A project of mine started in 1992 - to build a home made Scanning Tunnelling Electron Microscope.
After a purchase of two Peizo tubes at the time, with much of the electronics plus most of the software done, the project was put on hold for a few years (as one does with these things). Starting again in 1996 to rebuild the electronics, and started yet again in late 1998 to build the "mechanics" of the device... which finally led to the completion, and the first crude image (in constant altitude mode) on December 3, 1998.

The mechanics:
Image 1
Shown above is an early version of the mechanics. Built only for getting something going temporarily while I finalized the software and electronics, it nevertheless worked in a crude low-res topological mode (with no Z movement) with many quirks and limitations. I really wouldnt recommend it. ;)
Current version shown above, NOT TO SCALE, has the fixed sample holder replaced by a second piezo tube, which does the Z movement for constant-current feedback (as well as holding the sample). The tip is a wire placed down the centre of the other piezo tube. Enough plastic tape is wrapped around the wire to make it a snug fit against the inside of the tube.
The sample is merely carbon... or at least thats what I suspect black pencils are made of. Yeah, its just a bit of pencil "lead". Perhaps someone can fill me in on what else goes into these things. This sample is at ground potential while the tip is at 1 or 2 volts. Thin wires lead to the sample and the tip from the pre-amp board mounted on the same base.
The tip is a gold coated [steel?] wire from an audio plug. Since the wire is merely cut at a sharp angle to hopefully get a sharp tip, I cannot be sure exactly what type of atom is actually at the tip itself.
Why did I choose this as the tip? No reason really, it just happened to be a bit of wire that was handy and the right diameter. The plan was always to just use this temporarily, and switch to a PtIr (which seems to be the one of choice) later on.
The contraption is mounted on a thick wooden base, with the tip moved towards the sample by means of a micrometer (soviet made, from a second hand shop). Vibration elimination is done by hanging the frame/mount device via rubber bands from a tripod, which in turn is sitting on my bed. Yes, my bed. I told you this thing was HOME made.
Its all very rough. Fine tuning of the distance between tip and sample is almost impossible.
Piezo tubes are EBL #2, from Stavely Sensors, CT, USA.

The software:

Back in 1992, the only computer I had was a Amiga 2000, the software was written for it in C. Since I can capture screens to images, and port them over to this borrowed notebook PC if I wish, I havnt seen a great need to rewrite the software for a PC... so on the Amiga it stays. (For now, anyway)
I'll supply the source code if anyone really wants it, but it doesnt really do much clever stuff but read data coming in the parallel port, convert the data to greyscale pixels and plot them, 256 pixels across and 256 pixels down. Another minor change to the software gave each scan line a "height" offset from the scan line to give an illusion of a 3D representation of the sample.
The biggest reason I'd have in switching computers, is that the Amiga can only read data from the parallel port and display it on the screen without losing bytes, at a rate which give a line-scan rate of about 2 per second. That is, a whole frame of 256 x 256 pixels takes a few minutes to show.

The electronics:
Image 3
The tip was initially driven in a standard raster pattern (which incidentally makes the tube give a tiny "click" as it flips back at the end of each line), but extra circuitry was added later to give a smoother sawtooth pattern.
The sawtooth pattern is generated by a Digital-to-Analog chip which is fed data that counts up by one for each "clock" pulse. It counts up to 255 and then the Overflow signal triggers a "switch" that starts it counting down to 0 at which time the Borrow signal triggers the count to start up again... repeating ad infinitum. At the end of each "X" scan, the "Y" DAC chip is incremented from the "X" count Overflow.
This really doesnt work too well, since I also now have to filter out the overflow and underflow "pulses" with extra circuitry. I really need to do all this again from scratch.
It is freerunning in this first version. Scanning starts from the moment the box is turned on.
The voltage from each of the DAC chips is run through op amps (via potentiometers to reduce the voltage for a "magnification" effect) to give both a positive and negative going voltage of the same magnitude, which is then fed to each side of the piezo tube via cabling. Again problems here - the potentiometers are starting to self destruct and become "noisy". Mods need to be made to add switches instead.
The tunnelling current is amplified by means of a FET amplifier mounted on the wooden base. The "feedback" Z voltage is created and fed back to the other piezo tube that the sample sits on. It is also sent to an ADC chip and fed via a parallel cable into the Amiga 2000. The "strobe" signal for the parallel port is the same as the clock data for the DAC chip counters, thus ensuring that we get 256 bytes of data for each scan line.
As far as sychronising the start of scan lines and scan frames with the display on the Amiga, two of the 8 parallel port data lines are reserved for this purpose. One for scan lines, one for scan frames. This leaves 6 lines ( 64 values ) for the actual data itself. Since I only have 16 colors on the screen to display the image, this isnt a big worry to me at this time.
One more problem is 50 Hz ripple on the images. Not caused by being picked up at the tip/sample, but instead because the power supply is running at its limit, and the ripple is on the 5 Volt line. Gawd damn!.
In the above image, the leftmost board is the power supply, that produces both +-18 volts, and +5 volts. The larger board in the centre produces the scanning sawtooth waveforms. The rightmost board converts the "Z" feedback signal to digital, and the socket for the parallel cable is on the panel directly behind it.

The images:
Fairly ordinary resolution. It is clear that during the scan, the sample moves (or else the tube does) since the scan never shows the same area twice. Also, the mechanics still arent strong enough. The tip moves in or out during and between each scan. Sometimes in, sometimes out, maybe due to flexing, maybe due to temperaure effects. Like I said, this is an early version.
Image 1
One of the first images.
Since the scanning voltages were +/- 15 volts at the time, my calculations (based on the FAQ) imply that the image is 130 nanometers across.
Sample is carbon. Tip voltage: 2V

Image 2
Another early image.
Scanning voltages were +/- 2 volts, image is 18 nanometers across.
Sample is carbon. Tip voltage: 1V

Image 3
Scanning voltages were +/- 2 volts, image is 18 nanometers across.
Sample is carbon. Tip Voltage: 1V

Image 4
Scanning voltages were +/- 2 volts, image is 18 nanometers across.
Sample is carbon. Tip Voltage: 1V

Image 5
Scanning voltages were +/- 2 volts, image is 18 nanometers across.
Sample is carbon. Tip Voltage: 1V

I'm not at all sure the circuitry for the constant-current part of the STM is done correctly. It simply takes the current through the sample, amplifies it one hell of a lot, and sends the resultant voltage back to the Z movement piezo. (with the polarity of the voltage correct to provide contraction at higher current)
Is this it? Sure the tip will contract away from the sample as more current starts to flow, and move back towards the sample as the current drops away, but it certainly isnt referenced against any "standard" current, like the 1 or 2 nanoamps commonly mentioned. There appears to be no way of knowing what the tip-sample current actually is. Does anyone have a better circuit?

Atomic resolution? No, not on this first version. A few (probably solvable) reasons.
1. The tip hasnt been constructed with any degree of real care.
2. Vertical resolution. The voltage being fed back to the Z axis ranges from +13 to -12 volts. This range of 25 volts, 16 colours on the screen, changes the length of the tube by a height of about 975 Angstroms. That is, each colour represents a height change of 60 Angstroms at best. Since riding the crests over individual atoms is a change of about 3 Angstroms, I need to hype up the electronics sensitivity of over an order of magnitude. This will be tricky, but probably not impossible.

The future. By early 1999, I'll have a decent PC with soundcard. ( Believe it or not, despite working in the computer industry, I do not and never have owned any kind of IBM compatable PC.) This will let me run the scan at a much higher frequency. It'll also let me get a higher Z resolution.
Stay tuned.

The Links:

Block Diagram of the Electronics.

Back to my Home Page