Potential bug: missing a
in the denominator,
u[n+1] = (1 - (1-theta)*a*dt)/(1 + theta*dt)*u[n]
Running decay_convrate.py
gives same rates.
Why? The value of \( a \)... (\( a=1 \))
0 and 1 are bad values in tests!
Better:
Terminal> python decay_convrate.py --a 2.1 --I 0.1 \
--dt 0.5 0.25 0.1 0.05 0.025 0.01
...
Pairwise convergence rates for theta=0:
1.49 1.18 1.07 1.04 1.02
Pairwise convergence rates for theta=0.5:
-1.42 -0.22 -0.07 -0.03 -0.01
Pairwise convergence rates for theta=1:
0.21 0.12 0.06 0.03 0.01
Forward Euler works...because \( \theta=0 \) hides the bug.
This bug gives \( r\approx 0 \):
u[n+1] = ((1-theta)*a*dt)/(1 + theta*dt*a)*u[n]