Breakdown Analysis of a Body-Contacted Submicron High Electron Mobility Transistor

 

Introduction

Interest continues to grow in the development of high electron mobility transistor (HEMT) technologies for micrometer and millimeter wave power applications. A primary concern of device designers working with such technologies is the breakdown behavior in both the on- and off-states. As is the case for most field-effect transistors, reducing device dimensions results in a larger internal electric field near the drain-end of the device?s channel. The presence of such a field within the device can affect many areas of device performance including the breakdown characteristics.

In recent years, several studies [1-3] have examined the on-state and off-state breakdown mechanisms in submicron HEMT technologies. Both experimental and theoretical analysis has shown that a primary mechanism affecting both on-state and off-state breakdown is the accumulation in the channel, buffer, and supply layers of excess holes generated by impact ionization. The accumulated holes act as a positive fixed charge, lowering the energy barrier at the source/channel junction and enhancing the injection of electrons into the channel [2]. This phenomenon is often referred to as the parasitic bipolar effect (PBE).

The additional electrons injected into the channel region due to PBE are accelerated towards the drain further contributing to impact ionization, excess hole generation, and enhanced injection. This positive feedback mechanism ultimately leads to device breakdown and burn-out. In some devices, it has also been shown to cause a kink in the device?s output characteristics similar to that seen in silicon-on-insulator (SOI) technologies. The breakdown voltage of the HEMT is reduced as a consequence of the enhanced injection process limiting its power output [2]. Continued improvement of HEMT performance for power applications requires a better understanding of breakdown behavior.

Several approaches have been proposed for improving the breakdown characteristics of power HEMTs including the use of composite channel layers [5], quantum confinement [6], and an additional contact on the device body [4] to suppress impact ionization. This article follows the work of Sleiman et al [4] and examines the use of a body contact to improve the breakdown performance of a 250 nm psuedomorphic high electron mobility transistor. Further, this article will illustrate how Blaze, Silvaco's 2D device simulator for advanced materials, can be used to analyze the internal mechanisms of a common device technology. The analysis methods shown are applicable to a wide range of common device technologies.

Simulated Device

A schematic diagram of the simulated body-contacted HEMT structure is shown in Figure 1. The primary dimensions and doping values used for this study are shown. The dashed line extending across the bottom of the AlGaAs Supply layer represents a region of delta-doping. The delta-doping lies 2 nm from the Supply/Channel interface and has a peak density of 1x10 cm. Highly n-doped regions (peak concentration = 5x10 cm) also exist directly beneath the source and drain electrodes, but are not shown. These regions extend to the top of the InGaAs Channel layer.


Figure 1. Schematic cross section of the simulated
submicron psuedomorphic body-contacted HEMT

 

For comparison, a second HEMT structure without a body contact was also simulated. This structure, referred to as the base HEMT, is identical to the body-contacted structure shown in Figure 1 with the following exceptions: 1) the body contact was removed, and 2) the GaAs substrate is lowly doped (NA = 1x10 cm). The breakdown characteristics of both devices were simulated using the standard drift-diffusion approach. Appropriate models for Shockley-Read-Hall, Auger, and direct recombination were included. Carrier generation resulting from impact ionization was accounted for using the Selberherr impact ionization model. A full description of each device model is available within the ATLAS User's Manual [7].

It should be noted that the simulation approach used in this study represents a simplified model set. As applied, the drift-diffusion solution includes the following assumptions: 1) carrier velocity and mobility are functions of the local electric field, and 2) carrier temperature is equal to the lattice temperature which was 300 K for this study. As such, non-local transport effects such as velocity-overshoot were not accounted for.

A more advanced approach including non-local affects and carrier energy variations would be more accurate, however as shown by [8], a drift-diffusion approach neglecting non-local transport effects is adequate for the qualitative analysis of HEMT breakdown resulting from impact ionization. For an in-depth discussion of the advanced transport models available in Blaze please refer to the article "2D Simulation of Psuedomorphic Heterojunction Devices Using the Fully-Coupled Carrier Energy Balance Model" as published in the April 1995 Simulation Standard newsletter.

It has been proposed by some that off-state breakdown is dominated by a combination of thermionic emission and tunneling [9]. Investigation of this idea was determined to be beyond the scope of this work and will be left to future discussion.

 

Results and Discussion

Figure 2 presents the breakdown characteristics of both the base and body-contacted HEMT structures. The on-state and off-state breakdown curves were extracted at VG = 0.8V and VG = -0.4V, respectively. The body-contacted HEMT shows an improvement in breakdown performance of nearly 2V. This figure agrees well with the results published in [4]. The on-state breakdown simulation does highlight one potential limitation of the body-contact approach. The saturation current of the body-contacted device is substantial lower than that of the conventional HEMT. At VD = 5V, the drain current in the body-contacted device is 60 uA/um less than that of the base HEMT. This is a direct result of the holes being removed from the active device through the body contact.


Figure 2. Simulated HEMT breakdown characteristics for both the
Base and Body-Contacted HEMT structures.
On-state simulations conducted at VG = -0.4 V;
Off-state simulations conducted at VG = 0.8 V.

 

Figure 3 shows the contours of hole concentration across the supply, channel, and buffer regions of the base and body-contacted devices under on-state breakdown conditions. As can be seen, the concentration of holes near the source-end of the channel in the back-contacted device is substantially lower than in the base HEMT. This is more clearly seen in Figure 4 which shows the hole concentration across the lateral width of the channel layer 5nm beneath the supply/channel interface. The hole concentration at the source-end of the channel has been reduced by an order of magnitude. The difference at the middle of the channel is even more dramatic.


Figure 3. 2D device cross section taken from base and body-contacted
HEMT structures with overlain contours of hole concentration (cm-3).

 


Figure 4. Hole concentration across device channel at on-state breakdown

 

 

The off-state breakdown characteristics of the body-contacted device do not show as significant an improvement as the on-state characteristics. Again for the body-contacted device, the drain current prior to breakdown has been reduced in comparison to that of the conventional device, however the breakdown voltage itself is only improved by about 1V. This indicates that off-state breakdown is not as dependent on hole accumulation as on-state breakdown. It may also be possible that under off-state conditions a portion of the accumulated holes are removed through the gate electrode thereby reducing the impact of the back contact.

Figure 5 shows the hole current density vectors across the supply, channel, and buffer regions of the body-contacted device under both on- and off-state breakdown. As can be seen, the is a considerable amount of hole current flowing through the gate electrode under the off-state breakdown conditions. This supports the idea that the presence of the body contact has less of an effect on off-state breakdown behavior. Figure 6 shows the subthreshold characteristics of both the base and body-contacted HEMT structures. Extracted at VD = 2V, the subthreshold characteristics do not indicate any other significant design trade-offs resulting from the presence of the body contact. Both the drain current and conductance values are shifted to a slightly higher gate voltage.


Figure 5. Hole current density vectors for the body-contacted HEMT under
a) on-state breakdown (VG = 0.8V, VD = 7V)
b) off-state breakdown (VG = -0.4V, VD = 15V).

 


Figure 6. Subthreshold characteristics for base and body-Contacted
HEMT Structures. Characteristics extracted at VD = 2V.

 

Summary

An important component of HEMT design for power applications is the breakdown behavior of the device in both the on- and off-states. Similar to other field effect devices, the breakdown behavior of HEMTs is adversely affected by the large electric fields present within submicron structures. To further improve HEMT performance, a better understanding of the on-state and off-state breakdown mechanisms and methods for controlling these mechanisms is needed. Device simulation provides an excellent tool for analyzing these mechanisms and developing new device structures for minimizing their effects.

This article examined the effect of hole accumulation and the parasitic bipolar effect on the on-state and off-state breakdown behavior of a 250 nm psuedomorphic body-contacted HEMT structure through the use of a standard drift-diffusion approach. To obtain a better understanding of these mechanisms, the breakdown behavior of the body-contacted device was compared to that of a conventional HEMT structure of similar build. The presence of the body contact was shown to increase the breakdown voltage of the device under both on- and off-state conditions at the expense of a lower drain saturation current.

 

References

  1. M. Somerville et al., "A New Gate Current Extraction Technique for Measurement of On-State Breakdown Voltage in HEMT's," IEEE Electron Device Letters, vol. 19, pp.405-407, November 1998.
  2. A. Di Carlo et al., "Monte Carlo Study of the Dynamic Breakdown Effects in HEMTs," IEEE Electron Device Letters, vol. 21, pp. 149-151, April 2000.
  3. R. Shigekawa et al., "Electroluminescence of InAlAs/InGaAs HEMTs Lattice-Matched to InP Substrates," IEEE Electron Device Letters, vol. 16, pp. 515-517, November 1995.
  4. A. Sleiman et al., "Breakdown Quenching in High Electron Mobility Transistor by Using Body Contact," IEEE Transactions on Electron Devices, vol. 48, pp. 2188-2191, October 2001.
  5. T. Enoki et al., "Design and Characteristics of InGaAs/InP Composite-Channel HFETs," IEEE Transactions on Electron Devices, vol. 42, pp. 1413-1418, August 1995
  6. C. R. Bolognesi et al., "Impact Ionization Suppression by Quantum Confinement: Effects on the DC and Microwave Performance of Narrow-Gap Channel InAs/AlSb HFETs," IEEE Transactions on Electron Devices, vol. 46, pp. 826-832, May 1999.
  7. ATLAS User's Manual, Silvaco International, November 1998.
  8. K. Eisenbeiser et al., "Theoretical Analysis of the Breakdown Voltage in Psuedomorphic HFETs," IEEE Transactions on Electron Devices, vol. 43, pp. 1778-1787, November 1996.
  9. M. Somerville et al., "Off-State Breakdown in Power pHEMTs: The Impact of the Source," IEEE Transactions on Electron Devices, vol. 45, pp. 1883-1889, September 1998.