bremenfallturm said:
Ah yes, of course it's by definition. I think I'm so used to hearing "a number is prime if only divisible by itself and 1" but of course it follows from that what you wrote.
That is the definition of irreducible numbers, numbers that cannot be reduced to proper factors (except ##\pm 1## of course). Irreducible integers are prime and vice versa. This is not true for some other number systems so it makes sense to distinguish the two properties. The true definition of a prime number ##p\neq \pm 1## is
$$
p\,|\,(a\cdot b) \Longrightarrow p\,|\,a \text{ or } p\,|\,b
$$
which is more convenient in proofs about congruences.
bremenfallturm said:
Well, we can write ##9=3\cdot 3## but ##9\nmid 3##, as a simple example.
A better example is ##36.## We have ##9\,|\,36=6\cdot 6## but ##9\,\nmid\,6.## Hence ##9## is not prime.
We cannot know whether ##d## is prime or not. But partitioning the the prime factors of ##n## into arrays of prime factors of ##k## and prime factors of ##l## and the rest works.
It is usually more convenient in algebra to use indirect proofs. We want to show
$$
k\,|\,n \text{ and }l\,|\,n \text{ and }\operatorname{gcd}(k,l)=1 \Longrightarrow (k\cdot l)\,|\,n
$$
Assume that ##(k\cdot l)\,\nmid \,n.## Then there is a factor ##d\,|\,(k\cdot l)## such that ##d\,\nmid\,n.## Then there is a prime power ##p^q## that divides ##d## but does not divide ##n.## From ##p\,|\,p^q\,|\,d\,|\,(k\cdot l)## we get by the definition of a prime number ##p## that ##p\,|\,k## or ##p\,|\,l.## Now, if even ##p^q\,|\,k## then ##p^q\,|\,k\,|\,n## and likewise ##p^q\,|\,l\,|\,n## which we excluded. Hence ##p^r\,|\,k## for some ##q>r>0## and ##p^s\,|\,l## for some ##q>s>0.## But then ##p\,|\,k## and ##p\,|\,l## which contradicts the condition that ##k## and ##l## are coprime.
I know I got dizzy, too, while writing it. But that is how such proofs in abstract algebra usually work. The problem here was to reduce the situation from ##d## to a prime ##p## so that I could use the definition of primality which ##p## has and ##d## does not. The principle I used here is called
the pigeonhole principle. It is about distributing more pigeons than there are holes (or the other way around).
bremenfallturm said:
I see, thanks to your trivial example as well.
Is that what was missing from my reasoning or is there more I was missing?
I tried to reason again and ended up with this reasoning:
View attachment 351568
(^looks like the "t" in "not" slipped of and fell down in the image above, sorry about that!)
View attachment 351569
So I can see that we can see that ##P_{n^2}## consists of all factors in both ##P_k## and ##P_l## and I wrote down the definition of a prime number. But I am now just a little unsure about how I can apply the definition to conclude that ##P_n## contains all factors in both ##P_k## and ##P_l##.
Keep the example of ##9\,|\,36## in mind. I have proven a different statement, namely that if two coprime numbers divide another number ##n## then their product also divides ##n##. This is enough to conclude
\begin{align*}
10\,|\,n^2 &\Longrightarrow 2\,|\,10\,|\,n^2=n\cdot n\text{ and }5\,|\,10\,|\,n^2=n\cdot n\\
&\Longrightarrow 2\,|\,n\text{ and } 5\,|\,n \text{ since } 2 \text{ and }5\text{ are prime}\\
&\Longrightarrow (2\cdot 5)\,|\,n\text{ since } 2 \text{ and }5 \text{ are coprime}
\end{align*}
Operating with ##n^2## instead makes the situation unnecessarily complicated. We only need to consider the prime factors of ##n.##
The other proof is with
Bézout's lemma. It is shorter and far less dizzy. It states that there are always integers ##s,t## such that
$$
\operatorname{gcd}(k,l)= s\cdot k + t\cdot l
$$
Are we allowed to use this theorem?