40.1.7 Options for computing Hessians

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. 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.

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 40.2.9 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.