ScalingStacks

3. Semi-flat SYZ [0204]

Original official author HTML, exact retained edition. Historical TeX conversion verdicts remain unchanged. Cited-edition alignment and mathematical self-containment are not assessed.

Complete original source context · Original author HTML

3. Semi-flat SYZ

In case the special Lagrangian torus fibrations do not admit any singular fibers, the SYZ picture is particularly nice. McLean’s classic results [126] give us two naturally defined integral affine structures11 1 An integral affine structure on a manifold is an atlas of charts whose transition maps are all integral affine linear transformations. on the base manifold BB of a special Lagrangian torus fibration: the symplectic and complex affine structures, and mirror symmetry can be explained neatly via these structures. More specifically, a normal vector field vv to a fiber Lb:=μ−1​(b)L_{b}:=\mu^{-1}(b) determines a 1-form α:=−ιv​ω∈Ω1​(Lb,ℝ)\alpha:=-\iota_{v}\omega\in\Omega^{1}(L_{b};\mathbb{R}) and an (n−1)(n-1)-form β:=ιv​Im ​Ω∈Ωn−1​(Lb,ℝ)\beta:=\iota_{v}\textrm{Im }\Omega\in\Omega^{n-1}(L_{b};\mathbb{R}), where ω\omega and Ω\Omega are the Kähler form and holomorphic volume form on XX respectively. McLean [126] proved that the corresponding deformation is special Lagrangian if and only if both α\alpha and β\beta are closed. By identifying T​BTB with H1​(Lb,ℝ)H^{1}(L_{b};\mathbb{R}) using the cohomology class of α\alpha, we get the symplectic affine structure on BB, while identifying T​BTB with Hn−1​(Lb,ℝ)H^{n-1}(L_{b};\mathbb{R}) using the cohomology class of β\beta gives us the complex affine structure on BB. We also have the McLean metric defined by

g(v1,v2):=−∫Lbιv1ω∧ιv2Im Ω.g(v_{1},v_{2}):=-\int_{L_{b}}\iota_{v_{1}}\omega\wedge\iota_{v_{2}}\textrm{Im }\Omega.

In his illuminating paper [86], Hitchin explains how these structures are all related through the Legendre transform. If we denote by x1,…,xnx_{1},\ldots,x_{n} the local affine coordinates on BB with respect to the symplectic affine structure, then locally the McLean metric can be written as the Hessian of a convex function ϕ\phi on BB, i.e. g⁡(∂∂xi,∂∂xj)=∂2ϕ∂xi​∂xjg\left(\frac{\partial}{\partial x_{i}},\frac{\partial}{\partial x_{j}}\right)=\frac{\partial^{2}\phi}{\partial x_{i}\partial x_{j}}. Furthermore, setting xˇi:=∂ϕ/∂xi\check{x}_{i}:=\partial\phi/\partial x_{i} (i=1,…,ni=1,\ldots,n) gives precisely the local affine coordinates on BB with respect to the complex affine structure, and if

ϕˇ:=∑i=1nxˇi​xi−ϕ⁡(x1,…,xn)\check{\phi}:=\sum_{i=1}^{n}\check{x}_{i}x_{i}-\phi(x_{1},\ldots,x_{n})

is the Legendre transform of ϕ\phi, then we have xi=∂ϕˇ/∂xˇix_{i}=\partial\check{\phi}/\partial\check{x}_{i} and g⁡(∂∂xˇi,∂∂xˇj)=∂2ϕˇ∂xˇi​∂xˇjg\left(\frac{\partial}{\partial\check{x}_{i}},\frac{\partial}{\partial\check{x}_{j}}\right)=\frac{\partial^{2}\check{\phi}}{\partial\check{x}_{i}\partial\check{x}_{j}}.

If additionally we assume that the fibration μ:X→B\mu:X\to B admits a Lagrangian section, then a theorem of Duistermaat [41] implies that there are global action-angle coordinates so that we can write

X=T∗​B/Λ∨,X=T^{*}B/\Lambda^{\vee},

where the lattice Λ∨⊂T∗​B\Lambda^{\vee}\subset T^{*}B is locally generated by d​x1,…,d​xndx_{1},\ldots,dx_{n}, and ω\omega can be identified with the canonical symplectic form

ω=∑i=1nd​xi∧d​ui\omega=\sum_{i=1}^{n}dx_{i}\wedge du_{i}

on T∗​B/Λ∨T^{*}B/\Lambda^{\vee}. Here u1,…,unu_{1},\ldots,u_{n} are the fiber coordinates on T∗​BT^{*}B.

In this case, the mirror of XX is simply given by

Xˇ:=T​B/Λ,\check{X}:=TB/\Lambda,

where the lattice Λ⊂T​B\Lambda\subset TB is locally generated by ∂/∂x1,…,∂/∂xn\partial/\partial x_{1},\ldots,\partial/\partial x_{n}. The quotient Xˇ\check{X} has a natural complex structure whose holomorphic coordinates are given by zi:=exp⁡(xi+𝐢​yi)z_{i}:=\exp(x_{i}+\mathbf{i}y_{i}), where y1,…,yny_{1},\ldots,y_{n} are the fiber coordinates on T​BTB dual to u1,…,unu_{1},\ldots,u_{n}. This constructs the mirror of XX as a complex manifold with a nowhere vanishing holomorphic volume form

Ωˇ:=d​log⁡z1∧⋯∧d​log⁡zn.\check{\Omega}:=d\log z_{1}\wedge\cdots\wedge d\log z_{n}.

Moreover, there is an explicit fiberwise Fourier–type transform, which we call the semi-flat SYZ transform ℱsemi-flat\mathcal{F}^{\textrm{semi-flat}}, that carries exp⁡𝐢​ω\exp\mathbf{i}\omega to Ωˇ\check{\Omega}; see [22, Section 2] for more details.

Now if we switch to the complex affine structure on BB, then we get a symplectic structure on Xˇ\check{X} which is compatible with its complex structure so that the mirror Xˇ\check{X} becomes a Kähler manifold. Furthermore, if the function ϕ\phi above satisfies the real Monge-Ampère equation

det​(∂2ϕ∂xi​∂xj)=constant,\textrm{det}\left(\frac{\partial^{2}\phi}{\partial x_{i}\partial x_{j}}\right)=\textrm{constant},

then we obtain TnT^{n}-invariant Ricci-flat metrics on both XX and its mirror Xˇ\check{X}. The induced metric on BB is called a Monge-Ampère metric and BB is called a Monge-Ampère manifold. This links mirror symmetry to the study of real Monge-Ampère equations and affine Kähler geometry, where Cheng and Yau had made substantial contributions [31, 32, 33] before even mirror symmetry was discovered. The construction of Monge-Ampère metrics on affine manifolds with singularities has since been an important question in both affine geometry and the study of the SYZ conjecture. The highly nontrivial works of Loftin, Yau and Zaslow [121, 122] constructed such metrics near the “Y” vertex, a typical type of singularity in the 3-dimensional case. But other than this, not much is known.

So the SYZ conjecture indeed paints an appealing picture for mirror symmetry in the semi-flat case; many more details on semi-flat SYZ mirror symmetry were worked out by Leung in [110]. Unfortunately, this nice picture can hold true only at the large complex structure/volume limits where all instanton corrections are suppressed. Away from the limits, special Lagrangian fibrations will have singular fibers and the mirror can no longer be obtained simply by dualizing a fibration.

Original mathematics by the credited authors. Source-backed reader collection; mathematical self-containment is not assessed.