What is the infant mortality period for RF semiconductor devices and how does burn-in testing address it?

The infant mortality period is the early portion of the bathtub reliability curve where the failure rate is elevated due to manufacturing defects that escape production testing: (1) Bathtub curve: the failure rate of electronic components follows a bathtub shape over time: infant mortality period (decreasing failure rate): the first 100-1000 hours. Defective components (weak die attach, contamination, partial wire bond, oxide defects) fail early. Useful life period (constant, low failure rate): 1000 to 10^6+ hours. Only random failures occur at the intrinsic reliability level. Wear-out period (increasing failure rate): beyond the useful life. Degradation mechanisms (electromigration, metal interdiffusion, dielectric breakdown) cause increasing failures. (2) Infant mortality causes in RF semiconductors: gate oxide defects (pinholes, contamination): cause early gate leakage failure. Partial wire bond lift-off: the bond appears good during visual inspection but has reduced adhesion. It fails under thermal cycling. Die attach voids: air pockets in the die attach solder or epoxy. They create localized hot spots that accelerate degradation. Contamination (mobile ions, metallic particles): causes early parametric drift or intermittent shorts. Epitaxial defects (threading dislocations in GaN): create localized current paths that degrade under stress. (3) Burn-in testing: burn-in is the process of operating devices at elevated stress (temperature and/or voltage) for a specified time to precipitate infant mortality failures before the device reaches the customer. High Temperature Operating Life (HTOL) burn-in: operate the device at elevated junction temperature (typically T_j = 125-175°C) for 48-168 hours. Apply rated DC bias and RF drive. Any device that fails during burn-in is discarded. The surviving devices have been screened of early defects and have a much lower failure rate. (4) Burn-in conditions: temperature: T_j = 125-175°C (for GaN; higher temperatures provide more screening in less time but risk damaging good devices). Duration: 48-168 hours (shorter for commercial, longer for military). Voltage: rated V_DS and V_GS (to stress the gate oxide and channel). Some burn-in protocols apply over-voltage stress (e.g., 110% of rated V_DS) for more aggressive screening. (5) Effect on reliability: without burn-in: the initial failure rate may be 10-100× higher than the steady-state rate. With burn-in: the infant mortality failures are removed. The field failure rate immediately starts at the low steady-state level. The MTBF of burned-in devices is 10-100× higher than unscreened devices during the first year of operation.