TL;DR: A surface is cycled between a high vacuum chamber, where it is chilled to condense gas molecules, and a low vacuum chamber, where it is heated to drive them off. This should result in the high vacuum chamber being highly evacuated.
(This is on my list right above the other vacuum pump idea I just posted [link], so I think I had this idea first. It makes sense that I could have thought of this one first, because it's more closely related to the existing vacuum pumps, specifically cryopumps, I was thinking about at the time.)
This is a type of high vacuum pump intended to lower costs by not requiring much high-precision hardware. Like the differential solubility pump (the previously posted idea), it's used to evacuate a high vacuum chamber when you already have low vacuum provided by a conventional low vacuum pump. Unlike the differential solubility pump, this one doesn't need a vacuum-compatible gas-dissolving liquid, but it needs more intense heat pumping.
On the other hand, while the differential solubility pump can work with a high vacuum chamber and a low vacuum chamber (or the low vacuum coil option, or the gaseous elution option) separated by a distance limited only by the length of the solvent pipes and heat pump pipes between them, the C/C/C-CCCC needs the two chambers to share a wall. This shared wall needs an opening in it. The opening is occupied by the circulator, leaving only the smallest possible gap so it doesn't rub as it rotates. In the 'circular' version, the opening is a slot, and the circulator is a circular disc, with nearly half of the disc protruding into each chamber, either perpendicularly to the wall, at an oblique angle, or integrated into a side wall/floor/ceiling with only one side exposed. In the 'cylindrical' version, the opening is a rectangle, and the circulator is a cylindrical drum, in an analogous orientation to the disc, with nearly half protruding into each chamber. In the 'conveyorbeltlike' version, the opening is either one slot or two slots, and the circulator is a belt that has part of itself in each chamber, able to circulate by passing through the slot(s).
In all versions, this kind of cryopump works similarly to a traditional cryopump [link] on the high vacuum side. With a conventional cryopump, a plate is chilled and, therefore, when gas molecules strike it, they are likely to freeze there due to the cold and become stuck to it, condensing out of the gas phase in the high vacuum chamber. In a traditional cryopump, this is all that happens, and it can become saturated with gas molecules, meaning its pumping speed gradually decreases as it fills up with frozen molecules, until it eventually can't pump any more and will just hold onto the ones it already has, at which point it's no more effective as a vacuum pump than a similarly sized section of chamber wall.
That's where this pump's circulation aspect comes in. The circulator is continuously slowly rotating, exposing new condensant surface to the high vacuum chamber. At the same time, condensant surface with molecules trapped on it is moved into the low vacuum chamber and heated to drive off those molecules, so that the surface will be ready for its next cycle through the high vacuum chamber. In traditional cryopumps, this bakeout must be done to the whole pump at once on a scheduled or as-needed basis, and necessitates either reducing the high vacuum chamber to low vacuum while it's happening or building in a system to isolate the cryopump from the high vacuum chamber for the duration of the bakeout.
A further advantage over traditional cryopumps is that it can take better advantage of the 'cryotrapping' mode of operation, not just the 'cryopumping' mode. Cryopumping refers to the gas molecules being more or less permanently frozen onto the condensant surface (staying there until bakeout), and requires the surface to be colder than the boiling point of the molecules. Cryotrapping, on the other hand, is a temporary effect where gas molecules stay on the surface for a time, without actually freezing, between impact and separation, and works even when the surface is warmer than the boiling point (but still pretty cold). In a conventional cryopump, which has nowhere to dispose of molecules, these cryotrapped molecules eventually just jump off back into the high vacuum chamber, meaning that cryotrapping isn't as effective at long-term vacuum production as cryopumping is. But, with a circulating condensant surface, these molecules too are disposed of in the low vacuum chamber during bakeout.
Being able to dispose of molecules on a continuous basis makes this pump more suitable for pumping a chamber down to high vacuum from low vacuum, so it can be a substitute for a turbomolecular pump (expensive) or a diffusion pump (messy) where a regular cryopump can't.
There will necessarily be a small gap between the circulator and the edge of the opening in the wall, but, due to how gas molecules move in vacuum, I expect this won't cause much trouble. The wall thickness can be increased (or a thick ring can be added around the opening) to increase the distance the gas molecules must travel to get from one chamber to the other, if necessary. Even if some do make it from the low vacuum side to the high vacuum side without striking the opening edge or the circulator, there's a pretty good chance they'll immediately strike the cold condensant surface of the circulator as soon as they get out of the gap.
To increase efficiency, a heat recovery system can be integrated to recover heat/cold in each half of the circulator:
In the 'circular' and 'cylindrical' versions, there can be heatpipes integrated inside the rigid circulator, configured to equalize the temperature between the sector of circulator going from low vacuum (hot) to high vacuum (cold) before being heated or cooled by the heat pump, reducing the load on the heat pump. There can be just one heatpipe, or several in parallel (each with endpoints on a line parallel to the wall, not each with endpoints on a diameter of the circulator), analogous to a counter-flow heat exchanger for fluids but discrete rather than continuous.
Alternatively, the circulator can be filled with a gas to convect heat internally, ideally a gas with high heat transfer capability such as high-pressure helium. In this case, there would just have to be a wall inside the circulator, notionally coplanar with the wall in which the pump is set, to keep the hot gas and the cold gas separate. Unfortunately, it seems that convection would be forward in one side and backward in the other side, if the circulator rotated about a horizontal axis. With a vertical axis, it would be neutral in both sides; some sort of stirring mechanism inside may be beneficial.
In the 'conveyor' version, if the two runs of belt through the wall share a single slot, there can be a cylindrical roller between them. This not only fills the gap to block flow from the low vacuum side to the high vacuum side, but also transfers heat from the section of belt going from low vacuum (hot) to high vacuum (cold), reducing the load on the heat pump. Again, there can be multiple rollers in parallel, again analogous to a counter-flow heat exchanger. In this case, though, they can be just solid rollers rather than heatpipes, transferring heat from one part of the belt to the other by rolling from one side to the other.
As with a traditional cryopump, this kind of pump is amenable to coating the condensant surface with a sorbent material to increase trapping. It's also more obviously worthy of the name 'pump', directly transferring matter against a pressure gradient.
70/480 [between 2017-01-08 and late April 2017]-- notexactly, Oct 02 2019 Differential solubility vacuum pump by me. Mentioned in idea body. The other vacuum pump idea I just posted. [notexactly, Oct 02 2019] I had an immediate mental image of “Maxwell’s Demon” as I read this.
Looks like it will work, good idea [+]-- Frankx, Oct 04 2019 Thanks for the opinion; I may attempt to build this at some point. Also, I added a TL;DR at the top, because I think you're an outlier in that you seem to have read the whole thing.-- notexactly, Oct 05 2019 random, halfbakery