Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Let $G$ be any abelian group and $a\in{G}$. Show there exists a homomorphism $f:G\rightarrow{\mathbb{Q}/\mathbb{Z}}$ such that $f(a)\neq{0}$.

I can prove this question (I think) if I use the fact that $\mathbb{Q}/\mathbb{Z}$ is an injective abelian group: just define $f$ on the cyclic subgroup generated by $a$ and extend to $G$ via injectivity. However, I feel there should be a more 'elementary' way to prove this result, I just can't see one yet. One of my friends suggested using Zorn's Lemma, but I haven't done much with this information yet.

share|improve this question
    
What do you mean by "elementary"? –  Makoto Kato Oct 15 '12 at 20:44
    
maybe that was the wrong word to use, basically it's from an exercise in a book that asks this question as part (a) and then asks to prove $\mathbb{Q}/\mathbb{Z}$ is injective later in the question, so I really mean without the use of any homological algebra. A proof using group theory or set theory (like the Zorn's lemma suggestion) is what I meant when I said elementary. –  Paul Gilmartin Oct 15 '12 at 20:47
1  
Of course, the group $\,G\,$ must be non-trivial and also $\,a\neq 0\,$ –  DonAntonio Oct 15 '12 at 21:40
add comment

2 Answers

up vote 7 down vote accepted

Here is a proof using Zorn's lemma: Define a non-trivial map $f_0: \langle a \rangle \to \mathbb Q/\mathbb Z$ on the cyclic subgroup generated by $a$, and consider pairs $(U,f)$ where $U$ is a subgroup of $G$ containing $a$ and $f: U \to \mathbb Q/\mathbb Z$ is a homomorphism extending $f_0$. Those pairs form a partially ordered set with the relation $$(U,f) \leq (V,g) \Leftrightarrow U \subseteq V \text{ and } g\vert_U = f.$$ Clearly, every chain of such pairs has an upper bound. So, by Zorn's lemma, there is a maximal element $(U,f)$. We will show $U = G$. Suppose for contradiction that there is a $x \in G\setminus U$. We are finished if we can extend $f: U \to \mathbb Q/\mathbb Z$ to a homomorphism $\hat f: U + \langle x \rangle \to \mathbb Q/\mathbb Z$ since that would contradict the maximality of $(U,f)$. If $U \cap \langle x \rangle = 0$, the sum $U + \langle x \rangle$ is direct and we can just define $\hat f$ to be trivial on $x$. Otherwise, we have $U \cap \langle x \rangle = \langle nx \rangle$ for some integer $n \neq 0$ dividing the order of $x$. Our map $\hat f$ has to satisfy $$n \hat f(x) = \hat f(nx) = f(nx),$$ so we take any $\alpha \in \mathbb Q/\mathbb Z$ satisfying $n \alpha = f(nx)$ (which exists as $\mathbb Q/\mathbb Z$ is divisible) and define $\hat f(x) := \alpha$. Using that $n$ divides the order of $x$, check that this extends to a unique homomorphism $\hat f: \langle x \rangle \to \mathbb Q/\mathbb Z$. It coincides with $f: U \to \mathbb Q/\mathbb Z$ on the intersection $U \cap \langle x \rangle = \langle nx \rangle$, so the two can be put together to a homomorphism $\hat f: U + \langle x \rangle \to \mathbb Q/\mathbb Z$. This is then a well-defined homomorphism extending $f$, giving the desired contradiction.

share|improve this answer
    
Very nice, thanks! –  Paul Gilmartin Oct 15 '12 at 22:32
    
It seems to be true (I think) that this proof only really needs the divisibility of $\mathbb{Q}/\mathbb{Z}$, which is equivalent to injectivity for abelian groups, so should hold for all injective abelian groups. –  Paul Gilmartin Oct 15 '12 at 22:38
add comment

Let $M$ be an abelian group. Every integer $n$ defines an endomorphism $f_n\colon M \rightarrow M$ such that $f_n(x) = nx$. $M$ is called divisible if $f_n$ is surjective for every nonzero integer $n$. Clearly $\mathbb{Q}/\mathbb{Z}$ is divisible.

Lemma 1 Let $M$ be a divisible abelian group. Let $I$ be an ideal of $\mathbb{Z}$. Then the canonical homomorphism $Hom(\mathbb{Z}, M) \rightarrow Hom(I, M)$ induced by the canonical injection $I \rightarrow \mathbb{Z}$ is surjective.

Proof: If $I = 0$, the assertion is clear. Hence we assume $I \neq 0$. There exists a nonzero integer $n$ such that $I = \mathbb{Z}n$. Let $f \in Hom(I, M)$. Since $M$ is divisible, there exists $a \in M$ such that $f(n) = na$. Let $x \in I$. There exists an integer $m$ such that $x = mn$. $f(x) = f(mn) = mf(n) = mna = xa$. Hence $f$ is in the image of the map$\colon Hom(\mathbb{Z}, M) \rightarrow Hom(I, M)$. QED

Lemma 2 Let $T$ be a divisible abelian group. Let $M$ be an abelian group. Let $N$ be a subgroup of $M$. Let $f\colon N \rightarrow T$ be a homomorphism. Let $x \in M - N$. Then there exists a homomorphim $g\colon N + \mathbb{Z}x \rightarrow T$ extending $f$.

Proof: Let $I = \{a \in \mathbb{Z}\colon ax \in N\}$. Let $h\colon I \rightarrow T$ be the map defined by $h(a) = f(ax)$. Since $h$ is a homomorphism, by Lemma 1, there exists $z \in T$ such that $h(a) = az$ for all $a \in I$.

Suppose $y + ax = y' + bx$, where $y, y' \in N, a, b \in \mathbb{Z}$. $y - y' = (b - a)x$ Hence $b - a \in I$. Hence $h(b - a) = f((b - a)x) = (b - a)z$. Hence $f(y - y') = (b - a)z$. Hence $f(y) + az = f(y') + bz$. Therefore we can define a map $g\colon N + \mathbb{Z}x \rightarrow T$ by $g(y + ax) = f(y) + az$. Clearly $g$ is a homomorophism extending $f$. QED

Theorem Let $T$ be a divisible abelian group. Then $T$ is injective.

Proof: This follows immediately from Lemma 2 and Zorn's lemma.

Corollary $\mathbb{Q}/\mathbb{Z}$ is injective.

share|improve this answer
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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