A negative-capacitance FET (NCFET) inserts a thin ferroelectric layer (here HfO₂-style) in series with the ordinary gate oxide. The ferroelectric's free energy follows the Landau-Devonshire form U(P) = a·P² + b·P⁴, and in its ferroelectric phase a < 0, giving a double-well energy landscape and a region of negative slope dV/dQ — a genuine negative capacitance.
V_FE(Q) = t_FE·(a·Q + b·Q³) (Landau equation of state)
V_MOS(Q) = Q / C_ox (linear channel capacitor)
V_g = V_FE(Q) + V_MOS(Q) (series charge continuity)
Solving this cubic for Q at each V_g (Newton's method, warm-started from the previous point — which is why sweeping up/down can trace slightly different paths, a real "hysteretic NCFET" artifact seen when the stack isn't perfectly capacitance-matched) gives the internal voltage gain:
A = dψ_s/dV_g = 1 / (1 + C_ox·t_FE·(a + 3b·Q²))
Because a < 0, the denominator drops below 1 and A > 1: the ferroelectric layer amplifies the gate's pull on the channel surface potential ψ_s. Since subthreshold current follows I_ds ∝ exp(ψ_s/V_T), the swing becomes SS = (60 mV/dec) / A — below the fundamental Boltzmann limit that traps ordinary CMOS at 60 mV/decade at room temperature, letting a transistor turn on with a smaller voltage swing and less energy per switch.
- Vg — sweeps the gate bias; the dipole domains in the ferroelectric slab switch as the local charge crosses each domain's coercive threshold.
- NC strength — the magnitude of the Landau a-coefficient; stronger negative capacitance gives a bigger internal gain and steeper SS, up to the point the stack becomes unstable and snaps (visible as a hysteresis jump).
- tFE — ferroelectric layer thickness; thicker layers couple more strongly to the MOS capacitor.
- Mode toggle — switches the ferroelectric layer out of the circuit entirely, leaving a baseline MOSFET pinned to the 60 mV/decade limit (A = 1) for comparison.