Miniature Optical Particle Counter and Analyzer Involving a Fluidic-Optronic CMOS Chip Coupled with a Millimeter-Sized Glass Optical System †

sensors

Article Miniature Optical Particle Counter and Analyzer Involving a Fluidic-Optronic CMOS Chip Coupled with a Millimeter-Sized Glass Optical System †

Gabriel Jobert 1,2,*, Pierre Barritault 1, Maryse Fournier 1, Cyrielle Monpeurt 1, Salim Boutami 1,Cécile Jamois 2, Pietro Bernasconi 3, Andrea Lovera 3, Daniele Braga 3 and Christian Seassal 2

1 CEA-LETI Minatec, Université Grenoble-Alpes, F-38000 Grenoble, France; [email protected] (P.B.); [email protected] (M.F.); [email protected] (C.M.); [email protected] (S.B.) 2 Institut des Nanotechnologies de Lyon, UMR CNRS 5270 Ecole Centrale de Lyon, Université de Lyon, F-69134 Ecully, France; [email protected] (C.J.); [email protected] (C.S.) 3 FEMTOprint, CH-6933 Muzzano, Switzerland; [email protected] (P.B.); [email protected] (A.L.); daniele.braga@fluxim.com (D.B.) * Correspondence: [email protected] † This manuscript is extension version of the conference paper: Jobert, G.; Fournier, M.; Boutami, S.; Jamois, C.; Lovera, A.; Braga, D.; Seassal, C. Millimeter-Sized Particle Sensor Using a Wide Field of View Monolithic Lens Assembly for Light Scattering Analysis in Fourier Domain. In Proceedings of the Photonic Instrumentation Engineering VII; Soskind, Y., Busse, L.E., Eds.; SPIE: San Francisco, CA, USA, 2020; p. 24. Abstract: Our latest advances in the field of miniaturized optical PM sensors are presented. This sen- Citation: Jobert, G.; Barritault, P.; sor combines a hybrid fluidic-optronic CMOS (holed retina) that is able to record a specific irradiance Fournier, M.; Monpeurt, C.; Boutami, pattern scattered by an illuminated particle (scattering signature), while enabling the circulation of S.; Jamois, C.; Bernasconi, P.; Lovera, particles toward the sensing area. The holed retina is optically coupled with a monolithic, millimeter- A.; Braga, D.; Seassal, C. Miniature sized, refracto-reflective optical system. The latter notably performs an optical pre-processing of Optical Particle Counter and signatures, with a very wide field of view of scattering angles. This improves the sensitivity of the Analyzer Involving a Fluidic-Optronic CMOS Chip sensors, and simplifies image processing. We report the precise design methodology for such a sensor, Coupled with a Millimeter-Sized as well as its fabrication and characterization using calibrated polystyrene beads. Finally, we discuss Glass Optical System . Sensors 2021, its ability to characterize particles and its potential for further miniaturization and integration. 21, 3181. https://doi.org/10.3390/ s21093181 Keywords: air quality; particulate matter sensor; light-scattering; CMOS image sensor; miniature optics; micro-fabrication Academic Editor: Stefan Hippler

Received: 11 March 2021 Accepted: 28 April 2021 1. Introduction Published: 3 May 2021 1.1. Particulate Matter and Air Quality Monitoring Among air pollutants, there is a class defined as Particulate Matter (PM), which Publisher’s Note: MDPI stays neutral consists of a set of particles suspended in the air. The size of these particles can range with regard to jurisdictional claims in published maps and institutional affil- from a few nanometers up to a few tens of micrometers. PM comes in a wide variety of iations. shapes and chemical compositions, originating from natural or anthropogenic processes. Anthropogenic PM poses serious sanitary issues [1,2]; for instance, it can cause cardiac, cardiovascular, respiratory and neuro-degenerative diseases. In regulatory texts, PM is ranked into subclasses according to granulometry. Here we focus on the PM2.5 subclass (particles smaller than 2.5 µm) that is a known carcinogenic Copyright: © 2021 by the authors. agent (group 1, IARC classification) [3,4]. It has been shown that exposure to PM2.5 can Licensee MDPI, Basel, Switzerland. cause premature death [5]. Additionally, the PM subclass (smaller than 1 µm), which is This article is an open access article 1 unregulated at the moment, can cause more harm, due to its capacity to penetrate deeply distributed under the terms and conditions of the Creative Commons into the human respiratory system [6–8]. Attribution (CC BY) license (https:// Beyond health considerations, anthropogenic PM pollution appears to have an adverse creativecommons.org/licenses/by/ effect on climate [9,10], especially for Black-Carbon (BC) soot. Indeed, meteorological 4.0/). models suggest that controlling the emissions of BC could be one of the most efficient

Sensors 2021, 21, 3181. https://doi.org/10.3390/s21093181 https://www.mdpi.com/journal/sensors Sensors 2021, 21, 3181 2 of 23

ways to mitigate global warming, even more than controlling carbon monoxide or methane emissions [11–14]. Air quality monitoring (PM, gaseous pollutants, relative humidity, etc.) is classically carried out by stations with laboratory-grade instruments [15]. The major disadvantage of such a monitoring model is that it does not allow the collection of pollution levels with sufﬁcient spatial resolution (e.g., street-to-street air quality, or indoor air quality [16–18]). With the emergence of connected on-board systems (IoT, Internet of Things) equipped with geo-localization modules, we see the possibility that a network of portable environmental sensors can measure pollution levels with high spatio-temporal resolution [19]. Such monitoring networks could also achieve great accuracy by sharing and correcting measurements with data from weather conditions (humidity, temperature) [20] and measurements provided by climate monitoring satellites [21–25]. However, an air quality monitoring model based on a broad sensor network is only possible with widely distributed sensors. This underlines the need for inexpensive and portable air quality monitors [26,27].

1.2. Toward Further Miniaturization of Optical PM Sensors

Many techniques can be used to measure PM2.5 concentrations [28]; however, the optical method is preferred for inexpensive and portable PM2.5 sensors thanks to their simplicity and good sensitivity, especially light-scattering sensors [29,30]. The general setup involves a light source that illuminates a sensing volume where particles can scatter light. An off-axis photodetector then captures this scattered light, while avoiding the illumination beam [31]. Numerous research groups are actively working on down-scaling optical light-scattering PM2.5 & PM1 sensors. The methodology usually involves silicon microfabrication techniques and co-integration of optical and electro-optical elements and air micro-fluidics, the end goal being the full integration of a PM sensor ‘on-a-chip’ [32–35]. A simple light-scattering setup is able to perform an indirect measurement of the total mass concentration of the sampled aerosol. However, its composition and size distribution is unknown. An inertial element can filter PM2.5 from PM10 before the optical sensing stage [36]. For instance, this can be achieved by fluidic devices like cyclones [37] or Virtual Impactors (VI) [38]. For a particle of given size, morphology and optical index, we can describe the theoretical angular distribution of the scattered light (what we call the ‘scattering signature’); for example, using the Lorenz–Mie theory applied to spherical particles, which will be discussed in the next subsection. Retrieving an experimental signature enables the estimation of optical and geometrical parameters of particles (and subsequent classification) by performing an inverse problem. For example, precise scattering angular photometry can be achieved by using a rotating detection arm or periscope [39,40] with an optical Fourier transform imaging setup that uses an array of photodetectors coupled with mirrors [41] or lenses [42]. Optical Particle Counters (OPC) can usually be described as a light-scattering sensor [43] that can detect only one particle at once. The use of focusing beam optics and/or particle focusing nozzles is an efficient way to restrict the sampling volume [44]. In this document, we report the development of an OPC that records a particle’s angular scattering efficiency (scattering signature) by using a Fourier imaging setup. This setup is designed to enable the retrieval of the particle’s diameter and refractive index [45] for further classification, as explained earlier.

1.3. Lorenz-Mie Theory and Scattering Signatures Modelling and understanding the phenomenon of elastic light scattering by particles is of high importance and enables the design of analysis systems capable of classifying particles (particle number, size distribution, material, geometry etc.) in order to, for example, characterize ambient air and identify environmental or sanitary hazards, or Sensors 2021, 21, x FOR PEER REVIEW 3 of 23

Modelling and understanding the phenomenon of elastic light scattering by particles is of high importance and enables the design of analysis systems capable of classifying Sensors 2021, 21, 3181 particles (particle number, size distribution, material, geometry etc.) in order to, for exam3 of 23‐ ple, characterize ambient air and identify environmental or sanitary hazards, or to create alarm systems that are less sensitive to false positives, in order to operate in contaminated environments. to createThere alarm are many systems models that that are lesstake sensitive into account to false several positives, parameters in order such to as operate lighting in conditions,contaminated particle environments. morphology, orientation, size range, multiplicity of particles, etc. Here,There we will are mainly many modelsfocus on that the takeLorenz–Mie into account theory several [46,47]. parameters This is an suchelectromagnetic as lighting theoryconditions, valid particle for a homogeneous morphology, sphere orientation, with sizean arbitrary range, multiplicity diameter that of particles, is illuminated etc. Here, by we will mainly focus on the Lorenz–Mie theory [46,47]. This is an electromagnetic the- a monochromatic plane wave. It shows, notably, that larger particles (when compared to ory valid for a homogeneous sphere with an arbitrary diameter that is illuminated by a the illumination wavelength) exhibit oscillations in the scattering signature which are not monochromatic plane wave. It shows, notably, that larger particles (when compared to seen for smaller particles. The width and spacing of those oscillations is related to both the illumination wavelength) exhibit oscillations in the scattering signature which are not the sphere’s diameter and refractive index. seen for smaller particles. The width and spacing of those oscillations is related to both the Other, broader theories can be applied to model scattering by non‐spherical particles sphere’s diameter and refractive index. or aggregates, as reported in references [48,49]. Other, broader theories can be applied to model scattering by non-spherical particles or aggregates, as reported in references [48,49]. 2. Designing the Optical System 2.1.2. Designing Holed Retina, the an Optical Opto‐Fluidic System CMOS Chip 2.1. HoledIn order Retina, to properly an Opto-Fluidic record particle CMOS Chip scattering signatures, a CMOS (or ’Complemen‐ tary metalIn order oxide to properly semi‐conductor’, record particle which scattering refers to signatures, the manufacturing a CMOS (ortechnology) ’Complementary imager chipmetal (referred oxide semi-conductor’, to as the ‘holed retina’) which refers has been to the developed manufacturing and customized technology) for imagerour specific chip requirements(referred to as [50] the on ‘holed the basis retina’) of the has CreaPYX been developed standard and platform customized (by Pyxalis) for our [51]. specific The specificityrequirements of our [50 ]design on the is basis to allow of the the CreaPYX formation standard of a fluidic platform channel (by Pyxalis)through [the51]. chip. The Thespecificity latter were of our manufactured design is to allow first the through formation a 200 of mm a fluidic CMOS channel shuttle through run (multi the chip.‐project The wafer)latter were of an manufactured undisclosed firstindustrial through fab a 200(see, mm Figure CMOS 1a). shuttle The fabrication run (multi-project is followed wafer) by ofa specificallyan undisclosed developed industrial deep fab etching (see, Figure post1‐a).process The fabrication performed is in followed‐house (200 by a mm specifically MEMS foundry).developed This deep allowed etching us post-process to drill vertical performed fluidic in-house channels (200 at wafer mm MEMS scale. foundry).The chips This are thenallowed diced us and to drill wire vertical‐bonded fluidic (ball bonding) channels onto at wafer a ceramic scale. carrier The chips that areis also then drilled diced (see, and Figurewire-bonded 1b). (ball bonding) onto a ceramic carrier that is also drilled (see, Figure1b). TheThe overall overall footprint footprint of of the the system system is is that that of of the the ceramic ceramic package package (29.2 (29.2 mm, mm, square). square). NoteNote that that the chip is only sized 6 mmmm ×× 5.95.9 mm. mm. In In the the future, future, the the CMOS CMOS chip will be bondedbonded directly directly onto PCB, in order to further reduce the overall footprint.

Figure 1. (a) Photography of a 200 mm wafer that contains the custom image sensors. ( b)) Photography of a holed retina bonded on the ceramic carrier.

InIn Figure Figure 2a,2a, we we present present an an optical optical image image of ofthe the manufactured manufactured holed holed retina. retina. It is com It is‐ posedcomposed of an of uncommon an uncommon dual matrix dual matrix core part core that part features that features two detection two detection areas (sub areas‐retina (sub- 1retina and 2) 1 andspaced 2) spaced by an empty by an emptyarea where area wherea fluidic a ﬂuidicchannel channel is drilled is drilled (oblong (oblong profile, proﬁle, 2 mm ×2 0.5 mm mm).× 0.5 The mm). size The of both size of matrices both matrices is by 75 is × by 281 75 pixels,× 281 pixels,with a with10 μm a 10pixel’sµm pixel’s pitch. pitch.

Sensors 2021, 21, 3181 4 of 23 Sensors 2021, 21, x FOR PEER REVIEW 4 of 23

FigureFigure 2. 2.(a)( aOptical) Optical image image of ofthe the fabricated fabricated holed holed retina. retina. (b ()b Image) Image example example obtained obtained focus focus mode. mode.

Image examplesImage are examples shown in Figureare shown2b. Lines in Figure 1–75 and 2b. 76–150Lines 1–75 correspond and 76–150 to the correspond first to the and second sub-retinae,first and second respectively. sub‐retinae, In the respectively. image, we see In the a discontinuous image, we see image a discontinuous with two image with separate fieldstwo that separate have been fields stitched that have together, been causedstitched by together, the blind caused central by part. the blind The pixel central part. The has a 5T architecture,pixel has anda 5T allowsarchitecture, for a global-shutter and allows for low-noise a global‐shutter readout low mode,‐noise with readout two mode, with gains high dynamictwo gains range high [ 52dynamic]. More range details [52]. on theMore holed details retina on concerningthe holed retina the design, concerning the de‐ fabrication andsign, characterization fabrication and are characterization provided in reference are provided [31]. in reference [31]. 2.2. Difficulties Encoutered Using a Lens-less Setup 2.2. Difficulties Encoutered Using a Lens‐less Setup In reference [31], we have used the holed retina in a first generation of optical PM In reference [31], we have used the holed retina in a first generation of optical PM sensor, which is the base frame for the second generation presented here. The former is sensor, which is the base frame for the second generation presented here. The former is able to count particles and record scattering signatures that are projected onto the retina able to count particles and record scattering signatures that are projected onto the retina using a lens-less configuration. With this first-generation PM sensor, the Field of View using a lens‐less configuration. With this first‐generation PM sensor, the Field of View (FoV) is limited by the geometry as we can evaluate scattering signatures within only a (FoV) is limited by the geometry as we can evaluate scattering signatures within only a few tens of degrees. There is little margin of improvement that can be made using lens-free few tens of degrees. There is little margin of improvement that can be made using lens‐ planar projection. Moreover, the uncertain position of the particle within the illumination free planar projection. Moreover, the uncertain position of the particle within the illumi‐ beam can result in an affine transformation of the image as well as a blurring effect, with a nation beam can result in an affine transformation of the image as well as a blurring effect, shifted FoV. This last part was taken into account by our image analysis procedure but one with a shifted FoV. This last part was taken into account by our image analysis procedure may want to simplify the computing component as much as possible for future portable applications.but one may want to simplify the computing component as much as possible for future We addressportable those applications. identified problems through the design and fabrication of a miniature glass optical system,We address which those will identified be described problems in the through following. the design and fabrication of a minia‐ ture glass optical system, which will be described in the following. 2.3. Achieving Position Insensitivity In most2.3. non-miniaturized Achieving Position light-scattering Insensitivity photometers, the position of the particle has little impact onIn themost measurements. non‐miniaturized Indeed, light the‐scattering range of photometers, positions accessible the position by the of the particle particle is veryhas small little asimpact compared on the to measurements. the optical path Indeed, of scattered the range rays. of In positions our case, accessible the by the impacts of theparticle particle is displacement very small as include compared translation, to the optical magnification path of andscattered rotation rays. of theIn our case, the recorded image,impacts as it of was the seen particle in our displacement reference [31 include]; a superposition translation, of magnification shifted signatures and rotation of the can be generatedrecorded by a image, distribution as it was of particles, seen in our resulting reference in a [31]; position a superposition blur; as a corollary, of shifted signatures the detectioncan of a be moving generated particle by a (with distribution respect of to particles, the image resulting integration in a time) position results blur; in aas a corollary, motion blur. the detection of a moving particle (with respect to the image integration time) results in a We willmotion propose blur. a way to achieve position insensitivity. A setup equivalent to a far-field setup is theWe common will propose d-f lens a way system to achieve (or Fourier-domain position insensitivity. imaging A setup), setup whereequivalent to a far‐ the retina isfield placed setup at the is the Image common Focal d Plane‐f lens (IFP) system of a(or thin Fourier lens‐ (simplifieddomain imaging model) setup), so where the that the objectretina can beis placed seen at at infinity the Image [53]. Focal Consequently, Plane (IFP) the of retina a thin can lens record (simplified an angular model) so that the image by refocusingobject can parallel be seen rays. at infinity Indeed, [53]. the Consequently, image is not modified the retina by can displacing record an the angular image particle becauseby refocusing the same scattered parallel rays rays. are Indeed, always the seen image under is thenot same modified angles by (see displacing Figure the particle 3). We will showbecause that the this same optical scattered pre-treatment rays are always of the scattering seen under signature the same dramatically angles (see Figure 3). We simplifies thewill analysis show of that the this image optical taken pre by‐treatment the retina, of removing the scattering the need signature for energy-costly dramatically simplifies image processing.the analysis of the image taken by the retina, removing the need for energy‐costly image processing.

Sensors 2021 21 Sensors 2021, , 21Sensors,, 3181 x FOR 2021 PEER, 21 REVIEW, x FOR PEER REVIEW 55 ofof 2323 5 of 23

Figure 3. Illustration of the d-f setup that performs angular imaging by refocusing parallel scat- Figure 3. IllustrationFigure 3. of Illustration the d‐f setup of thethat d performs‐f setup that angular performs imaging angular by refocusing imaging by parallel refocusing scattered parallel rays. scattered rays. tered rays. Note that suchNote a thatsetup such can a achieve setup can position achieve insensitivity position insensitivity if we consider if we a thinconsider lens a thin lens Note that such a setup can achieve position insensitivity if we consider a thin lens with infinitewith lateral infinite extension. lateral For extension. a real lens For witha real a lensfinite with pupil, a finite the scattering pupil, the signature scattering signature with inﬁnite lateral extension. For a real lens with a ﬁnite pupil, the scattering signature is is contained within the projection of the lens’ pupil on the retina. However, one must keep containedis contained within within the the projection projection of theof the lens’ lens’ pupil pupil on on the the retina. retina. However, However, one one must must keep keep in in mind that the projection of the lens’ pupil depends on the particle’s position, which will mindin mind that that the the projection projection of the of the lens’ lens’ pupil pupil depends depends on theon the particle’s particle’s position, position, which which will will be be discussed later. discussedbe discussed later. later.

2.4.2.4. AssymetricAssymetric2.4. CrownCrown Assymetric AssemblyAssembly Crown andand Assembly SignatureSignature and Reconstruction Reconstruction Signature Reconstruction LetLet usus considerconsiderLet us aa lens lensconsider systemsystem a lens coupledcoupled system toto a acoupled detector detector to array.array. a detector A A major major array. factor factor A major that that limits limitsfactor that limits thethe FoVFoV isis thethethe pupilpupil FoV ofisof the the pupil system.system. of OurtheOur system. ideaidea behindbehind Our idea improvingimproving behind the theimproving FoV FoV for for light thelight FoV scattering scattering for light scattering analysisanalysis isis totoanalysis mergemerge severalisseveral to merge ofof these these several subsystems subsystems of these subsystems around around a a perpendicular perpendicular around a perpendicular air air channel channel in airin a achannel in a crowncrown shape, shape,crown as as illustrated illustrated shape, as in in illustrated Figure Figure4 a.4a. in Figure 4a. TheThe edgesedges ofofThe eacheach edges subsystemsubsystem of each may maysubsystem contain contain may aberrations aberrations contain or aberrationsor unwanted unwanted or artefacts artefacts unwanted that that artefacts we we that we callcall ‘blind‘blind fields’. fields’.call ‘blind ByBy takingtaking fields’. advantageadvantage By taking ofof advantage thethe symmetrysymmetry of the ofof symmetry thethe scatteringscattering of the signaturesignature scattering around around signature around thethe optical optical axis, axis,the oneoptical one can can axis, sacrifice sacrifice one acan bita bit sacrifice of of redundancy redundancy a bit of by redundancy by facing facing half half ofby the offacing lensesthe lenseshalf toward of toward the the lenses toward blindthe blind fields fields ofthe the blindof other the fields other halves. of halves. the The other resultingThe halves. resulting design The design resulting is an asymmetricis an design asymmetric is crown an asymmetric crown shape shape with crown shape anwith uneven an uneven numberwith number an of uneven optical of optical number subsystems subsystems of optical arranged subsystems arranged around around a arranged perpendicular a perpendicular around air a channel. perpendicular air chan An‐ air chan‐ examplenel. An example is shownnel. An is in shownexample Figure in4 , isFigure with shown three 4, within subsystemsFigure three 4 ,subsystems with (modelled three subsystems (modelled as thin lenses). as (modelled thin lenses). as thin lenses).

FigureFigure 4. 4.( a(a)) Principle FigurePrinciple 4. of (ofa) blind blindPrinciple ﬁelds fields of asymmetrically asymmetrically blind fields asymmetrically coupled coupled to to clear clear coupled ﬁelds. fields. to ( b(clearb)) Lenses Lenses fields. coupled coupled (b) Lenses with with coupled imagers imagers with in in Fourier Fourier imagers in Fourier planes.planes. ( (cc)) Principle Principleplanes. of of(c overlapping) overlapping Principle of sub-FoV suboverlapping‐FoV for for signature signaturesub‐FoV reconstruction.for reconstruction. signature reconstruction.

InIn the the IFP IFP of ofIn each each the subsystem, IFPsubsystem, of each one subsystem,one can can reconstruct reconstruct one can an reconstructan experimental experimental an scattering experimental scattering signature signa scattering‐ signa‐ Stureexp( θ𝑆)(because𝜃ture (because there𝑆𝜃 is there an (because injunctive is an injunctive there law is linking an law injunctive alinking position lawa position on linking IFP and on a the positionIFP scattering and theon IFP scatter angle) and‐ the scatter‐ alonging angle) each associatedingalong angle) each sub-FoV. alongassociated each Figure subassociated4c‐FoV. schematizes Figure sub‐FoV. the4c signatureschematizesFigure 4c seen schematizes the through signature all the three seensignature seen subsystemsthrough all through (FS,three S, subsystems and all BS,three which subsystems (FS, stand S, and for (FS,BS, For-Side, whichS, and Side, standBS, which and for Back-Side).For stand‐Side, for Side, For As ‐ explainedandSide, Back Side,‐ and Back‐ earlier, we design the sub-FoVs with an overlapping region (grey areas), in order to mitigate the defects at the edge of each sub-FoV by taking advantage of the asymmetric design.

Sensors 2021, 21, x FOR PEER REVIEW 6 of 23

Sensors 2021, 21, 3181 Side). As explained earlier, we design the sub‐FoVs with an overlapping region6 of (grey 23 ar‐ eas), in order to mitigate the defects at the edge of each sub‐FoV by taking advantage of the asymmetric design. As explainedAs explained in the in introduction, the introduction, we try we totry retrieve to retrieve certain certain particle’s particle’s parameters, parameters, such 𝐷 𝑛 such asas diameter diameterD pand and complex complex refractive refractive index index n.. To To do do so,so, wewe havehave toto compare an exper‐ experimentalimental signaturesignature S𝑆exp(𝜃θ) withwith a a predictionprediction from a a model, model, usually usually based based on on the the Lorenz– Lorenz–Mie,Mie, and, and, finally, ﬁnally, perform perform the inverse problem.problem. The The optimization optimization schematic schematic used used is is pre‐ presentedsented in Figure in Figure5. 5.

FigureFigure 5. 5.SchematicSchematic of of the the optimization optimization procedure procedure used used for for particle particle characterization. characterization.

A wayA to way compare to compare data data retrieved retrieved from from an experimentalan experimental image image with with data data from from a a mod‐ modelledelled one one has has to to be be defined.defined. For For the the lens lens-less‐less setup, setup, we’ve we’ve developed developed a procedure a procedure to reduce to reducean image an image (either (either experimental experimental or modelled) or modelled) in order in order to compare to compare those. those. This This procedure procedurerequires requires a prior a prior computation computation of the of modelled the modelled image. image. All these All steps, these which steps, are which relatively are relativelyexpensive expensive in calculation, in calculation, had to be had called to be regularly called regularly in an optimization in an optimization loop. Moreover, loop. as Moreover,the image as the obtained image obtained depends depends on geometrical on geometrical parameters parameters such as such the position as the position of the particle of the particle𝑦, thesey pmust, these be must taken be into taken account into account in the optimization in the optimization (see, reference (see, reference [31]). [31]). In the Fourier-domainIn the Fourier‐domain setup, setup, the procedure the procedure is simplified. is simplified. It requires It requires performing performing the the simple step of signature reconstruction Sexp(θ), and𝑆 this𝜃 only once. The computed signature simple step of signature reconstruction , and this only once. The computed signa‐ S θ, Dp, n is retrieved directly from the Lorenz–Mie theory, and it is the only step within ture 𝑆𝜃, 𝐷, 𝑛 is retrieved directly from the Lorenz–Mie theory, and it is the only step the optimization loop. The minimization criterion e Dp, n is also much simpler than within the optimization loop. The minimization criterion 𝜖𝐷, 𝑛 is also much simpler previously,than andpreviously, is a simple and mean is a simple of a one-dimensional mean of a one‐dimensional deviation function: deviation function: Z e Dp, n ∝ Sexp(θ) − S θ, Dp, n dθ 𝜖𝐷θ∈FoV, 𝑛 ∝ 𝑆𝜃 𝑆𝜃, 𝐷, 𝑛𝑑𝜃 ∈ D n In the future,In the future, one could one testcould different test different definitions definitions for e forp, 𝜖𝐷, using,, 𝑛, forusing, instance, for instance, an an EuclideanEuclidean norm insteadnorm instead of a simple of a simple deviation. deviation. It could It alsocould be also interesting be interesting to implement to implement ponderationsponderations at the center at the ofcenter each of sub-FoV each sub where‐FoV thewhere aberrations the aberrations are softer. are softer. ApartApart from the from simplified the simplified image image processing, processing, such a such sampling a sampling geometry geometry allows allows for for collectingcollecting the most the of most the scattered of the scattered flux, especially flux, especially around low around scattering low scattering and backscattering and backscatter‐ anglesing where angles the where scattering the efficiencyscattering isefficiency usually high.is usually This high. increases This the increases sensitivity the sensitivity of the of sensorthe to small sensor particles to small compared particles compared to a lens-less to a setup. lens‐less setup. Note thatNote the that illumination the illumination path must path be must completely be completely free of optical free of elements, optical elements, so that the so that sensitivity to scattering by diopters and ghost reflections is mitigated. This part is critical the sensitivity to scattering by diopters and ghost reflections is mitigated. This part is crit‐ in order to ensure a good signal to offset ratio, as it was highlighted in a previous version ical in order to ensure a good signal to offset ratio, as it was highlighted in a previous of this design, with five subsystems (see, reference [54]). version of this design, with five subsystems (see, reference [54]). 2.5. Compact Optical Subsystems with Coincidental Fourier Planes 2.5. Compact Optical Subsystems with Coincidental Fourier Planes Each optical subsystem must meet a certain set of specifications: it must be as compact as possible;Each it must optical also subsystem fold its Fourier must meet plane a (i.e., certain IFP) set onto of specifications: a plane perpendicular it must be to as com‐ the fluidicpact channelas possible; so thatit must all subsystems’also fold its Fourier Fourier plane planes (i.e., are IFP) coincidental. onto a plane This perpendicular set-up to enables the use of a single holed retina to image all Fourier planes on separate regions of interest (ROIs). As in the previous example, we select an assembly made with three subsystems: a first 90◦ centered system called S (‘Side’) subsystem and two symmetrical 90 ± 30◦ called FS and BS (‘Front-Side’ & ‘Back-Side’). Those subsystems are composed with a front spherical

Sensors 2021, 21, x FOR PEER REVIEW 7 of 23

the fluidic channel so that all subsystems’ Fourier planes are coincidental. This set‐up en‐ ables the use of a single holed retina to image all Fourier planes on separate regions of interest (ROIs). As in the previous example, we select an assembly made with three subsystems: a Sensors 2021, 21, 3181 7 of 23 first 90° centered system called S (‘Side’) subsystem and two symmetrical 90 30° called FS and BS (‘Front‐Side’ & ‘Back‐Side’). Those subsystems are composed with a front spherical diopter, an internal reflector and a flat back diopter. The reflecting surface can diopter,be coated an with internal a metallic reflector reflective and a flat layer back such diopter. as gold. The In reflectingthis section, surface the design can be and coated anal‐ withysis atools metallic are made reflective using layer ray suchtracing as (RT) gold. tools In this (Zemax section, Optic the designStudio, and Sequential analysis mode). tools areLet made us focus using first ray on tracing the S subsystem (RT) tools (Zemaxillustrated Optic in Figure Studio, 6a Sequential (note that mode). the (X,Y) Let coordinate us focus firstsystem on the is used S subsystem only in this illustrated subsection, in Figure and is6 specifica (note thatto a subsystem the (X,Y) coordinate alone. A new system coordi is ‐ usednate onlysystem in this will subsection, be used later and when is specific each subsystem to a subsystem will be alone. arranged A new around coordinate the air system chan‐ willnel). be used later when each subsystem will be arranged around the air channel).

FigureFigure 6. 6.( a()a) Design Design of of the the S S subsystem subsystem alongside alongside its its truncated truncated model. model. (b ()b Design) Design of of the the FS/BS FS/BS subsystem subsystem alongside alongside its its truncatedtruncated model. model.

TheThe system system is is designed designed with with full full compatibility compatibility with with a a sub-retina sub‐retina and and has has a a reasonably reasonably longlong Back Back Focal Focal Length Length (BFL (BFL = = 1 1 mm). mm). This This allows allows us us to to have have an an efficient efficient cloaking cloaking device device withwith efficient efficient stray stray light light protection protection that that fits fits within within the the BFL BFL (between (between the the optical optical piece piece and and thethe retina). retina). Only Only the the front front surface surface is is curved, curved, the the back back surface surface being being a a simple simple flat flat surface surface unprocessedunprocessed from from the the substrate, substrate, thus thus removing removing the the need need for for a a backside backside process. process. The The front front ◦ diopterdiopter is is slanted slanted by by 10 10°in in order order to to facilitate facilitate an an eventual eventual unmolding unmolding process. process. Then, Then, we we cancan improve improve the the uniformity uniformity of of the the PSF PSF (Point (Point Spread Spread Function) Function) by by engineering engineering an an internal internal reflectivereflective surface surface that that corrects corrects aberrations. aberrations. This This reflector reflector is is a freeforma freeform surface surface defined defined by by a Chebycheva Chebychev 2D-polynomial 2D‐polynomial [55 [55].]. Due Due to to the the relative relative position position of of the the front front surface surface and and the the ◦ sub-retina,sub‐retina, the the reflector reflector is is slanted slanted at at 33.5 33.5°so so that that the the optical optical axis axis is is rightfully rightfully aimed aimed at at the the ◦ centercenter of of the the sub-retina. sub‐retina. The The sub-FoV sub‐FoV is is as as wide wide as as 60 60°.. TheThe subsystem subsystem takes takes advantage advantage of of three three surfaces surfaces out out of of a a same same volume, volume, thus thus can can be be manufacturedmanufactured out out of of a a unitary unitary piece. piece. For For this this prototype, prototype, the the material material used used for for the the design design is fusedis fused silica silica glass, glass, but but it can it alsocan also operate operate with with most most optical optical polymers. polymers. Compactness Compactness is then is improvedthen improved by truncating by truncating such optical such optical surfaces, surfaces, as seen as in seen Figure in Figure6. 6. ForFor the the FSFS (or BS) subsystem subsystem (Figure (Figure 6b),6b), the the same same rules rules were were applied. applied. However, However, the theimage image plane plane (half (half a sub a sub-retina,‐retina, the thesecond second sub sub-retina‐retina being being shared shared by the by theFS and FS and BS sub BS ‐ subsystems) is rotated by 30◦ with respect to the front surface because a different viewing systems) is rotated by 30° with respect to the front surface because a different viewing angle of scattering signatures is evaluated. angle of scattering signatures is evaluated. In AppendixA, we plot the modulation transfer function of a subsystem, which is In Appendix A, we plot the modulation transfer function of a subsystem, which is very useful to quantify the resolution of an optical system; it may however be difficult for a very useful to quantify the resolution of an optical system; it may however be difficult for non-specialist to visualize the effects of the optical system on an image. Thus, we perform image simulations from an object at infinity (angular object) by the three subsystems (see, Figure7).

Sensors 2021, 21, x FOR PEER REVIEW 8 of 23

Sensors 2021, 21, 3181 a non‐specialist to visualize the effects of the optical system on an image. Thus, we8 ofper 23‐ form image simulations from an object at infinity (angular object) by the three subsystems (see, Figure 7).

FigureFigure 7. ((a)) Example Example picture picture as as a a source source object object at at infinity. inﬁnity. (b ()b Simulated) Simulated images images obtained obtained with with the theS subsystem S subsystem on the on first the ﬁrstsub‐ sub-retinaretina and and(c) with (c) with the FS the and FS andBS subsystems BS subsystems sharing sharing the second the second sub‐ sub-retina.retina. (d) Schematic (d) Schematic top‐view top-view of the of three the three sub‐ subsystemssystems aligned aligned with with the the two two sub sub-retinae.‐retinae.

The object isis aa squaresquare angularangular exampleexample picturepicture sizedsized 60 60°◦ ×× 6060°◦ (see, Figure7 7a).a). We We compute the spatially variant PSF of the optical subsystem. Chromatic aberrations areare alsoalso taken into account as thethe PSF is calculated with RGB colors, thisthis informationinformation cancan bebe usefuluseful if one wantswants toto scatterscatter lightlight withwith differentdifferent wavelengths.wavelengths. The simulationsimulation imageimage isis finallyfinally obtained by convoluting thethe objectobject andand thethe PSF.PSF. Figure7 7bb shows shows the the image image simulation simulation obtained obtained with with the the S S optical optical subsystem subsystem and and Figure7 7cc with with both both the the FS FS and and BS BS subsystems. subsystems. As As expected, expected, the the FS FS and and BS BS images images are are shared fromfrom aa first first sub-retina, sub‐retina, and and rotated rotated by by± 3030°◦, as, as explained explained earlier. earlier. The The magnification magnifica‐ istion rightfully is rightfully adapted adapted to our to holed our holed retina retina and the and PSF the appears PSF appears good good enough enough for scattering for scat‐ signaturetering signature imaging. imaging. Figure7 7dd presents presents a a top-down top‐down view view of of the the geometry, geometry, showing showing the the relative relative position position and tilt of the subsystems with respect to the sub-retinae.sub‐retinae. For readability reasons, onlyonly thethe front surfaces,surfaces, internalinternal reflectorsreflectors and retinaeretinae areare drawn.drawn. The FS andand BSBS subsystemssubsystems areare symmetrical withwith respect respect to to the the separation separation plane plane of theof twothe two halves halves of the of second the second sub-retina. sub‐ retina. 2.6. Merging the Subsystems and Coupling with the Holed Retina 2.6. MergingAll three the optical Subsystems subsystems and Coupling (S, FS with and the BS) Holed are merged Retina into a unitary piece. Front surfaces are at an equal distance (Front Focal Length, FFL = 0.5 mm) from the center of the All three optical subsystems (S, FS and BS) are merged into a unitary piece. Front channel. The channel is designed to have a section similar to that of the holed retina. The surfaces are at an equal distance (Front Focal Length, FFL = 0.5 mm) from the center of the resulting 3D model is rendered in Figure8. It contains the merged subsystems around a Sensors 2021, 21, x FOR PEER channel.REVIEW The channel is designed to have a section similar to that of the holed retina. The9 of 23 vertical fluidic channel. We also show, in the figure, the arrangement of the optical piece in resulting 3D model is rendered in Figure 8. It contains the merged subsystems around a relation to the holed retina. vertical fluidic channel. We also show, in the figure, the arrangement of the optical piece in relation to the holed retina.

FigureFigure 8. 8. 3D3D view view of of the the monolithic monolithic optical optical system system aligned aligned with with the the holed holed retina. retina. We have designed a cloaker optimized for stray‐light management. In particular, it is built with stray light protection features than can fit within the BFL of the optical piece. Such a cloaker, which is presented in Figure 9, is printed with a SLA (Stereo‐Lithography Apparatus) 3D‐printer (Form3 by Formlabs) using adaptive voxel resolution (down to 25 μm) and made with a black photopolymer resin. It is sized to fit within the socket of the Creapyx motherboard. It contains housing for the optical piece and shows optimized ap‐ ertures in front of all three subsystems (S, FS and BS subsystems).

Figure 9. (a) 3D view of the optical piece with the cloaker and the holed retina mounted on ceramic. (b) Cross‐sectional view of the optical setup, showing the optimized apertures.

Note that the illumination system is not directly integrated with the optical setup. This choice is motivated by the fact that such an optical piece can be studied separately from the illumination system, and that consequently, the latter can be interchanged easily, within the framework of an iterative approach to development. Moreover, separating the illumination system from the optical piece allows for implementing elements that can pro‐ tect the detectors from stray light coming from the source. Thus, we chose to implement a 45° mirror (3 mm wide), which sits on the cloaker and is separated from the optical piece by a diaphragm printed with the cloaker (see Figure 9a). By doing so, one can insert an interchangeable illumination module, as long as it pro‐ vides a downward beam that recovers the pupil of the mirror. Preferably, such an illumi‐ nation system would be designed in a similar way to the miniature refractive‐refractive optics present in the optical piece itself, and would allow forming a beam over the fluidic channel from a bare laser diode (LD). In our case, for reasons of supply, we use a system consisting of a commercially available bare LD (637 nm, 5 mW), and aspherical lenses contained in a standard 0.5 inch optical tube. Further details on the illumination module will be provided in Appendix B.

Sensors 2021, 21, x FOR PEER REVIEW 9 of 23

Sensors 2021, 21, 3181 9 of 23

Figure 8. 3D view of the monolithic optical system aligned with the holed retina.

WeWe have have designed designed a a cloaker cloaker optimized optimized for for stray stray-light‐light management. InIn particular,particular, it it is isbuilt built withwith straystray light protection features than than can can fit ﬁt within within the the BFL BFL of of the the optical optical piece. piece. SuchSuch a acloaker, cloaker, which which is is presented presented in in Figure Figure 9,9 ,is is printed printed with with a a SLA SLA (Stereo (Stereo-Lithography‐Lithography Apparatus)Apparatus) 3D 3D-printer‐printer (Form3 (Form3 by by Formlabs) Formlabs) using using adaptive adaptive voxel voxel resolution resolution (down (down to 25 to 25 µm) and made with a black photopolymer resin. It is sized to ﬁt within the socket of μm) and made with a black photopolymer resin. It is sized to fit within the socket of the the Creapyx motherboard. It contains housing for the optical piece and shows optimized Creapyx motherboard. It contains housing for the optical piece and shows optimized ap‐ apertures in front of all three subsystems (S, FS and BS subsystems). ertures in front of all three subsystems (S, FS and BS subsystems).

FigureFigure 9. 9. (a()a 3D) 3D view view of of the the optical optical piece piece with with the the cloaker cloaker and and the the holed holed retina retina mounted mounted on on ceramic. ceramic. (b (b) )Cross Cross-sectional‐sectional viewview of of the the optical optical setup, setup, showing showing the the optimized optimized apertures. apertures.

NoteNote that that the the illumination illumination system system is isnot not directly directly integrated integrated with with the the optical optical setup. setup. ThisThis choice choice is is motivated motivated by by the the fact fact that that such such an an optical optical piece piece can can be be studied studied separately separately fromfrom the the illumination illumination system, system, and and that that consequently, consequently, the the latter latter can can be be interchanged interchanged easily, easily, withinwithin the the framework framework of of an an iterative iterative approach approach to to development. development. Moreover, Moreover, separating separating the the illuminationillumination system system from from the the optical optical piece piece allows allows for implementing for implementing elements elements that can that pro can‐ tectprotect the detectors the detectors from from stray stray light lightcoming coming from from the source. the source. Thus,Thus, we we chose chose to to implement implement a a45° 45 ◦mirrormirror (3 (3 mm mm wide), wide), which which sits sits on on the the cloaker cloaker and and isis separated separated from from the the optical optical piece by aa diaphragmdiaphragm printed printed with with the the cloaker cloaker (see (see Figure Figure9a). 9a).By By doing doing so, so, one one can can insert insert an an interchangeable interchangeable illumination illumination module, module, as longas long as itas provides it pro‐ videsa downward a downward beam beam that that recovers recovers the pupil the pupil of the of mirror.the mirror. Preferably, Preferably, such such an illumination an illumi‐ nationsystem system would would be designed be designed in a similar in a similar way toway the to miniature the miniature refractive-refractive refractive‐refractive optics opticspresent present in the in optical the optical piece piece itself, itself, and would and would allow formingallow forming a beam a overbeam the over ﬂuidic the fluidic channel channelfrom a from bare lasera bare diode laser (LD). diode In (LD). our case, In our for case, reasons for ofreasons supply, of we supply, use a we system use a consisting system consistingof a commercially of a commercially available bare available LD (637 bare nm, LD 5 mW),(637 nm, and 5 aspherical mW), and lenses aspherical contained lenses in a containedstandard in 0.5 a inch standard optical 0.5 tube. inch Further optical detailstube. Further on the illumination details on the module illumination will be module provided willin Appendixbe providedB. in Appendix B. Below the optical piece, the shell has optimized apertures that can select the scattered rays that were transformed by the optical system, as shown in Figure9b. These apertures are slanted with the angle deﬁned by the folding mirror.

2.7. Angular Response at the Coincidental Fourier Plane We will now evaluate the angular response of the merged system. In Figure 10, we present a RT simulation (Zemax Optic Studio, non-sequential mode, the non-sequential mode is the Monte–Carlo ray-tracing method where rays can interact with any surface of an arbitrary geometry) of the silica glass piece associated with the cloaker and its optimized apertures. For a better comprehension, we have simpliﬁed the scattering particle down to a simple conical source point, which has a given irradiance angle, and is axially symmetrical around the beam axis. An image monitor is placed at the coincidental Fourier plane, and can record the irradiance pattern induced by the scatterer. Sensors 2021, 21, x FOR PEER REVIEW 10 of 23 Sensors 2021, 21, x FOR PEER REVIEW 10 of 23

Below the optical piece, the shell has optimized apertures that can select the scattered Below the optical piece, the shell has optimized apertures that can select the scattered rays that were transformed by the optical system, as shown in Figure 9b. These apertures rays that were transformed by the optical system, as shown in Figure 9b. These apertures are slanted with the angle defined by the folding mirror. are slanted with the angle defined by the folding mirror.

2.7.2.7. AngularAngular ResponseResponse at at the the Coincidental Coincidental Fourier Fourier Plane Plane WeWe willwill nownow evaluateevaluate the the angular angular response response of of the the merged merged system. system. In In Figure Figure 10, 10, we we presentpresent aa RTRT simulationsimulation (Zemax (Zemax Optic Optic Studio, Studio, non non‐sequential‐sequential mode, mode, the the non non‐sequential‐sequential modemode isis thethe Monte–CarloMonte–Carlo ray ray‐tracing‐tracing method method where where rays rays can can interact interact with with any any surface surface of of anan arbitraryarbitrary geometry)geometry) ofof the the silica silica glass glass piece piece associated associated with with the the cloaker cloaker and and its its opti opti‐ ‐ mizedmized apertures.apertures. ForFor aa betterbetter comprehension, comprehension, we we have have simplified simplified the the scattering scattering particle particle Sensors 2021, 21, 3181 downdown toto aa simplesimple conicalconical source source point, point, which which has has a agiven given irradiance irradiance angle, angle, and and is isaxially axially10 of 23 symmetricalsymmetrical around around the the beam beam axis. axis. An An image image monitor monitor is is placed placed at at the the coincidental coincidental Fourier Fourier plane,plane, andand cancan recordrecord the the irradiance irradiance pattern pattern induced induced by by the the scatterer. scatterer.

FigureFigure 10.10. ((aa)) 3D 3D viewview of ofof the thethe RT RTRT simulation. simulation.simulation. ( (b(bb)) )Close Close-upClose‐up‐up cross cross-sectionalcross‐sectional‐sectional view. view.view.

◦ ◦ ByBy sweepingsweeping the the the angle angle angle of of ofthe the the conical conical conical source source source from from from 1° 1° to to 1180°, 180°,to one 180 one can, can one obtain obtain can the obtain the inten inten the‐ ‐ intensitysitysity mapmap maponon thethe on coincidentalcoincidental the coincidental Fourier Fourier Fourier plane plane plane for for each foreach each scattering scattering scattering angle angle angle and and and then, then, then, associate associate associate a a ascatteringscattering scattering angleangle angle (and(and (and intensity)intensity) intensity) to to toeach each each pixel pixel pixel of of of the the the monitor. monitor. monitor. In In In Figure Figure 11a, 1111a, a,we we plot plotplot the thethe angularangular response response of of the the optical optical system system that thatthat evaluates evaluates evaluates a a acentered centered centered scatterer scatterer scatterer 𝑆S 𝑆(0,0,00,0,0,0 0, 0.) .The. TheThe angleangle map map is is plotted plotted using usingusing a aa colored coloredcolored scale, scale,scale, and andand the thethe associated associatedassociated intensity intensityintensity is is isplotted plottedplotted using usingusing a aa transparent-blacktransparenttransparent‐‐blackblack (bright (bright-dark)(bright‐‐dark)dark) overlay. overlay.overlay.

Figure 11. Angular response (in degrees) of the optical system evaluating (a) a centered scatterer and (b) a displaced FigureFigurescatterer. 11.11. AngularAngular response (in degrees)degrees) ofof thethe opticaloptical systemsystem evaluatingevaluating ((aa)) aa centeredcentered scattererscatterer andand ((bb)) aa displaceddisplaced scatterer.scatterer.

We have drawn white rectangles that correspond to the two sub-retinae. We recognize the three regions of interest (ROIs) associated with the three optical subsystems (FS, S, BS). We verify that a point within a ROI is associated to a single scattering angle. These ROIs allow us to evaluate the scattering signature thanks to the associated overlapping FoVs. Again, we verify that the S-ROI falls on a first sub-retina and that the FS and BS ROIs share the second sub-retina. The total FoV is as wide as 110◦. In order to evaluate the impact of the position of the scatterer, we perform the same RT simulation but the scatterer is displaced at S xp, yp, zp , using reasonable positions when compared with the incident beam (xp = 100 µm, yp = 15 µm, zp = 10 µm). Note that the range of accessible positions is highly elongated along the axis of the illumination beam (X-axis). The result of such a simulation is presented in Figure 11b. We verify that every pixel is still associated with the same scattering angle when compared to the centered scatterer case (pixels are colored the same in both subfigures). This gives us confidence that our strategy of measuring a scattering signature with position insensitivity is relevant. Sensors 2021, 21, 3181 11 of 23

The main difference between the two images is that the images of the pupils of each subsystem are slightly transformed due to the displacement of the scatterer, which was to be expected. This should not present many difﬁculties as it takes effect at the edge of every subsystem (blind ﬁelds), which can be easily corrected thanks to our asymmetric assembly of subsystems

3. Fabrication Process 3.1. Micro-Machining on Glass The prototype of the monolithic optical piece is a challenging piece to manufacture. It has small dimensions and the surfaces of the diopters must have optical-grade surface qualities. Precision micro-tooling of molds [56] is still very expensive but can be cost- efficient if optical pieces are molded at volume fabrication (using for instance micro- injection molding [57]). Direct laser 3D printing by two-photon polymerization shows great promise printing miniature optics [58,59] but is not yet suited for printing millimeter- sized optical pieces: the lateral dimensions are limited to a few hundred micrometers. For our prototype, we have selected a manufacturing process based on an innovative laser micro-machining of glass [60,61] developed by FEMTOprint that relies upon three main steps: laser exposure, wet/chemical etching and polishing. The outline of the 3D shape of the device inside a bulk piece of fused silica is generated by sweeping a tightly focused laser beam at a wavelength of 1030 nm. The exposure to femtosecond laser pulses triggers a non-linear absorption process that causes a glass densification within the laser voxel (the laser voxel is the volume at the vicinity of the focusing point where the optical power density has reached a certain value, allowing for the non-linear densification effect) that eventually induces a drastic change in selectivity (up to 1000 times) when immersed in an etching solution [62]. The excess material is then removed by a KOH-based solution and the unexposed volume of the device is released from the wafer. Machined surfaces show a typical roughness Ra (Ra stands for Roughness average, and is the average value of the profile heigh deviations) larger than approx. 100 nm, i.e., not low enough for optical elements such as the channel-facing front diopters, the reflecting surfaces, and the back diopters. For this reason, a proprietary polishing process has been applied to selectively reduce the roughness of the optical elements by locally reflowing the material surfaces while minimizing the deformation. To this goal, the slanted freeform surfaces have been designed to be easily accessible to the polishing tool and a reduction of Ra to values between 5 and 15 nm has been achieved, which is much better that the roughness initially targeted (λ/10), and allows for minimal surface scattering. There was no need to polish the flat top and bottom surfaces, as these were not machined and therefore maintained a pristine surface quality. An optical image showing a top view of the fabricated glass piece is shown in Figure 12. Figure 12c shows an optical view of the polished glass piece. Note that the rough surfaces (machined surfaces without polishing) appear bright because they scatter light, whereas the polished and pristine surfaces (e.g., the slanted freeform surfaces) appear dark, which testifies to an excellent surface quality. As a comparison, Figure 12d is a top-down view of the original drawing, which shows the great fidelity of the fabricated piece compared with the model. Sensors 2021, 21, x FOR PEER REVIEW 12 of 23

Sensors 2021, 21, 3181 12 of 23 An optical image showing a top view of the fabricated glass piece is shown in Figure 12.

Figure 12. (a) Photography of the fabricated optical system next to a one one-euro‐euro cent coin. ( (bb)) 3D 3D view view of of the the optical optical system. system. (c) Optical view of the fabricated glass piece. (d) Associated view of the original drawing. (c) Optical view of the fabricated glass piece. (d) Associated view of the original drawing.

Figure 12c shows an optical view of the polished glass piece. Note that the rough Sensors 2021, 21, x FOR PEER REVIEWNote that a molding and replication process was explored for a previous design 13 of 23 surfacesof this optical (machined piece surfaces in reference without [54]. polishing) The use ofappear molded bright polymer because pieces they wouldscatter allowlight, whereasformanufacturing the polished the sensorand pristine inexpensively, surfaces at (e.g., volume the fabrication.slanted freeform surfaces) appear dark, which testifies to an excellent surface quality. As a comparison, Figure 12d is a top‐ down3.2. Mirror view3.2. Deposition of theMirror original Deposition drawing, which shows the great fidelity of the fabricated piece comparedIn this with section, Inthe this model. we section, describe we how describe the mirror how the deposition mirror deposition step is performed. step is performed. A metallic A metallic layerNote has tothatlayer coat a hasmolding at leastto coat the and at slanted leastreplication the freeform slanted process surfaces, freeform was explored while surfaces, avoiding for while a previous the avoiding first design diopters the firstof diopters thisfacing optical thefacing fluidicpiece thein channel. referencefluidic channel. Thus, [54]. The a Thus, specific use a ofspecific area molded must area polymer bemust covered be pieces covered prior would prior to anyallow to any metal for metal‐ depo‐ manufacturingdepositionsition step: the astep: stripe sensor a abovestripe inexpensively, theabove fluidic the channel.atfluidic volume channel. The fabrication. remaining The remaining surfaces thatsurfaces do not that have do not have optical functionalityoptical functionality can be metallized can be metallized without affecting without the affecting operation the ofoperation the system. of the We system. We cover thosecover surfaces those using surfaces a stencil using made a stencil from made a 1-mm-wide from a 1‐ stripemm‐wide of Kapton stripe of (polyimide Kapton (polyimide adhesive film)adhesive as seen film) in as Figure seen in 13 Figurea. In 13 a preliminarya. In a preliminary process, process, we cut we the cut said the said Kapton Kapton stripe stripe manuallymanually with with a blade a blade and and a pair a pair of binoculars. of binoculars. We position,We position, again again manually, manually, the the stripe stripe on theon opticalthe optical piece piece using using a pair a pair of tweezers. of tweezers. In a In future a future process, process, the the fabrication fabrication of of such a such a stencilstencil should should be be computer-assisted; computer‐assisted; for for instance, instance, via via CNC CNC cutting, cutting, laser laser ablation ablation or lithog‐ or lithography.raphy.

Figure 13. (a) PhotographyFigure 13. ( aof) Photography the optical piece of the with optical the Kapton piece with stencil. the Kapton(b) Optical stencil. piece (b after) Optical the metallization piece after the process and stencil removal.metallization (c) Optical process top‐view and of stencil the final removal. optical (c piece.) Optical top-view of the ﬁnal optical piece.

The metallic layer is deposed by physical vapor deposition (PVD): first with a 10 nm titanium adhesion layer, followed by a 100 nm gold layer. Preliminary studies have shown

that the silica/metal interface is well preserved, in terms of surface quality, which justifies the use of such process to manufacture internal reflectors. The use of gold as the coating material is motivated by its reflectivity on a range from red to NIR (>90%), which is weakly impacted by the titanium layer. After this step, the parts that were not to be metallized are revealed, by simply removing the Kapton stencil with tweezers (see Figure 13b). The final optical piece, with its selectively deposited mirrors is presented in Figure 13c. In the future, a specific mirror deposition process will have to be developed in order to be compatible with our polymer pieces.

4. Characterization 4.1. Experimental Setup The characterizations of the sensor are performed in the same fashion as for the lens‐ less setup [31], by using the Creapyx motherboard that drives the holed retina, and a cal‐ ibrated aerosol test bench (Constant output atomizer model 3076 by TSI) that provides a continuous flow of calibrated particles, which are commercially available monodisperse PSL beads in our case. A photography of the characterization setup is given in Figure 14a. The optical piece is sitting on the cloaker that properly aligns it with the holed retina below. As explained earlier, the downward illumination is provided by the illumination module (standard 0.5 inch lens tube, with a LD and aspheric lenses, see Appendix B), and redirected horizon‐ tally toward the muzzle of the air channel using a 45° mirror.

Sensors 2021, 21, 3181 13 of 23

The metallic layer is deposed by physical vapor deposition (PVD): first with a 10 nm titanium adhesion layer, followed by a 100 nm gold layer. Preliminary studies have shown that the silica/metal interface is well preserved, in terms of surface quality, which justifies the use of such process to manufacture internal reflectors. The use of gold as the coating material is motivated by its reflectivity on a range from red to NIR (>90%), which is weakly impacted by the titanium layer. After this step, the parts that were not to be metallized are revealed, by simply removing the Kapton stencil with tweezers (see Figure 13b). The final optical piece, with its selectively deposited mirrors is presented in Figure 13c. In the future, a specific mirror deposition process will have to be developed in order to be compatible with our polymer pieces.

4. Characterization 4.1. Experimental Setup The characterizations of the sensor are performed in the same fashion as for the lens- less setup [31], by using the Creapyx motherboard that drives the holed retina, and a calibrated aerosol test bench (Constant output atomizer model 3076 by TSI) that provides a continuous ﬂow of calibrated particles, which are commercially available monodisperse PSL beads in our case. A photography of the characterization setup is given in Figure 14a. The optical piece is sitting on the cloaker that properly aligns it with the holed retina below. As explained earlier, the downward illumination is provided by the illumination module (standard 0.5 Sensors 2021, 21, x FOR PEER REVIEWinch lens tube, with a LD and aspheric lenses, see AppendixB), and redirected horizontally14 of 23

toward the muzzle of the air channel using a 45◦ mirror.

FigureFigure 14. 14.( (aa)) Photography Photography of of the the characterization characterization setup. setup. ( (bb)) Reference Reference image image without without particle. particle.

InIn thethe photographyphotography wewe seesee that, that, despite despite the the diaphragm, diaphragm, a a large large portion portion of of stray stray light light comescomes towardtoward thethe opticaloptical piece,piece, hittinghitting somesome ofof its its diopters. diopters. Then,Then, wewe taketake aa referencereference image,image, using using the the holed holed retina retina when when no no aerosol aerosol is is injected injected in in the the channel. channel. Such Such a reference a reference is presentedis presented in in Figure Figure 14 14b,b, and and is is automatically automatically subtracted subtracted from from future future images. images. We We notice notice thatthat a a portionportion of of thethe secondsecond sub-retina sub‐retina is is impacted impacted by by stray stray light light (FS (FS and and BS BS ROIs), ROIs), but but not not toto aa greatgreat extent. extent. GivenGiven thethe visiblevisible amountamount ofof straystray lightlight hittinghitting thethe optical optical piece, piece, and and the the reasonable reasonable amountamount thatthat endsends upup onon thethe retina,retina, one one mightmight argue argue that that the the anti-stray-light anti‐stray‐light designs designs were were successful,successful, atat leastleast for for our our given given illumination illumination setup. setup. WeWe record record bursts bursts of of 32 32 images, images, in in global global shutter shutter (GS) (GS) mode, mode, and and we we follow follow the the standard stand‐ deviationard deviation of the of full the image full image (with (with its reference its reference subtracted). subtracted). Then, Then, an image an image is saved is whensaved itswhen standard its standard deviation deviation reaches reaches a given a threshold. given threshold. More information More information on the experimentalon the experi‐ setup,mental notably setup, innotably terms ofin theterms sampled of the aerosol sampled and aerosol the driving and the of thedriving retina, of is the reported retina, in is referencereported [in31 reference]. [31].

4.2. Experimental Images of Scattering Signatures We were able to detect single PSL particles of 3 μm and 830 nm. Examples of repre‐ sentative signatures that can be experimentally recorded are given in Figure 15a,b. The images obtained seams quite repeatable for a given aerosol, both in terms of brightness and patterns. A precise study on repeatability should be conducted in the future (in order to quantify the deviation), but this can already give us confidence that the signature is weakly impacted by particle position, due to the Fourier‐domain imaging setup.

Sensors 2021, 21, x FOR PEER REVIEW 14 of 23

Figure 14. (a) Photography of the characterization setup. (b) Reference image without particle.

In the photography we see that, despite the diaphragm, a large portion of stray light comes toward the optical piece, hitting some of its diopters. Then, we take a reference image, using the holed retina when no aerosol is injected in the channel. Such a reference is presented in Figure 14b, and is automatically subtracted from future images. We notice that a portion of the second sub‐retina is impacted by stray light (FS and BS ROIs), but not to a great extent. Given the visible amount of stray light hitting the optical piece, and the reasonable amount that ends up on the retina, one might argue that the anti‐stray‐light designs were successful, at least for our given illumination setup. We record bursts of 32 images, in global shutter (GS) mode, and we follow the stand‐ ard deviation of the full image (with its reference subtracted). Then, an image is saved Sensors 2021, 21, 3181 when its standard deviation reaches a given threshold. More information on the experi14 of 23‐ mental setup, notably in terms of the sampled aerosol and the driving of the retina, is reported in reference [31].

4.2. Experimental Images of Scattering Signatures We were able to detect single PSL particles of 3 μµm and 830 nm. Examples of of repre repre-‐ sentative signatures that can be experimentally recorded are given inin FigureFigure 1515a,b.a,b. The images obtainedobtained seamsseams quite quite repeatable repeatable for for a givena given aerosol, aerosol, both both in terms in terms of brightness of brightness and andpatterns. patterns. A precise A precise study study on repeatability on repeatability should should be conductedbe conducted in thein the future future (in (in order order to toquantify quantify the the deviation), deviation), but thisbut canthis alreadycan already give usgive conﬁdence us confidence that the that signature the signature is weakly is weaklyimpacted impacted by particle by particle position, position, due to the due Fourier-domain to the Fourier‐domain imaging imaging setup. setup.

Figure 15. Experimental images of scattering signatures of PSL spheres: (a) 3 µm and (b) 830 nm; compared with associated computed signatures: (c,d) respectively.

We observe three separated areas that correspond to the S-ROI on the ﬁrst sub-retina (rows 76 to 150); the BS and FS-ROIs on the second sub-retina (rows 1 to 75). We notice that the positions of such ROIs are slightly translated compared to what was expected, due to the small alignment problems that will have to be addressed in the future. The white pixels correspond to pixels from the reference image that reached a certain threshold, that were removed from the image processing, in order to be less sensitive to Poissonian-type noises. The scattering signatures of large spheres, such as the 3 µm PSL, exhibit a number of Mie’s oscillations, whereas the 830 nm PSL does not (at least within the evaluated FoV). This effect is predicted by the Lorenz–Mie theory (spheres with refractive index nPSL = 1.5875 [53]), and be can be found on the RT simulations for the associated PSL diameters (see, subﬁgures (c,d)). Note that the images were normalized, but the same color axis was kept between (a,b) on the one hand and (c,d) on the other hand, in order to compare the relative brightness of scattering signatures obtained with different diameters.

4.3. Reconstruction of Scattering Signatures The oscillations of 3 µm PSL signatures facilitate the interpretation of such a recorded signature. For didactic reasons, we are going to focus in particular on the specific image shown in Figure 15a. We observe several vertical oscillations on the S-ROI, whereas those on the BS and FS ROIs are slanted by ±30◦, which was expected by design. Those oscillations follow what we call the iso-θ, which are defined as the trajectories on the retina that correspond to the rays scattered with the same scattering angle θ. The procedure to reconstruct a signature is relatively simple: let us first consider one of the three ROI separately. First, every pixel in the ROI is normalized using the simulated intensity map calculated previously using a RT simulation (intensity map in Figure 11a). Sensors 2021, 21, 3181 15 of 23

Then, by using an angle map, also computed by RT (see, Figure 11a), one can associate each pixel of the retina to a scattering angle. For a given scattering angle (with a tolerance of 1◦), a mean value of the intensity carried by those associated pixels can be calculated which then results in a sub-scattering signature reconstructed within the sub-FoV of each ROIs. In this case, the ROIs are misaligned with respect to the retina, thus, we had to manually translate the angle map to best fit with the recorded image. Due to this effect, the BS ROI was cropped, resulting in a loss of FoV for very high angles. Still, the total FoV is relatively wide, at about 85◦. Before plotting those sub-signatures together, we had to perform a correction that consists in having a multiplicative factor for each sub-signature. This step is motivated by the fact that the relative brightness of the ROIs are not consistent to what was expected: this can be seen especially with the BS-ROI (see Figure 15a) that is much brighter than expected (see Figure 15c). Hopefully, this correction appears to be very repeatable between the different recorded images, and could be easily found through a calibration campaign. We believe that it could be attributed to misalignment errors of the apertures of the cloaker with respect to the subsystems’ pupils, which may create an uneven loss of photometry between the ROIs. One notices that the sub-signatures are not perfectly overlapping: the edges of the individual curves correspond to the blind field and should not be over-trusted. Again, this overlapping mismatch will not overly affect results as it appears to be very repeatable, and thus easy to correct in post-process; for example, with a single calibration. To connect the Sensors 2021, 21, x FOR PEER REVIEW 16 of 23 curves, we make averages on the overlapping regions. The averaging weight varies linearly from 1 to 0, dropping at the edge of each curve. With this step, one can have a continuous merged curve (drawn in black), which is more representative of real scattering signatures. We insist on the fact that the reconstruction step is particularly resource-efficientresource‐efficient in termsterms of computing computing capabilities, especially when applying only matrix operations (which are optimized by MATLAB, our data processing tool), at least compared to the planar projection set-upset‐up (see,(see, reference reference [ 31[31]).]). This This gives gives us us confidence confidence that that procedure procedure is compatibleis compat‐ iblewith with low-energy low‐energy portable portable applications. applications. InIn Figure 1616,, wewe plotplot thethe mergedmerged scattering scattering signatures signatures alongside alongside the the theoretical theoretical signa- sig‐ ture of a 3 µm PSL sphere (refractive index n = 1.5875 [53]) illuminated by a 637 nm nature of a 3 μm PSL sphere (refractive indexPSL 𝑛 1.5875 [53]) illuminated by a 637 nmwavelength wavelength beam, beam, computed computed using using the Lorenz–Miethe Lorenz–Mie theory. theory.

Figure 16. (a) Reconstructed signature from experimental image plotted with the theoreticaltheoretical signature.signature. (b) Experimental image of a 3 μµm PSL used for the reconstruction.

Based on on such a a representative plot, plot, we we find ﬁnd that that the the reconstructed reconstructed signature signature is is sur sur-‐ prisingly faithful to the theoretical one, especially with the position and frequency of the Mie’s oscillations in the plot (which were not corrected). This gives us confidence that such an architecture of the sensor and reconstruction procedure is well suited for our application. However, to conclude on this point, we will have to conduct a proper statistical study, over a wide number of PSL diameters, with blind reconstructions. To do so, we will have to make our setup more robust. In the near future, the 3D printing of the cloaker will be optimized; then, the next step would be to integrate directly both the cloaker (with its optimized apertures) and the optical piece on‐ top the CMOS at wafer scale, in order to achieve optimal alignments. Note that a study on a larger number of particles, and on a larger diameter range, could allow us to conclude on the detection limit in terms of particle size. We can already determine with which SNR (Signal to Noise Ratio) we have detected 3 μm and 830 nm PSL. The PSL of 3 μm were detected with an SNR of about 825, while the PSL of 830 nm (which scatter less intensely) were detected with an SNR of about 145 (these values were quite constant between images obtained from the same PSL diameter, meaning that the brightness and patterns of the recorded signatures are very repeatable). Those values are quite high already, and thus we can realistically expect that we can measure smaller particle sizes using such a setup. In addition, the SNR should be further improved by optimizing the illumination system so that the amount of stray light is min‐ imized.

5. Discussion about Particle Identification We recall that the principle of measuring a scattering signature is to find certain pa‐ rameters of the particles that compose an aerosol, such as diameter 𝐷 and refractive in‐ dex 𝑛. We compare the image recorded experimentally, with a prediction from a model (Lorenz–Mie theory), and a geometrical transformation in order to perform the inverse

Sensors 2021, 21, 3181 16 of 23

This gives us conﬁdence that such an architecture of the sensor and reconstruction procedure is well suited for our application. However, to conclude on this point, we will have to conduct a proper statistical study, over a wide number of PSL diameters, with blind reconstructions. To do so, we will have to make our setup more robust. In the near future, the 3D printing of the cloaker will be optimized; then, the next step would be to integrate directly both the cloaker (with its optimized apertures) and the optical piece on-top the CMOS at wafer scale, in order to achieve optimal alignments. Note that a study on a larger number of particles, and on a larger diameter range, could allow us to conclude on the detection limit in terms of particle size. We can already determine with which SNR (Signal to Noise Ratio) we have detected 3 µm and 830 nm PSL. The PSL of 3 µm were detected with an SNR of about 825, while the PSL of 830 nm (which scatter less intensely) were detected with an SNR of about 145 (these values were quite constant between images obtained from the same PSL diameter, meaning that the brightness and patterns of the recorded signatures are very repeatable). Those values are quite high already, and thus we can realistically expect that we can measure smaller particle sizes using such a setup. In addition, the SNR should be further improved by optimizing the illumination system so that the amount of stray light is minimized.

5. Discussion about Particle Identiﬁcation

Sensors 2021, 21, x FOR PEER REVIEW We recall that the principle of measuring a scattering signature is to ﬁnd certain17 of pa- 23

rameters of the particles that compose an aerosol, such as diameter Dp and refractive index n. We compare the image recorded experimentally, with a prediction from a model (Lorenz– Mie theory), and a geometrical transformation in order to perform the inverse problem. problem. To do so, we define the minimization criterion R 𝜖𝐷 , 𝑛 ∝ 𝑆 𝜃 To do so, we deﬁne the minimization criterion e Dp, n ∝ θ∈FoV Sexp(θ) −∈S θ, Dp, n dθ, which𝑆𝜃, 𝐷 evaluates, 𝑛𝑑𝜃, which the deviation evaluates between the deviation an experimental between an signature experimentalSexp(θ signature) and a modelled 𝑆𝜃 oneand Sa modelledθ, Dp, n . Here,one 𝑆𝜃 we, 𝐷 only, 𝑛 study. Here, the we case only of study a non-absorbing the case of particle,a non‐absorbing thus we consider particle, thethus refractive we consider index the as refractive a real quantity. index as In a the real case quantity. of an absorbing In the case particle of an absorbing (such as a particle carbon particle),(such as a the carbon imaginary particle), part theIm (imaginaryn) must be part added 𝐼𝑚 to𝑛 the must list ofbe parameters added to the to belist optimized, of param‐ alongeters to with be optimized,Dp and Re( alongn). with 𝐷 and 𝑅𝑒𝑛. The optimizationoptimization procedure, procedure, which which was was shown shown in in Figure Figure5, is5, used. is used. Here, Here, we we are are going go‐ to study its accuracy, in its capacity to ﬁnd the correct values of D and n. For didactic ing to study its accuracy, in its capacity to find the correct values of 𝐷p and 𝑛. For didactic reasons,reasons, we are onlyonly consideringconsidering the case of thethe 33 μµmm PSL.PSL. TheThe testtest imagesimages areare foundfound inin Figure 15a. The minimization criterion is plotted in Figure 17 as a function of Dp and n. Figure 15a. The minimization criterion is plotted in Figure 17 as a function of 𝐷 and 𝑛.

Figure 17. MapMap of of our our minimization minimization criterion criterion versus versus particle particle diameter diameter and andrefractive refractive index index in loga in‐ logarithmicrithmic scale. scale.

Sensors 2021, 21, 3181 17 of 23

We draw the targeted parameters as black lines, and the reasonable local minima as a white contour. The accuracy of the procedure is judged by the extent of such a contour, and its distance from the targeted parameters. In addition, we are also interested in local minima, i.e., their number and prominence, for the reason that they can induce important errors in the estimation of the desired parameters. We have already discussed that the procedure used in the lens-less setup was less than optimal in view of these different judgment criteria (see reference [31]). In comparison, the procedure for the Fourier-domain setup appears more successful, at least for the test case of a 3 µm PSL. We judge, however, that such a procedure is to be improved, not least because the precision is still insufficient, in particular by the fact that the valley of the local minimum is very elongated in one direction, and by the presence of another local minimum. The direction of elongation of the local minima seems to come from the fact that similar signatures can be obtained for different couples (Dp, n), and that a simple measurement of the scattering signature only allows to find a kind of optical diameter (that combines both Dp and n). We believe that it is possible to solve this limitation by simultaneously evaluating the optical diameter of the same particle (of fixed geometric diameter Dp) in two different ways, for example by recording a dichromatic scattering signature. If one knows the type of particle evaluated and its refractive index, the accuracy on the geometric diameter is given by the horizontal width of the minimum in the plot, which results in an error of about 10% of the value obtained. It is likely that the accuracy is dependent on the diameter. In future work, it will be necessary to perform this accuracy study for different particle diameters. In particular, this will be necessary for small particles, such as those of 830 nm, in order to know the result of the method when no Mie oscillation is found in the signature. Then, another way to improve this procedure is to make the assembly more robust, as explained previously. Other elements could also help to reduce the uncertainty in the retrieval for parameters Dp and n, such as, for example, using the overall image brightness as a starting point for the optimization. Again, the simple use of a 3 µm PSL example is far from sufficient to conclude on the accuracy of the procedure. To do so, we will have to perform a proper statistical study on a large number of particles, and over a wide range of diameters and types of particles (organic, inorganic, aqueous, and irregular).

6. Conclusions In conclusion, we have designed a PM sensor that features a miniature, monolithic refracto-reflective system suited for the optical pre-treatment of scattering signatures. This design is applied to our family of PM sensor prototypes that uses holed retinae. It addresses the various areas of improvements for miniature PM sensors that were identified in a previous lens-less setup [31], such as the ability to collect more flux from the scatterer (with viewing angles where scattering is usually strong), particle’s position insensitivity, ultra-wide field of view, and simplified image processing. It features a monolithic assembly of three compact optical subsystems, arranged around an air channel, that are folded onto a coincidental focal plane. The optical piece includes a free-form surface and shows greater performances in all fields. In particular, it features elements that prevent stray- light sensitivity and is fully compatible with the holed retina: a custom fluidic-optronic CMOS chip. We present a state-of-the-art fabrication process for a glass prototype that was developed by FEMTOprint, which consists of the direct, 3D writing on glass, using a femtosecond laser inscription and selective etching and local surface polishing, followed by a selective mirror deposition process. The glass piece was assembled with an optimized cloaker and the holed retina, and tested with calibrated aerosols. We were able to measure the scattering signature (optically processed) of single particles of polystyrene beads of different sizes. We were then able to test the reconstruction procedure of the scattering signature, which is particularly faithful to Sensors 2021, 21, 3181 18 of 23

a theoretical signature. This gave promising results in solving the inverse problem, which consists of ﬁnding the size and refractive index of a particle using the Lorenz–Mie theory. However, in order to conclude on the performance of such a system, we will have to carry out more in-depth tests (on an assembly that needs to be made more robust). For example, it would be interesting to record signatures from a larger quantity of particles in order to perform a statistical study. Such a study would allow us, for example, to relate the detection rate to a particle concentration. We would also test our device on calibrated particles of various sizes, which would allow us to experimentally estimate a Limit of Detection (LoD) in size. Then, in order to be more representative of a real aerosol, we would perform those tests on a larger variety of particle types (such as organic, inorganic, aqueous, irregular particles etc.).

7. Patents Some of the designs reported in this article are ﬁled under the following patents: • Boutami, S. and Nicoletti, S., Optical detector of particles, 2018, Patent Application EP3574301, WO2018138223, FR3062209, KR20190112049 • Jobert, G., Optical particle detector, 2018, Patent Application EP3598102, US20200033246, FR3083864.

Author Contributions: Conceptualization: G.J.; Methodology: G.J., P.B. (Pierre Barritault), C.M. and M.F.; Software: G.J. and P.B. (Pierre Barritault); Validation: P.B. (Pierre Barritault), M.F., S.B., C.J. and C.S.; Formal analysis: G.J.; Investigation: G.J., P.B. (Pierre Barritault) and M.F.; Resources: M.F., P.B. (Pietro Bernasconi), A.L., D.B.; Data curation: P.B. (Pierre Barritault), M.F.; Writing—original draft preparation: G.J.; Writing—review and editing: S.B., P.B. (Pierre Barritault), P.B. (Pietro Bernasconi); Visualization: G.J.; Supervision: P.B. (Pierre Barritault), M.F., S.B., C.J. and C.S.; Project administration: M.F. All authors have read and agreed to the published version of the manuscript. Funding: This research was funded by the Department of Optics and Photonics of the CEA-LETI (France) and received no external funding. Institutional Review Board Statement: Not applicable. Informed Consent Statement: Written informed consent has been obtained from the patient(s) to publish this paper. Acknowledgments: The authors are thankful toward the fabrication team of CEA-LETI for their help concerning the development and post-process of the holed retina. Lara Boutafa for her work on the packaging. The Pyxalis team for the development of the holed retina, in particular to Benoit Dupont for his involvement and support toward the project. The FEMTOprint team for the fabrication of the miniature optical glass system, in particular to Giulia Bottarini who has been a privileged interlocutor throughout the collaboration. Sylvain Gout for his work on the mirror deposition process. Brice Poirier from CEA’s prototyping workshop for his assistance on the 3D printing of the cloaker. Arnaud Guiot from the CEA’s Nano-Safety Platform for his assistance on experiments with laboratory aerosols. Sergio Nicoletti and Laurent Duraffourg for supporting the topic within our laboratory. Finally, we would like to acknowledge Marine Garcia, who gave the permission to use a picture of her in order to carry out our image simulations. Conﬂicts of Interest: The authors declare no conﬂict of interest.

Appendix A In this appendix, we analyze the performances of an optical system. To do so, we commonly plot the Modulation Transfer Function (MTF) [53,63], which describes the loss of contrast created by the optical system as a function of the spatial frequency (usually given in cycles/mm). Here, we compute the PSF using Huygens’ wavelet direct integration algorithm. Then, the MTF is obtained by taking the amplitude of the Fast Fourier Transform (FFT) of Huygens’ PSF. We present in Figure A1, Huygens’ MTF of the optical system. Sensors 2021, 21, x FOR PEER REVIEW 19 of 23

Author Contributions: Conceptualization: G.J.; Methodology: G.J., P.B. (Pierre Barritault), C.M. and M.F.; Software: G.J. and P.B. (Pierre Barritault); Validation: P.B. (Pierre Barritault), M.F., S.B., C.J. and C.S.; Formal analysis: G.J.; Investigation: G.J., P.B. (Pierre Barritault) and M.F.; Resources: M.F., P.B. (Pietro Bernasconi), A.L., D.B.; Data curation: P.B. (Pierre Barritault), M.F.; Writing—original draft preparation: G.J.; Writing—review and editing: S.B., P.B. (Pierre Barritault), P.B. (Pietro Ber‐ nasconi); Visualization: G.J.; Supervision: P.B. (Pierre Barritault), M.F., S.B., C.J. and C.S.; Project administration: M.F. All authors have read and agreed to the published version of the manuscript. Funding: This research was funded by the Department of Optics and Photonics of the CEA‐LETI (France) and received no external funding. Institutional Review Board Statement: Not applicable Informed Consent Statement: Written informed consent has been obtained from the patient(s) to publish this paper Acknowledgments: The authors are thankful toward the fabrication team of CEA‐LETI for their help concerning the development and post‐process of the holed retina. Lara Boutafa for her work on the packaging. The Pyxalis team for the development of the holed retina, in particular to Benoit Dupont for his involvement and support toward the project. The FEMTOprint team for the fabrica‐ tion of the miniature optical glass system, in particular to Giulia Bottarini who has been a privileged interlocutor throughout the collaboration. Sylvain Gout for his work on the mirror deposition pro‐ cess. Brice Poirier from CEA’s prototyping workshop for his assistance on the 3D printing of the cloaker. Arnaud Guiot from the CEA’s Nano‐Safety Platform for his assistance on experiments with laboratory aerosols. Sergio Nicoletti and Laurent Duraffourg for supporting the topic within our laboratory. Finally, we would like to acknowledge Marine Garcia, who gave the permission to use a picture of her in order to carry out our image simulations. Conflicts of Interest: The authors declare no conflict of interest.

Appendix A In this appendix, we analyze the performances of an optical system. To do so, we commonly plot the Modulation Transfer Function (MTF) [53,63], which describes the loss of contrast created by the optical system as a function of the spatial frequency (usually given in cycles/mm). Here, we compute the PSF using Huygens’ wavelet direct integration Sensors 2021, 21, 3181 algorithm. Then, the MTF is obtained by taking the amplitude of the Fast Fourier19 of Trans 23 ‐ form (FFT) of Huygens’ PSF. We present in Figure A1, Huygens’ MTF of the optical sys‐ tem.

◦ ◦ FigureFigure A1. A1.Tangential Tangential and and Sagittal Sagittal MTF MTF of of the the S S subsystem subsystem at at 0 0°and and 25 25°..

We seeWe insee the in the plot plot that that the the MTF MTF curves curves with with the the free-formfree‐form reflector reflector seem seem almost almost iden‐ identical. The likeness of the sagittal and tangential MTF curves suggests that we have a tical. The likeness of the sagittal and tangential MTF curves suggests◦ that we have a nice nicecircular blur.blur. ThisThis is is maintained maintained both both with with an an on-axis on‐axis and and with with a 25 a 25°inclination, inclination, which which testifies for a uniform blur spot throughout the surface of the retina. Note that, without the testifies for a uniform blur spot throughout the surface of the retina. Note that, without free-form reflector, our subsystem would have suffered from two main types of aberrations: strong astigmatism and field curvature (not shown here). The resolution of the subsystem has to be compared with the resolution of the retina. To do so, we introduce the Nyquist frequency fN = 1/2ppix, which quantifies the geometrical sampling of the image plane, through the pixel’s pitch, denoted ppix. For a 10 µm pixel, like the holed retina, fN = 50 cycles/mm. Here, the contrast is completely lost at about 30 cycles/mm. It means that our blur spot is a bit larger than a pixel, which is quite advantageous for our application. Indeed, we don’t need to record scattering signatures with a pixel’s resolution, we even blur the image in post-process (Gaussian blur, σ = 2 pixels) to attenuate noise and other artefacts.

Appendix B In this appendix, we will describe the downward illumination module. Its main goal is to provide a focused beam just above the muzzle of the air channel, using a 45◦ mirror. Such an illumination system is realized by assembling commercially available optical elements such as a Laser Diode (LD) that provides 5 mW of optical power at a wavelength of 637 nm (Thorlabs, L637P5); a collimation aspheric lens L1 (Thorlabs, C330TMD-B, f1 = 3.1 mm); and a focusing aspheric lens L2 (Thorlabs, C280TMD-B, f1 = 18.4 mm). All those elements are contained within a standard lens tube for 0.5 inch optics (Thorlabs, SM05M10), that has an outer diameter of 17.78 mm. This module is presented in Figure A2. The L1,2 lens system is designed to properly form a focused beam above the ﬂuidic channel from the bare LD. First, the L1 collimation asphere is chosen with a high numerical aperture, in order to capture and collimate most of the radiation from the source; then, L2 focuses the beam in the aerosol sampling zone. Sensors 2021, 21, x FOR PEER REVIEW 20 of 23

the free‐form reflector, our subsystem would have suffered from two main types of aber‐ rations: strong astigmatism and field curvature (not shown here). The resolution of the subsystem has to be compared with the resolution of the retina. To do so, we introduce the Nyquist frequency 𝑓 1/2𝑝, which quantifies the geomet‐ rical sampling of the image plane, through the pixel’s pitch, denoted 𝑝. For a 10 μm pixel, like the holed retina, 𝑓 50 cycles/mm. Here, the contrast is completely lost at about 30 cycles/mm. It means that our blur spot is a bit larger than a pixel, which is quite advantageous for our application. Indeed, we don’t need to record scattering signatures with a pixel’s resolution, we even blur the image in post‐process (Gaussian blur, 𝜎2 pixels) to attenuate noise and other artefacts.

Appendix B In this appendix, we will describe the downward illumination module. Its main goal is to provide a focused beam just above the muzzle of the air channel, using a 45° mirror. Such an illumination system is realized by assembling commercially available optical el‐ ements such as a Laser Diode (LD) that provides 5 mW of optical power at a wavelength of 637 nm (Thorlabs, L637P5); a collimation aspheric lens ℒ (Thorlabs, C330TMD‐B, 𝑓 3.1 𝑚𝑚); and a focusing aspheric lens ℒ (Thorlabs, C280TMD‐B, 𝑓 18.4 𝑚𝑚). All Sensors 2021, 21, 3181 those elements are contained within a standard lens tube for 0.5 inch optics20 of 23(Thorlabs, SM05M10), that has an outer diameter of 17.78 mm. This module is presented in Figure A2.

Figure A2. (a) 3D Figureview of A2. the (fulla) 3D assembly view of theinvolving full assembly the illumination involving tube, the illumination a cloaker and tube, the a holed cloaker retina and the(partial holed cross sectional view). (bretina) RT simulation (partial cross of the sectional formation view). of the (b illumination) RT simulation beam. of the(c) Photography formation of of the the illumination illumination beam. tube. (c) Photography of the illumination tube. The ℒ, lens system is designed to properly form a focused beam above the fluidic Anchannel important from matter the bare to take LD. into First, account the ℒ in collimation the choice asphere of lenses is is chosen the angular with a mag- high numer‐ nificationical of aperture, the system: in order if θ1 toand captureθ2 are and polar collimate angles most from of the the source radiation and from to the the focal source; then, point respectively,ℒ focuses the then beam the relationin the aerosol between sampling the two zone. angles is given by the equation (demonstrationAn is important trivial): matter to take into account in the choice of lenses is the angular mag‐ tan θ f nification of the system: if 𝜃 and 2𝜃= are1 polar angles from the source and to the focal point respectively, then the relationtan θ1 betweenf2 the two angles is given by the equation The(demonstration LD provides a is Gaussian trivial): beam with a typical divergence of 8◦ (FWHM) for its ◦ slow axis and 34 (FWHM) for its fast axis. At thetan focal𝜃 point,𝑓 the divergence becomes 1.36◦ (FWHM) for its slow axis and 6.48◦ (FWHM) for its fast axis. tan 𝜃 𝑓 One can reasonably consider that when passing through the aspherical lens system, the beam is stillThe Gaussian. LD provides We recall a Gaussian that, for beam a Gaussian with a beam,typical the divergence steeper the of divergence 8° (FWHM) for its is, the smallerslow axis the waistand 34° and (FWHM) the shorter for the its Rayleighfast axis. distance At the focal are. Thepoint, waist the and divergence Rayleigh becomes 1.36° (FWHM) for its slow axis and 6.48° (FWHM) for its fast axis. distance roughly define the particle sampling volume (denoted Vs). A small volume allows to have a high optical power density and thus a bright scattering phenomena (which can allow to detect small particles). On the other hand, a low sampling volume reduces the probability of detecting a particle. The lens system used, and its consequence on the extend of sampling volume seems to be experimentally a good compromise, pending further study to determine an optimal configuration with a custom optical system.

References 1. Hamra, G.B.; Guha, N.; Cohen, A.; Laden, F.; Raaschou-Nielsen, O.; Samet, J.M.; Vineis, P.; Forastiere, F.; Saldiva, P.; Yorifuji, T.; et al. Outdoor Particulate Matter Exposure and Lung Cancer: A Systematic Review and Meta-Analysis. Environ. Health Perspect. 2014.[CrossRef][PubMed] 2. Anderson, J.O.; Thundiyil, J.G.; Stolbach, A. Clearing the Air: A Review of the Effects of Particulate Matter Air Pollution on Human Health. J. Med. Toxicol. 2012, 8, 166–175. [CrossRef] 3. Cohen, A.J.; Samet, J.M.; Straif, K.; International Agency for Research on Cancer. Air Pollution and Cancer; International Agency for Research on Cancer: Lyon, France, 2013; ISBN 978-92-832-2161-6. 4. Vineis, P.; Husgafvel-Pursiainen, K. Air Pollution and Cancer: Biomarker Studies in Human Populations †. Carcinogenesis 2005, 26, 1846–1855. [CrossRef][PubMed] 5. United States Environmental Protection Agency. Integrated Science Assessment for Particulate Matter (External Review Draft, 2018); U.S. Environmental Protection Agency: Washington, DC, USA, 2018; p. 1879. 6. Cormier, S.A.; Lomnicki, S.; Backes, W.; Dellinger, B. Origin and Health Impacts of Emissions of Toxic By-Products and Fine Particles from Combustion and Thermal Treatment of Hazardous Wastes and Materials. Environ. Health Perspect. 2006, 114, 810–817. [CrossRef] Sensors 2021, 21, 3181 21 of 23

7. Howard, V. Statement of Evidence: Particulate Emission and Health (An Bord Plenala, Proposed Ringaskiddy Waste-to-Energy Facility). Available online: http://www.durhamenvironmentwatch.org/Incinerator%20Health/CVHRingaskiddyEvidenceFinal1 .pdf (accessed on 3 January 2018). 8. Sioutas, C.; Delfino, R.J.; Singh, M. Exposure Assessment for Atmospheric Ultrafine Particles (UFPs) and Implications in Epidemiologic Research. Environ. Health Perspect. 2005, 113, 947–955. [CrossRef] 9. Fuzzi, S.; Baltensperger, U.; Carslaw, K.; Decesari, S.; Denier van der Gon, H.; Facchini, M.C.; Fowler, D.; Koren, I.; Langford, B.; Lohmann, U.; et al. Particulate Matter, Air Quality and Climate: Lessons Learned and Future Needs. Atmos. Chem. Phys. 2015, 15, 8217–8299. [CrossRef] 10. Clouds and Aerosols. In Climate Change 2013—The Physical Science Basis; Intergovernmental Panel on Climate Change, Ed.; Cambridge University Press: Cambridge, UK, 2014; pp. 571–658, ISBN 978-1-107-41532-4. 11. Jacobson, M.Z. Control of Fossil-Fuel Particulate Black Carbon and Organic Matter, Possibly the Most Effective Method of Slowing Global Warming. J. Geophys. Res. 2002, 107, 4410. [CrossRef] 12. Bond, T.C.; Doherty, S.J.; Fahey, D.W.; Forster, P.M.; Berntsen, T.; DeAngelo, B.J.; Flanner, M.G.; Ghan, S.; Kärcher, B.; Koch, D.; et al. Bounding the Role of Black Carbon in the Climate System: A Scientific Assessment: Black Carbon in the Climate System. J. Geophys. Res. Atmos. 2013, 118, 5380–5552. [CrossRef] 13. Ramanathan, V.; Carmichael, G. Global and Regional Climate Changes Due to Black Carbon. Nat. Geosci. 2008, 1, 221–227. [CrossRef] 14. Wang, C. A Modeling Study on the Climate Impacts of Black Carbon Aerosols: Climate Aspects of Black Carbon. J. Geophys. Res. 2004, 109.[CrossRef] 15. McMurry, P.H.; Shepherd, M.F.; Vickery, J.S. (Eds.) Particulate Matter Science for Policy Makers: A NARSTO Assessment; Cambridge University Press: Cambridge, UK, 2004; ISBN 978-0-521-84287-7. 16. Singer, B.C.; Delp, W.W. Response of Consumer and Research Grade Indoor Air Quality Monitors to Residential Sources of Fine Particles. Indoor Air 2018, 28, 624–639. [CrossRef][PubMed] 17. Lowther, S.D.; Jones, K.C.; Wang, X.; Whyatt, J.D.; Wild, O.; Booker, D. Particulate Matter Measurement Indoors: A Review of Metrics, Sensors, Needs, and Applications. Environ. Sci. Technol. 2019, 53, 11644–11656. [CrossRef][PubMed] 18. Kaliszewski, M.; Włodarski, M.; Młyńczak,J.; Kopczyński,K. Comparison of Low-Cost Particulate Matter Sensors for Indoor Air Monitoring during COVID-19 Lockdown. Sensors 2020, 17, 7290. [CrossRef] 19. Budde, M.; El Masri, R.; Riedel, T.; Beigl, M. Enabling Low-Cost Particulate Matter Measurement for Participatory Sensing Scenarios. In Proceedings of the 12th International Conference on Mobile and Ubiquitous Multimedia—MUM ’13; ACM Press: Lule¸, Sweden, 2013; pp. 1–10. 20. Jayaratne, R.; Liu, X.; Thai, P.; Dunbabin, M.; Morawska, L. The Influence of Humidity on the Performance of a Low-Cost Air Particle Mass Sensor and the Effect of Atmospheric Fog. Atmos. Meas. Tech. 2018, 11, 4883–4890. [CrossRef] 21. Snider, G.; Weagle, C.L.; Martin, R.V.; van Donkelaar, A.; Conrad, K.; Cunningham, D.; Gordon, C.; Zwicker, M.; Akoshile, C.; Artaxo, P.; et al. SPARTAN: A Global Network to Evaluate and Enhance Satellite-Based Estimates of Ground-Level Particulate Matter for Global Health Applications. Atmos. Meas. Tech. 2015, 8, 505–521. [CrossRef] 22. Wendt, E.A.; Quinn, C.W.; Miller-Lionberg, D.D.; Tryner, J.; L’Orange, C.; Ford, B.; Yalin, A.P.; Pierce, J.R.; Jathar, S.; Volckens, J. A low-cost monitor for simultaneous measurement of fine particulate matter and aerosol optical depth – Part 1: Specifications and testing. Atmos. Meas. Tech. 2019, 12, 5431–5441. [CrossRef] 23. Nguyen, T.N.T.; Luu, V.H.; Pham, V.H.; Bui, Q.H.; Nguyen, T.K.O. Particulate Matter Concentration Mapping from Satellite Imagery. In TORUS 3—Toward an Open Resource Using Services; Laffly, D., Ed.; Wiley: Hoboken, NJ, USA, 2020; pp. 103–130, ISBN 978-1-78630-601-2. 24. Liu, H.; Li, Q.; Shi, T.; Hu, S.; Wu, G.; Zhou, Q. Application of Sentinel 2 MSI Images to Retrieve Suspended Particulate Matter Concentrations in Poyang Lake. Remote Sens. 2017, 9, 761. [CrossRef] 25. van Donkelaar, A.; Martin, R.V.; Brauer, M.; Hsu, N.C.; Kahn, R.A.; Levy, R.C.; Lyapustin, A.; Sayer, A.M.; Winker, D.M. Global Estimates of Fine Particulate Matter Using a Combined Geophysical-Statistical Method with Information from Satellites, Models, and Monitors. Environ. Sci. Technol. 2016, 50, 3762–3772. [CrossRef][PubMed] 26. Snyder, E.G.; Watkins, T.H.; Solomon, P.A.; Thoma, E.D.; Williams, R.W.; Hagler, G.S.W.; Shelow, D.; Hindin, D.A.; Kilaru, V.J.; Preuss, P.W. The Changing Paradigm of Air Pollution Monitoring. Environ. Sci. Technol. 2013, 47, 11369–11377. [CrossRef] [PubMed] 27. Borghi, F.; Spinazzè, A.; Rovelli, S.; Campagnolo, D.; Del Buono, L.; Cattaneo, A. Domenico Cavallo Miniaturized Monitors for Assessment of Exposure to Air Pollutants: A Review. Int. J. Environ. Res. Public Health 2017, 14, 909. [CrossRef] 28. Amaral, S.; de Carvalho, J.; Costa, M.; Pinheiro, C. An Overview of Particulate Matter Measurement Instruments. Atmosphere 2015, 6, 1327–1345. [CrossRef] 29. Li, J. Recent Advances in Low-Cost Particulate Matter Sensor: Calibration and Application. Engineering and Applied Science Theses & Dissertations. Ph.D. Thesis, Washington University, St. Louis, MO, USA, 2019. 30. Kelly, K.E.; Whitaker, J.; Petty, A.; Widmer, C.; Dybwad, A.; Sleeth, D.; Martin, R.; Butterfield, A. Ambient and Laboratory Evaluation of a Low-Cost Particulate Matter Sensor. Environ. Pollut. 2017, 221, 491–500. [CrossRef] Sensors 2021, 21, 3181 22 of 23

31. Jobert, G.; Barritault, P.; Fournier, M.; Boutami, S.; Jobert, D.; Marchant, A.; Michelot, J.; Monsinjon, P.; Lienhard, P.; Nicoletti, S. Miniature Particulate Matter Counter and Analyzer Based on Lens-Free Imaging of Light Scattering Signatures with a Holed Image Sensor. Sens. Actuators Rep. 2020, 100010. [CrossRef] 32. Li, X.; Iervolino, E.; Santagata, F.; Wei, J.; Yuan, C.A.; Sarro, P.M.; Zhang, G.Q. Miniaturized Particulate Matter Sensor for Portable Air Quality Monitoring Devices. In Proceedings of the IEEE SENSORS 2014 Proceedings; IEEE: Valencia, Spain, 2014; pp. 2151–2154. 33. Dong, M.; Iervolino, E.; Santagata, F.; Zhang, G.; Zhang, G. Silicon Microfabrication Based Particulate Matter Sensor. Sens. Actuators A Phys. 2016, 247, 115–124. [CrossRef] 34. Qiao, Y.; Tao, J.; Zhang, Y.; Qiu, J.; Hong, X.; Wu, J.; Chen, C.-H. Sub-Micro Particle Matter Detection for Metal 3-D Printing Workshop. IEEE Sens. J. 2019, 19, 4932–4939. [CrossRef] 35. Nicoletti, S. Particle Detector and Method for Manufacturing Such a Detector 2012. Available online: https://permalink.orbit.com/ RenderStaticFirstPage?XPN=B5lJYIF%252BxRQ1zTc2PWcWRHfDUqlXTJ5uwQdFuycu4uk%3D%26n%3D1&id=0&base= (accessed on 3 May 2021). 36. Vincent, J.H. Aerosol Sampling; John Wiley & Sons, Ltd.: Chichester, UK, 2007; ISBN 978-0-470-06023-0. 37. Tryner, J.; Quinn, C.; Windom, B.C.; Volckens, J. Design and Evaluation of a Portable PM2.5 Monitor Featuring a Low-Cost Sensor in Line with an Active Filter Sampler. Environ. Sci. Process. Impacts 2019, 21, 1403–1415. [CrossRef] 38. Mahdavipour, O.; Fahimi, D.; Paprotny, I. Microfabricated Air-Microfluidics Virtual Impactor with Groove-Based Envelope-Flow Particle Focusing System. In Proceedings of the 2019 20th International Conference on Solid-State Sensors, Actuators and Microsystems & Eurosensors XXXIII (TRANSDUCERS & EUROSENSORS XXXIII); IEEE: Berlin, Germany, 2019; pp. 805–808. 39. Mulholland, G.W.; Choi, M.Y. Measurement of the Mass Specific Extinction Coefficient for Acetylene and Ethene Smoke Using the Large Agglomerate Optics Facility. Symp. Combust. 1998, 27, 1515–1522. [CrossRef] 40. Kullenberg, G. Scattering of Light by Sargasso Sea Water. Deep Sea Res. Oceanogr. Abstr. 1968, 15, 423–432. [CrossRef] 41. Bartholdi, M.; Salzman, G.C.; Hiebert, R.D.; Kerker, M. Differential Light Scattering Photometer for Rapid Analysis of Single Particles in Flow. Appl. Opt. 1980, 19, 1573. [CrossRef][PubMed] 42. Agrawal, Y.C.; Pottsmith, H.C. Instruments for Particle Size and Settling Velocity Observations in Sediment Transport. Mar. Geol. 2000, 168, 89–114. [CrossRef] 43. Heim, M.; Mullins, B.J.; Umhauer, H.; Kasper, G. Performance Evaluation of Three Optical Particle Counters with an Efficient “Multimodal” Calibration Method. J. Aerosol Sci. 2008, 39, 1019–1031. [CrossRef] 44. Pan, Y.-L.; Kalume, A.; Wang, C.; Santarpia, J.L. Opto-Aerodynamic Focusing of Aerosol Particles. Aerosol Sci. Technol. 2018, 52, 13–18. [CrossRef] 45. Weinert, D.; Cleary, T.G.; Mulholland, G.W.; Beever, P. Light Scattering Characteristics and Size Distribution of Smoke and Nuisance Aerosols. Fire Saf. Sci. 2003, 7, 209–220. [CrossRef] 46. Bohren, C.F.; Huffman, D.R. Absorption and Scattering of Light by Small Particles; Wiley: Hoboken, NJ, USA, 1983; ISBN 978-0-471-29340-8. 47. Hulst, H.C. van de Light Scattering by Small Particles; Dover Publications: Mineola, NY, USA, 1981; ISBN 978-0-486-64228-4. 48. Gouesbet, G.; Gréhan, G. Generalized Lorenz-Mie Theories; Springer: Berlin/Heidelberg, Germany, 2011; ISBN 978-3-642-17194-9. 49. Mishchenko, M.I.; Travis, L.D.; Mackowski, D.W. T-Matrix Computations of Light Scattering by Nonspherical Particles: A Review. J. Quant. Spectrosc. Radiat. Transf. 1996, 55, 535–575. [CrossRef] 50. Fournier, M.; Barritault, P.; Jobert, G.; Marchant, A.; Boutami, S.; Michelot, J.; Lienhard, P.; Nicoletti, S.; Duraffourg, L. A Miniaturized Optical Sensor for Particulate Matter Detection. In Proceedings of the Photonic Instrumentation Engineering VII; Soskind, Y., Busse, L.E., Eds.; SPIE: San Francisco, CA, USA, 2020; p. 43. 51. Michelot, J.; Monsinjon, P.; Caranana, J.; Menard, A.; Dubois, M.; Lesire, A.; Bouvier, C.; Cohet, S.; Caranhac, S. Creapyx: An Innovative Pixel Evaluation Platform; French Space Agency: Toulouse, France, 2015. 52. Dupont, B.; Caranana, J.; Pinoncely, P.A.; Michelot, J.; Bouvier, C.; Cohet, S.; Jourdain, P.; Monsinjon, P. A Dual-Core Highly Programmable 120dB Image Sensor. Electron. Imaging 2016, 2016, 1–3. [CrossRef] 53. Goodman, J.W. Introduction to Fourier Optics; Roberts and Company Publishers: Greenwood Village, CO, USA, 2005; ISBN 978-0-9747077-2-3. 54. Jobert, G.; Fournier, M.; Boutami, S.; Jamois, C.; Lovera, A.; Braga, D.; Seassal, C. Millimeter-Sized Particle Sensor Using a Wide Field of View Monolithic Lens Assembly for Light Scattering Analysis in Fourier Domain. In Proceedings of the Photonic Instrumentation Engineering VII; Soskind, Y., Busse, L.E., Eds.; SPIE: San Francisco, CA, USA, 2020; p. 24. 55. Ye, J.; Chen, L.; Li, X.; Yuan, Q.; Gao, Z. Review of Optical Freeform Surface Representation Technique and Its Application. Opt. Eng. 2017, 56, 1. [CrossRef] 56. Li, L.; Yi, A.Y. Design and Fabrication of a Freeform Microlens Array for a Compact Large-Field-of-View Compound-Eye Camera. Appl. Opt. 2012, 51, 1843. [CrossRef] 57. Sortino, M.; Totis, G.; Kuljanic, E. Comparison of Injection Molding Technologies for the Production of Micro-Optical Devices. Procedia Eng. 2014, 69, 1296–1305. [CrossRef] 58. Thiele, S.; Gissibl, T.; Giessen, H.; Herkommer, A.M. Ultra-Compact on-Chip LED Collimation Optics by 3D Femtosecond Direct Laser Writing. Opt. Lett. 2016, 41, 3029. [CrossRef] Sensors 2021, 21, 3181 23 of 23

59. Thiele, S.; Arzenbacher, K.; Gissibl, T.; Giessen, H.; Herkommer, A.M. 3D-Printed Eagle Eye: Compound Microlens System for Foveated Imaging. Sci. Adv. 2017, 3, e1602655. [CrossRef][PubMed] 60. Bellouard, Y. The Femtoprint Project. J. Laser Micro/Nanoeng. 2012, 7, 1–10. [CrossRef] 61. Krol, D.M. Femtosecond Laser Modiﬁcation of Glass. J. Non-Cryst. Solids 2008, 354, 416–424. [CrossRef] 62. Ross, C.A.; MacLachlan, D.G.; Choudhury, D.; Thomson, R.R. Optimisation of Ultrafast Laser Assisted Etching in Fused Silica. Optics Express 2018, 26, 24343. [CrossRef][PubMed] 63. Boreman, G.D. Modulation Transfer Function in Optical & Electro-Optical Systems; Tutorial Texts in Optical Engineering; SPIE Press: Bellingham, WA, USA, 2001; ISBN 978-0-8194-4143-0.