{
  "id": "differentiation-under-the-integral-sign",
  "title": "Differentiation Under the Integral Sign",
  "lang": "en",
  "source_lang": "en",
  "summary": "Differentiation under the integral sign is the interchange of parameter differentiation and integration, justified by regularity and convergence conditions.",
  "category": "mathematics",
  "subcategory": "analysis",
  "type": "concept",
  "aliases": [
    "Leibniz integral rule",
    "differentiation of parameter-dependent integrals",
    "Feynman's trick"
  ],
  "generated_at": "2026-11-05",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/differentiation-under-the-integral-sign/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/differentiation-under-the-integral-sign/",
      "json": "https://aiwiki.page/zh/differentiation-under-the-integral-sign.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "Differentiation under the integral sign is a method in calculus and mathematical analysis for finding the derivative of an integral that depends on a parameter. For a function\n\n\\[F(t)=\\int_a^b f(x,t)\\,dx,\\]\n\nthe method seeks to justify the identity\n\n\\[F'(t)=\\int_a^b \\frac{\\partial f}{\\partial t}(x,t)\\,dx.\\]\n\nThis interchange is valid under suitable regularity or integrability conditions, but not automatically. When the integration limits also depend on the parameter, additional boundary terms appear. The resulting formula is commonly called the Leibniz integral rule, after Gottfried Wilhelm Leibniz.\n\nBesides computing derivatives, the method can evaluate definite integrals: introducing a parameter may turn a difficult integral into one whose derivative is simpler, after which the original expression is recovered by integration with respect to the parameter."
    },
    {
      "heading": "Fixed integration limits",
      "text": "An elementary version applies on a finite interval. Let \\(I\\) be an open interval of parameter values, and suppose that \\(f(x,t)\\) and its partial derivative \\(\\partial_t f(x,t)\\) are continuous on \\([a,b]\\times I\\). Then\n\n\\[F(t)=\\int_a^b f(x,t)\\,dx\\]\n\nis continuously differentiable on \\(I\\), and\n\n\\[F'(t)=\\int_a^b \\partial_t f(x,t)\\,dx.\\]\n\nContinuity on a compact rectangle around each parameter value provides the uniform control needed to exchange differentiation and integration. These hypotheses are convenient sufficient conditions, not necessary ones: the identity can remain valid when the integrand is discontinuous or unbounded.\n\nFor example,\n\n\\[F(t)=\\int_0^1 e^{tx}\\,dx\\]\n\nsatisfies\n\n\\[F'(t)=\\int_0^1 xe^{tx}\\,dx.\\]\n\nThe integral definition remains well behaved at \\(t=0\\), even though its explicit expression\n\n\\[F(t)=\\frac{e^t-1}{t},\\qquad t\\ne0,\\]\n\nhas an apparent singularity there."
    },
    {
      "heading": "A dominated-convergence theorem",
      "text": "A more general formulation uses measure theory and the Lebesgue integral. Let \\((X,\\mathcal A,\\mu)\\) be a measure space, let \\(I\\subseteq\\mathbb R\\) be an open interval, and fix \\(t_0\\in I\\). Suppose that, on an open interval \\(J\\subseteq I\\) containing \\(t_0\\):\n\n- For every \\(t\\in J\\), the function \\(x\\mapsto f(x,t)\\) is measurable.\n- Outside a single set of measure zero, \\(t\\mapsto f(x,t)\\) is differentiable throughout \\(J\\).\n- There is an integrable function \\(g\\geq0\\) such that\n\n\\[|\\partial_t f(x,t)|\\leq g(x)\\]\n\nfor every \\(t\\in J\\), outside that same exceptional set.\n- The function \\(f(\\,\\cdot\\,,t_0)\\) is integrable.\n\nThen \\(f(\\,\\cdot\\,,t)\\) is integrable for every \\(t\\in J\\), and\n\n\\[\\left.\\frac{d}{dt}\\int_X f(x,t)\\,d\\mu(x)\\right|_{t=t_0} = \\int_X \\partial_t f(x,t_0)\\,d\\mu(x).\\]\n\nThe crucial condition is a common integrable bound on the derivatives in a neighborhood of the parameter value. The bound may depend on \\(x\\), but not on \\(t\\) within that neighborhood.\n\nWhy the theorem works\n\nDifferentiation is a limit of difference quotients:\n\n\\[\\frac{F(t_0+h)-F(t_0)}{h} = \\int_X \\frac{f(x,t_0+h)-f(x,t_0)}{h}\\,d\\mu(x).\\]\n\nFor almost every \\(x\\), the integrand converges to \\(\\partial_t f(x,t_0)\\). For real-valued \\(f\\), the mean value theorem bounds its absolute value by \\(g(x)\\). The dominated convergence theorem therefore permits the limit to pass through the integral.\n\nMore generally, an integrable bound directly on the difference quotients is sufficient, even if no derivative bound has been established throughout a neighborhood.\n\nFor complex-valued integrands with a real parameter, the same result follows by considering real and imaginary parts."
    },
    {
      "heading": "Variable integration limits",
      "text": "If the limits depend on \\(t\\), the derivative includes contributions from the moving endpoints. Under suitable continuity and differentiability assumptions,\n\n\\[F(t)=\\int_{a(t)}^{b(t)} f(x,t)\\,dx\\]\n\nsatisfies\n\n\\[F'(t) = f(b(t),t)b'(t) - f(a(t),t)a'(t) + \\int_{a(t)}^{b(t)}\\partial_t f(x,t)\\,dx.\\]\n\nThe first two terms describe the change caused by moving the endpoints; the final term describes the change in the integrand over the interval.\n\nThis rule combines differentiation under the integral sign with the fundamental theorem of calculus and the chain rule. For example, if\n\n\\[F(t)=\\int_0^t e^{tx}\\,dx,\\]\n\nthen\n\n\\[F'(t)=e^{t^2}+\\int_0^t xe^{tx}\\,dx.\\]\n\nThe boundary values in this formula require appropriate regularity. At singular endpoints or infinite limits, the ordinary endpoint formula cannot simply be applied without a separate convergence argument."
    },
    {
      "heading": "Evaluating integrals by introducing a parameter",
      "text": "Differentiation under the integral sign is particularly useful when a derivative simplifies an integrand. The usual procedure is to introduce a parameter, differentiate, evaluate the simpler integral, and then recover the parameter-dependent expression from an initial value.\n\nConsider\n\n\\[F(a)=\\int_0^1\\frac{x^a-1}{\\ln x}\\,dx, \\qquad a>-1,\\]\n\nwhere \\(\\ln\\) denotes the natural logarithm. The singularity at \\(x=1\\) is removable. Differentiating with respect to \\(a\\) gives\n\n\\[F'(a)=\\int_0^1 x^a\\,dx=\\frac{1}{a+1}.\\]\n\nTo justify the interchange near any fixed \\(a>-1\\), choose a lower bound \\(\\alpha>-1\\) for nearby parameter values. Since \\(0<x<1\\),\n\n\\[x^a\\leq x^\\alpha,\\]\n\nand \\(x^\\alpha\\) is integrable on \\((0,1)\\).\n\nBecause \\(F(0)=0\\), integration with respect to \\(a\\) yields\n\n\\[\\int_0^1\\frac{x^a-1}{\\ln x}\\,dx=\\ln(a+1), \\qquad a>-1.\\]\n\nThe name Feynman's trick is often used informally for this parameter-introduction technique. It is not a separate theorem and does not dispense with the conditions required for the interchange."
    },
    {
      "heading": "Improper integrals and repeated derivatives",
      "text": "For an integral over an unbounded interval,\n\n\\[F(t)=\\int_0^\\infty f(x,t)\\,dx,\\]\n\ncontinuity of \\(f\\) and \\(\\partial_t f\\) on every bounded rectangle is not enough. Additional control is needed near infinity and near any singular endpoint. A dominating integrable function provides one such control; suitable uniform convergence conditions provide another.\n\nThe gamma function illustrates repeated differentiation:\n\n\\[\\Gamma(s)=\\int_0^\\infty x^{s-1}e^{-x}\\,dx, \\qquad s>0.\\]\n\nFor every nonnegative integer \\(n\\),\n\n\\[\\Gamma^{(n)}(s) = \\int_0^\\infty x^{s-1}e^{-x}(\\ln x)^n\\,dx.\\]\n\nTo justify this locally, restrict \\(s\\) to a compact interval \\([\\alpha,\\beta]\\subset(0,\\infty)\\). Near zero, the absolute value of the integrand is bounded by\n\n\\[x^{\\alpha-1}|\\ln x|^n,\\]\n\nwhich is integrable. For \\(x\\geq1\\), it is bounded by\n\n\\[x^{\\beta-1}e^{-x}(\\ln x)^n,\\]\n\nwhich is also integrable.\n\nMore generally,\n\n\\[F^{(n)}(t)=\\int_X\\partial_t^n f(x,t)\\,d\\mu(x)\\]\n\nholds when the necessary interchange conditions are satisfied at each differentiation step. Justifying one derivative does not automatically justify every higher derivative."
    },
    {
      "heading": "Failure without adequate control",
      "text": "The existence of pointwise derivatives does not guarantee that differentiation commutes with integration.\n\nFor \\(0<x<1\\), define\n\n\\[f(x,t)=\\frac{t^2}{x^2+t^2}.\\]\n\nFor every fixed \\(x>0\\),\n\n\\[\\partial_t f(x,0)=0,\\]\n\nso\n\n\\[\\int_0^1\\partial_t f(x,0)\\,dx=0.\\]\n\nHowever, for \\(t\\ne0\\),\n\n\\[F(t)=\\int_0^1\\frac{t^2}{x^2+t^2}\\,dx = |t|\\arctan\\left(\\frac{1}{|t|}\\right),\\]\n\nwhile \\(F(0)=0\\). Consequently,\n\n\\[\\lim_{t\\downarrow0}\\frac{F(t)}{t}=\\frac{\\pi}{2}, \\qquad \\lim_{t\\uparrow0}\\frac{F(t)}{t}=-\\frac{\\pi}{2}.\\]\n\nThus \\(F\\) is not differentiable at zero, although every pointwise derivative of the integrand at zero exists.\n\nThe failure results from concentration near \\(x=0\\): the difference quotients approach zero at every fixed \\(x>0\\), but retain a nonvanishing integral. Pointwise convergence alone does not control this behavior."
    },
    {
      "heading": "Applications",
      "text": "Integral transforms\n\nFor a Laplace transform,\n\n\\[L(s)=\\int_0^\\infty e^{-sx}u(x)\\,dx,\\]\n\nsuitable integrability conditions give\n\n\\[L^{(n)}(s) = (-1)^n\\int_0^\\infty x^n e^{-sx}u(x)\\,dx.\\]\n\nDifferentiation in the transform variable therefore corresponds to multiplication by powers of the original variable.\n\nFor the Fourier transform convention\n\n\\[\\widehat u(\\xi)=\\int_{\\mathbb R}u(x)e^{-i\\xi x}\\,dx,\\]\n\nthe assumptions \\(u\\in L^1(\\mathbb R)\\) and \\(xu\\in L^1(\\mathbb R)\\) imply\n\n\\[\\widehat u'(\\xi) = \\int_{\\mathbb R}(-ix)u(x)e^{-i\\xi x}\\,dx.\\]\n\nExpectations and statistical models\n\nAn expectation under a fixed probability law is an integral, so the same principles yield\n\n\\[\\frac{d}{dt}\\mathbb E[h(X,t)] = \\mathbb E[\\partial_t h(X,t)]\\]\n\nwhen the derivative is appropriately dominated.\n\nIf the probability density also depends on the parameter, its derivative must be included:\n\n\\[\\frac{d}{dt}\\int h(x,t)p(x,t)\\,dx = \\int\\left[ \\partial_t h(x,t)\\,p(x,t) + h(x,t)\\,\\partial_t p(x,t) \\right]dx.\\]\n\nThis formula assumes a fixed integration domain and valid interchange conditions. Parameter-dependent support may require additional boundary terms. Such distinctions matter when differentiating normalization identities and deriving properties of Fisher information.\n\nSensitivity analysis and integral representations\n\nIn sensitivity analysis, differentiation under the integral sign expresses how an integrated quantity changes as a model parameter varies. For a vector parameter \\(\\theta\\), componentwise conditions can justify\n\n\\[\\nabla_\\theta\\int_X f(x,\\theta)\\,d\\mu(x) = \\int_X\\nabla_\\theta f(x,\\theta)\\,d\\mu(x).\\]\n\nThe method also appears in integral representations of solutions to partial differential equations, where derivatives with respect to space or time are transferred to an integral kernel. The required bounds depend on the kernel's singularities, the integration domain, and the regularity of the data."
    }
  ],
  "links": [
    "calculus",
    "mathematical-analysis",
    "derivative",
    "integral",
    "gottfried-wilhelm-leibniz",
    "partial-derivative",
    "measure-theory",
    "lebesgue-integral",
    "limit",
    "mean-value-theorem",
    "dominated-convergence-theorem",
    "fundamental-theorem-of-calculus",
    "chain-rule",
    "natural-logarithm",
    "uniform-convergence",
    "gamma-function",
    "pointwise-convergence",
    "laplace-transform",
    "fourier-transform",
    "expected-value",
    "probability-density-function",
    "fisher-information",
    "sensitivity-analysis",
    "partial-differential-equation"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}