Take the 2-minute tour ×
MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

$\exists p, q \in \mathbb{P}: p^4+1 = 2q^2$? I suspect there is some simple proof that no such p, q can exist, but I haven't been able to find one.

Solving the Pell equation gives candidates for p^2=x and q=y, with x=y=1 as the first candidate solution and subsequent ones given by x'=3x+4y, y'=2x+3y; chances of a prime square seem vanishingly unlikely as x increases, but I don't have a proof.

Meta: how do you search for a question like this? I looked for a searching HOWTO here and on meta, and couldn't find one. That the search appears to strip '^' and '=' makes it all the harder.

share|improve this question
2  
If you're looking for a simple proof, you probably have to a) factor the left hand side over Z[i]; b) subtract 1 and factor the right hand side over Z[sqrt{2}], or c) multiply by 2 and use the classification of Pythagorean triples (or subtract a square and factor the right hand side, which is a difference of squares). –  Franz Lemmermeyer May 14 '10 at 12:31
    
Yes, $\mathbb{P}$ is the primes. That is standard isn't it? –  Hugo van der Sanden May 14 '10 at 12:32
1  
@Franz: This is a variation of Pell's equation, so $p^2+q\sqrt2=(1+\sqrt2)^{2k+1}$ (it can be shown by modular argument that $k$ is multiple of 4). Thus (a) can't work, (b) is used in derivation of the above formula (there are useless recursions mentioned by the author), and for (c) I have no idea of what do you mean (how to relate the equation with Pythagorean triples?) –  Wadim Zudilin May 14 '10 at 13:45
3  
Interestingly enough, the similar equation p^2 + 1 = 2q^4 DOES have an interesting solution, namely (p,q) = (239,13). This is related to Machin's identity pi/4 = 4 arctan(1/5) - arctan(1/239). (See Ribenboim's book on Catalan Conjecture for this.) See p.7-8 of www.math.wisc.edu/~ellenber/MCAV.pdf for an "explanation" of this solution via a congruence mod 5 between a weight-2 cuspform in conductor 1024 and an Eisenstein series. –  JSE May 14 '10 at 20:45
1  
@Wadim: I can't see why a) cannot work. As for c), 2p^4 + 2 = (p^2+1)^2 + (p^2-1)^2 = (2q)^2. –  Franz Lemmermeyer May 16 '10 at 16:10
show 2 more comments

3 Answers 3

up vote 20 down vote accepted

This is not my solution, but I don't remember where I learned it.

Square both sides, subtract $4p^4$, and divide by 4 to obtain $({p^4-1\over 2})^2=q^4-p^4$.

However, $z^2=x^4-y^4$ has no solutions in non-zero integers. This is Exercise 1.6 in Edwards's book on Fermat's Last Theorem. The proof uses the representation of Pythagorean triples and infinite descent.

So you must have $p=\pm 1$.

share|improve this answer
    
Thanks, this does solve it. I also found the infinite descent proof of the lemma at planetmath.org/encyclopedia/… –  Hugo van der Sanden May 14 '10 at 15:42
add comment

I think -- correct me if I am wrong -- that it is known that the equation $x^4+1=Dy^2$ with given squarefree $D$ has at most one solution in integers, primes or no primes. See for example J. H. E. Cohn., Math. Comp. 66 (1997), 1347-1351. (http://www.ams.org/journals/mcom/1997-66-219/S0025-5718-97-00851-X/home.html) The article cites an original proof by Ljunggren in 1942, which I can't find online.

share|improve this answer
    
Ah, excellent. That suggests x=y=1 is the only solution; I'll take a look at the paper to see if it constitues a proof. –  Hugo van der Sanden May 14 '10 at 14:07
    
Oh, it is the Ljunggren paper I'd need. –  Hugo van der Sanden May 14 '10 at 15:40
add comment

I don't know if there is a simple proof, but I know one which is easy to do because it lets a computer do all the work (but the work is perhaps complicated): you simply ask a computer to solve Y^2=2X^4+2 in integers for you, like this (in MAGMA, but other packages will do it too):

> IntegralQuarticPoints([2,0,0,0,2]);
[
    [ 1, 2 ],
    [ -1, 2 ]
]

so the only solution with p,q integers is p,q=+-1 and that's it.

share|improve this answer
2  
How can Magma be used with confidence, their code is not open. –  teil May 14 '10 at 13:39
    
Negative's point is a valid one, but I think Kevin's answer points to the fact that there is a standard algorithm to solve these kinds of problems. Maybe it is also implemented in other packages. –  Felipe Voloch May 14 '10 at 14:59
1  
This is only semi-true. The outer code of SIntegralPoints in Magma can be viewed easily. As Felipe Voloch says in part, this lists the algorithm. Low-level functions (eg. Height) cannot be seen, though "open" is technically true as you can decompile. From the philosophy of science, no system is used with confidence, and only independent verification can be partially satisfactory. We still have at least two chip manufacturers, so hardware independence is possible. For using software, too many packages throw eggs into a large basket (like GMP), and so I see a scarcity of true independence often. –  Junkie May 15 '10 at 6:19
add comment

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.