When is the sum of three reduced rationals equal to an integer

34 Views Asked by At

When is the sum of three reduced rationals equal to an integer?

This may be a duplicate of Question 1550437 but even if it is, there is no answer associated with this other question.

Given three moduli, $m1$, $m2$, and $m3$, not necessarily distinct or coprime, but $m_i > 1$ $\forall m_i$, when does the sum of rationals: \begin{equation} \left( \frac{r_1}{m_1} + \frac{r_2}{m_2} + \frac{r_3}{m_3} \right) \end{equation} admit integer solutions? The remainders, $r_1$, $r_2$, and $r_3$, are all coprime to the corresponding modulus; i.e. $(r_1, m_1) = (r_2, m_2) = (r_3, m_3) = 1$. The rationals in the sum are in reduced form: i.e. $r_i < m_i$ $\forall r_i$

The question can be restated for what constellation of modulii do coprime remainders exists such that: \begin{equation} \left( \frac{r_1}{m_1} + \frac{r_2}{m_2} + \frac{r_3}{m_3} \right) \in \mathbb{Z} \end{equation} is true for at least one constellation of remainders?

I have found some solutions: \begin{align} \left( m_1, m2 \right) = \left( m_1, m3 \right) = \left( m_2, m3 \right) = 1 &\implies \nexists r_1, r_2, r_3 \text{ $\mid$ } \left( \frac{r_1}{m_1} + \frac{r_2}{m_2} + \frac{r_3}{m_3} \right) \in \mathbb{Z} \\ \left( m_1, m2 \right) = 1 \text{ and } m_3 \ne m_1 m_2 &\implies \nexists r_1, r_2, r_3 \text{ $\mid$ } \left( \frac{r_1}{m_1} + \frac{r_2}{m_2} + \frac{r_3}{m_3} \right) \in \mathbb{Z} \\ \left( m_1, m2 \right) = 1 \text{ and } m_3 = m_1 m_2 &\implies \exists r_1, r_2, r_3 \text{ $\mid$ } \left( \frac{r_1}{m_1} + \frac{r_2}{m_2} + \frac{r_3}{m_3} \right) \in \mathbb{Z} \end{align}

But this is too piece-meal for my tastes.

There must be some application of the Chinese remainder theorem that can tell me when: \begin{equation} m_2 m_3 r_1 + m_1 m_3 r_2 + m_1 m_2 r_3 \equiv 0 \text{ mod } \text{ lcm}\left( m_1, m_2, m3 \right) \end{equation} does or does not admit a solution?

Can anyone point me to an article or text which discusses under what conditions the above congruence does or does not have solutions in the three variables: $\left(r_1, r_2, r_3\right)$?

Do I solve for the an unrestricted solution for $\left(r_1, r_2, r_3\right)$ and then see if the co-prime condition is met? Or can I use the co-primality of the $r_i$'s to my advantage?