Left-Invariant Vector Fields
When the manifold carries a
Lie group
structure, vector fields acquire a restricted subclass: those invariant under left translation. The
space of all such fields is finite-dimensional, closed under the Lie bracket, and entirely
determined by its values at the identity. It is called the Lie algebra of the group, and it
captures the infinitesimal symmetries that govern how the group multiplies near the identity. The
present page constructs this object in general and identifies it as a vector space with the tangent
space at the identity; the explicit identification of its bracket for matrix groups is taken up
separately.
Translation and the invariance condition
Throughout this section, \(G\) denotes a Lie group with identity element \(e\). For each \(g \in G\),
the
left translation
is the smooth map \(L_g : G \to G\) defined by \(L_g(h) = gh\). It is a diffeomorphism with smooth
inverse \(L_{g^{-1}}\), and its
differential
\(d(L_g)_{g'} : T_{g'}G \to T_{gg'}G\) is therefore a linear isomorphism at every point \(g'\). Since
\(L_g\) is a diffeomorphism, the
pushforward
\((L_g)_* X\) of any \(X \in \mathfrak{X}(G)\) is a well-defined smooth vector field on \(G\).
Definition: Left-Invariant Vector Field
A vector field
\(X \in \mathfrak{X}(G)\)
on a Lie group \(G\) is left-invariant if, for every \(g \in G\), the field
\(X\) is \(L_g\)-related to itself; equivalently,
\[
d(L_g)_{g'}\bigl( X_{g'} \bigr) = X_{gg'}
\qquad \text{for all } g, g' \in G .
\]
Since \(L_g\) is a diffeomorphism, this is equivalent to the pushforward identity
\[
(L_g)_* X = X \qquad \text{for every } g \in G .
\]
The two forms of the invariance condition record the same data in two languages: the pointwise form
expresses left-invariance as an \(F\)-relatedness condition with \(F = L_g\), while the pushforward
form expresses it at the level of vector fields globally. We move freely between them, using
whichever is more convenient in a given argument.
Linear subspace structure
The set of left-invariant smooth vector fields on \(G\) is a linear subspace of \(\mathfrak{X}(G)\)
over \(\mathbb{R}\): if \(X, Y\) are left-invariant and \(a, b \in \mathbb{R}\), then for every
\(g, g' \in G\),
\[
d(L_g)_{g'}\bigl( (aX + bY)_{g'} \bigr)
= a\, d(L_g)_{g'}(X_{g'}) + b\, d(L_g)_{g'}(Y_{g'})
= a X_{gg'} + b Y_{gg'}
= (aX + bY)_{gg'} ,
\]
where the first equality uses the
linearity of the differential
and the second uses left-invariance of \(X\) and \(Y\). Thus \(aX + bY\) is left-invariant, and the
subspace property follows.
More is true: the set of left-invariant smooth vector fields is closed under the Lie bracket of
vector fields. This is the first substantive fact about left-invariant fields, and the one on which
everything that follows on this page depends.
Closure under the Lie bracket
Proposition (Bracket Closure of Left-Invariant Vector Fields)
Let \(G\) be a Lie group, and let \(X, Y \in \mathfrak{X}(G)\) be left-invariant. Then the Lie
bracket \([X, Y]\) is also left-invariant.
Proof:
Fix \(g \in G\). Since \(L_g\) is a diffeomorphism, the
pushforward commutes with the Lie bracket:
\[
(L_g)_* [X, Y] = [(L_g)_* X, (L_g)_* Y] .
\]
By left-invariance, \((L_g)_* X = X\) and \((L_g)_* Y = Y\), so the right-hand side equals
\([X, Y]\). Therefore
\[
(L_g)_* [X, Y] = [X, Y]
\qquad \text{for every } g \in G ,
\]
which is the pushforward form of left-invariance applied to \([X, Y]\). Hence \([X, Y]\) is
left-invariant.
The argument occupies one line of computation, but it relies on two non-trivial ingredients: the
naturality of the Lie bracket under smooth maps, established in the previous page, and the
diffeomorphism property of left translations, which is built into the definition of a Lie group.
Together these make the bracket a well-defined operation on the set of left-invariant smooth vector
fields, turning a linear subspace of \(\mathfrak{X}(G)\) into an object carrying its own algebraic
structure. The classification of that algebraic structure occupies the rest of this page.
The Lie Algebra Structure
Before continuing with left-invariant vector fields, we record the abstract algebraic structure that
they will instantiate: a real vector space with a bilinear, antisymmetric bracket satisfying the
Jacobi identity. This abstraction was developed in the linear-algebra track in the matrix setting,
and we use the existing
abstract Lie algebra
definition there as the governing notion on this page as well. The point of restating the
associated vocabulary — subalgebra, homomorphism, isomorphism — is to fix the language
used in the remainder of this section and the next.
Abstract Lie algebras
A Lie algebra over a field \(\mathbb{F}\) is a vector space \(\mathfrak{g}\) over \(\mathbb{F}\)
equipped with a bilinear bracket \([\,\cdot\,, \cdot\,] : \mathfrak{g} \times \mathfrak{g} \to
\mathfrak{g}\) that is antisymmetric and satisfies the Jacobi identity. The three properties —
bilinearity, antisymmetry, and Jacobi — match those established for the Lie bracket of vector
fields on the previous page, so the abstract definition simply axiomatizes the structure already in
hand.
The Jacobi identity plays the role of a substitute for associativity, which does not hold for
brackets in general; this is the structural reason the bracket is not the multiplication of an
associative algebra in the ordinary sense. Although Lie algebras over the complex numbers and other
fields are useful in many contexts, every Lie algebra on this page is assumed to be a real Lie
algebra, in keeping with the convention adopted throughout the smooth-manifold track. The complex
matrix algebra \(\mathfrak{gl}(n, \mathbb{C})\) appears below, but it is treated as a real Lie
algebra of real dimension \(2n^2\), not as a complex Lie algebra.
Subalgebras, homomorphisms, and isomorphisms
A Lie subalgebra of a Lie algebra \(\mathfrak{g}\) is a linear subspace
\(\mathfrak{h} \subseteq \mathfrak{g}\) closed under the bracket; with the restriction of the
bracket, \(\mathfrak{h}\) is itself a Lie algebra. A
Lie algebra homomorphism
is a linear map \(A : \mathfrak{g} \to \mathfrak{h}\) between Lie algebras that preserves brackets,
\(A[X, Y] = [AX, AY]\); when invertible it is a Lie algebra isomorphism, and the
two Lie algebras are then said to be isomorphic. The kernel and image of a Lie
algebra homomorphism are Lie subalgebras of the source and target, respectively, by a direct
check on the defining properties.
Examples
The next collection of examples records the instances of the Lie algebra structure used on this
page and elsewhere in the series. The first two arise on a smooth manifold, the next three are
matrix examples that connect to the linear-algebra track, and the final one is the trivial bracket
on an arbitrary vector space.
Example (Lie Algebras):
(a) Vector fields on a smooth manifold. For any smooth manifold \(M\), the
\(\mathbb{R}\)-vector space
\(\mathfrak{X}(M)\)
is a Lie algebra under the Lie bracket of vector fields, by the
bilinearity, antisymmetry, and Jacobi identity
established in the previous page.
(b) Left-invariant vector fields on a Lie group. If \(G\) is a Lie group, the
set of all smooth left-invariant vector fields on \(G\) is a Lie subalgebra of
\(\mathfrak{X}(G)\), by the
bracket closure for left-invariant fields
of the previous section. This is the example the rest of the page focuses on; we shall see in
the next section that it is finite-dimensional with dimension equal to \(\dim G\).
(c) Real \(n \times n\) matrices with the commutator. The \(n^2\)-dimensional
vector space \(M(n, \mathbb{R})\) of real \(n \times n\) matrices is a Lie algebra under the
matrix commutator
\([A, B] = AB - BA\). Bilinearity and antisymmetry are immediate from the formula, and the
Jacobi identity follows from a direct expansion using associativity of matrix multiplication;
the full verification is recorded in the linear-algebra track as
properties of the matrix Lie bracket.
When regarded as a Lie algebra in this way, \(M(n, \mathbb{R})\) is denoted
\(\mathfrak{gl}(n, \mathbb{R})\).
(d) Complex \(n \times n\) matrices with the commutator. The same construction
applied to \(M(n, \mathbb{C})\), with the matrix commutator \([A, B] = AB - BA\), yields a Lie
algebra of real dimension \(2n^2\), which we denote \(\mathfrak{gl}(n, \mathbb{C})\). Although
\(M(n, \mathbb{C})\) is naturally a complex vector space, we treat it here as a real vector
space, in line with the convention that all Lie algebras on this page are real.
(e) Endomorphisms of a vector space. If \(V\) is a finite-dimensional real
vector space, the space of all linear maps \(V \to V\) is a Lie algebra under the commutator
\([A, B] = A \circ B - B \circ A\), which we denote \(\mathfrak{gl}(V)\). Choosing a basis for
\(V\) and identifying linear maps with matrices yields a Lie algebra isomorphism
\(\mathfrak{gl}(V) \cong \mathfrak{gl}(n, \mathbb{R})\), where \(n = \dim V\); under the
standard identification of \(n \times n\) real matrices with linear maps from \(\mathbb{R}^n\)
to itself, \(\mathfrak{gl}(\mathbb{R}^n)\) is the same Lie algebra as
\(\mathfrak{gl}(n, \mathbb{R})\).
(f) Abelian Lie algebras. Any vector space \(V\) becomes a Lie algebra if every
bracket is set equal to zero: bilinearity is trivial, antisymmetry and the Jacobi identity hold
because every term is zero. A Lie algebra in which \([X, Y] = 0\) for all \(X, Y\) is called
abelian. The terminology reflects the fact that in the matrix examples above,
the commutator \([A, B] = AB - BA\) vanishes precisely when the underlying matrix product is
commutative; the connection between abelian Lie algebras and abelian Lie groups is taken up in
a later chapter of the development of the theory.
Examples (c)–(e) recall a development carried out in the linear-algebra track without any
smooth-manifold input: \(\mathfrak{gl}(n, \mathbb{R})\), \(\mathfrak{gl}(n, \mathbb{C})\), and the
classical matrix Lie algebras
\(\mathfrak{so}(n)\),
\(\mathfrak{su}(n)\),
\(\mathfrak{sl}(n, \mathbb{R})\)
were constructed there as specific subspaces of \(M(n, \mathbb{R})\) or \(M(n, \mathbb{C})\) cut out
by linear or algebraic conditions, with the bracket given directly by the matrix commutator. Example
(b) introduces a parallel construction of a Lie algebra from a Lie group through left-invariant
vector fields, with a bracket whose definition makes no reference to matrix multiplication. The
development of the present and following pages reverses the relationship between the two
constructions: the matrix commutator structure of \(\mathfrak{gl}(n, \mathbb{R})\) is recovered from
the left-invariant-vector-field construction applied to \(GL(n, \mathbb{R})\), as a theorem rather
than a definition, and the classical matrix Lie algebra \(\mathfrak{so}(n)\) is identified with
\(\mathrm{Lie}(O(n))\) by the same general mechanism, which applies equally to \(\mathfrak{su}(n)\),
\(\mathfrak{sl}(n, \mathbb{R})\), and every other Lie subalgebra arising from a matrix Lie subgroup.
The two viewpoints describe the same Lie algebras from different starting points: the linear-algebra
construction is concrete and admits direct calculation but applies only to groups already presented
as matrix subgroups of \(GL(n, \mathbb{C})\), while the smooth-manifold construction applies to any
Lie group but requires the technical apparatus of vector fields and the Lie bracket of vector fields
on a smooth manifold to produce a calculable bracket.
The Lie Algebra of a Lie Group
The set of smooth left-invariant vector fields on a Lie group \(G\), with the Lie bracket of vector
fields restricted to it, is a Lie subalgebra of \(\mathfrak{X}(G)\) by the bracket closure
proposition of the first section. A key structural fact about this Lie algebra is that it is
finite-dimensional: every left-invariant vector field is uniquely determined by its value at the
identity, and conversely every tangent vector at the identity extends to a left-invariant vector
field. The evaluation at the identity is therefore a vector space isomorphism, identifying the Lie
algebra of \(G\) with the tangent space \(T_eG\) and so giving it the same finite dimension as the
underlying group.
The Lie algebra of a Lie group
Definition: The Lie Algebra of a Lie Group
Let \(G\) be a Lie group. The Lie algebra of \(G\), denoted \(\mathrm{Lie}(G)\),
is the set of all smooth left-invariant vector fields on \(G\), equipped with the Lie bracket
inherited from \(\mathfrak{X}(G)\):
\[
\mathrm{Lie}(G)
= \{ X \in \mathfrak{X}(G) :
(L_g)_* X = X \text{ for every } g \in G \} ,
\]
with bracket \([X, Y]\) defined by the Lie bracket of vector fields. By the bracket closure
proposition, this is a Lie subalgebra of \(\mathfrak{X}(G)\).
The smoothness condition built into the definition turns out to be redundant: a corollary of the
next theorem shows that every left-invariant vector field on a Lie group is automatically smooth,
so the same Lie algebra is obtained by relaxing the definition to allow rough left-invariant fields.
We record this point after the main result.
Evaluation at the identity is an isomorphism
Theorem (Evaluation Isomorphism)
Let \(G\) be a Lie group with identity \(e\). The evaluation map
\[
\varepsilon : \mathrm{Lie}(G) \to T_eG ,
\qquad \varepsilon(X) = X_e ,
\]
is a vector space isomorphism. Consequently \(\mathrm{Lie}(G)\) is finite-dimensional with
dimension equal to \(\dim G\).
Proof:
Linearity. For \(X, Y \in \mathrm{Lie}(G)\) and \(a, b \in \mathbb{R}\), the pointwise
definition of \(aX + bY\) gives \((aX + bY)_e = aX_e + bY_e\), so
\(\varepsilon(aX + bY) = a\varepsilon(X) + b\varepsilon(Y)\).
Injectivity. Suppose \(X \in \mathrm{Lie}(G)\) satisfies \(\varepsilon(X) = X_e = 0\).
By left-invariance,
\[
X_g = d(L_g)_e( X_e ) = d(L_g)_e( 0 ) = 0
\qquad \text{for every } g \in G ,
\]
so \(X\) is the zero vector field. Hence \(\ker \varepsilon = \{0\}\) and \(\varepsilon\) is
injective.
Surjectivity. Given \(v \in T_eG\), we construct a smooth left-invariant vector field
\(v^L\) with \(v^L|_e = v\), forced by the left-invariance condition to be the only such field.
Define a rough vector field \(v^L\) on \(G\) by
\[
v^L|_g := d(L_g)_e( v )
\qquad \text{for every } g \in G .
\tag{1}
\]
Since \(d(L_g)_e\) sends \(T_eG\) into \(T_{L_g(e)}G = T_gG\), the assignment \(g \mapsto v^L|_g\)
produces a tangent vector to \(G\) at every \(g\), so \(v^L\) is a rough vector field on \(G\).
Three points remain: that \(v^L\) is smooth, that \(v^L\) is left-invariant, and that
\(\varepsilon(v^L) = v\).
Smoothness of \(v^L\). By the
smoothness criterion via the action on functions,
it suffices to show that \(v^L f \in C^\infty(G)\) for every \(f \in C^\infty(G)\). Choose a
smooth curve \(\gamma : (-\delta, \delta) \to G\) with \(\gamma(0) = e\) and
velocity
\(\gamma'(0) = v\). Using the defining property of the differential, for each \(g \in G\),
\[
\begin{align*}
(v^L f)(g)
&= d(L_g)_e(v)\, f
= v( f \circ L_g ) \\
&= \gamma'(0)( f \circ L_g )
= \left. \frac{d}{dt} \right|_{t = 0}
\bigl( f \circ L_g \circ \gamma \bigr)(t) .
\end{align*}
\]
Define \(\varphi : (-\delta, \delta) \times G \to \mathbb{R}\) by
\[
\varphi(t, g) = f\bigl( L_g \gamma(t) \bigr) = f\bigl( g \gamma(t) \bigr) ,
\]
so that \((v^L f)(g) = \partial \varphi/\partial t(0, g)\). The map \(\varphi\) is the
composition of the smooth group multiplication \(\mu : G \times G \to G\), the smooth curve
\(\gamma\), and the smooth function \(f\), and is therefore smooth on
\((-\delta, \delta) \times G\). Its partial derivative \(\partial \varphi/\partial t\) is smooth
on the same domain, and the evaluation \((v^L f)(g) = \partial \varphi/\partial t(0, g)\) is a
smooth function of \(g\) on \(G\). Hence \(v^L f \in C^\infty(G)\), and \(v^L\) is a smooth
vector field on \(G\).
Left-invariance of \(v^L\). Fix \(g, h \in G\). The group axiom
\(L_h(L_g(k)) = h(gk) = (hg)k = L_{hg}(k)\), valid for every \(k \in G\), gives the identity
\(L_h \circ L_g = L_{hg}\) of smooth maps \(G \to G\). Applying the
chain rule for differentials
and the definition (1),
\[
d(L_h)_g\bigl( v^L|_g \bigr)
= d(L_h)_g \circ d(L_g)_e( v )
= d( L_h \circ L_g )_e( v )
= d(L_{hg})_e( v )
= v^L|_{hg} ,
\]
which is the pointwise form of left-invariance for \(v^L\). Hence
\(v^L \in \mathrm{Lie}(G)\).
Recovery. Since left translation by the identity is the identity map
\(L_e = \mathrm{Id}_G\), its differential at \(e\) is the identity of \(T_eG\), and so
\[
\varepsilon( v^L ) = v^L|_e = d(L_e)_e( v ) = v .
\]
Thus every \(v \in T_eG\) lies in the image of \(\varepsilon\), and \(\varepsilon\) is
surjective.
Linearity, injectivity, and surjectivity together make \(\varepsilon\) a vector space
isomorphism, and the dimension assertion follows from the equality of dimensions of isomorphic
finite-dimensional vector spaces; \(T_eG\) is finite-dimensional with dimension \(\dim G\) by
the definition of a smooth manifold.
The construction \(v \mapsto v^L\), inverse to evaluation at the identity, is the mechanism by which
a tangent vector at \(e\) is propagated across the entire group by left translation. We continue to
use the notation \(v^L\) for the smooth left-invariant vector field defined by (1) for the rest
of this page.
Smoothness is automatic
Corollary (Left-Invariant Rough Vector Fields Are Smooth)
Every left-invariant
rough vector field
on a Lie group is smooth.
Proof:
Let \(X\) be a left-invariant rough vector field on a Lie group \(G\), and set \(v := X_e \in
T_eG\). The left-invariance of \(X\) gives \(X_g = d(L_g)_e(X_e) = d(L_g)_e(v) = v^L|_g\) for
every \(g \in G\), so \(X = v^L\) as rough vector fields on \(G\). Since \(v^L\) is smooth by
the surjectivity construction in the proof of the evaluation isomorphism, \(X\) is smooth.
Looking back at the proof of the evaluation isomorphism, the smoothness step never used smoothness
of \(v^L\) as an input; it derived smoothness from left-invariance alone, given a single tangent
vector \(v \in T_eG\) at the identity. The corollary therefore costs nothing extra, and the
smoothness condition in the definition of \(\mathrm{Lie}(G)\) is not a substantive constraint.
Parallelizability of Lie groups
The evaluation isomorphism makes \(\mathrm{Lie}(G)\) a finite-dimensional Lie algebra of dimension
\(\dim G\), and the inverse construction \(v \mapsto v^L\) supplies an explicit recipe for
propagating a basis of \(T_eG\) to a globally defined collection of vector fields. That collection
is automatically a smooth global frame, with the consequence that every Lie group is parallelizable
— a property held by only a small number of smooth manifolds in general.
Definition: Left-Invariant Frame
Let \(G\) be a Lie group. A
local or global frame
on \(G\) consisting of left-invariant vector fields is called a left-invariant
frame.
Corollary (Lie Groups Are Parallelizable)
Every Lie group admits a left-invariant smooth global frame, and is therefore
parallelizable.
Proof:
Let \(G\) be a Lie group of dimension \(n\), and choose a basis \((E_1, \dots, E_n)\) of
\(\mathrm{Lie}(G)\); such a basis exists because \(\dim \mathrm{Lie}(G) = n\) by the evaluation
isomorphism. Each \(E_i\) is a smooth vector field on \(G\), so the ordered tuple
\((E_1, \dots, E_n)\) is a smooth tuple of vector fields. At every \(g \in G\), the values
\((E_1)_g, \dots, (E_n)_g\) span \(T_gG\) and are linearly independent: applying the linear
isomorphism \(d(L_g)_e : T_eG \to T_gG\) to the basis \((E_1)_e, \dots, (E_n)_e\) of \(T_eG\)
yields the values \((E_1)_g, \dots, (E_n)_g\) by left-invariance, and an isomorphism carries a
basis to a basis. Hence \((E_1, \dots, E_n)\) is a left-invariant smooth global frame for \(G\),
and \(G\) is parallelizable.
The previous page noted in passing that all Lie groups are parallelizable; the argument above is the
one promised there. Parallelizability is uncommon among smooth manifolds — among the spheres,
only \(\mathbb{S}^1\), \(\mathbb{S}^3\), and \(\mathbb{S}^7\) admit smooth global frames — so
the corollary restricts which manifolds can carry a Lie group structure compatible with their
smooth structure. The sphere \(\mathbb{S}^2\), for instance, is not parallelizable, and therefore
admits no Lie group structure as a smooth manifold.
Examples of Lie algebras of Lie groups
The three Lie groups \(\mathbb{R}^n\), \(\mathbb{S}^1\), and \(\mathbb{T}^n\) admit elementary
Lie-algebra computations from their additive group structure. In each case the Lie algebra turns
out to be abelian.
Example (Lie Algebras of \(\mathbb{R}^n\), \(\mathbb{S}^1\), and \(\mathbb{T}^n\)):
(a) Euclidean space \(\mathbb{R}^n\) under addition. Left translation by
\(b \in \mathbb{R}^n\) is the affine map \(L_b(x) = b + x\), whose differential in standard
coordinates is the identity matrix at every point. A smooth vector field
\(X = X^i\, \partial/\partial x^i\) is therefore left-invariant if and only if its component
functions \(X^i\) are constants, since
\[
d(L_b)_x\bigl( X^i(x)\, \partial/\partial x^i|_x \bigr)
= X^i(x)\, \partial/\partial x^i|_{x + b} ,
\]
and left-invariance equates this with \(X^i(x + b)\, \partial/\partial x^i|_{x + b}\),
forcing \(X^i(x) = X^i(x + b)\) for all \(b\). The
bracket of constant-coefficient vector fields
is zero, so the Lie algebra of \(\mathbb{R}^n\) is the \(n\)-dimensional abelian Lie algebra
consisting of constant-coefficient vector fields, isomorphic to \(\mathbb{R}^n\) itself with the
trivial bracket: \(\mathrm{Lie}(\mathbb{R}^n) \cong \mathbb{R}^n\) as abelian Lie algebras.
(b) The circle group \(\mathbb{S}^1\). In appropriate angle coordinates, each
left translation on \(\mathbb{S}^1\) is locally a translation \(\theta \mapsto \theta + c\) by
a constant, whose differential is the \(1 \times 1\) identity. The angle coordinate vector field
on the circle, introduced earlier in the previous-page examples, is therefore left-invariant,
and forms a basis for the one-dimensional Lie algebra \(\mathrm{Lie}(\mathbb{S}^1)\). A
one-dimensional Lie algebra is automatically abelian, since the bracket of any field with itself
is zero by antisymmetry. Hence \(\mathrm{Lie}(\mathbb{S}^1) \cong \mathbb{R}\) as abelian Lie
algebras.
(c) The \(n\)-torus \(\mathbb{T}^n = \mathbb{S}^1 \times \cdots \times \mathbb{S}^1\).
A parallel analysis on each circle factor shows that the tuple
\((\partial/\partial \theta^1, \dots, \partial/\partial \theta^n)\) is a basis for
\(\mathrm{Lie}(\mathbb{T}^n)\), where \(\partial/\partial \theta^i\) is the angle coordinate
vector field on the \(i\)th factor. Brackets of coordinate vector fields vanish, so
\(\mathrm{Lie}(\mathbb{T}^n) \cong \mathbb{R}^n\) as abelian Lie algebras.
The three groups in the example above are all abelian, and their Lie algebras turn out to be abelian
as well; this is no coincidence. Every abelian Lie group has an abelian Lie algebra. The converse
holds when the Lie group is connected, but not in general; a Lie group with abelian Lie algebra need
not itself be abelian if it has multiple components. Both directions are taken up later in the
development of the theory, where the relationship between connectedness and the Lie group/Lie
algebra correspondence is examined.
The Lie algebra as a linear model
Lie Algebras as Linear Models of Lie Groups
At a single point of a smooth manifold, the
tangent space
is the linear model: a finite-dimensional vector space that captures the first-order behavior of
the manifold near that point. For a Lie group, the analogous role is played globally by the
Lie algebra. The evaluation isomorphism identifies \(\mathrm{Lie}(G)\) with \(T_eG\) at the level
of vector spaces, so the linear-model viewpoint already applies pointwise at the identity.
What the Lie group structure adds is that the bracket on \(\mathrm{Lie}(G)\) is determined by
the group multiplication, and conversely many properties of the group can be read off from the
Lie algebra. The classification of connected Lie groups up to local isomorphism reduces to the
classification of finite-dimensional real Lie algebras — a strictly linear-algebraic
problem — and many global constructions on the group (one-parameter subgroups,
exponentials, conjugation) acquire concrete linear descriptions at the Lie algebra level. The
Lie algebra is the linear model of the Lie group, and it inherits algebraic structure from the
group rather than only vector-space structure from a single tangent space.
The examples treated so far are all abelian, and the bracket has played essentially no role in the
computations. The bracket structure becomes active and concrete on the general linear group and its
Lie subgroups, where every construction of this page can be carried out by explicit matrix
calculation, and where the abstract bracket of left-invariant vector fields can be identified with
the matrix commutator already developed in the linear-algebra track. That identification is the
content of the next stage of the development.