As scientific and technological advancements proceed, humanity's ambition to venture into the outer space expand continually: range from launching spacecraft to constructing space stations, and even extends to the exploration of distant celestial stars such as Moon or Mars. Whether traveling to Moon, Mars or beyond, the mission of exploring outer space has stringent and severe requirements for integrated circuits (ICs)^{[1, 2]}. On the one hand, the complex electronic integrated circuits used in space need to be energy efficient, which are capable of completing the long-time space exploration with limited energy reserves. On the other hand, the complex electronic circuits used in space need to be resistant to cosmic radiation and high-energy solar, which ensures the reliability of ICs in deep space exploration^{[3, 4]}.

To satisfy these requirements, a variety of advanced low-dimensional materials and devices are currently be explored, such as MoS₂^{[5, 6]}, carbon nanotube (CNT)^[7−11], black phosphorus^[12] and so on. Among of them, MoS₂ emerges as a prime contender for the next generation of energy-efficient ICs^[13−15]. Due to its high carrier mobility and outstanding scaling behavior characteristics, the scaled 10 nm-channel-length MoS₂ field-effect transistors (FETs) exhibit higher intrinsic performance and lower power consumption than Si-based FETs with similar supply voltage (V_ds) and gate length^[16]. In addition, MoS₂ exhibits ultra-thin body and nanoscale sensitivity volume, which give them strong radiation tolerance. Very recently, significant progress has been made in the field of radiation-hardened MoS₂ ICs^{[5, 17−19]}, but these studies have primarily concentrated on the total ionizing dose (TID) effects, which is not sufficient to characterize the radiation tolerance of an IC when exposed to the actual conditions of the space radiation environment.

Compared to the TID effect, the displacement damage (DD) effect is equally crucial in practical applications. When high energy particles strike the MoS₂ FETs, non-ionizing processes result in energy dissipation and the lattice defects, including interstitial atoms and vacancies within the material^{[20, 21]}, which can influence the electrical characteristics of the devices. Although various studies have confirmed the remarkable tolerance of MoS₂ FETs to swift heavy ions^[22−24], all of these works failed to address the impact of MoS₂ thickness on its DD effects, which is directly related to the radiation absorption cross-section of MoS₂.

In this study, the DD degradation mechanism in back-gated MoS₂ FETs is systematically investigated using 30 keV proton irradiation with various fluences from 10¹² to 10¹⁴ p/cm². Both materials and devices showed notable DD radiation tolerance, demonstrated by stability in Raman spectra and slight degradation in electrical characterization. In addition, SRIM software toolkit was employed to simulate the DD effects of 30 keV proton irradiation on the exposed channel region. By combining simulation results with the electrical measurements, we proved that the higher tolerance to proton irradiation is attributed to the nanoscale cross-section in monolayer MoS₂. These MoS₂-based FETs exhibited high DD radiation tolerance up to 1.56 × 10¹³ MeV/g (10¹¹ MeV/g for outer space exploration), which provides novel perspectives on the radiation-hardening technology of MoS₂-based FETs.

Monolayer MoS₂ was grown on a two-inch c-sapphire substrate using a three-zone temperature-controlled chemical vapor deposition (CVD) system. Elemental sulfur (S) powder was positioned in a silica crucible and heated to 170 °C using a separate heating mantle, with Ar at a flow rate of 100 sccm facilitating its transport. Molybdenum (Mo) sheets, serving as the molybdenum sources, were introduced into the growth chamber in a separate stream, which consisted of a mixture of 5−10 sccm of O₂ (to promote the oxidation of Mo) and 50 sccm of Ar, alongside the S vapor. Mo was heated at 650 °C and the growth temperature was 950 °C for the monolayer MoS₂ growth.

The MoS₂ layer was transferred using a wet process assisted by polymer. Firstly, Poly (methyl methacrylate) (PMMA A4, 950K) was applied onto the MoS₂/sapphire substrate by spinning at 2000 rpm and baked at 150 °C to enhance the adhesion between the PMMA and MoS₂. Subsequently, the thermal release tape (TRT) was affixed onto PMMA and then immersed inside a 3 mol/L NaOH solution maintained at 40 °C. The TRT/PMMA/MoS₂ stack separated from the sapphire substrate was transferred onto the target substrate and subjected to a baking process at 120 °C for 10 min to ensure the TRT was successfully removed. Finally, the PMMA was removed by acetone for 30 min.

For FET fabrication, the Si substrate covered by 285 nm SiO₂, with the transferred MoS₂ layer above, was coated with bilayer photoresist (MMA EL9 and PMMA A4, 4000 rpm for 45 s) and baked at 150 °C for 3 and 7 min, respectively. Next, the channel regions were defined using electron-beam lithography (EBL) followed by etching processes utilizing SF₆ and O₂ plasma. Finally, EBL was again employed to pattern the source and drain, followed by e-beam evaporation deposition of 20 nm Sb and 30 nm Au, and finally the lift-off process. Fig. 1(d) shows the process flow of device fabrication. All electrical measurements were conducted using a Keysight B1500 semiconductor parameter analyzer within a vacuum probe station environment.

The 30 keV proton irradiation was carried out at low energy proton irradiator. The beam diameter was 20 to 150 mm with a high uniformity (fluence variation ≤(±5%)). The average flux was 8 × 10⁸ p/cm²·s. The irradiation fluences were set to be 10¹², 10¹³ and 10¹⁴ p/cm², respectively.

The simulated proton fluence in the SRIM simulation was set to 10¹², 10¹³, and 10¹⁴ p/cm². In the simulated cross-section, the Au (30 nm, 19.32 g/cm³) and Sb (20 nm, 6.70 g/cm³) metal electrode, MoS₂ (1.20 g/cm³) channel, and SiO₂ (285 nm, 2.15 g/cm³) and /Si (725 μm, 2.33 g/cm³) substrate were included. The thickness of each layer of MoS₂ is defined as 0.65 nm. The default material parameters of antimony (Sb), Au, and SiO₂ were applied to the corresponding materials. The DD caused by non-ionization energy loss (NIEL) is the main factor leading to the failure of devices in space radiation environment^{[25, 26]}. The NIEL is defined as:

1$\text{NIEL}=E_{\text{vac}}\frac{V\cdot \text{100}}{\rho }\mathrm{.}$

Here, $E_{\text{vac}}$ is the displacement threshold energy, and V is the number of vacancies per proton per Å, which can be extracted from TRIM simulation. The damage produced by different irradiation environments is defined using displacement per atom (dpa), which is the amount of displaced atoms divided by the total number of atoms^[27]. This is a method for measuring the degree of radiation damage and dpa is defined as:

2$\text{dpa}=\left(\frac{\text{vacancies}}{\text{p}\times \text{Å}}\right)\times \left(\frac{{10}^8\left(\frac{\text{Å}}{{\text{cm}}^2}\right)\times \text{fluence}\left(\frac{\text{p}}{{\text{cm}}^2}\right)}{\text{Atomdensity}\left(\frac{\text{atoms}}{{\text{cm}}^3}\right)}\right)=(\frac{\#\text{of vacancies}}{\text{atom}})\mathrm{.}$

Here, $\frac{\text{vacancies}}{\text{p}\times \text{Å}}$ is the data for TRIM simulation results, meaning the number of vacancies generated per incident ion within a unit target distance. Atom density $\left(\frac{\text{atoms}}{{\text{cm}}^{\text{3}}}\right)$ is the atomic density of the target material, and fluence $\left(\frac{\text{p}}{{\text{cm}}^{\text{2}}}\right)$ is the incident particle dose in the experience.

The growth of MoS₂ was conducted using a CVD system, as illustrated in Fig. 1(a). The MoS₂ was then transferred onto a 285-nm SiO₂/highly doped Si substrate employing a wet transfer process. In the AFM measurement, the height of monolayer MoS₂ on SiO₂/Si substrate is about 0.65 nm (Fig. S1). More detailed information are available in the Experimental Section. To characterize the CVD-grown MoS₂, Raman and photoluminescence (PL) spectroscopy were employed, with all tests conducted using a 488 nm laser. Fig. 1(b) presents the Raman spectra line scan, which features two peaks that correspond to the out-of-plane vibrational mode of the S atoms (${\text{A}}_{\text{1g}}$ mode at approximately 399.46 cm⁻¹) and in-plane vibrational mode involving both Mo and S atoms (${\text{E}}_{\text{2g}}^{\text{1}}$ mode at approximately 381.94 cm⁻¹) within the MoS₂ structure. The average difference of 17.53 ± 0.50 cm⁻¹ between these two peaks effectively indicates the presence of a monolayer^[28]. The PL spectrum, shown in Fig. 1(c), displays an averaged peak at 1.85 eV, which corresponds to the characteristic direct-bandgap emission for monolayer MoS₂^[29].

Based on the CVD-grown monolayer MoS₂, the back-gate MoS₂-based FETs were fabricated. The process flow plans are provided in Fig. 1(d). Semi-metallic Sb contacts were used to overcome the inherent van der Waals gap and reduce contact resistance^[13]. All FETs were designed with the same width (W = 4 µm) and length (L = 3 µm), as shown in the inset of Fig. 1(e). Fig. 1(e) plots the transfer characteristics of 20 MoS₂ FETs at V_ds = 3 V, with a highlighted curve representing the characteristics of a typical FET. All devices exhibited n-type transport behavior with an on/off ratio (I_on/I_off) of approximately 10⁵, an off-state current (I_off) of 30 pA, and ~10% subthreshold swing (SS) variation. The output characteristics of back-gate MoS₂ FETs are shown in Fig. 1(f) presenting both Ohmic contacts (a linear I−V at low bias) and an on-state current (I_on) of approximately 4 μA (at V_ds = 3 V). These properties are attributed to the use of semi-metallic Sb contacts, which can overcome the inherent van der Waals gap and reduce contact resistance^[13].

Following the growth of the MoS₂ and the fabrication of the FETs, the radiation tolerance of DD effect was measured. The MoS₂ FETs were subsequently irradiated by protons with the energy of 30 keV at the National Space Science Center, Chinese Academy of Sciences for DD tests. The quality of the MoS₂ in the channel region was evaluated before and after irradiation using Raman spectroscopy. Fig. 2(a) illustrates the variation of the Raman spectra corresponding to different levels of proton fluence, normalized by the maximum peak intensity of the monolayer MoS₂. A minor shift in position can be observed for the two peaks, with the ${\text{E}}_{\text{2g}}^{\text{1}}$ peak at 0.95 cm⁻¹ and the ${\text{A}}_{\text{1g}}$ peak at 1.98 cm⁻¹. The changes in peak positions and their difference with the irradiation fluence were further compared (Fig. 2(b)). It was found that both ${\text{E}}_{\text{2g}}^{\text{1}}$ and ${\text{A}}_{\text{1g}}$ peaks displayed blueshift with a low fluence (10¹² p/cm²). As the proton fluence increased up to 10¹³ p/cm², the shift of both peaks switched their direction, exhibiting a redshift. As the irradiation dose increases, the distance between the peaks expands.

This phenomenon has been further verified by Raman mapping measurements. Figs. 2(c) and 2(d), respectively, show the spatial mappings of peak-to-peak separation and peak intensity ratio between the ${\text{E}}_{\text{2g}}^{\text{1}}$ and ${\text{A}}_{\text{1g}}$ active modes, measured at 30 distinct points on the MoS₂ sample. The ${\text{E}}_{\text{2g}}^{\text{1}}$ and ${\text{A}}_{\text{1g}}$ peaks have an average peak distance of 17.53 cm⁻¹ and an average peak intensity ratio of 1.10 before irradiation. The Raman spatial mappings of peak-to-peak distance and peak intensity ratio between the ${\text{E}}_{\text{2g}}^{\text{1}}$ and ${\text{A}}_{\text{1g}}$ active modes with fluence conditions of 10¹⁴ p/cm² are shown in Figs. 2(e) and 2(f). A blueshift of 0.96 cm⁻¹ was noted for the average peak distances, and a redshift of 0.30 cm⁻¹ was noted for a mean peak intensity ratio.

The observed Raman shifts indicate the presence of strain within the MoS₂, which is due to the generation of vacancies, as reported in past studies^{[30, 31]}. The formation of S vacancies is more readily facilitated than Mo vacancies, primarily due to the lower displacement threshold energy and atomic mass of S atoms^{[21, 32]}. Initially, the fluence was maintained at a modest level of 10¹² p/cm², causing partial displacement of S atoms in MoS₂. This preferentially creates S vacancies, leading to the creation of point and line defects that induce compressive strain, resulting in a blueshift in the Raman peak position. As the proton fluence increased, additional Mo vacancies were produced, leading to the dissociation of Mo−S bonds. This process results in confinement of phonons due to the localized stretching of bonds (tensile strain) in the vicinity of defect sites, consequently causing a redshift of the ${\text{E}}_{\text{2g}}^{\text{1}}$ and ${\text{A}}_{\text{1g}}$peaks.

To further investigate the effects of DD on MoS₂, we subsequently conducted a comparative analysis of the electrical properties of MoS₂ FETs under different proton irradiation conditions, both before and after proton irradiation (Fig. 3(a)). The tested devices exhibited distinct DD responses at different irradiation fluences. Figs. 3(b)−3(d) present the transfer curves of 20 MoS₂ field-effect transistors (FETs) subjected to proton irradiation at distinct fluences of 10¹², 10¹³, and 10¹⁴ p/cm², respectively. The typical MoS₂ FET (highlighted in Fig. 1(e) and Figs. 3(b)−3(d)) was influenced noticeably following the proton irradiation fluence of 10¹² p/cm². For example, the V_th and SS of the device were measured as 14.6 V and 1.07 V/dec before proton irradiation, while after 10¹² p/cm² proton irradiation, the V_th and SS were measured as 15.9 V and 1.15 V/dec under identical testing conditions. However, following the proton fluence increased above 10¹³ p/cm² (10¹³−10¹⁴ p/cm²), there was no significant change in the electrical properties of the MoS₂ FET. For example, the V_th and SS of the device were measured as 15.7 V and 1.26 V/dec after 10¹³ p/cm² proton irradiation, while after 10¹⁴ p/cm² proton irradiation, the V_th and SS were measured as 13.2 V and 1.01 V/dec.

Figs. 3(e)−3(g) show the statistical results for V_th, SS, and I_on/I_off extracted from the transfer curves of MoS₂ FETs exposed to proton irradiation at three distinct fluence levels. The SS values of the MoS₂ FETs do not experience significant changes at fluences of 10¹² and 10¹³ p/cm² (from ~1.10 V/dec in the pristine sample to ~1.16 V/dec at 10¹² p/cm² and ~1.20 V/dec at 10¹³ p/cm²), while exhibit a noticeable decline at 10¹⁴ p/cm² influence (from ~1.20 V/dec at 10¹³ p/cm² to ~0.98 V/dec at 10¹⁴ p/cm²). The I_on/I_off values of irradiated MoS₂ FETs change slightly at 10¹² and 10¹³ p/cm² and enlarge at 10¹⁴ p/cm² compared with the pristine MoS₂ FETs (from ~4.05 × 10⁵ in the pristine sample to ~4.79 × 10⁵ at 10¹² p/cm², ~4.64 × 10⁵ at 10¹³ p/cm² and ~2.18 × 10⁶ at 10¹⁴ p/cm²). The I_on current shows no significant change before and after proton radiation, while primary reason for the enhanced I_on/I_off values is attributed to the reduction in the I_off current.

Comparing with SS and I_on/I_off, the V_th is an important metric describing the critical voltage at which the device turns on and its fluctuation will affect the power consumption of MoS₂ FETs. As shown in Fig. 3(g), with a small fluence of 10¹² p/cm², the V_ths of the MoS₂ FETs shifted in the positive direction (from ~14.7 V in pristine sample to ~16.2 V at 10¹² p/cm²). With the fluence of 10¹³ and 10¹⁴ p/cm², the V_ths of the FETs shifted in the negative direction (from ~14.8 V at 10¹³ p/cm² to ~13.7 V at 10¹⁴ p/cm²). Combined with the characterization in Fig. 2, the modification in the electrical properties of MoS₂ FET devices following proton irradiation is not inherently related to the MoS₂ itself. Instead, it might be attributed to DD-irradiation-generated interface trap states at the SiO₂/MoS₂ interface or the oxide traps in the SiO₂ layer.

Compared to silicon-based radiation-hardened FETs, the MoS₂ FETs fabricated in this study demonstrated smaller ΔSS (~ 45.8% at 2.7 × 10¹² p/cm² at silicon-based FETs^[33], ~8.99% at 10¹³ p/cm² in this work). To further investigate the mechanism, the SRIM Monte Carlo toolkit was utilized to model the penetration depth and displacement damage in the MoS₂ FETs subsequent to 30 keV proton irradiation. The SRIM simulations consisted of the source/drain region (Sb/Au/MoS₂/SiO₂/Si, Fig. 4(a)) and the MoS₂ channel region (MoS₂/SiO₂/Si, Fig. S2). The simulation model took into account the thickness and density of each layer, including Sb (20 nm, 6.70 g/cm³), Au (30 nm, 19.32 g/cm³) metal electrode, MoS₂ (1.20 g/cm³) channel, SiO₂ (285 nm, 2.15 g/cm³) and Si (725 μm, 2.33 g/cm³) substrate. The track distribution of 30 keV protons in monolayer MoS₂ with a SiO₂/Si substrate is shown in Fig. S3. To facilitate observation, only the top 350 nm depth of the device is displayed in Fig. 4(a). It can be seen from the enlarged view in Fig. 4(a) that fewer vacancies form in the MoS₂ compared with metal and SiO₂/Si substrate. In addition, a higher count of vacancies in the source/drain area compared to the channel area indicates that the DD effects to the metal-covered MoS₂ is more pronounced.

The displacement per atom^{[7, 27, 34]} (dpa, as detailed in the Experimental Section) indicates that under 30 keV proton irradiation, on average, only one vacancy is produced for every 244 atoms. This is attributed to the ultra-thin body and nanoscale sensitivity volume of MoS₂. To confirm the relationship between the sensitivity volume and DD induced damage, the SRIM simulations of MoS₂ with different thicknesses were performed. As shown in Fig. 4(b), with the thickness of the samples increasing, the number of vacancies induced by each proton gradually increases. Specifically, the number of vacancies per proton for monolayer, bilayer, and trilayer MoS₂ are 0.0025, 0.0040, and 0.0068, respectively. To verify the simulation results, we subsequently fabricated bilayer and trilayer MoS₂-based FETs which were procured via the mechanical exfoliation method and tested the electrical properties before and after proton irradiation (Fig. S4 and Figs. 4(c)−4(d)). After a 10¹⁴ p/cm² proton irradiation, the trilayer samples exhibit larger ΔV_th and ΔSS (3.19 V and 0.30 V/dec) than the monolayer MoS₂ FETs (2.98 V and 0.17 V/dec), which is corresponded to the simulation results.

The discussion above indicates that our MoS₂-based devices can withstand proton radiation at a level of 10¹⁴ p/cm², causing very minimal changes, which far exceeds the requirements for space applications. To benchmark the tolerance against the DD effect, the MoS₂ FETs have been compared with reported radiation-hardened FETs, such as SiC-based FETs^[35], CNT-based FETs^{[7, 11]}, graphene-based FETs^[36] or other FETs based on metal−oxide semiconductors^[37−39]. Considering the different channel materials and irradiation conditions employed in previous works, the radiation’s nonionizing energy loss (NIEL) values^{[25, 40]}, which represent the rate of energy loss attributed to atomic displacements as particles pass through a material^{[11, 20, 21, 35−39, 41]} (shown in Table S1), were selected as metrics for a fair comparison. The details about computational method are discussed in the Experimental Section. In this work, the NIEL of monolayer MoS₂ subjected to 30 keV proton bombardment was calculated as 0.156 MeV·cm²/g. The irradiation’s DD dose (D_d)^[42] provides a unified quantitative standard for assessing the degree of damage in different materials under various radiation fluences and it can be calculated through the following equation:

3$D_{\text{d}}\text{=NIEL}\times \text{fluence}\text{.}$

D_d is utilized to evaluate the MoS₂ FET DD tolerance^{[11, 20, 21, 35−39]}, as summarized in Figs. 4(e) and 4(f). Figs. 4(e) and 4(f) illustrates the percentage change in SS and ΔV_th at various D_d, respectively. It is clear that our MoS₂ device located at the preferred level with low variation in SS (less than 20% in Fig. 4(e)) and small change in V_th (~1.50 V in Fig. 4(f)). It is noted that CNT-based FETs can withstand higher D_d (e.g., 150 keV protons with a higher D_d of 1.51 × 10¹⁵ MeV/g)^[11], which is largely due to the strong carbon−carbon bonds. However, our devices exhibit a smaller variation in SS and V_th (11.5% of SS change rate and 0.98 V/dec of ΔV_th) under similar D_d conditions range from 10¹³ to 10¹⁴ MeV/g. We attribute the high tolerance against proton irradiation of MoS₂ to the smaller cross-section than CNTs, which should also be reproduced in two-dimensional materials-based devices. These results indicate the promising prospects of MoS₂-based FETs for electronic applications in Earth's orbital regions and even in deep space.

In this study, the DD degradation mechanism in MoS₂-based devices is systematically investigated using a 30 keV with various fluences from 10¹² to 10¹⁴ p/cm². Using the SRIM software, we simulated the DD radiation effects on the exposed channel region, demonstrating that the enhanced proton irradiation tolerance is due to the nanoscale cross-section of monolayer MoS₂. The monolayer MoS₂-based FETs fabricated in this work showed high DD radiation tolerance up to 1.56 × 10¹³ MeV/g (compared to 10¹¹ MeV/g required for outer space exploration), demonstrated MoS₂-based ICs as a prospective radiation-hardened technology for deployment in space exploration missions.