Structure of HDF5 file

Basics

All input and output data are stored in HDF5 format. The file names are specified by parameters

[chiq_common]
input = "dmft_bse.h5"
output = "dmft_bse.out.h5"

One may specify the same file for both input and output.

Note

One can print the stored data using h5ls command as shown below. However, it is recommended to use the class h5BSE to save/load data. See Access to HDF5 file for details.

dmft_bse.h5

One can print the structure of the stored data using h5ls command.

$ h5ls dmft_bse.h5
bse                      Group

All data are stored in the bse group. One can print subgroups in bse group by

$ h5ls dmft_bse.h5/bse
info                     Group
input                    Group

info subgroup contains basic information

$ h5ls dmft_bse.h5/bse/info
beta                     Dataset {SCALAR}
block_name               Group
inner_name               Group

input subgroup contains input data necessary for BSE calculations

$ h5ls dmft_bse.h5/bse/input
X0_loc                   Group
X0_q                     Group
X_loc                    Group
chi0_loc_in              Group
chi_loc_in               Group
gamma0                   Group

dmft_bse.out.h5

The output file dmft_bse.out.h5 contain output subgroup as follows.

$ h5ls dmft_bse.out.h5/bse
info                     Group
output                   Group

$ h5ls dmft_bse.out.h5/bse/output
I_q                      Group
chi0_loc                 Group
chi0_q                   Group
chi_loc                  Group
chi_q                    Group
chi_q_rpa                Group
chi_q_rrpa               Group

Structure of data

The structure of the data is classified into the following three types.

dict A

{(block1, block2): ndarray.complex[N_in1, N_in2, N_w, N_w] }

dict B

{(block1, block2): ndarray.complex[N_in1, N_in2, N_w] }

dict C

{(block1, block2): ndarray.complex[N_in1, N_in2] }

  • block1, block2: block index

  • N_in1: number of inner_indices in block1

  • N_in2: number of inner_indices in block2

  • N_w: number of fermionic Matsubara frequencies

dict A is used for two-particle Green’s function such as \([X_\text{loc}(i\omega_n, i\omega_{n'})]_{12,34}\), where \(1 \equiv (a_1, \sigma_1, m_1)\). The bosonic frequency has been omitted. This quantity is reformatted as \(X_\text{loc}(a_1 \sigma_1 \sigma_2, m_1 m_2, n; a_3 \sigma_3 \sigma_4, m_3 m_4, n')\). Here, we used \(a_1=a_2\) and \(a_3=a_4\). The indices before and after the semicolon represent row and column, respectively. Then, this is expressed as a block matrix as X_loc[(block1, block2)][i1, i2, w1, w2], where \(\mathtt{block1} = (a_1 \sigma_1 \sigma_2)\), \(\mathtt{block2} = (a_3 \sigma_3 \sigma_4)\), \(\mathtt{i1} = (m_1 m_2)\), \(\mathtt{i2} = (m_3 m_4)\), \(\mathtt{w1}=n\). and \(\mathtt{w2}=n'\).

dict B is used for bare two-particle Green’s function \([X_\text{0,loc}(i\omega_n)]_{12,34}\). The corresponding block matrix is given by X0_loc[(block1, block2)][i1, i2, w], where \(\mathtt{w}=n\).

dict C represents the structure of the susceptibility such as \([\chi(\boldsymbol{q})]_{12,34}\). There is no fermionic Matsubara frequency. Therefore, the block matrix is given by chi_q[(block1, block2)][i1, i2].

List of Data

info subgroup

The table below shows the data in info subgroup.

type

dtype

beta

float

block_name

list[string]

inner_name

list[string]

input subgroup

The table below shows the data in input subgroup.

The columns from ‘bse’ to ‘scl’ indicate whether the data is referred to by chiq_main.py and chiq_post.py when the corresponding approximation scheme (selected by ‘type’ parameter) is specified. The last two columns indicate data access by calc_Iq.py and calc_Iq_scl.py.

type

w

q

dtype

‘bse’

‘chi0’

‘rpa’

‘rrpa’

‘scl’

calc_Iq

calc_Iq_scl

X_loc

✓

–

dict A

◇

–

–

(△*1)

(△*1)

(△*1)

(△*1)

X0_loc

✓

–

dict B

◇

◇

–

–

◇

–

◇

X0_q

✓

✓

dict B

◇

◇

–

–

◇

–

◇

chi_loc_in

✓

–

dict C

△*1

–

–

△*1

△*1

△*1

△*1

chi0_loc_in

✓

–

dict C

△*2

△*2

–

–

△*2

–

–

gamma0

–

–

dict C

–

–

◇

–

–

–

–

Phi

✓

–

dict B

–

–

–

–

◇

–

–

  • ✓: Indicates required elements to get/save the data by h5BSE class (see Access to HDF5 file). For example, X_loc can be accessed with key=('X_loc', w), whereas X0_q is accessed with key=('X0_q', w, q).

  • ◇: Input (mandatory)

  • △*1: Input (recommended). Generated from X_loc if data is not provided.

  • (△*1): Used if chi_loc_in is not provided.

  • △*2: Input (recommended). Generated from X0_loc if data is not provided.

‘output’ subgroup

The table below shows the data in output subgroup and which mode output them.

type

w

q

dtype

‘bse’

‘chi0’

‘rpa’

‘rrpa’

‘scl’

calc_Iq

calc_Iq_scl

chi0_loc

✓

–

dict C

◆

◆

–

◇*3

◆

–

–

chi0_q

✓

✓

dict C

◆

◆

◇*3

◇*3

◆

–

–

chi_loc

✓

–

dict C

◆

–

–

◆

◆

–

◆

chi_q

✓

✓

dict C

◆

–

–

–

–

◇*4

–

chi_q_rpa

✓

✓

dict C

–

–

◆

–

–

–

–

chi_q_rrpa

✓

✓

dict C

–

–

–

◆

–

–

–

chi_q_scl

✓

✓

dict C

–

–

–

–

◆

–

◆

I_q

✓

✓

dict C

◆

–

–

–

–

◆

–

I_r

✓

✓*6

dict C

◆*5

–

–

–

–

–

–

I_q_scl

✓

✓

dict C

–

–

–

–

◆

–

◆

I_r_scl

✓

✓*6

dict C

–

–

–

–

◆*5

–

–

  • ◆: Output

  • ◇*3: Used as input. chiq_main.py with mode=’chi0’ should be run in advance.

  • ◇*4: Used as input. chiq_main.py with mode=’bse’ should be run in advance.

  • ◆*5: Output by chiq_fft.

  • ✓*6: q is used for the real space coordinate r.