Let $\overline{a_n \ldots a_1a_0}$ be the decimal representation of $65^k$ for some $k \geq 2$. Prove that the polynomial $a_nx^n+\cdots+a_1x+a_0$ has no rational roots.

I thought about using the rational root theorem. We know that $a_0 = 5$ since $65$ ends in a $5$, so any rational root must be of the form $-\dfrac{5}{a}$ or $-\dfrac{1}{a}$ where $a$ is an integer factor of $a_n$. How can we continue?

  • 1
    $\begingroup$ Potentially useful: a quick computer experiment shows that all the digits $\{1, \dots, 9\}$ occur for $a_n$. $\endgroup$
    – erfink
    Mar 31, 2017 at 23:06
  • 2
    $\begingroup$ That follows from $\log_{10} 65$ being irrational, @erfink $\endgroup$ Mar 31, 2017 at 23:09
  • $\begingroup$ you have $a_1=2$ too, though that probably won't be helpful. $\endgroup$
    – dezdichado
    Mar 31, 2017 at 23:31
  • $\begingroup$ Similar to @dezdichado comment, for $k\geq 6$, we seem to have that all the decimal representations end with $\dots 890625$. For $k \geq 7$, we seem to have endings of $\dots 2890625$ or $\dots 7890625$. $\endgroup$
    – erfink
    Mar 31, 2017 at 23:42
  • 1
    $\begingroup$ @астонвіллаолофмэллбэрг What do you mean by reduced it to 1 case? $\endgroup$ Apr 1, 2017 at 0:14

2 Answers 2


An incomplete answer, but I hope it may help in getting the final answer itself.

Let $\frac pq$ be a rational root of $f$. Since $f-f(10)$ has $10$ as a root, it follows that $f-f(10) = (x-10)g(x)$ for some $g$. Now, let $x = \frac pq$: $$ \left(\frac pq - 10\right) g\left(\frac pq\right) = -f(10) \neq 0 $$ Multiplying by $q^n$ : $$ \left(p-10q\right)q^{n-1}g\left(\frac pq\right) = -f(10)q^n $$ From here, we know that $\deg g = n-1$, so that $q^{n-1}g\left(\frac pq\right)$ is an integer. Hence, $p-10q | q^n f(10)$, and by co-primality, $p-10q | f(10) = 65^k$.

We have that $p|5$ and $q | a_n$. So, $p = 1/5$, and $q$ must satisfy $p-10q | 65^k$, where $-9 \leq q \leq -1$ (the polynomial does not have positive roots by Descartes' rule of signs).

Note that if $p=1$, then for the above $q$, not case matches. However, if $p=5$, then $q=-6$ fits the criteria. So all we have to do is to check that $\frac{-5}{6}$ is not a rational root, then we are done. In fact, this shows that if the leading coefficient $a_n \neq 6$, then we are done.

  • $\begingroup$ have not written down it rigorously, but if you let $f(x)=(6x+5)h(x)$ and manage to prove $h(x)\in\mathbb{Z}[x]$, then $h(10) = 65^{k-1}$ and $h$ is monic, which is a contradiction I think. $\endgroup$
    – dezdichado
    Apr 1, 2017 at 3:16
  • $\begingroup$ Can't you also have $q=-2$? $\endgroup$
    – Micah
    Apr 1, 2017 at 4:39
  • $\begingroup$ Let $f_n(x)=(6x+5)f_{n-1}(x)+(x-10)R(x)$. Then $f_n$ has $-\frac{5}{6}$ as a root iff $R$ does. But if $n \geq 2$ then $f_{n-1}(0)=f_n(0)=5$, which means we must have $R(0)=2$. So $R(x)$ can't possibly have $-\frac{5}{6}$ as a root. I still don't see any way to rule out $-\frac{5}{2}$ (which your argument does not do; if $p=5$ and $q=-2$ then $p-10q=25$ is a factor of $65^k$ for $k \geq 2$). $\endgroup$
    – Micah
    Apr 1, 2017 at 6:26
  • $\begingroup$ The argument with the Descartes-sign rule is correct, but it is obvious without this rule that a polynomial with positive coefficients cannot have non-negative roots. $\endgroup$
    – Peter
    Apr 1, 2017 at 8:21
  • $\begingroup$ Oh, I forgot about $q=-2$! All right, thank you all for reminding me of that. However, my feeling is that we can get a trivial bound, from the fact that all coefficients are bounded by $10$, and the first and last are $6$ and $5$ respectively. I am not sure, but I'll look for some bounds and tell you. $\endgroup$ Apr 1, 2017 at 10:44

Let $f_n$ be the polynomial obtained from the digits of $65^n$. The other answer shows that if $p/q$ is a root of this polynomial in lowest terms, then $p-10q|65^n$; along with the bounds on $f_n$, this implies that we must have $p=-5$ and either $q=2$ or $q=6$. That is, the only possible rational roots are $-\frac{5}{6}$ and $-\frac{5}{2}$.

Note that $(6x+5)^n$ is also a polynomial which evaluates to $65^n$ at $x=10$. So we can write $$ f_n(x)=(6x+5)^n+(x-10)R(x) $$ for some polynomial $R$.

Setting $x=0$ in this expression, we have $ 5=5^n-10R(0) $, and so $R(0)=\frac{5^{n-1}-1}{2}$, which is not a multiple of $5$.

Now, suppose $n \geq 3$ and $x$ is either $-\frac{5}{6}$ or $-\frac{5}{2}$. Then $(6x+5)^n$ has $5$-adic valuation at least $3$ (that is, its numerator is a multiple of $5^3$ when it is written in lowest terms). In contrast, the $5$-adic valuation of $R(x)$ is $0$ (because the $5$-adic valuation of $R(0)$ is $0$) and $(x-10)$ is either $-\frac{5}{6}-10=-\frac{65}{6}$ or $-\frac{5}{2}-10=-\frac{25}{2}$, neither of which has $5$-adic valuation greater than $2$. So $(6x+5)^n$ and $(x-10)R(x)$ have unequal $5$-adic valuations, which means that $f_n(x)=(6x+5)^n+(x-10)R(x)$ cannot possibly be $0$.

It only remains to check that $f_2$ has no rational roots, which is an easy calculation.


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.