Hi again Ed,
Is this Hart et al study, then their conculsion is: "Field measurements clearly show a correlation between gap width deflection and temperature,
but it is difficult to draw any definitive conclusions from our data regarding the correlation..." which can be interpreted as a warning against relying solely on that study, and particularly deriving one 'optimal' air gap to fit all.
also they seem to think that it is worse to have a smaller gap than to have a bigger one: "Units designed with smaller than optimal gaps may have significant U factor changes associated with inherent gap variability compounded with temperature induced gap reduction." & "At gap width greater than 4-5 mm the heat transfer rate decreases steadily then levels off at approximately 11 mm for Argon (13 mm for Air and 9 mm for Krypton) at which point boundary layer dominated convection starts. Boundary layer convection creates an isothermal core in the middle of the cavity, making further increases in gap width essentially inconsequential to overall heat transfer"
But with all things considered, there are many other far more important points like installation that will render a difference of 2-10mm in the air gap negligible. Consider a situation where a perfect window is installed in such a way where there is a gap AROUND the window which the builder can cover up with a bit of architrave for aethetics. That gap is not visible, but it is still there, and is hardly condusive to optimal thermal insulation. So it is much more important to take a wholesome approach, rather than focusing so much on the air gap that one forgets about all the other aspects of window performance.