Is there a way to use negative times in corr_3_op

[Perry Rice]

I'm looking at assymetries in things like <adag(0)(a(tau)+adag(tau))a(0)>

its a quadrature measurement conditioned on a click, and for negative times is is a detection probability conditioned on  a quadrature measurement

[Éric Giguère]

Only for closed systems.

mesolve / sesolve, used by correlation, accept times that are all decreasing or all increasing:

```

qutip.correlation_3op_1t(adag @ a, qutip.basis(4, 2), [0, -1, -2, -3], [], a, a @ adag, adag, options={"method": "adams"})

```

But for open systems, the states will not be physical most of the time

`qt.mesolve(qt.num(4), qt.basis(4, 2), [0, -1], [a]).final_state`

```
Quantum object: dims=[[4], [4]], shape=(4, 4), type='oper', dtype=Dense, isherm=True
array([
    [ 2.952,  0.,  0., 0.],
    [ 0., -9.341,  0., 0.],
    [ 0.,  0.,  7.389, 0.],
    [ 0., 0., 0., 0.]
])

```

So unless you have a way to ensure your results are right, it's not 

ps. not all ode methods support negative times, "adams", "diag", "dop853", should work.

[Perry Rice]

OK. These things are well defined in the quantum regression theorem, you just evolve "backwards" with the adjoint. Liouvillian

<A(0)B(-t)C(0)>=<A(t)B(0)C(t)> by stationarity, is there a way to use core_3op_1t for this?

[Éric Giguère]

<A(t)B(0)C(t)> can't be done directly.

You could compute <A(t)B(t+tau)C(t)> with tau=-t.

Or you could use mesolve twice:

```

solver = MESolver(H, c_ops, options=options)

states = solver.run(rho_ss, times).states

corr = []

for state, t in zip(states, times):

    corr.append(solver.run(C @ state @ A, [0, -t], e_ops=B).expect[0][1])

```

[Perry Rice]

thanks!