Birkhoff–Tartar Fixed-Point Results

• Birkhoff–Tartar fixed-point results are a generalization of the Poincaré–Birkhoff theorem that provide purely topological criteria ensuring the existence of at least two fixed points.
• The approach utilizes universal covering spaces and brick decompositions to analyze orbit divergence, establishing fixed-point guarantees through key topological obstructions.
• By significantly weakening traditional area and boundary-twist conditions, these results offer simplified proofs and broader applications in dynamical systems on annuli.

The Birkhoff–Tartar fixed-point results address the existence of fixed points for homeomorphisms of the closed annulus $A = \mathbb{S}^1 \times [-1,1]$ that are isotopic to the identity. These results generalize and unify various classical theorems—including the Poincaré–Birkhoff theorem—by introducing purely topological criteria that guarantee the presence of at least two fixed points, without requiring area preservation or boundary twist conditions. The framework isolates necessary topological obstructions and significantly weakens assumptions traditionally imposed, providing new proofs for several well-known fixed-point results (Bonino, 2010).

1. Purely Topological Generalization of the Poincaré–Birkhoff Theorem

Let $h \colon A \to A$ be a homeomorphism of the closed annulus, isotopic to the identity, with at most one fixed point. The central result states that under these conditions, one of two mutually exclusive topological configurations must occur:

1. There exists an essential Jordan curve $J \subset A$ such that $J \cap h(J) = J \cap \operatorname{Fix}(h)$.
2. There exists an arc $\alpha$ crossing the annulus (joining one boundary to the other) such that $$\alpha \cap h(\alpha) = \alpha \cap \operatorname{Fix}(h)$$ and $h(\alpha)$ does not intersect either of the local sides of $\alpha$ in $A\setminus\alpha$.

If neither obstruction is present, $h$ necessarily possesses at least two fixed points. The proof utilizes Le Calvez–Sauzet brick decompositions in the universal cover $\widetilde A = \mathbb{R} \times [-1,1]$ and a repeller–attractor argument following Franks. These decompositions and arguments identify fundamental topological barriers preventing the existence of nontrivial wandering subannuli and thereby ensure the fixed-point property holds in the absence of such obstructions (Bonino, 2010).

2. Universal Cover and Unbounded Orbits

In the universal covering space $\widetilde A = \mathbb{R} \times [-1,1]$, orbits of a lift $H \colon \widetilde A \to \widetilde A$ of $h$ can be characterized by their behavior along the horizontal coordinate. A point $\tilde x$ has a forward orbit unbounded on the right if $p_1(H^n(\tilde x)) \to +\infty$ as $n \to +\infty$, and unbounded on the left if $p_1(H^n(\tilde x)) \to -\infty$. If there exist points with orbits diverging in both directions, then—by contradiction via Theorem 1—the presence of both types of unboundedness precludes the possibility of only one fixed point, enforcing the existence of at least two (Bonino, 2010).

3. Weakening of the Boundary-Twist Condition

The classical Poincaré–Birkhoff theorem relies on an area-preservation and strict twist condition on the boundaries; that is, the existence of a lift $H$ with

$$H(\theta,-1) = (\varphi_-(\theta),-1), \quad H(\theta,+1) = (\varphi_+(\theta),+1),$$

where

$$\left(\varphi_-(\theta)-\theta\right)\left(\varphi_+(\theta)-\theta\right) < 0, \qquad \forall \theta.$$

Corollary 1 substantially weakens this by requiring only that $H$ has two orbits diverging to $+\infty$ and $-\infty$, provided there is no essential subannulus $B \subset A$ with $h(B) \subsetneq B$ or $h^{-1}(B) \subsetneq B$. This orbit criterion requires no quantitative twist or rotation, rendering it strictly weaker than classical assumptions. If the existence of such orbits is established, $h$ must possess at least two distinct fixed points, specifically in the Nielsen class $\Pi(\operatorname{Fix}(H))$ (Bonino, 2010).

4. Area-Preserving Case and the Conley–Zehnder Theorem

When $h$ preserves the normalized Lebesgue area $\lambda$ on $A$, the framework gives a new proof of a variant of the Conley–Zehnder theorem. Given a lift $H$ and the horizontal-displacement function

$$D_H(x) = p_1(H(\tilde x)) - p_1(\tilde x), \qquad \tilde x \in \Pi^{-1}(x),$$

the mean horizontal displacement $\overline D_H$ is

$\overline D_H = \int_A D_H(z)\,d\lambda(z).$

If $\overline D_H = 0$, $h$ must have at least two fixed points, with the conclusion holding in the Nielsen sense. This proof eschews generating-function and symplectic-capacity techniques, instead relying on a purely topological argument involving zero-measure crossing arcs and area-counting via Lemma 5.5 (Bonino, 2010).

5. Comparison with Classical Results and the Role of Topological Obstructions

A comparison with classical Poincaré–Birkhoff and Birkhoff–Tartar theorems reveals the scope and strength of the present approach. The Poincaré–Birkhoff theorem assumes area-preservation and strict boundary twisting, guaranteeing two fixed points. The Birkhoff–Tartar and Le Calvez–Wang frameworks weaken area preservation to require only the absence of wandering essential subannuli or invariant Brouwer foliations. The central theorem reviewed here removes both area and twist conditions, isolating the one-fixed-point case to two concrete topological alternatives. In doing so, it recovers and generalizes earlier results—including Franks’s twist-orbit criterion and Flucher’s approach to the Conley–Zehnder theorem—into a unified topological statement. This framework demonstrates that, under an at-most-one-fixed-point assumption, topological obstructions are both necessary and sufficient for the failure of the two-fixed-point phenomenon (Bonino, 2010).

6. Significance and Broader Implications

The isolation of topological obstructions in this context clarifies the dichotomy between conservative dynamics exhibiting global fixed-point phenomena and maps that evade such constraints via specific geometric structures. This approach provides a versatile toolkit for classifying annulus homeomorphisms according to their fixed-point structure, extending the reach of old results and simplifying the assumptions required for new ones. The characterization in terms of purely topological invariants supports further generalizations to more complex dynamical systems on surfaces (Bonino, 2010).