Qutip mesolve gives different results depending on number of points in time list

Question

I get different results from mesolve when I change the number points in tlist. The Hamiltonian and initial vector are the same. The results from mesolve are also incorrect (however, essolve gives me the correct answer from any number of points in tlist)... Am I missing something? In the QuTip tutorial demoing this, they use tlist = np.linspace(0.0, 10.0, 100), but when I use this I get ODE integration error.

H = 2 * np.pi * 0.1 * sigmax()
tlist = np.linspace(0.0, 10.0, 1000)
psi0 = basis(2, 0)
result = mesolve(H, psi0, tlist, [], [sigmax(), sigmay(), sigmaz()])
fig, axes = plt.subplots(1,1)
axes.plot(tlist, result.expect[2], label=r'$\left<\sigma_z\right>$')
axes.plot(tlist, result.expect[1], label=r'$\left<\sigma_y\right>$')
axes.plot(tlist, result.expect[0], label=r'$\left<\sigma_x\right>$')
axes.set_xlabel(r'$t$', fontsize=20)
axes.legend(loc=2);
plt.show()

tlist = np.linspace(0, 10, 150)
result = mesolve(H, psi0, tlist, [], [sigmax(), sigmay(), sigmaz()])
fig, axes = plt.subplots(1,1)
axes.plot(tlist, result.expect[2], label=r'$\left<\sigma_z\right>$')
axes.plot(tlist, result.expect[1], label=r'$\left<\sigma_y\right>$')
axes.plot(tlist, result.expect[0], label=r'$\left<\sigma_x\right>$')
axes.set_xlabel(r'$t$', fontsize=20)
axes.legend(loc=2);
plt.show()

The output from essolve for any time list

Answer 1

This is what I get when I run the same exact code as your first bloc.

Note that even only using 100 data points in tlist it gives the correct answer. So my naive intuition is that either something is horribly wrong with your installation, or there is some issue with your code. Perhaps are you using a jupyter notebook with stored kernel variables that are messing up your simulation? Your top plot kinda looks like there is some Lindblad operators causing the decay behavior of your expectation values.

I would suggest trying a completely new kernel, and just running that section of code.