Peak width parameters:
======================

see, for instance, http://www.mail-archive.com/rietveld_l@ill.fr/msg03908.html
from where this text has been extracted.

 
Have a look at Young, R. A. & Desai, P. 1989, 'Crystallite Size and
Microstrain Indicators in Rietveld Refinement', Archiwum Nauki o
Materialach, vol. 10, no. 1-2, pp. 71-90. 
 
They talk about the Thompson, Cox and Hastings model, which explicitly
separates the gaussian and lorentzian components of a psuedo-Voight peak
shape.
 
FWHM(G)^2 = U tan^2(T) + V tan(T) + W
FWHM(L) = X tan(T) + Y/cos(T)
 
The coefficients can be broken into instrumental and sample (size, strain)
components.
 
U = U_inst + U_strain
V = V_inst
W = W_inst
X = X_inst + X_strain
Y = Y_inst + Y_size
 
You can fix the instrument components with your standard, and then
refine the difference with your sample.
 
 
If you want to stick with the straight UVW symbolism, Young and Desai
also state that you can use the size broadening term FHWM(G)^2 =
Z/cos^2(T), which yields:
 
FWHM(G)^2 = Z/cos^2(T) + (U_inst + U_strain) tan^2(T) + V_inst tan(T) +
W_inst
 
which can be re-written as
 
FWHM(G)^2 = (U_inst + U_strain + Z_size) tan^2(T) + V_inst tan(T) +
(W_inst + Z_size) 
as long as you constrain the two Z_size's to be the same.