Molecular geometry optimization


Three types of the geometry can be optimized: the most stable (minimum energy) geometry, conical intersections between the electronic states, and the transition state geometry (or the saddle point on the potential energy surface).

In the optimizations, rather than using the exact Hessian, one can start using the approximate Hessian and update it according to the step taken. The advanced quasi-newton optimization methods, eigenvector following (EF) algorithm and rational functional optimization (RFO) are implemented. In the minimum energy conical intersection (MECI) optimization, the molecular gradient is replaced by the sum of the energy difference gradient and the upper state gradient after projecting the degeneracy lifting vectors out (gradient projection). The minimum distance conical intersection (MDCI) can be also optimized by replacing the upper state gradient in MECI optimization with the distance vector to the reference geometry. In addition, the minimum energy path to the reactants and products from the saddle point can be calculated using the second order algorithm, without mass weighting.

The optimizer in BAGEL has been interfaced with an external molecular mechanics program, TINKER, using which mixed quantum mechanics/molecular mechanics (QM/MM) optimization can be performed. To perform this, the TINKER input files (keyword file tinkin.key and initial coordinate file should be provided in tinker1 and tinker2 subdirectories, respectively. The testgrad program in the TINKER package should be installed in $PATH.

The output contains the gradient evaluation progress at the first step of the optimization, and the status of the optimization. The other information, including the quantum chemistry calculations at the optimization steps, are deposited in the file opt.log. The history of the optimization and the final geometry are also saved in the MOLDEN files opt_history.molden and opt.molden, and can be read by MOLDEN.


Required Keywords


Description: Request geometry optimization.
Datatype: string
Values: (optimize)
optimize: Optimize geometry.
Default: N/A


Description: Type of the optimization calculations.
Datatype: string
energy: find the most stable geometry.
conical or meci: find the minimum energy conical intersections, according to gradient projection method.
mdci: find the minimum distance conical intersections, according to modified gradient projection method.
transition: find the transition state geometry (saddle point on the PES).
mep: find the minimum energy path using the second-order algorithm, starting from the transition state geometry.
Default: energy.


Description: The target state to optimize.
Datatype: int
0: the ground state.
1: the first excited state, and so on.
Default: 0


Description: The second target state to optimize in the conical intersection optimization.
Datatype: int
0: the ground state.
1: the first excited state, and so on.
Default: 1


Description: The method array allows the user to specify one or more methods to be used in the Hessian calculation. See section on input structure for more information.

Convergence Criteria


Description: Maximum component of the gradient in Hartree / Bohr.
Datatype: double precision.
Default: 0.00001 (atoms in the molecule < 4); 0.0003 (larger).


Description: Maximum component of the displacement in Bohr.
Datatype: double precision.
Default: 0.00004 (atoms in the molecule < 4); 0.0012 (larger).


Description: The energy change in Hartree.
Datatype: double precision.
Default: 0.000001.

Optional Keywords (Universal)


Description: Algorithm for the optimization calculations.
Datatype: string
ef: Eigenvector-following (EF) algorithm.
rfo: Rational functional optimization algorithm.
nr: Newton–Raphson algorithm.
Default: ef.
Recommendation: use either “ef” or “rfo”.


Description: Maximum step. The unit is in the specifed coordinate.
Datatype: double precision.
Default: 0.3 (energy optimization); 0.1 (otherwise).


Description: Use internal coordinate or not.
Datatype: bool
true: use internal coordinates.
false: use Cartesian coordinates.
Default: true.
Recommendation: use default when you have a single molecule. If bond-breaking process is in consideration, use “false”.


Description: Use redunant internal coordinate or delocalized internal coordinate.
Datatype: bool
true: use redundant internal coordinate.
false: use delocalized internal coordinate.
Default: false.
Recommendation: use default.


Description: Maximum number of iteration for optimization.
Datatype: int
Default: 100.


Description: Maximum number of Z-vector iterations for gradient evaluation. Applies to SA-CASSCF, CASPT2, and MP2 calculations.
Datatype: int
Default: 100.
Recommendation: increase the value when the Z-vector equation does not converge.


Description: Use numerical gradient.
Datatype: bool
true: use numerical gradient.
false: use analytical gradient.
Default: false.


Description: Delta x for numerical gradient.
Datatype: double precision
Default: 0.001 (bohr).


Description: Hessian updating scheme.
Datatype: string
flowchart: use flowchart update. This automatically decides according to the shape of PES.
bfgs: use BFGS scheme.
psb: use PSB scheme.
sr1: use SR1 scheme.
Default: flowchart.


Description: Use approximate Hessian for the initial step of the optimization.
Datatype: bool
true: use approximate Hessian.
false: calculate numerical Hessian first, and start the optimization using the Hessian.
Default: true.


Description: Use adaptive stepsize in RFO algorithm.
Datatype: bool
true: use adaptive maximum stepsize.
false: use fixed maximum stepsize.
Default: true (algorithm is RFO); false (otherwise).

Optional Keywords (Conical Intersection Optimization)


Description: Type of non-adiabatic coupling matrix element to be used.
Datatype: string
full: use full nonadiabatic coupling.
interstate: use interstate coupling.
etf: use nonadiabatic coupling with built-in electronic translational factor (ETF).
noweight: use interstate coupling without weighting it by energy gap.
Default: noweight.


Description: Thiel’s C_3 parameter, which is multiplied to the full gradient.
Datatype: double precision
Default: 2.0 (MECI) or 0.01 (MDCI).


Description: Thiel’s C_4 parameter, which is multiplied to the gradient difference.
Datatype: double precision
Default: 0.5


Description: Specify reference geometry used in MDCI optimization.
Datatype: bool
true: specify reference geometry in the refgeom block.
false: the first geometry for optimization is considered as the reference geometry.
Default: false


Description: Reference geometry for MDCI optimization. The format is the same as the molecule block.

Optional Keywords (Minimum Energy Path)


Description: Direction of the MEP calculation from the transition state.
Datatype: int
1: use the direction of the lowest eigenvector.
-1: use the opposite direction of the lowest eigenvector.
Default: 1
Recommendation: run two calculations with “1” and “-1” to get the full path.

Optional Keywords (QM/MM)


Description: Do QM/MM energy optimization.
Datatype: bool
true: do QM/MM optimization.
false: do gas phase optimization.
Default: false


Description: Molecular mechanics program to do QM/MM.
Datatype: string
tinker: do QM/MM optimization with TINKER.
Default: tinker.


This optimizes the ground state geometry of benzophenone.

alternate text

The benzophenone molecule with carbon atoms in grey, oxygen in red, and hydrogen in white.

Sample input

 { "bagel" : [

   "title" : "molecule",
   "basis" : "cc-pvdz",
   "df_basis" : "cc-pvdz-jkfit",
   "angstrom" : false,
   "geometry" : [
   { "atom" : "C", "xyz" : [     -2.002493,     -2.027773,      0.004882 ] },
   { "atom" : "C", "xyz" : [     -2.506057,     -4.613700,      0.009896 ] },
   { "atom" : "C", "xyz" : [      0.536515,     -1.276360,      0.003515 ] },
   { "atom" : "C", "xyz" : [     -0.558724,     -6.375134,      0.013503 ] },
   { "atom" : "H", "xyz" : [     -4.396140,     -5.341490,      0.011057 ] },
   { "atom" : "C", "xyz" : [      2.478233,     -3.024614,      0.007049 ] },
   { "atom" : "H", "xyz" : [      0.959539,      0.714937,     -0.000292 ] },
   { "atom" : "C", "xyz" : [      1.936441,     -5.592475,      0.012127 ] },
   { "atom" : "H", "xyz" : [     -1.012481,     -8.367883,      0.017419 ] },
   { "atom" : "H", "xyz" : [      4.418042,     -2.380738,      0.005919 ] },
   { "atom" : "H", "xyz" : [      3.448750,     -6.968581,      0.014980 ] },
   { "atom" : "C", "xyz" : [     -6.758666,     -0.057378,      0.001157 ] },
   { "atom" : "C", "xyz" : [     -8.231109,     -2.241648,      0.000224 ] },
   { "atom" : "C", "xyz" : [     -8.022986,      2.269249,      0.001194 ] },
   { "atom" : "C", "xyz" : [    -10.853532,     -2.110536,     -0.000769 ] },
   { "atom" : "H", "xyz" : [     -7.410047,     -4.093049,      0.000224 ] },
   { "atom" : "C", "xyz" : [    -10.632155,      2.405932,      0.000369 ] },
   { "atom" : "H", "xyz" : [     -6.913797,      3.976253,      0.001805 ] },
   { "atom" : "C", "xyz" : [    -12.064741,      0.207004,     -0.000695 ] },
   { "atom" : "H", "xyz" : [    -11.941318,     -3.840822,     -0.001614 ] },
   { "atom" : "H", "xyz" : [    -11.548963,      4.232744,      0.000447 ] },
   { "atom" : "H", "xyz" : [    -14.107194,      0.302907,     -0.001460 ] },
   { "atom" : "C", "xyz" : [     -3.892311,      0.136360,      0.001267 ] },
   { "atom" : "O", "xyz" : [     -3.026383,      2.227189,     -0.001563 ] }

   "title" : "optimize",
    "method" : [ {
     "title" : "hf",
     "thresh" : 1.0e-12
   } ]

Using the same molecule block, a geometry optimization with XMS-CASPT2 can be performed. In this particular example as is often the case, the active keyword is used to select the orbitals for the active space that includes 4 electrons and 3 orbitals. Three sets of \(\pi\) and \(\pi^*\) orbitals localized on the phenyl rings are included along with one non-bonding orbital (oxygen lone pair). The casscf orbitals are state-averaged over three states. Since a multistate calculation is performed, the user must specify which state is to be optimized (the target). In this example, we optimize the ground state.

  "title" : "casscf",
  "nstate" : 2,
  "nclosed" : 46,
  "nact" : 3,
  "active" : [37, 44, 49]

  "title" : "optimize",
   "target" : 0,
   "method" : [ {
     "title" : "caspt2",
       "smith" : {
         "method" : "caspt2",
         "ms" : "true",
         "xms" : "true",
         "sssr" : "true",
         "shift" : 0.2,
         "frozen" : true
     "nstate" : 2,
     "nact" : 3,
     "nclosed" : 46
   } ]



Description of Reference Reference
Eigenvector following algorithm J. Baker, J. Comput. Chem. 7, 385 (1986).
Rational functional optimization algorithm A. Banerjee, N. Adams, J. Simons, and R. J. Shepard, J. Phys. Chem. 89, 52 (1985).
Second-order minimum energy path search C. Gonzalez and H. B. Schlegel, J. Chem. Phys. 90, 2154 (1989).
Gradient projection algorithm M. J. Bearpark, M. A. Robb, and H. B. Schlegel, Chem. Phys. Lett. 223, 269 (1994).
Flowchart method A. B. Birkholz and H. B. Schlegel, Theor. Chem. Acc. 135, 84 (2016).
ETF in nonadiabatic coupling S. Fatehi and J. E. Subotnik, J. Phys. Chem. Lett. 3, 2039 (2012).
Thiel’s conical intersection parameters T. W. Keal, A. Koslowski, and W. Thiel, Theor. Chem. Acc. 118, 837 (2007).