Académique Documents
Professionnel Documents
Culture Documents
R. Eugene Collins1
Received June 2, 2004; revised March 3, 2005
1. INTRODUCTION
This paper is addressed to all who share the desire to not simply know
physics but to understand physics, that is, to know why the laws of phys-
ics have the specic forms we discover through empirical studies. As an
undergraduate in physics more than 57 years ago, my reaction, when rst
exposed to the amazing capability of Newtons laws to predict the motions
of physical objects, was to ask the professor Why F = ma? Of course
he simply assured me that these were empirical relationships which cannot
be explained by more fundamental laws. At the time I accepted this but
as I continued my education I became convinced that not only Newtons
laws but all fundamental laws of physics must have a common basis linked
743
0015-9018/05/0500-0743/0 2005 Springer Science+Business Media, Inc.
744 Collins
Euclidean manifold of dimension 2n+1. Here closed means that all limit
points for point sets in M are also points of M. However, if the metric of
M is constrained in an appropriate manner M can be embedded closed in
E K with K < 2n + 1 but K > n. For example, the surface of a sphere in
E 3 is a two-dimensional Riemannian manifold embedded closed in E 3 .
Since all points of M are also points of E 2n+1 , the directed line
element dR, from a point P of M to a near-neighbor point P of M,
exists also in E 2n+1 . Specically, with the set of coordinates, j , j =
1, 2, . . . , 2n + 1, on E 2n+1 and x = x ( ), = 1, 2, . . . , n local rectilinear
coordinates on M, we may write dR, with the Einstein summation con-
vention, as the Euclidean displacement vector
R
dR = dx = 1 dx . (1)
x
This denes the vectors 1 (x) = R/x , = 1, 2, . . . , n at the point P
as local basis vectors on M in the directions of increasing x normal to
the coordinate hypersurfaces in E 2n+1 dened as x ( ) = constant =
1, 2, . . . , n. Thus, forming the Euclidean dot product, we have
dR dR = dR 2 = 1 1 dx dx , (2)
R
v= v = 1 (x)v . (5)
x
These x and v , being continuous, must be considered as random vari-
ables having a joint probability distribution dP = dP (x, v| ) conditioned
by a continuous, real scalar and constant parameters of the particle
and its environment whose values are xed in the preparation of the sys-
tem for observation, but these are not explicitly displayed here. However,
there are variable environmental parameters whose values are not xed by
the experimenter. Thus we have introduced the real scalar as a continu-
ous one-parameter label for the conguration of the external environment
dened such that if anything in the external environment changes then
the value of must increase. Thus must increase if the sun or moon,
or the hands on a clock, change position. Hence is time measured by
the experimenter with an appropriate clock in his frame of reference. This
frame of reference is the unchanging x and v manifolds. We will refer to
as world time.
We view dP = dP (x, v| ) as the distribution of results in multiple rep-
lications of an experiment predicted by mathematical theory, with pre-
dicted observed values for observables as expectation values under dP. By
the law of large numbers(4) this identication for observed values is sta-
tistically consistent with experimentally determined observed values com-
puted as the arithmetic means of measured values from replications of the
experiment. Here observables of the particle are real, single-valued, covar-
iant functions of the x and v . However, the x and v themselves are
not proper observables because the expectation values, as say v under
dP , are not covariant. In particular R is the observed value for the posi-
tion vector at time predicted by dP (x, v| ) so dR/d is the predicted
value for the observed velocity v in the experimenters frame of reference.
Thus with R and v under dP the identity
dR
v (6)
d
actually denes the vector v as velocity in this probability formalism.
The probability dP (x, v| ) is now expressed in a more useful repre-
sentation as the factored form
dP (x, v| ) = dPv (v|, x)x (x| ) gd n x (7)
748 Collins
with dPx = x gd n x the marginal probability for the x on the vol-
n
ume element gd x, conditioned only by and the xed parameters, and
dPv (v|, x) the conditional probability for the v which is conditioned by
the x as well as and the xed parameters. Thus dPv has unit integral
over the v-manifold for all and any values of the x D while dPx has
unit integral over the domain D of the x-manifold for all values.
By appropriate choice for the order of x and v integrations, as per-
mitted by Fubinis theorem,(5) this representation for dP (x, v| ) yields the
expectation value for R as
R = R(x)x (x| ) gd n x. (8)
D
This follows because the rst executed integral of dPv over the v is unity.
Similarly the expectation value for v under dP is the expectation value for
under dPx
v = = 1 (x) (x, )x (x| ) gd n x, (9)
D
Clearly these representations for R and v are covariant forms.
The dening relationship for v in Eq. (6) can also be represented in
terms of the vector eld (x, ), dened by Eqs. (9) and (10), in the form
dR
= v = = (x, )x (x| ) gd n x, (11)
d D
since the expectation value for (x, ) under dPx = x gd n x is equivalent
to the expectation value for v under dP and this is the observed value for
v at time . However, in Eq. (11) is unchanged if (x, ) is replaced
by u = (x, ) + q(x, ) with q(x, ) a vector eld, of the same charac-
ter as (x, ), but such that q under dPx is identically zero. With u thus
dened Eq. (11) can be written as
dR
= v = x gd n x = ux gd n x =u. (12)
d D D
The Mathematical Basis for Physical Laws 749
x R R (x ) + (x R), (14)
we may apply the divergence theorem to the integral of the last term on
the right obtaining a surface integral of x R that is zero because x is
zero exterior to the domain D. Thus with / taken inside the integral for
R from Eq. (8) on the left and some rearrangement, Eq. (12) becomes
the identity
x n
R + (x ) gd x 0, (15)
D
which states that the quantity in brackets is a scalar function, say Y (x, ),
x
+ (x ) = Y (16)
such that
RY gd n x 0. (17)
D
However, integrating each term of Eq. (16) over the domain D yields
x n
gd x + (x ) gd n x = Y gd n x, (18)
D D D
750 Collins
and both integrals on the left are identically zero, so the integral on the
right must also be identically zero. The rst of these integrals is zero
because any x must have unit integral over D for all values and the
second is zero since the divergence theorem, with x zero exterior to D,
yields a surface integral which is zero, and these are identities.
A form for Y which assures the right member of Eq. (18) identically
zero for any x is
Y = (x q) (19)
since by the divergence theorem the integral of this over all x D is zero
for any bounded, differentiable, vector eld, q(x, ), because x is zero
exterior to D. However, this form for Y is valid only if q satises the spe-
cic constraint imposed by Eq. (17). This constraint is exhibited in a more
informative representation as follows: Substituting Eq. (19) for Y into Eq.
(17) yields the requirement for q as
R (x q) gd n x 0 (20)
D
and here we again use the identity from Eq. (14) but now with q replacing
. This provides Eq. (20) in the form
(x qR) x q R gd n x 0 (21)
D
and again using the divergence theorem, the rst term yields a surface
integral which is zero because x is zero exterior to D. Finally, noting
that R is the identity dyadic, as discussed following Eq. (13), we see that
q R = q (1 1 ) = q so we have the remaining term of Eq. (21) as
qx gd n x 0 (22)
D
which must be an identity. Therefore the vector eld q and the scalar eld
x must be functionally related such that q 0 for any x . Thus q is
the same vector eld used to write in Eq. (11) as u in Eq. (12) with
u = + q.
Now, with Eq. (19) for Y in Eq. (16), we nd upon rearrangement
that
x
+ (x u) 0 (23)
The Mathematical Basis for Physical Laws 751
4. PROBABILISTIC DYNAMICS
u (x u) (x uu) + x u u (25)
to substitute for u (x u). The integral of the rst term on the right
is then zero by the divergence theorem because x is zero exterior to
D. Thus we obtain
dv u n du n
+ u u x gd x x gd x, (26)
d D D d
du n u 1
x gd x + u u + u u x gd n x (28)
D d D 2
u = + A = P (29)
with and real, scalar constants providing dimensions and units to the
dimensionless scalar elds, and P, and the solenoidal vector eld A. Here
the additive harmonic eld is included in Eqs. (29) and (30) as p + a
with p a harmonic component in P and a a harmonic component in
A. That is, p and a are identically zero while p and a
are zero everywhere except on isolated singularities of zero measure with
respect to gd n x. Note also that since A is not everywhere zero the inte-
gral term is path dependent such that the scalar eld P is multivalued.
This characteristic is examined in detail in another paper but need not be
addressed here because P is single valued.
These representations for u are now selectively introduced in Eq. (28)
with P for u in u/ and + A for u in u to arrive at the
identity
du n P 1
x gd x + u u + u A x gd n x (31)
D d D 2
P 1
+ uu =0+Q (32)
2
with Q(x ) such that Q 0 under any x , then the identity in Eq. (31)
assumes the form
du n
x gd x {0 + Q + u A}x gd n x (33)
D d D
du
= u A + 0 + Q (34)
is a valid equality for all x D and any and constitutes the dynamical
equation of evolution for u(x, ), Furthermore this dynamical equation can
actually be obtained directly as the gradient of Eq. (32) by making use of
Eq. (29) for u together with the identity for (u u/2) from Eq. (27) and
the denition of du/d as u/ + u u.
Now in the limit of absolute precision for observations of the x ,
with the limit for x (x| ) being the Dirac delta function (x xc ( )) and
Q 0 under all x , the identity in Eq. (33) assumes the limiting form
duc
= uc A + 0 (35)
d
in which uc (xc ( ), ) denotes the limiting form for the expectation value
for u(x, ) under x = (x xc ( )), and all functions of the x in this
equation are evaluated on the trajectory x = xc ( ), = 1, 2, . . . , n.
Thus, in the limit of absolute precision for observations of quantities
determined by the x , we obtain dynamics for a particle with Eq. (35) as
the deterministic equation of motion.
Some clarication of notation is called for here since the symbols 0,
A, and in Eq. (34) have been retained unchanged in Eq. (35) while in
fact these quantities are changed, but only in one key characteristic. While
0 and the A may have an explicit dependence on , as well as on the x
754 Collins
in Eq. (34), the basis vectors 1 , and therefore the operators and , are
functions only of the x , but in Eq. (35) all of these quantities have gained
an implicit dependence on through replacement of the x by the xc ( ).
This limit of absolute precision denes conditions necessary to jus-
tify the assumption of classical physics, held since the days of Galileo and
Newton, that the trajectory of a particle can be described by coordinates
xc ( ) as continuous functions of a time parameter.
Before we proceed to further details of this construction of particle
dynamics we show by direct construction that there does exist a form for
Q as a function of x such that the identity Q 0 is valid. Writing x
as Rx2 (x, ) with Rx a real, single-valued function of the x and , differ-
entiable to all orders, we dene Q as
2 Rx 2 x
Q = (36)
Rx x
with I the identity dyadic 1 1 . The integral of this over the domain D,
with the divergence theorem, yields a surface integral for the bracketed
dyadic which is zero because x and therefore both Rx and Rx are uni-
formly zero exterior to the domain D. Thus the function Q(x ) in Eq. (36)
assures Q identically zero for any x .
Observe that the precision limit for this probabilistic formalism yields
terms on the right in Eq. (35) having forms appropriate for identication
as electromagnetic forces on a moving charged particle. Specically, simply
raising and lowering the index we see the rst term as
uc A = 1 u
c F = 1 uc F ,
(38)
components of the proper four velocity of the particle in the Lorentz met-
ric. In this relativistic case there remains the term 0 yet to be identied
in terms of physical quantities. Later this will be shown to be the gravita-
tional force per unit mass acting on the particle for this manifold dimen-
sion and metric.
Note that the equation of motion in Eq. (35) is essentially an iden-
tity. Indeed, these laws of mechanics, as well as the continuity equation,
are simply consequences of the denition for the v by Eqs. (5) and (6),
and in this sense these are identities. Thus we nd that not only Newtons
second law, F = ma, appears here as a mathematical equation relating
two equivalent representations for the same mathematical quantity as F/m
and a, but also this purely mathematical construction yields specic laws
of force. However the elds appearing in the force terms of this equa-
tion of motion must be specied if this equation is to be applied to pre-
dict the motion of a particle. Here it will be shown that eld equations for
these elds are also dictated solely by mathematics, but rst we point out
other aspects of these constructions.
The continuity equation, Eq. (23),
x
+ (x u) = 0 (39)
and the dynamical equation, Eq. (34), with Eq. (36) for Q,
du 2 x
= u A + 0 , (40)
d x
are a pair of nonlinear eld equations for x and u. For specied elds
0 and A, with specied boundary and initial conditions for x and
u, these equations determine the scalar eld x (x| ) and the vector eld
u(x, ). In principle, for a given metric, these determine not only the evo-
lution for x and u but also the distribution of possible trajectories for a
particle as streamlines everywhere tangent to the local velocity u. However,
because of singularities noted below, this is not practical.
As already noted just following Eq. (34), this dynamical eld equation
can also be constructed directly from Eq. (32), with Eq. (36) for Q. With
u as P from Eq. (29) and Eq. (36) for Q we may write Eq. (32) as
P 1 2 x
+ |P| = 0
2
, (41)
2 x
756 Collins
h2 2
ih = + 0, (43)
2m
1
i = (i + A) (i + A) + (0 + / ) , (44)
2
The Mathematical Basis for Physical Laws 757
is the basic denition for the tensor F in this formalism. For manifold
dimension n, the tensor F is antisymmetric with n(n1)/2 distinct scalar
elements, say c1 , c2 , . . . , cN with N being n(n 1)/2. These can be viewed
as the components of one or more differentiable vector elds on the man-
ifold. Since A denes a pseudo-vector on the basis vectors within the
manifold only for n = 3, the smallest manifold dimension to be considered
is n = 3. For n = 3 the matrix [F ] is a 3 3 matrix having only three
distinct elements which are identied as components of a three vector, ci .
These must appear in the matrix [F ] with appropriate signs assuring this
matrix antisymmetric with zeros on the principal diagonal.
For n = 4 there are six distinct elements of F which are the three
components of cI plus the three components of a second three vector, cII .
The latter must appear, with appropriate signs and positions, in an added
row and column appended to the 3 3 matrix containing the components
of cI to form a 44 antisymmetric matrix with zeros on the principal diag-
onal. Proceeding in this manner, components of a four vector, cIII , with
appropriate signs and locations, must be introduced in another added row
and column to form the matrix F corresponding to n = 5. Continuation
758 Collins
ca = a + Ka with Ka 0 (46)
F F F
F, =
+
+ 0. (48)
x x x
0 ci3 +ci2 cii1 ciii1
+c3 0 ci1 ciii2 ciii2
i
[F ] = ci2 +ci1 0 cii3 ciii3 . (49)
+cii1 +cii2 +cii3 0 ciii4
+ciii1 +ciii2 +ciii3 +ciii4 0
Observe that in this manner equations for ci , cii , ciii , . . . , cn2
as in Eqs. (47), are generated, and assignments for all elements of [F ]
in terms of these vectors are determined. However, there remain the
eld equations for ci , cii , ciii , . . . , cn2 required by Eqs. (47),
yet to be specied. These constitute n additional scalar equations in
the elements F yet to be dened. Note also that the equations for
cii , ciii , . . . , cn2 deduced above are actually metric-independent iden-
tities simply because the Bianchi relationship is an absolute identity.
Before we consider constructions for these additional eld equations
in the F , we should note that the Bianchi identity in Eq. (48) can be
represented in the covariant vector notation as
| F = 1 1 1 F, 0 (50)
1
J = F = 1 ( gF ) (51)
g x
760 Collins
since 1 /x = {, }1 , 1 1 = and {, } = n g/x from
Section 2. Hence
2 gF
J = 0, (52)
gx x
determining the F tensor for any manifold dimension n and any metric.
This system of equations is based on just two denitions and two iden-
tities, F is dened as A and J is dened as F while | F 0 and
J 0 are identities. Thus the equations of evolution for x and u in
Eqs. (39) and (40), and the eld equations in Eqs. (53), are fully covariant
and based on identities.
However it must be noted that Eq. (53b) dening the vector eld J
is not unique; actually any antisymmetric, differentiable dyadic P denes a
vector eld J as J = P such that J 0 but only those dyadics
P having elements which are functions of the ca elds dened in F, such
that the equation P = J yields the eld equations for cI , cII , . . .
required by the HodgeDeRham theorem in Eqs. (47), are acceptable. This
alternative can be applied to construct eld equations in material media
but here, in matter-free space, we retain only F itself as this dyadic P.
Not only does Eq. (53a) provide the forms for ci , cii ,
ciii , . . . , cn2 required by Eqs. (47) but also Eq. (53b), provides the forms
for ci , cii , ciii , . . . , cn2 based on the HodgeDeRham theorem
in Eqs. (47). To show this we note that the dyadic F = A can be repre-
sented directly in terms of these eld vectors as
with each of the dyadics Fi , Fii , . . . containing only one vector eld. Spe-
cically, with the ca assigned in the matrix F as in Eq. (49), we have
(a) Fi = 1i ij k cik 1j
(b) Fii = 14 cii 1 1 cii 14 (55)
(c) Fiii = 15 ciii 1 1 ciii 15
with forms Fiv , Fv , . . . , Fn2 , just like Fiii for n > 5, each also containing
only one vector eld. This representation contains all F elements, so we
construct F by applying the linear operator, , onto F in Eq. (54)
which yields a sum of terms, Fi + Fii + Fiii + + Fn2 , each
containing only one of these vector elds.
With this construction for F Eq. (53b) does yield explicit equa-
tions for ci , cii , ciii , . . . , cn2 , just as called for by the Hodge
DeRham construction in Eqs. (47).
From the representation for A as the dyadic F, which has the form
Fi + Fii + + Fn2 in Eq. (54) with only one corresponding vector eld,
ci , cii , . . . , cn2 in each of the dyadics Fa , we nd that the general equation
of evolution from Eq. (34) can be written as the covariant form
du
= u Fi + u Fii + + u Fn2 + G + Q (56)
d
du
= [us ci + u4 cii us cii 14 + u5 ciii uciii 15 + un cn2
d
ucn2 1n ] + G + Q, (57)
where G denotes 0. Here, in the rst and third terms of Eq. (57), us
is the three vector of components u , = 1, 2, 3 of u because ci and cii
are three vectors. Also, with the equivalence ij k = ikj and renaming of
indices, j k and k j , Fi in Eq. (55) was modiedto arrive at the
j
cross-product form since us Fi then appears as 1 u 1i ij k ci 1k and
j
with the Einstein summation convention this is ijk ui ci 1k = us ci .
In this we see that each of the eld vectors ci , cii , . . . , cn2 con-
tributes separately to an additive force on the particle, but in addition
to these forces arising from F, there also appears the force due to the
n-dimensional eld vector G = 0. Of course all of these forces are
actually forces per unit particle mass.
762 Collins
(a) G = 0 0,
20 0
(b) G = 0 = g
g {, } = JG (58)
x x x
as appropriate eld equations for the vector eld G, with JG a scalar
source term just as in Eqs. (47) for the ca . However, the (a) equation here
is a trivial identity comparable to | F 0 for the dyadic eld F, while
in the (b) equation g = 1 1 is the
contravariant metric tensor and
g {, } represents 1 1 x which is clearly zero on any at
manifold. These two equations, together with Eqs. (53), form a complete
set of covariant eld equations for the covariant equation of evolution
in Eq. (57), determined by the scalar function JG and the n-component
vector eld J which must be prescribed.
However these covariant equations are useless without a prescribed
dimension and metric for the coordinate manifold. Since there is a distinct
classication for manifolds as at or curved we begin our comparisons of
this formalism to some elements of familiar physics by restricting the for-
malism to at manifolds. This will provide new insight into important ele-
ments of physics.
equations for this particle. In this diagonal metric we have assigned the
three-dimensional spatial manifold to be isotropic simply because this is
consistent with our local perception of space.
On this four-dimensional x-manifold the eld dyadic F = A contains
only the two three-vector elds cI and cII with elements congured as in
Eq. (49). Therefore we here adopt the notations
since B will be identied as the magnetic eld and E the electric eld
acting on the particle in constructions following.
With these elements specied, applying Eq. (53a),
| F 0, to the eld
dyadic yields four scalar equations in the components of B and E which
can be written in the Euclidean three-vector format as
(a) B 0
B . (60)
(b) E + 4 0
x
These are actually trivial identities. This is demonstrated by substituting
the denitions for the E and B as elements of the eld tensor F = A
into Eqs. (60) which yields both equations identically zero. These deni-
tions, based on Eqs. (49) and (59), can be represented in the three-vector
format as
(a) B = As
As , (61)
(b) E = A4 4
x
with As the three-vector of spatial components of the solenoidal eld A
and A4 the coordinate time component. Eqs. (60) have the familiar forms
of the metric-independent Maxwell equations.
The metric-dependent pair of Maxwells equations in free space fol-
low directly from application of Eq. (53b) to this dyadic F, but only with
additional consideration for the form of the metric tensor.
Since Eq. (53b) is equivalent to Eq. (51) and here the metric tensor is
diagonal as (gss , gss , gss , g44 ) with all g either +1 or 1 for the rectilin-
ear coordinates, the determinant g is either +1 or 1 so Eq. (51) becomes
F
F = 1 = J. (62)
x
Therefore we employ the identity
F = 1 F 1 = 1 g F g 1 = 1 F 1 (63)
764 Collins
to express the F in terms of the known F in Eq. (49) with the iden-
tications cI = B and cI I = E. Since g g 1 this equivalence can be
written as
1
F = g g F . (64)
For this particular metric all of these coefcients are either +1 or 1. For
2 )1 is positive,
all Fij , i, j = 1, 2, 3 in FI this coefcient is +1 since (gss
while for the F4i and Fi4 , i = 1, 2, 3 in FII the multiplier is (gss g44 )1
which is +1 for the gss and g44 having a common sign or 1 for different
signs.
Thus with FI and FII from Eq. (55) we form F as FI + FII to obtain
Eq. (62) as
1
1 1 i ij k
B j 1 k + (g ss g 44 ) (1 4 E E1 4 ) = 1 J (65)
x
in which we have raised and lowered all corresponding indices in F and
substituted cI = B and cII = E. Also we have used the equivalence ij k =
ikj and renamed indices as j k and k j . Thus executing the diver-
gence operation yields
E
B + (gss g44 )1 E1 4 = 1 J
(66)
x 4
1 2B
2B + = 0, (68)
gss g44 (x 4 )2
and by similar constructions E must satisfy this same equation.
The Mathematical Basis for Physical Laws 765
Now if (gss g44 ) > 0 this equation admits solutions for B, and simi-
larly for E, which are exponentially unbounded on the space manifold as
x 4 +, but if (gss g44 ) < 0 then only bounded solutions are admit-
ted. Thus to eliminate these exponentially unbounded solutions, we must
assign (gss g44 ) < 0. Then for |gss | and |g44 | both equal to unity we nd
the metric tensor on this four-dimensional, space-time manifold must be
either (+1, +1, +1, 1) or (1, 1, 1, +1), but there is a requirement
which determines that the latter is the only correct form.
Recall from Section 2 that the embedding theorem yielded the square
of the line element on the Riemannian manifold as the positive product
form in Eq. (2) and here, on this at, pseudo-Euclidean manifold, with
the diagonal Lorentz metric tensor (1, 1, 1, +1), in the precision limit,
x (x| ) (x xc ( )), the expectation value for this is
Thus since dxcs dxcs = |dxcs |2 for this diagonal metric, this
implies
that
the magnitude of the expectation value for particle velocity
dxcs dtc on the spatial manifold must be less than the speed c. Here the
notation employs the displacement vector as dxcs = 1i dxcs i on the E 3 spa-
tial manifold while dxc = cdtc is that on the coordinate time axis in the
4
precision limit, with all x xc ( ) in this limit. Were the alternative met-
ric selected the opposite would be implied, that is, the particle speed would
always exceed c in magnitude which is obviously not acceptable.
Thus it is the combination of requiring the space-time manifold at
and spatially isotropic, together with bounded solutions for the source-free
Maxwell equations as x 4 + and the constraint of dRc2 real and posi-
tive imposed by the embedding formalism, which determine that the met-
ric of this at space-time manifold must be the diagonal Lorentz metric
(1, 1, 1, +1).
Hence with gss = 1 and g44 = +1 these constructions, based on
the general covariant eld equations from Eqs. (53), yield eld equations
for the F equivalent in form to the Maxwell equations on this four-
dimensional manifold carrying the Lorentz metric, that is,
E
(a) B = 0 (c) B = Js
x 4
B
(b) E+ =0 (d) E = J 4 (70)
x 4
J 4
(e) Js + =0
x 4
766 Collins
with (a) and (b) from Eq. (53a), (c) and (d) from Eq. (53b) and (e) from
Eq. (53c). Also with J 4 as ce , e being the spatial density of electric
charge, and x 4 = ct, J 4 x 4 is e t so Eq. (70e) is the familiar con-
servation of electric charge on the spatial manifold which identies Js as
the electric current density. Certain factors of 4 and 1/c usually appear-
ing in the Maxwell equations, see Jackson,(15) p. 2, are absent in Eqs. (70)
because our vector eld A is the multiple c/(4 ) of the conventional four-
vector potential of the electromagnetic eld in Gaussian units.
Incidentally we also observe that with the Lorentz metric and x 4 = ct
in Eq. (68), this equation for B, and the corresponding equation for E,
have the form of the wave equation for waves of phase velocity c. Thus
this probability formalism for particle mechanics has identied c as the
constant, invariant speed of electromagnetic waves in free space, that is, c
is the speed of light.
All that remains for a complete set of eld equations is the appro-
priate eld equation for the scalar eld 0 which we tentatively identify as
the gravitational potential. Since this eld 0 must satisfy the inhomoge-
neous equation
in Eq. (58b) and on any at manifold 0 is simply
g 2 0 (x x ), such that in the diagonal Lorentz metric, with elements
(1, 1, 1, +1), x 4 as ct and JG as 4M , this is the inhomogeneous
DAlembertian equation,
1 20
20 = 4M , (71)
c2 t 2
with xs and xs position vectors on the at spatial manifold and G| the
causal form of the innite-domain Greens function for Eq. (71), as in
Mathews and Walker,(16) pp. 278280,
t t xs xs c
G| = . (73)
4 x x
s s
The Mathematical Basis for Physical Laws 767
du4 0 Q
= us E + 4 + 4 (77)
d x x
since I4 I4 = g 44 = +1 and u4 = g 44 u4 . Thus in addition to the
quantum mechanical force, Q, Eq. (76) includes the forces (per unit
mass) acting
on the relativistic particle, as the electromagnetic Lorentz
force us B + u4 E , with = e/(mc), and the force 0 which we iden-
tify as the gravitational force with 0 the gravitational potential.
A note on signs of certain terms is in order here; recall from Sec-
tions 2 and 3 that R, v and u are dened by components on the covariant
basis vectors 1 while the gradient operator is dened by components
on the contravariant basis vectors 1 . Thus doting 1 onto Eq. (75) yields
du /d having g 0/x in the right member because 1 1 = and
768 Collins
duc
uc = 0 uc , (79)
d
since uc A uc is identically zero simply because A is antisymmetric
in the coordinate indices. Hence, multiplying Eq. (79) by d and executing
the indenite integral provides
1 1
uc u c = 0 + (80)
2 2
with the constant of integration /2, independent of , having units of
energy per unit mass. This integration makes use of the fact that uc d is
equivalent to dRc from Eq. (13), with dR dened as in Eq. (1). This pro-
vides (dR/d )d as dRc in the precision limit x (x| ) (x xc ( ))
such that 0 dRc is simply d0 because in an isolated system 0 is an
explicit function of only the x and not .
Note that for any metric Eq. (80) can also be written as
g xc xc 20 = , (81)
with the uc = xc = dxc /d being components of the proper velocity of
the particle in the reference frame of the observer. For the relativistic par-
ticle on the space-time manifold the value of is determined as c2 by the
following considerations.
For the particle on the space-time manifold having the Lorentz metric
(1, 1, 1, +1), we have
2
dtc
uc uc = (u4c )2 + ucs ucs = (c2 vcs
2
), (82)
d
dtc 1 + 20 c2
= (86)
d 1 vcs
2 c2
with only positive signs on the radicals because by denition both t and
are non-decreasing parameters.
Now for 2 |0| /c2 essentially zero compared to unity, Eq. (86) yields
! "
d = dtc 1 vcs
2 c2 , (87)
rst expression in Eq. (88) is equivalent to that deduced in the metric the-
ory of general relativity, valid for small particle velocity and a weak grav-
itational eld, as described by Adler et al.(19) pp. 131137. However, their
roles for tc and are interchanged here because here it is the coordinate
time tc that is dened in the reference frame of the moving particle, not
, as just emphasized above. These authors(19) also described conrmation
of this relationship to an accuracy of better than one percent in experi-
ments based on the Mossbauer effect reported by Pound and Rebka(20)
and Pound and Snider.(21) It is commented that this illustrates how the
gravitational eld, in the form of a non-Lorentzian metric, inuences mat-
ter, (or light) in its vicinity, but here this is based explicitly on the Lo-
rentz metric. Thus the Einstein concept of a gravitationally determined
metric is not required to account for this particular phenomenon. Actually,
Einstein(22) rst predicted a gravitatational red shift in 1911 based on the
Doppler effect and the equivalence of a uniform, linear gravitational eld
gradient and a uniform, linearly accelerated frame of reference.
Einsteins 1911 paper also predicted the deection of light by a grav-
itational eld but this employed a very obtuse argument about the effect
of a gravitational eld on the operation of clocks, which he claimed to
yield a linear relationship between the speed of light at a point and the
potential 0 existing at this point. He then multiplied this speed by dt,
or d in our notation, to provide displacement distances for points of a
wave front corresponding to different values of 0. These were then used in
an analysis based on Huyghens principle and Euclidian geometry to com-
pute the displaced trajectory of a light ray passing near the Sun. However,
these wave front displacements are given in our constructions by simply
multiplying dt c in Eq.
(88) by the invariant speed of light c, thus yielding
cdtc cd 1 + 0 c2 for the displacement. One can then proceed with
Einsteins construction for the deection of the light ray using Huyghens
principle.
Einsteins 1911
calculation
yielded the angular displacement of the ray
as E = 2Ms / c2 rm , with Ms /rm = |0|, Ms the solar mass, New-
tons gravitational constant and rm the radius from the solar center at
nearest approach. This is incorrect by a missing factor of two; the cor-
rected deection, = 2E , predicted by general relativity theory,(19)
has been conrmed by numerous astronomical observations. In 1995 this
was conrmed to very great accuracy using very-long-baseline-interferom-
etry observations of radio waves from the quasar 3C279, at three differ-
ent frequencies, within a few days of it being occulted by the Sun. This
employed observations in Massachusetts and California, Lebach et al.(23)
Thus the deection of light passing near the Sun seems to conrm gravi-
tation as the determining factor for manifold geometry.
774 Collins
d 2 xc 0
2
= g , = 1, 2, 3, 4. (90)
d x
Thus on this at manifold no constraint forces are required.
In Einsteins general theory of relativity, the gravitational eld merely
determines the manifold metric while the particle moves with zero acceler-
ation along a geodesic. Specically, Einstein(24) assumed the equations of
motion to be
d 2 xc dxc dxc
+ {, } = 0, = 1, 2, 3, 4, (91)
d 2 d d
with the {, } determined by the gravitational eld of material bodies
acting on the particle whose motion is under study. Thus this conicts
with the principle of equivalence.
The Mathematical Basis for Physical Laws 775
1 g g g
{, } = g + , (92)
2 x x x
and, since g = g for any Riemannian metric, there always exists a lin-
ear transformation of coordinate differentials which diagonalizes the met-
ric tensor. Thus for a diagonal metric, g = 0, = , with Eq. (92)
for the {, } it is readily veried that the constraint forces in Eq. (89)
776 Collins
are represented by
1 g
{, }u
c uc = g
u u (uc g )uc (93)
2 x c c
for = 1, 2, 3, 4. We also note that, from the derivations of Eqs. (79) and
(80) from our equation of motion in Eq. (78), the relationships,
1 0
uc uc = , = 1, 2, 3, 4, (94)
x 2 x
are true equalities. Thus, since g uc uc uc uc , Eqs. (93) appear as
0 uc
{, }u
c uc = g
g u
c
(u c g )u
c
x x (uc g )uc
(95)
d 2 xc 0
2
= 2g , = 1, 2, 3, 4, (96)
d x
which reveal that at this point on the manifold with curvature the particle
appears to have a at manifold acceleration caused by a gravitational force
twice that seen on a at manifold with only the radial component nonzero.
Furthermore in the Schwartzschild metric g rr = (1 2/r) which for a
weak eld, |/2r|
1, is essentially minus one. Therefore Eq. (96) yields
d 2 rc /d 2 20/r. The above statement assumes the 0/x to be at
least approximately equivalent to those on a at manifold. In Section 11 it
is shown that on a manifold with the Schwartzschild metric the eld 0 is
The Mathematical Basis for Physical Laws 777
described by our Eq. (58) in this metric, which is well approximated by the
DAlembertian equation in Eq. (71) in the weak eld limit. Therefore the
0/x are well approximated by the at manifold values. This factor of
two on the gravity potential 0 is consistent with the factor of two required
to correct Einsteins 1911 calculation of light deection by the Sun. Thus,
if deection of light by gravitation arises because a ray of light is equiva-
lent to a beam of particles of negligible mass, then this is a plausible expla-
nation for the required factor of two for this correction and is equivalent
to that provided by general relativity theory.
With this background for Einsteins general relativity equations of
motion, and their relationship to our mathematical constructions, we now
show how the covariant vector calculus formalism yields eld equations
constraining the manifold metric.
because all terms involving derivatives of the A add to zero. Thus this,
(1) (2)
with dRs = 1 dx and dRs = 1 dx , yields Eq. (97) as
#
A = dA = A R
1 dx dx (99)
S
1 for the vector elds, dened by
with the notation R
1 1
R 1 =
, (100)
x x x x
which characterize curvature of the x-manifold. Indeed, considering any
area element corresponding to dx dx on the coordinate surface x =con-
stant, having local normal vector 1 (x) this reveals that if the vector elds
in Eq. (100) are zero then directions of 1 at diagonal corners of the sur-
face element are parallel so this surface element is at, but if these elds
are not zero then the directions of 1 at these corners are not parallel,
indicating curvature of this x surface. Executing the derivatives of 1 in
Eq. (100) with d1 /dx = {, } 1 we nd that in conventional tensor
notation
R = { , } {, }
x x
+ { , } {, } {, } {, } , (101)
However, these Einstein eld equations merely assure the existence of sin-
gle-valued vector elds, including the gravitational eld, 0, in matter-free
space. Since all { , } = n g x and n g are zero for g = 1,
Einstein (24) eliminated the rst and last terms of this equation by requir-
ing local coordinates such that g = 1, but here this requirement is not
imposed.
However, when used in conjunction with the Einstein equations of
motion in Eq. (91), these Einstein eld equations do determine the gravita-
tional eld in matter-free space, as dened by Einstein with his extensions
to include eld sources described below. Alternatively, when used with Eq.
(89) these Einstein free-space eld equations merely determine the metric
of the space-time manifold, with this constrained only by symmetry con-
ditions. An example is the Schwartzschild metric described in Section 11.
The gravitational eld 0 on a curved space-time manifold is determined
by our Eq. (58) with JG expressed as 4M , just as in Eq. (71) for the
at Lorentzian manifold. The metric in these equations for 0 is therefore
determined by the Einstein eld equations in Eq. (103) and this same met-
ric must also be applied on this space-time manifold in the electro-mag-
netic eld equations which are exhibited in Eq. (53) for arbitrary manifold
dimension and metric. Thus an alternative explanation for the deection
of light by the Sun may reside in this gravitationally altered metric in
the eld equations of electromagnetism.
Einsteins general relativity eld equations, in domains free of gravita-
tional sources in Eq. (103), are expressed in our covariant vector notation
as the dyadic form
R = 1 1 R = 0. (104)
G = kT, (105)
780 Collins
1
G = R 1 1 g R, with R = R g , (106)
2
then not only does T = 0 imply all R = 0 but also this provides the
identity G 0 which, with Eq. (105) implies the identity T
0. Indeed, the numerical coefcient, 1/2 in G, was selected to assure
G 0 by Einstein(25) because with his chosen form for T the iden-
tity T 0 provides conservation laws for the momentum, energy and
mass of the source material of the gravitational eld.
It was in the design of the tensor T , having divergence zero, that
Einstein called upon empirical conservation laws of physics. In the covar-
iant vector notation this form is
T = M u u , (107)
{, }xc xc 0, = 1, 2, 3, 4. (110)
782 Collins
Hence the result is Eq. (90) with all 0/x zero. This merely reverses the
process used by Einstein to arrive at Eq. (91), as described in our Section
9. However, this result has other consequences which are exposed with Eq.
(92) for the {, }. With this Eq. (110) are
1 g g g
g + xc xc 0, = 1, 2, 3, 4. (111)
2 x x x
Then since xc = (dxc /dtc )(dtc /d ) for = 1, 2, 3, 4 and from Eq. (86)
dtc /d [(1 + 20/c2 )/(1 vcs 2 /c2 )]1/2 , it follows that dt /d 1 for
c
|20/c |
1 and vcs /c
1. Thus since dxc4 /dtc = c and |dxc /dtc |
c
2 2 2
1 g44 2
{, }xc xc g c 0, = 1, 2, 3, 4, (112)
2 x
for a weak gravitational eld and low to moderate particle velocities.
In particular for spherical symmetry, as described by the Schwartzschild
metric in Eq. (108), g44 = (1 2/r), so this yields
g44
2
c2 c , = 1, 2, 3, 4, (113)
x x r
12. CONCLUSIONS
Based on the fact that continuous variables are not precisely observ-
able we have shown that the motion of a particle on a Riemannian man-
ifold is properly described by a joint probability distribution dP (x, v | )
for the position and velocity of the particle conditioned by a continuous
time parameter as well as other assigned parameters. Then predicted
observed values for all observables associated with the particle, dened as
covariant functions of the x and v , are provided as expectation values
under dP . Based solely on the existence of this dP (x, v ) and the deni-
tion of particle velocity v by the identity dR(x)/d v with R(x) the
position vector of the particle, basic laws of mechanics, quantum mechan-
ics, electromagnetism, gravitation and relativity are derived as mathemat-
ical identities. Specically, by rst limiting the manifold to be at and
four-dimensional, we have obtained mathematical equations having the
same forms as these laws with physical constants occurring in these equa-
tions which must be properly assigned to achieve complete equivalence.
This physical identication process must also be applied to eld quanti-
ties which appear in the formalism based solely on mathematics, as say the
four-vector eld A as the electromagnetic potential. Thus we have shown
that these familiar laws, established by experiment and empirical deduc-
tion, must have these familiar forms based solely on laws of mathematics.
Indeed, even the manifold metric was determined as the Lorentz metric
based on mathematical requirements for this at manifold case.
These constructions have added to our understanding of these laws by
exposing a new role for the gravitational eld in special relativity theory
with a new relationship between proper time and coordinate time which
includes the gravitational eld as well as particle velocity. This exposes an
alternative to conventional interpretations of coordinate time and proper
time and provides an improved understanding of gravitational effects on
temporal events in the formalism of special relativity.
Einsteins identication of manifold geometry as a manifestation of
gravitation is also given a new understanding by these constructions in
the form of two representations for general relativity theory. One of these
is Einsteins original form while the other is based on our equation of
motion for a particle which includes the gravitational force as the gra-
dient of a scalar potential 0. In Section 9 the differences in these two
784 Collins
formalisms are pointed out and this new formalism is shown to provide
an alternative explanation for the deviation of a light ray passing near the
Sun. In this second form for general relativity Einsteins free-space eld
equations do not determine the gravitational eld 0 acting on a material
particle, instead these determine the metric of the space-time manifold on
which particle motion occurs. The scalar gravitational eld 0 then satises
a generalized inhomogeneous DAlembertian-type equation, expressed in
the metric determined by Einsteins free-space eld equations. This is illus-
trated by the example of the Schwartzschild metric for a spherically sym-
metric source at the spatial origin. Before this new theory can be accepted
as a real alternative to the Einstein theory it must be demonstrated that
it correctly predicts precession of planetary perihelions. This will require
a proper physical identication for the parameter in the Schwartzschild
metric. Indeed, much work remains to really evaluate this new theory.
It is also pointed out that this manifold metric must be applied in the
generalized equations for the electromagnetic eld which were written for
an arbitrary manifold dimension and metric in Eq. (53) of the text. Thus
these Einstein free-space eld equations determine the space-time metric
for all elds which may act on a particle. Thus we have a unied eld the-
ory but we must defer development of this to a future study.
Last but not least these constructions expose the intimate relationship
of classical probability theory, classical relativistic mechanics and quan-
tum mechanics by exposing a proper basis for the hydrodynamic repre-
sentation for the Schrodinger equation including the so-called quantum
potential as named by David Bohm.
REFERENCES
1. R. E. Collins, The Continuum and Wave Mechanics, Ph.D. dissertation, Texas A&M
University (1954).
2. R. R. Dedekind, Essays on the Theory of Numbers trans. by W. W. Beman (1901)
(Open Court, La Salle, Illinois, 1924).
3. L. Auslander and R. E. Mackenzie, Introduction to Differentiable Manifolds
(McGraw-Hill, New York, 1963).
4. W. Feller, An Introduction to Probability Theory and Its Applications, Vols. I and II
(Wiley, New York, 1950, 1966).
5. P. R. Halmos, Measure Theory (Van Nostrand, Princeton, NJ, 1950).
6. G. DeRham, Varietes Differentiables, Formes, Courants, Formes Harmonique (Her-
mann, Paris, 1955).
7. W. V. D. Hodge, The Theory and Application of Harmonic Integrals (Cambridge Uni-
versity Press, London, 1963).
8. A. O. Barut, Electrodynamics and Classical Theory of Fields and Particles (McMillan,
New York, 1964).
The Mathematical Basis for Physical Laws 785