Ultramicroscopy 205 (2019) 49–56 Contents lists available at ScienceDirect Ultramicroscopy journal homepage: www.elsevier.com/locate/ultramic The new FAST module: A portable and transparent add-on module for time- T resolved investigations with commercial scanning probe microscopes ⁎ Carlo Dria,b, Mirco Panighelb, , Daniel Tiemannc, Laerte L. Paterab,d,1, Giulia Troianod,e,2, Yves Fukamorie, Fabian Knollere, Barbara A.J. Lechnere, Giuseppe Cauteroa, Dario Giuressia, Giovanni Comellib,d, Jordi Fraxedasc, Cristina Africhb, Friedrich Esche a Elettra-Sincrotrone Trieste, S.S. 14 km 163.5, Basovizza, I-34149 Trieste, Italy b CNR-IOM Laboratorio TASC, S.S. 14 km 163.5, Basovizza, I-34149 Trieste, Italy c Catalan Institute of Nanoscience and Nanotechnology (ICN2), CSIC and BIST, Campus UAB, Bellaterra, 08193 Barcelona, Spain d Department of Physics, University of Trieste, via A. Valerio 2, I-34127 Trieste, Italy e Chair of Physical Chemistry, Department of Chemistry & Catalysis Research Center, Technical University of Munich, Lichtenbergstr. 4, 85748 Garching, Germany A R T I C LE I N FO A B S T R A C T Keywords: Time resolution is one of the most severe limitations of scanning probe microscopies (SPMs), since the typical FastSTM image acquisition times are in the order of several seconds or even few minutes. As a consequence, the char- FastAFM acterization of dynamical processes occurring at surfaces (e.g. surface diffusion, film growth, self-assembly and Scanning tunneling microscopy chemical reactions) cannot be thoroughly addressed by conventional SPMs. To overcome this limitation, several Atomic force microscopy years ago we developed a first prototype of the FAST module, an add-on instrument capable of driving a Scanning probe microscopy commercial scanning tunneling microscope (STM) at and beyond video rate frequencies. Here we report on a fully redesigned version of the FAST module, featuring improved performance and user experience, which can be used both with STMs and atomic force microscopes (AFMs), and offers additional capabilities such as an atom tracking mode. All the new features of the FAST module, including portability between different commercial instruments, are described in detail and practically demonstrated. 1. Introduction experiment temperature. In the past, several attempts have been made in order to achieve high scanning frequencies, typically by designing The invention of scanning probe microscopes in the 1980s has dedicated instruments with high resonance frequency scanners [2–4], opened the unprecedented possibility of imaging and controlling matter in some cases even performing the slow and fast movements by separate at the molecular and atomic scale. The publishing rate of research ar- scan heads [5,6] or different scanner shapes and probe trajectories [7]. ticles where SPMs are adopted has grown steadily through the years, In combination with fast feedback electronics and preamplifiers, frame exceeding 7 k/year in 2017 [1]. The versatility of these instruments is rates up to 200 images/s could be reached in a hybrid mode between one of their main strengths: they can be operated in a multitude of the constant-height and constant-current [4,8]. Overall, several suc- environments, from vacuum, to air and high pressure cells, in liquids or cessful solutions have been proposed with dedicated systems for electrochemical cells. FastSTM [2–9], FastAFM [10–15] and atom tracking [16,17]. However, one of their fundamental limitations is the time resolu- In contrast to these approaches, we accelerate existing, slow in- tion, since the majority of commercially available instruments offers struments without modifying the existing scanner heads, exploiting the typical acquisition times per image of seconds or even minutes, de- fact that the majority of the mechanical problems encountered in fast, pending on imaging conditions. This hinders the observation of many constant-current measurements can be avoided when measuring in a dynamical phenomena of fundamental and technological interest that quasi-constant-height mode [2]. Based on this concept, back in 2011, commonly occur on much faster time scales and, in the case of rela- following several years of prototyping and experiments, we developed a tively large activation energies, cannot be slowed down by lowering the unique modular solution and introduced the first prototype of the FAST ⁎ Corresponding author. E-mail address: panighel@iom.cnr.it (M. Panighel). 1 Present address: Institute of Experimental and Applied Physics, University of Regensburg, D-93053 Regensburg, Germany. 2 Present address: Telit Communications Spa, Via Stazione di Prosecco 5b, Sgonico, I-34010 Trieste, Italy. https://doi.org/10.1016/j.ultramic.2019.05.010 Received 7 December 2018; Received in revised form 3 May 2019; Accepted 24 May 2019 Available online 25 May 2019 0304-3991/ © 2019 Elsevier B.V. All rights reserved. C. Dri, et al. Ultramicroscopy 205 (2019) 49–56 module [18,19], an add-on electronics applicable in principle to any the acquisition of the fast measurement signals (with an input range of existing scanning probe microscope. The basic concept behind the 1 V peak to peak) driven by the same clock as the DAC, thereby en- module was to simply “piggyback” the existing electronics of the suring stable and reliable synchronization between probe motion and commercial SPM in a completely transparent fashion, by generating the data acquisition. It is worth noting that the high-speed (16 bit @ fast scan signals and adding them to the original, slow scan signals. The 100 MS/s) of the DACs used to control the probe movement is crucial to probe signal was then acquired independently by the module itself. In avoid any scan distortion due to resonances caused by overtones. While this way, the module could be installed without modification of the the typical bandwidth required for fast SPM measurements stays in the existing scanner hardware and electronics and enabled seamless MHz frequency range, commercially available preamplifiers provide a switching between the fast and slow scan modes. rather smaller bandwidth (typically around 45 kHz in the 109 V/A Against the common belief that mechanical resonances of the SPM range and up to 200 kHz in the 108 V/A range), thus leading to a strong scanners pose an insurmountable limitation to the scan speed, with our suppression of the signal in the frequency range of interest. In this initial prototype, we successfully demonstrated that imaging at high context, the high sampling rate (14 bit @ 100 MS/s) of the FAST speed with fast scan frequencies among and above such resonances is module ADCs allows oversampling the input in order to recover the indeed possible with a resolution comparable to the slow scan mode measured signal even well above the preamplifier cutoff. When taking [19,20]. fast images with pixel frequencies above the preamplifier cutoff fre- quency, the preamplifier will increasingly act as an integrator. This 2. Limitations of the first prototype effect can partially be removed a posteriori by the use of a lateral gra- dient filter in order to recover images with standard constant-height The first prototype of the FAST module, described in detail in re- contrast. When studying surface dynamics, such as diffusional jumps or ference [18], suffered from a number of limitations. First of all, the molecule rotations, the accessible time resolution is, however, much whole module was based on an embedded personal computer (PC) (PC/ lower than the pixel frequency: if the statistical analysis strictly requires 104 architecture) hosting one Field Programmable Gate Array (FPGA) the detection of single events, moving features have to be detected, on and custom made (digital to analog converter) DAC and (analog to average, at least in two subsequent images; the achievable resolution is digital converter) ADC boards, mounted on two Industry Standard Ar- thus less or equal half the image frequency. chitecture (ISA) bus slots. The nominal data transfer rate of ISA bus, A slower, multi-functional input/output board (model NI PXIe-6341 about 8 MB/s, together with 10 Mbit/s of the Ethernet connection im- or NI PXI-6221) is used for the generation of the slow scan signals and posed two major bottlenecks to the data transfer rates; moreover, while the acquisition of other useful physical quantities, such as sample the generation of the scanning signals was carried out by the FPGA temperature or tunneling bias. The generated scan voltages are then controlling the DACs, the data acquisition chain was affected by un- routed into a new high voltage summation device based on APEX PA90 predictable interrupts from the operating system and by the small size amplifiers with a ± 160 V output range and up to 20 kHz bandwidth of the on-board memories, leading to random loss of data points. and then added to the corresponding signals from the SPM control unit. The 8 bit resolution of the DACs was also a strong limitation to the In addition to the x and y summation channels, a new channel is achievable imaging quality: the harmonic distortion of the generated dedicated to the z direction and the scan voltages can be optionally fast sine waves was large enough to cause instabilities and distortions in amplified (up to 4.8 times) to reach the desired scan range. The output the images, since clear improvements could be seen by “cleaning” the of the summation device is then sent to the piezoelectric motor of the scan waves with strong band-pass filtering. On the ADC side, the 12 bit, SPM head. The response signals from the SPM are finally acquired by 10 MS/s sampling was also a limiting factor: in fact, given that fast the control electronics of the FAST module. A passive attenuator is imaging requires pixel acquisition frequencies on the order of hundreds optionally used to reduce the typical ± 10 V output signal range of the of kHz, the bandwidth of commercial preamplifiers (e.g. about 200 kHz SPM control unit to comply with the 1 V peak to peak input of the ADCs. with 108 V/A gain) is well below this requirement, and thereby strong A high-level LabVIEW software running on a host PC drives the oversampling is required to recover a useful signal when working above control electronics and provides a dedicated graphical user interface the preamplifier cutoff. (GUI) for configuring the scan parameters and visualizing and storing Moreover, the sample tilt correction in the direction orthogonal to the acquired data. To limit the need of resources on the host PC, rather the sample surface was carried out manually with a custom-built analog than displaying the movie in real time, a selected frame rate (usually box, where correction parameters were tuned by means of analog po- few frames per second) is set for data visualization. A high-efficiency tentiometers, making it difficult to re-use parameter sets that were integrated software suite is needed to account for real time processing known to be working. and storing of the data acquired with the FAST module. As an example, We should also mention that a proper, open and extensible data the stream of data during the acquisition of a 200 × 200 pixels movie at analysis framework was also missing, making it difficult to implement an image rate of 100 frames per seconds with a resolution of 14 bit custom data exploration tasks. amounts to 56 Mbit/s. In order to conveniently and efficiently store this Finally, although the portability of the FAST module between dif- data, movies are saved into hierarchical data format (HDF5) together ferent STMs and its extension to AFMs was proposed as a possible de- with their relevant metadata. A dedicated open source Python-based velopment, its practical realization was never performed. software (see Section 8) was developed to visualize, process, analyze and export the movies. 3. Design and specifications of the new FAST module The control of the FAST module is entirely provided through an integrated hardware/software framework allowing the reproducibility The new FAST module, albeit based on the exact same concept, is and stability required for reliable data acquisition to be achieved. In completely new from the hardware, firmware and software perspective, particular, the adoption of an FPGA-based architecture guarantees a implementing in a reliable and user-friendly way many of the technical reproducible and stable delay between probe movement and data ac- solutions adopted in the previous prototype. The module is composed of quisition. At the same time, the availability of both slow and fast inputs an acquisition and control electronics and a high-voltage summation gives the full freedom to add fast dithering/scanning signals and to device (see Fig. 1). The control unit consists of a commercial PXI system track slow drift issues. by National Instruments hosting an FPGA-based high-speed re- As can be seen from the schematics of Fig. 1, the new FAST module, configurable input/output board (model NI PXI-7951). A NI-5781 as its predecessor, is completely transparent with respect to the existing adapter module provides two DACs for the generation of the fast scan electronics: when the output from the FAST control unit is set to zero, waveforms (with an output range of 2 V peak to peak) and two ADCs for the SPM can be driven by the conventional acquisition software. At the 50 C. Dri, et al. Ultramicroscopy 205 (2019) 49–56 Fig. 1. Schematics of the FAST module working principle (left). Connection scheme of the FAST module for the FastSTM/AFM and atom tracking operation modes (right). same time its design allows a seamless switch between the fast mode The fast x and z signals are identically generated, the latter with a and the traditional, slow imaging mode. variable amplification and an additional phase shift. In this case, the In the following sections, the functionality of the FAST module will control of the z signal amplification is used to provide a slope com- be described in detail for its different operation modes, namely pensation to overcome the unavoidable sample tilting, while the phase FastSTM, FastAFM and atom tracking. shift, as pointed out for the signal acquisition, is required to precisely match the actual movement along the x direction. In this new version of 4. FastSTM the module, the phase shift and amplitude of the z signal - which were previously manually controlled through an external analog module - are In the FastSTM mode, the high-speed DACs are used to output the integrated into the control software, allowing the user to easily re-use fast x and z scan waveforms, while the slow y scan signal is generated working parameters sets and thereby improving the reproducibility of by the conventional DAC board. As already pointed out previously [18], the measurements. the generated scan waveform for the fast signals, whose frequency can The frequency response of conventional feedback control electro- reach up to 20 kHz, has to describe a sine function as accurately as nics lies well below 20 kHz. For this reason, scans with the FAST possible in order not to excite uncontrolled resonances of the scanner. module are performed in quasi-constant height mode, i.e. a relatively Furthermore, to avoid inducing transient responses due to abrupt low value of the feedback loop is set. In this condition, along the fast changes in the probe movement, the scan amplitude is smoothly scan direction, the probe moves at a constant height (besides the ad- ramped up to the setpoint at the beginning, and down to zero at the end ditional slope correction), the feedback acting as an integrator, while of the probe scan. For the slow scan signal we instead use a simple along the slow scan direction the feedback accounts for any sample triangular waveform, since its frequency during the use of the FAST tilting and thermal drift. Once a region of interest has been found, the module usually does not exceed 50 Hz. scan in the conventional SPM software is paused and the FAST module When scanning at a speed much higher than the one for which the operation can be started. The high-speed ADCs are used to acquire the scanner has been designed, the phase of the probe's mechanical re- tunneling current, which is, in this case, the response signal of interest. sponse changes with respect to the phase of the excitation signal, due to Indeed, by acquiring the logarithm of the tunneling current, a good the combined effects of an increased hysteresis of the piezoelectric approximation of the actual topography can be obtained. Any phase scanner and the approaching of its resonance regimes. Moreover, the shift between the generated scan signals and the measured tunneling different paths of the signals inside the electronics can introduce ad- current response is manually corrected through a pixel shift control of ditional delays. To take these effects into account, the FAST module is the visualized data in the GUI. When using a high scan speed, the provided with a tunable phase shift between the signal acquisition and scanner is brought close to resonance conditions and the amplitude of the fast x signal generation, which is taken as a reference. its response is no longer linear with the applied voltage. A proper 51 C. Dri, et al. Ultramicroscopy 205 (2019) 49–56 calibration procedure is therefore required to know the correct image images with sufficient resolution is typically rather slow: 256 lines size, which, again, needs to be performed just once for a specific set of scanned at 1 Hz (1 s per line) result in more than 4 min per image. scanning parameters. Besides the conventional a posteriori recognition Sometimes, the surface can irreversibly evolve during image acquisition of features of known periodicity on the acquired image, the calibration due to the interaction with the tip especially in the case of soft samples. can be conveniently accomplished a priori by simply changing the Thus, there is a real need to reduce the acquisition time for practical scanner position by a known quantity through the conventional scan- purposes and also to open the possibility to study dynamical processes ning software while imaging a certain feature with the FAST module. in real time and operando. As mentioned above, one of the strengths of the FAST module is to Recent progresses in high-speed AFMs have been achieved since the enable time-resolved SPM studies, without stringent requirements on first report in 1991 [10] thanks to the development of ultra-short the microscope/scanner design. This allows, for example, its im- cantilevers [11], which exhibit resonance frequencies of 1 MHz and plementation in experimental setups already optimized for in-situ stu- above, higher than those typically used in dynamic mode, and to the dies, capable then to work at high temperatures and pressure, as well as development of dedicated high-speed scanners, including counter- in electrochemical environments. In this context, the new FastSTM has balanced [12], flexure-based [13] and MEMS-based scanners [14], to been already successfully exploited for the investigation of a variety of mention but a few, together with optimized fast amplitude detectors dynamical processes occurring at surfaces, such as catalytic reactions and feedback electronics [15]. Commercial AFMs designed to operate in [21,22], assembly of supramolecular networks [23] and diffusion of fast scan mode exhibit frame rates typically in the 10 frames per second supported metal clusters [24]. These recent results demonstrate the (fps) range, as claimed by the manufacturers. actual possibility to study surface processes at the atomic scale in real- We applied the FAST module to a commercial AFM instrument, time, providing new insight and understanding on the dynamics oc- Keysight Technologies’ 5500 (formerly Agilent Instruments [PicoPlus] curring in the millisecond time scale. 5500), normally working at 0.004 fps for a typical image with good The example reported in Fig. 2 shows two frames of a FastSTM resolution, to operate it in fast scan mode. Frame rates above 8 fps in movie acquired on a self-assembled monolayer of tetra-aminophenyl static (contact) mode and above 1 fps in dynamic (amplitude modula- porphyrin molecules on Au(111). The consecutive frames show the tion) mode were consistently achieved on silicon and PS-PMMA block collective dynamics of two rows of molecules at room temperature on a copolymer reference samples, with a pitch of about 38 nm [26]. Com- time scale of 100 ms. This example shows two important features of the pared to state-of-the-art instruments, our test system exhibits limited FAST module. First of all, it is possible to image organic molecules that performance in terms of (i) photodetector bandwidth (∼ 910 kHz), can be weakly bonded to a surface, with limited scanning-induced which hinders the use of cantilevers with resonance frequencies above structural perturbations, thanks to the ability to image also at relatively 1 MHz, and (ii) the use of a laser spot that is too large for ultra-short low tunneling currents. Secondly, it can be clearly seen that the imaging cantilevers. In spite of these technical limitations, it can be consistently resolution is comparable with that typically achievable in the slow operated in the fast scan mode, as shown below. scanning mode, allowing to observe also the intra-molecular structure In our experimental setup, a Keysight N9447A breakout box is used of the molecules. to exchange scan signals between the FAST module and the host AFM system using standard 50 Ω BNC coaxial connectors. The AFM high voltage positioning signals are branched out and fed into the FAST 5. FastAFM module, in which the fast scan signals are modulated onto the posi- tioning “carrier” signal and then routed back to the AFM system in AFMs, where the deflection of a cantilever beam due to the attrac- order to be applied to the piezoelectric motor position, just as the tive/repulsive interactions between a sharp tip and a surface is de- carrier signal alone would be. The measurement variable used for the tected, can be used in different environments (vacuum, air, gas and fast system depends on the chosen mode of AFM operation: (raw) de- liquid) for a plethora of materials with differentiated physical proper- flection for static/contact AFM or amplitude and/or phase signal for ties (stiff, soft, conductive, insulating, etc.), thus covering a vast range dynamic/tapping AFM, which are accessible via 50 Ω BNC connectors of scientific and technical applications in materials sciences, micro- located either on the Head Electronics Box or AC controller box of the electronics, biology and metrology, among others [25]. In the vast AFM system. In the setup arrangement one computer is dedicated to run majority of cases, atomic resolution is not a requirement in the multi- the Keysight AFM software (PicoView 1.20.3) thus providing a com- disciplinary AFM community. Instead, the focus lies on the meso-scale, plete, independent conventional AFM system. A second PC runs the with nanometer-size features (lateral dimensions typically above FAST control and acquisition software. 10 nm) and sampling areas usually above 1 × 1 µm2. Acquisition of We have used the Keysight N9524A multipurpose scanner which has nominal lateral and vertical scan ranges of 90 µm and 7 µm, respec- tively, for the available drive voltages of ± 215 V per axis that the Keysight controller can deliver. The piezoelectric actuator is mounted in a frame, which holds the optical system (laser and photodetector) and a nose cone with a window conceived to work in liquid environ- ment and the cantilever holder. The spectral response of the scanner to frequencies up to 1 kHz exhibits a resonance at about 420 Hz. The sample slope compensation is a key factor to consider for AFM since the target scan widths are noticeably larger than those used in STM under atomic resolution conditions. If such a slope cannot be ef- ficiently compensated and for samples with topography features in the low nanometer range, the feedback cannot be completely disabled, with a resulting penalty in terms of acquisition time. The FAST system does Fig. 2. Collective motion of two rows of tetra-aminophenyl porphyrin (TAPP) not provide a feedback loop for vertical position control of the scanned molecules on Au(111) (encircled in yellow) in consecutive FastSTM frames. Red probe. Just like in the FastSTM setup described in the previous section, lines serve as guides to better visualize the shift (110 × 110 pixels, 1.1 kHz fast the FastAFM measurements shown below were performed in quasi- scan, 10 fps, 8 × 8 nm2, −1.0 V, 0.15 nA, 300 K). (For interpretation of the constant height mode, in which the influence of the feedback loop of references to color in this figure legend, the reader is referred to the web ver- the original AFM system is minimized to make its response much slower sion of this article.) than the probe movement along the fast scan direction (x), while 52 C. Dri, et al. Ultramicroscopy 205 (2019) 49–56 Fig. 3. Frames from FastAFM movies of (a) a fingerprint-like silicon surface acquired in contact mode and (b) a fingerprint-like PS- PMMA block copolymer thin film acquired in tapping mode. The estimated scanned area is (a) 600 × 600 nm2 and (b) 385 × 385 nm2, respectively, and the line frequency of the fast scan axis is (a) 800 Hz and (b) 150 Hz, re- spectively. (c) Topography image acquired in standard tapping mode of the fingerprint-like PS-PMMA block copolymer thin film. simultaneously being fast enough to compensate for sample tilt along followed with time resolution down to the ms and sub-Å lateral re- the slow scan direction (y) and sample/scanner drift in vertical direc- solution, while being able to simultaneously measure the topographic z- tions. Obviously, the optimal feedback configuration varies with the channel and, optionally, a second channel such as the current/con- chosen frame rate and physical scan range. ductance. The low dithering frequencies and thus the low required Fig. 3(a) shows a frame from a FastAFM movie of the patterned bandwidth in the kHz range allow for measurements with currents silicon sample in air acquired in contact mode at 8 fps and a resolution down to the pA. Atom tracking can thus deliver precious additional of 100 × 100 pixels. In this configuration, the line frequency is information, as a complementary, low-signal technique to FastSTM/ equivalent to 800 Hz and the covered area corresponds to about AFM measurements. The immediate access to displacement rates allows 600 × 600 nm2. The image has been acquired using a PPP-CONTSCR for systematic investigation of the probe influence on the investigated cantilever (Nanosensors) with a nominal force constant of 0.2 N/m. particle dynamics. Bias and tip current dependencies, for example, help Above 8 fps, the images become unstable although the fingerprint to judge the effects of electric field and direct chemical interactions structure can still be identified. Fig. 3(b) shows a frame of a FastAFM [31]. Its seamless integration into the FAST module is therefore of movie obtained from the phase signal when scanning the PS-PMMA fundamental interest. block copolymer sample in tapping mode using a qp-fast cantilever To that purpose, a separate LabVIEW control program for atom (Nanosensors) with a resonance frequency of 871 kHz and a nominal tracking has been developed that uses the input/output connections in a force constant of 80 N/m. The explored area corresponds to about slightly different signal mapping than FastSTM/AFM, as described in 385 × 385 nm2 and was scanned at a frame rate of 1 Hz with a lateral Fig. 1: while the two fast analog output channels are used for gen- resolution of 150 × 150 pixels which corresponds to a line frequency of erating the dither signal, the two fast analog input channels detect the 150 Hz. Because of the variation of sensitivity of piezoelectric actuators topographic z-channel and current/conductance (in the following we with scanning rates, the periodicity of the fingerprint structures was refer just to the STM example, even if the technique can be equivalently used for lateral dimension calibration by comparing images acquired in used for AFM). Two slow analog output channels are used to apply the fast scan mode with those obtained for the same surface in the corre- offset voltages to the x and y piezoelectric actuators quadrants that are sponding standard AFM mode (see Fig. 3(c) acquired at 1 Hz line fre- delivered by the PID controller. Similar to the FastSTM/AFM program, quency). The scans on the patterned Si structures show no morpholo- the data are acquired with the same FPGA code, in blocks of 1 ms gical changes during acquisition (indicating a stable tip and surface). duration that are then analyzed by a digital lock-in. Typically the re- The integral (I) and proportional (P) gains of the feedback were I = 0.5 sulting error signals are subsequently averaged with a moving average and P = 0 in contact mode and I = 0.008 and P = 0 in tapping mode, over 10 values, leading to an overall time resolution of 10 ms. This time respectively. In tapping mode, the trade-off of sampling frequency constant could, in principle, be improved by using an analog lock-in, as versus resonant frequency is crucial so that the number of oscillations long as the interfacing is not conflicting with the user-friendliness. per pixel is satisfactory. In our case, with a cantilever resonance fre- User-friendliness is, however, extremely important for efficient quency of 871 kHz, a line frequency of 150 Hz and 150 pixels, the es- measurements. This concerns, first of all, the interconnection of the timated number of oscillations per pixel is 20. FAST module with the slow, standard STM controller via transistor- transistor logic (TTL) control signals: when repositioning the tip over a particular feature with the slow controller, the dither movement and, 6. Atom tracking after a certain delay of about 100 ms, the lateral feedback start auto- matically, making it possible to jump on fast moving objects to track While FastSTM/AFM measurements usually require elevated signal them. At the end, after the atom tracking measurement, the last applied levels (tunneling currents of several hundreds of pA) due to higher x and y offset voltages continue to be output by the FAST module, while noise levels at increased bandwidth of the preamplifier, a high time the lateral feedback is stopped, in order to keep the observed feature resolution can be obtained as well with low signal levels when well-centered for a subsequent imaging by the slow standard STM switching from the fast scan mode to the atom tracking mode, where controller when it comes to overlay diffusion trace and local topo- the tip simply follows the lateral and vertical movement of a local ex- graphy and to control the actual environment of the diffusing feature. tremum (protrusion or depression) on which it is centered by a lateral, User-friendliness concerns, secondly, the proper setup of the control two-dimensional feedback. This technique, proposed by Pohl and parameters. The frequencies at which the dithering movement is ex- Möller [17], has been implemented for the first time by Swartzentruber cited are close to bending eigenfrequencies of the piezoelectric scanner for the study of adatom diffusion [16]. It is currently included in ad- and might hence distort the movement from an accurate circular trace. vanced scanning probe controllers for accurate drift control [27–29]. In Therefore, a proper regulation of the excitation phase (between x and y this technique, circular dithering of the tip with a typical frequency movement), the excitation amplitudes and of the detection phase is below 1 kHz and a dithering radius of the order of 1 nm and below, necessary. When properly regulated, the error signal in x over a combined with a phase sensitive detection in x and y direction, allows hemispherical protrusion will show a vertical transition between posi- for determining the respective error signals that can be directly applied tive and negative errors and that in y a horizontal one. An example is as input for a lateral feedback via a proportional-integral-derivative given in Fig. 4(a) that shows a Pd12 cluster deposited on a graphene (PID) controller [30]. In this way, the dynamics of single species can be 53 C. Dri, et al. Ultramicroscopy 205 (2019) 49–56 Fig. 4(b) can be measured. In this example, the cluster diffuses over distances of 100 nm in about 30 min, before being caught on the lower terrace side of a step edge, where the cluster movement gets blocked, as indicated by the image taken after the tracking measurement. The diffusion trace is drift-corrected by subtracting a smooth drift vector gained from an analysis of the difference vectors between adjacent x and y values. The drift trace indicates that the diffusing cluster avoids the protruding, non-wettable areas of the moiré structure. Furthermore, the velocity-encoding shows that the diffusion occurs in a very dis- continuous way (with velocities smoothed with a moving average over 1 s), pointing to anomalous diffusion (superdiffusion): once the cluster starts to move, it diffuses notably, far more than one elementary cell of the moiré structure, having difficulties to slow down again. A detail of such a rare dissipative event is shown in Fig. 4(c) (enlargement of the white rectangle in Fig. 4(b)). The movement of the cluster apex that is indicated by the arrows can be resolved with sub-Å lateral resolution and 10 ms temporal resolution. As a guide to the eye, the periodicity of the carbon lattice of the graphene structure is indicated by white dots, while the Pd and Rh atom size is schematically drawn by the three circles - the cluster diameter (footprint of seven atoms, as obtained by calculations [32]) is therefore larger than the displayed image size. Summing up, the FAST module allows for a seamless integration of the atom tracking functionality with a maximum content of pre-con- figured routines that increase the user-friendliness in order to guarantee a successful outcome of demanding measurements. The com- plementarity of atom tracking and FastSTM/AFM measurements opens up the full panorama of SPM investigation techniques to capture par- ticle dynamics: while tracking unveils the dynamics of individual par- ticles, fast imaging unveils the dynamics of several particles and hence of the statistical differences between particles, e.g. due to surface het- erogeneities or isomer distributions. The low signal levels required for tracking open up measurement regimes where the probe influence on Fig. 4. Atom tracking experiment performed with the FAST module: Pd12 the investigated particle dynamics is minimal. cluster diffusing on a periodically wettable moiré lattice of graphene on Rh (111). Imaging conditions: +1 V, 1 pA. (a) Images of cluster topography (left) 7. Portability and transferability of the module and corresponding error signal in the x- (middle) and y-direction (right) upon dithering the tip (800 Hz, 1 nm radius). (b) Overlay of the velocity-encoded atom tracking trace on an STM image obtained after the diffusion ceased on the The FAST module can in principle upgrade any scanning probe lower terrace side of a step edge. The cluster could be traced over 30 min and microscope to video-rate imaging, the only requirement being the 100 nm distance while maintaining a temporal resolution of 10 ms and a lateral possibility to breakout the high voltage signals to the SPM head. The one below 1 Å. (c) Enlarged detail (area indicated by white rectangle in lower module has been initially developed and thoroughly tested on an right half of (b), arrows indicate the overall movement in the numbered se- Omicron VT-STM operated by SCALA electronics, on which a frame rate quence). up to 100 Hz at a resolution of 100 × 100 pixels can be routinely reached. The portability of the FAST module has been successfully moiré film grown on Rh(111). This surface of a periodicity of 3.2 nm is proven also on a similar Omicron VT-instrument including AFM optics periodically wettable, with two sticky sides in each elementary cell, and on an Omicron LT-STM, both driven by Matrix electronics. Atomic where the clusters bind preferentially [32,33]. resolution has been obtained up to a frame rate of 60 Hz (using the LT- Displayed are the topographic z-channel and the x and the y error STM at 77 K) and temperatures up to 780 K (using the VT-STM/AFM at signals of a well-corrected tip movement, such as they can be measured a frame rate of 12 Hz). Preliminary tests have been conducted also on a within the atom tracking program. These images have been taken with SPECS Aarhus VT-STM and an STM driven by Nanotec electronics at RT a specific imaging routine that can be called and that performs a slow reaching frame rates up to 40 Hz. In all cases the integration of the lateral scan, with which the tip and the setting qualities are tested prior FAST module proved to be an easy and rapid procedure, with the re- to the tracking measurement. Since the cluster in this example is two ported performance reached in a matter of hours (see Fig. 5). atomic layers high (i.e. itself a sharp tip), the measuring tip has to be of adequate cylindrical symmetry. The dithering movement of 1 nm dia- 8. Data conditioning and analysis meter is hardly recognizable in the topographic z-channel due to overlapping tip convolution effects. The FastSTM and FastAFM movies are recorded as a one-dimen- In a last step, the feedback parameters have to be optimized. Best sional (1D) raw data stream in an HDF5 file, in order to allow for full results have been obtained with a PI controller, where the integral time control on reconstruction and post-processing operations. To correctly constant was about 10 ms, on the order of the lock-in time constant. As transform the 1D data into a 3D movie stack, we use a specially written a test for the proper functioning and as a means for calibration, defined Python 3 software package, pyfast, which will be described in detail x and y offset steps of some nm width can be induced by the slow elsewhere. Pyfast assigns a location to each original pixel in order to controller. They have to be mapped adequately by the atom tracking map the data stream onto an equally spaced square mesh. To obtain the program - in particular, an apparent curvature of the traces indicates specific coordinates, the movement of the tip is described by a combi- erroneous phase settings. nation of a sinusoidal oscillation in the fast scan direction, a simulta- With a properly adjusted tracking setup, diffusion traces such as in neous triangular movement in the slow scan direction that additionally includes a piezoelectric creep at the turnaround point, and a continuous 54 C. Dri, et al. Ultramicroscopy 205 (2019) 49–56 same module has been successfully used to drive a commercial AFM, demonstrating the potential interest of the device for an extremely large user community. We achieved frame rates up to 100 fps with FastSTM and 8 fps with FastAFM. In the case of FastSTM the highest achievable speed is limited mainly by the cutoff frequency of the preamplifier, thus virtually al- lowing to achieve higher frame rates. In addition, the module can be used to track the motion of particles with the tip by atom tracking, i.e. dithering the tip in x and y to keep it on top of a protrusion or depression, thereby increasing the time re- solution even further. This versatile and simple approach paves the way for dynamical studies with standard SPM instruments down to the millisecond time scale for the general user community. This can be of real interest to many laboratories around the world which use these instruments on a daily basis, with the added advantage that the upgrade to the fast scan mode can be achieved without any modification of the existing hardware and of the control electronics. Acknowledgements The FAST project has received funding from the EU-H2020 research and innovation programme under grant agreement no. 654360 NFFA- Europe. The FAST module has been developed to be at public disposal and is hence also commercially available [35]. The ICN2 is funded by Fig. 5. Frames from FastSTM movies acquired with the FAST module coupled the CERCA program/ Generalitat de Catalunya. The ICN2 is supported to different instruments. (a) Self-assembly of tetra-aminophenyl porphyrins on by the Severo Ochoa program of MINECO (Grant SEV-2017-0706). We Au(111) at 77 K taken with an Omicron LT-STM at 61 fps. (b) Polymerization of acknowledge S. Gottlieb, M. Fernández-Regúlez and F. Pérez-Murano ethene taken with an Omicron VT-STM/AFM, 4 fps, 14 × 14 nm2 [34]. (c) for providing us with the silicon and block copolymer samples. Work at Graphite at RT taken with an STM driven by Nanotec electronics at 40 fps. TUM was supported by the Deutsche Forschungsgemeinschaft (Re- search grants ES 349/1-2 and HE 3454/18-2). BAJL acknowledges a thermal drift. The mapping then occurs via linear or cubic interpolation Research Fellowship from the Alexander von Humboldt Foundation and taking into account the lateral positions frame-by-frame. In this way, a Marie Skłodowska-Curie Individual Fellowship under grant Clus- movies can be created that utilize 100% of the recorded data points by terDynamics (no. 703972) from the EU Horizon 2020 Programme. interlacing - i.e. weaving into each other - the forward and backward lines into a single frame with twice the number of data points. Equally, Supplementary materials the frame rate is maximized by accurately correcting the distortion between upward and downward frames. Supplementary material associated with this article can be found, in In addition to the reconstruction of the image stack, pyfast includes the online version, at doi:10.1016/j.ultramic.2019.05.010. some 1D and 2D filters which remove noise and achieve plane correc- tion by filtering the frequencies of the fast and slow scan directions. References Care is taken not to reduce the time resolution of the movie when fil- tering. In a final step, the movie can be exported into an H.246/MP4 file [1] Data from Scopus retrieved on March 2017 with the following generic query string: with variable compression, or alternatively into images of single frames KEY (Scanning Tunneling Microscopy OR Scanning Probe Microscopy OR Atomic Force Microscopy) AND DOCTYPE (Ar). in common image file formats. [2] J. Wintterlin, J. Trost, S. Renisch, R. Schuster, T. Zambelli, G. Ertl, "Real-time STM Compared to the FastSTM/AFM evaluations, the evaluation of the observations of atomic equilibrium fluctuations in an adsorbate system: O/Ru atom tracking data is quite simple. The trajectory is recorded as a time- (0001), Surf. Sci. 394 (1–3) (1997) 159–169 https://doi.org/10.1016/S0039- sequence of x and y coordinates and requires exclusively a drift sub- 6028(97)00604-3. [3] M.J. Rost, G.J.C. van Baarle, A.J. Katan, W.M. van Spengen, P. Schakel, W.A. van traction for the x, y and z data. This drift correction is performed by Loo, T.H. Oosterkamp, J.W.M. Frenken, Asian J. Control 11 (2) (2009) 110–129 subtracting a continuous directed trend that can easily be discerned https://doi.org/10.1002/asjc.88. from random walk movements by appropriate smoothing of difference [4] M.J. Rost, L. Crama, P. Schakel, E. van Tol, G.B.E.M. van Velzen-Williams, C.F. Overgauw, H.t. Horst, H. Dekker, B. Okhuijsen, M. Seynen, A. Vijftigschild, values and thresholding. P. Han, A.J. Katan, K. Schoots, R. Schumm, W. van Loo, T.H. Oosterkamp, J.W.M. Frenken, Rev. Sc. Instrum. 76 (5) (2005) 053710https://doi.org/10.1063/1. 1915288. 9. Conclusions and outlook [5] R. Curtis, T. Mitsui, E. Ganz, Rev. Sci. Instrum. 68 (7) (1997) 2790–2796 https:// doi.org/10.1063/1.1148196. We have presented the new version of FAST, an electronics module [6] H.J. Mamin, H. Birk, P. Wimmer, D. Rugar, J. Appl. Phys. 75 (1) (1994) 161–168 capable of accelerating commercial scanning probe microscopes to https://doi.org/10.1063/1.355877. [7] G. Schitter, M.J. Rost, Scanning probe microscopy at video-rate, Mater. Today 11 video frame rates and beyond. The new module, as its prototype, (2008) 40–48 Elsevier BV https://doi.org/10.1016/s1369-7021(09)70006-9. transparently uses LabVIEW components and a summation board which [8] H. Junkes, H.-J. Freund, L. Gura, M. Heyde, P. Marschalik, Z. Yang, International are inserted between the standard control electronics and STM or AFM Conference on Accelerator and Large Experimental Control Systems (16th), 2017 https://doi.org/10.18429/JACoW-ICALEPCS2017-THPHA154. microscopes, but now the LabVIEW software package provides a user [9] F. Besenbacher, E. Lægsgaard, I. Stensgaard, Fast-scanning STM studies, Mater. friendly, simple yet comprehensive interface to measure video rate Today 8 (5) (2005) 26–30 https://doi.org/10.1016/S1369-7021(05)00843-6. movies at a location that is pre-defined in the standard control software. [10] R.C. Barrett, C.F. Quate, High-speed large-scale imaging with the atomic force microscope, J. Vac. Sci. Technol. B 9 (2) (1991) 302. American Vacuum Society The measured movies are processed and converted into common video https://doi.org/10.1116/1.585610. and image formats by a custom Python software package. We have [11] D.A. Walters, M. Viani, G.T. Paloczi, T.E. Schaeffer, J.P. Cleveland, M.A. Wendman, demonstrated the versatility of the FAST module, showing data re- G. Gurley, V.B. Elings, P.K. Hansma, Atomic force microscopy using small canti- levers, in: T.A. Michalske, M.A. Wendman (Eds.), Micromachining and Imaging, corded on different commercial STM instruments. Furthermore, the 55 C. Dri, et al. Ultramicroscopy 205 (2019) 49–56 edited by, SPIE, 1997, https://doi.org/10.1117/12.271227. investigating the energetics of small supported metal clusters, J. Phys. Chem. C 122 [12] I.S. Bozchalooi, K. Youcef-Toumi, D.J. Burns, G.E. Fantner, Compensator design for (39) (2018) 22569–22576 https://doi.org/10.1021/acs.jpcc.8b06866. improved counterbalancing in high speed atomic force microscopy, Rev. Sci. [25] G. Haugstad, Atomic Force Microscopy: Understanding Basic Modes and Advanced Instrum. 82 (11) (2011) 113712https://doi.org/10.1063/1.3663070. Applications, John Wiley & Sons, Inc, 2012, https://doi.org/10.1002/ [13] G.E. Fantner, G. Schitter, J.H. Kindt, T. Ivanov, K. Ivanova, R. Patel, N. Holten- 9781118360668.fmatter. Andersen, J. Adams, P.J. Thurner, I.W. Rangelow, P.K. Hansma, Components for [26] M. Fernández-Regúlez, L. Evangelio, M. Lorenzoni, J. Fraxedas, F. Pérez-Murano, high speed atomic force microscopy, Ultramicroscopy 106 (8–9) (2006) 881–887 Sub-10 Nm resistless nanolithography for directed self-assembly of block copoly- https://doi.org/10.1016/j.ultramic.2006.01.015. mers, ACS Appl. Mater. Interfaces 6 (23) (2014) 21596–21602. American Chemical [14] E.C.M. Disseldorp, F.C. Tabak, A.J. Katan, M.B.S. Hesselberth, T.H. Oosterkamp, Society (ACS) https://doi.org/10.1021/am506600m. J.W.M. Frenken, W.M. van Spengen, MEMS-based high speed scanning probe mi- [27] P. Rerkkumsup, M. Aketagawa, K. Takada, Y. Togawa, N.T. Thinh, Y. Kozuma, croscopy, Rev. Sci. Instrum. 81 (4) (2010) 043702https://doi.org/10.1063/1. Highly stable atom-tracking scanning tunneling microscopy, Rev. Sci. Instrum. 75 3361215. (2004), pp. 1061–1067. AIP Publishing https://doi.org/10.1063/1.1651637. [15] G. Schitter, F. Allgöwer, A. Stemmer, A new control strategy for high-speed atomic [28] P. Rahe, J. Schütte, W. Schniederberend, M. Reichling, M. Abe, Y. Sugimoto, force microscopy, Nanotechnology 15 (1) (2004) 108–114 https://doi.org/10. A. Kühnle, Flexible drift-compensation system for precise 3D force mapping in se- 1088/0957-4484/15/1/021. vere drift environments, Rev. Sci. Instrum. 82 (2011) 063704AIP Publishing [16] B.S. Swartzentruber, Direct measurement of surface diffusion using atom-tracking https://doi.org/10.1063/1.3600453. scanning tunneling microscopy, Phys. Rev. Lett. 76 (3) (1996) 459–462. American [29] http://www.specs-zurich.com/upload/cms/user/Specs-AT4.pdf. Physical Society (APS) https://doi.org/10.1103/physrevlett.76.459. [30] A. van Houselt, J. Harold, W. Zandvliet, Colloquium: time-resolved scanning tun- [17] D.W. Pohl, R. Möller, Rev. Sci. Instrum. 59 (6) (1987) 840–842 https://doi.org/10. neling microscopy, Rev. Mod. Phys. 82 (2) (2010) 1593–1605. American Physical 1063/1.1139790. Society (APS) https://doi.org/10.1103/revmodphys.82.1593. [18] F. Esch, C. Dri, A. Spessot, C. Africh, G. Cautero, D. Giuressi, R. Sergo, [31] B.S. Swartzentruber, A.P. Smith, H. Jónsson, Experimental and theoretical study of R. Tommasini, G. Comelli, The FAST module: an add-on unit for driving commercial the rotation of Si ad-dimers on the Si(100) surface, Phys. Rev. Lett. 77 (12) (1996) scanning probe microscopes at video rate and beyond, Rev. Sci. Instrum. 82 (5) 2518–2521. American Physical Society (APS) https://doi.org/10.1103/physrevlett. (2011) 053702https://doi.org/10.1063/1.3585984. 77.2518. [19] C. Dri, F. Esch, C. Africh, G. Comelli, How to select fast scanning frequencies for [32] B. Wang, B. Yoon, M. König, Y. Fukamori, F. Esch, U. Heiz, U. Landman, Size- high-resolution fast STM measurements with a conventional microscope, Meas. Sci. selected monodisperse nanoclusters on supported graphene: bonding isomerism, Technol. 23 (5) (2012) 055402https://doi.org/10.1088/0957-0233/23/5/055402. and mobility, Nano Lett. 12 (11) (2012) 5907–5912. American Chemical Society [20] Patents US8726409B2 and EP2428804B1. (ACS) https://doi.org/10.1021/nl303319f. [21] L.L. Patera, F. Bianchini, G. Troiano, C. Dri, C. Cepek, M. Peressi, C. Africh, [33] Y. Fukamori, M. König, B. Yoon, B. Wang, F. Esch, U. Heiz, U. Landman, G. Comelli, Temperature-driven changes of the graphene edge structure on Ni Fundamental insight into the substrate-dependent ripening of monodisperse clus- (111): substrate vs hydrogen passivation, Nano Lett. 15 (1) (2014) 56–62 https:// ters, ChemCatChem 5 (2013), pp. 3330–3341 Wiley https://doi.org/10.1002/cctc. doi.org/10.1021/nl5026985. 201300250. [22] L.L. Patera, F. Bianchini, C. Africh, C. Dri, G. Soldano, M.M Mariscal, M. Peressi, [34] B. Wang, M. König, C.J. Bromley, B. Yoon, M.-J. Treanor, J.A. Garrido Torres, G. Comelli, Real-time imaging of adatom-promoted graphene growth on nickel, M. Caffio, F. Grillo, H. Früchtl, N.V. Richardson, F. Esch, U. Heiz, U. Landman, Science 359 (6381) (2018) 1243–1246 https://doi.org/10.1126/science.aan8782. R. Schaub, Ethene to graphene: surface catalyzed chemical pathways, inter- [23] L.L. Patera, Z. Zou, C. Dri, C. Africh, J. Repp, G. Comelli, Imaging on-surface mediates, and assembly, J. Phys. Chem. C 121 (17) (2017) 9413–9423 https://doi. hierarchical assembly of chiral supramolecular networks, PCCP 19 (36) (2017) org/10.1021/acs.jpcc.7b01999. 24605–24612 https://doi.org/10.1039/C7CP01341H. [35] https://fastmodule.iom.cnr.it. [24] B.A.J. Lechner, F. Knoller, A. Bourgund, U. Heiz, F. Esch, A microscopy approach to 56