1. Determine all equitable bipartitions of the hypercube such that the corresponding quotient matrix has eigenvalues and .
\begin{proof}
Recall that hypercube has vertices , where iff for exactly one position.
We call the partition equitable if for all , where .
Link to original
原来 bipartition 是分两块的意思。。。
Let be a bipartition. Then , where , where and . Then the quotient matrix for some . Notice that is an eigenvector of with eigenvalue , which implies . Hence . If or , then has an entry , which contradicts with connected. So and .
Take and . Then is the adjacent matrix of and so , . Since and is a -regular graph, we have and so . Using similar construction, that is, fix one coordinate and
- take and , or
- take and .
Then we get bipartitions such that the corresponding quotient matrix has eigenvalues and .
It remains to check there does exist no other bipartition.
Since , we know each row of has exactly one and so for .
Then there exists a - correspondence between elements of set and .
Assume that and , where and for any .
If , then and for any , where is the standard basis.
Repeat this procedure, we know for all .
Now we finish the proof.
\end{proof}
2. Consider a polarity of a projective plane of order .
- (a) Define a graph with vertex set , two points adjacent if . Determine the spectrum of . Note that has loops.
- (c) Let be the induced subgraph of obtained by removing all vertices adjacent to themselves. Show that the spectrum of is in .
\begin{proof}
Recall that a polarity is a bijection that sends points to lines and lines to points, .
is called the order of the projective plane .
Link to original
a) Let be a adjacent matrix. Then satisfies and as two lines has exactly one common point. Hence .
Note that , then is an eigenvalue of . For any eigenvector of , we assume that . Since , we have . So . Therefore, .
Since , we have . If is not a perfect square, then by we have . If is a perfect square, then , where .
c) Let be a adjacent matrix of . Recall that
If is a principal submatrix of , then the eigenvalues of interlace the eigenvalues of .
Link to original
Then we know the eigenvalue of interlace the eigenvalues of . Let be all eigenvalues of . Then for .
It remains to show . GPT 告诉我这里有一个反例:
Take the projective plane . Define , and define . Since , we have . Since is non-degenerated, has dimension and so . Hence is a polarity.
There are three vertices adjacent to themselves, and so .
Note that .
Let be a adjacent matrix of .
Then .
Since , the spectrum of is NOT in .
\end{proof}
3.
- (a) Determine the spectrum of the point-line incidence graph of .
- (b) We call a collection of -subsets of points of a -uniform local arc if the are pairwise disjoint, and is an arc. Show that for fixed,
\begin{proof}
a) Let be a adjacent matrix of the point-line incidence graph.
Then , where is the point-line incidence matrix.
Note that both and have eigenvector , then is an eigenvector of with eigenvalue .
Assume that is an eigenvector of , then . It follows that and .
Since , the spectrum of is , where . So by ^vnqcpz.
b) 白给的尝试:Assume that and such that where iff . Let , where if . Then , where . We also have and so , where and . By ^ja6tar, we know and so .
Let be a -uniform local arc, and let . Then . Note that the point-line incidence graph of is -regular and by a). By ^dhoh0u, for any set of points and any set of lines, we have
where . Define . Since is an arc, is also an arc and then each determines exactly secant lines. Note that for any , which follows that . By we have
Furthermore, , then we have
So
Now we finish the proof.
\end{proof}
4. Consider the odd cycles , where .
- (a) Use the ratio bound to show that .
- (b) Use the inertia bound to show that .
- (c) Show that both bounds are tight.
\begin{proof}
a) Recall that is the independence number of .
Since the minimal eigenvalue of is by ^mt7vc7, there is by ^j2qp40.
Since , we have .
b) By ^s7pdtq, . Since there does not exist such that , we have . Therefore, .
c) Assume that , where and .
Then is an independent set and so both bounds are tight.
\end{proof}
5. Consider the Johnson graph , whose vertices are the -subsets of , with two vertices adjacent when their intersection has size . Let be its adjacency matrix.
- (a) Show that is regular of degree .
- (b) Determine the number of common neighbors of two vertices , depending on .
- (c) Write as a linear combination of .
- (d) Determine the spectrum of .
\begin{proof}
a) For each point , consider the neighborhoods of .
If and , then there are ‘s such that .
Similarly for and .
So is regular of degree .
b) It is easy to check that
- if , then the number of common neighbors of is ;
- if , then the number of common neighbors of is ;
- if , then the number of common neighbors of is ;
- if , then the number of common neighbors of is .
c) By b) we have
Recall that is the number of -walks from to , then we have
The cases of are easy to check. When , take and . Then
- with , and has choices;
- , and has choices;
- with , and has 4 choices;
- , and has choices.
Therefore, one obtains the following table:
Solving for constants such that , we get
d) Since is regular of degree , is an eigenvector with eigenvalue .
On the subspace orthogonal to the all-one vector, we have . So every non-principal eigenvalue satisfies
by c). Note that this factors as
Therefore, the eigenvalues are , , and .
Assume that these multiplicities are , respectively. Then we have ,
and
Therefore, we have
Now we finish the solution.
\end{proof}
6. Determine existence or nonexistence for each of the following SRG parameters:
- (a) ,
- (b) ,
- (c) ,
- (d) ,
- (e) ,
- (f) .
\begin{proof}
Recall that ^bebz6b and ^peg7jo.
a) Since and , such SRG does not exist.
b) is an example.
c) The Shrikhande graph is an example.
d) Eigenvalues are and . Then is not an integer and so such SRG does not exist.
e) One can compute that , , , and . Since does not hold, by absolute bound such SRG does not exist.
f) One can compute that , , , and .
Since does not hold, by absolute bound such SRG does not exist.
\end{proof}