Identity in Number Theory Paper

466 Views Asked by At

In this paper by Jerry Hu, he defines the function

$$f_{s,k,i}\left(u\right)=\prod_{p\mid u} \left(1-\frac{\sum_{m=i}^{k-1}{s \choose m}\left(p-1\right)^{k-1-m}}{\sum_{m=0}^{k-1}{s \choose m}\left(p-1\right)^{k-1-m}}\right)$$

for use in the main theorem.

In Lemma 4 (pages 3 & 4), it is claimed that

$$\frac{f_{s,k,i}\left(u_{i}\right)}{f_{s,k,i+1}\left(u_{i}\right)}=\sum_{d\mid u_{i}}\frac{\mu\left(d\right){s \choose i}^{\omega\left(d\right)}}{\alpha_{s,k,i}\left(d\right)}$$

for $i=1,2,\ldots k-2$, where $\mu$ is the Möbius function, $\omega\left(d\right)$ is the number of distinct prime factors of $d$, and

$$\alpha_{s,k,i}\left(d\right)=d^{i}\prod_{p\mid d}\sum_{m=0}^{i}{s \choose m}\left(1-p^{-1}\right)^{i-m}p^{-m}$$

for $i=1,2,\ldots k-1$.

Obviously $f_{s,k,i}$ is multiplicative and, hence, so is $\frac{f_{s,k,i}}{f_{s,k,i+1}}$, which means that it is enough to show this for $u_{i}=p^{a}$ for integers $a\geq0$.

I've only been able to go so far on either side of the equality, but I've hit a wall.

Starting from the left hand side, since the only prime dividing $p^{a}$ is $p$ itself, I have:

$$\frac{f_{s,k,i}\left(p^{a}\right)}{f_{s,k,i+1}\left(p^{a}\right)}=\frac{1-\frac{\sum_{m=i}^{k-1}{s \choose m}\left(p-1\right)^{k-1-m}}{\sum_{m=0}^{k-1}{s \choose m}\left(p-1\right)^{k-1-m}}}{1-\frac{\sum_{m=i+1}^{k-1}{s \choose m}\left(p-1\right)^{k-1-m}}{\sum_{m=0}^{k-1}{s \choose m}\left(p-1\right)^{k-1-m}}}= \\ \frac{\frac{\sum_{m=0}^{i-1}{s \choose m}\left(p-1\right)^{k-1-m}}{\sum_{m=0}^{k-1}{s \choose m}\left(p-1\right)^{k-1-m}}}{\frac{\sum_{m=0}^{i}{s \choose m}\left(p-1\right)^{k-1-m}}{\sum_{m=0}^{k-1}{s \choose m}\left(p-1\right)^{k-1-m}}} = \frac{\sum_{m=0}^{i-1}{s \choose m}\left(p-1\right)^{k-1-m}}{\sum_{m=0}^{i}{s \choose m}\left(p-1\right)^{k-1-m}}.$$

On the other hand, since $\mu\left(1\right)=1$, $\mu\left(p\right)=-1$, and $\mu\left(p^{a}\right)=0$ for $a>1$, as well as since $\omega\left(1\right)=0$, $\omega\left(p^{a}\right)=1$ for $a\geq1$, and $\alpha_{s,k,i}\left(1\right)=1$, I have:

$$\sum_{d\mid p^{a}}\frac{\mu\left(d\right){s \choose i}^{\omega\left(d\right)}}{\alpha_{s,k,i}\left(d\right)}=1-{s \choose i}\left(\alpha_{s,k,i}\left(p\right)\right)^{-1} = \\ 1-{s \choose i}\left(p^{i}\sum_{m=0}^{i}{s \choose m}\left(1-p^{-1}\right)^{i-m}p^{-m}\right)^{-1} = \\ 1-{s \choose i}\left(p^{i}\sum_{m=0}^{i}{s \choose m}\left(p-1\right)^{i-m}p^{-i}\right)^{-1} = \\ 1-{s \choose i}\left(\sum_{m=0}^{i}{s \choose m}\left(p-1\right)^{i-m}\right)^{-1}.$$

I can't see how to proceed from here. Assuming I haven't made any errors, what should I do to fill in the gap in

$$\frac{\sum_{m=0}^{i-1}{s \choose m}\left(p-1\right)^{k-1-m}}{\sum_{m=0}^{i}{s \choose m}\left(p-1\right)^{k-1-m}}=\overset{?}{\cdots}=1-{s \choose i}\left(\sum_{m=0}^{i}{s \choose m}\left(p-1\right)^{i-m}\right)^{-1}?$$

The second part of this lemma looks similar, so I think I can do that one once I've understood how to move forward in this one.

EDIT: I somehow missed an important part of the definition of $f_{s,k,i}$ when typing this question, and this continued throughout due to my copying and pasting. This has been fixed. I'm sorry for not having noticed this earlier.

1

There are 1 best solutions below

3
On BEST ANSWER

Here is one way to do the calculation:

\begin{align*} \frac{\sum_{m=0}^{i-1}\binom{s}{m}(p-1)^{k-1-m}}{\sum_{m=0}^{i}\binom{s}{m}(p-1)^{k-1-m}}&= \frac{\sum_{m=0}^{i-1}\binom{s}{m}(p-1)^{-m}}{\sum_{m=0}^{i}\binom{s}{m}(p-1)^{-m}}\tag{1}\\ &=\frac{\sum_{m=0}^{i}\binom{s}{m}(p-1)^{-m}-\binom{s}{i}(p-1)^{-i}}{\sum_{m=0}^{i}\binom{s}{m}(p-1)^{-m}}\tag{2}\\ &=1-\binom{s}{i}(p-1)^{-i}\left(\sum_{m=0}^{i}\binom{s}{m}(p-1)^{-m}\right)^{-1}\tag{3}\\ &=1-\binom{s}{i}\left(\sum_{m=0}^{i}\binom{s}{m}(p-1)^{i-m}\right)^{-1}\tag{4}\\ \end{align*}

Comment:

  • In (1) we divide by $(p-1)^{k-1}$

  • In (2) we add the summand $m=i$ to the sum in the numerator and subtract the corresponding value

  • In (3) we perform the division

  • In (4) there is a last small rearrangement

Note: According to your precise description you did most of the analysis of Lemma 4 and I suppose that this calculation will look therefore pretty easy to you.