Postcrystallization treatment.
Rationale- We have often obtained crystals of membrane
proteins but found the resolution limit to be on the
order of 10-30 A, so that we only see a few diffraction
spots around the beamstop.
While these crystals are useless for structure determination,
in principle they represent a tremendous step towrd
our goal in terms of arrangement of the molecules and
reduction of entropy. 10-30 A is small compared to the
dimensions of the molecules (100-200 A), so this level
of order means we have hundreds of millions of molecules
lined up in a three-dimensional array, but with a little
bit of rotational and translational "jitter" in their
positions. Now suppose instead of going back to repeat
the crystallization under different conditions in hope
of getting better order, we can take the already-formed
crystals and change conditions slightly so the the
pre-positioned molecules click into position with good 
order. This strategy also allows for the fact that the 
best conditions for growing a crystal may not be the
conditions giving the best order once the crystal
is grown.

As for dehydration, in the case of the orthorhombic bc1
crystals, we dehydrate by mounting the crystal in a loop,
holding the loop in the air at room temperature for
1-3 minutes, then freezing in liquid nitrogen.

This does two things- reduce the water activity
enough to prevent ice formation, and collapse the lattice
to the point where it is more rigid.

To make this more reproducible, so it doesn't depend
on things like relative humidity, air drafts, and size
of the crystal, we fish out the crystal on a loop
in a Hampton-Research "cryocap" pin base. Then screw
the cryocap into a cryovial with 200 ul of "humectant"
solution in the bottom to define humidity, and incubate
overnight at the temperature the crystals were grown
before freezing. In a typical experiment there would
be 5-10 crystals in vials with different humectant,
ranging from 25% to 70% w:w glycerol. At 70% the diffraction
is lousy and there is a powder-pattern ring at 4.37? A
from precipitated PEG. At 25% there are ice rings and
little or no diffraction. somewhere in between there is
good diffraction.
The unit cell parameters are extremely variable, with
the a edge going from 160 down to 120 A, best diffraction
at 128 A. And there is a new contact at 128, which was
well separated in the larger cells. So we think the
lattice collapses with dehydration until it hits a hard
stop when this contact is made, then further collapse
causes chaotic deformation and worse diffraction.

However the complex II crystals are damaged by any
dehydration, and the lattice parameters are pretty
constant. In that case I think the lattice is already
rigid. Maybe a slight increase in cell volume would help.
So we treat with mother-liquor plus glycerol and freeze

I think a 2-dimensional optimization is necessary:
The water activity has to be between the limits set by
ice formation and PEG precipitation. Then the cell volume
should be just right to give tight lattice contacts
without disruptive disordering. Dehydration varies
both simultaneously. By adjusting the glycerol
concentration before the dehydration step, the
curves can be shifted over so that a given water activity
corresponds to a larger or smaller cell volume.