Jacques Gélinas on Mon, 15 Apr 2019 14:28:33 +0200 |
[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]
Learning with GP: Verification of identities - Romik sequence |
The finite expression DT3 below has been found in [2] for derivatives at x=1 of the third Jacobi theta constant, theta3(x) := sum(n=-oo,oo,exp(-n^2*Pi*x)) This involves a new "Romik sequence" of integers dk defined via the infinite series dh(k) below [2,3]. 0. In order to verify identities for this sequence we use comparison routines from [1] (verif.gp). \\-------------------------------------------- Start cut eqr(x,y) = if(x==y,1,exponent(if(!x,y,1-y/x))/exponent(0.)); fleq(x,y,r=7/8) = if(x==y, 1, r <= \ eqr(0,normlp(x-y)/normlp(x+y))); IS(e,msg="") = print(if(e,"Pass: ","Fail: "),msg); ISF(x,y,msg="",r=7/8) = IS( fleq(x,y,r),msg); ISN(f,g,msg="",r=7/8,N=12) = IS( N == sum(n=1,N, fleq(f(n),g(n),r)), msg); /*1. The derivatives of theta3(x) at x=1 are equal to a rational polynomial in Pi and in T1 := theta(1) which can be computed in many different ways by GP.*/ T1 = Pi^(1/4) / gamma(3/4); \\ 1.0864348... oeis.org/A175573 ISF( T1, 1/sqrt(2*Pi)*gamma(1/4)/Pi^(1/4), "T1 gamma(1/4)" ); ISF( T1, 1+2*sumpos(n=1,exp(-n^2*Pi)), "T1 definition"); ISF( T1, 4/(1-2^(-1/4))*sumpos(n=1,exp(-(2*n-1)^2*Pi)),\ "T1 Whittaker-Watson, Modern Analysis p. 533" ); \\ L-functions (Bill Allombert) tzeta(x) = 1 + lfuntheta(1,sqrt(x)); \\ associated to zeta ISF( T1, tzeta(1), "T1=1+lfuntheta(1,sqrt(1))" ); Dtheta3(n=0,x=1) = lfuntheta(lfunqf(Mat(1)),x,n)/2 + (!n); ISF( T1, Dtheta3(0), "T1=Dtheta3(0,1) via lfuntheta(lfunqf)" ); \\ modular forms with q=exp(2*I*Pi*z) (Bill Allombert) mf = mfinit(Tm=mfTheta()); ISF( T1, mfeval(mf,Tm,I/2), "T1 via mfTheta" ); [mf,F] = mffromqf(Mat(2)); ISF( T1, mfeval(mf,F,I/2), "T1 via mffromqf" ); /*2. Each Romik integer can be computed by summing an infinite series of weighted Hermite polynomial values. At least 450D/25 terms are needed to verify the Zeilberger value [3] dk(100) = -4.7885...33551E336, and 1000D/35 terms for dk(200) = -1.2713...63551E792. We verify the first nine integers and also twice a (simplified) formula for the first twelve derivatives.*/ B = 2 * Pi^2 * T1^8; \\ 4*\Omega in [2]=38.31... dk(k) = round(dh(k)/T1/B^k); \\ [2, Proposition 10] dh(k) = my(z, pk=y->substpol(polhermite(4*k),x^2,2*y) );\ pk(0) + 2*suminf(n=1,z=n^2*Pi; exp(-z)*pk(z) ); IS( [1,1,-1,51,849,-26199,1341999,82018251,18703396449] == [dk(k)|k<-[0..8]], "Romik sequence d[0..8]" ); \\ DT3(n) = theta3^(n)(1) [2, Theorem 1] DT3(n) = (2*n)!/(-8)^n*sum(k=0,n/2, dh(k)/(n-2*k)!/(4*k)!); ISN( DT3, n->derivnum(x=1,tzeta(x),n), "DT3 vs derivnum" ); ISN( DT3, Dtheta3, "DT3 vs lfuntheta(lfunqf)" ); /*3. The functional equation theta3(1/x)=sqrt(x)*theta3(x), when differentiated and evaluated at x=1, yields a sequence of identities (such as -2*theta3'(1) = theta3(1)/2) that we use to verify our version DT3 of the Romik formula.*/ \\ Coffey identities [4] ISF( -2*DT3(1), T1/2, "-2*theta3'(1) = theta3(1)/2" ); ff(n) = prod(j=1,n,2*j-1); \\ !!=double factorial ISN({ n-> if(n==1,0,(1-(-1)^n)*DT3(n) - ff(n-1)*T1/(-2)^n), n-> sum(j=1,n-1,binomial(n,j)*DT3(j)* ((-1)^n*(n-1)!/(j-1)! + ff(n-j-1)/(-2)^(n-j))), "Coffey identities for theta3^(n)(1)");} \\-------------------------------------------- End paste ------------------------------ Features used !,Pi,my,prod,sum,sumpos,suminf,binomial,gamma, exponent,normlp,round,derivnum,exp,substpol, sqrt,polhermite,(anonymous closure),[f(n)|n<-[...]], lfuntheta,lfunqf,mfinit,mfeval,mfTheta,mffromqf ------------------------------ References [1] Learning with GP: Verification of identities 2019-04-08 pari.math.u-bordeaux.fr/archives/pari-users-1904/msg00008.html [2] Romik 2018 The Taylor coefficients of the Jacobi theta constant theta_3, arxiv.org/abs/1807.06130, to be published in The Ramanujan Journal. [3] Sequence related to the Jacobi theta_3 constant, oeis.org/A317651 Maple code from Doron Zeilberger and first 200 integers available at sites.math.rutgers.edu/~zeilberg/mamarim/mamarimhtml/Theta3RomikUM.html [4] Coffey 2002 A set of identities for a theta function at a unit argument, Physics Letters A, 300:367-369. [5] Tutorials: pari.math.u-bordeaux.fr/pub/pari/manuals/2.12.0/tutorial.pdf pari.math.u-bordeaux.fr/pub/pari/manuals/2.12.0/tutorial-mf.pdf [6] Reference cards: http://pari.math.u-bordeaux.fr/pub/pari/manuals/2.12.0/refcard.pdf pari.math.u-bordeaux.fr/pub/pari/manuals/2.11.1/refcard-mf.pdf Jacques Gélinas