bsemat.h5 file format

Group: /mf_header

Content: Header from the WFN file.

Group: /eps_header

Content: Header from the epsmat file.

Group: /bse_header

Content: Everything but the kernel matrix elements.

Dataset: /bse_header/versionnumber

Type: integer

Rank: 0

Value: Version number of this particular file.

Dataset: /bse_header/flavor

Type: integer

Rank: 0

Value: 1 for Real, 2 for CPLX.

Group: /bse_header/params

Content: General parameters.

Dataset: /bse_header/params/screening

Type: integer

Rank: 0

Value: What type of screening was used.

Dataset: /bse_header/params/icutv

Type: integer

Rank: 0

Value: Truncation flag. See Epsilon/inread.f90 for the meaning of each icutv.

Dataset: /bse_header/params/ecuts

Type: double

Rank: 0

Value: Cutoff for screened Coulomb interaction.

Dataset: /bse_header/params/ecutg

Type: double

Rank: 0

Value: Cutoff for the bare Coulomb interaction.

Dataset: /bse_header/params/efermi

Type: double

Rank: 0

Value: Fermi level determined by the code, in eV.

Dataset: /bse_header/params/theory

Type: integer

Rank: 0

Value: 0 for BSE, 1 for TDDFT.

Dataset: /bse_header/params/nblocks

Type: integer

Rank: 0

Value: How many transitions blocks are there in the kernel matrix? Possible values are: - 1 for restricted TDA kernel: vc -> vc - 2 for restricted non-TDA kernel: {vc,cv} -> {vc,cv} [not implemented] - 4 for extended kernel: {n1,n2} -> {n1,n2}

Dataset: /bse_header/params/storage

Type: integer

Rank: 0

Value: How we store the kernel matrix. Possible values are: 0 for no symmetries, i.e., {n1,n2} -> {n1,n2}

Dataset: /bse_header/params/nmat

Type: integer

Rank: 0

Value: Number of matrices stored in /mats

Dataset: /bse_header/params/energy_loss

Type: integer

Rank: 0

Value: 1 if this an energy-loss calculation. Otherwise, 0.

Group: /bse_header/bands

Content: Bands-related datasets.

Dataset: /bse_header/bands/nvb

Type: integer

Rank: 0

Value: Number of valence bands included in the kernel computation.

Dataset: /bse_header/bands/ncb

Type: integer

Rank: 0

Value: Number of conduction bands included in the kernel computation.

Dataset: /bse_header/bands/n1b

Type: integer

Rank: 0

Value: Number of bands of first kind we loop over in the kernel matrix. This is nvb for restricted kernels, and nvb+ncb for extended kernels.

Dataset: /bse_header/bands/n2b

Type: integer

Rank: 0

Value: Number of bands of second kind we loop over in the kernel matrix. This is ncb for restricted kernels, and nvb+ncb for extended kernels.

Dataset: /bse_header/bands/ns

Type: integer

Rank: 0

Value: Number of spins. This is 1 unless this is a spin-polarized calculation with a single spinor component, on which case this is 2.

Dataset: /bse_header/bands/nspinor

Type: integer

Rank: 0

Value: Number of spinor components. Either 1 or 2.

Group: /bse_header/kpoints

Content: K-points-related datasets.

Dataset: /bse_header/kpoints/nk

Type: integer

Rank: 0

Value: Number of k-points in the full zone.

Dataset: /bse_header/kpoints/kpts

Rank: 2

Dims(1): 3 #(the three crystal coordinates)

Dims(2): nk

Value: K-points.

Dataset: /bse_header/kpoints/kgrid

Type: integer

Rank: 1

Dims(1): 3

Value: K-grid used in epsilon calculation.

Dataset: /bse_header/kpoints/qflag

Type: integer

Rank: 0

Value: 0 for finite Q calculation with arb. Q (deprecated) 1 for Q=0 calculation (default) 2 for Q commensuare with WFN_co (under development)

Dataset: /bse_header/kpoints/center_mass_q

Type: double

Rank: 1

Dims(1): 3

Value: Center-of-mass q-shift for finite Q`s calculation.

Dataset: /bse_header/kpoints/patched_sampling

Type: logical

Rank: 0

Value: Was this bsemat calculated on a patch?

Group: /mats

Content: Matrix-elements-related datasets.

Dataset: /mats/head, wing, body, exchange, fxc

Type: double

Rank: 6

Dims(1): flavor

Dims(2:3): n1b

Dims(4:5): n2b

Dims(6:7): nk*ns

Value: Kernel matrix elements times V/8*Pi * Ry. The indices are stored as follows:

• For restricted kernels (default) n1b = nv and n2b = nc. The valence bands (dims 2 and 3) are counted downwards from the Fermi level, and the conduction bands (dims 4 and 5) are counted upwards from the Fermi level. So, restricted kernels separate valence-band from conduction-band indices. Without spin:
  $K^d(v,v',c,c',k,k') =$ $-\int d^3r d^3r' \; \psi_{ck}(r)^* \psi_{c'k'}(r) W(r,r') \psi_{vk}(r') \psi_{v'k'}(r')^*$
  $K^x(v,v',c,c',k,k') =$ $\int d^3r d^3r' \; \psi_{ck}(r)^* \psi_{vk}(r) v(r-r') \psi_{c'k'}(r') \psi_{v'k'}(r')^*$

• For extended kernels, n1b = n2b = nv + nc, so all bands (dims 2 to 5) are treated on the same footing. The first nv indices refer to valence states (counted downwards from the Fermi level), and the following nc indices refer to conduction states (counted upwards from the Fermi level). Without spin:
  $K^d(n,n',m,m',k,k') =$ $- \int d^3r d^3r' \; \psi_{mk}(r)^* \psi_{m'k'}(r) W(r,r') \psi_{nk}(r') \psi_{n'k'}(r')^*$
  $K^x(v,v',c,c',k,k') =$ $\int d^3r d^3r' \; \psi_{mk}(r)^* \psi_{nk}(r) v(r-r') \psi_{m'k'}(r') \psi_{n'k'}(r')^*$