Constrained Coordinate Scans

Introduction

When a reaction path is unknown, a researcher often begins by stretching a bond, rotating a dihedral or moving a proton through a series of fixed coordinate values. Such a coordinate scan can reveal where energy rises and suggest a transition-state geometry. It is an exploratory map, however, not proof of a minimum-energy path. The chosen coordinate can force an unnatural route, hide a coupled motion or miss a barrier entirely.

Core explanation

An unrelaxed scan varies a selected coordinate while leaving other coordinates fixed. It is fast but often produces high artificial energies because atoms cannot reorganise. In a relaxed scan , the selected bond length, angle or dihedral is constrained to a specified value and all remaining allowed coordinates are optimised. Relaxation gives a more useful constrained minimum at each point, yet it still prescribes the coordinate and can switch discontinuously between local conformational valleys.

Consider transfer of H from donor D to acceptor A. Scanning the D–H distance alone can reveal a rough energy crest. But the A–H distance, D···A approach angle and solvent alignment may all change. If a different coordinate is the true bottleneck, a D–H scan may pass over a high ridge rather than the lowest route. A simultaneous two-coordinate scan, such as D–H and A–H distances, can expose coupled changes but costs more calculations. A combination coordinate, for example r(D–H) − r(A–H), can sometimes describe transfer better, yet it still needs validation.

Scan direction can matter. Starting each constrained optimisation from the preceding point usually follows the nearest local minimum. A forward scan may remain in one conformer while a reverse scan remains in another, producing hysteresis . This is useful evidence that the surface has more than one valley, not merely a software nuisance. Separate starting guesses or constrained optimisations may uncover hidden branches. Coarse step spacing can leap over a narrow barrier or yield a misleading maximum; excessively fine steps cost time and may still follow the wrong valley.

A scan maximum is usually not a fully optimised first-order saddle, because its gradient along the constrained coordinate need not vanish exactly and all other curvature directions may not have the required signs. The ORCA documentation, for example, uses the highest relaxed-scan structure as a starting guess for a dedicated transition-state optimisation. After optimisation, a frequency calculation and both IRC branches should verify saddle order and connectivity. These checks separate exploratory scanning from a defensible mechanism.

Energy values from a scan also need careful interpretation. A vertical cut through fixed geometries differs from a relaxed profile, and both differ from a thermal free-energy profile. For reactions in solution, a constrained electronic geometry scan at one solvent configuration misses solvent fluctuations and configurational entropy. For a dissociation, forcing a finite bond distance may leave an artificial contact geometry rather than genuinely separated products. The curve's apparent barrier can therefore depend strongly on arbitrary constraints.

Step-by-step reasoning

Choose a coordinate suggested by plausible bond changes and inspect its endpoints. Perform a relaxed scan over a physically sensible range, using enough points to see changes without assuming smoothness. Repeat from the opposite side when branch switching is possible. Inspect every geometry near an energy crest and look for coupled motion or discontinuities. Use promising structures as saddle-search guesses, then run a Hessian check and IRC. Revise the coordinate if the saddle connects the wrong minima.

Visual explanation

Draw a two-dimensional contour map with axes r(D–H) and r(A–H). A straight line at fixed r(A–H) climbs over a ridge, while a curved minimum-energy valley bends around it. Mark dots for sampled scan points and a star for the true saddle. In a second graph show a forward and reverse scan tracing different energy curves because they occupy different conformational valleys.

Real-world analogy

Mapping a mountain by walking only due east can reveal a ridge but may miss an easier pass to the northeast. Allowing north–south movement at each eastward distance resembles a relaxed scan, yet forcing eastward progress can still miss a winding route. The analogy helps distinguish a selected coordinate from the full multidimensional surface.

Real-world example

For a molecular torsional isomerisation, scanning a dihedral while relaxing other geometry often finds a high-energy region that seeds a transition-state search. If an intramolecular hydrogen bond breaks during rotation, the forward scan might keep the original hydrogen-bonded arrangement longer than the reverse scan restores it. The mismatch signals multiple conformational branches that require individual optimisation and connectivity checks.

Why?

Why relax unused coordinates? Molecules normally redistribute bond lengths and angles as one coordinate changes. Why scan both directions? Local optimisation can follow different valleys depending on its initial geometry. Why refine a maximum? A constrained maximum need not satisfy the zero-gradient and one-negative-curvature conditions for a saddle. Why try multiple coordinates? Chemical motion can be strongly coupled.

Common misconception

The highest point of a relaxed scan is not automatically a transition state. It is a constrained-optimisation result that may contain residual forces or lie on an artificial route. Another misconception is that a smooth one-dimensional energy curve proves a single-step mechanism; hidden orthogonal minima and alternate channels can be invisible in that projection.

Worked example

Question: A relaxed scan of a forming C–N bond gives its largest energy at 2.1 Å. The structure still has a sizable force along another internal coordinate, and a saddle optimisation moves the N approach angle by 35°. Which geometry should define the barrier?

Reasoning: The scan fixed one distance while permitting a local optimisation at each value. Its maximum was a useful initial guess, but the residual force and substantial angular change show it was not a stationary saddle. The refined geometry must be checked for one imaginary mode and proper endpoints. Barrier energies should use the verified saddle and appropriate reactant reference, not the unrefined scan point.

Answer: Use the validated first-order saddle after frequency and IRC checks; the scan maximum is only a starting guess.

Quick check

1. What distinguishes a relaxed scan from an unrelaxed scan? Answer: A relaxed scan optimises all unconstrained coordinates at each prescribed value of the scanned coordinate.

Exam focus

State what is constrained, what relaxes and which energy is plotted. Explain why the coordinate choice can bias a path. Distinguish a scan maximum from a true first-order saddle and identify frequency plus IRC calculations as follow-up validation.

Advanced insight

In multidimensional optimisation, a fixed scan value can have several locally stable solutions for the remaining coordinates. An optimiser generally finds one solution near its starting geometry rather than a guaranteed global constrained minimum. Forward–reverse hysteresis and sudden jumps are therefore evidence of multiple branches. Adaptive scanning, two-dimensional surfaces or chain-of-states methods can improve exploration, but each still inherits the chosen electronic surface and sampling limitations.

Summary

Coordinate scans probe a candidate path by prescribing a geometric variable. Relaxed scans allow other coordinates to respond and are useful for locating saddle guesses, but an imposed coordinate can create artifacts or hide alternative routes. Inspect scan geometries and direction dependence, refine the peak to a first-order saddle, and confirm both endpoints before interpreting a mechanism or barrier.

Practice questions

1. Why can an unrelaxed scan overestimate the energy of a bond rearrangement? Answer: Other atoms and internal coordinates cannot relax as the selected coordinate changes.

2. What does different forward and reverse scan behavior suggest? Answer: The constrained surface may contain multiple conformational valleys or path branches.

3. What two calculations should follow a saddle guess from a scan? Answer: A transition-state optimisation and frequency calculation, then IRC tracing for endpoint verification.

4. How could a proton-transfer scan miss important solvent effects? Answer: One fixed solvent geometry and electronic-energy curve do not sample solvent rearrangement or solution entropy.

Sources: ORCA 6.1 Manual, surface scans; ORCA 6.1 Manual, transition-state searches.