B3.18

Partial Differential Equations | Part II, 2003

Consider the initial value problem

2ut2Δu=0\frac{\partial^{2} u}{\partial t^{2}}-\Delta u=0

to be solved for u:R×RnRu: \mathbb{R} \times \mathbb{R}^{n} \rightarrow \mathbb{R}, subject to the initial conditions

u(0,x)=f(x),ut(0,x)=0u(0, x)=f(x), \quad \frac{\partial u}{\partial t}(0, x)=0

for ff in the Schwarz space S(Rn)\mathcal{S}\left(\mathbb{R}^{n}\right). Use the Fourier transform in xx to obtain a representation for the solution in the form

u(t,x)=eixξA(t,ξ)f^(ξ)dnξu(t, x)=\int e^{i x \cdot \xi} A(t, \xi) \widehat{f}(\xi) d^{n} \xi

where AA should be determined explicitly. Explain carefully why your formula gives a smooth solution to (1) and why it satisfies the initial conditions (2), referring to the required properties of the Fourier transform as necessary.

Next consider the case n=1n=1. Find a tempered distribution TT (depending on t,xt, x ) such that (3) can be written

u=<T,f^>u=<T, \widehat{f}>

and (using the definition of Fourier transform of tempered distributions) show that the formula reduces to

u(t,x)=12[f(xt)+f(x+t)]u(t, x)=\frac{1}{2}[f(x-t)+f(x+t)]

State and prove the Duhamel principle relating to the solution of the nn-dimensional inhomogeneous wave equation

2ut2Δu=h\frac{\partial^{2} u}{\partial t^{2}}-\Delta u=h

to be solved for u:R×RnRu: \mathbb{R} \times \mathbb{R}^{n} \rightarrow \mathbb{R}, subject to the initial conditions

u(0,x)=0,ut(0,x)=0u(0, x)=0, \quad \frac{\partial u}{\partial t}(0, x)=0

for hh a CC^{\infty} function. State clearly assumptions used on the solvability of the homogeneous problem.

[Hint: it may be useful to consider the Fourier transform of the tempered distribution defined by the function ξeiξa\xi \mapsto e^{i \xi \cdot a}.]

Typos? Please submit corrections to this page on GitHub.