To prove a fundamental idea is "translation".
First, prove: there exists a primitive root g modulo p such that
gρ−1≡1(modp2)
In fact, take any primitive root g modulo p. If g does not satisfy (1), we say g+p is a primitive root modulo p that satisfies (1).
First, using the binomial theorem and the fact that g is a primitive root modulo p, we know that g+p is a primitive root modulo p.
Second, we have
(g+p)ρ−1≡gρ−1+p(p−1)gρ−2≡1−pgρ−2≡1(modp2)
Therefore, there exists a primitive root modulo p that satisfies (1).
Next, prove: if g is a primitive root modulo p that satisfies (1), then for any α∈N∗, g is a primitive root modulo pα.
First, we have the following conclusion: for any β∈N∗, we can set
gα(pβ)=1+pβ⋅kβ
where p∤kβ. In fact, when β=1, by the choice of g, (2) holds. Now assume (2) holds for β, then
gρ(pβ+1)=(gρ(pβ))p=(1+pβ⋅kβ)p≡1+pβ+1⋅kβ(modpβ+2)
Combining p∤kβ, we know that (2) holds for β+1. Therefore, (2) holds for β∈N∗.
Second, let δ=δpα(g), then by Euler's theorem, we know δ∣pα−1(p−1). And by g being a primitive root modulo p, and gδ≡1(modpα) (of course, also gδ≡1(modp)), we know (p−1)∣δ. Therefore, we can set δ=pβ−1(p−1), where 1⩽β⩽α.
Now using (2), we know
gρ(pβ)=≡1(modpβ+1), i.e., gδ=1(modpβ+1).
Combining gδ≡1(modpα), we know β⩾α.
In summary, β=α, i.e.,
δpα(g)=pα−1(p−1)=φ(pα)
Thus, g is a primitive root modulo pα.