Is $x^{14}+x^7+1$ irreducible over $\mathbb{Q}[x]$?
I think it is, but I'm not able to justify it using any existing criterion (e.g. Eisenstein). Any help?
Indeed, this is a question I encounter on a linear algebra one. The original question gives a $8 \times 8$ real matrix satisfying $A^{21}=I$ and asks to prove that $\mathbb{R}^8$ can be decomposed into the direct sum of 4 2-dimensional vector subspace invariant w.r.t. $A$.
My attempt was to find the minimal polynomial then discuss several cases, which need the factor of $x^{21}-1$. Any hint on the original question is also appreciated.
No, it is not irreducible. Indeed, as you observe, $X^{21}-1 = (X^{7}-1)(X^{14}+X^{7}+1)$. On the other hand, $X^{21}-1 = \prod_{d \mid 21} \Phi_{d}(X)$, where $\Phi_{d}(X)$ denotes the $d$th cyclotomic polynomial.
In this case, the divisors of $21$ are very nice: they are $1, 3, 7$ and $21$. It is easy to compute $\Phi_{d}$ for $d$ prime by a familiar Eisenstein computation; we find:
$$\Phi_{d}(X) = \sum_{i=0}^{d-1} X^{i}$$
for $d$ prime.
This gives $\Phi_{3}(X) = X^{2}+X+1, \Phi_{7}(X) = X^{6}+X^{5}+\cdots+1$. Since $\Phi_{1}(X) = X-1$, we can compute $\Phi_{21}(X)$ by long division, for example; it is a polynomial of degree $12$. By unique factorization, we must therefore have that $X^{14} + X^{7}+1$ is irreducible. By degree considerations (or just factoring $X^{7}-1$ as indicated above), one sees that $X^{14}+X^{7}+1 = (X^{2}+X+1)\Phi_{21}(X)$, which confirms the result in the comments by J.W. Tanner.