root / src / Step_GEDIIS.f90
Historique | Voir | Annoter | Télécharger (17,08 ko)
| 1 |
! Geom = input parameter vector (Geometry), Grad = input gradient vector. |
|---|---|
| 2 |
! HEAT is Energy(Geom) |
| 3 |
SUBROUTINE Step_GEDIIS(Geom_new,Geom,Grad,HEAT,Hess,NCoord,FRST) |
| 4 |
! This routine was adapted from the public domain mopac6 diis.f |
| 5 |
! source file (c) 2009, Stewart Computational Chemistry. |
| 6 |
! <http://www.openmopac.net/Downloads/Downloads.html> |
| 7 |
! |
| 8 |
!---------------------------------------------------------------------- |
| 9 |
! Copyright 2003-2014 Ecole Normale Supérieure de Lyon, |
| 10 |
! Centre National de la Recherche Scientifique, |
| 11 |
! Université Claude Bernard Lyon 1. All rights reserved. |
| 12 |
! |
| 13 |
! This work is registered with the Agency for the Protection of Programs |
| 14 |
! as IDDN.FR.001.100009.000.S.P.2014.000.30625 |
| 15 |
! |
| 16 |
! Authors: P. Fleurat-Lessard, P. Dayal |
| 17 |
! Contact: optnpath@gmail.com |
| 18 |
! |
| 19 |
! This file is part of "Opt'n Path". |
| 20 |
! |
| 21 |
! "Opt'n Path" is free software: you can redistribute it and/or modify |
| 22 |
! it under the terms of the GNU Affero General Public License as |
| 23 |
! published by the Free Software Foundation, either version 3 of the License, |
| 24 |
! or (at your option) any later version. |
| 25 |
! |
| 26 |
! "Opt'n Path" is distributed in the hope that it will be useful, |
| 27 |
! but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 28 |
! |
| 29 |
! MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 30 |
! GNU Affero General Public License for more details. |
| 31 |
! |
| 32 |
! You should have received a copy of the GNU Affero General Public License |
| 33 |
! along with "Opt'n Path". If not, see <http://www.gnu.org/licenses/>. |
| 34 |
! |
| 35 |
! Contact The Office of Technology Licensing, valorisation@ens-lyon.fr, |
| 36 |
! for commercial licensing opportunities. |
| 37 |
!---------------------------------------------------------------------- |
| 38 |
|
| 39 |
use Io_module |
| 40 |
|
| 41 |
IMPLICIT NONE |
| 42 |
|
| 43 |
INTEGER(KINT) :: NCoord |
| 44 |
REAL(KREAL) :: Geom_new(NCoord), Grad(NCoord), Hess(NCoord*NCoord) |
| 45 |
REAL(KREAL), INTENT(IN) :: Geom(NCoord) |
| 46 |
REAL(KREAL) :: HEAT ! HEAT= Energy |
| 47 |
LOGICAL :: FRST |
| 48 |
|
| 49 |
! MRESET = maximum number of iterations. |
| 50 |
INTEGER(KINT), PARAMETER :: MRESET=15, M2=(MRESET+1)*(MRESET+1) !M2 = 256 |
| 51 |
REAL(KREAL), ALLOCATABLE, SAVE :: GeomSet(:), GradSet(:) ! MRESET*NCoord |
| 52 |
REAL(KREAL), ALLOCATABLE, SAVE :: DX(:), GSAVE(:) !NCoord |
| 53 |
REAL(KREAL), SAVE :: ESET(MRESET) |
| 54 |
REAL(KREAL) :: ESET_tmp(MRESET), B(M2),BS(M2),BST(M2), B_tmp(M2) ! M2=256 |
| 55 |
LOGICAL DEBUG, PRINT, ci_lt_zero |
| 56 |
INTEGER(KINT), SAVE :: MSET ! mth Iteration |
| 57 |
REAL(KREAL) :: ci(MRESET), ci_tmp(MRESET) ! MRESET = maximum number of iterations. |
| 58 |
INTEGER(KINT) :: NGEDIIS, MPLUS, INV, ITERA, MM, cis_zero |
| 59 |
INTEGER(KINT) :: I, J, K, JJ, JNV, II, IONE, IJ, IX, JX, KX |
| 60 |
INTEGER(KINT) :: current_size_B_mat, MyPointer |
| 61 |
REAL(KREAL) :: XMax, XNorm, DET, THRES, tmp, ER_star, ER_star_tmp |
| 62 |
|
| 63 |
DEBUG=.TRUE. |
| 64 |
PRINT=.FALSE. |
| 65 |
|
| 66 |
IF (PRINT) WRITE(*,'(/,'' BEGIN GEDIIS '')') |
| 67 |
|
| 68 |
! Initialization |
| 69 |
IF (FRST) THEN |
| 70 |
! FRST will be set to False in SPACE_GEDIIS, so no need to modify it here |
| 71 |
IF (ALLOCATED(GeomSet)) THEN |
| 72 |
IF (PRINT) WRITE(*,'(/,'' In FRST, GEDIIS Dealloc '')') |
| 73 |
DEALLOCATE(GeomSet,GradSet,DX,GSave) |
| 74 |
RETURN |
| 75 |
ELSE |
| 76 |
IF (PRINT) WRITE(*,'(/,'' In FRST, GEDIIS Alloc '')') |
| 77 |
ALLOCATE(GeomSet(MRESET*NCoord),GradSet(MRESET*NCoord),DX(NCoord),GSAVE(NCoord)) |
| 78 |
END IF |
| 79 |
END IF ! IF (FRST) THEN |
| 80 |
|
| 81 |
! SPACE_GEDIIS SIMPLY LOADS THE CURRENT VALUES OF Geom AND Grad INTO THE ARRAYS GeomSet |
| 82 |
! AND GradSet, MSET is set to zero and then 1 in SPACE_GEDIIS at first iteration. |
| 83 |
CALL SPACE_GEDIIS(MRESET,MSET,Geom,Grad,HEAT,NCoord,GeomSet,GradSet,ESET,FRST) |
| 84 |
|
| 85 |
IF (PRINT) WRITE(*,'(/,'' GEDIIS after SPACE_GEDIIS '')') |
| 86 |
|
| 87 |
! INITIALIZE SOME VARIABLES AND CONSTANTS: |
| 88 |
NGEDIIS = MSET !MSET=mth iteration |
| 89 |
MPLUS = MSET + 1 |
| 90 |
MM = MPLUS * MPLUS |
| 91 |
|
| 92 |
! CONSTRUCT THE GEDIIS MATRIX: |
| 93 |
! B_ij calculations from <B_ij=(g_i-g_j)(R_i-R_j)> |
| 94 |
JJ=0 |
| 95 |
INV=-NCoord |
| 96 |
DO I=1,MSET |
| 97 |
INV=INV+NCoord |
| 98 |
JNV=-NCoord |
| 99 |
DO J=1,MSET |
| 100 |
JNV=JNV+NCoord |
| 101 |
JJ = JJ + 1 |
| 102 |
B(JJ)=0.D0 |
| 103 |
DO K=1, NCoord |
| 104 |
B(JJ) = B(JJ) + (((GradSet(INV+K)-GradSet(JNV+K))*(GeomSet(INV+K)-GeomSet(JNV+K)))/2.D0) |
| 105 |
END DO |
| 106 |
END DO |
| 107 |
END DO |
| 108 |
|
| 109 |
! The following shifting is required to correct indices of B_ij elements in the GEDIIS matrix. |
| 110 |
! The correction is needed because the last coloumn of the matrix contains all 1 and one zero. |
| 111 |
DO I=MSET-1,1,-1 |
| 112 |
DO J=MSET,1,-1 |
| 113 |
B(I*MSET+J+I) = B(I*MSET+J) |
| 114 |
END DO |
| 115 |
END DO |
| 116 |
|
| 117 |
! for last row and last column of GEDIIS matrix |
| 118 |
DO I=1,MPLUS |
| 119 |
B(MPLUS*I) = 1.D0 |
| 120 |
B(MPLUS*MSET+I) = 1.D0 |
| 121 |
END DO |
| 122 |
B(MM) = 0.D0 |
| 123 |
|
| 124 |
DO I=1, MPLUS |
| 125 |
!WRITE(*,'(10(1X,F20.4))') B((I-1)*MPLUS+1:I*(MPLUS)) |
| 126 |
END DO |
| 127 |
|
| 128 |
! ELIMINATE ERROR VECTORS WITH THE LARGEST NORM: |
| 129 |
80 CONTINUE |
| 130 |
DO I=1,MM !MM = (MSET+1) * (MSET+1) |
| 131 |
BS(I) = B(I) !just a copy of the original B (GEDIIS) matrix |
| 132 |
END DO |
| 133 |
|
| 134 |
IF (NGEDIIS .NE. MSET) THEN |
| 135 |
DO II=1,MSET-NGEDIIS |
| 136 |
XMAX = -1.D10 |
| 137 |
ITERA = 0 |
| 138 |
DO I=1,MSET |
| 139 |
XNORM = 0.D0 |
| 140 |
INV = (I-1) * MPLUS |
| 141 |
DO J=1,MSET |
| 142 |
XNORM = XNORM + ABS(B(INV + J)) |
| 143 |
END DO |
| 144 |
IF (XMAX.LT.XNORM .AND. XNORM.NE.1.0D0) THEN |
| 145 |
XMAX = XNORM |
| 146 |
ITERA = I |
| 147 |
IONE = INV + I |
| 148 |
ENDIF |
| 149 |
END DO |
| 150 |
|
| 151 |
DO I=1,MPLUS |
| 152 |
INV = (I-1) * MPLUS |
| 153 |
DO J=1,MPLUS |
| 154 |
JNV = (J-1) * MPLUS |
| 155 |
IF (J.EQ.ITERA) B(INV + J) = 0.D0 |
| 156 |
B(JNV + I) = B(INV + J) |
| 157 |
END DO |
| 158 |
END DO |
| 159 |
B(IONE) = 1.0D0 |
| 160 |
END DO |
| 161 |
END IF ! matches IF (NGEDIIS .NE. MSET) THEN |
| 162 |
|
| 163 |
! SCALE GEDIIS MATRIX BEFORE INVERSION: |
| 164 |
DO I=1,MPLUS |
| 165 |
II = MPLUS * (I-1) + I ! B(II)=diagonal elements of B matrix |
| 166 |
GSAVE(I) = 1.D0 / DSQRT(1.D-20+DABS(B(II))) |
| 167 |
!Print *, 'GSAVE(',I,')=', GSAVE(I)
|
| 168 |
END DO |
| 169 |
GSAVE(MPLUS) = 1.D0 |
| 170 |
DO I=1,MPLUS |
| 171 |
DO J=1,MPLUS |
| 172 |
IJ = MPLUS * (I-1) + J |
| 173 |
B(IJ) = B(IJ) * GSAVE(I) * GSAVE(J) |
| 174 |
END DO |
| 175 |
END DO |
| 176 |
|
| 177 |
! INVERT THE GEDIIS MATRIX B: |
| 178 |
DO I=1, MPLUS |
| 179 |
!WRITE(*,'(10(1X,F20.4))') B((I-1)*MPLUS+1:I*(MPLUS)) |
| 180 |
END DO |
| 181 |
|
| 182 |
CALL MINV(B,MPLUS,DET) ! matrix inversion. |
| 183 |
|
| 184 |
DO I=1, MPLUS |
| 185 |
!WRITE(*,'(10(1X,F20.16))') B((I-1)*MPLUS+1:I*(MPLUS)) |
| 186 |
END DO |
| 187 |
|
| 188 |
DO I=1,MPLUS |
| 189 |
DO J=1,MPLUS |
| 190 |
IJ = MPLUS * (I-1) + J |
| 191 |
B(IJ) = B(IJ) * GSAVE(I) * GSAVE(J) |
| 192 |
END DO |
| 193 |
END DO |
| 194 |
|
| 195 |
! COMPUTE THE NEW INTERPOLATED PARAMETER VECTOR (Geometry): |
| 196 |
ci=0.d0 |
| 197 |
ci_tmp=0.d0 |
| 198 |
|
| 199 |
ci_lt_zero= .FALSE. |
| 200 |
DO I=1, MSET |
| 201 |
DO J=1, MSET ! B matrix is read column-wise |
| 202 |
ci(I)=ci(I)+B((J-1)*(MPLUS)+I)*ESET(J) !ESET is energy set, yet to be fixed. |
| 203 |
END DO |
| 204 |
ci(I)=ci(I)+B((MPLUS-1)*(MPLUS)+I) |
| 205 |
!Print *, 'NO ci < 0 yet, c(',I,')=', ci(I)
|
| 206 |
IF((ci(I) .LT. 0.0D0) .OR. (ci(I) .GT. 1.0D0)) THEN |
| 207 |
ci_lt_zero=.TRUE. |
| 208 |
EXIT |
| 209 |
END IF |
| 210 |
END DO !matches DO I=1, MSET |
| 211 |
|
| 212 |
IF (ci_lt_zero) Then |
| 213 |
cis_zero = 0 |
| 214 |
ER_star = 0.D0 |
| 215 |
ER_star_tmp = 1e32 |
| 216 |
|
| 217 |
! B_ij calculations from <B_ij=(g_i-g_j)(R_i-R_j)>, Full B matrix created first and then rows and columns are removed. |
| 218 |
JJ=0 |
| 219 |
INV=-NCoord |
| 220 |
DO IX=1,MSET |
| 221 |
INV=INV+NCoord |
| 222 |
JNV=-NCoord |
| 223 |
DO JX=1,MSET |
| 224 |
JNV=JNV+NCoord |
| 225 |
JJ = JJ + 1 |
| 226 |
BST(JJ)=0.D0 |
| 227 |
DO KX=1, NCoord |
| 228 |
BST(JJ) = BST(JJ) + (((GradSet(INV+KX)-GradSet(JNV+KX))*(GeomSet(INV+KX)-GeomSet(JNV+KX)))/2.D0) |
| 229 |
END DO |
| 230 |
END DO |
| 231 |
END DO |
| 232 |
|
| 233 |
DO I=1, (2**MSET)-2 ! all (2**MSET)-2 combinations of cis, except the one where all cis are .GT. 0 and .LT. 1 |
| 234 |
ci(:)=1.D0 |
| 235 |
! find out which cis are zero in I: |
| 236 |
DO IX=1, MSET |
| 237 |
JJ=IAND(I, 2**(IX-1)) |
| 238 |
IF(JJ .EQ. 0) Then |
| 239 |
ci(IX)=0.D0 |
| 240 |
END IF |
| 241 |
END DO |
| 242 |
|
| 243 |
ci_lt_zero = .FALSE. |
| 244 |
! B_ij calculations from <B_ij=(g_i-g_j)(R_i-R_j)>, Full B matrix created first and then rows and columns are removed. |
| 245 |
DO IX=1, MSET*MSET |
| 246 |
B(IX) = BST(IX) !just a copy of the original B (GEDIIS) matrix |
| 247 |
END DO |
| 248 |
|
| 249 |
! Removal of KXth row and KXth column in order to accomodate cI to be zero: |
| 250 |
current_size_B_mat=MSET |
| 251 |
cis_zero = 0 |
| 252 |
! The bits of I (index of the upper loop 'DO I=1, (2**MSET)-2'), gives which cis are zero. |
| 253 |
DO KX=1, MSET ! searching for each bit of I (index of the upper loop 'DO I=1, (2**MSET)-2') |
| 254 |
IF (ci(KX) .EQ. 0.D0) Then !remove KXth row and KXth column |
| 255 |
cis_zero = cis_zero + 1 |
| 256 |
|
| 257 |
! First row removal: (B matrix is read column-wise) |
| 258 |
JJ=0 |
| 259 |
DO IX=1,current_size_B_mat ! columns reading |
| 260 |
DO JX=1,current_size_B_mat ! rows reading |
| 261 |
IF (JX .NE. KX) Then |
| 262 |
JJ = JJ + 1 |
| 263 |
B_tmp(JJ) = B((IX-1)*current_size_B_mat+JX) |
| 264 |
END IF |
| 265 |
END DO |
| 266 |
END DO |
| 267 |
|
| 268 |
DO IX=1,current_size_B_mat*(current_size_B_mat-1) |
| 269 |
B(IX) = B_tmp(IX) |
| 270 |
END DO |
| 271 |
|
| 272 |
! Now column removal: |
| 273 |
JJ=0 |
| 274 |
DO IX=1,KX-1 ! columns reading |
| 275 |
DO JX=1,current_size_B_mat-1 ! rows reading |
| 276 |
JJ = JJ + 1 |
| 277 |
B_tmp(JJ) = B(JJ) |
| 278 |
END DO |
| 279 |
END DO |
| 280 |
|
| 281 |
DO IX=KX+1,current_size_B_mat |
| 282 |
DO JX=1,current_size_B_mat-1 |
| 283 |
JJ = JJ + 1 |
| 284 |
B_tmp(JJ) = B(JJ+current_size_B_mat-1) |
| 285 |
END DO |
| 286 |
END DO |
| 287 |
|
| 288 |
DO IX=1,(current_size_B_mat-1)*(current_size_B_mat-1) |
| 289 |
B(IX) = B_tmp(IX) |
| 290 |
END DO |
| 291 |
current_size_B_mat = current_size_B_mat - 1 |
| 292 |
END IF ! matches IF (ci(KX) .EQ. 0.D0) Then !remove |
| 293 |
END DO ! matches DO KX=1, MSET |
| 294 |
|
| 295 |
! The following shifting is required to correct indices of B_ij elements in the GEDIIS matrix. |
| 296 |
! The correction is needed because the last coloumn and row of the matrix contains all 1 and one zero. |
| 297 |
DO IX=MSET-cis_zero-1,1,-1 |
| 298 |
DO JX=MSET-cis_zero,1,-1 |
| 299 |
B(IX*(MSET-cis_zero)+JX+IX) = B(IX*(MSET-cis_zero)+JX) |
| 300 |
END DO |
| 301 |
END DO |
| 302 |
|
| 303 |
! for last row and last column of GEDIIS matrix |
| 304 |
DO IX=1,MPLUS-cis_zero |
| 305 |
B((MPLUS-cis_zero)*IX) = 1.D0 |
| 306 |
B((MPLUS-cis_zero)*(MSET-cis_zero)+IX) = 1.D0 |
| 307 |
END DO |
| 308 |
B((MPLUS-cis_zero) * (MPLUS-cis_zero)) = 0.D0 |
| 309 |
|
| 310 |
DO IX=1, MPLUS |
| 311 |
!WRITE(*,'(10(1X,F20.4))') B((IX-1)*MPLUS+1:IX*(MPLUS)) |
| 312 |
END DO |
| 313 |
|
| 314 |
! ELIMINATE ERROR VECTORS WITH THE LARGEST NORM: |
| 315 |
IF (NGEDIIS .NE. MSET) THEN |
| 316 |
JX = min(MSET-NGEDIIS,MSET-cis_zero-1) |
| 317 |
DO II=1,JX |
| 318 |
XMAX = -1.D10 |
| 319 |
ITERA = 0 |
| 320 |
DO IX=1,MSET-cis_zero |
| 321 |
XNORM = 0.D0 |
| 322 |
INV = (IX-1) * (MPLUS-cis_zero) |
| 323 |
DO J=1,MSET-cis_zero |
| 324 |
XNORM = XNORM + ABS(B(INV + J)) |
| 325 |
END DO |
| 326 |
IF (XMAX.LT.XNORM .AND. XNORM.NE.1.0D0) THEN |
| 327 |
XMAX = XNORM |
| 328 |
ITERA = IX |
| 329 |
IONE = INV + IX |
| 330 |
ENDIF |
| 331 |
END DO |
| 332 |
|
| 333 |
DO IX=1,MPLUS-cis_zero |
| 334 |
INV = (IX-1) * (MPLUS-cis_zero) |
| 335 |
DO J=1,MPLUS-cis_zero |
| 336 |
JNV = (J-1) * (MPLUS-cis_zero) |
| 337 |
IF (J.EQ.ITERA) B(INV + J) = 0.D0 |
| 338 |
B(JNV + IX) = B(INV + J) |
| 339 |
END DO |
| 340 |
END DO |
| 341 |
B(IONE) = 1.0D0 |
| 342 |
END DO |
| 343 |
END IF ! matches IF (NGEDIIS .NE. MSET) THEN |
| 344 |
|
| 345 |
! SCALE GEDIIS MATRIX BEFORE INVERSION: |
| 346 |
DO IX=1,MPLUS-cis_zero |
| 347 |
II = (MPLUS-cis_zero) * (IX-1) + IX ! B(II)=diagonal elements of B matrix |
| 348 |
GSAVE(IX) = 1.D0 / DSQRT(1.D-20+DABS(B(II))) |
| 349 |
END DO |
| 350 |
GSAVE(MPLUS-cis_zero) = 1.D0 |
| 351 |
DO IX=1,MPLUS-cis_zero |
| 352 |
DO JX=1,MPLUS-cis_zero |
| 353 |
IJ = (MPLUS-cis_zero) * (IX-1) + JX |
| 354 |
B(IJ) = B(IJ) * GSAVE(IX) * GSAVE(JX) |
| 355 |
END DO |
| 356 |
END DO |
| 357 |
|
| 358 |
! INVERT THE GEDIIS MATRIX B: |
| 359 |
CALL MINV(B,MPLUS-cis_zero,DET) ! matrix inversion. |
| 360 |
|
| 361 |
DO IX=1,MPLUS-cis_zero |
| 362 |
DO JX=1,MPLUS-cis_zero |
| 363 |
IJ = (MPLUS-cis_zero) * (IX-1) + JX |
| 364 |
B(IJ) = B(IJ) * GSAVE(IX) * GSAVE(JX) |
| 365 |
END DO |
| 366 |
END DO |
| 367 |
|
| 368 |
DO IX=1, MPLUS |
| 369 |
!WRITE(*,'(10(1X,F20.4))') B((IX-1)*MPLUS+1:IX*(MPLUS)) |
| 370 |
END DO |
| 371 |
|
| 372 |
! ESET is rearranged to handle zero cis and stored in ESET_tmp: |
| 373 |
JJ=0 |
| 374 |
DO IX=1, MSET |
| 375 |
IF (ci(IX) .NE. 0) Then |
| 376 |
JJ=JJ+1 |
| 377 |
ESET_tmp(JJ) = ESET(IX) |
| 378 |
END IF |
| 379 |
END DO |
| 380 |
|
| 381 |
! DETERMINATION OF nonzero cis: |
| 382 |
MyPointer=1 |
| 383 |
DO IX=1, MSET-cis_zero |
| 384 |
tmp = 0.D0 |
| 385 |
DO J=1, MSET-cis_zero ! B matrix is read column-wise |
| 386 |
tmp=tmp+B((J-1)*(MPLUS-cis_zero)+IX)*ESET_tmp(J) |
| 387 |
END DO |
| 388 |
tmp=tmp+B((MPLUS-cis_zero-1)*(MPLUS-cis_zero)+IX) |
| 389 |
IF((tmp .LT. 0.0D0) .OR. (tmp .GT. 1.0D0)) THEN |
| 390 |
ci_lt_zero=.TRUE. |
| 391 |
EXIT |
| 392 |
ELSE |
| 393 |
DO JX=MyPointer,MSET |
| 394 |
IF (ci(JX) .NE. 0) Then |
| 395 |
ci(JX) = tmp |
| 396 |
MyPointer=JX+1 |
| 397 |
EXIT |
| 398 |
END IF |
| 399 |
END DO |
| 400 |
END IF |
| 401 |
END DO !matches DO I=1, MSET-cis_zero |
| 402 |
!Print *, 'Local set of cis, first 10:, MSET=', MSET, ', I of (2**MSET)-2=', I |
| 403 |
!WRITE(*,'(10(1X,F20.4))') ci(1:MSET) |
| 404 |
!Print *, 'Local set of cis ends:****************************************' |
| 405 |
|
| 406 |
! new set of cis determined based on the lower energy (ER_star): |
| 407 |
IF (.NOT. ci_lt_zero) Then |
| 408 |
Call Energy_GEDIIS(MRESET,MSET,ci,GeomSet,GradSet,ESET,NCoord,ER_star) |
| 409 |
IF (ER_star .LT. ER_star_tmp) Then |
| 410 |
ci_tmp=ci |
| 411 |
ER_star_tmp = ER_star |
| 412 |
END IF |
| 413 |
END IF ! matches IF (.NOT. ci_lt_zero) Then |
| 414 |
END DO !matches DO I=1, (2**K)-2 ! all (2**K)-2 combinations of cis, except the one where all cis are .GT. 0 and .LT. 1 |
| 415 |
ci = ci_tmp |
| 416 |
END IF! matches IF (ci_lt_zero) Then |
| 417 |
|
| 418 |
Print *, 'Final set of cis, first 10:***********************************' |
| 419 |
WRITE(*,'(10(1X,F20.4))') ci(1:MSET) |
| 420 |
Print *, 'Final set of cis ends:****************************************' |
| 421 |
Geom_new(:) = 0.D0 |
| 422 |
DO I=1, MSET |
| 423 |
Geom_new(:) = Geom_new(:) + (ci(I)*GeomSet((I-1)*NCoord+1:I*NCoord)) !MPLUS=MSET+1 |
| 424 |
! R_(N+1)=R*+DeltaR: |
| 425 |
DO J=1, NCoord |
| 426 |
tmp=0.D0 |
| 427 |
DO K=1,NCoord |
| 428 |
!tmp=tmp+Hess((J-1)*NCoord+K)*GradSet((I-1)*NCoord+K) ! If Hinv=.False., then we need to invert Hess |
| 429 |
END DO |
| 430 |
Geom_new(J) = Geom_new(J) - (ci(I)*tmp) |
| 431 |
END DO |
| 432 |
END DO |
| 433 |
|
| 434 |
DX(:) = Geom(:) - Geom_new(:) |
| 435 |
|
| 436 |
XNORM = SQRT(DOT_PRODUCT(DX,DX)) |
| 437 |
IF (PRINT) THEN |
| 438 |
WRITE (6,'(/10X,''DEVIATION IN X '',F10.4,8X,''DETERMINANT '',G9.3)') XNORM, DET |
| 439 |
!WRITE(*,'(10X,''GEDIIS COEFFICIENTS'')') |
| 440 |
!WRITE(*,'(10X,5F12.5)') (B(MPLUS*MSET+I),I=1,MSET) |
| 441 |
ENDIF |
| 442 |
|
| 443 |
! THE FOLLOWING TOLERENCES FOR XNORM AND DET ARE SOMEWHAT ARBITRARY! |
| 444 |
THRES = MAX(10.D0**(-NCoord), 1.D-25) |
| 445 |
IF (XNORM.GT.2.D0 .OR. DABS(DET) .LT. THRES) THEN |
| 446 |
IF (PRINT)THEN |
| 447 |
WRITE(*,*) "THE GEDIIS MATRIX IS ILL CONDITIONED" |
| 448 |
WRITE(*,*) " - PROBABLY, VECTORS ARE LINEARLY DEPENDENT - " |
| 449 |
WRITE(*,*) "THE GEDIIS STEP WILL BE REPEATED WITH A SMALLER SPACE" |
| 450 |
END IF |
| 451 |
DO K=1,MM |
| 452 |
B(K) = BS(K) ! why this is reverted? Because "IF (NGEDIIS .GT. 0) GO TO 80", see below |
| 453 |
END DO |
| 454 |
NGEDIIS = NGEDIIS - 1 |
| 455 |
IF (NGEDIIS .GT. 0) GO TO 80 |
| 456 |
IF (PRINT) WRITE(*,'(10X,''NEWTON-RAPHSON STEP TAKEN'')') |
| 457 |
Geom_new(:) = Geom(:) ! Geom_new is set to original Geom, thus DX = Geom(:) - Geom_new(:)=zero |
| 458 |
END IF ! matches IF (XNORM.GT.2.D0 .OR. DABS(DET).LT. THRES) THEN |
| 459 |
|
| 460 |
!******************************************************************************************************************* |
| 461 |
Geom_new(:) = 0.D0 |
| 462 |
DO I=1, MSET |
| 463 |
Geom_new(:) = Geom_new(:) + (ci(I)*GeomSet((I-1)*NCoord+1:I*NCoord)) !MPLUS=MSET+1 |
| 464 |
! R_(N+1)=R*+DeltaR: |
| 465 |
DO J=1, NCoord |
| 466 |
tmp=0.D0 |
| 467 |
DO K=1,NCoord |
| 468 |
tmp=tmp+Hess((J-1)*NCoord+K)*GradSet((I-1)*NCoord+K) ! If Hinv=.False., then we need to invert Hess |
| 469 |
END DO |
| 470 |
Geom_new(J) = Geom_new(J) - (ci(I)*tmp) |
| 471 |
END DO |
| 472 |
END DO |
| 473 |
!******************************************************************************************************************* |
| 474 |
|
| 475 |
IF (PRINT) WRITE(*,'(/,'' END GEDIIS '',/)') |
| 476 |
|
| 477 |
END SUBROUTINE Step_GEDIIS |
| 478 |
|