# Change History of GAMESS

Note that the version is not a number, but rather a date printed in the box appearing at the top of output files. This changes periodically when some new scientific capability appears. The weekly or even daily source code changes to fix bugs, support new machines, or add minor scientific capabilities are not noted here.

Sept 30, 2017:

OpenMP threaded RI-MP2 (energy) - see tests/rimp2omp for usage examples.

OpenMP threaded EFP Charge Transfer (energy) - see tests/efp-mpiomp for usage examples.

OpenMP threaded MAKEFP - see tests/efp-mpiomp/makefp for usage examples.

QMC-EFMO interface between QMCPACK and GAMESS - see tools/qmc_efmo/README for more information.

Apr 20, 2017:

GPU-accelerated resolution of the identity (RI) approximation
for RHF (energy and gradients) or UHF (energy) based MP2
calculations is added to LIBCCHEM. See new input group $CCHEMRI
in INPUT.DOC for more details. Input examples are available in
the libcchem/tests/rimp2Tests folder.

Aug 18, 2016:

Changes to solvent models can be found in EFP2 (where the
exchange repulsion for EFP2/quantum systems is coded, and
for which a R^(-7) term is added to the dispersion energy and
gradient), and in PCM which can be used with multireference
perturbation theory. For heavy element chemistry, the LUT-IOTC
method allows quick and accurate energy and gradient calculations,
for all ab initio runs, with Sapporo bases provided for all
6th row elements. There are substantial improvements made to
the bonding analysis using Singular Value Decomposition orbitals,
such as bond orders and energy breakdowns. FMO calculations
now include the following: UMP2 and DFTB energy and gradient;
Hessians, IR, Raman for RHF/FMO3, ROHF, DFT, and FMO/FDD;
PCM solvation with ROHF, UHF, and DFTB; SCZV gradient terms for
FMO3, DFTB, DFTB3; use of ECP potentials; RATTLE in MD, and
IRC runs. The ZAPT2 energy has been added to libcchem.

Dec 5, 2014:

Analytic hessians for UHF, R-DFT, and U-DFT are implemented.
Chemical bonding analyses include Valence Virtual Orbitals
following RHF (and DFT), ROHF, GVB, and MCSCF calculations,
forming atomic- or split-localized orbitals, and localization
of the external orbital space. Single and multi-reference
CEPA, ACPF, and AQCC correlation energies may be computed.
Frequency dependent alpha and beta polarizabilities may be
obtained for DFT cases. Solvation modeling changes include
EFP discrete solvation for CI functions and the CMIRS continuum
model. Density-functional tight-binding (DFTB) calculations
may be performed (energy and gradients). Major new FMO
capabilities are fully accurate gradients for R-DFT, ROHF,
and FMO-DFTB, approximate gradient for U-DFT, FMO hessians
for RHF and UHF (with IR or Raman spectra), and computation
of UTDDFT energies. Multiple open shell FMO fragments may
now be used in single reference methods. Dual basis FMO is
available (FMO/AP). SIMOMM is interfaced with FMO (FMO/MM).

May 1, 2013:

Several new electron correlation treatments are added:
Correlation Energy Extrapolation by Intrinsic Scaling (CEEIS)
for excited states, Multi-reference perturbation theory for Occupation
Restricted Multiple Active Space (ORMAS) references, and the Clusters
in Molecules (CIM) approach for large molecule coupled cluster energies.
Diabatic potential surfaces can be obtained at the MCSCF and multi-reference
perturbation levels. Solvent model changes include the use of EFP or
PCM models with spin-flip TD-DFT, dispersion in the EFP2 method, and
the use of PCM with semi-empirical MOPAC wavefunctions. Spin-restricted
"constrained UHF" (CUHF) open shell SCF calculations are possible. The
EFMO model (Effective Fragment Potential version of Fragment MO theory)
contains many improvements, including dispersion, charge transfer,
exchange repulsion, and charge penetration terms, and may be used for
fragments connected by covalent bonds, with multilayer capability and
frozen domains. The Fragment MO method has fully accurate gradients
for RHF and UHF, gradients for PCM with MP2, and an energy analysis for
EFP water solvent interactions. Gas phase energy calculations may use
h & i Gaussian basis functions, and thus any member of the correlation
consistent or polarization consistent basis sets.

**
The C++ add-on program "libcchem" now permits evaluation of the closed
shell MP2 and closed shell CCSD(T) energy on GPUs.
**

May 1, 2012:

Technical improvements such as Z-vector solutions to the MCSCF response
equation solver mean that state-averaged MCSCF gradients or NACME vectors
are available, for CAS-SCF and ORMAS. The state-averaging weights may
be dynamically adjusted. The multi-reference calculations (MCSCF, MRMP,
and MRCI) are interfaced to the MPQC program to obtain PT2-R12 corrections.
Valence Virtual Orbitals (for example, useful for starting MCSCF)
are available for all elements to Xenon. The VVOs may also be used to
speedup up charge-transfer in EFP2 calculations. The FMO program's
fully analytic gradient can be used with EFP particles present.
Changes to the DFT program include the availability of M11 functionals,
and two photon absorption cross-sections during TD-DFT. The source code
for the VB2000 valence bond program is now included, and compiled by
default on some machines.

August 11, 2011:

Solvation model changes include implementation of the SMD parameterization
for continuum solvation. The EFP discrete solvation model and the PCM,
COSMO, and SMD continuum models are fully integrated with all MP2 and
TDDFT gradient codes. QM/EFP energy analysis is possible. Excited state
calculations can be performed by spin-flip of open shell references at
the CIS and TDDFT levels, for nuclear gradients. The Tamm/Dancoff
approximation to TDDFT calculations is enabled. Conical intersections
may be found on MCSCF or TDDFT potential surfaces. The QuanPol program,
including polarizable aqueous solvation, is provided for QM/MM
or MM calculations such as MD simulations. Accuracy for the closed
shell Fragment Molecular Orbital gradients is improved, and geometrically
frozen domains can be used. The Local Response Dispersion correction
for DFT calculations may be evaluated. Sapporo-family basis sets for
all-electron scalar relativity are offered for the first 5 atomic periods.

A C++ program "libcchem" permitting closed shell Hartree-Fock computations by GPU processors is provided as an option.

October 1, 2010:

The EFP solvation program has been extended to TD-DFT gradients.
The PCM solvation program has been extended to open shell MP2 gradients.
The COSMO solvation program has been extensively revamped, and works
for RHF, UHF, and ROHF references at the HF, DFT, and MP2 levels.
Relativistic codes added are the Infinite Order Two Component tranformation
for scalar effects, and the ability to generate the density matrix and
properties for spin-orbit coupled states, along with a much wider range
of Model Core Potential basis sets.
TDDFT excitation energies can now use all available metaGGA functionals.
The semi-empirical RM1 parameterization has been added.
FMO changes include the EFMO method (a merger of EFP and FMO) and more
accurate gradient calculation.
A very efficient resolution of the identity approximation for RHF or UHF
based MP2 calculations is added.
A special program for integral evaluation and closed shell Fock matrix
construction on GPUs is provided, and support for compilation under Windows.

March 25, 2010:

New DFT functionals are allowed: the B97/HCTH family, range separated
functionals CAM-B3LYP and the wB97 family, and revised TPSS and M08-type
metaGGAs.
Additional Coupled-Cluster calculations are the excitation energy
left-eigenstate triples corrected CR-EOML, and EA-EOM or IP-EOM
methods for electron affinities and ionization potentials.
PCM solvation calculations can use heterogenous dielectrics, gradients
are allowed for closed shell MP2 and closed shell TD-DFT calculations,
and FIXPVA tesselation works for all PCM options.
Gas phase TD-DFT and ground state PCM nuclear gradients have been enabled in
the Fragment Molecular Orbital (FMO) method, in addition to the calculation
of molecular orbitals and their energies for the whole system (FMO/F).
An energy decomposition analysis using localized orbitals can be made,
for very general wavefunction types in each subsystem.

Special attention has been paid to simplification of the compilation process, particularly for high end machines and Infiniband clusters, together with improvements in the DDI messaging layer.

January 12, 2009:

The continuum PCM and the discrete EFP solvation models have been
updated, along with the PCM/EFP interface, with a new PCM tesselation
providing very accurate nuclear gradients. Non-adiabatic coupling
matrix elements between different MCSCF state-averaged states of
the same spin can be evaluated. The DFT program includes the TPSS
meta-GGA family of functionals, and solvation effects on vertical
TD-DFT excitation energies can be found using either PCM or EFP models.
Fragment Molecular Orbital (FMO) improvements include FMO/TD-DFT and
FMO/TD-DFT/PCM, increased accuracy in nuclear gradients, the use of
MCPs, and the Adaptive Frozen Orbital (AFO) alternative fragmentation.
A program for the calculation of SCF/DFT, MP2, and/or CCSD energies
by the divide+conquer method is included.

April 11, 2008:

The DFT program now permits the use of many additional GGA and meta-GGA
functionals, such as M06. The TD-DFT code will compute nuclear gradients for singlets, runs in parallel, and excited
states for a UHF reference can now be generated. Both ground and excited state DFT calculations can use Lebedev or the
"standard grid 1" grids. A parallel MP2 energy and gradient program for closed shell references
which stores all quantities on disk storage (rather than in distributed
memory) has been added.High spin open shell coupled cluster energies can be obtained, including
the novel triples correction CC(2,3), for spin-restricted references.
The GMCQDPT (general reference multiconfigurational quasi-degenerate
perturbation theory) code allows 2nd order energies to be found for
ORMAS and other non-complete MCSCF wavefunctions.
An automated procedure for extrapolation to full CI energies, the Correlation Energy Extrapolation by Intrinsic Scaling
procedure, is included. Changes to the EFP solvation model include treatment of open shells,
improved MD, and Ewald sums. The Vibrational SCF (VSCF) program is now capable of working in internal
coordinates for increased accuracy, and can utilize "group" parallel
computation. Thermochemical data can be computed using the G3(MP2,CCSD(T)) method.

March 24, 2007:

Analytic gradients are now provided for Model Core Potentials (MCP).
The Fragment Molecular Orbital (FMO) method now includes 3-body MP2
computations and the pair interaction energy decomposition analysis (PIEDA).
FMO has also been interfaced to the Effective Fragment Potential solvation
model. The VSCF method for anharmonicity can compute combination bands,
and generate its potential surface in internal coordinates. The density
functional program has been extended to excited states through the
TD-DFT procedure, and the number of ground state functionals is greatly
increased. The EFP model's screening has been extended to higher order,
and to its polarization. The ORMAS CI code has been enabled for parallel
computation. The NEO option has improvements to allow computing positron
wavefunctions.

September 7, 2006:

Parallel computation of the closed shell CCSD or CCSD(T) energy is
now enabled. The speed of MCSCF analytic hessians is greatly improved,
and the ability to compute ORMAS-type MCSCF hessians is added. The
minimum energy crossing point between two surfaces (different space
and/or spin symmetry) can now be located. The Nuclear-Electron Orbital
method (NEO) for treating nuclei quantum mechanically may be chosen as
a compile-time option, for nuclear computations at the HF, MP2, CI, or
MCSCF levels. The ful model for the Surface Volume Polarization for
Electrostatics method of continuum solvation is implemented. The
Elongation Method for polymer growth is included. Localized orbital
analysis of molecular hyperpolarizabilities may be performed.

These changes are based on the third version of the Distributed Data Interface (DDI), supporting "node-replicated" data structures, stored in SystemV shared memory.

February 22, 2006:

The left eigenstate's method of moments perturbative triples correction,
CR-CCSD(T)_L is included. Analytic computation of Raman and HyperRaman
spectra for RHF is possible. Analytic gradients for the multi-layer
solvent model EFP + PCM have been implemented. The Fragment MO method
now can be used together with the PCM solvation model. An extensive
library of model core potentials and basis sets has been provided.

June 27, 2005:

ZAPT2 parallel analytic gradients for open shell ROHF references may now
be computed. A determinant based multireference perturbation theory program
is provided, using direct CI techniques. Charge transfer and disperson
terms are added to the EFP2 model, together with a more efficient exchange
repulsion gradient. All Effective Fragment Potential computations may
now be executed in parallel. The Coupled Cluster program permits the
computation of CCSD(TQ) energies, and generation of the density matrix
for CCSD or EOM-CCSD states. The Surface and Volume Polarization for
Electrostatics continuum solvation model is included. The Fragment
Molecular Orbital scheme has been generalized to a multi-level treatment
of different fragments, and extended to include MCSCF and Coupled Cluster
computations. Two families of systematic basis sets have been provided,
namely the Polarization Consistent and Correlation Consistent sets.

November 22, 2004:

Two new codes for faster AO two electron integral evaluation are available,
using rotated axis techniques to do s,p,d,L integrals, and precursor
Hermite transfer equations for other high angular momentum cases. The
accuracy of the TEI programs has been improved. The QFMM code has been
extended to computation of nuclear gradients. The Quadratic Force Field
approximation has been added to reduce time requirements for computation
of anharmonic frequencies by vibrational SCF.

May 19, 2004:

Analytic hessians for full active space MCSCF wavefunctions are now coded,
and enabled for scalable parallel calculation. EOM-CCSD excitation energies,
and novel triples corrections to these, may be obtained for RHF references.
The Polarizable Continuum Model (PCM) for solvent computations has been
generalized for use by all SCF wavefunctions and their DFT counterparts,
and now defaults to Conductor PCM with Area Scaling tesselation. The PCM
interface with the EFP model has also been extended to all SCF types. The
Fragment Molecular Orbital (FMO) method for the computation of very large
molecules by linked computations of its subunits is included, for RHF or DFT
energies and gradient, or MP2 energies.

This version of GAMESS is constructed on top of the second version of the Distributed Data Interface (DDI). The new DDI is optimized for SMP clusters, using SystemV memory calls to implement distributed memory operations inside nodes, with network messaging between nodes. Support for processor subgroups is also included in DDI version two. The installation procedure for GAMESS now has a new, separate step for the compilation of DDI. Some systems may require reconfiguration to permit large SystemV memory allocations.

December 12, 2003:

Parallel computation of the CIS energy and its gradient for RHF references is
possible. Relativistic quantum chemistry using a third order Douglas-Kroll
transformation for the inclusion of scalar effects is included, along with
the ability to compute Spin-Orbit Coupling with Model Core Potentials.
Numerical differentiation of any of the available energy values to obtain
the nuclear gradient or the nuclear hessian has been coded. An NMR program
for RHF wavefunctions is also included.

July 3, 2003:

A distributed memory parallel UMP2 energy and gradient code is now
available. The Polarizable Continuum Model (PCM) solvation model now
offers accurate nuclear gradients, allowing for PCM geometry searches.
Finally, Model Core Potential (MCP) integrals are added, permitting
energy calculations with correctly shaped valence orbitals.

January 14, 2003:

Occupationally Restricted Multiple Active Space (ORMAS) can be
used as the CI step of MCSCF orbital optimizations, with fewer
determinants than FORS (CAS-SCF).

June 20, 2002:

Closed shell reference coupled-cluster energies such as the standard
CCSD and CCSD(T) models, and the completely renormalized CC-SD(T)
model for bond breaking systems may be computed.

February 16, 2002:

The nuclear gradient for UMP2 wavefunctions may now be computed in serial
fashion. Determinant-based direct second order CI computation is also
available.

September 6, 2001:

The Optimal Parameter Quantum Fast Multipole Method is included for fast
RHF, ROHF, and UHF Fock builds in large molecules. The multi-reference
MCQDPT peturbation theory code has been enabled for parallel execution.
A general determinental CI program permitting arbitrary specification of
the space products has been added. A Jacobi rotation program for MCSCF
orbital optimization is also included.

June 25, 2001:

A grid-based Density Functional Theory (DFT) for energy and gradients
has been implemented. Spin-orbit coupling using MultiConfigurational
Quasi-Degenerate Perturbation Theory (SO-MCQDPT) version has also been
included. The RESC integrals can optionally use the uncontracted
primitive basis set during resolution of the identity steps.

October 25, 2000:

The determinant CI step during MCSCF calculations can now exploit Abelian
point group symmetry. The IEF solver for PCM calculations has been added
to the original BEM solver, for more accurate PCM gradients. The energy
can be computed for the solvation model of EFP explicit waters surrounded
by a PCM continuum. Raman intensities can be predicted.

March 25, 2000:

The effective fragment potential methodology is extended to permit modeling
of a system joined by covalent bonding to the ab initio region. Vibrational
anharmonic corrections (VSCF) may be obtained. The analytic gradient for
the RESC and NESC relativistic corrections has been programmed.

January 10, 2000:

A fully general spin-orbit coupling package (1 or 2 electron operator,
arbitrary spin states, any active space dimension) is included. Two
options for inclusion of other spin-independent relativistic effects are
implemented, namely Nakajima's RESC and Dyall's NESC schemes for elimination
of small components. Multireference perturbation theory permits computation
of reference function weight, and energy analysis.

June 6, 1999:

A highly scalable parallel MP2 gradient program, based on distributed
memory storage, is included. The open shell second order perturbation
energy correction known as ZAPT is implemented, and computations in a
pure spherical harmonic variational space are now possible.

This version introduces the Distributed Data Interface (DDI), a parallel library supporting distribution of large matrices across all CPUs. DDI replaces the TCGMSG library, with the DDI source codes distributed with the GAMESS program. As part of this change, GAMESS will build for parallel execution on all Unix systems, although the resulting binary can still be run on just one CPU if desired.

December 1, 1998:

The effective core potential (ECP) integrals are rewritten for speed,
extension to spdfg basis sets, and to include analytic hessian computation.
Automatic generation of Delocalized Coordinates (DLC) for large molecules
is provided.

May 6, 1998:

A direct implementation of full CI using a determinantal basis may be used
for the CI optimization step within MCSCF calculations.

January 6, 1998:

The full Breit-Pauli spin-orbit coupling operator may be used with a general
active space, for singlet-triplet couplings only.

March 18, 1997:

The Polarizable Continuum Model developed at the University of Pisa for
treatment of solvation effects is included for RHF and MCSCF wavefunctions,
allowing computation of the nuclear gradient and solution phase
polarizabilities.

October 31, 1996:

Multiconfiguration Quasidegenerate perturbation theory (MCQDPT) energy
corrections for MCSCF wavefunctions can be evaluated to second order.
The analytic gradient for CI wavefunctions based on RHF orbitals is
implemented.

September 11, 1996:

The Effective Fragment Potential model for the treatment of weak
intermolecular interactions is released. A standard EFP for the treatment
of aqueous solvent effects is built into the program.

June 22, 1996:

The quasi-Newton orbital optimizer is extended to treat MCSCF functions.
MP2 gradients now allow frozen core orbitals to be present.

November 22, 1995:

Analytic gradients for closed shell MP2 can be computed. CONOPT, a new
geometry search scheme for locating saddle points is implemented. Two
algorithms for tracing gradient extremals on the potential energy surface
are included. Morokuma decomposition is enhanced, in particular to allow
up to ten monomers.

July 26, 1995:

The quasi-Newton SCF (SOSCF) convergence procedure is extended to include
ROHF and all GVB wavefunctions. The DIIS option for GVB is also enhanced.

March 10, 1995:

An approximate second order SCF (SOSCF) method is used for convergence of
RHF wavefunctions. Analytic hessian computation now includes IR intensities
and the optional computation of the polarizability tensor. The Huckel guess
for ECP basis sets is improved.

February 1, 1995:

The Morokuma decomposition of dimer interaction energies is implemented.
The spin-orbit coupling code now forms and diagonalizes the spin-orbit
Hamiltonian matrix, yielding total energy levels. There is also an option
for a simplistic scan of a potential energy surface.

November 17, 1994:

The Dynamic Reaction Coordinate, which is a classical trajectory on
the ab initio potential surface, can be computed.

August 11, 1994:

Time dependent Hartree-Fock is added, permitting the analytic computation
of the frequency dependent polarizability, and first and second
hyperpolarizabilities, for closed shell wavefunctions. These relate to
many interesting NLO properties, including the electro-optic Pockels
effect.

July 22, 1994:

A spdfg gradient package replaces the former analytic gradient integrals,
and runs 3-5 times faster than the former code. The localized charge
decomposition (LCD) model permits analysis of energy contributions from
each Ruedenberg type localized orbital.

Older versions going back to 1984 can be inferred by looking at the source code file GAMESS.SRC, for the string "new date in box in honor of ..." in the change history.