PHYSICAL REVIEW A, VOLUME 64, 052312

Measurement of qubits
Daniel F. V. James,1,* Paul G. Kwiat,2,3 William J. Munro,4,5 and Andrew G. White2,4
1
Theoretical Division T-4, Los Alamos National Laboratory, Los Alamos, New Mexico 87545
2
Physics Division P-23, Los Alamos National Laboratory, Los Alamos, New Mexico 87545
3
Department of Physics, University of Illinois, Urbana-Champaign, Illinois 61801
4
Department of Physics, University of Queensland, Brisbane, Queensland 4072, Australia
5
Hewlett-Packard Laboratories, Filton Road, Stoke Gifford, Bristol BS34 8QZ, United Kingdom
共Received 20 March 2001; published 16 October 2001兲
We describe in detail the theory underpinning the measurement of density matrices of a pair of quantum
two-level systems 共‘‘qubits’’兲. Our particular emphasis is on qubits realized by the two polarization degrees of
freedom of a pair of entangled photons generated in a down-conversion experiment; however, the discussion
applies in general, regardless of the actual physical realization. Two techniques are discussed, namely, a
tomographic reconstruction 共in which the density matrix is linearly related to a set of measured quantities兲 and
a maximum likelihood technique which requires numerical optimization 共but has the advantage of producing
density matrices that are always non-negative definite兲. In addition, a detailed error analysis is presented,
allowing errors in quantities derived from the density matrix, such as the entropy or entanglement of formation,
to be estimated. Examples based on down-conversion experiments are used to illustrate our results.

DOI: 10.1103/PhysRevA.64.052312 PACS number共s兲: 03.67.⫺a, 42.50.⫺p

I. INTRODUCTION trices with unit trace. The tomographically measured matri-

The ability to create, manipulate, and characterize quan-
tum states is becoming an increasingly important area of
physical research, with implications for areas of technology
such as quantum computing, quantum cryptography, and
communications. With a series of measurements on a large
enough number of identically prepared copies of a quantum
system, one can infer, to a reasonable approximation, the
quantum state of the system. Arguably, the first such experi-
mental technique for determining the state of a quantum sys-
tem was devised by George Stokes in 1852 关1兴. His famous
four parameters allow an experimenter to determine uniquely
the polarization state of a light beam. With the insight pro-
vided by nearly 150 years of progress in optical physics, we
can consider coherent light beams to be an ensemble of two-
level quantum mechanical systems, the two levels being the
two polarization degrees of freedom of the photons; the
Stokes parameters allow one to determine the density matrix
describing this ensemble. More recently, experimental tech-
niques for the measurement of the more subtle quantum
properties of light have been the subject of intensive inves-
tigation 共see Ref. 关2兴 for a comprehensive and erudite expo-
sition of this subject兲. In various experimental circumstances
it has been found reasonably straightforward to devise a
simple linear tomographic technique in which the density
matrix 共or Wigner function兲 of a quantum state is found from
a linear transformation of experimental data. However, there
is one important drawback to this method, in that the recov-
ered state might not correspond to a physical state because of
experimental noise. For example, density matrices for any
quantum state must be Hermitian, positive semidefinite ma-
*Corresponding author. Mailing address: Mail stop B-283, Los
Alamos National Laboratory, Los Alamos NM 87545. FAX: 共505兲
667-1931. Email address: dfvj@lanl.gov
ces often fail to be positive semidefinite, especially when
measuring low-entropy states. To avoid this problem the
‘‘maximum likelihood’’ tomographic approach to the estima-
tion of quantum states has been developed 关3–7兴. In this
approach the density matrix that is ‘‘mostly likely’’ to have
produced a measured data set is determined by numerical
optimization.
In the past decade several groups have successfully em-
ployed tomographic techniques for the measurement of
quantum mechanical systems. In 1990 Ashburn et al. re-
ported the measurement of the density matrix for the nine
sublevels of the n⫽3 level of hydrogen atoms formed fol-
lowing collision between H⫹ ions and He atoms, in condi-
tions of high symmetry which simplified the tomographic
problem 关8兴. Since then, in 1993 Smithey et al. made a ho-
modyne measurement of the Wigner function of a single
mode of light 关9兴. Other explorations of the quantum states
of single mode light fields have been made by Breitenbach
et al. 关10兴 and Wu et al. 关11兴. Other quantum systems whose
density matrices have been investigated experimentally in-
clude the vibrations of molecules 关12兴, the motion of ions
and atoms 关13,14兴, and the internal angular momentum quan-
tum state of the F⫽4 ground state of a cesium atom 关15兴.
The quantum states of multiple spin- 21 nuclei have been mea-
sured in the high-temperature regime using NMR techniques
关16兴, albeit in systems of such high entropy that the creation
of entangled states is necessarily precluded 关17兴. The mea-
surement of the quantum state of entangled qubit pairs, real-
ized using the polarization degrees of freedom of a pair of
photons created in a parametric down-conversion experi-
ment, was reported by us recently 关18兴.
In this paper we will examine in detail techniques for
quantum state measurement as it applies to multiple corre-
lated two-level quantum mechanical systems 共or ‘‘qubits’’ in
the terminology of quantum information兲. Our particular em-

1050-2947/2001/64共5兲/052312共15兲/$20.00 64 052312-1 ©2001 The American Physical Society

JAMES, KWIAT, MUNRO, AND WHITE
phasis is qubits realized via the two polarization degrees of
freedom of photons, data from which we use to illustrate our
results. However, these techniques are readily applicable to
other technologies proposed for creating entangled states of
pairs of two-level systems. Because of the central importance
of qubit systems to the emergent discipline of quantum com-
putation, a thorough explanation of the techniques needed to
characterize the qubit states will be of relevance to workers
in the various diverse experimental fields currently under
consideration for quantum computation technology 关19兴.
This paper is organized as follows. In Sec. II we explore the
analogy with the Stokes parameters, and how they lead natu-
rally to a scheme for measurement of an arbitrary number of
two-level systems. In Sec. III, we discuss the measurement
of a pair of qubits in more detail, presenting the validity
condition for an arbitrary measurement scheme and introduc-
ing the set of 16 measurements employed in our experi-
ments. Sec. IV deals with our method for maximum likeli-
hood reconstruction and in Sec. V we demonstrate how to
calculate the errors in such measurements, and how these
errors propagate to quantities calculated from the density
matrix.
II. THE STOKES PARAMETERS AND QUANTUM STATE
TOMOGRAPHY
As mentioned above, there is a direct analogy between the
measurement of the polarization state of a light beam and the
measurement of the density matrix of an ensemble of two-
level quantum mechanical systems. Here we explore this
analogy in more detail.
A. Single qubit tomography
The Stokes parameters are defined from a set of four in-
tensity measurements 关20兴: 共i兲 with a filter that transmits
50% of the incident radiation, regardless of its polarization;
共ii兲 with a polarizer that transmits only horizontally polarized
light; 共iii兲 with a polarizer that transmits only light polarized
at 45° to the horizontal; and 共iv兲 with a polarizer that trans-
mits only right-circularly polarized light. The number of
photons counted by a detector, which is proportional to the
classical intensity, in these four experiments is as follows:
N N
n 0⫽ 共 具 H 兩 ␳ˆ 兩 H 典 ⫹ 具 V 兩 ␳ˆ 兩 V 典 兲 ⫽ 共 具 R 兩 ␳ˆ 兩 R 典 ⫹ 具 L 兩 ␳ˆ 兩 L 典 兲 ,
2 2
n 1 ⫽N共 具 H 兩 ␳ˆ 兩 H 典 兲
N which may be written as follows:
⫽ 共 具 R 兩 ␳ˆ 兩 R 典 ⫹ 具 R 兩 ␳ˆ 兩 L 典 ⫹ 具 L 兩 ␳ˆ 兩 R 典 ⫹ 具 L 兩 ␳ˆ 兩 L 典 兲 ,
2 3
n 2 ⫽N共 具 D̄ 兩 ␳ˆ 兩 D̄ 典 兲
N
⫽ 共 具 R 兩 ␳ˆ 兩 R 典 ⫹ 具 L 兩 ␳ˆ 兩 L 典 ⫺i 具 L 兩 ␳ˆ 兩 R 典 ⫹i 具 R 兩 ␳ˆ 兩 L 典 兲 ,
2
n 3 ⫽N共 具 R 兩 ␳ˆ 兩 R 典 兲 .
PHYSICAL REVIEW A 64 052312
Here 兩 H 典 , 兩 V 典 , 兩 D̄ 典 ⫽( 兩 H 典 ⫺ 兩 V 典 )/ 冑2⫽exp(i␲/4)( 兩 R 典
⫹i 兩 L 典 )/ 冑2, and 兩 R 典 ⫽( 兩 H 典 ⫺i 兩 V 典 )/ 冑2 are the kets repre-
senting qubits polarized in the linear horizontal, linear verti-
cal, linear diagonal (45°), and right-circular senses respec-
tively, ␳ˆ is the (2⫻2) density matrix for the polarization
degrees of the light 共or for a two-level quantum system兲, and
N is a constant dependent on the detector efficiency and light
intensity. The Stokes parameters, which fully characterize
the polarization state of the light, are then defined by
S0 ⬅2n 0 ⫽N共 具 R 兩 ␳ˆ 兩 R 典 ⫹ 具 L 兩 ␳ˆ 兩 L 典 兲 ,
S1 ⬅2 共 n 1 ⫺n 0 兲 ⫽N共 具 R 兩 ␳ˆ 兩 L 典 ⫹ 具 L 兩 ␳ˆ 兩 R 典 兲 ,
S2 ⬅2 共 n 2 ⫺n 0 兲 ⫽Ni 共 具 R 兩 ␳ˆ 兩 L 典 ⫺ 具 L 兩 ␳ˆ 兩 R 典 兲 ,
S3 ⬅2 共 n 3 ⫺n 0 兲 ⫽N共 具 R 兩 ␳ˆ 兩 R 典 ⫺ 具 L 兩 ␳ˆ 兩 L 典 兲 .
共2.2兲
We can now relate the Stokes parameters to the density ma-
trix ␳ˆ by the formula
3
1 S
␳ˆ ⫽
2 兺 S
i⫽0 0
i
␴ˆ i , 共2.3兲
where ␴ˆ 0 ⫽ 兩 R 典具 R 兩 ⫹ 兩 L 典具 L 兩 is the single qubit identity opera-
tor and ␴ˆ 1 ⫽ 兩 R 典具 L 兩 ⫹ 兩 L 典具 R 兩 , ␴ˆ 2 ⫽i 兩 L 典具 R 兩 ⫺ 兩 R 典具 L 兩 , and ␴ˆ 3
⫽ 兩 R 典具 R 兩 ⫺ 兩 L 典具 L 兩 are the Pauli spin operators. Thus the
measurement of the Stokes parameters can be considered
equivalent to a tomographic measurement of the density ma-
trix of an ensemble of single qubits.
B. Multiple beam Stokes parameters: Multiple qubit
tomography
The generalization of the Stokes scheme to measure the
state of multiple photon beams 共or multiple qubits兲 is reason-
ably straightforward. One should, however, be aware that
important differences exist between the one-photon and the
multiple photon cases. Single photons, at least in the current
context, can be described in a purely classical manner, and
the density matrix can be related to the purely classical con-
cept of the coherency matrix 关21兴. For multiple photons one
has the possibility of nonclassical correlations occurring,
with quintessentially quantum mechanical phenomena such
as entanglement being present. We will return to the concept
of entanglement and how it may be measured later in this
paper.
An n-qubit state is characterized by a density matrix
1
␳ˆ ⫽
2n
兺
i 1 ,i 2 , . . . ,i n ⫽0
r i 1 ,i 2 , . . . ,i n ␴ˆ i 1 丢 ␴ˆ i 2 丢 ••• 丢 ␴ˆ i n ,
共2.4兲
where the 4 n parameters r i 1 ,i 2 , . . . ,i n are real numbers. The
normalization property of the density matrices requires that
r 0,0, . . . ,0 ⫽1, and so the density matrix is specified by 4 n
共2.1兲 ⫺1 real parameters. The symbol 丢 represents the tensor
052312-2
MEASUREMENT OF QUBITS PHYSICAL REVIEW A 64 052312

FIG. 1. Tree diagram representing number and type of measurements necessary for tomography. For a single qubit, the measurements
兵 ␮ˆ 0 , ␮ˆ 1 , ␮ˆ 2 , ␮ˆ 3 其 suffice to reconstruct the state, e.g., measurements of the horizontal, vertical, diagonal, and right-circular polarization
components, (H,V,D,R). For two qubits, 16 double-coincidence measurements are necessary ( 兵 ␮ ˆ 0␮
ˆ 0 ,␮
ˆ 0␮
ˆ 1 , . . . ,␮
ˆ 3␮
ˆ 3 其 ), increasing to 64
three-coincidence measurements for three qubits ( 兵 ␮ 0 ␮ 0 ␮ 0 , ␮ 0 ␮ 0 ␮ 1 , . . . , ␮ 3 ␮ 3 ␮ 3 其 ), and so on, as shown.
ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆ

冉 冊
product between operators acting on the Hilbert spaces asso- 1 0 0 0
ciated with the separate qubits.
1/2 1/2 0 0
As Stokes showed, the state of a single qubit can be de- ⌼⫽ . 共2.8兲
termined by taking a set of four projection measurements 1/2 0 1/2 0
which are represented by the four operators ␮ ˆ 0 ⫽ 兩 H 典具 H 兩 1/2 0 0 1/2
⫹ 兩 V 典具 V 兩 , ␮ ˆ 1 ⫽ 兩 H 典具 H 兩 , ␮
ˆ 2 ⫽ 兩 D̄ 典具 D̄ 兩 , ␮
ˆ 3 ⫽ 兩 R 典具 R 兩 . Simi-
larly, the state of two qubits can be determined by the set of Further, we have the relation Tr兵 ␴ˆ i ␴ˆ j 其 ⫽2 ␦ i j 共where ␦ i j is
16 measurements represented by the operators ␮ ˆ i丢␮ ˆ j (i, j the Kronecker delta兲. Hence Eq. 共2.7兲 becomes
⫽0,1,2,3). More generally the state of an n-qubit system can 3
be determined by 4 n measurements given by the operators
␮ˆ i 1 丢 ␮ˆ i 2 丢 ••• 丢 ␮ˆ i n (i k ⫽0,1,2,3 and k⫽1,2, . . . ,n). This
n i 1 ,i 2 , . . . ,i n ⫽N 兺
j 1 , j 2 , . . . , j n ⫽0
⌼ i 1 j 1 ⌼ i 2 j 2 •••⌼ i n j n r i 1 ,i 2 , . . . ,i n .

‘‘tree’’ structure for multiqubit measurement is illustrated in 共2.9兲

Fig. 1. Introducing the left inverse of the matrix ⌼, defined so that
The proof of this conjecture is reasonably straightforward. 3
兺 k⫽0 (⌼ ⫺1 ) ik ⌼ k j ⫽ ␦ i j and whose elements are

冉 冊
The outcome of a measurement is given by the formula
1 0 0 0
n⫽N Tr兵 ␳ˆ ␮
ˆ 其, 共2.5兲 ⫺1 2 0 0
⌼ ⫺1 ⫽ , 共2.10兲
⫺1 0 2 0
where ␳ˆ is the density matrix, ␮
ˆ is the measurement opera-
⫺1 0 0 2
tor, and N is a constant of proportionality which can be
determined from the data. Thus in our n-qubit case the out- we can find a formula for the parameters r i 1 ,i 2 , . . . ,i n in terms
comes of the various measurement are
of the measured quantities n i 1 ,i 2 , . . . ,i n , viz.,

n i 1 ,i 2 , . . . ,i n ⫽N Tr 兵 ␳ˆ 共 ␮ ˆ i 丢 ••• 丢 ␮
ˆ i 丢␮ ˆ i 兲其. 共2.6兲 Nr i 1 ,i 2 , . . . ,i n
1 2 n

3
Substituting from Eq. 共2.4兲 we obtain ⫽ 兺
j 1 , j 2 , . . . , j n ⫽0
共 ⌼ ⫺1 兲 i 1 j 1 共 ⌼ ⫺1 兲 i 2 j 2 ••• 共 ⌼ ⫺1 兲 i n j n
3
N
n i 1 ,i 2 , . . . ,i n ⫽
2n
兺
j 1 , j 2 , . . . , j n ⫽0
Tr兵 ␮
ˆ i ␴ˆ j 其
1 1
⫻n i 1 ,i 2 , . . . ,i n

⬅Si 1 ,i 2 , . . . ,i n . 共2.11兲
ˆ i ␴ˆ j 其 •••Tr兵 ␮
⫻Tr兵 ␮ ˆ i ␴ˆ j 其 r i ,i , . . . ,i .
2 2 n n 1 2 n
In Eq. 共2.11兲 we have introduced the n-photon Stokes param-
共2.7兲
eter Si 1 ,i 2 , . . . ,i n , defined in an analogous manner to the
single photon Stokes parameters give in Eq. 共2.2兲.
As can be easily verified, the single qubit measurement op-
Since, as already noted, r 0,0, . . . ,0 ⫽1, we can make the
erators ␮ ˆ i are linear combinations of the Pauli operators ␴ˆ j ,
identification S0,0, . . . ,0⫽N, and so the density matrix for the
i.e., ␮
ˆ i ⫽ 兺 3j⫽0 ⌼ i j ␴ˆ j , where ⌼ i j are the elements of the ma- n-qubit system can be written in terms of the Stokes param-
trix eters as follows:

052312-3
JAMES, KWIAT, MUNRO, AND WHITE PHYSICAL REVIEW A 64 052312

beams of light, whose quanta can be measured by means of

photodetectors. To project the light beams onto a polarization
state of the experimenter’s choosing, three optical elements
are placed in the beam in front of each detector: a polarizer
共which transmits only vertically polarized light兲, a quarter-
wave plate, and a half-wave plate. The angles of the fast axes
of both of the wave plates can be set arbitrarily, allowing the
兩 V 典 projection state fixed by the polarizer to be rotated into
any polarization state that the experimenter may wish.
Using the Jones calculus notation, with the convention

FIG. 2. Schematic illustration of the experimental arrangement.

QWP stands for quarter-wave plate, HWP for half-wave plate; the
冉冊0
1
⫽兩V典, 冉冊1
0
⫽兩H典, 共3.1兲
angles of both pairs of wave plates can be set independently giving
the experimenter four degrees of freedom with which to set the where 兩 V 典 ( 兩 H 典 ) is the ket for a vertically 共horizontally兲
projection state. In the experiment, the polarizers were realized us- polarized beam, the effects of quarter- and half-wave plates
ing polarizing prisms, arranged to transmit vertically polarized whose fast axes are at angles q and h with respect to the
light. vertical axis, respectively, are given by the 2⫻2 matrices

冑 冉 冊
1
3 Si 1 ,i 2 , . . . ,i n
i⫺cos共 2q 兲 sin共 2q 兲
␳ˆ ⫽ 兺
2 n i 1 ,i 2 , . . . ,i n ⫽0 S0,0, . . . ,0
␴ˆ i 1 丢 ␴ˆ i 2 丢 ••• 丢 ␴ˆ i n . 共2.12兲 Û QW P 共 q 兲 ⫽
1
,
2 sin共 2q 兲 i⫹cos共 2q 兲

冉 冊
This is a recipe for measurement of the density matrices
which, assuming perfect experimental conditions and the cos共 2h 兲 ⫺sin共 2h 兲
Û HW P 共 q 兲 ⫽ . 共3.2兲
complete absence of noise, will always work. It is important ⫺sin共 2h 兲 ⫺cos共 2h 兲
to realize that the set of four Stokes measurements
兵 ␮ˆ 0 , ␮ˆ 1 , ␮ˆ 2 , ␮ˆ 3 其 is not unique: there may be circumstances Thus the projection state for the measurement in one of the
in which it is more convenient to use some other set, which beams is given by
is equivalent. A more typical set, at least in optical experi-
ments, is ␮
⫽ 兩 R 典具 R 兩 .
ˆ 0⬘ ⫽ 兩 H 典具 H 兩 , ␮
ˆ ⬘1 ⫽ 兩 V 典具 V 兩 , ␮

In the following section we will explore more general

ˆ ⬘2 ⫽ 兩 D 典具 D 兩 , ␮
ˆ 3⬘
pro j 共 h,q 兲 典 ⫽Û QW P 共 q 兲 •Û HW P 共 h 兲 •
兩 ␺ (1) 冉冊0
1
schemes for the measurement of two qubits, starting with a ⫽a 共 h,q 兲 兩 H 典 ⫹b 共 h,q 兲 兩 V 典 , 共3.3兲
discussion, in some detail, of how the measurements are ac-
tually performed. where, neglecting an overall phase, the functions a(h,q) and
b(h,q) are given by
III. GENERALIZED TOMOGRAPHIC RECONSTRUCTION
OF THE POLARIZATION STATE OF TWO PHOTONS 1
a 共 h,q 兲 ⫽ 兵 sin共 2h 兲 ⫺i sin关 2 共 h⫺q 兲兴 其 ,
A. Experimental setup 冑2
The experimental arrangement used in our experiments is
shown schematically in Fig. 2. An optical system consisting 1
of lasers, polarization elements, and nonlinear optical crys- b 共 h,q 兲 ⫽⫺ 兵 cos共 2h 兲 ⫹i cos关 2 共 h⫺q 兲兴 其 . 共3.4兲
tals 共collectively characterized for the purposes of this paper 冑2
as a ‘‘black box,’’兲 is used to generate pairs of qubits in an
almost arbitrary quantum state of their polarization degrees The projection state for the two beams is given by
of freedom. A full description of this optical system and how
pro j 共 h 1 ,q 1 ,h 2 ,q 2 兲 典 ⫽ 兩 ␺ pro j 共 h 1 ,q 1 兲 典 丢 兩 ␺ pro j 共 h 2 ,q 2 兲 典
兩 ␺ (2)
such quantum states can be prepared can be found in Refs. (1) (1)
关22–24兴.1 The output of the black box consists of a pair of
⫽a 共 h 1 ,q 1 兲 a 共 h 2 ,q 2 兲 兩 HH 典

1 ⫹a 共 h 1 ,q 1 兲 b 共 h 2 ,q 2 兲 兩 HV 典
It is important to realize that the entangled photon pairs are pro-
duced in a nondeterministic manner: one cannot specify with cer- ⫹b 共 h 1 ,q 1 兲 a 共 h 2 ,q 2 兲 兩 VH 典
tainty when a photon pair will be emitted; indeed there is a small
probability of generating four or six or a higher number of photons. ⫹b 共 h 1 ,q 1 兲 b 共 h 2 ,q 2 兲 兩 VV 典 . 共3.5兲
Thus we can only postselectively generate entangled photon pairs:
i.e., one only knows that the state was created after if has been We shall denote the projection state corresponding to one
measured. particular set of wave plate angles 兵 h 1,␯ ,q 1,␯ ,h 2,␯ ,q 2,␯ 其 by

052312-4
MEASUREMENT OF QUBITS PHYSICAL REVIEW A 64 052312

the ket 兩 ␺ ␯ 典 ;2 thus the projection measurement is represented TABLE I. The tomographic analysis states used in our experi-
by the operator ␮ ˆ ␯ ⫽ 兩 ␺ ␯ 典具 ␺ ␯ 兩 . Consequently, the average ments. The number of coincidence counts measured in projection
measurements provides a set of 16 data that allow the density ma-
number of coincidence counts that will be observed in a
trix of the state of the two modes to be estimated. We have used the
given experimental run is
notation 兩 D 典 ⬅( 兩 H 典 ⫹ 兩 V 典 )/ 冑2, 兩 L 典 ⬅( 兩 H 典 ⫹i 兩 V 典 )/ 冑2, and 兩 R 典
⬅( 兩 H 典 ⫺i 兩 V 典 )/ 冑2. Note that, when the measurements are taken in
n ␯ ⫽N具 ␺ ␯ 兩 ␳ˆ 兩 ␺ ␯ 典 共3.6兲 the order given by the table, only one wave plate angle has to be
changed between measurements.
where ␳ˆ is the density matrix describing the ensemble of
qubits, and N is a constant dependent on the photon flux and ␯ Mode 1 Mode 2 h1 q1 h2 q2
detector efficiencies. In what follows, it will be convenient to
consider the quantities s ␯ defined by 1 兩H典 兩H典 45° 0 45° 0
2 兩H典 兩V典 45° 0 0 0
s ␯ ⫽ 具 ␺ ␯ 兩 ␳ˆ 兩 ␺ ␯ 典 . 共3.7兲 3 兩V典 兩V典 0 0 0 0
4 兩V典 兩H典 0 0 45° 0
B. Tomographically complete set of measurements 5 兩R典 兩H典 22.5° 0 45° 0
In Sec. II we have given one possible set of projection 6 兩R典 兩V典 22.5° 0 0 0
measurements 兵 兩 ␺ ␯ 典具 ␺ ␯ 兩 其 which uniquely determine the 7 兩D典 兩V典 22.5° 45° 0 0
density matrix ␳ˆ . However, one can conceive of situations in 8 兩D典 兩H典 22.5° 45° 45° 0
which these will not be the most convenient set of measure- 9 兩D典 兩R典 22.5° 45° 22.5° 0
ments to make. Here we address the problem of finding other 10 兩D典 兩D典 22.5° 45° 22.5° 45°
sets of suitable measurements. The smallest number of states 11 兩R典 兩D典 22.5° 0 22.5° 45°
required for such measurements can be found by a simple 12 兩H典 兩D典 45° 0 22.5° 45°
argument: there are 15 real unknown parameters that deter- 13 兩V典 兩D典 0 0 22.5° 45°
mine a 4⫻4 density matrix, plus there is the single unknown 14 兩V典 兩L典 0 0 22.5° 90°
real parameter N, making a total of 16. 15 兩H典 兩L典 45° 0 22.5° 90°
In order to proceed it is helpful to convert the 4⫻4 matrix 16 兩R典 兩L典 22.5° 0 22.5° 90°
␳ˆ into a 16-dimensional column vector. To do this we use a
set of 16 linearly independent 4⫻4 matrices 兵 ⌫ˆ ␯ 其 which Substituting from Eq. 共3.9兲 into Eq. 共3.6兲, we obtain the
have the following mathematical properties: following linear relationship between the measured coinci-
dence counts n ␯ and the elements of the vector r ␮ :
Tr兵 ⌫ˆ ␯ •⌫ˆ ␮ 其 ⫽ ␦ ␯ , ␮
16
16
n ␯ ⫽N 兺 B ␯,␮r ␮ 共3.11兲
Â⫽ 兺 ⌫ˆ ␯ Tr兵 ⌫ˆ ␯ •Â 其 ᭙Â,
␯ ⫽1
共3.8兲 ␮ ⫽1

where the 16⫻16 matrix B ␯ , ␮ is given by

where  is an arbitrary 4⫻4 matrix. Finding a set of ⌫ˆ ␯
matrices is in fact reasonably straightforward: for example, B ␯ , ␮ ⫽ 具 ␺ ␯ 兩 ⌫ˆ ␮ 兩 ␺ ␯ 典 . 共3.12兲
the set of 共appropriately normalized兲 generators of the Lie
algebra SU(2) 丢 SU(2) fulfill the required criteria 共for refer- Immediately we find a necessary and sufficient condition for
ence, we list this set in Appendix A兲. These matrices are of the completeness of the set of tomographic states 兵 兩 ␺ ␯ 典 其 : if
course simply a relabeling of the two-qubit Pauli matrices the matrix B ␯ , ␮ is nonsingular, then Eq. 共3.11兲 can be in-
␴ˆ i 丢 ␴ˆ j (i, j⫽0,1,2,3) discussed above. Using these matrices verted to give
the density matrix can be written as 16

16 r ␯ ⫽ 共 N 兲 ⫺1 兺
␮ ⫽1
共 B ⫺1 兲 ␯ , ␮ n ␮ . 共3.13兲
␳ˆ ⫽ 兺
␯ ⫽1
⌫ˆ ␯ r ␯ , 共3.9兲
The set of 16 tomographic states that we employed is
given in Table I. They can be shown to satisfy the condition
where r ␯ is the ␯ th element of a 16-element column vector,
that B ␯ , ␮ is nonsingular. By no means are these states unique
given by the formula
in this regard: these were the states chosen principally for
experimental convenience.
r ␯ ⫽Tr兵 ⌫ˆ ␯ • ␳ˆ 其 . 共3.10兲 These states can be realized by setting specific values of
the half- and quarter-wave plate angles. The appropriate val-
ues of these angles 共measured from the vertical兲 are given in
2
Here the first subscript on the wave plate angle refers to one of Table I. Note that overall phase factors do not affect the
the two photon beams; the second subscript distinguishes which of results of projection measurements.
the 16 different experimental states is under consideration. Substituting Eq. 共3.13兲 into Eq. 共3.9兲, we find that

052312-5
JAMES, KWIAT, MUNRO, AND WHITE PHYSICAL REVIEW A 64 052312

␳ˆ ⫽ 共 N 兲 ⫺1
16

␯ ⫽1
16

兺 M̂ ␯ n ␯ ⫽ ␯兺⫽1 M̂ ␯ s ␯ , 共3.14兲 Tr兵 M̂ ␯ 其 ⫽ 再 1

0
if ␯ ⫽1,2,3,4
if ␯ ⫽5, . . . ,16;
共3.18兲

where the sixteen 4⫻4 matrices M̂ ␯ are defined by hence the value of the unknown parameter N in our experi-
ments is given by
16

M̂ ␯ ⫽ 兺 共 B ⫺1 兲 ␯ , ␮ ⌫ˆ ␮ . 共3.15兲 4
␯ ⫽1 N⫽ 兺
␯ ⫽1
n␯
The introduction of the M̂ ␯ matrices allows a compact form
of linear tomographic reconstruction, Eq. 共3.14兲, which will ⫽N共 具 HH 兩 ␳ˆ 兩 HH 典 ⫹ 具 HV 兩 ␳ˆ 兩 HV 典
be most useful in the error analysis that follows. These M̂ ␯
⫹ 具 VH 兩 ␳ˆ 兩 VH 典 ⫹ 具 VV 兩 ␳ˆ 兩 VV 典 兲 . 共3.19兲
matrices, valid for our set of tomographic states, are listed in
Appendix B, together with some of their important proper- Thus we obtain the final formula for the tomographic recon-
ties. We can use one of these properties, Eq. 共B6兲, to obtain struction of the density matrices of our states:
the value of the unknown quantity N. That relationship im-
plies

兺␯ Tr兵 M̂ ␯ 其 兩 ␺ ␯ 典具 ␺ ␯兩 ␳ˆ ⫽ ␳ˆ . 共3.16兲
␳ˆ ⫽ 冉兺 冊冒冉兺 冊
16

␯ ⫽1
M̂ ␯ n ␯
4

␯ ⫽1
n␯ . 共3.20兲

As an example, the following set of 16 counts were taken

Taking the trace of this formula, and multiplying by N we
obtain
兺␯ Tr兵 M̂ ␯ 其 n ␯ ⫽N.
For our set of tomographic states, it can be shown that
for the purpose of tomographically determining the density
matrix for an ensemble of qubits all prepared in a specific
quantum state: n 1 ⫽34 749, n 2 ⫽324, n 3 ⫽35 805, n 4 ⫽444,
n 5 ⫽16 324, n 6 ⫽17 521, n 7 ⫽13 441, n 8 ⫽16 901, n 9
共3.17兲 ⫽17 932, n 10⫽32 028, n 11⫽15 132, n 12⫽17 238, n 13
⫽13 171, n 14⫽17 170, n 15⫽16 722, n 16⫽33 586. Applying
Eq. 共3.20兲 we find

冉 冊
0.4872 ⫺0.0042⫹i0.0114 ⫺0.0098⫺i0.0178 0.5192⫹i0.0380
⫺0.0042⫺i0.0114 0.0045 0.0271⫺i0.0146 ⫺0.0648⫺i0.0076
␳ˆ ⫽ . 共3.21兲
⫺0.0098⫹i0.0178 0.0271⫹i0.0146 0.0062 ⫺0.0695⫹i0.0134
0.5192⫺i0.0380 ⫺0.0648⫹i0.0076 ⫺0.0695⫺i0.0134 0.5020

This matrix is shown graphically in Fig. 3共left兲.
Note that, by construction, the density matrix is normal-
ized, i.e., Tr兵 ␳ˆ 其 ⫽1, and Hermitian, i.e., ␳ˆ † ⫽ ␳ˆ . However,
when one calculates the eigenvalues of this measured
density matrix, one finds the values 1.021 55, 0.068 123 8,
⫺0.065 274, and ⫺0.024 396; and also Tr兵 ␳ˆ 2 其 ⫽1.053. Den-
sity matrices for all physical states must have the property of
positive semidefiniteness, which 共in conjunction with the
normalization and Hermiticity properties兲 implies that all of
the eigenvalues must lie in the interval 关 0,1兴 , their sum being
1; this in turn implies that 0⭐Tr兵 ␳ˆ 2 其 ⭐1. Clearly, the density
matrix reconstructed above by linear tomography violates
this condition. From our experience of tomographic mea-
surements of various mixed and entangled states prepared
experimentally, this seems to happen roughly 75% of the
time for low-entropy, highly entangled states; it seems to
have a higher probability of producing the correct result for
states of higher entropy, but the cautious experimenter
should check every time. The obvious culprit for this prob-
lem is experimental inaccuracies and statistical fluctuations
of coincidence counts, which mean that the actual numbers
of counts recorded in a real experiment differ from those that
can be calculated by Eq. 共3.6兲. Thus the linear reconstruction
is of limited value for states of low entropy 共which are of
most experimental interest because of their application to
quantum information technology兲; however, as we shall see,
the linear approach does provide a useful starting point for
the numerical optimization approach to density matrix esti-
mation which we will discuss in the next section.
IV. MAXIMUM LIKELIHOOD ESTIMATION
As mentioned in Sec. III, the tomographic measurement
of density matrices can produce results that violate important
basic properties such as positivity. To avoid this problem, the
maximum likelihood estimation of density matrices may be
employed. Here we describe a simple realization of this tech-
nique.

052312-6
MEASUREMENT OF QUBITS PHYSICAL REVIEW A 64 052312

FIG. 3. Graphical representation of the den-

sity matrix of a state as estimated by linear to-
mography 共left兲 and by maximum likelihood to-
mography 共right兲 from the experimental data
given in the text. The upper plot is the real part of
␳ˆ , the lower plot the imaginary part.

A. Basic approach
具 ␺ 兩 T̂ † T̂ 兩 ␺ 典 ⫽ 具 ␺ ⬘ 兩 ␺ ⬘ 典 ⭓0, 共4.2兲
Our approach to the maximum likelihood estimation of
the density matrix is as follows. where we have defined 兩 ␺ ⬘ 典 ⫽T̂ 兩 ␺ 典 . Furthermore, (T̂ † T̂) †
共i兲 Generate a formula for an explicitly ‘‘physical’’ density ⫽T̂ † (T̂ † ) † ⫽T̂ † T̂, i.e., Ĝ⫽T̂ † T̂ must be Hermitian. To ensure
matrix, i.e., a matrix that has the three important properties normalization, one can simply divide by the trace: thus the
of normalization, Hermiticity, and positivity. This matrix will
be a function of 16 real variables 共denoted 兵 t 1 ,t 2 , . . . ,t 16其 ). matrix ĝ given by the formula
We will denote the matrix as ␳ˆ p (t 1 ,t 2 , . . . ,t 16).
共ii兲 Introduce a ‘‘likelihood function’’ which quantifies ĝ⫽T̂ † T̂/Tr兵 T̂ † T̂ 其 共4.3兲
how good the density matrix ␳ˆ p (t 1 ,t 2 , . . . ,t 16) is in relation
has all three of the mathematical properties that we require
to the experimental data. This likelihood function is a
for density matrices.
function of the 16 real parameters t ␯ and of the 16
For the two-qubit system, we have a 4⫻4 density matrix
experimental data n ␯ . We will denote this function as
with 15 independent real parameters. Since it will be useful
L(t 1 ,t 2 , . . . ,t 16 ;n 1 ,n 2 , . . . ,n 16).
to be able to invert relation 共4.3兲, it is convenient to choose a
共iii兲 Using standard numerical optimization techniques,
find the optimum set of variables 兵 t (opt) ,t (opt) (opt)
其 tridiagonal form for T̂:

冉 冊
1 2 , . . . ,t 16
for which the function L(t 1 ,t 2 , . . . ,t 16 ;n 1 ,n 2 , . . . ,n 16) has
its maximum value. The best estimate for the density matrix t1 0 0 0
is then ␳ˆ (t (opt)
1 ,t (opt)
2
(opt)
, . . . ,t 16 ). t 5 ⫹it 6 t2 0 0
The details of how these three steps can be carried out are T̂ 共 t 兲 ⫽ . 共4.4兲
t 11⫹it 12 t 7 ⫹it 8 t3 0
described in the next three subsections.
t 15⫹it 16 t 13⫹it 14 t 9 ⫹it 10 t 4
B. Physical density matrices
Thus the explicitly ‘‘physical’’ density matrix ␳ˆ p is given
The property of non-negative definiteness for any matrix by the formula
Ĝ is written mathematically as
␳ˆ p 共 t 兲 ⫽T̂ † 共 t 兲 T̂ 共 t 兲 /Tr兵 T̂ † 共 t 兲 T̂ 共 t 兲 其 . 共4.5兲
具 ␺ 兩 Ĝ兩 ␺ 典 ⭓0 ᭙兩␺典. 共4.1兲
For future reference, the inverse relationship, by which
Any matrix that can be written in the form Ĝ⫽T̂ † T̂ must be
non-negative definite. To see that this is the case, substitute the elements of T̂ can be expressed in terms of the elements
into Eq. 共4.1兲 of ␳ˆ , is as follows:

052312-7
JAMES, KWIAT, MUNRO, AND WHITE PHYSICAL REVIEW A 64 052312

冢 冣
⌬
0 0 0
M 11
(1)

冑M 11
(1)
M 12
(1)

M 11,22
(2) 冑 M 11
(1)

M 11,22
(2)
0 0

T̂⫽ . 共4.6兲
M 12,23 M 11,23
冑
(2) (2)
M 11,22
(2)
0
冑 冑
␳ 44 M 11,22
(2)
冑 冑
␳ 44 M 11,22
(2) ␳ 44

␳ 41 ␳ 42 ␳ 43
冑␳ 44
冑␳ 44 冑␳ 44 冑␳ 44

Here we have used the notation ⌬⫽Det( ␳ˆ ); M (1)i j is the first

where N⫽ 兺 ␯4 ⫽1 n ␯ .
minor of ␳ , i.e., the determinant of the 3⫻3 matrix formed
ˆ Rather than find the maximum value of P(t 1 ,t 2 , . . . ,t 16)
by deleting the ith row and jth column of ␳ˆ ; M (2) it simplifies things somewhat to find the maximum of its
i j,kl is the
logarithm 共which is mathematically equivalent兲.3 Thus the
second minor of ␳ , i.e., the determinant of the 2⫻2 matrix
ˆ
optimization problem reduces to finding the minimum of the
formed by deleting the ith and kth rows and jth and lth
following function:
columns of ␳ˆ (i⫽k and j⫽l).

C. The likelihood function

L共 t 1 ,t 2 , . . . ,t 16兲
The measurement data consist of a set of 16 coincidence
16
counts n ␯ ( ␯ ⫽1,2, . . . ,16) whose expected value is n̄ ␯ 关 N具 ␺ ␯ 兩 ␳ˆ p 共 t 1 ,t 2 , . . . ,t 16兲 兩 ␺ ␯ 典 ⫺n ␯ 兴 2
⫽N具 ␺ ␯ 兩 ␳ˆ 兩 ␺ ␯ 典 . Let us assume that the noise on these coin-
⫽ 兺
␯ ⫽1 2N具 ␺ ␯ 兩 ␳ˆ p 共 t 1 ,t 2 , . . . ,t 16兲 兩 ␺ ␯ 典
.
cidence measurements has a Gaussian probability distribu-
tion. Thus the probability of obtaining a set of 16 counts 共4.10兲
兵 n 1 ,n 2 , . . . n 16其 is
This is the ‘‘likelihood’’ function that we employed in our

P 共 n 1 ,n 2 , . . . ,n 16兲 ⫽
1
N norm 兿
16

␯ ⫽1
exp ⫺ 冋 共 n ␯ ⫺n̄ ␯ 兲 2
2 ␴ ␯2
册 ,
numerical optimization routine.

共4.7兲 D. Numerical optimization

where ␴ ␯ is the standard deviation for the ␯ th coincidence
measurement 共given approximately by 冑n̄ ␯ ) and N norm is the
normalization constant. For our physical density matrix ␳ˆ p
the number of counts expected for the ␯ th measurement is
n̄ ␯ 共 t 1 ,t 2 , . . . ,t 16兲 ⫽N具 ␺ ␯ 兩 ␳ˆ p 共 t 1 ,t 2 , . . . ,t 16兲 兩 ␺ ␯ 典 .
共4.8兲
Thus the likelihood that the matrix ␳ˆ p (t 1 ,t 2 , . . . ,t 16) could
produce the measured data 兵 n 1 ,n 2 , . . . ,n 16其 is
P 共 n 1 ,n 2 , . . . ,n 16兲
We used the MATHEMATICA 4.0 routine FINDMINIMUM
which executes a multidimensional Powell direction set al-
gorithm 共see Ref. 关25兴 for a description of this algorithm兲. To
execute this routine, one requires an initial estimate for the
values of t 1 ,t 2 , . . . ,t 16 . For this, we used the tomographic
estimate of the density matrix in the inverse relation 共4.6兲,
allowing us to determine a set of values for t 1 ,t 2 , . . . t 16 .
Since the tomographic density matrix may not be non-
negative definite, the values of the t ␯ ’s deduced in this man-
ner are not necessarily real. Thus for our initial guess we
used the real parts of the t ␯ ’s deduced from the tomographic
density matrix.
For the example given in Sec. II, the maximum likelihood

⫽
1
N norm 兿
16

␯ ⫽1
exp 冋 estimate is

册
3
关 N具 ␺ ␯ 兩 ␳ˆ p 共 t 1 ,t 2 , . . . ,t 16兲 兩 ␺ ␯ 典 ⫺n ␯ 兴 2 Note that here we neglect the dependence of the normalization
⫺ , 共4.9兲 constant on t 1 ,t 2 , . . . ,t 16 , which only weakly affects the solution
2N具 ␺ ␯ 兩 ␳ˆ p 共 t 1 ,t 2 , . . . ,t 16兲 兩 ␺ ␯ 典 for the most likely state.

052312-8
MEASUREMENT OF QUBITS PHYSICAL REVIEW A 64 052312

冉 冊
0.5069 ⫺0.0239⫹i0.0106 ⫺0.0412⫺i0.0221 0.4833⫹i0.0329
⫺0.0239⫺i0.0106 0.0048 0.0023⫹i0.0019 ⫺0.0296⫺i0.0077
␳ˆ ⫽ . 共4.11兲
⫺0.0412⫹i0.0221 0.0023⫺i0.0019 0.0045 ⫺0.0425⫹i0.0192
0.4833⫺i0.0329 ⫺0.0296⫹i0.0077 ⫺0.0425⫺i0.0192 0.4839

This matrix is illustrated in Fig. 3 共right兲. In this case, the

matrix has eigenvalues 0.986 022, 0.013 977 7, 0, and 0; and
Tr兵 ␳ˆ 2 其 ⫽0.972 435, indicating that, while the linear recon-
struction gave a nonphysical density matrix, the maximum
D ␯⫽
␭⫽1
4

兺 ␦ ␭, ␯ ⫽ 再 1

0
if 1⭐ ␯ ⭐4

if 5⭐ ␯ ⭐16.
共5.5兲

likelihood reconstruction gives a legitimate density matrix. Substituting from Eq. 共5.4兲 into Eq. 共5.2兲 and using Eq. 共5.3兲,
we obtain the result
V. ERROR ANALYSIS n␮ n ␯n ␮
␦ s ␯␦ s ␮⫽ ␦ ␯,␮⫹ 共 1⫺D ␮ ⫺D ␯ 兲 . 共5.6兲
In this section we present an analysis of the errors inher- N 2
N3
ent in the tomographic scheme described in Sec. III. Two
sources of error are found to be important: the shot noise In most experimental circumstances NⰇ1, and so the second
error in the measured coincidence counts n ␯ and the uncer- term in Eq. 共5.6兲 is negligibly small in comparison to the
tainty in the settings of the angles of the wave plates used to first. We shall therefore ignore it, and use the approximate
make the tomographic projection states. We will analyze expression in the subsequent discussion:
these two sources separately.
In addition to determining the density matrix of a pair of n␮ s␮
␦ s ␯␦ s ␮⬇ ␦ ␯,␮⬅ ␦ . 共5.7兲
qubits, one is often also interested in quantities derived from N 2 N ␯,␮
the density matrix, such as the entropy or the entanglement
of formation. For completeness, we will also derive the er-
rors in some of these quantities. B. Errors due to angular settings uncertainties
Using the formula 共3.7兲 for the parameters s ␯ we can find
A. Errors due to count statistics the dependence of the measured density matrix on errors in
the tomographic states. The derivative of s ␯ with respect to
From Eq. 共3.20兲 we see that the density matrix is specified
some generic wave plate setting angle ␪ is
by a set of 16 parameters s ␯ defined by

s ␯ ⫽n ␯ /N, 共5.1兲
⳵s␯
⳵␪
⫽
⳵
⳵␪ 再 冎
具 ␺ ␯ 兩 ␳ˆ 兩 ␺ ␯ 典 ⫹ 具 ␺ ␯ 兩 ␳ˆ
⳵
兩␺ 典 ,
⳵␪ ␯ 再 冎 共5.8兲

where n ␯ are the measured coincidence counts and N

where 兩 ␺ ␯ 典 is the ket of the ␯ th projection state 关see Eq.
⫽ 兺 ␯4 ⫽1 n ␯ . We can determine the errors in s ␯ using the fol-
共3.5兲兴. Substituting from Eq. 共3.14兲 we find
lowing formula 关26兴:

冋再 冎 再 冎册
16
⳵s␯ ⳵ ⳵
冉 冊冉 冊 兺
16
⳵s␯ ⳵s␮ ⫽ s 具 ␺ 兩 M̂ ␮ 兩 ␺ ␯ 典 ⫹ 具 ␺ ␯ 兩 M̂ ␮ 兩␺ 典 .
␦ s ␯␦ s ␮⫽ 兺
␭, ␬ ⫽1 ⳵n␭ ⳵n␬
␦ n ␭␦ n ␬, 共5.2兲 ⳵ ␪ ␮ ⫽1 ␮ ⳵␪ ␯ ⳵␪ ␯
共5.9兲

where the overbar denotes the ensemble average of the ran- For convenience, we shall label the four wave plate angles
dom uncertainties ␦ s ␯ and ␦ n ␭ . The measured coincidence 兵 h 1,␯ ,q 1,␯ ,h 2,␯ ,q 2,␯ 其 , which specify the ␯ th state by
counts n ␭ are statistically independent Poissonian random 兵 ␪ ␯ ,1 , ␪ ␯ ,2 , ␪ ␯ ,3 , ␪ ␯ ,4其 , respectively. Clearly the ␮ th state does
variables, which implies the following relation: not depend on any of the ␯ th set of angles. Thus we obtain
the following expression for the derivatives of s ␯ with re-
␦ n ␭ ␦ n ␬ ⫽n ␭ ␦ ␭, ␬ , 共5.3兲 spect to wave plate settings:
16
⳵s␯
where ␦ ␭, ␬ is the Kronecker delta.
Taking the derivative of Eq. 共5.1兲, we find that ⳵ ␪ ␭,i
⫽ ␦ ␯ ,␭
␮ ⫽1
兺
s ␮ f ␯(i), ␮ , 共5.10兲

where
⳵s␮ 1 n␮

再 冎 再 冎
⫽ ␦ ␮␯⫺ 2 D ␯ , 共5.4兲
⳵n␯ N N ⳵ ⳵
f ␯(i), ␮ ⫽ 具 ␺ ␯ 兩 M̂ ␮ 兩 ␺ ␯ 典 ⫹ 具 ␺ ␯ 兩 M̂ ␮ 兩␺ 典 .
⳵ ␪ ␯ ,i ⳵ ␪ ␯ ,i ␯
where 共5.11兲

052312-9
JAMES, KWIAT, MUNRO, AND WHITE PHYSICAL REVIEW A 64 052312

The 1024 quantities f ␯(i), ␮ can be determined by taking the
derivatives of the functional forms of the tomographic states
given by Eqs. 共3.4兲 and 共3.5兲, and evaluating those deriva-
tives at the appropriate values of the arguments 共see Table I兲.
The errors in the angles are assumed to be uncorrelated,
as would be the case if each wave plate were adjusted for
each of the 16 measurements. In reality, for qubit experi-
ments, only one or two of the four wave plates are adjusted
between one measurement and the next. However, the as-
sumption of uncorrelated angular errors greatly simplifies the
calculation 共which is, after all, only an estimate of the er-
rors兲, and seems to produce reasonable figures for our error
bars.4 Thus, with the assumption
␦ ␪ ␯ ,i ␦ ␪ ␮ , j ⫽ ␦ ␯ , ␮ ␦ i, j 共 ⌬ ␪ 兲 2
where M ␯ (i, j) is the i, j element of the matrix M̂ ␯ .
A convenient way in which to estimate errors for a maxi-
mum likelihood tomographic technique 共rather than a linear
tomographic technique兲 is to employ the above formulas,
with the slight modification that the parameter s ␯ should be
recalculated from Eq. 共3.7兲 using the estimated density ma-
trix ␳ˆ est . This does not take into account errors inherent in
the maximum likelihood technique itself.
C. Errors in quantities derived from the density matrix
When calculating the propagation of errors, it is actually
more convenient to use the errors in the s ␯ parameters 关given
by Eq. 共5.15兲兴, rather than the errors in the elements of den-
共5.12兲 sity matrix itself 共which have non-negligible correlations兲.

共where ⌬ ␪ is the rms uncertainty in the setting of the wave 1. Von Neumann entropy
plate, with an estimated value of 0.25° for our apparatus兲, we The von Neumann entropy is an important measure of the
obtain the following expression for the errors in s ␯ due to
angular settings: purity of a quantum state ␳ˆ . It is defined by 关27兴
4
4 16

␦ s ␯␦ s ␮⫽ ␦ ␯,␮ 兺 兺 f ␯(i), ⑀ f ␯(i),␭ s ⑀ s ␭ . 共5.13兲 S⫽⫺Tr兵 ␳ˆ log2 共 ␳ˆ 兲 其 ⫽⫺ 兺

a⫽1
p a log2 p a , 共5.17兲
i⫽1 ⑀ ,␭⫽1

Combining Eqs. 共5.13兲 and 共5.7兲 we obtain the following where p a is an eigenvalue of ␳ˆ , i.e.,
formula for the total error in the quantities s ␯ :
␳ˆ 兩 ␾ a 典 ⫽ p a 兩 ␾ a 典 , 共5.18兲
␦ s ␯␦ s ␮⫽ ␦ ␯,␮⌳ ␯ 共5.14兲
兩 ␾ a 典 being the ath eigenstate (a⫽1, . . . ,4). The error in this
where quantity is given by

冋 册 冉 冊
16
s␯
4 16
⳵S 2
⌳ ␯⫽ ⫹
N i⫽1 兺 ⑀ ,␭⫽1
兺 f ␯(i), ⑀ f ␯(i),␭ s ⑀ s ␭ . 共5.15兲 共 ⌬S兲 2 ⫽ 兺
␯ ⫽1 ⳵s␯
⌳␯ . 共5.19兲

These 16 quantities can be calculated using the parameters s ␯ Applying the chain rule, we find
and the constants f ␯(i), ⑀ . Note that the same result can be ob-
冉 冊 兺 冉 冊冉 冊
4
tained by assuming a priori that the errors in the s ␯ are all ⳵S ⳵pa ⳵S
⫽ . 共5.20兲
uncorrelated, with ⌳ ␯ ⫽ ␦ s ␯2 ; the more rigorous treatment ⳵s␯ a⫽1 ⳵ s ␯ ⳵pa
given here is necessary, however, to demonstrate this fact.
For a typical number of counts, say N⫽10 000 ⵮⬘ it is found The partial differential of an eigenvalue can be easily found
that the contribution of errors from the two causes is roughly by perturbation theory. As is well known 共e.g., 关28兴兲 the
comparable; for larger numbers of counts, the angular set- change in the eigenvalue ␭ a of a matrix Ŵ due to a pertur-
tings will become the dominant source of error. bation in the matrix ␦ Ŵ is
Based on these results, the errors in the values of the
various elements of the density matrix estimated by the lin-
␦ ␭ a ⫽ 具 ␾ a 兩 ␦ Ŵ 兩 ␾ a 典 , 共5.21兲
ear tomographic technique described in Sec. III are as fol-
lows:
where 兩 ␾ a 典 is the eigenvector of Ŵ corresponding to the
16
⳵␳ i, j ⳵␳ i, j
16 eigenvalue ␭ a . Thus the derivative of ␭ a with respect to
共 ⌬ ␳ i, j 兲 2 ⫽ 兺
␯ , ␮ ⫽1
␦s ␦s ⫽
⳵ s ␯ ⳵ s ␮ ␯ ␮ ␯ ⫽1 兺
共 M ␯ (i, j) 兲 2 ⌳ ␯ some variable x is given by
共5.16兲 ⳵␭a
⳵x
⫽ ␾a 冓 冏 冏冔
⳵ Ŵ
⳵x
␾a . 共5.22兲

4
In other experimental circumstances, such as the measurement of
Since ␳ˆ ⫽ 兺 ␯16⫽1 M̂ ␯ s ␯ , we find that
the joint state of two spin-1/2 particles, the tomography would be
realized by performing unitary operations on the spins prior to mea-
⳵pa
surement. In this case, an assumption analogous to ours would be ⫽ 具 ␾ a 兩 M̂ ␯ 兩 ␾ a 典 共5.23兲
wholly justified. ⳵s␯

052312-10
MEASUREMENT OF QUBITS PHYSICAL REVIEW A 64 052312

and so, taking the derivative of Eq. 共5.17兲, Eq. 共5.20兲 be- quantum state 关29兴. For two qubits,5 concurrence is defined
comes as follows: consider the non-Hermitian matrix R̂⫽ ␳ˆ ⌺̂ ␳ˆ T⌺̂
where the superscript T denotes the transpose and the ‘‘spin

冉 冊
4
⳵S 关 1⫹ln p a 兴

冉 冊
flip matrix’’ ⌺̂ is defined by
⳵s␯
⫽⫺ 兺
具 ␾ a 兩 M̂ ␯ 兩 ␾ a 典
ln 2
. 共5.24兲
a⫽1
0 0 0 ⫺1
0 0 1 0
Hence ⌺̂⫽ . 共5.29兲
0 1 0 0

共 ⌬S兲 ⫽ 2
16

兺 兺 冉
␯ ⫽1 a⫽1
4

具 ␾ a 兩 M̂ ␯ 兩 ␾ a 典
关 1⫹ln p a 兴
ln 2 冊 2

⌳␯ .
⫺1 0 0

Note that the definition of ⌺̂ depends on the basis chosen; we

0

共5.25兲
have assumed here the ‘‘computational basis’’
兵 兩 HH 典 , 兩 HV 典 , 兩 VH 典 , 兩 VV 典 其 . In what follows, it will be conve-
For the experimental example given above, S⫽0.106
nient to write R̂ in the following form:
⫾0.049.
16
1
2. Linear entropy R̂⫽
2 兺
␮ , ␯ ⫽1
q̂ ␮ , ␯ s ␮ s ␯ , 共5.30兲
The ‘‘linear entropy’’ is used to quantify the degree of
mixture of a quantum state in an analytically convenient where q̂ ␮ , ␯ ⫽M̂ ␮ ⌺̂M̂ T␯ ⌺̂⫹M̂ ␯ ⌺̂M̂ ␮T ⌺̂. The left and right
form, although unlike the von Neumann entropy it has no eigenstates and eigenvalues of the matrix R̂ we shall denote
direct information theoretic implications. In a normalized by 具 ␰ a 兩 , 兩 ␨ a 典 , and r a , respectively, i.e.,
form 共defined so that its value lies between zero and 1兲, the
linear entropy for a two-qubit system is defined by 具 ␰ a 兩 R̂⫽r a 具 ␰ a 兩 ,

4
P⫽ 共 1⫺Tr兵 ␳ˆ 2 其 兲 ⫽
3
4
3
1⫺
a⫽1
p 2a . 冉 兺
4

冊 共5.26兲
R̂ 兩 ␨ a 典 ⫽r a 兩 ␨ a 典 .
We shall assume that these eigenstates are normalized in the
共5.31兲

usual manner for biorthogonal expansions, i.e., 具 ␰ a 兩 ␨ b 典

To calculate the error in this quantity, we need the following ⫽ ␦ a,b . Further we shall assume that the eigenvalues are
partial derivative: numbered in decreasing order, so that r 1 ⭓r 2 ⭓r 3 ⭓r 4 . The
concurrence is then defined by the formula
C⫽max兵 0,冑r 1 ⫺ 冑r 2 ⫺ 冑r 3 ⫺ 冑r 4 其
4
⳵P 8 ⳵p
⫽⫺ 兺 p a ⳵ s ␯a
再 冉 冊冑 冎
⳵s␯ 3 a⫽1 4
3
8
4 ⫽max 0, 兺 sgn 2
⫺a ra , 共5.32兲
⫽⫺
3 兺
a⫽1
p a 具 ␾ a 兩 M̂ ␯ 兩 ␾ a 典
a⫽1

where sgn(x)⫽1 if x⬎0 and sgn(x)⫽⫺1 if x⬍0. The

8 tangle is given by T⫽C 2 and the entanglement of formation
⫽⫺ Tr兵 ␳ˆ M̂ ␯ 其 by
3

⫽⫺
8
3
16

兺 Tr兵 M̂ ␮ M̂ ␯ 其 s ␮ .
␮ ⫽1
共5.27兲
E⫽h 冉 1⫹ 冑1⫺C 2
2
, 冊 共5.33兲

where h(x)⫽⫺x log2 x⫺(1⫺x)log2 (1⫺x). Because h(x) is

Hence the error in the linear entropy is a monotonically increasing function, these three quantities
are to some extent equivalent measures of the entanglement

冉 冊 冉 冊
16 16 16 2 of a mixed state.
⳵P 2
8
共 ⌬P兲 2 ⫽ 兺
␯ ⫽1 ⳵s␯
⌳ ␯⫽ 兺␯ 3 兺
␮ ⫽1
Tr兵 M̂ ␮ M̂ ␯ 其 s ␮ ⌳␯ .
To calculate the errors in these rather complicated func-
tions, we must employ the perturbation theory for non-
共5.28兲 Hermitian matrices 共see Appendix C for more details兲. We
need to evaluate the following partial derivative:
For the example given in Secs. III and IV, P⫽0.037
⫾0.026. 5
The analysis in this subsection applies to the two-qubit case only.
Measures of entanglement for mixed n-qubit systems are a subject
3. Concurrence, entanglement of Formation, and tangle
of ongoing research: see, for example, 关30兴 for a recent survey. It
The concurrence, entanglement of formation, and tangle may be possible to measure entanglement directly, without quantum
are measures of the quantum coherence properties of a mixed state tomography; this possibility was investigated in 关31兴.

052312-11
JAMES, KWIAT, MUNRO, AND WHITE PHYSICAL REVIEW A 64 052312

冉 冊 National Security Agency, and Advanced Research and De-

4
⳵C 3 1 ⳵ra
⫽ 兺
⳵ s ␯ a⫽1
sgn ⫺a
2 2 冑r a ⳵ s ␯
velopment Activity 共ARDA兲, by the Los Alamos National
Laboratory LDRD program, and by the Australian Research

兺 冉 冊 冑 冓 冏 冏 冔
4 Council.
3 1 ⳵ R̂
⫽ sgn ⫺a ␰a ␨
a⫽1 2 2 ra ⳵s␯ a
ˆ MATRICES
APPENDIX A: THE ⌫

兺 兺 冉 冊 冑
4 16
3 1
⫽ sgn ⫺a 具 ␰ a 兩 q̂ ␮ , ␯ s ␮ 兩 ␨ a 典 , One possible set of ⌫ˆ matrices are generators of SU(2)
a⫽1 ␮ ⫽1 2 2 ra
丢 SU(2), normalized so that the conditions given in Eq.
共5.34兲 共3.8兲 are fulfilled. These matrices are

冉 冊 冉 冊
where the function sgn(x) is the sign of the quantity x: it
takes the value 1 if x⬎0 and ⫺1 if x⬍0. Thus sgn(3/2 0 1 0 0 0 ⫺i 0 0
⫺a) is equal to ⫹1 if a⫽1 and ⫺1 if a⫽2,3, or 4. Hence 1 1 0 0 0 1 i 0 0 0
the error in the concurrence is ⌫ˆ 1 ⫽ , ⌫ˆ 2 ⫽ ,
2 0 0 0 1 2 0 0 0 ⫺i

冉 冊
16
⳵C 2
共 ⌬C 兲 2 ⫽ 兺 ⳵s␯
⌳␯
0 0 1 0 0 0 i 0

冉 冊 冉 冊
␯ ⫽1

⫽
16

兺 兺 兺 冋 4

␯ ⫽1 a⫽1 ␮ ⫽1
16

⌫ˆ 3 ⫽
1
1
0 ⫺1
0 0
0
0
0
, ⌫ˆ 4 ⫽
1
0
0
0
0
1
0
0
1
,

冉 冊冑 册 2
2 0 0 1 0 2 1 0 0 0
3 1
⫻sgn ⫺a 具 ␰ a 兩 q̂ ␮ , ␯ s ␮ 兩 ␨ a 典 ⌳ ␯ . 0 0 0 ⫺1 0 1 0 0
2

冉 冊 冉 冊
2 ra
共5.35兲 0 0 0 1 0 0 0 ⫺i
For our example the concurrence is 0.963⫾0.018. 1 0 0 1 0 1 0 0 i 0
Once we know the error in the concurrence, the errors in ⌫ˆ 5 ⫽ , ⌫ˆ 6 ⫽ ,
2 0 1 0 0 2 0 ⫺i 0 0
the tangle and the entanglement of formation can be found
straightforwardly: 1 0 0 0 i 0 0 0

冉 冊 冉 冊
⌬T⫽2C⌬C, 共5.36兲
0 0 1 0 0 0 ⫺i 0

⌬E⫽
冑1⫺C 2
h ⬘
1⫹ 冑1⫺C 2
C
2
⌬C, 冉 冊 共5.37兲 ⌫ˆ 7 ⫽
1
2
0
1
0
0
0
0
⫺1
0
, ⌫ˆ 8 ⫽
1
2
0
i
0
0
0
0
⫺i
0
,

where h ⬘ (x) is the derivative of h(x). For our example the 0 ⫺1 0 0 0 i 0 0

冉 冊 冉 冊
tangle is 0.928⫾0.034 and the entanglement of formation is
0.947⫾0.025. 0 0 0 ⫺i 0 0 0 ⫺1
VI. CONCLUSIONS 1 0 0 ⫺i 0 1 0 0 1 0
⌫ˆ 9 ⫽ , ⌫ˆ 10⫽ ,
2 0 i 0 0 2 0 1 0 0
In conclusion, we have presented a technique for recon-
structing density matrices of qubit systems, including a full i 0 0 0 ⫺1 0 0 0

冉 冊 冉 冊
error analysis. We have extended the latter through to calcu-
lation of quantities of interest in quantum information, such
as the entropy and concurrence. Without loss of generality, 0 0 ⫺i 0 1 0 0 0
we have used the example of polarization qubits of entangled 1 0 0 0 i 1 0 1 0 0
photons, but we stress that these techniques can be adapted ⌫ˆ 11⫽ , ⌫ˆ 12⫽ ,
2 i 0 0 0 2 0 0 ⫺1 0
to any physical realization of qubits.
0 ⫺i 0 0 0 0 0 ⫺1

冉 冊 冉 冊
ACKNOWLEDGMENTS

The authors would like to thank Joe Altepeter, Mauro 0 1 0 0 0 ⫺i 0 0

d’Ariano, Zdenek Hradil, Susana Huelga, Kurt Jacobs, Poul 1 1 0 0 0 1 i 0 0 0
Jessen, James D. Malley, Michael Nielsen, Mike Raymer, ⌫ˆ 13⫽ , ⌫ˆ 14⫽ ,
Sze Tan, and Jaroslav Řeháček for useful discussions and
2 0 0 0 ⫺1 2 0 0 0 i
correspondence. This work was supported in part by the U.S. 0 0 ⫺1 0 0 0 ⫺i 0

052312-12
MEASUREMENT OF QUBITS PHYSICAL REVIEW A 64 052312

冉 冊 冉 冊 冉 冊
1 0 0 0 1 0 0 0 0 0 2 ⫺ 共 1⫹i 兲
1 0 ⫺1 0 0 1 0 1 0 0 1 0 0 ⫺ 共 1⫺i 兲 0
⌫ˆ 15⫽ , ⌫ˆ 16⫽ . M̂ 8 ⫽ ,
2 0 0 ⫺1 0 2 0 0 1 0 2 2 ⫺ 共 1⫹i 兲 0 0
0 0 0 1 0 0 0 1 ⫺ 共 1⫺i 兲 0 0 0
共A1兲 共B1兲

冉 冊 冉 冊
As noted in the text, this is only one possible choice for these 0 0 0 i 0 0 0 1
matrices, and the final results are independent of the choice.
0 0 ⫺i 0 0 0 1 0
M̂ 9 ⫽ , M̂ 10⫽ ,
0 i 0 0 0 1 0 0
APPENDIX B: THE M̂ MATRICES AND SOME OF THEIR
PROPERTIES ⫺i 0 0 0 1 0 0 0

冉 冊
The M̂ matrices, defined by Eq. 共3.15兲, are as follows:

冉 冊
0 0 0 i
2 ⫺ 共 1⫺i 兲 ⫺ 共 1⫹i 兲 1 0 0 i 0
M̂ 11⫽ ,
1 ⫺ 共 1⫹i 兲 0 i 0 0 ⫺i 0 0
M̂ 1 ⫽ , ⫺i 0 0 0
2 ⫺ 共 1⫺i 兲 ⫺i 0 0

冉 冊
1 0 0 0

冉 冊
0 2 0 ⫺ 共 1⫹i 兲
0 ⫺ 共 1⫺i 兲 0 1 1 2 0 ⫺ 共 1⫹i 兲 0
M̂ 12⫽ ,
1 ⫺ 共 1⫹i 兲 2 i ⫺ 共 1⫹i 兲 2 0 ⫺ 共 1⫺i 兲 0 0
M̂ 2 ⫽ , ⫺ 共 1⫺i 兲
2 0 ⫺i 0 0 0 0 0

冉 冊
1 ⫺ 共 1⫹i 兲 0 0

冉 冊
0 0 0 ⫺ 共 1⫹i 兲
0 0 0 1 1 0 0 ⫺ 共 1⫹i 兲 0
M̂ 13⫽ ,
1 0 0 i ⫺ 共 1⫹i 兲 2 0 ⫺ 共 1⫺i 兲 0 2
M̂ 3 ⫽ ,
2 0 ⫺i 0 ⫺ 共 1⫺i 兲 ⫺ 共 1⫺i 兲 0 2 0

冉 冊
1 ⫺ 共 1⫺i 兲 ⫺ 共 1⫹i 兲 2

冉 冊
0 0 0 ⫺ 共 1⫺i 兲
0 0 ⫺ 共 1⫹i 兲 1 1 0 0 ⫺ 共 1⫺i 兲 0
M̂ 14⫽ ,
1 0 0 i 0 2 0 ⫺ 共 1⫹i 兲 0 ⫺2i
M̂ 4 ⫽ ,
2 ⫺ 共 1⫺i 兲 ⫺i 2 ⫺ 共 1⫺i 兲 ⫺ 共 1⫹i 兲 0 2i 0

冉 冊
1 0 ⫺ 共 1⫹i 兲 0

冉 冊
0 ⫺2i 0 ⫺ 共 1⫺i 兲
0 0 2i ⫺ 共 1⫹i 兲 1 2i 0 共 1⫺i 兲 0
M̂ 15⫽ ,
1 0 0 共 1⫺i 兲 0 2 0 共 1⫹i 兲 0 0
M̂ 5 ⫽ ,
2 ⫺2i 共 1⫹i 兲 0 0 ⫺ 共 1⫹i 兲 0 0 0

冉 冊
⫺ 共 1⫺i 兲 0 0 0

冉 冊
0 0 0 1
0 0 0 ⫺ 共 1⫹i 兲 0 0 ⫺1 0
M̂ 16⫽ .
1 0 0 共 1⫺i 兲 2i 0 ⫺1 0 0
M̂ 6 ⫽ ,
2 0 共 1⫹i 兲 0 0 1 0 0 0
⫺ 共 1⫺i 兲 ⫺2i 0 0

冉 冊
The form of these matrices is independent of the chosen
0 0 0 ⫺ 共 1⫹i 兲 set of matrices 兵 ⌫ˆ 其 used to convert the density matrix into a
␯

1 0 0 ⫺ 共 1⫺i 兲 2 column vector. However, the M̂ ␯ matrices do depend on the

M̂ 7 ⫽ , set of tomographic states 兩 ␺ ␯ 典 .
2 0 ⫺ 共 1⫹i 兲 0 0 There are some useful properties of these matrices which
⫺ 共 1⫺i 兲 2 0 0 we will now derive. From Eq. 共3.15兲, we have

052312-13
JAMES, KWIAT, MUNRO, AND WHITE PHYSICAL REVIEW A 64 052312

r a⬘ ⫽r (0)
a ⫹␭r a ⫹␭ r a ⫹•••,
(1) 2 (2)
共C4兲
具 ␺ ␮ 兩 M̂ ␯ 兩 ␺ ␮ 典 ⫽ 兺 具 ␺ ␮ 兩 ⌫ˆ ␭ 兩 ␺ ␮ 典 共 B ⫺1 兲 ␭, ␯ . 共B2兲
␭
兩 ␨ a⬘ 典 ⫽ 兩 ␨ (0)
a 典 ⫹␭ 兩 ␨ a 典 ⫹␭ 兩 ␨ a 典 ⫹•••,
(1) 2 (2)
共C5兲
From Eq. 共3.12兲 we have 具 ␺ ␮ 兩 ⌫ˆ ␭ 兩 ␺ ␮ 典 ⫽B ␮ ,␭ ; thus we ob- 具 ␰ a⬘ 兩 ⫽ 具 ␰ (0)
a 兩 ⫹␭ 具 ␰ a 兩 ⫹␭ 具 ␰ a 兩 ⫹•••, 共C6兲
(1) 2 (2)
tain the result
具 ␺ ␮ 兩 M̂ ␯ 兩 ␺ ␮ 典 ⫽ ␦ ␮ , ␯ . 共B3兲 Writing R̂ ⬘ ⫽R̂ 0 ⫹␭ ␦ R̂, and comparing terms of equal pow-
ers of ␭ in the eigenequations, one obtains the following
If we denote the basis set for the four-dimensional Hilbert formulas:
space by 兵 兩 i 典 (i⫽1,2,3,4) 其 , then Eq. 共3.14兲 can be written as
a 典 ⫽r a 兩 ␨ a 典 ,
R̂ 0 兩 ␨ (0) 共C7兲
(0) (0)
follows:
具 ␰ (0)
a 兩 R̂ 0 ⫽r a 具 ␰ a 兩 ,
(0) (0)
共C8兲
具 i 兩 ␳ˆ 兩 j 典 ⫽ 兺 兺␯ 具 i 兩 M̂ ␯兩 j 典具 ␺ ␯兩 k 典具 l 兩 ␺ ␯ 典具 k 兩 ␳ˆ 兩 l 典 . 共B4兲
k,l
a Î 兲 兩 ␨ a 典 ⫽⫺ 共 ␦ R̂⫺r a 兲 兩 ␨ a 典 ,
共 R̂ 0 ⫺r (0) 共C9兲
(1) (1) (0)

Since Eq. 共B4兲 is valid for arbitrary states ␳ˆ , we obtain the

a 兩 共 R̂ 0 ⫺r a Î 兲 ⫽⫺ 具 ␰ a 兩 共 ␦ R̂⫺r a 兲 .
具 ␰ (1) 共C10兲
(0) (0) (1)
following relationship:
Equations 共C7兲 and 共C8兲 imply that, as might be expected,
兺␯ 具 i 兩 M̂ ␯兩 j 典具 ␺ ␯兩 k 典具 l 兩 ␺ ␯ 典 ⫽ ␦ ik ␦ jl . 共B5兲
a 典 ⫽ 兩 ␨ a典 ,
兩 ␨ (0) 共C11兲
Contracting Eq. 共B5兲 over the indices (i, j) we obtain
具 ␰ (0)
a 兩 ⫽ 具 ␰ a兩 , 共C12兲

兺␯ Tr兵 M̂ ␯ 其 兩 ␺ ␯ 典具 ␺ ␯兩 ⫽Î, 共B6兲 a ⫽r a .

r (0) 共C13兲

Taking the inner product of Eq. 共C9兲 with 具 ␰ a 兩 , and using the
where Î is the identity operator for our four-dimensional Hil- biorthogonal property Eq. 共C3兲, we obtain
bert space.
a ⫽ 具 ␰ a 兩 ␦ R̂ 兩 ␨ a 典 .
r (1) 共C14兲
A second relationship can be obtained by contracting Eq.
共B5兲, viz.,
This implies that
兺␯ 具 i 兩 M̂ ␯兩 j 典 ⫽ ␦ i j , 共B7兲
␦ r a ⬅r a⬘ ⫺r a ⬇ 具 ␰ a 兩 ␦ R̂ 兩 ␨ a 典 . 共C15兲

or, in operator notation, Thus, dividing both sides by some differential increment ␦ x
and taking the limit ␦ x→0, we obtain
兺␯ M̂ ␯ ⫽Î. 共B8兲
⳵ra
⳵x
⫽ ␰a 冓冏 冏冔
⳵ R̂
⳵x a
␨ . 共C16兲
APPENDIX C: PERTURBATION THEORY FOR NON-
HERMITIAN MATRICES Using the completeness property of the eigenstates,
兺 b 兩 ␨ b 典具 ␰ b 兩 ⫽Î, and the identity R̂ 0 ⫽ 兺 b r b 兩 ␨ b 典具 ␰ b 兩 , we obtain
Whereas perturbation theory for Hermitian matrices is the following formula
covered in most quantum mechanics textbooks, the case of
non-Hermitian matrices is less familiar, and so we will 1
present it here. The problem is as follows. Given the 共 R̂ 0 ⫺r a Î 兲 ⫺1 ⫽ 兺 兩 ␨ 典具 ␰ 兩 .
r b ⫺r a b b
共C17兲
b
eigenspectrum of a matrix R̂ 0 关32兴, i.e., b⫽a

具 ␰ a 兩 R̂ 0 ⫽r a 具 ␰ a 兩 , 共C1兲 Applying this to Eq. 共C9兲 we obtain

where
R̂ 0 兩 ␨ a 典 ⫽r a 兩 ␨ a 典 , 共C2兲
a 典 ⬅ 兩 ␨ a⬘ 典 ⫺ 兩 ␨ a 典 ⬇⫺
兩 ␦ ␨ (1) 兺
b
冉 具 ␰ b 兩 ␦ R̂ 兩 ␨ a 典
r b ⫺r a
冊 兩 ␨ b典 .
b⫽a
共C18兲
具 ␰ a 兩 ␨ b 典 ⫽ ␦ a,b , 共C3兲
Similarly, Eqs. 共C10兲 and 共C17兲 imply
we wish to find expressions for the eigenvalues r a⬘ and eigen-
states 具 ␰ a⬘ 兩 and 兩 ␨ a⬘ 典 of the perturbed matrix R̂ ⬘ ⫽R̂ 0 ⫹ ␦ R̂.
We start with the standard assumption of perturbation
theory, i.e., that the perturbed quantities r a⬘ , 具 ␰ ⬘a 兩 , and 兩 ␨ a⬘ 典
具 ␦ ␰ a 兩 ⬅ 具 ␦ ␰ a⬘ 兩 ⫺ 具 ␦ ␰ a 兩 ⬇⫺ 兺
b
冉 具 ␰ a 兩 ␦ R̂ 兩 ␨ b 典
r b ⫺r a
冊 具 ␰ b兩 .
b⫽a
can be expressed as power series of some parameter ␭: 共C19兲

052312-14
MEASUREMENT OF QUBITS
关1兴 G.C. Stokes, Trans. Cambridge Philos. Soc. 9, 399 共1852兲.
关2兴 U. Leonhardt, Measuring the Quantum State of Light 共Cam-
bridge University Press, Cambridge, 1997兲.
关3兴 Z. Hradil, Phys. Rev. A 55, R1561 共1997兲.
关4兴 S.M. Tan, J. Mod. Opt. 44, 2233 共1997兲.
关5兴 K. Banaszek, G.M. D’Ariano, M.G.A. Paris, and M.F. Sacchi,
Phys. Rev. A 61, 010304 共1999兲.
关6兴 Z. Hradil, J. Summhammer, G. Badurek, and H. Rauch, Phys.
Rev. A 62, 014101 共2000兲.
关7兴 J. Řeháček, Z. Hradil, and M. Ježek, Phys. Rev. A 63, 040303
共2001兲.
关8兴 J.R. Ashburn, R.A. Cline, P.J.M. Vanderburgt, W.B. Wester-
veld, and J.S. Risley, Phys. Rev. A 41, 2407 共1990兲.
关9兴 D.T. Smithey, M. Beck, M.G. Raymer, and A. Faridani, Phys.
Rev. Lett. 70, 1244 共1993兲; see also the discussion of polariza-
tion effects in M.G. Raymer, D.F. McAlister, and A. Funk, in
Quantum Communication, Computing, and Measurement ’98,
edited by P. Kumar 共Plenum, New York, 2000兲, pp. 147-162.
关10兴 G. Breitenbach, S. Schiller, and J. Mlynek, Nature 共London兲
387, 471 共1997兲.
关11兴 J.W. Wu, P.K. Lam, M.B. Gray, and H.-A. Bachor, Opt. Ex-
press 3, 154 共1998兲.
关12兴 T.J. Dunn, I.A. Walmsley, and S. Mukamel, Phys. Rev. Lett.
74, 884 共1995兲.
关13兴 D. Leibfried, D.M. Meekhof, B.E. King, C. Monroe, W.M.
Itano, and D.J. Wineland, Phys. Rev. Lett. 77, 4281 共1996兲; D.
Leibfried, T. Pfau, and C. Monroe, Phys. Today 51共4兲, 22
共1998兲.
关14兴 C. Kurtsiefer T. Pfau, and J. Mlynek, Nature 共London兲 386,
150-153 共1997兲.
关15兴 G. Klose, G. Smith, and P.S. Jessen, Phys. Rev. Lett. 86, 4721
共2001兲.
关16兴 I.L. Chuang, N. Gershenfeld, and M. Kubinec, Phys. Rev. Lett.
80, 3408 共1998兲.
关17兴 S.L. Braunstein, C.M. Caves, R. Jozsa, N. Linden, S. Popescu,
and R. Schack, Phys. Rev. Lett. 83, 1054 共1999兲.
052312-15
PHYSICAL REVIEW A 64 052312
关18兴 A.G. White, D.F.V. James, P.H. Eberhard, and P.G. Kwiat,
Phys. Rev. Lett. 83, 3103 共1999兲.
关19兴 A recent overview of many quantum computation technologies
is given by S. Braunstein and Ho.-K. Lo, Scalable Quantum
Computers: Paving the Way to Realization 共Wiley, New York,
2001兲; see also Fortschr. Phys. 48, 共9–11兲 共2000兲.
关20兴 E. Hecht and A. Zajac, Optics 共Addision-Wesley, Reading,
MA, 1974兲, Sec. 8.12.
关21兴 E. Wolf, Nuovo Cimento 13, 1165 共1959兲; L. Mandel and E.
Wolf, Optical Coherence and Quantum Optics 共Cambridge
University Press, Cambridge, 1995兲, Chap. 6.
关22兴 P.G. Kwiat, E. Waks, A.G. White, I. Appelbaum, and P.H.
Eberhard, Phys. Rev. A 60, R773 共1999兲.
关23兴 A.G. White, D.F.V. James, W.J. Munro and P.G. Kwiat, e-print
quant-ph/0108088.
关24兴 A. Berglund, Undergraduate thesis, Dartmouth College, 2000;
A. Berglund, e-print quant-ph/0010001.
关25兴 W.H. Press, S.A. Teukolsky, W.T. Vetterling, and B.P. Flan-
nery, Numerical Recipes in Fortran 77: The Art of Scientific
Computing, 2nd ed. 共Cambridge University Press, Cambridge,
1992兲, Sec. 10.5.
关26兴 A.C. Melissinos, Experiments in Modern Physics 共Academic
Press, New York, 1966兲, Sec, 10.4, pp. 467– 479.
关27兴 M.A. Nielsen and I.L. Chuang, Quantum Computation and
Quantum Information 共Cambridge University Press, Cam-
bridge, 2000兲, Chap. 11.
关28兴 L.I. Schiff, Quantum Mechanics, 3rd ed. 共McGraw-Hill, New
York, 1968兲, Eq. 共31.8兲, p. 246.
关29兴 W.K. Wootters, Phys. Rev. Lett. 80, 2245 共1998兲; V. Coffman,
J. Kundu, and W.K. Wootters, Phys. Rev. A 61, 052306 共2000兲.
关30兴 B.M. Terhal, e-print quant-ph/0101032.
关31兴 J.M.G. Sancho and S.F. Huelga, Phys. Rev. A 61, 042303
共2000兲.
关32兴 The properties of the eigenvectors and eigenvalues of non-
Hermitian matrices are discussed by P. M. Morse and H.
Feshbach, Methods of Theoretical Physics 共McGraw-Hill, New
York, 1953兲, Vol. I, p. 884 et seq.