6
$\begingroup$

Let $P\in{\mathbb R}[X]$ be a polynomial and $[a,b]$ be a bounded interval. Of course, the graph of $P$ is an algebraic set. I am interested in the lower convex envelope $\bar P$ of $P|_{[a,b]}$. It seems to me that its graph is semialgebraic. However I am not at all familiar with real algebraic geometry. Is there a clear justification ?

At least, the graph of $\bar P$ is the union of arcs of the graph of $P$, together with bi-tangents. The latter are obtained by an elimination in a system of two algebraic equations.

Same question when replacing $P$ by $\min(P,Q)|_{[a,b]}$ where $P$ and $Q$ are two polynomials.

Perhaps my question looks obvious for specialists, in which case I should be happy with a proper reference.

$\endgroup$
1
  • 1
    $\begingroup$ One dimension up: Ranestad, Kristian, and Bernd Sturmfels. "On the convex hull of a space curve." Advances in Geometry 12, no. 1 (2012): 157-178. arxiv abs: "The boundary of the convex hull of a compact algebraic curve in real 3-space defines a real algebraic surface." $\endgroup$ May 12, 2021 at 12:36

1 Answer 1

2
$\begingroup$

The answer is yes, but not in the most exciting way. The idea is to write the envelope in the first-order logic and then kill the problem with quantifier elimination for the theory of real-closed fields.

For example, we can describe the envelope as the set of points $(z, w)$ such that for all $x, y\in [a, b]$ we have $(z, w)$ is above or on the segment between $(x, P(x))$ and $(y, P(y))$ but for any $t > 0$ there are $x, y$ such that this is not true for $(z, w-t)$. One can write it out as a first-order formula of quantifier degree 6 or so. Then, running the elimination algorithm and then reducing the result to a disjunctive normal form we get the expression as a semi-algebraic set (small caveat here is that one should treat $a, b$ and the coefficients of $P$ as free parameters and the resulting semi-algebraic set of course depends on these parameters). Note that one can also bound the number of pieces of this semi-algebraic set in terms of the degree of $P$ only (something like $2^{2^{O(deg P)}}$ or whatever the complexity of the quantifier elimination algorithm).

Of course, this strategy can be easily modifed to two or more polynomials and many other situations.

$\endgroup$
2
  • $\begingroup$ Thank you. What is a practical reference, to learn about these techniques ? $\endgroup$ May 13, 2021 at 6:09
  • $\begingroup$ @DenisSerre my problem is that I've learned these things in my first year undergraduate logic class so I've long forgoten all the references (and all the proofs). The only link I can give you is Tarski-Seidenberg Theorem which states the result about the elimination of quantifiers, sorry. $\endgroup$ May 13, 2021 at 14:12

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.