ESTRUCTURA FAMILIAR DE LA COMUNIDAD CAMPESINA DE

Fig. 2.6 Conduction band diagram for a AlGaAs/GaAs MODFET

Since it was first investigated by Dingle et al. (1978) and Stormer et al. (1979) the MODFET has been the subject of a great deal of research, principally as a result of the extremely high carrier mobilities and novel physics (Harris et al. 1989) associated with such structures. Operating in a similar manner to the MOSFET (and the MESFET), the MODFET has a number of important advantages over its ubiquitous relative. The device is formed at the interface of two materials of differing electron affinity, most commonly GaAs/AlGaAs. The wide bandgap material (AlGaAs) is doped while the smaller bandgap (GaAs) is left nominally undoped (see Fig. 2.6). Due to the conduction band discontinuity at the interface (AEc) electrons migrate into the lower band gap material. The resulting band bending, caused by the high electric fields (~10^ V /cm in GaAs), and the conduction band discontinuity confine the free carriers in a Q2DEG. The confinement in the growth direction (z) is sufficient to cause quantisation of the electron energy (Solomon and Morkoc 1984), resulting in discrete

quantum-electric subbands being formed i.e. the carriers become governed by quantum mechanics. Mobility enhancement in these structures arises from a reduced ionised impurity scattering rate due to the spatially separated electrons and donors (Stormer et al. 1983). The addition of an undoped spacer layer further reduces this Coulomb interaction between the charged particles. Record mobilities of up to 1.2x10^ cm^/Vs (Pfeiffer et al. 1989) have been achieved at low temperatures (0.35K) for optimised structures (see Table 2.2). The high mobilities obtained in these structures have led to predicted values of fj up to 400GHz (Nguyen 1992).

Stormer (1983) Pfeiffer (1989) Schubert (1987) Hoke (1992) Brown (1988) material G aA s/ A lG aA s GaAs/ AlG aAs G aA s/ A lG aA s A lG a A s/ InGaAs^ A U n A s/ InGaAs n2D [xl0^2] 0 .1 0.24 1.5 3 3.4 P [cm^/Vs] 2x105(4K) 6.8x106(4K) 8900(300K) 7100(300K) 9500(300K) gate length [pm] ---- — 1 .2 0.25 0.3 gm [mS/mm] --- — 360 300 650 fm ax [GHz] --- — — 80 >80

However, the drawback of the MODFET structure is that although it can possess extremely high mobilities - and hence operating frequencies, this is at the expense of the carrier concentration i.e. the current in the Q2DEG - and thus the power output. The maximum carrier concentration is limited by three separate mechanisms. The first, and most significant mechanism, was studied by Witkowski et al. (1981). They examined the relationship between the carrier concentration, the mobility and spacer layer thickness and found, as one may expect, that increasing the spacer layer thickness increased the m obility, due to the decreased ionised im purity scattering, but decreased the carrier concentration. Hence structures optimised for high mobilities, i.e. having large spacer layers and

low doping densities, have carrier concentrations far below those required for FET devices (see Table 2.2). Even in structures optim ised for maximum charge transfer, the highest achievable n2o is -IxlO^^cm"^ for GaAs (Harris et a l 1986), this value is ultimately limited by the conduction band offset AEc at the heterojunction. 3-doping (Schubert et a l 1987, Shieh

et a l 1993a, 1993b ,Wu et a l 1994, Kudo 1994, Ploog et a l 1987) and puise doping (Hoke et a l 1992, Moll et a l 1988 & Roblin et a l 1989) of the barrier regions has been w idely used in an attempt to raise the maximum n2D- The highest carrier concentrations achieved for MODFETs have been 1.5xlQi2 cm"2 for A lG aAs/G aAs, 3x10^2 cm"2 for pseudom orphically strained AlGaAs/InxGai-xAs, utilising the larger AEc, and 3.4x1 0 ^ 2 cm‘2 for Ino.5 3Gao.4 7A s /A lo .4 8lno.5 2As (see Table 2.2). These techniques rely on increasing the doping levels in the wide bandgap material in an attempt to m aximise the charge transfer to the QW. Eventually, as the dopant concentration in the barrier increases, more carriers stay in the potential w ell formed by the dopant and the amount of charge transferred to the w ell saturates (Jogai et a l 94). Though these techniques have been relatively su ccessfu l they have yet to attain the sam e Q2DEG concentrations that are possible in doped channel structures with Zrenner

et a l (1988c) using 3-doping to achieve a m axim um free electron concentration of 1.2x10^^ cm"2 (§3.4).

In addition to the charge transfer limitations the maximum U2d is restricted by a second mechanism - the DX centre. This defect is related to the Si dopant atoms (see §3.6) and its energy relative to the conduction band (CB) edge is dependent on the A1 content of the material (Chadi & Chang 1989). For AlxGai_xAs with x>0.22 the DX centre is below the CB edge and so free electrons can become trapped in these states, decreasing the number of carriers in the channel, and resulting in noise and current instabilities. It is therefore advantageous to place the Si dopant in material

w ith a low A1 content to reduce DX centre trapping, however this limits AEc and hence n2D- In MODFETs, as we have seen, the dopant is normally located in the A1 rich material therefore DX centre trapping is a common trapping mechanism in these structures.

The third limiting mechanism, real space transfer, only occurs at the high fields found under the gate contact in an operating transistor device. Electrons in the conducting channel gain energy from the electric field and become hot; if they attain sufficient energy they can 'jump' into the barrier material. If we examine Fig. 2.6 w e can see that any hot- electrons injected into the AlGaAs barrier will be swept away by the field in the region (Littlejohn et a l 1983, Keever et a l 1981) thus decreasing the number of carriers in the well. In a theoretical treatment Foisy et a l

(1988), have determined maximum drain currents of ~250m A /m m and 350m A /m m for two identical (Ipm) AlG aAs/G aAs and pseudomorphic InGaAs / AlGaAs MODFET structures.

As w e have seen, MODFETs can achieve extremely high carrier mobilities, however these only occur at low fields, at higher fields they are dramatically reduced (Schubert et a l 1986). Masselink (1986) show s that for a series of MODFETs with a large range in zero-field (77K) mobility, the effective mobility for all the devices has decreased to approximately the same value by 1 kV /cm (typical of the fields found in FETs). Thus achieving large low-field mobilities is perhaps less important than other factors in designing high performance power FET structures (Inomata et a l 1986). The mobility cannot be totally ignored how ever as it has a significant impact in other areas of transistor operation:

• In reducing the access resistance [2.6] and thus the parasitic effects of the source region (Fig. 2.1), a high mobility can improve the extrinsic

• It helps reduce the saturation voltage by reducing the field required for electrons to reach their saturation drift velocity (Fig. 2.4).

Transconductances of between 275 and 1250 m S/m m (Morkoc and Solom on 1984, N guyen 1992) have been achieved for high m obility MODFET structures. However, the reduction in the gate-to-channel sp acin g required to achieve these transconductance valu es, has significantly reduced both the breakdown voltages (typically between - 2 and -5V) and the current densities (n%D < 0.5x10^^ cm“2) in these devices.

In conclusion although extremely high mobilities can be achieved in MODFETs there is evidence that this is perhaps of secondary importance to factors like carrier concentration, carrier confinement and gate-channel separation (Inomata et a l 1986) which define power output, transconductance and break dow n voltage and fm ax- Thus w e can see

although the MODFET is an extremely important device and has had som e success as a pow er FET, it has lim ited possibilities for the developm ent of the next generation of high performance pow er FET device structures.