However, afaics this doesn't seem to apply to 441, which is the product of
3exp2 and 7exp2 - both even powers.
Am I missing something obvious?
--
Mike Stone - Peterborough, England
"Freddie experienced the sort of abysmal soul-sadness which afflicts one of
Tolstoy's Russian peasants when, after putting in a heavy day's work
strangling his father, beating his wife, and dropping the baby in the
reservoir, he turns to the cupboard only to find the vodka bottle empty".
P G Wodehouse - Jill the Reckless
"Mike Stone" <mws...@aol.com> wrote in message
news:7o9uvaF...@mid.individual.net...
> I understood that FCT stated that a positive integer is expressible as the
> sum of two squares if all its odd prime factors are to an even power.
>
> However, afaics this doesn't seem to apply to 441, which is the product of
> 3exp2 and 7exp2 - both even powers.
>
> Am I missing something obvious?
>
Whoops! Should have been "- - factors of the form 4n+3 - -"
Hi,
and 441 = 21^2 + 0^2, that is sum of two squares ;-)
FCT doesn't say "strictly positive".
It is strictly positive when there is at least one 4k+1 prime factor,
or the factor 2 at an odd power.
(this doesn't prevent *some* of the decompositions to have nullsquares,
for instance 15^2 = 9^2 + 12^2 *or* 15^2 + 0^2)
This is also the drawback of other related theorems about sum of
squares :
Lagrange : any number is sum of 4 squares
Gauss : all numbers not of the form (8k+7)4^n are sum of 3 squares
All these including the possibility for some of the squares to be 0^2
For instance playing with the Lagrange theorem,
56 is sum of up to 4 squares in only one way = 6^2 + 4^2 + 2^2
(as all numbers in the form 7*2^(2k+1) this decomposition is unique)
therefore, 56 is sum of 4 squares as 6^2 + 4^2 + 2^2 + 0^2
*requiring* a 0^2
This results into complicated criteria (like the details about FCT)
to say when the squares are *non null*
BTW the "FCT" (in a letter from Fermat to Mersenne on 25 December 1640)
is exactly about :
the number of ways in which a number may be sum of two squares
(including the eventual + 0^2, and not just if it is or not)
Regards.
--
Philippe C., mail : chephip, with domain free.fr
site : http://mathafou.free.fr/ (mathematical recreations)
"Philippe 92" <nos...@free.invalid> wrote in message
news:mn.4c1d7d9cd...@free.fr...
Thanks.
What brought this to my attention was a little conundrum about Pythagorean
Triangles. I understand that the hypotenuse, being the sum of two squares,
is always of the form 4n+1, and never 4n+3. OK so far, but I notice that
some 4n+1 numbers - 9, 21, 33, 57, 69, 77, 81, 93 - -, are _not_ the sum of
two squares, and in every case that I've found, these numbers are ones whose
prime _factors_ are 4n+3. I can now see that 81 passes as a sum of two
squares in the same way as 441, but the others don't.
Is there some obvious reason (Obvious except to me, that is. You will gather
I am no mathematician) why a 4n+1 number which is the _product_ of two 4n+3
numbers (or the multiple of such a product) cannot be the sum of two
squares?
The problem is here in the chain :
"it is sum of two square => it is a 4k+1 number" (true, and obvious)
Which is not "it is a 4k+1 number => ..."
As you noticed not all of the N = 4k+1 are sums of two squares, and
it *is* the FCT [Fermat Christmas Theorem].
This can be understood in plenty of ways.
The most direct one is to consider prime decomposition in Z[i]
But as "not being a mathematician" this would be of little help !
To prove that, using elementary methods (without Gauss integers)
is tedious and quite long, so I won't try.
Some of the steps are however not obvious.
A 4n+3 *prime* can't obviously be the sum of two squares.
The important step is that *any* 4n+1 prime is the sum of two
squares (in only one way), and 2 = 1^2 + 1^2 also is.
One quite simple step along the proof :
- If N = a^2 + b^2 with GCD(a,b)=1, then all odd divisors
(prime or not) of N are of the form 4k+1
Proof :
To have a 4k+3 factor, it must have at least one 4n+3 *prime* factor.
Suppose N has a p = 4n+3 prime factor, and let m = 2n+1.
As GCD(a,b)=1 and p divides n, p doesn't divide a nor b.
"N multiple of p" is a^2 = -b^2 (mod p)
and because m is odd, (a^2)^m = -(b^2)^m (mod p) or
a^(2m) = - b^(2m) (mod p)
Little Fermat theorem says :
if 'a' is not a multiple of prime p, then a^(p-1) = 1 (mod p)
Here with 2m = p-1 : a^(2m) = b^(2m) = 1 (mod p)
Hence a contradiction and N can't have a 4n+3 prime factor.
We deduce of that :
- A square free number is the sum of two squares if and only
if it has only prime factors of the form 4k+1, or p=2.
The final result is FCT : (existence part)
A number is the sum of two squares if and only if all its prime
factors of the form 4k+3 have even exponents.
Then the number of ways in which N can be sum of two squares
depends only on the exponents of the 4k+1 primes.
(and for considering non null squares, on the exponent of 2)
In the case of hypothenuse of Pythagorean triangles, we want to get
the *square* of hypothenuse to be a sum of squares.
We deduce that the hypothenuse itself must have at least one 4k+1
prime factor to get a non degenerated triangle (c^2 = c^2 + 0^2 for
any c, uninterresting degenerated case).
For 4k+3 prime factors of hypothenuse, they don't care as they are
squared, hence the exponents in c^2 are always even !
A primitive pythagorean triangle must have only prime factors of the
form 4k+1 in its hypothenuse, and conversedly any product of primes of
the form 4k+1 is the hypothenuse of a primitive Pythagorean triangle.
You may get more detailed proofs (in books or on the web) about sum of
squares, but this requires a minimum mathematical skill.
A few conundrums, reflections and scripts about sum of squares
on my Website : http://mathafou.free.fr/pba_en/pb001.html
> The final result is FCT : (existence part)
> A number is the sum of two squares if and only if all its prime
> factors of the form 4k+3 have even exponents.
Express 9 that way. Or 49. Or 121.
The 'only if' holds but the 'if' fails.
It would seem that a necessary condition is that the number have a prime
factor that is expressible as a sum of two squares, i.e., the prime 2 or
a prime of form 4k+1.
Hi,
9 = 3^2 + 0^2 etc, as already said.
It is not required that the squares are non null !!!
And yes if you require non null squares, it requires the existence of
at least one 4k+1 prime factor, or an odd power of 2.
For instance 36 is only 6^2 + 0^2, just factor 2 doesn't suffice, it
requires odd power.
that's not FCT.
also 21 exp 2 + 0 exp 2,
any number having the form given above will itself be square, or square doubled.
> Virgil wrote :
> > In article <mn.4d2b7d9cc...@free.fr>,
> > "Philippe 92" <nos...@free.invalid> wrote:
> >
> >> The final result is FCT : (existence part)
> >> A number is the sum of two squares if and only if all its prime
> >> factors of the form 4k+3 have even exponents.
> >
> > Express 9 that way. Or 49. Or 121.
> >
> > The 'only if' holds but the 'if' fails.
> >
> > It would seem that a necessary condition is that the number have a prime
> > factor that is expressible as a sum of two squares, i.e., the prime 2 or
> > a prime of form 4k+1.
>
> Hi,
>
> 9 = 3^2 + 0^2 etc, as already said.
> It is not required that the squares are non null !!!
>
> And yes if you require non null squares, it requires the existence of
> at least one 4k+1 prime factor, or an odd power of 2.
>
> For instance 36 is only 6^2 + 0^2, just factor 2 doesn't suffice, it
> requires odd power.
As I merely said "necessary" but did not require "sufficient", my
statement was correct, if incomplete.
>Whoops! Should have been "- - factors of the form 4n+3 - -"
I have my suspicions about old Fermat, anyway. As in he seemed to have
a few ideas where he'd left the proof in his other jacket or whatever,
and IIRC at least one was simply wrong.
Still, it worked - he'll be remembered. Fermats last theorem will no
doubt always be Fermats last theorem, whereas the guy that actually
published a proof - well, what was his name again?