We present the design of a variable temperature setup that uses a pulse tube cryocooler to perform break-junction experiments at variable temperatures ranging from 12 K to room temperature. The use of pulse tube coolers is advantageous because they are easy to use, can be highly automatized, and used to avoid wastage of cryogenic fluids. This is the reason why dry cryostats are conquering more and more fields in cryogenic physics. However, the main drawback is the level of vibration that can be up to several micrometers at the cold-head. The vibrations make the operation of scanning probe-based microscopes challenging. We implemented vibration-damping techniques that allow obtaining a vibration level of 12 pm between the tip and sample. With these adaptations, we show the possibility to perform break junction measurements in a cryogenic environment and keep in place atomic chains of a few nanometers between the two electrodes.
INTRODUCTION
Transport measurements of heat and charge at atomic and molecular scales have benefited tremendously from scanning probe microscopy and related techniques. For example, to probe quantum transport properties, break junction (BJ) experiments based on scanning tunneling microscopy (STM) have become a major working horse to characterize atomic and molecular junctions. In STM-BJs, two electrodes, the metal tip and the metal surface, are brought in contact and then slowly separated. In this way, it is possible to slowly thin down their contact to the atomic level and pull out from the substrate a single atom or even linear chains of atoms, which can then be studied in terms of their physical properties. In break junctions, quantization effects are routinely used as models to study electrical, thermal, and mechanical properties down to the atomic scale at different temperatures.1–5 Because of a clear quantum nature and their atomic size, those junctions are often referred to as “quantum point contacts” (QPCs).
Transport physics involves various effects, ranging from hopping to tunneling, quantum interference, and magnetic correlations. To clearly separate and characterize these effects, it is often necessary to vary the temperature of the experiment over a wide range, from above room temperature to cryogenic temperatures. Bath cryostat setups allowing variable temperature operation consume considerable amounts of liquid helium, which is both costly and demanding for the operator. Pulse tube coolers (PTCs), in contrast, recycle helium in a closed-cycle operation. PTCs require less handling, are most cost-efficient, and can be remotely controlled, compared to standard wet refrigerators. However, even though they can be built without moving parts at low temperatures, the main drawback with pulse tubes is indeed the level of vibration, which, in most cases, is about several micrometers at the cold end. The formation and study of QPCs by STM-break junction techniques require a significant level of mechanical stability (below 1 Å) that is in contrast with the vibration levels of PTCs, for which micrometer level amplitudes are considered already ambitious,6,7 a prohibitive level in the case of sensitive techniques, such as scanning probes.
The reason is that pulse tubes rely on a periodic compression of the working gas, usually helium, inside the refrigerator. The periodic change in pressure causes a periodic elastic deformation (“breathing”) of the tubes inside the cryostat, which results in mechanical oscillations at the cold end.
To address this issue, mechanical decoupling schemes have been proposed, in which the cold end of the pulse tube is thermally coupled to the experimental site (such as the STM head) through compliant springs, minimizing mechanical interaction.8–10 A completely different approach is the employment of two pulse tube units that are 180° out of phase with each other to self-cancel the amplitude of the low-frequency oscillations.11 More sophisticated solutions rather rely on the combination of passive damping methods with active feedback control either directly at the experimental site, e.g., the feedback loop in a AFM system,9 or simply at the cold-head stage.12,13
Here, we report on the construction of a variable temperature setup that uses a pulse tube refrigerator to reliably perform STM-break junction experiments at vibration levels in the picometer regime by solely using passive damping techniques with a minimum working temperature of 12 K. The proposed design can be combined with conventional mechanical shielding, such as optical tables or concrete slabs (like in the present case).
DESCRIPTION OF THE SETUP
Setup
The overall schematic of the laboratory and the instrument is shown in Fig. 1, and photographs of the setup are shown in Fig. 2. The setup involves vibration reduction measures at different scales. First, it is important to physically decouple the noisy mechanical parts, i.e., the pulse-tube cooler and the helium pressure valves, from the sensitive parts, i.e., the microscope stage that is housed inside the vacuum chamber. This oftentimes necessitates considering the floor of the setup carrying the vacuum chamber as a sensitive part of the setup, from which the pulse tube should be decoupled. In the present case, the setup is located inside one of the IBM “Noise Free Labs” to guarantee a superior isolation from the main sources of mechanical and electromagnetic noise through the floor.14 Of particular interest, in this case, is the vibration isolation [see Fig. 1(a)]. The laboratory floor is split into two segments. The first one (5a) is a heavy slab of around 30 ton of concrete that is mechanically decoupled from the rest of the environment and suspended on actively controlled air pistons (not shown in Fig. 1). The concrete slab is used as a tool base for the experiment. The second segment (5b) is a wooden platform surrounding the concrete slab and attached to the inner walls of the laboratory, with no mechanical link to the concrete block. The wooden platform is walkable by the users and permits them to control the experiment without any interference from a close distance.
Schematic drawing of variable temperature STM-break junction setup, (a) including laboratory arrangement and (b) detailing the microscope head and attachment to the cryocooler. (1) Rotary valve separated from the rest of the cryogenic setup; (2) aluminum crossbar; (3) linear guide system; (4) optical table; (5) concrete slab and walkable wooden frame; (6) edge-welded bellow; (7) feedthrough flange; (8) glass-fiber rods for mechanical connection (yellow); (9) aluminum radiation shield; (10) springs and copper floppy spirals; (11) scanning stage; and (12) heater for variable temperature operation. The figure is not to scale, but reference flanges are DN200CF.
Schematic drawing of variable temperature STM-break junction setup, (a) including laboratory arrangement and (b) detailing the microscope head and attachment to the cryocooler. (1) Rotary valve separated from the rest of the cryogenic setup; (2) aluminum crossbar; (3) linear guide system; (4) optical table; (5) concrete slab and walkable wooden frame; (6) edge-welded bellow; (7) feedthrough flange; (8) glass-fiber rods for mechanical connection (yellow); (9) aluminum radiation shield; (10) springs and copper floppy spirals; (11) scanning stage; and (12) heater for variable temperature operation. The figure is not to scale, but reference flanges are DN200CF.
The selected pulse tube cooler is the PTD4200 from TransMIT GmbH, equipped with a compressor whose input power Pcomp = 2.4 kW and featuring a cooling power of 0.34 W at 4.2 K at the second stage. The main idea driving the whole design is to mechanically decouple the cooler from the microscope head while preserving a good thermal link between the two. In order to do that, first the rotary valve is physically separated from the cryocooler head and mounted on a suspended aluminum frame [Fig. 1, (1)]. A sound-absorbing material (ArmaSound) was used to cover the rotary valve and helium-supplying lines to reduce mechanical and acoustical noise. Second, the cryocooler itself is suspended to a transversal aluminum crossbar attached to the sidewalls of the laboratory (2). The suspension is realized through a linear guide system [Igus, (3)] that connects the cryocooler flange to the aluminum crossbar and allows changing the height of the cryocooler with respect to the vacuum chamber where it will be inserted for the experiment. The vacuum chamber, instead, is placed on a table (4) standing on the isolated concrete slab (5a). The only mechanical link connecting the cryocooler flange to the vacuum chamber is a flexible stainless-steel bellows (6), which provides the first stage of damping. Right below the bellows, there is a flange for the electrical feedthrough (7) and then a supporting flange to which an aluminum shield (9) is mechanically hung by means of four low thermal conductivity glass-fiber rods (8) (diameter ∅ = 6 mm, ). The aluminum shields are gold plated using electroplating to increase the emissivity. The shield is thermally connected to the first stage of the cryocooler at 50 K through flexible copper braids (OFHC grade, ). Like this, the shield is thermally strongly connected and mechanically weakly connected to the first stage. At the same time, the shield is mechanically strongly connected and thermally weakly connected to the supporting flange.
Inside the shield, a first copper platform is attached to the inner part of the shield through another set of the same glass-fiber rods. The copper platform is then thermally linked to the second stage of the pulse tube cooler at 4 K by means of a second series of copper braids like the ones above. This stage is used to thermalize all the wires [24× silver-plated copper twisted pair (34-AWG) plus 4× copper single wires (24-AWG)] by clamping them to six copper bobbins attached to the stage itself. Then, a second copper platform (11), situated below the first one, holds the microscope head and it is mechanically suspended to the upper platform via a combination of stiff rods and custom made Be–Cu springs (10) (spring constant ∼150 N/m at room temperature). In order to provide an optimal thermal link, a braided copper strip (thickness = 0.5 mm), shaped like a spiral to reduce stiffness, goes around each spring, connecting the two stiff rods (10). At the scanning stage (11), a sample carrier (FerroVac SHOMEC13) is placed on a stack of piezoelectric elements consisting of two positioners (Attocube ANPx101) for the coarse motion in the XY-plane and one open loop xyz-scanner (Attocube ANSxyz100) to guarantee sub-nanometer resolution during STM-break junction measurements. The thermalization of the sample carrier is guaranteed by the attachment of an additional flexible copper wire to the scanning stage. Finally, to allow for a variable operating temperature of the microscope, a heater [50 Ω resistor (12)] is attached to the bottom of the scanning copper stage and regulated via a temperature controller (Lakeshore 335). The total mass (mainly the copper base plate) of the load connected to the second stage is ∼1.7 kg.
Cooling operation
The system is able to reach a temperature of 12 K, overall limited by the thermal conduction of the cabling involved and the thermal coupling of the shield to the first cooling stage. However, in the selection of the cooler, the low level of vibration has been preferred over cooling capabilities. In the case under consideration, the anticipated experiments are also rather slow. Therefore, a large copper mass could be used to avoid thermal fluctuations, leading, in return, to relatively long cool down times. Usually, a minimum temperature of 12 K is steadily reached after 9 h of cooling from room temperature. The STM operation does not induce any significant additional heat load and no temperature rise is observed during the operation (Fig. 2).
Picture of the setup with the thermal radiation shield. In the box, a close-up view of the scanning stage suspended via Be–Cu springs.
Picture of the setup with the thermal radiation shield. In the box, a close-up view of the scanning stage suspended via Be–Cu springs.
CHARACTERIZATION METHODS
Vibration
Within the system, there is interest to measure the vibration levels at two locations, at the second stage of the PTC and at the STM head, to probe the relative motion of the tip with respect to a sample surface. The vibration levels at the PTC are measured using accelerometers (Bruel and Kjaer, Type 4506-B-003, BW = 0.3–1200 Hz) mounted directly on the second stage. Measurements were recorded at room temperature directly after starting the PTC operation because sensors mounting are not suited for UHV and low temperature operation. The measured power spectrum of vibration was then numerically integrated twice to calculate the displacement spectrum over a 20 Hz bandwidth [see Fig. 3(a)].
Vibration specifications of the pulse tube cooler measured at the second stage: (a) rms spectrum (the black arrow indicates the main frequency of vibration at 1.2 Hz). The resolution bandwidth is 0.1 Hz. (b) Amplitude of displacement at 1.2 Hz.
Vibration specifications of the pulse tube cooler measured at the second stage: (a) rms spectrum (the black arrow indicates the main frequency of vibration at 1.2 Hz). The resolution bandwidth is 0.1 Hz. (b) Amplitude of displacement at 1.2 Hz.
To measure the vibration amplitudes in the microscope head, we make use of the strong distance dependence of a tunneling current between the STM tip and surface. To this end, we use a gold STM tip and a 150 nm-thick gold layer deposited on a silicon chip. The tip is chemically etched15 from a 250 μm pure gold wire (Goodfellow AU005140), then rinsed in de-ionized (DI) water and isopropanol, and finally dried under nitrogen flow. The gold surface of the chip is cleaned by flame annealing, wire-bonded to a chip carrier, and then loaded into the vacuum chamber.
The net displacement between the tip and the sample is calculated inverting the equation , where z is the distance between the two electrodes, I is the tunneling current, I0 is a suitable parameter, while k is the inverse decay length specific to the material. As k is sensitive to the exact work function of the sample and tip, we measure m = 2k each time before and after performing a new measurement of vibration. The measured values of m range from 0.1 to 0.8 Å−1. A schematic of the measurement technique is depicted in Fig. 4(a). The amplifier used is a low noise current-to-voltage converter (SP983c from Basel Precision Instruments) with a gain of 107 V/A. The acquisition time interval is set to 10 s with a sampling time of 1 ms so that the accessible frequency range spans from 0.1 to 500 Hz. In this way, all frequencies relevant to STM-BJ operation, including the low frequencies typical of pulse tube oscillation of 1–2 Hz, can be probed.
(a) Schematic for the measurement of the net tip–surface displacement. (b) Raw displacement trace, including low frequency drift, from STM current recorded over an acquisition period of 10 s and its relative rms value (σ).
(a) Schematic for the measurement of the net tip–surface displacement. (b) Raw displacement trace, including low frequency drift, from STM current recorded over an acquisition period of 10 s and its relative rms value (σ).
Break junction measurements
Break junction traces are acquired right after vibration measurements by approaching the tip of the microscope into contact with the counter electrode (closing trace) and slowly retracting it apart (opening trace). The closing trace stops when a conductance of is reached. Here, is the quantum of electrical conductance. The speed for both approach and retraction from the surface is set to 7 nm/s.
CHARACTERIZATION OF VIBRATIONAL NOISE AND ITS EFFECTS
Vibration of the cryocooler
For comparison, Fig. 3 displays the measurement of the vibration level of the second stage of the PTC during operation of the PTC and in the idle stage. Figure 3(a) shows the rms-displacement spectrum for the z-axis over a 20 Hz bandwidth. A peak around ∼1.2 Hz, a signature of the pulse tube cooler, is clearly visible, together with higher frequency harmonics. In Fig. 3(b), the mechanical oscillations recorded by the accelerometer are plotted over a 3-s period. For this plot, the low-frequency drift visible in Fig. 3(a) was removed using a high-pass filter to visualize better the vibration level induced by the operation of the PTC. The peak amplitude of the signal vs time is equal to ±72.6 µm and the rms-amplitude to 35.6 µm.
Net displacement
Next, we turn to measurements of the vibration inside the microscope head. A schematic of the measurement technique is depicted in Fig. 4(a). For the measurements, we chose a temperature of 45 K after a week of cooling operation, which represents a typical experimental scenario. A vibration trace of 10 s, with the rms-value as low as 11.8 pm, could be recorded and is presented in Fig. 4(b). We note that vibration levels vary throughout the day but were never worse than 65 pm (see the supplementary material). Those levels are manageable for scanning probe measurements and compare well to other systems using PTC cryocoolers for sensitive measurements. Our system has vibration levels lower than other dry-cryostat setups using single-stage damping techniques, which feature rms-values of 65 pm or higher,8,9 and it is in the same range of vibrations of other setups utilizing more sophisticated multi-stage methods, showing a cumulated displacement of 60 pm over almost a 10 kHz bandwidth.10
We suppose that the reason for such high stability and small net displacement between the tip and sample can be attributed, in part, to the stillness of the lab and to the mechanical decoupling of the cold head from the rest of the setup but mainly to the series combination of springs, empirically selected to have the right elasticity to act as a “low-pass” filter for vibrations and the “high-pass” filter due to the rigid connection of the tip holder and sample carrier to the scanning stage. We note, however, that the comparison between the vibration of the pulse tube’s second stage and the tip–sample gap does not mean that the pulse-tube itself has been reduced in vibration. Instead, we isolated the tip–sample system from the noise created at the second stage of the pulse tube.
BREAK-JUNCTION OPERATION
To demonstrate the capability of the instrument to operate during break-junction experiments, we performed a break junction measurement at 45 K. During the experiment, the tip is repeatedly brought into (closing trace) and out of contact (opening trace) with the sample. The electrical conductance measured during such break-junction events demonstrates the hallmark of quantization steps and the thinning down of the junction down to single-atom diameters. An important criterion for the usability of good transport measurements is the temporal stability of the junction. Ideally, a junction plateau (the conductance being on a defined and narrow range of values) is long enough in time (or pulling distance) to enable detailed measurements.2,16
Because of the great stability of the system, the quantization of the electrical conductance is observable up to 8 G0 (see Fig. 5). Indeed, each quantization step represents the number of atoms (or monoatomic filaments) bridging the two electrodes, each one carrying a single quantum of electrical conductance G0. To check for spurious events, we performed 2300 measurements and combined them in a 1D histogram [see Fig. 6(a)]. More than 90% of the traces (2127 out of 2300) show a plateau of at least five datapoints in the range (0.9–1.1) G0. For those selected traces, we calculate the length of the plateau at G0, showing how the stability of the system allows the formation of long single-atom filaments up to several Angstroms in length with a good percentage of traces () going above 1 nm [Fig. 6(b)]. We note that this distance suggests pulling atomic chains. To estimate the number of atoms in such chains, the reported values for the radius of a gold atom, varying from 1.35 to 1.96 Å,17–19 can be considered, together with a stretched Au–Au bond distance around 3.6 Å.20 Figure 6(b) does not show obvious peaks at exact multiples of the aforesaid values but ultimately illustrates the possibility to uphold nanometric single-atom wires.
One-dimensional histograms showing (a) peaks at multiples of the electrical conductance quantum G0 and (b) length of the plateaus at 1 G0.
One-dimensional histograms showing (a) peaks at multiples of the electrical conductance quantum G0 and (b) length of the plateaus at 1 G0.
OUTLOOK AND CONCLUSION
We have designed a system to maintain a vibration level constantly below 65 pm between a STM tip and a sample, and that under optimal conditions of the lab, this level reduces further down to 11.8 pm. This allows for the measurement of break junctions at variable temperatures with a pulse tube cryostat.
Compared to previously reported systems, the present setup reaches competitive vibration levels with a single-stage damping system and without the need for more elaborated damping solutions, such as cascaded mass-spring systems. Further improvement appears possible using, for example, an eddy-current damper installed between the bottom of the experimental stage and the shield to further reduce vibrations. Possible improvements for reaching lower temperature consist of using more resistive wires for lines that do not carry the signal (piezos and temperature sensors) and/or reducing the diameter of the same. However, with the current modifications, we already reached a level of vibrations that allows us to run a break junction experiment with sub-Å mechanical noise at variable temperatures and extract, from a gold surface, single-atom filaments stretching over more than 1 nm before rupture.
We hope that the description of our STM break-junction experiment in a pulse tube-driven cryostat can also help design other cryogenic setups, especially with the present trend toward dry systems in cryogenics.
SUPPLEMENTARY MATERIAL
The supplementary material contains further datasets [like Fig. 4(b)] on vibration measurements during different times of the day.
ACKNOWLEDGMENTS
The authors gratefully acknowledge support from K. Moselund, R. German, and H. Riel and the Cleanroom Operations Team of the Binnig and Rohrer Nanotechnology Center (BRNC) for their help and technical support. Special thanks go to S. Karg and F. Albrecht for the useful discussions and practical support and to J. Falter and A. Schirmeisen, from TransMIT GmbH for their help and advice during the installation and first operation of the instrument.
This work has received funding from the European Community through the Horizon 2020 Research and Innovation Programs under Grant Agreement Nos. 767187 (QuIET) and 766853 (EFINED) and through the Swiss National Science Foundation under Grant Nos. 189924 and 134777.
AUTHOR DECLARATIONS
Conflict of Interest
The authors have no conflicts to disclose.
Author Contributions
A.G. and B.G. designed the setup with input from A.K.; A.G., A.Z., F.H., and B.G. built the setup; A.G. performed the measurements and analyzed them with input from B.G., M.C., and S.F.; A.G. and B.G. wrote the manuscript with input from all authors.
DATA AVAILABILITY
The raw data of the measurements discussed here are found in the supplementary material.