**2. Methods**

#### *2.1. Model*

We consider a parallel-plate capacitor in which the gap of thickness *d* between the metallic electrodes is filled with a nanocomposite dielectric. The rectangular electrodes are assumed identical to each other and have the lateral dimensions *L*<sup>1</sup> and *L*<sup>2</sup> along the *x* and *y* axes, respectively. The nanocomposite is represented by identical spherical nanoparticles (NPs) of radius *R* with the dielectric permittivity  *<sup>i</sup>* randomly dispersed in the host dielectric with the dielectric permittivity  *<sup>h</sup>* (see Figure 1).

**Figure 1.** (**a**) The model of a nanocomposite capacitor. (**b**) The model of an NP.

We assume that the NP radius is much less than the capacitor dimensions, i.e., *R d*, *L*1, *L*2, and treat the NPs as point dipoles which possess the polarizability of a dielectric sphere. Such an approach follows the assumption made by Maxwell Garnett [14] which has been widely used for the modeling of the dielectric properties of composites. The applicability of this approximation is limited to relatively small volume fractions of NPs (see Section 2.3).
