0
$\begingroup$

Consider a polynomial $$f(X) = X^{(p-1)p^{n-1}} + X^{(p-2)p^{n-1}} + \cdots + X^{p^{n-1}} + 1$$

Now I need to show $f(X+1)$ safeties Eisenstein's criterion. My argument is that
$$f(X+1) = (X+1)^{(p-1)p^{n-1}} + (X+1)^{(p-2)p^{n-1}} + \cdots + (X+1)^{p^{n-1}} + 1$$

I consider term by term, obviously, the coefficient of the highest order term is $1$, so $p$ does not divide $1$, the coefficient of the constant term is $p$, so $p^2$ does not divide $p$, for the middle terms, I observe that $p$ can mostly divide them, but for the first terms of each term of $f(X+1)$, coefficients for them are all $1$ ($p$ does not divide $1$) and I cannot easily find other equal order terms with them to observe the total coefficient, how should I solve it?

$\endgroup$
  • $\begingroup$ "but for the first terms of each term of $f(X+1)$, coefficients for them are all $1$ ($p$ does not divide $1$)", are you sure about that? Do you have some example of this? $\endgroup$ – Sil Apr 20 at 22:41
  • $\begingroup$ look at the second term, bu using binomial theorem, the first term of it is x^(p^n - 2p^(n-1)). so the coefficient of it is 1. p does not divide it. $\endgroup$ – Jonny Apr 20 at 22:46
  • $\begingroup$ I am not sure what you mean by second term and first term here, can you write down the full polynomial that satisfies this? For example for $p=3,n=2$ we have $f(x+1)=x^6+6x^5+15x^4+21x^3+18x^2+9x+3$, I don't see the thing you describe. $\endgroup$ – Sil Apr 20 at 22:47
  • $\begingroup$ I mean, if I take p=3, the first term of (X+1)$^{(P-2)P^{n-1}}$ is x$^{3^n - 2*3^{n-1}}$ $\endgroup$ – Jonny Apr 20 at 22:54
  • $\begingroup$ I see, well that is just a part of the polynomial, you need to collect all coefficients (possibly by Binomial theorem), then it won't be just $1$, and Eisenstein will be applicable. $\endgroup$ – Sil Apr 20 at 22:56
1
$\begingroup$

Here is something that could get you started (too long for a comment), write out the polynomial as:

\begin{align*} f(X+1) &= (X+1)^{(p-1)p^{n-1}} + (X+1)^{(p-2)p^{n-1}} + \cdots + (X+1)^{p^{n-1}} + 1\\ &= \sum_{k=0}^{p-1}(X+1)^{kp^{n-1}}\\ &= \sum_{k=0}^{p-1}\sum_{i=0}^{kp^{n-1}}\binom{kp^{n-1}}{i}X^i\\ &= \sum_{i=0}^{(p-1)p^{n-1}}\sum_{k=\lceil i/p^{n-1}\rceil}^{p-1}\binom{kp^{n-1}}{i}X^i\\ \end{align*} where we just swapped the indices by writing the conditions $0\leq k \leq p-1$, $0 \leq i \leq kp^{n-1}$, and rewrote into equivalent $0\leq i \leq (p-1)p^{n-1}$,$i/p^{n-1}\leq k \leq p-1$. So coefficient of $X^i$ for $i>0$ can be seen to be $\sum_{k=\lceil i/p^{n-1}\rceil}^{p-1}\binom{kp^{n-1}}{i}$. Now for Eisenstein you need to show divisibility of this sum by $p$ for $i<(p-1)p^{n-1}$.

However I have noticed that in some cases it does not seem to be straightforward to prove, for example for $p=3$, $n=2$ and coefficient at $x^3$ we get get sum $1+20=21$, which is divisible by $3$ but individual terms in sum are not. So something else is needed.

$\endgroup$
  • $\begingroup$ thanks for your comments! $\endgroup$ – Jonny Apr 21 at 4:01
2
$\begingroup$

My way of looking at this very special problem:

Showing that all lower coefficients are divisible by $p$ is the easy half of the problem.

Your original polynomial is $$\frac{X^{p^n}-1}{X^{p^{n-1}}-1}\,. $$ and after the substitution it becomes $$ \frac{(X+1)^{p^n}-1}{(X+1)^{p^{n-1}}-1}\,. $$ Now look at this modulo $p$, i.e. as a polynomial over $\Bbb F_p$, the prime field. The numerator becomes $X^{p^n}+1-1=X^{p^n}$, and similarly the denominator becomes $X^{p^{n-1}}$, so that the quotient has only one term not divisible by $p$.

To see that the constant term is $p$, look at numerator and denominator in the second display: both have constant term zero, so that you have a polynomial of the form $$ \frac{X^{p^n}+\cdots+p^nX}{X^{p^{n-1}}+\cdots+p^{n-1}X}\,, $$ and this polynomial clearly has constant term $p$.

(I’m guessing that you came across this as the cyclotomic polynomial $\Phi_{p^n}(X)$. )

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.