Compact demultiplexers and spectrometers for integrated photonics

Wavelength demultiplexing devices (WDD) disperse an input beam into a plurality of sub-beams, and are commonly used to separate optical channels at different wavelengths. Compact WDDs are of great interest for several integrated photonic applications including optical information processing, optical communications and networking, as well as integrated optical sensing such as `lab-on-a-chip’ biosensing systems. The compactness of these devices and their compatibility with integrated photonic and electronic platforms are among their main advantages when compared to other wavelength demultiplexing solutions. These advantages are the result of the wide range of controllable dispersive properties in a very compact structure enabled by the use of photonic crystals (PCs). Such dispersive properties are not available in conventional devices that usually rely on bulk or grating-based materials.

For efficient WDD implementation, we use PCs. These structures, which consist of sub-wavelength periodic features, can be used to control the optical properties of materials: they have unique dispersive properties that are not present in naturally occurring optical materials. In particular, properly designed PCs can angularly separate optical beams of different wavelengths in widely different directions. By analogy with conventional prisms, this phenomenon is called the superprism effect and represents the primary physical effect used for the separation of wavelength channels of an optical signal as shown in Figure 1(a).

Figure 1. Schematic representation of the three dispersive properties of photonic crystals used for wavelength demultiplexing: (a) the superprism effect, (b) negative diffraction, and (c) negative refraction.

Ideally, the compact spatial separation of different wavelengths would require a large angular difference between the directions of propagation of the different wavelengths. This would translate into a strong superprism effect for the selected PC. However, rapid variation of the angle with wavelength naturally occurs in the PC band structure along with strong divergence of the optical beams. As a result, the beams corresponding to different wavelengths have a large divergence angle, and thus undergo significant broadening during propagation. However, the complete spatial separation of optical beams at different wavelengths under such a large divergence, as performed in earlier implementations of superprism-based PC demultiplexers, requires large propagation lengths. The result is that the advantage of compactness is effectively lost: this is especially true for high-resolution applications.^1,2

An efficient, compact WDD can be realized if the effect of diffractive broadening can be avoided. Fortunately, PCs have the interesting property of showing diffraction effects negative to that of ordinary bulk materials. This effect, called negative diffraction, is illustrated in Figure 1(b).^3,4 Propagation in a suitable length of a negative diffraction material can eliminate the broadening problem and yield small spot sizes at the output plane of a WDD.^5–7 Another important requirement for these devices is the separation of stray signals, such as unwanted wavelengths or polarizations, at the output of the demultiplexer. This feature can also be added to PC-based WDDs because these crystals have a third dispersive property, namely the negative refraction illustrated in Figure 1(c), that, combined with the superprism and negative diffraction effects, can produce the desired signal.

We have shown that it is possible to engineer the band structure of PCs by geometrical optimization of their crystal lattice to achieve materials that simultaneously combine a strong superprism effect, negative diffraction, and negative refraction.⁷ PCs therefore represent perfect optical materials for the fabrication of compact WDDs. Figure 2(a) shows the overall device schematic with the three dispersive properties working together to achieve a compact and efficient design. A scanning electron microscope (SEM) image of a device fabricated on a silicon-on-insulator wafer is shown in Figure 2(b). Figure 3 shows the performance of a 70×100μm PC structure designed as a wavelength demultiplexer.⁷ This compact structure is capable of separating four wavelength channels with 8nm channel spacings with a crosstalk level better than 6.5dB. The device is also considerably smaller than all other proposed implementations of PC demultiplexers with similar performance characteristics.

Figure 2. (a) Schematic of a compact wavelength demultiplexing device formed by combining three dispersive properties of photonic crystals. (b) SEM image of a fabricated device on a silicon-on-insulator wafer. The output of the photonic crystal device is sampled using an array of ridge waveguides.

Figure 3. Spectral response of the four-channel wavelength demultiplexer shown in Figure 2(b). The wavelength spacing is 8nm with crosstalk reduction better than 6.5dB.

The use of PC demultiplexers for on-chip high-resolution spectrometers has also been considered. Integrated optical sensing platforms, such as lab-on-a-chip structures, would significantly benefit from this type of design. The use of a spectrometer provides more options on the type of light source being used in the sensors as well as additional flexibility for the configuration and sensing mechanism. The core element of such spectrometers is the WDD discussed above, which maps different wavelengths into different spatial distributions in the output plane (see for example, the device output waveguides shown in Figure 2). By calibrating this structure with a set of training data obtained at multiple known frequencies, it is possible to accurately locate spectral features in a reasonably wide bandwidth (e.g., 50nm). Such special-purpose spectrometers represent essential elements in current biological and environmental sensing applications in which the spectral signatures (e.g, Raman spectrum) of a given molecule (or a set of molecules) need to be accurately detected. We have shown that using a 70×100μm PC structure similar to that shown in Figure 2(b) allows us to estimate the location of a spectral peak with an accuracy better than 50pm over a 50nm-wavelength range.⁸ This level of accuracy, combined to the compact size of the structure, makes this spectrometer very attractive for lab-on-a-chip sensing applications with minimal sample requirements.

In summary, multiple applications significantly benefit from using a compact on-chip WDD. The unique dispersive properties of PCs effectively enable the realization of optical materials with controllable optical properties to fabricate compact and efficient WDDs. As an example of the unique properties of PC dispersive devices, we have described a device based on combining the superprism effect, negative diffraction, and negative refraction with great potential for compact on-chip spectrometers, and orders of magnitude smaller than other implementations with similar performance.

The authors would like to thank the Air Force Office of Scientific Research (G. Pomrenke) for supporting this research.