Vay current deposition (#1051)

* Added stub for current correction in RZ spectral solver

* Start implementation of Vay deposition

* Continue implementation of Vay deposition

* Correct deposition of D

* Add phase shift for staggered currents

* Small clean-up

* Fix units in deposition of D

* Implement average of cumulative sum (needs bug fix)

* Start fixing bug in average of cumulative sum

* Still debugging

* Cumulative sums should be correct now

* Subtract averages of cumulative sums:

- current implementation: cumulative sums, inverse Fourier transform, subtraction of averages
- needs to be tested (including units of D after Vay deposition)
- needs to be shortened (too many loops over boxes and ParallelFors)

* [skip CI] Clean up and fix units

* Still fixing units

* [skip CI] Remove temporarily averages of cumulative sums

* [skip CI] Remove distinction between staggered and nodal

* Vay and Esirkepov similar results on periodic single box:

TODO:
- debug (charge not conserved);
- try using compute_shifted_shape_factor as in Esirkepov deposition;
- clean up;
- try on multiple boxes and with correction of mode at 0 frequency.

* [skip CI] Clean up

* Fix bug in 3D deposition

* [skip CI] Clean up

* Fix 2D and 3D implementation:

- simulation results agree between direct and Vay deposition in both 2D and 3D
- Travis CI tests should pass except for check of charge conservation (debug)

* Small clean-up

* Fix bug when compiling in RZ geometry

* Add benchmark json files (will be reset later)

* Do not set zero current at zero frequency

* [skip CI] Revert last commit and clean up

* Fix small bug after reverting commit

* Set nodal test first on Travis

* Fix benchmark for nodal test in 3D

* Fix particle output for nodal test in 3D

* Fix bugs due to staggering

* Rename current nodal Travis tests

* Add Travis tests staggered in 2D and 3D

* Further clean-up after bug fix

* Abort when using Vay deposition with domain decomposition

* Add optional argument of index type to forward FFT

* Fourier shifts can be private members as before

* Small clean-up

* Clean up and improve Doxygen documentation

* Fix small bug in analysis script for 2D tests

* Fix tests (remove E and B fields from particle diags)

* Add option to fill guard cells and docs

* Fix value of last guard cell by enforcing periodicity

* Revert changes merged from #1121

* Clean up style

* Improve docs

* Fix forgotten alignment

* Improve docs

* Make base class functions VayDeposition pure

Co-authored-by: Dave Grote <dpgrote@lbl.gov>

1 files changed, 14 insertions, 0 deletions

diff --git a/Source/FieldSolver/SpectralSolver/SpectralSolver.H b/Source/FieldSolver/SpectralSolver/SpectralSolver.H
index 91cfc84f1..ba805c790 100644
--- a/Source/FieldSolver/SpectralSolver/SpectralSolver.H
+++ b/Source/FieldSolver/SpectralSolver/SpectralSolver.H
@@ -84,6 +84,20 @@ class SpectralSolver
              algorithm->CurrentCorrection( field_data, current, rho );
         };
 
+        /**
+         * \brief Public interface to call the virtual function \c VayDeposition,
+         * declared in the base class SpectralBaseAlgorithm and defined in its
+         * derived classes, from objects of class SpectralSolver through the private
+         * unique pointer \c algorithm.
+         *
+         * \param[in,out] current Array of unique pointers to \c MultiFab storing
+         *                        the three components of the current density
+         */
+        void VayDeposition (std::array<std::unique_ptr<amrex::MultiFab>,3>& current)
+        {
+            algorithm->VayDeposition(field_data, current);
+        }
+
     private:
         void ReadParameters ();