{
  "id": "minkowski-inequality",
  "title": "Minkowski Inequality",
  "lang": "en",
  "source_lang": "en",
  "summary": "The Minkowski inequality states that the \\(L^p\\) norm of a sum is at most the sum of the norms for \\(1\\le p\\le\\infty\\).",
  "category": "mathematics",
  "subcategory": "analysis",
  "type": "concept",
  "aliases": [
    "Minkowski's inequality",
    "Minkowski’s norm inequality"
  ],
  "generated_at": "2026-11-05",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/minkowski-inequality/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/minkowski-inequality/",
      "json": "https://aiwiki.page/zh/minkowski-inequality.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "The Minkowski inequality is the triangle inequality for L^p spaces: when \\(1\\le p\\le\\infty\\), the size of a sum, measured by its \\(p\\)-norm, does not exceed the sum of the sizes of its terms. Named after Hermann Minkowski, it applies to finite vectors, infinite sequences, and measurable functions. It is a foundational result in mathematical analysis, establishing that \\(p\\)-norms satisfy one of the defining properties of a norm."
    },
    {
      "heading": "Finite-dimensional form",
      "text": "Let \\(x=(x_1,\\ldots,x_n)\\) and \\(y=(y_1,\\ldots,y_n)\\) have real or complex coordinates. For \\(1\\le p<\\infty\\), the inequality states that\n\n\\[\\left(\\sum_{i=1}^{n}|x_i+y_i|^p\\right)^{1/p} \\le \\left(\\sum_{i=1}^{n}|x_i|^p\\right)^{1/p} + \\left(\\sum_{i=1}^{n}|y_i|^p\\right)^{1/p}.\\]\n\nWriting\n\n\\[\\|x\\|_p=\\left(\\sum_{i=1}^{n}|x_i|^p\\right)^{1/p},\\]\n\nthis becomes\n\n\\[\\|x+y\\|_p\\le \\|x\\|_p+\\|y\\|_p.\\]\n\nAt \\(p=\\infty\\), the corresponding norm is\n\n\\[\\|x\\|_\\infty=\\max_{1\\le i\\le n}|x_i|,\\]\n\nand the same triangle inequality holds.\n\nImportant special cases include:\n\n- \\(p=1\\): the inequality follows by summing the scalar inequalities \\(|x_i+y_i|\\le |x_i|+|y_i|\\).\n- \\(p=2\\): it is the usual triangle inequality for lengths in Euclidean space.\n- \\(p=\\infty\\): it bounds the largest absolute coordinate of a sum by the sum of the largest absolute coordinates of its terms.\n\nFor example, if \\(x=(1,0)\\), \\(y=(0,1)\\), and \\(p=2\\), then\n\n\\[\\|x+y\\|_2=\\sqrt{2}<2=\\|x\\|_2+\\|y\\|_2.\\]"
    },
    {
      "heading": "Integral and sequence forms",
      "text": "Let \\((X,\\Sigma,\\mu)\\) be a measure space, as studied in measure theory, and let \\(f,g:X\\to\\mathbb{R}\\) or \\(\\mathbb{C}\\) be measurable functions. For \\(1\\le p<\\infty\\), define\n\n\\[\\|f\\|_p= \\left(\\int_X |f|^p\\,d\\mu\\right)^{1/p},\\]\n\nwhere the integral is the Lebesgue integral. If \\(f\\) and \\(g\\) have finite \\(p\\)-norms, then\n\n\\[\\left(\\int_X |f+g|^p\\,d\\mu\\right)^{1/p} \\le \\left(\\int_X |f|^p\\,d\\mu\\right)^{1/p} + \\left(\\int_X |g|^p\\,d\\mu\\right)^{1/p}.\\]\n\nFor \\(p=\\infty\\), the norm is the essential supremum: the smallest bound on \\(|f|\\) that holds almost everywhere. The inequality becomes\n\n\\[\\operatorname*{ess\\,sup}_{x\\in X}|f(x)+g(x)| \\le \\operatorname*{ess\\,sup}_{x\\in X}|f(x)| + \\operatorname*{ess\\,sup}_{x\\in X}|g(x)|.\\]\n\nFunctions that agree almost everywhere represent the same element of \\(L^p\\). This identification is necessary because a function can have norm zero without being zero at every point.\n\nUsing counting measure on the positive integers gives the sequence version:\n\n\\[\\left(\\sum_{i=1}^{\\infty}|a_i+b_i|^p\\right)^{1/p} \\le \\left(\\sum_{i=1}^{\\infty}|a_i|^p\\right)^{1/p} + \\left(\\sum_{i=1}^{\\infty}|b_i|^p\\right)^{1/p}.\\]\n\nThus the finite-vector and infinite-sequence inequalities are special cases of the same measure-theoretic statement.\n\nA weighted form also follows immediately. For nonnegative weights \\(w_i\\),\n\n\\[\\left(\\sum_i w_i|a_i+b_i|^p\\right)^{1/p} \\le \\left(\\sum_i w_i|a_i|^p\\right)^{1/p} + \\left(\\sum_i w_i|b_i|^p\\right)^{1/p},\\]\n\nprovided the norms on the right are finite."
    },
    {
      "heading": "Proof using Hölder’s inequality",
      "text": "For \\(1<p<\\infty\\), the standard proof uses Hölder’s inequality. Let\n\n\\[q=\\frac{p}{p-1}, \\qquad \\frac1p+\\frac1q=1.\\]\n\nFirst, \\(f+g\\) has finite \\(p\\)-norm whenever \\(f\\) and \\(g\\) do. Indeed,\n\n\\[|f+g|^p \\le (|f|+|g|)^p \\le 2^{p-1}(|f|^p+|g|^p).\\]\n\nThe second inequality follows from the convexity of \\(t\\mapsto t^p\\) on the nonnegative real numbers.\n\nSet \\(h=f+g\\). The scalar triangle inequality gives\n\n\\[|h|^p\\le (|f|+|g|)|h|^{p-1}.\\]\n\nIntegrating and applying Hölder’s inequality to each term yields\n\n\\[\\begin{aligned} \\|h\\|_p^p &\\le \\int_X |f||h|^{p-1}\\,d\\mu + \\int_X |g||h|^{p-1}\\,d\\mu\\\\ &\\le (\\|f\\|_p+\\|g\\|_p) \\left(\\int_X |h|^{(p-1)q}\\,d\\mu\\right)^{1/q}. \\end{aligned}\\]\n\nSince \\((p-1)q=p\\), this is\n\n\\[\\|h\\|_p^p \\le (\\|f\\|_p+\\|g\\|_p)\\|h\\|_p^{p-1}.\\]\n\nIf \\(\\|h\\|_p=0\\), the result is immediate. Otherwise, division by \\(\\|h\\|_p^{p-1}\\) proves the inequality.\n\nThe cases \\(p=1\\) and \\(p=\\infty\\) follow directly from the pointwise triangle inequality. For \\(p=2\\), an alternative proof uses the Cauchy–Schwarz inequality to estimate the cross term in the squared norm."
    },
    {
      "heading": "Equality conditions",
      "text": "The case \\(1<p<\\infty\\)\n\nIf neither \\(f\\) nor \\(g\\) is zero as an \\(L^p\\) element, equality holds precisely when\n\n\\[f=cg \\quad\\text{almost everywhere}\\]\n\nfor a constant \\(c>0\\). Equality also holds if either function is zero almost everywhere.\n\nFor finite vectors, this means that the two nonzero vectors point in the same direction: one is a positive real multiple of the other. For complex-valued functions, positive proportionality requires both proportional magnitudes and matching complex phases wherever the functions are nonzero.\n\nThese conditions express the strict convexity of the \\(L^p\\) norm for \\(1<p<\\infty\\): distinct unit vectors have a midpoint of norm strictly less than one.\n\nThe case \\(p=1\\)\n\nEquality holds exactly when\n\n\\[|f+g|=|f|+|g| \\quad\\text{almost everywhere}.\\]\n\nFor real-valued functions, this is equivalent to\n\n\\[f(x)g(x)\\ge 0 \\quad\\text{almost everywhere}.\\]\n\nThe functions need not be proportional; they need only avoid cancellation at almost every point.\n\nFor \\(p=\\infty\\), equality does not generally imply proportionality, and it need not involve a point at which either essential supremum is attained."
    },
    {
      "heading": "Consequences and applications",
      "text": "Normed spaces and distances\n\nMinkowski’s inequality establishes the triangle axiom needed to make \\(L^p\\), for \\(1\\le p\\le\\infty\\), a normed vector space. These spaces are also complete and therefore are Banach spaces, although completeness requires a separate proof.\n\nThe formula\n\n\\[d(f,g)=\\|f-g\\|_p\\]\n\ndefines a metric on \\(L^p\\). In particular, Minkowski’s inequality gives\n\n\\[d(f,h)\\le d(f,g)+d(g,h).\\]\n\nIt also implies the reverse triangle inequality,\n\n\\[\\bigl|\\|f\\|_p-\\|g\\|_p\\bigr| \\le \\|f-g\\|_p,\\]\n\nso convergence in norm implies convergence of the norms themselves.\n\nSums and approximation errors\n\nRepeated application gives, for any finite collection \\(f_1,\\ldots,f_m\\),\n\n\\[\\left\\|\\sum_{k=1}^{m}f_k\\right\\|_p \\le \\sum_{k=1}^{m}\\|f_k\\|_p.\\]\n\nThis provides a basic estimate in approximation theory: if an error is decomposed into several terms, its total norm is bounded by the sum of their norms. Unlike squared-norm identities involving orthogonality, this bound requires no cancellation or independence assumptions.\n\nRandom variables\n\nOn a probability space, the integral form becomes\n\n\\[\\left(\\mathbb{E}|X+Y|^p\\right)^{1/p} \\le \\left(\\mathbb{E}|X|^p\\right)^{1/p} + \\left(\\mathbb{E}|Y|^p\\right)^{1/p}.\\]\n\nHere \\(X\\) and \\(Y\\) are random variables with finite \\(p\\)-th absolute moments, and \\(\\mathbb{E}\\) denotes expectation. No independence assumption is needed.\n\nFor \\(p=2\\), applying the inequality to centered variables gives\n\n\\[\\operatorname{SD}(X+Y) \\le \\operatorname{SD}(X)+\\operatorname{SD}(Y),\\]\n\nwhere \\(\\operatorname{SD}\\) is the standard deviation."
    },
    {
      "heading": "Minkowski’s integral inequality",
      "text": "A related form moves an integral outside an \\(L^p\\) norm. For suitable jointly measurable \\(F:X\\times Y\\to\\mathbb{R}\\) or \\(\\mathbb{C}\\), on sigma-finite measure spaces, and \\(1\\le p<\\infty\\),\n\n\\[\\left[ \\int_X \\left(\\int_Y |F(x,y)|\\,d\\nu(y)\\right)^p d\\mu(x) \\right]^{1/p} \\le \\int_Y \\left[ \\int_X |F(x,y)|^p\\,d\\mu(x) \\right]^{1/p} d\\nu(y).\\]\n\nIn particular, if the right-hand side is finite, then\n\n\\[\\left\\|\\int_Y F(\\,\\cdot\\,,y)\\,d\\nu(y)\\right\\|_{L^p(X)} \\le \\int_Y \\|F(\\,\\cdot\\,,y)\\|_{L^p(X)}\\,d\\nu(y).\\]\n\nThis is the integral counterpart of the triangle inequality for a finite sum. It is useful in estimating convolutions; for example, on \\(\\mathbb{R}^n\\),\n\n\\[\\|f*g\\|_p\\le \\|f\\|_1\\|g\\|_p, \\qquad 1\\le p\\le\\infty.\\]"
    },
    {
      "heading": "Why the range p\\ge 1 matters",
      "text": "For \\(0<p<1\\), the expression\n\n\\[\\|x\\|_p=\\left(\\sum_i |x_i|^p\\right)^{1/p}\\]\n\ndoes not generally satisfy the triangle inequality. Taking \\(x=(1,0)\\) and \\(y=(0,1)\\) gives\n\n\\[\\|x+y\\|_p=2^{1/p}>2=\\|x\\|_p+\\|y\\|_p.\\]\n\nIt therefore defines a quasi-norm, not a norm.\n\nNevertheless, the scalar inequality\n\n\\[|a+b|^p\\le |a|^p+|b|^p\\]\n\nimplies\n\n\\[\\|f+g\\|_p^p\\le \\|f\\|_p^p+\\|g\\|_p^p.\\]\n\nConsequently,\n\n\\[d(f,g)=\\int_X |f-g|^p\\,d\\mu\\]\n\ndoes define a metric on \\(L^p\\) equivalence classes for \\(0<p<1\\), even though the usual \\(p\\)-norm triangle inequality fails."
    }
  ],
  "links": [
    "triangle-inequality",
    "lp-space",
    "mathematical-analysis",
    "norm-mathematics",
    "complex-number",
    "euclidean-space",
    "measure-theory",
    "measurable-function",
    "lebesgue-integral",
    "almost-everywhere",
    "holders-inequality",
    "convex-function",
    "cauchy-schwarz-inequality",
    "normed-vector-space",
    "banach-space",
    "metric-space",
    "probability-space",
    "random-variable",
    "expected-value",
    "statistical-independence",
    "standard-deviation",
    "convolution"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}