2

Let $n$ be a positive integer. One can show (not that easy, but still via elementary methods) that the number of triples $(x,y,z)$ of positive integers satisfying $xyz + x + y = n$ is $O(n^{\frac{1}{3}+\varepsilon})$ for any $\varepsilon > 0$. (Use that $xz+1$ divides $n-x$, take without loss of generality $x< n^{\frac{1}{3}}$ or $z<n^{\frac{1}{3}}$, etc.)

Hence I was wondering - is it also true that this number is at least $Cn^{\frac{1}{3}}$ for some constant $C>0$?

DesmondMiles
  • 2,714
  • If you compute the number of solutions for the first several values of $n$, you can check to see if the sequence appears in the OEIS. If it does, you might get some insight into the likelihood of your conjecture (and possibly even an answer). – Barry Cipra Mar 10 '20 at 19:01
  • 1
    @BarryCipra That unfortunately leads to this page which is rather uninformative. – Servaes Mar 24 '20 at 21:36

4 Answers4

2

No. I'll show there are arbitrarily large $n$ with at most $c(\log n)^2$ solutions.

Let $s(n)$ be the number of solutions $(x, y, z)$ of $xyz + x + y = n$, and $t(n)$ the number of solutions of $xyz \leq n$, so we have $t(n) \geq \sum_{k=1}^n s(k)$, since $xyz + x + y \leq n$ implies $xyz \leq n$. Note that for given $x, y$, the number of $z$ with $xyz \leq n$ is $\lfloor n/xy \rfloor$, hence $$t(n) = \sum_{1 \leq x, y \leq n} \left \lfloor \frac{n}{xy} \right\rfloor \leq \sum_{1 \leq x, y \leq n} \frac{n}{xy} = nH_n^2 \leq 2n(\log n)^2$$ holds for sufficiently large $n$. If we also had $s(n) > 16(\log n)^2$ for large $n$, this would mean $$t(n) \geq \sum_{k=\lceil n/2 \rceil}^n s(k) > \sum_{k=\lceil n/2 \rceil}^n 16(\log k)^2 \geq (n/2) 16(\log (n/2))^2 \geq 2n(\log n)^2$$ for sufficiently large $n$, a contradiction, so there are arbitrarily large $n$ with $s(n) \leq 16(\log n)^2$.

user125932
  • 8,813
1

The equation is $$ x+y+z x y=n\tag 1 $$ I will find the number of solutions of (1) given a positive integer $n$, when $x,y,z$ are positive integers. For this done we assume $z$ is a parameter and rewrite (1) in the form $$ nz+1=(xz+1)(yz+1)\tag 2 $$ Set $N=nz+1$ and $AB=N$. Then the number of solutions of (2) when $x,y\geq 0$, $z>1$ is $$ r^{*}(z,n)=\sum_{ \begin{array}{cc} A,B>0\\ AB=nz+1\\ A\equiv 1(z)\\ B\equiv 1(z) \end{array} }1.\tag 3 $$ We assume $z\geq 2$. The case $z=1$ is easy (I will leave it). Hence equation $(1)$ have solutions when $x,y\geq1$ and $z\geq2$: $$ r(n)=-2(n-1)+\sum^{n}_{k=2}r^{*}(k,n)=-2(n-1)+\sum^{n}_{k=2}\sum_{ \begin{array}{cc} 0<d|(nk+1)\\ d\equiv 1(k) \end{array} }1 $$ The term $-2(n-1)$ in $r(n)$ is for removing the zero solutions $x=0$ or $y=0$. Hence the number of solutions of (1) is $$ r(n)=-2n+d(n+1)+\sum^{n}_{k=2}\sum_{ \begin{array}{cc} 0<d|(nk+1)\\ d\equiv 1(k) \end{array} }1 $$ where $d_a(n)=\sum_{d|n,d\equiv1(a)}1$.

0

$xyz+x+y=n \overset{x\to u+v\\y\to u-v}{\implies} (u z + 1)^2 - (vz)^2 = nz+1$

gp-code for test first small $n,z$:

xyzn()=
{
 for(n=1, 100,
  k= 0;
  for(z=1, 1000,
   T= thue('x^2-1, n*z+1);
   for(i=1, #T,
    X= T[i][1]; Y= T[i][2];
    if(X!=0, if(Y!=0,
     v= Y/z;
     if(v==floor(v),
      u= (X-1)/z;
      if(u==floor(u),
       x= u+v; y= u-v;
       if(x>0 && y>0,
\\        print("n = "n";    (x,y,z) = ("x","y","z")");
        k++
       )
      )
     )
    ))
   )
  );
  if(k, print("n = "n";    #sol = "k));
 )
};

Output:

n = 7;    #sol = 2
n = 10;    #sol = 2
n = 11;    #sol = 2
n = 13;    #sol = 2
n = 14;    #sol = 2
n = 15;    #sol = 2
n = 16;    #sol = 4
n = 19;    #sol = 4
n = 20;    #sol = 2
n = 21;    #sol = 2
n = 22;    #sol = 6
n = 23;    #sol = 4
n = 25;    #sol = 2
n = 26;    #sol = 4
n = 27;    #sol = 2
n = 28;    #sol = 4
n = 29;    #sol = 2
n = 30;    #sol = 2
n = 31;    #sol = 8
n = 32;    #sol = 4
n = 34;    #sol = 6
n = 35;    #sol = 2
n = 36;    #sol = 4
n = 37;    #sol = 4
n = 38;    #sol = 6
n = 39;    #sol = 4
n = 40;    #sol = 4
n = 41;    #sol = 2
n = 42;    #sol = 2
n = 43;    #sol = 6
n = 44;    #sol = 6
n = 45;    #sol = 2
n = 46;    #sol = 10
n = 47;    #sol = 6
n = 49;    #sol = 2
n = 50;    #sol = 4
n = 51;    #sol = 4
n = 52;    #sol = 8
n = 53;    #sol = 4
n = 54;    #sol = 4
n = 55;    #sol = 8
n = 56;    #sol = 8
n = 57;    #sol = 2
n = 58;    #sol = 8
n = 59;    #sol = 4
n = 61;    #sol = 6
n = 62;    #sol = 8
n = 63;    #sol = 4
n = 64;    #sol = 10
n = 66;    #sol = 6
n = 67;    #sol = 6
n = 68;    #sol = 6
n = 69;    #sol = 2
n = 70;    #sol = 6
n = 71;    #sol = 10
n = 72;    #sol = 4
n = 73;    #sol = 6
n = 74;    #sol = 6
n = 75;    #sol = 2
n = 76;    #sol = 12
n = 77;    #sol = 2
n = 78;    #sol = 6
n = 79;    #sol = 10
n = 80;    #sol = 6
n = 81;    #sol = 2
n = 82;    #sol = 12
n = 83;    #sol = 6
n = 84;    #sol = 2
n = 85;    #sol = 6
n = 86;    #sol = 10
n = 87;    #sol = 4
n = 88;    #sol = 4
n = 89;    #sol = 2
n = 90;    #sol = 4
n = 91;    #sol = 10
n = 92;    #sol = 12
n = 93;    #sol = 4
n = 94;    #sol = 14
n = 95;    #sol = 8
n = 96;    #sol = 4
n = 97;    #sol = 2
n = 98;    #sol = 6
n = 99;    #sol = 4
n = 100;    #sol = 10
Dmitry Ezhov
  • 1,653
-1

xyz+x+y=n. ----- (1)

Solution is,

(x,y,z,n)=((p-38),(p-11),(1),(p^2-47p+369))

for p>38

example, p=40, (x,y,z,n)=(2,29,1,89)

since 'p' can take many value's, equation (1) has infinite number of solutions.

Mathew
  • 21
  • 1