3

I'm currently working on the problem stated below, but i'm still kind of a beginner with proofs, so I would appreciate some tips (not a solution) on how to approach this proof.

A triangle is quasi-equilateral with error $e > 0$, if, for every angle $A$, $B$, $C$ in the triangle, we have that $\vert{A-60^{\circ}}\vert < e$, $\vert{B-60^{\circ}}\vert < e$ and $\vert{C-60^{\circ}}\vert < e $. Show that for every $e > 0$, there exist quasi-equilateral triangles with error $e$ that have vertices with integer coordinates in the plane.

I know how to prove that there are no equilateral triangles with integer coordinates, however, this looks much more difficult and I can't find a way to begin.

Blue
  • 75,673
Hlkwtz
  • 69
  • Probably by switching to a representation with vectors $\vec{OA}, \vec{OB}, \vec{OC}$ where $O$ is the origin... – Jean Marie Feb 16 '21 at 23:47
  • It seems impossible, as $e$ is a continuous intervalle in $ \mathbb{R}$ and integer coordinates are of order $\mathbb{N}\times \mathbb{N}$? – Toni Mhax Feb 17 '21 at 04:00
  • @ToniMhax What? The meaning of $e$ is very clearly stated, and (a) it's not a "continuous intervalle" (whatever that is supposed to be), and (b) it's not being compared to integer coordinates at any point. – user3482749 Feb 17 '21 at 13:36
  • @ToniMhax Sure, there are infinitely many such triangles. What makes you think that gives a bijection? In particular, why do you think that you get a different triangle for each $e$? – user3482749 Feb 17 '21 at 15:58
  • Well it is easy to make such error triangles in the continuum hypothesis $(\mathbb{R})$ take for example $A°=60+x$ and $B°=C°=60-\frac{x}{2}$ ($e=x$)... and these are one on one with integer coordinates triangles... Maybe something missed – Toni Mhax Feb 17 '21 at 16:25
  • @ToniMhax For information, there exists definitely a bijection between $\Bbb N$ and $\Bbb Q$ https://mathoverflow.net/a/200681/62193 , Because there exists also a bijection between $\Bbb Q \cap {(0,a)}$ and $\Bbb Q$, hence there exists une bijection between $N$ and $\Bbb Q \cap {(0,a)}$. – NN2 Feb 17 '21 at 16:35
  • FWIW, you can use the values $p/q$ which satisfy $p^2-3q^2=1$, which can be obtained from (alternating) convergents of the continued fraction of $\sqrt3$. I.e., $2/1, 7/4, 26/15, 97/56, 362/209, 1351/780, 5042/2911, 18817/10864, \dots$ – PM 2Ring Feb 18 '21 at 16:50

2 Answers2

2

Idea: set the origin $O$ be the circumcenter of the triangle $ABC$ and use the Pythagorean triple theorem.

Hint 1

Set the vertice $C$ on the axis $\vec{Ox'}$ (the coordinate of $C$ is $(-c,0)$ with $c\in \Bbb N^*)$. The origin $O$ is the circumcenter of the triangle $ABC$. The coordinate of points $A$ and $B$ are $(a,b)$ and $(a,-b)$ with $a,b \in \Bbb N^*$

Hint 2

The length $OA = OB = \sqrt{a^2+b^2}$ must in $\Bbb N^*$ because $OA = OB = OC$ with $C$ has integer coordinate on the axis $\vec{Ox}$

Hint 3

The solution of Pythagorean triple $a^2+b^2 = c^2$ (link) is, for example, $(a,b,c) = (m^2-n^2,2mn, m^2+n^2)$ with $m,n \in \Bbb N^*, m>n$.

Hint 4

Calculate the $\cos(\widehat{AOx}) = \frac{m^2-n^2}{m^2+n^2}$ which must be in the interval $(v_1,v_2)=\left(\cos(\frac{\pi}{3}+\epsilon), \cos(\frac{\pi}{3}-\epsilon) \right)$.

Hint 5

Prove there always exists a rational number $t\in \Bbb Q^+$ such that $$\sqrt{\frac{1+v_1}{1-v_1}} <t < \sqrt{\frac{1+v_2}{1-v_2}}$$ with $(v_1,v_2)$ defined in the hint 4. The couple $(m,n)$ will be defined by $t = \frac{m}{n}$. Hence, you deduce the value of $(a,b,c)$ from $(m,n)$

NN2
  • 15,892
1

First hint:

It suffices to show this for rational coordinates (then we can just scale up by the largest common denominator).

Second hint:

Since we're rescaling anyway, we might as well take two of our points to be exactly one unit apart: fix them and call them $x$ and $y$.

Third hint:

Consider the set $Z$ of all points $z$ that make an $e$-quasiequilateral triangle.

Fourth hint:

$Z$ is an open subset of the plane.

Fifth hint (if you're struggling with proving the fourth hint):

Take such a point $z$. Then let the angles of the triangle so formed be $A(z), B(z), C(z)$ (for future convenience, I'll make those the angles at $x,y,z$, in that order), and define $f(z) := \max(|A - 60^\circ|,|B - 60^\circ|, |C-60^\circ|)$. Show that $f$ is continuous, and you're done.

Sixth hint (if you're struggling with proving the fifth hint above):

It suffices to show that each of $A,B,C$ are continuous. To do this, take some $z'$ near $z$, and consider the triangle $x,z,z'$.

The end of the proof above:

The angle at $x$ of this triangle is $|A(z) - A(z')|$ by definition, and by the sine rule (with $\theta,\varphi$ the angles at $z,z'$ of this triangle), $|A(z) - A(z')| = |zz'|\frac{\theta}{|xz'|} = |z - z'|\frac{\varphi}{|xz|}$. But neither $\varphi$ nor $\theta$ can be more than $180^\circ$, so this is bounded above by $180^\circ|zz'|\left(\frac{1}{|xz|} + \frac{1}{|xz'|}\right)$. Note that (for $e$ smaller than $30^\circ$ - the final result for larger $e$ is easy): $z$ cannot lie within a distance of $\frac{1}{2}$ of $x$, as all points within that disk give angles at $y$ of no more than $30^\circ$, and similarly for $z'$. By the triangle rule, we have $|xz| \geq |xz'| - |zz'| \geq \frac{1}{2} - |zz'|$, so for $|zz'| < \frac{1}{4}$, we have $|xz| \geq \frac{1}{4}$, and so $\frac{1}{|xz|} \leq 4$. Thus, $|A(z) - A(z')| \leq 180^\circ|zz'|\left(4 + 2\right) = 1080^\circ|zz'| \to 0$ as $|zz'|\to 0$, so $A$ is continuous.

The proof for $B$ is the same as for $A$ with $y$ replacing $x$ throughout. For $C$, apply the cosine rule to the triangles $xyz$ and $xyz'$ to obtain $2|xz||yz|C(z) = |xz|^2 + |yz|^2 - 1$ and similarly for $z'$, and either combine these to obtain $$|C(z) - C(z')| =\left|\dfrac{|xz|^2 + |yz|^2 - 1}{2|xz||yz|} - \dfrac{|xz'|^2 + |yz'|^2 - 1}{2|xz'||yz'|}\right|$$ and proceed roughly as above with a moderately ridiculous sequence of approximations, or just notice that the functions $w \mapsto |xw|$ and $w \mapsto |yw|$ are continuous and neither is zero anywhere near either $z$ or $z'$, and hence $C$ is continuous by repeated applications of the product, sum, and quotient rules for continuity.

Fifth hint (if you're struggling with finishing after proving the fourth hint):

Since the rationals are dense in the reals, $\mathbb{Q}^2$ is dense in $\mathbb{R^2}$ (think of squares inside circles), and so there's a rational point in every open subset of $\mathbb{R^2}$, and in particular there's one in the above, so scaling everything by the lowest common multiple of the denominator gives an integer solution.

user3482749
  • 6,660
  • there may exists a sequence of rational points $q_n$ which convert to every point $z$ of the set $Z$. But it doesn't mean that there exists a $n_0$ such that $q_{n_0} \in Z$. In other words, the sequence $q_n$ may contain no element in $Z$ – NN2 Feb 17 '21 at 02:42
  • @NN2 $Z$ is open. – user3482749 Feb 17 '21 at 10:58
  • From the information $Z$ is open (or closed), we can say nothing whether $Z$ contains rational number ($\in \Bbb Q$) or not. In other words, $Z$ may contain only irrational number ($\not\in \mathbb{Q}$) – NN2 Feb 17 '21 at 11:24
  • @NN2 That is simply untrue. Every open set in the plane contains rational points. (And more to your original point, it is not possible for a sequence of points contained in the complement of an open set can converge to a point in the open set, since the complement is closed.) – user3482749 Feb 17 '21 at 12:25
  • First, you think "$Z$ is open" but we have no proof about that. And prove this statement is more difficult than the original problem. Besides, from this statement, as you wrote in your comment above, $Z$ contains rational points $\in \Bbb Q$. Hence, you can deduce directly that there are "equilateral triangles with integer coordinates", right? The,, your proof can be stopped at the hint 3. – NN2 Feb 17 '21 at 12:36
  • Second, Here is the definition of open set https://en.wikipedia.org/wiki/Open_set#Metric_space "A subset U of a metric space (M, d) is called open if, given any point x in U, there exists a real number ε > 0 such that, given any point $y\in M$ satisfying d(x, y) < ε, y also belongs to U" The point y here can be non-rational ($\not\in \Bbb Q$) – NN2 Feb 17 '21 at 12:36
  • @NN2 Again: every open subset of the plane contains rational points. The proof of this is very easy, and a standard first-year undergraduate exercise (hint: truncated decimal expansions). – user3482749 Feb 17 '21 at 12:42
  • The problem here is $Z$ is not an open subset of the plane but "set of all points z that make an e-quasiequilateral triangle"(1) (your third hint) and you assume later that " $Z$ is an open subset of the plane" (2) (your fourth hint). Hence, suppose your assumption (2) is correct, $Z$ will be "an open set of the plane that contains all points z that make an e-quasiequilateral triangle" and $Z$ is NOT simply "an open subset of the plane". $\implies$ you can't conclude $Z$ contains rational points. – NN2 Feb 17 '21 at 12:54
  • @NN2 No, we prove that $Z$ is an open subset of the plane. That's what the following four spoilers are there for. Any nonempty open subset of the plane contains rational points, regardless of any other conditions that may be put on it. – user3482749 Feb 17 '21 at 13:00
  • So, please prove explicitly in your answer : 1/ your assumption (2) (in your third hint): $Z$ is an open subset of the plane and 2/ why $Z$ contains rational point. – NN2 Feb 17 '21 at 13:04
  • (1) is proved in the answer: it's in hints four and five (you know, the ones that start with ("if you're struggling with proving the fourth hint"). (2) is elementary. And frankly: the OP explicitly doesn't want to be handed a full solution, which I've already pretty much done. Please stop making ridiculous comments on subjects which you clearly do not understand and have no intention of learning about. It's rude and it wastes my time. – user3482749 Feb 17 '21 at 13:07
  • I bet you can't prove what you said. And even if you can, your answer need to be stopped at the third hint, as I wrote in the 5th comment. PS: you are not as strong in math as you think you are, and worst, you don't know who I am. PS2: I stop here. I don't want to waste my time. – NN2 Feb 17 '21 at 13:18
  • @NN2 Again, the proof of the first point is very literally in the answer already. The proof of the second is trivial (prove it for $\mathbb{R}$ first, where it's immediate from truncated decimal expansions, as I already mentioned). I have no clue what you mean about stopping at some hint (and, more to the point, they're hints). Literally everything that I have said here is entirely correct, and I know the only relevant thing about you: your knowledge of mathematics is utterly inadequate for this discussion. (And that's what it is, so I'm moving it to chat). – user3482749 Feb 17 '21 at 13:32