Defect engineering of single- and few-layer MoS2 by swift heavy ion irradiation

We have investigated the possibility to use swift heavy ion irradiation for nano-structuring supported and freestanding ultra-thin MoS2 samples. Our comprehensive study of the ion-induced morphological changes in various MoS2 samples shows that depending on the irradiation parameters a multitude of extended defects can be fabricated. These range from chains of nano-hillocks in bulk-like MoS2, and foldings in single and bilayer MoS2, to unique nano-incisions in supported and freestanding single layers of MoS2. Our data reveals that the primary mechanism responsible for the incisions in the ultrathin supported samples is the indirect heating by the SiO2 substrate. We thus conclude that an energy of less than 2 keV per nm track length is sufficient to fabricate nano-incisions in MoS2 which is compatible with the use of the smallest accelerators.


Introduction
Two-dimensional(2D) materials are at the center of many novel technologies, including but not limited to energy conversion, opto-and flexible electronics and biomedical applications [1,2]. The most part of these concepts relies heavily on the possibility to engineer the 2D material in one way or the other. In particular, for potential breakthrough applications such as water desalination and DNA sequencing with the help of synthetic nano-pores [3,4], or manipulating catalytic [5], optical [6] and thermoelectric [7] properties based on specific sample nanostructures, defect engineering on the nanoscale is the key to success. Therefore several experimental and theoretical studies have looked into the effect of defects on the optical and electronic properties of transition metal dichalcogenides (TMDCs), seen as promising building blocks for future applications. While point-like defects can be introduced into TMDCs rather easily by ion [8] and electron bombardment [9], or plasma treatment [10], the artificial introduction of extended defects is not so easy. One of the most powerful tools for defect engineering of surfaces and bulk materials at the atomic scale is the irradiation with swift heavy ions (SHI) [11]. SHI are projectiles with a typical energy of around 1 MeV per atomic mass unit. Their interaction with matter is governed by electronic excitation rather than nuclear collisions (for a review see e.g. [12]). As a consequence, and in stark contrast to keV ions, a SHI may modify the material inside a very narrow (a few nanometres) and extended volume (up to several microns) centered around its trajectory before it is finally stopped. Therefore, using SHI irradiation under grazing incidence, extremely elongated modifications, so-called surface tracks, may be created at the surface of a given material. These may manifest themselves e.g.in the form of chains of nanosized hillocks in dielectrics such as SrTiO 3 [13], or as narrow, one-atom deep trenches as in SiC [14], or as foldings in graphene and MoS 2 , respectively [15,16]. Despite the recent successful applications of SHI irradiation in the exciting and rapidly growing field of nanostructured 2D materials, a systematic understanding of the relevant processes is still lacking. This deficiency hampers quantitative predictions as well as a systematic and efficient exploitation of SHI irradiation for nanostructuring 2D materials by a wider community.
In this work we present our efforts to overcome this problem. We provide evidence that grazing incidence SHI irradiation is a reliable, easy-to-use, and precise tool to fabricate various unique nanostructures in 2D materials. As a model system for our comprehensive study we used MoS 2 , arguably at present the most important representative of the class of TMDCs. We demonstrate that our approach offers a versatile tool-box to efficiently fabricate nanostructures with a great degree of control by showing that the number, length, and orientation of each of these nanostructures with respect to the MoS 2 lattice may be chosen at will. In the following, we will first describe in detail the conditions for fabrication of the various nanostructures. In the second part of the paper, we focus on elucidating the physical mechanism underlying the fabrication of nano-incisions in ultra-thin MoS 2 . This allows us to determine the minimum energy per path length required for an incision to occur which is an important factor with respect to the general feasibility of the technique.

Results
We start by presenting the basic types of modification which can be fabricated in ultra-thin MoS 2 sheets supported on SiO 2 by SHI irradiation under grazing incidence. Both of these modifications, i.e. nanoincisions and foldings, are shown in figure 1.
Here, two mechanically exfoliated samples have been irradiated using 91 MeV Xe ions under a grazing incidence angle of Q =  1 with respect to the surface. The corresponding energy deposited per track length into the target material is usually given in terms of the electronic stopping force , which amounts to 19 keV nm -1 in this case [17]. This energy deposition results either in the formation of incisions (which we will discuss in more detail further below), with an average length of 500 nm as shown in figure 1(b) or in the formation of foldings, figure 1(c) which appear to be similar to those found in graphene, see figure 1(e). However, in contrast to graphene, where foldings are the rule for grazing incidence irradiations [18], for foldings to be fabricated in ultra-thin MoS 2 , very specific conditions have to be met. Foldings do only occur if the azimuthal angle j between the SHI trajectory and a low-indexed direction of a MoS 2 flake is quite an exact multiple of 30°. Two examples of this phenomenon are shown in figures 2(a) and (b) in which j is 90°±3°. If the azimuthal angle j is just slightly off, as in figure 2(c) with62°±3°, the resulting modifications are incisions with two notable exceptions: a small folding which might have been caused by a local irregularity, e.g. a contamination, and one large folding who's origin is explained further below. Note, that there are two uncertainties in this analysis, which result in the relatively large error bar. First, as j is determined from AFM topography images, small drift effects due to the piezoelectric positioning system are always present and are difficult to avoid completely. Second, the edges of exfoliated MoS 2 flakes are not always straight, but can be heavily bent reducing the number of samples available for this analysis.
This unexpected dependence on the azimuthal angle is further corroborated by data obtained from bilayer MoS 2 as shown in figure 3. In basically every irradiated bilayer MoS 2 we studied, incisions were fabricated (see figure 3(a)). Only in a rare case, shown in figure 3(b), where the SHI ion hit the bilayer flake with an exact angle of j=60°, the formation of foldings did occur. The foldings are rather small, and exhibit a symmetry showing multiples of 30°with respect to the low-indexed MoS 2 edge. The latter finding being similar to what has been found in graphene for incidence angles Θ larger than 1° [18]. Even in case of exact alignment of the azimuthal angle j the folding efficiency is clearly less than one: in the inset of figure 3(b) one can see an example, where an SHI impact occurred, that did not result in a folding, instead it appears as if the folding process has stopped halfway through. A lower folding efficiency has been observed for bilayer graphene as well. In the case of graphene, this can be attributed to the fact that bilayer graphene is more resistant to ion irradiation [19] and without vacancy defects usually no foldings occur in the material due to its superior mechanical strength which overcomes the upwards push from the substrate material below. However, in bilayer MoS 2 , SHI do create incisions (see figure 3(a)) which should still facilitate the formation of foldings.
From the data presented above we deduce that the orientation of the SHI trajectory with respect to the MoS 2 crystallite is crucial when either foldings or incisions are to be created, and that foldings are basically limited to single layers. These findings are in striking contrast to graphene, where foldings can be fabricated even in trilayers (although with a strongly reduced efficiency) and in practically any direction with respect to the graphene lattice [15]. One obvious reason for the suppression of foldings in bilayer MoS 2 -which with S-Mo-S as a building block is actually a double trilayer of atoms-is the fact, that bilayer MoS 2 is already much heavier per unit area as bilayer graphene or single layer MoS 2 . The pressure exerted from the substrate during the folding process may thus simply be insufficient to push the material upwards. However, as even more important we consider the fact, that due to the additional inter-plane interactions, the de-facto trilayer MoS 2 is extremely resistant against out-ofplane bending compared to the one-atom thick graphene. Theoretical studies found a significantly larger elastic bending modulus of 9.61 eV for single layer MoS 2 compared to 1.4 eV for graphene [20], in agreement with experimental studies [21,22]. This is in perfect agreement with our findings here. In addition, our data clearly point towards a directional dependence of the bending resistance: folding does become possible when the bending occurs over a low-indexed direction.
As obviously the resistance of the material against bending sets an intrinsic limit to the folding efficiency, a way to overcome this and to force foldings nevertheless, is to have two incisions running very closely together as shown in figures 2(c) and (d). By analyzing the image carefully it can be seen that this folding is indeed due to a coincidental double impact. The first SHI impact had already weakened the single layer MoS 2 flake by creating an incision (which in fact continues behind the folding, figure 2(c)) and a second SHI running very close to the first was then able to cause the folding. The chance for overlapping incisions can be drastically enhanced by crossed beam irradiations, an example of which is shown in figure 2(e). Here, a bilayer of MoS 2 has been irradiated two times with 91 MeV Xe ions under a grazing incidence angle of Θ=1.5°. Between the two irradiations, the azimuthal angle has been rotated by j D =15°. The MoS 2 folds only and exactly at those positions, where the two incisions created by the SHI cross each other. Interestingly, in figure 2(e) two foldings are pointing backwards (events marked with 1), i.e. against the direction of the ion beam, and one is pointing forwards (event marked with 2). From a closer inspection one can learn that the position of the crossover between the two ion trajectories actually determines the direction of the folding. If the second ion crosses a incision at its beginning (event 2), the material folds in the direction of the ion trajectory and if it crosses the incision at the end (events 1), the material folds against the direction of the ion beam. Furthermore, it can be seen that the foldings pointing forwards appear flat at the folding edge, while the foldings pointing backwards exhibit pronounced folding edges. A possible reason for the flat folding edge is that the MoS 2 sheet in these forced foldings is not bent but rather broken, resulting in an open bilayer MoS 2 edge. Similar to grown MoS 2 nano-islands [23] or other artificial MoS 2 nanostructures [24,25], such a synthetic MoS 2 nano-flake would exhibit three chemically active edges, however not necessarily with the same crystallographic orientation.  Next, we will focus our attention on the incisions, a novel feature which may be fabricated by SHI grazing incidence irradiation. As a consequence of the current most commonly used production processes for SL-MoS 2 , i.e.mechanical exfoliation and chemical vapor deposition, respectively, very often also few layers of MoS 2 will be present at least in some parts of the sample. We have therefore investigated how the modifications due to SHI irradiation depend on the layer thickness. To this end, we have irradiated a series of  to trapped water in between the substrate and the Mos 2 , we observe an increased layer thickness of 1.6 nm for the first layer [26]. The next two layers show the typical height value of 0.7 nm [27].
In the case of bulk-like MoS 2 (measured height over 50 nm) chains of nanosized hillocks protrude from the surface. They are exactly aligned along the direction of the SHI beam and from the nominal fluence of 5 ions μm -2 we deduce that each chain corresponds to a single ion trajectory. In most cases, the hillocks are evenly spaced (average distance 60±10 nm), their height is up to 8nm, and their average diameter is 24±3nm. The length of this kind of surface track also varies with the angle of incidence (see below). At Q »  0.5 as shown in figure 4(a), the average length of the chain of nano-hillocks is about one micron (1040±275 nm). Elongated tracks have been observed in bulk MoS 2 also after irradiation with 4MeV C 60 ions at Q   20 , however individual hillocks were absent in these studies [28]. The striking periodicity of the hillocks found here has been observed in several crystalline materials and a possible explanation in terms of a varying electronic density in-between crystal layers has been discussed in detail in [29]. This feature disappears almost completely in thin, few layer MoS 2 sheets (thickness under 10 nm)-although individual hillocks are still created. The surface track is now a mixture of irregularly formed hillocks/protrusions and incisions, the latter of which resemble the ones found in single-and bilayer MoS 2 , see figures 1(b) and 3(a).
In even thinner MoS 2 crystallites, three layers and below, the major part of the surface track consists of a continuous incision. The length of the incision varies between trilayer (3L), bilayer (2L) and single layer (1L) as shown in figure 4(c). The surface tracks in tri-and bilayer MoS 2 are accompanied by protrusions before and after the central incision. Single layer MoS 2 surface tracks generally consist of incisions only and do not show any protrusions. While the complete surface track length in these three different MoS 2 samples is comparable, the length of the incisions is reduced by ≈25% for a bilayer and by ≈50% for a trilayer. Moreover, the morphology of the incisions changes as well. While the incisions in a single layer usually appear as a clean cut with straight edges (see figure 1(b)), bilayer and trilayer incisions appear more irregular. In particular for the trilayer in figure 4(c), it can be seen that the incisions appear more like a series of roundish holes with a diameter of about 10nm each.
The data presented above immediately leads to an interesting question from a physics as well as from an application point of view: How does the substrate influence the fabrication of incisions in SL-MoS 2 ? The substrate does indeed play a decisive role, as we will show further below, but it is by no means a prerequisite. In figure 5 we show scanning transmission electron microscopy (STEM) medium-angle-angulardark-field (MAADF) images from freestanding layers of MoS 2 irradiated with SHI under grazing incidence. Also here, extended incisions oriented along the direction of the incoming ion beam can be found ( figure 5(a)). In comparison they are however much smaller (length typically 100 nm, i.e. about one order of magnitude) and the chance for production is much less likely (here 1 track at a fluence of 15 ions μm -2 ). Atomically resolved images ( figure 5(b)) show that the incisions are extremely narrow (less than 10 nm) and with relatively straight edges on an atomic scale. Surprisingly, the surrounding lattice remains basically undisturbed despite the violent atomic displacements that have taken place inside the ion track core.
Summarizing this part, we have demonstrated that SHI irradiation of bulk-like MoS 2 (corresponding to SL-MoS 2 deposited onto a MoS 2 substrate) will result in nano-hillocks while incisions are reliably written into single-, bi-, and trilayers of MoS 2 deposited onto SiO 2 substrates, and to a lesser degree also into freestanding MoS 2 . In the transition regime of four to fifteen SiO 2 -supported layers, incisions still constitute a major part of the surface track.

Discussion
In the following, we will discuss the possible mechanisms for the fabrication of incisions in ultrathin MoS 2 , as they constitute the typical nanostructure following SHI grazing incidence irradiation. The large difference in incision sizes observed in supported and suspended MoS 2 , respectively, obviously points towards the substrate as a major driving force. The same can be deduced from the analysis of the incision length with respect to the angle of incidence. The corresponding experimental data is presented in figure 6 (blue dots). We assume that the traversing ion initiates the primary electronic excitations and that the subsequent processes will affect the material around the trajectory. Whatever changes are induced, they will be most pronounced in the immediate vicinity of the trajectory while further away the changes are less severe. At some given distance, basically the radius of the affected volume (on the order of a few nanometers), changes will no longer be effective and observable, respectively. If, like in our case, the ion traverses the surface under an angle, this radius will correspond to an effective depth d eff where the primary process can take place and its consequences may still be detected at the surface by an AFM for instance. We can then fit the data from figure 6 by the geometrical relation = Q L d tan eff . The best fit (blue line in figure 6) is obtained for = d 4.5 nm eff , which is a typical value for dielectrics [29,30].
The key point here however is, that d eff is significantly larger than the thickness of a single layer MoS 2 which is »1 nm. It is therefore evident that the processes yielding the incisions in supported SL-MoS 2 cannot take place exclusively in the MoS 2 but must occur mainly in the substrate material. We therefore need to consider what happens in the SiO 2 when the ion passes through, and how this affects the SL-MoS 2 . As already mentioned, the direct collision of a SHI with the target atoms is negligible in this energy range. Instead, during its passage through matter an SHI excites the electrons of the target material by direct ionization and secondary electron creation which then propagate through the target. Several mechanisms for material modifications related to the electronic excitation due to electronic stopping have been proposed. In particular, these are Coulomb explosion as a consequence of the electrostatic repulsive forces between the atoms in the ionized region [31][32][33], exciton self- trapping causing a local lattice distortion [34], nonthermal melting caused by significant changes of the interatomic potentials [35], or phase transitions such as melting due to a thermal spike [36,37]. These mechanisms occur on vastly different time scales, ranging from a few fs in the case of processes related to non-thermalized electrons up to ps in the case of phase transitions.
In non-metallic systems the resulting modifications are very often permanent in nature and are thus easily detectable (for a review see e.g. [38]), in the case of graphene see e.g. [39]. After irradiation under grazing incidence, hillock chains, or more general, protruding surface tracks, are frequently observed in crystalline dielectrics or graphite [13,29,[40][41][42][43]. The surface tracks which are shown in figure 4(a) in bulk MoS 2 bear a striking resemblance to the ones found in e.g.SrTiO 3 . For SrTiO 3 in particular, the origin of these hillock chains has been successfully explained in terms of a thermal spike model, i.e.the hillocks consist of re-solidified molten material. For MoS 2 a rather low decomposition temperature of » T 1460 c K has been reported [44,45]. This and the similarity in morphology leads us to believe that also the hillock chains in bulk-like MoS 2 consist of (potentially nonstoichiometric) MoS 2 which has re-solidified after a phase transition caused by the thermal spike. In thin layers of MoS 2 instead of hillocks, incisions are formed and therefore a different mechanism must be in effect. On the basis of the previous observation that the process cannot take place in the ultra-thin MoS 2 alone, we put forward the following hypothesis for the incision formation in supported MoS 2 : a thermal spike in the SiO 2 substrate results in a super heated surface track and in areas, where the temperature of the surface track in SiO 2 reaches or even exceeds the decomposition temperature of MoS 2 , and thus material is evaporating from the track region leaving nanoscaled incisions behind. To test our hypothesis, the corresponding track length in SiO 2 samples have been measured as well. The data (green squares in figure 6) are in good agreement with our hypothesis. Furthermore, the difference in the effective depth determined from the two fits corresponds roughly with the thickness of the MoS 2 layer, which further corroborates our hypothesis.
Note that preferred sublimation of one atom species may occur. Previous studies of thermally induced desorption found that S 2 vaporization starts already at 1140K and in the presence of water decomposition may occur at even lower temperatures (565 K) due to the vaporization of SO 2 . The lattice temperature in the spike is definitely high enough to decompose MoS 2 but in the border region, where temperatures are lower, preferentially sulfur atoms may evaporate which would be beneficial for any potential catalytic activity [46,47] of the incision but also with respect to the interesting optical and electronic properties of metallic edge states in MoS 2 [48,49]. In order to estimate the minimum energy required to fabricate an incision, one can simply look into the available data for track formation in SiO 2 , as e.g.found in [50]. From this we infer that a stopping force of 2keV nm -1 would be already sufficient. Note that this is the threshold for grazing incidence and it can thus not be directly compared to previous work on SHI and proton irradiation of MoS 2 field effect transistors, where it was shown that the devices withstand surprisingly high fluences at much higher stopping forces before deterioration [51,52]. Finally, we wish to point out that up-scaling to large area samples or down-scaling to localized single ion hits [53] is entirely possible with our approach which opens up ample possibilities for applications. In particular, it remains to be seen whether SHI causes direct damage in freestanding SL-MoS 2 also under non-glancing angles which would due to the resulting minute pore size immediately pave the way to exciting experiments such as the fabrication and study of synthetic nanopores for DNA sequencing. Furthermore, we expect that other 2D-TDMCs, in particular MoSe 2 , WS 2 and WSe 2 , will show a very similar response towards SHI irradiation.
In conclusion, we have shown that by choosing the proper irradiation conditions, two distinct morphological changes can be fabricated in single layers of MoS 2 by individual ions, namely foldings and incisions. Folding is significantly suppressed in MoS 2 due to its high bending resistance, but can still be routinely obtained by either exploiting the directional dependence of the bending resistance or by crossed-beam irradiations. The incisions represent a novel feature and are unique to the crystalline 2D semiconductor as surface tracks in the 3D form manifest themselves as material protruding from the surface which is the typical morphology for surface tracks in bulk dielectrics. The widths of the incisions in supported samples is around 10nm while their aspect ratio (width:length) can be tuned from more than 1:1000 down to nanopores with a ratio of 1:1. Incisions in suspended single layers are found to be one order of magnitude smaller in length. For supported MoS 2 the formation mechanism of the incisions could be linked to an ion induced thermal spike in the SiO 2 substrate.

Sample preparation
Supported samples have been prepared via mechanical exfoliation from a bulk single crystal of MoS 2 (Molykote). Suspended samples were produced by transferring CVD-grown MoS 2 onto a TEM grid. At first, single crystal MoS 2 flakes were grown directly on a 300nm SiO 2 /Si substrate by CVD. A 1% sodium cholate solution is initially spin coated onto the SiO 2 substrate to act as a growth promoter. A micro droplet of a saturated solution of Ammonium heptamolybate (AHM) is deposited onto the corner of the substrate.
The AHM will act as the Molybdenum feedstock. The substrate is placed in the center of a 1 inch Lindberg blue furnace and 15mg of sulfur is placed upstream at a distance of 18cm from the growth substrate. 500sccm of Nitrogen is flown through the chamber and the temperature of the furnace is ramped up to 800°C, the sulfur pellet is heated up to 150°C. After a 30min growth, the furnace is then stopped and rapidly cooled. The growth substrate is retrieved and MoS 2 flakes were grown across the SiO 2 substrate. Then, a thin Poly(methyl methacrylate) layer (PMMA, ARP 671.05; ALLRESIST GmbH) is spin coated on the MoS 2 sample (30 sec, 1000 rpm) and annealed under ambient conditions (150°C, 4 min). The wafer is then placed onto the surface of a KOH solution (3 g of KOH and 50 ml DI water) which etches the SiO 2 away so that the PMMA/MoS 2 detaches from the wafer. After replacing the etchant with DI water one can gradually reduce the amount of the liquid letting the PMMA/ MoS 2 stack sink onto the TEM grid. The remaining PMMA is dissolved in a bath of acetone for 45min.

Irradiation
The irradiation with SHI took place at the large scale facilities GANIL in Caen, France. MoS 2 on SiO 2 has been irradiated with 91MeV Xe ( = S e 19keV nm -1 ) and 84MeV Ta ( = S e 18keV nm -1 ) with typically 10ions μm -2 at varying angles of incidence in the range from  0.25 to 5 • with an accuracy of 0.2°(see [14] for details). Suspended MoS 2 samples have been irradiated with 93MeV Xe ions at a nominal fluence of 15ions μm -2 and at an angle of incidence of   0.2 5 . Due to the nature of these samples a more accurate control of the angle of incidence is currently not possible, which also prevents a more systematic analysis of the freestanding samples at this point.

Imaging
Atomic force microscopy images were taken at the University of Duisburg-Essen with a Veeco Dimension 3100 AFM using Nanosensors NCHR tips for the measurements in ambient conditions. The STEM experiments were carried out with an aberration corrected Nion UltraSTEM 100 microscope in Vienna. The microscope was operated at 60 kV in near ultrahigh vacuum (~´-2 10 7 Pa). The beam convergence semiangle was 30mrad and the semi-angular range of the MAADF detector was 60-200mrad.