Setup
  shapes:
    on the plane: a guide of ε 12 along x, 1 µm wide, centred on y = 0; a disc of radius 250 nm centred on its upper edge at x = 2.25 µm: a half disc on the guide
  medium: air, ε 1, in 2D (a = 1 µm)
  source: a dipole of E_z on the guide's axis at x = -2.25 µm, a pulse at 0.2 c/a, relative width 0.1
  window: x -3 to 3.75 µm, y -1.5 to 2 µm inside CPMLs of 1.5 µm; the scattered flux out of the box x 1.25 to 3.25 µm, y -1 to 1.5 µm
  grid: 10 to 30 pixels a, every 2; the reference smoothed at 50; Courant number 0.9; each run until E_z decays to 1e-11
  solver: 2D FDTD, with subpixel smoothing and with ε sampled; the scattering, the difference of runs with and without the bump
  tolerance: 0.6 on the smoothed order (2), 0.5 on the sampled one (1); smoothing better at every resolution

Oskooi et al.'s Fig. 7: a half disc of radius 0.25a on a guide of eps = 12, a wide,
E_z from a dipole 4.5a from it, f = 0.2 c/a; the scattered field's flux out of a
2a x 2.5a box about the bump, per unit dipole moment (a = 1 um)
reference: smoothed at 50 pixels/a, 1.21279e-3
pixels/a   smoothed: power, error   no smoothing: power, error
      10   1.30425e-3, 7.54e-2   2.13839e-3, 7.63e-1
      12   1.27077e-3, 4.78e-2   1.87428e-3, 5.45e-1
      14   1.25232e-3, 3.26e-2   1.92623e-3, 5.88e-1
      16   1.24242e-3, 2.44e-2   1.66174e-3, 3.70e-1
      18   1.23507e-3, 1.84e-2   1.66618e-3, 3.74e-1
      20   1.23245e-3, 1.62e-2   1.58018e-3, 3.03e-1
      22   1.22647e-3, 1.13e-2   1.63285e-3, 3.46e-1
      24   1.22582e-3, 1.07e-2   1.55356e-3, 2.81e-1
      26   1.22111e-3, 6.86e-3   1.54503e-3, 2.74e-1
      28   1.22199e-3, 7.58e-3   1.45723e-3, 2.02e-1
      30   1.21855e-3, 4.75e-3   1.51946e-3, 2.53e-1
Fig. 7, read: smoothed 0.08 at 10 and about 0.005 at 30; no smoothing 0.2 and 0.07
  order, smoothed (least squares)        2.370000   paper 2          ± 6e-1      ok
  order, no smoothing (least squares)     1.070000   paper 1          ± 5e-1     ok
  resolutions where smoothing is better           11   paper 11                  ok
all 3 within tolerance of the paper
