1. Introduction

Two-dimensional photonic crystal (2DPC) slab waveguides consisting of line-defects are promising for ultra-small photonic integrated circuits (PICs) composed of light sources, optical switches, and waveguide components. In 2DPC slab waveguides, light can be guided along the line defect while being confined laterally by the photonic band gap (PBG) and vertically by the refractive index difference. The former confinement allows the 2DPC slab waveguide to be bent sharply in the lateral plane [1–4], while the latter confinement means the top and bottom surfaces/interfaces of the slab must be very flat if they have a structure with a large refractive index step, such as an air-bridge, otherwise the waveguide will be very sensitive to light scattering along the surfaces/interfaces. For such a waveguide, demonstration of low loss is an important first step towards establishment of 2DPC-based PICs. So far, small propagation losses have been reported mostly for SOI (silicon-on-insulator) structures [5–8].

One of the unique features of a PC waveguide is its low group-velocity (V_g), often appearing in the vicinity of the PBG. From an application point-of-view, a low group-velocity provides not only compact delay lines and dispersion compensators, but also an ability to drastically enhance, for example, optical nonlinearities in the waveguide. A 2DPC-based symmetric Mach-Zehnder (PC-SMZ) ultrafast all-optical switch with buried quantum dots exhibiting optical nonlinearity is a typical example where a low V_g can contribute to enhancement of the optical nonlinearity, leading to a significant reduction of both the optical switching energy and the length of the nonlinear waveguide [9]. From a fabrication point-of-view, on the other hand, the propagation loss should be reduced sufficiently since this loss is likely to increase inversely proportional to V_g. Unlike the case for SOI, unfortunately, there have been few reports on the measured propagation loss for GaAs-based 2DPC waveguides.

This paper reports on an extremely low propagation loss of 0.76 dB/mm in a GaAs air-bridge type 2DPC slab waveguide up to 1 cm in length. A fabrication process and characterization of the transmittance spectra are also reported.

2. Sample preparations

The samples were fabricated in epitaxial hetero-structures grown by molecular beam epitaxy. A 250-nm-thick GaAs core layer was grown on top of a 2-µm-thick Al_0.6Ga_0.4As sacrificial clad layer on a GaAs substrate. An air-bridge waveguide was fabricated using high-resolution electron-beam (EB) lithography, dry etching, and selective wet-etching techniques. The lattice constant and air-hole diameter were 345 nm and 210 nm, respectively. A single-line defect was formed by leaving a row of perforated air-holes in the Γ–K direction. The 2DPC thus consisted of a periodic air-hole array etched into a planar GaAs slab. The sacrificial clad layer was removed by an HF solution via the air holes, as shown schematically by the fabricated structure in Fig. 1(a). Three straight-line-defect PC waveguides with lengths of 1, 4, and 10 mm were prepared by cleaving both ends. Each waveguide consisted of a missing row of air-holes surrounded by ten rows of air-holes on both sides, as shown in Figs. 1(b) and 1(c) at different scales. As shown schematically in Fig. 1(d), the top and bottom surfaces of a line-defect waveguide were observed by a scanning electron microscope (SEM). The resultant SEM images of cleaved edges viewed from above and below are shown in Figs. 1(e) and 1(f), respectively. The photographs show extremely smooth surfaces. It is also clear that there was no residual material on the rear surface of the slab. From these SEM images, the roughness of the top and bottom waveguide surfaces was roughly estimated as less than 10 nm.

The smooth surface, as shown above, has been achieved by the following three-step processes including surface treatment. The first step is to form high-precision air-hole structures by using a fine EB lithography and dry etching (chlorine-based reactive ion beam etching, that is, Cl₂-RIBE). Suppression of the proximity effect in the EB lithography and optimization of the plasma-etching parameters such as gas pressure and ion beam energy enabled us to realize waveguide-surface smoothness on the order of less than 10 nm, observed with the SEM as mentioned above. The second step is to optimize the composition of the AlGaAs layer for the selective wet-etching of the AlGaAs sacrificial lower-cladding layer. This process is important because a high Al composition gives us a high selectivity in AlGaAs/GaAs wet-etching, while it is in danger of accelerating Al oxidation on the AlGaAs surface, thus being in danger of the significant roughness on the bottom surface. Thanks to this optimization, the bottom surface was so smooth that no roughness or no residues were observed with an SEM on the bottom surface, as shown in Fig. 1(f). The third step is to precisely select an optimum oxygen-plasma condition for removal of the EB resist after dry etching of an air-hole array. Long-time exposure of the dry-etched surface to the oxygenplasma makes the resist removal more completely, while it is in danger of accumulating carbon-containing residues, originated from the plasma machine, on the waveguide surface after resist removal. Due to the optimized condition for the EB resist removal, the waveguide top-surface exhibited extremely smooth without any residues of the resist, as shown in Fig. 1(e).

 

Fig. 1. Schematic diagrams and SEM photographs of the cleaved edges of a sample viewed from several angles: (a) schematic diagram of an air-bridge structure, (b) and (c) SEM photographs viewed at different scales, (d) schematic diagram of the core layer, (e) and (f) SEM photographs viewed from above and below, respectively.

Download Full Size | PDF

3. AFM characterization

In order to investigate the surface roughness in more detail, we carried out atomic force microscope (AFM) measurements. Figures 2(a) and 2(b) show an AFM image and a surface-roughness profile, respectively, obtained with the AFM probe line-scanned along the white line indicated in Fig. 2(a). Paying particular attention to the profile corresponding to the 2DPC waveguide region (labeled DWG in Fig. 2(b)), that is, the region surrounded by the white dotted-line circle in Fig. 2(a), it should be noted that the roughness of the top surface of the 2DPC slab waveguide was less than 1 nm; this value was found to be typical. Nano-scale characterization using the AFM, as shown here, was valuable for optimizing the condition of the EB-resist removal, as mentioned above. That is, an SEM was useless any more for characterization of the surface roughness on the order of less than 1 nm in height, which was thought to be the height of the resist residues in this case. Taking into account that the optical field is concentrated at the center of the line-defect 2DPC waveguide rather than in the air-hole array neighboring the line-defect waveguide, it is important to suppress the top- and bottom-surface roughness of the line-defect waveguide as much as possible. Since the PC core layer is made very smooth in this way, the propagation loss along this waveguide is therefore expected to be sufficiently low.

 

Fig. 2. An AFM image (a) and a surface-roughness profile (b), obtained with the AFM probe line-scanned along the white line indicated in (a). The results indicate that the surface roughness is less than 1 nm.

Download Full Size | PDF

4. Transmission measurements

The band diagram of the 2DPC waveguide calculated by the 3D finite-difference time-domain method is given in Fig. 3(a). Two waveguide modes, for the even and odd symmetries, appear within the band gap of the TE-like mode (electric field in the slab plane). There is a single mode below the light line, as indicated by the double-headed arrow.

We observed the transmission spectrum over a broad wavelength region from 1050 to 1580 nm by using a spectroscopic method suited for very thin samples with a small input-face, like the present ones. As the incident light, we employed light from a white-light source comprising a halogen-lamp, a condenser lens, and optical filters, by coupling it to a polarization-preserving optical fiber with a core diameter of 9.9 µm, the opposite end of which was shaped by polishing to act as a lens. Output light from the fiber was focused onto the cleaved input-edge of the PC waveguide, and the signal appearing at the output edge was picked up by another identical lensed fiber. Then, the signal light was introduced to a monochromator equipped with a grating of 150 grooves/mm and was detected using a cooled multichannel InGaAs detector (OMA-V, Princeton Instruments Inc.) having sensitivity from 850 to 1680 nm. We adopted the above grating in this study so as to cover a broad spectral range of 600 nm at one time.

 

Fig. 3. (a) Calculated band structure of the present line-defect PC slab waveguide (an air-bridge type GaAs-based PC waveguide). The energy is given by a/λ with a and λ being the lattice constant and wavelength in vacuum, respectively; k is the wave vector. (b) Transmission spectra of the PC waveguides with different lengths. The spectra were plotted in such a way that each center wavelength was adjusted to 1279 nm, that of the shortest sample.

Download Full Size | PDF

Measured transmission spectra for different waveguide lengths are given in Fig. 3(b). Since output signals were normalized with respect to direct fiber-to-fiber reference signals, the ordinate value gives us the insertion loss of the waveguide. For fiber-to-waveguide coupling, we employed no particularly high-efficiency method other than butt-coupling. The measured insertion loss was as high as -20 to -30 dB. Although the insertion loss was high, the reproducibility in measurement of the transmittance was rather good, as shown by the plots in Fig. 4(a), described below. Looking closely at the spectra in Fig. 3(b), they gradually attenuate at longer wavelengths due to the band edge, while they change abruptly at shorter wavelengths, presumably due to their vicinity to the intersection of the even mode and the light line in Fig. 3(a). The transmission range was different for the different waveguide lengths and, as a whole, was narrower than the calculated range, as shown in Fig. 3(a). In addition, their center wavelengths deviated by 10 to 15 nm (0.8% to 1.1%). This is probably due to variations in the air-hole diameter or thickness of the core layer. Taking this into account, the spectra in Fig. 3(b) were plotted in such a way that each center wavelength was adjusted to 1279 nm, that is, the center wavelength of the shortest sample.

 

Fig. 4. (a) Transmission spectra of four identical samples for different waveguide lengths. Note that the spectra, in particular, those for 4- and 10-mm-long samples, are very similar. (b) Plot of the observed transmittances at the center wavelengths in Fig. 4(a) as a function of sample length. An extremely low propagation loss of 0.76 dB/mm is achieved.

Download Full Size | PDF

5. Propagation loss

Figure 4(a) shows transmittance spectra of four identical samples for different waveguide lengths. It was found that the transmission ranges and spectra shapes for each length, in particular, those for the 4- and 10-mm-long waveguides are very similar. In order to derive the propagation loss of these samples using a cut-back method, transmittances at the center wavelengths in Fig. 4(a) for each sample were plotted as a function of the waveguide length, as shown by the empty circles in Fig. 4(b). From the slope of the fitting, as shown by the red line in the figure, the propagation loss of (0.76±0.5) dB/mm was derived. This value is thought to be extremely low for the 2DPC slab waveguides [8]. The result suggests that the air-bridge structures with 2D air-hole arrays have been fabricated with high precision even for the waveguide up to 10 mm in length.

6. Conclusion

We have fabricated and characterized straight single-line-defect PC waveguides on GaAs films with lengths of 1, 4, and 10 mm. By controlling the Al content of the sacrificial AlGaAs clad layer and the wet etching duration, a PC core layer with a very smooth surface was obtained. AFM measurement results indicate that the surface roughness is less than 1 nm. An extremely low propagation loss of 0.76 dB/cm was achieved. These features indicate that the present results are promising for key passive elements, such as photonic-crystal symmetric Mach-Zehnder switches, needed in future optical communication applications.