Skip to content

Berny Transition State Optimization

Gaussian performs Transition State searches using the Berny Optimization Algorithm, an iterative optimization procedure specifically designed for molecular geometry optimization.

The Berny algorithm is responsible for controlling the overall optimization process. It repeatedly evaluates the molecular energy, computes gradients, updates the Hessian matrix, determines an improved molecular geometry, and checks for convergence.

For Transition State calculations, Berny optimization is combined with Rational Function Optimization (RFO), allowing Gaussian to locate a first-order saddle point rather than a local minimum.


What is Berny Optimization?

The Berny algorithm is an iterative optimization method that searches for stationary points on the Potential Energy Surface.

Rather than taking arbitrary geometry steps, it uses information from

  • the electronic energy,
  • the energy gradient,
  • the Hessian matrix,

to determine the next molecular geometry.

The optimization continues until all convergence criteria are satisfied.


Overall Optimization Cycle

Each optimization cycle consists of the following steps.

Current Geometry
SCF Calculation
Electronic Energy
Energy Gradient
Hessian Matrix
RFO Optimization Step
New Geometry
Convergence Test
   ┌────┴────┐
   │         │
  No        Yes
   │         │
   ▼         ▼
Repeat    Transition State

This cycle is repeated until the transition state has been located.


Internal Coordinates

Instead of optimizing directly in Cartesian coordinates,

Gaussian usually transforms the molecular geometry into redundant internal coordinates.

These include quantities such as

  • bond lengths,
  • bond angles,
  • dihedral angles.

For example,

Cartesian Coordinates

(x₁,y₁,z₁)
(x₂,y₂,z₂)
...


Internal Coordinates

Bond Lengths
Bond Angles
Dihedral Angles

Internal coordinates provide a more natural description of molecular motion and generally improve the stability and efficiency of the optimization.


Geometry Update

After each SCF calculation,

Gaussian evaluates

  • the electronic energy,
  • the nuclear gradient,
  • the Hessian matrix.

The Berny algorithm then requests an optimization step from the Rational Function Optimization (RFO) procedure.

The molecular geometry is updated,

and the process repeats.

Unlike ordinary geometry optimization,

the update is chosen so that

  • one direction moves toward higher energy,
  • all remaining directions move toward lower energy.

Hessian Updates

The Hessian matrix is expensive to calculate exactly at every optimization step.

Therefore,

Gaussian usually computes the Hessian only once (or occasionally),

then updates it during the optimization using approximate update formulas.

Initial Hessian
Geometry Update
Approximate Hessian Update
Next Optimization Step

These updates significantly reduce the computational cost while maintaining accurate optimization directions.


Convergence Criteria

After every optimization step,

Gaussian evaluates whether the geometry has converged.

The convergence criteria include

  • Maximum Force
  • RMS Force
  • Maximum Displacement
  • RMS Displacement

Only when all criteria satisfy their thresholds is the optimization considered complete.

This is the same convergence test used in standard Geometry Optimization.


Difference from Ordinary Geometry Optimization

The Berny algorithm is used for both Geometry Optimization and Transition State Search.

The major difference lies in the optimization objective.

Geometry Optimization Transition State Search
Locate a local minimum Locate a first-order saddle point
All directions minimized One direction maximized, remaining directions minimized
No imaginary frequencies Exactly one imaginary frequency

Thus, the overall optimization framework remains the same, but the mathematical target is different.


Relationship with RFO

The Berny algorithm and Rational Function Optimization work together.

Their responsibilities can be summarized as

Berny Optimization

• Controls optimization
• Stores geometry
• Updates Hessian
• Checks convergence


Rational Function Optimization

• Determines optimization step
• Handles negative curvature
• Moves toward saddle point

Berny manages the optimization,

while RFO determines the best direction for each optimization step.


Why Berny Optimization?

The Berny algorithm is widely used because it

  • efficiently handles large molecular systems,
  • converges reliably for most reactions,
  • works naturally with internal coordinates,
  • minimizes the number of expensive Hessian calculations,
  • integrates seamlessly with RFO.

These properties make it one of the most successful geometry optimization algorithms in computational chemistry.


Relation to Previous Mathematics

The Berny algorithm is also used during ordinary Geometry Optimization. The mathematical concepts of gradients, convergence criteria, and coordinate optimization were introduced previously.

Geometry Optimization

📘 Geometry Optimization Mathematics

The present chapter extends those ideas to the optimization of first-order saddle points.


Summary

The Berny Optimization Algorithm controls the entire Transition State search performed by Gaussian. During each optimization cycle, it evaluates the electronic energy, computes the gradient, updates the Hessian matrix, requests an optimization step from the Rational Function Optimization algorithm, and tests for convergence. By combining Berny optimization with RFO, Gaussian efficiently locates first-order saddle points while minimizing computational cost through the use of internal coordinates and approximate Hessian updates.


Next Section

The next chapter discusses Initial Hessian and CalcFC, explaining why an accurate initial Hessian is particularly important for Transition State searches and how the CalcFC keyword improves convergence.