Amit Rotem

ForwardDiff with complex numbers ode sometimes results in NaN derivatives

2

Found a small issue with ForwardDiff on complex numbers ode https://github.com/SciML/DiffEqBase.jl/pull/860 The Dual of time in the ode function has NaN values. Meaning, in the ode function `(u,p,t) -> ...`...

ForwardDiff fails on (schroedinger_dynamic with) TimeDependentOperator

6

derivative on `TimeDependentSum` fails; ``` using QuantumOptics import ForwardDiff as FD base=SpinBasis(1/2) ψi = spinup(base) ψt = spindown(base) function Ftdop(q) H= TimeDependentSum([q, abs2∘sinpi], [sigmaz(base), sigmax(base)]) _, ψf = timeevolution.schroedinger_dynamic(range(0,1,2), ψi,...

Adding ForwardDiff support for all solvers

1

Current fix for supporting Dual number in the Schroedinger solver https://github.com/qojulia/QuantumOptics.jl/pull/356 TODO; - Implement in other solvers Related issue; `Complex{Dual}` support in `DiffEqBase.jl` (see https://github.com/SciML/DiffEqBase.jl/pull/860)

``` using QuantumOptics import ForwardDiff as FD mb = NLevelBasis(5) s = randoperator(mb) h = [randoperator(mb),randoperator(mb)] LZtr(p) = abs2(tr(s'*(LazySum(p, h)*s))) LZtr_workaround(p) = abs2(tr(s'*(LazySum(p, h)*s))) function LZtr_workaround(p::Vector{

Missing details about default solver options

2

I'm wondering about the choice of default solver options; `DP5()` for the algorithm, https://github.com/qojulia/QuantumOptics.jl/blob/8012aa8cb224df75dcc0e1301b24c740c616d330/src/timeevolution_base.jl#L14-L18 and for the tolerance, https://github.com/qojulia/QuantumOptics.jl/blob/8012aa8cb224df75dcc0e1301b24c740c616d330/src/timeevolution_base.jl#L59-L66 I can relate to the tolerance choice, as the default in...