Predictor–Corrector Algorithm¶

The Minimum Energy Path (MEP) is generally a curved path on the Potential Energy Surface (PES). Because of this curvature, Gaussian cannot simply move in a straight line from one IRC point to the next.

Instead, Gaussian employs a Predictor–Corrector Algorithm, a numerical procedure that repeatedly predicts the next point along the reaction path and then corrects that point so that it lies precisely on the Minimum Energy Path.

This approach enables Gaussian to follow complex reaction pathways accurately while maintaining numerical stability.

Why is a Predictor–Corrector Algorithm Needed?¶

Suppose the reaction pathway were perfectly straight.

●────●────●────●────●

In this simple case,

moving directly from one point to the next would accurately describe the reaction.

However,

real reaction pathways are rarely straight.

Instead,

they typically curve across the Potential Energy Surface.

If Gaussian followed only straight-line displacements,

the calculation would gradually drift away from the true Minimum Energy Path.

The Predictor–Corrector algorithm prevents this from happening.

The Predictor Step¶

The first stage is the prediction.

Using information from the current IRC point,

including

the reaction coordinate,
the energy gradient,
the local curvature,

Gaussian estimates where the next IRC point should lie.

Current IRC Point
        │
        ▼
Predict Next Point
        │
        ▼
Estimated Geometry

This predicted geometry is only an approximation.

It does not usually lie exactly on the Minimum Energy Path.

The Corrector Step¶

Once the predicted geometry has been generated,

Gaussian performs a correction.

During this stage,

the geometry is adjusted using the local gradient and curvature until it lies on the Minimum Energy Path.

Predicted Geometry
        │
        ▼
Geometry Correction
        │
        ▼
Correct IRC Point

The corrected structure is then stored as the next point on the reaction pathway.

Predictor–Corrector Cycle¶

The complete procedure is repeated throughout the IRC calculation.

Current IRC Point
        │
        ▼
Predict Next Point
        │
        ▼
Correct Geometry
        │
        ▼
New IRC Point
        │
        ▼
Repeat

By repeatedly predicting and correcting,

Gaussian constructs the entire reaction pathway one point at a time.

Why Doesn't Gaussian Simply Use Geometry Optimization?¶

Geometry Optimization searches for a stationary point where

\[ \nabla E = 0. \]

An IRC calculation has a different objective.

Instead of finding a stationary point,

it follows a pathway connecting stationary points.

Consequently,

each IRC point must satisfy two conditions:

remain close to the reaction coordinate,
remain on the Minimum Energy Path.

The Predictor–Corrector algorithm enforces both conditions simultaneously.

Following Curved Reaction Paths¶

Because the Potential Energy Surface changes continuously,

the direction of the reaction coordinate also changes continuously.

Transition State
      ●
     /
    ●
   /
  ●
 /
●

Reactants

The Predictor–Corrector algorithm automatically adjusts the direction of motion at every IRC point,

allowing Gaussian to follow highly curved reaction pathways accurately.

Numerical Stability¶

Using a Predictor–Corrector procedure offers several advantages.

Reduces numerical errors.
Prevents deviation from the Minimum Energy Path.
Produces smooth reaction pathways.
Improves convergence for long IRC calculations.
Accurately follows curved Potential Energy Surfaces.

These properties make the algorithm robust for a wide variety of chemical reactions.

Relationship with Previous Mathematics¶

The prediction stage uses concepts introduced during the Transition State Search, including gradients and Hessian information.

Transition State Search¶

📘 Transition State Search Mathematics

The correction stage resembles a constrained Geometry Optimization, where the molecular geometry is adjusted while remaining on the reaction pathway.

Geometry Optimization¶

📘 Geometry Optimization Mathematics

The Predictor–Corrector algorithm combines these ideas to generate successive points along the Intrinsic Reaction Coordinate.

Summary¶

The Predictor–Corrector Algorithm is the numerical engine that allows Gaussian to trace the Intrinsic Reaction Coordinate accurately. At each step, Gaussian first predicts the next point along the reaction pathway and then corrects the predicted geometry so that it lies on the Minimum Energy Path. Repeating this process generates a smooth and physically meaningful sequence of molecular structures connecting the transition state with the reactants and products.

Next Section¶

The next chapter explains Step Size and Hessian Updates, describing how Gaussian determines the distance between successive IRC points and updates the local curvature information as the reaction pathway is followed.