{
  "id": "ellipse",
  "title": "Ellipse",
  "lang": "en",
  "source_lang": "en",
  "summary": "An ellipse is a closed conic defined by a constant sum of distances to two foci, fundamental to geometry, orbital mechanics, optics, and statistics.",
  "category": "mathematics",
  "subcategory": "geometry",
  "type": "concept",
  "aliases": [
    "Elliptic curve (elementary geometry)"
  ],
  "generated_at": "2026-11-03",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/ellipse/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/ellipse/",
      "json": "https://aiwiki.page/zh/ellipse.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "An ellipse is a closed plane curve consisting of all points whose distances from two fixed points, called its foci, have a constant sum. It is one of the conic sections: a suitable plane cutting a circular cone produces an ellipse. A circle is the special case in which the two foci coincide. Ellipses also arise as nonsingular affine images of circles, as level curves of positive-definite quadratic expressions, and as idealized orbits in the gravitational two-body problem.\n\nThe ellipse connects classical geometry with linear algebra, calculus, and physical modeling. Its focal definition explains its reflection properties; its quadratic equation supports precise calculations; and its matrix representation describes orientation and directional scale. These different descriptions underlie applications ranging from astronomical prediction and optical design to computer graphics and statistical uncertainty."
    },
    {
      "heading": "Foci, axes, and eccentricity",
      "text": "Let \\(F_1\\) and \\(F_2\\) be fixed points separated by a distance \\(2c\\). An ellipse is the set of points \\(P\\) satisfying\n\n\\[|PF_1|+|PF_2|=2a, \\qquad a>c\\geq 0.\\]\n\nThe condition \\(a>c\\) ensures a nondegenerate closed curve. By the triangle inequality, the distance sum cannot be less than \\(2c\\). If \\(a=c>0\\), the locus collapses to the line segment joining the foci; if \\(a<c\\), it is empty. In the usual terminology, “ellipse” means the nondegenerate curve, although degenerate cases are useful when classifying quadratic equations.\n\nThe midpoint of the foci is the center. For distinct foci, the line through them is the major axis, and its perpendicular through the center is the minor axis. The corresponding semiaxis lengths are \\(a\\) and \\(b\\), where\n\n\\[b^2=a^2-c^2.\\]\n\nThe major-axis endpoints, or major vertices, lie a distance \\(a\\) from the center; the minor-axis endpoints lie a distance \\(b\\) from it. The full axis lengths are therefore \\(2a\\) and \\(2b\\). Every noncircular ellipse has reflection symmetry in both axes and rotational symmetry through \\(180^\\circ\\). A circle has the additional property that every diameter is an axis of symmetry.\n\nThe dimensionless eccentricity is\n\n\\[e=\\frac{c}{a} =\\sqrt{1-\\frac{b^2}{a^2}}, \\qquad 0\\leq e<1.\\]\n\nIt measures departure from circularity independently of overall size. When \\(e=0\\), the curve is a circle. As \\(e\\) approaches \\(1\\) with \\(a\\) fixed, the minor semiaxis tends to zero and the curve approaches a segment. For example, an ellipse with \\(a=5\\) and \\(b=3\\) has \\(c=4\\) and \\(e=0.8\\). Eccentricity is not the same quantity as the axis ratio \\(b/a\\), although either determines the other.\n\nFor \\(e>0\\), an equivalent definition uses a focus and a directrix. With the center at the origin and the major axis horizontal, the foci are \\((\\pm ae,0)\\), and the corresponding directrices are \\(x=\\pm a/e\\). Every point on the ellipse satisfies\n\n\\[\\frac{\\text{distance to a focus}} {\\text{distance to its corresponding directrix}}=e.\\]\n\nThis definition places the ellipse in a common family with the parabola, whose eccentricity is \\(1\\), and the hyperbola, whose eccentricity exceeds \\(1\\). The circle is obtained as a limiting case: as \\(e\\to0\\), the directrices recede without bound."
    },
    {
      "heading": "Cartesian, parametric, and matrix equations",
      "text": "In a Cartesian coordinate system aligned with the axes, an ellipse centered at the origin has equation\n\n\\[\\frac{x^2}{a^2}+\\frac{y^2}{b^2}=1, \\qquad a\\geq b>0.\\]\n\nThis follows from the focal definition by isolating and squaring the distance expressions, using \\(c^2=a^2-b^2\\). Conversely, points satisfying this equation have focal distance sum \\(2a\\). For a center \\((h,k)\\) with the same axis directions, \\(x\\) and \\(y\\) are replaced by \\(x-h\\) and \\(y-k\\). The filled elliptical region is described by replacing equality with \\(\\leq\\).\n\nA useful parametrization is\n\n\\[x=a\\cos t,\\qquad y=b\\sin t, \\qquad 0\\leq t<2\\pi.\\]\n\nThe parameter \\(t\\) is the angle on an auxiliary circle before one coordinate is scaled. It is generally neither the polar angle of the ellipse point nor a parameter proportional to arc length. Indeed,\n\n\\[\\frac{ds}{dt} =\\sqrt{a^2\\sin^2t+b^2\\cos^2t},\\]\n\nwhich varies unless \\(a=b\\). A rational parametrization, obtained from the half-angle substitution \\(u=\\tan(t/2)\\), is\n\n\\[x=a\\frac{1-u^2}{1+u^2}, \\qquad y=b\\frac{2u}{1+u^2}.\\]\n\nAs \\(u\\) ranges over the real numbers, it traces the ellipse except for \\((-a,0)\\), which appears as a limit. This description is useful for exact calculations because the coordinates are rational functions of the parameter.\n\nA general second-degree equation is\n\n\\[Ax^2+Bxy+Cy^2+Dx+Ey+F=0.\\]\n\nThe inequality \\(B^2-4AC<0\\) identifies the elliptic type, but does not alone guarantee a real nondegenerate ellipse: the solution set can also be a point or empty. A more complete test uses a symmetric matrix:\n\n\\[\\mathbf{x}^{T}Q\\mathbf{x} +\\mathbf{d}^{T}\\mathbf{x}+f=0.\\]\n\nAfter multiplying the equation by \\(-1\\) if necessary, suppose \\(Q\\) is a positive-definite matrix. Completing the square gives\n\n\\[(\\mathbf{x}-\\mathbf{x}_0)^TQ(\\mathbf{x}-\\mathbf{x}_0)=k,\\]\n\nwhere\n\n\\[\\mathbf{x}_0=-\\frac12Q^{-1}\\mathbf{d}, \\qquad k=\\frac14\\mathbf{d}^TQ^{-1}\\mathbf{d}-f.\\]\n\nThere is a real nondegenerate ellipse exactly when \\(k>0\\). If \\(k=0\\), only the center remains; if \\(k<0\\), there are no real points.\n\nThe spectral theorem supplies an orthonormal basis of eigenvectors of \\(Q\\). These give the principal-axis directions. If its eigenvalues are \\(0<\\lambda_1\\leq\\lambda_2\\), then\n\n\\[a=\\sqrt{\\frac{k}{\\lambda_1}}, \\qquad b=\\sqrt{\\frac{k}{\\lambda_2}}.\\]\n\nThus the smaller eigenvalue corresponds to the longer semiaxis. This formulation handles translated and rotated ellipses without requiring separate geometric constructions."
    },
    {
      "heading": "Affine geometry, projections, and constructions",
      "text": "Every ellipse is the image of a circle under an invertible affine map. Starting with the unit circle, the transformation\n\n\\[\\begin{pmatrix}x\\\\y\\end{pmatrix} = \\mathbf{x}_0+ R \\begin{pmatrix}a&0\\\\0&b\\end{pmatrix} \\begin{pmatrix}\\cos t\\\\\\sin t\\end{pmatrix}\\]\n\nproduces an ellipse with center \\(\\mathbf{x}_0\\), semiaxes \\(a,b\\), and orientation determined by the rotation matrix \\(R\\). Conversely, translating to the center, rotating to the principal axes, and dividing the coordinates by the semiaxis lengths transforms any ellipse into the unit circle.\n\nAffine transformations preserve straight lines, parallelism, and ratios along a line, but do not generally preserve angles or lengths. Consequently, many circle theorems have affine counterparts for ellipses, while properties involving perpendicularity require special care. A diameter of an ellipse is a chord through its center. Two diameters are conjugate if each is parallel to the tangents at the endpoints of the other. With the parametrization above, the endpoint vectors\n\n\\[(a\\cos t,b\\sin t) \\quad\\text{and}\\quad (-a\\sin t,b\\cos t)\\]\n\ndefine conjugate diameters. They need not be perpendicular; the principal axes form a perpendicular conjugate pair.\n\nParallel projection gives a direct geometric realization. The orthogonal projection of a circle of radius \\(R\\), whose plane makes an angle \\(\\alpha\\) with the projection plane, has semiaxes \\(R\\) and \\(R|\\cos\\alpha|\\). For \\(0<\\alpha<90^\\circ\\), the result is a noncircular ellipse; at \\(90^\\circ\\), it degenerates to a segment. This explains why a circular rim usually appears elliptical in an orthographic drawing. Perspective projection instead belongs to projective geometry: it preserves conics under suitable nondegeneracy conditions, but not their Euclidean classification in every affine view.\n\nSeveral drawing methods implement different definitions. In the string-and-pins construction, pins mark the foci and a taut string constrains the sum of the two focal distances. The construction is exact in ideal geometry, although actual results depend on string stretch, pin size, and pencil thickness. An elliptic trammel uses a rigid bar with two points constrained to slide along perpendicular lines; another fixed point on the bar traces an ellipse. Its motion resolves into two sinusoidal coordinates with different amplitudes.\n\nThe connection with a cone can be demonstrated through Dandelin spheres. Two spheres tangent to both the cone and its cutting plane touch the plane at the ellipse’s foci. Equal tangent lengths from a point on the section establish the constant-sum property. Introduced by Germinal Pierre Dandelin in 1822, this construction supplies a geometric bridge between the spatial cone-section definition and the planar focal definition, which otherwise look quite different."
    },
    {
      "heading": "Tangents, normals, and curvature",
      "text": "At a point \\(P=(x_0,y_0)\\) on the standard ellipse, the tangent line is\n\n\\[\\frac{x_0x}{a^2}+\\frac{y_0y}{b^2}=1.\\]\n\nIt can be obtained by implicit differentiation or from the gradient of\n\n\\[F(x,y)=\\frac{x^2}{a^2}+\\frac{y^2}{b^2}-1.\\]\n\nThe vector\n\n\\[\\nabla F(P)= \\left(\\frac{2x_0}{a^2},\\frac{2y_0}{b^2}\\right)\\]\n\nis perpendicular to the tangent and gives an outward normal direction. In particular, an ellipse’s normal does not generally pass through its center. That familiar property of circles survives only at the principal-axis endpoints of a noncircular ellipse.\n\nThe tangent formula also identifies the support lines of the filled ellipse, a convex set. For a nonzero vector \\(\\mathbf{n}=(u,v)\\), the maximum of \\(ux+vy\\) over the centered ellipse is\n\n\\[h(\\mathbf{n})=\\sqrt{a^2u^2+b^2v^2}.\\]\n\nThis follows by inserting the parametrization and maximizing \\(au\\cos t+bv\\sin t\\), or by the Cauchy–Schwarz inequality. If \\(\\mathbf{n}\\) is a unit vector, \\(h(\\mathbf{n})\\) is the distance from the center to the corresponding support line. The width in that direction is \\(2h(\\mathbf{n})\\). For an ellipse expressed as \\(\\mathbf{x}^{T}Q\\mathbf{x}\\leq1\\), the same quantity is \\(\\sqrt{\\mathbf{n}^{T}Q^{-1}\\mathbf{n}}\\).\n\nThe curvature at parameter value \\(t\\) is\n\n\\[\\kappa(t)= \\frac{ab} {\\left(a^2\\sin^2t+b^2\\cos^2t\\right)^{3/2}}.\\]\n\nFor \\(a>b\\), curvature is greatest at the major-axis endpoints, where it equals \\(a/b^2\\), and smallest at the minor-axis endpoints, where it equals \\(b/a^2\\). Thus the long-axis ends are locally sharper, despite lying farther from the center. The corresponding radii of curvature are \\(b^2/a\\) and \\(a^2/b\\). Unlike an informal description of “flatness,” curvature gives a local, quantitative measure.\n\nThe centers of the osculating circles trace the ellipse’s evolute, given parametrically by\n\n\\[X=\\frac{a^2-b^2}{a}\\cos^3t, \\qquad Y=\\frac{b^2-a^2}{b}\\sin^3t.\\]\n\nFor a noncircular ellipse it has four cusps, corresponding to the four axis endpoints. The evolute records how the normal lines intersect locally and where the radius of curvature becomes stationary. These properties are important when constructing offset curves: moving a fixed distance along the normal generally does not produce another ellipse, and offsets can develop singularities when the displacement reaches a local radius of curvature."
    },
    {
      "heading": "Area, arc length, and elliptic integrals",
      "text": "The area enclosed by an ellipse is\n\n\\[A=\\pi ab.\\]\n\nOne proof uses the linear scaling \\((u,v)\\mapsto(au,bv)\\), which sends the unit disk to the elliptical region. Its determinant is \\(ab\\), so every area is multiplied by that factor. Another proof integrates the upper and lower halves of the Cartesian equation. The area formula concerns the enclosed region; the ellipse itself, regarded only as a curve, has zero planar area.\n\nThe perimeter is more complicated. From the parametric speed,\n\n\\[L= 4\\int_0^{\\pi/2} \\sqrt{a^2\\sin^2t+b^2\\cos^2t}\\,dt.\\]\n\nWriting \\(e^2=1-b^2/a^2\\) and changing the integration variable gives\n\n\\[L=4aE(e),\\]\n\nwhere\n\n\\[E(e)=\\int_0^{\\pi/2}\\sqrt{1-e^2\\sin^2\\theta}\\,d\\theta\\]\n\nis the complete elliptic integral of the second kind, using the modulus convention. Some references instead use the parameter \\(m=e^2\\), an important distinction when comparing formulas or software interfaces. Except in special cases such as the circle, ellipse arc length cannot be expressed using elementary functions.\n\nA small-eccentricity expansion is\n\n\\[L=2\\pi a \\left( 1-\\frac{e^2}{4} -\\frac{3e^4}{64} -\\frac{5e^6}{256} -\\cdots \\right).\\]\n\nThis power series is useful near a circle, though increasingly many terms can be needed as eccentricity approaches \\(1\\). An accurate approximation associated with Srinivasa Ramanujan is\n\n\\[L\\approx \\pi(a+b) \\left( 1+\\frac{3h}{10+\\sqrt{4-3h}} \\right), \\qquad h=\\left(\\frac{a-b}{a+b}\\right)^2.\\]\n\nIt is symmetric in \\(a,b\\), gives the exact circle circumference, and remains close to the true perimeter for very elongated ellipses. It is nevertheless an approximation, not a replacement for the elliptic integral when certified accuracy is required.\n\nA complete perimeter is simpler to specify than equal-length subdivision. Equal increments of \\(t\\) do not give equal arc lengths. To position points at uniform distances along the curve, numerical methods evaluate an incomplete arc-length integral and invert it. The positive speed makes this inversion monotone, allowing reliable bracketing methods as well as Newton’s method when suitable initial estimates are available.\n\nThe limiting cases provide useful checks. When \\(a=b\\), the perimeter is \\(2\\pi a\\). As \\(b\\to0\\) with \\(a\\) fixed, it tends to \\(4a\\), corresponding to traversal along the limiting segment in both directions. The appearance of elliptic integrals in this elementary geometric problem was historically important: attempts to understand such arc lengths helped motivate a much broader theory of special functions."
    },
    {
      "heading": "Elliptical orbits and time-dependent motion",
      "text": "An ellipse is the bound, noncollision orbit of the ideal gravitational two-body problem. In the relative motion of two point masses, the inverse-square force gives a conic with one focus at the force center. For negative specific orbital energy and nonzero angular momentum, that conic is an ellipse. In a barycentric description, both masses follow scaled ellipses about their common center of mass; neither body is generally fixed.\n\nWith a focus as the origin, polar coordinates give the orbit as\n\n\\[r=\\frac{p}{1+e\\cos\\nu}, \\qquad p=a(1-e^2)=\\frac{b^2}{a}.\\]\n\nHere \\(\\nu\\) is the true anomaly, measured from closest approach, and \\(p\\) is the semilatus rectum. The minimum and maximum distances are\n\n\\[r_{\\min}=a(1-e), \\qquad r_{\\max}=a(1+e).\\]\n\nThe force center is at a focus, not at the geometric center. This distinction is essential: a body moving at uniform angular speed about the ellipse’s center would not follow the gravitational timing law.\n\nKepler’s laws connect the geometry to time. Equal areas are swept out in equal times, expressing conservation of angular momentum. If \\(\\mu=G(m_1+m_2)\\), the period is\n\n\\[T=2\\pi\\sqrt{\\frac{a^3}{\\mu}}.\\]\n\nThus, for a fixed \\(\\mu\\), ellipses with the same semimajor axis have the same period even if their eccentricities differ. Their specific orbital energy and speed satisfy\n\n\\[\\varepsilon=-\\frac{\\mu}{2a}, \\qquad v^2=\\mu\\left(\\frac{2}{r}-\\frac{1}{a}\\right).\\]\n\nThe latter is the vis-viva equation. It shows that speed is greatest at closest approach and least at greatest distance.\n\nTime along the orbit is commonly computed through the eccentric anomaly \\(E\\):\n\n\\[x=a(\\cos E-e), \\qquad y=b\\sin E,\\]\n\nwhere the origin is the occupied focus. Defining the mean motion \\(n=\\sqrt{\\mu/a^3}\\) and the mean anomaly \\(M=n(t-\\tau)\\), with \\(\\tau\\) a time of closest approach, gives Kepler’s equation\n\n\\[M=E-e\\sin E.\\]\n\nFor \\(0\\leq e<1\\), the derivative of its right-hand side is \\(1-e\\cos E>0\\), so the equation has a unique continuous inverse once the orbital revolution is specified. That inverse is usually evaluated numerically.\n\nReal astronomical trajectories depart from fixed ellipses because of additional bodies, nonspherical mass distributions, atmospheric drag, radiation forces, or relativistic corrections. An osculating ellipse is the instantaneous Keplerian orbit matching a body's position and velocity when its instantaneous two-body energy is negative. Its parameters evolve under perturbations. The ellipse is therefore both an exact solution of an idealized model and a local reference geometry in more detailed celestial calculations."
    },
    {
      "heading": "Reflection, acoustics, and elliptical billiards",
      "text": "The ellipse has a focal reflection property: a ray traveling from one focus and reflecting specularly from the boundary passes through the other focus. This follows directly from the constant-distance definition. Let\n\n\\[S(P)=|P-F_1|+|P-F_2|.\\]\n\nOn the ellipse, \\(S(P)\\) is constant, so its gradient is normal to the curve:\n\n\\[\\nabla S(P)= \\frac{P-F_1}{|P-F_1|} + \\frac{P-F_2}{|P-F_2|}.\\]\n\nThis vector bisects the angle between the two outward focal directions. Consequently, the directions from the boundary point toward the foci make equal angles with the normal, giving the equal-angle law of reflection.\n\nThe result is exact in geometrical optics for an ideal reflecting boundary. In three dimensions, rotating an ellipse about its major axis creates a prolate ellipsoid, which has a corresponding point-to-point focusing property between its foci. An elliptical cylinder instead has focal lines: its cross-sectional focusing does not, by itself, imply focusing to one point in three-dimensional space. This distinction matters when moving from a plane diagram to an actual optical or acoustic device.\n\nElliptical reflectors can redirect radiation from a source near one focus toward the other. Analogous behavior occurs for sound when reflecting surfaces are sufficiently smooth and large relative to the wavelength. However, a geometrical construction does not predict all features of a wave field. Diffraction, interference, finite source size, absorption, and imperfect surfaces broaden or alter the concentration of energy. The equal focal path length \\(2a\\) is especially significant for coherent waves because ideal reflected paths accumulate the same propagation phase, apart from reflection effects.\n\nAn elliptical billiard is a mathematical dynamical system in which a point moves freely inside an ellipse and reflects elastically at the boundary. Its motion has a special invariant: generic trajectories have a confocal conic as a caustic, meaning that successive trajectory segments, or their extensions, are tangent to that same conic. The confocal family can be written\n\n\\[\\frac{x^2}{a^2-\\lambda} + \\frac{y^2}{b^2-\\lambda}=1.\\]\n\nFor \\(0<\\lambda<b^2\\), this is an inner ellipse; for \\(b^2<\\lambda<a^2\\), it is a hyperbola. The focal trajectory is a limiting special case.\n\nThis invariant makes the elliptical billiard a classical integrable system, unlike many billiard shapes that exhibit chaotic dynamics. Poncelet’s closure phenomenon adds another geometric feature: if a polygonal billiard path tangent to a suitable confocal caustic closes after a fixed number of reflections, corresponding paths starting elsewhere on the outer ellipse also close after that number. The reflection property thus extends beyond two-focus imaging to a structured family of repeated reflections."
    },
    {
      "heading": "Statistical ellipses and computational representation",
      "text": "Ellipses naturally describe directional variation in statistics. For a two-dimensional multivariate normal distribution with mean vector \\(\\boldsymbol{\\mu}\\) and positive-definite covariance matrix \\(\\Sigma\\), equal-density curves have the form\n\n\\[(\\mathbf{x}-\\boldsymbol{\\mu})^T \\Sigma^{-1} (\\mathbf{x}-\\boldsymbol{\\mu})=q.\\]\n\nThe expression on the left is the squared Mahalanobis distance. If \\(\\lambda_1,\\lambda_2\\) are the eigenvalues of \\(\\Sigma\\), the semiaxis lengths are \\(\\sqrt{q\\lambda_1}\\) and \\(\\sqrt{q\\lambda_2}\\), oriented along the corresponding eigenvectors. Unlike the coefficient matrix in an ellipse equation, covariance itself has its larger eigenvalue in the longer-axis direction.\n\nWhen the model parameters are known, the squared Mahalanobis distance has a chi-squared distribution with two degrees of freedom. A region containing probability \\(p\\) therefore uses\n\n\\[q=\\chi^2_{2,p}.\\]\n\nFor \\(p=0.95\\), \\(q\\) is approximately \\(5.991\\), giving semiaxes about \\(2.448\\) times the principal standard deviations. The contour \\(q=1\\), sometimes informally called a one-standard-deviation ellipse, contains only\n\n\\[1-e^{-1/2}\\approx0.393\\]\n\nof the probability in two dimensions, not the familiar one-dimensional proportion of about \\(0.683\\).\n\nA data ellipse is not automatically a confidence region for an unknown mean. The latter depends on sample size and on whether covariance is known or estimated; classical normal-theory constructions with estimated covariance involve a finite-sample adjustment. Likewise, a plotted covariance ellipse need not enclose a prescribed fraction of a nonnormal population. The eigenvectors still describe principal directions of second-moment variation, as in principal component analysis, but probability coverage requires additional distributional assumptions.\n\nFitting an ellipse to observed boundary points is another problem. A general conic has six coefficients determined only up to a common nonzero scale, so five suitably positioned exact points generically determine a conic. They do not guarantee that the result is an ellipse. With noisy observations, minimizing squared equation residuals is computationally convenient, but those residuals are algebraic rather than perpendicular geometric distances. Their scale and weighting depend on the equation’s coefficients. Geometric fitting instead minimizes distances to the curve, usually requiring iterative least-squares optimization and careful handling of eccentric or poorly sampled shapes.\n\nFor graphics, the trigonometric parametrization provides straightforward sampling, but uniform parameter steps produce unequal segment lengths and unequal approximation errors. Adaptive subdivision can use curvature or screen-space deviation. Polynomial Bézier curves approximate ellipse arcs but cannot represent a nondegenerate circular or elliptical arc exactly; rational quadratic Bézier curves can. Coordinate normalization and stable matrix decompositions also matter in computation, particularly when the semiaxes differ greatly or the ellipse is far from the coordinate origin."
    },
    {
      "heading": "Historical development and mathematical terminology",
      "text": "The ellipse belongs to the ancient Greek study of conic sections. Menaechmus, active in the fourth century BCE, is traditionally associated with early investigations of conics in connection with the duplication of the cube. Later Greek geometers developed systematic relations among sections, diameters, and tangents. Their methods did not use modern coordinate equations, but expressed relations through geometric constructions and proportions.\n\nApollonius of Perga, active around the late third and early second centuries BCE, made conic sections a unified subject in his Conics. The names ellipse, parabola, and hyperbola became established through this tradition. “Ellipse” derives from a Greek term associated with deficiency, reflecting an ancient construction involving the application of areas rather than simply the appearance of a flattened circle. Apollonius showed how the different conics could arise from sections of a single cone and developed extensive theories of their tangents and diameters. Archimedes studied the area of the ellipse and related results through methods of exhaustion.\n\nA decisive physical role emerged with Johannes Kepler’s Astronomia nova of 1609. Using Tycho Brahe’s observations, particularly those of Mars, Kepler concluded that planetary motion was better represented by an ellipse with the Sun at a focus than by combinations of uniform circular motions. His first two planetary laws appeared in that work; the period–semimajor-axis relation followed in 1619. Isaac Newton’s Principia of 1687 subsequently explained conic orbits and the area law through an inverse-square gravitational force, placing the ellipse within mathematical classical mechanics.\n\nThe seventeenth-century development of analytic geometry, associated especially with René Descartes and Pierre de Fermat, made polynomial equations central to the study of curves. Ellipses could then be treated alongside other quadratic loci by algebraic transformations. Later work on quadratic forms and matrix methods provided the modern principal-axis description, linking the geometry to eigenvalues and orthogonal diagonalization.\n\nThe perimeter problem led in a different direction. Eighteenth- and nineteenth-century investigations by mathematicians including Leonhard Euler, Adrien-Marie Legendre, Niels Henrik Abel, and Carl Gustav Jacob Jacobi developed elliptic integrals and their inverse functions. Here “elliptic” preserves a historical connection with ellipse arc length, although the resulting theory extends far beyond ellipses.\n\nAn ellipse should not be confused with an elliptic curve in modern algebraic geometry and number theory. A nonsingular projective conic has genus zero, whereas an elliptic curve has genus one together with a specified base point. The shared terminology reflects the history of elliptic integrals, not an identification of the two kinds of curve."
    }
  ],
  "links": [
    "conic-section",
    "circle",
    "geometry",
    "linear-algebra",
    "calculus",
    "triangle-inequality",
    "eccentricity",
    "parabola",
    "hyperbola",
    "cartesian-coordinate-system",
    "real-number",
    "matrix",
    "positive-definite-matrix",
    "spectral-theorem",
    "eigenvalues-and-eigenvectors",
    "affine-map",
    "projective-geometry",
    "gradient",
    "convex-set",
    "cauchy-schwarz-inequality",
    "determinant",
    "power-series",
    "newtons-method",
    "polar-coordinates",
    "keplers-laws",
    "dynamical-system",
    "chaos-theory",
    "statistics",
    "multivariate-normal-distribution",
    "covariance-matrix",
    "mahalanobis-distance",
    "chi-squared-distribution",
    "confidence-interval",
    "principal-component-analysis",
    "least-squares",
    "apollonius-of-perga",
    "archimedes",
    "classical-mechanics",
    "analytic-geometry",
    "pierre-de-fermat",
    "leonhard-euler",
    "elliptic-curve"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}