Tuning diamond electronic properties for functional device applications

Abstract

Because of its ultrahigh hardness, synthetic diamond has been widely used in advanced manufacturing and mechanical engineering. As an ultra-wide bandgap semiconductor, on the other hand, diamond recently shows a great potential in electronics industry due to its outstanding physical properties. However, like silicon-based electronics, the electrical properties of diamond should be well modulated before it can be practically used in electronic devices. In this work, we briefly review the recent progresses in producing high-quality, electronic grade synthetic diamonds, as well as several typical strategies, from the conventional element doping to the emerging “elastic strain engineering,” (ESE) for tuning the electrical and functional properties of microfabricated diamonds. We also briefly show some device application demonstrations of diamond and outline some remaining challenges that are impeding diamond’s further practical applications as functional devices and offer some perspective for future functional diamond development.

Keywords:

• Diamond
• electronic properties
• bandgap modulation
• elastic strain engineering
• wide bandgap semiconductor
• functional device

1. Introduction

Diamond is the hardest natural material. Due to its extreme hardness, diamond has been widely used as a precision cutting tool material in advanced manufacturing machinery equipment to achieve high machining quality [1–5]. In addition to its extensive use in mechanical area, diamond also shows great application potential as an electronic material.

As a member of ultra-wide bandgap semiconducting material, diamond owns many excellent physical and chemical properties (e.g. high thermal conductivity, high breakdown field, and high carrier mobility) [6–10]. These outstanding properties make diamond an appealing material for electronics application, especially under certain extreme conditions (such as high temperature, high radiation, etc.) where conventional silicon-based semiconductors can hardly serve [11–14]. Excellent mechanical properties and great electrical application potential make diamond one of the hottest research topics.

In this review, we will introduce some of the latest research work about diamond. The review will start with a brief introduction of diamond synthesis technology. Then we will introduce several strategies for modulating the electrical properties of diamond, including element doping, surface engineering, phase engineering, strain engineering, and defect engineering. Some devices applications of diamond and remaining problems impeding the application of diamond as electronic material are summarized in the last part.

2. Synthetic diamonds for device applications

As a material with excellent mechanical and functional properties, diamond is greatly needed in a wide range of fields. The high cost of natural diamond pushes forward the research of synthetic diamond. Diamond can be divided into single crystal diamond (SCD) and polycrystalline diamond (PCD). PCD has higher concentration of defects comparing to SCD due to the presence of grain boundary and the electrical properties of devices made by PCD are significantly limited as a result [15]. Therefore, if we want to exploit diamond’s potential as electronic and optical devices (UV/radiation detector, field-effect transistor, diode, etc.), SCD is a better choice than PCD. In the following part of this review, all the diamonds we talk about are SCDs if there is no specific explanation.

Diamond can be synthesized by two main methods: high pressure and high temperature (HPHT) and chemical vapor deposition (CVD) [16–19]. The HPHT method mimics the formation of natural diamond, which transforms graphite into diamond under HPHT [20,21]. HPHT method is of low cost and high yield. However, this method also has some problems. For example, impurities from air or the catalysts can be easily introduced into diamond during the synthesizing process [22–24]. Also, the synthetic diamond size is limited by the HPHT reaction chamber size, making HPHT diamond not an ideal choice for the semiconductor applications.

Another synthesizing method is CVD [25–28], which allows a lower temperature and pressure condition to synthesize diamond [29]. This technique, especially the micro plasma CVD (MPCVD), has been widely used for growing high-quality and large-size diamond. As shown in Figure 1(a–d), the research of MPCVD SCD growth has been conducted from several different aspects, including the preparation with high growth rate, low dislocation density, large size, and flat surface.

Figure 1. Research progresses of synthetic diamonds. (a) Nitrogen addition to increase diamond growth rate [35], diamond plate with (b) low dislocation density, (c) large size, and (d) flat surface [39,45,49]. (a) Reproduced with permission from ref 35, Copyright 2002, The National Academy of Sciences.(b) Reproduced with permission from ref 39, Copyright 2014 AIP Publishing LLC. (c) Reproduced with permission from ref 45, Copyright 2011 Elsevier B.V. All rights reserved. (d) Reproduced with permission from ref 49, Copyright 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim.

It is necessary to increase diamond growth rate as there is a strong demand in both industrial production and academic research [28,30]. The diamond growth mechanism has been investigated extensively [31,32]. One efficient strategy is by increasing plasma density during the growth process [33,34], which can be realized by increasing growth pressure or/and growth power. The addition of nitrogen can improve the diamond growth rate as well [30,35]. C. Yan et al. [35] achieved a diamond growth rate of 150 μm/h under a growth pressure of 200 Torr with nitrogen addition. The optical image of the synthesized diamond is shown in Figure 1(a). The sample is yellow due to the presence of single substitutional nitrogen. Q. Liang et al. [30] increased growth pressure to 300 Torr and achieved a growth rate of 165 μm/h.

Low dislocation density is important for the electronic applications of diamond [36,37]. A small number of defects would significantly weaken the electrical properties of diamond semiconductor devices. Some serious defects may even lead to the direct failure of the device. The intrinsic defect within diamond seed is the main defect source for the synthesized diamond. Several strategies can be used to minimize the influence from the seed defects. Takeuchi et al. [38] proposed a two-step growth method for growing diamond with low dislocation density, in which case a buffer layer of low dislocation density was synthesized at low growth rate first, followed by the high-speed growth of low-dislocation diamond on the buffer layer. The use of high-quality seed crystal is also important. Mokuno et al. [39] synthesized low-dislocation diamond using high-quality diamond substrate. The dislocation density is <100 cm⁻² for the diamond substrate seed and ∼400 cm⁻² for the synthetic diamond after growth (Figure 1(b)).

Large size is another important parameter [40–42]. Many microelectronics chips have a special demand of the size for integration of large numbers of transistors. Moreover, more chips can be fabricated from one single large-size wafer. Therefore, obtaining large (single crystal) diamond wafers with relatively low cost is another key point of the commercial application of diamond semiconductor devices. Mosaic growth is a popular strategy to realize large size diamond [43]. In this method, several small square diamond substrates are pieced together for growing one large diamond [42,44,45]. Figure 1(c) shows a 20 mm × 40 mm size diamond wafer prepared through this method [45]. However, defects such as stacking faults, twins, and dislocations can easily form in the junction region between the adjacent diamond seeds during the mosaic growth. Heteroepitaxial growth (i.e. growing diamond on silicon or sapphire) can also achieve the large-size diamond growth, but with a high dislocation density [46].

The flat surface of diamond is important as well [47–50], which is not only essential for diamond optics, but also ideal for the heterogeneous integration of diamond with other materials (e.g. GaN). Okushi et al. [48] successfully synthesized diamond with an atomically flat (001) surface, of which the roughness was reduced to 0.04 nm. However, the low growth rate of only tens of nanometers per hour cannot meet the requirement of industrial production. Bogdan et al. [49] improved diamond growth rate to 5 μm/h and guaranteed a low surface roughness of about 0.6 nm at the same time (Figure 1(d)).

Despite all these progresses, it is hard to achieve all the above requirements (high growth rate, low dislocation density, flat surface, etc.) in one sample, and tradeoffs are often made. For example, nitrogen addition is a common method to increase diamond growth rate [35], but this will introduce nitrogen impurity into diamond, which is not ideal for electronic-grade SCD. Another example is by using low growth temperature and methane content to grow diamond with low dislocation density or high-quality surface [39,48], but these growth parameters will result in low plasma density and reduce diamond growth rate significantly. Some strategies can be used to handle these conflicts. For example, it is feasible to achieve both high speed and high quality growth by using specially designed, high-power MPCVD device [33,34]. High device power has high requirement for MPCVD device. And it is more difficult to modulate plasma status under high power condition. With the development of MPCVD device, the improvement of seed crystal quality and diamond synthesizing technique, one can expect to see mass production of large-size and high-quality SCD in near future.

3. Modulating diamond’s electrical properties

Diamond has a great potential in semiconductor field as it owns excellent physical properties. However, intrinsic diamond shows an electrically insulating state due to its ultra-wide bandgap (∼5.5 eV). Researchers have been trying different methods in the hope to modulate its electrical properties. These methods are listed in Figure 2. In this section we will introduce the research progress regarding these strategies.

Figure 2. Strategies for modulating diamond’s electrical properties: element doping, surface engineering, phase engineering, strain engineering, and defect engineering.

3.1. Doping

Element doping can be divided into p-type or n-type depending on the electrical conductivity characteristic of the diamond after dopant incorporation. Element doping is the most conventional and popular way to modulate the electrical properties of diamond, which can be realized by two methods: ion-implantation and doping during diamond growth process [51,52]. In ion-implantation, the accelerating field accelerates the dopant ions to give them the kinetic energy. Then the dopant ions will be implanted into the diamond directly. With such method, the dopant can be introduced into the desired area and the doping concentration can be controlled [53,54].

However, it is hard to achieve a uniform doping in diamond through ion-implantation. Diamond has a rigid lattice structure with ultra-strong covalent bond, which makes it difficult to implant dopant ions into diamond. Also, the dopant concentration distribution for single energy level ion-implantation is close to normal distribution [55]. Multiple energy level or dopant dose ion implantations should be applied to get a relatively uniform dopant distribution. In addition, ion implantation will cause damage to diamond crystal structure and induce extra defects [51,56]. These dopant atoms cannot substitute carbon atoms efficiently and annealing heat treatment is often needed to eliminate these unwanted defects.

To avoid these problems occurred during ion-implantation, dopants are usually incorporated during the diamond growth process. There is a very limited room for the HPHT method to achieve this, although HPHT is effective to produce p-type boron-doped diamond. As shown in Figure 3(a), the incorporation of dopant is mainly conducted during the CVD diamond growth process [52], which uses methane and hydrogen as the main source gas for diamond growth [57], and dopant gas such as diborane or phosphine to achieve p- or n-type doping. Uniform dopant distribution can be realized by this method and the number of the unwanted crystallographic defects is much less than that in the ion-implantation.

Figure 3. (a) Schematic of dopant incorporation during CVD diamond growth process. (b) Heavily boron-doped diamond, showing dark color [60]. (c) Phosphorous-doped diamond, showing yellow color [66]. (b) Reproduced with permission from ref 60, Copyright 2010 American Institute of Physics. (c) Reproduced with permission from ref 66, Copyright 2016 Author(s). Published by AIP Publishing.

The most widely used and well-studied p-type dopant for diamond is boron (B) [58–60]. Werner et al. [52] realized B doped of diamond by using BCl₃ as boron source. However, the Hall mobility of doped diamond was very low. With the development of the B doping of diamond, the Hall mobility of B doped diamond reached 2200 cm²/(V·s) at room temperature, which is sufficient for semiconductor application use [61]. Figure 3(b) shows a heavily B-doped diamond, of which the dark color was caused by the high concentration of B [60]. By changing the B concentration, the electronic state of diamond can be tuned from semiconducting to metallic and even superconducting [62]. Although some problems remain to be solved, the preparation of p-type ­diamond is mature and practical enough for semi­conductor devices such as field-effect transistors and Schottky diodes.

On the other hand, n-type doping of diamond is extremely difficult. Various elements (e.g. phosphorus, nitrogen, lithium, and sulfur) have been studied but none of them has given a satisfactory result due to the higher activation energy and low doping efficiency. So far, the best candidate is phosphorus (P), which has been proved to form shallow donor level (∼0.6 eV below conduction band minimum) in diamond [63]. Koizumi et al. [64] realized n-type doping of diamond with Hall mobility of 23 cm²/(V·s) at 500 K by introducing P on diamond (111) surface, followed by a series of studies for increasing the Hall mobility of P-doped diamond [65–67]. Based on the successful attempt of the preparation of the P-doped n-type diamond, along with the B-doped p-type diamond, the first diamond pn junction diode was made in 2001 [68]. Kato et al. [66] realized the highest Hall mobility of 1060 cm²/(V·s) at room temperature for the P-doped diamond by reducing electron concentration to about 10¹⁰ cm⁻³. The optical image of P-doped diamond film can be seen in Figure 3(c). Although it is feasible to increase the Hall mobility of n-type diamond by using suitable doping strategies, a series of problems remain, including low solid solubility of doping atom, low ionization efficiency induced by deep donor level and low carrier concentration by charge compensation effect of crystal defects. These problems make it hard to obtain n-type diamond of high electrical conductivity at room temperature and restrict the development of diamond semiconducting devices.

As the single-element-doped diamond (especially n-type doped diamond) cannot fully satisfy the requirement for semiconductor applications, the researchers have been trying to dope diamond with multiple elements (co-doping) and also getting some promising results for both p- and n-type diamonds. For p-type doping, Li et al. [69] realized co-doping of boron and hydrogen in SCD under HPHT condition. The lattice structure of B-H co-doped diamond was more compatible and the carrier concentration and electrical conductivity were also improved compared with the B-doped diamond. Issaoui et al. [70] conducted boron and nitrogen co-doping in SCD under CVD condition, and demonstrated that a small amount of nitrogen addition can significantly limit the formation and propagation of defects and improve the diamond quality. For n-type doping, Hu et al. [71] obtained boron and phosphorus co-doped diamond by ion-implantation. The carrier concentration was similar with that of the P-doped diamond, but the Hall mobility and electrical conductivity were greatly improved. Li et al. [72] realized boron and sulfur co-doping in CVD diamond, of which the solubility of S was improved and the electrical conductivity appears higher than the S-doped diamond.

3.2. Surface engineering

Surface modification of diamond can be realized by using suitable treatments such as oxidation, hydrogenation, and fluorination [73–77]. Surface modification can drastically alter the functional properties of diamond, including electrical, tribological, and electrochemical properties [78–81]. The modulation of diamond’s functional properties by surface modification can be considered as surface engineering. In this section, we will focus on the effects of surface engineering on diamond’s electrical properties. Compared with the conventional doping of bulk diamond, surface modification changes the electrical properties of diamond by the modification on the surface level, but can improve the performance of diamond electronic devices significantly.

Recently, the research of surface engineering for diamond’s electrical property modulation focuses mainly on surface oxygen termination and surface hydrogen termination. Surface oxygen termination is a common treatment for diamond surface. Oxygen-terminated surface can be realized by carboxyl or hydroxyl termination [82–84]. As shown in Figure 4(a), carboxylated diamond surface can be realized by various methods [74]. Oxygen-terminated diamond surface shows positive electron affinity and is insulating in both vacuum and air conditions. The oxidation treatment of diamond surface can also improve the diamond purity by removing non-diamond carbon and metallic impurities, which is important in semiconductor field.

Figure 4. Surface engineering of diamond. (a) Carboxylated diamond surface can be realized by various methods [74]. (b) Hydrogenation of diamond by rotating plasma reactor [73]. (c) Schematic of graphite/hBN/hydrogen-terminated diamond [92]. (a) Reproduced with permission from ref 74, Copyright 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim. (b) Reproduced with permission from ref 73, Copyright 2010 Elsevier B.V. All rights reserved. (c) Reproduced with permission from ref 92, Copyright 2021 the author(s), published by Nature Publishing Group.

Surface hydrogen termination is another popular surface modification technique [85] which can be realized by hydrogen plasma (Figure 4(b)) or annealing at 500 °C under hydrogen atmosphere [73,86]. Hydrogen terminated diamond surface shows specific electrical properties, including p-type electronic conductivity and negative electron affinity [87–89]. A surface resistance of 10⁶ Ω*cm can be reached for hydrogen-terminated diamond [78], which is significantly less than that of normal undoped diamond (∼10¹⁶ Ω*cm). A commonly accepted reason for the conductivity of hydrogen-terminated diamond surface is the surface transfer doping mechanism [88]. The introduced hydrogen atoms work as acceptor impurity and increase the hole density near the diamond surface region. Also, the upward band bending occurs due to the C-H surface dipole which can facilitate the hole accumulation [88,90]. On the other side, the negatively charged acceptors can lead to hole scattering and reduce the carrier mobility [85,91]. To avoid this situation, high mobility p-channel wide-bandgap transistors can be made by reducing the surface acceptor density on hydrogen-terminated diamond surface [92]. One of the examples using such strategy is shown in Figure 4(c), in which hexagonal boron nitride and graphite were selected as the gate insulator and electrode, and a very high room-temperature Hall mobility (680 cm² V⁻¹ s⁻¹) was obtained [92].

Surface modification also has some limitations. It can only modulate the diamond properties within surface area in contrast with the conventional element doping of bulk diamond. Although p-type electrical conductivity of diamond can be achieved by hydrogen termination, there is no suitable surface modification strategy to realize n-type electrical conductivity of diamond, which limits its application range severely.

3.3. Phase engineering

Carbon material is a very big family, consisting of many allotropes including diamond, graphite, and fullerene which own distinct properties. Property difference is the result of different microstructures. The research of tuning material property by synthesizing material with different phases or conducting phase transformation can be thought as phase engineering. So far, the research of phase engineering for carbon materials has achieved some important progresses.

Generally, carbon materials can be divided into two types: crystalline carbon and amorphous carbon depending on the possession of long-range order in their lattices or not. Diamond is a typical crystalline carbon material while coal is an example of amorphous carbon material. Tang et al. [93] synthesized a new form of carbon material named “paracrystalline diamond.” The optical image of this “paracrystalline diamond” is shown in Figure 5(a). The “paracrystalline diamond” consists of sub-nanometer size paracrystallites which possess crystalline medium range order up to a few atomic shells. This material covers the gap between crystalline carbon and amorphous carbon materials.

Figure 5. Phase engineering of diamond. (a) Optical figure of paracrystalline diamond [93], (b) optical figure of ultrahard, diamond-like amorphous carbon [96]. (a) Reproduced with permission from ref 93, Copyright 2021 the author(s), published by Nature Publishing Group. (b) Reproduced with permission from ref 96, Copyright 2021 the author(s), published by Nature Publishing Group.

Pure diamond is a crystal consisting of sp³ hybridized carbon atoms. It has ultra-high hardness but is brittle and has a poor electrical conductivity. In contrast with diamond, graphite with sp² bonding is soft and electrically conductive. Researchers have been looking for new carbon materials which show good electrical conductivity like graphite and high hardness like diamond. Zhang et al. [94] synthesized amorphous carbon material with a high fraction of sp³. The new carbon materials are semiconducting with a bandgap range of 1.5–2.2 eV, which are smaller than that of diamond and comparable to that of the widely used amorphous silicon. These materials also show good mechanical properties such as high highness and strength, such that the newly synthesized carbon material can even scratch the (001) surface of diamond. The same year, S. Zhang et al. [95] obtained some other semiconducting amorphous carbon materials with band gaps of 0.1-0.3 eV and isotropic super hardness and toughness. Shang et al. [96] successfully synthesized millimeter-sized, transparent, and nearly pure sp³ (up to 97.1%) amorphous carbon for the first time, as shown in Figure 5(b). This material consists of many randomly oriented clusters with diamond-like short-/medium-range order and possesses the highest hardness, elastic modulus, and thermal conductivity among all the known amorphous material. These materials also exhibit a wide range of tunable optical bandgap from 1.85 to 2.79 eV.

The key point of phase engineering is to modulate material properties by phase synthetization or transformation. Simulation work can help a lot during this process, such as the prediction of potential material structure and property, the design and optimization of experimental parameters (pressure, temperature, the choice of ­precursor, etc.). Q. Li et al. [97] reported a monoclinic carbon allotrope named “M-carbon” by ab initio ­simulation. The calculated hardness and bulk modulus of this “M-carbon” were comparable to diamond, and the simulation result was consistent with experimental work [98]. X. Yang et al. [99] predicted a carbon allotrope by Crystal Structure Analysis by Particle Swarm Optimization (CALYPSO) method. This fully sp³-bonded carbon material was named “V-carbon.” Simu­lation results indicated that the material had a hardness of 90 GPa and bulk modulus 400 GPa, which was verified by experimental work.

3.4. Strain engineering

Although element doping, surface engineering, and phase engineering have achieved some progresses in tuning diamond’s electrical properties, there are still some limitations as discussed above. Strain engineering is emerging as a new strategy having some advantages to overcome or bypass these limitations.

For a defect-free crystal, the maximum elastic strain that it can sustain at 0 K without losing homogeneity is its ideal strain, which is around 10% [100]. However, the actual elastic strain of most bulk material is 2–3 orders of magnitude lower than the theoretical value due to the unavoidable defects existing within the material. When the geometrical size of the material decreases, the probability of the presence of defect inside the material reduces as well. As a result, the elastic strain of the material becomes closer to its ideal strain before the plastic yield appears. Therefore, nanometer materials can generally sustain far more elastic strain than the materials at macro size. This expands the tuning space for elastic strain engineering (ESE).

3.4.1. Growth-induced strain

Many experimental and theoretical works [101–105] have proved that the physical and chemical properties of materials can be improved or fundamentally changed by ESE. Strained Si technology is a typical example and has brought billions of dollars of product value. In this case, tensile strain is applied to a 10–100 nm width Si channel and the carrier mobility of Si is improved dramatically as a result [106]. One common way to obtain strained Si is by heteroepitaxially growing Si on a suitable substrate to induce lattice mismatch. As we can see in Figure 6(a,b), compressive or tensile strain can be introduced to the silicon film as the Si lattice is compressed or stretched to align with the underlying substrate material which has a different lattice constant [107,108].

Figure 6. The introduction of (a) compressive or (b) tensile strain to silicon by using different substrates [107]. (c) TEM image of GaN/diamond interface [113]. (a) and (b) Reproduced with permission from ref 107, Copyright 2004 IEEE. (c) Reproduced with permission from ref 113, Copyright 2020 American Chemical Society.

We wish to borrow the successful experiences of growing strained Si film and obtain locally or wholly strained diamond. Several experimental and theoretical work have demonstrated that the electrical properties of diamond can be modulated with elastic strain applied [109–112]. The research of heteroepitaxial growth of diamond has been conducted extensively as this method has the potential to produce large-size and high-quality diamond. A series of materials have been used as the substrates for diamond heteroepitaxial growth, including GaN, Si, sapphire and cubic boron nitride [113–120]. As the lattice constants of these materials are different from that of diamond, local stress/strains are introduced, especially at the interface between diamond and substrates. Figure 6(c) is the TEM image of GaN/diamond interface, the local strain/stress is introduced due to the different lattice constants of GaN (a = 318 pm, c =518 pm) and diamond (a = 357pm) [113]. At present, scientists are focused on the growth of diamond with reduced or eliminated interfacial stress in order to obtain high-quality and large-size diamond. Such undesired stress may be used for generating a fixed value of strain on diamond. More work needs to be conducted to obtain diamond with expected strain value.

3.4.2. Elastic strain engineering (ESE)

Diamond is generally thought to be hard, brittle and nearly non-deformable [121]. Deforming a diamond usually means irreversible damage to it. Several methods have been tried to investigate the mechanical properties of diamond since 1930s, including indentation and diamond anvil cell compression [122,123], as shown schematically in Figure 7(a,b). However, the deformation of diamond was small and the strain distribution was nonuniform. The research of diamond’s mechanical properties came to a standstill since then.

Figure 7. Experiment progresses on mechanically straining diamond. The development of diamond deformation strategy, from bulk-size diamond deformation ((a) indentation and (b) DAC compression) [122] to micro-size diamond deformation ((c) nanopillar compression, (d) & (e) nanoneedle bending, (f) microbridge and micro-array stretch) [109,110,124,129]. With the development of deformation strategy, larger and more uniform elastic deformation are being realized. (a) Reprinted courtesy of the National Institute of Standards and Technology, U.S. Department of Commerce. Not copyrightable in the United States. (c) Reproduced with permission from ref 124, Copyright 2015 American Chemical Society. (d) Reproduced with permission from ref 109, Copyright 2018 the author(s), published by American Association for the Advancement of Science (AAAS). (e) Reproduced with permission from ref 129, Copyright 2019 the author(s), published by Nature Communications. (f) Reproduced with permission from ref 110, Copyright 2021, American Association for the Advancement of Science.

Recently, with the development of advanced micro/nano fabrication and mechanical testing techniques, it becomes feasible to conduct mechanical testing for micro/nano diamond samples. A series of tests including nanopillar compression, nanoneedle bending, microbridge, and micro-array stretch [109,110,124–127] have been conducted for small-size diamond samples. Elastic strain or plastic strain of diamond was observed in these works.

In 2016, J. Wheeler et al. conducted uniaxial compression for micro/nano size diamond pillars for the first time [124]. SEM images of diamond nanopillars were observed successfully before and after compression, as shown in Figure 7(c). The diamond pillars showed elastic deformation without obvious plastic deformation before the brittle fracture. A maximum shear strength of 75 GPa was measured, which was close to the theoretical yield stress limit of diamond. However, the elastic deformation of diamond was small and mainly concentrated on the top area of the micro pillar.

A. Banerjee et al. [109] conducted side bending test of diamond nanoneedles. The bending deformation of a nanoneedle during loading process is shown in Figure 7(d). The researchers discovered ultralarge, fully reversible elastic deformation of nanoscale diamond needles for the first time. Simulation result indicated that a local (tensile) maximum principal strain of 8.88% was obtained on the tension side of the bent needle, which was much higher than the previously reported bulk diamond elastic limit [128]. A. Nie et al. further investigated the elastic limit of diamond nanoneedle [129]. The TEM image of diamond needle before and during compression is shown in Figure 7(e). Their research results showed that <100>-oriented nanoneedles with a diameter of 60 nm exhibited highest elastic tensile strain (13.4%) and tensile strength (125 GPa). These values are comparable with the theoretical elasticity and Griffith strength limit of diamond, thus open up the infinite possibility to modulate diamond’s electronic properties through nanomechanical straining.

3.4.3. Deep elastic strain engineering (DESE)

The realization of ultralarge elastic strain in nanoscale diamond has a significant meaning. On one hand, this opens the door for potential deep ESE (DESE) of diamond, that allows unprecedented properties under extreme strained states. On the other hand, it encourages theoretical calculations of diamond DESE as ultralarge deformation and even complicated coupled deformation modes of diamond were often considered to be impossible before.

Experimentally, the ultrahigh elastic strains achieved during the diamond nanoneedle bending tests were not uniform and limited in a relatively localized region. Moreover, bent nanoneedles have a complex strain condition that one side of the nanoneedle is under tensile strain while the other is under compressive strain. To solve these problems for practical device applications, C. Dang et al. [110] fabricated diamond microbridges and microbridge arrays and conducted in-situ uniform tensile tests, as shown in Figure 7(f). The sample size and deformation mode in this experiment were comparable to the actual semiconductor application. The diamond microbridge completely recovered to its original length after strain values of up to 7.5% in cyclic tensile tests. With some improvements, maximum tensile strain of up to 9.7% was achieved before fracture happened. This value approaches the ideal elastic limit [130]. Such a realization of large and uniform elastic strain in diamond microbridge and array is unprecedented. They also demonstrated a bandgap reduction of diamond under tensile strain by first principles calculation and confirmed it by electron energy-loss spectroscopy (EELS), which is important for diamond’s emerging DESE application in future semiconductor field.

On the simulation/modeling side, more possibility can be imagined for a diamond strained in a 3D space with unconventional behavior. However, for conventional computational power, it is hard to design the six-dimensional elastic strain tensor ε (ε₁ ≡ ε₁₁, ε₂ ≡ ε₂₂, ε₃ ≡ ε₃₃, ε₄ ≡ ε₂₃, ε₅ ≡ ε₁₃, ε₆ ≡ ε₁₂) through the most optimal combination of diamond’s normal and shear components due to the vast continua of choices available. In order to deal with this challenge, Z. Shi et al. [131] developed a method combining artificial intelligence and multiscale modeling to find out the most optimal pathway for the ESE of diamond. The results demonstrated that by straining diamond in the most optimal way, diamond can be transformed to a state with lower bandgap while its own uniqueness such as high strength and high thermal conductivity are preserved at the same time. Z. Shi et al. [132] improved their simulation strategies by combining machine learning method and a physics-informed, convolutional neural network technique. The achievable bandgap values of diamond through ESE are shown in Figure 8(a). Comparing to their previous work [131], intra-band and inter-band correlation are introduced for calculating energy band and the strategy is believed to be more accurate. This strategy is also more versatile such that a broader range of physical characteristics are considered, such as the mass of electrons and holes which are strong sensitive to noise. The simulation results of C. Liu et al. [111] indicated that when diamond is subjected to constrained shear (CS) strain condition, the stress response shows smooth structural flow characterization which is unprecedented in diamond. The anomalous structural deformation of diamond also has profound impacts on its electronic properties [133]. The bandgap of diamond reduced gradually under CS loading condition and a band gap closure appears at surprisingly modest CS stress condition, leading to diamond’s metallization.

Figure 8. Deep ESE simulations of diamond under straining. (a) Achievable bandgap values of diamond through ESE [132]. (b) Simulation prediction of the local strain distribution (left and middle needles, with compressive and tensile strains distribution) and bandgap distribution (right needle) of bent diamond nanoneedles [112]. (c) Conduction charge distribution (left figure) and electronic band structure (right figure) of diamond lattice under a CS strain ε value of 0.566, which is under superconducting condition [135]. (a) Reproduced with permission from ref 132, Copyright 2021 the author(s), published by Nature Publishing Group. (b) Reproduced with permission from ref 112, Copyright 2020 the author(s). Published by PNAS. (c) Reproduced with permission from ref 135, Copyright 2020 American Physical Society.

The simulation results indicated that when the applied strain reaches a certain level, fundamental changes will happen for diamond’s properties, such as metallization and superconductivity. Z. Shi et al. [112] demonstrated that it is possible to achieve metallization in diamond exclusively through the imposition of reversible elastic strains, without triggering phonon instability or phase transition from diamond to graphite. As can be seen in Figure 8(b), for <110>-oriented diamond nanoneedles bending along $\left[1\overline{1}0\right]$ direction, “safe” metallization can be realized on the local area of compression side with a compressive strain value of ∼10.8% which is less than the theoretically approachable maximum compressive strain of 14.5%. Also, the elastic strain energy density values for “safe” metallization are comparable to what has been demonstrated in experiments of reversible deformation of diamond nanoneedles [109,129]. Diamond can turn into a superconductor with a critical temperature T_c ≈ 4 K near 3% boron doping [134]. C. Liu et al. found that CS loading conditions may affect the superconductivity of diamond as well [135]. The simulation showed that diamond maintained a stable structure under the maximum CS strain of 0.566. The conduction charge distribution and electronic band structure of diamond lattice under this strain value are shown in Figure 8(c). The simulation results indicated that superconductivity occurred when the CS strain applied on diamond reached 0.349. A CS-strain driven critical temperature (T_c) of 2.4–12.4K can be reached for Coulomb pseudopotential μ* = 0.15–0.05. These surprisingly high T_c values under realistic μ* place deformed diamond among the most prominent elemental superconductors [136].

3.5. Defect engineering

Diamond often exists with defects such as vacancy, impurity, and structural dislocation. Although these defects make diamond not perfect crystal, they can be useful sometimes for some specific applications, if well controlled. The tuning of material property by adjusting/introducing specific defect in it can be considered as defect engineering. One of the most common impurities existing in diamond is N. In this section, we want to introduce a special N-related defect, nitrogen vacancy (NV) center.

NV center is composed of a substitutional N atom and an adjacent vacancy, as schematically shown in Figure 9(a) [137]. The spin state of NV center can be initialized or read-out by using suitable laser. The electron spin can be manipulated by external factors (e.g. microwave, magnetic and electric field) at room temperature. These unique properties give diamond great application potentials in quantum information and sensing. As NV center is an ideal solid qubit under room temperature, if NV center can be generated and manipulated, it is possible to build room-temperature quantum computers [138,139]. NV can also serve as an atomic-scale sensor with high sensitivity [140–142]. As we can see in Figure 9(b), in biological field, the nanoscale magnetic imaging of ferritins can be realized in a single cell by using the NV center as the sensor [140]. In future, the doctors may use NV center to diagnose the lesion of our body at early stage.

Figure 9. Defect engineering. (a) Schematic of NV center diamond [137]. (b) Magnetic imaging of ferritins in a single cell by NV center [140]. (c) Schematic of manipulating single NV spin by laser [14]. (a) Reproduced with permission from ref 137, Copyright 2021 by the authors. Licensee MDPI, Basel, Switzerland. (b) Reproduced with permission from ref 140, Copyright 2019 the author(s), published by American Association for the Advancement of Science (AAAS). (c) Reproduced with permission from ref 14, Copyright 2014 the author(s), published by Nature Communications.

As a nanoscale sensor, NV center is sensitive to the surrounding environment (magnetic, electrical, and strain field, etc.) [14,143,144]. It is possible to modulate the properties of NV center by mechanical strain and researchers have obtained some interesting results. S. Bennett et al. [145] proposed a strategy to realize phonon-induced spin-spin interactions by coupling them via strain introduced by vibrational mode of a diamond mechanical nanoresonator. The spin dephasing and relaxation are depressed significantly by generating ­substantial squeezed states of a spin ensemble. P. Ovartchaiyapong et al. [14] fabricated diamond cantilever with an embedded nitrogen-vacancy center. As shown in Figure 9(c), the couple of mechanical strain and the spin of the NV center is realized by the mechanical motion of the diamond cantilever. The NV center behaved as an atomic-scale sensor and the spin-based strain imaging reached a strain sensitivity of 3*10⁻⁶ strain Hz^−1/2. A. Barfuss et al. [146] presented a new approach to coherently manipulate a single electronic spin via internal strain. Time-varying stain was applied on a diamond cantilever to generate long-lasting, coherent oscillations of a NV center spin. The phonon-dressed states were observed by direct spectroscopy and the spin coherence time of the NV center was enhanced significantly by the continuous strain.

Overall, several strategies, from the conventional element doping to the ESE for modulating the electrical properties of diamond are reviewed in this section. These different methods are summarized in Table 1 to help readers have an overview. It is worth mentioning that none of these methods can fully use the outstanding physical properties of diamond and more work is needed to explore and accelerate its application in semiconductor devices.

Table 1. Summarization of all the modulation strategies discussed in this work.

Strategies	Advantages	Disadvantages	Current situation
Element doping	Uniform doping; fewer defects; effective p-type doping	No effective n-type doping; cannot achieve precise doping	n-type doping needed
Surface engineering	High p-type mobility	Cannot achieve n-type doping	n-type doping needed
Phase engineering	Changing properties fundamentally	Not suitable for mass production	Needs more exploration
ESE	Tuning electrical properties reversibly	Hard to maintain strain	Strain maintain needed; combination with element doping needed
NV center	Sensitive	Limited application fields	Quantum information

3.6. Electronic device applications of diamond

Nowadays, diamond shows increased application potentials in high frequency/high power semiconductor due to its outstanding properties (wide bandgap, high thermal conductivity, high breakdown field strength, etc.). M. W. Geis et al. [147] successfully fabricated Schottky diode based on B-doped diamond. The maximum work temperature of this device was 700 °C, far beyond that of SiC device (550 ∼ 600 °C). T. Zimmermann et al. [11] introduced a nitrogen doped ultra-nano-crystalline diamond layer into diamond diode. The device showed an extraordinary thermal stability, which worked well up to 1050 °C in vacuum. Besides high working temperature, diamond device can also sustain high electrical field strength. N. Tatsumi et al. [148] manufactured Pt/diamond Schottky diode with a breakdown voltage of 3.1 MV/cm, which was higher than its SiC counterpart.

Diamond also shows great potential in radiation detection field. The inherent properties (ultrawide bandgap, tight crystal lattice, and high covalent bonding energy) of diamond lead to its low radiation damage, rapid charge collection time. A. Balducci et al. [12] successfully increased the charge collection distance from 50 to 300 μm for micro-strip detector and pixel array detector made by diamond. S. Lagomarsino et al. [149] applied 3D concept to diamond detector. The charge collection efficiency reached 95% for their single crystal 3D diamond detector under a low voltage of 3 V and the device just worked fine after radiation.

4. Summary and outlook

As the hardest material in nature, diamond has been widely used in advanced manufacturing and tooling industry. Recent progresses in fabricating high-quality single crystalline diamond also bring unprecedented opportunities to open up functional device applications of diamond as a wide bandgap semiconductor, due to its outstanding electrical properties and figures of merit. The increasing demand of electronic grade diamond pushes forward the research and development (R&D) of synthetic diamonds, especially the latest CVD technique for growing wafer-scale, electronic-grade SCD.

Intrinsic diamond without doping shows electric insulation behavior due to its ultra-wide bandgap. The electrical properties of diamond need to be thus modulated before it being applied into electronic devices. A series of strategies including element doping, surface modification, phase engineering and strain engineering have been developed by researchers, and are reviewed in this paper. The device application of diamond is introduced as well. In addition, the specific properties of NV center in diamond bring a bright future for diamond’s application in quantum information and quantum sensing applications.

Lastly, although diamond has a lot of potentials as a promising electronic material, there are many challenges to be addressed. For the growth of large-size (up to wafer scale) and high-quality diamond, the challenges include improving crystal quality (single crystallinity), reducing the surface roughness and internal flaws, increasing the growth rate without compromising the size and quality, and developing uniform epitaxial growth techniques for strain engineering and device integration. For the development of diamond planar processing technology for strain-engineered devices, some difficulties include the control of crystal orientation, precise geometric shape, and surface morphology and topography. While for the realization of efficient n-type doping of diamond with high electrical conductivity is another challenge for diamond’s microelectronics application. As these problems are gradually going to be addressed with the rapid development in this field, we will witness the rising of functional diamond applications in the near future.

Abbreviations
B	=	boron
CALYPSO	=	Crystal Structure Analysis by Particle Swarm Optimization
CS	=	constrained shear
CVD	=	chemical vapor deposition
DAC	=	diamond anvil cell
DESE	=	deep elastic strain engineering
EELS	=	electron energy-loss spectroscopy
ESE	=	elastic strain engineering
hBN	=	hexagonal boron nitride
HPHT	=	high pressure and high temperature MPCVD micro plasma chemical vapor deposition
NV	=	nitrogen vacancy
P	=	phosphorus
PCD	=	polycrystalline diamond
R&D	=	research and development
SCD	=	single crystalline diamond
SEM	=	scanning electron microscope
TEM	=	transmission electron microscope
UV	=	ultraviolet.

Acknowledgments

We acknowledge the support from the Research Grants Council of the Hong Kong Special Administrative Region, China, under RFS2021-1S05 and the National Natural Science Foundation of China under project 11922215. H.Z. acknowledges the funding from the National Natural Science Foundation of China (Grant 11902200) and the Science and Technology Commission of Shanghai Municipality (Grant 19YF1433600). Also, we thank Drs. Maoyuan Li and Yuan Hou for fruitful discussions.

Disclosure statement

No potential conflict of interest was reported by the authors.

Funding

This work was supported by Research Grants Council of the Hong Kong Special Administrative Region, China, under RFS2021-1S05 and the National Natural Science Foundation of China under project 11922215 and Grant 11902200 and the Science and Technology Commission of Shanghai Municipality (Grant 19YF1433600), University Grants Committee, and Shanghai Science and Technology Committee.