Lie Algebras — Structure and Classification

⚡ Key Takeaways

One line: Every finite-dimensional Lie algebra over a field of characteristic zero decomposes into a solvable radical and a semisimple part (Levi decomposition), and the semisimple part is a direct sum of simple Lie algebras classified completely by Dynkin diagrams (four infinite families $A_n, B_n, C_n, D_n$ plus five exceptionals $G_2, F_4, E_6, E_7, E_8$ ). Key result: Structure theory reduces the study of all Lie algebras to (1) understanding solvable algebras and (2) the finite classification of simple algebras via root systems and Cartan matrices. Applies to: Differential geometry, representation theory, particle physics (gauge groups), integrable systems, algebraic groups, harmonic analysis. Who/when: Sophus Lie (1870s), Wilhelm Killing (1888–90), Élie Cartan (1894 thesis), Hermann Weyl, B. L. van der Waerden, E. B. Dynkin (1947).

💡 The Core Insight

The bracket is too unwieldy to attack directly, so one studies how $\mathfrak{g}$ acts on itself via the adjoint representation $\mathrm{ad}_x(y)=[x,y]$ . Choosing a maximal commuting set of diagonalizable elements (a Cartan subalgebra $\mathfrak{h}$ ), the algebra decomposes into simultaneous eigenspaces. The eigenvalues are linear functionals called roots, and they form a highly symmetric finite geometric configuration — a root system — invariant under reflections. The entire algebra, including its multiplication table, is reconstructible from this discrete combinatorial shadow. Classification of Lie algebras thus becomes classification of root systems, which becomes classification of certain integer matrices and connected diagrams.

DefinitionLie Algebra

A Lie algebra over a field $k$ is a $k$ -vector space $\mathfrak{g}$ with a bilinear map $[\cdot,\cdot]:\mathfrak{g}\times\mathfrak{g}\to\mathfrak{g}$ such that

1

\quad [x,x]=0 \quad\text{(alternating)}, \qquad [x,[y,z]]+[y,[z,x]]+[z,[x,y]]=0 \quad\text{(Jacobi)}

Derivation hint: Antisymmetry $[x,y]=-[y,x]$ follows by expanding $[x+y,x+y]=0$ ; the Jacobi identity is equivalent to $\mathrm{ad}_x$ being a derivation of the bracket. Here $x,y,z\in\mathfrak{g}$ . The adjoint map is $\mathrm{ad}_x:\mathfrak{g}\to\mathfrak{g}$ , $\mathrm{ad}_x(y)=[x,y]$ , and Jacobi says exactly

2

\quad \mathrm{ad}_{[x,y]}=[\mathrm{ad}_x,\mathrm{ad}_y]=\mathrm{ad}_x\mathrm{ad}_y-\mathrm{ad}_y\mathrm{ad}_x.

Derivation hint: Apply both sides to an arbitrary $z$ and reorganize the three Jacobi terms; this exhibits $\mathrm{ad}:\mathfrak{g}\to\mathfrak{gl}(\mathfrak{g})$ as a Lie algebra homomorphism.

DefinitionIdeals, Solvable, Nilpotent, Semisimple

A subspace $\mathfrak{a}\subseteq\mathfrak{g}$ is an ideal if $[\mathfrak{g},\mathfrak{a}]\subseteq\mathfrak{a}$ . Define the derived series and lower central series:

3

\quad \mathfrak{g}^{(0)}=\mathfrak{g},\;\; \mathfrak{g}^{(i+1)}=[\mathfrak{g}^{(i)},\mathfrak{g}^{(i)}]; \qquad \mathfrak{g}^0=\mathfrak{g},\;\; \mathfrak{g}^{i+1}=[\mathfrak{g},\mathfrak{g}^i].

Derivation hint: Each term is an ideal because the bracket of ideals is an ideal; nesting follows from $[\mathfrak{a},\mathfrak{a}]\subseteq\mathfrak{a}$ and induction. $\mathfrak{g}$ is solvable if $\mathfrak{g}^{(n)}=0$ for some $n$ , nilpotent if $\mathfrak{g}^n=0$ . The radical $\mathrm{rad}(\mathfrak{g})$ is the unique maximal solvable ideal. $\mathfrak{g}$ is semisimple if $\mathrm{rad}(\mathfrak{g})=0$ , and simple if it is non-abelian with no proper nonzero ideals.

DefinitionKilling Form and Cartan Subalgebra

The Killing form is the symmetric bilinear form

4

\quad \kappa(x,y)=\mathrm{tr}\!\big(\mathrm{ad}_x\,\mathrm{ad}_y\big).

Derivation hint: It is symmetric since $\mathrm{tr}(AB)=\mathrm{tr}(BA)$ , and <strong>invariant</strong> ( $\kappa([x,y],z)=\kappa(x,[y,z])$ ) follows from the trace identity $\mathrm{tr}([A,B]C)=\mathrm{tr}(A[B,C])$ together with [2]. A Cartan subalgebra (CSA) $\mathfrak{h}\subseteq\mathfrak{g}$ is a nilpotent self-normalizing subalgebra; for semisimple $\mathfrak{g}$ over $\mathbb{C}$ it is a maximal abelian subalgebra acting diagonalizably. Its dimension is the rank of $\mathfrak{g}$ .

TheoremCartan's Criterion

A finite-dimensional Lie algebra $\mathfrak{g}$ over a field of characteristic $0$ is:

solvable iff $\kappa(x,y)=0$ for all $x\in\mathfrak{g}$ , $y\in[\mathfrak{g},\mathfrak{g}]$ ;
semisimple iff the Killing form $\kappa$ is nondegenerate.

TheoremLevi Decomposition

Every finite-dimensional Lie algebra $\mathfrak{g}$ over a field of characteristic $0$ admits a decomposition

5

\quad \mathfrak{g}=\mathfrak{s}\ltimes \mathrm{rad}(\mathfrak{g}),

where $\mathfrak{s}$ is a semisimple subalgebra (a Levi subalgebra), unique up to conjugation by inner automorphisms (Malcev–Harish-Chandra). Derivation hint: Use the vanishing of the second cohomology $H^2(\mathfrak{s},\mathrm{rad})$ (Whitehead's lemma) to split the short exact sequence $0\to\mathrm{rad}\to\mathfrak{g}\to\mathfrak{s}\to0$ .

TheoremRoot Space Decomposition

Let $\mathfrak{g}$ be semisimple over $\mathbb{C}$ with CSA $\mathfrak{h}$ . Then

6

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

where $\Phi\subset\mathfrak{h}^*\setminus\{0\}$ is the finite set of roots, each $\dim\mathfrak{g}_\alpha=1$ , and $[\mathfrak{g}_\alpha,\mathfrak{g}_\beta]\subseteq\mathfrak{g}_{\alpha+\beta}$ . Derivation hint: Since $\mathrm{ad}_\mathfrak{h}$ is a commuting family of diagonalizable operators, simultaneously diagonalize; the eigenfunctionals are the roots, and $[\mathfrak{g}_\alpha,\mathfrak{g}_\beta]\subseteq\mathfrak{g}_{\alpha+\beta}$ follows from Jacobi.

TheoremClassification of Simple Lie Algebras

The simple finite-dimensional Lie algebras over $\mathbb{C}$ are in bijection with connected Dynkin diagrams, namely the four classical families and five exceptionals:

7

\quad A_n\,(n\ge1),\; B_n\,(n\ge2),\; C_n\,(n\ge3),\; D_n\,(n\ge4),\; G_2,\,F_4,\,E_6,\,E_7,\,E_8.

Derivation hint: Reduce to classifying irreducible root systems; the admissible Cartan matrices are constrained by the angle/length conditions [9], leaving precisely these connected diagrams (Killing–Cartan list).

ProofFrom CSA to Root System

Step 1 — Diagonalize. For semisimple $\mathfrak{g}$ , choose a maximal toral (diagonalizable abelian) subalgebra $\mathfrak{h}$ ; it equals its own centralizer, hence is a CSA. The commuting operators $\{\mathrm{ad}_h\}_{h\in\mathfrak{h}}$ are simultaneously diagonalizable, giving the decomposition [6].

Step 2 — Killing form pairs $\mathfrak{g}_\alpha$ with $\mathfrak{g}_{-\alpha}$ . Invariance gives $\kappa(\mathfrak{g}_\alpha,\mathfrak{g}_\beta)=0$ unless $\alpha+\beta=0$ . Nondegeneracy (Cartan's criterion) forces $-\alpha\in\Phi$ whenever $\alpha\in\Phi$ , and identifies $\mathfrak{h}$ with $\mathfrak{h}^*$ , transferring an inner product $(\cdot,\cdot)$ to the roots.

Step 3 — $\mathfrak{sl}_2$ triples. For each $\alpha$ , pick $e_\alpha\in\mathfrak{g}_\alpha$ , $f_\alpha\in\mathfrak{g}_{-\alpha}$ with $h_\alpha=[e_\alpha,f_\alpha]\in\mathfrak{h}$ normalized so that

8

\quad [h_\alpha,e_\alpha]=2e_\alpha,\quad [h_\alpha,f_\alpha]=-2f_\alpha,\quad [e_\alpha,f_\alpha]=h_\alpha,

giving a subalgebra $\cong\mathfrak{sl}_2(\mathbb{C})$ . Derivation hint: Rescale $e_\alpha,f_\alpha$ using $\kappa(e_\alpha,f_\alpha)\ne0$ guaranteed by Step 2.

Step 4 — Integrality. Representation theory of $\mathfrak{sl}_2$ forces every weight $\beta(h_\alpha)$ to be an integer. Hence for all roots

9

\quad \langle\beta,\alpha\rangle := \frac{2(\beta,\alpha)}{(\alpha,\alpha)}\in\mathbb{Z}, \qquad s_\alpha(\beta)=\beta-\langle\beta,\alpha\rangle\,\alpha\in\Phi.

Derivation hint: Apply $\mathrm{ad}_{e_\alpha},\mathrm{ad}_{f_\alpha}$ repeatedly to $\mathfrak{g}_\beta$ ; the resulting $\mathfrak{sl}_2$ -string forces integral eigenvalues and reflection-invariance.

Step 5 — Root system axioms. $\Phi$ is finite, spans $\mathfrak{h}^*$ , contains $\pm\alpha$ but no other multiple of $\alpha$ , and is closed under the reflections $s_\alpha$ with integer Cartan pairings. This is exactly the definition of a reduced root system.

Step 6 — Base, Cartan matrix, Dynkin diagram. Choose a set of simple roots $\Delta=\{\alpha_1,\dots,\alpha_n\}$ (a basis with every root a non-negative or non-positive integer combination). Form the Cartan matrix $A_{ij}=\langle\alpha_i,\alpha_j\rangle$ . The possible angles between simple roots are $90^\circ,120^\circ,135^\circ,150^\circ$ , encoded by $0,1,2,3$ edges in the Dynkin diagram.

Step 7 — Classification of connected diagrams. The positive-definiteness of $(\cdot,\cdot)$ severely limits admissible diagrams: no cycles, at most one branch point or one multiple edge, branch/edge length bounds. Enumerating yields precisely the list [7].

Step 8 — Reconstruction (Serre). Conversely each Cartan matrix yields a Lie algebra via the Serre relations on generators $e_i,f_i,h_i$ :

10

\quad [h_i,h_j]=0,\;\; [e_i,f_j]=\delta_{ij}h_i,\;\; [h_i,e_j]=A_{ji}e_j,\;\; (\mathrm{ad}\,e_i)^{1-A_{ji}}e_j=0,

and the analogous $f$ -relations. This proves existence and uniqueness, completing the bijection. $\blacksquare$ Derivation hint: The free Lie algebra modulo these relations is finite-dimensional exactly when $A$ is of finite (positive-definite) type; the Chevalley basis makes the structure constants integral.

∎

mathematician · Wilhelm Killing & Élie Cartan | 1847–1923, 1869–1951

Killing, a Gymnasium teacher working in near-isolation in Braunsberg, introduced the characteristic equation of $\mathrm{ad}_x$ , the rank, the Cartan subalgebra, root systems, and even discovered the exceptional algebras — in a series of four papers (1888–1890) Cartan later called "the most beautiful chapter of algebra." His work was error-prone and his proofs incomplete; Cartan's 1894 thesis corrected, rigorized, and extended everything, then went on to classify real forms, symmetric spaces, and spinors. Hermann Weyl unified the theory with compact group methods and the character formula; Dynkin (1947) gave the elegant diagram bookkeeping now universally used. The mismatch in credit ("Cartan subalgebra," "Killing form") reflects how much Cartan rescued from Killing's pioneering but flawed manuscripts.

🧠 Intuition

Think of a semisimple Lie algebra as a crystal. The Cartan subalgebra is a "ruler" — a maximal set of measurements you can take simultaneously without disturbing each other (commuting observables). Every other direction in the algebra is an eigenvector of this ruler, and its eigenvalue is a root: a vector recording "how the ruler sees it." The miracle is that these root vectors are not scattered randomly — they snap to a lattice with crystalline symmetry, invariant under a finite reflection group (the Weyl group). The angles between roots can only be $90^\circ$ , $120^\circ$ , $135^\circ$ , or $150^\circ$ — like the rigid bond angles of a molecule. Because the angles are so constrained, only finitely many "crystal shapes" exist, and that finiteness is exactly the classification theorem. The whole multiplication table of $\mathfrak{g}$ is just the geometry of how these root-vectors add ( $[\mathfrak{g}_\alpha,\mathfrak{g}_\beta]\subseteq\mathfrak{g}_{\alpha+\beta}$ ).

Figure

Root system of

A_2

(

\mathfrak{sl}_3

): six roots at

60^\circ

spacing, with simple roots

\alpha_1,\alpha_2

and the Weyl reflection hyperplanes

ExampleRoot System of

\mathfrak{sl}_3(\mathbb{C})

=

A_2

Given: $\mathfrak{g}=\mathfrak{sl}_3$ , traceless $3\times3$ matrices, $\dim=8$ . Step 1 — CSA: diagonal traceless matrices $\mathfrak{h}=\{\mathrm{diag}(a_1,a_2,a_3):\sum a_i=0\}$ , rank $=2$ . Step 2 — Roots: with $\varepsilon_i(\mathrm{diag}(a))=a_i$ , the matrix unit $E_{ij}$ ( $i\ne j$ ) satisfies $[h,E_{ij}]=(a_i-a_j)E_{ij}$ , so roots are $\alpha_{ij}=\varepsilon_i-\varepsilon_j$ , giving $6$ roots: $\pm(\varepsilon_1-\varepsilon_2),\pm(\varepsilon_2-\varepsilon_3),\pm(\varepsilon_1-\varepsilon_3)$ . Step 3 — Simple roots: $\alpha_1=\varepsilon_1-\varepsilon_2$ , $\alpha_2=\varepsilon_2-\varepsilon_3$ ; positive root $\alpha_1+\alpha_2=\varepsilon_1-\varepsilon_3$ . Step 4 — Cartan matrix: using $(\alpha_i,\alpha_j)$ with all roots same length, $\langle\alpha_1,\alpha_2\rangle=-1$ .

\boxed{\;A=\begin{pmatrix}2&-1\\-1&2\end{pmatrix},\quad \dim\mathfrak{sl}_3=2+6=8,\quad \text{type }A_2\;}

Check: dimension $=\dim\mathfrak{h}+|\Phi|=2+6=8$ . ✓ Weyl group $=S_3$ , order $6$ .

ExampleKilling Form of

\mathfrak{sl}_2(\mathbb{C})

Given: basis $e=\begin{pmatrix}0&1\\0&0\end{pmatrix},\,f=\begin{pmatrix}0&0\\1&0\end{pmatrix},\,h=\begin{pmatrix}1&0\\0&-1\end{pmatrix}$ , with $[h,e]=2e,\,[h,f]=-2f,\,[e,f]=h$ . Step 1 — adjoint matrices in basis $(e,h,f)$ : $\mathrm{ad}_h=\mathrm{diag}(2,0,-2)$ . Step 2 — $\kappa(h,h)=\mathrm{tr}(\mathrm{ad}_h^2)=2^2+0+(-2)^2=8.$ Step 3 — $\kappa(e,f)$ : compute $\mathrm{ad}_e\mathrm{ad}_f$ ; one finds $\mathrm{tr}=4$ . Also $\kappa(e,e)=\kappa(f,f)=0$ , $\kappa(h,e)=\kappa(h,f)=0$ .

\boxed{\;\kappa=\begin{pmatrix}0&0&4\\0&8&0\\4&0&0\end{pmatrix}\ \text{in basis }(e,h,f),\quad \det\kappa=-128\ne0\;}

Conclusion: nondegenerate $\Rightarrow$ $\mathfrak{sl}_2$ semisimple (here simple). General formula: $\kappa=2N\cdot\mathrm{tr}(XY)$ for $\mathfrak{sl}_N$ , here $N=2$ gives factor $4$ . ✓

⚠️ Common mistake

"Semisimple = simple + simple is the only structure." Semisimple means a direct sum of simple ideals, but a general Lie algebra is not semisimple — it has a solvable radical, and Levi decomposition is only a semidirect product, not a direct sum. The radical genuinely interacts with the Levi part.

"The Killing form classifies everything." Cartan's criterion is for char $0$ . Over fields of positive characteristic the Killing form can be degenerate even for "good" algebras, and the classification of simple modular Lie algebras (Block–Wilson–Strade–Premet) is vastly more complicated, including Cartan-type algebras with no classical analogue.

"Rank = dimension." Rank is $\dim\mathfrak{h}$ , the size of the Cartan subalgebra, far smaller than $\dim\mathfrak{g}$ . E.g. $E_8$ has rank $8$ but dimension $248$ .

Counterexample

Solvable but not nilpotent. Let $\mathfrak{g}=\mathrm{span}\{x,y\}$ with $[x,y]=y$ . Then $\mathfrak{g}^{(1)}=\mathrm{span}\{y\}$ , $\mathfrak{g}^{(2)}=0$ , so $\mathfrak{g}$ is solvable. But $\mathfrak{g}^1=\mathrm{span}\{y\}$ , $\mathfrak{g}^2=[\mathfrak{g},\mathfrak{g}^1]=\mathrm{span}\{[x,y]\}=\mathrm{span}\{y\}\ne0$ — the lower central series stabilizes at $\mathrm{span}\{y\}$ , so $\mathfrak{g}$ is not nilpotent. This shows the two notions differ, and that the abelian-vs-nilpotent-vs-solvable hierarchy is strict.

Killing form nondegenerate but algebra not semisimple — impossible in char 0, but in char $p$ : $\mathfrak{sl}_p$ over $\mathbb{F}_p$ has degenerate Killing form (the identity is central), illustrating why char-0 is essential.

connection

Structure theory of Lie algebras is the infinitesimal shadow of Lie Groups: the Lie algebra is $T_e G$ with bracket from the group commutator, and semisimplicity corresponds to compact or reductive groups. The classification feeds directly into Representation Theory of Semisimple Lie Algebras, where highest-weight theory and the Weyl character formula parametrize all irreducible modules by dominant weights. The combinatorial core — root systems and reflection groups — is exactly the subject of Coxeter Groups and Weyl Groups. On the geometric side, real forms classify Symmetric Spaces (Cartan's other great theorem), and the algebraic-group analogue is Linear Algebraic Groups, where Chevalley reproduced the classification over arbitrary fields. The infinite-dimensional generalization replacing positive-definite Cartan matrices with general ones gives Kac–Moody Algebras.

🔬 Where this appears

⏳ History

- 1873 — Sophus Lie founds the theory of continuous transformation groups, seeking a "Galois theory" for differential equations; infinitesimal generators give the bracket.

- 1884 — Friedrich Engel and Lie systematize the infinitesimal approach; Engel's theorem (every Lie algebra of nilpotent operators is nilpotent) emerges.

- 1888–1890 — Wilhelm Killing publishes four papers introducing rank, Cartan subalgebras, root systems, and discovering all exceptional algebras; riddled with gaps but visionary.

- 1894 — Élie Cartan's doctoral thesis rigorously classifies the simple complex Lie algebras, fixing Killing's errors and establishing the Killing form's role.

- 1905 — Cartan and others develop the theory of real forms and representations.

- 1925–1926 — Hermann Weyl proves complete reducibility (Weyl's theorem) using the "unitarian trick," and derives the character and dimension formulas.

- 1937 — Claude Chevalley constructs integral bases (Chevalley basis), enabling reduction mod

p

and the theory of finite groups of Lie type.

- 1947 — Eugene Dynkin introduces Dynkin diagrams, giving the classification its modern combinatorial form; Coxeter's parallel work links to reflection groups.

- 1955–1968 — Serre relations, Kostant's theorems, and Harish-Chandra's representation theory complete the structural picture; foundations laid for Kac–Moody generalizations (1968).

open-problem

Modular Lie algebras (char $p$ ): The classification of finite-dimensional simple Lie algebras over algebraically closed fields of characteristic $p>3$ was only completed in 2004–2007 (Block, Wilson, Strade, Premet) — they comprise classical, Cartan-type, and (for $p=5$ ) Melikian algebras. For $p=2,3$ the problem remains open, with sporadic exceptional algebras still being discovered.

Restricted and superalgebra extensions: Classification of finite-dimensional simple Lie superalgebras (Kac, 1977) introduces new families ( $\mathfrak{osp}$ , $\mathfrak{pq}$ , exceptional $G(3),F(4),D(2,1;\alpha)$ ); their representation theory (BGG category, atypical weights) is subtler than the classical case and not fully understood.

Infinite dimensions: For Kac–Moody algebras with indefinite (hyperbolic and beyond) Cartan matrices, there is no complete structural understanding — root multiplicities are unknown in general, and even $E_{10}$ and $E_{11}$ (conjectured symmetries of M-theory) resist a closed description.

Cohomology and deformations: Whether all deformations of a given solvable/nilpotent Lie algebra are classified is open in high dimension; the classification of nilpotent Lie algebras is "wild" (parametrized by moduli with continuous families) beyond dimension $7$ .

Lie Algebras — Structure and Classification

Formal Statement

Why It Holds / Derivation

Building Intuition

Worked Examples

Common Misconceptions

Connections

Real-World Applications

Historical Development

Subtleties & Limits