Device for Electrochemical Caffeine Analysis in Fluids (DECAF): A Stir‐Bar‐Embedded Sensor for Real‐Time Caffeine Analysis in Commercial and Homemade Beverages

Abstract

Caffeine (CAF) is the most widely consumed psychoactive compound worldwide; however, accurate labeling of its concentration in commercial beverages remains unregulated. To address this gap and minimize the risks of unintentional overconsumption, a portable electrochemical sensor—DECAF (Device for Electrochemical Caffeine Analysis in Fluids)—is developed for real‐time quantification of caffeine in both commercial and homemade beverages. The sensor incorporates Nafion‐modified laser‐induced graphene (LIG) electrodes and a phosphate‐buffered glass fiber (PBS‐GF) layer, facilitating rapid absorption and buffering of liquid samples to ensure stable electrochemical response. Designed in the form of a stir bar, the device operates using square wave voltammetry (SWV) and achieves a detection limit of 0.06 mM, with a linear detection range from 0.5 to 5.0 mM. The system demonstrates high selectivity under varying pH and temperature conditions. DECAF requires only 200 µL of sample and delivers results within 5 min, interfacing wirelessly with a smartphone for real‐time data visualization and storage. This compact, low‐cost, and user‐friendly platform enables on‐site caffeine monitoring and holds potential for extension to biofluid analysis and detection of other target analytes, offering a scalable solution for personal health tracking and dietary management.

A portable, stir bar‐based electrochemical sensor–DECAF–enables rapid and accurate caffeine detection in beverages. Using laser‐induced graphene electrodes and a phosphate‐buffered absorbent layer, DECAF quantifies caffeine in just 5 min over a 0.5 to 5.0 mM range with minimal sample volume. Its compact design, smartphone connectivity, and high selectivity offer a practical tool for personal health tracking and dietary monitoring.

1. Introduction

Caffeine (1,3,7‐trimethylxanthine) is a naturally occurring alkaloid found in various plant species, characterized by the chemical formula C₈H₁₀N₄O₂.^{[ 1 ]} Upon consumption, caffeine primarily exerts its physiological effects by blocking adenosine receptors in the brain.^{[ 2 ]} This inhibition leads to increased neuronal excitability and the subsequent release of stimulatory neurotransmitters such as dopamine and norepinephrine, thereby enhancing cognitive alertness, energy levels, and concentration.^{[ 3 ]} As the most widely consumed psychoactive substance globally, caffeine is ingested daily by approximately 80% of the world's population.^{[ 4 ]} Common dietary sources include coffee, tea, energy drinks, and soft drinks. The global market for caffeinated beverages was valued at over USD 200 billion in 2023 and is projected to exceed USD 300 billion by 2030, reflecting a significant upward trajectory.^{[ 5} , ^{6 ]} This substantial market expansion highlights the growing demands for devices capable of accurately measuring caffeine content in beverages. Another critical factor driving the need for such measurement tools is regulatory limitations. According to the U.S. Food and Drug Administration (FDA) regulation 21 CFR 101.4, manufacturers of caffeinated beverages are required to list caffeine as an ingredient on the nutrition facts label.^{[ 7 ]} However, there is no legal mandate to disclose the exact caffeine concentration in non‐drug consumables such as beverages. This lack of transparency emphasizes the necessity for increased consumer awareness regarding caffeine intake, particularly to mitigate potential adverse effects associated with excessive consumption.

According to the U.S. FDA and the European Food Safety Authority (EFSA), caffeine intake of up to 400 mg per day (approximately 5.7 mg kg⁻¹ body weight) distributed throughout the day does not pose safety concerns for healthy adults. This intake is roughly equivalent to the consumption of four 8‐ounce cups of brewed coffee, each containing approximately 95 mg of caffeine.^{[ 8} , ^{9 ]} However, the caffeine content per beverage can vary significantly, regardless of the source or similar volume.^{[ 10} , ^{11 ]} This variability makes it easy for consumers to inadvertently mismanage their caffeine intake. Exceeding the recommended intake may lead to adverse effects, including anxiety, insomnia, gastrointestinal disturbances, and dependence.^{[ 12} , ^{13 ]} Caffeine‐induced anxiety disorder is formally recognized in the Diagnostic and Statistical Manual of Mental Disorders (DSM), characterized by heightened nervousness, excessive worry, or fear, often accompanied by physiological symptoms such as increased heart rate, sweating, and tremors.^{[ 14} , ^{15 ]} Experimental studies examining acute caffeine effects have demonstrated a significant increase in subjective anxiety one hour after the ingestion of 6.6 mg/kg caffeine in human volunteers.^{[ 16 ]} Similarly, consumption of 10 mg kg⁻¹ has been associated with heightened anxiety and nervousness, as well as elevations in systolic and diastolic blood pressure.^{[ 17 ]} Due to caffeine's half‐life ranging between 2 and 10 h,^{[ 18 ]} its consumption in the evening can interfere with sleep quality, which is critical for cognitive function, emotional regulation, and overall well‐being.^{[ 19} , ^{20 ]} A polysomnography‐based study found that the ingestion of 400 mg of caffeine four hours before bedtime significantly reduced sleep quality and delayed sleep onset.^{[ 21 ]} Further investigations in human subjects corroborate these findings, linking caffeine intake to sleep disturbances and an increased risk of insomnia.^{[ 22} , ^{23 ]}

Caffeine is typically detected in beverages using sophisticated and resource‐intensive techniques such as High‐Performance Liquid Chromatography (HPLC)^{[ 24 ]} and Gas Chromatography‐Mass Spectrometry (GC‐MS).^{[ 25 ]} While highly accurate, these methods are expensive, complex, time‐consuming, and require bulky instrumentation, making them impractical for portable caffeine monitoring. Colorimetric, spectrophotometric, and fluorimetric detection has gained popularity due to its simplicity, sensitivity, and response time. However, it often requires optical equipment that not only increases the size and cost of the system but also demands careful calibration and environmental control, which limits portability and ease of use.^{[ 26} , ²⁷ , ²⁸ , ²⁹ , ^{30 ]} In contrast, electrochemical techniques enable direct, label‐free detection of caffeine with relatively simple instrumentation, lower power consumption, and easier miniaturization.^{[ 31} , ^{32 ]} Caffeine is an electrochemically active molecule that undergoes irreversible oxidation at positive potentials, which produces a well‐defined anodic peak, enabling its quantification through voltammetric techniques.^{[ 33} , ^{34 ]} The distinct electrochemical signature of caffeine highlights its suitability for direct electrochemical sensing, eliminating the need for complex preparation.

To support efficient electron transfer during detection, carbon‐based materials are widely employed as electrode substrates due to their excellent conductivity, chemical stability, and high surface area.^{[ 35} , ³⁶ , ^{37 ]} These materials can be fabricated through various carbonization techniques, such as laser‐induced conversion of polymers into porous, graphitic structures. This process not only creates high electrochemical activity but also allows for the direct patterning of conductive electrodes with tunable properties. As demonstrated in previous work,^{[ 38 ]} Laser‐Induced Graphene (LIG) electrodes offer significant advantages, particularly for portable devices,^{[ 39} , ⁴⁰ , ⁴¹ , ⁴² , ⁴³ , ⁴⁴ , ⁴⁵ , ⁴⁶ , ^{47 ]} the LIG electrodes can be fabricated efficiently by applying a laser to commercial polymers, making them a cost‐effective alternative.^{[ 48} , ^{49 ]} They exhibit excellent stability, allowing for prolonged usage without significant signal degradation.^{[ 50 ]} LIG electrodes offer exceptional versatility, enabling the detection of a wide range of analytes.^{[ 51} , ⁵² , ^{53 ]} Moreover, LIG electrodes integrate seamlessly with microfluidic devices and sensors, further enhancing their applicability in portable detection systems. These features collectively make LIG electrodes a more practical choice for caffeine quantification in real‐world POC applications.^{[ 54} , ^{55 ]}

In this study, we propose the development of a novel portable sensor, named the Device for Electrochemical Caffeine Analysis in Fluids (DECAF), for the real‐time detection of caffeine in everyday beverages that is both sensitive and selective. While electrochemical detection of caffeine in aqueous solutions has been previously demonstrated,^{[ 56} , ⁵⁷ , ⁵⁸ , ⁵⁹ , ⁶⁰ , ⁶¹ , ⁶² , ⁶³ , ⁶⁴ , ^{65 ]} significant challenges remain when transitioning to complex real‐world matrices such as commercial beverages. These include lack of supporting electrolytes, pH variability, and temperature fluctuations, all of which can impair detection sensitivity and reproducibility. Our approach integrates LIG electrodes with a glass fiber layer preloaded with phosphate‐buffered saline (PBS). Upon contact with the sample, the glass fiber rapidly absorbs and transports the liquid to the electrode surface while simultaneously introducing a controlled ionic environment. This preconcentration layer acts as both a buffer and stabilizer, mitigating the effects of pH and temperature variations commonly found in caffeinated beverages. Caffeine quantification is achieved via Square Wave Voltammetry (SWV), leveraging the high surface area and conductivity of the LIG for enhanced signal resolution. To our knowledge, this is the first system to integrate LIG electrodes with a preloaded buffer matrix for portable caffeine detection in commercial beverages. Improving upon previous approaches that relied on pretreatment procedures of beverages,^{[ 66 ]} thus eliminating the need for any sample treatment prior to analysis, as the device is capable of directly analyzing the beverage in its original form. The platform offers a low‐cost, compact, and user‐friendly solution that enables consumers to monitor their daily caffeine intake without the need for sophisticated laboratory instrumentation, bridging the gap between analytical chemistry and real‐world usability.

2. Results and Discussion

2.1. Device Overview

The Device for Electrochemical Caffeine Analysis in Fluids (DECAF) is a novel, paddle‐shaped sensor developed for real‐time quantification of caffeine in both self‐prepared and commercially available beverages (Figure  1 ). This device enables convenient, on‐site caffeine monitoring to support informed health decisions and mitigate adverse effects associated with caffeine overconsumption. To operate, the DECAF is immersed and stirred directly in the target beverage. An open paper microfluidic channel composed of Whatman grade 1 filter paper rapidly absorbs and wicks the liquid toward a preloaded PBS‐GF layer that interfaces with the electrochemical sensing zone. The sensing zone comprises a miniaturized three‐electrode electrochemical cell, in which the working electrode applies SWV to oxidize caffeine molecules. This redox process produces a current directly proportional to the caffeine concentration. For analytical detection, SWV is employed due to its high sensitivity and selectivity. The electrode contacts are exposed at the upper end of the device, allowing connection to a commercially available, smartphone‐compatible wireless potentiostat for real‐time electrochemical readout, data visualization, and storage. As shown in Figure S8 (Supporting Information), this setup enables seamless integration with wireless technology.

Figure 1.

Application of DECAF (Device for Electrochemical Caffeine Analysis in Fluid) in monitoring daily caffeine intake.

The DECAF is particularly valuable for individuals experiencing symptoms of excessive caffeine intake (Figure 1), offering a user‐friendly solution for proactive consumption tracking. Given the absence of standardized caffeine labeling and the variability in homemade caffeinated drinks, DECAF addresses a critical need for portable caffeine assessment tools to promote awareness and healthier dietary habits. The DECAF requires 200 µL of liquid to fully saturate the paper channel and PBS‐GF layer, achieving complete wicking and contact with the sensing zone within 5 minutes when positioned horizontally. The structural design of DECAF (Figure  2B) was inspired by the common daily use of coffee stirrer sticks, we engineered the device to be safely immersed in beverages by selecting biocompatible materials, (1) a wooden base shaped like a paddle for stirring, covered with (2) a hydrophobic adhesive polyethylene terephthalate (PET) layer to prevent fluid loss through absorption into the wood and to direct flow toward the sensing area (Figure S7, Supporting Information). The device also integrates (4, 6) dual Whatman grade 1 filter papers on both sides of (5) the PBS‐GF layer, enhancing capillary flow and reducing fluid transport time from the paper layer to (3) the sensing zone. The sensing zone is sealed with (7) an adhesive Tegaderm Transparent Film, which provides a waterproof, sterile barrier against external contaminants while preserving electrochemical quantification.

Figure 2.

The steps in the fabrication of DECAF. A) The fabrication of Nafion‐modified LIG electrodes. B) Layer‐by‐layer 3D model of DECAF. C) Real photo of elements for the assembly of the DECAF platform. D) Photo of the fully assembled DECAF.

2.2. Fabrication

Building upon our previous work,^{[ 38} , ³⁹ , ⁴⁰ , ⁴¹ , ⁴² , ⁴³ , ⁴⁴ , ⁴⁵ , ^{67 ]} we adopted a similar approach to optimize the fabrication of LIG for the electrochemical detection of caffeine (Figure 2A). LIG was fabricated via photothermal conversion of aromatic polyimide (PI) films using a 9.3 µm CO₂ laser beam, which transforms the polymer surface into a 3D, multilayered, and porous graphene network. Laser engraving was performed under a 15 psi compressed air stream, which is the standard setting recommended by the laser system manufacturer. This airflow serves to remove ablation byproducts during carbonization and prevent overburning. We note that unlike argon‐assisted LIG fabrication, which generates a hydrophobic surface, air‐assisted carbonization promotes moderate hydrophilicity that is favorable for fluid absorption and electrochemical interaction in aqueous environments, as demonstrated in our prior work.^{[ 68 ]} The structural and electrochemical quality of the resulting LIG is highly dependent on the energy surface density (or laser fluence) of the laser beam on PI film during the printing process, which is governed by the laser power, scan speed, and beam diameter. We controlled the fluence by adjusting the laser power and scan speed via the laser engraver's software interface (UCP v5.38.58.0, Universal Laser Systems Inc.). Beam diameter was modulated indirectly via defocusing, with increased defocus distance resulting in a larger spot size and lower fluence. To optimize LIG formation, we systematically varied the power between 7.4% and 8.6% and scan speed between 14% and 26% of the machine's maximum settings. Excessive fluence (high power, low speed) led to film burning, while insufficient fluence (low power, high speed) failed to induce graphene formation. Within the optimal window, we produced uniform, mechanically stable LIG with electrochemically active surface areas of 26.0–56.2 mm² (Figure  3A) and sheet resistances ranging from 9.58 to 19.9 Ω sq⁻¹ (Figure 3B). As LIG is a 2D carbon material, we report its electrical property in terms of sheet resistance (Ω/sq), which is the standard unit for planar conductive films. The surface area and sheet resistance measurements were performed in triplicates (n = 3), and corresponding standard deviations are provided in Tables  1 and 2 of the Supplementary Information. Based on these findings, we selected a power of 7.7% and scan speed of 20% as the optimal parameters, for subsequent sensor fabrication, ensuring high conductivity and surface area (13.6 Ω sq⁻¹ and 44.6 mm²) relative to 7.07 mm² geometric area, critical for sensitive and reproducible caffeine detection.

Figure 3.

Characterization of laser‐induced graphene (LIG) electrodes. A) Electrochemical active surface area with different parameters (n = 3). B) Sheet resistance with different parameters (n = 3). C) Top view of SEM image of LIG. D) Cross section view of SEM image of LIG. E) EDX analysis. F) Raman spectrum analysis. G) High‐resolution XPS spectra for C 1s. H) Cyclic voltammograms with solution of 3.0 mM [Fe(CN)₆]^3‐/4‐ in 100 mM KCl at a scan rates between 20 and 100 mV/s. I) The relationship of the oxidation and reduction peak currents with the square root of the scan rate of the cyclic voltammetry experiments.

Table 1.

Comparison of electrochemical sensors for caffeine detection.

Electrode	Technique	Solvent	Linear Range [µM]	Sensitivity [µA/mM]	LOD [µM]	Refs.
Nafion/LIG/PBS‐GF	SWV	Purified Water	500–5000	8.37	62.3	This work
		DI Water	500–5000	8.43	52.0
		PBS	500–5000	23.8	75.9
Nafion/RuO2Py/GCE	SWV	HClO4	0.005–0.2	507	2.0	[56]
Nafion‐Gr/GCE	DPV	Sulfuric acid	40.0–600	107	70.0	[57]
BDD	DPV	HClO4	0.4–25.0	3200	0.15	[58]
PCP	SWV	PBS	1.0–1000	2610	0.348	[59]
NCOMCP	DPV	Sulfuric acid	5.0–600	190	0.16	[60]
Poly‐AHNSA/GCE	SWV	Acetate buffer	0.06–400	503	0.137	[61]
PST/Nafion/GCE	SWV	PBS	0.3–1000	570	1.0	[62]
Surfactant/MWCNTs	AdSDPV	PBS	0.291–62.7	60.8	0.088	[63]
Nanocomposite MIP/PGE	SWV	PBS	0.002–10.0	309	0.001	[64]
SPCE/CNFS	AdSDPV	Sulfuric acid	200–1000	3900	0.056	[65]
PDA‐MWCNTs Agarose Hydrogel	Chronoamperometry	PBS	0.004–18.2	—	0.001	[66]

Table 2.

Interference results of common beverage components at 10:1 interferent‐to‐caffeine ratio (n = 3).

Name	Abbreviation	Ratio	Measured [mM]	Relative error	Relevance
Fructose	Fru	10:1	0.97 ± 0.27	3%	Natural Sugar
Dextrin	Dex	10:1	1.04 ± 0.08	4%	Flavor enhancer amino acid
Glycine	Gly	10:1	0.95 ± 0.21	5%	Artificial Sweetener
Aspartame	Asp	10:1	0.93 ± 0.13	7%	Texture/stabilizer additive
Benzoic Acid	Ben	10:1	0.92 ± 0.29	8%	Preservative
Acesulfame‐K	Ace	10:1	0.86 ± 0.19	14%	Artificial Sweetener
Glucose	Glu	10:1	0.85 ± 0.23	15%	Natural Sugar

2.3. Electrode Characterization

To characterize the surface morphology of our LIG, scanning electron microscopy (SEM) was performed. The top‐view (Figure 3C) and cross‐sectional (Figure 3D) SEM images reveal the formation of a 3D porous carbon network resulting from the laser ablation of the PI film. This highly porous architecture contributes to an increased electrochemically active surface area, thereby enhancing the sensitivity of the sensor platform. The electrochemical active surface area of LIG electrodes is typically much larger than their geometric area due to the material's 3D structure. While the geometric area refers to the projected visible flat area of the electrode, the electrochemical active surface area accounts for the actual surface accessible for electrochemical reactions, including internal pores, edge planes, and defects. As a result, electrochemical active surface area provides a more accurate measure of the electrode's functional performance in electrochemical sensing applications. Additionally, improved electron transfer kinetics and electrode conductivity were facilitated by the presence of edge‐plane sites on the LIG surface. To analyze the elemental composition, energy‐dispersive X‐ray spectroscopy (EDX) was performed (Figure 3E), confirming that the LIG is predominantly composed of carbon with only trace amounts of oxygen–evidence of successful carbonization of the polyimide film and the formation of high‐quality graphene‐like material. Figure S3 (Supporting Information) shows additional SEM and EDX characterization of LIG electrodes.

Raman spectroscopy was employed to further examine the structural features of the laser‐induced graphene formed under optimized engraving conditions. The Raman spectrum (Figure 3F) displays characteristic graphene peaks: the G band at 1575 cm⁻¹ associated with sp² carbon vibrations, the D band at 1350 cm⁻¹ indicative of lattice disorder and defects, and the 2D band at 2688 cm⁻¹ arising from a two‐phonon lattice vibration process. The intensity ratio I_2D /I _G was 0.485, suggesting the formation of 4–5 layers of graphene.^{[ 69 ]} X‐ray photoelectron spectroscopy (XPS) was performed to investigate the chemical state of the material. The survey spectrum exhibited distinct peaks for carbon (C 1s) and oxygen (O 1s), located at binding energies of 284.8 eV and 531.2 eV, respectively (Figure S1, Supporting Information). The deconvolution of the C 1s peak was performed using CasaXPS software to identify specific carbon bonding environments (Figure 3G). The sharp peak at 284.8 eV represents the C═C bond in the graphene structure, and the peak at 285.0 eV is assigned to the C–C bond. The other peaks at 285.9 and 289.2 eV indicate the C–O and C═O bonds, respectively.

The surface area and fabrication consistency of LIG‐based electrodes were evaluated using cyclic voltammetry (CV). Measurements were conducted in a potential window of –0.2 to 0.7 E vs. Ag/AgCl in a solution containing 2 mM [Fe(CN)₆]^{3 − /4 −} and 100 mM KCl. Scan rates of 20, 40, 60, 80, and 100 mV s⁻¹ were used (Figure 3H). As expected, both oxidation and reduction peak currents increased linearly with the square root of the scan rate, confirming a diffusion‐controlled electrochemical process at the electrode surface (Figure 3I). According to our previous work, the electroactive area was analyzed using the modified Randles–Ševčík equation (Equation 1) for the quasi‐reversible systems.

$$I_p^{\mathrm{quasi}}=\pm 0.436nFAC{\left(\frac{nFD\nu }{RT}\right)}^{1/2}$$ (1)

Where, i _p represents the peak current (A), n is the number of electrons transferred, F is the Faraday constant (C mol⁻¹), A is the electrode area (mm²), D is the diffusion coefficient (6.7 × 10⁻⁶ cm² s⁻¹), C is the electrolyte concentration (mol cm⁻³), ν is the scan rate (V s⁻¹), R is the gas constant (J K⁻¹ mol⁻¹), and T is the temperature (K). By this analysis, the calculated electroactive surface area was 26.6 ± 1.25 mm² (n = 4), representing an enhancement of approximately 3.7‐fold over the 7.07 mm² geometric area of the electrode, attributed to the porous and textured nature of the LIG surface.

2.4. Electrochemical Quantification of Caffeine

Caffeine is an electrochemically irreversible analyte that undergoes oxidation (Equation 2) at a positive potential between +1.1 and +1.3 E vs. Ag/AgCl, generating a measurable peak current via voltammetry‐based techniques that enables its quantification in solution.^{[ 33} , ^{34 ]} To enhance electrode performance and selectivity, a Nafion coating was applied to the working electrode. Nafion is a sulfonated tetrafluoroethylene‐based polymer composed of a hydrophobic PTFE backbone and hydrophilic sulfonic acid (–SO₃H) side chains, which facilitate proton conductivity and selective ion exchange.^{[ 70} , ^{71 ]} The coating serves to exclude negatively charged interferences, improves electrochemical stability, and reduces background current.^{[ 72} , ^{73 ]} We evaluated the performance of Nafion‐modified electrodes by drop‐casting varying concentrations (0.05–0.5% v/v) of Nafion solution onto the working electrode surface. Among these, the 0.1% (v/v) Nafion coating yielded the highest peak current‐to‐baseline ratio, indicating optimal performance for caffeine detection.

$$\underset{\text{caffeine}}{{\mathrm{C}}_8{\mathrm{H}}_{10}{\mathrm{N}}_4{\mathrm{O}}_2}+\underset{\text{water}}{{\mathrm{H}}_2\mathrm{O}}\to \underset{\text{oxidized caffeine}}{{\mathrm{C}}_8{\mathrm{H}}_{10}{\mathrm{N}}_4{\mathrm{O}}_3}+2\underset{\text{proton}}{{\mathrm{H}}^{+}}+2\underset{\text{electron}}{{\mathrm{e}}^{-}}$$ (2)

To assess signal consistency, we compared SWV responses of three independent electrodes with and without Nafion coating using 1 mM caffeine in PBS (Figure  4A). Although the Nafion‐coated electrodes exhibited a lower mean peak current (12.4 µA; range: 12.3–12.5 µA), they produced a significantly lower standard deviation (0.11 µA), indicating improved reproducibility. In contrast, bare electrodes showed a higher mean peak current (31.2 µA; range: 22.1–44.2 µA) but with a larger standard deviation (11.5 µA), highlighting greater variability in unmodified systems. We evaluated the contribution of the PBS‐GF layer to caffeine quantification in ion‐deficient solutions. SWV measurements of 1 mM caffeine in deionized (DI) water, with and without the PBS‐GF layer, revealed that a distinct oxidation peak was only observed when the PBS‐GF layer was present (Figure 4B). This confirms that the layer provides essential supporting electrolytes, enabling accurate caffeine quantification in low‐conductivity media.

Figure 4.

Caffeine quantification using DECAF. A) SWVs measured with bare LIG vs. Nafion‐modified LIG. B) SWVs measured with pre‐concentrated PBS‐GF layer vs PBS‐GF layer without buffer. C) Nafion % (v/v) optimization. D) Caffeine SWVs from 0.5 to 5 mM in PBS with non‐loaded PBS‐GF layer, E) in DI Water, and F) in Purified Water both with preloaded PBS‐GF layer. G) Calibration curve from 0.5 to 5 mM in PBS with non‐loaded PBS‐GF layer, H) in DI Water, and I) in Purified Water with preloaded PBS‐GF layer (n = 3).

Nafion‐modified LIG electrodes with PBS‐GF layer were employed to construct calibration curves for caffeine detection in PBS (Figure 4D,G) and DIW (Figure 4E,H) across a concentration range of 0.5–5 mM. This calibration range was selected to encompass typical caffeine concentrations found in commercially available beverages, ensuring relevance to both commercial and self‐prepared drinks. Additionally, a calibration curve was constructed in Purified Water (PW) (Figure 4F,I), the commercially used water in most caffeinated beverages, to better simulate real‐world conditions. Unlike DI water commonly used in laboratory settings, PW undergoes reverse osmosis filtration, which does not entirely remove ionic species.^{[ 74 ]} This approach allows assessment of the sensor's performance in a more representative ionic environment. Calibration curves are fundamental in sensor development, as they establish the quantitative relationship between the measured signal and known analyte concentrations, enabling accurate determination of caffeine levels in unknown samples.^{[ 75 ]} For caffeine detection in PW, the sensor exhibited a sensitivity of 8.37 µA mM⁻¹, with excellent linearity (R ² = 0.998) across the tested concentration range. The limit of detection (LOD) was determined to be 0.0623 mM, calculated as three times the standard deviation of the blank divided by the calibration slope (LOD = 3σ/s). Similarly, the limit of quantification (LOQ) was calculated as 0.208 mM, using ten times the standard deviation of the blank LOQ = 10σ/s. When compared to other modified electrodes (Table 1), the Nafion/LIG/PBS‐GF sensor demonstrated high sensitivity and a low detection limit, making it well‐suited for the quantification of caffeine concentrations in PW, typically found in popular commercial beverages, specially at high concentrations. While the applied SWV potential window (0.8–1.6 V vs Ag/AgCl) overlaps with the onset region of oxygen evolution in aqueous systems, no significant bubble formation, baseline instability, or signal distortion was observed across replicate measurements. The short pulse duration and low duty cycle of SWV minimize the accumulation of evolved oxygen at the electrode surface, reducing its impact on analytical performance. This is further supported by the reproducible caffeine oxidation peaks and stable calibration behavior across sample matrices (PBS, DI water and purified water). Thus, oxygen evolution does not appear to interfere with caffeine quantification under our operational conditions.

Stability studies were performed to evaluate the reliability and reproducibility of the LIG electrodes over time–an essential consideration for real‐world applications and batch‐to‐batch consistency. A batch of LIG electrodes and PBS‐GF layers were fabricated on a single occasion, then stored and tested on days 0, 2, 4, and 8 by measuring 3 mM caffeine in PW. As shown in Figure S4 (Supporting Information), the electrodes exhibited stable and consistent responses across all time points, with a mean peak current of 25.0 µA (range: 25.3–24.4 µA) and a low standard deviation of 0.44 µA.

2.5. Effect of pH on Caffeine Detection

When measuring caffeine in commercial beverages, it is crucial to consider the wide pH variability (pH 3–7) among products such as coffee, tea, and energy drinks.^{[ 76} , ^{77 ]} The oxidation peak current and potential of caffeine are strongly influenced by the pH of the electrolyte solution, primarily due to proton‐coupled electron transfer (PCET) mechanisms involved in its electrochemical oxidation.^{[ 78} , ^{79 ]} Variations in pH alter the availability of protons, which in turn affects the thermodynamics and kinetics of the oxidation process. At lower pH values, the abundance of protons facilitates caffeine oxidation, often resulting in higher peak currents. Conversely, at higher pH values, proton scarcity hinders the oxidation process, leading to lower peak currents and decreased sensitivity. To address this, we employed a PBS‐GF layer as a localized buffering matrix to stabilize pH during fluid flow across the sensor surface. The PBS‐GF layer maintains a consistent pH environment by leveraging the conjugate acid‐base pair (${\mathrm{H}}_2{\mathrm{PO}}_4^{-}/{\mathrm{HPO}}_4^{2-}$) present in PBS, which neutralizes incoming H⁺ and OH⁻ ions through acid‐base reactions. This buffering action ensures stable electrochemical performance at the electrode interface.

To evaluate the effect of pH stabilization, SWV measurements of 2.5 mM caffeine in PBS were conducted across pH values ranging from 3 to 7. Without the PBS‐GF layer (Figure  5A), we observed a increase in peak current as pH decreased, with values ranging from 43.4 to 77.2 µA (Figure 5D). In contrast, electrodes incorporating the PBS‐GF layer (Figure 5B) exhibited significantly stabilized peak currents, maintaining a narrow range of 42.2–44.2 µA with low standard deviation (Figure 5E) across the same range of pH values. Furthermore, the peak position of caffeine oxidation remained stable, with an average of 1.21 V vs. Ag/AgCl and a low standard deviation of 0.013 V, further validating the robustness of our sensing platform. These results highlight the influence of pH in caffeine quantification and demonstrate the effectiveness of the PBS‐GF layer in providing consistent and reliable electrochemical signals despite variations in sample pH.

Figure 5.

Caffeine measurements under varying conditions. A) Effect of pH on SWVs of 2.5 mM caffeine in LIG without PBS‐GF layer (B), and with PBS‐GF layer. C) Selectivity of 1 mM caffeine SWVs against interference molecules in PW. D) Effect of pH on peak current of 2.5 mM caffeine in LIG without PBS‐GF layer (E), with PBS‐GF layer. F) Selectivity of 1 mM caffeine peak currents against interference molecules in PW (n = 3).

2.6. Selectivity Studies

The selectivity of the DECAF device was evaluated by assessing the influence of common interfering species present in caffeinated beverages.^{[ 80} , ^{81 ]} The objective was to demonstrate the device's ability to accurately detect caffeine without signal distortion from other electroactive compounds. Potential interferents, including glucose (Glu), dextrin (Dex), glycine (Gly), fructose (Fru), aspartame (Asp), acesulfame‐K (Ace), and benzoic acid (Ben) were tested at 10 mM concentrations, representing a 10‐fold excess relative to caffeine (1 mM) in PW. SWV measurements were performed for each interferent‐spiked solution, with 1 mM caffeine added to all samples. The resulting peak currents were compared to those of an interference‐free caffeine solution. As shown in Figure 5C,F, the presence of interfering species did not significantly alter the caffeine oxidation signal, with peak currents remaining consistent with those obtained in pure caffeine solutions. To quantitatively assess selectivity, the Relative Error (%) was calculated using the ratio of the peak current from the interferent‐spiked sample to that of the interference‐free caffeine solution (Table 2).

Among the tested interferents, Fru, Dex, and Gly exhibited the lowest Relative Error (%) in caffeine detection, as they do not display significant redox activity within the caffeine oxidation window (+1.1 to +1.3 V vs. Ag/AgCl). In contrast, Asp, Ben, and Ace could be expected to interfere due to their potential electroactivity within this range or their ability to interact with the electrode surface. However, these effects are mitigated by the presence of Nafion, which electrostatically repels negatively charged species, thereby limiting their access to the electrode. Glu, despite being electrochemically inactive in this potential window, resulted in the highest Relative Error (%). This is likely due to its tendency to weakly adsorb onto the electrode surface at high concentrations (10 mM), partially blocking active sites and hindering electron transfer, leading to signal suppression. The resulting low interference percentages all below 15% confirm that the DECAF device exhibits excellent selectivity for caffeine detection, demonstrating its suitability for accurate quantification in complex beverage matrices. Additionally, it is important to consider other methylxanthines, such as theobromine and theophylline, which may oxidize at potentials similar to caffeine. However, these compounds were not included in our selectivity studies because they are either absent or present only in trace concentrations in the most common homemade and commercial caffeinated beverages.^{[ 82} , ⁸³ , ⁸⁴ , ⁸⁵ , ^{86 ]} Therefore, we did not consider them significant contributors to potential interference in our electrochemical measurements.

2.7. Fluidic Studies on Device for Sample Collection

The wicking components of the DECAF device consist of cellulose filter paper‐GF, which are assembled onto the device using double‐sided adhesive tape. The fluid transport in these materials is driven by capillary forces, which arise from the interaction between the liquid's surface tension and the porous structure of the material.^{[ 87} , ^{88 ]} Specifically, the interconnected pores within the paper and GF layers create capillary pressure that pulls the fluid into the void spaces, displacing the air originally present. This wicking mechanism allows the caffeinated beverage sample to travel through the wicking channel toward the sensing zone. To confine the fluid within the desired pathway, a hydrophobic PET layer is placed at the base of the device. This hydrophobic barrier prevents fluid from wicking outside the designated channel, ensuring that liquid flow remains directed through the paper‐GF pathway. Upon reaching the sensing zone, the channel is sealed with a Tegaderm transparent adhesive film, which serves as a protective barrier against polluting and external contamination. Tegaderm maintains optical clarity, provides a waterproof, sterile seal, while maintaining optical clarity, and has minimal impact on electrochemical performance.^{[ 89} , ^{90 ]} Additionally, it helps prevent fluid overflow beyond the sensing area, ensuring a controlled sample volume remains in contact with the electrodes.

To determine the dead volume of the microfluidic layer in the DECAF device, the minimum fluid volume required to saturate the channel and reach the sensing zone, we performed visualization experiments using PW dyed with dark blue food dye (Figure  6A,B). Sequential photographs were captured at 1‐minute intervals after introducing 100 or 200 µL of dyed solution into the channel. Results indicated that 100 µL was insufficient to fully saturate the channel and reach the sensing zone, even after extended time. In contrast, 200 µL successfully saturated the channel, fully reaching the sensing zone within 5 minutes of application. To further evaluate this behavior with caffeine solutions, we replaced the dye with a 3 mM caffeine solution in PW and performed SWV measurements at 1, 2, 3, 4, and 5‐min time points for both 100 and 200 µL sample volumes. Consistent with the dye experiments, no oxidation peaks were observed at any time point for the 100 µL trials (Figure 6C). However, a clear and reproducible oxidation peak corresponding to caffeine was detected in the 200 µL trials (Figure 6D), confirming the required volume for effective detection. Therefore, our dead volume studies demonstrate that 100 µL is insufficient to activate the DECAF system, as this volume does not fully saturate the microfluidic channel. Consequently, the fluid fails to reach the sensing zone containing the LIG electrode, preventing any measurable electrochemical response. In contrast, a minimum of 200 µL is required to completely saturate the channel and enable SWV measurements. The 200 µL saturation requirement is practical for real‐world applications, as it represents only a negligible fraction of a typical beverage serving, such as a 12 oz (∼ 355 mL) commercial or homemade drink. This volume would be sufficient to operate over 1700 DECAF devices, highlighting the feasibility of the platform in real‐world applications. Furthermore, the 5‐min activation and measurement window is reasonable for on‐site testing, though we recognize this could be further optimized in future iterations of the DECAF system.

Figure 6.

Flow and Temperature test with DECAF. A,B) Evaluation of dead volume in the microfluidic layer of DECAF by adding A) 100 µL or B) 200 µL dark blue dye aqueous solution at the DECAF paper base. C,D) SWVs measured at different times after adding C) 100 µL or D) 200 µL of 3 mM caffeine in PW solution to the base of DECAF. E,F) Temperature images of DECAF tested with E) cold water, and with F) hot water.

2.8. Temperature Studies on Device

When measuring caffeine in commercial beverages, it is essential to consider the broad temperature variations between products and their potential influence on caffeine quantification by the DECAF device. Caffeinated beverages can range from hot (up to 80 °C), as in coffee and tea, to cold (as low as 4 °C), as in iced coffee, energy drinks, and sodas.^{[ 91} , ^{92 ]} Temperature variations can significantly affect SWV measurements, as temperature influences the diffusion, reaction kinetics, and electron transfer rates of the analyte, potentially altering both the peak current intensity and peak potential.^{[ 93} , ⁹⁴ , ⁹⁵ , ^{96 ]} Since the calibration curve was established in PW at room temperature (approximately 25 °C), it is crucial to ensure that samples reaching the sensing zone are at a similar temperature for accurate quantification. To address this, DECAF's temperature‐stabilizing capability was evaluated in both hot and cold beverage scenarios as it is immersed directly, consistent with its intended use (Figure S5, Supporting Information). Thermal response was analyzed using a FLIR C3‐X infrared camera. For the cold test, a 15 oz ceramic mug filled with PW at 9 °C was used (Figure 6E). Temperature measurements along the device showed that after 1 min, the fluid at the base reached 18.2 °C; at the stem after 2 min, it was 18.4 °C; and at the top after 5 min, it reached 19.9 °C. For the hot test, the mug was filled with PW at 73.2 °C (Figure 6F). Measurements indicated that after 1 min, the fluid at the base was 43 °C; at the stem after 2 min, 27.4 °C; and at the top after 5 min, it equilibrated to 24.2 °C. These results demonstrate that DECAF effectively stabilizes the sample temperature, bringing both cold and hot fluids close to room temperature within minutes of entering the device. Therefore a correction factor is not required, as the DECAF system is designed to actively stabilize temperature fluctuations during fluid transport through the microfluidic channel. This temperature stabilization is attributed to the high surface area‐to‐volume ratio of the paper‐glass fiber microfluidic channel, which promotes efficient heat exchange with the surrounding air. The small fluid volume lacks sufficient thermal mass to retain its initial temperature, leading to rapid equilibration with ambient conditions through passive heat dissipation.

2.9. Electrochemical Detection of Caffeine in Real Samples

After completing comprehensive laboratory evaluations of the DECAF platform, including assessments in PW, selectivity studies, fluid flow analysis, pH variation, and both elevated and reduced temperatures to simulate a range of conditions found in commercial beverages, we then proceeded to test DECAF using real‐world samples. This phase aimed to evaluate the platform's performance in quantifying caffeine under non‐laboratory‐controlled conditions representative of day‐to‐day consumer products. To encompass a broad spectrum of commonly consumed caffeinated beverages, we categorized the test samples into two groups: (1) self‐prepared drinks, where the caffeine content can vary depending on user preparation (Starbucks medium roast coffee, latte, cold brew, and black tea), and (2) industrially prepared beverages, which have standardized caffeine levels due to consistent manufacturing processes (Diet Coca‐Cola, Celsius, and Red Bull). All samples were tested directly the same day they were purchased or within the expiration date, without any dilution or pretreatment.

Caffeine quantification was performed using SWV, and concentrations were determined using a calibration curve generated from standard caffeine solutions in PW. To validate the system's accuracy, recovery studies were conducted by spiking each beverage sample with 1 mM caffeine and reanalyzing under identical conditions. These recovery experiments are critical for assessing the sensor's reliability and accuracy in complex sample matrices. As shown in Table  3 , the recovery values for all tested samples were within 20%, demonstrating the platform's acceptable accuracy and robustness across diverse beverage types and minimal matrix interference on detection performance. Additionally, Figure S6 (Supporting Information), shows the peak currents measured for each real sample. While the recovery results were encouraging, some discrepancies were observed between the measured caffeine content and the manufacturer‐labeled values. These deviations may be attributed to batch‐to‐batch variability, differences in sample handling or preparation, or the gradual degradation of caffeine upon exposure to oxygen.^{[ 97} , ^{98 ]} Samples such as latte and red bull contain significant levels of electrolytes which can increase the ionic strength and overall conductivity of the solution. This enhanced conductivity reduces solution resistance and facilitates more efficient charge transfer at the electrode–electrolyte interface, thereby increasing the peak current. For future work, we would perform individual calibration curves for samples that show significant discrepancy from the label. In addition, benchmarking DECAF's measurements against gold standard analytical techniques such as HPLC and GC‐MS to further validate the accuracy and real‐world applicability of the platform. While we acknowledge the importance of comparisons with established methods such as HPLC, we chose to validate the accuracy and reliability of our device through recovery studies. Recovery analysis is a widely accepted approach for assessing the performance of analytical devices, particularly in real‐sample detection scenarios. As such, it provides a robust and practical validation of our sensor's applicability in real‐world conditions.

Table 3.

Caffeine quantification results in real beverage samples using the DECAF sensor (n = 3).

Name	Labeled [mM]	Measured [mM]	Spiked [mM]	Measured after spiking [mM]	Recovery rate
Medium Roast	3.32	1.01 ± 0.05	1.00	1.93 ± 0.30	92%
Diet Coke	0.66	0.68 ± 0.11	1.00	1.74 ± 0.13	106%
Cold Brew	2.24	2.73 ± 0.11	1.00	3.81 ± 0.56	108%
Red Bull	1.66	3.30 ± 0.35	1.00	4.44 ± 0.32	114%
Celsius	2.90	2.99 ± 0.22	1.00	4.13 ± 0.69	115%
Black Tea	1.02	0.63 ± 0.18	1.00	1.77 ± 0.77	115%
Latte	1.09	4.19 ± 0.17	1.00	5.35 ± 0.32	116%

Additionally, fluid viscosity can significantly influence the response of SWV. Increased viscosity reduces mass transport rates and slows diffusion of analytes toward the electrode surface, which typically results in decreased peak currents, broader peak shapes, and possible shifts in peak potentials.^{[ 99} , ^{100 ]} Most of the caffeinated beverages tested exhibit viscosities comparable to that of the PW used for the calibration curve ( 0.9 mPa·s at 25 °C) and thus are not expected to introduce significant deviations in electrochemical response. However, the latte sample warrants further investigation in future studies due to its higher viscosity ( 2.0 mPa·s), attributed to the presence of milk proteins, lipids, and sugars, which may influence both fluid transport and electrochemical behavior. Beyond its effect on the SWV signal, higher viscosity would also prolong the waiting time for our DECAF device, as the thicker fluid resists flow through the microfluidic channel.

3. Conclusion

In this work, we developed a portable, miniaturized sensor–DECAF–for convenient, on‐site caffeine monitoring aimed at supporting informed health decisions and reducing the risks associated with caffeine overconsumption. The device enables real‐time measurement of caffeine in both self‐prepared and commercially available beverages without requiring complex or expensive instrumentation. The sensor exhibited a detection limit of 0.0623 mM and a limit of quantification of 0.208 mM, well‐suited for the concentration ranges typically found in caffeinated beverages. It demonstrated high selectivity against common interferents present in popular drinks, with no disruption to the caffeine oxidation signal observed between +1.1 and +1.3 V vs. Ag/AgCl. The system also showed robust performance across varying pH and temperature conditions, maintaining reliability across a wide range of real beverage samples.

The DECAF platform leverages Nafion‐modified LIG electrodes and a preloaded PBS‐GF layer, requiring only 200 µL of sample and 5 minutes per test, making it highly suitable for on‐site quantification. DECAF represents a low‐cost, user‐friendly solution for monitoring daily caffeine intake, with potential to reduce reliance on clinical visits, support proactive health management, and improve overall well‐being. Its affordability is further enhanced by the mask‐free, scalable fabrication of LIG electrodes and the use of inexpensive, readily available materials. Looking ahead, this approach can be extended to detect other target molecules and be adapted for analysis in biofluids, opening avenues for broader applications in both consumer health and clinical diagnostics. Future work will focus on enabling direct analysis of caffeine metabolites from biological fluids to monitor real‐time physiological intake.

4. Experimental Section

Materials and Reagents

Caffeine, glucose, dextrin, glycine, fructose, aspartame, benzoic acid, acesulfame‐K, potassium chloride (KCl), potassium phosphate monobasic (KH₂PO₄), potassium phosphate dibasic (K₂HPO₄), and potassium ferricyanide/ferrocyanide ([Fe(CN)₆]^{3 − /4 −}) were all obtained from Sigma‐Aldrich (St. Louis, MO, USA). Phosphate‐buffered saline (1× PBS) with pH = 7.4 was purchased from Avantor (Radnor, PA, USA). Deionized water (Resistivity: 18.20 MΩ ·cm) was used throughout all experiments. Beverage samples, including Aquafina Purified Water (PW), Earl Grey black tea, Diet Coca‐Cola, Red Bull, CELSIUS Fizz Free Peach Mango Green Tea, Starbucks Cold Brew, Starbucks Pike Place Medium Roast, and Starbucks Latte with 2% milk were acquired from local stores in Los Angeles, CA, USA. Electrical‐grade Kapton polyimide film (12 in × 12 in × 0.005 in) was obtained from McMaster‐Carr (Elmhurst, IL, USA). Silver/silver chloride ink (AGCL‐1134) and compatible ink thinner (102‐03) were supplied by Kayaku Advanced Materials (Westborough, MA, USA).The glass fiber conjugate pad (20 cm × 30 cm × 0.43 mm) was obtained from EMD Millipore (Burlington, MA, USA). Wood stir bars were purchased from Amazon (Seattle, WA, USA). Transparent adhesive film (Tegaderm) was obtained from 3M Health Care (Brookings, SD, USA). Filter paper was obtained from Whatman (Little Chalfont, Buckinghamshire, UK), and PET adhesive sheets were acquired from Fellowes (Itasca, IL, USA).

Electrode Fabrication

Polyimide (PI) film was sequentially cleaned with acetone, isopropyl alcohol, and deionized (DI) water, then dried at 80°C for 10 min. The cleaned PI substrate was placed into a laser engraver (VLS2.30, Universal Laser System Inc., USA) equipped with a 30 W CO₂ laser source operating at a wavelength of 9.3 µm. Electrode patterns were designed using Adobe Illustrator (Adobe Inc., USA) and engraved onto the PI film under raster mode. After engraving, the electrodes were rinsed again with DI water and dried at 80°C for an additional 10 minutes. To define the electrode surface area, Kapton adhesive tape was applied as a masking layer. The reference electrode was fabricated by drop‐casting 0.2 µL Ag/AgCl ink, diluted 1:1 (v/v) with compatible ink thinner, onto a designated electrode area, followed by drying at 80°C for 30 min. The electrode was then oven‐dried following the manufacturer's protocol to stabilize the Ag/AgCl layer and ensure a reproducible reference potential, as demonstrated in our previous work.^{[ 38 ]} For the working electrode, 5 µL of a 0.1% (v/v) Nafion solution in DI water was drop‐cast and dried in a vacuum chamber for 1 hour. After drying, the Nafion‐modified electrode was cut and prepared for use. Glass fiber pads and filter paper were laser‐cut into 1.2 cm × 1.5 cm pieces using the same laser engraver under vector mode. Each GF pad was pre‐wetted with 100 µL of 1× phosphate‐buffered saline (PBS) with pH = 7.4 and left to dry at room temperature for later integration with the LIG electrodes in the DECAF system. The storage conditions for both the dried PBS‐GF layer and the LIG electrodes were identical. Each component was placed inside a transparent polystyrene Petri dish, sealed with Parafilm (Neenah, Wisconsin, USA), and stored in a dark drawer at room temperature, away from direct light. Both the PBS‐GF layers and LIG electrodes were designated for single use and were not reused after electrochemical testing.

Electrochemical Tests

All experiments in this work including Cyclic voltammetry (CV) and square wave voltammetry (SWV) measurements were performed using a CHI 760E electrochemical workstation (CH Instruments, TX, USA) by applying 100 µL solutions to PBS‐GF layer on top of Nafion‐modified LIG. Supporting Wireless in situ measurements on the DECAF platform were conducted using the Sensit BT device (PalmSens BV, Utrecht, Netherlands). SWV settings were optimized (Figure S2, Supporting Information) and included a potential window from 0.9 V to 1.6 V, with an increment of 0.004 V, amplitude of 0.1 V, frequency of 5 Hz, and a quiet time of 2 s. The pH of each solution was measured using an Orion Star A211 benchtop pH meter (Thermo Scientific Inc., USA).

Material Characterization

Raman spectroscopy analysis was conducted using a Horiba XploRA Raman Microscope System (Horiba, Japan). X‐ray photoelectron spectroscopy (XPS) was performed using a Kratos Axis Ultra DLD instrument (Kratos Analytical, UK). Scanning electron microscopy (SEM) imaging was carried out using a Apero 2 SEM (ThermoFisher Scintific, MA, USA) operated at an accelerating voltage of 10 keV and 0.80 nA current. Energy‐dispersive X‐ray spectroscopy (EDX) characterization was conducted with an Oxford UltimMax 170 Silicon Drift Detector (Oxford Instruments, UK) at 10 keV and a spot size of 4.0.

Data Processing and Statistical Analysis

The capacitive background current, an undesirable signal component in voltammetry tests, appears as the background signal and must be removed prior to peak height analysis. Therefore, baseline correction is essential for improving the accuracy of voltammetric sensors. A polynomial baseline correction algorithm was applied using PeakUtils, an open‐source python library.^{[ 101 ]} A linear background is fitted to the signal, and the peak height is measured after subtracting this background. Each measurement was performed using a minimum of three independently fabricated electrodes, and results are reported as the average value with the corresponding standard deviation.

Conflict of Interest

The authors declare no conflict of interest.

Supporting information

Supporting Information

Vargas Ramos D., Ma H., Khazaee Nejad S., et al. “Device for Electrochemical Caffeine Analysis in Fluids (DECAF): A Stir‐Bar‐Embedded Sensor for Real‐Time Caffeine Analysis in Commercial and Homemade Beverages.” Small 21, no. 45 (2025): e06439. 10.1002/smll.202506439

Data Availability Statement

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