The Lemma asserted in Herstein as given by $[a] = Ha$ seems very non intuitive to me. How do I think in order that this thing makes sense to me?

LEMMA 2.4.4 For all $a$ in $G$ , $$Ha = \{ x \in G : a \equiv x \mod H\}$$


  • $\begingroup$ The way I do it is purely algebraically. Just try it as the set with the properties provided and try to manipulate them. Work out some examples for yourself (like the Klein 4 group, or $\mathbb{Z}$ and $n\mathbb{Z}$ for some $n$'s, or $S_3$) and just write out the sets themselves (and the set of sets). $\endgroup$ – Eoin Mar 23 '15 at 7:13
  • $\begingroup$ Imagine lining up the elements of $H$ in a vertical column. At the top is the identity $e$. Multiplying by $a$ shifts everything to the right, so $Ha$ is a column of elements with the topmost element $a$. This illustrates that all of the cosets have the same cardinality and partition the group. $\endgroup$ – William Stagner Mar 23 '15 at 7:14
  • $\begingroup$ @K.Dutta Please give the statement of the Lemma in the question. Otherwise I guarantee you this question is on a short trip to "on hold: missing context". $\endgroup$ – Mario Carneiro Mar 23 '15 at 7:14
  • 1
    $\begingroup$ For any equivalence relation, represent elements of your original set as mushrooms (or favorite topping). Place them on pizza and slice it up. All mushrooms on a slice are in one partition/coset. $\endgroup$ – dls Mar 23 '15 at 15:13

It's easiest to see this if we start with the group of integers under addition (in this operation $Ha$ is wriiten $H + a$). Imagine the subgroup of multiples of $3$:

$3\Bbb Z = \{\dots,-9,-6,-3,0,3,6,9,12,\dots\}$

Then $3\Bbb Z + 1$ is $3\Bbb Z$ "shifted by $1$" (just add one to every element of $3\Bbb Z$):

$3\Bbb Z + 1 = \{\dots,-8,-5,-2,1,4,7,10,13,\dots\}$, and $3\Bbb Z + 2$ works the same way:

$3\Bbb Z + 2 = \{\dots,-7,-4,-1,2,5,8,11,14,\dots\}$.

Note that $3\Bbb Z + 3$ is just $3\Bbb Z$ again, since $3\Bbb Z$ has "gaps" of $3$, and extends infinitely negatively and positively. So we only get those $3$ cosets, which chops the integers into $3$ pieces, integers of the form $3k,3k+1$, or $3k+2$, and each "piece" of the integers is exactly the same "size".

So, now let's look at some finite group. Suppose we have a subgroup $H = \{e,h_1,\dots,h_n\}$. To get $Ha$ we "multiply everything by $a$ (on the right)":

$Ha = \{a,h_1a,\dots,h_na\}$.

We can do this for ANY $a \in G$, so it is natural to ask: can we have $Ha = Hb$?

If so, then for any $h_ka \in Ha$, it must be equal to some $h_mb \in Hb$. This means:

$h_ka = h_mb \implies h_k = h_mba^{-1} \implies h_m^{-1}h_k = ba^{-1}$.

So if $Ha = Hb$, then $ba^{-1} \in H$. So that is what we call a "necessary" condition for $Ha$ to equal $Hb$. When encountering a necessary condition, mathematicians like to inquire if it is "sufficient" as well. So we start with the OTHER assumption:

$ba^{-1} \in H$. What will this tell us about $Ha$ and $Hb$? Well, if $ba^{-1} = h \in H$, then $ha = b$.

Note we can re-write $h_k \in H$ as: $h_kh^{-1}h$. So:

$Ha = \{a,h_1a,\dots,h_na\} = \{h^{-1}ha, h_1h^{-1}ha,\dots,h_nh^{-1}ha\}$

$= \{h^{-1}b, h_1h^{-1}b,\dots,h_nh^{-1}b\}$

What we'd like to show now, is that the mapping $h_k \mapsto h_kh^{-1}$ ($H \to H$) is bijective.

So suppose $h_kh^{-1} = h_mh^{-1}$. Multiplying both sides on the right by $h$, gives $h_k = h_m$, so our mapping is injective. On the othr hand, given any $h_j \in H$, we see that: $h_jh \in H$, and our mapping takes:

$h_jh \mapsto (h_jh)h^{-1} = h_j(hh^{-1}) = h_j$, so the map $h_k \mapsto h_kh^{-1}$ is bijective.

THIS tells us the set $\{h^{-1}b, h_1h^{-1}b,\dots,h_nh^{-1}b\} = Hb$ since $h_kh^{-1}$ runs through ALL elements of $H$, just like $h_k$ does.

In other words $Ha = Hb$ if, and only if, $ba^{-1} \in H$ (since $H$ is a subgroup, this also means $(ba^{-1})^{-1} = ab^{-1} \in H$, as well).

So we now define a relation: $a \sim_H b$ means: $Ha = Hb$. As we saw above this means that $ba^{-1}$ (and $ab^{-1}$) are in $H$. It is not immediately apparent this is an equivalence relation. We have to PROVE it.

$\sim_H$ is reflexive: this states that $Ha = Ha$, or equivalently, that $aa^{-1} = e \in H$. This is obvious, since $H$ is a subgroup, and subgroups contain the identity.

$\sim_H$ is symmetric: this says if $Ha = Hb$, then $Hb = Ha$. Well that seems clear. Alternatively, if $ba^{-1} \in H$, then since $H$ is a subgroup, $(ba^{-1})^{-1} = ab^{-1} \in H$, that is: $a\sim_H b \iff b \sim_H a$.

$\sim_H$ is transitive: this says if $Ha = Hb$, and $hb = Hc$, then $Ha = Hc$. Again, this seems straight-forward. Using our "other definition" we need to show that if:

$ba^{-1}$ and $cb^{-1}$ are in $H$, so is $ca^{-1}$, But $H$ is a subgroup, and so is closed under multiplication, so:

$(cb^{-1})(ba^{-1}) = c(b^{-1}b)a^{-1} = ca^{-1} \in H$.

This tells us $\sim_H$ is a bona-fide equivalence relation. It should be clear now that if $b \in [a]_{\sim_H}$, that $Hb = Ha$, which means in particular, $b = eb \in Ha$. On the other hand, if $b \in Ha$, so that $b = ha$ for some $h \in H$, then $ba^{-1} = h \in H$, so that $b \in [a]_{\sim_H}$.

In summary, the equivalence class of $a$ (under the equivalence relation $\sim_H$), $[a]$ is PRECISELY the (right) coset, $Ha$.

  • $\begingroup$ Can you explain why if Ha=Hb implies $ba^{-1}$ in H . How does $(h_{m})^{-1}h_{k} = ba^{-1}$ proves it ? $\endgroup$ – Taylor Ted Mar 23 '15 at 9:02
  • $\begingroup$ @K.Dutta Because $H$ is a subgroup, it is closed under inverse and group operation, so $h_m\in H\implies (h_m)^{-1}\in H\implies (h_m)^{-1}h_k\in H$. $\endgroup$ – Mario Carneiro Mar 23 '15 at 10:03

The coset of a group I like to think of it as the set H multiplied by element a if your group is normal that is Ha = aH then multiplication by a will make both sets equal.

I will try to explain to you the whole story we have in normal $Z_n$ we have the following a $\equiv$ b (mod n) iff a - b = n q1*

we could look at this from another perspective that is we could instead of looking as numbers we could look at this in terms of sets that is we would have

a $\equiv$ b (mod Z_n) iff $a*b^{-1} \in Z_n$

we can show that this relation gives an equivalence relation and hence congruence class partitions your group G and the classes are given as follows

{b $\in$ G | b $\equiv$ a (mod K)} = {b $\in$ G | $b*a^-1$ = k, such that k $\in K$} = {ka | k $\in K$}. So the congruence class of a modulo K is our right coset.


It seems that the first $[a]$ is the equivalence class formed by all elements equivalent to $a$ under the relation $a\sim b$ when there is an $h \in H$ so that $a=hb$. While $Ha$ is the set formed by all products of $h \in H$ with a. These are the same because if $a = hb$ then $b = h^{-1}a$ and $h^{-1}\in H$ iff $h \in H$, so that the collection of all such elements $b$ is the same as that of all products in $Ha$. Intuitively one can think of a coset as some kind of orbit formed by the action of $H$ on an element $a \in G$, for example if $a$ is an element of $H$ then the orbit is $H$ itself, for $H$ is closed under multiplication (being a subgroup).

  • $\begingroup$ What do you mean by orbit ? $\endgroup$ – Taylor Ted Mar 23 '15 at 7:19

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.