14
$\begingroup$

Given a triangle T with sides a, b, and c, describe its "fitting set," the set of all points (x,y,z) in 3-dimensions for which a triangle with sides x, y, z exists that fits in T.

Such a set lies in the positive octant, is star-shaped with respect to the origin, and probably is a polyhedron; but it seems difficult to describe.

Improve this question
$\endgroup$
6
  • 1
    $\begingroup$ Welcome to MO! I tried to make your title a bit more informative, hope that's all right. $\endgroup$ – j.c. Oct 17 '17 at 17:58
  • 3
    $\begingroup$ This problem can be found in the book "100 problems" by Hugo Steinhaus, but without solution. $\endgroup$ – user82588 Oct 17 '17 at 18:22
  • 3
    $\begingroup$ This space is not convex. Consider a triangle with sides $(1,1,2-\epsilon_1)$ for $0 < \epsilon_1 <1$. For small $\epsilon_2$, a triangle with sides $(1,1,\epsilon_2)$ will fit inside, but an equilateral triangle with sides $(1,1,1)$ will not. Another way to see failure of convexity is that if triangles of sides $(3/2,3/2,\epsilon), (3/2,\epsilon, 3/2), (\epsilon,3/2,3/2)$ fit, this does not imply an equilateral $(1,1,1)$ fits. $\endgroup$ – Douglas Zare Oct 17 '17 at 18:29
  • 2
    $\begingroup$ It seems the case of two right angled triangles has been considered in the literature, see J.H. Jepsen, V. Vulpe "Fitting One Right Triangle in Another" maa.org/sites/default/files/jepsen6-200737759.pdf $\endgroup$ – user82588 Oct 17 '17 at 18:51
  • 1
    $\begingroup$ @j.c. The isosceles triangle with legs of length $1$ and apex angle $30^\circ$ doesn't fit into the equilateral $(1,1,1)$. $\endgroup$ – MTyson Oct 17 '17 at 20:45
17
$\begingroup$

Here's the abstract of K.A. Post, "Triangle in a triangle: On a problem of Steinhaus", Geom Dedicata (1993) 45: 115; this paper was cited in the one given in the comment by Nemo.

A necessary and sufficient condition on the sides $p, q, r$ of a triangle $PQR$ and the sides $a, b, c$ of a triangle $ABC$ in order that $ABC$ contains a congruent copy of $PQR$ is the following: At least one of the 18 inequalities obtained by cyclic permutation of $\{a, b, c\}$ and arbitrary permutation of $\{p, q, r\}$ in the formula: \begin{array}{l} Max\{ F(q^2 + r^2 - p^2 ), F'(b^2 + c^2 - a^2 )\} \\ + Max\{ F(p^2 + r^2 - q^2 ), F'(a^2 + c^2 - b^2 )\} \le 2Fcr \\ \end{array} is satisfied. In this formula $F$ and $F'$ denote the surface areas of the triangles, i.e. \begin{array}{l} F = {\textstyle{1 \over 4}}(2a^2 b^2 + 2b^2 c^2 + 2c^2 a^2 - a^4 - b^4 - c^4 )^{1/2} \\ F' = {\textstyle{1 \over 4}}(2p^2 q^2 + 2q^2 r^2 + 2r^2 p^2 - p^4 - q^4 - r^4 )^{1/2} . \\ \end{array}

Here's a picture for the case $a=b=c=1$, which I plotted in this SageMath Jupyter notebook. Note that in this case we need only consider 3 of the above inequalities (plus the triangle inequalities):

region for equilateral triangle

Improve this answer
$\endgroup$

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.