Chaos, Solitons and Fractals 19 (2004) 1323–1334
www.elsevier.com/locate/chaos
€dinger equation
Quantum derivatives and the Schro
Faycßal Ben Adda
a
a,*
, Jacky Cresson
b
Mathematical Sciences Department, Hail Community College, King Fahd University of Petroleum and Minerals,
P.O. Box 2440, 44 Avenue Bartholdi, 72000 le Mans, Hail, Saudi Arabia
b
Equipe de Math
ematiques de Besancßon, Universit
e de Franche-Comt
e, CNRS-UMR 6623, 16, route de Gray,
25030 Besancßon Cedex, France
Accepted 7 July 2003
Abstract
We define a scale derivative for non-differentiable functions. It is constructed via quantum derivatives which take
into account non-differentiability and the existence of a minimal resolution for mean representation. This justify
heuristic computations made by Nottale in scale-relativity. In particular, the Schr€
odinger equation is derived via the
scale-relativity principle and NewtonÕs fundamental equation of dynamics.
2003 Elsevier Ltd. All rights reserved.
1. Introduction
The aim of this article is to give a complete proof that the Schr€
odinger equation can be obtained from NewtonÕs
fundamental equation of dynamics in the scale-relativity setting developed by Nottale [1,11,14]. The articles [1,11,14]
contain only a sketch of proof and are based on informal arguments.
The main ‘‘ingredient’’ of NottaleÕs work is a new ‘‘derivative’’, 1 that he calls the scale derivative, which applies to
non-differentiable functions. Despite its importance, there exists no rigorous definition of this operator. In this article,
we give a precise definition of the scale derivative, as well as its geometric interpretation. As a consequence, we can
justify completely the computations made by Nottale in the articles [1,11,14].
The main problem is to define an extension of the classical differential calculus which has a clear physical meaning.
The starting point of our work is the following informal idea: Let c be a given curve. From the physical view-point,
we do not have access to c, but to a ‘‘representation’’ of it, denoted cs , which is always differentiable (up to a finite
number of points) at a given scale of observation s, and such that cs converge to c in C 0 topology. Of course, cs is not
always sufficient in order to describe the underlying physical process. In particular, if c is non-differentiable, the
fluctuations of cs when s goes to zero become non-negligible 2 contrary to what happens in the differentiable case.
This transition from a differentiable behaviour to a non-differentiable one when we follow cs must be quantified. In
this article, we introduce several concepts
R tþs in the special case of graphs of functions f ðtÞ and for a representation given
by the s-mean function fs ðtÞ ¼ ð1=2sÞ ts f ðsÞ ds.
In Section 2 we define the notion of s-differentiability, which leads to a natural transition quantity for the differentiable–non-differentiable behaviour of fs ðtÞ with respect to f ðtÞ, called minimal resolution and denoted sðf Þ.
*
Corresponding author. Address: Laboratoire dÕanalyse Numerique, tour 55-65, 5e etage, Universite Pierre et Marie-Curie, 4 Place
Jussieu, 75252 Paris Cedex, France.
E-mail addresses: fbenadda@kfupm.edu.sa, benadda@ann.jussieu.fr (F. Ben Adda), cresson@math.univ-fcomte.fr (J. Cresson).
1
We will see that this terminology is not appropriate.
2
Following Greene [8, p. 149] this is the basic reason for which differentiable (Riemanian) manifolds of EinsteinÕs relativity theory
cannot be used to describe the structure of space–time in quantum mechanics.
0960-0779/$ - see front matter 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0960-0779(03)00339-4
1324
F. Ben Adda, J. Cresson / Chaos, Solitons and Fractals 19 (2004) 1323–1334
Heuristically, we can say that for s > sðf Þ, the approximation of f ðtÞ by fs ðtÞ is sufficient to describe the behaviour of f
up to s small perturbations, which is not the case for s < sðf Þ.
When s < sðf Þ, we must take into account the non-differentiability of the limiting function f . We then define left and
right quantum derivatives, denoted þ f =t and f =t respectively,
which are nothing else than the derivatives of the
R tþrsðf Þ
right and left sðf Þ-mean function of f , i.e. fr ðtÞ ¼ ðr=sðf ÞÞ t
f ðsÞ ds, r ¼ . The fact to consider separately fþ ðtÞ
and f ðtÞ is due to the non-differentiability of f .
The scale derivative, defined in Section 3 and denoted by f =t, is a complex operator which takes into account the
two quantity r f =t, r ¼ , in such a way that when f is differentiable f =t ¼ f 0 ðtÞ. This ‘‘gluing’’ property of =t
to d=dt on the set of differentiable functions is a necessary constraint to be satisfied by all extended ‘‘differential’’
calculus. Paragraph 4 gives some properties of the scale derivative.
In Section 5, we generalize, following NottaleÕs scale relativity principle, NewtonÕs fundamental equation of dynamics, by replacing the classical derivative by our scale derivative. We prove that the new equation leads to a generalized Schr€
odinger equation. This justify heuristic computations made by Nottale [1,11,14].
2. Non-differentiable functions and minimal resolution
In the following, f is a continuous function, defined on an open set I of R. For all s 2 Rþ , we denote by fs the smean function defined by
Z tþs
fs ðtÞ ¼ ð1=2sÞ
f ðsÞ ds:
ð1Þ
ts
We call s-oscillation of f the quantity
oscs f ðtÞ ¼ supff ðt0 Þ f ðt00 Þ; t0 ; t00 2 ½t s; t þ sg:
ð2Þ
We say that the graph of f is fractal according to Tricot [12] if the quantity
oscs f ðtÞ
! þ1 when s ! 0;
s
ð3Þ
uniformly with respect to t, contrary to the differentiable case.
For 0 < a 6 1, we denote
jf ðtÞja;s ¼
sup
s;s0 2½ts;tþs; s6¼s0
jf ðsÞ f ðs0 Þj
:
js s0 ja
ð4Þ
Remark 1. We have
jf ja 6 jf ðtÞja;s 6 jf ja ;
ð5Þ
where
jf ja ¼ inf
t6¼t0 2I
jf ðtÞ f ðt0 Þj
;
jt t0 j
jf ja ¼ sup
t6¼t0 2I
jf ðtÞ f ðt0 Þj
;
jt t0 j
ð6Þ
and I is the closure of I.
For all t 2 I, we define aðtÞ such that
aðt; f ; sÞ ¼ supfa > 0 j jf ðtÞja;s 6¼ 0g:
ð7Þ
In the following, we do not write the explicit dependence of aðt; f ; sÞ with respect to f and s.
We denote
as f ðtÞ ¼
oscs f ðtÞ
:
2sjf ðtÞjaðtÞ;s
ð8Þ
We have
jfs ðtÞ f ðtÞj 6 2sjf ðtÞjaðtÞ;s as f ðtÞ;
so that as f ðtÞ is a measure of the approximation of f by the differentiable function fs .
ð9Þ
F. Ben Adda, J. Cresson / Chaos, Solitons and Fractals 19 (2004) 1323–1334
1325
Remark 2. For a differentiable function, we have aðtÞ ¼ 1 for all t 2 I and as f ðtÞ 6 1 for all t and all s.
We then have the following notion of s-differentiability at a point t:
Definition 1. Let f be a continuous function, defined on a compact set I of R, such that for all t 2 I,
ð10Þ
1 P aðtÞ > 0:
Let s > 0 be given. We say that f is s-differentiable at point t 2 I if
as f ðtÞ 6 1:
ð11Þ
A large class of continuous functions satisfy condition (10). At least, H€
olderian functions of order 0 < a < 1.
Moreover, there exists explicit construction of continuous functions with prescribed H€
older regularity, as discussed in
[16]. The simplest case is that of continuous function such that aðtÞ ¼ a for all t 2 I. As an example, we can consider the
Weierstrass function which possesses a uniform H€
older regularity [12].
Remark 3. The concept of s-differentiability can be understood as a way to characterize when the fluctuations of f are
‘‘small’’ with respect to the mean function fs . Of course, the notion of ‘‘smallness’’ depends on a normalization, which is
fixed. In our case, this normalization is contained in the choice of the number 1 for the upperbound of as;a f in (11), by
comparison with the differentiable case. Different kinds of normalization can be taken, leading to different notions of
non-differentiability. This is just a matter of choice.
We denote by sðf ÞðtÞ the minimal order of s-differentiability at point t:
sðf ÞðtÞ ¼ inffs P 0 j f is s-differentiable at point tg:
ð12Þ
Definition 2. Let f be a continuous real valued function defined on an open interval I R. We call minimal resolution
the quantity
sðf Þ ¼ inf sðf ÞðtÞ:
ð13Þ
t2I
We remark that for all k 2 R, we have
sðkf Þ ¼ sðf Þ and
sðf þ kÞ ¼ sðf Þ:
ð14Þ
If f is a fractal function (see (3)), then sðf Þ > 0. In fact, we have a more general result:
Lemma 1. If a continuous function f is differentiable then sðf Þ ¼ 0.
Proof. This follows from Remark 2.
As a consequence, if sðf Þ > 0, then f is an everywhere non-differentiable function. Of course, if f is differentiable on
a given subset J of I, then sðf Þ ¼ 0, despite the fact that sðf ÞðtÞ > 0 for all t 2 I n J . In the following, we consider only
continuous everywhere non-differentiable functions.
3. Quantum derivatives and the scale derivative
We define left and right quantum derivatives.
Definition 3. Let f : R ! R be a continuous function such that its minimal resolution sðf Þ satisfies sðf Þ > 0. We call
right and left quantum derivative the quantities
1326
F. Ben Adda, J. Cresson / Chaos, Solitons and Fractals 19 (2004) 1323–1334
þ f
f ðt þ sÞ f ðtÞ
ðtÞ ¼
;
t
s
f
f ðtÞ f ðt sÞ
ðtÞ ¼
;
t
s
ð15Þ
respectively (where s ¼ sðf Þ). If s ¼ 0, then
For a differentiable function, we have
r f
t
þ f
t
ðtÞ ¼ lims!0
f ðtþsÞf ðtÞ
.
s
ðtÞ ¼ t f ðtÞ ¼ f 0 ðtÞ.
Remark 4. We refer to [17] for a careful study of such kind of difference operators in the context of what they call time
scales calculus.
We define quantum functions fþ ðtÞ and f ðtÞ as
Z
1 tþs
f ðsÞ ds;
fþ ðtÞ ¼
s t
Z t
1
f ðtÞ ¼
f ðsÞ ds;
s ts
ð16Þ
respectively.
We have
r f
ðtÞ ¼ fr0 ðtÞ;
t
Remark 5. As
þ f
t
r ¼ :
ðtÞ and
f
t
ð17Þ
ðtÞ are continuous, the functions fþ ðtÞ and f ðtÞ are well defined.
Using the quantum functions fþ ðtÞ and f ðtÞ, we can write f as
f ðtÞ ¼ fþ ðtÞ þ nfþ ðtÞ;
f ðtÞ ¼ f ðtÞ þ nf ðtÞ;
ð18Þ
where nfþ and nf are non-differentiable functions representing fluctuations with respect to the right (resp. left) mean
function. In the following, we will denote nfþ and nf by nþ and n if no confusion is possible.
Lemma 2. Let f be a continuous function such that sðf Þ > 0. Then for h sufficiently small, we have
f ðt þ rhÞ ¼ f ðtÞ þ r
where
r f
t
r f
ðtÞh þ ðnr ðt þ hÞ nr ðtÞÞ þ oðhÞ;
t
ð19Þ
ÞÞf ðtÞ
, r ¼ .
ðtÞ ¼ f ðtþrsðf
rsðf Þ
Proof. We have f ðt þ rhÞ ¼ fr ðt þ rhÞ þ nr ðt þ rhÞ þ oðhÞ for r ¼ . A first order TaylorÕs expansion of fr ðt þ rhÞ
gives
fr ðt þ rhÞ ¼ fr ðtÞ þ rfr0 ðtÞh þ oðhÞ:
ð20Þ
Using (17), we obtain for h sufficiently small,
þ f
ðtÞh þ nþ ðt þ hÞ þ oðhÞ;
t
f
f ðt hÞ ¼ f ðtÞ þ
ðtÞh þ n ðt hÞ þ oðhÞ:
t
f ðt þ hÞ ¼ fþ ðtÞ þ
Replacing fr ðtÞ by f ðtÞ nr ðtÞ, we obtain (19).
In the following, we denote
rrf ðt; hÞ ¼ nfr ðt þ rhÞ nfr ðtÞ
for r ¼ , or rr ðt; hÞ if no confusion is possible.
ð21Þ
F. Ben Adda, J. Cresson / Chaos, Solitons and Fractals 19 (2004) 1323–1334
1327
Definition 4. Let f : R ! R be a continuous function and sðf Þ its minimal resolution. We call scale derivative of f at
point t the quantity
f
1 þ f
f
1 þ f
f
ð22Þ
ðtÞ ¼
ðtÞ þ
ðtÞ i
ðtÞ þ
ðtÞ ; i2 ¼ 1:
t
2 t
t
2 t
t
only to be coherent with NottaleÕs terminology
Remark 6. We use the terminology of scale derivative for the operator t
[1,11]. However, as quantum derivatives, the scale derivative is not a derivation on the set of continuous functions, i.e. it
does not satisfy the Liebniz rule. 3
When f is differentiable, we obtain the classical derivative. The real part of the scale derivative is the derivative of the
s-mean function of f , fs . We have
fþ ðtÞ þ f ðtÞ
¼ fs ðtÞ:
2
In our computations about the Schr€
odinger equation, we will need the following definition of the scale derivative for
complex valued functions:
Let f : R ! C, t 7! f ðtÞ, be a continuous function. We have f ðtÞ ¼ Reðf ðtÞÞ þ i Imðf ðtÞÞ where Re and Im are the
real and imaginary part of f ðtÞ respectively. The functions Reðf ðtÞÞ and Imðf ðtÞÞ are continuous real valued functions.
We define the scale derivative of f as
f
Reðf Þ
Imðf ÞðtÞ
ðtÞ ¼
ðtÞ þ i
:
t
t
t
ð23Þ
4. Consequences of non-differentiability
We keep the notations of Section 3.
4.1. Main lemma
In order to derive the Schr€
odinger equation from NewtonÕs fundamental equation of dynamics in Section 5, we need
to compute the scale derivative of a composed function of the form f ðxðtÞ; tÞ where f ðx; tÞ is a differentiable function,
and xðtÞ is not. The following lemma gives the formula.
Lemma 3. Let f ðx; tÞ be a C 3 function. Let xðtÞ be a continuous function such that sðxÞ > 0 and
xðt þ rhÞ ¼ xðtÞ þ r
r x
ðtÞh þ rrx ðtÞ
t
ð24Þ
for h > 0 sufficiently small, r ¼ .
We consider the function gðtÞ ¼ f ðxðtÞ; tÞ. We have
(i) sðgÞ > 0.
For all h sufficiently small, we denote
gðt þ rhÞ ¼ gðtÞ þ r
r g
ðtÞh þ rrg ðt; hÞ;
t
r ¼ :
ð25Þ
We assume that for r ¼ , we have
2
x
(*) the function
admits
g ðrr ðt; hÞÞ
a right derivative at point h ¼ 0 for all t,
of x
(**) we have rr ðt; hÞ ox rr ðt; hÞ h ! 0 when h ! 0r .
3
An operator D on an algebra A is a derivation if 8x; y 2 A we have Dðx yÞ ¼ Dx y þ x Dy which is usually called Liebniz identity
(see [10]).
1328
F. Ben Adda, J. Cresson / Chaos, Solitons and Fractals 19 (2004) 1323–1334
Then, we have
ðiiÞ
r g of
r x
of
1 o2 f
¼
ðxðtÞ; tÞ
ðtÞ þ ðxðtÞ; tÞ þ
ðxðtÞ; tÞar ðtÞ;
t
ox
t
ot
2 ox2
ð26Þ
where aþ ðtÞ (resp. a ðtÞ) is the right derivative at point h ¼ 0 of rþ2 ðt; hÞ (resp. r2 ðt; hÞ).
Remark 7. If f is a flat function at point ðxðtÞ; tÞ then ðof =oxÞðxðtÞ; tÞ ¼ 0 and assumption (**) is equivalent to the
differentiability of rr ðt; hÞ, r ¼ at h ¼ 0, which is not possible as sðxÞ > 0.
Proof. For (i) this follows easily from Lemma 1. For (ii), we have
þ x
ðtÞh þ oðhÞ þ rþx ðt; hÞ; t þ h ;
gðt þ hÞ ¼ f ðxðt þ hÞ; t þ hÞ ¼ f xðtÞ þ
t
using (19).
As rþx ðt; hÞ ! 0 when h ! 0, we have, doing a TaylorÕs expansion of f in the neighborhood of ðxðtÞ; tÞ,
of
þ x
of
ðtÞh þ oðhÞ þ rþx ðt; hÞ þ ðxðtÞ; tÞh
gðt þ hÞ ¼ gðtÞ þ ðxðtÞ; tÞ
ox
t
ot
2
2
1 of
þ x
þ
ðtÞh þ oðhÞ þ rþx ðt; hÞ
ðxðtÞ; tÞ
2 ox2
t
!
o2 f
þ x
o2 f
x
2
ðxðtÞ; tÞ
ðtÞh þ oðhÞ þ rþ ðt; hÞ h þ 2 ðxðtÞ; tÞh þ
þ2
oxot
t
ot
ð27Þ
ð28Þ
As ðrþx Þ2 ðt; hÞ is differentiable at point h, we have ðrþx Þ2 ðt; hÞ ¼ aþ ðtÞh þ oðhÞ. By factorizing terms of order 1 in h, we
obtain
of
þ x
of
1 o2 f
ðxðtÞ; tÞ
ðtÞ þ ðxðtÞ; tÞ þ
ðxðtÞ;
tÞa
ðtÞ
h þ oðhÞ þ Rþ
ð29Þ
gðt þ hÞ ¼ gðtÞ þ
þ
g ðt; hÞ;
ox
t
ot
2 ox2
where
Rþ
g ðt; hÞ ¼
of
ðxðtÞ; tÞrþx ðt; hÞ
ox
ð30Þ
is non-differentiable and rþg ðt; 0Þ ¼ 0.
We deduce
þ
of
þ x
of
1 o2 f
ðxðtÞ; tÞ
ðtÞ þ ðxðtÞ; tÞ þ
ðxðtÞ;
tÞa
ðtÞ
f ðxðtÞ; tÞ
þ
ox
t
ot
2 ox2
t
1 þ
8h > 0:
¼ ðrg ðt; hÞ þ oðhÞ Rþ
g ðt; hÞÞ
h
ð31Þ
By replacing Rg ðt; hÞ by (30) and taking the limit in (31), we obtain, using (**),
þ
of
þ x
of
1 o2 f
ðxðtÞ; tÞ
ðtÞ þ ðxðtÞ; tÞ þ
f ðxðtÞ; tÞ ¼
ðxðtÞ; tÞaþ ðtÞ;
ox
t
ot
2 ox2
t
ð32Þ
where aþ ðtÞ is the derivative of ðrþx Þ2 ðt; hÞ at point h ¼ 0.
Similar computations allow us to prove that
of
x
of
1 o2 f
ðxðtÞ; tÞa ðtÞ;
f ðxðtÞ; tÞ ¼
ðxðtÞ; tÞ
ðtÞ þ ðxðtÞ; tÞ þ
t
ox
t
ot
2 ox2
where a ðtÞ is the derivative of ðrx Þ2 ðt; hÞ at point h ¼ 0.
ð33Þ
About Ito’s stochastic calculus. As a consequence, the non-differentiability of xðtÞ introduces additional spatial terms
in the derivative of f ðxðtÞ; tÞ with respect to the differentiable case. This formula is similar to ItoÕs formula in stochastic
calculus [15]. However, ItoÕs formula is obtained under probabilistic assumptions. In Lemma 4, this follows from the
F. Ben Adda, J. Cresson / Chaos, Solitons and Fractals 19 (2004) 1323–1334
1329
geometric assumption that f is non-differentiable. For more details about our scale derivative and ItoÕs formula, we
refer to [2,3].
4.2. The complex case
Let
RR !
ðx; tÞ
7!
C:
C
:
Cðx; tÞ
We denote Aðx; tÞ ¼ ReCðx; tÞ and Bðx; tÞ ¼ ImCðx; tÞ, then Cðx; tÞ ¼ Aðx; tÞ þ iBðx; tÞ.
Let
x:
!
7!
R
t
R
xðtÞ
be a continuous function such that its minimal resolution sðf Þ satisfies sðf Þ > 0. We define the functions xþ ðtÞ, x ðtÞ,
nþ ðtÞ and n ðtÞ as in Section 3.
We denote by ar ðtÞ the right derivative of ðnr ðt þ rhÞ nr ðtÞÞ2 at point h ¼ 0.
Lemma 4. The scale derivative of the function
C:
R !
t 7!
C
CðxðtÞ; tÞ
is
C oC x oC 1
o2 C
¼
þ
þ aðtÞ 2 ;
t
ot
t ox 2
ox
ð34Þ
where
aðtÞ ¼
aþ ðtÞ a ðtÞ
2
i
aþ ðtÞ þ a ðtÞ
:
2
ð35Þ
Proof. We denote by AðtÞ and BðtÞ the functions AðxðtÞ; tÞ and BðxðtÞ; tÞ respectively. By Lemma 4, we have
r A oA r x oA
1
o2 A
¼
þ
þ r ar ðtÞ 2 :
t
ox t
ot
2
ox
We deduce
A oA 1
¼
t
ox 2
þ x x
þ
t
t
þ
oA 1 o2 A 1
oA 1 þ x x
1 o2 A 1
þ
ða
ða
þ
ðtÞ
a
ðtÞÞ
i
ðtÞ
þ
a
ðtÞÞ
;
þ
þ
ot 2 ox2 2
ox 2 t
t
2 ox2 2
that is
A oA x oA 1
o2 A
¼
þ
þ aðtÞ 2
t
ox t ot 2
ox
aþ ðtÞa ðtÞ
aþ ðtÞþa ðtÞ
with aðtÞ ¼
i
.
2
2
We obtain a similar formula for BðtÞ. Hence, by definition, we have
2
C x oA
oB
oA
oB
1
oA
o2 B
þ
i
:
¼
þi
þ
þi
þ aðtÞ
t
t ox
ox
ot
ot
2
ox2
ox2
We then obtain
C x oC 1
o2 C oC
¼
þ aðtÞ 2 þ
;
t
t ox 2
ox
ot
which concludes the proof.
1330
F. Ben Adda, J. Cresson / Chaos, Solitons and Fractals 19 (2004) 1323–1334
4.3. About the regularity assumption (*)
Feynman and Hibbs [7] have proved that generic trajectories of quantum particles are continuous non-differentiable
curves. However, there exists a quadratic velocity, that is, the quantity
lim0
x!x
ðf ðxÞ f ðx0 ÞÞ2
exists:
x x0
ð36Þ
Hence, the following quantities
ðf ðt þ hÞ f ðtÞÞ2
h
and
ðf ðtÞ f ðt hÞÞ2
h
ð37Þ
keep sense when h > 0þ , and are equals.
For all h > 0, we have
f ðt þ hÞ f ðtÞ
rþ2 ðt; hÞ
¼
h
h
þ f
t
h þ oðhÞ
2
:
ð38Þ
When h ! 0þ , we obtain
limþ
h!0
rþ2 ðt; hÞ
ðf ðt þ hÞ f ðtÞÞ2
:
¼ limþ
h!0
h
h
ð39Þ
Similar computations prove that
lim
h!0
r2 ðt; hÞ
ðf ðtÞ f ðt hÞÞ2
:
¼ lim
h!0
h
h
We denote by aþ ðtÞ (resp. a ðtÞ) the right derivative of rþ2 ðt; hÞ (resp. r2 ðt; hÞ). As the quadratic velocity is well
defined, we must have
aþ ðtÞ ¼ a ðtÞ:
ð40Þ
Assumption (*) is then satisfied by functions describing quantum trajectories.
Remark 8
• For the Browninan motion, Einstein [9] has proved that f ðt þ hÞ f ðtÞ h1=2 for h > 0, which is in accordance with
(*).
• The existence of a quadratic velocity is equivalent to 1/2-right differentiability of f following [4].
5. Scale relativity principle and Schrödinger equation
In [1,11,14], Nottale announce that the Schr€
odinger equation can be obtained from the classical NewtonÕs equation
of dynamics using a quantization procedure which comes from the scale relativity theory. The scale relativity theory is
developed by Nottale since 1980. Its aim is to generalize EinsteinÕs relativity principle in order to derive quantum
mechanics from a first principle. We refer to his work for more details [1]. The quantization procedure is based on a
generalized Euler–Lagrange equation coming from NottaleÕs theory and the use of the scale derivative instead of the
classical derivative. The computations made by Nottale in [1,11,14] are informal and based on heuristic arguments.
Using the scale derivative defined in the previous paragraph, we give a complete and detailed proof of his approach.
5.1. Action functional and wave function
Let x : R ! R, t 7! xðtÞ, be a continuous, non-differentiable function, describing the trajectory of a quantum particle
of mass m. Let
v:
R !
t 7!
C
x
vðtÞ ¼
t
1331
F. Ben Adda, J. Cresson / Chaos, Solitons and Fractals 19 (2004) 1323–1334
be its velocity. Let
U:
RR
ðx; tÞ
!
7!
C
Uðx; tÞ
ð41Þ
be a differentiable function, called scalar potential.
The action functional is then defined by
L:
RCR
!
ðx; v; tÞ
7!
C
1 2
:
mv Uðx; tÞ
2
ð42Þ
We note that the map Lðx; v; tÞ is differentiable with respect to x and v.
Scale assumption. We assume, following Nottale [1], that equation of motion for particles is given by the following
Euler–Lagrange generalized equation:
oL
oL
:
ð43Þ
¼
t ov
ox
Nottale deduce this equation informally via his scale relativity principle. We refer to his work [1,5,6] for more details.
We then have
m
v
oU
¼
:
t
ox
ð44Þ
We call this equation, fundamental equation of dynamics, by analogy with NewtonÕs classical equation.
The momentum is defined by p ¼ oL
, which gives p ¼ mv. We introduce an action A as
ov
A:
RR
ðx; tÞ
!
7!
C
;
Aðx; tÞ
ð45Þ
which is a differentiable function, related to the momentum via the relation p ¼ oAðx;tÞ
. We then obtain v ¼ m1
ox
We can introduce a function
w:
RR
ðx; tÞ
!
7!
C
;
wðx; tÞ
oA
.
ox
ð46Þ
differentiable, such that
iAðx;tÞ
wðx; tÞ ¼ e 2mc ;
ð47Þ
where c 2 R is a normalization constant to be determined.
This function is of course the wave function of a particle. We note that Aðx; tÞ ¼ 2mci lnwðx; tÞ and v ¼
lnw
, where ln is the complex logarithm.
vðx; tÞ ¼ i2c o ox
Remark 9. We obtain the classical correspondence principle of quantum mechanics for momentum and energy, that is
p ¼ 2imc
ow 1
;
ox w
E ¼ 2imc
ow 1
:
ot w
ð48Þ
5.2. Schr€odinger equation
Using the wave function, the fundamental equation of dynamics looks like
o
oU
2icm
ðln wÞ ¼
:
t ox
ox
Lemma 5. The fundamental equation of dynamics is equivalent to
2
aðtÞ
ow
1
o lnw
o2 w 1
þ i2cm
þ icaðtÞ 2 ¼ U þ aðxÞ:
i2cm ic þ
2
2
ox
ot
ox w
w
ð49Þ
ð50Þ
1332
F. Ben Adda, J. Cresson / Chaos, Solitons and Fractals 19 (2004) 1323–1334
Proof. The fundamental equation of dynamics is
o lnw
oU
2icm
¼
:
t
ox
ox
lnw
ðxðtÞ; tÞ. We have, using Lemma 5,
We denote NðtÞ ¼ nðxðtÞ; tÞ ¼ o ox
o
N on x on 1
o2 n
ðln wÞ ¼
¼
þ þ aðtÞ 2 :
t ox
t
ox t ot 2
ox
A simple computation gives
on
ot
¼ oxo
o lnw
ot
on x
on
¼ 2ic n;
ox t
ox
and
as function of w. We then have
by definition of v ¼ x
t
2 !
on x
on2
o
1 ow
¼ ic
¼ ic
:
ox t
ox w2 ox
ox
Moreover, we have
o2 n
o
¼
ox2 ox
o2 w 1
ox2 w
ow
ox
2
1
w2
!
:
We deduce, by gathering these terms
!
2
N
o
aðtÞ
ow
1 aðtÞ o2 w 1 o lnw
¼
þ
ic þ
:
þ
t
ox
2
ox
2 ox2 w
ot
w2
By replacing in the fundamental equation of dynamics, we obtain
!
2
o
aðtÞ
ow
1
o2 w 1
o lnw
oU
:
2icm ic þ
¼
þ icmaðtÞ 2 þ 2icm
ox
2
ox
ox w
ot
ox
w2
We conclude the proof by integration.
h
As a particular case, when the non-differentiability of xðtÞ is uniform, we obtain the classical Schr€
odinger equation.
Corollary 1. Let xðtÞ be a continuous, non-differentiable function such that
aðtÞ ¼ i2c:
ð51Þ
Then the fundamental equation of dynamics takes the form
i2cm
ow
o2 w
þ 2c2 m 2 ¼ ðU þ aðxÞÞw:
ot
ox
ð52Þ
We can always choose a solution of (52) such that aðxÞ ¼ 0. In this case, when
c¼
h
;
2m
ð53Þ
where h is the Planck constant, we obtain the classical Schr€
odinger equation
ih
ow h2 o2 w
¼ Uw:
þ
ot 2m ox2
ð54Þ
Þ2 w12 : In this case, by replacing aðtÞ and remarking that
Proof. The choice of aðtÞ allows us to cancel the term ðow
ox
o lnw
ow 1
,
we
obtain
Eq.
(52).
¼
ot
ot w
F. Ben Adda, J. Cresson / Chaos, Solitons and Fractals 19 (2004) 1323–1334
1333
Let w be a solution of (52). We search for a function w~ solution of the equation
i2cm
ow~
o2 w~
þ 2c2 m 2 ¼ Uw~
ot
ox
ð55Þ
of the form
Aðx;tÞ
w~ ¼ ei 2mc þhðxÞ ¼ wðx; tÞHðxÞ;
ð56Þ
where
HðxÞ ¼ ehðxÞ :
That is, we modify the phase of the wave function w.
We then have
ow~ ow
¼
H þ wH0 ;
ox
ox
o2 w~ o2 w
ow
¼ 2 H þ 2 H0 þ wH00 ;
ox2
ox
ox
ow ow
¼
H;
ot
ot
where H0 ðxÞ and H00 ðxÞ are the first and second derivative of HðxÞ.
By replacing in (55), we obtain
2
ow
ow
ow 0
00
2
H þ 2c m
i2cm
H þ 2 H þ wH ¼ UwH:
ot
ox2
ox
We deduce an ordinary differential equation in H of the form
ow
o2 w
ow 0
H þ 2c2 mwH00 ¼ 0:
þ 2c2 m 2 Uw þ 4c2 m
H i2cm
ot
ox
ox
ð57Þ
ð58Þ
ð59Þ
As w is a solution of (52), we have
HaðxÞw þ 4c2 m
ow 0
H þ 2c2 mwH00 ¼ 0:
ox
ð60Þ
This differential equation has always a solution. Hence, we can always choose a solution of (52) such that aðxÞ ¼ 0.
The choice of c in order to obtain Eq. (54) is then done by identification. h
Remark 10. Our derivation of the Schr€
odinger equation is done under the scale assumption, which follows from
NottaleÕs physical concept of scale relativity principle. We refer to [13, pp. 254–257] for a completely different proof.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
Nottale L. Scale-relativity and quantization of the universe I. Theoritical framework. Astronom Astrophys 1997;327:867–89.
Ben Adda F. Differentielles quantiques et formule dÕIto, preprint, 2001.
Ben Adda F, Cresson J. Divergence dÕechelle et differentiabilite. C R Acad Sci Paris, Serie I 2000;330:261–4.
Ben Adda F, Cresson J. About non-differentiable functions. J Math Anal Appl 2001;263:721–37.
Cresson J. Scale relativity for one dimensional non-differentiable manifolds. Chaos, Solitons & Fractals 2002;14(4):553–62.
Cresson J. Quelques problemes de physique mathematiques, Memoire dÕhabilitation a diriger des recherches, Univ. de FrancheComte, no. 99, Decembre 2001.
Feynman R, Hibbs A. Quantum mechanics and path integrals. MacGraw-Hill; 1965.
Greene B. In: R. Laffont (Ed.), LÕunivers elegant. Une revolution scientifique: de lÕinfiniment grand a lÕinfiniment petit,
lÕunification de toutes les theories de la physique, 2000.
Einstein A. Ann Phys 1905;17:549.
Jacobson A. Lie algebras. Intersci Tracts Pure Appl Math 1962;10.
Nottale L. Fractal space–time and microphysics. World Scientific; 1993.
Tricot C. Courbes et dimension fractale. 2nd ed. Springer; 1999.
1334
F. Ben Adda, J. Cresson / Chaos, Solitons and Fractals 19 (2004) 1323–1334
[13] Feynman R. LÕelectrodynamique quantique et lÕespace-temps, Conference Nobel dans ‘‘La nature de la physique’’, Ed. du Seuil,
1980.
[14] Nottale L. The scale-relativity program. Chaos, Solitons & Fractals 1999;10(2–3):459–68.
[15] Kahane J-P. Le mouvement Brownien. In: Maeriaux pour lÕhistoire des mathematiques au XX eme siecle, Seminaires et congres 3,
SMF, 1998, p. 123–57.
[16] Daoudy K, Levy Vehel J, Meyer Y. Construction of continuous functions with prescribed local regularity, preprint INRIA, 2763,
1995.
[17] Bohner M, Peterson A. Dynamic equations on time scales; an introduction with applications. Birkhauser; 2001.