1 /** @file inifcns_nstdsums.cpp
3 * Implementation of some special functions that have a representation as nested sums.
5 * classical polylogarithm Li(n,x)
6 * multiple polylogarithm Li(lst(n_1,...,n_k),lst(x_1,...,x_k)
7 * nielsen's generalized polylogarithm S(n,p,x)
8 * harmonic polylogarithm H(lst(m_1,...,m_k),x)
9 * multiple zeta value mZeta(lst(m_1,...,m_k))
12 * - All formulae used can be looked up in the following publication:
13 * Nielsen's Generalized Polylogarithms, K.S.Kolbig, SIAM J.Math.Anal. 17 (1986), pp. 1232-1258.
14 * This document will be referenced as [Kol] throughout this source code.
15 * - Classical polylogarithms (Li) and nielsen's generalized polylogarithms (S) can be numerically
16 * evaluated in the whole complex plane except for S(n,p,-1) when p is not unit (no formula yet
17 * to tackle these points). And of course, there is still room for speed optimizations ;-).
18 * - The calculation of classical polylogarithms is speed up by using Euler-MacLaurin summation.
19 * - The remaining functions can only be numerically evaluated with arguments lying in the unit sphere
20 * at the moment. Sorry. The evaluation especially for mZeta is very slow ... better not use it
22 * - The functions have no series expansion. To do it, you have to convert these functions
23 * into the appropriate objects from the nestedsums library, do the expansion and convert the
25 * - Numerical testing of this implementation has been performed by doing a comparison of results
26 * between this software and the commercial M.......... 4.1.
31 * GiNaC Copyright (C) 1999-2003 Johannes Gutenberg University Mainz, Germany
33 * This program is free software; you can redistribute it and/or modify
34 * it under the terms of the GNU General Public License as published by
35 * the Free Software Foundation; either version 2 of the License, or
36 * (at your option) any later version.
38 * This program is distributed in the hope that it will be useful,
39 * but WITHOUT ANY WARRANTY; without even the implied warranty of
40 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
41 * GNU General Public License for more details.
43 * You should have received a copy of the GNU General Public License
44 * along with this program; if not, write to the Free Software
45 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
55 #include "operators.h"
56 #include "relational.h"
63 // lookup table for Euler-MacLaurin optimization
65 std::vector<std::vector<cln::cl_N> > Xn;
69 // lookup table for Euler-Zagier-Sums (used for S_n,p(x))
71 std::vector<std::vector<cln::cl_N> > Yn;
73 //TODO EVIL MAGIC NUMBER !!! but first the transformations for S have to improve ...
74 const int initsize_Yn = 2000;
77 //////////////////////
78 // helper functions //
79 //////////////////////
82 // This function calculates the X_n. The X_n are needed for the Euler-MacLaurin summation (EMS) of
83 // classical polylogarithms.
84 // With EMS the polylogs can be calculated as follows:
85 // Li_p (x) = \sum_{n=0}^\infty X_{p-2}(n) u^{n+1}/(n+1)! with u = -log(1-x)
86 // X_0(n) = B_n (Bernoulli numbers)
87 // X_p(n) = \sum_{k=0}^n binomial(n,k) B_{n-k} / (k+1) * X_{p-1}(k)
88 // The calculation of Xn depends on X0 and X{n-1}.
89 // X_0 is special, it holds only the non-zero Bernoulli numbers with index 2 or greater.
90 // This results in a slightly more complicated algorithm for the X_n.
91 // The first index in Xn corresponds to the index of the polylog minus 2.
92 // The second index in Xn corresponds to the index from the EMS.
93 static void fill_Xn(int n)
95 // rule of thumb. needs to be improved. TODO
96 const int initsize = Digits * 3 / 2;
99 // calculate X_2 and higher (corresponding to Li_4 and higher)
100 std::vector<cln::cl_N> buf(initsize);
101 std::vector<cln::cl_N>::iterator it = buf.begin();
103 *it = -(cln::expt(cln::cl_I(2),n+1) - 1) / cln::expt(cln::cl_I(2),n+1); // i == 1
105 for (int i=2; i<=initsize; i++) {
107 result = 0; // k == 0
109 result = Xn[0][i/2-1]; // k == 0
111 for (int k=1; k<i-1; k++) {
112 if ( !(((i-k) & 1) && ((i-k) > 1)) ) {
113 result = result + cln::binomial(i,k) * Xn[0][(i-k)/2-1] * Xn[n-1][k-1] / (k+1);
116 result = result - cln::binomial(i,i-1) * Xn[n-1][i-2] / 2 / i; // k == i-1
117 result = result + Xn[n-1][i-1] / (i+1); // k == i
124 // special case to handle the X_0 correct
125 std::vector<cln::cl_N> buf(initsize);
126 std::vector<cln::cl_N>::iterator it = buf.begin();
128 *it = cln::cl_I(-3)/cln::cl_I(4); // i == 1
130 *it = cln::cl_I(17)/cln::cl_I(36); // i == 2
132 for (int i=3; i<=initsize; i++) {
134 result = -Xn[0][(i-3)/2]/2;
135 *it = (cln::binomial(i,1)/cln::cl_I(2) + cln::binomial(i,i-1)/cln::cl_I(i))*result;
138 result = Xn[0][i/2-1] + Xn[0][i/2-1]/(i+1);
139 for (int k=1; k<i/2; k++) {
140 result = result + cln::binomial(i,k*2) * Xn[0][k-1] * Xn[0][i/2-k-1] / (k*2+1);
149 std::vector<cln::cl_N> buf(initsize/2);
150 std::vector<cln::cl_N>::iterator it = buf.begin();
151 for (int i=1; i<=initsize/2; i++) {
152 *it = bernoulli(i*2).to_cl_N();
161 // This function calculates the Y_n. The Y_n are needed for the evaluation of S_{n,p}(x).
162 // The Y_n are basically Euler-Zagier sums with all m_i=1. They are subsums in the Z-sum
163 // representing S_{n,p}(x).
164 // The first index in Y_n corresponds to the parameter p minus one, i.e. the depth of the
166 // The second index in Y_n corresponds to the running index of the outermost sum in the full Z-sum
167 // representing S_{n,p}(x).
168 // The calculation of Y_n uses the values from Y_{n-1}.
169 static void fill_Yn(int n)
171 // TODO -> get rid of the magic number
172 const int initsize = initsize_Yn;
175 std::vector<cln::cl_N> buf(initsize);
176 std::vector<cln::cl_N>::iterator it = buf.begin();
177 std::vector<cln::cl_N>::iterator itprev = Yn[n-1].begin();
178 *it = (*itprev) / cln::cl_N(n+1);
181 // sums with an index smaller than the depth are zero and need not to be calculated.
182 // calculation starts with depth, which is n+2)
183 for (int i=n+2; i<=initsize+n; i++) {
184 *it = *(it-1) + (*itprev) / cln::cl_N(i);
190 std::vector<cln::cl_N> buf(initsize);
191 std::vector<cln::cl_N>::iterator it = buf.begin();
194 for (int i=2; i<=initsize; i++) {
195 *it = *(it-1) + 1 / cln::cl_N(i);
204 static cln::cl_N Li_series(int n, const cln::cl_N& x, const cln::float_format_t& prec)
206 // check if precalculated values are sufficient
208 for (int i=xnsize; i<n-1; i++) {
213 // using Euler-MacLaurin summation
215 // Li_2. X_0 is special ...
216 std::vector<cln::cl_N>::const_iterator it = Xn[0].begin();
217 cln::cl_N u = -cln::log(cln::complex(cln::cl_float(1, prec), 0)-x);
218 cln::cl_N factor = u;
219 cln::cl_N res = u - u*u/4;
221 for (int i=1; true; i++) {
223 factor = factor * u*u / (2*i * (2*i+1));
224 res = res + (*it) * factor;
225 it++; // should we check it? or rely on initsize? ...
226 if (cln::zerop(res-resbuf))
234 std::vector<cln::cl_N>::const_iterator it = Xn[n-2].begin();
235 cln::cl_N u = -cln::log(cln::complex(cln::cl_float(1, prec), 0)-x);
236 cln::cl_N factor = u;
239 for (int i=1; true; i++) {
241 factor = factor * u / (i+1);
242 res = res + (*it) * factor;
243 it++; // should we check it? or rely on initsize? ...
244 if (cln::zerop(res-resbuf))
246 // should not be needed.
247 // if (!cln::zerop(*it)) {
257 // forward declaration needed by function C below
258 static numeric S_num(int n, int p, const numeric& x);
261 // helper function for classical polylog Li
262 static cln::cl_N Li_projection(int n, const cln::cl_N& x, const cln::float_format_t& prec)
264 if (cln::realpart(x) < 0.5) {
265 return Li_series(n, x, prec);
268 return -Li_series(2, 1-x, prec) - cln::log(x) * cln::log(1-x) + cln::zeta(2);
270 cln::cl_N result = -cln::expt(cln::log(x), n-1) * cln::log(1-x) / cln::factorial(n-1);
271 for (int j=0; j<n-1; j++) {
272 result = result + (S_num(n-j-1, 1, 1).to_cl_N() - S_num(1, n-j-1, 1-x).to_cl_N())
273 * cln::expt(cln::log(x), j) / cln::factorial(j) ;
281 // helper function for classical polylog Li
282 static numeric Li_num(int n, const numeric& x)
286 return -cln::log(1-x.to_cl_N());
297 return -(1-cln::expt(cln::cl_I(2),1-n)) * cln::zeta(n);
300 // what is the desired float format?
301 // first guess: default format
302 cln::float_format_t prec = cln::default_float_format;
303 const cln::cl_N value = x.to_cl_N();
304 // second guess: the argument's format
305 if (!x.real().is_rational())
306 prec = cln::float_format(cln::the<cln::cl_F>(cln::realpart(value)));
307 else if (!x.imag().is_rational())
308 prec = cln::float_format(cln::the<cln::cl_F>(cln::imagpart(value)));
311 if (cln::abs(value) > 1) {
312 cln::cl_N result = -cln::expt(cln::log(-value),n) / cln::factorial(n);
313 // check if argument is complex. if it is real, the new polylog has to be conjugated.
314 if (cln::zerop(cln::imagpart(value))) {
316 result = result + conjugate(Li_projection(n, cln::recip(value), prec));
319 result = result - conjugate(Li_projection(n, cln::recip(value), prec));
324 result = result + Li_projection(n, cln::recip(value), prec);
327 result = result - Li_projection(n, cln::recip(value), prec);
331 for (int j=0; j<n-1; j++) {
332 add = add + (1+cln::expt(cln::cl_I(-1),n-j)) * (1-cln::expt(cln::cl_I(2),1-n+j))
333 * Li_num(n-j,1).to_cl_N() * cln::expt(cln::log(-value),j) / cln::factorial(j);
335 result = result - add;
339 return Li_projection(n, value, prec);
344 // helper function for S(n,p,x)
345 static cln::cl_N numeric_nielsen(int n, int step)
349 for (int i=1; i<n; i++) {
350 res = res + numeric_nielsen(i, step-1) / cln::cl_I(i);
360 // helper function for S(n,p,x)
362 static cln::cl_N C(int n, int p)
366 for (int k=0; k<p; k++) {
367 for (int j=0; j<=(n+k-1)/2; j++) {
371 result = result - 2 * cln::expt(cln::pi(),2*j) * S_num(n-2*j,p,1).to_cl_N() / cln::factorial(2*j);
374 result = result + 2 * cln::expt(cln::pi(),2*j) * S_num(n-2*j,p,1).to_cl_N() / cln::factorial(2*j);
381 result = result + cln::factorial(n+k-1)
382 * cln::expt(cln::pi(),2*j) * S_num(n+k-2*j,p-k,1).to_cl_N()
383 / (cln::factorial(k) * cln::factorial(n-1) * cln::factorial(2*j));
386 result = result - cln::factorial(n+k-1)
387 * cln::expt(cln::pi(),2*j) * S_num(n+k-2*j,p-k,1).to_cl_N()
388 / (cln::factorial(k) * cln::factorial(n-1) * cln::factorial(2*j));
393 result = result - cln::factorial(n+k-1) * cln::expt(cln::pi(),2*j) * S_num(n+k-2*j,p-k,1).to_cl_N()
394 / (cln::factorial(k) * cln::factorial(n-1) * cln::factorial(2*j));
397 result = result + cln::factorial(n+k-1)
398 * cln::expt(cln::pi(),2*j) * S_num(n+k-2*j,p-k,1).to_cl_N()
399 / (cln::factorial(k) * cln::factorial(n-1) * cln::factorial(2*j));
407 if (((np)/2+n) & 1) {
408 result = -result - cln::expt(cln::pi(),np) / (np * cln::factorial(n-1) * cln::factorial(p));
411 result = -result + cln::expt(cln::pi(),np) / (np * cln::factorial(n-1) * cln::factorial(p));
419 // helper function for S(n,p,x)
420 // [Kol] remark to (9.1)
421 static cln::cl_N a_k(int k)
430 for (int m=2; m<=k; m++) {
431 result = result + cln::expt(cln::cl_N(-1),m) * cln::zeta(m) * a_k(k-m);
438 // helper function for S(n,p,x)
439 // [Kol] remark to (9.1)
440 static cln::cl_N b_k(int k)
449 for (int m=2; m<=k; m++) {
450 result = result + cln::expt(cln::cl_N(-1),m) * cln::zeta(m) * b_k(k-m);
457 // helper function for S(n,p,x)
458 static cln::cl_N S_series(int n, int p, const cln::cl_N& x, const cln::float_format_t& prec)
461 return Li_series(n+1, x, prec);
464 // TODO -> check for vector boundaries and do missing calculations
466 // check if precalculated values are sufficient
468 for (int i=ynsize; i<p-1; i++) {
474 cln::cl_N resultbuffer;
475 for (int i=p; true; i++) {
476 resultbuffer = result;
477 result = result + cln::expt(x,i) / cln::expt(cln::cl_I(i),n+1) * Yn[p-2][i-p]; // should we check it? or rely on magic number? ...
478 if (cln::zerop(result-resultbuffer)) {
487 // helper function for S(n,p,x)
488 static cln::cl_N S_projection(int n, int p, const cln::cl_N& x, const cln::float_format_t& prec)
491 if (cln::abs(cln::realpart(x)) > cln::cl_F("0.5")) {
493 cln::cl_N result = cln::expt(cln::cl_I(-1),p) * cln::expt(cln::log(x),n)
494 * cln::expt(cln::log(1-x),p) / cln::factorial(n) / cln::factorial(p);
496 for (int s=0; s<n; s++) {
498 for (int r=0; r<p; r++) {
499 res2 = res2 + cln::expt(cln::cl_I(-1),r) * cln::expt(cln::log(1-x),r)
500 * S_series(p-r,n-s,1-x,prec) / cln::factorial(r);
502 result = result + cln::expt(cln::log(x),s) * (S_num(n-s,p,1).to_cl_N() - res2) / cln::factorial(s);
508 return S_series(n, p, x, prec);
512 // helper function for S(n,p,x)
513 static numeric S_num(int n, int p, const numeric& x)
517 // [Kol] (2.22) with (2.21)
518 return cln::zeta(p+1);
523 return cln::zeta(n+1);
528 for (int nu=0; nu<n; nu++) {
529 for (int rho=0; rho<=p; rho++) {
530 result = result + b_k(n-nu-1) * b_k(p-rho) * a_k(nu+rho+1)
531 * cln::factorial(nu+rho+1) / cln::factorial(rho) / cln::factorial(nu+1);
534 result = result * cln::expt(cln::cl_I(-1),n+p-1);
541 return -(1-cln::expt(cln::cl_I(2),-n)) * cln::zeta(n+1);
543 throw std::runtime_error("don't know how to evaluate this function!");
546 // what is the desired float format?
547 // first guess: default format
548 cln::float_format_t prec = cln::default_float_format;
549 const cln::cl_N value = x.to_cl_N();
550 // second guess: the argument's format
551 if (!x.real().is_rational())
552 prec = cln::float_format(cln::the<cln::cl_F>(cln::realpart(value)));
553 else if (!x.imag().is_rational())
554 prec = cln::float_format(cln::the<cln::cl_F>(cln::imagpart(value)));
558 if (cln::realpart(value) < -0.5) {
560 cln::cl_N result = cln::expt(cln::cl_I(-1),p) * cln::expt(cln::log(value),n)
561 * cln::expt(cln::log(1-value),p) / cln::factorial(n) / cln::factorial(p);
563 for (int s=0; s<n; s++) {
565 for (int r=0; r<p; r++) {
566 res2 = res2 + cln::expt(cln::cl_I(-1),r) * cln::expt(cln::log(1-value),r)
567 * S_num(p-r,n-s,1-value).to_cl_N() / cln::factorial(r);
569 result = result + cln::expt(cln::log(value),s) * (S_num(n-s,p,1).to_cl_N() - res2) / cln::factorial(s);
576 else if (cln::abs(value) > 1) {
580 for (int s=0; s<p; s++) {
581 for (int r=0; r<=s; r++) {
582 result = result + cln::expt(cln::cl_I(-1),s) * cln::expt(cln::log(-value),r) * cln::factorial(n+s-r-1)
583 / cln::factorial(r) / cln::factorial(s-r) / cln::factorial(n-1)
584 * S_num(n+s-r,p-s,cln::recip(value)).to_cl_N();
587 result = result * cln::expt(cln::cl_I(-1),n);
590 for (int r=0; r<n; r++) {
591 res2 = res2 + cln::expt(cln::log(-value),r) * C(n-r,p) / cln::factorial(r);
593 res2 = res2 + cln::expt(cln::log(-value),n+p) / cln::factorial(n+p);
595 result = result + cln::expt(cln::cl_I(-1),p) * res2;
600 return S_projection(n, p, value, prec);
605 // helper function for multiple polylogarithm
606 static cln::cl_N numeric_zsum(int n, std::vector<cln::cl_N>& x, std::vector<cln::cl_N>& m)
612 for (int i=1; i<n; i++) {
613 std::vector<cln::cl_N>::iterator be;
614 std::vector<cln::cl_N>::iterator en;
618 std::vector<cln::cl_N> xbuf(be, en);
622 std::vector<cln::cl_N> mbuf(be, en);
623 res = res + cln::expt(x[0],i) / cln::expt(i,m[0]) * numeric_zsum(i, xbuf, mbuf);
629 // helper function for harmonic polylogarithm
630 static cln::cl_N numeric_harmonic(int n, std::vector<cln::cl_N>& m)
636 for (int i=1; i<n; i++) {
637 std::vector<cln::cl_N>::iterator be;
638 std::vector<cln::cl_N>::iterator en;
642 std::vector<cln::cl_N> mbuf(be, en);
643 res = res + cln::recip(cln::expt(i,m[0])) * numeric_harmonic(i, mbuf);
649 /////////////////////////////
650 // end of helper functions //
651 /////////////////////////////
654 // Polylogarithm and multiple polylogarithm
656 static ex Li_eval(const ex& x1, const ex& x2)
662 return Li(x1,x2).hold();
666 static ex Li_evalf(const ex& x1, const ex& x2)
668 // classical polylogs
669 if (is_a<numeric>(x1) && is_a<numeric>(x2)) {
670 return Li_num(ex_to<numeric>(x1).to_int(), ex_to<numeric>(x2));
673 else if (is_a<lst>(x1) && is_a<lst>(x2)) {
674 for (int i=0; i<x1.nops(); i++) {
675 if (!is_a<numeric>(x1.op(i)))
676 return Li(x1,x2).hold();
677 if (!is_a<numeric>(x2.op(i)))
678 return Li(x1,x2).hold();
680 return Li(x1,x2).hold();
683 cln::cl_N m_1 = ex_to<numeric>(x1.op(x1.nops()-1)).to_cl_N();
684 cln::cl_N x_1 = ex_to<numeric>(x2.op(x2.nops()-1)).to_cl_N();
685 std::vector<cln::cl_N> x;
686 std::vector<cln::cl_N> m;
687 const int nops = ex_to<numeric>(x1.nops()).to_int();
688 for (int i=nops-2; i>=0; i--) {
689 m.push_back(ex_to<numeric>(x1.op(i)).to_cl_N());
690 x.push_back(ex_to<numeric>(x2.op(i)).to_cl_N());
695 for (int i=nops; true; i++) {
697 res = res + cln::expt(x_1,i) / cln::expt(i,m_1) * numeric_zsum(i, x, m);
698 if (cln::zerop(res-resbuf))
706 return Li(x1,x2).hold();
709 static ex Li_series(const ex& x1, const ex& x2, const relational& rel, int order, unsigned options)
712 seq.push_back(expair(Li(x1,x2), 0));
713 return pseries(rel,seq);
716 REGISTER_FUNCTION(Li, eval_func(Li_eval).evalf_func(Li_evalf).do_not_evalf_params().series_func(Li_series));
719 // Nielsen's generalized polylogarithm
721 static ex S_eval(const ex& x1, const ex& x2, const ex& x3)
726 return S(x1,x2,x3).hold();
729 static ex S_evalf(const ex& x1, const ex& x2, const ex& x3)
731 if (is_a<numeric>(x1) && is_a<numeric>(x2) && is_a<numeric>(x3)) {
732 if ((x3 == -1) && (x2 != 1)) {
733 // no formula to evaluate this ... sorry
734 return S(x1,x2,x3).hold();
736 return S_num(ex_to<numeric>(x1).to_int(), ex_to<numeric>(x2).to_int(), ex_to<numeric>(x3));
738 return S(x1,x2,x3).hold();
741 static ex S_series(const ex& x1, const ex& x2, const ex& x3, const relational& rel, int order, unsigned options)
744 seq.push_back(expair(S(x1,x2,x3), 0));
745 return pseries(rel,seq);
748 REGISTER_FUNCTION(S, eval_func(S_eval).evalf_func(S_evalf).do_not_evalf_params().series_func(S_series));
751 // Harmonic polylogarithm
753 static ex H_eval(const ex& x1, const ex& x2)
755 return H(x1,x2).hold();
758 static ex H_evalf(const ex& x1, const ex& x2)
760 if (is_a<lst>(x1) && is_a<numeric>(x2)) {
761 for (int i=0; i<x1.nops(); i++) {
762 if (!is_a<numeric>(x1.op(i)))
763 return H(x1,x2).hold();
766 return H(x1,x2).hold();
769 cln::cl_N m_1 = ex_to<numeric>(x1.op(x1.nops()-1)).to_cl_N();
770 cln::cl_N x_1 = ex_to<numeric>(x2).to_cl_N();
771 std::vector<cln::cl_N> m;
772 const int nops = ex_to<numeric>(x1.nops()).to_int();
773 for (int i=nops-2; i>=0; i--) {
774 m.push_back(ex_to<numeric>(x1.op(i)).to_cl_N());
779 for (int i=nops; true; i++) {
781 res = res + cln::expt(x_1,i) / cln::expt(i,m_1) * numeric_harmonic(i, m);
782 if (cln::zerop(res-resbuf))
790 return H(x1,x2).hold();
793 static ex H_series(const ex& x1, const ex& x2, const relational& rel, int order, unsigned options)
796 seq.push_back(expair(H(x1,x2), 0));
797 return pseries(rel,seq);
800 REGISTER_FUNCTION(H, eval_func(H_eval).evalf_func(H_evalf).do_not_evalf_params().series_func(H_series));
803 // Multiple zeta value
805 static ex mZeta_eval(const ex& x1)
807 return mZeta(x1).hold();
810 static ex mZeta_evalf(const ex& x1)
813 for (int i=0; i<x1.nops(); i++) {
814 if (!is_a<numeric>(x1.op(i)))
815 return mZeta(x1).hold();
818 cln::cl_N m_1 = ex_to<numeric>(x1.op(x1.nops()-1)).to_cl_N();
819 std::vector<cln::cl_N> m;
820 const int nops = ex_to<numeric>(x1.nops()).to_int();
821 for (int i=nops-2; i>=0; i--) {
822 m.push_back(ex_to<numeric>(x1.op(i)).to_cl_N());
825 cln::float_format_t prec = cln::default_float_format;
826 cln::cl_N res = cln::complex(cln::cl_float(0, prec), 0);
828 for (int i=nops; true; i++) {
829 // to infinity and beyond ... timewise
831 res = res + cln::recip(cln::expt(i,m_1)) * numeric_harmonic(i, m);
832 if (cln::zerop(res-resbuf))
840 return mZeta(x1).hold();
843 static ex mZeta_series(const ex& x1, const relational& rel, int order, unsigned options)
846 seq.push_back(expair(mZeta(x1), 0));
847 return pseries(rel,seq);
850 REGISTER_FUNCTION(mZeta, eval_func(mZeta_eval).evalf_func(mZeta_evalf).do_not_evalf_params().series_func(mZeta_series));