5
$\begingroup$

Let $P$ be a property of graphs such that if a graph $G$ has $P$, then any graph obtained from $G$ by removal of a vertex also has $P$.

Let $g(n)$ be the maximum size of a graph of order $n$ having $P$. It can be shown that for $n>2$, $$g(n) \leq \left\lfloor \frac{n}{n-2}\cdot g(n-1)\right\rfloor.$$ I suspect this inequality is well known, and would like to have a reference rather than proving it from scratch. Any pointers will be appreciated!

$\endgroup$
4
  • 2
    $\begingroup$ What is the size of a graph? The number of its vertices? $\endgroup$
    – LSpice
    Commented Feb 3, 2023 at 0:50
  • 2
    $\begingroup$ Order is number of vertices, and size is number of edges. $\endgroup$
    – RobPratt
    Commented Feb 3, 2023 at 1:17
  • $\begingroup$ Lots of papers use "size" for the number of vertices. Although the number of edges is the "standard" definition, this is abused so often that one can't really be sure without checking. $\endgroup$ Commented Mar 10, 2023 at 1:53
  • 1
    $\begingroup$ Such properties are called "hereditary", and this inequality is one of the standard proofs of Turan theorem (on graphs without $K_r$). So I would search for key words "hereditary graph property" and "Turan theorem" $\endgroup$ Commented Mar 10, 2023 at 5:28

1 Answer 1

1
$\begingroup$

Consider a graph $G$ with $g(n)$ edges. $G$ has average degree $2g(n)/n$ so there is a vertex $v$ with degree at most $$\delta = \left\lfloor \frac{2g(n)}{n}\right\rfloor.$$ Since by assumption $G-v$ has the property, $g(n-1) \ge g(n)-\delta$. Now solve this inequality for $g(n)$.

Write $2g(n)=an+b$ for $0\le b\le n-1$, then you can expand the floor and solve for $a$. Put that back into $an+b$ and you get (for $n\ge 3$) $$ g(n) \le \frac{n g(n-1)-b}{n-2},$$ which implies the given inequality.

You might think that the $-b$ in the last numerator allows for a slightly tighter inequality, but I think it doesn't. At the moment I have forgotten the reasoning there.

$\endgroup$
2
  • $\begingroup$ I have a similar proof, but it's not what I asked. I suspect this inequality is well-known and has published source(s), which I'd like to refer to. Thank you anyway. $\endgroup$ Commented Mar 10, 2023 at 1:05
  • $\begingroup$ @MaxAlekseyev Hi Max, I see what you mean. It is surely present in one of the many papers on explicit extremal graphs but I don't know of one off-hand. I've been using it in my programs for at least 30 years and I'm pretty sure I just made it up myself. $\endgroup$ Commented Mar 10, 2023 at 1:11

You must log in to answer this question.

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