TCAD Simulation Study of ESD Behavior of InGaAs / InP Heterojunction Tunnel FETs

: For the ﬁrst time, we investigated the electrostatic discharge (ESD) behavior of an InGaAs / InP heterojunction tunneling ﬁeld e ﬀ ect transistor (HTFET). The device structure in this study has a high on-state current without extra process steps. Under the positive transmission line pulse (TLP) simulation, the band-to-band tunneling (BTBT) current acts as an important initial current to accelerate the occurrence of impact ionization and the device is turned on quickly. Under the negative transmission line pulse (TLP) simulation, the operating principle of the HTFET is the same as for a poly-bounded diode. The ESD robustness of the device under TLP simulation are evaluated, and the impact factors, with regard to ESD robustness and failure mode, are discussed. Finally, the device behavior under very fast transmission line pulse (VFTLP) simulations with di ﬀ erent rise times and pulse widths is also investigated. The results show that this device may be used for the ESD protection of next-generation III–V technology.


Introduction
The phenomenon of band-to-band tunneling (BTBT) in tunnel field effect transistors (TFETs) enables them to achieve a subthreshold swing (SS) below 60 mV/dec at room temperature. TFETs have the potential to replace metal-oxide-semiconductor field effect transistors (MOSFETs) in next-generation integrated circuits [1][2][3]. Over the last decade, several device structures and materials, such as Si, Ge, SiGe, and III-V materials, have been proposed to fabricate TFETs with a high on-current (I on ) and low SS [4][5][6][7]. Among these material systems, In 0.53 Ga 0.47 As/InP heterojunction TFETs have a higher I on /I off ratio compared to TFETs that are based on silicon in low-voltage applications because of the small lattice mismatch, small direct energy bandgap, and small tunneling mass [8,9]. Electrostatic discharge (ESD) is one of the key reliability issues of integrated circuits (ICs), especially in nanoscale technologies [10,11]. Recently, TFETs, which are based on silicon and SiGe as ESD protection devices, were investigated [12][13][14][15]. However, the electrostatic-discharge capacity of TFETs limits their application as ESD devices. They can, however, replace the traditional reverse diode as a secondary discharge device to reduce the clamp voltage and enhance the ESD robustness of the whole-chip protection network [16]. In addition, the breakdown voltage of the gate-oxide decreases with decreasing thickness [10], which is necessary to search for a potential device with lower trigger voltage. Therefore, voltage. Therefore, InGaAs/InP heterojunction tunneling field effect transistor (HTFET) are promising candidates for ESD protection devices using the corresponding technology. In this paper, the direct-current (DC) and ESD characteristics of InGaAs/InP HTFETs are investigated using a technology computer aided design (TCAD) simulation. The impact of device parameters, such as doping concentration, channel length, and junction depth, on BTBT generation and on-state current is explored. Transmission line pulse (TLP) simulations are conducted and show that InGaAs/InP HTFETs have a low trigger-voltage under positive stress, which makes them suitable for future low-voltage ICs. Under negative TLP, InGaAs/InP HTFETs operate like poly-bounded diodes and can conduct a high current. Finally, the very fast TLP (VFTLP) characteristic of the InGaAs/InP HTFET is also discussed.

Model Calibration and Device Structure
The simulations for this work were performed using the Sentaurus-TCAD tool. We used the non-local dynamic BTBT model, the van Overstraeten-de Man avalanche-generation model, the high-field saturation model, the Philips unified mobility model, the thermionic emission model, band-gap narrowing, and the Shockley-Read-Hall recombination model. The quantum confinement has an insignificant impact when the device size is larger than 7 nm, and in turn was not considered in this paper [17,18]. The parameters that were used in this study were calibrated according to [7,9], and detailed values of the band structure are shown Table 1. A good match between the published data and our simulated result was obtained, as shown in Figure 1.
An HTFET structure was designed with In0.53Ga0.47As as a source material and InP as a substrate and drain material. The cross section of the device is shown in Figure 2. A metal gate (work function = 4.6 eV) with an HfO2 of 3 nm was used as a gate stack throughout the simulation. The doping concentrations of the source (P-type), drain (N-Type), and substrate (N-Type) were kept at 1 × 10 19 cm −3 , 1 × 10 19 cm −3 , and 1 × 10 17 cm −3 , respectively. The lengths of the source, channel, and drain were all set to 50 nm. Finally, the depth of the junction was set to Xj = 20 nm. An HTFET structure was designed with In 0.53 Ga 0.47 As as a source material and InP as a substrate and drain material. The cross section of the device is shown in Figure 2. A metal gate (work function = 4.6 eV) with an HfO2 of 3 nm was used as a gate stack throughout the simulation. The doping concentrations of the source (P-type), drain (N-Type), and substrate (N-Type) were kept at 1 × 10 19 cm −3 , 1 × 10 19 cm −3 , and 1 × 10 17 cm −3 , respectively. The lengths of the source, channel, and drain were all set to 50 nm. Finally, the depth of the junction was set to X j = 20 nm.  Figure 3a shows that when the doping concentration in the source region is increased (while other parameters of the HTFET remain unchanged), the Ion/Ioff ratio increases. The off-state current essentially represents the leakage of a reverse PN junction. Furthermore, the increase in doping concentration reduces the width of the space-charge region of the PN junction and enhances the drift motion of carriers. The energy band near the source/channel junction begins to bend with increasing gate voltage. When the doping concentration is increased, the band bends significantly (see Figure  3b) and, in turn, the on-state current increases. Finally, the value of the source doping concentration was set in this paper to 1 × 10 19 cm −3 , considering the process cost and complexity.

Effect of Channel Length
The long channel helps the carrier to participate in the recombination and then reduce the current in the off-state, as shown in Figure 4. In addition, once the current increases by a certain amount, the current that flows through the device depends on the BTBT generation rather than on the carrier recombination. Furthermore, increasing the size of the device facilitates heat dissipation as well as the maintenance of the high on-state current. Thus, the device channel length is set to a relatively large value (50 nm).  Figure 3a shows that when the doping concentration in the source region is increased (while other parameters of the HTFET remain unchanged), the I on /I off ratio increases. The off-state current essentially represents the leakage of a reverse PN junction. Furthermore, the increase in doping concentration reduces the width of the space-charge region of the PN junction and enhances the drift motion of carriers. The energy band near the source/channel junction begins to bend with increasing gate voltage. When the doping concentration is increased, the band bends significantly (see Figure 3b) and, in turn, the on-state current increases. Finally, the value of the source doping concentration was set in this paper to 1 × 10 19 cm −3 , considering the process cost and complexity.  Figure 3a shows that when the doping concentration in the source region is increased (while other parameters of the HTFET remain unchanged), the Ion/Ioff ratio increases. The off-state current essentially represents the leakage of a reverse PN junction. Furthermore, the increase in doping concentration reduces the width of the space-charge region of the PN junction and enhances the drift motion of carriers. The energy band near the source/channel junction begins to bend with increasing gate voltage. When the doping concentration is increased, the band bends significantly (see Figure  3b) and, in turn, the on-state current increases. Finally, the value of the source doping concentration was set in this paper to 1 × 10 19 cm −3 , considering the process cost and complexity.

Effect of Channel Length
The long channel helps the carrier to participate in the recombination and then reduce the current in the off-state, as shown in Figure 4. In addition, once the current increases by a certain amount, the current that flows through the device depends on the BTBT generation rather than on the carrier recombination. Furthermore, increasing the size of the device facilitates heat dissipation as well as the maintenance of the high on-state current. Thus, the device channel length is set to a relatively large value (50 nm).

Effect of Channel Length
The long channel helps the carrier to participate in the recombination and then reduce the current in the off-state, as shown in Figure 4. In addition, once the current increases by a certain amount, the current that flows through the device depends on the BTBT generation rather than on the carrier recombination. Furthermore, increasing the size of the device facilitates heat dissipation as well as the maintenance of the high on-state current. Thus, the device channel length is set to a relatively large value (50 nm).

Effect of Junction Depth
It is helpful to study the junction depth of the HTFET, with respect to technology operation and device characteristics. The on-state current, as shown in Figure 5, which flows through the device, increases significantly when Xj increases from 10 to 20 nm. The junction depth increase is equivalent to increasing the tunneling area at the source-channel interface, which in turn increases the tunneling current and lowers the on-resistance. When the junction depth is set to 30 nm, the on-state current does not change significantly. This is mainly because a large number of relaxed holes accumulates at the interface between the source and the channel, which lowers the potential near the interface and degrades the band bending [14]. The default value of the junction depth in this paper was set to 20 nm.

Effect of the Doping Concentration of the Substrate
Increasing the doping concentration of the substrate has undesired effects on the extension of the width of the space charge region of the PN junction and the enhancement of the electric field, as shown in Figure 6a,b. This greatly promotes the drift motion of carriers and increases the off-state current, as shown in Figure 7a. Subsequently, Figure 7b shows that when the gate bias voltage rises to 0.2 V, the substrate with a higher doping concentration will more likely lead to valence-band bending at the interface of the PN junction, which increases the BTBT current. Finally, if the VG continues to increase, the on-state current of the point-HTFETs with different substrate doping concentrations tends to saturate gradually. This is because the high voltage at the gate induces a large number of electrons underneath the gate oxide, which offset the benefits of the substrate with a high doping concentration.

Effect of Junction Depth
It is helpful to study the junction depth of the HTFET, with respect to technology operation and device characteristics. The on-state current, as shown in Figure 5, which flows through the device, increases significantly when X j increases from 10 to 20 nm. The junction depth increase is equivalent to increasing the tunneling area at the source-channel interface, which in turn increases the tunneling current and lowers the on-resistance. When the junction depth is set to 30 nm, the on-state current does not change significantly. This is mainly because a large number of relaxed holes accumulates at the interface between the source and the channel, which lowers the potential near the interface and degrades the band bending [14]. The default value of the junction depth in this paper was set to 20 nm.

Effect of Junction Depth
It is helpful to study the junction depth of the HTFET, with respect to technology operation and device characteristics. The on-state current, as shown in Figure 5, which flows through the device, increases significantly when Xj increases from 10 to 20 nm. The junction depth increase is equivalent to increasing the tunneling area at the source-channel interface, which in turn increases the tunneling current and lowers the on-resistance. When the junction depth is set to 30 nm, the on-state current does not change significantly. This is mainly because a large number of relaxed holes accumulates at the interface between the source and the channel, which lowers the potential near the interface and degrades the band bending [14]. The default value of the junction depth in this paper was set to 20 nm.

Effect of the Doping Concentration of the Substrate
Increasing the doping concentration of the substrate has undesired effects on the extension of the width of the space charge region of the PN junction and the enhancement of the electric field, as shown in Figure 6a,b. This greatly promotes the drift motion of carriers and increases the off-state current, as shown in Figure 7a. Subsequently, Figure 7b shows that when the gate bias voltage rises to 0.2 V, the substrate with a higher doping concentration will more likely lead to valence-band bending at the interface of the PN junction, which increases the BTBT current. Finally, if the VG continues to increase, the on-state current of the point-HTFETs with different substrate doping concentrations tends to saturate gradually. This is because the high voltage at the gate induces a large number of electrons underneath the gate oxide, which offset the benefits of the substrate with a high doping concentration.

Effect of the Doping Concentration of the Substrate
Increasing the doping concentration of the substrate has undesired effects on the extension of the width of the space charge region of the PN junction and the enhancement of the electric field, as shown in Figure 6a,b. This greatly promotes the drift motion of carriers and increases the off-state current, as shown in Figure 7a. Subsequently, Figure 7b shows that when the gate bias voltage rises to 0.2 V, the substrate with a higher doping concentration will more likely lead to valence-band bending at the interface of the PN junction, which increases the BTBT current. Finally, if the V G continues to increase, the on-state current of the point-HTFETs with different substrate doping concentrations tends to saturate gradually. This is because the high voltage at the gate induces a large number of electrons underneath the gate oxide, which offset the benefits of the substrate with a high doping concentration.

Source Drain
Gate

Breakdown Voltage
When the voltage between drain and substrate rises to the critical voltage (i.e., breakdown voltage), a large number of carriers, driven by the transverse electric field, begin to participate in the impact ionization and increase the drain current. As can be seen from Figure 8, the critical voltage is around 2 V, with ID = 2 nA/um.

Source Drain
Gate

Breakdown Voltage
When the voltage between drain and substrate rises to the critical voltage (i.e., breakdown voltage), a large number of carriers, driven by the transverse electric field, begin to participate in the impact ionization and increase the drain current. As can be seen from Figure 8, the critical voltage is around 2 V, with ID = 2 nA/um.

Breakdown Voltage
When the voltage between drain and substrate rises to the critical voltage (i.e., breakdown voltage), a large number of carriers, driven by the transverse electric field, begin to participate in the impact ionization and increase the drain current. As can be seen from Figure 8, the critical voltage is around 2 V, with I D = 2 nA/um.

Source Drain
Gate

Breakdown Voltage
When the voltage between drain and substrate rises to the critical voltage (i.e., breakdown voltage), a large number of carriers, driven by the transverse electric field, begin to participate in the impact ionization and increase the drain current. As can be seen from Figure 8, the critical voltage is around 2 V, with ID = 2 nA/um.

Forward Direction
In the whole-chip ESD protection network, TFETs and parasitic resistances form secondary discharge paths to assist the primary ESD discharge paths, resulting in a reduction in the clamp voltage [16]. However, the right-side boundary of the ESD window shows a decreasing trend. For example, the breakdown voltage of the thin oxide device, based on the FinFET and FDSOI, is around 2.5-3 V, which requires the clamp voltage of the ESD device to be lower than before. The HTFET is a promising candidate to help the primary ESD current path to discharge current quickly. This is because the trigger voltage and failure voltage are 1.09 and 1.71 V, respectively; see Figure 9. As a comparison, the trigger voltage and failure voltage of the silicon counterpart are around 6 and 8 V, respectively [13].

Forward Direction
In the whole-chip ESD protection network, TFETs and parasitic resistances form secondary discharge paths to assist the primary ESD discharge paths, resulting in a reduction in the clamp voltage [16]. However, the right-side boundary of the ESD window shows a decreasing trend. For example, the breakdown voltage of the thin oxide device, based on the FinFET and FDSOI, is around 2.5-3 V, which requires the clamp voltage of the ESD device to be lower than before. The HTFET is a promising candidate to help the primary ESD current path to discharge current quickly. This is because the trigger voltage and failure voltage are 1.09 and 1.71 V, respectively; see Figure 9. As a comparison, the trigger voltage and failure voltage of the silicon counterpart are around 6 and 8 V, respectively [13]. The gate of the HTFET is connected to its own drain as input, and the positive TLP current elevates the gate potential, which increases band bending between the source region and the channel. This increases the BTBT generation rate and, in turn, increases the current through the device. This BTBT current is not sufficient to turn the device on, but it acts as an initial current, which accelerates the occurrence of impact ionization. This, in turn, rapidly enhances the discharge robustness. Figure 10a,b show the BTBT generation rate and impact ionization rate of the HTFET with a 1 × 10 −4 A/um pulse, and the corresponding trigger voltage of the device is 1.09 V. As shown in Figure  11a,b, the increase in the BTBT generation rate with a large current-pulse results in an increase of the impact-ionization rate. The impact ionization of the carriers generates a large amount of heat, which will lead to the rapid increase of the lattice temperature of the HTFET. Figure 12 shows that the failure location appears on the source side. Lattice temperature, according to [19,20], is usually used as the criterion to determine device failure in simulations. The critical failure temperature is usually defined as a value that is hundreds of degrees Kelvin lower than the melting temperature. The melting temperature of materials used in this paper was selected according to [21,22]. The gate of the HTFET is connected to its own drain as input, and the positive TLP current elevates the gate potential, which increases band bending between the source region and the channel. This increases the BTBT generation rate and, in turn, increases the current through the device. This BTBT current is not sufficient to turn the device on, but it acts as an initial current, which accelerates the occurrence of impact ionization. This, in turn, rapidly enhances the discharge robustness. Figure 10a,b show the BTBT generation rate and impact ionization rate of the HTFET with a 1 × 10 −4 A/um pulse, and the corresponding trigger voltage of the device is 1.09 V. As shown in Figure 11a,b, the increase in the BTBT generation rate with a large current-pulse results in an increase of the impact-ionization rate. The impact ionization of the carriers generates a large amount of heat, which will lead to the rapid increase of the lattice temperature of the HTFET. Figure 12 shows that the failure location appears on the source side. Lattice temperature, according to [19,20], is usually used as the criterion to determine device failure in simulations. The critical failure temperature is usually defined as a value that is hundreds of degrees Kelvin lower than the melting temperature. The melting temperature of materials used in this paper was selected according to [21,22].     Figure 13 shows that when the source electrode serves as the current-pulse input, and the gate and drain electrodes are connected to a ground. HTFET operates like poly-bounded diodes and can conduct a high current. The device failure location, under TLP stress, is in the source region near the edge of the gate, as shown in Figure 14. In Figure 15, the current-density distribution plots for different current-pulses are (a) 1 × 10 −4 A/um, (b) 5 × 10 −4 A/um, (c) 1 × 10 −3 A/um, and (d) 2.5 × 10 −3 A/um. The high current-densities at the failure location can be clearly seen in Figure 15d. In addition, it can also be seen from Figure 16 that the electrostatic potential at the failure location is higher than in any other regions in the device. Therefore, we can conclude that occurring failures are mainly caused by high current and voltage.   Figure 13 shows that when the source electrode serves as the current-pulse input, and the gate and drain electrodes are connected to a ground. HTFET operates like poly-bounded diodes and can conduct a high current. The device failure location, under TLP stress, is in the source region near the edge of the gate, as shown in Figure 14. In Figure 15, the current-density distribution plots for different current-pulses are (a) 1 × 10 −4 A/um, (b) 5 × 10 −4 A/um, (c) 1 × 10 −3 A/um, and (d) 2.5 × 10 −3 A/um. The high current-densities at the failure location can be clearly seen in Figure 15d. In addition, it can also be seen from Figure 16 that the electrostatic potential at the failure location is higher than in any other regions in the device. Therefore, we can conclude that occurring failures are mainly caused by high current and voltage.  Figure 13 shows that when the source electrode serves as the current-pulse input, and the gate and drain electrodes are connected to a ground. HTFET operates like poly-bounded diodes and can conduct a high current. The device failure location, under TLP stress, is in the source region near the edge of the gate, as shown in Figure 14. In Figure 15, the current-density distribution plots for different current-pulses are (a) 1 × 10 −4 A/um, (b) 5 × 10 −4 A/um, (c) 1 × 10 −3 A/um, and (d) 2.5 × 10 −3 A/um. The high current-densities at the failure location can be clearly seen in Figure 15d. In addition, it can also be seen from Figure 16 that the electrostatic potential at the failure location is higher than in any other regions in the device. Therefore, we can conclude that occurring failures are mainly caused by high current and voltage.   Figure 13 shows that when the source electrode serves as the current-pulse input, and the gate and drain electrodes are connected to a ground. HTFET operates like poly-bounded diodes and can conduct a high current. The device failure location, under TLP stress, is in the source region near the edge of the gate, as shown in Figure 14. In Figure 15, the current-density distribution plots for different current-pulses are (a) 1 × 10 −4 A/um, (b) 5 × 10 −4 A/um, (c) 1 × 10 −3 A/um, and (d) 2.5 × 10 −3 A/um. The high current-densities at the failure location can be clearly seen in Figure 15d. In addition, it can also be seen from Figure 16 that the electrostatic potential at the failure location is higher than in any other regions in the device. Therefore, we can conclude that occurring failures are mainly caused by high current and voltage.

The Quasi-Static Characteristics of the HTFET under VFTLP Stress
Very fast transmission line pulse (VFTLP) simulation was carried out to investigate the device behavior under a charged-device model (CDM) event. Figure 17a shows the quasi-static curve, plotted by computing the average values between 1-2 ns for each positive VFTLP with a 200 ps rise time and a 2 ns pulse-width. As expected, the trend of the I-V curve was the same as for the TLP simulation. However, the failure current of the HTFET was much higher under VFTLP stress. This is mainly because an increasing pulse-width increases the lattice temperature quickly, which degrades the electrical properties. This explanation can be verified with Figure 17b, which shows the VFTLP curves with different rise times and pulse widths. It was found that pulse width has a very significant impact on the device temperature.

The Quasi-Static Characteristics of the HTFET under VFTLP Stress
Very fast transmission line pulse (VFTLP) simulation was carried out to investigate the device behavior under a charged-device model (CDM) event. Figure 17a shows the quasi-static curve, plotted by computing the average values between 1-2 ns for each positive VFTLP with a 200 ps rise time and a 2 ns pulse-width. As expected, the trend of the I-V curve was the same as for the TLP simulation. However, the failure current of the HTFET was much higher under VFTLP stress. This is mainly because an increasing pulse-width increases the lattice temperature quickly, which degrades the electrical properties. This explanation can be verified with Figure 17b, which shows the VFTLP curves with different rise times and pulse widths. It was found that pulse width has a very significant impact on the device temperature.

The Quasi-Static Characteristics of the HTFET under VFTLP Stress
Very fast transmission line pulse (VFTLP) simulation was carried out to investigate the device behavior under a charged-device model (CDM) event. Figure 17a shows the quasi-static curve, plotted by computing the average values between 1-2 ns for each positive VFTLP with a 200 ps rise time and a 2 ns pulse-width. As expected, the trend of the I-V curve was the same as for the TLP simulation. However, the failure current of the HTFET was much higher under VFTLP stress. This is mainly because an increasing pulse-width increases the lattice temperature quickly, which degrades the electrical properties. This explanation can be verified with Figure 17b, which shows the VFTLP curves with different rise times and pulse widths. It was found that pulse width has a very significant impact on the device temperature.

The Quasi-Static Characteristics of the HTFET under VFTLP Stress
Very fast transmission line pulse (VFTLP) simulation was carried out to investigate the device behavior under a charged-device model (CDM) event. Figure 17a shows the quasi-static curve, plotted by computing the average values between 1-2 ns for each positive VFTLP with a 200 ps rise time and a 2 ns pulse-width. As expected, the trend of the I-V curve was the same as for the TLP simulation. However, the failure current of the HTFET was much higher under VFTLP stress. This is mainly because an increasing pulse-width increases the lattice temperature quickly, which degrades the electrical properties. This explanation can be verified with Figure 17b, which shows the VFTLP curves with different rise times and pulse widths. It was found that pulse width has a very significant impact on the device temperature.

Conclusions
In this paper, the ESD characteristics of an InGaAs/InP HTFET was investigated. It was found that the InGaAs/InP HTFET has a low trigger voltage during positive TLP simulation, which makes it suitable to be used in low-voltage ICs. This became possible mainly due to the BTBT-assisted avalanche generation mechanism and its better material properties compared to its silicon counterpart. In a negative TLP simulation, the InGaAs/InP HTFET showed a similar characteristic as a poly-bounded diode. The results indicate that the InGaAs/InP HTFET can be used for the ESD protection of future III-V technology.