A dual-band RF exposure system for real-time electrophysiological recordings using microelectrode array device

Fifth-generation (5G) wireless communication systems operating in the FR1 band have increased interest in studying bioelectromagnetic interactions under controlled exposure conditions. MicroElectrode Array (MEA)-based electrophysiology provides a suitable approach for real-time investigations of neuronal activity, although its integration with RF exposure systems remains technically challenging. This work presents the design, integration, and numerical characterization of a dual-band RF exposure system for real-time electrophysiological recordings using MEA devices. The system operates at 0.7 GHz and 3.5 GHz and is based on an open-ended coaxial radiator, termed Localized Radiating-End Coaxial line (LoREC). Full-wave COMSOL simulations were performed to evaluate field confinement, dosimetric quantities (SAR), and compatibility with the MEA acquisition setup. The proposed exposure system achieved stable impedance matching and effective confinement of the electromagnetic field within the biological target region. SAR analysis showed frequency-dependent exposure patterns with reproducible and spatially consistent distributions across the active electrode area. Simulations also demonstrated negligible residual fields in proximity to the acquisition electronics, indicating limited electromagnetic interference with electrophysiological recordings. The proposed exposure system provides a controlled and reproducible exposure architecture suitable for real-time RF exposure and MEA electrophysiological measurements under physiological conditions. The numerical results demonstrated stable near-field confinement within the MEA culture region, predictable SAR distributions, and negligible residual fields on the acquisition electronics across the investigated frequency range. The system is therefore suitable for investigating frequency-dependent bioelectromagnetic effects associated with 5G-FR1 technologies.

Fifth generation (5G) wireless communication systems have introduced higher carrier frequencies and larger bandwidths than previous generations (Sanfilippo et al., 2018; Dangi et al., 2021; Liberti et al., 2023; Azcorra, 2025). 5G frequencies within the FR1 band (0.45 GHz–6 GHz, recently extended up to 7.125 GHz), with a focus on 0.7 GHz and 3.5 GHz, are among the most relevant for commercial use and exposure studies (Costanzo and Masotti, 2017; Rappaport et al., 2017; Persia et al., 2018; Shafique et al., 2020; Wang et al., 2021; Torres-Ruiz et al., 2024). In particular, 0.7 GHz enables deeper penetration, whereas 3.5 GHz corresponds to one of the most widely deployed 5G bands and is commonly adopted in experimental and dosimetric studies to ensure realistic and comparable exposure conditions. Compared to other frequencies within the FR1 band (e.g., 2.4 GHz or 5 GHz), this choice ensures consistency with current 5G deployment scenarios and with the broader research framework. This technological expansion has renewed interest in the study of bioelectromagnetic interactions and in the accurate characterization of RF exposure also in the sub-6 GHz range, where EM energy still penetrates biological tissue and exposure standards and time-modulation schemes differ from past generations (Simkó and Mattsson, 2019; Arnaud-Cormos et al., 2022; Jain and Jain, 2022; Joushomme et al., 2023; Torres-Ruiz et al., 2024). For these frequencies, international safety guidelines use the Specific Absorption Rate (SAR), as the basic (i.e., the reference dosimetric quantity) for assessing RF energy absorption and ensuring compliance with safety standards (ICNIRP, 2020).

Recent European initiatives, such as the GOLIAT project (GOLIAT Project, n.d.), have emphasized the importance of bridging engineering, biological, and epidemiological perspectives to characterize 5G-related exposures and their possible biological implications. Within this multidisciplinary framework, in vitro approaches play a fundamental role, allowing well-controlled and reproducible conditions that are crucial for correlating exposure metrics with potential biological effects (Romeo et al., 2022; Canovi et al., 2023; Eskandani and Zibaii, 2023; Joushomme et al., 2023; Torres-Ruiz et al., 2024). Compared with in vivo or epidemiological investigations, in vitro models enable precise control of environmental parameters and provide high spatial and temporal resolution for detecting subtle bioelectromagnetic responses (Simkó et al., 2016; Ravin et al., 2020).

Among the available experimental methods, MicroElectrode Array (MEA) systems have become a reference technology for studying the electrophysiological behaviour of neuronal networks. MEA devices allow the simultaneous recording and stimulation of extracellular activity across multiple sites, enabling the investigation of network connectivity, excitability, and plasticity under controlled conditions (Nefzi et al., 2019; Nefzi et al., 2022; Zhao et al., 2020)). Their non-invasive and long-term recording capability makes them particularly suited for monitoring dynamic changes in neuronal activity during controlled EM exposure. Through MEA-based electrophysiology, it is possible to explore whether RF fields may modulate cellular communication pathways or influence neural activity patterns, thus offering a sensitive model for identifying potential bioelectromagnetic effects at the cellular level (Obien et al., 2015; Nefzi et al., 2020; Mossink et al., 2021; Canovi et al., 2023).

However, integrating MEA systems within an RF setup for real-time exposure is technically challenging. The presence of metallic and electronic components required for signal acquisition can introduce electromagnetic interference (EMI) and distort the exposure field distribution, thereby affecting both dosimetric precision and signal quality (Chou et al., 1996; Paffi et al., 2010; Crouzier et al., 2012; Hansen et al., 2015; Engineering Shielding, 2023). Exposure systems that ensure homogeneous and reproducible fields, such as transverse electromagnetic (TEM) cells, are well established for biological dosimetry (Merla et al., 2011; Nefzi et al., 2022), but their design constraints and shielding configurations make them less suitable for real-time electrophysiological measurements. Furthermore, most MEA-based studies under RF exposure have been conducted at different frequencies (typically 0.9 GHz or 1.8 GHz), corresponding to previous communication standards, such as GSM or LTE (El Khoueiry et al., 2018; Zhao et al., 2020; Nefzi et al., 2022). While some of these advanced systems have demonstrated the feasibility of combining exposure and recording, they are not optimized for the frequency range and signal characteristics of 5G technologies, and the electrophysiological data acquisition is typically performed post-exposure rather than in real-time.

To address this limitation, this work presents a dedicated dual-band exposure architecture tailored for in vitro MEA electrophysiology at 0.7 GHz and 3.5 GHz. The proposed system integrates an open-ended coaxial structure, termed Localized Radiating-End Coaxial line (LoREC), designed to produce a confined and controlled near-field exposure while maintaining EM compatibility with MEA acquisition electronics. The LoREC system operates inside an incubator environment, ensuring physiological conditions and enabling real-time neuronal recordings during RF exposure. By combining field confinement and dosimetric reliability, the proposed exposure system aims to establish a controlled experimental platform for studying dynamic and frequency-dependent biological responses to 5G related EM fields. Numerical simulations confirm its performance and experimental validation of the prototype is currently ongoing and will be addressed in future work. This work contributes to the development of real-time bioelectromagnetic studies using integrated engineering and biological platforms that directly link RF dosimetry with electrophysiological outcomes.

CAD modelling of the MEA recording setup

The numerical reconstruction of the biological experimental setup was carried out in COMSOL Multiphysics v.6.2, reproducing the configuration available at the laboratory 1 where real-time electrophysiological experiments are routinely conducted. The model was designed to ensure geometric and EM consistency with the real setup, allowing accurate simulation of the exposure scenario.

The MEA recording setup is shown in Figure 1. Typically, the MEA device is positioned on the stage of an inverted optical microscope to allow simultaneous optical and electrophysiological recordings (Figure 1A). In this work, the MEA setup was positioned outside the microscope (Figure 1B). To numerically replicate the biological scenario, each component of the MEA recording setup was modelled individually and then assembled in COMSOL to obtain the “virtual” experimental scenario (Figure 2). A direct comparison between photos and 3D models was used to validate geometries and dimensions. The setup is composed of six main parts, reported in Figure 2: (1) the homemade metallic baseplate, (2) the stage top incubator for optical microscopes (Okolab s. r.l., Pozzuoli, NA, Italy), (3) the MEA (60MEA200/30iR-ITO, Multichannel System, HBio, Reutlingen, BW, Germany) with its glass chamber for cell culture, (4) the pogo-pin supporting structure, (5) the pogo-pins (metallic spring contacts), and (6) the multilayer PCB.

The incubator (2) consists of a rectangular metallic enclosure with a transparent glass top cover that allows visual inspection of the culture during measurements. Side openings allow the passage of cables and water tubes for thermal regulation. The bottom side is open to accommodate the baseplate, which is machined from aluminium and provides mechanical support and thermal stability. Its central square cavity hosts the MEA (3), while a circular aperture (24 mm diameter, 1 mm thickness) aligns with the MEA well to allow optical inspection of the culture through the microscope objective.

The MEAs electrodes are embedded in a very thin glass substrate on a robust ceramic carrier; 59 electrodes and one reference electrode are arranged in an 8 × 8 layout grid with electrode diameters of 30 μm and inter electrode distances of 200 μm creating an active area of 38 × 38 mm 2. The flat, round electrodes are made of titanium nitride (TiN). Tracks and contact pads are made of transparent indium tin oxide (ITO); insulation material is silicon nitride. The central glass chamber (I.D. 19 mm; O.D. 24 mm; 6 mm deep) contains the target culture immersed in physiological saline solution (≈1 mL), which defines the region of interest for EM dosimetry.

The pogo-pin support (4) is made of plexiglass and acts as a nonconductive frame housing the spring contacts. Its function is to maintain alignment and electrical insulation between the MEA and the acquisition electronics. The pogo-pins (5), separately modelled as individual metallic cylinders, ensure electrical continuity between the MEA pads and the PCB. 64 pins are distributed in a square pattern corresponding to the electrode layout. Each pin contacts the MEA pad through its lower tip (protrusion of 0.1 mm) and the PCB through its upper terminal (protrusion of 0.2 mm).

Finally, the PCB (6) is made of an FR4 substrate (εr = 4.5, σ = 0.004 S/m) with internal copper planes and surface-mounted connectors for signal routing. A central circular aperture (24 mm diameter) aligns with the MEA chamber, maintaining optical visibility through the vertical axis of the setup.

For the purpose of the present work, the virtual experimental scenario was treated as a standalone unit positioned apart from the microscope. The content of the MEA’s glass chamber was modeled as a physiological saline solution. The complete list of geometrical parameters and material properties employed in the CAD model is provided in Table 1.

LoREC design

Coaxial radiators are advantageous in this context because their intrinsic shielding and single-ended feed allow stable excitation even in environments with high metal content and limited space. The proposed LoREC was designed to act as a compact and controllable source capable of generating a localized electromagnetic field in the region of the culture sample, while preserving full compatibility with the geometry and materials of the MEA-incubator assembly. The LoREC operates as an open-ended coaxial radiator, excited in TEM mode and terminated by a free aperture that couples the EM energy into the biological medium. Its geometry, materials, and feeding configuration were optimized to achieve stable impedance across the two operating frequencies of 0.7 GHz and 3.5 GHz, ensuring reproducible and localized exposure conditions.

Figure 3A illustrates the geometry of the LoREC, which consists of concentric conductors made of copper, separated by a PTFE dielectric layer (εr = 2.07). The chosen inner-to-outer radius ratio (R/r = 3.35) satisfies the impedance matching requirement for a nominal 50 Ω line while preventing higher-order mode propagation within the operational frequency range.

This configuration satisfies the quasi-TEM condition and supports frequencies well below the cutoff of higher-order modes. The lowest non-TEM mode (TE 11) has a cutoff frequency (fc), defined as (Equation 1): resulting in f c = 15.25 GHz. This value confirms pure TEM propagation at both operating frequencies (0.7 GHz and 3.5 GHz). The canonical TEM configuration is represented in Figure 4A, where the electric and magnetic field lines are orthogonal and symmetrically distributed between conductors, following the expected analytical pattern. While only the fundamental TEM mode propagates along the coaxial structure within the frequency range of interest, higher-order modes excited at the open-ended aperture are evanescent and are inherently accounted for in the full-wave solution, contributing only locally to the near-field distribution.

A quantitative comparison between analytical predictions and numerical results is summarized in Table 2 in terms of the characteristic impedance Z0, the Voltage Standing Wave Ratio (VSWR) and the Scattering Parameter |S11|. Analytically, the VSWR quantifies the mismatch between the characteristic impedance Z0 and the load impedance ZL, and it can be expressed as (Equation 2): where Γ = Z L − Z 0 Z L + Z 0 is the reflection coefficient.

The simulated |S11| (Figure 4B for the whole range) remained below −15 dB at both 0.7 GHz and 3.5 GHz, indicating good matching with the analytical design (Table 2). The characteristic impedance and the VSWR extracted from the simulation results are in close agreement with the theoretical target value. These results confirm that the analytical design methodology provides accurate parameterization of the LoREC geometry and ensures consistent performance over the considered frequency range.

Finally, to reduce undesired lateral spreading and improve directivity, a circular aluminium flange was implemented at the open end, extending radially by Rf = 6 mm beyond the outer conductor and with a thickness of 1 mm (Figure 3A). The flange dimension (∼20 mm) was defined considering both mechanical constraints, imposed by the baseplate aperture (Ø 24 mm), and electromagnetic considerations. In particular, the selected size ensures that the flange remains electrically small with respect to the operating wavelengths (λ ≈ 42.9 cm at 0.7 GHz and 8.6 cm at 3.5 GHz) and does not support resonant behavior within the frequency range of interest. The flange acts as a passive reflector, reinforcing the vertical component of the radiated field while shielding the line from backward leakage and coupling to the incubator walls.

Integration within the exposure scenario

As illustrated in Figure 3B, for integration within the bio-experimental environment, the LoREC was positioned vertically below the baseplate of the incubator, aligned with the central axis of the MEA. The open aperture of the coaxial line faced upward, oriented toward the physiological saline volume (εr = 75, σ = 1.79 S/m at 0.7 GHz and 4.36 S/m at 3.5 GHz (Atkins and De Paula, 2017; Pisano et al., 2023) that fills the MEA chamber, representing the conductive culture medium. Hence, the LoREC was inserted into the central aperture of the baseplate until the metallic flange rested against its lower surface, allowing the coaxial aperture to directly contact the bottom glass layer of the MEA. The external feeding port of the coaxial line extended beyond the incubator wall, where the connection to the RF instrumentation is to be implemented via a standard SMA connector. The metallic body of the incubator was modelled to account for potential reflective and coupling effects, ensuring realistic boundary conditions for field confinement and resonance behavior.

Numerical modelling and simulation was performed in COMSOL Multiphysics v.6.2 (RF Module). The excitation was defined at the lower end of the LoREC through a Port boundary condition, which imposes an ideal TEM mode with a reference impedance of 50 Ω (Pin = 10 mW). This formulation neglects connector transitions, cable effects, and generator characteristics, thereby enabling the evaluation of the intrinsic electromagnetic behavior of the LoREC structure. The conductive elements of the LoREC were modelled as Perfect Electric Conductors (PEC), while all the other materials in the experimental scenario were implemented consistently with their real counterparts. The MEA electrodes were modelled consistently with their physical characteristics, adopting realistic geometrical dimensions and material properties (TiN), as described in Section 2.2. The external computational domain was defined as a parallelepiped air volume of 180 × 200 × 100 mm 3, bounded by Scattering Boundary Conditions (SBC) to emulate open-space propagation and suppress reflections.

A custom mesh was defined to ensure accurate discretization of all geometrical features, including the smallest elements of the MEA structure. The mesh was locally refined in proximity of the MEA electrodes, the coaxial aperture, the saline region, and material interfaces, with a minimum element size of 0.04 mm, while progressively coarser elements were adopted in regions farther from the exposure area, up to a maximum size of 4 mm. The resulting computational domain comprised approximately 22.7 million tetrahedral elements, guaranteeing numerical convergence across the studied frequency range.

The reported results are obtained from deterministic full-wave simulations and refer to the excitation power of 10 mW used in numerical simulations. Variability is assessed through sensitivity analysis within the regions of interest. In particular, the uncertainty budget accounts for the influence of geometrical and positional variations.

The EM performance of the LoREC-based exposure system was evaluated at 0.7 GHz and 3.5 GHz, i.e., the 5G-FR1 target frequencies. Figure 5 reports the electric (|E|) field distribution within the MEA-incubator assembly at both frequencies.

In both cases, the field remains concentrated around the exposure region while it rapidly decreases in the rest of the incubator (Figure 5A). The metallic flange suppresses lateral spreading, yielding a compact exposure footprint around the central cavity. Moreover, the field intensity decays within the first few centimeters along the z-axis, and the PCB experiences only residual fields below 1 V/m at 0.7 GHz and 7 V/m at 3.5 GHz. Two cross-sections were extracted to visualize the lateral and axial confinement: an XY plane coincident with the electrode surface, and a YZ plane crossing the central axis of the system as reported in panels B and C of Figure 5 for 0.7 GHz and 3.5 GHz, respectively. From visual inspection of both planes, at 0.7 GHz, |E| is smooth and symmetric around the coaxial aperture, with evident attenuation both radially and axially across the saline solution. In contrast, at 3.5 GHz, the pattern is less confined, and the field extends beyond the saline interface, forming a dome-shaped region above the liquid (highlighted in the YZ plane) with peak values above 30 V/m, while still not exceeding a few volts per meter in proximity to the PCB. Some lateral spread is also visible around the saline boundaries, confirming that at higher frequency the field is less confined but remains directionally constrained by the flange geometry.

The SAR was calculated within the physiological saline solution filling the MEA chamber and analyzed on two orthogonal sections: an axial XY plane taken 10 µm above the MEA glass bottom, corresponding to the approximate position of the adherent neuronal layer, and a sagittal YZ plane crossing the saline volume (Figure 6). The regions of interest (ROIs) defined in Figure 6A correspond to functional areas of the MEA: ROI1 represents the central cellular domain where neurons are cultured, while ROI2, ROI3, and ROI4 correspond to the two orthogonal active stripes and to the square electrode region involved in electrophysiological recordings, aligned with the active microelectrode arrays. At 0.7 GHz, the SAR distribution is smooth and radially symmetric, with moderate gradients along the z-axis. ROI1, corresponding to the cell domain, is only partially covered by the main lobe, whereas the others are fully exposed. At 3.5 GHz, the absorption pattern expands laterally and upward, showing a more complex distribution characterized by a prominent hot region reaching the edge of the saline solution. This behaviour reflects the increased conductivity of the saline solution and the higher spatial variability introduced by the shorter wavelength, resulting in higher local amplitudes but reduced confinement compared to 0.7 GHz. Table 3 provides the quantitative SAR analysis for the four ROIs. The 99th percentile (P99) identifies peak absorption, the mean captures overall energy deposition, the standard deviation quantifies spatial dispersion, and the coefficient of variation (CV = std/mean) reflects intra-ROI inhomogeneity. At 0.7 GHz, the mean SAR ranges from 1.12 W/kg in ROI1 to 7.23 W/kg in ROI4, with moderate variability and relatively homogeneous exposure across the electrode regions. At 3.5 GHz, both the mean and the local maxima increase by approximately one order of magnitude, reaching 23.73 W/kg in ROI4. Despite this increase, the CV values remain low, indicating that the absorbed power is concentrated in a reproducible and spatially consistent region. The higher SAR values at 3.5 GHz are consistent with the combined effect of increased conductivity and stronger local fields observed in the |E| maps.

The SAR distributions represented by the boxplots in Figure 7 are consistent with the trends observed in the SAR maps. At 0.7 GHz, the medians follow an increasing trend from the central ROI1 to the peripheral ROI4, reflecting the gradual rise in energy deposition closer to the LoREC axis. The interquartile ranges are relatively narrow, indicating limited dispersion and a uniform exposure within each ROI. The low variability, combined with comparable 99th percentile and mean values, confirms that the 0.7 GHz configuration achieves a homogeneous electromagnetic energy deposition throughout the active electrode region.

At 3.5 GHz, the distributions become broader and more asymmetric, showing a higher spread between the lower and upper quartiles. The median values increase significantly, while the whiskers extend toward higher SAR values, consistent with the stronger localized peaks visible in the maps. This indicates that, despite the reduced uniformity, absorption remains concentrated in predictable localized “hot regions” above the MEA electrodes. The overall trend demonstrates that increasing the frequency enhances the energy density within the culture area while maintaining a reproducible exposure pattern across the different ROIs.

The exposure efficiency (η) was quantified as the ratio between the induced SAR and the input power in each ROI, and the calculated values are reported in Table 3. For an input power of 10 mW, despite the local variations, ROI1 provides a representative estimate of the average exposure experienced by the cell population monitored by the MEA system. Accordingly, the system can be considered to have a efficiency of 112 W kg -1 /W at 0.7 GHz and 274 W kg -1 /W at 3.5 GHz, indicating that the exposure efficiency is approximately 2.5 times higher at the higher frequency.

An additional analysis was performed to identify potential sources of variability in the calculated exposure efficiency. The uncertainty budget included a 1% variation in medium permittivity (ε), a 2% variation in conductivity (σ) (Gabriel and Peyman, 2006), and positional variations of the LoREC within the incubator baseplate. Specifically, shifts of ±0.5 mm were considered along the x and y-axes. Tables 4, 5 summarize the resulting SAR variations at 0.7 GHz and 3.5 GHz, respectively, for each parameter, as well as the corresponding sensitivities, defined as the ratio between the percentage change in SAR and the percentage change in the parameter. The same results are graphically summarized in Figure 8.

At 0.7 GHz, variations in the dielectric properties of the saline solution produce relatively smaller effects on the SAR. Changes in the relative permittivity led to SAR variations between 7.8% and 10.2%, with corresponding sensitivity values ranging from 7.8 to 10.2, indicating a moderate but consistent influence across the ROIs. Similarly, variations in σ produce SAR changes between 5.7% and 7.6%, with sensitivity values between 2.8 and 3.8, suggesting a lower impact compared with permittivity. In contrast, the lateral displacements (Δx and Δy) show more significant contributions depending on the region of interest. In particular, Δx produces SAR variations up to 34.6%, with sensitivity values reaching 16.5, while Δy leads to variations up to 36.0% with sensitivities up to 17.1. These results highlight the stronger influence of geometrical positioning on the SAR distribution.

A similar trend is observed at 3.5 GHz, where the influence of the dielectric parameters remains relatively limited. Variations in ε produce SAR changes up to 12.3%, with sensitivity values up to 12.3, while σ leads to variations up to 11.9%, with sensitivities up to 6.0. Conversely, the lateral positioning parameters show the highest impact: Δx produces SAR variations up to 49.2% with sensitivity values reaching 23.4, and Δy produces variations up to 49.6% with sensitivities up to 23.6. Overall, both the percentage SAR variations and the associated sensitivity values indicate that geometrical positioning is the dominant source of variability in the SAR estimation, particularly at the higher frequency.

The combined uncertainty of the estimated SAR was evaluated by combining the individual SAR variations reported in Tables 4, 5 using a root-sum-square approach. Considering ROI1, which represents the biologically relevant exposure region, the resulting expanded uncertainty (k = 2) is approximately 19.5% at 0.7 GHz and 27.5% at 3.5 GHz, supporting a reliable dosimetric assessment at both frequencies. The expanded uncertainty is mainly driven by positional sensitivity, particularly lateral misalignment between the LoREC and the MEA, which represents the dominant contribution to the overall variance in the near-field exposure configuration.

The study was motivated by the increasing scientific interest in understanding potential bioelectromagnetic interactions within the frequency range used by 5G-FR1 technologies (Simkó and Mattsson, 2019; Joushomme et al., 2023; Liberti et al., 2023). In this context, in vitro methodologies represent a suitable approach to investigate exposure effects under controlled and reproducible conditions, while allowing direct correlation between EM parameters and measurable biological outcomes. Among the available techniques, MEA devices offer a unique opportunity to evaluate how RF exposure may influence neuronal network activity through real-time monitoring of extracellular signals (Nefzi et al., 2019; Nefzi et al., 2020; Canovi et al., 2023).

Designing an exposure system compatible with MEA platforms, however, requires addressing several engineering constraints (Paffi et al., 2010). The presence of metallic microelectrodes and embedded electronic components increases the susceptibility to EM coupling, potentially introducing artefacts in the acquired signals. Moreover, the restricted geometry of the MEA chamber demands precise spatial control of the electric field to achieve relevant SAR levels within the culture volume without perturbing the acquisition electronics or the surrounding environment. The present work addressed these challenges by developing the Localized Radiating-End Coaxial line, a compact open-ended coaxial structure conceived to deliver a confined and reproducible near-field exposure specifically tailored for MEA-based electrophysiological investigations. By combining the simplicity of a coaxial geometry with the field-shaping effect of a metallic flange, LoREC provides a directional and frequency-scalable architecture optimized for integration with MEA devices.

The numerical assessment of the LoREC-based platform confirmed that this design approach provides both spatial control of the exposure and EM safety for the measurement environment. It should be noted that the proposed system is intended as a research platform for controlled in vitro studies and operates at low input power, ensuring safe and localized exposure conditions within the experimental setup. The field distributions showed that the radiated energy remains confined around the culture region, with negligible residual values on the PCB and surrounding components. This implies that the exposure is effectively restricted to the biological target, minimizing the likelihood of unwanted coupling with the acquisition circuit. The clear decay of |E| within the first few centimetres along the vertical direction demonstrates that the metallic flange successfully confines the field to the desired region, while the differences observed between the two operating frequencies reveal the expected dependence on wavelength and saline conductivity. This assessment is primarily based on radiated field analysis, in line with standard dosimetric approaches in RF exposure system design (Schuderer et al., 2004; Paffi et al., 2010; Hansen et al., 2020). While the results indicate low field levels in the vicinity of the PCB and interconnection region, suggesting limited radiative interaction with the acquisition electronics, conducted coupling mechanisms through electrical interconnections (e.g., pogo-pins and PCB tracks) are not explicitly addressed, as the model focuses on the radiated coupling and dosimetric impact of the LoREC system. A more detailed characterization of electromagnetic coupling through such interconnects would require a dedicated multiphysics analysis, including the specific details of the acquisition circuitry, and could therefore be the subject of a future study. In this context, it is worth noting that the characteristic length of the MEA interconnects is much smaller than the wavelength at the operating frequencies, and they therefore behave as electrically short conductors with limited radiation efficiency (Balanis, 2005; Pfeiffer, 2017). Consistently, the simulated field distribution shows that the electric field intensity in the interconnection region remains significantly lower than in the exposure region, suggesting a limited contribution of unintended coupling, although this aspect may be further investigated in future studies.”

At 0.7 GHz, the exposure field extends uniformly within the MEA chamber, producing moderate gradients and relatively homogeneous SAR values across the culture layer. Such a configuration is well suited for experiments targeting global responses of neuronal populations under spatially uniform RF exposure. Conversely, at 3.5 GHz, the field distribution becomes more structured, with increased local amplitudes and a partial expansion of the field above the saline interface, consistent with the dome-shaped region observed in the numerical maps. The resulting SAR pattern reveals a stronger but less homogeneous energy deposition offering the possibility to induce spatially selective modulation or probe threshold-dependent phenomena in specific sub-regions of the neuronal culture. This frequency-dependent behaviour aligns with previous computational studies reporting enhanced field localization and reduced penetration depth in conductive biological media at higher frequencies (Fiedler et al., 2018; Nefzi et al., 2020; Canovi et al., 2023; Joushomme et al., 2023). Importantly, the dosimetric analysis performed across the four ROIs demonstrated that the LoREC enables a controlled and localized exposure within the MEA chamber, with distinct and quantifiable SAR levels associated with each ROI. This localized control allows mapping the distribution of the absorbed power over the active region of the MEA, corresponding to the spatial arrangement of the recording microelectrodes. This characteristic is particularly relevant for electrophysiological studies, where local SAR levels can be correlated with specific recording electrodes, enabling a direct link between RF exposure and neuronal network responses. From a practical perspective, this dosimetric characterization provides guidance for experimental planning, allowing researchers to select microelectrodes located in regions associated with well-defined exposure levels and to compare neuronal responses under different local EM doses. The statistical analysis across the four ROIs confirms exposure reproducibility: low CV at both frequencies indicates that absorption is evenly distributed within each functional region of the MEA, ensuring reliable correlation between SAR values and electrode geometry. The 99th percentile provides an upper bound on localized peaks and potential thermal load, while the mean SAR represents the effective delivered dose. Together, these parameters demonstrate that LoREC achieves the desired balance between exposure intensity and uniformity, an outcome not commonly achieved in prior in vitro platforms employing TEM cells or waveguide-based exposure systems (Merla et al., 2011; Nefzi et al., 2019; Nefzi et al., 2022; Merla et al., 2011; Nefzi et al., 2019; Nefzi et al., 2022).

In addition to the nominal dosimetric characterization, a sensitivity analysis was conducted to assess the robustness of the exposure conditions with respect to variations in material properties and system positioning. The results indicate that uncertainties in the dielectric properties of the saline solution (ε and σ) produce relatively limited SAR variations, typically below about 12%. In contrast, geometrical parameters represent the dominant source of variability. By combining the individual contributions, the expanded uncertainty of the estimated SAR was obtained. The resulting values are comparable with those commonly reported for in vitro numerical dosimetry, indicating that the LoREC platform provides reliable and well-controlled exposure conditions.

From an engineering perspective, the low electric field levels detected on the PCB confirm that the LoREC maintains high EM compatibility, reducing the risk of EMI with the recording electronics. This is a key requirement for real-time electrophysiological investigations, where artefacts induced by the exposure system must remain well below the intrinsic amplitude of biological signals. The negligible coupling to the metallic and electronic parts validates the LoREC as an architecture capable of supporting simultaneous exposure and measurement without compromising signal fidelity.

Overall, the numerical results indicate that the proposed LoREC architecture provides a robust and adaptable exposure solution for MEA-based experiments. The system maintains controlled and reproducible dosimetric conditions across the investigated frequency range, demonstrating its capability to operate effectively at both 0.7 GHz and 3.5 GHz. This flexibility makes the platform suitable for exploring frequency-dependent bioelectromagnetic effects while preserving compatibility with simultaneous electrophysiological recordings.

Limitations and future perspectives

This study focuses on the design and numerical characterization of the proposed exposure system, and some limitations should be noted. The results are based on full-wave simulations, and experimental validation of the prototype is currently ongoing. While the numerical approach enables a controlled assessment of field distribution and SAR, experimental measurements will be necessary to confirm the dosimetric performance under real operating conditions.

The analysis also indicates that exposure is influenced by geometrical and positional factors, highlighting the importance of controlled alignment in practical implementations. In this context, a dedicated mechanical anchoring system will be considered to improve positioning reproducibility and reduce the variability observed in the uncertainty budget.

The proposed system is specifically designed for in vitro investigations based on MEA devices, and its extension to more complex biological models would require different experimental configurations. Future work will include experimental validation and integration with the electrophysiological recording system.

This work presents the design and numerical characterization of the Localized Radiating-End Coaxial line, an exposure architecture conceived to enable controlled and electromagnetically compatible RF stimulation in conjunction with MEA devices. The simulations demonstrated that the proposed configuration provides stable near-field confinement within the MEA culture region while maintaining negligible residual fields on the surrounding components of the incubator and the acquisition electronics. The resulting distributions of the electric field and SAR revealed frequency-dependent exposure patterns that combine spatial localization with reproducibility across the 5G-FR1 band. The dosimetric analysis further shows that the system maintains predictable SAR distributions and controllable exposure efficiency, supporting reliable experimental planning across different operational conditions. The overall behaviour confirms that the structure achieves the required balance between exposure intensity, localization, and EM compatibility, thus addressing one of the main limitations of conventional in vitro RF systems.

Future developments will focus on the experimental realization and validation of the LoREC prototype and on its integration with functional MEA platforms for real-time RF exposure and electrophysiological recording. By bridging numerical and experimental dosimetry, the LoREC system may provide a useful platform for standardized MEA investigations of bioelectromagnetic interactions in the 5G-FR1 frequency range.