Channel Length
Channel length is the distance between the source and drain of a MOSFET — the length of the region the gate controls to switch the transistor on and off — and it is the single most important dimension in the whole story of transistor scaling. When people speak of a chip being built on a particular process node, they are referring, historically, to a measure closely tied to the transistor's channel length: the smaller that length, the more advanced the process. Channel length matters because it directly governs how fast the transistor switches, how densely transistors can be packed, and when the troublesome short-channel effects begin to appear.
Channel length sets transistor speed because it is the distance charge carriers must travel to cross the device, so a shorter channel means faster switching. The transistor turns on by forming a conducting channel through which carriers move from source to drain; the shorter that path, the less time the carriers need to traverse it and the faster the device can change state. This is the primary reason shrinking channel length has been so valuable — it buys speed directly. Each generation that reduced channel length produced transistors that could switch more quickly, which translated into faster circuits, and this relationship is one of the fundamental motivations for scaling the dimension down.
Channel length is the dimension most directly associated with the process node name, which is why node labels have historically tracked it even as the relationship grew looser. For much of the industry's history, a process generation was identified by a number corresponding roughly to the smallest channel length it could produce, and shrinking that number from generation to generation was the headline achievement of each advance. This is why channel length is often treated as a proxy for how advanced a manufacturing process is; it was the dimension that most cleanly captured the progress of scaling, even though modern node names have become marketing labels less tightly bound to any single physical measurement.
The lower limit on how short a channel can usefully be is set by short-channel effects, making channel length the dimension where the benefits and the limits of scaling collide. Shrinking the channel delivers speed and density, but once it becomes short enough that the source and drain begin to influence the channel and the gate loses full control, the transistor starts to misbehave — thresholds drift, leakage rises, and switching degrades. So channel length is simultaneously the dimension most worth reducing and the one whose reduction first runs into fundamental trouble, which is why so much device engineering focuses precisely on keeping transistors well-behaved as this length shrinks.
Because channel length is so central, controlling it precisely during manufacturing is a critical and difficult task. Small, unintended variations in channel length across a chip or between chips translate into variations in transistor speed and threshold, so the manufacturing process must reproduce this tiny dimension with great consistency. The difficulty of defining and controlling ever-shorter channel lengths is one of the main reasons advanced lithography and process control are so demanding; the economic value of the dimension and the physical difficulty of controlling it as it shrinks are two sides of why channel length dominates discussions of process technology.
Read channel length as the master dimension of scaling: the source-to-drain distance that sets how fast a transistor switches and how densely transistors pack, the number that historically named the process node, and the very dimension whose shrinking first collides with the short-channel effects that limit how small a transistor can go.