Also
The STM relies on the precise control of the probe-sample gap. Any external vibration has the potential to disrupt this, so efforts must be made to isolate the instrument. Low-frequency vibration isolation is generally done with large pneumatic or elastomer damping systems. I'm hoping to be able to avoid these by making the instrument as compact and stiff as possible, thereby maximising its resonance frequency (and therefore its response to low frequency vibrations). To isolate from high frequencies, I have chosen the 'stacked plate' approach, as described and characterised in "Low- and high-frequency vibration isolation for scanning probe microscopy" (A I Oliva, M Aguilar and Victor Sosa in Measurement Science and Technology, 9 (1998) 383-390)
The drawings for the isolator stack plates can be found here. The plates are each separated with 3 pieces of 5mm long, 5mm diameter pieces of viton rubber as described in the paper. I used an off-the-shelf viton o-ring from a local supplier and cut pieces off, securing with cyanoacrylate glue:
The assembled stack:
I haven't yet characterised the attenuation of the isolator stack - more on this after I've done some testing.
The isolator stack will rest on a tufnol base plate, which will provide connection points for the wires coming off the instrument. These connection points will run back to the control module via JST SR-series connectors on adapter PCBs in 3 recesses on the base plate. A shallow rectangular 'trench' in the plate runs around the base of the isolator stack. This is a mating recess for a steel fabric-lined hood that will cut down acoustic and temperature disturbances, and provide a mounting port for a USB microscope (used in the initial coarse approach phase of a measurement).
The drawing file can be found here.
The piezotube was purchased from EBL Products Inc. The material is EBL2 (PZT-5A). The dimensions and electrode configuration are as below:
I've used 36AWG telfon-coated wire (R36A705-00 from Hobby King) to connect to the electrodes, secured with epoxy glue. The electrical connection is done with Wire Glue conductive adhesive, as I didn't want to risk depolarising the piezotube by heating it with a soldering iron. The Wire Glue contacts were reinforced with a coating of cyanoacrylate. It's a bit ugly but the connections are very secure:
The microscope must have a way of bringing the probe to within tunnelling distance (i.e. nanometres) of the sample without the two crashing into each other. This requires a mechanism that possesses the following properties:
Two traditional approaches have been stepper motors and 'inchworm'-type piezoelectric devices. Stepper motors are heavy and require lever-reduction systems to achieve the required linear distance resolution. The inchworm devices typically use exotic materials and can be difficult to control.
I have opted to use a Piezo LEGS linear motor from PiezoMotor. This motor weighs only 23g, has nanometer resolution, and PiezoMotor supplies a control module that enables the motor to be controlled via USB. The drive rod is made from alumina, which means it will have a similar coefficient of thermal expansion to the ceramic holder and piezotube.
The piezotube and coarse positioner must be attached to a thermally stable, rigid platform to maintain their relative orientations. The platform must be thermally stable to avoid measurement errors due to thermal expansion and contraction, and it must be rigid to minimise the effects of vibration. I have opted for a simple 'L'-shaped design made from Macor, a machinable ceramic made by Corning Glass. Since the piezotube and coarse positioner rod are made from ceramic and alumina respectively, any expansion or contraction in these components that would alter the distance between the probe and sample will tend to be cancelled out by a similar expansion or contraction in the support platform. M1.6 mounting hardware is used to secure the electronics to the support, and the support to the isolator stack.
This module has 2 functions:
I have chosen a 12-bit bipolar DAC that will generate -10 to 10V. 12 bits gives an output resolution of 5mV, which is probably overkill. The DAC output is conditioned with a Sallen-Key filter to reduce noise bandwidth. The filter cutoff frequency is 1 kHz.
The DAC output is controlled via an SPI interface. The board is supplied with +/-15V rails, which are downregulated to +/-12V for the analogue output, and to 3.3V for the digital interface of the DAC.
A transimpedance amplifier is required to convert the tunnelling current to a voltage signal. I have opted to go with a high-bandwidth dual stage design as described in A low-noise and wide-band ac boosting current-to-voltage amplifier for scanning tunneling microscopy (Dae-Jeong Kim and Ja-Yong Koo, Review of Scientific Instruments, 76, 023703 (2005)). The design in the original paper uses a OPA111 op-amp as the input amplifier, and a OPA2111 dual op-amp for the ac recovery stage. I have opted to use an AD549 in a TO-99 case, as the TO-99 offers the ability to connect the metal case to ground potential, which should assist in screening out external noise. For the dual op-amp I have opted to use an OPA2209, as it is a much better performing device (lower noise, greater gain bandwidth, greater CMRR etc). The design files (in Kicad format) can be found here.
The amplifier will mount onto the back of the ceramic support, so the signal wire carrying the tunnelling current can be as short as possible.
The tunnelling current amplifier is an example of a transimpedance amplifier: it converts a current input into a voltage output. This amplifier is a critical circuit in the microscope, so I have taken advantage of available electronic simulation software (TINA-TI version 9, free to download from Texas Instruments) to test and tune the design before committing to a PCB. I have a choice between a simple, 'classic' transimpedance amplifier and a more complex - but possibly better performing - 'composite' amplifier.
In a transimpedance amplifier there is an unavoidable tradeoff between the bandwidth of the amplifier response and its output noise. The higher the bandwidth of the amplifier, the faster the response, but also the higher the output noise will be. These two parameters have the following effects on the microscope performance:
The design of the tunnelling current amplifier is guided by the following questions:
The fullscale output signal will be +/-10V, corresponding to +/-10nA (1G resistor). Assuming a 16-bit ADC, this means 1 LSB will be 20V / 65535 = 305uV. Ideally the noise floor of the amplifier would be below this level.
The 'classic' transimpedance amplifier design for small currents such as photodiodes etc is an op-amp with negative feedback via a single large value resistor, with a compensation capacitor to prevent oscillation due to noise gain:
R1 is the feedback resistor, C2 is the compensation capacitance (including the parasitic capacitance of the resistor itself). C1 represents the parasitic capacitance of the current source IG1. The output of the circuit is nominally is 1V/nA.
Assuming 10pF capacitance for the tunnelling signal input, and using only the parasitic capacitance of the feedback resistor for compensation (around 0.3pF for a leaded resistor), the 3dB bandwidth is about 520 Hz - not great, but possibly usable:
The output noise looks to be around 200uVrms, which comfortably meets the noise requirement:
An alternative, 'composite' amplifier is described in A low-noise and wide-band ac boosting current-to-voltage amplifier for scanning tunneling microscopy (Dae-Jeong Kim and Ja-Yong Koo, Review of Scientific Instruments, 76, 023703 (2005)). This amplifier has a 'classic' front end, followed by an 'ac-boosting' section to recover the high-frequency part of the signal:
This circuit needs careful tuning of the values of C9 and R5 for a flat response of the desired bandwidth. I have found that setting C9 = 1nF and R5 = 23kOhms gives a 3dB bandwidth of 5kHz, which is more than adequate:
The rms noise out to 5kHz is 300uV, which meets the noise requirement. Out to 100kHz or so it rises to 1000uV or so, but this that be mitigated with filtering:
The noise floor of the amplifier should practically be 1-2 LSB. Of course, this doesn't include the downstream conversion noise etc.
The composite amplifier yields a bandwidth an order of magnitude greater than the simple amplifier, for a moderate increase in noise. There is scope to make the bandwidth of the composite amplifier greater still (by tweaking C9 and R5) for applications that require faster scan times, and have less stringent noise requirements.