Changelog

Unstable

Changes on the unstable branch that are not part of a release yet. Rename this section to the version number when the next release is cut.

Note that calculations using dc_orb_shift on orbitals that are degenerate in DFT give different results than with 4.0.0, since the orbital dependent part of the shift was previously averaged out of the impurity problem, see the fix below.

General

• add the sigma_embedding argument to plot_correlated_bands.get_dmft_bands, giving the tight-binding orbitals each correlated shell occupies, for models that do not follow the wannier90 convention of the correlated orbitals coming first

test

• svo_cthyb_basic_tf now checks that the hybridization function passed to the solver is negative, which catches the dc_orb_shift bug below. The test only checked that the run completes, and had been passing intermittently with an average sign of about 1e-2 since the input was added in #74

fix

• dc_orb_shift no longer gets averaged out by the orbital symmetrization. The degenerate solver blocks are found from the DFT density matrix before the shift is applied, so orbitals that are shifted differently were still symmetrized: G0 lost the orbital dependent part of the shift while the impurity levels kept it. With delta_interface = true this put spikes of the wrong sign (of order 10 for shifts of 0.1 eV) into Delta(tau) at tau=0 and beta and gave cthyb an average sign of about 1e-2; with delta_interface = false the impurity problem silently did not see the orbital dependent part of the shift. Blocks now stay degenerate only if their orbitals are shifted by the same amounts, so up and down remain degenerate in a paramagnetic calculation

• plot_correlated_bands: proj_on_orb now returns the orbital-resolved spectral function, i.e. the masked diagonal of the lattice Green function in the Wannier basis, which is exact. It previously weighted the band-resolved spectral function with the orbital character |<orb|band>|^2, which drops the off-diagonal band components of the Green function and is only exact for an orbital independent self-energy; for an orbital dependent one the weight was misdistributed between orbitals by up to 66% of the peak of A(k,w) in a SrVO3 test. The trace was and is unaffected. Pass band_basis=True to get the orbital-character weighting back, and note that the tight-binding bands returned for fat-band plots are unchanged

• plot_correlated_bands no longer rejects DMFT archives converted with bloch_basis=True. The self-energy is written into the Wannier basis of seed_hr.dat directly instead of through SumkDFT.upfold, which embeds into the Kohn-Sham band basis and is therefore k-dependent in Bloch basis. The embedding is a k-independent index operation, so A(k,w) is unchanged no matter how strongly proj_mat varies along k, and results in Wannier basis are bit-identical to before

• plot_correlated_bands warns instead of aborting where the result may not be what is expected: band disentanglement (n_bands > n_orb), where A(k,w) is the spectral function of the Wannier model rather than of the DMFT lattice problem and the chemical potential was converged for the electron count of the full outer window; an archive from a non-wannier90 converter, where the correlated orbitals need not come first; and a mismatch between the crystal-field eigenvalues of the archive and of the tight-binding H(R=0), which catches a grossly wrong orbital mapping also in Wannier basis. Being eigenvalue based, that last check cannot see a wrong order within a shell, nor a wrong set of orbitals when both the selected and the correct block are internally degenerate. add_spin=True together with a self-energy read from the archive is refused rather than guessed, since the correlated shell then appears in both spin blocks

Version 4.0.0

solid_dmft version 4.0.0 is a major release, compatible with TRIQS 4.0, and updated to the latest app4triqs skeleton. This release ports solid_dmft to the TRIQS 4.0 API and adds support for new HPC platforms, alongside several bug fixes.

We thank all contributors: Jennifer Coulter, Thomas Hahn, Alexander Hampel, Henri Menke, Anthony Ramirez, Dylan Simon, Nils Wentzell, and Alberto Carta

Find below an itemized list of changes in this release.

General

• port to the TRIQS 4.0 API: run the port_to_triqs4 script and replace triqs.gf imports with triqs.gfs

• use the modern mesh constructor keyword arguments (statistic, n_iw, n_tau)

• add SLURM support for the OLCF CADES Baseline cluster (walltime, hostfile path, and hostname configuration)

• allow the [dft] section dft_exec to be a TOML table with per-executable overrides for QE (keys path, pw, pw2wan, bands, proj, win_pp, win), and validate unknown keys

• expand environment variables (e.g. {HOME}, {SLURM_NTASKS}) in dft_exec commands at run time, in addition to {number_cores}

fix

• fix a startup failure when correlated shells of different size were not unique (e.g. two equivalent Ni 3d and two equivalent O 2p shells)

• adapt to the updated TRIQS Hartree-Fock solver interface

doc

• update the Ce2O3 QE CSC tutorial to use the h_int_basis=qe tag and the natural Wannier90 f-orbital order

build

• build triqs_dftkit before dft_tools in the CI docker images and adapt the Dockerfiles for dftkit compatibility

• bump GCC to 14 and LLVM to 20 for TRIQS unstable compatibility

• skip TPRF in the openmpi docker image until it is stabilized for 4.0.0

• fix the source tarball name in the foss EasyBuild file

• bump the publish-to-pypi CI image tag to 4.0.x

• install sphinx_autorun for the Jenkins doc build

• bump actions/checkout from 6 to 7

Version 3.3.5

solid_dmft version 3.3.5 is a minor patch release that fixes several bugs and adds small features:

fix

• afm mapping with hartree solver was missing copy of Sigma_Refreq

• afm calc with hartree solver was missing copy of DC_energy

• fix ctseg afm calculations

• reading DFT energy and iteration count from vaspout.h5 was broken

• numpy 2.x compatibility: explicit int cast for single-element arrays in observables

• explicit float cast in eval_U_cRPA_Vasp.py to remove spurious complex warning

• invalid escape sequence warnings in eval_U_cRPA_Vasp.py

feat

• enable averaging of density correction by default for magnetic CSC calculations

• add mpi_exe config flag in [dft] section and update to latest VASP h5 interface

• add version printouts to stdout at the start of a calculation

doc

• update NiO tutorial for ctseg afm calculations

build

• use ctseg 3.3.x instead of unstable branch for 3.3.x docker image

• fix docker OpenMPI build by removing OMPI_MCA_btl_vader_single_copy_mechanism

• update GitHub Actions dependencies (docker actions, artifact upload/download actions)

We thank all contributors: Alexander Hampel

Version 3.3.4

solid_dmft version 3.3.4 is a minor patch release that fixes several bugs and improves documentation:

fix

• afm mapping did not work with G and Sigma moments and occupations attributes of cthyb

• make n_orb initialization for self-energy safer

• add check for n_iter_dmft not None and =>0

• fix deprecated ast features

doc

• change name of config.toml to dmft_config.toml in NNO afm tutorial

test

• add missing test for afm order and cthyb for better test coverage

build

• update GitHub Actions dependencies (actions/checkout from 5 to 6)

• merge dev app4triqs updates

We thank all contributors: Alexander Hampel

solid_dmft version 3.3.3 is a minor patch release that fixes a few small bugs:

fix

• missing imports in _ctseg_postprocessing

• magnetic calc enforced if h_field set

• hartree interface postprocessing stored unsymmetrized Gfimp

• fix tail fitting for ctseg

• add LCO ctseg test

build

• add ctseg to docker images for testing

We thank all contributors: jmmshn, Thomas Hahn, and Alexander Hampel

Version 3.3.2

solid_dmft version 3.3.2 is a minor patch release that fixes a few small bugs and improves documentation:

doc

• update Ce2O3 tutorials input file

fix

• wrong error message for tail treatment in cthyb

• pcb projection on orb feat for Fermi surfaces in TB only mode was not giving correct results

• cthyb measure_chi was not passed into solve params since 3.3.x

• small bug in SOC calculation

We thank all contributors: Thomas Hahn, Nils Wentzell, Ina Park, Alexander Hampel

Version 3.3.1

solid_dmft version 3.3.1 is a minor release that fixes bugs in the new triqs/ctseg interface and improves the interface to GW:

fix

• small bug in dc initialization

• dynamic U tensor corrections

• initial self-energy from dc for magnetic calculation

• align ctseg interface with triqs/ctseg release

• ctseg hartree_shift rename in triqs

feat

• add print of avg order for ctseg

• use symmetric meshes for DLR

• fix delta_interface in ctseg

build

• add missing default.toml to pypi packaging

We thank all contributors: Alexander Hampel, Nils Wentzell, Chia-Nan Yeh

Version 3.3.0

solid_dmft 3.3.0 is a major release, compatible with TRIQS 3.3, updated to the latest app4triqs skeleton, and bringing major changes to the code:

• the input parser is switched to a general toml parser, i.e. strings have to be passed in quotes, boolean parameters given without capitalization, and lists passed with brackets. Check below separate section for detailed changes.

• the new input parser allows to define for each impurity problem a different solver if wanted, i.e. d-shell with cthyb and p-shell with Hartree-Fock. See new text NiO-cthyb

• docker images are automatically build on each push for all major releases to ghcr.io

• switch from old ctseg to new ctseg_j solver

• allow CRM Dyson solver for both cthyb and ctseg to obtain Sigma_imp from G_tau: “crm_dyson_solver=true” and dlr_wmax and dlr_eps (see https://triqs.github.io/triqs/unstable/documentation/python_api/triqs.gfs.dlr_crm_dyson_solver.html#module-triqs.gfs.dlr_crm_dyson_solver for details)

• add new DC schemes ‘crpa_static’, ‘crpa_static_qp’, ‘crpa_dynamic’

• use cRPA calculated Uijkl as interaction via ‘crpa’, ‘crpa_density_density’, ‘dyn_density_density’, ‘dyn_full’ hint types

• read interaction tensor from AIMBES h5

• new experimental gw_embedding module. See gw_embedding/gw_flow.py for details allowing to run solid_dmft on top of AIMBES

• allow to use Pade for AC in post-processing

We thank all contributors: Sophie Beck, Thomas Hahn, Alexander Hampel, Henri Menke, Maximilian Merkel, and Nils Wentzell

Find below an itemized list of changes in this release.

General

• merge dev GW embedding (includes other fixes as well) (#78)

• pass gw params to all methods

• multiple solvers and toml input parser (#74)

• added toml to docker images

• Restore Python 3.8 compatibility for dictionary merge (#63)

• Allow mathematical expression to be passed for random_seed (#61)

• allow PCB to read from TRIQS TB object

• respack fit slater for p shell

• add Pade Sigma analytic continuation and refine tests

• add simple_intra interaction, for intro orbital only interaction

• add dc_orb_shift param to allow orbital dependent shift in impurity levels

• allow 0.0 mixing to perform stat sampling

• switch all pytest to unit tests

new toml input parser

• The following input parameters can now be a list per impurity:

  • general_params: U, J, U_prime, ratio_F4_F2, h_int_type, enforce_off_diag, dc_type

  • advanced_params: dc_U, dc_J, dc_fixed_occ, map_solver_struct, pick_solver_struct, mapped_solver_struct_degeneracies

• Multiple solvers can be used, which only solve the impurity problems specified in idx_impurities

  • general parameter solver_type moved to solver section and renamed to type

  • general parameter n_l moved to solver section

  • general parameter measure_chi moved to solver section

  • general parameter delta_interface moved to solver section

• All possible input parameters are defined in the python/solid_dmft/io_tools/default.toml

• according to toml format the config file is now called .toml (instead of .ini), and boolean are not capitalized, strings are given with quotes and lists are given with brackets.

• Documentation of the input is now generated from python/solid_dmft/io_tools/documentation.txt

• For an example, refer to the new integration test (see below)

• Updated interface to python scripts wrapping solid_dmft: new routine main.run_dmft that expects the params as python dictionaries, which are then supplemented with the defaults etc equivalent to what happens when reading in a toml file

• the existence of the parameter general_params['beta'] now determines if a imaginary- or real-frequency grid is used within solid_dmft

• Bug fix: Slater interaction for p orbitals can now be constructed

• Renaming of solver parameters for the different solvers is now moved to solver.py. The idea is that every other part of solid_dmft should care as little as possible what solvers are used, with the details abstracted by the SolverStructure class

  • In solver.py, all solver parameters that are passed to the triqs solver are transferred to a dict triqs_solver_params. When adding new triqs solver parameter to solid_dmft in the future, they also need to be added within solver.py.

• In the determination of the block structure, the largely unused parameter general_params['block_suppress_orbital_symm'] removed. Its behavior can be replaced by using advanced_params['mapped_solver_struct_degeneracies']

• Integration tests: previously existing tests updated, new tests added. One with ftps solver (requires installation of ftps, otherwise just passes without doing anything) and one with a combination of CT-HYB and Hartree solver

• Unit tests: added test for toml-related functionality

• read_config.py removed and the functionality for dealing with the dicts from reading a toml file moved to postproc_toml_dict.py

• io_tools/verify_input_params.py contains all checks of the input params that the code performs before starting the DMFT calculations

• Updated the documentation of the input parameters

doc

• add comment that proj in postprocessing is only correct for diag A(k,w)

• update NNO magnetic tutorial

• fix Vasp CSC tutorial for PNO after CSC fixes

build

• add new tests for CRM Dyson solver (requires triqs 3.3)

• add new GW embedding tests that run optionally with -DTest_GW_embedding=ON

• modify basic SVO test to do crm test instead of gl

• add useful apt packages to openmpi image

• use ghcr.io images when testing PR

• ci: build and cache base image separately (#70)

• use new auto build ghcr.io docker images

• add GitHub Actions workflow for Docker images (#66)

• simplify dockerfile for github ci

• trigger pypi build on tags

• add pypi workflow

• update Vasp patches for ver 6.4

• Cleaned up VASP diff files for CSC

• use cmake variable to determine max number of mpi ranks during testing

other fixes

• Added warning for matrix-valued selfenergy continuation

• draw colorbar only once in kslices

• PCB bug aprx Sigma as diagonal if interpolation is used

• broken FS: np.shape -> len

• fix small FTPS problems and introduce a different eta for FTPS

• maxent test precision fix and test dependency

• use of origin in Fermi surface

• fix calculation of Akw for off-diag Sigma

Version 3.2.3

solid_dmft version 3.2.3 and 3.2.2 are minor releases that fixes bugs in the post-processing routines and brings small new improvements:

• allow 0.0 mixing to perform stat sampling

• allow mathematical expression to be passed for random_seed

• fix broken FS plot in PCB: np.shape -> len

• fix PCB bug aprx Sigma as diagonal if interpolation is used

• fix PCB to draw coloarbar only once in kslices

We thank all contributors: Alexander Hampel, Henri Menke

Version 3.2.1

solid_dmft version 3.2.1 is a minor release that automatizes the pypi packaging release

Version 3.2.0

solid_dmft version 3.2.0 is a release that

• adds Jenkins CI support via flatiron-jenkins

• includes several fixes to match the latest triqs 3.2.x release

• changes the Z estimate to a correct linear fit of the first two Matsubara frequencies

• fixes for QE and Vasp CSC

• add option to add a magnetic field in DMFT

• add solid_dmft JOSS paper reference (doi.org/10.21105/joss.04623)

• add simple Ntot interaction

• allow Uprime!=U-2J in Kanamori

• updates the tutorials

• introduces input output documentation

• add support for the TRIQS Hartree Solver

• add RESPACK support

We thank all contributors: Sophie Beck, Alberto Carta, Alexander Hampel, Max Merkel, Harrison LaBollita, Nils Wentzell

Find below an itemized list of changes in this release.

General

• fix SzSz measurement in triqs unstable

• Updated mpich VASP5 docker file to include HF solver

• add hartree solver

• feat: add regular kmesh option to pcb postproc

• Fix to charge-self-consistency with Vasp (#48)

• removed QE fix files which are now in official release

• Modified dockerfile to add pmi support for cray supercomputing environments

• add RESPACK postprocessing routines (#38)

• Added correction to energy calculation

• add triqs logos to skeleton and include ico in install directive of doc

• change name of dft_mu to mu_initial_guess

• support different DFT cubic basis conventions (#36)

• allow magnetic calculation for CSC (output den correction is always averaged)

• fix sym bug in hubbardI postprocessing

• always calculate dft_mu at start of calculation

• add h_field_it to remove magnetic field after x iterations

• Write solid_dmft hash to h5

• fix delta interface of cthyb for multiple sites with different block structures

• correctly use tail fitted Sigma from cthyb not via double dyson equation

• add paper ref to toml

• minor addition of post-processing script: add_local Hamiltonian, separate from add_lambda. We might remove add_lambda

• update doc with JOSS references

• Bug fix for changes in sumk mesh definition in maxent_gf_latt

• adapt vasp patch files for ver6.3.2

• function to det n_orb_solver, fix test

• apply block picker before block mapping

• fix header writing for obs file

• add pick solver struct option to select specific blocks for the impurity problem

• fix print for failing comparison test

• allow different interaction Hamiltonians per impurity

• enforce PEP standard in interaction Hamiltonian

• print optimal alpha in other maxent scripts

• final corrections for PCB functions

• add proj_on_orb functionality to Akw

• fix bug in max_G_diff function ignoring norm_temp

• change Sigma_imp_iw / _w to Sigma_imp (DFTTools unstable)

• fix load Sigma with new gf_struct in triqs 3.1.x

• adapt to sumk mesh changes in dfttools

• Made the way mesh is stored in maxent_gf_latt consistent with maxent_gf_imp

fix

• fix deg shells in magnetic calculations

• fix parameter n_orb in hint construction

• doc strings of cRPA avering for Slater

• critical bug in hubbardI interface

• PCB fermi surface plot

• updates from triqs unstable

• simple Z estimate as linear fit

• PCB: removing “linearize” function, changing the model

• delta_interface with SOC and store solver options

• convert warmup cycles to int automatically

• problem with ish vs icrsh in PCB Thanks @HenryScottx for reporting!

• h_int uses now n_orb instead of orb_names

build

• adapt jenkins CI files

• simplify docker image

• update openmpi docker file with clang-15

• update CI dockerfile

• Updated docker file to ubuntu 22

feat

• enable MPI for maxent_gf_imp post-processing routines

• add possibility to specify Uprime in Kanamori interaction

• add loc_n_min / max arg for cthyb

• add additional support for hartree when computing DC from the solver

• add Ntot interaction

doc

• Added observables documentation for DMFT output

• Updated tutorial svo one-shot

test

• fix tests after Hartree additions

• add Hartree Solver test

• Integration test for maxent gf imp and latt, bug fixes to both scripts (#30)

• add new test for pcb get_dmft_bands function

Version 3.1.5

solid_dmft version 3.1.5 is a patch-release that improves / fixes the following issues:

• fix to charge-self-consistency with Vasp and QE

• feat add loc_n_min / max arg for cthyb

• fix simple Z estimate as linear fit

• adapt docker images for ubuntu 22.04

Contributors: Sophie Beck, Alberto Carta, Alexander Hampel, Max Merkel:

Version 3.1.4

solid_dmft version 3.1.4 is a patch-release that improves / fixes the following issues:

• fix and improve rootfinder in PCB for quasiparticle dispersion

• fix pypi package version.py module

Contributors: Sophie Beck, Alberto Carta, Alexander Hampel, Max Merkel:

Version 3.1.3

solid_dmft version 3.1.3 is a patch-release that improves / fixes the following issues:

• fix delta interface of cthyb for multiple sites with different block structures

• correctly use tail fitted Sigma from cthyb not via double dyson equation

• magnetic param not available in CSC crash PM calc

• improve PCB script from unstable branch

• convert warmup cycles to int automatically

• fix function calls in gap finder

• fix delta_interface with SOC and store solver options

• fix: update svo example for PCB test from unstable

Contributors: Sophie Beck, Alberto Carta, Alexander Hampel, Max Merkel

Version 3.1.2

solid_dmft version 3.1.1 is a patch-release that improves / fixes the following issues:

• fix deg shells in magnetic calculations

• fix bug in max_G_diff function ignoring norm_temp

• fix load Sigma with new gf_struct in triqs 3.1.x

• Made the way mesh is stored in maxent_gf_latt consistent with maxent_gf_imp

• adapt vasp patch files for ver6.3.2

• update README.md for Joss publication

• print optimal alpha in other maxent scripts

• update postprocessing routines for plotting spectral functions

• add new test for pcb get_dmft_bands function

• DOC: extend install instructions & improve readme for #21 #22

• DOC: update support & contribute section, bump ver to 3.1.1

• add proj_on_orb functionality to Akw

• Added observables documentation for DMFT output

• Added input documentation

• Added ETH logo to website, small fixes to documentation

• rename examples to debbuging_examples

• pip package build files

Contributors: Sophie Beck, Alberto Carta, Alexander Hampel, Max Merkel

Version 3.1.1

solid_dmft version 3.1.1 is a patch-release that improves / fixes the following issues:

• delete obsolete make_spaghetti.py

• SOC self energies can be continued in maxent

• run hubbardI solver on all nodes due to slow bcast performance of atomdiag object

• fix DFT energy read when running CSC QE

• updated documentation, small fixes to tutorials

• exposed params of maxent_gf_imp

• fix the way dft_mu is loaded in PCB

• fix executable in SVO tutorial

• fix shift in sigma continuator to remove dft_mu

• fix chemical potential in plot Akw and minor fixes

• correct plotlabels in postprocessing

• tiny modification of printing H_loc in postprocessing

Contributors: Sophie Beck, Alberto Carta, Max Merkel

Version 3.1.0

solid_dmft version 3.1.0 is a major release that provides tutorials in the documentation, changes to app4triqs skeleton, allows CSC calculations with QE, improves postprocessing routines, and add functionality for SOC calculations.

• all new tutorials

• generalize measure_chi functionality

• CSC with Vasp 6.3.0 works, examples updated

• fix two bugs in w90 interface in vasp

• Renamed files

• fix Fermi level print in mlwf.F LPRJ_WRITE call

• Automatic patching of vasp 6.3.0 with Docker

• Updated tutorial

• Added check on all mpi ranks if dmft_config exists at beginning of run

• fix small bug in convergence.py thanks @merkelm

• Rework convergence metrics

• remove gf_struct_flatten from solver in accordance with latest dfttools version

• Renaming to solid_dmft

• Update of maxent_gf_latt.py: more parameters exposed and spin averaging is not default anymore

• fix bug in afm calculation when measuring density matrix

• Add w90_tolerance flag for CSC

• use sphinx autosummary for module reference

• small changes in IO, additional mpi barriers in csc flow for better stability

• With SOC now program prints real and imag part of matrices

• Fixed creation of Kanamori Hamiltonian with SOC

• Improvements in plot_correlated_bands.py and updated tutorial

• change output name of MaxEnt Sigma to Sigma_maxent

• change to develop version of w90 because of mpi bug in openmpi dockerfile

• bugfix in plot_correlated_bands and cleaning up

• update OpenMPI Dockerfile to latest Ubuntu

• Tutorial to explore correlated bands using the postprocessing script

• check in CSC with QE if optional files are presesnt, otherwise skip calculation

• Updated maxent_sigma: mpi parallelization, continuator types, bug fixes, parameters exposed

• update installation instructions

• add workflow and code structure images

• Updated maxent sigma script

• W90 runs in parallel

• Fixing a bug related to measure_pert_order and measure_chi_SzSz for afm_order

• add vasp crpa scripts and tutorials

• add delta interface for cthyb

• fix get_dmft_bands and pass eta to alatt_k_w correctly

• allows to recompute rotation matrix even if W90 is used

• bugfix in initial_self_energies.py in case dc = False

• flatten gf_struct for triqs solvers to remove depracted warning

• add example files for SVO and LNO

• bump triqs and package version to 3.1

Contributors: Sophie Beck, Alberto Carta, Max Merkel

Version 3.0.0

solid_dmft version 3.0.0 is a compatibility release for TRIQS version 3.0.0 that

• introduces compatibility with Python 3 (Python 2 no longer supported)

• adds a cmake-based dependency management

• fixes several application issues