The continuous spectrum of sunlight can be approximated by using N
discrete wavelengths spaced evenly between 0.38 µm and 0.7 µm.
Normal colorimetric practice uses a relatively small number of wavelengths
(e.g. 7  30) but Mie calculations can give misleading results when only
a few wavelengths are used. It must be emphasised that this problem
is not caused by the spectral characteristics of the incident light: indeed,
such problems occur even if a flat spectrum is used.
The fundamental cause is that the fine ripple structure
visible in graphs of Mie scattering of monochromatic light is very dependent
on wavelength. As calculations for multiple wavelengths involve the addition
of intensity curves for different wavelengths, the ripple structure will
be reinforced if the maxima and minima for different wavelengths coincide:
for example, the maxima and minima of the ripple structure shown here
for 650 nm coincide approximately with those for 650.9 nm, 651.8 nm and
so on.
The following diagrams explore the effect of variations of N on simulations
of rainbows, glories and coronas.

Fig. 1 Simulation of the primary rainbow (for
unpolarised light) for a water drop of radius 200 µm
for the specified numbers of wavelengths spaced evenly between 0.38 µm
and 0.7 µm 

Fig. 1 indicates that simulations of the primary rainbow for r = 200
µm are almost identical when calculated using more than 15 wavelengths,
apart from minor changes in brightness. Hence, there is little value
in using more than 30 wavelengths for this condition. 
Fig. 2 Simulation of the primary and secondary
rainbows (for unpolarised light)
for a water drop of radius 100 µm for the specified number of wavelengths 

Fig. 3 Simulation of the secondary rainbow (for
unpolarised light) for
a water drop of radius 100 µm for the specified number of wavelengths 

For r = 100 µm, Fig. 2 shows minor variations in the appearance
of the primary rainbow (137°  150°) for 10 or more wavelengths,
but the secondary rainbow (120°  129°) shows significant
ripples when less than 100 wavelengths are used. The secondary rainbow
is shown in much greater detail in Fig. 3, indicating that perhaps 300
wavelengths are needed for consistent simulations of the secondary rainbow. 
Fig. 4 Simulation of the glory (for unpolarised
light) for a water drop
of radius 10 µm for the specified number of wavelengths 

Fig. 4 indicates that 10 µm drops result in a glory consisting
of a bright red circle centred on the antisolar point with a radius of
~2.1° (scattering angle of 177.9°) and a weaker red circle with
a radius of ~3.7° (176.3°).
The variations between the horizontal stripes of Fig. 4 show that MiePlot
simulations of the glory are very sensitive to the number of wavelengths
used in the calculations. Whereas the simulations of the primary
and secondary rainbows in Figs. 1  3 are fairly consistent with 30 or
more wavelengths, it is obvious from Fig. 4 that many more wavelengths
(more than 500) are required to achieve an accurate simulation of the glory.

Fig. 5 Simulation of the corona (for unpolarised
light) for a water drop
of radius 3 µm for the specified number of wavelengths 

Fig. 5 indicates that, unlike the glory and the rainbow, consistent
simulations of the corona can be produced using only a small number of
wavelengths. 