{
  "id": "elliptic-curve",
  "title": "Elliptic Curve",
  "lang": "en",
  "source_lang": "en",
  "summary": "An elliptic curve is a smooth projective curve of genus one with a distinguished point, linking geometry, number theory, complex analysis, and cryptography.",
  "category": "mathematics",
  "subcategory": "geometry",
  "type": "concept",
  "aliases": [
    "Elliptic algebraic curve"
  ],
  "generated_at": "2026-11-03",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/elliptic-curve/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/elliptic-curve/",
      "json": "https://aiwiki.page/zh/elliptic-curve.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "An elliptic curve is a smooth, projective algebraic curve of genus one together with a specified point that serves as the identity for a natural addition law. Although its equations can be elementary—for example, \\(y^2=x^3-x\\)—its points form an algebraic structure whose properties depend strongly on the number system in which the coordinates are taken. Over the complex numbers, an elliptic curve is a torus; over the rational numbers, its points form a finitely generated group; over a finite field, they form a finite group that can support cryptographic protocols.\n\nElliptic curves occupy a central position between algebraic geometry, number theory, and complex analysis. They helped connect equations in integers with modular forms, provided the setting for the proof of Fermat’s Last Theorem, and remain the subject of major open questions about rational solutions. Their name comes not from a resemblance to an ellipse but from their historical connection with integrals arising in the calculation of an ellipse’s arc length."
    },
    {
      "heading": "Definition, equations, and geometric invariants",
      "text": "The modern definition begins with a field \\(K\\). An elliptic curve over \\(K\\) is a smooth, projective, geometrically integral curve \\(E\\) of genus one, equipped with a distinguished \\(K\\)-rational point \\(O\\). “Geometrically integral” means that the curve remains irreducible and reduced after extending the field to an algebraic closure. The genus measures the intrinsic complexity of the curve: over the complex numbers, genus one corresponds to a compact surface with one handle. A genus-one curve without a specified rational point need not be an elliptic curve over its field of definition, even though it becomes one after a suitable field extension and choice of point.\n\nEvery elliptic curve admits a generalized Weierstrass equation\n\n\\[y^2+a_1xy+a_3y=x^3+a_2x^2+a_4x+a_6,\\]\n\nwith coefficients in \\(K\\). This equation describes an affine portion of the curve. Its completion in projective space adds the point \\(O=[0:1:0]\\), written using homogeneous coordinates. The added point is indispensable: it makes the curve complete and provides the identity for addition. The projective definition also prevents phenomena caused merely by omitting points at infinity from being mistaken for intrinsic geometric properties.\n\nWhen the characteristic of \\(K\\) is neither \\(2\\) nor \\(3\\), a change of coordinates reduces the equation to the short Weierstrass form\n\n\\[y^2=x^3+Ax+B.\\]\n\nThe condition for smoothness is\n\n\\[\\Delta=-16(4A^3+27B^2)\\ne0,\\]\n\nwhere \\(\\Delta\\) is the discriminant of this model. If the discriminant vanishes, the cubic has a singularity, such as a node or cusp, and is not an elliptic curve. For example, \\(y^2=x^3-x\\) is nonsingular, whereas \\(y^2=x^3\\) has a cusp at the origin. In characteristics \\(2\\) and \\(3\\), the generalized equation must be retained because the divisions used to obtain the short form are unavailable.\n\nA fundamental invariant is the j-invariant, which for the short equation is\n\n\\[j=1728\\frac{4A^3}{4A^3+27B^2}.\\]\n\nOver an algebraically closed field, two elliptic curves are isomorphic precisely when their \\(j\\)-invariants agree. Over the original field, equal \\(j\\)-invariants need not imply an isomorphism defined over that field. Distinct twists can become isomorphic only after extending the field. This distinction separates geometric classification from arithmetic classification: the same complex curve can have different forms over the rational numbers, with different rational points and different behavior at primes."
    },
    {
      "heading": "The addition law and the group of points",
      "text": "The defining point \\(O\\) turns an elliptic curve into an abelian group. For a short Weierstrass equation over the real numbers, addition has a geometric description called the chord-and-tangent construction. A line through two points \\(P\\) and \\(Q\\) meets the cubic in a third point \\(R\\), counting intersection multiplicity. Reflecting \\(R\\) across the \\(x\\)-axis gives \\(P+Q\\). When \\(P=Q\\), the tangent line replaces the chord. A vertical line meets the curve at two opposite points and at \\(O\\), so \\(P+(-P)=O\\).\n\nFor distinct affine points \\(P=(x_1,y_1)\\) and \\(Q=(x_2,y_2)\\), with \\(x_1\\ne x_2\\), the formulas are\n\n\\[m=\\frac{y_2-y_1}{x_2-x_1}, \\qquad x_3=m^2-x_1-x_2, \\qquad y_3=m(x_1-x_3)-y_1.\\]\n\nThen \\(P+Q=(x_3,y_3)\\). For doubling, provided \\(y_1\\ne0\\),\n\n\\[m=\\frac{3x_1^2+A}{2y_1}.\\]\n\nThe inverse is \\(-(x,y)=(x,-y)\\). In the generalized Weierstrass equation, the inverse becomes\n\n\\[-(x,y)=(x,-y-a_1x-a_3).\\]\n\nCases in which a denominator vanishes are handled by the identity and inverse rules rather than by attempting an undefined division.\n\nThese formulas are rational expressions in the coordinates, so they work over any field where the indicated denominators are nonzero. If \\(P\\) and \\(Q\\) have coordinates in \\(K\\), their sum also has coordinates in \\(K\\). Thus the set \\(E(K)\\) of \\(K\\)-rational points is a group. Associativity is the least obvious group axiom in the geometric construction. It can be verified algebraically, but its conceptual explanation comes from the theory of divisors: a point \\(P\\) corresponds to the degree-zero divisor class represented by \\((P)-(O)\\), and addition of divisor classes is associative. The resulting group is called the degree-zero Picard group.\n\nThe law creates new solutions from known ones. On\n\n\\[E:\\quad y^2=x^3-2,\\]\n\nthe rational point \\(P=(3,5)\\) doubles to\n\n\\[2P=\\left(\\frac{129}{100},-\\frac{383}{1000}\\right).\\]\n\nRepeated addition produces multiples \\(nP\\), but the coordinates can grow rapidly even when the starting point is small. A point is called torsion if \\(nP=O\\) for some positive integer \\(n\\). On a short Weierstrass model, nonidentity points of order two have \\(y=0\\), because they equal their own inverses. These elementary observations lead directly to arithmetic questions about finite-order points, generators, and the existence of infinitely many rational solutions."
    },
    {
      "heading": "Complex tori, elliptic functions, and elliptic integrals",
      "text": "Over the complex numbers, every elliptic curve has an analytic description as a quotient\n\n\\[E(\\mathbb C)\\cong\\mathbb C/\\Lambda,\\]\n\nwhere\n\n\\[\\Lambda=\\mathbb Z\\omega_1+\\mathbb Z\\omega_2\\]\n\nis a lattice and \\(\\omega_1/\\omega_2\\) is not real. Identifying opposite sides of a fundamental parallelogram produces a torus. This is an isomorphism of complex Lie groups, not merely a topological resemblance: addition of complex numbers modulo the lattice agrees with the algebraic addition law on the curve. The quotient is also a compact Riemann surface of genus one.\n\nThe bridge from the torus to a cubic equation is the Weierstrass \\(\\wp\\)-function,\n\n\\[\\wp(z;\\Lambda)=\\frac{1}{z^2} +\\sum_{\\substack{\\omega\\in\\Lambda\\\\\\omega\\ne0}} \\left(\\frac{1}{(z-\\omega)^2}-\\frac{1}{\\omega^2}\\right).\\]\n\nIt is an elliptic function, meaning a meromorphic function with two independent periods. Together with its derivative, it satisfies\n\n\\[\\wp'(z)^2=4\\wp(z)^3-g_2\\wp(z)-g_3.\\]\n\nThe map \\(z\\mapsto(\\wp(z),\\wp'(z))\\), extended by sending the lattice class of \\(0\\) to \\(O\\), identifies the complex torus with this cubic. The constants \\(g_2\\) and \\(g_3\\) are determined by convergent sums over the nonzero lattice points.\n\nThis description grew out of elliptic integrals, which include integrals of algebraic functions involving square roots of cubic or quartic polynomials. Such expressions arise in arc-length problems for ellipses and in mechanical systems. Eighteenth-century work by Leonhard Euler and Adrien-Marie Legendre developed their addition formulas and classification. In the 1820s, Niels Henrik Abel and Carl Gustav Jacob Jacobi transformed the subject by studying inverse functions. The inverse of a suitable elliptic integral is doubly periodic, explaining why a problem about integration leads to a torus and then to an algebraic curve.\n\nOn a short Weierstrass curve, the differential\n\n\\[\\omega=\\frac{dx}{2y}\\]\n\nis holomorphic everywhere on the complete curve, including at infinity. Integrating it around two independent closed cycles gives periods that generate the lattice, up to the normalization of the analytic identification. Thus algebraic coordinates, complex integration, and the topology of the torus describe the same object in different languages.\n\nAfter rescaling, a lattice can be written as \\(\\mathbb Z+\\mathbb Z\\tau\\), where \\(\\operatorname{Im}\\tau>0\\). Different choices of lattice basis replace \\(\\tau\\) by\n\n\\[\\frac{a\\tau+b}{c\\tau+d}, \\qquad ad-bc=1,\\]\n\nwith integer coefficients. The \\(j\\)-invariant identifies these equivalent descriptions. By contrast, the real locus \\(E(\\mathbb R)\\) has either one or two connected components. For \\(y^2=x^3+Ax+B\\), it has two components when the cubic has three distinct real roots and one when it has only one real root."
    },
    {
      "heading": "Rational points, torsion, and rank",
      "text": "For an elliptic curve over the rational numbers, the central structural result is the Mordell–Weil theorem:\n\n\\[E(\\mathbb Q)\\cong E(\\mathbb Q)_{\\mathrm{tors}}\\oplus\\mathbb Z^r.\\]\n\nThe torsion subgroup is finite, and \\(r\\), a nonnegative integer, is the rank. Louis Mordell proved finite generation over \\(\\mathbb Q\\) in 1922; André Weil extended the theorem to abelian varieties over number fields in his 1928 work. The theorem says that every rational point can be expressed using finitely many generators and integer coefficients. It does not supply a uniform procedure that immediately finds those generators.\n\nThe rank measures the number of independent infinite-order points. Rank zero means that only finitely many rational points exist, all torsion. Positive rank implies infinitely many rational points, obtained by taking integer combinations of independent generators. Two curves with very similar-looking equations can have different ranks. Moreover, finding many rational points is not the same as finding many independent points: a large list might consist entirely of multiples of a single generator.\n\nBarry Mazur’s torsion theorem, established in the 1970s, gives a complete list of possible torsion groups over \\(\\mathbb Q\\). They are\n\n\\[\\mathbb Z/n\\mathbb Z,\\qquad 1\\le n\\le10\\ \\text{or}\\ n=12,\\]\n\nand\n\n\\[\\mathbb Z/2\\mathbb Z\\oplus\\mathbb Z/2m\\mathbb Z, \\qquad 1\\le m\\le4.\\]\n\nThe list concerns curves over \\(\\mathbb Q\\); extending the field changes the possibilities. A complementary result, the Lutz–Nagell theorem, states that for a nonsingular short Weierstrass equation with integer coefficients, every rational torsion point has integer coordinates. Its \\(y\\)-coordinate is either zero or has square dividing \\(4A^3+27B^2\\). This turns torsion detection for such a model into a finite arithmetic problem.\n\nThe proof and computation of finite generation rely on heights, which quantify the arithmetic size of a rational point. A basic logarithmic height measures the numerator and denominator of its \\(x\\)-coordinate. The canonical height corrects this size function so that\n\n\\[\\widehat h(nP)=n^2\\widehat h(P).\\]\n\nIt vanishes exactly on torsion points and yields a positive-definite pairing on the free part of \\(E(\\mathbb Q)\\). The determinant of the pairing matrix for a basis is the regulator, an arithmetic measure of the spacing of independent rational points.\n\nInteger-coordinate solutions form a smaller set than rational solutions. Siegel’s theorem implies that a fixed elliptic curve given by an integral Weierstrass equation has only finitely many integral points, even when its rational points are infinite. Determining all of them requires more than simply generating many multiples of known points."
    },
    {
      "heading": "Reduction modulo primes and curves over finite fields",
      "text": "An integral equation for an elliptic curve can be reduced modulo a prime number \\(p\\). If the reduced equation remains nonsingular, the curve has good reduction at \\(p\\). In particular, a prime not dividing the discriminant of an integral model is a prime of good reduction. A model with unnecessarily large discriminant can obscure the intrinsic behavior, so arithmetic investigations use locally minimal models. At a prime of bad reduction, the special fiber is singular and the reduction is classified as multiplicative or additive.\n\nOver a finite field \\(\\mathbb F_q\\), the group \\(E(\\mathbb F_q)\\) is finite. Hasse’s theorem gives the sharp general estimate\n\n\\[\\left|\\#E(\\mathbb F_q)-(q+1)\\right|\\le2\\sqrt q.\\]\n\nThe integer\n\n\\[a_q=q+1-\\#E(\\mathbb F_q)\\]\n\nis the trace of Frobenius. The map raising coordinates to their \\(q\\)-th powers induces the Frobenius endomorphism, whose characteristic polynomial is\n\n\\[T^2-a_qT+q.\\]\n\nIf its roots are \\(\\alpha\\) and \\(\\beta\\), then \\(|\\alpha|=|\\beta|=\\sqrt q\\) under complex embeddings, and\n\n\\[\\#E(\\mathbb F_{q^n})=q^n+1-\\alpha^n-\\beta^n.\\]\n\nConsequently, one point count determines the counts over every finite extension of the base field.\n\nFor example, consider\n\n\\[E:\\quad y^2=x^3-x\\]\n\nover \\(\\mathbb F_5\\). The values of \\(x^3-x\\) for \\(x=0,1,2,3,4\\) are \\(0,0,1,4,0\\). These give seven affine points; adding \\(O\\) gives eight points in total. Thus \\(a_5=5+1-8=-2\\), in agreement with Hasse’s bound. Even this small example illustrates that counting points is a question about quadratic residues rather than the shape of a real graph.\n\nReduction also constrains rational arithmetic. At a prime of good reduction, the prime-to-\\(p\\) part of the rational torsion subgroup injects into \\(E(\\mathbb F_p)\\). Point counts at several good primes can therefore place strong restrictions on possible rational torsion. The local behavior can also be studied over the field of p-adic numbers, where reduction is refined by information about congruences modulo powers of \\(p\\).\n\nElliptic curves in characteristic \\(p\\) divide into ordinary and supersingular types. Over an algebraic closure, the group of points killed by multiplication by \\(p\\) has \\(p\\) elements in the ordinary case and only the identity in the supersingular case. This does not mean that the multiplication-by-\\(p\\) map has degree one in the supersingular case: its degree is always \\(p^2\\), and inseparability accounts for the difference between geometric points and algebraic degree."
    },
    {
      "heading": "Isogenies, endomorphisms, and complex multiplication",
      "text": "An isogeny is a nonconstant morphism between elliptic curves that preserves their identity points. It is a surjective group homomorphism with finite kernel. Isogenous curves need not be isomorphic: an isogeny can identify several points rather than provide a one-to-one correspondence. The basic example is multiplication by an integer,\n\n\\[[n]:E\\longrightarrow E,\\qquad P\\longmapsto nP,\\]\n\nwhose degree is \\(n^2\\) for \\(n\\ne0\\). When the characteristic does not divide \\(n\\), its geometric kernel is isomorphic to\n\n\\[(\\mathbb Z/n\\mathbb Z)^2.\\]\n\nFor a separable isogeny, the kernel determines the quotient curve up to isomorphism. Conversely, a finite subgroup that is appropriately defined over the base field gives rise to a quotient isogeny. An isogeny \\(\\phi:E\\to E'\\) of degree \\(d\\) has a dual isogeny \\(\\widehat\\phi:E'\\to E\\) satisfying\n\n\\[\\widehat\\phi\\circ\\phi=[d], \\qquad \\phi\\circ\\widehat\\phi=[d].\\]\n\nThese relations make isogenies useful for comparing arithmetic invariants. Elliptic curves over \\(\\mathbb Q\\) that are isogenous over \\(\\mathbb Q\\), for example, have the same rank and the same \\(L\\)-function, although their torsion subgroups and integral models may differ.\n\nAn endomorphism is a morphism from a curve to itself preserving the identity. Over \\(\\mathbb C\\), most elliptic curves have only the integer multiplication endomorphisms, so their endomorphism ring is \\(\\mathbb Z\\). Exceptional curves have complex multiplication: their endomorphism ring is an order in an imaginary quadratic field. For instance, the curve\n\n\\[y^2=x^3-x\\]\n\nhas an automorphism\n\n\\[(x,y)\\longmapsto(-x,iy),\\]\n\nwhich corresponds to multiplication by \\(i\\) in its complex-torus description. Its \\(j\\)-invariant is \\(1728\\). Curves with \\(j=0\\) likewise have additional automorphisms related to cube roots of unity.\n\nComplex multiplication creates a particularly explicit connection between elliptic curves and class field theory. Values of the \\(j\\)-function at imaginary quadratic arguments are algebraic integers, and they generate important abelian extensions of imaginary quadratic fields. These algebraic values also support algorithms for constructing elliptic curves over finite fields with prescribed point counts.\n\nIn characteristic \\(p\\), supersingular curves have a different endomorphism theory: their endomorphism algebras over an algebraic closure are quaternion algebras rather than imaginary quadratic fields. Another way to encode endomorphisms is through the Tate module\n\n\\[T_\\ell(E)=\\varprojlim_n E[\\ell^n],\\]\n\nfor a prime \\(\\ell\\) different from the characteristic. It is a rank-two module over the \\(\\ell\\)-adic integers. Endomorphisms and field automorphisms act on it, turning geometric operations into linear algebra and linking isogeny theory with Galois theory."
    },
    {
      "heading": "L-functions, modularity, and Fermat’s Last Theorem",
      "text": "An elliptic curve over \\(\\mathbb Q\\) produces a sequence of arithmetic data by counting points modulo primes. At a good prime \\(p\\), put\n\n\\[a_p=p+1-\\#E(\\mathbb F_p).\\]\n\nIts L-function packages these counts into an Euler product,\n\n\\[L(E,s)=\\prod_p L_p(E,s).\\]\n\nAt a good prime, the local factor is\n\n\\[L_p(E,s)=\\left(1-a_pp^{-s}+p^{1-2s}\\right)^{-1}.\\]\n\nAt multiplicative reduction it has the form \\((1-a_pp^{-s})^{-1}\\), with \\(a_p=1\\) for split and \\(a_p=-1\\) for nonsplit multiplicative reduction; at additive reduction it is \\(1\\). Hasse’s bound implies absolute convergence of this product for \\(\\operatorname{Re}(s)>3/2\\).\n\nA modular form is a highly symmetric analytic function on the upper half-plane. The modularity theorem states that every elliptic curve over \\(\\mathbb Q\\) corresponds to a weight-two newform of level equal to the curve’s conductor. If the newform has expansion\n\n\\[f(z)=\\sum_{n\\ge1}a_nq^n, \\qquad q=e^{2\\pi iz},\\]\n\nits coefficients at good primes agree with the point-counting numbers \\(a_p\\). The conductor is an integer supported on the primes of bad reduction, with exponents reflecting the local arithmetic.\n\nThis correspondence supplies analytic properties that are not apparent from the Euler product. The completed function\n\n\\[\\Lambda(E,s)=N^{s/2}(2\\pi)^{-s}\\Gamma(s)L(E,s)\\]\n\nextends to an entire function and satisfies\n\n\\[\\Lambda(E,s)=w_E\\Lambda(E,2-s),\\]\n\nwhere \\(N\\) is the conductor and \\(w_E\\in\\{1,-1\\}\\) is the root number. The point \\(s=1\\) is the center of the functional equation. Its sign constrains the parity of the order of vanishing there, although relating that order to rational points requires additional arithmetic results.\n\nThe theorem developed from ideas associated with Yutaka Taniyama, Goro Shimura, and André Weil. Andrew Wiles proved the semistable case in work completed with Richard Taylor and published in 1995. Christophe Breuil, Brian Conrad, Fred Diamond, and Taylor established the full theorem, published in 2001. These proofs use deformations of Galois representations and comparisons between deformation rings and Hecke algebras, rather than explicit formulas transforming every cubic into a modular form.\n\nThe semistable case was sufficient for Fermat’s Last Theorem. Following Gerhard Frey’s proposed connection and Jean-Pierre Serre’s formulation, Ken Ribet proved the crucial level-lowering implication: a hypothetical Fermat solution with prime exponent greater than two would produce a semistable elliptic curve incompatible with modularity. Wiles’s theorem ruled out that possibility. The argument illustrates the reach of elliptic curves: a question about integer powers was resolved by studying the compatibility of geometry, prime-by-prime arithmetic, and analytic symmetry."
    },
    {
      "heading": "The Birch–Swinnerton-Dyer conjecture and local–global questions",
      "text": "The Birch–Swinnerton-Dyer conjecture proposes that the rational points of an elliptic curve are governed by the behavior of its \\(L\\)-function at \\(s=1\\). Its rank statement is\n\n\\[\\operatorname{ord}_{s=1}L(E,s)=\\operatorname{rank}E(\\mathbb Q).\\]\n\nThe left side is the analytic rank, the order of the zero of the analytically continued function. The right side is the algebraic rank supplied by the Mordell–Weil theorem. Bryan Birch and Peter Swinnerton-Dyer developed the conjecture in the 1960s, using early computer calculations to investigate relationships between rational points and products of point counts modulo primes.\n\nA refined form predicts the leading coefficient:\n\n\\[\\lim_{s\\to1}\\frac{L(E,s)}{(s-1)^r} = \\frac{\\Omega_E\\,\\operatorname{Reg}_E\\, \\#\\Sha(E/\\mathbb Q)\\,\\prod_p c_p} {\\#E(\\mathbb Q)_{\\mathrm{tors}}^2}.\\]\n\nHere \\(r\\) is the rank, \\(\\Omega_E\\) is the real period in the standard normalization, \\(\\operatorname{Reg}_E\\) is the regulator, and \\(c_p\\) are local Tamagawa numbers. The formula also predicts finiteness of the Tate–Shafarevich group \\(\\Sha(E/\\mathbb Q)\\). Each factor captures a different aspect of the curve: real integration, the geometry of the rational-point lattice, finite torsion, or behavior at bad primes.\n\nThe Tate–Shafarevich group measures a failure of the local–global principle. A torsor under \\(E\\) is a genus-one curve on which \\(E\\) acts simply transitively after passing to an algebraic closure, but which may lack a rational point. Classes in \\(\\Sha\\) are represented by torsors that have points over the real numbers and over every \\(p\\)-adic field, yet may have no point over \\(\\mathbb Q\\). Thus solving an equation modulo arbitrarily high powers of every prime does not necessarily produce a rational solution.\n\nThe Selmer group is a finite, computable approximation that links local tests with global rational points. For an integer \\(n\\ge2\\), there is an exact sequence\n\n\\[0\\longrightarrow E(\\mathbb Q)/nE(\\mathbb Q) \\longrightarrow\\operatorname{Sel}_n(E/\\mathbb Q) \\longrightarrow\\Sha(E/\\mathbb Q)[n] \\longrightarrow0.\\]\n\nA descent calculation can therefore bound the rank from above. Any gap between this bound and the subgroup generated by known rational points can reflect either missing generators or nontrivial elements of \\(\\Sha\\).\n\nMajor theorems settle important cases without establishing the full conjecture. The Gross–Zagier formula of 1986 relates derivatives of \\(L\\)-functions to heights of Heegner points. Combined with Kolyvagin’s work, it implies that elliptic curves over \\(\\mathbb Q\\) of analytic rank zero or one have the predicted algebraic rank and finite Tate–Shafarevich group. The general rank statement, finiteness of \\(\\Sha\\), and the full leading-coefficient formula remain open in general; the conjecture is one of the Millennium Prize Problems."
    },
    {
      "heading": "Computational arithmetic and factorization",
      "text": "Elliptic-curve computation encompasses several distinct problems: finding rational points, proving a rank, counting points over finite fields, computing isogenies, and evaluating \\(L\\)-functions. Their difficulty varies considerably. Addition of two given points is straightforward, but proving that a proposed set generates all rational points can require deep arithmetic. A computer search through coordinates of bounded size gives a lower bound on the rank only after independence is established; failure to find more points is not an upper bound.\n\nDescent provides upper bounds by replacing the search for rational points with a finite collection of covering curves subject to local conditions. Two-descent is especially common, while descents associated with isogenies can be efficient when a suitable isogeny exists. Height pairings test independence of candidate generators. A further procedure called saturation checks whether the subgroup they generate omits a point whose multiple already lies in that subgroup. A reliable computation distinguishes a proven rank and full generating set from bounds or results conditional on unproved assumptions.\n\nPoint counting over finite fields has its own algorithmic theory. René Schoof’s 1985 algorithm computes \\(\\#E(\\mathbb F_q)\\) in time polynomial in \\(\\log q\\), for suitable standard representations, by determining the Frobenius trace modulo small primes and reconstructing it with the Chinese remainder theorem. The Schoof–Elkies–Atkin method improves practical performance by exploiting how Frobenius acts on torsion and by using modular polynomials. These methods make it possible to count points over fields far too large for enumerating every possible coordinate.\n\nElliptic curves also provide an integer factorization method introduced by Hendrik Lenstra in 1987. The method attempts elliptic-curve arithmetic modulo a composite integer \\(N\\). A denominator that cannot be inverted modulo \\(N\\) may reveal a factor through a greatest-common-divisor calculation. Its main stage multiplies a point by an integer divisible by many small prime powers. If the reduced point modulo an unknown prime factor \\(p\\) has sufficiently smooth order, this multiplication sends it to the identity modulo \\(p\\), creating an opportunity to expose \\(p\\).\n\nThe advantage over methods based on the multiplicative group modulo \\(p\\) is that changing the curve changes the relevant group order. For a fixed unknown factor \\(p\\), different elliptic curves have different orders near \\(p+1\\), as constrained by Hasse’s bound. Repeatedly choosing curves therefore gives fresh chances to encounter a smooth order. The elliptic-curve method is particularly effective at finding relatively small prime factors inside much larger composite integers.\n\nElliptic curves also underpin primality-proving algorithms. In elliptic-curve primality proving, carefully chosen curves and points supply certificates whose group-order properties imply that a candidate integer is prime. Verification relies on explicit arithmetic conditions and can be substantially simpler than the search that produced the certificate."
    },
    {
      "heading": "Cryptographic uses and security boundaries",
      "text": "Elliptic-curve cryptography uses finite groups of curve points to construct public-key systems. Victor Miller and Neal Koblitz independently proposed the approach in 1985. The basic operation is scalar multiplication: given a point \\(P\\) and an integer \\(k\\), compute\n\n\\[Q=[k]P.\\]\n\nRepeated doubling and addition require a number of group operations proportional to the bit length of \\(k\\). The reverse problem—recovering \\(k\\) from \\(P\\) and \\(Q\\)—is an instance of the discrete logarithm problem and is believed to be difficult in suitably chosen groups.\n\nIn elliptic-curve Diffie–Hellman key agreement, two parties publish points \\([a]P\\) and \\([b]P\\). Each can calculate \\([ab]P\\) using a private scalar and the other party’s public point. An observer is given the public points but not the scalars. Signature schemes such as ECDSA and EdDSA use related group operations to make signatures publicly verifiable. Actual protocols include encodings, hash functions, key-derivation procedures, and validation rules; the abstract curve group alone does not specify a complete secure system.\n\nFor a prime-order subgroup of size \\(n\\), generic classical attacks such as Pollard’s rho method take roughly \\(\\sqrt n\\) group operations. This explains why a group of order near \\(2^{256}\\) is commonly associated with about 128 bits of classical security. The estimate is conditional on the absence of faster attacks exploiting special curve properties. Curves with anomalous group order or unsuitable pairing behavior can be insecure even when their parameters appear large.\n\nSeveral equation forms are useful in implementations. Montgomery curves permit efficient arithmetic using only an \\(x\\)-coordinate representation, while twisted Edwards curves can provide efficient addition formulas with favorable completeness properties. Curve25519, introduced by Daniel Bernstein in 2006, uses the prime field of order \\(2^{255}-19\\); X25519 specifies a widely deployed key-agreement function based on it. Ed25519 is a signature system using a related twisted Edwards curve. Secp256k1, used in Bitcoin, has the short Weierstrass equation \\(y^2=x^3+7\\) over its specified prime field.\n\nImplementation security is distinct from mathematical hardness. Secret-dependent timing, power consumption, or memory access can leak scalar information. Invalid input points and mishandled subgroup structure can undermine protocols, while nonce reuse in ECDSA can reveal a signing key. Constant-time arithmetic and precise input-handling rules address threats that a large group order alone does not prevent.\n\nElliptic-curve cryptography is not resistant to a sufficiently capable quantum computer. Shor’s algorithm solves discrete logarithms in polynomial time, including those in elliptic-curve groups. Isogeny-based cryptography explores different computational problems involving elliptic curves, but security does not follow simply from replacing scalar multiplication with isogenies: the 2022 break of the SIKE construction demonstrated that the exact mathematical problem and the additional information exposed by a protocol are decisive."
    }
  ],
  "links": [
    "algebraic-geometry",
    "number-theory",
    "complex-analysis",
    "field-mathematics",
    "projective-space",
    "homogeneous-coordinates",
    "discriminant",
    "j-invariant",
    "isomorphism",
    "abelian-group",
    "complex-number",
    "riemann-surface",
    "rational-number",
    "prime-number",
    "finite-field",
    "homomorphism",
    "galois-theory",
    "fermats-last-theorem",
    "millennium-prize-problems"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}