Self-limiting growth of ultrathin Ga₂O₃ for the passivation of Al₂O₃/InGaAs interfaces

The passivation of high-k/III–V interfaces has long been investigated with the aim for realizing superior channel quality for next-generation metal–insulator–semiconductor field-effect transistor (MISFET) applications. The most important challenge is engineering the interface between the channel material and gate dielectrics. The native oxide of III–V MIS generate defects in the form of As–As bonding and Ga dangling bonds, which causes a high interface trap density (D_it), and introduces strong Fermi-level pinning and severe deterioration of channel mobility.^1,2) Therefore, several oxide removal techniques such as aqueous HCl and NH₄OH etching,³⁾ sulfur passivation,^4–6) selenium passivation,⁷⁾ self-cleaning with Al(CH₃)₃ (trimethylaluminum; TMA),^8–12) and H₂ plasma cleaning,¹³⁾ have been reported for the improvement of the GaAs and InGaAs channel properties. In addition to oxide removal, further surface passivation techniques such as plasma nitridation^14,15) and epitaxial Si passivation¹⁶⁾ have been proposed to control the anion composition at the MIS interface.

Although there have been many reports that removal of the Ga oxide component correlate with improvements in the MIS properties, some reports showed that Ga₂O was observed after TMA cleaning and Si passivation,^17,18) and an InGaAs MISFET with molecular beam epitaxy (MBE)-grown Al₂O₃/Ga₂O₃(Gd₂O₃) stacked gate dielectric exhibited superior characteristics.¹⁹⁾ Therefore, the effects of GaO_x species at III–V MIS interfaces are under debate at present. A scanning tunneling microscopy (STM) study showed that Ga₂O formation on InGaAs by MBE started with Ga₂O molecules bonding to As atoms, and disordered structures were formed after full coverage.²⁰⁾ Another group also reported that Ga₂O did not passivate nor unpin the interface and was unable to form on the InGaAs surface. The capacitance–voltage (C–V) characteristics of Al₂O₃/InGaAs with Ga₂O passivation also exhibited large frequency dispersion, which indicates high D_it and Fermi level pinning.²¹⁾ On the other hand, Ga₂O film passivation on GaAs using the MBE technique has also been demonstrated.^22,23) A monolayer of Ga₂O formed a crystalline interface that was electronically unpinned due to Ga₂O restoring the surface As and Ga atoms to near-bulk charge.²³⁾

The deposition technique to obtain GaO_x film should also be considered. Many deposition methods for GaO_x film growth have been successful, including MBE,^20–23) atomic layer deposition (ALD),²⁴⁾ plasma-enhanced ALD (PEALD),²⁵⁾ and electron beam evaporation.²⁶⁾ Among these techniques, ALD has received significant interest due to its intrinsic self-terminated growth nature.²⁷⁾ Although ALD of GaO_x films using Ga(CH₃)₃ (trimethylgallium; TMGa) and H₂O is known to be difficult due to the high decomposition temperature of TMGa (ca. 300 °C),^28,29) it has an advantage to form a thin passivation layer at the MIS interface from the viewpoint of equivalent oxide thickness (EOT) scaling. In this paper, we report that an ultrathin Ga₂O₃ layer can be grown in a self-limiting manner by utilizing the ALD chemistry of TMGa and H₂O. The impact of Ga₂O₃ layer insertion to the Al₂O₃/InGaAs interface on the MIS structure is also investigated.

Experiments were conducted using an ALD system with TMA, TMGa, and H₂O as precursors. In_0.53Ga_0.47As(001) samples employed for this study were epitaxial wafers prepared by metal organic chemical vapor deposition (MOCVD) on InP wafers. MIS capacitors and MISFETs were fabricated on n-type Si-doped and p-type Zn-doped InGaAs epitaxial layers, respectively. The thickness and doping concentration of the epitaxial layer were 0.5 µm and ca. 3 × 10¹⁶ cm⁻³, respectively, both for the n- and p-type InGaAs. To fabricate a capacitor, n-InGaAs(100) substrates were treated in NH₄OH solution for 1 min and rinsed in deionized (DI) water for 1 min prior to ALD. ALD of a 10-nm-thick Al₂O₃ layer was performed using a fixed deposition temperature of 250 °C and at a rate of 0.1 nm/cycle. Postdeposition annealing (PDA) was conducted in vacuum at 400 °C for 2 min. After gate electrode and back contact evaporation, postmetallization annealing (PMA) was performed in Ar at 350 °C for 2 min.

Figure 1 shows the change in the GaO_x film thickness on n-Si(001), n-GaAs(001), p- and n-InGaAs(001) substrates with the number of deposition cycles up to 150. All the III–V surfaces were prepared by NH₄OH treatment, and the Si surfaces were untreated prior to GaO_x deposition. The oxide thicknesses were measured and normalized using spectroscopic ellipsometry according to the reference refractive-index data for Ga₂O₃ and the Cauchy model. GaO_x growth was conducted by repeating the cycle of TMGa and H₂O supply at 250 °C. The TMGa dosage was set at 500 Pa·s, which was much higher than the TMA dosage for Al₂O₃ growth (ca. 1 Pa·s), to ensure the sufficient TMGa supply. GaO_x growth on the GaAs substrates could be detected after the deposition of more than 30 cycles, at which time the GaO_x film thickness became saturated at ca. 0.2 nm. GaO_x growth on the InGaAs substrates was similar to that on the GaAs substrates. Although GaO_x growth on the Si substrates was detected instantly after the first cycle, the thickness became saturated at ca. 1 nm. These results indicate that a thick GaO_x film cannot be formed using the deposition chemistry of TMGa and H₂O. However, monolayer GaO_x film can be grown on GaAs and InGaAs substrates in the manner of self-limiting growth after the initial incubation period. The difference in the initial formation of GaO_x on Si and GaAs substrates implies that the self-cleaning on the III–V surface by TMGa was carried out similarly to that for TMA.¹¹⁾ In addition, the thickness difference in the GaO_x film on native SiO₂ and GaAs substrates may be slightly distorted because the thickness was calculated using the same refractive index.

Zoom In Zoom Out Reset image size

To reveal the growth of the self-limiting GaO_x film, X-ray photoelectron spectroscopy (XPS) scans were obtained with and without GaO_x film. Figures 2(a) and 2(b) show As 2p_3/2 and Ga 2p_3/2 XPS spectra of the NH₄OH-treated n-InGaAs(001) surface and that following the deposition of ALD-Al₂O₃ (1 nm) and ALD-GaO_x (30 cycles)/ALD-Al₂O₃ (1 nm). In this experiment, all the samples were analyzed using ex situ XPS with an Al Kα X-ray source (1,486.7 eV). Figure 2(a) shows that the oxides on the InGaAs surface were effectively reduced by 1 nm Al₂O₃ capping due to the self-cleaning effect of TMA.^10,11) The As 2p peak was also reduced by insertion of the GaO_x layer. This behavior indicated that TMGa also has a self-cleaning effect analogous with TMA. On the other hand, the Ga 2p spectra in Fig. 2(b) after Al₂O₃ and GaO_x/Al₂O₃ depositions presented a contrastive characteristic. NH₄OH treatment left a surface oxide that exhibited a large shoulder on the high-binding-energy side of the bulk peak at 1.6 eV, which was possibly due to Ga(AsO₃)₃ or Ga(OH)₃.¹⁷⁾ While direct Al₂O₃ deposition reduced the GaO_x layer, ALD-GaO_x treatment realized distinct GaO_x growth on the InGaAs surface with chemical shift of 1.2 eV corresponding to Ga³⁺ oxidation state.⁹⁾ The chemical shift reduction from 1.6 to 1.2 eV indicates the removal of the higher oxidation component of Ga and the formation of the interfacial Ga₂O₃ layer between Al₂O₃ and InGaAs. The increase in thickness of the Ga₂O₃ layer by inserting the ALD-GaO_x was estimated from the Ga 2p peak to be 0.11 nm, which is consistent with the ellipsometry result shown in Fig. 1. These results indicate that TMGa had a selective self-cleaning effect on unnecessary AsO_x and GaO_x, similar to TMA, as well as significantly contributes to the formation of Ga₂O₃. Both Figs. 1 and 2 indicate monolayer termination on the InGaAs surface by ALD-Ga₂O₃ growth, which was eventually suitable to be an ultrathin passivation layer.

Zoom In Zoom Out Reset image size

A possible mechanism for Ga₂O₃ formation on InGaAs can be described as follows. AsO_x and GaO_x on the InGaAs surface are first reduced by the self-cleaning effect with exposure to TMGa, as shown in Fig. 2(a), and the surface is partially covered with Ga(CH₃)_n molecules (where n = 1 or 2). This selective absorption is limited by the bonding requirements and steric effects,²⁹⁾ and no Ga(CH₃)_n molecules are absorbed on Ga atoms at the surface.³⁰⁾ After surface oxide removal processing, (CH₃)_n is desorbed after the supply of H₂O and is decomposed into Ga-OH^29,31) or Ga₂O₃, as shown in Fig. 2(b). Repeated cycling of alternate TMGa and H₂O supply then provides almost complete surface coverage with Ga-OH because decomposition is difficult at low temperature. Ga(CH₃)_n molecules are rarely absorbed on this surface, so that Ga₂O₃ formation becomes more difficult. Therefore, the self-limiting growth of monolayer Ga₂O₃ on the InGaAs surface is achieved. This monolayer (1 ML) Ga₂O₃ layer could be chemically stable, with the potential to provide a low interface trap density (D_it) and prevent the intermixing of InAlGaAs at the MIS Al₂O₃/InGaAs interface.

Figures 3(a) and 3(b) show the C–V characteristics and conductance data for the (1) ALD-Al₂O₃ (10 nm)/InGaAs and (2) ALD-Al₂O₃ (10 nm)/30-cycle ALD-Ga₂O₃/InGaAs MIS capacitors (MISCAPs) with Au electrodes. The sample with the Ga₂O₃ passivation layer showed a reduction of capacitance under negative bias. Frequency dispersions around the flatband voltage (V_fb) and under charge accumulation were also slightly affected by the Ga₂O₃ passivation layer compared with the control sample, as shown in Figs. 3(a-1) and 3(a-2). A significant effect of Ga₂O₃ passivation was observed in the conductance data obtained at room temperature. The conductance at low frequencies was significantly reduced by the Ga₂O₃ passivation layer and the difference in the conductance peaks was clearly evident from a comparison of Figs. 3(b-1) and 3(b-2). The decrease in conductance may be caused by a reduction of the border trap in the Al₂O₃ film near the MIS interface by Ga₂O₃ passivation.³²⁾ D_it profiles for both the sample with the Ga₂O₃ passivation layer and the control sample from a bias voltage of V_fb = 0 to −0.5 V are shown in Fig. 4 . D_it was estimated using the conductance method assuming that D_it = 2.5/q × Gp/ω. D_it around the midgap of the InGaAs bandgap was extensively reduced over a wide range by Ga₂O₃ passivation. A minimum D_it as low as 2.4 × 10¹¹ cm⁻² eV⁻¹ was achieved.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

To investigate the impact of Ga₂O₃ passivation on MISFET performance, n-channel InGaAs MISFETs were fabricated using the Ni-alloy process with and without Ga₂O₃ passivation of NH₄OH- or (NH₄)₂S-treated surfaces. The p-InGaAs surface was first prepared by wet etching for 1 min in NH₄OH or 5 min in (NH₄)₂S and rinsed with deionized water for 1 min. Deposition of the Al₂O₃ gate dielectric layer with and without a 30-cycle Ga₂O₃ passivation layer was then performed. PDA was performed at 400 °C for 2 min. A 30-nm-thick TaN_x gate electrode was then deposited on the Al₂O₃ film by sputtering. TaN_x and Al₂O₃ gate were selectively removed by sulfur hexafluoride (SF₆) reactive ion etching and buffered hydrofluoric (BHF) wet etching, respectively. A 25-nm-thick Ni electrode layer was deposited by electron beam evaporation and annealed in Ar at 350 °C for 2 min. Finally, the fabrication process was completed by the etching of the unreacted Ni in HCl : DI water (1 : 5) for source, drain, and gate electrode isolation.

Figures 5 (a-1) and 5(a-2) show the drain current–gate voltage (I_D–V_G) characteristics of the NH₄OH- and (NH₄)₂S-treated Al₂O₃/InGaAs MISFETs, respectively. I_D–V_G curves with and without Ga₂O₃ passivation layers of 100-µm-gate-length devices were measured at drain voltages (V_D) of 50 mV and 1 V. The subthreshold slopes (SS) were 236 and 147 mV/decade for the NH₄OH- and (NH₄)₂S-treated control devices, which were improved to 198 and 141 mV/decade after Ga₂O₃ passivation, respectively. The threshold voltages (V_th) were not significantly changed after passivation with the Ga₂O₃ layer for the NH₄OH- and (NH₄)₂S-treated devices. The on/off ratios were improved for both treatments by Ga₂O₃ passivation. Lower off-currents were achieved with Ga₂O₃ passivation due to the reduction of surface leakage current. Although the SS values were reduced by Ga₂O₃ passivation, further improvement could be achieved by controlling the junction quality and channel metal connection.

Zoom In Zoom Out Reset image size

Figures 5(b-1) and 5(b-2) show the effective mobility (μ_eff) as a function of the inversion carrier density (N_s) for the NH₄OH- and (NH₄)₂S-treated devices, respectively. The N_s value was estimated by the split C–V method at 10 kHz. μ_eff for the Ga₂O₃ passivation devices were significantly higher than that for the control devices with both treatments over the entire N_s range. μ_eff for the NH₄OH- and (NH₄)₂S-treated devices with Ga₂O₃ passivation were increased by around 2.3 and 1.5 times that of the control devices, respectively. The improvement in µ_eff for the NH₄OH-treated device by Ga₂O₃ passivation was associated with MISCAP data. The Ga₂O₃ passivation layer improved the Al₂O₃/InGaAs MIS interface properties by reducing the native oxide and terminating Ga₂O₃ onto the surface. The combination of (NH₄)₂S treatment and Ga₂O₃ passivation layer revealed superior improvement of the Al₂O₃/InGaAs gate stack, which may be due to a reduction of D_it both near the conduction band edge^4–6) and around the midgap in the Al₂O₃/InGaAs MIS structure, respectively.

In conclusion, an ultrathin Ga₂O₃ interface layer was grown in a self-limiting manner using ALD with TMGa and H₂O precursors. Interaction of TMGa and H₂O with OH groups possibly restricted the formation of the Ga₂O₃ film. The self-cleaning effect of TMGa on AsO_x at the InGaAs surface was observed, similar to that for TMA. Insertion of the ultrathin Ga₂O₃ layer to the Al₂O₃/InGaAs interface was effective to reduce the trap density around the midgap. The MISFET performance characteristics also showed improvement of μ_eff by ultrathin Ga₂O₃ passivation for both NH₄OH- and (NH₄)₂S-treated devices. These experimental results demonstrate that Ga₂O₃ passivation does improve Al₂O₃/InGaAs MIS interface and FET performance.

Acknowledgements