We work over $\mathbb{C}$. Let $X$ be a normal projective irreducible variety, and let $\mathbb{C}^*$ act nontrivially on $X$. The fixed point locus of $X$, namely $X^{\mathbb{C}^*}$, can be decomposed into a disjoint union of connected fixed point components, let us call them $F_1,\ldots,F_s$. Moreover, for any $k=1,\ldots,s$, we can define the plus cell as

$$X^+(F_k)=\{p\in X\mid \lim_{t\to 0} t\cdot p \in F_k\}$$

and similarly also the minus cell, by considering the limit at $\infty$.

If $X$ is smooth, then the celebrated theorem of Bialynicki-Birula tell us that \begin{equation} X=\bigsqcup_{i=1\ldots k} X^+(F_k)\tag{*}\label{star} \end{equation} (and similarly for the minus cell), and moreover this is an affince cell decomposition.

Question: If $X$ is only normal, do we still have a decomposition as in \eqref{star} (but in this case the cell will not be afine, as explained in Białynicki-Birula decomposition for singular projective variety)?

I've tried to search on the literature, but what I've found is usually much more deep of what I'm asking and I have a hard time trying to translate that result in my specific case. Any help, or reference, would be much appreciated. Thanks in advance!

  • $\begingroup$ In your question, did you intend to write, "If $X$ is only normal, ..."? $\endgroup$ Commented Jun 7, 2022 at 10:11
  • $\begingroup$ Dear @JasonStarr, sure, thanks for spotting the typo! $\endgroup$ Commented Jun 7, 2022 at 10:29
  • 1
    $\begingroup$ Have you looked at Hausel-Hitchin arXiv:2101.08583, Section 2.1? $\endgroup$
    – Balazs
    Commented Jun 7, 2022 at 10:35
  • $\begingroup$ Such a generalization is I think due to Drinfeld and employed in several papers by Jelisiejew. $\endgroup$ Commented Jun 7, 2022 at 10:46
  • $\begingroup$ The references: arXiv:1308.2604 (Drinfeld), arXiv:1710.06124 (Jelisiejew, see Section 3), arXiv:1805.11558 (generalization to reductive groups, Jelisiejew-Sienkiewicz) $\endgroup$ Commented Jun 7, 2022 at 10:50

1 Answer 1


Yes, \eqref{star} is always a disjoint union (that's obvious). Moreover, each set $X^+(F_k)$ is locally closed and the map $x\mapsto\lim_{t\to0}t\cdot x$ induces an affine morphism $\pi_k:X^+(F_k)\to F_k$. In general, the morphism $\pi_k$ is not a fiber bundle anymore.

These assertions can be easily reduced to the smooth case by using a theorem of Sumihiro (Sumihiro, Hideyasu: Equivariant completion. J. Math. Kyoto Univ. 14 (1974), 1–28) according to which $X$ can be embedded equivariantly into a projective space.

The assertions are more generally valid for any normal complete variety. The paper "Konarski, Jerzy: Decompositions of normal algebraic varieties determined by an action of a one-dimensional torus. Bull. Acad. Polon. Sci. Sér. Sci. Math. Astronom. Phys. 26 (1978), 295–300" seems to be a good reference.

  • 1
    $\begingroup$ Dear Prof. Knop, thank you very much for the answer and the references. Before accepting it though, I have few questions. You say at the beginning "it's obvious": are you referring to the fact that the union is disjoint (and I agree with you on this), or are you referring to the fact that $X$ coincides with the union? I was precisely asking about the second point, but with your suggestion I think I know how to prove the second question (see the comment below): $\endgroup$ Commented Jun 8, 2022 at 6:27
  • 1
    $\begingroup$ Using Sumihiro's theorem, we embed $X$ $\mathbb{C}^*$-equivariantly into a $\mathbb{P}^n$. Now we use the BB-decomposition of the projective space, obtaining $$\mathbb{P}^n=\bigsqcup_{i=1}^k (\mathbb{P}^n)^+(F_i),$$ where $F_i$ are the connected fixed point components of $\mathbb{P}^n$ under the $\mathbb{C}^*$-action. Then since $X$ is $\mathbb{C}^*$-invariant we consider the intersection of the BB-decomposition with $X$, obtaining $$\mathbb{P}^n\cap X=X=\bigsqcup_{i=1}^k X^+(F_i\cap X),$$ where maybe some intersection will be empty, but we still obtain a $+$-decomposition of $X$. $\endgroup$ Commented Jun 8, 2022 at 6:28
  • 1
    $\begingroup$ @Mago: That's correct. That $(*)$ is a disjoint union follows from the fact that for every $x$ the limit $\lim_{t\to0} t\cdot x$ exists (since $X$ is complete) and is unique (since $X$ is separated) and that the limit point is a $\mathbb C^*$-fixed point. But all this follows also from the Sumihiro argument once you accept the assertions for the projective space. $\endgroup$ Commented Jun 8, 2022 at 14:44
  • 1
    $\begingroup$ @Mako: One more thing: Observe that $F_i\cap X$ might have several connected components. Then $X^+(F_i\cap X)$ decoposes accordingly. $\endgroup$ Commented Jun 8, 2022 at 14:54
  • $\begingroup$ Dear Prof. Knop, I see. Thank you very much for the clear explanation! $\endgroup$ Commented Jun 9, 2022 at 6:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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