Paper 1, Section II, A

Asymptotic Methods | Part II, 2009

Consider the integral

I(λ)=0Aeλtf(t)dt,A>0I(\lambda)=\int_{0}^{A} \mathrm{e}^{-\lambda t} f(t) d t, \quad A>0

in the limit λ\lambda \rightarrow \infty, given that f(t)f(t) has the asymptotic expansion

f(t)n=0antnβf(t) \sim \sum_{n=0}^{\infty} a_{n} t^{n \beta}

as t0+t \rightarrow 0_{+}, where β>0\beta>0. State Watson's lemma.

Now consider the integral

J(λ)=abeλϕ(t)F(t)dtJ(\lambda)=\int_{a}^{b} \mathrm{e}^{\lambda \phi(t)} F(t) d t

where λ1\lambda \gg 1 and the real function ϕ(t)\phi(t) has a unique maximum in the interval [a,b][a, b] at cc, with a<c<ba<c<b, such that

ϕ(c)=0,ϕ(c)<0\phi^{\prime}(c)=0, \phi^{\prime \prime}(c)<0

By making a monotonic change of variable from tt to a suitable variable ζ\zeta (Laplace's method), or otherwise, deduce the existence of an asymptotic expansion for J(λ)J(\lambda) as λ\lambda \rightarrow \infty. Derive the leading term

J(λ)eλϕ(c)F(c)(2πλϕ(c))12J(\lambda) \sim \mathrm{e}^{\lambda \phi(c)} F(c)\left(\frac{2 \pi}{\lambda\left|\phi^{\prime \prime}(c)\right|}\right)^{\frac{1}{2}}

The gamma function is defined for x>0x>0 by

Γ(x+1)=0exp(xlogtt)dt\Gamma(x+1)=\int_{0}^{\infty} \exp (x \log t-t) d t

By means of the substitution t=xst=x s, or otherwise, deduce Stirling's formula

Γ(x+1)x(x+12)ex2π(1+112x+)\Gamma(x+1) \sim x^{\left(x+\frac{1}{2}\right)} \mathrm{e}^{-x} \sqrt{2 \pi}\left(1+\frac{1}{12 x}+\cdots\right)

as xx \rightarrow \infty

Typos? Please submit corrections to this page on GitHub.