-
-
Couldn't load subscription status.
- Fork 650
Description
Based on the discussions in the 2022 PySCF developers' meeting, requested features are collected in the list below. If you are looking for features not available in PySCF or you are interested in contributing new features, this list can be used as a reference.
[Names] in the "[]" at the end of each statement are the people who are currently developing or have plans to develop the feature. The list will be updated roughly monthly.
People
AYS = Alex Sokolov or Sokolov group
CL = Chenghan Li
GC = Garnet Chan or Chan group
HY = Hongzhou Ye
HZ = Huanchen Zhai
JK = Janus Kristen
JY = Junjei Yang
KL = Ke Liao
MRH = Matthew R Hermes
NR = Nick Rubin
QS = Qiming Sun
SL = Seunghoon Lee
SS = Sandeep Sharma or Sharma group
TB = Tim Berkelbach or Berkelbach group
TZ = Tianyu Zhu
XW = Xiao Wang
XZ = Xing Zhang
YZ = Yu Zhang
ZC = Zhihao Cui
KB = Kyle Bystrom
XJW = Xiaojie Wu
Features
- GPU plugins for pyscf [QS,XJW,GC]
- Hartree-Fock (RHF, UHF, GHF) on GPU [QS,XJW,GC]
- Release pipelines of gpu plugins, mpi plugins [QS,XJW]
- DFT on GPU [QS,XJW]
- RCCSD on GPU [QS]
- AO evaluation on GPU would benefit PyQMC [QS]
- Integration of SymTensor/ctf-tensor/ctf-cc as addons [GC needs help with PR]
- Support vendors other than NVIDA [JK]
- HF and DFT nuclear gradients and Hessians on GPU [XJW]
- with density fitting [XJW]
- without density fitting [GC,XJW]
- Extension of molecular GPU HF to PBC
- pbc.df module in mpi4pyscf needs to be tested and updated against pyscf v2
- Benchmark GPU dispatch from JAX PySCFAD for many methods
- df ao2mo on GPU
- FCI on GPU
- FCI with MPI
- plane wave-based HF and post-HF [HY]
- Integral-direct DFT, HF and MP2 with Gaussian density fitting [HY]
- Nuclear Gradients with integral direct GDF
- PBC integrals evaluated in real space (the RSJK module) [QS]
- fast FFTDF J-build [GC,XZ]
- fast exchange from robust pseudospectral method [SS]
- Match popular PW-based DFT code
- k-point symmetry
- PW basis
- DFT
- k-point symmetry [XZ]
- Non-local xc (e.g., VV10) issue Efficiency improvements and analytical gradients for some (semi-)numerical integration routines #1471 [KB,QS]
- Pseudopotential (e.g., UPF with GTOs)
- 1st derivative: forces, stress and geom opt
- 2nd derivative: DFPT, e-ph coupling
- Charged unit-cell
- Bands interpolation
- Lower dimension systems [QS: 1D systems]
- Optimization integral grids [KB]
- Faster exchange
- k-point symmetry [XZ]
- FFTDF
- MO-RI, occ-RI, screening
- (RS)GDF
- better in-cell screening for large unit cell
- local density fitting
- Electron correlation
- k-point symmetry
- RPA (more quantum chemistry way?) [TZ]
- Response and excited states
- EOM-CC, CCGF [XW]
- GW [TZ]
- BSE (molecular version done; need k-points) [TB]
- ADC [AYZ]
- Reduced-scaling techniques
- Local correlation (PAO, OSV, LNO, PNO, etc.)
- QMC, AFQMC (better DF/CD)
- cRPA [TB]
- THC [NR]
- Fragment embedding
- Regional embedding
- DMET
- DMFT
- Unifying APIs for mol & pbc basis/pp/ecp [QS]
- Solvent models for periodic calculations
- Optimize auxiliary basis for GDF and release the basis database
- Optimize AFTDF performance [QS]
- Large scale pseudopotential DFT
- FFTDF2 or multigrid [XZ, CL]
- Gradients still under development
- MPI to be implemented
- Linear scaling SGX
- benchmarking [Bytedance]
- integral performance and grids improvement (issue Efficiency improvements and analytical gradients for some (semi-)numerical integration routines #1471) [KB]
- Fragment methods (update from developers?)
- DFT numerical integration optimization (issue Efficiency improvements and analytical gradients for some (semi-)numerical integration routines #1471) [KB]
- MPI for exc/vxc evaluation and numerical integration
- Large scale simulations at correlated level
- Cluster in molecule [TB]
- MPI4PySCF [ZC]
- SHCI with bosons, interfacing PySCF with electrodynamics packages (MEEP) [SS]
- enhanced interface QM/MM, [CL implemented gradient w.r.t. MM part]
- More semi-empirical methods? [YZ: currently we are moving our SEQ code into python platform (with GPU as well)]
- FMM or related methods?
- AO2MO with sparsity for localized orbitals
- PNO based methods
- density fitting gradients beyond mean-field methods. Better API of nuclear gradients (Post-SCF) of density-fitted integrals [MRH]
- Faster CCSD(T): improving existing module, implementing local CCSD(T), maybe PNO?
- Transition state search
- ctf-ccsd implementation [KL]
- QM/MM interface maintenance
- Failed to install pyscf on conda for newest Python versions
- Installing issue for new chips (e.g. apple M1)
- Single-reference methods for simulating one (ground or excited) state at a time
- State-specific MP2 (+ DF)
- State-specific CI (up to CISD)
- State-specific CC (up to CCSD(T)) with UHF, RHF, and GHF references, QCISD)
- Single-reference methods for simulating many states, spectra
- TD-SCF / TD-DFT
- RPA, GW
- ADC(n) (n <= 3) for IP, EA (core module + interface to adcc)
- AGF2
- EOM-CC for IP, EA, EE (up to EOM-CCSD)
- Multireference methods for excited states and strong correlation
- FCI / CASCI / CASSCF
- Interface to Block (DMRGCI / DMRGSCF)
- Interface to Block2 [HZ]
- DMRG-RCASSCF - energy, nuclear gradients, point group
- DMRG-UCASSCF - energy, <S^2> (PR Make spin_square work for a custom fcisolver in UCASCI/UCASSCF #1253)
- Interface to Dice (SHCI / SHCISCF?)
- Strongly-contracted state-specific NEVPT2 (sc-NEVPT2)
- Tailored, externally-corrected CCSD [SL/GC - PR soon]
- ADC for EE in the core module [AYS group]
- Partially-contracted state-specific NEVPT2 [AYS group]
- Partially-contracted state-specific NEVPT2 + REPT2 [HZ/GC, ready for PR]
- Partially-contracted quasidegenerate NEVPT2 [AYS group]
- Spin-orbit coupling for pc-NEVPT2 [AYS group]
- Interface to QC-Maquis [PR open, GC]
- Interface to Block2 [HZ]
- (DMRG-) UCASSCF - nuclear gradients and point group
- (DMRG-) GCASSCF/DCASSCF/KCASSCF
- DMRG-uncontracted MRCISD(T, Q …)
- Fully internally-contracted MRCISD (slow)
- EOM-tailored CCSD [SL/JY/GC]
- Lambda equation of tailored CCSD [JY/GC]
- Support of CSFs in CASSCF [MRH]
- Electron-phonon coupling ? (there is an eph module already)
- CCSD (and some relevant methods) with less IO [QS]
- Interface to MRCC [XW]
- CASPT2 (worth implementing) [GC]
- Tools for simulating core-electron excitations
- Spin-orbit coupling for core-excitations (Lan Cheng has some ways to do it)
- Versatile MCSCF wave functions: RASSCF/GASSCF
- TD-DFT Restricted Excitation Window for Core Excited States
- Oscillator strengths from EOM-CCSD
- x-ray core excitation, a better interface to run them use pyscf and block code
- Relativistic effects for spectroscopy of core electrons
- non-orthogonal CI, using QMC algorithms may be developed in the future
- Response algorithms for correlated methods (openRSP code, properties beyond linear-response)
- Tools for code generation (egraph, egg)
- More multireference perturbation theories, quasidegenerate perturbation theory
- Better interface to compute spin-orbit coupling (maybe some code from Dice can be incorporated in core PySCF?)
- CASSCF for DHF
- DMRG for periodic systems, can be useful with k-point symmetry (k-point CASSCF/DMRGSCF?)
- Orbital-optimized dft for state-specific excited states
- MP3 code from ADC can be moved to MP module for molecules and materials [AYS group]
- Remove the monkey patch to fix_spin_ function, Fix_spin_ MPI bug in SA-CASSCF, issues with conserving spin (MRH report)
- "pip install" to include everything [QS]
- Pyscf-forge [QS]
- Maintainer list for each module. [QS]