Everyman’s Turbomolecular Pump

What can you do with a very good vacuum pump? You can build an electron microscope, x-ray tubes, particle accelerators, thin films, and it can keep your coffee warm. Of course getting your hands on a good vacuum pump involves expert-level scrounging or a lot of money, leading [DeepSOIC] and [Keegan] to a great entry for this year’s Hackaday Prize. It’s the Everyman’s Turbomolecular Pump, a pump based on one of [Nikola Tesla]’s patents. It sucks, and that’s a good thing.

The usual way of sucking the atmosphere out of electron microscopes and vacuum tubes begins with a piston or diaphragm pump. This gets most of the atmosphere out, but there’s still a little bit left. To get the pressure down even lower, an oil diffusion pump (messy, but somewhat cheap) or a turbomolecular pump (clean, awesome, and expensive) is used to suck the last few molecules of atmosphere out.

The turbomolecular pump [DeepSOIC] and [Keegan] are building use multiple spinning discs just like [Tesla]’s 1909 patent. The problem, it seems, is finding a material that can be made into a disc and can survive tens of thousand of rotations per minute. It’s a very, very difficult build, and a mistake in fabricating any of the parts will result in a spectacular rapid disassembly of this turbomolecular pump. The reward, though, would be great. A cheap turbomolecular pump would be a very useful device in any hackerspace, fab lab, or workshop garage.

The HackadayPrize2016 is Sponsored by:

52 thoughts on “Everyman’s Turbomolecular Pump

  1. I wonder if 15K hard drive platters would be balanced and stable enough for the higher rotational speeds required by this device…and if they would have too much mass. (Purely speculative…I didn’t look at any of Tesla’s original design theory).

    1. 15k is far, far too low. I’m using the Edwards EXT250 series primarily, which are already very large and heavy (~8kg) and it spins at 60k; moreover, until it gets to about 85% of the full rotational speed, you are not going to see any shred of high vacuum. The EXT70 series, more popular because of its smaller size, spins at 90k. In a turbopump what matters is the linear speed on the outskirts of the rotor, which should be close to the average speed of the gas molecules, and even with 15k it’s possible to achieve such a linear speed, but the pump will be *very* large and massive, which will cause many problems: cost, bearings, shop time… Not to mention the amount of kinetic energy stored in the rotating rotor and then being liberated in an event of pump seizure.

      You can get perfectly functional second-hand Edwards pumps from eBay for as little as $500, and unlike with other vendors, they can be completely disassembled and serviced with just regular shop tools. In fact for most part you won’t need anything except a set of imperial allen keys.

      Building them? A complete non-starter. You need to balance the rotor. To give you a perspective on how difficult this is, turbopumps are a comparatively widely used industrial component, and there are entire *countries* without a single rotor balancing stand, as I’ve found out while searching for a vendor that will sell oil for my EXT250.

      Also, if you don’t have a drag stage in your turbopump, then you cannot use it with a cheapo rotary vane pump, such as sold for evacuating air conditioning systems. You need 10^-3 to 10^-2 torr at the outlet, which means you need a scroll pump or a decent two-stage rotary vane pump, which would easily blow through that $500 budget alone. (The EXT series has a drag stage and it easily achieves high vacuum at 10^-1 torr or even 10^0 torr outlet pressure.)

      1. I’m trying to put one of these back together (EXT-250H) but in the condition I received it in the oil sump parts and bearing were gone. All I could reconstruct is that It’s an 8mm x 22 mm hybrid bearing with a single side shield, PEEK cage and oil mist lubrication. The width ought to be around 10mm. Any help appreciated.

        1. It’s not oil mist… it’s lubricated by drawing oil from a set of felt discs (probably some synthetic replacement rather than actual felt). One of the felt discs have small protrusions that touch the shaft, which below the bearing is shaped like an inverted cone. The oil on the shaft is drawn up by centrifugal force. See the cross section in the datasheet, which shows that unit quite well. The bearing is ceramic but that’s about all I know about it.

          I’ve never taken out the bearing myself so I would not do that to one of my pumps just to show it to you, sorry. (I know that it’s easily possible since I talked to people who did that, I never did though.) With the bottom bearing being one of the hardest parts to source, I’m not sure if it makes a lot of sense to repair that pump vs get another one, maybe even a broken one, with a destroyed rotor or flange or something, and assembling one working from them.

          1. Thanks for the details on the lubrication assembly. There is a ~$100 service kit with the oil wicking parts sans bearing. I once even ordered a 10.319mm (13/32″) wide bearing to check how it seats but couldn’t find a proper spindle bearing with these dimensions and tolerances. There’s also a $500 kit with the bearing but of course they’re not telling you what the type is… sorry the write-up is not in english.

        2. I’m pretty sure the bearings are custom. I know for a fact the oil is a formulation custom made for Edwards; specifically Cambridge Mills CMP500. You can get it from AJVS (https://www.ajvs.com/sklep.new/product_info.php?products_id=7237) but I’ll gladly send you 15ml (that’s at least two refills) at nominal cost since I have far more than I need.

          That said… I also know people machine the pump case to fit off-the-shelf turbine bearings, so that’s probably your best chance. Find one that’s undersized for the shaft and oversized for the case, and machine away the difference, then fit it in.

          1. ​Hello Whitequark, I enjoyed your comments about being able to rebuild Edwards pumps with shop tools. I hope they apply to EXT70’s too. I see you have an EXT250 turbo? I am rebuilding an EXT70 so I can enjoy the benefits of good vacuum too. I have inherited two pumps, one in bits, and one complete. Both very old, so I’m going to tear them down and check what I can and see what happens. I’d like your advice on oil. I have a bottle of Varian Type GP oil which is specified for two stage rotary vane oil pumps in mass spectrometers. Do you think it is suitable for an EXT70. If not may I avail myself of your offer to supply 15mL or so?

          2. Andrea, you will have a relatively harder time with EXT70. On EXT250 the rotor can be easily disassembled, cleaned and reassembled if necessary; on EXT70 it is milled from a single piece of aluminium. Thus, if you do not want to lose the oil, you will have to do something like this to clean the pump: invert the EXT70 upside down, and then put it (on a suitable support) into a heated diethyl ether bath, such that vapors of ether condense inside and carry away contaminants. Ether is extremely flammable and somewhat toxic, so this is undesirable.

            However if you can afford losing the oil, then it is as simple as flushing the pump through with copious amounts of ether or R-141b or a similar organic solvent. I prefer HFCs/HCFCs for safety reasons but ether is greener and perhaps easier to obtain. In any case don’t use anything that would dissolve the lacquer on the motor windings or wires, like acetone; if in doubt try it on any induction motor you have around. Dunk the motor in oil and then apply the chosen solvent to it. If there is no traces of oil afterwards and the motor still works and isn’t visibly damaged, then this solvent is probably safe.

            Regarding the oil, rotary vane pump oil should UNDER NO CIRCUMSTANCES enter your turbopump. Nor even its vapor.

            I am out of country but write me at whitequark@whitequark.org and I will try to arrange a small vial of oil for you.

      2. Nearly all commercial turbo pumps are made to pump down really fast for the semiconductor manufacturing processes. For example, RTP – Rapid Thermal Processing – uses flat pancake chambers to minimize volume, and cycle quickly.

        It seems to me that 10K or 15K hard drive parts, though a lot slower, will still increase the probability of a molecule getting knocked out to the edge versus being in the center. The disks versus paddles of a typical turbo pump is totally different as well. It might be too slow to be practical compared to out-gassing, etc. But if you have the time, it might be OK as a replacement for a diffusion stage. I would be really surprised if there were not some research and test results somewhere out in the interwebs. Old servers are full of 15K drives – like 8 drives per server.

        Note to Tesla enthusiasts. Despite has many inventions, this is not based on a Tesla patent. The principle of operation is totally different.

          1. I don’t want to give the wrong impression — ooh! I have been listening to the live meteors all day — about knowing anything about vacuum. I’m not an expert. If I had to make a cryogenic system with no helium leaks it would be a lot of reading. But I was a research physicist at Sekidenko with a high vacuum system of my own with turbo-molecular pumps and all the right gauges. We developed devices for measuring the temperature of wafers in RTP by their IR signatures. Ever held a sapphire 6 inches long and 2 inches in diameter? The chain went Sekidenko –> AMAT –> Intel. And since Intel is Copy Exact, some research gear duplicated an Intel RTP line and other gear was more versatile. My last project was designing a “pocket” sized portable black body chamber for calibration. I made one with Shuttle tile material and one with B1 Bomber insulation. A 2mm sapphire rod brought the IR to a photo-diode. Each worked nicely. As far as I know, never went to production.

            Anyway it was all kinetic theory for RTP and the wee little low current end of the full Ebers-Moll junction model and radiative transfer stuff that gets too complicated.

            I think the closely space spinning disks and stator disks is worth testing and mightnot have to go very fast to get a measurable effect.

    2. Using parts from a HDD seems like a pretty good idea to me.

      The platters are amazingly well made (RMS roughness ~10 Angstrom) and very well balanced / centered about the spindle.

      HDD’s also have very nice air bearing based motors (so no bearing ware and capable of running at the 10’s of kRPM). In atmosphere the drive motors might struggle to get up to the required speed due to air resistance but in a low pressure environment this might be different?

      1. That is a good idea! They’d make perfect discs for sure. The reason I’ve avoided them so far is simply an issue of space. The gap between the discs needs to be pretty small for the rough vacuum supplied by the backing pump to approach molecular flow (about .016mm). But the available pumping throughput is directly related to gap size too. On a regular turbopump, a huge pump with a 10 inch diameter opening will pump a lot faster than a small 3 inch one. In my design, the “pump opening” is the surface area of all the gaps (2*pi*InletRadius*GapSize). But I need a small gap to get molecular flow, so to keep up the throughput I need a LOT of gaps (like 100+). That could get cumbersome with hard drive platters, so that’s why I’m trying for foil discs instead. Household aluminum foil happens to be about .016mm thick, so if I can reliably cut discs out of foil I can use it for the discs and for spacers between them. Since it’s just foil, I could put several hundred layers and it’d still be fairly light and thin. Thanks for the ideas!

        1. I expect household aluminum foil isn’t going to be strong enough, its nearly pure aluminum. However, if you can figure out the fabrication process with it, the process should be nearly identical with commercially available engineering foils such as stainless steel. A quick search turned up a supplier that offers not only stainless, but Titanium and other foils in various thicknesses: http://www.ulbrich.com/products/ulbrich-ultralite-foil-products/

  2. Most likely an idiotic thought brought about by the late hour but after reading the above I couldn’t help but think of hard drive platters.

    Yeah…..I definitely need sleep……

  3. Given the level of precision engineering needed, I suspect this is a non-starter for scratch fabrication in your garage.

    If you look at a gas centrifuge, just to look at the extreme case, the rotational velocity is so high that (assuming perfect vacuum bearings and balancing) you’re limited by the tensile strength of the rotor material. (Higher rotational velocity increases the single-stage separation ratio markedly, so it’s desirable to push it as high as possible.)

    In terms of tensile strength, aluminium alloys are OK, maraging steel is better (but it’s also export-controlled for just this reason, as the Pakistanis learned) and CF or similar composite materials are the best.

    But in a turbomolecular pump you’re not going to be pushing it that fast, and an aluminium alloy will be fine. Indeed, that is what is typically used AFAIK. The quality of the bearings, the balancing and the stabilization of the rotor, not the tensile strength of the rotor material itself, are going to be your first limiting factors here.

    1. “I suspect this is a non-starter”
      I actually totally agree myself, but I had the idea and my curiosity won’t let me drop it until I find out how close I can get, haha. Entered into the prize because why not! Also it gives me some deadlines and forces me to actually get around to building something hopefully.

      For bearings I’m hoping to keep it real simple and use permanent magnets. I don’t need any of the bearings to also be gas seals, so that makes it easier. Just hanging a needle from a permanent magnet makes a pretty decent low friction “bearing”. Electrodynamic bearings look simple too and might be a good option. Not sure yet honestly.

      Balancing and stabilization are problems I’m just hoping “go away” still, haha. I’ll cross that bridge once I actually get something real built and spinning… Thanks for reading!

    1. Tesla’s device is not a turbomolecular pump, did not produce high vacuum, and isn’t even particularly similar to a turbomolecular pump. A turbomolecular pump works in molecular flow conditions by repeatedly striking gas molecules via a series of counterrotating blades (usually implemented as a rotor and a stator), i.e. it works via impulse transfer; the concept of viscosity that Tesla’s patent relies on is not applicable at these pressures, and the rotating discs are not going to pump anything at all. Conversely, a turbopump is completely useless at non-molecular flow conditions i.e. over about 10^-2 torr, since that would result in intermolecular collisions in the gas happening more often than the collisions between gas molecules and blades.

      1. Well said.
        In my former R&D job we had one as part of an He- detector;
        we were advised by the manufacturer that it is imperative not to move the device during and also several minutes after operation. Nevertheless, despite handling the device carefully, we needed a new bearing after some time (and these are expensive air bearings…). So good luck with that pump, they will definitely need it. I wish them the best but prepare for the worst.

        1. I used one from Alcatel and it would still be spinning 20 to 30 minutes later. The techs using it tried to see if it was working right by pushing the intake tube right up to a bottle of He and giving it a direct shot. Burned out the sensor. They didn’t realize that there is a delay between when they went over a sample and the sample registered at the other end of the 6 foot of sniffer tubing.

          What I liked was the test bottle. It had no valve to shut it off, just a tiny orifice and a pressure gauge. Even with a continuous leak, the pressure did not drop over the week or so I was there, but the detector had no problem with it until the techs burned the sensor.

          1. I am very familiar with turbo pumps and He Leak detection, I used these while working on MOCVD deposition systems. The turbo pump is probably the worst type of pump to try to build at “home” or even in a very well equipped shop. They require a very good backing pump during operation and initial startup can only happen after the vacuum chamber is rough pumped to around 100 milliTorr 1000 milli Torr is about 15 PSI. the rotors and vanes of a real turbo pump are made of a type of foil and are extra fragile any dirt or foreign objects that find their way into the intake are 100% fatal to the pump. moving a spinning pump can be done in straight lines, but rotating the pump about its axis will often cause a failure sharp sudden motions are the worst. A failure caused by moving the pump is very destructive the bearing is destroyed, but the rotor is turning 75 thousand to 130,000 RPM so the rotor is destroyed in a blink of an eye and can prove explosive leaving very little to “repair” or salvage, and in certain models the rotor is driven by an inductive field, which when suddenly finds the load mismatched current surges and the drive motor is toast as well,, can make for a real bad day! (expensive!) The “test bottle” you describe above is actually called a calibrated leak, they come in different leak rates of He, they are not cheap the rate stated on the unit is a certified and traceable standard. the life span of one of these is measured in years from the day it is activated until it is no longer at the calibrated and stated rate. it may remain detectable as a “working” or “not working” check for leak detectors for many more years. The amount of He and the pressure found on a calibrated leak will not “burn out” a detector filament under any condition as the He emitted is actually a few molecules a minute ! The detector failure was most likely caused by some other condition that occurred during the test, like sudden exposure to room air at high (15 psi ) pressure ! The detector is used connected to a vacuum system under test ands kept at high vacuum (reason it has a turbo pump) leaks are traced by puffing tiny sprays of helium from a helium tank at the system under tests fittings. the leak He detector does not use a probe wand like refrigerant detection systems (which detect gas leaking from a system to the room, where High vac He detectors monitor for He leaking into a system )

          2. I’ve seen the insides of many turbo pumps both destroyed, and pristine and I’ve never seen any foil rotor fins. They are alloyed aluminum. When they grenade on failure I’ve seen the fins embed into chamber walls over a quarter inch. They CAN be rebuilt, you just have to have precision parts. 1000 millitorr is not 15 psi either, its about .002 psi atmospheric pressure is around 760 torr. I’ve always started them around 250 millitorr, and I have never had a problem in over 10 years.

          1. No, it’s a good idea. Commercially sold turbomolecular pumps use magnetic bearing at the top for over two decades. Unfortunately, no static system of permanent magnets can stably levitate an object (Earnshaw’s theorem), so you need active control for one of the bearings, and there is considerable complexity in that when the shaft is spinning at >60krpm and a tiny oscillation will cause a catastrophic failure.

            Newer pumps, like (if I remember correctly) the Edwards nEXT series use two magnetic bearings and the appropriate control system; in the EXT series the controller is little more than a bog-standard BLDC driver but the nEXT one is much more complex…

          2. @snow You are right about Earnshaw’s theorem here. This might well be a possible solution, but the rotational speeds involved sure ought to make ironing kinks out hard…

          3. That’s a great idea! That’s exactly what I’m thinking right now. I tried a rudimentary “Electrodynamic bearing”, they’re quite simple. Just a hollow circular magnet with my aluminum axle spinning inside (this is the inverse of the diagram on Wikipedia, but same principle). It didn’t work, the problem seems to be getting the axle up to speed enough for the effect to kick in. I’m not sure how fast it needs to go. The main issue is that when it is stationary, the magnet sucks the axle into the wall making it really hard to spin. I need some way of keeping the axle centered while allowing it to spin fast enough for the magnet effect to kick in. This essentially means another bearing, which is what I was trying to avoid in the first place with the magnets.

            Instead I’m thinking even simpler, just hanging a needle from a permanent magnet is really low friction and super easy to build. Of course the pump could only be used in one orientation, with the axle hanging straight down. But sometimes the simplest solution is best to start with. The other option for the electrodynamic bearing is to use a solenoid coil instead of a permanent magnet. This way it could be turned off until the axle got partially up to speed, then gradually turned on later. In the interim, a simple bushing with some graphite might suffice. This wouldn’t require any complex control circuitry, just a little microcontroller could do the job. Thanks for reading!

          4. i suspect your attempt was unsuccessful because of material and mass (aluminum instead of copper and your axle has not enough mass looks like you need some mass to create big enough “virtual magnets”) maybe you should try to glue the magnet to the axle and put a piece of copper tubing around it. as for the start up sequence i tough your original idea would probably work :make the axle into a point and put that into a groove so the groove stabilizes at low rpm and place the copper tube in such a way that the “self-centered position” is a mm or so higher. result (hopefully) the groove stabilizes the axle at low rpm and as soon as the electromagnetic properties kick in you have the axle “liftoff” by a mm or so and no more friction at the point/groove.

      2. Allegedly, Tesla did indeed build a version that he used for creating high vacuum in his workshop (emphasis on the allegedly, I can’t find my source anymore). His patent describes his invention as useful for “the rarefaction of air”.

        His device is only a turbomolecular pump to the same degree that you consider a Tesla Turbine a “turbine” I guess. To be honest, it all gets a little bit grey. In operation, my idea is probably closer to a molecular drag pump.

        1. Tesla used Mercury powered vacuum pumps to evacuate his glass tubes. (Leading to his mistakenly attributing the “blue glow” of the tubes as an attribute of the power he was feeding into them. Was actually an attribute of ionized Mercury vapor.)

          Anyway, his use of Mercury as a vacuum source is well documented in many of his writings.

          That said, just glad when I see credit given to him anywhere as it so seldom is and he so richly deserves.

  4. I have only the most elemental understanding of the mechanics involved but it did occur to me that it might be possible to set up a system of ionizing the gas within the pump and thereby make it responsive to a spinning electric field which might it exit the chamber as if driven by the spinning disks. This would eliminate the whole massive rotor mechanism. Just a thought.

    1. You are spot on. So-called ion pumps exist; instead of a spinning electric field they use a magnetic field, and charged particles have helical trajectories in one. However, the pumping rate is extremely low, and as such it is not realistic to construct an ion pump that has an inlet and an outlet, like usual; instead, they accelerate ions fast enough to be buried in an electrode. This includes even noble gases, atoms of which are mechanically injected and held inside a metal plate acting as an electrode.

      These pumps achieve very high vacuum, down to 10^-11 torr, but they’re very slow and have a finite, fairly short life, which is why they are not really a replacement for a diffusion or turbomolecular pump.

      1. They are used for tube applications a lot to keep tubes pumped down, extending their shelf life and as a getter. But unless there is already a pretty substantial vacuum, they don’t do much on their own.

  5. I’m troubled by the assumption that an extremely low-pressure gas will still act sufficiently like a continuous medium that it is susceptible to influence by Tesla-style discs.

    At the practical end, I don’t see how cascaded discs on the same axis can work unless there are also stators managing redirection of one stage’s gas back to the centre of the next stage, possibly with flow straightening etc.

    But I’d like my worries to be proven baseless :-)

    1. I actually had the same misgivings about stators and cascaded disc stages as well recently, so I’ve abandoned that avenue for now. Allegedly that is how Tesla did it, but I haven’t found enough details on the interwebs of his attempt to be sure. Currently I’m working towards a single stage design, no stators.

      Also, I agree with your skepticism that low-pressure gas would act like a continuous medium. It won’t. If Tesla’s version actually worked for vacuum applications (who knows?), I doubt he understood it the same way I do now. See the comment by Comedicles on the project page for a nice description of the competing theories. My version operates (ha, in theory) in the molecular flow region by virtue of extremely close disc spacing, about .016mm. At rough vacuum, this is much less than the mean free path of air, so most of the collisions will be between the discs and air molecules. When inelastic collisions happen, the discs should impart some of their momentum to the air molecules, sending them a little further towards the outer edge. At least that is my rudimentary understanding of the theory as best as I can describe now.

      This is very different from how Tesla would have understood it, and has nothing to do with “boundary layers”. As Ben Krasnow noted, this design is almost more of a molecular drag pump, but I think “turbomolecular” sounds cooler so I’m sticking with that :) This is, incidentally, why I think I can get away with a single stage design. I’m not really compressing the gas so much as moving it out. As long as the air in the disc gap is in the molecular flow region, it “should” pump gas outwards. But who knows until I can build it, haha. We’ll see. I will do my best to prove our worries baseless :) Thanks for reading!

      1. Any description of Tesla tinkering with vacuum has to be treated with caution, since I suspect that what was considered to be a good vacuum in the 1910s would be considered a very poor one today… apart from the mechanics, the available instrumentation has improved enormously.

        It’s interesting to wonder whether there’s a fundamental difference between a TM pump and a Tesla turbine, based on the angle that gas molecules impinge on the moving vanes/discs. +1 for whoever suggested “drag”, which suggests that it /might/ be possible to do something with a rotating layer which introduces a small amount of adsorbtion; my thinking being that if the molecules were “grabbed” for a short time it would improve the inertia transfer enormously.

        Hard disc platters might be worth investigating as a test rig, but I think the problem here is that they’re likely to have a solid axis assembly since any perforations would increase their drag. Making a new axis which actually encouraged gas flow and making sure that this was sufficiently precise that everything remained balanced would be decidedly non-trivial.

        I find myself musing on the fact that if a high vacuum is defined as the mean free path being larger than the vessel under consideration, then space doesn’t contain a high vacuum since it’s unbounded :-)

  6. Frankly I’m not even sure why they bother with a turbopump. Diffusion pumps are downright trivial to make if you have a lathe, work on a wide range of oils, and achieve pretty good ultimate vacuum. Their main downsides are that they’re slow to ramp up and down, and they contaminate the equipment with traces of oil, especially if the oil you’re using has a high vapor pressure. For the latter I would suggest a LN2-cooled baffle; LN2 is very cheap, and if you insist on doing everything yourself, can be DIY liquefied without much trouble (eg: http://hackaday.com/2014/05/23/homemade-liquid-nitrogen/). But if you’re just tinkering around, none of this is a problem…

    1. Good thoughts! Yeah, I’m really just tinkering around. I had the idea a little while ago, and my curiosity won’t let me drop it until I figure out why it won’t work. So I’m chasing it as far down the rabbit hole as I can. If it works, then great! If not, well then I’ve learned a lot of good math and fluid dynamics and physics along the way. Thanks for reading!

  7. With hard disk platters you’ll need to have holes cut around the center for gas to flow in so it can be accelerated outward. Water jet cutting is likely to be your best bet.

    With a 15K RPM hard drive you already have bearings that can handle at least 15K RPM.

    With holes cut in the platters and even higher RPM, I’d be wanting to buy a Kevlar vest to repurpose as a shrapnel container to put around the thing, or weld together a 1/2″ steel plate explosion container.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.