Laakso-type Fractal Spaces

• Laakso-type fractal spaces are compact, geodesic, and Ahlfors–regular metric measure spaces constructed via inverse limits or iterated graph systems, exhibiting non-integer Hausdorff dimensions.
• They support measurable differentiable structures and weak (1,1)–Poincaré inequalities, offering counterexamples to classical differentiability and Sobolev overlap results.
• These spaces serve as testbeds for fractal analysis, spectral computations, and conformal geometry, providing explicit models to probe embedding obstructions and quantum effects.

Laakso-type fractal spaces are a family of compact, geodesic, Ahlfors–regular metric measure spaces constructed via inverse limits or iterated graph systems, exhibiting fractional (non-integer) Hausdorff dimensions, supporting Poincaré inequalities, and displaying analytic and geometric phenomena absent in classical smooth spaces. Originating from Laakso’s 2000 construction, these spaces model rich fractal behaviors while retaining analytic tractability; they serve both as canonical counterexamples and as “toy models” for fractal analysis, differentiability theory, and geometric measure-theoretic phenomena. Such spaces are central in probing the boundaries of analysis on metric spaces, particularly in contexts where classical rectifiability, embedding properties, and Sobolev-type analysis fail.

1. Construction and Fundamental Properties

Laakso-type spaces are built by thickening the unit interval $I=[0,1]$ with a totally disconnected transversal, typically a Cantor-type set $K$ or more generally a fractal $\mathcal{F}$. At each “wormhole level”—a countable, dense set of heights in $I$—pairs of points along $K$ are identified according to a combinatorial gluing rule.

In the standard model, the construction proceeds as follows:

• The product space $I\times K$ is formed with the product of Euclidean and Cantor metrics.
• At each scale, indexed by $n$, specific “wormhole levels” $L_n$ subdivide $I$ into finer intervals; at each $t\in L_n$, certain copies of $K$ or its sub-cells are identified via a bijection (e.g., flipping a digit in the Cantor address).
• Taking the quotient by this identification yields a metric measure space $F = (I\times K)/\sim$ equipped with the shortest-path metric where paths may traverse “vertically” and utilize wormhole shortcuts.
• Variants employ iterated graph systems (IGS): starting from a finite connected generator graph $G_1$ with vertex and edge sets $(V_1,E_1)$ and a “gluing set” $I$, subsequent graphs $G_{n+1}$ are assembled by replacing each edge $e\in E_n$ by a new copy of $G_1$, glued along specified injections. The limit (in Gromov-Hausdorff sense) as $n\to\infty$ is the Laakso-type space $X$ (Anttila et al., 20 Jan 2026, Anttila et al., 2024, Capolli, 2022).

A key outcome is that for suitable choices, the resulting fractal space is $Q$-Ahlfors regular with

$Q = 1 + \frac{\log M}{\log(1/\theta)}$

for classical Laakso spaces with $M$ arms and subdivision parameter $\theta$ (cf. (Bate et al., 29 Oct 2025)), or more generally

$Q = \frac{\log |E_1|}{\log L_*}$

where $|E_1|$ is the number of edges in $G_1$, $L_*$ the scaling factor per level (Anttila et al., 2024, Anttila et al., 20 Jan 2026).

These spaces are compact, doubling, and carry a natural measure (push-forward of Lebesgue × Bernoulli on $I\times K$) which is Ahlfors–regular of dimension $Q$, and they support a weak $(1,1)$–Poincaré inequality (Capolli, 2022, Eriksson-Bique et al., 2024).

2. Analytical Structure and Differentiability

Laakso-type spaces admit measurable differentiable structures in the sense of Cheeger: every real-valued Lipschitz function $f$ on the space is differentiable almost everywhere with respect to a suitable family of singular, mutually singular, doubling measures $p_w$, parameterized by weighting $w$ on the Cantor side. This generalizes Rademacher’s theorem to these fractals (Eriksson-Bique et al., 2024).

The differentiable structure is carried by a single chart—the “height map” $h:F\to I$—so that for almost every $x\in F$, there is a unique $Df(x)$ with

$$\lim_{y\to x} \frac{ f(y) - f(x) - Df(x)\cdot \big( h(y)-h(x)\big)}{d(y,x)}=0.$$

Notably, maximality of the directional derivative in the $I$- (height-) direction only implies differentiability at a $\sigma$-porous set; this is in strong contrast to Euclidean and Carnot-group settings, where the set of such points is co-null or large (Capolli et al., 2022).

Moreover, the universal differentiability set (UDS) phenomenon appears: there exist sets $N\subset F$ of $\mathcal{H}^Q$–measure zero such that every Lipschitz $f:F\rightarrow\mathbb{R}$ is differentiable at some point in $N$ (Eriksson-Bique et al., 2024). Each measure $p_w$ supports a $(1,1)$–Poincaré inequality and almost everywhere differentiability.

3. Poincaré Inequalities, Energy Forms, and Sobolev Singularities

All Laakso-type fractals constructed via IGS or classic wormhole methods support a $(1,1)$–Poincaré inequality, by chaining and covering arguments reminiscent of discrete Poincaré on graphs (Capolli, 2022, Eriksson-Bique et al., 2024). Further, these spaces are testbeds for study of $p$-energy forms: $$\mathcal{E}_{p}^{(n)}(f) = \mathcal{M}_p^{-n} \mathcal{E}_{n,p}(V_n[f]),$$ where $V_n[f]$ is a discrete averaging and $\mathcal{M}_p$ is the $p$-capacity between boundary sets at level one. The self-similar limit

$$\mathcal{E}_p(f) = \lim_{n\to\infty} \mathcal{E}_{p}^{(n)}(f)$$

yields the Sobolev-type space $\mathscr{F}_p$ (Anttila et al., 17 Mar 2025).

A distinctive phenomenon—Sobolev singularity—arises: for certain explicit Laakso-type spaces (e.g., a “diamond” graph generator), the intersection $\mathscr{F}_{p_1} \cap \mathscr{F}_{p_2}$ is exactly the constants for all $p_1 \neq p_2$, and their $p$-energy measures are mutually singular (Anttila et al., 17 Mar 2025). This illustrates an “orthogonality” of Sobolev spaces not present on classical manifolds or Carnot groups.

4. Conformal and Moduli Theoretic Aspects

The Ahlfors regular conformal dimension $\dim_{AR}$ of a Laakso-type space is determined as the critical exponent $Q_*$ so that the $Q_*$-edge modulus $M_{Q_*}(\Theta^{(1)},G_1)=1$, where $\Theta^{(1)}$ is the set of paths connecting the distinguished boundaries in the generator graph (Anttila et al., 20 Jan 2026).

A crucial dichotomy arises:

• The conformal gauge contains an Ahlfors $Q_*$-regular metric if and only if the $Q_*$-optimal density on $E_1$ is positive on every edge (no removable edges). Non-attainment corresponds to the existence of a nontrivial porous subset obstructing the realization of the infimum (Anttila et al., 20 Jan 2026).
• These structures provide constructive counterexamples to conjectures such as Kleiner’s: some self-similar, approximately self-similar Laakso-type spaces are combinatorially Loewner (CLP) but not quasisymmetric to a Loewner space (Anttila et al., 2024, Anttila et al., 20 Jan 2026).

The combinatorial Loewner property becomes computable for such spaces, with discrete modulus at each graph scale; the critical exponent matches their Ahlfors dimension (Anttila et al., 2024).

5. Embedding Obstructions and Metric Geometry

Laakso-type spaces cannot be embedded bi-Lipschitzly into any Euclidean space or Hilbert space provided their dimension is non-integer or other more refined invariants fail (Capolli, 2022, Margaris et al., 2019). This is a consequence both of their non-integer Hausdorff dimension and specific failures in covering or coarea inequalities for curve families; such is witnessed already in the scaling of cycles and cut spaces in their finite graph approximations and persists in the fractal limit (Dilworth et al., 2020).

Spaces constructed with shortcut metrics $d_\eta$ (collapsing certain fibers at each scale by a factor $\eta_i$), can transition sharply between being PI-rectifiable and purely PI-unrectifiable: for $\sum \eta_i^s < \infty$, the space admits a PI structure; for divergence, the space is purely PI-unrectifiable and fails any positive-measure biLipschitz parametrization by a PI space (Bate et al., 29 Oct 2025). These provide models for distinguishing differentiability properties in Banach-valued targets according to uniform convexity.

6. Quantum, Spectral, and Physical Models

Laakso-type spaces have explicit Laplace operators constructed via limits of combinatorial graph Laplacians, with spectrum and multiplicities available in closed form, particularly for self-similar parameters (Kauffman et al., 2010, Kesler et al., 2012). The spectral zeta function

$$\zeta_\Delta(s) = \sum_{\lambda \in \sigma(\Delta)} (\mathrm{mult}\,\lambda) \lambda^{-s}$$

admits analytic continuation and encodes the spectral dimension. Boundary conditions and deformations (e.g., insertion of Dirichlet plates to model Casimir-type problems) yield modified spectra with visible physical consequences. For instance, the Casimir force on 1D Laakso models decays as $a^{-2}$ in plate separation, rather than as $a^{-(d_s+1)}$, and spectral zeta poles off the real axis introduce log-periodic oscillations in energy due to complex dimensions (Kesler et al., 2012). There exist parameter choices where the Casimir force is repulsive, a phenomenon not found in the classical (smooth) regime.

7. Applications and Open Directions

Laakso-type spaces are employed:

• As test cases for differentiability theory in metric measure spaces—providing explicit models with universal differentiability sets and singularity of Sobolev spaces (Eriksson-Bique et al., 2024, Anttila et al., 17 Mar 2025).
• As prototypes in the study of analysis on fractals, diffusions, and heat kernel estimates—supporting symmetric local Dirichlet forms, and admitting unique locally symmetric Brownian motions with Gaussian heat kernel bounds (Steinhurst, 2011).
• For the investigation of the Banach-space geometry of metric spaces: their transportation cost and Lipschitz-free spaces are far from $\ell_1$ (unlike trees), with projection constants and Banach-Mazur distances growing in the scale (Dilworth et al., 2020).
• In probing the limits of conformal dimension theory, the Loewner property, and critical exponent attainment—particularly illustrating the role of porosity and removable edges (Anttila et al., 20 Jan 2026, Anttila et al., 2024).
• For explicit calculation of spectral invariants, energies, and quantum effects not accessible in general fractal settings (Kauffman et al., 2010, Kesler et al., 2012).

A plausible implication is that further generalizations of Laakso-type constructions—by varying the replacement graphs, gluing rules, or metric families—may yield new insights into questions of rectifiability, modulus theory, and analytic singularity phenomena in fractal and higher complexity spaces.