function spaces
Besov Spaces
You should know: sobolev spaces, interpolation spaces
Overview
Besov spaces B^s_{p,q}(R^n) are a scale of function spaces that interpolate between Sobolev spaces and provide precise control of smoothness. The parameter s is the smoothness exponent, p controls integrability, and q is a fine-tuning parameter. Besov spaces arise naturally as real interpolation spaces between Sobolev spaces: B^s_{p,q} = (W^{k,p}, W^{m,p})_{theta,q} for appropriate theta. They contain the Sobolev spaces W^{s,p} and Holder spaces C^s as special cases, and are essential for sharp embeddings, trace theorems, and the theory of elliptic PDE.
Intuition
Sobolev spaces W^{k,p} measure smoothness by requiring derivatives of order up to k to be in L^p. Besov spaces allow fractional smoothness s and add a second exponent q that distinguishes between functions with the same 'average' smoothness but different fine structure. The definition uses differences: a function has Besov smoothness s if its s-th order differences are in L^p in a quantitative way measured by the q-norm. This is like asking that the function is smooth in a 'Littlewood–Paley block' decomposition.
Formal Definition
The first-order modulus of continuity is omega_1(f,t)_p = sup_{|h|<=t} ||Delta_h f||_p where Delta_h f(x) = f(x+h) - f(x). For integer m > s, the Besov space B^s_{p,q}(R^n) consists of functions f in L^p with finite Besov norm. Using the real interpolation method: B^s_{p,q} = (W^{k_0,p}, W^{k_1,p})_{theta,q} where s = (1-theta)k_0 + theta k_1. Equivalently, via Littlewood–Paley decomposition: if f = sum_j phi_j * f (dyadic blocks), then ||f||_{B^s_{p,q}} = (sum_j 2^{jsq} ||phi_j * f||_p^q)^{1/q}.
Notation
| Notation | Meaning |
|---|---|
| Besov space with smoothness s, integrability p, fine-tuning q | |
| m-th order modulus of continuity in L^p | |
| Sobolev space (B^s_{p,2} for integer s) | |
| Holder space (B^s_{inf,inf}) |
Theorems
Worked Examples
- 1
Using the Littlewood–Paley characterisation: ||f||_{B^s_{2,2}}^2 = sum_j 2^{2js} ||phi_j*f||_{L^2}^2.
- 2
By Parseval, ||phi_j*f||_{L^2}^2 = ||hat(phi_j) hat(f)||_{L^2}^2 ~ integral_{2^j <= |xi| < 2^{j+1}} |hat(f)(xi)|^2 d xi.
- 3
Summing: ||f||_{B^s_{2,2}}^2 ~ sum_j 2^{2js} integral_{2^j <= |xi| < 2^{j+1}} |hat(f)|^2 d xi ~ integral (1+|xi|^2)^s |hat(f)(xi)|^2 d xi.
- 4
So B^s_{2,2} = H^s with equivalent norms.
✓ Answer
B^s_{2,2}(R^n) = H^s(R^n) (Sobolev space) for all s >= 0, with equivalent norms via the Fourier characterisation.
Practice Problems
Explain why B^s_{p,1} embeds into L^inf for s = n/p (the borderline Sobolev case) while W^{n/p,p} does not.
Which inclusion is CORRECT for Besov spaces with fixed s and p?
Quiz
Summary
- Besov spaces B^s_{p,q} refine Sobolev spaces with a three-parameter scale (s, p, q).
- Special cases: B^s_{2,2} = H^s (Sobolev), B^s_{inf,inf} = C^s (Holder), B^0_{2,2} = L^2.
- Real interpolation: B^s_{p,q} = (W^{k,p}, W^{m,p})_{theta,q} with s = (1-theta)k + theta m.
- Littlewood–Paley: ||f||_{B^s_{p,q}} ~ (sum_j 2^{jsq}||phi_j*f||_p^q)^{1/q}.
- Inclusion: B^s_{p,1} subset B^s_{p,2} subset B^s_{p,inf}; larger q = larger (less restrictive) space.
References
- BookTriebel, H. — Theory of Function Spaces, Birkhauser, 1983
- BookBergh, J. & Lofstrom, J. — Interpolation Spaces: An Introduction, Springer, 1976
- Websiteen.wikipedia.org
Mathematics