Skip to content

Orbital Optimization

In the previous chapter, we solved the Configuration Interaction (CI) problem and obtained the optimal CI coefficients for a fixed set of molecular orbitals.

However, these orbitals were inherited from the previous iteration (or from the initial RHF calculation). If the orbitals themselves are not optimal, then the CI coefficients cannot represent the best possible electronic wavefunction.

The defining feature of CASSCF is that it optimizes both the CI coefficients and the molecular orbitals simultaneously until they become self-consistent.


Why Isn't CI Alone Enough?

Suppose we begin with molecular orbitals obtained from an RHF calculation.

RHF Orbitals


Perform CI


Best CI Coefficients

Although the CI coefficients are optimal for these orbitals, there is no guarantee that the orbitals themselves are optimal.

Imagine trying to build a house using perfectly arranged bricks, but the foundation is tilted. No matter how carefully the bricks are arranged, the final structure cannot be truly optimal.

Similarly,

  • CI optimizes the electronic configurations,
  • but the orbitals providing those configurations may still require improvement.

The Two Variables of CASSCF

Unlike RHF, which optimizes only the molecular orbitals, CASSCF optimizes two independent quantities.

CASSCF Wavefunction

├── Molecular Orbitals
└── CI Coefficients

Both influence the total electronic energy.

Changing either one changes the wavefunction.


The Variational Principle

According to the variational principle,

\[ E = \frac{\langle\Psi|\hat H|\Psi\rangle} {\langle\Psi|\Psi\rangle} \]

the best wavefunction is the one that minimizes the electronic energy.

Since

\[ \Psi= \sum_I c_I\Phi_I, \]

the energy depends on

  • the CI coefficients \(c_I\),
  • and the molecular orbitals used to construct the CSFs.

Therefore,

both must be optimized.


How Are Orbitals Changed?

The molecular orbitals are not replaced completely after each iteration.

Instead,

they are rotated.

Suppose we have two orbitals,

Orbital i

Orbital j

CASSCF mixes them slightly,

New Orbital i

=

Old Orbital i

+

Small Portion of Orbital j

and

New Orbital j

=

Old Orbital j


Small Portion of Orbital i

Only a small rotation is applied during each iteration.


Orbital Rotation Matrix

Mathematically,

orbital rotations are represented by a unitary transformation,

\[ \boxed{ \mathbf C' = \mathbf C e^{\kappa} } \]

where

  • \(\mathbf C\) is the molecular orbital coefficient matrix,
  • \(\kappa\) is the orbital rotation matrix,
  • \(\mathbf C'\) is the updated orbital coefficient matrix.

Because the transformation is unitary,

the orbitals remain orthonormal throughout the optimization.


Which Orbitals Are Rotated?

Not every orbital is allowed to mix with every other orbital.

CASSCF classifies orbitals into three groups.

Core

────────────

Active

────────────

Virtual

Only certain rotations lower the electronic energy.

The most important rotations are

  • Core ↔ Active
  • Active ↔ Virtual
  • Core ↔ Virtual

Rotations within the same subspace do not change the energy and are therefore unnecessary.


The Self-Consistent Optimization Cycle

The complete optimization proceeds iteratively.

Initial RHF Orbitals


Generate CSFs


Build Hamiltonian


Solve CI


Obtain CI Coefficients


Rotate Orbitals


Generate New CSFs


Rebuild Hamiltonian


Solve CI Again


Repeat

Each iteration improves

  • the orbitals,
  • the CI coefficients,
  • and the total energy.

Energy Convergence

After every orbital rotation,

the total energy is evaluated.

Iteration 1

Energy = -XXX.XXXX


Iteration 2

Energy = Lower


Iteration 3

Energy = Lower


Iteration N

No Significant Change

When the energy no longer decreases,

the calculation has reached self-consistency.


Simultaneous Optimization

Unlike RHF,

which optimizes only one quantity,

CASSCF alternates between two optimizations.

Optimize CI Coefficients


Optimize Orbitals


Optimize CI Again


Optimize Orbitals Again


Repeat Until Converged

This alternating procedure is the defining feature of the method.


Why Is It Called "Self-Consistent Field"?

The term Self-Consistent Field (SCF) means that the wavefunction is repeatedly updated until every part of it is internally consistent.

For RHF,

the orbitals are optimized until the electron density no longer changes.

For CASSCF,

both the orbitals and the CI coefficients are optimized until neither changes significantly.

Thus,

Orbitals


CI


Orbitals


CI


Orbitals


Converged

At convergence,

both quantities are mutually consistent.


What Happens Inside GAMESS?

When a CASSCF calculation is started,

GAMESS repeatedly performs the following steps:

  1. Read the current molecular orbitals.
  2. Generate all Configuration State Functions (CSFs).
  3. Construct the CI Hamiltonian.
  4. Diagonalize the Hamiltonian to obtain CI coefficients.
  5. Compute the energy gradient with respect to orbital rotations.
  6. Rotate the molecular orbitals.
  7. Rebuild the Hamiltonian using the updated orbitals.
  8. Repeat until the energy and wavefunction satisfy the convergence criteria.

This cycle is controlled by the $MCSCF input group and continues until a fully optimized multiconfigurational wavefunction is obtained.


Why Is Orbital Optimization Important?

Without orbital optimization,

the CI calculation would simply describe electron correlation using a fixed set of orbitals.

By optimizing the orbitals,

CASSCF allows the molecular orbital basis itself to adapt to the multiconfigurational wavefunction.

This leads to

  • lower electronic energies,
  • improved excited-state descriptions,
  • balanced treatment of near-degenerate states,
  • reliable potential energy surfaces,
  • and accurate wavefunctions for strongly correlated systems.

Complete CASSCF Workflow

The entire CASSCF procedure can now be summarized as

Initial RHF Calculation
Choose Active Space
Generate Configuration State Functions
Construct Hamiltonian Matrix
Diagonalize Hamiltonian
Obtain CI Coefficients
Rotate Molecular Orbitals
Update Hamiltonian
Repeat Until Converged
Final CASSCF Wavefunction

This is the algorithm followed by most modern multiconfigurational quantum chemistry programs, including GAMESS.


Key Takeaways

  • CASSCF optimizes both the CI coefficients and the molecular orbitals.
  • Orbital optimization is achieved through small unitary rotations of the molecular orbitals.
  • After each orbital update, the CI problem is solved again using the new orbitals.
  • The process is repeated until the total electronic energy no longer changes.
  • This iterative optimization is the reason the method is called Complete Active Space Self-Consistent Field (CASSCF).