Electron Irradiation of Metal Contacts in Monolayer MoS2 Field-Effect Transistors.

Metal contacts play a fundamental role in nanoscale devices. In this work, Schottky metal contacts in monolayer molybdenum disulfide (MoS2) field-effect transistors are investigated under electron beam irradiation. It is shown that the exposure of Ti/Au source/drain electrodes to an electron beam reduces the contact resistance and improves the transistor performance. The electron beam conditioning of contacts is permanent, while the irradiation of the channel can produce transient effects. It is demonstrated that irradiation lowers the Schottky barrier at the contacts because of thermally induced atom diffusion and interfacial reactions. The simulation of electron paths in the device reveals that most of the beam energy is absorbed in the metal contacts. The study demonstrates that electron beam irradiation can be effectively used for contact improvement through local annealing.

Introduction

Molybdenum disulfide (MoS2) is one of the most studied transition metal dichalcogenides, owing to its layered structure and useful mechanical, chemical, electronic, and optoelectronic properties.[1−4] A molybdenum (Mo) atomic plane sandwiched between two sulfur (S) planes constitutes the monolayer that is bonded to other monolayers by weak van der Waals forces to form the bulk material. MoS2 is a semiconductor suitable for several applications,[5−9] having a 1.2 eV indirect band gap in the bulk form that widens up to 1.8–1.9 eV and becomes direct in the monolayer.[3] Despite the lower field-effect mobility than graphene,[10,11] ranging from few tenths to hundreds[12−15] of cm2 V–1 s–1, MoS2 field-effect transistors (FETs) have recently become very popular as alternatives to graphene FETs[12−17] for next-generation electronics based on 2D materials.[18−25]

The fabrication and characterization of devices based on 2D materials greatly rely on the application of electron beam (e-beam) lithography or focused ion beam processing and on scanning electron microscopy (SEM) or transmission electron microscopy, which imply irradiation by charged particles. The exposure to low-energy electrons and/or ions can modify the electronic properties of the 2D materials or their interfaces.[9,17,26] Indeed, structural defects can locally modify the band structure and behave as charge traps, thereby changing the device characteristics both in the case of e-beam[27,28] and ion beam irradiation.[29,30] Conversely, electron beam, ion irradiation, or plasma treatments can be intentionally used for nanoincisions,[31] for pores,[32] or to purposely create defects, for instance, to reduce the contact resistance.[33−35] Choi et al. reported the effects of 30 keV electron beam irradiation of monolayer MoS2 FETs, showing that irradiation-induced defects act as trap sites by reducing the carrier mobility and concentration and shifting the threshold voltage.[36] A study of point defects in MoS2 using SEM imaging and first-principles calculations, by Zhou et al., demonstrated that vacancies are created by e-beam irradiation at low energies,[37] below 30 keV. Durand et al. studied the effects of e-beam on the MoS2-based FET, reporting an increase in carrier density and a decrease in mobility explained as irradiation-induced generation of intrinsic defects in MoS2 and as Coulomb scattering by charges at the MoS2–SiO2 interface, respectively.[38] Giubileo et al. reported a negative threshold voltage shift and a carrier mobility enhancement under 10 keV electron irradiation of few-layer MoS2 FETs attributed to beam-induced positive charge trapped in the SiO2 gate oxide.[27]

In this paper, we present the spectroscopic and electrical characterization of monolayer MoS2-based FETs, with Schottky Ti/Au contacts, focusing on the effects of low-energy e-beam irradiation. We show that the long exposure of the metal contacts to 10 keV e-beam in a SEM chamber enhances the transistor’s on-current. We explain such an improvement by radiation-induced lowering of the Schottky barrier at the metal contacts. We perform Monte Carlo simulation to track the e-beam through the device and show that when the beam is focused onto the contacts, most of the beam energy is absorbed within the metal. The local heat can induce atomic diffusion and interfacial reactions that change the chemical composition and structure of the metal–MoS2 interface or can generate or release tensile strain. Both effects cause the lowering of the Schottky barrier and the consequent increase in transistor current.

Our study shows that electron beam exposure during SEM imaging has non-negligible effects on MoS2 devices; however, it also highlights that a suitable exposure, with the e-beam focused on the contact region, can be conveniently exploited to reduce the contact resistance of the transistor. Compared to thermal annealing, our finding provides a way to improve the contact resistance by local conditioning, which avoids the exposure of the entire wafer to a high thermal budget.

Fabrication and Experimental Methods

The MoS2 monolayer flakes were grown via chemical vapor deposition in a three-zone split tube furnace, purged with 500 Ncm3/min Ar gas for 15 min to minimize the O2 content. The growth SiO2/Si substrate was spin-coated with a 1% sodium cholate solution; then, a saturated ammonium heptamolybdate (AHM) solution was first annealed at 300 °C under ambient conditions to turn AHM into MoO3 to be used as the source for molybdenum. The target material was placed in a three-zone tube furnace along with 50 mg of S powder, positioned upstream in a separate heating zone. The zones containing S and AHM were heated to 150 °C and 750 °C, respectively. After 15 min of growth, the process was stopped, and the sample was cooled rapidly.

We realized FETs using the SiO2/Si substrate (thickness of the dielectric: 285 nm) as the back gate and evaporating the drain and source electrodes on selected MoS2 flakes through standard photolithography and lift-off processes. The contacts were made of Ti (10 nm) and Au (40 nm) used as adhesion and cover layers, respectively. Ti was deposited in high vacuum, which could not exclude the formation of TiO2, contributing to the resistance and Schottky barrier at the contacts. Figure a,b shows the SEM top view of a typical device and its schematic layout and measurement setup. The channel is made up from a monolayer flake [as confirmed by Raman and photoluminescence (PL), see below] of width and length of 20 and 4 μm, respectively, and a nominal thickness of 0.7 nm. Atomic force microscope (AFM) images (Figure c,d,e) show that the flake has an average height of 1.2±0.3 nm (which is typical for single-layer MoS2 measured in air by AFM) and appears to be extremely flat (roughness rms < 0.25 nm) and structurally intact. There are some contaminants because of the lithography process, which are weakly bound and can be swept by the AFM tip. Contacted and noncontacted flake areas do not differ with respect to contamination density—spectroscopic data should thus be comparable.

Figure 1

(a) SEM image of the MoS2 device and contact labels. (b) MoS2 FET layout and schematic of the common source configuration used for electrical characterization. (c) AFM image of the MoS2 flake between the electrical contacts, which appear here in white as the scale has been adjusted to properly image the MoS2 flake. (d) Zoom-in into the upper region of (c), showing that the flake is flat and structurally intact. The rms roughness is 0.221 nm for the SiO2 substrate and 0.237 nm for MoS2. (e) Height distribution taken from image (d), yielding a flake height of ∼1.2 nm.

A total of seven MoS2 channels of identically prepared FETs have been characterized by Raman and PL spectroscopy just after processing. The measurements were performed with a Renishaw InVia Raman microscope at the Interdisciplinary Center for Analytics on the Nanoscale (ICAN). The excitation laser wavelength was 532 nm, and the power density was kept below 0.1 mW/μm2 to avoid damage to the MoS2 flake. Exemplary spectra of Raman characterization are shown in Figure . The chosen reference measurements are spectra obtained from MoS2 flakes on the same substrate, which were also in contact with the photoresist and various solvents during the processing and lift-off for the production of the FETs, but are not in contact with metal electrodes themselves. The shape of the PL spectra (Figure a) and the difference of the Raman modes (Figure b) differ significantly. The PL intensity (sum of all excitons and trions) for noncontacted MoS2 flakes is higher by a factor of 1.7 ± 0.8 than that for contacted MoS2. The mode differences for noncontacted and contacted MoS2 are 21.3 ± 0.7 cm–1 and 19.7 ± 0.7 cm–1, respectively. Both the changes in PL and Raman mode difference can be associated with built-in strain or changes in the electronic properties and the band structure of the MoS2 sheets.[39−43] From the linear dependencies of Raman mode positions on doping and strain,[39,40] we find a reduction of tensile strain by (0.46 ± 0.28) % and an increase in electron doping of 0.44 ± 0.36 × 1013 electrons per cm2 for the contacted 2D material in comparison with noncontacted MoS2 (details of the calculation method can be found in ref (44)). Hence, the significant alterations in the spectroscopic pre-characterization of the MoS2 channels can be clearly attributed to electronic and structural changes at the metal contact.

Figure 2

(a) PL and (b) Raman spectrum of monolayer MoS2 after FET processing. Blue: contacted MoS2 monolayer flake and red: noncontacted monolayer MoS2 flake.

In the following, most of the electrical characterization refers to the transistor between the contacts labeled C2 and C3 in Figure a. The contact C3 was used as the drain and C2 as the grounded source. The electrical measurements were carried out inside a SEM chamber (LEO 1530, Zeiss), endowed with two metallic probes with nanometer positioning capability, connected to a Keithley 4200 SCS (source measurement units, Tektronix Inc.), at room temperature and a pressure of about 10–6 mbar. The e-beam of SEM, set to 10 keV and 10 pA, was used for the time-controlled irradiation of specific parts of the device.

Results and Discussion

The output (Ids–Vds) and the transfer (Ids–Vgs) characteristics of the transistor are shown in Figure a,b, respectively. The output curve shows rectification with the forward current appearing at negative Vds, typical of a p-type Schottky diode, while the transfer characteristic shows an n-type transistor. This apparently contradictory behavior has been previously reported for MoS2 and WSe2 transistors and explained by the formation of two back-to-back and possibly asymmetric Schottky barriers at the contacts.[45,46] The forward current at negative Vds is caused by the different contact areas and by the image force barrier lowering of the forced junction (i.e., the drain, C3, in our case), while the reverse current at Vds > 0 V is limited by the grounded junction at the source (C2) contact. As the barrier lowering is more effective on the forced junction, the voltage being directly applied to it, the negative bias gives rise to the higher (apparently forward) current.

Figure 3

Output (a) and transfer (b) characteristics of the device between C2 and C3 contacts, with C3 used as the drain and C2 as the grounded source.

After the initial electrical characterization, we performed two sets of exposures to the SEM electron beam. Each exposure lasted 300 s, corresponding to a fluence of ∼180 e–/nm2, over a surface area of ∼100 μm2. The two sets of irradiations were carried out first on the drain contact (C3) and then on the grounded source contact (C2). A final exposure of the MoS2 channel to the e-beam was performed as well.

Figure summarizes the obtained results. The Ids–Vds curves were measured at the end of each irradiation, ∼120 s after the blanking of the e-beam, to allow cooling down. Starting from the bottom (black) line in Figure a, representing the output curve of the unexposed device, the current increases with the e-beam exposures. We note two major discontinuities in the sequence of Ids–Vds curves, corresponding to the start of the two irradiations sets. These gaps are likely due to the uncontrolled exposure of the whole device during the selection of the drain (C3) and grounded source (C2) contact areas for the respective irradiation sets.

Figure 4

(a) Output characteristics at Vgs = 0 V of the transistor formed by contacts C2–C3 exposed to two sets of electron irradiations performed first on contact C3 and then on C2. (b) Rectification ratio and (c) maximum forward and reverse current, at Vds = ±5 V, as a function of the irradiation number. (d) Zero-bias Schottky barrier variation at the contacts C2 and C3 as a function of the irradiation number.

A different behavior of the forward with respect to the reverse current can be observed in Figure a, and a distinction of the effects of the irradiations on the drain (C3) and the grounded source (C2) can be made. Although the irradiation of the drain increases both the forward and the reverse currents, keeping the rectification ratio almost constant (see Figure b), the irradiation of the source augments the reverse current in a faster way, rendering the output curves more symmetric. Figure b shows that repeated irradiations of the drain contact (C3) do not change the rectification ratio (at Vds = ±5 V), while the irradiation of the grounded source contact (C2) dramatically decreases the rectification ratio. Figure c shows that the maximum reverse and forward currents, at Vds = ±5 V, have different variation rates when the irradiation is either on the drain or source. Noticeably, Figure c shows that the increase in both the reverse and forward currents is an exponential function of the fluence, which is proportional to and can be parametrized by the irradiation number.

As the shape and the current intensity of the output characteristics are related to the Schottky barrier heights at the contacts, the exponentially increasing current and the changing rectification ratio point to radiation-induced Schottky barrier lowering. The energy release in the metal contacts can modify the chemistry of the metal–MoS2 interface or create stress and defects that can lead to a lowering of the barrier and a consequent contact resistance reduction. We note that the reduction of contact resistance by chemical reactions between the metal contacts and MoS2 channel has been reported for the metal deposited under ultrahigh vacuum[47] and contact laser annealing.[48] A disordered, compositionally graded layer, composed of Mo and TiS species, forms on the surface of the MoS2 crystal following the deposition of Ti, and thermal annealing in the 100–600 °C temperature range can cause Ti diffusion inducing further chemical and structural changes at the Ti–MoS2 interface.[49,50] It is also possible that diffusion of Au atoms to the interface with MoS2 occurs under the energetic electron beam irradiation. Au does not react with MoS2 but reduces the contact resistance and therefore the Schottky barrier height.[51]

Similarly, tensile strain has been demonstrated to induce considerable Schottky and tunneling barrier lowering.[52]

A Schottky barrier of ∼0.2 eV is formed by several metals on MoS2 because of Fermi level pinning below the MoS2 conduction band.[53−55] Density functional theory calculations have indicated that the pinning at the metal–MoS2 interface is different from the well-known Bardeen pinning effect, metal-induced gap states, and defect/disorder-induced gap states, which are applicable to traditional metal–semiconductor junctions. At metal–MoS2 interfaces, the Fermi level is pinned either by a metal work function modification due to interface dipole formation arising from the charge redistribution or by the production of gap states mainly of Mo d-orbitals, characterized by the weakened intralayer S–Mo bonding because of the interface metal–S interaction.[56,57] The observed decrease in the Schottky barrier by e-beam irradiation, up to its complete disappearance, supports the occurrence of interface modifications that cause Fermi level depinning.

As the forward current at Vds < 0 V is limited by the Schottky barrier at the drain contact (C3), while the reverse current at Vds > 0 V is limited by the Schottky barrier at the grounded source contact C2 (which are the reverse-biased junctions for negative and positive Vds, respectively), the output curves of Figure a, which correspond always to reverse current, can be used to extract the behavior of the Schottky barriers as a function of the fluence (i.e., the e-beam irradiation number). Let us consider the thermionic current through a reverse-biased Schottky barrier[58,59]where φB and Is are the barrier height and the reverse saturation current at the n-th e-beam irradiation, S is the junction area, A2D* is the 2D Richardson constant, k is the Boltzmann constant, T is the temperature, n is the ideality factor, and Va is the negative voltage across the barrier that makes . Let us define I as the reverse saturation current before e-beam exposure, that is, associated to the maximum barrier height φB0. To avoid the effect of bias which can induce image-force barrier lowering,[60] both I and I0 are obtained by extrapolating the measured currents to zero bias. Then, eq can be used to evaluate the variation of the Schottky barrier, ΔφB = φB – φB0, as a function of the irradiation number

The zero-bias Schottky barrier variation, ΔφB, is shown in Figure d for both source (C2) and drain (C3) contacts. The overall reduction of both barriers is comparable to the expected initial barrier height based on Fermi level pinning, meaning that the long irradiation can completely remove the barriers. The plot indicates that the two barriers behave differently for the irradiation of C2 or C3. Although the beam irradiation of either contact results in a lowering of both Schottky barriers, the barrier decrease is faster for the irradiation of the grounded source. Besides, the Schottky barrier at the source contact is the most affected by the irradiation of the source.

To explain these results, we propose the model based on the energy band diagrams, shown in Figure . A negative (positive) voltage applied to the drain contact (C3) causes an upward (downward) shift of the energy bands in the drain region. Electron beam irradiation of the contact lowers the Schottky barrier and the relative built-in potential, as shown by the red dashed lines in Figure . The reduction of a Schottky barrier and of its associated built-in potential, at the irradiated contact, results also in the lowering of the unexposed barrier, which can experience a stronger potential drop because of the reduced contact resistance of the first contact. Figure a represents the situation in which the e-beam is focused on the biased drain contact (C3). At Vds < 0 V, the current is limited mainly by the drain contact barrier which is lowered by the successive irradiations, causing the exponential increase in maximum forward current. At Vds > 0 V, the current is limited by the un-irradiated source contact (C2) barrier, and its dependence on the irradiation cycle is caused by the lowering of the built-in potential at the drain (C3). As the barrier and built-in lowering are the same, the rectification ratio remains almost constant. For irradiation of the grounded source (C2, Figure b), the current increases because of a similar mechanism, with the difference that the drain contact barrier limits the current for Vds > 0 V to a lesser extent, having been already irradiation-lowered. Therefore, the reverse current increases faster with the repeated irradiation and the rectification ratio decreases.

Figure 5

Low-bias energy band diagrams (black) and their modification under electron irradiation (red) of C3 (a) and of C2 (b) contacts resulting in barrier lowering (φ̅B).

The effect of irradiation on the transfer characteristic of the transistor is shown in Figure and confirms the radiation-induced increase in channel current. Besides, Figure a shows that the e-beam, independent of onto which contact it is focused on, causes a left shift of the transfer curve. Such a shift corresponds to a decrease in threshold voltage, defined as the x-axis intercept of the linear fit of the transfer curve on the linear scale. The threshold voltage as a function of the irradiation is displayed in Figure b. Although the e-beam exposure of the contacts provokes a left shift (the transfer curves are taken at the end of the two irradiation sets on the drain (C3) and grounded source (C2)), further left shift of the threshold voltage is observed when two successive irradiations are performed in the channel region.

Figure 6

(a) FET transfer characteristics at Vds = −4 V before and after e-beam irradiations of contacts C3 and C2 and of the channel. (b) Left shift of the threshold voltage extrapolated from the transfer characteristics over the e-beam exposure.

The observed negative shift of the threshold voltage has been reported and discussed before.[27] It can be explained by the pile-up of positive charge in trap states of the SiO2 gate dielectric or at the SiO2–Si interface. The e-beam exposures produce electron–hole pairs in the SiO2 gate oxide and in the Si substrate: although mobile electrons are easily swept by the applied bias, the positive charges can be stored for long times.[27] The positive charge storage acts as an extra gate (similarly to the gating effect under light irradiation[61,62]) and enhances the n-type doping of the channel.

Indeed, Figure shows that there is a slight recovery of the threshold voltage after 12 h of annealing at room temperature. However, we highlight that, as demonstrated by Figure a, the maximum channel current, which is limited by the contact resistances, remains unchanged after annealing, demonstrating that the irradiation-induced improvement of the contacts is permanent.

To further confirm our model, we performed a Monte Carlo simulation to track the path of the electrons under the contacts and in the channel region (Figure a,b), using the CASINO software package.[63−65] We simulated a 10 keV beam with one million electrons and a radius beam of 10 nm. The cathodoluminescence spectrum (Figure c) shows that electrons lose their energy and are stopped (Figure d) mostly in the Ti/Au metal stack, while they reach and are absorbed in the Si substrate when the irradiation is on the channel. The high release of energy in the metal contacts, similarly to thermal annealing,[66,67] induces Ti–MoS2 reactions and creates contact with the reduced Schottky barrier and contact resistance. Conversely, when we directly irradiate the MoS2 channel, energy is prevalently adsorbed in the Si bulk and its effect manifests only through the positive charge traps generated in the SiO2 layer.

Figure 7

Monte Carlo simulation using CASINO v2 of e-beam irradiation of the device (a) contacts and (b) of the MoS2 channel. (c) Simulated cathodoluminescence intensity through the sample, with the e-beam focused onto the contacts and the flake. (d) Simulation of the electrons' penetration depth through the sample.

Conclusions

We investigated the effects of 10 keV electron beam irradiation of the Schottky metal contacts in MoS2-based FETs. Spectroscopic analysis by Raman and PL shows that the presence of metal contacts changes the properties of monolayer MoS2 with respect to strain and doping. The electrical measurements revealed that electron beam irradiation improves the device conductance, reduces the rectification of the output characteristic, and causes a left shift of the threshold voltage. To explain such a feature, we propose that the energy absorbed in the metal contacts induces atomic diffusion and interfacial reactions that lower the Schottky barrier at the contacts and improve the contact resistance. We corroborate our model by direct measurement of the Schottky barrier height variation and by simulation of the electron trajectories in the contact regions.