Let $X \subset \mathbb R^d$ be a simplex and $x \notin X$

Prove that there exists a unique face $F \subseteq X $ with minimal dimension $\dim(F)=\dim(aff(F))$ such that $x \in K_F$ where $K_F:=\{x+\lambda(y-x):\lambda \ge 0 \;x\in F,y\in X \}$ denotes the tangent cone of $F$

Note: $\mathrm{aff}(F)$ is supposed to be the affine hull.

Would appreciate any help.

  • $\begingroup$ What are your approaches? Also where does the "tangent cone" from the title appear in the question? $\endgroup$ – M. Winter Jun 8 '18 at 21:48
  • $\begingroup$ @M.Winter The tangent cone in the title is $K_F$. At least that's my definition of a tangent cone given a face $F$ of $X$ $\endgroup$ – XPenguen Jun 8 '18 at 21:53

Here is one way to prove this by contradiction.

Let $F_1\neq F_2$ be two faces of $X$ that satisfy your constraints. Because $x\in K_{F_i}$, there are $p_i\in F_i$ and $q_i\in X$ such that $x$ belongs to the ray from $p_i$ through $q_i$. The properties of $p_i$ and $q_i$ are enough to derive a contradiction.

Hint 1:

$X$ is a simplex and is thus convex, it must fully contain the segment between $q_1$ and $q_2$.

Hint 2:

Because $x\notin X$, thus $x$ cannot be in the segment between $q_1$ and $q_2$. You can look at rays from $x$ and through points of that segment.

Hint 3:

The existence of $p_1$ and $p_2$ guarantee that for each rays mentioned in hint 2, the last point of $X$ that is met by the ray is not in the segment $[q_1q_2]$, but somewhere further. Look at those points.

The three hints above should be more than enough to get you started. If you're still struggling, here are some more.

Hint 4:

The points from hint 3 form a continuous path, let's call it $\gamma$. This path $\gamma$ entirely lies in the boundary of $X$. And each boundary point of $X$ belongs to some face of $X$ of dimension (strictly) less than $d$.

Hint 5:

If you look at the points in $\gamma$, you obtain a sequence of faces of $X$ that starts from $F_1$ and ends with $F_2$. Because you go from one face to another face of the same dimension, at some point $\gamma$ must cross the boundary of $F_1$ and $F_2$. What property do those boundary points (of $F_1$/$F_2$) have?

The last hint is one step away from the conclusion.

Hint 6:

Let $p$ be one of the points mentioned in hint 5. Point $p$ must belong to another face, $F_3$, whose dimension is (strictly) less than those of $F_1/F_2$. And it also belongs to $\gamma$.

  • $\begingroup$ Thanks for the reply. I just have one question. While this shows the uniqueness but what about existence ? I mean why does $x \in K_F$ for some face $F$ of $X$ ? $\endgroup$ – XPenguen Jun 9 '18 at 16:46
  • $\begingroup$ @XPenguen $X$ is the (unique) $d$-dimensional face of $X$, and $F_X=\mathbb R^d$. If you exclude $X$ as a face, the result is false. $\endgroup$ – N.Bach Jun 9 '18 at 16:48
  • $\begingroup$ Now I feel stupid :P Thanks for taking your time. I appreciate it $\endgroup$ – XPenguen Jun 9 '18 at 16:51

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.