{
  "id": "hyperbola",
  "title": "Hyperbola",
  "lang": "en",
  "source_lang": "en",
  "summary": "A hyperbola is a two-branched conic defined by a constant difference of distances from two foci, with applications in geometry, physics, and navigation.",
  "category": "mathematics",
  "subcategory": "geometry",
  "type": "concept",
  "aliases": [
    "Hyperbolic conic"
  ],
  "generated_at": "2026-11-03",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/hyperbola/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/hyperbola/",
      "json": "https://aiwiki.page/zh/hyperbola.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "A hyperbola is a plane curve consisting of two separate, unbounded branches. In Euclidean geometry, it is the set of points for which the absolute difference of the distances to two fixed points, called foci, is constant. It is also a conic section: an appropriate plane cutting both halves of a double cone produces a hyperbola. Together with the ellipse and parabola, it is one of the principal noncircular conics.\n\nThe standard equation \\(x^2/a^2-y^2/b^2=1\\) displays its central features: two branches opening in opposite directions, a center of symmetry, and two straight lines called asymptotes that the curve approaches at large distances. Hyperbolas connect classical geometric constructions with quadratic equations, projective transformations, and hyperbolic functions. They also describe inverse proportionality, idealized gravitational escape trajectories, and the position loci used in time-difference navigation."
    },
    {
      "heading": "Geometric definitions and fundamental elements",
      "text": "Let \\(F_1\\) and \\(F_2\\) be distinct points in the Euclidean plane, separated by a distance \\(2c\\). Choose a constant \\(2a\\) satisfying \\(0<a<c\\). The hyperbola with these foci is the locus of points \\(P\\) satisfying\n\n\\[\\bigl|PF_1-PF_2\\bigr|=2a.\\]\n\nThe restriction \\(a<c\\) follows from the triangle inequality and excludes degenerate limiting cases. Removing the absolute value separates the locus into two branches, according to which focal distance is larger. Each branch is connected, but the complete curve is not.\n\nThe midpoint of the foci is the center. The line through the foci is the transverse axis, and the perpendicular line through the center is the conjugate axis. With the center at the origin and the foci at \\((\\pm c,0)\\), the vertices are \\((\\pm a,0)\\). The transverse axis therefore meets the curve in a segment of length \\(2a\\). Defining \\(b>0\\) by\n\n\\[b^2=c^2-a^2\\]\n\ngives the canonical equation\n\n\\[\\frac{x^2}{a^2}-\\frac{y^2}{b^2}=1.\\]\n\nUnlike the semiminor axis of an ellipse, the length \\(b\\) is not a distance from the center to a point where the conjugate axis meets the curve: that axis does not intersect the hyperbola at all.\n\nA second characterization uses a focus and a fixed line, the directrix. The distance from a point of the curve to the focus is \\(e\\) times its perpendicular distance to the corresponding directrix, where the eccentricity is\n\n\\[e=\\frac ca>1.\\]\n\nFor the horizontal canonical hyperbola, the directrices are \\(x=\\pm a/e\\), paired with the foci \\((\\pm c,0)\\). The same focus–directrix definition gives an ellipse when \\(0<e<1\\) and a parabola when \\(e=1\\). Eccentricity therefore expresses a genuine geometric distinction among the conics, rather than merely describing how elongated a curve appears.\n\nA line through a focus perpendicular to the transverse axis cuts its branch in a chord called a latus rectum. Substituting \\(x=c\\) into the canonical equation gives endpoints with \\(y=\\pm b^2/a\\), so its length is \\(2b^2/a\\). For example, the hyperbola \\(x^2/9-y^2/16=1\\) has \\(a=3\\), \\(b=4\\), \\(c=5\\), eccentricity \\(5/3\\), directrices \\(x=\\pm9/5\\), and latus rectum length \\(32/3\\). These measurements determine both the focal geometry and the opening of the branches."
    },
    {
      "heading": "Cartesian equations and classification",
      "text": "The methods of analytic geometry express a general conic as\n\n\\[Ax^2+Bxy+Cy^2+Dx+Ey+F=0.\\]\n\nIts quadratic part can be represented by the symmetric matrix\n\n\\[Q= \\begin{pmatrix} A&B/2\\\\ B/2&C \\end{pmatrix}.\\]\n\nFor a nondegenerate real hyperbola, the decisive condition is\n\n\\[B^2-4AC>0.\\]\n\nEquivalently, \\(Q\\) has two nonzero eigenvalues of opposite signs. Geometrically, this means that the quadratic expression takes both positive and negative values along different directions.\n\nThe discriminant condition alone does not exclude a pair of intersecting lines. Nondegeneracy is tested using\n\n\\[M= \\begin{pmatrix} A&B/2&D/2\\\\ B/2&C&E/2\\\\ D/2&E/2&F \\end{pmatrix}.\\]\n\nThe conic is nondegenerate when \\(\\det M\\ne0\\), where \\(\\det\\) denotes the determinant. Thus \\(x^2-y^2=1\\) is a hyperbola, whereas \\(x^2-y^2=0\\) is the degenerate union of the lines \\(y=x\\) and \\(y=-x\\). The distinction matters whenever a conic is identified from coefficients rather than from an already normalized equation.\n\nThe center \\((h,k)\\) is obtained by solving\n\n\\[2Ah+Bk+D=0,\\qquad Bh+2Ck+E=0.\\]\n\nBecause \\(4AC-B^2\\ne0\\), this system has a unique solution. Translating the origin to \\((h,k)\\) removes the linear terms. A rotation then diagonalizes the quadratic part, eliminating the mixed product. The resulting equation takes one of the forms\n\n\\[\\frac{X^2}{a^2}-\\frac{Y^2}{b^2}=1 \\quad\\text{or}\\quad \\frac{Y^2}{b^2}-\\frac{X^2}{a^2}=1,\\]\n\nwith positive denominators. These transformations preserve distances and angles, so focal lengths and eccentricity can be read from the normalized equation without changing their geometric meaning.\n\nThe mixed term \\(Bxy\\) therefore indicates the orientation of the coordinate axes, not a fundamentally different kind of hyperbola. For instance, \\(xy=1\\) becomes a standard hyperbola after a rotation through \\(45^\\circ\\). More generally, a centered quadratic equation can be studied through the spectral theorem for real symmetric matrices: the eigenvectors give perpendicular principal-axis directions, and the eigenvalues determine the scaling along them.\n\nIn practical calculations, classification and normalization are distinct tasks. The discriminant identifies the type, the determinant detects degeneracy, and translation and rotation locate and orient the curve. Multiplying the entire equation by any nonzero constant changes none of these geometric properties."
    },
    {
      "heading": "Asymptotes, conjugate curves, and rectangular hyperbolas",
      "text": "An asymptote is a line that a branch approaches indefinitely. For\n\n\\[\\frac{x^2}{a^2}-\\frac{y^2}{b^2}=1,\\]\n\nthe asymptotes are\n\n\\[y=\\frac ba x,\\qquad y=-\\frac ba x.\\]\n\nThey arise by replacing the right-hand side of the canonical equation by zero and factoring the resulting homogeneous quadratic expression. Their intersection is the center, and their slopes describe the limiting directions of the branches. A nondegenerate hyperbola never meets either of its own asymptotes at a finite point.\n\nThe approach can be quantified. On the upper-right part of the curve,\n\n\\[y=\\frac ba x\\sqrt{1-\\frac{a^2}{x^2}}.\\]\n\nExpanding for large positive \\(x\\) gives\n\n\\[y=\\frac ba x-\\frac{ab}{2x}+O(x^{-3}).\\]\n\nThus the vertical difference between the curve and its asymptote decreases in proportion to \\(1/x\\) to leading order. This is a simple example of an asymptotic expansion. For a translated or rotated hyperbola, the asymptotes are found in the same way after centering the equation: the constant term is discarded, and the remaining quadratic equation describes two lines through the center.\n\nThe conjugate hyperbola of the canonical curve is\n\n\\[\\frac{y^2}{b^2}-\\frac{x^2}{a^2}=1.\\]\n\nIt has the same center and asymptotes, but its branches open into the opposite pair of angular regions. The transverse and conjugate axes exchange roles. The two curves are distinct and have no real intersection, yet together they show how a fixed pair of asymptotes can organize two complementary families of branches. More generally, centered equations with the same homogeneous quadratic part and different nonzero constant terms describe related hyperbolas sharing those asymptotes.\n\nA hyperbola is rectangular, or equilateral, when its asymptotes are perpendicular. In canonical coordinates this occurs precisely when \\(a=b\\), giving eccentricity \\(\\sqrt2\\). Rotating the coordinate axes transforms its equation into\n\n\\[uv=k,\\qquad k\\ne0.\\]\n\nConsequently, the familiar graph \\(y=k/x\\) is a rectangular hyperbola. Its coordinate axes are its asymptotes, and its two branches occupy either the first and third quadrants or the second and fourth, depending on the sign of \\(k\\).\n\nFor a general quadratic equation in orthonormal Cartesian coordinates, rectangularity is equivalent to \\(A+C=0\\), provided the equation represents a nondegenerate hyperbola. This trace condition is unchanged by rotation. Rectangular hyperbolas are especially important because inverse proportionality appears throughout elementary science: fixing the product of two variables produces precisely this geometry, although the numerical angle between plotted axes depends on the physical units and scale chosen."
    },
    {
      "heading": "Tangents, curvature, and reflection",
      "text": "At a point \\((x_0,y_0)\\) on the canonical hyperbola, the tangent line is\n\n\\[\\frac{xx_0}{a^2}-\\frac{yy_0}{b^2}=1.\\]\n\nThis follows by differentiating the defining equation or by using its gradient. If\n\n\\[f(x,y)=\\frac{x^2}{a^2}-\\frac{y^2}{b^2}-1,\\]\n\nthen \\(\\nabla f=(2x/a^2,-2y/b^2)\\) is normal to the curve. Substituting this normal into the point–normal equation of a line yields the tangent formula. It remains valid at the vertices, where the tangents are vertical and cannot be written using a finite slope.\n\nAway from the vertices, implicit differentiation gives\n\n\\[\\frac{dy}{dx}=\\frac{b^2x}{a^2y}.\\]\n\nA line written as \\(y=mx+d\\) is tangent precisely when substitution into the conic produces a repeated intersection point. The resulting condition is\n\n\\[d^2=a^2m^2-b^2.\\]\n\nA finite tangent of this form therefore requires \\(|m|>b/a\\). At equality the line is an asymptote, not a tangent at a finite point. This distinction illustrates why a limiting tangent direction need not correspond to an ordinary point of contact.\n\nThe local bending of the curve is described by curvature. Using\n\n\\[x=a\\cosh t,\\qquad y=b\\sinh t\\]\n\nfor the right branch gives\n\n\\[\\kappa(t)= \\frac{ab} {\\left(a^2\\sinh^2t+b^2\\cosh^2t\\right)^{3/2}}.\\]\n\nThe curvature is greatest at the vertex, where \\(\\kappa=a/b^2\\), and tends to zero along either end. The radius of curvature at a vertex is therefore \\(b^2/a\\), half the latus rectum length. This connects a local differential property with the focal dimensions introduced in elementary geometry.\n\nThe focal definition also produces an optical property. On a branch where the difference of focal distances is constant, let \\(\\mathbf u_1\\) and \\(\\mathbf u_2\\) be unit vectors from the two foci toward a point \\(P\\). Differentiating the distance difference shows that a normal is parallel to \\(\\mathbf u_1-\\mathbf u_2\\). Reflection across the tangent therefore interchanges the directions \\(\\mathbf u_1\\) and \\(\\mathbf u_2\\). A ray traveling from one focus to the surface is reflected along a line whose backward extension passes through the other focus, so that the latter acts as a virtual focus in this arrangement.\n\nA reflecting surface formed by rotating a suitable hyperbolic arc retains the corresponding axial focal property. Such hyperboloidal mirrors are used as secondary mirrors in classical Cassegrain telescope designs. Their function differs from that of an ellipsoidal mirror: the relevant focal construction involves a virtual continuation of a ray rather than simply joining two real focal points."
    },
    {
      "heading": "Parametrizations, hyperbolic functions, and area",
      "text": "The canonical identity\n\n\\[\\cosh^2t-\\sinh^2t=1\\]\n\nprovides a natural parametrization:\n\n\\[x=a\\cosh t,\\qquad y=b\\sinh t,\\qquad -\\infty<t<\\infty.\\]\n\nIt covers the right branch exactly once. Replacing \\(x\\) by \\(-a\\cosh t\\) covers the left branch. The hyperbolic functions are defined through the exponential function:\n\n\\[\\cosh t=\\frac{e^t+e^{-t}}2,\\qquad \\sinh t=\\frac{e^t-e^{-t}}2.\\]\n\nTheir names express the same connection with the unit hyperbola that ordinary sine and cosine have with the unit circle.\n\nThis parameter has an area interpretation. On the unit hyperbola \\(x^2-y^2=1\\), the signed area of the sector bounded by the origin, the point \\((1,0)\\), the point \\((\\cosh t,\\sinh t)\\), and the intervening arc is\n\n\\[\\frac12\\int_0^t \\left(x\\frac{dy}{du}-y\\frac{dx}{du}\\right)\\,du =\\frac t2.\\]\n\nThe parameter \\(t\\) is consequently called the hyperbolic angle. For the scaled hyperbola, the corresponding area is \\(ab\\,t/2\\). Unlike a circular angle, a hyperbolic angle is unbounded even while describing just one branch.\n\nA different area relation links the rectangular hyperbola with the natural logarithm:\n\n\\[\\int_1^x\\frac{du}{u}=\\ln x,\\qquad x>0.\\]\n\nThe region under \\(y=1/x\\) between positive abscissas \\(r\\) and \\(s\\) has area \\(\\ln(s/r)\\). Scaling both endpoints by the same positive factor leaves this area unchanged. Multiplication of endpoint ratios thus becomes addition of areas, explaining geometrically the logarithm law \\(\\ln(xy)=\\ln x+\\ln y\\). This relation was historically important in the development of logarithms and integral calculus.\n\nHyperbolas also admit rational parametrizations. Taking a line through the vertex \\((-a,0)\\) gives\n\n\\[x=a\\frac{1+t^2}{1-t^2},\\qquad y=\\frac{2bt}{1-t^2}, \\qquad t\\ne\\pm1.\\]\n\nHere both coordinates are rational functions of \\(t\\). The omitted vertex is recovered by the limiting parameter \\(t=\\infty\\), while \\(t=\\pm1\\) corresponds to the two directions at infinity. Rational parametrization is useful for exact constructions and for computations in algebraic geometry.\n\nNot every integral associated with a hyperbola is equally elementary. Its arc length involves\n\n\\[\\int\\sqrt{a^2\\sinh^2t+b^2\\cosh^2t}\\,dt,\\]\n\nwhich generally requires an elliptic integral. Thus the simple polynomial equation and elementary parametrizations do not imply an elementary formula for distance measured along the curve. Sector area, area under a branch, and arc length are different geometric quantities with different analytic behavior."
    },
    {
      "heading": "Affine and projective interpretations",
      "text": "An invertible affine transformation preserves lines, parallelism, and tangency, but not generally distances or angles. Under such transformations, every nondegenerate real hyperbola is equivalent to every other one. Translation moves its center, rotation aligns its axes, and independent scaling along the two axes converts its equation to \\(X^2-Y^2=1\\). Another linear change gives \\(UV=1\\). Rectangularity and eccentricity are therefore Euclidean properties, not affine invariants: a shear or unequal scaling can change them while leaving the curve within the affine class of hyperbolas.\n\nProjective geometry places this distinction in a broader setting. In homogeneous coordinates, the canonical equation becomes\n\n\\[\\frac{X^2}{a^2}-\\frac{Y^2}{b^2}-Z^2=0,\\]\n\nwhere ordinary affine coordinates are \\(x=X/Z\\) and \\(y=Y/Z\\) when \\(Z\\ne0\\). Setting \\(Z=0\\) finds the points on the line at infinity:\n\n\\[[X:Y:Z]=[a:b:0],\\qquad[a:-b:0].\\]\n\nThese are two distinct real points, representing the asymptotic directions of the hyperbola.\n\nThis gives a precise projective interpretation of the familiar conic classification. Relative to a chosen real line at infinity, an ellipse has no real intersection with that line, a parabola has one repeated intersection, and a hyperbola has two distinct intersections. The asymptotes are the projective tangents at the hyperbola’s two points at infinity. An arbitrary projective transformation may move the line at infinity, so ellipse, parabola, and hyperbola are not different types of nondegenerate real conic in unrestricted projective geometry. Their distinction depends on how the projective plane is divided into a finite affine region and its line at infinity.\n\nThe tangent formula extends to a useful dual construction. For a point \\((x_0,y_0)\\), whether or not it lies on the conic, the line\n\n\\[\\frac{xx_0}{a^2}-\\frac{yy_0}{b^2}=1\\]\n\nis its polar, when it is a finite line. If two real tangents can be drawn from that point, their points of contact lie on the polar. For a point on the curve, the polar is its tangent. The center is exceptional in affine coordinates: its polar is the line at infinity in the projective completion.\n\nThe apparent separation of the hyperbola into two branches also depends on the affine viewpoint. Each branch is homeomorphic to the real line, but after adding the two points at infinity, the complete real projective conic is homeomorphic to a circle. Ends of the affine branches meet through those added points. Projective geometry thereby explains both the two asymptotes and the change in connectedness without treating infinity as an ordinary distant location."
    },
    {
      "heading": "Historical development",
      "text": "The study of conic sections began in ancient Greek geometry. Menaechmus, active in the fourth century BCE, is traditionally associated with their discovery and their use in the problem of doubling the cube. In modern coordinates, the intersection of \\(y=x^2\\) and \\(xy=2\\) gives \\(x^3=2\\), illustrating how a parabola and a hyperbola can construct a cube root. The Greeks did not formulate the problem in Cartesian notation, but intersections of conics supplied geometric constructions unavailable through straightedge and compass alone.\n\nApollonius of Perga, working around the late third and early second centuries BCE, developed the subject systematically in his Conics. The names ellipse, parabola, and hyperbola are associated with his treatment and with relations arising from the application of areas. “Hyperbola” derives from a Greek term connected with excess. Apollonius’s geometry established properties of diameters, tangents, and conic sections without relying on coordinate equations. The modern identification of a hyperbola as a plane section meeting both nappes of a double cone belongs to this classical framework.\n\nArchimedes studied quadrature problems involving conics and related surfaces, although the area under a rectangular hyperbola proved fundamentally different from the parabolic area he famously evaluated. In the seventeenth century, Grégoire de Saint-Vincent’s 1647 work on quadrature developed important properties of hyperbolic areas. Alfonso Antonio de Sarasa subsequently made explicit their connection with logarithms. This development helped reveal why equal ratios of abscissas determine equal areas under \\(y=1/x\\), a property central to the integral interpretation of the logarithm.\n\nThe rise of coordinate methods transformed the subject. René Descartes’s La Géométrie of 1637 and the work of Pierre de Fermat established approaches in which equations could define curves and geometric problems could be translated into algebra. Hyperbolas became both geometric loci and examples of second-degree equations. Later developments in calculus made it possible to calculate their tangents, curvature, and associated areas systematically.\n\nDuring the eighteenth century, Vincenzo Riccati and Johann Heinrich Lambert developed the theory of hyperbolic functions, organizing relations that parallel circular trigonometry. Nineteenth-century projective geometry then clarified that ellipse, parabola, and hyperbola are different affine appearances of the same projective kind of curve. In the twentieth century, hyperbolic parametrizations acquired a further physical interpretation through Lorentz transformations. Across these developments, the curve retained its classical focal definition while gaining algebraic, analytic, and transformational descriptions."
    },
    {
      "heading": "Physical and technical applications",
      "text": "In the ideal Newtonian two-body problem, a particle moving under an attractive inverse-square force follows a conic with the force center at a focus. Positive specific orbital energy produces a hyperbolic orbit, zero energy a parabolic orbit, and negative energy a bound elliptic orbit. If \\(\\mu\\) is the gravitational parameter, \\(\\varepsilon\\) the specific orbital energy, and \\(h\\) the specific angular momentum, then\n\n\\[e=\\sqrt{1+\\frac{2\\varepsilon h^2}{\\mu^2}}.\\]\n\nFor \\(\\varepsilon>0\\), this gives \\(e>1\\). The incoming and outgoing asymptotes describe the directions of motion far from the attracting body.\n\nWriting \\(a>0\\) for the magnitude of the hyperbolic semimajor parameter, the excess speed at infinity satisfies\n\n\\[v_\\infty=\\sqrt{\\frac{\\mu}{a}}.\\]\n\nThe gravitational deflection angle is\n\n\\[\\delta=2\\arcsin\\!\\left(\\frac1e\\right).\\]\n\nThese relations are used in approximations to spacecraft flybys and escape trajectories. A real spacecraft is influenced by multiple bodies, so a hyperbola is usually a local two-body model rather than an exact description of its entire journey. The distinction between a geometric branch and a physical trajectory also matters: a single encounter follows one branch, with the attracting body at the appropriate focus.\n\nHyperbolas arise in time-difference-of-arrival positioning. Suppose two synchronized receivers at \\(F_1\\) and \\(F_2\\) detect a signal propagating at speed \\(v\\). A measured arrival-time difference \\(\\Delta t\\) implies\n\n\\[PF_1-PF_2=v\\Delta t.\\]\n\nIn a plane, a nonzero difference smaller in magnitude than the receiver separation defines one hyperbola branch. Several receiver pairs provide intersecting loci from which the source position can be estimated. The historical LORAN radio-navigation system used this principle. Measurement noise turns exact curve intersections into an estimation problem; spatial configurations use corresponding difference-of-distance surfaces rather than plane curves.\n\nIn special relativity, the invariant interval produces hyperbolic geometry in a spacetime diagram. With one spatial coordinate \\(x\\), events a fixed proper time \\(\\tau\\) from the origin satisfy\n\n\\[(ct)^2-x^2=c^2\\tau^2,\\]\n\nwhere \\(c\\) is the speed of light. Lorentz boosts preserve this hyperbola and can be written using a rapidity \\(\\eta\\), with coefficients \\(\\cosh\\eta\\) and \\(\\sinh\\eta\\). The associated velocity satisfies \\(v/c=\\tanh\\eta\\). The role played by ordinary rotations for circles is thus played by hyperbolic rotations for the Minkowski interval.\n\nInverse proportionality supplies a simpler but widespread application. For a fixed amount of ideal gas at constant absolute temperature, \\(pV\\) is constant, so a pressure–volume graph is a rectangular hyperbola. Similar curves arise whenever a fixed total is expressed as a product of two changing factors. Such examples are exact hyperbolas only under the stated model and choice of variables. More broadly, describing a trend as “hyperbolic” may indicate a reciprocal or rapid-growth form without establishing that the plotted relation is a geometric conic."
    }
  ],
  "links": [
    "euclidean-geometry",
    "conic-section",
    "ellipse",
    "parabola",
    "triangle-inequality",
    "analytic-geometry",
    "matrix",
    "eigenvalues-and-eigenvectors",
    "determinant",
    "spectral-theorem",
    "asymptote",
    "asymptotic-expansion",
    "gradient",
    "hyperbolic-functions",
    "exponential-function",
    "natural-logarithm",
    "algebraic-geometry",
    "affine-map",
    "projective-geometry",
    "homogeneous-coordinates",
    "archimedes",
    "pierre-de-fermat",
    "speed-of-light"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}