High-electron-mobility transistors (HEMTs) and heterojunction bipolar transistors (HBTs) have attracted many attentions in high speed and power applications due to the superior transport properties. As compared to AlGaAs pseudomorphic HEMTs (PHEMTs), InGaP-related devices have advantages, such as higher band gaps, higher valence-band discontinuity , negligible deep-complex (DX) centers , excellent etching selectivity between InGaP and GaAs, good thermal stabilities [3-5], higher Schottky barrier heights , and so on. Particularly, the use of an undoped InGaP insulator takes the advantages of its low DX centers and low reactivity with oxygen [6-10], which may still suffer from the high gate leakage issue. In order to inhibit the gate leakage issue, increase the power handling capabilities, and improve the breakdown voltages, a metal-oxide-semiconductor (MOS) structure has been widely investigated. However, it is still lacks a reliable native oxide film growing on InGaP, and very few papers have reported on InGaP/InGaAs MOS-PHEMTs. In addition, the MOS-PHEMT not only has the advantages of the MOS structure (e.g., lower leakage current and higher breakdown voltage) but also has the high-density, high-mobility 2DEG channel.
Over the past years, a study on the liquid phase oxidation (LPO) of InGaP near room temperature has been done [11-14]. The application of surface passivation to improve the InGaP/GaAs HBTs’ performance has also been first demonstrated . The InGaP/GaAs HBTs with surface passivation by LPO exhibit significant improvement in current gain at low collector current regimes due to the reduction of surface recombination current, as compared to those without surface passivation. Moreover, a larger breakdown voltage and a lower base recombination current are also obtained. In this chapter, the oxide film composition and some issues are addressed. Then a thin InGaP native oxide film prepared by the LPO as the gate dielectric for InGaP/InGaAs MOS-PHEMTs application are discussed, and the comparisons between devices with and without LPO passivation on the InGaP/GaAs HBTs are also reviewed.
2. Characterization of the oxide film
The root mean square (rms) value of surface roughness for the In0.49Ga0.51P sample is estimated to be 1.1 nm before oxidation (i.e., as received) by AFM measurement, and can be improved to 0.95 nm after oxidation (i.e., as grown), as shown in Fig. 1. Fig. 2 shows the SIMS depth profiles before and after liquid phase oxidation on In0.49Ga0.51P. Although LPO on InGaP material has a much slower oxidation rate which is less than 10 nm/h, as comparing to that of the GaAs material, however, it is still feasible to grow a thin oxide film without pH control [15, 16]. The oxidation rate becomes significantly saturated when the oxidation time is longer than an hour, which is measured using a Veeco Instrument DEKTAK and confirmed by SEM.
The XPS depth profiles of the LPO-grown oxide for In0.49Ga0.51P are shown in Fig. 3(a). Fig. 3(b) -(d)show the XPS surface spectra of the Ga-3d, In-3d, and P-2p core levels, respectively. The binding energies for all spectra are calibrated with the reference (as-received) signal. The as-received sample was dipped into a solution of HF:H2O = 1:200 for 30 s before measurement. From Fig. 3(c)-(d), in comparison with the previous paper , the spectrum is rather similar to that of InPO4. This is also confirmed by the values of the O-1s peak energy and energy separations between the main core levels (i.e., Ga-3d, In-3d, and P-2p) in the oxide phases . This clearly suggests that the oxide film is mostly composed of InPO4-like and Ga oxide. In addition, the oxide film may appear to be etched back in the growth solution after 2 h of oxidation. The thermal stability of the oxide layer is also important in
device fabrications because high-temperature processes are usually required. Again, XPS is utilized to also important in device fabrications because high-temperature processes are usually required. Again, XPS is utilized to analyze the surface chemistry of the oxide films, as shown in Fig. 3. After 2 h of oxidation, the RTA processes were performed in a furnace with N2 flowing at 300-700 oC for 1 min ; however, a peak of InPO4-like is still observed. InPO4 (bandgap energy = 4.5 eV) is chemically stable and has rather good dielectric properties . As a result, the InPO4 probably acts as a capping layer for the entire oxide film to enhance the thermal stability. However, the experimental results show that high-temperature treatments (700 oC) will change the properties of Ga2O3, since the XPS energy peak of Ga2O3 shifts to a lower binding energy, and the binding energy is inferred to form the GaOx or Ga2Ox.
3. InGaP/InGaAs MOS-PHEMT
Figure 4 schematically shows the PHEMT structure grown by the metallorganic chemical vapor deposition (MOCVD) on a semi-insulating GaAs substrate. Hall measurement indicates that the electron mobility is 4000 cm2/V s, and the electron sheet density is 2.2×1012 cm-2 at room temperature . The device isolation was accomplished by mesa wet etching down to the buffer layer. The ohmic contacts of the Au/Ge/Ni metal were deposited by evaporation and then were patterned by lift-off processes, followed by RTA. The depth of gate recess is 110 nm for reference PHEMT and 100 nm for MOS-PHEMT. After etching the capping layer and the partial Schottky layer, an LPO growth solution was used to generate the gate oxide for the MOS-PHEMT at 50 oC for 30 min. Finally, the gate electrode was formed with Au. Moreover, the oxide layer, as illustrated in the figure, also selectively and simultaneously passivated the isolated surface sidewall. The gate dimension is 2×100 μm2 with a drain-to-source spacing of 5 μm.
3.2. Results and discussion
Figure 5(a) compares the measured I-V characteristics of the MOS-PHEMT with those of the reference PHEMT fabricated under identical conditions. Clearly, good pinch-off and saturation current characteristics are obtained. Due to the higher energy barriers between the metal gate and the Schottky layer, the MOS-PHEMT can be operated at higher gate-to-source voltage (VGS) and drain-to-source voltage (VDS) than those of the conventional Schottky gate PHEMT, which can enhance the current driving capability. Fig. 5(b) compares the transconductance gm and the drain current density ID as a function of VGS at VDS = 4 V of the MOS-PHEMTs with those of the reference PHEMT. For MOS-PHEMT, the 1.8 V-wide gate voltage swing (defined by 10% reduction from the maximum gm) is higher than that of the PHEMT. The threshold voltage Vth of MOS-PHEMT shifts to the left, which is similar to the result of the one with oxide deposited on the Schottky layer [20, 21]. However, the separation region between the oxide-InGaP interface and the InGaAs channel for MOS-PHEMT is still larger than that of the reference PHEMT in this study, so the drain current density of the PHEMT is smaller than that of the MOS-PHEMT at the same bias VGS due to the decrease of the carrier concentration within the InGaAs 2DEG channel.
In addition, if the depth of gate recess is etched to be 120 nm, the Vth becomes more positive, -0.5 V, for MOS-PHEMT with the identical processing conditions including initial pH value (5.0), temperature (50 oC), and oxidation time (30 min). For Vth shifts to the right, the separation between the oxide-InGaP layer interface and the InGaAs channel layer is decreased due to the consumption of the InGaP during the processes of gate recess and the unique properties of the LPO with the reaction of InGaP, leading to the increase of the total effect of the gate bias on the control of Vth. However, a decrease in the maximum gm, 63 mS/mm, accompanies the degradation in the saturation current, 84 mA/mm at VGS = 1 V. The result is also confirmed by a longer oxidation time, i.e., a thicker oxide layer. This drawback can be overcome by suitable device structures, such as inserting a Si-planar doping layer under the InGaAs channel to increase the carrier density.
The oxide film provides an improvement in the breakdown voltage in terms of the gate leakage current of the MOS structure, supported by the typical gate-to-drain I-V characteristics, as shown in Fig. 6(a). For InGaP MOS-PHEMT, the turn-on voltage, 2.2 V, is obviously higher than that of InGaP PHEMT, 0.8 V, and the corresponding reverse gate-to-drain breakdown voltages, BVGD, are -14.1 V and -6.5 V, respectively. The turn-on voltage and the BVGD are defined as the voltage at which the gate current reaches 1 mA/mm. The gate leakage current can be suppressed at least by more than two orders of magnitude with an oxide film at VGD = -4 V. The smaller gate leakage current of MOS-PHEMT is due to the MOS structure and the elimination of sidewall leakage paths that are directly passivated during the oxidation, which is consistent with the result of Fig. 5. In addition, the gate leakage current observed in MOS-PHEMT comes from a gate leakage path at the edge of the mesa  that is not present in the MOS capacitor, which may contribute to the Schottky-like I-V characteristics for forward biases. Fig. 6(b) shows the gate current density as a function of reverse VGS at different VDS. Due to the high electric field existing in the gate-to-drain region, hot electron phenomena occur in the narrow band-gap InGaAs channel. Electrons can obtain higher energy to generate electron-hole pairs through the enhanced impact ionization, resulting in easy injection of the holes into the gate terminal . However, in InGaP-related devices, it is more difficult for the holes generated by the impact ionization to overcome the valence band discontinuity and to reach the gate , so the bell shaped behavior of the impact ionization does not appear in Fig. 6. Moreover, the gate current density of MOS-PHEMT is significantly improved, which is less than 0.5 μA/mm, as compared to that of PHEMT. In other words, the electrons and holes generated by the impact ionization are decreased to further reduce the drain and gate currents owing to the oxide layer with a high barrier height.
In order to have a better insight into the transient behavior of the studied devices, the gate pulse measurements were performed using a Tektronix 370A curve tracer . VGS was pulsed from the Vth to 0 V with a pulsewidth of 80 μs, while VDS was swept from 0 to 4 V. The comparisons between the static and pulsed I-V characteristics for PHEMT and MOS-PHEMT are shown in Fig. 7. The drain current of PHEMT decreased by 9.8%, while the MOS-PHEMT decreased by only 0.63%. To the best of our knowledge, if the pulsewidth is too short, electrons captured by the traps do not have enough time to be fully emitted. However, if the pulsewidth is long enough, all the trapped electrons are de-trapped and will contribute to the drain current. We believe that the differences between dc and pulsed I-V become evident by applying shorter voltage pulses to the gate such as less than 10-μs pulses for PHEMT and MOS-PHEMT. Therefore, it is clear that the oxide passivation on the Schottky layer can minimize the effect of surface traps, which is consistent with the lower gate leakage current in Fig. 6.
4. InGaP/GaAs HBT with LPO passivation
The structure used for HBT is given in Table 1. The epilayers were grown by a low-pressure MOCVD system on an (100)-oriented semi-insulating (S.I.) GaAs substrate. For InGaP/GaAs HBTs, device fabrication began with emitter definition. The emitter cap layer was removed and stopped at the InGaP active layer. After removing the InGaP layer, a growth solution was used to form the base oxide (passivation) on the exposed extrinsic surface of base and the base contact was then deposited. Finally, the mesa of base was defined and etched to sub-collector before the collector contact deposition. H3PO4-based etchant was used for GaAs and InGaP. The Au/Ge and Au/Be metals were deposited by evaporation and patterned by lift-off processing to form emitter, base and collector regions, respectively.
4.2. Results and discussion
Figure 8 shows the common-emitter I-V characteristics of the HBT with and without surface passivation by LPO. Clearly, the dc current gain (β) of HBTs with passivation is improved (increased) 15% when comparing to HBTs without passivation. The higher β with surface passivation is due to the reduction of the surface recombination current in the exposed extrinsic base regions by LPO method. The common-emitter I-V characteristics of the devices with and without surface passivation at low collector current regimes are shown
in Fig. 9. The devices with surface passivation have higher common-emitter β than those devices without passivation, due to the reduction of the surface combination velocity by using an oxide layer on the base surface. In addition, the β values with and without passivation are 13.3 and 2 at IB = 900 pA, respectively. The maximum increase of 7 fold in the current gain at collector current down to nA level.
Figure 10 illustrates the measured Gummel plots of the devices with and without LPO passivation. The collector currents are almost identical without being affected by the passivation treatment. However, a decrease of the base leakage current at low collector current levels is obviously observed after oxidation. Moreover, it is found that the recombination current at the extrinsic base region and the base-emitter perimeter are competed against one another, resulting in current reduction at lower base-emitter bias VBE = 0.4 V. The increasing β is owing to the reduction of the surface recombination current. It can also be indicated that the device with pasivation exhibits higher β than that without passivation at lower VBE bias. The comparison of β versus the collector current is shown in Fig. 11. The collector-base bias is maintained at 0 V. Clearly, the device with LPO passivation shows wider collector regimes from 10-10 A to 0.1 A. And the maximum shift of
5 fold in the current gain from collector current of 8.1×10-10 A to 1.6×10-10 A can be achieved. This is attributed to the surface state density are suppressed, i.e., the surface recombination current is effectively reduced. The inset shows the base-collector junction current against bias voltage for the devices with and without passivation. For the device with passivation, the breakdown voltage (23.5 V) is higher than that (21.9 V) without passivation at I = 50 μA. The smaller leakage current is owing to the reduction of the surface recombination by the native oxide passivation in the base region. Above results clearly indicate that the β at low (medium) collector current regimes and the breakdown voltage will be increased. Additionally, the base current is decreased for the devices with passivation when comparing to those without passivation, which will be beneficial to low-power electronics and communication applications.
The InGaP/InGaAs/GaAs MOS-PHEMT with the In0.49Ga0.51P oxide as the gate insulator prepared by LPO has been demonstrated. As compared to the counterpart of the conventional InGaP PHEMT, the proposed InGaP MOS-PHEMT can further reduce the gate leakage current at least by two orders of magnitude, increase the breakdown voltage by 200%, and enhance the gate voltage swing. Also, the pulse transient measurement shows much less impact of the surface trap effects for the InGaP MOS-PHEMT. In addition, as compared to the conventional InGaP/GaAs HBTs without surface passivation, the HBTs with LPO passivation possess the characteristics of lower surface recombination currents, higher breakdown voltage and improved higher dc current gain. The HBTs with LPO passivation exhibit 700% improvement in current gain at low collector current regimes by the reduction of surface recombination current, as compared to those without passivation. Therefore, the proposed low-temperature and low-cost LPO can easily be implemented and can provide new opportunities in device applications.