It seems that MEMS is very sensitive to helium, but only helium. This Link stated that hydrogen does not affect MEMS, which surprised me. Does anybody know why it's just helium, or what's really going on?

According to the link, an iPhone 8 will stop working after just a few minutes to hours of helium exposure, and will take days to recover, even using a vacuum.

According to this table from the linked article, a hydrogen molecule (120 pm radius) is actually smaller than a helium atom (140 pm radius), which is confusing as to why it is not more effusive. Is this correct?

enter image description here


An important "find" buried in the article was one solder bonding method that did keep out the Helium:

While the majority of the CSP (Chips Scale Packages) parts did show ingress of helium at varying rates, one type of metal solder seal did not show any indication of helium ingress. Unlike prior work, this study found a wafer-to-wafer bonding method that enabled diced, vacuum-bonded chip-scaled packages to pass 1000 hour, high-pressure helium testing with no change in sensor output due to helium ingress...

  • 1
    \$\begingroup\$ Your link says "These differences in Q impact suggests that classical effusion through a leakage path is not at work in hydrogen, or it would also diffuse into the MEMS cavities like helium. It is possible that at room temperature the hydrogen is reacting with adsorbed water vapor or hydroxyl molecules on the silicon, glass or oxide surfaces of these bonding interface imperfections and seams, while helium does not react with this surface, and that the effusion path through these surface or seal imperfections is too small for argon, nitrogen and other larger molecules." \$\endgroup\$
    – DKNguyen
    Commented Jun 9, 2023 at 5:56
  • \$\begingroup\$ @DKNguyen -- For them it suggests, but their explanation doesn't hold water for me (it's a leaky abstraction), because there should be some effusion, but there's none at all. Maybe there is yet another mechanism we have not yet discovered or recognized, that is only happening with Helium. Can Helium penetrate a crystal lattice? Or make it's own "pinholes"? Or sneak by using grain boundaries, where the crystals run into each other? I was hoping someone knew more from a different source. \$\endgroup\$ Commented Jun 9, 2023 at 6:31
  • 1
    \$\begingroup\$ @MicroservicesOnDDD gases penetrating crystal lattices is a pretty normal thing. It happens at a higher rate at higher temperature, but with helium atoms being very small and not easily captured by any polarity in the upper layer of a lattice, that's not surprising. As a matter of fact, some semiconductor production pathways include diffusion Doping, which is a fancy word for having the desired dopant in Gas phase next to the crystal lattice until you get embedded atoms. But there's more household-observable cases, as well: rock salt is a regular lattice, and it can absorb and store large... \$\endgroup\$ Commented Jun 9, 2023 at 8:25
  • 1
    \$\begingroup\$ ... amounts of water from atmospheric water vapor \$\endgroup\$ Commented Jun 9, 2023 at 8:31
  • 3
    \$\begingroup\$ The oblong orbital of a hydrogen molecule is much less "closed" than the spherical helium orbital and will interact with charge carriers on the silicon surface, essentially engaging in a loose surface alloy and consequently not getting deep. In contrast, helium's electrons are unavailable for any kind of loose or tight interaction. If the atoms fit through the crystalline grid, they won't get their electrons or a molecule binding tangled up in the process. \$\endgroup\$
    – user107063
    Commented Jun 9, 2023 at 12:18

2 Answers 2


It's a quartz oscillator replacement with 32,768Hz frequency (namely the realtime clock that does timekeeping even when the phone is off) that is affected here. There is a tiny silicon element vibrating in a sealed vacuum. Helium is an atomic gas like all noble gases, but with just a single, complete electron shell. The effective radius of its isolated atoms in a gas is much smaller than that of the two-atomic hydrogen molecules that share an oblong electron orbital. A hydrogen molecule essentially stores heat energy in three degrees of movement freedom and two degrees of rotational freedom (at higher energies, oscillation of the bond length also plays a role). Helium just stores heat energy in three degrees of movement, so at equal temperatures, the average collision energy is larger.

The atoms are small, self-contained, and don't interact. Silicon atoms are much larger even when arranged into a crystal. The surfaces are specially cleaned to allow doping with gas epitaxy where doping materials are suffused into the surface layers of a strained crystalline surface. Squeezing in a few helium atoms is like pouring fine sand on tightly packed peas: it will just sink in without much of a problem.

Now the problem is not one of electrical interaction (helium does not interact) but of the added weight impacting the MEMS element's oscillation, and the oscillator getting detuned enough to stop resonating with the electrical circuit driving it.

It should not get affected by air and actually oscillates in a sealed vacuum cavity, but when the bulk of the containment is stretched silicon, sealing off helium is a fool's errand. Hydrogen's oblong molecules have much more of a problem getting into deeper layers, particularly since its one-for-two orbital is a lot more interested in interacting with other electrons (which is why hydrogen is highly flammable and helium not at all).

  • 3
    \$\begingroup\$ So the helium also diffuses out of the "vacuum" cavity when it gets a chance, presumably to eventually match the partial pressure of He in the atmosphere. Crazy. \$\endgroup\$ Commented Jun 9, 2023 at 15:59
  • 1
    \$\begingroup\$ This is the first time I've ever heard of a 32.768 kHz watch crystal referred to as "MEMS." Not complaining about it or anything. It's just... interesting. Excellent answer though. +1. \$\endgroup\$ Commented Jun 10, 2023 at 2:19
  • 4
    \$\begingroup\$ @SolomonSlow it seems to be a crystal replacement IC with a MEMS resonator, not an actual crystal. The linked article says it's a SiT1532. \$\endgroup\$
    – kwc
    Commented Jun 10, 2023 at 2:30
  • 5
    \$\begingroup\$ @SolomonSlow It's not a crystal, it is a MEMS replacement of one with the oscillating element not being quartz but etched silicon. More compact and probably cheaper. I doubt that a quartz would be affected similarly by helium poisoning. My guess would be that the oscillator would not stop outright but it's possible that its accuracy may be affected as well. \$\endgroup\$
    – user107063
    Commented Jun 10, 2023 at 9:10

Helium is an atomic gas. Hydrogen, or rather di-Hydrogen, is a molecular gas, and so much larger, and slower to diffuse, through seals or materials.

It's for this reason that helium is used for testing the sealing, the hermicity, of packages, in a Helium Bomb. It's so much more effective at getting through small spaces than any other gas.

  • 1
    \$\begingroup\$ Please see my edited question. The linked article seems to say that a Hydrogen molecule is smaller than a Helium atom. \$\endgroup\$ Commented Jun 9, 2023 at 12:09
  • 1
    \$\begingroup\$ @MicroservicesOnDDD I don't know how they define "diameter" here but hydrogen is reactive. It's sort of like slow-pouring marbles vs glue-covered peas on a stack of coarse grids. The peas may be smaller but will not get as deep as the marbles because they are much more likely to stack up. \$\endgroup\$
    – user107063
    Commented Jun 9, 2023 at 12:39

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.