Skip to content

Self-Consistent Field (SCF) Algorithm

In the previous chapter, we transformed the Hartree–Fock equations into the Roothaan–Hall matrix equation

\[ \mathbf{FC}=\mathbf{SC}\boldsymbol{\varepsilon} \]

At first glance, this appears to be a standard matrix eigenvalue problem. However, there is one important complication:

The Fock matrix itself depends on the molecular orbitals that we are trying to determine.

This circular dependence prevents the equations from being solved directly. Instead, the Hartree–Fock equations must be solved iteratively until a self-consistent solution is obtained.

This iterative procedure is known as the Self-Consistent Field (SCF) method.

Why Is an Iterative Procedure Necessary?

The Fock matrix is

\[ \mathbf F = \mathbf H + \mathbf G(\mathbf P) \]

where

  • \(\mathbf H\) is the core Hamiltonian matrix,
  • \(\mathbf G(\mathbf P)\) represents the electron–electron interaction,
  • \(\mathbf P\) is the density matrix.

Notice that

\[ \mathbf G = \mathbf G(\mathbf P) \]

but the density matrix itself is computed from the molecular orbitals,

\[ \mathbf P \longrightarrow \mathbf C \]

while the molecular orbitals are obtained by solving

\[ \mathbf{FC} = \mathbf{SC}\boldsymbol{\varepsilon} \]

This creates a circular dependency:

Density Matrix
      │
      ▼
Fock Matrix
      │
      ▼
Solve Roothaan–Hall Equation
      │
      ▼
New Molecular Orbitals
      │
      ▼
New Density Matrix

The solution is to repeat this process until the density matrix no longer changes.

Step 1 – Initial Guess

Every SCF calculation begins with an initial estimate of the molecular orbitals.

Common choices include

  • Hückel guess
  • Core Hamiltonian guess
  • Superposition of Atomic Densities (SAD)
  • Orbitals from a previous calculation

In GAMESS, the initial guess is controlled using the $GUESS group.

For example,

$GUESS GUESS=HUCKEL $END

Although the initial guess is only approximate, it provides a starting point for the iterative procedure.

Step 2 – Construct the Density Matrix

Using the occupied molecular orbitals, the density matrix is constructed.

For a closed-shell RHF calculation,

\[ \boxed{ P_{\mu\nu} = 2 \sum_{i}^{occ} C_{\mu i}C_{\nu i} } \]

where

  • \(C_{\mu i}\) are the molecular orbital coefficients,
  • the factor of 2 accounts for paired electrons.

The density matrix describes how electrons are distributed throughout the molecule.

Step 3 – Build the Fock Matrix

The next step is to construct the Fock matrix.

Mathematically,

\[ \boxed{ \mathbf F = \mathbf H + \mathbf G(\mathbf P) } \]

where

  • \(\mathbf H\) is the core Hamiltonian,
  • \(\mathbf G\) contains Coulomb and exchange contributions.

Since the density matrix has changed,

the Fock matrix also changes.

Step 4 – Solve the Roothaan–Hall Equation

The updated Fock matrix is substituted into

\[ \boxed{ \mathbf{FC} = \mathbf{SC} \boldsymbol{\varepsilon} } \]

Diagonalizing this equation produces

  • new molecular orbital coefficients,
  • new orbital energies.

These updated orbitals are then used to construct a new density matrix.

Step 5 – Check for Convergence

The new density matrix is compared with the previous iteration.

If the change is sufficiently small,

the SCF calculation is considered converged.

Typical convergence criteria involve

  • total electronic energy,
  • density matrix elements,
  • RMS density change,
  • maximum density change.

Mathematically,

\[ \Delta P = P^{(n)} - P^{(n-1)} \]

If

\[ |\Delta P| < \text{threshold} \]

the calculation terminates successfully.

Otherwise,

another iteration begins.

The SCF Cycle

The complete SCF procedure can be summarized as

Initial Guess
      │
      ▼
Density Matrix
      │
      ▼
Build Fock Matrix
      │
      ▼
Solve Roothaan–Hall Equation
      │
      ▼
New Molecular Orbitals
      │
      ▼
New Density Matrix
      │
      ▼
Converged?
      │
 ┌────┴────┐
 │         │
No        Yes
 │         │
 ▼         ▼
Repeat   Finish

Each iteration gradually improves the molecular orbitals until they become self-consistent.

Convergence Acceleration

For difficult systems, simple iteration may converge very slowly or even diverge.

Modern quantum chemistry programs therefore employ convergence acceleration techniques.

The most widely used method is

DIIS (Direct Inversion in the Iterative Subspace).

Rather than using only the most recent Fock matrix, DIIS combines information from several previous iterations to predict an improved solution.

Other techniques include

  • damping,
  • level shifting,
  • Fermi smearing,
  • orbital mixing.

In GAMESS, DIIS is enabled automatically for most RHF calculations.

When Does SCF Fail?

Although RHF is generally robust, convergence problems may arise for

  • stretched chemical bonds,
  • transition metals,
  • near-degenerate orbitals,
  • poor initial guesses,
  • highly charged systems.

Common remedies include

  • improving the initial guess,
  • increasing the maximum number of iterations,
  • using level shifting,
  • changing the basis set,
  • switching to a different SCF procedure.

Computational Cost

The most computationally demanding step of the SCF cycle is constructing the Fock matrix.

For conventional Hartree–Fock calculations,

the computational cost scales approximately as

\[ O(N^4) \]

where

\(N\)

is the number of basis functions.

This scaling motivates the development of more efficient algorithms for large molecular systems.

Key Takeaways

  • The Fock matrix depends on the electron density.
  • The electron density depends on the molecular orbitals.
  • The molecular orbitals are obtained by diagonalizing the Fock matrix.
  • This circular dependence requires an iterative solution.
  • The SCF algorithm repeats until the electron density becomes self-consistent.

Looking Ahead

The previous chapters developed the mathematical framework underlying the Hartree–Fock method.

In the final chapter, we will apply these concepts to a real RHF calculation on the water molecule, connecting the mathematical equations with the actual input and output generated by GAMESS.