P Performance Analysis of f A l i f Ultra-Scaled InAs HEMTs - PowerPoint PPT Presentation

P Performance Analysis of f A l i f Ultra-Scaled InAs HEMTs Neerav Kharche 1 , Gerhard Klimeck 1 , Dae-Hyun Kim 2,3 , Jesús. A. del Alamo 2 , and Mathieu , , Luisier 1 1 Network for Computational Nanotechnology and 1 N t k f C t ti l N t h l d Birck Nanotechnology Center, Purdue University 2 Microsystems Technology Labs, Massachusetts c osyste s ec o ogy abs, assac usetts Institute of Technology 3 Teledyne Scientific & Imaging, LLC

Motivation: Towards III-V MOSFET • Strained channel • Device geometries • New gate • Channel materials dielectrics dielectrics • High-k dielectrics High k dielectrics 2015-2019 Research III-V channel devices Acknowledgement: Robert Chau, Intel Low-power & high-speed

Motivation: Why III-V HEMTs? • III-V : Extraordinary electron transport properties III V E t di l t t t ti • HEMTs : Very similar structure to MOSFETs except high- κ dielectric layer y • Excellent to Test Performances of III-V material without interface defects • Excellent to Test Simulation Models • Excellent to Test Simulation Models – Develop simulation tools and benchmark with experiments – Predict performance of ultra-scaled devices P di t f f lt l d d i 2007: 40nm 2008: 30nm D.H. Kim et al., EDL 29 , 830 (2008)

Outline • Motivation M i i • Modeling Approach – Real-space EM simulator including gate leakage – Atomistic tight-binding m* – Realistic description of simulation domain (gate Realistic description of simulation domain (gate geometry) • Comparison to Experiments L g =30, 40, 50nm Comparison to Experiments L g 30, 40, 50nm – Material parameters, I d -V gs , I d -V ds • Scaling Considerations for L g =20nm g g – Channel thickness, Insulator thickness, Gate metal work function • HEMT Simulator on nanoHUB.org • Conclusion and Outlook

Outline • Motivation M i i • Modeling Approach – Real-space EM simulator including gate leakage – Atomistic tight-binding m* – Realistic description of simulation domain (gate Realistic description of simulation domain (gate geometry) • Comparison to Experiments L g =30, 40, 50nm Comparison to Experiments L g 30, 40, 50nm – Material parameters, I d -V gs , I d -V ds • Scaling Considerations for L g =20nm g g – Channel thickness, Insulator thickness, Gate metal work function • HEMT Simulator on nanoHUB.org • Conclusion and Outlook

Device Geometry and Simulation Domain • Intrinsic device I t i i d i Extrinsic device – Near gate contact L g Source Drain – Self consistent 2D Self consistent 2D n+ Cap n+ Cap Schrodinger-Poisson InP etch stop Gate δ -doped InAlAs – Electrons injected layer R R R R Source e from all contacts Drain D S InAs InGaAs M. Luisier et. al., IEEE Transactions on Electron Devices Transactions on Electron Devices, Si Simulation Domain: l ti D i vol. 55, p. 1494, (2008). Intrinsic device • Extrinsic source/drain InP Substrate contacts t t – Series resistances R S and R D R S and R D R. Venugopal et.al., Journal of Applied Physics, vol. 95, p. 292, (2004).

Gate Geometry and Gate Leakage Current 1 (a) 3 Flat Gate 2 (b) 1) Include series resistances          ext int ext int V V I R V V I R R g gs gs g d s ds ds ds ds d d s s d d 2) Include gate leakage current            S D G S D G ( E H ) C ( S S S ) Curved Gate 3) Incl de the proper gate 3) Include the proper gate Gate leakage reduced geometry flat (a) or curved (b) in curved gate device

Accurate Effective Mass Calculation Full Band Transport: Full-Band Transport: Effective Mass Transport: Effective Mass Transport: •Strain, Disorder, Non- •Gate leakage, parabolicity, BTBT Computationally efficient •No gate leakage, •Parabolic bands, No Import Computationally very disorder, Wrong m* intensive quantization levels ~ 4 nm

Outline • Motivation M i i • Modeling Approach – Real-space EM simulator including gate leakage – Atomistic tight-binding m* – Realistic description of simulation domain (gate Realistic description of simulation domain (gate geometry) • Comparison to Experiments L g =30, 40, 50nm Comparison to Experiments L g 30, 40, 50nm – Material parameters, I d -V gs , I d -V ds • Scaling Considerations for L g =20nm g g – Channel thickness, Insulator thickness, Gate metal work function • HEMT Simulator on nanoHUB.org • Conclusion and Outlook

Transfer Characteristics: I d -V gs Parameter Parameter Initial Initial Final parameter set Final parameter set 30 40 50 L g [nm] L g [nm] 30, 40, 50 30, 40, 50 34.0 34.0 42.0 42.0 51.25 51.25 t ins [nm] 4 3.6 3.8 4.0 Φ M [eV] 4.7 4.66 4.69 4.68 m* ins (InAlAs) 0.075 0.0783 0.0783 0.0783 m* buf (InGaAs) 0.041 0.0430 0.0430 0.0430

Output Characteristics: I d -V ds Conclusion: • Good agreement for all L g ’s g • Less ballistic at L g =50nm • Use models and material parameters to design ultra- parameters to design ultra scaled devices (L g =20nm)

Outline • Motivation M i i • Modeling Approach – Real-space EM simulator including gate leakage – Atomistic tight-binding m* – Realistic description of simulation domain (gate Realistic description of simulation domain (gate geometry) • Comparison to Experiments L g =30, 40, 50nm Comparison to Experiments L g 30, 40, 50nm – Material parameters, I d -V gs , I d -V ds • Scaling Considerations for L g =20nm g g – Channel thickness, Insulator thickness, Gate metal work function • HEMT Simulator on nanoHUB.org • Conclusion and Outlook

What can be changed? • Gate geometry • Channel thickness Better control of scaling: t InAs scaling: t InAs surface potential • Insulator thickness scaling: t ins Gate leakage reduction • Metal work function and E-mode operation engineering: Φ M L g =20nm L 20 Gate Φ Φ M In 0.52 Al 0.48 As t ins urce In 0.53 Ga 0.47 As rain t t InAs InAs InAs Sou Dr

InAs (Channel) Layer Thickness InAs Channel Scaling: • Better electrostatic control – lower SS – larger I ON /I OFF ratio I I OFF • Increase of transport m* Increase of transport m* increases – reduced v inj , higher N inv => higher I ON g ON • Increase of gate leakage current Gate Gate – I ON /I OFF ratio saturates urce In 0.53 Ga 0.47 As rain t t InAs InAs InAs Sou Dr In 0.52 Al 0.48 As

InAlAs (Insulator) Layer Thickness InAlAs Insulator Scaling: • Better electrostatic control • Better electrostatic control (due to larger C ox ) • Increase of gate leakage g g gate current leakage – larger I OFF – larger SS l SS – smaller I ON /I OFF ratio Gate Gate t ins urce In 0.53 Ga 0.47 As rain InAs InAs Sou Dr In 0.52 Al 0.48 As

Work Function Engineering Work Function Increase: • Shift towards enhancement mode • Decrease of gate leakage current • Allows for thinner insulator layer – steeper SS – larger I ON /I OFF ratio

Parameters and Performances Summary (1) Gate (1) Gate (2) Channel (2) Channel (3) Insulator (4) Metal work (3) Insulator (4) Metal work geometry thickness thickness function Improved Higher gate Gate leakage gate control leakage reduction SS SS I I ON /I OFF /I 1 3 4 4 L g =20nm 2 1 1 3 3 2 2

Outline • Motivation M i i • Modeling Approach – Real-space EM simulator including gate leakage – Atomistic tight-binding m* – Realistic description of simulation domain (gate Realistic description of simulation domain (gate geometry) • Comparison to Experiments L g =30, 40, 50nm Comparison to Experiments L g 30, 40, 50nm – Material parameters, I d -V gs , I d -V ds • Scaling Considerations for L g =20nm g g – Channel thickness, Insulator thickness, Gate metal work function • HEMT Simulator on nanoHUB.org • Conclusion and Outlook

HEMT Simulator on nanoHUB.org OMEN_FET: • 2-D Schrödinger-Poisson solver • Real-space effective mass quantum transport model t t t d l • Injection (white arrows) from Source, Drain, and Gate contacts • HEMTs, Single- and Double-Gate HEMTs Single and Double Gate devices • Electron transport in Si and III-V • Ballistic transport (no Scattering) • Ballistic transport (no Scattering) • Current Flow Visualization h http://nanoHUB.org/tools/omenhfet // HUB / l / hf Run your own simulations!

Conclusion and Outlook Extrinsic device Extrinsic device • Multiscale Modeling Approach – EM transport including gate l leakage k Simulation Domain: – m* from tight-binding Intrinsic device • Good Agreement with • Good Agreement with Experiments • Scaling Considerations for 20nm Device • HEMT Simulator Deployed on nanoHUB org nanoHUB.org L g =20nm • Challenges and Future Directions – S/D contacts, high-k insulator, , g , scattering, interface traps V d =0.50 V V d =0.05 V

Thank You!

Transfer Characteristics: I d -V gs (2) L g [nm] L [nm] SS [mV/dec] SS [mV/dec] DIBL DIBL I ON /I OFF I /I V V inj [cm/s] [cm/s] [mV/V] 30 Expt. 107 169 0.47×10 3 Sim. 105 145 0.61×10 3 3×10 7 40 Expt. 91 126 1.38×10 3 Sim. Sim. 89 89 99 99 1.86 10 1.86×10 3 3.11×10 7 3.11 10 50 Expt. 85 97 1.80×10 3 Sim. 89 91 1.85×10 3 3.18×10 7