type;
  • status is Completed rather than Running, Failed, or Canceled — Completed only means the worker returned normally, not optimizer convergence, correct chemistry, a minimum, or thermodynamic comparability;
  • converged is true when the operation reports it;
  • final structure is the intended state, with method, charge/spin, environment, cell, and constraints matching the comparison protocol.

Read the authoritative field

Open result.json and use the operation-specific field:

OperationPrimary energy fieldInterpretation
Single Pointenergy_ev / energy_hartreemodel energy at the exact input geometry
Optimizationfinal_energy_ev / final_energy_hartreefinal accepted optimizer geometry; require converged: true
Reaction pathimage/profile fields and reaction_barrierpath electronic/model energies, not free-energy barriers
Molecular dynamicsfinal_energy_ev plus trace/frame energiesinstantaneous or sampled trajectory potential energy, not a minimum/free energy
Vibrationvibrations.zero_point_energy_evharmonic ZPE only; no electronic energy in the vibration result

The headline Calc panel can show a final energy for common operations, but result.json is authoritative for field identity, convergence, settings, and units. Preserve the Hartree value when available and convert only for presentation.

Read force convergence separately

Energy change alone is not an optimization convergence test. For Optimization, inspect final_fmax_ev_per_angstrom, the requested settings.fmax_ev_per_angstrom, steps, and converged; in variable-cell relaxation the reported maximum can come from a scaled cell row rather than an atom, so it is not guaranteed to be a Cartesian atomic force. Single Point forces diagnose whether a supplied geometry is stationary on that model surface — a low relative energy with large forces is a low sampled point, not a relaxed minimum.

Build a comparison table

Create one row per physical state:

StateFormula/cellCharge/spinMethod/settingsGeometry policyConvergedEnergy / eVRelative / eV

Choose one reference only after all compatibility checks pass, then calculate ΔEi=EiEref\Delta E_i=E_i-E_{\mathrm{ref}}. Do not overwrite absolute energies with shifted values; keep both so the comparison can be audited. When compositions differ, a simple minimum-relative table is normally invalid — build a balanced cycle instead. Cycle construction and its rationale (deformation, coverage, sign convention) are developed in Comparable Energies and Thermodynamic Cycles.

Add ZPE only from matching calculations

Open the matching Vibration result.json and read vibrations.zero_point_energy_ev, verifying that its input geometry is the same optimized state and uses the same model, charge, spin, and constraints. Calculate E0=E+ZPEE_0=E+\mathrm{ZPE} externally; Tako does not create this sum or a thermochemistry artifact. Do not use imaginary-mode magnitudes as stable harmonic contributions — a meaningful imaginary internal mode invalidates minimum RRHO thermochemistry.

Build reaction and binding cycles

For every term, record signed coefficient, composition, charge, spin, method, geometry, and correction level. Verify atom and charge balance before arithmetic, and never subtract a GFN2 total energy from an MLIP total energy — build the entire cycle within each method and compare derived differences. The interface has no controls or artifacts for temperature/pressure, partition functions, thermal/entropic corrections, enthalpy, Gibbs energy, standard-state conversion, or conformer-ensemble free energy, so report results as an electronic/model-energy difference (optionally with separately added harmonic ZPE) unless those terms were calculated externally. Binding, adsorption, and voltage sign conventions and the full model-energy-to-free-energy quantity hierarchy are developed in Comparable Energies and Thermodynamic Cycles.

Check units and normalization

Tako result fields state Hartree or eV. A comparison additionally needs a denominator: per molecule, atom, formula unit, surface area, adsorbate, inserted ion, or transferred electron. Do not compare “eV/atom” values obtained from cells with inconsistent formula definitions; for voltage, eV per transferred electron is numerically volts only after electron count and reaction direction are defined.

Reject invalid terms

Exclude a state when any of the following applies:

  • failed/canceled calculation;
  • explicit converged: false for a required optimized state;
  • wrong charge/spin/composition or malformed geometry;
  • mismatched calculator asset, dispersion, solvent, SCF, cell, or constraints;
  • missing reference species or unbalanced cycle;
  • reaction endpoint/path not representing the intended mechanism;
  • vibration correction from a different state/protocol;
  • meaningful imaginary mode in a claimed minimum;
  • trajectory endpoint presented as an equilibrium/free-energy state.

Do not replace a missing energy with zero, and do not keep a state that failed the checks above.

Report the comparison

State the named quantity and omissions. A precise minimal statement is:

Relative GFN2+D4 model energies at separately optimized geometries, reported in eV relative to conformer A; no ZPE, thermal, entropic, conformational-ensemble, or standard-state corrections.

For corrected data, list each contribution separately before the total. Preserve raw result files, final structures, vibration files, calculation settings, and the comparison table.

Diagnose common mistakes

ObservationProblemResponse
lower energy but converged: falseoptimizer stopped before force gatereject or continue optimization
two methods differ by huge absolute valueenergy zeros differcompare balanced differences within each method
binding value changes after fragment relaxationdeformation policy changedreport interaction and deformation components separately
MD endpoint ranks below optimized stateendpoint is an instantaneous configurationoptimize/average according to the scientific question
“free energy” table contains only energy_evthermal/entropy terms were omittedrename to model/electronic energy or calculate corrections externally

For reproducible arithmetic and exact helper behavior, use Energy Analysis with Tako Script.