The increasing deployment of millimeter-wave (mmWave) and sub-terahertz (THz) frequencies in next-generation 6G communication systems presents challenges for electromagnetic field (EMF) exposure assessment. Traditional methods often struggle with the high computational demand and complex modeling required for accurate assessment at these frequencies. This thesis introduces novel approaches to improve the efficiency and practicality of EMF exposure assessment in high-frequency scenarios, addressing both methodological and computational complexities.

Firstly, a boundary criterion for the application of incident power density (IPD) in the near field is proposed. Defined as the distance at which the influence of the imaginary part of the Poynting vector becomes notably significant with respect to the real part, the criterion offers a more precise transition between reactive and radiative near fields than traditional definitions. This reliable basis for IPD application in compliance assessments reduces dependency on absorbed power density (APD), which is difficult to measure within tissue, thus simplifying the evaluation process in near-field exposure scenarios and indirectly enhancing assessment efficiency.

Secondly, to overcome the limitations of existing methods in evaluating temperature rises in tissues at high frequencies, where multi-layer tissue models are essential due to heat dissipation across deeper layers, this thesis proposes a simplified approach. This approach initially assesses EMF without deep tissue layers, then using the results to estimate temperature distribution in a full-layer tissue model, thereby enabling faster and more practical assessments while maintaining accuracy, especially for complex exposure conditions above 100 GHz. 

Additionally, we examined the correlation between the temperature rise and APD across various distances, frequencies, averaging area for APD, array sizes of antennas designed for 6G communication using the proposed approach. This study fills the current lack of research on practical antennas operating at frequencies above 100 GHz. The results are in alignment with the international EMF guidelines, which further substantiates the effectiveness of the proposed method.

Thirdly, a machine learning (ML) framework is developed to predict temperature rises and APD on the skin surface of clothed human body models. This framework accounts for variations in factors such as incident angle, cross-polarization power ratio (XPR), clothing material, air gaps, and frequency, making it highly adaptable for diverse exposure scenarios. By incorporating a weights analyzer module to streamline training and a DNN optimization module that adaptively adjusts hyperparameters, this framework significantly enhances assessment efficiency—achieving a four-order magnitude increase in speed over traditional methods with minimal manual intervention.

The methodologies presented in this thesis provide approaches to enhancing the efficiency of mmWave and sub-THz EMF exposure assessment, enabling practical implementation in high-frequency environments and positioning it as a critical component of the safe and effective deployment of 6G wireless technology.

After the defence there will be a small reception at Fredrik Bajers Vej 7, A4-106