center for computational nanoscience - asuthe center for computational nanoscience ... enables large...
TRANSCRIPT
The Center for Computational Nanoscience Mark van Schilfgaarde, Director
•
A formal vehicle to forge links in research initiatives– Computational initiatives can be stronger when they can draw on a center– Facilitates finding areas of common interest, and links between ASU with
outside interested in collaborative research or supporting research.
•
A facility to invite visitors– Hope to draw leading experts as colloquium speakers– Possibly visitors can come for short states and collaborations
•
Better communication among existing faculty– Internal seminars inform the rest about what interesting is going on at ASU– Integrate Education component in computational techniques
•
Coordination with other centers, e.g. HPCI, Purdue Nanohub, IGERT, MRSEC
High-Performance Computing Initiative
ASU is fortunate to have an excellent center for high performance computing, initially sponsored by Ira Fulton.Prof. Stanzione
(director) has expanded the initial
center into a 1000 processor facility.Nodes are free
up to a certain maximum, after which
HPC charges $0.25/CPU hour ---
enables large scale calculations at minimal cost.HPC supplies software support of standard computer and scientific packages
(free) and custom support.
HPC offers classes on advanced programming techniques, e.g. code parallelization.Enables faculty to extend range of research, and frees them from maintaining their own facilities.
Classification of Expertise by length scaleASU’s
expertise in theory and computation spans many
disciplines,
ranging from atomic to the macroscopic scale:
First principles, many-electron Schrodinger equation applied to both ground state
properties (total energy,
phonons, etc) and excited state
properties (transport)
Semiclassical particle models for
electron transport
Model forms of the SE (electrons; nucleii
“classical”)
Increasing length scale
Classical atomistic models for: structure and transport
Constitutive relations and many other (>nano)
Representative Faculty (many departments)Ab
initio SE, ground state: Adams (thermochemistry),
Chizmeshya (carbon sequestration), Friesen (later), Rez (oxides), Seo (oxides)
Model SE , excited state: Ferry, Goodnick, Saraniti, Vasileska (later), Shumway (Quantum Monte Carlo, mostly dots)
Ab
initio SE, excited state: : Kotani (later), Sankey (molecular and biological systems), van Schilfgaarde (Magnetism, GW)
Classical atomistic methods: van der Vaart (later), Saraniti (later), Thorpe (rigid body dynamics of nano-bio molecules)
Semiclassical
electron transport: Ferry, Goodnick, Saraniti, Vasileska (later),
Grand Challenges for Computation:
1.
Extend reliability of theory: At the most fundamental level, this is solving the SE to high accuracy. (Example GW theory, Kotani). As length scales increase approximations must, also. A significant challenge at any level.
2.
Find alternatives that implement existing theories in more efficient ways.
3.
Develop techniques that bridge length scales.
Many ways to develop capacity to study new phenomena, or explain phenomena better. Three main classes:
–
Source: Physics Today 58, 49 (2005)
•
The 3 most cited
papers, and 6 of the 11 most cited papers in the Physical Review
series (Phys. Rev. B, Phys. Rev. Lett., Rev. Mod. Physics) all have to do with ab initio approaches to solving the fundamental (Schrodinger) equation
for the electrons.
•
Walter Kohn won the Nobel Prize
for Density Functional theory in 1998.•
Today, it is a major tool
in almost every branch of science and engineering.•
Applicable to whole periodic table; however, its reliability is imperfect, and it is too expensive
for even the mesoscale.
Success story: QSGWQuasiparticle
Self-Consistent GW (QSGW)
is the most accurate
and universally applicable ab initio theory yet developed for solids.
Example: semiconductor bandgaps
predicted to
~0.2 eV
of expt.
LDA
Example: Third-generation Photovoltaics•Growing acceptance that the U.S. must reduce its reliance on oil.–Photovoltaics
must be critical component of this, because their high conversion efficiencies.
–High conversion efficiencies
and low cost
are key barriers to practical realization today.
–A collection of new ideas
have been proposed to increase conversion efficiencies and surmount the standard Schockley-Quiesser
limit. They are called “third-generation”
technologies. They include:
•“selective energy”
contacts that collect electrons before thermalizing•Multiple e−h
generation / incident photon through impact ionization•Multiple e−h
generation / incident photon through impurity bandsBasic problem: none of these ideas has been realized in practice!Are they even feasible?
•
Are 3rd
generation proposals workable?A hard question
to answer experimentally.Theoretical treatment so far has been too
simpleGood modeling should: predict “best case”
scenarios, and the effect of nonidealities
that can occur under realistic conditions.
“Cellular Monte Carlo” simulator based on the Boltzmann equation
(Saraniti
and Goodnick). Models realistic devices
with realistic parameters. The most sophisticated tool of its kind, it can potentially give a reasonable answer to those questions, before spending large resources on trying them.
Example: GaAs MESFET
Bridging length scalesThe CMC
solves the Boltzmann equation
for realistic devices. But it requires as input parameters such as energy bands
and scattering matrix elements. If those inputs are good, the calculation should reliably predict
device behavior. Ab initio theory such as QSGW can supply accurate inputs.
Example: MESFET in GaAs. Much experimental data; reliable model to generate inputs (EPM) can be developed from this data.
Alternatively, use QSGW. Doesn’t rely on experiments! In this case, results almost identical. What about InN?
Ab
initio Electrochemistry (Friesen)
•
A key loss mechanism
in catalysts: The voltage required to generate a practically acceptable current (overpotential) is higher than the thermodynamic potential.
•
This difference occurs through losses
on the atomic scale as the ions cross the electrode/electrolyte
interface.
•
Details are poorly understood, particularly in protic
ionic liquids (little-studied)
•
Plan: use ab initio framework to model the electrodode, electrolyte, and interface; and the ion exchange mechanism.
•
Challenge: interfaces are too big
to realistically attack with good quality present-day modeling methods
based on the
local-density approximation to density-functional theory•
Project includes development of next-generation methods
that will make this possible.
Multi-modal, Bi-functional Electro-Catalysts
Renewable Energy Electrochemistry
Tuning Activity
-Tune hybridization w/ Adsorbates-strain-eigenstrains
Multi modal catalysts provide:-Enhanced transport kinetics-High surface area
EF
Strain
Reactivity
Charge transfer
Reactivity
d-bandDepth Strain
Constitutive Relation
CatalyticActivity
ConstitutiveRelation Constitu
tive
Relation
EF
Strain
Reactivity
Charge transfer
Reactivity
EFEF
Strain
Reactivity
Charge transfer
Reactivity
d-bandDepth Strain
Constitutive Relation
CatalyticActivity
ConstitutiveRelation Constitu
tive
Relation
Friesen Research GroupSchool of Materials, Arizona State University
Tuning catalytic activityTune activity by applying Physical strain: Max ~1%
Tune sp-d hybridization w/ physical strain.
-3 -2 -1 0 1 2 3
-4-202468
10
f_fin
ite (N
/m)
Strain (%)
Intrinsic f
-EigenstrainStrong adsorber
-3 -2 -1 0 1 2 3
-4-202468
10
f_fin
ite (N
/m)
Strain (%)
Intrinsic f
-EigenstrainStrong adsorber
Tune activity by modifying Eigenstrain:>4% achievable through adsorption
Constitutive relations exist between surface stress, strain, eigenstrain, and d-band position.
Map d-band position as a function of surface stress change w/ adsorption to tune in real- time.
Electronic structure calculations allow for the direct analysis of surface stress (~eigenstrain) and d-band position changes in the presence of adsorbed species. Experiment can then directly monitor catalytic activity and surface stress changes in real-time.
Molecular Dynamics of Folding/Unfolding/RefoldingConformational Changes
Conformational change = The change in shape ofa biomolecule
upon
binding other molecules
Arjan
van der
VaartCenter for Biological Physics
Department of Chemistry and Biochemistry
1. transport proteinse.g. maltose-binding protein
3.
molecular motorse.g. FO
F1
-ATPase
2. kinasestransfer phosphate groups from high-energy donor molecules (e.g. ATP), to specific target molecules. Very common in the body.
4. etc. etc.
Conformational changes are crucial
for the functioning of many proteins
Src tyrosine kinase
1. What interactions lead to the conformational behavior?
2. What are the pathways for the conformational change?
3. What is its biological function?
4. Can we block the conformational behavior?
1. High spatial resolution
2. All interactions are quantified
3. Thermodynamic properties can be calculated
Method:
Atomistic Molecular Dynamics Simulations
How do conformational changes work?
Atoms treated as classical point-masses moving in a model potential
– +
Bonds Angles DihedralsImproperDihedrals Electrostatics
van derWaals
Bonded terms Non-bonded terms
kb
(r–r0
)2 ka
(θ–θ0
)2 (q1
q2
)/(εr) E[(rm
/r)12–(rm
/r)6]
The model potential is fitted to experiments
and quantum mechanical
calculations
The potential must be transferable
Atomic motion propagated by classical (Newtonian) dynamics using a very small timestep
(10–15
seconds)
ki
(ω–ω0
)2kd
[1+cos(nφ–φ0
)]
Molecular Dynamics (MD)
protein domain motionunwinding of DNA helix
protein folding
molecular dynamics small systemmolecular dynamics large system
10-6 10-3 1 103 s10-15 10-12 10-9
Thus: need to develop “smart”
techniques!
Problem: Timescales of conformational motion
1
Takao Kotani,Mark
van Schilfgaarde
Quasiparticle
Self-consistentGW (QSGW) method
2
Our questions are
• What defects reduce the performance of HfO2 ? • Spin-polarized current for MRAM, Fe/MgO/Fe• Interpret EELS data to guess atomic structure• New “thermo-power material”?--- We have to know electronic structure ---
So many first-principle calculations are performed now.
3
Difficulties in first-principle methods.
(1) Computational costs(2)
We can not evaluate quantities
directly. E.g. Tc
for High-Tc
SC.(3)
Poorness of Standard approximations,
LDA (local density approx.) or soour QSGW method can solve this problem.
(3) is serious ---“reliability”
can be still problematic.
4
Poorness of standard approximations
Experimental band gap (eV)
CoO
is insulator (gap ~4 eV); LDA predicts metallicLDA+U works. But what U?
5
Quasiparticle self-consistent GW method
Optimum independent-particle picture. Thus, ready to be used for device simulators.
It looks like Hartree-Fock method, but it uses “Dynamical Screened Coulomb interaction *” instead of the bare Coulomb interaction .
*it is determined self-consistently.
6
Our Result by QSGW
7
Experimental band gapvs.
Calculated band gap
QSGW: Good!Almost along the diagonal line!
LDA, GW
Experiment
8
GaAs
QSGW: greenO: Experiment
m* (QSGW) = 0.073m* (LDA) = 0.022m* (expt) = 0.067
Ga d level well described
Typical case for semiconductor
9
ZnO
dielectric
0 10 20 300
1
2
3
4
5
ZnO
with LFCwithout LFCexpt.
Im ε
ω(eV)
10
0
20
40
60
80
100
Γ MX[ζ,ζ,ζ]
meV
[ζ,−ζ,−ζ]
QSGWLDAExperiment
Spinwave dispersion (MnO AF-II)
*Energy bands, and optical properties arealso described well.
11
•
The QSGW
method Self-consistent
perturbation theory
-
optimum independent-particle picture.
-
wide range of materials.
-
no parameters like LDA+U.
QSGW is the very basis for next-generation electronic structure calculation.
-----------------* How to make speed up?* Phonon should be included.* Make it more accurate.
Conclusions
"
TYR40
ARG42
ARG82
ARG132
TYR102
TYR106
ASP113
GLU117
Modeling and Design of Modeling and Design of BiomimeticBiomimetic NanoconductorsNanoconductorsMarco Saraniti, Sasha Smolyanitsky, and Prathibha Ramaprasad
Surface Charge Density on SiOSurface Charge Density on SiO22 PoresPores
bulk pHsu
rface
char
gede
nsity
[C/m
2 ]4 5 6 7 8 9
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
model, 0.2M KClmodel, 0.067M KCl
experiment, 0.2M KCl 1
experiment, 0.067M KCl 1
1. Patricia M. Dove and Colin M. Craven. Surface charge density on silica in alkali and alkaline earth cloride electrolyte solutions. Geochimica et Cosmochimica Acta, 69(21), 2005.
Conductance (pH=7)Conductance (pH=7)
KCl buffer concentration [M]
cond
ucta
nce
[nS
]
10-6 10-5 10-4 10-3 10-2 10-1 100
10-2
10-1
100
101
102
32 nm pore diameter64 nm pore diameter128 nm pore diameter
BD, 32 nm pore diameterBD, 64 nm pore diameterBD, 128 nm pore diameter
3D Brownian Dynamics Simulation3D Brownian Dynamics Simulation
x [nm]0
1020
y [nm]
010
20
z[n
m]
0
10
20
30
40
500.80.750.70.650.60.550.50.450.40.350.30.250.20.150.10.05
concentration, M
cations
x [nm]0
1020
y [nm]
010
20
z[n
m]
0
10
20
30
40
500.50.470.440.410.380.350.320.290.260.230.20.170.140.110.080.05
anions
x [nm]
z[n
m]
5 10 15 20
5
10
15
20
25
30
35
40
45
500.0880.0750.0630.0500.0380.0250.0130.000
potential [V] (slice at y = 10.5 nm)
pore bias [V]
pore
curr
ent[
nA]
-0.4 -0.2 0 0.2 0.4-10
-8
-6
-4
-2
0
2
4
6
8
10
0.6M buffer KCl
0.2M buffer KCl0.4M buffer KCl
pHpH--Dependent ConductanceDependent Conductance
Computational ElectronicsComputational Electronics
DragicaDragica VasileskaVasileskaProfessorProfessor
Arizona State UniversityArizona State University
PrePre--1990 Simulation Efforts1990 Simulation Efforts
SemiclassicalSemiclassical::Monte CarloMonte Carlo--Molecular Dynamics Simulations (FerryMolecular Dynamics Simulations (Ferry--LugliLugli))
Quantum: Quantum: Wigner Function Simulations (Wigner Function Simulations (KluksdahlKluksdahl, , RinghoferRinghofer, Ferry), Ferry)Quantum Hydrodynamic simulations without and with inclusion of Quantum Hydrodynamic simulations without and with inclusion of discrete impurity effects (JR Zhou)discrete impurity effects (JR Zhou)
PostPost--1990 Simulation Efforts1990 Simulation Efforts
SemiclassicalSemiclassical3D Drift3D Drift--diffusion simulations for discrete impurity effects diffusion simulations for discrete impurity effects examination in 50 nm MOSFET devices (Vasileskaexamination in 50 nm MOSFET devices (Vasileska--1996)1996)First 3D Monte Carlo device simulation code with a real First 3D Monte Carlo device simulation code with a real space treatment of the short range electronspace treatment of the short range electron--electron and electron and electronelectron--impurity interactions (Gross, impurity interactions (Gross, VasileskaVasileska, Ferry, Ferry--1997)1997)Development of a variety of 3D Efficient Poisson Equation Development of a variety of 3D Efficient Poisson Equation Solvers (Solvers (WiggerWigger, Speyer, , Speyer, VasileskaVasileska, , GoodnickGoodnick, Saraniti, Saraniti--1997)1997)Development of a fullDevelopment of a full--band CA Monte Carlo device band CA Monte Carlo device simulation code (simulation code (SaranitiSaraniti, , WiggerWigger, Goodnick, Goodnick--1997)1997)Inclusion of the effective potential in Inclusion of the effective potential in semiclassicalsemiclassical Monte Carlo Monte Carlo device simulation schemes (device simulation schemes (VasileskaVasileska, Ferry, Ringhofer, Ferry, Ringhofer--2000)2000)Development of the first 2D thermal device simulation Development of the first 2D thermal device simulation code (code (RalevaRaleva, , VasileskaVasileska, Goodnick, Goodnick--2007)2007)
MOSFETs - Role of the E-E and E-I
0
100
200
300
400
100 110 120 130 140 150 160 170 180
with e-e and e-imesh force only
Ele
ctro
n en
ergy
[m
eV]
Length [nm]
VD=1 V, V
G=1 V
channel drain
0
100
200
300
400
500
600
700
800
0.12 0.13 0.14 0.15 0.16 0.17 0.18
Ene
rgy
[meV
]
Length [nm]
0
100
200
300
400
500
600
700
800
0.12 0.13 0.14 0.15 0.16 0.17 0.18
Length [nm]
Ene
rgy
[meV
]
mesh forceonly
with e-e and e-i
Individual electron trajectories over
time
-1x107
-5x106
0
5x106
1x107
1.5x107
2x107
2.5x107
0 40 80 120 160
with e-e and e-imesh force only
Drif
t vel
ocity
[cm
/s]
Length [nm]
VD=V
G=1.0 V
source drainchannel
-1x107
-5x106
0
5x106
1x107
1.5x107
2x107
2.5x107
0 40 80 120 160
with e-e and e-imesh force only
Drif
t vel
ocity
[cm
/s]
Length [nm]
VD=V
G=1.0 V
source drainchannel
MOSFETs - Discrete Impurity Effects
0
20
40
60
80
100
0 1 2 3 4 5
σVth
=> approach 1σ
Vth => approach 2
σVth
=> our simulation results
σ Vth [
mV
]
Oxide thickness Tox
[nm]
05
10152025303540
1x1018 3x1018 5x1018 7x1018
σVth
=> approach 1σ
Vth => approach 2
σVth
=> our simulation results
σ Vth
[m
V]
Doping density NA [cm-3]
10
20
30
40
50
60
20 40 60 80 100 120 140
σVth
=> approach 1σ
Vth => approach 2
σVth
=> our simulation results
σ Vth
[mV
]Device width [nm]
effeff
A
oxox
ABSiBBSi
Vth WLNT
Nqq/Tkq 44 3
434
⎥⎦
⎤⎢⎣
⎡ε
+φε
φε≈σApproach 2 [2]:
⎟⎟⎠
⎞⎜⎜⎝
⎛=φ
εφε
=σiAB
Beffeff
A
oxoxBSi
Vth nN
lnqTk
;WL
NTq
44 3
2Approach 1 [1]:
[1] T. Mizuno, J. Okamura, and A. Toriumi, IEEE Trans. Electron Dev. 41, 2216 (1994).
[2] P. A. Stolk, F. P. Widdershoven, and D. B. Klaassen, IEEE Trans. Electron Dev. 45, 1960 (1998).
Lattice Heating and Scaling of Transistor Dimensions
25 nm FD SOI nMOSFET (Vgs=Vds=1.2V)
10 20 30 40 50 60 70
3
8
131345 nm FD SOI nMOSFET (Vgs=Vds=1.2V)
20 40 60 80 100 120
3
12
212160 nm FD SOI nMOSFET (Vgs=Vds=1.2V)
20 40 60 80 100 120 140 160 180
3
15
2727
300
400
500
300
400
500
300
400
500
source contact drain contact
source region drain region
80 nm FD SOI nMOSFET (Vgs=Vds=1.5V)
50 100 150 200
3
20
353590 nm FD SOI nMOSFET (Vgs=Vds=1.5V)
50 100 150 200 250
3
21
3939100 nm FD SOI nMOSFET (Vgs=Vds=1.5V)
50 100 150 200 250 300
3
23
4343
300
400
500
600
300
400
500
300
400
500
Conventional SOI Transistors vs. SOD transistors with different boundary con-dition for the temperature on the gate
0 25 50 7575
0
10
20
30
40
50
60
x (nm)
y (n
m)
300
350
400
450
500source drain
BOX
0 25 50 7575
0
10
20
30
40
50
60
x (nm)
y (n
m)
300
350
400
450
500source drain
BOX
0 25 50 7575
0
2
4
6
8
10
x (nm)
y (n
m)
300
305
310
315
320
325
330
335
source drain
0 25 50 7575
0
2
4
6
8
10
x (nm)
y (n
m)
300
320
340
360
380
400
source drain
0 25 50 7575
0
2
4
6
8
10
x (nm)
y (n
m)
300
310
320
330
340
source drain
0 25 50 7575
0
2
4
6
8
10
x (nm)
y (n
m)
300
320
340
360
380
400
source drain
SOI Device
SOD Device
SOAlN Device
Notice better heat spread in the SOD and the SOAlNdevices.
QuantumQuantumGreenGreen’’s function s function codelcodel for the lowfor the low--field mobility calculation field mobility calculation (1993(1993--Vasileska, Ferry) Vasileska, Ferry) –– For the first time we have For the first time we have included selfincluded self--consistent Born approximation in realistic consistent Born approximation in realistic transport calculationtransport calculationDevelopment of SCHRED (VasileskaDevelopment of SCHRED (Vasileska--1997)1997)Development of a variety of 2D/3D Schrodinger Poisson solvers Development of a variety of 2D/3D Schrodinger Poisson solvers for simulations of quantum dots in Si and for simulations of quantum dots in Si and GaAsGaAs technology technology ((VasileskaVasileska, Ahmed , Ahmed -- 1999)1999)Effective potential inclusion in Effective potential inclusion in semiclassicalsemiclassical simulators simulators ((VasileskaVasileska, Ferry, Ahmed, , Ferry, Ahmed, RinghoferRinghofer -- 2002)2002)2D Poisson2D Poisson--1D Schrodinger full1D Schrodinger full--band simulator for band simulator for transport in ptransport in p--channel Si/channel Si/SiGeSiGe devices (devices (KrischnanKrischnan, , VasileskaVasileska, , FischettiFischetti –– 2005)2005)Development of 2D/3D Contact block reduction method Development of 2D/3D Contact block reduction method for quantum transport of for quantum transport of nanoscalenanoscale devices (Khan, devices (Khan, MamaluyMamaluy, , VasileskaVasileska))
PostPost--1990 Simulation Efforts1990 Simulation Efforts
Universal mobility exploration in conventionalSi inversion layers.
Exchange-correlation effects
Starined-Si inversion layers
Green’s Functions
SCHRED was one of the first toolsthat was installed on PUNCH (nanoHUB) and is the most used non-comercial simulation module.
SCHRED
Conclusions From This Work …
• We developed the first full-band self-consistent MC device simulator that takes both band-structure and quantum-mechanical size quantization effects into account.
• We find that at low applied bias there is an improved performance of p-channel strained SiGe devices due to the mobility enhancement.
• The improvement becomes degradation effect at high gate biases due to increased importance of surface-roughness and alloy disorder scattering.
Modeling p-channel SiGe Devices
ELECTRON DENSITY: DG VS. TG
0
2
4
Y [n
m]
tox
0 2 4 6 8 10 120
2
4
6
8
Z [nm]
Y [n
m] tox
2.0E
186.
0E18
1.2E
191.
8E19
2.4E
193.
0E19
3.6E
194.
2E19
<2.0
E18
>
ON-state Electron density along the dotted line
VGS=0.2, VDS=0.4V
Electron density (TG) > electron density (DG)
CBR – FinFET Simulation
0.0 0.1 0.2 0.3 0.4
0
1
2
3
4
Dra
in c
urre
nt [μ
A]
Drain voltage [V]
DG FinFET TG FinFET
VGS = 0.1V
TG FinFET stronger control of gates pronounced saturationHigh output resistance compared to DG structure
Future WorkFuture Work
On the On the semiclassicalsemiclassical arena:arena:MC device simulator for modeling Solar CellsMC device simulator for modeling Solar CellsCoupled MC simulators for electrons and phonons to Coupled MC simulators for electrons and phonons to model heating effects in model heating effects in nanoscalenanoscale devicesdevices
On the quantum arena:On the quantum arena:Coupled 3DCBR and NEMO3D code for modeling Coupled 3DCBR and NEMO3D code for modeling quantum dot quantum dot photodetectorsphotodetectors and third generation and third generation quantumquantum--dot solar cellsdot solar cellsExtend 3DCBR to include magnetic domains and Extend 3DCBR to include magnetic domains and model spin currents.model spin currents.