THE THEORY OF SCALE

RELATIVITY∗
LAURENT NOTTALE
CNRS, Département d’Astrophysique Extragalactique et de Cosmologie,
Observatoire de Paris-Meudon, F-92195 Meudon Cedex, France

Received 18 April 1991

Revised 26 September 1991

Abstract
Basing our discussion on the relative character of all scales in nature and
on the explicit dependence of physical laws on scale in quantum physics, we
apply the principle of relativity to scale transformations. This principle, in
combination with its breaking above the Einstein-de Broglie wavelength and
time, leads to the demonstration of the existence of a universal, absolute and
impassable scale in nature, which is invariant under dilatation. This lower
limit to all lengths is identified with the Planck scale, which now plays for
scale the same role as is played by light velocity for motion. We get new
scale transformations of a Lorentzian form and generalize the de Broglie and
Heisenberg relations. As a consequence the high energy length and mass scales
now decouple, energy and momentum tending to infinity when resolution tends
to the Planck scale, which thus plays the role of the previous zero point.
This theory solves the problem of divergence of charge and mass (self-energy)
in electrodynamics, implies that the four fundamental couplings (including
gravitation) converge at the Planck energy, improves the agreement of GUT
predictions with experimental results, and allows one to get precise estimates
of the values of the fundamental coupling constants.

1 Introduction
Since the Galilean analysis about the nature of inertial motion, the theory of relativ-
ity has been developed by extending its application domain to coordinates systems
involved in more and more general states of motion: this was partly achieved in
Einstein’s special and general theories of relativity. Hence the principle of general
relativity states that “the laws of physics must be of such a nature that they apply
to systems of reference in any kind of motion” [1].
However, as pointed out by Levy-Leblond, [2] the abstract principle of relativity
should be distinguished from any of its possible realizations as concrete theories of
relativity. This point of view may still be generalized into a framework in which
c
∗ International Journal of Modern Physics A, Vol. 7, No. 20 (1992) 4899-4936. World

Scientific Publishing Company. Version complemented by notes and errata (15 May 2003).

1
relativity is considered as a general method of thinking in sciences [3]: it consists
in analysing how the results of measurements and (cor)relations between them are
dependent on the particular conditions under which the measurements have been
performed. Assuming that these results are measured in some “reference system”
(e.g. coordinate systems in case of position and time measurements), these condi-
tions may be characterized as “states” of the reference systems.
Such states of reference systems play a special role in physics: they being defined
as characteristics of the reference systems themselves, no absolute value can be
attributed to them, but only relative ones, since they can be defined and described
only with respect to another reference system. Two systems at least are needed to
define them. As a consequence the transformation laws between reference frames
will be of the greatest importance in a relativity theory.
In such a frame of thinking, the most fundamental relativity is the relativity of
positions and instants. It is usually expressed in terms of homogeneity and isotropy
of space and uniformity of time and actually makes up the basis of the whole of
physics. It states that there is no preferential origin for a coordinate system and is
finally included in special relativity through the Poincaré group.
Then Einstein’s relativity is, strictly, a theory of “motion relativity”, since the
particular relative state of coordinate systems which the special and general theories
of relativity have extensively analysed (in the classical domain) is their state of
motion.
We suggest in this paper that scale (i.e. resolution with which measurements
have been performed) may also be defined as a relative state of reference systems,
and that Einstein’s principle of relativity can be generalized by requiring that the
laws of physics apply to any systems of coordinates, whatever their state of scale.
In other words, we shall require scale covariance of the equations of physics. The
quantum behavior of microphysics may to some extent be reinterpreted as a man-
ifestation of scale relativity. But in its present form quantum field theory corre-
sponds, rather, to a Galilean version of such a scale relativity theory, especially in
the renormalization group approach.
Indeed we first recall how the renormalization group may be applied to space-
time itself, yielding an anomalous dimension1 for space and time variables. Then we
demonstrate in a general way that the principle of relativity alone, in its Galilean
form (i.e. without adding any extra postulate of invariance), is sufficient to derive
the Lorentz transformation as a general solution to the (special) relativity problem.
Once applied to scale, and owing to the fact that physical laws become explicitely
scale-dependent only for resolutions below the de Broglie length and time (namely,
that scale relativity is broken at the de Broglie transition), this reasoning leads to
the existence of an absolute, universal scale which is invariant under dilatations
and so cannot be exceeded. Then, after having identified this scale as the Planck
scale, we attempt to develop a theory based on this new structure: the Einstein-de
Broglie and Heisenberg relations are generalized, and first implications concerning
the domain of high energy physics are considered.
1 This dimension is ‘anomalous’ in the sense that it cannot be obtained from simple dimensional

analysis. It is actually a ‘scale dimension’, defined as the difference between the fractal dimension
and the topological dimension. It becomes a variable in scale-relativity, identified with a fifth
dimension that we have called ‘djinn’ in subsequent papers.

2
2 The Dependence on Scale of Microphysics
Starting from the Planck/Einstein work at the beginning of the century, the devel-
opment of the quantum theory has forced physicists to admit that the microscopic
world behaves in a radically new way compared to the classical way. One of the main
property of microphysics irreducible to being classical is described by the Heisenberg
relations, which imply a profound scale dependence of physical laws in the quan-
tum domain. When ∆p ≫ p0 (i.e. ∆x ≪ λdB , the de Broglie length), p becomes
of the order of ∆p and the position-momentum Heisenberg relation ∆p . ∆x ≈ ~
becomes p ≈ ~/∆x. Similarly when ∆E ≫ E0 (i.e. ∆t ≪ τdB , the de Broglie
time), E becomes of the order of ∆E and the time-energy Heisenberg relation
∆E . ∆t ≈ ~ becomes E ≈ ~/∆t. In the presently best accepted interpretation
of quantum mechanics, this behaviour is understood as a consequence of the un-
controllable interaction between the measurement apparatus and the system to be
measured. However it is remarkable that the Heisenberg relations are universal and
independent from any particular measurement process. They may be derived from
the general law that the momentum probability amplitude and position probability
amplitude are reciprocal Fourier transforms.
So we have proposed a different interpretation [3]: that the quantum behaviour
is a consequence of a fundamental and universal dependence of space-time itself on
resolution, which is revealed in any measurement: namely, that the quantum space-
time has fractal properties. This leads to the view that scale should be explicitely
introduced into the fundamental laws of physics, and that this goal may be achieved
by identifying it as a state of the reference system (i.e. of measurement apparatus
in the language of quantum mechanics). In this frame of thought, the Heisenberg
relations tell us that the results of measurements of momentum and energy are
relative to the state of scale of the reference system.
The scale dependence of microphysics has already taken on reinforced impor-
tance in the study of the asymptotic behaviour of quantum field theories, presently
best described by the renormalization group methods [4]-[8], which led to some
important results, like asymptotic freedom of QCD, the variation of coupling con-
stants with scale and their convergence at the “Grand Unified Theory” scale ≈ 1015
GeV. These methods have for the first time explicitly introduced scale into physical
equations. Indeed it is remarkable that below the Compton length of the electron,
there is a strong “degeneration” of space and time variables: the velocity becomes
disqualified as a pertinent mechanical variable (all velocities are close to the velocity
of light) and the classical laws of mechanics are actually replaced by dilatations in
terms of Lorentz γ factors. Then, at high energy, the laws of scale actually take the
place of the laws of motion.
There are, however, several additional elements in our proposal with respect to
the standard renormalization group approach. The first is that it is argued that
scale, like motion, may be considered as a state of coordinate systems which can
never be defined in an absolute way, and thus comes under a relativity theory. In
this respect we may identify the theory of the renormalization group as a Galilean
version of scale relativity. The second is that the renormalization group is, strictly,
only a semi-group (one integrates the small scales to get the larger ones) [9], while
one may hope it to be completed in the future by an inverse transformation, at least
for some elementary physical systems. This would mean being able to deduce the

3
small scale structure from the larger scale: this is exactly what a fractal generator
makes [3, 10, 11]. The third new element is that the relativistic analysis of scale,
once applied to space and time variables themselves, finally leads to a completely
new structure of physical laws, as will be demonstrated in the following.

3 Galilean Scale Relativity

One of the main characteristics of scale which point toward the need for a scale
relativity theory is the nonexistence of an a priori absolute scale. Just as one may
write for velocities in Galilean motion relativity :

v = v2 − v1 = (v2 − v0 ) − (v1 − v0 ), (1)

one may, in present physics, write for a scale ratio

∆x2 ∆x2 /∆x0
̺= = . (2)
∆x1 ∆x1 /∆x0
It is indeed clear that one can never define the length of an object without
comparing it to another object: only scale ratios, i.e. dilatations, have a physical
meaning. The expression (2) may be written under the same additive group form
as Eq. (1), in a logarithmic representation:
     
∆x2 ∆x2 ∆x1
ln ̺ = ln = ln − ln . (3)
∆x1 ∆x0 ∆x0

So the “scale state” V = ln(∆x2 /∆x1 ) appears like a “scale-velocity” or “zoom”,

in agreement with our principle that it should describe the state of resolution of
the coordinate system in the same way as the velocity describes its state of motion.
Just as one can speak only of the velocity of a system relative to another one, the
scale of a system can be defined only by its ratio to the scale of another system.
Eq. (3) may now be written in exactly the same form as Eq. (1):

V = V2 − V1 = (V2 − V0 ) − (V1 − V0 ). (4)

Concerning the problem of units, notice the difference of status between motion
and scale laws. While velocity is expressed in terms of a physical unit (e.g. m.s−1 ),
the scale state is expressed in terms of a mathematical unit, i.e. the adopted
logarithm base. Indeed the same behaviour is obtained (whatever base b is) using
the more general definition:
 
ln(∆x2 /∆x1 ) ∆x2
V = = logb . (5)
ln b ∆x1
It will be seen hereafter that this leads to a new kind of dimensional analysis.
Consider now a field ϕ which transforms under a dilatation q = ∆x/∆x′ follow-
ing a power law:
ϕ′ = ϕ q δ . (6)
In a renormalization group description, the power δ is identified with the anomalous
dimension of the field ϕ [9]. In a fractal interpretation of the same phenomenon, we

4
get δ = D −DT , where D is the fractal dimension and DT the topological dimension
[3, 11].
We are particularly interested here in the case where the “field” ϕ is space-time
itself. Let us briefly remind the present state of things concerning this approach [3].
Assume that we consider a system having first a de Broglie length λ0 = ~/p0 , and
that we perform successive measurements at given time intervals with a resolution
∆x in order to determine its velocity and then its average momentum and the length
of its trajectory. If ∆x ≫ λ0 , the momentum perturbation implied by Heisenberg’s
relation is ∆p ≪ p0 , so that the result will remain ≈ p0 , independent of scale. One
gets the usual classical trajectory whose length does not depend on resolution. On
the contrary when ∆x ≪ λ0 , ∆x . ∆p ≈ ~ implies that the measured momentum will
keep practically no trace of the initial one, i.e. ∆p ≫ p0 so that p = p0 + ∆p ≈ ∆p;
finally the momentum will be a direct function of resolution, p ≈ ~/∆x, and the
new de Broglie length of the system after the measurement becomes of the order
of ∆x. The length of the particle path now diverges as ∆x−1 [12, 13, 3]. This
means that the length L, integrated along the (fractal) path of a particle, diverges
for resolutions ∆x smaller than the de Broglie length (or time) λ as
λ
L = L0 , (7)
∆x
corresponding to the particular case δ = D − DT = 1 for D = 2 and DT = 1.
The same behaviour is found for the temporal coordinate around the de Broglie
time of the system [3]. This result is obtained when one takes into account not only
the transition to relativistic velocities, but also particle-antiparticle pair creations.
Owing to the fact that the whole set of virtual pairs contribute in the self-energy
of a particle (say, of the electron) and then in the nature of the particle itself,
and extending the Feynman-Wheeler-Stückelberg interpretation of antiparticles as
particles which run backward in time, we have suggested that, if one wants to
compute the full proper time T elapsed on the particle, one must add the proper
times elapsed on all the members of the virtual pairs to that of the “bare” particle.
Thus one finds a temporal coordinate diverging with energy, i.e. in an equivalent
way with the inverse of time resolution when ∆t < τ = ~/E as
τ
T = T0 . (8)
∆t
A similar result is obtained from localized solutions of the Dirac equation: from
the requirement that the solution should be localized into an interval ∆x ≈ c∆t ≈
~c/E, one may compute the rate of negative and positive energy solutions. One
finds P− /P+ ≈ (E − mc2 )/(E + mc2 ). Then considering that this set of positive
and negative solutions is nothing but the manifestation of a fractal trajectory which
runs backward in time for ∆t < τ leads to (8). This makes Lorentz covariant the
reinterpretation of the de Broglie scale as a universal space-time transition from
δ = 0 to an anomalous dimension δ = 1, since it applies to all four space-time
coordinates. In terms of the renormalization group, the de Broglie scale may be
identified as the correlation length of space-time.
Keeping all these results in mind, let us write Eq. (6) in a linear form by passing
once again to a logarithmic representation:
 ′    
ϕ ϕ ∆x
ln = ln + δ × ln , (9)
ϕ0 ϕ0 ∆x′

5
this being assumed to hold when ∆x ≪ λ. Our comparison with motion relativity
may then be pursued. The Galilean transform between two coordinate systems
reads
x′ = x + vt, (10)
t′ = t. (11)
We may now get a consistent description in which, as conjectured, resolution
acts as a “scale-velocity”, while the anomalous dimension (i.e. here the fractal
dimension minus 1) plays the role of a “scale-time”. Indeed, setting
 
ϕ
X = ln , (12)
ϕ0

from the linear relation  

λ
X = X0 + δ × ln , (13)
∆x
we may define the state of scale V as
 
λ d(ln ϕ) dX
V = ln = = , (14)
∆x dδ dδ

in the same way as the velocity of an object is defined as u = dx/dt in motion

relativity. Note the different approach with respect to the usual definition for the
fractal dimension δ = ∂ ln ϕ/∂ ln(λ/∆x).2 The scale law (13) is the equivalent for
scale of free motion at constant velocity, which is at the basis of the definition of
inertial motion. Likewise we suggest that a coordinate supersystem [3] (i.e. defined
by its states of motion and of scale) into which Eq. (13) holds, may be called
“scale-inertial”, and that we may set a principle of (special) scale relativity, which
states that the laws of nature are identical into all scale-inertial supersystems of
coordinates.
The anomalous dimension δ is assumed to be invariant (e.g. for space-time
coordinates we find the universal value δ = 1, itself coming from the universality
of the Heisenberg relations: in that case, δ is a constant), as time is invariant
in Galilean relativity. This is translated by the equations of the Galilean “scale-
inertial” transformation
X ′ = X + V δ, (15)
δ ′ = δ. (16)
In such a Galilean frame, the law of composition of scale states is the direct sum

W = U + V, (17)

which corresponds to the direct product ∆x′′ /∆x = (∆x′′ /∆x′ ) . (∆x′ /∆x) for
resolutions. Finally, with the three equations (15)-(17), we have put the scale
relativity problem in exactly the same mathematical form (Galileo group) as that
of motion relativity in classical mechanics.
But one should also keep in mind that the hereabove “inertial scaling” holds
only under some upper cut-off λ, contrarily to the motion case where it is universal.
2 More precisely the fractal dimension is DF = DT + δ, where DT is the topological dimension.

6
This may be expressed by writing, instead of (13), a formula including a transition
from scale dependence to scale independence, such as
"  2 #1/2
λ
X = X0 + δ × ln 1 + . (18)
∆x

More generally one may introduce a parameter k which characterizes the speed of
transition and replace V in (15) by Z, defined as
"  k #
1 λ
Z = ln 1 + = ln(1 + ekV )1/k . (19)
k ∆x

Then for ∆x ≫ λ, Z = 0 and for ∆x ≪ λ, Z = V . For k small, the transition

between these two regions is slow, while it is sudden (singular point) for k → ∞.
Strictly this description of the transition is only a model, since its details depend
on the physical system considered: for example, in many situations the transition
may imply exponentially decreasing “Yukawa-like” terms.

4 A New Derivation of the Lorentz Transforma-

tion
As remarked by Levy-Leblond [2], very little freedom is allowed for the choice of
a relativity group, so that the Poincaré group is an almost unique solution to the
problem [14]. In his original paper, Einstein derived the Lorentz transformation
from the (sometimes implicit) successive assumptions of (i) linearity; (ii) the invari-
ance of c, the light velocity in vacuum; (iii) the existence of a composition law; (iv)
the existence of a neutral element; and (v) reflection invariance.
But one may demonstrate that the postulate of the invariance of some abso-
lute velocity is not necessary for the construction of the special theory of relativity.
Indeed it was shown by Levy-Leblond [2] that the Lorentz transformation may be
obtained through six successive constraints: {1} homogeneity of space-time (trans-
lated by linearity of the transformation of coordinates), {2} isotropy of space-time
(translated by reflection invariance), {3} group structure (i.e. {3.1} existence of
a neutral element, {3.2} of an inverse transformation and {3.3} of a composition
law yielding a new transformation which is a member of the group, viz. which is
internal) and {4} the causality condition. The last group axiom, associativity, is in
fact straightforward in this case and leads to no constraint.
Actually this set of hypotheses is still overdetermined to derive the Lorentz trans-
formation. We shall indeed demonstrate hereafter that the Lorentz transformation
may be obtained from the only assumptions of {a} linearity; {b} internal composi-
tion law and {c} reflection invariance. All the other assumptions, in particular the
postulate of the existence of an inverse transformation which is a member of the
group, may be derived as consequences of these purely mathematical constraints.
The importance of this result, especially concerning scale relativity, is that we do
not have to postulate a full group law in order to get the Lorentz behavior: the
hypothesis of a semi-group structure is sufficient.

7
 Let us start from a linear transformation of coordinates:

x′ = a(v)x − b(v)t, (20)

t′ = α(v)t − β(v)x. (21)

In these equations and in the whole section, the coordinates x and t do not denote
a priori lengths and times, but may refer to any kind of variables having the mathe-
matical properties considered. Equation (20) may be written as x′ = a(v)[x−(b/a)t].
But we may define the “velocity” v as v = b/a, so that, without any loss of gener-
ality, linearity alone leads to the general form

x′ = γ(v) [x − vt], (22)

t′ = γ(v) [A(v)t − B(v)x], (23)

where γ(v) = a(v), and A and B are new functions of v. Let us now perform two
successive transformations of the form (22,23):

x′ = γ(u) [x − ut], (24)

t′ = γ(u) [A(u)t − B(u)x], (25)

′′ ′ ′
x = γ(v) [x − vt ], (26)
t′′ = γ(v) [A(v)t′ − B(v)x′ ]. (27)
This results in the transformation
 
′′ u + A(u)v
x = γ(u)γ(v) [1 + B(u)v] x − t , (28)
1 + B(u)v
 
′′ A(v)B(u) + B(v)
t = γ(u)γ(v) [A(u)A(v) + B(v)u] t − x . (29)
A(u)A(v) + B(v)u
Then the principle of relativity tells us that the composed transformation (28,29)
keeps the same form as the initial one (22, 23), in terms of a composed velocity w
given by the factor of t in (28). We get four conditions:

u + A(u)v
w= , (30)
1 + B(u)v

γ(w) = γ(u)γ(v)[1 + B(u)v], (31)

γ(w)A(w) = γ(u)γ(v)[A(u)A(v) + B(v)u], (32)
B(w) A(v)B(u) + B(v)
= . (33)
A(w) A(u)A(v) + B(v)u
Our third postulate is reflection invariance. It reflects the fact that the choice
of the orientation of the x (and x′ ) axis is completely arbitrary, and should be
indistinguishable from the alternative choice (−x, −x′ ). With this new choice, the
transformation (24,25) becomes {−x′ = γ(u′ )(−x−u′ t), t′ = γ(u′ )[A(u′ )t+B(u′ )x]}
in terms of the value u′ taken by the relative velocity in the new orientation. The

8
requirement that the two orientations be indistinguishable yields u′ = −u. This
leads to parity relations for the three unknown functions γ, A and B [2]:

γ(−v) = γ(v), A(−v) = A(v), B(−v) = −B(v). (34)

Combining Eqs. (30), (31) and (32) yields the relation

 
u + A(u)v A(u)A(v) + B(v)u
A = . (35)
1 + B(u)v 1 + B(u)v
Making v = 0 in this equation gives

A(u)[1 − A(0)] = uB(0). (36)

Making u = 0 yields only two solutions, A(0) = 0 or 1. The first case gives A(u) =
uB(0). B(0) 6= 0 is excluded by reflection invariance (34); then A(u) = 0. But (33)
becomes A(w) = B(w)u so that B(w) = 0 : this is a case of complete degeneration
to only one efficient variable since t′ = 0 ∀u, which can thus be excluded. We are
left with A(0) = 1, which implies B(0) = 0, and the existence of a neutral element
is demonstrated. Let us make now3 v = −u in (35) after accounting for (34), and
introduce a new even function F (u) = A(u) − 1, which verifies F (0) = 0. We obtain
 
1 + F (u)/2 uF (u)
2F (u) =F . (37)
1 − uB(u) 1 − uB(u)
We shall now use the fact that B and F are continuous functions and that B(0) = 0.
This implies that ∃η0 > 0 such that in the interval −η0 < u < η0 , 1−uB and 1+F/2
become bounded to k1 < 1 − uB(u) < k2 and k3 < 1 + F (u)/2 < k4 with k1 , k2 ,
k3 and k4 > 0. The bounds on 1 + F/2 and 1 − uB allow us to bring the problem
back to the equivalent equation, 2F (u) = F [uF (u)]. The continuity of F at u = 0
reads, owing to the fact that F (0) = 0:4 ∀ε, ∃η such that |u| < η ⇒ |(F (u)| < ε.
Start with some u0 < η yielding F (u0 ) = F0 = 2−n < ε. Then F (u0 F0 ) = 2F0 .
Set u1 = u0 F0 and iterate. After p iterations, one gets F (up ) = F [2p[(p−1)/2−n] u0 ] =
2p−n . In particular one gets after n iterations: F [2−n(n+1)/2 u0 ] = 1 if n is an integer.
(In the general case where n is not integer, one gets after Int[n] iterations a value
of F larger than 1/2 ). This is in contradiction with the continuity of F , since
un < u0 < η while F (un ) > ε. Then the only solution is F = 0 in a finite non null
interval around the origin, and from step to step whatever the value of u, so that

A(u) = 1. (38)

As a consequence (35) becomes B(u)v = B(v)u, a relation which finally constraints

the B function to be
B(v) = κv, (39)
where κ is a constant. At this stage of our demonstration, the law of transformation
of velocities is already fixed to the Einstein-Lorentz form:
u+v
w= , (40)
1 + κuv
3A misprint in the published version (v = u) has been corrected here.
4A misprint in the published version (F (u) = 0) has been corrected here.

9
and it is easy to verify that a full group law is verified, i.e. the existence of an identity
transformation and of an inverse transformation are ensured, without having been
presupposed. Consider now the γ factor. It verifies the condition
 
u+v
γ = γ(u)γ(v)(1 + κuv). (41)
1 + κuv

Let us consider the case u = −v. Equation (41) reads γ(0) = γ(v)γ(−v)(1 − κv 2 ).
For v = 0 it becomes γ(0) = [γ(0)]2 , implying γ(0) = 1, and we get
1
γ(v)γ(−v) = . (42)
1 − κv 2
The final step to the Lorentz transformation is straighforward from reflection in-
variance, which implies γ(v) = γ(−v) (Eq. 34) and fixes the γ factor in its Lorentz-
Einstein form:
1
γ(v) = √ . (43)
1 − κv 2
The case κ < 0 yields a non-ordered group (applying two successive positive veloci-
ties may yield a negative one), and we are left with the only two physical solutions,
the Galileo (κ = 0 ) and Lorentz (κ = c−2 > 0 ) groups. Three of their properties,
− existence of a neutral element, of an inverse element and commutativity (for one
space dimension) − have not been postulated, but deduced from our initial axioms.
Let us end this section by a brief but important comment. We have shown that,
once we have set the hypothesis of linearity, the Lorentz transformation may be ob-
tained through the only postulates of internal composition and reflection invariance.
Linearity is not a constraint by itself: indeed it corresponds to the simplest possible
choice (i.e. when searching for a transformation which would satisfy a given law,
one may first look for a linear one, and then look for non linearity only in case of
failure, or later as a generalization). With regard to the other two postulates, they
may be seen as a direct translation of the Galilean principle of relativity. Indeed
the hypothesis that the composed coordinate transformation (K → K ′′ ) and the
transformation in the reversed frame (−K → −K ′ ) must keep the same form as
the initial one (K → K ′ ) is nothing but an application of the Galilean principle of
relativity (“the laws of nature must keep the same form in different inertial reference
systems”) to the laws of coordinates transformation themselves, which are clearly
part of these laws to which the principle should apply. So the general solution to
the problem of inertial motion, without adding any postulate to the way it might
have been stated in the Galileo and Descartes epoch, is actually Einstein’s special
relativity, whose Galilean relativity is a special case (c = ∞ ).

5 Lorentzian Scale Transformation

In the preceding section, we have recalled that the general solution to the (special)
relativity problem is the Lorentz group. In the case of motion relativity the Lorentz
transformation for systems in inertial motion is now one of the most solid base of
physics. What about scale ?
We have argued in Secs. 1−3 that scale (resolution) also came under a relativity
theory. Set in a general way, the problem of scale transformation now consists

10
in looking for a two-variable transformation ln ϕ′ = f1 (ln ϕ, δ), δ ′ = f2 (ln ϕ, δ),
depending on one parameter, the scale state V = ln(λ/∆x).
Let us analyse how the mathematical axioms on which was founded the above
derivation of the Lorentz transformation are physically translated in the case of
scale. In the theory of motion relativity linearity may be derived from the homo-
geneity of space and time (which is itself an application of the principle of relativity
to positions and instants). In scale relativity, the things that play the role of lengths
and times are now respectively the logarithm of some field, ln ϕ, and the anomalous
dimension or fractal codimension δ (see Sec. 3). The uniformity of these variables
is not a priori straighforward, even though it is already assured in the scale laws
of present physics. But linearity, as already specified, may be inferred from a hy-
pothesis of simplicity. More precisely linearity is the simplest choice to make, and
so comes as a provisional specialization of the present theory. It is clear that a
generalization to nonlinear transformations must be considered in the future (we
have suggested that such an achievement would imply to use the tool of fractal
space-times) [3], but this departs from the frame of the present work.
The second axiom, the existence of an internal composition law, is a direct appli-
cation of the principle of relativity: there is no difference here between motion and
scale. Reflection invariance means that one may equally work with either ln(ϕ/ϕ0 )
or ln(ϕ0 /ϕ), to which would respectively correspond scale states ln(∆x/∆x0 ) and
ln(∆x0 /∆x); this is indeed straightforward. Finally the case κ < 0 is clearly also
excluded for scale, since when applying two successive dilatations we indeed expect
the final product not to be a contraction.
So from our result that the general solution to the linear relativity problem
is Lorentz, we conclude that the laws of scale transformation must also take a
Lorentzian form, instead of the Galilean form, which was up to now assumed to
be self-evident.
Let us now explicitly compare Lorentzian scale transformation to motion trans-
formations. While the composition of velocities follows an additive group law, the
composition of scales follows a multiplicative group law. It is easy to come back to
a multiplicative group by taking the logarithm of scale ratios, as shown in Sec. 3
(Galilean case, which is also the case of the standard renormalization group).
Start with the Einstein-Lorentz law of composition of “velocities”:
u+v
w= , (44)
1 + (uv/c2 )

where u, v, w are dimensioned quantities and c is an universal constant. This may

be written in a dimensionless way by setting U = u/c, V = v/c, W = w/c :

U +V
W = . (45)
1 + UV
Let us now write U , V and W , which are pure numbers, as logarithms of other
dimensionless quantities. This may be done into any base for the logarithms, say
K, by setting U = logK ν, V = logK ̺ , W = logK µ , i.e.
c
u= ln ν, (46)
ln K

11
and similar formulas for v and w. So (45) now becomes
logK ̺ + logK ν
logK µ = . (47)
1 + logK ̺ × logK ν
We may now divide both members of this equation by ln K and we get
ln ̺ + ln ν
ln µ = . (48)
1 + (ln ̺ ln ν/ ln2 K)

This formula is formally identical to the initial one, Eq. (44) [and to the general
structure (40)], with the difference that ln K is itself a pure number, while c was a
dimensioned quantity. Now identifying µ, ν, ̺ and K to scale ratios, we see that
(48) becomes the scale-relativistic generalization of the usual law of dilatation: this
means that the successive application of two dilatations ̺ and ν now yields the
dilatation µ instead of the usual product ̺ ν.
We get a new law for the transformation of the field ϕ, which generalizes (9):
 ′
ϕ logK (ϕ/ϕ0 ) + δ × logK ̺
logK = . (49)
ϕ0 (1 − log2K ̺)1/2
The anomalous dimension, which was previously invariant, becomes now a function
of the resolution and of the field:
δ + logK ̺ logK (ϕ/ϕ0 )
δ′ = . (50)
(1 − log2K ̺)1/2
However these laws still cannot be considered as the definitive laws of scale relativity,
since they do not incorporate the classical / quantum transition. This is done in
the following.

6 Scale Relativity Broken

As already specified, scale relativity5 , contrarily to motion relativity, is not a uni-
versal principle of nature. The fact that scales (or resolutions) can be defined only
by their ratios is indeed universal, but this is of no consequence in the classical
domain (∆x ≫ λdB ). There, resolution reduces to precision, and improving the
precision of measurements improves the precision of results, but does not change
the physics. The situation changes in the quantum and quantum-relativistic do-
mains, the transition to which corresponds to the de Broglie length and time (see
Ref. 3 and Sec. 3).
Hence scale relativity must be a broken principle above the de Broglie scales
λdB = (~/mv)(1 − v 2 /c2 )1/2 and τdB = (~/mc2 )(1 − v 2 /c2 )1/2 . In order to simplify
the argument, let us look at the high energy degenerated case, where only one space-
time variable may be considered, say ∆x ≈ c∆t, so that cτdB (= ~c/E) becomes
equal to the Compton length λ0 = ~/mc in the rest frame of a system of mass m.
Let us explicitely introduce this particular scale in the composition law (48).
5 In this section, “scale relativity” stands for “special scale relativity”. It is the special scale

relativity symmetry (involving the log-Lorentz scale transformations of previous section) whose
breaking toward Galilean-like scale relativity is analysed here.

12
 We start from the scale λ0 and first apply a contraction which leads to a new
scale λ, then apply another contraction by a factor ̺ leading to a final scale λ′ . The
“Galilean” character of λ0 allows us to take it as reference for all scale ratios (with
the exception of ̺ which relates λ to λ′ ). We thus set in (48) µ = λ′ /λ0 , ν = λ/λ0
and K = Λ/λ0 (this explicit writing of dilatations as scale ratios introduces a
Galilean structure), and the composition law now takes the form
 
 ′
λ ln λλ0 + ln ̺
ln =    . (51)
λ0 1 + ln ̺ ln λ / ln2 Λ
λ0 λ0

We verify that the dilatation which relates λ0 to any scale λ remains equal to
their ratio, as in the classical case, while it is no more true of two scales both
different from λ0 . In fact (51) may be inverted and understood as the function
̺(λ, λ′ ; λ0 ) which yields the dilatation allowing to go from one scale λ to another
scale λ′ [see Eq. (53)]. This dilatation factor now depends on the initial de Broglie
scale, λ0 .
But consider now the behaviour of the particular length Λ. Assume that we start
with this length, i.e. λ = Λ and that we apply to it the dilatation or contraction
̺. From (51), we find that this results into a length given by ln(λ′ /λ0 ) = ln(Λ/λ0 ),
i.e. λ′ = Λ whatever the value of the de Broglie scale λ0 . Starting from any scale
larger than Λ, and applying any finite contraction, we get a scale larger than Λ.
The scale Λ can be the result only of infinite contraction or of an infinite product of
contractions, i.e. it plays the same role as the zero point of the previous theory. In
terms of the renormalization group theory, it is a fixed point for space-time itself.
Hence the principle of relativity, once applied to scales, combined with the ex-
istence of the de Einstein-Broglie transition, leads to the existence of an absolute
length in nature, which is invariant under dilatations and contractions. Motion
relativity immediately ensures that this will be also true for time, and that an
invariant time interval T = Λ/c exists in nature. A particular case of scale trans-
formations is the Lorentz length contraction and time dilatation: as a consequence
it is straighforward that Λ and Λ/c will be also invariant under a Lorentz transfor-
mation, i.e. independent of the relative velocity of the reference system in which
they are observed.
One might be disturbed by the fact that K is not an absolute constant, contrar-
ily to the structure expected from a pure special relativity theory. However, once λ0
is fixed (and it is fixed by the state of motion of the system considered, since the de
Broglie length and time are Lorentz-covariant by construction), λ0 /Λ is a constant:
scale relativity relies on motion relativity. Conversely it is rather satisfactory that,
in the same way as motion relativity led to the existence of an absolute and un-
exceedable velocity, scale relativity leads to the existence of an absolute, invariant
limit for all lengths and times. The final point to be elucidated is the nature of Λ.
We suggest in the following that it is nothing but the Planck scale.

13
7 On the Absolute6 Character of the Planck Scale
The Planck length already plays a very special role in physics: it is the characteristic
scale for which all forces of nature are expected to become equivalent, while the
concept of a space-time continuum seems to lose its physical meaning for smaller
resolutions. It has been proposed [15, 16] that the topology of space-time may
become extremely complicated (foamlike) at that scale, the continuum itself being
broken.
Even though a bundle of physical arguments makes clear that the Planck scale
must play a central role in microphysics, all the previous approaches to the problem
were worked out in a frame where the scaling laws themselves were not questioned,
i.e. in which it was considered evident that applying a dilatation q to a scale ∆x
yields a new scale ∆x′ = q × ∆x. This is reminiscent of classical Galilean physics
in which it seemed also self-evident that throwing an object with a velocity v with
respect to a body moving with velocity u relative to the ground finally yields a
velocity w = u + v.
Let us now consider the question from the point of view, adopted here, of scale
relativity. The Planck length scale (~G/c3 )1/2 is particular in that its expression de-
pends on no particular physical object, but only on the three fundamental constants
of physics, G, ~ and c. While we have insisted at the beginning of this paper on the
relativity of all scales, the Planck scale is the only one which is in fact absolute in
its definition, i.e. independent from particular physical bodies or systems.
If we admit that the three constants G, ~ and c are indeed universal and un-
varying, even at the time and length scales of the Universe (~ is known to vary by
less than 4 × 10−13 per year and G by less than 10−11 per year, i.e. respectively
less than ≈ 0.4 % and 10 % over the age of the Universe) [17], one may be upset
by the fact that, in present physics, a “Planck rod” (~G/c3 )1/2 , when submitted to
a dilatation q becomes q(~G/c3 )1/2 in spite of its universality, and when observed
from a reference system inpwhich it moves with velocity v, is submitted to Lorentz
contraction and becomes (~G/c3 )(1 − v 2 /c2 ). Even if it is admitted that physics
may drastically change when the Planck scale is crossed, it is still admitted that
scales smaller than the Planck scale do exist in nature.
We take here a radically different position: based on the absolute character of the
definition of the Planck scale, we suggest identifying it as the invariant scale Λ, which
was derived above from the application to scale of the principle of relativity. The
Planck length becomes a universal scale which remains invariant upon dilatations.
It now plays for scale the same part as the velocity of light already plays for motion.
The concept of a resolution smaller than the Planck length also loses any physical
meaning, since the Planck length is now a limit which can not be exceeded (toward
lower resolutions).
Strictly Λ could be identified with the Planck length, times any pure and con-
stant number, but this would destroy the formal simplicity of the construction (only
the exchanges G → 2G and ~ → h remain uncertain, but ~ is preferable to h since
6 The adjective “absolute”, used here and throughout the text to characterize the properties

of the Planck length-scale in the new scale-relativistic interpretation, is incorrect. It should be

replaced by “universal” and/or “invariant”. Indeed the Planck scale, as any other physical object
or concept, is not defined in an absolute way, but relatively to other concepts (here, fundamental
constants). This is also clear from its value, which depends on units.

14
the actual transition lengths around which physics changes are indeed the reduced
Compton length ~/mc rather than h/mc); so we set, while waiting for possible
future experimental verification:
r
~G
Λ= . (52)
c3
It should be noted that, taken together, the three fundamental constants G, ~
and c do nothing but fix the arbitrary part of our units of length, time and mass.
Motion relativity has supplied us with a conceptual frame in which lengths and times
are logically related: one now deduces length units from time units and c fixed. (It
would be most consistent in fact to fix c = 1 and definitively measure lengths e.g. in
nanoseconds). Scale relativity, if confirmed by experiments, will achieve the same
for times themselves (at least in principle, since the bad present precision on G
prevents one from doing this explicitly for the moment; a precise determination of
the constant of gravitation now becomes an urgent task). Setting the Planck time
Λ/c = 1, all length and time intervals in nature become dimensionless real numbers
larger than one. In such a system, one would get G = 1/~. The final step to setting
also the Planck mass to 1 demands the determination of the ratios of the (low
energy) masses of all elementary particles over the Planck mass, and, if one wants
to be completely consistent, the understanding of the values of the three remaining
coupling constants at a given scale. It will be shown hereafter that scale relativity
allows to take some steps towards the achievement of this grand program.
Before going on, let us write the complete new transformation, in the case where
all scales considered are smaller than the de Broglie scale λ0 (assumed fixed) of the
system. Let ̺ be the dilatation which allows to go from ∆x to ∆x′ (∆x ≤ λ0 7
and ∆x′ ≤ λ0 ); let ϕ = ϕ(∆x) be some scale-dependent field and δ = δ(∆x) its
anomalous dimension; we set ϕ′ = ϕ(∆x′ ), δ ′ = δ(∆x′ ) and define an arbitrary
reference value for the field, ϕ0 ; the new transformations for the dilatation, the
field and the anomalous dimension read, in terms of the Planck scale Λ,
 ′
ln ∆x ∆x
ln ̺ = λ0

λ0

2 λ0 ,

(53)
1 − ln ∆x ′ ln ∆x / ln Λ
   
ϕ ϕ 2 λ0


 ′
ϕ ln ϕ0 + δ ln ̺ δ + ln ̺ ln ϕ0 / ln Λ
′
ln = q
, δ =
. (54)
ϕ0
q
1 − ln2 ̺/ ln2 λ0 Λ 1 − ln2 ̺/ ln2 λ0 Λ

Letting Λ → 0 yields back the standard (“Galilean”) scaling transformation ϕ′ =

δ
ϕ ̺ and δ ′ = δ. The standard transformation is also obtained as an approximation
in the limit ln ̺ ≪ ln(λ0 /Λ).
Equations (53, 54) hold only in the quantum case (∆x ≤ λ0 and ∆x′ ≤ λ0 ).
Going to a scale larger than the de Broglie scale leads to scale independence: δ = 0,
ϕ independent of scale and “Galilean” dilatation law ̺ = ∆x′ /∆x. As already
noted in Sec. 3, a precise description of this transition from scale dependence to
scale independence depends on the particular physical system considered. A useful
model (see Sec. 3, Eq. 19) consists in replacing, in Eqs. (53, 54), ln(λ0 /∆x) by
7A misprint in the published version (∆x ≤ ∆x0 ) has been corrected here and hereafter.

15
(1/k) ln[1 + (λ0 /∆x)k ] (and similar changes for all scales refered to λ0 ), where k is
a parameter which allows one to describe the steepness of the transition. Indeed
for fixed λ0 and ∆x ≪ λ0 , one gets (1/k) ln[1 + (λ0 /∆x)k ] ≈ ln(λ0 /∆x), while for
∆x ≫ λ0 , (1/k) ln[1 + (λ0 /∆x)k ] ≈ 0. To be complete, one should also replace δ
by some function ∆(∆x), with ∆ = δ for ∆x ≪ λ0 and ∆ = 0 for ∆x ≫ λ0 .
We also recall that Eqs. (53, 54) apply in the first place to the case where
ϕ represents either the length measured along a quantum particle trajectory (non
relativistic case), or its integrated proper time (relativistic case + reinterpretation of
particle-antiparticle pairs as part of a fractal trajectory running backward in time),
and more generally any of the four space-time coordinates (once integrated along
the fractal path). Then ∆x may represent any of the four coordinates’ resolution,
∆x i, i = 0 to 3, and more generally of some combination of them, in particular the
resolution of the classical invariant, ∆s.
To conclude this section, let us examine some of the consequences of Eqs. (53,54).
We have recalled in Sec. 3 that, in standard quantum mechanics, the (integrated)
coordinates were divergent as ϕ = ϕ0 (λ0 /∆x) when ∆x < λ0 , as a consequence of
Heisenberg’s relations. Equation (54) tells us that they will now tend to infinity,
not for ∆x → 0, but for ∆x → Λ. The standard relation corresponds to a constant
value δ = 1, i.e. to a fractal dimension D = 1 + δ = 2. In the new theory, the
anomalous dimension δ is now subject to scale relativistic effects: its expression
now implies the scale-relativistic equivalent of Lorentz so-called “γ-factors”. The
value δ = 1 holds only at ∆x = λ0 , then increases toward lower scales. Consider
the particular simplifying choice ϕ = ϕ0 = ϕ(λ0 ) : with δ(λ0 ) = 1, we get
1
δ(∆x) = q
, (55)
1 − ln2 λ0
/ ln2 λ0


∆x Λ

i.e. the scale γ-factor is directly given by the anomalous dimension at the new scale,
δ(∆x), so that one gets the new law
 δ(∆x)
′ λ0
ϕ = ϕ0 , (56)
∆x

to be compared to the standard one, ϕ′ = ϕ0 (λ0 /∆x)δ(λ0 ) .

A second scale-relativistic effect may occur even for δ factors close to 1. As
may be seen from Eq. (54), δ may increase with length ϕ even in the “non scale-
relativistic” case ln(λ0 /∆x) ≪ ln(λ0 /Λ) (i.e. the scale-relativistic equivalent of the
motion-relativistic relation v ≪ c ). Can this behaviour give rise to inconsistencies?
One expect to get a large effect (i.e. δ ′ − δ ≈ 1 ) for very large distances ϕ = L,
such that ln(L/λ0 ) ≈ ln2 (λ0 /Λ)/ ln(λ0 /∆x).8 This actually corresponds to large
macroscopic distances for which a quantum description has become inadequate for
long (e.g., even with the extreme choice ∆x = Λ, λ0 ≈ 10−13 m yields L ≈ 109
m since Λ ≈ 10−35 m),9 so that we expect this behaviour not to contradict well
established physical results.
Another unexpected effect is the decrease of the anomalous dimension when ϕ →
0. However this decrease is not inconsistent with the theory, since the singularity
8A misprint in the published version (the second log term was to the square) is corrected here.
9 Instead of L ≈ 1011 m as given in the published version.

16
δ = 0 would be reached for ϕ ≪ Λ, which is now excluded by the theory itself. Note
that these problems are avoided by the hereabove particular choice ϕ = ϕ0 for the
reference point of the field ϕ.
The new structures found in scale relativity imply profound changes of many
other fundamental basic relations. Indeed the requirement that any space and time
resolution in nature be larger than the Planck length and time imply that this
should be required also of any length and time interval. This is a radical change of
the nature of space-time itself, which is expected to have consequences for the whole
of physics. In particular, the scale transformations (53,54) rely on the concept of
de Broglie length and time. But it is immediately clear that the theory cannot be
self-consistent if their usual definition is kept. Indeed for masses larger than the
Planck mass, the Compton length (i.e. c times the rest frame de Broglie time) would
become smaller than Λ, which is a now forbidden behaviour. The next section is
devoted to this crucial problem of the mass (more generally energy-momentum)
scale in the new theory of scale-relativity.

8 Scale-Relativistic Mechanics
8.1 A new invariant
Let us attempt to clear up the problem which is now set before us. In the classical,
then relativistic theory of motion, the laws of mechanics set the relations between
energy-momentum and the essential variable in the inertial case, i.e., velocity. Our
claim here is that, in the quantum domain, the classical concept of motion becomes
inoperative, to the advantage of the concept of scale, and velocity is disqualified as
an essential variable, its place being taken by resolution. So it becomes logical to
expect the energy-momentum / velocity relations to be replaced in the quantum
theory by energy-momentum / scale relations: and this is precisely what the de
Broglie (< p > . λ = ~ ) and Heisenberg (σp . σx ≥ ~/2 ) relations are. The way
by which one may obtain these relations as consequences of the principle of scale
relativity and then generalize them in the Lorentzian case is clearly to construct a
scale relativistic mechanics.
In the frame of standard quantum mechanics, we have recalled that the Heisen-
berg relations ∆p . ∆x ≈ ~ and ∆E . ∆t ≈ ~ can be reinterpreted in terms of some
internal length L which becomes scale dependent (fractal) as L ≈ L0 (λ/∆x)δ for
∆x < λ ( λ being the de Broglie length ~/ < p > ), and of some internal time T
such that T ≈ T0 (τ /∆t)δ for ∆t < τ (τ being the de Broglie time ~/ < E > ), with
δ = 1 in both cases. Let us consider the one-dimensional case, with ϕ denoting
either L or T in the following. If we assume the classical coordinate system to be
fixed (origin, axes orientation and state of motion), d ln ϕ and dδ are independent
of each other and invariant in this “Galilean” frame.
Consider now the frame of scale relativity. The variables ln ϕ and δ respectively
play in scale relativity the roles played by position x and time t in motion relativity.
It is well known that a formulation of special (motion) relativity equivalent to the
requirement of Lorentz covariance is the requirement of invariance of the Minkowski
metric element ds2 = c2 dt2 − dx2 (we remain in the one-space-dimension case in
order to simplify the argument). In the same way, neither dδ nor d ln ϕ remains
invariant in scale relativity. The new scale invariant is (for λ0 fixed and resolution

17
∆x < λ0 ):
dϕ2
 
2 2 λ0
dσ = ln dδ 2 − . (57)
Λ ϕ2
Under this form a physical interpretation of the new invariant is difficult, since δ
is not a directly measurable quantity. However the Minkowski invariant may also be
expressed in terms of velocity as ds2 = c2 dt2 (1 − v 2 /c2 ) = (c2 /v 2 − 1)dx2 . In scale
relativity, the state of motion v is replaced by the state of scale ln(λ0 /∆x), so that
the new invariant may be expressed in terms of quantities which are measurable (at
least in principle): the length (or time) ϕ and the measurement resolution ∆x:
 2  2
ln (λ0 /Λ) dϕ
dσ 2 = − 1 . (58)
ln2 (λ0 /∆x) ϕ2

This result confirms our initial conjecture [3] that the space-time of microphysics is
of a radically new nature compared to the classical space-time: a proper description
of it implies an explicit intervention of resolution.
Let us proceed further in our construction of a relativistic “scale mechanics”.
The experience of special (motion) relativity may still be followed advantageously
[18]. We first assume that scale physical laws emerge from a least action principle.
Once the state of motion fixed, we expect the action to be the integral over dδ
of some Lagrange function L = L(ln ϕ, ln(λ0 /λ), δ), (to be compared with L =
L(x, v, t) in motion relativity) and its differential L dδ to be given, up to some
multiplicative constant, by the invariant dσ [18]. If we denote as V = ln(λ0 /λ) the
scale state, “conservative” quantities (prime integrals) ∂L/∂V and V ∂L/∂V − L
will emerge from the uniformity of ln ϕ and δ respectively. But note that these
quantities are not “conservative” here in terms of time independence: the uniformity
of δ implies that L does not depend explicitely on δ, so that here “conservative”
means that these quantities do not depend explicitely on the anomalous dimension
δ, which plays for scale the structural role played by time for motion.

8.2 Generalized de Broglie and Compton relations

Let us consider first the uniformity of ln ϕ. It implies the existence of a conservative
quantity, a “scale momentum” P, which is a function of the scale state ln(λ0 /λ) :

ln λλ0


P(λ) = µ q , (59)
ln2 (λ0 /λ)
1 − ln 2 (λ /Λ)
0

where µ is the constant, to be later determined, which comes from the fact that the
action and the metrics invariant are equal only to some proportionality factor (this
factor is equal to −mc in motion relativity [18]). A similar relation is obtained for
the time variable, in terms of de Broglie (τ0 ) and Planck (Λ/c) times:

ln ττ0


E(τ ) = µ q 2 (τ /τ )
. (60)
1 − lnln2 (cτ0
0 /Λ)

These two relations are the scale-relativistic

p equivalent of the motion relativistic
equation for momentum, p = mv/ 1 − v 2 /c2 .

18
 In order to know the meaning of this result, one first notes that physics must be
invariant under the choice of the logarithm base. Then the form of (59, 60) implies
that P is itself a logarithm of some dimensionless quantity. Now (59) has been
obtained from the uniformity of a space variable, from which the usual (motion)
momentum also derives as conservative quantity in classical mechanics, and (60)
from the uniformity of time, from which the concept of conservative energy derives
in classical mechanics. We then suggest that P is related to the classical momentum
(case of a space variable), leading to write P = ln(p/p0 ), and E to the classical
energy (case of time variable), so that E = ln(E/E0 ).
Consider now the limit Λ → ∞ : this limit should give us back standard quan-
tum mechanics, i.e. (59, 60) must be identifiable with already-known equations of
quantum mechanics. Indeed (59) becomes p/p0 = (λ0 /λ)µ for the space variable and
(60) becomes E/E0 = (τ0 /τ )µ for the time variable. We recognize in these equa-
tions the two Einstein-de Broglie relations, pλ = p0 λ0 = constant and Eτ = E0 τ0 =
constant, provided the constant µ is definitively set to the value µ = 1. Since λ and τ
are themselves defined up to some multiplicative factor, we may choose them in such
a way that the universal constants p0 λ0 and E0 τ0 are the same. This defines the
reduced Planck constant ~ (or h with a different choice for the remaining arbitrary
scale factor) and we get:
pλ = p0 λ0 = ~, (61)
Eτ = E0 τ0 = ~. (62)
Consider now the Lorentzian case where Λ 6= 0 : this leads us to infer that
the full equations (59, 60), in which must be set P = ln(p/p0 ), E = ln(E/E0 ) and
µ = 1, are the scale-relativistic generalization of the de Broglie relations which we
were seeking. They indeed own the expected property that momentum and energy
now tend to infinity when the generalized de Broglie length and time tend to the
Planck length and time. We may sum up these results by a comparison between
the four structures of Galilean / Lorentzian, motion / scale relativity: Galilean
motion relativity yields the momentum/velocity Descartes relation p = mv, whose
scale equivalent is the momentum/wavelength de Broglie relation P = µV , [with
P = ln(p/p0 ), V = ln(λp 0 /λ) and µ = 1 it reads p/p0 = λ0 /λ ]; Einstein motion
relativity yields p = mv/ 1 − v 2 /c2 while scale relativity generalizes the de Broglie
relation as:
 1/√1−V 2 /C02
λ0
p = p0 , (63)
λ
where we have set V = ln(λ0 /λ) and C0 = ln(λ0 /Λ), and where we have given up
the logarithmic form adopted in Eq. (59).
Actually, this result applies only to high energy physics (∆t < ~/me c2 , ∆x ≈
c∆t < ~/me c , where me is the electron mass). In this case λ may be identified
with the Compton length of a given system, (or more generally, to remain Lorentz
covariant, with cτdB , i.e., the Lorentz contracted Compton length). Equation (63)
provides us with a new relation between the momentum-energy scale and the length
scale. In standard high energy quantum mechanics, the length and mass scales
are directly inverse: there is an inverse correspondence between any mass scale m
(equivalently an energy mc2 ) and a length r through the relation mr ≈ ~/c. So the
asymptotic behavior of the various quantum theories, which is so crucial for their

19
ultimate understanding, corresponds to both r → 0 and p → ∞. In scale relativity
it now corresponds to r → Λ and p → ∞ : the experimental consequences of this
new length-momentum relation will be considered in Sec. 9.
The fact that we admit that (63) applies to the de Broglie lengths themselves,
while the de Broglie scale is used as “input” in the scale transformations, implies
some difficulty of interpretation. We are now comparing de Broglie lengths one with
another, rather than assuming λ0 fixed and then measuring the system at some res-
olution ∆x. So we need one new universal scale which will serve as reference for all
other scales. The Compton length of the electron clearly plays this role in micro-
physics. It corresponds to the less massive of all elementary particles and thus to
the transition from non-(motion-)relativistic to relativistic quantum behaviour. At
this scale, all velocities become relativistic, and the concept of well-defined position
loses its physical meaning, since the first occurrence of particle-antiparticle pair
creation-annihilation starts the domain of elementary particle physics: it is from
the electron Compton length λe onward that the fundamental coupling constants and
the particle rest masses begin to vary. Thus the electron Compton length clearly
plays the role of a zero point for the whole domain of relativistic quantum fields,
so that we can write (63) in terms of a new relation between mass m > me and
Compton length λc < λe :
 

m
 ln λλec
ln = q . (64)
me 2
1 − ln (λe /λc )
ln2 (λe /Λ)

This introduces the fundamental number Ce = ln(λe /Λ) = ln(mP /me ) [with Ce =
51.52797(7) from the presently known values of ~, c and G [19], which serve to define
the Planck mass mP = (~c/G)1/2 ; the number into parentheses following numerical
results is by convention the error on the last digits]. It is straightforward to verify
on (64) that now the Compton length is limited by Λ when the mass scale tends
to infinity. Applying a Lorentz transformation to (64), one finds, as expected, that
(63) is only an asymptotic formula (p ≫ me c ), so that the strict relation pλdB = ~
remains true in the non-relativistic domain, λdB > λe .

8.3 Generalized Heisenberg relations

The problem of a generalization of the Heisenberg relations is set in a somewhat
different way from the de Broglie and Compton problems since we now deal with in-
equalities rather than with strict equalities. However a similar behaviour is expected
for them, i.e. we expect σp to tend to infinity when σx now tends to the Planck scale.
Actually a full treatment of the problem would imply a proper generalization of the
whole structure of quantum mechanics: this huge technical problem goes outside
the scope of this paper. Our hope is that the setting of the self-consistency of our
generalization of the two basic quantum relations, de Broglie’s and Heisenberg’s,
will ensure the possibility of a self-consistent generalization of the whole formalism
of quantum mechanics. Let us briefly consider a possible way in that direction.
We have already shown in Ref. [3] that, starting from the hypothesis that the
quantum space-time is such that the lines which define the possible particle trajecto-
ries have a fractal dimension D = 1+δ, one gets generalized Heisenberg inequalities:
(σp /p0 )(σx /λ0 )δ ≥ 1, this holding for all four space-time coordinates. The usual

20
Heisenberg relation is recovered, as expected, for the particular value δ = 1. This
generalization, which was purely formal in Ref. [3], is endowed with physical mean-
ing now that we have introduced a fractal dimension (equivalently, an anomalous
dimension in a renormalization group approach) which is allowed to vary. Note
that such a relation is not incompatible with Heisenberg’s: the usual Heisenberg
inequality remains true even for δ > 1, though the inequality becomes stronger as
δ increases, i.e. σp . σx ≫ ~. Since we have demonstrated hereabove that, after a
dilatation (or contraction), the scale-γ-factor is precisely equal to the anomalous
dimension at the new scale, the generalized Heisenberg relations finally keep, as
expected, a form similar to de Broglie’s in the new theory (for σx ≤ λ0 ):
 
 
σp ln σλx0
ln ≥r , (65)
p0 λ
ln2 ( 0 )
1 − 2 σλx0
ln ( Λ )

with an equivalent expression holding for time and energy.

8.4 On the nature of charge

Let us come back to our construction of a scale-relativistic mechanics and build
the “conservative” quantity independent of δ which comes from the homogeneity
of the anomalous dimension itself, i.e., V ∂L/∂V − L. This is a completely new
quantity10 which had no theoretical existence in Galilean scale relativity, since there
δ either was itself undefined (classical case) or was an absolute constant (standard
quantum case). This is the scale equivalent (from the viewpointpof the mathematical
structure) of the relativistic expression for energy, E = mc2 / 1 − v 2 /c2 . It reads

ln2 λΛ0


E= r λ
. (66)
ln2 ( λ0 )
1 − 2 λ0
ln ( Λ )

This expression should involve a priori an arbitrary multiplicative factor µ, but this
factor was already set to 1 by the identification of the “scale-momentum” to the
logarithm of a ratio of motion momenta. Once again the requirement of invariance
under the logarithm form of this equation leads us to set E = ln2 (E/E0 ). Then we
get:
 δ1/2
E λ0
= , (67)
E0 Λ
q
where δ = 1/ 1 − ln2 (λ0 /λ)/ ln2 (λ0 /Λ). The remarkable result, which is reminis-
cent of what happened in motion special relativity, is the emergence of a non zero
value for this new physical quantity at large scale (δ = 1 ), i.e. a quantity which
must still exist in the classical scale-independent domain:
λ0 E0
E00 = . (68)
Λ
10 It has been named “complexergy” in subsequent works.

21
We do not know, a priori, what the dimensional equation of E is. Let us tentatively
express it in units of energy. It then seems logical to identify E0 as the rest energy
mc2 and λ0 as the Compton length ~/mc, so that we arrive at the conclusion that
any quantum system, in particular any particle, owns a physical property which is
in some way equivalent to an internal energy given by the Planck energy:
1/2
~c5

2
Eoo = mP c = . (69)
G

How can this be interpreted ? Let us consider the Newtonian gravitational force
between two bodies of masses m1 and m2 . It reads: Fg = Gm1 m2 /r2 . The Coulomb
force between two electric charges e1 and e2 reads: Fem = α~cZ1 Z2 /r2 in units
where Z1 and Z2 are dimensionless integers. But let us now write the Coulomb
force in a form√ such that √ the charges are expressed in units of masses. We get
Fem = G(Z1 α mP )(Z2 α mP )/r2 . A similar transformation may be (at least
formally) made for the weak and strong forces, with the fine structure constant α
replaced by the SU(2) and SU(3) coupling constants α2 and α3 . Under this form
we see that indeed the various charges in nature other than the gravitational one
correspond to an internal energy of order mP c2 : they would even be strictly equal
to the Planck energy provided the high energy common bare coupling constant was
equal to 1 [see Sec. 9 about the convergence of U(1), SU(2) and SU(3) couplings at
high energy].
This fact, often expressed in terms of the high value of the ratio of electric over
gravitational forces (≈ 4 × 1042 ), is a well known structure of present physics.
The new point here is the following: the mere existence of the electromagnetic field
presently relies, in its main lines, on experimental grounds [18] and does not seem
to be made necessary from fundamental principles. This is to be compared to the
present status of the gravitational field: the principle of relativity, once applied
to accelerated motion, leads to Einstein’s general relativity whose equations are
the simplest and most general equations which are invariant under continuous and
differentiable transformations of coordinates; they introduce space-time curvature as
a universal property of nature whose manifestations are what we call gravitation.
In this sense, one may say that the principle of (motion) relativity leads to the
demonstration of the unavoidable existence of gravitation in nature. We suggest
that the hereabove result is a first step towards a similar demonstration concerning
electromagnetism. Indeed one may interpret it by saying that, once applied to scale
dilatations, the principle of relativity implies the existence of some force of nature
additional to gravitation, of strength Fem ≈ Gm2P /r2 = ~c/r2 . However the full
understanding of these structures must clearly await a proper generalization of scale
relativity to fields, i.e. to nonlinear scale transformations.

9 Implications for High Energy Physics

9.1 Introduction
It is well known that Galilean motion relativity is recovered in the limit c → ∞ of
Einstein special relativity. Is it strictly true? Starting from special relativity, one
gets an expanded formula for energy given by E = mc2 + (1/2)mv 2 + (1/c2 ) · · · .

22
Taking the limit c → ∞ indeed makes all the last terms vanish and yields the
classical kinetic energy, but it also yields a term of infinite internal energy. Thus,
if one admits our argument of Sec. 4, according to which special relativity could
have been derived from the Galilean principle of relativity alone, classical mechan-
ics was already faced with a problem of energy divergence (even if this was not
explicitely realized) which the Einstein-Poincaré-Lorentz theory of relativity has
solved. Does scale relativity, which aims at understanding from first principles the
quantum behavior of microphysics, and which clearly has something to do with
electromagnetism (see previous section), solve the old problem of the divergences
of electric charge and self- electromagnetic energy ?
The present section is aimed at analysing this important issue and at first con-
sidering a not less important question: that of possible experimental verifications
of the theory. Having arrived at this point of our argument, the reader may indeed
legitimately ask himself whether scale relativity is a pure theoretical construction
whose consequences are only to be looked for only at the presently unobservable
Planck scale, or whether experimental consequences are to be expected in the en-
ergy range reached by existing particle accelerators. After a simplified reminder
of the current status of the charge and divergence problem in the standard model,
we shall show that scale relativity yields new predictions at observable energy yet
(E < 100 GeV), which may be used to test the theory.

9.2 The divergence of mass and charge : a reminder

In classical electrodynamics, the electrostatic energy of a system of point charges is
given in terms of the scalar potential φ by:
1X
U= ei φi . (70)
2 i

Once applied to the self-interaction of one electron, this gives one an electrodynam-
ical self-energy:
1 e2 1 α~c
Eem = = . (71)
2 r 2 r
So in classical electrodynamics, when r → 0, the electromagnetic contribution to
mass becomes infinite while the charge e (or in a similar way the coupling constant
α ) remains constant.
Let us now recall the state of the question in the frame of quantum electro-
dynamics (QED). For distances smaller than the Compton scale of the electron,
the nature of the problem of the interaction between two nearby charges radically
changes. Indeed, while the electromagnetic interaction was mediated only by pho-
tons at scales larger than the Compton length λc , this is no more the case when
the distance between charges becomes smaller than λc . The new behavior is due
to the phenomenon of creation and annihilation of virtual electron-positron pairs,
which mainly occurs, as expected from the Heisenberg relations, for time intervals
smaller than ∆t ≈ ~/2me c2 .
The most efficient way to get this high energy behaviour is the renormalization
group method [6, 7, 8, 9]. As discussed at the beginning of this paper, it is already
very close from the point of view adopted in scale relativity, namely that of a physics

23
explicitly dependent on scale; we additionally require in the present paper that
it be made scale-covariant. Consider the renormalization group Callan-Symanzik
equation [6, 7] for the electromagnetic coupling constant variation:

dα

= β(α) = β0 α2 + β1 α3 + . . . (72)
d ln λr

It may be obtained through very simple reasoning. One assumes that the coupling
α is explicitely scale-dependent and that it remains the only relevant parameter in
determining the physics at any given scale, so that even its infinitesimal variation
during an infinitesimal scale variation is a mere function of α itself: this yields the
first equality of Eq. (72). Then one assumes α ≪ 1, which allows one to expand the
β function in terms of powers of α. Finally the identification of the lowest order
terms with the result from the perturbative approach implies the vanishing of the
constant and linear terms, so that we deal with a marginal field: this yields the
second equality of Eq. (72). Now introducing the notation ᾱ = α−1 for the inverse
coupling, we get the differential equation it satisfies:11
dᾱ β
λ

= β0 + 1 + . . . , (73)
d ln r ᾱ

whose second order solution may be written as

    
λ0 β1 λ0
ᾱ = ᾱ0 + β0 ln + ln 1 + β0 α0 ln + ... (74)
r β0 r

The success of the renormalization group approach is demonstrated by the fact

that the lowest order solution automatically includes infinite sums of terms of the
form αn lnn (λ/r), which correspond to arbitrarily high orders in the “radiative
correction” perturbative method. These remarkable results, obtained from so simple
a method, apply to the coupling of QED (from the electron to the W/Z scale); to the
two couplings of the electroweak theory, α1 [U(1) group] and α2 [SU(2) group]; and
to the coupling α3 of Quantum ChromoDynamics (QCD) at high energies, for which
the condition α3 ≪ 1 remains fulfilled. [Note that to the second order, the actual
renormalization
P group equations for α1 , α2 and α3 are coupled: dᾱi /d ln(λ/r) =
βi + j βij αj ] [20, 21].
Let us come back to the divergence problem. The renormalization group results,
connected to the success of the electroweak and QCD theories allowed one to make
important progress in its understanding. The QED lowest order result:

α(λ)
α(r) = 2α(λ)
, (75)
λ
1− 3π ln r

still leads to a small scale divergence of the electric charge, but now at the “Landau
ghost”, the scale which makes the denominator to vanish. But one important result
has set the problem in a renewed way and led to the hope that it may be solved in
a frame requiring unification of electroweak and strong force at high energy [22, 23,
24]: the three running couplings have been found to converge at some high energy
11 A misprint in the published version on the second term (β0 instead of β1 ) is corrected here.

24
scale, the so-called Grand Unified Theory (GUT) scale. However, before turning to
this point, we remark that one may also remain dissatisfied if electrodynamics alone
can not be set as a self-consistent theory, which is the case in the present quantum
theory.
Let us first remind the lowest order QCD result [25]:
π
α3 = 11 nf
λs

(76)
2 − 3 ln r

where nf is the number of quark families (nf = 6 in the present standard model)
and where λs is an integration constant, whose experimental estimation presently
lies around 150 MeV [25]. This form for α3 , a consequence of the SU(3) group
structure (which has eight generators, identified as the eight intermediate gluons),
led to the important discovery of QCD asymptotic freedom [26, 27, 28], i.e. α3 → 0
when r → 0. The integration constant may also be expressed in terms of the value
of α3 at some given scale, often taken as the W boson scale (≈ 80 GeV). With this
choice and adopting nf = 6, one gets an inverse coupling given to lowest order by
 
7 λW
ᾱ3 = ᾱ3 (λW ) + ln , (77)
2π r
where ᾱ3 (λW ) = 9.35 ± 0.80 [29].
Concerning the electromagnetic and weak interactions, they are now unified in
the electroweak theory [30, 31, 32]. Its U(1) × SU(2) group structure implies 1 + 3
generators which, once mixed, yield the W + and W − bosons from one side and the
Z boson and the photon from the other side, and two couplings, α1 and α2 , which
are related to the electromagnetic coupling via the weak mixing angle θw (which
can be defined by cos θw = MW /MZ ):
3
ᾱ1 = ᾱ cos2 θw , ᾱ2 = ᾱ sin2 θw . (78)
5
So the inverse fine structure constant is (formally)
√ given by ᾱ = ᾱ2 + 5 ᾱ1 /3, while
the Fermi weak constant equals GF = πα/( 2MW 2
sin2 θw ).12 Current experimental
determinations of the basic parameters of the model from LEP are [33]: MZ =
91.177 (31) GeV, MW = 79.9 (4) GeV and sin2 θw = 0.2302 (21). Still unknown, as
well in the model as experimentally, are the number NH and the mass(es) of Higgs
boson doublets [25].
The two running couplings are given to the lowest order for r < λW by [29]
   
2 NH λW
ᾱ1 = ᾱ1 (λW ) − + ln , (79)
π 20π r
   
5 NH λW
ᾱ2 = ᾱ2 (λW ) + − ln , (80)
3π 12π r
while the variation of ᾱ from the electron to the W scale is given by [34]
 
2 X 2 MW 1
ᾱ(MW ) = ᾱ − Qf ln + , (81)
3π Mf 6π
f

12 In
√
the
√published version, the numerical factor 2/8 was erroneous. We give here the correct
factor π/ 2.

25
where the sum is done over all elementary particles, of charges Qf and mass Mf .
This yields ᾱ(MW ) − ᾱ = −9.2 ± 0.3 [35, 36]. Combining this result with the
measured value of Fermi’s constant, the values of the couplings at the W scale are
thus estimated to be ᾱ1 (λW ) = 59.17 ± 0.35 and ᾱ2 (λW ) = 29.07 ± 0.58 [29].
Let us finally come to grand unified theories. From Eqs. (77), (79) and (80) we
may plot the variation of the three couplings from the W scale to higher energies,
i.e. smaller resolutions. This yields the remarkable result of the convergence of the
three couplings at some high energy of the order of 1014 − 1015 GeV, which is a very
strong argument in favor of a complete unification of electromagnetic, weak and
strong (color) forces at this scale [22, 23, 24]. This convergence is ensured under
the “great desert hypothesis”, which assumes that there is no new particle (no new
physics) between the electroweak scale ≈ mW and the unification scale mX . Second
order terms in the solutions to the renormalization group coupled equations [20, 21]
do not change these conclusions, their contribution being presently smaller than the
errors on the couplings at the W scale.
GUTs achieved at first a lot of successes:
1. In their frame, the quantization of charge finds a natural explanation [23].
2. The value of the b quark mass may be predicted from its expected equality with
the τ lepton at energy mX , and from its evolution with scale deduced from the
renormalization group equations [37]; one finds Mb /Mτ (pred) = 2.75 ± 0.37,
to be compared with the observed ratio Mb /Mτ (obs)= 2.38 ± 0.06.
3. The number of generations is constrained to be n = 3 (a larger number would
have made the hereabove prediction unacceptably high) [37]: this has been
later definitively confirmed by primordial nucleosynthesis and by LEP [33].
4. The possible values of the low energy fine structure constant are constrained
to be < 1/25 [38], and even better 1/120 < α < 1/170 [39].
5. The value of Weinberg’s mixing angle may be predicted: at unification one
has α1 = α2 , and one may introduce a running effective angle, such that
sin2 θ̂W (mX ) = ᾱ2 (mX )/ᾱ(mX ) = 3/8, while the renormalization group
yields a scale variation given by [34]
  
2 3 109 α(mW ) mX
sin θ̂w (mW ) = 1− ln (82)
8 9 2π mW

The effective and measured values are related by sin2 θ̂w (mW ) = .9907 sin2 θw
[36]. This allows one to predict that sin2 θw (mW ) = 0.210 [36], which was
in good agreement with the measured value at the time of the prediction,
0.23 ± 0.02.
6. The decay of the proton was predicted [22], with a lifetime ≈ m4X /MP5 ≥ 1030
years for mX ≈ 1015 GeV, also in good agreement with known experiments
at that time.
We recall that most of these results hold in a large class of theories, not only
in the simplest GUT, based on the SU(5) group [23]. Unfortunately these great
hopes were soon dashed: the increase of precision of experimental results led to

26
unacceptable disagreements with the predictions. The present value of sin2 θw (mW )
as measured by LEP is sin2 θw (mW ) = 0.2302 ± 0.0021 [33], more than 10σ off the
theoretical prediction; the experimental proton lifetime is now known to be > 1032
years, this requiring that mX ≫ 1015 GeV, while agreement for the mixing angle
would require mX ≈ 1013 GeV.
This rather long reminder is intended to set the frame in which we shall now
consider some fundamental implications of scale relativity. We shall demonstrate
that the new structure of space-time we are proposing improves strongly the situ-
ation: it solves the divergence problem, yields a theoretical understanding on the
nature of the GUT scale which allows one to predict its value, and reconciles GUT
predictions with experimental results without introducing new particles.

9.3 Solution to the divergence problem

Because of the previous identity in standard quantum theory between length scale
and mass(-energy) scale [ln(m/m0 ) ≈ ln(r0 /r) for high energy in the rest frame],
the renormalization group equations are currently written indifferently in terms
of ln m or ln r. Actually the momentum representation, being easier to work out
and holding closer to experimental data, is systematically used in quantum field
theory rather than the position one (which may be obtained from the momentum
representation through a Fourier transform). In present quantum mechanics, the
momentum and position solutions to (72) differ only by some constants. This is no
more true in scale relativity, and one should now specify which changes are to be
brought to the renormalisation group equations.
One must in this respect distinguish between the cases of relevant fields and of
marginal fields. Fields which vary with scale as power laws (Eq. 6), for which we
have been able to establish a parallel with motion relativity laws, are precisely cases
of relevant fields. The lowest order term in their renormalization group equation is
linear:
dϕ

= δ × ϕ, (83)
d ln λr
this yielding the “Galilean” solution ϕ = ϕ0 (λ/r)δ . Note that the scale-relativistic
solution would correspond to the equation
3/2
ln2 (ϕ/ϕ0 )

dϕ

=δ×ϕ 1+ , (84)
d ln λr δ 2 ln2 (λ/Λ)

whose first non linear term is in (ϕ ln2 ϕ), so that such a form of the β function
could not have been guessed from the usual pure power expansion.
Conversely the lowest order term in the β function for marginal fields (which is
the case of coupling constants and masses) is to the square:
dα

= β0 α2 . (85)
d ln λr
This means that this lowest order term is beyond any order of the expansion of
Eq. (84), so that we conclude that the renormalization group equation for marginal
fields is unchanged in scale relativity (special, i.e. linear, case). However, an im-
portant point should be noted: this conclusion holds only for the renormalization

27
group equation expressed in terms of length scale: indeed the renormalization group
method, in its general definition [8, 9], as well as the scale relativistic approach, es-
sentially aims at describing the way physical laws change when going from one
spatio-temporal scale to a larger one. But the relation from length-time scale to
mass-energy-momentum scale is generalized in scale relativity (Eq. 64). So, while
we obtain the usual solution (Eq. 74) in terms of length scale, we get in terms of
mass scale the new relation (lowest order)

ln(m/mW )
ᾱ(m) = ᾱ(mW ) + β0 q (86)
1 + ln2 (m/mW )/C2W

where CW = ln(λW /Λ) = 39.876(6). At the absolute limit r = Λ, i.e. m → ∞, the

charge is now finite. So QED becomes a self-consistent theory in the frame of scale
relativity. In the same way, masses were previously divergent even to first order as
  
λ0
m = m0 1 + κα0 ln , (87)
r

while they now remain finite in scale relativity. We shall in the following let the
question of mass determination be open for future works, and shall focus mainly on
the coupling constant problem.

9.4 New predictions

In order to fix the ideas about the way scale relativity is expected to yield new
testable predictions, let us consider some numerical values. The amplitude of scale-
relativistic corrections will be given by Lorentz-like “scale γ factors” depending on
“V /C ” ratios, i.e. ln(λ0 /λ)/ ln(λ0 /Λ). For example, from the electron scale (0.511
MeV) to the W scale (79.9 GeV) , one already gets V /Ce = 0.232, i.e. a γ = 1.028,
which is not negligible. From the W scale to the GUT scale (≈ 1014 GeV), one gets
V /CW = 0.7, i.e. the large value γ = 1.4.
This last result allows us to introduce our first new prediction. In the standard
model, there is no understanding of why a new scale is needed in addition to the
W/Z one, and so no purely theoretical prediction of its value. Conversely scale
relativity naturally introduces a new fundamental scale in nature. Indeed the new
relation between the mass scale and the length scale (case r ≈ ct ) is such that the
Planck mass mP does not correspond any more to the Planck length Λ. We know
that to the Planck length now corresponds an infinite mass, and we thus expect a
new fundamental length λP to emerge (see Fig. 1). Let us compute it starting from
the W scale (we recall that in the new theory lengths are no more absolute; now
only scale dilatations from one scale to another have physical meaning):
 
λW ln(mP /mW )
ln = q . (88)
λP 1 + ln2 (mP /mW )/ C2W

A first estimate of this scale may be obtained by neglecting the fact the W length
scale should be itself subjected to a scale-relativistic correction. To this approxi-
mation we have CW = ln(λW /Λ) ≈ ln(mP /mW ), so that the denominator of (88),

28
 Λ 16
10

Length Scale ( λW / λ )
λ P 1012
Sch.

108

104

λW 1
80 107 1011 1015 1019 1023 1027 1031 1035
mW mP

Mass Scale (GeV)

Figure 1: The mass-energy / length relation in scale relativity. In the new theory the Planck
length, Λ = (~G/c3 )1/2 becomes an absolute and impassable limit to all lengths in nature. The new
relation is such that now the Planck space-time scale corresponds to infinite energy-momentum.
The Planck mass (mP = 1.22105(8) × 1019 GeV) then allows one to define a new universal length
scale, λP , which is found to be 10−12 times the W scale. This length-scale corresponds, in the
previous standard theory (dot-dashed line), to an energy of 1014 GeV, i.e. precisely the value of
the GUT scale deduced from the convergence of coupling constants. To the right of the diagram
the relation which corresponds to the Schwarzschild horizon rs = 2Gm/c2 is shown. It is now
reached at an energy of ≈ 1023 GeV.

√
i.e. the scale γ factor of the new length λP is found to be 2. So λP is expected
to be 1.41012 times smaller than the W scale, which corresponds to 1014 GeV in
the standard theory: in other words the new fundamental scale introduced by scale
relativity is precisely the GUT scale where the U(1), SU(2) and SU(3) couplings
are known to converge.
A more precise computation, accounting for the fact that the ratio λe /λW is
itself no more equal to mW /me , yields essentially the same result. Indeed at such
scales the correction is still small: one gets log(λe /λ
√ W ) = 5.059(3) , log(mW /me ) =
5.194(3), so that γ = 1.409 to be compared with 2 = 1.414....
This result has an important consequence. Indeed one may express it in another
way, by computing the energies corresponding to the unification scale. As expected
from the fact that the mass corresponding to λP is the Planck mass, we find that
the energy at which the α1 and α3 couplings cross themselves (Eqs. 77, 79 and 86)
is m13 = 1.1 × 1019 GeV/c2 , in excellent agreement with the value of the Planck
energy mP = 1.22105(8) × 1019 GeV/c2 . This means that now not only the three
electromagnetic, weak and color couplings converge at about the same energy, but
also the gravitational one (see Fig. 2). Indeed the gravitational coupling αg varies
with mass scale as αg = (m/mP )2 when m = (m20 + p2 /c2 )1/2 ≫√m0 ,13 so that ᾱg
reaches the common value ᾱ1 ≈ ᾱ2 ≈ ᾱ3 ≈ 40 at energy ≈ mP / 40 ≈ 1.9 × 1018
GeV. One may also directly study the crossing of the gravitational coupling with
the three others. We find that they cross at a scale λ = 1.8 × 10−12 λW , (which
corresponds in the standard theory to 4.4 × 1013 GeV), respectively at ᾱ = 42.4,
13 A misprint in the published version (≪ m0 ) has been corrected here.

29
42.8 and 39.7.
These results resolve the discrepancy concerning the GUT prediction of the
mixing angle and proton lifetime: the solution comes from the fact that it is length-
scale which occurs in the weak angle theoretical prediction, while the proton lifetime
prediction depends on mass-scale. The length-scale where α1 = α2 (Eqs. 79 and
80) is found to be 1011 times smaller than that of the W (previously 1013 GeV), so
that one gets a prediction sin2 θW (W ) = 0.232 ± 0.004 (see Eq. 82), which compares
well with the experimental value 0.230 ± 0.002 [33, 36]. (Note that this may imply
that there is no strictly common unification point, as recently remarked by some
authors [40]). But this unification range, 10−11 − 10−12 times the W scale, now
corresponds in mass to 1017 − 1019 GeV/c2 . Hence the proton life-time theoretical
expectation, which varies as m4X , becomes larger than 1037 years, far greater than
the present experimental lower limit.

Length Scale (λW / λ)

2 4 6 8 10 12
10 10 10 10 10 10

80 g
Inverse Coupling

40 2

20 3

10

5 7 9 11 13 15 17 19
80 10 10 10 10 10 10 10 10
Mass Scale (GeV)

Figure 2: Variation with length- and mass-scales of the inverse of the four coupling constants,
U(1), SU(2), SU(3) and gravitational (g). The mass-scale goes from the W boson mass (79.9 ± 0.4
GeV) to the Planck mass, mP = 1.22105(8) × 1019 GeV. The convergence point of couplings (i.e.
the GUT scale in the standard model), at a length-scale about 10−12 times the W length-scale,
now corresponds in scale relativity to the Planck energy (see Fig. 1 and text). In the new theory
the four fundamental couplings converge towards the same scale.

Another comment about the new structure is that it points toward a common
origin of all forces of nature. If we take Fig. 2 at face value, one starts at very high
energy from a purely gravitational regime, then going to lower energies one finds
a first decoupling with the color field, and finally with the electroweak field which
separates in U(1) and SU(2).
Let us point out another remarkable fact. In the standard theory, the Planck
length is equal to the Compton length of a Planck mass, but also to the Schwarzschild

30
radius of a Planck mass (to a factor of 2): Λ = ~/mP c = GmP /c2 . This implies
that a point mass larger than the Planck mass is a black hole of radius larger than
the Planck length. Hence the Planck scale is not only a domain where the quantum
and gravitational phenomena are expected to become of the same order, but also
a domain where the gravitational field itself is very strong: from this partly comes
the difficulty of elaborating a theory of quantum gravity. This is radically changed
in the new theory, so that the quantum gravity problem is now set in completely
different terms. Indeed the Planck mass mP now corresponds to the new scale λP ,
so that the gravitational potential at the Planck energy scale is in scale relativity
GmP /c2 λP ≈ 10−5 . This is a weak field situation, i.e. typically of the order of the
potential at the solar limb. The mass-length relation (Eq. 64) eventually crosses
the black hole horizon-mass relation (rs = 2Gm/c2 ), but at a far larger energy of
1.3 × 1023 GeV (see Fig. 1).
Let us now consider the question of the theoretical determination of the values of
the fundamental charges. The standard quantum theory, thanks to the renormaliza-
tion group approach, already arrived at a magnificent clarification of the problem:
the low energy charges (equivalently the low energy couplings) result from some
high energy unifying value at the GUT scale, and from a variation from high to low
energy which is completely determined by the renormalization group equations, in
terms of elementary particle masses. Hence one of the couplings may be estimated
from the other two: for example, starting from α1 (mW ) and α3 (mW ), one may
find, thanks to Eqs. (77) and (79), λX and αX at the GUT scale, and then predict
α2 (mW ) from Eq. (80).
Scale relativity brings several improvements to the solution of this problem.
The first one comes from the fact that we may now use the intersections of the
electroweak and color couplings with the gravitational coupling, which is completely
known, additionally to their own intersections. Starting, for example, from the
intersection point of αg and α3 , and admitting as a first step that it also yields the
zero point of α1 and α2 , allows us to deduce the low energy fine structure constant
to 6% of its measured value and the Fermi constant to 10%. We may also use
the fact that one of the free parameters of the standard theory, the GUT scale, is
theoretically known in scale relativity.
But the best improvement would come from an expectation for the value of the
“bare” common coupling at high energy. In this paper, we shall make a conjecture
about this value and we shall present some justifications of it based upon some new
structures which can be found in the (ᾱ, log r ) plane. Attempts at a theoretical
understanding of the origin of this value will be presented in a forthcoming work
[42]. Our conjecture is that there exist a universal high energy charge whose value
is equal to
√ 1
αP = . (89)
2π
This would yield a common inverse coupling ᾱP = 4π 2 = 39.478418.... Let us
consider the arguments in favor of this suggestion.14
One argument is the (αg , α3 ) intersection at which we have obtained ᾱ3g =
39.7 ± 0.8. Another argument is given by the electroweak couplings. At first sight
they do not support our conjecture, since they meet at an inverse coupling value
14 We have given in subsequent works theoretical arguments supporting this conjecture: see e.g.

L. Nottale, Chaos, Solitons and Fractals 7, 877 (1996); 16, 539 (2003).

31
of about ≈ 42. Consider, however, a formal electromagnetic inverse coupling ᾱ =
ᾱ2 + 53 ᾱ1 . Below the unification scale, it becomes equal to 8/3 of the common
inverse coupling ᾱ1 = ᾱ2 , and the solution to its renormalization group equation is
(lowest order)    
5 NH λW
ᾱ(r) = ᾱ(λW ) − + ln , (90)
3π 6π r
where we recall that NH is the (unknown) number of Higgs doublets. The value
of the fine structure constant at the W scale is known to 0.2%, [ᾱ(λW ) = 127.8 ±
0.3, see Eq. (81)], so that we may compute the value of ᾱ(Λ), the inverse formal
fine structure constant at the absolute limiting Planck scale, and then of ᾱ1 (Λ) =
ᾱ2 (Λ) = 3ᾱ(Λ)/8. We find, in this frame of a pure electroweak theory (1 Higgs
doublet assumed), that
3 5
ᾱ1 (Λ) = ᾱ2 (Λ) ≈ ᾱ(λW ) − CW = 40.0 ± 0.5, (91)
8 8π
which also supports the 4π 2 conjecture. Conversely the conjecture becomes testable,
since it allows us to predict the low energy value of the SU(3) coupling with a
precision which is a huge improvement on the presently known value. Starting
from the hypothesis that ᾱ3 (mP ) = 4π 2 and going back to the W scale from the
renormalization group equation, we find from a first order calculation ᾱ3 (λW ) =
0.113. However second order terms [20, 21] are not negligible to this precision.
Including them yields ᾱ3 (λW ) = 0.1165±0.0005, where the error comes from a rough
estimate of the contribution from third order terms. This prediction is compatible
with present experimental results: (0.120 ± 0.012 ) [33], (0.107 ± 0.011 ) [36], but
being far more precise, will allow one to test the theory when the experimental error
decreases.15
Equations (90, 91) do not lead to a similar prediction, owing to the fact that
the low energy fine structure constant is currently known to a high precision (ᾱ =
137.0359914(11) [41]). However it allows us to get its value to 0.5%, namely ᾱ =
137.7 ± 0.7 (first order, 1 Higgs doublet assumed), 138.1 ± 0.7 (second order, 1 Higgs
doublet), 136.0 ± 0.7 (second order, 0 Higgs doublet), and to predict that there can
be no more than 1 Higgs doublet, since each additional doublet would contribute
to +2.1 in the final result.16
Let us finally consider an additional argument in favor of the 1/2π conjecture.
It is remarkable that well definite structures seem to emerge in the new (ᾱ, log r)
plane. Hence the fundamental ratios of the Planck to W/Z mass scales, which are
now the two fundamental symmetry-breaking scales yield, from the current values
of weak boson masses (see Subsec. 9.2),
mP mP
ln = 39.567, ln = 39.436, (92)
mW mZ
15 Ten years later, the measurement of the SU(3) coupling has been improved by a factor of

6, and its value continue to support the theoretical prediction: α3 = 0.1172(20) [PDG2002] and
α3 = 0.1155(20) [NNLO].
16 Thanks to the improvement of the estimate of the variation of α between the electron scale

and the W/Z scale, we obtain now (in 2003) α−1 = 137.04 ± 0.03 for one Higgs doublet, which
continue to support the conjecture that the bare unified coupling (at Planck length-scale, that
is at infinite energy) is such that (3/8)ᾱ(∞) = 4π 2 . See references in note 14 and: L. Nottale,
Electromagnetic Phenomena T. 3, No.1 (9), 24 (2003).

32
which tightly enclose the ᾱP = 4π 2 value.17 Similar structures also seem to relate
the W/Z scale, but now in length, to the charge at the electron scale, which is the
third fundamental scale in the theory. Indeed we find:
λe λe
ln = 11.650, ln = 11.772, (93)
λW λZ
√
which enclose the inverse electric charge value, ē = ᾱ = 11.706.18 More tenta-
tively, note that the λP to Λ scale ratio may also be related to√the low energy electric
charge, since it is given to first approximation by CW (1 − 1/ 2) = 11.679. We shall
show elsewhere [42] that these remarks may be turned into precise formulas which
allow one to predict with high precision the W and Z boson masses, and which
brings new insights on the nature of the electric charge.19

10 Conclusion
What are the uncertainties in our construction ? Our theory is based on the fol-
lowing postulates:

1. Scale, as motion already does, may be defined as a relative state of reference

systems, so that scale transformations, i.e. dilatation and contractions, come
under the principle of relativity; the logarithm of the resolution with which a
measurement is performed is the measure of such a state, and plays in scale
relativity the part played by velocity in motion relativity.
2. The renormalization group method may be applied to space-time itself (in an
enlarged sense: it is applied to the length or time virtually “measured” along
a space or space-time particle path, i.e. to the internal quantum structure of
a particle).
3. The couple of variables (ln L, δ), i.e. the logarithm of length (or time) as
defined above, and the renormalization group anomalous dimension, play the
same role in scale laws as do length and time in motion laws.

Once these postulates are accepted, we believe we have demonstrated in a general

way that the general solution to the scale relativity problem implies the existence
of an impassable, absolute and limiting scale which is invariant under dilatations.
We could also have started from the postulate that the Planck length and time are
invariant under dilatation. This would have given us the same theory, but with
17 This relation has been borne out by the subsequent developments of the scale-relativity theory.

Indeed, the identification of gauge transformations with transformations of the internal resolution
variables has led to a proof of the quantization of charge based on the new meaning of the Planck
scale established in this paper, implying new relations between couplings and mass scales, of the
form α ln(mP /m) = 1. When one takes in this relation the bare value α = 1/4π 2 , one obtains
a mass-scale exp(−4π 2 )mP = 87.389 GeV, of the order of the W/Z mass scale. See references
quoted in notes 14 and 16.
18 This numerical value corresponds to a choice 4πǫ = 1.
0
19 Only the first of these remarks has subsequently acquired a profound physical meaning through

a theoretical development based on first principles (see note 17). The two other relations given
here (which involve the square-root of the coupling) have remained numerical coincidences devoid
of physical meaning.

33
the result that scale relativity should be broken above some particular transitional
scale, to be identified as the de Broglie length and time (otherwise one would get
an invariant dilatation instead of an invariant limiting scale).
Anyway the essence of our proposal may be traced back to the basic question:
Does such a limiting scale, invariant under dilatation, exist in nature ? If it does,
this is an universal law of nature, and the consequences of its existence must concern
the whole of physics. Even if the theory which is presented here was to be proved
insufficient in some of its aspects, one could not escape the need to build such a
theory and to make the whole of physics scale-relativistic.
Even in the restricted framework which has been considered in the present pa-
per, there is a lot of work still to be done. We have examined only the case of
one independent space or time variable, while proper account of the full space-
time should be taken: this must include in particular the questions of angles and
rotations. Solving the problem of the transformation of probability amplitudes is
also an urgent task;20 indeed our scale relativistic approach needs to be general-
ized to quantum systems less simple than those considered here (free particles), in
particular those to which no well defined de Broglie scale can be attributed.
We may, however, by now remark that there are some domains of physics which
would clearly be profoundly affected by this new structure, among them primor-
dial cosmology and unified theories (including gravitation). We shall particularly
consider the cosmological implications of scale relativity in a forthcoming work [42].
To conclude, we recall the encouraging successes which have been obtained:
(a) The theory solves the old problem of the divergence of charge and self-energy
of particles.
(b) It implies that the four fundamental coupling constants of physics converge on
about the same energy, which is now the Planck energy.
(c) It brings agreement, without introducing new interactions or particles, to the
predictions of GUT’s concerning the Weinberg mixing angle and the decay of
the proton, which were previously both mutually contradictory and inconsis-
tent with experimental results.
(d) It allows to fix to better than 10% the value of at least two out of the three
electromagnetic, weak and strong coupling constants and opens a new avenue
for the investigation of the nature of charge and the structure of elementary
particles: from the conjecture that the (now finite) “bare” charge is 1/2π, we
are able to predict a precise value for the strong coupling at the W scale.
Let us finally note that the precision of scale relativistic predictions is highly
dependent on the current error on the Planck length value, which is itself dominated
by the error on the gravitational constant. So it becomes urgent to have a precision
determination of G.
Acknowledgements: I am grateful to Y. Lachaud and E. Appert for pointing
out the incompleteness in the demonstration of Sec. 4 in the first version of the
manuscript, and acknowledge interesting discussions with Drs. S. Bonazzola, J.
Heyvaerts, P.Y. Longaretti, and C. Vilain.
20 See Ref. [42], p. 247, for an attempt of description of scale-relativistic transformations of

probability densities.

34
References
[1] Einstein, A., 1916, Ann. Phys. 35, 769
[2] Levy-Leblond, J.M., 1976, Am. J. Phys. 44, 271
[3] Nottale, L., 1989, Int. J. Mod. Phys. A4, 5047
[4] Gell-Mann, M., & Low, F.E., 1954, Phys. Rev. 95, 1300
[5] Stueckelberg, E.C.G., & Peterman, A., 1953, Helv. Phys. Acta, 26, 499
[6] Symanzik, K., 1970, Commun. Math. Phys., 18, 227
[7] Callan, C.G., 1970, Phys. Rev. D2, 1541
[8] Wilson, K.G., 1969, Phys. Rev. 179, 1499
[9] Wilson, K.G., 1983, Rev. Mod. Phys. 55, 583
[10] Mandelbrot, B., 1982, “The fractal Geometry of Nature” (Freeman)
[11] Nottale, L., & Schneider, J., 1984, J. Math. Phys. 25, 1296
[12] Abbott, L.F. & Wise, M.B., 1981, Am. J. Phys. 49, 37
[13] Feynman, R.P., & Hibbs, A.R., Quantum Mechanics and Path Integrals
(MacGraw-Hill, 1965)
[14] Bacry, H., & Levy-Leblond, 1968, J. Math. Phys. 9, 1605
[15] Fuller, F.W., Wheeler, J.A., 1962, Phys. Rev. 128, 919
[16] Hawking, S., 1978, Nucl. Phys. B144, 349
[17] Norman, E.B., 1986, Am. J. Phys. 54, 317
[18] Landau,L., & Lifshitz, E., Field Theory, Ed. Mir, Moscow
[19] Cohen, E. R. & Taylor, B.N., 1987, Rev. Mod. Phys. 59, 1121
[20] Jones, D.R.T., 1982, Phys. Rev. D25, 581
[21] Babu, K.S. & Ma, E., 1986, Z. Phys. C 31, 451
[22] Pati, J.C., & Salam, A., 1973, Phys. Rev. Lett. 31, 661
[23] Georgi, H. & Glashow, S.L., 1974, Phys. Rev. Lett. 32, 438
[24] Georgi, H., Quinn, H.R., Weinberg, S., 1974, Phys. Rev. Lett. 33, 451
[25] Hernandez, J.J., et al. (Particle Data Group), 1990, Phys. Lett. B239, 1
[26] Gross, D.J. & Wilczek, F., 1973, Phys. Rev. Lett. 30, 1343
[27] Politzer, H.D., 1973, Phys. Rev. Lett. 30, 1346
[28] Coleman, S. & Gross, D.J., 1973, Phys. Rev. Lett. 31, 851

35
[29] Marciano, W.J., 1989, Phys. Rev. Lett. 62, 2793
[30] Salam, A., in Elementary Particle Physics (Almqvist & Wiksell, Stockholm,
1968).
[31] Weinberg, S., 1967, Phys. Rev. Lett. 19, 1264
[32] Glashow, S., 1961, Nucl. Phys. 22, 579
[33] Miller, D.J., 1991, Nature 349, 379
[34] Marciano, W.J. & Sirlin, A., 1981, Phys. Rev. Lett. 46, 163
[35] Marciano, W.J. & Sirlin, A., 1988, Phys. Rev. Lett. 61, 1815
[36] Amaldi, U. et al., 1987, Phys. Rev. D36, 1385
[37] Buras, A.J., Ellis, J., Gaillard, M.K., Nanopoulos, D.V., 1978, Nucl. Phys.
B135, 66
[38] Glashow, S.L. & Nanopoulos, D.V., 1979, Nature 281, 464
[39] Ellis, J. & Nanopoulos, D.V., 1981, Nature 292, 436
[40] Amaldi, U., de Boer, W. & Füstenau, H., 1991, Phys. Lett. B260, 447
[41] Kinoshita, T., 1989, IEEE Trans. Instrum. Meas. 38, 172
[42] Nottale, L., Fractal Space-Time and Microphysics : Towards a Theory of Scale
Relativity (World Scientific, 1993).

36