6
$\begingroup$

We have a $d$-regular bipartite graph $G = (X,Y,E)$ with $|X| = |Y| = n$ and $|E| = nd$.

What is an upper bound on the number of perfect matchings of $G$?

$\endgroup$

2 Answers 2

12
$\begingroup$

To elaborate on Tony's answer, for a graph $G$ with an even number of vertices and degree sequence $d_1,\dots,d_n$ the number of matchings in $G$ is at most $\prod_{i=1}^n (d_i!)^{\frac{1}{2d_i}}$, with equality achieved only at graphs which are a disjoint union of complete bipartite graphs. The proof is a consequence of the Bregman inequality for permanents. See "The maximum number of perfect matchings in graphs with a given degree sequence", by Alon and Friedland

Notice that there is a conjecture that something similar holds for the number of matchings of size $t$ where $t$ is not necessarily half the number of vertices. It is conjectured that this is also maximized at disjoint union of complete bipartite graphs, but is open in general (even when we restrict to only bipartite graphs).

$\endgroup$
2
  • $\begingroup$ i just want to have a upper bound for the generally $d$-regular bipartite graph <math>$G = (X,Y,E)$</math> with <math>$|X| = |Y| = n$</math> and <math>$|E| = nd$</math>. <math>$n$</math> may be even or odd or whatever. I just want a formula that is true for all those case. <math>$#M \leq ....$</math> $\endgroup$
    – pnaky
    Commented May 21, 2013 at 10:04
  • $\begingroup$ @pnaky: Gjergji's answer is close to the best you can do. If $d$ is a divisor of $n$, it is exactly the best you can do. Just set all the $d_i$s equal to $d$. $\endgroup$ Commented May 21, 2013 at 14:22
4
$\begingroup$

In the case that $2d$ divides $n$, one can take $G$ to be $\frac{n}{2d}$ disjoint copies of $K_{d,d}$. This graph has $(d!)^{n/2d}$ perfect matchings, and as Gjergji's answer shows, this is the worst case.

$\endgroup$
6
  • $\begingroup$ it is only for the case that d divide n? $\endgroup$
    – pnaky
    Commented May 24, 2013 at 12:28
  • $\begingroup$ For your special case, your degree sequence is $d,d,\dots. d$, so Gjergji's answer shows that there are at most $(d!)^{n/d}$ perfect matchings. This works even if $d$ does not divide $n$. My answer shows examples where you can achieve this bound. Gjergji's answer says that these are the only graphs that achieve the upper bound. This is really the end of the story. Your question has been completely answered. $\endgroup$
    – Tony Huynh
    Commented May 24, 2013 at 19:19
  • $\begingroup$ Don't you mean "one can take $G$ to be $\frac{n}{2d}$ disjoint copies of $K_{d,d}$"? $\endgroup$
    – Untitled
    Commented Oct 9, 2014 at 7:21
  • $\begingroup$ Yes, thanks. I edited accordingly. $\endgroup$
    – Tony Huynh
    Commented Oct 9, 2014 at 9:34
  • 1
    $\begingroup$ @Evgenii.Balai Yes. Thanks! I edited the answer. $\endgroup$
    – Tony Huynh
    Commented Sep 8, 2021 at 1:30

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .