Chen, Hongmei and Chen, Min and Blower, Gordon and Chen, Yang (2017) Single uer MIMO system, Painleve transcendents and double scaling. Journal of Mathematical Physics, 58 (12): 123502. ISSN 0022-2488
chenchenblowerchenJmathPhysrev.pdf - Accepted Version
Available under License Creative Commons Attribution-NonCommercial.
Download (420kB)
Abstract
In this paper we study a particular Painlev\'e V (denoted ${\rm P_{V}}$) that arises from Multi-Input-Multi-Output (MIMO) wireless communication systems. Such a $P_V$ appears through its intimate relation with the Hankel determinant that describes the moment generating function (MGF) of the Shannon capacity. This originates through the multiplication of the Laguerre weight or the Gamma density $x^{\alpha} {\rm e}^{-x},\;x> 0,$ for $\alpha>-1$ by $(1+x/t)^{\lambda}$ with $t>0$ a scaling parameter. Here the $\lambda$ parameter ``generates" the Shannon capacity; see Yang Chen and Matthew McKay, IEEE Trans. IT, 58 (2012) 4594--4634. It was found that the MGF has an integral representation as a functional of $y(t)$ and $y'(t)$, where $y(t)$ satisfies the ``classical form" of $P_V$. In this paper, we consider the situation where $n,$ the number of transmit antennas, (or the size of the random matrix), tends to infinity, and the signal-to-noise ratio (SNR) $P$ tends to infinity, such that $s={4n^{2}}/{P}$ is finite. Under such double scaling the MGF, effectively an infinite determinant, has an integral representation in terms of a ``lesser" $P_{III}$. We also consider the situations where $\alpha=k+1/2,\;\;k\in \mathbb{N},$ and $\alpha\in\{0,1,2,\dots\}$ $\lambda\in\{1,2,\dots\},$ linking the relevant quantity to a solution of the two dimensional sine-Gordon equation in radial coordinates and a certain discrete Painlev\'e-II. \\ From the large $n$ asymptotic of the orthogonal polynomials, that appears naturally, we obtain the double scaled MGF for small and large $s$, together with the constant term in the large $s$ expansion. With the aid of these, we derive a number of cumulants and find that the capacity distribution function is non-Gaussian.