Lab Notes for the Scanning Tunneling Microscope

Rebuilt the experiment to the last state (oscillscope for live viewing, red pitaya for acquisition and x/y control, lab power supply).

Made a new tip to reproduce tunneling. Experimented a bit with different tip cutting strategies to see whether we can improve our results with better tips:

• Tried cutting a tip flat, without pulling
  • Seeing very weird offset currents when getting somewhat close to the sample. These are definitely not tunneling effects.
  • The offset seems to throw off the PI controller quite a bit
  • Not able to achieve much stable tunneling
• Tried cutting a tip at an angle, but with much less pulling force
  • Very very good results, achieving tunneling got considerably easier
  • Almost no hysteresis around the probe touching distance, making coarse approach very easy
  • Hypothesis: The tip is now much more stubby which makes it less prone to bending during sample contact.

Also started to put more effort into making sure the tips are inserted straight into the STM.

Made the C7 capacitor hack more permanent by soldering the capacitor to the respective terminals rather than wiring it in with alligator cables.

New address for the Red Pitaya Notebook: http://redpitaya.lab.fa-fo.de/jlab/lab/tree/RedPitaya/Fafo/STM_Control.ipynb

Wrote a new script for sweeping one of the red pitaya: http://redpitaya.lab.fa-fo.de/jlab/lab/tree/RedPitaya/Fafo/STM_Sweep.ipynb

Connected the output of the Red Pitaya (CH1) to the X input of the control box.

Added a 1k resistor parallel to R17 to increase the X-axis gain from 1 to 11. This means the input voltage is now -1V to 1V rather than -10V to 10V.

Managed to take some sweep recordings while the STM was tunneling:

We definitely can't see any surface here, but we could reliably observe that the X movement slowly moved a stable tunneling out of range.

Had troubles to get stable tunneling current, went to

• Probenvorspannung: Maximum
• Sollstrom: 0.144V (= 152pA)
• I: 5 (instead of 9 turns up (out of 10))
• P: 9 1/8 turns up (out of 10)

And now we get stable tunneling current for several minutes. We will for now continue with those settings.

Started integrating the Red Pitaya for X-Y scanning. As a first step, connected the Z-output of the control box to IN1. Added a BNC splitter so we can still see the control signal on the real oscilloscope. Once tuning the STM to tunneling range, we can now acquire the control-loop output from a Jupyter Notebook. Needed to update the firmware.

Access to the live notebook (only works from the lab): http://10.250.240.129/jlab/lab/tree/RedPitaya/Fafo/STM_Control.ipynb

A copy of the notebook is synchronized to git (please update!): https://git.fa-fo.de/fafo/k8ik-stm/src/branch/main/Control

New image of tunneling current after firmware update of the Red Pitaya and new script with the new Red Pitaya python API:

Right now the script is still very unusable, even for pure acquisition. We assume a problem is that the decimation factor is way too high at the moment. We should also look into what RP calls “Deep Memory Acquisition” to not be limited to the 16k sample buffer size.

Started making a bunch of new tips to see whether there is significant difference between them. There is, some tips are way easier to get to tunnel. ⇒ We need to work more on tip making skills to get reliable results here.

Rebooted the experiment, electronics still work. The old tip doesn't work anymore, we are unable to achieve reliable tunneling.

Installed a new tip. Got controlled tunneling again from time to time, but not for long periods.

Made yet another tip. Now hysteresis of the micrometer screw is gone, but we still only tunnel for short periods before drifting away. Assuming temperature drift.

Installed a PT100 RTD on the sample holder of the STM to monitor temperature. The PT100 is connected to a 3.3V power supply and a resistor divider with 100R. Unfortunately, the RTD responds way too slow for any useful results. Aborted the experiment. We should get a different RTD to build a useful temperature monitor.

Tried reproducing the results from yesterday → achieving tunneling was quickly done. The goal today is to tune the control loop better.

Achieved control for short periods of time, but the control behavior looks way too slow:

Assuming that the proportional control is not doing its job right now.

Added back an R8 of 100k to shut up the I-controller, to just look at the P-controller. It shows its familiar ringing when giving a factor using the potentiometer. This even couples back into the probe current, as can be seen below (at this time, the tip is very far away):

Frequency of the ringing is 1.47kHz.

Added another 1µF capacitor across C7. This gets rid of the ringing across the whole range of the P potentiometer.

With this, we are now able to see actual stable control over long periods of time with fast response from the control loop:

You can also see the piezo sweep using the I-controller until it finds the tunneling point.

What's still a bit unclear is how there can be situations like in the image above where tunneling stops again as the Z piezo moves even closer?

Part of a very stable tunneling over a very long period today:

During mechanical adjustment, the controller now sweeps until it finds something (left side of the picture below). On the right, it has gone into control.

Settings during the experiments above:

• Probenvorspannung: Maximum
• Sollstrom: 0.322V (= 322pA)
• I: Minimum
• P: Maximum

Now changing the settings somewhat in hopes of getting the control loop more tight:

• Probenvorspannung: Maximum
• Sollstrom: 0.144V (= 144pA)
• I: 9 turns up (out of 10)
• P: 9 1/8 turns up (out of 10)

You can see the faster I-response at least:

It seems the noise has increased since yesterday. We don't have values from before, but right now, seeing ~80mV of noise overlayed over the tunneling current signal.

After some time, the control loop settled down quite a bit, assumingly because process disturbances decreased for some unknown reason:

Managed to increase the setpoint current to 456pA:

Finally tried putting the lid onto the STM for noise shielding, but I think I crashed the tip way too hard during the process as I couldn't achieve stable tunneling again afterwards…

Should create a new tip next time and then retry more gently. Also X/Y control is probably a good next step.

Changed power supply from lab power supply to batteries to check source of 50Hz noise. Took four 9V block batteries to generate +18V/-18V and connected to power input on control unit. 50Hz noise on exact same level as with lab power supply. Current explanation: STM is still on ground of building which seems to be the source of the noise. STM is here grounded via the scope we use which is connected to ground. Currently checking whether it's possible to have it in floating ground mode.

When

• turned off, we measure 120mV with the multimeter (between exit of first amplification stage and ground)
• tunred on, we measure 2,8mV, ——“——
• turned on with connected ground we measure 5,5mV

Taking other scope makes no difference.

Adjusting the amplification steps with the rotary switch on the amplification board gives further questions:

• Highest amplification stage - to the very right (clockwise) - 10 MOhm - gives saturation of the 50Hz noise (rectangular, +/- 5V)
• Second amplification stage - second to the right (clockwise) - 1 MOhm - gives additional to 50Hz a 12kHz noise in saturation (+/- 5V)
• Third amplification stage - third to the right (clockwise) - 100 kOhm - gives 50Hz with 100mV amplitude with 12kHz in saturation (+/- 5V)
• Smallest amplification stage - to the very left (clockwise) - 10kOhm - gives a clear 50 Hz noise with 100mV amplitude

We now soldered in a very professional, Rahix-approved way, a 1 nF capacitor parallel to the second amplification stage into the amplifier board:

We got rid of the 12kHz!!

• 10MOhm: Saturated
• 1 MOhm: Saturated
• 100kOhm: 1V Amplitude
• 10kOhm: 100mV Amplitude

Without ground we measure ~250mV rms (completely isolated), with ground over control box ~400mV rms and with ground on STM ~600mV rms.

We now checked again with the lab power supply to see the difference to the batteries: Without ground we measure ~680mV rms (completely isolated), with ground approximately the same.

We now found a way to reduce the amplitude of the 50Hz noise from ~1V to 15mV by grounding the table!

Checking the table grounding effect with the power supply. Result: Grounding table leads to 15mV. Final explanation of main source of 50Hz noise: Cables in table induced 50Hz noise on metal parts of table. We do not understand however, why rahix observed the opposite effect last time he checked.

So far, the piezo cable has been left floating. We noticed that connecting it to the control box or otherwise grounding the shield of the cable gets rid of the last 15mV of 50Hz noise. So the cable was acting as an antenna…

After we replaced the IC 5/6, we achieved a first, short, tunneling current. I-part in control unit close to zero, high “Probenvorspannung”, low “Sollstrom”. We now replaced also IC 2/4 and the undocumented magic IC “Eleven”.

Tried measuring the travel of the piezos

• Z piezo connected to separate power supply
• Monitoring preamp current
• Approaching mechanically until contact
• adjusting Z piezo voltage to find contact point
• could not reliably find contact point after mechanical adjustments
• after playing, we managed to find a z voltage with seemingly stable tunneling current! (roughly 3.3V fwiw)

Reconnected the control loop now, managed to again achieve controlled tunneling for a while:

• Settings:
  • Probenvorspannung: 4.65V
  • Sollstrom: 0.359V (this means 359.0 pA of tunneling current)
  • I: 0.041V
  • P: 0.249V

[Pink: Amplified sample current; Yellow: Z-axis Piezo control voltage]

Preamplifier voltage to tunneling current mapping:

Got a DVD sample! Prepared a small section to put under the STM.

Looking at STM tips under the SEM:

• Sample 1: Tip created today that was used for the piezeo approach tests and successfull tunneling later today
• Sample 2: Tip created just now that has not been touched or used in any way

Achieved tunneling once again with the new tip and DVD sample installed. Very unstable though, quickly drops out every time. SEM was running at the same time, possibly partly related?

[Current measurement setup]

• SDS 1104X-E
  • CH1: BNC_Z output via BNC cable
  • CH2: Purple probe R3 (X1:3)
  • CH3: Yellow probe R7 (IC1B:7)
• Multimeter: R36:2

Looking into the behavior of the PI controller in more detail again:

• When tip is crashed, error input is -5V
• When tip is away, error input is -0.41V
• Sollstrom is positive 0.4V
• Voltage from the preamp is also positive?
• ⇒ The circuit is not actually calculating a difference, but rather a sum!
• Trying to invert the setpoint value (Sollstrom) by disconnecting R36 from C1/R1 and instead connecting it to -15V through a separate 10k resistor.
• This helped, now the error value switches sign when the sample current grows beyond the setpoint
• The integrator is still not integrating far enough → the reason is R8, a resistor across the feedback capacitor of the integrator
• Cut out R8 to make the integrator work at DC (with the resistor, it is essentially a low pass instead)
• The integrator now moves towards limiting values
• Questioning whether the piezo sign is correct or if it needs to be inverted?
• When the setpoint is not close to the actually achievable tunnel current, the PI controller now goes into saturation
• Once adjusting the setpoint accordingly, the loops starts controlling
• We swapped the piezo pins around in the cable plug to invert the sign
• But we think the original polarity was already correct
• Reverted piezo polarity in the cable plug again

After playing with the parameters a bit more, we got stable control for the first time!!!!!!!!

[magenta: tunneling current; cyan: control loop error input; yellow: piezo voltage]

You can see the piezo voltage adjusting over time to track the setpoint tunneling current (about 40pA in this case).

Control Values for this experiment:

• Probenvorspannung: 4.65V
• Sollstrom: 0.039V (this means 39 pA of tunneling current)
• I: 4.89V
• P: 9.44V

Played a bit more with it, but it was very hard to get it to become stable again. Assuming environmental factors now.

Cleaned up the setup from yesterday. Noticed a shift in the noise level after moving the equipment a bit…

Accidental discovery: Additional spikes show up in the waveform when touching bare screws of the table that the STM is sitting on.

• Actually bonding the table to PE leads to much worse noise, mostly in higher frequencies, showing up
• The additional noise seems to be ~500Hz and has a very erratic waveform
• Disconnected 0V from PE, same effects and noise levels as before
• Added a separate 0V connection which runs parallel to the PE bonding of the STM control box

Played a bit more with the calculations for the mass-spring-damper system to see where we are at, comparing to commercially available systems:

Finished builing the vibration isolation frame + EMI box.

Measuring the damping time of the STM block (without magnet damping). STM block is deflected by 4cm, we measure the time until the amplitude is half. Target is to measure the increase in damping by our eddy current brake. Time is 4,5 seconds.

Did not retry the experiment with the eddy current brake installed: The damping effect is so little that it would not be worth it…

Lesson learned: Our eddy current brake was not worth the effort. We'd need a lot more magnets to have any notable damping effect. It is probably better to solve the problem through more mass.

Calculated the damping behavior of our setup:

EMI Box:

• Need to install some handles/knobs to the upper box so it is easier to remove.
• Quick test of the EMI box by placing a phone inside and checking whether it still has signal → it does so the box is way too leaky still!

Electric bringup of the amplifier and control board:

• Powering on, verified the pinout of the DIN connector towards the amplifier board (Conector is labeled Signal out)
• Probenvorspannung knob: CCW smaller, CW bigger
• BNC Z is the controlled height value generated by the PI controller, so it is an output signal, not an input
• All knobs have 10 turns

Current setup:

• X and Y inputs are shorted to ground to prevent piezo movement.
• Z is connected to scope, Z piezo will be controlled by the PI controller
• Probenvorspannung is set to maximum (TODO: What was the actual voltage?)
• Sollstrom is minimal (probably, we are not sure!)
• I controller factor is half way (5 turns)
• P controller factor is half way (5 turns)
• A new tip was made and inserted into the holder
• The tip was positioned close to the brass surface by mechanical adjustment

Experiment:

• Turned on power
• Saw a ~5kHz waveform that flattens on probe contact, but voltage at Z stays consistently at ~12V
• No changes in Z voltage from any changes to P or I values
• Started looking into the PI controller behavior, got some strange reaction to Sollstrom changes

Trying to understand the PI controller:

• Disconnected the piezos from the control box to reduce influence factors while debugging the control loop
• Verified that the probe is indeed not touching by measuring resistance from probe to ground: >1M
• Turned Probenvorspannung to max and measured probe to ground: -4.63V
• Tuned Probenvorspannung such that a voltage of -0.5V is measured at the probe
• Measured short circuit current probe to PV (Probenvorspannung): -0.5µA
• Added a wire that shorts the probe to ground, making the -0.5µA current the effective tunnel current “measurement”
• Measured voltage at the output signal from the pre-amplifier (X1:3): 4.96V
• Sollstrom pot tops out at 4.7V
• Measured the error input to the PI controller at IC1:7 ↔ R7:
  • Sollstrom minimal: -4.96V
  • Sollstrom maximal: -9.66V
• Same measurement, but with the scope shows a rect wave @50 Hz. Depending on the Sollstrom it oscillates between:
  • Sollstrom minimal: -6.0V - 3.8V
  • Sollstrom maximal: -9.8V - 0.1V
• Scope on the Iststrom X1:3 (R3): Same 50 Hz rect, oscillating around 0V with an amplitude of 9.9V
• Adding the probe to PV bridge reliably changes this such that the signal is a constant 4.8V
  • This makes sense, the amplifier is outputting its maximum value when the probe is shorted
• Now, changing the Probenvorspannung to minimal leads to an amplifier output of -5.1V (minimum). Even the slightest notch upwards on Probenvorspannung raises the amplifier to its maximum (4.8V).
  • Also makes sense, as soons as there is voltage, the maximum current flows through the probe short, saturating the amplifier.
• Next question: why is the amplifier already saturated when the probe is far away? ⇒ Assuming 50Hz noise
• Added additional grounding connection to the bottom tray of the EMI box
• Tried measuring the voltage on the preamp input
  • We do see a 3.8mV of voltage noise which seems to grow and wane with a 50 Hz period
  • Adding the top EMI box does not have any measurable impact on this
• Measured the voltage betweenn the two TLC2202A preamplifier stages:
  • Seeing a nice 50 Hz sine at 60mV
  • Top EMI box again does not impact this at all
  • Even after turning off the STM power, a 7.6mV sine @50 Hz remains
• Adding a second 2.5mm² ground wire from PSU to preamp-starpoint has no visible impact
• Tried aluminum foil shielding, no change for 50 Hz or the visible higher frequency noise content
• It is possible to observe the 50 Hz noise increasing when moving an AC chord close (< 10mm) to the STM probe. This increases the amplitude by about 50% but admittedly the AC chord was not carrying a lot of current.
• Plugging the piezo cable back in drastically changes the noise pattern. This probably need additional investigation…

In conclusion:

• Leaving the piezos entirely disconnected, we can observe 50 Hz mains noise on the pre-amplifier.
• Between the two amplifier stages, this noise is nicely visible as a sine wave (after the second stage it is just a rectangle because it is completely saturated).
• We need to figure out how to reduce this noise to have a chance of measuring tunnel current.
• We need to figure out why the piezos introduce this much additional noise.

Removed the formwork from the conrete block we poured on friday. The block is so nice!

• Lessons learned:
  • Screwing the M5 threaded rods directly into the formwork meant we had to break apart the bottom panel. Maybe there is a better way?
  • We should have fixed the top ends of the M5 threaded rods into correct position somehow. The misalignment made it very hard to attach the aluminum profiles afterwards.
  • Not using any demoulding spray was totally fine, the OSB panels easily separated.
  • The OSB panels left a very nice texture :)
  • We should have put some edge liners into the formwork to chamfer the block edges — especially the ones on top, where the air surface of the pour was.

Continued bringup of the original control electronics.

Continued building the vibration isolation frame, up to the point where we are missing the remaining components. The concrete weight is still drying, did not want to touch it yet.

Prepared a lab setup for bringing the original electronics back to life. Turned on power again for the first time, with the power LED on the board lighting up! Got a spark in the enclosure, couldn't find a reason or fault from it. Did not yet verify any functionality of the board.

Poured a concrete weight for the vibration isolation system:

• Lessons learned:
  • You probably need more water then called for on the package to make the concrete pourable.
  • Something to vibrate/shake the form is extremely useful.
  • Don't try to make reinforced concrete if you don't have to :D

Started building the frame for the STM vibration isolation.