A Proof of Brouwers Conjecture - Vilmar
A Proof of Brouwers Conjecture - Vilmar
Manuscript Draft
Abstract. Building up from recent results, we present a short proof for Brouwer’s Conjecture.
1. Introduction
[n]
Let G ⊆ 2
be a simple undirected graph. We define its Laplacian as
L(G) = D(G) − A(G),
where A(G) is the n × n incidence matrix given by:
1, iff (i, j) ∈ G,
ai,j =
0, otherwise;
and D(G) is the diagonal n × n matrix:
di (G) = |{j : (i, j) ∈ G}|.
P
Note that tr(L(G)) = i di (G) = 2m(G), were m(G) is the number of edges in G. Moreover, a
standard argument shows that L(G) is nonnegative. We denote its eigenvalues as:
0 = µn (L(G)) ≤ µn−1 (L(G)) ≤ . . . ≤ µ1 (L(G)).
Brouwer conjectures bounds on intermediate sums of eigenvalues ([BH11]):
Brouwer’s Conjecture. The following inequalities are valid:
t
X t+1
µi (L(G)) ≤ m(G) + , t = 1, . . . , n.
i=1
2
It is known to be valid for t = 1, 2, n − 1, n and trees [HMTR10]; unicyclic and bicyclic
graphs [DZ12]; regular graphs [May]; for n ≤ 10 it was checked by A.Brouwer using computer
[BH11]; and it was proved to hold asymptotically almost surely [Roc19].
Brouwer’s Conjecture is also closely related to the study of the Laplacian energy and other
conjectures. In particular, combining the present result with [HT15] gives:
Theorem 1.1. For each graph G, there exists a threshold graph Ĝ with the same number of nodes
and edges whose Laplacian energy is greater or equal to the Laplacian energy of G.
Therefore, fixed the number of nodes and edges, the maximum Laplacian energy is always
attained by a threshold graph.
Following [HMTR10], we assume 2 ≤ t ≤ n − 2. Denote by dv (G) the entry in the matrix D(G)
corresponding to the vertex v. In what follows, we take a vertex v such that the degree of v, dv (G)
is maximum and recall the following interlacing property of the Laplacian (see [Lot07]):
(2) µ1 (L(G)) ≥ µ1 (L(G − v)) ≥ µ2 (L(G)) ≥ µ2 (L(G − v)) ≥ . . . ≥ µt (L(G − v)) ≥ µt+1 (L(G)).
We divide the proof in two cases:
Case 1: Assume dv (G) + t + 1 ≥ n. We have:
t
X t
X
µi (L(G − v)) ≥ µi+1 (L(G))
i=1 i=1
t+1
X t+2
= µi (L(G)) − µ1 (L(G)) > + m(G) − µ1 (L(G))
i=1
2
t+1 t+1
= + t + 1 + m(G) − µ1 (L(G)) ≥ + m(G) − dv (G).
2 2
The first inequality follows from interlacing property (2), the second follows from (1) and the
last inequality follows from the hypothesis dv (G) + t + 1 ≥ n, since µ1 (L(G)) ≤ n (recall that
µi (L(G)) = n − µn−i+2 (L(G)) ≥ 0, i = 2, . . . , n − 1). Since m(G − v) = m(G) − dv (G), we conclude
that G − v violates the Conjecture for the sum of the eigenvalues up to t.
Case 2: If t + 1 + dv (G) < n, we consider the complement Ḡ. Recall the following result (see, for
example, [HMTR10, Theorem 6] or the proof of [Che19, Theorem 3.1], observing that Ḡ ¯ = G):
References
[BH11] A. E. Brouwer and W. H. Haemers. Spectra of graphs. Springer Science & Business Media, 2011.
[Che19] X. Chen. On Brouwer’s conjecture for the sum of k largest Laplacian eigenvalues of graphs. Linear
Algebra and its Applications, 2019.
[DZ12] Z. Du and B. Zhou. Upper bounds for the sum of Laplacian eigenvalues of graphs. Linear Algebra and
its Applications, 436(9):3672–3683, 2012.
[HMTR10] W. H. Haemers, A. Mohammadian, and B. Tayfeh-Rezaie. On the sum of Laplacian eigenvalues of
graphs. Linear Algebra and its Applications, 432(9):2214–2221, 2010.
[HT15] C. Helmberg and V. Trevisan. Threshold graphs of maximal Laplacian energy. Discrete Mathematics,
338(7):1075–1084, 2015.
[Lot07] Z. Lotker. Note on deleting a vertex and weak interlacing of the Laplacian spectrum. Electron. J. Linear
Algebra, 16:68–72, 2007.
[May] M. Mayank. On variants of the Grone-Merris conjecture. Master’s Thesis, Eindhoven University of
Technology, https://pure.tue.nl/ws/files/46996128/693554-1.pdf.
[Roc19] I. Rocha. Brouwer’s conjecture holds asymptotically almost surely. arXiv:1906.05368, 2019.
Instituto de Ciência e Tecnologia – Unifesp, Avenida Cesare Mansueto Giulio Lattes, 1201,
12247-014, São José dos Campos, SP, Brazil, and, Institute for Information Transmission Prob-
lems,, B. Karetnyi 19, Moscow, Russia
E-mail address: [email protected]
Instituto de Ciência e Tecnologia – Unifesp, Avenida Cesare Mansueto Giulio Lattes, 1201,
12247-014, São José dos Campos, SP, Brazil
E-mail address: [email protected]