The specification of epoxy resin for electrical encapsulation: CTI, dielectric and thermal cycling begins not with a single property but with three coupled failure modes: surface tracking under polluted conditions, bulk dielectric breakdown under field stress, and interfacial fatigue under thermal excursion. Comparative tracking index is evaluated according to IEC 60112:2003 using plaques of 3 mm thickness, platinum electrodes, and 0.1% ammonium chloride electrolyte at 23 °C. The result places the material in an insulation material group under IEC 60664-1:2020: Material Group I requires CTI ≥600 V, Material Group II requires 400 V to 599 V, Material Group IIIa requires 175 V to 399 V, and Material Group IIIb requires 100 V to 174 V. This grouping directly changes allowable creepage distance; a power converter board designed for Material Group I cannot accept an encapsulant that drops to Material Group IIIa after filler substitution without respacing terminals.

Dielectric strength is measured separately under IEC 60243-1:2013 with 20 mm ball electrodes immersed in transformer oil. Rigid bisphenol A/anhydride systems typically fail at 18 kV/mm to 25 kV/mm at 23 °C, but thin sections around sharp conductor edges can initiate breakdown at 12 kV/mm to 16 kV/mm after 1000 h of 85 °C/85% RH conditioning. Relative permittivity and dissipation factor are measured per IEC 62631-2-1:2018; an unfilled bisphenol A/anhydride network typically shows εr of 3.9 to 4.2 at 1 MHz and tan δ of 0.025 to 0.030. Fused silica-filled cycloaliphatic grades used for gate-drive isolation may show tan δ below 0.015 after post-cure because residual polar groups are consumed.

Adding Alumina Trihydrate to Bisphenol A Anhydride Networks

Alumina trihydrate exerts the largest CTI shift in bisphenol A/anhydride systems. At loadings of 150 phr to 200 phr, CTI values move from 250 V–300 V for the unfilled resin to 500 V–600 V because endothermic dehydration cools the surface and dilutes carbonaceous tracking residue. The cost appears immediately in rheology: mixed viscosity on a Brookfield RV DV-II+ at 25 °C reaches 180,000 mPa·s to 400,000 mPa·s, above the stable intake range of many reciprocating dispense pumps. Vacuum degassing below 10 mbar is required, and pot life under static cup conditions is reduced to 30 min–50 min when the exotherm exceeds 150 J/g by dynamic differential scanning calorimetry per ISO 11357-2:2020. Compounding is performed in planetary mixers or twin-screw extruders with L/D ratios of 40:1 to 60:1; shear rates above 2000 s⁻¹ can raise batch temperature and shorten working time. Filler settlement must also be checked because a d50 shift from 12 μm to 25 μm can reduce CTI by one material group on production batches.

On a dispensing line using a 24-element static mixer and tip pressure of 40 bar to 60 bar, filled encapsulants remain processable only if dynamic viscosity stays below 150,000 mPa·s at 25 °C or if the material is preheated to 35 °C. Two-stage degassing at 5 mbar and mold preheating to 50 °C reduce void-related partial discharge; air trapped in 100 μm gaps under dice lowers corona inception voltage by 20% to 40% under IEC 60270:2000. In final inspection, CTI spot testing is commonly performed on every 10th batch because ATH pulverization variance is not visible from viscosity data alone.

Dielectric Loss and Permittivity Variability Across Frequency and Temperature

Below 60 Hz, the dissipation factor of anhydride-cured bisphenol A epoxy is governed by ionic conduction from hydrolyzable chloride, unreacted carboxyl groups, and absorbed water. When frequency rises above 100 kHz, dipole orientation of secondary hydroxyl and ester groups becomes dominant; the loss peak temperature measured at fixed frequency shifts from below -20 °C in undercured samples to above 60 °C after post-cure. For a fused silica-filled system post-cured at 150 °C for 2 h, tan δ at 1 MHz decreases from 0.030 to 0.015, while εr drops by 0.2 to 0.4 because free volume collapses. A 0.5% moisture uptake after 500 h at 85 °C/85% RH raises low-frequency tan δ by a factor of 2 to 3 and increases εr because water displays a relative permittivity of 80.4 at 20 °C. High-frequency planar transformers and gate-driver isolation require encapsulants with volume resistivity above 10¹⁵ Ω·cm at 23 °C when measured by IEC 62631-3-1:2016; cycloaliphatic epoxy with triglycidyl isocyanurate crosslinking is preferred over bisphenol A to reduce polar ester content.

Why Does Thermal Cycling Produce Delamination at the Copper-Leadframe Interface?

Thermal cycling qualification per IEC 60068-2-14:2009 test Na typically uses dwell times of 10 min to 30 min, transfer times below 15 s, and endpoints from -40 °C/+85 °C to -55 °C/+150 °C. The encapsulant fails primarily by interfacial delamination, not bulk cracking, because the coefficient of linear thermal expansion of unfilled bisphenol A epoxy is 65 ppm/K to 75 ppm/K below Tg, while copper is 17 ppm/K and silver is 19.6 ppm/K. Filling with fused silica at 68 wt% to 72 wt% reduces α1 to 25 ppm/K to 30 ppm/K when measured by thermomechanical analysis per ISO 11359-2:2020. After 500 cycles from -55 °C to 150 °C, scanning acoustic microscopy shows corner delamination in the unfilled control but no detectable voids in the fused silica grade. For hot-dwell stability, the DMA glass transition temperature per ASTM E1640-18 at 1 Hz is held above 165 °C in novolac-anhydride formulations.

Published data for the effect of ramp rates between 5 K/min and 15 K/min on delamination depth in high-Tg novolac encapsulants is limited; qualification programs therefore rely on step-stress cycling rather than a single accelerated curve. Preconditioning before thermal cycling for encapsulated modules follows JEDEC JESD22-A113 with 85 °C/85% RH soak for 96 h to 168 h and optional reflow passes at peak 245 °C to 260 °C. Absorbed water volatilizes during reflow; cohesive cracking appears when the DMA Tg is below 160 °C, and the package popcorning mode is suppressed by prebaking at 125 °C for 12 h or by selecting novolac grades with Tg above 175 °C. Rinse water conductivity above 5 μS/cm in component cleaning lines leaves ionic residue that can shift CTI by one material group under IEC 60112:2003.

Comparative Performance of Unfilled, ATH-Filled, and Fused Silica-Filled Encapsulants

The ranges below are assembled from publicly available manufacturer technical bulletins and independent test reports; they are indicative screening values, not specification limits.

When Sustained DC Voltages Exceed 800 V, Space Charge Accumulation Changes the Failure Mechanism

For electric vehicle traction inverters and high-voltage DC-DC stages with continuous bus voltages above 800 V, CTI and short-time dielectric strength do not capture long-term DC aging. Charge injected from bare copper busbars accumulates in the resin bulk until local fields at conductor edges exceed 5 kV/mm to 10 kV/mm, after which electrical tree growth is initiated. Pulsed electroacoustic measurements on 0.2 mm molded films show heterocharge accumulation near the anode when conversion by DSC is below 95%; post-curing at 140 °C for 8 h lowers charge density but does not eliminate it if the resin contains free chloride above 10 ppm. High-voltage DC encapsulants therefore use low-chloride cycloaliphatic epoxy with acid anhydride hardener and benzoxazine additions of 5 phr to 15 phr. These formulations maintain volume resistivity above 10¹³ Ω·cm after 85 °C/85% RH aging for 1000 h per IEC 62631-3-1:2016. Published data for space-charge distribution in filled commercial encapsulants at 120 °C and 10 kV/mm is limited; design verification must include direct time-to-breakdown testing on the actual bus geometry.

The following standards matrix consolidates the qualification measurements used for high-reliability encapsulated power electronics.