[RASMB] time to reach equilibrium

[Peter Schuck]

Patrick,

We usually let it go fairly long (at least 2 days for a 180ul sample) and 
take scans in time-intervals of 6 h.  As Jim already said, with WINMATCH, 
you can then subtract the last scan from all previous ones taken at several 
time-points earlier and look how the difference slowly approaches a 
constant (i.e. a difference only due to random noise) in the end.  In my 
experience, this can really take surprisingly long, in particular if you 
have a chemical reaction.  There's sometimes a long period when the 
profiles change relatively little during any few hours period, but they 
still keep going if observed over longer periods of time.

In my view, one of the critical problems is that one might be able to 
actually fit the data if they are really still relatively far from 
equilibrium with an apparently reasonable model. (Obviously you would get 
wrong parameters.)  That's because even in non-equilibrium, the profiles 
take the shape of a curve with monotonously increasing slope, which are 
notorious for looking just like noisy sums of exponentials.

If you can keep the temperature low and don't have to worry too much about 
degradation, it's frequently worth the wait.  In our hands, it's usually 
considerably faster after establishing equilibrium at one rotor speed when 
going to the next higher rotor speed.

Regards,
Peter