Nascent Project: Open Source Scanning Electron Microscope

I used to have access to some pretty nice Scanning Electron Microscopes (a SEM) at my day job. While they are a bit more complex than a 3D printer, they are awfully handy when you need them. [Adam Guilmet] acquired a scrapped unit and started trying to figure out how to breathe life into it. His realization was that a SEM isn’t all that complicated by today’s standards. So he has set out to take what he has learned and build one from scrap.

In all fairness, he has a long way to go and is looking for help. He currently says, “[T]his is being powered by fairy dust, unicorn farts, and a budget that would make the poorest of students look like Donald Trump.” Still, he’s collected a lot of interesting data and we hope he can build a team that can succeed.

One problem is that generating, focusing, and detecting electrons is probably the easy part. The hard parts are drawing a very heavy vacuum, the need to do things like gold-coat insulating samples (we used to put ours in an argon chamber and sputter gold on the sample; a gold-coated die in a plastic keychain always made a good gift). You also need high voltages, and to be careful of energetic electrons striking things and producing X-rays that aren’t good for you.

Then there are the other things you’d like to have once you have a SEM. EDS (Energy-dispersive X-ray spectroscopy) and Auger spectrometry come to mind. Another interesting use is to collect secondary electrons emitted from a CPU running a loop. If you collect the emissions at the same point in the loop long enough, you can determine if that area of the device has a positive or negative charge (that is, a one or a zero). Scan the beam in the XY axis and you can build a picture of the electrical state of the device. Repeat that at different points in the loop and you can build a movie and watch bits travel across the CPU circuitry. Sounds like a fantasy, but it has been done–but not in someone’s garage. At least not yet.

We’ve seen SEMs refurbished by hackers before. And there’s at least one homebrew SEM out there — of course we mean [Ben Krasnow]’s — the video of which is embedded below.

17 thoughts on “Nascent Project: Open Source Scanning Electron Microscope

  1. Im the guy with the hitachi in the first link.

    A diy SEM can be done (obviously since Ben has), thermionic emission is easy. High vacuum is pretty easy as well, old diff pumps and valving are on ebay.

    There are two books I would recommend looking through, both by the same guy, Nagamitsu Yoshimura of JEOL. “Historical Evolution Toward Achieving Ultrahigh Vacuum in JEOL Electron Microscopes” and “Vacuum Technology, Practice for Scientific Instruments” You should be able to find them in electronic form.

    You dont need to sputter most things. Some things will charge a bit but it can be worked with. The vast majority of the stuff I do and the stuff my college prof friend that owns a bunch of scopes is done uncoated.

    EDX (EDS) is probably not something you are going to find built from scratch due to things like the Si(Li) detector and the cryogenic temperatures it needs to run at. Also the toxic beryllium x-ray window adds more fun to the mix.

    1. We always gold coated passivated parts but maybe that was not to further damage the device or had some other function but I was taught to do it and taught others to do so (keep in mind the SEM was just one tool we used).

      As for EDS I agree it is a long shot but my point is that imaging is just one part of the value of a SEM.

      1. You coat non conductive parts otherwise they can build up a negative charge and deflect the beam since like charges repel. You can get some really interesting results from this, there is a picture where they have a nonconductive sphere in the chamber and it charges up and scatters the beam around the inside of the chamber and a distorted image of the inside of the chamber is “reflected” in the sphere.

        You might be able to put together a simple edx using a proportional radiation counter and a multichannel analyzer NIM module.

      2. ESEM is a term for low-pressure SEMs which use differential-pumping (high-vac up top for most of beam-path, low-vac immediately at sample) and the sample-chamber gas carries away excess charge. You can also do charge-matching (might not be the standard term), where you basically use very low voltage (or step it up progressively over scans), or scan slow enough to not over-charge any area before it bleeds down.

        I’ve been working out an x-ray detector based on a PIN photodiode, there is a good amount of info on these for x-ray spectroscopy… not as much resolution/speed… but it can get the job done. With that, I think cooling the amplifier would gain a decent amount of utility.

  2. Very cool project. If he can find a usable and cheap (may be mutually exclusive) turbo-molecular pump, it will save years of his life spent waiting for things to pump down (hey, its possible). So many of them are used in the semiconductor business that they have to be out there. Smaller ones from equipment for old processes perhaps? I can’t guess. Modern RTP uses such low volume clam shell vacuum chambers that they must pump down very quickly anyway.

    1. Only at startup will it be slow, once it is running a DP os the same physical size will pump down a chamber faster than a turbo. And for someone new to vacuum I would recommend a DP. Plus, there is one already on the guy’s column. Usually once the main valve opens on either pump the chamber is down to high vac pretty darn quickly.

      1. Interesting. I have never seen a DP as big as a turbo pump. They have always been maybe 2 or 3 inches diameter on old scopes and research gear without a lot of volume. Everything bigger had turbo pumps once people accepted that using vanes to swat molecules wasn’t crazy. Plus the oil handling (or mercury) and all that was a pain. Turbo’s were generally way too expensive for DIY, but with workable surplus units now?

        1. I had an ISI mini SEM that had a DP with about a 1″ bore. Tiny little guy. Some helium leak detectors used some really small pumps too.

          DPs have several advantages. You cant kill them, worst case is using hydrocarbon working fluid and exposing it to atmosphere at operating temp, the oil will turn into a carbon mess. Other silicone or fluorocarbon based fluids will not break down, it just gets everywhere which really sucks because it is pretty darn expensive. DPs are also have zero vibration, so if you have something that needs to be vibration free a DP or an Ion pump is the way to go. Lastly, DPs are faster for a given size. I have a Varian DP that came off one of my coaters, it is rated for 2400l/s with a 6″ throat. The turbo that replaced it has a 10″ throat and only 2000l/s capacity.

          For really, really big vacuum systems they still use diff pumps. the size of a turbo would be insane. These diff pumps have something like 36″ throats.

          Turbos on the used market are actually pretty reasonable. I picked up a couple Leybold TMP50 turbos off ebay for $100 each. Paid a couple hundred more for one control and some horse trading for a control for the seconds pump. I bough a larger Leybold turbo for work, around 700l/s, I think we paid $600 for it and it has a built in turbo drive. One way to get a cheap turbo is get an old mass spectrometer. They usually have turbos in them of various sizes.

          Turbos have a lot of advantages, they are clean, low maintenance (especially mag lev), but they are incredibly delicate. They can even just randomly grenade themselves. I knew a guy who worked at intel as a maintenance and they had a large turbo on a tool just suddenly shred itself. The resulting torque of the rotor going from 24krpm+ to a standstill ripped the turbo right off the tool snapping every bolt holding it in place. The manual for my big 2000l/s pump mentions the torque that is generated during a crash, I think it was somewhere in the 5000 ft/lb area. Turbos are nice but I would not recommend one for someone just getting into vacuum.

  3. I once took a job at a Fujitsu Japan fab making 4M DRAMS and I think 1M EPROMS, obviously a long time ago. I had fun doing the job for a few weeks; the pressure increase output forced life skills in time management on me from that time on, but it was crazy that asking questions outside my exact job description was basically forbidden, I quit when I got to the point of no more interesting stuff that I was allowed to learn or ask about.
    But, and maybe I have forgotten, I do not remember needing to gold coat the sample wafers we were scanning with the SEM. Is it because the deposited metal, gates, and Si already reflect well?

    1. Most likely you were looking at them before passivation. We were taking apart so once you took off the passivation with HF you didn’t gold coat anymore. The etch we used to take off the gold was some nasty cyanide compound I was always afraid of

  4. Awesome project! I would love to see more about the project, including the customized auxiliary systems.

    I would also love to see an itemized breakout of some kind. I think the open source world will flourish if we can start interconnecting different design patterns with similarly functional end-products, and build searchable repositories of options. An SEM isn’t too advanced by today’s tech standards, but is still very advanced for the open source world and could be a nice project to start consolidating an open source repository of tool designs like everhart-thornley detectors, oscilloscopes, etc.

Leave a 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.