Negative Capacitance FET Ncfet

# Negative Capacitance Field-Effect Transistors (NCFET): Physics, Materials & Scaling

The Negative Capacitance Field-Effect Transistor (NCFET) is a steep-slope solid-state device architecture engineered to overcome the fundamental thermodynamic switching limit of conventional MOSFETs—the Boltzmann Tyranny ($\text{SS} \ge 60\,\text{mV/dec}$ at $300\,\text{K}$). First proposed by Sayeef Salahuddin and Supriyo Datta in 2008, the NCFET integrates a thin ferroelectric layer in series with the conventional gate dielectric stack. By operating the ferroelectric in its intrinsic, thermodynamically stabilized negative differential capacitance regime, the gate stack generates internal step-up voltage amplification ($\partial \psi_s / \partial V_G > 1$), enabling sub-thermal subthreshold swing, aggressive supply voltage scaling ($V_{DD} < 0.5\,\text{V}$), and drastic reductions in dynamic switching energy.

Unlike hysteretic ferroelectric RAM (FeRAM) or ferroelectric field-effect transistors (FeFET) used for non-volatile storage, a logic NCFET operates in a strictly non-hysteretic, reversible equilibrium state. This condition is achieved through precise capacitance matching, wherein the positive capacitance of the underlying MOS dielectric and channel stabilizes the otherwise unstable negative capacitance of the ferroelectric layer across the full digital switching trajectory.

Landau Free Energy Landscape & Capacitance Matching thermodynamic stabilization of negative differential capacitance (C_FE < 0) in series with MOS dielectric 1. ISOLATED FERROELECTRIC (UNSTABLE NEGATIVE C) Polarization (P) Free Energy U(P) Negative Curvature (d²U/dP² < 0) -P_r +P_r Landau-Khalatnikov Energy Formulation: U(P) = α·P² + β·P⁴ + γ·P⁶ - E·P (α < 0) Intrinsic Negative Capacitance: C_FE = (d²U/dP²)⁻¹ < 0 around P ≈ 0 In isolated material, this state is unstable and spontaneously collapses into hysteretic domains (±P_r). 2. SERIES STABILIZED NCFET (|C_FE| > C_MOS) Charge (Q) Total Energy U_tot(Q) Single Stable Equilibrium (Q = 0 at V_G = 0) U_MOS = Q²/(2·C_MOS) Thermodynamic Stability Condition: |C_FE| > C_MOS ==> 1/C_tot = 1/C_MOS - 1/|C_FE| > 0 Internal Voltage Gain (A_V > 1): A_V = ∂ψ_s / ∂V_G = |C_FE| / (|C_FE| - C_MOS) > 1 Converts unstable negative curvature into non-hysteretic, sub-thermal voltage amplification across the silicon channel.

---

## 1. Thermodynamic Theory of Negative Capacitance

The physical basis of negative capacitance rests on the Landau-Devonshire-Khalatnikov (LDK) phenomenological theory of phase transitions.

### 1.1 Free Energy Potential & Polarization Curvature
The Helmholtz free energy density $U(P)$ of a ferroelectric thin film expanded in powers of polarization $P$ is expressed as:

$$U(P) = \alpha P^2 + \beta P^4 + \gamma P^6 - E P$$

Where $\beta > 0$, $\gamma \ge 0$, and $\alpha$ is temperature-dependent via the Curie-Weiss law:

$$\alpha = \frac{T - T_C}{2 \epsilon_0 C_{CW}}$$

For operating temperatures below the Curie temperature ($T < T_C$), the dielectric stiffness coefficient is negative ($\alpha < 0$). In this regime:
1. The electric field $E$ across the ferroelectric is the first derivative of the free energy:

$$E = \frac{\partial U}{\partial P} = 2\alpha P + 4\beta P^3 + 6\gamma P^5$$

2. The differential dielectric susceptibility $\chi_{FE}$ (and corresponding capacitance $C_{FE}$) relates directly to the second derivative:

$$\epsilon_0 \chi_{FE}^{-1} = \frac{\partial E}{\partial P} = \frac{\partial^2 U}{\partial P^2} = 2\alpha + 12\beta P^2 + 30\gamma P^4$$

3. In the vicinity of the unpolarized state ($P \approx 0$):

$$\left. \frac{\partial^2 U}{\partial P^2} \right|_{P \to 0} \approx 2\alpha < 0$$

Consequently, the intrinsic differential capacitance of the ferroelectric layer is strictly negative:

$$C_{FE} = \frac{dQ}{dV_{FE}} = \frac{A \cdot dP}{t_{FE} \cdot dE} = \frac{A}{t_{FE} \left(\frac{\partial^2 U}{\partial P^2}\right)} < 0$$

In an unconstrained standalone ferroelectric capacitor, states with negative curvature ($\partial^2 U / \partial P^2 < 0$) are thermodynamically unstable. The crystal spontaneously phase-separates into polarized domains at the double-well energy minima ($\pm P_r = \pm \sqrt{-\alpha / 2\beta}$), exhibiting classic hysteretic ferroelectric switching.

---

## 2. Capacitance Matching & Internal Voltage Amplification

NCFET Gate Architecture & Sub-Boltzmann Transfer Characteristics cross-sectional gate stack and semilogarithmic I_D–V_GS showing sub-60 mV/dec steep switching NCFET GATE STACK INTEGRATION Gate Electrode (TiN / Ru / W) Ferroelectric Layer (HZO, t = 3–5 nm) Orthorhombic Phase (Pca2₁) | C_FE < 0 Interfacial Dielectric (SiO₂ / Al₂O₃, t = 0.7 nm) Silicon / SiGe Channel (C_MOS) Surface Potential ψ_s amplified by A_V > 1 Series Capacitance Network: • Ferroelectric: C_FE < 0 (t_FE ~ 4 nm) • Interfacial + Semiconductor: C_MOS > 0 Total Effective Capacitance: C_tot = (|C_FE| · C_MOS) / (|C_FE| - C_MOS) > C_MOS Boosts effective gate control beyond physical EOT SUB-BOLTZMANN STEEP-SLOPE SWITCHING 10⁻³ 10⁻⁵ 10⁻⁷ 10⁻⁹ 10⁻¹¹ Drain Current I_D (A/μm) → 0.0 0.2 0.4 0.6 0.8 Gate-to-Source Voltage V_GS (V) → Baseline MOSFET SS = 65 mV/dec V_DD = 0.70 V NCFET (Steep-Slope) SS = 42 mV/dec V_DD = 0.40 V (-43%) ΔV_DD = 300 mV Boltzmann Limit (60 mV/dec)

### 2.1 Series Capacitance Stabilization
When the ferroelectric layer ($C_{FE} < 0$) is placed in series with a positive MOS dielectric layer and underlying silicon semiconductor body ($C_{MOS} > 0$), the total free energy $U_{tot}(Q)$ of the integrated gate stack becomes:

$$U_{tot}(Q) = U_{FE}(Q) + U_{MOS}(Q) = U_{FE}(Q) + \frac{Q^2}{2 C_{MOS}} - V_G Q$$

For the overall system to operate in a non-hysteretic, thermodynamically stable state, the curvature of the total free energy must be strictly positive:

$$\frac{\partial^2 U_{tot}}{\partial Q^2} = \frac{1}{C_{FE}} + \frac{1}{C_{MOS}} > 0$$

Defining $C_{FE} = -|C_{FE}|$, this stability requirement simplifies to the capacitance matching condition:

$$-\frac{1}{|C_{FE}|} + \frac{1}{C_{MOS}} > 0 \implies \frac{1}{C_{MOS}} > \frac{1}{|C_{FE}|} \implies |C_{FE}| > C_{MOS}$$

When $|C_{FE}|$ is designed to be slightly larger than $C_{MOS}$, the total gate capacitance $C_{tot}$ exceeds the baseline MOS capacitance:

$$C_{tot} = \left( \frac{1}{C_{FE}} + \frac{1}{C_{MOS}} \right)^{-1} = \frac{|C_{FE}| \cdot C_{MOS}}{|C_{FE}| - C_{MOS}} > C_{MOS}$$

This state achieves effective capacitance boosting without thinning the physical interfacial oxide, bypassing quantum mechanical gate dielectric tunneling leakage limits.

### 2.2 Internal Voltage Gain & Subthreshold Swing Derivation
The subthreshold swing ($\text{SS}$) of a field-effect transistor is defined as the change in gate voltage required to modulate the drain current by one order of magnitude:

$$\text{SS} = \frac{\partial V_G}{\partial (\log_{10} I_D)} = \left( \frac{\partial \log_{10} I_D}{\partial \psi_s} \cdot \frac{\partial \psi_s}{\partial V_G} \right)^{-1}$$

In conventional MOSFETs, the body factor $m = (\partial \psi_s / \partial V_G)^{-1} = 1 + C_{dep} / C_{ox} > 1$. Therefore, $\partial \psi_s / \partial V_G$ is always less than unity. In an NCFET:

$$\psi_s = V_G - V_{FE} = V_G - \frac{Q}{C_{FE}}$$

Differentiating with respect to gate voltage yields the internal voltage amplification factor $A_V$:

$$A_V = \frac{\partial \psi_s}{\partial V_G} = 1 - \frac{1}{C_{FE}} \frac{\partial Q}{\partial V_G} = 1 - \frac{C_{tot}}{C_{FE}} = \frac{|C_{FE}|}{|C_{FE}| - C_{MOS}} > 1$$

Substituting $A_V$ into the subthreshold swing expression:

$$\text{SS} = \frac{k_B T}{q} \ln(10) \left( 1 + \frac{C_{dep}}{C_{ox}} \right) \cdot \frac{1}{A_V}$$

Because $A_V > 1$, the subthreshold swing drops below the classical thermodynamic limit:

$$\text{SS} < \frac{k_B T}{q} \ln(10) \approx 59.6\,\text{mV/dec at } 300\,\text{K}$$

State-of-the-art laboratory NCFET demonstrations with optimized hafnium-zirconium oxide (HZO) routinely achieve $\text{SS} = 40\text{ to }50\,\text{mV/dec}$ over multiple decades of drain current at room temperature.

---

## 3. Ferroelectric Hafnia Materials Science: Orthorhombic $Pca2_1$ Phase

Historically, ferroelectric materials such as lead zirconate titanate ($\text{Pb}[\text{Zr}_x\text{Ti}_{1-x}]\text{O}_3$ or PZT) and barium titanate ($\text{BaTiO}_3$) were incompatible with CMOS manufacturing due to toxic heavy elements, low bandgaps ($E_g \approx 3.2\,\text{eV}$), high crystallization temperatures, and critical thickness scaling limits below $10\,\text{nm}$.

The discovery of ferroelectricity in doped hafnium dioxide ($\text{HfO}_2$) by Böscke et al. in 2011 revolutionized NCFET research, as hafnia is already the standard high-k dielectric in production fab lines:

+----------------------------------------------------------------------------------------------------+
|                                    HAFNIA CRYSTALLINE POLYMORPHS                                   |
+---------------------+-------------------+---------------------+------------------------------------+
| PHASE               | SPACE GROUP       | SYMMETRY            | FERROELECTRIC STATUS               |
+---------------------+-------------------+---------------------+------------------------------------+
| Monoclinic (m-phase)| P2₁/c             | Centrosymmetric     | Paraelectric (Dielectric k ~ 16);  |
|                     |                   | (Inversion center)  | Undesired parasitic phase          |
+---------------------+-------------------+---------------------+------------------------------------+
| Tetragonal (t-phase)| P4₂/nmc           | Centrosymmetric     | Antiferroelectric / High-k (k ~ 35)|
|                     |                   | (High symmetry)     | Precursor phase during annealing   |
+---------------------+-------------------+---------------------+------------------------------------+
| Orthorhombic (o-III)| Pca2₁             | Non-Centrosymmetric | FERROELECTRIC POLAR (k ~ 30-40);   |
|                     |                   | (Polar axis along c)| Active phase for NCFET operation   |
+---------------------+-------------------+---------------------+------------------------------------+

### 3.1 Stabilizing the Orthorhombic Phase
The polar orthorhombic $Pca2_1$ phase is metastable at room temperature. In production NCFET gate modules, it is stabilized through:
1. Doping and Stoichiometry: Alloying hafnia with equal parts zirconium to form $\text{Hf}_{0.5}\text{Zr}_{0.5}\text{O}_2$ (HZO) lowers the kinetic crystallization barrier for the orthorhombic phase. Other effective dopants include Si ($2\text{ to }4\,\text{mol}\%$), Al ($4\text{ to }6\,\text{mol}\%$), and La ($1\text{ to }3\,\text{mol}\%$).
2. Mechanical Encapsulation Stress: Annealing the amorphous ALD HZO film capped beneath a rigid metallic $\text{TiN}$ gate electrode (thickness $10\text{ to }20\,\text{nm}$) exerts strong in-plane tensile stress ($\sigma_{tensile} > 1.5\,\text{GPa}$), kinetically suppressing transformation into the thermodynamically favored monoclinic phase.
3. Rapid Thermal Annealing (RTA): Rapid millisecond flash annealing at $450^\circ\text{C}\text{ to }600^\circ\text{C}$ freezes the crystalline grains into the polar $Pca2_1$ structure without exceeding CMOS thermal budgets.

---

## 4. Performance, Power, and Area (PPA) Scaling Matrix

+----------------------------------------------------------------------------------------------------+
|                               MOSFET vs. FINFET vs. GAA vs. NCFET COMPARISON                       |
+-----------------------+-------------------+-------------------+------------------+-----------------+
| METRIC                | 7nm FINFET        | 3nm GAA NANOSHEET | 2nm GAA (N2)     | 2nm + NCFET     |
+-----------------------+-------------------+-------------------+------------------+-----------------+
| Nominal Supply (V_DD) | 0.75 V            | 0.70 V            | 0.65 V           | 0.40 to 0.45 V  |
+-----------------------+-------------------+-------------------+------------------+-----------------+
| Subthreshold Swing    | 70 mV/dec         | 64 mV/dec         | 62 mV/dec        | 42 to 48 mV/dec |
+-----------------------+-------------------+-------------------+------------------+-----------------+
| Body Factor (m)       | 1.15              | 1.06              | 1.04             | 0.70 to 0.80    |
+-----------------------+-------------------+-------------------+------------------+-----------------+
| Active Power Density  | Baseline (1.0×)   | 0.75×             | 0.62×            | 0.28× (-55%)    |
| (P_dyn ∝ C·V_DD²·f)   |                   |                   |                  |                 |
+-----------------------+-------------------+-------------------+------------------+-----------------+
| Hysteresis Window     | None (0 mV)       | None (0 mV)       | None (0 mV)      | < 5 mV (Matched)|
+-----------------------+-------------------+-------------------+------------------+-----------------+
| Max Frequency @ 0.5V  | Degraded (-60%)   | Degraded (-45%)   | Moderate (-30%)  | Full Clock (1.0×|
+-----------------------+-------------------+-------------------+------------------+-----------------+

Take negative capacitance fet ncfet further

Ask the copilot about this term, or have our engineers assess it against your process.

Ask CFSGPT about this → Request a technical assessment → Open the CFS app