On-wafer fabrication of etched-mirror UV-C laser diodes with the ALD-deposited DBR

UV-C (100–280 nm) laser diodes (LDs) are strong candidates for health care applications, such as bio-/chemical sensing and sterilization devices. Because of their shorter wavelength and potential for high efficiency through integration,¹ UV-C LDs have attracted attention as high-precision small-size laser light sources, which can be alternatives to conventional low-efficiency solid-state lasers. In 2019, the world's first stimulated emission of an AlGaN-based UV-C LD at room temperature under pulsed current injection was demonstrated.² The recent availability of 2-in., high-crystal-quality AlN substrates, grown using physical vapor transport (PVT), has made a significant contribution to the realization of AlGaN-based UV-C LDs.^3,4 A continuous increase in the AlN wafer size is expected as continuous improvements in production are realized.⁵ Thus, it is very attractive to develop fabrication methods for AlGaN-based UV-C LDs that can take advantage of these larger wafers and allow both low cost mass production and monolithic integration in the future.

In Fabry–Pérot (F–P) type LD fabrication processes, one of the key steps is the formation of two highly reflective two mirror facets parallel to each other, which induce optical feedback. Although crystallographic cleaving^6–10 is a common method of obtaining nearly ideal mirror facets for LDs, cleaving is generally considered unattractive for mass production and monolithic integration. This is particularly true for AlN where the wafers are hard and the laser bars are relatively delicate, leading to either a poor yield or much higher cost due to severe constraints on handling. On the other hand, in the case of group III nitrides, there has been extensive investigation on the formation of mirror facets for heteroepitaxial layers grown on substrates that have different crystal orientations from GaN, making facet cleaving impossible. Several alternative techniques for forming a mirror facet on heteroepitaxial substrates have been reported, e.g., focused-ion-beam etching,¹¹ chemically assisted ion-beam etching,^12,13 and reactive ion etching (RIE).^14,15 Mirror formation by an etching method on a whole wafer¹⁶ can significantly reduce the production costs. The etched-mirror technique also enables monolithic integration, for example, LD/photodetector monitoring systems,¹⁷ LD/optical waveguide modulation systems,¹⁸ and high-power LD array systems¹⁹ formed on the same wafer. However, this on-wafer etched-mirror process typically faces two major technical issues: (i) difficulty in achieving a vertical, dry-etched facet, which results in a low mirror quality, and (ii) difficulty in depositing a high-quality, distributed Bragg reflector (DBR) on the etched facet. To form the vertical facet, alkaline-based selective wet etching has been used to expose smooth (1$1¯$00) planes of GaN following RIE,^20,21 and reduction in the lasing threshold has been reported.²² This method has also been demonstrated to be applicable to AlGaN-based materials²³ and device structures.²⁴ For on-wafer DBR deposition onto a vertical etched facet, atomic layer deposition (ALD), which is conventionally used to cover a hill-and-valley structure,²⁵ is thought to be appropriate. In addition, the deposition of various kinds of dielectric materials onto III nitrides by ALD has been demonstrated. These candidate dielectric materials, which can be used to form DBRs for UV-C LDs, include $Al2O3, HfO2$⁠, and $SiO2$⁠.^26,27

In this paper, we extend these methods to high-Al-content, pseudomorphic epitaxial layers grown on AlN substrates² and demonstrate on-wafer AlGaN based F–P LD fabrication with the mirror facets formed using a dry and wet etching method and the deposition of a high-quality DBR by ALD.

A 2-in. epitaxial wafer of the UV-C LD structure on a single AlN substrate provided by Crystal IS was prepared for device fabrication. A detailed description of the LD structure and the epitaxial growth process could be found in Ref. 2.

First, two steps of RIE were carried out to (a) expose the n-side cladding layer and (b) form a 5-μm-wide stripe ridge-geometry structure by partly removing the contact layer. A vanadium-based n-electrode and a 4-μm-wide nickel/gold-based stripe p-electrode were formed on the exposed n-side cladding and the top contact layer, respectively. In order to define a 400-μm-long laser cavity along the $⟨11¯00⟩$ direction, the LD structure was then etched down into the AlN substrate with $Cl2$ and $BCl3$ inductive-coupled-plasma reactive-ion etching (ICP-RIE). A 500-nm-thick $SiO2$ layer deposited by plasma-enhanced chemical vapor deposition was used as the etching mask. The wafer was then immersed for 7.5 min in a 25% tetramethylammonium hydride (TMAH) solution kept at $80 °C$ to selectively expose the (1$1¯$00) facets. Next, the DBR was deposited over the entire wafer by ALD and then ICP-RIE was carried out to remove the DBR and $SiO2$ material that had been deposited on the p-type and n-type contact metals. These steps were followed by pad metal formation, and then the processed wafer was finally singulated to measure the edge emission. A schematic drawing of a cross section parallel to the cavity axis near the mirror facet of a fabricated UV-C LD is shown in Fig. 1. All processes before singulation were carried out on 2-in. wafers using the facilities of the Center for Integrated Research of Future Electronics, Transformative Electronics Facilities (C-TEFs).

Figure 2 shows a cross-sectional scanning electron microscope (SEM) image of the AlGaN facet after (a) the dry etching and (b) after the wet etching. As shown in Fig. 2(a), the dry-etched AlGaN facet showed a tilt angle of $30°$ to the $[0001]$ direction of the AlN substrate. This tilt could have been caused by the low selectivity of the $SiO2$ mask and AlGaN. As shown in Fig. 2(b), the (1$1¯$00) facet was exposed and became a mirror facet perpendicular to the (0001) face of the AlN substrate after anisotropic wet etching with TMAH. The wet etching was reasonably uniform in the lateral direction according to SEM observation, despite the different compositions of the $AlxGa1-xN$ layers in the LD structure. Even if the etching time was further extended, the sidewall etching was self-limited by the (1$1¯$00) facet formation [Fig. 2(c)].

Four periods of $HfO2$ and $Al2O3$ were deposited using ALD (Veeco/CNT, Fiji G2) as the DBR. $Al2O3$ was selected as the lower refractive index material because the deposition rate of it is remarkably higher than that of $SiO2$⁠. The deposition was carried out under a $250 °C$ plasma condition. Tetrakis(dimethylamido)hafnium(IV) (TDMAH) and trimethylaluminum (TMA) were used as precursors for the $HfO2$ and $Al2O3$ films, respectively. The DBR was designed to exceed 60% reflectivity at the center wavelength of 275 nm for this LD structure, on the basis of the transfer matrix method (TMM) calculation. In this TMM calculation, the measured refractive index spectra of the ALD-deposited films of $HfO2$ and $Al2O3$ shown in Fig. 3(a), which were acquired with spectroscopic ellipsometry, were used. The reflectivity spectra of the $HfO2$/$Al2O3$ DBR deposited on a (0001)-oriented sapphire substrate were measured by ultraviolet-visible spectroscopy and compared with the TMM-calculated spectrum in Fig. 3(b). Excellent agreement between the measured and calculated reflectivity was obtained.

Figure 4 shows a cross-sectional bright-field transmission electron microscope (TEM) image taken near the mirror facet of the fabricated LD. The tilt angle of the etched mirror facet was $1.5°$ from the vertical. We find that the ALD-deposited DBR uniformly covered the etched facet even though the etched facet was almost perpendicular to the (0001) face of the AlN substrate. The thickness of the deposited film of 280 nm on the etched facet was slightly less than the designed total thickness of 293 nm, whereas the thickness of the film deposited on the (0001) face was exactly as designed.

The electrical characteristics of the fabricated LDs were measured with 100 ns pulsed current injection in a time period of 0.5 ms (duty cycle: 0.02%) at room temperature. The edge emission was measured simultaneously with a photon multimeter and a spectrometer. Figure 5 shows the current–voltage (I–V) and the emission power features, evaluated using the photon multimeter, as a function of the pulsed forward current (I–L) of the fabricated device. Lasing was observed at a forward current above 0.3 A, which corresponded to $19.6 kA/cm2$⁠. The slope efficiency was 120 mW/A. A sharp lasing spectrum peak at around 278.9 nm clearly emerged above the threshold current. The inset in Fig. 5 shows the edge emission spectrum measured by the spectrometer under a 0.38 A forward current.

Controlling the tilt angle of the mirror facets and the ALD-deposited high-quality DBR were the key to achieve lasing in the LD. The tilted AlGaN facet is not suitable as a mirror facet for LDs because the reflectivity decreases as the tilt angle increases.^13,28 The reflectivity of a mirror facet ideally perpendicular to the cavity is calculated to be 18.7% by the method described in Ref. 28. With a larger tilt angle, reflectivity decreases, reaching 0% at around $25°$ for this LD structure. The result indicates that no reflection was expected by a $30°$ tilted mirror facet before wet etching. The fabricated mirror facet was tilted by $1.5°$⁠, and is expected to have a reflectivity of 18.3%, which is only slightly different from the ideal case, i.e., 18.7%. The present results indicate that wet etching using TMAH is a key technique in forming vertical mirror facets of AlGaN-based UV-C LDs. Nevertheless, even with highly vertical mirror facets, we could not achieve lasing for devices without a DBR on the same epitaxial wafer at injection current densities up to $100 kA/cm2$⁠. Thus, it appears that the increased reflectivity provided by the ALD-deposited DBR made a significant contribution to reducing the lasing threshold. The reflectivity of the fabricated DBR on the mirror facet at 278.9 nm was calculated to be 49.6%. However, as the actual thickness of each layer of the fabricated DBR is slightly different from the designed values, a reflectivity of 65.6% could be achieved. Further improvement of the threshold current can be expected by introducing an appropriately designed DBR structure which takes into account the thinner layers deposited on the side walls. Another possibility for improving reflectivity is to use a dielectric with a lower refractive index than that of $Al2O3$⁠, such as $SiO2$⁠. In addition, note that the thickness distribution of ALD-deposited dielectrics was less than 2% across the entire 2-in. wafer, leading to a distribution of reflectivity within 5%. The ALD technique could be a desirable candidate for DBR formation, especially for mass production on a larger AlN wafer.

In conclusion, we demonstrated an on-wafer AlGaN-based UV-C LD fabrication method. Vertical mirror facets, formed by a combination of dry and wet etching, and a DBR that was uniformly coated on the vertical facets contributed to lowering the lasing threshold. Lasing was observed at room temperature under pulsed current injection in the fabricated LDs. We propose this on-wafer fabrication approach as contributing to the development of a low-cost, highly integrated UV-C LD production technology based on AlGaN.

The authors would like to acknowledge Mr. Kazuhiro Nagase and Dr. Naohiro Kuze of Asahi Kasei Corporation for their invaluable discussion and considerable support. The authors also wish to acknowledge Mr. Tomohiro Shinagawa of C-TEFs for his significant contribution to the development of the ALD-deposited DBR process.