Sunday, April 3, 2011

DIY Scanning Electron Microscope - Image Quality Improvements 1

This weekend, I spent some time making incremental improvements to the image quality of my scanning electron microscope. I decided to abandon my own X/Y deflection amplifier in favor of the amplifier from a commercial analog oscilloscope. The oscilloscope's amplifier is extremely linear, easy to control with the front panel knobs, very fast, and has a "good enough" amount of deflection and offset, but it could be better. I will try to compensate for this by adjusting the deflection plate length and separation distance.

Here is a shot of the aluminum window screen that I have been using as a test target. The wires in the target are pretty straight, indicating that the system is relatively linear, and can produce an undistorted image.

This photo shows an integrated circuit -- this one is a MEMs gyroscope. The wild black/white pattern on the hard-edged object at lower right is the actual silicon die. I imagine the strange pattern is caused by the varying conductivity of the die's surface. The die's bond wires can be seen clearly as well as the metal pads that connect the bond wires to the chip's external leads.

I spent a good part of the day redesigning my filament power supply. It is now DC and fully regulated. The original design was AC, which caused huge focus problems (the beam would fluctuate at 60Hz, so I rectified it to unregulated DC. Later, I found that the ripple in the output voltage was also causing image problems. Since my image scan rate is on the order of 30Hz (in vertical), I would see the problems caused by 60Hz noise as rolling bars in the output image. The ripple in the DC filament voltage caused strange black bands to roll down the video image. I suspect this happened because of slight changes in the electron gun's bias. I am really surprised that the 2V filament voltage would have any effect on the bias voltage of hundreds of volts, but it apparently does.

The next problem I found was apparent at higher magnifications. This photo shows a close-up of one of the chip's bond wire pads. The pad in reality is very straight, but it is quite curvy in this image. At first, I thought this problem was caused by a ground loop between the X/Y amplifier and the rest of the microscope's power supplies. No, it turned out to be caused by the oscillating magnetic field created by the diffusion pump's cooling fan and the pump's heater.

I switched off the pump and fan, and the problem went away. I thought about adding magnetic shielding to the chamber, but I think I will just accept with the wavy image while focusing/panning, then turn off the pump to record an image.

My SEM design is ideal for teaching and exploring electron beam control, but has several major design flaws, which I will summarize for those interested in creating homebrew SEMs with actual utility value:

1. The photomultiplier tube is extremely sensitive to ambient light. In my SEM, I use a black plastic light-tight box to cover the glass bell jar and keep it protected from room lighting. However, the SEM's own filament produces enough light that escapes through the electron gun's vent holes to cause big problems. I can only run the photomultiplier tube at about 700V, before exceeding the average maximum rated anode current (.1mA). The SEM imaging system is AC coupled so the offset is not a deal-breaker, but I believe the DC offset also creates a lot of AC noise in the output image, and would also damage the photomultiplier if I should run it for a long time or greatly exceed the max average anode current. The solution is to build the microscope column from an opaque material (metal), and arrange the electron gun such that there is no optical path from its vent holes to the specimen chamber of the SEM (like all commercial designs).

2. Remove magnetic disturbances or build the SEM column from a magnetically-shielding material.

3. Use very clean, tightly regulated DC for all supplies involved with the project. This is a sensitive analog system, and it picks up every kind of noise.

4. The main accelerating voltage need not be very high (eg 3KV works fine). Lower-energy electrons are easier to deflect and focus, lowering the requirements for those respective power supplies. So far, I haven't seen any benefit to using higher accelerating voltages.

5. Minimize the use of all insulators inside the chamber. Try to build structures that stand on their own without insulator support, or shadow insulators from the electron beam with conductive surfaces. I will talk about this more in a future video, but it caused me many problems in the early weeks of the electron column design.

My next actions will target the overall softness of the image. A youtube commenter pointed out that my aperture disc is much too thick (.010" thick, with a 100um hole). In the pinhole photography world, this would be considered a silly setup and would lead to soft images. I am inclined to believe a similar thing is happening with the electron beam, and will try the same aperture size with a thinner sheet.

I will also try to reduce the amount of light coming out of the electron gun and/or reduce the amount of light entering the photomultiplier. This should reduce the grainy high-frequency noise in the image.

PS, I also think that all of my images are inverted with respect to the normal secondary-electron images from SEMs. All of the images that I have posted so far have dark regions where secondary emission was high. I will probably correct this by adding another inverting opamp the signal chain.


  1. Thanks for the update.
    Is there any reason why an electron gun assembly from an old oscilloscope or TV could not be salvaged for this kind of project?

  2. Anonymous, the cathodes in CRT guns are almost always coated with a metal oxide to increase electron yield. The problem is that the oxide must never be exposed to air even when cold or it will be damaged. This makes it nearly impossible to repurpose CRT electron guns. Another problem is that the emitting face of CRT cathodes is relatively large (eg 2mm in diameter), but SEMs work best with tiny electron sources -- the smaller the better, so just a tungsten wire is ideal for simple thermionic sources.

  3. Interesting about the oxide. Where can I find out more about this problem. I could not find out anything from google.

    I will be constructing a vacuum chamber soon and one of the things I wanted to experiment with was the electron gun from a CRT

  4. Anonymous, , also search for "cathode poisoning" and "cathode oxygen" Let everyone know how your vacuum project proceeds. Good luck -Ben

  5. Great stuff Ben! Please keep going - very interested in building one of these myself!

  6. Ben, mega cool. I can't wait to see your next update.

    Sacha Deangeli is another DIYer who is working on a scanning tunneling microscope kit. I think he's blogged about it at

    It's exciting to think that both of your projects might enable sub-$1000 nanoscopic imaging systems.

  7. In the second image, I think the insulating regions of the die are showing signs of charging - this creates scan-direction streaking as the secondary electron signal does not decay sufficiently rapidly. You'll need to build yourself a sample coater one day - the good news is that you already have the vacuum system :)

  8. Hello Ben!
    You could make a convenient diaphragm with a very thin gold leaf (available on ebay for decoration), glued with silver/epoxy conductive glue on a supporting small disk with a medium sized hole (0.5 mm or so). The gold leaf is pierced with a electro-etched tungsten tip that you can make this way :
    For this application, the process can be simplified : the etching KOH solution is just a drop inside a stainless steel ring, traversed by the tungsten wire. When the etching is complete, your tip fall in a rinsing water solution.

    The piercing itself is done by electro-erosion between the gold leaf and the tungsten tip(few volts inside a droplet of distillated water).

    Holes are very clean, and, with the help of an optical microscope, you could control holes up to 10 microns.

  9. Restorative to my sentiment on humanity to see people doing projects advanced as this. Way to go!

  10. sorry for double post, and double sorry if this is a stupid question, but regarding Ben's comment about not exposing CRT's to air, what if you built a chamber of another cheap gas with no O2 inside, such as CO2, N2, or He ? If it worked, the drawback would be anything you viewed would have to be "gassed" in a sealed box to flush the air out, but how much $ would it save to use a CRT ? Thanks for any replies.

  11. Anon, electron guns in CRTs are designed to emit lots of electrons from a relatively large area (eg 2mm dia). In contrast, SEM electron guns attempt to make the electron source as small as possible, ideally just a point source. Tungsten filaments are OK, but LaB6 emitters are better, and most high-quality modern SEMs use field emission, which is even better to generate the electron source.

    Many commercial SEMs use dry nitrogen to vent the specimen chamber. This lowers the amount of atmospheric water that gets into the chamber and improves the pump-down speed. It probably would not be good enough to prevent oxygen poisoning of metal-oxide cathodes though. If you are interested in building a SEM, I would definitely recommend plain tungsten cathodes. They are durable, cheap and easier to use than anything else.

  12. bricodada, thanks for the info!

  13. Hey- excellent work here! I don't know if you would feel that it's going against the spirit of this project, but I happen to literally have a storage unit full of old SEM parts and such that I not only don't need, but want to get rid of.

    You might not want most of it, but I do have some scan generators, etc. Most are parts from JSM-35 systems, which are nice in that almost everything is in small self-contained units.

    One thing I have which is small that you might want to use is a small molybdenum aperture. I have several that are of no use to me (They came from an instrument I no longer have) and I'd be happy to donate them to the cause. These are typically used for column apertures and final apertures in a commercial SEM, but are also used for other purposes, so they are not strictly SEM parts...


  14. kraftphysics, I'd be happy to pay for your time and shipping costs if you would like to send me some old SEM parts. Please email me at ben at magconcept dot com.

  15. Hello ... Congratulations ! I've good idea of the fantastic work you realized because I'm just finishing my homemade Time Of Flight Mass Spectrometer and I've acquiered experience (painfully !) in high vacuum, electromultiplier ... May be SEM will be my next project ... One more time : BRAVO !

  16. Hi, Have you ever tried to used electromagnetic lenses in your SEM ?

  17. Ravaner, no I have tried any magnetic lenses. I would have to find or make the pole pieces, which is pretty difficult either way. Commercial SEMs use magnetic lenses because they can be made to have a shorter focal length than electrostatic lenses, and apparently have lower optical aberrations. However building magnetic lenses from raw materials seems much more difficult. Electrostatic lenses are very easy to build, test and rebuild.

  18. Thanks for keep us updated. Your information is really informative and helpful.

  19. Hello! This is a very cool project! I've taken the liberty of referencing it in a small website I run for open/DIY projects that could make neuroscience more affordable and accessible. Let me know if the way it is referenced is ok for you!

    Best regards,
    André M Chagas

  20. Andre, sure that looks great! Thanks!

  21. Hi Ben.
    I have one question how you supply phosphor via 10kV supply to accelerate electrons? There is behind this phosphor any conductive glass and then connected to 10kV?

    Best regards
    Filip Orłowski

    1. Filip, the 10KV is connected to a ring that surrounds the phosphor disc. This creates an electric field that causes electrons to be accelerated toward the disc -- some electrons may hit the ring, but many will hit the disc because the acceleration created by the ring directs them mostly toward the disc. I believe the phosphor is mixed with a slightly conductive material to help spread the charge.