Improper Integrals: Comparison and p-Test Explained

An improper integral is a definite integral that breaks the usual rules: either a limit of integration is infinite, or the integrand blows up somewhere inside the interval. We handle it by replacing the problem spot with a variable limit and then taking a limit. The integral converges only if that limit exists as a finite number.

Improper integrals appear whenever models run to infinity — total probabilities, infinite time horizons, or areas under curves like 1/x² — and whenever densities spike, such as near a point charge in physics. Deciding convergence is a skill of its own, and the comparison test and p-test are the main tools.

Key Concepts

• Type 1 (infinite limit): ∫a∞ f(x) dx = limb→∞ ∫ab f(x) dx. Converges if the limit is finite.
• Type 2 (singularity): if f blows up at c inside [a, b], split there: ∫ab f = limt→c⁻ ∫at f + limt→c⁺ ∫tb f. Both one-sided limits must exist.
• p-test at infinity: ∫1∞ dx/xp converges if and only if p > 1.
• p-test at zero: ∫01 dx/xp converges if and only if p < 1.
• Comparison test: if 0 ≤ f(x) ≤ g(x) and ∫ g converges, then ∫ f converges; if f(x) ≥ g(x) ≥ 0 and ∫ g diverges, then ∫ f diverges.

Formulas

• p-integral at ∞: ∫1∞ dx/xp = 1/(p−1) for p > 1; diverges for p ≤ 1.
• p-integral at 0: ∫01 dx/xp = 1/(1−p) for p < 1; diverges for p ≥ 1.
• Exponential benchmark: ∫0∞ e−x dx = limb→∞ (1 − e−b) = 1, convergent.
• Limit comparison: if f(x)/g(x) → L with 0 < L < ∞, then ∫ f and ∫ g share the same convergence behaviour on the same improper interval.

Worked Example

Determine whether ∫1∞ dx/(x² + 1) converges.

1. For x ≥ 1, we have x² + 1 > x², so 0 < 1/(x² + 1) < 1/x².
2. The benchmark integral ∫1∞ dx/x² converges by the p-test with p = 2 > 1; its value is [−1/x]1∞ = 0 − (−1) = 1.
3. By the comparison test, since our integrand is smaller than 1/x² and non-negative, ∫1∞ dx/(x² + 1) also converges.
4. Indeed its exact value is [arctan x]1∞ = π/2 − π/4 = π/4 ≈ 0.7854, which is less than 1, consistent with the comparison.

Common Misconceptions

• "An infinite region always has infinite area." Gabriel's horn shows ∫1∞ dx/x² = 1 is finite; infinite extent does not force infinite area.
• "You can use symmetric limits for ∫−∞∞." The two halves must converge independently; lima→∞ ∫−aa x dx = 0 is the Cauchy principal value, but ∫−∞∞ x dx diverges as an improper integral.
• "If the integrand → 0 at infinity, the integral converges." ∫1∞ dx/x diverges (it equals ln b → ∞) even though 1/x → 0.

Real-World Applications

• Probability: total probability ∫−∞∞ of a density must equal 1, which is a convergence statement about an improper integral.
• Physics: electric potential and gravitational field of extended distributions involve integrals out to infinity that must converge.
• Engineering: Laplace transforms ∫0∞ e−stf(t) dt are improper integrals whose convergence determines which signals can be analysed.

Improper integrals extend integration past its comfort zone by taking limits, and convergence is never automatic. The p-tests give you instant answers for power-law behaviour at infinity and at zero, the comparison test handles everything that resembles a benchmark, and the two halves of a doubly-improper integral must each converge on their own. When in doubt, compare against 1/xp.