I do not understand the distinction between configuration state functions and Slater determinants. Is not every Slater determinant a CSF ?

Please explain the difference and the use/need of configuration state functions.


1 Answer 1


Slater determinants are not eigenfunctions of the $\hat S^2$ operator, but CSFs are.

The Hamiltonian commutes with the operators for total and projected spin \begin{align} [\hat H, \hat S^2] &= 0 \\ [\hat H, \hat S_z] &= 0 \end{align} Therefore, a set of common eigenfunctions to all three operators exists. As Slater determinants are eigenfunctions of $\hat S_z$, but not $\hat S^2$, using them as a basis for the electronic wave function (=eigenstate of $\hat H$), will not guarantee the (approximate) solution to be an eigenfunction of $\hat S^2$. The spin multiplicity of the found solution may not be a pure Singlet (or Doublet, Triplet, etc.).

As an example, consider a 2 electron system, with both electrons occupying different spatial orbitals. The situation with same spin for both electrons can be represented as the Slater determinant $|\alpha\alpha\rangle=\frac{1}{\sqrt{2}}[\alpha(1)\beta(2)-\alpha(2)\beta(1)]$. The spatial part is omitted here, to keep the notation short. The corresponding eigenvalue equations are then \begin{align} \hat S_z |\alpha\alpha\rangle &= 1 |\alpha\alpha\rangle \\ \hat S^2 |\alpha\alpha\rangle &= 2 |\alpha\alpha\rangle \end{align} Thus we have $S=1$, from $S(S+1)=2$, and the spin multiplicity is $2S+1=3$, i.e. a Triplet. Here the Slater determinant directly corresponds to a CSF.

For the case of opposite spin, we have \begin{align} \hat S_z |\alpha\beta\rangle &= 0 |\alpha\beta\rangle \\ \hat S^2 |\alpha\beta\rangle &= |\alpha\beta\rangle + |\beta\alpha\rangle \end{align} and therefore $|\alpha\beta\rangle$ is not an eigenfunction of $\hat S^2$. The expectation value for total spin would yield $\langle\hat S^2\rangle=1$, which is neither a Singlet nor a Triplet (not even a Doublet where $S(S+1)=0.75$).

This can be fixed by taking suitable linear combinations of Slater determinants. In this example we have 2 options: \begin{align} |^1\Psi\rangle &= \frac{1}{\sqrt{2}} \left( |\alpha\beta\rangle - |\beta\alpha\rangle \right) \\ |^3\Psi\rangle &= \frac{1}{\sqrt{2}} \left( |\alpha\beta\rangle + |\beta\alpha\rangle \right) \end{align} This yields the eigenvalue equations \begin{align} \hat S^2 |^1\Psi\rangle &= 0 |^1\Psi\rangle \\ \hat S^2 |^3\Psi\rangle &= 2 |^3\Psi\rangle \end{align} which correspond to a Singlet and a Triplet state respectively.

Overall, this yields 1 Singlet component $|^1\Psi\rangle = \frac{1}{\sqrt{2}} \left( |\alpha\beta\rangle - |\beta\alpha\rangle \right)$, and the 3 Triplet components $|^3\Psi\rangle = \frac{1}{\sqrt{2}} \left( |\alpha\beta\rangle + |\beta\alpha\rangle \right)$, $|\alpha\alpha\rangle$ and $|\beta\beta\rangle$.

For the arithmetic on how to apply the $\hat S_z$ and $\hat S^2$ operators to multi-electron systems, see Chapter 2.5 in Modern Quantum Chemistry by A. Szabo and N. Ostlund.

  • $\begingroup$ The kets are shorthand for determinants i assume ? So |ab> = |a(1)b(2)> - |a(2)b(1)>. And are alpha and beta spinorbitals or do you assume some factorization of spin and spatial part ? Sorry for the poor format of my comment but how can do line breaks in comments and can i do mathjax in the comments ? $\endgroup$
    – Hans Wurst
    Apr 1, 2019 at 16:19
  • 1
    $\begingroup$ @HansWurst Mathjax can be used in comments, e.g $|ab\rangle$ yields $|ab\rangle$. Linebreaks are unavailable, however you can do display math, e.g. $$|ab\rangle$$ yields $$|ab\rangle$$ $\endgroup$
    – Tyberius
    Apr 1, 2019 at 16:32
  • 1
    $\begingroup$ @HansWurst yes, the ket is a Slater determinant here. $\alpha$ and $\beta$ refers to the spin of one orbital each, so spin orbitals. The wave function can be factorized into spatial and spin part, however I omitted the spatial part. $\endgroup$
    – Feodoran
    Apr 1, 2019 at 16:39
  • 1
    $\begingroup$ Do the spatial orbitals alone also form a determinant or do they factorize as a simple product ? Like $$ |^1 \Psi \rangle = \phi_1(r1) \phi_1(r2) \cdot (|\alpha \beta \rangle - |\beta \alpha \rangle) $$ or does it take some other form ? $\endgroup$
    – Hans Wurst
    Apr 1, 2019 at 17:09
  • $\begingroup$ @HansWurst its always a Slater determinant, otherwise you get problem with the Pauli principle. $\endgroup$
    – Feodoran
    Apr 1, 2019 at 17:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.