Modulus of Continuity: t(1+|log t|) Analysis

• The modulus of continuity t(1+|log t|) is a quantitative measure defining intermediate smoothness that bridges uniform continuity and stronger (Lipschitz/Hölder) regularity.
• It is equivalent to Peetre’s K-functional and plays a central role in establishing optimal rates in approximation theory and operator estimates like the Hardy-Littlewood maximal function.
• Appearing in applications from nonlinear PDE boundary regularity to probabilistic and geometric settings, this modulus provides sharp, optimal estimates under minimal smoothness assumptions.

The modulus of continuity $t(1+|\log t|)$ is a quantitative rule describing the uniform continuity of functions and mappings, which appears in optimal estimates for several analytic, probabilistic, and geometric contexts. It characterizes situations where continuity is strictly better than mere uniform continuity, yet falls short of classical Hölder or Lipschitz regularity; the logarithmic factor reflects the presence of slow gain in smoothness at fine scales, frequently dictated by degenerate or critical phenomena in analysis. This modulus is explicitly realized in approximation theory, probabilistic regularity of stochastic processes, boundary regularity for nonlinear PDEs, and in geometric-functional correspondences such as the Kerov transition measure.

1. Definition and Construction of the $t(1+|\log t|)$ Modulus

The function $\omega(t) = t(1 + |\log t|)$ serves as a prototypical modulus of continuity, quantifying how uniformly continuous functions vary at small scales. In classical settings, the first-order modulus of continuity is defined by

$$\omega_1(f, t) = \sup_{|u| \le t} \|f(\cdot+u) - f(\cdot)\|_\infty,$$

but optimization over convolution or Steklov averages leads to the Boman–Shapiro special modulus

$\Omega_1(f, h) = \|f - f*\varphi_h\|_\infty,$

where $\varphi_h$ is the normalized hat kernel. The weighted version defines

$$\omega(f, t) = \sup_{0 < h \le t} \frac{\Omega_1(f, h)}{1 + h/t}.$$

As $t \to 0$, it is shown that

$$A\,t(1+|\log t|) \le \omega(f, t) \le B\,t(1+|\log t|),$$

for absolute constants $A,B > 0$ and all $f$ with $\|f'\|_\infty=1$ (Dolmatova, 2013). This is optimal for the Steklov setting and precise for capturing the worst possible behavior for continuous functions with bounded derivative.

2. Equivalence with Peetre’s $K$-Functional and Approximation Theory

This modulus of continuity has foundational equivalence with the first-order Peetre $K$-functional, defined as

$$K(f, t) = \inf_{g \in W^{1,\infty}(T)} \Big\{ \|f-g\|_\infty + t\|g'\|_\infty \Big\},$$

for $f$ continuous and 1-periodic. The equivalence is captured via

$$C_1\,\omega(f, t) \le K(f, t) \le C_2\,\omega(f, t),$$

where $C_1 = 1$, $C_2 = 2$ (Dolmatova, 2013). Through this bridge, one derives Jackson-type inequalities for approximation by trigonometric polynomials: $E_n(f)_\infty \le M\,\omega(f, 1/n), \qquad M = 3,$ with all constants explicit. The modulus $t(1+|\log t|)$ thus governs best uniform rates for linear approximation schemes in the absence of higher regularity.

3. Operator-Sensitive Modulus: Hardy-Littlewood Maximal Function

The $t(1+|\log t|)$ modulus also arises in the optimal bounds for the continuity of the Hardy-Littlewood maximal operator. For $f$ bounded and uniformly continuous, and $\omega(h) = h(1+|\log h|)$, it is shown that

$\omega_{Mf}(t) \le (\sqrt{2}-1)\,t(1+|\log t|),$

for all $t > 0$, where $\sqrt{2}-1$ is the best constant, attained via minimization over auxiliary parameters (Aldaz et al., 2010). As $t \to 0$, the modulus is sharp: $\omega_{Mf}(t) \sim (\sqrt{2}-1)\,t\,|\log t|,$ these optimal constants persist under cone-shaped extremal examples, confirming that the operator cannot improve regularity beyond the $t|\log t|$ regime.

4. Geometric Modulus in Representation Theory: Kerov’s Transition Measure

In asymptotic representation theory, particularly for symmetric groups, the modulus $t(1+|\log t|)$ describes the stability of the Kerov transition measure under perturbations of continual Young diagrams. For diagrams $\omega_1,\omega_2$ lying in a contracting region of a reference diagram $\Omega$,

$$\sup_{z \in [a_0, b_0]} |K_{\omega_1}(z) - K_{\omega_2}(z)| \le C\,t(1 + |\log t|),$$

where $t$ is the supremum-norm distance between diagram profiles and $C$ depends on geometric and measure-theoretic data (Śniady, 2024). The bound is realized via explicit barrier diagrams and exact Cauchy transform identities, and is optimal up to constants in staircase or "maximally shifted" diagram examples.

5. Boundary Modulus in Nonlinear PDE: Stefan Problem

Within degenerate parabolic PDE, such as the two-phase Stefan problem, the logarithmic-type modulus is structurally optimal for boundary regularity of weak solutions. Under minimal geometric and data regularity assumptions,

$|u(x,t) - u(y,s)| \le C\,\rho(1 + |\log \rho|),$

where $\rho = |x-y| + |t-s|^{1/2}$ and $C$ depends only on structural constants (Liao, 2021). This rate emerges from oscillation-decay iterations (De Giorgi-type), dichotomy of measure alternatives, and energy estimates up to geometric boundaries. Formal barrier arguments demonstrate that even for the linear problem, this logarithmic modulus cannot be improved.

6. Comparison with Log-Hölder and Related Moduli in Nonlinear and Probabilistic Settings

In settings of complex Monge–Ampère equations on Hermitian manifolds, the natural continuity is not as strong as the $t(1+|\log t|)$ modulus but is instead a log–Hölder modulus: $$|u(x) - u(y)| \le \frac{C}{|\log d(x,y)|^\alpha}, \qquad 0 < \alpha < \frac{p}{n} - 1,$$ noting that $t(1+|\log t|)$ is strictly better than log–Hölder continuity in small scales (Liu, 26 Oct 2025). In harmonizable fractional stable motion, the optimal modulus is of power–logarithmic type: $$|X(s) - X(t)| = O\left( |s-t|^H (\ln(1/|s-t|))^{1/\alpha+\varepsilon} \right),$$ with exponent dictated by the stable index and no improvement to $t(1+|\log t|)$ except in special cases (Ayache et al., 2023).

7. Asymptotic Optimality and Structural Necessity

The $t(1+|\log t|)$ modulus arises naturally from constructions involving convolution averages, oscillation decay via iterative schemes, and explicit counterexamples ("bump" or dyadic constructions). In Steklov averaging, integral telescoping arguments directly yield the logarithmic factor. In probabilistic and PDE contexts, the necessity is traced to degenerate decay of oscillations, limitations of energy propagation, and criticality at fine scales. For all previously enumerated applications, the $t(1+|\log t|)$ modulus is either sharp or the best possible up to constants, unless stricter regularity or compactness conditions are imposed.

---

Summary Table: Key Appearances of $t(1+|\log t|)$ Modulus

Context	Main Statement/Form	arXiv Reference
Boman–Shapiro modulus for periodic functions	$\omega(f,t) \sim t(1+|\log t|)$	(Dolmatova, 2013)
Peetre $K$-functional equivalence	$K(f,t) \asymp t(1+|\log t|)$	(Dolmatova, 2013)
Hardy-Littlewood maximal operator	$\omega_{Mf}(t) \le (\sqrt{2}-1)t(1+|\log t|)$	(Aldaz et al., 2010)
Kerov transition measure (diagram stability)	$$|K_{\omega_1}(z) - K_{\omega_2}(z)| \le Ct(1 + |\log t|)$$	(Śniady, 2024)
Stefan problem boundary regularity	$|u(x,t) - u(y,s)| \le C\rho(1+|\log \rho|)$	(Liao, 2021)

A plausible implication is that the $t(1+|\log t|)$ modulus represents an intermediate regularity regime demarcating the threshold between uniform continuity and genuinely stronger (e.g., Hölder or Lipschitz) regularity for a wide range of functional, geometric, and analytic constructs.