CoulombVertex

The CoulombVertex \(\Gamma^q_{rF}\) is a resolution-of-identiy decomposition of the electron repulsion integrals of the form

where \({\Gamma^\ast}^{pF}_s = \overline{\Gamma^s_{pF}}\) denotes the conjugate transpose. For periodic systems it is computed from the Fourier transform of the co-density of an orbital pair \({\psi^\ast}^q(\mathrm{r})\) and \(\psi_r(\mathrm{r})\), and the square-root of the Coulomb potential \(\sqrt{4\pi/\mathrm{G}^2}\), as specified below. The index order of \(\Gamma^q_{rF}\) in a Cc4s tensor is \((F,q,r)\).

The CoulombVertex needs to be provided by an interfaced electronic structure theory package and can be read into Cc4s using the Read algorithm.

- name: Read
  in:
    fileName: "CoulombVertex.yaml"
  out:
    tensor: CoulombVertex

For periodic systems the CoulombVertex \(\Gamma^q_{rF}\) is computed from the Fourier transform \(\tilde\Gamma^q_{rG}\) of the right vertex of the Coulomb interaction

where \(\mathbf{G}\) denotes the momentum vector associated to the momentum index \(G\) and where the indices \(q\) and \(r\) refer to the outgoing and incoming states at the right interaction vertex, respectively. Note, that the square root of the numerical integration weight \(w_G\) and of the Coulomb kernel \(4\pi/\mathbf{G}^2\) are contained in the above definition, such that the desired resolution-of-identity factorization of of the electron repulsion integrals holds

To reduce the size of the employed CoulombVertex while minimizing the error of the above factorization, the singular value decomposition of the Fourier transformed Coulomb interaction vertex \(\tilde\Gamma^q_{rG} = U^F_G \Sigma^F_F {W^\ast}^q_{rF}\) is considered. The CoulombVertex used by Cc4s is finally defined using only the \(N_\mathrm{F}\) largest singular values \(\Sigma^F_F\) and their respective left singular vectors \(U^F_G\)

The singular value index \(F\) is called AuxiliaryField index in Cc4s, as indicated by type field in the dimensions section of the CoulombVertex tensor specification. The coefficients \(U_F^G\) are needed by the FiniteSizeCorrection algorithm and they are contained in the object CoulombVertexSingularVectors. For more details, see (Hummel, Tsatsoulis, and Grüneis 2017).

An example CoulombVertex.yaml file is given below

version: 100
type: Tensor
scalarType: Complex64
dimensions:
- length:    356
  type: AuxiliaryField
- length:     96
  type: State
- length:     96
  type: State
elements:
  type: IeeeBinaryFile
unit: 0.1917011272153577       # = sqrt(Eh/eV)
metaData:
  halfGrid: 1