Scattering-type scanning near-field optical microscopy with Akiyama piezo-probes

Imaging a sample using conventional microscopy techniques relies on the collection of far-field light that has been scattered off the sample. The finest features of a material that can be observed using these techniques are limited by the far-field diffraction limit, which is approximately half of the incident light wavelength.¹ Scattering-type scanning near-field optical microscopy (s-SNOM), on the other hand, relies on collecting light that has been scattered in the immediate vicinity of the sample by a probing tip in the near-field regime. This enables imaging of materials on a scale far below the diffraction limit.^2–4 

Over the past two decades, s-SNOM imaging at ambient conditions has led to a plethora of scientifically important discoveries.^5–13 However, fundamentally interesting physics often occurs exclusively at cryogenic temperatures.^14–18 The reasons why constructing a cryogenic s-SNOM apparatus remains a challenging task are multi-fold: first of all, the apparatus requires simultaneous accommodation of three essential components in a limited space: (1) the sample stage including scanners, positioners, and optical access for sample/device observation, (2) focusing optics and clear pathways for the incident/scattered light used for near-field imaging with frequencies from near-infrared (near-IR) all the way to far-IR and THz ranges, and (3) an oscillating tip operated in the tapping mode, which so far has been predominantly based on laser-based feedback for a cantilever that is not self-sensing.¹⁹ With these components, one must maneuver several independently adjustable parts: the tip, the sample, and the steering optics for the incident light are individually controlled using micro-precision positioners. This often yields a large footprint for the instrument. The second challenge is that the cooling power must overcome enhanced radiative heating due to the incident light and environmental thermal radiation that enters through the windows needed for light coupling in the cryogenic chamber. The large number of mechanical and optical parts further adds to the mass that needs to be cooled, which impacts the base operating temperatures. The third important challenge is achieving control of the quality factor of the tip. This value tends to be high in vacuum and at cryogenic temperatures, preventing the use of a stable feedback based on amplitude modulation.

In the present paper, we address the above challenges by using a self-sensing Akiyama probe (A-probe). The A-probe is constructed from a quartz tuning-fork with a sharp micromachined Si cantilever glued to its prongs. This combination renders a compact and stable tip with a relatively soft spring constant (∼5 N/m, orders of magnitude smaller compared to bare tuning forks) that can be excited either electrically or mechanically. The resulting apparatus is capable of performing s-SNOM imaging and near-field photocurrent measurements with high spatial resolution and a good signal-to-noise ratio (S/N). We believe this is one of the most convenient and promising directions for achieving s-SNOM at extreme conditions (e.g., sub-K, magnetic field, or far-IR) in the near future. Through this paper, we demonstrate results originating from two different systems: an s-SNOM designed for room temperature, and a cryo-SNOM that can work down to T = 15 K. Schematic representations of the A-probe based s-SNOM systems are shown in Fig. 1.

An s-SNOM system consists of two main components: an atomic force microscope (AFM) [right panels in Figs. 1(a) and 1(b)] and an asymmetric Michelson interferometer [left panels in Figs. 1(a) and 1(b)]. The AFM provides a platform for probing the light-sample interaction in the near-field regime. By focusing a laser beam onto the AFM tip, one collects the light scattered off the tip, which encodes the local optical properties of the sample. The Michelson interferometer enables phase-sensitive detection of the scattered near-field signal. To suppress the background scattering off the tip shank and sample surface, the tip is operated in the tapping mode, oscillating harmonically close to its mechanical resonance frequency. The detected scattered signal is demodulated at higher harmonics of the tip tapping frequency to filter out the undesired far-field background.^20,21

To perform AFM measurements, measuring the probe's tapping amplitude and frequency is required. In conventional AFM, those parameters can be determined by quadrant or interferometry detection.²² In the A-probe based system, the tapping amplitude and frequency can be measured directly from the piezoelectric signal generated by the tuning fork since the A-probe is self-sensing.²³ This eliminates the need for separate optical alignment and detection schemes for the tip. As shown on the right panel of Fig. 1(a), the tip is fixed while the sample and focusing lens sit on XYZ piezo stages for sample scanning and optical alignment, respectively. Fixing the tip position increases the overall mechanical stability of the A-probe based s-SNOM system.

Another alternative to cantilever-based AFM is a tungsten wire attached to a tuning fork,^24,25 which is currently the only other experimentally realized piezo-probe. However, bare tuning forks have several disadvantages compared to the A-probe system. First, the effective spring constants of quartz tuning forks are about 25 000 N/m,²⁶ almost four orders of magnitude larger than those of A-probes. Thus, good mechanical coupling and substantial shaking power are required to achieve a tapping amplitude of 50–100 nm, which is usually required to effectively modulate the near-field interaction. Such strong dithering may affect the stability of the tuning fork system. Another disadvantage is related to the tip shank length being on the order of tens or hundreds of micrometers up to a few mm. As such, the dimensions of the shank are usually not controlled to below μm-scale length, which is relevant in the infrared regime. These characteristic sizes are mismatched compared to the incident infrared light wavelengths and, therefore, do not offer a good near-field antenna enhancement. Also, the low-temperature Q-factors for quartz tuning forks easily reach 50 000 and are typically higher than those of A-probes, impacting the measurement times. Last but not least, the large effective spring constant for the quartz tuning fork system implies that the tip-sample interaction enters the repulsive regime even at amplitude set-points that are very close to the free-space values. This “hammering” of the sample with a stiff cantilever may affect measurements of soft materials.

The Michelson interferometer, optimized for the mid-IR frequency range, consists of a ZnSe beam splitter that reflects 40% of the beam toward the tip and transmits 60% of the beam toward a reference arm. (The transmission arm and reflection arm are interchangeable in most cases.) The reflected light gets focused onto the tip through either an aspherical lens [Fig. 1(a)] or an off-axis parabolic (OAP) mirror [Fig. 1(b)]—whether an OAP or lens is used depends on the light source; the OAP has the advantage of being wavelength independent while the lens provides for easier alignment. The enhanced scattered signal is collected through the same lens or OAP, recombined with the reference arm, and focused onto a Mercury-Cadmium-Telluride (MCT) detector.

Further reduction in the overall sample-space size can be realized by employing a fixed OAP. This is done at the cost of not being able to improve the near-field S/N by moving the mirror while the AFM tip is in contact with the sample. However, straightforward alignment procedures have been put in place (see next paragraph for more details), which allow us to offset the lack of a moving mirror. A fixed mirror position, especially for larger mass OAPs, also allows for better thermal anchoring and avoiding drifts and shifts of the light focus spot as temperature is varied. Smaller size paraboloids can be, in principle, mounted on XYZ piezo-stages to work with or without an additional vibration isolation stage, depending on whether the optical alignment is done with the tip in contact.

In Figs. 2(a) and 2(b), we show actual images of the RT (room temperature) and LT (low temperature) systems, respectively. The tip in Fig. 2(a) sits on a piece of printed circuit board, which is attached to a 24° wedge block. The wedge block is held in place by a magnetic mount and does not move throughout the experiment. Typical resonance curves for both frequency and phase are shown in the left panel of Fig. 2(c). To maximize the s-SNOM signal in the RT system, the lens is scanned until a beam hotspot is found [right panel in Fig. 2(c)]. An optical microscope image of the Akiyama probe is shown in Fig. 2(d). A typical alignment procedure in the fixed OAP configuration involves two main steps. The first is achieving colinear alignment of a visible “pilot” laser, e.g., HeNe, with the IR laser, which is routinely achieved by using temperature sensitive liquid crystal sheets. The second step involves precise alignment and focusing of the pilot laser on the very apex of the AFM tip. This part is realized by a preliminary positioning of the tip in the vicinity of the pilot, as shown in Fig. 2(d), followed by high resolution direct tip imaging on a CCD camera by inserting a flip-mirror in the collection path after the beam splitter. This procedure usually allows for the detection of the true near-field signal during the first tip-sample approach. Further improvement in the SNOM signal can be achieved by parallel translation of the IR beam.

It is noticeable that the geometry of the tip shank of the A-probe is very distinctive from other commonly used AFM tips for s-SNOM such as the Arrow™ NCPt tip. Although both tips have a similar apex radius, the A-probe is almost twice as long (28 $μ$m vs 15 $μ$m). It is well known that the tip-antenna effect is closely related to the tip shank length; therefore, the A-probe should exhibit a good near-field scattering signal for longer incident light wavelengths. To verify this, we first perform full-wave numerical simulations using the method of moment (MoM) technique. Comparing to other full-wave numerical algorithms, MoM is especially suitable for such a simulation because only the tip surface needs to be discretized. This offers a monumental advantage in terms of computer memory and computation time. We follow the proposed simulation method in the recent study.²⁷ The A-probe and arrow tip are first imaged using a scanning electron microscope (SEM) as shown in Fig. 3(a) to precisely determine their geometries. Then the models are constructed accordingly as shown in Fig. 3(b). Figure 3(c) shows the amplitude of the scattered field demodulated at the second harmonic (⁠$S2$⁠), when the tip is placed on top of a metallic surface. The simulation is carried out in a broad frequency range from 1 terahertz (THz) to near-IR. As we can see in the mid- and near-IR spectral range, the scattering efficiency of both tips is comparable, while the A-probe exhibits significantly stronger scattering at longer wavelengths toward the THz regime. This simulation result demonstrates that the A-probe can be an ideal candidate for far-IR s-SNOM imaging (e.g., with an incident wavelength in the range of ∼20–100 μm), where geometry-optimized commercial tips are scarce.^28,29 In both spectra, multiple peaks are observed, which are attributed to the antenna resonances.²⁹ 

To characterize the signal-to-noise and near-field contrast of the s-SNOM measurements at room temperature and 15 K, we perform nano-imaging of standard 20 nm SiO₂/Si test samples from NT-MDT (RT system) and NanoAndMore (LT system). As shown in Fig. 4(a), we found the S/N of S₃ to be greater than 30 for the RT system. For the LT-SNOM, the S/N at 300 K is about a factor of two larger than at 15 K [Figs. 4(b) and 4(c)]. This is because at room temperature the Q-factor of the A-probe is more than an order of magnitude lower than at low temperatures, allowing for faster rastering and easier optimization of AFM feedback parameters. While electronic Q-control is very helpful, we find that usually it cannot completely compensate for the decrease in the resonance linewidth due to cooling. The near-field contrast between SiO₂ and Si, S₃(SiO₂)/S₃(Si), with a ∼10 μm incident light has a consistent value of ∼0.7 throughout the experiments and is comparable to previously reported values.³⁰ 

In the following, we present two case studies of the application of the A-probe system to two-dimensional materials. The first, shown in Fig. 5(a), is a near-field photocurrent measurement using the RT setup. The tip-enhanced IR light creates a local temperature gradient, which can induce a directional photocurrent with the presence of local inhomogeneities (and thus, local variations of the Seebeck coefficient). Instead of gathering the scattering signal, we detect the near-field photoinduced current between two closely placed electrodes on top of a bilayer graphene sample [left and middle panels in Fig. 5(a)]. The near-field photocurrent signal (P_n) is amplified by a preamplifier and demodulated at higher harmonics (⁠$n≥2$⁠) of the tip tapping frequency. We found a good signal-to-noise ratio of 20 for P₃ with approximately 13 mW incident power and a −20 V back-gate voltage. We observed the strongest signals in the vicinity of the electrodes as expected and observed fringes close to suspected line defects [Fig. 5(a), right panel]. The fact that the signal is strongest and reversely signed at the two electrodes suggests that we are observing the Seebeck effect. This phenomenon has been reported in the literature,³¹ and we do not discuss the physics in detail here.

The second case study, shown in Fig. 5(b), demonstrates AFM topography and λ = 10.6 μm SNOM data taken at ambient conditions in commercially available polycrystalline graphene. The scans reveal information beyond the typical optical contrast between graphene and the (probably insulating) islands associated with defects/adsorbates. We observe “halos” and “highways” surrounding and connecting topographic features, which arise from efficiently damped plasmonic scattering at grain boundaries or defect structures.³² A good S/N can be achieved routinely up to the fourth harmonic of the tapping frequency.

Our explorative studies demonstrate the particular strength of using a tuning fork-based Akiyama probe for room temperature s-SNOM and cryo-SNOM. We have demonstrated its nano-imaging capability at room and low temperatures and its potential for near-field photocurrent mapping. In principle, this method is not limited to mid-IR and can be extended to near- and far-IR, as demonstrated using full-wave simulation. Due to the much-simplified experimental layout, further experiments on magnetic field-dependent s-SNOM and cryo-SNOM at lower temperatures are possible and under way.

The authors of Stony Brook University acknowledge support from the National Science Foundation under Grant No. DMR-1904576. A.G. acknowledges support from the DOE Early Career Research program (Grant No. 2005410), from the Yale West Campus Materials Characterization Core for SEM imaging of AFM tips, and from Attocube Systems in the design of a custom LT-AFM system based on the Akiyama probe.

M.K.L. acknowledges support from the NSF Faculty Early Career Development Program under Grant No. DMR - 2045425. X.D. acknowledges support from the National Science Foundation under Grant No. DMR-1808491.