Design spec

1. Time marching
2. Update dot via Bloch equation 1. Dot must know current pseudospin 2. Dot must compute "Omega matrix" (Bloch RHS operator) to produce a derivative 3. Integrate derivative by exploiting prolates?
3. Communicate dot polarizations through Green's function 1. Construct dyadic green's function (polarization signal + 2 time derivatives) 1. Representation of signal via some basis function (prolates)
   1. Dots maintain knowledge of own past (Structures of arrays layout)
   2. Query signals at arbitrary (past) times
      1. Handle signal edge-cases; steady-state extrapolation near beginning of simulation, disfuturification near end
      2. Compute appropriate set of basis functions to sum (greater than/less than half timestep calculations) 1. Cache "basis function distance array"; evaluating basis functions at non-integral points corresponding to r = c*dt particle separations; likely a global structure.
4. Field samples
5. Same logic as what's used to communicate fields between dots except this is a "dot -> ghost particle" evaluation
6. Object to represent collection of ghost particles? 1. Can have multiple samplers that way (high res, low res, volumetric/"surfimetric" point collections)
7. SILO/visit output! 1. Can likely use a very similar interface to the one written for the F90 code 2. Spin components at each dot (point mesh) 3. E-field components at each sample point (quadmesh, perhaps other meshes later?)
8. Calculation of physical quantities 1. Quantities TBD, likely "integrals" over all points/pairs of points (correlations)