4

I am, just for fun, looking for long and complicated proofs for statements which can be proven rather easily and much faster. The proof itself still has to be correct however.

While the proof should be obfuscated, all parts should have some relevance. So do not prove Fermat‘s last theorem and end with "ah, by the way: 1+1=2, so the statement follows.

It is also boring to obfuscate simple arithmetic; one can prove "1+1=2" in 100 pages only using addition, subtraction, multiplication and division – but that is not fun.

I rather look for some very interesting obfuscation of a proof. Maybe a statement of elementary number theory can be proven in a "nice" complicated way. Or maybe one can use functional analysis to prove basic analysis stuff etc.


Here is one (not that good) example: Theorem: For $a, b \in \mathbb{R}$ it holds that $(a-b)^2 = a^2 - 2ab + b^2$.

Proof: Let $f: \mathbb{R} \rightarrow \mathbb{R}, x \mapsto x^2$. As $f$ is analytical the Taylor expansion of $f$ converges. Therefore

$$ \begin{align*} x^2 &= f(x) \\ &= Tf(x,b) \\ &= \sum_{n=0}^\infty \frac{f^{(n)}(b)}{n!} (x-b)^n \\ &= \frac{b^2}{0!} + \frac{2b}{1!} (x-b) + \frac{2}{2!} (x-b)^2 + \sum_{n=3}^\infty \frac{0}{n!} (x-b)^n\\ &= b^2 + 2bx - 2b^2 + (x-b)^2 \\ &= -b^2 + 2bx + (x-b)^2, \end{align*} $$ ie. $$ x^2 + b^2 - 2bx = (x-b)^2 $$ and for $x = a$ the theorem follows.

Keba
  • 2,485
  • 13
  • 28
  • I feel like any proof can be made arbitrarily complicated. –  May 29 '15 at 21:06
  • @avid19: Yes, and I tried to make clear that I am not looking for an arbitrarily complication, but an interesting one. – Keba May 29 '15 at 21:07
  • Perhaps a source of such proofs would be examining the Theorem -> Corollary relationship of most textbooks. Often the corollary would have been a known result, but is proven using an unnecessarily strong theorem. – Joel May 29 '15 at 21:08
  • 1
    I love this question. Of course, you can't dismiss Fermat's Last Theorem, for Fermat himself could have whittled Wiles' overly complicated proof to the margin of one of his books! If only we could have seen it...or just look at a generalization for something that was easily proved in Ottoman Arabia, like $a^n + b^n = c^n$ has no solutions for $n = 7$; why not just prove Fermat in the general case instead of the single page needed for this hypothesis?! – Alex May 29 '15 at 21:12
  • 4
    Somewhat similar question on MathOverflow: http://mathoverflow.net/questions/42512/awfully-sophisticated-proof-for-simple-facts – Hans Lundmark May 29 '15 at 21:15
  • 1
    What you might look for is a difficult proof of some rather general statement, which is then applied to a particular case that would have been easier to tackle on its own. For example, existence of a solution of $ax = b$ for $a \ne 0$ as a corollary of the Fundamental Theorem of Algebra. – Robert Israel May 29 '15 at 21:16
  • You presumably don't want examples like this: Show that $3$ is a quadratic non-residue of $7$. We first prove Quadratic Reciprocity, and then the result follows immediately. – André Nicolas May 29 '15 at 21:23
  • 1
    $2^n > n$ $\forall n \in \Bbb N$ by the diagonal argument. – wlad May 29 '15 at 21:34
  • "1+1=2" is yet obfuscated enough using ZFC. – Blex May 29 '15 at 21:42
  • @Alex: You got that Fermat reference. ;) However, if you are able to prove that theorem as short as Fermat claimed, I will want to see that proof. Of course your ideas are then very good answers (although nobody would care for my question, but only for your proof of Fermat‘s last theorem.) – Keba May 30 '15 at 00:53
  • There are famous examples of a proof much less difficult than others. The very celebrated Mordell's theorem of the finite basis of an elliptic curve was simplified (and largely generalized) by A. Weil; more recently in the time, the Conjecture of Mordell about rational points of curves of genus ≥ 2 was proved to be true by Faltings whose very acclaimed and intricate proof (Fields Medal) was simplified by an italian I don't remember his name now. – Piquito May 30 '15 at 01:22
  • The italian I mentioned, a very good mathematician, is Enrico Bombieri. – Piquito May 30 '15 at 01:44

3 Answers3

1

Maybe Wilson's theorem. For any odd prime $p$, $(p-1)!=-1$ mod $p$.

Easy proof : Among $2,..., p-2$ the numbers can be gathered by pairs of inverse mod $p$ so that :

$$2\times...\times p-2=1\text{ mod }p $$

Hence : $(p-1)!=-1$ mod $p$.

Slighty more complicated proof : Take $G:=\mathfrak{S}_p$ the symmetric group of $p$ elements. Set $n_p$ the number of $p$-Sylows of $G$.

We know that its $p$-Sylows are cyclic groups of order $p$. By a standard argument we know that there are $(p-1)!$ $p$-cycles. A $p$-cycle will have exactly $p-1$ $p$-cycles in the $p$-Sylow it generates so that $n_p=\frac{(p-1)!}{p-1}=(p-2)!$.

Now we apply second Sylow's theorem to get $n_p=1$ mod $p$. Finally $(p-1)!=-1$ mod $p$ by multiplying both sides by $p-1$.

  • I was looking for a question of this type for 15 minutes JUST to share this hilarious argument I stumbled on today!! – Andres Mejia Sep 23 '19 at 00:15
1

Every function from $\mathbb{N}$ to $\mathbb{R}$ is continuous, where the topologies are the canonical.

Proof 1: [Short]

$\mathbb{N}$ is a discrete space. $\blacksquare$

Proof 2:

Take a function $f: \mathbb{N} \rightarrow \mathbb{R}$. Now, consider the following function $g:\mathbb{R} \rightarrow \mathbb{R}$:

$g(x)=\left(f(\lfloor x \rfloor +1)-f(\lfloor x \rfloor) \right)(x-\lfloor x \rfloor)+f(\lfloor x \rfloor)$

In each interval $[n,n+1]$, where $n \in \mathbb{N}$, we have that the function $g$ is a linear function, hence continuous. Hence, by the pasting lemma, $g$ is a continuous function. Now, it is easy to see that $g|_{\mathbb{N}}=f$. Since the restriction of a continuous function is continuous, we have our result. $\blacksquare$

Also, see https://mathoverflow.net/questions/42512/awfully-sophisticated-proof-for-simple-facts

and more specifically:

https://mathoverflow.net/a/44742/48745 (which is a gem)

0

Proof: $0a = 0$

$$(a+0)^2 = a^2 = a^2 + 2(0) + 0^2$$

$$a^2 = a^2 + 2(0) + 0^2$$ $$0 = 2(0) + 0^2$$ $$2(0) = (1+1)0 = 1(0) + 1(0) = 0 + 0 = 0$$ $$\dfrac{0^2}{2} >= \sqrt{0^2} = 0$$ $$0^2 >= 2(0)$$

We have proven that $$2(0) = 0$$

Therefore, $$(a+0)^2 = a^2 = a^2 + 2(0) + 0^2 = a^2 + 0$$ $$a^2 = a^2 + 0$$ $$0 = 0$$

Q.E.D.

Jimmy360
  • 649