Lie Groups and Lie Algebras Lecture 4 notes

General Properties of Lie Groups

Definition: Lie Group

A Lie Group GG is a group which is a smooth manifold such that the group operations are smooth
Standard notation for the identity: eGe \in G
Multiplication in local coordinates: Let UU be a neighborhood of eGe \in G with some local coordinates. Assume that the coordinates of ee are (0,0,,0)(0, 0, \dots, 0).
Take an element xUx \in U with coords (x1,x2,,xn)(x_1, x_2, \dots, x_n) and yy with coords (y1,y2,,yn)(y_1, y_2, \dots, y_n). Then, multiplying them via x,yz=xyx, y \mapsto z = x \ast y, then the components of zz are smooth functions of xx and yy:
zk=zk(x1,,xn;y1,,yn)z_k = z_k(x_1, \dots, x_n; y_1, \dots, y_n)
It is interesting and useful to look at the taylor expansion of zkz_k.

Theorem: Taylor Expansion of Z_k

zk=xk+yk+ijbij,kxiyj+z_k = x_k + y_k + \sum_{ij} b_{ij,k} x_i y_j + \dots (terms of degree 3\geq 3) #-> Taylor expansion of zkz_k
similarly, if we consider the inversion map xu=x1x \to u = x^{-1} then the first terms for the Taylor expansion are:
uk=xk+ijbij,kxiyj+u_k = -x_k + \sum_{ij}b_{ij,k} x_i y_j + \dots (terms of degree 3\geq 3)
The constants bij,kb_{ij,k} are specific to keep multiplication associative. If bij,k=bji,kb_{ij,k} = b_{ji,k} then the group will be commutative #-> because then swapping xx and yy doesn't change the sum

In general, the taylor expansion of an arbitrary smooth function zk=zk(x1,,xn;y1,,yn)z_k = z_k(x_1, \dots, x_n; y_1, \dots, y_n) can be written as:
zk=a0+ljxj+mjyj+aijxixj+bijxiyj+cijyiyj+z_k = a_0 + l_j x_j + m_j y_j + a_{ij} x_i x_j + b_{ij} x_i y_j + c_{ij} y_i y_j + \dots
with summation over ii and jj.
If we substitute x=ex = e (=1= 1), then we have z=ey=yz = ey = y, i.e.
yk=a0+mjyj+cijyiyj+y_k = a_0 + m_j y_j + c_{ij} y_i y_j + \dots
hence a0=0a_0 = 0, mj=0m_j = 0 for all jj except for j=kj = k and mk=1m_k = 1, and cij=0c_{ij} = 0.
Similarly, substituting y=ey = e gives lj=0l_j = 0 for all jj except j=kj = k and lk=1l_k = 1, and aij=0a_{ij} = 0.
So, zk=xk+yk+ijbijxiyj+...z_k = x_k + y_k + \sum_{ij} b_{ij} x_i y_j + ...

Component of the identity
Remember that any smooth manifold MM can be presented as a disjoint union of connected manifolds called connected components of MM:
M=M1M2,Mi{},MiMj=M = M_1 \bigcup M_2 \bigcup \dots, M_i \neq \{\}, M_i \bigcap M_j = \varnothing
E.g. M=S1S1M = S^1 \bigcup S^1 is two disjoint circles; M=S1R1M = S^1 \bigcup \mathbb{R}^1 is a circle and an independent line
The second example is not a lie group because all connected components need to have the same topological structure #-> Because no operations between elements on a line and elements on a circle?

Theorem: Connected Components in Lie Groups

Let GG be a Lie group, G=G0G1G = G_0 \bigcup G_1 \bigcup \dots be the decomposition into connected components, and G0GG_0 \subset G be the connected component that contains eGe \in G. In other words, G0 is a connected open-closed subset which contains e.
Notice that such a subset is unique and can be characterized as the set of those points which can be joined with e by a continuous path. #-> Unique because of the disjoint path-connected clause
.#-> In manifolds, connected = path connected
Theorem:
- G0G_0 is a normal subgroup in GG;
- G0G_0 itself is a Lie group;
- any other connected component GiGG_i \subset G is diffeomorphic to G0G_0.

Proof:

Why subgroup: verify following conditions:
(i) G0G_0 is closed under multiplication
(ii) G0G_0 is closed under inversion
Proof of (i): If a,bG0a, b \in G_0 and G0G_0 is path-connected, then there are continuous curves a(t)a(t) and b(t)b(t) which connect ee with aa and bb respectively, i.e. a(0)=ea(0) = e, a(1)=aa(1) = a, and b(0)=eb(0) = e, b(1)=bb(1) = b.
Then, c(t)=a(t)b(t)c(t) = a(t) b(t) is a continuous curve connecting ee with abab. Thus abab belongs to the connected component that contains ee, i.e. G0G_0. #-> Continuity ensures c(t)c(t) is a path, and path-connectedness of G0G_0 ensures that c(t)c(t) is on G0G_0
Proof of (ii) is similar; it suffices to consider the continuous curve a(t)a(t) and c(t)=a(t)1c(t) = a(t)^{-1}.

Why G0G_0 is normal:
We just need to verify that for any gGg \in G, we have gG0g1=G0gG_0 g^{-1} = G_0.
Proof: Since the left multiplication by gg and right multiplication by g1g^{-1} are both diffeomorphisms, gG0g1gG_0 g^{-1} is a certain connected component of GG. #-> Why are they diffeomorphisms?
We know this because the left and right diffeomorphisms means that gG0g1gG_0 g^{-1} have the same topological structure as G0G_0. In addition, because geg1=egeg^{-1} = e, we know that egG0g1e \in gG_0 g^{-1}
Then, we can ensure that gG0g1=G0gG_0 g^{-1} = G_0 (because it is path-connected to ee).

Why any GiG_i is diffeomorphic to G0G_0:
If we take G0G_0 as the subgroup of GG, then you can see that GiG_i is a left or right coset of G0G_0, so taking an arbitrary element gGig \in G_i, Gi=gG0G_i = gG_0. #-> Why are they left/right cosets?

Theorem: Connected Component Generation

Let UU be any connected neighborhood of eGe \in G. Then, G0G_0 is generated by UU in the sense that any element xG0x \in G_0 can be represented as x=u1u2u3ukx = u_1 \ast u_2 \ast u_3 \ast\dots\ast u_k with uiUu_i \in U.
Remark:
- If UU is a neighborhood of ee, then aUaU and UaUa are both neighborhoods of aGa \in G. These neighborhoods are naturally homeomorphic (and diffeomorphic) to UU, but in general do not coincide.
- The set U1={u1uU}U^{-1} = \{u^{-1} | u \in U\} is a neighborhood of the identity ee.
Proof:
Consider subsets U1=U,U2=UU={u1u2u1,u2U},,UkU_1 = U, U^2 = U\ast U = \{u_1 \ast u_2 | u_1, u_2 \in U\}, \dots, U^k and take L=L = infinite union of UkU^k over kk. We need to prove L=G0L = G_0.
Because G0G_0 is a connected component, it's a set that's open, closed, and connected.
1) Clearly, LL is an open set because it's a union of open subsets.
2) LL is closed, because GLG\setminus L is open. Assume aGLa \in G\setminus L and consider its neighborhood aU1aU^{-1}. Check that this neighborhood does not intersect with LL.
By contradiction: if aU1LaU^{-1} \bigcap L \neq \varnothing then there must exist uu and u1,,uku_1, \dots, u_k such that au1=u1ukau^{-1} = u_1 \ast \dots \ast u_k or equivalently a=u1ukua = u_1 \ast \dots \ast u_k \ast u.
That is, aUk+1La \in U^{k + 1} \subset L. This is a contradiction, as we assumed aGLa \in G\setminus L. #-> If every point in a set has an open neighborhood contained in the set, then the set is open?
3) LL is connected, because for any product x=u1u2ukLx = u_1 \ast u2 \ast \dots u_k \in L, there exists a continuous path γ(t)\gamma(t) on LL which connects xx with the identity element ee. This is obvious, because we can simply take γ(t)=γ1(t)gamma2(t)γk(t)\gamma(t) = \gamma_1(t) \ast gamma_2(t) \ast \dots \ast \gamma_k(t) where γi(t)\gamma_i(t) is a continuous path that connects ee and uiu_i, i.e. γi(0)=e\gamma_i(0) = e, γi(1)=ui\gamma_i(1) = u_i. Then obviously γ(0)=e\gamma(0) = e and γ(1)=x\gamma(1) = x
Thus, LL is a connected open-closed subset which contains the identity eGe \in G, so by definition LL is the connected component G0G_0.

Definition: Algebraic Linear Group

An algebraic linear group GG is a subgroup in GL(n,R)GL(n, \mathbb{R}) which is defined by a system of polynomial equations pm(a11,a12,,ann)=0p_m(a_{11}, a_{12}, \dots, a_{nn}) = 0 where aija_{ij} are matrix components of AGL(n,R)\mathbf{A} \in GL(n, \mathbb{R}).
Examples:
- SL(n,R)SL(n, \mathbb{R}) is given just by one equation, det A=1\text{det } \mathbf{A} = 1.
- O(n)O(n) is given by one matrix equation AAT=id\mathbf{AA}^{T} = \text{id}, which amounts to n(n+1)2\frac{n(n + 1)}{2} usual equations. In the case of SO(n)SO(n), one additional equation appears: det A=1\text{det } \mathbf{A} = 1.
etc.

Theorem: Subgroup of General Linear Group Is Linear Group

Any algebraic linear group GGL(n,R)G \subseteq GL(n, \mathbb{R}) is a Lie subgroup in GL(n,R)GL(n, \mathbb{R}). In particular, GG is a Lie group.
Proof sketch:
The main problem is whether a subgroup GGL(n,R)G \subseteq GL(n, R) has any singular points.
Take VV as a neighborhood of e=idGL(n,R)e = \text{id} \in GL(n, \mathbb{R}), and Ue=GVU_e = G \bigcap V as the neighborhood of eGe \in G. Then, for any xGx \in G, Ux=xUe=x(GV)=G(xV)U_x = xU_e = x(G \bigcap V) = G\bigcap (xV) is the neighborhood of xGx \in G.
Then, because the left translation by xx is a diffeomorphism of GL(n,R)GL(n, \mathbb{R}), UxU_x and UeU_e are absolutely "isomorphic" from both topological and diffeomorphic viewpoints.
Because of this, if locally GG has the structure of a smooth submanifold of GL(n,R)GL(n, \mathbb{R}), then this condition should hold at any other point in GG.