Punch-Through

Punch-through is the failure of a device's blocking region when two depletion regions grow until they meet, so that a current path opens across a region that was meant to be insulating. In a bipolar transistor, the collector-base depletion region spreads across the thin base until it touches the emitter-side depletion region; in a short-channel MOSFET, the drain and source depletion regions merge beneath the gate. In both cases the device is limited by geometry rather than by a material property, and the result is a current that the control electrode can no longer fully govern.

In a bipolar transistor, punch-through sets the minimum safe base width for a given voltage. The base is made thin, often a fraction of a micrometre or less, so carriers cross it quickly, which is what gives high gain and speed. Applying a reverse bias to the collector widens its depletion region, and because the base is more lightly doped than the emitter, the region extends mostly into the base, narrowing the neutral part that carries the control. If the bias is high enough, the depletion layer reaches the emitter junction and the emitter injects carriers directly into the collector, so collector current rises sharply with voltage regardless of base current. Before that point, the same narrowing already causes the Early effect, in which the collector current drifts upward with voltage. Raising base doping delays punch-through but lowers gain and increases emitter-base capacitance, a classic trade-off.

In a MOSFET, punch-through is the shrinking-channel version of the same story, and it limited how far transistors could be scaled. In a long device the gate alone controls the channel. As gate length shrinks, the depletion regions of source and drain approach each other under the surface, and the drain voltage starts to lower the barrier that the gate is supposed to hold, a short-channel behaviour called drain-induced barrier lowering. Current then flows below the surface even when the gate is off, increasing leakage and making the threshold voltage depend on drain bias. Designers treated punch-through as one of the walls to scaling, along with oxide tunnelling and doping fluctuations, and Moore's law could continue only because each generation brought new ways to push the wall back.

Anti-punch-through implants are the standard process answer. A MOSFET built with a deeper, more heavily doped layer under the channel stops the depletion regions from spreading there, and extra "halo" or "pocket" implants near the source and drain ends raise doping at the edges that would otherwise merge first. Retrograde wells put the heavier doping below the surface where it is needed and leave a lightly doped channel for mobility. These fixes have costs: more doping increases junction capacitance and random dopant fluctuation, where the small number of dopant atoms in a tiny channel makes transistors differ. Eventually the industry changed structures, using thin fins in FinFETs, introduced in volume production around 2011, and later nanosheet gate-all-around devices, where the gate surrounds the channel and the subsurface leakage path no longer exists.

Punch-through is often confused with avalanche breakdown, but the two are different failures. Avalanche is carrier multiplication driven by the peak field, so it shows a sharp knee with a fixed voltage; punch-through involves merging depletion regions, usually gives a softer rise of current with voltage, and is not itself destructive if the current is limited. A device can be punch-through limited in one design and avalanche limited in another, depending on doping and geometry, and designers deliberately arrange for avalanche to come first in rugged power parts. Another pitfall is thinking the term applies only to old bipolar devices. It remains a live concern in every advanced process node, in the form of short-channel effects, and in related structures such as reach-through avalanche photodiodes where the effect is deliberately engineered.

Read punch-through as depletion regions meeting in the wrong place: a geometric limit that appears when a base or channel becomes too thin for its voltage, and that explains why every shrinking transistor needed new doping tricks and eventually a new shape.