Asymptotic expansion

In mathematics an asymptotic expansion, asymptotic series or Poincaré expansion (after Henri Poincaré) is a formal series of functions which has the property that truncating the series after a finite number of terms provides an approximation to a given function as the argument of the function tends towards a particular, often infinite, point.

If φn is a sequence of continuous functions on some domain, and if L is a (possibly infinite) limit point of the domain, then the sequence constitutes an asymptotic scale if for every n, \varphi_{n%2B1}(x) = o(\varphi_n(x)) \ (x \rightarrow L). If f is a continuous function on the domain of the asymptotic scale, then f has an asymptotic expansion of order N with respect to the scale as a formal series \sum_{n=0}^\infty a_n \varphi_{n}(x) if

f(x) - \sum_{n=0}^{N-1} a_n \varphi_{n}(x) = O(\varphi_{N}(x)) \ (x \rightarrow L)

or

f(x) - \sum_{n=0}^{N-1} a_n \varphi_{n}(x) = o(\varphi_{N-1}(x)) \ (x \rightarrow L).

If one or the other holds for all N, then we write

f(x) \sim \sum_{n=0}^\infty a_n \varphi_n(x) \ (x \rightarrow L).

See asymptotic analysis and big O notation for the notation.

The most common type of asymptotic expansion is a power series in either positive or negative powers. Methods of generating such expansions include the Euler–Maclaurin summation formula and integral transforms such as the Laplace and Mellin transforms. Repeated integration by parts will often lead to an asymptotic expansion.

Since a convergent Taylor series fits the definition of asymptotic expansion as well, the phrase "asymptotic series" usually implies a non-convergent series. Despite non-convergence, the asymptotic expansion is useful when truncated to a finite number of terms. Typically, the best approximation is given when the series is truncated at the smallest term. This way of optimally truncating an asymptotic expansion is known as superasymptotics.[1] The error is then typically of the form \sim\exp\left(-\frac{c}{\epsilon}\right) where ε is the expansion parameter. The error is thus beyond all orders in the expansion parameter. It is possible to improve on the superasymptotic error, e.g. by employing resummation methods such as Borel resummation to the divergent tail. Such methods are often referred to as hyperasymptotic approximations.

Examples of asymptotic expansions

\frac{e^x}{x^x \sqrt{2\pi x}} \Gamma(x%2B1) \sim 1%2B\frac{1}{12x}%2B\frac{1}{288x^2}-\frac{139}{51840x^3}-\cdots \ (x \rightarrow \infty)
xe^xE_1(x) \sim \sum_{n=0}^\infty \frac{(-1)^nn!}{x^n} \ (x \rightarrow \infty)
\zeta(s) \sim \sum_{n=1}^{N-1}n^{-s} %2B \frac{N^{1-s}}{s-1} %2B N^{-s} \sum_{m=1}^\infty \frac{B_{2m} s^{\overline{2m-1}}}{(2m)! N^{2m-1}}
where B_{2m} are Bernoulli numbers and s^{\overline{2m-1}} is a rising factorial. This expansion is valid for all complex s and is often used to compute the zeta function by using a large enough value of N, for instance N > |s|.
\sqrt{\pi}x e^{x^2}{\rm erfc}(x) \sim 1%2B\sum_{n=1}^\infty (-1)^n \frac{(2n)!}{n!(2x)^{2n}}.

Detailed example

Asymptotic expansions often occur when an ordinary series is used in a formal expression that forces the taking of values outside of its domain of convergence. Thus, for example, one may start with the ordinary series

\frac{1}{1-w}=\sum_{n=0}^\infty w^n.

The expression on the left is valid on the entire complex plane w\ne 1, while the right hand side converges only for |w|< 1. Multiplying by e^{-w/t} and integrating both sides yields

\int_0^\infty \frac{e^{-w/t}}{1-w}\, dw = \sum_{n=0}^\infty t^{n%2B1} \int_0^\infty e^{-u} u^n\, du,

after the substitution u=w/t on the right hand side. The integral on the left hand side, understood as a Cauchy principal value, can be expressed in terms of the exponential integral. The integral on the right hand side may be recognized as the gamma function. Evaluating both, one obtains the asymptotic expansion

e^{-1/t}\; \operatorname{Ei}\left(\frac{1}{t}\right) = \sum _{n=0}^\infty n! \; t^{n%2B1}.

Here, the right hand side is clearly not convergent for any non-zero value of t. However, by truncating the series on the right to a finite number of terms, one may obtain a fairly good approximation to the value of \operatorname{Ei}(1/t) for sufficiently small t. Substituting x=-1/t and noting that \operatorname{Ei}(x)=-E_1(-x) results in the asymptotic expansion given earlier in this article.

References

  1. ^ Boyd, John P. (1999). "The Devil's Invention: Asymptotic, Superasymptotic and Hyperasymptotic Series". Acta Applicandae Mathematicae 56 (1): 1–98. doi:10.1023/A:1006145903624.