Single-crystalline aluminum film for ultraviolet plasmonic nanolasers

Significant advances have been made in the development of plasmonic devices in the past decade. Plasmonic nanolasers, which display interesting properties, have come to play an important role in biomedicine, chemical sensors, information technology, and optical integrated circuits. However, nanoscale plasmonic devices, particularly those operating in the ultraviolet regime, are extremely sensitive to the metal and interface quality. Thus, these factors have a significant bearing on the development of ultraviolet plasmonic devices. Here, by addressing these material-related issues, we demonstrate a low-threshold, high-characteristic-temperature metal-oxide-semiconductor ZnO nanolaser that operates at room temperature. The template for the ZnO nanowires consists of a flat single-crystalline Al film grown by molecular beam epitaxy and an ultrasmooth Al2O3 spacer layer synthesized by atomic layer deposition. By effectively reducing the surface plasmon scattering and metal intrinsic absorption losses, the high-quality metal film and the sharp interfaces formed between the layers boost the device performance. This work should pave the way for the use of ultraviolet plasmonic nanolasers and related devices in a wider range of applications.


Nanolaser fabrication
A. Single-crystalline Al films epitaxy: The single-crystalline Al films were grown on a GaAs (100) substrate by Varian Gen II solid-source molecular beam epitaxy system (MBE). We first grew a 200-nmthick undoped GaAs buffer layer serve as the template for the following Al growth. The surface was turned from As-rich into Ga-rich before cooling down to room temperature.
The wafer was then kept in the ultra-high vacuum chamber to prevent the surface from oxidation. When the residual arsenic pressure was pumped down to less than 1x10 -10 torr, a 100-nm-thick Al layer was grown at ~ 0 °C with a growth rate of 0.05 nm/s.

B. Poly-crystalline Al films evaporation:
The poly-crystalline Al films were evaporated on the GaAs (100) substrate by an e-gun evaporator. Before the GaAs substrate was loaded into the evaporator chamber, a de-oxidation process using HCl etching solution was performed to remove the native oxide on the GaAs surface. After the standard clean process, the GaAs substrate had been loading in to the electron-gun chamber immediately. As the chamber pressure was less than torr, we evaporated a 100-nm-thick Al film at room temperature with an evaporation rate of 0.3 nm/s.

Complex dielectric constants extraction for the two Al films
The optical properties such as absorption, refraction and transmission of material can be described by its complex dielectric constants. As shown in Figure 3 (c) in the main text, the reflectivity of the PC-Al film is lower than that of the SC-Al one, especially in the wavelength range below 450 nm. To quantitatively explain the reflectivity discrepancy, we extracted the optical parameters of these two films by 3  10 6 5 fitting the measured reflectivity spectra with the Drude-Lorentz model. The complex dielectric constant for metals can be expressed as below 1, 2, 3 where and are the plasmon resonance frequency of Al and Lorentz model resonance frequency, respectively. is the angular frequency of the incident light.
is the background permittivity. and are the damping coefficients in Drude and Lorentz models, respectively. The refractive index n and κ can then be obtained accordingly by the following relations. (2) And the reflectivity (R) can be calculated from Eqs. (2) and (3) with .
By using Eq. (4), we fitted the whole spectrum range of Figure 3 (c) in the main text.
The fitting parameters and their values are listed in Table S1. Table S1. Drude-Lorentz fitting parameters of SC-Al and PC-Al films.
The fitting results are plotted in Figure S1. Figure S1 (a) shows the SC-Al fitting result.
The measured reflectivity spectrum (red) can be well fitted by Drude-Lorentz model fitting equation (blue) in the whole spectrum range. However, as Figure S1  [S1] are also plotted. Figure S2 (a) shows the real part of complex dielectric constant of our SC-Al film is very close to theirs but, as shown in Figure S2 (b), our imaginary part is slightly higher. [S1] as a function of wavelength.

Nanolaser numerical simulation
We used the commercial software COMSOL, the mode solver of the finite-  As the Figure S3 (b) shows, once the one-side length (d) of the hexagonal crosssection of the ZnO nanowire is below 55 nm, only the fundamental plasmonic mode can sustain in the cavity. Since the typical one-side hexagon length of our ZnO nanowire was 30 nm, only the F mode can be observed in our case. In addition, far-field polarization measurement in the Fig. 6 (b) of main text showed that the polarization direction was along the nanowire, which is the distinct feature of the fundamental surface plamonic mode in our nanolaser structures.
10 Figure S3 (b). The confinement factors of various plasmonic cavity modes. Figure S4 (a) shows the AFM surface morphology of a nanolaser template consisting of PC-Al film and a 5-nm-thick SiO2 spacer layer. The surface is quite rough with an RMS roughness of 2.14 nm. Figure S4 (b) shows the measured powerdependent emission spectra from a nanolaser at 77 K, indicating a lasing threshold power density of 54.5 mJ cm -2 . The inset shows that, below the lasing threshold, a broad emission spectrum of the 1.24-μm-long ZnO nanowire was observed. A clear lasing peak at 368 nm with the linewidth narrowing down to 0.8 nm was seen above threshold.

Experimental results of nanolasers with SiO 2 spacer layer
Owing to the rough surface morphology, the SPs suffer serious extrinsic scattering loss and intrinsic ohmic damping loss. Even more importantly, the rough surface could 11 drastically reduce the coupling efficiency between the excitons and SPs so a very high threshold power density was obtained.  better than the previous one. Figure S5 (b) shows the measured power-dependent emission spectra from a 1.11-μm-long ZnO nanolaser at 77 K. The SC-Al film could reduce the SPs intrinsic damping loss and grain boundaries scattering loss but the surface roughness caused considerable extrinsic SPs scattering loss. The inset to the right of Figure S3 shows the fluctuated spontaneous and lasing spectra with two obvious lasing peaks at 369 and 371 nm. The integrated emission intensity suggests a lasing threshold power density of 10.5 mJ cm -2 , which is about 5-fold smaller than ZnO nanowire lying on the PC-Al/SiO2 template.