Asymptotic expansions for oscillatory integrals using inverse functions (Q2391026): Difference between revisions

The authors consider asymptotic expansions of finite oscillatory integrals of the form \[ F(k):=\int_a^b f(x)e^{ikg(x)}dx, \quad k\to\infty. \] The functions \(f\) and \(g\) are real for real \(x\), analytic in \((a,b)\) and may have algebraic singularities at either of both interval end points. Under certain monotonicity conditions over \(g\), the authors introduce the change of variable \(g(x)=t\) and then, in the new variable \(t\), the integral becomes a finite Fourier transform: \[ \int_A^B \Phi(t)e^{ikt}dt, \quad \Phi(t):=f(g^{-1}(t))(g^{-1})'(t), \] where \(A:=g(a)\), \(B:=g(b)\) and \(g^{-1}\) denotes the inverse function of \(g\): \(x=g^{-1}(t)\). A straightforward application of the well known asymptotic expansion of a Fourier transform led the authors to derive an asymptotic expansion of \(F(k)\) in inverse powers of \(k\). The coefficients \(\varphi_n\) of this expansion are given in terms of the coefficients of the asymptotic expansion of \(\Phi(t)\) at \(t=A\). The definition of this function involves an inverse function and a composite function. The use of Lagrange-Bürmann's and Faá di Bruno's formulas led the authors to writhe the coefficients \(\varphi_n\) in terms of the coefficients of the asymptotic expansions of \(f(x)\) and \(g(x)\) at \(x=a\).