I am looking at an algorithm in the book, 'Algorithms for Computer Algebra' by Geddes/Czapor/Labahn, and the algorithm on p. 345, 'Finite Field Square-Free Factorization', which square -free factorizes univariate polynomials over Galois Fields, of order '$q$'. Where, '$q$' here is a prime, or a power of a prime, so that $q = p^m$, where '$p$' is prime, and '$m$' is a positive integer.
On page 344, there is a theorem that states that if $a(x) = a_0 + a_1x + \ldots + a_n x^n$ be a polynomial of degree '$n$' in $GF(q)[x]$ satisfying $a'(x) = 0$. Then $a(x) = b(x)^p$, for some polynomial b(x), and that each coefficient of $b(x)$, is given by :-
$$b_i = a_{ip}^{1/p} = a_{ip}^{p^{m-1}}$$
Now, I have myself produced an example, to test this theorem, so I have my $a(x)$ as :-
$$a(x) = 2 x^8 + 4 x^4 + 8 x^2 + 7$$
where here, I have chosen $q = 16, p = 2, m = 4$
Now, $a'(x) = 0$ (Hope you agree with that?)
So, if I am interpreting this theorem correctly, then I make my corresponding coefficients for $b(x)$, as:-
$b_0 = a_0^8 = 7^8 = 1$, $b_1 = a_2^8 = 8^8 = 0$, $b_2 = a_4^8 = 4^8 = 0,$ $b_3 = a_6^8 = 0^8 = 0,$ $b_4 = a_8^8 = 2^8 = 0$
All the other coefficients will be zero, so there's no point going any further - so my final result for $b(x)$ is:- $b(x) = 1$
Yet this cannot possibly be correct, as $b(x)^p$ does not equal $a(x)$.
Does anyone else have this book, or know this theorem. Can you see what's going wrong here please?
Thanks Jeremy