An asymptotic method for estimating integrals with highly oscillatory phases of the form ∫ A(x) e^{iλ φ(x)} dx for large parameter λ, focusing on contributions from points where the phase derivative φ'(x) vanishes (stationary points).

Principle

The dominant contributions to an oscillatory integral as the frequency grows come from neighborhoods of stationary points of the phase; elsewhere rapid oscillation causes cancellation. Local quadratic approximations of φ near stationary points give leading asymptotic terms.

Demonstration

Estimate I(λ)=∫_{−1}^1 e^{iλ x^2} dx for large λ by locating the stationary point at x=0, approximating φ(x)=x^2 ≈ x^2, and applying the Gaussian integral asymptotic to get I(λ) ~ e^{iπ/4} √(π/λ) as leading behavior.

Misapplication

Applying the stationary phase formula when stationary points are degenerate (higher-order vanishing), when amplitude A has singularities at stationary points, or neglecting boundary contributions in finite-interval integrals leads to incorrect asymptotics.

Consequence

Yields explicit leading-order asymptotic expansions for oscillatory integrals, clarifies phase-driven behavior in wave propagation and spectral problems, and offers computationally efficient approximations in high-frequency regimes.

Reversal

Instead of isolating stationary-phase neighborhoods, one could average or numerically sample the integrand globally; this reversal misses analytic asymptotic structure and often loses accuracy in the λ→∞ regime.

Boundary

Requires smooth phase and amplitude, isolated nondegenerate stationary points or explicitly treated degeneracies, and control of endpoints; fails or needs modification for non-smooth φ, dense stationary sets, or when contributions from saddle coalescence dominate.

Semantic Tension

Closely related to the method of steepest descent and to Fourier transform asymptotics: stationary phase emphasizes real stationary points and oscillatory cancellation, while steepest descent often uses complex contour deformation to capture equivalent contributions.

Synthesis

The method of stationary phase isolates stationary points of the phase and approximates the integrand locally to convert rapid oscillation into computable Gaussian-type leading terms, producing reliable asymptotic expansions for large frequency parameters.