Does there exist a sequence of strongly regular graphs with parameters $(n,d,\lambda,\mu)$ (so every pair of adjacent vertices have $\lambda$ common neighbours, and every pair of non-adjacent ones have $\mu$ common neighbours), with $d - \mu = \Omega(n)$ and $\mu - \lambda = \Omega(n)$?
-
$\begingroup$ Could you define $\ n\ $ and $\ d$? $\endgroup$– Wlod AACommented Jan 26, 2023 at 8:31
-
$\begingroup$ $n$ is the number of vertices and $d$ is degree/valency of each vertex. $\endgroup$– Gordon RoyleCommented Jan 26, 2023 at 8:47
-
$\begingroup$ I meant the formulation of the problem. $\endgroup$– Wlod AACommented Jan 26, 2023 at 8:51
-
$\begingroup$ It seems that the last two equations can be replace simply by one: $\ 2\cdot\mu= \lambda + d.$ $\endgroup$– Wlod AACommented Jan 26, 2023 at 9:03
-
1$\begingroup$ @WlodAA it is a less known as $O(n)$ but quite useful notation for "at least constant times $n$" $\endgroup$– Fedor PetrovCommented Jan 26, 2023 at 9:31
1 Answer
I think, the answer is negative (I expect that by $\Omega(n)$ you mean something positive of order $n$). Denote $\mu-\lambda=k,d-\mu=\ell$.
The eigenvalues of the adjacency matrix of such a graph $G$ are $d$ and $t_{1,2}$ which are the roots of a quadratic equation $t^2+kt-\ell=0$. The multiplicities $n_{1},n_2$ enjoy the equations $n_1+n_2=n-1$, $n_1t_1+n_2t_2=-d$. Consider two cases.
$n_1=n_2$. Then $n_1=n_2=(n-1)/2$ and $-d=(t_1+t_2)\frac{n-1}2=-k\frac{n-1}2$ which is too large (of order $n^2$).
$n_1\ne n_2$. Then from $t_1+t_2=-k$ and $n_1t_1+n_2t_2=-d$ we get $t_2(n_2-n_1)=kn_1-d$ and so $t_2$ is rational, i.e., the discriminant $k^2+4\ell$ of our quadratic equation is a perfect square, let it be equal to $k^2+4\ell=(k+2x)^2$, then $x^2+kx=\ell$. It yields that $x$ is bounded. So, the roots are $t_2=x$ and $t_1=-k-x$. And $n_1t_1+n_2t_2=-d$ reads as $n_2x+(n-1-n_2)(-k-x)=-d$, so $n_2(k+2x)=(n-1)(k+x)-d$ and $n_2=(n-1)-(n-1)\cdot\frac{x}{k+2x}-\frac{d}{k+2x}=n-O(1)$. In other words, the matrix $A-xI$ has bounded rank. But by Ramsey theorem it contains a large minor with $-x$ along diagonal and either 0's or 1's outside diagonal. This minor does not have full rank only if $x\in \{0,-1\}$ which is impossible.
-
$\begingroup$ For what it worth, if $k=\mu-\lambda$ is negative but of order $n$, this is possible only for complete graphs: it remains to consider the case $x=-1$, which reads as $\lambda=d-1$ that holds only for complete graphs. $\endgroup$ Commented Jan 27, 2023 at 4:52