You are on page 1of 34

What could not be in Feynman lectures:

Bell inequalities

V. Černý
Abstract

Lecture notes written in the spirit of Feynman lectures on quantum mechanics.


Feynman extensively uses Stern-Gerlach filters to describe basic principles of
QM using thought experiments with polarized electron beams. We add the
discussion on the EPR paradox, hidden parameters and Bell inequalities. This
is not meant as an introductory text from quantum mechanics. It is rather a
supplement to be read by students after they learned the basic course of quantum
mechanics and in particular the formalism of spin-1/2 particles. However it may
be useful also for students who are interested in, say, quantum computers and
the logical aspects of the problems of measurements in quantum mechanics.
They should then rely on the mathematical formalism and accept the fact that
some pieces of ”physical argumentation” will not be completely clear to them.
1 Stern Gerlach filters

In this section we briefly repeat what was (at least ”in spirit”) written in the
Feynman lectures. We shall speak on beams of electrons passing through various
Stern Gerlach filters. This kind of language has only symbolical meaning. We
do not care whether the described experiments are really technically feasible.
We use the philosophy of the thought experiments heavily exploited by Einstein,
Bohr and Feynman. Thought experiments are performed virtually, just in our
imagination. There might be technical problems if we tried to perform them in
reality. However we carefully check that on a principal level there is nothing to
prevent them working. The principal components used in our experiments will
be an electron gun producing a beam of unpolarized electrons, Stern Gerlach
filters and electron detectors which detect electrons with ideal efficiency irre-
spective of their polarization. Electrons are fermions with spin 1/2. Their spin
states are described by vectors in two-dimensional Hilbert space. A base in the
Hilbert space can be chosen arbitrarily: most often one chooses the base formed
by the eigenstates of the operator of the projection of the spin to the z-axis. W
shall denote these base vectors as

|↑i and |↓i

In the state |↑i the projection of the electron’s angular momentum on the z-axis
is h̄/2 while in the state |↓i it is −h̄/2.

To simplify the discussion we shall not consider real electrons but certain imag-
inary particles. What concerns their spatial motion we shall consider just a
(quasi-classical) uniform motion in the direction of the y-axis. What concerns
polarization, we shall consider only polarizations perpendicular to the beam
axis. Our ”electrons” will be polarized in the xz-plane and we shall completely
neglect the polarization in the direction of the y-axis (which is possible in real
world). So we shall consider particles living in a one dimensional space along
the y-axis with a two-dimensional space of ”internal degrees of freedom” given
by the plane xz. Mathematically it will mean that we shall take into account
not all the vectors from the Hilbert space spanned on the two basis vectors |↑i
and |↓i but only those superpositions

c↑ |↑i + c↓ |↓i (1)

where the coefficients c↑ and c↓ are real and satisfying the normalization condi-
tion
|c↑ |2 + |c↓ |2 = 1

The Stern Gerlach Filter (SGF) is extensively discussed in the ”Feynman lec-
tures on physics” so we skip even a schematic description of its internal con-
struction here and we just describe its function as a black box.

SGF is device used to filter electrons in the electron beam. Each SGF is char-
acterized by the direction and orientation of its ”principal axis” (denoted by
an arrow on our figures) which is perpendicular to the direction of the electron

1
z

x
Figure 1: Stern Gerlach filter

beam going through. The default axial orientation of the SGF is with its arrow
pointing vertically upwards.

The function of the SGF in its default position is simple. The electron in the
state |↑i passes through the filter without any change of state, electron in the
state |↓i is absorbed in the filter and so does not appear on the output side od
the SGF.

Electron in a pure quantum state is always polarized, however, not necessarily


in the direction of the z-axis. In general its spin state can be expressed as a
superposition 1. Suppose we hav a gun which produces a beam of electrons
each of them being in the state (1). Then according to the basic teaching
of quantum mechanics these electron pass through the SGF (in the default
orientation state) with the probability |c↑ |2 . So in the experimental setup given
in Fig.2 the detector fires with the rate |c↑ |2 relatively to the gun fire rate.

G
Figure 2: SGF in a beam produced by the electron gun G. The electrons passing
through the SGF are registered by the detector D

If we add an additional SGF oriented in the same way as the first one (Fig.3),
we find that the detector rate remains the same as it was with only one filter. It
means that all the electrons which pass through the first filter pass also through

2
the second filter. We conclude that after passing through the first filter the
electrons are in the polarization state which guarantees that they pass through
the second filter: it is the state |↑i. So whatever is the electron polarization
before entering the (default-oriented) SGF, it becomes |↑i after passing through
it.

G
Figure 3: Two identically oriented SGF’s in the beam

Now what happens if we turn the second SGF around the beam axis by an angle
ϑ, so that we get the experimental setup presented in Fig.4. Experimentally we
find that the firing rate of the detector decreases with respect to the situation
with just one SGF by a factor
cos2 (ϑ/2)

Figure 4: SGF in a default orientation followed by a SGF rotated by angle ϑ

We now rewrite what we just said in the language of mathematics. We described


the SGF devices as filters, but we can consider them to be measuring devices
as well. They measure a yes/no type of a physical quantity: the outcome of

3
the measurement by the SGF is either 1 or 0 according to whether the electron
passes through the filter or not. It is easy to see that the hermitean operator
corresponding to the SGF in its default orientation is

P↑ = |↑ih↑|

Really, this operator has two eigenstates |↑i and |↓i corresponding to the eigen-
values 1 and 0.
P↑ |↑i = |↑ih↑|↑i
P↑ |↓i = |↑ih↑|↓i = 0
Now let us discuss in the same way the SGF rotated axially by the angle ϑ.
We shall denote this SGF as SGF(ϑ). Obviously if we take the state |↑i and
rotate it by angle ϑ in the zx-plane we get the state which is guaranteed to pass
through SGF(ϑ). Formally it is the rotation with respect to the beam (y-axis).
The corresponding transformation is performed by the unitary operator

exp(−ih̄ϑσy /2)

where σy is the Pauli matrix


 
0 −i
σy =
i 0

Using this rotation operator we get for the rotated state

|ϑi = |↑i cos(ϑ/2) + |↓i sin(ϑ/2) (2)

So the hermitean operator corresponding to SGF(ϑ) is

|ϑi hϑ|

On the other hand the state which is guaranteed to be absorbed by the SGF(ϑ)
is the state obtained from |↓i by rotation by angle ϑ) It is the state

|ϑ⊥ i = − |↑i sin(ϑ/2) + |↓i cos(ϑ/2)

This state is, of course, orthogonal to the state |ϑi

hϑ⊥ | ϑi = 0

Now we can express the state vectors of our original basis in terms of the rotated
vectors. We get
|↑i = |ϑi cos(ϑ/2) − |ϑ⊥ i sin(ϑ/2) (3)
|↓i = |ϑi sin(ϑ/2) + |ϑ⊥ i cos(ϑ/2)

The interpretation of the Eq.3 is the following

If the system is in the state |↑i then it is simultaneously with the probability
cos2 (ϑ/2) in the state |ϑi and with the probability sin2 (ϑ/2) in the state
|ϑ⊥ i.

4
Now we see the mathematical background behind the experimental setup pre-
sented in Fig.4. After the SGF(↑) the system is in the state |↑i so with the
probability cos2 (ϑ/2) it pases through the SGF(ϑ) and with the probability
sin2 (ϑ/2) it is absorbed by it.

Now what happens if we add a third SGF oriented in the default direction
(SGF(↑)), see Fig.5

Figure 5: The sequence SGF(↑), SGF(ϑ) and SGF(↑) in the beam

Electrons after passing through the SGF(ϑ) forget about the state they were
occupying just before entering it. They are set to state |ϑi. What happens with
them in the third filter SGF(↑)? This we recognize if we express the state |ϑi
in the basis ”natural” to the filter SGF(↑). We have this expression in Eq.2

|ϑi = |↑i cos(ϑ/2) + |↓i sin(ϑ/2)

So the rate after the third filter with respect to that after the second filter is
reduced by the factor
cos2 (ϑ/2)
So irrespective of the fact that the electron already passed through the first filter
SGF(↑), adding another SGF(↑) after the SGF(ϑ) further reduces the detector
firing rate. Had we instead inserted the SGF(↑) in between the first SGF(↑)
and the SGF(ϑ) the detector rate would be the same as without the insertion,
because everything what passes through the first SGF(↑) passes through the
identical SGF(↑) immediately following.

5
2 Exercises

In this section we present simple exercises presented in the following figures as


different experimental setups. The problem is always the same. The electron
gun produces a the beam of electrons in the state

c↑ |↑i + c↓ |↓i

The firing rate of the gun is r (firing rate is expressed in units electron per
second). The task is to evaluate the detector firing rate.

Figure 6: Exercise 1

Figure 7: Exercise 2

Figure 8: Exercise 3

6
D

Figure 9: Exercise 4

Figure 10: Exercise 5

Figure 11: Exercise 6

Figure 12: Exercise 7

7
D

Figure 13: Exercise 8

Figure 14: Exercise 9

Figure 15: Exercise 10

Figure 16: Exercise 11

Figure 17: Exercise 12

8
D

Figure 18: Exercise 13

Figure 19: Exercise 14

Figure 20: Exercise15

Figure 21: Exercise 16

Figure 22: Exercise 17

9
D

Figure 23: Exercise 18

Figure 24: Exercise 19

10
3 Logic of quantum states

The discussion of the first section and the following exercises have provided us
with the ”experimental evidence” for the following statement.

If an electron passes through two SGF devices which are axially rotated to
each other the electron forgets the state it had coming out from the first
SGF and its state after the second SGF is completely determined by that
second SGF. Each SGF imprints on the electron a state given by that SGF
only, not depending on the previous history of the electron states.

The boxed statement is really strange, if we reexpress it in the language of SGF’s


as measuring devices.

Let us consider the SGF(↑). We can imagine a small display on the SGF which
shows ”1” when an electron entered the device and went through it and it
shows ”0” when the electron entered but did not go through (it was absorbed,
diverted to some other direction, whatever). The electron entered the SGF in
some (perhaps even unknown) in-state and after the interaction with the SGF
an out-state was imprinted on it. If the reading of the display was ”1” the
imprinted out-state is ket ↑, if the reading was ”0” did not get out of the SGF.

So the reading of the measuring apparatus is directly related to the out-state of


the electron, not that much to the in-state. In fact the only thing we can infer
about the in-state using the information from the display is that the in-state
was not orthogonal to the imprinted out-state. Saying it differently: the in
state was such that there was a non-zero probability that the outcome of the
measurement would be that one which was really observed. Nothing more.

This is completely contradictory to the creed of the experimental gurus from


the pre-quantum era. They believed that the purpose of a measurement is to
determine the in-state and that the ideal apparatus shows the relevant informa-
tion from which the in-state can be inferred without substantially influencing
that in-state. In other words they believed that one can determine the in-state
by the measurement and the out-state after the measurement will be identical
to the in-state.

All the experimental evidence of the quantum mechanics shows that kind of
belief was wrong. The only exception is the situation when we use the two
identical measuring devices one after the other within an infinitesimal time
interval. The the reading of the second apparatus would be the same as that of
the first one and the out-state after the second device would be the same as the
in-state, because that in-state relevant for the second apparatus is the out-state
coming from the first measuring device.

It is difficult to digest this fact. We know that Einstein felt very uneasy about
it and continued inventing thought experiments trying to beat the teaching
of Bohr about mutual incompatibility of measuring both the position and the

11
momentum of a particle. (The correspondence between Einstein and Bohr on
that matter is well known and reading it is a quite fascinating experience.) It
seems that Einstein finally accepted the possibility of the mutually incompatible
measurements.

Mathematically the incompatibility of two measurements is expressed via non-


commutability of the operators corresponding to the measuring devices.

In the Hilbert space of the electron polarizations any two different non-trivial
hermitean operators do not commute. In the language of SGF’s it means that
any two filters axially rotated with respect to each other are mutually incom-
patible. The corresponding hermitean operators do not commute. For example

|↑i h↑| does not commute with |ϑi hϑ| for ϑ 6= 0

In more complicated systems there may exist mutually commuting systems of


independent hermitean operators. The corresponding measurements are then
compatible what means that in a sequence of measurements like

A followed by B followed again by A

the system ”remembers” the value of the physical quantity A found in the first
measurement and the renewed measurement of A following the intermediate
measurement of B reveals exactly the same value of A as found by the first
measurement if the operators corresponding to A and B commute.

So let us return to our simple space of the electron polarizations and discuss the
consequences of the mutual incompatibility of measurements.

Suppose we have an electron coming out of the electron gun in an unknown


state. We have several SGF’s on stock with different axial orientation and we
want to perform some measurement on the incoming state. The main point is
that we have just one trial. Whatever measuring device we use, the incoming
state is lost forever. If we subsequently take another apparatus we perform the
measurement on a different state, not on the original one.

So before making the measurement we have to make a decision which apparatus


to take (which orientation of the SGF). The decision is final: once the mea-
surement is performed it cannot be undone to reconsider the situation or make
another choice. We make the decision, make the measurement, read the result
and from then on we know the state of the system. We might be curious to know
what would have been the result if we had chosen some other apparatus but
we cannot resolve that question. At least this is the teaching of the orthodox
quantum mechanics.

So given the incoming particle we can get the result of the measurement by any
single apparatus available but not by two (or more) of them: simultaneous
values of two (non-commuting)physical quantities are inaccessible.

12
This opens the way for situations quite contradictory to ”standard logical think-
ing”. Suppose we have an electron in an unknown state we choose to use the
SGF(↑) and we find that this particular electron did go through. And after
that we ask the question: If we had decided to use the SGF(π) apparatus in-
stead, would it go through or not? A very strange question indeed, because the
two devices in question, the SGF(↑) and the SGF(π) are not only incompati-
ble (non commuting) they are ”mutually exclusive”: any electron which went
through SGF(↑) will not go through the subsequent SGF(π) for sure. And vice
versa.

Our intuition then forces us to the conclusion that an electron which can go
through the SGF(↑) cannot go through the SGF(π). This conclusion is not
logically justified: mutual exclusivity of subsequent events does not inhibit their
concurrent possibility.

The concurrent possibility of mutually exclusive states in quantum mechanics is


mathematically expressed by the superposition principle: the existence of two
arbitrary states implies the existence of any state which can be expressed as a
superposition of the corresponding vectors in the Hilbert space. The fact that
the two considered states are mutually exclusive means that their corresponding
vectors are orthogonal. There is nothing strange in making superpositions of
mutually orthogonal vectors. Just the opposite is true, we prefer to use orthog-
onal bases for making superpositions.

In this way an arbitrary polarization state |ϑi can be expressed as a superposi-


tion of the mutually exclusive states |↑i and |πi.(See Eq.2)

|ϑi = |↑i cos(ϑ/2) + |πi sin(ϑ/2)

What we see here is in fact typical for quantum state: two mutually exclusive
states |↑i and |πi as ”concurrent possibilities” within a single quantum state
|ϑi.

To illustrate more clearly what is strange on quantum states let us discuss a


simple classical system a single die. The die has six faces labelled as 1, 2, 3,
4, 5, 6. We shall consider some measuring devices reading the state of the die.
To have some resemblance with our quantum filters we shall consider now only
devices which show on their displays 1 or 0 and the interpretation depends on
the type of the device which can be easily inferred from its name. So for example
the measuring device labelled as

GREATER-THEN-4

shows on its display 1 if the number read on the die is 5 or 6 and it shows 0 in
other cases.

Let us forget for the moment that we know that it is the die what we observe

13
with our devices and that we even do not understand the meaning of the English
words on the labels of the measuring devices. We just throw the die then mea-
sure it with different devices (we are in the classical world, so the measurement
does not change the state!) prepare another state by throwing the die, measure
it again and so on. Gaining more and more experimental data we find certain
rules to hold. Something like the following.

GREATER-THEN-4 is not mutually exclusive with LESS-THEN-6 because


there are states for which both the devices show 1 on their displays.

GREATER-THEN-4 excludes EQUAL-TO-3

ODD-NUMBER is mutually exclusive with EQUAL-TO-FOUR

After all we recognize that the devices labelled by ”EQUAL-TO” form a special
recognizable subset from all the devices we have. These devices are special since

• They form a complete set what means that for any state at least one of
them shows 1 on its display.
• They are all mutually exclusive, what means that never more then one of
them shows 1.
• They are irreducible what means that for any two states such that for
both of them the same ”EQUAL-TO” device show 1 then the readings on
all the other devices for these two states are equal.

It follows from these properties that any other measuring device can be con-
structed from the elementary ”EQUAL-TO” by the logical XOR-gates. For
example

LESS-THEN-3 = EQUAL-TO-1 XOR EQUAL-TO-2

We can use OR-gates instead of exclusive XOR-gates, sice the elementary de-
vices are mutually exclusive anyhow.

The states corresponding to the irreducible complete mutually exclusive mea-


surements are in the same way irreducible, complete and mutually exclusive.
These elementary states can be represented by the sets
{1}, {2}, {3}, {4}, {5}, {6}
The elementary states are ”microstates”: the term borrowed from the statis-
tical physics denoting the state which is completely determined (no missing
information or ”zero” entropy). All the other classical states are ”macrostates”,
incompletely determined states (non-zero entropy) which can be constructed
from the elementary state by the set operation ∪. For example the state
{1} ∪ {2} ∪ {3}

14
is the state recognized by the measuring device LESS-THEN-4 . In the classical
world the operator OR is therefore represented by the set operator ∪.

So in classical world we can also have ”concurrent mutually exclusive possi-


bilities” when allowing for macrostates. These are mixtures of the elementary
states. Different instances of the same macrostate can be treated (for a par-
ticular purpose) as equivalent, they are not ”ontologically” equal. Refining the
measurement proces I can always distinguish between different instances of the
same macrostate until I arrive to the level of elementary states which are not
further reducible. Therefore I assume that all the instances of the elementary
state are for all practical purposes equivalent. I can treat them as if they were
the same. So, referring to the Occam’s razor, they are the same.

In the classical world the ”concurrent mutually exclusive possibilities” implies


reducibility. In the quantum world, at least in the orthodox quantum world, we
have quite a different situation. As we have seen concurrent mutually exclusive
possibilities are realized not only as mixtures 1 but as pure states as well, the
pure states: the superpositions. The superposition is a trick how to get both
irreducibility and ”concurrent mutually exclusive possibilities”.

If the ”concurrent mutually exclusive possibilities” are labelled by the logical


symbol ”XOR” than this quantum mechanical XOR cannot be mathematically
modelled as unions of disjoined sets because disjoined sets imply reducibility
(each element of the union can be uniquely identified as belonging to exactly
one of the disjoined subsets). So if we believe to the irreducibility of the pure
quantum state, we must have different model of logic: the quantum mechanical
XOR is modelled as linear superposition of mutually orthogonal vectors.

So the question is why we believe to the irreducibility of the states such as


|ϑi = |↑i cos(ϑ/2) + |πi sin(ϑ/2)
To investigate this problem we have to make experiments with instances of the
state |ϑi. Let us imagine that we are the pioneers who are discovering the
quantum world and are playing the same game with the quantum devices as
was described above about the pioneers in the classical world investigating the
game of dice.

So the state |ϑi is the polarization state of any electron which passed through the
SGF(ϑ). So if the electron is found behind the filter (See Fig.25) it is assumed
to be in the state |ϑi, more exactly it is assumed to be in some instance of
the state |ϑi. We try to be unbiased and allow for possible reducibility of the
state prepared by the filter SGF(ϑ) and therefore we allow for a possibility of
unequivalent instances of the state |ϑi.
1 We have not discussed mixtures of pure quantum states in this text. The mixture states

can be created in meny different ways. The most transparent one is for example to produce
an electron beam by randomly switching between two electron guns. If the guns produce
electrons in different states the beam is the statistical mixture of electrons: some of them
are the instances of the quantum state produced by the first gun, the other are instances of
the state produced by the second gun. The resulting state is a ”macrostate”. Without going
to details we just say that such a macrostate is mathematically described byt the density
matrix.

15
D

Figure 25: Preparing instances of the state |ϑi

When we start resolving the question of the irreducibility of the state |ϑi we
immediately get suspicious that the state is reducible: if we measure the state
(See Fig.26) prepared by the SGF(ϑ) by the SGF(↑) we find that we are not
getting the unique reading from the SGF(↑). Some instances of |ϑi give 1, other
give 0 on the display of the SGF(↑).

Figure 26: Measuring the state |ϑi by SGF(↑)

So we decide to change the experimental setup to a complementary one and to


measure the state |ϑi with the filter SGF(π)- see Fig.27. We again find that
some instances of |ϑi give 1, other give 0 on the display of the SGF(π). We

Figure 27: Measuring the state |ϑi by SGF(π)

also find that (at least statistically when we do large number of observations on
both experimental setups) number of instances that pass through the SGF(↑)
is equal to the number of instances that are stopped by the SGF(π) and, com-
plementarily, the number of instances which are stopped by the SGF(↑) is the
same as the number of instances which pass through the SGF(π).2
2 The second part of the statement is the logical consequence of the firs part. This is

16
It seems that we have the classification of the instances of the state |ϑi into
two distinct classes: those which pass through the SGF(↑) and those which
pass through the SGF(π). However, this is an unjustified conclusion. From
the observed numbers we cannot infer that the instances which pass through
the SGF(↑) are those which are stopped by the SGF(π). Such a hypothesis is
consistent with what was said, but is not necessary.

Here is an example with specific numbers. Suppose we had 200 electrons passed
through the first SGF(ϑ). 100 of them was subsequently measured by the
SGF(↑): 60 passed through and 40 were stopped. The other 100 electrons
were measured byt the SGF(π), 40 passed and 60 were stopped. One solu-
tion is obviously that we have two classes of states: one (pass, stop) the other
(stop, pass) and we had 120 states of the first class and 80 states of the second
class. Another solution may be that we have four classes (pass,pass),(pass,stop),
(stop,pass) and (stop,stop). And we had for example 30 states of the first class,
30 states of the second class, 10 states of the third class and 30 states of the
fourth class. And it seems we cannot resolve which reduction is the correct
one because we would need the result of the simultaneous measurement (on the
same instance of the state) with both the filters, which is not possible.

Experimenting further we find that even if such a reduction was true it cannot
be the full story. Because we find that we can make very similar reduction with
a different pair of filters, for example with SGF(π/2) and SGF(3π/2). And we
need even more classes of states.

The orthodox quantum mechanics offers an elegant solution of the problem: all
possible reductions but no elementary states: linear superposition opens the way
for having all possible reductions simultaneously. The choice of base vectors in
a linear space is arbitrary. No decomposition into disjoint sets ever happens, so
no elementary states. But no ”classical logic” XOR operation either! Quantum
mechanical XOR is not the union of disjoint sets, it is a linear superposition. But
the prediction of the outcome of a particular measurement is non-deterministic,
instead of unique predictions quantum mechanics offers probabilities.

4 Hidden parameters

We have just described two possible solutions of the state classification problem
in the quantum world. One is the orthodox quantum mechanical description
with essential probabilities. The other option is to have much richer space of
states: each quantum state is decomposed into disjoint sets of more elementary
states. There are many classification schemes of this type consistent with the
result of measurements, at least of the measurement of the type described in
the exercises in Section 2. Disjoint elementary states allow for classical logic
and determinism. Such theories are called theories with hidden parameters. We
shall present one specific example of such a theory further below.
because the readings of the measuring devices are classical states, therefore standard logic is
applicable. The readings 1 and 0 are elementary mutually exclusive complete events.

17
To illustrate the essential difference between the orthodox quantum mechanics
and theories with hidden parameters us use a slightly different experimental
setup (see Fig.28). Two filters are put on a movable support so that we can
choose which filter to use very flexibly. And we do an experiment in which
we make a ”late choice” of the filter. We postpone the decision until after the
electron is already out of the first filter SGF(ϑ) before it reaches the place where
the second filter is to be positioned. So in the moment when the state |ϑi is
established even we do not know which measurement device we are going to use.
So the electron”does not know it” either.

G ?

Figure 28: Late choice of the SGF

Now the crucial question is in which moment of time the electron decides
whether it will go through the detector when it hits it. There are two pos-
sibilities

• the electron makes its decision before we make the decision which filter
we use
• the electron makes its decision only after we make our decision about the
detector. So in this case the electron ”does not want to play our game”.
We want to fool the electron by making the late decision, but the electron
is smarter then we and as a ”countermeasure” it makes its decision only
after we do our one.

If the first option (”early electron decision”) is true, the electron decision must
take into account all the possibilities which could ever happen: it must carry a
list of all possible detectors which it can hit and the decision whether to pass
through or not.

If the second option (”late electron decision”) is true the electron need not
to consider all the possible options because it already knows which detector
stands in its way and so it makes the decision only with respect to that specific
detector. So only one decision is ever made and all the other questions like ”what
would that particular electron do, had we chosen another detector”, remain
unanswered. Actually such questions are illegal.

The orthodox quantum mechanics says that the world is constructed according
to the late decision principle and the particular decision is done by pure chance:
electron hits a particular detector and knowing its quantum state it calculates

18
the probability p to pass through it. It then plays the dice: it generates a
random number ξ ∈ (0, 1) and if ξ < p it goes through, else it is stopped.

Let us now describe in somewhat more detail a specific ”early electron decision
model”. The essential point is that everything is decided in the moment when
the electron is leaving the ”state preparing” filter. As we have already agreed
the filter ”imprints” on the leaving electron the out-state. In the early decision
model the out-state not only carries the information which particular filter has
prepared the state, but also the information from which one can deterministi-
cally deduce whether the electron passes through any thinkable filter which can
follow.

Let us assume that the piece of information in the out-state which codes the
state preparing measurement will be identical to that in the quantum mechan-
ics. So if the device was SGF(ϑ), the ”complete” out-state is formed by the
corresponding vector from the standard Hilbert space plus a single real param-
eter λ ∈ (0, 1). This parameter codes the ”behind the quantum mechanics”
information on the outcome of all the possible future experiments. In our case
of electron polarizations all the measurement are done with the SGF filters.
Since no two SGF filters can be applied simultaneously, the set of ”all the pos-
sible future experiments” is quite simple: it is the set of SGF filters arbitrarily
axially rotated. The parameter λ has to code the outcome of measurement by
arbitrarily rotated filter SGF(ϕ). Let us take a simple coding:
ϑ−ϕ
if λ < cos2 ( ) electron passes through, otherwise it is stopped
2

As an exercise the reader can verify, that if the parameter λ is a random number
distributed uniformly in the interval (0, 1), the results of all the experiments of
the type described in the Section 2 are the same as in the orthodox quantum
mechanics. The parameter λ is called the ”hidden parameter”: we do not have
any theory on its dynamics, we cannot measure it by any available apparatus.
It is just an arbitrary mathematical tool which changes the late-decision-type
quantum theory into the early-decision-type theory.

Introducing the hidden parameter we were able to decompose the quantum state
into elementary mutually exclusive states labelled by λ, we have classical ”XOR-
composition” of states modelled by set union operator. However, the difference
between the hidden-parameter theory and our classical example of the states of
the die is that the decomposition into the elementary disjoint states does not
have the corresponding counterpart in the space of measuring devices. The point
is that we cannot drop the ”typically quantum” feature of changing the state
after each measurement. The electron in the incoming hidden-parameter-state
(|ϑi , λ)
after passing through the filter SGF(ϕ) will be in the out-state
(|ϕi , λ0 )
where the new value of the hidden parameter λ0 is a new random number with
no correlation to the old value λ whatsoever. Otherwise the predictions of the

19
new hidden-parameter-theory would be different from those of pure quantum
mechanics and ruled out by the experiment.

Since we do not have the decomposition of measuring devices into ”elementary


mutually disjoint complete system” our theory with the hidden parameters is
still not ”completely classical”. Some questions which are legal in the classical
physics may be illegal in quantum theory even in the presence of the hidden
parameters, because the measuring devices are still mutually incompatible”. So
the crucial question

We have chosen to use the filter SGF(ϕ) and the electron passed through.
What would be the outcome had we chosen the filter SGF(ϕ0 )?

is illegal in the hidden-parameter theory in the same way as it is in the orthodox


quantum mechanics. We cannot attack experimentally such a problem because
of mutual incompatibility of different SGF’s.

Let us repeat briefly the argumentation leading to such a conclusion

• Different SGF’s are described by non-commuting operators therefore their


eigenstates and eigenvalues are mutually incompatible

• By physical construction two different SGF’s cannot be used simultane-


ously but only in sequence.
• The state of the electron leaving the filter is changed with respect to what
it was before entering the filter.
• Subsequent measurement is performed on the changed state.

• Therefore it is (experimentally) illegal to speak about simultaneous values


of mutually incompatible physical quantities.

On the other hand in our particular model of the hidden-parameter theory the
question is legal an can be answered by the theory. Knowing the parameter λ we
can answer the question what would be the outcome of all thinkable experiments.
Unfortunately, the parameter λ is hidden from us. It should be also stressed that
it is possible to formulate theories having more hidden parameters for which the
answers to questions concerning alternative choice of measuring devices would be
quite different. However experiments we have discussed so far cannot distinguish
between various hidden-parameter theories nor between the orthodox quantum
mechanics and any reasonable hidden parameter theory.

However Einstein, Rosen and Podolski realized that there is a flaw in the above
argumentation concerning experimental inaccessibility of simultaneous values of
mutually incompatible quantities.

We said already that Einstein was very much willing to have simultaneous val-
ues of mutually incompatible quantities in the theory, but by force of the Bohr’s

20
argumentation,he had to retreat from his position and accept mutual incompat-
ibility of operators and the change of state by measurement.

The theory with hidden parameters provides the way to have both determin-
ism and indeterminism and both simultaneous values of physical quantities and
mutual incompatibility of measurements. The prize one has to pay for it is the
incompatibility between the ontological status and the gnoseological status of
the theory: the state of the system cannot be recognized as it really is. Onto-
logically the outcome of the measurement is deterministic, gnoseologically the
predictions of the theory are only probabilistic.

Such a status of the theory could be hardly acceptable for Einstein. (Just
recall that it was the very careful analysis of the measurement of time intervals
and clock synchronization which paved the way for abandoning the notion of
absolute time.)

The message of the EPR paper was that there is a way how to make the simul-
taneous values of incompatible quantities experimentally accessible while still
accepting all the Bohr’s arguments about incompatible measurements.

5 The EPR idea

The primer interest of authors of the EPR paper was the question of simultane-
ous measurement of the position and momentum with a precision higher then
that compatible with the uncertainty principle. We shall rewrite the story for
two incompatible SGF filters.

EPR realized that there might be a way to experimentally access incompatible


quantities simultaneously. The task is to measure two incompatible quantities3
SGF(ϑ) and SGF(ϕ) on a given (arbitrary but unknown) electron state. The
EPR trick is that we measure one quantity, say SGF(ϑ), on the given electron
and the other measurement we perform indirectly. We do not apply the device
SGF(ϕ) on the given electron, this is not possible without interfering with the
measurement of SGF(ϑ).

Instead we measure some other suitable quantity on some other system and
this other quantity is carefully chosen in such a way that from its value we
can deduce what would be the value found by the measurement of SGF(ϕ)
on the given electron had we chosen to measure SGF(ϕ) instead of SGF(ϑ).

It should be stressed that whatever the EPR suggestion is, it cannot work in
the orthodox quantum mechanics: in the late-decision-type theory the question
of simultaneous values of incompatible quantities is strictly illegal. Whatever
3 Here we again use the language treating SGF’s as measuring devices: the SGF(ϑ) mea-

sures the quantity represented by the operator |ϑi hϑ|. Then instead of saying ”the quantity
measured by the device SGF(ϑ)” we simply say the quantity SGF(ϑ).

21
the EPR procedure is, the result of it cannot and must not be interpreted as the
simultaneous measurement of quantities within the orthodox quantum theory.

The EPR paper wanted to show that in the early-decision-type of theory with
hidden parameters theoretically existing simultaneous values can be measured
(at least in some cases).

The observation of the EPR is that in some cases we expect strong correlations
between different physical quantities. That opens a way how to deduce the value
of some quantity without measuring it directly: we can measure some other
quantity with which it is correlated and from the result of that measurement
we can calculate the value which interests us.

A typical example of strong correlation is represented by any conservation law.


For example if we know the total energy of some system which consists of two
subsystems and later on we measure the energy of one of the subsystems, then
we do not need to measure the energy of the other subsystem. We can calculate
its value using the energy-conservation law.

In our case of electron polarizations we shall use angular momentum conser-


vation. We shall consider two particles (the electron and the positron) which
are created by decay of somr object with angular momentum zero at rest. The
two particles move in the opposite directions. Similarly as it was in the single
particle case we shall be interested only in the polarization states of the parti-
cles. We shall assume that the movement of the two particles can be treated
classically. Actually we just assume that the two particles move in the opposite
directions from the point of their common origin. A short time after their birth
their distance is large enough and we shall assume that any polarization mea-
surement using some SGF filter can be performed separately on any of these
two particles. This means that the particle does not interact with any SGF
filter which is being used on the other side to measure the other particle there.
To simplify the discussion we shall consider two independent experimentalists,

Bob

D Alice

Figure 29: Laboratories of Alice and Bob on the opposite sides of the point of
origin of the two particles

Alice and Bob, who have independent laboratories on the opposite sides, so for
each of them exactly one particle of our particle pair is available for making
measurements. Both Alice and Bob can perform any experiment with SGF de-
vices on their particle (See Fig.29). The SGF devices belonging to Alice will
be denoted as SGFA and those belonging to Bob as SGFB . Similarly (by the

22
appropriate index) we shall distinguish the single particle states.

6 EPR in orthodox quantum theory

We shall start our discussion within the framework of the orthodox quantum
mechanics. We shall call the initial polarization two-particle state as the EPR
state. Without going into details4 we just write the EPR state in terms of single
particle states as
1 1
|EP Ri = √ |↑A i |↓B i − √ |↓A i |↑B i
2 2

In the above equation we have used specific base states |↑i and |↓i. However, the
|EP Ri state is symmetric with respect to axial rotation, and so its expression
using any other (rotated) base will be the same5 , for example
1 1
|EP Ri = √ |ϑA i |ϑ⊥B i − √ |ϑ⊥A i |ϑB i
2 2

To worm up we first do a few easy exercises. First we calculate the probabil-


ity pA (↑) that the Alice’s particle passes through SGFA (↑). The appropriate
operator is |↑A ih↑A |. We get
1
p(↑A ) = hEP R |↑A ih↑A | EP Ri =
2

Now calculate the probability that Bob’s particle passes through SGFB (π/3).The
appropriate operator is |(π/3)B i|(π/3)B i. Realizing that

h↑B | (π/3)B i = cos(π/6)

h↓B | (π/3)B i = sin(π/6)


We get
1  1
p((π/3)B ) = hEP R | (π/3)B i h(π/3)B | EP Ri = cos2 (π/6) + sin2 (π/6) =
2 2

In general we get
1
p(ϑB ) =
2
4 The reader not familiar with rules of angular momentum composition can find the relevant

information in any textbook on quantum mechanics, where he shall look for the keyword
”Clebsh-Gordan coefficients.
5 Because of this axial symmetry of the EPR state every result we shall obtain will depend

only on relative angles between the devices in Alice’s laboratory with respect to those in Bob’s
laboratory.

23
for arbitrary angle ϑ. And the same is true, of course, for Alice’s measurements.

However, things start to be more interesting when we look for correlations. For
example let us calculate the joint probability that Alice’s particle passes through
SGFA (↑) and Bob’s particle passes through SGFB (π/3). The appropriate op-
erator now is
|↑A i |(π/3)B ih(π/3)B | h↑A |
and we get
1 1
p(↑A , (π/3)B ) = hEP R |↑A i |(π/3)B i h(π/3)B | h↑A | EP Ri = sin2 (π/6) =
2 8

So we see there is a correlation between the outcomes of the measurements on


the Alice’s and Bob’s sides since
1 1
= p(↑A , (π/3)B ) 6= p(↑A )p((π/3)B ) =
8 4

Let us now calculate the joint probability that Alice’s particle passes through
SGFA (↑) and Bob’s particle does not pass through SGFB (π/3). The appropriate
operator now is   
|↑A i h↑A | 1 − |(π/3)B i h(π/3)B |

and we get
 
p(↑A , N OT (π/3)B ) = hEP R| |↑A i h↑A | 1 − |(π/3)B i h(π/3)B | |EP Ri =
1 1 3
= − sin2 (π/6) =
2 2 8

Similarly we can calculate the remaining two possibilities of the joint probability
and we get
3
p(N OT ↑A , (π/3)B ) =
8
1
p(N OT ↑A , N OT (π/3)B ) =
8

Total correlation between Alice’s and Bob’s measurements using SGFA (↑) and
SGFB (π). We get
p(↑A , πB ) = 1/2
p(↑A , N OT πB ) = 0
p(N OT ↑A , πB ) = 0
p(N OT ↑A , N OT πB ) = 1/2

To fact of the total correlation is better seen from the conditional probabilities.
For example
p(↑A , πB ) = p(πB | ↑A )p(↑A )

24
Since p(↑A ) = 1/2 and p(↑A , πB ) = 1/2, we get

p(πB | ↑A ) = 1

where p(πB | ↑A ) is the conditional probability that Bob’s particle passes through
the SGFB (π) if we know that Alice’s particle passed through the SGFA (↑).

We used the formulation ”if we know that Alice’s particle passed...” which seems
to suggest that we are speaking about correlation in pieces of knowledge, but
the correlation is rather about real events. To stress this aspect we can have
the experimental setup as presented in Fig.30. There the Alice’s and Bob’s
laboratories are positioned asymmetrically with respect to the common point
of origin of the particles. Bob’s laboratory is much farther from the origin then
Alice’s one. So when Alice’s particle passes through a filter in her laboratory,
Bob’s particle is still far from his device. So we speak about genuine conditional
probability: the event in Alice’s laboratory already happened. So the probability
statement is not about lack of our knowledge what happened at Bob’s side.
Nothing happened yet there. But the mere fact that an event happened at
Alice’s side changes the expectations (probabilities) at Bob’s side. Before Alice’s
measurement the probability for the Bob’s particle to pass through SGFB (π)
was 1/2. After the Alice’s particle did pass through SGFA (↑), the probability
for the Bob’s particle to pass through SGFB (π) changes to 1.

Bob

D Alice

Figure 30: Alice’s and Bob’s laboratories positioned asymmetrically with respect
of the point of origin of the particles they measure.

Our statement about ”probability changing to 1” is relevant to the late-decision-


type theory of the orthodox quantum mechanics. The statement is about
the interpretation of the experimental results. Experimentally we do not
directly observe any probabilities. What we observe is the total correlation be-
tween the results of Alice’s measurement by SGFA (↑) and Bob’s measurements
by SGFB (π).

In the orthodox quantum mechanics we interpret the experiment in the language


of late decisions and probabilities in the following way.

The system is initially in the |EP Ri state. It is a two-particle state. When


Alice performs her measurement, we interpret it as a a measurement performed
on the two-particle system. So we know that the state of the two particle system
must change as a consequence of that measurement.

25
The only measuring devices we consider in these notes are the SGF filters. The
operators corresponding to such devices are projectors. In that case the rule how
to write down the out-state (the state of the system after the measurement) from
the in-state (the state of the system before the measurement) is quite simple.
It is described in the following box.

• We begin with properly normalized |in-statei


• The operator corresponding to SGF(ϑ) is the projection operator

P̂ = |ϑi hϑ|

• We apply the projection operator to the in-state and get an auxiliary


vector
|auxi = P̂ |in-statei

• We get the out-state of the particle which passed through the filter from
the auxiliary vector by normalizing it to 1:
1
|out-statei = p |auxi
haux | auxi

Applying this algorithm to the situation after the Alice found that her particle
did pass through the filter SGFA (↑)
|auxi = |↑A i h↑A | EP Ri =
 
1 1
= |↑A i h↑A | √ |↑A i |↓B i − √ |↓A i |↑B i =
2 2
1
= √ |↑A i |↓B i
2
|out-statei = |↑A i |↓B i

The |out-statei is the two-particle state! It is the state of the system just after
Alice’s experiment and before the Bob’s experiment. So it is the in-state from
the point of view of Bob. By inspection it is now clear why the Bob’s particle
passes through SGFB (π) with certainty.

Similar absolute correlation we get for any pair of SGF’s relatively rotated to
each other by the angle π.

As an exercise the reader can prove that there is a total autocorrelation between
Alice and Bob if they use a pair of equally oriented SGF’s.

7 EPR and hidden parameters

The theory with hidden parameters is the early-decision theory. It means that
the outcome of any future experiment must be decided in the moment of ”birth”

26
of the two particles. Both Alice’s and Bob’s particle therefore ”carry with them”
the list of decisions (coded into hidden parameters) whether they pass through
any possible SGF filter on their way. Both Alice’s and Bob’s laboratory may be
equipped by changers providing the flexibility to make the final decision what
filter to use when particles are already on they way (See Fig.31) Alice and Bob

Bob

Alice
D

Figure 31: Alice’s and Bob’s laboratories with changers providing flexibility to
make the final decision which filter to use when particles are already on their
way. .

can use any filter SGF(ϑ) axially rotated by arbitrary angle ϑ. For simplicity
we shall discuss just a limited discrete set of filters

SGF(↑), SGF(π/3), SGF(2π/3), SGF(π), SGF(4π/3), SGF(5π/3)

Both Alice and Bob can decide (in principle randomly) which one of these
filters they use for the measurement. The corresponding theory with hidden
parameters must be able to determine (to decode from the hidden-parameters
values) whether or not the particular particle would pass through any of these
detectors.

There are many possible theories6 with hidden parameters. Without loss of
generality we can imagine that the particles carry with them also the table
of 6 binary digits (flags) providing the information already decoded. The bits
correspond to the 6 possible SGF filters. If the bit has value 1 it means the
particular particle would go through the corresponding detector, 0 means the
particle would be stopped.

An example of a possible list of flags is presented int the following table. Now

↑ π/3 2π/3 π 4π/3 5π/3


1 0 1 0 1 0

we should realize that our theory with hidden parameters has to respect the
6 It should be stressed that this is a textbook-type of text. The primary goal here is to

explain some idea. A more rigorous meaning of this statement is more like ”At this stage of
discussion it seems that there are many possible theories with hidden parameters...”. Devel-
oping the arguments further in this and the following section we shall find that simple theories
are ruled out by the experiment. And, who knows, maybe no satisfactory theory with hidden
parameters can be formulated at all.

27
correlations between the Alice’s and Bob’s findings as discussed in the previous
section, otherwise the theory would be quickly ruled out by experiments. After
a little thinking we find that if the table as presented above is carried, say, by
Alice’s particle, the the table of hidden flags carried by the Bob’s particle is
completely determined: each flag is just opposite to that of the Alice’s particle
flag.7 The complete anticorrelation between the Alice’s and Bob’s flags is essen-

↑ π/3 2π/3 π 4π/3 5π/3


0 1 0 1 0 1

tial for the EPR proposal. Suppose that Alice decides to measure her particle
with the SGFA (π/3). Whatever results she gets she is (if she believes to the
early decision model) allowed to ask the question: what would be the result
had I chosen SGFA (2π/3) instead? She cannot answer the question by means
of her measuring devices, but she can ask Bob for help if they believe that
the flags of the Bob’s particle are not influenced by the fact that Al-
ice did perform already some measurement. It is enough then that Bob
measures his particle with his device SGFB (5π/3). Because of the conservation
law whatever he gets is equal to what Alice would have got had she decided to
use the SGFA (2π/3).

Now we see that Alice has the results of two incompatible measurements for the
same state of her particle, the first result she got directly through her measure-
ment, the other she gets indirectly, inferring it from the result of the correlated
Bob’s measurement.8

This is a vary strong statement indeed. If true, it would mean that (at least
in this case) the hidden parameters are not quite hidden: at least two of them
can be determined for any particular state. So the hidden parameters acquire
”ontological status”.

It should be stressed again, that the EPR trick is worthless if the late deci-
sion theory (like the orthodox quantum mechanics) is the true theory. There,
even if we are aware of the correlations between appropriate Alice’s and Bob’s
experiments when they are really performed, we cannot argue that the
correlation remains to exist between theperformed Bob’s experiment and
unperformed Alice’s experiment. In the early-decision theory such a cor-
relation remains to be true because in fact it is a correlation between the values
of flags and only through them it is a correlation between the experiments.
Since the existence of flags in the early-decision theory does not depend on the
fact whether the corresponding experiments were performed or not, the correla-
tion between the results of possible experiments is the same as the correlation
7 A little more thinking and one realizes that even the flags in the Alice’s table are not

independent from each other. They are correlated via flags in Bob’s table. For example
Alice’s flag for π/3 must be exactly opposite to Bob’s flag for 4π/3 which in turn must be
equal to Alice’s flag for 4π/3 so Alice’s flags for π/3 and 4π/3 must be opposite to each
other. So from the 6 considered Alice’s flags only 3 are independent and the other 3 are their
opposites.
8 The authors of the EPR paper were considering position and momentum of the particle

as the pair of incompatible quantities. The EPR trick, if found consistent, thus allows to
overcome the obstacle of the uncertainty principle.

28
between the experiments really performed.

8 Bell inequalities

The theory with hidden parameters seems to be very attractive. The question is
to what extend it is compatible with the results of existing experiments. Since
no experiment so far performed was found to be inconsistent with the predictions
of the orthodox quantum mechanics, the crucial question is to what extent the
predictions of the hidden-parameter theory can be made close to the predictions
of the orthodox quantum mechanics.

Let us first consider the single particle experiments. The question is whether
we can assign probabilities to all the possible flag tables in such a way that
all the single particle experiments give the same results as in the orthodox
quantum theory. Since we know that for the |EP Ri state the probability of
a single particle to pass through any filter is 1/2, it is enough to set the flags
just randomly to reproduce all the single particle experiments.9 So there exists
a hidden-parameter theory which gives for the single particle experiments the
same results as the orthodox quantum mechanics. Therefore it is not possible
to distinguish between the late-decision and early-decision theories by single
particle experiments with the |EP Ri state.

So we have to turn to correlation experiments performed simultaneously in


Alice’s and Bob’s laboratories and try to design a strategy which would help
to distinguish between the early-decision and late-decision theories. So the
question is whether there exists a hidden parameter theory the results of which
are undistinguishable from the orthodox quantum theory for both single particle
and correlation type of experiments.

The answer is due to J.Bell and is negative. It can be proved that no hidden
parameter theory10 can reproduce the results of quantum mechanics for all
correlation experiments. The proof is surprisingly simple. Let us consider the
experimental situation as presented in Fig.31. We shall consider 3 experimental
setups as presented in the following table The experimental procedure will be

Experiment Alice Bob


1 SGFA (↑) SGFB (5π/3)
2 SGFA (2π/3) SGFB (π/3)
3 SGFA (4π/3) SGFB (π)

the following. Particle pairs are sequentially produced in the point of origin. We
randomly change the experimental setup choosing randomly between setups ”1”,
”2”, and ”3”. Then the cycle is repeated indefinitely. For each pair Alice and
Bob write down their observation, that is they write down ”1” if they observed
9 Of course, theory with randomly assigned flags would be immediately ruled out by the

correlation experiments.
10 More exactly no local hidden parameter theory, see the discussion later.

29
their particle to pass through or write down ”0” if their particle was stopped.
In this way we obtain a record consisting of of pairs of observation, for example

{1, 0}1 , {1, 1}3 , {0, 1}1 , {0, 1}2 , {1, 0}2 , {1, 0}3 {1, 1}1 , {0, 1}1 , {0, 0}3 , · · ·

Now we shall be specially interested in pairs of observations consisting of equal


flags. To simplify the formulations we shall call any event which looks like
{0, 0} or {1, 1} as Bell event. So we shall be interested in Bell’s events. More
specifically we shall calculate the mean number n(Bell) of Bell events observed.
In the record presented above there are 3 Bell events.

Within the framework of the orthodox quantum mechanics the mean number of
Bell’s events can be easily calculated. Using the results presented in Section 6
we find
1
n(Bell) =
4

Now let us investigate the case of hidden parameter theory. Surprisingly enough
even without specifying the hidden-parameter-model in any detail we can easily
find a reasonable lower estimate for n(Bell).

It is enough to realize (see Fig.31) that Bob is using devices which are com-
plementary to Alice’s ones. Because of angular momentum conservation (total
correlations between the complementary devices) the outcome of Bob’s exper-
iment can be determined using just the hidden parameters (hidden flags as
discussed above) carried by Alice’s particle. Specifically, the outcome of Bob’s
experiment with SGFB (5π/3) is given by the Alice’s particle flag corresponding
to SGFA (2π/3). Within the EPR philosophy the experimental setups 1,2, and
3 can be reexpressed in the language of ”Alice’s measurements” only like

Experiment Alice Bob


1 SGFA (↑) SGFA (2π/3)
2 SGFA (2π/3) SGFA (4π/3)
3 SGFA (4π/3) SGFA (↑)

30
Now for example assume that the three relevant hidden flags carried by Alice’s
particle are as follows

↑ 2π/3 4π/3
1 1 0

If that particular particle was a member of particle pair undergoing the experi-
ment in the setup ”1”, the outcome would be (1,1), in the setup ”2” we get the
outcome (1,0) and in the setup ”3” the outcome would be (0,1).

So we see that the outcome of the experiment in any possible setup (1,2, or 3)
is uniquely determined byt the triplet of hidden flags carried by Alice’s particle.
Since from three flags at least two must be equal (each flag is 0 or 1!) we
conclude that in any case at least in one of the three possible experimental
setups we must observe the Bell’s event. Therefore the mean number of Bell’s
events 1/3. So we derived the Bell’s inequality11

n(Bell) ≥ 1/3

The inequality is called Bell inequality and should be compared with the unique
prediction of quantum mechanics
1
n(Bell) =
4

So the quantum mechanics prediction violates the Bell inequality predicted by


any reasonable hidden parameter theory. This result is valid for a very general
class of hidden parameter models. Actually we needed just two assumptions:

• the model is of the early decision type


• once the hidden parameter values are set at the moment of particle cre-
ation, they are not changed by interactions of the companion-particle at
a distant location

There exist many different experimental set-ups for which a similar result can
be obtained. The general scheme is as follows.

• the experiment (with possibly changing experimental set-ups) is repeated


many times and some statistical characteristics (a number) is calculated

• if a hidden parameter theory is valid the obtained number should satisfy


some inequality (called Bell inequality)
11 Similar inequalities can be derived for suitable combinations of correlation experiments

for other two-particle systems. These inequalities are commonly known as Bell’s inequalities
and their common attribute is that they do not depend on detailed specification of the hidden-
parameter model considered.

31
• If quantum mechanics is valid, than the Bell inequality should be violated
by the number obtained

Of the order of ten experiments of the kind just described were performed and
the majority of them claims finding violation of the relevant Bell inequality.
However, the problem is not considered as settled. Many objections were for-
mulated by critics suggesting possible flaws in the performed experiments. The
details of the discussions are beyond the scope of this notes.

32

You might also like