Diamond Electronic Devices

11 downloads 0 Views 134KB Size Report
RD, which is an important figure of merit, is given by [17]: ... S. Koizumi, and M. Suzuki, Physica Status Solidi (A) Applications and Materials 203, 3358-3366 (2006). ... T. Makimoto, M. Schwitters, D. J. Twitchen, G. A. Scarsbrook, and S. E. Coe,.
Diamond Electronic Devices J. Isberga a

Division for Electricity, Uppsala University, Box 534, S-751 21 Uppsala, SWEDEN

Abstract. For high-power and high-voltage applications, silicon is by far the dominant semiconductor material. However, silicon has many limitations, e.g. a relatively low thermal conductivity, electric breakdown occurs at relatively low fields and the bandgap is 1.1 eV which effectively limits operation to temperatures below 175 ˚C. Wide-bandgap materials, such as silicon carbide (SiC), gallium nitride (GaN) and diamond offer the potential to overcome both the temperature and power handling limitations of silicon. Diamond is the most extreme in this class of materials. By the fundamental material properties alone, diamond offers the largest benefits as a semiconductor material for power electronic applications. On the other hand, diamond has a problem with a large carrier activation energy of available dopants which necessitates specialised device concepts to allow room temperature (RT) operation. In addition, the role of common defects on the charge transport properties of diamond is poorly understood. Notwithstanding this, many proof-ofprinciple two-terminal and three-terminal devices have been made and tested. Two-terminal electronic diamond devices described in the literature include: p-n diodes, p-i-n diodes, various types of radiation detectors, Schottky diodes and photoconductive or electron beam triggered switches. Three terminal devices include e.g. MISFETs and JFETs. However, the development of diamond devices poses great challenges for the future. A particularly interesting way to overcome the doping problem, for which there has been some recent progress, is to make so-called delta doped (or pulse-doped) devices. Such devices utilise very thin (~1 nm) doped layers in order to achieve high RT activation. Keywords: CVD diamond, Single crystalline diamond, delta doping, diamond devices. PACS: 72.20.Fr, 72.20.Jv, 72.80.Cw, 73.40.Lq

DOPING DIAMOND The desire for electronic devices with higher power throughput, wider frequency bandwidth and higher operational temperatures is driving research and development in WBG materials. Diamond is extreme in this group of materials that includes silicon carbide (SiC) and gallium nitride (GaN). Judging by the physical properties of diamond, which include high carrier mobilities and high breakdown field, diamond electronic devices, such as power diodes and high-frequency field effect transistors, can be expected to deliver outstanding performance. TABLE 1. Material properties for Si, 4H-SiC, GaN and diamond. Si 4H-SiC GaN 1.1 3.2 3.4 Band gap (eV) 1.5 5 1.3 Thermal conductivity (W/cmK) 480 120 200 Hole mobility (cm2/Vs) 1450 900 440 Electron mobility (cm2/Vs)

Diamond 5.5 24 3800 4500

Undoped diamond is an excellent insulator due to the wide band-gap of 5.47 eV. The intrinsic carrier concentration is very low at room temperature and consequently the resistivity of pure diamond can exceed 1016 Ωcm [1]. Dopants in WBG semiconductors tend to have higher ionisation energies than in narrow-bandgap semiconductors, resulting in low activation at room temperature. For example, 4-H SiC has shallow donors (ndopant) but lacks a really shallow acceptor (p-dopant), the shallowest being Al with ionisation energy of 0.19 eV [2]. In the case of diamond known dopants have even higher ionisation energies. Boron (B), which is an acceptor in diamond, has an ionisation energy of 0.37 eV, and extrinsic p-type conduction can be observed at doping concentrations below 3·1019 atoms/cm3. At very high boron concentrations the activation energy approaches zero and conduction becomes metallic. A resistivity of about 10-3 Ωcm at 300K can be achieved for boron concentrations above 1021 cm-3 [3]. Extrinsic n-type conductivity in diamond can also be achieved [4-6], e.g., by doping with phosphorus (P) with an ionisation energy of 0.52 eV. If both acceptors and compensating donors are present with concentrations NA and ND, respectively, and with NA> ND, then in a non-degenerate semiconductor the hole concentration p can be calculated from

p( p + N D ) − ni2 N = V exp(− E A / kT ) 2 N A − N D − p − ni / p g a

(1)

where NV the valence band effective density of states, ni is the intrinsic carrier concentration, ga the spin degeneracy factor, k the Boltzmann constant, EA the acceptor ionisation energy, and T the absolute temperature. In this expression the influence of excited impurity states have been neglected. Expressions for more complex models, including several acceptor and donor levels and excited states, can be found in e.g. [7]. Figure 1 below shows the hole concentration in boron doped diamond at room temperature for different doping concentrations and different compensating donor concentrations. The degree of boron activation at 300 K is plotted in Figure 2, which shows the hole concentration for different acceptor and compensating donor concentrations. However, if the boron concentration is increased above 1019 cm-3 and the insulator-metal transition is approached, the activation increases rapidly to reach almost complete activation [1, 3, 8].

FIGURE 1. Hole concentration vs. dopant concentration of boron doped diamond at room temperature for different concentrations of compensating donors. At higher boron concentrations higher than 1019 cm-3 the activation increases rapidly to full activation and conduction becomes metallic-type.

FIGURE 2. Hole concentration vs. temperature of boron doped diamond with dopant concentration NA =4·1016 cm and compensating donor concentration ND =2·1014 cm-3. -3

The lack of known shallow dopants in diamond has led some investigators to conclude that the prospects of using diamond for electronic devices are poor. It is true that the lack of known shallow dopants restricts the capability of diamond, but on the other hand the extreme material properties of diamond could expand the capability in other directions. Therefore, conventional device designs cannot automatically be expected to work well in diamond. Instead a more creative approach is required and one should develop new types of devices exploiting the material advantages of diamond, rather than to adopt conventional device designs. Examples of this approach are delta-doped FETs [9] and Schottky diodes with intrinsic layers [10]. In these types of devices the charge carriers are supplied from a highly doped region, while the active region consists of undoped diamond where carrier mobilities are high. For devices operating at elevated temperatures high dopant ionisation energy is less of a problem and conventional designs work better. In some high-end power electronic applications there is also the option to run devices at elevated temperatures, if it improves device performance.

DIAMOND DEVICES In many device applications a semiconductor material with a high electric breakdown field is desirable. This is true not only for power devices, such as diodes and switches intended to block several kilovolts, but also for high frequency field-effect transistors. The reason being that if higher electric fields can be tolerated, the devices can be designed with shorter gate lengths, which results in faster switching. Intrinsic breakdown in semiconductors is inherent to the material. It results from impact ionisation and subsequent avalanche breakdown. On the other hand, extrinsic breakdown at defects is dependent on crystalline quality and improves with better quality material. Diamond exhibits the highest breakdown field (Ebr) of any semiconductor (or indeed any insulator). Values of Ebr in the range 10-20 MV/cm, have been reported. To reach high electric fields in the active regions of devices, it is necessary to use clever device designs to reduce undesirable field enhancement. Such field enhancement can occur at junctions in the semiconductor and also at interfaces, e.g. with a surrounding dielectric. It is therefore necessary to develop efficient edge termination concepts for diamond. The lack of a shallow n-dopant in diamond presents an added difficulty here. However, the dielectric constant of diamond, εr = 5.7, is one of the lowest among semiconductors, about half of the values in Si, SiC or GaN. This is an advantage for diamond when it comes to field control at interfaces with a surrounding dielectric. This is because a difference in dielectric constant between two materials results in a field enhancement in the material with the lower εr, which is usually the dielectric. Thus, with its low εr, diamond is better matched to common dielectrics such as SiO2 (εr =3.9) than most other semiconductors.

Three Terminal Devices The first transistor in diamond was reported in 1981 by Prins [11], who demonstrated a bipolar n-p-n transistor made from p-type natural diamond. The two n-type regions were made by carbon ion-implantation, creating zones with a high defect concentration, exhibiting n-type behaviour. However, the current gain in this transistor was very low. More promising, both for RF and power applications are diamond unipolar devices, such as MISFETs (metalinsulator-semiconductor field effect transistor), MESFETs (metal epitaxial semiconductor field effect transistor) and JFETs. In a conventional FET device current transport takes place in a channel below the gate between the source and drain contacts. The channel current is limited to qpdvsat, where p is the hole concentration, d is the channel thickness and vsat the hole saturation velocity. In order to reach a high current, a high carrier concentration is thus necessary. Due to the high ionization energy of boron-doped diamond, very high doping concentrations are required, which results in material with a high concentration of scattering centres and consequently poor transport properties. Therefore conventional FET designs cannot be expected to yield RF devices with good performance. For this reason, other device designs which make use of the superior transport properties of intrinsic diamond have been suggested. Several groups have made use of so-called ‘surface transfer doping’ where a hole channel is created near the surface by hydrogen termination of the diamond surface. Impressive progress has been made by several groups, which include a drain current density of 650 mA/mm and a cut-off frequency of 42 GHz [12], 2.1 W/mm at 1 GHz for a

gate width of 100 µm [13], cut off frequency fT=45 GHz and fmax=120 GHz [14]. Although these results are promising, the long-term stability of such surface transfer doped FETs remains an issue, especially when these devices are used at high temperatures. Another design, which do not suffer from stability problems, utilizes so-called ‘delta doping’ in order to achieve a spatial separation between ionized acceptors and holes. In such devices the delta layer is a thin (a few nanometres), highly doped layer surrounded by intrinsic diamond. The idea is that conduction mainly occurs in the intrinsic layers (with good transport properties), while the charge carriers are supplied by diffusion from the narrow delta-doped layer (where ionization is almost complete). This technique was first developed in silicon devices and it is today used in III–V semiconductor devices. Creating nanometre-thin delta layers in diamond presents difficult synthesis challenges. Principal among these is the requirement to prepare atomically abrupt defect-free interfaces between intrinsic diamond and the doped layer. Nevertheless, some progress in synthesising such layers have been reported and small-signal RF measurements have been performed on a ‘proof of concept’ diamond delta-doped MESFET device[15]. This device exhibited a cut-off frequency, fT above 1 GHz, and a maximum frequency of operation, fmax of 4 GHz. It seems feasible that further improvements in the delta-layer synthesis should yield substantial improvements in device performance possibly leading to commercially attractive devices [16].

Two Terminal Devices In power diodes it is desirable to combine a high breakdown voltage (Vbr) with a low specific on-state resistance of the drift layer (RD). There is a trade off between these two properties depending on the choice of doping concentration and thereby the depletion layer width at breakdown. For Schottky barrier diodes (SBDs) the ratio Vbr / RD, which is an important figure of merit, is given by [17]:

Vbr 1 = qEbr µ p p RD 2

(2)

where Ebr is the breakdown field, µp the hole mobility and p the hole concentration in the drift layer. This expression assumes a homogeneous p-doping in the drift layer and neglects field enhancement effects at the contact edge. As mentioned previously, diamond has a very high breakdown field and also a high hole mobility. However, due to the incomplete activation of the shallowest known dopant, boron, the hole concentration is much lower than the doping concentration (i.e. p