1. Introduction

Photonic crystals (PCs) influence the flow of light on the wavelength scale, and will enable the ultimate miniaturization of photonic integrated circuits (PIC) [1]. Post-production tuning is of importance for PIC as a means of correcting fabrication deviations, active device control and realization of additional functionality [2]. In particular, the precise control of resonant wavelengths of cavities is of importance. PC cavities are essential components in PIC. The spectral features of cavities are generally narrow, thus, large tuning effects can be realized when the effective refractive index of the cavity is varied.

Recently, significant progress has been made on post-production tuning of photonic crystals and cavities. As a means of trimming or tuning, complete infiltration of the PC was carried out with materials such as liquid crystals (LC) [3, 4, 5, 6, 7, 8, 9], polymers [10, 11, 12] and chalcogenide glass [13]. With these methods, it is possible to red-shift the spectral features of the original PC device, since the air inside the holes is replaced with a material with a higher refractive index. In contrast, a digital etching approach [14, 15, 16], i.e. cyclic oxidation and oxide removal, results in blue-shifting of spectral features as high index material is removed.

A more attractive method of post-processing is to address a small area of a device. In this way, e.g. one cavity or one of its resonant modes can be modified, without affecting the overall properties of the PC. Recently, several of such methods were reported. These include infiltration of several holes by micro-pipetting [17, 18, 19], photo-darkening of chalcogenide glass [13], local in-situ polymerization of globally infiltrated monomers [20], nano-oxidation by atomic force microscope [21] and controlled deposition of carbonaceous nano-dots [22].

Here, we present a novel lithographic technique of tuning PC devices, based on local mask opening for individual holes. This enables the use of any post production method on the exposed area of the PC, while not affecting other holes in the PC. None of the other techniques is capable of being both local on the single hole level and free, regarding the choice of postprocessing technique. The process is demonstrated by application of both local digital etching and local LC infiltration of the nearest-neighbor-holes of a point defect (H1) cavity in the technologically important InP system [23].

2. Experimental

The PCs were fabricated by etching a triangular lattice of air holes of <250 nanometer (nm) diameter, through a 500 nm InP/500 nm InGaAsP layerstack on an InP substrate using Cl₂=O₂ chemistry [24]. The hole depth is at least 2.5 µm, i.e. aspect ratio of >10. A so-called H1-cavity is created by eliminating one hole in the PC. The cavities were probed through access ridge waveguides (RWG) on the chip. The airfilling fraction was determined from Scanning Electron Microscope (SEM) images to be 0.27. Transmission spectroscopy was carried out using an end-fire technique using a tunable laser in the range of 1.47 µm to 1.57 µm in TE polarization.

After PC and waveguides are made, the transmission spectrum of the device was measured. Next, a SiN_x masking layer with a thickness of several hundred nm is deposited by plasma enhanced chemical vapor deposition (PECVD). This SiNx masking layer covers, but does not significantly fill the holes, see figure 1a). After the holes are covered by the mask the device is remeasured to verify that deposition of the mask layer does not affect the spectrum. For the infiltration experiments, the nematic liquid crystal 4-n-penthyl-4’-cyanobiphenyl (5CB, Merck) is used.

Holes in the PC lattice can be targeted by focussed ion beam (FIB) milling in a dual beam FIB system. First, shallow markers are created in the masking layer in the vicinity of the area of interest. Next, using SEM (at 20 kV energy) it is possible to detect the holes of the PC lattice through the masking layer. Thus, the PC lattice can be related to the markers. In the final step individual holes can be targeted for FIB milling on the basis of the markers, see figure 1b). The precision, both in x- and in y-direction, of the FIB milling is 30–40 nm (standard deviation).

 

Fig. 1. a) SEM picture of a cross-section of a PC covered by a SiN_x masking layer. b) SEM picture of an array of individually targeted holes by FIB milling. SEM allows detection of the PC holes through the masking layer. On the right side the alignment markers are visible. The scale in both figures is indicated by the white bar, which has a length of 1 µm in both cases.

Download Full Size | PDF

Post-processing of a single hole will only yield a small shift of the resonant frequency for the H1 cavity. Since the Q-factors of small cavities fabricated in the deeply etched InP/InGaAsP/InP system are low, this would result in shifts within the linewidth. Therefore, to optically demonstrate the selective mask opening mechanism, local post-processing of the six holes adjacent to an H1 cavity was performed.

To open the selected holes, a hexagonal pattern encompassing the six holes next to the H1 cavity was fabricated by FIB milling (see figure 2a), leaving a SiN_x layer of some tens of nm at the target position. It is necessary not to open the masking layer completely as the FIB etchrate on the underlying InP is high. Attempting to fully open the masking layer in a single step effectively destroys the PC. The final step consists of thinning down the remaining SiNx layer by wet chemical etching using a diluted solution of HF in water (1:200), opening the holes adjacent to the defect completely (see figure 2b). The wet chemical etching step can be controlled to only open the pre-thinned region, while the rest of the mask remains sufficiently thick to protect the holes. The wet etching does not affect the InP.

 

Fig. 2. a) 20 kV SEM picture after FIB opening of a hexagonal shape covering the holes adjacent to the H1-defect. The vertical arm of the cross-like structure is the access RWGs used for transmission spectroscopy. b) SEM picture of the hexagonal opening after the wet chemical etching process, showing the six opened PC holes. The four dark squares are the markers used for the alignment procedure. The holes situated below the thicker part of the mask remain closed. The scale in both figures is indicated by the white bar, which has a length of 1 µm in both cases.

Download Full Size | PDF

The work of Topol’anc̆ik et al. [25] describes the opening of large areas of PC masked by SiO_x using electron beam lithography (EBL) for subsequent infiltration. However, since we use high aspect-ratio RWGs we avoid planarization required for EBL, due to the use of the FIB. FIB milling is also used as a method of fabrication of PCs in diverse materials [26, 27, 28, 29, 30] by direct milling, i.e. without masking layer. Also, FIB modification of conventionally fabricated PC devices by direct milling is reported [31]. However, direct FIB milling of InP is severely challenging because of increased surface roughness due to InGa deposition [31], material redeposition inside the holes and self-focusing [32]. Although these direct writing methods by FIB are a viable route for post fabrication modification, they are neither precise, with respect to the impact on the spectral features, nor reversible.

3. Results & Discussion

Figure 3a shows a wide-band spectrum of the H1 cavity, measured in transmission for three conditions: as fabricated (”after fabrication”), after local digital etching (”after etching”) and after local LC infiltration (”LC filled”). The resonance after fabrication is clearly visible at a=λ=0.24. The spectrum is modulated by a complex pattern of Fabry-Perot (FP) fringes arising from reflections at the end-facets of the access RWGs and the PC boundaries. Since only the envelope of the signal is relevant, these were not analyzed in detail. Some discrete jumps are visible in the spectrum which are due to minor deviations between structures of the lithographically tuned series. Three dimensional Finite Difference Time Domain (3D-FDTD) calculations simulate the resonances, under the same three conditions as in the experiment, see figure 3b. In figure 3b an additional spectrum is given for the case that all the holes would be infiltrated, i.e. a failed experiment. The Q-factor of the cavity for these deeply etched structures is known to be low due to the low vertical index contrast. Comparison with the 3D-FDTD calculations shows that extra losses are incurred due to fabrication imperfections. The Q-factor of the experimental cavity was determined from the envelope to be approximately 50, by a factor of 3 smaller than expected from the simulations.

 

Fig. 3. a) Transmission spectrum of the H1 defect resonance. The spectra are shown stacked for clarity. The ”after fabrication” line represents the measurement after fabrication. The ”after etching” line displays the resonance after 15 steps of digital etching increasing the hole diameter from 190 nm to 210 nm, giving rise to a blue-shift of 40 nm (Δ a/λ~0.01). After filling (”LC filled” line) the opened and enlarged holes with LC, the resonance displays a red-shift of 40 nm, approximately returning the resonance frequency to its original position. The change of the shape of the resonance for both ”after etching” and ”LC filled” with respect to the reference is due to the presence of the SiN_x layer, which also covers the end-facets and acts as an anti-reflection coating. b) 3D-FDTD calculations of all the steps in the post-processing scheme. A ”completely filled” line was added to indicate the result of a failed experiment, i.e. all the holes of the PC are infiltrated. The calculations use the following parameters: a=380 nm, reference diameter=203 nm, after etch diameter=226 nm, LC filled effective refractive index of the holes=1.41, corresponding to a LC filling efficiency of ~75%.

Download Full Size | PDF

After every step in the opening process (mask deposition, FIB etching) the spectrum was re-measured and remained unchanged. Thus the spectrum ”after fabrication” in figure 3a also represents the cavity spectrum after the holes are opened. By repeatedly subjecting the selectively opened holes to alternating an oxygen plasma and an etchant, which removes the oxide formed by the plasma, the diameter of the selected holes is enlarged. As a result of the larger holes the resonance is blueshifted (figure 3a, after etching line) by 40 nm. This process of ”digital etching” is a precise and effective method of tuning, since both the thickness of the oxide layer and the number of etching steps can be accurately controlled [14, 15, 16]. In this case, 15 successive oxidizing and etching steps were applied to achieve the 40 nm shift. The increased hole size could be directly verified from SEM inspection, yielding an increased hole diameter from 190 nm to 210 nm. For the 3D-FDTD calculations an increase of 20±5 nm of the hole diameter was needed to obtain the measured shift. The uncertainty is quite high for this determination, due to the large linewidth. Since the main part of the PC is unexposed to the digital etching treatment, due to the protective masking layer, see figure 2b, the bulk properties of the PC away from the cavity under modification, remain unchanged, apart from some possible influence of the masking layer.

In the last step the selectively opened and enlarged holes are filled with LC (figure 3a, LC filled line) by deposition of a droplet on the sample surface. The introduction of the LC into the opened holes gives rise to a large redshift of 40 nm, consistent with the increased refractive index of holes directly surrounding the defect. 3D-FDTD calculations yield a infiltration efficiency of 70–80% to match the resonance frequency. Filling factors of the same magnitude were also found in other LC infiltration experiments [8]. As only the six holes adjacent to the defect are filled, the PC retains its bulk properties, thus the band edges of the PC remain unaffected. The Q-factor remains approximately constant. In contrast, results from global infiltration techniques consistently yield a decrease of the Q-factor [6]. Apart from the linewidth, the calculations are in very good agreement with the experimental data. The high frequency transmission band onset occurs near a/λ≫ 0.26 in all three experimental conditions. If all the holes of the PC were infiltrated, the high frequency band edge would show a considerable shift as well, see figure 3b, bringing the resonance and high frequency band edge closer together [8, 10]. This is not observed experimentally, showing that only the six holes are infiltrated indeed. It independently confirms optically what is already evident from the SEM-data in figure 2b.

Since the LC is easily dissolved in common solvents, the LC may be removed from the holes to continue the digital etching process after which the LC can be re-introduced into the holes. In this way, the resonance may be positioned close to the desired location by post-processing. The resonance frequency can subsequently be actively tuned, exploiting the relatively large dependence of the LC refractive index on an external electric field or temperature. Better precision of the blue-shift of the resonance may be achieved by using an infiltration method with more control over the refractive index of the infill, such as Atomic Layer Deposition (ALD) [33]. Since different resonant modes have spatially distinct electromagnetic field distributions, local post-processing can address specific modes for tuning.

4. Conclusion

In conclusion, we have demonstrated for the first time tuning of resonant modes by local post-processing of individual holes in photonic bandgap structures. The six holes adjacent to an H1-defect cavity were selectively opened using FIB milling. Using digital etching the H1 cavity resonance was blue-shifted by 40nm, and subsequently 40 nm red-shifted by local infilling with liquid crystal. This method can be applied to precisely control the resonant frequency by both blue- and red-shift, and can also be used for mode selective tuning of resonant states by targeting the appropriate holes.