The following is a detailed description of our helium recovery system along with some thoughts related to questions I had when planning this system and things I would do differently now that I have more experience. I hope it’s useful for those looking into setting up a recovery system in their own facility. For information about installation and operating costs, see the page on cost analysis.
The CryoMech System:
The overall footprint of the CryoMech system is a bit less than 250 square feet (roughly 14 ft by 18 ft).
- Atmospheric pressure recovery bag (300 ft3)
300 CF of He gas is equivalent to 11.2 L of LHe; but, the recovery compressor kicks on before the bag fully inflates. The amount of gas that’s pumped out of the bag between the times when the compressor automatically turns on and off is equivalent to ~8-9 L of LHe. The plume collected while filling most of the magnets on my system exceeds this volume, so I turn the compressor on just before a fill and adjust the flow rate to keep the plume < 4.5 on the flowmeter (i.e., 4.5 CFM air; the actual flow rate will be different for cold He gas)); for my specific setup, this is the value that’s just slower than the rate of the recovery compressor.
- Recovery compressor (5.5 kW, 30A 3Ph)
This is the loudest part of the setup; for this reason, I don’t recommend placing it in the NMR facility. I put mine in an adjacent room and had holes drilled in the wall for the control and transfer lines. The compressor generates a lot of heat, so make sure you have a space that can deal with the heat load. I initially tried putting this compressor in a doghouse to reduce the noise; but, that led to overheating, even with fans built in to add cooling air.
The compressor has a water adsorber on it to protect the purifier. The air here in Utah is pretty dry and the only way air can enter my system is when I disconnect/reconnect the vent lines to the transfer dewars; because of this, I’ve only needed to regenerate the adsorber once every 3 months. Depending on humidity and how you use your setup, you may need to regenerate the adsorber every month.
- Medium pressure storage tanks (x8)
At the max. P (400 psig) these tanks hold the equivalent of ~16 L LHe, so that’s ~128 L for my setup. I had little floor space available, but tall ceilings; so, I put the recovery bag on a shelf above the storage tanks.
- Automatic purifier (PT60 cold head)
This uses a cold head instead of a LN2 bath for the purifier cold trap and the convenience is quite nice; it automatically warms up to purge and pump the trap as necessary.
Initially, I had issues with the trap regeneration happening too frequently; this happens when the ground pin gets corroded (ground pin replacement is the most frequent maintenance needed for the automatic purifier). I’ve also had to pump down the outer vacuum space around the cold head several times. So far, I've had to pump this down once every 6 months.
It’s worth keeping in mind that it takes ~6 hours for the purifier to cool down and ~2 hours to transition to steady state purification, so it’s only worth turning off the purifier if you have a sufficiently long gap where you aren’t producing LHe. I have tried putting the purifier in “standby” mode when not actively liquifying He; this keeps the trap cold, but doesn’t push He to the LHe plant and should resume liquefying without the long cooldown time. However, I found the purifier would usually switch to the warm-up/trap-purge mode after resuming from standby (which then requires a full cool down period), so there was no time benefit. If I want to prevent the purifier from sending He to the LHe plant (e.g., to keep some pressure in the storage tanks for an upcoming transfer of LHe to a transfer dewar), I find it’s better to simply close the on/off valve on the LHeP inlet.
Note: the purifier also has a water adsorber. So far, the water adsorber on the recovery compressor has done its job well and kept the purifier adsorber dry; there’s no indication of water in this adsorber after more than a year of operation.
- LHeP22 (PT415 cold head) with 150 L collection dewar
The nominal liquefaction rate for this LHe plant is 22 L/day, but mine has consistently performed at ~26-27 L/day (1.1 L/hour) over the past year. It’s not as loud as the recovery compressor, but it is significantly louder than the typical background noise of the NMR lab (it’s loud enough that it sometimes startles students in the NMR facility when the cold head suddenly turns on).
Starting with a warm collection dewar and a warm cold head, it takes ~50 hours to cool the system down to the point where LHe begins collecting. Because of this, it’s simpler to keep the LHeP cold rather than letting it warm all the way up between fills. For this purpose, the LHeP has an “on/off” mode that allows the liquid in the collection dewar to passively cool the system through boil off until the head pressure reaches 8 psig; at that point, the cold head turns on until the head pressure reaches ~0.6 psig where it turns off again. This runs the LHeP at a duty cycle of ~40%.
Unfortunately, the LHeP does not have a heater to help push the LHe out when filling up a transfer dewar. There is a port you can attach to a compressed gas cylinder to use as push gas; however, trace impurities icing up on the cold head are the easiest way to reduce the performance, so you’d need to use expensive UHP He gas as a push gas. Instead, I make sure I keep enough gas in the storage tanks that I can use the pressure supplied by the purifier to help push the LHe out. That does require fiddling with the inlet regulator to gradually increase the pressure during the fill: I start at ~3 psig and gradually ramp up to ~5 psig and then turn it back down to ~4 psig for standard operation.
Equipment that’s used with the LHe plant:
- Floor scale (Optima 917)
Be sure to plan ahead so you have enough space for a good quality floor scale that sits next to the LHeP for monitoring transfer dewar fills. Somehow this completely escaped my attention during planning. As a result, I was very limited in the space and funds I had available after everything else was installed. My scale works fine and was not expensive, but there’s two significant annoyances: First, the ramp is steeper than ideal, so it takes some wrestling to get a transfer dewar on/off the scale. Second, this scale only reads ±0.5 lb which is ±2 L when recording volume readings; it would be much easier to monitor the filling of transfer dewars with a resolution of at least ±0.1 lb (±0.2 L) and my volume estimates of the LHe in the transfer dewar would be more accurate.
- Vacuum pump (Edwards nXDS6i + Kurt Lesker 275i Pirani guage)
In my initial planning, I didn’t realize that I would need a vacuum pump for regenerating the water adsorbers and pumping down the vacuum space in the purifier. I suspect many of you already have several pumps, so this is probably not an issue for most people planning a recovery system.
Given my limited funds, I purchased a simple scroll pump (max. vacuum achievable ~15 mTorr). This pump works well enough, but it takes > 2 weeks to regenerate a water adsorber (the manual says it should only take 2-3 days!) In theory, I should be using a turbo pump to get the vacuum space to a much lower pressure, but if it’s just going to lose vacuum within a few months, I haven’t felt the need to borrow a turbo pump. If I did pump it down further, that should shorten the cool down period for the purifier; but again, if that only lasts several months, I’m not sure it’s worth the trouble.
- LHe transfer dewar (Cryofab CMSH-100)
I have two 100 L transfer dewars. The standard one has an average boil-off of ~1.3 L/day whereas the one with a built-in heater has an average boil-off of ~2.1 L/day. These can go 5-6 weeks between topping them off to keep them cold. Depending on your fill schedule, most of this can be taken care of by simply topping off the dewars after each fill with an extra transfer or two when there’s a long gap between magnet fills. This approach, however, requires having enough excess He in your recovery system that you have LHe on hand to top off the transfer dewar after filling a magnet.
Note: to capture the boil-off from the transfer dewars, I put ½” F NPT to ½” JIC flare (3/4-16) adapters on the vent ports. The large ID allows for normal venting of the dewar during a fill while allowing quick connecting/disconnecting with typical extreme temperature flex lines which often use ½” JIC flare fittings (e.g., McMaster-Carr 50675K176 and 5676T12).
- He leak detector
A leak detector is essential for testing the collection circuit, connection manifolds, and the connections on the various parts of the CryoMech system. This is a good tool to have from the beginning as you test the system. You'll want to test the connections regularly, especially the connections between the magnet and the collection circuit that undergo thermal cycling during the He fills. A variety of portable leak detectors are available based on thermal conductivity detector (TCD) probes. I have a simple Restek 28500 that gives qualitative measurements to help find leaks; there are a number of other similar detectors at a similar price point. There are also more sophisticated (i.e., more expensive) models that offer quantitative measurement, data logging, etc.
The He collection circuit:
- Main collection circuit
The main collection circuit is assembled from 1” capped and cleaned Cu tube connected with Pro-Press fittings. The recovery system is in the basement in the NMR facility. Another magnet is down the hallway in the basement and two more magnets are up on the 3rd floor. Consequently, this required a lot of Cu tubing. Even using 1” tube, there’s a noticeably higher back pressure on the magnets upstairs as well as the one in the basement. I would recommend 2” or 2.5” tube for more distant magnets (if you can afford it) to reduce this back pressure as that increases the transfer loss when filling the magnet (i.e., you can see quite a bit more He gas escaping out of the inlet port on these magnets vs. filling with the exhaust open to the atmosphere).
I was initially skeptical of using Pro-Press fittings. Our physical facility people prefer Pro-press because it’s faster and cleaner than welding. For example, I needed our phys. fac. people to weld some brass KF25 nipples onto Cu pipe and the amount of black crap I had to clean out of the pipe was astounding! So, cleaned and capped Cu tube + Pro-Press results in a very clean collection circuit. Pro-Press does have a specific o-ring rated for He gas; these had to be ordered separately and swapped out for all the standard o-rings during assembly. There will still be water adsorbed to the surface and air in the lines, so I did several rounds of pumping and purging the collection circuit with He gas followed by pumping for a couple weeks. I then tested for any major leaks by closing the system and monitoring the pressure for a couple weeks. I started at 25 mTorr and saw no significant change in pressure.
As another check on the collection circuit, I recorded the total amount of He in the recovery system (transfer dewars, collection dewar, storage tanks, with an empty collection bag) before and after every fill. I compared the volume of He collected between fills to the predicted He boil-off for each of those time periods and integrated those values over the past year (after accounting for transfer losses) and I see no evidence of any He loss beyond the transfer losses.
I do have some concerns about how the o-rings in the Pro-Press fittings closest to the magnet will hold up over time due to the thermal cycling. I’m not sure I have any basis for this concern – after all, the o-rings in the magnet turrets handle thermal cycling every week from LN2 fills and every few months from LHe fills – but, it’s probably worth checking with a He leak detector periodically (once a quarter?)
- Manifolds for connecting magnets to the collection circuit
I copied Yale’s design for the manifolds. Here is some info for the major parts for quick reference (along with some thoughts, when relevant):
- Flow meter, 0-8 CFM (Omega FL7611) – this works beautifully for monitoring the plume during a fill. Typical flow rates during a magnet fill result in the plume reading 3.5-4.5 CFM on this meter; a rate that’s compatible with the recovery compressor. When the magnet is full, the flow quickly jumps to > 7.0 on the meter.
- Back-pressure regulator (Control Air Inc., 700BP; McMaster-Carr 4783K53)
- One way valve (Eclipse 1006A Disc-type Check Valve) – these “swinging disc valves” have very low cracking pressure, but they’re not guaranteed to be gas tight against backflow. I have them in the manifold as a safety measure – if a valve or KF25 fitting was left open, this check valve would turn a big leak into a slow leak.
- Heat exchanger (Zephair ZAS040NZA-F) – These are perfect for use with Pro-Press because of the 1” Cu ports. Note, however, that the 8x8 size is too small, in my opinion, despite what the BTU rating claims. I would get at least the 15x24 or 20x24, if I were building these again (e.g., see the image of filling the transfer dewar – the line after the heat exchanger is completely frozen).
- The magnet connection is made with a KF25 tee, a KF25 bellows, and KF25 nipple welded to Cu tube which is then connected via Pro-Press. – The goal here is to send normal boil-off through the usual pathway (a small check valve followed by a flow meter) and then into the back pressure regulator when the valve after the flex line is closed and the T-valve connects the back pressure regulator to the collection circuit. During a fill, the valve after the flex line is opened and the T-valve connects the heat exchanger to the collection circuit.
Update: Now that I have consistent access to a leak detector, I have started checking the ProPress fittings more regularly before, during, and after a fill. The fittings that are Cu connected to Cu show no detectable leak even when frozen (e.g., several ProPress fittings were frozen for ~45 minutes while filling a transfer dewar and no detectable leak was observed). However, on the transfer dewar manifold, there is a pair of brass ProPress adapters connected to Cu pipe (i.e., a ProPress to NPT connector) - these positions have no detectable leak before or after the fill, but have a clear leak when frozen (not surprising considering we have different metals at this joint). I suspect a Cu NPT nipple with a Cu ProPress fitting would work better here, but I need to get this modified and tested before I can say for sure. For those planning their systems, this is only a problem at positions that freeze over during a fill (i.e., before the heat exchanger).