{
  "id": "gauss-bonnet-theorem",
  "title": "Gauss–Bonnet Theorem",
  "lang": "en",
  "source_lang": "en",
  "summary": "The Gauss–Bonnet theorem equates integrated curvature with the Euler characteristic, linking local geometry to global topology.",
  "category": "mathematics",
  "subcategory": "geometry",
  "type": "concept",
  "aliases": [
    "Gauss–Bonnet formula",
    "Gauss–Bonnet–Chern theorem"
  ],
  "generated_at": "2026-11-05",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/gauss-bonnet-theorem/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/gauss-bonnet-theorem/",
      "json": "https://aiwiki.page/zh/gauss-bonnet-theorem.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "The Gauss–Bonnet theorem is a fundamental result connecting differential geometry with topology. For a compact surface without boundary, it states that the integral of Gaussian curvature equals \\(2\\pi\\) times the surface’s Euler characteristic. Curvature describes the geometry near individual points; the Euler characteristic describes the global topological structure. The theorem therefore identifies a quantity computed from local measurements that remains unchanged when the surface’s metric is altered.\n\nThe classical theorem concerns two-dimensional surfaces, including surfaces with boundaries and corners. Its higher-dimensional extension, usually called the Gauss–Bonnet–Chern theorem, expresses the Euler characteristic of an even-dimensional manifold as an integral of a polynomial in its curvature. Related formulations involve angle defects on polyhedral surfaces, indices of vector fields, and the heat equation on differential forms.\n\nThe theorem bears the names of Carl Friedrich Gauss and Pierre Ossian Bonnet. Its development from formulas for geodesic triangles to intrinsic curvature identities helped establish the modern relationship between geometry, topology, and analysis."
    },
    {
      "heading": "Intrinsic curvature and the closed-surface theorem",
      "text": "Let \\(M\\) be a smooth, compact, two-dimensional manifold without boundary, equipped with a Riemannian metric \\(g\\). The metric specifies lengths and angles in each tangent space and determines an area element \\(dA_g\\). If \\(K_g\\) is the Gaussian curvature, the theorem states\n\n\\[\\int_M K_g\\,dA_g=2\\pi\\chi(M),\\]\n\nwhere \\(\\chi(M)\\) is the Euler characteristic. For a finite triangulation, this integer is\n\n\\[\\chi(M)=V-E+F,\\]\n\nwith \\(V\\), \\(E\\), and \\(F\\) denoting the numbers of vertices, edges, and faces. Different triangulations yield the same integer. The theorem thus equates a metric-dependent integral with a topological invariant.\n\nFor a surface embedded in three-dimensional Euclidean space, Gaussian curvature is the product of the two principal curvatures:\n\n\\[K=\\kappa_1\\kappa_2.\\]\n\nA sphere has positive Gaussian curvature, a plane has zero curvature, and a saddle-shaped patch typically has negative curvature. Nevertheless, the theorem does not require an embedding. Gauss’s Theorema Egregium establishes that \\(K\\) can be determined entirely from the metric tensor and its derivatives. Accordingly, Gauss–Bonnet applies to abstract Riemannian surfaces, including ones for which a particular ambient-space picture is unavailable or misleading.\n\nThe metric affects both factors in the integrand. If lengths are multiplied by a constant \\(c>0\\), Gaussian curvature is multiplied by \\(c^{-2}\\), while area is multiplied by \\(c^2\\). Their product remains unchanged. More strongly, the total curvature is unchanged under arbitrary smooth changes of the metric, not just uniform rescaling. A sphere can be deformed into a highly irregular smooth shape, with regions of negative curvature, while its total Gaussian curvature remains \\(4\\pi\\).\n\nFor a connected, closed, orientable surface of genus \\(g\\),\n\n\\[\\chi(M)=2-2g, \\qquad \\int_M K\\,dA=4\\pi(1-g).\\]\n\nHere genus counts handles: a sphere has genus zero, a torus genus one, and a double torus genus two. For a connected, closed, nonorientable surface obtained as the connected sum of \\(k\\) projective planes, \\(\\chi(M)=2-k\\). The theorem still holds, interpreting \\(K\\,dA\\) as a density rather than an oriented differential form. Alternatively, it follows by passing to the orientable double cover, on which both total curvature and Euler characteristic double.\n\nCompactness and the absence of boundary are substantive hypotheses. Compactness ensures that the integral is finite for a smooth metric. A boundary introduces a further contribution from the way its tangent direction turns within the surface."
    },
    {
      "heading": "Boundaries, corners, and noncompact surfaces",
      "text": "For a compact oriented surface \\(M\\) with smooth boundary, Gauss–Bonnet becomes\n\n\\[\\int_M K\\,dA+\\int_{\\partial M}k_g\\,ds =2\\pi\\chi(M).\\]\n\nThe quantity \\(k_g\\) is the signed geodesic curvature of the boundary. With the boundary traversed so that the surface lies to the left, let \\(T\\) be its unit tangent and let \\(J\\) rotate tangent vectors counterclockwise through \\(\\pi/2\\). A consistent convention is\n\n\\[k_g=\\langle\\nabla_TT,JT\\rangle,\\]\n\nwhere \\(\\nabla\\) is the Levi-Civita connection. This measures turning within the surface, not the full curvature of the boundary as a curve in an ambient Euclidean space.\n\nIf the boundary consists of finitely many smooth arcs meeting at corners, the formula includes signed exterior turning angles \\(\\theta_j\\):\n\n\\[\\int_M K\\,dA+ \\int_{\\partial M}k_g\\,ds+ \\sum_j\\theta_j =2\\pi\\chi(M).\\]\n\nAt a corner with interior angle \\(\\alpha_j\\), the exterior angle is \\(\\theta_j=\\pi-\\alpha_j\\). Reflex corners therefore contribute negatively. A simple polygonal disk in the Euclidean plane has \\(K=0\\) and straight edges with \\(k_g=0\\), so its exterior angles sum to \\(2\\pi\\).\n\nFor a disk-shaped geodesic triangle, the edge integrals vanish. If its interior angles are \\(\\alpha,\\beta,\\gamma\\), the theorem gives\n\n\\[\\alpha+\\beta+\\gamma-\\pi=\\int_TK\\,dA.\\]\n\nOn a sphere of radius \\(R\\), the right side is \\(\\operatorname{Area}(T)/R^2\\). On a surface of constant curvature \\(-1\\), it is \\(-\\operatorname{Area}(T)\\). Spherical triangles consequently have angle sum greater than \\(\\pi\\), whereas hyperbolic triangles have angle sum less than \\(\\pi\\). These statements concern geodesic edges; arbitrary curved sides add geodesic-curvature terms.\n\nA planar round disk gives another normalization check: \\(K=0\\), its boundary has \\(k_g=1/R\\), and its perimeter is \\(2\\pi R\\), yielding \\(2\\pi\\). For a planar annulus, the outer boundary contributes \\(2\\pi\\) and the oppositely oriented inner boundary contributes \\(-2\\pi\\), consistent with \\(\\chi=0\\).\n\nOn a noncompact surface, curvature at infinity cannot simply be ignored. For an exhaustion by compact subsurfaces \\(M_R\\), the valid starting point is\n\n\\[\\int_{M_R}K\\,dA+\\int_{\\partial M_R}k_g\\,ds =2\\pi\\chi(M_R),\\]\n\nwith corner terms when necessary. Passing to a limit requires control of all terms. The Euclidean plane has zero total curvature but Euler characteristic one: expanding circular boundaries retain a geodesic-curvature integral of \\(2\\pi\\). By contrast, a complete finite-area hyperbolic surface of genus \\(g\\) with \\(n\\) cusp ends has\n\n\\[\\operatorname{Area}(M)=2\\pi(2g-2+n).\\]\n\nFor standard truncations of its cusps, the boundary contributions tend to zero."
    },
    {
      "heading": "Local curvature identities and global proof mechanisms",
      "text": "One classical proof first establishes a local curvature formula and then combines it across a triangulation. The local calculation uses an oriented orthonormal frame \\(e_1,e_2\\) on a surface patch. Its connection one-form \\(\\omega\\) can be defined by\n\n\\[\\nabla_Xe_1=\\omega(X)e_2, \\qquad \\nabla_Xe_2=-\\omega(X)e_1.\\]\n\nWith the corresponding curvature convention, the structural equation is\n\n\\[d\\omega=-K\\,dA.\\]\n\nThus Gaussian curvature measures the failure of the frame’s infinitesimal rotations to fit together without accumulated rotation.\n\nAlong a smooth curve, write the unit tangent as\n\n\\[T=\\cos\\varphi\\,e_1+\\sin\\varphi\\,e_2.\\]\n\nDifferentiation gives\n\n\\[k_g=\\frac{d\\varphi}{ds}+\\omega(T).\\]\n\nFor the positively oriented boundary of a disk contained in the patch, the tangent makes one full turn relative to a frame extending over the disk. If there are corners, their tangent jumps must also be counted. Applying Stokes’ theorem therefore yields\n\n\\[\\int_{\\partial D}k_g\\,ds+\\sum_j\\theta_j =2\\pi+\\int_{\\partial D}\\omega =2\\pi-\\int_DK\\,dA.\\]\n\nThis is the disk version of Gauss–Bonnet.\n\nTo obtain the closed-surface theorem, choose a sufficiently fine triangulation whose faces lie in suitable coordinate patches. Each face satisfies the disk formula. When the face formulas are summed, geodesic-curvature integrals along shared edges cancel because those edges are traversed in opposite directions. The angles around each smooth interior vertex sum to \\(2\\pi\\). For a closed triangular mesh, \\(3F=2E\\), so the remaining count becomes\n\n\\[\\int_MK\\,dA =2\\pi V-\\pi F =2\\pi(V-E+F).\\]\n\nThe appearance of the Euler characteristic is therefore not an extra assumption: it emerges from the cancellation of internal boundary terms and the combinatorics of gluing triangles.\n\nA second proof uses a vector field with isolated zeros. Away from its zeros, the normalized field supplies one vector of an orthonormal frame. Small disks around the zeros are removed, and the curvature identity is integrated on the remainder. In the limit, the turning around each removed disk records the integer index of the zero. The Poincaré–Hopf theorem identifies the sum of these indices with \\(\\chi(M)\\), giving the same result.\n\nThese proofs also reveal an obstruction. If a closed oriented surface had a global orthonormal frame, its connection form would be globally defined, making \\(K\\,dA\\) exact. Its integral would then vanish by Stokes’ theorem. Consequently, a surface with nonzero Euler characteristic cannot have such a frame. Local frames always exist, but their transition rotations encode global information that cannot be removed by choosing better coordinates."
    },
    {
      "heading": "Restrictions on geometry and conformal change",
      "text": "The most immediate consequences concern the possible signs of curvature. If a closed surface has \\(K>0\\) everywhere, its total curvature is positive and hence \\(\\chi(M)>0\\). Among connected closed orientable surfaces, only the sphere has positive Euler characteristic. If \\(K<0\\) everywhere, then \\(\\chi(M)<0\\), so an orientable surface must have genus at least two. If \\(K=0\\) everywhere, then \\(\\chi(M)=0\\).\n\nThese implications are necessary conditions, not a pointwise classification. A torus may have positive and negative curvature as long as their integrals cancel. The usual ring torus embedded in three-dimensional Euclidean space has positive curvature on its outer region and negative curvature on its inner region. It is therefore not intrinsically flat, even though its total curvature is zero. An abstract torus obtained by identifying opposite sides of a Euclidean parallelogram does carry a flat metric.\n\nFor constant curvature \\(K=c\\), Gauss–Bonnet reduces to\n\n\\[c\\,\\operatorname{Area}(M)=2\\pi\\chi(M).\\]\n\nEvery closed orientable hyperbolic surface of genus \\(g\\ge2\\), normalized to curvature \\(-1\\), has area\n\n\\[\\operatorname{Area}(M)=4\\pi(g-1).\\]\n\nDifferent hyperbolic metrics on the same topological surface can have different lengths of closed geodesics and different local arrangements, but their areas agree under this normalization. For a sphere of radius \\(R\\), \\(K=R^{-2}\\) and area \\(4\\pi R^2\\), again yielding total curvature \\(4\\pi\\).\n\nThe behavior under conformal changes makes metric invariance especially explicit. Let\n\n\\[\\widetilde g=e^{2u}g\\]\n\non a closed surface, where \\(u\\) is a smooth real-valued function. With the convention \\(\\Delta_g=\\operatorname{div}_g\\operatorname{grad}_g\\), the curvature transformation law is\n\n\\[\\widetilde K=e^{-2u}(K-\\Delta_gu), \\qquad dA_{\\widetilde g}=e^{2u}dA_g.\\]\n\nTherefore,\n\n\\[\\int_M\\widetilde K\\,dA_{\\widetilde g} =\\int_MK\\,dA_g-\\int_M\\Delta_gu\\,dA_g.\\]\n\nThe last integral vanishes by the divergence theorem. Total curvature is consequently unchanged, although curvature can be redistributed substantially from point to point.\n\nPrescribing a desired curvature function \\(\\widehat K\\) within a conformal class leads to the nonlinear equation\n\n\\[-\\Delta_gu+K=\\widehat K e^{2u}.\\]\n\nGauss–Bonnet supplies an integral compatibility condition for this partial differential equation, but does not by itself guarantee a solution. The uniformization theorem provides the broader existence result that every closed orientable surface admits a constant-curvature metric in each conformal class, with sign determined by its Euler characteristic. The distinction is important: Gauss–Bonnet gives the topological constraint, while uniformization establishes the existence of suitable metrics."
    },
    {
      "heading": "Polyhedral curvature and conical singularities",
      "text": "On a piecewise Euclidean surface assembled from flat triangles, Gaussian curvature vanishes in the interior of every face. Intrinsically, the interior of an edge is also flat: two adjacent triangles can be unfolded into a common plane. Curvature is concentrated at vertices, where the total angle may differ from \\(2\\pi\\).\n\nFor an interior vertex \\(v\\), define its angle defect by\n\n\\[\\delta(v)=2\\pi-\\sum_{f\\ni v}\\alpha_{f,v},\\]\n\nwhere \\(\\alpha_{f,v}\\) is the angle of face \\(f\\) at \\(v\\). The discrete Gauss–Bonnet theorem for a closed triangulated surface is\n\n\\[\\sum_v\\delta(v)=2\\pi\\chi(M).\\]\n\nIndeed, the sum of all face angles is \\(\\pi F\\), so the left side equals \\(2\\pi V-\\pi F\\). Using \\(3F=2E\\) converts this into \\(2\\pi(V-E+F)\\). No limiting process or smooth embedding is needed.\n\nFor a regular tetrahedron, three angles of \\(\\pi/3\\) meet at each vertex. Each vertex has defect \\(\\pi\\), and the four vertices contribute \\(4\\pi\\). At each vertex of a cube, three right angles meet, giving defect \\(\\pi/2\\); its eight vertices again contribute \\(4\\pi\\). Both polyhedra have the topology of a sphere. A saddle-type vertex can instead have total angle greater than \\(2\\pi\\), producing negative discrete curvature.\n\nFor a surface with boundary, define the defect at a boundary vertex by\n\n\\[\\delta_{\\partial}(v)=\\pi-\\sum_{f\\ni v}\\alpha_{f,v}.\\]\n\nThe corresponding formula is\n\n\\[\\sum_{v\\text{ interior}}\\delta(v) +\\sum_{v\\text{ boundary}}\\delta_{\\partial}(v) =2\\pi\\chi(M).\\]\n\nThe boundary terms combine intrinsic boundary turning with any polygonal corners. This form is useful in triangulated surface models because it preserves the global identity exactly, without estimating derivatives of a smooth metric.\n\nThe same principle applies to a smooth metric with isolated conical singularities. If the total angle around the \\(i\\)-th singular point is \\(\\beta_i\\), its concentrated curvature is \\(2\\pi-\\beta_i\\). Under the usual conical regularity assumptions,\n\n\\[\\int_{M_{\\mathrm{reg}}}K\\,dA+ \\sum_i(2\\pi-\\beta_i)=2\\pi\\chi(M),\\]\n\nwhere \\(M_{\\mathrm{reg}}\\) excludes the singular points.\n\nThese identities underlie curvature calculations in discrete differential geometry and surface processing. A vertex defect represents integrated curvature, not curvature per unit area. Dividing by a chosen local area produces a pointwise curvature estimate, but that estimate depends on the area convention and mesh quality. Likewise, bending a triangulated sheet along edges can change its appearance in space without changing its intrinsic angle defects. Gauss–Bonnet concerns intrinsic curvature, not merely visible bending."
    },
    {
      "heading": "The Gauss–Bonnet–Chern theorem in higher dimensions",
      "text": "For a closed oriented Riemannian manifold \\(M\\) of even dimension \\(2m\\), the higher-dimensional theorem reads\n\n\\[\\int_M e(TM,\\nabla)=\\chi(M).\\]\n\nHere \\(TM\\) is the tangent bundle, and \\(e(TM,\\nabla)\\) is the Euler form determined by its Levi-Civita connection. Unlike the surface formula, the integrand is generally not a single sectional curvature multiplied by volume. It is a specific degree-\\(m\\) polynomial in the curvature.\n\nIn a local oriented orthonormal frame, the curvature is represented by a skew-symmetric matrix \\(\\Omega\\) of two-forms. With compatible curvature and orientation conventions,\n\n\\[e(TM,\\nabla)= \\operatorname{Pf}\\!\\left(\\frac{\\Omega}{2\\pi}\\right),\\]\n\nwhere \\(\\operatorname{Pf}\\) is the Pfaffian. For a skew-symmetric \\(2m\\times2m\\) matrix \\(A\\), the Pfaffian is a polynomial satisfying\n\n\\[\\operatorname{Pf}(A)^2=\\det A.\\]\n\nFor a matrix of two-forms, multiplication is interpreted using exterior products. In dimension two, this construction reduces to \\(e(TM,\\nabla)=K\\,dA/(2\\pi)\\), recovering the classical theorem.\n\nThe Euler form represents the Euler class of the tangent bundle. This explains metric independence conceptually: changing the metric or compatible connection changes the differential-form representative but not its cohomology class. In particular, the difference between two Euler forms is exact, so its integral over a closed manifold is zero. This is a central example of Chern–Weil theory, which constructs topological characteristic classes from curvature.\n\nIn dimension four, a standard tensor expression is\n\n\\[\\chi(M)=\\frac{1}{32\\pi^2} \\int_M \\left( |\\mathrm{Rm}|^2-4|\\mathrm{Ric}|^2+R^2 \\right)dV.\\]\n\nHere \\(\\mathrm{Rm}\\) is the Riemann curvature tensor, \\(\\mathrm{Ric}\\) is its Ricci contraction, and \\(R\\) is scalar curvature; the norms are full tensor norms in an orthonormal frame. The combination matters. Integrating scalar curvature alone does not generally yield the Euler characteristic of a four-dimensional manifold.\n\nFor a unit round four-sphere, the three squared curvature quantities are \\(24\\), \\(36\\), and \\(144\\), respectively. The integrand is therefore \\(24\\). Since the sphere has volume \\(8\\pi^2/3\\), the formula gives \\(\\chi(S^4)=2\\).\n\nIn odd dimensions, every closed oriented manifold has Euler characteristic zero, as follows from Poincaré duality. There is no corresponding nonzero top-degree Euler curvature polynomial of the same kind. For even-dimensional manifolds with boundary, however, a higher-dimensional Gauss–Bonnet formula does exist. It requires a boundary correction involving both curvature and the boundary’s second fundamental form, generalizing the geodesic-curvature term for surfaces."
    },
    {
      "heading": "Differential forms, heat kernels, and index theory",
      "text": "Gauss–Bonnet also has an analytic interpretation through differential forms. On a closed Riemannian manifold, the exterior derivative \\(d\\) and its formal adjoint \\(d^*\\) define the Hodge Laplacian\n\n\\[\\Delta=dd^*+d^*d.\\]\n\nThe Hodge theorem identifies harmonic \\(p\\)-forms, those satisfying \\(\\Delta\\alpha=0\\), with classes in degree-\\(p\\) de Rham cohomology. Thus the \\(p\\)-th Betti number satisfies\n\n\\[b_p=\\dim\\ker\\Delta_p,\\]\n\nand the Euler characteristic is\n\n\\[\\chi(M)=\\sum_{p=0}^{n}(-1)^p b_p.\\]\n\nThis already turns a topological invariant into information about solutions of differential equations.\n\nThe operator\n\n\\[D=d+d^*\\]\n\nexchanges even-degree and odd-degree forms. Let \\(D^+\\) denote its restriction from even forms to odd forms. Its analytic index is\n\n\\[\\operatorname{ind}(D^+) =\\dim\\ker D^+-\\dim\\ker(D^+)^*.\\]\n\nSince \\(D^2=\\Delta\\), the two kernels consist of even and odd harmonic forms. Consequently,\n\n\\[\\operatorname{ind}(D^+)=\\chi(M).\\]\n\nGauss–Bonnet therefore computes the index of a particular elliptic differential operator by integrating curvature.\n\nThe heat-kernel proof explains how the local curvature expression arises. For \\(t>0\\), form the alternating heat trace\n\n\\[\\operatorname{Str}(e^{-t\\Delta}) =\\sum_{p=0}^{n}(-1)^p \\operatorname{Tr}(e^{-t\\Delta_p}).\\]\n\nThe contributions from every positive eigenvalue cancel between even and odd degrees. Only the zero modes remain, so this supertrace equals \\(\\chi(M)\\) for every \\(t>0\\). As \\(t\\) becomes large, that equality is visible from the survival of harmonic forms. As \\(t\\) tends to zero, the same quantity can be evaluated using the local asymptotic expansion of the heat kernel.\n\nIndividual heat kernels have singular leading terms as \\(t\\to0\\). In the alternating sum, the relevant lower-order contributions cancel, leaving the Euler curvature density. Integrating that local density gives the Gauss–Bonnet–Chern formula. This local cancellation is a substantive part of the proof, not a consequence of the ordinary heat equation on scalar functions alone.\n\nThe result is an important special case of the Atiyah–Singer index theorem, which relates analytic indices of elliptic operators to topological characteristic classes. For the de Rham operator, the topological expression is the Euler characteristic and its local curvature representative is the Euler form. This places Gauss–Bonnet within a larger framework connecting geometry, spectral theory, and topology."
    },
    {
      "heading": "Historical development",
      "text": "The theorem developed from several initially distinct subjects: triangle geometry, the intrinsic geometry of surfaces, polyhedral angle defects, and topological counting. Spherical trigonometry already related the area of a spherical triangle to its angular excess. On a sphere of radius \\(R\\), that relation is\n\n\\[\\operatorname{Area}(T) =R^2(\\alpha+\\beta+\\gamma-\\pi).\\]\n\nThe later theorem replaced constant spherical curvature with variable Gaussian curvature and replaced particular polygonal regions with surfaces of arbitrary topology.\n\nGauss’s 1827 Disquisitiones generales circa superficies curvas established the intrinsic nature of Gaussian curvature and developed the relation between curvature and geodesic triangles. The decisive conceptual shift was that curvature could be computed from measurements made within the surface. An ambient normal vector could assist calculations, but the curvature itself did not depend on how the surface was placed in three-dimensional space.\n\nBonnet’s 1848 work extended the relation to regions bounded by more general curves, introducing the boundary-turning contribution represented by geodesic curvature. The familiar closed-surface identity, expressed using the Euler characteristic, emerged through the subsequent development of global surface theory and topology. Walther von Dyck’s work in 1888 is an important stage in this global formulation. The present name therefore joins contributions made at different points in the development rather than identifying a single jointly published theorem.\n\nThe polyhedral side also had a long history. Descartes investigated total angular defect for polyhedra, while Leonhard Euler developed the vertex–edge–face relation. In the modern formulation, the total angle defect of a Euclidean polyhedral surface and the total Gaussian curvature of a smooth surface are parallel realizations of \\(2\\pi\\chi\\). Their unification became clearer as the Euler characteristic was understood as a topological invariant rather than merely a property of familiar convex solids.\n\nGeneralization beyond surfaces required an appropriate curvature integrand. Heinz Hopf obtained higher-dimensional results in the 1920s for hypersurfaces. Carl Allendoerfer in 1940, and Allendoerfer with André Weil in 1943, established broader higher-dimensional formulations. In 1944, Shiing-Shen Chern published an intrinsic proof that made the characteristic-form structure central and avoided dependence on an ambient Euclidean embedding.\n\nLater developments in characteristic classes, Hodge theory, and elliptic operator theory revealed further interpretations of the same identity. The heat-equation and index-theoretic approaches did not replace the triangle and boundary-turning proofs; they identified a common mechanism behind them: local differential data assemble into a global invariant, while changes in local choices contribute terms that cancel or integrate to zero."
    }
  ],
  "links": [
    "differential-geometry",
    "topology",
    "carl-friedrich-gauss",
    "manifold",
    "tangent-space",
    "euler-characteristic",
    "euclidean-space",
    "metric-tensor",
    "levi-civita-connection",
    "geodesic",
    "stokes-theorem",
    "vector-field",
    "conformal-map",
    "divergence-theorem",
    "partial-differential-equation",
    "tangent-bundle",
    "riemann-curvature-tensor",
    "differential-form",
    "leonhard-euler"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}