For a prime integer p, how many ordered pairs of positive integers (a, b) are there that satisfy $$\frac{1}{a} + \frac{1}{b} =\frac{1}{p}$$

For example, for p = 5, $$\frac{1}{6} + \frac{1}{30}$$ and $$\frac{1}{30} + \frac{1}{6}$$ are two different ways of getting $\frac{1}{5}$.

Ok, so I tried this for prime numbers from 2 to 19. It seems like that for any prime number there are only 3 such ordered pairs.

(Question: $\frac{1}{10} + \frac{1}{10}$ is just one ordered pair for p=5, right?)

But I don't know how to go about proving this. I can see that:

After $\frac{1}{p}$, take the next smallest fraction of the form $\frac{1}{a}$. Now we always get a fraction $\frac{1}{a} + \frac{1}{b}$. Also, now b is the upper bound for the numbers we need to check. But there is always just one other integer between a and b which satisfies for this property, this integer is is 2p

The question does not specifically ask for a prove, but they always expect a prove for everything.


If we multiply through by $pab$, we find that for $a, b \ne 0$, our equation is equivalent to $pb+pa=ab$, which we can rewrite as $$(a-p)(b-p)=p^2.$$ We are looking for non-zero solutions of this equation. Conveniently, $p^2$ does not have many factorizations!

We can have $a-p=-1$, $b-p=-p^2$, which yields a negative $b$, or $a-p=-p^2$, $b-p=-1$, which yields a negative $a$.

We could have $a-p=-p$, $b-p=-p$, but that yields the impossible $a=b=0$.

We can have $a-p=1$, $b-p=p^2$, or $a-p=p^2$, $b-p=1$, which respectively yield the solutions $a=p+1$, $b=p^2+p$, and $a=p^2+p$, $b=p+1$.

Finally, we can have $a-p=p$, $b-p=p$, which yields $a=b=2p$.

Comment: The same idea can be used to find all integer solutions of the equation $\frac{1}{x}+\frac{1}{y}=\frac{1}{n}$, where $n$ is a given positive integer. The factorizations of $n^2$, with the exception of $n^2=(-n)(-n)$, supply the integer solutions.


$1/a+1/b=1/p$, $bp+ap=ab$, $(a+b)p=ab$, so $p$ divides $a$ or $p$ divides $b$ (or both). Let's say $p$ divides $a$, so $a=cp$ for some $c$. Now $cp+b=bc$, $b=bc-pc=(b-p)c$, so $c$ divides $b$. Let $b=cd$. Then $d=cd-p$, so $p=cd-d=(c-1)d$. So either $c-1=p$ and $d=1$, or $c-1=1$ and $d=p$. Now work your way back up to $a$ and $b$.


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.