I want to prove the following lemma.

Let $p$ be a prime number and $G$ a finite abelian group such that $p\mid\#G$, then $G$ has an element of order $p$.

Here's the proof my professor presented.

Let $|G|=pm$ for some $m\in\mathbb{N}$. We will prove the result by induction on $m$.

The base case, $m=1$, is clear (every group of prime order is cyclic).

For the inductive step, choose $x\in G$ such that $\text{ord}(x)=t>1$.

If $p\mid t$ (don't we need to require that $p\neq t$?), then $\text{ord}\big((x^{\frac{t}{p}})^{p}\big)=1$ which implies that $\text{ord}(x^{\frac{t}{p}})=p$, and we are done.

Now suppose that $p\nmid t$. Since $G$ is a abelian then every subgroup is normal, in particular $\langle x\rangle\trianglelefteq G$ and $|G/\langle x\rangle|=\frac{pm}{t}$.

By hypothesis, $\exists y\in G$ such that $\text{ord}(y\langle x\rangle)=p$. (why?) Since for the homomorphism $f:G\rightarrow G/\langle x\rangle$ we have that $\text{ord}\big(f(y)\big)\mid \text{ord}(y)$ it follows that $\text{ord}(y\langle x\rangle)=p\mid \text{ord}\;y$ and we go back to the case where $p$ divides the order or a nontrivial element of $G$.

My questions:

1) the first questions appear in bold in the above text.

2) I don't really get the structure of this induction proof. When the inductive step is presented I don't see in what way we assume that the results holds for $m$ and how we show it is true for $m+1$.

My apologies if these questions are too basic but I am a former physics student and lack some background and experience with proofs.

  • $\begingroup$ Is $p$ a prime number? $\endgroup$ – paw88789 Oct 17 '17 at 16:20
  • $\begingroup$ Yes, I forgot to include this. Thanks. $\endgroup$ – aadcg Oct 17 '17 at 16:21
  • $\begingroup$ For the first bold part: Not needed since if $p=t$ all the statements that follow hold by definition of $x$. $ord((x^{t/p})^p)=ord(x^p)=ord(e)=1$. For the second bold part: Observe that $G/(x)$ has order $p(m/t)$ and $m/t < m$. Therefore, $G/(x)$ has an element of order $p$. That element is being called $y(x)$, the coset of some element $y\in G$. $\endgroup$ – Hellen Oct 17 '17 at 16:29
  • $\begingroup$ In finite abelian groups,Converse of Lagrange's theorem holds good. $\endgroup$ – Sumit Mittal Oct 17 '17 at 17:26

don't we need to require that $p \neq t$?

Not that I can see from your argument. If $\operatorname{ord}(x) = p$, then there is nothing to prove. That said, the way you currently have structured your argument, you will need to ensure $p \nmid t$ when applying the inductive hypothesis to the group $G / \langle x \rangle$.

One way around this would be to replace your current definition of $x$ with something more specific. For instance, you could say

Choose any $x \in G \setminus \{ 1 \}$. If $p \mid \operatorname{ord}(x)$, say $\operatorname{ord}(x) = pq$ for $q \in \mathbb{N}$, then by definition $y = x^q \in G$ is an element of order $p$. Conversely, suppose $p \nmid \operatorname{ord}(x)$ [$\ldots$]

and take things from there. This seems to be what you mean, but calling out the two possible cases more explicitly goes a long way towards clarifying the rest of the argument (for both the reader and you!).

By hypothesis, $\exists y \in G$ such that $\operatorname{ord}(y\langle x\rangle)=p \operatorname{ord}(y\langle x\rangle)=p$. (why?)

If you assume $p \nmid t$, then this is because $|G/\langle x \rangle|$ is of the form $pm'$ where $m' < m$ (and so the inductive hypothesis applies).

Regarding your last question: It sounds like you're assuming that this is a proof by weak induction, when this is really a proof by strong induction. Suppose $P$ is a property of the natural numbers, and you know that $P(1)$ is true. The difference in proving that $P(N)$ is true for all $N \in \mathbb{N}$ using weak and strong induction is the following:

  • Using weak induction, you assume for a fixed $N \in \mathbb{N}$ that $P(N)$ holds and show that this implies $P(N + 1)$.
  • Using strong induction, you assume for a fixed $N \in \mathbb{N}$ that $P(n)$ holds for all $n \leq N$ and show that these statements together imply $P(N + 1)$ holds.

See this set of notes for examples of both types of proofs. Each of these is a valid form of induction.

Your proof relies on strong induction, and you can see this from how you apply your inductive hypothesis. First, you choose an arbitrary $x \in G$ with $x \neq 1$ and $p \nmid t := \operatorname{ord}(x)$. You then apply the inductive hypothesis to the group $G/\langle x \rangle$. Supposing that $|G/\langle x \rangle| = pk$, you don't know the value of $k$, since you don't know enough about $x$. All you know is that it is less than $m$ where $|G| = pm$. To cover all bases, you have to assume the claim holds for all finite abelian groups of order $pn$ for $n < m$, not just for those of order $p(m - 1)$.

  • $\begingroup$ Very clear and informative. $\endgroup$ – aadcg Nov 10 '17 at 0:25

1) If $p=t$, $t/p=1$, so $x$ has order $p$ anyway and is the element we want.

2) Your second piece of bold is where the inductive step is used: the key point is that $G/\langle x \rangle$ is a group of order $pm/t$, and if $p \not\mid t$, then there is an integer $n$, smaller than $m$, so that $pm/t = pn$. The induction hypothesis then says the group $G/\langle x \rangle$ has an element of order $p$; and then the structure of the quotient group means that there is $y \in G$ so that this element can be written as $y\langle x \rangle$.


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.