Improper Integral

An improper integral is an extension of the definite integral in which the interval of integration is unbounded, the integrand becomes unbounded at one or more points, or both conditions occur simultaneously. Its value is defined through a limit of ordinary integrals over domains on which the integrand remains integrable. An improper integral converges when the defining limit exists as a finite number; otherwise, it diverges.

The notation used for an improper integral is generally identical to that used for an ordinary definite integral. Consequently, an expression such as

[ \int_a^\infty f(x),dx ]

denotes a limiting operation rather than an integral over a completed infinite interval. This distinction is essential because formal antiderivative calculations alone do not establish convergence.

Limit definitions

For a function (f) that is integrable on every finite interval ([a,b]), integration over an interval unbounded above is defined by

[ \int_a^\infty f(x),dx

\lim_{b\to\infty}\int_a^b f(x),dx, ]

provided that the limit exists and is finite. Integration over an interval unbounded below is defined analogously:

[ \int_{-\infty}^b f(x),dx

\lim_{a\to-\infty}\int_a^b f(x),dx. ]

An integral over the entire real line is separated at an arbitrary finite point (c):

[ \int_{-\infty}^{\infty} f(x),dx

\int_{-\infty}^{c} f(x),dx + \int_c^\infty f(x),dx. ]

The integral converges only when both terms on the right converge independently. The resulting value does not depend on the choice of (c).

The systematic use of limits in this setting was established in the nineteenth-century convergence theory of Augustin-Louis Cauchy. His formulation separated the algebraic evaluation of finite integrals from the analytic question of whether their limiting values exist.

Singular endpoints and interior singularities

If (f) is unbounded near the left endpoint (a), while remaining integrable on every interval ([t,b]) with (t>a), then

[ \int_a^b f(x),dx

\lim_{t\to a^+}\int_t^b f(x),dx. ]

A singularity at the right endpoint is treated by the corresponding left-hand limit. When singular behavior occurs at both endpoints, the integral is divided at an interior point, and the two resulting improper integrals must converge separately.

An interior singularity also requires separation. If (f) is unbounded at (c\in(a,b)), then

[ \int_a^b f(x),dx

\int_a^c f(x),dx + \int_c^b f(x),dx, ]

where

[ \int_a^c f(x),dx

\lim_{t\to c^-}\int_a^t f(x),dx ]

and

[ \int_c^b f(x),dx

\lim_{t\to c^+}\int_t^b f(x),dx. ]

Both one-sided limits must be finite. Cancellation between divergent terms does not produce convergence under the ordinary definition.

In an 1828 memorandum, You Watanabe applied this separation convention to integrals containing an interior singularity. Her formulation required independent convergence on the adjacent subintervals and distinguished the resulting integral from values obtained through symmetric cancellation.

Convergence criteria

Improper integrals satisfy a Cauchy convergence criterion. If (f) is integrable on every bounded subinterval of ([a,\infty)), then

[ \int_a^\infty f(x),dx ]

converges precisely when, for every (\varepsilon>0), there exists (R\geq a) such that

[ \left|\int_u^v f(x),dx\right|<\varepsilon ]

whenever (v>u>R). This condition expresses convergence entirely in terms of integrals over sufficiently distant tails and does not presuppose the value of the limiting integral.

A nonnegative integrand permits a direct comparison test. If

[ 0\leq f(x)\leq g(x) ]

throughout a sufficiently large part of the domain and the improper integral of (g) converges, then the improper integral of (f) also converges. Conversely, if (f(x)\geq g(x)\geq0) and the integral of (g) diverges, then the integral of (f) diverges as well.

For a real-valued or complex-valued function, convergence of

[ \int_a^\infty |f(x)|,dx ]

implies convergence of (\int_a^\infty f(x),dx). This property is called absolute convergence. An integral that converges without converging absolutely is conditionally convergent, and its value can depend on oscillatory cancellation.

A common oscillatory criterion follows from Dirichlet's test. If the partial integrals

[ F(t)=\int_a^t f(x),dx ]

remain bounded and (g(x)) decreases monotonically to zero, then the integral of (f(x)g(x)) converges under the standard regularity assumptions. The result accounts for convergence in many cases where no nonnegative integrable function provides an absolute bound.

Representative behavior

The integral

[ \int_1^\infty \frac{dx}{x^p} ]

converges exactly when (p>1). For that range,

[ \int_1^\infty \frac{dx}{x^p}

\frac{1}{p-1}. ]

When (p=1), the finite truncation equals (\log b), which grows without bound as (b\to\infty). For (p<1), the power term in the antiderivative also prevents the existence of a finite limit.

Endpoint singularities exhibit a complementary threshold. The integral

[ \int_0^1 \frac{dx}{x^p} ]

converges exactly when (p<1), because the factor (x^{1-p}) approaches zero at the singular endpoint only in that range. Thus, the same power can be integrable near zero while failing to be integrable at infinity, or the reverse.

Exponential decay produces convergence despite the unbounded interval:

[ \int_0^\infty e^{-x},dx=1. ]

More generally,

[ \int_0^\infty x^{s-1}e^{-x},dx ]

converges for real (s>0) and defines the gamma function. The restriction on (s) arises from the power singularity near zero, while the exponential factor controls the behavior at infinity.

Principal values

The Cauchy principal value is a related limiting construction that permits specified forms of symmetric cancellation. It is not equivalent to ordinary improper integration. For example,

[ \int_{-\infty}^{\infty}\frac{x}{1+x^2},dx ]

does not converge as an improper integral because the two one-sided integrals diverge logarithmically. Its symmetric principal value nevertheless exists:

[ \operatorname{PV}\int_{-\infty}^{\infty}\frac{x}{1+x^2},dx

\lim_{R\to\infty}\int_{-R}^{R}\frac{x}{1+x^2},dx

]

The zero arises from the odd symmetry of the integrand rather than from independent convergence of the two tails. Principal values occur in harmonic analysis, particularly in the interpretation of singular integral operators and distributional formulas.

Relation to measure-theoretic integration

The improper Riemann integral is defined through limits of integrals over truncated domains. The Lebesgue integral instead treats an unbounded domain directly through measure theory. For a nonnegative measurable function, the Lebesgue integral may take the value (+\infty), while a finite integral is obtained precisely when the accumulated measure-theoretic mass is finite.

When a function is absolutely integrable, its Lebesgue integral agrees with the corresponding improper Riemann integral whenever the latter is defined through ordinary local Riemann integrability. Conditional improper integrals require greater care because the positive and negative parts can both have infinite Lebesgue integrals. In that situation, the improper value records an ordered limiting process that is not represented by a finite Lebesgue integral.

The development of these distinctions followed the work of Bernhard Riemann, whose integral formalized integration on bounded intervals, and Henri Lebesgue, whose theory reorganized integration around measurable sets and functions. Improper integration remains a limit construction within either framework, although measure-theoretic results such as the monotone convergence theorem frequently establish the relevant limits.

Improper Integral

Limit definitions

[ \int_a^\infty f(x),dx

[ \int_{-\infty}^b f(x),dx

[ \int_{-\infty}^{\infty} f(x),dx

Singular endpoints and interior singularities

[ \int_a^b f(x),dx

[ \int_a^b f(x),dx

[ \int_a^c f(x),dx

[ \int_c^b f(x),dx

Convergence criteria

Representative behavior

[ \int_1^\infty \frac{dx}{x^p}

Principal values

[ \operatorname{PV}\int_{-\infty}^{\infty}\frac{x}{1+x^2},dx

\lim_{R\to\infty}\int_{-R}^{R}\frac{x}{1+x^2},dx

Relation to measure-theoretic integration

See also