a

r
X
i
v
:
1
3
1
2
.
3
3
6
6
v
1


[
q
u
a
n
t
-
p
h
]


1
1

D
e
c

2
0
1
3
Quantum uctuations from a local-causal information dynamics
Agung Budiyono

Jalan Emas 772 Growong Lor RT 04 RW 02 Juwana,

Pati, 59185 Jawa Tengah, Indonesia
(Dated: December 13, 2013)
Abstract
We shall show that the abstract and formal rules which govern the quantum kinematic and
dynamics can be derived from a law of change of the information content or the degree of uncer-
tainty that the system has a certain conguration in a microscopic time scale, which is singled
out uniquely, up to a free parameter, by imposing the condition of Macroscopic Classicality and
the principle of Locality. Unlike standard quantum mechanics, however, the system always has
a denite conguration all the time as in classical mechanics, following a continuous trajectory
uctuating randomly in time. Moreover, we shall show that the average of the relevant physical
quantities over the distribution of the conguration is equal to the quantum mechanical average of
the corresponding quantum mechanical Hermitian operators over a quantum state.
PACS numbers: 03.65.Ta; 03.65.Ud; 05.20.Gg
Keywords: Reconstruction of quantum mechanics; Physical origin of quantum uctuations; Information
dynamics; Principle of Locality; Macroscopic Classicality

Electronic address: agungbymlati@gmail.com

1
I. MOTIVATION
The violation of Bell inequality by quantum mechanics is widely believed to lead to a
bizarre conclusion that quantum mechanics allows the statistical results of a pair of mea-
surement events spacelike separated from each other to have a stronger correlation than that
is allowed by any local-causal theory [13]. The nonlocal correlation has been claimed to
be veried in numerous experiments [416], in spite of the fact that no experiment hitherto
conducted is free from loopholes [17, 18]. Such a nonlocality prima facie contradicts the spirit
of the special theory of relativity which presumes a nite maximum velocity of interaction
given by the velocity of light in vacuum. Further careful investigation however showed that
the quantum mechanical nonlocal correlation can not be exploited by one party to inuence
the statistical results of measurement performed by the other distantly separated party, thus
prohibits signaling, in accord with the assertion of the special theory of relativity [1923].
The co-existence of nonlocal correlation and no-signaling in quantum mechanics has in-
spired some authors to ask if quantum mechanics can be derived from a certain balance
between some kind of nonlocality and the principle of no-signaling [2428]. While it is
shown that the constraints put by the nonlocal correlation and no-signaling are not su-
ciently strong to single out quantum mechanics [2628], it has renewed an interest in an
approach to clarify the meaning of quantum mechanics by deriving its formal mathematical
structures and numerous abstract postulates from a set of conceptually simple and phys-
ically transparent axioms. In such a program, one attempts to directly answer the most
tantalizing foundational question: why the quantum? [29]. One of the advantages of
the program to reconstruct quantum mechanics is that it might provide physical insights for
possible natural extensions of quantum mechanics either by modifying the axioms or varying
the free parameters that are left unxed by the axioms. Extension of quantum mechanics
is not only necessary to set up precision tests against quantum mechanics, but might turn
out to be the necessary step to solve some of the foundational problems of the latter.
A lot of works along this line has been reported recently by regarding information as the
basic ingredient of Natural phenomena [3047] : all things physical is information-theoretic
in origin thus It from Bit [29]. In those works, one searches for a set of basic features of
information processing which can be promoted as axioms to reconstruct quantum mechanics.
Such an approach is partly motivated by the advancement of quantum information science
2
[48]: the fact that quantum mechanics allows information processing tasks that can not be
performed at least as eciently within classical mechanics suggests an intimate relationship
between the the foundation of quantum mechanics and the basic features of information
processing. This approach thus plays with information within operational-instrumentalist
theoretical framework in which the notion of preparation and measurement play central
role. Another dierent way to reconstruct quantum mechanics is to assume that quantum
uctuations is objectively real thus should be properly modeled by a stochastic processes.
A lot of eorts have been made within this realist theoretical framework to derive the
Schr odinger equation from a stochastic processes [4962]. The greatest challenge of such an
approach is how to explain the nonlocal correlation predicted by quantum mechanics.
In the present paper, we shall follow the above second point of view. We shall rst
propose a statistical model of stochastic deviation from classical mechanics in microscopic
regime based on a stochastic uctuations of innitesimal stationary action. We shall then
show that the abstract and strange [63] rules of quantization of classical systems can
be derived from a specic law of innitesimal change of the information content or the
degree of uncertainty that the system has a certain conguration along an innitesimally
short path, induced by the stochastic uctuations of the innitesimal stationary action.
This law for the dynamics of information is shown to be singled out uniquely, up to a free
parameter, by imposing the condition of Macroscopic Classicality and the principle of Local-
Causality. Note that here, as will be detailed later, information is used to quantify an actual
degree of of uncertainty referring directly to an event regardless of measurement. It is then
imperative to ask: how to explain the violation of Bell inequality in experiments? Putting
the problem aside, we will show that the local-causal statistical model thus developed leads
to the derivation the linear Schr odinger equation with Borns statistical interpretation of
wave function and quantum mechanical uncertainty relation, two of the cornerstones of
standard quantum mechanics.
We shall thus argue that quantization is physical and Planck constant acquires physical
interpretation as a statistical average of the stochastic deviation from classical mechanics in
a microscopic time scale. Two concrete examples of the application of the statistical model
will be given. In the main text, we shall apply the model to quantize a system of particles
subjected to external potentials and in the appendix we shall apply the model to quantize
a classical mechanical model of measurement of physical quantities. Both reproduce the
3
prediction of the standard canonical quantization. Unlike the latter, however, the system
always has a denite conguration all the time as in classical mechanics, uctuating ran-
domly with time. We shall also show that the average of the relevant physical quantities
over the distribution of the conguration is equal to the quantum mechanical average of
quantum mechanical Hermitian operators corresponding to the physical quantities over a
wave function.
II. A STATISTICAL MODEL OF MICROSCOPIC RANDOMNESS AND THE
DYNAMICS OF UNCERTAINTY OBEYING THE PRINCIPLE OF LOCALITY
A. A class of statistical models of microscopic stochastic deviation from classical
mechanics
There is a wealth of empirical evidences that phenomena in microscopic regime involve a
universal stochastic element. Yet, unlike the Brownian motion, hitherto there is no consensus
on the nature and origin of its randomness. Moreover, the prediction of quantum mechanics
on the AB (Aharonov-Bohm) eect [64] and its experimental verication [65] suggest that
the randomness in microscopic regime can not be adequately described by introducing some
kinds of conventional random forces as in Brownian motion. The force has to act at a
distance.
To discuss the universal randomness in microscopic regime, let us consider the following
class of statistical models. Let q denotes the conguration of the system and t is time param-
eterizing the evolution of the system. Let us assume that the Lagrangian is parameterized
by a random variable uctuating in a microscopic time scale dt, whose origin is not our
present concern: L = L(q, q; ), where q
.
= dq/dt. Let us then consider two innitesimally
close spacetime points (q; t) and (q + dq; t + dt) such that is constant. Let us assume
that xing , the principle of stationary action is valid to select a path, denoted by (),
that connects the two points. One must then solve a variational problem (Ldt) = 0 with
xed end points. This leads to the existence of a function, the Hamiltons principal function
denoted by A(q; t, ), whose dierential along the segment of trajectory is given by [66], for
a xed ,
dA = Ldt = p dq Hdt, (1)
4
where p( q) = L/ q is the classical momentum and H(q, p)
.
= p q(p) L(q, q(p)) is the
classical Hamiltonian. The above relation implies the following Hamilton-Jacobi equation:
p =
q
A,
H(q, p) =
t
A. (2)
Varying , the principle of stationary action will therefore pick up various dierent paths
(), all connecting the same two innitesimally close spacetime points, each might have
dierent values of innitesimal stationary action dA(). dA() thus is randomly uctuating
due to the random uctuations of . The system starting with a conguration q at time t
may therefore take various dierent paths randomly to end up with a conguration q + dq
at time t + dt. We have thus a stochastic processes driven by the random uctuations of
in a microscopic time scale. Hence a complete description of a single event is impossible.
Instead, one has to rely on a statistical approach.
One can see that the randomness enters into the dynamics in a microscopic time scale
in a fundamentally dierent way from that of the Brownian motion. In the model, it is
the innitesimal stationary action that is randomly uctuating in a microscopic time scale.
By contrast, the randomness in the Brownian motion is induced by some random forces.
We have thus assumed that the Lagrangian schema based on energies is more fundamental
than the Newtonian schema based on forces. We expect that this will lead to a local-causal
explanation of the AB eect. To see another implication of such a dierence, let us consider
a compound composed of two interacting subsystems. Within the formalism of Brownian
motion, it is then possible to introduce a joint-probability for two random forces each act-
ing locally to a subsystem. By contrast, since action is evaluated in conguration space
rather than in ordinary space, then in the statistical model based on a random uctua-
tions of innitesimal stationary action, one can not dene a joint-probability density for the
uctuations of innitesimal stationary action of each subsystem.
We have thus a class of stochastic models which dier fundamentally from the conven-
tional Brownian motion. In the following subsections, we shall select one of them by imposing
the constraints that the statistical model has a smooth classical limit in macroscopic regime
and respects the principle of Locality demanded by the theory of relativity.
5
B. The dynamics of information or uncertainty with a smooth Macroscopic
Classicality
To develop a statistical description of the stochastic processes, let us denote the joint-
probability density that at time t the conguration of the system is q and a random value of
is realized as (q, ; t). We would like to nd an equation which describes how changes
along an innitesimally short trajectory (). To do this, instead of working directly with
, we shall below consider a quantity dened as
I(q; t, )
.
= ln (q, ; t). (3)
This quantity is introduced by Shannon as a measure of information content or the degree
of uncertainty of an event. Within the context of the stochastic model under study, xing
, it is the information content or the degree of uncertainty that the conguration of the
system is q for the following intuitive reasons: i) it is vanishing if the system denitely has
a conguration q so that (q) = 1; ii) it is increasing monotonically as the probability that
the system has a conguration q is decreasing and iii) it is additive for independent events.
Let us rst note that the information content or the degree of uncertainty dened above
is objective referring directly to the conguration, thus the factual state, of the system. It
is not the information that one obtains by performing some measurements over the system
of interest. Hence, we shall in the present paper work with information within a realist
rather than instrumentalist-operational theoretical model. The latter approach is however
followed by most works in the reconstruction of quantum mechanics based on information
theory, which is apparently motivated by the central role of measurement in the formalism
of standard quantum mechanics. Moreover, let us note that the information quantied by
I(q) refers to a single event that the system has a particular conguration q, rather than
the whole possible events of the system distributed according to (q). The information with
regard to the whole possible events is usually quantied by the average of I(q) given by the
Gibbs-Shannon entropy which is central in information theory [67].
The interpretation of I(q) as the amount of information or degree of uncertainty that
the system has a conguration q may also be argued within the concept of microcanonical
ensemble as follows. First, given the parameters of the system, let N denotes the total
number of the microstates accessible by the system. Let us assume that the system may be in
6
one of the microstates equally probably. Let us then assume that q is a macroscopic coarse-
grained variable of the microstates. Let N
q
denotes the number of microstates compatible
with q. The probability that the system has a conguration q is thus given by = N
q
/N.
One therefore has ln N = I(q) + ln N
q
. Interpreting ln N as the amount of information or
uncertainty that the system lies in one of the N possible microstates, and ln N
q
as the amount
of information or uncertainty that the system lies in one of the N
q
microstates compatible
with q, then it is natural to interpret I(q) as the amount of information or uncertainty that
the system has a conguration q.
To avoid confusion, below we shall only use the term uncertainty to refer to information
content or degree of uncertainty of an event. Let us proceed to again consider two innites-
imally close spacetime points (q; t) and (q + dq; t + dt) such that is constant, connected
by an innitesimally short path (). Let us then assume that as the conguration evolves
along (), the uncertainty that the system has a conguration q also changes according to
the following balance equation:
dI(q; t, ) = d ln (q, ; t) = (q; t, )dt, (4)
where is a function of q, and t. Our main goal in the present section is then to nd a
unique functional form of and express it in terms of the physical properties of the system,
by imposing a set of conceptually simple and physically transparent axioms.
First, it is instructive to impose the condition of Macroscopic Classicality which demands
that in a physical regime corresponding to macroscopic world, one should regain the clas-
sical mechanics. Since the deviation from classical mechanics, as assumed in the previous
subsection, is due to the uctuations of innitesimal stationary action induced by the uc-
tuations of , then in the classical limit of macroscopic regime, such uctuations must be
ignorable. In the macroscopic regime, one must therefore regain the dynamics of ensemble
of classical trajectories driven by the deterministic ow of classical velocity eld. The in-
nitesimal change of the uncertainty must in this case solely be given by the ux due to the
deterministic classical velocity eld. Notice then that the uncertainty should increase if the
velocity divergence along the innitesimally short trajectory () is positive and vice versa.
On the other hand, since the system under consideration is closed, then probability has to
be conserved. These two conditions combined suggest that in the macroscopic regime whose
mathematical formulation will be claried later, on the right hand side of Eq. (4) must
7
reduce to

q
v
c
=
c
, (5)
where v
c
is the classical velocity eld which is related to the classical Hamiltonian and
the Hamiltons principal function via the kinematic part of the Hamilton equation and the
Hamilton-Jacobi equation of (2) as
v
c
=
H
p

p=qA
. (6)
Indeed, inserting Eq. (5) into Eq. (4), dividing both sides by dt and taking the limit dt 0,
one obtains the continuity equation

t
+
q

_
v
c
_
= 0, (7)
which guarantees the conservation of probability.
Hence the demand of Macroscopic Classicality suggests that the right hand side of Eq.
(4) should be given by the following terms:
dI(q; t, ) = (q; t, )dt + (q; t, ),
with
.
=
q
v, (8)
where v is a velocity eld which in the classical limit of macroscopic regime must approach
v
c
v v
c
, (9)
and is a function of q, and t which must be vanishing in the classical limit
(q; t, ) 0. (10)
(q; t, ) may thus be regarded as the rate of production of uncertainty along the random
path () due to the uctuations of in microscopic regime.
From the discussion above, especially the demand that must be vanishing in the classical
limit, it is then natural to assume that is a function of a quantity that measures the
deviation from classical mechanics in microscopic regime due to the uctuations of . To
identify such a quantity, let us rst assume that is the simplest random variable with two
possible values, a binary random variable. Without losing generality let us assume that the
two possible values of dier from each other only by their signs, namely one is the opposite
8
of the other, = [[. Suppose that both realizations of lead to the same path so that
dA() = dA(). Since the stationary action principle is valid for both values of , then
such a model must recover the classical mechanics. Hence, the non-classical behavior must
be measured by the dierence of dA() at [[, dA() dA().
Now let us proceed to assume that may take continuous values. Let us assume that
even in this case the dierence of the values of dA at ,
Z(q; t, )
.
= dA(q; t, ) dA(q; t, ) = Z(q; t, ), (11)
measures the non-classical behavior of the stochastic processes, namely the larger the dif-
ference, the stronger is the deviation from classical mechanics. (q; t, ) should therefore be
a function of Z(q; t, ):
=
_
Z(q; t, )
_
. (12)
For later purpose, let us introduce a new stochastic quantity S(q; t, ) so that the dier-
ential along the segment of path () is given by
dS(q; t, ) =
dA(q; t, ) + dA(q; t, )
2
= dS(q; t, ). (13)
Subtracting dA(q; t, ) from both sides, one gets
dS(q; t, ) dA(q; t, ) =
dA(q; t, ) dA(q; t, )
2
= Z(q; t, )/2. (14)
of Eq. (12) may thus be written as a function of dS() dA()
=
_
dS() dA()
_
. (15)
Note that the assumed universality of the law of physics demands that the functional form
of must be independent from the details of the system of interest: the number of particles,
masses, etc. It must only depend on dS dA.
Let us then express the condition of macroscopic classicality of Eqs. (9) and (10) in term
of S dened above. Let us rst assume that the sign of is uctuating randomly in a time
scale dt. Let us then denote the time scale for the uctuations of [[ as

, and assume that

it is much larger than dt:

dt. (16)
9
Within a time interval of length

, the magnitude of is thus eectively constant while its

sign uctuates randomly. In order for the stochastic system to have a smooth classical limit
for all time, then it is necessary that the classical mechanics is recovered in a time interval
of length

during which the magnitude of is eectively constant while its sign uctuates
randomly. As discussed above, for this binary random variable, the classicality is regained
when dA() = dA(). In this case, one also has dS() = dA() by the virtue of Eq. (14),
so that due to Eq. (1), S satises the Hamilton-Jacobi equation of (2). Taking into account
this fact, rst, the condition of Macroscopic Classicality of Eq. (10) should be rewritten as
lim
dSdA
(dS dA) = 0. (17)
Moreover, the condition of Macroscopic Classicality of Eq. (9) is attained by assuming that
v in Eq. (8) is related to S as follows
v =
H
p

p=qS
. (18)
One can see that in the limit dS dA one has
q
S
q
A so that v v
c
as expected. Let
us emphasize that the above condition is sucient to recover the classical mechanics only
within the time interval of length

in which [[ is constant. While it is also a necessary

condition to recover the classical dynamics for the whole time, it is not sucient. One needs
to have more conditions to recover classical mechanics for the whole time. This problem will
be discussed later.
C. An innitesimal change of uncertainty respecting the principle of Locality
Let us then proceed to show that imposing the principle of Locality will pick up uniquely,
up to a free parameter, the functional form of (dS dA). To do this, let us consider a
compound of two subsystems, say two particles whose conguration are denoted respectively
by q
1
and q
2
. Let us assume that they are spacelike separated from each other so that due
to the principle of Locality, there is no mechanical interaction between the two particles.
The total Lagrangian is thus decomposable as L(q
1
, q
2
, q
1
, q
2
) = L
1
(q
1
, q
1
) + L
2
(q
2
, q
2
) and
accordingly, dA(q
1
, q
2
) and dS(q
1
, q
2
) are also decomposable: dA(q
1
, q
2
) = dA
1
(q
1
) +dA
2
(q
2
)
and dS(q
1
, q
2
) = dS
1
(q
1
) + dS
2
(q
2
). can thus be written as
(dS dA) =
_
(dS
1
dA
1
) + (dS
2
dA
2
)
_
. (19)
10
Further, in this case, the classical Hamiltonian is also decomposable H(q
1
, q
2
, p
1
, p
2
) =
H
1
(q
1
, p
1
) + H
2
(q
2
, p
2
), where p
i
, i = 1, 2, is the momentum of iparticle. Putting this
into Eq. (18) and recalling that dS is decomposable, then dened in Eq. (8) is also
decomposable
(q
1
, q
2
) =
1
(q
1
) +
2
(q
2
). (20)
The change of the uncertainty that the compound system has a conguration q = (q
1
, q
2
)
moving along the path () thus reads, by the virtue of Eq. (19) and (20),
dI(q
1
, q
2
) =
_

1
+
2
_
dt +
_
(dS
1
dA
1
) + (dS
2
dA
2
)
_
. (21)
On the other hand, since the two subsystems are spacelike separated from each other,
the principle of Locality demands that the change of the uncertainty that the rst (second)
subsystem has a conguration q
1
(q
2
), when the compound system moves along an innitesi-
mally short trajectory (), must be independent from what happens with the second (rst)
subsystem. Otherwise, the uncertainty that one subsystem has a certain conguration can
be inuenced by the state of the other distantly separated subsystem by varying the control
parameters of the latter despite of no interaction. Hence, the change of the uncertainty that
each subsystem has a certain conguration must only depend on the corresponding single
particle Lagrangian. One therefore has the following pair of decoupled relations:
dI
1
(q
1
) =
1
dt + (dS
1
dA
1
),
dI
2
(q
2
) =
2
dt + (dS
2
dA
2
), (22)
where dI
i
= d(ln
i
), and
i
(q
i
), i = 1, 2, is the probability density for the conguration
of the iparticle. dI
i
is thus the change of the uncertainty that the isubsystem has a
conguration q
i
. Let us note again that the assumed universality of the law of physics
demands that the functional form of for the whole compound system on the right hand
side of Eq. (21) must be the same as those for each subsystem on the right hand side of Eq.
(22).
Let us rst assume that the probability distribution of the conguration of the com-
pound system is separable: (q
1
, q
2
) =
1
(q
1
)
2
(q
2
). In this case, the total change of the
uncertainty that the compound system has a conguration q = (q
1
, q
2
) as the congura-
tion evolves along () is then decomposable as dI(q
1
, q
2
) = dI
1
(q
1
) + dI
2
(q
2
). Now let us
11
consider a general case when the distribution of the conguration of the two spacelike sepa-
rated subsystems are correlated. One thus has (q
1
, q
2
) =
12
(q
1
[q
2
)
2
(q
2
), where
12
(q
1
[q
2
)
is the conditional probability that the conguration of the rst subsystem is q
1
when the
conguration of the second subsystem is q
2
. As the conguration of the compound system
evolves along an innitesimally short path (), the total change of the uncertainty that
the compound system has a conguration q = (q
1
, q
2
) is then
dI(q
1
, q
2
) = dI
12
(q
1
[q
2
) + dI
2
(q
2
), (23)
where dI
12
= d ln
12
is the innitesimal change of the uncertainty that the conguration
of the rst subsystem is q
1
when the conguration of the second subsystem is q
2
. The
principle of Locality however demands that, since the two subsystems are spacelike separated
from each other, the innitesimal change of the uncertainty that the rst subsystem has a
conguration q
1
must be independent from the conguration of the second subsystem q
2
.
One must thus have dI
12
(q
1
[q
2
) = dI
1
(q
1
). Inserting into Eq. (23), one therefore concludes
that in general the total innitesimal change of the uncertainty that the two non-interacting
subsystems have a conguration q = (q
1
, q
2
) is decomposable as
dI(q
1
, q
2
) = dI
1
(q
1
) + dI
2
(q
2
). (24)
Finally inserting Eqs. (21) and (22) into Eq. (24), in general (dSdA) must also satisfy
the following decomposability condition:

_
(dS
1
dA
1
) + (dS
2
dA
2
)
_
= (dS
1
dA
1
) + (dS
2
dA
2
). (25)
The above functional equation together with the necessary condition for Macroscopic Clas-
sicality of Eq. (17) can then be solved to give the following linear solution:
(dS dA) = (; t)(dS dA), (26)
where is a real-valued function independent from dS dA, yet might depend on t and ,
hence is randomly uctuating with in a microscopic time scale. Let us emphasize that Eq.
(26) now applies for general cases, not only for a compound of non-interacting subsystems.
For the reason that will be clear later, let us introduce a new non-vanishing random vari-
able (; t) = 2/. The change of the uncertainty that the system has a certain conguration
along an innitesimally short path () of Eq. (4) is thus given by
dI = d ln = dt +
2

(dS dA). (27)

12
Further, since is xed during the time interval dt, one can expand all the dierentials as
dF =
t
Fdt +
q
F dq to have the following pair of coupled partial dierential equations:

q
ln =
2

q
S p( q)
_
,

t
ln =
2

_
H(q, p) +
t
S
_
+ (S), (28)
where we have made use of Eq. (1). The spatial and temporal changes of the uncertainty
that the system has a certain conguration are thus related to the momentum and energy
of the system, respectively. Let us emphasize that the above pair of equations are valid only
for a time interval in which is constant.
For later convenient let us write the pair of coupled equations of Eq. (28) as follows:
p( q) =
q
S +

2

,
H(q, p) =
t
S +

2

+

2
(S), (29)
where the momentum and energy are put on the left hand side. The above pair of relations
must not be interpreted that the momentum or velocity and energy of the system are deter-
mined causally by the change of the uncertainty, which is physically absurd. Rather both the
momentum and energy provide the source of change of the uncertainty that the system has
a certain conguration along an innitesimally short trajectory () as shown explicitly by
Eq. (27). Further, it is evident that in the formal limit 0 whose physical meaning will
be claried in the next subsection, Eq. (29) reduces back to the Hamilton-Jacobi equation
of (2). In this sense, Eq. (29) can be regarded as a generalization of the Hamilton-Jacobi
equation. Unlike the Hamilton-Jacobi equation in which we have a single unknown function
A, however, to calculate the velocity or momentum and energy, one now needs a pair of
unknown functions S and .
D. A stochastic processes with a transition probability given by an exponential
distribution of deviation from innitesimal stationary action
We have started from a stochastic processes in which the system with a conguration
q at time t can take one of many possible random paths () selected by the principle of
stationary action with dierent random values of , to end up with a conguration q +dq at
13
time t+dt. We then derived a law of innitesimal change of the uncertainty of an event along
an innitesimally short path by imposing the condition of Macroscopic Classicality and the
principle of Locality. It is then tempting to investigate if the law of change of uncertainty
given by Eq. (27) completely determines the stochastic processes. To see this, xing , let

_
(q + dq; t + dt), (q; t)

()
_
denotes the conditional joint-probability density that the
conguration of the system is q at time t, tracing the trajectory () and end up with a
conguration q + dq at time d + dt. Using this quantity, the change of probability density
d due to the transport along the path () is given by
d(q, ; t) =
_
(q + dq; t + dt), (q; t)

()
_
(q, ; t). (30)
Inserting into Eq. (27) one therefore has

_
(q + dq; t + dt), (q; t)

()
_
=
_
1 (S)dt
2

(dS dA)

(q, ; t). (31)

Let us then consider the case when [(dSdA)/[ 1. Equation (31) can then be written
approximately as

_
(q + dq; t + dt), (q; t)

()
_
e

(dS()dA())(S)dt
(q, ; t). (32)
The above relation can obviously be read within the conventional probability theory as
follows: the joint-probability density that the system initially at (q; t) traces the segment
of trajectory () and end up at (q + dq; t + dt),
_
(q + dq; t + dt), (q; t)

()
_
, is equal
to the probability that the conguration of the system is q at time t, (q, ; t), multiplied
by a transition probability between the two innitesimally close spacetime points via the
segment of trajectory () given by
P((q + dq; t + dt)[(), (q; t)) e

(dS()dA())
/:, (33)
where : = exp((S)dt).
Some notes are instructive. First, to guarantee the normalizability of the above transition
probability, then the exponent (dS()dA())/() must be non-negative for any spacetime
point (q, t). This demands that dS() dA() must always have the same sign as (). On
the other hand, from Eq. (14), one can see that dS() dA() changes its sign as ips its
14
sign. Hence, to guarantee the non-negativity of (dS() dA())/(), must also change
its sign as ips its sign. This allows us to assume that the sign of is always the same
as that of . The time scale for the uctuations of the sign of must therefore be the same
as the time scale for the uctuations of given by dt. Moreover, the second term on the
right hand side of Eq. (27), , namely the rate of production of uncertainty due to the
uctuations of innitesimal stationary action is always non-negative.
It is also evident that for the distribution of Eq. (33) to make sense mathematically, the
time scale for the uctuations of [[, denoted by

, must be much larger than that of [[.

One thus has

dt. (34)
In other words, [[ uctuates much faster than [[, yet both and always have the same
sign uctuating randomly in the time scale dt. Hence, within a time interval of length

during which [[ is eectively constant, one may assume that dS() dA() is randomly
uctuating due to the uctuations of [[ distributed according to the exponential law of Eq.
(33) characterized by [[.
Next, there is no a priori reason on how the sign of the values of dS dA should be
distributed. Following the principle of indierence (principle of insucient reason) [68], it is
then advisable to assume that the sign of dS dA is distributed equally probably. Further,
since the sign of dS() dA() changes as ips its sign, then the sign of must also be
uctuating randomly with equal probability so that the probability density of the value of
at any given time, denoted below by P
H
(), must satisfy the following unbiased condition:
P
H
() = P
H
(). (35)
Since the sign of is always the same as that of then the probability distribution function
of must also satisfy the same unbiased condition. Further, since P
H
() =
_
dq(q, ; t),
then Eq. (35) demands the following symmetry relation:
(q, ; t) = (q, ; t). (36)
One also has, from Eq. (13), the following symmetry relations for the spatiotemporal gra-
dient of S(q, ; t):

q
S(q; t, ) =
q
S(q; t, ),

t
S(q; t, ) =
t
S(q; t, ), (37)
15
which together with Eq. (36) will play important role later.
Recall that the pair of relations in Eqs. (28) or (29) are valid when is xed. However,
since as discussed above, the second term on the right hand side of Eq. (27) is insensitive
to the sign of which is always the same as the sign of , then the pair of equations in
(29) are valid in a microscopic time interval of length

during which the magnitude of

is constant while its sign may change randomly. To have an evolution for a nite time
interval t >

, one can proceed along the following approximation. First one divides the
time into a series of microscopic time intervals of length

: t [(k 1)

, k

), k = 1, 2, . . . ,
and attributes to each interval a random value of (t) =
k
according to the probability
distribution P
H
k
(
k
) = P
H
k
(
k
). Hence, during the interval [(k 1)

, k

), the magnitude
of (t) =
k
is kept constant while its sign changes randomly in an innitesimal time scale
dt. One then applies the pair of relations in Eqs. (28) or (29) during each interval of time
with xed [(t)[ = [
k
[, consecutively.
Since dA is just the innitesimal stationary action along the short path (), [dS dA[
may be regarded as the deviation from innitesimal stationary action, the distribution of
which is given by Eq. (33). Such an exponential distribution was rstly suggested heuristi-
cally in Ref. [69] to model a microscopic stochastic deviation from classical mechanics. An
application of the statistical model to model quantum measurement is given recently in Ref.
[70]. For a xed value of [[ which is valid during a time interval of length

, one can see

from Eq. (33) that the average deviation from innitesimal stationary action is given by
[dS dA[ = [[/2. (38)
It is then evident that in the regime where the average deviation is much smaller than the
innitesimal stationary action itself, namely [dA/[ 1, or formally in the limit [[ 0,
Eq. (33) reduces to
(dS dA), (39)
so that dS() dA(). Such a regime thus must be identied as the macroscopic regime.
This fact suggests that [[ must have a very small microscopic value. In this regard, the pair
of equations in (29) may be regarded as a generalization of the Hamilton-Jacobi equation of
(2) due to the exponential distribution of deviation from innitesimal stationary action of
Eq. (33). Let us also note that since [[ in general may depend on time, then the transition
probability is in general not stationary except when = Q all the time, where Q is a
16
constant. We shall in the next section and appendix consider a stationary case of empirical
interest when Q = .
One can also see that the decomposability of the innitesimal change of the uncertainty
for a pair of spacelike separated subsystems given by Eq. (24), which is demanded by the
principle of Local-Causality, implies directly the separability of the transition probability of
Eq. (33) for the non-interacting subsystems. Namely, for non-interacting two subsystems
such that dA and dS are decomposable as dA(q
1
, q
2
) = dA
1
(q
1
) + dA
2
(q
2
) and dS(q
1
, q
2
) =
dS
1
(q
1
) +dS
2
(q
2
), respectively, so that is also decomposable (q
1
, q
2
) =
1
(q
1
) +
2
(q
2
), the
transition probability of Eq. (33) is separable as
P
S
(dS
1
+ dS[dA
1
+ dA
2
) = P
S
(dS
1
[dA
1
)P
S
(dS
2
[dA
2
). (40)
Hence, the transition probability which determines the stochastic behavior of one subsystem
is independent from that of the other subsystem as intuitively expected for spacelike sepa-
rated non-interacting subsystems. Otherwise, the dynamics and statistics of one subsystem
is inuenced by the other subsystem, which contradicts the principle of Locality. See also
Ref. [71] for a dierent approach to single out Eq. (33) by imposing the principle of Locality
without directly employing the concept of information.
III. QUANTIZATION
A. The Schr odinger equation and Borns statistical interpretation of wave function
Let us apply the above general formalism to stochastically modify a classical system of
a single particle subjected to external potentials so that the classical Hamiltonian takes the
following general form:
H(q, p) =
g
ij
(q)
2
(p
i
a
i
)(p
j
a
j
) + V, (41)
where a
i
(q), i = x, y, z and V (q) are vector and scalar potentials respectively, the metric
g
ij
(q) may depend on the position of the particle, and summation over repeated indices
are assumed. The application to many particles system with dierent kind of classical
Hamiltonians can be done in the same way by following exactly all the steps that we are
going to take below. In the appendix we shall give another example of the application of the
statistical model to a system of two interacting particles which is commonly used to model
17
the measurement of physical quantities, the detail elaboration of which is reported recently
in Ref. [70].
Let us rst consider a time interval of length

during which the magnitude of is

eectively constant while its sign uctuates randomly together with the random uctuations
of the sign of in a time scale dt. Let us then divide it into a series of microscopic time
intervals of length

, [(k 1)

, k

), k = 1, 2, . . . and attribute to each interval a random

value of (t) =
k
according to the probability distribution P
H
k
(
k
) = P
H
k
(
k
). Hence, in
each interval, the magnitude of (t) =
k
is constant while its sign changes randomly and
the pair of equations in (29) with xed [
k
[ apply.
Let us rst consider a microscopic time interval [(k 1)

, k

). Within this interval of

time, using Eq. (41) to express q in term of p via the (kinematic part of the) usual Hamilton
equation q = H/p, one has, by the virtue of the upper equation of (29)
q
i
() = g
ij
_

q
j
S() +

2

q
j
()
()
a
j
_
. (42)
Assuming the conservation of probability one thus obtains the following continuity equation:
0 =
t
+
q
( q)
=
t
+
q
i
_
g
ij
(
q
j
S a
j
)
_
+

2

q
i
(g
ij

q
j
). (43)
On the other hand, from Eq. (41), (S) dened in Eqs. (8) and (18) is given by
(S) =
q
i
g
ij
(
q
j
S a
j
). (44)
Using the above form of (S), the lower equation of (29) thus becomes
H(q, p) =
t
S +

2

+

2

q
i
g
ij
(
q
j
S a
j
). (45)
Plugging the upper equation of (29) into the left hand side of Eq. (45) and using Eq. (41)
one has, after arrangement

t
S +
g
ij
2
(
q
i
S a
i
)(
q
j
S a
j
) + V

2
2
_
g
ij

q
i

q
j
R
R
+
q
i
g
ij

q
j
R
R
_
+

2
_

t
+
q
i
_
g
ij
(
q
j
S a
j
)
_
+

2

q
i
(g
ij

q
j
)
_
= 0, (46)
where we have dened R
.
=

and used the identity:

1
4

q
i

q
j

=
1
2

q
i

q
j

q
i

q
j
R
R
. (47)
18
Inserting Eq. (43), the last line of Eq. (46) vanishes to give

t
S +
g
ij
2
(
q
i
S a
i
)(
q
j
S a
j
) + V

2
2
_
g
ij

q
i

q
j
R
R
+
q
i
g
ij

q
j
R
R
_
= 0. (48)
We have thus a pair of coupled equations (43) and (48) which are parameterized by . Recall
that the above pair of equations are valid in a microscopic time interval of length

during
which the magnitude of is constant while its sign changes randomly with equal probability.
Moreover, recall also that the sign of is always the same as the sign of . Keeping this in
mind, averaging Eq. (43) for the cases , thus is also over , one has, by the virtue of
Eqs. (36) and (37),

t
+
q
i
_
g
ij
(
q
j
S a
j
)
_
= 0. (49)
Similarly, averaging Eq. (48) over the cases will not change anything. We have thus
nally a pair of coupled equations (48) and (49) which are now parameterized by a constant
[[, valid during a microscopic time interval of length

characterized by a constant [[.

Next, since [[ is non-vanishing, one can dene the following complex-valued function:

.
=

exp
_
i
S
[[
_
. (50)
Using , recalling the assumption that [[ is constant during the time interval of interest
of length

, the pair of Eqs. (48) and (49) can then be recast into the following modied
Schr odinger equation:
i[[
t
=
1
2
(i[[
q
i
a
i
)g
ij
(q)(i[[
q
j
a
j
) + V . (51)
Notice that the above equation is valid only for a microscopic time interval [(n 1)

, n

)
during which the magnitude of =
n
is constant. For nite time interval t >

, one must
then apply Eq. (51) consecutively to each time intervals of length

with dierent random

values of [
n
[, n = 1, 2, 3, . . . .
Let us then consider a specic case when [[ is given by the reduced Planck constant
for all the time, namely = , so that the exponential distribution of the deviation from
innitesimal stationary action of Eq. (33) is stationary and the average deviation is given
by
/2. (52)
19
Let us further assume that the uctuations of [[ around its average is suciently narrow.
In this case, one may approximate (q, [[; t) and S(q; t, [
n
[) by the corresponding zeroth
order terms of their Taylor expansion around the average of [[, respectively denoted by

Q
(q; t) and S
Q
(q; t). The zeroth order approximation of Eq. (51) therefore reads
i
t

Q
(q; t) =

H
Q
(q; t),

Q
(q; t)
.
=
_

Q
(q; t)e
i

S
Q
(q;t)
, (53)
where

H is just the quantum Hamiltonian given by

H =
1
2
( p
i
a
i
)g
ij
(q)( p
j
a
j
) + V, (54)
with p
i
.
= i
q
i
is the quantum mechanical Hermitian momentum operator. Unlike Eq.
(51), Eq. (53) is now deterministic and is parameterized by . Moreover, from Eq. (53), one
can see that the Borns statistical interpretation of wave function is valid by construction

Q
(q; t) = [
Q
(q; t)[
2
. (55)
Hence, for the specic type of classical Hamiltonian with a quadratic momentum given
in Eq. (41), one regains the results of canonical quantization. The rules of canonical
quantization thus eectively arise from a statistical modication of classical mechanics in
microscopic regime based on a specic law of change of the uncertainty that the system has a
certain conguration of Eq. (27) chosen uniquely by imposing the condition of Macroscopic
Classicality and the principle of Local-Causality. Unlike the canonical quantization which is
formal-mathematical with obscure physical meaning, the statistical model of quantization
is thus physical. As shown above, Planck constant acquires a physical interpretation as
the average deviation from classical mechanics in a microscopic time scale. Moreover, unlike
canonical quantization which in general leads to an innite number of possible quantum
Hamiltonians if g
ij
in Eq. (41) depends on q due to operators ordering ambiguity, the
statistical model gives a unique quantum Hamiltonian with a specic ordering of operators
of Eq. (54) where g
ij
(q) is sandwiched by p a = i
q
a.
B. Conguration as beable and quantum mechanical average
Recall that in standard quantum mechanics, the state of the system is assumed to be de-
termined completely by specifying the wave function. The wave function is thus regarded as
20
fundamental. Statements about position and momentum are then relegated operationally to
certain acts of measurement over the state of the system represented by the wave function.
The canonical uncertainty relation between the statistical results of position and momentum
measurement is usually mentioned to support the argumentation that it is in general impos-
sible to attribute a pair of denite values of position and momentum to a system, nor such an
attribution is useful. By contrast, in the statistical model of quantization developed in the
present paper, one assumes the objective ontology of particles with a denite conguration
for all the time as in classical mechanics. Hence, conguration of the system is regarded as
the beable of the theory in Bells sense [72]. The wave function, on the other hand, is
argued to be emergent articial convenient mathematical tool to describe the dynamics and
statistics of ensemble of trajectories. The objective ontology of the trajectories guarantees
a conceptually smooth classical correspondence.
For general types of classical Hamiltonian, the velocities of the particles are then obtained
by rst solving p( q) on the left hand side of the upper equation of (29) in term of q to have
q() =
H
p

_
p=qS()+
()
2
q()
()
_
. (56)
Hence, the velocities of the particles are uctuating randomly due to the uctuations of .
The conguration of the system therefore follows a continuous and randomly uctuating
trajectory. Since, as argued in the previous section, in the limit [[ 0 which is interpreted
physically to correspond to classical regime, one has dS dA, then one can see that in this
limit, Eq. (56) reduces to the classical relation q =
H
p
[
p=qA
. We have thus a formally and
conceptually smooth classical correspondence.
Note again that xing [[, Eq. (56) is valid only within a time interval of length

during
which the sign of is uctuating randomly. It is then natural to dene an eective velocity
as

q([[)
.
=
q() + q()
2
. (57)
For the type of classical Hamiltonian given by Eq. (41), Eq. (56) reduces to Eq. (42). Let
us now consider the case when the classical Hamiltonian is given by Eq. (41) with g
ij
= 1/m
and a
i
= 0, describing a particle of mass m subjected to an external scalar potential V (q).
Let us again consider a specic case when = so that the average of the deviation from
innitesimal stationary action distributed according to Eq. (33) is given by /2, and as
shown in the previous subsection, one regains the prediction of canonical quantization. In
21
this case, the zeroth order approximation of Eq. (42) then reads
q =

q
S
Q
m


2m

Q
, (58)
where the signs change randomly with equal probability. Moreover, calculating the
eective velocity one gets, recalling that the sign of is the same as that of , and taking
into account Eqs. (36) and (37),

q =
q
S
Q
/m, (59)
which, unlike Eqs. (42) and (58), is now deterministic due to the deterministic time evolution
of S
Q
given by the Schr odinger equation of (53). Hence, q in general uctuates randomly
around
q
S
Q
/m except when the particle happens to lie at the extremum points of
Q
so
that the second term of Eq. (58) vanishes and one has q =
q
S
Q
/m.
One thus expects that the actual trajectory of the particle is in general uctuating ran-
domly around the integration over time of the eective velocity

q =
q
S
Q
/m. The latter is
just the Bohmian trajectory of the particle in pilot-wave theory [73]. Hence, we have a physi-
cal picture that the actual trajectory is uctuating randomly around the Bohmian trajectory
while the latter moves as if it is guided by the wave function evolving deterministically ac-
cording to the Schr odinger equation. Yet, unlike the pilot-wave theory, the wave function
in the statistical model is not physically real but an articial mathematical construct, and
the Schr odinger equation and the guidance relation of Eqs. (53) and (59) are derived from
rst principle rather than ad-hoc-ly postulated as in pilot-wave theory. Recall that the fun-
damental assumption in pilot-wave theory that the wave function is a physical eld, living
in conguration space rather than in ordinary space, is known to lead to a conceptual di-
culty, and furthermore implies rigid nonlocality in direct conict with the special theory of
relativity. By contrast, the present statistical model is developed based on the principle of
Locality. In this sense, the upper equation in (29) can not be regarded as a causal-dynamical
guidance relation as in pilot-wave theory, but a kinematical relation. Moreover, unlike the
pilot-wave theory which is deterministic and relegates the microscopic randomness to our
ignorance of the initial condition, the statistical model is strictly stochastic.
It is then imperative to calculate the statistical averages of the relevant physical quantities
over all possible conguration distributed according to (q, ; t). It is natural to ask how
such statistical averages are related to quantum mechanical averages. For this purpose, we
22
should again assume that = with equal probability in case of which the statistical
model reproduces the results of canonical quantization as the zeroth order approximation.
First, the average of any function of the conguration O(q) at any time is given by
O(q)
.
=
_
dqdO(q) =
_
dqd

O(q)

_
dq

Q
O(q)
Q
.
=
Q
[O( q)[
Q
, (60)
where we have again counted only the zeroth order terms. Numerically, it is thus equal to
the quantum mechanical average of a quantum mechanical observable O( q) = O(q) when
the state of the system is given by the wave function
Q
. In particular, for the cases
O(q) = q and O(q) = (q q)
2
, the left hand side of Eq. (60) are the average and standard
deviation of the uctuations of q, which are numerically equal to the quantum mechanical
average of position operator and its standard deviation over the state
Q
given by the right
hand side.
Let us further calculate the average of the actual value of momentum p at a given time.
One directly gets, from the upper equation in (29)
p =
_
dqd
_

q
S +

2

=
_
dqd(
q
S) =
_
dqd

(i[[
q
)

_
dq

Q
(i
q
)
Q
.
=
Q
[ p[
Q
, (61)
where in the second equality we have used Eqs. (36) and (37) and taken into account the
fact that the sign of is the same as that of , and in the last approximate equality we have
imposed = and counted only the zeroth order terms. Again, it is numerically given by
the quantum mechanical average of momentum operator p = i
q
over the state
Q
.
Let us proceed to calculate the average of a quantity which is a function of momentum
up to second degree. Such a quantity can always be put into a quadratic form. It is thus
sucient to calculate the average of a quantity of the type O(q, p) = a(p f(q))
2
, where a
is constant and f(q) is a vector-valued function. One directly gets
O(q, p) =
_
a
_

q
S +

2

f(q)
_
2
_
=
_
a
_

2
4

_
2
_
+a(
q
S f(q))
2

=
_
dqd

a
_
i[[
q
f(q)
_
2

Q
[a
_
p f(q)
_
2
[
Q
. (62)
23
Here, in the second equality we have again used Eqs. (36) and (37), and the fact that the
sign of is always the same as that of , in the third equality we have used the identity of
(47), and nally we imposed = and counted only the zeroth order terms. One can see
that it is numerically again given by the quantum mechanical average of the corresponding
Hermitian quantum operator O( q, p) = a( p f( q)
_
2
over the state
Q
.
Two specic cases are of importance. First, let us consider the case when a = 1 and f is
constant given by the average of momentum f = p. Using Eq. (61), one gets
(p p)
2
=
Q
[
_
p
Q
[ p[
Q

_
2
[
Q
. (63)
Hence, the standard deviation of the actual momentum p is also numerically given by the
quantum mechanical standard deviation of p over the state
Q
. Next, let us consider the
case when a = 1/(2m), where m is the mass of a particle. One then has
_
(p f)
2
2m
_
=
_

( p f)
2
2m

Q
_
. (64)
The left hand side is just the average of the actual kinetic energy of a particle subjected to
vector potential f. It is shown to be numerically equal to the quantum mechanical average
of the quantum mechanical Hamiltonian for a particle subjected to the vector potential f
over the state
Q
given by the right hand side.
Combining Eqs. (60) and (64), taking O(q) = V (q) in Eq. (60), one directly gets the
average of the actual energy H = (p f)
2
/(2m) + V (q):
H =
Q
[

H[
Q
, (65)
which is numerically equal to the quantum mechanical average of the corresponding quantum
Hamiltonian over the wave function
Q
.
One can also show easily that the average of the actual angular momentum l = q p
over the distribution of the conguration is numerically equal to the quantum mechanical
average of quantum mechanical angular momentum operator

l
.
= q p = q (i
q
) over
a quantum state represented by a wave function:
l =
Q
[

l[
Q
. (66)
One may thus conclude that for the case of a system of spin-less particles we are con-
sidering, the average of the actual values of all the relevant physical quantities O(q, p( q))
24
over all possible conguration of the system are equal to the quantum mechanical average
of the corresponding quantum observables represented by some Hermitian operators

O over
a wave function
Q
representing the corresponding quantum mechanical state:
O
.
=
_
dqdO(q, p)
Q
[

O[
Q
, (67)
where means that we have considered the case when = and counted only the zeroth
order terms. It is then tempting to guess that the above conclusion applies for all quantities
of a function of position and momentum O(q, p). One can however show that this is not
the case. For example, averaging O = p
3
, regardless of its physical meaning and proceeding
as before, one can check that p
3
, =
Q
[ p
3
[
Q
. Let us further emphasize however that
while
Q
[

O[
Q
in the standard quantum mechanics refers to the average of the outcomes
of measurement over an ensemble of identically prepared state represented by
Q
, O
in the statistical model refers to the objective properties of the ensemble independent of
measurement.
C. The principle of Locality, linearity of quantum dynamics and uncertainty re-
lation
Notice that as argued in the previous section, the unique form of the law of innitesimal
change of the degree of uncertainty that the system has a certain conguration along a short
path given by Eq. (27) is singled out by imposing the condition of Macroscopic Classicality
and the principle of Locality. In particular, the principle of Locality is decisive in selecting
the linear form of the second term on the right hand side of Eq. (27) which describes the
production of information or uncertainty due to the uctuation of innitesimal stationary
action. Since the stochastic processes based on such a change of the information leads to
the derivation of the linear Schr odinger equation, one may thus argue that the principle of
Locality expressed in Eq. (25) is a necessary condition for the linearity of the Schr odinger
equation. To support this argumentation, let us mention that a nonlinear extension of
quantum dynamics [74] may lead to signaling [75, 76] thus violating the principle of Locality.
The importance of the linearity of the Schr odinger equation can not be over emphasized.
It gives the mathematical basis for the superposition principle which plays very crucial
roles for the explanation of particle interference in double slits experiment, quantum entan-
25
glement, and also in quantum measurement. The linearity of the Schr odinger equation is
also responsible for another important characteristic trait of quantum mechanics that it is
impossible to copy an unknown quantum state, the no-cloning theory [7779].
Let us proceed to show that the principle of Locality also plays a very important role in the
derivation of the quantum mechanical uncertainty relation. First, as argued in the previous
subsection, the actual trajectory of the system in conguration space is uctuating randomly
around the Bohmian trajectory. It is then imperative to ask how the uctuations around the
Bohmian trajectory is distributed. First, from the normalization of ,
_
dqd(q, ) = 1,
and the assumption that [
q
= 0 which is valid for arbitrary value of , one has
1 =
_
dqd =
_
dqd(q q
0
)
q

=
_
dqd(q q
0
)

_
, (68)
where q
0
is an arbitrary real number and the integration over spatial coordinate is taken
from q = to q = . Applying the Schwartz inequality one gets
_
dqd(q q
0
)
2

_
dqd
_

_
2
1. (69)
Substituting Eq. (42), considering the case when g
ij
= 1/m and a
i
= 0, one directly obtains
_
dqd(q q
0
)
2

_
dqd(m q
q
S)
2


2
4
. (70)
Let us again consider a specic case when = . The zeroth order approximation of
the above inequality therefore reads [80]
_
dq(q q
0
)
2

Q
(q)
_
dq(m q
q
S
Q
)
2
(q)

2
4
. (71)
One then sees that the width of the uctuations of m q around
q
S
Q
is bounded from below
by the inverse of the width of the distribution of q and vise-versa, in a similar fashion as the
standard quantum mechanical uncertainty relation.
On the other hand, from Eqs. (60) and (63), one has
(q q)
2
(p p)
2

=
Q
[( q q)
2
[
Q

Q
[( p p)
2
[
Q


2
4
, (72)
where the inequality is due to [ q, p] = i. Hence, the width of the distribution of the
actual momentum is bounded from below by the inverse of the width of the distribution of
26
the actual position in the ensemble, in exactly the same manner as the standard quantum
mechanical uncertainty relation. Let us emphasize again that unlike the former which is
objective referring to no measurement, the latter is referring to the statistical results of
measurement of position and momentum over an ensemble of identically prepared system.
The above uncertainty relation is related to the uncertainty relation of Eq. (71) via the
fact that
(p p)
2

__

_
2
_
+(
q
S
Q

q
S
Q
)
2

__

_
2
_
=

(m q
q
S
Q
)
2
_
, (73)
where in the rst approximate equality we have used Eqs. (58) and (61) and taking into
account Eqs. (36) and (37) and the fact that the sign of is always the same as the sign of ,
and the last equality is due again to Eq. (58). Multiplying both sides with (qq
0
)
2
, taking
q
0
= q and imposing Eq. (71), one obtains Eq. (72). Hence, the quantum mechanical
uncertainty relation can be derived starting from the upper equation in Eq. (29). Since,
as argued in the previous section, the pair of equations in (29) are derived by imposing
the principle of Locality, then one may also conclude that the latter is necessary for the
derivation of quantum mechanical uncertainty relation.
IV. CONCLUSION AND REMARKS
We rst developed a stochastic processes for a microscopic stochastic deviation from
classical mechanics in which the randomness is modeled by a stochastic uctuations of the
innitesimal stationary action, thus is physically dierent from that of the Brownian motion
based on random forces. Such a stochastic processes leads to a production of the uncertainty
that the system has a certain conguration in a microscopic time scale, which is assumed
to be vanishing in the classical limit of macroscopic regime. We then showed that imposing
the principle of Locality, which requires the innitesimal change of the uncertainty that a
subsystem has a certain conguration to be independent from the conguration of the other
spacelike separated subsystem, will select a unique law of innitesimal change of uncertainty,
up to a free parameter. We then further showed that such a law of innitesimal change of
uncertainty determines a stochastic processes with a transition probability between two
innitesimally close spacetime points along a randomly chosen path that is given by an
27
exponential distribution of deviation from innitesimal stationary action.
Given a classical Hamiltonian, we showed that the statistical model leads to the deriva-
tion of Schr odinger equation with Borns statistical interpretation of wave function and
uncertainty relation, two of the cornerstones of standard quantum mechanics. Unlike the
canonical quantization, however, in the statistical model, the system always has a denite
conguration all the time as in classical mechanics, uctuating randomly along a continuous
trajectory. We have also shown, for a system of spin-less particles, that the average of the
relevant physical quantities over the distribution of the conguration is numerically equal
to the quantum mechanical average of the corresponding quantum mechanical Hermitian
operators over a quantum state represented by a wave function. Since the principle of Lo-
cality is derived from our conception of spacetime structure, then one may conclude that
the dynamics and kinematics of quantum mechanics is intimately related to the former.
Some problems are left for future study. It is rst imperative to ask how such a local-
causal statistical model would explain the violation of Bells inequalities predicted by the
quantum mechanics and veried in numerous experiments which is widely believed to give
strong evidences that Nature is nonlocal? This is a crucial problem needed an explanation
within the statistical model. Recall that Bells inequalities are derived by assuming 1) the
separability of probability of outcomes in a pair of spacelike joint-measurements (Bells lo-
cality assumption) and 2) the so-called measurement independence or experimental free-will,
that the distribution of the hidden variables underlying the measurement outcomes is inde-
pendent from the setting parameters of the apparatus chosen freely by the observer [8184].
It is tempting to guess that the objective locality of the statistical model implies the Bells
locality assumption so that the model must somehow violate measurement independence. It
is therefore instructive to study the above two fundamental hypothesis within the statistical
model by rst applying the model to develop quantum measurement in realistic physical
systems and derive the Borns rule.
Next, it is also tempting to ask why canonical quantization corresponds to a specic
case when [[ in Eq. (27), the free parameter of the statistical model, is given by so
that the average deviation from innitesimal stationary action distributed according to the
exponential law of Eq. (33) is given by /2. Is the relation [[ = exact? Or whether
Nature allows for a small uctuations of [[ around ? Recall also that the Schr odinger
equation is derived as the zeroth order approximation of the statistical model. It is then
28
imperative to study the higher orders corrections. These last two cases may thus provide
precision tests against quantum mechanics.
Acknowledgments
Appendix A: Quantization of classical mechanical model for the measurement of
angular momentum
To show the robustness of the statistical model, let us give an application of the model to a
system of two interacting particles. Let us assume that the interaction classical Hamiltonian
is given by
H
I
= gl
z
1
p
2
, with l
z
1
= x
1
p
y
1
y
1
p
x
1
. (A1)
Here g is an interaction coupling, and l
z
1
is the zpart angular momentum of the rst
particle. Let us further assume that the interaction is impulsive (g is suciently strong) so
that the single particle Hamiltonians, of the type given by Eq. (41), are ignorable.
The interaction Hamiltonian above can be used as a classical mechanical model of mea-
surement of the angular momentum l
z
1
of the rst particle by regarding the position of the
second particle as the pointer of the apparatus of measurement. To see this, rst, in such
a model l
z
1
is conserved:

l
z
1
= l
z
1
, H
I
= 0 where , is the usual Poisson bracket. The
interaction Hamiltonian of Eq. (A1) thus correlates the value of l
z
1
with the momentum
of the apparatus p
2
while keeping the value of l
z
1
unchanged. On the other hand, one also
has q
2
= q
2
, H
I
= gl
z
1
, which, noting the fact that l
z
1
is a constant of motion, can be
integrated to give
q
2
(t
M
) = q
2
(0) + gl
z
1
t
M
, (A2)
where t
M
is the time span of the interaction. The value of l
z
1
prior to the interaction can
thus in principle be inferred from the observation of the initial and nal values of q
2
. In
this way, the measurement of the physical quantity l
z
1
of the rst particle is reduced to the
measurement of the position of the second particle q
2
. In the model, q
2
(t) therefore plays
the role of the pointer of the apparatus of measurement.
Now let us apply the statistical model discussed in the main text to stochastically modify
the above classical mechanical model of measurement. We shall repeat all the steps and
manipulations performed in the main text. What we need to do is to put the Hamiltonian
29
of Eq. (A1) in place of Eq. (41). Again, let us rst consider a time interval of the scale

in which the absolute value of is eectively constant while its sign uctuates randomly
together with the random uctuations of the sign of . Let us then divide it into a series of
microscopic time intervals of length

, [(k 1)

, k

), k = 1, 2, . . . and attribute to each

interval a random value of (t) =
k
according to the probability distribution P
H
k
(
k
) =
P
H
k
(
k
) so that in each interval, the magnitude of is constant while its sign changes
randomly. During each time interval [(k 1)

, k

), the pair of equations in Eq. (29), each

with constant value of [
k
[, thus apply.
Let us rst consider a microscopic time interval [(k 1)

, k

). Using the form of H

I
given by Eq. (A1) to express q in term of p via q = H/p, the upper equation of (29)
becomes
x
1
= gy
1
_

q
2
S +

2

q
2

_
, y
1
= gx
1
_

q
2
S +

2

q
2

_
,
q
2
= g
_
x
1
_

y
1
S +

2

y
1

_
y
1
_

x
1
S +

2

x
1

__
, (A3)
and z
1
= 0. Again, assuming that the probability is conserved, one gets, after a simple
calculation, the following continuity equation:
0 =
t
+
q
( q)
=
t
gy
1

x
1
(
q
2
S) + gx
1

y
1
(
q
2
S) + gx
1

q
2
(
y
1
S)
gy
1

q
2
(
x
1
S) g(y
1

x
1

q
2
x
1

y
1

q
2
). (A4)
On the other hand, from Eq. (A1), (S) dened in Eqs. (8) and (18) is given by
(S) = 2g(x
1

q
2

y
1
S y
1

q
2

x
1
S). (A5)
Substituting this into the lower equation of (29), one then obtains
H
I
(q, p) =
t
S +

2

+g(x
1

y
1

q
2
S y
1

x
1

q
2
S). (A6)
Inserting the upper equation of (29) into the left hand side of the above equation, and using
30
Eq. (A1), one has, after arrangement

t
S + g
_
x
1

y
1
S y
1

x
1
S
_

q
2
S g
2
_
x
1

y
1

q
2
R
R
y
1

x
1

q
2
R
R
_
+

2
_

t
gy
1

x
1
(
q
2
S)
+gx
1

y
1
(
q
2
S) + gx
1

q
2
(
y
1
S) gy
1

q
2
(
x
1
S)
g(y
1

x
1

q
2
x
1

y
1

q
2
)
_
= 0, (A7)
where R =

and we have again used the identity of Eq. (47). Substituting Eq. (A4), the
last term of Eq. (A7) in the bracket vanishes to give

t
S + g
_
x
1

y
1
S y
1

x
1
S
_

q
2
S
g
2
_
x
1

y
1

q
2
R
R
y
1

x
1

q
2
R
R
_
= 0. (A8)
One thus has a pair of coupled equations (A4) and (A8) which are parameterized by .
This pair of equations are valid in a microscopic time interval of length

during which the

magnitude of is constant while its sign changes randomly with equal probability. Averaging
Eq. (A4) over the cases , recalling that the sign of is the same as that of , one has, by
the virtue of Eqs. (36) and (37),

t
gy
1

x
1
(
q
2
S
Q
) + gx
1

y
1
(
q
2
S
Q
)
+gx
1

q
2
(
y
1
S
Q
) gy
1

q
2
(
x
1
S
Q
) = 0. (A9)
On the other hand, averaging Eq. (A8) over the cases does not change anything. We
nally have a pair of Eqs. (A8) and (A9) which are now parameterized by [[ valid for a
microscopic time interval of duration

characterized by a constant [[.

Next, using dened in Eq. (50), and recalling that [[ is constant during the microscopic
time interval of interest with length

, the pair of Eqs. (A8) and (A9) can then be recast

into the following compact form:
i[[
t
=

2

H
I
. (A10)
Here

H
I
is a dierential operator dened as

H
I
.
= g

l
z
1
p
2
, (A11)
where p
i
= i
q
i
, i = 1, 2, are the quantum mechanical momentum operator referring to
the iparticle and

l
z
1
.
= x
1
p
y
1
y
1
p
x
1
is the zpart of the quantum mechanical angular
31
momentum operator pertaining to the rst particle; all are Hermitian. Recall that Eq.
(A10) is valid only for a microscopic time interval [(k 1)

, k

) during which [[ = [
k
[
is constant. For nite time interval, t >

one must then apply Eq. (A10) to each time

intervals, each is parameterized by a random value of [
k
[, k = 1, 2, . . . , consecutively.
Let us again consider a specic case when [[ = . Moreover, let us further assume
that the uctuations of [[ around its average is suciently narrow so that (q, [[; t) and
S(q; t, [[) can be approximated by the corresponding zeroth order terms, given respectively
by
Q
(q; t) and S
Q
(q; t). In this case, Eq. (A10) can be approximated as
i
t

Q
(q; t) =

H
I

Q
(q; t), (A12)
where
Q
is dened as in Eq. (53). Unlike Eq. (A10), Eq. (A12) is now deterministic and is
parameterized by the reduced Planck constant . Equation (A12) together with Eq. (A11)
is just the Schr odinger equation for the von Neumann model of measurement of angular
momentum operator

l
z
1
of the rst particle using the second particle as the apparatus.
Without giving the detail, let us mention that the same results will be obtained for the
measurement of position and linear momentum. In these cases, the interaction classical
Hamiltonian will be given by
H
I
= gO
1
(q
1
, p
1
)p
2
, (A13)
where O
1
= q
1
and O
1
= p
1
, respectively. Proceeding as before, repeating all the steps and
manipulations performed above, one will arrive at the Schr odinger equation with a quantum
Hamiltonian given by

H
I
= g

O
1
p
2
, (A14)
where

O
1
= q
1
= q
1
and

O
1
= p
1
= i
q
respectively for the measurement of position and
momentum. The measurement of energy should be reduced to the measurement of position,
momentum and angular momentum.
It is then instructive to discuss whether such a model of quantum measurement can
reproduce the prediction of quantum mechanics. It is also interesting to ask how the model
addresses the measurement problem. The above two problems are discussed in detail in Ref.
[70].
[1] J. S. Bell, Physics 1, 195 (1964).
32
[2] J. F. Clauser, M. A. Horne, A. Shimony, R. A. Holt, Phys. Rev. Lett. 23, 880 (1969).
[3] J. S. Bell, Speakable and Unspeakable in Quantum Mechanics (Cambridge University Press,
Cambridge, 1987).
[4] J. Freedman and J. F. Clauser, Phys. Rev. Lett. 28, 938 (1972).
[5] A. Aspect, P. Grangier and G. Roger, Phys. Rev. Lett. 47, 460 (1981).
[6] Z. Y. Ou and L. Mandel, Phys. Rev. Lett. 61, 50 (1988).
[7] Y. H. Shih and C. O. Alley, Phys. Rev. Lett. 61, 2921 (1988).
[8] J. G. Rarity and P. R. Tapster, Phys. Rev. Lett. 64, 2495 (1990).
[9] Z. Y. Ou, S. F. Pereira, H. J. Kimble and K. C. Peng, Phys. Rev. Lett. 68, 3663 (1992).
[10] J. Brendel, E. Mohler and W. Martienssen, Europhys. Lett. 20, 575 (1992).
[11] P. G. Kwiat, K. Mattle, H. Weinfurter, A. Zeilinger, A. V. Sergienko and Y. H. Shih, Phys.
Rev. Lett. 75, 4337 (1995)
[12] G. Weihs, M. Reck, H. Weinfurter and A. Zeilinger, Phys. Rev. Lett. 81, 5039 (1998).
[13] W. Tittel, J. Brendel, H. Zbinden and N. Gisin, Phys. Rev. Lett. 81, 3563 (1998).
[14] M. A. Rowe, D. Kielpinski, V. Meyer, C. A. Sackett, W. M. Itano, C. Monroe and D. J.
Wineland, Nature 149, 791 (2001).
[15] D. N. Matsukevich, P. Maunz, D. L. Moehring, S. Olmschenk and C. Monroe, Phys. Rev.
Lett. 100, 150404 (2008).
[16] D. Salart, A. Baas, C. Branciard, N. Gisin and H. Zbinden, Nature 454, 861 (2008).
[17] E. Santos, Found. Phys. 34, 1643 (2004).
[18] N. Brunner, D. Cavalcanti, S. Pironio, V. Scarani, and S. Wehner, arXiv:1303.2849v1.
[19] P. Eberhard, Nuovo Cimento B 46, 392 (1978).
[20] G. C. Ghirardi, A. Rimini, T. Weber, Lett. Nuovo Cimento 27, 293 (1980).
[21] D. Page, Phys. Rev. A 91, 57 (1982).
[22] J. Jarrett, Nous 18, 569 (1984).
[23] A. Shimony in J. Ellis and D. Amati (eds.), Quantum Reections (Cambridge University Press,
Cambridge, 2000).
[24] A. Shimony in Kamefuchi et al. (eds.), Foundations of Quantum Mechanics in Light of the
New Technology (Tokyo, Japan Physical Society, 1983).
[25] A. Shimony in R. Penrose and C. Isham (eds.), Quantum Concepts in Space and Time (Oxford,
Claredon Press, 1986).
33
[26] S. Popescu and D. Rohrlich, Found. Phys. 24, 379 (1994).
[27] J. Grunhaus, S. Popescu and D. Rohrlich, Phys. Rev. A, 53, 3781 (1996).
[28] S. Popescu and D. Rohrlich, arXiv:quant-ph/9709026v2.
[29] J. A. Wheeler in W. H. Zurek (ed.), Complexity, information content and the physics of
information (Westview Press, 1990).
[30] C. Rovelli, Int. J. Theor. Phys. 35, 1637 (1996).
[31] A. Zeilinger, Found. Phys. 29, 631 (1999).
[32] L. Hardy, Quantum theory from ve reasonable axioms, quant-ph/0101012.
[33] C. Simon, V. Buzek and N. Gisin, Phys. Rev. Lett. 87, 170405 (2001).
[34] R. Clifton, J. Bub and H. Halvorson, Found. Phys. 33, 1561 (2003).
[35] W. van Dam, Thesis, Univ. Oxford (1999); quant-ph/0501159.
[36] G. Brassard, H. Buhrman, N. Linden, A. A. Methot, A. Tapp and F. Unger, Phys. Rev. Lett.
96, 250401 (2006).
[37] P. Goyal, Phys. Rev. A 78, 052150 (2008).
[38] M. Pawlowski, T. Paterek, D. Kaszlikowski, V. Scarani, A. Winter and M. Zukowski, Nature
461, 1101 (2009).
[39] J. Oppenheim and S Wehner, Science 330, 1072 (2010).
[40] H. Barnum, S. Beigi, S. Boixo, M. B. Elliott and S. Wehner, Phys. Rev. Lett. 104, 140401
(2010).
[41] M. Navascues and H. Wunderlich, Proc. R. Soc. A 466, 881 (2010).
[42] D. Borivoje and C. Brukner in H. Halvorson (ed.), Deep beauty: understanding the quan-
tum world through mathematical innovation (Cambridge University Press, Cambridge, 2011);
(arXiv:0911.0695)
[43] L. Masanes and M. P. M uller, New J. Phys. 13, 063001 (2011).
[44] A. Caticha, J. Phys. A: Math. and Theor. 44, 225303 (2011).
[45] G. Chiribella, G. M. DAriano and P. Perinotti, Phys. Rev. A 84, 012311 (2011).
[46] G. de la Torre, L. Masanes, A. J. Short and M. P. M uller, Phys. Rev. Lett. 109, 090403
(2012).
[47] D. I. Fivel, Found. Phys. 42, 291 (2012).
[48] M. A. Nielsen and I. Chuang, Quantum Computation and Quantum Information (Cambridge
University Press, Cambridge, 2000).
34
[49] I. Fenyes, Z. Phys. 132, 81 (1952).
[50] W. Weizel, Z. Phys. 134, 264 (1953); 135, 270 (1953); 136, 582 (1954).
[51] D. Kershaw, Phys. Rev. 136B, 1850 (1964).
[52] E. Nelson, Phys. Rev. 150, 1079 (1966).
[53] L. de la Pe na-Auerbach, J. Math. Phys. 10, 1620 (1969).
[54] M. Davidson, Lett. Math. Phys. 3, 271 (1979).
[55] L. de la Pe na-Auerbach and A. M. Cetto, Found. Phys. 12, 1017 (1982).
[56] Ph. Blanchard, Ph. Combe and W. Zeng, Lecture Notes Physics 281, 1 (1987).
[57] F. Guerra and L. M. Morato, Phys. Rev. D 27, 1774 (1983).
[58] P. Garbaczewski and J. -P. Vigier, Phys. Rev. A 46, 4634 (1992).
[59] L. de la Pe na-Auerbach and A. M. Cetto, The quantum dice: An introduction to stochastic
electrodynamics (Kluwer Academic, Dordrecht, 1996).
[60] F. Markopoulou and L. Smolin, Phys. Rev. D 70, 124029 (2004).
[61] E. Santos, Phys. Lett. A 352, 49 (2006).
[62] L. de la Pe na-Auerbach, A. Valdes-Hernandes, A. M. Cetto, and H. M. Franca, Phys. Lett. A
375, 1720 (2011).
[63] D. Giulini, Lect. Notes Phys. 631, 17 (2003); arXiv:quant-ph/0304202v1.
[64] Y. Aharonov and D. Bohm, Phys. Rev. 115, 485 (1959).
[65] M. Peshkin and A. Tonomura, Lec. Notes. Phys. 340 (Springer, Berlin, 1989).
[66] H. Rund, The Hamilton-Jacobi theory in the calculus of variations: its role in mathematics
and physics (Van Nostrand, London, 1966).
[67] I and its average given by the Gibbs-Shannon entropy also play important roles in optimal
coding theory. See for example T. M. Cover and J. A. Thomas, Elements of Information
Theory (John Wiley & Sons, New Jersey, 2006).
[68] E. T. Jaynes, Probability theory: the logic of science (Cambridge University Press, Cambridge,
2003).
[69] A. Budiyono, Physica A 392, 307 (2013).
[70] A. Budiyono, J. Stat. Mech.: Theory and Experiment P11007, 1 (2013).
[71] A. Budiyono, Int. J. Theor. Phys. (2013), in press, DOI:10.1007/s10773-013-1925-4.
[72] J. S. Bell, Epistemological Letter 9, 11 (1976).
[73] D. Bohm and B. Hiley, The undivided universe: an ontological interpretation of quantum
35
theory (Routledge, London, 1993).
[74] S. Weinberg, Phys. Rev. Lett. 62, 485 (1989).
[75] N. Gisin, Phys. Lett. A 143, 1 (1990).
[76] J. Polchinski, Phys. Ref. Lett. 66, 397 (1991).
[77] W. K. Wootters and W. H. Zurek, Nature 299, 802 (1982).
[78] D. Dieks, Phys. Lett. A 92, (1982) 271.
[79] P. W. Milonni and M. L. Hardies, Phys. Lett. A 92, 321 (1982).
[80] A. Budiyono, Physica A 392, 43 (2013).
[81] J. S. Bell, Epistemol. Lett. 9, 11 (1976).
[82] A. Shimony, M. A. Horne, J. S. Clauser, Epistemol. Lett. 13, 9 (1976).
[83] B. d Espagnat, Phys. Rep. 110, 201 (1984).
[84] M. J. W. Hall, Phys. Rev. Lett. 105, 250404 (2010).
36