Totik-Ditzian Modulus of Continuity

• Totik-Ditzian modulus of continuity is a quantitative measure that adapts step-sizes to match domain geometry and weight structures for optimal error bounds.
• It underpins direct and inverse approximation theorems by linking the modulus with weighted Peetre K-functionals to precisely characterize function smoothness.
• It is applied in estimating errors for positive linear operators, refining classical estimates while effectively handling endpoint singularities and unbounded domains.

The Totik-Ditzian-type modulus of continuity (commonly referred to as the Ditzian–Totik modulus of smoothness or continuity) is a central tool in quantitative approximation theory for positive linear operators, especially when working on finite or infinite intervals and when capturing fine local regularity in the presence of endpoint singularities or weights. Invented by Z. Ditzian and V. Totik, it generalizes the classical modulus by introducing a variable step-size function that adapts its scale to the geometry or weight structure of the domain, producing estimates that are optimal in both local and global senses for the order of convergence, characterization of function spaces, and direct/inverse approximation theorems.

1. Formal Definitions and Step-Weight Schemes

For $r\geq1$ and an admissible step-weight $\phi\colon I\to[0,\infty)$ on a domain $I\subset\mathbb R$, the $r$th-order Ditzian–Totik modulus of smoothness for $f\in C(I)$ is defined as

$$\omega_\phi^r(f,t) = \sup_{0<h\le t} \sup_{x,k\,h\,\phi(x)\in I} \left| \Delta_{h\phi}^r f(x) \right|,$$

where

$$\Delta_{h\phi}^r f(x) = \sum_{k=0}^r (-1)^{r-k} \binom{r}{k}\, f(x + k\,h\,\phi(x)).$$

For the classical case on $[0,1]$ one sets $\phi(x)=\sqrt{x(1-x)}$; for unbounded domains like $[0,\infty)$ one employs $\phi(x)=\sqrt{x}$ or $\phi(x)=\sqrt{x(1+x)}$ (Rao et al., 2015, Deo et al., 4 Mar 2025, Pratap et al., 2018, Yadav et al., 2019, Wafi et al., 2015).

The first-order modulus is particularly common,

$$\omega_{\phi}(f,t) = \sup_{0<h\le t} \left\{\big|f(x+\tfrac{h\phi(x)}{2}) - f(x-\tfrac{h\phi(x)}{2})\big| : x \pm \tfrac{h\phi(x)}{2} \in I \right\},$$

with analogous forms on $[0,1]$, $[0,\infty)$, or weighted spaces.

2. Equivalence with Weighted Peetre K-Functionals

Central to Totik–Ditzian theory is the tight equivalence between their modulus of smoothness and an appropriately weighted Peetre $K$-functional,

$$K_{r,\phi}(f, t) = \inf_{g\in W^r(I)} \big\{ \|f-g\| + t^r \|\phi^r g^{(r)}\| \big\},$$

where $W^r(I)$ comprises functions having sufficient smoothness and whose weighted derivatives are bounded. The equivalence,

$$C_1\, K_{r,\phi}(f,t)\leq \omega_\phi^r(f,t)\leq C_2\,K_{r,\phi}(f,t),$$

with $C_1, C_2 > 0$ independent of $f$ and $t$, allows the modulus and $K$-functional to be used interchangeably for direct/inverse theorems and regularity classification (Kajla et al., 2018, Pratap et al., 2018, Yadav et al., 2019, Rao et al., 2015). For weighted approximation, the $K$-functional is often adapted further with a weight $w(x)$, as in

$$K_{r,\phi}(f, t)_w = \inf_{g} \left\{ \|w(f-g)\| + t^r \|w\phi^r g^{(r)}\| \right\},$$

which is fundamental for endpoint singularities (Lu et al., 2010).

3. Application to Operator Error Estimates: Direct and Inverse Theorems

Most refined error estimates for positive approximation operators (Bernstein, Baskakov, Szász–Mirakjan, Durrmeyer variants, Bézier variants, etc.) are expressible in Totik–Ditzian terms: $$|L_n(f;x) - f(x)| \leq C\, \omega_\phi^r \left( f; \gamma_n(x)\right),$$ where $\gamma_n(x)$ reflects a local scale determined by moments or the kernel structure, often $\gamma_n(x)\sim n^{-1/r}\phi(x)^\alpha$ for some $\alpha$; this sharpens the error estimate compared to classical moduli, particularly near endpoints (Kajla et al., 2018, Deo et al., 4 Mar 2025, Rao et al., 2015, Singh et al., 2020). For example, for Bézier-variants of Bernstein–Durrmeyer operators on $[0,1]$,

$$|\widetilde D_{n,M}^{(2),p}(f;x)-f(x)| \leq C\, \omega_\varphi^2 \left(f, \frac{1}{\sqrt{n+2}}\right),$$

with $\varphi(x)=\sqrt{x(1-x)}$ (Singh et al., 2020).

The inverse theorem typically has the contrapositive form: if the error $|L_n(f;x)-f(x)|$ decays sufficiently fast (e.g., $O(n^{-\lambda})$), then $f$ necessarily lies in the Lipschitz–Zygmund or weighted space determined by $\omega_\phi^r(f,t) = O(t^\lambda)$ (Lu et al., 2010).

4. Modulus of Continuity for Endpoint Singularities and Weighted Spaces

Totik–Ditzian moduli natively accommodate singularities at endpoints through the step-weight $\phi$ and auxiliary weight $w(x)$,

$$\omega_\phi^r(f,t)_w := \sup_{\cdots} \|w \Delta_{h\phi}^r f(x)\|_{I^\prime},$$

where $I^\prime$ may be an interior or boundary strip and $\Delta_{h\phi}^r$ transitions between weighted and ordinary differences as $x$ approaches the endpoint. This mechanism allows functions with $f(x)\sim x^{-\alpha}$ or $f(x)\sim (1-x)^{-\beta}$, provided $w(x)$ absorbs the singularity (Lu et al., 2010).

5. Specialized Moduli and Adaptations to Unbounded Domains

On $[0,\infty)$, the choice $\phi(x)=\sqrt{x(1+x)}$ or $\phi(x)=\sqrt{x}$ matches the scaling of kernel variances for Szász, Baskakov, or Kantorovich–Charlier operators, yielding

$$\omega_\phi^2(f; t) = \sup_{0<h\le t} \sup_{x\pm h\phi(x)\in [0,\infty)} |f(x+h\phi(x))-2f(x)+f(x-h\phi(x))|.$$

For the Bézier–Gupta–Srivastava family one obtains

$$|L_n^{m,c,a}(f;x)-f(x)| \leq C \omega_\phi^{\beta}\left(f, \sqrt{M_{n,2}(x)}\right), \qquad \phi(x)=\sqrt{x(1+cx)},$$

with the second central moment $M_{n,2}(x)\sim x(1+cx)/n$ providing spatially adaptive rates (Pratap et al., 2018, Deo et al., 4 Mar 2025, Yadav et al., 2019).

6. Quantitative Voronovskaja-Type Theorems

Totik–Ditzian moduli underpin generalized Voronovskaja results, which characterize the main term in the asymptotic expansion and provide error bounds in the same scale. For generalized Bernstein–Durrmeyer operators,

$$\lim_{n\to\infty} n\left[\mathcal G_{n,\rho}^{(\alpha)}(f;x)-f(x)\right] = \frac{1-2x}{\rho}f'(x) + \frac{(1+\rho)x(1-x)}{2\rho}f''(x),$$

with remainder decay governed by $\omega_\phi^2(f; n^{-1/2})$ (Kajla et al., 2018). For modified Szász–Mirakjan–Kantorovich or (p,q)-Bernstein–Stancu operators, analogous expansions and quantitative remainders in DT-modulus scale are derived (Yadav et al., 2019, Khan et al., 2016).

7. Numerical Implications and Extensions

Maple or other computational illustrations confirm that operators designed to fit Totik–Ditzian estimates converge locallly at rates predicted by the modulus. Pointwise error curves are sharper and more uniform when measured in the DT-modulus rather than the classical modulus, especially for functions of bounded variation or with endpoint behavior (Kajla et al., 2018, Singh et al., 2020). The flexibility in step-weight and order allows the theory to be extended to weighted Lipschitz-type spaces, almost-Lipschitz classes, and function spaces characterized by smoothness in the DT sense (Deo et al., 4 Mar 2025, Lu et al., 2010).

Table: Canonical Step-Weights in Totik–Ditzian Modulus

Domain	Step-Weight $\phi(x)$	Operator Class
$[0,1]$	$\sqrt{x(1-x)}$	Bernstein, Bernstein–Durrmeyer
$[0,\infty)$	$\sqrt{x}$, $\sqrt{x(1+x)}$	Szász–Mirakjan, Baskakov, Charlier

These choices ensure that the modulus adapts to the intrinsic scaling of the kernels, leading to optimal direct and inverse approximation theorems for the relevant operator families (Rao et al., 2015, Wafi et al., 2015, Yadav et al., 2019, Deo et al., 4 Mar 2025).

Papers referenced in the above exposition include:

• "Generalized Bernstein-Durrmeyer Operators of Blending Type" (Kajla et al., 2018)
• "Durrmeyer type operators linked with Boas-Buck type polynomials" (Deo et al., 4 Mar 2025)
• "Stancu-variant of generalized Baskakov operators" (Rao et al., 2015)
• "Statistical approximation by $(p,q)$-analogue of Bernstein-Stancu Operators" (Khan et al., 2016)
• "Optimal bounds on the modulus of continuity of the uncentered Hardy-Littlewood maximal function" (Aldaz et al., 2010)
• "Approximation properties by some modified Szasz-Mirakjan-Kantorovich operators" (Yadav et al., 2019)
• "Kantorovich form of generalized Szasz-type operators with certain parameters using Charlier polynomials" (Wafi et al., 2015)
• "Direct and Inverse Estimates for Combinations of Bernstein Polynomials with Endpoint Singularities" (Lu et al., 2010)
• "Bézier Variant of generalized Bernstein-Durrmeyer type operators" (Singh et al., 2020)
• "Rate of convergence of Gupta-Srivastava operators based on certain parameters" (Pratap et al., 2018)

The Totik-Ditzian modulus remains indispensable as the canonical smoothness measure for modern approximation theory, providing the exact framework for optimal convergence rates, regularity characterizations, and analysis of positive linear operators on diverse domains and function spaces.