Élie Cartan

⚡ Key Takeaways

Élie Cartan (1869–1951) was a French mathematician whose work reshaped Lie theory, differential geometry, and the calculus of differential forms. One-line statement: Cartan classified the simple real and complex Lie algebras, invented the exterior calculus and the method of moving frames, developed the theory of connections on principal and fiber bundles, and discovered spinors. His ideas underpin modern gauge theory, general relativity, and the geometry of homogeneous and symmetric spaces. He worked primarily in Paris (École Normale Supérieure, Sorbonne) from the 1890s through the 1940s.

💡 The Core Insight

Cartan's unifying vision was that local geometric structure is encoded in differential forms valued in a Lie algebra, and that the obstruction to "flattening" such a structure — its curvature — is the failure of an integrability condition expressed by his structure equations. Whether the object is a Lie group, a Riemannian manifold, a connection on a bundle, or a system of PDEs, one attaches a moving frame (a field of bases adapted to the geometry), reads off the Maurer–Cartan forms or connection forms, and computes their exterior derivatives. The resulting torsion and curvature are the complete local invariants. This single mechanism subsumes Lie's theory of groups, Riemann's curvature, and the gauge fields of physics.

DefinitionLie Group, Lie Algebra, and the Maurer–Cartan Form

A Lie group $G$ is a smooth manifold with smooth group operations. Its Lie algebra $\mathfrak{g} = T_eG$ is the tangent space at the identity, equipped with the bracket $[X,Y]$ given by the commutator of left-invariant vector fields. The Maurer–Cartan form $\omega$ is the unique $\mathfrak{g}$ -valued left-invariant 1-form on $G$ that is the identity on $\mathfrak{g} = T_eG$ :

1

\quad \omega_g = (L_{g^{-1}})_* : T_gG \xrightarrow{\;\sim\;} \mathfrak{g}.

Derivation hint: $\omega_g(v) = (dL_{g^{-1}})_g(v)$ pulls each tangent vector back to the identity by left translation; left-invariance is immediate since $L_{g^{-1}}\circ L_h = L_{(hg)^{-1}}^{-1}$ composes consistently.

TheoremMaurer–Cartan Structure Equation

The Maurer–Cartan form of any Lie group satisfies

2

\quad d\omega + \tfrac{1}{2}[\omega,\omega] = 0,

where $[\omega,\omega](X,Y) = [\omega(X),\omega(Y)] - [\omega(Y),\omega(X)] = 2[\omega(X),\omega(Y)]$ . Conversely, a $\mathfrak{g}$ -valued 1-form $\omega$ on a simply connected manifold satisfying [2] and pointwise isomorphism with $\mathfrak{g}$ is locally the pullback of the Maurer–Cartan form, integrating to a map into $G$ (Cartan's integration / "fundamental theorem"). Derivation hint: Evaluate on left-invariant fields $X,Y$ where $\omega(X),\omega(Y)$ are constant; then $d\omega(X,Y) = -\omega([X,Y]) = -[\omega(X),\omega(Y)]$ , which rearranges to [2].

DefinitionCartan Connection on a Principal Bundle

Let $\pi:P\to M$ be a principal $G$ -bundle with right $G$ -action. A connection 1-form (Ehresmann–Cartan connection) is a $\mathfrak{g}$ -valued 1-form $A$ on $P$ such that

3

\quad A(\xi^*) = \xi \text{ for } \xi\in\mathfrak{g}, \qquad R_g^*A = \mathrm{Ad}(g^{-1})A,

where $\xi^*$ is the fundamental vector field of $\xi$ and $R_g$ is right translation. The curvature 2-form is

4

\quad F = dA + \tfrac12[A,A] = dA + A\wedge A.

Derivation hint: Condition [3] declares fundamental vertical directions to map to their generators and demands equivariance; the curvature [4] is the covariant exterior derivative $D A = dA \circ h$ projected horizontally, with $A\wedge A$ packaging the bracket.

TheoremCartan's First and Second Structure Equations

For an affine/orthonormal frame with coframe $\theta^a$ and connection forms $\omega^a{}_b$ on a manifold with a linear connection, the torsion $T^a$ and curvature $\Omega^a{}_b$ satisfy

5

\quad T^a = d\theta^a + \omega^a{}_b \wedge \theta^b,

6

\quad \Omega^a{}_b = d\omega^a{}_b + \omega^a{}_c \wedge \omega^c{}_b.

Taking exterior derivatives yields the Bianchi identities $dT^a + \omega^a{}_b\wedge T^b = \Omega^a{}_b\wedge\theta^b$ and $d\Omega^a{}_b + \omega^a{}_c\wedge\Omega^c{}_b - \Omega^a{}_c\wedge\omega^c{}_b = 0$ . Derivation hint: Define $T^a, \Omega^a{}_b$ by [5]–[6]; differentiate using $d^2=0$ and $d(\alpha\wedge\beta)=d\alpha\wedge\beta-\alpha\wedge d\beta$ , substituting the defining equations back to eliminate $d\theta, d\omega$ .

ProofKilling–Cartan Classification of Complex Simple Lie Algebras

Goal. Show every complex simple Lie algebra $\mathfrak{g}$ is one of $A_n,B_n,C_n,D_n$ or $G_2,F_4,E_6,E_7,E_8$ .

Step 1 — Cartan subalgebra. Choose a maximal abelian subalgebra $\mathfrak{h}$ of semisimple (ad-diagonalizable) elements; its dimension is the rank $\ell$ . Cartan proved $\mathfrak{h}$ is self-normalizing (a Cartan subalgebra) and that all are conjugate.

Step 2 — Root space decomposition. Since $\mathrm{ad}(\mathfrak{h})$ is a commuting family of semisimple operators,

7

\quad \mathfrak{g} = \mathfrak{h} \oplus \bigoplus_{\alpha\in\Phi}\mathfrak{g}_\alpha, \qquad \mathfrak{g}_\alpha = \{X : [H,X]=\alpha(H)X\ \forall H\in\mathfrak{h}\}.

Derivation hint: simultaneous eigenspace decomposition of commuting semisimple operators on a finite-dimensional space; $\alpha\in\mathfrak{h}^*$ are the roots.

Step 3 — Killing form. The symmetric bilinear form $B(X,Y)=\mathrm{tr}(\mathrm{ad}X\,\mathrm{ad}Y)$ is nondegenerate iff $\mathfrak{g}$ is semisimple (Cartan's criterion). It identifies $\mathfrak{h}\cong\mathfrak{h}^*$ and induces an inner product on $\mathrm{span}_\mathbb{R}\Phi$ .

Step 4 — Root system axioms. Each $\dim\mathfrak{g}_\alpha=1$ ; if $\alpha\in\Phi$ then $-\alpha\in\Phi$ and $2\alpha\notin\Phi$ ; the Cartan integers $\langle\beta,\alpha\rangle = 2(\beta,\alpha)/(\alpha,\alpha)\in\mathbb{Z}$ ; reflections $s_\alpha$ permute $\Phi$ . This makes $\Phi$ an abstract root system.

Step 5 — Cartan matrix & Dynkin diagram. Choose simple roots $\{\alpha_1,\dots,\alpha_\ell\}$ ; form the Cartan matrix $C_{ij}=\langle\alpha_i,\alpha_j\rangle$ with $C_{ii}=2$ , $C_{ij}\in\{0,-1,-2,-3\}$ . The angles between simple roots are constrained to $\{90°,120°,135°,150°\}$ .

Step 6 — Classification of admissible diagrams. A combinatorial argument (positive-definiteness of $C$ ) shows the only connected admissible diagrams are the four infinite families and the five exceptionals. Each diagram reconstructs $\mathfrak{g}$ uniquely (Serre relations), completing the classification. $\blacksquare$

∎

ÉC

Élie Cartan

1869–1951

Born to a blacksmith in the alpine village of Dolomieu, Isère, Cartan rose through a meritocratic scholarship system, entered the École Normale Supérieure in 1888, and defended his landmark thesis in 1894. He worked in relative obscurity for decades — his abstract style was ahead of its time — before being recognized as a giant. He held chairs at Montpellier, Lyon, Nancy, and finally the Sorbonne (1912–1940). Famously generous and modest, he corresponded with Einstein on teleparallel gravity and influenced Weyl, Chevalley, Ehresmann, and Chern. His son Henri Cartan became a founding member of Bourbaki and a major mathematician in his own right (sheaf theory, homological algebra). Élie's work was so deep that much of twentieth-century differential geometry consists of unpacking it.

🧠 Intuition

Imagine standing on a curved surface holding a small rigid frame (a set of orthonormal axes). As you walk an infinitesimal loop and return to your start, the frame may come back rotated — that rotation, per unit area, is the curvature. Cartan's structure equation [6] is exactly the bookkeeping of how the frame must rotate ( $\omega^a{}_b$ ) to stay adapted to the geometry, and the failure of that rotation to be "integrable" (path-independent) is $\Omega^a{}_b$ . Torsion [5] is a complementary defect: walk along two directions in either order, and torsion measures whether the resulting infinitesimal parallelogram fails to close. The genius of the moving frame method is to stop fighting with coordinates and instead let the frame itself carry all the geometric information, encoded entirely in 1-forms and their exterior derivatives.

Figure

Holonomy: parallel transport of a frame around a loop on a curved surface yields a rotation equal to the integrated curvature

ExampleCurvature of the round 2-sphere via moving frames

Given. Sphere of radius $R$ with metric $ds^2 = R^2(d\vartheta^2 + \sin^2\vartheta\, d\varphi^2)$ .

Step 1 — Orthonormal coframe.

8

\quad \theta^1 = R\, d\vartheta, \qquad \theta^2 = R\sin\vartheta\, d\varphi.

Derivation hint: read coefficients so that $ds^2=(\theta^1)^2+(\theta^2)^2$ .

Step 2 — Solve first structure equation with zero torsion ( $d\theta^a = -\omega^a{}_b\wedge\theta^b$ ). $d\theta^1 = 0$ , $d\theta^2 = R\cos\vartheta\, d\vartheta\wedge d\varphi = \cot\vartheta\, d\vartheta\wedge\theta^2$ . With the antisymmetry $\omega^1{}_2=-\omega^2{}_1$ , one finds

9

\quad \omega^1{}_2 = -\cos\vartheta\, d\varphi.

Derivation hint: try $\omega^1{}_2 = f\,d\varphi$ ; matching $d\theta^2 = -\omega^2{}_1\wedge\theta^1=\omega^1{}_2\wedge\theta^1$ fixes $f=-\cos\vartheta$ .

Step 3 — Second structure equation.

10

\quad \Omega^1{}_2 = d\omega^1{}_2 = \sin\vartheta\, d\vartheta\wedge d\varphi = \frac{1}{R^2}\,\theta^1\wedge\theta^2.

Derivation hint: $d(-\cos\vartheta\,d\varphi)=\sin\vartheta\,d\vartheta\wedge d\varphi$ ; convert to the coframe using [8].

Result. Gaussian curvature $K = \boxed{1/R^2}$ (units: $\text{length}^{-2}$ ), constant and positive, recovered with no Christoffel symbols.

ExampleMaurer–Cartan form of

SU(2)

Given. $g\in SU(2)$ , $g = \begin{pmatrix} a & -\bar b\\ b & \bar a\end{pmatrix}$ , $|a|^2+|b|^2=1$ . Compute $\omega = g^{-1}dg$ .

Step 1 — Matrix Maurer–Cartan form.

11

\quad \omega = g^{-1}\,dg \in \mathfrak{su}(2),

a traceless anti-Hermitian $2\times2$ matrix. Derivation hint: for matrix groups the left-invariant form is literally $g^{-1}dg$ ; $d(g^{-1}g)=0$ gives $d(g^{-1})=-g^{-1}dg\,g^{-1}$ .

Step 2 — Expand in Pauli basis. Writing $\omega = \tfrac{i}{2}(\sigma_1\omega^1+\sigma_2\omega^2+\sigma_3\omega^3)$ defines real 1-forms $\omega^j$ , the components of the coframe on the group manifold $S^3$ .

Step 3 — Verify structure equation.

12

\quad d\omega^1 = -\omega^2\wedge\omega^3,\quad \text{(and cyclic)},

the explicit form of $d\omega+\tfrac12[\omega,\omega]=0$ for $\mathfrak{su}(2)\cong\mathbb{R}^3$ with the cross product. Derivation hint: $[\sigma_a,\sigma_b]=2i\epsilon_{abc}\sigma_c$ converts the bracket term to $-\epsilon^a{}_{bc}\omega^b\wedge\omega^c$ .

Result. $SU(2)\cong S^3$ carries a left-invariant coframe satisfying $\boxed{d\omega^a = -\tfrac12\epsilon^a{}_{bc}\,\omega^b\wedge\omega^c}$ , exhibiting it as a (constant-curvature) Lie group geometry — the structure constants $\epsilon_{abc}$ are dimensionless.

⚠️ Common mistake

Torsion is not curvature, and a "Cartan connection" need not be the Levi-Civita one. Many students conflate the two structure equations. Torsion $T^a$ [5] measures the failure of infinitesimal parallelograms to close and depends on the antisymmetric part of the connection; curvature $\Omega^a{}_b$ [6] measures holonomy. The Levi-Civita connection is the unique metric-compatible, torsion-free one — but Cartan deliberately studied connections with torsion (e.g. teleparallel gravity, Einstein–Cartan theory), where spin sources torsion. Setting $T^a=0$ is a choice, not a theorem.

Counterexample

A flat connection can still have nontrivial holonomy. Consider a flat connection ( $\Omega=0$ ) on a non-simply-connected base, e.g. a flat line bundle over the circle $S^1$ with monodromy $g\in U(1)\setminus\{1\}$ . Locally $A$ is pure gauge and curvature vanishes everywhere, yet parallel transport around the loop gives a nontrivial rotation. This shows curvature is only the local obstruction to flatness; global holonomy is governed by $\pi_1$ , a distinction Cartan's structure equations alone cannot detect. (This is the genesis of the connection between holonomy and the fundamental group.)

connection

Cartan's machinery sits at the crossroads of modern geometry. The classification of Lie Algebras feeds directly into Representation Theory and the structure of Symmetric Spaces, which Cartan classified using involutions. His Exterior Derivative and Differential Forms provide the calculus behind de Rham Cohomology and Stokes' Theorem. The connection 1-form [3] and curvature [4] are precisely the Gauge Theory potentials and field strengths, linking him to Yang–Mills Theory and the Principal Bundle formalism. His structure equations [5]–[6] are the frame-bundle expression of Riemannian Geometry and feed into General Relativity via the tetrad/Cartan formalism. Spinors connect to Clifford Algebra and the Dirac Equation.

🔬 Where this appears

⏳ History

- 1888–1891: Cartan enters the École Normale Supérieure; absorbs the work of Sophus Lie and Wilhelm Killing on continuous groups.

- 1894: Doctoral thesis Sur la structure des groupes de transformations finis et continus — rigorously establishes and corrects Killing's classification of complex simple Lie algebras; proves the conjugacy of Cartan subalgebras and Cartan's solvability criterion.

- 1899–1902: Develops the exterior calculus and the theory of Pfaffian systems, introducing the exterior derivative in its modern form.

- 1904–1908: Work on infinite-dimensional ("infinite continuous") pseudogroups; foundational papers on the general theory of moving frames (repère mobile).

- 1913: Discovers spinors while classifying the irreducible representations of the orthogonal Lie algebras.

- 1914–1925: Classifies the real forms of simple Lie algebras and develops the theory of symmetric spaces (essentially complete by 1926–27 in joint context with the theory of Riemannian holonomy).

- 1922–1925: Introduces the general notion of a connection (affine, projective, conformal, "Cartan geometries") generalizing Levi-Civita; correspondence with Einstein on teleparallelism (1929).

- 1925–1930: La théorie des groupes finis et continus et l'Analysis Situs and work linking topology of Lie groups to their algebra (with influence on de Rham and Hopf).

- 1945: Les systèmes différentiels extérieurs et leurs applications géométriques consolidates the Cartan–Kähler theory of exterior differential systems.

open-problem

Equivalence problems and the reach of the moving-frame method. Cartan's general equivalence method seeks complete invariants determining when two geometric structures (G-structures, PDE systems, Cartan geometries) are locally equivalent. While powerful, its full algorithmic content — managing the prolongation and involutivity of exterior differential systems (Cartan–Kähler theorem requires real-analyticity; the smooth case is governed by the harder Cartan–Kuranishi prolongation) — remains computationally formidable, and a fully effective classification of singular/non-involutive systems is open. Likewise, the classification of pseudogroups (Cartan's "infinite continuous groups") was not put on fully rigorous modern footing until much later (Kuranishi, Singer–Sternberg, and ongoing work in the geometry of jets and PDEs), and a complete intrinsic theory of parabolic geometries and their invariants (the modern descendant via Tanaka–Morimoto) is still actively developed. Determining sharp existence/regularity for general Cartan-geometric structures continues to generate open questions.

Formal Statement

Why It Holds / Derivation

Building Intuition

Worked Examples

Common Misconceptions

Connections

Real-World Applications

Historical Development

Subtleties & Limits

See Also