Geometry optimization (OPTG)

Automatic geometry optimization is invoked using the OPTG command. The OPT command available in previous MOLPRO versions is no longer needed and not available any more.

OPTG[, key1=value, key2=value,……]

The OPTG command can be used to perform automatic geometry optimizations for all kinds of wavefunctions. For minimum searches, it is usually sufficient to give just the OPTG command without further options or directives, but many options are available which are described in the following sections.

The OPTG command must be given after the energy calculation to which it refers. If the commands for the energy calculation (e.g. HF, KS, MP2, etc.) are in a procedure, the OPTG must also be in the procedure. Furthermore, no procedures without an energy calculation must directly precede OPTG.

Various optimization methods can be selected as described in section selecting the optimization method (METHOD). Molpro allows minimization (i.e. search for equilibrium geometries), transition state optimization (i.e. search for saddle points on energy surfaces), and reaction path following. Since Molpro 2023.1 the optimizer developed by J. Baker and P. Pulay (PQSOPT) is available in Molpro and used by default. The method is described in J. Baker et al., J. Chem. Phys. 105, 192 (1996); J. Baker and P. Pulay, J. Chem. Phys. 105, 11100 (1996); J. Baker, J. Comput. Chem. 18, 1079 (1997); J. Baker and P. Pulay, J. Comput. Chem. 21, 69 (2000). It uses delocalized internal coordinates and converges often better than previous optimisers in Molpro. Furthermore, it allows for constraint optimisations. Other available algorithms are based on the rational function approach and the geometry DIIS approach. Also available is the quadratic steepest descent following method of Sun and Ruedenberg (see J. Sun and K. Ruedenberg, J. Chem. Phys. 99, 5257 (1993)). This method is often advantageous in Transition State searches. For a detailed discussion of the various older minimization algorithms see F. Eckert, P. Pulay and H.-J. Werner, J. Comp. Chem 18, 1473 (1997). Reaction path following is described in F. Eckert and H.-J. Werner, Theor. Chem. Acc. 100, 21, (1998). Please refer to the references section for citations of the analytic gradient methods.

When analytical gradients are available for the optimized energy these will be used. See section analytical energy gradients for a list of methods with analytical gradients. Otherwise the gradient will be computed numerically from finite energy differences. Normally, the last computed ground-state energy is used. But the VARIABLE directive or option can be used to optimize, e.g., Davidson corrected energies, excited states, or counterpoise corrected energies.

By default the program repeats in each geometry optimization step those commands in the input that are needed to compute the last energy and any variable=value assignments in between the last energy calculation and the OPTG command. For example, for MP2 gradients the commands HF and MP2 are needed. The MP2 gradients will then be computed automatically. Variables set by these programs as, e.g., ENERGY, are set as usual in each optimization step. In addition, there is a variable OPTSTEP, which is incremented in each optimization step. Its initial value is zero. To save the results of the initial energy calculation, one can use, e.g.,

hf
if(optstep.eq.0) e1=energy
optg

It is also possible to define procedures for the energy calculation, or to specify the first command from which the input should be repeated in each step (see section options to select the wavefunction and energy to be optimized). The section of the input which is needed for the geometry optimization must not modify variables that are used in the geometry definition (changes of such variables are ignored, and a warning message is printed).

Most parameters can be given as options on the OPTG command line, as described in this section. Alternatively, directives can be used, which will be described in section directives for OPTG.

By default, the last computed energy is optimized, and all commands on which the last energy calculation depends are automatically executed. For certain purposes, e.g., optimization of counter-poise corrected energies or Davidson corrected energies, the following options can be used to alter the default behaviour.

• STARTCMD=command Specifies a start command. In each geometry optimization step all input beginning with command to the current OPTG is processed. This input must not include numerical gradient or Hessian calculations. If numerical gradients are needed, these will be computed for the final energy (or specified variable) by OPTG. It is assumed that these commands have been executed before entering the OPTG program.
• PROC=procname specifies a procedure to be executed in each geometry optimization step. This must define a complete energy calculation (orbital optimization and correlation treatment), and must not include numerical gradient of Hessian calculations (numerical gradients will be computed automatically for the optimized energy or variable). However, the procedure can include the calculation of analytical gradients, for instance for counter-poise corrected optimizations in which a linear combination of several gradient calculations is needed.
• VARIABLE=varname Optimize the value of variable varname. This implies numerical gradients.

• METHOD=PQS|RF|AH|DIIS|QSD|QSDPATH|SRMIN|SRTRANS|STSTEEP
  

Optimization method to be used. See section selecting the optimization method (METHOD) for details.

• ROOT=1$|$2 Minimum search (1, default) or transition state search (2).
• DIRECTION|IDIR=idir Determines initial step length and direction in reaction path following, see section reaction path following options (OPTION).
• STPTOL=value For reaction path following, see section reaction path following options (OPTION).
• SLMAX=value For reaction path following, see section reaction path following options (OPTION).
• STEPMAX=value Max step length in one optimization step. This also affects the step length in reaction path following. For more detailed specifications see section setting a maximum step size (STEP).
• ROTATE (logical). If .true., the Cartesian coordinates are transformed to minimize rotations (default=.true.)

The following options are not available in PQSOPT:

• TRUST=value Trust ratio for Augmented Hessian method (default 0.5).
• AHMAX=value Maximum step size allowed in the Augmented Hessian procedure. This refers to the scaled parameter space (default 0.5).
• CUT=value Threshold for orthonormalization used in conjugate gradient update of Hessian (default 1.d-3).

Specific options for PQSOPT (for details see options for the PQS optimizer):

• OPTC=value This determines the coordinate type (default -1)
• IGEN=value This determines whether the internal coordinates are regenerated in each step (default 1)
• TYPE=value Special flag for surface/cluster optimizations (default 0)
• ICONS=value Special flag for constraint optimisations (default 0)

Convergence criteria can be specified as for other methods using the ENERGY, GRADIENT, and STEP options (see options to modify convergence criteria).

The maximum stepsize can be specified using the STEPMAX option (see options for optimization methods)

Transition state optimization can be activated by setting option ROOT=2.

Options for numerical gradients and Hessian options work as for other optimization programs (see options for numerical gradients and options for computing Hessians).

There are many further options, but only the following may be sometimes relevant:

OPTC:

• OPTC= 1 - generate and optimize in internal coordinates; if this fails abort
• OPTC=-1 - generate and optimize in internal coordinates; if this fails switch to Cartesians and continue (default)
• OPTC= 2- optimize in Z-Matrix internal coordinates (not supported in Molpro)
• OPTC=-2 - optimize in Z-Matrix internal coordinates (not supported in Molpro)
• OPTC= 3 - large-molecule optimization in Cartesian coordinates (never tested so far)
• OPTC= 4 - large-molecule optimization in delocalized internals (never tested so far)

IGEN:

• IGEN=0 - generate a set of non-redundant internals and use throughout the optimization
• IGEN=1 - generate a new set of non-redundant internals from the same primitive space every cycle (default)
• IGEN=2 - generate a new set of underlying primitives and a new set of non-redundant internals every cycle (not recommended)

ITYPE:

• ITYPE=0 - standard molecular optimization (default)
• ITYPE=1 - optimization of molecule reacting/adsorbed on model surface
• ITYPE=2 - optimization of molecular clusters (see below)

ICONS:

• ICONS=0 - full optimization (no constraints, default)
• ICONS=1 - constrained optimization using Lagrange multipliers (default if constraints are given see constraint optimisations using the PQS optimizer)
• ICONS=2 - constrained optimization using penalty functions

Cluster optimizations:

Cluster optimizations use inverse distance coordinates for intermolecular degrees of freedom.

• MOLECULES=[n1,n2,n3,…] - For cluster optimisations. The molecules must come one after the other in the geometry input. n1, n2,… are the number of atoms in each molecule. The sum of these must be equal to the total number of atoms. This option sets automatically ITYPE=2

• CUTOFF=value - Distance cutoff for intermolecular coordinates in Angstrom. The program will generate inverse distance coordinates for between all atoms of different molecules, unless the distance is larger than this value. Default: 5 Angstrom.

• SCALE=value - Scaling factor for inverse distance coordinates. Typical values are between 1 and 10. Default is 1.0

Unfortunately, according to our current limited experience, the cluster optimization in pqsopt is neither well working in the original PQS program nor in Molpro and should therefore be used with care.

Optimizations with z-matrix input:

By default, optimizations with z-matrix geometry input are done with the old RF optimizer, and only the active variables in the z-matrix are optimised (cf. Defining active geometry parameters). It is possible, however, to force pqs optimization by giving the optg option PQS or METHOD=PQS. In this case, all cartesian coordinates will be optimised, and in each iteration the new structure is back-transformed to the z-matrix. If the z-matrix contains constants, these may change and will be overwritten. If the same variable is used in several places of the z-matrix and the corresponding new parameters are not the same, the variable is only kept in the first place, and the further ones are overwritten by the new fixed parameters. The updated z-matrix is printed in the log-file in each iteration.

Constraints can be be defined in an input block (with the OPTG command block) beginning with CONSTRAINT and ending with END. The possible coordinate types are:

* STRE I J value (Angstrom)
* BEND I J K value (degrees)
* TORS I J K L value (degrees)
* OUTP I J K L value (degrees)
* LINC I J K L value (degrees)
* LINP I J K L value (degrees)

where I, J, K, L refer to row numbers in the XYZ input. Alternatively, atom symbols like, e.g. H1, O, H2, can be given, but these must be unique in the input geometry. The coordinate types are:

STRE distance constraint between any two (different) atoms.
BEND planar bend constraint between any three (different) atoms. J is the middle atom of the bend.
TORS dihedral angle (proper torsion) constraint between any four (different) atoms. The connectivity is I-J-K-L, with J-K being the central “bond”. The angle is between the planes I-J-K and J-K-L.
OUTP out-of-plane-bend constraint between any four (different) atoms. The angle is made by bond I-L with the plane J-K-L (L is the central atom)
LINC colinear bending constraint between any four (different) atoms. Bending of I-J-K in the plane J-K-L.
LINP perpendicular bending constraint between any four (different) atoms. Bending of I-J-K perpendicular to the plane J-K-L.

LINC/LINP are special angles used when four atoms are near-linear.

The constraint values have not to be satisfied in the initial geometry. If value is not given, the initial value will be kept. The number of constraints is in principle arbitrary, but currently restricted to 20.

Example: Keep the distance between atoms 2 and 3 to 1.5 ANG:

{OPTG
CONSTRAINT
STRE,2,3,1.5
END}

Composite constraints:

It is possible to define linear combinations of primitive internal coordinates as constraints. These must be defined after the simple constraints (if present). Each composite constraint begins with the keyword COMPOSITE. The input of each primitive is as described above, but the weight (the factor in the linear combination) replaces value. The value is given at the end of the first line. Angles in the linear combinations are in radians. The general structure is:

{OPTG,options
CONSTRAINT
simple constrains (if present)
COMPOSITE
type atom_list weight value
type atom_list weight
…
END}

value is the total value of the combined constraint. Note that the weights are normalised. Example:

COMPOSITE
STRE I J 1.0
STRE J K -1.0
BEND I J K 2.0
END

will constrain the value of $6^{-1/2}(R_{IJ} - R_{JK} + 2 A_{IJK})$ to whatever it is in the starting geometry.

Currently, the total number of composite constraints is limited to 20. Note that there is no checking for incompatible or impossible constraints, and it is the user's responsibility to make correct definitions.

The standard Molpro convergency criterion requires the maximum component of the gradient to be less then $3 \cdot 10^{-4}$ [a.u.] and the maximum energy change to be less than $1 \cdot 10^{-6}$ [H] or the maximum component of the gradient to be less then $3 \cdot 10^{-4}$ [a.u.] and the maximum component of the step to be less then $3 \cdot 10^{-4}$ [a.u.].

It is also possible to use the convergency criterion of the Gaussian program package. It is somewhat weaker than the Molpro criterion and requires the maximum component of the gradient to be less then $4.5 \cdot 10^{-4}$ [a.u.] and the root mean square (RMS) of the gradient to be less then $3\cdot 10^{-4}$ [a.u.] as well as the maximum component of the optimization step to be less then $0.0018$ [a.u.] and the RMS of the optimization step to be less then $0.0012$ [a.u.].

• MAXIT=maxit maximum number of optimization cycles. The default is 100. It cannot be increased above 100. If an optimization did not work, restart from e.g. the structure with the lowest energy using the information in the correspoing log-file (suffix .log).
• GRADIENT=thrgrad required accuracy of the optimized gradient. The default is $3 \cdot 10^{-4}$.
• ENERGY=threnerg required accuracy of the optimized energy. The default is $1 \cdot 10^{-6}$.
• STEP=thrstep convergence threshold for the geometry optimization step. The default is $3 \cdot 10^{-4}$.
• BAKER (logical). Use Baker’s convergency criteria (see J. Baker, J. Comp. Chem. 14,1085 (1993)).
• GAUSSIAN (logical). Use Gaussian convergency criteria.
• SRMS=thrsrms sets (for Gaussian convergency criterion) the required accuracy of the RMS of the optimization step. The default is $0.0012$.
• GRMS=thrgrms sets (for Gaussian convergency criterion) the required accuracy of the RMS of the gradient. The default is $3 \cdot 10^{-4}$.
• FREEZE=thrfreez Freeze DFT grid and domains in local calculations if the step length is smaller than thrfreez (default 0.01).

Note: The defaults for the convergence parameters can also be changed by using a global GTHRESH directive, i.e.

GTHRESH, OPTSTEP=step, OPTGRAD=grad, ENERGY=energy;

If the geometry is given as Z-matrix, the default is to optimize the variables on which the Z-matrix depends. In case of xyz input, always all 3N coordinates are optimized, even if the xyz input depends on fewer variables. If Cartesian z-matrix input is used, optimization in full space is only enforced if automatic orientation is requested using the MASS, or CHARGE options on the ORIENT directive. See opt_space in section optimization coordinates (COORD) for details.

• SPACE=ZMAT|3N Specifies the coordinates to be used in the optimization. Z-matrix optimization is only possible if the geometry is given as Z-matrix.
• OPT3N|3N (logical). Same as SPACE=3N
• ZMAT (logical). Same as SPACE=ZMAT

These options specify the coordinates in which the optimization takes place. The default is to use local normal coordinates. See opt_coord in section optimization coordinates (COORD) for details.

• COORD=DELO[CALIZED]|NORMAL|NONORMAL|BMAT
• NORMAL (logical). Same as COORD=NORMAL.
• NONORMAL (logical). Same as COORD=NONORMAL.
• BMAT (logical). Same as COORD=BMAT.

DELO[CALIZED] enforces full 3N optimization in delocalized internal coordinates using the PQSOPT program (available in Molpro 2023.1 or later)

BMAT uses an older version of optimisations using internal coordinates.

Numerical gradients can be computed with respect to variables on which the Z-matrix depends or with respect to Cartesian coordinates. In the latter case, it is most efficient to use symmetrical displacement coordinates. These do not change the symmetry of the molecule and the number of displacements is minimal. Alternatively (mainly for testing purpose) the gradients can be computed using symmetry unique Cartesian displacements or all 3N Cartesian displacements. In these cases the symmetry of the molecule can be reduced by the displacements and using such displacements is normally not recommended.

• DISPLACE=ZMAT|SYMM|UNIQUE|CART
  

Displacement coordinates to be used for numerical gradient. The default is ZMAT if the geometry is given as a zmatrix which depends on variables, and SYMM (symmetrical displacement coordinates) otherwise. The use of UNIQUE or CART is not recommended.

• SYMMETRY=AUTO|NOSYM Symmetry to be used in wavefunction calculations of numerical gradients. This option is only relevant if DISPLACE=UNIQUE|CART. If AUTO is given, the maximum possible symmetry is used for each displacement. This implies that the energy is independent of the symmetry used. Note that this often not the case in MRCI or CASPT2 calculations. The option can also not be used in local correlation calculations.
• AUTO (logical). Same as SYMMETRY=AUTO
• NOSYM (logical). Same as SYMMETRY=NOSYM
• RSTEP=rstep Step length for distances in numerical gradient calculations (in bohr). The default is 0.01.
• DSTEP=dstep Step length for symmetrical displacements (in bohr). The default is 0.01.
• ASTEP=astep Step length for angles in numerical gradient calculations (in degree). The default is 1.
• FOURPOINT (logical). Use 4-point formula for accurate numerical gradient.
• NUMERICAL (logical). Force the use of numerical gradients, even if gradients are available.

In special cases the energy calculations may require several steps with different occupations and/or symmetries. In this case the FORCEINP directive must be given (see section miscellaneous options:).

By default, an approximate Hessian (model Hessian) is used. Optionally, a Hessian can be computed in the optimization or read from a previous Hessian or frequency calculation.

• NUMHESS=hstep If given, a numerical Hessian is computed in each hstep’th iteration. If hstep=0 or not given, only an initial Hessian is computed.
• HESSREC=record Read initial Hessian from the given record. If record is not given or zero, the last computed Hessian is used.
• READHESS (logical). Same as HESSREC=0.
• HESSPROC=procname specifies a procedure to be used for computing the Hessian. This procedure must be define a complete energy calculation (orbital optimization and correlation treatment). A different method can be used than for the optimized energy, but the basis must not be redefined in this procedure. For instance, an MP2 Hessian can be used for CCSD(T) optimizations, or a CASPT2 Hessian for MRCI optimizations. By default, the same procedure is used for the Hessian as for the optimized energy. Note: If a hessian procedure is used and two or more optimizations are done after each other in the same input, the complete energy calculation (including orbital optimization) must be defined before each optg command or in the optg procedure.
• HESSVAR=varname Compute Hessian for variable varname. This implies numerical calculation of the Hessian from energies. The default is to use the same variable as for the energy and gradient.
• HESSCENT Use central gradient differences for computing Hessian (only effective if gradients are available)
• HESSFORW Use forward gradient differences for computing Hessian (only effective if gradients are available). This effectively computes the Hessian at a slightly displaced geometry, but needs only half the number of displacements. This is the default.
• UPDATE=BFGS|IBFGS|CGRD|PMS|POWELL|MS|NONE
  

Hessian update method to be used. See section hessian update (UPDATE) for details.

• MAXUPD=maxupd Max number of Hessian updates. The count is reset to zero each time a Hessian is computed.
• NUMDIAG If true, replace diagonal elements of model hessian by diagonal numerical hessian (if available). This sometimes improves convergence, but since it may lead to symmetry breaking it is no the default.

Note that there are restrictions for computing Hessians for multireference methods (MCSCF, MRCI, ACPF,AQCC,RS2). For these methods the symmetry must not change by any displacements, since this could change the occupations and states and may lead to non-contiguous potential energy surfaces. One of the following three options can be used in these cases:

• Use no symmetry from the beginning (NOSYM).
• Use symmetric displacement coordinates. This is the default if the optimization is done in 3N cartesian coordinates. One can use OPTG,DISPLACE=SYMM to force the use of symmetrical displacements (this creates 3N cartesian coordinates if a Z-matrix is used in the geometry input).
• Use a Z-matrix with the restriction that no variable in the Z-matrix may change the symmetry. For example, geometry={O;H1,O,r;H2,O,r,H1,theta} would work, but geometry={O;H1,O,r1;H2,O,r2,H1,thetai} would not work. In this case the program prints a warning message. If an incorrect Z-matrix is used and the symmetry changes, the program will crash.

• VARSAVE Save Cartesian gradients in variables GRADX, GRADY, GRADZ.
• NONUC Do not compute gradients at lattice points.
• DEBUG Set debug print options.
• PRINT=iprint Print option for optimization.
• SAVEXYZ[=file] Save the optimized coordinates in an xyz-file. One file is written for each step. The filename is file_nn.xyz, where nn is the iteration number. For the final geometry, _nn is omitted. If filename is not given, file is taken to be the root name of the input, i.e. test.inp creates test_1.xyz in the first iteration and test.xyz for the converged geometry. By default, the xyz information is written to the log file in each step.
• SAVEACT[=file] Save optimized variables in each step. The file name is file.act. If file is not given the root name of the input is used. The file can be read later using the READVAR command or copied into new input.
• SAVEGRD[=file] Write in each step the Cartesian coordinates and gradients. The file name is file.grd. If file is not given, the root name of the input appended by .grd is used.
• APPEND (logical). If given, existing SAVEACT and/or SAVEGRD files are appended.
• REWIND (logical). If given, the SAVEACT and/or SAVEGRD files are rewound at each step, i.e. only the last geometry or gradient is saved, previous values are overwritten.
• FORCEINP (logical) Disables the mechanism that freezes the occupations during geometry optimizations. If this option is given, the occupations and wave function definitions given in the input for the energy calculations are strictly obeyed.

An alternative way to specify options is to use directives, as described in this section. In some cases this allows more detailed specifications than with the options on the OPTG command. In particular, directives ACTIVE or INACTICE can be used to define the optimization space in more detail.

METHOD,key;

key defines the optimization method.

For minimization the following options are valid for key:

• PQS PQS optimizer (default for full (3N) optimisations).
• RF Rational Function method (default for Z-matrix optimisations).
• AH Augmented Hessian method. This is similar to RF algorithm but uses a more sophisticated step restriction algorithm.
• DIIS Pulay’s Geometry DIIS method. As an an additional option you may add the number of geometries to be used in GDIIS interpolation (default 5) and the interpolation type (i.e. the subspace in which the GDIIS interpolation is made.
  

METHOD,DIIS, number, type
type may be GRAD interpolation using the gradients (default), working good for rigid molecules, STEP interpolation using Quasi-Newton steps which could be advantageous in dealing with very floppy molecules, ENER interpolation using energies, which is an intermediate between the above two.

• QSD Quadratic steepest descent method of Sun and Ruedenberg.
• SRMIN Old version of QSD.

For transition state searches (invoked with the ROOT option, see section transition state (saddle point) optimization (ROOT)) key can be

• PQS PQS optimizer using internal coordinates (default for full (3N optimisations).
• RF Rational Function method (default for Z-matrix optimisations).
• DIIS Pulay’s Geometry DIIS method (see above).
• QSD Quadratic Steepest Descent Transition State search using the image Hessian method (see J. Sun and K. Ruedenberg, J. Chem. Phys. 101, 2157 (1994)) The use of this option is recommended for transition state searches – especially in complicated cases. The optimization step is checked and the Hessian is recalculated when approaching a troublesome region of the PES. Thus this method is somewhat safer (and often faster) in reaching convergence than the RF or DIIS method. The Hessian recalculation safeguard may be turned off using the METHOD,QSD,NOHESS input card.
• SRTRANS Old version of QSD.

For reaction path following the input key is

• QSDPATH Quadratic Steepest Descent reaction path following. This methods determines reaction paths (intrinsic reaction coordinates, IRCs) by following the exact steepest descent lines of subsequent quadratic approximations to the potential energy surface. The Hessian matrix is calculated numerically at the first optimization step and subsequently updated by Powell or BFGS update. If a given arc length of the steepest descent lines is exceeded, the Hessian is recalculated numerically (see OPTION section reaction path following options (OPTION)). For details see J. Sun and K. Ruedenberg, J. Chem. Phys. 99, 5269 (1993) It is also possible to recalculate the Hessian after each m steps using the NUMHES,m command (see section numerical Hessian (NUMHESS)). If the Hessian matrix is recalculated in every optimization step (NUMHES,$1$) a algorithm different to the one with updated Hessians is used, which is very accurate. Using the PRINT,OPT card, this algorithm prints in every optimization step a reaction path point r which is different from the point where the energy and the gradient is calculated but closer to the real reaction path (for further details of the algorithm see J. Sun and K. Ruedenberg, J. Chem. Phys. 99, 5257 (1993)). For further input options of the QSD reaction path following see OPTION section reaction path following options (OPTION).
• SRSTEEP Old Version of QSDPATH.

It is possible to use various coordinate types and algorithms for the optimization. This can be controlled by additional subcommands as described in this and the following subsections.

COORD,[opt_space],[opt_coord],[NOROT]

These options choose the optimization space and the coordinate system in which the optimization takes place.

opt_space defines the parameters to be optimized. By default, if the geometry input is given in Z-matrix format, all variables on which the Z-matrix depends are optimized. Subsets of the variables on which the Z-matrix depends can be chosen using the ACTIVE or INACTIVE subdirectives. If the Z-matrix depends on no variables or xyz input is used, all $3N$ cartesian coordinates are optimized.

opt_space can be one of the following:

• ZMAT Optimize all variables on which the Z-matrix depends (default if the geometry is given as Z-matrix).
• 3N Optimize all $3N$ cartesian coordinates (default if the Z-matrix depends on no variables, or if xyz-input is used). Z-Matrix input coordinates will be destroyed if 3N is used.

opt_coord determines the coordinates in which the optimization takes place. Optionally, delocalized internal coordinates, local normal coordinates, cartesian coordinates, or natural internal coordinates can also be used.

opt_coord can be one of the following:

• DELO[CALIZED] Optimization in delocalized internal coordinates using PQSOPT (available in Molpro 2023.1 or later). This is the default for full space (3N) optimisations. If the option is given, 3N optimization is enforced (but constraints can be defined).
• NORMAL Optimization in local normal coordinates.
• NONORM Don’t use local normal coordinates.
• BMAT[=filename] Use Pulay’s natural internal coordinates, see G. Fogarasi, X. Zhou, P. W. Taylor and P. Pulay J. Am. Chem. Soc. 114, 8191 (1992); P. Pulay, G. Fogarasi, F. Pang, J. E. Boggs J. Am. Chem. Soc. 101, 2550 (1979)). Optionally, the created coordinates as well as additional informations about this optimization are written to the specified file. These coordinates resemble in part the valence coordinates used by vibrational spectroscopist, and have the advantage of decreasing coupling between different modes. This often increases the speed of convergence. Similar coordinates are also used in the PQSOPT program (DELO), and mostly this converges even better.

If the option [NOROT] is given, the cartesian coordinates are not transformed to minimize rotations.

DISPLACE,displacement_type

see section choice of coordinates (COORD) for details.

ACTIVE,param;

Declares variable name param to be active in the optimization. By default, initially all variables on which the geometry depends are active; inclusion of an ACTIVE card makes all parameters inactive unless explicitly declared active (see also INACTIVE).

INACTIVE,param;

Declares variable name param to be inactive in the optimization. If any ACTIVE card appears in the input, this card is ignored! (see also ACTIVE)

By default, the Molpro geometry optimization utilizes a force field approximation to the hessian (“Model Hessian”, see R. Lindh, A. Bernhardsson, G. Karlström and P. Malmqvist Chem. Phys. Lett. 241, 423 (1995)), which speeds up convergence significantly. The Model Hessian is parameterized for the elements up to the third row. Alternatively, the model Hessian of Schlegel can be used, or the Hessian can be computed numerically (see also section numerical Hessian (NUMHESS)).

HESSIAN,options

where options can be

• MODEL Use Lindh’s Model Hessian in optimization (default).
• MODEL=SCHLEGEL Use Schlegel’s Model Hessian.
• MODEL=VDW Add vdW terms to Lindh’s Model Hessian.
• SCHLEGEL same as MODEL=SCHLEGEL.
• VDW same as MODEL=VDW.
• NOMODEL Don’t use Model Hessian approximation to the hessian.
• NUMERICAL=hstep Recompute Hessian after hstep iterations. This disables the use of a model hessian. If hstep=0, the Hessian is only computed in the first iteration. Default parameters are used for computing the numerical Hessian, unless modified using options as described for the NUMHESS directive, see Sect. numerical Hessian (NUMHESS). Any option valid for the NUMHESS directive may also follow the NUMERICAL option on the HESSIAN directive.
• READ|RECORD|HESSREC=record Read Hessian from given record. If record is not given or zero, the last computed hessian will be read. See section numerical Hessian (NUMHESS) for more details about numerical Hessians.
• UPDATE=type Method used for hessian update. See section hessian update (UPDATE) for possibilities and details.
• MAXUPD=maxupd Max number of hessian updates. The count is reset to zero each time a hessian is computed.

If the Model Hessian is disabled (NOMODEL) and no Hessian is read or computed, the initial hessian is assumed to be diagonal, with values 1 hartree*bohr**(-2) for all lengths, 1 hartree*radian**(-2) for all angles. Additional matrix elements of the hessian can be defined using the HESSELEM directive, see section hessian elements (HESSELEM).

In transition state searches the Hessian is evaluated numerically in the first iteration by default. Alternatively, if READ is specified, a previously computed hessian is used.

NUMHESS,options

or

NUMHESS,hstep,options If this directive is present a numerical Hessian is computed using finite differences. If analytical gradients are available, one can use forward gradient differences (needs one gradient calculation for each coordinate) or central differences (more accurate, needs two gradient calculations for each coordinate). For transition state optimizations it is usually sufficient to use forward differences. If analytical gradients are not available for the optimized method, the energy is differentiated twice. In this case only central differences are possible.

The following options can be given:

HSTEP=hstep hstep=-1: Don’t calculate numerical hessian (default for minimization);
hstep=0 Calculate numerical hessian only once at the start of the optimization (default for transition state searches).
hstep=n Calculate numerical hessian after each n optimization steps. This is useful for difficult transition state optimizations (e.g. if the eigenvalue structure of the hessian changes during the optimization). FORWARD Use forward differences (default). CENTRAL Use the more accurate central differences. RSTEP=rstep Step length for distances (in bohr). The default is 0.01. ASTEP=astep Step length for angles (in degree). The default is 0.5 or 1 for angles below and above 90 degree, respectively. DSTEP=dstep Step length for symmetrical displacements (in bohr). The default is 0.01. VARIABLE=varname Use given variable for numerical calculation of the Hessian. Note that this disables the use of gradients, and Hessian evaluation can be very expensive. PROCEDURE=procname Procedure to be used for computing Hessian. This procedure must be define a complete energy calculation (orbital optimization and correlation treatment). A different method can be used than for the optimized energy. For instance, an MP2 hessian can be used for CCSD(T) optimizations, or a CASPT2 hessian for MRCI optimizations. By default, the same procedure is used for the hessian as for the optimized energy. DISPLACE=type type can be one of the following:

SYMM Use symmetric displacement coordinates (default). This is the only recommended option. CART Use $3N$ cartesian displacements (not recommended). This requires many more energy calculations than necessary and does not preserve the molecular symmetry. UNIQUE Use symmetry-unique cartesian displacements (not recommended)

Note that the displacement type for gradient and hessian must be the same. CALC=icalc icalc=0: Recalculate the complete Hessian matrix numerically after each hstep optimization steps (default).
icalc=1 (currently disabled): Recalculate selected Hessian matrix elements if the relative deviation of this element before and after update (see UPDATE, section hessian update (UPDATE)) is larger than thresh. If thresh is not specified, a default value of $thresh=0.05$ (i.e. a maximum deviation of $5 \%$) is used.
icalc=2 (currently disabled): Recalculate complete Hessian matrix if the RMS deviation of the Hessian matrix before and after update is larger than $thresh$. If thresh is not specified a default value of $thresh=0.5$ $a.u.$ is used. THRESH=thresh Threshold for partial or dynamical update of hessian, see above

[optgeo:hesselem]

HESSELEM,value, active1,active2,…

sets the starting value for hessian matrix element between active variables active1, active2 to value. If active2 is omitted it defaults to active1 (diagonal element). As many HESSELEM directives as needed may be given.

UPDATE,[TYPE=]type,MAXUPD=maxupd

This directive chooses the update type and limits the number of points used for the hessian update to maxupd. The default number of steps used in hessian update procedures is 5. If there are symmetry constraint in the coordinates of the optimization, the default number may be lower than five.

In minimizations type may be

• BFGS Use BFGS update of hessian (default).
• IBFGS Use BFGS update of the inverse hessian.
• CGRD Use Conjugate Gradient update (see also CUT,TRUST).
• NONE Don’t do any update.

In transition state optimizations type may be

• PMS Combined Powell/Murtagh-Sargent update of hessian (default).
• POWELL Use Powell’s update of the hessian.
• MS Use update procedure of Murtagh and Sargent.
• NONE Don’t do any update.

NUMERICAL,options,active$_1$=step$_1$, active$_2$=step$_2$ …;

With this directive the gradients are computed by finite differences. step$_i$ is the increment for the active geometry parameter $active_i$. For active parameters which are not specified, the default values are used. By default, the increment is 0.01 bohr for bond distances and 0.5 or 1 degree for angles less than or greater than 90 degrees, respectively. These defaults can be modified by specifying RSTEP or ASTEP. DSTEP is the length of symmetrical displacements, which are used if the optimization is performed in 3N coordinates.

For each active variable, two energy calculations are necessary in each geometry optimization step – so numerical optimizations may be expensive! In optimizations of 3N coordinates symmetrical displacement coordinates are normally used to minimize the number of energy calculations. (see section choice of coordinates (COORD)).

For optimization of special energies see VARIABLE section optimizing energy variables (VARIABLE).

The following options can be given:

• RSTEP=rstep Step length for distances (in bohr). The default is 0.01.
• ASTEP=astep Step length for angles (in degree). The default is 0.5 or 1 for angles below and above 90 degree, respectively.
• DSTEP=dstep Step length for symmetrical displacements (in bohr). The default is 0.01.
• CENTRAL Use central differences for gradient (default)
• FORWARD Use forward differences (not recommended for gradient).
• FOURPOINT Use four-point formula for very accurate numerical gradients.
• PROCEDURE=procname Use given procedure for numerical calculation of the gradient. This procedure must define a complete energy calculation (orbital optimization and correlation treatment).
• VARIABLE=varname Use given variable for numerical calculation of the gradient.
• DISPLACE=type The displacement type. Note that the displacement type for gradient and hessian must be the same. type can be one of the following:
  • SYMM Use symmetric displacement coordinates (default). This is the only recommended option.
  • CART Use $3N$ cartesian displacements (not recommended). This requires many more energy calculations than necessary and does not preserve the molecular symmetry.
  • UNIQUE Use symmetry-unique cartesian displacements (not recommended)

ROOT,root [root]

Specifies the eigenvector of the hessian to be followed.

• root=1 specifies a minimization (default).
• root=2 specifies a transition state (saddle point) optimization.

In the present implementation a saddle point search is possible with the rational function method (METHOD,RF), the geometry DIIS method (METHOD,DIIS) and the quadratic steepest descent method of Sun and Ruedenberg (METHOD,SRTRANS).

Note that convergence is usually much more difficult to achieve than for minimizations. In particular, a good starting geometry and a good approximation to the hessian is needed. The latter is achieved by evaluating the hessian numerically (see section numerical Hessian (NUMHESS)) or using a precomputed hessian (see section hessian approximations (HESSIAN)).

Example:

examples/butane_opt_transition.inp

 ***, Bicyclo-[1.1.0]-butane Transition State
 basis=3-21G

 L1=1.495  ang                ! Define Active Variables
 L2=1.418  ang
 L3=1.463  ang
 L4=1.093  ang
 L5=1.111  ang
 L6=1.098  ang
 L7=1.097  ang
 L8=1.110  ang
 L9=1.106  ang
 A1=92.1   degree
 A2=62.1   degree
 A3=136.0  degree
 A4=123.5  degree
 A5=122.4  degree
 A6=124.7  degree
 A7=126.7  degree
 A8=117.9  degree
 D1=-120.4 degree
 D2=4.4    degree
 D3=108.8  degree
 D4=-107.5 degree
 D5=84.2   degree
 D6=109.3  degree
 D7=-106.1 degree

 geometry={C1                 ! Geometry Specification Z-Matrix
           C2  1 L1
           C3  2 L2 1 A1
           C4  1 L3 2 A2 3 D1
           H5  1 L4 2 A3 3 D2
           H6  2 L5 1 A4 4 D3
           H7  3 L6 2 A5 1 D4
           H8  3 L7 2 A6 1 D5
           H9  4 L8 1 A7 2 D6
           H10 4 L9 1 A8 2 D7}
 int
 rhf
 optg
 root,2                       ! Transition State search
 method,qsd                   ! Use Quadratic Steepest Descent Method

STEP,steplength,drmax,damax,drmax1,damax1 [optgeo:step]

steplength is the initial step length in the scaled parameter space (default 0.3). In the AH-method this is dynamically adjusted, and can have a maximum value ahmax (see TRUST).

drmax is the initial max change of distances (in bohr, default 0.3). In the AH-method this is dynamically adjusted up to a maximum value of drmax1 (default 0.5 bohr).

damax is the initial max change of angles (in degree, default 2). In the AH-method this is dynamically adjusted up to a maximum value of damax1 (default 10 degrees).

TRUST,ratio,ahmax

ratio determines the radius around the current minimum in which points are used to update the Hessian with the conjugate gradient method (default 0.5; see also UPDATE).

ahmax is the maximum step size allowed in the Augmented Hessian procedure. This refers to the scaled parameter space (default 0.5). The initial step size is stepmx (see STEP card).

CUT,threshold

Specifies a threshold for ortho-normalization used in conjugate gradient update of hessian (default 1.d-3; see also UPDATE).

LINESEARCH,iflag,thrlmin,thrlmax

Interpolate the geometry of the stationary point (minimum or saddle point) by a quartic polynomial between the current and the previous geometry. If iflag=0 or no iflag is set, the next optimization step will be taken from the interpolated geometry using the interpolated energy and gradient. If iflag=1 the energy and gradient will be recalculated at the interpolated geometry before taking the new optimization step. Note though, that the additional effort of recalculating the energy and gradient is usually not met by the increase of the convergence rate of the optimization. thrlmin and thrlmax are min and max thresholds for the recalculation of the energy and the gradient in case iflag=1. I.e. the recalculation just takes place if the interpolated geometry isn’t too close to the actual geometry thrlmin and isn’t too remote from the actual geometry thrlmax. Default values are thrlmin=0.001 and thrlmax=0.05 in the scaled parameter space of the optimization.

[optgeo:option] OPTION,key=param;
where key can be

• IDIR If starting at a transition state (or near a transition state) determine where to take the first step. If IDIR=0 is chosen, the first step will be towards the transition state. This is the default. If IDIR=1 is given in the input the first optimization step will be along the ”transition vector” i.e. the hessian eigenvector to the smallest eigenvalue which points down towards the minimum. If using a larger IDIR parameter, the first step will be larger; if using a negative value, the first step will be in the opposite direction.
• STPTOL If using an updated hessian matrix, this parameter determines what update to take. If the step size between two subsequent points on which the steepest decent lines are puzzled together is smaller than stptol (i.e. if we are close to a minimum) the BFGS update is used, otherwise it is Powell update. The default value of stptol is $1.d-6$. Note that the stepsize is also affected by the STEPMAX option.
• SLMAX This option is only valid with the old version of the reaction path following algorithm (i.e. METHOD,SRSTEEP). In this algorithm slmax determines the frequency of the recalculation of the numerical hessian. If the total step size of the last steps exceeds slmax the hessian will be recalculated, otherwise it will be updated. By default slmax is two times the maximum step size of the optimization step steplength (see STEP section setting a maximum step size (STEP)). If you are using METHOD,QSD, the SLMAX option is obsolete and the NUMHES command (see above) should be used instead.

These options can also be given on the OPTG command line.

VARIABLE,name; [optgeo:var]

Defines a variable name which holds the energy value to be optimized in using finite differences. By default, this is ENERGY(1) as set by the most recent program. Other variables which can be used are

• ENERGY(i) holds last energy for state i.
• ENERGR(i) holds last reference energy for state i.
• ENERGD(i) holds last Davidson corrected energy for state i.
• ENERGP(i) holds last Pople corrected energy for state i.
• ENERGC holds CCSD (QCI, BCCD) energy in CCSD(T) [QCI(T), BCCD(T)] calculations (single state optimization).
• ENERGT(1) holds CCSD(T) energy in CCSD(T) calculations (single state)
• ENERGT(2) holds CCSD[T] energy in CCSD(T) calculations (single state).
• ENERGT(3) holds CCSD-T energy in CCSD(T) calculations (single state).

These variables are set automatically by the CI and/or CCSD programs. It is the user’s responsibility to use the correct variable name; an error exit occurs if the specified variable has not been defined by the last program or the user.

Note: The use of the VARIABLE option triggers NUMERICAL, so optimization can be very inefficient!

PRINT,code=level,…;

Enables printing options. Usually level should be omitted or 0; values of $level>0$ produce output useful only for debugging. code can be

• HESSIAN prints the updated hessian matrix. Note that its diagonal elements are printed anyway.
• HISTORY prints the complete set of previous geometries, gradients and energies.
• GRADIENT prints extended gradient information
• OPT prints detailed information about the optimization process (mainly for debugging).

Several print options can be specified with one PRINT command.

[conical]

To optimize a conical intersection between two electronic states having the same spin, three vectors must be evaluated at SA-CPMCSCF level:

• Non-Adiabatic Derivative Coupling (DC).
• Gradient of the lower state (LSG).
• Gradient of the upper state (USG).

This requires three different CPMCSCF directives in the MULTI input:

CPMCSCF, NACM, $S_i$, $S_j$, ACCU=1.0d-7, record=record1.file
CPMCSCF, GRAD, $S_i$, SPIN=Spin of state $S_i$, ACCU=1.0d-7, record=record2.file
CPMCSCF, GRAD, $S_j$, SPIN=Spin of state $S_j$, ACCU=1.0d-7, record=record3.file

where $S_i$,$S_j$ are the electronic states in the usual format istate.istsym, and record[n].file specifies the name and the file number where CPMCSCF solutions should be stored. Parameter SPIN is half of the value in the WF card used to define the electronic state.

Things to remember:

• Specify always three different record.file on the CPMCSCF directives.
• Evaluate the CPMCSCF for USG always last.
• Skip the DC evaluation if the conical intersection involves states with different spin (e.g., a Singlet/Triplet crossing) because the coupling is then zero.

Three sets of FORCE commands (only two for Singlet/Triplet intersection) follow the MULTI input. They will be like:

FORCE
SAMC,record[n].file
CONICAL,record4.file[,NODC]

where record.file is one of the records containing CPMCSCF info and record4.file points to a free record used for internal storage by the CONICAL code. record4.file must be the same on all the CONICAL directives. Furthermore, the present implementation works properly only if file=1 on the CONICAL directive. The optional keyword NODC must be used in case of different spins (e.g., S/T crossing) when DC is not needed.

The actual optimization is performed using OPTG,STARTCMD=MULTI The example below optimizes the conical intersection in $LiH_2$ (ground and excited states are both doublets).

examples/lih2_D0D1.inp

***, LiH2

basis=sto-3g
print,orbitals,civector

symmetry,x              !use only molecular plane. Both states must be in the same symmetry.
geometry={
          Li;
          h1,Li,r;
          h2,Li,r,h1,theta}

r=3.0
theta=35

{hf;wf,4,1,0}

{multi;occ,6,1;wf,5,1,1;state,2                       !state averaged casscf
CPMCSCF,NACM,1.1,2.1,accu=1.0d-7,record=5100.1        !cpmcscf for non-adiabatic couplings
CPMCSCF,GRAD,1.1,spin=0.5,accu=1.0d-7,record=5101.1   !gradient for state 1
CPMCSCF,GRAD,2.1,spin=0.5,accu=1.0d-7,record=5102.1}  !gradient for state 2

{Force
SAMC,5100.1             !compute coupling matrix element
CONICAL,6100.1}         !save information for optimization of conical intersection

{Force
SAMC,5101.1             !compute gradient for state 1
CONICAL,6100.1}         !save information for optimization of conical intersection

{Force
SAMC,5102.1             !compute gradient for state 2
CONICAL,6100.1}         !save information for optimization of conical intersection

optg,startcmd=multi     !find conical intersection

This second example optimizes the singlet-triplet intersection in $LiH_2(+)$ (ground state is Singlet, excited state is Triplet).

examples/lih2+_S0T0.inp

***, LiH2

basis=sto-3g

symmetry,nosym
geometry={
          Li;
          H1,Li,r;
          H2,Li,r,H1,theta}

r=3.7
theta=160

{hf;wf,4,1,0}

{multi;
 occ,7;
 wf,4,1,0;   !singlet state
 wf,4,1,2;   !triplet state
 CPMCSCF,GRAD,1.1,spin=0,accu=1.0d-7,record=5101.1   !cpmcscf for gradient of singlet state
 CPMCSCF,GRAD,1.1,spin=1,accu=1.0d-7,record=5100.1   !cpmcscf for gradient of triplet state
}

{Force
 SAMC,5101.1                !state averaged gradient for singlet state
 CONICAL,6100.1,NODC}       !save information for OPTCONICAL

{Force
 SAMC,5100.1                !state averaged gradient for triplet state
 CONICAL,6100.1,NODC}       !save information for OPTCONICAL

optg,startcmd=multi,gradient=1.d-6  !find singlet-triplet crossing point

It is optionally possible to use the SLAPAF program written by Roland Lindh for geometry optimizations. This is done by prepending the optimization method with ’SL’. The following methods are supported:

• SLRF Use the rational function approximation;
• SLNR Use the Newton-Raphson method;
• SLC1 Use the C1-DIIS method;
• SLC2 Use the C2-DIIS method.

When using DIIS methods (SLC1 or SLC2), the DIIS parameters are specified in the same way as in standard molpro optimizer.

There are some differences when using the SLAPAF program:

• It is not possible to use Z-matrix coordinates in the optimization.
• Instead, one can explicitly define internal coordinates to be varied or fixed.
• Additional constraints can be imposed on the converged geometry in a flexible way.

Constraints and internal coordinates (see below) can be linear combinations of bonds, angles etc. The latter, called here primitive internal coordinates, can be specified before the constraints definition, or directly inside. The general definition of a primitive coordinate is:

PRIMITIVE,[NAME=]symbolic name, explicit definition;

or

PRIM,[NAME=]symbolic name, explicit definition;

Here symbolic name is the name given to the primitive coordinate (if omitted, it will be generated automatically). This name is needed for further reference of this primitive coordinate.

explicit definition has the form:

type,atoms

type can be one of the following:

• BOND Bond length, defined by 2 atoms.
• ANGLE Bond angle, defined by 3 atoms (angle 1–2–3).
• DIHEDRAL Dihedral angle, defined by 4 atoms (angle between the planes formed by atoms 1,2,3 and 2,3,4, respectively).
• OUTOFPLANE Out-of-plane angle, defined by 4 atoms (angle between the plane formed by atoms 2,3,4 and the bond 1–4).
• DISSOC A dissociation coordinate, defined by two groups of atoms (Not permitted with constraints).
• CARTESIAN Cartesian coordinates of an atom.

For all types except DISSOC and CARTESIAN, atoms are given as:

ATOMS=[a1,a2,a3,…]

where the number of atoms required varies with type as specified above, and the atomic names a1,a2,a3,… can be either atomic tag names from the Z-matrix input, or integers corresponding to Z-matrix rows. Note that the square brackets are required here and do not indicate optional input.

For DISSOC the specification is as follows:

DISSOC,GROUP1=[a1,a2,…],GROUP2=[b1,b2,…];

The corresponding internal coordinate is the distance between the centres of mass of the two groups.

For CARTESIAN the definition is

CARTESIAN, I, atom;

where I can be one of X,Y,Z or 1,2,3 and atom can be a z-matrix atom name or an integer referring to the z-matrix row.

With this definition, the constraints are defined as

CONSTRAINT,[VALUE=]value,[unit],[[FACTOR=]fac,prim,[[FACTOR=]fac],prim,…;

where value is the value imposed to the constraint, and prim is either the name of the primitive defined before this constraint, or an explicit definition; and fac is a factor of the corresponding primitive in the constraint. If fac is omitted it is taken to be 1.

If value is specified in Angstrom or Radian, unit must be given.

Examples for H$_2$O in $C_s$ symmetry:

Constraining the bond angle to 100 degrees:

constraint,100,deg,angle,atoms=[h1,o,h2];

which is equivalent to

primitive,a1,angle,atoms=[h1,o,h2];

constraint,100,a1;

Keeping the two OH distances equal:

constraint,0,bond,atoms=[h1,o],-1.,bond,atoms=[h2,o];

which is equivalent to

primitive,b1,bond,atoms=[h1,o];

primitive,b2,bond,atoms=[h2,o];

constraint,0,b1,-1.,b2;

By default SLAPAF optimizes in force-constant weighted normal coordinates that are determined automatically. However, the user can define his own coordinates. The definition of internal coordinates, similar to constraints, is based on primitive coordinates. The input is:

INTERNAL,[[NAME=]name],[[FACTOR=]fac],prim,[[FACTOR=]fac],prim,…;

FIX, [[NAME=]name],[[FACTOR=]fac],prim,[[FACTOR=]fac],prim,…;

Internal coordinates that are specified using INTERNAL are varied and those using FIX are fixed to their initial values.

An important point for the definition of internal coordinates is that their total number must be equal to the number of degrees of freedom of the molecule . Otherwise an error message is generated. Only symmetry independent coordinates need to be given.

Some options can be passed to the SLAPAF program. Options are specified with SLOPT subdirective:

{opt;method=slnr;{slopt;opt1;opt2,par1,par2;opt3;...}}

The available options are

• CART Use eigenvectors of the approximate Hessian, expressed in cartesian coordinates, as the definition of internal coordinates;
• NOMA Don’t impose any restrictions on the step size;
• UORD Order the gradients and displacement vectors according to Schlegel prior to the update of the Hessian. Default is no reordering;
• HWRS Use force field weighted internal coordinates (default);
• RS-P Activate RS-P-RFO as default for transition state search; default is RS-I-RFO;
• NOHW Use unweighted internal coordinates;
• PRBM Print B-matrix;
• RTHR,Thra,Thrb,Thrt Thresholds for redundant coordinate selection for bonds, bends and torsions, respectively. Default 0.2, 0.2, 0.2
• MODE,index Hessian vector index for mode following when calculating transition states.
• FIND Enable unconstrained optimization for constrained cases, when looking for transition states (see MOLCAS manual).
• GNRM,thr Threshold for FIND, default 0.2 (see MOLCAS manual).

For more information, please consult the MOLCAS manual.

examples/h2o_optmp2.inp

***,H2O                      !title
include procedures
basis=vtz                    !use cc-pVTZ basis
geometry={O;                 !Z-matrix for water
          H1,O,R;
          H2,O,R,H1,THETA}
R=0.96 Ang                   !start bond distance
Theta=104                    !start bond angle
optmp2                       !do MP2 geometry optimization
show,R,Theta                 !show optimized geometry parameters

examples/h2o_optmp2_runccsdt.inp

include procedures
geometry={O;                 !Z-matrix for water
          H1,O,R;
          H2,O,R,H1,THETA}
basis=vtz                    !use VTZ basis
R=0.96 Ang                   !start bond distance
Theta=104                    !start bond angle

hf
mp2
optg    !optimize MP2 energy
mp4     !do single-point MP4 at MP2 geometry
ccsd(t) !do single-point CCSD(T) at MP2 geometry

There is an additional example in the introductory examples.

examples/allene_optscf.inp

 ***, Allene geometry optimization using Z-Matrix
 basis=sto-3g

 rcc=1.32 ang
 rch=1.08 ang
 acc=120  degree
 Geometry={C1                       !Z-matrix input
           C2,c1,rcc
           X1,c1,rcc,c2,45
           C3,c2,rcc,c1,180,x1,0
           h1,c1,rch,c2,acc,x1,0
           h2,c1,rch,c2,acc,h1,180
           h3,c3,rch,c2,acc,h1,90
           h4,c3,rch,c2,acc,h2,90}
 hf
 optg,saveact='allene.act',savexyz='allene.xyz' !default optimization
                                                !using model hessian.
                                                !Save optimized variables
                                                !in file allene.act
                                                !Save optimized geometry
                                                !in xyz style in
                                                !file allene.xyz

examples/allene_opt_bmat.inp

 ***, Allene geometry optimization using natural internal coordinates
 basis=sto-3g

 rcc=1.32 ang
 rch=1.08 ang
 acc=120  degree
 symmetry,nosym
 Geometry={C1;                       !Z-matrix input
           C2,c1,rcc
           X1,c1,rcc,c2,45
           C3,c2,rcc,c1,180,x1,0
           h1,c1,rch,c2,acc,x1,0
           h2,c1,rch,c2,acc,h1,180
           h3,c3,rch,c2,acc,h1,90
           h4,c3,rch,c2,acc,h2,90}
 hf;
 optg                !default optimization using model hessian
 coord,bmat          !use natural internal coordinates

 optg,coord=bmat     !same as above

examples/allene_optmp2.inp

 ***, Allene geometry optimization using Z-Matrix
 basis=vdz

 rcc=1.32 ang
 rch=1.08 ang
 acc=120  degree
 Geometry={C1                       !Z-matrix input
           C2,c1,rcc
           X1,c1,rcc,c2,45
           C3,c2,rcc,c1,180,x1,0
           h1,c1,rch,c2,acc,x1,0
           h2,c1,rch,c2,acc,h1,180
           h3,c3,rch,c2,acc,h1,90
           h4,c3,rch,c2,acc,h2,90}

optg,procedure=runmp2     !use procedure optmp2

runmp2={hf;mp2}         !procedure definition

examples/caffeine_opt_diis.inp

 ***, CAFFEINE cartesian coordinates (XYZ format)
 basis=sto-3g
 geomtyp=xyz
 geometry={
 24
            CAFFEINE CARTESIAN COORDINATES
 C          0.8423320060       -0.3654865620        0.0000000000
 C         -0.2841017540       -1.1961236000        0.0000000000
 N          2.0294818880       -1.1042264700        0.0000000000
 N          0.0774743850       -2.5357317920        0.0000000000
 N         -1.6472646000       -0.6177952290        0.0000000000
 C          1.4531962870       -2.3678913120        0.0000000000
 C          0.6373131870        1.1735112670        0.0000000000
 C         -1.7812691930        0.7688916330        0.0000000000
 N         -0.6771444680        1.6306355000        0.0000000000
 O          1.6106752160        1.9349693060        0.0000000000
 O         -2.9202890400        1.2510058880        0.0000000000
 C         -0.9202462430        3.1094501020        0.0000000000
 C         -2.8623938560       -1.4824503660        0.0000000000
 C          3.4552156930       -0.6811094280        0.0000000000
 H          2.0878150460       -3.2451913360        0.0000000000
 H         -1.4989252090        3.4222116470       -0.8897886280
 H         -1.4989252090        3.4222116470        0.8897886280
 H          0.0071905670        3.7148499490        0.0000000000
 H         -3.4903070930       -1.2888938190       -0.8907763360
 H         -3.4903070930       -1.2888938190        0.8907763360
 H         -2.6289534570       -2.5638654230        0.0000000000
 H          4.1360211370       -1.5529079440        0.0000000000
 H          3.6817059520       -0.0685850980        0.8931597470
 H          3.6817059520       -0.0685850980       -0.8931597470
 }

 hf
 optg,savexyz=caffeine.xyz   !save optimized geometry in file caffeine.xyz
 coord,bmat                  !Optimization in natural internal coordinates
 method,diis                 !Optimization method: Geometry DIIS

 optg,coord=bmat,method=diis,savexyz=caffeine.xyz   !same as above

The first example shows how to do a MP2 transition state optimization. The initial Hessian is taken from a previous HF frequency calculation.

examples/hcn_mp2_ts.inp

***, HCN <--> NHC Isomerization - Transition State Optimization and Frequencies

l1=1.18268242 ang
l2=1.40745082 ang
a1=55.05153416 degree

basis=3-21G

symmetry,nosym
geometry={
          C
          N,1,l1
          H,2,l2,1,a1}

hf                       ! HF-SCF

frequencies,analytical   ! Vibrational frequencies for HF-SCF (analytical Hessian)

mp2                      ! MP2

optg,root=2,method=rf,readhess     ! Transition State Search using Rational Function
                                   ! Optimizer and HF hessian

frequencies              ! Vibrational frequencies for MP2 (numerical Hessian)
---

The second example shows how to do a CCSD(T) optimization with an MP2 hessian. Note that currently the CCSD(T) gradient is computed numerically using finite energy differences, and this can take long time for larger molecules. The calculation of the MP2 hessian finite differences of analytical gradients.

examples/hcn_ccsd_ts.inp

***, HCN <--> NHC Transition State Optimization and Frequencies

rcn=1.18 ang
rnh=1.40 ang
alpha=55 degree

basis=vtz

geometry={
          C
          N,1,rcn
          H,2,rnh,1,alpha}

hf
ccsd(t)
optg,numerical,root=2,hessproc=runmp2  !Transition state optimization for ccsd(t)
                                       !using mp2 hessian
frequencies                  !CCSD(T) frequencies (using numerical second derivatives)

runmp2={hf;mp2}    !procedure definition
---

The last example shows how to do a MRCI+Q (MRCI with Davidson correction) optimization with an CASPT2 hessian. As for CCSD(T), the MRCI+Q gradient as computed numerically, while the CASPT2 hessian is obtained using finite differences of analytical CASPT2 gradients.

examples/hcn_mrci_ts.inp

***, HCN <-> NHC Isomerization - Transition State Optimization and Frequencies
print,orbitals,civector
rcn=1.18 ang
rnh=1.40 ang
alpha=55 degree

basis=vtz

geometry={
          C
          N,1,rcn
          H,2,rnh,1,alpha}

closed,4      ! global setting for casscf inactive space

hf            ! HF-SCF
multi         ! CASSCF
mrci          ! MRCI
optg,root=2,variable=energd,hessproc=runrs2   !optimize mrci+q transition state
                                              !and caspt2 for hessian

runrs2={multi;rs2}       !procedure definition for caspt2
---

The following input first optimizes the transition state, and then performs reaction path calculations in both directions. The results are plotted.

examples/hcn_isomerization.inp

 ***, HCN <---> NHC Isomerization Reaction Path
 basis=3-21G

 rcn=1.18282 ang      ! Starting geometry is transition state
 rnh=1.40745 ang
 alpha=55.05 degree

 symmetry,x           ! Cs Symmetry
 geometry={
           C
           N,1,rcn
           H,2,rnh,1,alpha}

 int
 rhf,gap_min=0     !set default shift to zero to avoid convergence to wrong state
 optg,root=2,saveact=hcn_ts,rewind            ! Find and store the TS
 {optg,method=qsdpath,dir=1, numhess=5,hesscentral,saveact=hcn_path}
                                              ! find IRC in positive direction

 readvar,hcn_ts.act                           ! Reset geometry to TS
 {optg,method=qsdpath,dir=-1,numhess=5,hesscentral,saveact=hcn_path,append}
                                              !find IRC in negative direction

 readvar,hcn_path.act

 alpha=alpha*pi/180    !convert angle to radian

 table,irc,rcn,rnh,alpha,e_opt   !tabulate results

 {table,irc,e_opt                !plot energy profile as function of irc
  plot,file='hcn_eopt.plot'}

 {table,irc,rcn,rnh,alpha        !plot distances and angle as function of irc
  plot,file='hcn_dist.plot'}

This produces the plots

Geometry optimization of counterpoise corrected energies is possible by performing for the total system as well as for each individual fragment separate FORCE calculations. The gradients and energies are added using the ADD directive. This requires that NOORIENT has been specified in the geometry input, in order to avoid errors due to unintended rotation of the system. This default can be disabled using the NOCHECK option, see ADD above.

The way a counterpoise corrected geometry optimization works is shown in the following example. Note that the total counterpoise corrected energy must be optimized, not just the interaction energy, since the interaction energy depends on the monomer geometries and has a different minimum than the total energy. The interaction energy could be optimized, however, if the monomer geometries were frozen. In any case, the last calculation before calling OPTG must be the calculation of the total system at the current geometry (in the example below the dimer calculation), since otherwise the optimizer gets confused.

examples/hfdimer_cpcopt1.inp

***,HF dimer MP2/CP optimization with relaxed monomers

basis=avtz
gthresh,energy=1.d-8

! INITIAL VALUES OF GEOMETRY VARIABLES

RFF=      5.3
R1=       1.76
R2 =      1.75
THETA1 =  7.0
THETA2 = 111

symmetry,x
orient,noorient
geometry={
      f1
      f2  f1  rff
      h1  f1  r1   f2  theta1
      h2  f2  r2   f1  theta2  h1  180.}

label:

text, CALCULATION AT LARGE SEPARATION

rff_save=rff            !save current rff distance
rff=1000                !dimer calculation at large separation

text, HF1

dummy,f2,h2;            !second hf is now dummy
{hf;accu,16}            !scf for first monomer
mp2;                    !mp2 for first monomer
ehf1inf=energy          !save mp2 energy in variable
forces;                 !compute mp2 gradient for first monomer

text, HF2

dummy,f1,h1;            !first hf is now dummy
{hf;accu,16}            !scf for second monomer
mp2;                    !mp2 for second monomer
ehf2inf=energy          !save mp2 energy in variable
forces;                 !compute mp2 gradient for second monomer
add,1                   !add to previous gradient
einf=ehf1inf+ehf2inf    !total energy of unrelaxed momomers

rff=rff_save            !reset HF - HF distance to current value

text, CP calculation for HF1 MONOMER

dummy,f2,h2;            !second hf is now dummy
{hf;accu,16}            !scf for first monomer
mp2;                    !mp2 for first monomer
ehf1=energy             !save mp2 energy in variable
forces;                 !compute mp2 gradient for first monomer
add,-1                  !subtract from previous gradient

text, CP calculation for HF2 MONOMER

dummy,f1,h1;            !first hf is now dummy
{hf;accu,16}            !scf for second monomer
mp2;                    !mp2 for second monomer
ehf2=energy             !save mp2 energy in variable
forces;                 !compute mp2 gradient for first monomer
add,-1                  !subtract from previous gradient

text, DIMER CALCULATION
dummy                   !reset dummies
{hf;accu,16}            !scf for dimer
mp2;                    !mp2 for dimer
edimer=energy           !save mp2 energy in variable
forces;                 !compute mp2 gradient for dimer
add,1                   !add to previous gradient

optg,gradient=1.d-4,startcmd=label:   !find next energy


text, compute optimized monomer energy

rhf=r1
geometry={h1
          F1,H1,rhf}

{hf;accu,16}            !scf for relaxed monomer
mp2;                    !mp2 for relaxed monomer
ehf=energy              !save mp2 energy in variable
optg                    !optimize HF structure

text,optimized geometry parameters
show,r1,r2,rhf,rff,theta1,theta2

text,computed interaction energies
decpc=(einf-ehf1-ehf2)*tocm    !counter poise correction
de=(edimer-ehf1-ehf2)*tocm     !CPC corrected interaction energy relative to unrelaxed monomers
erelax=(2*ehf-einf)*tocm       !relaxation energy
derel=de-erelax                !CPC corrected interaction energy relative to relaxed monomoers

The next example shows how the same calculations can be done using numerical gradients. In this case, first the total counter-poise corrected energy is formed and then optimized. Note that the ADD command does not work for numerical gradients.

examples/hfdimer_cpcopt1_num.inp

***,HF dimer MP2/CP optimization with relaxed monomers

basis=avtz
gthresh,energy=1.d-8

! INITIAL VALUES OF GEOMETRY VARIABLES

RFF=      5.3
R1=       1.76
R2 =      1.75
THETA1 =  7.0
THETA2 = 111

symmetry,x
orient,noorient
geometry={
      f1
      f2  f1  rff
      h1  f1  r1   f2  theta1
      h2  f2  r2   f1  theta2  h1  180.}

label:

text, CALCULATION AT LARGE SEPARATION

rff_save=rff            !save current rff distance
rff=1000                !dimer calculation at large separation

text, HF1

dummy,f2,h2;            !second hf is now dummy
{hf;accu,16}            !scf for first monomer
mp2;                    !mp2 for first monomer
ehf1inf=energy          !save mp2 energy in variable

text, HF2

dummy,f1,h1;            !first hf is now dummy
{hf;accu,16}            !scf for second monomer
mp2;                    !mp2 for second monomer
ehf2inf=energy          !save mp2 energy in variable
einf=ehf1inf+ehf2inf    !total energy of unrelaxed momomers

rff=rff_save            !reset HF - HF distance to current value

text, CP calculation for HF1 MONOMER

dummy,f2,h2;            !second hf is now dummy
{hf;accu,16}            !scf for first monomer
mp2;                    !mp2 for first monomer
ehf1=energy             !save mp2 energy in variable

text, CP calculation for HF2 MONOMER

dummy,f1,h1;            !first hf is now dummy
{hf;accu,16}            !scf for second monomer
mp2;                    !mp2 for second monomer
ehf2=energy             !save mp2 energy in variable

text, DIMER CALCULATION
dummy                   !reset dummies
{hf;accu,16}            !scf for dimer
mp2;                    !mp2 for dimer
edimer=energy           !save mp2 energy in variable
etot=edimer-ehf2-ehf1+ehf1inf+ehf2inf  !total BSSE corrected energy

optg,numerical,variable=etot,gradient=1.d-4,startcmd=label:   !optimize geometry


text, compute optimized monomer energy

rhf=r1
geometry={h1
          F1,H1,rhf}

{hf;accu,16}            !scf for relaxed monomer
mp2;                    !mp2 for relaxed monomer
ehf=energy              !save mp2 energy in variable
optg                    !optimize HF structure

text,optimized geometry parameters
show,r1,r2,rhf,rff,theta1,theta2

text,computed interaction energies
decpc=(einf-ehf1-ehf2)*tocm    !counter poise correction
de=(edimer-ehf1-ehf2)*tocm     !CPC corrected interaction energy relative to unrelaxed monomers
erelax=(2*ehf-einf)*tocm       !relaxation energy
derel=de-erelax                !CPC corrected interaction energy relative to relaxed monomoers

In the last example the monomer structures are kept fixed, and the interaction energy is optimized.

examples/hfdimer_cpcopt2.inp

***,HF dimer MP2/CP optimization without monomer relaxation

basis=avtz
gthresh,energy=1.d-8

! INITIAL VALUES OF GEOMETRY VARIABLES

RFF=      5.3
THETA1 =  7
THETA2 = 111

symmetry,x
orient,noorient
geometry={
      f1
      f2  f1  rff
      h1  f1  1.74764059   f2  theta1
      h2  f2  1.74764059   f1  theta2  h1  180.}   !using fixed HF distances of isolated HF

label:

text, CP calculation for HF1 MONOMER

dummy,f2,h2;            !second hf is now dummy
{hf;accu,16}            !scf for first monomer
mp2;                    !mp2 for first monomer
ehf1=energy             !save mp2 energy in variable
forces;                 !compute mp2 gradient for first monomer
scale,-1                !multiply gradient by -1

text, CP calculation for HF2 MONOMER

dummy,f1,h1;            !first hf is now dummy
{hf;accu,16}            !scf for second monomer
mp2;                    !mp2 for second monomer
ehf2=energy             !save mp2 energy in variable
forces;                 !compute mp2 gradient for first monomer
add,-1                  !subtract from previous gradient

text, DIMER CALCULATION
dummy                   !reset dummies
{hf;accu,16}            !scf for dimer
mp2;                    !mp2 for dimer
edimer=energy           !save mp2 energy in variable
forces;                 !compute mp2 gradient for dimer
add,1                   !add to previous gradient

optg,gradient=1.d-5,startcmd=label:   !find next energy

text,optimized geometry parameters
show,rhf,rff,theta1,theta2

text,computed interaction energies
de=(edimer-ehf1-ehf2)*tocm     !CPC corrected interaction energy with fixed monomers

Geometry optimization using CCSD(T)-F12 method including CABS singles corrections is possible by combining numerical gradients for HF+CABS singles method and analytical gradients for DF-CCSD(T)-F12 method obtained in separate FORCE calculations. The gradients are added using the ADD directive. Similarly to the example optimizing counterpoise corrected energies, NOORIENT is required. The following example shows geometry optimization for the hydrogen fluoride molecule using DF-CCSD(T)-F12b method without and with CABS singles correction.

examples/dfccsdtf12_3stc_tight_opt.inp

***, geometry optimization (tight thresholds) for HF using DF-CCSD(T)-F12b/3*C(FIXC,HY1) without and with CABS singles
gthresh,energy=1.d-10,gradient=1.d-08

nosym
noorient
angstrom
geometry={
    2

 F          0.0000000000        0.0000000000       -0.0463705158
 H          0.0000000000        0.0000000000        0.8745302823
}


basis=avdz

! perform geometry optimization using CCCSD(T)-F12b/3*C(FIXC,HY1) without CABS singles by using analytical derivatives
{df-hf;accu,18;}
{df-ccsd(t)-f12,ansatz=3*C(fixc,hy1),cabs_singles=0,cabs=0,ri_basis=optri,thrcabs=1.d-10,thrcabs_rel=1.d-10,thrabs=1.d-10,thrabs_rel=1.d-10,gem_beta=1.0d0;cphf,thrmin=1.d-9;}
{optg,energy=1.d-08,gradient=1.d-06;}

! save geometry parameters for printing
struct,distvar=ccf12distwocabs,angvar=ccf12angwocabs,dihedvar=ccf12dihedwocabs

! perform geometry optimization using CCCSD(T)-F12b/3*C(FIXC,HY1) with CABS singles by using numerical and analytical derivatives
label1
! compute CABS singles forces by using numerical derivatives
{df-hf;accu,18}
{df-mp2-f12,cabs_singles=-1,ri_basis=optri,thrcabs=1.d-10,thrcabs_rel=1.d-10,thrabs=1.d-10,thrabs_rel=1.d-10,gem_beta=1.0d0}
{forces,variable=ef12_singles,varsav=on,fourpoint}
! compute CCCSD(T)-F12b/3*C(FIXC,HY1) without CABS singles forces by using analytical derivatives
{df-hf;accu,18;}
{df-ccsd(t)-f12,ansatz=3*C(fixc,hy1),cabs_singles=0,cabs=0,ri_basis=optri,thrcabs=1.d-10,thrcabs_rel=1.d-10,thrabs=1.d-10,thrabs_rel=1.d-10,gem_beta=1.0d0;cphf,thrmin=1.d-9;}
{forces,varsav=on;add,1} ! add the forces
{optg,startcmd=label1,energy=1.d-08,gradient=1.d-06;}

! save geometry parameters for printing
struct,distvar=ccf12dist,distdef=distance,angvar=ccf12ang,angdef=angle,dihedvar=ccf12dihed,diheddef=dihedral

! compute CABS singles correction for bond length
distcorr=ccf12dist-ccf12distwocabs

! print results
table,distance,ccf12distwocabs,ccf12dist,distcorr