A series of secants

Jordan Bell
November 3, 2014

Let ℌ={τ∈ℂ:Im⁢τ>0}. Define C:ℌ→ℂ by

C⁢(τ)=2⁢∑n=-∞∞1eπ⁢i⁢n⁢τ+q-π⁢i⁢n⁢τ=∑n=-∞∞sec⁡π⁢n⁢τ,τ∈ℌ.

We take as granted that C is holomorphic on ℌ.

First we calculate the Fourier transform of x↦sech⁢π⁢x.11 1 Elias M. Stein and Rami Shakarchi, Complex Analysis, p. 81, Example 3.

Lemma 1.

For ξ∈R,

∫-∞∞e-2⁢π⁢i⁢ξ⁢x⁢sech⁢π⁢x⁢𝑑x=sech⁢π⁢ξ.
Proof.

Let ξ∈ℝ and define

f⁢(z)=e-2⁢π⁢i⁢z⁢ξcosh⁡π⁢z.

The poles of f are those z at which cosh⁡π⁢z=0, thus z=n⁢i+i2, n∈ℤ. Taking γR to be the contour going from -R to R, from R to R+2⁢i, from R+2⁢i to -R+2⁢i, and from -R+2⁢i to -R, the poles of f inside γR are i2 and 3⁢i2. Because (cosh⁡π⁢z)′=π⁢sinh⁡π⁢z, we work out

Resz=i/2⁢f⁢(z)=e-2⁢π⁢i⋅i2⁢ξπ⁢sinh⁡π⁢i2=eπ⁢ξπ⁢i⁢sin⁡π2=eπ⁢ξπ⁢i

and

Resz=3⁢i/2⁢f⁢(z)=e-2⁢π⁢i⋅3⁢i2⁢ξπ⁢sinh⁡π⁢3⁢i2=e3⁢π⁢ξπ⁢i⁢sin⁡3⁢π2=e3⁢π⁢ξ-π⁢i.

We bound the integrals on the vertical sides as follows. For z=-R+i⁢y,

|cosh⁡π⁢z|=|eπ⁢z+e-π⁢z|2≥||eπ⁢z|-|e-π⁢z||2=|e-R⁢π-eR⁢π|2=eR⁢π-e-R⁢π2,

and, for 0≤y≤2,

|e-2⁢π⁢i⁢z⁢ξ|=e2⁢π⁢y⁢ξ=e4⁢π⁢ξ.

For z=R+i⁢y,

|cosh⁡π⁢z|=|eπ⁢z+e-π⁢z|2≥||eπ⁢z|-|e-π⁢z||2=|eR⁢π-e-R⁢π|2=eR⁢π-e-R⁢π2,

and, for 0≤y≤2,

|e-2⁢π⁢i⁢z⁢ξ|=e2⁢π⁢y⁢ξ=e4⁢π⁢ξ.

Therefore

|∫-R-R+2⁢if⁢(z)⁢𝑑z|≤∫-R-R+2⁢i|f⁢(z)|⁢𝑑z≤2⋅e4⁢π⁢ξ⋅2eR⁢π-e-R⁢π=e4⁢π⁢ξeR⁢π-e-R⁢π

and likewise

|∫RR+2⁢if⁢(z)⁢𝑑z|≤e4⁢π⁢ξeR⁢π-e-R⁢π.

As R→∞, each of these tends to 0. Therefore,

∫-∞∞f⁢(z)⁢𝑑z+∫∞+2⁢i-∞+2⁢if⁢(z)⁢𝑑z=2⁢π⁢i⁢(eπ⁢ξπ⁢i+e3⁢π⁢ξ-π⁢i)=-2⁢e2⁢π⁢ξ⁢(eπ⁢ξ-e-π⁢ξ),

i.e.,

∫-∞∞f⁢(z)⁢𝑑z=∫-∞+2⁢i∞+2⁢if⁢(z)⁢𝑑z-2⁢e2⁢π⁢ξ⁢(eπ⁢ξ-e-π⁢ξ).

For the top horizontal side,

∫-R+2⁢iR+2⁢if⁢(z)⁢𝑑z =∫-RRe-2⁢π⁢i⁢(x+2⁢i)⁢ξcosh⁡(π⁢x+2⁢π⁢i)⁢𝑑x
=∫-RRe-2⁢π⁢i⁢x⁢ξ⁢e4⁢π⁢ξcosh⁡(π⁢x)⁢cosh⁡(2⁢π⁢i)+sinh⁡(π⁢x)⁢sinh⁡(2⁢π⁢i)⁢𝑑x
=e4⁢π⁢ξ⁢∫-RRe-2⁢π⁢i⁢x⁢ξcosh⁡π⁢x⁢𝑑x
=e4⁢π⁢ξ⁢∫-RRf⁢(x)⁢𝑑x.

Writing

I=∫-∞∞f⁢(z)⁢𝑑z,

this gives us

I=e4⁢π⁢ξ⁢I-2⁢e2⁢π⁢ξ⁢(eπ⁢ξ-e-π⁢ξ),

and so

I=-2⁢e2⁢π⁢ξ⁢eπ⁢ξ-e-π⁢ξ1-e4⁢π⁢ξ=2⁢eπ⁢ξ-e-π⁢ξe2⁢π⁢ξ-e-2⁢π⁢ξ=2⁢eπ⁢ξ-e-π⁢ξ(eπ⁢ξ-e-π⁢ξ)⁢(eπ⁢ξ+e-π⁢ξ)=sech⁢π⁢ξ,

which is what we wanted to show. ∎

Corollary 2.

For t>0 and a∈R,

∫-∞∞e-2⁢π⁢i⁢ξ⁢x⁢e-2⁢π⁢i⁢a⁢x⁢sech⁢π⁢xt⁢𝑑x=t⁢sech⁢(π⁢(ξ+a)⁢t),ξ∈ℝ.
Proof.
∫-∞∞e-2⁢π⁢i⁢x⁢ξ⁢e-2⁢π⁢i⁢a⁢x⁢sech⁢π⁢xt⁢𝑑x =∫-∞∞e-2⁢π⁢i⁢(ξ+a)⁢x⁢sech⁢π⁢xt⁢𝑑x
=t⁢∫-∞∞e-2⁢π⁢i⁢(ξ+a)⁢t⁢x⁢sech⁢π⁢x⁢𝑑x
=t⁢sech⁢(π⁢(ξ+a)⁢t).

∎

Theorem 3.

For all τ∈H,

C⁢(τ)=iτ⁢C⁢(-1τ).
Proof.

For f∈L1⁢(ℝ), we define f^:ℝ→ℂ by

f^⁢(ξ)=∫ℝe-2⁢π⁢i⁢ξ⁢x⁢f⁢(x)⁢𝑑x,ξ∈ℝ.

Following Stein and Shakarchi, for a>0, define 𝔉a to be the set of those functions f defined on some neighborhood of ℝ in ℂ such that f is holomorphic on the set {z∈ℂ:|Im⁢z|<a} and for which there is some A>0 such that

|f⁢(x+i⁢y)|≤A1+x2,x∈ℝ,|y|<a,

and we set 𝔉=⋃a>0𝔉a. The Poisson summation formula22 2 Elias M. Stein and Rami Shakarchi, Complex Analysis, p. 118, Theorem 2.4. states that for f∈𝔉,

∑n∈ℤf⁢(n)=∑n∈ℤf^⁢(n).

For z=x+i⁢y with |y|<12,

|sech⁢π⁢zt| =2|eπ⁢(x+i⁢y)-e-π⁢(x+i⁢y)|
≤2||eπ⁢(x+i⁢y)|-|e-π⁢(x+i⁢y)||
=2|eπ⁢x-e-π⁢x|
=sech⁢π⁢|x|.

Let t>0. Because the zeros of cosh⁡π⁢z are n⁢i+i2, n∈ℤ, the function f⁢(z)=sech⁢π⁢zt belongs to 𝔉t2. Corollary 2 with a=0 gives us

f^⁢(ξ)=t⁢sech⁢π⁢ξ⁢t,

so applying the Poisson summation formula we get

∑n∈ℤsech⁢π⁢nt=t⁢∑n∈ℤsech⁢π⁢n⁢t,

or,

∑n∈ℤsec⁡π⁢i⁢nt=t⁢∑n∈ℤsec⁡π⁢i⁢n⁢t,

i.e.,

C⁢(it)=t⁢C⁢(i⁢t).

For τ=i⁢t this reads

C⁢(τ)=iτ⁢C⁢(-1τ).

But τ↦C⁢(τ) and τ↦iτ⁢C⁢(-1τ) are holomorphic on ℌ, so by analytic continuation this identity is true for all τ∈ℌ. ∎

Theorem 4.
C⁢(1-1τ)∼4⁢τi⁢eπ⁢i⁢τ2,Im⁢τ→+∞.
Proof.

Let t>0 and define f⁢(z)=e-π⁢i⁢z⁢sech⁢π⁢zt, which we check belongs to 𝔉t2. Corollary 2 with a=12 tells us that for t>0,

f^⁢(ξ)=∫-∞∞e-2⁢π⁢i⁢ξ⁢x⁢e-π⁢i⁢x⁢sech⁢π⁢xt⁢𝑑x=t⁢sech⁢(π⁢(ξ+12)⁢t),ξ∈ℝ.

Thus the Poisson summation formula gives, as (-1)n=e-i⁢π⁢n,

∑n∈ℤ(-1)n⁢sech⁢π⁢nt=t⁢∑n∈ℤsech⁢(π⁢(n+12)⁢t),

or

∑n∈ℤ(-1)n⁢sec⁡π⁢i⁢nt=t⁢∑n∈ℤsec⁡(π⁢i⁢(n+12)⁢t).

For τ=i⁢t this reads

∑n∈ℤ(-1)n⁢sec⁡π⁢nτ=τi⁢∑n∈ℤsec⁡(π⁢(n+12)⁢τ).

Now,

sec⁡(π⁢n⁢(1-1τ))=1cos⁡π⁢n⁢cos⁡-π⁢nτ-sin⁡π⁢n⁢sin⁡-π⁢nτ=(-1)n⁢sec⁡π⁢nτ,

so the above states that for τ=i⁢t, t>0,

C⁢(1-1τ)=τi⁢∑n∈ℤsec⁡(π⁢(n+12)⁢τ). (1)

We assert that both sides of (1) are holomorphic on ℌ, and thus by analytic continuation that (1) is true for all τ∈ℌ.

Write τ=σ+i⁢t. For ν>0,

sec⁡π⁢ν⁢τ=2ei⁢π⁢ν⁢τ+e-i⁢π⁢ν⁢τ=2e-i⁢π⁢ν⁢τ⁢(e2⁢π⁢i⁢ν⁢τ+1)=2⁢ei⁢π⁢ν⁢τ⁢(1+O⁢(|e2⁢π⁢i⁢ν⁢τ|)),

or,

sec⁡π⁢ν⁢τ=2⁢ei⁢π⁢ν⁢τ+O⁢(|e3⁢π⁢i⁢ν⁢τ|).

Now,

|e3⁢π⁢i⁢τ2|=e-3⁢π⁢t2,

so,

sec⁡π⁢ν⁢τ=2⁢ei⁢π⁢ν⁢τ+O⁢(e-3⁢π⁢t2).

For ν<0,

sec⁡π⁢ν⁢τ=sec⁡(-π⁢ν⁢τ)=2⁢e-i⁢π⁢ν⁢τ+O⁢(e-3⁢π⁢t2).

For ν=12,

sec⁡π⁢ν⁢τ=2⁢ei⁢π⁢τ2+O⁢(e-3⁢π⁢t2),

and for ν=-12,

sec⁡π⁢ν⁢τ=2⁢ei⁢π⁢τ2+O⁢(e-3⁢π⁢t2).

It follows that

∑n∈ℤsec⁡(π⁢(n+12)⁢τ)=2⁢ei⁢π⁢τ2+2⁢ei⁢π⁢τ2+O⁢(e-3⁢π⁢t2)=4⁢ei⁢π⁢τ2+O⁢(e-3⁢π⁢t2).

Using this with (1) yields

C⁢(1-1τ)=4⁢τi⁢ei⁢π⁢τ2+O⁢(|τ|⁢e-3⁢π⁢t2),τ=σ+i⁢t,

proving the claim. ∎

Define θ:ℌ→ℂ by

θ⁢(τ)=∑n∈ℤeπ⁢i⁢n2⁢τ,τ∈ℌ.

By proving that Cθ2 is a modular form of weight 0, it follows that it is constant, and one thus finds that C=θ2.33 3 Elias M. Stein and Rami Shakarchi, Complex Analysis, p. 304. One reason that θ is significant is that, for q=ei⁢π⁢τ,

θ⁢(τ)2 =(∑n1∈ℤqn12)⁢(∑n2∈ℤqn22)
=∑(n1,n2)∈ℤ×ℤqn12+n22
=∑n=0∞r2⁢(n)⁢qn,

where r2⁢(n) denotes the number of ways that n can be expressed as a sum of two squares. We can write C⁢(τ) as

C⁢(τ) =2⁢∑n=-∞∞1qn+q-n
=2⁢∑n=-∞∞qn1+q2⁢n
=1+4⁢∑n=1∞qn1+q2⁢n
=1+4⁢∑n=1∞qn⁢1-q2⁢n1-q4⁢n
=1+4⁢∑n=1∞(qn1-q4⁢n-q3⁢n1-q4⁢n).

Therefore the identity θ⁢(τ)2=C⁢(τ) can be written as

∑n=0∞r2⁢(n)⁢qn=1+4⁢∑n=1∞(qn1-q4⁢n-q3⁢n1-q4⁢n).

We write

∑n=1∞qn1-q4⁢n=∑n=1∞qn⁢∑m=0∞(q4⁢n)m=∑n=1∞∑m=0∞qn⁢(4⁢m+1)=∑k=1∞a⁢(k)⁢qk,

where a⁢(k) denotes the number of divisors of k of the form 4⁢m+1, and

∑n=1∞q3⁢n1-q4⁢n=∑n=1∞q3⁢n⁢∑m=0∞(q4⁢n)m=∑n=1∞∑m=0∞qn⁢(4⁢m+3)=∑k=1∞b⁢(k)⁢qk,

where b⁢(k) denotes the number of divisors of k of the form 4⁢m+3. Thus for n≥1,

r2⁢(n)=4⁢(a⁢(n)-b⁢(n)).