modify minimization
   criterion value   # 1
      method meth       # 2
   #
      archive arch      # 3
      update_list upda  # 4
      print prin        # 5
      restart rest      # 6
   #
      reset rese        # 7
   #
      energy ener       # 8
      ceiling ceil      # 9
      gradient grad     # 10
   #
      step_size step    # 11
      up_step upst      # 12
      down_step down    # 13
   #
      max_step maxs     # 14
      #
   step_rate rate    # 15
      step_limit limt   # 16
      direction dire    # 17
      target targ       # 18
   end

Modify the energy minimizers. (1) The variable MIN_PASSED will be set to true when the potential energy at the end of the minimization is less than value (preset value 1.0) times the potential energy at the start. (2) Select an energy minimization method; two Conjugate Gradient methods are available, cgpr (preset value) for Polak-Ribiere and cgfr for Fletcher-Reeves; steepest selects the Steepest Descent method, gradient the Gradient method which is used for programming experiments; the sniffer method has been deprecated and will be removed in future versions. The Conjugate Gradient and Gradient methods make use of the linesearch methods set in the previous block.

(3) During minimization, the coordinates are archived every arch steps; the preset value of 0 means that no periodic archiving will be done. (4) The non-bond lists are updated every upda steps, preset value 25. (5) The energies are printed every prin steps, preset value 1. (6) In the conjugate gradient methods, the direction vectors are reset in the opposite direction of the gradient every rest steps; the preset value is three times the number of atoms in the system.

(7) The direction vector is also reset if the energy rises if rese is yes (preset value) or not if the value is no.

Normally the energy minimization will proceed for as many steps as is specified (see Simple Commands) but when certain conditions are met the iteration will terminal prematurely. This is not necessarily a bad thing; further minimization may not be productive. These conditions are: (8) minimum absolute energy change per step in the Conjugate Gradient methods, the preset value is 10 times the machine epsilon in units of kcal/mol; (9) maximum absolute energy increase in Conjugate Gradient, preset value is one-tenth the largest floating point value in units of kcal/mol that can be represented on the machine; (10) root mean square (RMS) gradient, preset value is 10^-3 kcal/mol/Å. Note that the preset values in (8) and (9) are such that they are unlikely to be satisfied; to make use of these stopping conditions you need to set realistic numbers.

(11) If the Steepest Descent method is used or if the Conjugate Gradient method is used with the nosearch linesearch method (see modify linesearch), the iterations start at an initial step size of step, preset value is 10^-9 Ų/CEU. (11) The step size is scaled by a factor upst, preset value 1.2, if the energy drops or (12) scaled by down, preset value 0.5, if the energy increases.

(14) In the Steepest Descent method, the maximum allowable change in energy per step can be set to maxs; the preset value of 0.0 Ų/CEUmeans that this maximum is not imposed.

The following are settings for the Sniffer method: (15) step size per unit energy µ, preset value 1.0 Å/CEU; (16) maximum step size µ^, preset value 2.0 Ų/CEU; (17) search direction offset, epsilon, preset value 2.0; (18) target scale factor c, preset value 0.01.