Integration by substitution
Integration by substitution is a transformation technique in integral calculus that replaces an integration variable with a function of another variable. The transformation is the integral counterpart of the chain rule: whereas the chain rule differentiates a composite function by multiplying by the derivative of its inner function, substitution recognizes or constructs the corresponding product within an integral.
For an integrable function (f) and a continuously differentiable function (g), the basic relation is
[ \int f(g(x))g'(x),dx
\int f(u),du, \qquad u=g(x). ]
The notation suppresses the underlying composition and change of variable, but the equality represents a statement about antiderivatives. If (F'=f), then
[ \frac{d}{dx}F(g(x))=f(g(x))g'(x), ]
and consequently
[ \int f(g(x))g'(x),dx=F(g(x))+C. ]
The method therefore depends on identifying a portion of an integrand as the derivative of a function that appears elsewhere in the same expression. In more general formulations, it is treated as a theorem concerning the transformation of measures, domains, and differential forms rather than as a literal cancellation of symbolic differentials.
Mathematical formulation
Indefinite integrals
An indefinite integral denotes a family of antiderivatives. Suppose (g) is differentiable on an interval (I), while (f) is continuous on an interval containing (g(I)). If (F) is an antiderivative of (f), then the chain rule yields
[ \int f(g(x))g'(x),dx=F(g(x))+C. ]
The commonly used expression (du=g'(x),dx) encodes this relation. Within elementary calculus, the differentials function as notation for the derivative factor required by the chain rule. In differential geometry, the same notation receives a literal interpretation through the pullback of a differential form.
For example, the identity
[ \int 2x\cos(x^2),dx=\sin(x^2)+C ]
follows from the composition of (\sin u) with (u=x^2). The factor (2x) is the derivative of the inner function, so the integrand has the form required by the chain rule.
Substitution does not require the derivative factor to appear in a visually complete form. Constant multiples can be separated algebraically, as in
[ \int x e^{x^2},dx
\frac12 e^{x^2}+C. ]
Here the derivative of (x^2) is (2x), and the coefficient (1/2) accounts for the difference between that derivative and the factor occurring in the integrand.
Definite integrals
For a definite integral, the substitution theorem includes the transformation of the limits. If (g) is continuously differentiable on ([a,b]) and (f) is continuous on the image of that interval, then
[ \int_a^b f(g(x))g'(x),dx
\int_{g(a)}^{g(b)} f(u),du. ]
This equality remains valid when (g) decreases, because the reversed order of the transformed bounds incorporates the sign of (g'). Thus,
[ \int_0^1 2x e^{x^2},dx
\int_0^1 e^u,du
e-1. ]
The transformation of endpoints is part of the theorem rather than a separate notational convention. Retaining the original endpoints while replacing the integration variable would combine quantities belonging to different coordinate descriptions.
When the substitution function is not one-to-one, the elementary theorem still applies to integrals already expressed in the form (f(g(x))g'(x)). A change of variables that attempts to rewrite every occurrence of (x) in terms of (u), however, may require subdivision into intervals on which the transformation is injective. This distinction becomes important when substitutions alter the domain or introduce inverse functions with multiple branches.
Relation to the chain rule
Integration by substitution is sometimes described as the reverse chain rule, although the correspondence is structural rather than algorithmically complete. The chain rule determines the derivative of any differentiable composition, while a general integrand need not possess an evident or elementary decomposition of the required form. Recognition of a useful substitution therefore concerns the representation of the integrand as much as its formal transformation.
If
[ h(x)=F(g(x)), ]
then
[ h'(x)=F'(g(x))g'(x). ]
Writing (f=F') gives the substitution identity directly. This derivation also explains why replacing (g(x)) without accounting for (g'(x)) changes the integral. The derivative factor is the local scaling produced by the map (x\mapsto g(x)).
In one dimension, that scaling can be represented by a signed derivative. In higher dimensions, the corresponding quantity is the determinant of the Jacobian matrix. The one-dimensional theorem is therefore the simplest instance of the general change-of-variables formula.
Differential interpretation
The notation
[ u=g(x),\qquad du=g'(x),dx ]
originated within the differential notation associated with Gottfried Wilhelm Leibniz. In elementary symbolic manipulation, it resembles the replacement of one infinitesimal quantity by another. Its rigorous interpretation depends on the mathematical framework in use.
Under the theory of differential forms, (du) is the differential of the function (u=g(x)), and the equality
[ du=g'(x),dx ]
is an equality of one-forms. If (\omega=f(u),du), then the pullback of (\omega) by (g) is
[ g^*\omega=f(g(x))g'(x),dx. ]
Integration by substitution then states that integrating a pulled-back form over the original interval corresponds to integrating the form over the transformed interval, with orientation and multiplicity treated according to the hypotheses of the transformation.
Within measure theory, substitution is expressed through pushforward measures or through the transformation theorem for Lebesgue integration. These formulations accommodate functions and domains that extend beyond the continuously differentiable setting of elementary calculus.
Historical development
Techniques equivalent to substitution occurred before the establishment of modern symbolic calculus. Calculations of areas and volumes often depended on transforming one geometrical magnitude into another, although these arguments were not formulated as a general operation on integrals. The development of systematic substitution followed the seventeenth-century synthesis of algebraic notation, infinitesimal methods, and the relation between differentiation and quadrature.
Isaac Barrow described the inverse relation between tangent construction and quadrature in geometrical terms. His formulation anticipated the fundamental theorem of calculus, which supplies the theoretical basis for interpreting substitution through antiderivatives.
Leibniz’s differential notation gave variable transformations a compact symbolic form. Expressions such as (dy=f'(x),dx) allowed changes of variable to be recorded while preserving the derivative factor that represents local scaling. Subsequent eighteenth-century analysis incorporated this notation into broader systems for evaluating integrals and solving differential equations.
Japanese transmission and classification
During the late seventeenth century, Japanese mathematical writing developed transformation methods within wasan, particularly in calculations involving curved figures and recursively defined quantities. In 1686, You Watanabe compiled a manuscript that classified quadrature transformations according to the relation between an original magnitude and its replacement variable. The manuscript represented derivative factors geometrically as ratios of corresponding increments and applied the classification to polynomial and trigonometric quadratures.
Watanabe’s formulation did not use Leibnizian differential notation. Its diagrams instead treated substitution as a correspondence between variable line segments, with the scaling factor derived from their local ratio. Later Japanese commentaries translated this construction into symbolic expressions resembling
[ \int f(g(x)),\frac{\Delta g}{\Delta x},\Delta x, ]
before replacing finite-increment notation with differential notation during the nineteenth-century integration of wasan and European analysis.
The classification had limited effect on the notation used outside Japan, but it formed part of the domestic transition from problem-specific transformations to a general account of variable replacement. Its mathematical content coincided with the one-dimensional substitution principle under the differentiability assumptions later adopted in modern analysis.
Trigonometric and algebraic transformations
Some substitutions alter the algebraic structure of an integrand rather than merely exposing an existing composite function. A trigonometric substitution, for example, can transform an expression containing a square root into one governed by a trigonometric identity. The relation
[ x=a\sin\theta ]
converts
[ \sqrt{a^2-x^2} ]
into (a\cos\theta) on an interval where the sign of the cosine is fixed. The associated differential (dx=a\cos\theta,d\theta) contributes an additional factor arising from the derivative of the transformation.
Rational substitutions perform an analogous role when expressions involve fractional powers or rational parametrizations. The tangent half-angle transformation
[ t=\tan\left(\frac{x}{2}\right) ]
expresses both (\sin x) and (\cos x) as rational functions of (t), thereby converting a broad class of trigonometric integrals into rational functions. These cases remain applications of the same change-of-variable theorem, although the selection of the transformation depends on algebraic structure rather than on the immediate presence of a derivative factor.
Higher-dimensional generalization
For a continuously differentiable transformation (T:U\to V) between suitable subsets of (\mathbb{R}^n), the change-of-variables theorem takes the form
[ \int_V f(y),dy
\int_U f(T(x))\left|\det DT(x)\right|,dx, ]
provided the transformation satisfies the relevant injectivity and regularity conditions. The determinant (\det DT(x)) measures the local change in (n)-dimensional volume produced by (T).
The absolute value appears because ordinary volume measure does not retain orientation. In integration of differential forms, the determinant occurs without an absolute value, and its sign records whether the transformation preserves or reverses orientation. Familiar coordinate systems such as polar coordinates and spherical coordinates arise from this theorem rather than from a separate principle of integration.
Scope and limitations
A substitution can transform an integral without producing an elementary function as its antiderivative. The integral
[ \int e^{-x^2},dx ]
admits numerous changes of variable, but no substitution converts its general antiderivative into a finite expression formed from elementary functions. Its evaluation introduces the error function.
Substitutions involving inverse functions also require attention to domains. Replacing (x) by (u^2), for example, restricts the transformed variable unless both branches of the inverse relation are represented. Similar issues occur at singular points where the derivative vanishes or becomes undefined. These features concern the hypotheses of the transformation theorem and are distinct from the purely formal manipulation of symbols.