MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I am trying to understand the theory of cubical structures and am interested in knowing if a disconnected commutative group variety whose identity component is a semi-abelian variety satisfies the theorem of the cube.

Recall that if $X$ is an abelian variety and $L$ is a line bundle on $X$ that is rigidified along the identity sectionn, then the Theorem of the Cube implies that the line bundle

$$ \Theta(L) := m_{123}^{*}(L) \otimes m_{12}^{*}(L^{-1}) \otimes m_{13}^{*}(L^{-1}) m_{23}^{*}(L^{-1}) \otimes m_{1}^{*}(L) \otimes m_{2}^{*}(L) \otimes m_{3}^{*}(L) $$

admits a unique trivialization that makes

$$ \Lambda(L) := m^{*}(L) \otimes p_1^{*}(L^{-1}) \otimes p_{2}^{*}(L^{-1}) $$

into a symmetric biextension of $X \times X$ by $\mathbb{G}_{m}$.

Here $m$ denotes the multiplication map, $p_i$ the projection maps, and $m_{\underline{i}}$ the morphism $X \times X \times X \to X$ given by summing the coordinates whose indices are in $\underline{i}$.

More generally, Breen proved this fact remains true when $X$ is a semi-abelian variety.

Does this statement remain true if we allow $X$ to have non-trivial component group?

If not, what is a example of a rigidified line bundle that does not have canonical cubical structure.

Does the theorem remain valid you rigidify along $1$ fixed point on every component (rather along the identity element)?

Added As BCnrd notes, over a more general base the formulas should be modified by adding the term $0^{*}(L^{\otimes \pm 1})$, which should be thought of as $m_{\emptyset}^{*}(L^{\otimes \pm 1})$. This (rigidified) line bundle is trivial when the base if a field, but not in general.

share|cite|improve this question
You've changed notation there a bit. Is $G=X$? And I'm not sure all your pullbacks end up over the same space - $m_i$ and $m_{ij}$ $i,j=1,2,3$ don't have the same codomain. – David Roberts Dec 6 '10 at 0:04
The $m_\underline{i}$ really do have the same codomain. For example, in coordinates, $m_3(x,y,z)=z=p_3(x,y,z)$ and $m_{2,3}(x,y,z)=y \cdot z$, just as described by 'summing the coordinates whose indices are in $\underline{i}$'. – Eric Peterson Dec 6 '10 at 0:51
Dear jlk: It is more natural to tack on an additional tensor factor of $0^{\ast}(L)^{-1}$ where $0:G^3 \rightarrow G$ is the constant map to the origin. When working over a field this has no serious effect, but in the relative setup (which makes various Yoneda-type arguments more effective) it is clearly necessary. – BCnrd Dec 6 '10 at 1:14
@BCnrd: Thank you for the comment. @David Roberts: Thanks. Fixed. – jlk Dec 6 '10 at 1:18
Not every component of $X$ necessarily has a $k$-rational point (the component group could be non-constant). Can you work out something reasonable when $X$ is a finite etale commutative group? If not, there seems little point in considering the more general case. – BCnrd Dec 6 '10 at 3:41
up vote 4 down vote accepted

Here is an explanation why connectedness is important. Let's work over ${\mathbb C}$. The Theorem of the Cube can be stated as follows: If $s:X\to X$ is a shift by a fixed element $g\in X$, then $s^*L\otimes L^{-1}$ satisfies the Theorem of the Square. The reason it holds is because $s^*L\otimes L^{-1}$ is topologically trivial, which in turn is true because $g$ can be continuously deformed into $0$. The last step depends on $X$ being connected.

For an explicit counterexample, take $X=E\times ({\mathbb Z}/{2\mathbb Z})$ where $E$ is an elliptic curve, and let $L$ be a line bundle whose degrees on the two components differ. I am not sure how rigidifying at any number of points is going to change this.

share|cite|improve this answer
@t3suji: thanks! – jlk Dec 6 '10 at 21:12

Uniqueness may fail if $X$ is a finite group: any two cube structures "differ" by a "quadratic" map $X\to \mathbb{G}_m$. If, say, $X=\mathbb{Z}/n\mathbb{Z}$ and $\zeta$ is an $n$-th root of unity, then $m\mapsto \zeta^{m^2}$ is such a map.

This of course implies non-uniqueness whenever the component group of $X$ admits nontrivial quadratic maps to $\mathbb{G}_m$.

share|cite|improve this answer
@Laurent Moret-Bailly: Thanks! – jlk Dec 6 '10 at 21:15

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.