DANIEL TSUI LECTURE

BEIJING 2005
SANKAR DAS SARMA
UNIVERSITY OF MARYLAND
CONDENSED MATTER THEORY CENTER
WWW.PHYSICS.UMD.EDU/CMTC

My connection to Dan
1976
1987-88
1995-2005
2004
My connection to China

Lai WY 1983-85 Beijing

Xie XC 1983-87; 88-91 USTC

Zhang FC 1984-86 Fudan

He S 1988-92 USTC

Li Q 1989-93 USTC

Lai ZW 1990-92 USTC(?);Chicago

Liu DZ 1990-94 USTC

Zheng L (1995-98); Hu J (1997-99) Indiana

Hu XD 1998-2003 Beijing;Michigan

Zhang Y 2002- USTC; Yale

Wang DW Taiwan 1996-2002

Tse GW Hong Kong 2004-

Also B.Y.K Hu; K.E. Khor,

SC Zhang (Stanford); R. Zia (Virginia Tech.);

DC Tsui (Princeton).

More than 100 publications with these collaborators!

I was in China (Beijing, Shanghai)

in 1986 as a guest of the Chinese
Academy of Sciences with the Institute of
Physics being my host!
 TIDBITS ABOUT QUBITS
Sankar Das Sarma

QUBITS = TWO-LEVEL
QUANTUM SYSTEM
LINEAR SUPERPOSITION
QUANTUM ENTANGLEMENT
QUANTUM PARALLELISM
TOPOLOGICAL QUANTUM COMPUTATION
www.physics.umd.edu/cmtc
 A (VERY) BRIEF HISTORY
OF COMPUTATION
UNARY: 10,000 YEARS AGO
BINARY: 1,000 YEARS AGO; BITS
ANALOG COMPUTERS: ~ 1000 years
BOOLEAN ALGEBRA: BITS
DIGITAL COMPUTERS: ~ 100 years
QUANTUM MECH.: 100 YEARS AGO
QUBITS: NOW (PERHAPS)
QUANTUM COMPUTERS: ??
 Spin
Quantum Computation in
Semiconductor Nanostructures
Localized Spin 1\2 qubits in Semiconductor
Nanostructures
(Heisenberg Coupling)

X. Hu;R. de Sousa;B. Koiller;V. Scarola;

W.Witzel
ARDA, ARO, UMD, LPS, NSA
 SPINTRONICS
SPIN MATERIALS
Diluted magnetic semiconductors (DMS): ferromagnetic

SPIN DEVICES
Active control of (nonequilibrium) spin AND charge

SPIN QUBITS
Scalable solid state spin quantum computation
 SPINTRONICS
SPIN + ELECTRONICS

Killer app. : SPIN QUANTUM COMPUTATION!

 QUANTUM COMPUTERS
HOW TO BUILD A QC
PHYSICS OF QC ARCHITECTURE

SCALABLE and ROBUST

FAULT TOLERANT
100-10,000 COUPLED QUBITS

Qubit dynamics
Qubit coupling, entanglement
Qubit decoherence
 What can a QC do?
Why build a QC?
Prime factorization Quantum parallelism
Shor algorithm Entanglement
Universal one and
Exponential speedup
two-qubit gates
Database search
Quantum error
Grover algorithm correction
Algebraic speedup Boolean vs. Quantum
Quantum simulation P/NP some day??
Feynmans dream Topological QC
Minimal QC Requirements
Qubits: 2-level quantum systems
Initialization of qubits
Control and manipulation of qubits
Quantum coupling of 2-qubits
1- and 2-qubit gates
Quantum error correction
High fidelity
Qubit specific measurement
Long quantum coherence
 PROPOSED QC
ARCHITECTURES (far too many)
ION TRAPS
LIQUID STATE NMR
NEUTRAL ATOM OPTICAL LATTICE
CAVITY QED
SQUIDS, JOSEPHSON JUNCTIONS
COOPER PAIR BOXES
ELECTRON SPINS IN SOLIDS (GaAs, Si)
SOLID STATE NMR
ELECTRON STATES ON HE-4 SURFACE
QUANTUM HALL STATES
 Quantum computing with spins
0
Electron/nuclear spin: An ideal qubit?
1

Quantum algorithms: Factoring, searching...

U N U1 01 0 an n
Measurement
10 1
n

Input Output

1-qubit: Spin rotation

Quantum gates:
2-qubit: Exchange interaction

S1 S 2 2

2
 Spin relaxation and manipulation of
localized states in semiconductors:
Considerations for solid state quantum computer
architectures

Quantum Dot Si Donor Nuclear Spin

QC Architecture QC Architecture
 Semiconductor
implementations
GaAs quantum dots
D. Loss and D.P. DiVincenzo,
PRA 1998

Silicon donors (P)

B. Kane, Nature 1998
R. Vrijen et al., PRA 2000

Fault tolerant if coherence time TM 10 4 gate

 Experiments

GaAs

Neighboring quantum dots

Single electron in each dot

Does a model of this system reproduce the Heisenberg

Spin Transitions in Few Electron Quantum Dots
Exact Diagonalization Theory
Going beyond perturbative/Heitler-London
exchange gate calculations in coupled dot
QC architectures ATOM to MOLECULE
Vito Scarola

WHEN IS THE 2-ELECTRON

QUNTUM DOT A MOLECULE WITH
TUNABLE EXCHANGE COUPLING?
WHEN IS IT JUST AN ARTIFICIAL 2-
ATOM SYSTEM?
Model
 Electron Mitosis 40 B=0 T

HOMOPOLAR BINDING IN AN

Y [nm]
ARTIFICIAL MOLECULE
0

-40
-40 0 40
X [nm]

30 B=9T

-30
-30 0 30
 Schematic Parameter Space
Cyclotron energy
Parabolic confinement

Dot separation
Modified magnetic length

Small Exchange

Spin
Hamiltonian
Vortex
1 Mixing

Level
Crossings

0 1
(magnetic field)
 Three electrons-Three Dots

B=5T
R=20nm
0=3meV
 Conclusion

Exact diagonalization allows accurate

treatment of strongly interacting regime

Exchange splitting (J) oscillates with

magnetic field

Trial state analysis implies singlet-triplet

transitions of Composite Fermions
Artificial Atom to Artificial Molecule
 Two Spins in Two Quantum
Dots:Quantum Gates
B S1 S2

Single spin qubits

Qubit #1 Qubit #2

Heisenberg
Hamiltonian:

Quantum gates:

Heisenberg interaction + local magnetic field gives

Validity of Heisenberg Exchange Hamiltonian
For Spin-Based Quantum Dot Quantum Computers
Our system
Energy spectrum

Exchange splitting
 Validity of Heisenberg Exchange Hamiltonian
For Six-Electron Double Quantum Dot

Six electron double dot

Energy spectrum

Exchange splitting
 Adiabatic Condition
When the system Hamiltonian is changed
adiabatically, the system wavefunction can be
expanded on the instantaneous eigenstates:
(t ) Ci (t )ui (t ),
i

System evolution is governed by the

Schroedinger equation:
ck N
ci H i t
k i exp ( Ek Ei )d
t i k Ek Ei t

Instantaneous eigenvalues and eigenstates are

needed to integrate the Schroedinger
equation.
 Loss due to non-adiabaticity
In an exchange gate for a double dot
 Exchange in silicon-based quantum computer architecture
MOTIVATION
Kanes proposal for a silicon-based quantum computer

B.E.Kane,
Nature (1998)

P donors in Si
Concern with donor positioning:
University of New South Walws
From the website of SNF at the

Each 31P in the array must be

exactly under the A-gate.
Sydney, Australia
 BUILDING BLOCKS OF KANES PROPOSAL
qubits are the 31
P nuclear spins (I=)
Spin interactions in Si:31P
1-qubit operations 2-qubit operations

R=
H ( R) H ( B) A1 1z e 1z n A2 z2 e z2 n
H e n B B ze g n n B zn A ze zn
J(R) 1z e z2e
EXCHANGE
Hydrogenic model for P donors in Si

~ +
_

e2
Si (IV) 14 e P (V) 15 e U (r )
14 p+ r
[ 2 2 / 2m U (r )] (r ) 15 Ed* (r )
*

p+ o
(r ) (1 / a * ) exp(r / a*) , a* a0 (m0 / m*) 30 A
3

Asymptotic exchange coupling of two hydrogen atoms

(Herring&Flicker, 1964) 2 5
e R
J ( R ) ~ E0 ( ) exp(2 R / a*)
2
a* a*
Electrons in Si ( beyond m* and )

CONDUCTION
REAL SPACE: RECIPROCAL
BAND MINIMUM:
Diamond SPACE:
Anisotropic and
structure Brillouin zone
six-fold degenerate
 Exchange between 31
P donors in Si
6
1
Ground state (r) F (r) (r)
1 6
1 [( x 2 y 2 ) / a 2 z 2 / b 2 ]
Envelope functions: F z (r ) e
a 2b
, u (r ) K c e
Bloch wave ik r
u (r )e K

iK r
functions:

Heitler-London triplet-singlet splitting

J (R ) Et Es

c 2 2 i ( K K ')R 2 2
K c
K' e (R ) cos(k k ) R
, K ,K '
Exchange calculated for two donors
along [100]

*
 Exchange versus donor displacements
within the Si unit cell

2nd neigh.
(12)

(4)
*
1st neigh.
PRL 88, 027903 (2002).

3rd neigh.
(6)

*
 The extreme sensitivity of the
exchange coupling to the relative
positioning of the substitutional donor
pair in Si is entirely due to the

six-fold degeneracy of the Si

conduction band minimum.

Dipolar spin coupling ? Dipolar gates?

Qubits are dipolar coupled single electron spins

R. de Sousa et al., cond-mat/031140, PRA 70, 052304 (2003)

Si:P SPIN DIPOLAR GATE QC ARCHITECTURE
 Gate imperfection in the presence of exchange

Long-range dipolar ~1/R3 is much stronger than short-range

exchange for large inter-donor separation; How large should be
the separation so that J can be neglected?

J0 leads to error of the order of (J/D)2; Hence the

criterium for gate error to be within p is:
 Gate times and donor separation

Separations of the order of 300 allows easier lithography;

Gates are 106 times slower than exchange coupling;
however there is no need for exchange control and donor
positioning with atomic precision.
Using 28Si we expect T2~T1~ seconds for B~1T
 Si Dipolar QC
Long range couplings are corrected with no overhead in
gate time (ability to -pulse within 5s is required).
Dipolar implementation is reliable, its
advantages/disadvantages should be compared with other
proposals without exchange (for example, Skinner,
Davenport, Kane, PRL 2003, which requires electron
shuttling between donors);
Dipolar coupling insensitive to electronic structure: No
inter-valley interference, interstitial defects are also good
qubits;
Top-down construction schemes based on ion
implantation can be used even though they lack atomic
precision in donor positioning.
Can be scaled up
 Electron spin coherence in
semiconductor QCs
Bound orbital states T1 ~ 1ms (GaAs Quantum dot)
(B=1T, T<<1K) 10 s (Si:P)
Decoherence is dominated by spin-spin interactions:
SPECTRAL DIFFUSION

Electrons Zeeman frequency

fluctuates due to
B nuclear dipolar flip-flops
RESULTS:
T2 ~ 50 s GaAs-QD
>1000 s Si:P
 Spin-orbit + phonons
Blochs equation
Hyperfine + phonons
Spin-orbit + photons

dM 1 1
g B B M M M || Spectral diffusion
(nuclear spins,
dt T2 T1 time dependent
magnetic fields)
Dipolar / exchange
coupling between like
* spins
T2 TM T1 Unresolved hyperfine
structure
Different g-factors
Inhomogeneous fields
Dipolar / exchange
between unlike spins.
 Spectral diffusion of a Si:P spin

B
Nuclear induced spectral diffusion
Nuclear spins flip-flop due to
their dipolar interaction;
Electrons Zeeman frequency
fluctuates in time due to nuclear
hyperfine field.

Theory
Nuclear pairs are described by
Poisson random variables;
Flip-flop rates are calculated
using the method of moments, a
high temperature expansion.
The Hamiltonian

An
S I n S I n
2
Dependency with 29Si density,
sample orientation

TM increases very fast when we remove 29Si !

 Spin-1/2 theory of nuclear spectral
diffusion: Comparison with experiment
700 Theory: R. de Sousa, S. Das Sarma, PRB 68, 115322 (2003)
600 Experiment: E. Abe, K.M. Itoh, J. Isoya, S. Yamasaki,
cond-mat/0402152.
500 29
Natural [4.67% Si]
400 Theory/2.5
300 Experiment
TM [ s]

200
60
50
40 Enriched [99.22% Si]
29

Theory/2.5
30 Experiment
20
0 10 20 30 40 50 60 70 80 90
[100] [111] [011]
[degrees]
 GaAs quantum dots

Spectral diffusion is very important: Ga and As do

not have I=0 isotopes !
DYNAMIC NUCLEAR POLARIZATION ?
Quantum theory of spectral diffusion: Cluster
expansion results
[W.M. Witzel, R. de Sousa, S. Das Sarma, cond-mat/0501503]
1.0 10

(a) (b)
0
1 =60
0.8

0
=10
Hahn echo

0.1
0.6
0
stochastic ( =60 )
0.1 1
0.4
0 0
=60 =0
0.2

0.0
0.0 0.1 0.2 0.3 0.4 0.5
[ms]
 Conclusions

Electrical control of single spin dynamics is

promising for III-V quantum dots because of spin-
orbit coupling;
The spin of localized states interact weakly with
the phonons at low T: Nuclear induced spectral
diffusion if the dominant decoherence mechanism;
Isotopically purified Si:P donor spins can be
coherent for ~1000 s (B = 0.3 Tesla); 60 ms
already measured ! (S.A. Lyon, 2003)
GaAs quantum dots (or donors) coherent for only
1 100 s, but TM /J > 106 !
Making a quantum computer
What is the right analogy?

Aviation?
Manhattan project?
Controlled fusion?
Integrated circuits (chips)?

Toy paper airplane to 747 jumbo aircrafts

10-15 YEARS FOR <100 QUBITS RESEARCH QC

PERHAPS 50 YEARS FOR A COMMERCIAL QC
BASED ON LINEAR EXTRAPOLATION
 Solid state spin quantum
computation in semiconductors
CMTC/UMD Spin Quantum Computation Group
Sankar Das Sarma
Vito Scarola
Kwon Park SDS, Michael Freedman, Chetan Nayak
cond-mat/0412343 (PRL 2005)
Belita Koiller
Xuedong Hu
Rogerio De Sousa, Wayne Witzel
Juan Delgado, Magdalena Constantin
Supported by NSA, LPS, ARO, ARDA, UMD
 Quantum theory of spectral diffusion: Two
possible series expansions
[W.M. Witzel, R. de Sousa, S. Das Sarma, cond-mat/0501503]

1
Exact expression: v( ) Tr U U e H n / kBTU U
M
rr
U e iH
with
1

H An I nz bnm I n I m 3I nz I mz
2 n nm

Essential condition: Max bnm = 1

Expansion in : Expansion in :

v( ) 1 4 c bnm O cnm bnm

4 6
2 4 1 4bnm
nm
powers of =1 !
nm cnm An Am
An Am
converges when cnm =1
4bnm
 Quantum theory of spectral diffusion:
non-perturbative cluster expansion
Define set D contribution recursively: vD ( ) vD ( ) vS ( )
S D
k0
Additive version of cluster exp.: v( ) 1 vD ( ) O( k0 1 )
k 2 | D| k

Large sets D are mainly composed of disconnected clusters

Si; If clusters are far enough, neglect inter-cluster coupling to vD vSi
i
get

Examples of |D|=10:

Need 10-site exact solution Need 2-site solution only!

 Lowest order cluster expansion: Product of
pairs

v( ) 1 vnm
( ) exp vnm ( )
n m nm
configurations
Exact solution for pair nm
: 110 10

: 1109 ( )
vnm ( ) 1 vnm

: 4 107 2
4cnm
1 sin 4
(nm )
: 3 107 1 c
2
2

: 1106
nm
nm bnm 1 cnm
: 4 103
2

| An Am |
cnm = 1 v( ) 1 4 cnm bnm 1 4
4
= 1 v( ) 1 2 sin 4
2

n m cnm c
n m nm 4

Cluster expansion interpolates between and

expansions at the lowest order!
 Spin-orbit coupling in semiconductor
heterojunctions
z0
V (z )
eEz z0
1 *
V ( x, y ) m 0 ( x 2 y 2 )
2

Spin-orbit coupling

Dresselhaus:

Rashba:
 Spin-flip + phonon

Z Coupling
energy

T1 ~ 10 ms for GaAs (=30 nm, B=1 T)

Spin relaxation in III-V quantum dots

4
10 GaSb
Spin-flip rate 1/T1 [s ]
-1

3
10 InSb GaAs
2
10
1
10 InAs

0
10
-1
10
0.01 0.1 1 10
Longitudinal Magnetic field [Tesla]
 Electrical control of g-factor

Electron penetration D.D. Awschalom, Nature 2001

into AlGaAs barrier E. Yablonovitch, PRB 2001

But even without barrier penetration:

g g0 ( E 2 4/3
E ) O( B )
2 2 2 2

Dresselhaus! Rashba!

( g g0 ) / g0 ~ 1 for E~105 Volt/cm !

 Electrical manipulation of g-factor in GaAs
E=104 V/cm Dresselhaus dominated!
105 V/cm

1.0 1.0
GaAs
0.8
0.9
0.6
g g0

0.4 0.8

g g0
0.2
0.7
0.0
-0.2 0.6 GaAs
-0.4 0.5
20 40 60 80 100 120 140 0 2 4 6 8 10 12 14
Quantum dot radius l0 [nm] Longitudinal magnetic field [Tesla]
 Electrical manipulation of g-factor in InAs
Rashba dominated!

1.7 1.8
1.6 InAs InAs
1.5 1.6
1.4
1.4
g g0

1.3

g g0
1.2
1.2
1.1
1.0 1.0
0.9
0.8 0.8
10 15 20 25 30 35 40 0 1 2 3 4
Quantum dot radius l0 [nm] Longitudinal magnetic field [Tesla]

E=104 V/cm
 g-factor control T1 control !!
6
10
5
10
Spin-flip rate 1/T1 [s ]
-1

4
10
3
10
2
10 4
1 X 10 Volt/cm
5
1 5 X 10
10 5
7 X 10
0
10
0 50 100 150 200
Quantum dot radius l0 [nm]