|0||Default (same as 1).|
|1||Read the integrals off disk.|
|2||Compute 2e integrals.|
|3||Compute 2e integrals and store in-core.|
|4||Compute 2e integrals and forbid in-core.|
|NNNNx||Use option NNNNN in control of 2e integral calculation.|
|000000||Default — incremental Fock matrix formation only for direct SCF.|
|100000||Form full Fock matrix every time.|
|200000||Form delta-F each iteration.|
|1000000||Clear in-core integrals for testing.|
Convergence (RMS density except in L506 (SQCDF), L508 (rms rotation gradient), and L510 (Energy)).
|0||10-8, except 10-7 for PBC, and 10-10 for SQCDF.|
Maximum number of iterations.
|0||128 (except 512 in L503 and L508, and 64 in L506).|
|-1||Do CI only in L510.|
|-2||Do CI and density matrices only in L510.|
|-3||Do a single iteration in L510.|
|0||Steepest descent with search parameters default.|
|1||Steepest descent with search parameters read (see below).|
|2||Classical SCF (Roothaan’s method of repeated diagonalization).|
|4||Conjugate gradients with search parameters default.|
|5||Conjugate gradients with search parameters read (see below).|
|The search parameters are max. number of search points (I1).|
|Min. number of search points (I1).|
|Initial step size, TAU (G18.5).|
|Scaling factor for subseq. TAU (G20.5)|
|2||Scaled steepest descent.|
|3||Quadratic convergence (after rotation gradient is sufficiently small).|
|4||Exit when NR point is reached, so L502 can take over.|
|00||Default linear search (full search).|
|10||Do a full linear search to locate a minimum.|
|20||Do a linear search only if the energy goes up after the initial step.|
|000||Default handling of wrong curvature (switch direction).|
|100||Reverse direction if curvature in NR step direction is wrong.|
|200||Take pure NR steps, even if curvature is wrong.|
|1000||Turn on linear search and variable step logic.|
|2000||Turn off linear search and variable step logic.|
|1||IRdF2, read damping coefficients.|
|10||IFrzCI, freeze CI coefficients after 1st iteration.|
|100||Read unformatted symbolic matrix elements from NDATA instead of RWF.|
|1000||Read in damping factors from cards.|
|10000||Use Levy damping.|
|100000||Read Fock matrix restriction matrix.|
|1||Yes, criterion default 10-3.|
|2||Yes, criterion read in (Format G16.10).|
|-1||None; coefficients are frozen at initial values (L504 only, causes coefficients to be read in order 11 12 22).|
|-N||Dynamic level shifting to achieve a gap of -0.001*N.|
|-2||Dynamic level shift to a default goal (same as -200).|
|-1||No level shifting.|
|0||Default: -200 for diagonalization calculations, -1 for sparse diagonalization replacements, and if energy DIIS is turned on.|
|N||Shift by 0.001*N.|
|0||Both three-point and four-point extrapolation are performed when applicable.|
|1||Three-point extrapolation is inhibited, but the program will still perform four-point extrapolation when possible.|
|2||Both three-point and four-point extrapolation schemes are ‘locked out’ (IE. disabled).|
|10||Do Camp-King always, taking one step using if |Lpred-1| ³ Thresh.|
|20||Do not do Camp-King.|
|30||Do Camp-King only if the energy rises by at least Thresh.|
|40||Same as 1, but CK also done if the energy goes up.|
|NNN00||Threshold for CK is NNN/1000 (step for 10, Hartrees for 30). Default is 0.3 for 10,0.001 for 30; with 30,999 implies 10-10.|
Whether to allocate only two N2 arrays for RHF.
Number of GVB pairs (L506). If non-zero, the number of orbitals in each pair is read in format (30I2). Each pair consists of the highest available occupied from the guess (after high spin orbs are accounted for) and the lowest available virtuals. If <0, pair coefficients are read; otherwise standard initial values are used.
|1||Continue the run even on non-convergence. The ILSW flag for convergence failure is set.|
|2||Terminate on non-convergence.|
|1||Turn the current RHF run into a uhf run at the end of this link.|
|10||Terminate after computing the 2e terms at the first iteration.|
|20||Just recompute band structure from stored real-space Fock matrix.|
|100||ADMP/FOSimult, later cycles: transform the density from L103/L121 before calculating the energy and Fock matrices.|
|200||ADMP/FOSimult, first cycle: use initial AO densities.|
|1000||Use Generalized energy-weighted density routines regardless.|
|2000||Do not use GEW routines even for CP.|
|10000||Fit the converged density even if fitting is not in use during the SCF. Also redoes the fit at the end even if using fits during SCF.|
|0||Calculation is performed (provided of course that enough space exists in the RW-files).|
|1||Calculation is bypassed.|
|2||Calculation is performed, contingent on space, and the system RW-files for the appropriate density matrices are updated (useful if one wants a population analysis).|
|0||On. Bessel criterion.|
|1||On. Stronger individual-overlap criterion.|
Flags for MCSCF.
|1||Skip valence-valence Fock matrix elements.|
|10||Skip core-valence Fock matrix elements.|
|100||Skip valence-virtual Fock matrix elements.|
|1000||Skip core-valence Fock matrix elements.|
|10000||Use full diagonalization method rather than Lanczos. (Obsolete; use IOp(5/17)).|
|100000||State average density matrices.|
|0||Default (1 for L502, 2 for L501 and L506).|
|2||Yes, keep occupation of each irreducible representation the same as the initial guess.|
|3||Yes, keep overall wavefunction the same as the initial guess, but doing the minimal amount of orbital switching to accomplish this.|
|00||Default (use Abelian symmetry in diagonalization).|
|10||Use Abelian symmetry in diagonalization.|
|20||Do not use Abelian symmetry in diagonalization.|
|1||TAU is kept fixed.|
|0||Default (1 for full matrices HF/DFT, -30 for semi-empirical, 4 for sparse).|
|-N||Pseudo-diagonalization with real diagonalization every Nth cycle.|
|-1||Same as 3.|
|3||Pseudo-diagonalization whenever allowed by internal tests.|
|7||Sign Matrix Method.|
|1xx||Force formation of the Fock matrix using full storage.|
|2xx||Force formation of the Fock matrix using sparse storage.|
|1||By energy least change.|
|2||By orbital least change.|
|-1||Read in eigenvector.|
Set to zero if Abs(F(I,J)).LE.FUZZY; delete coupling terms between almost degenerate (DELTA E .LE. DEGEN) M.O. vectors.
|1||FUZZY and DEGEN read in (2D20.14).|
|0||Virtuals obtained by diagonalization of Hamiltonians.|
|1||Virtuals obtained by Schmidt orthogonalization to occupieds.|
|0||Default (1032 for 502, 1012 for 508).|
|1||Choose LinEq convergence based on orbital gradient.|
|2||Always use tight convergence.|
|3||Tighten convergence by an extra factor of 10.|
|10||If 2E symmetry is on, symmetrize Fock matrices and require proper density matrix symmetry.|
|20||If 2E symmetry is on, replicate integrals so that density matrices and wavefunctions need not be symmetric.|
|30||If 2E symmetry is on, choose between replicating integrals and symmetrizing the Fock matrix based on whether the current density matrix is symmetric.|
|40||Same as 30 in 502 but 20 in 508.|
|100||Force the density matrix to have full symmetry at the first iteration.|
|200||Force the density matrix to have full symmetry at every iteration.|
|1000||If the density matrices pass the symmetry test, symmetrize them to ensure that they are exactly symmetric.|
|2000||Do not symmetrize the density matrices.|
|00000||Default (20000, except if IOp(8) requests old algorithm).|
|10000||Always pseudocanonicalize in L508.|
|20000||Only pseudocanonicalize in L508 when doing a Newton-Raphson step.|
|0||Orthogonalize C,O,V by separate Lowdin, then schmidt.|
|1||Lowdin orthogonalize C+O and V, then schmidt.|
|10||Don’t use natural orbitals each iteration. Bad for 1st order method.|
|100||Use full 2nd order convergence.|
|200||2nd order iteration at end, in preparation for CPMCSCF.|
|1000||Generate data for multi-reference MP2?|
|10000||Attempt to control root flipping in CI.|
|100000||Read CI vector and use it every iteration.|
|1000000||Use full diagonalization method rather than Lanczos.|
|10000000||Use State Average density matrices (the weights 8F10.8)|
|20000000||Do SA and prepare for SA-CPMCSCF.|
|30000000||Do SA and prepare for Gradient of Energy difference.|
|40000000||Do SA and prepare for SA Second Derivative Computation (terms involving 2nd order orbital rotation derivatives not included).|
|-3||MO damping at all iterations.|
|-2||Turn off damping.|
|-1||Dynamic selection of density damping based on band gap and DIIS error.|
|0||Default (-1 unless re-optimizing during Stable=Opt).|
|N||N/100 new density, (100-N)/100 old density.|
and SPAN read in (2D20.14).
|-1||Choose the best given amount of memory available.|
|0||2 if possible, otherwise 1.|
|1||Forbid in-core: force re-reading of integrals even if they fit in 2 buffers if conventional, do not convert to in-core if direct and enough memory for in-core is available.|
|2||Force allocation for 1 or 2 buffer case conventional case (VV¹IBuf2E).|
|3||Force Lower-triangular in-memory storage.|
|1x||Save generated integrals on disk (file 610).|
|2x||Force computation of raff 1 and 2 integrals even for RHF.|
|3x||Do not save integrals (same as 0x).|
|0||Default (only if doing pseudo-diagonalization or QNDMS).|
|1||Yes, do a final unextrapolated diagonalization after convergence is reached.|
|2||No, just quit when extrapolated convergence is reached.|
|3||Do a full diagonalization at the end without recomputing a new Fock matrix.|
|0||On all the time.|
|N||Rotations are turned on when SQCDF is below 10-N.|
DIIS error for density damping, maximum virtual mixing for MO damping.
For density damping.
|0||Default (Damp if error > 0.001).|
|N||Damp if error > 10-N.|
For MO damping:
|0||Default, no more than 1/3 virtual component for any occupied at each iteration.|
|N||Maximum N/1000 virtual component.|
|0||Abort run via LNK1E.|
|2||Check orthonormality at first iteration.|
|2||Generate MOs using UHF natural orbitals.|
|0||Default (1042) for calculations using diagonalization (2) for calculations using sparse diagonalization replacements.|
|3||Yes, with Fermi broadening as well, deciding on the fly between the two forms.|
|4||Yes, with "pFON" version of Fermi broadening.|
|5||Yes, with "FON" version of Fermi broadening.|
|30||Energy DIIS when DIIS error has increased significantly or is above threshold.|
|40||Energy DIIS when DIIS error has increased significantly, otherwise, mixture of energy and commutator.|
|1xx||Use energy DIIS when commutator gives huge coefficients.|
|Nxxx||Switch from energy to commutator when error is 10-N in method 3; use (DIIS error/10-N) for weight of energy DIIS in method 4.|
|Mxxxx||Use print level M in DIIS.|
|0||Read from input stream.|
|1||Read from RWF.|
|2||Read from checkpoint.|
|0||Optimize all orbitals.|
|1||Freeze all closed, high spin and first natural orbitals. Optimize only 2nd and higher naturals.|
|0||Default (exponentiate rotation angles).|
|1||Apply rotations sequentially.|
Number of Hamiltonians to read in (L506). If zero, the unpaired orbitals are assumed to be high spin. If -1, an open-shell singlet is assumed.
Root of CI to use in MCSCF.xxx
|0||Defaults to 1.|
|N||Use Nth root.|
Use of Raffenetti integrals during direct SCF.
|-1||All integrals done as Raffenetti.|
|0||Default: let FoFJK decide. It will never use Raffenetti for SCF.|
|1||All integrals are done as regular integrals.|
|N||Integrals with degree of contraction greater than or equal to N are done at regular integrals.|
|3||Symmetrize even if symmetry blocking was done, and print symmetries.|
|0||Default (3 in L509).|
|1||Yes, use loose integral cutoffs, convergence on either energy or density and always do incremental Fock formation.|
|3||Thresholds similar to DGauss for convergence and integrals.|
|4||Yes, doing an inexpensive pass 0 and then full accuracy in pass 1.|
|5||Decide between 1 and 4 based on details of the calculation.|
|6||Do iterations with sleazy XC grid, then one iteration with next grid up. The default is CoarseGrid for iterations and SG1 for final energy.|
|00||No longer used.|
|N00||No longer used.|
|I000||Use approximation I, 0=normal 1=Linear approximation to Xc.|
|00000||Use general DBF logic only if the DBF RWF is present.|
|10000||Force use of 1c instead of general DBF logic.|
|20000||Force use of general DBF logic.|
|0||Only summary results are printed (with possible control from the ‘no-print’ option).|
|1||The eigenvalues and the M. O. coefficients are printed at the end of the SCF.|
|2||Same as IOp(5/33)=1, but additionally the density matrix is printed.|
|3||Same as IOp(5/33)=2, but at the end of each iteration.|
|4||Same as IOp(5/33)=3, but all matrix transactions are printed (Beware: Lots of output.)|
Dump option. Regular system defaults apply here.
Whether to checkpoint after every SCF cycle.
|0||Default (checkpoint only if direct).|
|-1||Do not do incremental Fock formation.|
|0||Default (every 20 for direct, except 40 if Camp-King is on).|
|N||Every Nth cycle.|
Whether to vary integral cutoffs during direct SCF.
|2||Yes, do integrals 3 digits more accurately than current convergence.|
|3||Yes, do integrals at same accuracy as convergence until final iteration, then 2 digits more accurately.|
|4||Converge to 10-5 with integrals good to 10-6 first, then full convergence.|
|5||VarAcc allowed, decide based on details of problems.|
|6||VarAcc forbidden because of Guess=Read; allows different default actions for PBC.|
|7||Full accuracy for 2e part, but do pass 0 with cheaper XC grid.|
|8||Full grid throughout, but do pass 0 with cheaper integrals.|
On the fly symbolic matrix element generator.
Use of reaction field; only used now for Onsager and control of details of SCIPCM.
|-N||Multipoles of order N, increment field in Gen(2-4).|
|N||Multipoles of order N, store field in Gen(2-4).|
|00000||Default (for SCIPCM, same as 10000).|
|10000||Update surface every iteration.|
|20000||Update surface every iteration in pass 1 only.|
|30000||Update surface on pass 2 iterations only.|
|40000||Same as 3, but re-use 1e matrix instead of surface terms.|
|50000||Update surface and restart DIIS when within 10-2 of convergence.|
Whether to converge on maximum density change as well or instead of RMS.
|N||Maximum allowed change is 10N larger than RMS.|
|-1||Maximum allowed changed is same as RMS (i.e., convergence only on maximum).|
|-2||Converge only on RMS density change.|
|N0||Converge on energy to 10N*RMS-density-accuracy.|
|xx||Maximum dimension of reduced Hamiltonian used as guess is 100*xx. Default=Min(NSec,500).|
|yy00||Maximum dimension of iterative subspace is 10*yy. Default=Max(50*NStates,200).|
|zz0000||Number of guess vectors generated: Default= NStates*k.|
|k000000||Reduction factor between number of guess vectors provided and number of vectors wanted at the end (1 ≤ k ≤ 9). Default: 1 if reading guess vectors from prev. calc for all states, otherwise 2.|
|ll0000000||Davidson iteration after which to scale back the number of vectors. Warning: For overflow reasons, value must be 0 ≤ ll ≤ 20. Default=2.|
|1||Localize all active orbitals.|
|N||Localize first N (strongly occupied!) orbitals.|
|1||Set up for CAS-MP2.|
|2||Do spin-orbit calculation.|
Options to be passed to CalDFT.
|N||Control flag for CalDFT is N.|
Use of sparse storage and Conjugate Gradient optimization instead of N2 memory and diagonalization.
|0||Default (11, or 22 if sparse is set in ILSW).|
|10||Square storage (only in Fock formation if CG).|
|20||Linear storage (only in Fock formation if diagonalization).|
|2||Use Lanczos except for the last iteration.|
PCM input and solvent type.
|N>0||Solvent type N, default parameters.|
|N<0||Dielectric constant |N|/1000.|
How many HOMOs and LUMOs to solve for after CG.
|N||N of each.|
A0for Onsager SCRF.
First iteration at which to level shift and do FON.
|0||Default= 1 unless doing Stable=Opt, then start after instability searches.|
Override standard values of IRadAn.
Override standard values of IRanWt
Override standard values of IRanGd.
Whether to do FMM.
|0||Use global default.|
|1||Turn off FMM here regardless.|
|100||Turn off both FMM and FoFCou here.|
Override default value of FMFlags.
|N||Yes, use N.|
Override NFx parameter.
|N||Yes, use N.|
Override the choice of XC functional.
|0||Use global values.|
|N||Use functional N, with the same values as for IOp(5/74) in overlay 3.|
Maximum initial temperature for FON (non-PBC), or temperature for broadening (PBC and IOp(5/74)=[1-4]xx).
|-1||Start at a high temperature (limited only by DIIS error).|
|0||Default (3000K = 10 milliHartrees for non-PBC, 600K for PBC).|
|0||Default — 10 steps FON / 20 steps PFON.|
Whether L510 should save a state density as the SCF density and whether it should save any excited state densities as CI/TD densities. Requires that Slater determinants be used so that spin densities can be computed, and cannot be used when doing forces or frequencies.
|1||Read the pair of states to use to compute the density saved as the SCF density. Only one number gives the total rather than transition density.|
|-1||Store the lowest state as the SCF density and the higher states, if any, as CIS/TD excited state densities.|
Options for ADMP.
|0||Default (2 for ADMP, 1 for QNDMS).|
|1||Use Lowdin basis for CP orthonormal transform.|
|2||Use Cholesky basis for CP orthonormal transform.|
Type of k-point integration.
|0||Default (911, should be 193 for metals).|
|1||Use LT method (interpolation).|
|2||Occupy entire points (used together with broadening).|
|3||Full points for insulators, temperature broadening for metals.|
|9||Occupy lowest NE at each k point regardless of the energies.|
|10||Improved LT with quadratic corrections.|
|20||Original LT method.|
|90||No concern for corrections.|
|100||Smearing Marzari method I.|
|200||Smearing Marzari method II.|
|300||First order Hermite-Gaussian of Paxton and Methfessel.|
|500||Classical Fermi-Dirac broadening.|
|900||No broadening (this will be Gaussian broadening with small T).|
Number of alpha electrons, alpha orbitals, beta electrons, and beta orbitals for fractional occupation.
Range around Fermi level around which temperature distribution will be applied if broadening is turned on for PBC.
|0||Default, a value will be chosen in ZInLT1.|
The maximum conjugate gradient step size.
|-1||No maximum step size.|
|0||Default maximum (.8).|
|MMNN||Step size of MM.NN.|
|MM||Maximum Number of CG cycles per SCF iteration. (defaults to 4 CG cycles).|
|NN00||Maximum Number of purification cycles per CG iteration. (defaults to 3 cycles).|
|00000||Don’t use CG DIIS.|
|10000||Use CG DIIS.|
|000000||Polak-Ribiere CG minimization.|
|100000||Fletcher-Reeves CG minimization.|
|0000000||Use diagonal preconditioning in Conjugate-Gradient.|
C.G. Convergence criterion.
|0||Defaults to 10-7.|
Maximum SCF DIIS vectors.
|0||Default (20, except 40 if Camp-King is on).|
|N||Use SCF DIIS with N vectors.|
Over-riding of maximum cycles for XQC.
|-1||Default for first step (32).|
|N||Limit is N cycles.|
|0||Default (1 except during Stable=Opt, then 4).|
|1||Just continue as usual if energy goes up.|
|2||Reduce DIIS space when energy rises from previous cycle.|
|3||Reduce DIIS space when energy goes above the lowest energy.|
|4||Reduce DIIS space whenever energy is above the lowest energy.|
|10||Turn on dynamic level shift from the beginning.|
|20||Turn on dynamic level shift only after FON is over.|
|100||Keep level shift after energy rises.|
|200||Turn off level shift after energy rises.|
|1000||Level shift to a maximum of the goal.|
|2000||Level shift to a maximum of 2*goal.|
|3000||Level shift as much as necessary for HOMO > LUMO.|
|4000||Level shift only if the HOMO-LUMO gap is zero.|
|5000||Level shift only if the HOMO-LUMO gap is zero or insignificant (> -0.1)|
|6000||Level shift only if the gap is zero or insignificant (> -0.1), up to twice the goal.|
|N0000||No longer used.|
|100000||Turn off 3 and 4 point extrapolation if DIIS is on.|
|200000||Retain 3 and 4 point extrapolation if DIIS is on.|
|The energy is only checked after FON has been turned off.|
Accuracy criterion in Fock matrix formation.
|0||Default, set in FoFCou/CalDSu based on accuracy part of IOp(5/5). Typically 10-10 for molecules and 10-12 for periodic systems.|
No longer used.
Linearly dependent basis control for PBC; this and ZFormV should be moved to L302.
Whether to generate sparse guess here.
|1||Yes, do preliminary AM1 calculation.|
|2||Yes, do preliminary AM1 calculation and compare with guess from previous step in geometry optimization.|
Control option for Chebyshev sparse control.
How to do exact exchange.
|0||Default (Normal processing based on FMM for non-PBC, separate Coulomb and NFx exchange for PBC).|
|1||FoFCou for Coulomb, separate FoFCou/NFx for exchange.|
Number of initial iterations for which damping is allowed.
|N||Use grid N.|
|ww||ww = Number of Ras1 orbitals.|
|xx00||xx = Maximum number of holes in Ras1.|
|yy0000||yy = Number of Ras3 orbitals.|
|zz000000||zz = Maximum number of electrons in Ras3.|
Whether to update precomputed grid data with timing information.
|0||Default (Yes, if available).|
Grid for numerical k-integration in FT-LT method.
|0||Default: 32,12,8 for 1,2,3d.|
|N||Number of points in the grid.|
Tight convergence during CGDMS.
SDif test on numerical accuracy of PBC diagonalization.
|N>0||Abort if SDif is larger than N.|
Maximum number of configurations for CAS-MP2.
Number of occupied and virtual orbitals to print for each k-point.
|-1||Default of 5 occupieds and 5 virtuals.|
|0||Default is 5 if printing turned up; otherwise 0.|
|N||N occupieds and N virtuals.|
|0||Default (6 digits on coefficients).|
|N||10-N on coefficients.|
|0||Default (huge, number of CI configurations).|
Minimum number of iterations at which to damp density.
|0||Default (1 if transition metals present, otherwise 0).|
Whether to store nuclear repulsion energy as total energy.
IDoVI for HarFok in test calculations.
|xx0||Ones digit always set to 4 here.|
Whether to do Hirshfeld analysis of spin orientations and compute spin each iteration.
|0||Default (yes for GHF/GKS, no otherwise).|
Number of iterations for Pseudodiagonalization.
|0||Default (1 for semi-empirical, -1 for HF/DFT).|
|-1||Sweep until variable convergence is reached.|
|N||Do a maximum of N sweeps.|
Variable convergence in pseudodiagonalization:
|0||Default (1 for ab initio, 2 for semi-empirical).|
|N||Off-diagonals larger than the initial max/10N are swept.|
|-N||Off-diagonals larger than OVMax*10-N are swept, with OVMax updated each iteration.|
Scale factor for Diag/PseudoDiag tradeoff. Roughly related to the ratio of Diag to Fock formation for large systems.
|0||Default (30 for ab intio, 15 for semi-empirical).|
|N||Pseudodiag must be estimated to be at least N/10 times faster than full diag to be used.|
Threshold for trying alternate initial guess:
|-1||Same as 0.|
|-2||Ignore the alternate guess.|
|N||Try the alternate if the DIIS error >eq; 10-N.|
Whether to match phases with reference orbitals (useful during numerical differentiation).
|1||Match if file of reference orbitals exists.|
|2||Do not match.|
L510: Large-scale CI method
|0||Default (Shaopeng for CAS, Klene for RAS).|
|1||Shaopeng (NYI for RAS).|
Last updated on: 21 October 2016. [G16 Rev. B.01]