The development of modern integrated circuits (ICs) has been hindered by further downscaling the physical size of transistors. Single-walled carbon nanotubes (SWCNTs) are promising to replace silicon as a new generation semiconductor material to continue Moore’s Law due to the ultrathin body and excellent electrical properties^[1-5]. In particular, possessing strong C–C bonds, nanoscale cross-section and low atomic number, SWCNTs have great potential in fabricating radiation-hardened field-effect transistors (FETs), which are required in aerospace applications^[6-11].

With the advancement of space exploration, severe challenges have been put forward for the radiation resistance of ICs, which directly affects the lifetime of spacecrafts^{[12, 13]}. Space radiation sources mainly come from the particles in the geomagnetic trapping radiation belt, solar cosmic rays and galactic cosmic rays, all of which contain protons. Compared with high energy proton radiation, low energy proton (< 1 MeV) radiation often means higher displacement damage, which plays an important role in the degradation of transistors performance^[14]. Additionally, protons with energies ranging from 50 to 200 keV have uncertain effects on the transistors^[15-17]. Therefore, it is of great significance for the reliability of spacecrafts to study the effect of low energy proton radiation on transistors^[18].

However, studies are mostly carried out under high energy proton irradiation, where the resistivity of CNTs decreased (under 8–12 MeV)^[19], the structure of CNTs was considered unchanged (under 3.3 MeV)^[20] and the tensile strength of CNTs was strengthened (under 2 MeV with the fluence of 1 × 10¹³ p/cm²)^[21]. The lack of study on low energy proton radiation effect limits further understanding of radiation effects on CNTs. Thus, it is necessary to explore the radiation effects on CNTs under low energy proton irradiation.

In this article, semiconducting SWCNT FETs with a back-gate structure were prepared by the solution-deposited method, and the impact of 150 keV protons with different fluences up to 1 × 10¹⁵ p/cm² on them were investigated. The electrical behaviors of SWCNT FETs were studied before and after irradiation. Especially, Stopping and Ranges of Ions in Matter (SRIM)^[22] and Geometry and Tracking (GEANT 4)^{[23, 24]} software toolkits were used to simulate the radiation effects of 150 keV proton irradiation on metal/CNT contact and the exposed channel region, for the first time to our knowledge. By combing simulation results with the electrical measurements, the mechanism of SWCNT FET degradation by proton radiation was thoroughly revealed, which provides new insight for the radiation-harden technology of SWCNT film-based FETs.

Shown in the inset of Fig. 1(a) is the SWCNT FET applied in the irradiation experiment. The n-doped silicon substrate was employed as a back-gate. The semiconducting SWCNTs channel was fabricated from 99.9% arc-discharged CNTs dispersions purchased from Suzhou CINK Nano Materials Co. Ltd. The diameters of SWCNTs we used range from 1.3 to 1.7 nm, the lengths were between 0.8–2.5 μm, and the initial concentration was more than 0.2 mg/mL. The silicon substrate with 300 nm silica was ultrasonically treated with acetone, isopropanol and deionized water for 10 min, and after that the substrate was baked at 120 °C for 30 min. Then the SWCNTs were diluted for 20 times by o-xylene. Followed by 24 h immersion in diluted SWCNTs dispersions, the Si/SiO₂ substrate was rinsed with o-xylene, purged with N₂ and baked at 150 °C for 30 min at atmosphere.

In addition to the conventional FETs, the transmission line model (TLM) test structures were fabricated on the SWCNT film to study the radiation effects on the contact resistance (R_C) and the sheet resistance (R_SH) of SWCNTs. Active regions were etched by oxygen plasma to isolate the devices. Electrodes of Ti/Pd/Au (2/30/50 nm) were fabricated with electron beam lithography (EBL), electron beam evaporation and lift-off.

Surface morphology and quality of SWCNT film were characterized by atomic force microscopy (AFM) and Raman spectroscopy (LabRAM HR Raman system with a laser wavelength of 473 nm). Electrical properties were measured using the Keithley 4200 electrometer system.

The proton radiation experiment was carried out at the low-energy charged particle irradiation simulation test device of Harbin Institute of Technology. The irradiation ion energy is 150 keV, the beam current is 80 nA/cm², the proton fluences are 5 × 10¹², 5 × 10¹³, 5 × 10¹⁴ and 1 × 10¹⁵ p/cm², respectively, and the vacuum degree of sample chamber is 10^–4 Pa during the irradiation.

The AFM morphologic image of SWCNT films deposited on a silicon substrate with 300 nm silica is shown in Fig. 1(a) , where the inset is the optical image of SWCNT-film-based FET with a channel length of 10 μm and a width of 20 μm. The electrical properties of the SWCNT FETs were measured before irradiation. The transfer characteristics of the SWCNT FET shown in Fig. 1(b) indicates a typical p-type FET behavior, mainly because the absorption of water and oxygen molecules results in p-doping of the CNTs when exposed to air^[25]. The on/off ratio (I_on/I_off) of the SWCNT FETs before irradiation is more than 10⁴, which is suitable for a logic circuit. The mobility is about 10 cm²/(V·s). The low-energy proton irradiation device with an ionization chamber, accelerator and irradiation chamber is sketched in Fig. 1(c), in which the samples were irradiated with four different proton fluences of 5 × 10¹², 5 × 10¹³, 5 × 10¹⁴ and 1 × 10¹⁵ p/cm². Proton irradiation results in two main effects: total ionizing dose (TID) and displacement damage effect. TID effect accumulates positive oxide-trapped charges in the back-gate oxide (SiO₂), which will introduce negative induced charges into the CNTs, while displacement damage effect causes irreversible damage to the CNTs, thereby affecting the interface state (Fig. 1(d)).

SRIM software toolkit was employed to simulate the penetration depth and displacement damage of the CNT FET after 150 keV proton irradiation^{[15, 22]}. The simulation was divided into the source/drain region (inset of Fig. 2(a) : Au/Pd/Ti/CNT/SiO₂/Si) and the CNT channel region (inset of Fig. 2(c): CNT/SiO₂/Si). The simulation model was based on the thickness and density of each layer, including Au (50 nm, 19.32 g/cm³), Pd (30 nm, 12.02 g/cm³), Ti (2 nm, 4.54 g/cm³), CNT (1.5 nm (the deposited CNT film can be considered as a single layer according to the linear density of 20 lines/μm), 1.20 g/cm³), SiO₂ (300 nm, 2.15 g/cm³) and Si (500 μm, 2.32 g/cm³). The proton distribution in the metal-CNT contact region and the channel region is concentrated in the depth ranging from 1000 to 1500 nm, which corresponds to the Si substrate layer, indicating that the 150 keV protons penetrate the contact and the channel region during the irradiation process, and finally stopped in the Si substrate (Figs. 2(a) and 2(c)). Although most of the protons eventually stopped in the Si substrate, vacancies are generated in each layer along the trace during the proton irradiation, which is known as displacement damage. The number of vacancies in each layer is simulated, showing that the displacement damage caused by the proton irradiation is different in the CNT-layer covered by metal and that exposed to air (Figs. 2(b) and 2(d)). It can be seen from the enlarged image in Fig. 2(b) that the vacancy number of the CNT-layer covered by metal (red marking line: x = 82 nm) is much larger than that in the CNT-layer exposed to air (Fig. 2(d): x = 0 nm), which is of great significance for studying the properties of CNTs before and after irradiation.

GEANT 4 software was used to further explore the effect of proton irradiation on the contact and channel region of the CNT-layer^[23], computing the energy deposition of 150 keV protons passing through the CNT-layer with different fluences of 5 × 10¹², 5 × 10¹³, 5 × 10¹⁴ and 1 × 10¹⁵ p/cm². As shown in Fig. 3, the energy loss increases with irradiation fluence in both the source/drain and channel region of the CNT-layer. In addition, the deposited energy in the source/drain region of the CNT-layer is greater than that in the channel region at each fluence, revealing that the displacement damage of the metal-covered CNT is more serious.

A Raman spectrum was conducted to characterize the displacement damage of the CNT-film-based channel after irradiation^{[9, 26]}. The D peak intensity of the CNT channel increases with the increase of the irradiation fluence as plotted in Fig. 4(a), in which the G peak was normalized. The statistics result of the ratio of the D peak intensity to the G peak intensity (I_D/I_G) by Raman mapping is shown in Fig. 4(b), of which the test area is 5 × 5 μm² with a step of 0.5 μm. I_D/I_G increases from 0.10 (before irradiation) to 0.12 (5 × 10¹² p/cm²), 0.16 (5 × 10¹³ p/cm²), 0.22 (5 × 10¹⁴ p/cm²) and 0.30 (1 × 10¹⁵ p/cm²), indicating that displacement damage of the CNT-film-based channel increases with the increase of the irradiation fluence^[27], which is consistent with the simulation results in Fig. 3.

The electrical properties of the radiation-hardened CNT FET, represented by the transfer characteristics, were measured at room temperature in ambient air and found to be in line with the simulation results. The back-gate FET with 300 nm oxide is more sensitive to TID effect when compared to the top-gate FET with a thin oxide^{[8, 28]}. Fig. 5(a) shows the typical transfer characteristics of the CNT FET before and after irradiation with four different fluences in linear (left) and logarithmic (right) coordinates at V_DS = –10 V, from which the threshold voltage (V_th), subthreshold swing (SS), mobility, on-state current (I_on), off-state current (I_off) and I_on/I_off with four different proton irradiation fluences were calculated. The simplified Y-function method, expressed as^{[29, 30]}

$ Y=\frac{{I}_{\rm{DS}}}{\sqrt{{g}_{\rm{m}}}}=\left({\mu {C}_{\rm{OX}}{V}_{\rm{DS}}W/L }\right)^{0.5} \left({V}_{\rm{BG}}-{V}_{\rm{th}}\right){,} $  (1)

was used to extract V_th (Fig. 5(b)), where g_m is the transconductance defined as , C_ox is the capacitance density of 300 nm silica, W is the channel width and L is the channel length of the CNT FET. Obviously, V_th moves to the negative direction with the increasing irradiation fluences (from –17.24 to –30.04 V under the irradiation with the fluence of 1 × 10¹⁵ p/cm²), which was caused by TID effect (Fig. 6(a) showing the statistics measurements of V_th). When the samples are exposed to proton irradiation, the ionization effect generates electron–hole pairs in the oxide layer. The electrons escape to the upper surface at a very high speed (within picoseconds). During the escape process, they may recombine with the holes, and the holes that are not recombined will move to the carbon tube/silicon oxide interface at a low speed. In this process, some holes will be trapped, forming positive oxide-trapped charges^[31]. The positive oxide-trapped charges induce negative charges in the CNT channel (Fig. 1(d)), which can result in the negative shift of V_th.

In terms of the energy band, the p-type FET doped by water and oxygen has a Schottky barrier (SB) with metal, which plays a decisive role in the performance of the CNT FET^{[32, 33]}. The negative charges induced in the CNT channel cause the Fermi level to move toward the conduction band. Due to the pinning effect between CNT and metal^[34], the energy band of CNT bends downward and the SB will be thicker (Fig. 5(c)). In case of the on-state, holes are blocked by the thicker SB, and thus I_on decreases with the increasing fluence (Fig. 6(b)). Due to the obvious decrease of I_on, I_on/I_off decreases a little with the increasing irradiation fluence (Fig. 6(d)). In order to ensure the reliability of the change evaluation, the measurement of the device performance change was carried out in situ.

According to the energy band diagram in Fig. 5(c), the variation of I_off characteristics should be the same as that of I_on, but the result of the rate of I_off change in Fig. 6(c) is smoothly fluctuating, which results from the interface traps between CNT and SiO₂^{[35, 36].} After irradiation, a small amount of displacement damage was produced in the CNT and SiO₂, forming new interface states. As plotted in Fig. 5(d), the hysteresis voltage (H), which is generally defined as the absolute value of the difference between gate voltages (of positive and negative sweep) at half I_on, is relatively stable under the increasing irradiation fluence (H is 70, 71, 68, 73,70 and 69 V, respectively), indicating that the interface state changed little.

The trend of the rate of mobility change is similar to that of I_on under the four irradiation fluences (Fig. 6(e)), indicating that holes in the CNT channel were affected by the Coulomb scattering of the positive oxide-trapped charges in SiO₂^[37]. Besides, the degradation of mobility is also related to the displacement damage caused in the CNT layer.

The transfer characteristics in the logarithmic coordinate in Fig. 5(a) illustrate the radiation hardness of the CNT FET, where the subthreshold swing (SS) is almost the same with the fluences up to 1 × 10¹⁵ p/cm² (Fig. 6(f) showing the statistics measurements of SS). SS can be expressed as^[38]

$ {\rm{S}}{\rm{S}}={\rm{ln}} 10 \times kT/q \times (1+{C}_{\rm{it}}/{C}_{\rm{OX}}){,} $  (2)

$ {D}_{\rm{it}}={C}_{\rm{it}}/{q}^{2}{,} $  (3)

where k is the Boltzmann’s constant, T is the temperature, q is the elementary charge, C_ox is the capacitance density of 300 nm silica, C_it is the interface states capacitance, and D_it is the interface states density. The stability of SS further verifies that the interface states density (D_it) did not change significantly.

To go a further step, the contact resistance (R_C) and the sheet resistance (R_SH) were explored under four different irradiation fluences. The transmission line model (TLM) test structures with the same channel width (W = 7 μm) and different channel length (L = 4, 6, 10, 14, 18, 22, 28 and 35 μm, respectively) were fabricated to measure R_C and R_SH (the inset of Fig. 7(a) showing the optical image of a TLM structure). The total resistance (R_T) in a CNT FET consists of R_C and R_SH as

$ {R}_{\rm{T}}={2R}_{\rm{C}}+\frac{{R}_{{\rm{S}}{\rm{H}}}L}{W}{,} $  (4)

where R_T can be extracted from the I–V characteristics of a complete CNT TLM test structure as shown in Fig. 7(a)^[39-45]. The measured total resistances before and after irradiation with four fluences are plotted as a function of the channel length between two electrodes (Fig. 7(b)). 2R_C and R_SH/W are determined by the intercept and slope of the data fitting line in Fig. 7(b), respectively. On one hand, both R_SH and R_C increase with the increasing fluence, which is due to the thicker SB and the decreased mobility. On the other hand, the results show that R_SH is 1.1 (5 × 10¹² p/cm²), 1.1 (5 × 10¹³ p/cm²), 1.7 (5 × 10¹⁴ p/cm²) and 6.3 (1 × 10¹⁵ p/cm²) times of that before irradiation, and R_C is 2.2 (5 × 10¹² p/cm²), 6.9 (5 × 10¹³ p/cm²), 15 (5 × 10¹⁴ p/cm²) and 22 (1 × 10¹⁵ p/cm²) times of that before irradiation, indicating that R_C is more severely affected by proton irradiation, which agrees well with the simulation result that the displacement damage caused in the source/drain region of the CNT layer is more serious than that in the channel region.

In conclusion, we have fabricated SWCNT film-based FETs with back-gate structure, of which the electrical behavior was explored under low-energy proton irradiation. It is found that the TID effect caused the negative shift of V_th and the decrease of I_on, while other electrical parameters such as SS, I_off would not change obviously with the increasing fluence, revealing that the displacement damage caused in the SWCNT FET is not serious. More interestingly, the displacement damage in the metal/CNT and channel region was simulated and found to be different, explaining the various changes of R_C and R_SH. Combining the simulation results and electrical measurements, we have analyzed the low-energy proton irradiation mechanism of the CNT FETs, which provides meaningful guidelines for the radiation hardness technology of CNT film-based ICs for aircraft application in outer space.

This work was financially supported by the National Natural Science Foundation of China (No. 61704189), the Common Information System Equipment Pre-Research Special Technology Project (31513020404-2), Youth Innovation Promotion Association of Chinese Academy of Sciences and the Opening Project of Key Laboratory of Microelectronic Devices & Integrated Technology, and the Key Research Program of Frontier Sciences, CAS (Grant ZDBS-LY-JSC015)