Quantum Open Baker's Maps

• Quantum open baker's maps are finite-dimensional non-unitary quantizations of classical baker dynamics, modeling open quantum chaos and scattering resonances.
• They are constructed using discretized Fourier transforms and smooth cutoffs that select allowed phase-space branches, connecting spectral properties to classical trapped Cantor sets.
• The framework underpins spectral gap estimates, fractal Weyl laws, and resonance counting through fractal uncertainty principles and additive combinatorics.

Quantum open baker’s maps are finite-dimensional non-unitary quantizations of classical open baker dynamics on the torus. They function as analytically tractable models of open quantum chaos and of scattering resonances: the quantum propagator has spectrum in the unit disk, eigenvalues of large modulus are interpreted as long-lived resonances, and the dominant semiclassical structures are determined by the classical trapped set, typically a product Cantor set of dimension $\delta=\frac{\log|\mathcal A|}{\log M}$. The subject has developed along three closely connected lines: precise operator constructions for open baker dynamics, spectral-gap results based on fractal uncertainty principles, and fractal Weyl laws controlling the number of eigenvalues in annuli; later work has extended the framework to continuous openings, decohered projections, and anisotropic higher-dimensional models (Dyatlov et al., 2016, Li, 2022, Cunningham, 23 Feb 2026).

1. Canonical construction of the open quantum baker map

A standard formulation starts from a triple

$$(M,\mathcal A,\chi), \qquad M\in\mathbb N,\quad \mathcal A\subset\{0,\dots,M-1\},\quad \chi\in C_0^\infty((0,1);[0,1]).$$

Here $M$ is the base of the baker map, $\mathcal A$ is the alphabet of allowed branches, and $\chi$ is a smooth cutoff used to localize in the allowed region. In one common normalization one assumes $N=KM$ and works on $\ell_N^2=\ell^2(\mathbb Z_N)$, while in the fractal uncertainty principle setting one often takes $N=M^k$ (Li, 2022, Dyatlov et al., 2016).

The quantum open baker’s map is

$$B_N=\mathcal F_N^* \begin{pmatrix} \chi_{N/M}\mathcal F_{N/M}\chi_{N/M} & & \ & \ddots & \ & & \chi_{N/M}\mathcal F_{N/M}\chi_{N/M} \end{pmatrix} I_{\mathcal A,M},$$

where $\mathcal F_N$ is the unitary discrete Fourier transform on $$(M,\mathcal A,\chi), \qquad M\in\mathbb N,\quad \mathcal A\subset\{0,\dots,M-1\},\quad \chi\in C_0^\infty((0,1);[0,1]).$$0, $$(M,\mathcal A,\chi), \qquad M\in\mathbb N,\quad \mathcal A\subset\{0,\dots,M-1\},\quad \chi\in C_0^\infty((0,1);[0,1]).$$1 is the discretized cutoff, and $$(M,\mathcal A,\chi), \qquad M\in\mathbb N,\quad \mathcal A\subset\{0,\dots,M-1\},\quad \chi\in C_0^\infty((0,1);[0,1]).$$2 is the diagonal projector selecting the branches indexed by $$(M,\mathcal A,\chi), \qquad M\in\mathbb N,\quad \mathcal A\subset\{0,\dots,M-1\},\quad \chi\in C_0^\infty((0,1);[0,1]).$$3. Equivalently,

$$(M,\mathcal A,\chi), \qquad M\in\mathbb N,\quad \mathcal A\subset\{0,\dots,M-1\},\quad \chi\in C_0^\infty((0,1);[0,1]).$$4

This operator is the discrete quantum analogue of the open classical baker map on $$(M,\mathcal A,\chi), \qquad M\in\mathbb N,\quad \mathcal A\subset\{0,\dots,M-1\},\quad \chi\in C_0^\infty((0,1);[0,1]).$$5 (Li, 2022).

The underlying classical map is

$$(M,\mathcal A,\chi), \qquad M\in\mathbb N,\quad \mathcal A\subset\{0,\dots,M-1\},\quad \chi\in C_0^\infty((0,1);[0,1]).$$6

defined on the strips

$$(M,\mathcal A,\chi), \qquad M\in\mathbb N,\quad \mathcal A\subset\{0,\dots,M-1\},\quad \chi\in C_0^\infty((0,1);[0,1]).$$7

Thus the openness is implemented by allowing only the branches indexed by $$(M,\mathcal A,\chi), \qquad M\in\mathbb N,\quad \mathcal A\subset\{0,\dots,M-1\},\quad \chi\in C_0^\infty((0,1);[0,1]).$$8, with the cutoff $$(M,\mathcal A,\chi), \qquad M\in\mathbb N,\quad \mathcal A\subset\{0,\dots,M-1\},\quad \chi\in C_0^\infty((0,1);[0,1]).$$9 smoothing the corresponding phase-space localization (Dyatlov et al., 2016).

This finite-dimensional construction is the canonical model for open quantum baker dynamics. It is non-unitary by design, and $M$0 plays the role of the semiclassical parameter $M$1 (Li, 2022).

2. Trapped sets, resonances, and semiclassical organization

The central classical invariant is the trapped set. For the open baker map it is described in terms of the limiting Cantor set

$M$2

where

$M$3

The forward and backward trapped sets are

$M$4

and the full trapped set is $M$5. Its fractal dimension is

$M$6

This $M$7 is the basic exponent in the fractal Weyl law (Dyatlov et al., 2016).

The quantum spectrum lies in the unit disk since $M$8. The eigenvalues are interpreted as model resonances, with the correspondence

$M$9

so that a strip $\mathcal A$0 corresponds to an annulus $\mathcal A$1. The associated counting function is

$\mathcal A$2

with multiplicities. These are the eigenvalues not too small in modulus, i.e. the long-lived part of the non-unitary spectrum (Li, 2022).

A recurrent semiclassical heuristic is that long-lived quantum states should localize near the trapped set, so the number of such states should be proportional to the number of semiclassical states that can fit near a set of dimension $\mathcal A$3, namely $\mathcal A$4. Later rigorous results show that this heuristic captures the correct leading exponent for broad classes of annuli, but not the full structure of spectral gaps and counting exponents (Li, 2022, Cunningham, 23 Feb 2026).

3. Spectral gaps and the fractal uncertainty principle

A major development was the introduction of a discrete fractal uncertainty principle for open quantum baker’s maps. Dyatlov–Jin study the operator

$\mathcal A$5

and define the fractal uncertainty exponent

$\mathcal A$6

The key estimate is

$\mathcal A$7

and it leads to an essential spectral gap

$\mathcal A$8

for some

$\mathcal A$9

This strictly improves the standard pressure bound for all $\chi$0, including the regime $\chi$1 where the pressure bound is trivial (Dyatlov et al., 2016).

The same work also connects spectral improvement to additive combinatorics. Using additive energy, one obtains a lower bound of the form

$\chi$2

provided the additive energy of the discrete Cantor sets satisfies the stated decay condition. This shows that the spectral gap is not determined by $\chi$3 alone: arithmetic structure matters (Dyatlov et al., 2016).

This point became sharper in later work. "Improved fractal Weyl bounds matching improved spectral gaps for hyperbolic surfaces and open quantum maps" (Cunningham, 23 Feb 2026) proves an improved fractal Weyl bound for quantum open baker’s maps that matches both the improved FUP gap $\chi$4 and the additive-energy gap $\chi$5. The resulting exponent

$\chi$6

shows explicitly that improved resonance-free annuli and improved resonance counting are governed by the same fractal-analytic mechanisms (Cunningham, 23 Feb 2026).

The FUP framework has also been extended to genuinely anisotropic settings. For the 2D anisotropic quantum open baker’s map, with base $\chi$7, $\chi$8, and trapped set a Bedford–McMullen carpet, one has an anisotropic discrete FUP

$\chi$9

under non-full-row or non-full-column hypotheses, and consequently an essential spectral gap

$N=KM$0

This extends the 1D discrete FUP strategy to self-affine trapped sets with anisotropic scaling (Jin et al., 22 Jun 2026).

4. Fractal Weyl laws and eigenvalue counting in annuli

The counting problem asks how many eigenvalues lie in the annulus $N=KM$1. Dyatlov–Jin proved that for every $N=KM$2 and $N=KM$3,

$N=KM$4

For $N=KM$5, this gives the standard fractal Weyl exponent $N=KM$6; for smaller $N=KM$7, the exponent decreases linearly and vanishes at the pressure gap threshold $N=KM$8 (Dyatlov et al., 2016).

A sharper result was obtained in "Weyl Laws for Open Quantum Maps" (Li, 2022). For each fixed $N=KM$9,

$\ell_N^2=\ell^2(\mathbb Z_N)$0

This removes the previous $\ell_N^2=\ell^2(\mathbb Z_N)$1-loss and improves the earlier $\ell_N^2=\ell^2(\mathbb Z_N)$2 bound to the sharp exponent $\ell_N^2=\ell^2(\mathbb Z_N)$3. If the cutoff has Gevrey regularity,

$\ell_N^2=\ell^2(\mathbb Z_N)$4

then for all $\ell_N^2=\ell^2(\mathbb Z_N)$5 and all sufficiently large $\ell_N^2=\ell^2(\mathbb Z_N)$6,

$\ell_N^2=\ell^2(\mathbb Z_N)$7

The Gevrey case therefore yields explicit dependence on the annulus depth $\ell_N^2=\ell^2(\mathbb Z_N)$8 (Li, 2022).

The proof strategy has become structurally standard. It begins with nonstationary phase estimates and one-step propagation of singularities, which imply that $\ell_N^2=\ell^2(\mathbb Z_N)$9 pushes mass toward the next Cantor-like strip while $N=M^k$0 propagates in reverse. Iterating this propagation yields an approximate inverse

$N=M^k$1

where $N=M^k$2 is a finite-rank localizer onto a neighborhood of the trapped set. The rank estimate for $N=M^k$3 is precisely where $N=M^k$4 enters. Choosing propagation time $N=M^k$5 makes the rank scale like $N=M^k$6, and Jensen’s formula applied to a determinant built from the parametrix converts this rank control into eigenvalue counting (Li, 2022).

The 2026 refinement modifies the determinant stage. Instead of the older determinant, it uses

$N=M^k$7

which allows sharper trace-class estimates and brings the improved FUP and additive-energy exponents directly into the counting argument. A common misconception is that trapped-set dimension alone fixes the optimal counting exponent in every annulus; the improved results show that the fine fractal arithmetic encoded in $N=M^k$8 and $N=M^k$9 can further sharpen the bound near the spectral gap (Cunningham, 23 Feb 2026).

5. Continuous openings, tribaker maps, and periodic-orbit organization

Not all open baker models use a fully absorbing opening. In the continuously open quantum tribaker map, the opening is described by a reflectivity function

$$B_N=\mathcal F_N^* \begin{pmatrix} \chi_{N/M}\mathcal F_{N/M}\chi_{N/M} & & \ & \ddots & \ & & \chi_{N/M}\mathcal F_{N/M}\chi_{N/M} \end{pmatrix} I_{\mathcal A,M},$$0

so trajectories are partially reflected rather than completely removed. The quantum map is

$$B_N=\mathcal F_N^* \begin{pmatrix} \chi_{N/M}\mathcal F_{N/M}\chi_{N/M} & & \ & \ddots & \ & & \chi_{N/M}\mathcal F_{N/M}\chi_{N/M} \end{pmatrix} I_{\mathcal A,M},$$1

and long-lived resonances are those with $$B_N=\mathcal F_N^* \begin{pmatrix} \chi_{N/M}\mathcal F_{N/M}\chi_{N/M} & & \ & \ddots & \ & & \chi_{N/M}\mathcal F_{N/M}\chi_{N/M} \end{pmatrix} I_{\mathcal A,M},$$2 close to $$B_N=\mathcal F_N^* \begin{pmatrix} \chi_{N/M}\mathcal F_{N/M}\chi_{N/M} & & \ & \ddots & \ & & \chi_{N/M}\mathcal F_{N/M}\chi_{N/M} \end{pmatrix} I_{\mathcal A,M},$$3. The corresponding classical object is a continuous repeller obtained from finite-time forward and backward trapped intensity distributions rather than a strict escape/no-escape set (Prado et al., 2017).

Two reflectivity profiles were studied: a Fermi-Dirac-like step smoothing and a sinusoidal reflectivity. The central semiclassical conclusion is that the shortest periodic orbits belonging to the classical repeller of the fully open map remain robust in a perturbative regime and continue to support the long-lived resonances. Scar functions built from such short periodic orbits form an efficient nonorthogonal basis, and the overlap between the exact quantum continuous-repeller distribution and the semiclassical one satisfies

$$B_N=\mathcal F_N^* \begin{pmatrix} \chi_{N/M}\mathcal F_{N/M}\chi_{N/M} & & \ & \ddots & \ & & \chi_{N/M}\mathcal F_{N/M}\chi_{N/M} \end{pmatrix} I_{\mathcal A,M},$$4

in all tested cases. For step-like reflectivity, the number of scar functions needed is significantly reduced, similarly to the completely open situation; for sinusoidal reflectivity, the reduction is less pronounced and the spectral behavior deviates more strongly from the discontinuous case (Prado et al., 2017).

The same work also emphasizes that continuous openings alter spectral scaling qualitatively. The strong oscillations typical of discontinuous openings in the scaling of the number of long-lived resonances are almost absent, and the authors suggest this may indicate a different Weyl-law regime. Thus the fully open repeller remains the organizing semiclassical structure in the perturbative regime, but continuous reflectivity changes the detailed resonance statistics (Prado et al., 2017).

6. Adjacent open-system formulations and operator-theoretic viewpoints

A closely related construction is the quantum Bernoulli map, defined as a projection of the quantum baker map with instant decoherence in the position basis after each step: $$B_N=\mathcal F_N^* \begin{pmatrix} \chi_{N/M}\mathcal F_{N/M}\chi_{N/M} & & \ & \ddots & \ & & \chi_{N/M}\mathcal F_{N/M}\chi_{N/M} \end{pmatrix} I_{\mathcal A,M},$$5 This produces a dissipative, non-unitary effective map on density matrices. The paper constructs quantum decaying states represented by density matrices, derives decay laws

$$B_N=\mathcal F_N^* \begin{pmatrix} \chi_{N/M}\mathcal F_{N/M}\chi_{N/M} & & \ & \ddots & \ & & \chi_{N/M}\mathcal F_{N/M}\chi_{N/M} \end{pmatrix} I_{\mathcal A,M},$$6

and shows that the evolving quantum Bernoulli polynomials develop a quasi-fractal structure down to the resolution scale

$$B_N=\mathcal F_N^* \begin{pmatrix} \chi_{N/M}\mathcal F_{N/M}\chi_{N/M} & & \ & \ddots & \ & & \chi_{N/M}\mathcal F_{N/M}\chi_{N/M} \end{pmatrix} I_{\mathcal A,M},$$7

The construction is explicitly framed as conceptually close to the broader class of open quantum baker maps, since the openness comes from repeated decoherence and irreversible coarse-graining rather than from a spatial hole (Ordonez et al., 2011).

Another operator-theoretic viewpoint comes from truncations of the closed quantum baker propagator. In the OTOC formulation with projector observables, the projected evolution

$$B_N=\mathcal F_N^* \begin{pmatrix} \chi_{N/M}\mathcal F_{N/M}\chi_{N/M} & & \ & \ddots & \ & & \chi_{N/M}\mathcal F_{N/M}\chi_{N/M} \end{pmatrix} I_{\mathcal A,M},$$8

is a subunitary truncated matrix, and

$$B_N=\mathcal F_N^* \begin{pmatrix} \chi_{N/M}\mathcal F_{N/M}\chi_{N/M} & & \ & \ddots & \ & & \chi_{N/M}\mathcal F_{N/M}\chi_{N/M} \end{pmatrix} I_{\mathcal A,M},$$9

is determined by the squared singular values $\mathcal F_N$0 of the truncated propagator. This is exactly the structure familiar from open quantum baker maps: a projector turns unitary evolution into a truncated non-unitary operator whose singular values encode contraction, escape, and scrambling (Lakshminarayan, 2018).

A more semiclassical but technically adjacent representation appears in the transfer-matrix approach to baker traces. There the baker transfer matrix $\mathcal F_N$1 is non-unitary, while the reflected baker transfer matrix $\mathcal F_N$2 is exactly unitary; in the circuit representation, the nonunitarity of $\mathcal F_N$3 is restricted to a single one-qubit gate

$\mathcal F_N$4

Although this is not itself an open baker-map construction, it isolates a controlled nonunitary core inside a baker-map circuit and provides a useful comparison point for open and truncated models (Abreu et al., 2010).

Taken together, these adjacent formulations clarify that “openness” in baker systems can be realized in several mathematically distinct ways: by deleting branches in phase space, by continuous reflectivity, by repeated decoherence, or by projector-induced truncation. The shared themes are nonunitarity, resonance-like decay, and localization on dynamically defined fractal or symbolic structures.