Let us consider the function:


Here $f(x,y)$ is the division of two quadratic polynomials with the same degree

where $a,b,c,d,r,,s,h,w$ are non zero integers.

My question is:

(1) Given $a,b,c,d,r,,s,h,w$ as integers, can we find $x$ and $y$ rational numbers such that $f(x,y)$ is a strictely positive integer.

(2) Given $a,b,c,d,r,,s,h,w$ as integers, can we find $x$ and $y$ positive integers such that $f(x,y)$ is a strictely positive integer.


We can assume that $f$ is bijective with respect to $x$ if $y$ is a constant and bijective with respect to $y$ if $x$ is a constant.

  • $\begingroup$ Are there any restrictions whatsoever on $(a,b,c,d,r,s,h,w)$? (For instance, we will obviously have a hard time if $a=b=c=r=s=h=0$, because then our choice for $(x, y)$ doesn't even matter.) $\endgroup$ – Lynn Jan 7 '18 at 14:27
  • $\begingroup$ @Lynn: No. Only they are non zero integers. $\endgroup$ – China Jan 7 '18 at 14:32
  • $\begingroup$ With bijectivity condition, it is easy. Fix some $x$, then choose $y$ such that $f(x, y) =1$ (you can because of surjectivity). Maybe the condition you want is "non constant" instead of "bijective"? $\endgroup$ – Andrea Marino Jan 12 '18 at 11:50

The denominator can be the opposite of the numerator, so that the function is a negative constant. Take for instance $$\frac{xy+x+y+1}{-xy-x-y-1}$$ in this way $f(x, y) = - 1$ and you have no chance of finding such $(x, y) $!

Let me enounce a simple fact: For every fixed integers $u, v$, it exists an integer N of the same sign of $u$ such that for every $n$ of the same sign of $u$ with $|n|>|N|$ we have

$$un+v >0 $$

As a corollary, we have : If u, u' have the same sign, for every v, v' it exists n such that un+v, u'n+v' are positive. Proof. Take N, N' given by the previous fact: we know they have the same sign. Thus if we take a n of the same sign of N, N', but greater in modulus, we have $un+v, u'n+v'$ both positive.

Now i will show you that if $a, r$ have the same sign, the answer is affirmative in both cases. In fact, choose $x$ of the same sign of $a$, big enough such that both $ax+b, rx+s$ are positive. Then you can find $y$ such that both the linear forms $(ax+b) y + (cx+d), (rx+s) y+(hx+w) $ are positive, because the coefficient of $y$ have the same sign. In conclusion, the quotient

$$\frac{axy+by+cx+d}{rxy+sy+hx+w} $$

Is positive!

  • $\begingroup$ @ Andrea Marino: Can you explain on more detaills. $\endgroup$ – China Jan 7 '18 at 15:24
  • $\begingroup$ I have filled in more details! Is it clearer now? $\endgroup$ – Andrea Marino Jan 8 '18 at 17:49
  • $\begingroup$ @ Andrea Marino: But the last expression is not necessarly a positive integer. $\endgroup$ – China Jan 9 '18 at 12:42
  • $\begingroup$ hahah, ok, I didn't read that condition. The problem became more interesting! Let me make some changes. $\endgroup$ – Andrea Marino Jan 9 '18 at 22:45
  • $\begingroup$ I have a solution to both the questions I will write in a few days! I have a complete characterization of point (1), while just a counterexample to (2) more refined than the constant function. $\endgroup$ – Andrea Marino Jan 15 '18 at 0:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.