Preliminary Results
Let us consider the ascending Landen sequence of moduli ⋯<k−n<k−(n−1)<⋯<k−2<k−1<k0=k<k1<k2<⋯<kn<⋯ where kn+1=2√kn1+kn,kn=1−k′n+11+k′n+1 Then it can be checked easily that the sequence of complementary moduli in reverse order ⋯<k′n<k′n−1<⋯<k′2<k′1<k′0=k′<k′−1<k′−2<⋯<k′−n<⋯ also forms an ascending Landen sequence.Also let us put Kn=K(kn),K′n=K(k′n). We then have Kn+1=(1+kn)Kn for all n. Hence we can write K0=(1+k−1)K−1=(1+k−1)(1+k−2)K−2=(1+k−1)(1+k−2)⋯(1+k−n)K−n⇒K0K−n=(1+k−1)(1+k−2)⋯(1+k−n) and letting n→∞ and noting that k−n→0 so that K−n→π/2 we get 2Kπ=(1+k−1)(1+k−2)(1+k−3)⋯ and more generally if λ0>λ1>λ2>⋯ forms a decreasing Landen sequence then 2Λπ=2K(λ0)π=(1+λ1)(1+λ2)(1+λ3)⋯ In particular we have 2K′π=(1+k′1)(1+k′2)(1+k′3)⋯ Again we note that (1+k′1)K1=2K(1+k′2)K2=2K1(1+k′3)K3=2K2(1+k′4)K4=2K3(1+k′n)Kn=2Kn−1 Multiplying theabove equations we get (1+k′1)(1+k′2)(1+k′3)⋯(1+k′n)Kn=2nK or Kn2nK=1(1+k′1)(1+k′2)(1+k′3)⋯(1+k′n) Taking limits as n→∞ we get limn→∞Kn2nK=π2K′
Infinite Product Expansion of Elliptic Functions
We shall now use the standard Landen's ascending transformation sn(u,k)=(1+k′1)sn((1+k)u2,k1)cn((1+k)u2,k1)dn((1+k)u2,k1)=(1+k′1)sn(K1u2K,k1)cn(K1u2K,k1)dn(K1u2K,k1)=(1+k′1)sn(K1u2K,k1)sn(K1u2K+K1,k1)=(1+k′1)sn(K1u2K,k1)sn(K12K(u+2K),k1) Using the same formula again we get sn(K1u2K,k1)=(1+k′2)sn(K22K1K1u2K,k2)sn(K22K1(K1u2K+2K1),k2)=(1+k′2)sn(K2u22K,k2)sn(K222K(u+4K),k2) and by the same token sn(K12K(u+2K),k1)=(1+k′2)sn(K222K(u+2K),k2)sn(K222K(u+6K),k2) Hence we have sn(u,k)=(1+k′1)(1+k′2)2sn(K2u22K,k2)sn(K222K(u+2K),k2)sn(K222K(u+4K),k2)sn(K222K(u+6K),k2) We can write the last term as sn(K222K(u+6K),k2)=sn(2K2−K222K(u+6K),k2)=sn(K222K(2K−u),k2) and thus we can write sn(u,k)=(1+k′1)(1+k′2)2sn(K2u22K,k2)sn(K222K(2K+u),k2)sn(K222K(2K−u),k2)sn(K222K(4K+u),k2) This can be continued further in the same way to yield sn(u,k)=(1+k′1)(1+k′2)2(1+k′3)4sn(K3u23K,k3)sn(K323K(2K+u),k3)sn(K323K(2K−u),k3)sn(K323K(4K+u),k3)sn(K323K(4K−u),k3)sn(K323K(6K+u),k3)sn(K323K(6K−u),k3)sn(K323K(8K+u),k3) Finally if we continue till the modulus kn we get sn(u,k)=(1+k′1)(1+k′2)2(1+k′3)4⋯(1+k′n)2n−1sn(Knu2nK,kn)sn(Kn2nK(2K+u),kn)sn(Kn2nK(2K−u),kn)sn(Kn2nK(4K+u),kn)sn(Kn2nK(4K−u),kn)sn(Kn2nK(6K+u),kn)sn(Kn2nK(6K−u),kn)⋯⋯sn(Kn2nK((2n−2)K+u),kn)sn(Kn2nK((2n−2)K−u),kn)sn(Kn2nK(2nK+u),kn) Now we can take limits as n→∞ (assuming that we have convergence for the entities we are concerned with here, this assumption being justified at a later stage) and note that kn→1 and the ratio Kn/(2nK)→π/(2K′). Thus we arrive at sn(u,k)=A′sn(πu2K′,1)∞∏n=1sn(π2K′(2nK+u),1)sn(π2K′(2nK−u),1) where A′=limn→∞(1+k′1)(1+k′2)2(1+k′3)4⋯(1+k′n)2n−1 Since sn(v,1)=tanhv=ev−e−vev+e−v we can write sn(π2K′(2nK+u),1)sn(π2K′(2nK−u),1)=tanh(nπKK′+πu2K′)tanh(nπKK′−πu2K′) Let us put q′=exp(−πKK′),x=πu2K′ and then the above expression becomes (q′−nex−q′ne−xq′−nex−q′ne−x)(q′−ne−x−q′nexq′−ne−x+q′nex)=q′−2n+q′2n−(e2x+e−2x)q′−2n+q′2n+(e2x+e−2x)=q′−2n+q′2n−2cos(2ix)q′−2n+q′2n+2cos(2ix)=1−2q′2ncos(2ix)+q′4n1+2q′2ncos(2ix)+q′4n=1−2q′2ncos(πiuK′)+q′4n1+2q′2ncos(πiuK′)+q′4n Hence we finally have sn(u,k)=A′tanh(πu2K′)∞∏n=11−2q′2ncos(πiuK′)+q′4n1+2q′2ncos(πiuK′)+q′4n Noting that tanhx=−itan(ix) and sn(u,k)=−isc(iu,k′) we get sn(iu,k′)cn(iu,k′)=A′tan(πiu2K′)∞∏n=11−2q′2ncos(πiuK′)+q′4n1+2q′2ncos(πiuK′)+q′4n Replacing iu by u and k′ by k we get sn(u,k)cn(u,k)=Atan(πu2K)∞∏n=11−2q2ncos(πuK)+q4n1+2q2ncos(πuK)+q4n where A is some constant and q=exp(−πK′/K).The above equation marks the birth of the parameter q and all things q-otic (not chaotic!!) like the q-series, q-products. We shall have occasion to deal with lot of such q-otic stuff in later posts.
We can again go back to the equation sn(u,k)=A′sn(πu2K′,1)∞∏n=1sn(π2K′(2nK+u),1)sn(π2K′(2nK−u),1) Replacing u by u+K we get sn(u+K,k)=A′sn(π(u+K)2K′,1)∞∏n=1sn(π2K′((2n+1)K+u),1)sn(π2K′((2n−1)K−u),1)⇒cn(u,k)dn(u,k)=A′∞∏n=1sn(π2K′((2n−1)K+u),1)sn(π2K′((2n−1)K−u),1) and like the previous case we have cn(u,k)dn(u,k)=A′∞∏n=11−2q′2n−1cos(πiuK′)+q′4n−21+2q′2n−1cos(πiuK′)+q′4n−2⇒1dn(iu,k′)=A′∞∏n=11−2q′2n−1cos(πiuK′)+q′4n−21+2q′2n−1cos(πiuK′)+q′4n−2⇒dn(iu,k′)=1A′∞∏n=11+2q′2n−1cos(πiuK′)+q′4n−21−2q′2n−1cos(πiuK′)+q′4n−2 Replacing iu by u and k′ by k we get dn(u,k)=1A∞∏n=11+2q2n−1cos(πuK)+q4n−21−2q2n−1cos(πuK)+q4n−2 To evaluate the constant A we put u=0 in the above equation and get A=∞∏n=11+2q2n−1+q4n−21−2q2n−1+q4n−2=∞∏n=1(1+q2n−11−q2n−1)2 Again putting u=K we get k′=1A∞∏n=11−2q2n−1+q4n−21+2q2n−1+q4n−2=1A2 Thus we get 1A=√k′,k′=∞∏n=1(1−q2n−11+q2n−1)4 It now follows that dn(u,k)=√k′∞∏n=11+2q2n−1cos(πuK)+q4n−21−2q2n−1cos(πuK)+q4n−2sn(u,k)cn(u,k)=1√k′tan(πu2K)∞∏n=11−2q2ncos(πuK)+q4n1+2q2ncos(πuK)+q4n⇒sn(u,k)cn(u,k)tan(πu2K)=1√k′∞∏n=11−2q2ncos(πuK)+q4n1+2q2ncos(πuK)+q4n Taking limits as u→0 we get 2Kπ=1√k′∞∏n=11−2q2n+q4n1+2q2n+q4n=1√k′∞∏n=1(1−q2n1+q2n)2⇒2√k′Kπ=∞∏n=1(1−q2n1+q2n)2 Multiplying by √k′ (and its infinite product equivalent) we get 2k′Kπ=∞∏n=1(1−qn1+qn)2 Again 2Kπ=1√k′∞∏n=1(1−q2n1+q2n)2=A∞∏n=1(1−q2n1+q2n)2=∞∏n=1(1+q2n−11−q2n−1)2(1−q2n1+q2n)2⇒2Kπ=∞∏n=1{(1−q2n)(1+q2n−1)2}2 To summarize: sn(u,k)cn(u,k)=1√k′tan(πu2K)∞∏n=11−2q2ncos(πuK)+q4n1+2q2ncos(πuK)+q4ndn(u,k)=√k′∞∏n=11+2q2n−1cos(πuK)+q4n−21−2q2n−1cos(πuK)+q4n−22Kπ=∞∏n=1{(1−q2n)(1+q2n−1)2}22k′Kπ=∞∏n=1(1−qn1+qn)22√k′Kπ=∞∏n=1(1−q2n1+q2n)2k′=∞∏n=1(1−q2n−11+q2n−1)4 As we can see above the elliptic functions have been expressed as ratio of two infinite products. Each of the infinite products in numerator and denominator belongs to a class of functions known as the Theta Functions which will form the main topic of the next post. In that post we will also justify the results which we have obtained here (by assuming the desired convergence of the infinite products involved).
Print/PDF Version
0 comments :: Elliptic Functions: Infinite Products
Post a Comment