Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I'm trying to understand normal subgroups and kernels of homomorphisms. Normal subgroups are defined as such:

$gHg^{-1}=H~~~\forall g \in G$

While i was trying to see which subgroups are normal, to verify the above statment, i should run through all elements of $G$.

To find a shorter way, I come up with the below:

$hGh^{-1}\cap H=\emptyset~~~\forall h \in H $ (wrong)

$h(G \setminus H)h^{-1}\cap H=\emptyset~~~\forall h \in H $ (correct)

This is basically telling that, if H is normal, then $hGh^{-1}$ is in $G \setminus H$.

Assume the contrary,

$h_{0}Gh_{0}^{-1}=h_{1}$ implies $h_{0}G=h_{3}$ implies $G=h_{4}$ or $G=H$ which is a contradiction.

Could anybody prove the above statement from the first, or possible vice versa?

Second one cheaper since H has fewer elements to run through, so i would prefer to write tests over the second one.


share|cite|improve this question
There are other ways to check whether a given subgroup $H$ is normal in a given group $G$. You can check whether the left cosets of $H$ in $G$ are also right cosets. You can try to find a homomorphism from $G$ to some other group $K$ with kernel $H$. If you already know the conjugacy classes of $G$ you can check whether $H$ is a union of conjugacy classes. – Gerry Myerson Mar 5 '12 at 22:39
The statement $h(G \setminus H)h^{-1} \cap H = \phi$ for all $h \in H$ is true for all subgroups $H$ of $G$, not just normal subgroups. – Derek Holt Mar 5 '12 at 23:44
... and you can just check $gHg^{-1}=H$ where $g$ runs over a set of generators of $G$ (you don't need to check every element of $G$). – Arturo Magidin Mar 6 '12 at 4:08

You want to show that $H$ is normal if and only if $h(G\setminus H)h^{-1} = G\setminus H$ for every $h\in H$.

This is incorrect. The condition $h(G\setminus H)h^{-1}=G\setminus H$ for all $h\in H$ holds whenever $H$ is a subgroup, whether or not it is normal: because the map $x\longmapsto hxh^{-1}$ is a bijection from $G$ to itself for every elements of $G$ (if $hxh^{-1}=hyh^{-1}$, then cancelling $h$ on the left and $h^{-1}$ on the right we get $x=y$; and for every $z\in G$, $z$ is the image of $h^{-1}zh$ under this map). If $H$ is a subgroup, and $h\in H$, then $hyh^{-1}\in H$ for every $y\in H$; that is, $hHh^{-1}=H$; and since we have a bijection and it maps $H$ to itself, it must map the complement of $H$ to itself, so $h(G\setminus H)h^{-1} = G\setminus H$. This does not require $H$ to be normal.

In fact, the condition may hold for sets that are not even subgroups. For instance, if $h\neq\{e\}$ then $X=\{h,h^{-1}\}$ is not a subgroup, but $hXh^{-1} = h^{-1}Xh = X$, so $X$ satisfies your condition.

So the condition does not characterize normality; it does not even characterize subgroup.

If you want a "cheaper" way to test normality, there are plenty: if you can find a homomorphism that has $H$ as a kernel, then $H$ is necessarily normal. In fact,

Let $G$ be a group. A subgroup $H$ of $G$ is normal in $G$ if and only if there exists a homomorphism $f$ with domain $G$ such that $\mathrm{ker}(f) = H$.

Proof. Kernels are normal; if $H$ is normal, then the canonical map $\pi\colon G\to G/H$ given by $\pi(g) = gH$ is a homomorphism, and $\mathrm{ker}(\pi)=H$. $\Box$

Another useful set of facts:

Let $G$ be a group, $H$ a subgroup, and $S$ a subset of $G$ such that $G=\langle S\rangle$. The following are equivalent:

  1. $H$ is normal in $G$.
  2. $gHg^{-1} = H$ for every $g\in G$.
  3. $gHg^{-1}\subseteq H$ for every $g\in G$.
  4. $sHs^{-1} = H$ for every $s\in S$.
  5. $sHs^{-1}\subseteq H$ and $s^{-1}Hs\subseteq H$ for every $s\in S$.

If $H$ is finite, then in 5 it suffices to check $sHs^{-1}$ for every $s\in S$.

Proof. 1 and 2 are equivalent by definition. 2 clearly implies 3; if 3 holds and $x\in G$, then $xHx^{-1}\subseteq H$; we only need to show that $H\subseteq xHx^{-1}$ also holds; by 2, $(x^{-1})H(x^{-1})^{-1} = x^{-1}Hx\subseteq H$; multiplying on the left by $x$ and on the right by $x^{-1}$ we obtain $H = xx^{-1}Hxx^{-1}\subseteq xHx^{-1}$. This proves $xHx^{-1}=H$ for every $x\in G$, which means 2 holds.

2 implies 4; to see that 4 implies 3, first note that since $sHs^{-1}=H$ for every $s\in S$, then multiplying on the left by $s^{-1}$ and on the right by $s$ we get $H=s^{-1}Hs$ for every $s\in S$ as well. Now let $g\in G$ be an arbitrary element. Then $g$ can be written as a product of elements of $S$ and its inverses, $g = s_1^{\epsilon_1}\cdots s_n^{\epsilon_n}$, where $\epsilon_i\in\{1,-1\}$ for every $i$; we prove that $gHg^{-1}=H$ by induction on $n$. If $n=1$, then the result holds by assumption or by our observation. If the result holds for elements that can be expressed as the product of $k$ elements of $S$ and their inverses, and $g=s_1^{\epsilon_1}\cdots s_{k+1}^{\epsilon_{k+1}}$; then by the induction hypothesis we have $$s_2^{\epsilon_2}\cdots s_{k+1}^{\epsilon_{k+1}}H(s_2^{\epsilon_2}\cdots s_{k+1}^{\epsilon_{k+1}})^{-1} = H.$$ Multiplying on the left by $s_1^{\epsilon_1}$ and on the right by $(s_1^{\epsilon_1}){-1}$, we get $$gHg^{-1} = s_1^{\epsilon_1}s_2^{\epsilon_2}\cdots s_{k+1}^{\epsilon_{k+1}}H(s_2^{\epsilon_2}\cdots s_{k+1}^{\epsilon_{k+1}})^{-1}(s_1^{\epsilon_1})^{-1} = s_1^{\epsilon_1}H(s_1^{\epsilon_1})^{-1}.$$ But by 4, $s_1^{\epsilon_1}H(s_1^{\epsilon_1})^{-1}=H$, so $gHg^{-1}=H$, as desired.

Finally, 4 implies 5 via our observation above about inverses of elements of $S$; and if 5 holds and $s\in S$, then $sHs^{-1}\subseteq H$; to show $H\subseteq sHs^{-1}$, we note that $s^{-1}Hs\subseteq H$, and multiplying on the left by $s$ and on the right by $s^{-1}$ we obtain the desired inclusion.

If $H$ is finite and 5 holds for elements $s\in S$, then since $sHs^{-1}\subseteq H$ but both are of the same size (since $x\mapsto sxs^{-1}$ is one-to-one), then they are equal, so we obtain 4 directly. $\Box$

So it suffices to check the condition for a generating set of $G$, instead of for every element of $G$. Moreover, since $hHh^{-1}=H$ for every $h\in H$, it sufffices to check the condition for every element of a generating set that is not in $H$.

share|cite|improve this answer

$hGh^{-1}$ contains $H$ if $H$ is a subgroup of $G$.

share|cite|improve this answer
Oh yes, i missed that point. in $hGh^{-1}$, $G$ is actually $G \setminus H$. Otherwise, it would not make sense as you have pointed. – foobar Mar 5 '12 at 22:40
@foobar: It's still wrong in that case; you don't get normality out of this condition. – Arturo Magidin Mar 6 '12 at 4:08

You let $g \in G$. Then show that $gHg^{-1} = H$. Since $g$ was arbitrarily chosen, $H$ is normal in $G$.

You do not need to run through all elements in $G$.

share|cite|improve this answer
This reminds me of the old joke: to prove that every positive integer is prime, let $n$ be an arbitrary positive integer. 17 is pretty arbitrary; and it is prime. Therefore, every positive integer is prime. – Arturo Magidin Mar 6 '12 at 4:25
When I was tutoring an Abstract Algebra class for CS department many years ago, I ran into quite a few students who believed that they need to check every element in a set if there is a $\forall$ in a problem. – scaaahu Mar 6 '12 at 4:35
My point, I guess, is that the phrasing "arbitrarily chosen" is somewhat ambiguous. What you really mean, of course, is that if one can prove that $gHg^{-1}=H$ using only the fact that $g$ is an element of $G$ and not using any other property that some elements of $G$ may have but others lack, then the argument suffices to establish the result for all elements of $G$. Much like proving $A\subseteq B$ by showing that if $a\in A$, then $a\in B$: if all you use about $a$ is that it is in $A$, you're fine. – Arturo Magidin Mar 6 '12 at 4:38
@Arturo, your explanation is much more clearer than mine. I put my answer there because I was afraid OP might just have a simple misunderstanding. He was asking a question looking like from the first chapter of an Abstract Algebra textbook. – scaaahu Mar 6 '12 at 4:48

Your Answer


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.