Let $A_{i, j}, B_{i, j}, C, D \in \mathbb{Q}$, and consider the following pair of equations $$ A_{1, 1} x_1 y_1 + A_{1, 2} x_1 y_2 + A_{2, 1} x_2 y_1 + A_{2, 2} x_2 y_2 = C $$ $$ B_{1, 1} x_1 y_1 + B_{1, 2} x_1 y_2 + B_{2, 1} x_2 y_1 + B_{2, 2} x_2 y_2 = D. $$ I was interested in figuring out if this system of equations satisfies the local to global principle. In other words, suppose I can find a solution $(\mathbf{x}, \mathbf{y})$ of the system in $\mathbb{R}^4$ and in $\mathbb{Q}^4_p$ for all primes $p$, then there exists a solution $(\mathbf{x}, \mathbf{y}) \in \mathbb{Q}^4$.

I would appreciate any comments or suggestions or counterexamples. Thank you very much!

  • 2
    $\begingroup$ Intersections of two quadrics in $\mathbb{P}^4$ are del Pezzo surfaces of degree 4. For those there are counter-examples to the Hasse principle going back to Birch and Swinnerton-Dyer. I don't know if any known example has this particular form. There are more knowledgable people active here who will give a better answer. $\endgroup$ Commented May 1, 2020 at 7:28
  • 1
    $\begingroup$ Actually, what the OP is looking at is the intersection of 2 planes with a smooth quadric $Q\subset \mathbb{P}^3$ -- that is, the intersection of a line with $Q$. This is an ordinary second degree equation, which certainly satisfies the local to global principle. $\endgroup$
    – abx
    Commented May 1, 2020 at 8:06
  • 2
    $\begingroup$ @abx, I had the same thought until I realised that C and D are not necessarily zero... the projective closure is a dP4 with four singular points at infinity. $\endgroup$ Commented May 1, 2020 at 8:15
  • 1
    $\begingroup$ May be I'm extremely stupid and missing something obvious, but jf we just pick $y$'s at random we get two linear equations in $x$. If the determinant is zero for all $y$'s then either equations are proportional in which case we have only one equation which always have solutions or both of our equations factor as $(ax_1+bx_2)(cy_1+dy_2)$ and moreover $a$ and $b$ are the same for them. But doing the same thing switching $x$ and $y$ shows that $c$ and $d$ are the same as well and we are back to the proportionality case. All in all solution either exists or not in all fields of characteristics $0$ $\endgroup$ Commented May 1, 2020 at 9:58
  • 2
    $\begingroup$ I think that Aleksei may have a point; The blow-up of $X$ in a rational point not on a line is a cubic surface with $2$ skew lines; these are all well-known to be rational, so it seems that $X$ is actually rational, rather than just unirational. This means that there should actually be a complete parametrisation of all the solutions, which I expect is what Aleksei's approach is leading to (writing down an explicit complete parametrisation). $\endgroup$ Commented May 1, 2020 at 11:13

1 Answer 1


In fact such system of equations always have a solution (at least if the coefficients are general).

Let $X$ denote the closure of your variety in $\mathbb{P}^4_{\mathbb{Q}}$. Explicitly: \begin{align*} &A_{1, 1} x_1 y_1 + A_{1, 2} x_1 y_2 + A_{2, 1} x_2 y_1 + A_{2, 2} x_2 y_2 = C z^2 \\ &B_{1, 1} x_1 y_1 + B_{1, 2} x_1 y_2 + B_{2, 1} x_2 y_1 + B_{2, 2} x_2 y_2 = D z^2. \end{align*} As remarked in comments, this defines a del Pezzo surface of degree $4$ with $4$ singular points at infinity (for general choices of coefficients).

This contains the $2$ lines at infinity $$x_1=x_2=z=0, \quad y_1=y_2=z=0.$$ In particular $X$ contains a smooth rational point (choose a point on a line which is not one of the $4$ singular points). But it is well-known that a singular del Pezzo surface of degree $4$ with a smooth rational point is unirational (see e.g. [1, Theorem B]). Thus the rational points are Zariski dense, so there is a rational point with $z \neq 0$, which gives a solution to the original equations.

Here by "general", to apply [1, Theorem B] one needs to know that $X$ is irreducible with only finitely many singular points and is not a cone.

[1] Coray, Tsfasman - Arithmetic on singular del Pezzo surfaces.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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