All nano-composites were prepared with unpigmented LDPE powder (additive-free - density: 0.922 g/cm³) and MgO nano-particles (1% weight-nominal mean diameter: 30 nm). The nanoparticles were first dried in vacuum during 24 h at 130˚C, which is not expected to change the surface chemistry as this temperature is much lower than the thermal decomposition temperature of surface hydroxide groups [10] . They were then ultrasonically dispersed in ethanol until a homogenous suspension was formed and subsequently a solution of (3-Glycidyloxypropyl) trimethoxysilane was slowly added to the suspension with the mixture stirred at 50˚C for 48 h in vacuum. The resulting slurry was washed with fresh ethanol several times and dried in a vacuum oven at 80˚C for 24 h. This type of surface treatment is expected to improve the compatibility between the organic matric and the inorganic particles and has been reported to improve the dielectric properties for several metallic oxides polyolefins composites [13] [14] .

Following the nanoparticles treatment, LDPE/MgO masterbatches were produced using a conventional hot mixer. The mixing was conducted at a temperature of 140˚C, a rotation speed of 60 rpm, and for 30 min. This was followed by micro-extrusion conducted by the means of a mini-extruder.

Pellets obtained from the extrusion were finally pressed at 140˚C and under a pressure of 10 MPa to obtain films with thicknesses ranging from 175 to 200 μm. The after-fabrication concentration was assessed by Thermo Gravimetric Analysis (TGA) using the following procedure: the temperature was risen from 50˚C to 600˚C at the rate of 20˚C/min under nitrogen and then maintained at 600˚C for 30 min under air. The residual weights averaged on several samples are listed in Table 1.

A SEM Hitachi SU3500 was used for the microstructural characterization of the samples. A reasonable dispersion and distribution of the MgO particles was observed with the mixing procedure previously described, as shown in Figure 1.

The short-term AC dielectric breakdown strength was measured by using a classical experimental set-up composed of a 50 Hz high voltage transformer fed by a motorized autotransformer. The voltage rising rate was fixed at 1 kV/s. The experimental set-up is fully automatized, i.e.: the autotransformer is stopped when the current flowing through the sample reaches 10 mA and the corresponding breakdown voltage is then recorded. The measuring cell made of plexiglass (Figure 2) was equipped with two stainless steel electrodes: one plane electrode (diameter: 25 mm) and one semi-spherical (diameter: 1 mm). Samples were sandwiched in between the two electrodes and the measuring cell was filled with an insulating surrounding medium (dielectric liquid: GALDEN^© HT55) to avoid any flashover or surface dielectric breakdown.

The thickness of each specimen was estimated after each experiment in the breakdown area, i.e.: four measurements around the breakdown point channel in order to calculate an average thickness. The variability in samples thickness was less than 50 μm. The Weibull’s statistical analysis was used to treat the recorded data according to the IEEE std-930 [15] . The two-parameter Weibull’s distribution cumulative probability of failure is given by:

P ( x ) = 1 − exp [ − ( x α ) β ] (1)

where α and β are respectively the scale and shape parameters and x is the breakdown field. α and β were obtained by using the weighted least-squares regression technique. Since the estimated values obtained from a single experiment involving a finite number of specimens are unlikely to equal the true values of α and β, the range of conceivable values for α and β, given the experimental data, was obtained by computing the 90% confidence bounds using the procedure described in [15] .

Lifetime measurements have been performed by applying square pulse voltages (duty cycle: 50%). The frequency has been adjusted from 1 to 20 kHz, the voltage at ±1 kV. All measurements were performed at room temperature (i.e.: 20˚C; RH = 55%). The accuracy of the thermal chamber was about ±0.5˚C. The rise and fall times of the square voltage have been estimated to be about 30 and 50 kV/μs respectively. During these tests, the level of applied voltages was always higher than the partial discharge inception voltage (PDIV) which has been estimated to be ±800 V at 20˚C. The experimental lifetime bench is shown in Figure 3.

Samples have been fixed on grounded stainless-steel plates. The square wave high voltage has been applied on stainless steel spheres (diameter: 4 mm) deposited on the sample surface (Figure 4). An automatic system detects any sample short-circuit, extracts the broken sample for the test bench and record its lifetime. Partial discharges take place between the spherical electrode and the sample surface. Submitted to ions bombardment, heat and oxidation mainly due to ozone formation, the sample surface erodes progressively until the sample final dielectric breakdown. As partial discharges appear during the rising and falling fronts of the square voltage, it is expected that the higher is the frequency, the lower is the lifetime.

Since partial discharges generate heat around the discharge area, all the samples have been put into a climatic chamber (at the left of Figure 3), rather than in a simple oven, in order to maintain the medium temperature at 20˚C.

The dielectric strength results of both pure and nano-filled LDPE samples are reported in Table 2. The electrical field at breakdown is simply calculated by

dividing the breakdown voltage by the sample thickness. This is an average breakdown field across the sample since the electrical field is not uniform due to the electrode geometry. For each breakdown test, the sample was examined in order to assess that the breakdown occurred through the sample and was not due to surface breakdown or flashover. Values reported in Table 2 correspond to the Weibull’s shape and scale parameters, obtained from measurements on 8 samples for each material. The slight improvement observed is in agreement with reports found in the literature where similar tests were reported [16] [17] .

The lifetime results of both pure and nano-filled LDPE samples are reported in Figure 5. Each dot in this figure corresponds to the Weibull’s scale parameter obtained from 8 measured lifetime.

The AC dielectric strength of pure and nano-filled LDPE measurements showed data very similar, i.e. a marginal difference of about 6% with an overlap of the confidence bounds. As expected, the lifetime data reported in Figure 5 clearly shows that the inclusion of the nano-fillers improves the resistance to partial discharges. The lower the frequency (i.e.: the lower the number of PD per second), the higher was this enhancement of PD resistance. These results show that PE/MgO composites are potentially interesting material to be used for the insulation of high voltage AC cable with an improved resistance to local PD initiated in defects (i.e.: voids) that may appear during the LDPE insulator ageing. This is expected to lead to improved resistance to electrical treeing. Since PD is the main mechanism leading to the propagation of electrical trees in polymeric dielectrics, as well as much longer inception times. This behavior is also in good agreement with other reports in the literature on the dielectric properties of MgO-based nanocomposites. As an example, it was found that the lapse time to tracking failure increases with the content of nano-MgO in the 0 to 5 wt% range in the case of epoxy [18] . More complete reviews can be found in the literature on the improvement of erosion resistance in polymeric nanocomposites (see [1] [19] for example) that also are in good agreement with the observed trend in this report.