Connection between Theta Functions and Hypergeometric Functions
Let's recall the Gauss Transformation formula from an earlier post: F(a,b;2b;4x(1+x)2)=(1+x)2aF(a,a−b+12;b+12;x2) where F is the hypergeometric function 2F1. Putting a=b=1/2 we get 2F1(12,12;1;4x(1+x)2)=(1+x)2F1(12,12;1;x2) or 2F1(12,12;1;1−(1−x1+x)2)=(1+x)2F1(12,12;1;x2)If we set 1−x1+x=ϕ2(−q)ϕ2(q) where ϕ(q)=∑∞−∞qn2 is one of the theta functions defined by Ramanujan (see this post) so that x=ϕ2(q)−ϕ2(−q)ϕ2(q)+ϕ2(−q)=4qψ2(q4)ϕ2(q2)⇒1+x=2ϕ2(q)ϕ2(q)+ϕ2(−q)=ϕ2(q)ϕ2(q2) and x2=16q2ψ4(q4)ϕ4(q2)=ϕ4(q2)−ϕ4(−q2)ϕ4(q2)=1−ϕ4(−q2)ϕ4(q2) Therefore we finally obtain: 2F1(12,12;1;1−ϕ4(−q)ϕ4(q))=ϕ2(q)ϕ2(q2)2F1(12,12;1;1−ϕ4(−q2)ϕ4(q2)) The beauty of this formula is that it can be applied recursively. The hypergeometric function on the right is the same as that on the left except that the parameter q is replaced by q2. Also the first factor on the right consists of parameter q in numerator and q2 in denominator so that the recursion can be utilized quite nicely to lead us to the following result: 2F1(12,12;1;1−ϕ4(−q)ϕ4(q))=ϕ2(q)ϕ2(q2n)2F1(12,12;1;1−ϕ4(−q2n)ϕ4(q2n)) When n→∞,q2n→0 and therefore ϕ(q2n)→1,ϕ(−q2n)→1 so that we obtain ϕ2(q)=2F1(12,12;1;1−ϕ4(−q)ϕ4(q)) We can use the Gauss transformation formula again by setting x=ϕ2(−q)ϕ2(q) and noting that 4x(1+x)2=4ϕ2(−q)ϕ2(q)ϕ4(q)(ϕ2(q)+ϕ2(−q))2=ϕ2(q)ϕ2(−q)ϕ4(q2)=ϕ4(−q2)ϕ4(q2) and 1+x=2ϕ2(q2)ϕ2(q) Thus we arrive at 2F1(12,12;1;ϕ4(−q2)ϕ4(q2))=2ϕ2(q2)ϕ2(q)2F1(12,12;1;ϕ4(−q)ϕ4(q)) or 2F1(12,12;1;ϕ4(−q)ϕ4(q))=ϕ2(q)2ϕ2(q2)2F1(12,12;1;ϕ4(−q2)ϕ4(q2)) By applying recursion we get 2F1(12,12;1;ϕ4(−q)ϕ4(q))=ϕ2(q)2nϕ2(q2n)2F1(12,12;1;ϕ4(−q2n)ϕ4(q2n)) Dividing (1) by (3) and replacing 2n by m we get 2F1(12,12;1;1−ϕ4(−q)ϕ4(q))2F1(12,12;1;ϕ4(−q)ϕ4(q))=m2F1(12,12;1;1−ϕ4(−qm)ϕ4(qm))2F1(12,12;1;ϕ4(−qm)ϕ4(qm)) Mutiplying by −π and taking exponentials on both sides we get F(ϕ4(−q)ϕ4(q))={F(ϕ4(−qm)ϕ4(qm))}m where m is a positive integral power of 2 and F(x) represents the Ramanujan's function defined in previous post.
The same equation (4) on taking reciprocals leads us to {F(1−ϕ4(−q)ϕ4(q))}m=F(1−ϕ4(−qm)ϕ4(qm))
The Inversion Fomula
Using the above result we can prove the fundamental inversion formula F(1−ϕ4(−q)ϕ4(q))=q Clearly if we set xm=ϕ4(−qm)ϕ4(qm) then xm→1 as m→∞ and hence 1−xm→0. Clearly then we have F(1−xm)∼1−xm16 and hence F(1−ϕ4(−q)ϕ4(q))=m√F(1−xm)=limm→∞m√F(1−xm)=limm→∞m√1−xm16=limm→∞m√1−xm=A Then we have logA=limm→∞1mlog(1−xm)=limm→∞1mlog(1−ϕ4(−qm)ϕ4(qm))=limm→∞1mlog(ϕ4(qm)−ϕ4(−qm)ϕ4(qm))=limm→∞1mlog(16qmψ4(q2m)ϕ4(qm))=limm→∞logq+log16m+log(ψ4(q2m)ϕ4(qm))=logq And thus we have A=q and the inversion formula is established. In other words if x=1−ϕ4(−q)ϕ4(q) then q=F(x) Using (6) in (2) we get ϕ2{F(1−ϕ4(−q)ϕ4(q))}=2F1(12,12;1;1−ϕ4(−q)ϕ4(q)) or ϕ2(F(x))=2F1(12,12;1;x) where x=1−ϕ4(−q)ϕ4(q) Using the Ramanujan's notation x,y,z we now have ϕ(F(x))=ϕ(e−y)=√zTransformation Formula for ϕ(q)
If we have a look at the definitions of x,y and F(x) then we can see that logF(x)=−y and logF(x)⋅logF(1−x)=π2 so that logF(1−x)=−π2/y or F(1−x)=e−π2/y Let α,β be positive numbers such that αβ=π and let y=α2 so that π2/y=β2. Then we have ϕ2(F(1−x))ϕ2(F(x))=2F1(12,12;1;1−x)2F1(12,12;1;x)=yπ=α2αβ=αβ or ϕ2(e−β2)ϕ2(e−α2)=αβ or √αϕ(e−α2)=√βϕ(e−β2) In the classical notation this is √sθ3(e−πs)=θ3(e−π/s) (see this post) and therefore the above constitutes a proof of the transformation formula for theta functions without the use of Poisson Summation formula.In the next post we will have a look at various identities relating the theta functions of q and qn and these will be finally transcribed in the form of modular equations. Most of the identities will be derived using Lambert series for the theta functions.
Print/PDF Version
0 comments :: Elementary Approach to Modular Equations: Ramanujan's Theory 3
Post a Comment