LORENE
et_rot_lambda_grv2.C
1 /*
2  * Method Etoile_rot::lambda_grv2.
3  *
4  * (see file etoile.h for documentation)
5  *
6  */
7 
8 /*
9  * Copyright (c) 2000-2001 Eric Gourgoulhon
10  *
11  * This file is part of LORENE.
12  *
13  * LORENE is free software; you can redistribute it and/or modify
14  * it under the terms of the GNU General Public License as published by
15  * the Free Software Foundation; either version 2 of the License, or
16  * (at your option) any later version.
17  *
18  * LORENE is distributed in the hope that it will be useful,
19  * but WITHOUT ANY WARRANTY; without even the implied warranty of
20  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21  * GNU General Public License for more details.
22  *
23  * You should have received a copy of the GNU General Public License
24  * along with LORENE; if not, write to the Free Software
25  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
26  *
27  */
28 
29 
30 
31 
32 /*
33  * $Id: et_rot_lambda_grv2.C,v 1.8 2016/12/05 16:17:54 j_novak Exp $
34  * $Log: et_rot_lambda_grv2.C,v $
35  * Revision 1.8 2016/12/05 16:17:54 j_novak
36  * Suppression of some global variables (file names, loch, ...) to prevent redefinitions
37  *
38  * Revision 1.7 2014/10/13 08:52:58 j_novak
39  * Lorene classes and functions now belong to the namespace Lorene.
40  *
41  * Revision 1.6 2014/10/06 15:13:09 j_novak
42  * Modified #include directives to use c++ syntax.
43  *
44  * Revision 1.5 2013/06/05 15:10:42 j_novak
45  * Suppression of FINJAC sampling in r. This Jacobi(0,2) base is now
46  * available by setting colloc_r to BASE_JAC02 in the Mg3d constructor.
47  *
48  * Revision 1.4 2008/08/27 08:47:17 jl_cornou
49  * Added R_JACO02 case
50  *
51  * Revision 1.3 2003/10/27 10:53:16 e_gourgoulhon
52  * Changed variable name mp --> mprad in order not to shadow member mp.
53  *
54  * Revision 1.2 2002/09/09 13:00:39 e_gourgoulhon
55  * Modification of declaration of Fortran 77 prototypes for
56  * a better portability (in particular on IBM AIX systems):
57  * All Fortran subroutine names are now written F77_* and are
58  * defined in the new file C++/Include/proto_f77.h.
59  *
60  * Revision 1.1.1.1 2001/11/20 15:19:28 e_gourgoulhon
61  * LORENE
62  *
63  * Revision 2.1 2001/10/10 13:52:21 eric
64  * Modif Joachim: suppression caractere invisible en fin de fichier.
65  *
66  * Revision 2.0 2000/11/19 18:52:30 eric
67  * *** empty log message ***
68  *
69  *
70  * $Header: /cvsroot/Lorene/C++/Source/Etoile/et_rot_lambda_grv2.C,v 1.8 2016/12/05 16:17:54 j_novak Exp $
71  *
72  */
73 
74 // Headers C
75 #include <cmath>
76 
77 // Headers Lorene
78 #include "etoile.h"
79 #include "proto_f77.h"
80 
81 namespace Lorene {
82 double Etoile_rot::lambda_grv2(const Cmp& sou_m, const Cmp& sou_q) {
83 
84  const Map_radial* mprad = dynamic_cast<const Map_radial*>( sou_m.get_mp() ) ;
85 
86  if (mprad == 0x0) {
87  cout << "Etoile_rot::lambda_grv2: the mapping of sou_m does not"
88  << endl << " belong to the class Map_radial !" << endl ;
89  abort() ;
90  }
91 
92  assert( sou_q.get_mp() == mprad ) ;
93 
94  sou_q.check_dzpuis(4) ;
95 
96  const Mg3d* mg = mprad->get_mg() ;
97  int nz = mg->get_nzone() ;
98 
99  // Construction of a Map_af which coincides with *mp on the equator
100  // ----------------------------------------------------------------
101 
102  double theta0 = M_PI / 2 ; // Equator
103  double phi0 = 0 ;
104 
105  Map_af mpaff(*mprad) ;
106 
107  for (int l=0 ; l<nz ; l++) {
108  double rmax = mprad->val_r(l, double(1), theta0, phi0) ;
109  switch ( mg->get_type_r(l) ) {
110  case RARE: {
111  double rmin = mprad->val_r(l, double(0), theta0, phi0) ;
112  mpaff.set_alpha(rmax - rmin, l) ;
113  mpaff.set_beta(rmin, l) ;
114  break ;
115  }
116 
117  case FIN: {
118  double rmin = mprad->val_r(l, double(-1), theta0, phi0) ;
119  mpaff.set_alpha( double(.5) * (rmax - rmin), l ) ;
120  mpaff.set_beta( double(.5) * (rmax + rmin), l) ;
121  break ;
122  }
123 
124  case UNSURR: {
125  double rmin = mprad->val_r(l, double(-1), theta0, phi0) ;
126  double umax = double(1) / rmin ;
127  double umin = double(1) / rmax ;
128  mpaff.set_alpha( double(.5) * (umin - umax), l) ;
129  mpaff.set_beta( double(.5) * (umin + umax), l) ;
130  break ;
131  }
132 
133  default: {
134  cout << "Etoile_rot::lambda_grv2: unknown type_r ! " << endl ;
135  abort () ;
136  break ;
137  }
138 
139  }
140  }
141 
142 
143  // Reduced Jacobian of
144  // the transformation (r,theta,phi) <-> (dzeta,theta',phi')
145  // ------------------------------------------------------------
146 
147  Mtbl jac = 1 / ( (mprad->xsr) * (mprad->dxdr) ) ;
148  // R/x dR/dx in the nucleus
149  // R dR/dx in the shells
150  // - U/(x-1) dU/dx in the ZEC
151  for (int l=0; l<nz; l++) {
152  switch ( mg->get_type_r(l) ) {
153  case RARE: {
154  double a1 = mpaff.get_alpha()[l] ;
155  *(jac.t[l]) = *(jac.t[l]) / (a1*a1) ;
156  break ;
157  }
158 
159  case FIN: {
160  double a1 = mpaff.get_alpha()[l] ;
161  double b1 = mpaff.get_beta()[l] ;
162  assert( jac.t[l]->get_etat() == ETATQCQ ) ;
163  double* tjac = jac.t[l]->t ;
164  double* const xi = mg->get_grille3d(l)->x ;
165  for (int k=0; k<mg->get_np(l); k++) {
166  for (int j=0; j<mg->get_nt(l); j++) {
167  for (int i=0; i<mg->get_nr(l); i++) {
168  *tjac = *tjac /
169  (a1 * (a1 * xi[i] + b1) ) ;
170  tjac++ ;
171  }
172  }
173  }
174 
175  break ;
176  }
177 
178 
179  case UNSURR: {
180  double a1 = mpaff.get_alpha()[l] ;
181  *(jac.t[l]) = - *(jac.t[l]) / (a1*a1) ;
182  break ;
183  }
184 
185  default: {
186  cout << "Etoile_rot::lambda_grv2: unknown type_r ! " << endl ;
187  abort () ;
188  break ;
189  }
190 
191  }
192 
193  }
194 
195 
196  // Multiplication of the sources by the reduced Jacobian:
197  // -----------------------------------------------------
198 
199  Mtbl s_m(mg) ;
200  if ( sou_m.get_etat() == ETATZERO ) {
201  s_m = 0 ;
202  }
203  else{
204  assert(sou_m.va.get_etat() == ETATQCQ) ;
205  sou_m.va.coef_i() ;
206  s_m = *(sou_m.va.c) ;
207  }
208 
209  Mtbl s_q(mg) ;
210  if ( sou_q.get_etat() == ETATZERO ) {
211  s_q = 0 ;
212  }
213  else{
214  assert(sou_q.va.get_etat() == ETATQCQ) ;
215  sou_q.va.coef_i() ;
216  s_q = *(sou_q.va.c) ;
217  }
218 
219  s_m *= jac ;
220  s_q *= jac ;
221 
222 
223  // Preparations for the call to the Fortran subroutine
224  // ---------------------------------------------------
225 
226  int np1 = 1 ; // Axisymmetry enforced
227  int nt = mg->get_nt(0) ;
228  int nt2 = 2*nt - 1 ; // Number of points for the theta sampling
229  // in [0,Pi], instead of [0,Pi/2]
230 
231  // Array NDL
232  // ---------
233  int* ndl = new int[nz+4] ;
234  ndl[0] = nz ;
235  for (int l=0; l<nz; l++) {
236  ndl[1+l] = mg->get_nr(l) ;
237  }
238  ndl[1+nz] = nt2 ;
239  ndl[2+nz] = np1 ;
240  ndl[3+nz] = nz ;
241 
242  // Parameters NDR, NDT, NDP
243  // ------------------------
244  int nrmax = 0 ;
245  for (int l=0; l<nz ; l++) {
246  nrmax = ( ndl[1+l] > nrmax ) ? ndl[1+l] : nrmax ;
247  }
248  int ndr = nrmax + 5 ;
249  int ndt = nt2 + 2 ;
250  int ndp = np1 + 2 ;
251 
252  // Array ERRE
253  // ----------
254 
255  double* erre = new double [nz*ndr] ;
256 
257  for (int l=0; l<nz; l++) {
258  double a1 = mpaff.get_alpha()[l] ;
259  double b1 = mpaff.get_beta()[l] ;
260  for (int i=0; i<ndl[1+l]; i++) {
261  double xi = mg->get_grille3d(l)->x[i] ;
262  erre[ ndr*l + i ] = a1 * xi + b1 ;
263  }
264  }
265 
266  // Arrays containing the data
267  // --------------------------
268 
269  int ndrt = ndr*ndt ;
270  int ndrtp = ndr*ndt*ndp ;
271  int taille = ndrtp*nz ;
272 
273  double* tsou_m = new double[ taille ] ;
274  double* tsou_q = new double[ taille ] ;
275 
276  // Initialisation to zero :
277  for (int i=0; i<taille; i++) {
278  tsou_m[i] = 0 ;
279  tsou_q[i] = 0 ;
280  }
281 
282  // Copy of s_m into tsou_m
283  // -----------------------
284 
285  for (int l=0; l<nz; l++) {
286  for (int k=0; k<np1; k++) {
287  for (int j=0; j<nt; j++) {
288  for (int i=0; i<mg->get_nr(l); i++) {
289  double xx = s_m(l, k, j, i) ;
290  tsou_m[ndrtp*l + ndrt*k + ndr*j + i] = xx ;
291  // point symetrique par rapport au plan theta = pi/2 :
292  tsou_m[ndrtp*l + ndrt*k + ndr*(nt2-1-j) + i] = xx ;
293  }
294  }
295  }
296  }
297 
298  // Copy of s_q into tsou_q
299  // -----------------------
300 
301  for (int l=0; l<nz; l++) {
302  for (int k=0; k<np1; k++) {
303  for (int j=0; j<nt; j++) {
304  for (int i=0; i<mg->get_nr(l); i++) {
305  double xx = s_q(l, k, j, i) ;
306  tsou_q[ndrtp*l + ndrt*k + ndr*j + i] = xx ;
307  // point symetrique par rapport au plan theta = pi/2 :
308  tsou_q[ndrtp*l + ndrt*k + ndr*(nt2-1-j) + i] = xx ;
309  }
310  }
311  }
312  }
313 
314 
315  // Computation of the integrals
316  // ----------------------------
317 
318  double int_m, int_q ;
319  F77_integrale2d(ndl, &ndr, &ndt, &ndp, erre, tsou_m, &int_m) ;
320  F77_integrale2d(ndl, &ndr, &ndt, &ndp, erre, tsou_q, &int_q) ;
321 
322  // Cleaning
323  // --------
324 
325  delete [] ndl ;
326  delete [] erre ;
327  delete [] tsou_m ;
328  delete [] tsou_q ;
329 
330  // Computation of lambda
331  // ---------------------
332 
333  double lambda ;
334  if ( int_q != double(0) ) {
335  lambda = - int_m / int_q ;
336  }
337  else{
338  lambda = 0 ;
339  }
340 
341  return lambda ;
342 
343 }
344 }
const Map * get_mp() const
Returns the mapping.
Definition: cmp.h:901
const Grille3d * get_grille3d(int l) const
Returns a pointer on the 3D mono-grid for domain no. l.
Definition: grilles.h:517
Component of a tensorial field *** DEPRECATED : use class Scalar instead ***.
Definition: cmp.h:446
const double * get_alpha() const
Returns the pointer on the array alpha.
Definition: map_af.C:604
int get_np(int l) const
Returns the number of points in the azimuthal direction ( ) in domain no. l.
Definition: grilles.h:479
Multi-domain array.
Definition: mtbl.h:118
Lorene prototypes.
Definition: app_hor.h:67
const Mg3d * get_mg() const
Gives the Mg3d on which the mapping is defined.
Definition: map.h:777
int get_etat() const
Returns the logical state.
Definition: cmp.h:899
void coef_i() const
Computes the physical value of *this.
int get_etat() const
Gives the logical state.
Definition: tbl.h:414
double * x
Array of values of at the nr collocation points.
Definition: grilles.h:215
void set_beta(double beta0, int l)
Modifies the value of in domain no. l.
Definition: map_af.C:768
int get_etat() const
Returns the logical state.
Definition: valeur.h:760
virtual double val_r(int l, double xi, double theta, double pphi) const =0
Returns the value of the radial coordinate r for a given in a given domain.
Coord dxdr
in the nucleus and in the non-compactified shells; \ in the compactified outer domain.
Definition: map.h:1575
double * t
The array of double.
Definition: tbl.h:176
const double * get_beta() const
Returns the pointer on the array beta.
Definition: map_af.C:608
Mtbl * c
Values of the function at the points of the multi-grid.
Definition: valeur.h:309
Base class for pure radial mappings.
Definition: map.h:1551
int get_nzone() const
Returns the number of domains.
Definition: grilles.h:465
static double lambda_grv2(const Cmp &sou_m, const Cmp &sou_q)
Computes the coefficient which ensures that the GRV2 virial identity is satisfied.
void set_alpha(double alpha0, int l)
Modifies the value of in domain no. l.
Definition: map_af.C:757
Coord xsr
in the nucleus; \ 1/R in the non-compactified shells; \ in the compactified outer domain...
Definition: map.h:1564
int get_nr(int l) const
Returns the number of points in the radial direction ( ) in domain no. l.
Definition: grilles.h:469
Multi-domain grid.
Definition: grilles.h:279
Affine radial mapping.
Definition: map.h:2042
bool check_dzpuis(int dzi) const
Returns false if the last domain is compactified and *this is not zero in this domain and dzpuis is n...
Definition: cmp.C:718
int get_nt(int l) const
Returns the number of points in the co-latitude direction ( ) in domain no. l.
Definition: grilles.h:474
int get_type_r(int l) const
Returns the type of sampling in the radial direction in domain no.
Definition: grilles.h:491
Tbl ** t
Array (size nzone ) of pointers on the Tbl &#39;s.
Definition: mtbl.h:132
Valeur va
The numerical value of the Cmp.
Definition: cmp.h:464