Let $a,b \in {\mathbb{Z_+}}$ such that $$a|b^2, b^3|a^4, a^5|b^6, b^7|a^8 \cdots$$

Prove $a=b$

  • 9
    $\begingroup$ Show that if $a<b$ then $b^n>a^{n+1}$ for large values of $n$. $\endgroup$ – Thomas Andrews Mar 23 '12 at 12:22

consider the prime factorisations of $a = \prod_p p^{\nu_p(a)}$ and $b = \prod_p p^{\nu_p(b)}$. Your assumptions yield

  • $\nu_p(a) \le \nu_p(b^2) = 2\nu_p(b)$ for each $p$
  • $3\nu_p(b) \le 4\nu_p(a)$
  • $\ldots$
  • $(4n+1)\nu_p(a) \le (4n+2)\nu_p(b)$ for each $p$, $n$
  • $(4n+3)\nu_p(b) \le (4n + 4)\nu_p(a)$, each $p$, $n$

So we have for each $p$, $n$ $$ \frac{4n + 3}{4n+4} \cdot \nu_p(b) \le \nu_p(a) \le \frac{4n+2}{4n+1}\nu_p(b) $$ letting $n\to \infty$ yields $\nu_p(a) = \nu_p(b)$ for each $p$, so $a = b$.



If $a > b$ then $\frac{a}{b}>1$ and hence there is an $n$ such that $\left(\frac{a}{b}\right)^n > b$, thus $a^n > b^{n+1}$. This contradicts $a^{4k+1} | b^{4k+2}$. The case $a<b$ works in just the same way.

  • $\begingroup$ I like that this solution makes it clear that divisibility is a red herring---the problem would work just as well with the much weaker hypothesis that $a \leq b^2$, $b^3 \leq a^4$, etc., for real values of $a$ and $b$. $\endgroup$ – Dan Petersen Mar 23 '12 at 15:08
  • $\begingroup$ @Dan Divisibility isn't necessarily a red herring, since the proof I gave employs divisibility, and it works in rings where the (archimedean) order-based proofs do not apply, e.g. the Gaussian integers $\mathbb Z[i]$ or the ring of integers of any (complex) algebraic number field. $\endgroup$ – Bill Dubuque Mar 24 '12 at 21:50

Let $p$ be a prime, and suppose $p^7$ divides $a$, but $p^8$ doesn't. Then from $a^9\mid b^{10}$ you can deduce $p^7\mid b$ (why?). Now if $p^8\mid b$, then from $b^{11}\mid a^{12}$ you would get $p^8\mid a$ (why?), contradiction. So same power of $p$ divides $a$ and $b$.

Can you see how to generalize this argument?


Hint $\rm\ \ \forall n\in\mathbb N:\ \ a\:\!\left(\dfrac{a}b\right)^{4n+3}\!\in\mathbb Z,\:\ b\:\!\left(\dfrac{b}a\right)^{4n+1}\!\in \mathbb Z\ \ \Rightarrow\ \dfrac{a}b,\:\dfrac{b}a\in\mathbb Z\ \ \Rightarrow\ \ a = \pm b\ \ \ $ QED

Remark $\ $ This is true far more generally. Suppose $\rm\:D\:$ is any Noetherian integrally closed domain, e.g. any PID. Suppose that $\rm\:w\:$ is a fraction over $\rm\:D\:$ such that some unbounded sequence of powers of $\rm\:w\:$ has a common denominator $\rm\:0 \ne d\in D,\:$ i.e. $\rm\:d\!\:w^{n_i}\in D\:$ for all $\rm\:n_i.\:$ Then $\rm\:w\in D.$

Proof $\ $ By ACC the sequence of ideals $\rm (d, dw^{n_1}, dw^{n_2},\ldots)$ eventually stabilizes, which implies that for some $\rm\:k\:$ we have $\rm\: dw^{n_k}\in (dw^{n_{k-1}},\ldots, dw^{n_1}, d),\:$ which implies

$$\rm d\: w^{n_k} + c_{n_{k-1}} d\: w^{n_{k-1}} +\:\! \cdots +\: c_{n_1} d\: w^{n_1} + d\: =\: 0$$

Cancelling $\rm\:d\:$ yields $\rm\:w\:$ is integral over $\rm\:D,\:$ hence $\rm\:w\in D,\:$ since $\rm\:D\:$ is integrally closed. $\ $ QED

Elements whose powers have such a common denominator are called almost integral. It is clear that integral elements are almost integral. The above shows that the converse holds true in Noetherian domains.


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.