Little question about gcd and Fermat pseudoprimes.

Question

From Wikipedia:

...a Carmichael number is a composite number $n$ which satisfies the modular arithmetic congruence relation:

$1)$ $b^{n-1}\equiv 1{\pmod {n}}$

for all integers $b$ which are relatively prime to $n$...

Equivalently, a Carmichael number is a composite number $n$ for which

$2)$ $b^{n}\equiv b{\pmod {n}}$

for all integers $b$.

It's clear to me that $2) \Rightarrow 1)$ if $\gcd(b,n)=1$, but I can't rule out the possibility of $\gcd(b,n) \neq 1$ (so that $1)$ is emptily satisfied) and $2)$ not holding.

Is it true that if $\gcd(b,n) \neq 1$ then $b^{n}\equiv b{\pmod {n}}$?

Also, regarding Fermat's pseudprimes, many sources define them using the congruence $b^{n}\equiv b{\pmod {n}}$. Other (like Wikipedia) uses the congruence $b^{n-1}\equiv 1{\pmod {n}}$, which seems more restrictive. I know theoretically this is a minor difference, my only doubt is about the available statistics about pseudoprime, to which definition they conform?

Answer 1

Suppose $\ (1)\ $ holds

Let $\ p\ $ be a prime factor of $\ n\ $

• If $\ p\ $ divides $\ b\ $, then $\ b^n \equiv b\mod p\ $ is obvious
• If $\ p\ $ does not divide $\ b\ $ , we have $\ b^{n-1}\equiv 1\mod p\ $ implying $\ b^n\equiv b\mod p\ $

Since $\ n\ $ must be squarefree ( a Carmichael number is always squarefree ) , the chinese remainder theorem allows to conclude that with all prime factors of $\ n\ $ , $\ n\ $ itself must divide $\ b^n-b\ $ , showing $\ (2)\ $