betaFeedback

WorksPhysics and Astronomy

Calibrated Microwave Reflectance in Low-Temperature Scanning Tunneling Microscopy

Abstract

We outline calibrated measurements of the microwave reflection coefficient from the tunnel junction of an ultra-high vacuum low temperature scanning tunneling microscope. The microwave circuit design is described in detail, including an interferometer for enhanced signal-to-noise and a demodulation scheme for lock-in detection. A quantitative, in-situ procedure for impedance calibration based on the numerical 3-error-term model is presented. Our procedure exploits the response of the microwave reflection signal due to the change of the tunneling conductance caused by sub-nm variation of the tunneling distance. Experimental calibration is achieved by a least-squares numerical fit of simultaneously measured conductance and microwave reflection retraction curves at finite conductance. Our method paves the way for nanoscale microscopy and spectroscopy of dielectric surface properties at GHz frequencies and cryogenic temperatures. This opens a promising pathway even for dielectric fingerprinting at the single molecule limit.

Introduction

Recent developments of near-field scanning microwave microscopy[1] have enabled the characterization of materials with nanoscale dimensions, surfaces, and nanoelectronics devices[2] up to GHz frequencies and with sub-micron lateral resolution. In particular, this includes measuring electrostatic force[3, 4], electrical capacitance[5, 6] and electrochemical activity[7]. Each of these techniques has its specific limitations, such as the frequency of operation[6, 8, 9]. Recently, microwave reflection spectroscopy with sub-micron lateral resolution has been demonstrated up to 19 GHz[10, 11]. Various material properties, including relative dielectric permittivity[12] and dopant concentration[2] have been successfully detected by a transmission line implementation realized in an atomic force microscope operated at ambient conditions. Although closely related to the atomic force microscope, the scanning tunneling microscope (STM) has been scarcely demonstrated as platform for calibrated microwave reflection spectroscopy to date. The present work outlines a straightforward approach, i.e. utilizing the STM tip as a probe for calibrated microwave reflection spectroscopy. In particular, we employ a low-temperature STM instrument, which is well-known to provide ultimate stability and spatial resolution[13] by operating at ultra-high vacuum conditions and cryogenic temperatures. Our approach offers great potential for achieving microwave impedance spectroscopic fingerprinting even at the scale of single molecules – thus complementing related methods that utilize spin-resonance[14, 15] or dielectric relaxation[16] at GHz frequencies. A major advantage of STM over similar AFM based techniques is the ability to select the bias and current setpoint at which the microwave reflection spectroscopy is performed. This opens the door to, for example, studying transient molecular redox reactions[7].

Commonly, the tunnel junction of an STM exhibits an impedance ZjunZ_{\mathrm{jun}} of 10810^{8} to 10910^{9} Ω. In contrast, the impedance of the microwave transmission line built from standard microwave components is Zref=50Z_{\mathrm{ref}} = 50 Ω. The interface of the tunnel junction to the microwave transmission line thus represents a massive change of impedance, which affects the degree of reflection of an incoming microwave at the junction. The latter is described by the complex-valued scattering parameter[17]:

S11,jun=Zjun−ZrefZjun+Zref(1)S_{11,\mathrm{jun}} = \frac{Z_{\mathrm{jun}} - Z_{\mathrm{ref}}}{Z_{\mathrm{jun}} + Z_{\mathrm{ref}}} \tag*{(1)}

Accordingly, the STM junction will almost completely reflect incoming microwaves, i.e. S11≈1S_{11} \approx1. Notice that S11S_{11} is frequency-dependent, since the complex impedance of the tunnel junction depends on the frequency ff of the microwave as well as on the conductance GG and capacitance CC of the junction[17]:

Zjun=1Yjun=1G+i 2πfC(2)Z_{\mathrm{jun}} = \frac{1}{Y_{\mathrm{jun}}} = \frac{1}{G + i\,2\pi f C} \tag*{(2)}

where YjunY_{\mathrm{jun}} is the complex-valued admittance and i=−1i = \sqrt{-1} is the imaginary unit. Setting well-defined values for ZrefZ_{\mathrm{ref}} and ff leaves S11S_{11} dependent on GG and CC of the junction only. This dependency lies at the heart of microwave reflectance spectroscopy implemented on a scanning probe instrument[18, 19], relying on the accurate quantitative determination of CC from measured S11S_{11} data. Such quantitative determination requires accurate calibration, because S11,junS_{11,\mathrm{jun}} cannot be measured directly. Instead, the experimental measurement yields the reflection coefficient S11,m≠S11,junS_{11,\mathrm{m}} \ne S_{11,\mathrm{jun}}. Notice that the subscript m indicates that S11S_{11} is measured at the so-called measurement plane. S11,junS_{11,\mathrm{jun}} refers to the value of S11S_{11} at the tunnel junction, also known as the calibration plane. The measurement plane is separated from the calibration plane by various circuit components and the transmission line. The amplitude and phase of the complex-valued S11,mS_{11,\mathrm{m}} systematically deviate from S11,junS_{11,\mathrm{jun}} due to microwave absorptions and reflections in the transmission line. Even though the deviation of S11,mS_{11,\mathrm{m}} from S11,junS_{11,\mathrm{jun}} is systematic, due to the complex nature of the transmission line it is not practical to use an analytic model to describe the relationship between S11,mS_{11,\mathrm{m}} and S11,junS_{11,\mathrm{jun}}. The well-known numerical 3-error-term model[10, 20, 21] relates S11,junS_{11,\mathrm{jun}} to S11,mS_{11,\mathrm{m}} via three complex-valued parameters e00e_{00}, e01e_{01}, and e11e_{11}:

S11,m=e00+e01S11,jun1−e11S11,jun(3)S_{11,\mathrm{m}} = e_{00} + e_{01}\frac{S_{11,\mathrm{jun}}}{1 - e_{11}S_{11,\mathrm{jun}}} \tag*{(3)}

Importantly, the phase and amplitude of S11S_{11} depend strongly on the exact frequency chosen. Therefore, the ee-parameters have to be determined separately for every measurement frequency.

Quantitative GG and CC values can only be derived from S11,mS_{11,\mathrm{m}} if the values of the ee-parameters are precisely known. The ee-parameters can be obtained from an in-situ retraction curve calibration. Retraction curve calibration requires that both S11,m(z)S_{11,\mathrm{m}}(z) and S11,jun(z)S_{11,\mathrm{jun}}(z) are known [zz is the tip-sample separation, see Figure 1(a)]. In an early approach by Farina et al. 21, C(z)C(z) was calculated using an assumed analytic function with absolute tip-sample distance and ex-situ measured tip-radius as input parameters. In AFM, a recent retraction calibration procedure relies on simultaneously measuring ∂C(z)/∂z\partial C(z)/\partial z and S11,m(z)S_{11,\mathrm{m}}(z) during tip retraction[10]. Using Eqs. (1) and (2) and the assumption that G=0G = 0, S11,jun(z)S_{11,\mathrm{jun}}(z) is calculated from the independently measured C(z)C(z). Inserting S11,jun(z)S_{11,\mathrm{jun}}(z) and S11,m(z)S_{11,\mathrm{m}}(z) into Eq. (3) allows to determine the three ee-parameters via a least-squares fit routine. In STM, the force detection required to measure ∂C(z)/∂z\partial C(z)/\partial z is not available and an alternate strategy should be used.

Image not extracted

Figure 1. The experimental set-up. (a) Schematic overview of the electronic circuit. Low frequency components for STM operation and GG detection are outlined in grey. Microwave specific components are outlined in black. The microwave voltage is added to VDCV_{\mathrm{DC}} using a bias-tee (colored yellow). (b) Block diagram of the microwave interferometer circuit (shaded orange) and the down-conversion circuit (shaded green). Pink and cyan arrows indicate the measurement and difference paths of the microwaves through the interferometer, respectively. Specific ports of the components are labeled. Black lines indicate connections via a single adapter piece, blue lines are dedicated microwave cables, and red lines are low-noise BNC cables. For abbreviations and details of the components, see Table I.

In this work, we present a retraction calibration procedure that uses simultaneous G(z)G(z) and S11(z)S_{11}(z) measurements upon the well-controlled variation of the tunnel distance zz. We determine the numerical values of the ee-parameters of Eq. (3) that define the calibration. Additionally, we present our implementation of microwave reflection spectroscopy in a low-temperature ultra-high vacuum STM. Our STM instrument has been upgraded by a cryogenic microwave transmission line[20, 22]. The microwave circuit design described herein incorporates a custom interferometer and lock-in detection, forgoing the use of a vector network analyzer.

Methods

Microwave reflectance measurements were performed on our upgraded Createc low-temperature ultra-high vacuum STM system (p<4⋅10−11p < 4 \cdot10^{-11} mbar, T<8.5T < 8.5 K)[22, 23], shown schematically in Figure 1(a). All experiments herein were performed over a clean Au(111) surface (Surface Preparation Laboratory), prepared by repeated cycles of Ar ion sputtering (0.61 kV, 10 min) and annealing (703 K, 10 min). A tungsten tip (Bruker DTT10) was used as the STM and microwave probe. The bias voltage VDCV_{\mathrm{DC}} was applied to the tip. The microwave signal was added to VDCV_{\mathrm{DC}} via a bias-tee (SHF BT45R-B). Inside the vacuum chamber, the microwaves are transmitted via microwave compatible cables (Elspec MK5001 and Elspec Stormflex 047Cryo). At the sample side, a second bias-tee (Tektronix PSPL5541A) separates the DC line, which is grounded at the tunnel current amplifier (Femto DLPCA 200), and the microwave line, which is terminated with a 50 Ω\Omega load. Note that, as shown in Figure 1(a), the bias-tees are the only microwave related components inserted in the DC line, which does increase the tunnel current noise by approximately a factor of <2< 2. As the rest of the microwave circuit is fully separated from the DC line, this does not affect current noise or the signal-to-noise ratio of GG spectra.

We detect amplitude changes and phase shifts of the microwaves that have been reflected at the tunnel junction. In order to increase the signal-to-noise ratio, the background resulting from the near full microwave reflection at the tunnel junction [see Eq. (1)] was subtracted using a Mach-Zehnder type interferometer[11]. The orange shaded section of the schematic in Figure 1(b) depicts the components and layout of the interferometer. Table I lists the details of the indicated components. The transmission line to the STM tip acts as the measurement signal of the interferometer [see pink arrows in Figure 1(b)], while the variable voltage attenuator (VVA) acts as the difference signal (see cyan arrows in Figure 1(b)). Figure 2(a) shows a schematic representation of the interferometer, including an indication of the microwave amplitude, AA, and phase, φ\varphi, at different parts of the interferometer. An optimum (i.e. highest) signal-to-noise ratio for S11,mS_{11,\mathrm{m}} detection is achieved for destructive interference between the two paths in the interferometer. This happens when the signals from the two parts have the same amplitude (A3=A2A_3 = A_2) and are antiphase (φ2=φ3±(2n+1)π\varphi_2 = \varphi_3 \pm(2n+1)\pi, where nn is an integer). Amplitude matching is performed by tuning the attenuation of the difference signal A3A_3 by the VVA input voltage. Antiphase conditions are obtained at fixed frequencies where the total path length of the two arms differ by exactly (n+1/2)λ(n+1/2)\lambda, where nn is an integer and λ\lambda is the microwave wavelength. The frequencies for which this condition is satisfied are referred to as notches in the frequency spectrum[11], as shown in Figure 2(b). Calibration measurements are performed at a frequency ff and VVA attenuation that maximizes destructive interference in the interferometer at the tunneling setpoint.

Image not extracted

Figure 2. Details of the interferometer. (a) Schematic representation of the interferometer. The purple and cyan lines represent the measurement and difference paths of the microwaves through the interferometer, as in Figure 1(b). Amplitudes AiA_i and phases φi\varphi_i of VMWV_{\mathrm{MW}} are indicated at different points in the interferometer. In the directional coupler where the two paths meet, the two waves interfere and the resulting VMWV_{\mathrm{MW}} depends on the amplitudes and phases of the interfering waves. (b) Microwave scattering parameter S11,m(f)S_{11,\mathrm{m}}(f) amplitude measured without (black curve 1) and with (orange curve 2) the interferometer. Curve 1 is obtained with A3=0A_3 = 0, so no interference takes place and the output amplitude is solely determined by the loss in the microwave circuit. For curve 2, A3A_3 is tuned to match A2A_2. The relative phase of the interfering waves is determined by the frequency ff. When ff is such that the two waves are out of phase, i.e. φ2=φ3±(2n+1)π\varphi_2 = \varphi_3 \pm(2n + 1)\pi, destructive interference is maximized, as indicated with a red dashed line.

The green shaded section of the schematic in Figure 1(b) depicts the down-modulation circuit and Table I lists the details of the indicated components. Before down-modulation, the microwaves coming from the interferometer are amplified using a low-noise amplifier (LNA). Then, the microwave signal is demodulated to an intermediate frequency fIFf_{\mathrm{IF}}, since the microwave frequency ff is too high to detect with the lock-in amplifier directly. The demodulated signal was detected using a lock-in amplifier, at a reference frequency of fIFf_{\mathrm{IF}} obtained by demodulating a fraction of the microwaves going to the interferometer. With this arrangement, continuous phase-stable detection is achieved. In the experiments presented herein, f=0.908f = 0.908 GHz and fIF=500f_{\mathrm{IF}} = 500 kHz. The peak-to-peak amplitude of the incident microwaves at the tunnel junction[22] was determined to be Vpp,MW=69V_{\mathrm{pp,MW}} = 69 mV. GG was measured by a lock-in amplifier (EG&G 5210) using a sinusoidal modulation of VDCV_{\mathrm{DC}} (Vrms,lock−in=8V_{\mathrm{rms,lock-in}} = 8 mV, flock−in=775f_{\mathrm{lock-in}} = 775 Hz).

LabelComponentBrand/type
DiCDirectional CouplerMarki C13-0126
HyCHybrid CouplerMarki QH0R518
PDPower DividerMarki PD0126
M3-Port MixerMini Circuits ZX05-63LH-S+
LNALow-Noise AmplifierLow Noise Factory LNR1_15A_SV
VVAVariable Voltage AttenuatorUniversal Microwave Corp. AG-U000-60V
Gen 1Generator 1Keysight E8257D
Gen 2Generator 2Rohde & Schwarz SMA100A
LIA2Lock-In AmplifierZurich Instruments MFLI (500 kHz)

Table 1. Details of the microwave circuit components depicted in Figure 1.

Results

Figure 1(a) outlines the schematic of our retraction curve measurement set-up for the calibration of S11S_{11} based on the simultaneous measurement of G(z)G(z) and S11,m(z)S_{11,\mathrm{m}}(z) during tip retraction. Figure 3(a) shows G(z)G(z) recorded while retracting the tip a distance of Δz=51\Delta z = 51 nm away from the Au(111) surface. Prior to recording this retraction curve, the tip was approached by Δz=−1\Delta z = -1 nm from the tunneling setpoint (51 pA and 20 mV), representing z=0z = 0, to the starting point of the retraction curve at z=−1z = -1 nm. This was done in order to achieve a sufficiently large change of S11,m(z)S_{11,\mathrm{m}}(z), which is found to occur at a conductance close to atomic point contact, 1 G01\,G_0, while retaining an approximately exponential G(z)G(z) relation in the tunneling regime, as shown in the lower panel of Figure 3(a). Here, G0=2e2/h≈77.48G_0 = 2e^2/h \approx77.48 μS is the conductance quantum, e≈+1.6022×10−19e \approx+1.6022 \times10^{-19} C is the elementary charge and h≈6.6261×10−34h \approx6.6261 \times10^{-34} Js is the Planck constant. The red curves in Figure 3(a) represent a numerical fit to the G(z)G(z) curve with an exponential decay function, as is well-established for tunneling in STM[13]

Experimental retraction curves

Figure 3. Experimental retraction curves. (a) Conductance retraction curve G(z)G(z) recorded over clean Au(111). The lower panel shows the portion of G(z)G(z) at z<0z < 0. The red curves represent a fit with an exponential decay function (R2=0.9986R^2 = 0.9986). (b) zz-dependence of amplitude (upper panel) and phase (lower panel) of the complex S11,m(z)S_{11,\mathrm{m}}(z) during the microwave reflection retraction curves. (c) S11,mS_{11,\mathrm{m}} plotted in the complex plane. The color code is non-linear with respect to zz, with red being points taken closest to the surface and green/cyan furthest away from the surface. All zz-values are reported relative to the tunneling setpoint, with negative numbers meaning closer to the surface.

G(z)=G0e−2κeff(z−z0)(4)G(z) = G_0 e^{-2\kappa_{\mathrm{eff}}(z-z_0)} \tag*{(4)}

where κeff\kappa_{\mathrm{eff}} and z0z_0 are parameters of the fit. From the fit, we find κeff=3.96±0.01\kappa_{\mathrm{eff}} = 3.96 \pm0.01 nm−1^{-1}, which is significantly lower than the value of κ≈11.9\kappa\approx11.9 nm−1^{-1} expected from the work function of Au(111) of ϕ≈5.4\phi\approx5.4 eV alone[13]. We remark that our κeff\kappa_{\mathrm{eff}} value is consistent with literature values for similarly small tunnel distances. In particular, originating from different effects, such as a non-trapezoidal shape of the tunneling barrier[24], finite input impedance of the tunneling current amplifier[25], and the setting-in of bond formation between tip and sample close to point contact.

In our concurrent microwave reflection measurement, the amplitude of S11,m(z)S_{11,\mathrm{m}}(z) decreases approximately linearly over the entire retraction range, as shown in the upper panel of Figure 3(b). In contrast, the phase of S11,m(z)S_{11,\mathrm{m}}(z), shown in the lower panel of Figure 3(b), increases sharply in the first nm above the surface, followed by a nearly linear increase over the rest of the retraction curve. Figure 3(c) shows the same S11,m(z)S_{11,\mathrm{m}}(z) data in a different representation, i.e. plotted in the complex plane and the zz-dependence shown as color gradient. Based on the shape of the S11,m(z)S_{11,\mathrm{m}}(z) curve, we distinguish two different regimes, labeled high GG (red marker) and G≈0G \approx0 (cyan marker). The high GG regime ranges from z=−1z=-1 nm to z≈−0.8z\approx-0.8 nm and is characterized by a conductance of about G>G0/3G>G_{0}/3. The G≈0G\approx0 regime starts at z≈4z\approx4 nm.

In the next step, the values of the parameters e00e_{00}, e01e_{01}, and e11e_{11} are determined from the experimental S11,m(z)S_{11,\mathrm{m}}(z) data with the help of a numerical least-squares fit (R2≥0.91R^{2}\ge0.91) using Eq. (3). We obtain the values e00=(−1.7±0.4)+(0.0±0.1)ie_{00}=(-1.7\pm0.4)+(0.0\pm0.1)i, e01=(0.41±0.08)+(−2.3±0.4)ie_{01}=(0.41\pm0.08)+(-2.3\pm0.4)i, and e11=(0.91±0.02)+(0.79±0.07)ie_{11}=(0.91\pm0.02)+(0.79\pm0.07)i. Here, the numerical error intervals of the ee-parameters are obtained from numerical fitting (see also supplementary material Figure 1). The value of e00e_{00} depends on the constant CC in the calculation of S11,jun(z)S_{11,\mathrm{jun}}(z), as it compensates for the offset between S11,m(z)S_{11,\mathrm{m}}(z) and S11,jun(z)S_{11,\mathrm{jun}}(z). Possible interdependencies between e01e_{01} and e11e_{11} can be analysed from the mathematical form of Eq. (3). Since all quantities are complex valued, simple interdependencies are obtained only in certain cases. For instance, in the limit of ∣S11,jun∣→±∞\lvert S_{11,\mathrm{jun}}\rvert\to\pm\infty, Eq. (3) reduces to S11,m=e00−e01e11S_{11,\mathrm{m}}=e_{00}-\frac{e_{01}}{e_{11}}, revealing a trivial interrelation between e01e_{01} and e11e_{11}. For the general case both e01e_{01} and e11e_{11} are required for a correct description. Notice that the numerical fit procedure concerns the difference between the experimental data (Figure 3c) and the respective values predicted by Eq. (3), as follows: The scattering parameter at the tunnel junction S11,jun(z)S_{11,\mathrm{jun}}(z) is calculated using Eqs. (2) and (1). The G(z)G(z) measured by the lock-in amplifier is the starting point in the calculation of S11,jun(z)S_{11,\mathrm{jun}}(z). The exponential fit to G(z)G(z) of Eq. (4), as shown in Figure 3(a), is substituted in for the conductance term in Eq. (2). For the unknown capacitance term, we apply a Taylor series approximation of C(z)≈C(d)+C′(d)(z−d)+⋯C(z)\approx C(d)+C'(d)(z-d)+\cdots. Since we perform the numerical fitting in a narrow zz-range of 0.540.54 nm, we consider only the constant and linear terms of the series and neglect higher terms. Notice that in the high GG regime, the exponential decay of G(z)G(z) dominates the linear C′(d)(z−d)C'(d)(z-d) term in the Taylor expansion of C(z)C(z). Therefore, we also neglect the linear term and assume C(z)=const.C(z)=\mathrm{const.} in Eq. (2) for the rest of our analysis. We remark that after successful calibration the validity of the assumption is confirmed, i.e. the zz-range of z=−1z=-1 to −0.46-0.46 nm used for calibration is sufficiently narrow (see discussion).

Figure 4(a) shows our result of the calibration, where the experimental S11,m(z)S_{11,\mathrm{m}}(z) data from Figure 3(c) has been transformed into calibrated admittance Yjun(z)Y_{\mathrm{jun}}(z) using the values of the ee-parameters obtained from the numerical fit procedure above. The entire measured zz-range between z=−1z=-1 to 5050 nm is calibrated this way. Notice that according to Eq. (2), Yjun(z)=G+i 2πf CY_{\mathrm{jun}}(z)=G+i\,2\pi f\,C and thus the real and imaginary parts of Yjun(z)Y_{\mathrm{jun}}(z) correspond to GG and CC, respectively. After calibration, the different conduction regimes can still be clearly distinguished [compare Figure 3(c) and Figure 4(a)]. As expected, the high GG regime (red, z<−0.8z<-0.8 nm) is dominated by a change of GG, while CC remains almost constant. Conversely, the low-GG regime (cyan, z>4z>4 nm) is dominated by a significant change of CC, while the value of GG is almost constant at G≈0G\approx0.

Image not extracted

Figure 4. The result of calibration. (a) Calibrated admittance YjunY_{\mathrm{jun}} (reciprocal impedance) plotted in the complex plane. (b) The conductance GMW(z)=Re⁡[Yjun(z)]G^{\mathrm{MW}}(z) = \operatorname{Re}[Y_{\mathrm{jun}}(z)] derived from S11,jun(z)S_{11,\mathrm{jun}}(z) after calibration. The lower panel shows the portion of GMW(z)G^{\mathrm{MW}}(z) at z<0z < 0. The red curves represent the fit to G(z)G(z) as measured with the conventional lock-in method. These curves are included to directly compare G(z)G(z) with GMW(z)G^{\mathrm{MW}}(z); in a successful calibration the conductance curves coincide. (c) The capacitance change ΔCMW(z)=Im⁡[Yjun(z)]/(2πf)\Delta C^{\mathrm{MW}}(z) = \operatorname{Im}[Y_{\mathrm{jun}}(z)]/(2\pi f) derived from S11,jun(z)S_{11,\mathrm{jun}}(z) after calibration. The green line is a linear fit (R2=0.9957R^2 = 0.9957) to ΔCMW(z)\Delta C^{\mathrm{MW}}(z). All zz-values are reported relative to the tunneling setpoint, with negative numbers meaning closer to the surface.

Figure 4(b) and (Figure 4c) show the calibrated data of Figure 4(a) in a different representation, i.e. GMW(z)G^{\mathrm{MW}}(z) and ΔCMW(z)\Delta C^{\mathrm{MW}}(z) respectively. To avoid confusion with the conventionally obtained GG data, the superscript MW indicates that the respective data has been obtained from the microwave reflection procedure. Note that a capacitance shift ΔCMW(z)\Delta C^{\mathrm{MW}}(z) is measured, since the magnitude of the capacitance between the sample and the tip-holder (stray capacitance) is not exactly known[10].

Discussion

The quality of the obtained calibration can be judged by comparing the conductance data, as shown in Figure 4(b), that has been obtained in two different manners. Firstly, the exponential fit to G(z)G(z), directly measured by the common dI/dVdI/dV lock-in method [see also Figure 3(a)], which was used to obtain the ee-parameters, is shown as the red curve. Secondly, GMW(z)G^{MW}(z) obtained from the microwave reflection according to the procedure described above is shown as the blue curve. As can be seen, both curves are in reasonably good agreement. Notice that the agreement is particularly good in the calibration interval of z<−0.46z < -0.46 nm, as illustrated in the lower panel of Figure 4(b). From this good quantitative agreement, we conclude that the calibration was successful. We remark that the minor discrepancy between the fit to G(z)G(z) (red) and the GMW(z)G^{MW}(z) data (blue) in the calibration interval cannot be explained by deviations in CC due to the tip apex. Instead, it may be rationalized by a frequency dependence of GG[26, 27] that is not explicitly included in the calibration procedure and goes beyond the scope of the present work.

In order to verify the validity of our assumption that the change in capacitance is negligible on the calibration interval of z=−1z = -1 to −0.46-0.46 nm, we compare the relative impact of changes in GG and CC on Yjun(z)Y_{\mathrm{jun}}(z). We obtain a respective change of ΔG(z)=ΔRe⁡[Yjun(z)]≈1.6 G0≈126\Delta G(z) = \Delta\operatorname{Re}[Y_{\mathrm{jun}}(z)] \approx1.6\,G_0 \approx126 μS, see lower panel of Figure 4(b). In comparison, a linear fit over the whole zz-range from −1-1 to +50+50 nm of the ΔCMW(z)\Delta C^{MW}(z) curve, see Figure 4(c), yields a value for the slope ∂CMW(z)/∂z\partial C^{MW}(z)/\partial z of −2.3-2.3 fF/nm, resulting in an absolute change of ΔCMW≈1.2\Delta C^{MW} \approx1.2 fF over the respective zz-range (Δz=1−0.46=0.54\Delta z = 1 - 0.46 = 0.54 nm), and finally resulting in a value of ΔCMW 2πf=ΔIm⁡[Yjun(z)]≈7\Delta C^{MW}\,2\pi f = \Delta\operatorname{Im}[Y_{\mathrm{jun}}(z)] \approx7 μS. This value is circa 20 times smaller (!) than the above value of 126 μS, demonstrating that YjunY_{\mathrm{jun}} is indeed dominated by GG (and not by CC), in agreement with the above assumption. Notice that within the calibration interval the slope ∂G(z)/∂z\partial G(z)/\partial z derived from Figure 4(b) is found to be always larger than the respective slope of 2πf ∂CMW(z)/∂z2\pi f\,\partial C^{MW}(z)/\partial z (see above), which is again in agreement with the above assumption. Repeating the calibration for slightly larger and smaller zz-ranges (±0.05\pm0.05 nm) yields values that are in good agreement [see supplementary material Figure 2 for deviations in GMW(z)G^{MW}(z) and ΔCMW(z)\Delta C^{MW}(z)] to the above values. Either enforcing the G=0G = 0 condition for z>10z > 10 nm or accounting for the linear term in C(z)C(z) in a self-consistent optimization loop, again yields consistent values. This shows that the calibration procedure yields robust values for the capacitance.

The ee-parameters obtained in this calibration procedure numerically describe non-idealities, such as absorptions and reflections, of microwave propagation through the transmission line from the calibration plane to the measurement plane. As such, these ee-parameters are independent from zz and from tip and sample properties[10]. The calculation of S11,jun(z)S_{11,\mathrm{jun}}(z) involves an exponential fit to G(z)G(z), so the tip must be stable to give an exponential G(z)G(z) dependence during calibration. The ee-parameters can be used to calibrate microwave reflectance measurements to obtain quantitative ΔCMW\Delta C^{MW} values, as shown in Figure 4(c). It is important to note that a particular set of ee-parameters is only valid for the frequency and VNA settings at which the calibration was performed. To obtain a frequency spectrum, a calibration should be performed for each frequency. Currently, this is done with a separate approach curve at each frequency notch, but we envision using a full frequency sweep during a single approach curve in the future.

Conclusion

We present a calibration procedure for microwave reflectance spectroscopy based on the acquisition of conductance retraction curves in an ultra-high vacuum low-temperature STM instrument. While the retraction curve calibration concept based on capacitive force detection has been very successfully employed for microwave reflection measurements in an AFM[10] in the past, in STM the force detection is not available. Therefore, we acquire the tunneling conductance, which changes exponentially with distance, instead of the capacitance, which has a long-range interaction and changes only slowly with tip sample distance. We show that in the high GG regime, for approximately G>G0/3G > G_0/3 occurring at sub-nm tip-sample distances, the change of the measured scattering parameter S11,mS_{11,\mathrm{m}} with respect to the tunnel distance is dominated by changes in the conductance signal. In this narrow zz-range of 0.54 nm in the high-conduction regime, the change of the junction capacitance with tunnel distance can be neglected. This allows for a successful calibration of the error parameters describing the distortion of S11,mS_{11,\mathrm{m}} relative to the scattering parameter at the tunnel junction S11,junS_{11,\mathrm{jun}}. We remark that the employed ee-parameter procedure captures all non-idealities of the transmission line and detection circuit and can be readily extended to calibration of an entire range of frequencies.

The calibration of microwave reflectance spectra presented here will pave the way for quantitative capacitance and dielectric spectroscopy measurements with a lateral resolution of a few nm and ultimately single molecule fingerprinting in low-temperature ultra-high vacuum STM. Specifically, we expect that the tip-sample contribution to ΔCMW\Delta C^{\mathrm{MW}} can be isolated either in constant height measurements or with an appropriate capacitance model. This may allow us to determine material properties, such as the dielectric permittivity, in future experiments[12]. The imaging capabilities of STM together with frequency dependent microwave reflectance, could then be utilized to achieve full spectroscopic imaging.

Supplementary material

See the supplementary material for additional details on the dependence of the ee-parameters, GMW(z)G^{\mathrm{MW}}(z), and ΔCMW(z)\Delta C^{\mathrm{MW}}(z) on the zz-range included in the calibration procedure.

Acknowledgments

The authors acknowledge the financial support the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (Grant Agreement No. 771193). This work was additionally supported by funding from EMPIR project 20IND12 Elena.

Author declarations

Conflict of Interest

The authors have no conflicts to disclose.

Author Contributions

Bareld Wit: Conceptualization, Formal Analysis, Methodology, Validation, Visualization, and Writing. Georg Gramse: Conceptualization, Funding Acquisition, Methodology, Validation, and Writing. Stefan Müllegger: Conceptualization, Validation, Funding Acquisition, Project Administration, Supervision, and Writing.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

References

  1. [1]A. Imtiaz, T. M. Wallis, and P. Kabos, IEEE Microw. Mag. 15, 52 (2014).
  2. [2]H. P. Huber, I. Humer, M. Hochleithner, M. Fenner, M. Moertelmaier, C. Rankl, A. Imtiaz, T. M. Wallis, H. Tanbakuchi, P. Hinterdorfer, et al., J. Appl. Phys. 111, 014301 (2012).
  3. [3]Y. Martin, D. W. Abraham, and H. K. Wickramasinghe, Appl. Phys. Lett. 52, 1103 (1988).
  4. [4]S. Gómez-Moñivas, L. S. Froufe-Pérez, A. J. Caamaño, and J. J. Sáenz, Appl. Phys. Lett. 79, 4048 (2001).
  5. [5]J. R. Matey, and J. Blanc, J. Appl. Phys. 57, 1437 (1985).
  6. [6]P. J. Hansen, Y. E. Strausser, A. N. Erickson, E. J. Tarsa, P. Kozodoy, E. G. Brazel, J. P. Ibbetson, U. Mishra, V. Narayananmurti, S. P. DenBaars, and J. S. Speck, Appl. Phys. Lett. 72, 2247 (1998).
  7. [7]S. Grall, I. Alić, E. Pavoni, M. Awadein, T. Fujii, S. Müllegger, M. Farina, N. Clément, and G. Gramse, Small 17, 2101253 (2021)
  8. [8]C. Gao, T. Wei, F. Duewer, Y. Lu, and X.-D. Xiang, Appl. Phys. Lett. 71, 1872 (1997).
  9. [9]A. Imtiaz, T. M. Wallis, S.-H. Lim, H. Tanbakuchi, H.-P. Huber, A. Hornung, P. Hinterdorfer, J. Smoliner, F. Kienberger, and P. Kabos, J. Appl. Phys. 111, 093727 (2012).
  10. [10]G. Gramse, M. Kasper, L. Fumagalli, G. Gomila, P. Hinterdorfer, and F. Kienberger, Nanotechnol. 25 145703 (2014).
  11. [11]S.-S. Tuca, M. Kasper, F. Kienberger, and G. Gramse, IEEE Trans. Nanotechnol. 16, 991 (2017).
  12. [12]M. C. Biagi, G. Badino, R. Fabregas, G. Gramse, L. Fumagallie, and G. Gomila, Phys. Chem. Chem. Phys. 19, 3884 (2017).
  13. [13]C. J. Chen Introduction to Scanning Tunneling Microscopy, Oxford University Press (1993).
  14. [14]S. Müllegger, S. Tebi, A. K. Das, W. Schöfberger, F. Faschinger, and R. Koch, Phys. Rev. Lett. 113, 133001 (2014).
  15. [15]S. Baumann, W. Paul, T. Choi, C. P. Lutz, A. Ardavan, and A. J. Heinrich Science 350, 417 (2015).
  16. [16]V. Stetsovych, S. Feigl, R. Vranik, B. Wit, E. Rauls, J. Nejedlý, M. Šámal, I. Starý, and S. Müllegger, Sci. Rep. 12, 2865 (2022).
  17. [17]D. M. Pozar, Microwave Engineering, John Wiley & Sons Ltd (2011)
  18. [18]F. Kremer, A. Schönhals, Broadband Dielectric Spectroscopy, Springer (2002)
  19. [19]S. M. Anlage, V.V. Talanov, and A. R. Schwartz Scanning Probe Microscopy: Electrical and Electromechanical Phenomena at the Nanoscale ed S. V. Kalinin and A. Gruverman, Springer (2006).
  20. [20]Applying Error Correction to Network Analyzer Measurements, Application Note, Keysight Technologies, Santa Clara, CA (2022). Retrieved from Keysight website: https://www.keysight.com/us/en/assets/7018-06761/application-notes/5965-7709.pdf.
  21. [21]M. Farina, D. Mencarelli, A. Di Donato, G. Venanzoni, and A. Morini, IEEE Trans. Microw. Theory Tech. 59, 2769 (2011).
  22. [22]S. Feigl, R. Vranik, B. Wit, and S. Müllegger, Rev. Sci Instr. 92, 043710 (2021).
  23. [23]B. Wit, R. Vranik, and S. Müllegger, Sci. Rep. 12, 6183 (2022).
  24. [24]N. D. Lang, Phys. Rev. B 37, 10395 (1988).
  25. [25]L. Olesen, M. Brandbyge, M. R. Sørensen, K. W. Jacobsen, E. Lægsgaard, I. Stensgaard, and F. Besenbacher, Phys. Rev. Lett 76, 1485 (1996).
  26. [26]T. Brandes, D. Weinmann, and B. Kramer, Europhys. Lett. 22, 51 (1993).
  27. [27]G. Goppert and H. Grabert, C. R. Acad. Sci. Ser. II 327, 885 (1999).

Paper details

vlnc:W_ejgzur8z
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

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