Fabrication Methods and Etch Depth

The photonic crystal structure is etched in the GaInAsP/InP heterostructure by e-beam lithog-raphy and chemically assisted ion beam etching (CAIBE). 150nm of SiO₂ sputtered on the sample yields an etch mask suitable for high-resolution patterning. The triangular lattice is written in 500nm spin coated polymethylmethacrylate (PMMA) resist by e-beam exposure. The PMMA is developed in 1:3 methylisobutylketone/propanol and the hole pattern is transferred into the SiO₂ layer using CHF₃/Ar-based reactive ion beam etching (RIE). This two-step writing process pro-vides higher selectivity in the mask patterning.

Since out-of-plane losses strongly depend on the hole morphology [Lalanne, Ph., et al. (2001)], optimizing the etching process is of primary importance for obtaining high quality samples. In fact, insufficient hole depth with respect to the vertical extent of the guided mode profile [Benisty, H., et al. A (2002)] and/or conical hole shape [Ferrini, R., et al. (2002)] increase light scattering into the substrate. In order to minimize out-of-plane losses, the etch depth has to be larger than 1.5µm, with holes as straight as possible. This implies high anisotropy of the etching process and high aspect ratios. CAIBE based on Ar/Cl₂ has been found to yield vertical profiles with high aspect ratios, in comparison to the standard methane-based RIE. For this reason, Ar/Cl₂ CAIBE is used to etch the photonic crystal pattern into the InP-based heterostructure, using the e-beam

Figure 3.5 SEM micrographs of a photonic crystal with lattice constant a=400nm fabricated using CAIBE etching. The images were taken before the SiO₂ mask removal. Courtesy of Ferrini, R., EPFL, Switzerland, Mulot, M., KTH, Sweden, and Talneau, A., LPN - CNRS, France.

exposed SiO₂ mask. Thus, the sample is sputtered by an energetic argon ion beam with 5sccm flow and 400eV ion energy. At the same time, a chemical attack is generated by a 1sccm chlorine flow.

Argon ions sputter phosphorus atoms, whereas chlorine enhances the removal of indium atoms by forming volatile products (InCl_x). Since the vapor pressure of InCl_x is quite low at room temper-ature, the sample needs to be heated to achieve efficient removal of the etch products. Optimal samples are obtained for a temperature of about 200^◦C and for an etching time of 20min..

Fig. 3.5 shows a selection of scanning electron microscopy (SEM) micrographs of InP-based photonic-crystal test structures. The air holes are conical, with nearly vertical walls close to the surface and strongly tapered at the bottom, resulting in a carrot-like profile. Notice also that the bottom tails are bent with respect to the hole vertical axis. The mechanism responsible for this bending is not well understood yet. The etch depth is about 2.5µm for hole diameters larger than 220nm.

However, for smaller diameters, the hole depth decreases with the hole diameter.

Having carrot-like hole shapes is equivalent, with good approximation, to having straight cylin-drical holes with reduced etch depth [Ferrini, R., et al. (2002)], which results in increased out-of-plane losses. Thus, from the modelling point of view, the effect of conical hole shape is accounted for by simply increasing ²⁰⁰. On the other hand, if losses are too large, several photonic-crystal functionalities are disrupted and the modelling itself becomes nonsense. That is why being able to fabricate high quality samples with low propagation losses is so important not only in view of

applications, but also at the characterization and modelling levels.

State of the art InP-based photonic crystals exhibit out-of-plane losses that can be modelled assum-ing²⁰⁰'0.1, which is already an acceptable loss level for characterization and modelling purposes.

If one wants to have better performances, the fabrication process has to increase the etch depth while keeping the hole shape cylindrical. Once that the etch depth has reached a critical value, which depends on the waveguide and photonic-crystal geometry, the amount of out-of-plane losses becomes equal to the intrinsic loss level, corresponding to infinite etch depth [Benisty, H., et al.

A (2002)], as highlighted in Sec. 2.5.3. However, one has always to deal with roughness-induced scattering losses, which are unavoidable in real samples. Nevertheless, also this kind of loss mech-anism can be included in the two-dimensional approximation through the ²⁰⁰ parameter.

Another efficient etching technique for InP-based photonic crystals is electron cyclotron resonance / reactive ion etching (ECR/RIE), which provides 3.5µm-deep holes with vertical sidewalls over 2µm at lattice constanta∼380nm. Recently, the application of induced coupled plasma (ICP) etching in the fabrication of InP-based photonic crystals has given very promising results as etch depth and hole shape are concerned. The method is still under optimization within the PCIC collaboration.

As to GaAs-based photonic crystals, literature has been plenty of results since the pioneering works of Krauss T. F., et al. (1996) (fabrication) and Labilloy, D.,et al. A (1997) (characteriza-tion). For this reason, a detailed description of the fabrication process is not reported here. It is worth to mention, however, that the waveguide geometry designed in the PCIC project is similar to the InP system, with AlGaAs and GaAs in place of InP and GaInAsP, respectively, and with the quantum well layers replaced by quantum dots. Due to the different index profile of the GaAs waveguide with respect to InP, the effective dielectric constant of the fundamental TE mode is now

²_eff = 11.56, instead of 10.5. Therefore, as far as two-dimensional modelling is concerned, moving from InP- to GaAs-based photonic crystals is resolved in changing the effective dielectric constant, and, as necessary, the loss parameter ²⁰⁰. Moreover, since the index contrast between GaAs and AlGaAs is higher than that between GaInAsP and InP, the field is more confined and a smaller etch depth is enough for reaching intrinsic out-of-plane losses.