1. Introduction

InP-based devices have started to dominate opto-electronics because lasers and related devices with InGaAsP/InP heterostructures are suitable for integrated optics and low-loss fibre communications around the 1.5μm wavelength region where attenuation is about 0.2dB/km. The fabrication of integrated optoelectronic devices necessitates pattern transfer techniques with a high degree of precision and a variable anisotropy. Wet chemical etching is a simple technique that offers high selectivity and prevents deep damage to QW layers when compared with dry-etching techniques. However, features with dimensions less than 3μm are usually patterned using dry etching techniques because wet etching rates can be the same in all directions (isotropic) which causes undercut, and thus wet etching can hardly be used reliably below 3μm feature size.

It is necessary to use anisotropic (vertical) dry-etching in order to produce a uniform etch profile along the entire structure because sidewall verticality (anisotropy) is extremely important for applications such as etched laser facets. Smooth structures are also required because surface roughness can cause short-circuit between patterns that can fail the operation of lasers. Therefore, various dry etching techniques such as plasma etching [1,2], reactive ion etching (RIE) [3,4], ion beam etching (IBE) [5,6], reactive ion beam etching (RIBE) [7,8] and chemically assisted ion beam etching (CAIBE) [9,10] have been successfully used to fabricate InP-based devices to date.

Of these methods, reactive ion etching (RIE) and chemically assisted ion beam etching (CAIBE) are the most frequently used ones. RIE of InP-based structures has been reported widely in the literature using non-corrosive/non-toxic hydrocarbon chemistries (CH₄/H₂) that provide residue-free morphologies over Cl₂-based halogen chemistries [11,12]. Smooth etching of InP has also been demonstrated using N₂ and N₂/O₂ [13] plasmas. But these plasmas not only reduce etch rates significantly (3 to 8nm/min) when compared with CH₄/H₂ (typically 40–50nm/min) mixtures but also generate a certain amount of sidewall etching.

The second method used for dry etching is CAIBE in which flexibility for profile definition is possible by varying the angle of incidence of beams. In this process, the inadequate performance, i.e. the surface and sidewall roughness, of pure Ar ion beam etching (IBE) can be eliminated by using a mixture of Ar/CH₄/H₂[14], N₂/H₂/CH₄[14], Ar/H₂[15] or N₂/O₂ [10] gas(es). CAIBE of InP-based structures has been performed using Ar/H₂/CH₄ and N₂/H₂/CH₄ atmospheres by using an rf inductively coupled plasma (ICP) ion source [14]. Although the roughness in the Ar process was more severe than that for the nitrogen counterpart, the anisotropy was higher for the Ar process [14]. Vollrath and others produced good quality laser facets by CAIBE using mixtures of nitrogen and oxygen [10].

Here, fabrication of long wavelength (1.5μm) ridge-waveguide semiconductor lasers that overcomes short lifetime and low efficiency of oxide-stripe lasers are presented from InGaAs/InGaAsP/InP material system. Comparative experimental results of the lasers are discussed using the three etching (RIE, CAIBE and wet) methods. Anisotropy and surface roughness of the etched structures are investigated using scanning electron microscopy (SEM) and Atomic Force Microscopy (AFM) techniques. The fabricated lasers with 2μm waveguide-width are characterized using light-current-voltage (L-I-V) and spectra measurements. Finally, wet and RIE etched lasers with a waveguide-width of 20μm are studied comparatively.

2. Experiment

2.1 Material structure

The wafer was grown by metal organic vapour phase epitaxy (MOVPE) at the EPSRC III-V central facility at the University of Sheffield. The epitaxial layers were grown on an n⁺ Si doped InP substrate. From the substrate, the layer specifications are as follows: a 1μm n-type InP lower cladding layer, a 295nm waveguide core, a 1.2μm p-type InP upper cladding layer, a 50nm InGaAsP transition layer, and finally an In_0.53Ga_0.47As contact layer.

The active layer contains five 60Å In_0.53Ga_0.47As quantum wells with six 120Å InGaAsP barriers. The wells are surrounded in both directions with a step graded index region, consisting of a 50nm InGaAsP and an 80nm InGaAsP quaternary layers. All the layers are lattice matched to InP.

2.2 Fabrication of InGaAs/InGaAsP/InP lasers using RIE and CAIBE

We prepared ridge-waveguide InGaAs/InGaAsP/InP lasers in three steps: Etch mask fabrication, etch process to define ridge and metal contact deposition. The etch mask was formed by classical photolithography-defined patterns (2μm straight waveguides) of resist on 160nm and 50nm thick layers of sputter-deposited SiO₂ and Ni. These patterns were then transferred to the substrate by RIE with C₂F₆, followed by an O₂ plasma to strip the resist. In the second step, the ridge waveguide was formed by etching 1.2μm of the top layers with RIE or CAIBE as described below.

RIE of InGaAs/InGaAsP/InP material system was carried out in a Plasma Therm dual electrode system operating at a frequency of 13.56MHz. The rf power was applied to the bottom electrode while the other electrode and the chamber wall were kept at ground potential. Both electrodes and the chamber wall are made of aluminum. During etching, the samples were placed on the bottom electrode which was un-cooled. CH₄/H₂ gases were introduced through small holes around the lower electrode and evacuated via a 3.2cm diameter hole in the center. In CH₄/H₂ discharge, hydrogen atoms, organic radicals, and ions are produced that interact on InP surfaces to etch the material [16].

CAIBE was done with an Ion Tech Ltd ion beam etching system where a saddle-field DC discharge ion gun was used to generate the ion beam. Ions were generated in the discharge chamber by electron bombardment of the neutral Ar atoms. To ignite the discharge, a potential difference was applied between cathode and anode. A combination of reactive gases (H₂/CH₄) were introduced directly to the sample to improve etch quality. Samples were rotated during etching to enhance etch uniformity. In both etching techniques, several parameters were optimized to produce the highest level of anisotropy and surface smoothness without compromising the etch rate. The optimized parameters are tabulated in Table 1.

Table 1. Optimized parameters during RIE and CAIBE etching of InGaAs/InGaAsP/InP

View Table | View all tables in this article

While a maximum etch rate of 50nm/min was observed in RIE, it was 40nm/min in the CAIBE. The etch depth was measured to be 1.2μm for the both processes. In RIE, a polymer layer accumulated in the chamber, requiring oxygen plasma (rf power 150W, 50sccm) cleaning after each run. In RIE and CAIBE, etching occurs by both chemical due to the formation of volatile products and physical because of sputtering of surface by the ions.

In the final stage of the fabrication, the two-layered mask (SiO₂/Ni) was removed by H₂SO₄ and HF prior to the deposition of SiO₂. After applying second photolithography and contact window mask, the SiO₂ layer on the waveguide was removed by HF. Finally, a 20nm titanium, 20nm platinum and 150nm gold metal alloy were used in the p-contact recipe, followed by the deposition of a 14nm gold, 14nm germanium, 14nm gold, 11nm nickel and 200nm gold on the n-side. Annealing at 400°C for 1 min. was the optimum condition to obtain the best lasing characteristics since otherwise the devices gave very poor characteristics.

2.3 Fabrication of InGaAs/InGaAsP/InP lasers using wet etching

Two etching steps were applied to fabricate wet etched ridge waveguide lasers. The constituent of the first etch (etch 1) is a mixture of acetic acid, hydrochloric acid (HCl), hydrogen peroxide (H₂O₂) and water with a ratio of 40:20:3:15. To form the ridge, a second etch step (etch 2) containing orthophosphoric acid (H₃PO₄) and HCl (3:1) mixture was used.

To fabricate the devices, first 5μm wide waveguide mask was applied followed by S1805 photoresist spinning as an etch mask. Then, etch 1 was used to remove the contact layer. Prior to Al₂O₃ (300nm) deposition, the ridge (etch depth=1.2μm) was defined by etch 2. Following the removal of the resist in the acetone, p and n-contact metal deposition was performed by using the same compositions as that of the dry-etched ones.

3. Results and discussion

3.1 SEM and AFM measurements of the fabricated devices

SEM and AFM images of the RIE and CAIBE etched structures are given in Fig. 1 and Fig. 2.

 

Fig. 1. (a) Surface/sidewall structure, (b) AFM measurement, and (c) etch profile of the RIE etched lasers

Download Full Size | PDF

 

Fig. 2. (a) Surface/sidewall structure, (b) AFM measurement, and (c) etch profile of the CAIBE etched lasers

Download Full Size | PDF

Surface roughness in both RIE and CAIBE was quantified by AFM technique. The rms surface roughness was obtained to be 10.8±1.7nm with the CAIBE in Fig. 2(b). A dramatic improvement of the surface morphologies of the etched surfaces was recorded in the case of RIE where the rms roughness was only 2.1±0.4nm as illustrated in Fig. 1(b). This was very close to the initial roughness of an un-etched InP surface which was about 1.3±0.4nm.

Sidewall profiles of the RIE and CAIBE etched structures are shown in Fig. 1(c) and Fig. 2(c) respectively. In Fig. 1(c), a highly anisotropic (~90°) sidewall profile to the surface parallel in the RIE was achieved while sidewall angle (anisotropy) was 83° in the CAIBE process (Fig. 2(c)). As a result, the best surface quality and anisotropy were obtained in the case of RIE. How these results affect the L-I-V characteristics is discussed below.

3.2 L-I-V characteristics of the fabricated lasers etched by RIE, CAIBE and wet chemical

The light-current-voltage (L-I-V) characteristic of the RIE and CAIBE etched 2μm wide and 400μm long ridge-waveguide lasers, shown in Fig. 3(a) and (b), was measured in order to characterize the operation of the fabricated lasers. A pulse generator, which generates electrical pulses with pulsewidth of 500ns and pulse period of 100μs, was used for driving the lasers. The output of the lasers was coupled either to an Anritsu optical powermeter or an Advantest optical spectrum analyzer via an optical fiber.

 

Fig. 3. L-I-V characteristics of 2μm wide and 400μm long (a) RIE and (b) CAIBE etched lasers, and (c) spectral characteristic for both

Download Full Size | PDF

A spectral characteristic of the device is also illustrated in Fig. 3(c) in which center wavelength was recorded to be 1.56 μm. In Fig. 3(a), a threshold current as low as 30mA and 13mW maximum optical power at room temperature was observed for the RIE etched laser. However, the CAIBE laser yielded a threshold current of 65mA and output power of just about 7.5mW. This shows that CAIBE etched lasers have a higher threshold current than the RIE counterparts which may be caused by roughness induced loss in the CAIBE process. In the case of CAIBE, there was also a decrease by 12.5% in the external quantum efficiency. It was assumed that the RIE process caused lower damage than that of CAIBE possibly due to less physical sputtering.

The forward bias characteristics (I-V) of the RIE and CAIBE etched lasers were also taken as depicted in Fig. 3(a) and (b). The typical turn-on voltage of 1.2V in InGaAsP/InP laser diodes [18] was recorded for the RIE etched lasers (Fig. 3(a)) in our measurements. However, this was reduced to 0.75V in the CAIBE etched lasers (Fig. 3(b)) which implies that either the optimized RIE etch did not cause the hydrogen passivation effect (the passivation of electrical activity) on the sidewalls or it was removed by the annealing process. The above results shows that RIE etched lasers are superior to the CAIBE etched ones.

Since the fabricated wet etched lasers whose feature sizes were less than 3μm produced very poor characteristics due to probably undercut caused by the wet etch, a comparison of wet and RIE etched semiconductor lasers with 20μm wide waveguide is given in Table 2.

Table 2. Material parameters of wet and RIE etched InGaAs/InGaAsP/InP lasers with 20μm width and 400μm length

View Table | View all tables in this article

Table 2 demonstrates that a slight increase in both the internal loss (α_i) and the threshold current density was observed in the case of RIE. This implies that RIE caused a slight damage, which was not removed by the post etch annealing process. However, higher internal quantum efficiency (η_in) and slightly better turn-on voltage were obtained from the RIE etched lasers. Overall results obtained shows that RIE can be preferable to the other etching techniques to produce highly vertical and smooth structures that are required for laser mirrors.

4. Summary and conclusions

Fabrication and characterization of RIE and CAIBE etched 2μm wide ridge-waveguide long wavelength (1.5μm) semiconductor lasers from InGaAs/InGaAsP/InP material system have been presented comparatively using some characterization techniques such as SEM, AFM, spectra and L-I-V. In addition, 20μm wide devices were fabricated by using wet and RIE etching to obtain comparative results. Prior to the fabrication of the dry-etched lasers, much attention was paid to optimize the etching conditions of RIE and CAIBE processes to produce efficient and reliable lasers. Highly smooth and anisotropic surface morphologies were obtained by using RIE that produced just 2.1±0.4nm rms surface roughness and ~90° sidewall angle while CAIBE resulted in an rms roughness of 10.8±1.7nm and an anisotropy of 83°. Thus, for 2μm wide devices performance of the RIE etched devices was observed to be better than the CAIBE etched ones. This is likely due to the higher damage and surface roughness caused by the CAIBE. Since 2μm-wide wet etched lasers failed, a comparison was made for 20μm RIE and wet etched devices where wet etching resulted in a slightly better performance over RIE because of probably a slight damage caused by the RIE. However, as a result, highly vertical and smooth structures are required for applications such as corner mirrors in optical crosspoint switches and etched laser mirrors, and therefore the advantage of using dry etching techniques preferably RIE largely outweighs the risk of damaging the material.