You are on page 1of 58

Solid State

Quantum Computation.
1. Spins

David DiVincenzo, IBM

Can. Summer School, 5/2004
Physical systems actively considered
for quantum computer implementation
Liquid-state NMR
NMR spin lattices
Linear ion-trap
spectroscopy
Neutral-atom optical
lattices
Cavity QED + atoms
Linear optics with single
photons
Nitrogen vacancies in
diamond
Electrons on liquid He
Small Josephson junctions
charge qubits
flux qubits
Spin spectroscopies,
impurities in semiconductors
Coupled quantum dots
Qubits:
spin,charge,excitons
Exchange coupled, cavity
coupled
Concept device: spin-resonance transistor
R. Vrijen et al, Phys. Rev. A 62, 012306 (2000)
Kane (1998)
Quantum-dot array proposal
Five criteria for physical implementation of
a quantum computer
1. Well defined extendible qubit array -stable
memory
2. Preparable in the 000 state
3. Long decoherence time (>10
4
operation time)
4. Universal set of gate operations
5. Single-quantum measurements
D. P. DiVincenzo, in Mesoscopic Electron Transport, eds. Sohn, Kowenhoven,
Schoen (Kluwer 1997), p. 657, cond-mat/9612126; The Physical
Implementation of Quantum Computation, Fort. der Physik 48, 771 (2000),
quant-ph/0002077.
5. Measurement requirement
Ideal quantum measurement for quantum computing:
For the selected qubit:
if its state is |0, the classical outcome is always 0
if its state is |1, the classical outcome is always 1
(100% quantum efficiency)
If quantum efficiency is not perfect but still large (50%),
desired measurement is achieved by copying (using
cNOT gates) qubit into several others and measuring all.
If q.e. is very low, quantum computing can still be
accomplished using ensemble technique (cf. bulk NMR)
Fast measurements (10
-4
of decoherence time) permit
easier error correction, but are not necessary
Loss & DiVincenzo
quant-ph/9701055
Single spin measurement now achieved by Delft
group (Elzerman, Vandersypen, Hansen,
Kouwenhoven, Dec. 2003)
Variant on spin-charge conversion mechanism.
QPC detector
1 electron
quantum dot
Quantum dot attached to spin-polarized leads*
P. Recher, E.V. Sukhorukov, D. Loss, Phys. Rev. Lett. 85, 1962 (2000)

Spin Read-out via Spin Polarized Leads
* - magnetic semiconductors [R. Fiederling et al., Nature 402, 787 (1999); Y. Ohno et al., Nature 402, 790 (1999)]
- Quantum Hall Edge states [M. Ciorga et al., PRB 61, R16315 (2000)],


Thus: I
s
= 0
spin up ( I
c
<< I
s
)
I
s
> 0
spin down
=> single-spin memory device, US patent PCT/GB00/03416
4. Universal Set of Quantum Gates
Quantum algorithms are specified as sequences of unitary
transformations U
1
,U
2
, U
3
, each acting on a small number of qubits
Each U is generated by a time-dependent Hamiltonian:
) / ) ( exp(
}
= t dtH i U o o
Different Hamiltonians are needed to generate the desired
quantum gates:
yi xi H o o ,
zj zi H cNOT o o
1-bit gate
many different repertoires possible
integrated strength of H should be very precise, 1 part in 10
-4
,
from current understanding of error correction
(but, see topological quantum computing (Kitaev, 1997))
Quantum-dot array proposal
Gate operations with quantum dots (1):

--two-qubit gate:

Use the side gates to move electron positions
horizontally, changing the wavefunction overlap
Pauli exclusion principle produces spin-spin interaction:
) (
2 1 2 1 2 1 2 1 z z y y x x
J S JS H o o o o o o + + = =
Model calculations (Burkard, Loss, DiVincenzo, PRB, 1999)
For small dots (40nm) give J 0.1meV, giving a time for the
square root of swap of
t 40 psec
NB: interaction is very short ranged, off state is accurately H=0.
Making the CNOT from exchange:
) a |
) a |
) b |
) b |
Exchange generates the
SWAP operation:
More useful is the square root of swap, S
S S
=
Using SWAP:
S S
z
o

CNOT
Quantum-dot array proposal
Gate operations with quantum dots (2):

--one-qubit gate:

Desired Hamiltonian is:
) (
z z y y x x B B
B B B g B S g H o o o + + = =
One approach: use back gate to move electron
vertically. Wavefunction overlap with magnetic
or high g-factor layers produces desired Hamiltonian.
If B
eff
= 1T, t 160 psec
If B
eff
= 1mT, t 160 nsec
Can we get CNOT with just
Heisenberg exchange?
Conventional answer NO:

--because Heisenberg interaction has too much symmetry
--it cannot change
S (total angular momentum quantum number)
S
z
(z component of total angular momentum)
Correct answer (Berkeley, MIT, Los Alamos) YES:

--the trick: encode qubits in states of specific angular momentum
quantum numbers
Specific scheme to get quantum gates with just
Heisenberg exchange:
Most economical coding scheme:
1 qubit = 3 spins:
+|| |+|
L
0
( ) | +| |+ =
(i.e., singlet times spin-up)
+|| |+| ||+ 2 1
L
(triplet on first two
spins)
Because quantum numbers are fixed (S=1/2, S
z
=+1/2), all gates on
These logical qubits can be performed using SWAP:
Economical coded-gate implementations
results of simulations
By varying interactions times shown,
all 1-qubit gates on coded qubits can be obtained with no more
than 4 exchange operations (if only nearest-neighbor interactions)
or 3 exchange interactions (if interactions between spin 1 and spin 3
are possible)
CNOT on two coded qubits
-minimal solution
19 interactions,
doable in 13 time steps

-essentially unique

-gate accuracy 10
-5

with precision shown

-nearest-neighbor
seems best

k
k
k k k k
k k k
k
k k
k k
k
k
k
k
k
k
k
k
k
k
k
k
k k
Simple features of scheme
for coded computation
--Initialization: turn on uniform B field and strong antiferromagnetic
Heisenberg exchange between spins 1 and 2. Then

L
0
( ) | +| |+
is the ground state of the system.
--Measurement: coded qubit is measured by determining whether
spins 1 and 2 are in a relative singlet or triplet. Somewhat easier
than single-spin measurements.
Spin Orbit
first-order effect (Dzyaloshinski-Moriya)
Bonesteel, Stepanenko, DiVincenzo, Phys. Rev. Lett. 87, 207901 (2001)
see also G. Burkard and D. Loss, Phys. Rev. Lett. 88, 047903 (2002)
x
y
y
z
z
R
Heuristic: tunneling motion produces effective B field
( ) ( )
2 1 2 1 2 1
S JS S S S S J H R ~ + = |
Kavokin, cond-mat/0011340
Cancellation of spin-orbit effects
Exchange
( ) ) ( ) (
2 1
t A t J H + = S S
Spin-orbit
) ) ( )( ) ( ( ) ( ) ( ) ( ) (
2 1 2 1
S S S S + = t t t t t A
(Moriya-Dzyaloshinski; Kavokin 01; Bonesteel et al., q-phys/0106161)
Burkard & Loss, Phys. Rev. Lett. 88, 047903 (2002)
|
|
|
.
|

\
|
= =
+

) 1 (
0 0 0
0 1 0 0
0 0 1 0
0 0 0
) 1 (
2 2 1
2 / 2 /
o
o
t t t
i
ie
i
ie
V Ve e e U
z z z
S i S i S i
g
Quantum gate
|
|
|
.
|

\
|
|
|
|
.
|

\
|
=
|
|
.
|

\
|

+

+
}

o
|
|
o

t
t
1 0 0 0
0 1 0
0 1 0
0 0 0 1
2
exp ) ( exp
2 /
2 /
i
i
i dt t H i T V
s
s
XOR g
U U dt t J
s
s
= + =
}

) 1 ( 2 / ) (
2 /
2 /
o t
t
t
i.e. for ideal case with identical pulse shapes ( ),
spin-orbit cancels exactly in XOR Gate
2
| o =
const. = = |
Physics Issues for Spin Qubits
Big issue with solid state qubits: do we have
effective and robust controls available (does the
electrical engineering work)?
Theoretical focus on fundamental decoherence
problems: hyperfine effects (nuclear spins) and
spin-orbit
For these two, there are many possible problems
and (almost) as many theoretically proposed
solutions.
Solid State
Quantum Computation.
2. Superconductors

David DiVincenzo, IBM

Can. Summer School, 6/2004
Typical flux qubit (Delft) Chioescu, Nakamura, and Mooij,
Science (2004).
Not just a Josephson junction, but a complex electric circuit, operating
quantum-mechamically.
Recent progress Josephson junction qubit


Science 296, 886 (2002)
Simple electric circuit
small
L
C
harmonic oscillator with resonant
frequency
LC / 1
0
= e
Quantum mechanically:
x is any circuit variable
(capacitor charge/current/voltage,
Inductor flux/current/voltage)
That is to say, it is a
macroscopic variable that is
being quantized.
But we will need to learn to deal
with
small
--Josephson junctions
--current sources
--resistances and impedances
--mutual inductances
--non-linear circuit elements?
Josephson junction circuits
small
Practical Josephson
junction is a combination
of three electrical
elements:
Ideal Josephson junction (x in circuit):
current controlled by difference in
superconducting phase phi across the
tunnel junction:
Completely new electrical circuit
element, right?
not really
small
Whats an inductor (linear
or nonlinear)?
( ) u =
u =
= u

1
1
,
L I
L I
LI
u
is the magnetic flux
produced by the
inductor
V = u
-
(Faraday)

Ideal Josephson junction:


is the superconducting phase
difference across the barrier
V =
-
u

t 2
0
(Josephsons second law)
e h/
0
= u
flux quantum
(instantaneous)
not really
small
Whats an inductor (linear
or nonlinear)?
( ) u =
u =
= u

1
1
,
L I
L I
LI
u
is the magnetic flux
produced by the
inductor
V = u
-
(Faraday)

Ideal Josephson junction:


is the superconducting phase
difference across the barrier
V =
-
u

t 2
0
(Josephsons second law)
Phenomenologically, Josephson junctions
are non-linear inductors.
So, we now do the systematic
quantum theory
small
Strategy: correspondence principle
small
--Write circuit equations of motion: these are equations of classical
mechanics
--Technical challenge: it is a classical mechanics with constraints;
must find the unconstrained set of circuit variables
--find a Hamiltonian/Lagrangian from which these classical
equations of motion arise
--then, quantize!

NB: no BCS theory, no microscopics this is phenomenological,
But based on sound general principles.
First, the graph formalism (6 steps)
small
1. Identify a tree of the graph maximal subgraph containing
all nodes and no loops
graph
tree
Branches not in tree are called chords; each chord completes a loop
graph formalism, continued
small
graph
2. Enumerate the loops 1,2,F, assign each a direction
3. Enumerate the branches 1,2,B, assign each a direction
4. Construct fundamental loop matrix F
ij
, i=1..F, j=1..B
Branch in loop, same direction
Branch in loop, opposite direction
Branch not in loop
graph formalism, continued
small
graph
Branch in loop, same direction
Branch in loop, opposite direction
Branch not in loop
NB: numbering of branches and loops can always
be organized so that F
(L)
has the form

(each chord defines one loop)
Define also cutset matrix:
F = -M
T
M
Circuit equations in the graph formalism:
small
Kirchhoffs current laws:
Kirchhoffs voltage laws:
V: branch voltages
I: branch currents
u: external fluxes threading
loops
graph formalism, continued
small
5. Identify submatrices of F:
No C chords.
graph formalism, continued
small
e.g.,
NB: this introduces
submatix of F labeled
by branch type
graph formalism, continued
small
6. Write down inductance matrix L and capacitance matrix C.
NB: inductances are both in the tree (K) and among the chords (L),
L should be partitioned into these submatrices:
With all this, the equation of motion:
small
The tricky part: what are the independent degrees of freedom?


If there are no capacitor-only loops (i.e., every loop has an inductance),


then the independent variables are just the Josephson phases, and the
capacitor phases (time integral of the voltage):
the equation of motion (continued):
small
All are complicated but straightforward functions of
the topology (F matrices) and the inductance matrix
the equation of motion (continued):
small
The lossless parts of this equation arise from a simple Hamiltonian:
the equation of motion (continued):
small
The lossy parts of this equation arise from a bath Hamiltonian,
Via a Caldeira-Leggett treatment:
Connecting Cadeira Leggett to circuit theory:
small
Overview of what weve accomplished:
small
We have a systematic derivation of a general
system-bath Hamiltonian. From this we can proceed to obtain:
system master equation
spin-boson approximation (two level)
Born-Markov approximation -> Bloch Redfield theory
golden rule (decay rates)
leakage rates
For example:
Application to Delft qubit
small
Gives particle in 5-dimensional potential coupled to environment
--reducible to usual two-level description; exposes approximations
Application to Delft qubit (cont.)
small
Previous work (Wilhelm, van der Wal, ) quantum theory as particle
in 2D potential

Present approach: 5 JJs, so 5D quantum mechanics?

Relation between these?

5D theory to 2D theory:
small
1. One of measuring SQUID degrees of freedom is heavily damped
by external impedance, not really quantum mechanical
2. 4D theory is really right! But, because effective inductances of the
SQUID and qubit are fairly small,
3. the particle is fairly tighly confined to a 2D sub-plane of the 4D
space.
4. ALSO this plane is not quite flat, it is slightly warped because of
the nonlinearity of the qubit potential.
The warping is very important,
because it is the only thing that
couples the 2D particle to the
external impedance tells us
the decoherence operator
of the qubit.
Overview:
small
1. A user friendly procedure: automates the assessment of
different circuit designs
2. Gives some new views of existing circuits and their analysis
3. A meta-theory aids the development of approximate theories
at many levels
4. BUT it is the orthodox theory of decoherence exotic effects
like nuclear-spin dephasing not captured by this analysis.
New application:
May, 2004, cond-mat
Chioescu, Nakamura, and Mooij, Science (2004).
--intended to have top-bottom symmetry.
--observed departure from symmetry is very large. Why?
Ideal circuit:
Actual circuit, with shadow effects:
Blue=bottom
Red=top layer
Our circuit model:
Formal network graph:
Actual external impedance:
Bottom/top effect causes big asymmetry in I
B
:
-- positive and negative bias asymmetric
--line of least decoherence is at large positive bias
-- 3uA optimal point very close to value seen experimentally
T1 & T2 times along line of minimum qubit frequency
good agreement with measurement
Conclusions:
small
A user friendly procedure: automates the assessment of
different circuit designs
Gives some new views of existing circuits and their analysis
Aided interpretation of latest Delft qubit, guiding design of
new IBM qubits
A meta-theory aids the development of approximate theories
at many levels
BUT it is the orthodox theory of decoherence exotic effects
like nuclear-spin dephasing not captured by this analysis.
THANKS to Guido Burkard and Roger Koch at IBM Watson,
and to TU Delft hospitality!