[Ben Krasnow] Gasses MEMS Chips, For Science

Why in the world does helium kill iPhones and other members of the Apple ecosystem? Enquiring minds want to know, and [Ben Krasnow] has obliged with an investigation of the culprit: the MEMS oscillator. (YouTube, embedded below.)

When we first heard about this, courtesy in part via a Hackaday post on MRI-killed iPhones, we couldn’t imagine how poisoning a micro-electromechanical system (MEMS) part could kill a phone. We’d always associated MEMS with accelerometers and gyros, important sensors in the smartphone suite, but hardly essential. It turns out there’s another MEMS component in many Apple products: an SiT 1532 oscillator, a tiny replacement for quartz crystal oscillators.

[Ben] got a few from DigiKey and put them through some tests in a DIY gas chamber. He found that a partial pressure of helium as low as 2 kPa, or just 2% of atmospheric pressure, can kill the oscillator. To understand why, and because [Ben] has a scanning electron microscope, he lapped down some spare MEMS oscillators to expose their intricate innards. His SEM images are stunning but perplexing, raising questions about how such things could be made which he also addresses.

The bottom line: helium poisons MEMS oscillators in low enough concentrations that the original MRI story is plausible. As a bonus, we now understand MEMS devices a bit better, and have one more reason never to own an iPhone.

61 thoughts on “[Ben Krasnow] Gasses MEMS Chips, For Science

      1. I think what was meant was a list of common electronics which use MEMS instead of much less exotic quartz oscillators. But yes, I’d assume all MEMS have a problem with helium until further notice.

    1. If it is a MicroElectroMechanical Systems (MEMS), surely is would depend on the physical dimensions of structures inside the device, large structures, minimal effect, tiny devices large effect. But you would also have longer MTBF with then smaller structures due to physically less area for friction. So it would be swings and roundabouts either way.

        1. I mean it has at least something to do with the dimensions. It’s not like a tuning fork is susceptible to helium. That’s basically what the scaled-up version of a MEMS oscillator would be. It’s an interesting and relevant question: Where’s the line drawn? At what scale does this start to become a problem?

  1. Very nice tear-down and explanation of what is happening in these devices. The only question I have that was left unanswered is the actual chemical and/or mechanical reason why the He prevents the device from functioning properly. If friction was the cause, I would have expected the frequency to shift lower before failing, not higher.

          1. Inhaling helium makes your voice high because of the 3x higher speed of sound in helium compared to air. It’s nothing to do with friction, your meaty vocal cords aren’t slowed down by minute friction with some pathetic gas.

            I imagine MEMS oscillators run in vacuum, since they’re so tiny and accurate. It might be that, at that scale, any gas is like running in treacle, and that helium is one of the gases small enough to slip into any tiny gap, which it’s known for. But then yes, that would be friction, which Rhys’s point seems to rule out.

            I wonder what other gases would do?

          1. Reading up a little more, it does seem like the oscillator was originally vacuum sealed. Thus, any gas presence would cause issues. It looks like only helium is capable of diffusing through silicon. Hydrogen molecules, though lighter, are actually larger than helium atoms.

    1. I worked with a device once that measured the molecular density of hydrocarbon gasses by passing them thru a chamber that had a bell that was caused to ring by a feedback circuit – as the density of the gas changed, the pitch/frequency of the bell changed, [the higher the density, the lower the pitch.

      I suspect the oscillator circuit that drives the MEMS device is tuned to operate in a very narrow range [pahse looked loop?], so that the device doesn’t oscillate at a harmonic off 32KHz. when the gas density changes due to the hydrogen infusion, the mechanical resonance of the MEMS shifts too far and the oscillator circuit tries to correct the drift, looses lock and stops working.

      1. Yeah but what would the oscillator circuit have to correct the drift against? If it is it’s own single source of timing. If it’s part of a PLL wouldn’t changing the MEMS change the loop?

        I’m surprised, too, that the bell thing you mention worked. Presuming you mean like an electric bell where the moving contact cuts off it’s own electricity supply? I thought that’d be more to do with the springiness of the contact, than the speed of sound. I suppose air has to get out of the way in order for the contact to swing, but still, compared to the solid springy metal, a bit of air isn’t much of a force to reckon with, surely?

        But then you say that it worked, I can’t argue with that. Usually it’s only wind instruments, where a column of (usually) air resonates at a frequency to do with it’s length, and the speed of sound in that (usually) air; that are affected like that. A flute, say, played with helium ought to rise a lot in pitch, but not a bell, or come to think of it a guitar or xylophone or anything else that doesn’t rely on resonating waves in air.

  2. Helium diffusion through copper is one of the wearout mechanisms of heatpipes.

    New heatpipe => vacuum + small amount of refrigerant (water, ammonia, etc.).
    Old heatpipe => same as new heatpipe + small amount of He matching the partial pressure of He in the atmosphere.

    The He atoms get in the way of the refrigerant gas on its way to the condenser, increasing the thermal resistance. (This is based on things I’ve read – I haven’t actually done the experiments myself.)

    1. There’s just about no helium in the atmosphere, in nature it floats to the top of the atmosphere and away into space. The helium we have comes as a by-product of gas and oil wells, where it’s been forming and building up for millions of years. As the result of alpha radiation from radioactive minerals, since alpha particles basically are helium.

      For something as large-scale as a heatpipe I’d be surprised to see it make any sort of difference you could measure, outside perhaps of extremely well-controlled circumstances.

      1. There’s about a 5 parts per million (by volume) of helium in the air at sea level. That’s enough to be annoyingly high when doing over-pressure helium leak testing. When doing low/vacuum pressure leak testing on things like heatpipes or vacuum systems there are systems to detect into the parts per billion region. And not just because we can, but because for some systems it makes a difference. In a vacuum system, that little bit of helium can become a relatively high relative pressure to whatever other gasses are in there, since helium diffuses much more readily. So yes, when comparing to a near complete vacuum I can easily see helium making a lot of difference.

      2. Very rare is not the same as nonexistent. There are plenty of helium atoms in the room you’re in right now, probably a few in your lungs. Remember that unlikely means rare, not impossible. Unlikely still happens at exactly the rate it should and shouldn’t be discounted.

        That said I’d like to see the experimental data as well. And why is hydrogen not a factor? Too reactive maybe?

    1. Because they’re not microelectromechanical systems. They’re just quartz crystals of a specific size that deform when a voltage is applied to them and swing back when the voltage is removed. It’s why they need those capacitors on each side of them, they help the crystals to oscillate.

        1. So in other words it is a physical piece of something that moves but is for some unknown reason not micro mechanical?

          More elucidating is needed for me to draw this line because to me both of those replies imply it is MEMS whether you specify it as such or you say it isn’t. Saying it is not is fine if that is the way it is specified I suppose.

          1. Its the piezoelectric effect causing the oscillations.

            from wikipedia: A crystal oscillator is an electronic oscillator circuit that uses the mechanical resonance of a vibrating crystal of piezoelectric material to create an electrical signal with a precise frequency.

          2. Also I suppose, in a crystal the whole thing is the moving part. There only is one part, the chunk of quartz, with wires attached. MEMS have little levers and things, they’re more complex. They deserve “system” more than a simple crystal does, and even “mechanical” really. Although a crystal has physical oscillations, there’s no “parts” or mechanisms. Same way you wouldn’t really call a doorstop a machine.

            I don’t know if the actual rules for all this are written down anywhere.

          3. It seems to me that it is either ill-defined or not defined or the definition is hidden away in those huge documents that standards definitions get put in. If one moving part defines it as not being then that means a speaker is not an electro-mechanical system and I can’t see that being true.

          4. It’s not a legal term, it’s just ordinary English, if perhaps a bit specialist. Most words and phrases don’t have direct, exclusive, fully encompassing definitions. I’m giving my impression but I don’t actually set the rules, and neither does anybody.

            FWIW, while we’re having a pointless discussion, a speaker has several parts, several of which (coil, cone, spring, etc) move. The frame and magnet don’t (in a moving-coil speaker, that is). A quartz oscillator crystal is literally a solid crystal. It has wires attached like most components, and a case, to keep the dirt off. But it’s not really one “moving” part, it’s just one entire part. The same as a cap or resistor.

      1. It’s some kind of pedantry. Sure, on an atomic level almost everything is microelectromechanical. The distinction is still useful to us, though. There are perfectly good reasons why, say, chemistry is not lumped in with mechanical engineering or physics or whatever.

    2. I’ts probably a matter of scale (quartz oscillators are not that small) and manufacturing methods (MEMS devices are fabricated or built up where-as a quartz crystal is actually cut down.)

  3. So we now know that Mister Spock would never use an iPhone, because he has to beam down to foreign planets regularly, and their atmospheres can contain a much higher amount of helium, than earth’s does.

    AFAIK, the only reason why you would use a MEMS oscillator instead of a quartz is price.
    It seems that Apple is even cheaper than we thought.

    1. The issue isn’t MEMS vs bulk quartz resonators, it’s hermetic seal vs non-hermetic seal. That SiTime part appears to be “in” a chipscale package without any encapsulation.

      1. If you watch the video, the leakage mechanism has nothing to do with packaging. The MEMS resonator is fully encased in silicon, hermetically sealing the resonator. It doesn’t GET any more hermetic than that. The helium diffuses through the silicon crystal lattice itself.

        1. Hermetic means “gas tight” although I’m not sure that’s possible with helium. (I would appreciate being corrected here.)

          BTW, I had watched the video, and that Si that you mentioned is part of the MEMS chip. Most people would not call a layer of a bare die “packaging”.

          1. Yes, light gasses like helium and hydrogen leak through hermetic seals due to atomic diffusion. A rubber balloon is quite gas-tight, but as we all know the helium eventually gets out anyway.

            It’s a big pain in the ass for space propulsion. The most ideal propellants slowly escape from any container over time.

    2. I don’t know the specific parts in question well, but: the SiT oscillators we’ve used cost about 10x what a quartz crystal would cost, and almost 20x what a ceramic resonator would cost. The reason we use them is that their oscillation frequency is programmable: we can use the same part number across multiple systems and just instruct it to be a 16mhz or 20mhz or 24mhz, so for us it’s a material handling issue. But we’re way more sensitive to inventory management cost than purchase price, and I don’t think that’s the case for Apple. I wonder if they’re doing something with EMI that means they have to program very precise but slightly different frequencies into otherwise nearly identical units. That seems like it would be ridiculously expensive to test, but if there’s an EMI spec you have to pass it might be worthwhile.

    3. I don’t like Apple at all, and have not purchased any of their products, but I really dislike this sort of uninformed bashing and assumption that they’re cheap because they use this component that behaves strangely.

      You have clearly not watched the linked video, which explains this. I kind of wish HaD actually mentioned the very clear reasons why you would use a MEMS oscillator in a high end portable device.

      The SiT oscillators are in fact generally more expensive, which is why most people have not looked at them. 32kHz crystal oscillators are simple, commodity items, and they are probably easily 10x cheaper than MEMS oscillators as another poster mentioned, even in extreme volumes.

      If you are designing very small portable devices (i.e. Apple watch) or care about standby battery life (fitness trackers, etc.), then you may have run into this sort of issue. When you are looking at shaving microwatts, standard crystal oscillator circuits take a lot of power during standby. These MEMS oscillators use quite a bit less. Also, when you care about space, the SiT is maybe 1/2 the size of the absolute smallest, specially packaged SMD 32kHz crystal you can buy.

      Either way, I’d agree with Ben – this is a pretty extreme failure mode. Helium is not a cheap gas (not to mention, a nonrenewable resource that pretty much can only be obtained practically as a by-product of natural gas production). The event that vented enough helium in the MRI machine likely vented easily thousands of dollars if not $10k+ of helium to raise the nearby helium levels to be high enough to damage MEMS oscillators.

      The other thing to keep in mind is that most other MEMS devices in other phones likely had similar accuracy problems for some time after the event.

      1. I have a hard time believing that helium is a non renewable element, surely new atoms are produced in the core of our planet by the nuclear reactions daily as alpha particles, and some of that must eventually make it to the surface through the molten layers. At the very least there should be slightly higher concentrations where volcanoes vent into the atmosphere. OK it will not be as cheap to recover as the 0.3% to 8% currently found in natural gas, which will eventually run out.

      2. People accuse it of being cheap because they presume, without cause, that everything not only has a reason but a good reason.

        My own estimation is that everything has a reason, and that those reasons are of average quality, so most things have mediocre reasons.

        Here, probably they’re just squeezing size and power everywhere they can reach, and there isn’t something as clear as “cheaper” that says it was a good tradeoff. They’re squeezing everything and if it didn’t cause a problem, alternatives might not really have been looked at that closely.

        It seems perfectly reasonable to dislike that process, or to like it. It seems more reasonable to be for or against Apple’s practices than to be for or against people being for or against it.

        I saw some people on the internet arguing about the cost of the helium, one person was waving their hands and saying $10k, another who also wasn’t there was throwing around numbers saying maybe only $2500. That’s a lot for a gag at the coffee shop, but it might still get used in various situations.

  4. Some time ago I was browsing through some datasheets for polymer capacitors.
    I found a bunch of warnings about susceptability to methyl bromide (Bromomethane).

    Methyl bromide seems to be used on a large scale (relatively, you don’t need much) to fumigate shipping containers.

    1. Depends how much, but the Mickey Mouse voice is 100%. I suppose you might hear the difference at 20%. More an issue of “noticable”, really, than the physics. Any particular gas mixture produces a different pitch in a wind instrument like a human throat.

      If you’re thinking of filling a room with it, nah, there’s better ways of spending thousands of quid for childish pranks.

  5. In case it’s relevant to the discussion: Diffusion of gasses through materials is not always related solely to atomic mass of the gas. I messed about with some high altitude balloons for a while and although foil balloons are much better for retaining helium, hydrogen actually diffuses through the foil faster than through normal latex balloons so you use rubber for hydrogen and foil for helium.
    That could mean that the packaging for these devices is more permeable to helium than to hydrogen, even though hydrogen is lighter.

  6. Very cool video. I was curious how the gyro would cause the iphone to die and then recover, interesting to see that there is another mems device that cause the shut down, not some funky software issue that shuts you down after getting a bad input from the gyro. Thank you.

  7. My guess about the different time scales is this. At 2% Helium after 30 minutes the partial pressure in the sensor will be much lower then outside the sensor. So the diffusion from the inside to the outside will be much slower then the other direction.

  8. Diffusion is a random process. On the way in there is a much larger number of atoms on the outside available to enter the package, and only a few need to be there to interfere with it. The odds of them then leaving is much lower because the number of them is much lower. I expect the rate of entry relative to the level required to disrupt it is such that by the time the effect is noticed the infusion is far higher than would cause it. It is probably a two stage process; in the first part the case is filled with helium. In the second part the helium diffuses into the mechanical parts, changing their masses. So not only do the parts need to see the helium exit the free volume, that volume needs to be free in order for a net migration out of the mechanical parts to take place.

    One big difference is between the mentioned gases is that hydrogen is diatomic while helium is monatomic, so that the hydrogen gas molecule is larger in one dimension than helium atoms are. This likely means that the hydrogen cannot permeate the silicon and therefore either not enter in the first place and not affect the inertia/density of the tiny structure. Individual hydrogen atoms are also slightly larger than individual helium atoms, but the difference is much smaller than the difference in size from being a molecule. I expect the heating is to cause the crystal structure to expand enough to let the hydrogen out.

    I am surprised that the cap is not placed and diffusion bonded instead of growing it.

  9. Thanks for sharing the very pretty blog post for roulette game. Your thoughts for playing roulette game online is very pretty. it’s always compulsory to set your limit at the time of playing game. As you know this era is famous for android operating system, Everyone are using android mobile phone in this fast forwarding life. Most of people are like to play the game in their phone. They download a application to playing game online according to their interest. Most of rich people like to play roulette game. That is the main region to develop a roulette game for the android and IOS platform. Because these are most usable platform that’s people use to playing the game. Thanks you for nice sharing.

Leave a Reply to Allan HCancel reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

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