3D-Printed Micro-Optofluidic Slug Flow Detector

Non-Newtonian fluids analysis in microdevices is challenging both in biological and chemical applications. In this context, the flow velocity evaluation is crucial. This work presents a portable and disposable micro-optofluidic detector ( $\mu $ OFD), in which microoptical and microfluidic components are integrated and used for the real-time characterization of a sequenced flow generated by two immiscible fluids, called slug flow. The 3D-printed approach was chosen for the device fabrication, being simple, flexible, fast, and low-cost, and for the possibility of exploring wider channel geometries as compared to soft lithography. In the micro optofluidic detector, the light interacts with the flow in two observation points, 1 mm apart from each other, placed along the microchannel at 26 mm far from the T-junction. The optical signal variations, correlated with the fluids’ optical properties, were used for real-time tracking of slug frequency passage, velocity, and length by an ad hoc signal processing procedure. Two $\mu $ OFD prototypes were presented. One prototype was entirely made in poly-dimethyl-siloxane (PDMS), while in the second, the microoptical component was made of VeroClear and the microfluidic part in PDMS. Both prototypes were successfully characterized in different hydrodynamic conditions as proof of concept of their validity as flow velocity detectors. The advantage of realizing, by using a low-cost and easy-to-use fabrication process, a micro-optofluidic device that embeds the optical monitoring elements and the microchannels, without constraints on the localization of the observation point and microchannel height, opens the way to the design of a great variety of lab-on-a-chip (LOC) microdevices for complex fluids investigation.

non-Newtonian fluids [4].In the presence of non-Newtonian flows, e.g., immiscible fluids or particle suspensions, the real-time flow velocity monitoring and inner fluids dynamics in the microdevices plays a fundamental role in fluids property characterization, as demonstrated in many biological [5], [6], [7] and chemical applications [8], [9].
In particular, the slug flow (or segmented flow) is a non-Newtonian process obtained by an interlaced sequence of two immiscible fluids (e.g., air-water) [10].The use of slug flow in microchannels has resulted in a variety of cost-effective and interesting applications, such as electronics cooling [11], micromixing of liquid samples in chemical syntheses [12], [13], refrigeration (e.g., air-conditioning for thermal management of integrated circuits) [14], boiling [15],, and solvent extraction [16].Extensive research has demonstrated that, for these applications, slug flow presents notable advantages over single-phase flows [17], [18].The main reason for these advantages is the presence of re-circulation zones within and between the slugs, which facilitate efficient mixing.In the context of the aforementioned applications, it is essential to carefully monitor flow parameters, such as slug velocity, slug length, and slug alternation, throughout the process.Then, a primary objective is to integrate a detection system directly into the microfluidic chip to facilitate the development of portable microdevices with heightened sensitivity and rapid response capabilities.Several methodologies are employed for flow monitoring, such as electrical [19] or optical [20] detection techniques.Among those, the use of optical technology in microfluidic process monitoring is very often necessary and offers the advantage of reduced invasiveness [21], [22].However, its use entails the need to guarantee optical access within the device.For this reason, both in the academic and industrial fields, the challenge is to develop transparent miniaturized devices having integrated optical and fluidic features by using time-saving and low-cost manufacturing techniques [23], [24], [25].Poly-dimethyl-siloxane (PDMS) is a widely used material for LOC device fabrication, which combines a good resolution, up to a few micrometers, with the advantages of biocompatibility, optical transparency, impermeability to liquids, and high chemical resistance [26].The traditional fabrication technique for the PDMS realization of microfluidic and microoptical components is soft-lithography [27], [28].This manufacturing process requires access to clean-room facilities and time-consuming manufacturing procedures [27], [29].Another critical limitation of this technology is the size of manufacturable channels, whose heights are typically limited to less than 100-150 µm.This presents a significant drawback in the study of slug-flow and other non-Newtonian fluids in microchannels, as the dynamic of the fluids need to be characterized in real working conditions and equivalent environments with larger microchannels.In the last ten years, 3D-printing technology has emerged and has offered the possibility to rapidly and cost-effectively realize complex 3D structures made of different materials, with microchannels up to 1 mm in height.Since then, the quality and quantity of the 3D-printed microfluidics devices [30], [31], [32], [33] and LOC applications [34], [35], [36], [37] have not stopped increasing.Nowadays, many 3D-printing technologies have demonstrated their suitability in fabricating nano-or microoptical components and devices (e.g., optical waveguides, lenses, and optical fibers) through one-step, repeatable, and high-resolution processes [38].Table I illustrates some common 3D-printing technologies used in these fields with evidence in their build volume capability.For these purposes, used materials range from optical glass to crystal, polymers, and metals [38], [39].Table S1 (reported in supplementary materials) reports examples of microoptical components realized by using different 3D-printing technologies pointing to their dimension, the used materials, and the application fields such as imaging, sensing, and photonics [40].Table S2 (reported in supplementary materials) points to the 3D-printed micro-optofluidic devices, that are still in a very early stage, with an indication of the used material and dimensions.In previous works, the authors investigated the potential of using the inkjet 3D-printing technology via a master-slave procedure in order to manufacture both PDMS microfluidics [41], [42] and microoptical elements, i.e., microwave guides [43] and a microsplitter [44].
In this work, we present a micro-optofluidic detector (labeled µOFD), fully realized in 3D-printing technology, integrating the microfluidic T-junction and the microopticalsplitter (labeled µSPT) for real-time detection and characterization of the slug flow inside a microchannel with a height of 400 µm.Particularly, two prototypes of the microoptical detector were realized using two materials: PDMS and VeroClear RGD810 [45].In the first prototype, both the microoptical splitter and the microfluidic components were realized by the PDMS master-slave 3D-printing approach [41] and then integrated.In the second prototype, a hybrid strategy was carried out: the microoptical splitter was realized by VeroClear RGD810, using direct 3D-printing, and then integrated with the PDMS microfluidic component.The optical signal analysis procedure, developed by the authors in the previous work [46], was here extended and implemented to compute multiple flow information in a designed microchannel observation area: the frequency, velocity, and length of the slug passage.Both prototypes were successfully characterized by comparing their performance in different hydrodynamic conditions.A wide analysis of the difference between the slug velocity and length inside the microchannel (obtained by the µOFD) and outside in the tube (measured by external sensors) has evidenced the relevance of the in situ detection, as a consequence of the non-Newtonian characteristic of the slug flow.
The significance of the proposed study relies both on the µOFD design, i.e., an easy-to-use portable device suitable for LOC integration, and on the proposed PDMS-based 3D-printing fabrication approach, i.e., adaptable to different microchannel dimensions, needed for complex fluids investigation, whilst maintaining the biocompatibility property.In addition, real-time optical signal monitoring eliminates the need for image processing analysis or external sensors common in biochemical applications.
The article is structured as follows.The design of the µOFD, supported by ray-tracing simulations, is presented in Section II.Section III describes the 3D-printing manufacturing procedure.Section IV reports the experimental set-up and the performed experimental campaigns.Section V shows and compares the performance of two µOFDs together with the description of the implemented signal analysis procedure.

II. DEVICE DESIGN
The ad hoc µOFD was designed to be a compact system integrating both the microoptic and microfluidic components, thus reducing the need for bulky optical equipment suitable for further ON-chip implementation.To achieve this aim, it was crucial to confine and transport the light as close as possible to the sample, by guiding it into a selected area of the microfluidic channel and then collecting in situ the transmitted radiation.

A. Operation Principle
The micro-optofluidic device integrates a microfluidic T-junction to generate an air-water slug flow, the microsplitter to reorient a light beam and split it into two paths, and three optical fiber insertions.Through the input optical fiber insertion, the light source is conveyed to the µSPT and consequently split into two paths directed toward two close segments on the microchannel's investigation area.After the interaction with the fluid (air or water), the two optical paths are detected by the two output optical fibers aligned at the opposite side of the microchannel.The two optical signals are then analyzed to extract information about the fluids' passage.
The schematic of the microdevice working principle is shown in Fig. 1(a) and the upper view of the CAD design, with a zoomed-in view of the investigation area and the observations points, is presented in Fig. 1(b) and (c).In particular, the microfluidic T-junction is composed of two inlets through which the two fluids, i.e., air and water, are injected, and an outlet through which the fluid exits.
The design of µSPT was previously studied in detail in [44].It was realized by including a micromirror (µMR) between two microwaveguides (µWGs) of length L with a squared section of 1 × 1 mm 2 as shown in Fig. 2(a).Briefly, the operative principle to bend the light beam coming from the input source is as follows.An input optical fiber guides the light, coming from a light source, through a first waveguide (µWG1) toward a µMR.The light, after being angled by the µMR, is guided to the output optical fiber by a second waveguide (µWG2).The µMR geometry consists of two angled surfaces M1 and M2, respectively tilted of the angles {β = 57.35 • ; γ = 26 • } with respect to the light input section [see Fig. 2(b)].After the light is bent by the µMR, two light beams are obtained with a direction angled of about 36 • with respect to the output section of the µMR, as shown in Fig. 2(c).Considering both the 3D-printing fabrication requirements and the hydrodynamic and optical properties of the process, the µOFD was designed as follows.
1) A T-junction microchannel with a square cross section was set to 400 × 400 µm 2 and the total length of the microchannel to 6.4 cm. 2) To ensure a slug flow process stabilization, the investigation area was set at a distance of 2.6 cm from the T-junction [10].
3) The investigation area extends 1.7 mm within the microchannel, from point A to B, as shown in Fig. 1(c).4) The µSPT diameter was set to 1 mm due to the 3D-Printing limitations [44].5) The diameter of the input and output optical fibers were 400 µm equal to the microchannel width to maximize the acquisition [43].6) The distance between the µSPT and the microchannel was set to 1 mm, to optimize the ray-tracing distribution.7) The distance of the microchannel and the two output optical fibers insertions was set to 500 µm, to reach an optimal trade-off among the device manufacturing by avoiding fluid leakage and light losses.

B. Ray-Tracing Optical Simulations
The performance of the micro-optofluidic device was studied by ray-tracing simulations (TracePro, Lambda Research Corporation, Westford, MA, USA).The primary objective of this simulation was to compare two versions of µOFD detectors by analyzing the trajectory, spatial distribution, and intensity dispersion of light rays.More specifically, the two versions of µOFD detectors, named PDMS-µOFD and VeroClear-µOFD, differ for the µSPT integrated with the PDMS microfluidic device.The two microoptical splitters are as follows: one was made using PDMS (named as PDMS-µSPT) and one using VeroClear RGD810 (named as VeroClear-µSPT), with refractive index value n PDMS = 1.41 and n VC = 1.53 [45], respectively.Both the µSPT were already analyzed independently by ray-tracing simulations in [44], assuming that they are surrounded by air (n 0 = 1), where it is proved that the performance of the PDMS-µSPT and the VeroClear-µSPT are almost equivalent.In both cases, 60% of the rays reaches the output fibers surface with a variation up to 8%, depending on the waveguide length (L).
The PDMS-µSPT, having to be surrounded by PDMS in the µOFD integration, was designed to be gold shielded (refractive index value n Au = 0.47) to convey the light within the microoptical component, by mean of the total internal reflection phenomenon.Indeed, according to the waveguides working principle [47], a cladding (i.e., gold) characterized by a lower refractive index value than the core one (i.e., PDMS) allows for achieving better performance for the optical component.Conversely, regarding the VeroClear-µSPT, with its refractive index value greater than PDMS one (in this case, the surrounding material works directly as cladding), the transmitted light travels from the source through the microoptical component without creating any additional cladding for the core.
The amount of light transmission from the input optical fiber through the µSPT to the microchannel investigation area and, subsequently, to the two output optical fibers was evaluated.With this aim, a study was conducted to investigate the spatial distribution of incident rays across four specific surfaces: the input and the output surfaces of the µSPT (labeled IS 1 and IS 2 , respectively), and the surfaces of the two output optical fibers (OF 1 and OF 2 ), as shown in Fig. 3(a).For the simulation, a light source of 100 rays with 10 mW of power was used.The complete path of the light inside the PDMS-µOFD is shown in Fig. 3(b).
The radiance maps obtained for each surface of interest of the PDMS-µOFD are reported in Fig. 4. Thus, the percentage of incident rays evaluated in an area of 1 mm 2 in {IS 1 } and {IS 2 } is 100%, with a power loss lower than 1%.Indeed, the entirety of the light rays introduced into the IS 1 surface was able to reach the IS 2 surface, even though this was done with a different spatial distribution; the rays are not focused as in IS 1 surface, as shown in Fig. 4(a) and (b).This observation indicates that the µSPT transmits light rays efficiently.Since we are investigating the slug flow between two immiscible fluids (i.e., deionized water and air characterized by the refractive index values n water = 1.33 and n air = 1, respectively) for the light rays transmissions between the IS 2 surface and the two output surfaces {OF 1 and OF 2 }, two different cases were considered.The first one is with an air-filled microchannel, while the second is with a water-filled microchannel.The transmitted ray power is attenuated when it reaches the output optical fibers due to: the distance between IS 2 and the output surfaces {OF 1 , OF 2 }; the absorption caused by fluids (i.e., air and water) and chip material (i.e., PDMS).It was observed that there is a loss of 70%-90% of ray power using both PDMS-µOFD and VeroClear-µOFD.If the microchannel is air-filled, there is a power loss of 90%, and 10% of power gets transmitted at the output surfaces of OF 1 and OF 2 , as shown in Fig. 4(c) and (d).If the microchannel is water-filled, there is a power loss of 70%, and 30% of power gets transmitted at the output surfaces of OF 1 and OF 2 , as shown in Fig. 4(e) and (f).The difference in power losses in the two conditions is due to the difference between the refractive index values between the PDMS-water and PDMS-air.
Based on the previous simulations, it can be concluded that the µSPT is capable of efficiently splitting light rays and directing them to two closely spaced points within the microchannel area.Additionally, the difference in power losses between the two scenarios considered (air-filled and waterfilled microchannel) enables the discrimination of the fluid passing through the microchannel at any time.

III. DEVICE MANUFACTURING
In the fabrication of the two µOFDs prototypes, three different phases can be distinguished.
1) The realization of the PDMS-µOFD, in which the microfluidic T-junction, the slots for the µSPT and the optical fibers insertion are integrated.2) The realization of the µSPT using either PDMS or VeroClear.
3) The final assembly of the µOFD by the insertion of either the PDMS-µSPT or VeroClear-µSPT in the designed slot and the device closing by bonding the PDMS-µOFD with bulk PDMS.In the first phase, the PDMS-µOFD was fabricated using a specific 3D-printing-based master-slave procedure, implemented by Cairone et al. [41].The fabrication procedure, which is schematized in Fig. 5, consists of five different steps: 1) design of the CAD model for the masters; 2) its manufacturing through an inkjet 3D-printing technique; 3) the surface UV treatment to avoid the issue of leaving surface area not fully cured, by compromising the final surface finish of the mold; 4) the PDMS pouring within the master and its curing process; and 5) PDMS demolding from the master.In particular, the master was 3D-printed using an inkjet 3D-printer (Objet260 Connex1, Stratasys, Los Angeles, CA, USA).Once the 3D-printing procedure was accomplished, the support material used (FullCure705, OVERMACH S.p.A, Parma, Italy) was washed out through a water jet.Through the afore-described procedure, the µOFD in PDMS is obtained, where the desired geometry is patterned.Then, a flat base, namely a bulk layer of PDMS, is used to close the µOFD.However, before the closing procedure, it is necessary to insert the µSPT inside the µOFD's designed slot.Therefore, in the second phase, the two versions of the µSPT were realized by using different procedures.In the PDMS-µSPT, the already described 3D-printing-based master-slave procedure was used.Then, the PDMS-µSPT was gold-sputtered up to a thickness of 20 nm, using a sputter coater (AGB7340, Agar Scientific, London, U.K.).The VeroClear-µSPT was directly 3D-printed using the inkjet 3D-printer Objet260 Connex1.The VeroClear RGD810 resin has a proprietary formulation developed by Stratasys for the PolyJet1 3D-printing technique.According to the safety data sheets (SDSs), it is an acrylic liquid photopolymer made of a complex mixture of photo activators and acrylate monomers.To fully remove the support material's residues from the 3D-printed VeroClear-µSPT, thus achieving a cleaner and smoother surface, the printed part was soaked in a 1% solution of sodium hydroxide, as suggested by Stratasys postprinting process guide.More details on the µSPT fabrication are reported in [44].Finally, the third phase includes the integration of the µSPT in the designed slot of the µOFD and the bounding with a 0.5 mm thick bulk by a reversible bound procedure.More specifically, the µSPT integration was made manually, by placing the splitter in the slot of the µOFD.The pictures of the two realized µOFDs are shown in Fig. 6: the PDMS-µOFD obtained with insertion of the gold-spattered PDMS-µSPT [in Fig. 6(a)] and the VeroClear-µOFD with the VeroClear-µSPT [in Fig. 6(b)].

IV. EXPERIMENTAL SET-UP AND CAMPAIGN A. Experimental Set-Up
An experimental set-up based on the simultaneous monitoring of the process through optical and flowmeter detectors has been realized and it is shown in the block scheme of Fig. 7.The optical sensors give information about the process inside the microchannel without direct access to the flow, while the flowmeter monitors the flow in correspondence with the microchannel outlet.A CCD camera was placed above the device for a visual inspection of the process in the investigation area.
The picture of the real experimental set-up is shown in Fig. 8(a).The continuous slug flow is generated by simultaneously pumping deionized water and air at the inlets of the T-junction by means of two syringe pumps (neMESYS by Cetoni Gmbh, Münster, Germany) connected to the two-channel inlets [see Fig. 8(b)].The input light source is a laser system (Rgb NovaPro Laser 660-125, Lasersystems, Kelheim, Germany) which generates a light beam with a wavelength of 600 nm and a maximum output power of 128 mW.The light is collected at the two output optical fibers connected with two photodiodes (PDA 100A, Thorlabs, Newton, NJ, USA, gain used 70 dB) and then acquired by a PC oscilloscope (Picoscope 2204A, Pico Technology, Cambridgeshire, U.K.),  with a sampling frequency of 1.5 kHz.The diameter of the input and output optical fibers at the insertions is equal to 365 µm.A flowmeter (SF1300, Sensirion, Stäfa, Switzerland), with a sampling frequency of 200 Hz, is connected at the microchannel outlet, through a Tygon tube having an inner diameter of 1.3 mm.

B. Optical Signal Monitoring
The optical monitoring system, composed of the two photodiodes, captures the light in two points of the microchannel, in correspondence of the investigated area of the µOFD [see Fig. 1(b)].The variation in the luminosity during the slug passage is due to the difference of the refraction index values {n PDMS , n water , n air }, as discussed in Section II-B.As a consequence of this phenomenon, the air and liquid passages are detected in the optical signal in correspondence of two brightness levels [46].The top level reveals the water presence, while the low level reveals the air passage.Additionally, two peaks can be recognized at the air front and rear.The signal levels at the slug passage associated with the video frames acquired using the VeroClear-µOFD and PDMS-µOFD are shown in Fig. 9(a) and (b), respectively.In Fig. 9(a), in addition to the optical signal acquired during the slug flow (PH), the reference voltage levels are also reported (blue and red lines).These values were determined for the air-filled and water-filled microchannel.It is possible to notice how, during the passage of the water-slug, the signal is at the same level as the water-reference, while at the passage of the air-slug, the signal does not overlap the air-reference.This highlights the presence of a small layer of water sticking to the walls during the passage of the air slug, which is correlated with the recirculation effect.
Even though during the simulation phase no differences between the two µOFD used were detected, the PDMS-µOFD is affected by a greater light dispersion compared with the VeroClear-µOFD, as highlighted by the difference in the optical signal range: [1.4; 1.7] V for the VeroClear-µOFD [see Fig. 9(a)] and [0.03; 0.13] V for the PDMS-µOFD [see Fig. 9(b)].This result is justifiable by the fact that a lower accuracy is obtained for the fabricated PDMS-µSPT using the master-slave approach since it is an operator-dependent protocol when compared to the direct 3Dprinting approach, which is a completely automated procedure.The same dispersion was observed in the experimental characterization of the VeroClear-µSPT and PDMS-µSPT [44].
Two reference conditions (water-filled and air-filled microchannels), and four laser power levels ({1, 5, 10, 25} mW) were considered to compare the performance of VeroClear-µOFD with ray-tracing simulation.By using the direct characterization of laser-photodiodes, reported in [43], and the voltage values acquired by the optical signals, the transmission index was computed, as shown in Fig. 10.Considering the laser level of 10 mW, as in the ray-tracing simulation, for the air-filled microchannel, the transmission index is about 4.7% with losses of about 96.3%, while in the case of water-filled microchannel, the transmission index is about 8% with losses of about 92%.Comparing these values with those obtained from simulations (see Fig. 4), we observe a decrease in the actual experimental performance: the average transmission in water-filled conditions was approximately 30%, and in air-filled conditions, it was to around 10%.This difference can be correlated with the input-output fiber insertions that affect the optical signal amplitude [43].Still, it does not    The signals acquired by both the photodiodes and the flowmeter were preprocessed.Initially, a low-pass filter with a 40 Hz cut-off frequency was applied to remove high-frequency harmonics.Then, a smoothing procedure was performed to remove the noise from the signal and unveil the main square wave pattern, as shown in Fig. 11(a)-(c).
Additionally, the flow monitoring was obtained by the flowmeter placed at the outlet of the devices (see Fig. 1) which measures the velocity of the fluid in the tube.Two levels are distinguished in the flow signal: the flow value different from zero reveals the water presence (being the sensor sensitive only to liquid) and the flow value equal to zero is for the air presence.The negative and positive peaks are related to the air-slug front and the air-slug rear, respectively.The flow signal ranges between 0 ml/min and the water flow-rate imposed at the inlet, that is 0.1 ml/min, and the signal levels associated with the acquired video frames are shown in Fig. 9(c).

C. Experimental Campaign
In the experimental campaign, the hydrodynamic characteristic of the slug flow was investigated using different powers of the laser light and pressure strength at the inlet of the microchannel, as summarized in Table II.A total of 18 experiments were run to obtain a device characterization and quantify their performance.As expected, due to the highest sensitivity to the light of the VeroClear-µOFD, a lower value of the laser power was sufficient to have a significant difference in air-water range during the optical detection of slug flow.Based on that, 15 experiments were conducted varying the laser power levels {1; 5; 10; 25} mW and labeled per laser power as {Exp-1; Exp-2; Exp-3; Exp-4} (see Table II).Three experiments were carried out with the PDMS-µOFD by setting the laser power value at 25 mW and labeled as Exp-5 (see Table II).In correspondence with each laser power condition, the input flow rate (Fl in ) was set equal for the two fluids used {air, water} and varied on three different levels, i.e., {0.05; 0.1; 0.2} mL/min.The duration of each experiment was set to 60 s.

V. RESULTS AND DISCUSSION
The µOFD dynamical characterization of both detectors is reported presenting their performance.In the processing phase, the optical and flow signal were analyzed both in frequency and time domain.The spectral analysis was carried out to detect the frequency of the slug passages.The correlation-based time domain analysis of the optical signals was used to compute the slug velocity [48].Both values were necessary to compute the slug length.

A. Detection of Slug Frequency Passage
Thanks to the spectral analysis of the optical and flowmeter signals, it was possible to evaluate the slug frequency, i.e., the slug passage duration in time.The optical sensors monitor the process inside the microchannel, while the flow sensor at the outlet.Since the spectra of the optical signals acquired by the two photodiodes {PH 1 ; PH 2 } equivalent, only the information of one photodiode {PH 1 } was used.
In Fig. 11, the trends and the spectra of the signals acquired by the two types of sensors during the Exp-3 are shown.In particular, both the filtered optical and flow signals were constructed by a square wave (in red) as shown in Fig. 11(a)-(c), respectively.The square waves were used to identify the slug period corresponding to the passage of water (T w ) and air (T a ).Consequently, the period of the entire air-water passage was determined by considering the sum of the latter two parameters [see (1)].By the analysis of the optical and flow signals in the frequency domain, the spectra were obtained [see Fig. 11(b)-(d)], where it is possible to identify the frequency of the dominant peak ( f p ).The f p is equal to the inverse of the mean air-water passage (T ) as in the following equation: It can be verified from the f p in the spectra [Fig.11(b)-(d)] where f p−tube = 0.12 Hz (T tube = 8 s) and f p−µch = 0.24 Hz (T µch = 4 s).The spectral analysis approach offers the advantage of a fast real-time implementation, easily adaptable to slow and fast slug-flow processes [48].The spectral analysis was extended to all the experiments using the VeroClear-µOFD (from Exp-1 to Exp-4 in Table II) by varying the input flow rates {0.05; 0.1; 0.2} mL/min.In Fig. 12, the mean period   Empirically, it is also possible to quantify a relation, reported in (2), where k T ≃ 2. This equation relates the mean slug period of the flow within the microchannel (T µch ) and in the tube (T tube ).As known by the Venturi relation, the velocities in the microchannel and the tube are affected by the change in the area of their cross sections as in (3).In this case, the ratio between the perimeter of the microchannel and the tube cross-section {P tube ; P µch } is 2.5, i.e., very close to the value empirically obtained for k T Additionally, in Fig. 12, it can also be observed that the slug-flow process is more sensitive to the intensity of the laser power when the flow rate is set at its slowest value, i.e., 0.05 mL/min.The sensitivity to this variation is higher using the optical measurement inside the microchannel.

B. Slug Frequency: µOFDs in Comparison
In Fig. 13, the values of the mean period of slug flow detected by the optical signals using both the VeroClear-µOFD and PDMS-µOFD are compared.Despite the difference in the signal range (as discussed in Section III), the values obtained for the slug flow mean period are almost equivalent, which confirms the robustness of the µOFD devices for the two selected manufacturing processes.This variation could be a consequence of process nonperiodicity, as discussed in [49].To rationalize the repeatability of the developed µOFD devices' ability to monitor the slug process inside a microchannel, a design of experiment (DoE) approach was exploited.A replicated general factorial design was studied, in which the investigated factors (independent variables) for the experimental design are reported below.III.The investigated response (dependent variables) for the experimental plan is the mean period (T µch ) associated with a complete slug flow passage.An analysis of variance (ANOVA) was carried out on the collected observations for the investigated responses in Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.

TABLE IV SLUG FLOW PROCESS: ANOVA TABLE FOR THE RESPONSE MEAN PERIOD (T µch ) ASSOCIATED WITH A COMPLETE SLUG-FLOW
order to analyze the statistical significance of each investigated factor and their possible interaction.
The ANOVA table for the response mean period T µch associated with a complete air-water slug is reported in Table IV.The obtained results proved that the two investigated factors, i.e., µSPT material (factor A) and flow rate (factor B), and even their interaction are influential factors ( p − value < 0.0001).Next, by considering the model adequacy checking, no anomalies were identified for the residuals.In the end, it can also be reasonably assumed that most of the variability found for the collected observations is associated with the variation of the considered independent variables, i.e., µSPT material and flow rate.Both the evaluated R-squared and the adjusted R-squared values are very high (R 2 = 0.9985; R 2 adj = 0.9978).In accordance with the trend for the mean period (T µch ) associated with a slug flow (see Fig. 13) for the two µOFD by varying the factor A, i.e., the material of which the µSPT is made, a significant variability for the investigated response is found.This is due to the variability correlated with the nonperiodicity of the slug flow and the typical variability of the chosen manufacturing process (PDMS-based master-slave approach) for the microfluidic T-junction.A similar result, characterized by an analogous dispersion for the collected observation T µch , was found in a previous study [49].Thus, the factor A should not be considered as an influent factor.Furthermore, as expected, the investigated response decreased by increasing the set flow rate, which is consistent with the significance of the factor the B. So, the unique influent parameter is factor B, which is strictly related to the hydrodynamic process, consistent with previous results [49], [50].

C. Detection of Slug Flow Velocity
The slug velocity was evaluated by processing the optical signals acquired from both photodiodes {PH 1 ; PH 2 }.The two optical signals, monitoring the same process at two observation points in the investigation area, exhibit identical waveforms but vary in terms of temporal displacement.Being known the distance between the observation points, and using crosscorrelation analysis, it is possible to determine the delay between the signals and thus the slug velocity inside the microchannel.The optical signals {PH 1 ; PH 2 } acquired by the VeroClear-µOFD in Exp-3 with laser power of 10 mW and input flow rate of 0.1 mL/min are shown in Fig. 14.In Fig. 14(b), their cross-correlation function the time delay T d was plotted.Indeed, the peak in the cross correlation can be In Fig. 15(a), the slug velocity values using the VeroClear-µOFD were computed for all the experiments from Exp-1 to Exp-4, as reported in Table II.The values are quite consistent and no significant variation can be observed when increasing the laser power.As a consequence, the difference in the slug passage period, discussed in Section V-A, can be associated with the air/water slug shrink.In Fig. 15(b), the values of the slug velocity, obtained using the VeroClear-µOFD and the PDMS-µOFD, are compared.As in the case of the detection of the slug passage period, the results are consistent, but the velocity in the PDMS-µOFD is always lower than the one obtained for the VeroClear-µOFD.Conversely, the period of the slug passage is always greater than the PDMS-µOFD (Fig. 13).This difference arises from the fact that velocity is determined as an average value observed over an extended duration.As discussed in the previous section, a significant contributing factor to the observed variability is the different manufacturing process employed for the two devices.
Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.

D. Slug Velocity Inside Microchannels
In correspondence with the experimental scenario Exp-1 (lower power level) using the VeroClear-µOFD, three other experiments were carried out considering the input flow rate {0.025; 0.3; 0.4} mL/min.The slug velocity inside the microchannel and in the tube was estimated analytically by ( 5) and (6), where Fl in is the input flow rate expressed in m 3 /s, while s µch and s tube are the area of cross section of the microchannel and of the tube expressed in m 2 In the two equations above, the input flow rate was doubled, due to the T-junction structure with two incoming equal flows joining in the microchannel.In Fig. 16, the values of the slug velocity determined were compared with those computed analytically inside the microchannel (v µch ) and in the tube (v tube ).It can be noticed that, by varying the input flow rate, the slug velocity values inside the microchannel experimentally and analytically differ.Moreover, defining ( kv ) as the ratio between the analytical velocity computed inside the microchannel (v µch ) and the analytical velocity in the tube (v tube ), and (k v ) as the ratio between the experimental velocity inside the microchannel (v µOFD ) and the analytical velocity in the tube (v tube ), as reported in (7), a comparison can be computed for all the input flow rate conditions, as shown in Fig. 16.Based on the Venturi effect [see (3)], the expected value is equal to 8.29 for all the input flow rates, but as shown in Fig. 16, the condition cannot be considered valid Based on the results obtained, the value of the slug velocity computed by (5), cannot describe properly the process inside the microchannel.This result underlines the need to use the µOFD device and the proposed noninvasive approach for slug flow detection and characterization.Indeed, in the context of a highly nonlinear microfluidic process a direct process monitoring becomes imperative.This is attributed to the inadequacy of predicted values in providing a comprehensive depiction of the process within the microchannel and the dynamics of fluids interaction.Finally, the mean length of a slug passage (airwater) L can be calculated by ( 8) using the values obtained for the mean period of the slug passage T µch and the mean slug velocity v µOFD Fig. 17.Slug length obtained using the VeroClear-µOFD in Exp-1 compared with the slug length obtained analytically in the tube per the input flow rates.
A comparison of the evaluated mean slug length values in the microchannel and the tube is presented in Fig. 17.The length of the slug in the tube was computed using v tube and the T µch extracted by the flow signals.A greater variation can be observed in the length of the slug inside the microchannel than in the tube, as a consequence of both variations in the slug period and velocity due to the nonlinearity of the process.

VI. CONCLUSION
In this article, a micro-optofluidic device realized by 3D-printing-based approach, which integrates a microfluidic T-junction and a µSPT, is presented to detect slug flow in microchannels.The possibility of monitoring the process inside the microchannels plays a crucial role in different chemical and biological applications, especially in the presence of non-Newtonian fluids.
The microdevice novelty relies both on its design, i.e., an easy-to-use portable device suitable for LOC integration, and on the low-cost 3D-printing-based manufactured approach, i.e., that involves unmanned operation for its realization.In particular, two prototypes of the microoptical detector were realized: the PDMS-µOFD, entirely in PDMS, and the VeroClear-µOFD, a hybrid device where the microoptical splitter was made of VeroClear, using the direct 3D-printing approach, while only the microfluidic component is in PDMS.Both prototypes were successfully characterized by comparing their performance in different hydrodynamic conditions.They were able to perform a real-time tracking of the slug passage by an ad hoc signal analysis procedure evaluate the slug's frequency passage, length, and velocity.
Particularly, the hybrid version offers the advantage of improving the optical performance in signal detection by maintaining the biocompatibility of the microdevice.A reduced light dispersion was obtained in the experimental results using the VeroClear-µOFD rather than the PDMS-µOFD.This result is correlated with the direct 3D-printing process, which allowed higher accuracy for the microoptical component.Additionally, due to the light sensitivity of the two materials used, it is worth highlighting that a lower value of the input laser power was required to achieve a significant difference in air-water level range for the optical detection of slug flow when the VeroClear-µOFD is used (i.e., 1 mW) rather than when the PDMS-µOFD is selected (i.e., 25 mW).Finally, this microdetector offers optimal performance in terms of cost-effectiveness as proved by the cost analysis reported in the Supplementary Materials section.Moreover, the final cost can be further reduced by using cheaper photocurable resins suitable 3D-printing techniques [51], [52].
The results obtained in this study evidence the suitability of µOFDs for a wide range of applications involving non-Newtonian fluids investigation and control in a chip: such as the characterization of immiscible fluids, or of different fluid regimes by properly tailoring the existing difference between the fluid refractive index values, or such as the evaluation of the fluids' viscosity (i.e., blood or cell/microparticle suspensions) as an indirect measure from the detected velocity.Moreover, thanks to the simplicity of its design, the micro-optofluidics detector could also be easily adapted to complex microchannel structures to implement fluid monitoring in specific areas inside the microchannel.An ad hoc integrated control strategy could also be developed using the acquired optical signals to tune the fluids' velocity.
Future research activities will be driven toward methodological aspects involving its use with particle suspensions, as well as the ON-chip fluid control, and toward the assessment of technological issues for the realization of a µOFD detector fully fabricated using a direct 3D-printing approach, taking into account a multimaterial strategy.Ultimately, the µOFD represents a proof of the concept of the high potential of the proposed direct 3D-printing fabrication approach, based on quite low-cost technologies, for the creation of a new class of micro-optofluidic LOC devices.She is currently a Postdoctoral with the University of Catania.Her research interests include the multifunctional additive manufacturing relying on 3D printing, polymer, and composite materials.
Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.

Fig. 1 .
Fig. 1.(a) Operation principle of the micro-optofluidic slug detector.(b) CAD top view of the micro-optofluidic device: the microfluidics T-junction, the input and output optical fibers insertions, and the microsplitter.(c) Zoomed-in view of the CAD in the microchannel investigation area.

Fig. 2 .
Fig. 2. (a) Design of the microsplitter: the µMR between the two µWGs of length L. (b) CAD top view of the µMR.(c) Optical path obtained by ray-tracing simulations shows the light split at the output section.

Fig. 3 .
Fig. 3. (a) Enlargement in the µOFD CAD of the four test surfaces investigated in the ray-tracing simulations.(b) Complete ray-tracing path for PDMS-µOFD obtained with water-filled microchannel.

Fig. 4 .
Fig. 4. Radiance maps at the four test surfaces in the PDMS-µOFD, showing the spatial distribution and the percentage of incident light rays, at the surface IS 1 (a) and IS 2 (b) of the µSPT, and at the two output optical fibers {OF 1 and OF 2 }.In the presence of air inside the microchannel {OF 1 and OF 2 } (c)-(d), and in the presence of water {OF 1 and OF 2 } (e)-(f).The radiance maps were evaluated in an area of 1 mm 2 along the X -Z -axes.The color map indicates the light power expressed in W/m 2 .

Fig. 5 .
Fig. 5. Phases of the 3D-printing master-slave protocol PDMS-based to fabricate the µOFD in which T-junction, the µSPT, and the three areas for the fiber insertions are integrated.

Fig. 8 .
Fig. 8. Pictures of (a) complete experimental set-up, (b) zoomed-in view of the µOFD connected with the inlet and outlet tubing, and the input and output optical fibers, and (c) zoomed-in view of the µOFD connected with the flowmeter.

Fig. 9 .
Fig. 9. Trend of the optical signal acquired by (a) VeroClear-µOFD during the slug passage, with reference values obtained by the water-filled (blue line) and air-filled (red line) microchannel and (b) PDMS-µOFD during the slug passage.(c) Trend of the flow signal in the tube.In the three plots, signal levels are associated with the video frames showing the slug passage.

Fig. 10 .
Fig. 10.VeroClear-µOFD transmission index for air-filled and waterfilled microchannel per input laser power.limit the capability of the µOFDs in the slug characterization (frequency passage, velocity, and length) based on the analysis of the signal trend (fluid level differentiation air-slug versus water-slug).

Fig. 11 .
Fig. 11.Trends and spectra of detected signals using the VeroClear-µOFD in the experimental condition with a laser power of 10 mW and an input flow rate of 0.1 ml/min (Exp-3).The filtered signals are overlapped with a reconstructed square wave (in red line) (a) for the optical acquisition in the micro-channel (side size 400 µm) and (c) for the flow-meter acquisition in the tube (diameter size 1 mm).The passage of water is labeled by T w and the air with T a .Spectrum of the optical signal (b) and flow signal (d).

Fig. 12 .
Fig. 12. Mean period of slug flow in the VeroClear-µOFD computed by the spectral analysis of (a) photodiode signals and (b) flowmeter signal, per input flow rates, and laser powers.
of slug flow detected by optical (T µch ) and flow (T tube ) signals are reported.

Fig. 13 .
Fig.13.µOFDs in comparison: performance of the VeroClear-µOFD and the PDMS-µOFD in the detection of the mean period of slug flow by the optical signal in the experimental condition using the power laser of 25 mW (Exp-4, Exp-5).

1 )
µSPT Material (Factor A)-Categorical factor varied among two levels (a = 2), which are PDMS and VeroClear resin.2) Flow Rate (Factor B)-Quantitative factor varied among three levels (b = 3) corresponding to {0.05, 0.1, 0.2} mL/min.The number of replications was fixed at n = 3, for a total of N = a × b × n = 18 experimental runs.The experimental plan is reported in Table

Fig. 14 .
Fig. 14.Trends of the optical signals acquired by {PH 1 ; PH 2 } using the VeroClear-µOFD in the experimental condition Exp-3 with a laser power of 10 mW and an input flow rate of 0.1 mL/min.(a) Filtered signal (black line) overlapped with the square-wave signal (red line).(b) Cross correlation between the optical signals of PH 1 and PH 2 to compute their time delay.

Fig. 15 .
Fig. 15.(a) Slug velocity values obtained for VeroClear-µOFD.(b) Comparison between the slug velocity values obtained using the VeroClear-µOFD and PDMS-µOFD in the experimental conditions Exp-4 and Exp-5 with a laser power of 25 mW.

Fig. 16 .
Fig. 16.(a) Slug velocity values obtained using the VeroClear-µOFD in Exp-1 compared with the slug velocity values obtained analytically inside the microchannel and in the tube per input flow rates.(b) Values of K v , the ratio between v µOFD and v µch .In red is the value expected computed by the Venturi relation.

Giovanna
Stella was born in Ragusa, Italy, in 1994.She received the master's degree in automation engineering and control of complex systems and the Ph.D. degree in systems, energy, computer and telecommunications engineering from the University of Catania, Catania, Italy, in 2019 and 2022, respectively.During the Ph.D. degree, she spent an Erasmus period for research activities with the Microfluidic Laboratory of the Institute Clement Ader, a Research Laboratory of the University of Toulouse (INSAT), Toulouse, France.Lorena Saitta was born in Basel, Switzerland, in 1993.She received the master's degree in mechanical engineering and the Ph.D. degree in systems, energy, computer and telecommunications engineering from the University of Catania, Catania, Italy, in 2019 and 2022, respectively.

TABLE II EXPERIMENTAL
CAMPAIGNS CARRIED OUT USING THE TWO µOFDS VARYING THE LASER POWER AND THE HYDRODYNAMIC PRESSURE V air = V water

TABLE III EXPERIMENTAL
PLAN FOR MICROPARTICLES SUSPENDED PROCESS: FACTORS AND LEVELS