**15. Conclusions**

Limits of wide wavelength tuning, FWHM and stopband widths are investigated for the InP multiple air-gap system. Wavelength tuning efficiency for said sensor system is found to be the closest to 1, and additionally with much smaller size compared to other systems and different sensor methodologies.

In our comparison, the tunable chirped fiber Bragg grating reveals by far the smallest FWHM of 0.007 nm at 1.5 μm. However, the space requirement is the largest in this case. The next lowest FWHM of typically 0.1 nm at *λ* = 1.3 μm are measured in MEMS tunable PC filters. Potentially, the space requirement is also very small. However, in order to obtain larger spectral spans of e.g., 400 nm, the combination of several neighboring spectral tuning ranges is required. For that purpose, the QD have to be adjusted and varied in all the PC crystal in a defined way. This might be quite challenging concerning the spectral adjustment and defined variation from array to array. On the other hand, AWGs also provide small linewidths, and the arrangemen<sup>t</sup> of several arrays next to each other is very easy. However, the fabrication of AWGs for the VIS spectral range is still an enormous challenge.

The typical linewidths measured for FP filter arrays are higher than the typical values of PC or AWG sensors. Although 0.1 nm linewidth could be achieved at *λ* = 1.5 μm in an InP/multiple air-gap FP filter in a single case, it was not reproducible in tunable filters or in static FP filter arrays. Using the InP multiple airgap MEMS technology, very small linewidths are extremely challenging in the tunable technology, but on the other hand, it is very attractive in terms of scalability. Since various static FP filter arrays can be manufactured next to each other within a single 3D nanoimprint step, this technology enables lowest tentative price per spectral range for visible spectral ranges.

A chance to achieve small linewidths in MEMS tunable filters lies in the application of stable cavities. A small linewidth of <0.15 nm was reported over the whole tuning range, although the tuning range is limited and only thermal tuning was applied [124].

The current version of the plasmonic MEMS sensor evaluates charge carriers induced by the SPP resonance into a diode structure and transforming this angle dependent current into the spectral information. At the moment this is limited to a rather high FWHM value of 10 nm or more. On the other hand, this concept is very promising due to its wide tuning range.

Thermally tuned chirped FBG shows narrow band filter lines with FWHM of 0.007 nm and can potentially be implemented in any working spectral range of an optical fiber. The small tuning rang of only 16.5 nm requires, however, an array of many individual chirped FBG to cover broader spectral spans.

The classical grating spectrometer is definitively the best in our comparison in terms of the efficiency in making most out of available light. However, the grating spectrometer suffers considerably from strongly reduced spectral resolution when downscaling the devices, whereas such limitations are not relevant for all the other sensor types compared in our review. In all these cases, the resolution is very high and independent from miniaturization. The AWG uses available light much more efficiently than the static and tunable FP filter arrays and the tunable PC filter array. The latter three own rather low efficiencies, but the efficiencies can be boosted by spectral preselection as shown in this review.

Fourier spectroscopy in the infrared spectral range is using the amount of light much more efficiently than grating spectrometers (Multiplex and Jacquinot advantages) in large set-ups. The disadvantage is that the miniaturization achieved up to now is by far less than that obtained for PC and FP-based sensors.

Nanoimprint can be applied to all the compared sensors, except the chirped fiber Bragg grating and the classical grating spectrometer. Transmission gratings could be fabricated by nanoimprint lithography. Notwithstanding, nanoimprint technology can only reveal its full potential in manufacturing static FP filter arrays. Here, 192 different filter lines were demonstrated using a single 3D nanoimprint step to define accurate and

diverse 3D cavity layers. In a proof-of-concept, 192 spectrally different filter lines were successfully demonstrated, which is far better than the three different broad filter lines used in modern digital cameras. There are no limits on the principle to considerably increase these values in static FP filter arrays. At the same time, nanoimprint substantially reduces fabrication time, cost and effort.
