Papers
Topics
Authors
Recent
Search
2000 character limit reached

Vanishing Mean Oscillation (VMO)

Updated 7 February 2026
  • VMO functions are a subspace of BMO where the local mean oscillation vanishes at small (and sometimes large) scales, ensuring refined regularity.
  • They are characterized through quantitative criteria such as Carleson measures, semigroup smoothing, and translation continuity, providing robust operator invariance.
  • VMO plays a crucial role in analysis, influencing the study of PDE regularity, spectral analysis, and various function space decompositions.

Vanishing Mean Oscillation (VMO) functions are a central concept in harmonic analysis, PDE, and modern function space theory. VMO is the natural subspace of bounded mean oscillation (BMO) where the local mean oscillation “vanishes” at small scales and, in global contexts, at large scales or far from a basepoint. VMO arises in quantitative regularity theory, singular integral operator theory, spectral analysis on metric spaces, and complex/real analysis. The precise delineation of VMO as the closure of continuous (or uniformly continuous) functions in the BMO-norm is due to Sarason, and multiple operator and geometric characterizations (involving Carleson measures, singular integrals, semigroups, maximal functions, and tent spaces) have been established.

1. Foundational Definitions and Characterizations

Space Definition (Euclidean Setting): For fLloc1(Rn)f\in L^1_{\text{loc}}(\mathbb{R}^n),

fBMO=supQRn1QQf(x)fQdx<,fQ=1QQf,\|f\|_{BMO} = \sup_{Q\subset\mathbb{R}^n} \frac{1}{|Q|} \int_Q |f(x)-f_Q|\,dx <\infty,\qquad f_Q = \frac{1}{|Q|}\int_Q f,

and fVMO(Rn)f\in VMO(\mathbb{R}^n) if in addition

lim(Q)01QQf(x)fQdx=0.\lim_{\ell(Q)\to0} \frac{1}{|Q|} \int_Q |f(x)-f_Q|\,dx = 0.

The space is defined modulo additive constants.

Sarason's Theorem: VMO(Rn)VMO(\mathbb{R}^n) is the closure in BMO of the set of uniformly continuous functions with bounded mean oscillation: VMO(Rn)=UC(Rn)BMO(Rn)BMO.VMO(\mathbb{R}^n) = \overline{UC(\mathbb{R}^n)\cap BMO(\mathbb{R}^n)}^{\|\cdot\|_{BMO}}. Analogous definitions and density/closure results hold on general domains Ω\Omega and for function spaces of John-Nirenberg type (Martell et al., 2016, Butaev et al., 2022, Korte et al., 2023).

Alternative Local Formulation: On domains (possibly unbounded), fVMO(Ω)f\in VMO(\Omega) iff the modulus of mean oscillation ωΩ(f,t)0\omega_\Omega(f, t)\to0 as t0t\to0, where

fBMO=supQRn1QQf(x)fQdx<,fQ=1QQf,\|f\|_{BMO} = \sup_{Q\subset\mathbb{R}^n} \frac{1}{|Q|} \int_Q |f(x)-f_Q|\,dx <\infty,\qquad f_Q = \frac{1}{|Q|}\int_Q f,0

Equivalent characterizations use balls instead of cubes or uniform small-scale estimates (Butaev et al., 2022, Butaev et al., 2018).

2. Quantitative and Operator-theoretic Criteria

Carleson Measure Criteria: A function fBMO=supQRn1QQf(x)fQdx<,fQ=1QQf,\|f\|_{BMO} = \sup_{Q\subset\mathbb{R}^n} \frac{1}{|Q|} \int_Q |f(x)-f_Q|\,dx <\infty,\qquad f_Q = \frac{1}{|Q|}\int_Q f,1 lies in fBMO=supQRn1QQf(x)fQdx<,fQ=1QQf,\|f\|_{BMO} = \sup_{Q\subset\mathbb{R}^n} \frac{1}{|Q|} \int_Q |f(x)-f_Q|\,dx <\infty,\qquad f_Q = \frac{1}{|Q|}\int_Q f,2 if and only if, for the Poisson (or generalized elliptic fBMO=supQRn1QQf(x)fQdx<,fQ=1QQf,\|f\|_{BMO} = \sup_{Q\subset\mathbb{R}^n} \frac{1}{|Q|} \int_Q |f(x)-f_Q|\,dx <\infty,\qquad f_Q = \frac{1}{|Q|}\int_Q f,3-Poisson) extension fBMO=supQRn1QQf(x)fQdx<,fQ=1QQf,\|f\|_{BMO} = \sup_{Q\subset\mathbb{R}^n} \frac{1}{|Q|} \int_Q |f(x)-f_Q|\,dx <\infty,\qquad f_Q = \frac{1}{|Q|}\int_Q f,4 of fBMO=supQRn1QQf(x)fQdx<,fQ=1QQf,\|f\|_{BMO} = \sup_{Q\subset\mathbb{R}^n} \frac{1}{|Q|} \int_Q |f(x)-f_Q|\,dx <\infty,\qquad f_Q = \frac{1}{|Q|}\int_Q f,5 to the upper half-space,

fBMO=supQRn1QQf(x)fQdx<,fQ=1QQf,\|f\|_{BMO} = \sup_{Q\subset\mathbb{R}^n} \frac{1}{|Q|} \int_Q |f(x)-f_Q|\,dx <\infty,\qquad f_Q = \frac{1}{|Q|}\int_Q f,6

where fBMO=supQRn1QQf(x)fQdx<,fQ=1QQf,\|f\|_{BMO} = \sup_{Q\subset\mathbb{R}^n} \frac{1}{|Q|} \int_Q |f(x)-f_Q|\,dx <\infty,\qquad f_Q = \frac{1}{|Q|}\int_Q f,7 is the Carleson box above fBMO=supQRn1QQf(x)fQdx<,fQ=1QQf,\|f\|_{BMO} = \sup_{Q\subset\mathbb{R}^n} \frac{1}{|Q|} \int_Q |f(x)-f_Q|\,dx <\infty,\qquad f_Q = \frac{1}{|Q|}\int_Q f,8 (Martell et al., 2016). This extends to complex-valued, vector-valued, and system Poisson kernels and is the canonical Dirichlet "vanishing Carleson" characterization (Martell et al., 2016, Lu et al., 2023, Liu et al., 2021).

Semigroup and Singular Integral Invariance: For any semi-Calderón–Zygmund operator fBMO=supQRn1QQf(x)fQdx<,fQ=1QQf,\|f\|_{BMO} = \sup_{Q\subset\mathbb{R}^n} \frac{1}{|Q|} \int_Q |f(x)-f_Q|\,dx <\infty,\qquad f_Q = \frac{1}{|Q|}\int_Q f,9 (i.e., with fVMO(Rn)f\in VMO(\mathbb{R}^n)0),

fVMO(Rn)f\in VMO(\mathbb{R}^n)1

Moreover, fVMO(Rn)f\in VMO(\mathbb{R}^n)2 if and only if fVMO(Rn)f\in VMO(\mathbb{R}^n)3 for every such fVMO(Rn)f\in VMO(\mathbb{R}^n)4; equivalently, for Riesz transforms or other algebra-generating singular integrals (Martell et al., 2016, Lu et al., 2023, Cao et al., 2020).

Poisson and Translation Continuity: For fVMO(Rn)f\in VMO(\mathbb{R}^n)5, the following are equivalent to fVMO(Rn)f\in VMO(\mathbb{R}^n)6 (Lu et al., 2023):

  • fVMO(Rn)f\in VMO(\mathbb{R}^n)7 as fVMO(Rn)f\in VMO(\mathbb{R}^n)8 (translation continuity).
  • fVMO(Rn)f\in VMO(\mathbb{R}^n)9 as lim(Q)01QQf(x)fQdx=0.\lim_{\ell(Q)\to0} \frac{1}{|Q|} \int_Q |f(x)-f_Q|\,dx = 0.0 (Poisson smoothing).

Tent Space and Operator-adapted VMO: If lim(Q)01QQf(x)fQdx=0.\lim_{\ell(Q)\to0} \frac{1}{|Q|} \int_Q |f(x)-f_Q|\,dx = 0.1 is a metric-measure space operator with Davies–Gaffney estimates and bounded lim(Q)01QQf(x)fQdx=0.\lim_{\ell(Q)\to0} \frac{1}{|Q|} \int_Q |f(x)-f_Q|\,dx = 0.2 functional calculus, lim(Q)01QQf(x)fQdx=0.\lim_{\ell(Q)\to0} \frac{1}{|Q|} \int_Q |f(x)-f_Q|\,dx = 0.3 consists of functions lim(Q)01QQf(x)fQdx=0.\lim_{\ell(Q)\to0} \frac{1}{|Q|} \int_Q |f(x)-f_Q|\,dx = 0.4 whose normalized local oscillations (defined via lim(Q)01QQf(x)fQdx=0.\lim_{\ell(Q)\to0} \frac{1}{|Q|} \int_Q |f(x)-f_Q|\,dx = 0.5 smoothing) vanish on small and large balls and outside large balls centered at a basepoint. Characterization is via vanishing tent space conditions on a suitable conical square function lim(Q)01QQf(x)fQdx=0.\lim_{\ell(Q)\to0} \frac{1}{|Q|} \int_Q |f(x)-f_Q|\,dx = 0.6 (Liang et al., 2011).

3. Structure, Duality, and Extension Properties

Density and Closure:

  • lim(Q)01QQf(x)fQdx=0.\lim_{\ell(Q)\to0} \frac{1}{|Q|} \int_Q |f(x)-f_Q|\,dx = 0.7 is the closure in BMO of lim(Q)01QQf(x)fQdx=0.\lim_{\ell(Q)\to0} \frac{1}{|Q|} \int_Q |f(x)-f_Q|\,dx = 0.8 or lim(Q)01QQf(x)fQdx=0.\lim_{\ell(Q)\to0} \frac{1}{|Q|} \int_Q |f(x)-f_Q|\,dx = 0.9 (Butaev et al., 2022, Lu et al., 2023, Butaev et al., 2018).
  • For John-Nirenberg generalization VMO(Rn)VMO(\mathbb{R}^n)0, both vanishing subspaces VMO(Rn)VMO(\mathbb{R}^n)1 and VMO(Rn)VMO(\mathbb{R}^n)2 coincide and equal the closure of relevant smooth/differentiable classes in the VMO(Rn)VMO(\mathbb{R}^n)3-norm (Korte et al., 2023).

Extension Operator Characterization: On a domain VMO(Rn)VMO(\mathbb{R}^n)4, there is a bounded linear extension VMO(Rn)VMO(\mathbb{R}^n)5 if and only if VMO(Rn)VMO(\mathbb{R}^n)6 is uniform (Jones' condition) (Butaev et al., 2018, Butaev et al., 2022).

Duality: VMO(Rn)VMO(\mathbb{R}^n)7 is the predual of the respective Hardy space (VMO(Rn)VMO(\mathbb{R}^n)8 for classical VMO(Rn)VMO(\mathbb{R}^n)9; VMO(Rn)=UC(Rn)BMO(Rn)BMO.VMO(\mathbb{R}^n) = \overline{UC(\mathbb{R}^n)\cap BMO(\mathbb{R}^n)}^{\|\cdot\|_{BMO}}.0 for Orlicz–Hardy VMO(Rn)=UC(Rn)BMO(Rn)BMO.VMO(\mathbb{R}^n) = \overline{UC(\mathbb{R}^n)\cap BMO(\mathbb{R}^n)}^{\|\cdot\|_{BMO}}.1-adapted settings), i.e.

VMO(Rn)=UC(Rn)BMO(Rn)BMO.VMO(\mathbb{R}^n) = \overline{UC(\mathbb{R}^n)\cap BMO(\mathbb{R}^n)}^{\|\cdot\|_{BMO}}.2

where the dual pairing is via VMO(Rn)=UC(Rn)BMO(Rn)BMO.VMO(\mathbb{R}^n) = \overline{UC(\mathbb{R}^n)\cap BMO(\mathbb{R}^n)}^{\|\cdot\|_{BMO}}.3 inner product and VMO(Rn)=UC(Rn)BMO(Rn)BMO.VMO(\mathbb{R}^n) = \overline{UC(\mathbb{R}^n)\cap BMO(\mathbb{R}^n)}^{\|\cdot\|_{BMO}}.4 is the Banach completion of the molecular Orlicz–Hardy space (Liang et al., 2011, Cao et al., 2020).

4. VMO Associated with Operators, Weights, and General Structures

VMO for General Operators: For VMO(Rn)=UC(Rn)BMO(Rn)BMO.VMO(\mathbb{R}^n) = \overline{UC(\mathbb{R}^n)\cap BMO(\mathbb{R}^n)}^{\|\cdot\|_{BMO}}.5 satisfying Davies–Gaffney and VMO(Rn)=UC(Rn)BMO(Rn)BMO.VMO(\mathbb{R}^n) = \overline{UC(\mathbb{R}^n)\cap BMO(\mathbb{R}^n)}^{\|\cdot\|_{BMO}}.6 functional calculus, VMO(Rn)=UC(Rn)BMO(Rn)BMO.VMO(\mathbb{R}^n) = \overline{UC(\mathbb{R}^n)\cap BMO(\mathbb{R}^n)}^{\|\cdot\|_{BMO}}.7 encompasses operator-adapted oscillation vanishing; molecular Hardy and tent-space machinery provide representation and duality (Liang et al., 2011).

Neumann Laplacian Setting: VMO(Rn)=UC(Rn)BMO(Rn)BMO.VMO(\mathbb{R}^n) = \overline{UC(\mathbb{R}^n)\cap BMO(\mathbb{R}^n)}^{\|\cdot\|_{BMO}}.8 is characterized by vanishing VMO(Rn)=UC(Rn)BMO(Rn)BMO.VMO(\mathbb{R}^n) = \overline{UC(\mathbb{R}^n)\cap BMO(\mathbb{R}^n)}^{\|\cdot\|_{BMO}}.9-oscillation of Ω\Omega0 over cubes as side Ω\Omega1, Ω\Omega2, or far from the origin, extending reflection principles and BMO–Hardy duality (Cao et al., 2020).

Muckenhoupt Ω\Omega3-Weights and VMO: For weights Ω\Omega4, Ω\Omega5 if and only if a certain associated geometric Carleson measure on the upper half-plane is vanishing and Ω\Omega6 is vanishing-doubling, parallel to Ω\Omega7–Ω\Omega8 duality but with vanishing mass at small scales; this supports characterizations of symmetric homeomorphisms and weight theory (Liu et al., 2021).

5. Special Function Classes and Geometric/Analytic Perspectives

Plurisubharmonic Functions and Lelong Numbers: A plurisubharmonic function Ω\Omega9 on a bounded domain fVMO(Ω)f\in VMO(\Omega)0 is in fVMO(Ω)f\in VMO(\Omega)1 if and only if its Lelong number vanishes at every point (i.e., absence of logarithmic singularities). This characterization connects VMO to singularity order analysis in complex variables and pluripotential theory (Biard et al., 2024).

Function Rearrangement and VMO: The decreasing rearrangement fVMO(Ω)f\in VMO(\Omega)2 and symmetric decreasing rearrangement preserve VMO, and continuity of the rearrangement map holds at fVMO(Ω)f\in VMO(\Omega)3 points (under BMO and fVMO(Ω)f\in VMO(\Omega)4 convergence) but not in general for BMO (Burchard et al., 2022).

Maximal Operators on Metric Spaces: In bounded doubling metric measure spaces, the fractional maximal operator fVMO(Ω)f\in VMO(\Omega)5 maps fVMO(Ω)f\in VMO(\Omega)6 into itself (under an annular decay condition on the measure). However, fVMO(Ω)f\in VMO(\Omega)7 is not continuous as an operator fVMO(Ω)f\in VMO(\Omega)8 or fVMO(Ω)f\in VMO(\Omega)9 (Gibara et al., 2023).

Semigroup and Analytic Function Spaces: For ωΩ(f,t)0\omega_\Omega(f, t)\to00 and ωΩ(f,t)0\omega_\Omega(f, t)\to01 on the disc, Sarason's theorem uses strong continuity with respect to rotation semigroups to characterize ωΩ(f,t)0\omega_\Omega(f, t)\to02, generalizing to classes of holomorphic semigroups via logarithmic vanishing oscillation conditions on the generator (LVMO/LBB), and linking the structure of vanishing mean oscillation to the Bloch and little Bloch spaces (Chalmoukis et al., 2021).

Representation and Decomposition: Every ωΩ(f,t)0\omega_\Omega(f, t)\to03 function admits a decomposition analogous to the Fefferman–Stein or Carleson representation: on ωΩ(f,t)0\omega_\Omega(f, t)\to04, ωΩ(f,t)0\omega_\Omega(f, t)\to05 with ωΩ(f,t)0\omega_\Omega(f, t)\to06 continuous and ωΩ(f,t)0\omega_\Omega(f, t)\to07 a vanishing Carleson measure; in ωΩ(f,t)0\omega_\Omega(f, t)\to08, as sums of bounded uniformly continuous functions and Riesz transforms applied to such functions (Lu et al., 2023).

6. Applications and Examples

Critical Points and Carleson Measures: Using VMO multipliers, any Carleson measure on the disc can be made vanishing Carleson, enabling factorization of analytic functions so that critical points are preserved with boundary values in VMO, with applications to Hardy and BMOA spaces, Volterra operators, and invariant subspace theory (Bellavita et al., 30 Sep 2025).

Regularity for Nonlocal PDE: For nonlocal double-phase equations with VMO coefficients, VMO regularity assumptions permit approximation by translation-invariant comparison problems, yielding higher Hölder regularity of solutions and showing the critical role of mean-oscillation decay in variable coefficient elliptic and nonlocal regularity (Byun et al., 2023).

Examples:

  • The constant function and all uniformly continuous functions lie in ωΩ(f,t)0\omega_\Omega(f, t)\to09.
  • Oscillatory functions (e.g., t0t\to00 on small intervals) may be in t0t\to01 with infinite oscillation frequency near t0t\to02.
  • Any t0t\to03 extends to t0t\to04, demonstrating flexibility compared to BMO.

7. Broader Context and Further Developments

VMO is the endpoint regularity class for functions whose mean oscillation decays at small scales, with extensions to metric settings and operator-adapted spaces. VMO controls critical boundary behavior in regularity theory, is stable under a wide class of singular integrals, and supports powerful approximation, extension, and decomposition structures. The theory admits deep connections to Hardy spaces, Carleson measures, Muckenhoupt weights, pluripotential theory, function rearrangement, and maximal operators. Quantitative and geometric criteria—via tent spaces, vanishing Carleson measures, and operator-smoothness—provide tools for ongoing research in PDE, harmonic analysis, complex analysis, and geometric function theory (Martell et al., 2016, Lu et al., 2023, Liu et al., 2021, Korte et al., 2023, Liang et al., 2011, Bellavita et al., 30 Sep 2025).

Topic to Video (Beta)

No one has generated a video about this topic yet.

Whiteboard

No one has generated a whiteboard explanation for this topic yet.

Follow Topic

Get notified by email when new papers are published related to Vanishing Mean Oscillation (VMO) Functions.