On Schemes for Exponential Decay

Hans Petter Langtangen [1, 2]

[1] Center for Biomedical Computing, Simula Research Laboratory
[2] Department of Informatics, University of Oslo

Mar 1, 2021






Goal

The primary goal of this demo talk is to demonstrate how to write talks with DocOnce and get them rendered in numerous HTML formats.
Layout

This version utilizes reveal slides with the theme beigesmall.

Notice

Speaker notes show up by pressing "s".

Methods

Mathematical problem

 
$$ \begin{align} u'(t) &= -au(t), \tag{1}\\ u(0) &= I, \tag{2} \end{align} $$

 

  • \( t\in (0,T] \)
  • \( a \), \( I \), and \( T \) are prescribed parameters
  • \( u(t) \) is the unknown function





Numerical solution method

  • Mesh in time: \( 0= t_0 < t_1 \cdots < t_N=T \)
  • Assume constant \( \Delta t = t_{n}-t_{n-1} \)
  • \( u^n \): numerical approx to the exact solution at \( t_n \)

Numerical scheme:  
 
$$ u^{n+1} = \frac{1 - (1-\theta) a\Delta t}{1 + \theta a\Delta t}u^n, \quad n=0,1,\ldots,N-1 $$  
 

Forward Euler explained

Implementation

The numerical method is implemented in a Python function:

def solver(I, a, T, dt, theta):
    """Solve u'=-a*u, u(0)=I, for t in (0,T]; step: dt."""
    dt = float(dt)           # avoid integer division
    N = int(round(old_div(T,dt)))     # no of time intervals
    T = N*dt                 # adjust T to fit time step dt
    u = zeros(N+1)           # array of u[n] values
    t = linspace(0, T, N+1)  # time mesh

    u[0] = I                 # assign initial condition
    for n in range(0, N):    # n=0,1,...,N-1
        u[n+1] = (1 - (1-theta)*a*dt)/(1 + theta*dt*a)*u[n]
    return u, t

Selected results

The Crank-Nicolson method





The artifacts can be explained by some theory

Exact solution of the scheme:

 
 
$$ u^n = A^n,\quad A = \frac{1 - (1-\theta) a\Delta t}{1 + \theta a\Delta t}\thinspace .$$  
 

  • Stability: \( |A| < 1 \)
  • No oscillations: \( A>0 \)
  • Always for Backward Euler (\( \theta=1 \))
  • \( \Delta t < 1/a \) for Forward Euler (\( \theta=0 \))
  • \( \Delta t < 2/a \) for Crank-Nicolson (\( \theta=1/2 \))

Concluding remarks:

Only the Backward Euler scheme is guaranteed to always give qualitatively correct results.