1. Introduction

Reliable low cost and wideband wavelength tunable semiconductor laser is a key enabling technology for optical networks. To meet the requirement of the optical systems, several types of techniques have been developed to improve the performance of wavelength tunable lasers, including external cavity lasers [1], distributed Bragg reflector (DBR) grating based lasers [2,3], and arrayed DFB lasers [4,5]. Although the performance of these devices can satisfy the system requirements, the cost of the devices is very high due to the expensive fabrication or assembling processes.

To reduce the cost, Engelstaedter et al. [6,7] developed a wavelength tunable laser using periodically etched slots as distributed reflectors. Unlike the previously reported DBR based lasers [2,3], the grating of the laser is defined by introducing etched slots on the ridge of the laser waveguide. The periodical slots can generate a comb like reflection spectrum with a specific Free Spectral Range (FSR) determined by the optical length between the adjacent slots. By introducing two groups of slots with different FSRs as the front and the rear reflection mirrors in the laser, one can utilize the Vernier effect between the slot mirrors to enlarge the total FSR of the device and the tuning range. Compared with the Bragg grating based laser such as SG-DBR or DS-DBR [2,3], the merit of the slot approach is its fabrication simplicity. While epitaxial regrowth is needed in the fabrication of DBR lasers, the slots can be fabricated by standard UV lithography and no epitaxial regrowth is required. However, the surface etched slots of about 1µm wide limited by standard contact lithography introduce a large scattering loss which increases the laser threshold and degrades its performance. For example, in Byrne’s simulation, to generate about 1% intensity reflection with a slot of about 1µm wide, only 50% of the optical power is transmitted to the other side, which implies a nearly 3dB loss per slot [7]. A possible way to reduce the optical loss of the slots is to reduce the slot width to the deep-submicron domain [8]. However, one needs to use a high resolution lithography method such as the electron-beam lithography to fabricate deep-submicron structure, which increases the cost.

In this paper, we first analyze the reflection, transmission and loss of a slot as a function of its geometrical parameters, showing the low-loss advantage of deep-submicron slots. We then report a newly developed pillar reversing process for fabricating deep-submicron slots by using standard UV-lithography. Based on the process, slotted Fabry-Perot and tunable slotted grating lasers are fabricated. Experimental results show that the devices have good performance in terms of threshold, side mode suppression ratio and tuning range. The process can also be used for fabricating other multi-functional photonic integrated devices such as the Q-modulated laser [9] and V-coupled-cavity laser [10].

2. Numerical analysis of the deep-submicron slot

Figure 1(a) shows the schematic of a slotted grating laser. Two distributed slot gratings with slightly different periods are used as cavity mirrors for the laser. Each of the gratings comprises of several periodically distributed slots by dry etching process [7]. Figure 1(b) is the optical microscope picture of the laser. The device is controlled by three independent injection currents: gain current I_g, the front and rear grating current I_f and I_r. The wafer used for fabrication is grown by metal organic chemical vapor deposition (MOCVD) on n-InP substrate. The wafer comprises of an active region with five compressively strained In_0.8Ga_0.2As_0.8P_0.2 quantum wells sandwiched between 50nm 1.25Q p-InGaAsP and 50nm n-InGaAsP Graded Index Separate Confinement Heterostructure (GRINSCH) layers from 1.25Q down to 1.05Q gradually. The p-InP cladding layer is 1.65µm thick and covered by 200nm InGaAs contact layer. The slot is etched through the cladding but do not penetrate into the QW layers of the wafer. In a slotted grating, the total reflection coefficient r_t can be expressed as [7]

 

Fig. 1 Schematic (a) and optical microscope picture (b) of a slotted distributed reflector laser.

Download Full Size | PDF

Figure 2(a) depicts contour map of the intensity reflection R(d,w) by varying the etching depth d and the slot width w, calculated by using two-dimensional Method of Lines (MOL) [11]. The merit of MOL is that the time complexity is proportional to the number of sections in light propagation direction rather than the total length. Hence, the length of the left section and the right section can be long enough so that the reflection and transmission field are convergent to the fundamental mode of the waveguide. The polarization in the calculation is TE and the wavelength is set to be 1530nm. The oscillation of the contour is caused by the interference between the two facets of the slot which results in a weak width dependence of the reflectivity [12]. For a specific contour curve under which the reflectivity of the slot equals to the contour value R₁, the coordinate [w,d] of the curve points satisfy some implicit function w(d) which is defined by R(d,w) = R₁. Hence, to obtain a fixed reflection value R₁, the transmission T(d,w) of the slot can be considered as a function $T_R^{}\text{(w)}$of single variable w. Figure 2(b) sketches the function $T_{R_{}}^{}\text{(w)}$ with different reflection values from 0.8% to 1.4%. One can see that to generate sufficient intensity reflection around 1.4%, the transmission of a deep-submicron slot (width <300nm) is very desirable compared to a wide slot that can be fabricated by conventional photolithography (width >800nm). The narrower the slot, the lower the optical loss. Here the loss is defined as the scattering loss other than the reflection and transmission, i.e. (1-R-T), or 10log₁₀(R + T) in dB. For example, the transmission of a 800nm-slot is about 50% while it is 85% for a slot width of 120nm. A 2.3dB loss reduction per slot is therefore achieved by using deep-submicron slots.

 

Fig. 2 (a) Contour map of the intensity reflection with respect to the slot width and depth (rectangular slot); (b) Comparing of the transmission of rectangular and non-rectangular slots as a function of the slot width with the reflection varies from 0.8% to 1.4%.

Download Full Size | PDF

In the deep-submicron domain, it is very difficult to fabricate a rectangular shaped slot by dry etching methods due to its small feature size and high aspect ratio [13]. To make the simulation closer to the actual situation, we also calculated the reflectivity R(d,w) of a non-rectangular slot with a shape fitted to the experimental profile. Figure 3 shows a typical profile of an etched slot. A fitting shape is constructed by using arc tangent curves at the bottom half of the slot, which are smoothly connected to straight lines at the top half of the slot. Here, w denotes the width at the top of the slot and d is the vertical distance from the top of the slot to its bottom tip. As we can see in Fig. 3, the fitting shape matches well the actual profile.

 

Fig. 3 SEM picture of an etched slot of about 240nm with a fitting profile (dashed line).

Download Full Size | PDF

Figure 4(a) depicts the contour map of the intensity reflection of a non-rectangular slot. Compared with the results in Fig. 2(a), one can see that, each contour curve has moved towards the right side because the narrower tip reduces the effective depth of the slot. The transmission $T_{\text{R}}^{\text{n}}\text{(w)}$of the non-rectangular slot and $T_R^r\text{(w)}$ of the rectangular slot are plotted together in Fig. 4(b) for comparison.

 

Fig. 4 (a) Contour map of the intensity reflection with respect to the slot width and depth (non-rectangular slot); (b) Comparing of the transmission of rectangular and non-rectangular slots as a function of the slot width with the reflection varies from 0.8% to 1.4%; (c) Effective width of a non-rectangular slot.

Download Full Size | PDF

From the Fig. 4(b), it is clear that to obtain the same reflectivity R₁ and transmission T values, the width w of a non-rectangular slot is much larger than that of a rectangular slot. For example, to achieve a reflectivity of 1.4% and a transmission of 0.85, the width of the rectangular slot is about 120nm which is half of that of a non-rectangular one.

Therefore, we can define an effective width w_e(w) of the non-rectangular slot implicitly through the equation $T_{{\text{R}}_{}}^{\text{r}}\text{(}{w}_e\text{)}\text{=}{T}_{\text{R}}^{\text{n}}\text{(}w\text{)}$, where r denotes the rectangular slot shape and n denotes the non-rectangular slot shape. Figure 4(c) shows the relation between the effective width and the actual top width at reflectivity of 0.8% and 1.4%. We can see that the effective width of the non-rectangular slot is almost half of the top width of the slot, and it is independent of the reflectivity R₁ within the range of 0.8%-1.4%. Therefore, the non-rectangular nature of the deep-submicron slot does not affect the optical characteristics of the slots in the low reflectivity regime. The wider top width and non-stringent requirement on the shape are very helpful to reduce the fabrication difficulties.

3. Device fabrication

The fabrication steps of the deep-submicron slots are shown in Fig. 5 . First, a 1.5µm thick SiO₂ layer is deposited on the wafer by plasma-enhanced chemical vapor deposition (PECVD). According to the designed positions of the slots, ridge patterns are formed on the SiO₂ layer by standard UV-lithography followed by a CF₄/CHF₃ inductively coupled plasma (ICP) dry etching step. The etching depth of SiO₂ is 500nm. The resulting profile of the SiO₂ layer is shown in Fig. 5(a). Then a 400nm thick chromium (Cr) film is sputtered all over the SiO₂ ridges as shown in Fig. 5 (b). A Cl₂/O₂ ICP etching is then performed to remove the Cr film on the top surface. Due to the anisotropic nature of the ICP etching, Cr on the sidewalls of the SiO₂ ridges remains after the ICP etching, as shown in Fig. 5(c). A further step of SiO₂ ICP etching is carried out to the bottom of the SiO₂ layer, leaving pillars of Cr/SiO₂ on the InP wafer_, as shown in Fig. 5(d). Subsequently, a thick photoresist is spin-coated on the surface of the wafer which planarizes the pillars, as shown in Fig. 5(e). After an O₂ etch-back step (Fig. 5(f)), the SiO₂ pillars are removed by buffered oxide etcher (BOE: NH₄F/HF) together with the Cr pillars over them, as shown in Fig. 5 (g). This forms a photoresist pattern for the deep-submicron slots. The slots are then etched into the InP multi quantum well (MQW) wafer by a Cl₂/H₂/CH₄/Ar ICP etching using the photoresist mask which is subsequently removed, as shown in Fig. 5 (h). The dry etching is stopped above the MQW layer. By controlling the thickness of the Cr film on the side walls of the SiO₂ ridges, which is relatively easy, the slot width can be accurately controlled to tens of nanometers.

 

Fig. 5 Fabrication steps of the deep-submicron slots.

Download Full Size | PDF

Figure 6 shows the SEM images at different stages of the fabrication process. Figure 6(a) is the deep-submicron pillar comprised of Cr over SiO₂. Figure 6(b) shows the deep-submicron slot pattern formed on photoresist mask. The final etched slot is shown in Fig. 6(c). The top width of the slot is about 240nm, corresponding to an effective width of 120nm.

 

Fig. 6 (a) Deep submicron pillars comprised of Cr over SiO2; (b) Deep submicron slot pattern formed on photoresist mask; (c) Deep submicron slot on InP MQW wafer. All figures have the same scale.

Download Full Size | PDF

After the formation of the deep-submicron slots, a 400nm SiO₂ layer is deposited by PECVD as an etching mask for the 3µm wide ridge laser waveguide. After a conventional photo-lithography and mask etching step, ridge waveguides are patterned using a Cl₂/H₂/CH₄/Ar ICP etching. The SiO2 mask can protect the slot during the ridge waveguide etching. Su8 is then spin-coated over the wafer for planarization. After curing and etch-back, p-type metal contact comprising Ti/Pt/Au is deposited by sputtering with standard lift-off process. Su8 in the slots is cleaned by ICP dry etching after p-type metal contact patterning. The wafer is then thinned and polished down to 120µm. Back side metallization of Ti/Pt/Au is performed and the wafer is cleaved into laser chips. No coatings are applied on the cleaved facets.

4. Device characteristics

To verify the quality of the deep-submicron slot, we first fabricated and tested slotted FP lasers [13] with 9 slots distributed along a 427µm long cavity. In the slotted FP laser, the optical length between any two adjacent slots are all equal to an integer multiple times of quarter wavelength. The threshold gain spectrum of the laser is calculated by using the Transfer Matrix Method (TMM) [14]. The reflection and transmission coefficients of the slots used in the TMM calculation are obtained from the 2D method of lines model. The reflection coefficient of the facets is set to be 0.5255 corresponding to the actual value of the uncoated cleaved facets. The positions of the slots are optimized using the genetic algorithm [15] for maximizing the side mode suppression ratio (SMSR), with the fitness function chosen as the threshold gain difference between the main mode and the side mode within a wavelength range of 100nm.

The cleaved laser chips are mounted on an AlN heat sink and then loaded onto a thermoelectric cooler (TEC) for test. Figure 7(a) shows the power of the laser output as a function of the injection current. The threshold current is 22mA and the slope efficiency of the device is about 0.18W/A. The performances are better than the device of [16]. with similar slot number and cavity length in which a multi-wavelength laser array with a threshold value of 39mA and a slope efficiency of 0.1W/A was reported. The emission spectrum of the laser biased at 60mA is shown in Fig. 7(b). A SMSR of 39dB is achieved. The lower threshold and higher slope efficiency of our device confirms that the deep-submicron slots have much smaller optical loss.

 

Fig. 7 (a)Output power of the laser as a function of the injection current; (b) Emission spectrum of the laser biased at 60mA.

Download Full Size | PDF

A tunable laser based on periodically distributed slots is then fabricated and characterized based on the proposed process. The length of the laser gain section is 400µm. The period of the front grating is 108 µm and the one of the rear grating is 98 µm, corresponding to a free spectral range of 3nm and 3.3nm respectively. There are 6 periods in each of the gratings. Figure 8(a) shows the emission spectrum of the laser. A SMSR of 43dB is achieved. The current injected into the front grating, rear grating and the gain section are s achieved. The current injected into the front grating, rear grating and the gain section are I_f = 77mA, I_r 75mA, and I_g = 35mA respectively. By changing the injection current of the front or rear grating, the laser wavelength can be switched discretely for 11 channels. The total tuning range is about 38nm and the overlapped emission spectrum is shown in Fig. 8(b). The SMSR of all channels are larger than 36dB. The side mode suppression performance of the deep-submicron laser is better than the device reported in [7], in which the SMSR of 30dB (40dB max) within 20nm tuning range is achieved by using micron-sized slots.

 

Fig. 8 (a) Emission spectrum of the laser of one channel; (b) Overlapped spectrum of all channels.

Download Full Size | PDF

When only one of the electrodes in the front and rear reflectors is tuned, the spacing between the discretely switched channels is not uniform, which varied from 1.8nm to 4.5nm, as shown in Fig. 8(b). This is because the lasing wavelength also needs to satisfy the resonant condition of the cavity comprising the middle gain section. To access all wavelengths matching the ITU grid, it is necessary to control all three electrodes, similar to the case of SG-DBR lasers. This is a disadvantage of this type of lasers as compared to simpler V-coupled-cavity laser in which uniformly spaced discrete channel switching can be controlled with a single electrode [17].

5. Conclusions

We have introduced a new process for fabricating deep-submicron etched-slots using standard UV-lithography. Both single mode slotted FP laser and tunable slotted grating laser are designed and fabricated. The high SMSR and low threshold of the slotted FP laser confirmed the quality of the deep-submicron slots. Excellent performance of the slotted grating laser is also obtained with a large tuning range of 38nm and a maximum SMSR of 43dB. The results demonstrate that our method can offer an excellent solution for low cost fabrication of photonic devices with deep-submicron features without using expensive tools such as an e-beam lithography.