Semiconductor Technology from A to Z

The failure rate of electronic components over operating time typically follows the so-called bathtub curve: early in the service life, infant mortality failures caused by manufacturing defects dominate and the failure rate decreases; this is followed by a long period of low, roughly constant random failures – often expressed in FIT (Failures In Time, failures per 10⁹ device-hours) and typically in the range of a few to a few hundred FIT for qualified semiconductor devices –, before wear-out failures driven by aging mechanisms such as electromigration, TDDB, or BTI cause the failure rate to rise again toward the end of the useful life.

Burn-in refers to a stress test in which components are operated for a defined period, often 24 to 168 hours, under elevated voltage and temperature in order to accelerate through the infant mortality phase of the bathtub curve and screen out defective units before they are shipped to the customer. Because burn-in adds cost and test time, it is mainly applied to applications with high reliability requirements, such as automotive or industrial electronics, while mature consumer products with well-controlled processes are often covered by statistical sampling instead.

Highly Accelerated Temperature and Operating Life (HTOL) testing is a standardized qualification test in which a sample of components – per JEDEC JESD22-A108 often 77 units drawn from at least three different production lots – is stressed for typically 1000 hours at an elevated junction temperature (often 125 °C, or 150 °C for some technologies) and elevated operating voltage under active switching conditions. The units under test are electrically characterized before, during, and after the test to detect parameter drift or functional failures.

To pass the test, it is usually required that no unit fails within the sample (a zero-fail criterion), which for a sample of 77 units statistically corresponds to a failure rate of less than about 0.3 % at 60 % confidence. HTOL primarily covers diffusion-driven and field-dependent aging mechanisms such as electromigration, TDDB, and BTI, complementing ESD and latchup qualification with time-dependent operating reliability.

Because testing under nominal conditions for the required field lifetimes of ten years or more is not practical, HTOL tests are run at elevated temperature and the results are extrapolated to the operating temperature using the Arrhenius model. The model describes the rate of a thermally activated aging mechanism as an exponential function k(T) = A · e^-E_a/(k·T) of the inverse absolute temperature, weighted by an activation energy E_a characteristic of the respective failure mechanism.

The ratio of failure times at test temperature and operating temperature yields an acceleration factor AF = e^{(E_a/k) · (1/T_use – 1/T_test)}. For a typical TDDB mechanism with E_a ≈ 0.7 eV, a test temperature of 125 °C (398 K), and an assumed operating temperature of 55 °C (328 K), the acceleration factor works out to roughly 40 to 50 – under these assumptions, 1000 test hours therefore correspond to an equivalent field operating time of about 4 to 5 years. Because different failure mechanisms have different activation energies – TDDB and electromigration are typically in the range of 0.6 to 1 eV –, a reliable lifetime prediction requires knowledge of the dominant mechanism.