Gaussian 26 Release Notes
- New Capabilities
- Improvements
- Changes to input and output
- Extensions for GauOpen
- Improvements to Utilities
New Dynamics Code/Capabilities:
- These provide consistent input and use of initial conditions, thermostats, and post-processing across different types of dynamics.
- Real-Time dependent Hartree-Fock and DFT, which propagates the electronic degrees of freedom in accord with the time-dependent Schroedinger equation, with fixed nuclei.
- Ehrenfest dynamics, in which the electronic degrees of freedom are propagated using the time-dependent Schroedinger equation at the HF or DFT level. By default, propagation of nuclei is treated classically.
- Path Integral Molecular Dynamics, including Ring Polymer Molecular Dynamics (RPMD), is available. This treats the propagation of the nuclei quantum-mechanically. This can be combined with either the Born-Oppenheimer treatment of the electronic degrees of freedom or with Ehrenfest propagation of the electronic wavefunction.
- Non-adiabatic dynamics using surface hopping, Ehrenfest, or coherent switching with decay of mixing (CSDM) methods are available for two-state systems. Nuclei can be treated either as classical particles or as ring polymers for nuclear quantum effects using independent bead approximation.”
- Dynamics trajectories can be distributed among CPUs and/or nodes in a cluster.
General Internal Coordinates:
General Internal Coordinates (GICs) are useful in all types of geometry optimization and are the default:
- By default overall coordinates for translation and rotation of disjoint fragments are used.
- For larger molecules, coordinates for overall rotation of XY3 groups are used.
- Geom=GICOneSymm, Geom=GICAllSymm request symmetrized linear combinations of the default coordinates. The former includes only combinations in the totally symmetric representation of the point group and hence constrains the symmetry, while the latter includes a full set of coordinates.
- Geom=Modify allows very general specification of GICs.
- Refer to $GAUSS_EXEDIR/g26/doc/GICKeywords.pdf for a full list of predefined coordinates and other input.
Interface With Tinker:
- There is an interface with the Tinker program for QM/MM dynamics using the Amoeba polarizable force field. Tinker controls the dynamics and calculates the bonded and van der Waals contributions to the energy and force and runs Gaussian to solve for the classical and quantum charge interactions and their contribution to the energy and forces.
EOM/CCSD Additions:
- Gradients of EOM-CCSD excited state energies are computed analytically.
- Ionization potentials can be computed using EOM-CCSD, for both closed and open-shell systems.
CASSCF Calculations:
- CASSCF calculations can be conducted allowing the CI coeficients to be complex and the orbitals to be either complex spin-aligned or complex mixtures of alpha and beta spins. Energies and analytic gradients can be computed, but frequencies are numerical. One-electron spin-orbit terms can be included self-consistently with GCASSCF.
- The on-top DFT (PDFT) dynamic correlation correction to CASSCF is available. It is requested by the PDFT option to CAS: CASSCF(…,PDFT) requests PDFT using the default BLYP correlation functional. CASSCF(…,PDFT=N) requests PDFT using XC functional number N. This is the same number set in option 74 of overlay 3 for that functional in a regular DFT calculation. The Int=Grid=… option can be used to change the quadrature grid from the default of UltraFine.
Scan Calculations:
- Scan=List automates performing energy, gradient, or frequency calculations on an arbitrary list of geometries. Scan=Addpoints adds additional points to a previous Scan=List. Geom=Input can be used to select points to be included from checkpoint files from previous optimizations, relaxed scans, IRCs, or dynamics trajectories.
Semi Direct CIS(D) And TDA(D):
- CIS(D) and TDA(D) corrections are now evaluated using semi-direct procedures. This extends the use of double-hybrid functionals to excited states.
NMR Shieldings With Double Hybrid Functionals:
- NMR shieldings can now be computed with double-hybrid functionals.
Nuclear Electronic Orbitals:
Nuclear Electron Orbitals (NEOs) can be used to treat hydrogen nuclei quantum-mechanically at the SCF and DFT levels of theory (including XC terms for the proton-electron cusp). By default, all protons are quantum. If a general NEO basis is read in, those protons with basis functions are quantum. Energies and forces are available options:
- ProtSP; put a single S=P shell on each quantum proton.
- [PB4D,PB4F1,PB4F2,PB5D,PB5F,PB5G,PB6D,PB6F,PB6G,PB6H]: put the named basis on quantum protons.
- Gen: read in a general basis input section giving the basis on each quantum proton.
- Chk: Read the NEO basis from the chk file.
- NShell=N: put N even-tempered basis functions on each proton.
- LMax=L: The even-tempered basis set will go up to angular momentum L. The default is 0 (S functions only).
- EStart=E: The even-tempered basis set will have the tightest exponent be E. The default is 24.0.
- EMiddle=E: The even-tempered basis set will have equal numbers of functions tighter and more diffuse than E.
- ERatio=R: The even-tempered basis will have ratio R between consecutive exponents. The default is Sqrt(2).
- ERatSq=R: The even-tempered basis will have ratio Sqrt(R) between consecutive exponents.
- Protons: Give each quantum proton the mass of a hydrogen nucleus, regardless of the atomic weight given or defaulted on its atom specification line.
- Deuterons: Give each quantum proton the mass of a deuterium nucleus, regardless of the atomic weight given or defaulted on its atom specification line. Note that the quantum nuclei are still treated as fermions.
- Tritons: Give each quantum proton the mass of a tritium nucleus, regardless of the atomic weight given or defaulted on its atom specification line.
- [LSDA,EPC17]: Requests that the EPC17 local electron-nuclear cusp functional be used.
- [LYP,EPC19]: Requests that the EPC19 GGA electron-nuclear cusp functional be used.
- Grid=grid-specification: Specifies the quadrature grid to be used when evaluating the cusp functional. The default is to use the same grid as the electronic XC quadrature (which defaults to Grid=UltraFine).
Structural Changes:
- Gaussian 26 is a single executable, and many utilities are soft links to the same file.
- The %nproc, %cpus, %workers and %gpucpu link0 commands are no longer available and the equivalent command-line switches -p, -c, -w and -g must be used instead. G16 input files will need to have these removed and the corresponding command-line used instead.
Changes to optimizations:
- Opt=NoQMMMLinear disables the combined QM/MM linear search during ONIOM(MO:MM) optimizations with mechanical embedding.
- Opt=MaxEver=N sets the maximum number of iterations that can be done in an optimization even using Opt=Restart to increase the number of steps. This can be useful to limit memory usage during optimization of very large systems.
- Opt=MaxStep=N now defaults to 0.1 for excited state optimizations, 0.3 for ground state optimizations.
- Opt=NGoUp=N, the number of steps which raise the energy during a minimization before doing only linear searches, defaults to 1 (linear search after the second rise in energy) except for ONIOM and excited state minimizations, for which it is -1 (linear searches only after a step which raises the energy).
- Opt=NGoDownHill=N, the number of Hessian eigenvectors having negative eigenvalues to mix when stepping away from a saddle point, now defaults to 1 for excited state minimizations, 3 for ground state minimizations. NoDownHill always takes an RFO step and does not try to step away from saddle points.
- Opt=MaxEStep=N, the maximum step to take when moving away from a saddle point, defaults to a maximum change in any internal coordinate of 0.1 atomic units. N>0 makes this limit N/1000 atomic units. N<0 makes the limit |N|/1000 atomic units in the length of the overall step, so N=-600 corresponds to the Gaussian 16 default.
- Opt=NoStop causes a relaxed potential surface scan to continue the scan even if the optimization at a scan point does not converge.
New or changed options for the Opt keyword for harmonic constraints with respect to the initial structure:
- InitHarm=N/ChkHarm=N now Uses N/1000000 for the force constant rather than N/1000 as done in G16.
- ReadFCHarm causes Cartesian coordinates for the constraint structure to be read from input, followed by force constants for each Cartesian coordinate.
- ReadAtHarm=N requests constraints with respect to the input structure but reads an input section with a list of the atoms to be constrained. The constraining force constants are all set to N/(1000000*NAtCon), where NAtCon is the number of atoms which are constrained.
- NoHarm turns off the default reading of constraints from the chk file when doing Geom=AllCheck.
Changes to PCM:
- The rotation of the molecular geometry between input and standard orientation is now applied also to the spherical grids and polyhedra used to discretize the PCM cavity. This ensures a consistent relative orientation of the molecular geometry and the discretized PCM cavity during optimizations, IRC, etc.. The additional PCM keywords NoRotateGrids and NoRotatePolyhedra restore the G16 behavior.
- The normal vectors to the discretized PCM cavity surface are now smoothly rotated in the vicinity of the intersection of two (or multiple) spheres so that they approximately align to the normal direction to the actual SES (Solvent Excluded Surface). Additionally, the normal vectors are also smoothly damped (switched off) as they get close to nearby intersecting spheres. This is similar to the smooth switching off of the integration weights. The additional keyworks NoRotateNormalVectors and DampNormalVectors=No restore the G16 behavior.
More robust parser:
- A new more robust parser is used for input to vibro-rotational analysis, anharmonic vibrational calculations, and excited state Franck-Condon/Herzberg-Teller analysis.
Changes to Vibro-Rotational analysis:
- New default behavior for the construction of the cylindrical axis Z. The cylindrical axis Z for the computation of rotational constants is now set to be the same as the Z axis used by Gaussian. This can be changed to Gaussian’s default with Print=ZAxisSymm=Auto where the Z axis was chosen based on the inertia moment to get the closest-match symmetric-top representation.
- There is a new generator of input for SPFIT/SPCAT for harmonic and anharmonic vibro-rotational analysis.
- There are new input options for harmonic vibro-rotational analysis. Users can now choose the top representation, the normal mode ordering (ascending or descending, by irrep or not), the representation of the wave function for systems with degenerate modes (spherical vs Cartesian), or provide a constant scaling factor to be used for the vibro-rotational analysis. The harmonic rotational analysis can be requested via Freq=VibRot or Freq=ReadVibRot if options are to be read from an input section.
Changes to anharmonic vibrational calculations:
- Improved automatic identification of resonances, especially for the intensities. In some conditions, the scheme used in G16 could fail to identify resonances between closely lying bands, potentially resulting in excessively intense bands or in unexpected intensity redistributions.
- New state representation when the variational correction in GVPT2 is too large to maintain a direct correspondence between harmonic and anharmonic states. Gaussian now prints a description of the states, numbered sequentially, and uses it as reference. Besides avoiding the risk of misleading users on the nature of the anharmonic states, this new behavior fixes a problem where Gaussian could fail to assign states after variational correction, resulting in an error termination. Users can deactivate this feature with Print=NoVarStates, but this can lead to error terminations.
- Added support of Wang-Landau density of states to compute the vibrational partition function instead of simple perturbation theory (SPT), providing more accurate thermodynamic quantities. The construction of the density of states can be relatively long, so SPT remains the default.
- Selective deactivation of anharmonic contributions in intensities. Users can now include only mechanical or electrical (property) anharmonicity and see the effect on the band-shape.
- Vibrational averaging of the ground to excited state non-adiabatic coupling is done during anharmonic excited-state frequency calculations.
- Reporting on the consistency checks performed on the anharmonic constants has been made clearer. The new table provides more useful information on the magnitude of the inconsistent data to assess the reliability of the working data.
- Reporting on the resonance analysis has been more complete. Users can now see the results from the different checks, even with resonances added in input, and which resonances will be included or ignore (passive/inactive modes).
- There is new output for the variational corrections in GVPT2. Gaussian now groups interconnected states as polyads and show the resulting matrix for the variational correction.
- Chirality is now preserved when a different top representation is used than that of the standard orientation.
Changes to vibronic spectra:
- For large-amplitude modes, the mode can be projected onto a user-specified General Internal Coordinate.
- Computation of one-photon (OPA, OPE, ECD, CPL) and resonance Raman spectra can be done starting from selected hot bands. Users can now provide in input stream a list of initial states and choose the weight to be applied to each state.
- There are additional integration grids for the path integral formalism (TD). The supported grid specifications are: Rectangle (default for one-photon) and Gauss-Legendre (default for resonance Raman).
- Automatic generation of multiple resonance Raman or ROA spectra over a range of incident light frequencies and/or damping constants. Gaussian will construct all possible combinations and generate the intensities or Raman tensors for each one. The ranges can be automatically built or based on specific lists.
- Users can request that the superposition of the ground and excited state structures be carried out over a specific set of atoms provided in the input stream rather than all atoms.
- The states of interest in the sum-over-states formulation of resonance Raman can be controlled by the user, The minimum and maximum number of excited modes, the number of quanta per mode and the total number of quanta can be specified.
- There is an improved method to compute the spectral progression of chiroptical spectra after each class. The spectral progression is estimated to ensure a smooth evolution and avoids possible large variations from sign changes over the first classes. The actual progression is shown at the end.
- A new warning message is printed whenever the printed spectrum is below 99%. This message is in addition to the existing message when less than 80% of the spectrum was shown.
- An overflow which could occur during prescreening in systems having more than 600 normal modes was fixed.
- Rare cases of chiroptical spectra using FCHT could invert the sign(s) if FC and HT contributions had opposite signs were fixed.
CIS/TD/TDA(D):
- CIS=Read and TD=Read can now read amplitudes when the basis set changes.
- Optimizations, anharmonic frequencies, IRCs, and dynamics on excited states now try to match the state of interest at a new geometry to that at the previous geometry. By default, if no state at the new geometry has a good overlap with the state of interest at the previous geometry, the job terminates with an error message. This can be controlled with options to the CIS or TD keyword:
- NoFollow turns off the attempt to match states during the current job.
- NoMatchFail attempts to match states but proceeds with the new state of interest having the largest overlap with the old state even if that overlap is small.
- MatchThresh=N sets the minimum acceptable overlap of new and old states to N/1000. The default is N=500 (0.5)
- NoSwap turns off matching of states at the initial geometry when doing CIS=Read or TD=Read.
- New options for CIS(D) and TD(D):
- ReadDStates causes a list of states for which the (D) correction is calculated. The default is to do (D) for all excited states done in the CIS/TD calculation.
- Semidirect=N to specify at most N occupieds per pass in both the ground state E2 calculation and the (D) calculation on the excited states, analogous to MP2=Semidirect=N when doing just the ground state calculation.
Improvements to ONIOM:
- Disjoint model systems are now supported. ONIOM=Fragment=N specifies the number of fragments, which are specified on the atom specification input lines as for counterpoise or Guess=fragment calculations. Separate model system calculations are done for each fragment of each model system. The low-level calculation on the real system is always done with all atoms.
- External(UseGaussian,…) requests that Gaussian be run as the external program. IOBAF is implied. This can be used to automate some unusual calculations which are not automated within one Gaussian calculation, for example doing Stable=opt in each ONIOM sub-calculation.
CPHF=Gamma:
- Polarizabilities, harmonic and anharmonic vibrational Raman and ROA spectra and resonant Raman and ROA spectra can be computed with a broadening term (for the instrument) included. CPHF=Gamma=NNNN adds an (imaginary) broadening term with magnitude NNNN/10000 atomic units to the polarizability part of second and third-order property calculations.
Additional density functionals:
A complete list of new exchange and correlation functionals is in the appendix below. Here we list some added DFT model chemistries which are of particular interest.
- M11Plus is a Rung 3.5 functional which has good accuracy across a wider range of properties than most functionals.
There are additional or improved dispersion corrections (options to the EmpiricalDispersion keyword) available:
- GD4 requests Grimme’s 4th dispersion correction. Analytic second derivatives are available.
- The Smoother option can be combined with GD3 or GD4 to request the more continuous versions. Analytic second derivatives are available.
- 3Body or No3Body can be combined with GD3 or GD4 to include or exclude the 3-body terms. The default is to include them with GD4 but not with GD3. Analytic second derivatives are available.
- XDM requests that dispersion correction. Analytic gradients are available but not second derivatives.
- VV10 requests that dispersion correction. Analytic gradients are available but not second derivatives.
- None (or NoEmpiricalDispersion) turns off VV10 from models that default to including it.
Addtional model chemistries:
- The HF3c, PBEh3c, B973c, R2Scan3C, and wB97x3c models of Grimme are available. These imply the standard basis set for each model, but these basis sets (Minix, mSVP, mTZVP, mTZVPP, and vDZP, respectively) have keywords and can be combined with other methods.
Additional basis sets:
- The maSVP, maTZVPP, maTZVP, maQZVPP, maQZVP AO basis sets of Truhlar.
- cc-pV[D,T,Q,5]-PP AO basis sets.
- Dummybasis puts a single s function on each atom. This is useful in setting up systems when a QM calculation will not be done.
- The CRENBL basis sets and associated ECPs.
- The TZVPfit density fitting set.
- Various NEO proton basis sets, described in the section on NEO.
- Some ECPs now include spin-orbit.
GPU support:
- There is now support for X86_64 CPUs with Volta, Ampere, Hopper and Blackwell GPUs.
X2C relativistic model:
- The X2C two-component relativistic model is available. Like DKHSO, this includes one-electron spin-orbit when doing GHF or GKS in the SCF. By default, Boettger’s scaling factors are used for the spin-orbit terms with either X2C or DKHSO.
Changes to SCF procedures and defaults:
- SCF=XQC is the default except for PBC, ROHF, GHF/GKS, and NEO calculations. SCF=NoXQC reverts to using only the regular SCF.
- The regular SCF (l502) saves the density corresponding to the lowest energy so far for the subsequent QC stop and SCF=Restart.
- The regular SCF now checks for convergence problems and moves on to the QCSCF or aborts early if the iterations are not progressing. This includes checking for repeated iterations which fail to lower the energy significantly.
There is an additional combination of SCF strategies available. The full set of strategies is:
- SCF=NoXQC (the default in G16): Do regular SCF and terminate if it does not converge.
- SCF=XQC (the default): Try to converge using the regular SCF and if this doesn’t work, do a combination of linear searches and quadratic steps.
- SCF=YQC: Initially do linear searches as in SCF=QC but when the rotation gradient is small switch to regular SCF instead of quadratic steps. If the regular SCF does not converge, switch to quadratic steps.
- SCF=ZQC: Start with the regular SCF. If this does not converge, then switch to linear searches as in XQC but when these have reduced the rotation gradient switch back to regular SCF. If this fails, switch to quadratic steps.
- The default number of regular SCF cycles (SCF=MaxConventional=N) is 48 when XQC, YQC, or ZQC is being done.
- SCF=NonVariational turns off checks in the regular SCF for multiple iterations failing to lower the energy. It is needed for unusual models in which the SCF is not variational, such as some solvation models with external iterations.
- SCF=Average=(N,M) do an SCF in which the N highest energy electrons are averaged over M orbitals. The resulting wavefunction can be used for gradients but not second derivatives. This is primarily useful when the SCF results will be fed into a larger embedding model, and may also be useful in some cases to generate initial orbitals for a CAS. The choice of which initial guess orbitals are selected for averaging is done in the same way as the selection of active space orbitals for a CAS(N,M).
- SCF=JustEnergy computes the Fock matrix and energy without iterating during a regular SCF
- CAS(…,OnlyCI) does the active-space CI using the initial orbitals and computes the energy and exits without iterating or updating the orbitals. CAS(…,OnlyCIDen) does the active-space CI using the initial orbitals, computes the energy, forms the one and two particle density matrices and exits without iterating or updating the orbitals.
Changes to the External interface and keyword:
- InFAF or InBAF to request that a faf or baf be provided as input to the external program.
- IOFAF or IOBAF to provide a faf/baf as input to the external program and to recover results from a faf/baf written by it.
- Any text written by the external program to the output or message files named on the command line generated for the external program is copied to the Gaussian log file.
- If the string giving the command to run specified with the External keyword starts with a #-sign, G26 is run as the external program and IOBAF is implied. The route line passed to the external calculation is the string provided by the user with ChkBasisGeom=AllCheck InExternal appended. The first two keywords ensure that the structure and basis set are those set up in the controlling job (for the current model and sub-system for ONIOM jobs) and the third turns off various consistency checks on the input data which Gaussian normally performs but which would fail for a ONIOM model system or Counterpoise sub-calculation.
- The ReadInputSection option tells Gaussian to read an input section to provide as the input file for the external program, instead of creating the default file with atomic numbers, coordinates, etc.
ECD rotational strength:
- Electronic Circular Dichroism rotational strengths can be computed with giaos.
Real to complex stability testing:
- Real->Complex stability can be tested for DFT as well as HF.
Higher order point groups:
- Higher-order non-crystallographic point groups (i.e., C30) are supported.
Link 608, which evaluates a list of functionals given a density, has been extended:
- The input is more general and allows for range-separation, Rung 3.5 functionals and functionals requiring 6d quadrature.
- It can take input from the chk file (i.e. one created by the -ib command line switch and hence originally in a faf/baf) and it stores its results on disk for inclusion in an output baf.
- Both PCM solvation and VV10 correlation can be included.
Appendix 1: New DFT exchange and correlation functionals
- Additional exchange functionals include M11 exchange with parametrized R35 term and the exchange parts of the wB97X-D3, wB97X-V, B97M-V, wB97M-V, wB97M(2), revPBE, SCAN, r1SCAN, r2SCAN, revSCANx, r1SCAN-14, r2SCAN-14, CF22D, modified PBEc (for PBEh-3c), and B97-3c functionals.
- Additional correlation functionals include M11 correlation with parametrized R35 term, GAMc, LYP+R3.5, the correlation parts of revM06-L, revM06, revM11, wB97X-D3, wB97X-V, B97M-V, wB97M-V, wB97M(2), SCAN, r1SCAN, r2SCAN, revSCAN, r1SCANc-14, r2SCANc-14, modified PBEc (for PBEh-3c), and B97-3c, and local and non-local functionals for the electron-nuclear cusp.
- Additional hybrid and double-hybrid functionals, including m11plus, m11po, m11pz, coach, wb97m-d3bj, wb97x-d3bj, b97m-d3bj, b3v5lyp, revdsd-pbep86-d4, revdsd-pbep86-d3, wb97x3c, r2scanxc3c, b973c, pbeh3c, hf3c, cf22d, r2scan0xc14, r2scan0xc, r2scanxc14, r1scanxc14, revscanxc. r2scanxc, r1scanxc, scanxc, scs-mp2, xygjos, xyg3, mpw2plyp-d3, revtpss1revtpss, tpss1tpss, mpw1b95, b1p86, rsx-qidh, rsx-0dh, wb97m(2), wb97m-v, b97m-v, wb97x-v, wb97x-d3, vv10, revm11, revm06, revm06-l, and gamxc.
New or notable command line (and Default.Route) options:
- -m=N. Sets the amount of dynamic memory allocated by the job (equivalent to the %mem link0 command). If N is an integer it is in units of real*8 words, but it can also have a unit attached, e.g. -m=1000mb for 1000 megabytes or equivalently -m=125mw for 125 megawords. kb, kw, gb, gw, tb, and tw can also be used.
- -c=list of processors gives a list of SMP processors to run on, e.g. -c=”3,6-9″ would run on processors 3, 6, 7, 8, and 9. On systems which cannot bind threads to CPUs this is equivalent to -p=5. It is usually necessary to quote the list if ranges are used to avoid attempts by the shell to expand the string.
- -p=N says to run across N threads without attempting to bind threads to specific CPUs.
- -g=list=list gives a list of GPUs to use and a list of the CPUs to bind to each GPU. For example, -c=”0-9” -g=”0-1,4=8,3,9″ would use 10 CPUs and 3 GPUs, with GPU 0 being used by the thread on CPU 8, GPU 1 used by the thread on CPU 3, and GPU 4 used by the thead on CPU 9.
- -ic=cfile: copy the contents of checkpoint file cfile to the jobs’s chk file (equivalent to the %oldchk link0 line).
- The -ib, -ob, -baf, -x, and -rr switches described above in connection with the GauOpen interface. If both -ib and -ic are used, the chk file is copied before the baf, so data in the baf overrides corresponding data from the chk file.
- -oc=cfile sets the job (output) checkpoint file name to cfile. This is used by all steps in the job unless there is a %chk command in the input for a later job step.
- -ob=file/ -of=file/ -ob4=file/ -of4=files requests that the named baf, faf, baf with 4-byte integers, or faf with 4-byte integers, respectively, be written at the end of each step in the input file. Compound job steps such as Opt freq or Cbs-qb3 write the file once.
- -w=list gives a list of node names to use for distributed parallelization. Gaussian 26 must have been built for Linda parallel processing.
- -lf=string provides a string of switches which are passed to ntsnet when running with cluster based parallelism, e.g. -lf=”-vv” to turn on debug print when starting the Linda workers.
- -z=rfile/ -zn=rfile. Names the rwf file to be used (equivalent to the %rwf link0 command). With -z the file is retained at the end of the job; -zn causes the named file to be deleted at the end of a job which completes but retained for restarting if the job is killed or fails.
- -fchk=fname/ -fck=fname/ -baf=fname/ -faf=fname. Requests that the named file be written each time the job checkpoints (i.e., at each step of a geometry optimization). Normally the same file is over-written each time, so if the job finishes normally the final file is the same as would be left by -oc, -ob, etc. However, this switch allows a controlling program to monitor the progress of a job which works on multiple geometries. In addition, it causes relaxed potential surface scans to leave behind a separate file (with a name derived from fname) for each converged point during the scan.
Molecule specification enhancements:
- Geom=Quiet turns off print of input which is NAtoms in length. Useful when running Gaussian on systems with a very large number of atoms, e.g. when doing QM/MM dynamics with Tinker.
- Geom=ReadFrag causes an input section to be read assigning atoms to fragments, over-riding what was read from a chk file or specified on atom specification lines or defaulted based on connectivity.
- Geom=NoDefFrag turns off assignment of atoms to fragments based on the connectivity (the default if fragments were not specified on the atom specification lines).
- Geom=NoReadInternal over-rides the default of reading internal coordinates from the chk file when doing Geom=AllCheck.
- The charge and multiplicity specification for the molecule, oniom layers, or counterpoise fragments can now have a negative multiplicity, which implies more beta spin than alpha-spin electrons. For example 0,3 means charge 0 and multiplicity 3 with 2 more alpha than beta electrons, while 0,-3 means charge 0, multiplicity 3 with 2 more beta than alpha electrons.
- For ONIOM calculations with disjoint model systems, the charge and multiplicity can be specified for each ONIOM system and then the charge and multiplicity for each fragment, separated by a slash.
- There are addition possible specifications with Geom=input when reading from chk files from optimizations or IRCs:
- Step=Original for the first geometry done.
- IRCForward=N for the Nth point in the forward direction.
- IRCReverse=N for the Nth point in the reverse direction.
- IRCTab=N for the Nth point in the table printed at the end of the IRC.
IRC changes
- IRC=Report causes internal coordinates to be generated (if they were not read in) and their values reported for each point in the summary table. Use Geom=Modify or Geom=ReadAllGIC to add coordinates to report or to specify exactly which coordinates to report.
- IRC=KeepTrRot to turn off projection of overall translations and rotations along the IRC.
Post-SCF changes:
- ModifyWindow option for the Window keyword and post-SCF method keywords. This sets up a type of frozen-core or full window as specified or defaulted (by Full, FC, FC1, etc.) but first an input section is read in which changes the number of core orbitals frozen for selected elements or atoms. For example
Ga=5 2=0
would freeze 5 orbitals (10 electrons) on all Gallium atoms and would not freeze any core orbitals on atom 2. - The Tight option for CIS, TD, CCSD, BD, and EOMCCSD causes single-point calculations with these methods to use the same tighter convergence threshold as is used during geometry optimizations.
- The TWInCore and NoTWInCore options for CCSD, BD, and EOMCCSD control whether the amplitudes and product vectors are kept in memory. By default, this is done if there is sufficient memory provided. TWIncore will cause the job to terminate if not enough memory was provided.
MM options:
- By default MM input is checked to make sure that any atoms having non-zero MM charges also have van der Waals repulsive terms and that the connectivity which is provided or defaulted implies reasonable valencies for all atoms. Several options to the MM keywords (UFF, Amber, etc.) modify this behavior:
- NoCheck turns off error termination if either of these tests fail. Instead, warnings are printed.
- NoCheckVDW turns the test for vdW terms on charged atoms into a warning.
- NoCheckConnectivity turns the test for reasonable connectivity into a warning.
- TightCheckConnectivity makes the check on implied valencies more restrictive. This can be useful if all the bond orders provided in the connectivity are correct (i.e. put in carefully when a structure is drawn in GaussView), but often gives errors if the bond orders are guessed based on distances by either GaussView or Gaussian.
- CheckAllVDW and WarnAllVDW cause all atoms to be checked for vdW parameters even they are uncharged and generates an error or a warning, respectively, if parameters are missing.
- The Amoeba keyword specifies that this polarizable MM model is to be used. Gaussian performs only the classical and quantum electrostatic part of the Amoeba — this is for use along with the interface to Tinker, which is assumed to handle both the bonded terms in the force field and to propagate the dynamics. The PMM option to the Amoeba keyword controls details of the polarizable MM electrostatics.
Population analysis options:
- Pop=Quiet turns of all print in links 601 and l602 (Population and Property evaluation), but still computes atomic charges and properties which are stored on the chk file and in a baf if the latter is written.
- Pop=FitFrag and Pop=ReadFrag requests that atomic charges fit to electrostatic potentials (Pop=MK,CHelpG,etc.) constrain the total charge on each disjoint fragment be constrained to the specified overall charge on that fragment. FitFrag uses the assignment of atoms to fragments specified in the atom specification input or defaulted based on the connectivity. ReadFrag reads an input section which lists the atoms in each fragment to be used in the fit (with fragments separated by a /), followed by an input section listing the total charge to be constrained for each fragment.
- Pop=DCT requests Adamo’s charge-transfer diagnostic for whichever excited state densities are requested via the Density keyword.
- Pop=Grid=quadrature-grid-name specifies the quadrature grid used in analysis procedures which require a 3d quadrature, such as Pop=Hirshfeld, HGLY, or DCT. The default is the grid set for general use via the Grid keyword, or the default of UltraFine.
- Pop=NBO8, Pop=NBO8Read to run the NBO version 8 (not provided with Gaussian) via the external interface. The corresponding NBO6 and NBO7 options work as in Gaussian 16.
AllQuiet keyword:
- AllQuiet. This keyword sets quiet (terse) mode everywhere and is equivalent to Geom=Quiet Pop=Quiet NoArchive as well as NoDiagFull for a frequency job. It removes most print which scales with the number of atoms and is useful for keeping the output short for calculations involving large numbers of atoms where the job is being run in order to produce results in a chk file or baf.
Initial guess options:
- Guess=CopyChk causes the MOs and/or density to be copied from the chk file without any projection or renormalization. This is useful when a density generated externally should be passed as-is to the SCF and other steps.
- Guess=(CopyChk,NoSCF) copies the MOs and/or density from the chk file and do not do any SCF. This is useful in computing properties from a specific density or MO 2e integrals from a specific set of orbitals provided via a baf. Note that when forming a Fock matrix from an un-normalized density it will be necessary to disable some of the usual usual consistency checks with Int=NoXCTest and if the density (normalized or not) does not come from the MOs with unit occupations to disable the use of MOs during DFT quadrature with Int=NoXCMO.
- Guess=Dummy indicates that a dummy guess — dummy matrices of the correct size — are to be stored for the orbitals and density. This is not useful on the route line but is sometimes useful with Guess=Input.
- Guess=None skips generating or storing an initial guess. This is not useful on the route line but can sometimes be useful with Guess=Input.
- Guess=MOLCASCards,NWCards reads guess orbitals from input as with Guess=Cards but assumes the order of basis functions is that used by the specified program.
PBC options:
- PBC=NRecip=N requests approximately N k-points between 0 and pi.
Data for Antechamber:
- Prop=Antechamber does Merz-Kollman fitting of the electrostatic potential and writes the corresponding data file for the Antechamber fitting code, which is part of the Amber suite of programs.
Changes to NMR:
- NMR=AICD and NMR=ReadAICD does an NMR shielding calculation using the CSGT model and writes the results to an input file for the AICD program. Since this is also referred to as ACID, the synonyms ACID and ReadACID are also accepted. ReadAICD causes a list of the occupied orbitals to include; the default is to include all orbitals.
- NMR calculations with MP2 or double-hybrid functionals now default to including all orbitals rather than frozen-core.
Additional options for the Integral keyword:
- Int=ExtraMem=N requests that the amount of memory normally reserved for evaluation of 2e integrals (which is estimated given the basis set and order of derivatives) be increased by N words, if N>0. If N<0 then |N| words are reserved regardless of the basis or derivative
- Int=NoTestBasis turns off the check for a suspicious basis (QM atoms with no basis functions, atoms with too few basis functions for even the core electrons, etc.)
- Int=NoXCMO turns off use of MO coeficients when computing the density during DFT XC quadrature. By default, the MOs are used during the SCF for those grid points for which this is less expensive than using the density matrix. This option is needed when supplying a density which does not arise from unit occupation of the occupied MOs.
- Int=ThrOvl=N sets the initial threshold on singular values of the overlap matrix to retain when producing a linearly independent sst of basis functions to 10^-N. The threshold may be made larger if the initial set fails various numerical tests. The default is N=6 (10^-6).
Counterpoise keyword:
- Counterpoise=[Only]InputFiles requests that input files for each sub-calculation of a counterpoise calculation be printed in the output file. OnlyInput just prints the input files and does not do any calculations.
Additional options for excited state energy transfer (EET):
- EET=Densities runs the EET sub-calculations but stops before computing the couplings.
- EET=Couplings computes the couplings from information on the chk file from a prior EET=Densities calculation.
- EET=Read reads an input section giving the fragments whose excited states whose couplings are to be computed.
- EET=DoK/NoK/DoXC/NoXC specifies whether or not to include exact exchange and/or DFT XC exchange-correlation in the couplings.
Fragment input:
- When an input section is read for a list of atoms, the specification and now include Fragments=i,j,… to include all atoms in the fragments listed, or NotFragments=i,j,… to exclude all atoms in the listed fragments.
The interface files are now called Binary Array Files and Fortran Array Files, rather than raw and Fortran matrix element files. The new names make the unique characteristics of the format clearer. The two file formats are used in the same way by Gaussian and are generally referred to as baf/faf below. The discussion below focuses on the capabilities added to Gaussian 26 to generate and use these files.
These files can be used to provide input to Gaussian and receive output from Gaussian, removing the need for other programs to write text input files or to parse output files. This is done in a way which provides for upward compatibility, so that programs which use this interfacen to Gaussian 26 will continue to work with future versions.
Both bafs and fafs written by Gaussian 26 default to using 8-byte integers on machines where G26 uses 8-byte integers internally, which is almost every platform. Files using 4-byte integers are generated if requested and can also be used as input to Gaussian.
Gaussian 26 command-line switches related to bafs:
- -IB=filename, -IF=filename specify a baf/faf as input (similar to -IC=filename to specify an existing chk file to copy to the current chk file). Gaussian can determine whether a file is raw or fortran unformatted and whether it has 4 or 8 byte integers by looking at the file, so the only difference between -IB and -IF is that the former defaults the extension to .baf and the latter defaults to .faf. If both -IB and -IC are used, then the contents of the input chk file are copied first, so data in the baf/faf overrides corresponding data from the input chk file.
- -[OB,OB4,OF,OF4]=filename. These name a baf/faf to be written at the conclusion of each job step. -OB/ -OF defaults to 8-byte integers except on machines which have only 4-byte integers.
- -baf=name/ -faf=name are similar to -OB/-OF except that the file is written at each point of a geometry optimiation, and a relaxed scan writes a separate file with a distinct modification of name for each optimized point.
- -x=“route-string” specifies the route line to be used. No route section is read from the input file.
- -rr — the route is taken from the checkpoint file (and typically has been imported there from a baf/faf specified by -IB). It is also not an error if the regular input file does not exist. Normally the route will include Geom=Allcheck so that all input is taken from the input baf/faf/chk file.
Additions to data stored in a baf and the output keyword:
- 8-byte integers are the default for both bafs and fafs.
- Complex, anti-symmetric, and anti-Hermitian matrices are supported in the bafs/fafs and are stored there by Gaussian when appropriate. Refer to the GauOpen documentation for details of how these are stored.
- Strings and arrays of strings are supported in bafs/fafs and string objects are stored by Gaussian for route, title, and names of representations of the point group. Refer to the GauOpen documentation for details on how string objects are stored.
- The per-point information from geometry optimizations, IRCs, dynamics, and excited state properties is now split up into separately labelled arrays (coordinates, forces, etc.)
- Density fitting and NEO proton basis sets are included if used.
- Parameters and results from PBC calculations are now included.
- MCSCF active space orbitals and the active space two-particle density matrix are included.
- If the DFT XC quadrature grid was stored on disk using the Int=PCXCGrid option, this is included on the baf.
- Information related to PCM solvation, such as the cavity and polarization charges, is included.
New options for the Output keyword:
- If an output file was named on the command line (-OB=name, -OF4=name, etc.) then the file name is not read from the input stream.
- If not requested on the command line, an output faf or baf can be requested via the baf or faf option to the output keyword. These are equivalent to the old matrix/raw options in G16, except that the default for a faf is now 8-byte integers.
- Minimal: include only input information and results which are scalar or whose dimensions depend only on the number of atoms, omitting operator matrices and other objects whose dimensions depend on the size of the basis set.
- Tinker: write just the information required for the Tinker/Gaussian interface (for QM/MM dynamics using the Amoeba polarizable force field).
- NoTrajectory: do not include information on previous points in a dynamics trajectory. This is useful when using a baf to communicate with an external program at each point of a trajectory.
- AllAtoms/ActiveAtoms: whether to include all atoms or only the active ones when writing a baf during a calculation on an ONIOM model system. ActiveAtoms is the default.
- ReadScalars: when writing a baf at the end of a calculation, read an input section with additional values to include in the Gaussian Scalars object. These are pairs of element # and floating-point value. 1000 scalars are allocated and Gaussian uses only the lowest numbers. Elements 900-1000 are reserved for user values.
- ReadNotes: when writing a baf at the end of a calculation, read an input section with text to include as the Notes object in the file.
SepLines: read a notes section and store it as an array of strings with each line as an element of the array.
MarkLines: read a notes section and store it with lines delimited by backslashes.
General Information:
- formchk, unfchk, cubegen, newzmat, and freqchk default file types based on the extension if this is specified, I.e.:
formchk t
reads t.chk and writes t.fchk, as in G16, while
formchk t t.baf
reads t.chk and writes binary array file (BAF) t.baf. Similarly,
cubegen dens np
reads np.fchk by default, while
cubegen dens np.chk
reads binary checkpoint file np.chk.
All four can handle BAFs/FAFs in addition to the file types handled previously.
- Most utilities obey the -m command line switch to set memory usage.
Cubegen:
- Electrostatic potential evaluation is parallelized, and controlled by the -c or -p command-line switches as for Gaussian.
- The input and/or output file can be BAFs. The points at which the requested property is computed can be taken from a cube specification or a 3d quadrature grid in a BAF.
- Additional properties which can be requested by the first command-line argument:
- Basis[=N]: evaluate the AO basis functions, up to Nth derivatives, at each grid point.
- VSpin: evaluate the spin vector at each grid point. This will be a vector in the z direction for a spin-aligned density but a general vector for a GHF/GKS density.
- AtDens: evaluate the total density from a superposition of spherical atomic densities (the same ones used for the Harris initial guess and for Hirshfeld partitioning of properties).
- RawPotential=N: evaluate the electrostatic potential integrals over pairs of AO basis functions at each grid-point, producing NB*(NB+1)/2 values at each grid-point, where NB is the number of basis functions, or the potential integrals involving one basis function and Nth order derivatives of a basis function, producing an (NB,NB,3) matrix at each grid-point for N=1 and an (NB,NB,15) matrix at each grid-point for N=2.
- RawSpinOrbit: return raw one-electron spin-orbit integrals over AO basis functions, producing 3*NB*(NB+1)/2 values at each grid-point.
- VCurrentDensity: produce the vector current density at each grid-point. This is only non-zero for complex or non-spin-aligned (GHF/GKS) densities.
- AllXC=density: Uses the specified density to evaluate all the intermediate variables used in XC functionals. If the density is spin-aligned, 12 variables are evaluated (alpha and beta density, 3 components of grad-Rho alpha and beta, tau and the Laplacian). For general densities, there are 24 variables: the 12 above followed by the corresponding variables evaluated using the real symmetric and imaginary symmetric parts of the alpha-beta density.
- EXC=IExCor[,N]: Evaluate the XC functional number IExCor at each grid-point, or the functional value and up to Nth derivatives. The derivatives are with respect to the variables defining the functional (2 for local functionals, 5 for GGAs, 7 for tau-functionals, etc.) The values of IExCor are those set by the route generator for IOp(74) in overlay 3 for the functional.
Newzmat:
- baf/faf files can be input or output.
- xmol files can be input or output.
- -deffrag switch requests that fragment information be generated from the connectivity if it is not already present.
- -restrs restores the ONIOM real system on an input chk file.
- When the output file is a Gaussian input file and the input file is either a baf/faf or a chk file -rdgen n m requests that elments n through m of the Gen array (the Gaussian Scalars object in a baf/faf) be copied as an input section for the ReadScalars route keyword.
Formchk:
For an input chk file from an ONIOM calculation:
- -rs to restore the ONIOM real system
- -HS, -LM set up the chk file and produce an output file for the high-level on the small system, low-level on the medium system, etc.
- For an ONIOM calculation with multiple disjoint fragments in the high level, -onfrag N specifies fragment number N.
For an output baf/faf:
- -allatoms to include inactive atoms (in an ONIOM model system).
- -min to include just the molecule specification and results such as energies and forces, omitting matrices whose dimensions depend on the basis set.
Dumpbaf:
- prints the contents of a baf/faf, or
- with -l switch, just lists the names and dimensions of the objects and not their contents.
Cubman:
- action and file names can now be specified on the command line.
Solname:
- solname -s stoich prints all solvents matching stoichiometry stoich
- solname -n name prints all the synonyms for the named solvent.
Copychk:
- uses -m for memory size. an integer as the first command-line argument is ignored.
- can do files file1,file2,… to file1n,file2n,… to copy only the specified files but to different numbers in the output file.
- -restrt, -hs, etc. to set up the output file for a particular ONIOM sub-calculation as for formchk.
Mm:
- -out name.baf or -out name.faf writes a baf/faf with the final structure and energy.
Trjbaf:
- This is a new utility which extracts information from any of the new types of dynamics into a baf which is suitable for post-processing. Refer to the documentation and tutorial paper on the new dynamics code for details.
