This action might not be possible to undo. Are you sure you want to continue?
r
X
i
v
:
0
8
0
8
.
2
6
2
3
v
1
[
g
r

q
c
]
1
9
A
u
g
2
0
0
8
Diﬀeomorphism Invariance in the Hamiltonian formulation of
General Relativity
N. Kiriushcheva
∗
and S.V. Kuzmin
†
Faculty of Arts and Social Science, Huron University College,
N6G 1H3 and Department of Applied Mathematics,
University of Western Ontario, N6A 5B7, London, Canada
C. Racknor
‡
and S.R. Valluri
§
Physics and Astronomy Department and Department of Applied Mathematics,
University of Western Ontario, N6A 5B7, London, Canada
Abstract
It is shown that when the EinsteinHilbert Lagrangian is considered without any noncovariant
modiﬁcations or change of variables, its Hamiltonian formulation leads to results consistent with
principles of General Relativity. The ﬁrstclass constraints of such a Hamiltonian formulation, with
the metric tensor taken as a canonical variable, allow one to derive the generator of gauge trans
formations, which directly leads to diﬀeomorphism invariance. The given Hamiltonian formulation
preserves general covariance of the transformations derivable from it. This characteristic should
be used as the crucial consistency requirement that must be met by any Hamiltonian formulation
of General Relativity.
(Published in Phys. Lett. A 372 (2008) 5101)
∗
Electronic address: nkiriush@uwo.ca
†
Electronic address: skuzmin@uwo.ca
‡
Electronic address: cracknor@uwo.ca
§
Electronic address: valluri@uwo.ca
1
I. INTRODUCTION
The Hamiltonian formulation of General Relativity (GR) is more than a half century old.
To cast GR into Hamiltonian form, a method which deals with singular systems is required.
A way of dealing with this problem, the constrained dynamics, was proposed by Dirac in
1949 [1]. Dirac’s procedure was almost immediately applied by Pirani, Schild, and Skinner
[2] to GR using the covariant metric tensor as the canonical variable. However, the analysis
of [2] was not complete; the time development of secondary constraints was not considered
and, strictly speaking, the closure of the Dirac procedure was not demonstrated. Neither the
Dirac conjecture that all ﬁrstclass constraints generate gauge symmetries [3] nor explicit
methods for derivation of gauge transformations were known at that time.
A few years later, Dirac revisited this problem himself [4] with the same choice of a
canonical variable, the covariant metric tensor, and made a modiﬁcation of the Einstein
Hilbert (EH) Lagrangian to simplify the primary constraints. This modiﬁcation does not
aﬀect equations of motion but, according to Dirac, “can be achieved only at the expense of
abandoning fourdimensional symmetry” [4]. He explicitly demonstrated that the canonical
Hamiltonian is proportional to a linear combination of secondary constraints. However, as
in [2], it was not possible at that time to derive the gauge invariance which results from the
presence of ﬁrstclass constraints.
Shortly after the appearance of Dirac’s article [4], a new set of variables was introduced
by Arnowitt, Deser, and Misner (ADM) (see [5] and references therein) and the geometrical
interpretation of these new variables was widely emphasized and discussed (e.g. see [6]).
The Dirac conjecture [3] was converted into an algorithm [7] (see also [8, 9]) long after the
appearance of [2, 4, 5]. It was applied only to the most widely accepted formulation based
on the ADM variables. The complete derivation of the gauge transformations from the full
set of ﬁrstclass constraints in the ADM formulation actually appeared only recently [10].
However, a ﬁelddependent redeﬁnition of the gauge parameters found in [10] is required
to present the result in the form of diﬀeomorphism. These transformations [10] have been
known for a long time [11] and were partially derived in [7] for the ADM formulation as
an illustration of a general procedure for the derivation of gauge transformations. In [11]
the distinction between these transformations and the diﬀeomorphism transformation was
pointed out. In [10] this distinction was called “the unity of the diﬀerent symmetries” (our
2
Italic).
Why is it in the case of GR, where the Lagrangian and the equations of motion are
invariant under the diﬀeomorphism transformation
1
δ
(diff)
g
µν
= −ξ
µ;ν
−ξ
ν;µ
, (1)
does the Hamiltonian formulation lead to “unity”, instead of giving the equivalent result?
The Lagrangian and Hamiltonian formulations are supposed to give the same results; in
particular, the same gauge invariance. This is wellknown equivalence for ordinary gauge
ﬁeld theories. In addition, we refer to the result of Samanta [15] where (1) was derived using
the Lagrangian method [16] without introducing any new variables or modifying the EH
Lagrangian.
The ADM change of variables, or rather the geometrical meaning related to it, was
criticized by Hawking on general grounds as contradictory to “the whole spirit of relativity”
[17]. In our view, the disagreement between the results of [10] and [15] is a conﬁrmation
of Hawking’s criticism. To eliminate such a discrepancy, we reconsider the Hamiltonian
formulation of GR by neither using new variables (as was done by ADM) nor by making
additional modiﬁcations of the original action (as was done by Dirac). We revert to the ﬁrst
canonical treatment of GR in [2] and prove by explicit calculations that the Hamiltonian
formulation of GR with the metric tensor as the canonical variable leads to the symmetry
of (1), as derived in the Lagrangian approach of [15]. This is expected from a consistent
Hamiltonian formulation of GR.
1
In mathematical literature the term diﬀeomorphism refers to a mapping from one manifold to another,
which is diﬀerentiable, onetoone, onto, with diﬀerentiable inverse. However, in the literature on GR, the
word “diﬀeomorphism” is often used as equivalent of the transformation (1) (the semicolon “;” means a
covariant derivative) and in our article the latter meaning is employed. By a gauge invariance one usually
understands the invariance in the same coordinate frame of reference [12, 13] and with this respect the
transformation (1) serves as a gauge invariance of GR and can be written in variety of forms (e.g. (25) of
Section V is more suitable for comparison with the results there). Note that the righthand side of (1) is
in fact a Lie derivative of the metric tensor [14] which is obviously a true tensor, so it immediately follows
that it is not aﬀected by any change of coordinates (in other words, (1) is generally covariant).
3
II. THE HAMILTONIAN
The starting point of the Hamiltonian formulation in [2, 4] is the Γ − Γ part [14, 18] of
the EinsteinHilbert Lagrangian which is quadratic in ﬁrst order derivatives of the metric
tensor
L =
√
−gg
αβ
_
Γ
µ
αν
Γ
ν
βµ
−Γ
ν
αβ
Γ
µ
νµ
_
=
1
4
√
−gB
αβγµνρ
g
αβ,γ
g
µν,ρ
(2)
where
B
αβγµνρ
= g
αβ
g
γρ
g
µν
−g
αµ
g
βν
g
γρ
+ 2g
αρ
g
βν
g
γµ
−2g
αβ
g
γµ
g
νρ
.
To ﬁnd momenta π
αβ
conjugate to g
αβ
, we rewrite (2) to explicitly separate terms with
time derivatives of the covariant metric tensor (the “velocities”)
L =
1
4
√
−gB
αβ0µν0
g
αβ,0
g
µν,0
+
1
2
√
−gB
(αβ0µνk)
g
αβ,0
g
µν,k
+
1
4
√
−gB
αβkµνt
g
αβ,k
g
µν,t
(3)
where Latin alphabet is used for spatial components, 0 for temporal one, and () or (...  ...)
brackets indicate symmetrization in two indices or two groups of indices, i.e.,
B
(αβ)γµνρ
=
1
2
_
B
αβγµνρ
+ B
βαγµνρ
_
, B
(αβγµνρ)
=
1
2
_
B
αβγµνρ
+ B
µνραβγ
_
.
Using (3) we obtain
π
γσ
=
δL
δg
γσ,0
=
1
2
√
−gB
((γσ)0µν0)
g
µν,0
+
1
2
√
−gB
((γσ)0µνk)
g
µν,k
. (4)
The explicit form of the ﬁrst term is (see also [4])
1
2
√
−gB
((γσ)0µν0)
g
µν,0
=
1
2
√
−gg
00
E
µνγσ
g
µν,0
(5)
where
E
µνγσ
= e
µν
e
γσ
−e
µγ
e
νσ
, e
µν
= g
µν
−
g
0µ
g
0ν
g
00
. (6)
Note that both e
µν
and E
µνγσ
are zero, unless all µ, ν, γ and σ diﬀer from 0, and so
from (5) it immediately follows that we have d primary constraints (d is the dimension of
spacetime)
4
φ
0σ
= π
0σ
−
1
2
√
−gB
((0σ)0µνk)
g
µν,k
. (7)
For γσ = pq, equation (4) is invertible and giving
g
mn,0
= I
mnpq
1
g
00
_
2
√
−g
π
pq
−B
((pq)0µνk)
g
µν,k
_
(8)
where I
mnpq
is inverse of E
pqkl
:
I
mnpq
=
1
d −2
g
mn
g
pq
−g
mp
g
nq
, I
mnpq
E
pqkl
= δ
k
m
δ
l
n
. (9)
The appearance of the singularity in (9) for d = 2 corresponds to the fact, that in two
dimensions Eq. (4) cannot be solved for all velocities. So the number of primary constraints
equals the number of independent components of the metric tensor.
The Hamiltonian is deﬁned to be H = π
αβ
g
αβ,0
− L, which after elimination of the
velocities using (8) gives the total Hamiltonian
H
T
= H
c
+ g
0σ,0
φ
0σ
(10)
where the canonical Hamiltonian H
c
is
H
c
=
1
√
−gg
00
I
mnpq
π
mn
π
pq
−
1
g
00
I
mnpq
π
mn
B
(pq0µνk)
g
µν,k
(11)
+
1
4
√
−g
_
1
g
00
I
mnpq
B
((mn)0µνk)
B
(pq0αβt)
−B
µνkαβt
_
g
µν,k
g
αβ,t
.
III. CLOSURE OF THE DIRAC PROCEDURE
The fundamental PB are deﬁned by
{g
αβ
, π
µν
} =
1
2
_
δ
µ
α
δ
ν
β
+ δ
ν
α
δ
µ
β
_
≡ ∆
µν
αβ
and, following the standard procedure (see [3, 16, 19]), we have to calculate time development
of primary constraints, {φ
0σ
, H
T
}. Using (7) we now ﬁnd the PB among primary constraints
_
φ
0σ
, φ
0α
_
= 0 (12)
5
while, because H
c
is independent of π
0α
(see (11)), {φ
0σ
, H
c
} cannot be proportional to
primary constraints. This leads to the secondary constraints:
χ
0σ
= −
1
2
1
√
−g
g
0σ
g
00
I
mnpq
π
mn
π
pq
(13)
+
1
2
g
0σ
g
00
I
mnpq
π
mn
A
(pq0µνk)
g
µν,k
+ δ
σ
m
_
π
mk
,k
+
_
π
pk
e
qm
−
1
2
π
pq
e
km
_
g
pq,k
_
−
1
8
√
−g
_
g
0σ
g
00
I
mnpq
B
((mn)0µνk)
B
(pq0αβt)
−g
0σ
B
µνkαβt
_
g
µν,k
g
αβ,t
+
1
4
√
−g
1
g
00
I
mnpq
B
((mn)0µνk)
g
µν,k
g
αβ,t
_
g
σt
_
g
00
g
pα
g
qβ
+ g
pq
g
0α
g
0β
−2g
αq
g
0p
g
0β
_
−g
σp
_
2g
00
g
qα
g
tβ
−g
00
g
αβ
g
qt
+ g
αβ
g
0q
g
0t
−2g
qα
g
0β
g
0t
−2g
tα
g
0β
g
0q
+ 2g
qt
g
0α
g
0β
_
+g
0σ
_
2g
βt
g
αp
g
0q
−2g
pα
g
qβ
g
0t
−2g
pq
g
tβ
g
0α
+ 2g
pt
g
qβ
g
0α
+ g
pq
g
αβ
g
0t
−g
tp
g
αβ
g
0q
_¸
−
1
4
√
−gg
µν,k
g
αβ,t
_
g
σt
_
g
αµ
g
βν
g
0k
+ g
µν
g
αt
g
0β
−2g
µα
g
kν
g
0β
_
+ g
0σ
_
2g
αt
g
βµ
g
νk
−3g
tµ
g
νk
g
αβ
−2g
µα
g
νβ
g
kt
+ g
µν
g
kt
g
αβ
+ 2g
µt
g
νβ
g
kα
_
+g
σµ
__
g
αβ
g
νt
−2g
να
g
tβ
_
g
0k
+ 2
_
g
βν
g
kt
−g
βk
g
tν
_
g
0α
+
_
2g
kβ
g
αt
−g
αβ
g
kt
_
g
0ν
_¸
−
1
2
√
−gg
00
E
pqtσ
_
1
g
00
I
mnpq
B
((mn)0µνk)
g
µν,k
_
,t
−
1
2
√
−gB
((σ0)kαβt)
g
αβ,kt
where
A
αβ0µνk
= B
(αβ0µνk)
−g
0k
E
αβµν
+ 2g
0µ
E
αβkν
. (14)
We observe that the primary constraints (7) are the same as those derived in [2] (see Eq.
(6) of [2]) and provide the correct limit of linearized GR [20], but our (11) is diﬀerent from
H
c
of [2] (see Eq. (8) of [2]). While Eq. (11) gives the correct limit of linearized GR, the
H
c
of [2] does not.
Of course, Eq. (13) can be presented in diﬀerent forms, however this form is convenient
when performing an additional consistency check. Having H
c
proportional to the secondary
constraints is a common feature of all generally covariant theories (e.g., see [21]) and this
can be seen as a good veriﬁcation of the correctness of (13). Comparing (11) and (13) we
obtain
H
c
= −2g
0σ
χ
0σ
+
_
2g
0m
π
mk
−
√
−g
_
g
0n
B
((nk)0αβt)
+ g
0γ
B
((0γ)kαβt)
_
g
αβ,t
_
,k
. (15)
6
Let us continue to apply the Dirac procedure and consider the time derivatives of sec
ondary constraints, {χ
0σ
, H
T
}. We ﬁnd that
_
χ
0σ
, φ
0γ
_
= −
1
2
g
σγ
χ
00
(16)
and
_
χ
0σ
, H
c
_
= −
2
√
−g
I
mnpq
π
mn
g
σq
g
00
χ
0p
+
1
2
g
σk
g
00,k
χ
00
+ δ
σ
0
χ
0k
,k
(17)
+
_
−2
1
√
−g
I
mnpk
π
mn
g
σp
g
00
+ I
mkpq
g
αβ,t
g
σm
g
00
A
(pq)0αβt
_
χ
0k
−
_
g
0σ
g
00,k
+ 2g
nσ
g
0n,k
+ g
nσ
g
0m
g
00
(g
mn,k
+ g
km,n
−g
kn,m
)
_
χ
0k
.
This completes the proof of closure of the Dirac procedure, i.e., no further constraints
arise. According to (12), (15), (16), and (17), all constraints are ﬁrstclass. This is suﬃcient
to derive the gauge transformations. A number of algorithms of such a derivation are
available [8, 9]. We follow the approach that was applied for the ﬁrst time to ﬁeld theories
by Castellani [7].
IV. THE GENERATORS
The Castellani procedure [7] is based on a derivation of generators of gauge transfor
mations which are deﬁned by chains of ﬁrstclass constraints. One starts with primary
ﬁrstclass constraint(s), i = 1, 2, ..., and from them constructs chain(s) ξ
(n)
i
G
i
(n)
. Here ξ
(n)
i
is
a time derivative of the ith gauge parameter of order n (n = 0, 1, ...) and n corresponds to
the length of chain. The number of gauge parameters is equal to the number of ﬁrstclass
primary constraints. Note, that the chains are unambiguous when the primary constraints
are deﬁned. The functions G
i
(n)
are calculated as follows:
G
σ
(1)
(x) = φ
0σ
(x) , (18)
G
σ
(0)
(x) = −
_
φ
0σ
(x) , H
T
_
+
_
α
σ
γ
(x, y) φ
0γ
(y) d
3
y, (19)
where functions α
σ
γ
(x, y) have to be chosen so that the chain ends on the surface of primary
constraints
7
_
G
σ
(0)
, H
T
_
= primary (20)
and the generator G(ξ
σ
) is given by
G(ξ
σ
) = ξ
σ
G
σ
(0)
+ ξ
σ,0
G
σ
(1)
. (21)
To construct the generator (21), we have to ﬁnd functions α
σ
γ
(x, y) using the condition (20)
_
G
σ
(0)
, H
T
_
= −
_
χ
0σ
(x) , H
T
_
+
_
_
α
σ
γ
, H
T
_
φ
0γ
(y) d
3
y +
_
α
σ
γ
_
φ
0γ
(y) , H
T
_
d
3
y. (22)
The second term of (22) is irrelevant because it is zero on a surface of primary constraints,
according to the condition (20). The rest of the PBs being known (see (16) and (17)) allows
one to just read oﬀ the functions α
σ
γ
(x, y) from (22) and obtain an explicit expression for
G
σ
(0)
G
σ
(0)
(x) = −χ
0σ
−
_
g
00,0
1
2
g
0σ
+ g
0m,0
g
σm
−
1
2
g
σk
g
00,k
_
φ
00
−δ
σ
0
φ
0k
,k
(23)
+
_
−2
1
√
−g
I
mnpk
π
mn
g
σp
g
00
+ I
mkpq
g
αβ,t
g
σm
g
00
A
(pq)0αβt
_
φ
0k
−
_
g
0σ
g
00,k
+ 2g
nσ
g
0n,k
+ g
nσ
g
0m
g
00
(g
mn,k
+ g
km,n
−g
kn,m
)
_
φ
0k
.
This completes the calculation of the generator (21).
V. TRANSFORMATION OF THE METRIC TENSOR
Transformations of ﬁelds can be found by calculating the following PB
δg
µν
= {G, g
µν
} =
_
ξ
σ
G
σ
(0)
+ ξ
σ,0
φ
0σ
, g
µν
_
. (24)
Let us compare the result of this PB with (1) which we present in equivalent but more
suitable form for comparison with (24)
δ
(diff)
g
µν
= −ξ
µ,ν
−ξ
ν,µ
+ g
αβ
(g
µβ,ν
+ g
νβ,µ
−g
µν,β
) ξ
α
. (25)
For the timetime component, g
00
, (24) gives
8
δg
00
= {G, g
00
} = −
δ
δπ
00
_
ξ
σ
G
σ
(0)
+ ξ
σ,0
φ
0σ
_
so that nonzero contributions come only from those terms proportional to the primary
constraint φ
00
δg
00
= ξ
σ
_
g
00,0
1
2
g
0σ
+ g
0k,0
g
σk
−
1
2
g
σk
g
00,k
_
−ξ
0,0
. (26)
Putting µ = ν = 0 and partially separating the space and time indices in Eq. (25), we have
δ
(diff)
g
00
= −2ξ
0,0
+ g
σ0
g
00,0
ξ
σ
+ g
σk
(2g
0k,0
−g
00,k
) ξ
σ
which is equivalent to (26) up to a numerical factor
2 {g
00
, G} = δ
(diff)
g
00
(27)
that can be incorporated into a parameter ξ
σ
→2ξ
σ
.
For the transformation of the spacetime components, g
0k
, we need the corresponding
part of the generator depending on φ
0s
δg
0k
= −
1
2
ξ
0,k
−
1
2
ξ
k,0
−
1
2
ξ
σ
_
−2
1
√
−g
I
mnpk
π
mn
g
σp
g
00
+ I
mkpq
g
αβ,t
g
σm
g
00
A
(pq)0αβt
_
(28)
+
1
2
ξ
σ
_
g
0σ
g
00,k
+ 2g
nσ
g
0n,k
+ g
nσ
g
0m
g
00
(g
mn,k
+ g
km,n
−g
kn,m
)
_
.
In the ﬁrst bracket of (28) we substitute momenta in terms of the metric (4) so that using
(14) we obtain
−
1
2
ξ
σ
g
σm
g
00
_
−g
0n
g
mk,n
+ g
0α
g
αk,m
+ g
0α
g
αm,k
_
and ﬁnally, after a simple rearrangement of terms in (28), we have
δg
0k
= −
1
2
ξ
0,k
−
1
2
ξ
k,0
+
1
2
ξ
σ
g
0σ
g
00,k
+
1
2
ξ
σ
g
nσ
(−g
0k,n
+ g
0n,k
+ g
nk,0
) . (29)
This is equivalent to the transformations of the corresponding components of (25) up to the
same numerical factor that occurs in (27).
For the transformation of the spacespace components, g
nm
, we need to keep only terms
in G with π
pq
dependence, so that
9
δg
nm
= {G, g
nm
} =
δ
δπ
nm
_
ξ
σ
χ
0σ
−ξ
σ
2
√
−g
I
abpk
π
ab
g
σp
g
00
φ
0k
_
.
The second term produces contributions proportional to the primary constraints and
can be neglected on the surface of primary constraints. In χ
0σ
we need only momentum
dependent terms appearing in (13). We obtain then
δg
nm
= −
1
2
ξ
n,m
−
1
2
ξ
m,n
+ ξ
σ
δ
σ
a
_
e
qa
∆
pk
nm
−
1
2
∆
pq
nm
e
ka
_
g
pq,k
+ ξ
σ
g
0σ
g
00
1
2
_
−
1
√
−g
I
abnm
2π
ab
+ I
nmcd
A
(cd)0µνk
g
µν,k
_
.
After substitution of π
ab
from (4) and using (6), (9), and (14), we have
δg
nm
= −
1
2
ξ
n,m
−
1
2
ξ
m,n
+
1
2
ξ
0
_
g
00
(g
0m,n
+ g
0n,m
−g
nm,0
) + g
0p
(g
pm,n
+ g
pn,m
−g
nm,p
)
_
(30)
+
1
2
ξ
p
_
g
qp
(g
nq,m
+ g
mq,n
−g
nm,q
) + g
0p
(g
0m,n
+ g
0n,m
−g
nm,0
)
_
which is again equivalent with (25) up to the same numerical factor of 2.
The transformations of the components of g
µν
in (26), (29), (30) can be combined into
one covariant expression (25) or (1). This completes the proof of our original statement that
the Hamiltonian formulation gives the same result for the gauge invariance of GR as the
Lagrangian formulation [15], as it should. Moreover, this also demonstrates that we obtained
the consistent Hamiltonian formulation of GR because the transformation (1) derivable from
it is generally covariant.
VI. CONCLUSION
We would like to point out some peculiarities of the Hamiltonian formulation of GR.
First of all, the transformation (30) is valid on the surface of primary constraints. In
ordinary gauge ﬁeld theories (e.g., Maxwell or YangMills) this does not happen, because
the righthand side of (20) equals zero exactly. Another peculiarity of GR is the PB among
primary and secondary constraints is not zero by (16). Such deviations from ordinary
gauge theories actually can be expected. In ordinary ﬁeld theories, equations of motion are
10
exactly invariant under a gauge transformation, whereas the Einstein equations of motion,
G
µν
= R
µν
−
1
2
g
µν
R = 0, transform into a combination of the same equations [12], i.e. are
invariant only onshell
δ
(diff)
G
µν
= −ξ
ρ
G
µν,ρ
−ξ
ρ
,µ
G
νρ
−ξ
ρ
,ν
G
µρ
. (31)
In particular, because of this, the “crucial test” of having oﬀshell closure of the constraint
algebra [22] seems as unreasonably strong for GR. Moreover, it is impossible to expect
oﬀshell closure for generators of a transformation which by itself produces only onshell
invariance of the equations of motion (31).
Going back to the “spiritual” statement of Hawking, we can make the conjecture or
possibly draw the conclusion that if diﬀeomorphism is derivable through the Lagrangian
or Hamiltonian approach, then the spirit of GR is “alive”, as the mathematical essence of
the spirit of GR is to retain general covariance and that diﬀeomorphism can be derived
from its structure. Vice versa, if diﬀeomorphism cannot be obtained in the Lagrangian or
Hamiltonian formulation, then such a formulation is not the equivalent to GR. In particular,
our Hamiltonian formulation of GR which does not resort to any change of variables or
modiﬁcations gives the same diﬀeomorphism invariance as the Lagrangian approach [15].
On the contrary, the ADM Hamiltonian formulation gives diﬀerent transformations which
can be presented in the form of diﬀeomorphism only after ﬁelddependent and noncovariant
redeﬁnition of parameters [10]:
ε
⊥
ADM
=
_
−g
00
_
−1/2
ξ
0
diff
, ε
k
ADM
= ξ
k
diff
−
g
0k
g
00
ξ
0
diff
. (32)
Moreover, the equivalence of the Lagrangian and Hamiltonian formulations implies that
the ADM Lagrangian should give the same ﬁelddependent redeﬁnition of gauge parameters
(32) if treated by the Lagrangian method of [15], and it is the expected result [23]. The
equivalence of results in both formalisms for asymptotically ﬂat spacetimes was discussed
in [24].
In this paper we demonstrated that the consistent Hamiltonian formulation of GR can
be obtained by considering the metric tensor as the canonical variable and not performing
any change of variables. Following the Dirac conjecture and applying the Castellani proce
dure, we derived the gauge generators from the full set of ﬁrstclass constraints. From these
11
generators, without redeﬁnition of gauge parameters, we explicitly derived diﬀeomorphism
invariance in the Hamiltonian formulation of GR for the ﬁrst time. In the Hamiltonian for
mulation of GR, the covariance is not manifest but the ﬁnal result on gauge transformation
is presented in manifestly covariant form as in the Hamiltonian formulation of ordinary ﬁeld
theories. Our result is another illustration of the resiliency of Einstein’s General Relativity
and any approach, if correctly used, cannot violate its symmetry and leads to the generally
covariant transformation (1).
Note added in proof
After completion of this work it came to our attention that, by a diﬀerent from [9] method
which was used in [10], the “speciﬁc metricdependent diﬀeomorphism” (32) of the ADM
formulation was also obtained and extensively discussed in [25].
VII. ACKNOWLEDGEMENTS
We would like to thank A.M. Frolov, D.G.C. McKeon, V.I. Rupasov and A.V. Zvelin
dovsky for discussions and W.G. Cliﬀ for inspiration and support. The helpful correspon
dence with A.N. Petrov and S. Samanta is gratefully acknowledged. The research was
partially supported by the Huron University College Faculty of Arts and Social Science
Research Grant Fund.
[1] P.A.M. Dirac, Can. J. Math. 2 (1950) 129.
[2] F.A.E. Pirani, A. Schild and R. Skinner, Phys. Rev. 87 (1952) 452.
[3] P.A.M. Dirac, Lectures on Quantum Mechanics, Belfer Graduate School of Sciences, Yeshiva
University, New York, 1964.
[4] P.A.M. Dirac, Proc. Roy. Soc. A 246 (1958) 333.
[5] R. Arnowitt, S. Deser and C.W. Misner, in: L. Witten (Ed.), Gravitation: An Introduction
to Current Research, Wiley, New York, 1962, p. 227.
[6] K. Kuchar, in: W. Israel (Ed.), Relativity, Astrophysics and Cosmology, Reidel Publ. Comp.,
Dordrecht, 1973, p. 237.
[7] L. Castellani, Ann. Phys. 143 (1982) 357.
12
[8] M. Henneaux, C. Teitelboim and J. Zanalli, Nucl. Phys. B 332 (1990) 169.
[9] R. Banerjee, H.J. Rothe and K.D. Rothe, Phys. Lett. B 479 (2000) 429; R. Banerjee, H.J.
Rothe and K.D. Rothe, Phys. Lett. B 462 (1999) 248.
[10] P. Mukherjee and A. Saha, arXiv: 0705.4358 [hepth].
[11] P.G. Bergmann and A. Komar, Int. J. Theor. Phys. 1 (1972) 15.
[12] L.P. Grishchuk, A.N. Petrov and A.D. Popova, Commun. Math. Phys. 94 (1984) 379.
[13] A.D. Popova and A.N. Petrov, Int. J. Mod. Phys. A 3 (1988) 2651.
[14] M. Carmeli, Classical Fields: General Relativity and Gauge Theory, John Wiley & Sons, 1982.
[15] S. Samanta, arXiv: 0708.3300 [hepth].
[16] D.M. Gitman and I.V. Tyutin, Quantization of Fields with Constraints, Springer, Berlin,
1990.
[17] S.W. Hawking, in: S.W. Hawking and W. Israel (Eds.), General Relativity. An Einstein
Centenary Survey, Cambridge University Press, Cambridge, 1979, p. 746.
[18] L.D. Landau and E.M. Lifshitz, The Classical Theory of Fields, fourth ed., Pergamon Press,
Oxford, 1975.
[19] K. Sundermeyer, Constrained Dynamics, Lecture Notes in Physics, vol. 169, Springer, Berlin,
1982.
[20] K.R. Green, N. Kiriushcheva and S.V. Kuzmin, arXiv: 0710.1430 [grqc].
[21] G. F¨ ul¨ op, D. M. Gitman and I. V. Tyutin, Int. J. Theor. Phys. 38 (1999) 1941.
[22] H. Nicolai, K. Peeters and M. Zamaklar, Class. Quantum Grav. 22 (2005) R193.
[23] S. Samanta, private communication.
[24] A.N. Petrov, Int. J. Mod. Phys. D 4 (1995) 451; A.N. Petrov, Int. J. Mod. Phys. D 6 (1997)
239.
[25] J.M. Pons, D.C. Salisbury, L.C. Shepley, Phys. Rev. D 55 (19970 658; J.M. Pons, Class.
Quantum Grav. 20 (2003) 3279.
13
4. Schild. the analysis of [2] was not complete. according to Dirac. as in [2]. Dirac revisited this problem himself [4] with the same choice of a canonical variable. Shortly after the appearance of Dirac’s article [4]. and Skinner [2] to GR using the covariant metric tensor as the canonical variable. a ﬁelddependent redeﬁnition of the gauge parameters found in [10] is required to present the result in the form of diﬀeomorphism. This modiﬁcation does not aﬀect equations of motion but. Deser.I. and made a modiﬁcation of the EinsteinHilbert (EH) Lagrangian to simplify the primary constraints. the closure of the Dirac procedure was not demonstrated. In [10] this distinction was called “the unity of the diﬀerent symmetries” (our 2 . Dirac’s procedure was almost immediately applied by Pirani. strictly speaking. He explicitly demonstrated that the canonical Hamiltonian is proportional to a linear combination of secondary constraints. “can be achieved only at the expense of abandoning fourdimensional symmetry” [4]. was proposed by Dirac in 1949 [1]. Neither the Dirac conjecture that all ﬁrstclass constraints generate gauge symmetries [3] nor explicit methods for derivation of gauge transformations were known at that time. INTRODUCTION The Hamiltonian formulation of General Relativity (GR) is more than a half century old. see [6]). and Misner (ADM) (see [5] and references therein) and the geometrical interpretation of these new variables was widely emphasized and discussed (e. 9]) long after the appearance of [2. a method which deals with singular systems is required. A way of dealing with this problem. To cast GR into Hamiltonian form. However. the covariant metric tensor. In [11] the distinction between these transformations and the diﬀeomorphism transformation was pointed out. a new set of variables was introduced by Arnowitt. The Dirac conjecture [3] was converted into an algorithm [7] (see also [8. the time development of secondary constraints was not considered and. it was not possible at that time to derive the gauge invariance which results from the presence of ﬁrstclass constraints. A few years later. It was applied only to the most widely accepted formulation based on the ADM variables. The complete derivation of the gauge transformations from the full set of ﬁrstclass constraints in the ADM formulation actually appeared only recently [10]. the constrained dynamics. 5]. These transformations [10] have been known for a long time [11] and were partially derived in [7] for the ADM formulation as an illustration of a general procedure for the derivation of gauge transformations. However. However.g.
the same gauge invariance. This is expected from a consistent Hamiltonian formulation of GR. (1) is generally covariant). By a gauge invariance one usually understands the invariance in the same coordinate frame of reference [12. where the Lagrangian and the equations of motion are invariant under the diﬀeomorphism transformation1 δ(dif f ) gµν = −ξµ.” means a covariant derivative) and in our article the latter meaning is employed. so it immediately follows that it is not aﬀected by any change of coordinates (in other words. To eliminate such a discrepancy. we reconsider the Hamiltonian formulation of GR by neither using new variables (as was done by ADM) nor by making additional modiﬁcations of the original action (as was done by Dirac). the word “diﬀeomorphism” is often used as equivalent of the transformation (1) (the semicolon “. in the literature on GR. as derived in the Lagrangian approach of [15]. onto. In addition. with diﬀerentiable inverse. instead of giving the equivalent result? The Lagrangian and Hamiltonian formulations are supposed to give the same results. Why is it in the case of GR. (25) of Section V is more suitable for comparison with the results there). or rather the geometrical meaning related to it. 1 In mathematical literature the term diﬀeomorphism refers to a mapping from one manifold to another.Italic). in particular. However. we refer to the result of Samanta [15] where (1) was derived using the Lagrangian method [16] without introducing any new variables or modifying the EH Lagrangian. was criticized by Hawking on general grounds as contradictory to “the whole spirit of relativity” [17].ν − ξν. 3 .g. Note that the righthand side of (1) is in fact a Lie derivative of the metric tensor [14] which is obviously a true tensor. onetoone. the disagreement between the results of [10] and [15] is a conﬁrmation of Hawking’s criticism. which is diﬀerentiable. The ADM change of variables. This is wellknown equivalence for ordinary gauge ﬁeld theories. (1) does the Hamiltonian formulation lead to “unity”.µ . In our view. 13] and with this respect the transformation (1) serves as a gauge invariance of GR and can be written in variety of forms (e. We revert to the ﬁrst canonical treatment of GR in [2] and prove by explicit calculations that the Hamiltonian formulation of GR with the metric tensor as the canonical variable leads to the symmetry of (1).
eµν = g µν − g 0µ g 0ν . B (αβ)γµνρ = 1 1 B αβγµνρ + B βαγµνρ . g 00 (4) The explicit form of the ﬁrst term is (see also [4]) (5) (6) Note that both eµν and E µνγσ are zero. i. 0 for temporal one. unless all µ.e.0 + −gB (αβ0µνk) gαβ.0 2 2 where E µνγσ = eµν eγσ − eµγ eνσ . γ and σ diﬀer from 0. δgγσ. 2 2 Using (3) we obtain π γσ = 1√ 1√ δL = −gB ((γσ)0µν0) gµν. 18] of the EinsteinHilbert Lagrangian which is quadratic in ﬁrst order derivatives of the metric tensor √ −gg αβ Γµ Γν − Γν Γµ = αν βµ αβ νµ 1√ −gB αβγµνρ gαβ.. To ﬁnd momenta π αβ conjugate to gαβ .ρ 4 L= where (2) B αβγµνρ = g αβ g γρg µν − g αµ g βν g γρ + 2g αρ g βν g γµ − 2g αβ g γµ g νρ.. 4] is the Γ − Γ part [14.k . and so from (5) it immediately follows that we have d primary constraints (d is the dimension of spacetime) 4 . ν.t 4 2 4 (3) where Latin alphabet is used for spatial components.0 = −gg 00 E µνγσ gµν. B (αβγµνρ) = B αβγµνρ + B µνραβγ . we rewrite (2) to explicitly separate terms with time derivatives of the covariant metric tensor (the “velocities”) L= 1√ 1√ 1√ −gB αβ0µν0 gαβ..) brackets indicate symmetrization in two indices or two groups of indices.. THE HAMILTONIAN The starting point of the Hamiltonian formulation in [2.k gµν.0 2 2 1√ 1√ −gB ((γσ)0µν0) gµν.0 gµν.k + −gB αβkµνt gαβ.0 gµν.0 + −gB ((γσ)0µνk) gµν.γ gµν. and () or (.II.  ..
0 − L.0 = Imnpq where Imnpq is inverse of E pqkl : Imnpq = 1 g 00 2 √ π pq − B ((pq)0µνk) gµν. HT }. Imnpq E pqkl = δm δn . Using (7) we now ﬁnd the PB among primary constraints φ0σ . The Hamiltonian is deﬁned to be H = π αβ gαβ.k −g (8) The appearance of the singularity in (9) for d = 2 corresponds to the fact. CLOSURE OF THE DIRAC PROCEDURE The fundamental PB are deﬁned by {gαβ .0 φ0σ where the canonical Hamiltonian Hc is 1 k l gmn gpq − gmp gnq . d−2 (9) (10) Hc = √ 1 1 I π mn π pq − 00 Imnpq π mn B (pq0µνk) gµν. which after elimination of the velocities using (8) gives the total Hamiltonian HT = Hc + g0σ. φ0α = 0 (12) 5 .t . that in two dimensions Eq. (4) cannot be solved for all velocities. following the standard procedure (see [3. + 4 g (11) III. 19]).φ0σ = π 0σ − 1√ −gB ((0σ)0µνk) gµν. we have to calculate time development of primary constraints. So the number of primary constraints equals the number of independent components of the metric tensor. π µν } = 1 µ ν ν µ δ δ + δα δβ ≡ ∆µν αβ 2 α β and. 2 (7) For γσ = pq. 16.k 00 mnpq −gg g 1 1√ −g 00 Imnpq B ((mn)0µνk) B (pq0αβt) − B µνkαβt gµν. {φ0σ . equation (4) is invertible and giving gmn.k .k gαβ.
While Eq. (11) gives the correct limit of linearized GR.t g σt g αµ g βν g 0k + g µν g αt g 0β − 2g µα g kν g 0β 4 + g 0σ 2g αt g βµ g νk − 3g tµ g νk g αβ − 2g µα g νβ g kt + g µν g ktg αβ + 2g µt g νβ g kα +g σµ − where Aαβ0µνk = B (αβ0µνk) − g 0k E αβµν + 2g 0µ E αβkν .k gαβ.t − −g 8 g 00 1√ 1 + −g 00 Imnpq B ((mn)0µνk) gµν.kt 2 (14) We observe that the primary constraints (7) are the same as those derived in [2] (see Eq.k + π pk eqm − π pq ekm gpq. the Hc of [2] does not. however this form is convenient when performing an additional consistency check.k . Having Hc proportional to the secondary constraints is a common feature of all generally covariant theories (e. (8) of [2]).k + δm π. see [21]) and this can be seen as a good veriﬁcation of the correctness of (13).k gαβ.k + 00 mnpq 2g 2 0σ 1√ g Imnpq B ((mn)0µνk) B (pq0αβt) − g 0σ B µνkαβt gµν.t (13) − 1√ −gB ((σ0)kαβt) gαβ.k g 00 . because Hc is independent of π 0α (see (11)).k gαβ. This leads to the secondary constraints: 1 1 g 0σ χ0σ = − √ Imnpq π mn π pq 2 −g g 00 1 1 g 0σ σ mk I π mn A(pq0µνk) gµν.t g σt g 00 g pαg qβ + g pq g 0α g 0β − 2g αq g 0p g 0β 4 g − g σp 2g 00 g qαg tβ − g 00 g αβ g qt + g αβ g 0q g 0t − 2g qα g 0β g 0t − 2g tα g 0β g 0q + 2g qt g 0α g 0β +g 0σ 2g βtg αp g 0q − 2g pαg qβ g 0t − 2g pq g tβ g 0α + 2g ptg qβ g 0α + g pq g αβ g 0t − g tp g αβ g 0q 1√ − −ggµν. {φ0σ . Hc } cannot be proportional to primary constraints.. but our (11) is diﬀerent from Hc of [2] (see Eq. Comparing (11) and (13) we obtain Hc = −2g0σ χ0σ + 2g0m π mk − √ −g g0n B ((nk)0αβt) + g0γ B ((0γ)kαβt) gαβ. (15) . Eq. 1√ −gg 00 E pqtσ 2 g αβ g νt − 2g ναg tβ g 0k + 2 g βν g kt − g βk g tν g 0α + 2g kβ g αt − g αβ g kt g 0ν 1 Imnpq B ((mn)0µνk) gµν. (6) of [2]) and provide the correct limit of linearized GR [20].while.t 6 .g. (13) can be presented in diﬀerent forms. Of course.
e. HT }.. One starts with primary ﬁrstclass constraint(s).k + g nσ 00 (gmn.k + 2g nσ g0n. A number of algorithms of such a derivation are available [8.. i. all constraints are ﬁrstclass... The number of gauge parameters is equal to the number of ﬁrstclass primary constraints. According to (12). (19) σ where functions αγ (x. φ0γ = − g σγ χ00 2 and 1 g σq 2 σ χ0σ . Here ξi (n) (n) (n) is a time derivative of the ith gauge parameter of order n (n = 0. that the chains are unambiguous when the primary constraints are deﬁned.m) χ0k . 9]. (16). We follow the approach that was applied for the ﬁrst time to ﬁeld theories by Castellani [7].k + gkm.. {χ0σ . i = 1. (1) (18) Gσ (x) = − φ0σ (x) .k χ00 + δ0 χ0k . and (17).Let us continue to apply the Dirac procedure and consider the time derivatives of secondary constraints. and from them constructs chain(s) ξi Gi . . This is suﬃcient to derive the gauge transformations. y) φ0γ (y) d3 y. The functions Gi are calculated as follows: (n) Gσ (x) = φ0σ (x) . g (16) (17) This completes the proof of closure of the Dirac procedure. IV. no further constraints arise.. Note.) and n corresponds to the length of chain. . 2. HT + (0) constraints 7 σ αγ (x. We ﬁnd that 1 χ0σ . THE GENERATORS The Castellani procedure [7] is based on a derivation of generators of gauge transformations which are deﬁned by chains of ﬁrstclass constraints. Hc = − √ Imnpq π mn 00 χ0p + g σk g00. y) have to be chosen so that the chain ends on the surface of primary .n − gkn.t 00 A(pq)0αβt χ0k −g g g g 0m − g 0σ g00. (15). 1.k −g g 2 1 g σm g σp + −2 √ Imnpk π mn 00 + Imkpq gαβ.
(23) V. we have to ﬁnd functions αγ (x.µ − gµν. y) using the condition (20) σ G(0) . (24) gives 8 (24) Let us compare the result of this PB with (1) which we present in equivalent but more (25) . (0) suitable form for comparison with (24) δ(dif f ) gµν = −ξµ.0 g 0σ + g0m.0 Gσ . HT φ0γ (y) d3 y + σ αγ φ0γ (y) . HT = − χ0σ (x) . The rest of the PBs being known (see (16) and (17)) allows σ one to just read oﬀ the functions αγ (x.ν + gνβ. gµν .µ + g αβ (gµβ. (22) The second term of (22) is irrelevant because it is zero on a surface of primary constraints.k + gkm. y) from (22) and obtain an explicit expression for σ G(0) 1 1 σ Gσ (x) = −χ0σ − g00. HT d3 y. (0) (1) (20) (21) σ To construct the generator (21). gµν } = ξσ Gσ + ξσ.k + g nσ 00 (gmn.k (0) 2 2 σp g σm (pq)0αβt 1 mn g φ0k + Imkpq gαβ. g00 . g This completes the calculation of the generator (21).Gσ .β ) ξα .0 g σm − g σk g00.m) φ0k . according to the condition (20).n − gkn. TRANSFORMATION OF THE METRIC TENSOR Transformations of ﬁelds can be found by calculating the following PB δgµν = {G. HT = primary (0) and the generator G (ξσ ) is given by G (ξσ ) = ξσ Gσ + ξσ.0 φ0σ .k + 2g nσ g0n.t 00 A + −2 √ Imnpk π 00 −g g g 0m g − g 0σ g00. HT + σ αγ .ν − ξν. For the timetime component.k φ00 − δ0 φ0k .
0 + g σ0 g00.0 + ξσ g 0σ g00.0 g 0σ + g0k.k ) ξσ which is equivalent to (26) up to a numerical factor 2 {g00 . 2 2 2 2 same numerical factor that occurs in (27).k 2 g and ﬁnally. For the transformation of the spacespace components.δg00 = {G.m) .k 2 2 − ξ0.k + gnk. after a simple rearrangement of terms in (28).k + gkm.0 .0 ξσ + g σk (2g0k. so that 9 (29) This is equivalent to the transformations of the corresponding components of (25) up to the .0) .n + g 0α gαk.k + ξσ g nσ (−g0k. g00 } = − constraint φ00 δ ξσ Gσ + ξσ. (25).t 00 A(pq)0αβt 2 2 2 −g g g 0m 1 g + ξσ g 0σ g00.0 φ0σ (0) δπ 00 so that nonzero contributions come only from those terms proportional to the primary 1 1 δg00 = ξσ g00. g0k .k − ξk.0 − g00.n − gkn.0 − ξσ −2 √ Imnpk π mn 00 + Imkpq gαβ. 2 g (28) In the ﬁrst bracket of (28) we substitute momenta in terms of the metric (4) so that using (14) we obtain 1 g σm − ξσ 00 −g 0n gmk.0 g σk − g σk g00.k + g nσ 00 (gmn. we need the corresponding part of the generator depending on φ0s 1 1 1 g σp g σm 1 δg0k = − ξ0.n + g0n. (27) For the transformation of the spacetime components.k + 2g nσ g0n.k − ξk. (26) Putting µ = ν = 0 and partially separating the space and time indices in Eq. G} = δ(dif f ) g00 that can be incorporated into a parameter ξσ → 2ξσ . gnm .m + g 0α gαm. we have δ(dif f ) g00 = −2ξ0. we need to keep only terms in G with π pq dependence. we have 1 1 1 1 δg0k = − ξ0.
k nm 2 2 2 nm 1 g 0σ 1 − √ Iabnm 2π ab + Inmcd A(cd)0µνk gµν.n − gnm. VI. we have 1 1 1 δgnm = − ξn. + ξσ 00 g 2 −g After substitution of π ab from (4) and using (6).m − gnm.n + g0n.n + ξ0 g 00 (g0m. (30) can be combined into one covariant expression (25) or (1). gnm } = δ δπ nm The second term produces contributions proportional to the primary constraints and can be neglected on the surface of primary constraints. In ordinary gauge ﬁeld theories (e.m − ξm.m − ξm. (9). In χ0σ we need only momentumdependent terms appearing in (13).m − gnm. because the righthand side of (20) equals zero exactly.n + g0n. this also demonstrates that we obtained the consistent Hamiltonian formulation of GR because the transformation (1) derivable from it is generally covariant. Maxwell or YangMills) this does not happen.n + ξσ δa eqa ∆pk − ∆pq eka gpq. We obtain then 2 g σp ξσ χ0σ − ξσ √ Iabpk π ab 00 φ0k . This completes the proof of our original statement that the Hamiltonian formulation gives the same result for the gauge invariance of GR as the Lagrangian formulation [15]. The transformations of the components of gµν in (26).0 ) 2 which is again equivalent with (25) up to the same numerical factor of 2. equations of motion are 10 . the transformation (30) is valid on the surface of primary constraints. In ordinary ﬁeld theories..k . Such deviations from ordinary gauge theories actually can be expected. CONCLUSION We would like to point out some peculiarities of the Hamiltonian formulation of GR. First of all. as it should. and (14). −g g 1 1 1 σ δgnm = − ξn.p ) 2 2 2 (30) 1 + ξp g qp (gnq.n + gpn.δgnm = {G. (29). Another peculiarity of GR is the PB among primary and secondary constraints is not zero by (16).g.q ) + g 0p (g0m.m − gnm.0 ) + g 0p (gpm.m + gmq. Moreover.
then the spirit of GR is “alive”. the equivalence of the Lagrangian and Hamiltonian formulations implies that the ADM Lagrangian should give the same ﬁelddependent redeﬁnition of gauge parameters (32) if treated by the Lagrangian method of [15].µ Gνρ − ξ.exactly invariant under a gauge transformation. the “crucial test” of having oﬀshell closure of the constraint algebra [22] seems as unreasonably strong for GR. we derived the gauge generators from the full set of ﬁrstclass constraints.ρ − ξ.ν Gµρ . are invariant only onshell ρ ρ δ(dif f ) Gµν = −ξ ρ Gµν. Going back to the “spiritual” statement of Hawking. In this paper we demonstrated that the consistent Hamiltonian formulation of GR can be obtained by considering the metric tensor as the canonical variable and not performing any change of variables. if diﬀeomorphism cannot be obtained in the Lagrangian or Hamiltonian formulation. the ADM Hamiltonian formulation gives diﬀerent transformations which can be presented in the form of diﬀeomorphism only after ﬁelddependent and noncovariant redeﬁnition of parameters [10]: ⊥ εADM = −g 00 −1/2 0 k ξdif f . transform into a combination of the same equations [12]. because of this. i. Following the Dirac conjecture and applying the Castellani procedure.e. 1 Gµν = Rµν − 2 gµν R = 0. we can make the conjecture or possibly draw the conclusion that if diﬀeomorphism is derivable through the Lagrangian or Hamiltonian approach. as the mathematical essence of the spirit of GR is to retain general covariance and that diﬀeomorphism can be derived from its structure. On the contrary. whereas the Einstein equations of motion. From these 11 . In particular. εk ADM = ξdif f − g 0k 0 ξ . Vice versa. our Hamiltonian formulation of GR which does not resort to any change of variables or modiﬁcations gives the same diﬀeomorphism invariance as the Lagrangian approach [15]. it is impossible to expect oﬀshell closure for generators of a transformation which by itself produces only onshell invariance of the equations of motion (31). (31) In particular. and it is the expected result [23]. then such a formulation is not the equivalent to GR. The equivalence of results in both formalisms for asymptotically ﬂat spacetimes was discussed in [24]. g 00 dif f (32) Moreover. Moreover.
G. Wiley.A. p. 87 (1952) 452. Schild and R. Note added in proof After completion of this work it came to our attention that. Comp. Belfer Graduate School of Sciences. Dirac. [2] F. Cliﬀ for inspiration and support.). [3] P.A. Proc. [1] P. In the Hamiltonian formulation of GR. 1973. Ann. Dirac. Rev. 143 (1982) 357. Arnowitt. the covariance is not manifest but the ﬁnal result on gauge transformation is presented in manifestly covariant form as in the Hamiltonian formulation of ordinary ﬁeld theories.A. [5] R.M.E.). Gravitation: An Introduction to Current Research.G. New York. Kuchar. [6] K. Frolov. by a diﬀerent from [9] method which was used in [10]. Rupasov and A. Deser and C. Relativity. Dirac. Witten (Ed. A 246 (1958) 333. 227. Israel (Ed.W. in: L. Math. 1962.M. Skinner. The research was partially supported by the Huron University College Faculty of Arts and Social Science Research Grant Fund. V. Our result is another illustration of the resiliency of Einstein’s General Relativity and any approach.V. [4] P. D. 237. the “speciﬁc metricdependent diﬀeomorphism” (32) of the ADM formulation was also obtained and extensively discussed in [25]. Yeshiva University. A. Zvelindovsky for discussions and W.generators. Lectures on Quantum Mechanics. [7] L. Petrov and S. Castellani. The helpful correspondence with A. Pirani. Samanta is gratefully acknowledged.I.N.. 1964. Phys.M. J. we explicitly derived diﬀeomorphism invariance in the Hamiltonian formulation of GR for the ﬁrst time. Misner.A. Astrophysics and Cosmology. 12 . Reidel Publ. p. McKeon. Soc. VII. in: W. Roy. New York. Dordrecht.M. S. ACKNOWLEDGEMENTS We would like to thank A.C. without redeﬁnition of gauge parameters. Phys. cannot violate its symmetry and leads to the generally covariant transformation (1). if correctly used. Can. 2 (1950) 129.
J.M. arXiv: 0708. Gitman and I.G. J. 169. Berlin.C. Phys.J. 13 . vol. Cambridge University Press. Berlin. 20 (2003) 3279. Theor.1430 [grqc]. Kuzmin. Int. V. [21] G. D 6 (1997) 239. Phys. K. Nicolai. Classical Fields: General Relativity and Gauge Theory. General Relativity.3300 [hepth]. [12] L. Petrov.V. private communication. [10] P. Carmeli.D.D. 1979. Phys. Phys. Quantization of Fields with Constraints. [15] S. [11] P. Kiriushcheva and S.M. Class. J.P. Green. Quantum Grav. Rothe and K. C. Hawking and W. p. Phys. Lifshitz. [25] J. Phys.W.[8] M. The Classical Theory of Fields.. J. 94 (1984) 379. Oxford. D.N. Pons.C. Israel (Eds. Mod.). Rothe and K. Peeters and M. 38 (1999) 1941. Hawking. Grishchuk. [13] A. Theor. Teitelboim and J. Quantum Grav. A. D. Henneaux. Shepley. An Einstein Centenary Survey.D. [16] D. in: S. Petrov. Int.4358 [hepth]. N. Rothe.M. Zamaklar.R. 746. Banerjee. Lett. Phys. Springer. [20] K. 1982. Phys. Banerjee. Lett. Samanta. Nucl. Mod. arXiv: 0710. [19] K. Commun. Gitman and I. B 332 (1990) 169. Saha. [14] M. [24] A. [17] S. Class. D 55 (19970 658. John Wiley & Sons.J. B 479 (2000) 429. Sundermeyer.D. 1 (1972) 15.M. Landau and E. Petrov and A. H. Phys. Pergamon Press. R. 1982.D. A. L. J. [9] R. Rothe. F¨l¨p. J.N. Bergmann and A. Popova.N.N. B 462 (1999) 248. Int. Rev. [23] S. fourth ed. uo [22] H. arXiv: 0705. Springer. Pons. Tyutin. Komar. D 4 (1995) 451. Constrained Dynamics. 22 (2005) R193. 1975. Mod. Phys. Math. Salisbury. Samanta.W. Lecture Notes in Physics. Popova and A. Cambridge. 1990. Petrov. H.V. Int. Mukherjee and A. [18] L. Zanalli. Int. A 3 (1988) 2651. Tyutin. M.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue reading from where you left off, or restart the preview.