11
$\begingroup$

Given $n$ quadratic polynomials in $n-1$ variables over the complex field, what is the maximum number of common zeros? Can we have $2^{n-1}-1$ common zeros? I assume that a linear combination of the polynomials is always different from zero and the number of zeros is finite.

With $4$ polynomials, the maximum is not smaller than $6$. Using a projective space $(x_1,x_2,x_3,x_4)$, an example with $6$ roots is given by the polynomials.
$P_1=x_1 x_2$,
$P_2=x_1 x_3$,
$P_3=L_1 x_2+L_2 x_3$,
$P_4=(\text{a general quadratic polynomial})$,
$L_k$ being general linear polynomials. Indeed, if $x_1=0$, then the first two polynomials are equal to zero and the remaining two polynomials in $x_2, x_3, x_4$ give four roots, which are distinct for a general choice of $L_k$ and $P_4$. If $x_2=x_3=0$, then the first three polynomials are equal to zero and the fourth polynomial in $x_1,x_4$ gives 2 additional roots.

This construction has a natural generalisation to $n$ polynomials, giving $2^{n-2}+2^{n-3}$ roots, which is about $3/4$ of the desired bound $2^{n-1}-1$.

$\endgroup$
10
  • 1
    $\begingroup$ By "polynomials over an $n-1$-dimensional space," do you mean polynomials in $n-1$ variables? Are we working over the reals, the complex numbers, some other field? What's an "intersection point" of two or more polynomials? Is it a common zero? $\endgroup$ – Gerry Myerson Sep 26 '20 at 1:22
  • $\begingroup$ Thanks for the comment. I modified the question accordingly. $\endgroup$ – Alberto Montina Sep 26 '20 at 7:07
  • 2
    $\begingroup$ I think this is an interesting question, let me rewrite it more geometrically: can we find $n+1$ quadrics in $\mathbb{P}^n_{\Bbb{C}}$ which intersect exactly in $2^n-1$ distinct points? For example, 4 quadrics in $\mathbb{P}^3$ intersecting in 7 points? $\endgroup$ – abx Sep 26 '20 at 7:17
  • 1
    $\begingroup$ You need to make some 'general position' assumption to get a good answer. For example, when $n=5$, there are 6 linearly independent quadratic polynomials in 4 variables that vanish identically on the rational curve $(t,t^2,t^3,t^4)$, so the number of common zeros of $5$ linearly independent quadratic polynomials in 4 variables can be infinite. $\endgroup$ – Robert Bryant Sep 27 '20 at 10:37
  • 1
    $\begingroup$ You are right. I was assuming implicitly that the number of zeros is finite. Let me state explicitly in the question $\endgroup$ – Alberto Montina Sep 27 '20 at 15:57
14
$\begingroup$

There is a bound for the multiplicity of a (homogenous) almost complete intersection in Theorem 1 of this paper by Engheta. In case of $n$ quadrics in $n-1$ variables, that bound is $2^{n-1}-(n-2)$. So for $n\geq 4$, you can not get $2^{n-1}-1$.

(In Theorem 1 there was a condition that the first $n-1$ generators form a complete intersection, but I don't think that this is serious in your case, one could always make a linear change to turn the first $n-1$ generators into a regular sequence, as the whole ideal has maximal height)

There are further improvements and examples of Engheta's bound in this work and this very recent work.

$\endgroup$
3
  • $\begingroup$ Is the bound sharp? I looked through the papers and did not see examples in the form here. $\endgroup$ – Matt F. Sep 28 '20 at 21:21
  • $\begingroup$ @MattF.: I don't think it is sharp for large $n$, but I am not an expert on this topic. $\endgroup$ – Hailong Dao Sep 28 '20 at 23:09
  • 1
    $\begingroup$ Btw, I noted that bound given in the last paper cited by @HailongDao gives 3,6,12,24,51 common zeros for 3,4,5,6,7 polynomials. The example in my question gives 3,6,12,24,48 common zeros. Thus, the example is optimal for up to 6 polynomials. If the next improvement of the bound will give 48 common zeros for n=7, I will start thinking that my example is optimal. $\endgroup$ – Alberto Montina Oct 19 '20 at 13:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.