5
$\begingroup$

Let $x$ and $y$ be real numbers. Prove that if $x\leq y + \epsilon$ for every positive real number $\epsilon$, then $x\leq y$.

I would like a hint as to how to prove this. Thank you. Pictorial proof would be nice too.


So this is how I word-smithed the answer given by the author of my accepted answer:

We will proceed by showing the contrapositve. We are given that $$x\leq y+\epsilon \implies x\leq y,$$ so the contrapositive is resolved as $$x > y \implies x>y+\epsilon,$$ or by substituting $x-y = \omega$ simply $$\omega > 0 \implies \omega > \epsilon,$$ but if $\omega >0$, then $\epsilon = \frac{\omega}{2} > 0$; however, $\omega \leq \frac{\omega}{2}$ is false. Therefore, we can assert that for all $\epsilon > 0$ $$\omega\leq\epsilon \implies \omega \leq 0;$$ quod erat demonstrandum.

Here is my pictorial representation:

enter image description here

$\endgroup$
4
  • 2
    $\begingroup$ Suppose it weren't so, i.e. $y < x$. Look for an $\epsilon$ violating the constraints. $\endgroup$ Jul 7, 2013 at 22:55
  • 4
    $\begingroup$ What is "discrete" about this? $\endgroup$
    – Pedro
    Jul 7, 2013 at 22:57
  • $\begingroup$ It is from a discrete mathematics textbook. Should a student studying discrete mathematics also have this problem in their discrete mathematics textbook, then they will find this of interest. $\endgroup$
    – Trancot
    Jul 7, 2013 at 22:57
  • 1
    $\begingroup$ @Trancot I have seen textbooks for linear give a definition and some properties of injective functions, should a question of this kind have a title "linear algebra:If we have an injective function bla bla"? $\endgroup$
    – clark
    Jul 7, 2013 at 23:01

2 Answers 2

11
$\begingroup$

The claim is equivalent to showing that if $\omega\leq \epsilon$ for each $\epsilon >0$, then $\omega\leq 0$.

But, if $\omega>0$, then $\epsilon=\frac \omega 2 >0$ and $\omega \leq \frac\omega 2$ does not hold. Having proven the contrapositive, we can assert that $$(\forall\epsilon >0\;:\;\omega\leq\epsilon )\implies \omega \leq 0$$

Now let $\omega =x-y$.

Pictorially If for any $\color{green}{\epsilon >0}$ we choose, $\omega$ is to the left of $\color{green}{\epsilon}$, it must be the case $\omega$ is to the left of the green bar, that is, on the red side $\color{red}{\omega <0}$ (strictly negative numbers) or that it is on the breaking point, that is $\color{orange}{\omega=0}$.

enter image description here

ADD The contrapositive of the assertion is $$\omega >0\implies (\exists \epsilon >0:\omega\not\leq \epsilon)$$ or, which is the same, $$\omega >0\implies (\exists \epsilon >0:\omega> \epsilon)$$

We proved the contrapositive with $\epsilon =\omega /2$.

$\endgroup$
6
  • $\begingroup$ Should not $\omega \leq 0$ be $\omega < 0$? Actually, no I guess not. $\endgroup$
    – Trancot
    Jul 7, 2013 at 23:08
  • $\begingroup$ The claim is true if $\omega =0$. See how the proof of the contrapositive breaks down in such case, @Trancot. $\endgroup$
    – Pedro
    Jul 7, 2013 at 23:09
  • $\begingroup$ See my most recent edit. $\endgroup$
    – Trancot
    Jul 7, 2013 at 23:26
  • $\begingroup$ Fancy answer...+1. $\endgroup$
    – Julien
    Jul 7, 2013 at 23:34
  • $\begingroup$ @Trancot I added something. $\endgroup$
    – Pedro
    Jul 7, 2013 at 23:36
3
$\begingroup$

Consider $x,y$ be universally quantified, then

$$\forall \epsilon\ (x\leq y+\epsilon) \implies x\leq y$$ $$\neg (x\leq y)\implies \neg\forall \epsilon\ (x\leq y+\epsilon)$$ $$\neg (x\leq y)\implies \exists \epsilon\ \neg(x\leq y+\epsilon)$$ $$x>y\implies \exists \epsilon\ (x>y+\epsilon)$$

For $x>y$, we know $x>\tfrac{x+y}{2}=y+\tfrac{x-y}{2}$ and $\tfrac{x-y}{2}>0$, so it fits the bill.

$\endgroup$
6
  • $\begingroup$ @Angelo I edited it back out of principle. Can you please not randomly go and edit my answer without a warning? Yes, OP speaks of the situation the quantifier over $\epsilon$ goes over just positive numbers. $\endgroup$
    – Nikolaj-K
    May 21 at 14:58
  • $\begingroup$ Really I do not understand what you mean. Anyway, your answer is correct and I like it, in fact I upvoted it. $\endgroup$
    – Angelo
    May 21 at 17:45
  • $\begingroup$ @Angelo I don't think it's a common habit on this site to edit other people's words, unless it's a buggy or unreadable opening question $\endgroup$
    – Nikolaj-K
    May 22 at 16:32
  • $\begingroup$ @Angelo Grumpy un-vote now? Well it's not a Wiki with textbook information, it's people posting text in their own name. $\endgroup$
    – Nikolaj-K
    May 22 at 17:03
  • $\begingroup$ I just wrote that $\varepsilon>0\;$ and $\;\varepsilon=\frac{x-y}2$. I apologise if it bothered you so much. I did not have the slightest intention of annoying you. Sorry ! $\endgroup$
    – Angelo
    May 22 at 17:27

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .