Non-decomposability of the de Rham complex and non-semisimplicity of the Sen operator
Abstract: We describe the obstruction to decomposing in degrees $\leq p$ the de Rham complex of a smooth variety over a perfect field $k$ of characteristic $p$ that lifts over $W_2(k)$, and show that there exist liftable smooth projective varieties of dimension $p+1$ whose Hodge-to-de Rham spectral sequence does not degenerate at the first page. We also describe the action of the Sen operator on the de Rham complex in degrees $\leq p$ and give examples of varieties with a non-semisimple Sen operator. Our methods rely on the commutative algebra structure on de Rham and Hodge-Tate cohomology, and are inspired by the properties of Steenrod operations on cohomology of cosimplicial commutative algebras. The example of a non-degenerate Hodge-to-de Rham spectral sequence relies on a non-vanishing result on cohomology of groups of Lie type. We give applications to other situations such as describing extensions in the canonical filtration on de Rham, Hodge, and \'etale cohomology of an abelian variety equipped with a group action. We also show that the de Rham complex of a smooth variety over $k$ is formal as an $E_{\infty}$-algebra if and only if the variety lifts to $W_2(k)$ together with its Frobenius endomorphism.
Paper Prompts
Sign up for free to create and run prompts on this paper using GPT-5.
Top Community Prompts
Explain it Like I'm 14
Overview
This paper studies a basic tool in algebraic geometry called the de Rham complex, which helps mathematicians compute “shape-like” information (cohomology) of geometric objects called varieties. The paper focuses on varieties over a field with characteristic (think arithmetic done modulo a prime number ). A famous result says that if such a variety can be “lifted” slightly (from mod to mod ), then the de Rham complex breaks into simple pieces in low degrees. The big questions here are: does this breaking continue one step further, and what stops it if it doesn’t?
The author shows:
- In general, the breaking (or “decomposition”) stops at the next step: there is a precise obstruction that prevents it.
- There are concrete examples where the usual shortcut through the de Rham complex fails, so a standard spectral sequence (a multi-step calculation tool) does not collapse early.
- A related “Sen operator,” which acts like a degree-counting device on the de Rham complex, can be non-simple (it mixes data in a way that can’t be fully untangled by a change of coordinates).
What questions did the paper ask?
In simple terms, the paper asks:
- If a smooth variety over a field of characteristic can be lifted a little bit (from mod to mod ), does its de Rham complex continue to split into simple parts one degree further than previously known?
- If not, what exactly is the “roadblock” (the obstruction) that stops the splitting?
- How does a special operator (the Sen operator), which watches how the complex behaves with respect to , act? Can it fail to be “nice” (non-semisimple)?
- Can we find examples that clearly show the failure of splitting and the non-simplicity of the Sen operator?
- What does this tell us about other cohomology theories and about when the de Rham complex behaves as simply as possible (is “formal”)?
How did the author approach the problem?
The methods use ideas that can be pictured with everyday analogies:
- Decomposing in low degrees: Think of the de Rham complex as a layered machine that processes information. A famous theorem (by Deligne and Illusie) says if the variety lifts to (this is like upgrading your arithmetic from mod to mod ), then in degrees less than the machine can be split into simpler independent parts. The paper asks if this splitting continues one more layer (at degree ).
- The obstruction: The author identifies a specific “blocker” that prevents the split at degree . It’s built from two ingredients: 1) Whether the Frobenius map (raising to the th power) can be lifted along with the variety (this gives an “obstruction class” ). 2) A characteristic class called , built from the cotangent bundle (the package of “infinitesimal directions” on the variety). These combine, and then a “Bockstein” operation (which measures how information changes between mod and mod ) turns their product into the final obstruction.
- The Sen operator: When a lift to exists, a canonical operator acts on the de Rham complex. On the th layer (degree ), it acts like multiplication by . But the interesting part is how it acts on extensions between layers. The paper computes the class measuring this action in degree , and links it to the same two ingredients above.
- Algebraic structure and symmetry: A key technical idea is that the de Rham complex is not just a list of groups; it has a multiplication (like a ring). The author uses this structure carefully, along with tools from “derived” algebra (which manage hidden homotopies), and analogies to classical “Steenrod operations” (special moves that only appear in characteristic ).
- Concrete constructions: To show the obstruction is really nonzero, the author builds examples using:
- Group cohomology of groups like and (matrix groups).
- Abelian varieties (higher-dimensional analogs of elliptic curves) with group actions.
- Special choices where coherent cohomology and de Rham cohomology behave differently because Frobenius acts differently on them.
What are the main results?
Here are the key findings, stated plainly:
- Splitting stops at degree : Even if the variety lifts to , the de Rham complex does not, in general, split into simple parts one step beyond what was known. The author computes the exact obstruction class preventing this, and shows it can be nonzero.
- A concrete non-degenerating example: There are smooth projective varieties of dimension that lift all the way to , but whose Hodge-to-de Rham spectral sequence does not collapse at the first page. In other words, you really need multiple steps to compute the de Rham cohomology—no shortcut.
- The obstruction formula: The degree- extension class is exactly
- “Bockstein of (obstruction to lifting Frobenius × the α-class of the cotangent bundle).”
- In symbols: the only possibly nonzero component is in and equals
For , the class can be written down very concretely using a short exact sequence involving symmetric and exterior squares.
- The Sen operator can be non-semisimple: The paper provides examples where the Sen operator is not “diagonalizable,” meaning it has a nilpotent part and mixes information in a way that cannot be untangled. The degree- class controlling its action is computed as
When this is nonzero, the operator is not semisimple.
- Formality criterion: The de Rham complex is as simple as it can be as an -algebra (this is called “formal”) if and only if the variety lifts to together with its Frobenius map. So lifting the variety is not enough—you must also be able to lift Frobenius.
- Applications to abelian varieties with group actions: The paper uses its general machinery to describe extension classes in de Rham, Hodge, and étale cohomology when an abelian variety carries a group action. It also proves that for certain abelian schemes, coherent cohomology splits while de Rham cohomology does not—showing the subtle role of Frobenius.
- A neat side result about elliptic curves: For large enough , the splitting behavior of -equivariant coherent cohomology of (where is an elliptic curve) detects whether is supersingular. This ties decomposition phenomena directly to the nature of Frobenius on .
Why is this important?
- Sharpening a classic theorem: The Deligne–Illusie theorem gave splitting up to degree . This paper shows that the next step is blocked in general, and gives the exact formula for the blockage. That’s a precise and valuable refinement.
- Understanding -phenomena: Many surprising features occur only in characteristic (e.g., Steenrod operations, Frobenius behavior). This paper uncovers a new, concrete connection among Frobenius, the Bockstein map, and the structure of the de Rham complex.
- Operator insight: The Sen operator appears in -adic Hodge theory and related areas. Showing it can be non-semisimple on de Rham cohomology, and computing the class that measures this, gives clearer understanding of -adic structures on cohomology.
- Practical criteria: The formality criterion and the supersingularity test give clean, usable conditions that link deep geometry (lifting with Frobenius, Frobenius action on ) to the behavior of entire complexes.
- Broader methods: The techniques—using multiplicative structures, derived symmetric powers, and comparisons to Steenrod operations—are widely applicable and may inform future work on higher-degree obstructions and other cohomology theories.
In short, the paper explains exactly why and how the de Rham complex fails to split one step further in characteristic , computes the operator that detects this failure, and builds compelling examples that make the phenomena visible and useful.
Collections
Sign up for free to add this paper to one or more collections.