Perforating freestanding molybdenum disulfide monolayers with highly charged ions

TL;DR

This study demonstrates that irradiation with highly charged ions can efficiently create well-defined, narrow-sized pores in freestanding molybdenum disulfide monolayers, with potential applications in filtration and sensing.

Contribution

It introduces a novel ion irradiation technique using highly charged ions to fabricate precise pores in MoS₂ monolayers, highlighting the role of electronic excitation in defect formation.

Findings

Pore creation efficiency increases linearly with ion potential energy.

Pore sizes depend on the potential energy of the ions used.

Irradiation produces pores with radii between 0.3 and 3 nm.

Abstract

Porous single layer molybdenum disulfide (MoS$_2$) is a promising material for applications such as DNA sequencing and water desalination. In this work, we introduce irradiation with highly charged ions (HCIs) as a new technique to fabricate well-defined pores in MoS$_2$. Surprisingly, we find a linear increase of the pore creation efficiency over a broad range of potential energies. Comparison to atomistic simulations reveals the critical role of energy deposition from the ion to the material through electronic excitation in the defect creation process, and suggests an enrichment in molybdenum in the vicinity of the pore edges at least for ions with low potential energies. Analysis of the irradiated samples with atomic resolution scanning transmission electron microscopy reveals a clear dependence of the pore size on the potential energy of the projectiles, establishing irradiation with highly charged ions as an effective method to create pores with narrow size distributions and radii between ca. 0.3 and 3 nm.

Equations

$$E^{\text{dep}}(q)={\left(\frac{dE}{dx}\right)}_n(q)+{\left(\frac{dE}{dx}\right)}_e(q)+E_{\text{pot}}^{\text{dep}}(q),$$

Full text

\alsoaffiliation

TU Wien, Institute of Applied Physics, A-1040 Vienna, Austria

\alsoaffiliationDepartment of Applied Physics, Aalto University, P.O. Box 11100, FI-00076 Aalto, Finland

{tocentry}

1 Introduction

Since the isolation of a single sheet of graphene and the discovery of its extraordinary properties, two-dimensional (2D) materials have attracted much attention 1, 2, 3. Due to its electronic 4, chemical 5 and optical properties 6, molybdenum disulfide (MoS2), a member of the family of 2D transition-metal dichalcogenides is one of the most prominent candidates in this field. These properties, combined with available synthesis of large-scale MoS2 single layers by chemical vapour deposition (CVD) 7 open a wide range of potential applications including solar cells 8, catalysts 9, 10, power generators 11, water desalination 12, protein translocation 13, 14, and DNA sequencing 15, 16, 17, 18.

Some of these applications require the introduction of well-defined openings with radii of a few nm in the 2D material. For example, a pore radius of $\leq$ 4 nm is suitable for the differentiation of nucleotides 17, whereas water desalination requires pores of $\leq$ 1 nm radius 19. For both applications the created pores would preferably have a narrow size distribution. This is typically achieved through irradiation with energetic particles, i.e. electrons or ions 20. While high-energy electrons are able to perforate freestanding MoS2 easily enough, their potential in this context remains limited because mostly single or double vacancies are created 21. This limitation can be overcome by drilling each hole separately with a focussed beam, but this is not suitable for mass production. Another way is to expose the whole membrane to a high flux of electrons, which however will widen all defects initially present (and not only the artificially introduced ones) into larger pores and finally result in an uncontrollable size distribution and pore density as shown for single-layer hexagonal boron nitride 22, 23. Because large-area MoS2 samples are typically grown via CVD and thus contain a non-negligible number of intrinsic defects, this represents a severe drawback for up-scaling. The same is also true for methods relying on plasma treatment or chemical etching 24.

Individual impacts of ions with kinetic energies of a few keV have been shown to also yield mainly single and double vacancies in MoS2, 25 and again, by focussing the beam, larger pores may be drilled into the membranes. In this way, pores with radii of 20 nm have been created 26. For ion beams this is, however, not the limit, because in contrast to electron beams, a much wider range of experimental parameters exist for tuning the defect creation mechanism. For example, just recently it was shown that the bombardment of MoS2 with swift heavy ions under grazing incidence leads to the fabrication of foldings and incisions 27. The key feature is the energy deposition into the MoS2 by electronic excitations rather than by nuclear collisions due to the very high kinetic energy of the projectile. This activates completely different defect creation mechanisms through electronic excitations resulting in structural changes far beyond single and double vacancies. The generation of a sufficiently strong electronic excitation does not necessarily require swift ions (and thus a large accelerator), but may also be achieved by using slow highly charged ions (HCIs) with velocities below the Bohr velocity $v_{0}$ 28. The deposition of their potential energy, i.e. the sum of the ionisation energies of all electrons stripped from the projectile, can result in various structural changes, ranging from hillocks 29 over pits 30 in 3D materials, and holes in ultrathin carbon nanomembranes 31, to non-topographical features like frictional changes in graphene 32 and hexagonal boron nitride 33. Due to the energy deposition in the shallow depth range of a HCI during the interaction with surfaces, these projectiles are well suited for the modification of 2D materials. Furthermore, multiple pores can be created by exposing the whole sample to a scanned beam. This enables parallel writing and thus provides an avenue for mass production.

In this letter, we provide experimental evidence that highly charged ions can be used to manufacture well-defined openings in freestanding MoS2 with sizes suitable for applications such as DNA sequencing and water desalination. We show that the pores are produced by individual ion impacts, and determine, for a wide range of beam parameters the probability with which a pore is created. Our approach ensures a controlled pore density that is independent of any intrinsically present defects. The pore radius can be tuned with great accuracy in the range from 0.5 nm to 5 nm. In addition, we demonstrate that in particular for the smaller technologically important pores, the size distribution is extremely narrow and thus fulfills even the severe requirements of water desalination. By theoretical modelling of the ion-solid interaction, we can clearly exclude momentum transfer between the impinging ion and the atoms of the target material as the physical mechanism for pore creation. This result underlines that it is the energy associated with the charge state of the ion which gives rise to the efficient removal of atoms from a nanometer-sized area. Both the simulations as well as the experiments hint towards molybdenum enrichment at the pore edges which could be advantageous for DNA sequencing and catalysis.

2 Experiment

For the irradiation of MoS2 monolayers with HCIs, freestanding samples were used to ensure that the ions deposit their energy exclusively within the MoS2, and to avoid substrate effects on the modification of the 2D material. Large-scale monolayers of MoS2 were grown on SiO2 via CVD in a microcavity set-up in Dublin as decribed in reference 34 and in a custom-made process setup as an extended version of the system reported by Lee et al. 7 in Duisburg. The 2D material was transferred onto perforated membranes of amorphous carbon via wet transfer as described in reference 27. This QUANTIFOIL®, with a regular array of 1.2 µm pores, was supported by a 400 mesh gold transmission electron microscopy (TEM) grid. This sample system allows us to exclude substrate-induced irradiation damage and to image the structure of the monolayer on the atomic scale by means of scanning transmission electron microscopy (STEM).

The irradiation of MoS2 monolayers was carried out at the HICS beamline 35 at the University of Duisburg-Essen. This facility is equipped with an electron beam ion source (EBIS), which can deliver xenon ions with a maximum charge state of $q=48+$. Several charge states, ranging from $q=20+$ up to $q=40+$ were used in this work corresponding to a potential energy range of 4.6 keV $\leq E_{\textrm{pot}}\leq$ 38.5 keV. The irradiation took place under perpendicular incidence with fluences ranging from $\Phi=2.800$ µm*-2* for Xe40+ to up to $\Phi=18.000$ µm*-2* for Xe20+. To study the role of the charge state $q$, the kinetic energy was kept fixed at a constant value of $E_{\textrm{kin}}=180$ keV for all irradiations. The sample was at room temperature and the vacuum pressure during the irradiation was approximately $1\times 10^{-9}$ mbar.

The STEM measurements were carried out in Vienna using an aberration-corrected Nion UltraSTEM 100 microscope with a high-angle annular dark-field (HAADF) detector. The electron acceleration voltage was 60 kV and the HAADF detector semiangle was $80-300$ mrad. A typical STEM image of a pristine MoS2 monolayer is presented in Figure 1a. It shows a regularly arranged honeycomb lattice, without pores or other structural defects within the crystalline domains (the bright features are Mo, darker ones correspond to the two-atomic S column). After the interaction with HCIs, however, the STEM images reveal the presence of pores, examples of which are shown for four different charge states in Figures 1b-e. Overview images are provided in the Supplementary Figure S1.

To investigate the dependency of the pore size on the charge state of the ion, an average number of over 200 pores per irradiation were surveyed and for each pore its area $A$ was determined. Since the majority of the pores shows a roughly round shape, the areas have been converted to radii $r$ using $A=\pi r^{2}$ and the obtained radii collected in histograms. As an example, the normalized pore size histograms for Xe20+, Xe31+ and Xe40+ are shown in Figure 2a. All histograms have been fitted using a Gaussian distribution.

Comparing the three histograms in Figure 2a, one can clearly see that for decreasing charge state, the distribution shifts to smaller pore sizes. The mean value of the pore radius for Xe20+ is ca. 0.55 nm, whereas pores larger than 2.65 nm in radius are created for Xe40+. For comparing the histograms, the mean value of each distribution was determined and plotted in Figure 2b. The standard deviation and the uncertainty of the mean are represented by the red bars and the error bars, respectively. This diagram summarizes the STEM data for all charge states, showing a clear correlation between the potential energy and the pore size. This dependency was best fitted using a linear function $r(E_{\textrm{pot}})=aE_{\textrm{pot}}+b$ with the slope $a=(6.04\pm 0.32)\cdot 10^{-2}$ nm/keV and the offset $b=(0.27\pm 0.06)$ nm.

When analysing our data, it became apparent that the probability to create a pore upon impact (pore density/fluence) decreases with decreasing $E_{\textrm{pot}}$ (Figure 2c). The uncertainties correspond to the limited accuracy of the current measurement in the picoampere regime. In the case of Xe40+, an efficiency of $(93\pm 14)$% was determined, meaning that the observed pore density is nearly identical to the applied fluence. This implies that practically every HCI created a pore in the sample. However, decreasing the potential energy of the projectile, the defect creation efficiency decreases down to $(15\pm 2)$%. A linear fit of these data reveals a slope of $(2.1\pm 0.4)$ %/keV and an offset of $(6.0\pm 8.3)$%.

3 Discussion

Let us now discuss the experimental results, starting with the pore size. Figures 2a and b clearly show that nanometer-sized pores are induced into the 2D material due to the interaction with HCIs. The pore radius can be adjusted in the range of $0.3-3$ nm by varying the potential energy $E_{\textrm{pot}}$ of the incoming ion. Comparing the pore radius with previous experimental data of HCI-irradiated MoS2 monolayers supported on KBr 37, it stands out immediately that the pore area is about one order of magnitude smaller in our experiment. Although the kinetic energy of our projectiles is clearly lower (180 keV instead of 260 keV), this cannot explain the large discrepancy between the two systems. Instead, a probable reason for the reduction of the pore size is the lack of a substrate. It is well known that the substrate can have a significant influence of the modification of 2D materials 38, 39. In addition to the direct interaction of the HCI with MoS2, atoms sputtered from the substrate may lead to an additional removal of atoms from the material, resulting in an increase of the pore size for the supported sample 40.

By decreasing $E_{\textrm{pot}}$, the radius can be reduced down to the offset value of the linear fit, as shown in Figure 1, representing the extrapolated limiting size of a pore that would be induced into a MoS2 monolayer after interaction with a neutral Xe atom. Without any potential energy, the pore creation process would be driven by the deposition of kinetic energy. This pore of $0.27\pm 0.06$ nm would correspond to ca. 9 missing molybdenum and twice as many missing sulphur atoms, and represents the lower limit of the achievable pore size in suspended MoS2 by Xe irradiation for the given kinetic energy. Although theoretical simulations predicted comparable pore production after irradiation with neutral Xe projectiles 25, to our knowledge, no experimental data on Xe irradiation of MoS2 monolayers has been published so far.

Coming to the pore size evolution presented in Figure 2b, the linear increase of the pore radius with the potential energy of the ion proves that this parameter plays a crucial role in the defect creation process. The experimental data even suggests that the number of sputtered atoms per keV of potential energy increases significantly with $E_{\textrm{pot}}$ of the HCI. While approximately 7 atoms per keV were removed with Xe20+ (overall 33 atoms), nearly three times as many (ca. 20 atoms/keV or 765 atoms in total) have left the 2D material due to an impact with a Xe40+. Qualitatively, similar findings have been obtained for other materials, including graphene supported by SiO2 32. A possible explanation could be found in the emission characteristics of the sputtered particles. For increasing potential energy, it becomes more likely to sputter clusters of atoms 41. This consumes less energy than sputtering single atoms, since the energy cost for breaking the bonds within the clusters is saved. As a consequence, more atoms (individually or in clusters) per energy unit can be emitted from the surface for higher charge states.

Next, let us try to shed light on the defect creation process. This is driven by the energy $E^{\textrm{dep}}$, which is deposited into MoS2 during the HCI-MoS2 interaction. This energy can be described as

[TABLE]

with $E_{\textrm{pot}}^{\textrm{dep}}(q)$ being the amount of potential energy of the HCI deposited into the electronic subsystem of the target and $\left(\frac{dE}{dx}\right)_{n,e}(q)$ being the nuclear and electronic stopping power. Whereas the dependency of $E_{\textrm{pot}}^{\textrm{dep}}$ on the charge state $q$ is quite intuitive, it has just recently been shown that the stopping power of slow HCIs in graphene shows a strong dependency on $q$ 42.

As sputtering is typically driven by nuclear stopping, it is instructive to assess the impact of the charge-dependent nuclear stopping $\left(\frac{dE}{dx}\right)_{n}(q)$ of the ion on the defect creation process. To this end, we carried out molecular dynamics simulations using the large-scale atomic/molecular massive parallel simulator (LAMMPs) package 43. Additional details on the MD simulation are provided in the Supporting information.

To model the ionic charge state in the ion irradiation, a charge state-dependent potential in MD simulations based on a modified electrostatic potential for nuclear energy transfer was used 42. The electronic-stopping power $\left(\frac{dE}{dx}\right)_{e}$ was kept constant and charge-independent in this simulation. A value of 41 eV/Å was calculated from SRIM 44(irradiation of bulk MoS2 using 180 keV Xe) and was adapted for the monolayer. The energy brought in by the ion impact was estimated as a product of electronic stopping power and layer thickness (6.15 Å). It was deposited within a circular area around the impact point with a radius of 10 Å. Simulations with different physically meaningful radii gave qualitatively similar results, with the radius being within 50% of what is presented in Figure 3 (for details, see Supplementary Figure S2). In any case, as evident from the following discussion, charge-independent electronic stopping cannot explain the experimental observations, as the deposited potential energy $E_{\textrm{pot}}^{\textrm{dep}}(q)$ was not taken into account.

The results of the simulation (orange) together with the experimental results (black) are presented in Figure 3. It is first of all apparent that the MD simulation also shows a perforation of the MoS2 single layer, exhibiting similar pore shapes compared to the experimental results. However, the calculated pore radii, ranging from about 0.35 nm for Xe20+ to 0.37 nm for Xe40+, are significantly smaller than the experimental results. Whereas for Xe20+, the simulated pores are by a factor of ca. 1.5 smaller compared to the experiment, the discrepancy increases further to a factor of more than seven in the case of Xe40+. This indicates that considering just the nuclear stopping of the projectiles to be charge-state-dependent and accounting for charge-state-independent electronic stopping is not sufficient to reproduce the experimental results. Instead, one must consider the electronic stopping to be charge-state-dependent as well, and even more importantly, implement the deposition of the potential energy of the HCI in the calculation. This, however, is not easily achieved, as e.g. not all energy stored in the projectile is deposited in the layer. For graphene, where efficient charge exchange can be expected, it has been shown that most projectiles retain a significant amount of potential energy after transmission 29. Clearly, further experiments to obtain more information about the interaction of the HCIs with monolayer MoS2 (i.e. energy loss, charge transfer, etc.) need to be carried out to help improving the computational models. This, however, is a major undertaking and thus beyond the scope of this paper.

The impact of an HCI on a 2D material may trigger the emission of electrons 29, 45 as well as a local temperature increase 33. In the case of a MoS2 monolayer, heat treatment and electron bombardment have both been shown to result in a depletion of sulfur atoms 46, 47. Interestingly, indications for this are also found in our calculations. In the inset of Figure 3, a result of the MD simulation is depicted, showing a pore in the MoS2 that is created after the impact of a Xe40+ ion. Besides the pore creation, the simulation additionally shows amolybdenum enrichment in the direct vicinity of the pore. By comparing the amount of sputtered ions (see Supplementary Figure S3), the simulation suggests a depletion of sulfur, while sputtering of molybdenum appears to be negligible. Furthermore, the simulation suggests a displacement of molybdenum atoms from the lattice to the surface, resulting in an aggregation of molybdenum clusters at the pore edges (see Supplementary Figure S4), which has also been reported for MoS2 monolayers irradiated with electrons 48, 49. Although a direct comparison between the simulations and experiments is not warranted due to the prevailing limitations in the computational methods, we note that a similar molybdenum aggregation was in many cases also observed experimentally. For an example see Supplementary Figure S5.

Next, let us discuss the linear evolution of the defect creation efficiency with the charge state of the ion. This linear increase without a threshold is in contrast to earlier observations with HCIs on surfaces of bulk materials and other 2D materials 50, 32. There, a threshold value for the energy deposition can be determined in a narrow range, separating a domain with a certain finite constant efficiency for the modification of the target material from the regime where the energy deposition is not sufficient to modify the irradiated system resulting in zero efficiency. In a number of experiments, a decrease of the defect creation efficiency was observed in a narrow range near the threshold value 51. In our experiment, however, the evolution of the efficiency spans an order of magnitude in potential energy. Whereas approximately every fifth ion creates a pore for $E_{pot}\approx 4.6$ keV (Xe20+), the efficiency increases linearly up to a value close to 100% for $E_{pot}\approx 38.5$ keV (Xe40+). This is a truly surprising finding, which indicates that the probability for energy deposition from the HCI to the sample must depend either on the charge state or the potential energy in a manner that has not been previously observed. In a bulk sample this is obscured, because the penetration depth of the HCI is limited and thus all potential energy will be deposited close to the sample surface with 100% likelihood. Clearly, in order to improve our understanding on the interaction between HCI and 2D materials, systematic study of HCI irradiation of structures with different properties is necessary.

Finally, we wish to point out that we see no reason why the pore creation process itself should be specific for MoS2. While in other TMDCs details of the defect structure and their exact size might differ, we are confident that pore creation of similar size range is entirely feasible.

4 Conclusion

In summary, this paper presents the first experiments focused on the interaction of highly charged ions with freestanding monolayers of MoS2. The successful creation of pores in the otherwise pristine lattice of the 2D material was observed, and the radius of the pores ranging from 0.55 nm to 2.65 nm was controlled by changing the charge state of the projectile. To disentangle the contribution of the charge dependent deposition of the kinetic energy and the deposition of the potential energy of the highly charged ion to the defect formation process, MD simulations were carried out showing that considering only the nuclear stopping to be charge-state-dependent and electronic stopping to be charge-independent is not sufficient to reproduce the experimental results, suggesting the charge state-dependent electronic stopping and the potential energy to be the driving force. Although additional experiments need to be carried out to clarify the underlying defect creation process, irradiation with highly charged ions clearly represents a novel technique for fabricating porous MoS2 monolayers with pores in the size regime meeting the strict requirements of water desalination or DNA sequencing.

{acknowledgement}

We acknowledge financial support from the DFG within the SFB 1242 ("Non-Equilibrium Dynamics of Condensed Matter in the Time Domain") Project C5, within the Project NU-TEGRAM (SCHL 384/16-1), SCHL 384/20-1, KR 4866/2-1 and WI-4691/1-1, the Austrian Science Fund projects I3181, P31605 and P28322 as well as the Wiener Wissenschafts-, Forschungs- und Technologiefonds (WWTF) project MA14-009. G.S.D acknowledges support of the Science Foundation Ireland under grant No 12/RC/2278 and 15/IA/3131. The computational support from the HZDR computing cluster is gratefully appreciated.

{suppinfo}

The following files are available free of charge.

• •

Details on the molecular dynamics simulation

• •

Figure S1: STEM overview and close-up images of pristine and ion irradiated MoS2

• •

Figure S2: Results of the MD simulation, showing the pore size dependency on the energy deposition radius $r_{d}$

• •

Figure S3: Results of the MD simulation, showing the average amount of sputtered sulphur and molybdenum atoms.

• •

Figure S4: Results of the MD simulation, showing the atomic lattice and the molybdenum density after ion impact.

• •

Figure S5: STEM images of HCI irradiated MoS2 showing a aggregation of Mo atoms on the pore edges.