Coarse-Grained Ellipticity in Elliptic PDEs

• Coarse-grained ellipticity is a scale-dependent generalization of uniform ellipticity defined via variationally constructed effective diffusion matrices across spatial blocks.
• It extends classical regularity and homogenization theories to PDEs with degenerate, singular, fractal, or random coefficient fields.
• It underpins rigorous functional inequalities, renormalization analyses, and multiscale numerical methods, impacting materials science and theoretical PDE studies.

Coarse-grained ellipticity is a modern, scale-dependent generalization of the classical notion of uniform ellipticity for divergence-form elliptic operators. In contrast to uniform ellipticity, which imposes pointwise lower and upper bounds on the symmetric coefficient field, coarse-grained ellipticity is defined through variationally constructed effective diffusion matrices on spatial blocks (typically triadic cubes) across all scales, using only minimal integrability conditions on the coefficients and their inverses. This framework allows for rigorous regularity and homogenization results for equations with highly degenerate, singular, or even fractal and random coefficient fields, significantly broadening the reach of the De Giorgi–Nash–Moser theory and stochastic homogenization beyond classical integrability constraints.

1. Mathematical Definition and Variational Construction

Let $U \subset \mathbb{R}^d$ be a Lipschitz domain and $a(x)$ a symmetric, nonnegative matrix-valued field with $a, a^{-1} \in L^1(U)$. Coarse-grained ellipticity is defined via quadratic functionals associated with test-cones $\mathcal{C}(Q) \subset H^1_a(Q)$ on triadic cubes $Q$:

$a(Q, p; \mathcal{C}) = \sup_{u \in \mathcal{C}(Q)} \fint_Q \{-\nabla u \cdot a \nabla u + 2p \cdot a \nabla u \}$

$(Q, q; \mathcal{C}) = \sup_{u \in \mathcal{C}(Q)} \fint_Q \{-\nabla u \cdot a \nabla u + 2q \cdot \nabla u \}$

These are quadratic in $p$ and $q$, respectively, and induce effective diffusion matrices $a(Q)$ and their duals $a_*^{-1}(Q)$, with the ordering $a_*(Q) \le a(Q)$.

To encode scale dependence, one introduces scale-discounted sums over all subcubes of a reference cube $_m$, parameterized by $s, t \in (0, 1)$:

$\Lambda_s(_m; \mathcal{C}) = \Bigl( c_s \sum_{k=-\infty}^m 3^{-s(m-k)} \max_{z \in 3^k \mathbb{Z}^d \cap\,_m} \|a(z+_k; \mathcal{C})\|^{1/2} \Bigr)^2$

$\lambda_t(_m) = \Bigl( c_t \sum_{k=-\infty}^m 3^{-t(m-k)} \max_{z \in 3^k \mathbb{Z}^d \cap\,_m} |a_*^{-1}(z+_k)|^{1/2} \Bigr)^{-2}$

The coarse-grained ellipticity ratio is then

$$\Theta_{s,t}(_m; \mathcal{C}) = \frac{\Lambda_s(_m; \mathcal{C})}{\lambda_t(_m)}$$

A coefficient field $a$ is coarse-grained elliptic in $_m$ if there exist exponents $s,t$ with $s+t<1$ and admissible cone $\mathcal{C}$ such that $\Theta_{s,t}(_m; \mathcal{C}) < \infty$ (Armstrong et al., 9 Jan 2026).

2. Functional Inequalities and Regularity Consequences

Under coarse-grained ellipticity, the classical regularity results for elliptic equations extend to highly degenerate or singular settings:

• Local boundedness: Weak subsolutions $-\nabla \cdot (a \nabla u) \leq 0$ obey

$\sup_{\frac12\,_0} u \leq C(s,t,d) \Bigl[ \Theta_{s,t}(_0; \mathcal{C}) \Bigr]^{\frac{d}{4\sigma}} \|u_+\|_{L^2(_0)}$

where $\sigma = 1 - s - t$.

• Harnack inequality: Nonnegative weak solutions of $-\nabla \cdot (a \nabla u) = 0$ satisfy

$\sup_{\frac18\,_0} u \leq \exp \Bigl( C(s, t, d) \Theta_{s,t}(_0; \mathcal{C})^{1/2} \Bigr) \inf_{\frac18\,_0} u$

The dependence on $\Theta_{s,t}^{1/2}$ is proven optimal in the constant-coefficient case (Armstrong et al., 9 Jan 2026).

• Refined operator and energy estimates: Poincaré and Caccioppoli inequalities, as well as stability and flux estimates, are established on all scales, with constants quantitatively depending on the coarse-grained ellipticity parameters (Armstrong et al., 29 Sep 2025).

3. Negative Sobolev Regularity and Integrability Criteria

Coarse-grained ellipticity can be enforced via negative Sobolev regularity assumptions rather than just integrability. Let $a, a^{-1} \in L^1(U) \cap W^{-s,p}(U)$, $s < 1 - 1/p$, and $a^{-1} \in L^1(U) \cap W^{-t,q}(U)$, $t < 1 - 1/q$, together with the balance condition

$$\frac{s + t}{2} + \frac{d}{2} \Bigl( \frac{1}{p} + \frac{1}{q} \Bigr) < 1$$

Then $a$ is coarse-grained elliptic with explicit quantitative bounds on $\Theta_{s,t}$, and the regularity theory (including Harnack inequalities) holds on unit cubes (Armstrong et al., 9 Jan 2026).

Setting $s = t = 0$ recovers the classical result (Trudinger’s condition):

$$a \in L^p, \quad a^{-1} \in L^q, \quad \frac{1}{p} + \frac{1}{q} < \frac{2}{d} \implies \Theta_{0,0}(_0) < \infty$$

with the sharp Harnack constant scaling

$$\exp \Bigl( C(d, p, q) \|a\|_{L^p(_0)}^{1/2} \|a^{-1}\|_{L^q(_0)}^{1/2} \Bigr)$$

4. Examples Beyond Classical Integrability: Singular and Fractal Coefficients

Coarse-grained ellipticity admits coefficient fields well beyond integrable regimes:

• Singular fractal measures: For sets $F$ Ahlfors-$\alpha$ regular with $\alpha \in (d-2, d)$ and mollified Frostman measures $\mu_F^\delta$, the field $a_\delta(x) = 1 + \mu_F^\delta(x)$ satisfies $a_\delta, a_\delta^{-1} \notin L^{1+\varepsilon}$ for any $\varepsilon > 0$, yet is coarse-grained elliptic for all $s > (d-\alpha)/2$, uniformly in $\delta$ (Armstrong et al., 9 Jan 2026).
• Gaussian multiplicative chaos: For log-correlated fields $X$, the measure $\mu_\gamma = e^{\gamma X} dx$, after mollification, yields $a_\delta = 1 + \mu_\gamma * \eta_\delta$. Almost surely, $\sup_\delta \|a_\delta\|_{L^1} < \infty$ but $\|a_\delta\|_{L^p} \to \infty$ for all $p > 1$, yet coarse-grained ellipticity holds for $s > \gamma / c_d$ (Armstrong et al., 9 Jan 2026).

This demonstrates the applicability to settings where neither $a$ nor $a^{-1}$ possess higher integrability, including random and fractal environments.

5. Stochastic Homogenization under Coarse-Grained Ellipticity

Coarse-grained ellipticity permits rigorous stochastic homogenization without uniform ellipticity, provided certain scale-discounted negative regularity norms are bounded. For stationary, ergodic random coefficient fields $a(x) = s(x) + k(x)$, with symmetric $s$ and skew-symmetric $k$ parts:

• Define scale-discounted Besov-type norms for cubes $\Box_n$, e.g.,

$\snorm{f}_{\underline{B}^{-2s}_{\infty,1}(\Box_n)} = \sum_{k=-\infty}^n 3^{2s(k-n)}\, \max_{z \in 3^k \mathbb{Z}^d \cap \Box_n} |\int_{z+\Box_k} f(x)\,dx|$

• Require that the scale-discounted supremal averages of $s^{-1}$ and $b = s + k^t s^{-1} k$ remain bounded as $n \to \infty$.

Under these conditions, almost-sure (quenched) homogenization holds in negative-Sobolev norms for solutions of the rescaled equation $$-\nabla \cdot a^{\varepsilon}(x) \nabla u^{\varepsilon} = \nabla \cdot \nabla h$$, yielding convergence to a homogenized effective matrix $\bar a$ (Lau, 20 Dec 2025).

6. Renormalization, Iterability, and Scale-Local Ellipticity

The coarse-grained theory provides a framework for renormalizing elliptic operators across scales:

• Effective block matrices $A(\cu_m),~B(\cu_m)$ are constructed from dual variational principles and encode Dirichlet and Neumann responses on blocks of size $3^m$.
• Scale-local ellipticity constants:

$\Lambda_{s,1}(\cu_m) = \Bigl( \sum_{k=-\infty}^{m} 3^{-s(m-k)} \max_{z \in 3^k\mathbb{Z}^d\cap\cu_m} |A(z+\cu_k)|^{1/2}\Bigr)^2$

$\lambda_{t,1}(\cu_m) = \Bigl( \sum_{k=-\infty}^{m} 3^{-t(m-k)} \max_{z \in 3^k\mathbb{Z}^d\cap\cu_m} |B(z+\cu_k)^{-1}|^{1/2} \Bigr)^{-2}$

One says $a$ is coarse-grained elliptic if these remain finite for $s+t < 1$.

• Scale-local contrast $\Theta_m = |A(\cu_m) B(\cu_m)^{-1}|$ iterates across scales. In the high-contrast regime, homogenization is reached within $O(\log^2(1+\Theta_0))$ scales, where $\Theta_0$ is the initial contrast (Armstrong et al., 29 Sep 2025).

The iterability of the coarse-grained ellipticity constants facilitates rigorous renormalization group analyses and quantitative homogenization even in the presence of microstructural degeneracies or high local contrast.

7. Applications, Extensions, and Comparative Perspectives

Coarse-grained ellipticity subsumes classical uniform ellipticity and integrability conditions, with broad implications:

• Materials science: Accommodates composite media, polycrystalline grains, and localized defects that induce degenerate or singular conductivities at fine scales, provided their spatial distribution is sufficiently dilute in the scale-discounted sense (Lau, 20 Dec 2025).
• Numerical analysis: Guides a posteriori error estimation in heterogeneous multiscale finite element methods, with negative Sobolev rates reflecting that fine-scale oscillatory coefficients exert negligible influence on effective large-scale behavior.
• PDE theory: Extends the reach of classical De Giorgi–Nash–Moser iteration, reverse Hölder inequalities, and logarithmic Caccioppoli techniques to settings governed by blockwise effective parameters.

Coarse-grained ellipticity thus unifies the functional-analytic, probabilistic, and variational approaches to regularity and homogenization of degenerate and singular elliptic equations, with explicit multiscale control. This suggests potential future directions in the analysis of PDEs with fractal, chaotic, or critical coefficients and in the rigorous development of multiscale numerical algorithms in complex media.