Lie Groups, Lie Algebras, and the Exponential Map

⚡ Key Takeaways

One-line statement: A Lie group is a smooth manifold with compatible group structure; its tangent space at the identity is a Lie algebra, and the exponential map provides a canonical local diffeomorphism translating the linear (algebraic) data into the curved (group) data. Key result in words: The infinitesimal structure (Lie algebra with bracket) determines the group near the identity, and for connected, simply connected groups, globally — the exponential map being the bridge. Where it applies: Differential geometry, representation theory, quantum mechanics, gauge theory, control theory, harmonic analysis, integrable systems. Who/when: Sophus Lie (1870s–1880s), Élie Cartan, Wilhelm Killing, Hermann Weyl, Claude Chevalley (1880s–1950s).

💡 The Core Insight

Linearize the group at the identity. A Lie group is a nonlinear, curved object — but group multiplication lets you transport tangent data everywhere via left translations, so the single tangent space $\mathfrak{g} = T_eG$ controls the whole structure. The failure of group multiplication to commute is captured infinitesimally by the commutator, which becomes the Lie bracket. The exponential map integrates these infinitesimal generators (left-invariant vector fields) into one-parameter subgroups, packaging "infinitesimal symmetry → finite symmetry."

DefinitionLie Group

A Lie group is a smooth manifold $G$ equipped with a group structure such that the maps

1

\quad \mu: G \times G \to G,\ (g,h) \mapsto gh \quad\text{and}\quad \iota: G \to G,\ g \mapsto g^{-1}

are smooth. Equivalently, the single map $G\times G \to G$ , $(g,h)\mapsto gh^{-1}$ , is smooth. Derivation hint: Smoothness of $(g,h)\mapsto gh^{-1}$ implies smoothness of $\mu$ and $\iota$ separately by restriction (fix $g=e$ for $\iota$ ; compose for $\mu$ ). The dimension of $G$ as a manifold is the dimension of the Lie group. Examples: $\mathrm{GL}(n,\mathbb{R})$ (dim $n^2$ ), $\mathrm{SO}(n)$ , $\mathrm{SU}(n)$ , the torus $\mathbb{T}^n$ , the Heisenberg group.

DefinitionLeft-Invariant Vector Fields and the Lie Algebra

For $g \in G$ , let $L_g: G \to G$ , $L_g(h) = gh$ be left translation, a diffeomorphism. A vector field $X$ on $G$ is left-invariant if

2

\quad (dL_g)_h\, X_h = X_{gh} \quad \text{for all } g,h \in G.

Derivation hint: This says $X$ is $L_g$ -related to itself; pushforward by the derivative of $L_g$ preserves $X$ . A left-invariant vector field is determined by its value $X_e \in T_eG$ , giving a vector-space isomorphism between left-invariant fields and $T_eG$ . The Lie algebra of $G$ is

3

\quad \mathfrak{g} := T_eG, \qquad [X,Y]_e := [\tilde X, \tilde Y]_e,

where $\tilde X,\tilde Y$ are the left-invariant extensions and $[\tilde X,\tilde Y]$ is the vector-field commutator (Lie bracket). Derivation hint: The commutator of two left-invariant fields is again left-invariant, so it has a well-defined value at $e$ ; transfer that bracket to $T_eG$ .

DefinitionLie Algebra (abstract)

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

4

\quad [X,X]=0 \quad(\text{antisymmetry}),\qquad [X,[Y,Z]] + [Y,[Z,X]] + [Z,[X,Y]] = 0 \quad(\text{Jacobi}).

Derivation hint: Antisymmetry $[X,Y]=-[Y,X]$ follows from $[X,X]=0$ by expanding $[X+Y,X+Y]=0$ . The Jacobi identity for vector fields follows from associativity of composition of derivations.

DefinitionThe Exponential Map

For $X \in \mathfrak{g}$ , let $\gamma_X: \mathbb{R} \to G$ be the unique one-parameter subgroup with $\gamma_X'(0) = X$ , i.e. the integral curve through $e$ of the left-invariant field $\tilde X$ . Define

5

\quad \exp: \mathfrak{g} \to G, \qquad \exp(X) := \gamma_X(1).

Derivation hint: Left-invariant fields are complete (their flow is defined for all time by the group-homomorphism property $\gamma_X(s+t)=\gamma_X(s)\gamma_X(t)$ ), guaranteeing $\gamma_X(1)$ exists. It satisfies $\gamma_X(t) = \exp(tX)$ and the scaling identity

6

\quad \exp(sX)\exp(tX) = \exp((s+t)X), \qquad \frac{d}{dt}\Big|_{t=0}\exp(tX) = X.

Derivation hint: $\gamma_X(t)$ and $\gamma_{tX}(1)$ solve the same ODE with the same initial velocity after reparametrization $s\mapsto ts$ .

TheoremMatrix Exponential Formula

For a matrix Lie group $G \subseteq \mathrm{GL}(n,\mathbb{R})$ with $\mathfrak{g} \subseteq \mathfrak{gl}(n,\mathbb{R})$ , the abstract exponential coincides with the matrix exponential:

7

\quad \exp(X) = e^X = \sum_{k=0}^{\infty} \frac{X^k}{k!} = I + X + \frac{X^2}{2!} + \cdots

Derivation hint: $\gamma(t)=e^{tX}$ solves $\gamma'=\gamma X$ , $\gamma(0)=I$ — exactly the ODE for the one-parameter subgroup generated by the left-invariant field with value $X$ ; uniqueness of ODE solutions gives equality.

TheoremLocal Diffeomorphism / Lie's Fundamental Theorems

1. $d(\exp)_0 = \mathrm{id}_{\mathfrak{g}}$ , so $\exp$ is a diffeomorphism from a neighborhood of $0 \in \mathfrak{g}$ onto a neighborhood of $e \in G$ . 2. Lie's Third Theorem: Every finite-dimensional real Lie algebra is the Lie algebra of some (simply connected) Lie group. 3. The functor $G \mapsto \mathfrak{g}$ is an equivalence of categories between simply connected Lie groups and finite-dimensional Lie algebras.

8

\quad \exp(X)\exp(Y) = \exp\!\Big(X + Y + \tfrac{1}{2}[X,Y] + \tfrac{1}{12}\big([X,[X,Y]] - [Y,[X,Y]]\big) + \cdots\Big)

Derivation hint: This is the Baker–Campbell–Hausdorff formula; derive by differentiating $\exp(tX)\exp(tY)$ and matching the Dynkin series; it shows the group product near $e$ is determined entirely by iterated brackets.

ProofConstruction of the Bracket and Exponential

Step 1 (Left-invariant fields ↔ $T_eG$ ). Given $v \in T_eG$ , define $\tilde v_g := (dL_g)_e v$ . By the chain rule and $L_{gh}=L_g\circ L_h$ , $\tilde v$ is left-invariant, smooth, and $\tilde v_e = v$ . Conversely any left-invariant field equals $\widetilde{X_e}$ . This gives the linear isomorphism $\mathfrak{g}\cong T_eG$ .

Step 2 (Bracket is well-defined). For left-invariant $\tilde X,\tilde Y$ , the field commutator

9

\quad [\tilde X,\tilde Y] f = \tilde X(\tilde Y f) - \tilde Y(\tilde X f)

is a derivation, hence a vector field. Since $L_g$ -relatedness is preserved by the bracket (naturality: $\phi_*[A,B]=[\phi_*A,\phi_*B]$ for diffeomorphisms), $[\tilde X,\tilde Y]$ is left-invariant. Set $[X,Y]:=[\tilde X,\tilde Y]_e$ . Derivation hint: Use that $\tilde X$ is $L_g$ -related to itself and the bracket of related fields is related.

Step 3 (One-parameter subgroups exist and are complete). The flow $\Phi_t$ of $\tilde X$ satisfies $\Phi_t\circ L_g = L_g \circ \Phi_t$ by left-invariance, so $\gamma(t)=\Phi_t(e)$ obeys $\gamma(s+t)=\gamma(s)\gamma(t)$ . This functional equation lets us extend any local solution to all $t\in\mathbb{R}$ : completeness.

Step 4 ( $d\exp_0 = \mathrm{id}$ ). Compute $\frac{d}{dt}\big|_{t=0}\exp(tX) = \gamma_X'(0) = X$ . So the derivative of $\exp$ at $0$ along $X$ is $X$ itself; by the Inverse Function Theorem $\exp$ is a local diffeomorphism. $\blacksquare$

∎

ProofBaker–Campbell–Hausdorff (sketch via Dynkin)

Step 1. Define $Z(t)$ by $e^{tX}e^{tY} = e^{Z(t)}$ for small $t$ ; $Z(t)$ is analytic with $Z(0)=0$ .

Step 2. Differentiate using the derivative-of-exp formula

10

\quad \frac{d}{dt}e^{Z(t)} = e^{Z(t)}\cdot \frac{1 - e^{-\mathrm{ad}_{Z}}}{\mathrm{ad}_{Z}}\,\dot Z(t), \qquad \mathrm{ad}_X(Y):=[X,Y].

Derivation hint: Use the standard identity $\frac{d}{dt}e^{Z}=e^{Z}\,\frac{1-e^{-\mathrm{ad}_Z}}{\mathrm{ad}_Z}\dot Z$ , proved by introducing $F(s,t)=e^{-sZ}\partial_t e^{sZ}$ and solving an ODE in $s$ .

Step 3. Match against $\frac{d}{dt}(e^{tX}e^{tY})=(X+ \mathrm{Ad}_{e^{tX}}Y)e^{Z}$ and invert the operator series $\frac{\mathrm{ad}_Z}{1-e^{-\mathrm{ad}_Z}}$ , yielding $Z$ as a series in nested brackets — the BCH formula [8]. The key consequence: the entire local group law is encoded by $[\cdot,\cdot]$ . $\blacksquare$

∎

mathematician · Sophus Lie | 1842–1899

Norwegian mathematician who, seeking a "Galois theory for differential equations," introduced continuous transformation groups in the 1870s. Working with Friedrich Engel, he developed the three "fundamental theorems" relating local groups to their infinitesimal generators. Lie thought locally and analytically; the global, manifold-theoretic formulation came later (Weyl, Cartan, Chevalley). His insight — that symmetries of differential equations form continuous groups whose infinitesimal version is linear — reshaped geometry, physics, and analysis. The bracket, algebras, and the exponential bridge all trace to his program.

🧠 Intuition

Think of a Lie group as a "curved space of symmetries." Standing at the identity $e$ , the tangent space $\mathfrak{g}$ is the set of infinitesimal motions — velocities of curves through $e$ . Left translation lets you carry any such infinitesimal motion to every other point, weaving a left-invariant vector field (a globally consistent "flow direction"). Following such a flow for unit time is exactly what $\exp$ does: it converts an infinitesimal generator into a finite group element. The bracket $[X,Y]$ measures the failure of two flows to commute: flow along $X$ , then $Y$ , then back along $X$ , then back along $Y$ — you don't return to start, and the second-order discrepancy is $[X,Y]$ . This is why abelian groups have zero bracket and $\exp$ is a homomorphism there.

HW

Hermann Weyl

1885–1955

Weyl unified Lie's local theory with topology, proving that compact semisimple Lie groups have finite fundamental group (the "Weyl unitary trick") and connecting representation theory to global geometry. In physics, his work made Lie groups the language of gauge symmetry: the conserved currents of Noether's theorem are valued in $\mathfrak{g}$ , and gauge fields are $\mathfrak{g}$ -valued connection 1-forms whose exponentials are the finite gauge transformations.

Figure

The exponential map sends a tangent vector at the identity to a one-parameter subgroup curve on the group manifold

ExampleThe Rotation Group SO(2) and Its Exponential

Given: $G = \mathrm{SO}(2) = \{R(\theta)\}$ , the $2\times 2$ rotation matrices. Its Lie algebra $\mathfrak{so}(2)$ consists of antisymmetric $2\times 2$ matrices, spanned by

11

\quad J = \begin{pmatrix} 0 & -1 \\ 1 & 0 \end{pmatrix}, \qquad J^2 = -I.

Derivation hint: Differentiate $R(\theta)R(\theta)^T=I$ at $\theta=0$ to get $X + X^T=0$ .

Steps: Compute $\exp(\theta J)$ via the matrix series, splitting even/odd powers using $J^2=-I$ :

12

\quad e^{\theta J} = \sum_{k} \frac{(\theta J)^k}{k!} = \Big(\sum \frac{(-1)^m\theta^{2m}}{(2m)!}\Big)I + \Big(\sum\frac{(-1)^m\theta^{2m+1}}{(2m+1)!}\Big)J.

Derivation hint: $J^{2m}=(-1)^m I$ , $J^{2m+1}=(-1)^m J$ ; the two series are $\cos\theta$ and $\sin\theta$ . This yields

13

\quad \exp(\theta J) = \cos\theta\, I + \sin\theta\, J = \begin{pmatrix} \cos\theta & -\sin\theta \\ \sin\theta & \cos\theta\end{pmatrix} = R(\theta).

Derivation hint: Recombine the cosine and sine series with the matrices $I,J$ .

Boxed answer: $\boxed{\exp(\theta J) = R(\theta)\ \text{(rotation by angle } \theta \text{ radians)}}$ . The bracket is trivial ( $\dim 1$ abelian), so here $\exp$ is a surjective homomorphism, with kernel $2\pi\mathbb{Z}$ .

ExampleNon-commuting Generators in SU(2): BCH at Second Order

Given: Pauli matrices yield a basis of $\mathfrak{su}(2)$ . Set $X = \tfrac{i}{2}\sigma_x$ , $Y=\tfrac{i}{2}\sigma_y$ , with

14

\quad [\,\tfrac{i}{2}\sigma_x,\ \tfrac{i}{2}\sigma_y\,] = i\,\tfrac{i}{2}\sigma_z \cdot(\text{from }[\sigma_x,\sigma_y]=2i\sigma_z) = -\tfrac{1}{2}\sigma_z \cdot i = \tfrac{i}{2}\sigma_z = Z.

Derivation hint: Use $[\sigma_a,\sigma_b]=2i\epsilon_{abc}\sigma_c$ ; with the $\tfrac i2$ normalization the structure constants become $[X,Y]=Z$ cyclically.

Steps: Take small parameter $t=0.1$ and compute the leading BCH correction for $e^{tX}e^{tY}$ . By [8],

15

\quad Z(t) = tX + tY + \tfrac{1}{2}[tX,tY] + O(t^3) = t(X+Y) + \tfrac{t^2}{2}Z + O(t^3).

Derivation hint: Substitute $tX,tY$ into BCH; the quadratic term is $\tfrac12[tX,tY]=\tfrac{t^2}{2}[X,Y]=\tfrac{t^2}{2}Z$ . Numerically: $\tfrac{t^2}{2} = \tfrac{0.01}{2} = 0.005$ . So the product is not $e^{t(X+Y)}$ ; there is a $0.005\,Z$ generator's worth of extra "twist" about the $z$ -axis.

Boxed answer: $\boxed{e^{0.1X}e^{0.1Y} = \exp\!\big(0.1(X+Y) + 0.005\,Z + O(10^{-3})\big)}$ — concretely, composing a $0.1$ rad rotation about $x$ then about $y$ induces a $\approx 0.005$ rad spurious rotation about $z$ (a tangible non-commutativity used in geometric phase / gyroscope analysis).

⚠️ Common mistake

$\exp$ is neither surjective nor a homomorphism in general.

It is a homomorphism $\exp(X+Y)=\exp(X)\exp(Y)$ only when $[X,Y]=0$ (e.g. abelian groups, or $X,Y$ in a common abelian subalgebra). The BCH formula [8] makes the obstruction explicit.
It is not surjective for many connected groups. In $\mathrm{SL}(2,\mathbb{R})$ , the element $\begin{pmatrix}-1 & 1\\ 0 & -1\end{pmatrix}$ is not in the image of $\exp$ .
For compact connected groups $\exp$ is surjective; for connected nilpotent groups it is a diffeomorphism. Don't extrapolate from these to the general case.

Counterexample

A non-surjective exponential on $\mathrm{SL}(2,\mathbb{R})$ . Any $e^X$ with $X\in\mathfrak{sl}(2,\mathbb{R})$ (trace zero) has eigenvalues $e^{\pm\lambda}$ (if $X$ has eigenvalues $\pm\lambda$ ). For real $X$ , the eigenvalue pair is either $\{e^\lambda,e^{-\lambda}\}$ with $\lambda$ real (positive product, possibly repeated $1$ ) or $\{e^{i\theta},e^{-i\theta}\}$ . A matrix like $A=\begin{pmatrix}-2 & 0\\ 0 & -1/2\end{pmatrix}$ has $\det=1$ , $\mathrm{tr}=-5/2<-2$ , distinct negative eigenvalues — impossible to realize as $e^X$ with $X$ real trace-free (that would force a complex-conjugate or positive-product spectrum). Hence $A\notin \exp(\mathfrak{sl}(2,\mathbb{R}))$ , even though $A$ lies in the connected component of the identity (since $\mathrm{SL}(2,\mathbb{R})$ is connected). Surjectivity fails.

connection

The Lie group / Lie algebra correspondence sits at the crossroads of geometry and algebra. The bracket is the infinitesimal version of the adjoint action $\mathrm{Ad}_g = d(C_g)_e$ where $C_g(h)=ghg^{-1}$ ; differentiating gives $\mathrm{ad}_X = [X,\cdot]$ , linking to Representation Theory of Lie Algebras. Left-invariant fields and their flows tie directly to Vector Fields and Flows on Manifolds and the Frobenius Integrability Theorem (subalgebras ↔ subgroups). The exponential map is a special case of the Riemannian Exponential Map when $G$ carries a bi-invariant metric, with one-parameter subgroups as geodesics. Structure theory (root systems, Killing form) connects to Semisimple Lie Algebras and Root Systems, and the global classification to Covering Spaces and Fundamental Groups via Lie's third theorem.

🔬 Where this appears

⏳ History

- 1873–1874: Sophus Lie begins his theory of continuous transformation groups, seeking a Galois theory for differential equations.

- 1888–1893: Lie and Friedrich Engel publish Theorie der Transformationsgruppen (3 volumes), establishing the three fundamental theorems linking local groups and infinitesimal generators.

- 1888–1890: Wilhelm Killing classifies simple complex Lie algebras (with errors), introducing root systems and the "Killing form."

- 1894: Élie Cartan's thesis corrects and completes Killing's classification, rigorously establishing the simple Lie algebras

A_n,B_n,C_n,D_n

and the exceptionals

G_2,F_4,E_6,E_7,E_8

.

- 1925–1926: Hermann Weyl proves complete reducibility for semisimple groups (unitary trick) and the character/Weyl integration formulas, globalizing the theory.

- 1929: John von Neumann and others clarify the relation between Lie groups and topological groups (closed subgroup theorem era).

- 1930s–1940s: Claude Chevalley reframes the theory in modern manifold/global terms; his Theory of Lie Groups (1946) is foundational.

- 1950s: The Hilbert's fifth problem resolution (Gleason, Montgomery–Zippin, Yamabe, 1952) shows locally Euclidean topological groups are Lie groups.

- 1960s onward: Harish-Chandra develops harmonic analysis on semisimple groups; Lie theory pervades physics (gauge theory), number theory (Langlands program), and geometry.

open-problem

Where the clean correspondence frays:

Infinite dimensions: For infinite-dimensional (Fréchet/Banach) Lie groups — diffeomorphism groups, loop groups, gauge groups — Lie's third theorem can fail; $\exp$ may not be a local diffeomorphism (the diffeomorphism group of the circle is a notorious example where $\exp$ is not locally surjective onto any neighborhood). A complete theory of "Lie's third problem in infinite dimensions" (integrating a given infinite-dimensional Lie algebra to a group) remains delicate.
Lie algebroids and groupoids: Not every Lie algebroid integrates to a Lie groupoid — the Crainic–Fernandes obstructions (2003) quantify this, a genuinely modern resolution of an "infinite-dimensional Lie's third theorem."
Surjectivity & logarithms: Determining the precise image of $\exp$ for a given non-compact group, and choosing canonical logarithms (branch issues), is subtle; the exponential's failure to be onto complicates global parametrizations and certain proofs.
Convergence of BCH: The BCH series converges only for sufficiently small $X,Y$ in general; global control laws require care.

Lie Groups, Lie Algebras, and the Exponential Map

Formal Statement

Why It Holds / Derivation

Building Intuition

Worked Examples

Common Misconceptions

Connections

Real-World Applications

Historical Development

Subtleties & Limits

See Also