ViewVC Help
View File | Revision Log | Show Annotations | View Changeset | Root Listing
root/group/trunk/OOPSE/libmdtools/NPTf.cpp
Revision: 772
Committed: Fri Sep 19 16:01:07 2003 UTC (20 years, 9 months ago) by gezelter
File size: 13271 byte(s)
Log Message:
fixed bugs in NPTf, found (nearly) conserved quantities for both NPTi
and NPTf

File Contents

# Content
1 #include <cmath>
2 #include "Atom.hpp"
3 #include "SRI.hpp"
4 #include "AbstractClasses.hpp"
5 #include "SimInfo.hpp"
6 #include "ForceFields.hpp"
7 #include "Thermo.hpp"
8 #include "ReadWrite.hpp"
9 #include "Integrator.hpp"
10 #include "simError.h"
11
12 #ifdef IS_MPI
13 #include "mpiSimulation.hpp"
14 #endif
15
16 // Basic non-isotropic thermostating and barostating via the Melchionna
17 // modification of the Hoover algorithm:
18 //
19 // Melchionna, S., Ciccotti, G., and Holian, B. L., 1993,
20 // Molec. Phys., 78, 533.
21 //
22 // and
23 //
24 // Hoover, W. G., 1986, Phys. Rev. A, 34, 2499.
25
26 template<typename T> NPTf<T>::NPTf ( SimInfo *theInfo, ForceFields* the_ff):
27 T( theInfo, the_ff )
28 {
29 int i, j;
30 chi = 0.0;
31 integralOfChidt = 0.0;
32
33 for(i = 0; i < 3; i++)
34 for (j = 0; j < 3; j++)
35 eta[i][j] = 0.0;
36
37 have_tau_thermostat = 0;
38 have_tau_barostat = 0;
39 have_target_temp = 0;
40 have_target_pressure = 0;
41
42 have_chi_tolerance = 0;
43 have_eta_tolerance = 0;
44 have_pos_iter_tolerance = 0;
45
46 oldPos = new double[3*nAtoms];
47 oldVel = new double[3*nAtoms];
48 oldJi = new double[3*nAtoms];
49 #ifdef IS_MPI
50 Nparticles = mpiSim->getTotAtoms();
51 #else
52 Nparticles = theInfo->n_atoms;
53 #endif
54
55 }
56
57 template<typename T> NPTf<T>::~NPTf() {
58 delete[] oldPos;
59 delete[] oldVel;
60 delete[] oldJi;
61 }
62
63 template<typename T> void NPTf<T>::moveA() {
64
65 // new version of NPTf
66 int i, j, k;
67 DirectionalAtom* dAtom;
68 double Tb[3], ji[3];
69 double A[3][3], I[3][3];
70 double angle, mass;
71 double vel[3], pos[3], frc[3];
72
73 double rj[3];
74 double instaTemp, instaPress, instaVol;
75 double tt2, tb2;
76 double sc[3];
77 double eta2ij;
78 double press[3][3], vScale[3][3], hm[3][3], hmnew[3][3], scaleMat[3][3];
79 double bigScale, smallScale, offDiagMax;
80 double COM[3];
81
82 tt2 = tauThermostat * tauThermostat;
83 tb2 = tauBarostat * tauBarostat;
84
85 instaTemp = tStats->getTemperature();
86 tStats->getPressureTensor(press);
87 instaVol = tStats->getVolume();
88
89 tStats->getCOM(COM);
90
91 //calculate scale factor of veloity
92 for (i = 0; i < 3; i++ ) {
93 for (j = 0; j < 3; j++ ) {
94 vScale[i][j] = eta[i][j];
95
96 if (i == j) {
97 vScale[i][j] += chi;
98 }
99 }
100 }
101
102 //evolve velocity half step
103 for( i=0; i<nAtoms; i++ ){
104
105 atoms[i]->getVel( vel );
106 atoms[i]->getFrc( frc );
107
108 mass = atoms[i]->getMass();
109
110 info->matVecMul3( vScale, vel, sc );
111
112 for (j=0; j < 3; j++) {
113 // velocity half step
114 vel[j] += dt2 * ((frc[j] / mass) * eConvert - sc[j]);
115 }
116
117 atoms[i]->setVel( vel );
118
119 if( atoms[i]->isDirectional() ){
120
121 dAtom = (DirectionalAtom *)atoms[i];
122
123 // get and convert the torque to body frame
124
125 dAtom->getTrq( Tb );
126 dAtom->lab2Body( Tb );
127
128 // get the angular momentum, and propagate a half step
129
130 dAtom->getJ( ji );
131
132 for (j=0; j < 3; j++)
133 ji[j] += dt2 * (Tb[j] * eConvert - ji[j]*chi);
134
135 // use the angular velocities to propagate the rotation matrix a
136 // full time step
137
138 dAtom->getA(A);
139 dAtom->getI(I);
140
141 // rotate about the x-axis
142 angle = dt2 * ji[0] / I[0][0];
143 this->rotate( 1, 2, angle, ji, A );
144
145 // rotate about the y-axis
146 angle = dt2 * ji[1] / I[1][1];
147 this->rotate( 2, 0, angle, ji, A );
148
149 // rotate about the z-axis
150 angle = dt * ji[2] / I[2][2];
151 this->rotate( 0, 1, angle, ji, A);
152
153 // rotate about the y-axis
154 angle = dt2 * ji[1] / I[1][1];
155 this->rotate( 2, 0, angle, ji, A );
156
157 // rotate about the x-axis
158 angle = dt2 * ji[0] / I[0][0];
159 this->rotate( 1, 2, angle, ji, A );
160
161 dAtom->setJ( ji );
162 dAtom->setA( A );
163 }
164 }
165
166 // advance chi half step
167 chi += dt2 * ( instaTemp / targetTemp - 1.0) / tt2;
168
169 // calculate the integral of chidt
170 integralOfChidt += dt2*chi;
171
172 // advance eta half step
173
174 for(i = 0; i < 3; i ++)
175 for(j = 0; j < 3; j++){
176 if( i == j)
177 eta[i][j] += dt2 * instaVol *
178 (press[i][j] - targetPressure/p_convert) / (NkBT*tb2);
179 else
180 eta[i][j] += dt2 * instaVol * press[i][j] / (NkBT*tb2);
181 }
182
183 //save the old positions
184 for(i = 0; i < nAtoms; i++){
185 atoms[i]->getPos(pos);
186 for(j = 0; j < 3; j++)
187 oldPos[i*3 + j] = pos[j];
188 }
189
190 //the first estimation of r(t+dt) is equal to r(t)
191
192 for(k = 0; k < 4; k ++){
193
194 for(i =0 ; i < nAtoms; i++){
195
196 atoms[i]->getVel(vel);
197 atoms[i]->getPos(pos);
198
199 for(j = 0; j < 3; j++)
200 rj[j] = (oldPos[i*3 + j] + pos[j])/2 - COM[j];
201
202 info->matVecMul3( eta, rj, sc );
203
204 for(j = 0; j < 3; j++)
205 pos[j] = oldPos[i*3 + j] + dt*(vel[j] + sc[j]);
206
207 atoms[i]->setPos( pos );
208
209 }
210
211 if (nConstrained) {
212 constrainA();
213 }
214 }
215
216
217 // Scale the box after all the positions have been moved:
218
219 // Use a taylor expansion for eta products: Hmat = Hmat . exp(dt * etaMat)
220 // Hmat = Hmat . ( Ident + dt * etaMat + dt^2 * etaMat*etaMat / 2)
221
222 bigScale = 1.0;
223 smallScale = 1.0;
224 offDiagMax = 0.0;
225
226 for(i=0; i<3; i++){
227 for(j=0; j<3; j++){
228
229 // Calculate the matrix Product of the eta array (we only need
230 // the ij element right now):
231
232 eta2ij = 0.0;
233 for(k=0; k<3; k++){
234 eta2ij += eta[i][k] * eta[k][j];
235 }
236
237 scaleMat[i][j] = 0.0;
238 // identity matrix (see above):
239 if (i == j) scaleMat[i][j] = 1.0;
240 // Taylor expansion for the exponential truncated at second order:
241 scaleMat[i][j] += dt*eta[i][j] + 0.5*dt*dt*eta2ij;
242
243 if (i != j)
244 if (fabs(scaleMat[i][j]) > offDiagMax)
245 offDiagMax = fabs(scaleMat[i][j]);
246 }
247
248 if (scaleMat[i][i] > bigScale) bigScale = scaleMat[i][i];
249 if (scaleMat[i][i] < smallScale) smallScale = scaleMat[i][i];
250 }
251
252 if ((bigScale > 1.1) || (smallScale < 0.9)) {
253 sprintf( painCave.errMsg,
254 "NPTf error: Attempting a Box scaling of more than 10 percent.\n"
255 " Check your tauBarostat, as it is probably too small!\n\n"
256 " scaleMat = [%lf\t%lf\t%lf]\n"
257 " [%lf\t%lf\t%lf]\n"
258 " [%lf\t%lf\t%lf]\n",
259 scaleMat[0][0],scaleMat[0][1],scaleMat[0][2],
260 scaleMat[1][0],scaleMat[1][1],scaleMat[1][2],
261 scaleMat[2][0],scaleMat[2][1],scaleMat[2][2]);
262 painCave.isFatal = 1;
263 simError();
264 } else if (offDiagMax > 0.1) {
265 sprintf( painCave.errMsg,
266 "NPTf error: Attempting an off-diagonal Box scaling of more than 10 percent.\n"
267 " Check your tauBarostat, as it is probably too small!\n\n"
268 " scaleMat = [%lf\t%lf\t%lf]\n"
269 " [%lf\t%lf\t%lf]\n"
270 " [%lf\t%lf\t%lf]\n",
271 scaleMat[0][0],scaleMat[0][1],scaleMat[0][2],
272 scaleMat[1][0],scaleMat[1][1],scaleMat[1][2],
273 scaleMat[2][0],scaleMat[2][1],scaleMat[2][2]);
274 painCave.isFatal = 1;
275 simError();
276 } else {
277 info->getBoxM(hm);
278 info->matMul3(hm, scaleMat, hmnew);
279 info->setBoxM(hmnew);
280 }
281
282 }
283
284 template<typename T> void NPTf<T>::moveB( void ){
285
286 //new version of NPTf
287 int i, j, k;
288 DirectionalAtom* dAtom;
289 double Tb[3], ji[3];
290 double vel[3], myVel[3], frc[3];
291 double mass;
292
293 double instaTemp, instaPress, instaVol;
294 double tt2, tb2;
295 double sc[3];
296 double press[3][3], vScale[3][3];
297 double oldChi, prevChi;
298 double oldEta[3][3], prevEta[3][3], diffEta;
299
300 tt2 = tauThermostat * tauThermostat;
301 tb2 = tauBarostat * tauBarostat;
302
303 // Set things up for the iteration:
304
305 oldChi = chi;
306
307 for(i = 0; i < 3; i++)
308 for(j = 0; j < 3; j++)
309 oldEta[i][j] = eta[i][j];
310
311 for( i=0; i<nAtoms; i++ ){
312
313 atoms[i]->getVel( vel );
314
315 for (j=0; j < 3; j++)
316 oldVel[3*i + j] = vel[j];
317
318 if( atoms[i]->isDirectional() ){
319
320 dAtom = (DirectionalAtom *)atoms[i];
321
322 dAtom->getJ( ji );
323
324 for (j=0; j < 3; j++)
325 oldJi[3*i + j] = ji[j];
326
327 }
328 }
329
330 // do the iteration:
331
332 instaVol = tStats->getVolume();
333
334 for (k=0; k < 4; k++) {
335
336 instaTemp = tStats->getTemperature();
337 tStats->getPressureTensor(press);
338
339 // evolve chi another half step using the temperature at t + dt/2
340
341 prevChi = chi;
342 chi = oldChi + dt2 * ( instaTemp / targetTemp - 1.0) / tt2;
343
344 for(i = 0; i < 3; i++)
345 for(j = 0; j < 3; j++)
346 prevEta[i][j] = eta[i][j];
347
348 //advance eta half step and calculate scale factor for velocity
349
350 for(i = 0; i < 3; i ++)
351 for(j = 0; j < 3; j++){
352 if( i == j) {
353 eta[i][j] = oldEta[i][j] + dt2 * instaVol *
354 (press[i][j] - targetPressure/p_convert) / (NkBT*tb2);
355 vScale[i][j] = eta[i][j] + chi;
356 } else {
357 eta[i][j] = oldEta[i][j] + dt2 * instaVol * press[i][j] / (NkBT*tb2);
358 vScale[i][j] = eta[i][j];
359 }
360 }
361
362 for( i=0; i<nAtoms; i++ ){
363
364 atoms[i]->getFrc( frc );
365 atoms[i]->getVel(vel);
366
367 mass = atoms[i]->getMass();
368
369 for (j = 0; j < 3; j++)
370 myVel[j] = oldVel[3*i + j];
371
372 info->matVecMul3( vScale, myVel, sc );
373
374 // velocity half step
375 for (j=0; j < 3; j++) {
376 // velocity half step (use chi from previous step here):
377 vel[j] = oldVel[3*i+j] + dt2 * ((frc[j] / mass) * eConvert - sc[j]);
378 }
379
380 atoms[i]->setVel( vel );
381
382 if( atoms[i]->isDirectional() ){
383
384 dAtom = (DirectionalAtom *)atoms[i];
385
386 // get and convert the torque to body frame
387
388 dAtom->getTrq( Tb );
389 dAtom->lab2Body( Tb );
390
391 for (j=0; j < 3; j++)
392 ji[j] = oldJi[3*i + j] + dt2 * (Tb[j] * eConvert - oldJi[3*i+j]*chi);
393
394 dAtom->setJ( ji );
395 }
396 }
397
398 if (nConstrained) {
399 constrainB();
400 }
401
402 diffEta = 0;
403 for(i = 0; i < 3; i++)
404 diffEta += pow(prevEta[i][i] - eta[i][i], 2);
405
406 if (fabs(prevChi - chi) <= chiTolerance && sqrt(diffEta / 3) <= etaTolerance)
407 break;
408 }
409
410 //calculate integral of chidt
411 integralOfChidt += dt2*chi;
412
413 }
414
415 template<typename T> void NPTf<T>::resetIntegrator() {
416 int i,j;
417
418 chi = 0.0;
419
420 for(i = 0; i < 3; i++)
421 for (j = 0; j < 3; j++)
422 eta[i][j] = 0.0;
423
424 }
425
426 template<typename T> int NPTf<T>::readyCheck() {
427
428 //check parent's readyCheck() first
429 if (T::readyCheck() == -1)
430 return -1;
431
432 // First check to see if we have a target temperature.
433 // Not having one is fatal.
434
435 if (!have_target_temp) {
436 sprintf( painCave.errMsg,
437 "NPTf error: You can't use the NPTf integrator\n"
438 " without a targetTemp!\n"
439 );
440 painCave.isFatal = 1;
441 simError();
442 return -1;
443 }
444
445 if (!have_target_pressure) {
446 sprintf( painCave.errMsg,
447 "NPTf error: You can't use the NPTf integrator\n"
448 " without a targetPressure!\n"
449 );
450 painCave.isFatal = 1;
451 simError();
452 return -1;
453 }
454
455 // We must set tauThermostat.
456
457 if (!have_tau_thermostat) {
458 sprintf( painCave.errMsg,
459 "NPTf error: If you use the NPTf\n"
460 " integrator, you must set tauThermostat.\n");
461 painCave.isFatal = 1;
462 simError();
463 return -1;
464 }
465
466 // We must set tauBarostat.
467
468 if (!have_tau_barostat) {
469 sprintf( painCave.errMsg,
470 "NPTf error: If you use the NPTf\n"
471 " integrator, you must set tauBarostat.\n");
472 painCave.isFatal = 1;
473 simError();
474 return -1;
475 }
476
477
478 // We need NkBT a lot, so just set it here: This is the RAW number
479 // of particles, so no subtraction or addition of constraints or
480 // orientational degrees of freedom:
481
482 NkBT = (double)Nparticles * kB * targetTemp;
483
484 // fkBT is used because the thermostat operates on more degrees of freedom
485 // than the barostat (when there are particles with orientational degrees
486 // of freedom). ndf = 3 * (n_atoms + n_oriented -1) - n_constraint - nZcons
487
488 fkBT = (double)info->ndf * kB * targetTemp;
489
490 return 1;
491 }
492
493 template<typename T> double NPTf<T>::getConservedQuantity(void){
494
495 double conservedQuantity;
496 double Energy;
497 double thermostat_kinetic;
498 double thermostat_potential;
499 double barostat_kinetic;
500 double barostat_potential;
501 double trEta;
502 double a[3][3], b[3][3];
503
504 Energy = tStats->getTotalE();
505
506 thermostat_kinetic = fkBT* tauThermostat * tauThermostat * chi * chi /
507 (2.0 * eConvert);
508
509 thermostat_potential = fkBT* integralOfChidt / eConvert;
510
511 info->transposeMat3(eta, a);
512 info->matMul3(a, eta, b);
513 trEta = info->matTrace3(b);
514
515 barostat_kinetic = NkBT * tauBarostat * tauBarostat * trEta /
516 (2.0 * eConvert);
517
518 barostat_potential = (targetPressure * tStats->getVolume() / p_convert) /
519 eConvert;
520
521 conservedQuantity = Energy + thermostat_kinetic + thermostat_potential +
522 barostat_kinetic + barostat_potential;
523
524 cout.width(8);
525 cout.precision(8);
526
527 cerr << info->getTime() << "\t" << Energy << "\t" << thermostat_kinetic <<
528 "\t" << thermostat_potential << "\t" << barostat_kinetic <<
529 "\t" << barostat_potential << "\t" << conservedQuantity << endl;
530
531 return conservedQuantity;
532 }