Tunnel field-effect transistors (TFETs) are studied as a route to low-voltage switching because band-to-band tunneling can, in principle, produce a subthreshold swing below the conventional 60 mV/dec thermal limit. In practice, defects can interfere with tunneling and weaken device performance. Calibrated TCAD analysis of an InAs/InGaAsSb nanowire TFET shows that TFET trap-assisted tunneling near the heterojunction is a major source of degraded swing, while oxide-interface traps affect performance through a different mechanism.
The findings matter to semiconductor device engineers and process teams evaluating III-V heterojunctions: realistic material profiles, trap locations, and gate geometry all influence the simulated and measured current-voltage behavior. The analysis also demonstrates why matching a transfer curve requires understanding the physical mechanisms behind it, not simply adjusting model parameters.
Why defects matter in a heterojunction TFET
A TFET controls current by using the gate electric field to enable band-to-band tunneling (BTBT). In the device examined, a p-doped InGaAsSb source meets an intrinsic InAs nanowire channel, with an n-doped InAs drain. The heterojunction reduces the effective tunneling gap, supporting higher ON-current, but defects in and around this junction can create alternative paths for charge transport.
The simulated structure included a compositionally graded alloy region near the interface. Because composition affects band gap and effective masses—and those parameters influence tunneling—the model used position-dependent material properties derived from physical characterization. It also accounted for strain and the change in InAs crystal structure. This approach avoided fitting the band parameters directly to the measured current-voltage curves.
Measured and simulated temperature-dependent transfer characteristics showed good agreement. That agreement supports the model setup, although the simulated ON-current remained about 30% below the measured value. The analysis notes that uncertainty in determining alloy composition from microscopy could contribute to this difference.
TFET trap-assisted tunneling at the heterojunction
The simulations identify bulk traps adjacent to the heterojunction as the primary cause of subthreshold-swing degradation. In this process, an electron tunnels from the valence band to a trap and then tunnels onward to the conduction band. The trap provides an intermediate step that enables current under conditions where direct BTBT would otherwise be limited.
In the modeled device, this junction-adjacent trap-assisted tunneling (TAT) both worsened the swing and increased leakage. The result points to a practical device-design priority: reducing defects in the material region near the tunneling junction may improve switching characteristics. The location matters because a trap’s energy relative to the tunneling window changes as its position shifts. Single-trap simulations showed that moving a trap within the junction region could substantially alter its contribution to current.
This sensitivity also helps explain why nominally similar nanowire devices may exhibit variation. The study suggests that statistical differences in the location of an individual defect near the junction could contribute to variation in swing. It does not establish that defects elsewhere are absent; rather, it indicates that defects at this particular location have a pronounced effect.
How oxide-interface traps affect electrostatic control
Traps at the InAs/oxide interface affect the device differently. The nanowire’s small diameter and operating conditions inhibit the band bending needed for strong TAT through these interface traps. Instead, the modeled traps produce a weak surface Shockley-Read-Hall generation current and alter the channel electrostatics.
By screening the channel, oxide-interface traps reduce electrostatic coupling between the gate and the channel. The weaker gate control further degrades the subthreshold swing, even though oxide-interface TAT is not the dominant current mechanism in this device. The simulations also reproduced a measured shift in threshold voltage as drain-to-source voltage changed. In the model, the shift resulted from changing trap occupancy and interface charge; it disappeared when the oxide-interface traps were deactivated.
Explaining negative transconductance
At higher gate bias, the measured drain current first rises and then decreases over part of the transfer curve, producing negative transconductance. The simulations reproduced this behavior and linked it to the changing BTBT generation profile.
At lower gate bias, tunneling begins at the InAs/InGaAsSb heterointerface and strengthens as gate bias increases. At a higher bias, the source region beneath the gate enters volume inversion. The resulting change progressively quenches BTBT generation at the heterojunction, so drain current falls as gate voltage continues to rise. This mechanism explains why the transfer curve cannot be interpreted as a simple, continually increasing response to gate bias.
Gate alignment can improve performance—but demands precision
Simulations of an alternative geometry found that aligning the gate with the InAs/InGaAsSb heterojunction improved both ON-current and subthreshold swing compared with the source-overlapped geometry studied. In the aligned structure, the electric fields at the junction combine, strengthening the field that supports tunneling.
The benefit was sensitive to placement. A 5 nm overlap on the source reversed the improvement, while a 5 nm underlap caused a less severe degradation than the overlap. These are simulation results for the studied device, not a universal tolerance specification. They nevertheless show why gate placement should be treated as a critical design variable when evaluating this geometry.
Practical implications for device modeling
For engineers assessing TFET structures, the study highlights several useful modeling and design considerations:
- Represent material variation: composition-dependent band structure and strain can influence predicted tunneling behavior.
- Separate trap mechanisms: bulk traps near a heterojunction and oxide-interface traps affect current and gate control in different ways.
- Consider defect location: a localized trap’s position relative to the tunneling window can change its impact.
- Model geometry carefully: gate alignment may improve simulated swing and ON-current, but small shifts can reduce the gain.
For industrial automation decision-makers, these results are relevant upstream: the performance of electronic components used in control and computing hardware depends in part on semiconductor device design and manufacturing. The analysis does not establish system-level benefits, but it does clarify the device-level tradeoffs that engineers must resolve.
Conclusion
In the analyzed InAs/InGaAsSb nanowire TFET, traps near the heterojunction were the strongest source of subthreshold-swing degradation, while oxide-interface traps primarily weakened gate-to-channel coupling. TCAD also explained the observed negative transconductance and predicted improved swing and ON-current with gate alignment. The practical lesson is to evaluate defect location, material profiles, and gate geometry together: each can shape the device’s tunneling behavior, and geometric gains may be sensitive to placement.