4

The textbook I am reading (A Friendly Introduction to Number Theory) is proving that when $m$ and $n$ are coprime $ \phi(mn) = \phi(m)\phi(n)$ by defining two sets:

  1. All of the numbers a where $a$ and $mn$ are coprime (where $1 \le a < mn$)
  2. All pairs of numbers (b, c) where $b$ and $m$ are coprime and $c$ and $n$ are coprime ($1 \le b < m$ and $1 \le c < n$)

Image of textbook

Image of textbook

and then proving that their sizes are equal.

It does this in two steps:

  1. "Different numbers in the first set get sent to different pairs in the second set."
  2. "Every pair in the second set is hit by some number in the first set."

The second part is the one I don't understand. To prove it, the book just defines the Chinese Remainder Theorem and declares the proof complete. But as far as I understand, the Chinese Remainder Theorem defines a bijection between $\mathbb Z_{pq}$ and $\mathbb Z_p \times \mathbb Z_q$, and not the two sets they've defined above.

How is it proven that the number "hit" in the second set actually came from the first set and not some other element in $\mathbb Z_{pq}$ ?

Image of textbook

thea
  • 45
  • Hi, welcome to Math SE. Thanks for trying to use MathJax. Please put a dollar sign on each side of each intended inline math environment (use two per side for display-line environments). – J.G. Sep 20 '22 at 06:48
  • $\phi(n)$ is not the set of all numbers coprime to $n$ , it is the number of those numbers in the range $[1,n-1]$ – Peter Sep 20 '22 at 06:49
  • The community appreciates your efforts to type in mathjax. Please note that you need to enclose the formulae in dollar (“$$…$$”) signs for the formatting to work. If you haven’t already, you can refer here for further syntax: https://math.meta.stackexchange.com/questions/5020/mathjax-basic-tutorial-and-quick-reference – insipidintegrator Sep 20 '22 at 06:49
  • @Peter Yes, sorry I meant the numbers in that range – thea Sep 20 '22 at 06:50
  • 3
    should we guess what $\varphi $ is ? I guess it's Euler totient function... but in this case, what means : "numbers that are in $\varphi (nm)$" or "numbers that are in $\varphi (n)$" ? $\varphi (n)$ is not a set... – Surb Sep 20 '22 at 06:51
  • Yes it is Euler's totient function. Sorry, I know it isn't a set, I will rephrase it. hopefully you know what I mean. – thea Sep 20 '22 at 06:55
  • 2
    We also need $n$ and $m$ to be coprime. – PM 2Ring Sep 20 '22 at 06:57
  • @Will Sherwood what are $m'$ and $n'$ in your comment? – thea Sep 20 '22 at 07:00
  • Bah yeah, I mixed up the definition of $\phi$. My comment above is for divisors, not # of coprime. Sorry for the confusion – Will Sherwood Sep 20 '22 at 07:12
  • I deleted the comment since it was wrong – Will Sherwood Sep 20 '22 at 07:13
  • 2
    The key fact about the Chinese remainder theorem is that the map $\mathbb{Z}{mn}\to\mathbb{Z}{m}\times\mathbb{Z}{n}$, $[a]{mn}\mapsto([a]_m,[a]_n)$ is not only well defined, but also a ring isomorphism, so it induces a bijection between the sets of invertible elements. Note: $[a]_m$ denotes the residue class of $a$ modulo $m$. – egreg Sep 20 '22 at 07:21

2 Answers2

5

Elementary Proof

Let $p$ and $q$ be coprime. The Chinese remainder theorem defines a bijection $\psi:\mathbb{Z}_p\times\mathbb{Z}_q\rightarrow\mathbb{Z}_{pq}$ given by $\psi(a,b)=qa+pb\pmod{pq}$. The crux, and missing detail of the proof in your book is that (LEMMA) for any $(a,b)\in\mathbb{Z}_p\times\mathbb{Z}_q$, we have that $\gcd(qa+pb,pq)=1$ iff $\gcd(a,p)=\gcd(b,q)=1$. This lemma eqivalently tells us that $\psi$ restricts to a bijection between from the set $\{(a,b)\in\mathbb{Z}_a\times\mathbb{Z}_b:\gcd(a,q)=\gcd(b,p)=1\}$ to the set $\{c\in\mathbb{Z}_{pq}:\gcd(c,pq)=1\}$, and thus that $\varphi(p)\varphi(q)=\varphi(pq)$. We prove this lemma below.

First, we prove $(\Rightarrow)$. Let $\gcd(qa+pb,pq)=1$, then if $n$ is a divisor of both $a$ and $p$, it must also divide $qa+pb$ and $pq$, so $n$ would have to divide $\gcd(qa+pb,pq)=1$, meaning that $n=1$. Since the only natural $n$ that divides both $a$ and $p$ is $n=1$, then $\gcd(a,p)=1$. We can repeat this same argument to show that $\gcd(b,q)=1$.

We now prove $(\Leftarrow)$. Let $\gcd(a,p)=\gcd(b,q)=1$. Since $q,a$ are both coprime to $p$ and $p,b$ are both coprime to $q$, then $\gcd(qa,p)=\gcd(pb,q)=1$. Since $p,q$ are coprime, then \begin{equation} \begin{split} \gcd(qa+pb,pq)&=\gcd(qa+pb,p)\gcd(qa+pb,q)\\ &=\gcd(qa,p)\gcd(pb,q)\\ &=1 \end{split} \end{equation} by the properties of the $\gcd$. This completes our proof of the lemma.

More Abstract Proof

Let's first define a useful concept. For any ring $R$, we may define $R^\times$ to be the set of units (elements with an inverse) of $R$; this is in fact a group under the multiplication operator inherited from $R$, and is known as the unit group of $R$.

It is not hard to see that the units of $\mathbb{Z}_n$ are simply the elements $[k]\in\mathbb{Z}_n$ such that $k$ is coprime to $n$. In other words, $|\mathbb{Z}_n^\times|=\varphi(n)$.

Now, the Chinese remainder theorem tells us that if $p,q$ are coprime, then there exists a ring isomorphism $\psi:\mathbb{Z}_{pq}\rightarrow\mathbb{Z}_p\times\mathbb{Z}_q$. Since $\psi$ is a ring isomorphism, then it must also restrict to a group isomorphism $\mathbb{Z}_{pq}^\times\rightarrow(\mathbb{Z}_p\times\mathbb{Z}_q)^\times$. Combining this with the fact that $(\mathbb{Z}_p\times\mathbb{Z}_q)^\times\cong\mathbb{Z}_p^\times\times\mathbb{Z}_q^\times$, we have that $\mathbb{Z}_{pq}^\times\cong\mathbb{Z}_p^\times\times\mathbb{Z}_q^\times$, and thus \begin{equation} \varphi(pq)=|\mathbb{Z}_{pq}^\times|=|\mathbb{Z}_p^\times|\cdot|\mathbb{Z}_q^\times|=\varphi(p)\varphi(q) \end{equation} as desired.

  • 1
    But the cited book (and proof) is far more elementary and makes no mention of rings (as is often true for textbooks on elementary number theory). So, technically, the above goes far beyond the knowledge of the reader. – Bill Dubuque Sep 20 '22 at 18:25
  • 1
    I appreciate your response but unfortunately Bill is correct in that your proof takes a more background knowledge than I currently have. For now, I am just trying to understand the missing part of the proof I have before moving on to other concepts. – thea Sep 20 '22 at 18:43
  • @thea Sorry, I didn't consider that. When you mentioned the Chinese remainder theorem, I completely forgot there was an elementary version of the theorem that your book was probably referencing. I'll delete or improve my answer a bit later in the day. – Christian E. Ramirez Sep 20 '22 at 19:16
  • 1
    No worries! I am still just a newbie. Hopefully in a few chapters I'll be able to come back and understand this "real" proof – thea Sep 20 '22 at 19:23
  • 1
    I have opted to add an elementary answer, and leave the more abstract proof in case you do want to come back to it in the future, or in case it's useful to anyone else. – Christian E. Ramirez Sep 20 '22 at 21:19
  • 1
    @thea Don't worry, both proofs are real. – J.G. Sep 20 '22 at 21:21
  • 1
    @C-RAM thank you!! Quick question though: why is it (,)=+(mod) and not $\psi (a,b) = qq_1a + pp_1b$ (mod $pq$) like on this website? https://crypto.stanford.edu/pbc/notes/numbertheory/crt.html $q_1 = q^{-1}$ (mod $p$) and $p_1 = p^{-1}$ (mod $q$). Are they the same thing? – thea Sep 21 '22 at 00:20
  • @thea There are multiple possible bijections (group isomorphisms) that work. That link uses a more standard one, but I used this one because I think it makes things more clear in the proof. If you only know that $\psi^(a,b)=qq_1a+pp_2b\pmod{pq}$ is a bijection, then all you need to know is that since $\gcd(q,p)=1$, then the map $a\rightarrow pa\pmod{q}$ is a bijection from $\mathbb{Z}_q$ to itself, and similarly the map $b\rightarrow qb\pmod{p}$ is a bijection from $\mathbb{Z}_p$ to itself; therefore my map given by $\psi(a,b)=qa+pb\pmod{pq}=\psi^(qa,pb)$ is also a bijection. – Christian E. Ramirez Sep 21 '22 at 02:30
  • 1
    @thea The "elementary" proof above is quite different than the proof Silverman presents in his book. I added an answer showing how to complete SIlverman's proof in a way that is faithful to the book. – Bill Dubuque Sep 21 '22 at 02:56
  • @BillDubuque You're probably right. I think the biggest issue with both my previous and my updated answer is that I do not own the book... I hope the "elementary" proof I presented is still understandable at that level. $\tag*{}$Anyways, it seems strange that an elementary number theory text would handwave that away. Neither of our proofs seem like something I would expect someone at that level to just instantly intuit. – Christian E. Ramirez Sep 21 '22 at 03:27
  • 1
    I don't think it was meant to be handwaved away. Rather, it was likely simply an oversight. One of the biggest challenges of teaching is mastering how to forget advanced knowledge beyond the student's level (since that works against decades of training one's mind to exploit such expertise). That said, it is surprising that such a serious pedagogical gap still exists even in the 4th edition. – Bill Dubuque Sep 21 '22 at 03:34
1

Your critique is correct. Silverman's proof has a gap: it doesn't prove the bijection restricts to the sets of coprimes, i.e. that the coprimality constraints are preserved. But the gap is easily filled via

$\qquad\color{#c00}{\rm L1}\!:\ \ \ (a,mn)\! =\! 1\iff (a,m)\! =\! 1 \!=\! (a,n).\ $ Proof

$\qquad\color{#0a0}{\rm L2}\!:\ \ \ j\equiv k\pmod{\! n}\Rightarrow (j,n) = (k,n).\qquad $Proof

In step $1\!:\ $ $a\in {\rm set_1}\!\Rightarrow (a,mn)\!=\!1 \overset{\rm\color{#c00}{L1}}\Rightarrow 1\!=\!(a,m)\overset{\rm\color{#0a0}{L2}}=(a\bmod m,m);\;$ by $\,m\leftrightarrow n\,$ symmetry this also yields $\,(a\bmod n,n)\!=\!1,\,$ so $\,(a\bmod m,\,a\bmod n) \in {\rm set}_2$

In step $2\!:\ $ if $\,(b,c) \in {\rm set_2}$ then by CRT there is $\,1\le a < mn\,$ with $\,\begin{align}&a\equiv b\!\!\!\pmod{\! m}\\ &a\equiv c\!\!\!\pmod{\! n}\end{align}\,$ so $\,(a,m)\overset{\rm\color{#0a0}{L2}}=(b,m)\!=\!1\!=\!(c,n)\overset{\rm\color{#0a0}{L2}}=\!(a,n)\overset{\rm\color{#c00}{L1}}\Rightarrow (a,mn)\!=\!1\Rightarrow a \in {\rm set}_1$

Remark $ $ As explained in the linked proof, an arithmetically conceptual way to view Lemma $\rm\color{#c00}{L1}$ is via the obvious fact that units (invertibles) $\!\bmod n\,$ are clearly preserved under products and divisors (vs. the common brute-force computational Bezout-based proof that is - alas - usually unmotivated, i.e. pulled out of a hat like magic).

The innate arithmetical structure at the heart of the matter will be better understood structurally when one learns the ring-theoretic view of CRT in terms of product rings, where the above becomes $$(m,n)=1\Rightarrow \Bbb Z_{mn}\cong \Bbb Z_m\times \Bbb Z_n\,\Rightarrow\, U(\Bbb Z_{mn})\cong U(\Bbb Z_m)\times U(\Bbb Z_n)\qquad$$ where $U(R)$ denote the group of units (invertibles) of the ring $R.\,$ But this is far beyond the level of Silverman's book (which never uses any ring theory).

Bill Dubuque
  • 272,048