Eigen values interlacing result

for various types of graphs

Supervised By-
Dr. Jitendra Kumar Presented by-
Assistant Professor DIPALI SWAIN
Department of Mathematics CUJ/M/2017/MMA/08
Central University of Jharkhand
 Graph Theory
The Konigsberg Bridge Problem:- The problem was proposed and one of the
most notable problem in the history of mathematics.
The negative resolution was given by Leonhard Euler in 1736 that there exists no
such trip from any point such that every bridge is travelled exactly once and return
to the same point.

C
c g
d
A e D
a b
f
B
Resolving the problem into a graph theoretic problem
Graph:
A graph G is a pair of sets denoted by G(V, E), where V is
the set of vertices and E is the set of edges, connecting the
pairs of vertices.
v1 e1 v2 Degree:-
deg(v1)= 4
e3 deg(v2)= 2
e5 deg(v3)= 5
deg(v4)= 2
e9 e2 deg(v5)= 5 (self loop counted twice)
Walk:-
Trail:-
v5 Path:-

e4 v3 Circuit:-

e8 e7 e6

v4
 Complete graph Connected graph Minimally Connected graph

Disconnected Euler Graph Hamiltonian path

Connected Euler graph
v2 v1
v1 v2
v3 v4 v3 v4
v6 v5
v8 v5 v8 v7
v7 v6
Bipartite graph

v1 v6

v2 v5

v3 v4
Complete bipartite graph
 Matrices associated with Graph Theory
• Incidence matrix: Denoted by I with entries defined
as the following:-
1 if ej is incident on vi
aij=
v1 0 otherwise

e1 e5 e1 e2 e3 e4 e5 e6
e4 v4 e6 v1 1 0 0 0 1 0
v5
v2 v2 1 1 0 1 0 0
e2 e3 v3 0 1 1 0 0 0
v4 0 0 1 1 1 0
v3
v5 0 0 0 0 0 1
• Adjacency matrix: Denoted by A with entries defined
as the following:-
1 if vi adjacent to vj
aij =
0 otherwise

v1

e1 e5 v1 v2 v3 v4 v5
e4 v4 e6 v1 0 1 0 1 0
v5 v2 1 0 1 1 0
v2
v3 0 1 0 1 0
e2 e3
v4 1 1 1 0 1
v3 v5 0 0 0 1 0
• Laplacian matrix: Denoted by L with entries defined
as the following:-
di if vi = vj and d(vi) ≠ 0

-1 if vi is adjacent to vj
aij =
v1 0 otherwise

e1 e5 v1 v2 v3 v4 v5
e4 v4 e6 v1 2  1 0 1 0
v5
v2 v2  1 3  1 1 0
e3 v3 0  1 2 1 0
e2
v4  1  1  1 4 1
v3
v5 0 0 0 1 1
• Normalised Laplacian matrix: Denoted by L with
entries defined as the following:-
1 if vi = vj and d(vi) ≠ 0
1
aij =  if if vi is adjacent to vj
di d j
v1 0 otherwise
e1 v1 v2 v3 v4 v5
e5 1 1
1 0 0
e4 v4 e6 v1 6 2 2
v5 1 1 1
1 0
v2 v2 6 6 2 3
1 1
v3 0 1 0
e2 e3 1
6
1 1
2 2
1
v4 1
2 2 2 3 2 2 2
v3 v5 1
0 0 0 1
2
 Properties of Laplacian Matrix L of a graph.
• For a simple graph G with non isolated vertices, L = D - A
where D is the diagonal matrix with diagonal entries dii as
the degree of the ith vertex.
• The Laplacian matrix L is positive semi definite therefore
it has non negative eigen values.
• The row sum in a Laplacian matrix L is 0 thus (smallest
eigen value)λ1= 0 is always an eigen value for L.
• The second eigenvalue λ2 is non zero only for connected
graphs and is called Friedler value and the associated
eigenvector is called Friedler vector.
• The first(smallest) non zero eigen value λk is called the
spectral gap for L.
Properties of Normalised Laplacian Matrix

• The notion of Normalised Laplace matrix was given by Fan

R. K. Chung in her book Spectral Graph Theory.
• The normalised Laplacian matrix L for a graph G with non
isolated vertices can also be written as
L = D-1/2 L D-1/2 = I- D-1/2 A D-1/2
• A reason why Normalised Laplacian matrix is preferred over
the Laplacian Matrix is its consistency of eigenvalue.
• L is a semi definite matrix thus (smallest eigen value)λ1= 0
n
• The largest eigenvalue λn,  n  2 and λn= 2 if G is
n 1
bipartite.(The proof of which is given in the “Journal of Inequalities and
Applications2014” Jianxi Li, Ji-Ming Guo, Wai Chee Shiu)
 Interlacing
Definition: Consider two sequences of real numbers
λ1 ≥ λ2 ≥ ... ≥ λn and μ1≥μ2≥ ... ≥μm with m < n.
The second sequence is said to interlace with the first whenever
λi ≥μi ≥ λn-m+i i=1, 2, 3, ..., m
The interlacing is called tight if there exists
k∊ {0, 1, ..., m} such that λi =μi 1 ≤ i ≤ k
and μi=λn-m+i when k+1≤ i ≤ m

Lemma: (Interlacing Theorem):Let A be an n × n symmetric

matrix with eigen values |λ1| ≥ |λ2|≥ ... ≥ |λn|.
Let B be an (n-k) × (n-k) symmetric minor of A with
eigenvalues |μ1| ≥ |μ2| ≥ ... ≥ |μn-k |
Then |λi| ≥ |μi| ≥ |λi+n-(n-k)|
 Known interlacing results
• Result 1
Let G be a graph of order n and let H = G − v, where v is a
vertex of G. Then Cauchy’s interlacing theorem gives an
interlacing property of the eigenvalues of A(G) and the
eigenvalues of A(H), which we refer to as the vertex version
of the interlacing property.
Let G be a graph and H = G − v, where v is a vertex of G. If
|λ1| ≥ |λ2|≥ ... ≥ |λn| and |θ1| ≥ |θ2| ≥ ... ≥ |θn-1| are the
eigenvalues of A(G) and A(H), respectively, then λi ≥ θi ≥
λi+1 for each i = 1, 2, 3, 4,...,n − 1.
• Result 2
Let G be a graph and H = G − e, where e is an
edge of G.
If |λ1| ≥ |λ2|≥ ... ≥ |λn|= 0 and |θ1|≥ |θ2|≥ ...
≥|θn|= 0 are the eigen values of L(G) and
L(H), respectively,
then |λi|≥ θi ≥ λi+1 for each i= 1, 2, 3, 4,...,n − 1.
• Result 3
Let G be a graph of order n and H = G−v,
where v is a vertex of G of degree r.
If |λ1| ≥ |λ2| ≥ ... ≥ |λn| = 0 and |θ1| ≥ |θ2| ≥ ... ≥
|θn| = 0 are the eigen values of L(G) and L(H),
respectively,
then λi ≥ θi ≥ λi+1 for each i = 1, 2, …, n − 1
where λi = 0 for i ≥ n + 1.
 Aim of the Project
• One of our goals is to describe all eigenvalue interlacing results for A
(G), L (G), and L (G), associated with the removal of an edge or
vertex. Three of the six possible cases have been resolved:
• eigenvalue interlacing result on the adjacency matrix when a vertex
is removed.
• eigenvalue interlacing result on the Laplacian when an edge is
removed
• eigenvalue interlacing result on the Laplacian when a vertex is
removed.

The aim of my project is eigenvalue interlacing result on the

normalised Laplacian L (G) when a vertex is removed.
 Applications of eigen values of
Graph in real life Problems
• One other major use is by social networking sites like
Facebook so as to decide how to make a friend
community for a particular user, which ads, which
friends suggestion to show etc.
• In quantum chemistry, the skeleton of a non-saturated
hydrocarbon is represented by a graph. The energy
levels of the electrons in such a molecule are the
eigenvalues of the graph. The stability of molecules is
closely related to the spectrum of its graph.
• In Shannon information theory(1948), the channel capacity,
which characterizes the maximum amount of information that is
transmitted over a channel or stored into a storage medium per
bit, can be expressed in terms of the eigenvalues of its channel
graph.
• In coding theory, the minimum Hamming distance of a linear
code can be represented by the second largest eigenvalue of a
regular graph. (Hamming distance is the number of entries in
which two codeword differ.)
• In geographic studies, the eigenvalues and eigenvectors of a
transportation network provide information about its
connectedness. It is proven that the more highly connected in a
transportation network G is, the larger is the largest eigenvalue.
This gives rise to an open problem, given vertices, can one
construct a graph with the larger eigenvalue λ1 compared to its
other forms?
 References
• “Journal of Inequalities and Applications-2014” Jianxi Li, Ji-
Ming Guo, Wai Chee Shiu (Published Aug 2014)
• Fan R.K. Chung, Spectral Graph Theory, American Mathematical
Society, 1996
• Narsingh Deo, Graph Theory with applications to engineering and
Computer Science; 1974
• Chris Godsil, Gordon Royle , Algebraic Graph Theory Springer,
2000
• Owen Jones, Spectra of Simple Graphs; Whitman College; May 13,
2013
• Jiaqi Jiang, An Introduction to Spectral Graph Theory
THANK YOU.