The SYZ Program and Gross–Siebert Log / Toric Degenerations¶
Job: develop the Strominger–Yau–Zaslow conjecture (special Lagrangian torus fibrations, the dual affine structure, instanton corrections), the Gross–Siebert algebro-geometric reconstruction via toric degenerations / log geometry / scattering diagrams, and the cluster structure of the mirror — with the SYZ mirror of \(\mathbb{P}^2\) (the Hori–Vafa superpotential) as the worked example.
Core¶
The SYZ conjecture (Strominger–Yau–Zaslow)¶
Conjecture (SYZ). Near a large complex structure limit (LCSL) point of the complex-structure moduli of a Calabi–Yau \(n\)-fold \(X\), the manifold \(X\) admits a special Lagrangian torus fibration $$ f:X\to B $$ over an \(n\)-dimensional base \(B\) with a discriminant locus \(\Delta\subset B\) (codim \(\ge1\)), the generic fibers \(F_b=f^{-1}(b)\cong T^n\) for \(b\in B_0=B\setminus\Delta\) being special Lagrangian tori, and the mirror \(\check X\) is the dual torus fibration \(\check f:\check X\to B\) (the dual \(T^n\)-fibers over \(B_0\), with instanton / wall-crossing corrections near \(\Delta\)).
Special Lagrangian condition. A Lagrangian \(L\subset(X,\omega,\Omega)\) (CY with the holomorphic \(n\)-form \(\Omega\)) is special Lagrangian if $$ \omega|_L=0\quad\text{(Lagrangian)},\qquad \mathrm{Im}\,e^{i\theta}\Omega|_L=0\quad\text{(calibrated by }\mathrm{Re}\,e^{i\theta}\Omega\text{)}, $$ i.e. \(L\) is calibrated by \(\mathrm{Re}\,e^{i\theta}\Omega\) (Harvey–Lawson). The fibers \(F_b\) are volume-minimizing in their homology class (the calibrated geometry), and the SYZ duality is the metric duality of the special Lagrangian tori (the dual torus is the moduli of flat \(U(1)\) connections on \(F_b\), the Fourier–Mukai dual).
| SYZ data | Description |
|---|---|
| Base \(B\) | \(n\)-dim (real) base of the torus fibration; \(B_0=B\setminus\Delta\) the smooth locus |
| Discriminant \(\Delta\) | where the torus fibers degenerate (nodal / singular); codim \(\ge1\) |
| Affine structure | on \(B_0\): the action coordinates \(\int_{\gamma_i}\lambda\) (the symplectic action \(\lambda\) over a basis \(\gamma_i\) of \(\pi_1(F_b)\)) give \(B_0\) a torsion-free \(\mathrm{GL}(n,\mathbb{Z})\)-structure |
| Dual fibers | \(\check F_b=\mathrm{Hom}(\pi_1(F_b),U(1))\) (the dual torus, the moduli of flat \(U(1)\) local systems on \(F_b\)) |
| Instanton corrections | holomorphic disk instantons bounding \(F_b\) (near \(\Delta\)) correct the dual fibration — the wall-crossing |
The affine / integral affine structure¶
On the smooth locus \(B_0=B\setminus\Delta\), the action coordinates $$ y_i(b)=\int_{\gamma_i}\lambda\in\mathbb{R},\qquad \gamma_i\text{ a basis of }\pi_1(F_b)\cong\mathbb{Z}^n, $$ (the symplectic action \(\lambda\) — the Liouville form, well-defined up to exact — integrated over cycles in the fiber) give \(B_0\) an integral affine structure: the transition functions between the action-coordinate charts are in \(\mathrm{GL}(n,\mathbb{Z})\ltimes\mathbb{R}^n\) (the affine group, torsion-free). The angle coordinates \(\theta_i\) (the dual coordinates on the fiber, \(\theta_i(b,x)=\int_{\gamma_i}\lambda\) on the fiber) give the torus \(F_b\) its \(T^n\) structure, and the action-angle coordinates \((y,\theta)\) identify \(f^{-1}(U)\cong U\times T^n\) for small \(U\subset B_0\).
The mirror \(\check X\) over \(B_0\) is the dual torus fibration: the fibers \(\check F_b=\mathrm{Hom}(\pi_1(F_b),U(1))=T^n\) (the dual torus), with the dual affine structure (the affine structure on \(B_0\) from the dual action coordinates). Over \(B_0\), the SYZ duality is the Fourier–Mukai duality of the torus fibers. Near the discriminant \(\Delta\), the duality is corrected by holomorphic disk instantons (the wall-crossing), and the dual fibration is not the naive dual but the instanton-corrected one.
The Gross–Siebert program (log / toric degenerations)¶
The transcendental SYZ (special Lagrangian fibrations) is hard: the existence of the fibration is open in general. The Gross–Siebert program is the rigorous algebro-geometric reconstruction, replacing the special-Lagrangian picture by a toric degeneration.
A toric degeneration is a flat family \(\mathcal{X}\to\mathrm{Spec}\,R\) (over a DVR \(R\), say \(R=\mathbb{C}[[t]]\)) with generic fiber \(X_t\) the CY \(n\)-fold and special fiber \(X_0=\bigcup_i X_i\) a union of toric varieties (glued along toric divisors). The intersection complex \(B\) of \(X_0\) (the dual complex of the components, an integral affine manifold with singularities) replaces the SYZ base; the affine structure on \(B_0=B\setminus\Delta\) is the combinatorial one (the fan / polytope structure of the toric pieces), and the singularities (the discriminant) encode the degenerate fibers.
| Transcendental SYZ | Gross–Siebert algebraic |
|---|---|
| special Lagrangian fibration \(f:X\to B\) | toric degeneration \(\mathcal{X}\to\mathrm{Spec}\,R\) with special fiber \(X_0\) |
| base \(B\) with affine structure | intersection complex \(B\) of \(X_0\) (integral affine manifold with singularities) |
| discriminant \(\Delta\) (singular fibers) | singularities of \(B\) (the non-smooth locus of the affine structure) |
| instanton corrections (holomorphic disks) | scattering diagrams (the wall-crossing functions) |
| mirror \(\check X\) = dual fibration | mirror \(\check X\) reconstructed from \(B\) via log geometry + scattering |
| special Lagrangian condition | log-structure / Kato–Nakayama / Ogus log geometry |
The log-structure (Kato): the special fiber \(X_0\) with its reduced structure carries a canonical log-structure (the divisorial log-structure from the components); the Kato–Nakayama real locus \((X_0)_{\log}^{\mathrm{top}}\) (the real blow-up / the skeleton) is the SYZ base \(B\) with the affine structure. The mirror \(\check X\) is reconstructed from \(B\) by the log-geometric gluing: the toric pieces of \(X_0\) dualize to toric pieces of \(\check X_0\), glued by the scattering data, and the generic fiber \(\check X_t\) is the SYZ mirror.
Scattering diagrams and the cluster structure¶
A scattering diagram \(\mathfrak{D}\) over the affine manifold \(B\) is a collection of walls — codimension-1 polyhedral cells \(\mathfrak{w}\subset B\) (with rational / integral slopes, the "support" of the wall) — each equipped with an attached wall-crossing function $$ f_\mathfrak{w}\in\mathbb{C}[x_1^{\pm},\dots,x_n^{\pm}]\quad(\text{the monoid algebra of the dual torus}), $$ encoding the instanton correction across the wall (the wall-crossing: a path crossing \(\mathfrak{w}\) acts by \(f_\mathfrak{w}\) on the local system / the Floer data). The consistency condition: for any path in \(B_0\) around the singularities, the path-ordered product of the wall-crossing functions is trivial (the wall-crossing is consistent — the mirror is well-defined). The scattering diagram is consistent if this holds.
Theorem (Gross–Siebert; Kontsevich–Soibelman). The consistent scattering diagrams over \(B\) produce the cluster algebra of the mirror: the theta functions / the canonical basis of the cluster algebra are the broken lines (the piecewise-linear paths in \(B\) refracted by the walls, the instanton-corrected geodesics), and the cluster transformations are the wall-crossing functions. The mirror coordinate ring is the algebra of theta functions \(\mathbb{C}[\check X]=\bigoplus\vartheta_m\) (the Gross–Hacking–Keel canonical basis).
| Scattering diagram | Cluster algebra |
|---|---|
| walls \(\mathfrak{w}\) (codim-1 cells) | cluster \(\mathcal{A}\)-transformations (the mutations) |
| wall-crossing function \(f_\mathfrak{w}\) | the cluster mutation \(\mu_k\) (the rational transformation) |
| consistency (path-ordered product trivial) | the cluster Poisson-commutation / the Laurent phenomenon |
| broken lines (refracted paths) | theta functions \(\vartheta_m\) (the canonical basis) |
| the support of \(\mathfrak{D}\) | the scattering fan / the cluster fan |
The connection to the Fock–Goncharov cluster / higher Teichmüller: the same cluster structure (the theta functions / the scattering diagrams) appears on the Hitchin / mirror moduli — the Fock–Goncharov positive structure on \(\mathrm{Rep}(\pi_1 X,G)\) is the cluster structure of the SYZ mirror of the Hitchin system (see Hitchin & Geometric Langlands and higher Teichmüller frontiers).
The log Calabi–Yau and the mirror reconstruction¶
A log Calabi–Yau variety is a pair \((X,D)\) with \(D\) an anticanonical divisor (a normal crossings / toric boundary divisor, \(D\in|-K_X|\)); the log-structure from \(D\) makes \((X,D)\) a log-CY, the algebro-geometric avatar of the SYZ picture. The mirror is reconstructed via the punctured Gromov–Witten invariants of \((X,D)\) (the Gross–Siebert / Gross–Pandharipande–Siebert logarithmic GW theory).
Theorem (Gross–Siebert mirror theorem). The mirror family \(\check X\to\mathrm{Spec}\,R\) is reconstructed from the log GW invariants of \((X,D)\) (the punctured invariants counting curves meeting \(D\) with prescribed tangencies), and the mirror map / the Yukawa coupling is the log-period computation (the periods of the log-structure). The scattering diagrams encode the wall-crossing / the instanton corrections to the mirror family.
See it¶
SYZ mirror of the elliptic curve / the log CY \((\mathbb{P}^2,D)\).
(a) Elliptic curve. The elliptic curve \(E\) (the mirror of a symplectic \(T^2\)) viewed as a SYZ fibration \(f:E\to S^1\) over \(S^1\) with fiber \(S^1\) (the special Lagrangian fibration, no discriminant on \(B_0=S^1\)): the dual fibration is the dual torus \(\check E\to S^1\) (the mirror elliptic curve, the Fourier–Mukai dual). The SYZ duality over \(B_0\) is exact (no instanton corrections, no discriminant).
(b) Log CY \((\mathbb{P}^2,D)\) with \(D\) the toric boundary (three lines). The Gross–Siebert scattering diagram over the dual intersection complex \(B\) (the triangle, the moment polytope of \(\mathbb{P}^2\)) has walls (the legs / rays from the vertices of the triangle, the singularities of the affine structure at the vertices) with attached functions $$ f_\mathfrak{w}=(1+z)^{n_w}\quad\text{with }n_w=#{\text{holomorphic disks of class }\beta_w\text{ bounding the torus fiber}} $$ (the disk counts \(n_w\) — the Auroux superpotential coefficients, here \(n_w=1\) for the three coordinate-line disks on \(\mathbb{P}^2\)). The consistent scattering (the path-ordered product around the vertices is trivial, a non-trivial constraint fixing the wall functions) reconstructs the mirror as the Hori–Vafa superpotential $$ \check Y=(\mathbb{C}^)^2,\qquad W(x,y)=x+y+\frac{1}{xy}, $$ (the LG mirror of \(\mathbb{P}^2\) — the sum over the three toric boundary disks, with the cluster structure the theta functions of the 2D scattering). The log GW invariants of \((\mathbb{P}^2,D)\) (the punctured invariants, the counts of rational curves meeting \(D\) with prescribed tangencies) give the *mirror map** (the identification of the Kähler parameter of \(\mathbb{P}^2\) with the log-period of the mirror), and the Yukawa coupling / the quantum cohomology of \(\mathbb{P}^2\) is recovered from the log-period computation. This is the operational SYZ / Gross–Siebert: the mirror is built from the scattering (the instanton-corrected dual fibration), and the log GW invariants of \((\mathbb{P}^2,D)\) give the mirror map — the algebro-geometric realization of the SYZ duality, avoiding the transcendental special-Lagrangian fibration.
Check¶
Why is the SYZ duality exact over \(B_0=B\setminus\Delta\) but requires instanton corrections near the discriminant \(\Delta\)?
Over \(B_0\) the torus fibers are smooth special Lagrangian \(T^n\) and the SYZ duality is the Fourier–Mukai duality of the flat tori (the dual torus = the moduli of flat \(U(1)\) local systems, a metric duality, no corrections). Near \(\Delta\) the fibers degenerate (nodal / singular), and the naive dual fibration has singularities that are corrected by holomorphic disk instantons bounding the singular fibers — these disks contribute to the superpotential / the Floer products and change the dual fibration (the wall-crossing). The corrected dual is the Gross–Siebert / scattering-corrected SYZ mirror; without the corrections, the naive dual has the wrong complex structure (it misses the instanton contributions).
Verify that the Hori–Vafa superpotential \(W=x+y+1/(xy)\) on \((\mathbb{C}^*)^2\) is the sum over the three toric boundary disks of \(\mathbb{P}^2\).
The toric boundary \(D=D_1+D_2+D_3\) of \(\mathbb{P}^2\) (the three coordinate lines \(\{x_i=0\}\)) gives three disk classes \(\beta_1,\beta_2,\beta_3\) (the disks bounded by the Lagrangian torus fiber, each meeting one boundary component once). The Auroux superpotential is \(W=\sum_i n_i z^{\beta_i}=z^{\beta_1}+z^{\beta_2}+z^{\beta_3}\); identifying the monomials \(z^{\beta_1}=x\), \(z^{\beta_2}=y\), \(z^{\beta_3}=1/(xy)\) (the relation \(\beta_1+\beta_2+\beta_3=0\) from the toric relation \(\sum D_i=-K\), so \(z^{\beta_3}=z^{-\beta_1-\beta_2}=1/(xy)\)) gives \(W=x+y+1/(xy)\), the Hori–Vafa superpotential — the three monomials are the three toric disk classes.
Pitfalls¶
- The SYZ conjecture is a heuristic near the LCSL — the existence of the special Lagrangian fibration is open in general (constructed in special cases: the elliptic curve, the SYZ fibration of toric CY, the almost-toric fibrations of Symington). Do not state "SYZ is a theorem".
- The Gross–Siebert program is the algebro-geometric reconstruction, not the transcendental SYZ — the toric degeneration replaces the special Lagrangian fibration; the two are conjecturally equivalent but the GS is the rigorous version.
- The scattering diagrams produce the cluster structure — the consistency is a non-trivial constraint (the path-ordered product trivial), not automatic; not every collection of walls gives a consistent scattering diagram.
- The cluster algebra / canonical basis (GHK) is for log CY surfaces and some higher-dim cases; the general (higher-dim, non-toric) canonical basis is partially conjectural.
- The log GW invariants (punctured invariants) are a recent construction (GPS, Gross–Siebert); the full mirror theorem (the equality of the log-period and the mirror map) is proven in a class of cases but not uniformly.
- The instanton corrections (the wall-crossing) change the Floer products / the superpotential across walls — the naive dual fibration is only correct in the "large complex structure limit" (no disks); away from the LCSL, the corrections are essential.
- The SYZ mirror is not unique — different toric degenerations / different affine structures give different (but conjecturally derived-equivalent) mirrors.
Refs¶
- Strominger–Yau–Zaslow, "Mirror symmetry is \(T\)-duality" (Nucl. Phys. B 1996)
- Gross–Siebert, "Mirror symmetry via logarithmic degeneration data I–IV" (J. Diff. Geom. 2007–2011); "From real geometry to complex geometry" (Annals 2011)
- Gross–Hacking–Keel, "Mirror symmetry for log Calabi–Yau surfaces I" (Publ. Math. IHÉS 2015); Gross–Pandharipande–Siebert, "The mirror theorem for log CY surfaces" (Duke 2018)
- Kontsevich–Soibelman, "Affine structures and non-archimedean analytic spaces" (2006); Fock–Goncharov, "Moduli spaces of local systems and higher Teichmüller theory" (Publ. Math. IHÉS 2006)
- Hori–Vafa, "Mirror symmetry" (hep-th/0002222); Auroux, "Special Lagrangian fibrations, wall-crossing, and mirror symmetry" (Surv. Diff. Geom. 2013)
Next¶
MS index · Foundations · Gromov–Witten & Quantum Cohomology · Fukaya Category & HMS · Hitchin & Geometric Langlands · Tropical geometry · TG / scattering & mirror — the Gross–Siebert scattering / theta functions / tropical mirror · Higher Teichmüller frontiers