The theory of Macdonald integrals. II. Asymptotic expansions
A. A. Tuzhilin
Submitted 1967-01-01 | RussiaRxiv: ru-196701.85488 | Translated from Russian

Abstract

The article is a continuation of the author's work [1] and is devoted to the derivation of two types of asymptotic expansions for Macdonald integrals $n$ of order $M_n(x,y)$. Two expansions of the first type are asymptotic as $|y|\sqrt{1+x^2}\to\infty$. The first expansion is an expansion in terms of the asymptotic sequence $(H^1_{m+\alpha}(y\sqrt{1+x^2})/(y\sqrt{1+x^2})^m){m\in N}$, where $H^{(1)}\nu(z)$ is the Hankel function of the first kind; $\alpha$ is a real number; $N$ is the set of natural numbers. The second asymptotic expansion of the first type is an expansion in terms of the asymptotic sequence $((y\sqrt{1+x^2})^{-m}){m\in N}$, which follows from the first one. Both expansions are uniform with respect to $x$ for $|x|\ge\varepsilon>0$, where $\varepsilon$ is an arbitrarily small number. The asymptotic expansion of the second type is valid for $|y|\to\infty$ and in the case where the parameter $y(\sqrt{1+x^2}-1)$ remains finite as $|y|\to\infty$, i.e., $x$ must simultaneously tend to zero such that the expression $y(\sqrt{1+x^2}-1)$ remains finite. The presented expansion of the second type is an expansion in terms of the asymptotic sequence $(y^{-m})$. Bibliography: 4 items.

Full Text

Preamble

In this section, we refine the asymptotic properties of the function $M_n(x, y)$ introduced in \cite[1]. We consider the behavior of the function as $|y| \sqrt{1 + x^2} \to \infty$ and $|y| \sqrt{1 + x^2} \to 0$. Specifically, for $|x| > \epsilon > 0$, we establish estimates for the remainder terms and provide expansions in terms of the parameter $y$.

§ 1. Asymptotic Definitions and Notation

Let $x_0$ be a limit point in the domain $U$. We say that $\phi(x) = O(\psi(x))$ as $x \to x_0$ if there exists a constant $A$ such that $|\phi(x)| < A |\psi(x)|$ for all $x in a neighborhood $U_\delta$ of $x_0$. Similarly, $\phi(x) = o(\psi(x))$ if $\phi(x)/\psi(x) \to 0$ as $x \to x_0$.

An asymptotic series is defined as:
$$f(x) \sim \sum_{k=0}^{\infty} \phi_k(x), \quad x \to x_0 \tag{1.1}$$
if for every $n \in N$, the following condition holds:
$$f(x) = \sum_{k=0}^{n} \phi_k(x) + o(\phi_n(x)), \quad x \to x_0 \tag{1.2}$$
Equivalently, this can be expressed using the $O$-notation as:
$$f(x) = \sum_{k=0}^{n} \phi_k(x) + O(\phi_{n+1}(x)) \tag{1.3}$$
In the context of our analysis, we often encounter expansions of the form:
$$f(x) = \phi(x) + \psi(x) \left{ \sum_{k=0}^{n} a_k \phi_k(x) + o(\phi_n(x)) \right} \tag{1.4}$$
which implies the asymptotic equivalence:
$$f(x) \sim \phi(x) + \psi(x) \sum a_k \phi_k(x) \tag{1.5}$$

§ 2. Asymptotic Expansion for Large Arguments

We consider the domain $-\pi < \arg z < 2\pi$ as $|z| \to \infty$. Let the sequence be defined by $(H_{n+a}(z)/z^n){n \in N}$. We assume the property:
$$\frac{H
$$}(z)}{z^{n+1}} = o\left(\frac{H_{n+a}(z)}{z^n}\right) \tag{2.1
For the function $M_n(x, y)$, as $|y| \sqrt{1+x^2} \to \infty$, we derive the following representation:
$$M_n(x, y) \sim \frac{H_{n-1/2}^{(1)}(y)}{\sqrt{\pi} (1+x^2)^{n/2}} \sum_{k=0}^{\infty} \frac{(-1)^k \Gamma(k+1/2)}{2^k \Gamma(1/2)} \frac{|x|^{2k+1}}{(1+x^2)^k} \frac{H_{n-k-1}^{(1)}(y \sqrt{1+x^2})}{(y \sqrt{1+x^2})^{k+1}} \tag{2.2}$$
The remainder term $R_k(x, y)$ for the expansion in (2.2) is given by:
$$R_k(x, y) = \frac{(-1)^{k+1}}{2^{k+1} \Gamma(k+3/2)} \int_1^{\infty} H_{n-k-2}^{(1)}(y u \sqrt{1+x^2}) (u^2-1)^{k+3/2} du \tag{2.3}$$
For $x > 0$ and $x < 0$, the behavior of the integral depends on the sign of the argument. Using the properties of Hankel functions $H_n^{(1)}(z)$, we can bound the remainder $|R| < A$ for $|x| > \epsilon > 0$.

By applying the integral representation from \cite[1]:
$$M_n(x, y) = \int_0^{\text{arsh } x} H_n^{(1)}(y \cosh \xi) (\cosh \xi)^{n-1} d\xi \tag{2.5}$$
and substituting the series for the Hankel function, we obtain the coefficients for (2.2). Specifically, using the relation:
$$H_{n-1}^{(1)}(z) = \exp(iz) \sum \dots \tag{2.9}$$
we verify that the $k$-th term in (2.3) matches the required asymptotic order.

§ 3. Expansion for Small Arguments

As $|y| \to 0$, we investigate the behavior of $M_n(x, y)$ in the domain $0 < \arg y < \pi$. The expansion takes the form:
$$M_n(x, y) \sim \frac{H_n^{(1)}(y) \text{sign}(x)}{\sqrt{y (\sqrt{1+x^2}-1)}} \int e^{iu^2} du + \dots \tag{3.1}$$
where the coefficients $c_n$ are determined by the recurrence relations:
$$c_n = \frac{(-1)^l (2l+1)!! (2p-1)!! \Gamma(n+m-l)}{(2m+1)!! 2^{2p} (l+1-p)! p! (m-l)!} \tag{3.2}$$
For $x > 0$, we use the substitution $u = \sqrt{2y \sinh(\xi/2)}$. The integral (3.3) is then evaluated using the properties of the Gamma function and the asymptotic expansion of $H_n^{(1)}(z)$ as $z \to 0$.

Applying Lemma 2, we find that for $|x| > \epsilon > 0$:
$$R_k(x, y) = O\left(\frac{\exp(iy)}{y^{k+3}}\right) \tag{3.5}$$
The final asymptotic form for $M_n(x, y)$ as $|y| \sqrt{1+x^2} \to 0$ is:
$$M_n(x, y) \approx \frac{\text{sign}(x)}{\sqrt{\pi} (1+x^2)^{n/2-1}} \sum_{m=0}^k \frac{(n+m-1)!}{m! (n-m-1)!} \frac{1}{(2iy)^m} \tag{3.12}$$
This result is consistent with the general theory of confluent hypergeometric functions and the specific bounds established in \cite[4].

References

  1. T. A. A., Journal of Mathematical Physics, No. 10, 1967.
  2. D. A. A., Asymptotic Expansions, 1962.
  3. I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and Products, 1965.
  4. Higher Transcendental Functions, Vol. 2, 1966.

Submission history

The theory of Macdonald integrals. II. Asymptotic expansions