Generalized concurrence measure for faithful quantification of multiparticle pure state

entanglement using Lagranges identity and wedge product

Vineeth S. Bhaskara1, and Prasanta K. Panigrahi2,
1

arXiv:1607.00164v1 [quant-ph] 1 Jul 2016

Indian Institute of Technology Guwahati, Guwahati 781 039, Assam, India

Indian Institute of Science Education and Research Kolkata, Mohanpur 741 246, West Bengal, India
(Dated: July 4, 2016)
Concurrence, introduced by Hill and Wootters [Phys. Rev. Lett. 78, 5022 (1997)], provides an
important measure of entanglement for a general pair of qubits that is faithful: strictly positive for
entangled states and vanishing for all separable states. Such a measure captures the entire content of entanglement, providing necessary and sufficient conditions for separability. We present an
extension of concurrence to multiparticle pure states in arbitrary dimensions by a new framework
using the Lagranges identity and wedge product representation of separability conditions, which
coincides with the I-concurrence of Rungta et al. [Phys. Rev. A 64, 042315 (2001)] who proposed by extending Wootterss spin-flip operator to a so-called universal inverter superoperator.
Our framework exposes an inherent geometry of entanglement, and may be useful for the further
extensions to mixed and continuous variable states.
PACS numbers: 03.65.Ud, 03.67.-a, 02.10.Ud

I.

INTRODUCTION

A deeper understanding of inseparability or entanglement is of fundamental importance for the understanding of intrinsic quantum correlations. It has far reaching
applications in quantum computation and information
theory [1]. Entanglement forms an elementary resource
in quantum computation and various quantum communication protocols [2, 3]. Detecting and quantifying this
resource is of great practical application.
Quantifying entanglement faithfully in a multiparticle
scenario is central to quantum information theory so that
one can estimate how close quantum states are to classical ones, and characterize the efficiency of protocols deterministically, which use entanglement as a resource [4
6]. Recent interest on the connections between quantum
entanglement and the emergence of space-time [7, 8] also
calls for a systematic study of the geometry-entanglement
relationship with the quantification of entanglement playing a subtler role in the context of quantum gravity.
For the two-qubit case, an important measure of entanglement is the concurrence [9], which is strictly positive for entangled states, and vanishing for separable
states. It provides the necessary and sufficient conditions of separability for a general pair of qubits. An extension of concurrence for multiparticle pure states is the
I-concurrence introduced by Rungta et al. [10]. They
generalized the spin-flip superoperator to act on quantum
systems of arbitrary dimensions, and introduced the corresponding generalized concurrence for joint pure states
of bipartite quantum systems.
In this paper, we present a similar generalization of
concurrence to multiparticle pure states of arbitrary di-

bhaskaravineeth@gmail.com
pprasanta@iiserkol.ac.in

mensions that is faithful by a new framework using the

Lagranges identity and wedge product representation,
leading to a measure of entanglement identical to the
I-concurrence. This framework reveals an essential geometry of entanglement and may be useful for further
extension of concurrence to other complex systems of interest.
There have been works on a similar spirit, of which
some include the study by Sawicki et al. [11] on the
symplectic geometry of entanglement, Nielsen [12] on the
connection between the algebra of majorization and entanglement transformations, Zhu [13] on the structure of
quantum correlations of many-body systems, Duan et al.
[14] and Simon [15] on the entanglement in continuous
variable systems.

II.

SEPARABILITY FOR PURE MULTI-QUDIT

STATES, AND THE CENTRAL RESULT

For future convenience, we define separability for pure

multi-qudit (d dimensional) states. Consider, a nparticle, d-dimensional quantum system. Let P |Q be a
bipartition of this composite(whole) system P Q, with
respective Hilbert spaces HP and HQ for the states of the
sub-systems P and Q, then the state space of the composite system is given by the tensor product H = HP HQ .
If a pure state |i H of the composite system can be
written in the form
|i = |iP |iQ ,
where |iP HP and |iQ HQ are the pure states of
the sub-systems P and Q respectively, then the system is
said to be separable across the bipartition P |Q. Alternatively, the sub-system P is separable from the composite
system P Q. Otherwise, the sub-systems P and Q are
said to be entangled.

2
To state the central result of the paper, consider a nqudit system with qudits labelled by (k), k = 1, 2, ..., n.
Suppose |i is any pure state of the system and =
|ih| be its density matrix. Let M be the set of the
particular qudits whose bipartite separability from the
composite system is of interest with cardinality m (< n).
Then the generalized concurrence, EM , for the bipartition M| M is given, in equivalent forms, as (M being
the complementary set of M):
2
EM
=4

X
i<j

M
M M
M
ii jj ij ji



 M 2
= 2 1 tr ( ) ,

=4

i j

i<j

a b =

def P
where M =
j hj|M (|ih|) |jiM = TrM () is the
reduced density matrix on the subsystem M of qudits
obtained by tracing out the subsystem M, and i are
the eigenvalues of M .
EM vanishes iff the system is separable across the bipartition M| M and takes the maximum value iff M
is maximally mixed. A measure of global entanglement
would then be the sum of measures for different bipartitions of the system. Evidently, a composite system
is fully-separable if and only if all the single-qudit bipartitions are separable. Therefore, the necessary and
sufficient
criterion for full-separability of the system is
Pn
2
(k)
E
is the single-qudit reduced
k=1 (k) = 0, where
density matrix of on qudit (k).
One can arrive at the result by considering the simple
case of a two-qubit system, and subsequently generalizing
the framework to multiple qudit systems.

III. TWO-QUBIT CONCURRENCE USING

LAGRANGES IDENTITY AND WEDGE
PRODUCT FRAMEWORK

Consider, a two-qubit system with qubits A and B. Let

|i be a normalized pure state of the system with
|i = p|0A 0B i + q|0A 1B i + r|1A 0B i + s|1A 1B i
(p, q, r, s C). Rewriting the state as:




|i = |0A i p|0B i + q|1B i + |1A i r|0B i + s|1B i (1)
= |0A i h0A |i + |1A i h1A |i,

the bipartition A|B is separable if and only if the vectors

h0A |i = p|0B i + q|1B i and h1A |i = r|0B i + s|1B i (or,
equivalently h0B |i and h1B |i) are parallel, i.e., if and
only if
p
q
= ps qr = 0.
r
s

using the notation of a wedge product, which generalizes

easily to multiparticle systems in arbitrary d dimensions,
as we show subsequently.
In geometric algebra [16], the wedge product of two
vectors is seen as a particular generalization of cross
product to higher dimensions, and is defined as follows.

Consider, any two vectors

a and b in Cm written
in the orthonormal basis {
ei }m
i=1 . Their wedge product is a bivector in the m C2 dimensional exterior space
with basis {
ei }m
ej }m
i=1 {
j=1 defined, by stipulating that
ei ej =
ej ei and ei ei = 0, as:

(2)

Then the modulus of ps qr is a faithful measure of entanglement for two qubits, which vanishes only for separable states. This condition may be elegantly written

m1
X

m
X

i=1 j=i+1

(ai bj aj bi ) ei ej ,

(3)

with
a
a = 0 and
a b = (1) b
a . In the

coordinate notation
a b may be written as:
(a1 b2 a2 b1 , a1 b3 a3 b1 , ..., a1 bm am b1 ,
a2 b3 a3 b2 , ..., a2 bm am b2 , ..., am1 bm am bm1 ).
This representation allows one to write the separability conditions in a compact and useful form. We note
||h0A |ih1A |i|| = ||h0B |ih1B |i|| = |psqr|, which
is the measure of entanglement for the case of a two-qubit
pure state. ||h0A |i h1A |i|| geometrically represents
the area of the complex parallelotope formed by the vectors h0A |i and h1A |i in the Hilbert space of qubit B.
We write the two-qubit measure of entanglement as E =
2||h0A |i h1A |i|| = 2||h0B |i h1B |i|| = 2|ps qr|,
which is the concurrence [9] for two-qubit pure states de = 2|psqr|,
fined by Hill and Wootters 
as C() = |h|i|
= y | i, y = 0 i and | i is the comwhere |i
i 0
plex conjugate of |i.
For maximal entanglement by this measure, the area
of the parallelotope, |ps qr|, must be maximum, which
implies that the parallelotope must be a square with its
sides taking the maximum possible value. As the sum
of the squares of the sides is constrained to be 1 (by
normalization), i.e., |h0A |i|2 + |h1A |i|2 = 1, the area is
maximized when each of the side of the square equals 12 .
Then, E(max) = 1, 0 E 1. Therefore, for maximal
entanglement,
1
1
|h0A |i| = |h1A |i| = , |h0B |i| = |h1B |i| = ,
2
2
(h0A |i) h1A |i = 0, (h0B |i) h1B |i = 0.
These conditions for maximal entanglement are identical to the condition of the reduced density matrix being
maximally mixed.
This representation enables us to solve for the conditions of seperability, i.e., conditions for the composite
Hilbert space to be expressible as a tensor product of two
smaller Hilbert spaces corresponding to pure states of a

3
seperable bipartition, using a generalized form of the Lagranges identity, which becomes non-trivial for complex
systems consisting of multiple particles in d dimensions.
Recall the generalized Lagranges identity [17] for vectors in Cm , which is a generalization of the BrahmaguptaFibonacci identity [18] and a special form of the Binet-

m
X

k=1

|ak |2

m
X

k=1

|bk |2

2

m1
m
m

X
X X


|ai bj aj bi |2

ak b k  =


k=1

where bk represents the complex conjugate of bk (see Appendix A for proof). The norm of the wedge product

a b calculated by RHS of Eq. (4) takes O(m2 ) steps,

while calculating the same using the LHS takes only
O(m) steps. Therefore, this identity when applied to the
wedge product representation of the separability conditions results in a computationally lesser intensive expression, asymptotically with increasing number of particles
and dimensions, in terms of the traces of the squared
reduced density matrices of the pure state.
2
By this identity, one may write EA
= 4||h0A |i
h1A |i||2 = 4||(p, q) (r, s)||2 as 4(|p|2 + |q|2 )(|r|2 +

2
2
|s| ) 4|pr + qs| . By noting this to be the determi-

nant of the reduced density matrix on

definition, as in this case is
2
|p| pq pr ps
qp |q|2 qr qs
=
rp rq |r|2 rs
sp sq sr |s|2

qubit A (A ), by

and therefore the reduced density matrix on A, , takes

the form:
A = h0B ||0B i + h1B ||1B i

 2
|p| + |q|2 pr + qs
,
=
rp + sq |r|2 + |s|2

one may, thus, rewrite

the two-qubit
measure of entanp
p
glement as E = 2 det(A ) = 2 det(B ). This may be
rewritten as
X


A
A A
2
A
EA
= 4 det(A ) = 4
ii jj ij ji
i<j

X


1
A
A A
A
= 4
ii jj ij ji
2 i,j

 
1
A 2
A 2
[tr( )] tr[( ) ]
=4
2


A 2
= 2 1 tr[( ) ] ,

Cauchy identity [19, 20]. Consider, two vectors

a, b
Cm . Then the Lagranges identity takes the form:

k
a k 2 k b k 2 |
a b |2 = k
a b k2 (k k representing the
norm of a vector and | |, the modulus of a scalar), i.e.,

(5)
(6)

(4)

i=1 j=i+1

since the trace of a valid P

density matrix is unity and
for any square matrix M , i,j Mij Mji = tr(M 2 ), and
P
P
P
2
j Mjj = tr(M ) .
i Mii
i,j Mii Mjj =
The characteristic polynomial of a m m matrix M in
t is given by:


1
tm (tr M )tm1 +
tr(M )2 tr(M 2 ) tm2
2
+ + (1)m (det M ).

So Eq. (5) is the tm2 coefficient (except for a constant)

of the characteristic polynomial of the m m reduced
density matrix A . This can be thought of as the first
step, interpolating between the trace of A (which is the
tm1 coefficient) and the determinant of A (which is
the constant coefficient). The roots of the characteristic
polynomial are precisely the eigenvalues of A . If the
eigenvalues of A are 1 , . . . , m then
!#
"
X
X
1
2
2
i j
(7)
=4
i
1
EA = 4
2
i<j
i
[21]. This mathematical setting extends in a straightforward way to more general cases in higher dimensions, and
a global faithful measure of entanglement may be written
down by summing over the entanglement contribution of
each of the independent bipartitions of the general pure
state as we show subsequently.
IV.

EXTENSION TO MULTIPARTICLE,
d-DIMENSIONAL STATES

a. Three-qubit states: Consider the three-qubit case.

Let a normalized pure state of the three-qubit system
be |i with density matrix = |ih| and with qubits
labelled by A, B and C. Let
|i = p|0A 0B 0C i + q|0A 0B 1C i + r|0A 1B 0C i + s|0A 1B 1C i
+ t|1A 0B 0C i + u|1A 0B 1C i + v|1A 1B 0C i + w|1A 1B 1C i


= |0A i p|0B 0C i + q|0B 1C i + r|1B 0C i + s|1B 1C i


+ |1A i t|0B 0C i + u|0B 1C i + v|1B 0C i + w|1B 1C i
= |0A i h0A |i + |1A i h1A |i

4
(p, q, r, s, t, u, v, w C). Similar to the two-qubit case,
for separability of qubit A (i.e., the bipartition A|BC)
here, the vectors h0A |i, h1A |i must be parallel. This
yields the condition for separability of qubit A to be:
q
r
s
p
= = = .
t
u
v
w

(8)

The separability condition may be written in the wedge

product representation as h0A |i h1A |i = 0, which
is equivalent to the relations in Eq. (8) on crossmultiplying, since:
(p, q, r, s) (t, u, v, w) = (pu qt, pv rt, pw st,
qv ru, qw su, rw sv),
by the coordinate notation of the wedge product defined
previously. Therefore, the bipartite separability A|BC
h0A |i h1A |i = 0. Hence, its norm is a deterministic measure of entanglement of qubit A with system BC.
By the Lagranges identity, kh0A |i h1A |ik2 turns out
to be equal to the determinant of qubit As reduced density matrix A by definition, similar to the previous case.
Therefore, one can write the global measure of entanglement for a three-qubit system, considering independent
bipartitions, as:
E = EA + EB + EC
= 2||h0A |i h1A |i|| + 2||h0B |i h1B |i||+
2||h0C |i h1C |i||

q
q
q
det(A ) + det(B ) + det(C ) .
=2
This can be rewritten in terms of eigenvalues of the reduced density matrices by the derivation in Eq. (7).
The maximum norm of each of the wedge products
above is = 12 .
Therefore, 0 E 3.
The

2
EAB
=4

||h0A 0B |i h0A 1B |i||2 + ||h0A 0B |i h1A 0B |i||2 + ||h0A 0B |i h1A 1B |i||2 +


||h0A 1B |i h1A 0B |i||2 + ||h0A 1B |i h1A 1B |i||2 + ||h1A 0B |i h1A 1B |i||2 .

Therefore, separability of bipartition AB|CD EAB =

0. Again by the Lagranges identity, the expression Eq.
2
(9) for EAB
simplifies to the similar form as:
2

2
EAB

=4

2
X

i,j=1,i<j

|GHZi3 state is maximally entangled three-qubit state

with E = 3, by this measure, and for the |W i3
state, E = 2 2 2.828, which suggests that it is
highly entangled but lesser than |GHZi3 state, where:
|GHZi3 = 12 (|0A 0B 0C i + |1A 1B 1C i) and |W i3 =
1 (|1A 0B 0C i + |0A 1B 0C i + |0A 0B 1C i) .
3
Analogously, for a n-qubit system with pure state |i
and density operator , separability of qubit labelled by
i ( n) h0i |i h1i |i = 0. By Lagranges identity
this simplifies to: det(i ) = 0. Therefore, a particular
qubit is separable from a n-qubit system if and only if
its corresponding single-qubit reduced density matrix is
singular. For full-separability of the system, every singlequbit reduced density matrix being singular is necessary
and sufficient.
b. Four-qubit states: We analyze the above construction for the case of a four-qubit system and then
for the case of two-qutrits to assess the generalization
and arrive at our result for pure multiparticle states
in arbitrary dimensions. Consider, a four-qubit system with qubits labelled by A, B, C and D. Let
|i be its pure state with density
p matrix = |ih|.
EA = 2kh0A |i h1A |ik = 2 det(A ) determines the
separability of qubit A or qubit system (BCD) from
the composite system (ABCD), similar to the previous
cases. Analogous to the previous construction, for the
separability of qubits (AB) or (CD) from the system,
the vectors h0A 0B |i, h0A 1B |i, h1A 0B |i, h1A 1B |i in
the Hilbert space HCD of qubit system (CD)
must be parallel.
This can be seen by writ
ing |i as |0A 0B i h0A 0B |i + |0A1B i h0A 1B |i +
|1A 0B i h1A 0B |i + |1A 1B i h1A 1B |i . Therefore, a
non-vanishing wedge product of one of the vectors
with any other among h0A 0B |i, h0A 1B |i, h1A 0B |i and
h1A 1B |i, indicates entanglement of the sub-systems
(AB) and (CD). Therefore, define EAB as:

X


AB
AB AB
i j
=4
AB
ii jj ij ji
i<j

= 2 1 tr[(AB )2 ] ,
where i are the eigenvalues of AB . Note that the term


P22
AB AB
AB AB
above is not the deteri,j=1,i<j ii jj ij ji

(9)

minant of AB . Therefore, the generalizing expression

is in terms of the traces of the squared reduced density
matrices but not in terms of their determinants for gen2
eral cases (n-qudits). Similar expressions follow for EAC
2
and EAD . Considering independent bipartitions one can
write the global measure of entanglement for the fourqubit system as:
E = EA + EB + EC + ED + EAB + EAC + EAD .
Evidently, E takes the maximum value only when the reduced density matrices
are maximally mixed. Therefore,

E(max) = 4 + 3 2 6 7.674 for maximal entanglement,

5
by this measure. This cannot be attained for the case
of a four-qubit system, as shown by Higuchi et al. [22].
Therefore, 0 E < 4 + 3 2 6 .
For |GHZi4 = 12 (|0A 0B 0C 0D i + |1A 1B 1C 1D i) state,
E = 7, and for the four-qubit Higuchi-Sudbery state
found numerically by Higuchi et al. [22]:
1
|HSi = [|0011i + |1100i + (|1010i + |0101i)
6
+ 2 (|1001i + |0110i)],

where = e2i/3 , E = 4 + 2 3 7.464, which is close

to the unattainable bound of 7.674, showing that it is
more entangled than the |GHZi4 state, by this measure.
c. Two-qutrit state: One can analyze the case of a
two-qutrit system to construct a similar measure of entanglement in arbitrary dimensions, i.e., for qudits (dlevel systems), that fits into the above generalizing structure observed for arbitrary number of qubits. Consider,
a two-qutrit system with levels |0i, |1i, |2i and qutrits labeled by A and B. Let |i be its pure state and its density matrix. Similar to the previous reasoning, for separability of qutrit A, the vectors h0A |i, h1A |i, h2A |i
must be parallel. This is clear once the state is written
as: |i = |0A i (h0A |i) + |1A i (h1A |i) + |2A i (h2A |i).
Therefore, define the measure of entanglement of qutrit
A with qutrit B as:

2
EA
= 4 ||h0A |ih1A |i||2 + ||h0A |i h2A |i||2

+ ||h1A |i h2A |i||2 .
Applying the Lagranges identity, one discovers the generalizing expression for entanglement in qutrits to be of
the similar form:
2
EA

=4

3
X

A
ii

A
jj

i,j=1,i<j

A 2

A
ij

A
ji

=4

sential geometry and mathematical structure, and is of

relevance to various related problems like separability of
mixed states and continuous variable systems, classification of entanglement transformations, and entanglement
characterization. This framework gives a faithful, computable measure of entanglement for pure states, and
may further be useful in generalizing concurrence for
mixed and continuous variable states. The measure may
also be used in numerical searches for highly entangled
multiparticle states [2325], without missing any useful state, to improve existing and discover new quantum
information processing protocols [26, 27].
Acknowledgments

Bhaskara is thankful to Prof. Oliver Knill (Harvard)

for pointing out the Binet-Cauchy identity. This work
was supported by the National Initiative on Undergraduate Science (NIUS) undertaken by the Homi Bhabha
Centre for Science Education, Tata Institute of Fundamental Research (HBCSETIFR), Mumbai, India. The
authors acknowledge Prof. Vijay Singh and Dr. Praveen
Pathak for continuous encouragement.
Appendix: Proof of Lagranges identity

Consider,

RHS = ||
a b ||2
=
=

i j

i<j

= 2 1 tr[( ) ] ,
where i are the eigenvalues of A .
One thus arrives at the result for pure multi-qudit
states by noting the generalizing structure from the various cases above. A global measure of entanglement for
the multi-qudit system can be constructed by summing
over the measures for different bipartitions of the system.

=
=

CONCLUSION

We hope our work provides new insights into the deeply

interesting phenomenon of entanglement, exposing its es-

[1] M. A. Nielsen and I. L. Chuang, Quantum Computation and Quantum Information, Cambridge Univ. Press

m
X

i=1 j=i+1
m X
m
X

1
2

i=1 j=1

|ai bj aj bi |2
|ai bj aj bi |2

m m
1 XX

(ai bj aj bi )(ai bj aj bi )

1
2

(|ai |2 |bj |2 2Re(ai bj aj bi ) + |aj |2 |bi |2 )

i=1 j=1
m X
m
X
i=1 j=1

m
X
i=1

=
V.

m1
X

! m
m X
m
X
X
(ai bj aj bi )
|bj |2 Re
|ai |2
j=1

i=1 j=1


2
! m
m
m
X

X
X


ai b i 
|bj |2 
|ai |2


i=1

j=1

i=1

= k
a k 2 k b k 2 |
a b |2 = LHS.
Hence the identity.

(2002).

6
[2] D. Bouwmeester, J. W. Pan, K. Mattle, M. Eibl, H. Weinfurter and A. Zeilinger, Nature 390, p. 575 (1997).
[3] Q. Zhang, A. Goebel, C. Wagenknecht, Y. A. Chen, B.
Zhao, T. Yang, A. Mair, J. Schmiedmayer and J. W. Pan,
Nature Phys. 2, p. 678 (2006).
[4] V. Vedral, Nature Phys. 10, p. 256 (2014); and references
therein.
[5] B. Zhang, X. Liu, J. Wang and C. Tang, Int. J. Theor.
Phys. 55, p. 1601 (2016).
[6] X. Tan, X. Zhang and J. Fang, Inf. Proc. Lett. 116, Issue
5, p. 347 (2016).
[7] R. Cowen, Nature 527, p. 290 (2015).
[8] M. Van Raamsdonk, Gen. Relativ. Grav. 42, p. 2323
(2010).
[9] S. Hill and W. K. Wootters, Phys. Rev. Lett. 78, 5022
(1997).
[10] P. Rungta, V. Buzek, C. M. Caves, M. Hillery and G. J.
Milburn, Phys. Rev. A 64, 042315 (2001).
[11] A. Sawicki, A. Huckleberry and M. Kus, Commun. Math.
Phys. 305, Issue 2, p. 441 (2011).
[12] M. A. Nielsen, Phys. Rev. Lett. 83, 436 (1999).
[13] J. Zhu, Int. J. Theor. Phys., p. 1 (2016).
[14] L. M. Duan, G. Giedke, J. I. Cirac and P. Zoller, Phys.
Rev. Lett. 84, 2722 (2000).
[15] R. Simon, Phys. Rev. Lett. 84, 2726 (2000).

[16] C. J. L. Doran and A. N. Lasenby, Geometric Algebra for

Physicists, Cambridge Univ. Press (2003).
[17] L. V. Ahlfors, Complex Analysis (3rd ed.), New York,
McGraw-Hill Book Co., p. 9 (1978).
[18] J. Stillwell, Mathematics and Its History (2nd ed.),
Springer, p. 72 (2002).
[19] M. Marcus and H. Minc, Introduction to Linear Algebra,
New York, Dover Publications (1965).
[20] O. Knill, Linear Algebra Appl. 459, p. 522 (2014).
[21] B. L. van der Waerden, Algebra, Vol. 1, New York,
Springer-Verlag (1993).
[22] A. Higuchi and A. Sudbery, Phys. Lett. A 273, 213
(2000).
[23] M. Enriquez, I. Wintrowicz and K. Zyczkowski, J. Phys.:
Conf. Ser. 698, 012003 (2016).
[24] I. D. K. Brown, S. Stepney, A. Sudbery and S. L. Braunstein, J. Phys. A 38, 1119 (2005).
[25] A. Borras, A. R. Plastino, J. Batle, C. Zander, M. Casas
and A. Plastino, J. Phys. A 40, 13407 (2007).
[26] S. Muralidharan and P. K. Panigrahi, Phys. Rev. A 77,
032321 (2008).
[27] C. Gao, S. Ma and W. Chen, Int. J. Theor. Phys. 55,
Issue 5, p. 2643 (2016).