8

A question from Advanced Modern Algebra by Joseph J.Rotman.

Let $n=(p^r)m $ such that the prime $p\nmid m$.Prove that $p\nmid \dbinom{n}{p^r}$.HINT: Assume otherwise, cross multiply and apply Euclid's lemma.

I followed the hint and got

$n!=p^r!(n-p^r)!\dbinom{n}{p^r}$

Now using Euclid's lemma, we conclude that since p divides the left side,it must also divide the right side. But we already know from our primary assumption that p divides $\dbinom{n}{p^r}$. But it must also divide $p^r!(n-p^r)!$. I am lost at this point and cannot seem to derive the contradiction as hinted in the hint.

I would greatly appreciate it if you guys could nudge me in the right direction.

rah4927
  • 3,864

3 Answers3

12

Let's try another proof using polynomials that I think is cool :

Consider the polynomial $$ (1+X)^p - (1+X^p) $$ over the field $\mathbb{Z}_p$. We know that $a^p \equiv a\pmod{p}$ for all $a\in \mathbb{Z}_p$, so this is a polynomial of degree $p-1$, that has $p$ roots. This is only possible if it is identically zero. Hence, $$ (1+X)^p \equiv (1+X^p)\pmod{p} $$ By induction, we see that $$ (1+X)^{p^r} \equiv (1+X^{p^r})\pmod{p} $$ Thus $$ (1+X)^{p^rm} \equiv (1+X^{p^r})^m\pmod{p} $$ Hence the corresponding coefficients of both these polynomials must be equal $\pmod{p}$. Consider the coefficient of $X^{p^r}$ on both sides to conclude that $$ {p^r m \choose p^r } \equiv m\pmod{p} $$ which proves what you want.

  • Nice,but I was wondering why everyone is so desperate to avoid Euclid's lemma and use the hint given in the book.Also,how did you come up with this solution?What triggered the idea of considering that particular polynomial? – rah4927 Dec 03 '13 at 13:41
  • Because this is easier! You can of course write out $${n\choose p^r} = \frac{p^rm(p^rm - 1)(p^rm - 2)\ldots (p^rm - p^r + 1)}{p^r(p^r-1)(p^r-2)\ldots (p^r - p^r+1)}$$ and use the fact that the power of $p$ dividing $(p^rm -i)$ is the same as the one that divides $(p^r-i)$ – Prahlad Vaidyanathan Dec 03 '13 at 13:44
  • 1
    The fact that $(x+y)^p = x^p+y^p$ in $\mathbb{Z}_p$ is the real idea here (This is called the Frobenius homomorphism). The polynomial merely shows up as a result of that. – Prahlad Vaidyanathan Dec 03 '13 at 13:46
  • 1
    Nice proof (+1). Equation $(2)$ in my answer says that the number of factors of $p$ in $\binom{mp^r}{p^r}$ is the same as the number of factors of $p$ in $m$. This is only stronger than $\binom{mp^r}{p^r}\equiv m\pmod{p}$ when $p\mid m$; that is, not here :-) – robjohn Dec 03 '13 at 13:53
11

Here is a different approach.

For a prime $p$, the Corollary below says that the number of factors of $p$ that divide $\binom{n}{k}$ is $$ \frac{\sigma_p(k)+\sigma_p(n-k)-\sigma_p(n)}{p-1}\tag{1} $$ where $\sigma_p(n)$ is the sum of the digits in the base-$p$ represention of $n$.

For any $m$, $\sigma_p\left(mp^r\right)=\sigma_p\left(m\right)$. Therefore, $$ \frac{\sigma_p\left(p^r\right)+\sigma_p\left((m-1)p^r\right)-\sigma_p\left(mp^r\right)}{p-1} =\frac{\sigma_p(1)+\sigma_p(m-1)-\sigma_p(m)}{p-1}\tag{2} $$ Thus, if $p\not\mid m$, the low order digit of $m$ in base-$p$ is not $0$ and so the base-$p$ representation of $m-1$ is the same as the base-$p$ representation of $m$ with the lowest order digit reduced by $1$. Thus the quantity on the right side of $(2)$ is $0$.

QED


Lemma (Legendre's Formula): The number of factors of a prime $p$ that divide $n!$ is $$ \frac{n-\sigma_p(n)}{p-1}\tag{3} $$ Proof: Note that the number of multiples of $p^j$ not greater than $n$ is $\left\lfloor\frac n{p^j}\right\rfloor$. Thus, to compute the number of factors of $p$ that divides $n!$ we count the number of multiples of $p$ not greater than $n$, then add the number of multiples of $p^2$ not greater than $n$ (we add them once since they have already been counted once with the multiples of $p$), then add the number of multiples of $p^3$ not greater than $n$ (we add them once since they have already been counted once with the multiples of $p$ and once with the multiples of $p^2$), and so forth. That is, the number of factors of $p$ in $n!$ is $$ \sum_{j=1}^k\left\lfloor\frac n{p^j}\right\rfloor\tag{4} $$ where $k$ is the number of digits in the base-$p$ representation of $n$.

Now suppose that $$ n=\sum_{i=0}^kd_ip^i\tag{5} $$ Then $$ \begin{align} \sum_{j=1}^k\left\lfloor\frac n{p^j}\right\rfloor &=\sum_{j=1}^k\sum_{i=j}^kd_ip^{i-j}\\ &=\sum_{i=1}^k\sum_{j=1}^id_ip^{i-j}\\ &=\sum_{i=1}^kd_i\frac{p^i-1}{p-1}\\ &=\sum_{i=0}^kd_i\frac{p^i-1}{p-1}\\ &=\frac{n-\sigma_p(n)}{p-1}\tag{6} \end{align} $$ QED

Corollary (Kummer's Theorem): The number of factors of $p$ that divide $\binom{n}{k}$ is $$ \frac{\sigma_p(k)+\sigma_p(n-k)-\sigma_p(n)}{p-1}\tag{7} $$ Proof: Note that $$ \binom{n}{k}=\frac{n!}{k!(n-k)!}\tag{8} $$ and apply the Lemma to each term in the numerator and denominator of $(8)$.

QED

Formula $(7)$ says that the number of factors of $p$ in $\binom{n}{k}$ is the number of carries when adding $k$ and $n-k$ in base-$p$.

robjohn
  • 345,667
4

Using rational numbers and rearranging the way to multiply, it is immediate that:

$$ {p^rm\choose p^r} = \frac {p^rm} {p^r} \cdot \frac {p^rm -1} {p^r -1} \cdot ... \cdot \frac {p^rm -p} {p^r -p} \cdot ... \cdot \frac {p^rm -kp^a} {p^r -kp^a}\cdot ... \cdot\frac {p^rm -p^r+1} {1}$$

$$ = m \cdot \frac {p^rm -1} {p^r -1} \cdot ... \cdot \frac {p^{r-1}m -1} {p^{r-1} -1}\cdot ... \cdot \frac {p^{r-a}m -k} {p^{r-a} -k} \cdot ... \cdot (p^rm -p^r+1) $$

$$ = q_0 \cdot q_1 \cdot ... \cdot q_p \cdot ... \cdot q_{kp^a} \cdot ... \cdot q_{p^r-1}$$

All $q_n$ are free of $p$ in the numerator and denominator when p does not divide m or those k. And therefore $p$ does not divide that combinatorial.