Improper Integrals (Rigorous)

Core Idea

An improper integral extends the Riemann integral to unbounded intervals or unbounded integrands by taking limits. For infinite intervals, ∫ₐ^∞ f(x) dx = lim_{t→∞} ∫ₐᵗ f(x) dx; for unbounded integrands near a point c, ∫ₐᵇ f(x) dx = lim_{ε→0⁺} ∫ₐ^{c−ε} f(x) dx + lim_{ε→0⁺} ∫_{c+ε}ᵇ f(x) dx. The integral converges if these limits exist and are finite. Convergence criteria mirror those for series: comparison tests, limit comparison, and absolute convergence all apply. An integral can converge conditionally (like ∫₁^∞ sin(x)/x dx) without converging absolutely. These integrals arise naturally in probability, Fourier analysis, and Laplace transforms.

How It's Best Learned

Work through the classic examples: ∫₁^∞ 1/xᵖ dx (converges iff p > 1), then ∫₀¹ 1/xᵖ dx (converges iff p < 1). These two cases build the intuition that convergence depends on how fast the integrand decays or blows up relative to the interval.

Common Misconceptions

Students sometimes evaluate improper integrals by plugging in ∞ directly, skipping the limit process. This can produce correct-looking answers but obscures conditional convergence issues. Also, the two limits in a doubly improper integral must be taken independently—they cannot be combined into a single symmetric limit.

Explainer

The Riemann integral ∫ₐᵇ f(x) dx, as you studied it, requires both a finite interval [a, b] and a bounded integrand f. Improper integrals extend this framework to two situations where these conditions fail: integrating over an infinite interval (like ∫₁^∞ 1/x² dx) or integrating a function with an unbounded singularity (like ∫₀¹ 1/√x dx). In both cases, the extension is made through limits — you compute a proper Riemann integral on a truncated domain and then take a limit as the domain grows to its full extent.

For integrals over infinite intervals, the definition is ∫ₐ^∞ f(x) dx = lim_{t→∞} ∫ₐᵗ f(x) dx. The integral converges if this limit exists and is finite; otherwise it diverges. The classic reference is ∫₁^∞ 1/xᵖ dx, which converges if and only if p > 1. When p = 2, the antiderivative is −1/x, and the limit gives ∫₁^∞ 1/x² dx = lim_{t→∞} (1 − 1/t) = 1. When p = 1, the antiderivative is ln(x), and lim_{t→∞} ln(t) = ∞, so the integral diverges. For integrals with singularities, a similar limit handles the blow-up: ∫₀¹ 1/xᵖ dx = lim_{ε→0⁺} ∫_ε¹ 1/xᵖ dx, which converges if and only if p < 1. The two conditions (p > 1 for infinity, p < 1 for zero) are complementary, reflecting the different nature of the two problems.

A critical subtlety arises with doubly improper integrals like ∫₋∞^∞ f(x) dx. The definition requires two independent limits: ∫₋∞^∞ f(x) dx = lim_{s→−∞} ∫_s⁰ f(x) dx + lim_{t→∞} ∫₀ᵗ f(x) dx. Both limits must exist independently. The symmetric limit lim_{T→∞} ∫₋ᵀᵀ f(x) dx — called the Cauchy principal value — is a weaker notion that can give a finite answer even when the integral diverges. For f(x) = x, the principal value is 0 (by symmetry), but ∫₀^∞ x dx = ∞, so the integral diverges. Conflating the principal value with the integral is a common and serious error.

The convergence theory for improper integrals closely parallels that of infinite series. Comparison tests work the same way: if 0 ≤ f(x) ≤ g(x) and ∫ g converges, then ∫ f converges. Absolute convergence (convergence of ∫ |f(x)| dx) implies convergence, but the converse fails — ∫₁^∞ sin(x)/x dx converges conditionally, meaning the oscillating cancellation produces a finite value even though ∫₁^∞ |sin(x)/x| dx diverges. This parallel is not a coincidence: both series and improper integrals are limits of accumulating quantities, and the same structural issues — decay rates, cancellation, comparison — govern convergence in both settings.