Lp Spaces and Norms

Understanding Lp spaces is fundamental to modern analysis, providing the rigorous language needed to discuss function size, convergence, and approximation. These spaces form the bedrock for solving differential equations, developing probability theory, and analyzing signals, turning vague notions of "well-behaved functions" into precise, quantifiable tools. Mastering their structure—the norms, inequalities, and completeness properties—unlocks the powerful techniques of functional analysis that underpin advanced mathematics and theoretical physics.

Defining Lp Spaces and the Lp Norm

At its core, an Lp space is a collection of functions where a specific kind of "average size" is finite. To define it rigorously, we start with a measure space $(X,\mathcal{F},\mu )$, where $X$ is a set, $\mathcal{F}$ a sigma-algebra, and $\mu$ a measure (common examples are ${\mathbb{R}}^n$ with Lebesgue measure). For a real number $1\le p<\infty$, we consider all measurable functions $f:X\to \mathbb{R}$ (or $\mathbb{C}$) whose p-th power is integrable.

Formally, we first define the set of functions with finite p-th power integral: $${\mathcal{L}}^p(X,\mu )=\{f\text{ measurable}:{\int }_X|f{|}^{p}d\mu <\infty \}.$$ However, if we define a "norm" on these functions directly, we encounter a problem: a function that is zero almost everywhere would have zero norm, but it is not the zero function. To create a proper normed vector space, we must identify functions that are equal almost everywhere. This leads to the definition of $L^p(X,\mu )$ as the set of equivalence classes of functions in ${\mathcal{L}}^p$, where two functions are equivalent if they are equal almost everywhere. This quotient construction is crucial; in $L^p$, we work with "point" that are actually families of functions, which simplifies the theory immensely.

On this space of equivalence classes, we define the Lp norm. For $1\le p<\infty$, the $L^p$ norm of a function (class) $f$ is given by: $$||f|{|}_p:={\left({\int }_X|f{|}^{p}d\mu \right)}^{1/p}.$$ For the borderline case $p=\infty$, we define the $L^{\infty }$ or "essential supremum" norm: $$||f|{|}_{\infty }:=\inf \{M\ge 0:|f(x)|\le M\text{ for almost every }x\in X\}.$$ This norm measures the smallest constant $M$ that bounds $|f|$ almost everywhere. The $L^p$ norm satisfies all the axioms of a norm: it is absolutely homogeneous ($||\alpha f|{|}_p=|\alpha |||f|{|}_p$), non-negative, zero only for the zero equivalence class, and, as we will see, satisfies the triangle inequality.

Foundational Inequalities: Hölder and Minkowski

The geometry and calculus of $L^p$ spaces are governed by two fundamental inequalities. These are not just technical tools but provide the very structure that makes these spaces workable.

Hölder's Inequality is a generalization of the Cauchy-Schwarz inequality (which is the special case $p=q=2$). It relates the integrals of products of functions. Let $1<p,q<\infty$ be conjugate exponents, meaning they satisfy the relation $\frac{1}{p}+\frac{1}{q}=1$. This convention also includes the case $p=1,q=\infty$. For $f\in L^p$ and $g\in L^q$, Hölder's inequality states: $$||fg|{|}_1={\int }_X|fg|d\mu \le ||f|{|}_p||g|{|}_q.$$ This inequality tells us that if $f$ is in $L^p$ and $g$ is in the conjugate space $L^q$, then their product is integrable ($L^1$). It is indispensable for estimating integrals and proving the triangle inequality for the $L^p$ norm.

That triangle inequality is precisely Minkowski's Inequality. For $1\le p\le \infty$ and $f,g\in L^p$, it states: $$||f+g|{|}_p\le ||f|{|}_p+||g|{|}_p.$$ Minkowski's inequality confirms that $||\cdot |{|}_p$ is indeed a norm and is the key to showing that $L^p$ is a vector space. The proof for $1<p<\infty$ typically relies on clever algebraic manipulation and an application of Hölder's inequality, showcasing how these two results are intertwined.

Completeness: Banach Spaces and the Riesz-Fischer Theorem

A normed space is called complete if every Cauchy sequence converges to a limit within the space. Completeness is a vital property because it guarantees that limits of convergent processes remain in the space, enabling the use of powerful analytical techniques. A complete normed space is called a Banach space.

The Riesz-Fischer theorem asserts that for any $1\le p\le \infty$, the space $L^p$ is complete; it is a Banach space. The proof is a classic construction in analysis. One takes a Cauchy sequence $\{f_n\}$ in $L^p$ and cleverly extracts a subsequence that converges pointwise almost everywhere to some limit function $f$. Using the Cauchy condition and Fatou's lemma, one then shows that this limit $f$ is indeed in $L^p$ and that the original sequence converges to $f$ in the $L^p$ norm. This completeness is non-trivial: the rational numbers are not complete under the usual absolute value norm, but $L^p$ spaces, built from equivalence classes, are.

This property makes $L^p$ spaces, particularly the all-important Hilbert space $L^2$, the perfect setting for solving integral equations and partial differential equations (PDEs) using methods of successive approximation.

Dual Spaces and Their Characterization

In functional analysis, the dual space of a Banach space $B$, denoted $B^{\ast }$, is the space of all continuous linear functionals on $B$. Understanding the dual of $L^p$ spaces provides deep insight into their structure and is critical for PDE theory, where solutions are often found by applying functionals to test functions.

For $1\le p<\infty$, the dual space of $L^p$ is isometrically isomorphic to $L^q$, where $q$ is the conjugate exponent ($1/p+1/q=1$). This means every continuous linear functional $\varphi :L^p\to \mathbb{R}$ can be uniquely represented by a function $g\in L^q$ via integration: $$\varphi (f)={\int }_{X}fgd\mu \text{for all }f\in L^p,$$ and furthermore, the operator norm of $\varphi$ equals $||g|{|}_q$. This beautiful result, a consequence of the Radon-Nikodym theorem, shows that Hölder's inequality is sharp and that $L^p$ spaces are reflexive for $1<p<\infty$ (meaning $(L^p{)}^{\ast }=L^q$ and $(L^q{)}^{\ast }=L^p$).

The case $p=\infty$ is different. The dual of $L^{\infty }$ is not $L^1$; it is a much larger space of finitely additive measures. This asymmetry underscores why the $1<p<\infty$ case is often more tractable in analysis.

Common Pitfalls

1. Ignoring the Equivalence Class Nature: Treating elements of $L^p$ as specific functions rather than equivalence classes leads to errors. Statements about "value at a point" are generally meaningless in $L^p$ for $p<\infty$. You must always reason up to sets of measure zero. For example, the function that is 1 at the origin and 0 elsewhere is in the same $L^p(\mathbb{R})$ equivalence class as the constant zero function, so its $L^p$ norm is zero.
2. Misapplying Hölder's Inequality: A common mistake is to use exponents $p$ and $q$ that are not conjugate ($1/p+1/q\ne 1$). This breaks the inequality. Always verify the exponent relationship first. Another error is forgetting that the inequality guarantees $fg\in L^1$, but it does not say that $f\in L^p$ and $g\in L^q$ is necessary for $fg\in L^1$; it is only a sufficient condition.
3. Confusing Dual Space Identification: Assuming $(L^{\infty }{)}^{\ast }=L^1$ is a critical error. This identification holds only for $1\le p<\infty$. For $p=\infty$, the dual space is more complex. Similarly, while $L^2$ is self-dual (a key feature of Hilbert spaces), this is a special case and does not generalize to other $p$.
4. Overlooking Completeness in Subsets: It's easy to assume a subset of a complete space is complete. The space $C([0,1])$ of continuous functions on $[0,1]$ with the $L^2$ norm is a subspace of $L^2([0,1])$, but it is not complete. A sequence of continuous functions can converge in $L^2$ norm to a discontinuous function, which is in $L^2$ but not in $C([0,1])$.

Summary

• Lp spaces are complete normed vector spaces (Banach spaces) whose elements are equivalence classes of measurable functions with finite p-th power integral, where two functions are equivalent if they are equal almost everywhere.
• The Lp norm $||f|{|}_p=(\int |f{|}^p{)}^{1/p}$ for $1\le p<\infty$ governs the "size" of a function, with the essential supremum defining the $L^{\infty }$ norm.
• Hölder's Inequality ($\int |fg|\le ||f|{|}_p||g|{|}_q$) and Minkowski's Inequality ($||f+g|{|}_p\le ||f|{|}_p+||g|{|}_p$) are the essential inequalities that define the space's analytic and geometric properties.
• The Riesz-Fischer Theorem guarantees the completeness of $L^p$ spaces, making them reliable settings for taking limits and solving equations.
• For $1<p<\infty$, the dual space $(L^p{)}^{\ast }$ is isometrically isomorphic to $L^q$ (with $1/p+1/q=1$), a cornerstone result for analysis in PDEs and optimization. The $L^1$ and $L^{\infty }$ duality is more complex.