Lie Groups and Lie Algebras Lecture 1 notes

Definition: Lie Group

A Lie Group is a space that possesses two structures:
1) structure of a group
2) structure of a smooth manifold

Examples (main are matrix groups):
- General linear group GL(n,R)GL(n, \mathbb{R})
- Special linear group SL(n,R)SL(n, \mathbb{R})
- Orthogonal groups O(n,R)O(n, \mathbb{R}), SO(n,R)SO(n, \mathbb{R})
etc.

Simplest examples:

SO(2)SO(2) = {2×2\{2\times 2 matrix A\mathbf{A} s.t. AAT=id\mathbf{AA}^T = \text{id}, det A=1}\mathbf{A} = 1\}
Then, it's not hard to verify that SO(2)SO(2) is the set of all rotation matrices.
As a smooth manifold, SO(2)SO(2) is a circle (trivial to show)

Similarly, the orthogonal group:
O(2)={2×2O(2) = \{2\times 2 matrix A\mathbf{A} s.t. AAT=id}\mathbf{AA}^T = \text{id}\}
This ends up to be similar to SO(2)SO(2), except the polarity of the right side of the matrix is flipped (determinant can be either 1 or -1)
This is a disjoint union of two circles

SL(2,R)={2×2SL(2, R) = \{2\times 2 matrix A\mathbf{A} s.t. det A=1}A = 1\}
Because there are 4 dimensions (2×22\times 2), this defines a second order hypersurface in R4\mathbb{R}^4 #-> Second order hypersurface? Defined by second order polynomial

Theorem: Proposition 1

SL(2,R)SL(2, \mathbb{R}) is three-dimensional and diffeomorphic to the direct product S1×R2S^1 \times \mathbb{R}^2.
Sketch of proof:
If we write a=x+y,d=xy,b=u+v,c=uva = x + y, d = x - y, b = u + v, c = u - v, we get
adbc=1x2y2u2+v2=1ad - bc = 1 \to x^2 - y^2 - u^2 + v^2 = 1
then, x2+v2=1+y2+u2x^2 + v^2 = 1 + y^2 + u^2
If (y,u)R2(y, u) \in \mathbb{R}^2, then (x,v)(x, v) lies on a circle S1S^1 (of variable radius). Hence, ((y,u),(x,v))R2×S1((y, u), (x, v)) \in \mathbb{R}^2 \times S^1.

Unitary group:

U(2)={2×2U(2) = \{2\times 2 matrix where coeffs in C\mathbb{C}, AA=id}\mathbf{AA}^* = \text{id}\} #-> Implication of conjugate?

Special unitary group:
SU(2)={2×2SU(2) = \{2\times 2 matrix where coeffs in C\mathbb{C}, AA=id\mathbf{AA}^* = \text{id}, and det A=1}\mathbf{A} = 1\}

Theorem: Proposition 2

SU(2)SU(2) is three-dimensional and diffeomorphic to 3 dimensional sphere S3S^3. U(2)U(2) is four-dimensional and diffeomorphic to the direct product S3×S1S^3 \times S^1.
Sketch of proof:
In terms of a,b,c,da, b, c, d, AA=1\mathbf{AA}^* = 1 can be represented as three equations.
The first, aa+bb=1a\overline{a} + b\overline{b} = 1, means that (a,b)(a, b) in C2=R4\mathbb{C}^2 = \mathbb{R}^4 belongs to the sphere of radius 1 (because a12+a22+b12+b22=1a_1^2 + a_2^2 + b_1^2 + b_2^2 = 1)
The second, ca+db=0c\overline{a} + d\overline{b} = 0, means that (c,d)(c, d) is proportional (b,a)(-\overline{b}, \overline{a}). This means that c=λbc = -\lambda\overline{b}, d=λad = \lambda\overline{a} for some λ\lambda.
The third, cc+dd=1c\overline{c} + d\overline{d} = 1, means that if we plug in our above expressions for cc and dd, we get λ=1|\lambda| = 1, which means that λ=eiϕ\lambda = e^{i\phi} for some ϕ\phi.
Then, matrices in U(2)U(2) are [abλba]\begin{bmatrix}a & b \\ -\lambda\overline{b} & \overline{a}\end{bmatrix}.
Because the determinant of matrices in U(2)U(2) are exactly λ\lambda, setting λ=1\lambda = 1 gives us matrices in SU(2)SU(2): [abba]\begin{bmatrix}a & b \\ -\overline{b} & \overline{a}\end{bmatrix}.
Then, there is a natural bijection between the 3-sphere S3S^3 = {(a,b,c,d)R4 s.t. a2+b2+c2+d2=1}={(a,b)C2 s.t. a2+b2=1}\{(a, b, c, d) \in \mathbb{R}^4 \text{ s.t. } a^2 + b^2 + c^2 + d^2 = 1\} = \{(a, b) \in \mathbb{C}^2 \text{ s.t. } |a|^2 + |b|^2 = 1\} and SU(2)SU(2).
In the case of U(2)U(2), we have an additional parameter λ=eiϕ\lambda = e^{i\phi}, which defines a circle, so U(2)=S3×S1U(2) = S^3 \times S^1.

Definition: Abelian Group

A group is called Abelian if the binary operation is commutative.
Examples:
- Any vector space VV over R\mathbb{R} can be an abelian lie group, with the binary operation being addition: (u,v)u+v(u, v) \mapsto u + v.
- A torus Tn=S1×S1× n times ×S1T^n = S^1 \times S^1 \times \dots \text{ n times } \dots \times S^1 is also an abelian lie group. Representing each point on TnT^n as an nn-tuple (ϕ1,ϕ2,,phin)(\phi_1, \phi_2, \dots, phi_n) where each ϕk\phi_k is an "angle" mod 2π2\pi. The binary operation then is the addition mod 2π2\pi: (ϕ1,dots,ϕn)+(ψ1,,ψn)=((ϕ1+ψ1)mod2π,,(ϕn+ψn)mod2π)(\phi_1, dots, \phi_n) + (\psi_1, \dots, \psi_n) = ((\phi_1 + \psi_1) \mod 2\pi, \dots, (\phi_n + \psi_n) \mod 2\pi)
Note that TnT^n can be represented as a matrix lie group, if we assign each nn-tuple (ϕ1,,ϕn)(\phi_1, \dots, \phi_n) to the diagonal matrix with each component as eiϕke^{i\phi_k}.

On a group, we can define many other structures of Lie groups (not all commutative). Ex. on R3\mathbb{R}^3 we can define:
(x1,x2,x3)(y1,y2,y3)=(x1+y1,x2+y2,x3+y3+x1y2)(x_1, x_2, x_3) \cdot (y_1, y_2, y_3) = (x_1 + y_1, x_2 + y_2, x_3 + y_3 + x_1y_2)
The smoothness of this operation is evident, even though it's non-linear. #-> Why? Is it because each component is a sum or product?
We have a natural matrix representation, the group of upper triangular matrices: G={A=[1x1x301x2001]}G = \{\mathbf{A} = \begin{bmatrix}1 & x_1 & x_3 \\ 0 & 1 & x_2 \\ 0 & 0 & 1\end{bmatrix}\}. We can see that a product of two of such groups yields the corresponding group.

General Linear Group:
GL(n,R)GL(n, \mathbb{R}) = all n×nn\times n matrices with non-zero determinants
It's smooth because it's an open subset in the vector space Rn2\mathbb{R}^{n^2}.
Because GL(n,R)GL(n, \mathbb{R}) is not connected, it consists of {A: det A>0}\{\mathbf{A}: \text{ det } \mathbf{A} > 0\} and {A: det A<0}\{\mathbf{A}: \text{ det } \mathbf{A} < 0\}

Theorem: Proposition 3

In the simplest case GL(2,R)GL(2, \mathbb{R}), each of these components (refering to {A: det A>0}\{\mathbf{A}: \text{ det } \mathbf{A} > 0\} and {A: det A<0}\{\mathbf{A}: \text{ det } \mathbf{A} < 0\}) is diffeomorphic to S1×R3S^1 \times \mathbb{R}^3.
Rough proof sketch:
We can express any matrix in GL(2,R)GL(2, \mathbb{R}) with positive determinant as a product of some positive constant and a matrix in SL(2,R)SL(2, \mathbb{R}).
Then, as SL(2,R)S1×R2SL(2, \mathbb{R}) \cong S^1 \times \mathbb{R}^2 and R+R\mathbb{R}_+ \cong \mathbb{R}, the extra scaling coefficient adds a dimension, and thus GL(2,R+)S1×R3GL(2, \mathbb{R}^+) \cong S^1 \times \mathbb{R}^3. The same argument can be made for GL(2,R)GL(2, \mathbb{R}_-).