Geometry Optimization and Transition-State Keywords in the Jaguar Input File

Many of the keyword settings for optimization of minimum-energy structures and transition states described in this subsection can be made from the GUI, as described in Jaguar Optimizations and Scans, which also contains more details about the methods used for optimizations.

Table 1 contains general optimization keywords that apply to all kinds of optimizations. Most default values for the integer keywords are indicated in bold italics, and only the values listed in the table are allowed. In cases where the default is different for optimizations to minimum-energy structures than it is for transition-state optimizations, both defaults are in bold italics, and the cases for which each is a default are explained in the keyword description.

Table 1. General geometry optimization keywords

Keyword

Value

Description

igeopt

0

Do not optimize molecular geometry

 

1

Optimize minimum-energy structure

 

−1

Calculate forces but do not perform geometry optimization

 

2

Optimize transition-state geometry

nogas

0

For optimizations in solution, perform gas phase geometry optimization first (to get accurate solvation energy). If nogas is set to 0 for single point energy (igeopt=0) or force calculations (igeopt=-1) in solution, nogas=2 is used instead.

 

1

For optimizations in solution, skip gas phase geometry optimization and compute solvation energies using esolv0 value (from input file) as gas phase energy (should yield same structure as nogas=0). This option can also be used with single point energy (igeopt=0) or force calculations (igeopt=-1) in solution.

 

2

For optimizations in solution, skip gas phase geometry optimization and compute solvation energies using energy of initial structure as gas phase energy (should yield same structure as nogas=0). This is the default for the PCM solvation model, as well as single point energy (igeopt=0) or force calculations (igeopt=-1) in solution.

maxitg

>0

Maximum number of optimization iterations (maximum number of structures generated); default is 100.

nmder

0

If calculating forces, compute analytic derivatives of energy

 

1

If calculating forces, compute numerical derivatives of energy (obtained from calculations on 6 Natom perturbed geometries by moving each atom pertnd bohr in positive or negative x, y, or z direction)

 

2

Calculate frequencies numerically

intopt

0

Use Cartesian coordinates for optimization

 

1

Use internally generated redundant internal coordinates for optimization (including any from coord or connect sections, if they exist)

 

2

Use internal coordinates from input Z‑matrix for optimization (note: if geometry input is in Cartesian format or contains a second bond angle rather than a torsional angle for any atom, intopt is reset to 1)

intopt_switch

0

Do not switch from internal coordinates to Cartesian coordinates during optimization.

 

1

Switch from internal coordinates to Cartesian coordinates during optimization if the optimization gets stuck.

 

2

Switch from internal coordinates to Cartesian coordinates during optimization if the calculation is converged to within a factor of intopt_switchfac of the convergence thresholds.

intopt_switchfac

10.0

Factor to be applied to convergence threshold to determine when to switch from internal coordinates to Cartesian coordinates during optimization. This applies for intopt_switch=2.

optcoord_newhist

0

Do not discard optimization history when switching coordinates.

 

1

Discard optimization history when switching coordinates.

 

2

Discard optimization history when switching coordinates and reset the Hessian if the calculation is an IRC calculation.

optcoord_update

0

Do not run checks for changing the internal coordinate set used for optimization.

 

1

Run checks for changing the internal coordinate set used for optimization.

nooptr

0

Optimize all bond lengths not specifically constrained in zmat section

 

1

Constrain (freeze) all bond lengths for optimization

noopta

0

Optimize all bond angles not specifically constrained in zmat section

 

1

Constrain (freeze) all bond angles for optimization

nooptt

0

Optimize all torsional angles not specifically constrained in zmat section

 

1

Constrain (freeze) all torsional angles for optimization

ip472

1

Do not save intermediate structures in the .mae output file.

 

2

Save intermediate structures in the .mae output file. This allows you to plot the energy during the course of the geometry optimization, using the Workflow Action Menu in Maestro.

props_each_step

0

Do not calculate properties at each step in a geometry optimization.

 

1

Calculate specified properties at each step in a geometry optimization. The properties you can calculate are ESP, Mulliken, and stockholder charges, multipole moments, and Fukui indices. The relevant keywords for the properties must also be included in the gen section (icfit, mulken, stockholder_q, ldips, fukui).

needgwd

0

Do not compute DFT grid weight derivatives

 

1

Compute DFT grid weight derivatives (and second derivatives for CPHF)

 

2

Compute DFT grid weight derivatives and gradient from grid translation (symmetry will be turned off)

 

3

Compute DFT grid weight derivatives and gradient from grid translation and rotation (symmetry will be turned off)

iaccg

2

Use convergence criteria shown in Table 7 This setting is used by default for gas-phase calculations, as well as SMD and PCM implicit solvent model calculations.

 

3

Use a looser criteria; a factor of 5 times the criteria in Table 7

 

4

A factor of 3 times the criteria in Table 7. This setting is

used by default for PBF implicit solvent model calculations.

 

5

Use tight criteria; a factor of 0.1 times the criteria in Table 7

scale_geoconv

1.0

Scaling factor for gradient convergence criteria. Each of the values in Table 7 is scaled by this factor.

geoconv_mode

standard

Use standard criteria for convergence, in which all five criteria must be met (energy difference, maximum gradient, rms gradient, maximum displacement, rms displacement).

 

flexible

Assess convegence based on a score, with one point for meeting a criterion, an extra point if a gradient or displacement criterion is met to 0.2 of the threshold, an extra point if the energy change is less than 0.05 of threshold and another if it is less than 0.005 of threshold. The optimization is converged if the score is 5 or more points.

ignore_displacement_criteria

0

Convergence criteria set for displacement RMS (gconv6) and maximum displacement (gconv5) are used when considering convergence.

 

1

Convergence criteria for displacement RMS (gconv6) and maximum displacement (gconv5) are not used when considering convergence.

nops_opt_switch

>0

Specifies convergence cutoff for switching from pseudospectral to analytic integral evaluation in a geometry optimization. The value multiplies the standard convergence criterion, and a switch to nops=1 is made when convergence is reached with the scaled convergence cutoff. The value must be greater than 1 for switching to take place. Not available with IRC or check_min > 0.

ngeorestart

≥0

Maximum number of automatic restarts used for a “stuck” geometry optimization. Restarting discards the Hessian history and sets iacc=1. Default is 1

nogdiis

0

Use GDIIS method (Geometry optimization by Direct Inversion in the Iterative Subspace) [203] to get new geometry

 

1

Don’t use GDIIS method

ilagr

<0

Zero out gradients along frozen coordinates

 

0 or 1

Project out gradient components in constrained coordinates

 

>1

Apply constraints using Lagrange multipliers

constraint_exponent

2

Exponent of monomial in flat-bottomed constraints. The default value produces a harmonic constraint. See Applying Harmonic Constraints in Jaguar Calculations.

die_on_opt_failure

0

Allow the calculation to continue if the optimization does not converge. This permits analysis and subsequent use of the results. Default if nofail=1.

 

1

Stop the calculation if the optimization does not converge.

nofail_lowest_e

0

Return the geometry from the final iteration on geometry convergence failure, when nofail=1. Default if check_min>0.

 

1

Return the geometry that gives the lowest energy on geometry convergence failure, when nofail=1.

Table 2 lists keywords for analyzing geometry convergence and checking whether a minimum has been obtained when requested. The checking feature is not supported with dynamic constraints, or for scans or IRC calculations.

Table 2. Geometry optimization keywords for analyzing and checking convergence

Keyword

Value

Description

check_min

0

Do not check if the stationary point is a minimum.

 

1

Check for a minimum-energy structure by perturbing the initial converged geometry using the QM Hessian, which is calculated. The geometry is perturbed and re-optimized only if the QM Hessian contains negative eigenvalues, i.e., a lower energy geometry is found.

 

2

Check for a minimum-energy structure by perturbing all rotatable proper and improper torsions (excluding those in rings and double bonds). If the geometry can be perturbed in this way, a re-optimization is performed from this perturbed geometry.

 

3

Check for a minimum-energy structure by perturbing overlapping torsions and planar inversions (excluding those in rings and double bonds), and torsions involving potential internal hydrogen bonds in which the hydrogen atom (donor) is symmetrically placed between two acceptors. If the geometry can be perturbed in this way, a re-optimization is performed from this perturbed geometry.

check_min_update

0

The initial converged geometry is written to the output .mae. During optimization of the perturbed geometry, the output .mae file’s geometry will not be updated at each optimization step. The output .mae file will only be updated after the convergence of the perturbed structure.

 

1

The initial converged geometry is written to the output .mae. During optimization of the perturbed geometry, the output .mae file’s geometry will be updated at each optimization step. The output .mae file will also be updated after the convergence of the perturbed structure.

check_min_return

0

Do not check that the optimization of the perturbed structure returned to the initial converged structure.

 

1

Check that the optimization of the perturbed structure returned to the initial converged structure. If optimization of a perturbed geometry does return to the initial minimum, report this then exit from the optimization of the perturbed geometry. The criteria for checking are given by the keywords check_min_ergtol and check_min_geotol.

check_min_eigcut

-5.0

Cutoff value for negative Hessian eigenvalues, in cm−1. The Hessian used for this check is not mass-weighted, i.e., all masses are set to unity. The Hessian is represented in coordinates set by the intopt keyword. Eigenvalues above this cutoff do not count towards the number of negative eigenvalues used to perturb the geometry when check_min=1.

check_min_ergtol

0.00008

Energy tolerance in hartrees for determining if the geometry is returning back to the initial converged structure. If the energy difference between the perturbed geometry at any optimization step and the initial minimum is less than this amount, the geometry is considered to have returned to the initial structure.

check_min_geotol

0.02

Geometry tolerance in bohrs/radians for determining if the geometry is returning back to the initial converged structure. If the rms displacement between the perturbed geometry at any optimization step and the initial minimum is less than this amount, the geometry is considered to have returned to the initial structure.

check_min_pert

0.6

Magnitude of perturbation along the Hessian eigenvalue for check_min=1, in atom ic units.

check_min_target

0

Add perturbation values to the current angles for check_min=2 or 3. Default for check_min=2.

 

1

Use the perturbation value as a target angle for check_min=2 or 3. Default for check_min=3.

check_min_p_pert

20.0

Perturbation in degrees for torsional perturbations. Interpreted as a target value or an increment, depending on check_min_target.

check_min_p_tol

8.0

Tolerance in degrees for checking whether to perturb a torsion for check_min=3. If the value of the torsion deviates from planarity by less than this amount, the torsion is perturbed, or for H-bond perturbations, if the two torsions differ by more than this amount.

check_min_amide

0

Do not perturb amide bonds when check_min=3.

 

1

Perturb amide bonds when check_min=3.

check_min_hb_tar

0

Add perturbation values to H-bond dihedrals for check_min=3.

 

1

Use the perturbation value as a target H-bond dihedral for check_min=3.

check_min_hb_lim

90.0

Upper limit for potential H-bond dihedral angles. If the H-bond dihedral angles are greater in absolute value than this amount, they are not considered potential H-bonds and are not perturbed.

optverdict

0

Do not run analysis of geometry convergence.

 

1

Run analysis of geometry convergence and print report:

0: Monotonic convergence. Convergence OK.

1: Non-monotonic convergence. No erratic convergence was detected and the structure converged to an optimal structure. Convergence OK.

2: Erratic convergence but the optimization converged to an optimal structure. Convergence OK.

 

 

3: Convergence to a non-optimal structure. The difference between the converged energy and the minimum energy is less than 0.1 kcal/mol. Convergence OK for solvation, OK for pseudospectral if the difference is less than 0.01 kcal/mol, but not otherwise.

4: Converged to a non-optimal structure, whose energy is above the minimum energy by more than 0.1 kcal/mol. Convergence not OK.

No convergence category is written if the optimization ran out of iterations.

 

2

Print report on convergence at each iteration.

Table 3 lists keywords specific to transition-state optimizations. The keywords for checking the transition state (starting with ts_vet) require that the products and reactants are specified with zmat2 and zmat3 sections. These keywords check the transition state at convergence, and the calculation can be terminated if the checks do not pass (for example, before proceeding with an IRC calculation). Table 4 lists keywords that are used to specify the initial Hessian, control Hessian updating, and modify the Hessian when using it to update the geometry.

Table 3. Keywords for transition-state optimizations

Keyword

Value

Description

iqst

0

Perform standard (non-QST) transition-state search

 

1

Use quadratic synchronous transit (QST) methods to guide transition-state search. Sets itrvec to −5

qstinit

0.5

Distance of LST transition-state initial guess between reactant and product geometries. Range is 0.0 to 1.0.

ifollow

0

For each optimization iteration, select a new eigenvector to follow

 

1

For each optimization iteration, follow eigenvector that most closely correlates with one followed previously

itrvec

0

Select lowest Hessian eigenvector as transition vector

 

>0

Select eigenvector number itrvec as the transition vector (see Transition-State Optimizations). Sets ifollow to 1.

 

−1

Select lowest non-torsional eigenvector as the transition vector

 

−2

Select the lowest stretching eigenvector as the transition vector

 

−5

Select the eigenvector that best represents the reaction path (the QST vector)

 

−6

Select the eigenvector that best overlaps the active coordinates specified in the coord section (see Active Coordinates in the Jaguar Input File)

ts_vet

0

Do not do any transition state sanity checks

 

1

Check overlap of the transition vector with the QST vector, and stop if the check fails.

 

−1

Check overlap of the transition vector with the QST vector, and display results of the check only.

  2 Check interatomic distances for breaking or forming bonds. The bonds that are broken or formed are determined from the connectivity of the reactant and product structures. The bonds should be in a range appropriate to a transition state, determined by two distance factors, ts_vet_dist_fac0 and ts_vet_dist_fac1. If the ratio of the interatomic distance to the sum of their van der Waals radii is not between these two factors, the test fails.

 

−2

Check interatomic distances for bonds breaking or forming, and display results of the check only.

  3 Check both overlap and interatomic distances, and stop if the check fails.

 

−3

Check both overlap and interatomic distances, and display results of the check only.

ts_vet_dist_fac0

1.15

Factor for checking bonded distance. If the distance between two atoms is less than this factor times the sum of their van der Waals radii, they are considered to be at a bonded distance, rather than an appropriate distance for a transition state, and the test fails.

ts_vet_dist_fac1

1.8

Factor for checking nonbonded distance. If the distance between two atoms is more than this factor times the sum of their van der Waals radii, they are considered to be not bonded, rather than at an appropriate distance for a transition state, and the test fails.

ts_vet_olap_cut

0.33

Cutoff for overlap of the transition vector with the QST vector or with the active coordinates. If the overlap is less than this value, the test fails.

no_mul_imag_freq 0 Do not attempt to remove or prevent multiple imaginary frequencies for transition state structures.
  1 Attempt to remove multiple imaginary frequencies in a transition state search. This is done by nudging the structure in the direction of the eigenvectors for the imaginary frequencies.
no_mul_nudge_fac 1.0 Factor used to determine the magnitude of the displacement used to nudge the structure away from the geometry with multiple imaginary frequencies.
no_mul_hess_redo 0 Do not recalculate an analytic Hessian when nudging the structure.
  1

Recalculate an analytic Hessian after nudging the structure.

  2

Recalculate an analytic Hessian before nudging the structure.

 

Table 4. Keywords for control of the Hessian

Keyword

Value

Description

inhess

−1

Use Fischer-Almlöf guess for Hessian

 

0

Use Schlegel guess for Hessian (default only if no hess section exists)

 

1

Use unit matrix for initial Hessian

 

2

Use Cartesian input Hessian found in hess section (inhess=2 automatically if non-empty hess section exists)

 

4

Compute and use quantum mechanical Hessian

  6 Compute and use GFN2-xTB Hessian

input_hessian

fischer

Use Fischer-Almlöf guess for Hessian. Equivalent to setting inhess=−1.

 

schlegel

Use Schlegel guess for Hessian (default only if no hess section exists). Equivalent to setting inhess=0.

 

unit_matrix

Use unit matrix for initial Hessian. Equivalent to setting inhess=1.

 

input_cartesian

Use Cartesian input Hessian found in hess section (input_hessian=input_cartesian automatically if non-empty hess section exists). Equivalent to setting inhess=2.

 

true_hessian

Compute and use quantum mechanical Hessian. Equivalent to setting inhess=4.

  xtb Compute and use GFN2-xTB Hessian. Equivalent to setting inhess=6.

nhesref

>0

Number of lowest-frequency Hessian eigenvectors used in Hessian refinement (default is 0)

 

−1

Use the active coordinates specified in the coord section to refine the Hessian (see Active Coordinates in the Jaguar Input File).

pertnd

0.05

Displacement (in atomic units) used for Hessian refinement or calculations of numerical forces or frequencies.

irefhup

2

Refine initial Hessian using Powell updates [205]

 

3

Refine initial Hessian using mixed Murtagh-Sargent/Powell updates [206]

 

4

Refine initial Hessian using Murtagh-Sargent updates [207]

ihuptyp

0

Don’t update Hessian

 

1

Update Hessian each iteration using BFGS (Broyden-Fletcher-Goldfarb-Shanno) method [208] (default for minimum-energy structure optimizations)

 

2

Update Hessian using Powell method [205]

 

3

Update Hessian using mixed Murtagh-Sargent/Powell method [206] (default for transition-state optimizations and IRC calculations)

 

4

Update Hessian using Murtagh-Sargent method [207] (not recommended)

 

−1

Compute quantum mechanical Hessian at each geometry; sets inhess=4

irfo

0

Before using Hessian to update geometry, modify it by sign-flipping or reverting to an older Hessian [204]

 

1

Before using Hessian to update geometry, modify it by RFO (rational function optimization) level shifting [209]. Default for geometry optimizations that do not use dynamic constraints.

 

2

Before using Hessian to update geometry, modify it by P-RFO (partitioned rational function optimization) level shifting [209]. Default for transition-state searches. Automatically set for geometry optimizations that use dynamic constraints.

In order to avoid changing the geometry too much because of an unusually shaped potential well or an inaccuracy in the Hessian, Jaguar restricts the norm of the changes to the Cartesian or internal coordinates to be less than a certain trust radius, which is defined in atomic units (bohrs, radians). The trust radius can vary from one iteration to another (itradj=1) or it can be fixed (itradj=0). Keywords controlling the use of a trust radius are listed in Table 5.

Table 5. Keywords for trust radius adjustment

Keyword

Value

Description

itradj

0

Use same trust radius throughout optimization (default for minimum-energy structure optimizations)

 

1

Adjust trust radius using Culot/Fletcher heuristic [[208], [210]] (default for transition-state optimizations)

 

−1

Adjust trust radius using Simons’ cubic potential model [211] (not recommended with Jaguar)

itrcut

0

Apply trust radius by truncating Newton-Raphson step(s)

 

1

Apply trust radius by level shifting of Hessian to reduce resultant step size

trust

0.3

Initial trust radius, in atomic units (bohr and/or radians): if norm of proposed displacements exceeds trust radius, step size is reduced as described by itrcut (default is 0.3, except for transition-state optimizations, when it is 0.1)

tradmx

0.3

Maximum trust radius allowed during optimization for itradj>0; see trust information (default is 0.3, except for solvated cases, when it is 0.1)

tradmn

0.01

Minimum trust radius allowed during optimization for itradj>0; see trust information

tremx

0.25

Trust radius reduction criterion; if relative error between actual and predicted energy changes is more than tremx and itradj>0, trust radius is reduced

trgmx

0.0

Trust radius reduction criterion; for itradj>0 and trgmx>0, if absolute error in a component of predicted gradient exceeds trgmx hartrees/bohr, trust radius is reduced

treok

0.2

Criterion for increasing trust radius; if itradj=1 and relative error between actual and predicted energy changes is less than treok, trust radius is increased (treok default is 0.2)

trescal

2.0

Scale factor for trust radius adjustment; used only when itradj=1

If the trust radius is fixed, it remains the same throughout the optimization except when Jaguar determines that changing it will lead to better convergence for problem jobs. This setting is the default for optimizations to minimum-energy structures. If the trust radius is allowed to vary (the default for transition-state optimizations), Jaguar keeps geometry changes within the region that is well-described by the Hessian by increasing the trust radius when the Hessian is correctly predicting energy changes and decreasing it when the predictions are inaccurate.

Prevention of chemical reactions during a geometry optimization can be done with the stop_rxn keyword. This keywords turns on checking of the connectivity after each geometry step. If bonds have been broken or formed, the optimization is restarted with additional repulsive or harmonic restraint potentials to prevent the bonding change, based on the value of stop_rxn. The repulsive potential for keeping atoms apart has the form

while the potential for keeping atoms together is a flat-bottomed harmonic potential,

where r is the distance between the two atoms whose bonding change is to be prevented, r0 is the initial distance, and a is the half-width of the flat bottom. The parameters can be set by keywords, which are described in Table 6. The energies from these added terms are not included in the reported total energies, but are used, along with the forces from these terms, for optimization purposes. As the added energies are artificial and are often added to a region of the potential surface that is quite steep, the results can be sensitive to the parameters used and to other circumstances of the calculations.

Table 6. Keywords for prevention of chemical reactions

Keyword

 

Description

stop_rxn

0

Do not prevent chemical reactions during geometry optimizations.

 

1

Add repulsive terms if bonds are formed and harmonic constraints if bonds are broken.

 

2

Add repulsive terms if bonds are formed only

 

3

Add harmonic constraints term if bonds are broken only.

stop_rxn_epsilon

35.0

Maximum strength of repulsive potential ε in kcal mol−1 Å2.

stop_rxn_sigma

0.15

Width parameter for repulsive potential σ in Å−2.

stop_rxn_kval

1000.0

Force constant for harmonic restraint k in kcal mol−1 Å−2.

stop_rxn_offset

0.0

Width of flat bottom a in Å.

The keywords shown in Table 7 may be used to specify the geometry convergence criteria. The values may be scaled to five times their default values with the keyword setting iaccg=3 for a quicker, coarser calculation, or to a tenth of the default values for tighter convergence with iaccg=5. The first two keywords in Table 7 have units of hartrees/bohr, gconv5 and gconv6 have units of bohrs, and gconv7 has units of hartrees.

Table 7. Geometry convergence criteria keywords

Keyword

Default value

Convergence Criterion For

gconv1

4.5×10−4

Maximum element of gradient

gconv2

3.0×10−4

rms of gradient elements

gconv5

1.8×10−3

Maximum element of nuclear displacement

gconv6

1.2×10−3

rms of nuclear displacement elements

gconv7

5.0×10−5

Difference between final energies from previous and current geometry optimization iterations

SCF calculations performed for each new structure generated during an optimization are judged to be converged when they meet the criterion for the root mean square of the change in density matrix elements, which is controlled by the keyword dconv; the usual SCF energy convergence criterion (econv) is ignored for optimizations.

If you want to save the structure at each step of a geometry optimization in a Maestro-formatted file, set ip472=2. You can also extract the structures from the output file to a Maestro file with the command

$SCHRODINGER/utilities/jagconvert -ijout jobname.out -omultimae filename.mae

If you import these structures into Maestro, you should set the color scheme manually: it is not set correctly on import.

To calculate certain properties (charges, multipole moments, Fukui indices) at each step in a geometry optimization, set props_each_step=1. This allows you to examine these properties during the course of a geometry optimization for points at which they change significantly.

Geometry Optimization Examples

Transition State Geometry Optimization Examples