Lie Algebras Constructed From Graphs, by A.J. Butler

Graphs

Graphs have nodes, and edges that connect nodes
We are interested in graph isomorphisms specifically

Definition: Graph Isomorphism

Graphs GG and GG' are isomorphic if we can construct a bijection between vertex sets f:VVf: V\to V' s.t. u,vVu, v \in V are adjacent iff f(u),f(v)Vf(u), f(v) \in V' are adjacent
i.e. adjacent vertices are mapped to adjacent vertices

Turns out it's not easy to see whether graphs are isomorphic at first glance, esp with large number of nodes/vertices #-> this is mostly recap from Ted GT class

Lie brackets and Lie Algebras

Definition: Lie Algebra

A Lie Algebra g\frak{g} is a vector space with a bilinear operation [,]:g×gg[, ]: \frak{g}\times \frak{g} \to \frak{g} called the Lie bracket with
1) [x,x]=0[x, x] = 0 for all xgx \in \frak{g}
2) [x,y]=[y,x][x, y] = -[y, x] (implied by 1 and bilinearity)
3) [x,[y,z]]+[y,[z,x]]+[z,[x,y]]=0[x, [y, z]] + [y, [z, x]] + [z, [x, y]] = 0 for all x,y,zgx, y, z \in g

Example: We see that R3\mathbb{R}^3 with cross product is a Lie algebra
A×B=B×AA \times B = -B \times A according to right hand rule, and we can validate that the Jacobi identity works

Another example: If AA is an associative F\mathbb{F}-algebra (i.e. an F\mathbb{F}-vector space equipped with bilinear associative operation \ast).
For all v,wAv, w\in A, define the commutator [v,w]vwwv[v, w] \coloneqq v\ast w - w\ast v.
An explicit example of this would be gln(R)\frak{gl}_n(\mathbb{R}), the set of all n×nn\times n matrices over R\mathbb{R} with the Lie brackets defined as [A,B]=ABBA[A, B] = A\ast B - B \ast A.

Lie Algebras are an approdimation of the notions of symmetry, which are easy to work in.

Definition: Just Some Notations/ideals

The lie algberas we look at are nilpotent and solvable.

Definition: Lie Algebra Isomorphisms

A **Lie Algebra Isomorphism is a Lie Algebra Homomorphism that is also a Vector Space Isomorphism.
Formally, ϕ:gh\phi: \frak{g}\to \frak{h} s.t. [ϕ(x),ϕ(y)]=ϕ([x,y])[\phi(x), \phi(y)] = \phi([x, y]) for all x,ygx, y \in \frak{g}
Vector space isomorphism: Linear, bijection, thus dim equal and ϕ(v)=0\phi(v) = 0 implies v=0v = 0

Examples:
- Identity map xxx \mapsto x is a Lie Algebra to itself
- If the brackets are trivial ([v1,v2]=0v1,v2V[v_1, v_2] = 0 \forall v_1, v_2 \in V and [w1,w2]=0w1,w2W[w_1, w_2]=0\forall w_1, w_2\in W), then they are isomorphic iff dim are the same

Theorem: Isomorphism On N X N Matrices

If we take gln(k)\frak{gl}_n (k) as the Lie algebra given by all n×nn\times n matrices over field kk with the commutator as the Lie Bracket, then the map

AAT \begin{align*} A \mapsto -A^{T} \end{align*}


is a Lie algebra isomorphism from gln(k)\frak{gl}_n (k) to itself.
Proof:
If we let ϕ:AAT\phi: A\mapsto -A^{T}, then we have

ϕ[A,B]=ϕ(ABBA)=(ABBA)T=((AB)T(BA)T)=(BTATATBT)=BTAT+ATBT \begin{align*} \phi[A, B] &= \phi(A\ast B - B\ast A) \\ &= -(A\ast B - B\ast A)^T \\ &= -((A\ast B)^T - (B\ast A)^T) \\ &= -(B^TA^T - A^TB^T) \\ &= -B^TA^T + A^TB^T \\ \end{align*}


We also have

[ϕ(A),ϕ(B)]=[AT,BT]=AT(BT)(BT)(AT)=ATBTBTAT \begin{align*} [\phi(A), \phi(B)] &= [-A^T, -B^T] \\ &= -A^T(-B^T) - (-B^T)(-A^T) \\ &= A^TB^T - B^TA^T \\ \end{align*}


QED

Lie Algebras on Graphs

Definition: Lie Algebra Associated On a Graph

Let G=(V,E)G = (V, E) be a simple graph (no self loops). Let g=spanR(VE)\frak{g} = \text{span}_{\mathbb{R}}(V\bigcup E) be a vector space.
We define the bracket by

[vi,vj]={ekif i<j and ek is an edge between vi and vjekif i>j and ek is an edge between vi and vj0otherwise [v_i, v_j] = \begin{cases} e_k &\text{if } i < j \text{ and } e_k \text{ is an edge between } v_i \text{ and } v_j \\ -e_k &\text{if } i > j \text{ and } e_k \text{ is an edge between } v_i \text{ and } v_j \\ 0 & \text{otherwise}\\ \end{cases}


.#-> Related to cohomology? Holden: Yes, but not in the way you think it is
.#-> A.J.: The graph structure informs the cohomology of the Lie Algebra of the graph

All other brackets not defined by linearity are zero.
The vector space g\frak{g} with this braket operation is the Lie algebra associated with GG.
Because it acts as a vector space, we can have linear combinations and it's more powerful than just an adjacency matrix

We can compare graphs via the Lie Algebras as well! cool beans.
Because the Lie Algebra structure is mapped to the graph structure, we can reduce graph isomorphisms to Lie Algebra isomorphisms.

Theorem: Mainkar's Theorem

Let G1,G2G_1, G_2 be graphs. Then, G1G2G_1 \cong G_2 iff g1g2\frak{g}_1 \cong \frak{g}_2.
Forward direction is intuitive, because Lie Algebra structure relies on graph structure
We see that we can embed graph theory into lie theory.

Applications:
- Embeds simple graphs into Lie Algebras
- Helps categorize a subgroup (2 step nilpotent, solvable) of Lie Algebras
- May allow us to take results from Graph Theory (like Hall's Marriage Problem) and restate them in terms from Lie Theory, or vice versa