Function spaces, norms, completeness, and the Hölder and Minkowski inequalities
The Lp spaces are the function spaces that emerge naturally from the Lebesgue integral. For 1 ≤ p < ∞, the space Lp(X, μ) consists of all measurable functions f : X → ℝ (or ℂ) such that ∫ |f|ᵖ dμ < ∞, with the norm ‖f‖_p = (∫ |f|ᵖ dμ)^{1/p}. For p = ∞, L∞ consists of essentially bounded functions with the essential supremum norm ‖f‖_∞ = ess sup |f|.
Technically, elements of Lp are not individual functions but equivalence classes of functions that agree almost everywhere. Two functions f and g with f = g a.e. represent the same element of Lp. This identification is necessary to make ‖·‖_p a true norm (not just a seminorm): ‖f‖_p = 0 implies f = 0 a.e., which means f = 0 in Lp.
Lp spaces are the natural habitat for studying convergence, approximation, and duality in analysis. They appear throughout mathematics: in PDE theory (weak solutions live in Sobolev spaces built from Lp), in probability (moments of random variables are Lp norms), in signal processing (L² is the space of finite-energy signals), and in quantum mechanics (L² is the Hilbert space of wave functions).
The parameter p controls how the norm weights large versus small values of |f|. For small p (close to 1), the norm is relatively tolerant of large spikes — it integrates |f|ᵖ, and raising to a low power dampens peaks. For large p, the norm heavily penalizes large values, and as p → ∞, the Lp norm converges to the essential supremum ‖f‖_∞ = ess sup |f|. This interpolation between "average size" (L¹) and "worst case" (L∞) is one of the most useful features of the Lp scale.
The unit ball {f : ‖f‖_p ≤ 1} changes shape with p. In finite dimensions, the L¹ unit ball is a cross-polytope (diamond), the L² ball is the Euclidean ball, and the L∞ ball is a cube. As p increases, the ball "inflates" from the diamond toward the cube. This geometric picture helps build intuition for how different Lp norms measure size.
Key insight: The choice of p reflects a trade-off. L¹ treats all values proportionally (good for sparse functions). L² minimizes squared error (good for energy and least-squares). L∞ controls the maximum deviation (good for uniform approximation). Choosing the right p for your problem is an art.
The Riesz-Fischer theorem states that for 1 ≤ p ≤ ∞, every Cauchy sequence in Lp converges to an element of Lp. In other words, Lp is a Banach space — a complete normed vector space. Completeness is the single most important structural property of Lp spaces, and it is a direct consequence of the Lebesgue integral's convergence theorems.
The proof uses a clever subsequence argument: given a Cauchy sequence (fₙ) in Lp, extract a subsequence (fₙₖ) with ‖fₙₖ₊₁ − fₙₖ‖_p < 2⁻ᵏ. The telescoping series fₙ₁ + Σₖ (fₙₖ₊₁ − fₙₖ) converges absolutely in Lp (by comparison with a geometric series), and the Monotone Convergence Theorem shows it converges almost everywhere to some f ∈ Lp. Then ‖fₙ − f‖_p → 0 by the triangle inequality.
Completeness is essential for applying the major theorems of functional analysis — the Banach-Steinhaus theorem, the open mapping theorem, the closed graph theorem — to Lp spaces. It is also what makes Lp spaces suitable as solution spaces for differential equations: completeness guarantees that limit procedures (e.g., Galerkin approximation) converge to actual solutions, not phantom limits that escape the space.
Hölder's inequality: If 1 ≤ p, q ≤ ∞ with 1/p + 1/q = 1 (conjugate exponents), then for f ∈ Lp and g ∈ Lq, the product fg is in L¹ and ∫ |fg| dμ ≤ ‖f‖_p · ‖g‖_q. The special case p = q = 2 is the Cauchy-Schwarz inequality. Hölder's inequality is the fundamental estimate for products of functions in different Lp spaces.
Minkowski's inequality: For 1 ≤ p ≤ ∞ and f, g ∈ Lp, ‖f + g‖_p ≤ ‖f‖_p + ‖g‖_p. This is the triangle inequality for the Lp norm, and it is what makes ‖·‖_p a genuine norm (not just a function). The proof for 1 < p < ∞ uses Hölder's inequality applied to |f + g|ᵖ = |f + g|^{p-1} · |f + g| ≤ |f + g|^{p-1} · (|f| + |g|).
∫|fg| ≤ ||f||_p · ||g||_q
||f+g||_p ≤ ||f||_p + ||g||_p
Key insight: Hölder's inequality tells us which function spaces pair naturally: Lp and Lq are "dual" when 1/p + 1/q = 1. This duality is made precise by the Riesz representation theorem, which identifies (Lp)* with Lq for 1 < p < ∞. The interplay between a space and its dual is central to modern analysis.
Among all Lp spaces, L² is special: it is the only one that is a Hilbert space. The inner product ⟨f, g⟩ = ∫ f · g̅ dμ (where g̅ denotes complex conjugation) satisfies all the axioms of an inner product, and the induced norm ‖f‖₂ = ⟨f, f⟩^{1/2} = (∫ |f|² dμ)^{1/2} is the L² norm. The Cauchy-Schwarz inequality |⟨f, g⟩| ≤ ‖f‖₂ · ‖g‖₂ is the p = 2 case of Hölder.
The inner product structure gives L² geometric features that other Lp spaces lack: orthogonality, projections, and orthonormal bases. Every separable Hilbert space is isometrically isomorphic to ℓ² (the space of square-summable sequences), so L²(X, μ) is "the same space" as ℓ² up to a choice of orthonormal basis. This is the content of the Riesz-Fischer theorem in its Hilbert space form.
Fourier analysis lives naturally in L². The Fourier transform is a unitary operator on L²(ℝ) (Plancherel's theorem), and Fourier series provide orthonormal bases for L²([0, 2π]). Parseval's identity ‖f‖₂² = Σₙ |⟨f, eₙ⟩|² expresses the L² norm in terms of Fourier coefficients, connecting the time and frequency domains.
In quantum mechanics, the state space of a quantum system is a separable Hilbert space, typically L²(ℝ³). Observables are self-adjoint operators, and measurements correspond to projections onto eigenspaces. The entire mathematical framework of quantum theory rests on the Hilbert space structure of L² — a direct descendant of Lebesgue's integral.