We have written a light curve synthesis code that makes direct use of model atmosphere specific intensities , in particular the NextGen model atmosphere grid for cool giants ( T _ { eff } \leq 6800 K and \log ( g ) \leq 3.5 , Hauschildt et al . [ 1999 ] ) .
We point out that these models ( computed using spherical geometry ) predict a limb darkening behaviour that deviates significantly from a simple linear or two-parameter law ( there is less intensity at the limb of the star ) .
The presence of a significantly nonlinear limb darkening law has two main consequences .
First , the ellipsoidal light curve computed for a tidally distorted giant using the NextGen intensities is in general different from the light curve computed using the same geometry but with the black body approximation and a one- or two-parameter limb darkening law .
In most cases the light curves computed with the NextGen intensities have deeper minima than their black body counterparts .
Thus the light curve solutions for binaries with a giant component obtained with models with near linear limb darkening ( either black body or plane-parallel model atmosphere intensities ) are biased .
Observations over a wide wavelength range ( i.e .
both the optical and infrared ) are particularly useful in discriminating between models with nearly linear limb darkening and the NextGen models .
Second , we show that rotational broadening kernels for Roche lobe filling ( or nearly filling ) giants can be significantly different from analytic kernels due to a combination of the nonspherical shape of the star and the radical departure from a simple limb darkening law .
As a result , geometrical information inferred from V _ { rot } \sin i measurements of cool giants in binary systems are likewise biased .