image/svg+xml

GSoC2015: Post Mid-term update

Hello everyone,

During the past few weeks we have made some decent progress and here is a gist of updates :

  1. All the PR’s from the previous blog post are now in, which implies we have support for a. Solver expmv from both the packages ExpmV.jl and Expokit.jl b. Liouville von Neumann Equation types and constructs.

  2. In addition to the above we have support for Lindblad Master Equation and most of the blog post would be about the related constructs.

  3. There have also been few additions to QuBase.jl a. From the previous blog, the PR on vec is now in. b. We had a fix on multiplication, PR, we had spotted this as we were testing out Lindblad based examples. c. PR has been opened to support Base.complex function for QuArray (this requirement was spotted while working on ODE related solvers)

To begin with, Lindblad Master Equation type QuLindbladMasterEq has fields lindblad, hamiltonian and collapse_ops which are the lindblad operator, hamiltonian and the collapse operators of the system. The lindblad operator is constructed from the function lindblad_op which accepts the hamiltonain and vector. The vector here is of type Any and we still have some design issues which will be summarized at the end, it would be great to hear from the community on the possibilities of solving these. The liouvillian_op is constructed from lindblad_op by passing an empty array to the vector parameter. The reason for caching Lindblad operator in type construct is that we do not need to construct it over and over again by the use of operator or other means, this turns out to be advantageous for time-independent hamiltonian cases. This is also consistent with the QuLiouvillevonNeumannEq type where we cache the liouvillian in the type construct itself. The two issues we currently have are the following :

a. Design of time dependent and time independent lindblad operator constructs, we need to think of ways to support time-dependent nature. b. Promotion rule for collapse operators, as of now we cannot construct something like

c_ops_1 = [sigmax, sigmay]
c_ops_2 = [sigmax, lowerop(2)]

We need to have promotion rules to support such kind of constructions.

Moving on, we have added the additional constructs for QuPropagator which accepts

  1. QuLindbladMasterEq and AbstractQuMatrix.
  2. AbstractQuMatrix and Vector{AbstractQuMatrix}, constructs a QuLindbladMasterEq from AbstractQuMatrix and Vector{AbstractQuMatrix}.

We have worked on porting Jaynes Cummings by QuTiP to Julia. We would like to thank QuTiP developers for having shared the notebook, below are both the versions

  1. Jaynes - Cummings using QuTiP
  2. Jaynes - Cummings using QuDynamics

We have also done benchmarking analysis for various solvers using the package BenchmarkLite.jl. We have added the code under perf in the repo which has the benchmarking analysis for Jaynes Cummings model and Quantum Harmonic Oscillator. The results have been presented in the notebook, but there is still a need to refine it.

A quick summary on the PR’s :

  1. PR 16 (merged) [Equation types, expo solvers, tests and docs] has PR 13 (closed) [expo solvers, tests and docs] and PR 14 (closed) [revamped tests]
  2. PR 17 (merged) had an updated README.md
  3. PR 18 (merged) has the constructs of Lindblad Master Equation
  4. PR 20 (merged) has performance analysis of Jaynes Cummings Model and Quantum Harmonic Oscillator which has been done using BenchmarkLite.jl

On QuBase, we have an open PR on adding Base.complex to QuArray.

The next targets would be to start working on the implementation of Quantum Monte Carlo Wavefunction method and then extend the current framework to support parallelization techniques.