Modulair ontwerp voor microgolf diëlektrische karakterisatie van watergebaseerde materialen in de levenswetenschappen

Microwave-microfluidic systems offer a compelling route toward integrated, label-free sensing and contactless heating of liquid and biological samples. Their ability to simultaneously monitor dielectric properties and induce localized thermal actuation is particularly attractive for lab-on-chip platforms. However, existing solutions often remain limited in frequency range, address only single samples, or require complex calibration procedures that hinder their adaptability across different sample types. This thesis presents the development and validation of a modular microwave platform capable of concurrent multi-sample dielectric sensing and selective heating within the gigahertz frequency range. The system operates across several independent channels, allowing simultaneous measurements and thermal control of microliter-scale liquid samples. The sensor-heater units were designed to resonate at distinct frequencies between 3.8 and 8 GHz, enabling spectral isolation and parallel operation. Each unit was configured to measure the dielectric properties of individual samples with volumes on the order of 10-20 uL. Through resonance analysis, both the real and imaginary components of the permittivity were extracted with high precision across a wide range of ethanol-water mixtures. Simultaneously, independent heating of selected channels was demonstrated using controlled microwave power delivery. Temperature control was achieved over a range from room temperature up to 60 °C with heating rates as high as 25 °C and overshoot consistently below 5%, enabling reproducible thermal cycling suitable for thermo-responsive material studies. To extend the dielectric characterization beyond narrowband measurements, a broadband calibration method was introduced based on two-port scattering parameters and a minimal set of liquid standards. This approach provided accurate retrieval of complex permittivity within localized spans of several hundred megahertz to a few gigahertz around each resonance without the need for extensive electromagnetic simulations or full de-embedding processes. In addition, machine learning techniques were employed to model the concentration-dependent permittivity behavior of ethanol-water mixtures. A neural network was trained on a sparse dataset to predict unknown concentrations, achieving errors below 5%. Furthermore, a two-stage algorithm combining classification and regression was used to estimate both temperature and concentration, successfully disentangling their overlapping effects on the dielectric spectrum. The developed platform was also applied to the study of thermo-responsive hydrogels. Broadband measurements up to 50 GHz enabled the extraction of Cole-Cole relaxation parameters across different gel concentrations. For dynamic analysis, the narrowband sensor-heater units monitored the temperature-dependent dielectric response of hydrogels under controlled heating and cooling cycles, covering a range from approximately 20 °C to 40 °C, relevant for many biomedical applications. A clear hysteresis was observed in the dielectric properties as a function of temperature, which was effectively captured using a memory polynomial model. By accounting for dynamic thermal conditions, this approach improves the interpretation of dielectric sensing and enhances the control and efficiency of microwave-based heating in the presence of hydrogels. Additionally, the platform was used to investigate non-homogeneous suspensions of polystyrene microspheres, selected for their size similarity to biological cells. Suspensions containing 10 um and 15 um beads at varying concentrations were monitored over time as sedimentation occurred. The resulting dielectric response showed measurable shifts in resonance parameters (up to 5.9 MHz in frequency for the largest beads at the highest concentration), demonstrating sensitivity to both particle size and redistribution dynamics. This application highlights the system's ability to track time-dependent particle movement through label-free microwave sensing, providing a baseline for future studies on more complex biological suspensions. In summary, this thesis introduces a compact and scalable microwave platform that enables real-time, multi-parameter characterization of liquid and biological samples, as well as dynamic processes such as thermal hysteresis in hydrogels and particle sedimentation in suspensions. By combining multi-resonant sensor-heater units with efficient calibration and machine learning based modeling, the system extends the capabilities of microwave-microfluidic devices. These contributions pave the way for advanced lab-on-chip solutions in diagnostics, material analysis, and biomedical engineering, where simultaneous sensing and actuation are critical.