A Question from Murray - Asymptotic Analysis

77 Views Asked by At

I'm stuck on one of the questions related to the method of stationary phase in Murray's book on Asymptotic analysis. The question is as follows;

If $h(t)$ has a single stationary point at $t_0$, where $a < t_0 < b$, and $h'(t_0) = h''(t_0) = 0$, $h'''(t_0) > 0$, show that $$\int_a^b e^{i \lambda h(t)} dt \approx \Gamma\left(\frac43\right) \left\lbrace \frac{48}{\lambda h'''(t_0)}\right\rbrace^{\frac13} e^{i \lambda h(t_0)} \cos(\frac{\pi}{6}) \text{ as } \lambda \to \infty.$$

I don't really think this is meant to be particularly hard, but I'm kind of confused by a few of the terms in that expression, especially the gamma function term and the cosine term.

Regardless, I started with the Taylor expansion of $h(t)$; $$h(t) \approx h(t_0) + \frac{1}{3!}h'''(t_0)(t-t_0)^3$$

Which then gives;

$$\int_a^b e^{i \lambda h(t)} dt \approx \int_{t_0 - \epsilon}^{t_0 + \epsilon} e^{i \lambda \left[h(t_0) + \frac16 h'''(t_0) (t-t_0)^3 \right]}dt$$

Expanding out the right hand side integral, we get (I think...); $$ e^{i \lambda h(t_0)} \int_{t_0 - \epsilon}^{t_0 + \epsilon} e^{i \lambda \left[\frac16 h'''(t_0) (t-t_0)^3\right]} dt$$

At this stage, I'm a bit confused with what to do. In class, we're generally left with a squared term in the integral, to which we make a substitution $\pm s^2 = \text{stuff inside the square brackets multiplied by } \lambda$, but in this case, I don't think that that will work. We've taken the substitution to be $s^2$, because then we get an integral that is easy to evaluate.

However, as I said, I'm a bit unsure as to how I should proceed. I've tried a couple of substitutions (such as $s^2$, $s^3$, etc...), but none of these really seem to give me what I want.

Any hints as to how to proceed would be very much appreciated :).