3

I'm struggling a bit with Sweedler notation. Let $(H,∆,ε,S,m,u)$ be a Hopf algebra over a commutative ring $k$ and let $X,Y$ be right $H$-comodules which are finitely generated projective as $k$-modules (if you like, let $k$ be a field and let $X,Y$ be finite-dimensional vector spaces).

Then $X^*$ is also an $H$-comodule via

$$π_{X^*}(f)(x) = (f_{(1)}⊗f_{(2)})(x) = (f⊗S)∘π_X(x) = f(x_{(1)})⊗S(x_{(2)})$$

and $X⊗Y$ is an $H$-comodule via

$$π_{X⊗Y}(x⊗y) = x_{(1)}⊗y_{(1)}⊗m(x_{(2)}⊗y_{(2)}).$$

The source I am using says that the coaction on $X^*⊗X$ is given by

$$π_{X^*⊗X}(f⊗x) = f⊗x_{(1)}⊗m(S(x_{(2)})⊗x_{(3)}) $$

which I think means that symbols $x_{(2)},x_{(3)}$ refer to (sets of) elements of $H$ that "split off" from $f$ and $x$ respectively under the coactions, but this is confusing to me because there also seems to be an implicit identification $X^*⊗H = \mathrm{Hom}_k(X,H)$ floating around. And then to show that the evaluation map is an $H$-comodule homomorphism we have

$$ (\mathrm{ev}_X⊗\mathrm{id}_H)∘π_{X^*⊗X}(f⊗x) = f(x_{(1)})⊗S(x_{(2)})x_{(3)} \\ = f(x_{(1)})⊗ε(x_{(2)}) = f\big(x_{(1)}ε(x_{(2)})\big)⊗1_H = f(x). $$

Why should the components $x_{(1)},x_{(2)},x_{(3)}$ reassemble like this to produce $x$ when it looks like they didn't all come from $x$ to begin with? In other words, this is completely fine if $x_{(1)}⊗x_{(2)}⊗x_{(3)} = (π_X⊗\mathrm{id}_H)∘π_X(x)$, but I don't see why that should be the case.

  • What is your source? I don't agree with the expression either. – Jo Mo Jun 08 '22 at 07:54
  • @JoMo It's a dissertation from a few years ago, so I don't want to embarrass anyone. Plus I did change the indexing from 0,1,2 to 1,2,3 because I'm learning this on my own and I didn't know any better. – Elizabeth Henning Jun 08 '22 at 17:49
  • I see, hope the explanation in my answer makes sense. Also: did you maybe do a copying mistake, and in the thesis it actually says $\pi_{X^* \otimes X}(f \otimes x) = f_{(0)} \otimes x_{(0)} \otimes m(S(f_{(1)}) \otimes x_{(1)})$, or something like that? – Jo Mo Jun 09 '22 at 06:13
  • @JoMo It's very likely that I'm misunderstanding something (another reason I don't want to cite the source), but it does in fact say $(ev⊗1)∘π{X^*⊗X}(f⊗x) = (ev⊗1)∑f⊗x{(0)}⊗S(x_{(1)})x_{(2)}$ and also $π{X⊗X^*}∘coev(1_k) = π{X^*⊗X}(∑x_i⊗x^i) = ∑x_{i(0)}⊗x^i⊗x_{i(1)}S(x_{i(2)})$. – Elizabeth Henning Jun 10 '22 at 02:37

1 Answers1

3

Long story short: it seems there is a mistake in your source for $\pi_{X^* \otimes X}$, and also the employed Sweedler notation was quite obfuscating.


Let me use the (very common) Sweedler notation $\pi_X(x) = x_{[0]} \otimes x_{[1]}$ for the right coactions. Here the 0-indexed factor is always in the comodule, while the ones indexed with 1, 2, ... are in $H$. For iterated coactions, I write $x_{[0,0]} \otimes x_{[0,1]} \otimes x_{[1]} = x_{[0][0]} \otimes x_{[0][1]} \otimes x_{[1]}$, and with this we can express coassociativity as $x_{[0,0]} \otimes x_{[0,1]} \otimes x_{[1]} = x_{[0} \otimes x_{[1](1)} \otimes x_{[1](2)}$, where the indicies in parentheses are the usual Sweedler notation for the Hopf algebra $H$.

As you said, on tensor products, the coaction is defined as $$ \pi_{X \otimes Y}(x \otimes y) = x_{[0]} \otimes y_{[0]} \otimes x_{[1]} y_{[1]}. $$ In terms of maps, $$ \pi_{X \otimes Y} = X \otimes Y \otimes m \circ X \otimes \tau_{H, Y} \otimes H \circ \pi_X \otimes \pi_Y, $$ where $\tau$ is the tensor factor swap in the category of vector spaces, i.e. $\tau_{X, Y}(x \otimes y) = y \otimes x$. And for the dual space, $\pi_{X^*}(f) \in X^* \otimes H \cong Hom_k(X, H)$ via $$ \pi_{X^*}(f)(x) = f(x_{[0]}) S(x_{[1]}),^{1} $$ so that $\pi_{X^*}(f) = f \otimes S \circ \pi_X$. Putting this together, we have \begin{align*} \pi_{X^* \otimes Y}(f \otimes y) &= X \otimes Y \otimes m \circ X \otimes \tau_{H, Y} \circ (f \otimes S \circ \pi_X) \otimes \pi_Y(y). \end{align*} Thus, identifying $X^* \otimes Y \otimes H \cong Hom_k(X, Y \otimes H)$, we have $$ \pi_{X^* \otimes Y}(f \otimes y)(x) = f(x_{[0]}) y_{[0]} \otimes S(x_{[1]}) y_{[1]} $$

And then we find for example \begin{align*} (\operatorname{ev}_X \otimes H \circ \pi_{X^* \otimes X})(f \otimes x) &= \operatorname{ev}_X \otimes H \circ X^* \otimes X \otimes m \circ X^* \otimes \tau_{H, X} \circ (f \otimes S \circ \pi_X) \otimes \pi_X(x) \\ &= m \circ \operatorname{ev}_X \otimes H \otimes H \circ X^* \otimes \tau_{H, X} \otimes H \circ (f \otimes S \circ \pi_X) \otimes \pi_X(x) \\ &= f(x_{[0,0]}) S(x_{[0,1]}) x_{[1]} \\ &\overset{*}{=} f(x_{[0]}) S(x_{[1](1)}) x_{[1](2)} \\ &= f(x_{[0]}) \varepsilon(x_{[1]}) 1_H \\ &= f(x) 1_H \\ &= \pi_{k} \circ \operatorname{ev}_X(f \otimes x) , \end{align*} so that evaluation indeed is a comodule map (the mysterious-seeming third equality is best seen using string diagrams, really${}^2$). The step marked $*$ uses coassociativity of the coaction as described above.


${}^1$ For completeness: In terms of maps, we have $$ \pi_{X^*} = \tau_{H, X^*} \circ S \otimes X^* \circ \operatorname{ev}_X \otimes H \otimes X^* \circ X^* \otimes \pi_X \otimes X^* \circ X^* \otimes \operatorname{coev}_X. $$


EDIT

${}^2$ To be more transparent, I'll prove it algebraically using the zig-zags of evaluation and coevalutation. I sometimes suppress the tensor product symbol. \begin{align*} &\operatorname{ev}_X \otimes H \circ \pi_{X^* \otimes X} \\ &= \operatorname{ev}_X \otimes H \circ X^* X \otimes m \circ X^* \otimes \tau_{H, X} \otimes H \circ \pi_{X^*} \otimes \pi_X \\ &= m \circ \operatorname{ev}_X \otimes H H \circ X^* \otimes \tau_{H, X} \otimes H \circ \pi_{X^*} \otimes \pi_X \\ &= m \circ \operatorname{ev}_X \otimes H H \circ X^* \otimes \tau_{H, X} \otimes H \circ \tau_{H, X^*} \otimes X H \\ & \quad \circ S \otimes X^* X H \circ \operatorname{ev}_X \otimes H X^* X H \circ X^* \otimes \pi_X \otimes X^* X H \\ & \quad \circ X^* \otimes \operatorname{coev}_X \otimes X H \circ X^* \otimes \pi_X \end{align*} Up until here, I have just inserted things, and pulled the evaluation past the multiplication. Now note that $$ \operatorname{ev}_X \otimes H H \circ X^* \otimes \tau_{H, X} \otimes H \circ \tau_{H, X^*} \otimes X H = H \otimes \operatorname{ev}_X \otimes H. $$ Then we get the following, to which we apply the zigzag (underlined, and after pulling the left-most evaluation past a lot of things) \begin{align*} &= m \circ H \otimes \operatorname{ev}_X \otimes H \\ & \quad \circ S \otimes X^* X H \circ \operatorname{ev}_X \otimes H X^* X H \circ X^* \otimes \pi_X \otimes X^* X H \\ & \quad \circ X^* \otimes \operatorname{coev}_X \otimes X H \circ X^* \otimes \pi_X \\ &= m \circ S \otimes H \circ \operatorname{ev}_X \otimes H H \circ X^* \otimes \pi_X \otimes H \\ & \quad \circ \underline{X^* X \otimes \operatorname{ev}_X \otimes H \circ X^* \otimes \operatorname{coev}_X \otimes X H} \circ X^* \otimes \pi_X \\ &= m \circ S \otimes H \circ \operatorname{ev}_X \otimes H H \circ X^* \otimes \pi_X \otimes H \circ X^* \otimes \pi_X \\ &= m \circ S \otimes H \circ \operatorname{ev}_X \otimes H H \circ X^* \otimes \pi_X \otimes H \circ X^* \otimes \pi_X . \end{align*} This maps \begin{align*} f \otimes x &\mapsto f \otimes x_{[0]} \otimes x_{[1]} \\ & \mapsto f \otimes x_{[0,0]} \otimes x_{[0,1]} \otimes x_{[1]} \\ & \mapsto f (x_{[0,0]}) x_{[0,1]} \otimes x_{[1]} \\ & \mapsto f (x_{[0,0]}) S(x_{[0,1]}) x_{[1]}, \end{align*} as claimed above.

Jo Mo
  • 2,145
  • 1
    First of all, I really appreciate your being so thorough and taking the time to write this out. I'm terrible at this kind of bookkeeping, so it's extremely helpful to see it in excruciating detail. – Elizabeth Henning Jun 10 '22 at 02:40
  • Out of curiosity, what's the indexing convention for left comodules? – Elizabeth Henning Jun 10 '22 at 02:48
  • For left comodules, people commonly write (and it weirds me out a bit haha) "$\pi_X(x) = x_{[-1]} \otimes x_{[0]}$", so that the negative indices belong to $H$. And for what it's worth, the bookkeeping is a lot easier once you know how to work with string diagrams in (braided) monoidal categories. What I did was, essentially, write out the computation in diagrams, and then I translated it into algebraic expressions. Which took me longer than expected :D – Jo Mo Jun 11 '22 at 13:44
  • Thanks! Yeah, this is brutal. It took me several passes to understand what was going on. I have only a vague familiarity with string diagrams, but this is a good argument for learning them. – Elizabeth Henning Jun 11 '22 at 14:46