Sobolev Spaces and Weak Derivatives

Motivation: Derivatives for "Rough" Functions

Classical derivatives require smoothness. But problems in mechanics and physics have solutions with discontinuities: a load with a jump creates a "kink" in the deflection of the beam. Sobolev spaces generalize the notion of derivative to $L^2$-functions via the integration by parts formula. This is the language of the theory of weak solutions to differential equations and the finite element method.

Weak (Generalized) Derivatives

Motivation: For $u \in C^1$ and $\varphi \in C_0^\infty$: $\int u' \cdot \varphi, dx = -\int u \cdot \varphi', dx$. If $u$ is not smooth, we define the weak derivative $v$ as a function from $L^1$ such that:

$ \int v \cdot \varphi, dx = -\int u \cdot \varphi', dx\quad \text{for all } \varphi \in C_0^\infty(\Omega). $

Examples:

• $u = |x|$: $v(x) = \operatorname{sign}(x)$ is the weak derivative. Classically, $u'(0)$ does not exist.
• $u = H(x)$ (Heaviside): $v = \delta(x)$ — the delta function. No longer a function, but a distribution.

Sobolev Spaces

$W^{k,p}(\Omega)$: Functions $u \in L^p(\Omega)$, all weak derivatives $D^\alpha u$ ($|\alpha| \leq k$) are also in $L^p$. Norm: $ |u|{W^{k,p}} = \left(\sum{|\alpha|\leq k} |D^\alpha u|_p^p \right)^{1/p}. $

$H^k(\Omega) = W^{k,2}(\Omega)$: Sobolev Hilbert space.

$H_0^k(\Omega)$: Closure of $C_0^\infty(\Omega)$ in $H^k$. Functions from $H_0^1$ satisfy zero boundary conditions.

Sobolev embedding theorems:

• $W^{1,p}(\Omega) \hookrightarrow L^q(\Omega)$ for $q \leq np/(n-p)$.
• $H^1(-1,1) \hookrightarrow C^{0,1/2}[-1,1]$ (Hölder functions, $n=1$).

Lax–Milgram Theorem: If $a: H\times H\to\mathbb{R}$ is continuous and coercive ($a(u,u) \geq \alpha|u|^2 > 0$), then for any $F \in H^*$ there exists a unique $u \in H: a(u,v) = F(v)$ for all $v \in H$. Application: the weak Dirichlet problem ($-\Delta u = f$) always has a unique solution in $H_0^1(\Omega)$.

Numerical Example

Problem: Find the weak derivative of $u(x) = |x|$ on $(-1,1)$ and check whether $u \in H^1(-1,1)$.

Step 1. Weak derivative $v$: $ \int_{-1}^1 |x| \cdot \varphi'(x)dx = -\int_{-1}^1 v(x)\cdot \varphi(x)dx \quad \text{for all } \varphi \in C_0^\infty. $

Step 2. Split: $ \int_{-1}^0 (-x)\varphi' dx + \int_0^1 x\varphi' dx. $ Integrate each part by parts: $ \int_{-1}^0 (-x)\varphi'dx = [-x\varphi]{-1}^0 + \int{-1}^0 \varphi dx = 0 + \int_{-1}^0 \varphi dx. $ $ \int_0^1 x\varphi'dx = [x\varphi]_0^1 - \int_0^1 \varphi dx = 0 - \int_0^1 \varphi dx. $

Step 3. $ \int_{-1}^0 \varphi dx - \int_0^1 \varphi dx = -\int_{-1}^1 \operatorname{sign}(x) \varphi dx. $ Thus, $v(x) = \operatorname{sign}(x)$.

Step 4. Check $u \in H^1:$ $ |u|^2_{H^1} = |u|^2_{L^2} + |u'|^2_{L^2}. $ $||x||^2 = \int_{-1}^1 x^2 dx = 2/3$. $|\operatorname{sign}(x)|^2 = \int_{-1}^1 1 dx = 2$. $|u|^2_{H^1} = 2/3 + 2 = 8/3 < \infty$ $\rightarrow$ $u \in H^1(-1,1)$ ✓.

Step 5. Check $u \notin H^2:$ $u' = \operatorname{sign}(x)$ — weak derivative $= 2\delta(x)$ (from the next article) $\notin L^2$. Therefore $u \in H^1\setminus H^2$ ✓.

Step 6. Weak problem: $-u'' = 1$ on $(-1,1)$, $u(\pm1) = 0$. In $H_0^1$: $a(u,v) = \int u'v'dx = \int v dx$ for all $v \in H_0^1$. Exact solution $u = (1 - x^2)/2 \in H_0^1 \cap H^2$ ✓.

Real Application

Mechanics of deformable bodies: the deformation of a beam under load is described by the equation $d^4 u/dx^4 = q(x)$. The weak formulation in $H^2$ allows working with discontinuous loads. The FEM is constructed precisely on the basis of Sobolev spaces and the variational formulation.

Additional Aspects

Variational methods translate differential equations into the language of minimization of functionals. Classical example: the Dirichlet problem $-\Delta u = f$ with $u|_{\partial \Omega} = 0$ is equivalent to the minimization of the energy functional $ J(u) = \frac{1}{2}\int |\nabla u|^2 - \int f \cdot u. $ A solution exists and is unique due to the convexity and coercivity of $J$ and belongs to the Sobolev space $H^1_0(\Omega)$. This is the weak (or variational) formulation — the foundation of the theory of nonlinear elliptic equations (Lax–Milgram theorem), numerical FEM methods, and shape optimization. In physics, the variational principle underlies classical mechanics (Hamilton's principle), general relativity (the Einstein–Hilbert action), and quantum field theory (Feynman's formalism).

Connection with Other Branches of Mathematics

In the theory of differential equations, Sobolev spaces are the natural environment for elliptic and parabolic equations. The classical results of Hadamard, Agallera, and De Giorgi–Nash–Moser on regularity of weak solutions are formulated precisely in terms of $H^k$ and $W^{k,p}$. Sobolev embedding and the Rellich–Kondrachov compactness theorem are key to the existence of solutions in variational problems of the Dirichlet and Neumann types.

In functional analysis, the spaces $W^{k,p}$ are model reflexive Banach spaces; their dual spaces are described through spaces of distributions and Besov spaces. The Riesz representation theorem for linear functionals in a Hilbert space, together with Lax–Milgram, yields a unique weak object corresponding to the right-hand side of the equation.

The connection with topology manifests itself through the concept of trace on the boundary and trace spaces. Trace theory (Guber, Lions–Magenes) shows how Dirichlet and Neumann boundary conditions are realized as restrictions of functions from $H^1$ onto multidimensional manifolds. This forms the basis of the theory of boundary-value problems on Riemannian manifolds and spectral theory of the Laplace operator.

In probability theory, Sobolev gradients and Dirichlet form describe symmetric Markov processes (works by Fukushima and Ma). For Brownian motion on a manifold, the energy of a function coincides with the norm in $H^1$, and the weak Laplacian is associated with the generator of the Markov semigroup.

Numerical methods, primarily the finite element method (Ciarlet, Brenner–Scott), use $H^1$ and $H^2$ as target spaces for approximation. Céa-type estimates and FEM convergence theorems are formulated in the Sobolev norm, and the theory of a posteriori error estimates relies on local norms $W^{1,p}$.

Historical Background and Development of the Idea

The foundations of the theory go back to the works of Serra and Gateaux on generalized derivatives at the beginning of the 20th century. The terminology and systematic use of weak derivatives are associated with the name of Sergei Lvovich Sobolev. In 1935–1938, he published in "Doklady AN SSSR" and the monograph "Some Applications of Functional Analysis in Mathematical Physics" a rigorous concept of spaces now bearing his name, and applied it to problems of hyperbolic equations. In Western literature, the ideas became widely known through the works of Laurent Schwartz, Jacques-Louis Lions, and Enzo Magenes in the 1950s–1960s. The books by Lions–Magenes "Non-homogeneous Boundary Value Problems and Applications" and the monograph by Jürgen Stumpf "Boundary Value Problems for Elliptic Equations" established the language of Sobolev spaces for elliptic and parabolic problems. The work of Lax and Milgram in 1954 in "Communications on Pure and Applied Mathematics" played an important role, where the abstract variational principle for the existence of weak solutions of elliptic problems was formulated and proved.