A robust molecular probe for Ångstrom-scale analytics in liquids

Traditionally, nanomaterial profiling using a single-molecule-terminated scanning probe is performed at the vacuum–solid interface often at a few Kelvin, but is not a notion immediately associated with liquid–solid interface at room temperature. Here, using a scanning tunnelling probe functionalized with a single C60 molecule stabilized in a high-density liquid, we resolve low-dimensional surface defects, atomic interfaces and capture Ångstrom-level bond-length variations in single-layer graphene and MoS2. Atom-by-atom controllable imaging contrast is demonstrated at room temperature and the electronic structure of the C60–metal probe complex within the encompassing liquid molecules is clarified using density functional theory. Our findings demonstrates that operating a robust single-molecular probe is not restricted to ultra-high vacuum and cryogenic settings. Hence the scope of high-precision analytics can be extended towards resolving sub-molecular features of organic elements and gauging ambient compatibility of emerging layered materials with atomic-scale sensitivity under experimentally less stringent conditions.

S ensing objects below the resolution limit of the eye began with the advent of water-based microscopes in 2000 BCE. Following a series of ground-breaking advances, instrument development eventually led to the modern electron and scanning probe microscopes which have not only advanced resolution capability but have enabled new research disciplines from quantum cryptology to single-molecular nanotechnology. Recently, high-resolution transmission electron microscopy 1 , scanning tunnelling microscopy (STM) [2][3][4] and scanning tunnelling spectroscopy 5 have made possible the analysis of a new class of ultra-thin materials. The atomic-scale structure, extrinsic doping, bonding states and chemical composition can be precisely measured for single-atom-thick electronic materials from the two-dimensional (2D) form of carbon, graphene 6 to more recent transition metal dichalcogenides (TMDs) 7 with potential for triggering a new wave of 2D nanodevice technologies.
However, preserving the surface integrity of single-atom-thick materials outside of ultra-high vacuum (UHV) settings, and extracting information with highest possible resolution, where each atom is directly exposed to contaminants is not trivial. Yet the payoff is immense. To take one illustrative example, it would allow rapid fingerprinting of a vast body of theoretically predicted 2D layered materials 8 and simultaneously verify the ambient compatibility of such novel materials 9 and 2D patterned structures 10 . A known alternative to vacuum to protect air-sensitive surfaces, is the liquid-solid interface in which a STM can be operated. This field of research has an almost 30-year history, from early reports on achieving atomic resolution on solid surfaces immersed in water 11 , liquid nitrogen 12 and acidic solution 13 environments, observation of molecular dynamics 14 , decoding molecular layer-underlying surface epitaxial relationship 15 , capturing oxidation catalysis reactions 16 and in investigating through high-resolution STM images the supramolecular chemistry of molecules 17 and pattern formation during molecular self-assembly 18 at the liquid-solid electrical interface with a non-functionalized metal probe. More recently it has been demonstrated that by terminating the apex of a scanning probe (including STM and AFM) with a single molecule, it is possible to further the limits of spatial resolution and enhance chemical contrast of the low-dimensional materials under study [19][20][21][22][23][24][25][26] . However, these single-molecule-terminated scanning probes have been mainly demonstrated to operate at UHV in a temperature scale ranging from 4 to 100 K (refs [20][21][22]25,27,28). The protocols for engineering such molecular probes, in particular bonding effects between the molecule and the metal-apex, have been documented in detail 22,24,[27][28][29][30] . Conversely, the benefits of the molecular probe have not been fully exploited under standard laboratory conditions, mostly because of instabilities in moleculemetal coupling at room temperature. In the present work, we demonstrate that by engineering and operating a single-moleculeterminated Au STM probe in high-density liquids, it is possible to control random fluctuations of the molecule at the metal apex, maintain a clean interface protected from ambient contaminants thereby resulting in a robust single-molecular probe with prolonged lifetime.

Results
Electron tunnelling in a high-density liquid environment. The central concept of our experiment is depicted in Fig. 1, a metallic (Au) STM probe functionalized with a C 60 molecule is brought close to the surface (epitaxial graphene on SiC 31 ) held in a Teflonbased liquid cell filled with silicone oil ([-Si(CH 3 ) 2 O)-] n ). C 60 geometry permits easier lateral manipulation and transfer to the tip from the surface at ambient conditions than, for example, CO, and the protocol for C 60 transfer is well established 19,22,24,32 . Single C 60 -terminated STM tips are routinely used to probe organic materials and C 60 clusters anchored on metal STM tips have also been demonstrated previously, to investigate low-dimensional surface defects 33 . Such molecular cluster-based tips are relatively robust (loss of a single molecule at the apex can be compensated by its nearest neighbour) but interpretation of the observed chemical contrast improvements is not trivial. The mechanism of electron tunnelling from the C 60 -terminated Au probe through a non-polar liquid medium into the conductive graphene layer may be modelled as tunnelling through a vacuum gap where the tunnelling current (I tunnel ) is exponentially dependent on the distance (Z) between the probe and the conductive surface using an effective energy barrier height (f eff ) as shown below, However, the effective energy barrier height (f eff ) has been shown to fluctuate for STM measurements at the conductive liquid-solid electrical interface 34 . In addition to fluctuations related to f eff , the tunnelling probability (both elastic and inelastic) during STM measurements in polar liquids can be influenced by the barrier dielectric response. To exclude these effects, it is necessary to work in a liquid that is electrochemically inert, has a low affinity towards the sample and a low vapour pressure. The dense molecular structure of silicone oil prevents moisture build-up on the solid surface, protects air-sensitive samples 12,35 and minimizes tip and surface contamination 36 . Note that during the tunnelling process (irrespective of bias polarity), the liquid can be squeezed out of the gap and this holds for both polar 37 and non-polar tunnel barriers 12,35,38 .
Engineering and testing the robustness of the molecular probe. Single C 60 -terminated metal STM tips have been fabricated in the past through lateral manipulation and transfer of a surface bound C 60 molecule to a metal apex under cryogenic conditions 22 and later applied to resolve intramolecular details of molecules 19 and applied in atom-scale engineering of single-molecule contacts 24 . We attempt a challenging lateral manipulation technique to engineer a C 60 functionalized Au probe in a liquid environment at room temperature. For this we first locate an isolated C 60 molecule on Au(111) surface in silicone oil liquid cell and the specific molecule was later transferred to the metal tip by controlled lateral manipulation as shown from the schematic steps (step 1-step 4) in Fig. 2a. The tunnel current set point is reduced by nearly 30% during transfer attempt in comparison to the set point applied during normal imaging experiments. To verify whether the lateral manipulation approach was successful, the same region is scanned again. The target molecule indicated by the black arrow in the STM image (Fig. 2b) and after the lateral manipulation (Fig. 2c) experiment in the presence of the liquid medium, serves as a first proof of the transfer of a C 60 molecule from the surface to the Au tip. However, we did not observe any immediate apparent change in the structure of the near-lying C 60 molecule when imaged with a C 60 -Au probe as shown in Fig. 2c, which was recorded at a high speed (scan rate of 30.5 Hz, tip velocity of 15.3 mm s À 1 and 256 Â 256 lines per scan area). This is in contrast to previous low-temperature STM studies of C 60 -terminated tungsten tips 22 , where directly after termination of the surface bound molecule to the metal apex resulted in visualization of the molecular orbitals of the C 60 when imaged over Au adatoms, which served as sharp reverse imaging sites. In our study, upon reducing the scan speed (scan rate of 1.51 Hz, tip velocity of 0.7 mm s À 1 and 512 Â 512 lines per scan area), we obtained an improved visualization of the structural change incurred on the surface bound C 60 molecule ( Fig. 2e), when compared with the featureless structural shape imaged previously for the same molecule with a non-functionalized Au tip (Fig. 2d). After passing this control test, the molecular probe (C 60 -Au STM tip) was then directly transferred to a liquid-cell holder with the epitaxial graphene sample encompassed in silicone oil. Here again we conduct an additional control test to confirm the presence of the molecular probe, which is a reverse imaging experiment at local defect sites on graphene. The observed structural changes were comparable to early C 60 -terminated STM-based reverse imaging studies on defects on a graphite surface 39 , which further validates the presence of the C 60 molecule at the Au tip-apex. (see 'Methods' section and Supplementary Fig. 1, Fig. 2a,b). Note: In total, we have conducted 135 attempts to fabricate the C 60 molecular STM probe through lateral manipulation, of which B80 were successful. This nearly 65% success rate includes tests where we were in the process of finding the optimal tip-sample height distances and bias energy ranges. Generally the failed attempts were due to incidents where the bias energy (Z1.8 V) and/or the current set point (Z250 pA) was too high, during the manipulation attempt, suggesting that the molecule was displaced rather than transferred. Note: The optimal bias energy range and tunnel current set point values apply only during manipulation Step 3 Liquid-cell Step 4 Liquid-cell experiments, the molecular probe after fabrication can be operated in a bias energy range of À 2 V to þ 2 V and under a tunnel current set point range of a few pA to B1.5 nA. However, the best imaging on 2D materials was obtained at low bias and low tunnel current set points and this also extended the lifetime of the molecular probes.
Similar transfer experiments in the absence of the silicone oil liquid medium had a much lower success rate of 10% and even the few molecular probes that we were able to synthesize did not last for more than a few scans under dry conditions, indicating the importance of the encompassing liquid medium in reducing the fluctuations of the molecule at the Au tip-apex. To further quantify the bonding mechanism of the C 60 molecule to the Au tip, we rely on quantum mechanical models (representative structures shown in Fig. 2g-i, also see Supplementary Methods 1 for details of the density functional theory (DFT) calculations) which indicate that the C 60 molecule exhibits a significant preference of (0.35±0.03) eV per molecule for bonding to low-coordination defect sites rather than terrace sites on Au and reveal that the strongest covalent character for C 60 attachment occurs at the apex site (by contrast, terrace sites form predominantly van der Waals attachments to C 60 ).
Atom-scale metrology of graphene systems. Figure 3a is a largearea STM image acquired on a graphitic region. The interface (marked by the black dotted line) separates the monolayer region (marked by the blue arrow) and bilayer region (marked by the green arrow). The area marked exactly at the interface (indicated by the white rectangle in Fig. 3a) is projected in a three-dimensional format as shown in Fig. 3b. The content of the imaged structure (Fig. 3a,b) is in agreement with previous UHV-STM experiments performed on similar graphene systems at 4 K (ref. 40) and high-resolution AFM studies at room temperature 41 . Zooming into the separate regions indicates clearly the bilayer structure (Fig. 3c) and monolayer graphene with a honeycomblike architecture (Fig. 3d). A three-dimensional representation of a similar monoatomic graphene lattice (Fig. 3e) highlights the atomic hillock-and-valley structures, signifying that these 2D materials are not perfectly planar and can have differences in the local corrugations. Imaging on a separate region on the monoatomic graphene lattice, we also observed point defects exhibiting a two-fold symmetry (Fig. 3f) as previously observed on monolayer graphene imaged in a UHV-based STM using nonfunctionalized metal tips 42 . The occurrence of corrugations and point defects, appearing as protrusions on the atomic lattice, is dependent not only on the synthesis mechanism but relates strongly with the registry of these materials to the underlying surface and its roughness 43 . The coupling of the C 60 molecule to the metal-apex within the silicone oil medium is key to the stability of these molecular probes under ambient conditions. Control experiments where the molecular probes were used to image graphene (with similar bias and tunnel current set points) but under dry-conditions lasted for only a few scans (five to eight frame scans at a rate of B15 s per frame, after which the molecule was lost, confirmed through experiments conducted on isolated C 60 molecules and reverse imaging 33 over lattice defects on graphene, see section 1 Supplementary Information). This is in contrast to the longer lifetime of the molecular probes in silicone oil (B60 frame scans of comparable scan areas at a rate of B15 s per frame, and in a few cases longer than 60 frame scans when applied for imaging over graphene regions free of defects and protrusions).
Quantifying bond lengths and molecular fluctuations. liquids. We calculated the mean bond length connecting two adjacent carbon atoms within an atomic hexagon to be 1.41 Å, consistent with previous high resolution STM reports performed under cryogenic conditions 44 . The measured bond length is double the atomic radius of carbon (0.7 Å). Factors such as substrate curvature, topological defects and surface corrugations can induce strain, leading to variations in the experimentally measured bond lengths. The upper limit of the bond length values we report (coded in green in the histogram, Fig. 4b) was mainly measured near point defects on the graphene lattice, particularly in the vicinity of a topological defect region (similar to the constant current STM data on a point defect in Fig. 3f). This is consistent with previous theoretical calculations predicting an asymmetric bond length distribution in the graphene lattice at room temperature 45 and with previous high-resolution transmission electron microscopy analysis of C-C bond length fluctuations in single-layer graphene 46 .
To better understand the role of silicone oil in the formation of a stable molecular probe, we performed classical molecular dynamics simulations to examine the motion of C 60 on the metal-apex in the presence and absence of the liquid. Single C 60 dynamics on the tip-apex is expressed as root-mean-square fluctuations in the C 60 centre of mass, averaged over 4,000 structures sampled (every 10 ps) during the final 40 ns of 50 ns of equilibrated room-temperature dynamics. The molecular dynamics model cell was based on DFT calculations of C 60 bonding to an Au apex site via two Au-C bonds of length 2.2 Å (Fig. 4d). The C-Au bond force constants of 55 N m À 1 are estimated from DFT by shifting the C 60 molecule 0.25 Å closer and nearer to Au in steps of 0.05 Å, consistent with previous reports 30 . Figure 4d is a snapshot of the final computed structure of the C 60 -functionalized Au probe in silicone oil. The computed C 60 dynamics on the Au probe revealed significant reduction of C 60 motion on the Au tip-apex in silicone oil. There is a threefold damping when silicone oil covers the molecular probe (see Supplementary Methods 2 for more details of the molecular dynamics simulations).
Tuning atom-by-atom imaging contrast. Having established a route to resolve with high spatial clarity the covalent bonding sequence at a fixed bias energy, we then posed the question, is it possible to control the imaging contrast at a single-atom level? Figure 5a-d is a series of constant current STM images acquired over a single atomic hexagon spanning an area of 4.2 Å 2 represented in a three-dimensional format, imaged at a fixed tunnel current set point (15 pA) and increasing the bias energy in steps. We observe a marked improvement in the ability of the molecular probe to identify the exact location of the atoms (Fig. 5a-d) within a six-membered carbon hexagonal ring. The location of the atomic sites can be controllably revealed and masked by increasing and decreasing the bias energy at fixed tunnel current. This level of control and ability to tune the imaging contrast, permits direct visualization of the electronic variations in embedded atoms, in this case the inter-atomic bonds within a single hexagon. We believe the high spatial information content upon terminating the Au tip with a C 60 molecule and bias-dependent control of contrast can be attributed to the corresponding alterations in the tip states, magnitude of the bias energy and the resulting changes in the spatial orbital overlap between the molecular probe and monoatomic graphene. The local electrostatic interactions 47 between the molecule and the graphene surface is also a factor that contributes to the enhancement in imaging contrast. To gain a deeper quantitative understanding of the impact of tunnelling parameters on the STM images in a liquid medium, it is requisite to establish a first principle-guided study of electron tunnelling between a molecular probe and graphene lattice in the presence of a silicone oil mesh similar to previous studies on electronic tunnelling through polar liquids 48 .
Although we have been able to previously image the atomic lattice of single-layer graphene in liquids using non functionalized metallic tips 36 , with resolution comparable to the data shown in Fig. 3d, it has not been possible to achieve the resolution and imaging control shown in Fig. 4a and Fig. 5a-d using a bare metallic Au tip in liquids at room temperature. Such high resolution imaging where the exact atomic positions can be identified allows better quantification of bond atomic distances in 2D crystals that is closer to the expected theoretical estimates. To check for any electronic interference from the silicone oil medium in the observed variations in imaging contrast, we performed dispersion-corrected electronic structure calculations on the C 60terminated Au probe in the presence of silicone oil molecules. The C 60 -Au electronic structure does not change when the solvent layer is included (Fig. 5e).
Probing atomic landscape of 2D MoS 2 . In addition to studying the lattice structure of monoatomic graphene with a molecular STM probe in liquids, we demonstrate that this robust probe can also be extended to provide a quantitative analysis at a singleatom level of the structural composition of other 2D layered materials such as TMDs. For this, we use a MoS 2 sample grown by chemical vapour deposition (CVD) on Au(111) (for further details on growth protocols, see 'Methods' section). The 2D-MoS 2 is an electronic material gaining immense interest as a transistor channel material 49 , in biomolecular sequencing 50 , information storage 51 and in analytical chemistry 52 for its built-in semiconducting nature and atomic structure. The as-grown MoS 2 -Au(111) sample is then transferred to the liquid-cell holder filled with silicone liquid and imaged using a single C 60 -Au STM probe engineered using protocols identical to those used to image graphene systems. Figure 6a is a large-area STM image showing regions of MoS 2 layers grown on Au(111) terraces. Single-layer regions are identified by measuring a height of B6.8 Å at the MoS 2 layer and gold interface (within the region marked by the black dotted rectangle in Fig. 6a), as shown in the line section analysis (Fig. 6b). This is consistent with the expected layer thickness of TMD materials that ranges from 6.5 Å to 7.1 Å with a layer of metal atoms trapped between two layers of chalcogen atoms (atomic model, Fig. 6b). The measured layer thickness is also in agreement with previous low temperature STM reports on CVD grown MoS 2 (ref. 53) imaged using a nonfunctionalized metal probe. Positioning the molecular STM probe in an interface region between single-layer MoS 2 and the gold surface reveals the moiré superstructure that stems from the lattice mismatch between the underlying Au(111) surface and the as-grown MoS 2 . A close-up view of this moiré superstructure presented in Fig. 6d and the atomic structure of the moiré patterns is further resolved by zooming within the same region as shown in Fig. 6e. The periodicity of the moiré patterns is measured to be 32.5 Å (Fig. 6e) and the presence of line-type surface defects was also observed (Fig. 6f) consistent with recent STM studies of single-layer MoS 2 epitaxially grown on Au(111) surface 54 . A spatially magnified constant-current STM image of the honeycomb type MoS 2 atomic structure is shown in Fig. 6g. The position of the S atoms in the hexagonal structure is better resolved by the molecular STM probe than the Mo atom sites suggesting that the electron density is dominated by toppositioned S atoms 55 . On the basis of several such highresolution images captured at low scan speeds (1-4 Hz at 512 Â 512 lines), we were able to measure a MoS 2 lattice constant of (3.2±0.1) Å, consistent with previous STM reports under UHV conditions using a non-functionalized STM probe. Coupled with the graphene data, this quantitative information on MoS 2 indicates that a single C 60 -terminated STM tip can be reliably applied for atom-by-atom analytics of 2D nanomaterials, even when operated in liquids at room temperature.

Discussion
Taken collectively, the scanning tunnelling microscopic observations and the molecular simulations provide compelling evidence for the damping of C 60 motion at the probe metal-apex in silicone oil, resulting in a stable molecular probe. Given the growing interest in analysing atomic materials under practical conditions 9,10,41,56 , our findings provide a significant step forward in atom-mapping technology in liquids at room temperature, which can be applied to examine lattice geometry, ambient compatibility and probe the impact of solvents on next generation of 2D semiconducting solids 57 . However, a deeper understanding of the influence of rotational motion of the molecule at the probe-apex will be required before these findings can be fully extended to resolve the intramolecular architecture of organic molecules with an assortment of heterogeneous chemical bonds. To address these open issues, it will require bold efforts in the future in a liquid environment, to decode the orientation of the molecule at the probe-apex 19 to establish a clearer interpretation on the origin of the visualized enhancement of information content from scanning probe images.

Methods
C 60 solution synthesis. C 60 molecules in powder form (purchased from Sigma-Aldrich, 99% purity) were solubilized (bath sonication 30 s) in toluene, and a solution with 0.1 mM concentration was prepared. One millilitre of this solution was drop-casted on a Au(111) surface. The excess solvent was blown dry using N 2 gas and the sample was placed in a liquid cell and immediately covered with silicone oil. The silicone oil used in this study having a chemical formula of ([-Si(CH 3 ) 2 O)-] n ), where n ¼ 22, was purchased from Sigma-Aldrich. The CAS number is 63148-62-9.
STM tip preparation. The tips (Au) used in this work was just mechanically cut with hand-held pliers, followed by rinsing in acetone (5 min bath sonication) and isopropyl alcohol (2 min bath sonication), blow drying with N 2 gas (until dry) and followed by annealing at 50-55 degrees Celsius for 30 min in vacuum oven.
Fabrication and analysis of the molecular STM probe. We compared the current-voltage curves recorded using non-functionalized Au tips and C 60 -terminated STM tips (see Supplementary Fig. 1) and thus confirm (based on the evolution from metal-metal contact to a molecular signature in the IV curve) that a C 60 molecule is adsorbed on the Au STM probe apex. The imaging is then temporarily halted, the molecular STM probe is withdrawn and the Au(111) sample is replaced by the epitaxial graphene on SiC sample within the liquid cell and the molecular probe is brought back into contact with the surface. An additional step to check whether the molecule is still anchored to the apex in between scans of the graphene surface is through reverse imaging in liquids. Where the molecular probe is raster scanned over a sharp defect feature acting as a secondary site on the graphene surface. This is usually a topological defect that acts as a surface tip array which then reflects the image of the molecule adsorbed on the Au tip apex. One such STM image is shown in Supplementary Fig 2a and Fig. 2b, where the point defects on the graphene surface aid in imaging the molecule on the tip. The underlying graphene architecture is visible and the C 60 molecular shape is reflected similar to Supplementary Fig. 2a.
In addition to the lateral manipulation experiment, we have also attempted vertical manipulation, through which the molecule on surface is transferred to the tip. For this, we place the Au tip on top of the C 60 molecule, and vary the bias voltage when keeping the tip-sample height distance constant (fixed tunnel current set point). However, for the current work, we report only the molecules transferred using the lateral manipulation approach. The success of both these experiments, lateral and vertical manipulation, is strongly dependent on the low mobility of the underlying molecule and this is achieved by trapping the molecules in a highdensity liquid such as silicone oil. To remove all other possible sources of environmental fluctuations, the measurements were conducted in next-generation noise-free laboratories, which offer vibration isolation together with precise temperature and humidity control. The specifications used include, active mechanical damping of the platform supporting the STM equipment down to 0.5 Hz, mechanical vibration with velocity less than 500 nm s À 1 (x,y,z) below 16 Hz, electromagnetic fields with flux density less than 5 nT RMS in an integral spectrum between 0 and 625 Hz and temperature stability of 0.1°C h À 1 with fluctuations of 0.5°C per day.
Instrument and materials details. The STM measurements were conducted in constant-current mode using Nanoscope IIIa, E-scanner, Digital Instruments. A low-current pre-amplifier circuit capable of B1 pA sensitivity was used. The lateral STM tip drift rate is r1 nm min À 1 in silicone oil medium. The STM instrument is allowed to scan with bare Au tips on lithographically designed calibration samples (Ted Pella Inc, 677-STM) for 6-8 h to reduce scanner hysteresis before mounting molecular STM probe for actual measurements on graphene. Nanoscale calibration of the scanner was performed using a freshly cleaved highly oriented pyrolytic graphite surface (x, y axes) and a Au(111) on mica substrate (z axis). A Teflonbased custom-built liquid-cell setup was used to support the sample substrates. based on the gas-phase reaction between MoO 3 and high-pure sulfur evaporated from solid phase. Crucible containing B5 mg MoO 3 (Z99.998% Alfa Aesar) with Au/mica substrates was loaded into a three-zone furnace with a 32 mm outer diameter quartz tube. Located upstream from the growth substrates, a second crucible contained B350 mg of sulfur (Z99.99% purity, Sigma-Aldrich). Ultrahigh-purity argon (Ar) was used as the carrier gas. CVD growth was performed at atmospheric pressure. The recipe of MoS 2 growth is as follows: ramp the temperature to 300°C and set for 10 min at 200 s.c.c.m. of Ar flow, ramp to 700°C with 50°C min À 1 rate and set for 10 min at 10 s.c.c.m. of Ar, cool down to 500°C and further open the furnace for rapid cooling with the increasing of Ar flow to 200 s.c.c.m. As a result, monolayers MoS 2 flakes of the order of 1 mm have been grown on Au/mica substrates.
Data availability. The data that support the findings of this study are available from the corresponding author on request.