1
$\begingroup$

There are two positive integer sets $(b,a),(a,b)$ which satisfy $a^2+b^2=128^2+25^2$ and neither $a$ nor $b$ equals to $128$ and $25$. Find these two sets.

I don't even have a idea how to do this so I would love some help. I have tried to find nearby large numbers such as $130$ and $129$ but both of them don't work. Also will inequality (AM-GM or Cauchy work here?) This is a contest question so I need a preferably fast solution.

  • 3
    "This is a contest question so I need to solve this fast". Usually it is against contest rules to ask for outside help.....2017-01-05
  • 2
    I would assume that the contest guidelines don't allow you to seek help online...2017-01-05
  • 0
    i covered this while studying for amc8 ummm 80^2 + 103^2 works2017-01-05
  • 0
    um actually it is not a contest. I just mean that I hope to find a rather faster way of doing this since my plug and chuk is really slow. Saketh can u show your solution? Maybe we can discuss about it2017-01-05
  • 0
    @alexzhang I have edited your tags; this question has nothing to do with algebraic topology...2017-01-05
  • 2
    " This is a contest question" "um actually it is not a contest" Um... why should we believe you?2017-01-05
  • 3
    @fleablood if it were a contest, he wouldn't have access to his phone. when he wants a fast solution, he doesnt want us to give him a solution asap, he wants us to give him a solution that uses minimal steps. if he really snuck into the bathroom of whatever to use his phone, he wouldn't ask for a fast solution because he has limited time2017-01-05
  • 0
    @SakethMalyala, some math contests take place online where the participants are on the honor system not to cheat. AFAIK none of these "matter" (in the same sense that, say, the IMO matters).2017-01-05
  • 0
    @tilper What kind of contest is that?!2017-01-05
  • 0
    @SimpleArt, [here's one place](http://internetolympiad.org/) that occasionally hosts such contests. I've seen at least one or two others but I'm having trouble recalling them. It's been a while so they may not even exist anymore.2017-01-05
  • 2
    Or it could be _from_ a contest that has concluded.2017-01-05

4 Answers 4

3

Here is one way to proceed:

Render $25^2=5^4, 128^2=4\times 8^4$. Then:

$u^4+4v^4=(u^2-2uv+2v^2)(u^2+2uv+2v^2); u=5, v=8$

$25^2+128^2=5^4+4(8^4)$

$=(5^2-2(5)(8)+2(8^2))(5^2+2(5)(8)+2(8^2))=73\times 233$

Then $73=8^2+3^2$, $233=13^2+8^2$, and we get two product expressions. One gives the original sum of squares and is included as a check, the other gives a second pair:

$(8^2+3^2)(13^2+8^2)$

$=((8\times 13)+(3\times 8))^2+((8\times 8)-(3\times 13))^2=128^2+25^2$

$=((8\times 13)-(3\times 8))^2+((8\times 8)+(3\times 13))^2=80^2+103^2$

Since the factors $73$ and $233$ are prime, this solution is unique apart from trivial operations.

  • 0
    thanks! this helped me more compare to the previous stuff2017-01-05
  • 0
    Wait, $(25,128)$ is not allowed...2017-01-05
  • 0
    That was included as a check.2017-01-05
2

$25^2+128^2$ is of the form

$$a^4+(a^2+b)^2=2a^4+2a^2b+b^2=2(a^2+b)a^2+b^2$$

As luck (or intelligent design) would have it, in this case we have

$$2(a^2+b)=2\cdot128=16^2$$

Thus

$$25^2+128^2=5^4+(5^2+103)=16^2\cdot5^2+103^2=80^2+103^2$$

This doesn't rule out the possibility of a third pair of squares summing to $17009$, but the problem, as stated, doesn't ask for that. If you do want to rule out other possibilities, you can give a proof based on the fact that the only primes less than $\sqrt[3]{17009}\approx25$ congruent to $1$ mod $4$ are $17$, $13$, and $5$, none of which divide $17009$. (It's immediately clear that $5$ and $17$ aren't divisors. As for $13$, I reduced $17009=13000+4009$, then $4009=3900+109$.) Fermat's sum-of-two-squares theory does the rest.

Note, this approach gets an answer without factoring $17009$, but only because we got lucky in that $a^2+b$ in this case is twice a square.

To give credit where credit is due, I got the key idea for this approach from a deleted answer by user Xam.

  • 0
    You can test $17009$ for divisibility by $13$ (also $7$) by taking the difference between the last three digits and the remaining digits. Thus $17-(00)9=8=$ not a multiple of $13$.2017-01-05
  • 0
    @OscarLanzi, nice! I'll have to remember this!2017-01-05
  • 0
    You're casting out multiples of $1001$ which is divisible by $13$.2017-01-06
1

One may note that by taking both sides modulus $3$, it is clear that $x,y$ must not be a multiple of $3$.

Taking both sides modulus $2$, it is clear either $x$ is odd and $y$ is even or the other way around.

Taking modulus $5$, it is clear that either $x$ is a multiple of $5$ and $y$ is equivalent to $2$ or $3$ modulus $5$.

Putting these hints together, and assuming $x$ is the multiple of $5$, we find

$$x\stackrel?=5,10,20,25,35,40,50,55,65,\dots,130$$

One may also note that if $x<91$, then $y>90$, likewise, if $x>90$, $y<91$.

From here, checking term by term isn't very hard, and one can deduce solutions $(80,103)$.

  • 0
    But there is supposed to be another solution ... .2017-01-05
  • 0
    @OscarLanzi Sorry, wasn't a fully finished answer. Was seeing if I could come up with any good ideas for higher modulus'.2017-01-05
  • 0
    Please complete the analysis. Alas, I suspect that this is not really efficient as we know now that the only other odd number that works is $103$.2017-01-05
  • 0
    @OscarLanzi I haven't fully finished, and alas, I have to go to bed (T_T I have school to get to tomorrow), but I've added some results that make this a better answer.2017-01-05
  • 0
    Taking mod $8$ you can conclude one of them is divisible by $4$.2017-01-05
1

Since this is a contest question, I think I am doing no harm in showing a solution that is the Gaussian integers version of that of Oscar Lanzi.

$128^2+25^2 = 17009$ factorizes as $73 \cdot 233$. Both primes are congruent to $1$ modulo $4$ so they can be written as the sum of two squares $$ 73 = 8^ 2 + 3^ 2 = (8 + 3 i)(8 - 3 i), \quad 233 = 13^2 + 8^2 = (13 + 8 i) (13 - 8 i). $$ If we group $$ (8 + 3 i)(13 - 8 i) = 128 + 25 i $$ we obtain $17009 = 128^2 + 25^2$. If we group $$ (8 + 3 i)(13 + 8 i) = 80 + 103 i $$ we obtain $17009 = 80^2 + 103^2$.