Fun with Logarithms

My concentration is better today, and I spent three and a half hours ironing out the wrinkles in my logarithm proof. I’ve had the general outline written down for some time, but there turned out to be a lot of details that needed attention, so that I think I’m justified in counting the effort as my study for the day, even though I don’t usually include time spent blogging.

The proof uses several supporting facts that I’ve listed below as Notions 1-3. I haven’t decided yet whether I want to prove them, but I may at some point.1 I’ve also stated and used the Fundamental Theorem of Arithmetic without proving it. The course I took my first semester at college in which we developed the tools to prove that theorem, was one of the turning points in my life. I can no longer remember how it’s done, however.

I’m not entirely happy with this proof. I think it’s sound, but unclear in places. I had fun making it, though, and I hope someone enjoys it.


Definition: Two integers are relatively prime if they have no common factors besides $1$.

Notion 1: Every rational number can be written as the ratio of two integers that are relatively prime and at least one of which is positive.

Notion 2: If $a>1$, then $a^n>1$ if and only if $n$ is positive.

Notion 3: If $a$ divides $bc$ and $a$ and $b$ are relatively prime, then $a$ divides $c$.


The Fun(damental) Theorem of Arithmetic: Every integer greater then $1$ can be expressed as a product of primes in exactly one way, ignoring order.


Proposition: Given two natural numbers $a$ and $b$ that are both greater than $1$, if $\log_a b$ is rational, then both $a$ and $b$ are powers of a natural number $c$.

Suppose that $a$ and $b$ are natural numbers greater than $1$ and that $\log_a b$ is rational.

Then, using Notion 1, there exist integers $n$ and $m$ such that $a^{\frac{n}{m}}=b$, $n$ and $m$ are relatively prime, and $n$ or $m$ is positive.

A bit of algebra yields $a^n=b^m$.

Since both $a$ and $b$ are greater than $1$ and $n$ or $m$ is positive, the equal expressions $a^n$ and $b^m$ are both greater than $1$ by Notion 2. It follows, again by Notion 2, that both $n$ and $m$ are positive.

Consequently, because the integers are closed under multiplication, the value $a^n=b^m$ is an integer.

Let it be called $d$.

By the Fundamental Theorem of Arithmetic, $a$, $b$, and $d$ can all be expressed as a product of primes in exactly one way.

If the prime factorization of $a$ is $(a_1)(a_2)…(a_i)$, with some values possibly equal, then since $d=a^n$, the prime factorization of $d$ is $(a_1)^n(a_2)^n…(a_i)^n$.

Likewise, if the prime factorization of $b$ is $(b_1)(b_2)…(b_j)$, then since $d=b^m$, the prime factorization of $d$ is $(b_1)^m(b_2)^m…(b_j)^m$.

Notice that the prime factorization of $d$ contains the same factors as the prime factorization of $a$ and that the prime factorization of $d$ also contains the same factors as the prime factorization of $b$. It follows that the prime factorizations of $a$ and $b$ contain the same factors as one another, as well, though not necessarily the same number of times.

[This relationship between $a$ and $b$ is one I found very early on. I thought there was probably a name for it, but inquiries on Math Stack Exchange only turned up the notion of a radical, which is the product of one copy of every one of an integer’s prime factors. In those terms, what I’m saying is that $a$ and $b$ have the same radical.]

Now consider $p$, an arbitrary prime factor of $a$, $b$, and $d$. Let the number of occurrences of $p$ in the prime factorizations of each number be called $O_{a,p}$, $O_{b,p}$, and $O_{d,p}$, respectively.

Because an exponent represents repeated multiplication by the same value, it multiplies the number of occurrences of each factor. Thus, the relationships among $a$, $b$, and $d$ mean that $(O_{a,p})(n)=O_{d,p}=(O_{b,p})(m)$.

By Notion 3, since $(O_{a,p})(n)=(O_{b,p})(m)$ and $n$ and $m$ are relatively prime, $m$ divides $O_{a,p}$ and $n$ divides $O_{b,p}$. Thus, for the arbitrary prime factor $p$, $\frac{O_{a,p}}{m}=\frac{O_{b,p}}{n}$ is an integer. Furthermore, since all values involved are positive, it is a positive integer.

Therefore, it is possible to construct a natural number $c$ such that, for each prime factor $p$ in the prime factorizations of $a$ and $b$, the prime factorization of $c$ contains $\frac{O_{a,p}}{m}=\frac{O_{b,p}}{n}$ occurrences of $p$.

Now, by construction, the prime factorization of $c$ contains the same factors as the prime factorization of $a$.

Furthermore, by the reasoning used above, the prime factorization of $c^m$ contains the same factors as the prime factorization of $c$. Thus the prime factorization of $c^m$ also contains the same factors as the prime factorization of $a$.

Since an exponent multiplies the number of occurrences of each factor, given any prime factor $p$ in the prime factorization of $a$, the number of occurrences of $p$ in the prime factorization of $c^m$ is $\frac{O_{a,p}}{m}(m)=O_{a,p}$.

Recall that this is the number of occurrences of $p$ in the prime factorization of $a$. Hence $c^m$ has the same prime factors as $a$ with the same number of occurrences for each factor. It follows that $c^m=a$.

The same reasoning can be used to show that $c^n=b$, so the proposition is proven.

[Phew.]

  1. To do. ↩︎

8 Replies to “Fun with Logarithms”

  1. Three comments:

    (0) Very nice and careful, as always. And that you can propose questions like this and then answer them is wonderful.

    (1) I wouldn’t use as many ( ) in the products. For instance, I’d say $a_1^na_2^n\cdots a_i^n$.

    (2) Is the converse of your proposition true? Proving that the implication works in both directions would wrap the result up in a very neat package.

    1. I’m very pleased you liked my proof. You and Yerpa are the main audience for the more involved posts, and you have much more math background.

      Thank you for the style tip, too. I’ll bear it in mind.

      Asking “what about the converse?” is a good habit that I’m glad to be reminded of. I’m pretty sure a little algebra is all that would be needed to prove the converse here. The proof definitely wouldn’t be as long as this one. (Honestly, I wonder if this proof needs to be as long as it is. It seems like there should be a neater way to prove this proposition.)

Leave a Reply to Olly Cancel reply

Your email address will not be published. Required fields are marked *