Reaction-path calculations ask for the lowest-energy connected route between two declared states under one potential-energy model. They do not merely compare the endpoints. The output is a path in configuration space and, when it contains an appropriate first-order saddle point, a barrier relative to a stated reactant basin. It is not automatically a finite-temperature rate, a mechanism proved in experiment, or a guarantee that every alternative route has been found.
Open the image chain before trusting the barrier
Start with separately verified endpoint minima under one compatible model. Open both endpoints together, identify the atoms and periodic images that correspond, and generate more than one physically plausible interpolation when mechanisms can differ. Load the complete image chain in a structure or trajectory viewer and scrub through it before and during optimization. Look for atom swaps, close contacts, broken molecules, discontinuous boundary crossings, changing spin/charge identity, or a collective motion that the interpolation omitted.
Plot energy and projected force for every image, not only the fitted barrier. Refine the high-energy region and validate a candidate transition state by local curvature and connectivity to the intended basins. Converge the barrier against path initialization, image resolution, electronic settings, finite-size model, and saddle verification. ASE GUI, OVITO, code-specific NEB routes, and manuals are grouped under specialist tools. Visual continuity is necessary evidence but does not prove a first-order saddle. This overview does not claim an executed NEB path.
Make the path inspectable on disk: keep the two endpoint structures, the original interpolation, every numbered image, the common Hamiltonian settings, and one energy/force table keyed to those image numbers. ASE can open the complete image sequence in its GUI before it launches or parses a path. In a Quantum ESPRESSO route, preserve neb.in, neb.out, and neb.err separately and invoke the documented file-input form, neb.x -in neb.in > neb.out 2> neb.err, rather than ordinary stdin redirection. If images cross, atom mapping changes, one image switches electronic state, the climbing image runs to an endpoint, or the projected force stalls, save the adverse chain and revise the interpolation, image density, or parent states before quoting a barrier.
Endpoints define the question before a path can answer it
The initial and final structures must each be relaxed and identified as the intended states under compatible cell, composition, charge, spin, boundary, Hamiltonian, and numerical choices. For an adsorption event, for example, a different site, coverage, surface face, or reference proton/electron state changes the reaction being represented. A path connecting two local minima is conditional on those minima; it cannot decide which reactant population is available or whether a different product is thermodynamically preferred.
Let be a continuous path of atomic coordinates, parametrized by , with endpoints and . Its energy profile is . The forward potential-energy barrier is
Here is the highest relevant saddle-point configuration on the chosen minimum-energy path and all energies use the same normalization. This subtraction answers a zero-temperature potential-energy question. Vibrational free energies, field work, solvent, electrode potential, entropy, and populations are additional model terms, not implicit properties of a static barrier.
A nudged elastic band is a path optimization, not an ordinary relaxation
NEB represents the path with intermediate images. Springs maintain their order along the path while the physical force perpendicular to the local tangent relaxes each image toward the minimum-energy path. A naïve interpolation is only an initial guess: atom exchange, a poor collective coordinate, an abrupt bond rearrangement, or an unphysical close contact can place the chain on the wrong basin or cause it to cut a corner.
The number and placement of images control the resolution of the path; the spring treatment, tangent definition, optimizer, and image force criterion control its numerical behavior. There is no generally valid image count, spring constant, or force threshold. Refine where geometry or energy changes rapidly, compare physically distinct initial paths, and check that the final profile is stable to path representation rather than only to electronic SCF tolerance.
Climbing images refine a candidate saddle, but must still be checked
In climbing-image NEB, a selected high-energy image loses its spring force along the path and reverses the parallel component of the potential force. When neighbouring images give a reliable tangent, this drives that image toward a saddle point. It should normally be enabled only after a non-climbing path has become a reasonable representation; otherwise the selected image can climb an artefact of the initial chain.
A converged CI-NEB image is evidence for a saddle-point candidate, not by itself proof of a transition state. Verify its residual force in the appropriate projected sense and analyse the Hessian or a validated local curvature calculation. A first-order saddle has one unstable mode that connects the intended basins; multiple negative modes, a mode leading elsewhere, or an unconverged orthogonal direction changes the interpretation. The path maximum, the climbing image, and a validated first-order saddle are related but not interchangeable labels.
From an energy profile to a scientific claim
Report the endpoint identities, each image lineage, coordinate convention, energy reference, path initialization, optimizer and force definitions, image refinement tests, maximum-energy image, saddle validation, and every correction or thermodynamic term. Keep forward and reverse barriers separate when endpoints differ in energy. If a rate is claimed, state the kinetic theory, temperature, prefactor treatment, free-energy surface, recrossing assumptions, and whether nuclear quantum effects or multiple pathways have been considered.
Common failures include moving the wrong atoms during interpolation, inconsistent endpoint electronic states, allowing images to change cell or composition without defining that model, accepting a sparse path maximum as a saddle, and treating a lower barrier from one guessed path as global mechanism discovery. A static DFT NEB can support a conditional pathway and potential-energy barrier within its declared model. It cannot establish an experimental rate, selectivity, catalytic turnover, environmental mechanism, or exhaustive reaction network without further evidence.