Add gmx convert-trj
[gromacs.git] / src / gromacs / gmxana / gmx_dos.cpp
blob602e11db34e81d615b5dcd71483995fffde36a84
1 /*
2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2011,2012,2013,2014,2015,2016,2017,2018, by the GROMACS development team, led by
5 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
6 * and including many others, as listed in the AUTHORS file in the
7 * top-level source directory and at http://www.gromacs.org.
9 * GROMACS is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public License
11 * as published by the Free Software Foundation; either version 2.1
12 * of the License, or (at your option) any later version.
14 * GROMACS is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with GROMACS; if not, see
21 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
22 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
24 * If you want to redistribute modifications to GROMACS, please
25 * consider that scientific software is very special. Version
26 * control is crucial - bugs must be traceable. We will be happy to
27 * consider code for inclusion in the official distribution, but
28 * derived work must not be called official GROMACS. Details are found
29 * in the README & COPYING files - if they are missing, get the
30 * official version at http://www.gromacs.org.
32 * To help us fund GROMACS development, we humbly ask that you cite
33 * the research papers on the package. Check out http://www.gromacs.org.
35 #include "gmxpre.h"
37 #include <cmath>
38 #include <cstdio>
39 #include <cstdlib>
40 #include <cstring>
42 #include "gromacs/commandline/pargs.h"
43 #include "gromacs/commandline/viewit.h"
44 #include "gromacs/correlationfunctions/autocorr.h"
45 #include "gromacs/correlationfunctions/integrate.h"
46 #include "gromacs/fft/fft.h"
47 #include "gromacs/fileio/confio.h"
48 #include "gromacs/fileio/gmxfio.h"
49 #include "gromacs/fileio/trxio.h"
50 #include "gromacs/fileio/xvgr.h"
51 #include "gromacs/gmxana/gmx_ana.h"
52 #include "gromacs/math/functions.h"
53 #include "gromacs/math/units.h"
54 #include "gromacs/math/utilities.h"
55 #include "gromacs/math/vec.h"
56 #include "gromacs/topology/index.h"
57 #include "gromacs/topology/topology.h"
58 #include "gromacs/trajectory/trajectoryframe.h"
59 #include "gromacs/utility/arraysize.h"
60 #include "gromacs/utility/fatalerror.h"
61 #include "gromacs/utility/futil.h"
62 #include "gromacs/utility/pleasecite.h"
63 #include "gromacs/utility/smalloc.h"
65 enum {
66 VACF, MVACF, DOS, DOS_SOLID, DOS_DIFF, DOS_CP, DOS_S, DOS_A, DOS_E, DOS_NR
69 static int calcMoleculesInIndexGroup(const t_block *mols, int natoms, const int *index, int nindex)
71 int i = 0;
72 int mol = 0;
73 int nMol = 0;
74 int j;
76 while (i < nindex)
78 while (index[i] > mols->index[mol])
80 mol++;
81 if (mol >= mols->nr)
83 gmx_fatal(FARGS, "Atom index out of range: %d", index[i]+1);
86 for (j = mols->index[mol]; j < mols->index[mol+1]; j++)
88 if (index[i] != j)
90 gmx_fatal(FARGS, "The index group does not consist of whole molecules");
92 i++;
93 if (i > natoms)
95 gmx_fatal(FARGS, "Index contains atom numbers larger than the topology");
98 nMol++;
100 return nMol;
103 static double FD(double Delta, double f)
105 return (2*std::pow(Delta, -4.5)*std::pow(f, 7.5) -
106 6*std::pow(Delta, -3.0)*std::pow(f, 5.0) -
107 std::pow(Delta, -1.5)*std::pow(f, 3.5) +
108 6*std::pow(Delta, -1.5)*std::pow(f, 2.5) +
109 2*f - 2);
112 static double YYY(double f, double y)
114 return (2*gmx::power3(y*f) - gmx::square(f)*y*(1+6*y) +
115 (2+6*y)*f - 2);
118 static double calc_compress(double y)
120 if (y == 1)
122 return 0;
124 return ((1+y+gmx::square(y)-gmx::power3(y))/(gmx::power3(1-y)));
127 static double bisector(double Delta, double tol,
128 double ff0, double ff1,
129 double ff(double, double))
131 double fd, f, f0, f1;
132 double tolmin = 1e-8;
134 f0 = ff0;
135 f1 = ff1;
136 if (tol < tolmin)
138 fprintf(stderr, "Unrealistic tolerance %g for bisector. Setting it to %g\n", tol, tolmin);
139 tol = tolmin;
144 f = (f0+f1)*0.5;
145 fd = ff(Delta, f);
146 if (fd < 0)
148 f0 = f;
150 else if (fd > 0)
152 f1 = f;
154 else
156 return f;
159 while ((f1-f0) > tol);
161 return f;
164 static double calc_fluidicity(double Delta, double tol)
166 return bisector(Delta, tol, 0, 1, FD);
169 static double calc_y(double f, double Delta, double toler)
171 double y1, y2;
173 y1 = std::pow(f/Delta, 1.5);
174 y2 = bisector(f, toler, 0, 10000, YYY);
175 if (std::abs((y1-y2)/(y1+y2)) > 100*toler)
177 fprintf(stderr, "Inconsistency computing y: y1 = %f, y2 = %f, using y1.\n",
178 y1, y2);
181 return y1;
184 static double calc_Shs(double f, double y)
186 double fy = f*y;
188 return BOLTZ*(std::log(calc_compress(fy)) + fy*(3*fy-4)/gmx::square(1-fy));
191 static real wCsolid(real nu, real beta)
193 real bhn = beta*PLANCK*nu;
194 real ebn, koko;
196 if (bhn == 0)
198 return 1.0;
200 else
202 ebn = std::exp(bhn);
203 koko = gmx::square(1-ebn);
204 return gmx::square(bhn)*ebn/koko;
208 static real wSsolid(real nu, real beta)
210 real bhn = beta*PLANCK*nu;
212 if (bhn == 0)
214 return 1;
216 else
218 return bhn/std::expm1(bhn) - std::log1p(-std::exp(-bhn));
222 static real wAsolid(real nu, real beta)
224 real bhn = beta*PLANCK*nu;
226 if (bhn == 0)
228 return 0;
230 else
232 return std::log((1-std::exp(-bhn))/(std::exp(-bhn/2))) - std::log(bhn);
236 static real wEsolid(real nu, real beta)
238 real bhn = beta*PLANCK*nu;
240 if (bhn == 0)
242 return 1;
244 else
246 return bhn/2 + bhn/std::expm1(bhn)-1;
250 int gmx_dos(int argc, char *argv[])
252 const char *desc[] = {
253 "[THISMODULE] computes the Density of States from a simulations.",
254 "In order for this to be meaningful the velocities must be saved",
255 "in the trajecotry with sufficiently high frequency such as to cover",
256 "all vibrations. For flexible systems that would be around a few fs",
257 "between saving. Properties based on the DoS are printed on the",
258 "standard output.",
259 "Note that the density of states is calculated from the mass-weighted",
260 "autocorrelation, and by default only from the square of the real",
261 "component rather than absolute value. This means the shape can differ",
262 "substantially from the plain vibrational power spectrum you can",
263 "calculate with gmx velacc."
265 const char *bugs[] = {
266 "This program needs a lot of memory: total usage equals the number of atoms times 3 times number of frames times 4 (or 8 when run in double precision)."
268 FILE *fp, *fplog;
269 t_topology top;
270 int ePBC = -1;
271 t_trxframe fr;
272 matrix box;
273 int gnx;
274 real t0, t1;
275 t_trxstatus *status;
276 int nV, nframes, n_alloc, i, j, fftcode, Nmol, Natom;
277 double rho, dt, Vsum, V, tmass, dostot, dos2;
278 real **c1, **dos, mi, beta, bfac, *nu, *tt, stddev, c1j;
279 gmx_output_env_t *oenv;
280 gmx_fft_t fft;
281 double cP, DiffCoeff, Delta, f, y, z, sigHS, Shs, Sig, DoS0, recip_fac;
282 double wCdiff, wSdiff, wAdiff, wEdiff;
283 int grpNatoms;
284 int *index;
285 char *grpname;
286 double invNormalize;
287 gmx_bool normalizeAutocorrelation;
289 static gmx_bool bVerbose = TRUE, bAbsolute = FALSE, bNormalizeDos = FALSE;
290 static gmx_bool bRecip = FALSE;
291 static real Temp = 298.15, toler = 1e-6;
292 int min_frames = 100;
294 t_pargs pa[] = {
295 { "-v", FALSE, etBOOL, {&bVerbose},
296 "Be loud and noisy." },
297 { "-recip", FALSE, etBOOL, {&bRecip},
298 "Use cm^-1 on X-axis instead of 1/ps for DoS plots." },
299 { "-abs", FALSE, etBOOL, {&bAbsolute},
300 "Use the absolute value of the Fourier transform of the VACF as the Density of States. Default is to use the real component only" },
301 { "-normdos", FALSE, etBOOL, {&bNormalizeDos},
302 "Normalize the DoS such that it adds up to 3N. This should usually not be necessary." },
303 { "-T", FALSE, etREAL, {&Temp},
304 "Temperature in the simulation" },
305 { "-toler", FALSE, etREAL, {&toler},
306 "[HIDDEN]Tolerance when computing the fluidicity using bisection algorithm" }
309 t_filenm fnm[] = {
310 { efTRN, "-f", nullptr, ffREAD },
311 { efTPR, "-s", nullptr, ffREAD },
312 { efNDX, nullptr, nullptr, ffOPTRD },
313 { efXVG, "-vacf", "vacf", ffWRITE },
314 { efXVG, "-mvacf", "mvacf", ffWRITE },
315 { efXVG, "-dos", "dos", ffWRITE },
316 { efLOG, "-g", "dos", ffWRITE },
318 #define NFILE asize(fnm)
319 int npargs;
320 t_pargs *ppa;
321 const char *DoSlegend[] = {
322 "DoS(v)", "DoS(v)[Solid]", "DoS(v)[Diff]"
325 npargs = asize(pa);
326 ppa = add_acf_pargs(&npargs, pa);
327 if (!parse_common_args(&argc, argv, PCA_CAN_VIEW | PCA_CAN_TIME,
328 NFILE, fnm, npargs, ppa, asize(desc), desc,
329 asize(bugs), bugs, &oenv))
331 sfree(ppa);
332 return 0;
335 beta = 1/(Temp*BOLTZ);
337 fplog = gmx_fio_fopen(ftp2fn(efLOG, NFILE, fnm), "w");
338 fprintf(fplog, "Doing density of states analysis based on trajectory.\n");
339 please_cite(fplog, "Pascal2011a");
340 please_cite(fplog, "Caleman2011b");
342 read_tps_conf(ftp2fn(efTPR, NFILE, fnm), &top, &ePBC, nullptr, nullptr, box, TRUE);
344 /* Handle index groups */
345 get_index(&top.atoms, ftp2fn_null(efNDX, NFILE, fnm), 1, &grpNatoms, &index, &grpname);
347 V = det(box);
348 tmass = 0;
349 for (i = 0; i < grpNatoms; i++)
351 tmass += top.atoms.atom[index[i]].m;
354 Natom = grpNatoms;
355 Nmol = calcMoleculesInIndexGroup(&top.mols, top.atoms.nr, index, grpNatoms);
356 gnx = Natom*DIM;
358 /* Correlation stuff */
359 snew(c1, gnx);
360 for (i = 0; (i < gnx); i++)
362 c1[i] = nullptr;
365 read_first_frame(oenv, &status, ftp2fn(efTRN, NFILE, fnm), &fr, TRX_NEED_V);
366 t0 = fr.time;
368 n_alloc = 0;
369 nframes = 0;
370 Vsum = 0;
371 nV = 0;
374 if (fr.bBox)
376 V = det(fr.box);
377 Vsum += V;
378 nV++;
380 if (nframes >= n_alloc)
382 n_alloc += 100;
383 for (i = 0; i < gnx; i++)
385 srenew(c1[i], n_alloc);
388 for (i = 0; i < gnx; i += DIM)
390 c1[i+XX][nframes] = fr.v[index[i/DIM]][XX];
391 c1[i+YY][nframes] = fr.v[index[i/DIM]][YY];
392 c1[i+ZZ][nframes] = fr.v[index[i/DIM]][ZZ];
395 t1 = fr.time;
397 nframes++;
399 while (read_next_frame(oenv, status, &fr));
401 close_trx(status);
403 if (nframes < min_frames)
405 gmx_fatal(FARGS, "You need at least %d frames in the trajectory and you only have %d.", min_frames, nframes);
407 dt = (t1-t0)/(nframes-1);
408 if (nV > 0)
410 V = Vsum/nV;
412 if (bVerbose)
414 printf("Going to do %d fourier transforms of length %d. Hang on.\n",
415 gnx, nframes);
417 /* Unfortunately the -normalize program option for the autocorrelation
418 * function calculation is added as a hack with a static variable in the
419 * autocorrelation.c source. That would work if we called the normal
420 * do_autocorr(), but this routine overrides that by directly calling
421 * the low-level functionality. That unfortunately leads to ignoring the
422 * default value for the option (which is to normalize).
423 * Since the absolute value seems to be important for the subsequent
424 * analysis below, we detect the value directly from the option, calculate
425 * the autocorrelation without normalization, and then apply the
426 * normalization just to the autocorrelation output
427 * (or not, if the user asked for a non-normalized autocorrelation).
429 normalizeAutocorrelation = opt2parg_bool("-normalize", npargs, ppa);
431 /* Note that we always disable normalization here, regardless of user settings */
432 low_do_autocorr(nullptr, oenv, nullptr, nframes, gnx, nframes, c1, dt, eacNormal, 0, FALSE,
433 FALSE, FALSE, -1, -1, 0);
434 snew(dos, DOS_NR);
435 for (j = 0; (j < DOS_NR); j++)
437 snew(dos[j], nframes+4);
440 if (bVerbose)
442 printf("Going to merge the ACFs into the mass-weighted and plain ACF\n");
444 for (i = 0; (i < gnx); i += DIM)
446 mi = top.atoms.atom[index[i/DIM]].m;
447 for (j = 0; (j < nframes/2); j++)
449 c1j = (c1[i+XX][j] + c1[i+YY][j] + c1[i+ZZ][j]);
450 dos[VACF][j] += c1j/Natom;
451 dos[MVACF][j] += mi*c1j;
455 fp = xvgropen(opt2fn("-vacf", NFILE, fnm), "Velocity autocorrelation function",
456 "Time (ps)", "C(t)", oenv);
457 snew(tt, nframes/2);
459 invNormalize = normalizeAutocorrelation ? 1.0/dos[VACF][0] : 1.0;
461 for (j = 0; (j < nframes/2); j++)
463 tt[j] = j*dt;
464 fprintf(fp, "%10g %10g\n", tt[j], dos[VACF][j] * invNormalize);
466 xvgrclose(fp);
468 fp = xvgropen(opt2fn("-mvacf", NFILE, fnm), "Mass-weighted velocity autocorrelation function",
469 "Time (ps)", "C(t)", oenv);
471 invNormalize = normalizeAutocorrelation ? 1.0/dos[VACF][0] : 1.0;
473 for (j = 0; (j < nframes/2); j++)
475 fprintf(fp, "%10g %10g\n", tt[j], dos[MVACF][j] * invNormalize);
477 xvgrclose(fp);
479 if ((fftcode = gmx_fft_init_1d_real(&fft, nframes/2,
480 GMX_FFT_FLAG_NONE)) != 0)
482 gmx_fatal(FARGS, "gmx_fft_init_1d_real returned %d", fftcode);
484 if ((fftcode = gmx_fft_1d_real(fft, GMX_FFT_REAL_TO_COMPLEX,
485 dos[MVACF], dos[DOS])) != 0)
487 gmx_fatal(FARGS, "gmx_fft_1d_real returned %d", fftcode);
490 /* First compute the DoS */
491 /* Magic factor of 8 included now. */
492 bfac = 8*dt*beta/2;
493 dos2 = 0;
494 snew(nu, nframes/4);
495 for (j = 0; (j < nframes/4); j++)
497 nu[j] = 2*j/(t1-t0);
498 dos2 += gmx::square(dos[DOS][2*j]) + gmx::square(dos[DOS][2*j+1]);
499 if (bAbsolute)
501 dos[DOS][j] = bfac*std::hypot(dos[DOS][2*j], dos[DOS][2*j+1]);
503 else
505 dos[DOS][j] = bfac*dos[DOS][2*j];
508 /* Normalize it */
509 dostot = evaluate_integral(nframes/4, nu, dos[DOS], nullptr, int{nframes/4}, &stddev);
510 if (bNormalizeDos)
512 for (j = 0; (j < nframes/4); j++)
514 dos[DOS][j] *= 3*Natom/dostot;
518 /* Now analyze it */
519 DoS0 = dos[DOS][0];
521 /* Note this eqn. is incorrect in Pascal2011a! */
522 Delta = ((2*DoS0/(9*Natom))*std::sqrt(M_PI*BOLTZ*Temp*Natom/tmass)*
523 std::pow((Natom/V), 1.0/3.0)*std::pow(6.0/M_PI, 2.0/3.0));
524 f = calc_fluidicity(Delta, toler);
525 y = calc_y(f, Delta, toler);
526 z = calc_compress(y);
527 Sig = BOLTZ*(5.0/2.0+std::log(2*M_PI*BOLTZ*Temp/(gmx::square(PLANCK))*V/(f*Natom)));
528 Shs = Sig+calc_Shs(f, y);
529 rho = (tmass*AMU)/(V*NANO*NANO*NANO);
530 sigHS = std::cbrt(6*y*V/(M_PI*Natom));
532 fprintf(fplog, "System = \"%s\"\n", *top.name);
533 fprintf(fplog, "Nmol = %d\n", Nmol);
534 fprintf(fplog, "Natom = %d\n", Natom);
535 fprintf(fplog, "dt = %g ps\n", dt);
536 fprintf(fplog, "tmass = %g amu\n", tmass);
537 fprintf(fplog, "V = %g nm^3\n", V);
538 fprintf(fplog, "rho = %g g/l\n", rho);
539 fprintf(fplog, "T = %g K\n", Temp);
540 fprintf(fplog, "beta = %g mol/kJ\n", beta);
542 fprintf(fplog, "\nDoS parameters\n");
543 fprintf(fplog, "Delta = %g\n", Delta);
544 fprintf(fplog, "fluidicity = %g\n", f);
545 fprintf(fplog, "hard sphere packing fraction = %g\n", y);
546 fprintf(fplog, "hard sphere compressibility = %g\n", z);
547 fprintf(fplog, "ideal gas entropy = %g\n", Sig);
548 fprintf(fplog, "hard sphere entropy = %g\n", Shs);
549 fprintf(fplog, "sigma_HS = %g nm\n", sigHS);
550 fprintf(fplog, "DoS0 = %g\n", DoS0);
551 fprintf(fplog, "Dos2 = %g\n", dos2);
552 fprintf(fplog, "DoSTot = %g\n", dostot);
554 /* Now compute solid (2) and diffusive (3) components */
555 fp = xvgropen(opt2fn("-dos", NFILE, fnm), "Density of states",
556 bRecip ? "E (cm\\S-1\\N)" : "\\f{12}n\\f{4} (1/ps)",
557 "\\f{4}S(\\f{12}n\\f{4})", oenv);
558 xvgr_legend(fp, asize(DoSlegend), DoSlegend, oenv);
559 recip_fac = bRecip ? (1e7/SPEED_OF_LIGHT) : 1.0;
560 for (j = 0; (j < nframes/4); j++)
562 dos[DOS_DIFF][j] = DoS0/(1+gmx::square(DoS0*M_PI*nu[j]/(6*f*Natom)));
563 dos[DOS_SOLID][j] = dos[DOS][j]-dos[DOS_DIFF][j];
564 fprintf(fp, "%10g %10g %10g %10g\n",
565 recip_fac*nu[j],
566 dos[DOS][j]/recip_fac,
567 dos[DOS_SOLID][j]/recip_fac,
568 dos[DOS_DIFF][j]/recip_fac);
570 xvgrclose(fp);
572 /* Finally analyze the results! */
573 wCdiff = 0.5;
574 wSdiff = Shs/(3*BOLTZ); /* Is this correct? */
575 wEdiff = 0.5;
576 wAdiff = wEdiff-wSdiff;
577 for (j = 0; (j < nframes/4); j++)
579 dos[DOS_CP][j] = (dos[DOS_DIFF][j]*wCdiff +
580 dos[DOS_SOLID][j]*wCsolid(nu[j], beta));
581 dos[DOS_S][j] = (dos[DOS_DIFF][j]*wSdiff +
582 dos[DOS_SOLID][j]*wSsolid(nu[j], beta));
583 dos[DOS_A][j] = (dos[DOS_DIFF][j]*wAdiff +
584 dos[DOS_SOLID][j]*wAsolid(nu[j], beta));
585 dos[DOS_E][j] = (dos[DOS_DIFF][j]*wEdiff +
586 dos[DOS_SOLID][j]*wEsolid(nu[j], beta));
588 DiffCoeff = evaluate_integral(nframes/2, tt, dos[VACF], nullptr, nframes/2., &stddev);
589 DiffCoeff = 1000*DiffCoeff/3.0;
590 fprintf(fplog, "Diffusion coefficient from VACF %g 10^-5 cm^2/s\n",
591 DiffCoeff);
592 fprintf(fplog, "Diffusion coefficient from DoS %g 10^-5 cm^2/s\n",
593 1000*DoS0/(12*tmass*beta));
595 cP = BOLTZ * evaluate_integral(nframes/4, nu, dos[DOS_CP], nullptr,
596 int{nframes/4}, &stddev);
597 fprintf(fplog, "Heat capacity %g J/mol K\n", 1000*cP/Nmol);
598 fprintf(fplog, "\nArrivederci!\n");
599 gmx_fio_fclose(fplog);
601 do_view(oenv, ftp2fn(efXVG, NFILE, fnm), "-nxy");
603 return 0;