The standard back-gated structure depicted in Figure 1a is commonly used to study the electrical transport properties of SWCNTs and graphene. The structure allows for rapid fabrication of devices with variable dimensions. Furthermore, any semiconductor or metal substrate with an insulating dielectric layer may serve as the global back-gate electrode to modulate the Fermi level of the carbon active layer. The devices studied here use thermally grown SiO₂ on p⁺-Si as the starting substrate.

Though versatile, these back-gated structures are not optimized for high performance since they typically have thick dielectric layers (>1000 Å) resulting in low gate capacitance, and thus, require biases of tens-of-volts to fully saturate or pinch-off the drain current. Furthermore, thick dielectric layers are highly susceptible to radiation induced charge build-up, which is known to cause threshold voltage shifts and increased leakage in metal oxide semiconductor (MOS) devices [16]. To mitigate these effects, we locally etch the dielectric layer in the active region of the back-gated FET, which is depicted in Figure 1b. We then deposit a gate dielectric material (e.g., Al₂O₃ or Si₃N₄; depicted in red in Figure 1b) over the entire substrate with a thickness controlled by growth. This design feature is important since high quality dielectric layers deposited directly onto SWCNTs or graphene for use as a top gate electrode are still under development. Using the locally etched back-gated (LEBG) structure allows us to maintain the properties of the transferred carbon nanomaterial channel, and have flexibility in the choice and processing performed on the dielectric layer prior to transferring the channel material and fabricating devices.

We pattern the channel regions of the LEBG devices using standard photolithographic procedures. These substrates are etched for 1–3 min in buffered oxide etch (BOE) (7:1 NH₄F:HF), followed by a short ~1 min rinse in flowing deionized water. This etchant produces smooth hydrogen-terminated Si terraces [17,18], thereby providing a controlled surface for subsequent dielectric layer deposition.

We deposit Si₃N₄ films on LEBG substrates and BOE etched control wafers for optical and electrical characterization using an Oxford Instruments PlasmaPro™ System 100. Freshly prepared substrates are loaded, pumped to 5 × 10⁻⁶ Torr and preheated to 350 °C for 10 min. Si₃N₄ deposition proceeds at a pressure of 2 Torr with an N₂ carrier gas, and SiH₄ and NH₄ as the Si and N precursors, respectively. We obtain an index of refraction of 1.97 at 633 nm based on fitted reflectance measurements indicating nearly stoichiometric Si₃N₄. Following SWCNT thin-film deposition, annealing the structures in air at 300 °C for 16 h introduces oxygen into the nitride layer transforming it into silicon oxynitride (SiON) with an index of refraction of 1.75 at 633 [11]. On control samples, we deposit Ti/Au contact pads to perform capacitance and electrical breakdown measurements. For a 23 nm SiON film we obtained an average dielectric constant of ε_r = 5.5 at 1 MHz and dc-breakdown fields in excess of 8 MV/cm [11].

We deposit Al₂O₃ films on LEBG substrates and BOE etched control wafers for optical and electrical characterization using an Oxford Instruments FlexAL^® system. In a similar fashion as the Si₃N₄ we deposit Al₂O₃ on freshly prepared substrates, which are loaded, pumped, and preheated to 300 °C. The growth proceeds at this temperature as trimethyl aluminum (TMA) the Al precursor and oxygen plasma are alternatingly pulsed into the growth chamber. We deposited Ti/Au contact pads to perform capacitance and electrical breakdown measurements. For a 32 nm Al₂O₃ film we obtained an average dielectric constant of ε_r = 8.3 at 1 MHz and dc-breakdown fields in excess of 10 MV/cm.

We recently observed that the response of SWCNT-TFTs to Co-60 exposure is highly dependent on the local environment of the device during irradiation [8]. We report here a similar experiment on a standard back-gated graphene FET with a 250 nm SiO₂ gate oxide, channel length of 15 µm, channel width of 60 µm, and a positive gate bias of 5 V during irradiation. Figure 2a illustrates the TID dependent transfer characteristics of the wire bonded standard back-gated graphene FET measured in static vacuum. We measured the pre-irradiation transfer characteristics (black trace labeled 0 krad(Si)) immediately before the first TID exposure (1 krad = 10 Gy). We performed control measurements which yielded constant transfer characteristics for the device held in the sample chamber under static vacuum prior to irradiation. We observe a large shift in the transfer characteristics (towards negative V_g) following a TID of 20 krad(Si), which proceeds with further dosing. Following a TID of 40 krad(Si), the Dirac point becomes apparent at V_g = 73 V along with the onset of the electron transport for V_g > 73 V. These results are consistent with hole trapping in the SiO₂, which increases the Fermi level closer to mid-gap. Following the initial Co-60 irradiation up to 5 Mrad(Si), we exposed the device to air for 15 m then repeated the experiment with the same device in an air ambient as shown in Figure 2b. During the 15 m repose, the transfer characteristics shift towards positive gate voltage, resulting from room temperature annealing of the radiation-induced trapped charges and due to adsorption of molecular species from the air. Additional irradiation in an air ambient causes the transfer characteristics to shift further towards positive gate bias, the opposite of what we observed when irradiated under vacuum. This indicates that molecular species from the air, namely oxygen and water, overpower the effects of SiO₂ trapped holes and decrease the Fermi level promoting hole doping in the channel [14].

Using the standard long-channel field effect mobility equation: , where is the channel length, is the width, is the oxide capacitance, is the drain current, and is the gate voltage, we calculate the peak from the transfer curves as plotted verses TID in Figure 3a for the graphene FET irradiated in vacuum and air. When irradiated in vacuum, the hole initially degrades from 1090 cm²/Vs to 980 cm²/Vs following a TID of 200 krad(Si), recovers slightly then degrades again reaching a minimum mobility of 896 cm²/Vs after a TID of 5 Mrad(Si). During the 15 m repose, the mobility recovers considerably from the low of 896 cm²/Vs increasing to 1126 cm²/Vs exceeding the pre-irradiation mobility. With increasing TID exposure in air, the mobility decreases to 907 cm²/Vs after a TID of 100 krad(Si), then after a brief plateau, degrades to 663 cm²/Vs following a total (additional) TID of 5 Mrad(Si) in air.

We also plot the variation in with TID for the transfer curves measured in vacuum in Figure 3a (lower right y-axis). In this figure, displays non-monotonic behavior that approximately follows that of the mobility (measured in vacuum). The , for devices of these dimensions, provides a relative measure of the charge inhomogeneity at the graphene-substrate interface resulting from trapped charges in SiO₂ [19] and adsorbed impurities including oxygen, moisture, and photoresist residues [20]. The initial decrease in with TID reflects the increasing trapped charge density and magnitude of charge potential fluctuations (i.e., electron-hole puddles) within the graphene channel. These fluctuations restrict current flow in the graphene channel, which favors transport through regions of unperturbed potential, and is expected to reduce the mobility as we observe. Following a TID of 200–500 krad(Si), the and begin to recover. We attribute this behavior to a reorganization of the potential fluctuations in the graphene, potentially resulting in a correlated charge distribution in the SiO₂ [21] or rearrangement of mobile surface adsorbates, though additional work is needed to confirm this mechanism.

In Figure 3b we provide characteristic pre- and post 10 Mrad(Si) Co-60 irradiation Raman spectra of the graphene device measured in Figure 2. We measure a 2D-band (2670 cm⁻¹) to G-band (1580 cm⁻¹) peak intensity ratio of 1.5, which is characteristic of single layer graphene on SiO₂ [22]. The D-band mode (~1345 cm⁻¹) results from disorder in the graphene, potentially arising from defects, edges, wrinkles, or residual PMMA resist, but the relatively low intensity indicates the disorder is minimal. Furthermore, we observe no significant difference in the D-band intensity following irradiation indicating that the changes observed in the graphene FETs due to Co-60 irradiation result from trapped charges in the SiO₂ gate oxide and not due to changes in lattice defect concentration. These results are consistent with our recent study on SWCNTs-TFTs and emphasize the need to control the environmental conditions of carbon electronics, especially when investigating the basic radiation response mechanisms [8]. Furthermore, the controlled environment results (vacuum data) confirm that, like Si MOS-FETs, carbon electronics are susceptible to the oxide trapped charge effects necessitating the development of radiation-hardened gate dielectrics.

In Figure 4a we compare the transfer characteristics of two LEBG SWCNT-TFTs with either a 32 nm SiON or a 32 nm Al₂O₃ gate dielectric layer, 2 µm channel lengths, and 32 µm channel widths. Both devices are representative of the devices with drain current I_d,on/I_d,off greater than 10⁴ (about 20 devices for each gate dielectric). For both dielectric layers, we see a maximum drain current of 1 × 10⁻⁷ A and a consistently lower minimum drain current for the SiON (about 10× lower). The larger off-current may result from the onset of electron transport in the Al₂O₃ device reflecting the ambipolar transport properties of SWCNTs [23]. A different level of adsorbed water due to the different surface energies of Al₂O₃ and Si₃N₄ may cause this, which is supported by the fact that similar ambipolar behavior is observed in LEBG SWCNT-TFTs with a SiON gate dielectric following vacuum annealing at 125 °C and 1 × 10⁻⁵ Torr (see Figure 4b).

In Figure 4b we investigate the effects of Co-60 irradiation for a LEBG SWCNT-FET with a 23 nm SiON gate oxide and channel length and width of 2 µm and 16 µm, respectively, biased with a 0.25 MV/cm gate field during irradiation. By scaling the gate dielectric material to 23 nm we observe essentially no effect of Co-60 irradiation on the transfer characteristics up to a TID of 2 Mrad(Si). We observe a maximum shift in the threshold voltage of −0.25 V at a TID of 100 krad(Si) after which the transfer characteristics begin to shift back towards positive V_g values. The stability in V_g results from lower trapped charge accumulation in the thin SiON layer—the thinner layer allows more carriers to escape by tunneling or through field assisted transport [24]. Furthermore, the trapping characteristics of SiON are distinct from SiO₂ and favor electron trapping over hole trapping [16]. Therefore, we attribute the positive shift in threshold voltage to electron traps in the SiON [11], although it is not possible to completely rule out some molecular doping during irradiation in static vacuum conditions.

The data we have presented here is promising evidence towards hardening carbon nanoelectronics against TID exposure. Therefore, our future investigations are aimed at maintaining the radiation tolerance levels while simultaneously improving device performance and reducing the susceptibility to atmospheric doping effects.