1 | // Special functions -*- C++ -*-  |
2 |   |
3 | // Copyright (C) 2006-2019 Free Software Foundation, Inc.  |
4 | //  |
5 | // This file is part of the GNU ISO C++ Library. This library is free  |
6 | // software; you can redistribute it and/or modify it under the  |
7 | // terms of the GNU General Public License as published by the  |
8 | // Free Software Foundation; either version 3, or (at your option)  |
9 | // any later version.  |
10 | //  |
11 | // This library is distributed in the hope that it will be useful,  |
12 | // but WITHOUT ANY WARRANTY; without even the implied warranty of  |
13 | // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the  |
14 | // GNU General Public License for more details.  |
15 | //  |
16 | // Under Section 7 of GPL version 3, you are granted additional  |
17 | // permissions described in the GCC Runtime Library Exception, version  |
18 | // 3.1, as published by the Free Software Foundation.  |
19 |   |
20 | // You should have received a copy of the GNU General Public License and  |
21 | // a copy of the GCC Runtime Library Exception along with this program;  |
22 | // see the files COPYING3 and COPYING.RUNTIME respectively. If not, see  |
23 | // <http://www.gnu.org/licenses/>.  |
24 |   |
25 | /** @file tr1/bessel_function.tcc  |
26 | * This is an internal header file, included by other library headers.  |
27 | * Do not attempt to use it directly. @headername{tr1/cmath}  |
28 | */  |
29 |   |
30 | /* __cyl_bessel_jn_asymp adapted from GNU GSL version 2.4 specfunc/bessel_j.c  |
31 | * Copyright (C) 1996-2003 Gerard Jungman  |
32 | */  |
33 |   |
34 | //  |
35 | // ISO C++ 14882 TR1: 5.2 Special functions  |
36 | //  |
37 |   |
38 | // Written by Edward Smith-Rowland.  |
39 | //  |
40 | // References:  |
41 | // (1) Handbook of Mathematical Functions,  |
42 | // ed. Milton Abramowitz and Irene A. Stegun,  |
43 | // Dover Publications,  |
44 | // Section 9, pp. 355-434, Section 10 pp. 435-478  |
45 | // (2) The Gnu Scientific Library, http://www.gnu.org/software/gsl  |
46 | // (3) Numerical Recipes in C, by W. H. Press, S. A. Teukolsky,  |
47 | // W. T. Vetterling, B. P. Flannery, Cambridge University Press (1992),  |
48 | // 2nd ed, pp. 240-245  |
49 |   |
50 | #ifndef _GLIBCXX_TR1_BESSEL_FUNCTION_TCC  |
51 | #define _GLIBCXX_TR1_BESSEL_FUNCTION_TCC 1  |
52 |   |
53 | #include <tr1/special_function_util.h>  |
54 |   |
55 | namespace std _GLIBCXX_VISIBILITY(default)  |
56 | {  |
57 | _GLIBCXX_BEGIN_NAMESPACE_VERSION  |
58 |   |
59 | #if _GLIBCXX_USE_STD_SPEC_FUNCS  |
60 | # define _GLIBCXX_MATH_NS ::std  |
61 | #elif defined(_GLIBCXX_TR1_CMATH)  |
62 | namespace tr1  |
63 | {  |
64 | # define _GLIBCXX_MATH_NS ::std::tr1  |
65 | #else  |
66 | # error do not include this header directly, use <cmath> or <tr1/cmath>  |
67 | #endif  |
68 | // [5.2] Special functions  |
69 |   |
70 | // Implementation-space details.  |
71 | namespace __detail  |
72 | {  |
73 | /**  |
74 | * @brief Compute the gamma functions required by the Temme series  |
75 | * expansions of @f$ N_\nu(x) @f$ and @f$ K_\nu(x) @f$.  |
76 | * @f[  |
77 | * \Gamma_1 = \frac{1}{2\mu}  |
78 | * [\frac{1}{\Gamma(1 - \mu)} - \frac{1}{\Gamma(1 + \mu)}]  |
79 | * @f]  |
80 | * and  |
81 | * @f[  |
82 | * \Gamma_2 = \frac{1}{2}  |
83 | * [\frac{1}{\Gamma(1 - \mu)} + \frac{1}{\Gamma(1 + \mu)}]  |
84 | * @f]  |
85 | * where @f$ -1/2 <= \mu <= 1/2 @f$ is @f$ \mu = \nu - N @f$ and @f$ N @f$.  |
86 | * is the nearest integer to @f$ \nu @f$.  |
87 | * The values of \f$ \Gamma(1 + \mu) \f$ and \f$ \Gamma(1 - \mu) \f$  |
88 | * are returned as well.  |
89 | *   |
90 | * The accuracy requirements on this are exquisite.  |
91 | *  |
92 | * @param __mu The input parameter of the gamma functions.  |
93 | * @param __gam1 The output function \f$ \Gamma_1(\mu) \f$  |
94 | * @param __gam2 The output function \f$ \Gamma_2(\mu) \f$  |
95 | * @param __gampl The output function \f$ \Gamma(1 + \mu) \f$  |
96 | * @param __gammi The output function \f$ \Gamma(1 - \mu) \f$  |
97 | */  |
98 | template <typename _Tp>  |
99 | void  |
100 | __gamma_temme(_Tp __mu,  |
101 | _Tp & __gam1, _Tp & __gam2, _Tp & __gampl, _Tp & __gammi)  |
102 | {  |
103 | #if _GLIBCXX_USE_C99_MATH_TR1  |
104 | __gampl = _Tp(1) / _GLIBCXX_MATH_NS::tgamma(_Tp(1) + __mu);  |
105 | __gammi = _Tp(1) / _GLIBCXX_MATH_NS::tgamma(_Tp(1) - __mu);  |
106 | #else  |
107 | __gampl = _Tp(1) / __gamma(_Tp(1) + __mu);  |
108 | __gammi = _Tp(1) / __gamma(_Tp(1) - __mu);  |
109 | #endif  |
110 |   |
111 | if (std::abs(__mu) < std::numeric_limits<_Tp>::epsilon())  |
112 | __gam1 = -_Tp(__numeric_constants<_Tp>::__gamma_e());  |
113 | else  |
114 | __gam1 = (__gammi - __gampl) / (_Tp(2) * __mu);  |
115 |   |
116 | __gam2 = (__gammi + __gampl) / (_Tp(2));  |
117 |   |
118 | return;  |
119 | }  |
120 |   |
121 |   |
122 | /**  |
123 | * @brief Compute the Bessel @f$ J_\nu(x) @f$ and Neumann  |
124 | * @f$ N_\nu(x) @f$ functions and their first derivatives  |
125 | * @f$ J'_\nu(x) @f$ and @f$ N'_\nu(x) @f$ respectively.  |
126 | * These four functions are computed together for numerical  |
127 | * stability.  |
128 | *  |
129 | * @param __nu The order of the Bessel functions.  |
130 | * @param __x The argument of the Bessel functions.  |
131 | * @param __Jnu The output Bessel function of the first kind.  |
132 | * @param __Nnu The output Neumann function (Bessel function of the second kind).  |
133 | * @param __Jpnu The output derivative of the Bessel function of the first kind.  |
134 | * @param __Npnu The output derivative of the Neumann function.  |
135 | */  |
136 | template <typename _Tp>  |
137 | void  |
138 | __bessel_jn(_Tp __nu, _Tp __x,  |
139 | _Tp & __Jnu, _Tp & __Nnu, _Tp & __Jpnu, _Tp & __Npnu)  |
140 | {  |
141 | if (__x == _Tp(0))  |
142 | {  |
143 | if (__nu == _Tp(0))  |
144 | {  |
145 | __Jnu = _Tp(1);  |
146 | __Jpnu = _Tp(0);  |
147 | }  |
148 | else if (__nu == _Tp(1))  |
149 | {  |
150 | __Jnu = _Tp(0);  |
151 | __Jpnu = _Tp(0.5L);  |
152 | }  |
153 | else  |
154 | {  |
155 | __Jnu = _Tp(0);  |
156 | __Jpnu = _Tp(0);  |
157 | }  |
158 | __Nnu = -std::numeric_limits<_Tp>::infinity();  |
159 | __Npnu = std::numeric_limits<_Tp>::infinity();  |
160 | return;  |
161 | }  |
162 |   |
163 | const _Tp __eps = std::numeric_limits<_Tp>::epsilon();  |
164 | // When the multiplier is N i.e.  |
165 | // fp_min = N * min()  |
166 | // Then J_0 and N_0 tank at x = 8 * N (J_0 = 0 and N_0 = nan)!  |
167 | //const _Tp __fp_min = _Tp(20) * std::numeric_limits<_Tp>::min();  |
168 | const _Tp __fp_min = std::sqrt(std::numeric_limits<_Tp>::min());  |
169 | const int __max_iter = 15000;  |
170 | const _Tp __x_min = _Tp(2);  |
171 |   |
172 | const int __nl = (__x < __x_min  |
173 | ? static_cast<int>(__nu + _Tp(0.5L))  |
174 | : std::max(0, static_cast<int>(__nu - __x + _Tp(1.5L))));  |
175 |   |
176 | const _Tp __mu = __nu - __nl;  |
177 | const _Tp __mu2 = __mu * __mu;  |
178 | const _Tp __xi = _Tp(1) / __x;  |
179 | const _Tp __xi2 = _Tp(2) * __xi;  |
180 | _Tp __w = __xi2 / __numeric_constants<_Tp>::__pi();  |
181 | int __isign = 1;  |
182 | _Tp __h = __nu * __xi;  |
183 | if (__h < __fp_min)  |
184 | __h = __fp_min;  |
185 | _Tp __b = __xi2 * __nu;  |
186 | _Tp __d = _Tp(0);  |
187 | _Tp __c = __h;  |
188 | int __i;  |
189 | for (__i = 1; __i <= __max_iter; ++__i)  |
190 | {  |
191 | __b += __xi2;  |
192 | __d = __b - __d;  |
193 | if (std::abs(__d) < __fp_min)  |
194 | __d = __fp_min;  |
195 | __c = __b - _Tp(1) / __c;  |
196 | if (std::abs(__c) < __fp_min)  |
197 | __c = __fp_min;  |
198 | __d = _Tp(1) / __d;  |
199 | const _Tp __del = __c * __d;  |
200 | __h *= __del;  |
201 | if (__d < _Tp(0))  |
202 | __isign = -__isign;  |
203 | if (std::abs(__del - _Tp(1)) < __eps)  |
204 | break;  |
205 | }  |
206 | if (__i > __max_iter)  |
207 | std::__throw_runtime_error(__N("Argument x too large in __bessel_jn; "   |
208 | "try asymptotic expansion." ));  |
209 | _Tp __Jnul = __isign * __fp_min;  |
210 | _Tp __Jpnul = __h * __Jnul;  |
211 | _Tp __Jnul1 = __Jnul;  |
212 | _Tp __Jpnu1 = __Jpnul;  |
213 | _Tp __fact = __nu * __xi;  |
214 | for ( int __l = __nl; __l >= 1; --__l )  |
215 | {  |
216 | const _Tp __Jnutemp = __fact * __Jnul + __Jpnul;  |
217 | __fact -= __xi;  |
218 | __Jpnul = __fact * __Jnutemp - __Jnul;  |
219 | __Jnul = __Jnutemp;  |
220 | }  |
221 | if (__Jnul == _Tp(0))  |
222 | __Jnul = __eps;  |
223 | _Tp __f= __Jpnul / __Jnul;  |
224 | _Tp __Nmu, __Nnu1, __Npmu, __Jmu;  |
225 | if (__x < __x_min)  |
226 | {  |
227 | const _Tp __x2 = __x / _Tp(2);  |
228 | const _Tp __pimu = __numeric_constants<_Tp>::__pi() * __mu;  |
229 | _Tp __fact = (std::abs(__pimu) < __eps  |
230 | ? _Tp(1) : __pimu / std::sin(__pimu));  |
231 | _Tp __d = -std::log(__x2);  |
232 | _Tp __e = __mu * __d;  |
233 | _Tp __fact2 = (std::abs(__e) < __eps  |
234 | ? _Tp(1) : std::sinh(__e) / __e);  |
235 | _Tp __gam1, __gam2, __gampl, __gammi;  |
236 | __gamma_temme(__mu, __gam1, __gam2, __gampl, __gammi);  |
237 | _Tp __ff = (_Tp(2) / __numeric_constants<_Tp>::__pi())  |
238 | * __fact * (__gam1 * std::cosh(__e) + __gam2 * __fact2 * __d);  |
239 | __e = std::exp(__e);  |
240 | _Tp __p = __e / (__numeric_constants<_Tp>::__pi() * __gampl);  |
241 | _Tp __q = _Tp(1) / (__e * __numeric_constants<_Tp>::__pi() * __gammi);  |
242 | const _Tp __pimu2 = __pimu / _Tp(2);  |
243 | _Tp __fact3 = (std::abs(__pimu2) < __eps  |
244 | ? _Tp(1) : std::sin(__pimu2) / __pimu2 );  |
245 | _Tp __r = __numeric_constants<_Tp>::__pi() * __pimu2 * __fact3 * __fact3;  |
246 | _Tp __c = _Tp(1);  |
247 | __d = -__x2 * __x2;  |
248 | _Tp __sum = __ff + __r * __q;  |
249 | _Tp __sum1 = __p;  |
250 | for (__i = 1; __i <= __max_iter; ++__i)  |
251 | {  |
252 | __ff = (__i * __ff + __p + __q) / (__i * __i - __mu2);  |
253 | __c *= __d / _Tp(__i);  |
254 | __p /= _Tp(__i) - __mu;  |
255 | __q /= _Tp(__i) + __mu;  |
256 | const _Tp __del = __c * (__ff + __r * __q);  |
257 | __sum += __del;   |
258 | const _Tp __del1 = __c * __p - __i * __del;  |
259 | __sum1 += __del1;  |
260 | if ( std::abs(__del) < __eps * (_Tp(1) + std::abs(__sum)) )  |
261 | break;  |
262 | }  |
263 | if ( __i > __max_iter )  |
264 | std::__throw_runtime_error(__N("Bessel y series failed to converge "   |
265 | "in __bessel_jn." ));  |
266 | __Nmu = -__sum;  |
267 | __Nnu1 = -__sum1 * __xi2;  |
268 | __Npmu = __mu * __xi * __Nmu - __Nnu1;  |
269 | __Jmu = __w / (__Npmu - __f * __Nmu);  |
270 | }  |
271 | else  |
272 | {  |
273 | _Tp __a = _Tp(0.25L) - __mu2;  |
274 | _Tp __q = _Tp(1);  |
275 | _Tp __p = -__xi / _Tp(2);  |
276 | _Tp __br = _Tp(2) * __x;  |
277 | _Tp __bi = _Tp(2);  |
278 | _Tp __fact = __a * __xi / (__p * __p + __q * __q);  |
279 | _Tp __cr = __br + __q * __fact;  |
280 | _Tp __ci = __bi + __p * __fact;  |
281 | _Tp __den = __br * __br + __bi * __bi;  |
282 | _Tp __dr = __br / __den;  |
283 | _Tp __di = -__bi / __den;  |
284 | _Tp __dlr = __cr * __dr - __ci * __di;  |
285 | _Tp __dli = __cr * __di + __ci * __dr;  |
286 | _Tp __temp = __p * __dlr - __q * __dli;  |
287 | __q = __p * __dli + __q * __dlr;  |
288 | __p = __temp;  |
289 | int __i;  |
290 | for (__i = 2; __i <= __max_iter; ++__i)  |
291 | {  |
292 | __a += _Tp(2 * (__i - 1));  |
293 | __bi += _Tp(2);  |
294 | __dr = __a * __dr + __br;  |
295 | __di = __a * __di + __bi;  |
296 | if (std::abs(__dr) + std::abs(__di) < __fp_min)  |
297 | __dr = __fp_min;  |
298 | __fact = __a / (__cr * __cr + __ci * __ci);  |
299 | __cr = __br + __cr * __fact;  |
300 | __ci = __bi - __ci * __fact;  |
301 | if (std::abs(__cr) + std::abs(__ci) < __fp_min)  |
302 | __cr = __fp_min;  |
303 | __den = __dr * __dr + __di * __di;  |
304 | __dr /= __den;  |
305 | __di /= -__den;  |
306 | __dlr = __cr * __dr - __ci * __di;  |
307 | __dli = __cr * __di + __ci * __dr;  |
308 | __temp = __p * __dlr - __q * __dli;  |
309 | __q = __p * __dli + __q * __dlr;  |
310 | __p = __temp;  |
311 | if (std::abs(__dlr - _Tp(1)) + std::abs(__dli) < __eps)  |
312 | break;  |
313 | }  |
314 | if (__i > __max_iter)  |
315 | std::__throw_runtime_error(__N("Lentz's method failed "   |
316 | "in __bessel_jn." ));  |
317 | const _Tp __gam = (__p - __f) / __q;  |
318 | __Jmu = std::sqrt(__w / ((__p - __f) * __gam + __q));  |
319 | #if _GLIBCXX_USE_C99_MATH_TR1  |
320 | __Jmu = _GLIBCXX_MATH_NS::copysign(__Jmu, __Jnul);  |
321 | #else  |
322 | if (__Jmu * __Jnul < _Tp(0))  |
323 | __Jmu = -__Jmu;  |
324 | #endif  |
325 | __Nmu = __gam * __Jmu;  |
326 | __Npmu = (__p + __q / __gam) * __Nmu;  |
327 | __Nnu1 = __mu * __xi * __Nmu - __Npmu;  |
328 | }  |
329 | __fact = __Jmu / __Jnul;  |
330 | __Jnu = __fact * __Jnul1;  |
331 | __Jpnu = __fact * __Jpnu1;  |
332 | for (__i = 1; __i <= __nl; ++__i)  |
333 | {  |
334 | const _Tp __Nnutemp = (__mu + __i) * __xi2 * __Nnu1 - __Nmu;  |
335 | __Nmu = __Nnu1;  |
336 | __Nnu1 = __Nnutemp;  |
337 | }  |
338 | __Nnu = __Nmu;  |
339 | __Npnu = __nu * __xi * __Nmu - __Nnu1;  |
340 |   |
341 | return;  |
342 | }  |
343 |   |
344 |   |
345 | /**  |
346 | * @brief This routine computes the asymptotic cylindrical Bessel  |
347 | * and Neumann functions of order nu: \f$ J_{\nu} \f$,  |
348 | * \f$ N_{\nu} \f$.  |
349 | *  |
350 | * References:  |
351 | * (1) Handbook of Mathematical Functions,  |
352 | * ed. Milton Abramowitz and Irene A. Stegun,  |
353 | * Dover Publications,  |
354 | * Section 9 p. 364, Equations 9.2.5-9.2.10  |
355 | *  |
356 | * @param __nu The order of the Bessel functions.  |
357 | * @param __x The argument of the Bessel functions.  |
358 | * @param __Jnu The output Bessel function of the first kind.  |
359 | * @param __Nnu The output Neumann function (Bessel function of the second kind).  |
360 | */  |
361 | template <typename _Tp>  |
362 | void  |
363 | __cyl_bessel_jn_asymp(_Tp __nu, _Tp __x, _Tp & __Jnu, _Tp & __Nnu)  |
364 | {  |
365 | const _Tp __mu = _Tp(4) * __nu * __nu;  |
366 | const _Tp __8x = _Tp(8) * __x;  |
367 |   |
368 | _Tp __P = _Tp(0);  |
369 | _Tp __Q = _Tp(0);  |
370 |   |
371 | _Tp __k = _Tp(0);  |
372 | _Tp __term = _Tp(1);  |
373 |   |
374 | int __epsP = 0;  |
375 | int __epsQ = 0;  |
376 |   |
377 | _Tp __eps = std::numeric_limits<_Tp>::epsilon();  |
378 |   |
379 | do  |
380 | {  |
381 | __term *= (__k == 0  |
382 | ? _Tp(1)  |
383 | : -(__mu - (2 * __k - 1) * (2 * __k - 1)) / (__k * __8x));  |
384 |   |
385 | __epsP = std::abs(__term) < __eps * std::abs(__P);  |
386 | __P += __term;  |
387 |   |
388 | __k++;  |
389 |   |
390 | __term *= (__mu - (2 * __k - 1) * (2 * __k - 1)) / (__k * __8x);  |
391 | __epsQ = std::abs(__term) < __eps * std::abs(__Q);  |
392 | __Q += __term;  |
393 |   |
394 | if (__epsP && __epsQ && __k > (__nu / 2.))  |
395 | break;  |
396 |   |
397 | __k++;  |
398 | }  |
399 | while (__k < 1000);  |
400 |   |
401 | const _Tp __chi = __x - (__nu + _Tp(0.5L))  |
402 | * __numeric_constants<_Tp>::__pi_2();  |
403 |   |
404 | const _Tp __c = std::cos(__chi);  |
405 | const _Tp __s = std::sin(__chi);  |
406 |   |
407 | const _Tp __coef = std::sqrt(_Tp(2)  |
408 | / (__numeric_constants<_Tp>::__pi() * __x));  |
409 |   |
410 | __Jnu = __coef * (__c * __P - __s * __Q);  |
411 | __Nnu = __coef * (__s * __P + __c * __Q);  |
412 |   |
413 | return;  |
414 | }  |
415 |   |
416 |   |
417 | /**  |
418 | * @brief This routine returns the cylindrical Bessel functions  |
419 | * of order \f$ \nu \f$: \f$ J_{\nu} \f$ or \f$ I_{\nu} \f$  |
420 | * by series expansion.  |
421 | *  |
422 | * The modified cylindrical Bessel function is:  |
423 | * @f[  |
424 | * Z_{\nu}(x) = \sum_{k=0}^{\infty}  |
425 | * \frac{\sigma^k (x/2)^{\nu + 2k}}{k!\Gamma(\nu+k+1)}  |
426 | * @f]  |
427 | * where \f$ \sigma = +1 \f$ or\f$ -1 \f$ for  |
428 | * \f$ Z = I \f$ or \f$ J \f$ respectively.  |
429 | *   |
430 | * See Abramowitz & Stegun, 9.1.10  |
431 | * Abramowitz & Stegun, 9.6.7  |
432 | * (1) Handbook of Mathematical Functions,  |
433 | * ed. Milton Abramowitz and Irene A. Stegun,  |
434 | * Dover Publications,  |
435 | * Equation 9.1.10 p. 360 and Equation 9.6.10 p. 375  |
436 | *  |
437 | * @param __nu The order of the Bessel function.  |
438 | * @param __x The argument of the Bessel function.  |
439 | * @param __sgn The sign of the alternate terms  |
440 | * -1 for the Bessel function of the first kind.  |
441 | * +1 for the modified Bessel function of the first kind.  |
442 | * @return The output Bessel function.  |
443 | */  |
444 | template <typename _Tp>  |
445 | _Tp  |
446 | __cyl_bessel_ij_series(_Tp __nu, _Tp __x, _Tp __sgn,  |
447 | unsigned int __max_iter)  |
448 | {  |
449 | if (__x == _Tp(0))  |
450 | return __nu == _Tp(0) ? _Tp(1) : _Tp(0);  |
451 |   |
452 | const _Tp __x2 = __x / _Tp(2);  |
453 | _Tp __fact = __nu * std::log(__x2);  |
454 | #if _GLIBCXX_USE_C99_MATH_TR1  |
455 | __fact -= _GLIBCXX_MATH_NS::lgamma(__nu + _Tp(1));  |
456 | #else  |
457 | __fact -= __log_gamma(__nu + _Tp(1));  |
458 | #endif  |
459 | __fact = std::exp(__fact);  |
460 | const _Tp __xx4 = __sgn * __x2 * __x2;  |
461 | _Tp __Jn = _Tp(1);  |
462 | _Tp __term = _Tp(1);  |
463 |   |
464 | for (unsigned int __i = 1; __i < __max_iter; ++__i)  |
465 | {  |
466 | __term *= __xx4 / (_Tp(__i) * (__nu + _Tp(__i)));  |
467 | __Jn += __term;  |
468 | if (std::abs(__term / __Jn) < std::numeric_limits<_Tp>::epsilon())  |
469 | break;  |
470 | }  |
471 |   |
472 | return __fact * __Jn;  |
473 | }  |
474 |   |
475 |   |
476 | /**  |
477 | * @brief Return the Bessel function of order \f$ \nu \f$:  |
478 | * \f$ J_{\nu}(x) \f$.  |
479 | *  |
480 | * The cylindrical Bessel function is:  |
481 | * @f[  |
482 | * J_{\nu}(x) = \sum_{k=0}^{\infty}  |
483 | * \frac{(-1)^k (x/2)^{\nu + 2k}}{k!\Gamma(\nu+k+1)}  |
484 | * @f]  |
485 | *  |
486 | * @param __nu The order of the Bessel function.  |
487 | * @param __x The argument of the Bessel function.  |
488 | * @return The output Bessel function.  |
489 | */  |
490 | template<typename _Tp>  |
491 | _Tp  |
492 | __cyl_bessel_j(_Tp __nu, _Tp __x)  |
493 | {  |
494 | if (__nu < _Tp(0) || __x < _Tp(0))  |
495 | std::__throw_domain_error(__N("Bad argument "   |
496 | "in __cyl_bessel_j." ));  |
497 | else if (__isnan(__nu) || __isnan(__x))  |
498 | return std::numeric_limits<_Tp>::quiet_NaN();  |
499 | else if (__x * __x < _Tp(10) * (__nu + _Tp(1)))  |
500 | return __cyl_bessel_ij_series(__nu, __x, -_Tp(1), 200);  |
501 | else if (__x > _Tp(1000))  |
502 | {  |
503 | _Tp __J_nu, __N_nu;  |
504 | __cyl_bessel_jn_asymp(__nu, __x, __J_nu, __N_nu);  |
505 | return __J_nu;  |
506 | }  |
507 | else  |
508 | {  |
509 | _Tp __J_nu, __N_nu, __Jp_nu, __Np_nu;  |
510 | __bessel_jn(__nu, __x, __J_nu, __N_nu, __Jp_nu, __Np_nu);  |
511 | return __J_nu;  |
512 | }  |
513 | }  |
514 |   |
515 |   |
516 | /**  |
517 | * @brief Return the Neumann function of order \f$ \nu \f$:  |
518 | * \f$ N_{\nu}(x) \f$.  |
519 | *  |
520 | * The Neumann function is defined by:  |
521 | * @f[  |
522 | * N_{\nu}(x) = \frac{J_{\nu}(x) \cos \nu\pi - J_{-\nu}(x)}  |
523 | * {\sin \nu\pi}  |
524 | * @f]  |
525 | * where for integral \f$ \nu = n \f$ a limit is taken:  |
526 | * \f$ lim_{\nu \to n} \f$.  |
527 | *  |
528 | * @param __nu The order of the Neumann function.  |
529 | * @param __x The argument of the Neumann function.  |
530 | * @return The output Neumann function.  |
531 | */  |
532 | template<typename _Tp>  |
533 | _Tp  |
534 | __cyl_neumann_n(_Tp __nu, _Tp __x)  |
535 | {  |
536 | if (__nu < _Tp(0) || __x < _Tp(0))  |
537 | std::__throw_domain_error(__N("Bad argument "   |
538 | "in __cyl_neumann_n." ));  |
539 | else if (__isnan(__nu) || __isnan(__x))  |
540 | return std::numeric_limits<_Tp>::quiet_NaN();  |
541 | else if (__x > _Tp(1000))  |
542 | {  |
543 | _Tp __J_nu, __N_nu;  |
544 | __cyl_bessel_jn_asymp(__nu, __x, __J_nu, __N_nu);  |
545 | return __N_nu;  |
546 | }  |
547 | else  |
548 | {  |
549 | _Tp __J_nu, __N_nu, __Jp_nu, __Np_nu;  |
550 | __bessel_jn(__nu, __x, __J_nu, __N_nu, __Jp_nu, __Np_nu);  |
551 | return __N_nu;  |
552 | }  |
553 | }  |
554 |   |
555 |   |
556 | /**  |
557 | * @brief Compute the spherical Bessel @f$ j_n(x) @f$  |
558 | * and Neumann @f$ n_n(x) @f$ functions and their first  |
559 | * derivatives @f$ j'_n(x) @f$ and @f$ n'_n(x) @f$  |
560 | * respectively.  |
561 | *  |
562 | * @param __n The order of the spherical Bessel function.  |
563 | * @param __x The argument of the spherical Bessel function.  |
564 | * @param __j_n The output spherical Bessel function.  |
565 | * @param __n_n The output spherical Neumann function.  |
566 | * @param __jp_n The output derivative of the spherical Bessel function.  |
567 | * @param __np_n The output derivative of the spherical Neumann function.  |
568 | */  |
569 | template <typename _Tp>  |
570 | void  |
571 | __sph_bessel_jn(unsigned int __n, _Tp __x,  |
572 | _Tp & __j_n, _Tp & __n_n, _Tp & __jp_n, _Tp & __np_n)  |
573 | {  |
574 | const _Tp __nu = _Tp(__n) + _Tp(0.5L);  |
575 |   |
576 | _Tp __J_nu, __N_nu, __Jp_nu, __Np_nu;  |
577 | __bessel_jn(__nu, __x, __J_nu, __N_nu, __Jp_nu, __Np_nu);  |
578 |   |
579 | const _Tp __factor = __numeric_constants<_Tp>::__sqrtpio2()  |
580 | / std::sqrt(__x);  |
581 |   |
582 | __j_n = __factor * __J_nu;  |
583 | __n_n = __factor * __N_nu;  |
584 | __jp_n = __factor * __Jp_nu - __j_n / (_Tp(2) * __x);  |
585 | __np_n = __factor * __Np_nu - __n_n / (_Tp(2) * __x);  |
586 |   |
587 | return;  |
588 | }  |
589 |   |
590 |   |
591 | /**  |
592 | * @brief Return the spherical Bessel function  |
593 | * @f$ j_n(x) @f$ of order n.  |
594 | *  |
595 | * The spherical Bessel function is defined by:  |
596 | * @f[  |
597 | * j_n(x) = \left( \frac{\pi}{2x} \right) ^{1/2} J_{n+1/2}(x)  |
598 | * @f]  |
599 | *  |
600 | * @param __n The order of the spherical Bessel function.  |
601 | * @param __x The argument of the spherical Bessel function.  |
602 | * @return The output spherical Bessel function.  |
603 | */  |
604 | template <typename _Tp>  |
605 | _Tp  |
606 | __sph_bessel(unsigned int __n, _Tp __x)  |
607 | {  |
608 | if (__x < _Tp(0))  |
609 | std::__throw_domain_error(__N("Bad argument "   |
610 | "in __sph_bessel." ));  |
611 | else if (__isnan(__x))  |
612 | return std::numeric_limits<_Tp>::quiet_NaN();  |
613 | else if (__x == _Tp(0))  |
614 | {  |
615 | if (__n == 0)  |
616 | return _Tp(1);  |
617 | else  |
618 | return _Tp(0);  |
619 | }  |
620 | else  |
621 | {  |
622 | _Tp __j_n, __n_n, __jp_n, __np_n;  |
623 | __sph_bessel_jn(__n, __x, __j_n, __n_n, __jp_n, __np_n);  |
624 | return __j_n;  |
625 | }  |
626 | }  |
627 |   |
628 |   |
629 | /**  |
630 | * @brief Return the spherical Neumann function  |
631 | * @f$ n_n(x) @f$.  |
632 | *  |
633 | * The spherical Neumann function is defined by:  |
634 | * @f[  |
635 | * n_n(x) = \left( \frac{\pi}{2x} \right) ^{1/2} N_{n+1/2}(x)  |
636 | * @f]  |
637 | *  |
638 | * @param __n The order of the spherical Neumann function.  |
639 | * @param __x The argument of the spherical Neumann function.  |
640 | * @return The output spherical Neumann function.  |
641 | */  |
642 | template <typename _Tp>  |
643 | _Tp  |
644 | __sph_neumann(unsigned int __n, _Tp __x)  |
645 | {  |
646 | if (__x < _Tp(0))  |
647 | std::__throw_domain_error(__N("Bad argument "   |
648 | "in __sph_neumann." ));  |
649 | else if (__isnan(__x))  |
650 | return std::numeric_limits<_Tp>::quiet_NaN();  |
651 | else if (__x == _Tp(0))  |
652 | return -std::numeric_limits<_Tp>::infinity();  |
653 | else  |
654 | {  |
655 | _Tp __j_n, __n_n, __jp_n, __np_n;  |
656 | __sph_bessel_jn(__n, __x, __j_n, __n_n, __jp_n, __np_n);  |
657 | return __n_n;  |
658 | }  |
659 | }  |
660 | } // namespace __detail  |
661 | #undef _GLIBCXX_MATH_NS  |
662 | #if ! _GLIBCXX_USE_STD_SPEC_FUNCS && defined(_GLIBCXX_TR1_CMATH)  |
663 | } // namespace tr1  |
664 | #endif  |
665 |   |
666 | _GLIBCXX_END_NAMESPACE_VERSION  |
667 | }  |
668 |   |
669 | #endif // _GLIBCXX_TR1_BESSEL_FUNCTION_TCC  |
670 | |