Dear Peter, the most correct constraint would be to fix a sum of coordinates of all atoms along polar axis, but not only of the first two. It gives the best estimation of the standard uncertainties.
Yaroslav Filinchuk PZ> Bob, PZ> Thank you for the clarification. I never had doubts that GSAS handles fixed origin issue properly but also never understand how it PZ> is done. This way has one big advantage over user fixed origin -- it yields standard uncertainties for all atoms and therefore for PZ> all distances, etc. PZ> Peter Zavalij PZ> -----Original Message----- PZ> From: Von Dreele, Robert B. [mailto:[EMAIL PROTECTED] PZ> Sent: Wednesday, March 31, 2004 10:17 AM PZ> To: [EMAIL PROTECTED] PZ> Dear Stephen (& others), PZ> I know there have been a number of replies to this since P2/n does have an inversion center which is positioned at the unit cell PZ> origin. However, if the space group is Pn (or P2) then the location of the origin is arbitrary on one or more axes. GSAS does PZ> "automatically" handle this by forcing the first and second atoms to have the opposite shifts for each of the arbitrary axes. So in PZ> Pn the shifts for atom #1 will be sx,sy,sz and for atom #2 they will be -sx,sy',-sz, i.e. the sx & sz shifts will be opposite but PZ> identical for these two atoms while each will have it's own sy. Beware, different damping applied to these atoms will make it drift PZ> as the applied shifts won't be equal & opposite. If you do not want this to happen then put in your own "hold" on the appropriate PZ> coordinates for one atom or else put in your own constraint. PZ> Bob Von Dreele PZ> ________________________________ PZ> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] PZ> Sent: Wed 3/31/2004 12:02 AM PZ> To: [EMAIL PROTECTED] PZ> Dear Rietvelders, PZ> I used GSAS to refine my structure in P2/n, however the program automatically PZ> fixed the the XYZ positions of the first atom during refinement. To my PZ> surprise, this atom is not seen in the "fixed atom list" in the atom parameters PZ> menu of GSAS. PZ> Please advise and many thanks, PZ> stephen