Skip to content

Complete Geometry Optimization Algorithm

In the previous chapters, we introduced the mathematical foundations of geometry optimization, including the Potential Energy Surface, energy gradients, Hessian matrices, Taylor series expansion, optimization algorithms, the Berny optimization method, and convergence criteria.

This chapter combines these individual concepts into a single algorithm that describes how Gaussian determines an optimized molecular geometry.

Although the underlying mathematics can be sophisticated, the overall procedure follows a logical sequence that is repeated until the molecular structure satisfies the convergence criteria.


Overview

Geometry optimization is an iterative energy minimization process.

Starting from an initial molecular geometry, Gaussian repeatedly

  • evaluates the electronic energy,
  • calculates the energy gradient,
  • estimates the local curvature of the Potential Energy Surface,
  • predicts an improved molecular geometry,
  • checks whether the optimization has converged.

The process continues until a stationary point is reached.


Complete Geometry Optimization Workflow

                 Initial Molecular Geometry
                 Read Cartesian Coordinates
              Construct Internal Coordinates
                 Generate Initial Hessian
              Perform SCF Electronic Structure
                     Calculation
               Obtain Electronic Energy
                Calculate Energy Gradient
                Update Hessian Estimate
          Construct Local Taylor Expansion
          of the Potential Energy Surface
      Berny / Rational Function Optimization
             Predict New Molecular Geometry
             Calculate New Electronic Energy
          Evaluate Convergence Criteria
                 ┌─────────┴─────────┐
                 │                   │
                No                  Yes
                 │                   │
                 ▼                   ▼
      Repeat Optimization      Optimized Geometry
           Cycle

Step 1 — Read the Initial Geometry

The optimization begins with the molecular structure supplied by the user.

The geometry may originate from

  • GaussView,
  • experimental crystal structures,
  • previous calculations,
  • molecular databases.

This structure serves only as the starting point for the optimization.


Step 2 — Generate Internal Coordinates

Rather than optimizing Cartesian coordinates directly, Gaussian converts the molecular structure into a set of chemically meaningful internal coordinates.

These include

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

Internal coordinates usually produce smoother optimization paths and faster convergence.


Step 3 — Generate an Initial Hessian

An initial estimate of the Hessian matrix is generated.

The Hessian describes the local curvature of the Potential Energy Surface.

Since calculating the exact Hessian is computationally expensive, Gaussian normally begins with an approximate Hessian.


Step 4 — Perform an SCF Calculation

Using the current molecular geometry, Gaussian performs a complete electronic structure calculation.

Depending on the selected method, this may involve

  • Hartree–Fock,
  • Density Functional Theory,
  • or another electronic structure method.

The result is the electronic energy corresponding to the current geometry.


Step 5 — Calculate the Energy Gradient

After the SCF calculation converges, Gaussian computes the energy gradient.

The gradient determines

  • the direction of steepest energy decrease,
  • the magnitude of the forces acting on the atoms.

The optimization algorithm uses this information to determine how the molecular geometry should be modified.


Step 6 — Update the Hessian

The Hessian is updated using the new gradient information.

Instead of recalculating every second derivative, Gaussian improves its previous estimate.

As the optimization progresses, the Hessian becomes an increasingly accurate representation of the local Potential Energy Surface.


Step 7 — Construct the Local Potential Energy Surface

Using

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

Gaussian constructs a local quadratic approximation of the Potential Energy Surface through the Taylor series expansion.

This local model predicts how the energy will change for small displacements of the atoms.


Step 8 — Determine the Optimization Step

The Berny optimization algorithm applies Rational Function Optimization (RFO) to determine the next molecular geometry.

The optimization step

  • follows the downhill direction,
  • accounts for the local curvature,
  • avoids excessively large atomic displacements,
  • maintains numerical stability.

This produces a new set of atomic coordinates.


Step 9 — Check Convergence

After updating the molecular geometry, Gaussian evaluates the convergence criteria.

The following quantities are examined:

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

If every criterion satisfies its threshold, the optimization terminates successfully.

Otherwise, the new geometry becomes the starting point for another optimization cycle.


Iterative Optimization Cycle

The complete optimization may be represented as

Current Geometry
SCF Calculation
Electronic Energy
Gradient
Hessian Update
Local Taylor Expansion
Berny Optimization
New Geometry
Convergence?
   No ──┴── Yes
 Repeat       Optimized Geometry

Each iteration moves the molecule closer to a stationary point on the Potential Energy Surface.


Mathematical Foundations

Every optimization cycle combines the concepts introduced throughout this section.

Mathematical Concept Role in Geometry Optimization
Potential Energy Surface Defines the energy landscape
Energy Gradient Determines the direction of atomic motion
Hessian Matrix Describes the local curvature
Taylor Series Approximates the local energy surface
Optimization Algorithm Determines the optimal displacement
Berny Method Efficiently updates the geometry
Convergence Criteria Determines when optimization is complete

Together, these components form the mathematical framework underlying every geometry optimization performed by Gaussian.


After Geometry Optimization

A successful geometry optimization provides the equilibrium molecular structure.

In most computational studies, the optimized geometry is subsequently used for

  • Frequency calculations,
  • Single Point Energy calculations using larger basis sets,
  • NMR calculations,
  • Excited-state calculations,
  • Intrinsic Reaction Coordinate (IRC) calculations,
  • Reaction energy calculations.

To confirm that the optimized structure corresponds to a true minimum on the Potential Energy Surface, a Frequency Calculation should always be performed.


Summary

Geometry optimization in Gaussian is an iterative procedure that combines quantum chemical calculations with numerical optimization techniques to locate stationary points on the Potential Energy Surface. Beginning with an initial molecular geometry, Gaussian repeatedly performs an SCF calculation, evaluates the energy gradient, updates the Hessian matrix, constructs a local quadratic approximation of the energy surface, and determines an improved molecular geometry using the Berny optimization algorithm. This cycle continues until the convergence criteria are satisfied, yielding the optimized molecular structure. The optimized geometry then serves as the starting point for further computational studies such as frequency analysis, thermochemical calculations, and excited-state investigations.


Conclusion

With this chapter, the mathematical treatment of Geometry Optimization is complete.

The next major topic in the Gaussian documentation is Frequency Analysis, where the optimized geometry is used to calculate molecular vibrational frequencies, confirm the nature of the stationary point, and obtain thermodynamic properties such as zero-point energy, entropy, enthalpy, and Gibbs free energy.