Velocity Saturation

Classical MOSFET theory assumes that carrier velocity increases linearly with the electric field: v = u * E. This assumption holds for long-channel devices where the electric field in the channel is relatively modest. However, as channel lengths shrink to deep submicron and nanometer scales, the lateral electric field becomes extremely large, and carriers can no longer accelerate indefinitely. Instead, their velocity saturates at a maximum value called saturation velocity vsat. This is the fundamental short-channel effect known as velocity saturation.

Core Concept: Physics of Velocity Saturation

At high electric fields, carriers (electrons or holes) undergo frequent collisions with the crystal lattice (optical phonon scattering). The energy gained from the field is rapidly dissipated in these collisions, preventing further acceleration. The velocity approaches an asymptotic limit: the saturation velocity. For electrons in silicon, vsat is approximately 10^7 cm/s, and for holes it is slightly lower at about 6*10^6 cm/s.

The critical electric field Ec is defined as vsat / u, where u is the low-field mobility. For electrons in Si with u_n approximately 480 cm^2/Vs, Ec = 10^7 / (480 * 10^2) = approximately 2 * 10^4 V/cm. For a 100 nm channel with VDS = 1 V, the average field is 1/(100*10^-7) = 10^5 V/cm, which far exceeds Ec. This is why velocity saturation dominates in all modern short-channel devices.

Mathematical Expression: Modified Saturation Current

In the velocity-saturated regime, the drain current no longer follows the square-law relationship. Instead of ID proportional to (VGS - Vt)^2, the current becomes approximately linear in overdrive voltage. Using the piecewise linear velocity model v = u*E / (1 + E/Ec), the saturation current with velocity saturation is ID_sat = W * Cox * (VGS - Vt - VDSsat) * vsat, where VDSsat reduces from the long-channel value of VGS - Vt.

A simpler and widely used approximation gives: ID_sat = W * Cox * (VGS - Vt)^2 / (2*(1 + (VGS-Vt)/(Ec*L))). When Ec*L is very large (long channel), this reduces to the standard square law. When Ec*L is small (short channel), ID_sat approaches W*Cox*vsat*(VGS - Vt), a linear dependence on overdrive.

The VDSsat under velocity saturation is also modified. It is given by VDSsat = Ec*L*(VGS-Vt)/(Ec*L + VGS-Vt), which is always less than VGS-Vt (the long-channel value). Transistors enter saturation at a lower VDS in short channel devices, meaning they can operate at lower supply voltages.

Practical Understanding

Velocity saturation has important consequences for circuit design. Since the current no longer scales quadratically with VGS - Vt, increasing W/L or overdrive gives diminishing returns in very short channel devices. The transistor transitions from being a square-law device to being approximately a linear transconductance device.

The transconductance gm in the velocity-saturated limit approaches W*Cox*vsat, independent of bias. This means gm no longer increases proportionally with increasing overdrive, which limits amplifier gain at very short channel lengths. It also means that power-delay product does not scale as favorably with channel length shrinking as the ideal model predicts.

For digital design, velocity saturation means that drive current and propagation delay do not scale in proportion with channel length reduction. Designers must account for the actual transconductance efficiency when estimating circuit performance. Technology CAD (TCAD) tools use sophisticated mobility models that capture velocity saturation, mobility degradation, and other short-channel effects simultaneously.

Given:
VGS = 1.0 V, Vt = 0.4 V, VGS - Vt = 0.6 V
L = 0.1 um = 0.1e-6 m
Ec = 4e6 V/m (for short channel NMOS)
W = 1 um, Cox = 10 fF/um^2 = 10e-3 F/m^2
vsat = Ec * un / 1 = Ec * un (using Ec = vsat/un)

Why this formula applies:
Short channel device: Ec*L = 4e6 * 0.1e-6 = 0.4 V. Since Vov = 0.6 V > Ec*L = 0.4 V, velocity saturation is significant.

Formula:
ID_sat = W*Cox*(VGS-Vt)^2 / (2*(1 + (VGS-Vt)/(Ec*L)))

Substitution:
ID_sat = 1e-6 * 10e-3 * (0.6)^2 / (2*(1 + 0.6/0.4))
= 1e-6 * 10e-3 * 0.36 / (2 * 2.5)
= 3.6e-9 / 5

Final Answer:
ID_sat = 0.72 nA scaled to unit width. For W=1um: 0.72 uA

Long-channel prediction: ID = W*Cox*(VGS-Vt)^2/2 = 1.8 uA
Velocity saturation reduces current to 0.72 uA (60% reduction).

Exam Tip: In GATE, the key indicator of velocity saturation is a short channel length (deep submicron). The long-channel ID proportional to (VGS-Vt)^2 becomes approximately ID proportional to (VGS-Vt) in the fully velocity-saturated limit. If the question gives vsat explicitly, use ID = W*Cox*vsat*(VGS-Vt) directly.

• Velocity saturation occurs when the lateral field E exceeds the critical field Ec = vsat/u in short channel MOSFETs.
• Carrier velocity saturates at vsat approx 10^7 cm/s for electrons in Si due to optical phonon scattering.
• Short channel ID_sat is proportional to Vov (linear), not Vov^2 (square law) as in long channel.
• VDSsat is lower in short channel devices, so transistors enter saturation at lower VDS.
• gm in velocity saturation limit: gm = W*Cox*vsat, independent of bias current.

Quick Revision

• Velocity saturation: carrier velocity v = u*E / (1 + E/Ec) saturates at vsat = u*Ec.
• Long channel: ID = (kn/2)*(VGS-Vt)^2. Short channel (vel. sat.): ID = W*Cox*vsat*(VGS-Vt).
• Critical field Ec = vsat / u. For electrons: Ec approx 2*10^4 V/cm.
• Short channel VDSsat = Ec*L*(VGS-Vt)/(Ec*L + VGS-Vt), always less than long channel VDSsat.
• gm (vel. sat.) = W*Cox*vsat. Constant, not proportional to Vov.
• Trap: Do not apply long-channel square-law formula to sub-100nm devices without modification.
• Velocity saturation is why transistor speed improvement slows down at deep submicron nodes.