Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

There was an interesting problem asked about triples $(x,y,z)$ which are solutions of

$$x! = y! + z!.$$

Here $(2,1,1)$ is a solution because $2! = 1! + 1!$, as are $(2,1,0)$ and $(2,0,1)$.

Now I wanted to analyze this a bit further and thought of using the gamma function definition of a factorial, to see where it led and this is what I got:

$x! = y! + z!$

$ \Gamma(x) = \Gamma(y) + \Gamma(z) $

$\int_{-\infty}^{\infty}t^xe^{-t}dt = \int_{-\infty}^{\infty}t^ye^{-t}dt + \int_{-\infty}^{\infty}t^ze^{-t}dt $

$\int_{-\infty}^{\infty}t^xe^{-t}dt - \int_{-\infty}^{\infty}t^ye^{-t}dt - \int_{-\infty}^{\infty}t^ze^{-t}dt = 0$

$\int_{-\infty}^{\infty}(t^x - t^y - t^z)e^{-t}dt = 0$

Now my line of thinking was that, in a manner similar to the fundamental lemma of the calculus of variations, this should imply that the above integral can only be true for arbitrary values of $x, y$ and $z$ if $t^x - t^y - t^z = 0$.

I'm not really sure if that is justifiable so I'd appreciate some comment on this.

The reason I continued on despite the uncertainty is that when you're left with the polynomial $t^x - t^y - t^z = 0$ you encounter a strange fact. First off, plugging in a triple like $(2,1,1)$ results in $t^2 - t^1 - t^1 = 0$ which implies $t^2 = 2t$ thus $t = 2$. Now plugging in $t = 2$ gives $2^2 = 1^2 + 1^2$. In other words you get what you'd expect. However plugging in a triple you know is inconsistent, like $(0,0,0)$, gives

$t^0 = t^0 + t^0$

$1 = 1 + 1$

$1 = 2$.

You get absurdities when you plug in inconsistent triples.

So while $(2,1,0)$, which works, gives

$t^2 - t^1 - t^0 = 0$

$t^2 - t - 1 = 0$

whose roots are

$\frac{1}{2} + \frac{\sqrt{5}}{2}$

$\frac{1}{2} - \frac{\sqrt{5}}{2}$

and on plugging these into $t^x - t^y - t^z = 0$ gives a consistent equality for both roots, an inconsistent triple like $(3,2,1)$ leaves us with $t^3 - t^2 - t = 0$, one of whose roots are $t = 0$ (which gives a consistent equality $0^3 - 0^2 - 0 = 0$ whereas another one of it's roots, $t = \frac{1}{2} + \frac{\sqrt{5}}{2}$ does not give a consistent equality

$(\frac{1}{2} + \frac{\sqrt{5}}{2})^3 - (\frac{1}{2} + \frac{\sqrt{5}}{2})^2 - (\frac{1}{2} + \frac{\sqrt{5}}{2}) = 0$.

If all of the above is magically okay you can use this idea to show that no triples $(n, n - 1, n - 2)$ will work for $n$ greater than $2$. I'd like to go further with it but I'm afraid it's all just conceptually flawed, is it?

A more interesting question stems from something a great lecturer told me, she basically analyzed the whole question geometrically and said something about certain triples having to do with the area of those gamma integrals canceling out. I thought I understood what she was saying but I actually don't so I'd appreciate any comment on what this could mean.

I hope I was clear enough, thanks for your time.

share|improve this question
4  
The reasoning after and including calculus of variations doesn't make sense to me. Also I would note: this isn't the usual definition of the gamma function, which interpolates as $\Gamma(x) = (x-1)!$. –  Cocopuffs Oct 3 '12 at 17:23
1  
Wow, I'm such an idiot. Based on forgetting the - 1 term I went in circles doing a lot of nonsense. Just delete this post, no marks, complete & utter failure... Thanks for your time. –  sponsoredwalk Oct 3 '12 at 18:13
5  
no! upvote for you for forging ahead in muddy waters! –  im so confused Oct 3 '12 at 20:37

1 Answer 1

$x! = y! + z!$ does not have any solutions in integers with $x \geq 3.$ As soon as $x \geq 3,$ we have $(x-1)! \leq x! / 3.$ With the necessary $y,z < x,$ we get $y! \leq x! / 3, \; \; z! \leq x! / 3,$ so $y! + z! \leq 2 x! / 3$ and $y! + z! \neq x!$

share|improve this answer
    
Neat solution. You can rewrite the solution this way: $y,z <x$ hence $y,z \leq x-1$, and thus $x!=y!+z! \leq (x-1)!+(x-1)!$ which implies $x \leq 2$. –  N. S. Oct 3 '12 at 17:46
2  
The more interesting problem is $x! = y! \; z!.$ For any $n,$ we know $n! = (n-1)! \; n.$ Taking $n = z!,$ we get the infinite family $ (z!)! = (z! - 1)! \; z!.$ I think there is at least one other solution, the question is to find others. It's on MO somewhere. –  Will Jagy Oct 3 '12 at 18:07
    
@Will. It's here too. –  Rick Decker Oct 3 '12 at 18:21
    
@RickDecker thanks, I think that is actually where I saw it. Also I had not remembered $6! \; 7! = 10!$ –  Will Jagy Oct 3 '12 at 18:34

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.