Safe Operating Area
The safe operating area (SOA) is the boundary on a plot of voltage against current inside which a power device can be operated without destroying itself, and that boundary is not one straight line but several different physical limits stacked together, whichever one is most restrictive at a given operating point wins. A power MOSFET, IGBT, or GaN or SiC device can fail from several completely different mechanisms depending on how hard and how long it is pushed, so the SOA curve a datasheet shows is really four or five separate failure boundaries drawn on the same axes, and reading it correctly means knowing which wall you are actually about to hit.
The power limit is the one engineers hit most often, and it changes shape with time. A device dissipating V times I as heat has to stay below the junction temperature its package and cooling can handle, which traces out a constant-power line with a slope of negative one on the log-log V-I plot a datasheet actually uses — but that line sits much higher for a brief pulse than for continuous DC operation, because a short pulse doesn't give heat time to build up in the silicon before it can spread into the package, so a real SOA chart shows several nested power-limit curves for different pulse widths (10 microseconds, 100 microseconds, 1 millisecond, DC) rather than one fixed boundary.
Second breakdown is the limit that makes SOA a genuinely dangerous corner to misjudge, not just a thermal-margin number. At high voltage and moderate current, current flow inside some device structures can collapse into a localized hot spot rather than spreading evenly across the die, and that hot spot heats further, draws more current, and heats further still in a positive-feedback runaway that destroys the device in microseconds — this failure mode folds the SOA boundary sharply inward at high voltage, well below where a simple power calculation would otherwise suggest the device is fine, which is exactly why datasheets draw it as its own separate curve rather than folding it into the power limit.
Designers use the SOA curve directly when choosing how a device gets switched and protected. Gate-drive engineers shape turn-off slew rate specifically to keep the device's instantaneous operating trajectory inside the SOA boundary during a short-circuit fault, and application engineers for automotive and industrial power modules size heatsinks, snubbers, and fault-detection response time against the same curve a device vendor like Infineon, onsemi, or STMicroelectronics publishes for that specific part — the SOA chart is one of the first things a power-circuit designer checks, before efficiency numbers, because it is the one chart that tells you exactly how the part dies.
Read the SOA curve through a *whichever-wall-you-hit-first* lens rather than a *single-spec* lens — a device operating safely at one voltage and current can be destroyed at the same power level simply by moving to a different point on the curve, because the limit that actually matters changes depending on where you are standing on it.