Exploring the Surface Oxidation and Environmental Instability of 2H‐/1T’‐MoTe2 Using Field Emission‐Based Scanning Probe Lithography

An unconventional approach for the resistless nanopatterning 2H‐ and 1T’‐MoTe2 by means of scanning probe lithography is presented. A Fowler–Nordheim tunneling current of low energetic electrons (E = 30–60 eV) emitted from the tip of an atomic force microscopy (AFM) cantilever is utilized to induce a nanoscale oxidation on a MoTe2 nanosheet surface under ambient conditions. Due to the water solubility of the generated oxide, a direct pattern transfer into the MoTe2 surface can be achieved by a simple immersion of the sample in deionized water. The tip‐grown oxide is characterized using Auger electron and Raman spectroscopy, revealing it consists of amorphous MoO3/MoOx as well as TeO2/TeOx. With the presented technology in combination with subsequent AFM imaging it is possible to demonstrate a strong anisotropic sensitivity of 1T’‐/(Td)‐MoTe2 to aqueous environments. Finally the discussed approach is used to structure a nanoribbon field effect transistor out of a few‐layer 2H‐MoTe2 nanosheet.


Introduction
Molybdenum ditelluride (MoTe 2 ) is a van der Waals layered material appearing in three different crystal phases, the hexagonal 2Hphase (semiconductor), the monoclinic 1T'-phase (semimetal), and the orthorhombic T d -phase (Weyl-semimetal) (Figure 1a). [1][4][5][6] Remarkably, the semiconducting DOI: 10.1002/adma.2023108879][10][11][12][13] Furthermore, it is possible to integrate the 2H-phase directly in plane with the semimetallic 1T'-phase, either already during synthesis or via subsequent laser irradiation, allowing the fabrication of transistor structures with very low Schottkybarrier heights ( B ≈10 meV). [14,15]Next to the practical aspects of the 2H-and 1T'-phase, the T d -phase is a type II Weylsemimetal, a superconductor (T c up to 7.6 K) and has been recently found to exhibit ferroelectric behavior as well. [16,17,18][21][22] In general, MoTe 2 is known to be the material being the most sensitive to phase transitions out of all group VI transition metal dichalgonides (TMDCs). [23]The most common phase transition appears from the thermodynamically more stable 2H-phase toward the 1T'-phase.][26][27][28] For instance, high energy electron beam irradiation generates tellurium vacancies which then can induce enough stress to the crystal lattice to enable a 2H-to 1T'-phase transition. [26,27]ompared to the 2H-to 1T'-transition, the transition from the 1T'-to the T d -phase can be rather easily initiated for bulk MoTe 2 by cooling it down below 250 K.The 1T'-and T d -phase have a similar monolayer crystal structure and the 1T'-phase represents a distorted version of the T d -phase's unit cell as shown in Figure 1a.Thus, by straining the crystal (e.g., through cooling) a transition from 1T'-to T d -MoTe 2 can be achieved.Nevertheless, by protecting the MoTe 2 from oxidation via encapsulation using hexagonal boron nitride, the T d -phase was also observed in thin sheets of MoTe 2 (thickness <12 nm) at room temperature. [29]s demonstrated by the previous examples, MoTe 2 offers great potential for the exploration of fundamental physical effects, but suffers from its rather fragile nature.In order to preserve its initial physical properties, the technological processing should therefore be as gentle and minimal invasive as possible.Nanostructuring of 2D materials like MoTe 2 is typically done using Figure 1.a) Ball-and-stick representations of the different crystal phases of MoTe 2 .Crystal dimensions after. [66,67]b) Sketch of the used setup for probebased patterning.Details about the used system can be found. [65]The water adlayer is actually also covering the MoTe 2 surface, but was omitted here for a better visibility.The displayed equation for the Fowler-Nordheim emission current is taken from. [60]c) AFM image showing the FE-SPL based patterning on a 1T'-MoTe 2 nanoribbon at V Bias = 45 V, line dose ranging from 5 to 20 nC cm −1 (RH = 30%).d) Topographic image of the resulting structures after dissolution of the patterned oxide in DI-water.The natural surface oxidation led to an increased surface roughness as well.e) Crosssectional profiles extracted from (c) and (d).The displayed patterning at 5 nC cm −1 resulted in a removal of a single layer of MoTe 2 (the increased roughness due to surface oxidation after water immersion occasionally lead to depths >0.7 nm).f) Oxide height versus resulting structure depth after dissolution in water for various exposure doses.The data were extracted from tests on bulk (about 100 nm in thickness) 2H-MoTe 2 , but are comparable for few-layer 2H-and 1T'-MoTe 2 for a water immersion time of 30 s.
[35][36][37] Apart from the conventional electron beam lithography-based nanopatterning, atomic force microscopy (AFM) is widely used to determine the actual number of atomic layers present in the investigated 2D nanosheet.Knowing this thickness is crucial, since one of the most distinct features of van der Waals materials is the thickness-dependent bandgap.In case of 2H-MoTe 2 , the bandgap varies from about 0.8 eV (indirect) for bulk material to ≈1.15 eV (direct) for monoatomic and bilayer MoTe 2 . [2,3]herefore, the direct combination of atomic force microscopy and nanolithography appears to be an attractive approach for a controlled nanostructuring of van der Waals-layered materials.0] Nevertheless, scanning probe-based patterning is a sequential and thus inherently slow writing process, which therefore is not able compete with the throughput of an advanced state-of-theart electron beam lithography system.However, certain unique properties of this technology make it an attractive candidate for an application in fundamental material science.In contrast to EBL, no high energy electrons are involved, which potentially alter or damage the investigated material.Furthermore, the AFM imaging allows a pattern alignment with an accuracy well below 100 nm even onto 3D features.43][44][45][46][47][48][49][50][51][52][53][54] Rather than exposing a resist film on top of the 2D material, these works directly structured the nanosheet itself via a tipinduced nanoscale oxidation.This is typically achieved by applying a bias voltage (DC or pulsed) between the nanosheet and the AFM tip while both are in soft contact with each other.In addition, this is carried out under ambient conditions (relative humidity RH≈30-40%), causing a water adlayer to be present on the nanosheets' surface which furthermore also immerses the cantilever tip into a water meniscus. [55,56]Thus, the patterning is achieved through a nanoscale electrochemical oxidation between tip and sample and is therefore typically referred to as local anodic oxidation (LAO).In addition to this, recently also the scanning probe-based patterning of 2H-MoTe 2 was demonstrated using heated AFM tips (tip temperature about 500-900 °C) for a thermomechanical nanostructuring instead of an anodic oxidation. [57,58][61][62] Because of this, the tip is depending on the chosen emission current setpoint a few nanometers to a few tens of nanometers above the sample surface and thus, not immersed into the water adlayer present on the nanosheets' surface.Furthermore, the fixed emission current I (typically 0.5-20 pA) in combination with a defined lateral tip velocity v Tip (typically 0.1-2 μm s −1 ) results in a fixed electron exposure dose, allowing a homogenous and reproducible exposure (e.g., 1 pA at 1 μm s −1 equals a line dose of 10 nC cm −1 ).
The general scheme is illustrated in Figure 1b: a bias voltage between 30 and 60 V is applied at the sample (i.e., the 2D nanosheet) and the cantilever tip is brought in close proximity to it by a proportional-integral (PI) feedback loop controlling the AFM's z-piezo, regulating the tip position on a fixed emission current setpoint.As the specific emission current is according to the Fowler-Nordheim equation dependent on the electric field (see adapted eq. in Figure 1b, taken from [60] ) and therefore the distance between tip and sample due to the fixed V Bias , the maintenance of a fixed current emission therefore also equals maintaining a fixed distance between tip and sample.Thus, writing a pattern across the surface of the nanosheet surface also means in general the tip is tracking the sample's topography.The structuring process is therefore to a certain degree independent of the surface morphology.A detailed overview over the used AFM/SPLsystem can be found. [65]s the patterning is carried out under ambient conditions at a relative humidity (RH) of 30-40%, the strong electric field between tip and sample leads in addition to the Fowler-Nordheim emission current to a dissociation of water molecules, generating H + as well as O 2 − and OH − . [44]Due to the ambient humidity, water molecules are not only present within the air between tip and sample, but at a certain threshold value (RH ≈15-25%) also condensation of water on the sample occurs, resulting in a typically less than 1 nm thin H 2 O-adlayer. [55,56]urthermore, since a positive bias is applied to the 2D nanosheet while the tip is on ground potential, holes are accumulating below the surface of the nanosheet within the area underneath the tip.The general anodic oxidation reaction for group VI TMDCs can be found as: [68] MX 2 + 9H 2 O + 14h where M represents the transition metal and X the chalcogen atom, respectively.Hence, in case of MoTe 2 , the oxidation leads to the generation of MoO 3 as well as TeO 3 2− and furthermore potentially also TeO 4 2− . [68]As can be seen from this equation, the presence of a sufficient concentration of water is crucial for the oxidation reaction, which is used for patterning.At a relative humidity below ≈25%, the tip-based patterning tends to become instable, resulting in inhomogeneous patterned ("dotted") lines or the ab-sence of any oxidized structures at all, whereas with increasing humidity the written oxide features widen.The experimental results presented in this work were therefore acquired at a relative humidity between 30% and 40%, as this allowed a continuously stable patterning while maintaining the smallest possible structure width.
The generated oxide compound is water soluble and thus, a direct pattern transfer of the previously oxidized structure into the MoTe 2 is possible through an immersion in deionized (DI) water (Figure 1d).For this reason, the entire tip-based patterning process has an overall low contamination risk, because no resists or chemicals other than DI water are required.We characterized the oxidation reaction and the subsequent pattern transfer for the 2H-phase in Figure 1f (the reaction for the 1T'-phase is very similar for an immersion time of 30 s).By using exposure doses as low as 5 nC cm −1 (I = 0.5 pA, v Tip = 1 μm s −1 ) it was possible to oxidize only the topmost layer (the oxidation and subsequent removal of a single layer of 1T'-MoTe 2 can be seen in the profile in Figure 1d,e), while on the other hand high line doses of 200 nC cm −1 lead to trenches with a depth of around 6 nm after DI-H 2 O immersion.The resulting structure depth was typically around one third of the previously present oxide height.As illustrated in Figure 1f, the oxide growth and the resulting trenches show a quite linear relationship with respect to the exposure dose in the range below 100 nC cm −1 .At higher doses the oxide growth and thus the resulting trenches after pattern transfer are becoming slightly less proportional to the used electron exposure.We assume this originates from an alteration of the electric field distribution between tip and sample that is caused by the increased oxide growth toward the tip, which affects the electron emission as well as the surface oxidation.Furthermore, with increasing oxide height growth, it is becoming more and more challenging for the PI feedback loop to avoid contact between the cantilever tip and the oxide, due to the rapid topographic changes during oxidation in the vertical z-direction.As discussed later on (Figure 2e), the tip-grown oxide is also nearly an insulator, which furthermore potentially affects the patterning at high emission doses in a negative manner.Therefore, exposure doses any higher than 400 nC cm −1 were not explored to prevent tip crashes.
Next to the selected emission current and the present relative humidity, the patterning is affected by the electric field between the cantilever tip and the sample.A positive bias voltage V Bias is applied to the sample, while the tip is always kept on ground potential through a transimpedance amplifier (TIA) which also converts the emission current into a proportional voltage signal.[62] On this basis, it seems advisable to use rather lower bias voltages to achieve smaller structures.However, a lower bias voltage also means that it is required to bring the tip in closer proximity to the surface to enable the field electron emission.Due to the oxide growth in z-direction toward the tip, we observed that the patterning at low voltages is in our particular setup therefore indirectly restricted to low exposure dosages (<50 nC cm −1 ), since at higher doses it will likely initiate a contact between the tip and the rapidly growing oxide.We therefore remained using bias voltages between 30 and 60 V, although, a reliably stable patterning at V Bias = 25 V at low emission setpoints (<1 pA) was achievable as well.Another important effect which has to be considered, is that the surface adsorbed water used for the oxidation-based patterning is a polar molecule and thus affected by the electric field between tip and sample.In general, the discussed approach operates the cantilever tip on a fixed emission current setpoint above the nanosheets' surface and also above the adsorbed water layer on top of it.However, due to the voltage biasing, water molecules are attracted toward the tip and can even bridge onto the tip once a certain threshold potential is reached. [63] Since in the case of water bridging onto the tip it is immersed into a water meniscus, the tipbased patterning is leaving the field emission-based regime and enters the local anodic oxidation (LAO) regime instead.An example of how this effect influences the patterning will be discussed later in Figure 3.
Summarizing, the parameters governing the dimensions of the tip-induced oxide are the relative humidity RH, the bias voltage V Bias and the electron emission line dose, which in turn is a product of the emission current setpoint I and the lateral tip speed v Tip .Increasing the relative humidity or the line dose results in the generation of more oxide and thus, a structuring reaching deeper into the MoTe 2 after oxide removal.An increased bias voltage leads just to a wider oxide growth, but does not noticeably affect the height of it.
Typical parameters for achieving very small features (sub-20 nm FWHM, only single MoTe 2 layer structured) are a bias voltage of 25-30 V, a relative humidity of 30% and an exposure line dose of around 5 nC cm −1 (e.g., I = 0.5 pA at v Tip = 1 μm s −1 ).At such parameters, parallel structures can have a pitch of down to around 50 nm.If one intends to structure thicker MoTe 2 (up to 10 nm), suitable parameters would be a bias voltage of 50-60 V, a relative humidity of 40% or even higher and an exposure dose of 400 nC cm −1 (e.g., I = 4 pA at v Tip = 0.1 μm s −1 ).
In addition to the parameters dominating the oxidation reaction, the utilized AFM cantilever tip plays an important role for a successful patterning.Throughout this work, we used silicon tips which were able to reliably write dozens of different pat-terns.However, tips can also degrade and become blunt, which can result in an instable patterning and thus also in inhomogeneous oxidized structures (cf. Figure S2, Supporting Information).Moreover, degraded tips can furthermore suffer from multitip emission, where multiple field emission spots are present on the tip and lead to multiple lines being oxidized in parallel onto the nanosheet.A simple but also more elaborate way to minimize such effects is to use more wear-resistent tips for patterning, such as gallium nitride or diamond. [69,70]

Tip-Induced Oxide Characterization on 2H-MoTe 2
While the results presented in Figure 1 were acquired on rather thick nanosheets, we additionally performed similar SPL experiments on few layer 2H-MoTe 2 .Initially, we intended to fabricate monolayer samples, but switched to few-layer nanosheets due to the very poor stability of monolayer MoTe 2 . [21]An example of a dosage test on trilayer 2H-MoTe 2 is shown in Figure 2a-d.The MoTe 2 nanosheet was transferred onto a Si/SiO 2 -substrate which also featured prepatterned 10/50 nm Ti/Pt film as contact electrode for biasing (see also Figure S5a, Supporting Information).The tip-structured region was located upon the 300 nm SiO 2 of the chip in order to simulate the conditions present when modifying a FET channel suspended on a SiO 2 gate dielectric.For this dose test, a constant current setpoint of I = 0.5 pA was used and the different line dosages were achieved by variating the writing speed v Tip from 2 μm s −1 (corresponds to 2.5 nC cm −1 ) down to 0.15 μm s −1 (corresponds to 33.3 nC cm −1 ).After pattern transfer in DI-H 2 O, there were no noticeable changes in the nanosheets' topography for exposures below 5 nC cm −1 detectable (Figure 2b).In the simultaneously recorded phase shift image (Figure 2c), however, a modification of the MoTe 2 is still clearly observable.It is likely that in this case just the outer tellurium atoms of the topmost MoTe 2 layer were oxidized.At doses of 5 nC cm −1 , a tip-induced oxidation and removal of the topmost layer was observed, similar to the patterning of 1T'-MoTe 2 displayed in Figure 1c-e, where the same dose of 5 nC cm −1 also resulted in the removal of the topmost layer.Furthermore, line dosages of 10 nC cm −1 then also resulted in a modification of the topmost two layers of 2H-MoTe 2 .As soon as the used emission dose reaches a level sufficient to oxidize all present MoTe 2 layers, the oxidation reaction saturates along the z-direction and is growing in lateral directions instead.This can especially be seen at the dose of 25 nC cm −1 when compared 33.3 nC cm −1 in Figure 2d.The height and depth values are the same for both, whereas the feature width increased for the 33.3 nC cm −1 exposure.
Conductive AFM measurements using an Asylum Research Cypher AFM revealed that the tip-induced oxide features a very low conductivity, as the measured current was always less than 100 pA, as displayed Figure 2e.The high resistance of the oxide is potentially also contributing to the instabilities we observed during the tip-based patterning at high emission setpoints (I > 20 pA).Furthermore, the sample shown in Figure 2e was prior to the SPL structuring kept under ambient conditions (T = 20 °C, RH = 30-50%) for 20 days to enable a natural surface oxidation of the 2H-MoTe 2 nanosheet.Compared to the tip-grown oxide, the naturally oxidized surface appears to be well conductive.Therefore, it is likely that the two kinds of oxide have very different compositions.During CAFM measurements we used a high contact force of up to 100 nN to assure a proper electrical contact between the cantilever (boron-doped diamond tip, TipsNano DEP01) and the MoTe 2 surface.This force was sufficient to break off pieces of MoTe 2 at the edges of the nanosheet, while the tip-grown oxide remained unaffected.Therefore, it appears to have a quite high mechanical stability.After pattern transfer in DI-H 2 O, there is almost no difference in the conductance between tip-modified and the pristine MoTe 2 observable (Figure 2e, bottom images).This indicates that the probe-induced oxide completely dissolves from the surface.Furthermore, it confirms that the tip-based lithography does not induce any distinguishable changes within the 2H-MoTe 2 , since for instance a phase transition toward the 1T'-phase should display an increased conductivity due to the semimetallic nature of 1T'-MoTe 2 .
As the presented field emission-based tip oxidation is restricted to line patterns which typically have spatial dimensions rarely exceeding 100 nm and a z-height of less than 50 nm, an adequate material analysis beyond scanning probe microscopy remains challenging.To examine the structured surfaces, micro-Raman and Auger electron spectroscopy (AES) were used, as both methods feature a spatial resolution well below one micrometer and a z-resolution of basically between one atomic layer for Raman and down to about 10 nm for the Auger spectroscopy.
Pristine MoTe 2 compared to patterned MoTe 2 featured decreased characteristic Auger peaks of molybdenum and tellurium, while simultaneously the oxygen peak increased (Figure 2a).The Auger peak ratios given in Figure 2a indicate that the probe structured oxide potentially consists of a mix of MoO 3 /MoO x and TeO x , which agrees well with the previously mentioned anodic oxidation equation.A study using X-ray photoelectron spectroscopy (XPS) by Jaegermann et al. on in air and photoelectrochemically oxidized MoTe 2 featured binding energies of MoO 3 /MoO 4 2− and TeO 3 /TeO 4 2− . [22]More recent XPSbased studies on naturally and in O 2 -atmosphere oxidized 1T'-MoTe 2 revealed a similar composition, showing that oxidized MoTe 2 surfaces featured MoO 3 /MoO x as well as TeO 2 /TeO x . [21,71]hus, based on our results and others from literature, we conclude that the probe grown oxide species mainly consists of amorphous MoO 3 and TeO 2 , while other oxidation states are likely present as well.
A tip-induced growth of MoO 3 /MoO x combined with a subsequent dissolution in DI-water has previously been reported for MoS 2 and MoSe 2 . [52]The presence of TeO 2 /TeO x makes the pattern transfer slightly more challenging, due to the considerably lower water solubility of the tellurium oxide (TeO 2 : 0.0025 g L −1 ) compared to MoO 3 (1 g L −1 ). [72,73]The dissolution of the oxidized patterns in DI-water can in principle be improved by using gentle ultrasonication or using heated water (e.g., T≈70 °C).In case of the presence of TeO 3 for instance, the heating eventually enables the dissolution in water due to the formation of the telluric acid H 6 TeO 6 . [74]owever, an immersion in DI water for <5 min combined with gentle manual agitation of the immersed substrate was usually sufficient in our experiments to completely dissolve the oxide with respect to the 2H-phase.For the 1T'-phase such treatment was already too harsh, as will be discussed.
Raman measurements were performed on a SPL-oxidized ≈16 nm thick 2H-MoTe 2 nanosheet (Figure 2g).As shown in Figure 2g, Raman peaks are increased for the in-plane E 1 G (115 cm −1 ) and E 1 2G (230 cm −1 ) modes, as well as for the out-ofplane A 1G (170 cm −1 ) and B 1 2G (285 cm −1 ) modes in tip-oxidized regions. [75]No modes of MoO 3 or TeO 2 were present in the recorded spectra (Figure S2 for wideband Raman spectra, Supporting Information). [76,77]Therefore, we assume that in this case the probe-structured oxide is of amorphous nature and the measured Raman signal originates from the underlaying buried pristine MoTe 2 , which was not exposed to the oxidizing environmental atmosphere like the referenced top-most nonpatterned MoTe 2 .Moreover, the increased Raman response can also be assigned to the effectively reduced thickness of the MoTe 2 buried underneath the tip-induced oxide, which provides for the present thickness of about 10-12 nm in general a stronger Raman response (Figure S3, Supporting Information).
After immersion in DI-H 2 O (Figure 2h) and thus dissolving and removing the SPL-grown oxide, the Raman response was very similar, showing a stronger response of all modes for the regions which were previously covered by the nanopatterned oxide.However, compared to the spectra in Figure 2g, a peakshift of +0.7 cm −1 of the E 1 2g mode within the patterned regions is visible.We suppose this originates from the thinning of the material (from ≈16 nm in Figure 2b to ≈10 nm in Figure 2h), which leads to a blueshift of the E 1 2g mode for such thicknesses as reported elsewhere. [75,78,79]However, the simultaneously reported redshift of the A 1g mode was not observed by us.Besides the absence of any oxide peaks, the Raman spectra recorded on patterned 2H-MoTe 2 also do not feature any signature of modes typical for 1T'-MoTe 2 (for reference see Figure 4d,f).Thus, we conclude that the SPL-based patterning induces no measurable phase transitions to the 2H-MoTe 2 crystal.

Nanoscale Oxidation and Structural Sensitivity of 1T'-MoTe 2
Following the 2H-phase, we investigated the 1T'-phase in a similar manner.During dosage tests on 1T'-MoTe 2 , the tip-induced oxide growth was comparable to the 2H-phase, however, we noticed an anisotropy of the remaining grooves after pattern transfer in DI-water, as well as a generally increased sensitivity of the material when being exposed to water.For short water immersions of 30 s the resulting structure depths are comparable to 2H-MoTe 2 , but with increasing time of water exposure the structure depth and width increases.
A first example of this is given in Figure 3.The dosage test (V Bias = 50 V, v Tip = 1 μm s −1 , I = 0.5-20 pA) in Figure 3a was taken on an ≈30 nm thick, mechanically exfoliated nanosheet of 1T'-MoTe 2 , which was transferred onto a Si/SiO 2 -substrate featuring a 50 nm thick palladium film used as electrical contact.Furthermore, the sample was not thermally annealed nor wet chemically cleaned prior to SPL to avoid any additional degradation of the material.During the patterning of the exposure test, we noticed on certain occasions the previously mentioned field induced water bridging effect, which can be seen more detailed in the pattern written at 70 nC cm −1 (I = 7 pA at v Tip = 1 μm s −1 ) in Figure 3b,c.As the cantilever is approaching the surface toward the emission setpoint of 7 pA, the starting point of the pattern is slightly overexposed, since the field emission gradually starts from 0 pA and has first to reach near the desired emission current setpoint (around t = 1 s in Figure 3c).In our particular setup, a threshold value is used to determine when the xy-movement for the actual pattern writing is enabled, which was typically as soon as 75% of the setpoint was reached.As soon as this condition is met, the cantilever is moved in xy-directions over the surface to write the pattern.In Figure 3c at around t = 1.5 s, however, the emission current suddenly changes to a much higher value, which is then compensated again by the feedback loop which retracts the cantilever by around 100 nm from the surface (the actual values for the current spike and subsequent z-retraction are in reality higher than in the graph, because the displayed values are only sampled with 20 data points per second in the software user interface).In this very moment, the water attracted by the electric field toward cantilever eventually snapped around the tip and immersed it in a water meniscus, which then caused the rapidly increased current level and consequently also a much stronger oxidation reaction.Hence, in this moment the cantilever accidently leaves the field emission-based SPL regime and enters the local anodic oxidation (LAO) SPL regime.The retraction of the cantilever by 100-200 nm then again pulls the tip out of the water meniscus and allows it to return to the desired field emission current setpoint.For the pattern written at 70 nC cm −1 , the unwanted water meniscus bridging happened in total seven times, where each oxide dot in the topography (Figure 3b) can directly be correlated to spikes in the emission current recorded over time (Figure 3c).Additional curves of the emission current and z-position over time can be found in Figure S6 (Supporting Information).A direct visualization of the water meniscus on an (unbiased) AFM tip can furthermore be found in works by Weeks et al. and Schenk et al. [80,81] After the pattern transfer through immersion in DI-water, the probe-fabricated structures appeared homogenous, however, for those written at lower line dosages (<50 nC cm −1 ) the oxide did not dissolve properly (Figure 3d).As discussed previously, we suppose this reduced solubility originates from the incorporated TeO 2 /TeO x , which in general has a significantly reduced water solubility compared to MoO 3 /MoO x .Moreover, since no chemical cleaning was carried out prior to the patterning test, it is likely that some residuals from the exfoliation and nanosheet transfer decreased the oxide solubility.In addition, also airborne hydrocarbon contaminations could be adsorbed at the 2D surface, affecting the oxidation reaction as well as the dissolution of the oxide. [82]Because of the remaining oxide, we immersed the sample a second time in DI-water, which then revealed an anisotropy of the patterned grooves, depending on the writing direction of the cantilever (Figure 3e).
Along both axes the patterned grooves had grown in width during the second DI-water immersion.However, the growth approximately along the [010] crystal axes of the 1T'-MoTe 2 was pronounced.In Figure 3f, the cross-section profiles, extracted from the same positions in Figure 3a,d,e are displayed and provide an overview of the dimensional changes.While for the patterning roughly following the [100] axis of the MoTe 2 , the trenches had the same width as the oxide previously, whereas along the [010] axis the trenches even exceeded the width of the initial oxide.Noteworthy, independent from the scanning probe-induced oxidation, a rupture in the nanosheet showed the same behavior, being visible especially after the second immersion and showing a significant growth in width along the [010] direction.This indicates that the pattern growth after the second immersion does not simply originate from remaining tip-induced oxide being dis-solved, rather it seems to be due to an oxidative dissolution during the water immersion.Furthermore, the oxide thickness versus the resulting structure depth after oxide dissolution was very similar to the trend given for 2H-MoTe 2 in Figure 1f for the first water immersion of about 30 s.
The 1T'-phase is well-known for a preferential breakage during exfoliation along the Mo-Mo zigzag chain parallel to the [010] axis.[85][86] To further investigate the anisotropy, we patterned such an exfoliated 1T'-MoTe 2 nanoribbon exactly along the [100] and [010] axis and subsequently removed the oxide in DIwater (Figure 4a-c; and Figure S4, Supporting Information).Additionally, the crystal orientation was also verified afterward using angle resolved Raman measurements (see Figures S8 and S10 in the Supporting Information). [84,85]In Figure 4b, the anisotropy is clearly visible on lines oxidized along the [010] direction.Those oxidized patterns, which mainly consisted of parallel lines along the [100] axis, even resulted in a complete areal removal of MoTe 2 .Within the latter regions, the surface appears to be very homogenous and shows the same depth for each individual pattern of the same dose (Figure 4b).Therefore, the anisotropic behavior in combination with the tip-induced oxidation could potentially be employed for a 3D structuring and thinning of 1T'-MoTe 2 .
Cross-sectional profiles of the trenches shown in Figure 4b are given in Figure 4c along with illustrations of the crystal orientation and the anisotropy direction.It can be seen that the material removal during water immersion occurs preferentially along the crystal direction perpendicular to the Mo-Mo zigzag edge.[89] For instance, a publication by Naylor et al. using density functional theory showed that oxygen first dissociates and then adsorbs at tungsten atoms in 1T'-WTe 2 , while simultaneously displacing the adjacent tellurium atom (the preferred O 2 adsorption site according to this is highlighted in Figure 4f). [88]Moreover, the presence of water even promotes this reaction, as it lowers the energetic barrier required for the dissociation of oxygen. [89]According to this, the surface oxidation of 1T'-WTe 2 is in water even stronger than within a pure oxygen atmosphere.Overall, this provides a reasonable explanation which is projectable on MoTe 2 for the observed anisotropic sensitivity to DI water.
To evaluate the 1T'-phase in the same fashion like the 2H-phase in Figure 2b,c, Raman measurements as shown in Figure 4d were recorded on a 1T'-MoTe 2 nanosheet (thickness 55 nm) featuring a 3 × 3 array of 4 μm 2 patterns written at a line dose of 200 nC cm −1 (V Bias = 35 V).Even though the exposure dose was the same as used previously for the comparable experiments on 2H-MoTe 2 (Figure 2b), the Raman response in the patterned regions was less distinct compared to pristine 1T'-MoTe 2 , being only slightly higher and showing no shifts of Raman-modes (Figure 4d).
An example of the strong structural sensitivity of 1T'-MoTe 2 to aqueous environments is given in Figure 4e,f.Taking the anisotropic behavior into account, we oxidized several patterns using the same parameters (V Bias = 40 V, dose = 50 nC cm −1 ) on an ≈50 nm thick 1T'-MoTe 2 nanosheet.Two patterns were aligned in parallel to the orientation of the crystal and the remaining patterns were shifted by 15°with respect to each other.Due to the rectangular structures of the written patterns, any angle between 0°and 180°was covered with a step size of 15°.After immersion in DI-water for about 5 min, we observed that independent from our patterning, multiple layers of MoTe 2 were removed on the nanosheet's entire surface (Figure 4e,f).Initially, after oxidation the written structures were about 6 nm high and thus, following the relationship given in Figure 1f, the depth after immersion should have been around 2-3 nm.However, the delamination of MoTe 2 was so intense that no traces of the patterns remained.Even though the same dose was used for the exposure of all patterns, in some regions pits with a depth of up to about 12 nm were formed during the DI-water immersion.These pits had no clear relationship to the positions of the originally intended patterns, except for the fact that they occurred mostly near their original positions.Instead, they had an orientation parallel to the [010] crystal axis.The same behavior was also observed for other samples being immersed in DI-water for about 2-10 min.The material removal appeared to be proportional to the immersion time (see the Supporting Information for additional examples).We therefore conclude that this effect originates from an oxidative dissolution of 1T'-MoTe 2 during the immersion in DI water. [88,89]For a further analysis, we also carried out Raman intensity mapping on this sample (Figure 4e,f).This measurement revealed a distinct increase of all modes within the pits featuring a significant MoTe 2 removal and particularly the A g modes at 77 and 161 cm −1 were the highest.The Raman spectra representing the response inside a pit (red) and about 2 μm next to it (black) are displayed in Figure 4f as well as a close-up AFM image of a pit showing the degraded MoTe 2 surface.The distinct changes of especially the modes around 110 and 255 cm −1 in Figure 4f compared to Figure 4d originate mainly from the different crystal orientations of the nanosheets (Figure 4e,f has a rotation of around +50°compared to Figure 4d). [85]he crystal structures of the 1T'-and T d -phase are very similar and furthermore do not feature the symmetry of the 2H-MoTe 2 crystal, as was already illustrated in Figure 1a.Consequently, for both phases comparable anisotropies have been reported, for instance of the mechanical properties as well as for the interlayer coupling strengths. [83,90]Because of this, we expect a very similar behavior of the T d -phase when being exposed to water.Furthermore, the lower interlayer coupling strengths of 1T'-and T d -MoTe 2 in z-direction, being only about half of that of the 2Hphase, potentially also promotes the observed large area delamination during water immersion. [90]ummarizing our experiments on the 1T'-phase, we observed three different reactions after oxide removal in DI-water, which directly relate to the immersion time as well as the time passed between patterning and immersion (the latter being mostly important for case (I)).(I) a (nearly) 1:1 pattern transfer such as shown in Figures 1c,d,e and 3d for an immersion of <30 s done within a few hours after tip-induced oxidation.(II) a clearly visible anisotropy for samples immersed for >30 s (Figure 4a,b) which then transitions into (III), a removal of multiple layers everywhere on the nanosheet for immersions over about 2 min and beyond (Figure 4e,f, additional examples for each stage: Supporting Information).Based on these results, the exposure of 1T'-MoTe 2 to an aqueous environment should be kept as short as possible to avoid the reported water-induced surface degradation.f) The nanoribbon FET channel after tip lithography.g) The final nanoribbon FET channel after pattern transfer in deionised water.h) Cross-sectional profiles of the FET channel from (f) and (g).

Fabrication of a Few Layer 2H-MoTe 2 Nanoribbon FET
After the in-depth characterizion of the tip-induced patterning for the 2H-and 1T'-phase, we finally employed it to structure a nanoribbon FET out of a three atomic layers thick 2H-MoTe 2 nanosheet (Figures 5-7).First, 2H-MoTe 2 was mechanically exfoliated, and then a nanosheet of interest was transferred onto a Si/SiO 2 substrate featuring prepatterned Ti/Pt (10/50 nm) electrodes.This prestructured electrode is used to provide the required bias voltage for the tip-based nanopatterning to the MoTe 2 .The very first test patterns were oxidized onto the MoTe 2 quite close to the electrode to verify the presence of a proper electrical contact between the bias electrode and the nanosheet (Figure 5a).Furthermore, since the nanosheet is quite large (> 40 μm), very small fractures could potentially be present in the MoTe 2 , which could in turn cause that certain parts of it are electrically insulated and thus could not be tip-patterned.A direct inspection of this, as well as a verification that tip-structured regions are fully insulated, could potentially be done in situ using electrostatic force microscopy (EFM), however, at our particular SPL setup this mode was not available.Throughout the test patterns written over almost the entire nanosheet surface (Figure S5, Supporting Information), the field emission current from the AFM tip was always absolutely stable, therefore a proper electrical contact was confirmed.Additionally, a dosage test was performed (see Figure 2a-d) to once again determine a line dose which is sufficient to oxidize through all of the three layers of MoTe 2 present.After this, we finally patterned the FET structure indicated by the white dashed lines in Figure 5a at an emission dose of 33.3 nC cm −1 (I = 0.5 pA at v Tip = 150 nm s −1 ) (Figure 5b).A detailed discussion of the patterning will be discussed later in Figure 6.Once the tip-based structuring was finished, the FET channel was encapsulated with hexagonal boron nitride (hBN) to suppress any further degradation of the MoTe 2 and to provide also a mechanical stabilization for the fragile nanoribbons.Prior to the transfer, we cleaned the nanosheet in aceton/IPA/DI-H 2 O and heated the chip for 10 min to 120 °C to reduce the risk of residuals or water being trapped between the MoTe 2 and the hBN.After transfer, the chip was additionally annealed at 180 °C in vacuum for 90 min under Ar/N 2 gas flow (250 sccm each).For a proper placement of the hBN, the larger oblong patterns oxidized into the MoTe 2 prior to the actual FET were used.They also allowed an easier layout alignment for the structuring of the source/drain electrodes, which were eventually patterned using maskless photolithography and subsequent electron beam evaporation of 10/50 nm chromium/palladium for low Schottky barrier contacts. [91,92]During the lift-off of the con-tact electrodes, a large fraction of the MoTe 2 nanosheet delaminated from the Si/SiO 2 substrate and thus insulated the FET structure from the already existing electrode used for the voltage biasing during scanning probe patterning.This could otherwise have been achieved by an additional poststructuring of the MoTe 2 using plasma etching.Furthermore, the biasing electrode could have also been used as for instance a side gate electrode to the unconnected excess MoTe 2 surrounding the FET channel.
The actual structured FET channel, consisting of six 2 μm long nanoribbons with a width of 130 nm, separated by gaps of 35 nm can be found detailed in Figure 6.Since the tip-induced oxide appears to be nearly insulating (see Figure 2e), we first structured the FET-channel and regions between the source and drain electrode (marked as #1 in Figure 6a), before finally structuring lines separating the intended FET structure from the rest of the MoTe 2 nanosheet (#2 in Figure 6a).Furthermore, we oxidized multiple parallel lines and additionally subdivided larger to be insulated MoTe 2 areas (such as the region between source and drain in Figure 6d,e), to fully guarantee an insulation of drain and source electrodes in case at certain points the MoTe 2 was not completely oxidized through.During patterning, we typically performed some AFM imaging in-between, to verify the success of the oxidation and proper alignment to the nanosheet outlines (Figure 6d).
An AFM image of the final FET structure is shown in Figure 7a.Additional fractures in the MoTe 2 , likely caused by strain induced during electrode fabrication, are clearly visible.Also, several smaller segments of the tip-structured MoTe 2 were displaced, demonstrating that an encapsulation of the nanoribbon structure is vital.][95] Notably, the off-state current of the FET reached quite decently low values, being in the range of 100 fA up to 1 pA, which is also about the lowest current level the utilized sourcemeter unit (Keithley 2450 SMU) can measure.
In order to verify that the tip-based structuring of the FET was successful and the measured behavior can be solely addressed to the nanoribbon channel, we performed electrostatic force microscopy (EFM) on the device (Figure 7d-f) using another AFM system (Asylum Research Cypher).A bias voltage of +3 V was applied to the source electrode and the cantilever was scanned in dual pass EFM mode with a z-shift of +50 nm over the FET channel.In the EFM phase shift image (Figure 7f) the tip-written FET structure buried under the hBN can clearly be resolved.The excess MoTe 2 islands surrounding the nanoribbon channel have no electronic connection to any bias potential anymore.Thus, due to the absence of the electrostatic force in these regions, different forces are acting on the cantilever, leading to different phase shifts which eventually prove proper operation of the device.

Conclusion
We discussed in depth an unconventional and resist-free method to nanopattern the semiconducting (2H-) as well as the semimetallic (1T'-) phase of MoTe 2 .It was shown that the presented technological approach does not induce any measurable phase transitions and is therefore highly suitable for exploring the fundamental physical properties of MoTe 2 at the nanometer scale.
The tip-induced patterning combined with AFM imaging revealed an anisotropic instability of the 1T'-phase as well as a general structural sensitivity to aqueous environments.Even though we did not investigate the T d -phase within this work, we presume a behavior comparable to the 1T'-phase, since it features a very similar crystal structure (see Figure 1a for reference).
Moreover, not only has tungsten ditelluride (WTe 2 ) also similar crystal phases like MoTe 2 , it furthermore has also been demonstrated that tip-grown WoO 3 on WSe 2 can be dissolved in DIwater. [51,87]Thus, based on the results shown here on MoTe 2 , a tip-based resistless structuring of WTe 2 using DI-H 2 O for pattern transfer should potentially be possible as well.In addition, due to the similar crystal structure and the theoretically predicted susceptibility to a water promoted oxidation, [88,89] WTe 2 potentially shows the same behavior as presented for 1T'-MoTe 2 when being exposed to aqueous environments.To our best knowledge, the work presented here was furthermore the first tip-based oxide patterning of a TMDC featuring tellurium as the chalcogen atom.In combination with other previous works on MoS 2 , MoSe 2 , and WSe 2 , it becomes therefore evident that all group VI TMDCs can be resistless structured using a tip-based oxidation and a subsequent pattern transfer in DI-water. [46,51,52]inally, we utilized the minimal-invasive character of the presented patterning approach to structure a nanoribbon FET out of a large few-layer 2H-MoTe 2 nanosheet.The patterned nanoribbons had a width of 130 nm, were separated by gaps of 35 nm and later contacted using maskless photolithography.Thus, we successfully demonstrated a way of fabricating a nanoscale MoTe 2 device without the need for an electron beam lithography system.
Contact electrodes were structured using maskless UV photolithography using a direct laserwriter (LW405d, MICROTECH srl, Italy and MLA150, Heidelberg Instruments Mikrotechnik GmbH, Germany) and subsequent metal deposition by electron beam evaporation (CS400ES, Von Ardenne GmbH, Germany) of chromium (10 nm) and palladium (50 nm).After metal deposition, the contacts were thermally annealed in vacuum for 1 h under a forming gas flow (250 sccm Ar + 250 sccm N 2 ) at a temperature of 180 °C.
Sample Characterization: Atomic force microscopy as well as scanning probe lithography was done using the referenced in-house built AFM/SPLsystem using self-actuated piezoresistive cantilevers (f res ≈ 90-130 kHz, Nano Analytik GmbH, Germany). [65]The cantilevers were additionally sharpened by Ga + ion milling in a focused ion beam system (FEI Helios Nanolab 600) (Figure S1, Supporting Information).Also, always the same cantilever has been used for patterning and AFM imaging.The humidity and temperature within the AFM chamber was constantly monitored using a USB hygrometer (Hygrosens Instruments GmbH, one sample per second, accuracy ±0.1% and ±0.1 °C) and kept on a constant value by blowing compressed air through a DI-water reservoir.Conductive AFM (CAFM) and electrostatic force microscopy (EFM) were performed using an Asylum Research Cypher AFM with soft (1-3 N m −1 ) and highly conductive diamond cantilevers (TipsNano DEP01).The AFM image data were processed using the Gwyddion software environment (www.gwyddion.net).
Raman measurements were performed using an alpha300 apyron system (WITec GmbH, Germany) using a 532 nm excitation (P Laser = 1-5 mW) using an x100/0.9NA objective in combination with an optical grating featuring 1800 grooves per millimeter.The integration time for Raman mapping was 0.25 and 10 s for single spectra, respectively.
Auger spectroscopy was done using an AES Microlab 350 from Thermo Fisher Scientific, Inc. (USA).
Electrical measurements were made using a Keithley Instruments 2450 SMU (Tektronix, USA) under ambient conditions with the samples being shielded from light.Between measurements and nanopatterning, samples were kept in a dry environment (RH 5-10%) to reduce exposure to oxygen and water.
The 3D atomic models included in several figures were originally downloaded from the Materials Project (www.materialsproject.org) and further edited using the VESTA software environment (www.jp-minerals.org/vesta/en/). [97,98]

Figure 2 .
Figure 2. a) Dosage test ranging from 2.5 to 33.3 nC cm −1 (V Bias = 50 V, RH = 35%, I = 0.5 pA, v Tip = 0.15-2 μm s −1 ) taken on three atomic layers thick 2H-MoTe 2 .b) The same structure after immersion in DI-water for 1 min.c) AFM Phase shift image recorded in parallel to (b), showing that down to an exposure of 2.5 nC cm −1 the material has been altered.d) Topographic profiles of (a) and (b) for 5, 10, 25, and 33.3 nC cm −1 .Similar to the 1T'-MoTe 2 example from Figure 1c-e, a dose of around 5 nC cm −1 resulted in a removal of the topmost layer and 10 nC cm −1 to the removal of the topmost two layers, respectively.e) Conductive AFM (CAFM) measurements on tip-induced oxide as well as on patterns transferred into the 2H-MoTe 2 surface.Left-hand side are topographic images and on the right-hand side the simultaneously recorded current flow.f) Auger spectra taken on pristine and SPL-oxidized 2H-MoTe 2 (5×5 patterns, 900×900 nm 2 -similar to Figure 2e, V Bias = 35 V, dose = 200 nC cm −1 ).g) Raman measurements on SPL oxidized 2H-MoTe 2 featuring 16 square shaped patterns (each 900×900 nm 2 , line dose of 200 nC cm −1 , V Bias = 50 V).h) The same pattern measured after immersion in DI-H 2 O which removed the oxide and therefore uncovered the MoTe 2 underneath.In the optical image the underlying palladium electrode shines through the thinned down regions.

Figure 3 .
Figure 3. a) Line dose test taken on an exfoliated 1T'-MoTe 2 nanosheet.The number next to each pattern represents the used electron exposure dose in nC cm −1 (V Bias = 50 V, RH = 40%).b) Close-up AFM image of the pattern written at 70 nC cm −1 featuring several large oxide dots.c) The emission current and relative cantilever z-position over time recorded during the patterning of (b).Each oxide dot shown in (b) can clearly be addressed to an unwanted water film bridging which immerses the tip into a water meniscus and thus accidentally enters the LAO SPL regime.d) The sample after immersion in DI-water.e) After a second immersion, ≈5 h after d), both immersions were for about 30 s. f) Topographic profiles extracted along the colored lines in (a,d,e).The upper graph approximately follows the [100] 1T'-MoTe 2 crystal axes, the lower the [010] axes shown in (a,d,e).The lateral differences of the structure positions originate from drift during AFM imaging.

Figure 4 .
Figure 4. a) Tip-induced oxidation parallel and perpendicular to the crystal axes of a 1T'-MoTe 2 nanosheet.The oxide written 400 nC cm −1 was about 25-30 nm high.Image distortions along the x-axis originate from the used line-wise image correction algorithm required for AFM data reconstruction.b) The same sample after DI-water immersion, clearly showing an anisotropic pattern transfer.An additional ball-and-stick model illustrates the crystal orientation of the nanosheet.c) Cross-sectional profiles extracted from (b) and the present crystal orientation.The trench patterned along the [100] direction had a width about twice of the lines along the [010] direction (the lower graph shows the profile of two parallel lines).The insets illustrate the present crystal orientation.d) Tip-oxidized test patterns (V Bias = 40 V, dose = 50 nC cm −1 ) on an about 55 nm 1T'-MoTe 2 nanosheet and corresponding Raman mapping.The Raman response seems to be negligibly stronger in the structured area.e) Test patterns with 15°of rotation to each other oxidized (50 nC cm −1 , oxide height ≈ 6 nm) on an ≈45 nm nanosheet and the resulting topography after DI-water immersion (previous pattern position indicated by dashed black lines).Multiple layers delaminated over the entire nanosheet surface, deep pits were created at some locations of the patterns with an anisotropy along the [010] axis.f) Zoomed-in AFM image of a pit from (e), also showing the damaged MoTe 2 surface.The Raman spectra in (f) are showing the signal inside a pit and next to it on the unpatterned but damaged surface (corresponding Raman intensity map in (e)).

Figure 5 .
Figure 5. Optical images of a field effect device structured out of trilayer 2H-MoTe 2 using the discussed probe lithography.a) At the beginning of the fabrication, showing first test structures as well as larger oblong patterns for optical alignment.b) After the actual transistor structure was patterned into the MoTe 2 .The dosage test discussed in Figure 2a-d can be seen as well.c) The final device after encapsulation in hexagonal boron nitride and electrode deposition.A large part of the MoTe 2 delaminated during electrode lift-off, thus the MoTe 2 was disconnected from the initially used bias electrode.

Figure 6 .
Figure 6.a) AFM topography image of a three atomic layer thick 2H-MoTe 2 nanosheet in which a nanoribbon FET structure was oxidized (33.3 nC cm −1 , V Bias = 50 V, RH = 35%).b) The sample after oxide dissolution in DI-H 2 O.The white dashed line marks the outlines of the FET structure.c) Optical image of a), previous tip-written test structures and alignment markers can be seen well where the oxide had already been removed.d) AFM image recorded during patterning showing the MoTe 2 region between source and drain which is supposed to be isolated from both.e) The same position after the patterning was finished and the oxide removed in DI-water.Multiple cuts were patterned in order to assure an insulation of the drain and source region.f) The nanoribbon FET channel after tip lithography.g) The final nanoribbon FET channel after pattern transfer in deionised water.h) Cross-sectional profiles of the FET channel from (f) and (g).

Figure 7 .
Figure 7. a) AFM image of the final MoTe 2 nanoribbon field effect device.Additional fractures in the MoTe 2 nanosheet can be seen.b,c) Electrical measurements showing a purely p-type behavior of the structured FET, likely caused by the surface oxidation of the nanosheet.The inset shows an optical image of the device.d) AFM topography image of the hBN-encapsulated FET channel.The white dashed line indicates the different electrode geometries patterned into the MoTe 2 nanosheet.e) AFM Phase shift image of the same position.f) Electrostatic force microscopy phase shift image recorded simultaneously to (d) and (e) with a bias voltage of +3 V applied to the FET source electrode.The MoTe 2 FET channel underneath the hBN can clearly be distinguished.