1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
|
C Copyright (c) 2003-2010 University of Florida
C
C This program is free software; you can redistribute it and/or modify
C it under the terms of the GNU General Public License as published by
C the Free Software Foundation; either version 2 of the License, or
C (at your option) any later version.
C This program is distributed in the hope that it will be useful,
C but WITHOUT ANY WARRANTY; without even the implied warranty of
C MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
C GNU General Public License for more details.
C The GNU General Public License is included in this distribution
C in the file COPYRIGHT.
SUBROUTINE TKSTEP(GRD, HES, SCRATCH, HESMOD, GRDMOD, STEP,
& GRDHES, DIAGHES, VEC, RFAMAT, EVRFAMAT,
& BMATRIX, HES_INTACT, FSCR, IADITNL)
C
C Control all the optimiztion alogrithms. The old EFOL (John Stanton)
C subroutine was not flexible enough to add new optimization
C algorithms. This routine act as a front end, and call appropriate
C routines depending on user's choice of a particular algorithm.
C
C Ajith Perera, March, 2000
C
IMPLICIT DOUBLE PRECISION (A-H, O-Z)
C
DOUBLE PRECISION LMBDAN, LMBDAP
LOGICAL MORSE, TS, NRORMANR, RFA, EVFTS, IGTS, QSD,
& QSTLST_CLIMB
C
#include "mxatms.par"
#include "machsp.com"
#include "coord.com"
#include "flags.h"
#include "jodaflags.com"
C
PARAMETER (STPTOL = 1.0D-12, LUOUT = 6,
& LUDONE = 80,
& EPS = 1.0D-8)
COMMON /USINT/ NX, NXM6, IARCH, NCYCLE, NUNIQUE, NOPT
C
COMMON /OPTCTL/ IPRNT, INR, IVEC, IDIE, ICURVY, IMXSTP, ISTCRT,
& IVIB, ICONTL, IRECAL, INTTYP, IDISFD, IGRDFD,
& ICNTYP,ISYM, IBASIS, XYZTol
C
DIMENSION HES(NXM6,NXM6), GRD(NXM6), SCRATCH(NX*NX),
& HESMOD(NOPT,NOPT), GRDMOD(NOPT), STEP(NXM6),
& IQFIX(3*NATOMS,3*NATOMS), DIAGHES(NOPT, NOPT),
& GRDHES(NOPT),VEC(NOPT), EVRFAMAT(NOPT+IADITNL,
& NOPT+IADITNL), RFAMAT(NOPT+IADITNL, NOPT+IADITNL),
& BMATRIX(NXM6*NX), HES_INTACT(NX*NX), FSCR(NXM6*NX),
& ENERGY(2)
C
C Form modified Hessian containing optimized coordinates
C only, and transform it to totally symmetric coordinates. Do the
C same for gradient vector as well. Do all other initilizations
C and printing. Note that STPMAX is a user defined flag. It is
C initially set in mkvmol.F but is reprocessed in setup.f.
C
IBREAK = 0
IF (iFlags2(h_IFLAGS2_geom_opt) .eq. 1) THEn
C
C This is a user defined internal optimization
C
CALL SETUP(HES, GRD, HESMOD, GRDMOD, IQFIX, STPMAX, LUOUT)
C
ELSE IF (iFlags2(h_IFLAGS2_geom_opt) .eq. 2) THEN
C
C
C This is a Cartesian optimization
CALL DCOPY(NOPT, GRD, 1, GRDMOD, 1)
C
C Vibrations and rotation must be projected out from the Hessian. In
C this particular instance unit masses at the points where atoms are
C located.
C
Do Iatoms = 1, NX/3
Fscr(Iatoms) = 1.0D0
Enddo
#ifdef _DEBUG_LVLM1
Write(6,*) "@-TKSTEP, The Hessian before the projection"
call output(Hes, 1, NXM6, 1, NXM6, NXM6, NXM6, 1)
#endif
CALL PROJEC_FC(Q, Hes, Fscr, GrdMOD, HesMOd, Scratch,
& EPS, Nx/3, .True., .True., .False.)
#ifdef _DEBUG_LVLM1
Write(6,*) "@-TKSTEP The Hessian after the projection"
call output(Hes, 1, NX, 1, NX, NX, NX, 1)
#endif
CALL SETUP(HES, GRD, HESMOD, GRDMOD, IQFIX, STPMAX, LUOUT)
CALL DCOPY(NOPT*NOPT, HES, 1, HESMOD, 1)
C
ELSE IF (iFlags2(h_IFLAGS2_geom_opt) .gt. 2) THEN
C
C This is a redundant internal optimization
C
CALL PROJEC_IFC(HES, FSCR, HESMOD, HES_INTACT, DIAGHES,
& GRD, NXM6)
CALL SETUP(HES, GRD, HESMOD, GRDMOD, IQFIX, STPMAX, LUOUT)
ENDIF
C
C Diagonalize the symmetrized Hessian. First, save the
C original symmetrized Hessian in HES to be used in QST/LST
C procedure. See the dependents of ANLYSHES that deal
C with the QST/LST procedure. Note that HESMOD and DIAGHES
C will be changed for QST/LST steps (see modfy_hessian.F).
C
CALL DCOPY(NOPT*NOPT, HESMOD, 1, HES, 1)
CSSS CALL OUTPUT(HES, 1, NOPT, 1, NOPT, NOPT, NOPT, 1)
CALL EIG(HESMOD, DIAGHES, NOPT, NOPT, 1)
IF (iFlags2(h_IFLAGS2_geom_opt) .eq. 2) CALL EVEC_SHIFT(HESMOD,
& DIAGHES, HES_INTACT, NOPT)
C
C Analyse the eigenvalues and eigenvectors of the current Hessian
C to decide the direction to follow. Note that ANALYSHES does
C meaningful work only for the transition state searches.
C
C ANLYSHES was modified to incorporate linear and quadratic
C synchronous (LST & QST) methods(?) of Schlegel et al. (Israel
C Journal of Chemistry, vol. 33, pg. 449, 1993). Both LST & QST
C approaches control the direction of the steps being taken until
C the quadratic region of the transition state is reached. Then
C the step direction is along the vector that corresponds to the
C smallest eigenvalue of the Hessian or a user-specified
C eigenvector of the Hessian. See ANALYSHES for further comments.
C Ajith Perera, 07/04.
C
C Important Note: RFAMAT and EVRFAMAT are used in ANLYSHES to
C keep the LST and QST directional vectors. They
C revert back to zero after their use in ANLYSHES so that
C there intended use in GETLAMBDA is not upset.
C
CALL ANLYSHES(HESMOD, DIAGHES, RFAMAT, EVRFAMAT, HES,
& GRDMOD, SCRATCH, VEC, NOPT, NX,
& NCYCLE, INR, IVEC, IMODE, IDIE, TS,
& NRORMANR, RFA, EVFTS, IGTS, QSD, LUOUT,
& QSTLST_CLIMB, NATOMS, IPRNT)
C
C
C Calculate the gradients along Hessian eigenvectors (GRDHES). This
C is all you need to do bare NR (MANR) optimizations. Also fill the
C scratch vector with the internal coordinates of the previous step
C (first cycle it is identical to the user given starting values!)
C
CALL ZERO(GRDHES, NOPT)
DO 1180 I = 1, NOPT
C
SCRATCH(I) = R(NOPTI(I))
C
CSSS WRITE(6,*) "The value of Grd-Mod", GRDMOD(I)
DO 1179 J = 1, NOPT
GRDHES(I)=GRDHES(I)+GRDMOD(J)*DIAGHES(J,I)
C
1179 CONTINUE
1180 CONTINUE
C
C Do the Newton-Raphson or Morse adjusted Newton-Raphson minima search.
C
MORSE = (INR .EQ. 3 .OR. INR .EQ. 5)
CSSS Write(6,*) "Debug Info: The QST and LST climb", QSTLST_CLIMB
C
C IF (NCYCLE .GE. 2) THEN
C
C The following stuff is for the line search. God I need some help
C from litrature in order to finish this. Let's comment this out
C until, I have some help from others.
C
C CALL DGETREC(20, 'JOBARC', 'TOTENERG', 1, CURR_ENRG)
C CALL DGETREC(20, 'JOBARC', 'OLDENERG', 1, PREV_ENRG)
C CALL DGETREC(20, 'JOBARC', 'OLDGRADS', NXM6,
C & SCRATCH(3*NXM6+1))
C CALL DCOPY(NXM6, GRD, 1, SCRATCH(2*NXM6+1), 1)
C CALL DCOPY(NXM6, R, 1, SCRATCH(1), 1)
C CALL DGETREC(20, 'JOBARC', 'OLDGEOMT', NXM6,
C & SCRATCH(NXM6 +1))
C ENERGY(1) = CURR_ENRG
C ENERGY(2) = PREV_ENRG
C Print*, (SCRATCH(I), I=1, 2*NOPT)
C Print*, (SCRATCH(2*NOPT+I), I=1, 2*NOPT)
C CALL LN_SEARCH(ENERGY, SCRATCH(1), SCRATCH(2*NXM6+1),
C & SCRATCH(4*NXM6+1), NXM6, 2, DQHDQ, STPTOL)
C CALL XDCOPY(NXM6, SCRATCH(2*NXM6+1), 1, GRD, 1)
C Print*, "The New Gradients", (GRD(I), I=1, NXM6)
C CALL ZERO(GRDHES, NOPT)
C DO I=1, NOPT
C Write(6,*) GRD(NOPTI(I)), NEQ(NOPTI(I))
C GRDMOD(I)= DSQRT(DFLOAT(NEQ(NOPTI(I))+1))*GRD(NOPTI(I))
C ENDDO
C
C Print*, "The Grad mod", (GRDMOD(I), I=1, NOPT)
C DO IOPT = 1, NOPT
C SCRATCH(IOPT) = SCRATCH(NOPTI(IOPT))
C DO J = 1, NOPT
C GRDHES(IOPT)=GRDHES(IOPT)+GRDMOD(J)*DIAGHES(J,IOPT)
C ENDDO
C ENDDO
C Print*, "After a line search:", (SCRATCH(i), i=1,nopt)
CC Plese zero out the SCRATCH before leave this block so what's done
CC here can not interfere with the rest.
C
C ENDIF
C
IF (NRORMANR) THEN
C
CALL NEWRAPH(SCRATCH, GRDHES, HESMOD, DIAGHES, MORSE, NOPT,
& NX, LUOUT)
C
ELSE IF (TS .OR. RFA) THEN
C
C First get the Lambda(P) and Lambda(N) parameters: Jon Baker
C J. Comput. Chem. 7, 385, 1986 and references their in. In general
C this is an eigenvector following method based on Rational Function Approximation
C that can apply for both minimas and transition states. In the case of minima
C search Lambda(N) correspond to the lowest eigenvalue of RFA matrix. There
C is no Lambda(p) parameter for minima search.
C
CALL GETLAMBDA(HESMOD, GRDHES, RFAMAT, NOPT, EVRFAMAT,
& LMBDAN, LMBDAP, IADITNL, IMODE, LUOUT,
& TS, RFA, QSTLST_CLIMB)
C
IF (RFA) CALL FLOWEVFMIN(SCRATCH, GRDHES, HESMOD, DIAGHES,
& LMBDAN, MORSE, NX, NOPT)
C
IF (EVFTS) CALL FLOWEVFTS(SCRATCH, GRDHES, HESMOD, DIAGHES,
& LMBDAN, LMBDAP, STPMAX, MORSE,
& IMODE, NX, LUOUT, IBREAK, NOPT,
& QSTLST_CLIMB)
C
ELSE IF (IGTS .OR. QSD) THEN
C
CALL QSDMINORTS(SCRATCH, GRDMOD, HESMOD, DIAGHES, RFAMAT,
& EVRFAMAT, STPMAX, EPS, IGTS, QSD, NX, NOPT,
& NCYCLE)
C
ENDIF
C
C Process the step taken: Convert to internal from symmetry coodinates,
C filter, apply the requested step size control, update the Geometry
C vector (R).
C
C Initialize the starting trust radius. At this point it is tied to
C the maximum step size, controlled by the user by setting MAX_STEP.
C STPMAX is set in mkvmol.F and setup.F.
C
BEGN_TRUST_RAD = STPMAX
CALL PROCESTEP(SCRATCH, HES, GRDMOD, SCALE, STPMAG, STEP,
& STPTOL, STPMAX, IQFIX, ISTCRT, BEGN_TRUST_RAD,
& EPS, QSTLST_CLIMB, TS)
C
C Let's do the printing. Test to see if calculation has converged.
C
CALL SUMMARY(SCRATCH, RFAMAT, GRD, SCALE, STPMAG, IQFIX,
& NOPT, NX, NXM6, IBREAK, ICONTL, LUOUT, NCYCLE,
& LUDONE, BMATRIX, HES_INTACT, FSCR, VEC,
& STEP)
C
RETURN
END
|
