Manipulation of the electrical and memory properties of MoS2 field-effect transistors by highly charged ion irradiation

Field-effect transistors based on molybdenum disulfide (MoS2) exhibit a hysteresis in their transfer characteristics, which can be utilized to realize 2D memory devices. This hysteresis has been attributed to charge trapping due to adsorbates, or defects either in the MoS2 lattice or in the underlying substrate. We fabricated MoS2 field-effect transistors on SiO2/Si substrates, irradiated these devices with Xe30+ ions at a kinetic energy of 180 keV to deliberately introduce defects and studied the resulting changes of their electrical and hysteretic properties. We find clear influences of the irradiation: while the charge carrier mobility decreases linearly with increasing ion fluence (up to only 20% of its initial value) the conductivity actually increases again after an initial drop of around two orders of magnitude. We also find a significantly reduced n-doping (≈1012 cm−2) and a well-developed hysteresis after the irradiation. The hysteresis height increases with increasing ion fluence and enables us to characterize the irradiated MoS2 field-effect transistor as a memory device with remarkably longer relaxation times (≈ minutes) compared to previous works.


Introduction
Molybdenum disulde (MoS 2 ), a member of the family of the socalled transition metal dichalcogenides (TMDCs), is one of the most studied two-dimensional (2D) materials right aer graphene.While in its bulk (3D) form it has an indirect bandgap of around 1.2 eV, 1 it develops a direct bandgap of 1.8-1.9eV (ref.2)  when reduced to its covalently bonded S-Mo-S monolayer structure.This bandgap allows the utilization as typical building blocks for modern electronics like e.g., eld-effect transistors (FETs) based on atomically thin 2D materials. 3ecause of that, it was quickly realized that monolayer MoS 2 might be an excellent candidate for future electronic and optoelectronic applications, especially when large scale production techniques such as chemical vapor deposition (CVD) are used.The on-going reduction of device dimensions poses critical problems for traditional semiconductor devices e.g. based on silicon, as the carrier mobility degrades rapidly for channel thicknesses reaching the scale of only a few nm, [4][5][6] which is not the case for MoS 2 and other 2D materials. 7Indeed it was demonstrated that MoS 2 FETs with a small gate length (#10 nm) a simultaneously reasonable mobility and high on-currents can be achieved. 8,9Open challenges for 2D-TMDC FETs to date are Schottky barriers at the metal-TMDC interface, 10,11 nonsufficient doping techniques 12 and structural defects either in the channel material or in the underlying oxide. 13These structural defects can trap charges and act as scattering centres, modifying the electrical properties of the devices.5][16][17][18] The trapped charges inuence the charge carriers in the 2D material channel and shi the transfer characteristics depending on the gate voltage sweep direction.][21] Defects can be articially and controllably introduced into 2D materials by particle irradiation, e.g. using electrons or ions as projectiles.6][27][28] These defects have been proposed or even utilized for a broad variety of applications, e.g.ultraltration, 29,30 DNA sequencing 31,32 or catalysis. 33By ne tuning the energy of the ions, irradiation can even be used to clean the surface of 2D materials from process residues stemming from transfer and a Faculty of Physics and CENIDE, University of Duisburg-Essen, Lotharstraße 1, D-47057 Duisburg, Germany.E-mail: stephan.sleziona@uni-due.delithography steps, without damaging the underlying 2D material too much. 34,35The irradiation of MoS 2 with electrons or ions with a moderate kinetic energy in the keV range can lead to single or double vacancy defects in the TMDC lattice, [36][37][38] like e.g., a sulphur vacancy V S shown schematically in Fig. 1(f).In this work, we use highly charged ions (HCI) with a charge state q = 30+ at a kinetic energy E kin = 180 keV to deliberately introduce defects.We chose HCIs since their potential energy (i.e.their charge state) and kinetic energy can be tuned independently and by that control the defect creation in our devices.In contrast to singly charged ions with keV kinetic energy, the defect creation mechanism by HCIs in 2D materials is still under discussion. 24For free-standing MoS 2 , the formation of nm-sized pores was observed aer irradiation with HCIs with the same kinetic energy used in this work.The size of the pores depends on the charge state of the ions 39 and no vacancy-type defects were reported.Since there are, to the best of our knowledge, no corresponding imaging experiments for substrate-supported MoS 2 proving the contrary, it is feasible to assume that the irradiation with HCIs also creates these nmsized holes in substrate-supported MoS 2 .Recent time-of-ight mass spectrometry experiments show that the kinetic energy of the HCIs has to be taken into account to account for ioninduced damage of substrate supported MoS 2 and that the type of the substrate is important. 40With the parameters used in this work we thus expect pores in the MoS 2 and a substantial amount of defects in the underlying substrate.
Although there has been done some prior work with particle irradiation of 2D FETs [41][42][43][44] these articles focus mostly only on standard electrical performance like conductivity, mobility, and irradiation hardness [45][46][47] but not on the hysteretic properties of the irradiated devices.Also, compared to previous studies, where MoS 2 devices were irradiated with singly charged ions (like e.g.Ar + , N + , He + ), [48][49][50] in this work, we take a new approach and use HCI with signicantly lower irradiation uences to realize strong modications of the devices properties.Furthermore, we pay specic attention to the manipulation of the hysteretic properties of 2D MoS 2 FETs by ion irradiation.
To this end we fabricate CVD-grown single layer MoS 2 FETs on a Si/SiO 2 substrate via photolithography and characterize their electrical properties by measuring their output and transfer characteristics.Aer this initial characterization, the devices are irradiated with HCI to deliberately introduce defects.We show that the irradiation leads to distinct modications of the electrical properties and especially causes a strongly reduced n-doping of the devices.Most importantly, we demonstrate that the irradiation leads to the opening of a hysteresis, most likely caused by additional defects in the underlying oxide.The height of the hysteresis scales with the introduced ion uence, enabling the realisation of a memory device.

Results and discussion
We begin by describing our devices and the general course of our work.In Fig. 1(a) an optical microscopy image of one of the MoS 2 -FETs used in this work is shown.The four metal contacts are labeled with numbers (1-4).For most devices, we employed The inset shows the PL spectrum measured at the same spatial position, displaying one strong peak at an energy of 1.83 eV attributed to the A exciton and a smaller peak at an energy of 1.98 eV attributed to the B exciton.Both observations clearly proof that our samples are indeed monolayers of MoS 2 . 2,51Fig. 1(c) outlines the general course of our experiment: the FET structure is used for standard electrical characterization of the MoS 2 , in particular the output (I DS (V DS )) and transfer (I DS (V GS )) characteristics.Aer this initial characterization, the devices are irradiated with Xe 30+ ions at a kinetic energy of 180 keV to deliberately introduce defects into the devices.Aerwards, the devices are again characterized to observe the inuence of the introduced defects on their electrical behavior.We irradiated four different devices with four different uences, namely 100 ions/mm 2 , 200 ions/mm 2 , 400 ions/mm 2 and 1600 ions/mm 2 .
The output and transfer characteristics of one of our device before the irradiation is shown exemplary in Fig. 1(d) and (e), respectively.3][54][55][56][57] The transfer characteristics in Fig. 1(e) reveals the behavior of a normally-on n-type transistor with very strong n-doping.9][60][61] Consequently, the off-state of the transistor can not be reached in the applied gate voltage range, so the ratio between the minimum and maximum current is only 10 3 .We note that the devices in this work exhibit small differences in their overall electrical behavior, which is a typical observation for 2D devices in literature and can be explained by the contact resistance and Fermilevel pinning being delicately dependent on microscopic details in the contact formation at the metal-TMDC interface. 55,62evertheless, all our devices have in common that they have a low Schottky barrier and exhibit strong n-doping.
In Fig. 2(a) and (b) the output and transfer characteristics of the MoS 2 device before (blue) and aer irradiation (red) with a uence of 400 ions/mm 2 are shown.Let us discuss the output characteristic in Fig. 2(a) rst.It displays a reduction of the current by around two orders of magnitude aer the irradiation.Besides that, the Schottky type behavior is not modied.This nding is within our expectations: particle irradiation of contacts can modify the metal-2D material interface in such devices and lead to reduced contact resistance and Schottky barriers. 63,64In our work however, the kinetic energy of the ions is not high enough to penetrate through the metal contacts, as supported by SRIM calculations that demonstrate that all ion collision events occur only in the metal and not at the interface (see Fig. S1a †).This means that all the energy of the ions is deposited into the metal and not at the metal-TMDC interface, from which follows, that the interface can not be modied by the irradiation.
The strong reduction of I DS indicates a signicant increase in scattering centers.Regardless of the specic defect type created by the irradiation (vacancies, holes, strained/chemically modi-ed lattice), all of them will pose scattering centres for the charge carriers and thus reduce the conductivity of our device.We will now discuss the transfer characteristics in Fig. 2(b) which shows striking differences between the measurement before (blue) and aer (red) the irradiation.As discussed for Fig. 1(e), the transfer characteristics before the irradiation displays a strong n-doping behavior.When sweeping the gate voltage between −30 V and +30 V and back, no hysteresis effect is observed.This might be caused by the strong n-doping of our devices, since the saturation region of the transfer characteristics usually does not show a signicant hysteretic behavior when measured in high vacuum conditions. 14,16,21Aer the irradiation, the most striking difference is the appearance of the off-state region and a hysteresis in the transfer characteristics.As mentioned before, this hysteresis is generally attributed to either defects in the MoS 2 lattice, at the MoS 2 /oxide interface or in the oxide itself. 16,21,65As the occurrence of the hysteresis is clearly related to the irradiation it seems straightforward to claim that it is a result of additional defects introduced by the irradiation.The right-shi of the transfer aer the forward gate voltage sweep, which leads to the clockwise hysteresis, is indicative of negative charge trapping.A further discussion of the properties of the hysteresis and which type of defect is likely the reason for its occurrence will be conducted later on.Additionally, the transfer curve shows, that the n-doping of the device is strongly reduced aer the irradiation, since the threshold voltage (V th ) shis towards positive gate voltages.It is now even possible to reach the off-state of the transistor in the applied gate voltage range and a high on-off ratio of nearly 6 orders of magnitude can be derived.
This reduction of n-type doping may be attributed to several possible causes.Oxygen molecules capture electrons in MoS 2 (ref.0][71] Although the amount of adsorbed molecules should be reduced under vacuum conditions, ion irradiation can lead to the formation of chemically adsorbed MoO 3 at the defect edges, which would be very resistant to vacuum assisted desorption 33 and could also explain the observed reduction in ndoping.Since the HCIs will not only deposit their energy in the MoS 2 monolayer, but also in the underlying SiO 2 substrate, defects in the oxide could also play a role in p-doping the device.In fact, electron-trapping defect states for MoS 2 on a SiO 2 substrate have already been reported 65,72,73 and would lead to an effective p-doping of our devices by an additional gating effect (see schematic representation of oxide traps in Fig. 1(f)).The trapped negative charges in the oxide inuence the electric eld generated by the applied gate voltage and thereby shiing the transfer curve towards positive gate voltages.
In the following, we will compare the results of the electrical characterization before and aer the irradiation of our devices in dependence of the irradiation uence.For this, we will address the conductivity, mobility, and charge carrier density, starting with the conductivity.In Fig. 3(a) we display the remaining conductivity (s/s 0 ) using the output characteristics before and aer irradiation.That is, we normalized the conductivity aer irradiation to its value before irradiation to compare the different devices with each other.Interestingly, the devices show an increasing remaining conductivity with increasing ion uence, aer the initial drop in conductivity already discussed above (see Fig. 2(a)).This unintuitive increase of electrical conductivity with increasing ion uence was rst reported by Fox et al. 74 for the irradiation of bilayer MoS 2 with He-ions.This peculiar behavior could be connected to an irradiation induced activation of an additional transport mechanism.For 2D TMDCs it has already been shown, that an increase in chalcogen vacancies or interface defects can lead to hopping transport via localized states and, as a consequence, lead to an increasing conductivity. 58,75Nevertheless, temperature-dependent conductivity measurements would be needed to conrm a change in the transport mechanism due to the irradiation. 76,77ote, that the FET irradiated with 200 ions/mm 2 is the only exception to the otherwise linear behavior.For this device we used the contacts 1 and 3 as source and drain contact with one contact (2) in between on the MoS 2 channel (see Fig. 1(a)).By Fermi-level pinning this contact can modify the electrical behavior of the 2D material channel, which is mirrored in all electrical characterizations (Fig. 3 the devices used in this work, which is in the range typically measured for such devices. 14,78,79Aer the irradiation, the mobility was examined again and then normalized to the value measured before the irradiation (m/m 0 ).The results of this analysis are shown in Fig. 3(b) displaying a monotonous decrease of the mobility with increasing ion uence.The defects introduced by the HCI irradiation either in the MoS 2 or in the oxide can lead to increased Coulomb scattering for the charge carriers in our devices by charge trapping.This leads to shorter scattering times and therefore an overall reduced mobility, despite the enhanced remaining conductivity.While nm-sized holes created in the MoS 2 lattice by HCI irradiation would certainly act as scattering centres for the charge carriers in the device, the SRIM calculations in Fig. S1b † demonstrate, that for a MoS 2 monolayer on top of a SiO 2 substrate, most of the collisions happen in the SiO 2 when this system is irradiated with Xe + ions at a kinetic energy of 180 keV.Around 90% of the collisions within the rst few nm happen in the oxide, pointing towards oxide defects playing the dominant role for the reduction of the mobility due to the irradiation.
Lastly, we discuss the change in charge carrier density.We quantitatively evaluate the change in doping for our devices using the transfer characteristics of each device before and aer irradiation.Because of the initially high n-doping of our devices we use V DS to t the transfer curve and extract the value for V th (see green-streaked line in Fig. 1(e)). 16rom this we calculated the change in charge carrier concentration with Dn ¼ C ox Â DV th q : As can be seen in Fig. 3(c), the irradiation leads to less n-doping (i.e.effective p-doping) in our devices in the order of 10 12 cm −2 and increases with increasing ion uence up to 3.0 × 10 −12 cm −2 without any indication of a saturation behavior.
To further conrm this nding, we performed Raman spectroscopy of a CVD-grown MoS 2 sample between the different irradiation steps (see Fig. 3(e)).The qualitative behavior of the Raman spectra, a constant position for the E ′ mode, while the A 0 1g mode shis to higher wavenumbers, points to decreasing n-doping, 80 as it was also derived from the transfer characteristics.We also nd a stronger reduction of the n-doping with increasing ion uence.For a quantitative analysis we used the procedure from ref. 81.The result of this is shown in Fig. 3(d).Obviously, both methods, electrical characterization and evaluation of Raman spectra, yield the same result, a linearly decreasing n-doping with increasing irradiation uence, supporting our previous ndings.
Possible mechanisms for a decrease of n-doping, like the adsorption of oxygen molecules at defect sites or electron capturing defects in the oxide, were already discussed above.With increasing ion uence, the density of these defects increases and we therefore infer, that the irradiation does create electron capturing defects.This is also conrmed by the fact, that the absolute values extracted from the FET transfer characteristics are somewhat lower than those extracted from the Raman data, as the FETs were measured under high vacuum conditions (defects in the MoS 2 are not saturated by oxygen and thus do not contribute to the reduction), while the Raman spectra were collected in ambient conditions (both types of defects contribute).
We note, that the change in charge carrier density as derived by both methods is z10 12 -10 13 cm −2 and thus 1-2 orders of magnitude higher than the irradiation uence.As already discussed above, defect sites will facilitate p-doping.The defects we induce here are not point-like, but have a spatial extension on the order of nm.We therefore expect a high number of dangling bonds at the defect edges, which are prone to the adsorption or even bonding of gas molecules, explaining the high efficiency in terms of p-doping per ion.For the other possible cause of the observed doping effect, electron trapping defects in the SiO 2 substrate, there will also several defects per ion be created (see discussion below).This is also consistent with the high efficiency in terms of doping per ion.Therefore, the observed doping effect can be explained satisfactory by both possibilities: either defects in the MoS 2 channel or in the underlying oxide.Since the measurements were performed under high vacuum conditions, defects in the oxide seem more likely.
Finally, we want to address the manipulation of the hysteresis' properties of the MoS 2 devices by ion irradiation.As it was already shown in Fig. 2(b) aer the irradiation, a hysteresis can be observed in the devices transfer characteristics, which was absent before the irradiation.3][84] From an application point of view, such a device may be used as a nonvolatile storage element.Favorable properties are two stable and clearly distinguishable memory states, the so-called memory window, a sufficiently high hysteresis to prevent unwanted switching and long time-constants when switched by erase/write voltage pulses.
For our analysis, we rst evaluate the memory window, i.e., the height of the hysteresis (i.e.DI DS at the same V GS ) for the different irradiation uences.The results found for our different devices are summarized in Fig. 4(a).The hysteresis' height increases linearly with increasing ion uence reaching up to around two orders of magnitude for the device irradiated with the highest uence.We note, that even for the smallest irradiation uence of only 100 ions/mm 2 there is already a fully developed hysteresis observable even though it was nearly absent prior to the irradiation (see Fig. S2 †).These results clearly demonstrate, that ion induced defects are responsible for the observed hysteresis.
To study the switching behavior of our device we applied ±30 V gate pulses and recorded the transient behavior of the device irradiated with a uence of 200 ions/mm 2 .The result is shown in Fig. 4(b).In this device we can reach two distinct memory states at a gate voltage of V GS = 0 V with a current separation of around one order of magnitude.The current separation prevails and is stable for the entire observed pulse interval (around 30 min), which is comparable with the rentention times observed in few-layer MoS 2 charge-trapping memory devices. 85,86The transients observed in Fig. 4(b) can be interpreted in terms of charge trapping/detrapping mechanisms.The time constants have been evaluated for the read (s 1 ) and the erase (s 2 ) conguration by tting an exponential function f ðtÞ ¼ c Â e À t s þ A 0 to the data.Compared to previous studies, the time constants s 1 = 1600 s and s 2 = 90 s, respectively, are rather long. 15This nding points towards oxide defects playing a major role because charges trapped in deep oxide defects have considerably longer relaxation times than e.g.traps in the 2D material itself or at the 2D material/oxide interface. 21,87,88The time constant for the positive gate pulse (s 1 ) is much longer than the time constant for negative gate pulse (s 2 ).This is also another indicator of negatively charged oxide defects being the main contribution to the hysteresis in our work.These defects lie in the vicinity of the conduction band of MoS 2 (ref.65) and would therefore be charged when applying positive gate voltages, but would not be charged for the negative gate pulse.Additionally, we show in Fig. 4(c) that the observed hysteresis is independent of the applied V DS within the range of 1 V-5 V.This is in contrast to recent observations for black phosphorous FETs, where the dependence of the hysteresis on V DS was ascribed to defects in the 2D material channel itself. 89Considering our irradiation conditions, signicant defect generation in the substrate is to be expected.
As already discussed above, the SRIM calculations in Fig. S1(b) † demonstrate, that most of the collisions caused by the ion irradiation at this kinetic energy occur in the oxide, since the MoS 2 channel is atomically thin.We therefore conclude, that while the doping effect may be due to both, ioninduced defects in the MoS 2 and the substrate, the hysteresis observed in our devices is caused by negatively charged defects in the underlying oxide induced by the HCI irradiation.With Fig. 4(d) we prove that the separation of the two memory states is stable for several memory cycles.

Conclusion
We have investigated the manipulation of the electrical properties of MoS 2 FETs by the irradiation with HCIs.While we found a decreasing mobility in the devices with increasing ion uence, the conductivity aer an initial drop actually increases with higher defect density suggesting that at hopping-like transport takes over with increasing defect density.This further proves, that the devices are rather resistant to ion irradiation, an important factor for the possible use in high radiation environments like e.g.space applications.
Additionally, we have shown that HCI irradiation can be used for deliberate and controlled manipulation of the doping density of MoS 2 devices.In our case we found a strong decrease in n-doping.Most notably, the irradiation leads to a hysteresis in the transfer characteristics of the device which we successfully exploited for a non-volatile memory device with two stable memory states and a long retention time.We demonstrated that the memory window can be tuned by the irradiation uence, opening up new possibilities to boost the performance of MoS 2 based memory devices.We also believe, that this procedure can be applied to other similar 2D devices since the hysteresis most likely originates from defects induced into the underlying oxide.
Since the strong modications of the devices' properties already happen at comparatively low uences, we observe no notable damage to the active channel and the surrounding substrate in atomic force microscopy (AFM) images aer the irradiation (see Fig. S3 †).In particular, we nd no evidence of hydrocarbon deposition, which would hinder further processing steps of the devices aer the irradiation.This is a big advantage to previous works, where MoS 2 FETs have been irradiated with He-ions. 49,50urther, we like to point out that Chen et al. 48succeeded in realizing a MoS 2 based non-volatile memory device by seeding defects in the oxide via irradiation with Ar + and N + ions before the MoS 2 was deposited on the substrate, while here, we were able to realize a memory device by irradiating the MoS 2 aer device fabrication.Our approach thus opens up the possibility to ne-tune the electrical and memory properties of devices by choosing the appropriate ion uence.This, together with the independent control of potential and kinetic energy, will allow to precisely manipulate the electrical properties of the irradiated devices in future experiments.

Experimental procedure
MoS 2 akes were grown via chemical vapor deposition (CVD) on a highly doped p-type Si substrate (resistivity 0.001-0.005U cm) covered by 285 nm thermal SiO 2 .At rst a 1% sodium cholate solution was spin coated onto the substrate working as a seeding promoter.The growth was performed in a three-zone (ThermConcept ROK 70/750/12-3z) tube furnace.By 10 min purging with 500 sccm Ar gas (99.9%) ow, the O 2 content of the furnace was minimized.40 mg of S powder (99.98% Sigma Aldrich) were placed in the upstream heating zone at 150 °C.MoO 3 , used as the source for molybdenum, was obtained from a aqueous ammonium heptamolybdate (AHM) solution (ratio 1 : 1) initially annealed at 300 °C for 24 min under ambient conditions and positioned in the next downstream zone at 750 °C.During the whole process 500 sccm of Ar gas ows through the quartz tube.The growth process lasted 30 min and was followed by a rapid cooling.At a temperature of around 100 °C the samples were retrieved from the CVD furnace.The resulting MoS 2 akes are mostly single layers with triangular shape.
For device fabrication the freshly grown samples were investigated via optical microscopy to select suitable akes for photolithography processing.Aer the standard photolithography process 10 nm of Cr and 100 nm of Au were deposited by electron-beam (Cr) and thermal evaporation (Au) at a process pressure of 1 × 10 −5 mbar to electrically contact the MoS 2 akes.
Electrical characterization of the devices was performed with a cryogenic probe station with pressure control and four metallic nanoprobes, which are connected to a Keithley 4200 SCS.The metallic sample plate was used to apply the backgate voltage to the Si substrate.All electrical measurements in this work are performed under a vacuum of 1 × 10 −4 mbar and the samples were le there for at least 12 hours before starting the measurements.
To irradiate the samples, highly charged xenon ions were generated in an electron beam ion source (EBIS) commercially available from Dreebit GmbH, Germany. 90A kinetic energy of 180 keV (1.4 keV amu −1 ) and an ion charge state of 30+ with a potential energy of 15.4 keV (0.1 keV amu −1 ) was selected via a sector magnet and used for all experiments.Ion irradiation was performed under ultra-high vacuum conditions (pressure about 4 × 10 −9 mbar), and each sample was irradiated with a total uence between 100 and 1600 ions/mm 2 .During the irradiation, the entire devices, including their electrical contacts, are impacted by ions due to the spatial extent (around 1 mm 2 ) of the ion beam.

Paper
Nanoscale Advances

Fig. 1
Fig. 1 (a) Optical microscopy image of monolayer MoS 2 on SiO 2 contacted with four Cr/Au leads.(b) Raman spectrum of monolayer MoS 2 with corresponding photoluminescence spectrum as inset.(c) Schematic of the device and measurement setup used in this work.In (f) we show a schematic representation of the different possible defect types that might be introduced in our devices due to the irradiation.Large and complex defects in the MoS 2 lattice are represented by D, while V S denotes sulphur vacancies.The white lines in the oxide are a schematic representation of electron trap states.(d) Output and (e) transfer characteristics of one of our devices before the irradiation.

Fig. 2
Fig. 2 (a) Output and (b) transfer characteristics of a MoS 2 FET before (blue) and after (red) irradiation with Xe 30+ ions with a fluence of 400 ions/mm 2 .
Fig. 3 (a) Remaining conductivity of the MoS 2 FETs after the irradiation with different fluences normalized to the respective value of the device before the irradiation (s/s 0 ).(b) Effective mobility of the MoS 2 FETs after the irradiation with different fluences normalized to the respective value of the device before the irradiation (m/m 0 ).(c) Change in charge carrier concentration of the MoS 2 FETs for the different fluences calculated by the shift of V th .Negative values indicate a decrease of the electron density, meaning increased p-doping.(d) Change in charge carrier concentration for the different fluences calculated from the Raman spectra in (e) which are taken from a MoS 2 sample at the different irradiation fluences.

Fig. 4
Fig. 4 (a) Hysteresis height (maximum DI DS at the same V GS , see also inset in (b)) evaluated from the transfer characteristics of each device after the irradiation with different fluences.(b) Transient behavior of the device irradiated with a fluence of 200 ions/mm 2 for a single set-read-resetread cycle.The dashed orange curves correspond to fits of exponential decays from which the trapping times s 1 and s 2 are evaluated.Inset shows the corresponding hysteresis curve with DI DS highlighted.(c) Transfer characteristics of the device from (b) for different values of V DS .(d) Several set-read-reset-read cycles of the same device used in (b).