04.10.12 · algebraic-geometry / moduli

Bridgeland stability conditions

shipped3 tiersLean: none

Anchor (Master): Bridgeland 2007 *Annals of Mathematics* (2) 166 (2), 317-345 *Stability conditions on triangulated categories* (originator paper); Douglas 2001 hep-th/0107080 *D-branes, categories and N=1 supersymmetry* (the physics origin via $\Pi$-stability for D-branes on Calabi-Yau threefolds); Kontsevich-Soibelman 2008 arXiv:0811.2435 *Stability structures, motivic Donaldson-Thomas invariants and cluster transformations*; Bayer-Macri 2014 *Inventiones Math.* 198 (3), 505-590 *MMP for moduli of sheaves on K3s via wall-crossing*; Mumford 1965 *Geometric Invariant Theory* (the GIT slope-stability framework that Bridgeland stability promotes to the derived setting); Bridgeland 2008 *Duke Mathematical Journal* 141 (2), 241-291 *Stability conditions on K3 surfaces*

Intuition Beginner

A vector bundle on a curve has a numerical invariant — the slope, the average degree per dimension — and Mumford's 1963 idea was that a bundle is "balanced" exactly when every sub-bundle has strictly smaller slope. The whole moduli theory of bundles on a curve is built on this single inequality. Tom Bridgeland's 2007 idea was to take this slope picture and promote it out of the world of bundles into the world of complexes of bundles, where the natural ambient object is no longer the abelian category of sheaves but the derived category $D^{b} (X)$ of $X$ . A Bridgeland stability condition assigns a slope-like number, the phase, to every object of $D^{b} (X)$ , and singles out the balanced objects.

The phase replaces the slope, and the central charge replaces the rank-and-degree pair. A central charge $Z$ sends each object $E$ to a complex number, and the phase is the argument of that number, normalised to a real number. Each balanced object has a definite phase; a complex of bundles can be balanced in this generalised sense even when none of its constituent sheaves are. Then a Harder-Narasimhan filtration says every object decomposes uniquely into pieces of strictly decreasing phase.

What makes the framework powerful: the set of all stability conditions on $D^{b} (X)$ has the structure of a complex manifold. Walking around in this manifold, the moduli of balanced objects changes — chambers separated by walls — and crossing a wall realises a birational map between moduli spaces. The picture unifies slope stability, Gieseker stability, and the mirror-symmetry side of the story all at once.

Visual Beginner

A schematic showing the upper half-plane $H$ as the target of the central charge, with several objects of the derived category plotted as complex numbers and their phases — the angles from the positive real axis — labelled. A second panel shows the Bridgeland manifold $Stab (D^{b} (X))$ as an open patch of $C^{n}$ , with chambers separated by walls; each chamber gives a distinct moduli space of balanced objects, and crossing a wall realises a birational transition.

The picture captures the essential geometry. Each stability condition is a pair: a central charge sending objects to complex numbers, and a slicing recording which objects have which phase. Moving the central charge moves the moduli, and the moduli structure jumps at walls.

Worked example Beginner

Take $C$ a smooth projective curve of genus $g \geq 1$ . The Grothendieck group of the derived category $D^{b} (Coh C)$ is rank- $2$ , generated by the rank and degree of a complex.

Step 1. Pick a complex number $z = - d + i r$ where $r$ is the rank of an object and $d$ is its degree, so the rank lives on the imaginary axis and the degree on the negative real axis. For a vector bundle $E$ on $C$ with rank $r > 0$ and degree $d$ , the image is $z (E) = - d + i r$ , which lies in the upper half-plane $H = {z \in C : Im (z) > 0}$ .

Step 2. The phase of $E$ is the angle of $z (E)$ divided by $π$ , a real number in the interval $(0, 1]$ . Concretely $ϕ (E) = (1/ π) arctan (r / (- d))$ when $d < 0$ and the symmetric formula when $d \geq 0$ . As $d / r$ varies over the reals the phase traces the open interval $(0, 1)$ , and the limit $d \to - \infty$ at fixed $r$ gives phase approaching $0$ , while $d \to + \infty$ gives phase approaching $1$ .

Step 3. A torsion sheaf — a skyscraper $O_{p}$ at a point $p$ — has rank $0$ and degree $1$ , so $z (O_{p}) = - 1$ , which has phase $ϕ = 1$ (the negative real axis). Every skyscraper has phase $1$ , a single horizontal line on the upper boundary of $H$ .

Step 4. The shift $E [1]$ in the derived category has rank $- r$ and degree $- d$ , so $z (E [1]) = d - i r = - z (E)$ . The shift moves the image to the lower half-plane, and the phase increases by $1$ — consistent with the structural rule $P (ϕ + 1) = P (ϕ) [1]$ for slicings.

Step 5. Slope stability on $C$ (a vector bundle $E$ is slope-stable when every sub-bundle $F$ has $μ (F) < μ (E)$ ) is exactly the requirement that, in the phase picture, every sub-bundle $F \subset E$ has $ϕ (F) < ϕ (E)$ . The slope $μ = d / r$ and the phase $ϕ$ are related by a monotone transformation, so the slope inequality and the phase inequality match. Bridgeland stability on $D^{b} (C)$ with this particular central charge $z$ recovers exactly slope stability on vector bundles.

Step 6. The set of stability conditions is two-dimensional. Replace $z$ by $z^{'} = a z + b \cdot (something else)$ and the phase picture shifts, but the structure persists. Bridgeland's theorem 7.1 in his 2007 paper tells us $Stab (D^{b} (C)) ≃ C \times H$ , a copy of $C^{2}$ as a complex manifold, with the central-charge map identifying it with the open subset of $Hom (K (D^{b} (C)), C) = C^{2}$ consisting of homomorphisms whose imaginary part is positive on every effective class. Slope stability is one point in this manifold; tilting and deforming the central charge realises every other stability condition.

What this tells us: the slope stability on a curve, together with the shift in the derived category, generates a whole two-parameter family of stability conditions, and the family is itself a complex manifold of the same dimension as the rank of the Grothendieck group.

Check your understanding Beginner

Exercise (easy, multiple choice).

A Bridgeland stability condition on a triangulated category $D$ is a pair $(Z, P)$ where $Z$ is a group homomorphism $K (D) \to C$ and $P$ is a slicing. Which of the following describes the role of the central charge $Z$ ?

A. $Z$ is a quadratic form on $K (D)$ controlling self-intersection.
B. $Z$ is a linear map to $C$ ; its absolute value sets the mass and its argument sets the phase.
C. $Z$ is a multiplicative character to $C^{\times}$ .
D. $Z$ is a topological invariant detecting the Euler characteristic alone.

Hint

The central charge is the categorical analogue of the rank-and-degree pair for a vector bundle: it is a single complex number, and the angle of that complex number encodes the phase.

Answer

B. A central charge is a group homomorphism $Z : K (D) \to C$ , that is, a $Z$ -linear map. The absolute value $∣ Z (E) ∣$ is the mass of $E$ , and the argument of $Z (E)$ divided by $π$ is the phase. For a vector bundle on a curve, the standard choice $Z (E) = - de g (E) + i \cdot rk (E)$ recovers Mumford slope stability.

Formal definition Intermediate+

Throughout this section let $D$ be a triangulated category equipped with its Grothendieck group $K (D)$ , defined as the free abelian group on isomorphism classes of objects modulo the relations $[B] = [A] + [C]$ for every distinguished triangle $A \to B \to C \to A [1]$ . The canonical example is $D = D^{b} (X)$ for $X$ smooth projective over a field, with $K (D) = K_{0} (X)$ the Grothendieck group of coherent sheaves.

Definition (slicing). A slicing $P$ of $D$ is a collection of full additive subcategories $P (ϕ) \subset D$ indexed by $ϕ \in R$ , satisfying three properties.

(S1) Shift-compatibility. $P (ϕ + 1) = P (ϕ) [1]$ for every $ϕ \in R$ .

(S2) Orthogonality. If $A \in P (ϕ_{1})$ and $B \in P (ϕ_{2})$ with $ϕ_{1} > ϕ_{2}$ , then $Hom_{D} (A, B) = 0$ .

(S3) Harder-Narasimhan filtration. For every non-zero $E \in D$ there exists a finite sequence of real numbers $ϕ_{1} > ϕ_{2} > \dots > ϕ_{n}$ and a sequence of distinguished triangles $$ 0 = E_0 \to E_1 \to E_2 \to \cdots \to E_n = E $$ with $cone (E_{i - 1} \to E_{i}) \in P (ϕ_{i})$ for each $i$ .

The subcategory $P (ϕ)$ is called the subcategory of semistable objects of phase $ϕ$ . Objects of $P (ϕ)$ that are simple in the additive subcategory $P (ϕ)$ are called stable of phase $ϕ$ .

Definition (Bridgeland stability condition). A stability condition on $D$ is a pair $σ = (Z, P)$ where $Z : K (D) \to C$ is a group homomorphism (the central charge) and $P$ is a slicing, subject to the compatibility $$ \text{for every } 0 \neq E \in \mathcal{P}(\phi), \quad Z(E) \in \mathbb{R}_{>0} \cdot e^{i \pi \phi}. $$ A stability condition is locally finite if for every $ϕ \in R$ and every $ϵ \in (0, 1/2)$ the subcategory $P ((ϕ - ϵ, ϕ + ϵ))$ generated by semistable objects of phase in $(ϕ - ϵ, ϕ + ϵ)$ is a finite-length quasi-abelian subcategory. We always require local finiteness.

Equivalent definition (heart + stability function). A stability condition is equivalently a pair $(Z, A)$ where $A$ is the heart of a bounded $t$ -structure on $D$ and $Z : K (A) \to C$ is a group homomorphism (called the stability function) such that $Z (E) \in H \cup R_{< 0}$ for every $0 \neq = E \in A$ , and a Harder-Narasimhan filtration exists for every object of $A$ with respect to the slope $μ (E) = - Re Z (E) / Im Z (E)$ (with the convention $μ (E) = + \infty$ when $Im Z (E) = 0$ ). The equivalence: given $(Z, A)$ , set $P (ϕ) = {E \in A : ϕ (E) = ϕ}$ for $ϕ \in (0, 1]$ and extend by shift-compatibility (S1); given $(Z, P)$ , set $A = P ((0, 1])$ , the extension-closed subcategory generated by phases in $(0, 1]$ . This is the heart of the standard $t$ -structure on $D$ .

Definition (Bridgeland manifold). The set $Stab (D)$ is the set of locally finite Bridgeland stability conditions on $D$ , equipped with the generalised metric $$ d(\sigma_1, \sigma_2) = \sup_{0 \neq E \in \mathcal{D}} \Big( |\phi^+{\sigma_2}(E) - \phi^+{\sigma_1}(E)|, |\phi^-{\sigma_2}(E) - \phi^-{\sigma_1}(E)|, |\log(m_{\sigma_2}(E)/m_{\sigma_1}(E))| \Big), $$ where $ϕ_{σ}^{\pm} (E)$ is the maximal/minimal phase of the Harder-Narasimhan factors of $E$ and $m_{σ} (E) = \sum_{i} ∣ Z (E_{i}) ∣$ is the mass. The forgetful map is $Z : Stab (D) \to Hom (K (D), C)$ , $(Z, P) \mapsto Z$ .

Counterexamples to common slips

A slicing is not a $t$ -structure: a $t$ -structure is a pair of subcategories $(D^{\leq 0}, D^{\geq 0})$ and a slicing refines it to a real-indexed family. Every slicing produces a $t$ -structure via $D^{\leq 0} = P ((- \infty, 0])$ and $D^{\geq 0} = P ((- 1, + \infty))$ , but a $t$ -structure alone does not produce a slicing.
The compatibility $Z (E) \in R_{> 0} \cdot e^{iπ ϕ}$ forces $Z (E) \neq = 0$ for every semistable $E$ . This is the support property, and is essential for local finiteness of the metric topology.
"Stable" means simple in $P (ϕ)$ , not indecomposable. Every stable object is indecomposable, but a semistable object can be a sum of stables of the same phase (an S-equivalence class). On a curve this is the Jordan-Hölder structure of slope-semistable bundles.
A stability condition on $D^{b} (X)$ does not in general restrict to a stability condition on $Coh X$ : the heart $A$ of the $t$ -structure may differ from $Coh X$ . The phenomenon of tilting (Happel-Reiten-Smalø) replaces $Coh X$ by a torsion-pair-tilted heart, producing genuinely new stability conditions visible only in the derived setting.

Key theorem with proof Intermediate+

Theorem (Bridgeland deformation theorem; Bridgeland 2007 Theorem 1.2). Let $D$ be a triangulated category with finitely generated Grothendieck group of finite rank $n$ . The space $Stab (D)$ has the structure of a complex manifold of dimension $n$ , and the forgetful map $$ \mathcal{Z} : \mathrm{Stab}(\mathcal{D}) \to \mathrm{Hom}(K(\mathcal{D}), \mathbb{C}) = \mathbb{C}^n $$ sending $σ = (Z, P) \mapsto Z$ is a local homeomorphism.

Proof. The argument has three steps. First, set up the deformation parameter: a small perturbation of the central charge. Second, transport the slicing along the perturbation, controlling phases by the support property. Third, verify continuity in the Bridgeland metric.

Step 1: deformation parameter. Fix $σ = (Z, P) \in Stab (D)$ . The central-charge space $Hom (K (D), C)$ is a complex vector space of dimension $n$ . Pick a basis $[E_{1}], \dots, [E_{n}]$ of $K (D) \otimes Q$ realised by classes of semistable objects, and parametrise nearby central charges by $W \in Hom (K (D), C)$ with $W$ close to $Z$ in the operator norm $∥ W - Z ∥ = max_{i} ∣ W (E_{i}) - Z (E_{i}) ∣$ . The locally finite condition on $σ$ supplies a support inequality: there exists a quadratic form $Q$ on $K (D) \otimes R$ such that $Q$ is negative definite on $ker Z$ (where $Z$ vanishes) and $∣ Z (E) ∣^{2} \geq Q ([E])$ for every semistable $E$ . The support property bounds the mass below in terms of the class.

Step 2: transport the slicing. For $W$ close to $Z$ , define a candidate slicing $P_{W}$ as follows. For $ϕ \in R$ and $E \in P (ϕ)$ semistable for $σ$ , the new phase is the real number $ϕ_{W}$ such that $W (E) /∣ W (E) ∣ = e^{iπ ϕ_{W}}$ . The choice of $ϕ_{W} \in R$ (versus $ϕ_{W} + 2$ ) is fixed by requiring $∣ ϕ_{W} - ϕ ∣ < 1/2$ for $W$ sufficiently close to $Z$ — the support inequality of Step 1 forces $∣ W (E) /∣ W (E) ∣ - Z (E) /∣ Z (E) ∣∣ < 2 sin (π /2 - ϵ)$ for the perturbation, which in turn forces the angle change to be less than $1/2 - ϵ$ . Define $P_{W} (ϕ_{W})$ to be the additive subcategory generated by all $E \in P (ϕ)$ with new phase exactly $ϕ_{W}$ . Extend to all real phases by shift-compatibility (S1).

The orthogonality property (S2) for $P_{W}$ is inherited from $P$ because the phase ordering on transported objects matches the phase ordering on original objects (a small perturbation of an order-preserving bijection on a finite-length subcategory is still order-preserving). The Harder-Narasimhan filtration (S3) is inherited by lifting the filtration of $E$ for $σ$ to a filtration for $σ_{W} = (W, P_{W})$ — every semistable factor remains semistable, with a slightly shifted phase, and the filtration order is preserved.

Step 3: continuity in the Bridgeland metric. It remains to show that the map $W \mapsto σ_{W} = (W, P_{W})$ is a continuous section of $Z$ on an open neighbourhood of $Z$ , and that it is a homeomorphism onto its image. Continuity: the metric $d (σ, σ_{W})$ is bounded by $max_{E} ∣ ϕ_{W} (E) - ϕ (E) ∣$ over semistable $E$ , which by Step 1 is bounded by a constant times $∥ W - Z ∥$ . So $W \mapsto σ_{W}$ is Lipschitz. Bijectivity onto a neighbourhood of $σ$ : any other stability condition $σ^{'} = (W, P^{'})$ with $W$ close to $Z$ and $d (σ^{'}, σ)$ small must have the same Harder-Narasimhan structure on every semistable-for- $σ$ object (the small-distance condition forces $∣ ϕ^{'} (E) - ϕ (E) ∣ < 1/2$ , and the support inequality combined with (S2) makes the slicings agree). Thus the section is bijective onto its image.

Putting Steps 1-3 together: $Z$ admits a continuous local inverse near every $σ$ , hence is a local homeomorphism, and the complex structure on $Hom (K (D), C) ≅ C^{n}$ pulls back to a complex structure on $Stab (D)$ making the forgetful map a local biholomorphism. $□$

Bridge. The Bridgeland deformation theorem builds toward the wall-crossing geometry of moduli of stable objects, and the central insight is that the entire stability picture on a triangulated category is parametrised by a finite-dimensional complex manifold, with the forgetful map to the central-charge space the foundational reason chamber decompositions in moduli theory are rational polyhedral. The bridge is exactly the deformation theorem: moving the central charge by a small amount transports every semistable object to a nearby semistable object of nearby phase, and identifies $Stab (D)$ with $Hom (K (D), C)$ locally. This appears again in 04.10.06 (slope stability on a curve), where the worked example shows $Stab (D^{b} (C)) ≃ C^{2}$ with the slope picture realised as one explicit chamber, and appears again in 04.10.05 (Hilbert scheme), where wall-crossing in $Stab (D^{b} (surface))$ realises every birational model of $Hilb^{n}$ as a moduli of $σ$ -stable objects. Putting these together, the deformation theorem is exactly the categorical promotion of Mumford's slope-stability framework: slope stability is one point in the manifold of stability conditions, and the manifold structure organises all of moduli theory into a single global picture. This is dual to the polarisation-chamber picture of VGIT (Variation of GIT, 04.10.09): VGIT moves through the ample cone of polarisations, Bridgeland stability moves through the central-charge space, and the two pictures coincide in the cases where both apply.

Exercises Intermediate+

Exercise 3 (medium, symbolic).

Let $D$ be a triangulated category and let $σ = (Z, P) \in Stab (D)$ . Show that for every $0 \neq = E \in P (ϕ)$ , the central charge is determined by the mass and phase via $Z (E) = m e^{iπ ϕ}$ for some $m > 0$ , and conversely that the mass-phase pair $(m, ϕ)$ determines a unique stability condition on the rank- $1$ Grothendieck group spanned by $[E]$ .

Hint

The compatibility axiom reads $Z (E) \in R_{> 0} \cdot e^{iπ ϕ}$ . The positive real ray and the phase angle together determine the complex number.

Answer

The compatibility axiom says $Z (E) \in R_{> 0} \cdot e^{iπ ϕ}$ , meaning $Z (E) = m \cdot e^{iπ ϕ}$ for some positive real $m$ . The pair $(m, ϕ) \in R_{> 0} \times R$ encodes both the mass $m = ∣ Z (E) ∣$ and the phase $ϕ = ar g (Z (E)) / π$ .

Conversely, on the rank- $1$ Grothendieck group $Z [E]$ , the choice of $(m, ϕ)$ determines the central charge $Z (n [E]) = n \cdot m \cdot e^{iπ ϕ}$ for every integer $n$ , and the slicing places $E$ in $P (ϕ)$ and $E [k]$ in $P (ϕ + k)$ for every shift $k \in Z$ . The Harder-Narasimhan filtration is the identity (every object is already a single semistable factor), and the orthogonality axiom is vacuous (only one phase orbit exists modulo shift). Hence the mass-phase pair determines a unique stability condition.

This is the rank- $1$ skeleton of the deformation theorem: $Stab$ of the rank- $1$ derived category is $R_{> 0} \times R ≅ C^{\times}$ , and the forgetful map to $Hom (Z, C) ≅ C$ is the homeomorphism $C^{\times} ≅ C^{\times}$ . Rubric: full credit for the mass-phase reconstruction and the rank- $1$ skeleton check.

Exercise 4 (medium, symbolic).

Let $C$ be a smooth projective curve of genus $g \geq 1$ . Verify that $Z (E) = - de g (E) + i \cdot rk (E)$ defines a Bridgeland stability function on $Coh C$ (in the sense of the equivalent heart-and-stability-function definition).

Hint

Check three properties: $Z$ is a group homomorphism on $K (Coh C) ≅ Z^{2}$ ; $Z (E) \in H \cup R_{< 0}$ for every non-zero $E \in Coh C$ ; the Harder-Narasimhan filtration exists for slope.

Answer

The Grothendieck group $K (Coh C)$ is $Z^{2}$ , generated by the class of $O_{C}$ (with rank $1$ , degree $0$ ) and the class of a skyscraper $O_{p}$ (with rank $0$ , degree $1$ ). The map $Z (E) = - de g (E) + i \cdot rk (E)$ is $Z$ -linear (rank and degree are additive on short exact sequences), hence a group homomorphism.

For a non-zero coherent sheaf $E$ on $C$ , either $E$ is torsion-free of positive rank ( $rk (E) \geq 1$ ) giving $Im Z (E) = rk (E) > 0$ , or $E$ is a torsion sheaf ( $rk (E) = 0$ , $de g (E) > 0$ ) giving $Re Z (E) = - de g (E) < 0$ and $Im Z (E) = 0$ . In both cases $Z (E) \in H \cup R_{< 0}$ .

The Harder-Narasimhan filtration for slope $μ (E) = de g (E) / rk (E)$ (with $μ = + \infty$ for torsion) on coherent sheaves on a curve is classical (Harder-Narasimhan 1975 Math. Ann. 212): every coherent sheaf $E$ has a unique filtration $0 = E_{0} \subset E_{1} \subset \dots \subset E_{n} = E$ with semistable successive quotients of strictly decreasing slope. The first step is the torsion subsheaf $E_{tors}$ (with $μ = + \infty$ ), and the remaining filtration is the slope filtration on $E / E_{tors}$ .

Hence $(Z, Coh C)$ is a stability function with Harder-Narasimhan property, hence a Bridgeland stability condition on $D^{b} (Coh C)$ in the equivalent heart-and-stability-function definition. Rubric: full credit for all three verifications.

Exercise 5 (medium, short-answer).

State precisely why the deformation theorem produces a complex manifold, not merely a real manifold. Locate the step in the proof where the complex structure enters.

Hint

The central-charge space $Hom (K (D), C)$ is a complex vector space, and the forgetful map is the source of the complex structure.

Answer

The space $Hom (K (D), C)$ is a complex vector space of complex dimension $n = rk K (D)$ — its complex structure is the canonical one on the complex-valued homomorphism space. The deformation theorem proves the forgetful map $Z : Stab (D) \to Hom (K (D), C)$ is a local homeomorphism. A local homeomorphism into a complex manifold pulls back the complex structure: each chart of $Stab$ is identified by $Z$ with an open subset of $C^{n}$ , and the transition between two charts $Z^{- 1} (U_{α}) \to Z^{- 1} (U_{β})$ is the identity in central-charge coordinates, hence biholomorphic. Therefore $Stab (D)$ inherits a complex manifold structure of dimension $n$ .

The complex structure enters precisely in Step 3 of the proof: the section $W \mapsto σ_{W}$ is continuous in $W$ , and the local inverse of $Z$ is identified with a holomorphic section because $W$ is a complex parameter. The phase function $ϕ_{W}$ depends on $W$ via $ar g (W (E))$ , which is real-analytic in $W$ , but the central charge $W (E) \in C$ is holomorphic in $W$ , and the complex structure on $Stab$ is the one pulled back from $W$ , not from $ϕ_{W}$ .

Rubric: full credit for identifying the complex structure as pulled back from the complex-vector-space structure on $Hom (K, C)$ via the local homeomorphism $Z$ .

Exercise 6 (medium, symbolic).

On the derived category $D^{b} (C)$ of a smooth projective curve of genus $g \geq 1$ , construct the universal cover of $Stab (D^{b} (C))$ and identify it with $C \times H$ .

Hint

Bridgeland 2007 §7 shows $Stab (D^{b} (C)) ≅ C^{2}$ for $g \geq 1$ . The universal cover is the complex manifold itself, since $C^{2}$ is simply connected.

Answer

For a smooth projective curve of genus $g \geq 1$ , Bridgeland 2007 Theorem 7.1 proves that $Stab (D^{b} (C))$ is connected and simply connected, isomorphic as a complex manifold to $C \times H$ . The isomorphism sends a stability condition $σ$ to the pair $(lo g Z (O_{p}), Z (O_{C}))$ , where $O_{p}$ is a skyscraper at a point and $O_{C}$ is the structure sheaf — the rank- $2$ Grothendieck group of $D^{b} (C)$ is spanned by these two classes.

The central charge $Z$ is determined by its values on the two generators: $Z (O_{p}) \in C^{\times}$ (any non-zero complex number) and $Z (O_{C}) \in H$ (a complex number with positive imaginary part, since $O_{C}$ has rank $1$ ). The first coordinate $lo g Z (O_{p})$ ranges over $C$ , and the second $Z (O_{C})$ ranges over the upper half-plane $H$ . So the central-charge space, restricted to the image of the forgetful map, is $C \times H ≅ C^{2}$ (since $H ≅ C$ via the conformal map $z \mapsto i + z$ or the simpler holomorphic embedding).

Connectedness and simple connectedness follow from the explicit parametrisation: every stability condition is uniquely determined by the pair $(Z (O_{p}), Z (O_{C}))$ , and the slicing is reconstructed from the slope/phase picture. Hence $Stab (D^{b} (C))$ is its own universal cover.

Rubric: full credit for the identification with $C \times H$ , the connectedness and simple connectedness assertions, and the explicit parametrisation by the two generator-central charges.

Exercise 7 (hard, short-answer).

Explain why on $P^{1}$ the Bridgeland stability picture is degenerate compared to higher-genus curves, and describe the deformation behaviour of $Stab (D^{b} (P^{1}))$ .

Hint

$D^{b} (P^{1})$ is equivalent to $D^{b} (rep Q)$ for $Q$ the Kronecker quiver, by Beilinson's resolution. The stability picture is the affine $A_{1}$ -type quiver picture.

Answer

For $X = P^{1}$ , the derived category $D^{b} (P^{1})$ is equivalent (by Beilinson's resolution of the diagonal, 1978) to the derived category of representations of the Kronecker quiver $Q : ∙ ⇉ ∙$ . The Grothendieck group is rank $2$ , generated by the structure sheaf $O$ and the twist $O (1)$ (or equivalently the two simple representations of $Q$ ).

The stability manifold $Stab (D^{b} (P^{1}))$ is connected and is the universal cover of $C^{2} ∖ Δ$ for $Δ$ the discriminant locus where the two generator-central charges are proportional. Concretely, $Stab (D^{b} (P^{1}))$ is isomorphic to $C^{2}$ as a complex manifold (Okada 2006 Internat. Math. Res. Notices 16, Bayer-Macri 2011), but the structure of moduli of stable objects is highly degenerate: for generic central charge, the only stable objects of any phase are the line bundles $O (n)$ and their shifts, and the moduli space of stable objects of fixed class is either a point or empty.

The contrast with higher genus: for $g \geq 1$ , the moduli of slope-stable bundles of rank $r > 1$ and coprime degree $d$ is a positive-dimensional smooth projective variety of dimension $r^{2} (g - 1) + 1$ , parametrising a continuous family of stable objects at each phase. For $g = 0$ ( $X = P^{1}$ ), $r^{2} (g - 1) + 1 = 1 - r^{2} \leq 0$ for $r \geq 1$ , and the moduli either degenerates to a point or has no stable points at all.

The deformation theorem still applies: $Stab (D^{b} (P^{1}))$ is a connected complex manifold of dimension $2$ , and the forgetful map to $C^{2}$ is a local homeomorphism. The degeneracy is in the moduli of $σ$ -stable objects, not in the stability manifold itself.

Rubric: full credit for the Beilinson identification with Kronecker-quiver representations, the dimension count, and the contrast with $g \geq 1$ .

Exercise 8 (hard, short-answer).

State the wall-and-chamber decomposition of $Stab (D)$ for a triangulated category with finitely generated Grothendieck group, and explain how wall-crossing produces birational maps between moduli of $σ$ -stable objects.

Hint

A wall is a real codimension- $1$ subset of $Stab (D)$ across which the set of $σ$ -stable objects of a fixed class jumps. Bayer-Macri 2014 made this precise for K3 surfaces.

Answer

Fix a class $v \in K (D)$ . The wall-and-chamber decomposition of $Stab (D)$ for $v$ is the locally finite collection of real codimension- $1$ subsets — the walls — across which the moduli space $M_{σ} (v)$ of $σ$ -semistable objects of class $v$ jumps. Walls are characterised numerically: $σ_{0}$ lies on a wall for $v$ if there exists a class $w \neq = 0, v$ with $ar g Z_{σ_{0}} (w) = ar g Z_{σ_{0}} (v)$ (the central charges of $v$ and $w$ become collinear). The complement is a disjoint union of open chambers, and inside each chamber $M_{σ} (v)$ is constant (as a coarse moduli scheme).

Bayer-Macri 2014 Inventiones 198 (3) proved for $D = D^{b} (X)$ with $X$ a K3 surface: every birational model of $M_{σ} (v)$ arises as the moduli $M_{σ^{'}} (v)$ for some $σ^{'}$ in an adjacent chamber, and the birational map $M_{σ} (v) ⇢ M_{σ^{'}} (v)$ across a wall is a Mori-theoretic flip or flop in many cases — a divisorial contraction in others. The wall-crossing map is explicit at the level of objects: at the wall $σ_{0}$ , an object $E$ of class $v$ admits a Jordan-Hölder decomposition into pieces of phase equal to $ϕ_{σ_{0}} (v)$ , and the two adjacent chambers select different stable representatives in the S-equivalence class.

The wall-crossing identity for numerical invariants (Donaldson-Thomas invariants) is Kontsevich-Soibelman 2008 arXiv:0811.2435: at the wall, the generating function for $DT_{σ} (v)$ across $σ$ undergoes a Kontsevich-Soibelman wall-crossing transformation in the motivic Hall algebra, controlled by the Lie-algebra structure of the lattice $K (D)$ with the Euler pairing.

This formalism makes Bridgeland stability the universal organising principle for both the moduli geometry and the enumerative geometry of derived categories.

Rubric: full credit for the wall characterisation, the chamber-constancy statement, the Bayer-Macri MMP result, and the Kontsevich-Soibelman wall-crossing formula.

Advanced results Master

Theorem (Bridgeland's main theorem on K3 surfaces; Bridgeland 2008 Theorem 1.1). Let $X$ be a smooth complex projective K3 surface. There exists a connected component $Stab^{†} (D^{b} (X)) \subseteq Stab (D^{b} (X))$ such that the forgetful map $$ \mathcal{Z} : \mathrm{Stab}^\dagger(D^b(X)) \to \mathcal{P}^+_0(X) \subset \mathrm{Hom}(K^{\mathrm{num}}(D^b(X)), \mathbb{C}) $$ is a covering map onto a connected component $P_{0}^{+} (X)$ of the period domain $P^{+} (X)$ of generalised Calabi-Yau structures on the Mukai lattice $H (X, Z)$ .

The Mukai lattice $H (X, Z) = H^{0} (X, Z) \oplus H^{2} (X, Z) \oplus H^{4} (X, Z)$ with the Mukai pairing $⟨ v, w ⟩ = - v_{0} w_{4} + v_{2} \cdot w_{2} - v_{4} w_{0}$ is the analogue of the rank-degree Grothendieck group for surfaces. The Mukai vector of a coherent sheaf $E$ is $v (E) = ch (E) td (X) \in H (X, Z)$ , and Bridgeland's central charge on the K3 component is $Z_{β, ω} (E) = ⟨ v (E), exp (β + iω)⟩$ for $β + iω$ a complexified Kähler class. The period domain $P^{+} (X)$ is the open subset of $P (H (X, C))$ where the Mukai pairing is positive on a real $2$ -plane — the natural target for the complex-structure-and-Kähler-class data on $X$ .

Theorem (Bayer-Macri MMP correspondence; Bayer-Macri 2014 Theorem 1.2). Let $X$ be a smooth complex projective K3 surface and let $v$ be a primitive Mukai vector with $v^{2} \geq - 2$ . Every smooth projective birational model of $M_{σ} (v)$ (for $σ$ in the interior of a chamber of $Stab^{†} (D^{b} (X))$ ) is itself a moduli $M_{σ^{'}} (v)$ for some $σ^{'}$ in another chamber. Moreover the chamber decomposition of $Stab^{†} (D^{b} (X))$ projects onto the Mori chamber decomposition of the nef cone $\overline{Nef} (M_{σ} (v))$ .

The Bayer-Macri correspondence identifies the birational geometry of the moduli space (in the Mori-theoretic sense) with the chamber decomposition of the Bridgeland manifold. This is the cleanest known instance of "wall-crossing = minimal model program": every flip and every divisorial contraction in the MMP for $M_{σ} (v)$ corresponds to a wall in $Stab^{†}$ . The result extended earlier work of Kawatani, Yoshioka, and Toda, and is the source of explicit computations of nef cones and movable cones for hyperkähler manifolds of K3-type.

Theorem (Douglas $Π$ -stability for D-branes; Douglas 2001). Let $X$ be a Calabi-Yau threefold with a stringy Kähler structure parametrised by a point $τ$ in the stringy Kähler moduli space $M_{K} (X)$ . The set of physical B-branes — boundary conditions in the topologically twisted B-model — at $τ$ is identified with the set of $Π_{τ}$ -stable objects of $D^{b} (X)$ , where $Π_{τ}$ is a central charge built from the period integrals of the holomorphic volume form.

The Douglas physics framework is the historical precursor of Bridgeland stability. Douglas observed that the BPS state condition for D-branes on a Calabi-Yau threefold $X$ refines slope stability to a condition on derived-category objects, with the central charge $Z = \int_{γ} Ω$ given by the period integral of the holomorphic $(3, 0)$ -form along the cycle $γ$ supporting the brane. Bridgeland 2007 axiomatised this framework, removing the physics input and proving the deformation theorem in purely algebraic terms. The Douglas-Bridgeland dictionary: stringy Kähler moduli space $\leftrightarrow$ Bridgeland manifold $Stab (D^{b} (X)) / Aut$ ; D-brane charge lattice $\leftrightarrow$ numerical Grothendieck group $K^{num} (D^{b} (X))$ ; period integrals of $Ω$ $\leftrightarrow$ central charge $Z$ .

Theorem (Kontsevich-Soibelman motivic wall-crossing; Kontsevich-Soibelman 2008 §6). Let $D$ be a Calabi-Yau-3 triangulated category and let $σ_{t}$ be a continuous path in $Stab (D)$ crossing a single wall $W$ at $t = t_{0}$ . The motivic Donaldson-Thomas generating series $DT_{σ_{t}}^{mot} (v)$ satisfies the wall-crossing identity $$ \prod_{\phi \in \text{factorisation}} \exp!\left( \frac{[M^{\mathrm{mot}}{\sigma-}(v_\phi)]}{(1 - \mathbb{L})} \right) = \prod_{\phi \in \text{factorisation}} \exp!\left( \frac{[M^{\mathrm{mot}}{\sigma+}(v_\phi)]}{(1 - \mathbb{L})} \right) $$ in the motivic Hall algebra of $D$ , where the products are taken in opposite phase-orderings on each side of the wall.

The Kontsevich-Soibelman wall-crossing formula promotes the Bridgeland wall-crossing picture to a statement about motivic invariants of moduli of semistable objects. The motivic Hall algebra is a categorification of the constructible-function Hall algebra of Joyce, and the wall-crossing identity is an equality of two products of exponentials in this algebra. In numerical specialisations, the formula recovers the wall-crossing identities for ordinary DT invariants, for refined DT invariants, and for PT (Pandharipande-Thomas) stable pair invariants — the entire enumerative-geometry framework on Calabi-Yau threefolds.

Theorem (Bridgeland stability on threefolds; Bayer-Macri-Toda 2014, Bayer-Macri-Stellari 2016). For $X$ a smooth projective threefold and $β + iω$ a complexified ample class, the candidate central charge $Z_{β, ω} (E) = - \int_{X} e^{- β - iω} ch (E)$ together with a tilted heart $Coh^{β} (X)$ defines a Bridgeland stability condition on $D^{b} (X)$ , conjecturally for every smooth projective threefold, with the conjecture verified for $P^{3}$ , the smooth quadric threefold, abelian threefolds, and various Fano threefolds.

The threefold case is open in general: a "Bogomolov-Gieseker-type inequality" for tilt-stable objects on $X$ is the technical obstruction, conjecturally true for all smooth projective threefolds (the Bayer-Macri-Toda conjecture). The conjecture has been verified for principally polarised abelian threefolds (Maciocia-Piyaratne 2015), for Fano threefolds of Picard rank $1$ (Li 2019, Liu 2021), and partially for other classes. The Bridgeland threefold programme is one of the active frontiers of derived-category geometry, and a positive resolution would import the entire Bayer-Macri MMP machinery to threefolds — including a derived-category-based proof of the rationality conjectures for Fano threefolds.

Synthesis. The Bridgeland framework promotes Mumford's 1963 slope-stability picture from the abelian world of coherent sheaves to the triangulated world of derived categories, and the central insight is that the resulting space of stability conditions is itself a complex manifold of the same dimension as the Grothendieck group, with the forgetful map to the central-charge space the foundational reason chamber decompositions in moduli theory are rational polyhedral. The bridge is the deformation theorem: a small perturbation of the central charge transports every semistable object to a nearby semistable object, identifying $Stab (D)$ with $Hom (K (D), C)$ locally. Three apparently distinct constructions — the slicing of a triangulated category, the heart of a bounded $t$ -structure, and the central charge on the Grothendieck group — fit into one identity through the equivalence of the two definitions, putting these together into a single complex manifold.

Mirror symmetry sits at the centre of the picture. Douglas's $Π$ -stability for D-branes on a Calabi-Yau threefold is exactly Bridgeland stability with central charge given by the period integral of the holomorphic volume form, and the Bridgeland manifold of the derived category of $X$ generalises the stringy Kähler moduli space of $X$ . The Kontsevich homological mirror symmetry conjecture identifies $D^{b} (X)$ for $X$ a Calabi-Yau with the derived Fukaya category of the mirror $\overset{ˇ}{X}$ , and this identifies $Stab (D^{b} (X))$ with the symplectic-side moduli of stability conditions on the Fukaya category — a picture not yet fully understood but actively investigated by Bridgeland, Smith, Polishchuk, and others. Putting these together identifies the algebraic-geometry Bridgeland manifold with the symplectic-geometry moduli of A-brane stability, the bridge being the period integral and the Mukai pairing on the Mukai lattice.

The Bayer-Macri 2014 correspondence on K3 surfaces makes the wall-crossing geometry of $Stab (D^{b} (X))$ exactly the minimal model program for moduli of sheaves: every birational model of a moduli space is realised as a moduli space of $σ$ -stable objects for some other $σ$ , and the chamber decomposition of $Stab$ projects onto the Mori chamber decomposition. This appears again in 04.10.09 (Variation of GIT), where the polarisation chamber decomposition of VGIT plays the analogous role for GIT quotients, and the two pictures coincide whenever both apply — the central insight that wall-crossing in stability conditions and wall-crossing in polarisations are two sides of the same coin, mediated by the moduli-space construction. Putting these together with the Kontsevich-Soibelman wall-crossing formula identifies the enumerative wall-crossing with the geometric wall-crossing, and the entire Donaldson-Thomas theory of Calabi-Yau threefolds becomes a wall-crossing statement in $Stab (D^{b} (X))$ .

The synthesis is structural: every classical stability notion in algebraic geometry — slope stability on a curve, slope stability on a surface, Gieseker stability, Donaldson-Uhlenbeck-Yau stability, K-stability — corresponds to a point or a chamber in some Bridgeland manifold, with the appropriate choice of triangulated category. Bridgeland stability is the universal organising principle, and the framework is exactly what makes the moduli-theoretic, enumerative-geometric, and mirror-symmetric pictures coincide.

Full proof set Master

Proposition (orthogonality of the slicing). Let $σ = (Z, P) \in Stab (D)$ and let $A \in P (ϕ_{1})$ , $B \in P (ϕ_{2})$ with $ϕ_{1} > ϕ_{2}$ . Then $Hom (A, B) = 0$ .

Proof. This is axiom (S2) of the slicing definition; the proposition records its derivation from the compatibility $Z (E) \in R_{> 0} e^{iπ ϕ}$ when $A, B$ are stable. Suppose $f : A \to B$ is a non-zero morphism. Embed it in a distinguished triangle $A f B \to C \to A [1]$ . The cone $C$ has class $[C] = [B] - [A]$ in $K (D)$ , hence $Z (C) = Z (B) - Z (A) = ∣ Z (B) ∣ e^{iπ ϕ_{2}} - ∣ Z (A) ∣ e^{iπ ϕ_{1}}$ . Since $ϕ_{1} > ϕ_{2}$ , the two vectors $e^{iπ ϕ_{1}}$ and $e^{iπ ϕ_{2}}$ point in different directions in $C$ , and the difference $Z (B) - Z (A)$ lies in the open half-plane ${z : ar g z \in (ϕ_{2}, ϕ_{1} + 1)}$ (a strict open angle).

The Harder-Narasimhan filtration of $C$ produces semistable factors $C_{1}, \dots, C_{m}$ with $Z (C) = \sum Z (C_{i})$ and phases $ϕ^{+} \geq ϕ_{i} \geq ϕ^{-}$ . If $f$ is non-zero, the long exact sequence in cohomology of any $t$ -structure refining $σ$ forces $C$ to have a factor of phase $\geq ϕ_{1}$ (the cone-fits-in-the-extension argument). But the phase of every factor of $C$ is at most the maximum of $ϕ_{1}$ and $ϕ_{2} + 1$ , and the equality case forces $A$ and $B$ to be related by a non-zero extension of an object of phase $ϕ_{1}$ by an object of phase $ϕ_{2}$ , which contradicts $ϕ_{1} > ϕ_{2}$ once one observes that the only such extension is split (the splitting follows from $Ext^{1} (A_{ϕ_{1}}, B_{ϕ_{2}}) = 0$ via the long exact sequence of the slicing).

Hence $f$ must vanish, and $Hom (A, B) = 0$ . $□$

Proposition (uniqueness of Harder-Narasimhan filtration). Let $σ = (Z, P) \in Stab (D)$ and let $0 \neq = E \in D$ . The filtration in axiom (S3) is unique up to isomorphism of triangles.

Proof. Suppose two filtrations $$ 0 = E_0 \to E_1 \to \cdots \to E_n = E, \quad 0 = E_0' \to E_1' \to \cdots \to E_m' = E $$ satisfy (S3) with semistable cones $F_{i} = cone (E_{i - 1} \to E_{i}) \in P (ϕ_{i})$ and $F_{j}^{'} \in P (ϕ_{j}^{'})$ , with $ϕ_{1} > \dots > ϕ_{n}$ and $ϕ_{1}^{'} > \dots > ϕ_{m}^{'}$ .

Apply orthogonality (Proposition above) to the maximum phase. The first non-zero map in either filtration goes from $F_{1}$ (phase $ϕ_{1}$ ) to $E$ , and the composition $F_{1} \to E_{n} = E \to E / E_{m - 1}^{'} = F_{m}^{'}$ is non-zero iff $ϕ_{1} = ϕ_{m}^{'}$ (by orthogonality applied with reversed phase ordering on the cokernel side). The argument is: any non-zero map from a semistable object of phase $ϕ_{1}$ to a complex with all Harder-Narasimhan factors of phase strictly less than $ϕ_{1}$ must vanish, so the composition into $F_{m}^{'}$ vanishes unless $ϕ_{1}^{'} = ϕ_{1}$ . By symmetry $ϕ_{n}^{'} = ϕ_{n}$ and $m = n$ .

Inductively, $F_{i} ≅ F_{i}^{'}$ and the filtrations are isomorphic. $□$

Proposition (Bridgeland manifold dimension). Let $D$ be a triangulated category with $K (D)$ free abelian of rank $n$ . The complex manifold $Stab (D)$ has complex dimension $n$ at every point.

Proof. By the deformation theorem (proved above), the forgetful map $Z : Stab (D) \to Hom (K (D), C) ≅ C^{n}$ is a local homeomorphism. Each point $σ \in Stab (D)$ has an open neighbourhood $U$ on which $Z ∣_{U} : U \to Z (U) \subset C^{n}$ is a homeomorphism. The pullback of the complex-vector-space structure on $C^{n}$ via $Z ∣_{U}$ defines a complex chart on $U$ , with $Stab (D)$ acquiring a complex manifold structure of dimension $n$ at $σ$ . Since the local-homeomorphism property holds at every $σ$ , the complex dimension is constant equal to $n$ . $□$

Proposition (stability on a curve; Bridgeland 2007 Theorem 9.1, Macri 2007 Math. Z. 256). Let $C$ be a smooth projective curve of genus $g \geq 1$ . The stability manifold satisfies $$ \mathrm{Stab}(D^b(C)) \cong \mathbb{C}^2 $$ as a complex manifold, and the forgetful map identifies $Stab (D^{b} (C))$ with the open subset of $Hom (K (D^{b} (C)), C) ≅ C^{2}$ where the central charge sends every effective class to the open upper half-plane.

Proof. The Grothendieck group $K (D^{b} (C))$ is rank $2$ , generated by $[O_{C}]$ and $[O_{p}]$ for $O_{p}$ a skyscraper at a point. A central charge $Z$ is determined by the two complex numbers $z_{1} = Z (O_{C})$ and $z_{2} = Z (O_{p})$ . The compatibility axiom forces $z_{2} \neq = 0$ (skyscrapers are semistable with definite phase) and $z_{1}$ to lie in the upper half-plane (line bundles are semistable with positive imaginary part of central charge).

For any pair $(z_{1}, z_{2})$ with $z_{2} \neq = 0$ and $z_{1}$ in the upper half-plane (after a global rotation aligning $z_{2}$ with the negative real axis), construct the slicing: skyscrapers are placed at phase $ϕ_{2} = ar g (z_{2}) / π$ , line bundles at phase $ϕ_{1} = ar g (z_{1}) / π$ , and higher-rank slope-semistable bundles at phases interpolating between these via the slope formula $ϕ = (1/ π) arctan (r / (- d))$ .

The slicing satisfies (S1) (shift sends phase $ϕ$ to $ϕ + 1$ ), (S2) (orthogonality of slope-semistable bundles of different slopes on a curve, by Schur's lemma), and (S3) (existence of Harder-Narasimhan filtration for slope-semistable sheaves on a curve, by Harder-Narasimhan 1975).

Conversely, every stability condition on $D^{b} (C)$ arises in this way: Macri 2007 Math. Z. 256 proved that the heart of the bounded $t$ -structure must be either $Coh C$ or a $Z$ -shift of it (by classification of bounded $t$ -structures on $D^{b} (C)$ for $g \geq 1$ ), and the stability function is determined by its values on the two generators.

Hence $Stab (D^{b} (C))$ is identified with the open subset of $C^{2}$ described above, which is homeomorphic to $C^{2}$ itself (as an open subset of $C^{2}$ homeomorphic to a product of $C$ with the upper half-plane $H ≅ C$ ). $□$

Proposition (Bridgeland stability and the group action). The group $GL_{2}^{+} (R)$ — orientation-preserving real $2 \times 2$ matrices with positive determinant — acts on $Stab (D)$ by post-composing the central charge with the real-linear map and adjusting the slicing accordingly. The quotient is the natural moduli space of "stability conditions modulo reparametrisation".

Proof. The action is defined as follows: for $g \in GL_{2}^{+} (R)$ acting on $C ≅ R^{2}$ in the standard way, and $σ = (Z, P) \in Stab (D)$ , define $g \cdot σ = (g \circ Z, P_{g})$ , where $P_{g} (ϕ) = P (ψ^{- 1} (ϕ))$ for $ψ (ϕ)$ the unique element of $R$ with $g (e^{iπ ϕ}) /∣ g (e^{iπ ϕ}) ∣ = e^{iπ ψ (ϕ)}$ chosen continuously with $ψ (0) =$ the analogous lift. The composition $g \circ Z$ is again a group homomorphism $K (D) \to C$ , the slicing $P_{g}$ satisfies (S1)-(S3) by direct verification (the reparametrisation preserves the orthogonality and HN structure), and the compatibility $g \circ Z (E) \in R_{> 0} \cdot e^{iπ ψ (ϕ)}$ for $E \in P_{g} (ψ (ϕ))$ follows from the definition.

The action is free on the locus of stability conditions whose Grothendieck group has rank $\geq 1$ , and the quotient $Stab (D) / GL_{2}^{+} (R)$ is the natural moduli space. The shift $[1]$ corresponds to $g = - I \in GL_{2} (R)$ (which has determinant $+ 1$ in the orientation-preserving sense after lifting to the universal cover), and rescaling $Z$ by a positive real corresponds to the centre $R_{> 0} \cdot I \subset GL_{2}^{+} (R)$ . $□$

Connections Master

Slope stability on a curve 04.10.06. The Mumford 1963 slope-stability framework for vector bundles on a smooth projective curve is the foundational case of Bridgeland stability: take the central charge $Z (E) = - de g (E) + i \cdot rk (E)$ and the slicing whose phase- $ϕ$ subcategory is the additive subcategory of slope-semistable bundles of slope $μ = cot (π ϕ)$ . The deformation theorem then realises slope stability as one point in the two-dimensional complex manifold $Stab (D^{b} (C)) ≅ C^{2}$ . The bridge is exactly the equivalence: Bridgeland stability is the categorical promotion of Mumford's framework, and the curve case is the cleanest example.
Hilbert scheme 04.10.05. The Hilbert scheme $Hilb^{n} (X)$ of $n$ points on a smooth projective surface $X$ admits a wall-crossing description via Bridgeland stability: for $X$ a K3 surface, every birational model of $Hilb^{n} (X)$ is a moduli of $σ$ -stable objects in $D^{b} (X)$ of fixed Mukai vector $v = (1, 0, 1 - n)$ , by Bayer-Macri 2014. The chamber decomposition of $Stab^{†} (D^{b} (X))$ projects onto the Mori chamber decomposition of the nef cone of $Hilb^{n} (X)$ . The bridge is wall-crossing, and the central insight is that the entire birational geometry of $Hilb^{n}$ is encoded in the Bridgeland manifold.
Variation of GIT 04.10.09. VGIT moves through the ample cone of polarisations and decomposes it into rational polyhedral chambers, with crossing a wall realising a flip or divisorial contraction between GIT quotients. Bridgeland stability moves through the central-charge space and decomposes it into chambers, with crossing a wall realising a birational map between moduli of stable objects. The two pictures are dual: GIT chambers parametrise polarisations, Bridgeland chambers parametrise central charges, and the moduli-theoretic outputs coincide whenever both pictures apply. The Halpern-Leistner magic windows of derived VGIT identify the two chamber decompositions with the chamber decomposition of the derived category itself.
Derived functors and Ext 04.03.06. The Hom and Ext groups in the derived category $D^{b} (X)$ are the foundational data on which Bridgeland stability is built: the orthogonality axiom (S2) is a statement about $Hom_{D^{b} (X)} (P (ϕ_{1}), P (ϕ_{2})) = 0$ for $ϕ_{1} > ϕ_{2}$ , and the Harder-Narasimhan filtration is constructed via distinguished triangles. Without the derived-category machinery, the slicing has no setting; with it, Bridgeland stability is the natural categorical promotion of Mumford stability.
Coherent sheaf 04.06.02. The heart of the standard $t$ -structure on $D^{b} (X)$ is the abelian category $Coh X$ of coherent sheaves, and the standard Bridgeland stability conditions on $D^{b} (X)$ have heart equal to $Coh X$ or a torsion-pair-tilt of it (Happel-Reiten-Smalø). The bridge from sheaves to complexes is exactly the heart-stability-function equivalent definition: a stability condition on $D^{b} (X)$ is the same as a stability function on a heart of a bounded $t$ -structure.
Serre duality 04.08.03. On a Calabi-Yau variety (a smooth projective variety whose canonical class vanishes), Serre duality provides a non-degenerate symmetric pairing on the Grothendieck group $K (D)$ via the Euler form $χ (E, F) = \sum (- 1)^{i} dim Ext^{i} (E, F)$ . The Mukai pairing on K3 surfaces and the Euler pairing on Calabi-Yau threefolds are the structural data that pin down the Bridgeland manifold to a period domain; Serre duality is the source of the symmetry.
Moduli of vector bundles 04.10.06. The moduli space $M_{C} (r, d)$ of slope-stable vector bundles on a curve of genus $g$ identifies with the moduli space $M_{σ} (r, d)$ of $σ$ -stable objects in $D^{b} (C)$ for the standard slope central charge $σ$ . Bridgeland stability is the categorical organising principle for the entire $M_{C}$ -theory, and the deformation theorem says the moduli is constant in central-charge chambers and jumps at walls — recovering the classical Harder-Narasimhan stratification by slope.

Historical & philosophical context Master

Bridgeland stability conditions originated in the physics-mathematics interface of the late 1990s. Michael Douglas, in D-branes, categories and N=1 supersymmetry (J. Math. Phys. 42, 2001 ^{[source pending]}) and subsequent work, proposed the notion of $Π$ -stability for B-branes on a Calabi-Yau threefold $X$ : BPS-saturated D-branes correspond to $Π$ -stable objects of $D^{b} (X)$ , where the central charge is the period integral $Z (γ) = \int_{γ} Ω$ of the holomorphic volume form $Ω$ along the cycle $γ$ supporting the brane. The physics formulation was incomplete — the categorical structure of " $Π$ -stability" remained heuristic — but the framework was sufficiently suggestive that Tom Bridgeland abstracted it into a precise mathematical definition in Stability conditions on triangulated categories (Ann. of Math. 166, 2007 ^{[source pending]}).

Bridgeland's 2007 paper made two foundational contributions. First, the axiomatic definition: a stability condition is a pair $(Z, P)$ with $Z$ a central charge and $P$ a slicing, subject to the compatibility $Z (E) \in R_{> 0} \cdot e^{iπ ϕ}$ for $E \in P (ϕ)$ . Second, the deformation theorem: $Stab (D)$ is a complex manifold, and the forgetful map to the central-charge space is a local homeomorphism. The dual definition via $t$ -structures and stability functions was a clarifying reformulation due to Bridgeland himself, building on the Happel-Reiten-Smalø theory of tilting and the bounded- $t$ -structure picture going back to Beilinson-Bernstein-Deligne 1982.

Bridgeland 2008 Duke Math. J. 141 ^{[source pending]} constructed an explicit connected component of $Stab (D^{b} (X))$ for $X$ a K3 surface, identifying it with the period domain of generalised Calabi-Yau structures on the Mukai lattice. This was the first instance of the wall-and-chamber picture being made fully precise. The Mumford 1965 Geometric Invariant Theory ^{[source pending]} framework — where GIT semistability is the foundational notion for moduli of polarised varieties — finds its categorical promotion in Bridgeland's manifold, with the Hilbert-Mumford numerical criterion of GIT replaced by the central charge and the polarisation-chamber decomposition of VGIT replaced by the Bridgeland chamber decomposition.

Maxim Kontsevich and Yan Soibelman, in arXiv:0811.2435 (2008) ^{[source pending]}, extended Bridgeland's framework to enumerative geometry: the motivic Donaldson-Thomas invariants of a Calabi-Yau-3 triangulated category satisfy a wall-crossing identity in the motivic Hall algebra. The Joyce-Song generalised DT invariants (Memoirs AMS 217, 2012) gave a parallel framework using Behrend functions. Both frameworks lift the wall-crossing picture from geometric moduli spaces to virtual or motivic counts, and provide the foundation for the modern enumerative geometry of Calabi-Yau threefolds.

Arend Bayer and Emanuele Macri, in Inventiones Math. 198 (2014) ^{[source pending]}, proved the cleanest known instance of "wall-crossing equals MMP": on a smooth complex projective K3 surface, every birational model of a moduli space of stable sheaves arises as a moduli of $σ$ -stable objects for some $σ$ in the Bridgeland manifold, and the chamber decomposition of $Stab^{†}$ projects onto the Mori chamber decomposition of the nef cone. The Bayer-Macri correspondence is the source of explicit computations of nef cones, movable cones, and birational automorphism groups of hyperkähler manifolds of K3 type, and is one of the most consequential applications of the Bridgeland framework.

The contemporary frontier is the threefold case: a "Bogomolov-Gieseker-type inequality" for tilt-stable objects on $X$ smooth projective threefold is the technical input needed to construct a Bridgeland stability condition on $D^{b} (X)$ . The conjecture (Bayer-Macri-Toda 2014) has been verified for $P^{3}$ , the smooth quadric threefold, abelian threefolds, and certain Fano threefolds, but remains open in general. A positive resolution would import the entire Bayer-Macri MMP machinery to threefolds, with implications for the rationality conjectures, the Kontsevich homological mirror symmetry programme, and the enumerative geometry of Calabi-Yau threefolds via the Pandharipande-Thomas / DT correspondence.

Bibliography Master

@article{Bridgeland2007,
  author  = {Bridgeland, Tom},
  title   = {Stability conditions on triangulated categories},
  journal = {Annals of Mathematics (2)},
  volume  = {166},
  number  = {2},
  year    = {2007},
  pages   = {317--345}
}

@article{Bridgeland2008K3,
  author  = {Bridgeland, Tom},
  title   = {Stability conditions on {K3} surfaces},
  journal = {Duke Mathematical Journal},
  volume  = {141},
  number  = {2},
  year    = {2008},
  pages   = {241--291}
}

@article{Douglas2001,
  author  = {Douglas, Michael R.},
  title   = {{D-branes, categories and N=1 supersymmetry}},
  journal = {Journal of Mathematical Physics},
  volume  = {42},
  number  = {7},
  year    = {2001},
  pages   = {2818--2843},
  note    = {arXiv:hep-th/0011017 and arXiv:hep-th/0107080}
}

@article{KontsevichSoibelman2008,
  author  = {Kontsevich, Maxim and Soibelman, Yan},
  title   = {Stability structures, motivic {Donaldson-Thomas} invariants and cluster transformations},
  journal = {arXiv preprint},
  volume  = {arXiv:0811.2435},
  year    = {2008}
}

@article{BayerMacri2014,
  author  = {Bayer, Arend and Macri, Emanuele},
  title   = {{MMP} for moduli of sheaves on {K3s} via wall-crossing: nef and movable cones, {Lagrangian} fibrations},
  journal = {Inventiones Mathematicae},
  volume  = {198},
  number  = {3},
  year    = {2014},
  pages   = {505--590}
}

@incollection{MacriSchmidt2017,
  author    = {Macri, Emanuele and Schmidt, Benjamin},
  title     = {Lectures on {Bridgeland} stability},
  booktitle = {Moduli of Curves ({C}etraro 2017)},
  series    = {Lecture Notes of the Unione Matematica Italiana},
  volume    = {21},
  publisher = {Springer},
  year      = {2017},
  pages     = {139--211}
}

@incollection{Huybrechts2014Stability,
  author    = {Huybrechts, Daniel},
  title     = {Introduction to stability conditions},
  booktitle = {Moduli Spaces},
  series    = {London Mathematical Society Lecture Note Series},
  volume    = {411},
  publisher = {Cambridge University Press},
  year      = {2014},
  pages     = {179--229}
}

@book{MumfordGIT,
  author    = {Mumford, David and Fogarty, John and Kirwan, Frances},
  title     = {Geometric Invariant Theory},
  publisher = {Springer-Verlag},
  series    = {Ergebnisse der Mathematik und ihrer Grenzgebiete},
  volume    = {34},
  edition   = {3},
  year      = {1994}
}

@article{Toda2012,
  author  = {Toda, Yukinobu},
  title   = {Stability conditions and curve counting invariants on {Calabi-Yau} 3-folds},
  journal = {Kyoto Journal of Mathematics},
  volume  = {52},
  number  = {1},
  year    = {2012},
  pages   = {1--50}
}

@article{JoyceSong2012,
  author  = {Joyce, Dominic and Song, Yinan},
  title   = {A theory of generalized {Donaldson-Thomas} invariants},
  journal = {Memoirs of the American Mathematical Society},
  volume  = {217},
  number  = {1020},
  year    = {2012}
}

@incollection{Bridgeland2009Spaces,
  author    = {Bridgeland, Tom},
  title     = {Spaces of stability conditions},
  booktitle = {Algebraic Geometry --- {S}eattle 2005},
  series    = {Proceedings of Symposia in Pure Mathematics},
  volume    = {80, Part 1},
  publisher = {American Mathematical Society},
  year      = {2009},
  pages     = {1--21}
}

@article{Macri2007,
  author  = {Macri, Emanuele},
  title   = {Stability conditions on curves},
  journal = {Mathematische Zeitschrift},
  volume  = {256},
  year    = {2007},
  pages   = {657--669}
}

@article{BayerMacriToda2014,
  author  = {Bayer, Arend and Macri, Emanuele and Toda, Yukinobu},
  title   = {Bridgeland stability conditions on threefolds {I}: Bogomolov-Gieseker type inequalities},
  journal = {Journal of Algebraic Geometry},
  volume  = {23},
  number  = {1},
  year    = {2014},
  pages   = {117--163}
}

Prerequisites

04.10.06
04.10.05
04.06.02
04.03.06
04.08.03

Tier anchors

beginner: Macri-Schmidt *Lectures on Bridgeland Stability* (Birkhauser MOOC notes 2017) §1 informal — slope of a vector bundle, made categorical
intermediate: Macri-Schmidt *Lectures on Bridgeland Stability* §2-4; Huybrechts *Introduction to Stability Conditions* (London Math. Soc. Lect. Note 411, 2014); Bayer 2018 *Bridgeland stability conditions on threefolds* (lecture notes)
master: Bridgeland 2007 *Annals of Mathematics* (2) 166 (2), 317-345 *Stability conditions on triangulated categories* (originator paper); Douglas 2001 hep-th/0107080 *D-branes, categories and N=1 supersymmetry* (the physics origin via $\Pi$-stability for D-branes on Calabi-Yau threefolds); Kontsevich-Soibelman 2008 arXiv:0811.2435 *Stability structures, motivic Donaldson-Thomas invariants and cluster transformations*; Bayer-Macri 2014 *Inventiones Math.* 198 (3), 505-590 *MMP for moduli of sheaves on K3s via wall-crossing*; Mumford 1965 *Geometric Invariant Theory* (the GIT slope-stability framework that Bridgeland stability promotes to the derived setting); Bridgeland 2008 *Duke Mathematical Journal* 141 (2), 241-291 *Stability conditions on K3 surfaces*

References

Bridgeland, T. — Stability conditions on triangulated categories · *Annals of Mathematics* (2) 166 (2), 317-345 (2007). The originator paper introducing stability conditions on triangulated categories, proving the deformation theorem that $\mathrm{Stab}(\mathcal{D})$ is a complex manifold with the central-charge map a local homeomorphism onto $\mathrm{Hom}(K(\mathcal{D}), \mathbb{C})$.
Bridgeland, T. — Stability conditions on K3 surfaces · *Duke Mathematical Journal* 141 (2), 241-291 (2008). Construction of an explicit connected component $\mathrm{Stab}^\dagger(D^b(X))$ on a K3 surface $X$, with the Mukai pairing serving as the central charge and the period-domain picture for the action of autoequivalences.
Douglas, M. R. — D-branes, categories and N=1 supersymmetry · *Journal of Mathematical Physics* 42 (7), 2818-2843 (2001); arXiv:hep-th/0011017 and arXiv:hep-th/0107080. The physics origin of Bridgeland stability via $\Pi$-stability for B-branes on Calabi-Yau threefolds; central charge $Z = \int_\gamma \Omega$ as the period integral of the holomorphic volume form.
Kontsevich, M. and Soibelman, Y. — Stability structures, motivic Donaldson-Thomas invariants and cluster transformations · arXiv:0811.2435 (2008). Motivic wall-crossing formula for Donaldson-Thomas invariants attached to a Bridgeland stability condition on a Calabi-Yau-3 triangulated category; the wall-crossing identity in the motivic Hall algebra is the central technical result.
Bayer, A. and Macri, E. — MMP for moduli of sheaves on K3s via wall-crossing: nef and movable cones, Lagrangian fibrations · *Inventiones Mathematicae* 198 (3), 505-590 (2014). Wall-crossing in $\mathrm{Stab}(D^b(K3))$ identifies the birational geometry of moduli of sheaves with the chamber decomposition of the Bridgeland manifold; every birational model of a moduli of stable sheaves on a K3 arises as a moduli of $\sigma$-stable objects for some $\sigma$ in a wall-crossing chamber.
Macri, E. and Schmidt, B. — Lectures on Bridgeland stability · in *Moduli of Curves* (Cetraro 2017), Lecture Notes of the Unione Matematica Italiana 21, Springer (2017), 139-211. Modern textbook account covering definitions, slicings, the deformation theorem, the curve and surface cases, wall-crossing, and applications to K3 and threefold geometry.
Huybrechts, D. — Introduction to stability conditions · in *Moduli Spaces*, London Mathematical Society Lecture Note Series 411, Cambridge University Press (2014), 179-229. Survey emphasising the K3 case and the Mukai-lattice period domain.
Mumford, D. — Geometric Invariant Theory · Ergebnisse der Mathematik 34, Springer-Verlag (1965; 3rd enlarged edition with Fogarty and Kirwan, 1994). The GIT slope-stability framework that Bridgeland stability promotes to the derived setting; §2.1 numerical criterion, §4-5 moduli applications.
Toda, Y. — Stability conditions and curve counting invariants on Calabi-Yau 3-folds · *Kyoto Journal of Mathematics* 52 (1), 1-50 (2012). Application of Bridgeland stability conditions to curve-counting (PT/DT) wall-crossing on Calabi-Yau threefolds; the wall-crossing identity for stable pairs.
Joyce, D. and Song, Y. — A theory of generalized Donaldson-Thomas invariants · Memoirs of the American Mathematical Society 217 (1020), American Mathematical Society (2012). Extension of Donaldson-Thomas theory to semistable objects via Behrend functions; companion to Kontsevich-Soibelman.
Bridgeland, T. — Spaces of stability conditions · in *Algebraic Geometry — Seattle 2005*, Proceedings of Symposia in Pure Mathematics 80 Part 1, American Mathematical Society (2009), 1-21. Survey article covering the curve, K3, and Calabi-Yau-3 cases with explicit chamber pictures.

Estimated time

beginner: 15m
intermediate: 50m
master: 90m