QST Theory: Quantum Information (WiSe 22/23) Last updated: 2022-12-01

Chapter 2: Entanglement and Non-Locality

Instructor: Dr. Christian Schilling Scriber: Ignacio Cirac and Sirui Lu

1 Introduction
Quantum Physics is an old theory, more than 120 years old. Soon after being developed, some
scientists noticed that it has consequences beyond quantum physics. In particular, the existence of
superpositions was very astonishing, since they imply a non-realist view of Nature. That is, that,
before measurement, the properties of objects may not be defined, and only after we measure they
do. Some scientists attributed that to the fact that the theory makes statistical predictions, and
thus it should apply to ensembles. In fact, at that time it was impossible to carry out experiments
with single photons, atoms, or electrons, so that there was no need to ask what happens if one
has just one and wants to apply quantum physics to it. But some scientists (like Schrödinger
or Einstein) realized that applying quantum physics to few-particle systems would lead to a big
change in the way we view Nature. They came up with so-called Gedankenexperiments, which
could not be realized in practice, but showed some apparently paradoxical predictions of quantum
physics.
In particular, Einstein, Podolsky, and Rosen (EPR), in a seminal paper, argue that Quantum
Physics cannot be complete and needs to be complemented with some other variables, which they
called hidden variables. In that case, quantum predictions can be understood with a realistic
theory, without having to impose that in some cases properties are not defined. Later on, Bell
figured out that such a theory would have to be non-local if it has to reproduce all predictions of
quantum physics. This observation led to his famous theorem that states that quantum physics is
incompatible with all local hidden variable theories. From his ideas, one can devise an experiment
where local hidden variable theories and quantum physics make different predictions. Those exper-
iments have been carried out in the last 40 years, yielding results that are incompatible with local
hidden variable theories. This, apart from having some philosophical consequences, highlights the
power of quantum physics since it can give rise to results that cannot be obtained by more natural
theories. And, in particular, the role of entanglement in all that. In fact, all those arguments gave
rise to the field of Quantum Information and all its applications.
In this chapter, we will revise the arguments given by EPR, Bell, and others and show how one
can harness entanglement to carry out tasks that are impossible in classical setups.

2 Einstein–Podolsky–Rosen paradox
Entangled states raised controversy in the 30s. The maximum exponent was the article of Einstein,
Podolsky, and Rosen (EPR) “Can Quantum Mechanical description of physical reality be consid-
ered complete?”. In that article they argue, using an apparent paradox, that Quantum Mechanics
does not give a complete description of a system1 .

Elements of reality EPR define this concept as follows: If, without in any way disturbing a
system, we can predict with certainty the value of a physical quantity, then there exists an element
of physical reality corresponding to this physical quantity. This condition basically means that
this physical quantity has a value independently of whether we measure it or not.

Locality By locality we mean that instantaneous actions at a distance are not possible. More
precisely, that an action that takes some time t in a place cannot affect another place at a distance
L, if L > ct, where c is the speed of light.

1 Complete theory: deterministic measurement outcomes can be predicted exactly even before the measurement.

1
Chapter 2 QST Theory: Quantum Information (WiSe 22/23)

Singlet state Consider the following Bell state of two qubits, A and B,
1
|Ψ− i = √ (|0iA |1iB − |1iA |0iB ). (2.1)
2
In the context of spins, this is also called a singlet state. This has the property that, for any ~n
(σ~nA ⊗ σ~nB )|Ψ− i = −|Ψ− i. (2.2)
That is, if we measure and and
σ~nA and denote by
σ~nB and
m~na the outcomes, then
m~nb m~na = −m~nb .
One can say that the two measurement results are anti-correlated.

Scenario We consider two qubits in a singlet state and separated by a long distance such that
during the measurements no signal can go from A to B or vice versa. We consider measuring the
qubit A, and then making predictions about the outcomes if we would measure qubit B.

Measurements Suppose A and B measure the state of their respective particles.

• σz : First consider the case where they measure the observable σz . We denote by mA z and
mBz the results of the measurements. If mz = 1 then mz has to be mz = −1. On the
A B B

other hand, if mAz = −1 then mz = 1. In other words, if one measures σz in particle A, the
B A

result of σz for the second particle is completely determined without disturbing it (because
B

of locality). Thus, σzB is an element of reality.

• σx and σy : For the same arguments, σxB and σyB are elements of reality.

EPR Paradox σx,y.z B

are elements of reality, i.e. they are well defined before a measurement. How-
ever, according to quantum mechanics, one cannot assign definite values to these three observables
(σxB , σyB , σzB ) simultaneously, since their corresponding operators do not commute.

What is wrong with QM? According to EPR, “it becomes evident that the paradox forces us
to relinquish one of the following assertions: (1) the description by means of a wavefunction is
complete; (2) the real states of spatially separated systems are not independent”. According to
Einstein, the second assertion was indisputable since “the situation of system B is independent
of what is done with the system A, which is spatially separated from the former”. This physical
principle has received the name Einstein locality.

Hidden variables According to EPR, quantum mechanics is not complete and hidden variables
are needed to complete it. These are variables that cannot be measured (at least in principle),
and that determine the result of any particular experiment. If one averages statistically over the
possible values of the hidden variables, one would obtain the same results as the expectation values
of quantum physics.

Bohr response Bohr answered that the experiments used to determine mz , mx and my are
mutually incompatible. If we measure σxA , then we can determine what would be the result if
one measures σxB , but one cannot predict anything about the value of σyB (and analogously for
σzB ). This requires a different experimental setup, and these experiments have to be analyzed
independently. Therefore, this is a counter-factual experimental proposal, and no conclusions can
be made about elements of reality.

Realist theory is that which assumes that the observables (elements of reality) are defined
(whether we measure or not) and can be described by a set of hidden variables. Local hidden
variable theories (LHVT) or local realist theories are those in which, in addition, locality is ful-
filled.

3 Bell theorem
Bell proved a mathematical theorem that shows the incompatibility between LHVT and quantum
mechanics. He did that by proposing a Gedankenexperiment in which those theories would have
different predictions.

2
Chapter 2 QST Theory: Quantum Information (WiSe 22/23)

3.1 Gedankenexperiment
Setup To describe the setup, it is very important not to fall into the trap of defining an experiment
using the language of quantum physics, since we want to treat that theory and any LHVT on equal
footing. The experiment has Alice and Bob separated a long distance, so that we can assume
locality. They share many pairs of particles. Then they measure one out of two observables in each
pair, choosing randomly and independently the observables. The outcomes of each observable can
be ±1. One can imagine that they have three boxes, each of them corresponding to one observable,
and they put one of the particles of a pair in one of the boxes. Then a white or a red light shines,
indicating if the outcome of the experiment is ±1.
We call the possible observables a, b for Alice and b, c for Bob, and the outcomes mA a and mb
A

for Alice’s measurements and mb and mc for Bob’s.

B B

Assumptions We assume that the conditions of the experiment are such that:
1. The values of the outcomes m can take the values +1 or −1.
2. If for a pair of particles, Alice and Bob perform the same measurement, the result is opposite.
That is, mA b mb = −1.
B

Experiment After they have obtained all the values in the different runs of the experiments,
they take all the ones in which Alice has measured a and Bob b, multiply them, and compute the
average, which we call E(a, b). Then they do the same for the other combinations and compute
E(a, c) and E(b, c). We are interested in the relations between these quantities when the number
of experiments is very large. (Note that E(b, b) = −1 according to the second assumption).

3.2 Analysis: hidden variables theory and Bell inequality

Hidden variables Let us describe the above experiment assuming the existence of hidden vari-
ables. We, therefore, assume that the outcome of each single realization of any experiment depends
on a set of hidden variables λ, which change from experiment to experiment. These hidden variables
are distributed statistically according to an unknown distribution ρ(λ) ≥ 0, which is normalized
to one Z
dλρ(λ) = 1. (3.1)

Note that there may be many variables, λ1 , λ2 , . . ., but we collect all of them and call them λ to
simplify the notation.

Prediction An LHVT imposes that the results of the measurements in a given single realization of
the experiment depend on the observable they measure as well as on the hidden variables λ, which
may differ from experiment to experiment. Moreover, the following properties must be fulfilled:
1. Measurement in A: We denote by A(a, λ) = ±1 the outcome of the experiment if Alice
measures the observable a and the value of the hidden variables is λ. In the same way we
define A(b, λ). The fact that a and λ determine the value of the measurement is connected
to the assumption of realism. The fact that this value is independent of the choice of Bob’s
observable is due to locality. The fact that it is ±1 is due to the first assumption.
2. Measurement in B: Similarly, we denote by B(b, λ), B(c, λ) = ±1.
3. Correlations: Because of the second assumption
A(b, λ)B(b, λ) = −1. (3.2)

Expectation values If we perform several times the experiments for observables a and b for Alice
and Bob, respectively, multiply them and average over all the experiments, we will obtain the
result Z
E(a, b) = dλρ(λ)A(a, λ)B(b, λ), (3.3)

where ρ(λ) is the statistical distribution of the hidden variables, which is unknown. In the same
way, if we run another set of experiments choosing other directions, we will have similar expressions
for E(a, c) and E(b, c).

3
Chapter 2 QST Theory: Quantum Information (WiSe 22/23)

Derivations
Z
E(a, b) − E(a, c) = dλρ(λ)[A(a, λ)B(b, λ) − A(a, λ)B(c, λ)]
Z (3.4)
= − dλρ(λ)A(a, λ)A(b, λ)[1 + A(b, λ)B(c, λ)]

where we have used that A(b, λ)2 = 1, and that B(b, λ) = −A(b, λ). Calculating the modulus, and
taking into account that |A(a, λ)A(b, λ)| = 1, we find the Bell inequality.

Bell inequality
|E(a, b) − E(a, c)| ≤ 1 + E(b, c). (3.5)
This inequality is predicted by any local realist theory in any experiment as the one described
above.

3.3 Quantum prediction

σ~nAa/b
a/b
A mA

Singlet state |Ψ− i σ~nBb/c

b/c
B mB

Figure 1: Gedankenexperiment of Bell.

Now, we want to describe a specific experiment fulfilling all assumptions, and predict the results
with quantum theory. Thus, we are now allowed to use the language of such a theory. Thus,
we consider Alice and Bob have pairs of qubits, each of them in the single state (2.1) and the
observables a, b, and c are σ~na , σ~nb and σ~nc , respectively. We can easily compute

E(x, y) = hΨ|σ~nx ⊗ σ~ny |Ψi = −~nx · ~ny . (3.6)

Note that the first assumption is fulfilled, since the outcomes of the experiments will be ±1.
Furthermore, the second assumption is fulfilled as well, since if they measure the same observable
the results will be opposite. In particular E(b, b) = −1.

Violations of Bell inequality Simply select

~na = (1, 0, 0), ~nb = (cos(π/3), 0, sin(π/3)), ~nc = (cos(2π/3), 0, sin(2π/3)). (3.7)

We have E(a, b) = E(b, c) = −1/2, E(a, c) = 1/2. Bell inequality (3.5) would give 1 ≤ 1/2.
This proves Bell’s theorem, i.e., quantum physics is incompatible with local realist theories (or,
equivalently, local realist theories).

4 CHSH inequalities
Bell inequalities cannot be tested experimentally. The reason is that the second assumption (or,
equivalently, that E(b, b) = −1) is not robust against any experimental imperfection. The CHSH
(Clauser, Horne, Shimony, and Holt) inequalities circumvent this problem by waiving that assump-
tion. The setup of CHSH inequalities is similar, but now, Alice and Bob can choose two observables
each, a and d for Alice, and b and d for Bob. The only requirement is that the absolute values of
the observables obtained by A and B do not exceed one: |A(a, λ)|, |A(d, λ)|, |B(b, λ)|, |B(c, λ)| ≤ 1.

4
Chapter 2 QST Theory: Quantum Information (WiSe 22/23)

Derivation
Z
E(a, b) − E(a, c) = dλρ(λ)[A(a, λ)B(b, λ) − A(a, λ)B(c, λ)]
Z
= dλρ(λ)A(a, λ)B(b, λ)[1 ± A(d, λ)B(c, λ)] (4.1)
Z
− dλρ(λ)A(a, λ)B(c, λ)[1 ± A(d, λ)B(b, λ)],

where we have added and subtracted the same quantity in the integral. The sign ± is a shorthand
notation. One can read this equation as two different equations, one with the plus sign + and
another one with the minus sign −. Using |A|, |B| ≤ 1, we have

|E(a, b) − E(a, c)| ≤ 2 ± [E(d, c) + E(d, b)] (4.2)

from which the CHSH inequality directly follows.

CHSH Inequality
S ≡ |E(a, b) − E(a, c)| + |E(d, c) + E(d, b)| ≤ 2. (4.3)
This inequality has to be fulfilled by any local realist theory. For the particular case d = c and
E(c, c) = −1, it reduces to Bell inequality (3.5).

Violations of the CHSH inequality In quantum physics, the CHSH inequality can be violated
by selecting

1 1 1 1
   
~na = (1, 0, 0), ~nb = √ , 0, √ , ~nc = − √ , 0, √ , ~nd = (0, 0, 1), (4.4)
2 2 2 2
which gives √ √
E(a, b) = E(d, c) = E(b, d) = −1/ 2, E(a, c) = 1/ 2. (4.5)
√
We find a violation of the CHSH inequality, 2 2 > 2.

~nd

~nc ~nb

~na

Figure 2: A set of measurement angles (in the x − z plane) that gives maximal violation of the
CHSH inequality.

violation of Bell’s inequality requires entanglement. Imagine we have a separable state

The P
ρ = k p k ρA
k ⊗ ρk and measure σ~
B
a , σ~b , the expectation value of the measurement would be
A B

E(~a, ~b) = tr(ρσ~aA ⊗ σ~bB )

X
= pk tr(ρA
k σ~ a )tr(ρk σ~b )
A B B
(4.6)
X
= pk A(~a, k)B(~b, k).
k

Here we can think of k as a hidden variable λ in the LHV model. Therefore all separable states
will fulfill Bell’s inequality.

5
Chapter 2 QST Theory: Quantum Information (WiSe 22/23)

5 Bell theorem without inequalities

So far, we have seen some proofs of the Bell theorem based on statistical inequalities. That is,
if one performs many experiments, the correlations predicted by local realist theories must fulfill
some inequalities. Bell theorem can also be proved without the use of inequalities: that is, showing
that a given outcome of a given experiment is incompatible with local realist theories.

5.1 GHZ states

Consider 3 particles A, B, and C, separated a long distance from each other, in a three–particle
entangled GHZ state
1
|Ψi = √ (|0i|0i|0i − |1i|1i|1i). (5.1)
2
Using the following properties
σx |0i = |1i, σy |0i = −i|1i, (5.2)
σx |1i = |0i, σy |1i = i|0i, (5.3)
one can readily check that the GHZ state (5.1) is an eigenstate of several operators. In particular,
σxA σyB σyC |Ψi = |Ψi, (5.4)
σyA σxB σyC |Ψi = |Ψi, (5.5)
σyA σyB σxC |Ψi = |Ψi, (5.6)
σxA σxB σxC |Ψi = −|Ψi. (5.7)

Measurements If one observer measures in A σxA or σyA , he can obtain two results for each of
them, namely mAx = ±1 and my = ±1. The same applies to B and C, i.e. mx , my , mx , my = ±1.
A B B C C

On the other hand,

• If in an experiment one measures σxA , σyB , and σyC , then, according to (5.4), the results will
fulfil

x my my = 1.
mA (5.8)
B C

• If in an experiment one measures σyA , σxB , and σyC , then, according to (5.5), the results will
fulfil
mAy mx my = 1. (5.9)
B C

• If in an experiment one measures σyA , σyB , and σxC , then, according to (5.6), the results will
fulfil
mAy my mx = 1. (5.10)
B C

• If in an experiment one measures σxA , σxB , and σxC , then, according to (5.7), the results will
fulfil
mA x mx mx = −1. (5.11)
B C

Elements of reality In view of (5.8, 5.9, 5.10), σx,y

A,B,C
are elements of reality, and therefore,
according to a local realist theory, they would have well defined values independently of if one
measures them. For example, without disturbing A one can guess the value of σxA by measuring
σyB and σyC and using (5.8). Similarly, one can determine the value of σxB without disturbing the
second system by measuring σyA and σyC and using (5.9). Analogously, one can determine the value
of σxC without disturbing C.

5.2 Bell Theorem

Gedankenexperiment Consider three parties, Alice, Bob, and Charlie, separated by long dis-
tances so that locality applies. They have many sets, and in each set, there are three particles,
one for each of them. Each of them can measure two observables, x and y, and obtain an outcome
mAx , . . . , my .
C

6
Chapter 2 QST Theory: Quantum Information (WiSe 22/23)

Assumptions We assume that the conditions of the experiment are such that:

1. The values of the outcomes m can take the values +1 or −1.

2. For any set, if they multiply the outcomes mA
x my my , my mx my or my my mx , they always
B C A B C A B C

obtain +1.

Local Hidden variables theories For any value of the hidden variables, λ, if we multiply r =
mA x my my my mx my my my mx they must obtain 1, according to the second assumptions. The
B C A B C A B C

first assumption implies that (mA

y ) = (my ) = (my ) = 1, thus we obtain
2 B 2 C 2

x mx mx = 1.
mA (5.12)
B C

Thus, any LHVT fulfilling the assumptions must give that value.

Quantum prediction If they choose the GHZ state and the measurements x, y correspond to σx
and σy , then the assumptions are fulfilled. However, the prediction (5.11) contradicts the LHVT
prediction (5.12).

6 XOR Game
The XOR game uses CHSH inequalities to show that it is possible to get an advantage by using
quantum physics (as opposed to classical) in some particular scenarios.

x = 0, 1 y = 0, 1

Alice Bob

a = 0, 1 b = 0, 1

Figure 3: The setup of XOR game.

Setup XOR game is a game that involves two players, Alice and Bob. They meet to plan a
strategy, and then they are isolated in two rooms, where they cannot communicate with each
other. Alice is given a random bit x ∈ {0, 1} and Bob is given a random bit y ∈ {0, 1}. Then Alice
produces a bit a ∈ {0, 1} and Bob produces a bit b ∈ {0, 1} without communicating with each
other. Based on the bits they choose, they are assigned a score s = +1 if a ⊕ b = xy and a s = 0
score if a ⊕ b 6= xy. The game is repeated for many rounds and the average score s̄ determines
their final score.

x y x·y a b a⊕b
0 0 0 0 0 0
0 1 0 0 1 1
1 0 0 1 0 1
1 1 1 1 1 0

Table 1: The truth table of x · y. Table 2: The truth table of a ⊕ b.

Classical strategy Let us denote by ax and by the bits produced by Alice and Bob given the
input bits x and y. Of course, Alice and Bob can use different strategies in different rounds. They
can agree on different strategies, denoted by λ, so that ax (λ) and by (λ). They can choose to use a
given strategy in each round. Or we can consider a more general scenario, where the strategy λ is

7
Chapter 2 QST Theory: Quantum Information (WiSe 22/23)

chosen randomly (i.e., assuming that Alice and Bob can draw the same set of random numbers).
This means that we can just assign a probability distribution ρ(λ).
For a given strategy, the score if they are given x, y will be
1 + (−1)ax (λ)+by (λ)−x·y 1 1
 
S̄x,y (λ) = = 1 + (−1) ax (λ)
(−1) by (λ)
(−1) xy
. (6.1)
2 2 4

Let us introduce A(x, λ) = (−1)ax (λ) and B(y, λ) = (−1)ay (λ) and average with respect to the
values of x, y = 0, 1 (each of them with probability px,y = 1/4) and the distribution ρ(λ)
1 1
 Z 
S̄ = 1+ dλρ(λ) [A(0, λ)B(0, λ) + A(0, λ)B(1, λ) + A(1, λ)B(0, λ) − A(1, λ)B(1, λ)]
2 4
1 1
= [1 + (E(0, 0) + E(0, 1) + E(1, 0) − E(1, 1))]
2 4
1 1
≤ [1 + (|E(0, 0) + E(0, 1)| + |E(1, 0) − E(1, 1)|)]
2 4
1 1 3
≤ (1 + ) = = 0.75.
2 2 4
(6.2)
In the last step we have used the CHSH inequality (4.3), since the derivation is the same. Therefore,
in the classical setting the winning probability of Alice and Bob is upper bounded by S̄ = 0.75.
This bound can be saturated by a simple strategy - Alice and Bob always produces 0, ax = by = 0
for all inputs x and y. In this way A(0, λ) = A(1, λ) = B(0, λ) = B(1, λ) = 1, and indeed,
!
1 1X 1 1 3
S̄ = 1+ (−1)xy = (1 + ) = . (6.3)
2 4 x,y 2 2 4

Quantum strategy If Alice and Bob had known quantum physics and CHSH inequalities, they
could use qubits to get a better score. They just have to share entangled pairs and make measure-
ments depending on whether they are given x, y = 0, 1. In that case,
√
1 2
S̄ = [1 + ] ≈ 0.85 > 0.75. (6.4)
2 2
XOR game is an example in which quantum physics unlock things that are not possible classically.

7 *Non-contextual hidden variable theory

In principle, any system can be described by a set of hidden variables. What Bell showed is that,
if additionally, one requires locality, then there are quantum mechanics that cannot be described
by hidden variables. Other scientists showed that if one does not assume locality but makes other
assumptions, there are also contradictions between a (realist) theory based on hidden variables
and quantum mechanics. Or, simply, those theories cannot exist. Here we will see those examples
of these contradictions. In contrast to Bell inequalities, they do not depend on a specific state of
a system. The situation can be summarized as
1. Local hidden variable theory: contradicts quantum physics (Bell theorem).
2. Hidden variable theory: no contradiction with quantum physics.
3. Noncontextual hidden variable theory: contradicts quantum physics.

Compatible observables Formally, HVT is described by a set of maps ωλ : L(H) → R that map
observables to real numbers, together with a normalized probability distribution p(λ). One of the
assumptions that may seem “logical” is the following. Imagine we have a set of observables A1 , A2 ,
· · · , AN , which quantum mechanically are represented by operators. For a given system, a hidden
variable theory assigns certain values
ωλ (A1 ), ωλ (A2 ), . . . , ωλ (AN ) (7.1)
to these observables. If the quantum mechanical operators fulfill a certain algebraic relation, like
f (A1 , A2 , . . . , AN ) = 0, (7.2)

8
Chapter 2 QST Theory: Quantum Information (WiSe 22/23)

one could assume that the same relation should be fulfilled by the values assigned by the hidden
variable theory, that is,
f (ωλ (A1 ), ωλ (A2 ), . . . , ωλ (AN )) = 0. (7.3)
This assumption, although it may seem logical, leads to a contradiction.

Von √Neumann example For example, for a two–level system we define an operator σr = (σx +
σy )/ 2. Since values taken by σx,y are ωλ (σx ) = ±1, then the mentioned assumption would imply
that σr can take values

ωλ (σx ) + ωλ (σy ) √ √
ωλ (σr ) = √ = {+ 2, 0, − 2}. (7.4)
2
However, σr only has eigenvalues ±1 so we should assign ωλ (σr ) = ±1. Thus a LHV theory in
which (7.3) is true for all functions (7.2) is not consistent.
Von Neumann concludes that this implies that quantum mechanics is, in general, not compatible
with hidden variable theories (not only LHVT). However, later people realized that Von Neumann’s
argument could be refuted in a similar way as to how Bohr responded to the EPR paradox. The key
point is that if the operators do not commute, then (according to quantum mechanics) we cannot
assign specific values to the results of measurement for any state (there is no way of measuring
non-compatible observables on a single system). Thus, there is no reason in principle to impose
the same condition on the operators and the values assigned by a realist theory. However, we may
impose this condition if the observables are compatible (i.e., they commute with each other). In
that case, the observables can be measured simultaneously, and indeed they (will) fulfill the same
algebraic equations as the operators.

Non–contextual theories The contradictions that we will analyze in this section have, apart from
the assumption explained above, another one that is related and that is called non–contextuality.
It basically means that the result of a measurement (if we would perform it) would not depend
on whether we measure some other variables or not, as long as those correspond to operators that
commute; that is, a measurement does not depend on its context. To be more specific, let us
consider three observables A, B and C, such that A and B can be measured simultaneously, and
also A and C can be measured simultaneously (using the language of quantum mechanics, the
operators associated to these observables A, B, and C fulfill [A, B] = [A, C] = 0, but nothing is
assumed regarding the commutator [B, C]). A non–contextual theory is the one for which the result
of the measurement of A is independent of whether we measure it together with B or together
with C. Note that these observables can refer to the same system (they do not need to belong to
two distant systems, as in the case of the Bell theorem).
Example 7.1 (A qutrit example). Let us consider a three level (qutrit) system {|1i , |2i , |3i}. One
can measure the projectors on |1i, |2i, |3i or the projectors on |1i, |ai, |bi, where |ai = √12 (|2i+|3i)
and |bi = √12 (|2i − |3i),

{P1 = |1i h1| , P2 = |2i h2| , P3 = |3i h3|} , (7.5)

{P1 = |1i h1| , Pa = |ai ha| , Pb = |bi hb|} . (7.6)

These projectors commute with each other, [Pi , Pj ] = 0, i 6= j and sum up to the identity

P1 + P2 + P3 − 1 = 0, (7.7)
P1 + Pa + Pb − 1 = 0. (7.8)

From this identity (7.7), hidden variable theory will assign the values

ωλ (P1 ) + ωλ (P2 ) + ωλ (P3 ) = 1, (7.9)

ωλ (P1 ) + ωλ (Pa ) + ωλ (Pb ) = 1, (7.10)

in term of the noncontextual HVT. The HVT is called ”noncontextual” because we are demanding
that ωλ (P1 ) is independent of the measurement outcome of other observables P2 , P3 or Pa , Pb (the
context). The value is assigned independent of the context.

9
Chapter 2 QST Theory: Quantum Information (WiSe 22/23)

7.1 The Kochen–Specker theorem

The Kochen–Specker theorem asserts that there is no consistent way of ascribing “noncontextual”
definite answers (or results) to sets of yes–no questions (measurements) regarding an individual
physical system. As explained above, non–contextual means that the results of the measurements
are independent on which other observables are being measured jointly.

Premises Given an individual system, let v(u) denote the answer (1 =yes, 0 =not) to the proposi-
tion Pu (mathematically represented by the projector |uihu|). The premises of the Kochen–Specker
theorem (also known as Bell–Kochen–Specker theorem) can be formulated as follows:
• In an individual system, each proposition Pu has a unique answer 0 or 1, which is independent
of which other observables are being considered jointly (non–contextuality).
• For each set ofP
one–dimensional projectors whose sum is the unit matrix in the n–dimensional
Hilbert space u Pu = 1 with n ≥ 3, the answer to one and only one of the projectors is 1,
and the answer to the other n − 1 projectors is 0.

The theorem can be enunciated as follows: there is no set of answers satisfying these premises for
all sets of propositions. To prove the theorem, one has to find a set of propositions that does not
fulfill the premises. We will illustrate the gist of Kochen–Specker theorem in a simplified setting -
Mermin’s example.

7.2 Mermin’s example

Apart from non–contextuality, let us assume that the result of any set of measurements fulfills the
same multiplication rule as the corresponding quantum mechanical operators (that is, if XY = C,
then the values assigned to the observables also fulfill these rules).
With an example, Mermin showed that this conditions are inconsistent. Mermin considered a
two-qubit system H = C2 ⊗ C2 and two operators x and y, which are non-contextual, i.e, they
commute [x, y] = 0. As a consequence, the hidden variable contextual theory, described by the
assigned values wλ , we impose
wλ (x · y) = wλ (x)wλ (y) (7.11)
and
wλ (−x) = −wλ (x). (7.12)

Operators Consider the following 9 operators and make a Mermin table

1A ⊗ σzB σzA ⊗ 1B σzA ⊗ σzB

σxA ⊗ 1B 1A ⊗ σxB σxA ⊗ σxB (7.13)
σxA ⊗ σzB σzA ⊗ σxB ???

Hidden variables In each row and each column, the three operators commute (i.e. according
to non–contextuality, we can assign a single value to these operators, regardless of whether we
measure the observables of the row or of the column). Then, the value assigned to the first two
columns of the last row has to be the product of the values associated to the first two rows on that
columns. Analogously, the values assigned to the fist two rows of the last column are the products
of those assigned to the fist two columns. But, what should we assign to the entry of the last row
and column? On the one hand, it should be the product of the first two rows of the last columns,
i.e. ω = ω(σzA ⊗ σzB )ω(σxA ⊗ σxB ), which, according to the assumption (7.11), should be

ω = ω(σzA ⊗ σzB )(σxA ⊗ σxB ) = ω(−σyA ⊗ σyB ) = −ω(σyA ⊗ σyB ) (7.14)

where in the last step we have used (7.12). Analogously, it should be obtained by multiplying the
first two columns of the last row, i.e. ω = ω(σxA ⊗ σzB )ω(σzA ⊗ σxB ). Using again (7.11) we have

ω = ω(σxA ⊗ σzB )(σxA ⊗ σzB ) = ω(σyA ⊗ σyB ). (7.15)

Thus, we get a contradiction so that such a theory cannot exist.

10
Chapter 2 QST Theory: Quantum Information (WiSe 22/23)

On the other hand, each operator is the product of the other two, except for the third column,
where a minus sign would be needed, i.e.

σyA ⊗ σyB = (σxA ⊗ σzB )(σzA ⊗ σxB )

= −(σzA ⊗ σzB )(σxA ⊗ σxB ).

Hidden variables Each of these operators has ±1 eigenvalues. Therefore, if we would measure
one of them, we would get ±1. A hidden variable theory would tell us that we can assign values ±1
to all these variables, independently of whether they are measured or not, or what it is measured

ωzB ωzA ωzA ωzB

ωxA ωxB ωxA ωxB (7.16)
A B
ωx ωz ωzA ωxB ωyA ωyB

These variables have to be ±1, and have to fulfill the same rules as the products of the corresponding
operators. Thus, we give values to ωx,z
A,B
and everything else should be determined. However, this
is impossible since it would be required that (ωzA ωzB )(ωxA ωxB ) = −(ωxA ωzB )(ωzA ωxB ).

8 Remarks on causality
Instantaneous action at distances Performing a measurement in one of the EPR pair (2.1)
apparently changes the state of the other particle. For example, if in the system A one obtains
the result +1, the state of the other particle is projected onto |0i. One might think that this may
lead to supraluminal communication, but this is not so. Some observer that possesses particle B
cannot find out an observer with particle A has performed a measurement or not. The reason is
that with a single particle, one cannot measure the state of its particle [whether it is a pure state
0i or a mixture ρB = 21 (|0ih0| + |1ih1|)]. More precisely, since he does not know the result of the
measurement in A all, the probabilities for his measurements are given by traces of observables
with the operator
1
ρB = (|0ih0| + |1ih1|), (8.1)
2
which does not depend on whether A measured an observable or not.

Quantum cloning is not possible If there exists a unitary cloning operation, U , that can copy
two states, |ψ1 i and |ψ2 i, in the following way,
U
|ψ1 i |0i |Ei −
→ |ψ1 i |ψ1 i |Eψ1 i ,
(8.2)
U
|ψ2 i |0i |Ei −
→ |ψ2 i |ψ2 i |Eψ2 i .

Then the inner product between |ψ1 i and |ψ2 i satisfy

hψ1 | ψ2 i = | hψ1 | ψ2 i|2 hEψ1 | Eψ2 i ≤ | hψ1 | ψ2 i|2 , (8.3)

because hEψ1 | Eψ2 i ≤ 1. This inequality can only be true if | hψ1 | ψ2 i| = 0 or 1. This means
that quantum cloning is only possible when the two states are orthogonal (eg, the cloning of two
classical bits |0i and |1i is possible) or parallel.
From the causality perspective, if quantum cloning were possible (which is not the case), one
could make many copies of the particle in B and then measure whether it is in the state |0i or |1i
or in a mixture. Thus, if one would be able to copy unknown quantum one could violate causality.

Nonlinear quantum mechanics There have been some attempts to generalize QM and construct
a nonlinear theory. If Quantum Mechanics were nonlinear, causality would be violated. The
justification of this statement can be summarized as follows: consider two observers A and B,
separated by a long distance, who share several EPR pairs in a singlet state. The observer in A
wants to send one of two possible messages to the observer in B. The message X one consists of
measuring σzA on all his particles, and the message number Y σxA . In the first case, each particle
of the second observer will be either in |0i or on |1i, whereas in the second one, they will be in

11
Chapter 2 QST Theory: Quantum Information (WiSe 22/23)

√
(|0i ± |1i)/ 2. Thus, if observer B can distinguish among these two sets, he will be able to know
the message sent A instantaneously. The idea is that if Quantum Mechanics were not linear, one
could find an evolution that would give different density operators for these two messages.
If we consider a non-unitary quantum mechanics

U (a |0i + b |1i) = aU |0i + bU |1i . (8.4)

This would imply that when we apply the unitary operation to ρ1 and ρ2 , we will obtain different
density matrices.
1
ρ1 = (|0i h0| + |1i h1|),
2 (8.5)
1
ρ2 = (|+i h+| + |−i h−|).
2
♣♣♣

Acknowledgement.— We thank Bennet Windt, Yilun Yang and Adrian O. Paulus for providing
their handwritten notes. We thank Franz Silva-Tarouca, Adrian O. Paulus, and Jan Geiger for
their proofreading.

These lecture notes were kindly provided by Ignacio Cirac and Sirui Lu.

12