Fourier
∇2P − 1 ∂P ∂ 2 P ∂ 2 P 1 ∂P =0 → + − =0 D ∂t ∂x 2 ∂y 2 D ∂t
The diffusion equation can be derived from theprobabilistic nature of Brownian motion described as random walks (speak with me if you really want to see the derivation). The constant D is the diffusion coefficient whose nature we will explore in a moment, but for now we are solving a math problem. Notice we have no powers or complicated functions of any derivative of the function P. Since the derivatives of P only appear to first power, we callthis a linear PDE. Also notice that the equation does not mix x’s, y’s, or time within a single term, that is there are no multiplicative, divisional, or functional terms in the equation between the independent variables, we call this feature separability of the PDE. These two properties, linearity and separability, make the problem vastly more tractable than many other PDE’s encountered in physics.I am going to attempt to walk you through a challenging but very powerful method for solving PDE’s using the Fourier Transform (FT). Many of you may know that the FT is used in signal analysis and manipulation, but it was first used by Fourier to solve this problem. Consider the ˆ following integral relations that define the 2D FT in Cartesian coordinates. We will call the function P the FT of ouroriginal function P:
∞ ∞
ˆ P(k x , k y , t ) = P ( x, y , t ) =
−∞ −∞ ∞ ∞ i 2π ( k x ⋅ x + k y ⋅ y ) −∞ −∞
∫ ∫e
− i 2π ( k x ⋅ x + k y ⋅ y )
P( x, y, t )dxdy
∫ ∫e
ˆ P (k x , k y , t )dk x dk y
Notice the symmetry in going forward and backward in the transform. This is because switching between the normal form of the problem and what we call Fourier Space, where theproblem exists after the FT, are physically identical. We will require one more fact familiar from calculus, namely integration by parts:
∫ udv = uv − ∫ vdu
b a a a
b
b
Using these facts, let’s examine the spatial derivatives of the diffusion equation, where we consider the second derivative to be the function of interest. We can integrate these second derivatives by parts, identifying u...
Regístrate para leer el documento completo.