ProofComplete

The Exponential Map - Key Proof

We prove that the exponential map is a local diffeomorphism near the identity, establishing one of its most important properties. This result underpins much of the theory of Lie groups.

Proof

Theorem: Let GG be a Lie group with Lie algebra g\mathfrak{g}. The exponential map exp:gG\exp: \mathfrak{g} \to G is a local diffeomorphism in a neighborhood of 0g0 \in \mathfrak{g}.

Proof:

We apply the inverse function theorem. To do this, we must show that the differential d(exp)0:gTeGd(\exp)_0: \mathfrak{g} \to T_e G is an isomorphism.

Step 1: Identify the differential at zero.

For XgX \in \mathfrak{g}, consider the curve γ(t)=exp(tX)\gamma(t) = \exp(tX) in GG. This is a one-parameter subgroup, and by definition: γ(0)=exp(0)=e,γ(0)=X\gamma(0) = \exp(0) = e, \quad \gamma'(0) = X

The differential d(exp)0d(\exp)_0 sends XX to the tangent vector of γ\gamma at t=0t = 0, which is precisely XX when we identify TeGgT_e G \cong \mathfrak{g} via left-invariant vector fields.

Step 2: Show d(exp)0=idd(\exp)_0 = \text{id}.

More formally, the exponential map satisfies: ddtt=0exp(tX)=X\frac{d}{dt}\bigg|_{t=0} \exp(tX) = X for all XgX \in \mathfrak{g}.

Using the definition of the differential, for any XgX \in \mathfrak{g}: d(exp)0(X)=limt0exp(tX)etd(\exp)_0(X) = \lim_{t \to 0} \frac{\exp(tX) - e}{t}

In the identification TeGgT_e G \cong \mathfrak{g}, this limit is exactly XX. Therefore: d(exp)0=idgd(\exp)_0 = \text{id}_{\mathfrak{g}}

Step 3: Apply the inverse function theorem.

Since d(exp)0:gTeGd(\exp)_0: \mathfrak{g} \to T_e G is an isomorphism (in fact, the identity under our identification), the inverse function theorem guarantees that there exist neighborhoods UU of 00 in g\mathfrak{g} and VV of ee in GG such that: expU:UV\exp|_U: U \to V is a diffeomorphism.

Step 4: Explicit local inverse.

In the neighborhood where exp\exp is bijective, the inverse map is called the logarithm: log:VU\log: V \to U defined by log(exp(X))=X\log(\exp(X)) = X for XUX \in U.

For matrix groups, this coincides with the matrix logarithm: log(g)=n=1(1)n1n(gI)n\log(g) = \sum_{n=1}^\infty \frac{(-1)^{n-1}}{n}(g - I)^n which converges for gI<1\|g - I\| < 1. □

Remark

The size of the neighborhood where exp\exp is a diffeomorphism depends on the group. For compact groups, the injectivity radius is positive (exp is injective on a ball of fixed radius), while for some non-compact groups, the injectivity can fail arbitrarily close to the identity along certain directions.

Example

For G=SU(2)G = SU(2), which is diffeomorphic to S3S^3, the exponential map is injective on the open ball Bπ={Xsu(2):X<π}B_\pi = \{X \in \mathfrak{su}(2) : \|X\| < \pi\}. On the sphere X=π\|X\| = \pi, we have exp(X)=exp(X)=I\exp(X) = \exp(-X) = -I, so injectivity fails at the boundary.

Proof

Corollary: Every connected Lie group is generated by any neighborhood of the identity.

Proof: Let V=exp(U)V = \exp(U) be a neighborhood of ee in GG where UU is a neighborhood of 00 in g\mathfrak{g}. Since exp\exp is a local diffeomorphism, VV contains an open neighborhood of ee.

Consider the subgroup HH generated by VV. Since VV is open, HH is open in GG. But HH is also closed (being a subgroup of a Lie group). Since GG is connected, the only subset that is both open and closed is GG itself. Therefore H=GH = G, proving that VV generates GG. □

This corollary is powerful: to understand a connected Lie group, it suffices to understand an arbitrarily small neighborhood of the identity!