Abstract-We present a new model for the the kink effect in InAlAs/InGaAs HEMT's. The model suggests that the kink is due to a threshold voltage shift which arises from a hole pile-up in the extrinsic source and an ensuing charging of the surface and/or the buffer-substrate interface. The model captures the many of the observed behaviors of the kink, including the kink's dependence on bias, time, temperature, illumination, and device structure. Using the model, we have developed a simple equivalent circuit, which reproduced well the kink's dc characteristics, its time evolution in the nanosecond range, and its dependence on illumination.

No abstract

We report what we believe is the first InAs modulationdoped field-effect transistor (MODFET) using an epitaxial heterostructure based entirely on arsenides. The heterostructure was grown by MBE on InP and contains a 30 A InAs channel. An Lc=2 pm device displays well behaved characteristics, showing sharp pinch-off (Vt,=-0.8 V) and small output conductance (5 mS/mm) at 300 K. The maximum tansconductance is 170 mS/mm with a maximum drain current of 312 mA/mm. Strong channel quantiration results in an unprecedented breakdown voltage of -9.6 V, a several-fold improvement over previous InAs MODFETs based on antimonides. Lowtemperature magnetic field measurements show strong Shubnikov-de Haaa oscillations from which the electron channel is con6rmed to be the InAs layer.I

No abstract

Increasing the I d s mole fraction in InGaAs-base heterostructure FET's W E T ' S ) leads to improved device performance due to the superior carrier transport properties of these materials. At the same time, however, the use of narrow band-gap semiconductors results in enhanced impactionization, with severe detrimental effects like excessive shot noise in ID and large gate current IG even at regular bias points.Detailed physical understanding of impact-ionization and of the behaviour of the copious amount of holes that are generated in InGaAs channels is crucial to developing guidelines for designing high-performance devices. Gate current measurements 'and electroluminescence spectra have been widely adopted to evaluate hot-electron effects and impact-ionization in GaAs-based MESFET's and HEMT's, but no agreement has been found as of the origin of the different spectral components of the emitted radiation. In any case, no work has been presented, up to now, in InGaAsbased WET'S. In this paper we cam' out a detailed study of gate current and its correlation w i t h the various spectral components of light emitted in InAlAs/InGaAs HFET's at regular bias points.Our work reveals that light emitted in the visible portion of the spectrum is a good signature of impact-ionization in the channel as impact-ionized holes recombine with channel electrons. On the other hand. light emitted in the infrared portion of the spectrum is found to originate in conduction band-to-conduction band transitions of the hot electrons in the channel. These findings establish electroluminescence in the appropriate spectral range as an ideal tool to characterize hot carrier phenomena in InP-based HFET's, and allowed us for the first time to quuntitavely separate the gate current into its electron and hole components.The devices characterized in ths work are n-channel normally-on L = 1 pm InAIAs/lnGaAs HFET's, with an 100 A n+ -In0 ;jGaO 47As Si-doped channel (Nsi = 8 x lo1* ~m -~) , a 300In0.41Alo. j9As strained insulator'and a '50 A In0 j3Ga0.47As cap layer. When these devices are biased at high Vds (23 V), sigmficant impact-ionizkon takes place in the channel. A detailed study of the gate current reveals that, for negative Vgs, IG is dominated by collection of impact-ionized holes, while for positive Vgs, IG is dominated by electron real-space-transfer at low V h , and by hole collection at high Vds. Light emission both in the infrared and visible region takes place at high Vds. The intensity of the emitted light increases remarkablv by increasing Vds or by decreasing T. For energies higher than 1.5 eV all spectra exhibit nearly hfa?rwellian distributions. with effective temperatures Teff in the 1170 K -1360 K range. Teff increases with increasing Vds and decreasing T. The intensity of the integrated light in the 1.1 eV -1.25 eV range is found to be proportional to ID, thus suggesting conduction band-to-conduction band transitions as the dominant light emission mechanism in the infrared range. This mechanism, although predicted by simulation...

No abstract

InAlAs/InGaAs metamorphic High Electron Mobility Transistors (HEMT) hold promise for power-millimeter wave applications. A major reliability concern in some of these devices is the degradation of the drain resistance that is observed when the device is electrically stressed for a long time at bias conditions necessary for power applications. The goal of this thesis was to find the physical origin of this reliability problem and to suggest solutions to it. State-of-the-art InAlAs/InGaAs metamorphic HEMTs, provided by our sponsor, Hewlett Packard, were stressed under different bias schemes. It was found that most figures of merit associated with the drain-side of the device degrade under severe bias stress. In particular, the drain resistance, RD, has been found to increase significantly. In order to understand the physical origin of this degradation, we have studied the degradation of simpler Transmission Line Model (TLM) structures. We have found that in TLMs and HEMTs there appear to be two different degradation modes, both associated with hot electrons. In the first degradation mechanisms, we postulate that hot electrons are trapped by defects at the interface between the GaAs etch-stopper and the AlInAs Schottky barrier layer, depleting the carrier concentration in the channel underneath. In the second mechanism hot electrons degrade the InGaAs ohmic contacts. No degradation mechanism associated with the metamorphic nature of the structure has been identified. AcknowledgementsI would like to thank Prof. Jesus del Alamo for giving me the chance to work on this interesting project. He has guided me through this research with a huge amount of patience, ideas and red ink, teaching me a methodology in research and reporting. He always tried to make time available for me, even if he was extremely busy. I would like to thank him for offering me the opportunity to be a teaching assistant for his very interesting class and for giving me a place at MIT from the time I arrived. This research has been funded by Hewlett-Packard. I would like to thank Don D'Avanzo for starting this project, appropriating the funding and giving me the opportunity to work as a SEED student in Santa Rosa. Larry Studebaker, thanks for the invaluable help and the samples during the year and the supervision this Summer. I am very happy that I have found such a great co-advisor. I had a great time in Santa Rosa and I learned a lot there, also thanks to Dan Scherrer, Fred Sughiwo and Bob Yeats. HP Labs also helped me a lot through this thesis. Hans Rohdin, thank you for answering so many questions, reading this thesis on time and sending me the samples that helped pull things together. Thanks to Arlene Wakita and Nick Moll for making these devices available to me.Roxann Blanchard, thanks for all the discussions we held and the questions you answered, even when you were busy with your own thesis. You have proven to be an infinite well of knowledge and I can only hope I have been able to soak up enough of it to continue this research.Tassan...

We have investigated the hydrogen sensitivity of InP HEMTs with a gate stack containing a thick Ti-layer (order of 1000 A). We have found that the hydrogen-induced piezoelectric effect in these devices is an order of magnitude smaller than conventional Ti/Pt/Au-gate HEMTs with a thin Ti layer (order of 250 A). This markedly improved reliability can be explained through the diffusion mechanism of H in Ti which limits hydrogenation of the Ti layer to a thin sheet at the top. Using Auger Electron Spechoscopy, we have confumed that under the studied conditions, TiH, is only formed in the top 250 A of the Ti-layer. In some devices located in the periphery of the wafer, we have observed a second hydrogen degradation mechanism that induces a large positive AV,. This appears to be related to an improperly fabricated gate recess. The use of a thick Ti layer in the gate stack allows for a simple and effective mitigation of the H-induced piezoelectric effect in InP and other 111-V HEMTs. 2