My new dual-beam SEM - FEI Versa 3D

As some of you have probably already heard, some time ago I got myself a FEI Versa 3D - dual-beam Schottky field-emission SEM, with EDS/EBSD detectors and platinum deposition.

This has been my dream for quite some time, and it is surreal that it is finally here.

Schottky field-emission SEM is a good middle ground (between cold field-emission and tungsten filament) - quite robust and stable, yet will require quite significant expenses for electron source maintenance. At high electron energies, Versa 3D reaches 1nm resolution, which is close to the SEM practical limit, only at low acceleration voltages it falls behind SEMs with immersion final lens (which has its own disadvantages). It has a beam deceleration / stage bias option, which gives quite a lot of flexibility. The only missing feature is a low-vacuum option, but there are workarounds.

Let’s have a look:



When I was looking for a SEM, I realized I must find one in working order, as I won’t have enough time and dedication to restore one from storage. SEM was purchased from a university in Austria, where it was used for a decade for studying metal alloys. I went to the university in person for SEM disassembly and to make HDD backups before the road trip. Austria matches expectations exactly:


After some challenges on the Swiss-Austria border, it was time to move in. Unfortunately, the elevator in my office was ~90mm too narrow, so I had to organize the move-in through a window. For that, - central street in Zug had to be partially blocked for about 2 hours, with police and safety inspectors on site. It was expected that the window height was tight even with partial SEM disassembly. At the end, it fit with about 10mm of space left. I was mentally prepared to cut the window frame if needed; it was cheaper than postponing the operation. This whole process was quite stressful, but went relatively flawlessly.


I got to assemble and bake-out the microscope myself with remote guidance from the service engineer, which also went flawlessly. SEM runs on relatively modern computers - both on Windows 7. I did some upgrades on the computers: HDD’s were replaced to Samsung 850 PRO SSD’s (which are among the most reliable SATA ones), CPU’s were upgraded to final Haswell’s, memory increased to 32/48Gb. SEM software is not hardware locked, license for EDS/EBSD software is hardware locked, but was fortunately reissued by EDAX in my name (and new hardware). To make installation tidier, I used 10m optical DisplayPort extenders and 5m FireWire extender which allowed me to move the computers to the main hardware rack.



SEM is pumped by nEXT240 turbo pump, which I consider a lucky coincidence, as this is the one that I can fully service myself. I’ll add a temperature sensor on a bearing just in case.


On the vacuum side, I was never a fan of plastic fore vacuum hoses, so they both were replaced with hydro-formed metal ones. One concern is the transmission of vibration / acoustic noise, but as this is a field-emission SEM, it has a fore vacuum buffer so that 99% of the time the low-vacuum pump is not running. After replacement of the hoses, - buffer cycle length reached ~3 hours (i.e. turbo pump is running for 3 hours with no backing pump).


One issue I’ve noticed was very suspicions high-vacuum gauge readings. They were not moving regardless of what I did. It’s a miracle bake-out worked with no issues. I discovered that in early 2024 during service, - vacuum gauge was replaced to incorrect type: PKR251 instead of IKR251. PKR gauge is for ESEM’s, and IKR is for high-vacuum only SEM’s (like this one). They have different conversion curves. Then, in late 2024, - gauge apparently failed and started displaying even more unrealistic readings (could be platinum contamination?). By this time I’ve already ordered and received a replacement PKR251 gauge, which worked but showed incorrect pressure reading (as expected). At the end, I patched the pressure conversion formula in binary and got the correct reading. I’ve also got PKR360 gauge with a much weaker stray magnetic field.


As its dual-beam SEM, all of its hardware does not fit in the base, and there is a separate rack with HV power supplies, which look quite intimidating:


Originally, the SEM was powered by an 8kW UPS which had 20 lead-acid batteries, and weight >150kg. Batteries were, of-course hopelessly dead. UPS was also loud. I replaced it with a much smaller 2.2kW unit, which is enough to keep the microscope at idle for ~1 hour (field emission source should never be switched off, especially abruptly). I will need to figure out automatic shutdown from the UPS signal in the future.


Now let's see how it works:

Sample is mounted on a 5-axis stage:


This is the business end of the microscope: you can see the main electron column, the final conical lens, and on the left, - ion column final lens. On the right - secondary electron detector (recognizable by a grid in front of it). Purple color is caused by platinum deposition from Gas Injection System.


Now we can take a selfie: (SU3800 selfie for reference). You can also see more SEM photos from SU3800 here.


On the back-scattered electron detector, I see it has subdivisions for topographic mode, but these are not selectable in the UI. Unlike SU3800, I see no secondary electron signal here at all (which is expected, but how SE was detected on the SU3800 BSE detector is a mystery).


To get this selfie, a dielectric (glass hemisphere on fused silica insulator) is charged by an electron beam to a high potential, which then reflects electrons and allows to observe SEM chamber from the inside:


Now we can briefly compare the image of a 180nm SRAM cell on Versa 3D:


vs the same cell on a Hitachi SU3800 (tungsten filament source). Not only resolution is much worse, it also much more painful to use because of low brightness of the electron source (I.e. much lower frame rate at high magnification):


Resolution target:


First ion beam milling tests:





Optical disk:


Electron backscatter diffraction detector - single crystal silicon sample:


The most useful external detector - EDS X-ray spectrometer. Fortunately, it is TEC-cooled. It has a polymer window, so I will have to be quite careful with it as it’s more fragile than a silicon one which is commonly used.



And finally, 0.5mm lead from an automatic pencil - these folded graphite sheets look quite a bit more complicated than expected:


It was quite a journey to get all of this installed and working. Still, there is much more ahead. Maybe some readers can help?:

• Does anyone know if software later than 6.3.0 can be installed? Does it make sense?
• Does anyone know of or have libraries for automation? I need stage control, stigmator/focus/gun alignment and real-time image... I see that it’s all going through the network, so it should be possible, but I would very much prefer not to reverse engineer it.
• What is the part number for the ICE detector? Any hints where it can be found?
• How to switch between CBS/ABS BSD mode (concentric/topographic)?
• Does anyone have experience with ICD detector? I got it working only once, and even then it worked poorly.
• If anyone got a good deal on a 100 amu quadrupole mass spectrometer with electron multiplier, I am interested, both for leaks and sample analysis.
• If you have old aperture strips (both ion and electron) - I am interested in getting these to make custom ones.
• Any service manuals would help.

If you are working on a hobby project that you will publish as an article or a video and have something that requires inspection under a SEM (ether image or EDS spectra), I may be able to assist you. It can also be done commercially with no publication / reference. Feel free to contact me.

Thanks for continued support on Patreon - knowing that people are interested prevented me from being too lazy countless times! You can also follow us on Twitter @Zeptobars or subscribe to our RSS feed and finally on Telegram @zeptobars_com - we'll continue opening chips.