Quantum Zeno effect for mesolve with non-uniform dt?

[neago]

mesolve should be able to take arbitrary times as the tlist input, but it seems to give wrong results if tlist is not linear. For e.g. a quadratic spacing of the times, the decay of a cavity field seems to be inhibited - including more time steps even leads to less decay!

The sample code below produces this graph:

Black is for linspace times, red for quadratic times, and blue for a linspace with every 6th element removed. I guess all curves should follow the black one.

a = destroy(30)
H = a.dag() * a
psi0 = coherent(30, 4)
tlist1a = linspace(0, 1., 50)
tlist1b = linspace(0, 1., 10)
tlist2a = linspace(0, 1., 50)**2
tlist2b = linspace(0, 1., 10)**2
tlist3a = tlist1a[arange(1, len(tlist1a)+1) % 6 != 0]

result1a = mesolve(H, psi0, tlist1a, [a], [a.dag()*a])
result1b = mesolve(H, psi0, tlist1b, [a], [a.dag()*a])
result2a = mesolve(H, psi0, tlist2a, [a], [a.dag()*a])
result2b = mesolve(H, psi0, tlist2b, [a], [a.dag()*a])
result3a = mesolve(H, psi0, tlist3a, [a], [a.dag()*a])

plot(result1a.times, result1a.expect[0], 'k', result1b.times, result1b.expect[0], 'ko')
plot(result2a.times, result2a.expect[0], 'r', result2b.times, result2b.expect[0], 'ro-')
plot(result3a.times, result3a.expect[0], 'b')

xlabel('t')
ylabel('<n>')

[neago]

Yikes - it's already been fixed!
That will teach me to pull the latest version before filing an issue (＞＜)

[jrjohansson]

Hi Jonas. Thanks for the bug report! Yes, this should be fixed already. It was only a few days we fixed this bug, so no wonder you hadn't seen it.

[neago]

Hi Robert. Yes, I saw that, but still annoying - I could have saved an hour this morning if I had updated beforehand :)