Again I am confused about something regarding the cobarconstruction of a dga coalgebra $(C,d_C)$. The cobar construction of $C$ is the dga algebra $(T(s^{-1}\bar{C}),d_1+d_2)$ where $d_2$ is induced by the reduced coproduct of $C$ and $d_1$ is induced by $d_C$. This is how I would try and define $d_1$:

(Below when I take tensor products of linear maps I use the Koszul sign convention.)

First of all we get a differential $1\otimes d_C$ on $s^{-1} \bar{C}=\mathbb{K}s^{-1}\otimes \bar{C}$. Then we get a differential $$\sum (1\otimes 1)\otimes...\otimes (1\otimes d_C)\otimes... \otimes (1\otimes 1)$$ on $(\mathbb{K}s^{-1}\otimes \bar{C})^{\otimes n}$ and taking the direct sum of all of these maps gives us a differential $d_1$ on all of $T(s^{-1}\bar{C})$.

Can someone tell me if this is the correct way to go about it?

The reason I started doubting myself is because I fail trying to prove that $d_1$ and $d_2$ anti-commute. In fact I get the feeling they commute.

Here is what I did. I want to check that $d_1d_2+d_2d_1$ is zero on $s^{-1}\bar{C}$ to begin with. We have $$d_1|_{s^{-1}\bar{C}}=1\otimes d_C:\mathbb{K}s^{-1}\otimes \bar{C}\to \mathbb{K}s^{-1}\otimes \bar{C},$$ $$d_1|_{(s^{-1}\bar{C})^{\otimes 2}}=1\otimes d_C\otimes 1\otimes 1+1\otimes 1\otimes 1\otimes d_C:(\mathbb{K}s^{-1}\otimes \bar{C})^{\otimes 2}\to (\mathbb{K}s^{-1}\otimes \bar{C})^{\otimes 2}$$ and $$d_2|_{s^{-1}\bar{C}}=(1\otimes \tau\otimes 1)\circ (\Delta_s\otimes \bar{\Delta}):\mathbb{K}s^{-1}\otimes \bar{C}\to (\mathbb{K}s^{-1}\otimes \bar{C})^{\otimes 2}$$ where $\Delta_s:s^{-1}\mapsto -s^{-1}\otimes s^{-1}$ and $\bar{\Delta}$ is the reduced coproduct in $C$ and $\tau$ is the signed switching map, in our case $\tau:\mathbb{K}s^{-1}\otimes \bar{C}\to \bar{C}\otimes \mathbb{K}s^{-1}$, $s^{-1}\otimes x\mapsto (-1)^{|x|}x\otimes s^{-1}$. Then this is what I get

$$d_2\circ d_1|_{s^{-1}\bar{C}}=(1\otimes \tau\otimes 1)\circ(\Delta_s\otimes \bar{\Delta})\circ(1\otimes d_C)=$$ $$(1\otimes \tau\otimes 1)\circ(\Delta_s\otimes (\bar{\Delta}\circ d_C))$$ Using that $d_C$ is a coderivation we get $$=(1\otimes \tau\otimes 1)\circ(\Delta_s\otimes ((1\otimes d_C+d_C\otimes 1)\circ\bar{\Delta}))=$$ $$(1\otimes \tau\otimes 1)\circ (1\otimes 1\otimes 1\otimes d_C+1\otimes 1\otimes d_C\otimes 1)\circ (\Delta_s\otimes \bar{\Delta})=$$ $$(1\otimes 1\otimes 1\otimes d_C+1\otimes d_C\otimes 1\otimes 1)\circ(1\otimes \tau\otimes 1)\circ(\Delta_s\otimes \bar{\Delta})=$$ $$d_1\circ d_2|_{s^{-1}\bar{C}}.$$

If someone could help me find my error either in how i define $d_1$ or if I defined it correctly in the computations above I would be very grateful!

Edit: I see now that I am missing signs from when maps "jump over each other". Maybe this fixes things..


1 Answer 1


I will use the notation $[c]$ for $s^{-1}c$, as usual, and, more generally, the bar notation for tensor monomials. The equation $d_1d_2+d_2d_1=0$ follows from the fact $d$ is a coderivation for $\Delta$. The suspension signs are what do the trick, so you must carefully follow them, using the Koszul sign rule.

Recall that $d_1[c] = -[dc]$ and that $d_2[c] = -(-1)^{|c_{(1)}|} [c_{(1)}\vert c_{(2)}]$, where I'm using Sweedler notation. That the differential $d$ of $C$ be a coderivation for $\Delta$ means that

$$\Delta d(c) = (d\otimes 1+1\otimes d)\Delta(c)$$

Writing this is Sweedler notation, you get that

$$\tag{1} (dc)_{(1)}\otimes (dc)_{(2)} = d(c_{(1)})\otimes c_{(2)} +(-1)^{|c_{(1)}|}c_{(1)}\otimes d(c_{(2)}).$$

Applying $s^{-1}\otimes s^{-1}$ on both sides, you obtain from $(1)$ the equation

$$\tag{2} (-1)^{|(dc)_{(1)}|}[(dc)_{(1)}\vert (dc)_{(2)}] = (-1)^{|d(c_{(1)})|}[d(c_{(1)})\vert c_{(2)}] +[c_{(1)}\vert d(c_{(2)})].$$

Now I will compute $d_1d_2$ and $d_2d_1$, and observe that your desired equation follows from $(2)$. To compute $d_1d_2$, I will use that $d_1$ is, by definition, extended as a derivation for the product (concatenation of bars) in $\Omega C$, so that (the degree of $[c]$ is $|c|-1$!)

$$d_1[c_1\vert c_2] = d_1[c_1]\cdot [c_2] +(-1)^{|c_1|-1} [c_1]\cdot d[c_2].$$

This gives that

\begin{align} d_1d_2[c]&= -(-1)^{|c_{(1)}|}d_1[c_{(1)}\vert c_{(2)}]\\ &= (-1)^{|c_{(1)}|}[d(c_{(1)})\vert c_{(2)}] + (-1)^{|c_{(1)}|+|c_{(1)}|-1}[c_{(1)}\vert d(c_{(2)})]\\ &= (-1)^{|c_{(1)}|}[d(c_{(1)})\vert c_{(2)}] -[c_{(1)}\vert d(c_{(2)})] \\ &= (-1)^{|d(c_{(1)})|+1}[d(c_{(1)})\vert c_{(2)}] -[c_{(1)}\vert d(c_{(2)})] \\ &= -((-1)^{|d(c_{(1)})|}[d(c_{(1)})\vert c_{(2)}] +[c_{(1)}\vert d(c_{(2)})]) \end{align}

On the other hand, we have that

\begin{align} d_2d_1[c] &= -d_2[dc] \\ &= (-1)^{|(dc)_{(1)}|} [(dc)_{(1)} \vert (dc)_{(2)} ] \end{align}

Equation $(2)$ now shows $d_1d_2+d_2d_1$ vanishes on $s^{-1}C$, which is what you wanted.

  • $\begingroup$ Thanks! It now makes sense to me=) $\endgroup$
    – budwarrior
    Sep 17, 2018 at 16:55

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.