Piezoresistive AFM cantilevers surpassing standard optical beam deflection in low noise topography imaging

Optical beam deflection (OBD) is the most prevalent method for measuring cantilever deflections in atomic force microscopy (AFM), mainly due to its excellent noise performance. In contrast, piezoresistive strain-sensing techniques provide benefits over OBD in readout size and the ability to image in light-sensitive or opaque environments, but traditionally have worse noise performance. Miniaturisation of cantilevers, however, brings much greater benefit to the noise performance of piezoresistive sensing than to OBD. In this paper, we show both theoretically and experimentally that by using small-sized piezoresistive cantilevers, the AFM imaging noise equal or lower than the OBD readout noise is feasible, at standard scanning speeds and power dissipation. We demonstrate that with both readouts we achieve a system noise of ≈0.3 Å at 20 kHz measurement bandwidth. Finally, we show that small-sized piezoresistive cantilevers are well suited for piezoresistive nanoscale imaging of biological and solid state samples in air.


SUPPLEMENTARY INFORMATION Piezoresistive AFM cantilevers surpassing standard optical beam detection in low noise topography imaging
Maja Dukic, Jonathan D. Adams where is the cantilever length, is the position along the cantilever length (starting from the fixed end), is Young's modulus of the cantilever material along its length and is the moment of inertia of the cantilever cross section about its neutral axis. The laser beam used in OBD readout is a Gaussian beam, and it is common to define the laser beam diameter as a point where the laser intensity falls to a fraction 1/ ! of its initial intensity. We will denote ! as the laser beam diameter along the cantilever length. In order to reflect most of the laser power off of the cantilever surface, the optimal position of the center of the laser beam spot, along the cantilever length is ! ≈ − ! 2. Inputting ! in equation (S2) and expressing it in terms of ∆ we get the equation II Measurement setup AFM imaging and noise measurements were performed using a custom made cantilever holder (see Figure S1a-c). The cantilever holder was designed in order to enable simultaneous measurements with both the OBD and the piezoresistive readout. A stack piezo actuator (PL022.30, Physik Instrumente, USA) was integrated in the holder to excite the cantilever resonance. The custom made electronics setup was used for electrical readout of the self-sensing cantilevers. The electrical readout consisted of a full Wheatstone bridge of piezoresistors located on the cantilever chip and subsequent amplification stages (see Figure S1d). The flexible printed circuit board (PCB) was used to provide signals to and from the cantilever chip. A low noise instrumentation amplifier AD8429 (Analog Devices, USA) was positioned on the flexible PCB, close to the cantilever chip to reduce noise and stray capacitances of the electrical lines. The rest of the amplification stages were located on the readout electronics PCB. An ultra-precision, low noise voltage reference ADR420 (Analog Devices, USA) was used to bias the Wheatstone bridge.

III Estimated parameters for the OBD readout
For the case of the OBD readout, parameters for a custom made optical AFM head designed for AFM imaging with the small sized cantilevers 2 are given in Table S1. All calculations were performed for room temperature = 23 ! .

IV MDD derivation for OBD readout
The overall deflection noise of OBD readout can be calculated by adding deflection power spectral densities (PSDs) of all relevant noise sources and then integrating this sum over the frequency range of an AFM lock-in measurement bandwidth. To perform this calculation, the deflection sensitivity of the OBD readout method also needs to be determined in order to scale electrical noises from amperes to distance units.
The mean square deflection at the cantilever free end, for fundamental resonance mode, as seen with OBD readout is 3-5 where ! is the Boltzmann constant, is the temperature and ! = 1.875 for the first resonance mode. Equation (S4) differs from equation (3) given in the paper because OBD readout measures angular changes, rather than deflection, so the correction factor is introduced [3][4][5] . From equation (S4) and the cantilever amplitude transfer function 6 we obtain the deflection noise PSD of the first resonance mode of the cantilever: where ! and are the cantilever fundamental mode resonance frequency and the quality factor, respectively. Assuming that we excite the cantilever oscillations at the frequency ! ≅ ! and that the lock-in measurement bandwidth is , we obtain the power of the deflection noise coming from the thermomechanical noise 7,8 : In the OBD readout electronics, a transimpedance amplifier is commonly used as the first stage amplifier, to convert the photodiode current ! to voltage. Noise sources present in this readout include the photodiode current shot noise !,! , the amplifier input referred current noise !,!"# and the feedback resistor noise !,! . These noise sources can all be treated identically as the current noise sources, which add onto the measured photodiode current 9 . These noise sources have the same gain, which is constant in the amplifier flat-band.
Additional noise source also present in the readout is the influence of the amplifier noninverting input voltage noise !,!"# . This noise, contrary to other noise sources, varies with frequency in the amplifier flat-band. !,!"# is multiplied by the amplifier's non-inverting closed loop gain !" ( ).
The referred-to-input (RTI) voltage amplifier noise !,!"# sees the amplifier circuit as presented in Figure S2, where ! = ! || ! represents the total feedback impedance and ! represents the total stray capacitance which includes the photodiode capacitance, the amplifier input capacitance and other stray capacitances (e.g. such as the ones coming from the traces).
If !" ( ) is the amplifier open loop gain frequency response, then !, !" is amplified by the amplifier's non-inverting closed loop gain: Figure S2. The amplifier circuit seen by the RTI amplifier voltage noise !,!"# In the amplifier flat-band we can assume that By substituting ! = ! 1 + ! ! in equation (S7) and assuming that ! ≫ ! we obtain From (S9) we see that the closed loop gain transfer function has a zero at ! = 1 2 ! ! (where the gain starts to rise) and a pole at ! = 1/2 ! ! (where the gain levels off). For low frequencies !" ( ) ≈ 1 which will result in the input referred current noise of !,!"# ! . Finally, the total input referred current noise PSD of the transimpedance readout in the gain flat-band can be written as where is an electron charge and ! is the resistance of the feedback resistor. In equation (S10) we take into account the entire current of the photo sensitive detector ! but for the rest of the noise sources we consider only noise sources coming from one transimpedance amplifier. Depending on the number of photodiode quadrants !" in the detector, the remaining noise terms need to be multiplied by !" to obtain the total readout noise. Finally, the current noise needs to be scaled to distance units by the deflection sensitivity (in units). The deflection sensitivity of the OBD readout is 10 where ! is the diameter of the short axis of the collimated laser beam, ! is the focal length of the focusing lens, is the efficiency of the light-to-current conversion at the photodiode, is the correction factor correcting for the assumed rectangular shape of the laser spot 10 , ! is the laser power and is the total laser power attenuation factor (coming from the optical path loss, the laser light spillage and the cantilever absorption). Finally, the total deflection noise or MDD of OBD readout equals to In this section, we ignored laser noise sources for two main reasons: the laser intensity fluctuations are mostly eliminated by the differential amplifier present in the OBD readout electronics, and laser mode fluctuations that cause fluctuations in spatial distribution are very hard to estimate and strongly depend on the OBD setup 11 .
Finally, in most cases the noise term coming from the photodiode shot noise will be the dominant one ( ! ≈ 2 ! ) and for any well designed system it will determine the lower limit of deflection noise 10,11 . Therefore, we used this assumption in the noise calculations presented in the paper.

V Estimated parameters for the piezoresistive readout
For the case of the piezoresistive readout, the estimated readout parameters for the 70×30 µm sized piezoresistive silicon cantilevers (PRS probes, SCL-Sensor.Tech. Fabrication GmbH, Austria) are given in Table S2. The differential amplifier used in the calculations was the low noise instrumentation amplifier AD8429 (Analog Devices, USA), set to a gain of 10. This amplifier was also used in the noise measurements and AFM imaging. All calculations were performed for room temperature = 23 ! .

VI MDD derivation for piezoresistive readout
The deflection noise of piezoresistive strain-sensing readout can be calculated in similar fashion as was the case for the OBD readout. To perform this calculation, the deflection sensitivity of the piezoresistive readout method needs to be determined in order to scale the electrical noises from volts to distance units.
The mean square deflection at the free end of the cantilever for the fundamental resonance mode, as seen with piezoresistive readout is 14 Equation (S12) differs from equation (3) given in the paper because piezoresistive strain-sensing readout measures changes in the induced strain, rather than deflection, so a correction factor is introduced. Assuming that we excite cantilever oscillations at a frequency ! ≅ ! and that the lock-in measurement bandwidth is , we obtain the power of the deflection noise coming from the thermomechanical noise: The PSD of the piezoresistor Johnson noise, from the Wheatstone bridge is ! = 4 ! . Johnson noise coming from the differential amplifier is where !,!"# and !,!"# are the input-referred Johnson voltage and current noise of the amplifier. Finally, some of the noise coming from the bridge voltage reference will affect the readout, where the level of the influence depends on the common mode rejection ratio (CMRR) of the differential amplifier. The PSD of the bridge voltage reference noise, referred to the amplifier input is where !" is the amplifier CMRR expressed in decibels and !"# is the voltage noise spectral density of the bridge voltage reference. Usually, with a well-chosen differential amplifier and a low noise bridge reference (e.g. such as battery), this noise term is negligible. The total electrical noise PSD is then In order to calculate the total deflection noise, the electrical noise needs to be scaled by the deflection sensitivity (in / units, calculated for the case of two active resistors on the cantilever) [14][15][16] !"#$% = where and are the cantilever length and thickness, ! and ! are the piezoresistor length and thickness, is Young's modulus of the cantilever material, along its length, ! is the longitudinal piezoresistive coefficient, and ! is the bridge supply voltage. Finally, total deflection noise or MDD of the piezoresistive readout equals to !"# = !!,! + !"#$% ! ⋅ ⋅ ! (S19)

VII Estimation of the cantilever mechanical bandwidth
We estimated the cantilever mechanical bandwidth as ! / where ! and are the first mode resonance frequency and the quality factor, respectively. ! was calculated as 17 where and are the cantilever length and thickness, is the Young's modulus of the cantilever material along the cantilever length and is the cantilever material density. The quality factor in air was calculated as 18,19 = 4 where is the cantilever width, is the gas constant, and , , and are air dynamic viscosity, molar mass, temperature and pressure, respectively.
VIII Dependence of the cantilever spring constant on the cantilever dimensions The spring constants of the analysed cantilever dimensions are presented in Figure S3.