Finite-size correction in units of 2 pi sigma e b^2 (Aspden's term = 1):
  point quon in its void (Aspden)            1.0
  spread quon, no void (our earlier 1/6)     0.16666667
  spread quon in its void (consistent)       1.0
  -> inside a spherical cavity the outside charge gives a constant potential, so any spherical
     quon charge distribution has exactly the energy of a point quon at the centre.

Coefficient F -> lattice 1/alpha, size ratio N, odd integer, final 1/alpha:
  F=0.1667  lattice 1/alpha= 137.220233  N= 1828.196 -> 1827  1/alpha=137.235205  (1454.0 ppm)
  F=   0.5  lattice 1/alpha= 137.139212  N= 1834.686 -> 1833  1/alpha=137.160233  (906.6 ppm)
  F=   0.9  lattice 1/alpha= 137.041414  N= 1842.556 -> 1841  1/alpha=137.060715  (180.4 ppm)
  F=  0.95  lattice 1/alpha= 137.029144  N= 1843.546 -> 1843  1/alpha=137.035915  (-0.6154 ppm)
  F=   1.0  lattice 1/alpha= 137.016865  N= 1844.538 -> 1843  1/alpha=137.035915  (-0.6154 ppm)
  F=  1.05  lattice 1/alpha= 137.004576  N= 1845.531 -> 1845  1/alpha=137.011146  (-181.4 ppm)
  F=   1.1  lattice 1/alpha= 136.992277  N= 1846.525 -> 1845  1/alpha=137.011146  (-181.4 ppm)
  F=   2.0  lattice 1/alpha= 136.769164  N= 1864.673 -> 1863  1/alpha=136.789622  (-1798.0 ppm)
His rule selects 1843 (N in [1843, 1845)) for F from 0.92241 to 1.0233, i.e. the term must be right to within -7.76% / +2.33%

void volume fraction per cell v = 0.0001978 (d/b = 27.6659); if it entered the finite-size term alone, F would change by about 0.000396
