linear PDE
solutions obey principle of superposition ie y1,y2 solutions implies y = Ay1 + By2 solution therefore solution space is vector space ( ∞ dim)
second order linear PDE
if u = u(x,t)
Auxx + 2Buxt + Cutt + Dut + Eux + Fu = 0
A, B, C, D, E, F can be any functions of x we shall take generally as constants
PDEs are classified into 3 types - sign of det(M)
M = (A B
B C)
detM = AC - B^2
3 types of PDE
boundary conditions for ODEs
suppose we know all at t = 0
want to know solution to nth order linear ODE at t>0
y = A1y1 + A2y2 + … + Anyn
specify n pieces of info to fix Ai’s and get particular solution
eg y(x), dy/dy (0)… d^(n-1)y/dt^(n-1) (0)
boundary conditions for PDEs
as solution space is ∞ dim need ∞pieces of info to fix Ai’s - generally in form of functions specified at t = 0 eg u(x,0) = R(x), du/dt (x,0) = S(x)
can have boundary at t= 0 but also in space need to specify boundary here as well
boundary condition examples PDE
boundary conditions in space
can have a fixed end u(a,t) = 0 - dirichlet
or vibrating vertically ux(a,t) = 0 - neumann
method of separation of variables
u(x,t) = ∑ Ai ui(x,t) look for ui in nice form
ui(x,t) = X(x)T(t)
- sub into differential eq
- sub in position boundary conditions and solve
- sub in time boundary conditions
Bn and An in Ancos(t) +Bn sin(t) term
Bn control initial velocity
An control initial position
finding An method 1 - fourier series
R(x) = ∑An sin(nx)
sum is odd/ even periodic extension of the R(x)
An = 1/pi ∫R(x)sin(nx) dx from pi to -pi
finding An method 2 - orthogonality
R(x) = ∑An sin(nx)
sin(nx) is a solution to x’’ = λx so sin(x) is a eigenfunction of d^2/dx^2 with eigenalue λ
if d^2/dx^2 is self adjoint then sin(nx) orthogonal basis so easy to find components
chose and inner product (f,g) = ∫fg dx from pi to 0 - show is self adjoint - then sin(nx) orthogonal so (sin(nx),sin(mx)) = 0 for n≠m
so Am = (R(x),sin(mx))/(sin(mx), sin(mx))