For sintered die attach on 1200 V SiC MOSFET half-bridge modules, the electronic-grade silver paste is dispensed or printed onto active-metal-brazed Si₃N₄ substrates with a wet film thickness of 75–125 µm depending on die camber and substrate bow. The paste typically contains 88–91 wt% silver, a D50 particle size of 2–4 µm, and a metal tap density above 4.5 g/cm³. Solvent removal is carried out as a stepped ramp from 25°C to 140°C at no more than 3°C/min under nitrogen, using a convection oven or a vacuum-assisted dryer with platen temperature uniformity of ±2°C. Residual organic content below 1.5 wt% at the end of the drying segment is verified by thermogravimetric analysis per ASTM E1131-08. After die placement, sintering at 230–260°C under 10–20 MPa for 2–5 min in a press with ±2°C platen uniformity produces a silver joint with 10–15% porosity and die shear strength above 30 MPa when tested per MIL-STD-883 TM 2019.9. Laser flash measurement per ASTM E1461-13 gives thermal conductivity in the 150–200 W/mK range for this porosity band. Manufacturing lines report die-lift failures when the drying ramp is pushed above 5°C/min, because vapor release at the die edge outpaces the center by 2–3°C, leaving carbon-rich residues near the joint center. Scanning acoustic microscopy is used to reject modules with single-void area above 5% of the die attach region; this threshold is an internal line-control limit rather than a published standard.

On production-scale sintering presses, die tilt is controlled by clamping force feedback and a pressure head flatness better than 5 µm across 100 mm. Peak temperature overshoot at the paste joint is limited to +5°C because temperatures above 265°C decompose residual organics too quickly and create porosity channels that reduce thermal conductivity by roughly 20 W/mK. A qualification lot with paste silver content lower by 0.5 wt% from nominal can drop die shear below 25 MPa after 1000 thermal cycles; incoming paste is therefore screened by ASTM E1131-08 for metal residue and by rotational rheometry at 25°C and 10 s⁻¹. For modules with 5 mm × 5 mm SiC dies, press cycle times below 2 min are not recommended because thermal diffusion through the substrate lags by 20–30 s, leaving the center region under-sintered.

How Do Ni/Pd/Au and Bare Cu Metallizations on Si₃N₄ AMB Substrates Shift Sintered Silver Adhesion?

Electroless nickel/immersion gold and Ni/Pd/Au finishes introduce distinct diffusion barriers and oxide formation kinetics on direct-bond copper or active-metal-brazed substrates during silver sintering. A nickel thickness of 3–7 µm with palladium at 0.05–0.15 µm and a gold flash of 0.03–0.10 µm is typical; X-ray fluorescence thickness checks per ISO 3497 are performed at incoming inspection because gold below 0.03 µm permits nickel oxidation before sintering and lowers die shear by 15–25% after 250°C assisted densification. Bare copper surfaces are pre-dried at 150°C for 2 h in nitrogen with oxygen content below 100 ppm, and X-ray photoelectron spectroscopy is used to confirm copper oxide thickness below 25 nm before print. Plasma treatment with CF₄/O₂ at 100–200 W for 60–120 s increases surface energy on Ni/Au finishes to above 60 mN/m, measured with contact-angle goniometry. Thermal cycle testing per JEDEC JESD22-A104G from -55°C to 175°C for 1000 cycles generally requires die shear retention above 80%; on bare copper, shear loss after cycling is often 10–15% higher than on Ni/Pd/Au because copper oxidation at the joint periphery weakens the silver network. A production bottleneck occurs when immersion gold bath aging shifts phosphorus content in the underlying nickel above 10 wt%; the sintered silver still bonds, but the nickel layer becomes brittle under thermomechanical stress and can crack within 500 cycles.

Power module substrates with bare copper lands require an additional forming gas treatment at 250°C for 30 min prior to sintering when copper oxide thickness measured by XPS exceeds 25 nm. On Ni/Pd/Au finishes, low-temperature silver sintering does not form brittle intermetallic compounds, but silver diffuses into gold above 250°C and can cause Kirkendall void formation if gold thickness exceeds 0.30 µm. Gold thickness is therefore specified at 0.03–0.10 µm rather than the thicker ENIG used for solderable PCB finishes. Shear testing per IEC 60749-19 after JEDEC JESD22-A104G thermal cycling is supplemented by cross-sectional analysis to verify failure location: cohesive failure inside sintered silver is acceptable; adhesion failure at the nickel/silver interface is not. Field data from line audits show that substrate lot acceptance improves when the immersion gold bath is replenished at 0.2 g/L increments and pH is maintained within 5.8–6.2; published data for exact bath chemistry interactions with pressure-assisted silver sintering remains limited.

Clip Bonding Paste Rheology, Stencil Release, and Thixotropic Recovery

Stencil printing for copper-clip attach on power modules requires the paste to maintain a viscosity of 20–50 Pa·s at 10 s⁻¹ and 25°C with a thixotropic index above 3.5 when measured per ISO 3219. A metal squeegee angle of 15° and blade speed of 25–60 mm/s on a stencil printer with vacuum hold-down and 0.3–0.8 mm snap-off typically yields aperture release above 95% for 70 µm stencil openings. After an idle time of 5 min on the stencil surface, viscosity recovery must exceed 85% of the initial value; otherwise paste thickens at edge apertures and produces missed prints or insufficient clip wetting. A three-step rotational rheometry test at 0.1 s⁻¹, 50 s⁻¹, and 0.1 s⁻¹ is used on production lots to quantify this recovery. Printed slump is held below 10% width expansion after 15 min at 23±2°C and 50±5% RH; this is measured by optical profilometry on a glass test coupon. The process conflict in clip bonding is that high thixotropy improves stencil definition but slows self-leveling on the copper clip landing area. If leveling is incomplete, void formation under the clip is observed by X-ray computed tomography with 5 µm voxel resolution, particularly when clip placement force is below 5 N and the paste does not wet the clip underside. In high-volume lines, stencil undersize of 10–15% relative to clip area is applied to avoid paste squeeze-out and shorting between adjacent gate pads.

Clip bonding lines with automatic optical inspection set a paste skip threshold of 0.3 mm in length and 0.1 mm in width because smaller skips fill during clip wetting and do not create open voids. Paste lot release includes a print test on an 80 µm stencil with aperture area of 2 mm × 2 mm; the printed deposit is measured with a non-contact laser profilometer within 1 min after separation. Recovered viscosity after 10 min idle, measured at 0.1 s⁻¹, must remain above 70 Pa·s but below 120 Pa·s; higher values indicate solvent loss and can produce incomplete clip wetting. X-ray computed tomography of the clip attach is set to detect voids above 0.1 mm diameter; typical acceptance is less than 2% cumulative void area under the clip, with no single void above 1%. On manual lines, squeeze-out is reduced by using a stencil aperture reduction of 60–80 µm along the clip edge nearest the gate pad.

Because power discrete packages such as TO-247 and D²PAK use silver paste on bare copper or spot-silvered leadframes and are subsequently overmolded, the paste must tolerate both die-attach thermal processing and mold compound cure without excessive outgassing or delamination. Automated dispensing through needle diameters of 0.25–0.51 mm deposits 2–10 mg dots at wet thicknesses of 100–250 µm for die sizes between 3 mm and 6 mm. When die length exceeds 6 mm, multiple dots at 0.6–0.8 mm pitch are standard; otherwise die tilt above 5 µm across the diagonal creates nonuniform joint thickness after clip or leadframe attach. The organic carrier is removed at 125–150°C with a 30 min dwell in a convection oven, and residual solvent after pre-dry is held below 0.5 wt% by ASTM E1131-08. Leadframe silver spot roughness is maintained at Ra 0.4–0.7 µm using contact profilometry, because lower roughness reduces mechanical interlocking and higher roughness traps organics at the interface. JEDEC moisture sensitivity classification per JESD22-A113 Level 1 requires no package cracking after 260°C reflow; paste-related delamination at die edges appears when printed area coverage is below 75% of the die base. After assembly, ionic cleanliness must be below 0.1 µg/cm² NaCl equivalent per IPC-TM-650 2.3.25, because chloride residues above this level promote silver migration and wire-bond pad corrosion under biased humidity testing.

Overmold compatibility is a stricter constraint for discretes than for open power modules. The coefficient of thermal expansion mismatch between copper leadframe at 17 ppm/K, sintered silver at 19–21 ppm/K, and silica-filled epoxy mold compound at 8–12 ppm/K concentrates shear at the die attach during thermal cycling. For TO-247 packages, die shear retention after 500 cycles from -65°C to 150°C is often specified above 85% of initial to avoid field delamination. Ionic cleanliness after die attach is tested before mold by extraction resistometer per IPC-TM-650 2.3.25; acceptance is ≤0.1 µg/cm² NaCl equivalent. Silver paste residues on the leadframe shoulder above 0.5 mm outside the die footprint are removed by plasma descum before wire bonding, because organic films cause non-stick on pull testing and increase wire bond lift-off.

When Pressure-Free Sintering Replaces Assisted Densification for Thin-Die Power Discrete Packages

Pressure-free sintered silver is selected for dies below 100 µm thickness because non-uniform platen force can fracture the silicon carbide or silicon edge during assisted densification. Isothermal hold at 250°C for 60 min in air or nitrogen with oxygen below 500 ppm relies on particle surface diffusion and slow organic burnout rather than viscoplastic flow. The resulting joint has higher porosity, typically 20–30%, and lower thermal conductivity of 80–130 W/mK measured by laser flash per ASTM E1461-13. Die shear strength on silver-plated leadframes is 10–25 MPa per IEC 60749-19, which is sufficient for low-current power discretes but marginal for traction inverters with high power cycling amplitude. The process window is constrained to 240–260°C; below 235°C, porosity exceeds 25% and shear strength falls below 10 MPa, while above 265°C silver particle growth is rapid but residual carbon concentrates at the joint center. Oven chamber uniformity of ±5°C or better is mandatory, and batch-to-batch variation in paste solvent content above 0.3 wt% shifts the required dwell time by 10–15 min, causing under-sintered edges in high-volume reflow ovens with six-zone profiles.

Pressure-free sintering of large dies above 4 mm × 4 mm exhibits a non-uniform density profile: edges may reach 90% relative density while centers remain below 80%, especially when a chip top clip acts as a heat sink during ramp. This is measured by cross-section image analysis and is one reason pressure-free silver is usually restricted to die sizes below 3 mm × 3 mm for power discretes or to low-current converter applications. Power cycling per AQG 324 on modules with pressure-free joints shows acceptable performance only when junction temperature swing ΔTⱼ is below 80 K and maximum junction temperature is below 150°C; above this, the porous silver network coarsens and thermal impedance rises by 10–15% after 10,000 cycles. Production ovens with insufficient nitrogen coverage report dark silver color and shear strength loss of 20–30%, which is traced to oxidation of the fine particle fraction below 1 µm.

Electrochemical migration in sintered silver die attach is evaluated by applying 1000 V bias at 85°C and 85% RH per JEDEC JESD22-A101, with optical inspection at 100× for dendritic growth between adjacent pads on test coupons. Ionic contamination from upstream die-attach solvent, flux residues, or substrate cleaning must be held below 0.2 µg/cm² NaCl equivalent before molding; values above this level produce silver dendrite formation after 168 h in biased humidity on non-passivated copper surfaces. Silicone encapsulation compounds with extractable chloride below 10 ppm are preferred for modules subjected to condensation cycling, because higher extractables release mobile ions at the gate pad edge. Operational boundaries include avoidance of ammonia-based cleaning agents after sintering, since they can dissolve residual silver oxide and leave amine residues that increase surface leakage. Packages stored above 60% RH require 125°C for 4 h pre-mold bake to prevent moisture entrapment at the sintered joint periphery. X-ray inspection after high-humidity reverse bias commonly locates darkening around the die attach fillet before electrical failure; this is used as an early line-screen indicator.

Surface insulation resistance testing on sintered silver couplers with 0.5 mm gap under 40 V bias and 85% RH per JIS Z 3197 or IPC-TM-650 2.6.3 detects leakage currents above 10 nA when organic residues are not fully volatilized. Sintered silver pastes formulated for power packaging are not compatible with low-temperature bismuth-based solders used in rework; contact between the two produces galvanic potentials that accelerate silver dissolution under humid conditions. Rework of sintered silver is limited to full die removal and substrate cleaning; partial re-sintering over an existing joint is not recommended because the residual sintered layer has lower surface energy than the original substrate and cannot achieve full densification. Published data for rework reliability on pressure-sintered silver in automotive modules is limited; most manufacturers replace the substrate rather than attempt local joint repair.