Achieving damage-free, uniform, abrupt, ultra-shallow junctions while simultaneously controlling the doping concentration on the nanoscale is an ongoing challenge to the scaling down of electronic device dimensions. Here, we demonstrate a simple method of effectively doping ΙΙΙ-V compound semiconductors, specifically InGaAs, by a solid phase doping source. This method is based on the in-diffusion of oxygen and/or silicon from a deposited non-stoichiometric silicon dioxide (SiOx) film on InGaAs, which then acts as donors upon activation by annealing. The dopant profile and concentration can be controlled by the deposited film thickness and thermal annealing parameters, giving active carrier concentration of 1.4 × 1018 cm−3. Our results also indicate that conventional silicon based processes must be carefully reviewed for compound semiconductor device fabrication to prevent unintended doping.
In the unceasing pursuit of scaling down semiconductor devices, high mobility compound semiconductors have gained great interest as future candidate materials beyond silicon.1–5 In order to develop devices in the post-silicon regime, effective doping of these compound semiconductors is essential. It is often more challenging to remove ion implantation-induced damage in compound semiconductors than in silicon due to incongruent evaporation during annealing, which leads to increased junction leakage and interferes with dopant activation.6–10 In addition, there are challenges with ion implantation for doping of 3-D structures, such as FinFETs, where shadowing effects of the ion beam result because ion implantation is a line-of-sight process. Such shadowing effects require changes in structure optimization for FinFETs, which, in turn, requires increases in process complexity, cost, and time. To overcome these limitations, alternative doping strategies have been studied for compound semiconductors, such as monolayer doping (MLD), solid phase regrowth (SPR), and surface charge transfer doping.11–14 In this letter, we propose an alternative method for solid phase doping of compound semiconductors, in this case InGaAs. With this method, we demonstrate that silicon or oxygen atoms from a deposited SiOx layer on InGaAs can diffuse into InGaAs during post-deposition annealing, as shown in Figure 1.
The InGaAs samples were cleaned prior to deposition. The samples consist of GaAs (300 nm), InP (850 nm), InAlAs (300 nm), and In0.53Ga0.47As film (300 nm) grown on silicon substrates by molecular beam epitaxy (MBE). The deposition of SiOx films was carried out by plasma enhanced chemical vapor deposition (PECVD) at 285 °C and 900 mTorr with RF power of 30 W and various N2O/SiH4 gas flow ratios.15 The samples were then annealed in a N2 ambient for 10 minutes at 700 °C. The SiOx films were then etched off with HF, and metal contacts (Mo/Ti/Au = 20/40/100 nm) were immediately deposited on the InGaAs samples using electron beam evaporation patterned via metal shadow mask for electrical transport and Hall effect measurements.
Figure 2 shows the Rutherford backscattering spectrometry (RBS) analysis results for the various SiOx films which were deposited on silicon substrates in parallel with the InGaAs films. RBS measurement was done with a NEC 6SDH-2 accelerator and ion beam analysis system. In order to enhance the oxygen scattering cross-section, He2+ incident ions were used at an energy of 3.05 MeV. In addition, the RBS data were taken at a laboratory scattering angle of 170° with a detector of 14 keV nominal energy resolutions. These SiOx films used for characterization were deposited on silicon substrates that were put in the deposition chamber simultaneously with InGaAs film substrates. From the RBS measurements, we were able to determine the composition and thickness of each SiOx film. The RUMP code was used for quantitative analysis of the measured RBS spectra.16 Table I summarizes the Si-to-O ratio for three different gas flow deposition conditions. Depending on the gas flow ratio, the deposited SiOx films resulted in silicon-rich silicon oxide (SRSO, SiOx, x < 2) or oxygen-rich silicon oxide (ORSO, SiOx, x > 2).17
N2O/SiH4 ratio . | Element ratio . | |
---|---|---|
Silicon . | Oxygen . | |
0.6 | 1 | 1.75 |
6 | 1 | 2.1 |
10 | 1 | 2.1 |
N2O/SiH4 ratio . | Element ratio . | |
---|---|---|
Silicon . | Oxygen . | |
0.6 | 1 | 1.75 |
6 | 1 | 2.1 |
10 | 1 | 2.1 |
In order to verify the doping efficacy of our technique, current–voltage measurements were made for InGaAs samples doped by SRSO and ORSO. Figure 3 shows the current flow in the InGaAs layer after SiOx film deposition, annealing, and etch. A higher doping level was obtained with ORSO compared to SRSO. In addition, the current in InGaAs from the ORSO-deposited sample was two orders of magnitude higher in comparison to the SRSO-deposited sample, when the thickness of the ORSO and SRSO films was similar. When a thicker (2 μm) ORSO film was deposited (compared to the thinner, 0.5 μm, film), the current flow in the InGaAs layer was three times greater than in the current in the thinner ORSO film. These results indicate that oxygen in-diffusion is related to InGaAs doping.
Figure 4 shows the secondary ion mass spectrometry (SIMS) depth profiling results for the InGaAs layers after doping. The analysis shows that (1) more oxygen than silicon is incorporated in the InGaAs and (2) the oxygen concentration is further increased in the InGaAs layer with a thicker ORSO film. This is consistent with the findings from the I–V measurements. To confirm the role of oxygen inside InGaAs, ion implantation was conducted for comparison. 20 keV oxygen ions were implanted into InGaAs at three different doses (3 × 1013, 3 × 1014, and 3 × 1015 cm−2), giving an implant depth of 41 nm.18 Following implantation, the samples were capped with 20 nm of Al2O3 by atomic layer deposition (ALD) and annealed with the same conditions as the SiOx treated samples. A control InGaAs sample was deposited with Al2O3, annealed, and etched to remove Al2O3, and tested as a control to confirm that the Al2O3 capping layer does not play any role in InGaAs doping.
Figure 5(a) shows that there is an increase in active dopant concentration in oxygen-implanted samples in proportion to oxygen dose, and a reduction in sheet resistance. To show that ion implantation did not otherwise alter the structure of the InGaAs sample, depth profiles of the oxygen-implanted InGaAs samples were taken with time-of-flight SIMS (TOF SIMS) as shown in Figure 5(b). The depth profiles show that no silicon or other outdiffusion of underlying layers occurred to impact the InGaAs film. This indicates that oxygen indeed appears to act as an n-type dopant at active dopant concentrations that have not been reported to date. The SIMS depth profile shows the dopant distribution with respect to depth, whereas the current–voltage measurement reflects an average dopant concentration. The InGaAs sample doped with thick ORSO film showed an active doping concentration of 1.4 × 1018 cm−3, which is comparable to other doping methods including ion implantation.
It has been previously reported that an increase in compressive stress within a GaAs substrate can generate vacancies via enhanced gallium diffusion and arsenic–antisite defects.19,20 This extrinsic stress occurs because of the difference in the coefficient of thermal expansion (CTE) between SiO2 (0.55 × 10−6 K−1)and GaAs (6.86 × 10−6 K−1). Similarly, vacancies can be formed due to the thermal stress between SiOx and InGaAs in the InGaAs substrate (Figure 6(b)). In addition, the amount of vacancies generated within InGaAs is greater in the thicker ORSO (2 μm) than in the thinner ORSO (0.5 μm). This is because increasing the film thickness generates more stress, which leads to an enhancement of the vacancy formation phenomenon. This, in turn, increases the probability for dopants to diffuse into InGaAs (Figure 6(c)).
In addition to the conventional annealing temperature and time process parameters, our method provides an additional parameter available to control doping (i.e., solid source film thickness and stoichiometry). The high concentration of oxygen compared to silicon inside InGaAs can be understood by considering the diffusivity of each element, which is dictated by an Arrhenius relation. Oxygen has a lower activation energy for diffusion (EA = 1.1 eV) compared to silicon (EA = 2.2–2.8 eV) in GaAs, an analogue of InGaAs.21 In the case that oxygen has a lower energy barrier for diffusion into InGaAs, it allows a higher doping concentration at a given temperature.
Oxygen impurities in III-V semiconductors have typically been considered as contaminants that create energy levels near mid-gap.22–25 However, oxygen has also been shown to increase the free carrier density in GaAs and InGaAs in some cases, presumably due to differences of the where the oxygen sits in the lattice and how it affects the bandstructure.25–27 Although the fundamental mechanism of how oxygen acts as a dopant was not elucidated in our work, the use of ORSO demonstrates an alternative approach for effectively doping ΙΙΙ-V compound semiconductors. It also points out a potential unintentional doping effect when applying silicon processes to compound semiconductors, such as for silicon dioxide-based device isolation. While this isolation method is effective for conventional silicon-based technology, unintended doping can occur for compound semiconductors during a similar process steps and can negatively affect device performance.
In conclusion, we were able to demonstrate a simple technique for doping compound semiconductors through PECVD SiOx. As oxygen has a lower activation energy for diffusion, it diffuses more easily than silicon into InGaAs, leading to n-type doping. This solid phase doping method can be applied to other semiconductors by choosing appropriate films. Our method provides an alternative way to overcome one of the challenges of doping compound semiconductors, such as InGaAs.
This was supported, in part, by a grant from Samsung and by the NSF NNCI program.