VLS! DESIGN 1998, Vol. 8, Nos. (1-4), pp. 349-353 Reprints available directly from the publisher Photocopying permitted by license only (C) 1998 OPA (Overseas Publishers Association) N.V. Published by license under the Gordon and Breach Science Publishers imprint. Printed in India. A New HEMT Breakdown Model Incorporating Gate and Thermal Effects LUTFI ALBASHA*, CHRISTOPHER M. SNOWDEN and ROGER D. POLLARD Institute of Microwaves and Photonics, School of Electronic and Electrical Engineering, University of Leeds, Leeds, LS2 9JT, UK This paper presents a comprehensive physical model for the breakdown process in HEMTs. The model is integrated into in a fast quasi-two-dimensional HEMT physical simulator. The work is based on a full study of the complex interactions between the different breakdown mechanisms and the influence of design parameters. The model takes account of tunnelling effects in the region of the gate metallization, and of the thermal effects in the active channel under the gate region. Keywords." Gate tunneling, thermal modelling, breakdown, quasi-two-dimensional modelling 1. INTRODUCTION Microwave circuits such as power amplifiers operate under large-signal conditions. Their ability to perform efficiently is limited by the devices’ breakdown characteristics, which limit the transis- tor’s performance and power output. The accuracy of the large-signal design relies on the availability of suitable breakdown models. Popular break- down theories have not been adequate to indepen- dently explain the full picture of the breakdown process in HEMTs. The effects of the gate leakage and substrate conduction in HEMTs on this have not been simulated. Analytical models for ava- lanche breakdown, such as Frensley’s [1], were based on physical simulations of the active channel around the gate. The effects of the gate leakage and substrate conduction on this have not been included. This paper considers breakdown more comprehensively including the effects of reverse gate conduction, thermal fluctuations within the active region and substrate conduction [2]. 2. MODEL DESCRIPTION The Quasi-Two-Dimensional (Q2D) model used here is based on the earlier work of Snowden [3]. * Corresponding author. Tel." + 44 113 2332082, Fax: + 44 113 2332032, email: [email protected]. 349
Transcript
VLS! DESIGN1998, Vol. 8, Nos. (1-4), pp. 349-353Reprints available directly from the publisherPhotocopying permitted by license only
(C) 1998 OPA (Overseas Publishers Association) N.V.Published by license under
the Gordon and Breach SciencePublishers imprint.
Printed in India.
A New HEMT Breakdown Model IncorporatingGate and Thermal Effects
LUTFI ALBASHA*, CHRISTOPHER M. SNOWDEN and ROGER D. POLLARD
Institute of Microwaves and Photonics, School of Electronic and Electrical Engineering,University of Leeds, Leeds, LS2 9JT, UK
This paper presents a comprehensive physical model for the breakdown process inHEMTs. The model is integrated into in a fast quasi-two-dimensional HEMT physicalsimulator. The work is based on a full study of the complex interactions between thedifferent breakdown mechanisms and the influence of design parameters. The modeltakes account of tunnelling effects in the region of the gate metallization, and of thethermal effects in the active channel under the gate region.
Microwave circuits such as power amplifiersoperate under large-signal conditions. Their abilityto perform efficiently is limited by the devices’breakdown characteristics, which limit the transis-tor’s performance and power output. The accuracyof the large-signal design relies on the availabilityof suitable breakdown models. Popular break-down theories have not been adequate to indepen-dently explain the full picture of the breakdownprocess in HEMTs. The effects of the gate leakageand substrate conduction in HEMTs on this havenot been simulated. Analytical models for ava-
lanche breakdown, such as Frensley’s [1], werebased on physical simulations of the active channelaround the gate. The effects of the gate leakageand substrate conduction on this have not beenincluded. This paper considers breakdown morecomprehensively including the effects of reversegate conduction, thermal fluctuations within theactive region and substrate conduction [2].
2. MODEL DESCRIPTION
The Quasi-Two-Dimensional (Q2D) model usedhere is based on the earlier work of Snowden [3].
The Q2D method is based on the assumption thatthe fundamental driving force for electron trans-port is the x-directed electric field. The HEMTQ2D model important features have been reportedpreviously [4]. In contrast to MESFETs, spurioussubstrate current occurs in HEMT buffer layersdue to the lateral Ex field component. This currentis drawn around the depleted channel and reducesthe magnitude of the field. It is assumed in thispaper that, as a result of the reduction in theelectric field, an increase in the breakdown voltageis possible. Figure shows a schematic diagram ofa HEMT device.The computational interpretation of the break-
down model presented in this paper is based on theinteractions between the avalanche and gateleakage mechanisms. The relation that bothprocesses simultaneously have with the devicedesign parameters and power dissipation insidethe device is included. The flow chart of the modelis shown in Figure 2. The gate leakage, conven-tionally termed ’soft breakdown’, is assumed toalways occur in devices prior to avalanche. Theflow of electrons from the gate into the semicon-ductor would then influence the impact ionisationprocess. Adding the leaked electrons to thechannel electrons constituting the increasing draincurrent imposes this effect. The leakage is allowedin the three pinchoff stages.
Source Gate Drain
Electrons
FIGURE Schematic diagram of a HEMT device showingelectrons leaking form gate.
Thermal model calculates fluctuations in
Tunneling gate(confirmed by light surface, wilh bias)
FIGURE 2 Flow chart of the gate breakdown model.
3. THERMALLY DRIVEN GATE LEAKAGECURRENT
The gate leakage model is principally based on acombination of Padovani [5] and Rideout [6]equations. These equations involve complex func-tions of temperature, barrier height and semicon-ductor parameters. The form that these equationsare presented in the above references is quitecomplicated. In order to maintain the numericalefficiency of the physical device simulator, newapproximations are introduced in this paper,which simplify the tunneling functions and main-tain the accuracy of the solution. Tunnelingthrough the gate metal-semiconductor barrierbecomes significant in the reverse direction thanin the forward direction because the bias voltagesinvolved are usually greater. This cause thepotential barrier to become thin enough such thattunneling is dominant. The Thermionic (T),Thermionic-Field (T-F) and Field Emission (FE)gate leakage currents are dependent on the latticetemperature and material specifications. Whatdetermines the current mechanism is the tempera-ture of the channel and the applied bias. A newthermal model [7] incorporated into the Q2Dphysical model computes the instantaneous tem-perature of the active channel. The output of thismodel is linked with the breakdown model. Thisassists the dynamic update of the gate leakagecurrent mechanism.
A NEW HEMT BREAKDOWN MODEL 351
3.1. Thermionic-field Tunneling Current
This mechanism is the major form of tunneling.The tunneling current is defined by the equation
IT-F Ags exp(E/e’) (1)
Where A is the gate area, E is applied energycalculated from q Vr, Vr being the applied reversebias voltage given a positive sign throughout thiswork. q is the electron charge and e’ is an energyterm defined as:
eoo[eOO-1
tanh ( KTJ 1 (2)
The term Eoo quantifies the diffusion potentialfrom metal into semiconductor. It has two equi-valent definitions given in [5] and [6]. The latterdefinition is adopted in this work, however, slightnumeric alteration is needed for consistency ofunits:
Eoo 18.57 x 10-5 (N) 1/2(3)
mrEr
N is the doping density per m-3. In order to useequation (3) in (2), Eoo must be converted back bymultiplying by q, the electron charge. The type ofHEMT used connects the gate metalisationdirectly to the N-doped A1GaAs layer.
In equation (1), Js is defined as the saturationcurrent given in [5] as:
R(TrEoo) 1/2
KTqVr
csh2KTJ
1/2
(4)
where bb is the Schottky barrier height. Eo is aterm defined as:
( oo)Eo Eoo coth (5)
It was observed during the course of this workthat in the saturation current equation (4), thesquare root term was dominated by the first term.Hence a simpler approximate numerical expres-sion for the saturation current was deduced:
Js [yrEooqVr]]/2 exp ( -qb )eo (6)
which is integrated into the model. An empiricaldifference limit between equations (4) and (6) wasreached after some experimentation’s beyondwhich if this limit is exceeded, the solutionobtained from equation (6) was observed to affectthe numerical accuracy of the tunneling current.The model then switches to the more stringentexpression of equation (4). The thermionic andfield emission currents are calculated using similarequations. Threshold equations are incorporatedwhich, according to the thermal status of thedevice, the appropriate leakage mechanisms isinvoked.
4. SIMULATION RESULTS
Figure 3 shows the effect of varying the tem-perature on the tunnelling currents at various
70-"
3"10-"-
2qO"
0q)
N,202.3 m
N2e22
Temperalure (K)
FIGURE 3 Tunneling current versus channel temperature atvarious doping concentrations.
352 L. ALBASHA et al.
doping levels. An increase in the T-F current isobserved with an increase in the dopingconcentration. Figure 4 compares between thethermionic-field and field emission currents withrespect to temperature. The T-F current is clearlyof more prominent effect. The simulation of theDC I-V characteristics of a HEMT device enablinggate current tunnelling mechanism is shown inFigure 5. The soft breakdown mechanism isinfluenced by the design parameters and applied
7-
6-
0 2 5 4 5Reverse bias voltage (v)
FIGURE 4 Comparison between the thermionic-field andfield-emission leakage currents.
DC Charactersfics of a HEMT
0.08
0.06 -,
0.04-
0.02
2 3 4 5 6 7
Vos v
0.00 8 9 10
FIGURE 5 I-V characteristics of a HEMT device incorporat-ing the gate model.
bias. It can be practically and numericallyobserved when the stimulating conditions aremet. Comparisons with measured data in [8] showsgood agreement between simulated and measuredresults of a MESFET case.
5. CONCLUSION
The Q2D physical device simulation ofHEMTs hasbeen expanded to include soft thermal breakdowneffects using tunnelling current mechanisms. Theinfluence of soft breakdown mechanism on the DCcharacteristics is influenced by the design para-meters and applied bias. The thermal effects in theactive channel under the gate region are included.This is conducted using an incorporated thermalmodel, which calculates the channel temperatureand updates the thermal breakdown model.
References
[1] Frensley, W. R. (1981). "Power-limiting breakdowneffects in MESFETs, IEEE Trans., ED-28(8), 962-970.
[2] Morton, C. G., Atherton, J. S., Snowden, C. M., Pollard,R. D. and Howes, M. J. "A large-signal physical HEMTmodel ," Int. Microwave Symp. MTT-S, 3, 1759-1762,San Francisco.
[3] Snowden, C. M. and Pantoja, R. R. (1989). "Quasi-two-dimensional MESFET simulation for CAD", IEEETrans., ED-36(9), 1564-1574.
[4] Morton, C. G. and Snowden, C. M. "HEMT physicalmodel for MMMIC CAD", Proceedings of 25th EuropeanMicrowave Conference, Sept. 1995.
[5] Padovani, F. A. and Stratton, R. (1966). "Field andthermionic-field emission in schottky barriers", Solid-State Electronics, 9, 695-707.
[6] Crowell, C. R. and Ridout, V. L. (1969). "Normalisedthermionic-field (T-F) emission in metal-semiconductor(schottky) barriers", Solid-State Electronics, 12, 89-105.
[7] Johnson, R. G., Snowden, C. M. and Pollard, R. D.(1997). "A Physics Based Electro-Thermal Model ForMicrowave And Millimetre Wave HEMTs", Internationalmicrowave symposium, MTT-S, Denver, USA.
[8] Albasha, L., Snowden, C. M. and Pollard, R. D. "Break-down Characterization of HEMTs and MESFETs Basedon A New Thermally Driven Gate Model", InternationalConference on Simulation of Semiconductors processes andDevices SISPAD, 8-10 Sep. 1997, Cambridge, MA, USA(To be published).
Authors’ Biographies
Lutfi Albasha received his B.Eng. (Hons.) degree inElectronic and Electrical Engineering from the
A NEW HEMT BREAKDOWN MODEL 353
University of Leeds in 1990. In 1991 he obtainedan M.Sc. (with distinction) in RF communica-tions. In 1992, He joined the Microwave andTerahertz Technology Group at the University ofLeeds and obtained his Ph.D. in 1995. His thesiswas on the electromagnetic modelling of micro-wave circuits and its feasibility as a CAD tool. In1994 he was employed by Filtronics ComponentsLtd., UK. Currently he is a research fellow at theUniversity of Leeds conducting new studies on thehigh frequency breakdown modelling in HEMTsand its measurement techniques at W-bandfrequencies. His research interests include electro-magnetic numerical modelling and the design andmeasurements of microwave circuits and devices.
Christopher M. Snowden received the B.Sc.,M.Sc. and Ph.D. degrees from the University ofLeeds. After graduating in 1977 he worked as an
Applications Engineer for Mullard, Mitcham. HisPh.D. studies were conducted in association withRacal-MESL and were concerned with the large-signal characterisation of MESFET microwaveoscillators. In 1982 he was appointed Lecturer inthe Department of Electronics at the University ofYork. He joined the Microwave Solid State Groupin the Department of Electrical and Electronic atthe University of Leeds in 1983. He now holds theChair of Microwave Engineering in the Micro-wave and Terahertz Technology Research Groupand is also currently Head of the Department ofElectronic and Electrical Engineering. During1987 he was a Visiting Research Associate at theCalifornia Institute of Technology. He has been aConsultant to M/A-COM Inc., Corporate Re-search and Development since 1989, where he wason sabbatical leave during the period 1990-91.During this year he represented M/A-COM asSenior Staff Scientist. He was Chairman of the1995 international Microwaves and RF Confer-ence. He is a Member of the MIT Electromag-netics Academy. He is also a Top Scientist at theInternational Research Centre for Telecommuni-
cations-Transmission and Radar, Delft Universityof Technology, Netherlands. Professor Snowden isa Fellow of the IEEE and a Fellow of the IEE. Heis a Distinguished Lecturer (1996/7) for the IEEE(Electron Devices Society). He is co-Chairman ofthe MTT-1 Committee and a Member of the 1997IEEE MTT-S Technical Program Committee. Hismain research interests include compound semi-conductor device modelling, microwave, terahertzand optical nonlinear subsystem design andadvanced semiconductor devices. He has written7 books and over 190 papers.Roger D. Pollard was born in London, England
in 1946. He received his technical education,graduating with the degrees of BSc and PhD inElectrical and Electronic Engineering, at theUniversity of Leeds, Leeds, UK.He holds the Hewlett-Packard Chair in High
Frequency Measurements in the School of Elec-tronic and Electrical Engineering at the Universityof Leeds where he has been a faculty member since1974. He is Deputy Director of the Institute ofMicrowaves and Photonics which has over 40active researchers, a strong graduate program andhas made contributions to microwave passive andactive device research. The activity has significantindustrial collaboration as well as a presence incontinuing eduction through its Microwave Sum-mer School. Professor Pollard’s personal researchinterests are in microwave network measurements,calibration and error correction, microwave andmillimetre-wave circuits and large-signal and non-linear characterization. He has been a consultantto the Hewlett-Packard Company, Santa Rosa,CA since 1981. He has published over 100technical articles and hold 3 patents.
Roger Pollard is a Chartered Engineer, a memberof the Institution of Electrical Engineers (UK) anda Fellow of the IEEE. He is 1998 President of theIEEE Microwave Theory and Techniques Societywhere he is serving his second term as an electedmember Administrative Committee.
