# Problem related to Frobenius coin problem

Let's say that the linear form $ax+by$ represents $n$ if $ax+by=n$ for some positive integer $x$ and $y$.

Call a pair $(a,b)\in\Bbb N\times\Bbb N$ with $\mathsf{gcd}(a,b)=1$ good if, for any $r,s,u,v>1$ with each of $rs,uv,ru,sv,rv,su<(a-1)(b-1)$ (the Frobenius number of $(a,b)$), there is at most one set from among $\{rs,uv\}$, $\{ru,sv\}$ and $\{rv,su\}$ with both components representable by $ax+by$.

Do good pairs exist at all?

If they do, then is it true that for every sufficiently large integer $l$, there is a good pair $(a,b)$ with $a,b\in[l,2l]$?

$a=22,b=21,s = 16, t = 17,r = 19,u = 15$

$$10a+4b=rs$$ $$8a+7b=rt$$ $$9a+2b=su$$ $$3a+9b=tu$$

• A piece of advice. To receive proper attention and get answered, a problem must be presented in a comprehensible way. For your specific question, you can consider, for instance, defining the notion of a "good pair" $(a,b)$, presenting and explaining some examples of good and "bad" pairs, and only then asking whether there is a good pair in any interval $[n,2n]$. – Seva Nov 30 '15 at 20:01
• @Seva Actually I thought about it, I am not sure $n_0$ has a good upper bound at all. – Brout Nov 30 '15 at 21:04
• @Seva you think it is comprehensible now? I do not have an example for good pairs. – Brout Nov 30 '15 at 23:08
• Doesn't the Frobenius problem itself immediately imply that the answer to your question is negative? All numbers $n$ greater than $(a-1)(b-1)$ have a representation $n=av+bw$ with $v,w$ positive, so as soon as your $r,s,t,u$ are sufficiently large (or more simply as soon as their products are) you'll inevitably have a positive representation. – Steven Stadnicki Dec 1 '15 at 7:56
• @Turbo You should really clarify the quantifiers in your original question, then - as it currently stands, the natural reading is 'if the condition on the left side of the implication arrow holds, then the condition on the right side holds' (this is what an implication usually means!) which is generally equivalent to saying 'for ALL values which satisfy the condition on the LHS, the condition on the right holds'. – Steven Stadnicki Dec 1 '15 at 8:16

I don't totally understand your definition of "good" so I don't really know exactly what makes your example bad. I filled some idle time trying to generalize your example of a bad pair. I realize (now) that that is not relevant to your actual question. But having found it anyway, here it is:

Your example of a bad pair $a=2\cdot 11,b=a+1=3\cdot 7$ suggests something like $a=n(4n+3),b=(n+1)(4n-1)$ The rest is nicely patterned as well with (possibly not on the first try) $u,s,t,r=u,u+1,u+2,u+2+n.$ A little fiddling gives

$$a=4n^2+3n,b=4n^2+3n-1,$$ $$s = 4n^2+n-2, t = 4n^2+n-1,r = 4n^2+2n-1,$$$$u = 4n^2+n-3=(n+1)(4n-3)$$

$$(2n^2+n)a+(2n^2-n-2)b=rs$$ $$(2n^2)a+(2n^2-1)b=rt$$ $$(3n+3)a+(4n^2-4n-6)b=su$$ $$(n+1)a+(4n^2-2n-3)b=tu$$

--

• Thank you for the time taken. The idea is simple. Lets say a linear form $ax+by$ represents $n$ if $ax+by=n$ for some $x,y\geq0$. Good pair $a,b$ essentially forms a linear form such that if $ax+by$ represents $rs$ and $tu$, it should not represent $rt,su$ and $ru,st$. That is all there is to it. I am looking for an infinite collection of good pair above a certain universal constant $n_0$. – Brout Dec 1 '15 at 6:54
• In example above we have $a=22,b=21$ represent $rs=19\cdot16$ and $tu=17\cdot 15$ but also representing $rt$ and $su$ so $a,b$ is a bad pair. – Brout Dec 1 '15 at 7:06