Photon Inner-Product and The Gauss Linking Number
Photon Inner-Product and The Gauss Linking Number
Photon Inner-Product and The Gauss Linking Number
gr-qc/9608017
Abstract
2
More precisely, the situation is the following. Given a loop and a weighting function f
on R3, we can dene a \canonical thickening" f of the loop. The self-linking number of a
thickened loop can be computed by adding the Gauss linking numbers of the loops involved
in the thickening with weights given by f . The measure that dictates the inner product
is just the exponential of this self-linking number of f . Now, as is well-known, the Fock
inner product depends on the Minkowski metric. It is quite interesting that one can put all
the information about the space-time metric in the construction that associates quantum
states with functionals of loops and then express the inner-product itself in terms of the
Gauss linking number which is a topological invariant. A further striking fact is that this
\coding" of the inner-product information in a topological invariant works only in the loop
representation based on self-dual (or anti-self-dual) connections.
The plan of the paper is as follows. In section II, we will collect some mathematical
preliminaries. These are used in section III to construct the loop representation based on
self-dual connections. The main result then follows in section IV. Since our primary goal
is to bring out the interplay between self-duality of connections, loop representations and
the linking number, we will keep the functional analytic details to a minimum. However,
it should be rather straightforward to see how one can complete our discussion to obtain a
rigorous treatment. Throughout the article, we use units where c = 1, but write h and e
explicitly.
This section is divided in to two parts. In the rst, we recall the phase space formulation
of the Maxwell eld using self-dual variables and in the second we introduce the notion of
\form factors" associated with loops and thickened loops.
3
A. Self-Dual Variables for the Maxwell Field
Let us begin with a brief summary of the standard phase space formulation of Maxwell
elds. Denote by a space-like three-plane in Minkowski space, and by qab, the induced
positive denite (
at) metric thereon. The conguration variable for the Maxwell eld is
generally taken to be the connection one-form Aa(x) (the vector potential for the magnetic
eld) on . Its canonically conjugate momentum is the electric eld E a(x) on . (E a(x)
naturally arises as a vector density. However, since we have an underlying metric, qab, which
can be used to add or remove density weights, we will ignore density weights in this paper.)
The fundamental Poisson bracket is:
The system has one rst class constraint, @aE a(x) = 0. One can therefore pass to the reduced
phase space by xing transverse gauge. The true degrees of freedom are then contained in
the pair (ATa (x); ETa (x)) of transverse (i.e. divergence-free) vector elds on . Denote by ?
the phase space spanned by these elds. On ?, the only non-vanishing fundamental Poisson
bracket is:
where is the Laplacian operator compatible with the
at metric qab. It is convenient to
write ATa (x) and ETa (x) in terms of their Fourier decomposition. Then, the true degrees of
freedom are contained in the new dynamical variables qj (k), pj (k) with j = 1; 2:
where ma and m a are transverse (complex) vectors satisfying: maka = 0, and mam a = 1.
The Poisson brackets (2.2) for the transverse components are,
In order to construct the self dual connection, we will use dTa (x), the transverse vector
potential of the electric eld (ETa (x) = abc@bdTc (x))
We want to use the pair ("ATa (x), ETa (x)) as the basic variables. In terms of the (qj ; pj )
coordinates, the self dual connection takes the form,
"AT (x) = ? 1
Z
d3k eikx[(?p (k) + ijkjq (k))m (k) + (p (k) + ijkjq (k))m (k)]
a a
a (2)3=2 jkj 1 1 2 2
1
= ? (2)3=2
Z
d3k eikx[z (k)m (k) + z (k)m (k)] (2.9)
a a
jkj 1 2
where
The basic Poisson brackets for the pairs (zi(k); pj (k)), { the Fourier components of the
self-dual connection and the real electric eld{ are:
Finally, for later convenience, let us examine the self-dual magnetic eld B a := abc@b"Ac.
Its Fourier components of the magnetic eld are given by
These conditions will play an important role in the selection of the inner-product in the loop
representation.
B. Loops
Let us begin with some denitions. By a loop we shall mean a continuous and piecewise
smooth mapping
from S 1 to , where s 2 [0; 2]. Two loops
and will be said
to be holonomically equivalent if, for every smooth connection Aa, we have
Aadsa =
H
Aads . It turns out that two holonomically equivalent loops,
and , can dier from
H
a
each other only through: i) reparametrization,
(s) = (s0) for some (orientation-preserving)
reparametrization s ! s0 of the curve (s); ii) retracing identity,
= l l?1, where l
is a line segment and indicates composition of segments [5]. Each equivalence class will
be referred to as a holonomic loop. Since loops will primarily enter our discussions through
holonomies, it is these equivalence classes {rather than individual loops{ that will be directly
relevant to our discussion. To keep the notation simple, we will use the same symbols {say
{ to denote both an individual loop and the holonomic loop it denes; the context should
suce to resolve the resulting ambiguity.
An analytic characterization of holonomic loops can be given through certain distribu-
tional vector densities, called form factors. Given a loop , its form factor, F a(; x), is
dened via:
Z I
6
where (s) is a point on the loop at parameter value s and _ a(s) the tangent vector to
at (s). Note that the form factor F a(; x) is automatically divergence free,
@aF a(; x) = 0 ; (2.17)
because @af dsa = 0. It is often convenient to perform a Fourier transform to obtain the
H
(2.18)
Let us note a few properties of these form factors. First, two loops and will have
the same form factors if and only if they are holonomic. Thus, F a(; x) can be used to
characterize holonomic loop . Next, since F a(; x) is divergence-free its Fourier transform
is transverse (kaF a(; k) = 0). We can write the two independent components as:
F1(; k) F +(; k) = (21)3=2 ds _ a(s)m a(k)e?ik(s)
I
1 I
Fj (# ); k) = Fj (; k) + Fj (; k) (2.22)
In order to construct the quantum theory in the loop representation, we will need to
thicken the loops appropriately. We will conclude this section by indicating how this can
be done. Fix an averaging function fr (x) such that R3 d3xfr (x) = 1, and goes to a delta
R
To make the discussion concrete, we will make this choice and thus characterize the thick-
ening completely by a real parameter r. (However, the generalization to arbitrary smearing
functions is obvious.) Now, given a loop we take the loop + y obtained by rigidly shifting
the loop by the vector ya,
Next, we can average over y using the weight fr (y) and dene a \smeared form factor" via:
Z
Fra(; k) = exp ? r 2k F a(; k)
!
2 2
(2.26)
We will see that these Fra(; k) can be used as \generalized coordinates" for loops. More
precisely, once the weight functions fr (y) are chosen, we can associate with any loop a
transverse, smooth vector eld,
This section is divided in to three parts. In the rst, we recall a general quantization
program (for details, see [6], [7]), in the second we construct a ?-algebra of operators based
on loop variables and in the third we construct the loop representation.
A. Quantization Program
Consider a classical system with phase space ?. To construct the quantum theory, we
can proceed in the following steps.
i) Choose a subspace S of the space of complex valued functions on ? which is closed
under the Poisson bracket operation and large enough so that any well behaved function on
? can be expressed as (possible the limit of) a sum of products of elements of S . Elements of
S are called elementary classical variables and are to have unambiguous quantum analogs.
ii) Associate with each f in S an elementary quantum operator f^ and consider the
free associative algebra generated by these abstract operators. Impose on this algebra the
(generalized) canonical commutation relations
for all f and g in S . In addition, if the set S is over-complete, impose on the algebra also
`anti-commutation relations', namely the relations that capture the algebraic relations that
exist between elements of S . For instance if f , g and h = fg are all in S , then f^g^+^g f^ = 2^h.
Denote the resulting associative algebra by A.
9
iii) Introduce an involution, , on A by setting
(f^) = f^ (3.2)
for all elementary variables f (the bar denotes complex conjugation as before) and requiring
that satises the dening properties of an involution: (A^+B^ ) = A^+B^ ; (A^B^ ) = B^ A^
and (A^) = A^, for all A^, B^ in A and complex numbers . Denote the resulting -algebra
by A.
iv) Choose a linear representation of A on a complex vector space V . (The -relations
are ignored at this step).
v) Introduce on V an inner product h ; i so that the \quantum reality conditions" are
satised
h ; A^i = hA^ ; i (3.3)
for all , in V and A^ in A. Thus, it is the -relations that are to select the inner product.
The program requires two external inputs: the choice of S in step (i) and the choice of
the carrier space V of the representation in step (iv). If the choices are viable, i.e. if the
program can be completed at all, the resulting inner product is unique on each irreducible
sector of the representation of A on V [8]. In the framework of this program, the textbook
treatments of eld theories correspond to choosing for elements of A the smeared eld
operators, and, for V , the Fock space or, alternatively, suitable functionals of elds. In the
loop quantization, on the other hand, one changes this strategy. both S and V are now
constructed from holonomic loops.
Let us now implement this program for the Maxwell eld using loop variables. Let us
dene the smeared holonomy of self-dual connections as:
Z I
hr [] := exp 1
e d3 x ds _ a(s) "Aa(x)fr (x ? )
Z
= exp 1 3 a "
d x Fr (; x) Aa(x) (3.4)
e
10
or equivalently,
d3k (z (k)F (k) + z (k)F (k)) exp ? r2k2
" !#
Z
hr [] = exp ? 1e jkj 1 1 2 2
2
: (3.5)
Being a function of the self dual connection it can be regarded as a \conguration variable".
As a momentum variable we will take the (real) electric eld E a(x), or its Fourier transform
E a(k). (Strictly speaking we should take the smeared observable E [f ] = E afad3x, but
R
this smearing will not be relevant for our results.) Hence, hr [] and E a(k) provide us with
a (over-) complete coordinatization of the phase space. The space S of elementary classical
variables required in the rst step of the quantization program shall be the vector space
generated by the hr [] and E a(k). It is closed under Poisson-bracket operation because
The next step in the quantization program is the construction of the algebra A of quan-
tum operators. Let us associate with each hr [] in S an operator h^ r [] and with each E a(k)
an operator E^ a(k) and consider the associative algebra generated by nite sums of products
of these elementary quantum operators. On this algebra impose the commutation relations:
Furthermore, we must incorporate in this quantum algebra the fact that hr [] is over-
complete. i.e. there are algebraic relations among them; hr []hr [ ] = hr [# ]. This is
achieved by imposing on the algebra the relations h^ r []^hr [ ] = ^hr [# ] for all holonomic
loops and . The result is the algebra A of quantum operators.
C. Loop representation
The next step in the program is to choose a vector space V and a representation of the
quantum operators. The procedure involved is generally exploratory. Thus, one does not
11
specify all the required regularity conditions right in the beginning; the precise denition of
spaces considered becomes clear only at the end of the construction. This will also be the
case in our construction.
We wish to choose for V a vector space of suitable functionals of loops. As noted at the
end of section II B, in the standard electric eld representation, one can choose states as
suitably regular functionals [F ] of smooth, vector elds F a(k) which are transverse, i.e.,
satisfy F a(k)ka = 0. Now, in section II B, (for each choice of a smearing function fr ) we set
up a mapping 7! Fra(; k) from loops to smooth transverse vector elds in the momentum
space. We can just pull back the functionals (F ) via this map to obtain certain functionals
() on the loop space:
(Using the regularity conditions on that come from the standard electric-eld representa-
tion, it is not dicult to check that the map has no kernel, i.e., () = 0 for all if and
only if [F ] = 0.) Since the transverse vector elds F a(k) have only two components F (k),
from now on we will regard as functionals of the two elds F .
Thus, for the representation space V , we will use the functionals on the loop space of
the form (3.8). Using the procedure that was successful in the loop representation adapted
to the positive-frequency connections [3], the action of the basic operators ^hr [] and E^ a(k)
will be taken to be:
^hr [] (
) = (
)
E^ a(k) (
) = he F a(
; k) (
) (3.9)
As is usual in the loop representation, the electric eld is diagonal in the representation.
The only non-vanishing commutator between the basic operators is
Finally, for later convenience, we note the action of the magnetic eld operators B^ on these
12
states.
which is nothing but the \loop derivative" evaluated at Fr (see, e.g. [9]).
Our next task is to nd an inner-product on V so that the \quantum reality conditions"
(3.3) are satised. Let us begin with an inner product of the form
Z
Using the form of the operators (3.9) and (3.11) for p^ (k) and B^ (k), we conclude that the
reality condition (2.14) is satised if and only if
T = 2h F (k) : (3.14)
F (k) e2jkj
The solution to this equation is:
h d3k jF +(k)j2 ? jF ?(k)j2
T [F ] = ? 2
Z h i
(3.15)
e2 jkj r r
Hence, the explicit form of the inner product (3.12) is given by:
h i
1Although F (k) are complex-valued, they arise as Fourier components of a real eld F a (x) and
13
dierences. First, our states are functionals of loops rather than of a conguration eld
variable (such as the connection or the electric eld). Second, for the positive component,
the Gaussian is exponentially growing rather than damping. Hence, while we can take the
states to be polynomials in F ? as in the Schrodinger representation, we have to assume that
they are exponentially damped in their dependence on F +. Thus, for example, we can take
elements of V to be the functionals () on the loop space of the form:
? eh2 djkkj jFr+(k)j2
" #
3 Z
() = P [Fr(; k)] exp (3.17)
where P [Fr(; k)] is a polynomial in F . As usual, the Cauchy completion will enlarge
this space; the Hilbert space of all states will contain more general functionals. In this
description, F + captures positive helicity while F ? captures the negative helicity of the
photon. Thus, as one might have expected from our use of only the self-dual part of the
connection, the description is asymmetric in the two helicities.
To summarize, the elementary operators are ^hr () and E^ a. The space of quantum states
is given by functionals () of holonomic loops which are normalizable with respect to the
inner-product (3.16) and the action of the elementary operators is given by (3.9). For every
r > 0, this loop representation is naturally isomorphic to the Fock representation2 (where
the isomorphism, however, depends on the value of r.) The fact that we are using a loop
representation adapted to self-dual connections is re
ected in the measure that dictates the
inner product (3.16). In the loop representation adapted to positive-frequency elds [3], for
example, the measure has the same form but the squares of both jF j appear with negative
hence satisfy the reality conditions F (k) = ?F (?k). The conguration space underlying our loop
representation is thus real and states (F ) are arbitrary complex-valued functions of F (i.e.,
not subject o any \holomorphicity" condition.)
2If we let r go to zero, the smearing function fr (x) tends to the -distribution and the thickend
loop r reduces to the loop . However, now the exponent T in the measure diverges and the loop
representation ceases to exist.
14
signs in the exponent.
Recall that our quantum states are functionals of thickened loops r , or equivalently, of
their form factors Fr(; k); it is for technical convenience that in the intermediate stages
of calculations that we extended them to functionals on the vector space of all elds F (k).
Therefore, it is instructive to examine the measure that dictates the inner-product also
directly in terms of the thickened loops. This is easy to achieve: we can just pull-back the
\Gaussian" exp ?T that dictates the inner-product to the space of thickened loops. The
result is trivially given by:
exp (?T [Fr(; k)]) = exp 2h2 d k jFr+(; k)j2 ? jFr?(; k)j2
" #
Z 3
(4.1)
e jkj
We will now show that this loop functional can be expressed in terms of the Gauss linking
number.
Let us begin by recalling the denition of the linking number. Given non-intersecting
loops and the Gauss linking number GL(; ) between them can be expressed in terms
of their form factors as:
Z
where F a(; x) is the form factor for and wa(; x) is a potential for the form factor
of : abc@bwc(; x) = F a(; x). The integral is independent of the specic choice of the
potential !a (; x) because F a(; x) is divergence free. Note that neither the denition of
the form factor F a(; x) nor that of the potential !a(; x) requires any background elds on
the underlying oriented 3-manifold R3; in particular, there is no reference to the 3-metric.
(Since F a is a vector density, the abc in the denition of !a(; x) is the Levi-Civita density
which is naturally available on any oriented 3-manifold.) This is to be expected since the
Gauss linking number is a topological invariant.
15
Nonetheless, one can use the
at metric qab on R3 to express the linking number in more
familiar terms. First, we have the well-known form used by Gauss himself [10]:
GL(; ) := 41 ds dt abc_ a(s) _ b(t) j((ss))??(t()tj)3
Z Z c c
(4.3)
For our purposes, a more convenient form is the one involving the Fourier transforms of the
form factors. The Fourier transform of the potential has the form:
F a(; k) = iwc(; k)kbabc
= wc(; k)jkj(mam c ? m amc)
= jkj(maw+ (; k) ? m aw? (; k)) (4.4)
whence,
F +(; k) = jkjw+(; k); and F ?(; k) = ?jkjw?(; k): (4.5)
Therefore, the Gauss linking number takes the form
Z
(4.7)
jkj r r r r
16
This is the result we were seeking. (Note, incidentally, that the coecient of the linking
number is 2 over the ne structure constant.)
We conclude with a remark. Had we used positive frequency connections [3], for example,
the loop functional (4.1) would have been replaced by
Z
d3k (F +(; k)F +(; k) + F ?(; k)F ?(; k))
jkj r r r r
which has no obvious interpretation in terms of the topology of loops. Similarly, if we had
worked in the self-dual connection representation, the measure would have been dictated by
a \Gaussian" on the space of connections (see chapter 11.5, especially Eq 42b in [4] and [11])
and would therefore also have had no relation to topological invariants of loops. We need
both self-duality of the connection and the loop representation to relate the photon inner
product with the Gauss linking number.
Acknowledgments
This work was supported in part by the NSF grant 95-14240, by the Eberly research
fund of Penn State University and by DGAPA of UNAM.
17
REFERENCES
[1] Ashtekar A and Rovelli C 1992 Class. Quantum Grav. 9 S3
[2] Ashtekar A and Corichi A 1996 \Gauss linking number and the uncertainty principle"
CGPG Preprint
[3] Ashtekar A and Rovelli C 1992 Class. Quantum Grav. 9 1121
[4] Ashtekar A 1991 Lectures on Non-perturbative Canonical Gravity (Notes prepared in
collaboration with R S Tate) (Singapore: World Scientic)
[5] Ashtekar A and Lewandowski J 1994 Representation theory of analytic holonomy C
algebras Knots and Quantum Gravity ed J Baez (Oxford: Oxford University Press)
[6] Ashtekar A and Tate R S 1994 J. Math. Phys. 35 6434
[7] Ashtekar A, Lewandowski J, Marolf D, Mour~ao J and Thiemann T 1995 J. Math. Phys.
36 6456
[8] Rendall A 1993 Class Quantum Grav. 10 2261
[9] Gambini R and Pullin J 1996 Loops, Knots, Gauge Theory and Quantum Gravity (Cam-
bridge: Cambridge University Press)
[10] Gauss C F 1877 Werke vol V (Gotingen: Konigliche Gesellschaft der Wissenschaften)
605
[11] Ashtekar A, Rovelli C and Smolin L 1992 J. Geom. Phys. 8 7
18