Mid-infrared (mid-IR) spectroscopy offers high sensitivity and selectivity for molecular analysis in gas or liquid phases, making it valuable for studying chemical reactions, particularly in biomedical applications like drug production. However, monitoring dynamic processes in liquids often relies on bulky systems requiring time-consuming offline analysis. Existing techniques face challenges in addressing broad absorption bands in liquids, especially at low analyte concentrations or with rapidly changing concentrations during reactions or conformational changes. Ideal sensors for dynamic liquid-phase processes require rapid response times, high sensitivity and specificity, and the ability to analyze wide concentration ranges in microliter samples. Targeting the mid-IR spectral fingerprint region, particularly the protein amide I band (~1600–1700 cm⁻¹ for protein analysis), enhances sensor specificity. Improving sensitivity necessitates maximizing the light's interaction length within the sample, which is often limited to low micrometer scales in aqueous solutions. High-power light sources like quantum cascade lasers (QCLs) and high-performance detectors such as quantum cascade detectors (QCDs) are crucial for addressing these limitations. While sensor specificity and sensitivity have been addressed in prior work, this study focuses on advancing two critical features: (i) the ability to monitor dynamic processes with high temporal resolution using an in situ sensor for label-free real-time measurements, eliminating offline analytics; and (ii) miniaturization to analyze minute liquid amounts, enabling online microliter-sample measurements with minimal interference. This paper introduces a fully monolithic integrated mid-IR sensor combining these features. By integrating the laser, interaction region, and detector on a single chip, and leveraging plasmonic waveguides to overcome diffraction limitations, a fingertip-sized (<5 × 5 mm²) sensor is realized. Simulations confirm the plasmonic capabilities in a liquid environment and the suitability of spectrally optimized QCLDs. The sensor's performance is characterized using bovine serum albumin (BSA) in heavy water (D₂O) through concentration series and thermal denaturation experiments. The study quantitatively assesses the sensor's limit of detection (LOD), linearity, concentration range, sample volume, and robustness to direct analyte exposure. Finally, the sensor's operation is demonstrated in normal water (H₂O) to assess real-world applicability.

Several studies have explored mid-IR spectroscopy for chemical sensing, particularly using quantum cascade lasers (QCLs) for gas-phase analysis. However, liquid-phase detection using mid-IR techniques remains less developed due to the challenges posed by broad absorption bands and high densities in liquid media. Previous research has investigated the use of QCLs and other mid-IR technologies for protein analysis, focusing on aspects like sensitivity and specificity. However, many of these approaches involve bulky instrumentation and/or time-consuming offline analysis. The development of miniaturized, integrated devices for real-time, in situ monitoring of dynamic processes in liquids remains an active area of research. Some studies have explored methods such as attenuated total reflection (ATR) spectroscopy and surface-enhanced infrared absorption spectroscopy (SEIRAS), but these often have limitations in terms of sensitivity, the ability to perform real-time monitoring, and miniaturization potential. This study builds upon previous research by demonstrating a fully integrated chip-scale sensor that addresses many of these limitations.

The study utilizes a fully monolithic integrated mid-infrared (mid-IR) sensor based on quantum cascade laser detector (QCLD) technology. This sensor combines a quantum cascade laser (QCL) emitter, a dielectric-loaded surface plasmon polariton (DLSPP) waveguide, and a quantum cascade detector (QCD) on a single chip, measuring approximately 5 x 5 mm². The DLSPP waveguide, composed of a thin silicon nitride (SiN) layer on a gold (Au) layer, acts as the interaction region where the mid-IR light interacts with the analyte. Finite Element Method (FEM) simulations using COMSOL software were used to model the propagation of the plasmonic mode in both air and D₂O, confirming the waveguide's suitability for liquid spectroscopy. Two sets of experiments were conducted: a concentration series and a thermal denaturation experiment, both using bovine serum albumin (BSA) in D₂O. For the concentration series, a peristaltic pump continuously added BSA stock solution (150 mg/ml in D₂O) to a beaker of pure D₂O, while the QCLD sensor was submerged in the solution. The sensor's absorbance was measured at 1597 cm⁻¹ using a 350 MHz oscilloscope. The temperature was monitored using an on-chip temperature probe and an external Pt100 probe. For the thermal denaturation experiment, a custom-made 60 µl microfluidic flow cell was used, with the QCLD sensor positioned within. A BSA solution (20, 40, or 60 mg/ml in D₂O) was constantly heated from room temperature to 90 °C and pumped through the cell. The absorbance at 1620 cm⁻¹ was measured using the QCLD sensor. The data was fitted to a sigmoidal Boltzmann function to extract the transition temperature (x₀) of the protein unfolding. To investigate real-life conditions, the sensor was also submerged in deionized water (H₂O). This experiment primarily focused on assessing the crosstalk in the sensor, which was utilized for crosstalk correction in the main experiments. Attenuated total reflection Fourier-transform infrared (ATR-FTIR) spectroscopy was employed as a reference method for comparison and calibration. This method used a Bruker Tensor 37 FTIR spectrometer with a diamond ATR crystal and was utilized to obtain the molar decadic absorption coefficient (ε) and effective path length (d_eff) for quantitative analysis.

The QCLD sensor demonstrated excellent performance in monitoring BSA in D₂O. The concentration series showed a linear relationship between absorbance and concentration over more than three orders of magnitude, from 75 µg/ml to 92 mg/ml. The sensor exhibited a remarkably low limit of detection (LOD) of 75 ppm by weight (0.0075% m/v), approximately 120 times lower than that of the ATR-FTIR reference system. The effective path length (d_eff) of the sensor, determined by comparing its absorbance to the ATR-FTIR reference, was 43.1 µm, consistent with the 48 µm length of the plasmonic waveguide. The thermal denaturation experiments revealed the expected sigmoidal shape of absorbance changes with increasing temperature, confirming previous findings. The transition temperature (x₀) was found to be concentration-dependent, decreasing with increasing BSA concentration. Notably, the sensor operated stably for more than 40 hours in continuous liquid exposure, including measurements with and without applied bias, demonstrating its robustness. FEM simulations confirmed the sensor's excellent suitability for in situ operation in a D₂O matrix, and the experiment conducted in deionized water (H₂O) showed that the sensor remains functional even when the optical mode is fully absorbed by the water. The remaining detector signal in this scenario was employed for crosstalk correction.

The findings demonstrate the successful development and validation of a next-generation, fully integrated mid-IR lab-on-a-chip sensor for real-time, in situ analysis of dynamic processes in liquids. The sensor's exceptional sensitivity, wide dynamic range, and microliter-scale sample requirements represent significant improvements over traditional bulky systems. The low LOD and high absorbance compared to existing techniques highlight the potential of this technology for various applications. The concentration-dependent transition temperature observed in the thermal denaturation experiments aligns with literature and validates the accuracy of the measurements. The successful operation in both D₂O and H₂O demonstrates the sensor's versatility and potential for studying biophysical processes under native conditions. The high stability observed during prolonged liquid exposure highlights the robustness and reliability of the device for practical applications.

This research presents a miniaturized, highly sensitive mid-IR lab-on-a-chip sensor for real-time monitoring of liquid-phase reactions. The sensor's superior performance, demonstrated through concentration series and thermal denaturation studies, opens possibilities for diverse applications. Future work will focus on optimizing the device for measurements in aqueous solutions (H₂O) by adjusting the plasmonic waveguide geometry and selecting appropriate wavelengths to mitigate water absorption.

The current study primarily focuses on BSA in D₂O. While the D₂O matrix provides a good environment for spectroscopic analysis, it does not perfectly replicate the native biophysical conditions of H₂O. Further studies are needed to fully evaluate the sensor's performance in H₂O, requiring optimization of the waveguide geometry and wavelength selection. The sensor currently operates at specific wavelengths and does not provide full spectral information. Implementation of a QCL array could overcome this limitation.