Partial differential equation

A partial differential equation relates a multivariable function to its partial derivatives.

An obstacle constraint requires a solution to lie above a given function. In an optimal stopping problem, the value dominates the payoff, satisfies a generator equation where continuation is optimal, and a generator inequality where stopping is optimal. The complementary conditions can be expressed as a maximum equal to zero.

A first-contact argument compares a smooth solution with a subsolution or supersolution. At a first spatial minimum of their difference, the Hessian is nonnegative; a strictly favorable time derivative then contradicts contact. For $u_t=u^2{u}_{xx}+u^3$, a positive decreasing constant-in-space barrier proves that positive periodic initial data stay positive for the lifetime of the smooth solution.

The domain of dependence of a solution value consists of the initial and boundary locations whose data can influence it. For $u_t+a{u}_x=0$ on the whole line, $u(x,t)=u(x-at,0)$, so its initial-data domain of dependence is the foot of the characteristic. For a finite-difference scheme, repeated stencil dependencies determine a numerical domain of dependence. The Courant–Friedrichs–Lewy condition requires the numerical domain to cover the physical one for convergence; it is necessary under its usual hyperbolic hypotheses, but not in general sufficient for stability.

A linear partial differential equation is a partial differential equation in which the unknown function and its partial derivatives occur linearly, with coefficients independent of the unknown. Its homogeneous solutions obey the superposition principle.

An ill-posed problem fails existence, uniqueness or continuous dependence in the specified function spaces. Backward heat equation evolution is unstable because its Fourier modes grow like $e^{t|\xi {|}^2}$, making arbitrarily small high-frequency errors grow without a uniform bound.

A linear partial differential operator sends a function to a linear combination of its partial derivatives. If its coefficients are constant, the Fourier transform of a derivative turns it into multiplication by its polynomial symbol $P(\xi )$.

The elliptic Liouville equation $\Delta u=e^{2u}$ describes conformal metrics of constant negative Gaussian curvature under a common sign convention.

A radial function depends only on the distance from the origin: $f(x)=g(|x|)$. Differential operators then reduce to radial formulas such as $\Delta f=g^{\prime \prime }+(n-1)g^{\prime }/r$.

A periodic domain identifies opposite faces of a fundamental cell, producing a flat torus. Periodic integration by parts has no boundary term, and zero-mean functions have no zero Fourier mode.

An energy estimate tests a differential equation against the solution or one of its derivatives to control a norm by initial data and forcing. Coercive terms represent dissipation, while Young inequality and the Gronwall inequality control lower-order terms.

A bootstrap argument assumes an a priori bound on an initial time interval and derives a strict improvement using an energy estimate or another inequality. Continuity extends the improved bound to the entire interval under consideration. For example $A(t)\le C\epsilon +C{\int }_0^t(1+s{)}^{-1}A(s{)}^{2}ds$ closes $A\le 2C\epsilon$ when $\epsilon \log (1+t)$ is sufficiently small. The strict improvement is what prevents a first failure time.

If $\square u=F$ with d'Alembert operator $\square =-{\partial }_t^2+\Delta$, then $\Vert \partial u(t){\Vert }_2\le \Vert \partial u(0){\Vert }_2+{\int }_0^t\Vert F(s){\Vert }_{2}ds$, where $\partial u=(u_t,\nabla u)$. Test the wave equation against $u_t$, use integration by parts, and apply the Cauchy-Schwarz inequality. The same energy estimate applies to a finite system of wave equations.

For the unit-speed wave equation, the integral of $(u_t^2+|\nabla u{|}^2)/2$ over $B_r(x_0)$ at time $T\ge 0$ is at most its initial integral over $B_{r+T}(x_0)$. On a shrinking ball the outward energy flux minus the loss from its moving boundary is $-((u_t-{\partial }_{n}u{)}^2+|{\nabla }_{\tan }u{|}^2)/2$, hence nonpositive. This estimate establishes the domain of dependence and finite propagation speed without any assumptions at spatial infinity.

An estimate $E(R)\le C(E(2R)-E(R))$ bounds the energy in a ball by the energy in the surrounding annulus. Moving the smaller-ball energy to the left gives the strict contraction $E(R)\le \theta E(2R)$ with $\theta =C/(1+C)<1$. Iteration turns this contraction into decay at small scales.

If $E(R/2)\le \theta E(R)$ and $E$ is nondecreasing, then $E(2^{-k}R)\le {\theta }^{k}E(R)$. Between dyadic radii, monotonicity gives $E(r)\le 2^{\mu }(r/R{)}^{\mu }E(R)$ for any $0<\mu \le -{\log }_2\theta$. This transfers a discrete contraction into a uniform power bound.

For a smooth Dirichlet Laplacian eigenfunction, testing $-\Delta u=\lambda u$ with $|u{|}^{\gamma -2}u$ for $\gamma \ge 2$ gives the displayed energy estimate. The Sobolev chain rule for $v=|u{|}^{\gamma /2}$ then yields $\int |Dv{|}^2=\lambda {\gamma }^2/[4(\gamma -1)]\int |u{|}^{\gamma }$. The absolute value and the sign in the test function permit nodal changes. At $\gamma =2$, use the Lipschitz absolute-value chain rule and the fact that a gradient of a Sobolev function vanishes on a level set.

A Caccioppoli inequality is a local energy estimate obtained by testing an elliptic equation with a cutoff times the solution. It controls the gradient on a smaller ball by the function and forcing on a larger ball.

For a weak solution of a uniformly elliptic homogeneous divergence-form elliptic operator, testing with ${\eta }^2(u-c)$ and choosing the gradient of the cutoff function in $B_{2R}\setminus B_R$ gives ${\int }_{B_R}|\nabla u{|}^2\le C{R}^{-2}{\int }_{B_{2R}\setminus B_R}|u-c{|}^2$. The arbitrary constant $c$ can be chosen as an annular average, which makes the estimate suitable for a hole-filling argument.

For a nonnegative bounded weak subsolution and $\beta >1$, test with ${\eta }^2(u+\epsilon {)}^{\beta -1}$. Uniform ellipticity and Cauchy-Schwarz inequality give $\int |Du{|}^2{u}^{\beta -2}{\eta }^2\le 4{\Lambda }_{\ast }^2/[{\lambda }^2(\beta -1{)}^2]\int u^{\beta }|D\eta {|}^2$, with ${\Lambda }_{\ast }$ an operator norm bound for the coefficient matrix. The Sobolev chain rule and $\epsilon \downarrow 0$ justify fractional powers and zero values. An entrywise bound $\Lambda$ yields ${\Lambda }_{\ast }\le n\Lambda$.

For a nonnegative $H^1$ weak solution of a homogeneous divergence-form elliptic equation on $B_1\subset {\mathbb{R}}^{\ell }$, $\ell \ge 2$, with symmetric measurable $\lambda I\le A\le \Lambda I$, set $R=\Lambda /\lambda$ and $\alpha =1+1/\ell$. There is $D_{\ell }\ge 1$ independent of $q\ge 2$ such thatTest the equation with a cutoff times $u^{q-1}$, first regularized or truncated, to get $\int {\zeta }^2|\nabla u^{q/2}{|}^2\le [q/(q-1){]}^2{R}^2\int u^q|\nabla \zeta {|}^2$. Apply a fixed-domain Sobolev embedding theorem to $\zeta u^{q/2}$ and take the $2/q$ power. Keeping the constant inside that power is crucial for Moser iteration.

For a positive weak supersolution $u$ of a uniformly elliptic divergence-form equation, testing with ${\eta }^2/u$ givesThe constant depends only on the ellipticity bounds.

A second-order partial differential equation contains derivatives of the unknown function of order at most two and at least one second derivative. Its principal symbol determines its characteristic directions and elliptic, parabolic or hyperbolic type.

A quasilinear partial differential equation is linear in its highest-order derivatives, while its coefficients may depend on the independent variables, the unknown function, and lower-order derivatives.

For a $C^1$ flux $F$ and a weak solution of $\mathrm{div}F(Du)=0$, discrete integration by parts and the fundamental theorem of calculus along a line segment show that $w_h={\delta }_{\ell ,-h}u$ solves $\mathrm{div}(A_{h}D{w}_h)=0$, with $p_{-}=Du(x-h{e}_{\ell })$ and $p_{+}=Du(x)$. Coercive quadratic-form bounds and operator norm bounds for $DF$ on all segments between these gradients pass to $A_h$. This produces a divergence-form elliptic operator without initially assuming second partial derivatives of $u$.

The p-Laplacian is the nonlinear divergence-form operatorIt is the Euler-Lagrange operator of the p-energy. For $p>2$ it is degenerately elliptic where $Du=0$.

For a scalar differential operator of order $k$, the principal symbol replaces each order-$k$ derivative ${\partial }^{\alpha }$ by ${\xi }^{\alpha }$ and discards lower-order terms. For a quasilinear partial differential equation, the coefficients are evaluated at the prescribed lower-order data.

A hypersurface $\Sigma =\{\varphi =0\}$ is characteristic for a differential equation at $x\in \Sigma$ when its principal symbol vanishes on the conormal $d\varphi (x)$:It is non-characteristic where this quantity is nonzero.

A characteristic plane is a planar characteristic hypersurface. For $u_{zz}-m^2{u}_{xx}=0$, the planes $x\pm mz=\text{constant}$ carry the two travelling-wave families.

A non-characteristic hypersurface satisfies $p(x,d\varphi (x))\ne 0$. This lets the equation solve for the highest derivative normal to the hypersurface.

A characteristic coordinate is a function constant along one family of characteristics. For a second-order equation in two variables, two independent characteristic coordinates reduce the principal part to a mixed derivative.

The change of variables $X=\rho e^{-t}$, $S=e^{-t}>0$ turnsinto $w_{SS}-w_{XX}=0$. Indeed, with $D={\partial }_t+\rho {\partial }_{\rho }$, the original operator is $D^2+D-{\partial }_{\rho }^2$, while $D=-S{\partial }_S$ and ${\partial }_{\rho }=S{\partial }_X$. Its characteristic curves are $\rho =-1+C{e}^t$ and $\rho =1+C{e}^t$, including the characteristic lines $\rho =\pm 1$. General smooth solutions have the form $F(e^{-t}(\rho +1))+G(e^{-t}(\rho -1))$. Zero Cauchy data on $-1<\rho <1$ at $t=0$ force vanishing throughout that strip for $t\ge 0$, but not for all negative time: $u=(e^{-t}(\rho +1)-2{)}_{+}^3$ is a global $C^2$ counterexample. Thus the characteristic barriers describe a forward domain of dependence, not a barrier in both time directions.

A scalar second-order equationis hyperbolic where $B^2-AC>0$. Equivalently, its principal symbol has two distinct real characteristic directions.

For a one-dimensional first-order system $v_t+A(v,x,t)v_x=s$, hyperbolicity means that the coefficient matrix has real eigenvalues and a complete set of eigenvectors. Distinct real eigenvalues give strict hyperbolicity and separate characteristic curves. Along a curve of speed $\lambda$, a left eigenvector $\ell A=\lambda \ell$ gives the compatibility equation $\ell dv/dt=\ell s$.

For a one-dimensional wave equation whose principal symbol is ${\xi }_t^2-c(y{)}^2{\xi }_y^2$, a travel-time coordinate isIts two families of characteristic curves are $s(y)-t=\text{constant}$ and $s(y)+t=\text{constant}$ wherever the wave speed $c$ is positive.

For $c(y)=1-y^2$ in the strip $|y|<1$, the travel-time coordinate is $s(y)=\mathrm{artanh}y$. The characteristic curves areequivalently $y=\tanh (C\mp x)$. Outside the strip the same separated ordinary differential equation gives hyperbolic cotangent branches, and the equilibrium curves $y=\pm 1$ are also characteristic.

A Cauchy problem prescribes a solution and enough of its derivatives on a hypersurface to determine local evolution away from that hypersurface.

Continuous dependence means that convergence of the initial data in a specified topology implies convergence of the corresponding solutions in the specified solution topology. It is a stability requirement, not merely existence or uniqueness. Smooth-data instability of the Cauchy-Riemann Cauchy problem gives a counterexample even when the initial data converge uniformly with every fixed number of derivatives.

For a scalar equation of order $k$, Cauchy data consist of the first $k$ normal jets $u,{\partial }_{N}u,\dots ,{\partial }_N^{k-1}u$ on the initial hypersurface.

The boundary jet through order $m-1$ is the collection of traces ${\partial }^{\alpha }u{|}_{\partial \Omega }$ with $|\alpha |\le m-1$. Vanishing of the whole jet is stronger than vanishing of one derivative. On a smooth hypersurface, zero normal Cauchy derivatives through order $m-1$, together with their tangential derivatives, give a zero full boundary jet.

For an order-$k$ scalar quasilinear partial differential equation with real-analytic coefficients, real-analytic Cauchy data on a real-analytic non-characteristic hypersurface determine a unique real-analytic solution in a neighbourhood of each point of that hypersurface.

For a linear differential operator with real analytic coefficients, a classical solution with zero Cauchy data on a real analytic non-characteristic hypersurface vanishes locally near that hypersurface. The solution itself need not be real analytic. A proof solves the real analytic inhomogeneous formal adjoint problem against polynomial sources, annihilates all polynomial moments using boundary jets, and uses polynomial density to force the solution to vanish.

Near a non-characteristic plane $x=0$, use the small cap ${\omega }_{\epsilon }=\{x>0,x+y^2<\epsilon \}$. Its curved face is non-characteristic by continuity of the principal symbol. A polynomial coordinate change flattens that face. Uniform real analytic coefficient bounds and the uniform Cauchy radius for polynomial forcing solve $P^{\ast }v=g$ throughout a common cap, with zero boundary jet on the curved face. The complete boundary-jet condition for formal adjoints then gives ${\int }_{{\omega }_{\epsilon }}ug=0$ for every polynomial $g$. Uniform approximation of $\bar{u}$ gives $\int |u{|}^2=0$. Repeating on the other side of the plane proves two-sided local uniqueness.

For a linear real analytic non-characteristic Cauchy problem for a partial differential equation with zero data, the local real analytic radius may be chosen uniformly over polynomial forcing terms. Fix an analyticity exponent $R^{-1}$. Each polynomial, including a polynomial obtained by a polynomial coordinate change, satisfies the corresponding factorial derivative bound with some finite prefactor, uniformly over a compact parameter range. Divide the forcing by that prefactor to put all such data in one fixed majorant class. The Cauchy-Kovalevskaya theorem supplies a common radius, and linearity rescales the solution back without shrinking it. This argument does not bound the amplitudes of the resulting solutions uniformly.

A real analytic system solved for the highest normal derivatives, with real analytic Cauchy data, has a unique local real analytic solution. For common order $m$, its normal form is ${\partial }_t^{m}U=F(t,x,\{{\partial }_t^j{\partial }_x^{\alpha }U:j<m,j+|\alpha |\le m\})$. An invertible coefficient/Jacobian matrix for the highest normal derivatives is the system's non-characteristic hypersurface condition. A real analytic coordinate change flattens the initial hypersurface, and adjoining derivatives through order $m-1$ reduces the problem to a first-order real analytic system. Uniqueness here is in the real analytic class; it does not assert Hadamard well-posedness in smooth or Sobolev norms.

In null coordinates, the characteristic boundary problemis a Volterra integral equation. Compatible analytic boundary data determine a unique analytic solution.

On a compact characteristic rectangle, the Volterra iterates for the analytic Goursat problem are bounded by $C^m/(m!{)}^2$. Their series and differentiated series converge throughout the rectangle, extending the local analytic solution globally across it.

A solution of a partial differential equation is a function satisfying the equation in a specified sense and obeying its initial or boundary data.

A classical solution has enough ordinary derivatives for the differential equation and its data to hold pointwise.

The notation $C_b^{1,2}([0,\infty )\times \mathbb{R})$ ordinarily denotes functions with continuous time derivative and two continuous space derivatives, with the function and indicated derivatives bounded globally. A different condition, often intended in finite-horizon partial differential equations, is membership in $C_b^{1,2}([0,T]\times \mathbb{R})$ for each finite $T$, allowing bounds to grow with $T$. These conventions must be distinguished in a long-time growth problem: the second permits divergence with time; the first does not. There is no universal notation for the finite-horizon condition, so it should be stated explicitly.

A weak solution satisfies an integrated identity in which derivatives are transferred to test functions. It can therefore exist with fewer classical derivatives.

A weak-strong uniqueness principle says that a weak solution with the same data as a sufficiently regular classical solution agrees with that solution throughout its interval of regular existence, within the specified weak class. The conclusion is class-dependent: all bounded weak solutions of a prescribed $C^1$ linear transport field satisfy it, while nonlinear conservation laws require an admissibility condition such as an entropy solution condition.

For $u_{tt}+L(t)u=f$ with homogeneous Dirichlet boundary conditions, smooth uniformly elliptic principal coefficients, and initial data $({\psi }_0,{\psi }_1)\in H_0^1(U)\times L^2(U)$, a space-time $H^1$ weak solution has zero lateral trace, trace $u(0)={\psi }_0$, and satisfiesfor every space-time $H^1$ test function with zero lateral trace and $v(T)=0$. The velocity initial condition is encoded by the last term; an arbitrary space-time $H^1$ function need not have an $L^2$ trace of its time derivative. The equation itself gives $u_{tt}\in L^2(0,T;H^{-1}(U))$, so $u_t$ has a continuous $H^{-1}$ representative.