I. INTRODUCTIONThin metallic layers have been gaining significant importance, e.g., in the field of so-called active packaging [1] and electromagnetic shielding [2]. The behavior of a layer of metallic nanometer-scale spheres [3] and rings [4] has been investigated at optical frequencies. The response of thin metallic layers at microwave frequencies has been investigated by several authors. It was shown in [5] that even a very thin layer may absorb up to 50 % of the incident electromagnetic wave energy. The influence of layer thickness on effective conductivity was studied in [6].A thin sputtered gold layer was experimentally investigated at microwave frequencies in [7]. It was found that there is a small reflection from a layer of disconnected gold nano-particles for short sputtering times. For long sputtering times when the gold layer up to tens of nanometers in thickness is homogeneous, the reflection is high. The transition between these two states of reflection was recorded.The thin metallic layer is represented in this paper as a 2D array of spherical gold particles. The process of gradual depositing of metal by sputtering is simulated by decreasing the distance between particles, which is equivalent to increasing their number. Their interaction with an electromagnetic wave is interpreted in four ways. The first model taken from [8] analyses the electromagnetic field interaction with the electric dipole moment of the particles. The second, newly proposed model, is based on the capacities between spheres. Next we consider the layer to be homogeneous, and take into account the effect of decreasing conductivity in very thin films [6], which gives results valid for longer sputtering times. Finally, the layer of gold particles is modeled using the CST Microwave Studio (MWS). Unlike the above-mentioned three models, the MWS model accounts for a random distribution of particles over the analyzed aperture, which gives the most relevant results. All resulting curves of the reflection coefficient with respect to sputtering time are compared with the measured data. The aim of this work is to approximate the measured reflection coefficient [7] for the whole range of investigated sputtering times.