1. Introduction

Metallic films of high-quality find their uses in exploring condensed matter physics and in making industrial products. In the past decade, to reproducibly fabricate plasmonic devices/circuits in ultra-violet (UV) regime [1,2] and superconducting qubits/resonators [3,4], growing single-crystalline aluminum films on large-scale wafers have attracted increasing attentions. Epitaxial nano-scale aluminum films have been successfully grown on GaAs, Si, and sapphire substrates [5–8]. The material characterization on those films and their transport properties revealed their promising potential for further applications. The reported literatures using high-quality metallic films, such as the plasmonic nano-structures, [2,9], the beyond-diffraction-limit nano-lasers, [8,10,11] and the enhanced superconductivity [6,12,13] evidenced that the perfectly crystalline films could boost device performance and enhance design flexibility.

High-quality nano-scale aluminum films have also been applied to improve the performance of conventional photonic devices. By adjusting the film thickness, they can be wide-bandwidth highly-reflective mirrors for solar energy concentrator [14] or semi-transparent conducting layers for resonant cavity design to enhance light-matter interaction or photo-detection [15–18]. In this aspect, aside from the superior crystal quality, smooth surface morphology is also required to obtain ideal optical reflectance. However, at least two issues remain unsolved for their wide applications. (1) The preparation of high-quality aluminum films, particularly for high reflectance in blue-to-UV regime, requires thick film thickness (typically > 100 nm) and/or additional surface treatment [19]. (2) The reported optical constants of epitaxial aluminum film are different from one to the other [20–22] and from the conventional bulk values [23]. The inconsistence may arise from the epitaxy quality of aluminum films or the extraction methods. Further works are certainly needed to resolve this issue.

In this report, with careful surface treatment on pre-deposition template and proper growth recipe, we have demonstrated the aluminum epitaxy on GaAs (100) and Si (111) substrates by using molecular beam epitaxy (MBE) technique. The Al film thickness ranges from 3 to 50 nm with the surface root-mean-square roughness of about 0.2–0.7 nm. The θ-2θ X-ray diffraction (XRD) shows the normal-to-substrate direction is Al (111) and the rocking curves gives the narrowest peak with a linewidth ∼ 0.17°. To study their optical reflectance, we have measured normal-incidence reflectivity spectra in the range of 190–1000 nm. The observed reflectance, similar to the bulk data, is ∼ 90% for the 40-nm-thick samples, indicating the excellent optical quality suitable for plasmonic and optoelectronic device applications. We have also performed the simulation and found that, by using the commonly-used Palik’s optical parameters [23], the measured reflectivity spectra are well consistent with the simulated ones for all the films. This work paves the way for developing, designing, and optimizing aluminum-based optical, plasmonic and superconducting devices/circuits using high-quality epitaxial aluminum nano-films.

2. Sample growth and structural analysis

In this work, two sets of samples were grown in a Veeco Gen II solid source MBE system. The first set includes four aluminum samples on GaAs (100) substrate, named as samples A1, A2, A3, and A4, with aluminum film thickness of 40, 20, 10, and 3 nm, respectively. These samples were prepared as follows. After thermal desorption of the native oxide on epi-ready semi-insulating GaAs substrate at 600°C for 20 minutes, a 200-nm-thick undoped GaAs buffer layer was grown at 585°C. The substrate was then heated to 600 °C without arsenic flux for 3 minutes in order to transform the surface into the Ga-rich condition. After that, GaAs wafer was cooled down in the ultra-high vacuum chamber to about 0 °C. When the residual arsenic vapor in the growth chamber was pumped out and the background pressure was lower than 2 × 10⁻¹⁰Torr, aluminum films were finally deposited. The second set is consisted of three samples on Si (111) substrates, named as samples B1, B2, and B3, with the respective aluminum thickness of 50, 10, and 3 nm. The phosphorus-doped Si (111) wafer was first cleaned by the standard RCA procedure and then etched by dilute HF to remove the native oxide. In order to obtain an atomically smooth surface, Si substrate was then dipped in 40% NH₄F solution for 2 minutes. Si wafers were finally rinsed in deionized water for 20 seconds before loading into MBE chambers. In the growth chamber, Si wafer was heated to about 630 °C, at which Si 1 × 1 surface can be observed by the equipped high-energy electron diffraction (RHEED). Si wafer was then also cooled down in the ultra-high vacuum chamber to about 0 °C, and, finally Al films were grown. The aluminum growth rate was 0.1 nm/s for all the samples.

After growth, all samples were loaded out the MBE chamber for further characterizations. The surface morphology of samples was examined in air by atomic force microscopy (AFM, Bruker Edge) with the tapping mode. A Bede D1 high-resolution x-ray diffractometer, equipped with CuKα1 radiation (λ = 1.5406 Å), a two-bounce Si 220 channel-cut collimator crystal, and a dual channel Si 220 analyzer crystal, was used to characterize the crystal structure of the samples. Normal-incidence reflectance spectra in 190-1000 nm range with a step of 1 nm were performed with a thin film measurement system (n&k 1500, n&k Tech. Inc.). The approximate spot size of the reflectometry is 1 mm upon the sample.

The typical AFM surface images of 5 × 5 µm² are shown in Figs. 1(a) and 1(b) from sample A1 and sample B1, respectively. Very smooth surfaces without specific feature were observed. The root-mean-square (RMS) roughness of sample A1 is about 0.56 nm, and about 0.69 nm of sample B1. Figure 1(c) summarizes the RMS roughness as a function of film thickness for all the samples. The roughness are in the range of 0.2–0.7 nm but tend to increase with the increasing thickness, probably due to the low growth temperature.

 

Fig. 1. Surface images by atomic force microscopy from, (a) Sample A1, and (b) sample B1. (c) Root-mean-square surface roughness with error bar as a function of film thickness from all the samples on GaAs (black solid sphere) and Si (blue hollow circle) substrates.

Download Full Size | PDF

Figure 2 presents the measured XRD data from samples A1 and B1. In Figs. 2(a) and 2(b), the typical XRD 2θ-θ scans along surface normal of samples A1 and B1 reveal strong and sharp Al (111) and Al (222) reflections, together with substrate signals. The lattice constants calculated from the Al (111) peaks on GaAs and Si substrates are 0.4065 nm and 0.4049 nm, respectively. These values are very close to 0.405 nm of bulk aluminum [24] and tell that the aluminum films are basically relaxed. The rocking curves in the insets on Figs. 2(a) and 2(b) show that the full width at half-maximum (FWHM) of Al (111) reflections are about 0.17° and 0.55° for samples A1 and B1, respectively. This result suggests that epitaxial Al (111) films have been successfully deposited on GaAs and Si substrates. It is also noticed that, by comparing the FWHMs of the measured rocking curves from samples A1 and B1, the crystallinity of Al films grown on GaAs substrate is excellent and slightly better compared with that of Al films grown on Si substrates.

 

Fig. 2. Measured X-ray diffraction θ-2θ scan data, (a) from sample A1 and (b) from sample B1. Insets on (a) and (b): respective XRD rocking curves for Al (111).

Download Full Size | PDF

3. Reflectance measurement and simulation

Figure 3(a) illustrates the measured reflectance spectra, plotted in solid lines, of samples A1 – A4 and a GaAs substrate. Let us first focus on sample A1, the pink solid line in Fig. 3(a). The reflectance is large than 90% except at wavelength less than ∼250 nm and the interband transition dip around 820 nm [25,26]. This broad-band high reflectance, particularly in blue to UV regime, is attributed to the high quality of MBE-grown films as that is difficult to achieve with other film preparation methods, such as conventional e-gun evaporation [2,5,11]. In fact, even with 40 nm thickness, the measured reflectance has been very close to that of an ideal bulk aluminum as stated in the following. The orange solid line in Fig. 3(a) is plotted by using Palik’s optical constants of bulk aluminum and equation (1) below [27].

 

Fig. 3. Experimental (solid lines) and simulated (dashed lines) reflectance spectra in 190-1000 nm wavelength range for (a) samples A1-A4 and a GaAs substrate, and for (b) samples B1-B3 and a Si substrate.

Download Full Size | PDF

In Fig. 3(a), with the decreasing film thickness, the measured reflectance gets lower as expected. The aluminum film becomes semi-transparent at the thickness of 20, 10 and 3 nm. Comparing with the spectra of GaAs substrate (black solid line), we can see that the specific peak features around 400 and 220 nm can be easily attributed to the GaAs substrate. With the decreasing film thickness, the interband transition dips around 820 nm fade out and the dip wavelengths red shift slightly. The measured spectra of sample A4 (blue solid line) is very similar to that of the GaAs substrate because of its ultrathin aluminum layer of 3 nm. Figure 3(b) shows the measured reflectance spectra (solid lines) from samples B1 – B3 and a Si substrate. A close look can find that the spectra of sample B1 is very close to that of sample A1 and the ideal bulk aluminum calculated by Eq. (1). A similar trend has also been observed with samples B2 and B3. Therefore, we can conclude that the nano-scale aluminum films of excellent optical quality have been successfully prepared on GaAs and Si substrates.

To comprehensively understand the experimental reflectance spectra, we have performed the corresponding numerical simulations with the FDTD (finite difference time domain) module in the commercially-available software Lumerical [28]. The simulation results are also plotted in dashed lines in Fig. 3 above. It is clear that the simulated spectra are in good agreement with the experimental ones. More importantly, with the simulation, we have learned three points worth-noting. (1) It is necessary to add a thin aluminum oxide layer on top of aluminum films to well fit the measured spectra, particularly for the samples A2, A3 and B2 with 10 or 20 nm aluminum. This is reasonable because aluminum film does oxidize in air and the native oxide layer serves as a protection layer to prevent the underneath aluminum from further oxidation. The thickness of the oxide layer is usually 2-4 nm so we put 2.56 nm Al₂O₃ layer in our simulations for samples A1-A3 and B1-B2. At the same time, their aluminum thickness was subtracted by 2 nm assuming that every 1 nm metallic aluminum generates 1.28 nm Al₂O₃. However, for samples A4 and B3, the 3-nm-thick aluminum samples, the Al₂O₃ layer was 1.92 nm and the aluminum subtraction was 1.5 nm. (2) The effect of surface roughness shown AFM iamges on the simulated reflectance spectra is insignificant, even in the UV regime. Therefore, in our simulation, the surface roughness was ignored. (3) The aluminum optical constants, that is, the wavelength-dependent n and k, used in our simulation were from Palik’s database [23] without any correction. As we can see in Fig. 3, the consistency is very good saying that the film quality is quite reproducible with the aforementioned growth recipe. This is important for further applications using these aluminum films for designing and engineering photonic and opto-electronic devices. We have also examined the validity of the optical constants reported by Cheng et al. as their values are available [20]. However, although the reflectance spectra of the thick aluminum films (samples A1 and B1) can be fitted very well, significant inconsistency in the semitransparent films (particularly for samples A2, A3, and B2) between the measured and simulated spectra has been observed, even with slight adjustments on the roughness, native oxide thickness and film thickness. This indicates the high sensitivity of semitransparent film on the applied optical constants but further investigations are certainly needed to clarify this issue.

With the simulation tool, aside from the reflectance, we can also obtain the transmission (T) spectra by simultaneously setting a monitor just underneath the aluminum films. The absorption (A) spectra have also been calculated by using $\textrm{A} = 1 - \textrm{R} - \textrm{T}$. Figure 4 shows the simulated results for samples A1 – A4 as an example. As shown in Fig. 4(a), as expected, the transmission increases with the decreasing film thickness. For sample A1 with 40-nm aluminum, the transmission is nearly equal to zero at all wavelengths as the aluminum is thick enough to absorb all photons to block the light penetration. On the other hand, the absorption spectra in Fig. 4(b) tells a slightly different story. Let us focus on the absorption bump around 800 nm as the blue-to-UV regime was dominated by the underneath GaAs substrate. The four samples have the similar peak absorbance in the range of ∼0.11 - 0.15 even the thickness of sample A4 is much thinner than that of sample A1. The largest absorbance unexpectedly occurs at sample A2 rather than sample A1. These observations can be explained by simply considering the multiple reflections of the two interfaces, the upper air/aluminum and the lower aluminum/substrate. With the decreasing aluminum film thickness, the aluminum film itself becomes semi-transparent so the light would reflect/transmit at the lower aluminum/substrate interface and then at the upper air/aluminum interface. The multiple reflections reduce the total reflectance and increase the optical path length as well as the absorption of the aluminum layer.

 

Fig. 4. Simulated (a) transmission and (b) absorption spectra for samples A1-A4.

Download Full Size | PDF

4. Conclusions

In conclusion, on GaAs and Si substrates, we have successfully prepared nano-scale epitaxial aluminum films with very smooth surface and excellent crystallinity. Normal-incident reflectance from UV to near infrared regime of these films have been measured and analyzed. The simulation results show very good consistency with the experimental data while considering the native oxide layer. Simulated transmission and absorbance spectra reveal the importance of multiple reflections between the two interfaces of air/aluminum and aluminum/substrate.