betaFeedback

WorksPhysics and Astronomy

Low-temperature AFM with a microwave cavity optomechanical transducer

Abstract

We demonstrate AFM imaging with a microcantilever force transducer where an integrated superconducting microwave resonant circuit detects cantilever deflection using the principles of cavity optomechanics. We discuss the detector responsivity and added noise pointing to its crucial role in the context of force sensitivity. Through analysis of noise measurements we determine the effective temperature of the cantilever eigenmode and we determine the region of detector operation in which the sensor is thermal-noise limited. Our analysis shows that the force-sensor design is a significant improvement over piezoelectric force sensors commonly used in low-temperature AFM. We discuss the potential for further improvement of the sensor design to achieve optimal detection at the standard quantum limit. We demonstrate AFM operation with surface-tracking feedback in both amplitude-modulation and frequency-modulation modes.

Introduction

The past two decades have seen the emergence of a variety of remarkable microscopic and mesoscopic optomechanical devices. Through innovative design and fabrication, new record values of fundamental figures of merit have been reported at a surprisingly rapid rate. The field is now mature for the next challenge – application-ready devices. Optomechanical interaction has been proposed to enable or improve many applications [1], including accelerometers [2], tests of quantum gravity [3, 4, 5], force microscopy [6, 7], magnetometry [8] and quantum state transfer [9]. In some cases further improvement on fundamental figures of merit is required, while in other cases the difficulty lies in balancing trade-offs to find an optimal design which fulfills the specific requirements of the application. The latter is indeed the case for force sensing in atomic force microscopy (AFM). Force transduction at maximum sensitivity requires detecting the position of a “test mass”, while minimizing the added noise introduced by the detection itself [10, 11]. The challenge for high-resolution AFM is designing such a detector for a force transducer hosting a sharp tip and scanning over a surface in ultra-high vacuum and at ultra-low temperature.

In this paper we report on an AFM cantilever force sensor with an integrated superconducting microwave resonant circuit. Using the principles of cavity optomechanics [12], we detect the deflection of the cantilever from its equilibrium position as a shift of the microwave resonance. We briefly summarize the operating principle of the force sensor based on kinetic-inductive electromechanical coupling (KIMEC), whose design and fabrication were already presented in detail in previous publications by our group [13, 14]. Here we focus on the deployment of the force sensor, demonstrating force-gradient sensing and scanning over a test surface at 10 mK in a closed-cycle dilution refrigerator (DR). We operate the microscope with surface-tracking feedback using the two most common imaging modes of dynamic AFM: amplitude modulation (AM-AFM) and frequency modulation (FM-AFM).

One of the biggest challenges of operating an AFM in a closed-cycle DR is the pulse tube cryogenic head causing significant wideband mechanical vibrations which can propagate through the cold stages down to the measurement apparatus. However, force sensors with integrated position detectors, as the one presented in this work, do not require free-space optical alignment with a long optical path length. They can therefore be made light and compact, significantly reducing their susceptibility to external vibrations, the complexity of vibration isolation, and their thermal mass. For this prototype we implement a rather simple vibration isolation which, as we will show, is sufficient to demonstrate imaging features in a 1 μm\mu\mathrm{m} scan field, but has not been tested to atomic resolution.

Force sensitivity and imaging

The microcantilever in dynamic AFM operates as a resonant force transducer. Each individual eigenmode is described by a mechanical susceptibility χ\chi, expressing cantilever deflection ζ\zeta in response to a force FF at the frequency Ω\Omega

χ(Ω)=ζ(Ω)F(Ω)=1meff(Ωm2−Ω2+iΩΓ),(1)\chi(\Omega)=\frac{\zeta(\Omega)}{F(\Omega)}=\frac{1}{m_{\mathrm{eff}}(\Omega_{\mathrm{m}}^{2}-\Omega^{2}+\mathrm{i}\Omega\Gamma)}, \tag*{(1)}

where Ωm=k/meff\Omega_{\mathrm{m}}=\sqrt{k/m_{\mathrm{eff}}} is the resonance frequency and Γ=η/meff\Gamma=\eta/m_{\mathrm{eff}} the damping rate, with meffm_{\mathrm{eff}}, kk and η\eta respectively the effective mass, stiffness and damping coefficient of the eigenmode.

A key figure of merit of a force transducer is its force sensitivity determined by all sources of noise. The total noise can be expressed as the sum of the power spectral densities (PSD) of uncorrelated contributions such as the thermal fluctuations in cantilever deflection Sζζcant(Ω)S_{\zeta\zeta}^{\mathrm{cant}}(\Omega) [m2^{2}/Hz] at the effective mode temperature TT, and the frequency-independent added noise of the detection SVVS_{\mathrm{VV}} [V2^{2}/Hz] which is converted to an equivalent deflection noise SζζdetS_{\zeta\zeta}^{\mathrm{det}} via the detector responsivity α\alpha [V/m]

Sζζtot(Ω)=Sζζdet+Sζζcant(Ω)=SVVα2+2kBTη∣χ(Ω)∣2.(2)S_{\zeta\zeta}^{\mathrm{tot}}(\Omega)=S_{\zeta\zeta}^{\mathrm{det}}+S_{\zeta\zeta}^{\mathrm{cant}}(\Omega)=\frac{S_{\mathrm{VV}}}{\alpha^{2}}+2k_{\mathrm{B}}T\eta\lvert\chi(\Omega)\rvert^{2}. \tag*{(2)}

The noise-equivalent force for a given measurement bandwidth Δf\Delta f is given by,

FN(Ω)=(∫Ω−πΔfΩ+πΔfSζζtot(Ω)∣χ(Ω)∣2dΩ2π)1/2.(3)F_{\mathrm{N}}(\Omega)=\left(\int_{\Omega-\pi\Delta f}^{\Omega+\pi\Delta f}\frac{S_{\zeta\zeta}^{\mathrm{tot}}(\Omega)}{\lvert\chi(\Omega)\rvert^{2}}\frac{\mathrm{d}\Omega}{2\pi}\right)^{1/2}. \tag*{(3)}

This quantity is minimized at the mechanical resonance frequency, where ∣χ(Ωm)∣2=meff/(kη2)|\chi(\Omega_m)|^2 = m_{\mathrm{eff}}/(k\eta^2), resulting in a minimum detectable force,

Fmin⁡=FN(Ωm)≈(SVVkη2α2meff+2kBTη)Δf.(4)F_{\min}=F_N(\Omega_m)\approx\sqrt{\left(\frac{S_{VV}k\eta^2}{\alpha^2m_{\mathrm{eff}}}+2k_BT\eta\right)\Delta f}. \tag*{(4)}

It is worth noting that eq. (4) clarifies a recurring misconception. Decreasing stiffness kk does not necessarily reduce Fmin⁡F_{\min}. Rather, it relaxes the constraint on the detector’s added noise for the force sensor to operate in the thermal-limited regime Sζζdet≪Sζζcant(Ωm)S_{\zeta\zeta}^{\mathrm{det}}\ll S_{\zeta\zeta}^{\mathrm{cant}}(\Omega_m), where Fmin⁡≃2kBTηΔfF_{\min}\simeq\sqrt{2k_BT\eta\Delta f}. Achieving this limit can be especially challenging at cryogenic temperatures.

Mamin et al. achieved a record force sensitivity of 0.82 aN/Hz\sqrt{\mathrm{Hz}} using a cantilever with Ωm=5\Omega_m=5 kHz and stiffness k=260 μN/mk=260\,\mu\mathrm{N}/\mathrm{m} at T=110T=110 mK [15]. AFM usually requires stiffness on the order of 100 N/m to avoid the tip jumping to contact and sticking to the surface [16]. Such stiffness would necessitate five orders of magnitude improvement in detector noise to achieve an equivalent force sensitivity. Reduction of detection noise is therefore critical for improving low temperature AFM.

AFM sensor and detection principle

The advantages of cavity optomechanical detection for AFM sensors were already showcased by Liu et al. [7], who implemented a whispering gallery mode toroidal optical cavity coupled through its evanescent field to a doubly-clamped beam with a tip in the middle. We replace the optical cavity with a superconducting microwave resonant circuit which is comparatively easy to fabricate and to integrate on an AFM cantilever with a tip at the free end. Figure 1(a)–(Figure 1c) shows a scanning electron micrograph of the probe. The triangular cantilever is a Si-N plate released from a Si substrate. The microwave superconducting lumped-element resonant circuit consists of an inter-digital capacitor in series with a meandering nanowire inductor, both fabricated from a single layer of Nb-Ti-N deposited on the Si-N. The nanowire meanders across the clamping edge of the plate, giving rise to a modulation of its kinetic inductance due to surface strain generated by cantilever bending. A Pt-C tip with curvature radius <10< 10 nm at its apex is formed at the free end of the cantilever through a series of electron-beam-assisted depositions.

Scanning electron micrographs of the probe and schematic of the experimental setup

Figure 1. (a)–(c) Scanning electron micrograph of the probe, featuring a Si-N triangular cantilever released from the Si substrate, hosting an integrated lumped-element LC resonator realized from a single layer of Nb-Ti-N and a Pt-C tip deposited at the cantilever apex. (d) Experimental setup. The scanner box is suspended from the 10 mK stage with soft springs. Folded copper braids provide thermalization and additional mechanical damping. The digital synthesizer and analyzer (Presto platform by Intermodulation Products AB) provides the DC output to control the scanner, the radio-frequency drive for the piezo shaker and the microwave multifrequency lock-in modulation/demodulation.

Figure 1(d) shows a schematic of the experimental setup implementing a standard low-noise cryogenic microwave reflection measurement of the sensor, which is mounted in a prototype AFM suspended with three springs from the 10 mK plate of the DR. The microscope consists of a metallic box which supports the scanner and a printed-circuit board to which the sensor is mounted. This custom board also hosts a piezoelectric inertial actuator (shaker) which drives the cantilever oscillation. A test surface is mounted on top of the open-loop scanner stack consisting of a zz-positioner for coarse approach and an open loop xyzxyz-scanner for fine positioning (Attocube anpz 102/LT/HV, ansxyz 100/LT/HV). Twisted pairs and coaxial cables are coiled to make springy electrical connections to the AFM. Thermal anchoring of the various parts is achieved by copper braids with multiple zig-zag folds. At room temperature the resonance frequency of the AFM’s suspension system is roughly 1 Hz, with a quality factor of roughly 2.

Mechanical oscillation of the tip causes phase modulation of the reflected microwave pump, detected as motional sidebands in the signal spectrum. Measuring the microwave response at a sideband, the detection responsivity α\alpha [V/m] can be expressed as

α=Z0G2(Pinℏωpκexκ2/4+Δ2)κex(κ/2)2+(Δ+Ω)2=Z0G2ncκex(κ/2)2+(Δ+Ω)2(5)\alpha= \sqrt{Z_{0}G^{2}\left(\frac{P_{\mathrm{in}}}{\hbar\omega_{\mathrm{p}}}\frac{\kappa_{\mathrm{ex}}}{\kappa^{2}/4+\Delta^{2}}\right)\frac{\kappa_{\mathrm{ex}}}{(\kappa/2)^{2}+(\Delta+\Omega)^{2}}} = \sqrt{Z_{0}G^{2}n_{\mathrm{c}}\frac{\kappa_{\mathrm{ex}}}{(\kappa/2)^{2}+(\Delta+\Omega)^{2}}} \tag*{(5)}

where G=∂ω0/∂ζG = \partial\omega_{0}/\partial\zeta [Hz/m] is the electromechanical coupling coefficient, κex\kappa_{\mathrm{ex}} the external loss rate and κ\kappa the total loss rate of the microwave resonance, and Z0Z_{0} the characteristic impedance of the transmission line. The term in the first parenthesis is the circulating power or energy stored in the cavity, expressed as the intracavity photon number ncn_{\mathrm{c}}. We control ncn_{\mathrm{c}} through the drive power of the pump PinP_{\mathrm{in}} and the detuning from the cavity resonance frequency Δ=ωp−ω0\Delta= \omega_{\mathrm{p}}-\omega_{0}. In principle, arbitrarily large responsivity α\alpha is achieved for arbitrary weak coupling GG by increasing PinP_{\mathrm{in}}. In practice, nonlinear effects emerge at large ncn_{\mathrm{c}}, such as heating (nonlinear loss) and resonance-frequency shift (Kerr nonlinearity), which cause the linear response picture of cavity optomechanics to break down.

Detection noise in the cavity optomechanical scheme is determined by quantum fluctuations of ncn_{\mathrm{c}}, or photon shot noise. Using eq. (5) for the responsivity α\alpha we can convert this detection noise to an equivalent mechanical displacement noise or imprecision noise, inversely proportional to ncn_{\mathrm{c}},

Sζζimp(Ω)=ζzpf24g02nc(κ2/4+(Ω+Δ)2κex).(6)S_{\zeta\zeta}^{\mathrm{imp}}(\Omega) = \frac{\zeta_{\mathrm{zpf}}^{2}}{4g_{0}^{2}n_{\mathrm{c}}}\left(\frac{\kappa^{2}/4+(\Omega+\Delta)^{2}}{\kappa_{\mathrm{ex}}}\right). \tag*{(6)}

Here we introduce the single photon coupling rate g0=Gζzpfg_{0} = G\zeta_{\mathrm{zpf}}, corresponding to the microwave resonance frequency shift associated with the quantum zero-point motion of the mechanical mode ζzpf=ℏ/2meffΩm\zeta_{\mathrm{zpf}} = \sqrt{\hbar/2m_{\mathrm{eff}}\Omega_{\mathrm{m}}}. The cavity optomechanical measurement also introduces a fluctuating radiation pressure force on the test mass or back-action noise, proportional to ncn_c [17, 18].

SξξBA(Ω)=ℏ2g02ncxzpf2(κexκ2/4+(Ω+Δ)2)2∣χ(Ω)∣2.(7)S_{\xi\xi}^{\mathrm{BA}}(\Omega)=\frac{\hbar^2 g_0^2 n_c}{x_{\mathrm{zpf}}^2}\left(\frac{\kappa_{\mathrm{ex}}}{\kappa^2/4+(\Omega+\Delta)^2}\right)^2\lvert\chi(\Omega)\rvert^2 . \tag*{(7)}

When the microwave pump frequency is detuned by Δ=−Ωm\Delta=-\Omega_{\mathrm{m}}, and the mechanical mode is measured on resonance Ω=Ωm\Omega=\Omega_{\mathrm{m}} where ∣χ(Ωm)∣=2xzpf2/ℏΓ\lvert\chi(\Omega_{\mathrm{m}})\rvert=2x_{\mathrm{zpf}}^2/\hbar\Gamma, detector noise is minimized when Sξξimp=SξξBAS_{\xi\xi}^{\mathrm{imp}}=S_{\xi\xi}^{\mathrm{BA}}, where

nc=ncSQL=Γ4g02κ2/4κex.(8)n_c=n_c^{\mathrm{SQL}}=\frac{\Gamma}{4g_0^2}\frac{\kappa^2/4}{\kappa_{\mathrm{ex}}}. \tag*{(8)}

If the mechanical oscillator is cooled to a very low temperature, so that it is in its quantum ground state, a measurement at ncSQLn_c^{\mathrm{SQL}} gives SξξSQL(Ωm)=2xzpf2/ΓS_{\xi\xi}^{\mathrm{SQL}}(\Omega_{\mathrm{m}})=2x_{\mathrm{zpf}}^2/\Gamma, the so-called standard quantum limit (SQL). In an actual experiment components after the cavity in the measurement chain introduce additional imprecision noise, expressed as an added number of noise photons naddn_{\mathrm{add}}

Sξξdet=Sξξimp(1+2nadd)+SξξBA.(9)S_{\xi\xi}^{\mathrm{det}}=S_{\xi\xi}^{\mathrm{imp}}(1+2n_{\mathrm{add}})+S_{\xi\xi}^{\mathrm{BA}} . \tag*{(9)}

The added noise in our setup [Figure 1(d)] is determined by the cryogenic low-noise amplifier (LNA) mounted at the 4 K stage. We use Planck spectroscopy to calibrate our LNA, where we measure nadd≈14n_{\mathrm{add}}\approx14 at 4.5 GHz [19, 20]. The addition of a near quantum-limited amplifier (e.g. a Josephson parametric amplifier) as the first stage in the amplification chain, where nadd≈1/2n_{\mathrm{add}}\approx1/2, is a commonly adopted solution to reduce detection noise [21].

Sensor characterization

Table 1 summarizes the device parameters for three different probes employed in this work, measured at 10 mK in free space far from the surface.

ProbeΩm\Omega_{\mathrm{m}} (MHz)Γ\Gamma (Hz)meffm_{\mathrm{eff}} (pg)ω0\omega_{0} (GHz)κ\kappa (MHz)κex\kappa_{\mathrm{ex}} (MHz)GG (kHz/nm)
15.66915474.273.383.0424.92
24.9635614.641.030.6424.56
35.88145444.404.544.0428.98

Table 1.

We characterize the sensor by measuring the total noise of the undriven mechanical resonance, up-converted to the microwave spectrum. Figure 2(a) shows the data for Probe 1, obtained by convertTable 1: Device parameters Ωm\Omega_{\mathrm{m}} and Γ\Gamma are determined by fitting the measured mechanical resonance. ω0\omega_{0}, κ\kappa and κex\kappa_{\mathrm{ex}} are determined by fitting to the measured microwave resonance, while meffm_{\mathrm{eff}} and GG are determined from FEM simulations [14].

Detector output noise equivalent displacement power spectral density and noise contributions

Figure 2. (a) Detector output noise equivalent displacement PSD and individual contributions. (b) Noise contributions evaluated at the mechanical resonance as a function of the average circulating photon number ncn_{\mathrm{c}}. The detector noise (solid red line) accounts for the intracavity ground state fluctuations (dashed red line), back action noise (dash-dot red line) and the added noise of the first stage of amplification. The horizontal line marks the fluctuations of the cantilever at its effective mode temperature. The solid and dashed black curves describe the total noise of the sensor for various cantilever mode temperatures.

ing the measured upper motional sideband to an equivalent displacement PSD SζζtotS_{\zeta\zeta}^{\mathrm{tot}} through eq. (5) with Δ=−Ωm\Delta=-\Omega_{\mathrm{m}} and nc=1.98×106n_{\mathrm{c}}=1.98\times10^{6}. Fitting eq. (2) to the measured data we extract the detector (red) and cantilever (orange) contributions to the total noise. Through this fitting procedure we determine an effective temperature of the cantilever T=0.74 KT=0.74\ \mathrm{K}, which is substantially higher than the bath temperature Tbath=10 mKT_{\mathrm{bath}}=10\ \mathrm{mK}. The discrepancy is most likely due to inertial actuation by noise from the piezo shaker and decoupling of the cantilever eigenmode from the thermal bath.

Figure 2(b) shows various contributions to the noise of Probe 1 as a function of ncn_{\mathrm{c}}. The light blue area marks the range of operation of this sensor, where Kerr nonlinear response of the superconducting resonant circuit [22] limits nc<2.0×106n_c < 2.0 \times10^{6}, well below ncSQL≈1.4×108n_c^{\mathrm{SQL}} \approx1.4 \times10^{8}. It is possible to achieve operation closer to ncSQLn_c^{\mathrm{SQL}} by improving the sensor design.

For a given g0g_0 and Γ\Gamma we can decrease ncSQLn_c^{\mathrm{SQL}} by reducing the external loss rate κex\kappa_{\mathrm{ex}} and the total loss rate κ\kappa. We can control κex\kappa_{\mathrm{ex}} through design of the resonator coupling to the readout transmission line [14]. Decreasing κ\kappa also requires improvement on the internal loss rate of the cavity κ0=κ−κex\kappa_0 = \kappa-\kappa_{\mathrm{ex}}.

With typical parameters for our sensors κex/2π=1\kappa_{\mathrm{ex}}/2\pi= 1 MHz, Γ/2π=10\Gamma/2\pi= 10 Hz, g0/2π=0.14g_0/2\pi= 0.14 Hz, we find a value of ncSQL≈31.8×106n_c^{\mathrm{SQL}} \approx31.8 \times10^{6} photons, which would require a factor 15 improvement on κ0\kappa_0.

Achieving this improvement is quite reasonable as our microwave resonators exhibit a κ0\kappa_0 far larger than the typical values reported in the literature [23, 24, 25]. The main limitation to the internal quality factor of our microwave resonators is the highly-disordered low-stress Si-N layer, which causes significant dielectric losses [26].

As discussed above, the figure of merit of a displacement detector should be evaluated with respect to the fluctuations of the specific mechanical resonator that is measuring force. In this regard, it is informative to define the crossover temperature T∗T^*, or effective mode temperature of the mechanical resonator, at which Sζζcant(Ωm)=SζζdetS^{\mathrm{cant}}_{\zeta\zeta}(\Omega_{\mathrm{m}}) = S^{\mathrm{det}}_{\zeta\zeta}. When the mechanical mode temperature T>T∗T > T^*, the noise is dominated by mechanical displacement noise and the force measurement is at the thermal limit. When T<T∗T < T^*, the measurement is dominated by detector noise and improvements to displacement detection will improve force sensitivity.

Piezoelectric sensors with transimpedance amplifier detectors as reported by Giessibl et. al. [27] exhibit Sζζdet=62S^{\mathrm{det}}_{\zeta\zeta} = 62 fm/Hz\sqrt{\mathrm{Hz}} giving T∗=10T^* = 10 K for tuning forks, and Sζζdet=1.89S^{\mathrm{det}}_{\zeta\zeta} = 1.89 fm/Hz\sqrt{\mathrm{Hz}}, giving T∗=21T^* = 21 K for length-extensional resonators. Optimized optical-beam deflection detectors as reported by Fukuma et al. [28] exhibit Sζζdet=17S^{\mathrm{det}}_{\zeta\zeta} = 17 fm/Hz\sqrt{\mathrm{Hz}}, T∗=1.2T^* = 1.2 K. For comparison, from measurements on Probe 1 shown in Figure 2(a), we obtain a lower crossover temperature T∗=0.43T^* = 0.43 K, despite the fact that the sensor operates below its optimal point.

AFM operation

The noise-to-noise ratio Sζζcant/Sζζdet=T/T∗S^{\mathrm{cant}}_{\zeta\zeta}/S^{\mathrm{det}}_{\zeta\zeta}=T/T^{*} provides a qualitative measure of the pixel acquisition rate, or measurement bandwidth Δf\Delta f, above which force sensitivity is degraded by detector noise.

For FM-AFM this bandwidth is Δf=Γ(T/T∗)\Delta f=\Gamma(T/T^{*}) where as AM-AFM requires Δf≤Γ\Delta f\le\Gamma. Force sensitivity is however only one aspect of a high-performance AFM. Ultimately the image speed and resolution, both vertical and lateral, are influenced by many factors: equilibrium tip-surface distance, oscillation amplitude of the driven cantilever, sharpness of the tip, and pickup of environmental vibrations and noise, such as noise from the actuator driving the cantilever oscillation as well as feedback noise and sample topography.

We perform initial positioning of the microscope by driving the coarse zz-positioner with the Attocube ANC-300 control unit that provides the 60 V saw-tooth pulses to the piezo actuators. The stick-slip motion of the actuator under high-voltage saw-tooth pulses (≈50\approx50 nm/step) produces significant heating at the 10 mK stage. Under continuous operation, the heating power is of the order of 1 μ\muW for an approach speed of 1 μ\mum/min. This temperature change during coarse approach causes drift in the cantilever’s resonance frequency, requiring a couple of minutes to restore a stable configuration. Therefore, during coarse approach we monitor only the microwave resonance.

When approaching the sample surface we observe a shift (<2< 2%) of the microwave resonance frequency to lower frequency and a slight increase of the internal losses, as shown in Figure 3(a)–(Figure 3d). This red-shift suggests that proximity to the surface introduces a lossy capacitive load to the microwave resonant circuit. While slightly degrading the performance of the displacement detector, this effect has the advantage of providing an easily measurable indication of probe-surface separation during coarse approach. For the final stage of the approach, we drive the zz-scanner with a DC-voltage (100 nm/V) while monitoring both the microwave and cantilever resonances. Figure 3(e)–(Figure 3f) show the measured frequency shift versus distance, demonstrating force gradient sensing by the KIMEC AFM probe.

Magnitude and phase of the microwave resonator’s reflection coefficient, resonance frequency and loss rate, mechanical susceptibility, and resonance frequency shift

Figure 3. Magnitude (a) and phase (b) of the microwave resonator’s reflection coefficient S11S_{11} measured during coarse approach. Proximity to the surface causes a non-negligible shift in the microwave resonance frequency (c) and increase in the total loss rate (d). (e) Effect of force gradient on the mechanical susceptibility of the cantilever close to the surface. (f) Mechanical resonance frequency shift vs. tip-surface distance curve.

We tested the AFM by imaging a nano-fabricated calibration grating over a 1 μ\mum ×\times 1 μ\mum scan area with a resolution of 100 pixels/line. In order to track the surface topography while scanning, we developed a proportional-integral-derivative (PID) control firmware running on the embedded real-time processing unit (RPU) of the microwave multifrequency lock-in amplifier (Presto by Intermodulation Products AB). The code is released as open source under the MIT license, and is available online in a repository hosted on GitHub [29]. The feedback system must be able to follow rapid changes in topography while avoiding feedback oscillation. Depending on the mode of operation, AM-AFM or FM-AFM, the PID setpoint and error signal are determined from the amplitude or phase of the motional sideband which is the up-converted cantilever oscillation.

The PID control is developed in the Rust programming language, and features low-pass filtering of the derivative action, as well as clamping of the integrator state to prevent integral windup and of the output control to prevent damage to the AFM probe.

We operate Probe 2 in AM-AFM mode by driving the cantilever slightly above resonance Ω=Ωm+Γ/2\Omega= \Omega_{\mathrm{m}} + \Gamma/2. The PID loop controls the zz-extension of the scanner to keep oscillation amplitude constant at about 200 pm. The blue-detuned cantilever drive ensures that the PI controller retracts the zz-scanner in the event of a sudden drop in the measured signal, thus avoiding crashing the probe into the surface. The image in Figure 4(a) is acquired with a measurement bandwidth of Δf=1.5\Delta f = 1.5 Hz <Γ< \Gamma, corresponding to a scan rate of roughly 67 s/line. The choice of the integral gain is crucial for optimal surface tracking. Too large values of integral gain result in an overly responsive feedback that introduces noise on the topographic image and feedback oscillations. Too low values can lead to instability or loss of surface tracking with abrupt changes in topography. The latter occurred during the scan; in multiple instances the feedback temporarily loses track of the surface immediately after the tip encounters a step. The single scan line on the fast, i.e. xx-axis, shown in Figure 4(b)–(Figure 4c) clearly shows different behavior of trace and retrace scan directions.

AM-AFM imaging of the calibration grating and scan lines

Figure 4. (a) AM-AFM imaging of the calibration grating over a 1 μ\mum ×\times 1 μ\mum scan area containing a single vertical stripe with a 10 nm step height. The dashed lines in the image mark the location of the scan lines in (b)–(c).

We achieved faster imaging and improved stability with FM-AFM as shown in Figure 5. FM-AFM requires phase-sensitive detection of tip displacement using a scheme described in detail in earlier publications [13, 30]. The cantilever is now driven on resonance at the same ≃\simeq 200 pm oscillation amplitude while the microwave resonator is driven with two pump tones of equal intensity and symmetrically detuned with Δ=±Ωm\Delta= \pm\Omega_{\mathrm{m}}. The lower and upper motional sidebands from the two pumps interfere at ω0\omega_0, resulting in microwave response amplitude which depends on the phase of the mechanical motion [31, 32]. The feedback controls the zz-extension to maintain a constant microwave response amplitude, or constant mechanical response phase. The tracking of the surface is more consistent using this FM-AFM feedback scheme as shown in Figure 5(b)–(Figure 5c) by the trace and retrace signals. Figure 5(d)–(Figure 5f) shows the effect of the measurement bandwidth on a single scan line.

FM-AFM imaging and scan-line measurements

Figure 5. (a) FM-AFM imaging (trace) of the calibration grating over a 1 μm×0.8 μm\mu\mathrm{m} \times0.8\,\mu\mathrm{m} scan area. (b)–(c) Average and standard deviation over ten consecutive acquisitions of a single scan line at constant yy-position in both trace (b) and retrace (c) scan direction. (d)–(f) Single scan line measured with different pixel acquisition rates Δf\Delta f.

Finally, we tested the lateral resolution of our microscope by imaging a less trivial structure. We fabricated a second test sample by etching a pattern from a 10 nm Ti thin film deposited on a 5 nm Au layer on a silicon substrate. Each unit cell in the pattern contains different shapes as shown in the SEM image in Figure 6(a). The tapered arms of crosses and ribbons shrink to reach a nominal minimum feature size of 20 nm. Figure 6(b) shows a FM-AFM image of the 1 μm×1 μm1\,\mu\mathrm{m} \times1\,\mu\mathrm{m} scan area corresponding to the inset of Figure 6(a) acquired using Probe 3.

SEM image, FM-AFM images, Fourier transform, and scan line

Figure 6. (a) SEM image of the second test sample. The zoomed inset highlights the area imaged with FM-AFM in (b). (c) Fast Fourier transform of the error signal for a single scan line. The xx-axis of the plot is converted to frequency through the pixel acquisition rate Δf=112\Delta f = 112 Hz. (d) Single scan line showing the rippling effect caused by pulse-tube vibrations. The orange curve is obtained by applying a low-pass filter with cutoff frequency 0.75 Hz. (e) Effect of filtering and smoothing on the acquired image.

Although the various structures can be identified, several scanning artifacts appear in the image. In this particular experiment the probe was especially sensitive to pulse-tube vibrations. Figure 6(c) shows the fast-Fourier transform (FFT) of the error signal for a single horizontal (fast axis) line scan. The component at 0.8 Hz and its higher harmonics coincide with periodicity of the pulse-tube, resulting in ripples in the height image (feedback signal) as shown in Figure 6(d). In this image we also see distortion and misalignment between consecutive scan lines due to hysteresis and drift in the open loop x−yx-y scanners. These problems are more pronounced when attempting to resolve smaller features, and image processing techniques such as filtering and smoothing are not particularly effective in enhancing resolution, as shown by Figure 6(e).

Conclusion

We have demonstrated ultra-low-temperature AFM operation with a kinetic-inductive electromechanical (KIMEC) force sensor. In comparison with state-of-the-art low-temperature AFM, these initial AFM images are not of particularly high resolution. However, our AFM was built in to a standard pulse-tube dilution refrigerator which was not designed for preparation of pristine surfaces, nor was any attempt made to sharpen the tip in-situ.

Our analysis of force sensitivity showed that residual noise from the actuator and decoupling of the cantilever eigenmode from the thermal bath limited the sensor’s effective temperature to 0.74 K, well above the bath temperature of 10 mK. However, our analysis shows that the KIMEC principle of detecting cantilever deflection, as it is implemented in these sensors, is capable of operating at the thermal limit of force sensitivity down to a temperature of 0.43 K. This is a significant improvement over much lower frequency piezoelectric transduction methods commonly used for low-temperature AFM, which can work at the thermal limit down to some 10 K. We estimate that, with reasonable improvements on the current KIMEC sensor design, this detection principle will be capable of operating at the fundamental limit of minimum added noise. Achieving this performance limit is a prerequisite for improved force sensitivity with ultra-low temperature and ultra-low damping mechanical modes. Alternatively, the lower noise of the KIMEC detection principle allows for AFM imaging with increased pixel acquisition rate (measurement bandwidth) without degrading force sensitivity.

Acknowledgements

We acknowledge helpful discussions with Thilo Glatzel.

Funding

The European Union Horizon 2020 Future and Emerging Technologies (FET) Grant Agreement No. 828966 — QAFM and the Swedish SSF Grant No. ITM17–0343 supported this work.

Data availability

Data generated and analyzed during this study is openly available in Zenodo at 10.5281/zenodo.15744435.

References

  1. [1]Metcalfe, M. Applied Physics Reviews 2014, 1 (3), .
  2. [2]Krause, A. G.; Winger, M.; Blasius, T. D.; Lin, Q.; Painter, O. Nature Photonics 2012, 6 (11), 768–772.
  3. [3]Pikovski, I.; Vanner, M. R.; Aspelmeyer, M.; Kim, M.; Brukner, Č. Nature Physics 2012, 8 (5), 393–397.
  4. [4]Marletto, C.; Vedral, V. Physical review letters 2017, 119 (24), 240402.
  5. [5]Krisnanda, T.; Zuppardo, M.; Paternostro, M.; Paterek, T. Physical review letters 2017, 119 (12), 120402.
  6. [6]Hälg, D.; Gisler, T.; Tsaturyan, Y.; Catalini, L.; Grob, U.; Krass, M.-D.; Héritier, M.; Mattiat, H.; Thamm, A.-K.; Schirhagl, R. et al. Physical Review Applied 2021, 15 (2), L021001.
  7. [7]Liu, Y.; Miao, H.; Aksyuk, V.; Srinivasan, K. Optics express 2012, 20 (16), 18268–18280.
  8. [8]Forstner, S.; Sheridan, E.; Knittel, J.; Humphreys, C. L.; Brawley, G. A.; Rubinsztein-Dunlop, H.; Bowen, W. P. Advanced materials (Deerfield Beach, Fla.) 2014, 26 (36), 6348–6353.
  9. [9]Hill, J. T.; Safavi-Naeini, A. H.; Chan, J.; Painter, O. Nature communications 2012, 3 (1), 1196.
  10. [10]Braginskiĭ, V. B.; Vorontsov, Y. I. Soviet Physics Uspekhi 1975, 17 (5), 644.
  11. [11]Smith, D. Review of scientific instruments 1995, 66 (5), 3191–3195.
  12. [12]Aspelmeyer, M.; Kippenberg, T. J.; Marquardt, F. Reviews of Modern Physics 2014, 86 (4), 1391.
  13. [13]Roos, A. K.; Scarano, E.; Arvidsson, E. K.; Holmgren, E.; Haviland, D. B. Physical Review Applied 2023, 20 (2), 024022.
  14. [14]Roos, A. K.; Scarano, E.; Arvidsson, E. K.; Holmgren, E.; Haviland, D. B. Beilstein Journal of Nanotechnology 2024, 15 (1), 242–255.
  15. [15]Mamin, H.; Rugar, D. Applied Physics Letters 2001, 79 (20), 3358–3360.
  16. [16]Garcia, R.; Perez, R. Surface science reports 2002, 47 (6-8), 197–301.
  17. [17]Bowen, W. P.; Milburn, G. J. Quantum optomechanics; CRC press, 2015.DOI
  18. [18]Schliesser, A.; Arcizet, O.; Rivière, R.; Anetsberger, G.; Kippenberg, T. J. Nature Physics 2009, 5 (7), 509–514.
  19. [19]Jolin, S. W.; Andersson, G.; Hernández, J. R.; Strandberg, I.; Quijandría, F.; Aumentado, J.; Borgani, R.; Tholén, M. O.; Haviland, D. B. Physical Review Letters 2023, 130 (12), 120601.
  20. [20]Lingua, F.; Rivera Hernández, J. C.; Cortinovis, M.; Haviland, D. B. Phys. Rev. Lett. 2025, 134, 183602. doi:10.1103/PhysRevLett.134.183602.DOI
  21. [21]Teufel, J. D.; Donner, T.; Li, D.; Harlow, J. W.; Allman, M. S.; Cicak, K.; Sirois, A. J.; Whittaker, J. D.; Lehnert, K. W.; Simmonds, R. W. Nature 2011, 475 (7356), 359–363. doi:10.1038/nature10261.DOI
  22. [22]Scarano, E.; Arvidsson, E. K.; Roos, A. K.; Holmgren, E.; Haviland, D. B. Applied Physics Letters 2024, 124 (24), .
  23. [23]Samkharadze, N.; Bruno, A.; Scarlino, P.; Zheng, G.; DiVincenzo, D.; DiCarlo, L.; Vander-sypen, L. Physical Review Applied 2016, 5 (4), 044004.
  24. [24]Kroll, J. G.; Borsoi, F.; Van Der Enden, K.; Uilhoorn, W.; De Jong, D.; Quintero-Pérez, M.; Van Woerkom, D.; Bruno, A.; Plissard, S.; Car, D. et al. Physical Review Applied 2019, 11 (6), 064053.
  25. [25]Müller, M.; Luschmann, T.; Faltermeier, A.; Weichselbaumer, S.; Koch, L.; Huber, G. B.; Schumacher, H. W.; Ubbelohde, N.; Reifert, D.; Scheller, T. et al. Materials for Quantum Technology 2022, 2 (1), 015002.
  26. [26]Mittal, S.; Adachi, K.; Frattini, N. E.; Urmey, M. D.; Lin, S.-X.; Emser, A. L.; Metzger, C.; Talamo, L.; Dickson, S.; Carlson, D. et al. Physical Review Applied 2024, 21 (5), 054044.
  27. [27]Giessibl, F. J.; Pielmeier, F.; Eguchi, T.; An, T.; Hasegawa, Y. Physical Review B—Condensed Matter and Materials Physics 2011, 84 (12), 125409.
  28. [28]Fukuma, T.; Kimura, M.; Kobayashi, K.; Matsushige, K.; Yamada, H. Review of Scientific Instruments 2005, 76 (5), .
  29. [29]Intermodulation Products AB, rpu-qafm GitHub repository. https://github.com/intermod-pro/rpu-qafm. Accessed: 2024-10-08
  30. [30]Arvidsson, E. K.; Scarano, E.; Roos, A. K.; Qvarfort, S.; Haviland, D. B. Physical Review A 2024, 110 (4), 043524.
  31. [31]Suh, J.; Weinstein, A. J.; Lei, C. U.; Wollman, E. E.; Steinke, S. K.; Meystre, P.; Clerk, A. A.; Schwab, K. C. Science 2014, 344 (6189), 1262–1265. doi:10.1126/science.1253258.DOI
  32. [32]Braginsky, V. B.; Vorontsov, Y. I.; Thorne, K. S. Science 1980, 209 (4456), 547–557. doi:10.1126/science.209.4456.547.DOI

Paper details

vlnc:W_yaqmhepx
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 7

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents