Introduction
In the representation theory of finite-dimensional algebras, silting complexes play a pivotal role as a generalization of tilting modules, facilitating the comparison of derived categories. A fundamental structural question, known as the Complement Question, asks whether every presilting complex (a "partial" silting complex) can be completed into a full silting complex by adding appropriate direct summands. While this completion is always possible for classical tilting modules (Bongartz's completion), its validity for general silting complexes remained an open problem. This paper provides a definitive negative answer by constructing a specific counterexample.
Geometric Modeling of Derived Categories
To address the Complement Question, the authors utilize a geometric model for the derived categories of gentle algebras. This framework maps abstract algebraic objects to topological ones:
- Indecomposable complexes are represented by graded arcs on a marked surface.
- Morphisms between complexes correspond to intersections between these arcs.
- The presilting condition is satisfied if an arc lacks self-intersections with positive indices.
The specific algebra $\Lambda$ studied is a gentle algebra of global dimension two, modeled on a torus with a single boundary component containing four marked points ($p, q, r, s$).
The "Dead End" Construction
The core of the proof lies in identifying an "isolated" arc $\tilde{\gamma}$ on the torus. The authors demonstrate that this arc represents a presilting complex that cannot be extended.
The arc $\tilde{\gamma}$ is constructed such that it "blocks" the surface. Any additional arc $\sigma$ intended to complete the silting complex would necessarily intersect $\tilde{\gamma}$ in a way that creates a non-zero morphism in the forbidden degrees, thereby violating the presilting requirement for the direct sum $X(\tilde{\gamma}) \oplus S^\bullet$. This "geometric blocking" provides a visual and rigorous proof that some presilting complexes are indeed "dead ends."
Key Results
The paper's primary contribution is summarized in the following results:
- Theorem 0.1: There exists a presilting complex in the bounded derived category $D^b(\text{mod}\Lambda)$ that is not a direct summand of any silting complex.
- Proposition 1.4: For the specific complex $X(\tilde{\gamma})$ constructed on the marked torus, there is no indecomposable complex $S^\bullet$ such that $X(\tilde{\gamma}) \oplus S^\bullet$ is presilting.
This counterexample is a 3-term complex, highlighting that the failure of Bongartz-type completion occurs just beyond the well-understood 2-term case. Interestingly, while completion fails, the authors note that the "Rank Question"—whether a complex of the appropriate size is necessarily silting—holds true for this specific algebra, indicating that the issue lies in the extendability of small complexes rather than the properties of full-sized ones.
Conclusion and Future Directions
By settling the Complement Question in the negative, this work delineates the boundaries of silting theory. Future research may focus on characterizing the specific classes of algebras where completion remains possible, such as piecewise hereditary algebras, or exploring the behavior of "blocking" arcs on surfaces of higher genus. The reliance on gentle algebras also suggests investigating whether similar phenomena occur in broader algebraic contexts where geometric models are less established.
Summary updated