www.nature.

com/scientificreports

OPEN Mid‑infrared hyperspectral sensor

based on MEMS Fabry‑Pérot
interferometer for stand‑off
sensing applications
Abba Saleh1*, Mikhail Mekhrengin2, Timo Dönsberg2, Teemu Kääriäinen2,
Guillaume Genoud2 & Juha Toivonen1
We report a novel hyperspectral sensor employing a Fabry-Pérot interferometer based on micro-
electro-mechnical system and a custom mid-infrared supercontinuum laser. The Fabry-Pérot
interferometer allows on-axis filtering, of spectral components of supercontinuum light backscattered
from a target, with a spectral resolution of about 25 nm. We demonstrated hyperspectral
identification of black polypropylene (PP) and polyethylene (PE500) using the 3–3.5 µ m region of the
supercontinuum spectrum and a corresponding measurement rate of 62.5 spectra / s. The resulted
spectra show excellent agreement with the reference based on an FTIR spectrometer. Furthermore,
we showed that the coloring of the plastics has no effect on their identification at this wavelength
range.

A hyperspectral sensor measures transmitted or backscattered light from a target, probing multiple spectral
components of the light to enable analysis of the target. Hyperspectral sensors have been widely used in various
applications including a­ griculture1, ­medicine2 and mineral e­ xploration3. Conventional hyperspectral sensors and
cameras rely on passive target illumination, typically ambient lightening, which make them prone to misinterpre-
tation due to any fluctuations in the illumination spectrum as it undermines the integrity of the signal. However,
recent advances in nonlinear fiber optics have led to the development of spatially coherent yet broadband fiber
lasers, termed supercontinuum (SC) l­asers4,5, enabling active target illumination for robust multi-spectral6,7 as
well as ­hyperspectral8–10 sensing applications with very high signal-to-noise ratio.
Furthermore, significant progress in terms of the SC spectrum expanding into the mid-infrared (MIR), have
been reported in the ­literature11,12, with significantly high average output ­power13, high repetition ­rate14 and
femtosecond pulse ­durations15,16, opening the door for various applications in the MIR including ­spectroscopy17,
­imaging18 and optical coherence tomography (OCT)19. The MIR spectral region offers novel perspective for a
more accurate analysis of a target as molecules demonstrate strong and characteristic absorption in this region,
termed the molecular fingerprint, due to the strong fundamental vibrational t­ ransitions20. The aforementioned
properties of MIR SC have opened up new possibilities for hyperspectral sensing applications in the MIR spectral
region. However, a fast and robust filtering of various spectral components of the MIR SC spectrum is required
to meet the demands of hyperspectral sensing applications. Some promising solutions based on a spectrom-
eter comprising a miniaturized Fabry-Pérot interferometer have been proposed in the ­literature8. However, the
integrated nature of the spectrometer limits the sensor design flexibility and photodetector choice as both the
detector and the Fabry-Pérot interferometer are incorporated as a single unit. Thus, further study is required to
fully realize the commercial potential of MIR hyperspectral sensors.
Herein, we present an active hyperspectral sensor (AHS) using a combination of a MEMS-based Fabry-Pérot
interferometer (FPI) and a spectrally tailored SC light source covering up to 3.5 µ m of the MIR spectral region.
The voltage tunable FPI enables compact, cost effective and on axis non-dispersive filtering of spectral compo-
nents of the SC light reflected from a target. The standalone nature of the FPI allows flexible instrument design
as it can be placed at any desirable location on the instrument. We showed for the first time, to the best of our
knowledge, the suitability of FPI for hyperspectral sensing of plastics. We demonstrated hyperspectral identifica-
tion of black plastics as their detection is tedious in the near infrared (NIR) due to the carbon based additives
which significantly absorb the NIR light. Black polyethylene (PE500) and polypropylene (PP) are specifically

1
Photonics Laboratory, Tampere University, Tampere 33101, Finland. 2VTT Technical Research Centre of Finland
Ltd, Espoo 02150, Finland. *email: [email protected]

Scientific Reports | (2022) 12:19392 | https://doi.org/10.1038/s41598-022-23758-w 1

Vol.:(0123456789)
 www.nature.com/scientificreports/

Figure 1.  Tuning parameters of the Fabry-Pérot interferometer. (a) The blue line represents the corresponding
central wavelength with respect to the tuning voltage, and the overlay red (dashed) line shows the full-width-
half-maximum (FWHM). (b) Transmittance spectra of the Fabry-Pérot interferometer are shown as a function
of the operating voltage in the range of 11–29 V with constant voltage increments indicated by different colors in
the figure.

Figure 2.  Experimental setup for stand-off hyperspectral reflectance measurement. Abbreviations: MT—

mirror telescope, LP—longpass filter, DU—detection unit, FPI—Fabry-Pérot interferometer and DAQ—data
acquisition.

chosen as sample materials as they are one of the most important black plastic waste especially among waste
electrical and electronic equipment (WEEE)21. Additionally, they have very similar absorption features which
makes them very difficult to distinguish. We successfully measured their reflectance spectra using the 3 – 3.5 µ m
spectral band of the SC spectrum. The resulting spectra strongly correspond to their reference based on an
FTIR spectrometer. The technique is further extended to detection of white plastics to validate its applicability
to differently colored plastics. This emphasizes the great potential of the technique for plastic wastes sorting, in
recycling processes, and other hyperspectral sensing applications.

Methods and results
Operational parameters of the Fabry-Pérot interferometer are shown in Fig. 1. The miniaturized MEMS-based
tunable FPI filter is part of the MEMS-FPI solutions developed at VTT Technical Research Centre of Finland
­Ltd22–25. The FPI resonator comprises two highly reflective mirrors separated by an air gap. Constructively
interfered light between the mirrors is transmitted, and the transmission band is defined by the air gap between
the mirrors. The air gap is tuned by applying a voltage between the mirrors and, thus, the transmission band of
the FPI can be tuned to the range of interest. Figure 1a presents the tuning parameters of the FPI. The operating
voltage of the FPI is in the range of 0 – 29 V which corresponds to a wavelength tuning range of 3000 – 3600 nm.
The full width half maximum (FWHM) of the transmission band remains in the range of 22 – 28 nm over the
whole operating range of the FPI. An example transmittance spectra of the FPI at a constant voltage increment
is shown in Fig. 1b.
The experimental arrangement is presented in Fig. 2. A custom SC light s­ ource26 producing 10 kW peak
power sub-nanosecond pulses at a repetition rate of 100 kHz (more detailed description of the supercontinuum
light source can be found in Ref. 26) is directed towards a black plastic target having a thickness of 3–5 mm and
located at a distance of 2 m. The SC light is partially absorbed upon incidence on the target while being scat-
tered. The backscattered light is collected via a mirror telescope arrangement (MPD399V–M01 & PFE10–P01,
Thorlabs) then filtered by a longpass filter (SLWP-2989-000453,NOC) to cut out wavelengths below 3000 nm.

Scientific Reports | (2022) 12:19392 | https://doi.org/10.1038/s41598-022-23758-w 2

Vol:.(1234567890)
www.nature.com/scientificreports/

Figure 3.  Measured backscattered SC spectrum (– -) from a diffuser and the corresponding relative intensity
noise (—) based on ten consecutive measurements.

The FPI scans across 3050–3600 nm spectral range at a frequency of 62.5 Hz. The light transmitted through the
FPI is focused onto a photodetector (PV-3TE-5, VIGO Systems). The measured signal is amplified with a custom
bandpass-filtered amplifier having a voltage gain of 250 and passband of 3.5 kHz - 10 MHz. The measured sub-
nanosecond SC pulses are temporarily stretched due to the 10 MHz cut-off frequency of the amplifiers. They
are digitized using a 14-bit analog-to-digital converter (LTC2145-14, Analog Devices, Inc.) which is a part of
a development board (STEMlab 125-14, Red Pitaya d.d.). The digitized signal is processed using a system on a
chip (SoC) solution (Zynq 7010, Xilinx). The SoC contains field-programmable gate array (FPGA)-based pro-
grammable logic (Artix 7, Xilinx) and a processor (ARM Cortex-A9 MPCore, Arm Holdings). The FPGA part
of SoC removes constant component of the signal and integrates all data samples related to one SC pulse into
one value while the processor send the integrated data values to the PC via Ethernet. The FPI voltage modula-
tion and SoC acquisition software are synchronized using the same external trigger as applied for the SC laser.
Although the FPI scans continuously during the measurement, the measured spectra seem to show discrete
values as transmitted signal is sampled by FPGA with 125 MHz-sampling rate, which allows to distinguish and
integrate individual SC pulses. And they are then averaged over 100-µs-long periods of time, corresponding to
independent spectral channels.
The reflectance spectrum of a sample can be derived using
IS − N
R= , (1)
IR − N
where IS and IR are the measured intensity spectrum of the sample and reference target respectively, and N is
the background noise. As a reference target, we used a ground glass diffuser (DG10-120-M01, Thorlabs) having
spectrally flat reflectance of 97.5 % over the probed wavelength range.
Figure 3 shows a measured backscattered SC spectrum from the diffuser and the corresponding relative
intensity noise (RIN) based on 10 different measurements. The resulted spectral shape is mainly due to the
spectrum of the SC light source, as the diffuser and detector responses are flat over the probed wavelength
range. The RIN is calculated from 10 consecutive measurements, where a single measurement is an average of
two-hundred spectra corresponding to a measurement time of 3.2 s and an average of a thousand pulses per
spectral channel. The relative intensity variations between subsequent measurements can reach values as small
as 1 % at the wavelength of 3 µ m and increases to about 8 % at the wavelength of 3.5 µ m. The dramatic variation
in the 3.35–3.5 µ m region, which coincides with the long wavelength edge of the SC spectrum, is attributed to
the significant decrease in the power spectral density and large stochastic spectral power fluctuation of the SC
source. These intensity fluctuations are typical characteristics at both the short and long wavelength edges of a
supercontinuum ­spectrum27.
A black polypropylene (PP) sample is investigated using the developed active hyperspectral sensor (AHS),
and its corresponding reflectance is obtained using Eq. 1. The resulted spectrum is compared to a reference black
PP reflectance spectrum measured with an FTIR spectrometer (FT-MIR Rocket, ARCoptix) having a spectral
coverage of 2–6 µ m and a resolution of 4 cm−1 . The comparison of the black PP reflectance spectra, measured
with both the AHS and the FTIR spectrometer, is shown in Fig. 4a. The FTIR and AHS spectra demonstrate a very
close agreement. The two reflectance peaks in the FTIR spectra between 3.3 and 3.45 µ m are ascribed to the CH3
functional groups while the peak between 3.45–3.5 µ m is due to the CH 2 functional ­group28. Note that the AHS
reflectance peaks are slightly broader in comparison to that of the FTIR. This is due to the relatively low spectral
resolution of the FPI. Similarly, the FTIR and AHS reflectance spectra of black PE500 are compared in Fig. 4b.

Scientific Reports | (2022) 12:19392 | https://doi.org/10.1038/s41598-022-23758-w 3

Vol.:(0123456789)
 www.nature.com/scientificreports/

Figure 4.  (a), (b) and (c) compares the reflectance spectra of black polypropylene (PP), black polyethylene
(PE500) and white polyethylene (PE500), respectively. Solid blue line represents the spectra measured with FTIR
spectrometer and the red line with a marker represents that of the active hyperspectral sensor (AHS).

Both spectra are in excellent agreement as shown in the figure. The reflectance peak between 3.4–3.45 µ m spec-
tral region is assigned to the CH 2 functional group. Although both black PP and PE500 have somewhat similar
spectral absorption features, their reflectance spectra are clearly distinguishable as shown in Fig. 4a, b respectively.
The AHS is further extended to the measurement of white plastics to ascertain the reliability of the technique
as well as the impact of the additive coloring material. Fig. 4c presents a comparison between AHS and FTIR
reflectance spectra of a white PE500. We can see a very good agreement between them. Moreover, the reflectance
spectra of both the black and the white PE500 are very similar as can be seen in Fig. 4b,c, respectively. This is
because the additive coloring material has negligible impact on the optical properties of the polymer at this
wavelength range. This highlights the potential applicability of the sensor to differently colored plastics. There
is a slight variation in the AHS spectra of the plastics. The cause for this is attributed to the fact that samples
used in this work vary in surface quality (glossy, matte or rough surface), which leads to different light scattering
properties. Nonetheless, all the absorption peaks are clearly present in the measured spectra and can be used for
the identification and differentiation of the plastic species.

Conclusion
We developed a novel hyperspectral sensor using a mid-infrared supercontinuum light source and a tunable
MEMS-based Fabry-Pérot interferometer. The FPI enables robust wavelength selection across the probed spec-
tral range of the SC spectrum. The FPI is engineered to be standalone, thereby enabling robust sensor design

Scientific Reports | (2022) 12:19392 | https://doi.org/10.1038/s41598-022-23758-w 4

Vol:.(1234567890)
www.nature.com/scientificreports/

as positioning of the FPI on the sensor is flexible. For instance, placing the FPI in front of the laser would allow
target illumination with only the desired spectral components with relatively low power compared to the whole
SC spectrum, this is particularly important in applications where the target sample has very low damage thresh-
old. Our preliminary demonstration of line filtering with a FPI of similar principal mechanism operating in the
near infrared, with an aperture size of about 1.5 mm, showed a power handling of more than 10 W of continuous
wave of a 1–2 µ m SC laser. We demonstrated hyperspectral sensing of black polyethylene (PE500) and polypro-
pylene (PP) using the 3–3.5 µ m band of the SC spectrum. The measured reflectance spectra of the plastics are
compared to their reference measured with an FTIR spectrometer. An excellent agreement was observed between
the spectra. Furthermore, we measured the reflectance spectra of a white PE500 to emphasize the applicability
of the sensor to differently colored plastic samples. Our results showed, for the first time, the suitability of the
FPI for active hyperspectral identification of polymers. The sensor demonstrates a measurement rate of about
sixty-five spectra per second, limited by the tuning frequency of the FPI, and a relative intensity noise of 1–8 %
in the 3–3.5 µ m wavelength range. The main sources of noise in this work are the supercontinuum light source
spectrum instabilities and the readout noise associated with the detector, amplifier as well as the FPGA. It is
important to emphasize that the measurement speed and accuracy can be significantly improved by optimizing
the 0.5 duty cycle scanning frequency of the Fabry-Pérot interferometer to make use of the full cycle, which would
double the spectrum acquisition rate and enhance √ the signal level during the same acquisition time, and thus
increase the signal-to-noise ratio by a factor of 2 . Further improvement can be achieved by tailoring temporal
scanning profile so that noisy part of the spectrum would have longer integration time compared to other parts
of the spectrum. Furthermore, the biggest improvement can be realized by using better tailored supercontinuum
light source where the wavelength range of interest is not located at the far edge of the SC spectrum. This would
result in increase of average spectral density in that region, and significantly decrease shot to shot fluctuations
thereby enhancing the signal-to-noise ratio of measured spectra. The aforementioned optimizations would enable
real-time plastic wastes sorting and other hyperspectral sensing applications in the mid-infrared.

Data availibility
The datasets used and/or analysed during the current study are available from the corresponding author upon
reasonable request.

Received: 29 June 2022; Accepted: 4 November 2022

References
1. Adão, T. et al. Hyperspectral imaging: A review on uav-based sensors, data processing and applications for agriculture and forestry.
Remote Sens. 9, 1110 (2017).
2. Lu, G. & Fei, B. Medical hyperspectral imaging: A review. J. Biomed. Opt. 19, 010901 (2014).
3. Krupnik, D. & Khan, S. Close-range, ground-based hyperspectral imaging for mining applications at various scales: review and
case studies. Earth Sci. Rev. 198, 102952 (2019).
4. Dudley, J. M., Genty, G. & Coen, S. Supercontinuum generation in photonic crystal fiber. Rev. Mod. Phys. 78, 1135 (2006).
5. Genty, G., Coen, S. & Dudley, J. M. Fiber supercontinuum sources. JOSA B 24, 1771–1785 (2007).
6. Saleh, A., Aalto, A., Ryczkowski, P., Genty, G. & Toivonen, J. Short-range supercontinuum-based lidar for temperature profiling.
Opt. Lett. 44, 4223–4226 (2019).
7. Saleh, A., Ryczkowski, P., Genty, G. & Toivonen, J. Supercontinuum lidar for industrial process analysis. Opt. Express 29, 42082–
42089 (2021).
8. Kilgus, J., Duswald, K., Langer, G. & Brandstetter, M. Mid-infrared standoff spectroscopy using a supercontinuum laser with
compact fabry-pérot filter spectrometers. Appl. Spectrosc. 72, 634–642 (2018).
9. Kääriäinen, T., Jaanson, P., Vaigu, A., Mannila, R. & Manninen, A. Active hyperspectral sensor based on mems fabry-pérot inter-
ferometer. Sensors 19, 2192 (2019).
10. Kääriäinen, T. & Dönsberg, T. Active hyperspectral imager using a tunable supercontinuum light source based on a mems fabry-
perot interferometer. Opt. Lett. 46, 5533–5536 (2021).
11. Petersen, C. R. et al. Mid-infrared supercontinuum covering the 1.4–13.3 µ m molecular fingerprint region using ultra-high na
chalcogenide step-index fibre. Nat. Photonics 8, 830–834 (2014).
12. Zhao, Z. et al. Mid-infrared supercontinuum covering 2.0–16 µ m in a low-loss telluride single-mode fiber. Laser Photonics Rev
11, 1700005 (2017).
13. Liu, K., Liu, J., Shi, H., Tan, F. & Wang, P. High power mid-infrared supercontinuum generation in a single-mode zblan fiber with
upto 21.8 w average output power. Opt. Express 22, 24384–24391 (2014).
14. Kedenburg, S. et al. High repetition rate mid-infrared supercontinuum generation from 1.3 to 5.3 µ m in robust step-index tellurite
fibers. JOSA B 34, 601–607 (2017).
15. Salem, R. et al. Mid-infrared supercontinuum generation spanning 1.8 octaves using step-index indium fluoride fiber pumped by
a femtosecond fiber laser near 2 µ m. Opt. Express 23, 30592–30602 (2015).
16. Eslami, Z., Ryczkowski, P., Salmela, L. & Genty, G. Low-noise octave-spanning mid-infrared supercontinuum generation in a
multimode chalcogenide fiber. Opt. Lett. 45, 3103–3106 (2020).
17. Mikkonen, T. et al. Broadband cantilever-enhanced photoacoustic spectroscopy in the mid-ir using a supercontinuum. Opt. Lett.
43, 5094–5097 (2018).
18. Petersen, C. R. et al. Mid-infrared multispectral tissue imaging using a chalcogenide fiber supercontinuum source. Opt. Lett. 43,
999–1002 (2018).
19. Zorin, I. et al. Mid-infrared fourier-domain optical coherence tomography with a pyroelectric linear array. Opt. Express 26,
33428–33439 (2018).
20. Hollas, J. M. Modern Spectroscopy (John Wiley & Sons, 2004).
21. Signoret, C., Caro-Bretelle, A.-S., Lopez-Cuesta, J.-M., Ienny, P. & Perrin, D. Mir spectral characterization of plastic to enable
discrimination in an industrial recycling context: I. Specific case of styrenic polymers. Waste Manag. 95, 513–525 (2019).
22. Rissanen, A. et al. Vtt’s fabry-perot interferometer technologies for hyperspectral imaging and mobile sensing applications. . In
Moems and Miniaturized Systems Xvi Vol. 10116. pp. 119–130 (International Society for Optics and Photonics, 2017).
23. Akujärvi, A., Guo, B., Mannila, R. & Rissanen, A. Moems fpi sensors for nir-mir microspectrometer applications. In Moems and
Miniaturized Systems Xv Vol. 9760. pp. 117–124 (International Society for Optics and Photonics, 2016).

Scientific Reports | (2022) 12:19392 | https://doi.org/10.1038/s41598-022-23758-w 5

Vol.:(0123456789)
 www.nature.com/scientificreports/

24. Rissanen, A., Akujärvi, A., Antila, J. E., Blomberg, M. & Saari, H. K. MOESMs miniature spectrometers using tuneable fabry-perot
interferometers. J. Micro/Nanolithogr. MEMS MOEMS 11, 023003 (2012).
25. Rissanen, A., Mannila, R., Tuohiniemi, M., Akujärvi, A. & Antila, J. Tunable moems fabry-perot interferometer for miniaturized
spectral sensing in near-infrared. in MOEMS and Miniaturized Systems XIII Vol. 8977. pp. 243–250 (International Society for
Optics and Photonics, 2014).
26. Amiot, C., Aalto, A., Ryczkowski, P., Toivonen, J. & Genty, G. Cavity enhanced absorption spectroscopy in the mid-infrared using
a supercontinuum source. Appl. Phys. Lett. 111, 061103 (2017).
27. Kudlinski, A. et al. Control of pulse-to-pulse fluctuations in visible supercontinuum. Opt. Express 18, 27445–27454 (2010).
28. Becker, W., Sachsenheimer, K. & Klemenz, M. Detection of black plastics in the middle infrared spectrum (MIR) using photon
up-conversion technique for polymer recycling purposes. Polymers 9, 435 (2017).

Acknowledgements
A.S. acknowledges the support from Finnish Foundation for Technology Promotion, and Finnish Cultural Foun-
dation. The authors also acknowledge Academy of Finland (Flagship PREIN, 320165).

Author contributions
A.S. conducted the experiments. J.T. and G.G. supervised the project. All authors contributed to the data analysis
and writing of the manuscript.

Competing interests 
The authors declare no competing interests.

Additional information
Correspondence and requests for materials should be addressed to A.S.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note  Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.
Open Access  This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the
Creative Commons licence, and indicate if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by/4.​0/.

© The Author(s) 2022

Scientific Reports | (2022) 12:19392 | https://doi.org/10.1038/s41598-022-23758-w 6

Vol:.(1234567890)