Focus-Focus Singularities in Integrable Systems

• Focus–focus singularities are nondegenerate fixed points in integrable systems characterized by complex eigenvalue pairs and distinctive symplectic invariants.
• They locally adhere to the Eliasson normal form, providing canonical coordinates that reveal affine monodromy and obstructions to global action–angle coordinates.
• Their global topology displays multi-pinched torus structures and smooth invariants that are vital for classifying semitoric systems and advancing semiclassical analysis.

A focus–focus singularity is a prototypical nondegenerate fixed point in the theory of integrable Hamiltonian systems of two degrees of freedom, and occupies a central position in symplectic topology, the geometry of torus fibrations, and the global study of Lagrangian fibrations. Such singularities exhibit characteristic local and global phenomena—topologically nontrivial vanishing cycles, affine monodromy, obstructions to global action–angle coordinates, and intricate smooth invariants—making their classification and analysis foundational for advanced research in integrable systems, symplectic geometry, and semiclassical analysis.

1. Local Structure and Normal Forms

Let $(M^4,\omega)$ be a symplectic 4-manifold with an integrable system

$\mathcal{F} = (H, F) : M \to \R^2$

where $\{H, F\} = 0$ and $\mathrm{rank}\,d\mathcal{F} = 2$ generically. A point $x \in M$ is a rank-0 singular point if $dH(x) = dF(x) = 0$. The point $x$ is a focus–focus singularity if:

• The Hessians $d^2 H(x)$ and $d^2 F(x)$ are linearly independent.
• Some linear combination has symplectic Hamiltonian matrix with eigenvalues $\pm\alpha \pm i\beta$, $\mathcal{F} = (H, F) : M \to \R^2$0.

The local geometric structure is governed by the Eliasson (Vey–Ito) normal form theorem: there exist canonical coordinates $\mathcal{F} = (H, F) : M \to \R^2$1 so that

$\mathcal{F} = (H, F) : M \to \R^2$2

and

$\mathcal{F} = (H, F) : M \to \R^2$3

for

$\mathcal{F} = (H, F) : M \to \R^2$4

Up to local diffeomorphism of $\mathcal{F} = (H, F) : M \to \R^2$5, this normal form is unique and provides a symplectic model for all local considerations (Smirnov, 2013, Ngoc et al., 2011, Ratiu et al., 2017).

2. Global Topology and Complexity

A focus–focus fiber $\mathcal{F} = (H, F) : M \to \R^2$6 is defined as a connected fiber containing only rank-0 focus–focus points and no rank-1 points. The complexity $\mathcal{F} = (H, F) : M \to \R^2$7 of such a fiber is the number of distinct focus–focus points it contains:

• Simple focus–focus singularity: $\mathcal{F} = (H, F) : M \to \R^2$8
• Complex (multi-pinched) singularity: $\mathcal{F} = (H, F) : M \to \R^2$9

The singular fiber is topologically a torus pinched $\{H, F\} = 0$0 times, equivalently a chain of $\{H, F\} = 0$1 embedded Lagrangian $\{H, F\} = 0$2-spheres, each intersecting its neighbors transversely at the focus–focus points. Cohomological obstructions dictate that:

• $\{H, F\} = 0$3
• $\{H, F\} = 0$4 (or $\{H, F\} = 0$5 if $\{H, F\} = 0$6 is compact)

Each Lagrangian sphere in the chain has self-intersection number $\{H, F\} = 0$7; their Poincaré duals are linearly independent except for one relation in the noncompact case (Smirnov, 2013).

3. Monodromy and Affine Structure

Focus–focus singularities are the canonical source of nontrivial monodromy in Lagrangian torus fibrations:

• In the regular region, action–angle coordinates $\{H, F\} = 0$8 exist.
• A small loop around the focus–focus value induces a transformation of the action cycles: $\{H, F\} = 0$9 where $\mathrm{rank}\,d\mathcal{F} = 2$0 is the complexity. For the simple case ($\mathrm{rank}\,d\mathcal{F} = 2$1), this is the classical

$\mathrm{rank}\,d\mathcal{F} = 2$2

monodromy matrix (Smirnov, 2013, Wacheux, 2014, Ratiu et al., 2017, Pelayo et al., 2013).

The nontrivial monodromy obstructs the existence of global action–angle coordinates, as the period lattice bundle over the image of the moment map is nontrivial. The singular integral affine structure on the base acquires branch cuts at the focus–focus values, and the local multi-valuedness of action variables reflects the logarithmic branch structure in their asymptotics.

4. Classification, Moduli, and Invariants

Simple Focus–Focus Singularities

The saturated neighborhood of a simple focus–focus fiber is classified by the Taylor series invariant at the focus–focus point (Vu Ngọc), which, up to local symplectomorphism in the germ, uniquely distinguishes such singularities (Pelayo et al., 2013, Ngoc et al., 2011). In the context of semitoric systems, additional global invariants come into play (see below).

Complex (Multi-Pinched) Fibers

Multiple focus–focus points on a single fiber require refined invariants for full classification:

• The moduli space of such singularities is parameterized by collections of formal power series (action–Taylor series and transition data), up to the natural action of $\mathrm{rank}\,d\mathcal{F} = 2$3, where $\mathrm{rank}\,d\mathcal{F} = 2$4 is the dihedral group permuting and reflecting the $\mathrm{rank}\,d\mathcal{F} = 2$5 points (Pelayo et al., 2018).
• Smooth invariants are encoded via the Taylor expansions of gluing maps between standard normal charts; for double-pinched ($\mathrm{rank}\,d\mathcal{F} = 2$6), a single real parameter $\mathrm{rank}\,d\mathcal{F} = 2$7 (the modulus of the first-order term) suffices for the smooth classification (Bolsinov et al., 2017).

Notably, there exist homeomorphic but non-diffeomorphic focus–focus singularities, and product decomposition fails in the presence of nontrivial gluing invariants—this disproves Zung’s conjecture that any nondegenerate singularity is smoothly a product of elementary blocks (Bolsinov et al., 2017).

5. Occurrence in Mechanical Systems and Explicit Models

A strong constraint governs the appearance of complex focus–focus singularities:

• In cotangent bundles over closed orientable surfaces of genus $\mathrm{rank}\,d\mathcal{F} = 2$8 or on non-orientable surfaces, only simple focus–focus singularities occur for integrable systems; on $\mathrm{rank}\,d\mathcal{F} = 2$9, complexity-2 can arise if the magnetic term is exact (Smirnov, 2013).
• On $x \in M$0 and $x \in M$1 orbits, complexity is generically at most 2, with complexity-2 occurring only at special parameter values (e.g., vanishing magnetic area or symplectic areas in resonance).

Model systems exhibiting complexity-2 focus–focus fibers include:

• The singular orbit $x \in M$2 with magnetic term vanishing.
• The resonant orbit $x \in M$3, where the symplectic areas are equal (Smirnov, 2013).

6. Affine Geometry and Convexity Phenomena

The integral affine structure on the base of a Lagrangian torus fibration with focus–focus points exhibits both local convexity and, when total monodromy is large, global nonconvexity. Key features:

• Locally around a focus–focus point, “focus–boxes” are convex: for any two points in such a neighborhood, there is always a straight affine segment connecting them, for at least one branch of the multi-valued coordinate.
• Globally, when the monodromy group is nontrivial (multiple focus–focus points/curves), the local-global convexity principle can fail; for example, “integral affine black holes” can occur where geodesics are trapped and global convexity is broken (Ratiu et al., 2017).
• In simple semitoric systems (each focus fiber singly pinched), singular affine structures are classified up to fiber-preserving symplectomorphism by the polygon invariant and Taylor data; with multi-pinched fibers, further local invariants are essential (Tang, 2024).

7. Applications and Further Directions

Focus–focus singularities are indispensable in the symplectic classification of semitoric and almost-toric integrable systems, provide laboratory examples for quantum and semiclassical spectral analysis, and are central in understanding Lagrangian surgery theories, topological monodromy, Fukaya categories, and the SYZ mirror symmetry framework (Pelayo et al., 2013, Abouzaid et al., 2024).

Multiple explicit families with two or more focus–focus points in compact and semitoric manifolds have recently provided exact models for the study of the twisting index, height invariants, and the full range of semitoric invariants (Hohloch et al., 2017, Meulenaere et al., 2019, Alonso et al., 2020). Furthermore, higher index (multi-pair) focus–focus singularities arise in settings such as the Jaynes–Cummings–Gaudin model, where the topology and regularized symplectic invariants reflect subtle non-factorizable dynamical behavior (Babelon et al., 2013).

References: (Smirnov, 2013, Ngoc et al., 2011, Ratiu et al., 2017, Babelon et al., 2013, Bolsinov et al., 2017, Pelayo et al., 2018, Tang, 2024, Hohloch et al., 2017, Meulenaere et al., 2019, Alonso et al., 2020, Wacheux, 2014, Pelayo et al., 2013, Abouzaid et al., 2024)