I am currently trying to wrap my head around Fresnel propagation, and I understand it is mathematically linked to the Fractional Fourier Transform, but I'm having a hard time with the units and the interpretation. We can write out the equation for Fresnel propagation (ignoring the factors that ensure conservation of energy)

$$e^{i\pi f_y^2 \frac{z}{f}}\int{e^{-i 2\pi(f_y y-\frac{1}{2}f\frac{y^2}{z}) }g(y)dy}$$

where $f$ is the wavenumber in free space, $y$ is the coordinate in the plane we are propagating from, and $z$ is the propagation distance, and $f_y$ is the wavenumber in the $y$-direction.

The Fractional Fourier Transform, in contrast, can be written as (ignoring the thing out front that I believe is a normalization factor analagous to the $2 \pi$ in the regular FT):

$$ e^{i \pi u^2 cot(\alpha)}\int{e^{-i 2\pi(csc(\alpha)ux-\frac{cot(\alpha)}{2}x^2)}h(x)dx}$$

where I have tried to write the two as similarly to each other as I can. The problem I am having is that when I try to equate the arguments of the exponentials I get a set of equations which appears not to have a solution $$f_y^2 \frac{z}{f}=u^2cot(\alpha)$$ $$f_y y=csc(\alpha)u x$$ $$f \frac{y^2}{z}=cot(\alpha)x^2$$

When I try to solve them I get some nonsense like $cos^2(\alpha)=1$. Am I doing something wrong, and how does Fresnel Propagation map onto the Fractional Fourier Transform as it is typically defined?


I had not come across the fractional Fourier transform before but loking at the Wiki article suggests what is going on. Fresnel's transform provides a kind of representation of the action of the non-compact symplectic group that underlies Hamiltonian optics. (The "metaplectic" representation). Non compact means that the propagation distance $z$ is not a periodic variable: it can be as large as one likes.

The fractional fourier transform has a parameter $\alpha$ instead of $z$, and $\alpha$ is periodic, so that the fourth power of the usual Fourer is the identity opertor. Thus it is providing a representation of a compact group hence the ordinary $2\pi$-periodic trig functions. The algebra is indeed similar but I suspect one needs to make $\alpha$ complex to get the mapping between to two, so $\cot\alpha$ has to become $\coth \alpha$ and ${\rm cosec}\, \alpha $ becomes ${\rm cosech}\, \alpha$. Did you try this?

Note added: There is another possibility which I now think more likely. It comes from thinking of Fractional Fourier in terms of the Mehler Kernel. When one takes the coeffcient of the Hermite Polys as a pure phase this coincides with the propagator of the quantum harmonic oscillator. This gives a periodic transform in your $z$ direction but in its optics interpretation is is not free wave propagation. Instead it is propagation in an optic fibre with the a quadratic refractive index gradient $n \propto y^2$ so that the beam is periodically passes through a focus. This gradient alters the coefficient of the $y^2$ term in Fresnel so that it is no longer just $fy^2/z$ but is $fy^2/z+\beta$ for some parameter $\beta$ proprtional to the focusing effect. Perhaps this extra degree of freedom allows you to match all parameters?

  • $\begingroup$ I suspect the periodicity may be an issue that has to be resolved, but we know a priori that Fresnel propagation in the limit as z approaches infinity becomes Fraunhofer propagation (the fractional fourier transform becomes the Fourier Transform, and alpha becomes pi/2). So the range of z from 0 to infinity should map onto alpha from 0 to pi/2. I read a paper that claims to answer this question, but honestly I can't make heads or tails of it (Fractional Fourier Optics, 1995, Haldam Ozatkas). fy will also in general be bounded by the aperture of the optical system. $\endgroup$ – Jordan Jun 16 '19 at 19:09
  • $\begingroup$ I'll have a look at the Ozatkias paper tomorrow. I'll also try to work out the details via the interpretation of the FFT via Mehler evolution. $\endgroup$ – mike stone Jun 16 '19 at 19:11
  • $\begingroup$ You are also correct about the gradient index being a good basis for this transformation, there was a paper written on it in 1993: Mendlovic, David, and Haldun M. Ozaktas. "Fractional Fourier transforms and their optical implementation: I." JOSA A 10, no. 9 (1993): 1875-1881. $\endgroup$ – Jordan Jun 16 '19 at 19:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.