Access Official Technical PDF

Share Summary:LinkedInEmail

A useful TFET SPICE model needs to describe more than a transistor’s steep turn-on characteristic. For circuit analysis, it must also represent leakage, forward-bias behavior and transitions across the voltage conditions that devices encounter in a circuit. A compact modeling approach developed for tunneling field-effect transistors (TFETs) addresses this challenge by combining a tunneling-current model with BSIM4-based MOSFET behavior, then implementing the result in Verilog-A for simulation.

The distinction matters to design teams evaluating TFETs for ultra-low-power circuits. Their band-to-band tunneling operation differs from conventional MOSFET operation, so a model focused only on the on and off states may not capture the range of behavior needed for circuit-level analysis.

Why TFET circuit modeling needs broad voltage coverage

In the conventional n-type TFET structure considered here, a heavily doped p-type source, an intrinsic channel and an n-type drain form the device. Applying a positive gate-to-source voltage (Vgs) enables carriers to tunnel from the source edge into the channel through band-to-band (BTB) tunneling. With a drain-to-source voltage (Vds) applied, carriers can then reach the drain.

At Vgs = 0 V, the device is in its off state: tunneling from the source valence band to the channel conduction band is inhibited, even when Vds is positive. However, a practical circuit model must account for more than this basic switching description. The dominant current mechanism changes with the signs of Vgs and Vds:

  • Vds ≥ 0 and Vgs > 0: BTB tunneling is the dominant current mechanism, producing the TFET’s characteristic steep subthreshold behavior.
  • Vds ≥ 0 and Vgs ≤ 0: The device is off, and gate and junction leakage must be considered.
  • Vds < 0 and Vgs > 0: The source-to-drain p-n junction is forward biased, making diode current dominant.
  • Vds < 0 and Vgs ≤ 0: Diode current and gate leakage both contribute.

These operating regions explain why a model that describes only reverse-biased tunneling may be insufficient when simulating interconnected devices. Current paths that are secondary in one bias condition can become important in another.

How the TFET SPICE model combines tunneling and BSIM4

The compact model represents the device as two parallel current contributions: a TFET component and a MOSFET component based on BSIM4. In simplified form, the total drain current is the sum of the TFET-model current and the MOSFET-model current. The TFET component represents BTB tunneling and p-n diode current; the BSIM4 component provides a basis for modeling leakage, including gate and junction leakage.

This structure pairs a mechanism-specific description of tunneling with an established MOSFET model framework. It also supports representation of current in different voltage regions rather than treating the transistor as having only a single tunneling-controlled operating mode.

Representing charge and tunneling current

For the reverse-bias current formulation, the model describes total charge as a weighted combination of tunneling charge and oxide charge associated with the gate-to-source overlap region. A smoothing function sets the weighting between these contributions. The drain current is then expressed using the total charge, device width and an injection velocity at the source edge, within a drift-diffusion approach under the gradual-channel approximation.

The tunneling-charge formulation incorporates the band-gap energy and reverse bias at the source side. The charge-based approach is intended to represent the steep subthreshold swing while maintaining a continuous transition from weak to strong inversion. That continuity is important in circuit simulation, where a model must behave consistently as terminal voltages change.

Model parameters are not all interpreted like conventional MOSFET parameters. For example, the reported formulation identifies the tunneling-related parameter Ntun as strongly associated with subthreshold swing, while Vth0 is associated with threshold voltage. Parameters such as μtun and μth have dimensions resembling effective mobility, but the paper cautions that they do not have a straightforward proportional relationship to drain current through the charge terms.

Calibration and circuit-level evaluation

The model was implemented in a circuit simulator using a Verilog-A description. Parameters for a conventional TFET structure were extracted by comparing model behavior with in-house two-dimensional technology computer-aided design (2D TCAD) simulation results. The simulated device had a 120 nm channel length, a 2 nm oxide thickness and a substrate concentration of 1 × 1016 cm-3. Its structure was configured to demonstrate a steep subthreshold swing and the targeted on-current behavior.

Evaluation included current-voltage characteristics across gate and drain biases, leakage behavior and transistor-level SPICE convergence. The model’s results were reported to agree with the TCAD simulations over the targeted operating voltage range. This is a calibration result for the device structure and conditions examined; it should not be read as a guarantee that the same parameter set applies to other TFET designs.

To check circuit-level use, the researchers simulated an 81-stage ring oscillator using the proposed model. This step moved evaluation beyond individual transistor curves and demonstrated that the model could be used in a multi-device circuit simulation. It provides a relevant example for engineers assessing whether a compact model can support analysis involving repeated switching stages.

Practical implications for design teams

For engineers and system integrators assessing TFET-based circuits, the key modeling lesson is to match model coverage to the intended analysis. A steep subthreshold characteristic alone does not describe all circuit-relevant current paths. Gate leakage, junction leakage and forward-biased diode current can also affect simulated behavior, depending on terminal bias.

A combined TFET and BSIM4 approach offers a way to represent these mechanisms in a circuit simulator while retaining a charge-based formulation for tunneling operation. Parameter extraction against device-level TCAD results, followed by transistor- and circuit-level checks, connects the compact model to a defined device structure and intended use.

Conclusion

TFET circuit analysis calls for a model that spans tunneling-dominated operation, off-state leakage and forward-bias conditions. The described TFET SPICE model combines BTB tunneling and diode behavior with BSIM4-based leakage modeling, uses a weighted charge formulation for reverse-bias current, and is implemented in Verilog-A. Its TCAD comparison and 81-stage ring-oscillator simulation illustrate a practical validation path: calibrate the model for the target device, check its behavior across relevant voltage regions, and then assess its suitability in circuit-level simulations.

Access Official Technical PDF

About The Author