Abstract Korovkin-type Approximation Theorem

• Abstract Korovkin-type Approximation Theorem is a unifying principle in operator theory, showing that convergence on a small test set ensures convergence on an entire function space.
• The framework generalizes classical results by extending positive linear operators to ordered Banach lattices, weighted spaces, and non-linear settings through statistical and summability convergence methods.
• Key applications include numerical analysis with Bernstein and q-generalizations, employing explicit error estimates via moduli of smoothness and advanced summability techniques.

An abstract Korovkin-type approximation theorem provides a unifying principle in operator approximation theory: uniform (or strong) convergence of an operator sequence on a small “Korovkin set” of test functions ensures convergence on the entire function space. Classical Korovkin theory, originally formulated for positive linear operators on $C([0,1])$, has, over the past decades, been generalized to encompass a wide variety of settings—ordered Banach lattices, weighted spaces, $L^p$-spaces, non-positive or nonlinear operators, and multiple convergence notions such as summability, statistical, or modular convergence. These generalizations have not only clarified the intrinsic structure of approximation but have also facilitated quantitative and qualitative results relevant for numerical analysis, functional analysis, operator theory, and applications in applied mathematics.

1. Classical Formulation and Generalizations

The classical Korovkin theorem asserts that a sequence $(L_n)$ of positive linear operators $L_n:C([0,1]) \to C([0,1])$ converges uniformly to the identity if and only if

$$\lim_{n\to\infty} L_n(1) = 1, \quad \lim_{n\to\infty} L_n(t) = t, \quad \lim_{n\to\infty} L_n(t^2) = t^2,$$

for every $f \in C([0,1])$, it holds that $L_n(f) \to f$ uniformly.

The abstraction and generalization of this theorem, as developed by Altomare, Campiti, and collaborators (Altomare, 2010), consider ordered Banach lattices and broader spaces such as:

• $C_0(X)$: continuous functions vanishing at infinity, for $X$ locally compact.
• Weighted spaces: $C_0^w(X)$ with a weight $w$.
• $L^p(X,\mu)$ spaces, relying on the density of $C_0(X)$ in $L^p$ for finite regular Borel measure $\mu$.

A “Korovkin set” $K\subset C_0(X)$ satisfying that if an equibounded sequence of positive linear operators $(L_n)$ converges on $K$, then $L_n(f)\to f$ for all $f\in C_0(X)$, is typically constructed as

$\{f_0\} \cup f_0M \cup f_0M^2,$

with $f_0>0$ strictly positive, $M \subset C_0(X)$ separating points.

Further generalizations (Mahmudov, 2010, Mahmudov, 2011, Gal et al., 2022, Gal et al., 2021, Gal et al., 2022, Kumar et al., 2022) replace positivity and linearity with monotonicity, sublinearity, or strong translatability, and positivity is shown not to be strictly necessary for certain operator classes.

2. Extended Frameworks: Summability, Statistical, and Modular Convergence

To accommodate contexts where classical convergence fails or is inappropriate, substantial work has been devoted to replacing uniform convergence with weaker or alternative modes:

• Statistical Convergence: Approximation is controlled on a “density one” subsequence, with Korovkin-type theorems holding via statistical convergence for both classical and weighted spaces (Sontakke et al., 2016, Yadav et al., 2019).
• Summability (e.g., Power Series, Matrix Methods, Ideals): Various summability methods $R$ are considered; the key is that $R$ must preserve inequalities or order-inequalities for the Korovkin conclusion to hold (Listán-García et al., 2023). For instance, if

$$|L_n f(x) - f(x)| \leq C\big(|L_n 1(x) - 1| + |L_n t(x) - x| + |L_n t^2(x) - x^2|\big),$$

and all test functions converge to their target via $R$-summability, then the $R$-limit of $L_n f$ is $f$.

• Power Series Statistical Convergence: An even finer method, utilizing regular power series methods $P$, leads to $P$-statistical convergence results, foundational in (Söylemez et al., 2 Sep 2025), where operator convergence is shown via density notions defined by the power series coefficients.

Additional settings include:

• Modular Convergence: In modular spaces (e.g., Orlicz or $L^0$), convergence and quantitative theorems are obtained for nets of operators via modular functionals (Boccuto et al., 2021).
• Abstract axiomatic frameworks: The convergence mode is specified axiomatically, unifying classical, statistical, filter, and almost convergence.

3. Operator Classes and Non-Positivity

A key direction has been relaxing the requirement of operator positivity:

• Non-positive Operators: Results by Zeren et al. (Kumar et al., 2022) and Wulbert (Kumar et al., 2024) establish that, for uniformly bounded sequences of linear operators, Korovkin-type theorems can be established under suitable convergence on the test set, even in the absence of positivity.
• Nonlinear and Sublinear Operators: Sublinear, monotone, and (strongly) translatable or comonotone additive operators are shown to satisfy Korovkin-type theorems under sublinearity and monotonicity (Gal et al., 2021, Gal et al., 2022, Gal et al., 2022). The test function structure is retained (e.g., $\{1, x, x^2\}$, or appropriate projections), and rates of convergence are often expressed via (higher order) moduli of smoothness.

4. Quantitative Estimates and Moduli of Smoothness

Current theory emphasizes not only qualitative convergence but also explicit rates, with error estimates typically involving:

• First and Second modulus of continuity:

$$\omega(f, \delta) = \sup_{|x - y| < \delta} |f(x) - f(y)|$$

with corresponding higher-order moduli $\omega_2$, $\omega_3$.

• Weighted Moduli: For unbounded domains or weighted spaces, moduli adapted to the space (e.g., $\omega_{\varphi}(f, \delta)$) are used.
• Lipschitz Space Error Control: For $f$ in classes $\mathrm{Lip}_M(\nu)$, convergence rates are controlled by the moment structure of the operator kernel (Mursaleen et al., 2016, Çekim et al., 2016, Kaur et al., 2020).
• Peetre's $K$-functionals: Further refine quantitative analysis in terms of $$K_2(f, \delta) \leq C\{\min(1, \delta)\|f\| + \omega_2(f, \delta)\}$$ (Çekim et al., 2016).

5. Key Applications and Examples

Korovkin-type theory underpins much of constructive approximation and numerical analysis. Applications include:

• Bernstein, Szász–Mirakjan, Kanotorovich, Meyer–König and Zeller (MKZ), Stancu, and various $q$- and $(p,q)$-generalizations: These include classical results and modern quantum calculus extensions (Altomare, 2010, Mursaleen et al., 2016, Sontakke et al., 2016).
• Weighted Spaces and Unbounded Domains: Approximation in spaces $C_0^w(X)$, $L^p(X, \mu)$, and various generalized Lebesgue spaces (Kumar et al., 2022, Kaur et al., 2020).
• Operator Iterates and Projections: Results on convergence to projections and limit operators, including the Bernstein–Schnabl projection (Altomare, 2010, Mahmudov, 2010, Mahmudov, 2011).
• Convolution Operators: Fejér and Abel–Poisson operators for periodic functions, with ties to Fourier analysis (Altomare, 2010, Yavuz et al., 2017).
• Non-commutative/Operator-Algebraic Settings: Completely positive maps and spectral clustering convergence for operator approximation (notably in Toeplitz pre-conditioning) (Kumar et al., 2012).
• Extensions to Fuzzy Analysis: Power series summability methods applied in fuzzy Korovkin theory, with metrics given by the Hausdorff distance on fuzzy number level sets (Baxhaku et al., 2022).

6. Equivalence with Stone–Weierstrass and Banach Lattice Theory

A major structural insight is the equivalence between Korovkin-type theorems and the Stone–Weierstrass theorem. A subalgebra $A\subset C_0(X)$ that is dense (Stone–Weierstrass) is also a Korovkin set (Altomare, 2010). The use of Banach lattice structures induces powerful transfer principles, with lattice homomorphisms, positive projections, and measure-theoretic tools (e.g., Radon measures via Riesz representation) providing a unification across disparate approximation settings.

7. Recent Advances: $P$-Statistical Framework and Operator Generalizations

The most recent advances include:

• Power Series Statistical Korovkin Theory: The $P$-statistical Korovkin theorem (Söylemez et al., 2 Sep 2025) asserts: for $T_n$ a sequence of positive linear operators on $C(X)$, if

$$\mathrm{st}_P-\lim_n \|T_n(\psi_\ell) - \psi_\ell\| = 0 \qquad \forall \ell = 1,\dots,d$$

for a Korovkin set $\{\psi_\ell\}$, then

$$\mathrm{st}_P-\lim_n \|T_n(\psi) - \psi\| = 0 \qquad \forall \psi \in C(X)$$

Thus, approximation is guaranteed outside a set of indices negligible with respect to the $P$-density induced by the power series. A crucial decomposition result is that every $P$-statistically convergent sequence contains a classical convergent subsequence on a density-1 set, reducing the proof of operator approximation to classical arguments on large subsequences.

• Operators Beyond Positivity: The $r$-th order generalization of classical positive operators, often losing positivity, still enables Korovkin-type approximation in the $P$-statistical sense, provided suitable control on the higher derivatives of the target functions (e.g., $r$-th derivative in a Lipschitz class).

Table: Korovkin-type Theorems Across Settings

Setting	Test Set / Functions	Mode of Convergence
$C([0,1])$, positive linear	$\{1, t, t^2\}$	Uniform
$C_0(X)$, weighted spaces	$\{f_0\} \cup f_0M \cup f_0 M^2$	Uniform/Weighted
$L^p$, Banach lattices	$\{1, t, t^2\}$ (dense)	$L^p$-norm
Nonlinear/Monotone/Sublinear	$\{1, \pm pr_1, \dots, pr_N, pr_1^2, \dots\}$	Uniform/$L^p$/a.e./in measure
Summability/statistical	Korovkin set appropriate	Statistical, $P$-statistical, matrix, or Abel summability

Conclusion

Abstract Korovkin-type approximation theorems provide a singularly powerful paradigm for operator approximation. They show that checking convergence (in an appropriate sense) on a small, algebraically significant set of test functions suffices to ensure convergence for a much broader class of functions, across a diversity of operator types, function spaces, and convergence modes. Recent developments employing summability and statistical methods, as well as the accommodation of non-pos. and nonlinear operators, cement the deep flexibility of the framework and extend its reach to new domains within and beyond classical analysis (Altomare, 2010, Mahmudov, 2010, Mahmudov, 2011, Gal et al., 2021, Gal et al., 2022, Listán-García et al., 2023, Kumar et al., 2024, Söylemez et al., 2 Sep 2025).