2

I'm currently working on an exercise and the body of the text for the exercise is as follows. I have a first draft of the proof but am missing some things and am unsure about some things as well so any criticism and guidance is welcome.

Prove a nonempty set $G$ with an associative operation $\ast$ is a group iff $\forall g,h \in G$, $\exists x,y \in G$ such that $$g \ast x = h$$ $$y \ast g = h$$ Hint: Prove that if $e \ast g =g$ for some $g \in G$ then $e \ast h = h$ for all $h \in G$. Then appeal to the problem that a nonempty set with associative operation and a existence of left identity and inverse implies it is a group.

Being that this is an if and only if statement we have to prove the forward and reverse statements. I will start by proving the reverse.


Reverse Statement

Assume that $(G, \ast)$ is a group. Then for the equations $$y \ast g = h\ \text{and}\ g \ast x = h$$ we can take $y$ to be the left identity $e$ and $x$ to be the right identity $e$. So that we have $$g \ast e = g \in G\ \text{and}\ e \ast g = g \in G$$ Similarly we can take $x$ to be the right inverse of an arbitrary element $g$ and $y$ to be the left inverse of an arbitrary element $g$ - since groups are closed under inverses - to see that $$g \ast g^{-1} = e \in G\ \text{and}\ g^{-1} \ast g =e \in G$$ Finally. since groups are closed under the products determined by the operation $\ast$ if we take $x,y$ to be any other element in $G$ then we are guaranteed to have a solution in $G$ $$g \ast x = h \in G\ \text{and}\ y \ast g = h \in G$$ Thus $\forall g,h \in G$, $\exists x,y \in G\ \text{such that}\ g \ast x = h \in G\ \text{and}\ y \ast g = h \in G$ $\square$


Forward Statement

For this part I am having some trouble getting started. Can we just assume that since $G$ is a nonempty set and we have the equation $y \ast g =h$ then it must contain a left identity element? Otherwise how would we follow the hint? Furthermore, I'm not sure how to go from the statement "if $e \ast g =g$ for some $g \in G$" to the statement "then $e \ast h = h$ for all $h \in G$". Would we just argue that if $e$ fulfills the role of the identity for one element then it must for all elements by definition and uniqueness of the identity?

From there I think I would say that the existence of the left identity element implies the existence of the left inverse for an arbitrary element since $e \in G$ and so by the equation $y \ast g = h$ we have $$g^{-1} \ast g = e$$ Next assume that $g \ast g = g$ then we have by multiplication of left inverse: $$g^{-1} \ast g \ast g = g^{-1} \ast g$$ Now by associativity of $\ast$ $$(g^{-1} \ast g) \ast g = g^{-1} \ast g$$ By definition of left inverses $$e \ast g = e \rightarrow g = e$$ Now, using the equation $g \ast x = h$ and the fact that $\ast$ is associative we can say that $$(g \ast g^{-1})(g \ast g^{-1}) = g \ast (g^{-1} \ast g) \ast g^{-1} = g \ast e \ast g^{-1} = g \ast g^{-1}$$ Now if we define $g \ast g^{-1} \equiv j$. Then we have, $j \ast j = j$ which implies that $j = g \ast g^{-1} = e = g^{-1} \ast g$. Thus right inverses are equivalent to left inverses. Now, we take $$g \ast e = g \ast (g^{-1} \ast g) = (g \ast g^{-1}) \ast g = e \ast g$$ That is, that right identities equal left identities. Then all we have left to prove is closure under products but the given equations show just that. $\square$

  • 2
  • I don't understand the beginning of your proof of the "reverse statement": $x$ and $y$ should depend on $g$ and $h$, no? If $x=y=e$, then it would imply that $g=h$. Also, it is not necessary to speak about left and right identities, since the identity of a group is by definition left and right identity.
  • "if e fulfills the role of the identity for one element then it must for all elements by definition and uniqueness of the identity?" This is what you have to prove in this problem. You can't appeal to the definition of the identity since you don't know $e$ is the identity yet.
  • – Taladris May 23 '22 at 00:04
  • This is an oft-asked question. Because you are asking for feedback on your attempt I won't golden-hammer it as a duplicate... yet, but you can look around the site for solutions. – Arturo Magidin May 23 '22 at 00:12
  • @Taladris. I see, I was invoking the left and right identities mainly because the hint is suggesting us to appeal to the previous problem which was concerned with showing that a nonempty set with associative operation and left identity and inverse was a full fledged group. As well, the hint only made mention of the left identity when speaking about showing the identity for one element is the identity for all elements.

    So, how would I begin to follow the hint then? Can I just say that there exists an equation for $g$ such that $e \ast g = g$ and conclude that $e$ must be the identity then?

    – Numerical Disintegration May 23 '22 at 00:15
  • @ArturoMagidin. My apologies for the near duplicate question. I try and search the site beforehand, but I didn't see anything similar. Maybe I need to refine my search a bit. Thank you for the information though. – Numerical Disintegration May 23 '22 at 00:16
  • 1
    The hint is for the forward direction. Indeed, applying the given property to some $g=h$ in $G$ implies that $g$ a right-identity $x=e'_g$ and a left-identity $y=e_g$. The hint says that you may prove that $e_g=e_h$ for every $g$ and $h$ in $G$ (so $G$ has a left-identity), then appeal to some other problem (you probably already solved it) to finish the proof. The Reverse Statement is elementary: every element of a group has an inverse, so given $g$ and $h$, we can define $x=g^{-1}h$ and $y=hg^{-1}$ and check they satisfy the given property. – Taladris May 23 '22 at 00:35
  • 2
    One, two (the latter has a uniqueness clause, but the top voted answer does it without assuming uniqueness); this answer here is three, four, five... I searched first for "semigroup equations solutions" within the tag [group-theory], then without the tag. – Arturo Magidin May 23 '22 at 00:36
  • @Taladris Thank you for the clarification! – Numerical Disintegration May 23 '22 at 00:49
  • @ArturoMagidin I see. I hadn't heard the term "semigroup" before, even though that is apparently what we are dealing with here it was never given a name, so I was trying to search too specifically I think for the name of the problem. Things like "A nonempty subset with associative operation is a group if the equations..." and so forth. Thanks, I'll try to broaden my search and include any proper terms when I can. – Numerical Disintegration May 23 '22 at 00:52