Enhanced Laser‐Induced Graphene Microfluidic Integrated Sensors (LIGMIS) for On‐Site Biomedical and Environmental Monitoring

Abstract

The convergence of microfluidic and electrochemical biosensor technologies offers significant potential for rapid, in‐field diagnostics in biomedical and environmental applications. Traditional systems face challenges in cost, scalability, and operational complexity, especially in remote settings. Addressing these issues, laser‐induced graphene microfluidic integrated sensors (LIGMIS) are presented as an innovative platform that integrates microfluidics and electrochemical sensors both comprised of laser‐induced graphene. This study advances the LIGMIS concept by resolving issues of uneven fluid transport, increased hydrophobicity during storage, and sensor biofunctionalization challenges. Key innovations include Y‐shaped reservoirs for consistent fluid flow, hydrophilic polyethyleneimine coatings to maintain wettability, and separable microfluidic and electrochemical components enabling isolated electrode nanoparticle metallization and biofunctionalization. Multiplexed electrochemical detection of the neonicotinoid imidacloprid and nitrate ions in environmental water samples yields detection limits of 707 nm and 10^−5.4 m with wide sensing ranges of 5–100 µm and 10⁻⁵–10⁻¹ m, respectively. Similarly, uric acid and calcium ions are detected in saliva, demonstrating detection limits of 217 nm and 10^−5.3 m with sensing ranges of 10–50 µm, and 10⁻⁵–10^−2.5 m, respectively. Overall, this biosensing demonstrates the capability of the LIGMIS platform for multiplexed detection in biologically complex solutions, with applications in environmental water quality monitoring and oral cancer screening.

A laser‐induced graphene microfluidic with integrated sensors (LIGMIS) are demonstrated for practical use in agricultural and biomedical domains. The porous electrodes are modified with silver nanoplatelets and gold microstructured trees for increased electroactive surface area while ion electrodes are modified with solid‐contact membranes for analyte detection. The results discriminate between healthy and unhealthy samples, validating the utility of the LIGMIS.

1. Introduction

The convergence of microfluidic and electrochemical biosensor technologies is vitally important for in‐field and rapid analyte quantification for numerous applications including biomedical diagnostics and environmental monitoring due to their enhanced sensitivity and specificity, reduced sample and reagent volumes, portability for use in remote areas, and facile multiplexed sensing capabilities.^{[ 1 ]} Platforms like the micro total analysis system have evolved over three decades and encompass multiple chemical processing and sensing steps that can be performed on the same platform due to small‐scale channels or microchannels.^{[ 2 ]} Initial microfluidic designs were limited because they were typically crafted from silicon and glass using expensive techniques, such as photolithography, micromilling, and laser ablation, often necessitating fabrication by a third‐party manufacturer.^{[ 3 ]} Microfluidic sensors experienced burgeoning growth among researchers with the advent of soft lithography and poly(dimethylsiloxane) (PDMS)‐based microfluidics, which allowed researchers to quickly and cost‐effectively create microfluidics within their own laboratories.^{[ 4 ]} Although microfluidic systems created through soft lithography enabled advanced fluid manipulation, they were not suitable for scalable manufacturing and required pumps and valves to circulate fluids for sensing; traditional microfluidics are typically too costly for numerous in‐field monitoring and diagnostic applications. A turning point in the development of low‐cost microfluidics came when Whitesides and co‐workers presented a low‐cost, pumpless microfluidic for sensing in remote locations by developing a paper‐based microfluidic coined µPADs.^{[ 5 ]} However, the reliance on wax printing or photolithography to create hydrophobic channels hinders the scalable manufacturing of paper microfluidic sensors. Additionally, the nonconductive nature of paper necessitates interfacing metal or carbon electrodes with the paper to develop integrated electrochemical microfluidic systems, further reducing scalability and increasing costs.^{[ 6 ]} Moreover, paper microfluidics are typically single‐use devices as the cellulose‐based matrices can suffer from saturation, analyte retention, and biofouling, resulting in increased sensor costs and potentially increased environmental waste as compared to multiuse microfluidics.^{[ 7} , ^{8 ]} Hence, current integrated microfluidic/electrochemical biosensor systems typically fail to satisfy the “ASSURED” criteria determined by the World Health Organization for cheap diagnostic sensors in resource‐deficient settings.^{[ 9 ]} The ASSURED criteria require an affordable, sensitive, specific, user‐friendly, rapid and robust, equipment‐free, and deliverable or accessible device.

More recently, a new branch of microfluidic sensors has been demonstrated that we are naming laser‐induced graphene microfluidic integrated sensors (LIGMIS), with the promise to integrate both the electrical and microfluidic components of electrochemical microfluidic systems in a manner that meets most, if not all, of the ASSURED criteria. The concept of LIGMIS was introduced by our previous publication,^{[ 10 ]} in which both the microfluidics and connecting electrochemical electrodes were made of the same material, laser‐induced graphene (LIG). The practical nature of this concept is that both the microfluidics and the electrodes can be created simultaneously from the same standard CO₂ laser in ambient conditions on the same polyimide substrate. This one‐step, mask‐free process used different laser settings to create open microfluidic hydrophilic LIG tracks outlined with highly hydrophobic LIG walls that are capable of splitting and transferring fluid samples to electrically conductive LIG‐based electrodes for electrochemical sensing with both 2‐ and 3‐electrode configurations. The cross‐shaped LIGMIS pattern was used as proof of concept to split and transport fluid to three ion‐selective electrodes (ISEs) for potassium (K⁺), nitrate (NO₃ ⁻), and ammonium (NH₄ ⁺) monitoring and to an enzymatic pesticide sensor for organophosphate pesticide (parathion) monitoring in environmental water samples.^{[ 10 ]} Another research group followed this same LIGMIS cross‐shaped microfluidic pattern for the simultaneous monitoring of dopamine and uric acid in human serum samples.^{[ 11 ]} It is important to note that the concept of LIGMIS should not be confused with other LIG‐based systems that are integrated with polymeric microfluidics^{[ 12 ]} or transfer polyimide‐based LIG to polymeric microfluidics.^{[ 13 ]} Moreover, LIGMIS circumvents the need for photolithography with UV masks,^{[ 14 ]} deep reactive ion etching,^{[ 15 ]} or brush painting of a fluorine‐containing layer,^{[ 16 ]} which are common methods to create open microfluidics through wettability patterning on substrates besides paper.

In this paper, we address the limitations of the original LIGMIS approach to expand its utility to real‐world applications. First, one of the major drawbacks of the original cross‐pattern LIGMIS was that fluid transport did not flow evenly toward the sensors and consequently did not completely wet the four connected electrochemical sensors (see Movie S2 in the Supporting Information from Chen et al.^{[ 10 ]} and Video S1 in the Supporting Information from Chen et al.).^{[ 11 ]} In this work, this issue is circumvented by adding two exit reservoirs extending from each of the sensing wells via a Y‐shaped junction. These distal end exit reservoirs act as LIG pumps that assist in pulling fluid over and through the sensing wells to completely wet the wells in a similar fashion to how paper‐based capillary pumps pull fluid through paper microfluidics.^{[ 17 ]} Another major concern of the original LIGMIS was low shelf‐life due to the tendency of LIG to become more hydrophobic during dry storage, a phenomenon most likely caused by hydrocarbon contamination during dry storage as recently noted with LIG as well as other carbonaceous materials, such as highly ordered pyrolytic graphite.^{[ 18} , ¹⁹ , ^{20 ]} This change in surface wettability can render the LIG open microfluidic tracks inoperable as the hydrophilic tracks turn hydrophobic and impede fluid transport. To circumvent this hinderance, a hydrophilic polymer polyethyleneimine (PEI) was employed to modify the fluidic surface by using 1‐ethyl‐3‐(3‐dimethylaminopropyl)carbodiimide (EDC) chemistry, ensuring a stable, hydrophilic surface. Finally, another concern of the original LIGMIS design was that it was challenging to isolate the electrochemical sensors from the microfluidics during biofunctionalization as both components are positioned close to each other on the same polyimide swatch. Avoiding contamination of the microfluidics during modification with electrocatalysts (e.g., metallic nanoparticles) or functionalization with biorecognition agents (e.g., enzymes, antibodies, DNA/aptamers) proved challenging. This was especially difficult because protocols often require placing the sensor in a small, sealed container for mixing, sonication, or extended incubation to promote biofunctionalization and prevent evaporation, which could otherwise dry out the small, drop‐casted biorecognition agent solutions before incubation is complete.^{[ 21} , ^{22 ]} To address this concern, the LIG electrodes are separated from the microfluidics after fabrication and positioned face‐down on the sensing well during sensing. This configuration aligns with the LIGMIS philosophy of simultaneously creating both the sensors and microfluidics using LIG in a single step, while also enabling the user to functionalize the sensors in separate, sealed containers distinct from the microfluidics. Also, this configuration creates less waste than typical paper microfluidics as the LIG open microfluidics can be rinsed and reused. Hence, this innovative approach to LIGMIS advances the field toward achieving the ASSURED standards and overcoming the challenges associated with many of the current microfluidic biosensor technologies geared toward chemical/biological monitoring in remote locations.

Herein, the developed LIGMIS analyzed both environmental and biomedical samples to validate the performance and reliability of the device across varying sensing domains, ultimately aiming to validate its efficacy for real‐world deployment and widespread adoption. Unique metallic micro/nanostructures were electrodeposited onto the LIG surface to improve the electroactive surface area and increase sensitivity toward the target analytes. Interestingly, silver (Ag) nanoplatelets and gold (Au) microstructured trees can be grown on the LIG scaffold to increase the available edge sites that engage with the chemical targets, presenting a useful technique in nanoengineering graphitic surfaces. The neonicotinoid imidacloprid (IMD) and fertilizer ion NO₃ ⁻ were accurately quantified and detected simultaneously on the open microfluidic with detection limits of 707 nm and 10^−5.4 m, respectively. This enhanced LIGMIS for environmental monitoring offers a rapid, quantifiable test for monitoring IMD and NO₃ ⁻ contaminants in rivers, drinking water, and food. Moreover, the sensing of IMD and NO₃ ⁻ contamination is important as nontarget migration has been associated with long‐term health risks as well as impairment of aquatic invertebrates and honeybees.^{[ 23} , ^{24 ]} Similarly, the biomarkers, uric acid (UA) and calcium (Ca²⁺), were accurately detected with detection limits of 217 nm and 10^−5.3 m, respectively, with the enhanced LIGMIS platform. These biomarkers are important to the biomedical field, as various oral cancers can be diagnosed based on varying levels of UA and Ca²⁺.^{[ 25} , ^{26 ]} Moreover, fluid flow throughout the LIGMIS was observed and characterized using a numerical model based on the Washburn relationship to quantify the capillary‐driven transport dynamics and predict the flow behavior within the channels.^{[ 27 ]} All targets were validated in complex fluids (e.g., river water for agrochemicals and artificial saliva for biomarkers). This work enhances the utility of LIGMIS and demonstrates its potential for point‐of‐service biomedical and environmental sensing applications.

2. Results and Discussion

2.1. Fabrication and Detection Mechanism

Both the fluidic and biosensing portions of the LIGMIS platform are based on LIG created from the same CO₂ laser in a single manufacturing process. The multiplexed, open microfluidic device is capable of splitting and transporting a single fluid sample to two distinct LIG electrodes, as illustrated in Figure  1 . The LIG electrodes were made by laser irradiating the polyimide film using an Epilog Fusion M2 CO₂ laser at 15% speed, 7% power, and 2 mm defocus (Figure 1a–i). The laser irradiation of polyimide photothermally converts the precursor sp³ carbons into sp² hybridized carbons.^{[ 28 ]} After the lasing process, the LIG electrodes were finished with Ag leads to prevent contact damage, acrylic polish to guarantee a constant working area, and a silver/silver chloride (Ag/AgCl) pseudoreference electrode coated with a salt‐loaded polyvinyl chloride (PVC) membrane. Then, LIG electrodes were modified with nanomaterials and ion‐selective membranes (ISMs) for the detection of their corresponding target analytes (Figure 1a–ii). The open microfluidic patch was laser‐scribed in the desired pattern of the fluid flow during the same lasing process and using the same settings (Figure 1b–i). The laser settings created fluidic tracks with a porous structure and high wettability for enhanced fluid transport as further discussed in the Microfluidic Characterization section. To maintain hydrophilicity over extended periods of time, PEI was bonded to the surface of the LIG using EDC chemistry (Figure 1b–ii). Other reservoirs that are necessary for reliable fluid flow are shown (e.g., inlet, exit, and sensing reservoirs). The final device configuration (Figure 1b–iii) displays the attachment of the sensors over the open microfluidic. Figure 1c depicts the methods of detection for the target analytes NO₃ ⁻, IMD, Ca²⁺, and UA. The ion‐specific sensors for NO₃ ⁻ (Figure 1c–i) and Ca²⁺ (Figure 1c–iii) incorporate an ISM corresponding to each ion, which selectively interacts with the target ions while not engaging with other ionic species. Ag nanostructures (AgNSs) in the form of nanoplatelets were electrodeposited onto the LIG surface to catalytically detect IMD as they show an affinity toward reducing nitro‐containing aromatics (Figure 1c–ii); thus, AgLIG was used as an effective material to reduce the nitroguanidine group present in IMD, as shown in a previous report for detecting the neonicotinoid thiamethoxam.^{[ 29 ]} Gold nanostructures (AuNSs) with a tree‐like microstructure were electrodeposited onto LIG to directly oxidize UA into its quinoid diimine form (Figure 1c–iv).^{[ 30} , ^{31 ]} Images of an array of LIG electrodes, single microfluidic, and combined LIGMIS device are shown in Figure S1 (Supporting Information). In combining the LIG microfluidic with the various surface‐modified composites, an all‐in‐one LIGMIS platform is realized and operates as a plug‐and‐play system.

Figure 1.

Device fabrication scheme of the enhanced LIGMIS. a) Electrode fabrication through i) laser irradiation of the polyimide film to create LIG and ii) silver leads, Ag/AgCl pseudoreference electrode, acrylic passivation layer, and the corresponding NO₃ ⁻ ISM, AgNS, Ca²⁺ ISM, and AuNS superficial customizations; b) construction of fluidic through i) lasing the polyimide surface, ii) binding PEI onto the LIG surface to alter fluidic wettability, and iii) constraining the fluidic and sensing components into a 3D‐printed housing; and c) zoomed‐in view of LIG working electrode to monitor i) NO₃ ⁻ using NO₃ ⁻ ISM, ii) IMD using AgLIG, iii) Ca²⁺ using Ca²⁺ ISM, and iv) UA using AuLIG.

2.2. Microfluidic Characterization

LIG‐based materials have been underexplored as a platform for microfluidics, and their potential as an all‐in‐one multiplexed sensor system for both agro‐ and biomonitoring has yet to be fully realized. Some researchers have investigated various surface modifications and fabrication techniques to create LIG fluidic tracks, demonstrating their ability to transport small fluid volumes to targeted locations.^{[ 10} , ^{32 ]} However, it is important to highlight here that microfluidics in this work differ from the original cross‐shaped LIGMIS designs, where hydrophobic LIG sidewalls were used to guide fluid along hydrophilic tracks.^{[ 10} , ^{11 ]} Instead, we follow a more recent work in which we created purely hydrophilic LIG fern leaf patterned tracks to channel moisture to an apex.^{[ 33 ]} In this work, it was noted that menisci formed when first wetting the wedge‐shaped tracks with liquid quickly moving forward following Laplace pressure driven flow associated with open microfluidics while later in the track the fluid fully absorbed into the LIG, and fluid transport was much slower following capillary driven flow associated with paper microfluidics. Hence, in this work, we use smaller field sample sizes that become rapidly adsorbed by the LIG so that fluid flow is capillary driven flow that follows the Washburn model. To analyze the fluidic capabilities of the enhanced LIGMIS, fluid flow on an LIG strip was recorded at varying timestamps (Figure  2a). The flow modeling study was performed on tracks with widths of 0.6 mm. Fluid progression on porous media obeyed an analogous relationship to the Washburn model (Equation (1)) that relates the fluid front location (x) with the fluid viscosity (µ), surface tension (γ), and time of travel (t).^{[ 27 ]} Figure 2b indicated that as fluid propagated throughout the LIG, the distance traveled can be numerically modeled following a square root relationship with time. The numerical model was fitted to Equation (2), which collapses the surface tension and viscosity into coefficient D as described in a previous report.^{[ 34 ]} Average coefficient values of 12.8 ± 0.4 mm s^−1/2 were obtained, which highlights the reproducibility of the numerical model. The longevity of the device was studied over a 5‐day period (Figure 2c). The hydrophilic nature of LIG was confirmed, as shown in Figure 2c inset, with average contact angles of 29.5° ± 5.9° and 16.6° ± 2.3° for bare LIG and PEI–LIG, respectively. The variable surface wettability of the unmodified LIG was demonstrated by a decay in flowrate over the 5‐day period, resulting in no fluid progression on day 5. It is believed that the nature of the LIG transitions with respect to its wettability and becomes hydrophobic over the course of days due to potential hydrocarbon contamination, as previously mentioned, which explains the inability of fluid to flow by day 5.^{[ 18} , ^{20 ]} Conversely, the PEI‐modified track produced stable and consistent flowrates over the same period as it rendered the LIG hydrophilic over the course of this experiment. The longevity study agrees with the reported use of PEI as a hydrophilic polymer for drug delivery.^{[ 35 ]} Transport throughout the complete open microfluidic design is shown in Figure 2d‐i–iv, demonstrating the ability to use LIG for multichannel flow. The reported LIGMIS characterization introduces significant contributions to the microfluidic field and larger domain of point‐of‐care devices that require small sample volume transportation.

$$x\sim {\left(\frac{\gamma }{\mu }\right)}^{1/2}{t}^{1/2}$$ (1)

$$x\sim D{t}^{1/2}$$ (2)

Figure 2.

Fluid transport characterization. a) Time lapse (seconds) overlayed images converted to black and white to improve visualization of fluid flow on hydrophilic PEI–LIG track, b) numerical model (n = 4 samples) for 0.6 mm track widths, c) flowrate longevity on bare LIG and PEI–LIG over 5 days (n = 3), and d) i–iv) overhead timelapse (seconds) black‐and‐white images of open microfluidic platform.

2.3. Material Characterization

A suite of techniques was employed to investigate the morphological and chemical properties, including scanning electron microscopy (SEM), Raman spectroscopy, and X‐ray photoelectron spectroscopy (XPS). Images of bare LIG (Figure  3a–c), AgLIG (Figure 3d–f), and AuLIG (Figure 3g–i) depicted the various morphological structures. Bare LIG exhibited a porous and weblike architecture conducive to wetting and capillary wicking flow similar to other reports that have harnessed the properties of porous media for time‐dependent fluid models.^{[ 36 ]} The electrodeposited AgNSs formed plate/flower‐like clusters, while the AuNSs appeared urchin/tree‐like. The growth of the Ag plate/flower‐like clusters (Figure 3e,f) typically propagated from a nucleation point near edge defects of the LIG, which is supported by other reports that have decorated LIG with AgNSs using electrodeposition,^{[ 37} , ^{38 ]} drop‐coating (nanoparticles synthesized from the Lee–Meisel method),^{[ 39 ]} and doped polyimide with Ag salts.^{[ 40 ]} Overall, homogeneously distributed clusters were observed for both nanostructures. The LIG topography with Ag plate/flower‐like clusters likely increased the available interfaces to electrochemically engage with target probes and analytes, which is characterized by improved charge transfer and reaction kinetics, as later discussed in the Electrochemical Characterization section.^{[ 40 ]} In the case of AuNS LIG, urchin/tree‐like structures with rougher shapes were observed similar to another report that electrodeposited AuNSs onto LIG for detecting Her‐2.^{[ 41 ]} Since the electrodeposition potential for AuNS (−0.7 V vs Ag/AgCl) is significantly lower than the potential used for electrodepositing AgNS (slightly lower than 0.2 V vs Ag/AgCl) with respect to their theoretical reduction potentials, more nucleation sites developed and grew into rougher, tree‐like structures compared to the plate clusters. When AuNSs are deposited at a significantly lower potential relative to AgNS, a higher nucleation rate occurs, leading to increased surface coverage and the formation of rough, fractal‐like morphologies rather than smooth platelets. The high driving force at the more negative deposition potential for Au (−0.7 V) enhances the rate of ion reduction, favoring dendritic and irregular growth patterns over the more controlled and clustered deposits seen in AgNSs. These findings align with previous observations that the applied potential plays a critical role in governing the nucleation density and growth mechanism of nanostructures.^{[ 42} , ^{43 ]} Zhu et al.^{[ 44 ]} developed a AuNS LIG electrode and noted that the nanoparticles increased the roughness factor of the material, which provided more accessible sites to detect glucose. Other researchers have demonstrated AuNS modifications by coating the precursor polyimide film with Au‐salt‐loaded chitosan membranes prior to laser irradiation, which produced nanoparticles with remarkably low sizes that ranged from 9.6 to 15 nm.^{[ 45} , ^{46 ]} Although techniques that create nanoparticles through laser irradiation boast fine nanoparticle sizes, these works require the preparation of salt‐loaded polymers, which are arguably more arduous to fabricate compared to electrochemical plating solutions.

Figure 3.

Material characterization. a–c) Optical images and SEM images with 1000× and 10 000× magnification of bare LIG, d–f) plate/flower‐like Ag micro/nanoparticles electrodeposited on an LIG electrode, and g–i) Au urchin/tree‐like micro/nanoparticles electrodeposited on an LIG electrode; j) XPS and k) Raman spectroscopy of bare LIG, LIG treated with PEI (PEI–LIG), and LIG with electrodeposited Ag and Au micro/nanostructures (AgLIG and AuLIG).

The chemical composition of all materials was investigated by using XPS and Raman spectroscopy. The observed XPS peaks identified the various elemental components (Figure 3j). The C 1s and O 1s shells were identified in all the materials near 284.6 and 531.9 eV, respectively. The O 1s peak likely derives from material oxidation that is induced by the laser irradiation of film in ambient environments. As reported in the Supporting Information, the C 1s peak was further deconvoluted to reveal sp² carbon bonds (284.5 eV), epoxy (286.2 eV), carbonyls (288.1 eV), carboxyls (289.0 eV), and π–π interactions (291.4 eV) (Figure S2f, Supporting Information). The N 1s peak (400.8 eV) was notable in the PEI–LIG sample, indicating bound ─NH₂, which confirmed the successful attachment of PEI as a surface modifier for sustained hydrophilicity. The Ag 3d location (Figure S2d, Supporting Information) contained both the Ag 3d_5/2 (368.1 eV) and 3d_3/2 (374.5 eV) orbital shells and indicated metallic particles embedded on the LIG surface just as others have reported for AgLIG electrodes.^{[ 47} , ^{48 ]} The Au 4f location (Figure S2e, Supporting Information) contained the Au 4f_7/2 (84.4 eV) and 4f_5/2 (88.2 eV) shells similar to other studies that investigated Au‐ and graphene‐based nanocomposites, which verified the metallic state of Au.^{[ 45} , ^{49 ]} Raman spectroscopy (Figure 3k) showed the D (1338 cm⁻¹), G (1576 cm⁻¹), and 2D (2679 cm⁻¹) peaks in all samples, which are characteristic of graphitic materials. Each band clarifies the structure and chemical composition of LIG as the D band indicates defects, G band identifies in‐plane vibrations, and 2D band corresponds to second order zone‐boundary phonons.^{[ 50 ]} The ratios of the band intensities can be calculated between the D and G bands (I _D/I _G) to estimate the defect density. The ratio between I _D/I _G for the LIG, PEI–LIG, AgLIG, and AuLIG materials produced values of 0.9, 0.9, 1.0, and 1.1, respectively. Based upon these values, the surfaces modified with Ag and Au nano/micromaterials possessed a slighter degree of defects, which was consistent with the visual analysis (SEMs) of the materials. The presence of nano/microstructures on carbon‐based materials effectively increased the structural defects as the structures were typically observed at the LIG edge defects. Due to the interaction of these structures at the LIG edge defects, which govern the D band, the carbon lattice became more disordered and increased the I _D/I _G ratio. Additionally, comparing the 2D and G band intensities (I _2D/I _G) can quantify the graphene layers. The I _2D/I _G ratios for LIG, PEI–LIG, AgLIG, and AuLIG were calculated as 0.60 ± 0.08, 0.68 ± 0.07, 0.6 ± 0.1, and 0.48 ± 0.09, respectively. An I _2D/I _G ratio less than 1 indicates a few to multilayer surface.^{[ 51 ]}

2.4. Electrochemical Characterization

The electroactive surface area (ESA) and electrocatalytic properties of the modified LIG materials were characterized electrochemically to assess the nanostructured materials and biosensing capabilities. Cyclic voltammetry (CV) was used to investigate the charge transfer at the surface of LIG, AuLIG, and AgLIG during the reduction and oxidation process of [Fe(CN)₆]^3−/4− (Figure  4a). The higher peak current observed from the AuLIG electrode demonstrated superior performance compared to the bare LIG electrode. With respect to the AgLIG electrode, a plethora of anodic peaks were observed between 0 and 0.6 V, capturing three possible oxidation processes. A pronounced anodic peak emerged near 0.3 V, with an abnormal shoulder protruding from this peak followed by a second anodic peak at 0.45 V. The more defined peak at 0.3 V would typically indicate the oxidation peak of [Fe(CN)₆]⁴⁻ as evidenced by the CV scans performed on the AuLIG and bare LIG electrodes. However, we hypothesize that this is the direct oxidation of Ag⁰ to Ag₂O and/or AgOH as some have reported.^{[ 52 ]} Other researchers have asserted that this behavior is due to the various oxidation states of Ag in this potential range, which makes it difficult to estimate an ESA;^{[ 53 ]} therefore, the ESA was not calculated for AgLIG. The utility in calculating ESA reveals the ability to better access chemical species through enhanced surface area.

Figure 4.

Electrochemical characterization. a) Representative CV comparison of LIG, AuLIG, and AgLIG electrodes in 5 mm [Fe(CN)₆]^3−/4− with 0.1 m potassium nitrate (KNO₃) at 50 mV s⁻¹, b) CV of LIG in 5 mm [Fe(CN)₆]^3−/4− with 0.1 m KNO₃ at varying scan rates, c) CV of AuLIG in 5 mm [Fe(CN)₆]^3−/4− with 0.1 m KNO₃ at varying scan rates, d) the corresponding Randles Sevcik plots for n = 3 electrodes each, e) representative DPV comparison of LIG and AgLIG electrodes in 10 µm IMD, and f) representative CV comparison of LIG and AuLIG electrodes in 200 µm UA.

For both LIG and AuLIG, the ESA was calculated by performing CVs at varying scan rates in [Fe(CN)₆]^3−/4− (Figure 4b,c). The Randles–Sevcik plot (Figure 4d) displayed linearity in peak currents for both LIG and AuLIG as a function of the square root of the scan rate, indicating a diffusion‐controlled process.^{[ 54 ]} However, the peak‐to‐peak separation (ΔE _p) of both the LIG and AuLIG electrodes drifted with increasing scan rates and exceeded the reversible ΔE _p of 60 mV (117.9–337.5 mV); therefore, the electrodes were considered quasireversible. Thus, the Randles–Sevcik equation (Equation (3)) for quasireversible electrodes was used to estimate the ESA (A) by relating the peak current (i _p), diffusion coefficient (D), electrons transferred (n), scan rate (v), and redox species concentration (C).^{[ 55 ]} The acquired ESAs of the LIG and AuLIG were revealed to be 5.4 ± 0.5 and 7.9 ± 1.2 mm², constituting 172.6% and 251.2% of the geometric surface area. Such ESAs highlight the efficacy of LIG and nanostructured‐modified LIG and are explicated by the rich presence of edge site defects that provide a more readily available scaffold for engaging with redox species.

The electrodes were further tested in the presence of the target analytes and compared to bare LIG to reinforce the justification for incorporating nanostructured materials. Differential pulse voltammograms (DPVs) were taken of the bare LIG and AgLIG electrodes in 10 µm IMD (Figure 4e). A prominent reduction peak was observed near −1.15 and −1.05 V for the LIG and AgLIG electrodes, respectively. The benefits of the AgLIG are twofold compared to the bare LIG: the increased cathodic current and an improved overpotential, which revealed the efficiency of the AgLIG material at reducing IMD. In a similar manner, scans were taken of AuLIG and LIG in 200 µm UA (Figure 4f). The electrochemical fingerprint of UA was identified with the distinct oxidation peak near 0.2 V where the AuLIG electrode outperformed the bare LIG, suggesting that the increased ESA associated with the AuNSs improved electrochemical accessibility to UA.

(3)

2.5. Electrochemical Sensing

The DPV and open circuit potentiometry (OCP) techniques were executed to quantify agrochemicals and biomarkers. Detailed images of the 3‐electrode and 2‐electrode setups as well as the cable connections are shown in Figure S3 (Supporting Information). Voltametric sensors employed an all‐LIG, 3‐electrode setup, which included a pseudo‐Ag/AgCl reference, LIG counter, and nanoparticle‐modified LIG working electrode. The ion sensors that utilized OCP were designed in a 2‐electrode configuration, which included the pseudoreference as well as an LIG working electrode that was functionalized with the corresponding ISM. It is noteworthy to mention that the previous LIGMIS by Chen et al.^{[ 10 ]} did not include the pseudo‐Ag/AgCl reference and used an external Ag/AgCl electrode instead.

The AgLIG device performed DPV scans toward more negative potential, at which point, IMD was reduced, and a negative current was read. For increasing concentrations of IMD, peaks progressed toward more negative currents (Figure  5a). As previously discussed, the nitroguanidine group in IMD was reduced, and thus, the parent molecule gained electrons.^{[ 29 ]} The characteristic peak that is associated with this reduction was observed near −0.92 V. The IMD sensor displayed a linear sensing range between 5 and 100 µm (Figure 5b) and achieved a limit of detection (LOD) of 707 nm. The fluctuation and stability of the electrodes during calibration were monitored by observing the blank scans in phosphate buffer saline (PBS) without the target analyte. This baseline calibration is represented by the control data in Figure 5b, which indicated negligible baseline signal variation when compared to the expected signal and averaged 3.4% of the expected signal. The stable baseline signal confirmed that the change in current near −0.92 V is solely due to the direct reduction of IMD. In this electrochemical event, the electroactive group within imidacloprid, nitroguanidine, follows a 2‐step reduction where the nitro group is reduced to hydroxylamine followed by a final reduction to amine.^{[ 29} , ^{56 ]} The 2‐step reduction became more apparent at higher concentrations where two peaks were present at −0.92 and −1.05 V.

Figure 5.

Agrochemical sensor calibration. a) Representative DPV plot for increasing concentrations for IMD in 10× PBS, b) calibration plot presenting the change in current with respect to the IMD concentration with a linear range from 0 to 100 µm in 10× PBS (n = 5, R _adj ² = 0.9004, p _model < 0.001), c) representative OCP for half‐log step additions of NO₃ ⁻ in deionized (DI) water, and d) calibration plot presenting the change in OCP with half‐log step additions of NO₃ ⁻ with a linear range from 10⁻⁵ to 10⁻¹ m (n = 4, R _adj ² = 0.9626, p _model < 0.001).

The NO₃ ⁻ target was quantified by coupling its corresponding ISE with OCP. For increasing concentrations of NO₃ ⁻, decreasing OCPs were observed as the negatively charged ions were selectively read due to the interaction with the ISM (Figure 5c). The NO₃ ⁻ sensors displayed a linear range with a strong correlation within 10⁻⁵–10⁻¹ m and high reproducibility (Figure 5d) while yielding a sensitivity of −61.2 mV dec⁻¹ and LOD of 10^−5.4 m (0.27 ± 0.07 ppm). Both IMD and NO₃ ⁻ sensors can be used for repeated use to accurately monitor agrochemical concentrations, thus indicating that such sensors can be used multiple times or as a one‐time disposable sensor (Figure S4a,b, Supporting Information). The IMD sensor stability was further investigated by performing multiple scans at the same concentration to verify a consistent signal (Figure S5a, Supporting Information), which showed a minimal signal fluctuation of 4.8%. The calibrated models for IMD and NO₃ ⁻ establish a well‐defined linear range that is suitable for environmental monitoring as IMD concentrations have been noted to reach 17.08 mg L⁻¹ (67 µm) in agricultural runoff, and NO₃ ⁻ concentrations might exceed the maximum contaminant levels in drinking water of 10 ppm (0.16 mm).^{[ 57} , ^{58 ]}

Similarly, DPV scans were performed on that AuLIG electrode in order to quantify UA by performing oxidizing UA, resulting in a loss of electron and transformation into its quinoid diimine form.^{[ 30} , ^{31 ]} As the forward scan traversed beyond 0.2 V, UA relinquished electrons, and an anodic peak emerged (Figure  6a). For increasing concentrations of UA (10 to 200 µm), the current increased consistently in the positive direction, displaying a linear range from 10 to 50 µm (Figure 6b) with an LOD of 217 nm. Furthermore, these electrodes were compared to the blank scans to monitor signal drift over the course of testing in the absence of the target molecule, which is depicted by the control in Figure 6b. These results highlighted the stability of the UA sensors as the baseline control constituted an average of 1.8% of the expected signal.

Figure 6.

Biomedical sensor calibration. a) Representative DPV plot for increasing concentrations of UA in 10× PBS, b) calibration plot presenting the change in current versus UA concentration with a linear range from 10 to 50 µm (n = 4, R _adj ² = 0.9563, p _model < 0.001) in 10× PBS, c) representative OCP with half‐log step additions of Ca²⁺ in DI water, and d) calibration plot presenting the change in OCP with half‐log step additions of Ca²⁺ with a linear range from 10⁻⁵ to 10^−2.5 m (n = 4, R _adj ² = 0.9185, p _model < 0.001).

The Ca²⁺ target was quantified by coupling its corresponding ISE with OCP. As the concentration of Ca²⁺ increased, the positive charge on the surface of the electrode increased; the positively accumulated charge induced a positive change in the OCP (Figure 6c). The sensors operated within a linear range of 10⁻⁵–10^−2.5 m, which yielded a sensitivity of 22.8 mV dec⁻¹ and LOD of 10^−5.3 m. The acquired figures of merit indicated near‐Nernstian behavior for a divalent cation and a suitable LOD for detecting Ca²⁺ in saliva. Both calibrated models appeal to biomedical diagnostics as salivary UA varies between 184.9 to 278.1 µm among healthy and unhealthy patients, and salivary Ca²⁺ varies from 0.6 to 0.8 mm among dentulous and edentulous patients.^{[ 59} , ^{60 ]} For concentrations beyond the linear range of the sensors, saliva samples could be diluted to a suitable concentration. The reusability of each sensor was investigated (Figure S4c,d, Supporting Information), indicating sustained and reliable use for n = 4 cycles. The UA sensor stability was further investigated by performing multiple scans at the same concentration to verify a consistent signal (Figure S5b, Supporting Information), which showed a minimal signal fluctuation of 1.9%.

2.6. Interference Testing

The specificity of both the IMD and NO₃ ⁻ electrodes were investigated against common interferent species. Interferent data comparisons for IMD and NO₃ ⁻ sensors were represented as the average response ± standard deviation for n = 4 sensors. To test the agrochemical insecticide specificity of the IMD sensor, the AgLIG electrode was screened against molecules often found in agricultural environments including: the herbicides glyphosate, atrazine, and dicamba; the similarly structured insecticide chlorpyrifos; and fertilizer ion NO₃ ⁻. The responses to 40 µm of pesticide interferent species and 100 µm of ionic species were reported as the observed current changes at −0.92 V (Figure S6a, Supporting Information) and were compared to the reference current change of 40 µm IMD at the same potential location (Figure  7a). All possible interferences were designated as negligible since their signals represented 0.2–3.0% of the expected response to IMD. The findings indicate that none of these species are considered electroactive near the reduction potential of IMD. The NO₃ ⁻ sensor was tested against possible interferent ions (Cl⁻, NO₂ ⁻, CO₃ ²⁻, SO₄ ⁻², PO₄ ⁻³) using the fixed interferent method^{[ 61 ]} (Figure S6b, Supporting Information) and yielded the logarithm of the selectivity coefficients (log $K_{A,B}^{\mathrm{pot}}$), which were compiled in Table 1 ; lack of interference is indicated by more negative selectivity coefficients.^{[ 62 ]} Generally, the acquired selectivity coefficients indicated minimal interference similar to another report that detected NO₃ ⁻ in lake water.^{[ 63 ]} However, Cl⁻ and SO₄ ²⁻ yielded coefficients of −1.17 and −0.88, respectively, which present slight interference. It is possible to mitigate such interferences through desalination or reverse osmosis.^{[ 64 ]} Nevertheless, the reported data demonstrated that LIG can be functionalized and used for agricultural monitoring.

Figure 7.

Interference testing. Sensor response comparisons of the a) AgLIG electrode in IMD as a reference compared to other agrochemicals (n = 4) and b) AuLIG electrode in UA as a reference compared to other biomarkers. Data represent mean ± standard deviation (n = 4).

Table 1.

Ionic interferent data comparing the log of the selectivity coefficients for select ionic interferent species commonly found in river water and saliva (n = 4).

Target	Ion, j	$\log K_{\mathrm{target},j}^{\mathrm{POT}}$
NO3 −	Cl−	−1.17 ± 0.08
	NO2 −	−2.42 ± 0.08
	HCO3 −	−2.16 ± 0.10
	SO4 2−	−0.88 ± 0.04
	H2PO4 2−	−1.77 ± 0.12
Ca2+	Mg2+	−3.43 ± 0.87
	Na+	−1.73 ± 0.75
	K+	−3.30 ± 0.97

The biomarker sensors were screened against possible interferences present in salivary samples. The specificity of the UA sensor was investigated by scanning the AuLIG electrode in the presence of common salivary metabolites and ions including glucose, glycine, ascorbic acid, and Ca²⁺. The responses to 50 µm metabolite interferent species and 100 µm ionic species at 0.2 V were recorded (Figure S6c, Supporting Information) and compared to the expected response of 50 µm UA at the same potential location (Figure 7b). The metabolic and ionic interferences were considered negligible as their normalized signal represented minute percentages (0.6–3.0%) of the expected signal. Though it is established within the literature that ascorbic acid and glucose can be electrochemically read at low potentials, the peak occurring at 0.2 V in this work is solely attributed to the oxidation of UA.^{[ 65} , ^{66 ]} The interference studies in the works that detect ascorbic acid and glucose indicate lack of interference from UA, which further confirm that all three molecules are oxidized at different potential locations. The Ca²⁺ ISE was validated against common cations found in saliva including Mg²⁺, Na⁺, and K⁺ (Figure S6d, Supporting Information). These species were evaluated using the separate solution method^{[ 67 ]} and produced values for the logarithm of the selectivity coefficients (logK), which are reported in Table 1. Overall, minimal interference was observed, and the affinity of the Ca²⁺ ISM was validated.

2.7. Multiplexed Sensor Validation

To demonstrate the efficacy and practical application of the agricultural and biomedical devices, the sensors were combined with the open microfluidic to create the LIGMIS device and quantify spiked samples. The agricultural device combined the IMD and NO₃ ⁻ electrodes to properly quantify spiked river water samples. Two distinct solutions were prepared with known concentrations of both targets. The solutions were characterized by excess concentrations of both targets (solution A) and a moderate concentration of IMD with an excess concentration of NO₃ ⁻ (solution B). A small sample volume of 75 µL was introduced to the inlet of the microfluidic. Based upon the fluidic model previously discussed, scans were not performed until 20 s, at which point, the sensing reservoirs were fully wetted, and the exit reservoirs were operational. The responses were recorded in triplicates for each solution and were assessed using the calibrated models from Figure 5b for IMD and Figure S7a (Supporting Information) for NO₃ ⁻. As the sample traveled across each sensor, the appropriate DPV and OCP scans were performed (Figures S8 and S9, Supporting Information). As river water maintains ionic strength due to the various ions present, the IMD calibration in 10× PBS (Figure 5b) was used as an analogous predictive model. Since the NO₃ ⁻ ISE presents slight interferences from Cl⁻ and SO₄ ²⁻, a modified sensing model was developed by recalibrating ISEs in river water to account for the complex matrix (Figure S7a, Supporting Information). Using the models, the found concentrations could be calculated and were compared to the spiked concentrations (Figure  8a). Solution A was spiked with 1.6 mm NO₃ ⁻ and 100 µm IMD, and concentrations were detected with predicted concentrations of 1.85 ± 0.06 mm and 89.9 ± 16.3 µm, which yielded recovery percentages of 115.6% and 89.9%, respectively. Solution B was spiked with 0.16 mm NO₃ ⁻ and 50 µm IMD, presenting predicted concentrations of 0.36 ± 0.14 mm and 46.3 ± 7.0 µm, which yielded recovery percentages of 222.7% and 92.5%. In general, the found concentrations of solutions A and B were satisfactory with the exception of 0.16 mm NO₃ ⁻ in solution B. It was apparent that by detecting a diluted concentration of NO₃ ⁻, the Cl⁻ and SO₄ ²⁻ molecules present in river water competed against NO₃ ⁻ and induced an overprediction of NO₃ ⁻. However, the overall recovery percentages highlight the ability to use the open microfluidic device and electrodes for on‐site agricultural analysis.

Figure 8.

Complex fluid validation on the LIGMIS device. a) Agrochemical detection in spiked river water and b) biomedical analyte detection in spiked saliva. Data represent mean ± standard deviation (n = 4). The true concentrations are indicated by the written concentrations within each bar and the dashed lines that correspond to those written concentrations.

In a similar manner, the biomedical device integrated the UA and Ca²⁺ electrodes to analyze spiked saliva samples. Two distinct solutions with varying concentrations of UA and Ca²⁺ were prepared. The solutions were characterized by excess concentrations of both targets (solution A), which could indicate oral diseases (e.g., halitosis, periodontitis)^{[ 68 ]} and healthy concentrations of both targets (solution B). Again, 75 µL of the solution was injected at the inlet of the microfluidic, and measurements were performed 20 s after fluid injection. Multiple responses were recorded for each solution, and the found concentrations of each were calculated from the complex media models. The saliva matrix hindered the Ca²⁺ sensitivity compared to the calibrated model in deionized (DI) water, so a complex model in saliva was created (Figure S7b, Supporting Information). Additionally, saliva has a lower ionic strength than 10× PBS, which necessitated that the UA sensors were recalibrated in saliva showing a linear range of 50–300 µm (Figure S7c, Supporting Information). The found concentrations using the calibration models were compared to the spiked concentrations (Figure 8b). Solution A was spiked with 1 mm Ca²⁺ and 300 µm UA. The predicted concentrations were calculated as 0.98 ± 9 × 10⁻⁷ mm and 332.7 ± 48.6 µm, corresponding to recovery percentages of 97.9% and 110.9% for each respective target. Solution B was spiked with 0.5 mm Ca²⁺ and 180 µm UA, and the LIGMIS device accurately predicted 0.42 ± 3 × 10⁻⁷ mm and 229.1 ± 56.7 µm, yielding recovery percentages of 84.5% and 127.3%, respectively. The appropriate DPV and OCP scans are shown in Figures S10 and S11 (Supporting Information) for both solutions A and B. The complex media data validated the ability to use the LIGMIS device for point‐of‐care biomedical diagnostics. As a final experiment to prove the robustness of the electrode system, a bend cycle analysis was performed on the AgLIG and AuLIG electrodes and resulted in sustained conductivity with only a 12.4% and 22.7% increase in the 2‐point resistance by the 10th cycle, respectively (Figure S12, Supporting Information). This validation highlights the durability of the Au‐ and AgLIG electrodes, further bolstering their utility in field and biomedical analysis. It is important to note that these sensors will most likely operate on‐plane with the microfluidic within a housing and therefore would not experience the harsh conditions of the bending cycles (as performed in Figure S12 in the Supporting Information).

Microfluidics with integrated multiplexed sensing emerge as a strong candidate for point‐of‐care and lab‐on‐chip analysis. Various works have similarly established multiplexing microfluidic systems for environmental analysis of nitrate/nitrite ions, metal contaminants, and water pollutants through pump‐driven devices or µPADs.^{[ 69} , ⁷⁰ , ^{71 ]} Though cost effective and promising, these works are either limited by additional mechanical components (e.g., pumps) or employ colorimetric sensors that lack reusability. These aspects can be circumvented by capillary‐driven flow on LIGMIS devices that further utilize reusable LIG‐based sensors. Table  2 compares the works that have incorporated multiplex electrochemical systems into microfluidics that detect the agrochemicals and biomarkers reported in this work. Materials like PDMS, resin, and adhesives occupy the bulk of microfluidic literature, and these materials usually require a precursor material to be molded or 3D printed before shaping the microchannels. The polymeric‐styled fluidics are often combined with additional substrates and integrate other carbon‐based electrodes (e.g., glassy carbon, screen‐printed, carbon nanotubes). Paper‐based fluidics showcase cost‐effectiveness and possess the necessary properties for fluid transport but typically involve surface modifications achieved through printing hydrophobic polymers to act as a boundary along the fluidic track. Interestingly, there is a dearth of works that employ LIG‐based microfluidics with integrated electrochemical sensors for multiplex sensing. The research of Chen et al.^{[ 10 ]} began the groundbreaking work of fabricating open microfluidics with LIG by tuning the surface wettability based upon the laser settings, which interestingly affected the oxygen functionality of the surface. Nevertheless, the all‐LIG fluidics and electrodes were investigated separately as a proof of concept and failed to monitor targets on a fully integrated device. Finally, the flowrate stability and device longevity were not experimented, and the target analytes were limited to agrochemicals alone. Liu et al. developed an LIG‐based digital microfluidic, which impressively manipulated fluid droplets by moving, splitting, and mixing using LIG driving electrodes.^{[ 72 ]} Another group has recently investigated a similar approach by fabricating an LIG microfluidic that was paired with a UA and dopamine multiplex sensor.^{[ 11 ]} Though this application directly detected both targets on the microfluidic, microfluidic longevity was not explored, and the report was restricted to the biomedical domain. In comparison, the enhanced LIGMIS device in this study improves upon these works and allows the user to biofunctionalize/plate LIG sensors off‐board without contaminating the microfluidic, evenly wets the electrode surfaces through the addition of wells and reservoirs, and operates the device at extended time periods due to increased shelf‐life. In summary, neither of the previous works developed a microfluidic model that considered the wetting properties of LIG nor did either report or explore the longevity of the microfluidic, which our work has demonstrated that a hydrophilic polymer coating is necessary for sustained wettability and more accurately addresses the shelf‐life of such a device. A more impactful achievement of this work was the direct validation of healthy and unhealthy levels of the targets in spiked river water and saliva matrices.

Table 2.

Comparison table of integrated multiplexed electrochemical sensor and microfluidic systems for agrochemical and biomedical monitoring.

Target	Microfluidic material	Fabrication	Electrode material	Sample	LOD	Refs.
IMD	LIG	Laser engraved	LIG	River water	707 nm	This work
NO3 −	LIG	Laser engraved	LIG	River water	10−5.07 m	[10]
NO3 −	LIG	Laser engraved	LIG	River water	10−5.4 m	This work
UA	PDMS	3D printed	MWCNTs	Saliva	62.5 µm	[73]
UA	Adhesive	Laser engraved	LIG	Saliva, sweat	740 nm	[74]
UA	Paper‐based	Cutter printer	Graphene QDs	Urine	8.4 nm	[75]
UA	Photosensitive resin	3D printed	SPE	Serum	450 nm	[76]
UA	PDMS	Laser engraved	Glassy carbon	NA	NA	[77]
UA	Paper‐based	Cutter printer	Carbon ink	Urine	12 µm	[78]
UA	LIG	Laser engraved	LIG	Serum	48 nm	[11]
UA	LIG	Laser engraved	LIG	Saliva	217 nm	This work
Ca2+	PDMS	3D printed	Graphene composite	NA	NA	[79]
Ca2+	PDMS	Cut	Pt/Ti	Serum	1 nm	[80]
Ca2+	Photosensitive resin	3D printed	MWCNTs	Urine	NA	[81]
Ca2+	LIG	Laser engraved	LIG	Saliva	10−5.3 m	This work

IMD, imidacloprid; UA, uric acid; LIG, laser‐induced graphene; PDMS, polydimethylsiloxane; MWCNTs, multiwall carbon nanotubes; QDs, quantum dots; SPE, screen printed electrode.

3. Conclusions

The findings of this study demonstrate the utility of LIGMIS as a rapid, in‐field, multiplexed biosensing platform for biomedical and agricultural applications. This work introduces a scalable sensing solution through a reproducible, simple, and high‐throughput fabrication process. The platform is realized using a resource‐efficient laser‐scribing technique, which eliminates the need for traditional graphene synthesis methods and materials. By irradiating polyimide film with a laser, conductive and wettable LIG layers are created through the conversion of sp³‐ to sp²‐hybridized carbons, resulting in a porous, defect‐rich structure ideal for fluid transport. The study highlights the power of multiplexed sensing and open microfluidics, successfully demonstrating the simultaneous detection of IMD and NO₃⁻ in environmental samples, as well as UA and Ca²⁺ in saliva. The sensors were validated in real‐world samples, confirming the platform's effectiveness in detecting target analytes in surface water and saliva, even in the presence of common interferents.

From a broader perspective on the emerging field of LIGIMS, this work addresses three key challenges of the original LIGMIS design. First, the issue of uneven fluid transport to sensors was resolved by incorporating Y‐shaped exit reservoirs and wells, ensuring consistent fluid flow and coverage. Second, the challenge of decreased microfluidic shelf life—due to the increasing hydrophobicity of LIG during storage—was overcome by coating the LIG tracks with PEI, which preserved their hydrophilic nature over time. Third, the difficulty in isolating and preparing electrochemical sensors from the microfluidics during biofunctionalization was addressed by separating the microfluidic and electrochemical components after fabrication, allowing for isolated metallization and biofunctionalization of the LIG sensors.

Overall, this study introduces and enhances an LIG‐based, pump‐free, PDMS‐free microfluidic platform (i.e., LIGMIS) that integrates functionalized sensors with microchannels for fluid transport, offering a scalable, low‐cost solution for field‐deployable electrochemical sensing. LIGMIS represents a new branch of microfluidics that overcomes the limitations of traditional paper‐ and PDMS‐based systems by utilizing LIG to create both microfluidic channels and electrochemical sensors in a single step. This capillary‐driven, pump‐free system simplifies fabrication and operation, making it ideal for low‐cost, in‐field multiplexed biosensing applications. Furthermore, the study establishes a numerical model based on the Washburn relationship for capillary‐driven flow, providing a foundation for future research to develop an analytical model for fluid flow through LIG porous media. This innovative approach marks a significant advancement in the development of point‐of‐care diagnostic and monitoring systems, with broad potential for real‐world applications in both biomedical and environmental fields.

4. Experimental Section

Materials

PBS, citric acid, ethanol, IMD, glyphosate, atrazine, dicamba, chlorpyrifos, UA, glucose, glycine, ascorbic acid, potassium nitrate (KNO₃), sodium nitrate (NaNO₃), calcium nitrate tetrahydrate, silver nitrate (AgNO₃), chloroauric acid, hydrochloric acid, PVC, 2‐nitrophenyl octyl ether (2‐NPOE), potassium tetrakis(4‐chlorophenyl)borate, sodium tetrakis[3,5‐bis(trifluoromethyl)phenyl]borate (NaTFPB), calcium ionophore II (ETH 129), tetrahydrofuran (THF), isopropyl alcohol, ethanol, EDC, PEI, calcium nitrate hexahydrate (Ca(NO₃)₂·4H₂O), and potassium chloride (KCl) were purchased from MilliporeSigma (St. Louis, MO). Ag/AgCl ink for the reference electrode was obtained from Nagase ChemteX (Delaware, OH). Kapton (0.125 µm), the commercial polyimide film used in this work, was acquired from McMaster‐Carr (Elmhurst, IL). River water was collected from the South Skunk River in Ames, IA. Artificial saliva for pharmaceutical research was purchased from Pickering Laboratories (Mountain View, CA). All solutions were prepared using DI water with an ≈18.2 MΩ cm resistivity. All reagents were analytical grade unless otherwise specified.

LIG Fabrication

LIG microfluidics and electrodes were fabricated using a 75 W Epilog Fusion M2 CO₂ laser (10.6 µm wavelength).^{[ 82 ]} The 3‐electrode setup, containing a 2 mm diameter working electrode, and the microfluidic track were designed in CorelDRAW (Corel Corp., Ottawa, Canada). Polyimide films were first cleaned with isopropanol and then lased‐scribed in raster mode using 15% speed, 7% power, 50% frequency, 1200 dots in.⁻¹ (DPI), and at a +2 mm defocus based upon the previous work.^{[ 10 ]} The microfluidic was fabricated using the same settings. The 2‐electrode configuration used for ISEs incorporated a 3 mm working electrode and was twice lased using raster mode with the following settings. The first pass used the same settings as above (15% speed, 7% power, 50% frequency, and 1200 DPI) but at a +1 mm defocus. The second lase used similar parameters except the speed was changed to 20% and power to 3% to create a hydrophobic surface as described in the previous studies.^{[ 10} , ^{63 ]} To guarantee a constant working area throughout the experiments, the stems of the electrodes were coated with fast‐drying acrylic. Reference electrodes for the 2‐ and 3‐electrode setups were made as follows. The Ag/AgCl ink was painted onto the corresponding reference electrode. The samples were cured at 100 °C for 10 min to cure the ink. Subsequently, 4 mg KCl, 32 mg PVC, 61.5 µL NPOE, and 1 mL THF were combined into a mixture and bath sonicated for 1 min. The solution was drop‐cast over the cured Ag/AgCl ink and air dried for 2 h. Unless otherwise specified, all electrochemical potentials were reported with respect to this solid‐contact Ag/AgCl reference. The microfluidic was lased using raster mode at 15% speed, 7% power, 50% frequency, 1200 DPI, and a +2 mm defocus. To maintain the hydrophilicity of the LIG microfluidic, a solution of 50 mm EDC and 20% w/v PEI was prepared in DI water and drop‐coated onto the LIG tracks. After 10 min, any unbound PEI was rinsed with DI water, and the fluidic was allowed to air dry.

Ion‐Selective Membrane Fabrication

The ISEs were fabricated by modifying the solid‐contact LIG electrodes with an ISM specific for each ion. The NO₃ ⁻ membrane was prepared by mixing 7.5 mg tridodecylmethylammonium nitrate, 29 mg PVC, 78 µL 2‐NPOE, and 500 µL THF. The Ca²⁺ membrane was prepared by mixing 1.8 mg of calcium ionophore, 1 mg, NaTFPB, 32 mg PVC, 65.2 µL 2‐NPOE, and 750 µL THF. The ISM cocktails were vortexed individually until the solutes were completely dissolved. To functionalize the ISEs, two applications of 3 µL were drop‐coated onto the working area of the corresponding electrodes and allowed to dry for one day. Subsequently, the ISEs were submerged in a conditioning solution (1 mm NaNO₃ for NO₃ ⁻ electrodes and 1 mm Ca(NO₃)₂ 5H₂O for Ca²⁺ electrodes). Electrodes were submerged in their corresponding conditioning solutions until tested (≈24 h).

Fluidic Characterization

LIG fluidics were designed with 0.6 mm wide tracks. The experiments were conducted by placing the LIG fluidic sample on a glass slide directly under the camera (Photron Fastcam WX Mini with a maximum pixel resolution of 2048 × 2048 and a frame rate of up to 10 kHz) (San Diego, CA). Low‐flicker illumination was provided by using a 150 W fiber halogen lamp (RPS Studio CooLED 200 RS‐5620) (Wallingford, CT). A programmable Genie Touch syringe pump (Torrington, CT) was used to input a droplet into the end of the channel at an initial flowrate of 50 µL min⁻¹. The wicking of the water through the designed LIG track was captured at a frame rate of 250 frames s⁻¹ by using an EX Sigma 105 mm DG Macro lens (Aizu, Japan). This setup provided the temporal resolution to capture the fast movement of the fluid along the channel as well as the required spatial resolution to zoom into the region of interest. For each case, 3000 frames were recorded, and the images were analyzed using the proprietary software provided by the camera manufacturer (Photron Fastcam Viewer v4). Flowrate data were recorded using a high‐speed camera (250 fps). The individual frames were combined using Matlab (MathWorks, Natick, MA) and were inverted to improve the contrast between the fluid and the background. The fluid front was then analyzed using the Tracker Video Analysis and Modeling Tool by Physlets. For consistency, the hemiwicking front was tracked and was taken as the wetted portion of the material that slightly led the droplet. In order to track this front, the location was approximated as the point where the pixels became noticeably distorted or changed contrast due to fluid passing through the LIG. The Washburn relationship^{[ 27 ]} was used to develop a numerical model for fluid flow on the LIG microfluidic tracks. Data were plotted and residuals were minimized to yield the coefficient in Equation (1). Data in Figure 2b were represented by the average data on a 0.6 mm track. PowerPoint and Microsoft Paint were used to adjust the image contrast and coloring. A Rame‐Hart goniometer (model 90) quantified the static contact angle. The experiment was performed by drop‐coating 1 µL of DI water onto an LIG square (2 mm × 2 mm), and the contact angle of the droplet at the interface was analyzed using OMNIC software. Measurements were performed on bare LIG and PEI–LIG for n = 6 samples each.

Silver Nanostructure Electrodeposition

LIG electrodes were washed with 70% v/v ethanol 3 times, followed by DI water, to reduce surface tension on the electrode surface and allow the plating solution to penetrate the porous LIG structure. During electrodeposition, an LIG working, commercial external platinum counter, and commercial Ag/AgCl reference electrode (1 m KCl) were used. The Ag electrodeposition on the LIG working electrode was performed similar to the method used by Jeong et al.^{[ 38 ]} Briefly, the LIG electrodes were submerged in the Ag plating solution consisting of 5 mm AgNO₃ and 0.1 m KNO₃. Then, an initial cyclic voltammogram was taken from 0 to 0.7 V at a scan rate of 50 mV s⁻¹. The working electrode was then conditioned in a potential range beyond the reduction potential determined from the previous CV sweep in the plating solution by performing 5 CV cycles from 0.4 to 1 V at 50 mV s⁻¹. A potential step function was used to electrodeposit Ag nanostructures. While stirring at 700 rpm, an initial potential of 0.1 V was applied for 0.5 s followed by a potential step of 0.25 V for 50 s. The entire step function was performed 4 times, at which point, a color change was noticed on the surface of the working electrode. Electrodes were rinsed with 0.1 m citric acid thrice to clean oxidized Ag, followed by a DI water rinse. The electrodes were then rinsed with 70% ethanol thrice, followed by DI water. Stabilization CV scans were finally performed on the all‐LIG electrode setup in PBS pH 7.4 with the following settings: 30 cycles from −0.2 to −1.2 V at a scan rate of 50 mV s⁻¹.

Gold Nanostructure Electrodeposition

The electrodes were first cleaned with 70% ethanol thrice followed by DI water. A CV scan was performed from 0.6 to −0.8 V at 50 mV s⁻¹ to gain insight into the electrochemical reduction location of the gold salt. The AuNSs were electrodeposited onto LIG similar to the method utilized by Rauf et al.^{[ 41 ]} Briefly, while stirring at 700 rpm, an initial potential of −0.8 V was applied for 0.5 s followed by a pulse of −0.7 V for 10 s. This step function was performed 4 times. Electrodes were then rinsed with 70% ethanol followed by DI water and stabilized using the 3‐electrode LIG configuration in PBS pH 7.4 by performing 25 CV cycles from −0.3 to 0.4 V at 50 mV s⁻¹.

Raman Spectroscopy

Raman spectroscopy measurements were performed using a Horiba XploRA Plus confocal Raman microscope with a 532 nm laser operating at 1.2 mW and a 50× (0.5 numerical aperture) objective. The spectra were collected from 500–3500 cm⁻¹ with a 600 grooves mm⁻¹ grating. Seven Raman spectra were collected at seven randomly selected locations, and each Raman spectrum was collected with a 30 s acquisition and 3 accumulations. All Raman peaks in each spectrum were fitted to a Lorentzian function in Igor Pro 6.37 to calculate the average I _D/I _G and I _2D/I _G ratios.

Electrochemical Measurements

Electrochemical characterization was initially performed in 5 mm [Fe(CN)₆]^3−/4− redox probe and 0.1 m KNO₃ as the supporting electrolyte. AgLIG electrodes for IMD were stabilized in PBS using CV from −0.2 to −1.2 V at a scan rate of 50 mV s⁻¹ for 25 total CV cycles. For calibrations to IMD, electrodes were submerged in 5 mL of PBS. DPV was performed from −0.2 to −1.2 V with a 0.004 V potential increment, 0.1 V potential amplitude, 0.2 s pulse width, 0.4 s pulse period, and 5 s quiet time. IMD additions were made prior to the DPV scan. Scans were performed once for each concentration. In a similar manner, AuLIG electrodes for UA were stabilized in PBS following the same protocol listed above except in the range of −0.3 to 0.3 V. For calibrations to UA, electrodes were submerged in 2.5 mL of PBS pH 7.4. DPV was performed from 0.1 to 0.4 V with the same parameters previously described for the AgLIG electrodes. UA additions were made prior to scans while stirring at 400 rpm. DPV scans were performed without stirring. OCP measurements were performed in DI water with stirring at 300 rpm. Additions were made at half‐decade concentration steps.

Stability and Durability

Sensor stability was monitored by performing scans in 10× PBS without target analyte. The number of scans performed in a typical calibration was also performed in 10× PBS alone to monitor baseline fluctuations. Sensor stability at a single concentration was monitored by five successive scans in 50 µm for both IMD and UA. To test the durability of each sensor, sensors were bent at 90°, and 2‐point probe resistances were measured afterward.

Complex Media Sensing

Agrochemical sensors were validated in spiked river water collected from the South Skunk River, Ames, IA. River water was spiked as is without filtration. Two distinct river water solutions were prepared with spiked concentrations of NO₃ ⁻/IMD: solution A (1.6 mm/100 µm) and solution B (0.16 mm/50 µm). The acquired DPV measurements of the spiked samples were compared to the IMD calibration model in 10× PBS to calculate the found concentration and obtain recovery percentages with respect to the spiked concentrations. The acquired OCP measurements of the spiked samples were compared to the NO₃ ⁻ calibration model performed in river water (Figure S6a, Supporting Information). Biomedical sensors were validated in spiked artificial saliva, and the scans were compared to the Ca²⁺ and UA calibration models performed in artificial saliva (Figure S6b,c, Supporting Information). Two distinct artificial saliva solutions were prepared with spiked concentrations of Ca²⁺/UA: solution A (1 mm/300 µm) and solution B (0.5 mm/180 µm).

Interferent Testing

The appropriate interfering species were tested against the IMD and UA sensors in 10× PBS pH 7.4 and against the NO₃ ⁻ and Ca²⁺ sensors in DI water. The IMD sensor was tested in the presence of 40 µm of the pesticide species atrazine, glyphosate, dicamba, and chlorpyrifos as well 100 µm of the NO₃ ⁻. The NO₃ ⁻ sensor was tested using the fixed interferent method where a calibration of NO₃ ⁻ was performed in a constant concentration of the chosen interfering species (1 mm), which included Cl⁻, NO₂ ⁻, PO₄ ⁻³, CO₃ ²⁻, or SO₄ ⁻². The UA sensor was tested in the presence of 50 µm ascorbic acid, glucose, glycine, and 100 µm Ca²⁺. The Ca²⁺ sensor was also tested using the separate solution method in the presence of 1 mm Mg²⁺, Na⁺, or K⁺. Data for the IMD and UA voltametric sensors were reported as percentages with respect to the expected signal of the reference (no interfering species). Data for each reference and interfering species were reported as the average ± standard deviation for n = 4 sensors.

Data Analysis

A completely randomized design was used in this study with at least three independent replicates, and all results were reported as mean ± standard deviation. Data analysis was performed using JMP Pro (v.17, SAS, Cary, NC). The DPV baselines for each data set were fit using 4th order orthogonal least square functions. The final curves were obtained by subtracting the modeled baseline from the DPV original curve to normalize the data. For voltammetric sensors, the blank scans (no target analyte) were averaged among all reported sensors. The current noise (σ) was considered the standard deviation of the current baseline at −0.92 or 0.2 V for IMD and UA, respectively. Regression analysis with a significance level of 5% (α = 0.05) was performed to determine the linear sensing range and the relationship between each analyte concentration and the electrochemical signal. LODs were calculated using the 3 sigma method for voltametric sensors. The sensitivity was taken as the slope of the linear sensing region. For ion‐selective sensors, the LOD values were determined as the intersection point obtained from the extrapolation of the two linear segments observed in the ISE calibration plots (i.e., the nonsensing linear range at lower concentrations and the linear range). All figures were plotted in OriginPro (2022b Academic, MA, USA).

Conflict of Interest

The authors declare no conflict of interest.

Supporting information

Supporting Information

Johnson Z. T., Ellis G., Pola C. C., Banwart C., McCormick A., Miliao G. L., Duong D., Opare‐Addo J., Sista H., Smith E. A., Hu H., Gomes C. L., Claussen J. C., Enhanced Laser‐Induced Graphene Microfluidic Integrated Sensors (LIGMIS) for On‐Site Biomedical and Environmental Monitoring. Small 2025, 21, 2500262. 10.1002/smll.202500262

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.