MathGroup Archive 2005

[Date Index] [Thread Index] [Author Index]

Search the Archive

Re: Gaussian sums (Was: Speeding up simple Mathematica expressions?)

  • To: mathgroup at
  • Subject: [mg63307] Re: Gaussian sums (Was: Speeding up simple Mathematica expressions?)
  • From: Paul Abbott <paul at>
  • Date: Fri, 23 Dec 2005 05:08:31 -0500 (EST)
  • Organization: The University of Western Australia
  • References: <do8ioc$rvd$> <dodec4$65o$>
  • Sender: owner-wri-mathgroup at

In article <dodec4$65o$1 at>, AES <siegman at> 

> I originally wrote
> >  I'm evaluating a series of functions of which a typical example is 
> >  
> >  f[a_, x_] := Sum[
> >  Exp[-(Pi a)^2 n^2 -
> >  ((x - n Sqrt[1 - (Pi^2 a^4)])/a)^2],
> >  {n, -Infinity, Infinity}];
> and Paul Abbott <paul at>   replied:
> >  The Mathematica Journal 9(4) under the heading "Sum of Gaussians". 
> >  After completing the square of the exponent, such sums can be 
> >  expressed in closed form in terms of EllipticTheta functions. See 
> >  
> > 

Actually, that is a link to the definition of EllipticTheta[3,z]. I 
actually meant to post the link to

> >  For this particular example,
> >  
> >  f[a_, x_] = a Sqrt[Pi] E^(-(Pi a x)^2) * 
> >  EllipticTheta[3, -Pi Sqrt[1 - a^4 Pi^2] x, E^(-(Pi a)^2)] 
> Thanks for this, and for other replies which pointed me to NSum (now if 
> I just had any idea of what the EllipticTheta function is, or does!).
> In any event, it's possible some of those who replied would be 
> interested in where the interest in this topic came from.  There's a 
> coupled fiber optics concept of current practical interest which leads 
> to three such sums, namely
> f[a_, x_] = Sum[
> Exp[ -(Pi a)^2 n^2 - ((x - n Sqrt[1 - (Pi^2 a^4)])/a)^2], 
> {n, -Infinity, Infinity}];
> g[a_, x_] = Sqrt[ Pi a^2] Sum[
> Exp[ -(Pi a)^2 n^2 - (Pi a)^2 x^2 - I 2 Pi n Sqrt[1 - (Pi^2 a^4)] x], 
> {n, -Infinity, Infinity}];
> h[a_, x_] := Sqrt[ Pi a^2] Sum[
> Exp[ -(n/a)^2 - Pi^2 a^2 (x - (I n Sqrt[1 - (Pi^2 a^4)])/(Pi a^2))^2], 
> {n, -Infinity, Infinity}];
> with the properties that
> 0)  The functions g and h are obviously the same except for completing 
> the square in the exponent;

Actually, the summands are still different after completing the square 
-- symmetry of the summand under n -> -n implies symmetry under x -> -x.

> 1)  The function pairs f and g or h are Fourier transforms of each 
> other; and
> 2)  All three functions are also *identical* (but with no obvious way to 
> convert f into g or h by algebraic methods).
> Ergo, f and g or h are apparently a new (or at least not widely 
> recognized) family of self-Fourier-transforming functions; the 
> references are:
> [1]   C. J. Corcoran and K. A. Pasch, "Self-Fourier functions and 
> coherent laser combination," J. Phys. A, vol. 37, pp. L461--L469,  
> (2004).
> [2]   C. J. Corcoran and F. Durville, "Experimental demonstration of a 
> phase-locked laser array using a self-Fourier cavity," Appl. Phys. 
> Lett., vol. 86, pp. 201118,  (16 May 2005).

I note that there is a small entry on this in Section 1.8 of the 
Symbolics Volume of Michael Trott's Mathematica Guidebook.
> A colleague Adnah Kostenbauder has pointed out that this seems to be a 
> version of "Jacobi's imaginary transformation" given in Section 21.51 of 
> Whittaker and Watson.  Presumably it also has a connection to some 
> obscure property of the EllipticTheta functions.

Not an obscure property. It is a basic transformation. See

> In physical terms f corresponds an array of narrow, parallel, 
> transversely but equally displaced gaussian beams with gaussianly 
> decreasing amplitude across the array; g represents a set of wider, 
> increasingly tilted gaussian beams all convering onto a common spot; and 
> h has the appearance of an array of wide, nominally parallel gaussian 
> beams with equal but imaginary-valued transverse displacements.

Most interesting.


Paul Abbott                                      Phone:  61 8 6488 2734
School of Physics, M013                            Fax: +61 8 6488 1014
The University of Western Australia         (CRICOS Provider No 00126G)    

  • Prev by Date: Re: Multiple integration of UnitStep fails
  • Next by Date: Re: Re: NDSolve
  • Previous by thread: Re: What we from 1^Infinity, Infinity^0, and similar stuff
  • Next by thread: Re: Gaussian sums (Was: Speeding up simple Mathematica expressions?)