Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > comp.soft-sys.math.mathematica > #10440 > unrolled thread

Theta function integration bug

Started bySebastian Meznaric <meznaric@gmail.com>
First post2012-06-13 08:57 +0000
Last post2012-06-15 19:29 +0000
Articles 4 — 3 participants

Back to article view | Back to comp.soft-sys.math.mathematica


Contents

  Theta function integration bug Sebastian Meznaric <meznaric@gmail.com> - 2012-06-13 08:57 +0000
    Re: Theta function integration bug Robert Miller <robertelmiller@gmail.com> - 2012-06-14 09:32 +0000
    Re: Theta function integration bug derivatorb@gmail.com - 2012-06-14 09:33 +0000
    Re: Theta function integration bug Sebastian Meznaric <meznaric@gmail.com> - 2012-06-15 19:29 +0000

#10440 — Theta function integration bug

FromSebastian Meznaric <meznaric@gmail.com>
Date2012-06-13 08:57 +0000
SubjectTheta function integration bug
Message-ID<jr9km3$3kp$1@smc.vnet.net>
The bug can be found in the following analytic integral:
Integrate[
 EllipticTheta[3, \[Phi], \[Alpha]] Exp[I \[Phi]], {\[Phi], 0, 
  2 \[Pi]}]

Mathematica evaluates the integral to 0, which is not correct, which can be seen considering that for small value of alpha, the elliptic theta is close to the delta function. Numerical integration also confirms that this integral is non-zero. 

[toc] | [next] | [standalone]


#10455

FromRobert Miller <robertelmiller@gmail.com>
Date2012-06-14 09:32 +0000
Message-ID<jrcb3r$drd$1@smc.vnet.net>
In reply to#10440
Look at the series expansion of EllipticTheta[3,\[Phi], \[Alpha]] in the Documentation Center or find it on the web.

EllipticTheta[3, \[Phi], \[Alpha]]=1 + 2*Sum[\[Alpha]^n^2*Cos[2*n*\[Phi]], {n, 1, Infinity}]

Clear from this that if you multiply by E^I\[Phi]and perform your integral, every term in the series is equal to zero. I don't think there is a bug.

Robert
On Wednesday, June 13, 2012 1:57:39 AM UTC-7, Sebastian Meznaric wrote:
> The bug can be found in the following analytic integral:
> Integrate[
>  EllipticTheta[3, \[Phi], \[Alpha]] Exp[I \[Phi]], {\[Phi], 0, 
>   2 \[Pi]}]
> 
> Mathematica evaluates the integral to 0, which is not correct, which can be seen considering that for small value of alpha, the elliptic theta is close to the delta function. Numerical integration also confirms that this integral is non-zero.

[toc] | [prev] | [next] | [standalone]


#10457

Fromderivatorb@gmail.com
Date2012-06-14 09:33 +0000
Message-ID<jrcb5o$ds5$1@smc.vnet.net>
In reply to#10440
El mi=E9rcoles, 13 de junio de 2012 10:57:39 UTC+2, Sebastian Meznaric  escribiF3:> The bug can be found in the following analytic integral:
> Integrate[
>  EllipticTheta[3, \[Phi], \[Alpha]] Exp[I \[Phi]], {\[Phi], 0,
>   2 \[Pi]}]
>
> Mathematica evaluates the integral to 0, which is not correct, which can be seen considering that for small value of alpha, the elliptic theta is close to the delta function. Numerical integration also confirms that this integral is non-zero.

>From the definition
1 + 2 Sum[\[Alpha]^(n^2) Cos[2 n \[Phi]], {n, 1, Infinity}],
it is clear that
EllipticTheta[3, \[Phi], \[Alpha]] 
contains only Fourier modes
E^(I k \[Phi]
with even k. So the integral you propose is actually zero.
When \[Alpha] is zero EllipticTheta[3, \[Phi], \[Alpha]]  is equal to constant unity.

[toc] | [prev] | [next] | [standalone]


#10479

FromSebastian Meznaric <meznaric@gmail.com>
Date2012-06-15 19:29 +0000
Message-ID<jrg2f6$si7$1@smc.vnet.net>
In reply to#10440
On Wednesday, June 13, 2012 9:57:39 AM UTC+1, Sebastian Meznaric wrote:
> The bug can be found in the following analytic integral:
> Integrate[
>  EllipticTheta[3, \[Phi], \[Alpha]] Exp[I \[Phi]], {\[Phi], 0, 
>   2 \[Pi]}]
> 
> Mathematica evaluates the integral to 0, which is not correct, which can be seen considering that for small value of alpha, the elliptic theta is close to the delta function. Numerical integration also confirms that this integral is non-zero.

I was evaluating the integral of EllipticTheta[3, (\[Phi] - \[Pi])/2, E^(-(\[Sigma]^2/2))]/(2 \[Pi]) over 0 to 2Pi. You can show that this is equal to e^{-\sigma^2 / 2} as follows. 

Start with the sum  \sum_{k=-\infty}^{\infty} \int_0^{2\pi} \frac{1}{\sigma \sqrt{2 \pi}} e^{-(\phi+2\pi k)^2/2\sigma^2} e^{i \phi} d\phi.

You can exchange integral and sum by Weierstrass M-test of uniform convergence. Now the sum evaluates, according to Mathematica, to (EllipticTheta[3, \[Phi]/2, 
  E^(-(\[Sigma]^2/2))])/(2 \[Pi]) for positive \sigma. So the above is the same as the integral of the Elliptic Theta function. Now if you do the integration term by term, on the other hand, you get
\sum_{k=-\infty}^{\infty} \frac{1}{2} e^{-2i\pi i k - \sigma^2/2}((E_f(\frac{2(k+1)\pi - i \sigma^2}{\sqrt{2}\sigma}) - (E_f(\frac(2k\pi - i \sigma^2}{\sqrt{2}\sigma}), where E_f is the error function. 

This is a cascading sum of the error functions that evaluates to e^{-\sigma^2/2}. 

There are also physical reasons why the above integral should not vanish.

[toc] | [prev] | [standalone]


Back to top | Article view | comp.soft-sys.math.mathematica


csiph-web