PHYSICAL REVIEW A VOLUME 57, NUMBER 1 JANUARY 1998

Quantum computation with quantum dots

Daniel Loss1,2,* and David P. DiVincenzo1,3,†
1
Institute for Theoretical Physics, University of California, Santa Barbara, Santa Barbara, California 93106-4030
2
Department of Physics and Astronomy, University of Basel, Klingelbergstrasse 82, 4056 Basel, Switzerland
3
IBM Research Division, T.J. Watson Research Center, P.O. Box 218, Yorktown Heights, New York 10598
~Received 9 January 1997; revised manuscript received 22 July 1997!
We propose an implementation of a universal set of one- and two-quantum-bit gates for quantum compu-
tation using the spin states of coupled single-electron quantum dots. Desired operations are effected by the
gating of the tunneling barrier between neighboring dots. Several measures of the gate quality are computed
within a recently derived spin master equation incorporating decoherence caused by a prototypical magnetic
environment. Dot-array experiments that would provide an initial demonstration of the desired nonequilibrium
spin dynamics are proposed. @S1050-2947~98!04501-6#

PACS number~s!: 03.67.Lx, 89.70.1c, 75.10.Jm, 89.80.1h

I. INTRODUCTION namics of magnetic nanosystems and could point the way

towards more extensive studies to explore the large-scale
The work of the past several years has greatly clarified quantum dynamics envisioned for a quantum computer.
both the theoretical potential and the experimental challenges
of quantum computation @1#. In a quantum computer the II. QUANTUM-DOT IMPLEMENTATION
state of each bit is permitted to be any quantum-mechanical OF TWO-QUBIT GATES
state of a qubit ~quantum bit, or two-level quantum system!.
Computation proceeds by a succession of ‘‘two-qubit quan- In this paper we develop a detailed scenario for how
tum gates’’ @2#, coherent interactions involving specific pairs quantum computation may be achieved in a coupled
of qubits, by analogy to the realization of ordinary digital quantum-dot system @7#. In our model the qubit is realized as
computation as a succession of Boolean logic gates. It is now the spin of the excess electron on a single-electron quantum
understood that the time evolution of an arbitrary quantum dot; see Fig. 1. We introduce here a mechanism for two-
state is intrinsically more powerful computationally than the qubit quantum-gate operation that operates by a purely elec-
evolution of a digital logic state ~the quantum computation
can be viewed as a coherent superposition of digital compu-
tations proceeding in parallel!.
Shor has shown @3# how this parallelism may be exploited
to develop polynomial-time quantum algorithms for compu-
tational problems, such as prime factoring, which have pre-
viously been viewed as intractable. This has sparked inves-
tigations into the feasibility of the actual physical
implementation of quantum computation. Achieving the con-
ditions for quantum computation is extremely demanding,
requiring precision control of Hamiltonian operations on
well-defined two-level quantum systems and a very high de-
gree of quantum coherence @4#. In ion-trap systems @5# and
cavity quantum electrodynamic experiments @6#, quantum
computation at the level of an individual two-qubit gate has FIG. 1. ~a! Schematic top view of two coupled quantum dots
been demonstrated; however, it is unclear whether such labeled 1 and 2, each containing one excess electron (e) with spin
1/2. The tunnel barrier between the dots can be raised or lowered by
atomic-physics implementations could ever be scaled up to
setting a gate voltage ‘‘high’’ ~solid equipotential contour! or
do truly large-scale quantum computation, and some have
‘‘low’’ ~dashed equipotential contour!. In the low state virtual tun-
speculated that solid-state physics, the scientific mainstay of
neling ~dotted line! produces a time-dependent Heisenberg ex-
digital computation, would ultimately provide a suitable change J(t). Hopping to an auxiliary ferromagnetic dot ~FM! pro-
arena for quantum computation as well. The initial realiza- vides one method of performing single-qubit operations. Tunneling
tion of the model that we introduce here would correspond to (T) to the paramagnetic dot ~PM! can be used as a POV read out
only a modest step towards the realization of quantum com- with 75% reliability; spin-dependent tunneling ~through ‘‘spin
puting, but it would at the same time be a very ambitious valve’’ SV! into dot 3 can lead to spin measurement via an elec-
advance in the study of controlled nonequilibrium spin dy- trometer E. ~b! Proposed experimental setup for initial test of swap-
gate operation in an array of many noninteracting quantum-dot
pairs. The left column of dots is initially unpolarized, while the
*Electronic address: loss@ubaclu.unibas.ch right one is polarized; this state can be reversed by a swap operation
†
Electronic address: divince@watson.ibm.com @see Eq. ~31!#.

1050-2947/98/57~1!/120~7!/$15.00 57 120 © 1998 The American Physical Society

57 QUANTUM COMPUTATION WITH QUANTUM DOTS
trical gating of the tunneling barrier between neighboring
quantum dots rather than by spectroscopic manipulation as in
other models. Controlled gating of the tunneling barrier be-
tween neighboring single-electron quantum dots in patterned
two-dimensional electron-gas structures has already been
achieved experimentally using a split-gate technique @8#. If
the barrier potential is ‘‘high,’’ tunneling is forbidden be-
tween dots and the qubit states are held stably without evo-
lution in time (t). If the barrier is pulsed to a ‘‘low’’ voltage,
the usual physics of the Hubbard model @9# says that the
spins will be subject to a transient Heisenberg coupling,
H s ~ t ! 5J ~ t ! SW 1 •SW 2 ,
where J(t)54t 20 (t)/u is the time-dependent exchange con-
stant @10# that is produced by the turning on and off of the
tunneling matrix element t 0 (t). Here u is the charging en-
ergy of a single dot and SW i is the spin-1/2 operator for dot i.
Equation ~1! will provide a good description of the
quantum-dot system if several conditions are met. ~i! Higher-
lying single-particle states of the dots can be ignored; this
requires DE@kT, where DE is the level spacing and T is the
temperature. ~ii! The time scale t s for pulsing the gate po-
tential low should be longer than \/DE in order to prevent
transitions to higher orbital levels. ~iii! u.t 0 (t) for all t; this
is required for the Heisenberg exchange approximation to be
accurate. ~iv! The decoherence time G 21 should be much
longer than the switching time t s . Much of the remainder of
the paper will be devoted to a detailed analysis of the effect
of a decohering environment. We expect that the spin-1/2
degrees of freedom in quantum dots should generically have
longer decoherence times than charge degrees of freedom
since they are insensitive to any environmental fluctuations
of the electric potential. However, while charge transport in
such coupled quantum dots has received much recent atten-
tion @11,8#, we are not aware of investigations on their non-
equilibrium spin dynamics as envisaged here. Thus we will
carefully consider the effect of magnetic coupling to the en-
vironment.
If G 21 is long, then the ideal of quantum computing may
be achieved, wherein the effect of the pulsed Hamiltonian is
to apply a particular unitary time evolution operator U s (t)
5Texp$2i*t0Hs(t8)dt8% to the initial state of the two spins:
u C(t) & 5U s u C(0) & . The pulsed Heisenberg coupling leads
to a special form for U s : For a specific duration t s of the
spin-spin coupling such that * dtJ(t)5J 0 t s 5 p (mod2 p )
@12#, U s (J 0 t s 5 p )5U sw is the ‘‘swap’’ operator: If u i j & la-
bels the basis states of two spins in the S z basis with i, j
50,1, then U sw u i j & 5 u ji & . Because it conserves the total an-
gular momentum of the system, U sw is not by itself sufficient
to perform useful quantum computations, but if the interac-
tion is pulsed on for just half the duration, the resulting
square root of the swap operator is very useful as a funda-
mental quantum gate: For instance, a quantum XOR gate is
obtained by a simple sequence of operations
z z z
U XOR 5e i ~ p /2 ! S 1 e 2i ~ p /2 ! S 2 U 1/2
sw e
i p S 1 1/2
U sw ,
z
where e i p S 1 , etc., are single-qubit operations only, which can
be realized, e.g., by applying local magnetic fields ~see Sec.
121
III B! @13#. It has been established that XOR along with
single-qubit operations may be assembled to do any quantum
computation @2#. Note that the XOR of Eq. ~2! is given in the
basis where it has the form of a conditional phase-shift op-
eration; the standard XOR is obtained by a simple basis
change for qubit 2 @2#.
III. MASTER EQUATION
We will now consider in detail the nonideal action of the
swap operation when the two spins are coupled to a magnetic
environment. A master equation model is obtained that ex-
~1! plicitly accounts for the action of the environment during
switching, to our knowledge, the first treatment of this effect.
We use a Caldeira-Leggett–type model in which a set of
harmonic oscillators are coupled linearly to the system spins
by H int 5l ( i51,2SW i •bW i . Here b ij 5 ( a g ai j (a a ,i j 1a †a ,i j ) is a
fluctuating quantum field whose free motion is governed by
the harmonic-oscillator Hamiltonian H B 5 ( v ai j a †a ,i j a a ,i j ,
where a †a ,i j (a a ,i j ) are bosonic creation ~annihilation! opera-
tors ~with j5x,y,z) and v ai j are the corresponding frequen-
cies with spectral distribution function
J i j ( v )5 p ( a (g ai j ) 2 d ( v 2 v a ) @14#. The system and environ-
ment are initially uncorrelated with the latter in thermal equi-
librium described by the canonical density matrix r B with
temperature T. We assume for simplicity that the environ-
ment acts isotropically and is equal and independent on both
dots. We do not consider this to be a microscopically accu-
rate model for these as-yet-unconstructed quantum-dot sys-
tems, but rather as a generic phenomenological description
of the environment of a spin, which will permit us to explore
the complete time dependence of the gate action on the
single coupling constant l and the controlled parameters of
H s (t) @15#.
A. Swap gate
The quantity of interest is the system density matrix
r (t)5TrB r̄ (t), which we obtain by tracing out the environ-
ment degrees of freedom. The full density matrix r̄ itself
obeys the von Neumann equation
r̄˙ ~ t ! 52i @ H, r̄ # [2iL r̄ , ~3!
where
L5Ls ~ t ! 1Lint 1LB ~4!
denotes the Liouvillian @16# corresponding to the full Hamil-
tonian
H5H s ~ t ! 1H int 1H B . ~5!
Our goal is to find the linear map ~superoperator! V(t) that
connects the input state of the gate r 0 5 r (t50) with the
output state r (t) after time t. t s has elapsed, r (t)5V(t) r 0 .
V(t) must satisfy three physical conditions: ~i! trace preser-
~2! vation Trs Vr 51, where Trs denotes the system trace; ~ii!
Hermiticity preservation (Vr ) † 5Vr ; and ~iii! complete posi-
tivity, (V ^ 1 B ) r̄ >0. Using the Zwanzig master equation ap-
proach @16#, we sketch the derivation for V in the Born and
122 DANIEL LOSS AND DAVID P. DIVENCENZO 57

Markov approximations, which respects these three condi- @17,16#!. We also note that the above Born and Markov ap-
tions. The situation we analyze here is unusual in that H s is proximations could also be introduced in the master equation
explicitly time dependent and changes abruptly in time. It is in the more usual differential-integral representation. How-
this fact that requires a separate treatment for times t< t s and ever, it is well known from studies in noninteracting spin
t. t s . To implement this time scale separation and to pre- problems @18# that in this case the resulting propagator is in
serve positivity it is best to start from the exact master equa- general no longer completely positive.
tion in pure integral form Next, we evaluate the above superoperators more explic-
itly, obtaining
r ~ t ! 5Us ~ t,0! r 0 2 E sE t s
d t Us ~ t, s ! M~ s , t ! r ~ t ! ,
E
d
ts
(i
0 0
~6! K2 r 5 ~ G1iD ! d t @ SW i ~ t s ! ,SW i ~ t ! r # 1H.c., ~13!
0

where

Ui ~ t,t 8 ! 5Texp 2i H E t

t8
J
d t Li ~ t ! , ~7!
K3 r 5G 3 r 22 S (i SW i r •SW i D , ~14!

where i5s, B, int, or q. Here q indicates the projected where in the commutator in Eq. ~13! a dot product is under-
Liouvillian stood between the vector parts of the two factors, and where
G,D are real and given by
Lq 5 ~ 12 P ! L5 ~ 12 r B TrB ! L. ~8!

Also, the ‘‘memory kernel’’ is G5

l2
p
E E
0
`
dt
0
`
d v J ~ v ! cos~ v t ! coth S D
v
2k B T
, ~15!

M~ s , t ! 5TrB Lint Uq ~ s , t ! Lint r B . ~9!

We solve Eq. ~6! in the Born approximation and for t@ t s .

To this end the time integrals are split up into three parts: ~i!
D5
l2
p
E E
0
`
dt
`

0
d v J ~ v ! sin~ v t ! . ~16!

0< t < s < t s ,t, ~ii! 0< t < t s < s ,t, and ~iii! 0< t s < t
< s ,t. Keeping only leading terms in t s , we retain the
contribution from interval ~ii! as it is proportional to t s ,
whereas we can drop interval ~i!, which leads to higher-order
terms. However, note that terms containing J 0 t s must be
kept to all orders @12#. Interval ~iii! is independent of t s .
Rewriting the expressions and performing a Born ap-
proximation ~i.e., keeping only lowest-order terms in l 2 )
with subsequent Markov approximation we find, for t> t s ,
V~ t ! 5e 2 ~ t2 t s ! K3 Us ~ t s ! ~ 12K2 ! ,
where Ui ( t s )5Ui ( t s ,0), K2 describes the effect of the envi-
ronment during the switching,
In our model, the transverse and longitudinal relaxation or
decoherence rates of the system spins are the same and given
by G. For instance, for Ohmic damping with J( v )5 h v , we
get G5l 2 h k B T and D5l 2 h v c / p , with v c some high-
frequency cutoff. Requiring for consistency that G t s ,D t s
!1, we find that K2 is in fact a small correction. However,
we emphasize again that, to our knowledge, this is the first
time that any analysis of this K2 term, describing the action
of the environment during the finite time that the system
~10! Hamiltonian is switched on, has been given.
For further evaluation of V we adopt a matrix representa-
tion, defined by

K2 5U†s ~ t s ! E E0
ts
dt
`

0
dt TrB Lint Us ~ t ! UB ~ t !
Vab u cd 5 ~ e ab ,Ve cd ! [Trs e †ab Ve cd , ~17!

where $ e ab u a,b51, . . . ,4% is an orthonormal basis, i.e.,

3Lint r B Us ~ t s 2 t ! , ~11! (e ab ,e cd )5 d ac d bd . In this notation we then have

while
r ~ t ! ab 5 ( Vab u cd ~ r 0 ! cd , ~18!
K3 5 E 0
`
dt TrB Lint UB ~ t ! Lint r B ~12!
c,d

with V being a 16316 matrix.

is independent of H s . We also note that Us (12K2 ) has a Note that K2,3 and Us are not simultaneously diagonal.
simple interpretation as being the ‘‘transient contribution’’ However, since K3 (1,SW i )52G(0,SW i ) we see that exp$2(t
that changes the initial value r 0 at t50 to Us ( t s )(1 2ts)K3 % is diagonal in the ‘‘polarization basis’’ $ e ab p
2K2 ) r 0 at t5 t s . We show in the Appendix that, to leading 5e a e b ;e 1, . . . ,45(1/A2, A2S i , A2S i , A2S i ),i51,2% , while
1 2 i x y z
order, our superoperator V indeed satisfies all three condi- Ls and thus Us are diagonal in the ‘‘multiplet basis’’ $ e ab m

5 u a &^ b u , a , b 51, . . . ,4; u 1 & 5( u 01& 2 u 10& )/ A2, u 2 & 5( u 01&

tions stated above, in particular complete positivity. Such a
proof for spins with an explicit time-dependent and direct
1 u 10& )/ A2, u 3 & 5 u 00& , u 4 & 5 u 11& % , with
interaction ~1! is not simply related to the case of a master
equation for noninteracting spins ~and without explicit time m m
dependence! considered in the literature ~see, for example, Us ~ t ! ab u a 8 b 8 5 d aa 8 d bb 8 e 2it ~ E a 2E b ! , ~19!
57 QUANTUM COMPUTATION WITH QUANTUM DOTS 123

where E m1 523J 0 /4 and E 2,3,4 5J 0 /4 are the singlet and trip-

E
m 2
ts
let eigenvalues. Finally, K2 is most easily evaluated also in
0
s 5
dtH H ( v i t s S zi ,
i51
~23!
the multiplet basis; after some calculation we find that K2
5Kd2 2Knd 2 , with
with v i 5g m B H zi , where we assume that the H field acting
on spin i is along the z axis. The calculation proceeds along
~ Kd2 ! ab u g d 5 ( @ d ag ^ d u SW iu a 8 & • ^ a 8u SW iu b & k a*8a 8ud b
i, a 8
the same line as the one described above: Just as in Eq. ~10!,
the expression obtained for the superoperator is
1 d b d ^ a u SW i u a 8 & • ^ a 8 u SW i u g & k a 8 a 8 u ga # , ~20! VH ~ t ! 5e 2 ~ t2 t s ! K3 U H
s ~ t s !~ 12K2 ! . ~24!
H

~ Knd
2 ! ab u g d 5 (i ^ a u SW iu g & • ^ d u SW iu b & @ k abug d 1 ~ k baud g ! * # . K3 is exactly the same as before, Eq. ~14!. UH s ( t s ) is again
given by Eq. ~7! with the modification that the Liouvillian
~21! @see Eq. ~4!# corresponding to the magnetic-field Hamil-
tonian of Eq. ~23! is used rather than that for the exchange
Here Hamiltonian H s @Eqs. ~5! and ~1!#. The explicit matrix rep-

E
resentation is
ts

S D
m m m m
k ab u g d 5 ~ G1iD ! e i ~ E d 2E b ! t s d t e i ~ E a 2E g ! t 2
0

1
s ~ t s ! …rs u r 8 s 8 5 d rr 8 d ss 8 exp
„U H 2i ( ~ E ir 2E is ! t s
i51
.
5 @ Gc d b 2Ds d b 1i ~ Gs d b 1Dc d b !# ~25!
2 v ag
3 @ s ag 1i ~ 12c ag !# , Here we are employing another basis, the S z basis for the
two spins $ e zrs 5 u r &^ s u , r,s51,2,3,4; u s & 5 u 00& , u 01& , u 10& ,
c i j 5cos~ t s v i j ! , s i j 5sin~ t s v i j ! , v i j 5E mi 2E mj . u 11& % . The energies are

Using the above matrix notation, we can write explicitly v1

$ E 1r % 5 $ E 11,2,3,4% 5 $ 1,1,21,21 % ,
2
Vab u g d 5 (
a,b, a 8 , b 8
~ C ab u ab ! * ~ e 2 ~ t2 t s ! K3 ! ab u ab
v2
$ E 2r % 5 $ E 21,2,3,4% 5 $ 1,21,1,21 % . ~26!
m m 2
2i t s ~ E 2E !
3C ab u a 8 b 8 e a8 b8 ~ 12K2 ! a 8 b 8 u g d ,
~22! The KH 2 calculation also proceeds as before @see Eq. ~13!#
using the new Hamiltonian; the result is KH 2 5K2
H,d

where C ab u ab 5(e ab
p m
,e ab ) is the unitary basis change be- 2K2 , with
H,nd

tween the polarization and the multiplet basis.

B. One-bit gates
~ KH,d
2 ! rs u tu 5 ( @ d rt ^ u u SW i u r 8 & • ^ r 8 u SW i u s & ~ k r 8 r 8 u us ! *
i

i,r 8
We now repeat the preceding analysis for single-qubit ro- 1 d su ^ r u SW i u r 8 & • ^ r 8 u SW i u t & k r 8 r 8 u tr # ,
i
z ~27!
tations such as e i( p /2)S i as required in Eq. ~2!. Such rotations
can be achieved if a magnetic field H W i could be pulsed ex-
clusively onto spin i, perhaps by a scanning-probe tip. An ~ KH,nd
2 ! rs u tu 5 (i ^ r u SW iu t & • ^ u u SW iu s & @ k rsi utu 1 ~ k sri uut ! * # .
alternative way, which would become attractive if further
~28!
advances are made in the synthesis of nanostructures in mag-
netic semiconductors @19#, is to use, as indicated in Fig. 1~a!, Here
an auxiliary dot ~FM! made of an insulating, ferromagneti-
cally ordered material that can be connected to dot 1 ~or dot
2! by the same kind of electrical gating as discussed above
@8#. If the the barrier between dot 1 and dot FM were low-
i
k rs u tu 5 ~ G1iD ! e
i
i ~ E u 2E s ! t s
i
E 0
ts i i
d t e i ~ E r 2E t ! t

ered so that the electron’s wave function overlaps with the

1
magnetized region for a fixed time t s , the Hamiltonian for 5 @ Gc ius 2Ds ius 1i ~ Gs ius 1Dc ius !#
the qubit on dot 1 will contain a Zeeman term during that 2 v irt
time. For all earlier and later times the magnetic field seen by
the qubit should be zero; any stray magnetic field from the 3 @ s irt 1i ~ 12c irt !# , ~29!
dot FM at neighboring dots 1, 2, etc., could be made small
by making FM part of a closure domain or closed magnetic c ki j 5cos~ t s v ki j ! , s ki j 5sin~ t s v ki j ! , v ki j 5E ki 2E kj .
circuit.
In either case, the spin is rotated and the corresponding The E k ’s are from Eq. ~26!. Finally, the explicit matrix form
Hamiltonian is given by for VH may be written
124
FIG. 2. ~a! Swap polarization s[2 ^ S z1 (t) &
@see Eq. ~31!#, gate
fidelity F, and gate purity P vs Gt for ‘‘swap’’ using parameters
J 0 t s 5 p , G t s 50.017, and D t s 520.0145. ~b! Same for XOR ob-
tained using the four operations in Eq. ~2! ~the final two single-spin
operations done simultaneously!. The same parameters and scales
as in ~a! are used; the pulse-to-pulse time is taken to be 3 t s . Gt is
measured from the end of the fourth pulse.
(
~ e 2 ~ t2 t s ! K3 ! ab u ab ~ D rs u ab ! *
H
Vab u a 8 b 8 5
8 8
r,s,r ,s
3exp 2i S 2
( t s~ E ir 2E is !
i51
D ~ 12KH
3D r 8 s 8 u a 8 b 8 ,
where D rs u ab 5(e zrs ,e ab
p
) is now the unitary basis change be-
tween the S z basis and the polarization basis.
C. Numerical study for swap gate and XOR gate
Having diagonalized the problem, we can now calculate
any system observable; the required matrix calculations are
involved and complete evaluation is done with
MATHEMATICA. We will consider three parameters (s, F, and
P in Fig. 2! relevant for characterizing the gate operation.
We first perform this analysis for the swap operation intro-
duced above.
The swap operation would provide a useful experimental
test for the gate functionality: Let us assume that at t50 spin
2 is ~nearly! polarized, say, along the z axis, while spin 1 is
~nearly! unpolarized, i.e., r 0 5(112S z2 )/4. This can be
achieved, e.g., by selective optical excitation or by an ap-
plied magnetic field with a strong spatial gradient. Next we
apply a swap operation by pulsing the exchange coupling
such that J 0 t s 5 p and observe the resulting polarization of
spin 1 described by
1 quantum-dot quantum computation relies on simultaneous
^ S z1 ~ t ! & 5 V~ t ! 41u 14 ,
2
where V is evaluated in the polarization basis. After time t s
spin 1 is almost fully polarized ~whereas spin 2 is now un-
polarized! and, due to the environment, decays exponentially
with rate of order G. To make the signal ~31! easily measur-
able by conventional magnetometry, we can envisage a setup
DANIEL LOSS AND DAVID P. DIVENCENZO 57
consisting of a large array of identical, noninteracting pairs
of dots as indicated in Fig. 1~b!.
To further characterize the gate performance we follow
Ref. @20# and calculate the gate fidelity
F5 ^ c 0 u U† ( t s ) r (t) u c 0 & and the gate purity P5Tr s @ r (t) # 2 ,
where the overbar means an average over all initial system
states u c 0 & . Expressing V in the multiplet basis and using
trace and Hermiticity preservation we find
1 1
F ~ t ! 5 1 Re
6 24 F( a
Vaa u aa 1 (
a,b
Vab u ab e i t ~ E
s
m m
a 2E b !
G ,
~32!
P~ t !5
1
24 (
i,k,k 8
F u Vkk 8 u ii u 2 1 (j ~ Vkk 8uii V*kk 8u j j 1 u Vkk 8ui j u 2 ! G
~33!
@in fact, the expression for P(t) holds in any basis#. Evalu-
ations of these functions for specific parameter values are
shown in Fig. 2. For the parameters shown, the effect of the
environment during the switching, i.e., K2 in Eq. ~10!, is on
the order of a few percent.
The dimensionless parameters used here would, for ex-
ample, correspond to the following actual physical param-
eters: If an exchange constant J 0 580 m eV'1 K were
achievable, then pulse durations of t s '25 ps and decoher-
2 ! rs u r 8 s 8
ence times of G 21 '1.4 ns would be needed; such param-
eters, and perhaps much better, are apparently achievable in
solid-state spin systems @19#.
~30! As a final application, we calculate the full XOR by ap-
plying the corresponding superoperators in the sequence as-
sociated with the one on the right-hand side of Eq. ~2!. We
use the same dimensionless parameters as above, and as be-
fore we then calculate the gate fidelity and the gate purity.
Some representative results of this calculation are plotted in
the inset of Fig. 2~b!. To attain the p /2 single-bit rotations of
Eq. ~2! in a t s of 25 ps would require a magnetic field H
'0.6 T, which would be readily available in the solid state.
IV. DISCUSSION
As a final remark about the decoherence problem, we note
that the parameters that we have chosen in the presentation
of our numerical work, which we consider to be realistic for
known nanoscale semiconductor materials, of course fall far,
far lower than the 0.999 99 levels that are presently consid-
ered desirable by quantum-computation theorists @1#; still,
the achievement of even much lesser quality quantum gate
operation would be a tremendous advance in the controlled,
nonequilibrium time evolution of solid-state spin systems
and could point the way to the devices that could ultimately
be used in a quantum computer. Considering the situation
more broadly, we are quite aware that our proposal for
~31!
further advances in the experimental techniques of semicon-
ductor nanofabrication, magnetic semiconductor synthesis,
single electronics, and perhaps in scanning-probe techniques.
Still, we also feel strongly that such proposals should be
developed seriously, and taken seriously, at present since we
believe that many aspects of the present proposal are testable
in the not-too-distant future. This is particularly so for the
57 QUANTUM COMPUTATION WITH QUANTUM DOTS 125

demonstration of the swap action on an array of dot pairs. Obviously, such a state is achieved if the system is cooled
Such a demonstration would be of clear interest not only for sufficiently in a uniform applied magnetic field; acceptable
quantum computation, but would also represent a technique spin polarizations of electron spins are readily achievable at
for exploring the nonequilibrium dynamics of spins in quan- cryogenic temperatures. If a specific arrangement of up and
tum dots. down spins were needed as the starting state, these could be
To make the quantum-dot idea a complete proposal for created by a suitable application of the reverse of the spin
quantum computation, we need to touch on several other valve measurement apparatus.
important features of quantum-computer operation. As our
guideline we follow the five requirements laid out by one of ACKNOWLEDGMENTS
us @4#: ~i! identification of well-defined qubits, ~ii! reliable
state preparation, ~iii! low decoherence, ~iv! accurate quan- We are grateful to D. D. Awschalom, H.-B. Braun, T.
tum gate operations, and ~v! strong quantum measurements. Brun, and G. Burkard for useful discussions. This research
Items ~i!, ~iii!, and ~iv! have been very thoroughly considered was supported in part by the National Science Foundation
above. We would now like to propose several possible under Grant No. PHY94-07194.
means by which requirements ~ii! and ~v!, for state prepara-
tion ~read in! and quantum measurement ~read out!, may be APPENDIX: COMPLETE POSITIVITY
satisfied. OF TIME-EVOLUTION SUPEROPERATOR V
One scheme for qubit measurement that we suggest in-
volves a switchable tunneling @ T in Fig. 1~a!# into a super- Here we sketch the proof that the superoperator V in Eq.
cooled paramagnetic dot ~PM!. When the measurement is to ~10! is completely positive. We analyze the K3 term first. We
be performed, the electron tunnels ~this will be real tunnel- write
ing, not the virtual tunneling used for the swap gate above!
into PM, nucleating from the metastable phase a ferromag-
netic domain whose magnetization direction could be mea- N→`
t
e 2 t K3 5 lim 12 K3
N S D N
. ~A1!

sured by conventional means. The orientation ( u , f ) of this

magnetization vector is a ‘‘pointer’’ that measures the spin It is sufficient to prove that the infinitesimal operator is com-
direction; it is a generalized measurement in which the mea- pletely positive. It is straightforward to show, using Eq. ~14!,
surement outcomes form a continuous set rather than having that

S D
two discrete values. Such a case is covered by the general
formalism of positive-operator-valued ~POV! measurements t
12 K3 r 5Z †3 • r Z 3 1O„~ t /N ! 2 …. ~A2!
@21#. If there is no magnetic anisotropy in dot PM, then N
symmetry dictates that the positive measurement operators
would be projectors into the overcomplete set of spin-1/2 Here Z 3 is the seven-component vector operator

S At D
coherent states 6
t
u u Z 3 5 12 ( B †k B k ,
2N k51 N
B , ~A3!
u u , f & 5cos u 0 & 1e i f sin u 1 & . ~34!
2 2
where
A 75%-reliable measurement of spin up and spin down is
obtained if the magnetization direction ( u , f ) in the upper B5 ~ B 1 , . . . ,B 6 ! 5 A2G ~ SW 1 ,SW 2 ! . ~A4!
hemisphere is interpreted as up and in the lower hemisphere
as down; this is so simply because Note that for this case B †k 5B k and ( 6k51 B †k B k 53G.
We recall that it is easy to prove that any superoperator S
1
2p
EU
3
dV u ^ 0zu , f & z2 5 .
4
~35!
of the form

Sr 5Z † r Z ~A5!
Here U denotes integration over the upper hemisphere and
as in the first term of Eq. ~A2! is completely positive. Indeed,
2 p is the normalization constant for the coherent states.
considering its action on any state vector of the system plus
Another approach which would potentially give a 100%
environment f and taking a positive r we get
reliable measurement requires a spin-dependent, switchable
‘‘spin valve’’ tunnel barrier ~SV! of the type mentioned, e.g., ~ f ,Sr f ! 5 ~ f ,Z † r Z f ! 5 ~ Z f , r Z f ! >0 ; f . ~A6!
in Ref. @22#. When the measurement is to be performed, SV
is switched so that only an up-spin electron passes into semi- Next we consider the 12K2 term of Eq. ~10!. Starting
conductor dot 3. Then the presence of an electron on 3, mea- from Eq. ~13!, we put this term in a form corresponding to
sured by electrometer E, would provide a measurement that the completely positive form ~A5!. We find
the spin had been up. It is well known now how to create
nanoscale single-electron electrometers with exquisite sensi- ~ 12K2 ! r 5Z †2 • r Z 2 1O„l 4 , t 2s , ~ l 2 t s ! 2 …, ~A7!
tivity ~down to 1028 of one electron! @23#.
We need only discuss the state-preparation problem with Z 2 being the vector operator
briefly. For many applications in quantum computing, only a
simple initial state, such as all spins up, needs to be created. Z 2 5 ~ 11Y † •X † ,X2Y † ! , ~A8!
126 DANIEL LOSS AND DAVID P. DIVENCENZO 57

with that 12K2 is completely positive up to the order of accuracy

discussed in the text.
Finally, we note that the other two general conditions for
X52 ~ G1iD ! „SW 1 ~ t s ! ,SW 2 ~ t s ! …, ~A9! a physical superoperator also follow immediately: Trace
preservation of V follows from the fact that a Liouvillian L

Y5 E 0
ts
d t „SW 1 ~ t ! ,SW 2 ~ t ! …. ~A10!
appears to the left in the basic equations for K2 , Eq. ~11!, and
K3 , Eq. ~12!. Trace preservation is also reflected in the fact
that Z 2 •Z †2 51 and Z 3 •Z †3 51 to leading order. The form
~A5! also obviously preserves Hermiticity of the density op-
So, from the same arguments as above, Eq. ~A7! establishes erator; this is also clear from the forms of Eqs. ~13! and ~14!.

@1# S. Lloyd, Science 261, 1589 ~1993!; C. H. Bennett, Phys. To- @12# We assume for simplicity that the shape of the applied pulse is
day 48„10…, 24 ~1995!; D. P. DiVincenzo, Science 269, 255 roughly rectangular with J 0 t s constant.
~1995!; A. Barenco, Contemp. Phys. 37, 375 ~1996!. @13# We note that explicitly U XOR 5 21 1S z1 1S z2 22S z1 S z2 , with
@2# A. Barenco et al., Phys. Rev. A 52, 3457 ~1995!. the corresponding XOR Hamiltonian * t0 dt 8 H XOR
@3# P. Shor, Proceedings of the 35th Annual Symposium on the 5 p @ 122S 1 22S 2 14S 1 S 2 # /4. An alternative way to achieve
z z z z

Foundations of Computer Science ~IEEE Press, Los Alamitos, the XOR operation is given by U XOR
1994!, p. 124. z z
21/2 2i ( p /2 ) S 1 z
5e i p S 1 U sw e U sw e i ( p /2) S 1 U sw
1/2
. This form has the po-
@4# D. P. DiVincenzo, Report No. cond-mat/9612126; in Mesos-
tential advantage that the single-qubit operations involve only
copic Electron Transport, Vol. 345 of NATO Advanced Study
spin 1.
Institute, Series E: Applied Sciences, edited by L. Sohn, L.
@14# A simple discussion of the consequences of decoherence mod-
Kouwenhoven, and G. Schoen ~Kluwer, Dordrecht, 1997!.
@5# J.-I. Cirac and P. Zoller, Phys. Rev. Lett. 74, 4091 ~1995!; J.-I. els of this type may be found in I. L. Chuang, R. Laflamme, P.
Cirac, T. Pellizzari, and P. Zoller, Science 273, 1207 ~1996!; Shor, and W. H. Zurek, Science 270, 1633 ~1995!.
C. Monroe et al., ibid. 75, 4714 ~1995!. @15# For a microscopic discussion of dissipation in quantum dots
@6# Q. A. Turchette et al., Phys. Rev. Lett. 75, 4710 ~1995!. concerning the charge degrees of freedom see, e.g., H. Schoe-
@7# There has been some earlier speculation on how coupled quan- ller and G. Schön, Phys. Rev. B 50, 18 436 ~1994!; Physica B
tum wells might be used in quantum-scale information pro- 203, 423 ~1994!.
cessing: see R. Landauer, Science 272, 1914 ~1996!; A. @16# E. Fick, G. Sauermann, and W. D. Brewer, Quantum Statistics
Barenco et al., Phys. Rev. Lett. 74, 4083 ~1995!. of Dynamic Processes, Springer Series in Solid-State Sciences,
@8# C. Livermore et al., Science 274, 1332 ~1996!; F. R. Waugh edited by H. K. V. Lotsch, M. Cardona, P. Fulde, K. v. Klitz-
et al., Phys. Rev. B 53, 1413 ~1996!; Phys. Rev. Lett. 75, 705 ing, and H.-J. Queisser, Vol. 86 ~Springer-Verlag, Berlin,
~1995!. 1990!.
@9# N. W. Ashcroft and N. D. Mermin, Solid State Physics ~Saun- @17# E. B. Davies, Quantum Theory of Open Systems ~Academic,
ders, Philadelphia, 1976!, Chap. 32. New York, 1976!.
@10# We can also envisage a superexchange mechanism to obtain a @18# M. Celio and D. Loss, Physica A 150, 769 ~1989!.
Heisenberg interaction by using three aligned quantum dots @19# S. A. Crooker et al., Phys. Rev. Lett. 77, 2814 ~1996!.
where the middle one has a higher-energy level ~by the amount @20# J. F. Poyatos, J.-I. Cirac, and P. Zoller, Phys. Rev. Lett. 78,
e ) such that the electron spins of the outer two dots are also 390 ~1997!; e-print quant-ph/9611013.
Heisenberg coupled, but now with the exchange coupling be- @21# A. Peres, Quantum Theory: Concepts and Methods ~Kluwer,
ing J54t 40 (1/e 2 u11/2e 3 ). Dordrecht, 1993!.
@11# L. I. Glazman and K. A. Matveev, Zh. Éksp. Teor. Fiz. 98, @22# G. Prinz, Phys. Today 45„4…, 58 ~1995!.
1834 ~1990! @Sov. Phys. JETP 71, 1031 ~1990!#; C. A. @23# M. Devoret, D. Estève, and Ch. Urbina, Nature ~London! 360,
Stafford and S. Das Sarma, Phys. Rev. Lett. 72, 3590 ~1994!. 547 ~1992!.