Selection of a novolac-based epoxy molding compound for power module encapsulation begins at the compounding stage rather than at the molding press. In a co-rotating twin-screw extruder with an L/D ratio of 48:1, the solid epoxy novolac resin is melt-mixed with a phenolic novolac hardener, fused silica or alumina filler, carbon black, wax release agent, silane coupling agent, flame retardant, and triphenylphosphine accelerator. The epoxy-to-phenolic hydroxyl equivalent ratio is maintained between 0.9 and 1.1; a ratio below 0.9 leaves unreacted phenolic hardener that can exude at the mold surface and reduce wetting of the copper leadframe, while a ratio above 1.1 produces excessive epoxy homopolymerization that increases fracture brittleness and reduces resistance to thermal shock. Filler loading in power module grades is typically set between 70 wt% and 90 wt% because the coefficient of linear thermal expansion below the glass transition temperature must be brought close to that of copper, at 17 ppm/K. The filler particle size distribution is bimodal, with coarse particles in the range of 10 µm to 40 µm and fine particles below 5 µm, to maximize packing density without creating a viscosity cliff-edge at the transfer molding temperature. Residual volatile content after compounding is controlled to < 0.5 wt% by vacuum venting in the extruder barrel at −0.08 MPa gauge; failure to do so results in surface porosity after solder reflow because the cured matrix cannot vent evolved water and low-molecular-weight oligomers.

The filler surface is treated with a silane coupling agent, typically 3-(glycidyloxypropyl)trimethoxysilane, at 0.3 wt% to 0.8 wt% relative to filler mass, to reduce melt viscosity and improve flexural strength retention after moisture exposure. Without surface treatment, the filler-matrix interface acts as a moisture diffusion path, and the flexural strength measured by ISO 178 can drop by 30 % after 168 h exposure to 85 °C/85 % RH. The silane is added during high-shear mixing in the first barrel section of the twin-screw extruder to ensure monolayer coverage of the reactive silica surface, but excess silane above 1.0 wt% is avoided because methoxy hydrolysis produces methanol that volatilizes during compounding and can be detected as residual solvent above 50 ppm in the final pellet. The particle size distribution is controlled by laser diffraction per ISO 13320; the span, defined as (d90 − d10) / d50, is maintained between 1.2 and 1.8 to maximize packing and minimize the viscosity increase at the tight end of the distribution.

On a production transfer molding press, pellet preheating is performed at 35 °C to 45 °C by radio-frequency heating before the pellet is placed in the transfer pot. The mold is held at 175 °C ± 3 °C, and the transfer ram applies 8 MPa to 12 MPa on the compound to fill the cavity through a narrow gate of 0.5 mm to 1.0 mm thickness. A vacuum level of ≤ 5 kPa absolute is drawn in the cavity during mold closing to prevent air entrapment between the wire bonds and the leadframe. Under these conditions a spiral flow length of 100 cm to 150 cm per ASTM D3123 is normally required, but narrow-cavity modules with gate-runner lengths above 200 mm and wire-bond loop heights below 100 µm may require spiral flow above 130 cm. Because the cure accelerator is already active at the molding temperature, the viscosity rise is exponential after the first 15 s; therefore the transfer time from pot to cavity must be completed within 10 s to 20 s and the total cure time is set between 120 s and 240 s. Mold temperature variation of ±5 °C around the set point changes the gel time by 20 % to 35 % and can reduce spiral flow by 15 % to 25 % at the low end, producing incomplete wire sweep coverage and open areas at the die attach fillet. This is the primary processing window limitation for fast-cure novolac compounds.

What Rheological Indicators Control Transfer Molding Yield in High-Voltage Module Packages?

Rheological testing of a novolac molding compound is conducted before lot release with a capillary rheometer at 175 °C using a die diameter of 1 mm and a shear rate range of 100 s⁻¹ to 1,000 s⁻¹, because the material exhibits strong shear-thinning behavior. The apparent viscosity at 100 s⁻¹ should remain between 100 Pa·s and 400 Pa·s; values above 600 Pa·s at this shear rate are associated with incomplete filling around the gate and excessive wire sweep. Spiral flow measured per ASTM D3123 at 175 °C and 80 kgf/cm² injection pressure is recorded for each batch, and the moving-die rheometer torque at 175 °C is used to determine the time to reach 90 % of maximum torque. The t90 value is typically 45 s to 75 s for power module grades; a t90 below 30 s creates a processing window narrower than ±3 °C and is unsuitable for cavity layouts with long runner systems, whereas a t90 above 90 s extends clamp time and reduces units-per-hour throughput on a mold press with 200 t clamp force. Batch-to-batch variation of melt viscosity is controlled within ±10 % by limiting the moisture content of the silica filler to < 0.05 wt% before extrusion and by pelletizing the compound at a die face temperature below 50 °C to prevent pre-reaction. A lot that is exposed to 60 % relative humidity for more than 24 h must be re-dried at 60 °C for 4 h to 6 h before molding; otherwise the moisture vaporizes at the mold temperature and produces voids at the leadframe interface. Amine-based adhesion promoters and imidazole accelerators above 0.5 phr are avoided because they reduce the storage life at 5 °C from 12 months to less than 3 months and can gel the material in the transfer pot during dwell periods longer than 10 min.

Wire sweep severity in a transfer molded power module is determined by the ratio of the melt front velocity to the wire bond natural frequency. For aluminum wire bonds of 400 µm diameter and loop length of 8 mm, the natural frequency is approximately 200 Hz; if the melt front crosses the wire at a velocity above 0.3 m/s, the drag force exceeds the yield strength of the annealed wire and residual deformation exceeds 5 % of the loop height. The compound therefore requires a low-viscosity window during the first 5 s of filling, and the filler particle size must be kept below the bond pitch to avoid particle bridging between adjacent wires. A maximum particle size of 24 µm is typical for modules with wire-to-wire spacing above 500 µm, but for spacing below 300 µm the maximum particle size is reduced to 10 µm and the filler fraction is adjusted to maintain spiral flow above 120 cm. Insufficient filler packing because of the fine-particle restriction can raise the α1 above 22 ppm/K, which is a direct processing conflict that limits the use of ultra-fine fillers in high-voltage modules. Wire sweep is measured by 2D X-ray inspection with a measurement uncertainty of ±2 µm; acceptance is < 5 % of loop height.

For a 1.7 kV silicon carbide half-bridge module with sintered silver die attach and aluminum wire bonds up to 500 µm diameter, the encapsulant must combine a glass transition temperature above 200 °C with a coefficient of linear thermal expansion below the glass transition that does not exceed 20 ppm/K. The copper baseplate and leadframe expand at 17 ppm/K, and the silicon carbide die at 3.5 ppm/K; when the encapsulant α1 exceeds 20 ppm/K, tensile stress at the die edge after thermal cycling from −55 °C to 175 °C is sufficient to initiate passivation cracks and wire bond heel fractures. Filler loading is therefore increased from 82 wt% to 88 wt% in many formulations, but the corresponding melt viscosity increase is nonlinear. Spiral flow may remain acceptable at 110 cm at 86 wt% fused silica, then collapse to 65 cm at 90 wt% because the binder volume is no longer sufficient to wet the entire filler surface and particle-particle friction dominates. This is a property cliff-edge caused by the packing limit of the bimodal filler; spherical alumina and boron nitride mitigate the thermal conductivity penalty but add cost and mold wear. Thermal conductivity measured by ASTM E1461 increases from 1.0 W/(m·K) for fused silica to 3.5 W/(m·K) for spherical alumina and to 6.0 W/(m·K) for an aluminum nitride/boron nitride hybrid at equivalent filler volume fraction, but the same hybrid filler reduces spiral flow by 50 % compared with fused silica at identical loading.

Package warpage after post-mold cure is controlled by balancing the filler content and the epoxy network shrinkage against the copper leadframe stiffness. The chemical shrinkage of a novolac epoxy system during cure is typically 2 % to 4 % by volume, while the difference between the molding temperature and room temperature imposes thermal contraction strain of 0.3 % to 0.6 %. A filled novolac compound with a filler loading above 86 wt% reduces total volumetric shrinkage to 0.4 % to 0.7 %, but the flexural modulus increases to 22 GPa to 25 GPa per ISO 178, and the high modulus converts residual stress into warpage on thin substrates. Warpage measured by contact profilometry after a 175 °C post-mold cure should remain below 100 µm across a 50 mm × 50 mm package; values above 150 µm cause die attach delamination during subsequent wire bonding or reflow. The selection of filler shape and coupling agent therefore involves a tradeoff between low shrinkage and acceptable modulus, with spherical fused silica giving the lowest mix viscosity and warpage, while crystalline silica gives lower CTE at the cost of higher mold wear.

Flash formation at the mold parting line is controlled by the molding compound’s melt viscosity and the mold clamping force. A compound with a melt viscosity below 100 Pa·s at 175 °C produces flash thickness below 10 µm at a clamp pressure of 50 kg/cm², but excessive flow reduces the cavity pressure available for dense packing at the die attach fillet. The mold parting line is maintained at a clearance of 2 µm to 5 µm because the filled novolac compound contains hard filler particles that absorb clamp force and prevent metal-to-metal contact. After molding, the flash is removed by dry-ice blasting or water jet deflashing; the compound must have a flexural modulus above 15 GPa so that thin flash sections break cleanly without tearing into the package body. Softer compounds with filler loading below 75 wt% produce rubbery flash that resists low-strain fracture and decreases deflashing yield on an automated line.

When the Encapsulant Must Maintain Adhesion to Copper, Aluminum, and Polyimide Passivation After 1,000 Hours at 85 °C/85 % RH

High-humidity biased testing such as H3TRB per IEC 60068-2-67 applies a voltage bias at 85 °C and 85 % RH for 1,000 h and forces the encapsulant to resist both interfacial moisture ingress and electrochemical corrosion of the copper leadframe and aluminum wire bonds. In this environment, adhesion loss at the copper-epoxy interface is driven by hydration of the copper oxide layer and hydrolysis of the silane coupling agent. Molding compounds intended for H3TRB exposure use naphthalene-type epoxy novolac resins and mercaptosilane coupling agents, which reduce the adhesion loss after 1,000 h from greater than 50 % to less than 20 % when measured by button shear on a copper substrate. Aluminum wire bond corrosion is suppressed by limiting extractable chloride, sodium, and potassium to < 10 ppm each using aqueous extraction at 121 °C for 20 h and subsequent ion chromatography. For a 1.2 kV IGBT module, the leakage current under HTRB at 1,200 V and 150 °C must remain below 10 µA after 1,000 h; ionic impurities above 10 ppm are known to produce localized conductive paths and dendrite growth. Adhesion to a polyimide passivation layer is further complicated by the low surface energy of polyimide; O2 plasma treatment at 300 W for 60 s prior to molding is commonly used to raise the surface free energy from 40 mN/m to 60 mN/m, but the molding compound must also contain a silane that reacts with the activated polyimide surface without being consumed by the high surface area of the silica filler.

Flame Retardant Mechanisms and REACH-Constrained Additive Selection

Flame retardancy for power module encapsulants is normally specified as UL 94 V-0 at a thickness of 0.8 mm, but high-voltage applications also require a comparative tracking index above 600 V per IEC 60112 because carbonized paths can form on the package surface under contaminated environments. Halogen-free formulations use 9,10-dihydro-9-oxa-10-phosphaphenanthrene-10-oxide type phosphorus compounds at 2 wt% to 5 wt% phosphorus content, in combination with nitrogen synergists such as melamine polyphosphate and inorganic zinc borate as an afterglow suppressant. The phosphorus additive shifts the decomposition pathway toward char formation by reacting with the epoxy novolac hydroxyl groups during the early stages of thermal degradation; the resulting char layer has an oxygen index above 35 % measured by ISO 4589-2. Metal hydroxide fillers, while effective as endothermic flame retardants, are generally avoided in transfer molding because they increase the melt viscosity and release water at molding temperatures above 180 °C, creating void defects. Regulatory constraints through REACH and RoHS restrict halogenated flame retardants and antimony trioxide in many export markets; a formulation containing more than 0.1 wt% bromine by weight is considered non-compliant for most European power module shipments. The glass transition temperature of a halogen-free novolac system can be depressed by 8 °C to 15 °C when 5 wt% organophosphorus flame retardant is added, so the base resin must be adjusted to a higher crosslink density to maintain the Tg above 180 °C.

Void-free encapsulation is mandatory for modules rated above 1.2 kV because a spherical void of 100 µm diameter at the die edge can exceed the partial discharge inception voltage threshold of 10 pC under 1.5 kV sinusoidal excitation. The transfer mold vacuum is held below 5 kPa absolute for 5 s before the compound is injected, and the gate is positioned at the leadframe side rather than over the die to prevent jetting. Cured packages are inspected by scanning acoustic microscopy with a 30 MHz transducer per IPC/JEDEC J-STD-035; delamination between the molding compound and the die passivation greater than 200 µm at the die attach fillet is a rejection criterion. Molding compounds with high moisture absorption above 0.3 wt% after 85 °C/85 % RH for 168 h measured by ISO 62 are unsuitable for high-voltage modules because the absorbed water reduces the breakdown strength by 20 % to 40 % per IEC 60243-1 and increases the dielectric loss factor measured by IEC 62631-2-1 at 1 MHz.

For continuous junction temperatures of 200 °C to 250 °C in silicon carbide modules, the chemical structure of the cured novolac network must resist oxidative decomposition long enough to maintain insulation resistance above 10¹² Ω at 500 V after 1,000 h of high-temperature storage at 250 °C. The glass transition temperature of a fully cured power module grade is typically between 180 °C and 240 °C by differential scanning calorimetry per ISO 11357-2; a post-mold cure of 4 h at 175 °C raises the degree of conversion to > 95 %. If the conversion is below 90 %, the residual epoxide groups continue to react during the first 500 h of high-temperature operation, causing Tg to rise by 10 °C to 20 °C and shrinkage stress at the die attach fillet to increase by 15 % to 25 %. Thermogravimetric analysis at 300 °C for 8 h in air should show a mass loss of < 0.5 wt%, but published data for specific SiC module configurations above 250 °C are limited because standard novolac encapsulants are not generally rated for continuous use beyond 225 °C. The selection boundary is therefore set by the intersection of the glass transition, the oxidative induction time of the flame retardant package, and the wire-bond pull strength retention measured after temperature cycling. A module designed for 175 °C Tj may accept a standard novolac compound, whereas a module designed for 200 °C Tj requires a naphthalene-based novolac or a cyanate ester-modified novolac network with reduced aliphatic methylene oxidation sites.

Cure Kinetics Are Measured by Differential Scanning Calorimetry and Moving-Die Rheometry

During incoming inspection, the cure kinetics of each novolac molding compound lot are characterized by differential scanning calorimetry per ASTM E2160 at heating rates of 5 °C/min, 10 °C/min, and 20 °C/min to determine the activation energy of the curing reaction. The activation energy for a phenolic-novolac-cured epoxy system is typically between 70 kJ/mol and 90 kJ/mol; lower values indicate a higher accelerator concentration and a narrower processing window. The moving-die rheometer at 175 °C records the torque increase as the compound crosslinks, and the time to 50 % torque (t50) and 90 % torque (t90) are used for lot acceptance. A lot with t50 below 30 s may gel in the transfer pot before the cavity is filled when the press transfer time is longer than 20 s; a lot with t50 above 60 s requires extended clamp time and reduces the throughput of a multi-cavity press. The degree of cure after post-mold cure is measured by residual exotherm; a residual exotherm below 5 J/g indicates a degree of conversion above 95 % for most systems, and a residual exotherm above 10 J/g is rejected because the under-cured network releases additional heat during the first high-temperature reverse bias test. Vacuum-assisted transfer molding with a vacuum level of ≤ 5 kPa absolute removes entrapped air, water vapor, and low-molecular-weight volatiles from the mold cavity; residual water at molding temperature flashes to vapor and is extracted before the compound wets the leadframe. Pressure holding at the end of fill is maintained at 80 kgf/cm² to 120 kgf/cm² until the gate freezes, after which the mold is opened and the package is degated.