Disclaimer: these examples are mostly taken from my pre-Eikomagos career
The system described here is comprised of three parts:
To obtain rounded crystals the sample is plunge frozen to -100°C. This puts the sample in a glassy state. On heating it up, at around -30°C as the sample leaves the glassy state, a lot of tiny crystals are formed which grow rapidly while the temperature rises further to the target temperature (e.g. -5°C).
To support this procedure, the stage contains a liquid N₂ cooled block and a Peltier cooled block. The first is used to plunge freeze the sample; the second at the target temperature and is able to maintain its temperature very accurately (<0.01°C)
Apart from loading the sample and (dis)connecting N₂ gas at the appropriate time, the system is fully automated: the temperature, the sample's position, focus and neutral density filters, condenser aperture, field stop as appropriate (Köhler) for the objective. The output is a series of images and a log file that contains the images' name, microscope settings and temperature. These are taken for offline analysis.
The video shows a typical recrystallisation experiment taken under polarised light. Note the mobility of the ice crystals and the recrystallisation process: smaller crystals melt and larger crystals grow larger.
The use of polarised light makes for prettier pictures and visualises changes in orientation of the crystals, but actual the experiment are run in standard transmitted light mode: crystals with their main axis parallel to the viewing direction have the same colour as the background so the colours cannot be used to identify (all) the crystals. The dynamic range needed is also greater.