Releases

Added
- Internally contracted CASPT2, including density fitting and IPEA shift (#3401)
- Analytic nuclear gradients for spin-restricted DF-CCSD and DF-CCSD(T), including auxiliary basis response (#3426)
- Analytic nuclear gradients for GHF with spin-free Hamiltonians and GKS with collinear XC functionals (#3420)
- TREXIO interface transferred from PySCF-forge to pyscf.tools (#3454)
- Secondary grids for SCF response functions, configurable through set_second_grids (#3439)
- ASE density of states interface (#3376)
- pob-DZVP-rev2 and pob-TZVP-rev2 all-electron solid-state basis sets (#3337)
- Support specifying the solvent name in PCM, ddCOSMO and ddPCM (#3416, #3541)
- Restricted open-shell MP2 (ROMP2), including density fitting (#3471)
- Particle-particle RPA (ppRPA) module transferred from PySCF-forge to PySCF core (#3478)
- Frozen natural orbitals for molecular and periodic post-HF calculations through the ADC module (#3462)
- Spin-orbit coupling with GTH pseudopotentials (#3436)
- COACH functional (#3475)
- Hirshfeld and CM5 atomic charges (#3534)
- Mayer bond orders and Wiberg bond indices (#3533)
- EOM-EE-CCSD singlet transition dipoles, oscillator strengths and excited-state analysis (#3480)
- Transition multipoles, oscillator strengths, natural transition orbitals and excited-state analysis for GHF/GKS TDDFT (#3490)
- Relaxed density matrices and dipole moments from CCSD and CCSD(T) gradient objects (#3477)
- Triplet A and B matrices in RHF/RKS get_ab (#3482)
- Quasi-RRHO entropy and enthalpy options, including a molecule-specific average moment of inertia (#3491, #3544)
- Composite B97-3c, r2SCAN-3c and wB97X-3c methods through the dft3c interface (#3403)
- Outlying charge correction for C-PCM and COSMO in COSMO-RS exports (#3421)
Improvements
- Faster Gamma-point PBC RCCSD using the optimized real CCSD kernel (#3452)
- Improved efficiency of SA-CASSCF nuclear gradients, including density fitting (#3444)
- Use the optimized AFT algorithm when kpts_band is a subset of kpts (#3392)
- Updated pbc.tools.precompute_exx for exchange singularity corrections and band calculations (#3412)
- Updated SAP potentials from fractionally occupied exchange-only LDA calculations (#3367)
- Faster RCCSD(T) lambda and response-density intermediates using C kernels (#3470)
- Batched KMP2 kernel (#3498)
- Improved finite-difference Hessians, convergence checks and excited-state tracking (#3428)
- Pass Hessian logging settings to the CPHF solver (#3531)
- Report installed PySCF extension paths in system information (#3526)
- Removed obsolete nao and icmpspt installation extras (#3501)
Fixes
- Missing orbital-rotation response terms in ROHF and KROHF second-order SCF solvers (#3448)
- State-specific CASCI/CASSCF MC-PDFT, including excited-state density matrices with DMRG solvers (#3440)
- KMP2 frozen-orbital masks now exclude orbitals marked with INVALID_ORBITAL_ENERGY (#3451)
- Recover the full Monkhorst-Pack mesh for meta-Lowdin population analysis in KRHF, KUHF and KGHF (#3450)
- Corrected Ba polarization shells in def2 and ma-def2 basis sets, missing Mn f shells and I f/g exponents in def2-QZVPD and def2-QZVPPD (#3382)
- UHF non-convergence messages report the actual number of completed cycles
- Unsupported DF-UCCSD analytic gradients now raise NotImplementedError instead of returning gradients inconsistent with the density-fitted energy (#3426)
- Premature Krylov termination in CPHF and UCPHF for near-dependent right-hand sides (#3334)
- Krylov projected systems with nonorthogonal trial vectors and per-right-hand-side fallback for difficult multi-root solves (#3549, #3384)
- Basis projection and MINAO initial guesses with nearly linearly dependent basis sets (#3333)
- X2C analytic gradients and Hessians with finite nucleus models (#3397)
- X2C diagonalization fallback for nearly linearly dependent basis sets (#3405)
- EOM-GCCSD star corrections for degenerate roots (#3404)
- Mixed-spin block of the UCCSD(T) response density (#3387)
- KRPA/KURPA occupation handling, exact exchange and momentum conservation for shifted k-point meshes (#3385, #3408)
- Use the solvent optical dielectric constant in non-equilibrium TDDFT (#3386)
- L-PDFT CI vectors with state-average mix solvers and gradient initialization (#3432, #3345)
- CASCI/CASSCF gradients when the reference calculation has not yet been run (#3394)
- Normalize the initial CIAH trial vector to prevent premature termination (#3406)
- KUHF occupations when there are no beta electrons (#3348)
- Nuclear energy evaluation with the PBC RSJK builder (#3388)
- Integer overflow in AFT orbital-pair counts (#3356)
- Shape handling for spinor four-center integrals (#3370)
- Stack overflow in relativistic direct J/K screening (#3369)
- Incorrect broadcasting in ft_ao and ft_aopair for a single reciprocal-space vector (#3340)
- POSCAR reader (#3364)
- Reduce Coov symmetry to C2v (#3422)
- Allow get_fock without arguments when dynamic level shifting is enabled (#3063)
- PYSCF_DISABLE_EVAL environment variable to disable eval-based behavior (#3363)
- Non-canonical MP2 singles c 8000 ontributions to energies and one-particle density matrices (#3471)
- CCSD(T) gradients now solve the (T) lambda equations when multipliers are not supplied (#3469)
- Consistent range separation of auxiliary-basis metrics in DF gradients and Hessians (#3435, #3446)
- Empty shell blocks in MP2, CCSD and multiconfigurational gradients with small memory limits (#3545)
- Saturated DFT grid-response derivatives and small denominators (#3525)
- Unsupported grid response in RKS/UKS Hessians now raises NotImplementedError (#3502)
- Symmetrize the Hessian before diagonalization in harmonic analysis (#3507)
- PCM intermediates now respect max_memory (#3509)
- Density and energy corrections in CASSCF with solvent (#3503)
- Selected CI with an empty spin channel; exposed as_SCIvector (#3481)
- SCF conversions with fractional occupations and closed-shell ROHF (#3474)
- GHF get_jk for molecular and periodic calculations (#3524)
- Unsupported smearing in second-order SCF and KROHF/KROKS now raises an error; corrected UHF LUMO reporting (#3526)
- Preserve the MINAO initial-guess electron count for ECP atoms (#3543)
- ECP sanity checks for non-symbol atom keys and ECP replication in periodic supercells (#3542)
- KMP2 orbital padding (#3527)
- FFTDF and AFTDF meshes after reset (#3473)
- UGWAC Fermi energy with smearing (#3551)
- MP2 auxiliary-basis selection for 6-311G** (#3539)
- Linear dependence in CABS projection (#3460)
- TREXIO AO integral ordering, ECP handling and imported basis normalization (#3463, #3467)
- Cube-file box vectors, odd-electron molecule reads and overlapping output columns for large structures (#3488, #3499, #3506)
- Molden spherical d/f function declarations (#3497)
- MC-PDFT energy logger formatting (#3492)
- Linear-algebra exception handling in DIIS (#3516)
- Sigma-state reflection symmetry in the Dooh/Coov cylindrical-symmetry FCI solver (#3519)
- Clear stale ECP, pseudopotential, atom, basis and symmetry data when rebuilding Mole/Cell (#3532)
- Fix AGF2 complex matrix-vector products (#3532)
- Molecular GW diagonal self-energies now evaluate each frequency separately (#3532)
- Unrestricted exact-GW static corrections apply to every orbital (#3532)
- Periodic GW linearized density matrices now include the missing G0 contribution. (#3532)
- Periodic DF Coulomb matrices retain imaginary contributions from complex non-Hermitian Gamma-point transition densities (#3532)
- Reusable ddPCM solver helper calls and experimental SMD gradient imports (#3532)
- Prevent out-of-bounds factorial-table reads in high-angular-momentum ECP derivatives (#3532)
- Periodic meta-GGA kinetic-energy density with negative occupations (#3532)
API Updates
- Python 3.9 or newer is required (#3500)
- Updated libcint to 6.2.0 and LibXC to 7.1.2 (#3523, #3336)
- MP2 update_amps now takes (mp, t1, t2, eris) instead of (mp, t2, eris) and returns (t1, t2) (#3471)
- Non-canonical MP2 includes T1 singles by default (#3471)
- mp.MP2 with an ROHF reference now returns ROMP2 or DFROMP2 instead of UMP2 (#3471)
- Smearing SCF convergence uses the free-energy change and orbital gradient in the main SCF loop, instead of the total-energy change (#3410)
- Restored standard auxiliary-basis normalization for periodic range-separated density fitting (#3510)
- Density-matrix keyword for SCF response functions (#3415)
- Add NumInt.is_hybrid_xc and NumInt.is_nlc to replace the direct calls to functions in the libxc module (#3396)

Added
- Gamma-point G0W0 for GWAC
- Bethe-Salpeter equation (BSE) for molecular GW module. Support both spin-restricted and spin-unrestricted.
- KRPA/KURPA: new routine for spin-restricted/unrestricted k-point RPA
- PBC RPA for smeared systems
- RCCSDT(Q) correction implementation
- MC26 and COF26 on-top functionals
- normalize CF22D functional and enable its built-in D3 dispersion
- spin-orbit coupling (SOC) Hamiltonian for GCCSD
- Implement CABS Singles correction (#3265)
- symmetry breaking via homo-lumo-mixing (#3271)
- Nuclear Gradients of GTH PP for molecules
- qavg-vSZPs basis and ecp-q-vSZP
- Analytic gradients for CASCI with UHF/RKS/UKS orbitals
- self-consistent dipole corrections for slab/2D systems (#3260)
Improvements
- Windows support: DLL wrapper, compatibility
- GW improvements, including
  - PBC G0W0 (#3222). Improve CPU and memory efficiency for KRGWAC and KUGWAC.
  - GW improvements: self-consistent PBC GW routines (#3244)
  - molecular GW routines (#3236)
- Support complex orbitals in GCCSD
- gto/ecp: fix premature convergence in radial quadrature at large exponents (#3250)
- Refactoring high-order CC implementations
- KRPA and KURPA: improve the numerical stablity of ACFD EXX
- SCF logging and add HOMO-LUMO gap reporting (#3281)
- Make wb97x-d4 configurable. DFTD4 recommended variants can be enabled by the parameter dispersion.DFTD4_RECOMMENDATIONS.
- Molecular Pipek-Mezey localization code refactored.
- Pipek-Mezey Wannier functions for PBC k-point orbitals constructed using CIAH algorithm. Wannier interpolation for PBC k-point calculations.
- allow customizing functionals by param name.
Fixes
- Load basis sets from the Basis Set Exchange (BSE) when elements are missing from local .dat files
- init_guess_by_mod_huckel function can now be called without arguments
- fixed memory over allocation in PCM gradients (fixes #3246)
- Fixed KeyError for MM ECPs in project_to_atomic_orbitals (#3243)
- Molden parser now handles files without the Sym field
- pbc/df: chunk kpt-pairs in CCGDFBuilder.outcore_auxe2 to avoid OOM (#3257)
- Atomic approximation for GHF X2C
- meta-GGA second derivative ordering for custom XC
- bad atomic radii assigned to ghost atom, which causes a bad Becke partition and a negative BSSE (#3302)
- density fitting UKS hessian

Fixes
- Missing CP2K basis set data in wheel distributions
- Small-rotor error.
- Corrected SG1 grid radii handling for ghost atoms.
- Fixed ECP loading to correctly fall back to Basis Set Exchange when local data is unavailable.

Added
- ASE band structure interface
- SOC-ECP for PBC DFT methods
- get_rho_with_derivatives() function
- PCM ISWIG surface discretization method
- grid_response support for the Stratmann grid partition scheme
- Cartesian basis support in the PCM solvent module
- gw_ac and ugw_ac methods
- meta-GGA support in multigrid methods (currently computed in a slower mode)
- SGX gradients
Improvements
- Warnings for basis-element mismatch
- Updated GTH basis set and pseudopotential databases
- Enforced real orbitals for time-reversal symmetric KSCF
- Added sanity checks to prevent bugs in the Newton solver for PBC mean-field methods
- Updated ccECP from pseudopotentiallibrary.org
- Improved PBC 2c2e integral accuracy for better numerical stability in PBC-GDF methods
- SGX implementation for improved numerical accuracy and performance
- Canonical orthogonalization by default in the SCF eigenvalue solver
- Support specifying "autoaux" as a key for auxiliary basis set
Fixes
- RSJK initialization
- Skipped printing of velocity and magnetic dipole moments when ECP is detected
- mo_coeff reshape size in PBC SCF HF DF-JK
- XYZ file parser
- A bug where PCM assigned incorrect radii to ECP atoms
- Inconsistent symmetry axis assignment for D₂h molecules (#2773)
- Load default cell and k-points in KGHF.get_hcore
- POSCAR parser bug
- Handle ghost atoms in the QM/MM interface
- Errors in AO values caused by an incorrect diffuseness estimator
- Edge case involving empty arrays in DF-JK construction for PBC DF
- Incorrect DF-MP2 energy when using jkfit in HF and rifit in MP2
- Edge case for the frozen attribute (frozen=0)
- PBC nuclear gradients in fft_jk.get_k_e1 with exxdiv=ewald
API Updates
- Set multigrid2 as the default multigrid integrator
- Smearing functions are moved to a separate module
- Set default auxbasis_response level to 2 (full response) in density-fitting Hessian methods
- Dump nelecas in mcscf chkfile
- Allow to skip .run() for intermediate instances in method chains
- In method pipelines, .run() of intermediate steps is now executed automatically
- Linear dependency for orbital bases are removed by default in SCF

Fixes
- NumPy 2.4.0 compatibility issue caused by mixing of scalar and 1D array
Improvements
- Optimization function for improved convergence and stability in smearing.

New Features
- Velocity gauge integrals and oscillator strengths for TDSCF using GTH basis sets
- High Performance CCSDT and CCSDTQ Implementation (T1-dressed formalism)
- GHF stability_slow
- Density fitting implementation for the NEVPT2
Improved
- Adjust UHF and ROHF analyze() to be compatible with msym
- Evaluate pseudo potential nonlocal term gradient on real space, the same as corresponding Fock evaluation
- Improve GPU4PySCF interface
- In pseudopotential nonlocal term gradient, if large imaginary is found, print warning instead of raise error
- Improve PBC DFT instantiation
- support complex vector for fci.direct_spin1 solver
- Reduced memory usage for pbc RSGDF
- Configurable einsum backend, supporting pytblis, tlibs-einsum and tensordot
Fixes
- KRHF gradients when exxdiv='ewald'
- segfault for qcint backend
- fix a bug in x2c, where imag part has been deprecated.
New Features and Improvements in the Molecular ADC Module
- Expanded 1-RDM Functionality:
  - Introduced 1-RDM calculation functions for CVS-IP-RADC, CVS-IP-UADC, EE-RADC, and EA-UADC methods.
  - Enhanced reference 1-RDM calculation function in UADC.
  - Updated 1-RDM functions in IP-UADC and EE-UADC to include contributions up to ADC(3).
- Expanded oscillator strengths Functionality:
  - Enabled oscillator strengths calculation for EE-RADC.
  - Introduced get_properties function for EE-RADC.
  - Updated get_trans_moments function EE-UADC to include contributions from 3rd order singlet excited amplitudes
- Advanced Orbital Control:
  - Implemented frozen orbital functionality across all ADC cases, offering greater computational efficiency and control.
- Improved Initial Guess Strategies for ADC:
  - Added CIS initial guess options for EE-UADC and EE-RADC.
  - Introduced user-defined initial guess capabilities for all ADC methods, providing increased flexibility.
- Enhanced Analysis and Output:
  - Integrated analyze_eigenvector function into EE-UADC, EE-RADC, CVS-IP-RADC, and CVS-IP-UADC for detailed analysis.
  - Added analyze_spec_factor for EE-RADC.
  - Updated analyze function for CVS-IP-RADC, CVS-IP-UADC, EE-RADC, and EA-UADC methods.
  - Improved logging in kernel and kernel_gs functions (UADC and RADC) to report the number of frozen and active orbitals.
- Bug Fixes
  - Resolved a bug in 1-RDM functions that occurred when approx_trans_moments was set to True.
  - Fixed merge conflicts.
- General Improvements
  - Replaced long floating-point numbers with fractional representations for improved precision and readability.

Added
- MCPDFT, XMS-PDFT, CMS-PDFT, and L-PDFT
- Analytical Gradients for CASSCF-PDFT, SA-CASSCF-PDFT, CMS-PDFT, and L-PDFT
- Non-adiabatic coupling vector for CMS-PDFT
- QM/MM interface for PBC systems.
- Analytical nuclear gradients for DFT+U methods
- Gaussian charge models for MM particles in QM/MM interface
- MC25 functional parameters
- Stress tensor for semi-local functioanlsGGA,
- Fixed-Size APC active space selection
Improved
- libxc interface for customizing functional parameters
- Support the 'default' key in the input for the mol.ecp and grids.atom_grid
- Print imaginary frequency as negative frequency in dump_normal_mode()
- RHF exchange performance by preserving MO information
- Improve performance of PM localization with Becke charges
- Molecular density fitting RMP2 and UMP2 support PBC gamma-point mf instance
- ASE interface to support lattice optimization and geometry optimization for PBC systems
Fixes
- to_gpu interface for solvent modules
- The mismatch exx between k-point and super-cell calculations caused by the inconsistent treatment in get_coulG and super_cell functions
- cell.KRKS() instantiation for symmetry adapted k-points
- Return converged == False if extra cycle in SCF is not converged
API changes
- move state_average_mix to CASBase
- Remove the misleading warning exp_to_discard

Added
- Analytical hessian for vv10 functional
- vv10 response contributions in TDDFT
- vv10 functionals for GKS
- Spin-seperated 4-particle density matrix for FCI
- TDDFT with C-PCM/IEF-PCM solvents and the corresponding TDDFT analytic gradients
- DFT+U for molecules and a linear-response method for determining U
- Add multigrid_numint method to setup the multigrid algorithm in DFT instances.
- Frozen orbitals for TDDFT
- EE-ADC for spin-conserved electronic excitations
Improved
- Checks and warnings for missing ECP input on heavy elements
- Auxiliary basis set assignment based on BSE database
- Support for user-defined basis set aliases
- Multigrid performance for non-orthogonal lattice
- Multigrid APIs now follow the NumInt module; multigrid is decoupled from the
  FFTDF implementation, supporting mixing of GDF and multigrid
- The treatment of dimension=0 under PBC; the PBC dimension=0 can be called
  with FFTDF, AFTDF, GDF, Multigrid integrals now.
- Breit and Gaunt integrals for non-hermitian density matrices, supporting the
  TDDFT computation with Breit and Gaunt interactions.
- Low-Memory and Efficient Implementation of RCCSD(T) lambda-equation and RDM Intermediates
Fixes
- transform_xc for 6th order XC derivatives
- Subgroups of C2h
- EXX integral computation in RSH functionals for PBC DFT with AFTDF
- Application order of solvent models and density fitting
- Slow convergence in GHF SOSCF
- Bugs in the geomeTRIC interface: max iteration, convergence check, and non-standard element symbols

Added
- Analytical PCM Hessian
- Density fitting UCCSD
- A general driver for finite difference gradients and hessian
Improved
- Assign default auxiliary basis for def2-mtzvp and def2-mtzvpp
- Better initial guess for smearing
- libxc interface for passing omega in composite functionals
- The tolerance in geometry when identifying point group symmetry
- Adjust PBC GDF eigenvalue decomposition accuracy
- Special treatment in PBC GDF for dimension=0 systems
- The density matrices computation performance of fci_dhf_slow
Fixes
- Dimension bug in spinor X2C code
- C-PCM and SS(V)PE gradients
- Complex density matrices for DFHF
- Fixes COSMO-RS functionality
- DF auxiliary basis assignment for ghost atoms
- Lattice sum range issue for low-dimensional systems
- The precision when dumping molecular geometry in fromfile/fromstring function
- kpts_band in the rsdf method.

Added
- The ccpy extension (https://github.com/piecuch-group/ccpy) which enables various Coupled cluster methods
- COSMO-RS functionality
- DFMP2 and RPA for UHF reference
- Add functions to read and write Cell geometry, allowing Direct or Cartesian in VASP input
- Add def2-mtzvp and def2-mtzvpp basis sets for 3c methods
- libqcschema module to load qcschema json
Improved
- Integral screening for Gaunt and Breit term
- Performance and multi-threading efficiency of DFMP2 and RPA, gaining 4x - 8x speedups
- Stability analysis for extended system, matching with molecular version
- Update to Libxc 7.0
- Adjust TDDFT amplitudes to follow the CIS convention
- Accelerate DFT density and XC potential, especially for MGGA.
- Automatically apply SCF initial guess from existing wavefunction.
- Improve the first order derivatives of density fitting J/K matrices for non-hermitian density matrices.
- Improve TDDFT diagonalization numerical stability.
- Remove redundant integral computation for SR-only (such as HSE) and LR-only (such as wb97) RSH functionals
Fixes
- ASE interfaces regarding to the API changes in ASE v3.11.0
- Missing conj() for t2 amps in GCCSD and UCCSD routines
- UCCSD damping
- Fix biased implementation for the becke radi method.
- Fix DFT define_xc_ interface and examples.
- Fix SR-ERI integral screening estimator.
- Fix CCSD two-particle density matrix for complex orbitals.
- Take cell.rcut into account when generating becke grids for low-dimensional systems.

PySCF v2.15.0 release

Uh oh!

PySCF v2.14.0 release

Uh oh!

PySCF v2.13.1 release

Uh oh!

PySCF v2.13.0 release

Uh oh!

PySCF v2.12.1 release

Uh oh!

PySCF v2.12.0 release

Uh oh!

PySCF v2.11.0 release

Uh oh!

PySCF v2.10.0 release

Uh oh!

PySCF v2.9.0 release

Uh oh!

PySCF v2.8.0 release

Uh oh!