[Date Index] [Thread Index] [Author Index]

Re: Integers that are the sum of 2 nonzero squares in exactly 2 ways

• To: mathgroup at smc.vnet.net
• Subject: [mg125946] Re: Integers that are the sum of 2 nonzero squares in exactly 2 ways
• From: Dana DeLouis <dana01 at me.com>
• Date: Sun, 8 Apr 2012 04:18:25 -0400 (EDT)
• Delivered-to: l-mathgroup@mail-archive0.wolfram.com

> I don't know if this answers anything, but the factorization pattern of such numbers appears to be as follows. 
> All primes of the form 4n+3 > occur to even powers.

Hi.  I can't add much, but this article was very interesting...

http://mathworld.wolfram.com/SumofSquaresFunction.html


One of the codes along the line of 4n+3 was the following:

MySquaresRUnordered[n_Integer?NonNegative] := Module[
  {b, p, q, primes, powers},
  {primes, powers} = Transpose[FactorInteger[n]];
  mod1 = Flatten[Position[primes, _?(Mod[#, 4] == 1 &)]];
  mod3 = Flatten[Position[primes, _?(Mod[#, 4] == 3 &)]];
  If[Or @@ (OddQ /@ powers[[mod3]]),
   0,
   (*Else*)
   b = Times @@ (1 + # & /@ powers[[mod1]]);
   If[EvenQ[b], b, b + 1]/2
   ]
  ]

But this still considers zero a solution, which the Op does not want.

This is not faster, but just interesting.
A workaround might be that we want those solutions that equal length 2, but where the square root of the number is not an integer.

For example, 400 has two solutions:

PowersRepresentations[400,2,2]
{{0,20},{12,16}}

But... the square root of 400 is the integer 20, so this would not be a valid solution.
However, we also need those of length 3, but where the square of the number is an integer.
For example, 625 has 3 solutions.  but since 25^2 = 625, we can disregard this solution, and assume the remaining 2 solutions are valid.

PowersRepresentations[625,2,2]
{{0,25},{7,24},{15,20}}

Sqrt[625]
25

So, with two aux functions:

NotSquare[n_] :=  Not[IntegerQ[Sqrt[n]]]
IsSquare[n_] := IntegerQ[Sqrt[n]]

For this idea, you have a table of numbers....

tbl=Table[{n,MySquaresRUnordered[n]},{n,820}];

tt=Cases[tbl,{_,n_/;n==2}][[All,1]]
{25,50,65,85,100,125,130,145,169,170,185,200,205,.. ..}

But we disregard those that have an integer as a square root.

lst1=Select[tt,NotSquare]
{50,65,85,125,130,145,170,185,200,205,...}

However, the number 625 is missing from this small list.
We need to look at the few that have 3 as a solution, and also HAS an integer as a square root.

lst2=Select[Cases[tbl,{_,n_/;n==3}][[All,1]],IsSquare]
{625}


This is ok for small ranges, but slow when the range is large.

Only2[n_] := Piecewise[
   {{(SquaresR[2, n] - 4)/8 == 2,    IntegerQ[Sqrt[n]]}, 
    {(SquaresR[2, n] - 4)/8 == 1,    IntegerQ[Sqrt[n/2]]}, 
    {SquaresR[2, n]/8 == 2, True}}]

Select[Range[820],Only2]

{50,65,85,125,130,145,170,185,200,205,221,...}

= = = = = = = = = = = = = = =
Interesting subject :>)
Dana DeLouis
Mac & Math 8
= = = = = = = = = = = = = = =



On Mar 31, 4:44 am, d... at wolfram.com wrote:
> On Wednesday, March 28, 2012 5:00:45 AM UTC-5, Cisco Lane wrote:
> > I've been looking at integers that are the sum of 2 nonzero squares in exactly 2 ways. The smallest example is 50 = 5^2+5^2=7^2+1^2. The first few terms are 50, 65, 85, 125, 130, 145, ....  This is given in OEIS ashttps://oeis.org/A025285
> 
> > If I plot the pairs {1,50},{2,65},{3,85},... I get a more or less straight line with a slope of about 8.85... In other words, eventually, about one in 8.85 integers qualify.
> 
> > I wonder if there is a theoretical value for this approximate number of 8.85...?
> 
> I don't know if this answers anything, but the factorization pattern of such numbers appears to be as follows. All primes of the form 4n+3 occur to even powers. There are one, two or three other factors. They can take the form
> 
> 2^n*p*q where p and q are distinct primes of the form 4n+1, n an arbitrary nonnegative integer
> 
> 2^(2*n+1)*p^2 with n, p as above
> 
> 2^(2*n)*p^(2*k) for n as above and k satisfying some relation I cannot quite figure out other than it has to be at least 2. There may also be further restrictions on n that depend on k.
> 
> Daniel Lichtblau
> Wolfram Research

Re: Integers that are the sum of 2 nonzero squares in exactly 2 ways