4
$\begingroup$

Given a coalgebra $(C,\Delta,\varepsilon)$, over a field, the following is a well-known property:

the comultiplication $\Delta:C\to C\otimes C$ is a coalgebra map if and only if $C$ is cocommutative

My question is how we can prove the above with an explicit computation? (The main problem to be handled here seems to be the treatment of the emerging elements $c_{(1)(2)}$ and $c_{(2)(1)}$).

To the best of my knowledge, I have not found the details of the computation neither in some text nor in some article or some other question at this site. So, I have tried to devise the computation myself and I am posting the answer below. Is there some other approach ?

P.S. This is exercise 3, p.66, Ch. III, from Moss E.Sweedler's book on Hopf algebras.

$\endgroup$
7
  • $\begingroup$ IIRC, this is an exercise in chapter 1 of Vic Reiner's and my notes on hopf algebras in Combinatorics. The dual statement is explicitly proven element-free in the solutions section, so you should be able to just blindly revert all arrows. Sorry for lack of detais, I am on my phone standing in a line. $\endgroup$ Commented Sep 6, 2016 at 18:23
  • 2
    $\begingroup$ Ah, yes. Exercise 1.59 in arxiv.org/abs/1409.8356v4 . The solutions are in an ancillary file. $\endgroup$ Commented Sep 6, 2016 at 18:27
  • $\begingroup$ @ darij grinberg: thank you for mentioning. I was already aware of your (very interesting) notes. As you have already said, the dual statement is proven (with a very detailed and long proof). It would be nice if you find some time later to post a sketch of that proof, adapted (with all arrows reversed as you said) to the question as stated here. thanks again! $\endgroup$ Commented Sep 6, 2016 at 18:36
  • $\begingroup$ @darijgrinberg Hi Darij ! It seems that one must suppose that the coalgebra is coassociative with counit and then, no field is needed, am I right ? (see my argument with identities in the answer below) $\endgroup$ Commented Sep 7, 2016 at 2:44
  • $\begingroup$ @DuchampGérardH.E.: I'm working under the conventions that coalgebras are coassociative and counital by definition! Yes, it's probably false otherwise, and yes, the fieldness is not needed. $\endgroup$ Commented Sep 7, 2016 at 3:06

2 Answers 2

6
$\begingroup$

Given a (coassociative and counital) coalgebra $(C,\Delta,\varepsilon)$, over a field $k$, we can form the tensor product coalgebra $(C\otimes C,\Delta_{C\otimes C},\varepsilon_{C\otimes C})$ through: $$ \Delta_{C\otimes C}=(Id\otimes\tau\otimes Id)\circ(\Delta\otimes\Delta):C\otimes C\rightarrow C\otimes C \otimes C \otimes C , \\ \\ \\ \varepsilon_{C\otimes C}=\phi \circ (\varepsilon\otimes\varepsilon):C\otimes C\rightarrow k $$ where $Id$ is the identity map and $\phi:k\otimes k\stackrel{\cong}{\rightarrow} k$ the natural isomorphism.

The comultiplication $\Delta:C\rightarrow C\otimes C$ being a morphism of coalgebras, or a coalgebra map, by definition means that for an arbitrary $c\in C$ we have: $\varepsilon_C(c)=\varepsilon_{C\otimes C}\circ\Delta(c)=\varepsilon(c_1)\varepsilon(c_2)$ and $$ \Delta_{C\otimes C}\circ\Delta(c)=(\Delta\otimes\Delta)\circ\Delta(c) \Leftrightarrow \\ \\ \\ \Leftrightarrow \Delta_{C\otimes C}\big(\sum c_1\otimes c_2\big)=(\Delta\otimes\Delta)\big(\sum c_1\otimes c_2\big)\Leftrightarrow \\ \\ \\ \Leftrightarrow (Id\otimes\tau\otimes Id)\circ(\Delta\otimes\Delta)\big(\sum c_1\otimes c_2\big)=(\Delta\otimes\Delta)\big(\sum c_1\otimes c_2\big)\Leftrightarrow \\ \\ \\ \Leftrightarrow \sum c_1\otimes c_3\otimes c_2\otimes c_4=\sum c_1\otimes c_2\otimes c_3\otimes c_4 $$ In the last line of the above, we have made use of generalized coassociativity, expressing both sides in Sweedler's notation. Given that $\Delta(c)=\sum c_1\otimes c_2$, the last line of the above could have been written (without using generalized coassociativity) alternatively as: $$ \sum c_{1_1}\otimes c_{2_1}\otimes c_{1_2}\otimes c_{2_2}=\sum c_{1_1}\otimes c_{1_2}\otimes c_{2_1}\otimes c_{2_2} $$ Now, applying to both sides of the last line of the above, the map $(\varepsilon\otimes Id\otimes Id\otimes\varepsilon)$, we get $$ \sum \varepsilon(c_{1_1})\otimes c_{2_1}\otimes c_{1_2}\otimes \varepsilon(c_{2_2})=\sum \varepsilon(c_{1_1})\otimes c_{1_2}\otimes c_{2_1}\otimes \varepsilon(c_{2_2})\Leftrightarrow \\ \\ \\ \Leftrightarrow \sum\varepsilon(c_{2_2}) c_{2_1}\otimes \varepsilon(c_{1_1})c_{1_2}=\sum \varepsilon(c_{1_1})c_{1_2}\otimes \varepsilon(c_{2_2})c_{2_1} \Leftrightarrow \\ \\ \\ \Leftrightarrow \sum c_2\otimes c_1=\sum c_1\otimes c_2 $$ for any $c\in C$. In the last line, use has been made of the defining property of the counity map: $\sum\varepsilon(c_1)c_2=\sum c_1\varepsilon(c_2)=c$, for any $c\in C$.

Thus, we have shown that: If the comultiplication $\Delta:C\rightarrow C\otimes C$ is a morphism of coalgebras, or a coalgebra map then this implies the cocommutativity of $C$.

The converse implication, i.e. cocommutativity of $C$ implies that the comultiplication is a coalgebra map, comes from the fact that generalized coassociativity permits us to write
$\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ (\Delta\otimes\Delta)=(Id\otimes\Delta\otimes Id)\circ(Id\otimes\Delta)$
and thus, together with cocommutativity of $C$, they imply $$ (\Delta\otimes\Delta)=(Id\otimes\Delta\otimes Id)\circ(Id\otimes\Delta)= \\ =(Id\otimes\tau\circ\Delta\otimes Id)\circ(Id\otimes\Delta) = \\ =(Id\otimes\tau\otimes Id)\circ(Id\otimes\Delta\otimes Id)\circ(Id\otimes\Delta)= (Id\otimes\tau\otimes Id)\circ(\Delta\otimes\Delta)=\Delta_{C\otimes C}\Rightarrow \\ \Rightarrow (\Delta\otimes\Delta)\circ\Delta=\Delta_{C\otimes C}\circ\Delta $$

$\endgroup$
4
  • $\begingroup$ You need that your coalgebra (even unital) be coassociative, don't you ? $\endgroup$ Commented Sep 7, 2016 at 2:35
  • 1
    $\begingroup$ O.K. after reading your post I see that coassociativity is contextual. $\endgroup$ Commented Sep 7, 2016 at 3:21
  • $\begingroup$ @Duchamp Gérard H. E., you are right of course. The coalgebra is considered to be coassociative and counital. This is why I denote it $(C,\Delta,\varepsilon)$. And yes, the field does not seem to play some particular role here. I will edit the notation in the original post as well. $\endgroup$ Commented Sep 7, 2016 at 4:14
  • $\begingroup$ Thank you for your efforts. I think we'll have this completely clarified now. $\endgroup$ Commented Sep 7, 2016 at 6:15
3
$\begingroup$

Not an answer, but a clarification of the context and ``bording counterexamples''. In short, this property is true only in the context of counital coalgebras and NOT otherwise.

If it were true in general, it would be true in the subcategory $k-coalg^{fin}$ of finite dimensional coalgebras (we are over a field $k$) and the dual property (i.e. the multiplication is an algebra map iff the algebra is commutative) true in the dual category towards which we have an isofunctor (the vector dual). To see that this is not the case, take any finite dimensional non-commutative associative algebra $\mathcal{A}$ such that $\mathcal{A}^{(4)}$ (the space generated by the products of 4 elements) is zero.

Take, for example, a two-letter alphabet (i.e. a set $A=\{x,y\}$ of two variables). And consider the algebra $k[A^+]$ of the free semigroup $A^+$ (it is the algebra of non-commutative polynomials without constant term). Now, let $(A^+)_{\geq 4}$ be the set of words of length $\geq 4$. It generates the two-sided ideal $\mathcal{I}_{\geq 4}$ and the non-commutative algebra $\mathcal{A}=k[A^+]/\mathcal{I}_{\geq 4}$ is a counterexample. If one likes to stick to coalgebra, on can consider the dual coalgebra ($\mathcal{A}$ is finite dimensional).

Remark : It is transparent in terms of identities that the property is true for coassociative coalgebras with counits. The dual property is that the identity $$ xy=yx\qquad (1) $$
(commutativity) is true iff the identity $$ (x_1y_1)(x_2y_2)=(x_1x_2)(y_1y_2)\qquad (2) $$ is true, which is false in general and straightforward if you suppose associativity and unit.

Statement (after dualization) It seems that the full statement is:

Let $\mathcal{C}=(C,\Delta)$ be a coassociative coalgebra, then

  1. if $\mathcal{C}$ is cocommutative, $\Delta$ is a coalgebra map
  2. if $\Delta$ is a coalgebra map and if $\mathcal{C}$ admits a counit then $\mathcal{C}$ is cocommutative.
$\endgroup$
7
  • 1
    $\begingroup$ Always good to be reminded of, though in the context of Sweedler's book (and most any other context directly attached to Hopf algebras), all (co)algebras are assumed to have (co)units. $\endgroup$ Commented Sep 7, 2016 at 2:24
  • 1
    $\begingroup$ @zibadawatimmy You are right, that's precisely why I wrote "as stated" and elaborated my counterexamples in order to avoid false beliefs (the devil is in non-conscious contexts :) $\endgroup$ Commented Sep 7, 2016 at 2:38
  • 3
    $\begingroup$ Technically, "as stated" by the time you wrote this answer, the OP had already specified this was an exercise in Sweedler's book he was working on. So the existence of a counit (and coassociativity) was implicit from that context. Your parenthetical in your comment is definitely something I can agree with, though. Every so often these standard assumptions get violated, even in the "nicer" settings where you aren't specifically working on such topics, and lots can go wrong in your own thought process if you aren't aware of where those assumptions are required. $\endgroup$ Commented Sep 7, 2016 at 4:52
  • $\begingroup$ @zibadawatimmy Yes, again you're right. That's why I attacked by the necessity of a unity (hence the counterexample of finite dimensional algebra where the dual property is violated). Time passing and through interactions, I see my contribution more as clarifying the borders of the statement (or context). Too much (non-conscious) "contextual mathematics" is a trend which boosts the development of non cross understanding communities and I, myself, see no harm to clarify. $\endgroup$ Commented Sep 7, 2016 at 5:54
  • 1
    $\begingroup$ @KonKan In my understanding, it is always a value to clarify context and hunt hidden assumptions. $\endgroup$ Commented Sep 8, 2016 at 0:39

You must log in to answer this question.

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