Investigation of the thermal stability of Mo-In_0.45Ga_0.47As for applications as source/drain contacts

III–V materials, such as GaAs and InGaAs, show promise as alternative channel materials to Si for n-type metal-oxide-semiconductor field effect transistors (MOSFETs) due to their higher injection velocities and electron mobilities. Research has recently focused on In_0.53Ga_0.47As (referred to as InGaAs for the remainder of this report) due to promising improvements in the InGaAs/high-κ interface;^1–6 however, the issue of source/drain (S/D) contacts in InGaAs based MOSFETs remains. In order to achieve the increased bandwidth, operation speed, and device packing density required in future MOSFETs, the resistivity of these contacts must decrease. A possible solution is to find a self-aligned silicide-like material to act as the S/D contacts.^7,8 The optimum material would ideally display an abrupt ordered interface with InGaAs, low contact resistance (∼2 Ω μm²), as well as the thermal stability to withstand the temperatures involved in current MOSFET fabrication processes. The material chosen should also have a Schottky barrier height of less than 0.2 eV, as predicted by intrinsic interface state theory,⁹ in order to form an Ohmic contact with n-InGaAs.

Mo-InGaAs has been highlighted as a candidate system for this application. The thermal stability of Mo and the alignment between the work-function of Mo (4.65 eV) and the conduction band edge of InGaAs (electron affinity of 4.5 eV) make Mo a good candidate for the formation of an Ohmic contact with n-InGaAs.¹⁰ Contact resistance values of 1.1 to 4.7 Ω μm² and the ability to incorporate Mo-InGaAs into current device processing techniques have been previously demonstrated.^10–14 However, unlike one of the other candidate material systems, Ni-InGaAs,¹⁵ there has been no definitive study on the interface chemistry and stability as a function of anneal temperature. The aim of this preliminary work is to study the chemical and electrical stability of the Mo/InGaAs system at room temperature and following a range of post deposition thermal anneals between 250 and 500 °C.

The undoped InGaAs samples comprised of 30 nm thick In_0.53Ga_0.47As layers grown by molecular beam epitaxy (MBE) on InAlAs epi-ready layers on InP substrates. The samples were cleaned for 60 s in dilute hydrofluoric acid prior to being loaded into the metal deposition chamber. For hard X-ray photoelectron spectroscopy (HAXPES) and X-ray absorption spectroscopy (XAS) analysis, one sample was left bare to act as a clean InGaAs reference, one had 30 nm of Mo sputter deposited to act as a thick Mo reference, while all other samples had 5 nm of Mo deposited. One of these (5 nm) Mo capped samples was left unannealed, while the remaining samples were annealed in-situ in the deposition chamber in 100 °C steps ranging between 250 and 450 °C, and one sample annealed at 500 °C. In order to investigate the changes in the relative intensities of the core level photoemission peaks as a function of thermal annealing in the HAXPES study, which can assist in determining the extent of inter-diffusion and hence the location of the chemical species, an internal reference peak which remains at constant intensity is necessary. This was achieved by capping all of the HAXPES samples with a sputter deposited 3 nm SiN capping layer in a separate chamber, after the Mo deposition and anneal, to act both as a barrier to post processing oxidation and as an internal reference. As such, the Si 1s peak was acquired at the same time as each core level, with a constant ratio between the number of scans of the Si 1s and the relevant core level of 1:4. This was subsequently used to correct for any photon energy drift during the HAXPES measurements, and to normalize the intensity of the core level spectra across all samples. Using the Si 1s reference spectra for each core level, the change in peak intensity of a given chemical species throughout the anneal study can be determined, and thus the diffusion behavior of the different elements can be investigated. An equivalent sample set was prepared using identical conditions for the XAS measurements; however, the SiN layer was not deposited on these samples, as a reference layer is not required. Electrical sheet resistance (R_sh) measurements were performed on the Mo capped, annealed samples using a four-point probe apparatus.

HAXPES measurements were carried out on the National Institute of Standards and Technology (NIST) beamline X24A at the National Synchrotron Light Source (NSLS) at Brookhaven National Laboratory (BNL). A double Si (111) crystal monochromator allowed for photon energy selection in the range of 2.1 to 5.0 keV. A hemispherical electron energy analyzer was operated at a pass energy of 200 eV, giving an overall instrumental broadening of 0.52 eV at 4050 eV. The total sampling depth of the HAXPES measurement using a photon energy of 4050 eV is estimated to be 21 nm (Ref. 16), making it possible to detect photoemission signals for the InGaAs substrate even in the presence of the SiN overlayer. The XPS core level spectra were curve fitted using Voigt profiles composed of Gaussian and Lorentzian line shapes with a Shirley-type background. Fluorescence-yield XAS measurements were performed on the As, Ga, and Mo K-edges at room temperature on the NIST beamline X23A2 at the NSLS. A double Si (311) crystal monochromator allowed for photon energy selection in the range of 4.9 to 30 keV. XAS measurements were performed with the angle between the incident beam and sample surface at 0.5° (Ga and As K-edges) and 1.2° (Mo K-edge) to probe the entire sample. The grazing incidence angle chosen was at a critical angle of 0.15° where the x-ray photon penetration depth into the sample is between 2.5 and 5 nm. Below the critical angle, total external reflection occurs, where an incidence angle of 90° is normal to the sample surface.

The most important factor in determining the suitability of Mo-InGaAs for use as a source/drain material in future InGaAs-based MOSFETs is the electrical resistance and stability under thermal processing. The sheet resistance of Mo-InGaAs samples over a range of annealing temperatures was determined by four-point probe measurements. Figure 1 shows a plot of R_sh for a 5 nm thick Mo layer as-deposited and annealed in 100 °C steps between 250 °C and 450 °C, and also a sample after anneal at 500 °C. The R_sh value stays approximately constant (with an average value of 96 Ω/◻) throughout the anneal series, within ±15 Ω/◻, indicating that the R_sh of Mo-InGaAs is thermally stable up to the temperatures used in InGaAs MOSFET processing. These results are in contrast to the Ni-InGaAs system, which is an alternative candidate for source/drain contacts, where substantial fluctuation of R_sh is seen as a function of anneal temperature.^15,17–19

In order to explore the chemical stability of the Mo-InGaAs system, the interfacial region between the Mo and InGaAs was probed by HAXPES and XAS measurements. Characteristic HAXPES spectra are shown in Figure 2. Intensities are normalized to the Si 1s from the SiN capping layer, as shown in Figure S1 in the supplementary material. Note that for the As 2p_3/2, Ga 2p_3/2, and In 3d_5/2 core levels, the plotted intensities of the bare InGaAs (no Mo) sample spectra were divided by a factor of 5 for easy comparison with the core levels from the annealed samples. The InGaAs spectra (Figures 2(a)–2(c)) were all curve fitted (not shown) and show the expected InGaAs peaks (As-Ga in the As 2p_3/2, Ga-As in the Ga 2p_3/2, and In-As in the In 3d_5/2 spectra) with trace evidence of oxide-related features (primarily at 1323 eV in the As 2p) on the no Mo sample. The Mo 2p_3/2 spectra remain similarly stable throughout the anneal study, as do the Mo 3d in Figure S2 in the supplementary material. The curve fits of the 30 nm thick reference Mo sample (not shown) similarly show the expected metallic Mo 2p_3/2 signal with trace evidence of native oxide on the higher binding energy side (at 2521 eV) of the peak due to air exposure prior to the SiN deposition. These core-level spectra showed no substantive changes in peak profile throughout the annealing study, indicating little or no chemical interaction between the Mo and InGaAs in agreement with a previous study by Lin et al.²⁰ Slight changes in the binding energy position (<0.15 eV) and peak intensity after anneal in all core levels suggest there may be some minimal interaction between the Mo and InGaAs, or possible densification of the Mo film. However, the lack of any change in line-shape suggests no formation of additional chemical states at the interface.

Figure 3(a) shows the Ga K-edge XAS spectra of a bare InGaAs (no Mo) sample and a Mo-InGaAs sample as-deposited and after a 450 °C anneal. It is clear that there is no change in the spectral profile in either the as-deposited or annealed sample, suggesting there is no change in the local bonding environment with thermal anneal in agreement with the HAXPES data. K-edge data were also acquired and showed similar behavior in terms of thermal stability, shown in Figure S3(a) in the supplementary material. Figure 3(b) shows the Mo K-edge for a thick (30 nm) Mo reference sample and for the thin (5 nm) Mo-InGaAs sample, as-deposited and after a 450 °C anneal. Similar to the As and Ga K-edges, there is no change in the Mo K-edge spectra suggesting no interaction between Mo and InGaAs, in agreement with the HAXPES spectra. Surface-sensitive XAS spectra were also acquired (Figures S3 and S4 in the supplementary material) and demonstrate that there is no change in the chemical structure with depth, consistent with a chemically abrupt interface. Therefore, the combined HAXPES and XAS analysis of the samples suggests that the Mo-InGaAs interface chemistry is thermally stable.

To further investigate the thermal stability of the Mo-InGaAs system, it is important to understand the location and diffusion behavior of each chemical species throughout the anneal study. To do this, HAXPES peak areas normalized to photoionization cross sections and inelastic mean free paths and referenced to the 3 nm SiN capping layer are plotted as a function of anneal temperature for each sample in Figures 4(a)–4(d); the data were also normalized to the as-deposited peak areas. The error bars in Figure 4 are taken from the residual from the peak fitting using the XPS fitting software. All of the photoemission peaks display a gradual decrease in signal intensity as a function of annealing temperature by approximately the same amount. These near-identical trends suggest that this change cannot be attributed to inter-diffusion. If inter-diffusion occurs, then at least one of the photoemission signals should increase with respect to the intensity of the other peaks. This decrease in intensity as a function of increased anneal temperature could be attributed to the densification of the SiN film, which would result in an increased IMFP and thus a decrease in the peak area of all core levels originating from layers below the SiN as seen here for the Mo and InGaAs. The similar trends observed for all peaks again indicate that the Mo-InGaAs interface is structurally stable as a function of anneal temperature. These results are in contrast to a previous study of the Ni-InGaAs system which describes a large extent of inter-diffusion between the Ni and InGaAs as a function of thermal anneal, with the Ga diffusing through the Ni to the sample surface.¹⁵ This comparison further demonstrates the improved thermal stability of Mo-InGaAs over Ni-InGaAs.

In summary, a combination of sheet resistance, XAS, and HAXPES characterization has shown the electrical and chemical stability of Mo-InGaAs films as a function of post-deposition anneal up to 500 °C. No evidence of chemical interaction or inter-diffusion between the Mo and InGaAs layers was observed. This thermal stability combined with the results from previous studies, which found the contact resistance to be suitable for device fabrication, indicates that Mo is a strong candidate for use as a source/drain material in future InGaAs-based MOSFETs as it is stable up to the temperatures typically found in device processing.

The chemical and electrical stability of a Mo-InGaAs film was studied as a function of post-deposition anneal temperature. It was found that the sheet resistivity of 5 nm thick Mo films remains constant over the entire annealing temperature range. Combined XAS and HAXPES analysis of the films reveals no inter-diffusion between the Mo and InGaAs, indicating that the interface is chemically abrupt. This highly stable material system is therefore well suited to use as a future source/drain material in InGaAs MOSFET devices.

See supplementary material for additional x-ray absorption and hard x-ray photoelectron spectroscopy data.

The authors from Dublin City University acknowledge the financial support of SFI under Grant No. SFI/09/IN.1/I2633. The use of the National Synchrotron Light Source, Brookhaven National Laboratory, was supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under Contract No. DE-AC02-98CH10886. Additional support was provided by the National Institute of Standards and Technology.