{
  "id": "hyperbolic-functions",
  "title": "Hyperbolic Functions",
  "lang": "en",
  "source_lang": "en",
  "summary": "Hyperbolic functions are exponential-based functions that parametrize a hyperbola and occur in geometry, calculus, differential equations, and mathematical physics.",
  "category": "mathematics",
  "subcategory": "analysis",
  "type": "concept",
  "aliases": [
    "Hyperbolic trigonometric functions"
  ],
  "generated_at": "2026-11-05",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/hyperbolic-functions/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/hyperbolic-functions/",
      "json": "https://aiwiki.page/zh/hyperbolic-functions.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "Hyperbolic functions are a family of mathematical functions analogous to the ordinary trigonometric functions, but associated with a hyperbola rather than a circle. The basic functions, hyperbolic sine and hyperbolic cosine, are defined by combinations of the exponential function. They provide a natural language for describing exponential growth and decay together, parametrizing hyperbolas, and solving certain differential equations. Their names reflect geometric relationships, not a restriction to hyperbolic geometry."
    },
    {
      "heading": "Definitions",
      "text": "For a real or complex argument \\(z\\), the hyperbolic sine and hyperbolic cosine are defined by\n\n\\[\\sinh z=\\frac{e^z-e^{-z}}{2}, \\qquad \\cosh z=\\frac{e^z+e^{-z}}{2}.\\]\n\nThe other four standard hyperbolic functions are\n\n\\[\\tanh z=\\frac{\\sinh z}{\\cosh z}, \\qquad \\coth z=\\frac{\\cosh z}{\\sinh z},\\]\n\n\\[\\operatorname{sech}z=\\frac{1}{\\cosh z}, \\qquad \\operatorname{csch}z=\\frac{1}{\\sinh z},\\]\n\nwhere the denominators must be nonzero. Their names are hyperbolic tangent, hyperbolic cotangent, hyperbolic secant, and hyperbolic cosecant.\n\nFor real arguments, their domains and ranges are:\n\nFunction | Real domain | Real range\n\\(\\sinh x\\) | \\(\\mathbb R\\) | \\(\\mathbb R\\)\n\\(\\cosh x\\) | \\(\\mathbb R\\) | \\([1,\\infty)\\)\n\\(\\tanh x\\) | \\(\\mathbb R\\) | \\((-1,1)\\)\n\\(\\coth x\\) | \\(\\mathbb R\\setminus\\{0\\}\\) | \\((-\\infty,-1)\\cup(1,\\infty)\\)\n\\(\\operatorname{sech}x\\) | \\(\\mathbb R\\) | \\((0,1]\\)\n\\(\\operatorname{csch}x\\) | \\(\\mathbb R\\setminus\\{0\\}\\) | \\(\\mathbb R\\setminus\\{0\\}\\)\n\nThe functions \\(\\sinh\\), \\(\\tanh\\), \\(\\coth\\), and \\(\\operatorname{csch}\\) are odd; \\(\\cosh\\) and \\(\\operatorname{sech}\\) are even. In particular,\n\n\\[\\sinh 0=0,\\qquad \\cosh 0=1,\\qquad \\tanh 0=0.\\]\n\nUnlike sine and cosine, hyperbolic sine and cosine are not periodic on the real line. As \\(x\\to+\\infty\\),\n\n\\[\\sinh x\\sim\\frac{e^x}{2}, \\qquad \\cosh x\\sim\\frac{e^x}{2},\\]\n\nwhile \\(\\tanh x\\to1\\). As \\(x\\to-\\infty\\), \\(\\tanh x\\to-1\\)."
    },
    {
      "heading": "Geometric interpretation",
      "text": "The central identity is\n\n\\[\\cosh^2 t-\\sinh^2 t=1.\\]\n\nConsequently,\n\n\\[(X,Y)=(\\cosh t,\\sinh t)\\]\n\nparametrizes the entire right-hand branch of the unit hyperbola\n\n\\[X^2-Y^2=1,\\qquad X>0.\\]\n\nThis parallels the parametrization \\((\\cos t,\\sin t)\\) of the unit circle, whose equation is \\(X^2+Y^2=1\\). The change from a sum of squares to a difference of squares accounts for many of the sign differences between trigonometric and hyperbolic identities.\n\nThe parameter \\(t\\) is called a hyperbolic angle. Its geometric meaning is an area rather than an ordinary Euclidean angle: the signed area of the sector bounded by the ray through \\((1,0)\\), the ray through \\((\\cosh t,\\sinh t)\\), and the intervening hyperbola arc is \\(t/2\\). Indeed,\n\n\\[\\frac12\\int_0^t(XY'-YX')\\,du = \\frac12\\int_0^t \\left(\\cosh^2u-\\sinh^2u\\right)\\,du =\\frac t2.\\]\n\nSimilarly, an ordinary circular angle measured in radians is twice the corresponding sector area of a unit circle. However, hyperbolic angle is not Euclidean arc length along the hyperbola."
    },
    {
      "heading": "Fundamental identities",
      "text": "The exponential definitions immediately give\n\n\\[\\cosh x+\\sinh x=e^x, \\qquad \\cosh x-\\sinh x=e^{-x}.\\]\n\nBesides the fundamental identity, useful quotient identities are\n\n\\[1-\\tanh^2x=\\operatorname{sech}^2x, \\qquad \\coth^2x-1=\\operatorname{csch}^2x.\\]\n\nThe addition formulas are\n\n\\[\\sinh(x+y)=\\sinh x\\cosh y+\\cosh x\\sinh y,\\]\n\n\\[\\cosh(x+y)=\\cosh x\\cosh y+\\sinh x\\sinh y.\\]\n\nIt follows that\n\n\\[\\tanh(x+y) = \\frac{\\tanh x+\\tanh y} {1+\\tanh x\\tanh y},\\]\n\nwhenever the expressions are defined. For real \\(x\\) and \\(y\\), the denominator is always positive.\n\nSetting \\(y=x\\) gives the double-angle formulas\n\n\\[\\sinh 2x=2\\sinh x\\cosh x,\\]\n\n\\[\\cosh 2x =\\cosh^2x+\\sinh^2x =2\\cosh^2x-1 =1+2\\sinh^2x,\\]\n\nand\n\n\\[\\tanh 2x=\\frac{2\\tanh x}{1+\\tanh^2x}.\\]\n\nThus the algebra of hyperbolic functions closely resembles trigonometry, but the two systems cannot be interchanged without tracking their signs."
    },
    {
      "heading": "Calculus and power series",
      "text": "Derivatives and integrals\n\nThe derivatives of the two basic functions satisfy the particularly symmetric relations\n\n\\[\\frac{d}{dx}\\sinh x=\\cosh x, \\qquad \\frac{d}{dx}\\cosh x=\\sinh x.\\]\n\nThe remaining derivatives are\n\n\\[\\frac{d}{dx}\\tanh x=\\operatorname{sech}^2x, \\qquad \\frac{d}{dx}\\coth x=-\\operatorname{csch}^2x,\\]\n\n\\[\\frac{d}{dx}\\operatorname{sech}x =-\\operatorname{sech}x\\tanh x,\\]\n\n\\[\\frac{d}{dx}\\operatorname{csch}x =-\\operatorname{csch}x\\coth x.\\]\n\nIn particular, \\(\\sinh x\\) and \\(\\tanh x\\) are strictly increasing on the real line. The function \\(\\cosh x\\) decreases for \\(x<0\\), increases for \\(x>0\\), and has its unique minimum at \\(x=0\\).\n\nCorresponding antiderivatives include\n\n\\[\\int\\sinh x\\,dx=\\cosh x+C, \\qquad \\int\\cosh x\\,dx=\\sinh x+C,\\]\n\n\\[\\int\\tanh x\\,dx=\\ln(\\cosh x)+C,\\]\n\n\\[\\int\\coth x\\,dx=\\ln|\\sinh x|+C.\\]\n\nThe last formula applies on intervals not containing zero.\n\nPower series\n\nThe power series expansions are\n\n\\[\\sinh z=\\sum_{n=0}^{\\infty}\\frac{z^{2n+1}}{(2n+1)!} =z+\\frac{z^3}{3!}+\\frac{z^5}{5!}+\\cdots,\\]\n\n\\[\\cosh z=\\sum_{n=0}^{\\infty}\\frac{z^{2n}}{(2n)!} =1+\\frac{z^2}{2!}+\\frac{z^4}{4!}+\\cdots.\\]\n\nBoth converge for every complex \\(z\\). Unlike the corresponding sine and cosine series, their coefficients do not alternate in sign.\n\nNear zero,\n\n\\[\\tanh z=z-\\frac{z^3}{3}+\\frac{2z^5}{15}+\\cdots.\\]\n\nIts Taylor series has radius of convergence \\(\\pi/2\\), determined by the nearest complex singularities at \\(z=\\pm i\\pi/2\\)."
    },
    {
      "heading": "Inverse hyperbolic functions",
      "text": "An inverse function requires a one-to-one domain. Hyperbolic sine is one-to-one on all of \\(\\mathbb R\\), while hyperbolic cosine is conventionally restricted to \\([0,\\infty)\\) before inversion.\n\nUsing the natural logarithm, the principal real inverses are\n\n\\[\\operatorname{arsinh}x =\\ln\\left(x+\\sqrt{x^2+1}\\right), \\qquad x\\in\\mathbb R,\\]\n\n\\[\\operatorname{arcosh}x =\\ln\\left(x+\\sqrt{x^2-1}\\right), \\qquad x\\ge1,\\]\n\n\\[\\operatorname{artanh}x =\\frac12\\ln\\left(\\frac{1+x}{1-x}\\right), \\qquad |x|<1.\\]\n\nTheir ranges are respectively \\(\\mathbb R\\), \\([0,\\infty)\\), and \\(\\mathbb R\\). Common alternative notation includes \\(\\operatorname{asinh}\\), \\(\\operatorname{acosh}\\), and \\(\\operatorname{atanh}\\).\n\nThe notation \\(\\sinh^{-1}x\\) is also widely used for the inverse function, but must be distinguished from the reciprocal \\(1/\\sinh x=\\operatorname{csch}x\\).\n\nTheir derivatives are\n\n\\[\\frac{d}{dx}\\operatorname{arsinh}x =\\frac{1}{\\sqrt{1+x^2}},\\]\n\n\\[\\frac{d}{dx}\\operatorname{arcosh}x =\\frac{1}{\\sqrt{x^2-1}}, \\qquad x>1,\\]\n\n\\[\\frac{d}{dx}\\operatorname{artanh}x =\\frac{1}{1-x^2}, \\qquad |x|<1.\\]\n\nThese formulas explain the frequent appearance of inverse hyperbolic functions in integration. For example,\n\n\\[\\int\\frac{dx}{\\sqrt{x^2+a^2}} = \\operatorname{arsinh}\\left(\\frac{x}{a}\\right)+C, \\qquad a>0.\\]\n\nThe substitution \\(x=a\\sinh u\\) simplifies the square root because \\(1+\\sinh^2u=\\cosh^2u\\)."
    },
    {
      "heading": "Complex arguments",
      "text": "For a complex number \\(z\\), the exponential definitions remain valid. Hyperbolic and trigonometric functions are connected by\n\n\\[\\sinh(iz)=i\\sin z, \\qquad \\cosh(iz)=\\cos z,\\]\n\nor equivalently,\n\n\\[\\sin(iz)=i\\sinh z, \\qquad \\cos(iz)=\\cosh z.\\]\n\nThese relationships connect the two families within complex analysis, rather than merely establishing a formal analogy.\n\nBoth \\(\\sinh z\\) and \\(\\cosh z\\) are entire: they are complex differentiable throughout the complex plane. Their zeros are\n\n\\[\\sinh z=0\\quad\\Longleftrightarrow\\quad z=n\\pi i,\\]\n\n\\[\\cosh z=0\\quad\\Longleftrightarrow\\quad z=\\left(n+\\frac12\\right)\\pi i, \\qquad n\\in\\mathbb Z.\\]\n\nThe reciprocal and quotient functions are meromorphic, with poles at the relevant denominator zeros. Although the basic functions have no nonzero real period, they satisfy\n\n\\[\\sinh(z+2\\pi i)=\\sinh z, \\qquad \\cosh(z+2\\pi i)=\\cosh z.\\]\n\nThe function \\(\\tanh z\\) has period \\(\\pi i\\).\n\nComplex inverse hyperbolic functions involve choices of logarithm and square-root branches. Consequently, real inverse formulas cannot be extended to all complex arguments without specifying branch conventions."
    },
    {
      "heading": "Applications",
      "text": "Differential equations\n\nHyperbolic sine and cosine form a basis for the solutions of the ordinary differential equation\n\n\\[y''=y.\\]\n\nEvery solution has the form\n\n\\[y(x)=A\\cosh x+B\\sinh x,\\]\n\nwhere \\(A=y(0)\\) and \\(B=y'(0)\\).\n\nMore generally, for \\(k>0\\), solutions of \\(y''=k^2y\\) can be written using \\(\\cosh(kx)\\) and \\(\\sinh(kx)\\). These forms are often more convenient than exponentials when imposing symmetric or antisymmetric boundary conditions.\n\nHyperbolic functions also occur when separation of variables reduces a partial differential equation, such as Laplace's equation, to ordinary equations with exponential solutions.\n\nHanging chains and surfaces\n\nA uniform, perfectly flexible chain suspended between two supports under uniform gravity forms a catenary, with equation\n\n\\[y=a\\cosh\\left(\\frac{x-x_0}{a}\\right)+y_0, \\qquad a>0.\\]\n\nThis curve is not a parabola, although the two can look similar near their lowest points. Rotating a catenary about its appropriate axis produces a catenoid, a classical minimal surface.\n\nSpecial relativity\n\nIn special relativity, a one-dimensional velocity \\(v\\) can be expressed through the rapidity \\(\\phi\\):\n\n\\[\\frac vc=\\tanh\\phi, \\qquad \\gamma=\\cosh\\phi, \\qquad \\gamma\\frac vc=\\sinh\\phi,\\]\n\nwhere \\(c\\) is the speed of light and\n\n\\[\\gamma=\\frac{1}{\\sqrt{1-v^2/c^2}}.\\]\n\nCollinear rapidities add, so the hyperbolic tangent addition formula yields the relativistic velocity-addition law. Lorentz transformations act as hyperbolic rotations preserving a difference of squares.\n\nNonlinear models\n\nHyperbolic tangent is a bounded, smooth, odd function used in nonlinear models and artificial neural networks. It is related to the logistic function\n\n\\[\\sigma(x)=\\frac{1}{1+e^{-x}}\\]\n\nby\n\n\\[\\tanh x=2\\sigma(2x)-1.\\]\n\nIts derivative, \\(1-\\tanh^2x\\), becomes small for large \\(|x|\\). In neural networks, this saturation can contribute to vanishing gradients."
    },
    {
      "heading": "Numerical evaluation",
      "text": "Direct exponential formulas are mathematically simple but can be problematic in floating-point arithmetic.\n\nFor small \\(|x|\\), evaluating \\(e^x-e^{-x}\\) subtracts nearly equal numbers and can lose relative precision. Reliable implementations use local series, specialized elementary-function routines, or formulas based on \\(\\operatorname{expm1}(x)=e^x-1\\), which is designed for accurate evaluation near zero.\n\nFor large \\(|x|\\), direct evaluation of the quotient defining \\(\\tanh x\\) can overflow even though its result always lies between \\(-1\\) and \\(1\\). For \\(x\\ge0\\), the equivalent expression\n\n\\[\\tanh x=\\frac{1-e^{-2x}}{1+e^{-2x}}\\]\n\navoids exponentially large intermediate values; negative arguments can be handled by odd symmetry. Near zero, a different evaluation method is preferable to avoid cancellation in the numerator.\n\nInverse functions also require care. For example, directly evaluating the logarithmic formula for \\(\\operatorname{arsinh}x\\) at a large negative \\(x\\) can involve severe cancellation. Specialized routines select formulas according to the argument's magnitude and sign to improve numerical stability."
    },
    {
      "heading": "Historical development",
      "text": "The theory of hyperbolic functions developed in the eighteenth century, notably through the work of Vincenzo Riccati and Johann Heinrich Lambert. Their treatment established a systematic counterpart to circular trigonometry based on the geometry of the hyperbola.\n\nThe exponential definitions later made the functions standard tools in mathematical analysis. Their role expanded through differential equations, complex function theory, engineering, and the hyperbolic geometry of Lorentz transformations."
    }
  ],
  "links": [
    "mathematical-function",
    "trigonometric-functions",
    "hyperbola",
    "exponential-function",
    "real-number",
    "circle",
    "radian",
    "derivative",
    "antiderivative",
    "power-series",
    "inverse-function",
    "natural-logarithm",
    "complex-number",
    "complex-analysis",
    "ordinary-differential-equation",
    "partial-differential-equation",
    "laplace-equation",
    "parabola",
    "special-relativity",
    "speed-of-light",
    "artificial-neural-network",
    "logistic-function",
    "floating-point-arithmetic",
    "numerical-stability"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}