Quantum Entanglement

An Introduction by Andreas Gabriel

October 2009

1 Motivation
Although already discovered in 1935 by Erwin Schrödinger, quantum entanglement only
began to be taken seriously by the scientific community at the end of the twentieth century.
It soon gave rise to the several fields of research - most importantly to quantum information
theory. Today, quantum information technology is on the verge of making its way into
everyday life, being the first application of quantum mechanics which makes use of single
quanta. Concepts like quantum cryptography are already nearing practical use, while other
applications – like quantum computation – still need to work out many details.
Quantum entanglement is basically a form of correlation between two or more parties (which
may or may not be spatially separated) exceeding any classical correlation in magnitude.
Determining if a given state is entangled or not is in general a very difficult task.

2 Formalism
2.1 States
In quantum information theory, pure quantum states are described by finite-dimensional
complex vectors

|Ψi ∈ H = C n (2.1)

on the Hilbert space H. However, in reality, states are usually not pure, since they cannot be
fully isolated from the environment. Therefore, one has to consider mixed states, described
by density matrices of the form
X
ρ= pn |Ψn i hΨn | (2.2)
n

where the {pn } are a probability distribution for the states {|Ψn i} accordingly. Density
matrices always satisfy

ρ† = ρ Tr(ρ) = 1 ρ≥0 (2.3)

where the last inequality is to be understood as an operator-inequality, i.e. it means that all
eigenvalues of ρ are nonnegative.

1
2.2 Tensor Products
Multipartite systems can be described by elements of product spaces , which are obtained
by forming tensor products of unipartite Hilbert spaces:
HA,B,... = HA ⊗ HB ... (2.4)
In this letter, only bipartite systems will be discussed. Nevertheless, most concepts can be
extended to multipartite systems in a straightforward fashion.
Uncorrelated systems can be described by direct tensor products of states
   
|Ψi = ΨA ⊗ ΨB (2.5)
or, using the density matrix formulism
ρ = ρA ⊗ ρB (2.6)
However, product spaces also contain states which cannot be expressed as products of states
from the subsystems.

3 Defining Entanglement
A pure quantum state is called separable iff it can be written as a direct product, i.e.
   
|Ψsep i = ΨA ⊗ ΨB (3.1)
Else, it is called entangled and has the form
X    
|Ψent i = ci ΨA
i ⊗ ΨB
i (3.2)
i

with at least two non-vanishing complex coefficients ci 1 .

For density matrices, the definitions are a bit more subtle. A density matrix ρ is separable
iff there exists at least one decomposition of the form
X
ρ= pn ρ A B
n ⊗ ρn (3.3)
n

while for an entangled state there is no decomposition into separable states.

4 Detecting Entanglement
4.1 Pure States
For pure states, it is rather easy to find out if a given state is separable or entangled. any
given pure state |Ψi is entangled iff
S(ρA ) > 0 (4.1)
1
Note that separable states can also be written as a superposition of several product states, whereas
entangled states can never be written as direct products

2
where S is any suitable entropy2 and the reduced density matrix is defined by

ρA = TrB |Ψi hΨ| (4.2)

This can be understood by realising, that the entropy of a pure state always vanishes and
entanglement can be considered as information about a composite state, that does not apply
to any one of the single parties but only to the composite system. By tracing over one
subsystem, this information is lost and the remaining (reduced) state becomes mixed.

4.2 Mixed States

For mixed states, this task is much more complicated. Since up to now now method was
found capable of distinguishing between entangled or separable states in general, one has to
settle with necessary separability criteria. If one of these is violated by a state, is has to be
entangled, while nonviolation does not represent a conclusive result.

4.2.1 Entanglement Witnesses

An important example for such criteria is the so called entanglement witness theorem. Since
the set of all separable states S is a convex one3 , it can be enclosed within hyperplanes sep-
arating it from states outside. In particular, for every entangled state there exists at least
one such hyperplane separating it from all separable states (as illustrated in Fig. 1). Math-

Fig. 1: Illustration of the entanglement witness theorem

ematically this can be formulated by witness operators W, which are defined as operators
satisfying

Tr(W σ) ≥ 0∀σ ∈ S ∃ρ : Tr(ρW ) < 0 (4.3)

Although this method is in principle capable of detecting all entangled states, it can be very
difficult to find apropriate witness operators for a given state.
2
For example the von-Neumann-Entropy: S(ρ) = −Trρ ln ρ
3
This can easily be seen from the definition of separable density matrices.

3
4.2.2 The PPT Criterion
The partial transposition (i.e. transposition in one of two subsystems) of a state ρ is defined
as

ρTB = (1 ⊗ T )ρ (4.4)

where T is the transposition operator. If now ρ is separable, it immediately follows that

X
ρTB = pn ρA ⊗ ρBT ≥ 0 (4.5)
n

where the inequality is to be understood as an operator inequality again. The necessary

separability criterion can therefore be formulated like follows: Separable states necessarily
have to stay positive semidefinite under partial transposition.
This criterion has proven to be very valuable, since it is easily computable and comperatively
stong (i.e. it is capable of detecting many entangled states).

4.3 Examples
Consider the mixed state
 
1−α 0 0 0
 i
− 1 − α 1  0 1 + α −2α 0
ρ α = α Ψ Ψ  + 1=   (4.6)
4 4 0 −2α 1 + α 0
0 0 0 0

where |Ψ− i is the singlet-Bell-state, 1/4 is the maximally mixed separable state and α > 0
is a free parameter making this state range from pure and maximally entangled to separable
and maximally mixed.
Let us first apply the PPT-criterion. The partially transposed density matrix is
 
1−α 0 0 −2α
1 0 1+α 0 0 
ρTB =   (4.7)
4 0 0 1+α 0 
−2α 0 0 0

with the eigenvalues λ1,2,3 = 1 + α and λ4 = 1 − 3α. Thus, this state has to be entangled if
α > 1/3.
The same result can be obtained by means of entanglement witnesses. Using for example4
 
1 0 0 0
0 0 1 0
W = 0 1 0 0
 (4.8)
0 0 0 1

one obtains Tr(ρα W ) = (1 − 3α)/2, which is negative iff α > 1/3.

4
The proof that the given operator is indeed an entanglement witness will be omitted here.