Introduction
In the companion paper [10 ] , the authors have constructed knotted versions of the Oakley–Usher Lagrangians in [24 ] . The goal of this paper is to have an in-depth look at the situation in the three-dimensional quadric Q 3 Q^3 Q 3 .
Setup
We equip Q 3 = { z 0 2 + ⋯ + z 4 2 = 0 } ⊂ C P 4 Q^3 = \{z_0^2 + \cdots+ z_4^2 = 0\} \subset\mathbb{CP}^4 Q 3 = { z 0 2 + ⋯ + z 4 2 = 0 } ⊂ CP 4 with the Fubini–Study form normalized so that a line has area 2 π 2\pi 2 π . By [10 Proposition 3.10] , this space can be viewed as the symplectic cut of a the cotangent bundle T ∗ S 3 T^*S^3 T ∗ S 3 . This identifies D ∗ S 3 D^*S^3 D ∗ S 3 with the complement of a copy Q 2 Q^2 Q 2 of the two-dimensional quadric. We write Σ ⊂ Q 3 \Sigma\subset Q^3 Σ ⊂ Q 3 for the image of the zero-section under the symplectic cut; in this model Σ = { [ 1 : i q ] : q ∈ S 3 } \Sigma= \{[1 : iq] : q \in S^3\} Σ = {[ 1 : i q ] : q ∈ S 3 } .
The families P 1 , 1 ( r ) \mathcal{P}_{1,1}(r) P 1 , 1 ( r ) and P 1 , 1 C h ( a ) \mathcal{P}_{1,1}^{\mathrm{Ch}}(a) P 1 , 1 Ch ( a ) arise by lifting curves under the T 2 = S O ( 2 ) × S O ( 2 ) T^2 = \mathrm{SO}(2) \times\mathrm{SO}(2) T 2 = SO ( 2 ) × SO ( 2 ) symplectic reduction described in § 3 in [10 ] . The ordinary Oakley–Usher Lagrangians P 1 , 1 ( r ) \mathcal{P}_{1,1}(r) P 1 , 1 ( r ) are lifts of the noncontractible circles, while their Chekanov counterparts P 1 , 1 C h ( a ) \mathcal{P}_{1,1}^{\mathrm{Ch}}(a) P 1 , 1 Ch ( a ) are lifts of contractible circles bounding area a / 2 a/2 a /2 in the smooth reduced cylinder. We use the size convention of [10 ] : both families have their monotone member at r = a = 4 π / 3 r = a = 4\pi/3 r = a = 4 π /3 . Throughout this paper we abbreviate
P = P 1 , 1 ( 4 π / 3 ) , P C h = P 1 , 1 C h ( 4 π / 3 ) . \mathcal{P} = \mathcal{P}_{1,1}(4\pi/3), \qquad\mathcal{P}^{\mathrm{Ch}} = \mathcal{P}_{1,1}^{\mathrm{Ch}}(4\pi/3). P = P 1 , 1 ( 4 π /3 ) , P Ch = P 1 , 1 Ch ( 4 π /3 ) .
Both are Lagrangian three-tori. They are Lagrangian isotopic by [10, Theorem B`], but not Hamiltonian isotopic, as we shall see now.
Main results
As discussed in [24 Theorem 1.3] , the torus P \mathcal{P} P can be viewed as the Biran circle bundle lift of the Clifford torus in Q 2 ≅ S 2 × S 2 Q^2 \cong S^2 \times S^2 Q 2 ≅ S 2 × S 2 [24 Theorem 1.3] . Similarly, one can lift all Vianna tori from Q 2 Q^2 Q 2 [33 ] . Our first result shows that the Chekanov OU-torus is different from tori obtained in this way.
Theorem A. The torus P C h \mathcal{P}^{\mathrm{Ch}} P Ch cannot be mapped by a symplectomorphism to P \mathcal{P} P or the lift of any other Vianna torus from the divisor Q 2 ⊂ Q 3 Q^2 \subset Q^3 Q 2 ⊂ Q 3 .
The proof computes the versal deformations of the Ψ \Psi Ψ -invariant of Shelukhin–Tonkonog–Vianna [29 ] . To that end, we study a Gelfand–Cetlin fibration which has P C h \mathcal{P}^{\mathrm{Ch}} P Ch as its monotone fibre. Related lifting constructions are considered in earlier work by Chanda–Hirschi–Wang [12 ] and by Brendel [8 ] . This approach will be significantly generalized in upcoming work by Diogo–Tonkonog–Vianna–Wu [14 ] .
Our second result concerns P \mathcal{P} P . It computes the signed Maslov two disk-potential and identifies a finite set of generators for the monotone Fukaya category.
Theorem B. There is a spin structure on the Oakley–Usher torus P ⊂ Q 3 \mathcal{P} \subset Q^3 P ⊂ Q 3 , and a basis of π 1 ( P ) ≅ Z 3 \pi_1(\mathcal{P}) \cong\mathbb{Z}^3 π 1 ( P ) ≅ Z 3 for which the superpotential W P : Hom ( π 1 ( P ) , Λ × ) ≅ ( Λ × ) 3 → Λ \mathcal{W}_{\mathcal{P}} : \operatorname{Hom}(\pi_1(\mathcal{P}),\Lambda^\times) \cong(\Lambda^\times)^3 \to\Lambda W P : Hom ( π 1 ( P ) , Λ × ) ≅ ( Λ × ) 3 → Λ is
W P ( x 1 , x 2 , x 3 ) = T 2 π / 3 ( x 1 + x 2 + x 3 x 1 + x 3 x 2 + 1 x 3 ) . \mathcal{W}_{\mathcal{P}}(x_1,x_2,x_3) = T^{2\pi/3}\left(x_1 + x_2 + \frac{x_3}{x_1} + \frac{x_3}{x_2} + \frac{1}{x_3}\right). W P ( x 1 , x 2 , x 3 ) = T 2 π /3 ( x 1 + x 2 + x 1 x 3 + x 2 x 3 + x 3 1 ) . where Λ \Lambda Λ is the universal Novikov field over C \mathbb{C} C defined in (18). Furthermore, it carries three critical local systems, which, when taken together with a disjoint Lagrangian sphere Σ ⊂ Q 3 \Sigma\subset Q^3 Σ ⊂ Q 3 , split-generate the monotone Fukaya category of Q 3 Q^3 Q 3 .
In particular, the Oakley–Usher torus P \mathcal{P} P is non-displaceable, answering a question raised in [24 ] for this torus. The sphere Σ \Sigma Σ accounts for the zero-eigenvalue summand of quantum multiplication by c 1 ( Q 3 ) c_1(Q^3) c 1 ( Q 3 ) , while the three critical local systems on P \mathcal{P} P account for the three nonzero eigensummands. The proof of Theorem B relies on a toric degeneration of Q 3 Q^3 Q 3 and on the disk-counting results of Nishinou–Nohara–Ueda [22 , 23 ] .
Comparison to previous work
The split-generation result in Theorem B is inspired by work of Abouzaid–Diogo [1 ] , which treats the open case of T ∗ S 3 T^*S^3 T ∗ S 3 . Let us briefly compare the two results. Abouzaid–Diogo prove that the monotone Fukaya category is split-generated by the zero section equipped with an uncountable family of bounding cochains, together with either of the following two families:
a one-parameter family of copies of Lagrangian S 1 × S 2 S^1 \times S^2 S 1 × S 2 corresponding to the Oakley–Usher Lagrangians P 0 , 2 ( r ) ⊂ T ∗ S 3 \mathcal{P}_{0,2}(r) \subset T^*S^3 P 0 , 2 ( r ) ⊂ T ∗ S 3 , equipped with an uncountable family of suitable local systems;
a one-parameter family of Oakley–Usher tori P 1 , 1 ( r ) ⊂ T ∗ S 3 \mathcal{P}_{1,1}(r) \subset T^*S^3 P 1 , 1 ( r ) ⊂ T ∗ S 3 , equipped with an uncountable family of suitable local systems.
See [24 ] or [10 ] for more details on these families. After compactification, only one member of each of the families above is monotone in Q 3 Q^3 Q 3 . Interestingly, the monotone copy of S 1 × S 2 S^1 \times S^2 S 1 × S 2 is displaceable in Q 3 Q^3 Q 3 and therefore vanishes in the Fukaya category. The monotone copy of the tori is P ⊂ Q 3 \mathcal{P} \subset Q^3 P ⊂ Q 3 and thus appears as a generator by our result.
Kim [20 ] and Kawamoto [19 ] study a different torus in Q 3 Q^3 Q 3 . In fact, their torus can be viewed as the lift of the Chekanov torus in Q 2 Q^2 Q 2 . Enumerative properties of the OU-torus in T ∗ S 3 T^*S^3 T ∗ S 3 were studied by Chan–Pomerleano–Ueda [11 ] and Achig-Andrango [2 ] ; the latter also discusses the effect of compactification on the potential. Sheridan [30 ] established homological mirror symmetry for Fano hypersurfaces (including Q 3 Q^3 Q 3 ) in complex projective space. We identify explicit geometric generators: the Oakley–Usher torus equipped with three critical local systems, together with the Lagrangian sphere Σ \Sigma Σ .
Outline and conventions
Section 2 develops the Ψ \Psi Ψ -invariant, constructs the Gelfand–Cetlin fibration, computes the two OU Ψ \Psi Ψ -germs, and compares P C h \mathcal{P}^{\mathrm{Ch}} P Ch with lifted Vianna tori, proving Theorem A. Section 3 computes the superpotential, proves non-displaceability and split-generation, and studies the change under compactification, proving Theorem B. We use S 1 = R / Z S^1=\mathbb{R}/\mathbb{Z} S 1 = R / Z and integer coefficients for homology unless otherwise specified. Symplectic reduction is denoted by X / / μ = c G X//_{\mu=c}G X / / μ = c G . Induced moment maps on reduced or cut spaces retain the original notation. The symplectic normalization and size parameters agree with those in [10 ] ; the Novikov field and local-system conventions are specified in Section 3.
Acknowledgments
This paper and its companion [10 ] grew out of a Master’s thesis written by AP under the supervision of JB at ETH Zürich. We thank Paul Biran for orchestrating this. We warmly thank Leonid Polterovich for drawing our attention to Oakley–Usher Lagrangians, and Johanna Bimmermann, Marc Fares, Johannes Hauber, David Keren Yaar, Joel Schmitz, Mike Usher, and Renato Vianna for interesting discussions, comments and questions. We are particularly grateful to Felix Schlenk for asking many interesting questions. JB is supported by SNSF Ambizione Grant PZ00P2-223460.
Statement on the use of LLMs
An earlier version of this project comprising this paper and its companion [10 ] was written completely without the use of large language models. In subsequent versions, GPT was used to help split the project into two papers, and proofread the text. The authors take full responsibility for the final mathematical content.
Chekanov OU-torus in the three-dimensional quadric
In this section, we exclusively work with the case k = m = 1 k=m=1 k = m = 1 in the three-dimensional quadric Q 3 Q^3 Q 3 . For ease of notation, we drop the 1 , 1 1,1 1 , 1 -subscripts and sizes of the OU-Lagrangians everywhere. In other words, we write P = P 1 , 1 ( 4 π 3 ) \mathcal{P}=\mathcal{P}_{1,1}\left(\frac{4\pi}{3}\right) P = P 1 , 1 ( 3 4 π ) and P C h = P 1 , 1 C h ( 4 π 3 ) \mathcal{P}^{\mathrm{Ch}}=\mathcal{P}^{\mathrm{Ch}}_{1,1}\left(\frac{4\pi}{3}\right) P Ch = P 1 , 1 Ch ( 3 4 π ) . Recall that these are the monotone OU-Lagrangian and Chekanov OU-Lagrangian, respectively.
In order to distinguish tori up to Hamiltonian isotopy, we use the so-called Ψ \Psi Ψ -invariant, introduced by Shelukhin–Tonkonog–Vianna [29 ] . Let ( X , ω ) (X,\omega) ( X , ω ) be a symplectic manifold. For the general statements in this subsection, take X X X closed, and L L L closed, orientable and spin. These hypotheses suffice for all applications here; extensions to open manifolds require the appropriate compactness hypotheses for the Fukaya algebra.
The Ψ \Psi Ψ -invariant associates to every orientable spin Lagrangian L ⊂ X L\subset X L ⊂ X a number Ψ ( L ) ∈ ( 0 , + ∞ ] \Psi(L)\in(0,+\infty] Ψ ( L ) ∈ ( 0 , + ∞ ] which measures the smallest symplectic area of a J J J -holomorphic disk which contributes nontrivially to some algebraic disk-count in the Fukaya A ∞ A_{\infty} A ∞ -algebra of L L L .
As we will see, Ψ \Psi Ψ itself does not distinguish the tori we are interested in. For example, we will see that Ψ ( P ) = Ψ ( P C h ) = 2 π 3 \Psi(\mathcal{P})=\Psi(\mathcal{P}^{\mathrm{Ch}})=\frac{2\pi}{3} Ψ ( P ) = Ψ ( P Ch ) = 3 2 π . Therefore, we consider its versal deformation. We refer to [9 Section 3] for details on versal deformations.
Definition 2.1. Let L ⊂ X L \subset X L ⊂ X be an orientable spin Lagrangian. We call Ψ-germ (of L L L ) the germ, in the sense of [9 Definition 3.13] , of the invariant Ψ \Psi Ψ at L L L . We denote it by
ψ L = [ Ψ ] L : H 1 ( L ; R ) ⇢ R ∪ { + ∞ } , \psi_L = [\Psi]_L : H^1(L;\mathbb{R}) \dashrightarrow\mathbb{R} \cup\{+\infty\}, ψ L = [ Ψ ] L : H 1 ( L ; R ) ⇢ R ∪ { + ∞ } , where the dashed arrow denotes a map defined on a neighbourhood of the origin of H 1 ( L ; R ) H^1(L;\mathbb{R}) H 1 ( L ; R ) .
Proposition 2.2. The Ψ \Psi Ψ -germ is of the form
ψ L ( a ) = Ψ ( L ) + min { ⟨ a , ξ i ⟩ } , (1) \psi_L(a) = \Psi(L) + \min\{\langle a,\xi_i\rangle\}, \tag*{(1)} ψ L ( a ) = Ψ ( L ) + min {⟨ a , ξ i ⟩} , (1) for some ξ 1 , … , ξ N ∈ H 1 ( L ) \xi_1,\ldots,\xi_N \in H_1(L) ξ 1 , … , ξ N ∈ H 1 ( L ) , provided Ψ ( L ) < + ∞ \Psi(L)<+\infty Ψ ( L ) < + ∞ . In particular, the displayed piecewise-affine expression extends to ψ L : H 1 ( L ; R ) → R \psi_L:H^1(L;\mathbb{R})\to\mathbb{R} ψ L : H 1 ( L ; R ) → R .
Proof. It was proved in [29 ] that, in a small neighbourhood of L L L in the space of Lagrangians, the invariant Ψ \Psi Ψ can be written as the minimum of symplectic areas of relative classes β 1 , … , β N ∈ H 2 ( X , L ) \beta_1,\ldots,\beta_N\in H_2(X,L) β 1 , … , β N ∈ H 2 ( X , L ) with boundary on L L L . More precisely, the classes β i \beta_i β i are those of area Ψ ( L ) \Psi(L) Ψ ( L ) which contribute nontrivially to some symmetrized A ∞ A_\infty A ∞ -operation on odd-degree cohomology classes, in a classically minimal model of the Fukaya algebra of L L L ; see [29 Theorem 4.8 and the proof of Theorem 4.9] . By Gromov’s compactness theorem, this set is finite. Setting ξ i = ∂ β i \xi_i=\partial\beta_i ξ i = ∂ β i and computing the versal deformation yields the claim. □ \square □
Instead of working with ψ L \psi_L ψ L directly, it is often more convenient to work with its (super-)level sets, since these are easier to manipulate and can be represented visually.
Definition 2.3. Let L ⊂ X L\subset X L ⊂ X be an orientable spin Lagrangian with Ψ ( L ) < + ∞ \Psi(L)<+\infty Ψ ( L ) < + ∞ . Using the global piecewise-affine extension of ψ L \psi_L ψ L given by Proposition 2.2, we define
Δ Ψ ( L ) : = { a ∈ H 1 ( L ; R ) ∣ ψ L ( a ) ≥ 0 } . \Delta^\Psi(L):=\{a\in H^1(L;\mathbb{R})\mid\psi_L(a)\ge0\}. Δ Ψ ( L ) := { a ∈ H 1 ( L ; R ) ∣ ψ L ( a ) ≥ 0 } . This is a convex polyhedron, which we call the Ψ \Psi Ψ -polytope of L L L whenever it is bounded. If Ψ ( L ) = + ∞ \Psi(L)=+\infty Ψ ( L ) = + ∞ , the germ is identically + ∞ +\infty + ∞ by [29 Theorem 4.9] ; one may separately set Δ Ψ ( L ) = H 1 ( L ; R ) \Delta^\Psi(L)=H^1(L;\mathbb{R}) Δ Ψ ( L ) = H 1 ( L ; R ) .
Remark 2.4. For the Oakley–Usher torus considered in the next subsection, the relevant structure is a Gelfand–Cetlin fibration , meaning that the Ψ \Psi Ψ -polytope agrees with
the polytope obtained as the base of the fibration, after a translation that shifts the image of the monotone fibre L L L to the origin, if necessary.
the appropriately scaled polar dual of the Newton polytope of the superpotential of that fibre.
The fact that these three polytopes agree is [29 Theorem B] . We only use Gelfand–Cetlin fibrations once to compute ψ P \psi_P ψ P for the Oakley–Usher Lagrangian P ⊂ Q 3 P\subset Q^3 P ⊂ Q 3 . All other invariants will be computed by versal deformation techniques.
We use the following result to distinguish tori.
Proposition 2.5. Let L , L ′ ⊂ X L,L'\subset X L , L ′ ⊂ X be orientable spin Lagrangian manifolds. If a symplectomorphism ϕ : X → X \phi:X\to X ϕ : X → X maps L L L to L ′ L' L ′ , then there exists an isomorphism χ : H 1 ( L ; Z ) → H 1 ( L ′ ; Z ) \chi:H^1(L;\mathbb{Z})\to H^1(L';\mathbb{Z}) χ : H 1 ( L ; Z ) → H 1 ( L ′ ; Z ) such that χ R ( Δ Ψ ( L ) ) = Δ Ψ ( L ′ ) \chi_{\mathbb{R}}(\Delta^\Psi(L))=\Delta^\Psi(L') χ R ( Δ Ψ ( L )) = Δ Ψ ( L ′ ) , where χ R : H 1 ( L ; R ) → H 1 ( L ′ ; R ) \chi_{\mathbb{R}}:H^1(L;\mathbb{R})\to H^1(L';\mathbb{R}) χ R : H 1 ( L ; R ) → H 1 ( L ′ ; R ) denotes the extension to R \mathbb{R} R -coefficients.
Proof. Set χ = ( ϕ ∣ L ) − 1 \chi=(\phi|_{L})^{-1} χ = ( ϕ ∣ L ) − 1 . Pushing forward the almost complex structure and transporting the spin structure by ϕ \phi ϕ identifies the relevant disk moduli spaces and their symplectic areas, so the definition of Ψ \Psi Ψ in [29 ] gives Ψ ( ϕ ( K ) ) = Ψ ( K ) \Psi(\phi(K))=\Psi(K) Ψ ( ϕ ( K )) = Ψ ( K ) . The flux of a deformation is carried to its image under χ R \chi_{\mathbb{R}} χ R ; consequently ψ L ′ = ψ L ∘ χ R − 1 \psi_{L'}=\psi_L\circ\chi_{\mathbb{R}}^{-1} ψ L ′ = ψ L ∘ χ R − 1 . Taking nonnegative superlevel sets gives the claim. □ \square □
In practice, we usually identify H 1 ( L ; R ) H^1(L;\mathbb{R}) H 1 ( L ; R ) and H 1 ( L ′ ; R ) H^1(L';\mathbb{R}) H 1 ( L ′ ; R ) with R n \mathbb{R}^n R n and consider the polytopes up to GL ( n ; Z ) \operatorname{GL}(n;\mathbb{Z}) GL ( n ; Z ) . In that case, we obtain that the Ψ \Psi Ψ -polytopes of L L L and L ′ L' L ′ are related by an element in GL ( n ; Z ) \operatorname{GL}(n;\mathbb{Z}) GL ( n ; Z ) whenever an ambient symplectomorphism takes one to the other.
The Oakley–Usher Lagrangian
To compute the Ψ \Psi Ψ -germ of P \mathcal{P} P , we rely on [29 Proposition C] . It determines Ψ \Psi Ψ of fibres of so-called Gelfand–Cetlin fibrations which have a monotone fibre.
A continuous Lagrangian torus fibration is called Gelfand–Cetlin if it restricts to a toric moment map away from the codimension two faces. As it turns out, the moment map μ 1 , 1 \mu_{1,1} μ 1 , 1 defined by [10 Proposition 3.2] induces a Gelfand–Cetlin fibration on Q 3 Q^3 Q 3 in which P \mathcal{P} P appears as a monotone fibre, see Figure 1 for its polytope.
Figure 1. The Gelfand–Cetlin polytope Δ Q 3 \Delta_{Q^3} Δ Q 3
Proposition 2.6. There is a Gelfand–Cetlin fibration μ : Q 3 → Δ Q 3 \mu:Q^3\to\Delta_{Q^3} μ : Q 3 → Δ Q 3 over the polytope
Δ Q 3 = { x 1 , x 2 ≥ 0 , x 3 − x 1 ≥ 0 , x 3 − x 2 ≥ 0 , 2 π − x 3 ≥ 0 } ⊂ R 3 (2) \Delta_{Q^3}=\{x_1,x_2\ge0,\ x_3-x_1\ge0,\ x_3-x_2\ge0,\ 2\pi-x_3\ge0\}\subset\mathbb{R}^3 \tag*{(2)} Δ Q 3 = { x 1 , x 2 ≥ 0 , x 3 − x 1 ≥ 0 , x 3 − x 2 ≥ 0 , 2 π − x 3 ≥ 0 } ⊂ R 3 (2) such that
the fibre Σ = μ − 1 ( 0 ) \Sigma=\mu^{-1}(0) Σ = μ − 1 ( 0 ) is a Lagrangian sphere;
the fibre P = μ − 1 ( x P ) \mathcal{P}=\mu^{-1}(x_{\mathcal{P}}) P = μ − 1 ( x P ) of x P = ( 2 π 3 , 2 π 3 , 4 π 3 ) x_{\mathcal{P}}=\left(\frac{2\pi}{3},\frac{2\pi}{3},\frac{4\pi}{3}\right) x P = ( 3 2 π , 3 2 π , 3 4 π ) is the Oakley–Usher Lagrangian, and
the restriction μ : Q 3 ∖ Σ → Δ Q 3 ∖ { 0 } \mu:Q^3\setminus\Sigma\to\Delta_{Q^3}\setminus\{0\} μ : Q 3 ∖ Σ → Δ Q 3 ∖ { 0 } is a smooth toric moment map.
Proof. Recall that the continuous moment map
μ 1 , 1 : T ∗ S 3 → R 3 , ( q , p ) ↦ ( 2 π ∣ p ∣ , 2 π ( p 1 q 2 − p 2 q 1 ) , 2 π ( p 3 q 4 − p 4 q 3 ) ) (3) \mu_{1,1}:T^*S^3\to\mathbb{R}^3,\qquad(q,p)\mapsto\left(2\pi|p|,2\pi(p_1q_2-p_2q_1),2\pi(p_3q_4-p_4q_3)\right) \tag*{(3)} μ 1 , 1 : T ∗ S 3 → R 3 , ( q , p ) ↦ ( 2 π ∣ p ∣ , 2 π ( p 1 q 2 − p 2 q 1 ) , 2 π ( p 3 q 4 − p 4 q 3 ) ) (3)
generates a Hamiltonian T 3 T^3 T 3 -action away from the zero-section 0 S 3 0_{S^3} 0 S 3 . Here the last two scalar components use period-one generators of the two SO ( 2 ) \operatorname{SO}(2) SO ( 2 ) factors; this specifies their normalization. and sign in the identification with R 2 \mathbb{R}^2 R 2 . This moment map is smooth away from the zero-section, and its image is the cone defined by the following four inequalities
Δ ^ T ∗ S 3 = { x 1 ± x 2 ± x 3 ≥ 0 } ⊂ R 3 . \widehat{\Delta}_{T^*S^3}=\{x_1\pm x_2\pm x_3\ge0\}\subset\mathbb{R}^3. Δ T ∗ S 3 = { x 1 ± x 2 ± x 3 ≥ 0 } ⊂ R 3 .
Recall however from [10 Remark 3.1] that the element θ 1 / 2 = ( 1 / 2 , 1 / 2 , 1 / 2 ) ∈ T 3 \theta_{1/2}=(1/2,1/2,1/2)\in T^3 θ 1/2 = ( 1/2 , 1/2 , 1/2 ) ∈ T 3 acts trivially, meaning that the action is not effective and thus not toric. This can be remedied as follows: Viewing the standard torus T 3 T^3 T 3 as the quotient of R 3 \mathbb{R}^3 R 3 by the lattice Λ 0 = Z 3 \Lambda_0=\mathbb{Z}^3 Λ 0 = Z 3 , we can add θ 1 / 2 \theta_{1/2} θ 1/2 as a generator to the lattice Λ 0 \Lambda_0 Λ 0 to obtain Λ \Lambda Λ . The new torus R 3 / Λ \mathbb{R}^3/\Lambda R 3 /Λ acts effectively and has the same image under μ 1 , 1 \mu_{1,1} μ 1 , 1 , though with respect to a different lattice in R 3 = t ∗ \mathbb{R}^3=\mathfrak{t}^* R 3 = t ∗ (the target space of μ 1 , 1 \mu_{1,1} μ 1 , 1 ). The dual lattice of Λ \Lambda Λ is given by
Λ ∗ : = { ( x 1 , x 2 , x 3 ) ∈ Z 3 ∣ x 1 + x 2 + x 3 ∈ 2 Z } . \Lambda^*:=\{(x_1,x_2,x_3)\in\mathbb{Z}^3\mid x_1+x_2+x_3\in2\mathbb{Z}\}. Λ ∗ := {( x 1 , x 2 , x 3 ) ∈ Z 3 ∣ x 1 + x 2 + x 3 ∈ 2 Z } .
Since we are used to drawing the image of the moment map in R 3 = t ∗ \mathbb{R}^3=\mathfrak{t}^* R 3 = t ∗ equipped with the standard integral affine structure Λ 0 ∗ = Z 3 \Lambda_0^*=\mathbb{Z}^3 Λ 0 ∗ = Z 3 , we apply some M ∈ GL ( 3 ; Q ) M\in\operatorname{GL}(3;\mathbb{Q}) M ∈ GL ( 3 ; Q ) which induces an isomorphism from Λ ∗ \Lambda^* Λ ∗ to Z 3 \mathbb{Z}^3 Z 3 . It is not hard to show that the linear transformation
M = 1 2 ( 1 − 1 − 1 1 1 − 1 2 0 0 ) (4) M=\frac{1}{2}\begin{pmatrix}1&-1&-1\\1&1&-1\\2&0&0\end{pmatrix} \tag*{(4)} M = 2 1 1 1 2 − 1 1 0 − 1 − 1 0 (4)
has this property. We define a new moment map μ = M μ 1 , 1 \mu=M\mu_{1,1} μ = M μ 1 , 1 . Since the kernel of the original torus action is exactly { 0 , θ 1 / 2 } \{0,\theta_{1/2}\} { 0 , θ 1/2 } , this moment map induces an effective Hamiltonian T 3 T^3 T 3 -action and turns T ∗ S 3 ∖ 0 S 3 T^*S^3\setminus0_{S^3} T ∗ S 3 ∖ 0 S 3 into a toric manifold. Its continuous extension to 0 S 3 0_{S^3} 0 S 3 satisfies μ − 1 ( 0 ) = 0 S 3 \mu^{-1}(0)=0_{S^3} μ − 1 ( 0 ) = 0 S 3 and its image is the cone
Δ T ∗ S 3 = { x 1 , x 2 ≥ 0 , x 3 − x 1 ≥ 0 , x 3 − x 2 ≥ 0 } ⊂ R 3 . \Delta_{T^*S^3}=\{x_1,x_2\ge0,\ x_3-x_1\ge0,\ x_3-x_2\ge0\}\subset\mathbb{R}^3. Δ T ∗ S 3 = { x 1 , x 2 ≥ 0 , x 3 − x 1 ≥ 0 , x 3 − x 2 ≥ 0 } ⊂ R 3 .
Recall from [10 Proposition 3.10] that performing a symplectic cut at 2 π ∣ p ∣ = 2 π 2\pi|p|=2\pi 2 π ∣ p ∣ = 2 π yields the quadric Q 3 Q^3 Q 3 . The Hamiltonian 2 π ∣ p ∣ 2\pi|p| 2 π ∣ p ∣ is equal to the first component of the original moment map μ 1 , 1 \mu_{1,1} μ 1 , 1 and thus corresponds to the third component of μ \mu μ , as a computation using M M M shows. Therefore, performing a symplectic cut with respect to μ 3 = 2 π \mu_3=2\pi μ 3 = 2 π yields Q 3 Q^3 Q 3 . By abuse of notation, we denote the resulting moment map by μ : Q 3 → R 3 \mu:Q^3\to\mathbb{R}^3 μ : Q 3 → R 3 , too. By construction, its image is Δ Q 3 \Delta_{Q^3} Δ Q 3 as defined in (2) . Let Σ = μ − 1 ( 0 ) ⊂ Q 3 \Sigma=\mu^{-1}(0)\subset Q^3 Σ = μ − 1 ( 0 ) ⊂ Q 3 . This is the Lagrangian sphere obtained from 0 S 3 ⊂ T ∗ S 3 0_{S^3}\subset T^*S^3 0 S 3 ⊂ T ∗ S 3 after the symplectic cut.
Recall from [10 Proposition 3.4, equation (17)] that the Oakley–Usher Lagrangians P 1 , 1 ( r ) ⊂ T ∗ S 3 \mathcal{P}_{1,1}(r)\subset T^*S^3 P 1 , 1 ( r ) ⊂ T ∗ S 3 are given by μ 1 , 1 − 1 ( r , 0 , 0 ) \mu_{1,1}^{-1}(r,0,0) μ 1 , 1 − 1 ( r , 0 , 0 ) in the original moment map and thus we obtain
P 1 , 1 ( r ) = μ − 1 ( r 2 , r 2 , r ) , \mathcal{P}_{1,1}(r)=\mu^{-1}\left(\frac{r}{2},\frac{r}{2},r\right), P 1 , 1 ( r ) = μ − 1 ( 2 r , 2 r , r ) ,
with respect to the new moment maps for all r > 0 r>0 r > 0 in T ∗ S 3 T^*S^3 T ∗ S 3 and for r ∈ ( 0 , 2 π ) r\in(0,2\pi) r ∈ ( 0 , 2 π ) in Q 3 Q^3 Q 3 . Recall from [10 Proposition 3.16] that the Oakley–Usher Lagrangian in Q 3 Q^3 Q 3 is monotone for r = 4 π 3 r=\frac{4\pi}{3} r = 3 4 π . It follows that P = μ − 1 ( x P ) ⊂ Q 3 \mathcal{P}=\mu^{-1}(x_{\mathcal{P}})\subset Q^3 P = μ − 1 ( x P ) ⊂ Q 3 for x P = ( 2 π 3 , 2 π 3 , 4 π 3 ) x_{\mathcal{P}}=\left(\frac{2\pi}{3},\frac{2\pi}{3},\frac{4\pi}{3}\right) x P = ( 3 2 π , 3 2 π , 3 4 π ) . By construction, the moment map μ : Q 3 → Δ Q 3 \mu:Q^3\to\Delta_{Q^3} μ : Q 3 → Δ Q 3 is continuous and a smooth toric moment map away from Σ \Sigma Σ .
Let
T ( x ) = μ − 1 ( x ) , x ∈ int Δ Q 3 , T(x)=\mu^{-1}(x),\qquad x\in\operatorname{int}\Delta_{Q^3}, T ( x ) = μ − 1 ( x ) , x ∈ int Δ Q 3 ,
denote the corresponding Gelfand–Cetlin fibres and let d I A ( x , ∂ Δ Q 3 ) d_{\mathrm{IA}}(x,\partial\Delta_{Q^3}) d IA ( x , ∂ Δ Q 3 ) be the integral affine distance of the point x ∈ R 3 x \in\mathbb{R}^3 x ∈ R 3 to the boundary of Δ Q 3 \Delta_{Q^3} Δ Q 3 . Explicitly, it is given by
d I A ( x , ∂ Δ Q 3 ) = min { x 1 , x 2 , x 3 − x 1 , x 3 − x 2 , 2 π − x 3 } . (5) d_{\mathrm{IA}}(x,\partial\Delta_{Q^3})=\min\{x_1,x_2,x_3-x_1,x_3-x_2,2\pi-x_3\}. \tag*{(5)} d IA ( x , ∂ Δ Q 3 ) = min { x 1 , x 2 , x 3 − x 1 , x 3 − x 2 , 2 π − x 3 } . (5)
Together with Proposition 2.6, we use [29 Proposition C] to find the following.
Proposition 2.7. For every x ∈ int Δ Q 3 x \in\operatorname{int}\Delta_{Q^3} x ∈ int Δ Q 3 , the Ψ \Psi Ψ -invariant of a Gelfand–Cetlin fibre T ( x ) T(x) T ( x ) is equal to the integral affine distance of x x x to the boundary ∂ Δ Q 3 \partial\Delta_{Q^3} ∂ Δ Q 3 ,
Ψ ( T ( x ) ) = d I A ( x , ∂ Δ Q 3 ) . \Psi(T(x))=d_{\mathrm{IA}}(x,\partial\Delta_{Q^3}). Ψ ( T ( x )) = d IA ( x , ∂ Δ Q 3 ) .
Remark 2.8. The ease with which the results in [29 ] allow us to compute Ψ \Psi Ψ on all of the fibres of μ \mu μ is the reason why we use the invariant Ψ \Psi Ψ instead of displacement energy. In fact, we suspect that whenever Ψ \Psi Ψ is affine linear (as opposed to just piecewise affine linear) on a neighbourhood of a point x ∈ int Δ Q 3 x \in\operatorname{int}\Delta_{Q^3} x ∈ int Δ Q 3 , then the displacement energy agrees with Ψ \Psi Ψ . The upper bound on displacement energy readily follows from probes (see for example [21 ] ) in the Gelfand–Cetlin polytope, but the lower bound requires more thought.
In particular, Proposition 2.7 allows us to compute the Ψ \Psi Ψ -germ of P \mathcal{P} P . Let H 1 ( P ) ≅ Z 3 H_1(\mathcal{P}) \cong\mathbb{Z}^3 H 1 ( P ) ≅ Z 3 be the isomorphism induced by taking the generators of H 1 ( P ) H_1(\mathcal{P}) H 1 ( P ) defined by the (local) toric action generated by μ \mu μ . By [9 Proposition 3.15] , we obtain a versal deformation of P \mathcal{P} P by considering a ↦ μ − 1 ( x P + a ) a \mapsto\mu^{-1}(x_{\mathcal{P}}+a) a ↦ μ − 1 ( x P + a ) , where a ∈ R 3 ≅ H 1 ( P ; R ) a \in\mathbb{R}^3 \cong H^1(\mathcal{P};\mathbb{R}) a ∈ R 3 ≅ H 1 ( P ; R ) under the identification dual to the above isomorphism. Recall that P = μ − 1 ( ( 2 π 3 , 2 π 3 , 4 π 3 ) ) \mathcal{P}=\mu^{-1}\left(\left(\frac{2\pi}{3},\frac{2\pi}{3},\frac{4\pi}{3}\right)\right) P = μ − 1 ( ( 3 2 π , 3 2 π , 3 4 π ) ) in the Gelfand–Cetlin fibration. From Proposition 2.7 and (5) we obtain the following.
Corollary 2.9. There is an identification H 1 ( P ; R ) ≅ R 3 H^1(\mathcal{P};\mathbb{R}) \cong\mathbb{R}^3 H 1 ( P ; R ) ≅ R 3 such that for all a ∈ R 3 a \in\mathbb{R}^3 a ∈ R 3 ,
ψ P ( a ) = 2 π 3 + min { a 1 , a 2 , a 3 − a 1 , a 3 − a 2 , − a 3 } , \psi_{\mathcal{P}}(a)=\frac{2\pi}{3}+\min\{a_1,a_2,a_3-a_1,a_3-a_2,-a_3\}, ψ P ( a ) = 3 2 π + min { a 1 , a 2 , a 3 − a 1 , a 3 − a 2 , − a 3 } , Δ Ψ ( P ) = { a ∈ R 3 ∣ a 1 , a 2 , a 3 − a 1 , a 3 − a 2 , − a 3 ≥ − 2 π 3 } . \Delta^\Psi(\mathcal{P})=\{a\in\mathbb{R}^3\mid a_1,a_2,a_3-a_1,a_3-a_2,-a_3\geq-\frac{2\pi}{3}\}. Δ Ψ ( P ) = { a ∈ R 3 ∣ a 1 , a 2 , a 3 − a 1 , a 3 − a 2 , − a 3 ≥ − 3 2 π } .
As expected, the Ψ \Psi Ψ -polytope of P \mathcal{P} P agrees with the base polytope Δ Q 3 \Delta_{Q^3} Δ Q 3 of the Gelfand–Cetlin fibration after a translation by − μ ( P ) = ( − 2 π 3 , − 2 π 3 , − 4 π 3 ) -\mu(\mathcal{P})=\left(-\frac{2\pi}{3},-\frac{2\pi}{3},-\frac{4\pi}{3}\right) − μ ( P ) = ( − 3 2 π , − 3 2 π , − 3 4 π ) .
The Chekanov Oakley–Usher Lagrangian
On the other hand, we obtain the following Ψ \Psi Ψ -germ for the Chekanov Oakley–Usher Lagrangian P C h \mathcal{P}^{\mathrm{Ch}} P Ch .
Proposition 2.10. There is an identification H 1 ( P C h ; R ) ≅ R 3 H^1(\mathcal{P}^{\mathrm{Ch}};\mathbb{R}) \cong\mathbb{R}^3 H 1 ( P Ch ; R ) ≅ R 3 such that for all b ∈ R 3 b \in\mathbb{R}^3 b ∈ R 3 ,
ψ P C h ( b ) = 2 π 3 + min { ± b 1 ± b 2 − 2 b 3 , b 3 } , \psi_{\mathcal{P}^{\mathrm{Ch}}}(b)=\frac{2\pi}{3}+\min\{\pm b_1\pm b_2-2b_3,b_3\}, ψ P Ch ( b ) = 3 2 π + min { ± b 1 ± b 2 − 2 b 3 , b 3 } , Δ Ψ ( P C h ) = { b ∈ R 3 ∣ ± b 1 ± b 2 − 2 b 3 , b 3 ≥ − 2 π 3 } . \Delta^\Psi(\mathcal{P}^{\mathrm{Ch}})=\{b\in\mathbb{R}^3\mid\pm b_1\pm b_2-2b_3,b_3\geq-\frac{2\pi}{3}\}. Δ Ψ ( P Ch ) = { b ∈ R 3 ∣ ± b 1 ± b 2 − 2 b 3 , b 3 ≥ − 3 2 π } .
See Figure 2 for an illustration of Δ Ψ ( P C h ) \Delta^\Psi(\mathcal{P}^{\mathrm{Ch}}) Δ Ψ ( P Ch ) . The Ψ \Psi Ψ -polytopes Δ Ψ ( P ) \Delta^\Psi(\mathcal{P}) Δ Ψ ( P ) and Δ Ψ ( P C h ) \Delta^\Psi(\mathcal{P}^{\mathrm{Ch}}) Δ Ψ ( P Ch ) cannot be mapped to one another by an element in GL ( 3 ; Z ) \operatorname{GL}(3;\mathbb{Z}) GL ( 3 ; Z ) . One way of seeing this is as follows: The integral affine distance of the four-valent vertex of Δ Ψ ( P ) \Delta^\Psi(\mathcal{P}) Δ Ψ ( P ) to its opposite facet is 2 π 2\pi 2 π . For Δ Ψ ( P C h ) \Delta^\Psi(\mathcal{P}^{\mathrm{Ch}}) Δ Ψ ( P Ch ) this distance is π \pi π . This is well-defined and an invariant under GL ( n ; Z ) \operatorname{GL}(n;\mathbb{Z}) GL ( n ; Z ) . By Proposition 2.5, we obtain the following.
Figure 2. The Ψ \Psi Ψ -polytope of the Chekanov OU Lagrangian P C h \mathcal{P}^{\mathrm{Ch}} P Ch . We have applied a translation in the x 3 x_3 x 3 -direction for readability.
Theorem 2.11. No ambient symplectomorphism of Q 3 Q^3 Q 3 maps P \mathcal{P} P to P C h \mathcal{P}^{\mathrm{Ch}} P Ch .
All that is left to prove is Proposition 2.10 . Its proof follows the same pattern as the computation of displacement energy germs in [6 , 8 , 9 , 13 ] : on a small neighbourhood of P C h \mathcal{P}^{\mathrm{Ch}} P Ch in its versal deformation, there is an open dense set on which a deformation of P C h \mathcal{P}^{\mathrm{Ch}} P Ch is Hamiltonian isotopic to a deformation of P \mathcal{P} P . Note that the latter are just the fibres T ( x ) T(x) T ( x ) of the Gelfand–Cetlin fibration and thus we can use Proposition 2.7 to prove Proposition 2.10 . This computes ψ P C h \psi_{\mathcal{P}^{\mathrm{Ch}}} ψ P Ch and the Ψ \Psi Ψ -polytope on an open dense subset of a neighbourhood of the origin. By Proposition 2.2 , this determines ψ P C h \psi_{\mathcal{P}^{\mathrm{Ch}}} ψ P Ch and the polytope everywhere.
Proof of Proposition 2.10. The proof is very similar to Sections 4 and 5 in [8 ] , which we recommend for more details.
Step 0: Outline. Recall that, by construction, the Lagrangian torus P C h \mathcal{P}^{\mathrm{Ch}} P Ch is the lift of a curve Γ ( 4 π 3 ) \Gamma\left(\frac{4\pi}{3}\right) Γ ( 3 4 π ) bounding area 2 π 3 \frac{2\pi}{3} 3 2 π in the smooth part of the symplectic reduction
Q 3 / / μ 1 , 1 = 0 G 1 , 1 , Q^3 /\!/_{\mu_{1,1}=0}G_{1,1}, Q 3 / / μ 1 , 1 = 0 G 1 , 1 ,
see the discussion surrounding [10 Definitions 3.15 and 3.20] . Strictly speaking, this reduction differs from the space Z ^ \widehat{Z} Z considered in [10 ] : the actual quotient contains the interval Σ / G 1 , 1 ≅ [ 0 , 1 ] \Sigma/G_{1,1} \cong[0,1] Σ/ G 1 , 1 ≅ [ 0 , 1 ] , whereas Z ^ \widehat{Z} Z is obtained by collapsing this interval to a point. Since the curve Γ ( 4 π 3 ) \Gamma\left(\frac{4\pi}{3}\right) Γ ( 3 4 π ) and its enclosed disk lie away from this interval, the two descriptions agree on the region relevant to our construction.
Recall that a versal deformation of the torus P C h ⊂ Q 3 \mathcal{P}^{\mathrm{Ch}} \subset Q^3 P Ch ⊂ Q 3 is a continuous map
v P C h : H 1 ( P C h ; R ) ≅ R 3 ⇢ L P C h (6) v_{\mathcal{P}^{\mathrm{Ch}}}\colon H^1(\mathcal{P}^{\mathrm{Ch}};\mathbb{R}) \cong\mathbb{R}^3 \dashrightarrow\mathcal{L}_{\mathcal{P}^{\mathrm{Ch}}} \tag*{(6)} v P Ch : H 1 ( P Ch ; R ) ≅ R 3 ⇢ L P Ch (6)
defined on a neighbourhood of the origin 0 ∈ H 1 ( P C h ; R ) 0 \in H^1(\mathcal{P}^{\mathrm{Ch}};\mathbb{R}) 0 ∈ H 1 ( P Ch ; R ) which is a local section to the Lagrangian flux map. By L P C h \mathcal{L}_{\mathcal{P}^{\mathrm{Ch}}} L P Ch we have denoted the space of Lagrangian deformations of P C h \mathcal{P}^{\mathrm{Ch}} P Ch equipped with the C 1 C^1 C 1 topology. For a detailed exposition, we recommend [9 §3.3] , see in particular Definition 3.9 therein. To define such a versal deformation of P C h \mathcal{P}^{\mathrm{Ch}} P Ch explicitly, we will carry out symplectic reduction on a family of probes pointing in the direction ( 1 , 1 , 2 ) (1,1,2) ( 1 , 1 , 2 ) and identify the members of the versal deformation as certain lifts of curves in the family of reduced spaces. The deformation space of P C h \mathcal{P}^{\mathrm{Ch}} P Ch is three-dimensional, see (6) : one of the parameters will come from varying the area of the curve Γ ( ⋅ ) \Gamma(\mathord{\cdot}) Γ ( ⋅ ) we lift, and the two other parameters will come from varying the reduced space in the family that we lift the curve from.
Step 1: Symplectic reduction. Let us now set up the family of symplectic reductions. To that end, we define the following moment map
ν : Q 3 → R 2 , ν = ( μ 2 − μ 1 , μ 3 − μ 1 − μ 2 ) . (7) \nu: Q^3 \to\mathbb{R}^2,\qquad\nu=(\mu_2-\mu_1,\mu_3-\mu_1-\mu_2). \tag*{(7)} ν : Q 3 → R 2 , ν = ( μ 2 − μ 1 , μ 3 − μ 1 − μ 2 ) . (7)
Let us point out that the level sets (fibres) of ν \nu ν point in the direction ( 1 , 1 , 2 ) (1,1,2) ( 1 , 1 , 2 ) in the image of μ \mu μ , which justifies this choice. Since its components are integral linear combinations of the components of μ \mu μ , the moment map ν \nu ν defines a T 2 T^2 T 2 -subaction of the T 3 T^3 T 3 -action generated by μ \mu μ . In fact, the inclusion of tori T 2 ↪ T 3 T^2\hookrightarrow T^3 T 2 ↪ T 3 is given by ( θ 1 , θ 2 ) ↦ ( − θ 1 − θ 2 , θ 1 − θ 2 , θ 2 ) (\theta_1,\theta_2)\mapsto(-\theta_1-\theta_2,\theta_1-\theta_2,\theta_2) ( θ 1 , θ 2 ) ↦ ( − θ 1 − θ 2 , θ 1 − θ 2 , θ 2 ) . Let us compute the image of ν \nu ν . To that end, recall from Proposition 2.6 that the image Δ Q 3 \Delta_{Q^3} Δ Q 3 of μ \mu μ is defined by a set of five inequalities ℓ i ( x ) ≥ 0 \ell_i(x)\geq0 ℓ i ( x ) ≥ 0 for
ℓ 0 ( x ) = 2 π − x 3 , ℓ 1 ( x ) = x 1 , ℓ 2 ( x ) = x 2 , ℓ 3 ( x ) = x 3 − x 1 ℓ 4 ( x ) = x 3 − x 2 . (8) \ell_0(x)=2\pi-x_3,\qquad\ell_1(x)=x_1,\qquad\ell_2(x)=x_2,\qquad\ell_3(x)=x_3-x_1\qquad\ell_4(x)=x_3-x_2. \tag*{(8)} ℓ 0 ( x ) = 2 π − x 3 , ℓ 1 ( x ) = x 1 , ℓ 2 ( x ) = x 2 , ℓ 3 ( x ) = x 3 − x 1 ℓ 4 ( x ) = x 3 − x 2 . (8)
A computation shows that under these constraints the image of ν = ( x 2 − x 1 , x 3 − x 1 − x 2 ) \nu=(x_2-x_1,x_3-x_1-x_2) ν = ( x 2 − x 1 , x 3 − x 1 − x 2 ) is given by the lozenge
⋄ = { ( b 1 , b 2 ) ∈ R 2 ∣ ± b 1 ± b 2 ≤ 2 π } . (9) \diamond=\{(b_1,b_2)\in\mathbb{R}^2\mid\pm b_1\pm b_2\leq2\pi\}. \tag*{(9)} ⋄ = {( b 1 , b 2 ) ∈ R 2 ∣ ± b 1 ± b 2 ≤ 2 π } . (9)
Indeed, since ⋄ = im ( ν ) \diamond=\operatorname{im}(\nu) ⋄ = im ( ν ) is the image of a linear projection of the convex polytope Δ Q 3 \Delta_{Q^3} Δ Q 3 , it is itself a convex polytope spanned by the image of the vertices of Δ Q 3 \Delta_{Q^3} Δ Q 3 . Now let ( b 1 , b 2 ) ∈ int ( ⋄ ) (b_1,b_2)\in\operatorname{int}(\diamond) ( b 1 , b 2 ) ∈ int ( ⋄ ) and let
π ( b 1 , b 2 ) : ν − 1 ( b 1 , b 2 ) → S ( b 1 , b 2 ) (10) \pi_{(b_1,b_2)}:\nu^{-1}(b_1,b_2)\to S_{(b_1,b_2)} \tag*{(10)} π ( b 1 , b 2 ) : ν − 1 ( b 1 , b 2 ) → S ( b 1 , b 2 ) (10)
be the corresponding, possibly singular, symplectic reduction by the T 2 T^2 T 2 -action generated by ν \nu ν . We think of this as the reduction associated to the singular probe
P ( b 1 , b 2 ) = { ( x 2 − x 1 , x 3 − x 1 − x 2 ) = ( b 1 , b 2 ) } ∩ Δ Q 3 . (11) P_{(b_1,b_2)}=\{(x_2-x_1,x_3-x_1-x_2)=(b_1,b_2)\}\cap\Delta_{Q^3}. \tag*{(11)} P ( b 1 , b 2 ) = {( x 2 − x 1 , x 3 − x 1 − x 2 ) = ( b 1 , b 2 )} ∩ Δ Q 3 . (11)
Recall that the intersection of a probe with the boundary of a toric moment polytope yields a smooth point in the reduced space if and only if this intersection happens in the interior of a facet and satisfies the integral transversality condition. This condition asks that a lattice basis in the facet in question can be completed to a lattice basis of the full ambient lattice by the direction vector of the probe, see also [21 §2.1] or [7 Definition 2.3] for a generalization. The singular probes P ( b 1 , b 2 ) P_{(b_1,b_2)} P ( b 1 , b 2 ) turn out to be honest probes for all ( b 1 , b 2 ) (b_1,b_2) ( b 1 , b 2 ) with b 1 , b 2 ≠ 0 b_1,b_2\neq0 b 1 , b 2 = 0 . This follows from the fact that their direction vector ( 1 , 1 , 2 ) (1,1,2) ( 1 , 1 , 2 ) intersects all four facets F i = { ℓ i = 0 } F_i=\{\ell_i=0\} F i = { ℓ i = 0 } for i ∈ { 1 , 2 , 3 , 4 } i\in\{1,2,3,4\} i ∈ { 1 , 2 , 3 , 4 } integrally transversely. Here ℓ i \ell_i ℓ i is defined as in (8). Its intersection with F 0 = { ℓ 0 = 0 } F_0=\{\ell_0=0\} F 0 = { ℓ 0 = 0 } is not integrally transverse, leading to a Z 2 \mathbb{Z}_2 Z 2 -orbifold point in all reduced spaces (as in the case of Z ^ \widehat{Z} Z in general). Let us discuss what the reduced spaces look like in more detail.
In the case b 1 = b 2 = 0 b_1=b_2=0 b 1 = b 2 = 0 , we recover the reduction used in the construction of P C h \mathcal{P}^{\mathrm{Ch}} P Ch with one distinction. The actual quotient S ( 0 , 0 ) S_{(0,0)} S ( 0 , 0 ) contains the interval Σ / T 2 ≅ [ 0 , 1 ] \Sigma/T^2\cong[0,1] Σ/ T 2 ≅ [ 0 , 1 ] , parametrized by ∣ q 1 ∣ 2 + ∣ q 2 ∣ 2 |q_1|^2+|q_2|^2 ∣ q 1 ∣ 2 + ∣ q 2 ∣ 2 . The space Z ^ \widehat{Z} Z considered in [10 ] is obtained by collapsing this interval to a point. It is a two-sphere with a Z 2 \mathbb{Z}_2 Z 2 -orbifold point at the upper end and a distinguished point arising from this collapse. Since the curve defining P C h \mathcal{P}^{\mathrm{Ch}} P Ch and its enclosed disk lie in the smooth part away from the interval, the two descriptions agree on the region relevant to our construction.
In case only one of the b i b_i b i is zero, the reduced space is a two-sphere containing a Z 2 \mathbb{Z}_2 Z 2 -orbifold point and a distinguished point coming from the fact that the probe P ( b 1 , b 2 ) P_{(b_1,b_2)} P ( b 1 , b 2 ) intersects one of the (toric!) edges of Δ Q 3 \Delta_{Q^3} Δ Q 3 , whose fibre is a circle. Its smooth part is symplectomorphic to a standard cylinder.
(3) In case b 1 b 2 ≠ 0 b_1b_2 \ne0 b 1 b 2 = 0 , the reduced space is a two-sphere containing only the Z 2 \mathbb{Z}_2 Z 2 -orbifold point coming from the upper facet. The complement of that point is symplectomorphic to a standard symplectic two-disk. Again, the smoothness in the interior of the disk is due to the fact that the probe P ( b 1 , b 2 ) P_{(b_1,b_2)} P ( b 1 , b 2 ) intersects one of the facets F 1 , … , F 4 F_1,\ldots,F_4 F 1 , … , F 4 integrally transversely.
Step 2: Versal deformation. Let us now define the versal deformation of P C h \mathcal{P}^{\mathrm{Ch}} P Ch . We recall (for example from [9 §3.3] ) that any Weinstein chart φ : T ∗ L ⊃ U → X \varphi:T^*L\supset U\to X φ : T ∗ L ⊃ U → X of a Lagrangian torus L ⊂ X L\subset X L ⊂ X induces a versal deformation v L ( b ) = φ ( { b } × T n ) v_L(b)=\varphi(\{b\}\times T^n) v L ( b ) = φ ({ b } × T n ) . Here we have identified T ∗ L = R n × T n T^*L=\mathbb{R}^n\times T^n T ∗ L = R n × T n and H 1 ( L ; R ) = R n H^1(L;\mathbb{R})=\mathbb{R}^n H 1 ( L ; R ) = R n . To make our versal deformation compatible with the symplectic reduction set up above, we choose a T 2 T^2 T 2 -equivariant Weinstein chart of P C h \mathcal{P}^{\mathrm{Ch}} P Ch . On the model space T ∗ T 3 = { ( p i , q i ) } T^*T^3=\{(p_i,q_i)\} T ∗ T 3 = {( p i , q i )} , we pick the T 2 T^2 T 2 -action defined by standard rotation in the q 1 , q 2 q_1,q_2 q 1 , q 2 components generated by the moment map given by the projection to ( p 1 , p 2 ) (p_1,p_2) ( p 1 , p 2 ) . On Q 3 Q^3 Q 3 , we have the T 2 T^2 T 2 -action defined by ν \nu ν as in (7) . Although μ \mu μ is not smooth at Σ \Sigma Σ , the two components of ν \nu ν are the last two components of μ 1 , 1 \mu_{1,1} μ 1 , 1 . Hence ν \nu ν is smooth everywhere and generates the original T 2 T^2 T 2 -action. Since P C h \mathcal{P}^{\mathrm{Ch}} P Ch is invariant under the latter, we can choose a Weinstein chart φ \varphi φ which is T 2 T^2 T 2 -equivariant with respect to these two actions. Define a versal deformation
v = v P C h φ : H 1 ( T 3 ; R ) ≅ R 3 ⟶ L P C h , v ( b ) = φ ( { b } × T 3 ) . (12) v=v_{\mathcal{P}^{\mathrm{Ch}}}^{\varphi}:H^1(T^3;\mathbb{R})\cong\mathbb{R}^3\longrightarrow\mathcal{L}_{\mathcal{P}^{\mathrm{Ch}}},\qquad v(b)=\varphi(\{b\}\times T^3). \tag*{(12)} v = v P Ch φ : H 1 ( T 3 ; R ) ≅ R 3 ⟶ L P Ch , v ( b ) = φ ({ b } × T 3 ) . (12)
Lemma 2.12. The Lagrangians appearing in the versal deformation (12) are lifts of curves from the family of symplectic reductions (10) . More specifically, there exists a family of curves C ( b 1 , b 2 ) ( b 3 ) ⊂ S ( b 1 , b 2 ) C_{(b_1,b_2)}(b_3)\subset S_{(b_1,b_2)} C ( b 1 , b 2 ) ( b 3 ) ⊂ S ( b 1 , b 2 ) contained in the smooth part of the reduced spaces S ( b 1 , b 2 ) S_{(b_1,b_2)} S ( b 1 , b 2 ) and bounding a disk of symplectic area 2 π 3 + b 3 \frac{2\pi}{3}+b_3 3 2 π + b 3 such that
v ( b ) = π ( b 1 , b 2 ) − 1 ( C ( b 1 , b 2 ) ( b 3 ) ) v(b)=\pi^{-1}_{(b_1,b_2)}\bigl(C_{(b_1,b_2)}(b_3)\bigr) v ( b ) = π ( b 1 , b 2 ) − 1 ( C ( b 1 , b 2 ) ( b 3 ) ) for all b = ( b 1 , b 2 , b 3 ) ∈ H 1 ( T 3 ; R ) ≅ R 3 b=(b_1,b_2,b_3)\in H^1(T^3;\mathbb{R})\cong\mathbb{R}^3 b = ( b 1 , b 2 , b 3 ) ∈ H 1 ( T 3 ; R ) ≅ R 3 contained in a small enough neighbourhood V V V of the origin such that v v v is defined.
Proof. By equivariance of φ \varphi φ , the tori v ( b ) v(b) v ( b ) are T 2 T^2 T 2 -invariant with respect to the action generated by ν \nu ν . Therefore, these tori are lifts of curves in the corresponding reduced spaces.
To compute the areas of the disks bounded by these curves, we choose the first two generators of H 1 ( P C h ; Z ) H_1(\mathcal{P}^{\mathrm{Ch}};\mathbb{Z}) H 1 ( P Ch ; Z ) to be the T 2 T^2 T 2 -orbit classes. We choose the third generator to be the boundary of a section over the disk bounded by the original reduced curve. Such a section exists since the principal T 2 T^2 T 2 -bundle over this disk is trivial. With this choice of marking, the equivariant Weinstein chart can be normalized so that ν ∘ φ = ( p 1 , p 2 ) \nu\circ\varphi=(p_1,p_2) ν ∘ φ = ( p 1 , p 2 ) . For sufficiently small b b b , the reduced curves and their bounding disks remain in the smooth part of the reduced spaces, and the corresponding section disks can be chosen to vary smoothly.
Their symplectic areas agree with the reduced disk areas, and their variation is given by the Lagrangian flux evaluated on the third generator. By the definition of the versal deformation, this variation is b 3 b_3 b 3 . Since the original reduced disk has area 2 π / 3 2\pi/3 2 π /3 , we obtain
Area ( D b ) = 2 π 3 + b 3 . \operatorname{Area}(D_b)=\frac{2\pi}{3}+b_3. Area ( D b ) = 3 2 π + b 3 .
The argument is the same as in [8 Proposition 4.4] , with the above choice of marking and continuously varying section disks.
Step 3: Deforming the tori. The crucial observation which makes this proof work is: whenever b 1 , b 2 ≠ 0 b_1,b_2 \ne0 b 1 , b 2 = 0 , the reduced space S ( b 1 , b 2 ) S_{(b_1,b_2)} S ( b 1 , b 2 ) is a sphere containing a single Z 2 \mathbb{Z}_2 Z 2 -orbifold point (and no distinguished points), see point (3) in the enumeration above. Therefore the curve C ( b 1 , b 2 ) ( b 3 ) ⊂ S ( b 1 , b 2 ) C_{(b_1,b_2)}(b_3) \subset S_{(b_1,b_2)} C ( b 1 , b 2 ) ( b 3 ) ⊂ S ( b 1 , b 2 ) can be deformed by a Hamiltonian isotopy into a standard curve in S ( b 1 , b 2 ) S_{(b_1,b_2)} S ( b 1 , b 2 ) bounding the same area. By standard we mean an orbit of the residual S 1 S^1 S 1 -action on the reduced space, or, equivalently, the curve obtained as the image of the corresponding fibre T ( x ) = μ − 1 ( x ) T(x)=\mu^{-1}(x) T ( x ) = μ − 1 ( x ) under the reduction. This can be used to prove that v ( b ) v(b) v ( b ) is Hamiltonian isotopic to some T ( x ) T(x) T ( x ) , and thus, by Proposition 2.7, allows us to compute the Ψ \Psi Ψ -invariant Ψ ( v ( b ) ) \Psi(v(b)) Ψ ( v ( b )) whenever b 1 , b 2 ≠ 0 b_1,b_2 \ne0 b 1 , b 2 = 0 . More precisely, we will show the following.
Lemma 2.13. We have
v ( b ) ≅ T ( Φ ( b ) ) for all b = ( b 1 , b 2 , b 3 ) with b 1 , b 2 ≠ 0 in V v(b) \cong T(\Phi(b)) \qquad\text{for all } b=(b_1,b_2,b_3)\text{ with }b_1,b_2\ne0\text{ in }V v ( b ) ≅ T ( Φ ( b )) for all b = ( b 1 , b 2 , b 3 ) with b 1 , b 2 = 0 in V where Φ : R 3 → R 3 \Phi:\mathbb{R}^3\to\mathbb{R}^3 Φ : R 3 → R 3 is the continuous and bijective piecewise integral affine map defined by
Φ ( b 1 , b 2 , b 3 ) − ( 2 π 3 , 2 π 3 , 4 π 3 ) = { ( b 3 , b 1 + b 3 , b 1 + b 2 + 2 b 3 ) b 1 ≥ 0 , b 2 ≥ 0 ( b 3 − b 1 , b 3 , b 2 − b 1 + 2 b 3 ) b 1 ≤ 0 , b 2 ≥ 0 ( b 3 − b 1 − b 2 , b 3 − b 2 , 2 b 3 − b 1 − b 2 ) b 1 ≤ 0 , b 2 ≤ 0 ( b 3 − b 2 , b 1 − b 2 + b 3 , b 1 − b 2 + 2 b 3 ) b 1 ≥ 0 , b 2 ≤ 0. (13) \Phi(b_1,b_2,b_3)-\left(\frac{2\pi}{3},\frac{2\pi}{3},\frac{4\pi}{3}\right)=
\begin{cases}
(b_3,b_1+b_3,b_1+b_2+2b_3) & b_1\ge0,b_2\ge0\\
(b_3-b_1,b_3,b_2-b_1+2b_3) & b_1\le0,b_2\ge0\\
(b_3-b_1-b_2,b_3-b_2,2b_3-b_1-b_2) & b_1\le0,b_2\le0\\
(b_3-b_2,b_1-b_2+b_3,b_1-b_2+2b_3) & b_1\ge0,b_2\le0.
\end{cases} \tag*{(13)} Φ ( b 1 , b 2 , b 3 ) − ( 3 2 π , 3 2 π , 3 4 π ) = ⎩ ⎨ ⎧ ( b 3 , b 1 + b 3 , b 1 + b 2 + 2 b 3 ) ( b 3 − b 1 , b 3 , b 2 − b 1 + 2 b 3 ) ( b 3 − b 1 − b 2 , b 3 − b 2 , 2 b 3 − b 1 − b 2 ) ( b 3 − b 2 , b 1 − b 2 + b 3 , b 1 − b 2 + 2 b 3 ) b 1 ≥ 0 , b 2 ≥ 0 b 1 ≤ 0 , b 2 ≥ 0 b 1 ≤ 0 , b 2 ≤ 0 b 1 ≥ 0 , b 2 ≤ 0. (13)
Proof. Note that the condition b 1 , b 2 ≠ 0 b_1,b_2\ne0 b 1 , b 2 = 0 separates int ( ◊ ) \operatorname{int}(\Diamond) int ( ◊ ) into four disjoint open regions. Let us focus on the case b 1 , b 2 > 0 b_1,b_2>0 b 1 , b 2 > 0 . Since ν = ( x 2 − x 1 , x 3 − x 1 − x 2 ) \nu=(x_2-x_1,x_3-x_1-x_2) ν = ( x 2 − x 1 , x 3 − x 1 − x 2 ) , this implies x 2 > x 1 x_2>x_1 x 2 > x 1 and x 3 > x 1 + x 2 x_3>x_1+x_2 x 3 > x 1 + x 2 in Δ Q 3 \Delta_{Q^3} Δ Q 3 , meaning that this yields only probes P b 1 , b 2 P_{b_1,b_2} P b 1 , b 2 intersecting the facet F 1 = { ℓ 1 = 0 } F_1=\{\ell_1=0\} F 1 = { ℓ 1 = 0 } on the lower and the facet F 0 = { ℓ 0 = 0 } F_0=\{\ell_0=0\} F 0 = { ℓ 0 = 0 } on the upper end. As we have discussed, the lower intersection with Δ Q 3 \Delta_{Q^3} Δ Q 3 is integrally transverse whereas the upper one contributes a Z 2 \mathbb{Z}_2 Z 2 -orbifold point to the reduced space S ( b 1 , b 2 ) S_{(b_1,b_2)} S ( b 1 , b 2 ) . By Lemma 2.12, we find that v ( b ) v(b) v ( b ) (for b 1 , b 2 > 0 b_1,b_2>0 b 1 , b 2 > 0 and arbitrary b 3 b_3 b 3 such that ( b 1 , b 2 , b 3 ) ∈ V (b_1,b_2,b_3)\in V ( b 1 , b 2 , b 3 ) ∈ V ) projects to a curve C ( b 1 , b 2 ) ( b 3 ) C_{(b_1,b_2)}(b_3) C ( b 1 , b 2 ) ( b 3 ) bounding area 2 π 3 + b 3 \frac{2\pi}{3}+b_3 3 2 π + b 3 in the reduced space S ( b 1 , b 2 ) S_{(b_1,b_2)} S ( b 1 , b 2 ) . By a Hamiltonian isotopy we can map the curve C ( b 1 , b 2 ) ( b 3 ) C_{(b_1,b_2)}(b_3) C ( b 1 , b 2 ) ( b 3 ) to the standard circle S 1 ( 2 π 3 + b 3 ) S^1\left(\frac{2\pi}{3}+b_3\right) S 1 ( 3 2 π + b 3 ) bounding the same area in the disk obtained as the smooth part of S ( b 1 , b 2 ) S_{(b_1,b_2)} S ( b 1 , b 2 ) . After reduction by T 2 T^2 T 2 , the T 3 T^3 T 3 -action on Q 3 Q^3 Q 3 (we are away from Σ \Sigma Σ ) induces a residual S 1 S^1 S 1 -action on S ( b 1 , b 2 ) S_{(b_1,b_2)} S ( b 1 , b 2 ) , which allows us to identify the smooth part of the reduced space S 1 S^1 S 1 -equivariantly with the standard symplectic disk carrying the standard S 1 S^1 S 1 -action by rotation. The standard circle S 1 ( 2 π 3 + b 3 ) S^1\left(\frac{2\pi}{3}+b_3\right) S 1 ( 3 2 π + b 3 ) is an orbit of that residual S 1 S^1 S 1 -action. Its preimage π ( b 1 , b 2 ) − 1 ( S 1 ( 2 π 3 + b 3 ) ) \pi^{-1}_{(b_1,b_2)}\left(S^1\left(\frac{2\pi}{3}+b_3\right)\right) π ( b 1 , b 2 ) − 1 ( S 1 ( 3 2 π + b 3 ) ) under the reduction is a fibre torus T ( x ) = μ − 1 ( x ) T(x)=\mu^{-1}(x) T ( x ) = μ − 1 ( x ) . By lifting this Hamiltonian isotopy, we find that v ( b ) v(b) v ( b ) is Hamiltonian isotopic to T ( x ) T(x) T ( x ) .
Let us determine x = ( x 1 , x 2 , x 3 ) = Φ ( b ) x=(x_1,x_2,x_3)=\Phi(b) x = ( x 1 , x 2 , x 3 ) = Φ ( b ) in terms of b b b . Since by construction T ( x ) T(x) T ( x ) is contained in the level set ν − 1 ( b 1 , b 2 ) \nu^{-1}(b_1,b_2) ν − 1 ( b 1 , b 2 ) , we find that x 2 − x 1 = b 1 x_2-x_1=b_1 x 2 − x 1 = b 1 and x 3 − x 1 − x 2 = b 2 x_3-x_1-x_2=b_2 x 3 − x 1 − x 2 = b 2 . This constrains x x x to lie on the probe P ( b 1 , b 2 ) P_{(b_1,b_2)} P ( b 1 , b 2 ) defined in Equation (11) . Since T ( x ) T(x) T ( x ) projects to the circle S 1 ( 2 π 3 + b 3 ) S^1\left(\frac{2\pi}{3}+b_3\right) S 1 ( 3 2 π + b 3 ) in the reduced space, the integral affine distance of x x x to the facet F 1 = { ℓ 1 = 0 } F_1=\{\ell_1=0\} F 1 = { ℓ 1 = 0 } is 2 π 3 + b 3 \frac{2\pi}{3}+b_3 3 2 π + b 3 . Indeed, on the smooth part of the reduced space S ( b 1 , b 2 ) S_{(b_1,b_2)} S ( b 1 , b 2 ) , the residual S 1 S^1 S 1 -action is Hamiltonian and identifies this space S 1 S^1 S 1 -equivariantly with the standard symplectic disk of area equal to the affine length of the probe P ( b 1 , b 2 ) P_{(b_1,b_2)} P ( b 1 , b 2 ) . Under this identification, the moment map of the residual S 1 S^1 S 1 -action is exactly the integral affine coordinate along the probe, normalized to vanish at the facet where the probe enters the polytope. Combining these facts, we find
Φ ( b 1 , b 2 , b 3 ) = ( 2 π 3 , 2 π 3 , 4 π 3 ) + ( b 3 , b 1 + b 3 , b 1 + b 2 + 2 b 3 ) . (14) \Phi(b_1,b_2,b_3)=\left(\frac{2\pi}{3},\frac{2\pi}{3},\frac{4\pi}{3}\right)+(b_3,b_1+b_3,b_1+b_2+2b_3). \tag*{(14)} Φ ( b 1 , b 2 , b 3 ) = ( 3 2 π , 3 2 π , 3 4 π ) + ( b 3 , b 1 + b 3 , b 1 + b 2 + 2 b 3 ) . (14)
The same kind of argument shows the formula in the three remaining cases. The map defined in (13) is the unique continuous extension to the region b 1 b 2 = 0 b_1b_2=0 b 1 b 2 = 0 of these partially defined maps. □
With this in hand, we can finish the proof of Proposition 2.10. Combining Lemma 2.13 and Proposition 2.7, we find
Ψ ( v ( b ) ) = Ψ ( T ( Φ ( b ) ) ) = d I A ( Φ ( b ) , ∂ Δ Q 3 ) = min i = 0 , … , 4 { ℓ i ( Φ ( b ) ) } . (15) \Psi(v(b))=\Psi(T(\Phi(b)))=d_{I_A}(\Phi(b),\partial\Delta_{Q^3})=\min_{i=0,\ldots,4}\{\ell_i(\Phi(b))\}. \tag*{(15)} Ψ ( v ( b )) = Ψ ( T ( Φ ( b ))) = d I A ( Φ ( b ) , ∂ Δ Q 3 ) = i = 0 , … , 4 min { ℓ i ( Φ ( b ))} . (15)
Again, we only discuss the case b 1 , b 2 > 0 b_1,b_2>0 b 1 , b 2 > 0 , as the others are analogous. Using (13) , we find that the functions ℓ i ∘ Φ \ell_i\circ\Phi ℓ i ∘ Φ are given by
ℓ 0 ( Φ ( b ) ) = 2 π 3 − b 1 − b 2 − 2 b 3 , ℓ 1 ( Φ ( b ) ) = 2 π 3 + b 3 , ℓ 2 ( Φ ( b ) ) = 2 π 3 + b 1 + b 3 , \ell_0(\Phi(b))=\frac{2\pi}{3}-b_1-b_2-2b_3,\qquad\ell_1(\Phi(b))=\frac{2\pi}{3}+b_3,\qquad\ell_2(\Phi(b))=\frac{2\pi}{3}+b_1+b_3, ℓ 0 ( Φ ( b )) = 3 2 π − b 1 − b 2 − 2 b 3 , ℓ 1 ( Φ ( b )) = 3 2 π + b 3 , ℓ 2 ( Φ ( b )) = 3 2 π + b 1 + b 3 ,
ℓ 3 ( Φ ( b ) ) = 2 π 3 + b 1 + b 2 + b 3 , ℓ 4 ( Φ ( b ) ) = 2 π 3 + b 2 + b 3 . \ell_3(\Phi(b))=\frac{2\pi}{3}+b_1+b_2+b_3,\qquad\ell_4(\Phi(b))=\frac{2\pi}{3}+b_2+b_3. ℓ 3 ( Φ ( b )) = 3 2 π + b 1 + b 2 + b 3 , ℓ 4 ( Φ ( b )) = 3 2 π + b 2 + b 3 .
In that case, we clearly have ℓ 1 ≤ ℓ 2 , ℓ 3 , ℓ 4 \ell_1\leq\ell_2,\ell_3,\ell_4 ℓ 1 ≤ ℓ 2 , ℓ 3 , ℓ 4 , meaning that the minimum in (15) is attained by ℓ 0 \ell_0 ℓ 0 or by ℓ 1 \ell_1 ℓ 1 . Note that those are indeed two of the expressions appearing in the minimum in Proposition 2.10. The remaining three cases yield b 3 b_3 b 3 and the remaining three expressions in the minimum, in each case.
Step 4: Extension. The above determines ψ P C h \psi_{\mathcal{P}^{\mathrm{Ch}}} ψ P Ch on an open dense subset of a neighbourhood of the origin. By Proposition 2.2, this determines ψ P C h \psi_{\mathcal{P}^{\mathrm{Ch}}} ψ P Ch everywhere, and thus Δ Ψ ( P C h ) \Delta^{\Psi}(\mathcal{P}^{\mathrm{Ch}}) Δ Ψ ( P Ch ) . This step does not work in the case of displacement energy. □
Remark 2.14. In light of this proof, the fact that P \mathcal{P} P and P C h \mathcal{P}^{\mathrm{Ch}} P Ch are Lagrangian isotopic becomes very evident. Indeed, both P \mathcal{P} P and P C h \mathcal{P}^{\mathrm{Ch}} P Ch can be deformed into Lagrangian tori in a nearby reduced space with b 1 , b 2 ≠ 0 b_1,b_2\neq0 b 1 , b 2 = 0 , which only has one orbifold point. Since their curves are isotopic there, so are the respective deformations.
Lifted Vianna tori
The quadric Q 3 Q^3 Q 3 contains infinitely many monotone Lagrangian tori which are not Hamiltonian isotopic (and not symplectomorphic, either) to one another. These can be constructed by lifting Vianna tori from the two-dimensional quadric Q 2 = S 2 × S 2 Q^2=S^2\times S^2 Q 2 = S 2 × S 2 to Q 3 Q^3 Q 3 by
viewing Q 2 = S 2 × S 2 Q^2=S^2\times S^2 Q 2 = S 2 × S 2 as a suitable symplectic quotient of Q 3 Q^3 Q 3 and lifting Vianna tori from that quotient, or, equivalently, by
Biran’s circle bundle construction.
This is why we call them lifted Vianna tori . Both the construction using Biran’s approach and the fact that there are infinitely many tori in Q 3 Q^3 Q 3 are known to experts, though this result does not seem to have appeared in the literature. For example it can be seen as a variation of the case of C P 3 \mathbb{CP}^3 CP 3 which is treated in [12 ] , or as a special case of forthcoming work by Diogo–Tonkonog–Vianna–Wu [14 ] , where similar ideas will be treated in much greater generality. We will use the first approach and work with symplectic reduction.
As we will see below, the Oakley–Usher torus P ⊂ Q 3 \mathcal{P}\subset Q^3 P ⊂ Q 3 is the lift of the Clifford torus S e q 1 × S e q 1 ⊂ S 2 × S 2 S^1_{\mathrm{eq}}\times S^1_{\mathrm{eq}}\subset S^2\times S^2 S eq 1 × S eq 1 ⊂ S 2 × S 2 . What is of interest to us here is that the Chekanov Oakley–Usher Lagrangian is not of this form.
Theorem 2.15. The Chekanov Oakley–Usher Lagrangian P C h ⊂ Q 3 \mathcal{P}^{\mathrm{Ch}} \subset Q^3 P Ch ⊂ Q 3 is not symplectomorphic to any of the lifted Vianna tori.
Let us now discuss the Vianna tori T k V i a ⊂ S 2 × S 2 T_k^{\mathrm{Via}} \subset S^2 \times S^2 T k Via ⊂ S 2 × S 2 . Let Δ 0 = [ − 2 π 3 , 2 π 3 ] × [ − 2 π 3 , 2 π 3 ] ⊂ R 2 \Delta_0 = \left[-\frac{2\pi}{3},\frac{2\pi}{3}\right] \times\left[-\frac{2\pi}{3},\frac{2\pi}{3}\right] \subset\mathbb{R}^2 Δ 0 = [ − 3 2 π , 3 2 π ] × [ − 3 2 π , 3 2 π ] ⊂ R 2 be the Delzant polytope of the standard toric structure on S 2 × S 2 S^2 \times S^2 S 2 × S 2 equipped with the monotone symplectic form which evaluates to 4 π 3 \frac{4\pi}{3} 3 4 π on both standard generators of H 2 ( S 2 × S 2 ) H_2(S^2 \times S^2) H 2 ( S 2 × S 2 ) . It will become clear below why we choose this normalization.
Let { Δ k } k ∈ N \{\Delta_k\}_{k\in\mathbb{N}} { Δ k } k ∈ N denote the set of all possible almost toric base diagrams obtained from ATF-mutations of Δ 0 \Delta_0 Δ 0 . We take T k V i a T_k^{\mathrm{Via}} T k Via to be the fibre over 0 ∈ Δ k 0 \in\Delta_k 0 ∈ Δ k , after nodal sliding the nodes close enough to the vertices such that 0 0 0 does not lie on any branch cuts.
We can forget all of the almost toric decorations (nodes, branch cuts) of Δ k \Delta_k Δ k to obtain an ordinary polytope Δ k ⊂ R 2 \Delta_k \subset\mathbb{R}^2 Δ k ⊂ R 2 .
Proposition 2.18. The Ψ \Psi Ψ -polytope of each Vianna torus coincides with its ATF base diagram,
Δ Ψ ( T k V i a ) = Δ k . (16) \Delta^\Psi(T_k^{\mathrm{Via}}) = \Delta_k. \tag*{(16)} Δ Ψ ( T k Via ) = Δ k . (16)
Proof. The proof is based on the same approach as the proof of [8 Theorem 3.2] , to which we refer for details. Let μ S 2 × S 2 : S 2 × S 2 → □ \mu_{S^2 \times S^2}: S^2 \times S^2 \to\square μ S 2 × S 2 : S 2 × S 2 → □ be the standard toric fibration of S 2 × S 2 S^2 \times S^2 S 2 × S 2 over the square □ = [ − a , a ] × [ − a , a ] \square= [-a,a] \times[-a,a] □ = [ − a , a ] × [ − a , a ] , whose size a > 0 a > 0 a > 0 depends on the normalization of the symplectic form. For every y ∈ □ y \in\square y ∈ □ , denote by T y ⊂ S 2 × S 2 T_y \subset S^2 \times S^2 T y ⊂ S 2 × S 2 the corresponding toric fibre. Then Ψ ( T y ) = d I A ( y , ∂ □ ) \Psi(T_y) = d_{\mathrm{IA}}(y,\partial\square) Ψ ( T y ) = d IA ( y , ∂ □ ) by [29 Proposition C] . Since this agrees with the displacement energy, we can use the same arguments as in the proof of [8 Theorem 3.2] , to find that Ψ ( π k − 1 ( y ) ) = d I A ( y , ∂ Δ k ) \Psi(\pi_k^{-1}(y)) = d_{\mathrm{IA}}(y,\partial\Delta_k) Ψ ( π k − 1 ( y )) = d IA ( y , ∂ Δ k ) , for y y y in an open dense subset of a neighbourhood of the origin. Here π k : S 2 × S 2 → Δ k \pi_k:S^2 \times S^2 \to\Delta_k π k : S 2 × S 2 → Δ k is the almost toric projection map. Thus, Ψ T k V i a \Psi_{T_k^{\mathrm{Via}}} Ψ T k Via agrees with the integral affine distance to the boundary of Δ k \Delta_k Δ k on an open dense subset of a neighbourhood of the origin and applying Proposition 2.2 again, we find that it does so everywhere. This proves the claim.
We note that for the dual of the Newton polytope of the disk counting potential (instead of the Ψ \Psi Ψ -polytope), this result was proved by Vianna [33 Theorem 5.1] . It also follows from subsequent work by Pascaleff–Tonkonog [26 ] .
Let us now discuss the relevant lifts of these tori using symplectic reduction. To that end, we perform symplectic reduction on Q 3 Q^3 Q 3 with respect to the third component μ 3 : Q 3 → R \mu_3: Q^3 \to\mathbb{R} μ 3 : Q 3 → R of the Gelfand–Cetlin moment map μ \mu μ . More precisely we consider the symplectic quotient M = μ 3 − 1 ( 4 π 3 ) / S 1 M=\mu_3^{-1}\left(\frac{4\pi}{3}\right)/S^1 M = μ 3 − 1 ( 3 4 π ) / S 1 . Since μ \mu μ is toric away from Σ \Sigma Σ (see Proposition 2.6 ), it is easy to check that the S 1 S^1 S 1 -action on this level set is free and thus M M M is an honest symplectic manifold. To identify M M M , we note that it carries a residual toric T 2 T^2 T 2 -action which has the polytope
Δ Q 3 ∩ { x 3 = 4 π 3 } = [ 0 , 4 π 3 ] × [ 0 , 4 π 3 ] × { 4 π 3 } \Delta_{Q^3} \cap\left\{x_3=\frac{4\pi}{3}\right\}=[0,\tfrac{4\pi}{3}]\times[0,\tfrac{4\pi}{3}]\times\{\tfrac{4\pi}{3}\} Δ Q 3 ∩ { x 3 = 3 4 π } = [ 0 , 3 4 π ] × [ 0 , 3 4 π ] × { 3 4 π }
as its image. This polytope is a standard square and thus the reduced space is symplectomorphic to Q 2 = S 2 × S 2 Q^2=S^2\times S^2 Q 2 = S 2 × S 2 with an appropriately scaled monotone symplectic form. This is an instance of what is called toric reduction in [7 §2.2] , to which we refer for details.
Denote by p : μ 3 − 1 ( 4 π 3 ) → Q 2 p:\mu_3^{-1}\left(\frac{4\pi}{3}\right)\to Q^2 p : μ 3 − 1 ( 3 4 π ) → Q 2 the reduction map. For any Lagrangian L ⊂ Q 2 L\subset Q^2 L ⊂ Q 2 , we denote its lift via p p p by L ^ = p − 1 ( L ) ⊂ Q 3 \widehat{L}=p^{-1}(L)\subset Q^3 L = p − 1 ( L ) ⊂ Q 3 . This lift is automatically a Lagrangian submanifold.
Proposition 2.19. The lift T ^ C l \widehat{T}_{\mathrm{Cl}} T Cl of the Clifford torus T C l = S e q 1 × S e q 1 ⊂ M T_{\mathrm{Cl}}=S^1_{\mathrm{eq}}\times S^1_{\mathrm{eq}}\subset M T Cl = S eq 1 × S eq 1 ⊂ M coincides with the Oakley–Usher Lagrangian P ⊂ Q 3 P\subset Q^3 P ⊂ Q 3 .
Proof. The Clifford torus T C l = S e q 1 × S e q 1 T_{\mathrm{Cl}}=S^1_{\mathrm{eq}}\times S^1_{\mathrm{eq}} T Cl = S eq 1 × S eq 1 is the fibre over the central point ( 2 π 3 , 2 π 3 ) ∈ [ 0 , 4 π 3 ] × [ 0 , 4 π 3 ] \left(\frac{2\pi}{3},\frac{2\pi}{3}\right)\in[0,\frac{4\pi}{3}]\times[0,\frac{4\pi}{3}] ( 3 2 π , 3 2 π ) ∈ [ 0 , 3 4 π ] × [ 0 , 3 4 π ] in the moment polytope of M M M . Under the affine-linear isomorphism of this square to Δ Q 3 ∩ { x 3 = 4 π 3 } \Delta_{Q^3}\cap\left\{x_3=\frac{4\pi}{3}\right\} Δ Q 3 ∩ { x 3 = 3 4 π } , the point ( 2 π 3 , 2 π 3 ) \left(\frac{2\pi}{3},\frac{2\pi}{3}\right) ( 3 2 π , 3 2 π ) maps to the point ( 2 π 3 , 2 π 3 , 4 π 3 ) ∈ Δ Q 3 \left(\frac{2\pi}{3},\frac{2\pi}{3},\frac{4\pi}{3}\right)\in\Delta_{Q^3} ( 3 2 π , 3 2 π , 3 4 π ) ∈ Δ Q 3 , the fibre of which is precisely P P P , which proves the claim. See also [7 Proposition 2.8] for details. □
The Ψ \Psi Ψ -polytope of P P P is (a translate of) Δ Q 3 \Delta_{Q^3} Δ Q 3 itself by Corollary 2.9. Since the Ψ \Psi Ψ -polytope of T C l T_{\mathrm{Cl}} T Cl is the square, this suggests to view Δ Q 3 \Delta_{Q^3} Δ Q 3 as the (truncated) cone over the square as follows
Δ Q 3 = Cone ( S ) = { t x ∣ t ∈ [ 0 , 1 ] , x ∈ S } , \Delta_{Q^3}=\operatorname{Cone}(S)=\{tx\mid t\in[0,1],x\in S\}, Δ Q 3 = Cone ( S ) = { t x ∣ t ∈ [ 0 , 1 ] , x ∈ S } ,
for S = [ 0 , 2 π ] × [ 0 , 2 π ] × { 2 π } S=[0,2\pi]\times[0,2\pi]\times\{2\pi\} S = [ 0 , 2 π ] × [ 0 , 2 π ] × { 2 π } . In other words, the Ψ \Psi Ψ -polytope of the lifted Lagrangian is obtained by taking the cone over the Ψ \Psi Ψ -polytope of the Lagrangian which is lifted.
Lemma 2.20. Let S k = ( 3 2 Δ k ) × { 2 π } S_k=(\frac{3}{2}\Delta_k)\times\{2\pi\} S k = ( 2 3 Δ k ) × { 2 π } . Up to translation, the Ψ \Psi Ψ -polytope of T ^ k V i a \widehat{T}^{\mathrm{Via}}_k T k Via is the truncated cone over S k S_k S k . More precisely,
Δ Ψ ( T ^ k V i a ) + ( 0 , 4 π 3 ) = Cone ( S k ) . (17) \Delta^\Psi(\widehat{T}^{\mathrm{Via}}_k)+(0,\tfrac{4\pi}{3})=\operatorname{Cone}(S_k). \tag*{(17)} Δ Ψ ( T k Via ) + ( 0 , 3 4 π ) = Cone ( S k ) . (17) In particular, the integral affine distance from the apex to the opposite facet is 2 π 2\pi 2 π .
Proof. A similar statement was proved in [8 ] for the displacement energy of the lift to C 3 \mathbb{C}^3 C 3 of Vianna tori in C P 2 \mathbb{C}P^2 C P 2 . Since the proof is a straightforward adaptation of the proof of [8 Proposition 1.6] , we only give an outline here and refer the reader to [8 ] for details.
An explicit versal deformation of T ^ k V i a ⊂ Q 3 \widehat{T}^{\mathrm{Via}}_k\subset Q^3 T k Via ⊂ Q 3 can be constructed from a versal deformation of the Vianna torus T k V i a T^{\mathrm{Via}}_k T k Via downstairs. On an open dense subset, the members of this deformation downstairs are Hamiltonian isotopic to ordinary toric fibres (see [8 §3.3] ). The lifts of these toric fibres are fibres of the Gelfand–Cetlin moment map μ : Q 3 → Δ Q 3 \mu:Q^3\to\Delta_{Q^3} μ : Q 3 → Δ Q 3 , of which we know the Ψ \Psi Ψ -invariant. Tracking which member of the versal deformation is Hamiltonian isotopic to which μ \mu μ -fibre yields the result on an open dense subset of a neighbourhood of the origin. We again use Proposition 2.2 to deduce that the result holds everywhere. □
We can now prove the main theorem of this subsection.
Proof of Theorem 2.15 . It suffices to prove that Δ Ψ ( P C h ) \Delta^{\Psi}(\mathcal{P}^{\mathrm{Ch}}) Δ Ψ ( P Ch ) is not integral affine equivalent to any of the Δ Ψ ( T k V i a ) \Delta^{\Psi}(\mathcal{T}^{\mathrm{Via}}_{k}) Δ Ψ ( T k Via ) . By Lemma 2.20, we know that the latter Ψ \Psi Ψ -polytopes are cones over the Δ k \Delta_k Δ k , meaning cones over triangles or over quadrilaterals.
If Δ k \Delta_k Δ k is a triangle, then its cone cannot be integral equivalent to Δ Ψ ( P C h ) \Delta^{\Psi}(\mathcal{P}^{\mathrm{Ch}}) Δ Ψ ( P Ch ) , since the latter contains a four-valent vertex;
If Δ k \Delta_k Δ k is a quadrilateral, then the integral affine distance of its four-valent vertex to the opposite facet is 2 π 2\pi 2 π . The integral affine distance of the four-valent vertex in Δ Ψ ( P C h ) \Delta^{\Psi}(\mathcal{P}^{\mathrm{Ch}}) Δ Ψ ( P Ch ) to its opposite facet is π \pi π . Since this is an integral affine invariant, the claim follows.
Enumerative properties of the OU-torus in Q3 ^{3} 3
In this section, we use a toric degeneration of the quadric Q 3 Q^{3} Q 3 to study the disk-counting potential of the Oakley–Usher torus P ⊂ Q 3 \mathcal{P} \subset Q^{3} P ⊂ Q 3 . Additionally, we apply this to prove a split-generation result for the monotone Fukaya category of Q 3 Q^{3} Q 3 and compare it to Abouzaid–Diogo [1 ] .
Superpotentials and non-displaceability
In [24 ] , the question whether the monotone OU-Lagrangian P k , m ( r ∗ ) ⊂ Q n \mathcal{P}_{k,m}(r_{*}) \subset Q^{n} P k , m ( r ∗ ) ⊂ Q n is displaceable for k ≥ 1 k \ge1 k ≥ 1 is raised. In this subsection, we give a negative answer for the monotone toric case P = P 1 , 1 ( 4 3 ) ⊂ Q 3 \mathcal{P}=\mathcal{P}_{1,1}\left(\frac{4}{3}\right)\subset Q^{3} P = P 1 , 1 ( 3 4 ) ⊂ Q 3 . To show this, we use a method to reconstruct the superpotential of P \mathcal{P} P from the versal deformation of its Ψ \Psi Ψ -invariant, and use it to show that this Lagrangian supports critical local systems, in the sense of Biran-Cornea [5 ] .
Throughout this section, we work with the universal Novikov field over C \mathbb{C} C with real exponents, defined as follows:
Λ : = { ∑ j = 0 ∞ a j T λ j ∣ a j ∈ C , λ j ∈ R , λ j → + ∞ } . (18) \Lambda:=\left\{\left.\sum_{j=0}^{\infty}a_jT^{\lambda_j}\ \right|\ a_j\in\mathbb{C},\ \lambda_j\in\mathbb{R},\ \lambda_j\to+\infty\right\}. \tag*{(18)} Λ := { j = 0 ∑ ∞ a j T λ j a j ∈ C , λ j ∈ R , λ j → + ∞ } . (18)
We use multiplicative notation for Novikov weights: a holomorphic disk class β \beta β contributes the factor T ω ( β ) T^{\omega(\beta)} T ω ( β ) . All Floer categories and vector-space identifications below use Z / 2 Z \mathbb{Z}/2\mathbb{Z} Z /2 Z gradings.
Note that P \mathcal{P} P is a spin manifold: any n n n -torus is parallelizable (as all Lie groups are), and therefore its second Stiefel–Whitney class w 2 ( P ) ∈ H 2 ( P ; Z 2 ) w_2(\mathcal{P})\in H^2(\mathcal{P};\mathbb{Z}_2) w 2 ( P ) ∈ H 2 ( P ; Z 2 ) vanishes. Denote by Spin ( P ) \operatorname{Spin}(\mathcal{P}) Spin ( P ) the set of its spin structures. Let ∂ : π 2 ( Q 3 , P ) → π 1 ( P ) \partial:\pi_2(Q^3,\mathcal{P})\to\pi_1(\mathcal{P}) ∂ : π 2 ( Q 3 , P ) → π 1 ( P ) be the connecting homomorphism from the homotopy long exact sequence of the pair ( Q 3 , P ) (Q^3,\mathcal{P}) ( Q 3 , P ) . Let S ∈ Spin ( P ) S\in\operatorname{Spin}(\mathcal{P}) S ∈ Spin ( P ) . We define the superpotential of ( P , S ) (\mathcal{P},S) ( P , S ) to be the map W ( P , S ) : Hom ( π 1 ( P ) , Λ × ) → Λ \mathcal{W}_{(\mathcal{P},S)}:\operatorname{Hom}(\pi_1(\mathcal{P}),\Lambda^\times)\to\Lambda W ( P , S ) : Hom ( π 1 ( P ) , Λ × ) → Λ ,
W ( P , S ) ( χ ) : = ∑ A ∈ π 2 ( Q 3 , P ) m ( A ) = 2 n A , S T ⟨ ω , A ⟩ χ ( ∂ A ) , (19) \mathcal{W}_{(\mathcal{P},S)}(\chi):=\sum_{\substack{A\in\pi_2(Q^3,\mathcal{P})\\m(A)=2}}n_{A,S}T^{\langle\omega,A\rangle}\chi(\partial A), \tag*{(19)} W ( P , S ) ( χ ) := A ∈ π 2 ( Q 3 , P ) m ( A ) = 2 ∑ n A , S T ⟨ ω , A ⟩ χ ( ∂ A ) , (19)
where n A , S ∈ Z n_{A,S}\in\mathbb{Z} n A , S ∈ Z are signed open disk counts: for generic compatible almost complex structures J J J , they count the number of J J J -holomorphic disks representing the class A A A , and whose boundary lies on P \mathcal{P} P and passes through a generic point of P \mathcal{P} P . The choice of spin structure determines the orientations of the moduli spaces of Maslov-2 disks, and hence the coefficients n A , S n_{A,S} n A , S .
By exactness, such a class lies in the image of π 2 ( Q 3 ) \pi_{2}(Q^{3}) π 2 ( Q 3 ) . Write j ∗ ( A ) = β j_{*}(A)=\beta j ∗ ( A ) = β , for A ∈ π 2 ( Q 3 ) A\in\pi_{2}(Q^{3}) A ∈ π 2 ( Q 3 ) . Its Maslov index is then
m ( β ) = 2 c 1 ( Q 3 ) ( A ) ∈ 2 N Q 3 Z = 6 Z . m(\beta)=2c_{1}(Q^{3})(A)\in2N_{Q^{3}}\mathbb{Z}=6\mathbb{Z}. m ( β ) = 2 c 1 ( Q 3 ) ( A ) ∈ 2 N Q 3 Z = 6 Z .
Therefore, the constant term vanishes, and the superpotential of the Oakley–Usher torus is of the form
W ( P , S ) ( x 1 , x 2 , x 3 ) = c 1 x 1 + c 2 x 2 + c 3 x 3 x 1 + c 4 x 3 x 2 + c 5 1 x 3 , \mathcal{W}_{(P,S)}(x_{1},x_{2},x_{3})=c_{1}x_{1}+c_{2}x_{2}+c_{3}\frac{x_{3}}{x_{1}}+c_{4}\frac{x_{3}}{x_{2}}+c_{5}\frac{1}{x_{3}}, W ( P , S ) ( x 1 , x 2 , x 3 ) = c 1 x 1 + c 2 x 2 + c 3 x 1 x 3 + c 4 x 2 x 3 + c 5 x 3 1 ,
for non-vanishing coefficients c j ∈ Λ × c_{j}\in\Lambda^{\times} c j ∈ Λ × . We compute the full potential in Proposition 3.5 using methods which are independent of § 2 .
We now introduce the notion and techniques of toric degenerations. They will be used to compute the superpotential of P \mathcal{P} P exactly, not just its support.
Definition 3.3. A toric degeneration of a symplectic manifold ( X , ω ) (X,\omega) ( X , ω ) is a flat family Π : X → C \Pi:\mathcal{X}\to\mathbb{C} Π : X → C , whose fibres X t : = Π − 1 ( t ) X_{t}:=\Pi^{-1}(t) X t := Π − 1 ( t ) satisfy the following:
For t ≠ 0 t\ne0 t = 0 , the fibre X t X_{t} X t is smooth and symplectomorphic to X X X
X 0 X_{0} X 0 is a singular toric variety
The fibres X t X_{t} X t are projective subvarieties of the same projective space, i.e. there is a morphism φ : X → C P N \varphi:\mathcal{X}\to\mathbb{CP}^{N} φ : X → CP N , such that for every t ∈ C t\in\mathbb{C} t ∈ C , φ ∣ X t : X t → C P N \left.\varphi\right|_{X_{t}}:X_{t}\to\mathbb{CP}^{N} φ ∣ X t : X t → CP N is an embedding.
Toric degenerations have the following useful feature: symplectic parallel transport along a path in the base induces symplectomorphisms between the regular fibres of the degeneration. For a real path, this transport can be described by the gradient-Hamiltonian flow introduced by Ruan [28 ] . When the path approaches the singular fibre X 0 X_{0} X 0 , the flow yields a symplectomorphism away from the vanishing locus, see for example [22 , 17 ] .
We now fix a toric degeneration Π : X → C \Pi:\mathcal{X}\to\mathbb{C} Π : X → C by setting
X = { ( [ z 0 : z 1 : z 2 : z 3 : z 4 ] , t ) ∈ C P 4 × C | t z 0 2 + z 1 2 + z 2 2 + z 3 2 + z 4 2 = 0 } , \mathcal{X}=\left\{\left([z_{0}:z_{1}:z_{2}:z_{3}:z_{4}],t\right)\in\mathbb{CP}^{4}\times\mathbb{C}\mathrel{}\middle|\mathrel{}tz_{0}^{2}+z_{1}^{2}+z_{2}^{2}+z_{3}^{2}+z_{4}^{2}=0\right\}, X = { ( [ z 0 : z 1 : z 2 : z 3 : z 4 ] , t ) ∈ CP 4 × C t z 0 2 + z 1 2 + z 2 2 + z 3 2 + z 4 2 = 0 } ,
and taking Π \Pi Π to be the projection to the C \mathbb{C} C -component. The fibre X 1 = Π − 1 ( 1 ) X_{1}=\Pi^{-1}(1) X 1 = Π − 1 ( 1 ) is by definition just the smooth quadric threefold Q 3 Q^{3} Q 3 . Using the gradient-Hamiltonian vector field and a path avoiding the origin, we find that all fibres X t ≠ 0 X_{t\ne0} X t = 0 are symplectomorphic to Q 3 Q^{3} Q 3 . The central fibre is
X 0 = { [ z 0 : z 1 : z 2 : z 3 : z 4 ] ∈ C P 4 | z 1 2 + z 2 2 + z 3 2 + z 4 2 = 0 } , X_{0}=\left\{[z_{0}:z_{1}:z_{2}:z_{3}:z_{4}]\in\mathbb{CP}^{4}\mathrel{}\middle|\mathrel{}z_{1}^{2}+z_{2}^{2}+z_{3}^{2}+z_{4}^{2}=0\right\}, X 0 = { [ z 0 : z 1 : z 2 : z 3 : z 4 ] ∈ CP 4 z 1 2 + z 2 2 + z 3 2 + z 4 2 = 0 } ,
which we can think of as the cone in C P 4 \mathbb{CP}^{4} CP 4 over the two-dimensional quadric in { z 0 = 0 } ⊂ C P 4 \{z_{0}=0\}\subset\mathbb{CP}^{4} { z 0 = 0 } ⊂ CP 4 . We will use the gradient-Hamiltonian parallel transport for the path γ : [ 0 , 1 ] → C \gamma:[0,1]\to\mathbb{C} γ : [ 0 , 1 ] → C defined by γ ( t ) = 1 − t \gamma(t)=1-t γ ( t ) = 1 − t . By [17 Theorem A] , we find that there is a continuous map ϕ : Q 3 = X 1 → X 0 \phi:Q^{3}=X_{1}\to X_{0} ϕ : Q 3 = X 1 → X 0 which restricts to a symplectomorphism
ϕ ∣ X 1 ∖ V : X 1 ∖ V → X 0 ∖ { p } , (23) \left.\phi\right|_{X_{1}\setminus V}:X_{1}\setminus V\to X_{0}\setminus\{p\}, \tag*{(23)} ϕ ∣ X 1 ∖ V : X 1 ∖ V → X 0 ∖ { p } , (23)
where p = [ 1 : 0 : 0 : 0 : 0 ] p=[1:0:0:0:0] p = [ 1 : 0 : 0 : 0 : 0 ] is the singular point of X 0 X_{0} X 0 and V = ϕ − 1 ( p ) V=\phi^{-1}(p) V = ϕ − 1 ( p ) is the vanishing locus of the degeneration Π \Pi Π with respect to the path γ \gamma γ . Let us compute the vanishing locus. Recall from (25) in [10 ] that we have denoted by Σ ⊂ Q 3 \Sigma\subset Q^{3} Σ ⊂ Q 3 the Lagrangian sphere obtained as the image of the zero section in T ∗ S 3 T^{*}S^{3} T ∗ S 3 under the symplectic cut. It is explicitly given by Σ = { [ 1 : i q ] ∈ C P 4 ∣ q ∈ S 3 } ⊂ Q 3 \Sigma=\{[1:iq]\in\mathbb{CP}^{4}\mid q\in S^{3}\}\subset Q^{3} Σ = {[ 1 : i q ] ∈ CP 4 ∣ q ∈ S 3 } ⊂ Q 3 . In the Gelfand–Cetlin fibration introduced in Proposition 2.6 , it appears as the singular fibre Σ = μ − 1 ( 0 ) \Sigma=\mu^{-1}(0) Σ = μ − 1 ( 0 ) over the origin.
Lemma 3.4. The vanishing locus V V V of the degeneration Π \Pi Π is Σ ⊂ Q 3 \Sigma\subset Q^3 Σ ⊂ Q 3 .
Proof. Equip C P 4 × C \mathbb{C}P^4 \times\mathbb{C} C P 4 × C , and thus by restriction X \mathcal{X} X , with a product Kähler form
ω X : = ω F S ⊕ i c 2 d t ∧ d t ˉ , c > 0. \omega_{\mathcal{X}} := \omega_{\mathrm{FS}} \oplus\frac{ic}{2}dt \wedge d\bar{t}, \qquad c > 0. ω X := ω FS ⊕ 2 i c d t ∧ d t ˉ , c > 0. We use an antisymplectic involution σ \sigma σ on X \mathcal{X} X which preserves the fibres over the real locus and whose fixed locus in the central fibre X 0 X_0 X 0 consists of the singular point p ∈ X 0 p \in X_0 p ∈ X 0 . The fixed locus is preserved by symplectic parallel transport along the path [ 0 , 1 ] ⊂ C [0,1] \subset\mathbb{C} [ 0 , 1 ] ⊂ C . This strategy is standard; see, for example, [16 Lemma 1.20] .
In our setup, the antisymplectic involution is given by
σ : X ⟶ X , ( [ z 0 : … : z 4 ] , t ) ⟼ ( [ z ˉ 0 : − z ˉ 1 : … : − z ˉ 4 ] , t ˉ ) . \sigma: \mathcal{X} \longrightarrow\mathcal{X}, \qquad([z_0:\ldots:z_4],t) \longmapsto([\bar{z}_0:-\bar{z}_1:\ldots:-\bar{z}_4],\bar{t}). σ : X ⟶ X , ([ z 0 : … : z 4 ] , t ) ⟼ ([ z ˉ 0 : − z ˉ 1 : … : − z ˉ 4 ] , t ˉ ) . Its fixed locus in X t X_t X t , for real t > 0 t>0 t > 0 , is
Fix ( σ ) ∩ X t = { [ 1 : i t q ] : q ∈ S 3 } . \operatorname{Fix}(\sigma) \cap X_t = \{[1:i\sqrt{t}\,q] : q \in S^3\}. Fix ( σ ) ∩ X t = {[ 1 : i t q ] : q ∈ S 3 } . These spheres collapse to p p p as t → 0 t \to0 t → 0 . Since Fix ( σ ) ∩ X 1 = Σ \operatorname{Fix}(\sigma) \cap X_1 = \Sigma Fix ( σ ) ∩ X 1 = Σ , we obtain Σ ⊆ V \Sigma\subseteq V Σ ⊆ V .
To prove the reverse inclusion, consider the S O ( 4 ) \mathrm{SO}(4) SO ( 4 ) -action on the last four homogeneous coordinates, and let μ S O ( 4 ) \mu_{\mathrm{SO}(4)} μ SO ( 4 ) denote its moment map. The action preserves the degeneration and the Kähler form. Moreover, its infinitesimal generators are tangent to the fibres of Π \Pi Π . Since the parallel-transport vector field is symplectically orthogonal to the fibres, μ S O ( 4 ) \mu_{\mathrm{SO}(4)} μ SO ( 4 ) is preserved by parallel transport.
The point p p p is fixed by S O ( 4 ) \mathrm{SO}(4) SO ( 4 ) , so μ S O ( 4 ) ( p ) = 0 \mu_{\mathrm{SO}(4)}(p)=0 μ SO ( 4 ) ( p ) = 0 . Consequently,
V ⊆ μ S O ( 4 ) − 1 ( 0 ) ∩ X 1 . V \subseteq\mu_{\mathrm{SO}(4)}^{-1}(0) \cap X_1. V ⊆ μ SO ( 4 ) − 1 ( 0 ) ∩ X 1 . We now show that this set is precisely Σ \Sigma Σ . Indeed, writing z ′ = ( z 1 , … , z 4 ) = x + i y z'=(z_1,\ldots,z_4)=x+iy z ′ = ( z 1 , … , z 4 ) = x + i y , the moment map vanishes exactly when x ∧ y = 0 x \wedge y=0 x ∧ y = 0 . In the affine chart z 0 = 1 z_0=1 z 0 = 1 , the equation of X 1 X_1 X 1 gives
x ⋅ y = 0 , ∣ y ∣ 2 − ∣ x ∣ 2 = 1. x \cdot y=0, \qquad|y|^2-|x|^2=1. x ⋅ y = 0 , ∣ y ∣ 2 − ∣ x ∣ 2 = 1. Since x ∧ y = 0 x \wedge y=0 x ∧ y = 0 , the vectors x x x and y y y are linearly dependent. Together with x ⋅ y = 0 x \cdot y=0 x ⋅ y = 0 , this forces x = 0 x=0 x = 0 and ∣ y ∣ = 1 |y|=1 ∣ y ∣ = 1 . Thus the zero set of the moment map in this chart is exactly
{ [ 1 : i y ] : y ∈ S 3 } = Σ . \{[1:iy]:y\in S^3\}=\Sigma. {[ 1 : i y ] : y ∈ S 3 } = Σ. On the hyperplane z 0 = 0 z_0=0 z 0 = 0 , the quadric equation instead gives
x ⋅ y = 0 , ∣ x ∣ 2 = ∣ y ∣ 2 . x \cdot y=0, \qquad|x|^2=|y|^2. x ⋅ y = 0 , ∣ x ∣ 2 = ∣ y ∣ 2 . As x x x and y y y cannot both vanish, these conditions are incompatible with x ∧ y = 0 x \wedge y=0 x ∧ y = 0 . Therefore, there are no additional zeros of the moment map.
We conclude that
μ S O ( 4 ) − 1 ( 0 ) ∩ X 1 = Σ , \mu_{\mathrm{SO}(4)}^{-1}(0) \cap X_1=\Sigma, μ SO ( 4 ) − 1 ( 0 ) ∩ X 1 = Σ , and hence V ⊆ Σ V \subseteq\Sigma V ⊆ Σ . This proves V = Σ V=\Sigma V = Σ . □ \square □
Let us point out that the antisymplectic involution σ \sigma σ is very natural. It extends the antisymplectic involution induced on Q 3 Q^3 Q 3 , after the symplectic cut, by the standard antisymplectic involution ( q , p ) ↦ ( q , − p ) (q,p)\mapsto(q,-p) ( q , p ) ↦ ( q , − p ) on T ∗ S 3 T^*S^3 T ∗ S 3 . We are now in a position to prove the first part of Theorem B by computing the superpotential of the Oakley–Usher torus.
Proposition 3.5. There exists a reference spin structure S 0 ∈ Spin ( P ) S_0 \in\operatorname{Spin}(P) S 0 ∈ Spin ( P ) such that the five Maslov–2 disk classes contributing to the superpotential all have algebraic count + 1 +1 + 1 . With respect to this spin structure,
W ( P , S 0 ) ( χ ) = T 2 π / 3 ( x 1 + x 2 + x 3 x 1 + x 3 x 2 + 1 x 3 ) . \mathcal{W}_{(P,S_0)}(\chi) = T^{2\pi/3}\left(x_1+x_2+\frac{x_3}{x_1}+\frac{x_3}{x_2}+\frac{1}{x_3}\right). W ( P , S 0 ) ( χ ) = T 2 π /3 ( x 1 + x 2 + x 1 x 3 + x 2 x 3 + x 3 1 ) .
Proof. To prove this result, we prove that Π : X → C \Pi\colon\mathcal{X} \to\mathbb{C} Π : X → C is a toric degeneration and compute its moment polytope. Since P P P will turn out to be the monotone fibre of the toric degeneration, we can apply Nishinou–Nohara–Ueda [23 Theorem 1.2] to compute its superpotential.
Recall that by [17 Theorem A] and Lemma 3.4 , we have found the symplectomorphism (23) . Since Q 3 ∖ Σ = X 1 ∖ Σ Q^3 \setminus\Sigma= X_1 \setminus\Sigma Q 3 ∖ Σ = X 1 ∖ Σ is toric by Proposition 2.6 , we can transport its toric structure by the map ϕ \phi ϕ to find that X 0 ∖ { p } X_0 \setminus\{p\} X 0 ∖ { p } carries a symplectic toric structure.
To verify that the central fibre X 0 X_0 X 0 is algebraically toric, we argue directly. Make a unitary linear change of the four coordinates z ′ = ( z 1 , z 2 , z 3 , z 4 ) z'=(z_1,z_2,z_3,z_4) z ′ = ( z 1 , z 2 , z 3 , z 4 ) to coordinates w 00 , w 10 , w 01 , w 11 w_{00},w_{10},w_{01},w_{11} w 00 , w 10 , w 01 , w 11 in which its equation is
w 00 w 11 = w 10 w 01 . w_{00}w_{11}=w_{10}w_{01}. w 00 w 11 = w 10 w 01 .
Thus X 0 X_0 X 0 is the projective cone over the Segre embedding C P 1 × C P 1 ⊂ C P 3 \mathbb{CP}^1 \times\mathbb{CP}^1 \subset\mathbb{CP}^3 CP 1 × CP 1 ⊂ CP 3 , and in particular is an algebraic toric variety. The dense algebraic torus acts with weights
0 , ( 0 , 0 , 1 ) , ( 1 , 0 , 1 ) , ( 0 , 1 , 1 ) , ( 1 , 1 , 1 ) 0,\qquad(0,0,1),\qquad(1,0,1),\qquad(0,1,1),\qquad(1,1,1) 0 , ( 0 , 0 , 1 ) , ( 1 , 0 , 1 ) , ( 0 , 1 , 1 ) , ( 1 , 1 , 1 )
on z 0 , w 00 , w 10 , w 01 , w 11 z_0,w_{00},w_{10},w_{01},w_{11} z 0 , w 00 , w 10 , w 01 , w 11 , respectively. Therefore, with the normalization in which a projective line has symplectic area 2 π 2\pi 2 π , its moment polytope is exactly the Gelfand–Cetlin polytope Δ Q 3 \Delta_{Q^3} Δ Q 3 .
It remains to identify the transported torus with the specific Oakley–Usher torus P P P . To this end, observe that the degeneration is SO ( 4 ) \operatorname{SO}(4) SO ( 4 ) -equivariant, where SO ( 4 ) \operatorname{SO}(4) SO ( 4 ) acts on the last four homogeneous coordinates. Since the symplectic parallel transport is defined using an SO ( 4 ) \operatorname{SO}(4) SO ( 4 ) -invariant Kähler form, it is SO ( 4 ) \operatorname{SO}(4) SO ( 4 ) -equivariant and preserves the corresponding moment map.
The two chosen rotation components of this moment map are precisely the last two components of μ 1 , 1 \mu_{1,1} μ 1 , 1 . Moreover, the norm of the angular-momentum bivector recovers its radial component. Indeed, in the cut coordinates
[ 1 − ∣ p ∣ 2 : p + i q ] , [\sqrt{1-|p|^2}:p+iq], [ 1 − ∣ p ∣ 2 : p + i q ] ,
the appropriately normalized norm is 2 π ∣ p ∣ 2\pi|p| 2 π ∣ p ∣ . On X 0 X_0 X 0 , writing z ′ = x + i y z'=x+iy z ′ = x + i y , the equation of the central fibre implies ∣ x ∣ = ∣ y ∣ |x|=|y| ∣ x ∣ = ∣ y ∣ and x ⋅ y = 0 x\cdot y=0 x ⋅ y = 0 . Consequently, the same normalized function is
2 π ∣ z ′ ∣ 2 ∣ z 0 ∣ 2 + ∣ z ′ ∣ 2 , 2\pi\frac{|z'|^2}{|z_0|^2+|z'|^2}, 2 π ∣ z 0 ∣ 2 + ∣ z ′ ∣ 2 ∣ z ′ ∣ 2 ,
which is the toric height coordinate. It follows that, under parallel transport, the effective toric moment map on X 0 X_0 X 0 agrees with M μ 1 , 1 M\mu_{1,1} M μ 1 , 1 , up to the fixed choice of signs and ordering of the two rotation components. In particular, the fibre with rotation components zero and height 4 π / 3 4\pi/3 4 π /3 is the image of the Oakley–Usher torus P = μ − 1 ( x P ) P=\mu^{-1}(x_P) P = μ − 1 ( x P ) , as in Proposition 2.6 . Thus ϕ ( P ) \phi(P) ϕ ( P ) is the monotone toric fibre over x P x_P x P in X 0 X_0 X 0 .
The singularity of X 0 X_0 X 0 at p p p is a conifold , or ordinary double point singularity. This follows from the toric normal form of Δ Q 3 \Delta_{Q^3} Δ Q 3 at the origin, see [15 Example 4.32] . In the same example, Evans discusses why this singularity admits a small resolution. More explicitly, a global small resolution of X 0 X_0 X 0 is obtained by recording one ruling of the rank-one 2 × 2 2 \times2 2 × 2 matrix ( w i j ) (w_{ij}) ( w ij ) . The singular point p p p is replaced by a copy of C P 1 \mathbb{CP}^{1} CP 1 , and no exceptional divisor is introduced.
Furthermore, X 0 X_0 X 0 is Gorenstein Fano. Indeed, since it is a normal quadric hypersurface in C P 4 \mathbb{CP}^{4} CP 4 , the adjunction formula gives
K X 0 = ( K C P 4 + O C P 4 ( 2 ) ) ∣ X 0 = O X 0 ( − 3 ) . K_{X_0} = (K_{\mathbb{CP}^{4}} + \mathcal{O}_{\mathbb{CP}^{4}}(2))|_{X_0} = \mathcal{O}_{X_0}(-3). K X 0 = ( K CP 4 + O CP 4 ( 2 )) ∣ X 0 = O X 0 ( − 3 ) .
In particular, K X 0 K_{X_0} K X 0 is Cartier and − K X 0 = O X 0 ( 3 ) -K_{X_0} = \mathcal{O}_{X_0}(3) − K X 0 = O X 0 ( 3 ) is ample.
We have therefore verified the hypotheses needed to apply [23 Theorem 1.2] . We choose the reference spin structure S 0 S_0 S 0 on P \mathcal{P} P by transporting the standard toric spin structure from the corresponding toric fibre of X 0 X_0 X 0 .
Since the central fibre X 0 X_0 X 0 is toric and admits a small resolution, the Maslov–2 potential of a Gelfand–Cetlin torus fibre in the smooth fibre is determined by the toric facets of the polytope associated to X 0 X_0 X 0 . More precisely, if L ( y ) = μ − 1 ( y ) \mathcal{L}(y)=\mu^{-1}(y) L ( y ) = μ − 1 ( y ) for y ∈ Int ( Δ Q 3 ) y \in\operatorname{Int}(\Delta_{Q_3}) y ∈ Int ( Δ Q 3 ) , then, with respect to the reference spin structure S 0 S_0 S 0 ,
W ( L ( y ) , S 0 ) ( x ) = ∑ j = 0 4 T ℓ j ( y ) x v j , \mathcal{W}_{(\mathcal{L}(y),S_0)}(x) = \sum_{j=0}^{4} T^{\ell_j(y)}x^{v_j}, W ( L ( y ) , S 0 ) ( x ) = j = 0 ∑ 4 T ℓ j ( y ) x v j ,
where v j v_j v j are the primitive inward normals to the five facets of Δ Q 3 \Delta_{Q_3} Δ Q 3 , and ℓ j ( y ) \ell_j(y) ℓ j ( y ) are the corresponding affine distances to the facets. For the monotone fibre P \mathcal{P} P , all five affine distances are equal to 2 π / 3 2\pi/3 2 π /3 . The corresponding exponent vectors are
( 1 , 0 , 0 ) , ( 0 , 1 , 0 ) , ( − 1 , 0 , 1 ) , ( 0 , − 1 , 1 ) , ( 0 , 0 , − 1 ) , (1,0,0), \quad(0,1,0), \quad(-1,0,1), \quad(0,-1,1), \quad(0,0,-1), ( 1 , 0 , 0 ) , ( 0 , 1 , 0 ) , ( − 1 , 0 , 1 ) , ( 0 , − 1 , 1 ) , ( 0 , 0 , − 1 ) ,
from which the desired expression of W ( P , S 0 ) \mathcal{W}_{(\mathcal{P},S_0)} W ( P , S 0 ) follows. □
For completeness, we discuss the effect that changing spin structure has on the superpotential. The set Spin ( P ) \operatorname{Spin}(\mathcal{P}) Spin ( P ) is an H 1 ( P ; Z 2 ) H^1(\mathcal{P};\mathbb{Z}_2) H 1 ( P ; Z 2 ) -torsor: for any given S 0 ∈ Spin ( P ) S_0 \in\operatorname{Spin}(\mathcal{P}) S 0 ∈ Spin ( P ) , any other spin structure is of the form
S 0 + η , η ∈ H 1 ( P ; Z 2 ) ≅ ( Z 2 ) 3 . S_0+\eta,\qquad\eta\in H^1(\mathcal{P};\mathbb{Z}_2)\cong(\mathbb{Z}_2)^3. S 0 + η , η ∈ H 1 ( P ; Z 2 ) ≅ ( Z 2 ) 3 .
Therefore, there are eight inequivalent spin structures on P \mathcal{P} P .
Corollary 3.6. For a general spin structure S = S 0 + η S=S_0+\eta S = S 0 + η , with η = ( η 1 , η 2 , η 3 ) ∈ H 1 ( P ; Z 2 ) \eta=(\eta_1,\eta_2,\eta_3)\in H^1(\mathcal{P};\mathbb{Z}_2) η = ( η 1 , η 2 , η 3 ) ∈ H 1 ( P ; Z 2 ) , the superpotential is
W ( P , S ) ( χ ) = T 2 π / 3 ( ( − 1 ) η 1 x 1 + ( − 1 ) η 2 x 2 + ( − 1 ) η 1 + η 3 x 3 x 1 + ( − 1 ) η 2 + η 3 x 3 x 2 + ( − 1 ) η 3 1 x 3 ) . \mathcal{W}_{(\mathcal{P},S)}(\chi)=T^{2\pi/3}\left((-1)^{\eta_1}x_1+(-1)^{\eta_2}x_2+(-1)^{\eta_1+\eta_3}\frac{x_3}{x_1}+(-1)^{\eta_2+\eta_3}\frac{x_3}{x_2}+(-1)^{\eta_3}\frac{1}{x_3}\right). W ( P , S ) ( χ ) = T 2 π /3 ( ( − 1 ) η 1 x 1 + ( − 1 ) η 2 x 2 + ( − 1 ) η 1 + η 3 x 1 x 3 + ( − 1 ) η 2 + η 3 x 2 x 3 + ( − 1 ) η 3 x 3 1 ) . Proof. It follows immediately from the fact that, if we choose the spin structure S = S 0 + η S=S_0+\eta S = S 0 + η , then the contribution of a disk class β \beta β is multiplied by ( − 1 ) ⟨ η , ∂ β ⟩ (-1)^{\langle\eta,\partial\beta\rangle} ( − 1 ) ⟨ η , ∂ β ⟩ . □
For the remainder of the paper, we work with the spin structure S 0 S_0 S 0 . For our superpotential W ( P , S 0 ) \mathcal{W}_{(\mathcal{P},S_0)} W ( P , S 0 ) , we consider its Jacobian ring:
Jac ( W ( P , S 0 ) ) : = Λ [ x 1 ± 1 , x 2 ± 1 , x 3 ± 1 ] ⟨ x i ∂ x i W ( P , S 0 ) ∣ i = 1 , 2 , 3 ⟩ \operatorname{Jac}(\mathcal{W}_{(\mathcal{P},S_0)}) := \frac{\Lambda[x_1^{\pm1},x_2^{\pm1},x_3^{\pm1}]}{\langle x_i\partial_{x_i}\mathcal{W}_{(\mathcal{P},S_0)} \mid i=1,2,3\rangle} Jac ( W ( P , S 0 ) ) := ⟨ x i ∂ x i W ( P , S 0 ) ∣ i = 1 , 2 , 3 ⟩ Λ [ x 1 ± 1 , x 2 ± 1 , x 3 ± 1 ]
Proposition 3.7. The critical locus of W ( P , S 0 ) \mathcal{W}_{(\mathcal{P},S_0)} W ( P , S 0 ) consists of exactly three nondegenerate points. In particular,
Jac ( W ( P , S 0 ) ) ≅ Λ ⊕ Λ ⊕ Λ (24) \operatorname{Jac}(\mathcal{W}_{(\mathcal{P},S_0)}) \cong\Lambda\oplus\Lambda\oplus\Lambda \tag*{(24)} Jac ( W ( P , S 0 ) ) ≅ Λ ⊕ Λ ⊕ Λ (24)
is semi-simple.
Proof. The critical-point equations are
{ x 1 − x 3 x 1 = 0 , x 2 − x 3 x 2 = 0 , x 3 x 1 + x 3 x 2 − 1 x 3 = 0. \begin{cases}
x_1-\frac{x_3}{x_1}=0,\\
x_2-\frac{x_3}{x_2}=0,\\
\frac{x_3}{x_1}+\frac{x_3}{x_2}-\frac{1}{x_3}=0.
\end{cases} ⎩ ⎨ ⎧ x 1 − x 1 x 3 = 0 , x 2 − x 2 x 3 = 0 , x 1 x 3 + x 2 x 3 − x 3 1 = 0.
The first two equations imply x 1 2 = x 2 2 = x 3 x_1^2=x_2^2=x_3 x 1 2 = x 2 2 = x 3 , and hence
( x 1 − x 2 ) ( x 1 + x 2 ) = 0. (x_1-x_2)(x_1+x_2)=0. ( x 1 − x 2 ) ( x 1 + x 2 ) = 0.
On the other hand, the third equation gives
x 1 + x 2 = x 1 x 2 x 3 2 . x_1+x_2=\frac{x_1x_2}{x_3^2}. x 1 + x 2 = x 3 2 x 1 x 2 .
Since the right-hand side is invertible in the Jacobian ring, so is x 1 + x 2 x_1+x_2 x 1 + x 2 . Consequently, x 1 = x 2 x_1=x_2 x 1 = x 2 holds in the Jacobian ring itself. Substituting x 2 = x 1 x_2=x_1 x 2 = x 1 and x 3 = x 1 2 x_3=x_1^2 x 3 = x 1 2 into the third equation yields an isomorphism
Jac ( W ( P , S 0 ) ) ≅ Λ [ x 1 ± 1 ] ⟨ x 1 3 − 1 2 ⟩ . \operatorname{Jac}(\mathcal{W}_{(P,S_0)})\cong\frac{\Lambda[x_1^{\pm1}]}{\langle x_1^3-\frac{1}{2}\rangle}. Jac ( W ( P , S 0 ) ) ≅ ⟨ x 1 3 − 2 1 ⟩ Λ [ x 1 ± 1 ] .
Since x 1 3 − 1 2 x_1^3-\frac{1}{2} x 1 3 − 2 1 has three distinct roots in C ⊂ Λ \mathbb{C}\subset\Lambda C ⊂ Λ , the Chinese remainder theorem gives
Jac ( W ( P , S 0 ) ) ≅ Λ ⊕ Λ ⊕ Λ . \operatorname{Jac}(\mathcal{W}_{(P,S_0)})\cong\Lambda\oplus\Lambda\oplus\Lambda. Jac ( W ( P , S 0 ) ) ≅ Λ ⊕ Λ ⊕ Λ.
In particular, the critical scheme consists of three reduced points, which are therefore nondegenerate. This also proves that the Jacobian ring is semisimple. □ \square □
Hence there exist three local systems χ λ ∈ M ( P ) \chi_\lambda\in\mathcal{M}(P) χ λ ∈ M ( P ) , parametrized by
χ λ ↦ ( x 1 , x 2 , x 3 ) = ( 2 − 1 / 3 λ , 2 − 1 / 3 λ , 2 − 2 / 3 λ 2 ) ∈ ( C × ) 3 , λ 3 = 1. (25) \chi_\lambda\mapsto(x_1,x_2,x_3)=\left(2^{-1/3}\lambda,2^{-1/3}\lambda,2^{-2/3}\lambda^2\right)\in(\mathbb{C}^{\times})^3,\qquad\lambda^3=1. \tag*{(25)} χ λ ↦ ( x 1 , x 2 , x 3 ) = ( 2 − 1/3 λ , 2 − 1/3 λ , 2 − 2/3 λ 2 ) ∈ ( C × ) 3 , λ 3 = 1. (25)
such that H F ∗ ( ( P , χ λ ) , ( P , χ λ ) ) ≠ 0 HF^*((P,\chi_\lambda),(P,\chi_\lambda))\ne0 H F ∗ (( P , χ λ ) , ( P , χ λ )) = 0 . In particular, P ⊂ Q 3 P\subset Q^3 P ⊂ Q 3 is non-displaceable, proving the first claim in Theorem B.
Homological mirror symmetry and split-generation
The preceding superpotential computations may appear ad hoc, but they admit a natural explanation via the mirror symmetry picture. In this subsection, we make this precise by relating our construction to the standard Landau–Ginzburg mirror of the quadric. After that, we show that the OU-Lagrangian P P P with its three critical local systems and the Lagrangian sphere Σ \Sigma Σ of Proposition 2.6 split-generate the monotone Fukaya category of Q 3 Q^3 Q 3 . Throughout, we denote by F ( M ) \mathcal{F}(M) F ( M ) the monotone Fukaya category of M M M , and by D π F ( M ) D^\pi\mathcal{F}(M) D π F ( M ) its split-closed derived category.
We begin on the quantum cohomology side (or A-side) of the correspondence. By Beauville’s computation of the quantum cohomology of complete intersections [4 ] , applied to the quadric hypersurface Q 3 ⊂ C P 4 Q^3\subset\mathbb{CP}^4 Q 3 ⊂ CP 4 , we have
Q H ∗ ( Q 3 ) ≅ Λ [ h ] ⟨ h 4 − 4 q h ⟩ , ∣ h ∣ = 2. (26) QH^*(Q^3)\cong\frac{\Lambda[h]}{\langle h^4-4qh\rangle},\qquad|h|=2. \tag*{(26)} Q H ∗ ( Q 3 ) ≅ ⟨ h 4 − 4 q h ⟩ Λ [ h ] , ∣ h ∣ = 2. (26)
Here h ∈ H 2 ( Q 3 ; Z ) h \in H^{2}(Q^{3};\mathbb{Z}) h ∈ H 2 ( Q 3 ; Z ) is the hyperplane class, and q : = T ⟨ ω , A ⟩ = T 2 π q := T^{\langle\omega,A\rangle} = T^{2\pi} q := T ⟨ ω , A ⟩ = T 2 π is the quantum parameter associated to the generator A A A of H 2 ( Q 3 ) H_{2}(Q^{3}) H 2 ( Q 3 ) .
We now define the quantum multiplication operator ( c 1 ( Q 3 ) ⋆ − ) ∈ End ( Q H ∗ ( Q 3 ) ) (c_{1}(Q^{3}) \star-) \in\operatorname{End}(QH^{*}(Q^{3})) ( c 1 ( Q 3 ) ⋆ − ) ∈ End ( Q H ∗ ( Q 3 )) , given by
α ↦ c 1 ( Q 3 ) ⋆ α = 3 ( h ⋆ α ) . \alpha\mapsto c_{1}(Q^{3}) \star\alpha= 3(h \star\alpha). α ↦ c 1 ( Q 3 ) ⋆ α = 3 ( h ⋆ α ) .
From the ring relation, this operator satisfies c 1 ( Q 3 ) 4 = 108 q c 1 ( Q 3 ) c_{1}(Q^{3})^{4} = 108q\,c_{1}(Q^{3}) c 1 ( Q 3 ) 4 = 108 q c 1 ( Q 3 ) , so its spectrum is
Spec ( c 1 ( Q 3 ) ⋆ − ) = { 0 } ⊔ { ( 108 q ) 1 / 3 λ ∣ λ 3 = 1 } . \operatorname{Spec}(c_{1}(Q^{3}) \star-) = \{0\} \sqcup\{(108q)^{1/3}\lambda\mid\lambda^{3}=1\}. Spec ( c 1 ( Q 3 ) ⋆ − ) = { 0 } ⊔ {( 108 q ) 1/3 λ ∣ λ 3 = 1 } .
It follows that Q H ∗ ( Q 3 ) QH^{*}(Q^{3}) Q H ∗ ( Q 3 ) is semi-simple and splits into four eigensummands
Q H ∗ ( Q 3 ) ≅ Λ e 0 ⊕ ⨁ λ 3 = 1 Λ e λ , QH^{*}(Q^{3}) \cong\Lambda e_{0} \oplus\bigoplus_{\lambda^{3}=1}\Lambda e_{\lambda}, Q H ∗ ( Q 3 ) ≅ Λ e 0 ⊕ λ 3 = 1 ⨁ Λ e λ ,
where
e 0 : = 1 − h 3 4 q , e λ : = h 3 + ( 4 q ) 1 / 3 λ h 2 + ( 4 q ) 2 / 3 λ 2 h 12 q . (27) e_{0} := 1-\frac{h^{3}}{4q}, \qquad e_{\lambda} := \frac{h^{3}+(4q)^{1/3}\lambda h^{2}+(4q)^{2/3}\lambda^{2}h}{12q}. \tag*{(27)} e 0 := 1 − 4 q h 3 , e λ := 12 q h 3 + ( 4 q ) 1/3 λ h 2 + ( 4 q ) 2/3 λ 2 h . (27)
It follows from direct calculation that these eigenvectors of the quantum multiplication operator are idempotent and
e 0 + ∑ λ 3 = 1 e λ = 1. e_{0}+\sum_{\lambda^{3}=1}e_{\lambda}=1. e 0 + λ 3 = 1 ∑ e λ = 1.
Let us recall the categorical meaning of these eigenvalues and eigenvectors. For a monotone symplectic manifold M M M , its monotone Fukaya category decomposes as a direct sum of summands indexed by the possible values of m 0 m_{0} m 0 , the curvature term of the A ∞ A_{\infty} A ∞ -algebra:
F ( M ) = ⨁ w λ F ( M ) w λ . \mathcal{F}(M)=\bigoplus_{w_{\lambda}}\mathcal{F}(M)_{w_{\lambda}}. F ( M ) = w λ ⨁ F ( M ) w λ .
Every summand containing a nonzero object is indexed by an eigenvalue of quantum multiplication by c 1 ( M ) c_{1}(M) c 1 ( M ) . The converse is not automatic for a general monotone M M M ; however, it is the case for M = Q 3 M=Q^{3} M = Q 3 , as we will show below.
Indeed, for an object ( L , ξ ) (L,\xi) ( L , ξ ) , the (length-zero) closed-open string map is a unital algebra homomorphism
C O ( L , ξ ) 0 : Q H ∗ ( M ) ⟶ H F ∗ ( ( L , ξ ) , ( L , ξ ) ) \mathcal{CO}^{0}_{(L,\xi)}:QH^{*}(M)\longrightarrow HF^{*}((L,\xi),(L,\xi)) C O ( L , ξ ) 0 : Q H ∗ ( M ) ⟶ H F ∗ (( L , ξ ) , ( L , ξ ))
satisfying
C O ( L , ξ ) 0 ( c 1 ( M ) ) = m 0 ( L , ξ ) 1 ( L , ξ ) . \mathcal{CO}^{0}_{(L,\xi)}(c_{1}(M))=m_{0}(L,\xi)1_{(L,\xi)}. C O ( L , ξ ) 0 ( c 1 ( M )) = m 0 ( L , ξ ) 1 ( L , ξ ) .
Thus, if L L L is a non-trivial object of F ( M ) \mathcal{F}(M) F ( M ) (that is, if H F ∗ ( ( L , ξ ) , ( L , ξ ) ) ≠ 0 HF^{*}((L,\xi),(L,\xi))\ne0 H F ∗ (( L , ξ ) , ( L , ξ )) = 0 ), the scalar m 0 ( L , ξ ) m_{0}(L,\xi) m 0 ( L , ξ ) must lie in the spectrum of c 1 ( M ) ⋆ − c_{1}(M)\star- c 1 ( M ) ⋆ − .
We describe the mirror-symmetry picture in terms of the closed-open string map. The local system chart associated to the monotone torus ( P , S 0 ) (P,S_{0}) ( P , S 0 ) gives a Landau–Ginzburg chart with superpotential W ( P , S 0 ) \mathcal{W}_{(P,S_{0})} W ( P , S 0 ) .
On the B-side, the appropriate category is the category of matrix factorizations:
⨁ λ 3 = 1 MF ( ( Λ × ) 3 , W ( P , S 0 ) − w λ ) . \bigoplus_{\lambda^{3}=1}\operatorname{MF}((\Lambda^{\times})^{3},\mathcal{W}_{(P,S_{0})}-w_{\lambda}). λ 3 = 1 ⨁ MF (( Λ × ) 3 , W ( P , S 0 ) − w λ ) .
Its three isolated critical points give three Jacobian factors. The category of matrix factorizations of the unshifted W \mathcal{W} W on its zero fibre is trivial, since all its critical values are nonzero.
Since Q H ∗ ( Q 3 ) QH^*(Q^3) Q H ∗ ( Q 3 ) is generated as a Λ \Lambda Λ -algebra by h = 1 3 c 1 ( Q 3 ) h=\frac{1}{3}c_1(Q^3) h = 3 1 c 1 ( Q 3 ) , the length-zero closed–open map associated with each critical local system satisfies
C O ( P , χ λ ) 0 ( f ( h ) ) = f ( w λ 3 ) 1 ( P , χ λ ) . CO^0_{(\mathcal{P},\chi_\lambda)}(f(h))=f\left(\frac{w_\lambda}{3}\right)1_{(\mathcal{P},\chi_\lambda)}. C O ( P , χ λ ) 0 ( f ( h )) = f ( 3 w λ ) 1 ( P , χ λ ) .
As H F ∗ ( ( P , χ λ ) , ( P , χ λ ) ) ≠ 0 HF^*((\mathcal{P},\chi_\lambda),(\mathcal{P},\chi_\lambda))\ne0 H F ∗ (( P , χ λ ) , ( P , χ λ )) = 0 , its unit identifies the scalar subalgebra Λ 1 ( P , χ λ ) \Lambda1_{(\mathcal{P},\chi_\lambda)} Λ 1 ( P , χ λ ) with Λ \Lambda Λ . Thus, the images of all elements of Q H ∗ ( Q 3 ) QH^*(Q^3) Q H ∗ ( Q 3 ) under these closed–open maps lie in canonically identified scalar subalgebras.
Composing the three resulting scalar characters with the inverse of the critical-point evaluation isomorphism
Jac ( W P ) ⟶ ≅ ∏ λ 3 = 1 Λ \operatorname{Jac}(W_{\mathcal{P}})\overset{\cong}{\longrightarrow}\prod_{\lambda^3=1}\Lambda Jac ( W P ) ⟶ ≅ λ 3 = 1 ∏ Λ
defines a unital algebra homomorphism
C O P 0 : Q H ∗ ( Q 3 ) ⟶ Jac ( W P ) . CO^0_{\mathcal{P}}:QH^*(Q^3)\longrightarrow\operatorname{Jac}(W_{\mathcal{P}}). C O P 0 : Q H ∗ ( Q 3 ) ⟶ Jac ( W P ) .
We refer to C O P 0 CO^0_{\mathcal{P}} C O P 0 as the combined scalar closed–open homomorphism. It satisfies
C O P 0 ( c 1 ( Q 3 ) ) = [ W P ] , CO^0_{\mathcal{P}}(c_1(Q^3))=[W_{\mathcal{P}}], C O P 0 ( c 1 ( Q 3 )) = [ W P ] ,
where [ W P ] [W_{\mathcal{P}}] [ W P ] is the class of the superpotential in its Jacobian ring.
Proposition 3.8. The following holds:
For all λ , ζ ∈ C \lambda,\zeta\in\mathbb{C} λ , ζ ∈ C with λ 3 = ζ 3 = 1 \lambda^3=\zeta^3=1 λ 3 = ζ 3 = 1 ,
C O ( P , χ λ ) 0 ( e ζ ) = δ λ , ζ 1 ( P , χ λ ) , (28) CO^0_{(\mathcal{P},\chi_\lambda)}(e_\zeta)=\delta_{\lambda,\zeta}1_{(\mathcal{P},\chi_\lambda)}, \tag*{(28)} C O ( P , χ λ ) 0 ( e ζ ) = δ λ , ζ 1 ( P , χ λ ) , (28) In particular, C O P 0 CO^0_{\mathcal{P}} C O P 0 is surjective.
Consequently, C O P 0 CO^0_{\mathcal{P}} C O P 0 descends to an isomorphism of rings
Q H ∗ ( Q 3 ) / Λ e 0 ≅ Jac ( W ( P , S 0 ) ) . (29) QH^*(Q^3)/\Lambda e_0\cong\operatorname{Jac}(\mathcal{W}_{(\mathcal{P},S_0)}). \tag*{(29)} Q H ∗ ( Q 3 ) /Λ e 0 ≅ Jac ( W ( P , S 0 ) ) . (29)
Proof. Recall that C O ( P , χ λ ) 0 ( h ) = ( 4 q ) 1 / 3 λ 1 ( P , χ λ ) CO^0_{(\mathcal{P},\chi_\lambda)}(h)=(4q)^{1/3}\lambda1_{(\mathcal{P},\chi_\lambda)} C O ( P , χ λ ) 0 ( h ) = ( 4 q ) 1/3 λ 1 ( P , χ λ ) , therefore
C O ( P , χ λ ) 0 ( e ζ ) = 1 3 ( 1 + λ ζ 2 + λ 2 ζ ) = { 1 , λ = ζ , 0 , λ ≠ ζ . CO^0_{(\mathcal{P},\chi_\lambda)}(e_\zeta)=\frac{1}{3}(1+\lambda\zeta^2+\lambda^2\zeta)=
\begin{cases}
1, & \lambda=\zeta,\\
0, & \lambda\ne\zeta.
\end{cases} C O ( P , χ λ ) 0 ( e ζ ) = 3 1 ( 1 + λ ζ 2 + λ 2 ζ ) = { 1 , 0 , λ = ζ , λ = ζ .
The second bullet follows from
C O P 0 ( e 0 ) = 1 − C O P 0 ( h ) 3 4 q = 0. CO^0_{\mathcal{P}}(e_0)=1-\frac{CO^0_{\mathcal{P}}(h)^3}{4q}=0. C O P 0 ( e 0 ) = 1 − 4 q C O P 0 ( h ) 3 = 0.
As C O P 0 CO^0_{\mathcal{P}} C O P 0 is a map of rings, the result follows from the first isomorphism theorem for rings. □ \square □
We see that the image of the closed-open string map detects exactly the nonzero part of quantum cohomology, while its kernel detects the remaining zero eigensummand. Thus W ( P , S 0 ) \mathcal{W}_{(\mathcal{P},S_0)} W ( P , S 0 ) should not be regarded as the full global Landau–Ginzburg mirror potential of Q 3 Q^3 Q 3 , but rather as the potential on the Laurent chart associated to P \mathcal{P} P . See [32 ] for related discussions. With our construction, we can give a simple geometric representative of the missing summand. Namely, the Gelfand–Cetlin fibre over the vertex ( 0 , 0 , 0 ) ∈ Δ Q 3 (0,0,0)\in\Delta_{Q^3} ( 0 , 0 , 0 ) ∈ Δ Q 3 is the Lagrangian sphere Σ = μ − 1 ( 0 ) ≅ S 3 \Sigma=\mu^{-1}(0)\cong S^3 Σ = μ − 1 ( 0 ) ≅ S 3 . The next Proposition shows that the rank-one summand not seen by the Laurent chart of P \mathcal{P} P is naturally accounted for by the singular Gelfand–Cetlin sphere fibre, and that these pieces split-generate the monotone Fukaya category of Q 3 Q^3 Q 3 .
Proposition 3.9. The Lagrangian sphere Σ \Sigma Σ split-generates the summand D π F ( Q 3 ) 0 D^\pi\mathcal{F}(Q^3)_0 D π F ( Q 3 ) 0 , and each ( P , χ λ ) (\mathcal{P},\chi_\lambda) ( P , χ λ ) split-generates the summand D π F ( Q 3 ) w λ D^\pi\mathcal{F}(Q^3)_{w_\lambda} D π F ( Q 3 ) w λ . Consequently,
D π F ( Q 3 ) = ⟨ Σ ⟩ s p l i t ⊕ ⨁ λ 3 = 1 ⟨ ( P , χ λ ) ⟩ s p l i t . D^\pi\mathcal{F}(Q^3)=\langle\Sigma\rangle_{\mathrm{split}}\oplus\bigoplus_{\lambda^3=1}\langle(\mathcal{P},\chi_\lambda)\rangle_{\mathrm{split}}. D π F ( Q 3 ) = ⟨ Σ ⟩ split ⊕ λ 3 = 1 ⨁ ⟨( P , χ λ ) ⟩ split . Proof. We first show that Σ \Sigma Σ defines a nonzero object in the zero eigensummand. Since π 1 ( Σ ) = π 2 ( Σ ) = 0 \pi_1(\Sigma)=\pi_2(\Sigma)=0 π 1 ( Σ ) = π 2 ( Σ ) = 0 , the long exact sequence of the pair gives π 2 ( Q 3 , Σ ) ≅ π 2 ( Q 3 ) \pi_2(Q^3,\Sigma)\cong\pi_2(Q^3) π 2 ( Q 3 , Σ ) ≅ π 2 ( Q 3 ) . Thus Σ \Sigma Σ has minimal Maslov number N Σ = 2 N Q 3 = 6 N_\Sigma=2N_{Q^3}=6 N Σ = 2 N Q 3 = 6 . In particular, there are no Maslov-2 disks, therefore m 0 ( Σ ) = 0 m_0(\Sigma)=0 m 0 ( Σ ) = 0 . Since 6 > dim Σ + 1 = 4 6>\dim\Sigma+1=4 6 > dim Σ + 1 = 4 , Oh’s monotone spectral sequence collapses [25 ] . Thus, as graded vector spaces,
H F ∗ ( Σ ) ≅ H ∗ ( Σ ; Λ ) ≠ 0. HF^*(\Sigma)\cong H^*(\Sigma;\Lambda)\ne0. H F ∗ ( Σ ) ≅ H ∗ ( Σ ; Λ ) = 0. Therefore Σ \Sigma Σ is a nonzero object in the zero eigensummand F ( Q 3 ) 0 \mathcal{F}(Q^3)_0 F ( Q 3 ) 0 .
For split-generation, recall that each piece of the semi-simple decomposition of Q H ∗ ( Q 3 ) QH^*(Q^3) Q H ∗ ( Q 3 ) has rank 1. As Σ \Sigma Σ is a non-trivial object of F ( Q 3 ) 0 \mathcal{F}(Q^3)_0 F ( Q 3 ) 0 , and each ( P , χ λ ) (\mathcal{P},\chi_\lambda) ( P , χ λ ) is a nontrivial object of F ( Q 3 ) w λ \mathcal{F}(Q^3)_{w_\lambda} F ( Q 3 ) w λ , the result follows immediately from [30 Corollary 1.12] . □ \square □
Remark 3.10. This result is the quadric analogue of the Abouzaid–Diogo generation theorem for the compact-monotone Fukaya category of T ∗ S 3 T^*S^3 T ∗ S 3 [1 Theorem 1.4] . In the next subsection, we recall their results and discuss how the compactification changes the categorical picture.
Remark 3.11. This remark was pointed out to us by David Keren Yaar. Since Q 3 Q^3 Q 3 contains two disjoint monotone spin Lagrangians with non-vanishing Floer cohomology and minimal Maslov number at least two, a result of Alizadeh–Atallah–Cant–Shang [3 Corollary 21] , applied to Σ \Sigma Σ and ( P , χ λ ) (\mathcal{P},\chi_\lambda) ( P , χ λ ) , implies that the spectral diameter is infinite, and hence so is the Hofer diameter of Ham ( Q 3 ) \operatorname{Ham}(Q^3) Ham ( Q 3 ) . This was already known by other methods [19 Theorem A] .
Corollary 3.12. For each λ 3 = 1 \lambda^3=1 λ 3 = 1 , let w λ : = W ( P , S 0 ) ( χ λ ) w_\lambda:=\mathcal{W}_{(\mathcal{P},S_0)}(\chi_\lambda) w λ := W ( P , S 0 ) ( χ λ ) . Then there is an equivalence of split-closed Z / 2 Z \mathbb{Z}/2\mathbb{Z} Z /2 Z -graded triangulated categories
D π F ( Q 3 ) w λ ≃ D π MF ( O ^ ( Λ × ) 3 , χ λ , W ( P , S 0 ) − w λ ) , D^\pi\mathcal{F}(Q^3)_{w_\lambda}\simeq D^\pi\operatorname{MF}\left(\widehat{\mathcal{O}}_{(\Lambda^\times)^3,\chi_\lambda},\mathcal{W}_{(\mathcal{P},S_0)}-w_\lambda\right), D π F ( Q 3 ) w λ ≃ D π MF ( O ( Λ × ) 3 , χ λ , W ( P , S 0 ) − w λ ) , under which ( P , χ λ ) (\mathcal{P},\chi_\lambda) ( P , χ λ ) corresponds to the stabilized residue field at χ λ \chi_\lambda χ λ . Moreover, the zero-eigensummand satisfies
D π F ( Q 3 ) 0 ≃ D π ( mod − C l 1 ) . D^\pi\mathcal{F}(Q^3)_0\simeq D^\pi(\operatorname{mod}-\mathrm{Cl}_1). D π F ( Q 3 ) 0 ≃ D π ( mod − Cl 1 ) . Proof. The first assertion follows from [32 Theorem C] and Propositions 3.7 and 3.9. Indeed, the critical points of W ( P , S 0 ) \mathcal{W}_{(\mathcal{P},S_0)} W ( P , S 0 ) are isolated, and each corresponding torus brane split-generates its eigensummand.
The second assertion follows from the description of the zero-eigensummand of the Fukaya category of an odd-dimensional quadric in [31 Corollary 4.9] . □ \square □
The preceding corollary gives a geometric realization of homological mirror symmetry for Q 3 Q^3 Q 3 in terms of the Oakley–Usher torus and the Lagrangian sphere Σ \Sigma Σ . Indeed, the three nonzero eigensummands are identified with the matrix-factorization categories at the three nondegenerate critical points of the superpotential W ( P , S 0 ) \mathcal{W}(P,S_0) W ( P , S 0 ) . The canonical Landau–Ginzburg mirror of Q 3 Q^3 Q 3 , defined in [27 ] has an additional nondegenerate critical point of critical value zero, which is not contained in the Laurent chart associated with P \mathcal{P} P . Its local matrix-factorization category is graded Morita equivalent to the category of modules over C l 1 \mathrm{Cl}_1 Cl 1 , and hence corresponds to the zero-eigensummand generated by Σ \Sigma Σ . Therefore, the four eigensummands of the Fukaya category are accounted for by explicit Lagrangian submanifolds.
Monotone Fukaya categories under compactification
At the beginning of this subsection, we recall the setup and split-generation result of [1 Theorem 1.4] for the compact-monotone Fukaya category of T ∗ S n T^*S^n T ∗ S n (denoted F ( T ∗ S n ) \mathcal{F}(T^*S^n) F ( T ∗ S n ) ), and its special case n = 3 n=3 n = 3 . After that, we compare with our result for D π F ( Q 3 ) D^\pi\mathcal{F}(Q^3) D π F ( Q 3 ) . We slightly adapt the notation and terminology for consistency with our previous exposition. We denote its split-closed derived category by D π F ( T ∗ S n ) D^\pi\mathcal{F}(T^*S^n) D π F ( T ∗ S n ) . Throughout, we assume n ≥ 3 n \ge3 n ≥ 3 odd. The even-dimensional case requires different graded modules and bounding-cochain data; see [1 Sections 3 and 6] .
We use the same Novikov field Λ \Lambda Λ as defined in (18) . It carries the non-Archimedean valuation
ν : Λ ⟶ R ∪ { + ∞ } , \nu: \Lambda\longrightarrow\mathbb{R} \cup\{+\infty\}, ν : Λ ⟶ R ∪ { + ∞ } ,
defined by
ν ( ∑ j = 0 ∞ a j T λ j ) = min { λ j ∣ a j ≠ 0 } , ν ( 0 ) = + ∞ . \nu\left(\sum_{j=0}^{\infty} a_j T^{\lambda_j}\right)=\min\{\lambda_j\mid a_j\ne0\},\qquad\nu(0)=+\infty. ν ( j = 0 ∑ ∞ a j T λ j ) = min { λ j ∣ a j = 0 } , ν ( 0 ) = + ∞.
Following [1 ] , we call an element unitary if it is invertible and has zero valuation, and denote the group of unitary Novikov units by
U Λ ∗ : = { U ∈ Λ × ∣ ν ( U ) = 0 } . U_\Lambda^*:=\{U\in\Lambda^\times\mid\nu(U)=0\}. U Λ ∗ := { U ∈ Λ × ∣ ν ( U ) = 0 } .
Thus a rank-one unitary local system on a Lagrangian L L L is equivalently a holonomy homomorphism
χ : H 1 ( L ) ⟶ U Λ ∗ . \chi:H_1(L)\longrightarrow U_\Lambda^*. χ : H 1 ( L ) ⟶ U Λ ∗ .
Let F ⊂ T ∗ S n F\subset T^*S^n F ⊂ T ∗ S n be a cotangent fiber, and consider its wrapped A ∞ A_\infty A ∞ -algebra A : = C W ∗ ( F , F ; Λ ) \mathcal{A}:=CW^*(F,F;\Lambda) A := C W ∗ ( F , F ; Λ ) . The A ∞ A_\infty A ∞ -Yoneda functor associates to every compact monotone brane L L L in F \mathcal{F} F the right A ∞ A_\infty A ∞ -module
L ⟼ C W ∗ ( F , L ; Λ ) , L\longmapsto CW^*(F,L;\Lambda), L ⟼ C W ∗ ( F , L ; Λ ) ,
over A \mathcal{A} A . The cohomology algebra of A \mathcal{A} A is
A : = H ∗ ( A ) = H W ∗ ( F , F ; Λ ) ≅ H − ∗ ( Ω S n , Λ ) ≅ Λ [ u ] , ∣ u ∣ = 1 − n . A:=H^*(\mathcal{A})=HW^*(F,F;\Lambda)\cong H_{-*}(\Omega S^n,\Lambda)\cong\Lambda[u],\qquad|u|=1-n. A := H ∗ ( A ) = H W ∗ ( F , F ; Λ ) ≅ H −∗ ( Ω S n , Λ ) ≅ Λ [ u ] , ∣ u ∣ = 1 − n .
Abouzaid–Diogo prove that A = Λ [ u ] A=\Lambda[u] A = Λ [ u ] is intrinsically formal [1 Proposition 5.2] , and moreover that every graded right A A A -module is intrinsically formal [1 Proposition 5.4] . Consequently, cohomology determines the objects of the relevant A ∞ A_\infty A ∞ -module category up to quasi-isomorphism. We use the cohomological interpretation of [1 Corollary 5.5] , with morphisms given by graded Ext groups, throughout the following module discussion.
In particular, the Yoneda image of a compact monotone brane L L L may be studied through the ordinary A A A -module
Y ( L ) : = H W ∗ ( F , L ; Λ ) . Y(L):=HW^*(F,L;\Lambda). Y ( L ) := H W ∗ ( F , L ; Λ ) .
Let mod p r ( A ) \operatorname{mod}_{\mathrm{pr}}(A) mod pr ( A ) denote the Z / 2 Z \mathbb{Z}/2\mathbb{Z} Z /2 Z -graded cohomological category of finite-dimensional graded A A A -modules, with morphisms Ext A ∗ ( M , N ) \operatorname{Ext}^{*}_{A}(M,N) Ext A ∗ ( M , N ) and Yoneda composition. Transport its triangulated structure from the derived A ∞ A_{\infty} A ∞ -module category. With this convention the cohomological Yoneda functor is fully faithful, and the generation theorem identifies split-closed triangulated categories:
D π F ( T ∗ S n ) ≃ mod p r ( A ) . D^{\pi}\mathcal{F}(T^{*}S^{n}) \simeq\operatorname{mod}_{\mathrm{pr}}(A). D π F ( T ∗ S n ) ≃ mod pr ( A ) .
Thus, the geometric split-generation problem may be studied entirely on the module side. Since Λ [ u ] \Lambda[u] Λ [ u ] is a PID and Λ \Lambda Λ is algebraically closed, the simple proper graded Λ [ u ] \Lambda[u] Λ [ u ] -modules, up to parity shift, are precisely
S a : = Λ [ u ] / ⟨ u − a ⟩ , a ∈ Λ . S_{a}:=\Lambda[u]/\langle u-a\rangle,\qquad a\in\Lambda. S a := Λ [ u ] / ⟨ u − a ⟩ , a ∈ Λ.
Every proper A A A -module has finite length and is built from these simple modules by extensions. Equivalently, the modules S a S_{a} S a generate mod p r ( A ) \operatorname{mod}_{\mathrm{pr}}(A) mod pr ( A ) [1 Corollary 6.5] . Therefore, in order to split-generate D π F ( T ∗ S 3 ) D^{\pi}\mathcal{F}(T^{*}S^{3}) D π F ( T ∗ S 3 ) , it suffices to realize all the modules S a S_{a} S a geometrically, possibly as direct summands of Yoneda images.
The zero-section contributes the uncountable collection of objects ( 0 S n , α [ p t ] ) (0_{S^{n}},\alpha[\mathrm{pt}]) ( 0 S n , α [ pt ]) , where α [ p t ] \alpha[\mathrm{pt}] α [ pt ] is the family of bounding cochains used in [1 ] , with ν ( α ) ≥ 0 \nu(\alpha)\geq0 ν ( α ) ≥ 0 (including α = 0 \alpha=0 α = 0 ). These objects have modules
Y ( 0 S n , α [ p t ] ) ≅ S α , (30) Y(0_{S^{n}},\alpha[\mathrm{pt}])\cong S_{\alpha}, \tag*{(30)} Y ( 0 S n , α [ pt ]) ≅ S α , (30)
thus producing all the nonnegative-valuation modules. Here and below, we choose the orientation of [ p t ] [\mathrm{pt}] [ pt ] and the sign of u u u so that the displayed identification holds.
Second, the negative valuation part is realized by the Oakley–Usher family P 0 , n − 1 ( r ) ⊂ T ∗ S n P_{0,n-1}(r)\subset T^{*}S^{n} P 0 , n − 1 ( r ) ⊂ T ∗ S n , indexed by the area-parameter r ∈ ( 0 , + ∞ ) r\in(0,+\infty) r ∈ ( 0 , + ∞ ) . Since H 1 ( P 0 , n − 1 ( r ) ) ≅ H 1 ( S 1 × S n − 1 ) ≅ Z H_{1}(P_{0,n-1}(r))\cong H_{1}(S^{1}\times S^{n-1})\cong\mathbb{Z} H 1 ( P 0 , n − 1 ( r )) ≅ H 1 ( S 1 × S n − 1 ) ≅ Z , a rank-one local system is determined by one holonomy U ∈ U Λ ∗ U\in U_{\Lambda}^{*} U ∈ U Λ ∗ . Let α : = T − r U − 1 ∈ Λ \alpha:=T^{-r}U^{-1}\in\Lambda α := T − r U − 1 ∈ Λ . On the level of Λ [ u ] \Lambda[u] Λ [ u ] -modules, the brane ( P 0 , n − 1 ( r ) , U ) (P_{0,n-1}(r),U) ( P 0 , n − 1 ( r ) , U ) produces the rank-two module
Λ [ u ] / ( u 2 − α ) ≅ S α ⊕ S − α . \Lambda[u]/(u^{2}-\alpha)\cong S_{\sqrt{\alpha}}\oplus S_{-\sqrt{\alpha}}. Λ [ u ] / ( u 2 − α ) ≅ S α ⊕ S − α .
Since U U U is unitary, we have ν ( α ) = − r < 0 \nu(\alpha)=-r<0 ν ( α ) = − r < 0 . Since n n n is odd, the two summands have the same parity, and the decomposition is compatible with the chosen Z / 2 Z \mathbb{Z}/2\mathbb{Z} Z /2 Z -grading; see [1 Lemma 4.13 and Corollary 4.14] .
Thus the zero-section sector covers ν ( a ) ≥ 0 \nu(a)\geq0 ν ( a ) ≥ 0 , while the P 0 , n − 1 ( r ) P_{0,n-1}(r) P 0 , n − 1 ( r ) -sector covers ν ( a ) < 0 \nu(a)<0 ν ( a ) < 0 . Together they produce all simple modules S a S_{a} S a .
For n = 3 n=3 n = 3 , the family P 0 , 2 ( r ) P_{0,2}(r) P 0 , 2 ( r ) can be replaced by the torus family P 1 , 1 ( r ) ≅ T 3 P_{1,1}(r)\cong T^{3} P 1 , 1 ( r ) ≅ T 3 . In a suitable basis of H 1 ( T 3 ) ≅ Z 3 H_{1}(T^{3})\cong\mathbb{Z}^{3} H 1 ( T 3 ) ≅ Z 3 (described in [1 Lemma 2.13] ), their superpotential is given by
W P 1 , 1 ( r ) o p e n ( z 1 , z 2 , z 3 ) = T r / 2 z 1 ( 1 + z 2 ) ( 1 + z 3 ) . (31) \mathcal{W}^{\mathrm{open}}_{P_{1,1}(r)}(z_{1},z_{2},z_{3})=T^{r/2}z_{1}(1+z_{2})(1+z_{3}). \tag*{(31)} W P 1 , 1 ( r ) open ( z 1 , z 2 , z 3 ) = T r /2 z 1 ( 1 + z 2 ) ( 1 + z 3 ) . (31)
This superpotential admits critical local systems, given by χ U = ( U , − 1 , − 1 ) \chi_{U}=(U,-1,-1) χ U = ( U , − 1 , − 1 ) for U ∈ U Λ ∗ U\in U_{\Lambda}^{*} U ∈ U Λ ∗ . The objects ( P 1 , 1 ( r ) , χ U ) (P_{1,1}(r),\chi_{U}) ( P 1 , 1 ( r ) , χ U ) have modules
Y ( P 1 , 1 ( r ) , χ U ) ≅ S σ β [ ϵ 1 ] ⊕ S σ β [ ϵ 2 ] , Y(P_{1,1}(r),\chi_{U})\cong S_{\sigma\beta}[\epsilon_{1}]\oplus S_{\sigma\beta}[\epsilon_{2}], Y ( P 1 , 1 ( r ) , χ U ) ≅ S σ β [ ϵ 1 ] ⊕ S σ β [ ϵ 2 ] ,
where σ ∈ { 1 , − 1 } \sigma\in\{1,-1\} σ ∈ { 1 , − 1 } and ϵ 1 , ϵ 2 ∈ Z / 2 Z \epsilon_{1},\epsilon_{2}\in\mathbb{Z}/2\mathbb{Z} ϵ 1 , ϵ 2 ∈ Z /2 Z . Since ν ( β ) = − r / 2 < 0 \nu(\beta)=-r/2<0 ν ( β ) = − r /2 < 0 , varying r > 0 r>0 r > 0 and U ∈ U Λ ∗ U\in U_{\Lambda}^{*} U ∈ U Λ ∗ realizes all the negative-valuation parameters, independently of the sign and parity shifts.
We have ν ( β ) = − r 2 < 0 \nu(\beta)=-\frac{r}{2}<0 ν ( β ) = − 2 r < 0 . As r > 0 r>0 r > 0 and U ∈ U Λ ∗ U\in U_{\Lambda}^{*} U ∈ U Λ ∗ vary, the resulting simple summands, allowing their parity shifts and the fixed sign convention, exhaust all parameters of negative valuation.
In this sense, the S 1 × S 2 S^1 \times S^2 S 1 × S 2 -family and the T 3 T^3 T 3 -family describe the same negative-valuation part of the module-theoretic picture.
Our compactification result has a different form. In Q 3 Q^3 Q 3 , the monotone Fukaya category decomposes according to the finite spectrum of quantum multiplication by c 1 ( Q 3 ) c_1(Q^3) c 1 ( Q 3 ) :
Spec ( c 1 ( Q 3 ) ⋆ − ) = { 0 } ⊔ { ( 108 q ) 1 / 3 λ ∣ λ 3 = 1 } . \operatorname{Spec}(c_1(Q^3) \star-) = \{0\} \sqcup\{(108q)^{1/3}\lambda\mid\lambda^3 = 1\}. Spec ( c 1 ( Q 3 ) ⋆ − ) = { 0 } ⊔ {( 108 q ) 1/3 λ ∣ λ 3 = 1 } .
The singular Gelfand–Cetlin fibre Σ ≃ S 3 \Sigma\simeq S^3 Σ ≃ S 3 has m 0 ( Σ ) = 0 m_0(\Sigma) = 0 m 0 ( Σ ) = 0 and generates the zero eigensummand. The monotone Oakley–Usher torus P \mathcal{P} P , equipped with its three critical local systems
( x 1 , x 2 , x 3 ) = ( 2 − 1 / 3 λ , 2 − 1 / 3 λ , 2 − 2 / 3 λ 2 ) , λ 3 = 1 , (x_1,x_2,x_3) = (2^{-1/3}\lambda,\,2^{-1/3}\lambda,\,2^{-2/3}\lambda^2), \qquad\lambda^3 = 1, ( x 1 , x 2 , x 3 ) = ( 2 − 1/3 λ , 2 − 1/3 λ , 2 − 2/3 λ 2 ) , λ 3 = 1 ,
generates the three nonzero eigensummands. We compare the torus family before and after compactification. In Q 3 Q^3 Q 3 , only the torus r = 4 π 3 r = \frac{4\pi}{3} r = 3 4 π is monotone, thus relevant for the monotone Fukaya category F ( Q 3 ) \mathcal{F}(Q^3) F ( Q 3 ) . Indeed, the four disk classes visible in T ∗ S 3 T^*S^3 T ∗ S 3 have area r 2 \frac{r}{2} 2 r , while the compactifying disk has area 2 π − r 2\pi-r 2 π − r , and monotonicity after compactification forces r = 4 π 3 r = \frac{4\pi}{3} r = 3 4 π . This allows us to study how the superpotential is affected by compactification. For the tori P 1 , 1 ( r ) ⊂ T ∗ S 3 \mathcal{P}_{1,1}(r) \subset T^*S^3 P 1 , 1 ( r ) ⊂ T ∗ S 3 , the superpotential is given by (31). The change of basis
B = ( 1 0 0 − 1 1 0 − 1 − 1 1 ) ∈ GL ( 3 ; Z ) B = \begin{pmatrix} 1 & 0 & 0 \\ -1 & 1 & 0 \\ -1 & -1 & 1 \end{pmatrix} \in\operatorname{GL}(3;\mathbb{Z}) B = 1 − 1 − 1 0 1 − 1 0 0 1 ∈ GL ( 3 ; Z )
induces the following monomial change of coordinates on the holonomy variables:
( z 1 , z 2 , z 3 ) = ( x 1 , x 2 x 1 , x 3 x 1 x 2 ) . (z_1,z_2,z_3) = \left(x_1,\frac{x_2}{x_1},\frac{x_3}{x_1x_2}\right). ( z 1 , z 2 , z 3 ) = ( x 1 , x 1 x 2 , x 1 x 2 x 3 ) .
In these variables, the superpotential of P \mathcal{P} P becomes
W P ( z 1 , z 2 , z 3 ) = T 2 π / 3 ( z 1 ( 1 + z 2 ) ( 1 + z 3 ) + 1 z 1 2 z 2 2 z 3 ) = W P 1 , 1 ( 4 π / 3 ) o p e n ( z 1 , z 2 , z 3 ) + T 2 π / 3 1 z 1 2 z 2 2 z 3 . \begin{aligned}
\mathcal{W}_{\mathcal{P}}(z_1,z_2,z_3) = T^{2\pi/3}\left(z_1(1+z_2)(1+z_3) + \frac{1}{z_1^2z_2^2z_3}\right) \\
&= \mathcal{W}^{\mathrm{open}}_{\mathcal{P}_{1,1}(4\pi/3)}(z_1,z_2,z_3) + T^{2\pi/3}\frac{1}{z_1^2z_2^2z_3}.
\end{aligned} W P ( z 1 , z 2 , z 3 ) = T 2 π /3 ( z 1 ( 1 + z 2 ) ( 1 + z 3 ) + z 1 2 z 2 2 z 3 1 ) = W P 1 , 1 ( 4 π /3 ) open ( z 1 , z 2 , z 3 ) + T 2 π /3 z 1 2 z 2 2 z 3 1 .
We see that compactification adds exactly a term to the open superpotential.
As a result, the critical equations change. The family of unitary critical local systems χ U \chi_U χ U is replaced by three critical local systems, which in these coordinates are
χ ~ λ = ( 2 − 1 / 3 λ , 1 , 1 ) , λ 3 = 1. \widetilde{\chi}_{\lambda} = (2^{-1/3}\lambda,1,1), \qquad\lambda^3 = 1. χ λ = ( 2 − 1/3 λ , 1 , 1 ) , λ 3 = 1.
Since these coordinates are in C × ⊂ Λ × \mathbb{C}^{\times} \subset\Lambda^{\times} C × ⊂ Λ × , they have valuation zero and thus χ ~ λ \widetilde{\chi}_{\lambda} χ λ are still unitary. Moreover, compactification changes the role of the ( S 1 × S 2 ) (S^1 \times S^2) ( S 1 × S 2 ) -generators. The monotone element of this family in Q 3 Q^3 Q 3 is P 0 , 2 ( 4 π 3 ) ⊂ Q 3 \mathcal{P}_{0,2}\left(\frac{4\pi}{3}\right) \subset Q^3 P 0 , 2 ( 3 4 π ) ⊂ Q 3 , and recall from [10 Proposition 3.16] that it is displaceable. Therefore, P 0 , 2 ( 4 π 3 ) ≃ 0 \mathcal{P}_{0,2}\left(\frac{4\pi}{3}\right) \simeq0 P 0 , 2 ( 3 4 π ) ≃ 0 in F ( Q 3 ) \mathcal{F}(Q^3) F ( Q 3 ) . Every admissible brane supported on this displaceable compactified Lagrangian has zero Floer cohomology. Thus it contributes no nonzero object after compactification, while suitable branes on the open family contribute to the cotangent-bundle generating collection.
Appendix A. Solid mutations (joint with Soham Chanda)
The goal of this appendix is to give an alternate description of the Chekanov Oakley–Usher Lagrangian in the quadric Q 3 Q^3 Q 3 by presenting it as a double solid torus mutation of the OU Lagrangian, in the spirit of [12 ] . This allows us to compute its disk-potential using wall-crossing results. This requires making a judicious choice of basis for the first homology. To this end, and to simplify our disk-potential computations, we would use the disk-potential computation of [1 ] . We again denote by P \mathcal{P} P and P C h \mathcal{P}^{\mathrm{Ch}} P Ch the Oakley–Usher and the Chekanov Oakley–Usher tori in Q 3 Q^3 Q 3 , respectively.
Solid-Torus Mutations and Disk-Potential
We begin with recalling the notion of solid-torus surgery and how such surgeries affect disk-potentials from [12 ] . Geometrically, solid-torus surgery is essentially a circle-fibered version of the disk surgery.
Definition A.1 (Solid Mutation Configuration). A pair ( L , T ) (L,\mathcal{T}) ( L , T ) of Lagrangians in a symplectic manifold M M M is a solid mutation configuration (SMC) if
L L L is a Lagrangian torus;
T \mathcal{T} T is a Lagrangian solid torus, i.e. T \mathcal{T} T is diffeomorphic to D × T n − 2 D \times\mathbb{T}^{n-2} D × T n − 2 where D D D denotes the closed two-dimensional disk;
L L L and T \mathcal{T} T intersect cleanly along the boundary of T \mathcal{T} T ;
the pair ( L , T ∩ L ) (L,\mathcal{T} \cap L) ( L , T ∩ L ) is diffeomorphic to the standard pair ( T n , T n − 1 ) (\mathbb{T}^n,\mathbb{T}^{n-1}) ( T n , T n − 1 ) .
Example A.2. From Equation (34) we see that when α ∈ { 0 , 1 4 } \alpha\in\left\{0,\frac{1}{4}\right\} α ∈ { 0 , 4 1 } , the image of I I I restricted to ( 0 , r ) × { α } × S 1 × S 1 (0,r) \times\{\alpha\} \times S^1 \times S^1 ( 0 , r ) × { α } × S 1 × S 1 has a solid-torus closure. We denote these solid-tori as T 0 \mathcal{T}_0 T 0 and T 1 4 \mathcal{T}_{\frac{1}{4}} T 4 1 . We thus have two disjoint solid tori T 0 \mathcal{T}_0 T 0 , T 1 4 \mathcal{T}_{\frac{1}{4}} T 4 1 such that ( P , T ∗ ) (\mathcal{P},\mathcal{T}_*) ( P , T ∗ ) form a solid mutation configuration. In terms of Figure 2 in [10 ] , these solid tori project to line segments, at angles 0 0 0 , π 4 \frac{\pi}{4} 4 π with the horizontal axis, joining the center with the red circle. In coordinates, we have the following presentations,
T 0 = { q = ( x , 0 ) , p = ( 0 , v ) : x ∈ S 1 , ∣ v ∣ ≤ 2 / 3 } , \mathcal{T}_0 = \{q=(x,0),\ p=(0,v): x\in S^1,\ |v|\leq2/3\}, T 0 = { q = ( x , 0 ) , p = ( 0 , v ) : x ∈ S 1 , ∣ v ∣ ≤ 2/3 } , T 1 / 4 = { q = ( 0 , y ) , p = ( − v , 0 ) : y ∈ S 1 , ∣ v ∣ ≤ 2 / 3 } . \mathcal{T}_{1/4} = \{q=(0,y),\ p=(-v,0): y\in S^1,\ |v|\leq2/3\}. T 1/4 = { q = ( 0 , y ) , p = ( − v , 0 ) : y ∈ S 1 , ∣ v ∣ ≤ 2/3 } .
An application of a generalized Weinstein neighborhood theorem [12 Lemma 2.3][18, Theorem A] states that any smooth identification of L ∪ T L\cup\mathcal{T} L ∪ T as a stratified space can be upgraded to a symplectomorphism from a neighborhood of L ∪ T L\cup\mathcal{T} L ∪ T . This allows one to identify a neighborhood of a solid mutation configuration with neighborhood of a standard local model that we describe now. Let X : = C 2 ∖ { x 1 x 2 = 1 } X := \mathbb{C}^2 \setminus\{x_1x_2=1\} X := C 2 ∖ { x 1 x 2 = 1 } and endow it with the Lefschetz fibration π : X → C ∖ { 1 } \pi: X \to\mathbb{C}\setminus\{1\} π : X → C ∖ { 1 } defined by π ( x ) = x 1 x 2 \pi(x)=x_1x_2 π ( x ) = x 1 x 2 . Given a simple loop γ \gamma γ in C ∖ { 0 , 1 } \mathbb{C}\setminus\{0,1\} C ∖ { 0 , 1 } define the Lagrangian
T γ : = { ( x , y ) ∈ C 2 | ∣ x ∣ = ∣ y ∣ , π ( x , y ) ∈ im ( γ ) } . (32) T_\gamma:= \left\{(x,y)\in\mathbb{C}^2 \middle| |x|=|y|,\ \pi(x,y)\in\operatorname{im}(\gamma)\right\}. \tag*{(32)} T γ := { ( x , y ) ∈ C 2 ∣ x ∣ = ∣ y ∣ , π ( x , y ) ∈ im ( γ ) } . (32)
The standard solid-mutation configuration in C n \mathbb{C}^n C n is ( T S 1 × T C l n − 2 , T [ 0 , 1 ] × T C l n − 2 ) (T_{S^1}\times\mathbb{T}^{n-2}_{\mathrm{Cl}},T_{[0,1]}\times\mathbb{T}^{n-2}_{\mathrm{Cl}}) ( T S 1 × T Cl n − 2 , T [ 0 , 1 ] × T Cl n − 2 ) where T C l n − 2 \mathbb{T}^{n-2}_{\mathrm{Cl}} T Cl n − 2 is the unit-radius Clifford torus in C n − 2 \mathbb{C}^{n-2} C n − 2 .
Definition A.3. Let ( L , T ) (L,\mathcal{T}) ( L , T ) be a solid mutation configuration and identify a neighborhood U U U of L ∩ T L\cap\mathcal{T} L ∩ T to U s t d U_{\mathrm{std}} U std , a neighborhood of the standard solid mutation configuration, via a symplectomorphism φ : U → U s t d ⊂ C n \varphi:U\to U_{\mathrm{std}}\subset\mathbb{C}^n φ : U → U std ⊂ C n such that φ ( L ) = T S 1 × T n − 2 \varphi(L)=T_{S^1}\times\mathbb{T}^{n-2} φ ( L ) = T S 1 × T n − 2 and φ ( T ) = T [ 0 , 1 ] × T n − 2 \varphi(\mathcal{T})=T_{[0,1]}\times\mathbb{T}^{n-2} φ ( T ) = T [ 0 , 1 ] × T n − 2 . A
solid mutation for the data ( L , T , φ ) (L,T,\varphi) ( L , T , φ ) replaces L L L with L μ : = φ − 1 ( T γ × T n − 2 ) L_{\mu}:=\varphi^{-1}(T_{\gamma}\times T^{n-2}) L μ := φ − 1 ( T γ × T n − 2 ) where γ \gamma γ is a curve in C ∖ [ 0 , 1 ] \mathbb{C}\setminus[0,1] C ∖ [ 0 , 1 ] . If L L L is monotone, one can choose a curve γ \gamma γ to ensure L μ L_{\mu} L μ is monotone. A solid mutation is also called a solid torus surgery.
When one performs a solid mutation of a monotone torus, the disk-potential change can be explicitly computed. We state the main wall-crossing result needed for our computation without proof.
Proposition A.4. [26],[12, Theorem 3.5] Select a basis ( x ^ , y ^ , w ^ 1 , … , w ^ n − 2 ) (\hat{x},\hat{y},\hat{w}_{1},\ldots,\hat{w}_{n-2}) ( x ^ , y ^ , w ^ 1 , … , w ^ n − 2 ) of the first homology of T S 1 × T n − 2 T_{S^{1}}\times T^{n-2} T S 1 × T n − 2 using the natural splitting H 1 ( T S 1 ) ⊕ H 1 ( T n − 2 ) H_{1}(T_{S^{1}})\oplus H_{1}(T^{n-2}) H 1 ( T S 1 ) ⊕ H 1 ( T n − 2 ) and x ^ , y ^ \hat{x},\hat{y} x ^ , y ^ are chosen such that x ^ \hat{x} x ^ is the fiber circle in the Lefschetz fibration and y ^ \hat{y} y ^ projects to a counter-clockwise oriented circle in the base. Then we have the following wall-crossing result for change of disk potentials under solid torus surgery,
W L ( x , y , w 1 , … , w n − 2 ) = W L μ ( x , y ( 1 + x ) , w 1 , … , w n − 2 ) . W_{L}(x,y,w_{1},\ldots,w_{n-2})=W_{L_{\mu}}(x,y(1+x),w_{1},\ldots,w_{n-2}). W L ( x , y , w 1 , … , w n − 2 ) = W L μ ( x , y ( 1 + x ) , w 1 , … , w n − 2 ) . Here we use x , y , … , w n − 2 x,y,\ldots,w_{n-2} x , y , … , w n − 2 to denote the monodromy around x ^ , y ^ , … w ^ n − 2 \hat{x},\hat{y},\ldots\hat{w}_{n-2} x ^ , y ^ , … w ^ n − 2 .
Comparison with the Abouzaid–Diogo model
We first relate the coordinates used above to the affine-quadric model used by Abouzaid–Diogo in [1 Section 2.2] . Recall that the symplectic cut identifies the unit co-disk bundle with the affine chart { Z 0 ≠ 0 } ⊂ Q 3 \{Z_{0}\ne0\}\subset Q^{3} { Z 0 = 0 } ⊂ Q 3 by
( q , p ) ↦ [ 1 − ∣ p ∣ 2 : p + i q ] . (q,p)\mapsto\left[\sqrt{1-|p|^{2}}:p+iq\right]. ( q , p ) ↦ [ 1 − ∣ p ∣ 2 : p + i q ] .
Thus, writing w j = Z j Z 0 w_{j}=\frac{Z_{j}}{Z_{0}} w j = Z 0 Z j , we obtain affine coordinates
w = p + i q 1 − ∣ p ∣ 2 . w=\frac{p+iq}{\sqrt{1-|p|^{2}}}. w = 1 − ∣ p ∣ 2 p + i q .
Since Q 3 = { [ Z 0 : ⋯ : Z 4 ] ∈ C P 4 ∣ Z 0 2 + ⋯ + Z 4 2 = 0 } Q^{3}=\{[Z_{0}:\cdots:Z_{4}]\in\mathbb{CP}^{4}\mid Z_{0}^{2}+\cdots+Z_{4}^{2}=0\} Q 3 = {[ Z 0 : ⋯ : Z 4 ] ∈ CP 4 ∣ Z 0 2 + ⋯ + Z 4 2 = 0 } , they satisfy ∑ i = 1 4 w i 2 = − 1 \sum_{i=1}^{4}w_{i}^{2}=-1 ∑ i = 1 4 w i 2 = − 1 . It is convenient to pass to the usual model of affine quadric
X 3 = { ( ξ 1 , … , ξ 4 ) ∈ C 4 ∣ ξ 1 2 + ⋯ + ξ 4 2 = 1 } X_{3}=\{(\xi_{1},\ldots,\xi_{4})\in\mathbb{C}^{4}\mid\xi_{1}^{2}+\cdots+\xi_{4}^{2}=1\} X 3 = {( ξ 1 , … , ξ 4 ) ∈ C 4 ∣ ξ 1 2 + ⋯ + ξ 4 2 = 1 }
by setting ξ j : = − i w j \xi_{j}:=-iw_{j} ξ j := − i w j . Abouzaid–Diogo use the equivalent affine model
X A D = { ( z , u 1 , v 1 , u 2 , v 2 ) ∈ C 5 ∣ u 1 v 1 = z + 1 , u 2 v 2 = z − 1 } . X_{\mathrm{AD}}=\{(z,u_{1},v_{1},u_{2},v_{2})\in\mathbb{C}^{5}\mid u_{1}v_{1}=z+1,\quad u_{2}v_{2}=z-1\}. X AD = {( z , u 1 , v 1 , u 2 , v 2 ) ∈ C 5 ∣ u 1 v 1 = z + 1 , u 2 v 2 = z − 1 } .
The identification between X 3 X_{3} X 3 and X A D X_{\mathrm{AD}} X AD is given by
u 1 = 2 ( ξ 1 + i ξ 2 ) , v 1 = 2 ( ξ 1 − i ξ 2 ) , u 2 = 2 ( ξ 3 + i ξ 4 ) , v 2 = − 2 ( ξ 3 − i ξ 4 ) , z = 2 ( ξ 1 2 + ξ 2 2 ) − 1 = 1 − 2 ( ξ 3 2 + ξ 4 2 ) . (33) \begin{aligned}
u_{1}&=\sqrt{2}(\xi_{1}+i\xi_{2}), & v_{1}&=\sqrt{2}(\xi_{1}-i\xi_{2}),\\
u_{2}&=\sqrt{2}(\xi_{3}+i\xi_{4}), & v_{2}&=-\sqrt{2}(\xi_{3}-i\xi_{4}),\\
z&=2(\xi_{1}^{2}+\xi_{2}^{2})-1=1-2(\xi_{3}^{2}+\xi_{4}^{2}). \tag*{(33)}
\end{aligned} u 1 u 2 z = 2 ( ξ 1 + i ξ 2 ) , = 2 ( ξ 3 + i ξ 4 ) , = 2 ( ξ 1 2 + ξ 2 2 ) − 1 = 1 − 2 ( ξ 3 2 + ξ 4 2 ) . v 1 v 2 = 2 ( ξ 1 − i ξ 2 ) , = − 2 ( ξ 3 − i ξ 4 ) , (33)
This is an isomorphism of affine quadrics.
The projection
z : X A D ⟶ C z:X_{\mathrm{AD}}\longrightarrow\mathbb{C} z : X AD ⟶ C
is the Morse–Bott Lefschetz-type fibration considered by Abouzaid–Diogo. Its critical values are ± 1 \pm1 ± 1 . For a regular value z ≠ ± 1 z\ne\pm1 z = ± 1 , they consider the two vanishing circles
V 1 , z = { u 1 v 1 = z + 1 , ∣ u 1 ∣ = ∣ v 1 ∣ } , V 2 , z = { u 2 v 2 = z − 1 , ∣ u 2 ∣ = ∣ v 2 ∣ } . V_{1,z}=\{u_{1}v_{1}=z+1,\ |u_{1}|=|v_{1}|\},\qquad V_{2,z}=\{u_{2}v_{2}=z-1,\ |u_{2}|=|v_{2}|\}. V 1 , z = { u 1 v 1 = z + 1 , ∣ u 1 ∣ = ∣ v 1 ∣ } , V 2 , z = { u 2 v 2 = z − 1 , ∣ u 2 ∣ = ∣ v 2 ∣ } .
For any curve C ⊂ C C \subset\mathbb{C} C ⊂ C , the Lagrangian T C ⊂ X A D T_C \subset X_{AD} T C ⊂ X A D is defined by
T C : = ⋃ z ∈ C V 1 , z × V 2 , z . T_C := \bigcup_{z \in C} V_{1,z} \times V_{2,z}. T C := z ∈ C ⋃ V 1 , z × V 2 , z .
Proposition A.5. We have the following identifications
P = T C O U , T 0 = T [ 1 , 13 5 ] , T 1 4 = T [ − 13 5 , − 1 ] , \mathcal{P}=T_{\mathrm{COU}},\qquad T_0=T_{[1,\frac{13}{5}]},\qquad T_{\frac{1}{4}}=T_{[-\frac{13}{5},-1]}, P = T COU , T 0 = T [ 1 , 5 13 ] , T 4 1 = T [ − 5 13 , − 1 ] , where C O U ⊂ C C_{\mathrm{OU}} \subset\mathbb{C} C OU ⊂ C is an ellipse.
Proof. This is a straightforward computation using the above. □ \square □
Proposition A.7. The Oakley-Usher torus P \mathcal{P} P and the solid tori T 0 , T 1 4 T_0,T_{\frac{1}{4}} T 0 , T 4 1 are equal to the Abouzaid-Diogo Lagrangian tori T C O U , T [ 1 , 13 5 ] T_{C_{\mathrm{OU}}},T_{[1,\frac{13}{5}]} T C OU , T [ 1 , 5 13 ] and T [ − 13 5 , − 1 ] T_{[-\frac{13}{5},-1]} T [ − 5 13 , − 1 ] where C O U C_{\mathrm{OU}} C OU is an ellipse. Moreover, the Chekanov Oakley-Usher torus P C h \mathcal{P}^{\mathrm{Ch}} P Ch is equal to a double solid torus surgery of T C O U T_{C_{\mathrm{OU}}} T C OU along the solid tori T [ 1 , 13 5 ] T_{[1,\frac{13}{5}]} T [ 1 , 5 13 ] and T [ − 13 5 , − 1 ] T_{[-\frac{13}{5},-1]} T [ − 5 13 , − 1 ] .
Proof. We start with directly checking that the Oakley–Usher parametrization is adapted to precisely the circles V ∗ , z V_{*,z} V ∗ , z . Let x = ( x 1 , x 2 ) x=(x_1,x_2) x = ( x 1 , x 2 ) , y = ( y 1 , y 2 ) ∈ S 1 ⊂ R 2 y=(y_1,y_2)\in S^1\subset\mathbb{R}^2 y = ( y 1 , y 2 ) ∈ S 1 ⊂ R 2 . For k = m = 1 k=m=1 k = m = 1 , the Oakley–Usher parametrization takes the form
q = ( cos ( 2 π α ) x , sin ( 2 π α ) y ) , p = ( − ρ sin ( 2 π α ) x , ρ cos ( 2 π α ) y ) . (34) q=(\cos(2\pi\alpha)x,\sin(2\pi\alpha)y),\qquad p=(-\rho\sin(2\pi\alpha)x,\rho\cos(2\pi\alpha)y). \tag*{(34)} q = ( cos ( 2 π α ) x , sin ( 2 π α ) y ) , p = ( − ρ sin ( 2 π α ) x , ρ cos ( 2 π α ) y ) . (34)
In particular, ∣ p ∣ = ρ |p|=\rho ∣ p ∣ = ρ . Set X : = x 1 + i x 2 X:=x_1+ix_2 X := x 1 + i x 2 , Y : = y 1 + i y 2 ∈ S 1 Y:=y_1+iy_2\in S^1 Y := y 1 + i y 2 ∈ S 1 , and
a ( ρ , t ) : = cos ( 2 π α ) + i ρ sin ( 2 π α ) 1 − ρ 2 , b ( ρ , t ) : = sin ( 2 π α ) − i ρ cos ( 2 π α ) 1 − ρ 2 . a(\rho,t):=\frac{\cos(2\pi\alpha)+i\rho\sin(2\pi\alpha)}{\sqrt{1-\rho^2}},\qquad b(\rho,t):=\frac{\sin(2\pi\alpha)-i\rho\cos(2\pi\alpha)}{\sqrt{1-\rho^2}}. a ( ρ , t ) := 1 − ρ 2 cos ( 2 π α ) + i ρ sin ( 2 π α ) , b ( ρ , t ) := 1 − ρ 2 sin ( 2 π α ) − i ρ cos ( 2 π α ) .
Then (33) becomes
u 1 = 2 a ( ρ , α ) X , v 1 = 2 a ( ρ , α ) X − 1 , u 2 = 2 b ( ρ , α ) Y , v 2 = − 2 b ( ρ , α ) Y − 1 . (35) \begin{aligned}
u_1&=\sqrt{2}a(\rho,\alpha)X,\qquad v_1=\sqrt{2}a(\rho,\alpha)X^{-1},\\
u_2&=\sqrt{2}b(\rho,\alpha)Y,\qquad v_2=-\sqrt{2}b(\rho,\alpha)Y^{-1}. \tag*{(35)}
\end{aligned} u 1 u 2 = 2 a ( ρ , α ) X , v 1 = 2 a ( ρ , α ) X − 1 , = 2 b ( ρ , α ) Y , v 2 = − 2 b ( ρ , α ) Y − 1 . (35)
In particular, ∣ u 1 ∣ = ∣ v 1 ∣ |u_1|=|v_1| ∣ u 1 ∣ = ∣ v 1 ∣ and ∣ u 2 ∣ = ∣ v 2 ∣ |u_2|=|v_2| ∣ u 2 ∣ = ∣ v 2 ∣ . Hence, for fixed ( ρ , α ) (\rho,\alpha) ( ρ , α ) , varying ( x , y ) ∈ S 1 × S 1 (x,y)\in S^1\times S^1 ( x , y ) ∈ S 1 × S 1 parametrizes exactly V 1 , z ( ρ , α ) × V 2 , z ( ρ , α ) V_{1,z(\rho,\alpha)}\times V_{2,z(\rho,\alpha)} V 1 , z ( ρ , α ) × V 2 , z ( ρ , α ) inside the fibre of the Abouzaid–Diogo fibration.
It remains to determine the corresponding point in the base. By direct computation, we obtain
z ( ρ , α ) = 1 + ρ 2 1 − ρ 2 cos ( 4 π α ) + i 2 ρ 1 − ρ 2 sin ( 4 π α ) . (36) z(\rho,\alpha)=\frac{1+\rho^2}{1-\rho^2}\cos(4\pi\alpha)+i\frac{2\rho}{1-\rho^2}\sin(4\pi\alpha). \tag*{(36)} z ( ρ , α ) = 1 − ρ 2 1 + ρ 2 cos ( 4 π α ) + i 1 − ρ 2 2 ρ sin ( 4 π α ) . (36)
For fixed ρ \rho ρ , the Oakley–Usher Lagrangian therefore projects to the ellipse
C ρ : = { 1 + ρ 2 1 − ρ 2 cos ( 4 π α ) + i 2 ρ 1 − ρ 2 sin ( 4 π α ) ∣ α ∈ R / Z } . C_\rho:=\left\{\frac{1+\rho^2}{1-\rho^2}\cos(4\pi\alpha)+i\frac{2\rho}{1-\rho^2}\sin(4\pi\alpha)\mathrel{\Big|}\alpha\in\mathbb{R}/\mathbb{Z}\right\}. C ρ := { 1 − ρ 2 1 + ρ 2 cos ( 4 π α ) + i 1 − ρ 2 2 ρ sin ( 4 π α ) α ∈ R / Z } .
For the monotone value ρ = 2 / 3 \rho=2/3 ρ = 2/3 , this is C O U : = { 13 5 cos ( 4 π α ) + i 12 5 sin ( 4 π α ) } C_{\mathrm{OU}}:=\left\{\frac{13}{5}\cos(4\pi\alpha)+i\frac{12}{5}\sin(4\pi\alpha)\right\} C OU := { 5 13 cos ( 4 π α ) + i 5 12 sin ( 4 π α ) } . Thus the monotone Oakley–Usher torus is precisely the torus obtained by taking V 1 , z × V 2 , z V_{1,z}\times V_{2,z} V 1 , z × V 2 , z over the ellipse C O U C_{\mathrm{OU}} C OU .
We can now identify the two solid mutation configurations. The first solid torus is given by
T 0 = S 1 = { α = 0 , 0 ≤ ρ ≤ 2 3 } . T_0=S_1=\left\{\alpha=0,\quad0\leq\rho\leq\frac{2}{3}\right\}. T 0 = S 1 = { α = 0 , 0 ≤ ρ ≤ 3 2 } .
It is immediately checked that z ( S 1 ) = [ 1 , 13 5 ] z(S_1)=[1,\frac{13}{5}] z ( S 1 ) = [ 1 , 5 13 ] . Similarly, the second solid torus is
T 1 4 = S 2 = { α = 1 4 , 0 ≤ ρ ≤ 2 3 } . T_{\frac14}=S_2=\left\{\alpha=\frac14,\quad0\leq\rho\leq\frac{2}{3}\right\}. T 4 1 = S 2 = { α = 4 1 , 0 ≤ ρ ≤ 3 2 } .
Its image under the projection is z ( S 2 ) = [ − 13 5 , − 1 ] z(S_2)=[-\frac{13}{5},-1] z ( S 2 ) = [ − 5 13 , − 1 ] .
We have therefore identified the two solid mutation configurations with the two natural thimbles of the Abouzaid–Diogo fibration cut off by the monotone Oakley–Usher ellipse:
S 1 = ⋃ z ∈ [ 1 , 13 / 5 ] V 1 , z × V 2 , z , S 2 = ⋃ z ∈ [ − 13 / 5 , − 1 ] V 1 , z × V 2 , z , (37) S_1=\bigcup_{z\in[1,13/5]}V_{1,z}\times V_{2,z},\qquad S_2=\bigcup_{z\in[-13/5,-1]}V_{1,z}\times V_{2,z}, \tag*{(37)} S 1 = z ∈ [ 1 , 13/5 ] ⋃ V 1 , z × V 2 , z , S 2 = z ∈ [ − 13/5 , − 1 ] ⋃ V 1 , z × V 2 , z , (37)
where at z = 1 z=1 z = 1 the second circle collapses and at z = − 1 z=-1 z = − 1 the first circle collapses.
On the other hand, the base change equation (36) is a diffeomorphism from ( 0 , 1 ) × ( R / ( 1 2 Z ) ) (0,1)\times(\mathbb{R}/(\frac12\mathbb{Z})) ( 0 , 1 ) × ( R / ( 2 1 Z )) to C ∖ [ − 1 , 1 ] \mathbb{C}\setminus[-1,1] C ∖ [ − 1 , 1 ] . A direct check shows that the curve defining the Chekanov Oakley–Usher torus is mapped to a curve C C h C_{\mathrm{Ch}} C Ch in the Abouzaid–Diogo base that does not contain the line segment [ − 1 , 1 ] [-1,1] [ − 1 , 1 ] in its interior; see Figure 3 .
Figure 3. The curve Γ C h \Gamma_{\mathrm{Ch}} Γ Ch in the Brendel–Piccirilli model, drawn in polar coordinates w = ρ e 2 π i α w=\rho e^{2\pi i\alpha} w = ρ e 2 π i α (left), and its image C C h C_{\mathrm{Ch}} C Ch in the Abouzaid–Diogo base under the coordinate change (36) (right). Matching colours identify corresponding pieces: the teal and purple arcs have constant ρ \rho ρ , while the green and orange pieces have constant α \alpha α . The critical values ± 1 \pm1 ± 1 lie outside the region bounded by C C h C_{\mathrm{Ch}} C Ch . The curves are schematic, with corners understood to be smoothed.
The last claim of the Proposition follows by noticing that the curve C C h C_{\mathrm{Ch}} C Ch can be Hamiltonian isotoped, with respect to a correct choice of symplectic form such that Hamiltonian isotopy of base induces Hamiltonian isotopy of the Lagrangian torus above it, to a curve that is realized by a double solid tori surgery of T C O U T_{COU} T C O U along the solid tori T [ 1 , 13 5 ] T_{[1,\frac{13}{5}]} T [ 1 , 5 13 ] and T [ − 13 5 , − 1 ] T_{[-\frac{13}{5},-1]} T [ − 5 13 , − 1 ] .
Disk-Potential Computation
Theorem A.8. The disk potential of the Chekanov O U OU O U torus is
W P C h ( z 1 , z 2 , z 3 ) = z 1 + ( z 2 + 1 ) 2 ( z 3 + 1 ) 2 z 1 2 z 2 z 3 . W_{\mathcal{P}^{\mathrm{Ch}}}(z_1,z_2,z_3)=z_1+\frac{(z_2+1)^2(z_3+1)^2}{z_1^2z_2z_3}. W P Ch ( z 1 , z 2 , z 3 ) = z 1 + z 1 2 z 2 z 3 ( z 2 + 1 ) 2 ( z 3 + 1 ) 2 . Proof. From Example A.2 we know that ( P , T 0 ) (\mathcal{P},T_0) ( P , T 0 ) and ( P , T 1 4 ) (\mathcal{P},T_{\frac{1}{4}}) ( P , T 4 1 ) form a solid mutation configuration. Moreover, from an application of the generalized Weinstein neighborhood theorem [18 ] and Figure 2 in [10 ] , one can check that the Chekanov O U OU O U torus P C h \mathcal{P}^{\mathrm{Ch}} P Ch is the iterated solid torus mutation of P \mathcal{P} P along the solid tori T 0 T_0 T 0 and T 1 4 T_{\frac{1}{4}} T 4 1 .
P → T 0 -mutation P K i m → T 1 4 -mutation P C h \mathcal{P}\xrightarrow{T_0\text{-mutation}}\mathcal{P}^{\mathrm{Kim}}\xrightarrow{T_{\frac{1}{4}}\text{-mutation}}\mathcal{P}^{\mathrm{Ch}} P T 0 -mutation P Kim T 4 1 -mutation P Ch The disk-potential of P \mathcal{P} P can be written, using the homology conventions set in Remark A.6, as
W P = z 1 ( z 2 + 1 ) ( z 3 + 1 ) + 1 z 1 2 z 2 z 3 . W_{\mathcal{P}}=z_1(z_2+1)(z_3+1)+\frac{1}{z_1^2z_2z_3}. W P = z 1 ( z 2 + 1 ) ( z 3 + 1 ) + z 1 2 z 2 z 3 1 . Using the mutation formula from Proposition A.4 and homology identifications in Remark A.6, we have
W P C h ( z 1 , z 2 , z 3 ) = W P ( z 1 ( z 2 + 1 ) ( z 3 + 1 ) , z 2 , z 3 ) = z 1 + ( z 2 + 1 ) 2 ( z 3 + 1 ) 2 z 1 2 z 2 z 3 . W_{\mathcal{P}^{\mathrm{Ch}}}(z_1,z_2,z_3)=W_{\mathcal{P}}\left(\frac{z_1}{(z_2+1)(z_3+1)},z_2,z_3\right)=z_1+\frac{(z_2+1)^2(z_3+1)^2}{z_1^2z_2z_3}. W P Ch ( z 1 , z 2 , z 3 ) = W P ( ( z 2 + 1 ) ( z 3 + 1 ) z 1 , z 2 , z 3 ) = z 1 + z 1 2 z 2 z 3 ( z 2 + 1 ) 2 ( z 3 + 1 ) 2 .
References
D-MATH, ETH Zürich, Rämistrasse 101, 8092 Zürich, Switzerland Email address: joe.brendel@math.ethz.ch
Institut de mathématiques, Université de Neuchâtel, Rue Emile-Argand 11, 2000 Neuchâtel, Switzerland Email address: andrea.piccirilli@unine.ch