Hydrogen-terminated diamond surfaces, emerging as a promising two-dimensional (2D) electron platform with great thermal and electronic properties, hold great potential for the next-generation high power and high frequency field effect transistor (FET). However, ideal gate dielectrics with high crystallinity and defect-free surfaces are still largely elusive. In this work, using the contamination-free pickup transfer method, hexagonal boron nitride (h-BN) flakes were fabricated on top of the hydrogen-terminated diamond surface to serve as a gate material and the passivation layer. The morphological and optical characterizations revealed the formation of homogeneous and intimate interface between h-BN and diamond. Benefiting from the h-BN gate dielectric layer, the maximum drain current density, subthreshold swing, and on/off ratio of diamond FET are measured to be −210.3 mA mm−1, 161 mV/dec, and 106, respectively. Especially, the transport measurement shows an almost constant Hall mobility of around 260 cm2 V−1 s−1 in the hole density range of 2 − 6 × 1012 cm−2, suggesting the excellent gate controllability of h-BN. Our results indicate that h-BN could form high-quality interface with hydrogen-terminated diamond, paving the way for the development of diamond-based electronic applications.
Diamond has received much attention in recent years due to its wide bandgap (5.47 eV), high carrier mobility (4500 and 3800 cm2 V−1 s−1 for electrons and holes at room temperature, respectively), large carrier saturation velocity (2.7 × 107 and 1.1 × 107 cm s−1 for electrons and holes, respectively), high thermal conductivity (∼22 W cm−1 K−1), large breakdown field (>10 MV cm−1), etc.1,2 These extraordinary properties endow diamond with great potential in high power and high frequency devices, even in harsh environments like high temperature or strong radiation conditions.3,4 Nonetheless, the lack of reliable doping technology has significantly impeded the progress of diamond-based electronics. For instance, the activation energies of traditional acceptor and donor dopants are very large (0.37 and 0.57 eV for boron and phosphorus, respectively).5 Compared to the conventional bulk doping method, a layer of two-dimensional hole gas (2DHG) can be generated through exposing the hydrogen-terminated (H-terminated) diamond surface to electron acceptor materials such as molecular species or metal oxides.6,7 Generally, the 2DHG is considered to be induced by surface transfer doping of the acceptors, leading to the p-type conductivity.8 In virtue of relatively facile fabrication, large hole density, and high carrier mobility, the field effect transistors (FETs) with 2DHG channel have been intensively investigated and become one of the mainstream diamond device architectures.2,8–10
However, the 2DHG on H-terminated diamond surface tends to degrade even at moderate temperatures due to the desorption of adsorbates.11 To achieve thermal stabilization, various gate dielectrics or passivation layers have been explored for H-terminated diamond FETs, such as Al2O3,12 HfO2,13 ZrO2,14 La2O3,15 SnO2,16 or BaF2.17 At the same time, low-temperature film deposition processes are essentially employed for these dielectrics to preserve the 2DHG.18 Some dielectrics, such as MoO3, WO3, and V2O5, have transfer doping effects by themselves, which can be utilized to realize high sheet density of holes.8,19 Whatever the material of dielectric, the property of the dielectric/diamond interface is one of the key factors to dominate the device performance,16,20,21 e.g., carrier mobility (μ), threshold voltage (VTH), and subthreshold swing (SS). Because of weak carrier scattering from high frequency of optical phonons and high velocity of acoustic phonons, the intrinsic mobility of bulk diamond is very high.22 Analogously, the hole mobility of 2DHG is predicted to be 3000 cm2 V−1 s−1 when only phonon scattering is taken into account without any external disorders.23 Meanwhile, the sheet density of the 2DHG can also exceed the order of 1013 cm−2. However, it is actually difficult to achieve high hole mobility for H-terminated diamond FETs, especially under large hole density. Apart from the crystallinity and surface structure of diamond, the amorphous gate dielectric and its interface defects are primarily responsible for the low hole mobility.20 These non-ideal factors cause intense scattering of carriers or high gate leakage current, which hinders the improvement of diamond device performance.23,24
Hence, a gate insulator with high single crystallinity and low interface defect density is a prerequisite for H-terminated diamond FETs. Hexagonal boron nitride (h-BN) is a van de Waals crystal with atomically flat surface free of steps, charged impurities, and unsaturated bonds, which has been established as an indispensable substrate or dielectric for electronic or optoelectronic devices based on 2D materials.25 Furthermore, h-BN possesses large bandgap (5.9 eV), large breakdown field (∼12 MV cm−1), and high thermal conductivity (∼500 W m−1 K−1).26,27 Using the pickup transfer method, the bubble-free and intimate h-BN/target interface can be obtained.28 These properties make h-BN a competitive gate material in H-terminated diamond FETs.20,29–31 In this work, we assembled high-quality h-BN/diamond heterostack using the pickup method and subsequently fabricated diamond FET. The interface properties and electrical performances of the device were systematically investigated at room temperature.
The fabrication process of H-terminated diamond FET is schematically shown in Fig. 1(a). A polished type-IIa CVD-grown single crystalline diamond with (100) oriented (5 × 5 × 0.4 mm3) was used as substrate in our experiment. The diamond substrates are purchased from Hebei Plasma Diamond Technology Co., Ltd (China). After the cleaning procedure, the diamond surface was treated with hydrogen plasma at 800 °C for 30 min in a MPCVD chamber to generate hydrogen termination. The diamond was kept in pure H2 atmosphere until it was cooled down to room temperature. A 50 nm-thick gold layer was deposited on diamond by the electron beam (EB) evaporation system to protect the H-terminated surface and simultaneously form a good Ohmic contact. Through photolithography and wet etching in potassium iodide (KI) solution, the exposed region on diamond surface was treated with low power oxygen plasma to achieve device isolation. The electrodes of a Hall bar structure were then fabricated after the as-deposited Au film was removed by a second photolithography along with etching in KI solution, as reported previously.4
(a) Fabrication process of H-terminated diamond FET with h-BN gate insulator. (b) Top view schematic of the device. (The light blue region indicates the H-terminated surface, and the gate metal was not shown here for clarity).
(a) Fabrication process of H-terminated diamond FET with h-BN gate insulator. (b) Top view schematic of the device. (The light blue region indicates the H-terminated surface, and the gate metal was not shown here for clarity).
The H-terminated diamond with metal electrodes and patterned channel has to be annealed in vacuum to effectively remove surface adsorbates. In our experiments, the chip was first vacuum annealed at 250 °C for 1 h in a CVD chamber equipped in a N2-filled glovebox (named GI), which was the optimized condition based on previous experiments. Subsequently, the annealed diamond was transferred from the CVD chamber in GI to another N2-filled glovebox (named GII) equipped with a transfer stage without any air exposure. The h-BN/diamond stack was prepared using a typical pickup method.28 Bulk h-BN crystals are commercially available from HQ graphene (Netherlands). First, flakes of h-BN are exfoliated onto a Si/SiO2 (285 nm) wafer in air and examined by optical microscopy and atomic force microscopy (AFM). Before picking up h-BN, the flakes on SiO2/Si will be baked on a hot plate at ∼110 °C for 15–30 min in glovebox. Then, the chosen h-BN flake was picked up with a poly-propylene carbonate (PPC)/polydimethyl siloxane (PDMS) stamp affixed to a glass slide. The h-BN/PPC/PDMS is placed onto diamond substrate by a microcontroller. Then, the substrate is heated to 90 °C to soften the PPC, leaving h-BN on the surface of diamond. Principally, this transfer method relies on the competition of adhesive (van de Waals) forces between the h-BN flake, substrates, and the PPC film.32 As a consequence, the PPC film can directly pickup the h-BN on SiO2/Si and be released from the PDMS stamp at high temperatures.33 The h-BN/diamond stack was then transferred back to the CVD chamber in GI to further dimmish the residuals on h-BN by vacuum annealing at 300 °C for 2 h. Finally, 20/200 nm Ti/Au gate electrodes were deposited by photolithography, EB evaporation, and the liftoff process in acetone. The top view illustration of the Hall bar structure is displayed in Fig. 1(b).
Figure 2 shows the optical images of the diamond surface before and after transferring h-BN. The thickness of h-BN was determined to be 52.3 nm by a Bruker Dimension ICON AFM. Due to the large bandgap of diamond (5.47 eV)2,3 and h-BN (5.9 eV),26,27 these two single crystals are both transparent in the visible light range. Even so, the stack region can still be seen when covered with h-BN, as shown in Fig. 2(c). The optical image of the device after depositing the gate electrode is presented in the inset. To investigate the morphology of the h-BN/diamond stack, the white dashed rectangle regions in Figs. 2(a) and 2(c) were scanned using AFM. Compared to the bare diamond surface [Fig. 2(b)], no bubbles or particles can be observed in the channel region underneath the h-BN flake in Fig. 2(d). One common problem of van der Waals stacks is the bubbles between different layers due to the gas, water, or hydrocarbon molecules trapped during the transferring process.34 These bubbles have detrimental effects on the device performance. For the pickup process, the interface cleanliness highly depends on the operation speed and temperature. At a relatively high temperature, the bubbles can be pushed out under an ultraslow proceeding velocity in the approaching region, effectively removing the trapped impurities at the interface. This is one of the major advantages of the pickup technique,25,28 which can yield better h-BN/diamond interface with almost no bubbles compared to the PMMA or PDMS transfer methods.20,29,31 Such interface could substantially benefit the device performance.
(a) Optical and (b) AFM images of the diamond surface before transferring h-BN. (c) and (d) Corresponding images after transferring h-BN. The inset of (c) shows the image of the device after the gate electrode deposition. The scale bars are 30 μm in (a) and (c), 6 μm in (b) and (d), and 5 μm in the inset of (c).
(a) Optical and (b) AFM images of the diamond surface before transferring h-BN. (c) and (d) Corresponding images after transferring h-BN. The inset of (c) shows the image of the device after the gate electrode deposition. The scale bars are 30 μm in (a) and (c), 6 μm in (b) and (d), and 5 μm in the inset of (c).
The h-BN/diamond stack was further characterized using a confocal Raman system (WITec Alpha 300R) under a 532 nm excitation laser at room temperature, and the Raman mapping was also performed. In Fig. 3(a), the sharp peak observed at 1332.2 cm−1 with a narrow full width at half maximum (FWHM) of ∼2.5 cm−1 characterizes the diamond phase, and no any other non-diamond features appear in the range from 1000 to 1500 cm−1.35 The intensity of this typical diamond peak is higher at the h-BN/diamond region compared to the bare diamond surface, which could be attributed to the enhancement induced by h-BN layer.36,37 This can be seen more clearly from the Raman intensity mapping at 1332.2 cm−1 in Fig. 3(c), and homogeneity of interface is ought to be confirmed. Due to the huge difference of diamond and h-BN in characteristic peak intensity under the same test condition, the h-BN peak is usually obscured in linear scale curve [the blue line in the main figure of Fig. 3(b)]. Only in the semi-log plot, the weak E2g peak of h-BN (∼1365 cm−1) can be resolved,26 as indicated by the rectangle box in the inset of Fig. 3(b). However, on the electrode region, the h-BN feature can be easily distinguished. In Fig. 3(d), the Raman intensity mapping at 1365.7 cm−1 outlines the h-BN flake corresponding to the optical image of Fig. 2(c).
(a) Raman spectra in linear scale taken from the four points marked in the inset (optical image of the stack), showing features from pure diamond region (cyan), h-BN/diamond stack region (blue), and electrode/diamond region (red). (b) Comparison of the Raman spectra in linear scale recorded at the h-BN on diamond (blue) and on electrode (green) [inset in (a)]. The inset shows the same plots in log scale. Raman intensity mapping of the h-BN/diamond heterostructure at 1332.2 (c) and 1365.7 cm−1 (d). The scale bars are 10 μm in (c) and (d).
(a) Raman spectra in linear scale taken from the four points marked in the inset (optical image of the stack), showing features from pure diamond region (cyan), h-BN/diamond stack region (blue), and electrode/diamond region (red). (b) Comparison of the Raman spectra in linear scale recorded at the h-BN on diamond (blue) and on electrode (green) [inset in (a)]. The inset shows the same plots in log scale. Raman intensity mapping of the h-BN/diamond heterostructure at 1332.2 (c) and 1365.7 cm−1 (d). The scale bars are 10 μm in (c) and (d).
(a) Output and (b) transfer characteristics of the diamond FET with 52.3-nm-thick h-BN as the gate material. The channel length and width are 10.4 and 3.4 μm, respectively. The inset shows |IDS|1/2 vs VGS curve. (c) |IDS|/|IGS|–VGS plots in the logarithmic scale. The inset is the C–V characteristic of the device tested at 1 MHz.
(a) Output and (b) transfer characteristics of the diamond FET with 52.3-nm-thick h-BN as the gate material. The channel length and width are 10.4 and 3.4 μm, respectively. The inset shows |IDS|1/2 vs VGS curve. (c) |IDS|/|IGS|–VGS plots in the logarithmic scale. The inset is the C–V characteristic of the device tested at 1 MHz.
To investigate the charge transfer mechanism of the h-BN/diamond interface, the capacitance–voltage (C–V) characteristic of the FET was tested at 1 MHz, as shown in the inset of Fig. 4(c). Apparently, the charge depletion and accumulation regions can be observed. The maximum gate capacitance is 0.047 μF/cm2, corresponding to the effective h-BN thickness of 51.4 nm, which is similar to the AFM measurement. The flatband voltage, VFB, is determined to be 4.65 V.39 The positive VFB indicates that holes concentrate under the gate dielectric layer. Due to limited lateral dimension of h-BN flake, the space between h-BN and diamond in the vicinity of the contact electrodes could serve as the passage channel for air molecules. Actually, the threshold voltage will increase after leaving devices in air, suggesting the adsorption of molecule species on the channel over time. Further encapsulation methods should be developed for achieving the normally off operation.
Transport behavior of another diamond FET with 25.2-nm-thick h-BN as gate dielectric. The channel length and width are 4.9 and 3.2 μm, respectively. (a) Cartoon representation of the electric connections for Hall measurement. (b) Sheet conductance (σ), (c) hole density (p), and (d) field effect mobility (μFE, black curve)/Hall mobility (μFE, blue curve) as a function of the gate bias voltage (VGS).
Transport behavior of another diamond FET with 25.2-nm-thick h-BN as gate dielectric. The channel length and width are 4.9 and 3.2 μm, respectively. (a) Cartoon representation of the electric connections for Hall measurement. (b) Sheet conductance (σ), (c) hole density (p), and (d) field effect mobility (μFE, black curve)/Hall mobility (μFE, blue curve) as a function of the gate bias voltage (VGS).
In summary, the fabrication and characterization of H-terminated diamond FET with h-BN gate dielectric were demonstrated. The h-BN/diamond stack was assembled using the pickup method. The high quality of h-BN/diamond interface was confirmed by the AFM and Raman characterizations. IDmax is extracted to be −210.3 mA/mm from the output characteristic of the device. The transfer curve shows that the values of VTH, on/off ratio, SS, and sheet resistance are 5.7 V, 106, 150 mV/dec, and 4.2 kΩ, respectively. The Hall measurement indicates that the hole mobility of 2DHG stays almost unchanged at ∼260 cm2 V−1 s−1 in the hole density range of 2 × 1012–6 × 1012 cm−2. Our work suggests the great potential of h-BN gate dielectric for achieving high-performance diamond FET.
This work was supported by the National Key Projects for Research and Development of China (Grant No. 2022YFA1204700), the National Natural Science Foundation of China (Grant Nos. 12074173 and 61804117), the Natural Science Foundation of Jiangsu Province (Grant No. BK20220066), the Program for Innovative Talents and Entrepreneur in Jiangsu (Grant No. JSSCTD202101), the Natural Science Foundation of Shaanxi Province (Grant No. 2022JM-364), and the Guangdong Basic and Applied Basic Research Foundation (Grant No. 2022A1515111075). The authors would like to acknowledge the support from State Key Laboratory of Intelligent Manufacturing Equipment and Technology and National Laboratory of Solid State Microstructures.
AUTHOR DECLARATIONS
Conflict of Interest
The authors have no conflicts to disclose.
Author Contributions
Yan Huang, Junfeng Xiao, and Ran Tao contributed equally to this work.
Yan Huang: Investigation (lead); Methodology (equal); Writing – original draft (equal). Haolin Wang: Conceptualization (equal); Investigation (equal); Methodology (equal); Writing – original draft (equal); Writing – review & editing (equal). Lei Wang: Conceptualization (equal); Funding acquisition (equal); Project administration (equal); Resources (equal); Writing – review & editing (equal). Junfeng Xiao: Investigation (equal); Methodology (equal). Ran Tao: Investigation (equal); Methodology (equal). Zhi Liu: Investigation (equal); Methodology (equal). Yiran Mo: Investigation (supporting); Methodology (supporting). Xinxin Yu: Investigation (supporting); Methodology (supporting). Zhengyi Cao: Methodology (supporting). Yun Wu: Conceptualization (equal); Project administration (equal); Resources (equal). Zhonghui Li: Project administration (equal); Resources (equal).
DATA AVAILABILITY
The data that support the findings of this study are available from the corresponding authors upon reasonable request.