NEMO5 on Blue Waters - A Flexible Package for Nanoelectronics - - PowerPoint PPT Presentation

▶
nemo5 on blue waters a flexible package for
SMART_READER_LITE
LIVE PREVIEW

NEMO5 on Blue Waters - A Flexible Package for Nanoelectronics - - PowerPoint PPT Presentation

Dec 05, 2022 305 likes •509 views

Network for Computational Nanotechnology (NCN) NEMO5 on Blue Waters - A Flexible Package for Nanoelectronics Modeling Problems Jim Fonseca Network for Computational Nanotechnology PRAC - Accelerating Nano-scale Transistor Innovation PI:

slide-1
SLIDE 1

Network for Computational Nanotechnology (NCN)

NEMO5 on Blue Waters - A Flexible Package for Nanoelectronics Modeling Problems

Jim Fonseca Network for Computational Nanotechnology PRAC - Accelerating Nano-scale Transistor Innovation PI: Gerhard Klimeck Blue Waters Symposium May 2015

slide-2
SLIDE 2

2

NEMO5 - Bridging the Scales From Ab-Initio to Realistic Devices

Approach:

  • Ab-initio:
  • Bulk constituents
  • Small ideal superlattices
  • Map ab-initio to tight binding

(binaries and superlattices)

  • Current flow in ideal structures
  • Study devices perturbed by:
  • Large applied biases
  • Disorder
  • Phonons

Ab-Initio

Goal:

  • Device performance with realistic

extent, heterostructures, fields, etc. for new / unknown materials Problems:

  • Need ab-initio to explore new

material properties

  • Ab-initio cannot model non-

equilibrium.

  • TCAD uses quantum corrections

TCAD

slide-3
SLIDE 3

3

NEMO5 – A multiscale simulation tool for nanoelectronic modelling

  • Multiscale/multiphysics
  • Empirical tight binding
  • NEGF, DD, QTBM, EM
  • Electron core, k.p, mode space
  • Ohmic and Schottky contacts
  • Scattering optical and acoustic
  • Phonons
  • Strain models-VFF, Keating, Lazarenkova
  • External magnetic fields
  • Solves
  • Atomistic strain
  • Electronic band structures
  • Charge density
  • Potential
  • Current
  • 4-level MPI parallelization
  • bias, energy, momentum, space

30nm

slide-4
SLIDE 4

4

NEMO5 and nanoHUB

  • Distribution and Support Group on nanoHUB.org

» https://nanohub.org/groups/nemo5distribution » Source code, example, discussion forum, run NEMO5 on Purdue Resources

  • nanoHUB.org

» 330,000 annual users » 4,200 resources (video lectures, presentations, tutorials, etc.) » 330 simulation tools » Nanoelectronics, nanophotonics, materials science, molecular electronics, carbon-based systems, Microelectromechanical systems » 4,200 resources (video lectures, presentations, tutorials, etc.) » NEMO5 T

  • ols

 Quantum Dot Lab  Crystal Viewer  Bandstructure Lab  …

slide-5
SLIDE 5

Network for Computational Nanotechnology (NCN) Non-equilibrium Green's functions method: Non-trivial and disordered leads

Yu He, Yu Wang, Tillmann Kubis, Gerhard Klimeck

slide-6
SLIDE 6

6

Problem: assumption of periodic contacts in NEGF contradicts experiment

http://www.electroiq.com/articles/sst /2010/12/iedm-reflections_.html

semi-infinite periodic contacts. But in the real world… Roughness Common self- energy methods Sancho Rubio, transfer matrix Non-periodic geometries

How to solve non-periodic contacts?

Source Drain

  • S. Koenig et al, Appl. Phys. Lett,
  • Vol. 104, pp. 103106, 2014
  • Q. Liu, et al, IEDM p.229 2013

Random alloy Periodic assumption contradicts realistic contacts

slide-7
SLIDE 7

7

General Lead Method: problem & idea

Problem:

  • No exact solution for semi-infinite systems unless periodicity assumed
  • Approximate solution

 Physically correct  Numerically solvable for arbitrary contact structures Idea: extend complex absorbing potential (CAP) method

  • Non-periodic contact : Hamiltonian for explicit contact segments;
  • CAP serves as scattering : physical assumption of contacts;
  • Efficient, memory thin : converge within finite iterations;

CAP CAP

  • J. Driscoll et al, Phys.
  • Rev. B. Vol. 78, pp.

245118, 2008

slide-8
SLIDE 8

8

Example: SiGe random alloy

device length Si Si Si0.5Ge0.5 Example: 3x3nm Si0.5Ge0.5 nanowire in sp3d5s* tight binding Device length 20nm and 6nm Results averaged over 50 samples Si0.5Ge0.5 Si0.5Ge0.5 Si0.5Ge0.5 Justification: With same effective alloyed disorder in contacts, expected transmission has weak dependence of device length

Yu He, Yu Wang, Gerhard Klimeck, Tillmann Kubis, "Non-equilibrium Green's functions method: Non-trivial and disordered leads” Appl. Phys. Lett. 105, 213502 (2014)

slide-9
SLIDE 9

9

Example: SiGe random alloy

device length Si Si Si0.5Ge0.5 Si0.5Ge0.5 Si0.5Ge0.5 Si0.5Ge0.5 General lead approach works well for contacts with alloy randomness. Alloyed contact yield virtually device length independent transmission; DOS of contacts match device better  less reflections of electrons;

slide-10
SLIDE 10

10

device length 20nm Si Si Si0.5Ge0.5

Example: SiGe random alloy

1020 cm-3 1020 cm-3 Si0.5Ge0.5 Si0.5Ge0.5 Si0.5Ge0.5 gate length 8nm Non-trivial contacts critical in transport simulations 45% decrease in on-current.

slide-11
SLIDE 11

Network for Computational Nanotechnology (NCN) Bilayer Graphene: a Good candidate for Transistors?

Fan Chen, Hesameddin Ilatikhameneh, Rajib Rahman, Gerhard Klimeck

slide-12
SLIDE 12

12 12

Graphene Transistor

  • Graphene has a zero band gap
  • It has a good high ON current, but

it can’t be turned off

ON/OFF < 10

We need to achieve:

  • Large ON/OFF ratio needed in transistors (~105)
  • Small OFF current -> Low power consumption

http://www.jameshedberg.com/img/samples/ https://www.kth.se/en/ict/forskning/ickretsar/

slide-13
SLIDE 13

13 13

Bilayer Graphene

http://jarilloherrero.mit.edu/research/gated-bilayer-graphene/

2 4 6 8 10 50 100 150 200 250 300 Bilayer Graphene Band Gap with Dav Dav [V/nm] Band Gap [meV]

Bandstructure Small field Large field

Bilayer Graphene: Create a Band-Gap by Electric Field

  • Control band-gap by applying vertical electric field

Band Gap

Bandgap vs. E-Field

slide-14
SLIDE 14

14 14

NEMO5 – realistic atomic approach

Challenge:

  • Matrix size ~ 64 million
  • Inverse, Eigenvalues

http://chemwiki.ucdavis.edu/

. . . . . . . . . . . . . . . .

Daniel Mejia

① Limitation from fabrication technique, short channel effect, gate leakage … ② Device size is typically 100nm (thick) x 200nm (long) x 20nm (wide)

  • 3.2 million atoms in simulation

20x20

3.2 million 3.2 million

  • rbitals

Matrix

  • Tao. Chu, Prof. Zhihong Chen Purdue
slide-15
SLIDE 15

15 15

  • 1 -0.5 0 0.5 1 1.5 2 2.5 3

0.2 0.5 1 2 5 10 20 IdVg VTG(V) I( A/ m)

Bilayer Graphene: Open band gap

Band Gap opens through the change of Top Gate

VBG = -1.75 V Vds = 0.002 V Fermi = 0 eV

10 20 30 40 50

  • 0.4
  • 0.2

0.2 VTG = -1.4V x[nm] Band Edge [eV] 10 20 30 40 50

  • 0.4
  • 0.2

0.2 VTG = 1.15V x[nm] Band Edge [eV] 10 20 30 40 50

  • 0.4
  • 0.2

0.2 VTG = 3.6V x[nm] Band Edge [eV]

ON/OFF = 100

1 2 3 1 2 3 Image courtesy Gianluca Fiori

TG BG

slide-16
SLIDE 16

16 16

Vds = 0.002 V Fermi = 0 eV VBG

Dynamic Band gap

Dynamic band gap: |VBG| ↑ ⇒ E↑ ⇒ Eg↑⇒ ION/IOFF↑

EF,Source EF,Drain Ec Ev

S D

VBG VTG Back oxide Physical structure

Band Gap modulated by back gate

slide-17
SLIDE 17

17 17

VBG

NEMO5 – self consistent simulation

Construct Hamiltonian Add Potential Calculate charge density Calculate Potential Determine energy grid Convergence? 64 million x 64 million matrix 64 M x 64 M Eigenvalue 64 M x 64 M Inverse

One I-V data point

Hundreds of data points

slide-18
SLIDE 18

18 18

iNEMO Group

  • PI: Gerhard Klimeck
  • 3 Research Faculty: Tillmann Kubis, Michael

Povolotskyi, Rajib Rahman

  • Research Scientist: Jim Fonseca
  • 2 Postdocs: Bozidar Novakovic, Jun Huang
  • Students: Tarek Ameen, Robert Andrawis,

James Charles, Chin-Yi Chen, Fan Chen, Yuanchu (Fabio) Chen, Rifat Ferdous, Jun Zhe Geng, Yu He, Yuling Hsueh, Hesameddin Ilatikhameneh, Zhengping Jiang, Daniel Lemus, Pengyu Long, Daniel Mejia Padilla, Kai Miao, Samik Mukherjee, Harshad Sahasrabudhe, Prasad Sarangapani, Saima Sharmin, Yaohua Tan, Yui Hong (Matthias) Tan, Archana Tankasala, Daniel Valencia Hoyos, Kuang Wang, Yu Wang, Evan Wilson

  • Ryan Mokos
  • Intel, Samsung, Philips, TSMC