2
$\begingroup$

I'm currently studying abelian groups in Kurosh's The Theory of Groups. I'm trying to understand the proof of the theorem:

Let $B \leqq A$ be abelian groups. If $A/B \cong C$ and $C$ is a free group, then $A\cong B \oplus C$

Kurosh gives a proof on page 144, though I don't quite follow these steps, and am looking for some more detail. I have looked at Manipulating quotients and direct sums of abelian groups, and the answer is certainly helpful, but still does not include a full proof. Both seem to suggest constructing a subgroup of generating elements, but I can't seem to figure out what the isomorphism would look like. Any help would be greatly appreciated!

From Kurosh: enter image description here

$\endgroup$
3
  • $\begingroup$ You mean $C$ is a free abelian group, not $C$ is a free group. $\endgroup$
    – Derek Holt
    Commented Jun 22 at 10:25
  • $\begingroup$ Please do not rely on pictures of text. $\endgroup$
    – Shaun
    Commented Jun 22 at 11:51
  • $\begingroup$ @DerekHolt yes, thank you! $\endgroup$ Commented Jun 26 at 12:41

1 Answer 1

2
$\begingroup$

Since $C$ is free abelian, $C = \mathbb{Z}^{(I)}$ for some index set $I$. For each $i \in I$, let $e_i \in \mathbb{Z}^{(I)} = C$ be the element that’s $1$ at $i \in I$ and $0$ everywhere else. Since $A/B = C$, we may pick $a_i \in A$ s.t. $a_i + B = e_i$. Then the desired isomorphism $\pi: B \oplus C \to A$ is given by,

$$\pi(b \oplus (z_i)_{i \in I}) = b + \sum_{i \in I} z_ia_i$$

It is clearly a group homomorphism. To see it is injective, note that $\pi(b \oplus (z_i)_{i \in I}) = 0$ implies $\varphi(\pi(b \oplus (z_i)_{i \in I})) = 0$ where $\varphi: A \to A/B = C$ is the natural quotient map. But $B = \ker(\varphi)$ and $\varphi(a_i) = e_i$, so $0 = \varphi(\pi(b \oplus (z_i)_{i \in I})) = \sum_{i \in I} z_ie_i$. Thus, $z_i = 0$ for all $i$. But then $0 = \pi(b \oplus (z_i)_{i \in I}) = b$. Therefore, $\ker(\pi) = \{0\}$.

For surjectivity, let $a \in A$ be arbitrary. Then $\varphi(a) = (z_i)_{i \in I}$ for some $(z_i)_{i \in I} \in \mathbb{Z}^{(I)} = C$. Thus, $b = a - \sum_{i \in I} z_ia_i$ satisfies that $\varphi(b) = \varphi(a) - \sum_{i \in I} z_ie_i = 0$, i.e., $b \in B$. Now by definition of $\pi$, we see that $a = \pi(b \oplus (z_i)_{i \in I})$, so $a \in \text{ran}(\pi)$, whence $\pi$ is surjective.

$\endgroup$
7
  • $\begingroup$ Thanks for this! Some of this is going a little over my head. Why can we pick $a_i \in I$ s.t. $a_i+B=e_i$? Are we considering $a$ to be elements of $A$ which are indexed by $I$? And what happens when we choose $\mathbb{N}$ as our index set? I've also edited my post to include Kurosh's explanation which I would really appreciate some pointers on! $\endgroup$ Commented Jun 22 at 10:07
  • $\begingroup$ @MathematicallyUnsound The first one is a typo. It should have been $a_i \in A$. It doesn’t matter what set is the index set - that is determined by what $C$ is, but the proof goes through regardless. $\endgroup$
    – David Gao
    Commented Jun 22 at 17:29
  • $\begingroup$ @MathematicallyUnsound Kurosh’s explanation that you added to your post is the same proof - my proof is just a more detailed explanation, is all. $\endgroup$
    – David Gao
    Commented Jun 22 at 17:31
  • $\begingroup$ Thanks, this makes more sense! Though I'm still a bit confused by surjectivity. $0=\varphi(b)=\varphi(a)-\sum_{i \in I} z_ie_i$ makes sense, but $\varphi$ is not necessarily injective, so how does this imply $b = a- \sum_{i \in I}z_i a_i$? I don't see how $\pi(b\oplus(z_i)_{i \in I})$ can be equal to the arbitrary $a \in A$. $\endgroup$ Commented Jun 26 at 12:40
  • $\begingroup$ @MathematicallyUnsound There is no implication, $b$ is defined to be $a - \sum_{i \in I} z_ia_i$, and $\varphi(b) = \varphi(a) - \sum_{i \in I} z_ie_i = 0$ is the consequence of this definition. $\endgroup$
    – David Gao
    Commented Jun 26 at 22:13

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .