This is a question from a group theory exam. I have solved about 90% of the question. Actually, I have solved the whole question, but I want to solve it differently. There is no good reason to do it the the other way, except it feels like it should be easy and I can't stand being unable to solve it that way.

Let $G$ be a group and $H\preceq G$. Let $$f:G \rightarrow H$$ be a homomorphism such that the restriction $\left. f \right|_H$ is the identity $id_H$ on $H$.

For $a,\tilde{a} \in H$ and $b,\tilde{b} \in \text{ker}(f)$ show that:

a) If $H$ is normal then $ab=ba$.

Solution: If $H$ is normal then $$bab^{-1} \in H$$ for any $a \in H, b\in G$. It follows that $f(a) = a$ and $f(bab^{-1}) = bab^{-1}$. Because $f$ is a homomorphism and $f(b) = e_H$ we have that $$ bab^{-1} = f(bab^{-1}) = f(b) f(a) f(b^{-1}) = f(a) = a.$$ We conclude that $a = bab^{-1}$ as requested.

b) Explain why $$(a,b) \cdot (\tilde{a}, \tilde{b}) = (a\tilde{a}, b\tilde{b})$$ defines a group structure on $H \times \text{ker}(f)$.

Solution: This is just going trough the group axioms. I will spare you the details.

c) Let $\Psi: H \times \text{ker}(f) \rightarrow G$ be the map $$ \Psi(a,b) = ab.$$ Using part a), prove that $\Psi$ is an isomorphism.

Solution: It is not difficult to verify that $\Psi$ is a homomorphism. Furthermore, $\Psi$ is injective since if $\Psi(a,b) = \Psi(\tilde{a},\tilde{b})$ then $$ ab = \Psi(a,b) = \Psi(\tilde{a},\tilde{b}) = \tilde{a}\tilde{b}$$ so $$ ab = \tilde{a}\tilde{b}.$$ Applying $f$ we get $$ f(ab) = f(a) f(b) = a$$ and $$ f(\tilde{a}\tilde{b}) = \tilde{a}$$ So $a = \tilde{a}$. It follows automatically that $b = \tilde(b)$.

Problem: I can't find a proof that $\Psi$ is surjective. This is equivalent to proving that any $g\in G$ can be written $G = ab$ with $a \in H$ and $b \in \text{ker}(f)$. I can solve the question by constructing the inverse explicitly: $$ f^{-1}(g) = (f(g), f(g)^{-1}g).$$ But I don't want to. I want surjectivity :P.

edit: to be clear, I found the above inverse myself. I know it solves the question. But I want to show surjectivity without using the inverse. Actually, I'm just curious why every element $g \in G$ can be written $$ g = ab$$ with $a \in H$ and $b \in \text{ker}(f)$. This seems like a statement with content and the proof using the inverse does not help me understand it.

  • $\begingroup$ Given a function $f: X\to Y$, constructing a function $u:Y\to X$ such that $f\circ u=id_Y$ is one way to prove surjectivity: every element $y\in Y$ is $f(x)$, with $x=u(y)$. $\endgroup$ – user228113 Apr 9 '16 at 10:49
  • $\begingroup$ @G.Sassatelli I understand, but I want to prove it directly. $\endgroup$ – Timon van der Berg Apr 9 '16 at 10:51
  • $\begingroup$ Perhaps you want a way to guess why the inverse is that one? $\endgroup$ – user228113 Apr 9 '16 at 10:52
  • $\begingroup$ Given $g\in G$, define $a=f\left(g\right)\in H$ and $b=f\left(g\right)^{-1}g$. It is easy to check that $b\in\ker f$, and $\Psi\left(a,b\right)=g$. $\endgroup$ – Guy Apr 9 '16 at 11:26
  • $\begingroup$ @guy Thanks, but the point of my question was: how do I show surjectivity without explicitly constructing the inverse? As you you can see, I had found the inverse myself. $\endgroup$ – Timon van der Berg Apr 9 '16 at 12:36

For injectivity, since $\Psi$ is a homomorphism, you can just verify what it means that $\Psi(a,b)=1$. This means $ab=1$, so $a=b^{-1}\in\ker f$; therefore $a=f(a)=1$ and so also $b=a^{-1}=1$.

For surjectivity, consider $g\in G$. Then $a=f(g)\in H$, so $f(g)=a=f(a)$ and therefore $f(a^{-1}g)=1$. Hence, setting $b=a^{-1}g$, we have $$ g=aa^{-1}g=ab=\Psi(a,b) $$

  • $\begingroup$ Very neat answer, but this does apparently not answer his question. This is exactly his method. $\endgroup$ – Jori Apr 10 '16 at 15:32
  • $\begingroup$ @Jori Where am I using $f^{-1}$? I think I'm exactly showing how the inverse can be built, rather than guessed at. $\endgroup$ – egreg Apr 10 '16 at 15:40
  • $\begingroup$ Fair enough. The question is rather vague anyhow. In general I think it is hard to say whether two proofs are similar or not. $\endgroup$ – Jori Apr 10 '16 at 15:54
  • $\begingroup$ @Jori And there's not much more to do: the information is that $f(g)\in H$, so $f(g)=f(f(g))$ and therefore $f(g)^{-1}g\in\ker f$. After all, the inverse map is unique! ;-) $\endgroup$ – egreg Apr 10 '16 at 16:09
  • $\begingroup$ Sure, +1. My method is slightly different, but it boils down to the same things of course. For the folks (like me) that did not realize how easy it was to construct a specific (simple) inverse function, haha. $\endgroup$ – Jori Apr 10 '16 at 16:15

Maybe you are looking for something in which the properties of $f, G, H$ are more specifically used and the proof follows more naturally (but your approach is certainly valid, did you find it by trial and error?):

We must show that for any $g \in G$ we have for some $a \in H$ and $b \in \ker(f)$ such that $g = ab$. Because $\ker(f)$ is a subgroup of $G$ we know that we can write $g = \hat{g}b'$ for some $\hat{g} \in G$ and $b' \in \ker(f)$. Also since $H$ is normal in $G$ we can rewrite $a$ to $\hat{g}a'\hat{g}^{-1}$ for some $a' \in H$. Therefore we get that we must show that $\hat{g}a'\hat{g}^{-1}b = \hat{g}b' \Leftrightarrow a'\hat{g}^{-1}b = b'$. Remember that we can choose $a, b$ freely and therefore because of bijection also $a'$. If we could show now that $a'\hat{g}^{-1} \in \ker(f)$ for some $a'$ we would be done (again because of the bijection we create by multiplying freely with $b$ and remembering that $\ker(f)$ is a subgroup). This last statement is obvious because $f$ maps $G$ into $H$, and we are done

  • $\begingroup$ This is what I was looking for, but I accepted egrerg's answer since it is slightly more elegant (though he does not emphasize the part I am interested in, it is there). Now that I am typing this, I think my own proof contains the same information. I just did not realize that it did. To answer your question: I'm working on that problem as well. Let me know if you find anything. $\endgroup$ – Timon van der Berg Apr 10 '16 at 17:09

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.