2
$\begingroup$

If the following quadratic equation $$qx^2+(p+q)x+bp=0$$ always has rational roots for any non-zero integers $p$ and $q$ what will be the value of $b$?

My book's solution says the value of $b$ will be $0$ or $1$.

If we consider the discriminant of the equation, $$D=(p+q)^2-4bqp = p^2+2q(1-2b)p+q^2$$ then $D$ should be a perfect square of a rational number for the equation having rational roots. so the value of b should be 0 0r 1 for given conditions (p and q are non-zero integers and the equation always has rational roots)

But I am not sure whether we can conclude the discriminant $D=p^2+2q(1-2b)p+q^2$ is a perfect square of a rational number if only if b is 0 or 1.

I can see if b is 0 or 1, then $D$ will be $(p+q)^2$ or $(p-q)^2$. So $D$ will be a perfect square of a rational number.

But I can't figure out the other case: If $D$ is a perfect square of a rational number for any non-zero integer p and q , then the value of b will be 0 or 1.

EDIT 1: b is rational number.

$\endgroup$
1
  • $\begingroup$ Are there restrictions on $b$? (Is it an integer? a rational number?) $\endgroup$
    – Blue
    Jul 2 '13 at 22:05
3
$\begingroup$

The key is that it must always have rational roots for any $p, q$.

Claim: Let $f(x) = x^2 + bx + c$, where $b$ and $c$ are rational numbers. If $f(x)$ satisfies the condition that for all rational numbers $r$, $f(r)$ is the square of a rational number, then $f(x)$ must be the square of a polynomial.

Proof: By completing the square, we simply need to consider polynomials of the form $x^2 + \frac{n}{m}$, where $m,n$ are integers and $\gcd(n,m) = 1$.

Consider $x = 0$ and $\frac{1}{m}$, which gives us $\frac{n}{m}$ and $\frac{1+mn}{m^2}$ are both squares of rational numbers. Multiplying both terms by $m^2$, we get that $mn$ and $mn+1$ are both integers that are squares of a rational number, hence they are perfect squares, so $mn=0$.

Since $m\neq 0$, hence $n=0$ and thus $f(x) = x^2$ and we are done. $_\square$

Corollary: $b=0$ or $1$.

Proof: Consider the polynomial $f(x) = \left( \frac{p}{q} \right)^2 + 2(1-2b) \frac{p}{q} + 1$. From the conditions, for rational numbers, this always evaluates to a rational square. Hence, we must have $ f(x) = (ax+b)^2$.

Comparing coefficients, we see that $ a ^2 = 1, b^2 = 1$, which implies that $f(x) = (x+1)^2 $ or $(x-1)^2$, which hence correspond to $b=0$ and $b=1$ respectively. $_\square$


Note: Franklin shows that the claim is true for all polynomials, though it uses more machinery.

$\endgroup$
1
  • $\begingroup$ Nice! I had typed a horribly awkward argument, and decided, with some reluctance, to post it. Now don't need to. $\endgroup$ Jul 3 '13 at 1:40

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.