Undergraduate Lecture Notes in De RhamHodge Theory

arXiv:0807.4991v4 [math.DG] 15 May 2011

Vladimir G. Ivancevic Tijana T. Ivancevic

Abstract
These lecture notes in the De RhamHodge theory are designed for a 1semester
undergraduate course (in mathematics, physics, engineering, chemistry or biology).
This landmark theory of the 20th Century mathematics gives a rigorous foundation
to modern field and gauge theories in physics, engineering and physiology. The only
necessary background for comprehensive reading of these notes is Greens theorem
from multivariable calculus.

Contents
1 Exterior Geometrical Mmachinery 2
1.1 From Greens to Stokes theorem . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Exterior derivative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Exterior forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Stokes theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2 De RhamHodge Theory Basics 7

2.1 Exact and closed forms and chains . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 De Rham duality of forms and chains . . . . . . . . . . . . . . . . . . . . . 9
2.3 De Rham cochain and chain complexes . . . . . . . . . . . . . . . . . . . . . 9
2.4 De Rham cohomology vs. chain homology . . . . . . . . . . . . . . . . . . . 10
2.5 Hodge star operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.6 Hodge inner product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.7 Hodge codifferential operator . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.8 Hodge Laplacian operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.9 Hodge adjoints and selfadjoints . . . . . . . . . . . . . . . . . . . . . . . . 16
2.10 Hodge decomposition theorem . . . . . . . . . . . . . . . . . . . . . . . . . . 17

Land Operations Division, Defence Science & Technology Organisation, P.O. Box 1500, Edinburgh SA
5111, Australia (Vladimir.Ivancevic@dsto.defence.gov.au)

Tesla Science Evolution Institute, Adelaide, Australia (Tijana.Ivancevic@alumni.adelaide.edu.au)

1
3 Hodge Decomposition and Gauge Path Integral 18
3.1 Feynman Path Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.1.1 Functional measure on the space of differential forms . . . . . . . . . 23
3.1.2 Abelian ChernSimons theory . . . . . . . . . . . . . . . . . . . . . . 23
3.2 Appendix: Path Integral in Quantum Mechanics . . . . . . . . . . . . . . . 25

1 Exterior Geometrical Mmachinery

To grasp the essence of HodgeDe Rham theory, we need first to familiarize ourselves with
exterior differential forms and Stokes theorem.

1.1 From Greens to Stokes theorem

Recall that Greens theorem in the region C in (x, y)plane 2
H RR R connects a line integral
C (over the boundary C of C) with a double integral C over C (see e.g., [1])
ZZ  
Q P
I
P dx + Qdy = dxdy.
C C x y
H RR
In other words, if we define two differential forms (integrands of C and C ) as

1form : A = P dx + Qdy, and

 
Q P
2form : dA = dxdy,
x y

(where d denotes the exterior derivative that makes a (p + 1)form out of a pform, see
next subsection), then we can rewrite Greens theorem as Stokes theorem:
Z Z
A= dA.
C C

The integration domain C is in topology called a chain, and C is a 1D boundary of a 2D

chain C. In general, the boundary of a boundary is zero (see [11, 12]), that is, (C) = 0,
or formally 2 = 0.

1.2 Exterior derivative

The exterior derivative d is a generalization of ordinary vector differential operators (grad,
div and curl see [9, 10]) that transforms pforms into (p + 1)forms d (see next
subsection), with the main property: dd = d2 = 0, so that in R3 we have (see Figures 1
and 2)

2
Figure 1: Basis vectors and one-forms in Euclidean R3 space: (a) Translational case; and
(b) Rotational case [2]. For the same geometry in R3 , see [11].

3
 any scalar function f = f (x, y, z) is a 0form;

the gradient df = of any smooth function f is a 1form

f f f
= df = dx + dy + dz;
x y z

the curl = d of any smooth 1form is a 2form

     
R Q P R Q P
= d = dydz + dzdx + dxdy;
y z z x x y
if = df = ddf = 0.

the divergence = d of any smooth 2form is a 3form

 
A B C
= d = + + dxdydz; if = d = dd = 0.
x y z

In general, for any two smooth functions f = f (x, y, z) and g = g(x, y, z), the exterior
derivative d obeys the Leibniz rule [3, 4]:

d(f g) = g df + f dg,

and the chain rule:

d (g(f )) = g (f ) df.

1.3 Exterior forms

In general, given a socalled 4D coframe, that is a set of coordinate differentials {dxi } R4 ,
we can define the space of all pforms, denoted p (R4 ), using the exterior derivative
d : p (R4P
) p+1 (R4 ) and Einsteins summation convention over repeated indices (e.g.,
Ai dx = 3i=0 Ai dxi ), we have:
i

1form a generalization of the Greens 1form P dx + Qdy,

A = Ai dxi 1 (R4 ).

For example, in 4D electrodynamics, A represents electromagnetic (co)vector poten-

tial.

4
Figure 2: Fundamental twoform and its flux in R3 : (a) Translational case;
R R (b) Rotational
case. In both cases the flux through the plane u v is defined as i
uv c dpi dq and
measured by the number of tubes crossed by the circulation oriented by u v [2]. For the
same geometry in R3 , see [11].

5
2form generalizing the Greens 2form (x Q y P ) dxdy (with j = /xj ),

B = dA 2 (R4 ), with components

1
B = Bij dxi dxj , or
2
B = j Ai dxj dxi , so that
Bij = 2j Ai = i Aj j Ai = Bji .

where is the anticommutative exterior (or, wedge) product of two differential

forms; given a pform p (R4 ) and a qform q (R4 ), their exterior product
is a (p + q)form p+q (R4 ); e.g., if we have two 1forms a = ai dxi , and
b = bi dxi , their wedge product a b is a 2form given by

= a b = ai bj dxi dxj = ai bj dxj dxi = b a.

The exterior product is related to the exterior derivative d = i dxi , by

d( ) = d + (1)p d.

3form

C = dB ( = ddA 0) 3 (R4 ), with components

1
C = Cijk dxi dxj dxk , or
3!
C = k B[ij] dxk dxi dxj , so that
Cijk = 6k B[ij], where B[ij] is the skewsymmetric part of Bij .

For example, in the 4D electrodynamics, B represents the field 2form Faraday, or

the LienardWiechert 2form (in the next section we will use the standard symbol
F instead of B) satisfying the sourceless magnetic Maxwells equation,

Bianchi identity : dB = 0, in components k B[ij] = 0.

4form

D = dC ( = ddB 0) 4 (R4 ), with components

l i j k
D = l C[ijk] dx dx dx dx , or
1
D = Dijkl dxi dxj dxk dxl , so that
4!
Dijkl = 24l C[ijk].

6
1.4 Stokes theorem
Generalization of the Greens theorem in the plane (and all other integral theorems from
vector calculus) is the Stokes theorem for the pform , in an oriented nD domain C
(which is a pchain with a (p 1)boundary C, see next section)
Z Z
= d.
C C

For example, in the 4D Euclidean space R4 we have the following three particular cases
of the Stokes theorem, related to the subspaces C of R4 :
The 2D Stokes theorem: Z Z
A= B.
C 2 C2
The 3D Stokes theorem: Z Z
B= C.
C 3 C3
The 4D Stokes theorem: Z Z
C= D.
C 4 C4

2 De RhamHodge Theory Basics

Now that we are familiar with differential forms and Stokes theorem, we can introduce
HodgeDe Rham theory.

2.1 Exact and closed forms and chains

Notation change: we drop boldface letters from now on. In general, a pform is called
closed if its exterior derivative d = i dxi is equal to zero,

d = 0.

From this condition one can see that the closed form (the kernel of the exterior derivative
operator d) is conserved quantity. Therefore, closed pforms possess certain invariant
properties, physically corresponding to the conservation laws (see e.g., [6]).
Also, a pform that is an exterior derivative of some (p 1)form ,

= d,

is called exact (the image of the exterior derivative operator d). By Poincare lemma, exact
forms prove to be closed automatically,

d = d(d) = 0.

7
 Since d2 = 0, every exact form is closed. The converse is only partially true, by
Poincare lemma: every closed form is locally exact.
Technically, this means that given a closed pform p (U ), defined on an open
set U of a smooth manifold M (see Figure 3), 1 any point m U has a neighborhood on
which there exists a (p 1)form p1 (U ) such that d = |U . In particular, there is

Figure 3: Depicting the manifold concept (see footnote 1).

a Poincare lemma for contractible manifolds: Any closed form on a smoothly contractible
1
Smooth manifold is a curved nD space which is locally equivalent to Rn . To sketch it formal definition,
consider a set M (see Figure 3) which is a candidate for a manifold. Any point x M has its Euclidean
chart, given by a 11 and onto map i : M Rn , with its Euclidean image Vi = i (Ui ). Formally, a chart
i is defined by
i : M Ui x 7 i (x) Vi Rn ,
where Ui M and Vi Rn are open sets.
Any point x M can have several different charts (see Figure 3). Consider a case of two charts,
i , j : M Rn , having in their images two open sets, Vij = i (Ui Uj ) and Vji = j (Ui Uj ). Then we
have transition functions ij between them,

ij = j 1
i : Vij Vji , locally given by ij (x) = j (1
i (x)).

If transition functions ij exist, then we say that two charts, i and j are compatible. Transition
functions represent a general (nonlinear) transformations of coordinates, which are the core of classical
tensor calculus.
A set of compatible charts i : M Rn , such that each point x M has its Euclidean image in at least
one chart, is called an atlas. Two atlases are equivalent iff all their charts are compatible (i.e., transition
functions exist between them), so their union is also an atlas. A manifold structure is a class of equivalent
atlases.
Finally, as charts i : M Rn were supposed to be 1-1 and onto maps, they can be either homeomor-
phisms, in which case we have a topological (C 0 ) manifold, or diffeomorphisms, in which case we have a
smooth (C k ) manifold.

8
manifold is exact.
The Poincare lemma is a generalization and unification of two wellknown facts in
vector calculus:
(i) If curl F = 0, then locally F = grad f ; and (ii) If div F = 0, then locally F = curl G.
A cycle is a pchain, (or, an oriented pdomain) C Cp (M ) such that C = 0. A
boundary is a chain C such that C = B, for any other chain B Cp (M ). Similarly, a
cocycle (i.e., a closed form) is a cochain such that d = 0. A coboundary (i.e., an exact
form) is a cochain such that = d, for any other cochain . All exact forms are closed
( = d d = 0) and all boundaries are cycles (C = B C = 0). Converse is true
only for smooth contractible manifolds, by Poincare lemma.

2.2 De Rham duality of forms and chains

Integration on a smooth manifold M should be thought of as a nondegenerate bilinear
pairing h, i between pforms and pchains (spanning a finite domain on M ). Duality of
pforms and pchains on M is based on the De Rhams period, defined as [9, 7]
Z
Period := := hC, i ,
C

where C is a cycle, is a cocycle, while hC, i = (C) is their inner product hC, i :
p (M )Cp (M ) R. From the Poincare lemma, a closed pform is exact iff hC, i = 0.
The fundamental topological duality is based on the Stokes theorem,
Z Z
= d or hC, i = hC, di ,
C C

where C is the boundary of the pchain C oriented coherently with C on M . While the
boundary operator is a global operator, the coboundary operator d is local, and thus
more suitable for applications. The main property of the exterior differential,

d d d2 = 0 = 2 = 0, (and converse),

can be easily proved using the Stokes theorem (and the above period notation) as

0 = 2 C, = hC, di = C, d2 = 0.




2.3 De Rham cochain and chain complexes

In the Euclidean 3D space R3 we have the following De Rham cochain complex
d d d
0 0 (R3 ) 1 (R3 ) 2 (R3 ) 3 (R3 ) 0.
grad curl div

9
Using the closure property for the exterior differential in R3 , d d d2 = 0, we get the
standard identities from vector calculus

curl grad = 0 and div curl = 0.

As a duality, in R3 we have the following chain complex

0 C0 (R3 )C1 (R3 )C2 (R3 )C3 (R3 ) 0,

(with the closure property 2 = 0) which implies the following three boundaries:

C1 7 C0 = (C1 ), C2 7 C1 = (C2 ), C3 7 C2 = (C3 ),

where C0 C0 is a 0boundary (or, a point), C1 C1 is a 1boundary (or, a line),

C2 C2 is a 2boundary (or, a surface), and C3 C3 is a 3boundary (or, a hypersurface).
Similarly, the De Rham complex implies the following three coboundaries:
d d d
C 0 7 C 1 = d(C 0 ), C 1 7 C 2 = d(C 1 ), C 2 7 C 3 = d(C 2 ),

where C 0 0 is 0form (or, a function), C 1 1 is a 1form, C 2 2 is a 2form, and

C 3 3 is a 3form.
In general, on a smooth nD manifold M we have the following De Rham cochain
complex [9]
d d d d d
0 0 (M ) 1 (M ) 2 (M ) 3 (M ) n (M ) 0,

satisfying the closure property on M, d d d2 = 0.

2.4 De Rham cohomology vs. chain homology

Briefly, the De Rham cohomology is the (functional) space of closed differential pforms
modulo exact ones on a smooth manifold.
More precisely, the subspace of all closed pforms (cocycles) on a smooth manifold
M is the kernel Ker(d) of the De Rham dhomomorphism (see Figure 4), denoted by
Z p (M ) p (M ), and the sub-subspace of all exact pforms (coboundaries) on M is the
image Im(d) of the De Rham homomorphism denoted by B p (M ) Z p (M ). The quotient
space
Ker d : p (M ) p+1 (M )


p Z p (M )
HDR (M ) := = , (1)
B pM Im (d : p1 (M ) p (M ))
is called the pth De Rham cohomology group of a manifold M . It is a topological invariant
of a manifold. Two pcocycles , p (M ) are cohomologous, or belong to the same

10
Figure 4: A small portion of the De Rham cochain complex, showing a homomorphism of
cohomology groups.

cohomology class [] H p (M ), if they differ by a (p 1)coboundary = d

p
p1 (M ). The dimension bp = dim H p (M ) of the De Rham cohomology group HDR (M )
of the manifold M is called the Betti number bp .
Similarly, the subspace of all pcycles on a smooth manifold M is the kernel Ker()
of the homomorphism, denoted by Zp (M ) Cp (M ), and the sub-subspace of all
pboundaries on M is the image Im() of the homomorphism, denoted by Bp (M )
Cp (M ). Two pcycles C1 ,C2 Cp are homologous, if they differ by a (p 1)boundary
C1 C2 = B Cp1 (M ). Then C1 and C2 belong to the same homology class [C]
Hp (M ), where Hp (M ) is the homology group of the manifold M , defined as

Zp (M ) Ker( : Cp (M ) Cp1 (M ))
Hp (M ) := = ,
Bp (M ) Im( : Cp+1 (M ) Cp (M ))

where Zp is the vector space of cycles and Bp Zp is the vector space of boundaries on
M . The dimension bp = dim Hp (M ) of the homology group Hp (M ) is, by the De Rham
theorem, the same Betti number bp .
If we know the Betti numbers for all (co)homology groups of the manifold M , we can
calculate the EulerPoincare characteristic of M as
n
X
(M ) = (1)p bp .
p=1

For example, consider a small portion of the De Rham cochain complex of Figure 4

11
spanning a space-time 4manifold M ,
dp1 dp
p1 (M ) p (M ) p+1 (M )

As we have seen above, cohomology classifies topological spaces by comparing two sub-
spaces of p : (i) the space of pcocycles, Z p (M ) = Ker dp , and (ii) the space of
pcoboundaries, B p (M ) = Im dp1 . Thus, for the cochain complex of any space-time
4manifold we have,
d2 = 0 B p (M ) Z p (M ),
that is, every pcoboundary is a pcocycle. Whether the converse of this statement is
true, according to Poincare lemma, depends on the particular topology of a space-time
4manifold. If every pcocycle is a pcoboundary, so that B p and Z p are equal, then the
cochain complex is exact at p (M ). In topologically interesting regions of a space-time
manifold M , exactness may fail [13], and we measure the failure of exactness by taking
the pth cohomology group
H p (M ) = Z p (M )/B p (M ).

2.5 Hodge star operator

The Hodge star operator : p (M ) np (M ), which maps any pform into its dual
(n p)form on a smooth nmanifold M , is defined as (see, e.g. [8])

= = h, i , for , p (M ),
= (1)p(np) ,
(c1 + c2 ) = c1 ( ) + c2 ( ),
= 0 0.

The operator depends on the Riemannian metric g = gij on M 2 and also on the orien-
tation (reversing orientation will change the sign)3 [4]. The volume form is defined in
local coordinates on an nmanifold M as (compare with Hodge inner product below)
q
= vol = (1) = det(gij ) dx1 ... dxn , (2)
2
In local coordinates on a smooth manifold M , the metric g = gij is defined for any orthonormal basis
(i = xi ) in M by
gij = g(i , j ) = ij , k gij = 0.

3
Hodge operator is defined locally in an orthonormal basis (coframe) of 1forms ei dxi on a smooth
manifold M as:
(ei ej ) = ek , ()2 = 1.

12
and the total volume on M is given by
Z
vol(M ) = (1).
M

For example, in Euclidean R3 space with Cartesian (x, y, z) coordinates, we have:

dx = dy dz, dy = dz dx, dz = dx dy.

The Hodge dual in this case clearly corresponds to the 3D crossproduct.

In the 4Delectrodynamics, the dual 2form Maxwell F satisfies the electric Maxwell
equation with the source [11],

Dual Bianchi identity : d F = J,

where J is the 3form dual to the chargecurrent 1form J.

2.6 Hodge inner product

For any two pforms , p (M ) with compact support on an nmanifold M , we define
bilinear and positivedefinite Hodge L2 inner product as
Z Z
(, ) = h, i (1) = , (3)
M M

where is an nform. We can extend the product (, ) to L2 (p (M )); it remains

bilinear and positivedefinite, because as usual in the definition of L2 , functions that differ
only on a set of measure zero are identified. The inner product (3) is evidently linear in
each variable and symmetric, (, ) = (, ). We have: (, ) 0 and (, ) = 0 iff
= 0. Also, ( , ) = (, ). Thus, operation (3) turns the space p (M ) into an
infinitedimensional innerproduct space.
From (3) it follows that for every pform p (M ) we can define the norm functional
Z Z
2
kk = (, ) = h, i (1) = ,
M M

for which the EulerLagrangian equation becomes the Laplace equation (see Hodge Lapla-
cian below),
= 0.
For example, the standard Lagrangian for the free Maxwell electromagnetic field,
F = dA (where A = Ai dxi is the electromagnetic potential 1form), is given by [3, 4, 5]
1
L(A) = (F F ),
2

13
with the corresponding action
1
Z
S(A) = F F.
2

Using the Hodge L2 inner product (3), we can rewrite this electrodynamic action as
1
S(A) = (F, F ). (4)
2

2.7 Hodge codifferential operator

The Hodge dual (or, formal adjoint) to the exterior derivative d : p (M ) p+1 (M ) on
a smooth manifold M is the codifferential , a linear map : p (M ) p1 (M ), which
is a generalization of the divergence, defined by [9, 8]

= (1)n(p+1)+1 d , so that d = (1)np .

That is, if the dimension n of the manifold M is even, then = d .

Applied to any pform p (M ), the codifferential gives

= (1)n(p+1)+1 d , d = (1)np+1 d d.

If = f is a 0form, or function, then f = 0. If a pform is a codifferential of a

(p + 1)form , that is = , then is called the coexact form. A pform is coclosed
if = 0; then is closed (i.e., d = 0) and conversely.
The Hodge codifferential satisfies the following set of rules:

= 2 = 0, the same as dd = d2 = 0;

= (1)p+1 d; = (1)p d;

d = d; d = d.

Standard example is classical electrodynamics, in which the gauge field is an electro-

magnetic potential 1form (a connection on a U (1)bundle),

A = A dx = A dx + df, (f = arbitrary scalar field),

with the corresponding electromagnetic field 2form (the curvature of the connection A)

F = dA, in components given by

1
F = F dx dx , with F = A A .
2

14
Electrodynamics is governed by the Maxwell equations, which in exterior formulation read4

dF = 0, F = 4J, or in components,

F[,] = 0, F , = 4J ,

where comma denotes the partial derivative and the 1form of electric current J = J dx
is conserved, by the electrical continuity equation,

J = 0, or in components, J , = 0.

2.8 Hodge Laplacian operator

The codifferential can be coupled with the exterior derivative d to construct the Hodge
Laplacian : p (M ) p (M ), a harmonic generalization of the LaplaceBeltrami dif-
ferential operator, given by5

= d + d = (d + )2 .
4
The first, sourceless Maxwell equation, dF = 0, gives vector magnetostatics and magnetodynamics,

Magnetic Gauss law : div B = 0,

Faradays law : t B + curl E = 0.

The second Maxwell equation with source, F = J (or, d F = J), gives vector electrostatics and
electrodynamics,

Electric Gauss law : div E = 4,

Amp`eres law : t E curl B = 4j.

5 2
Note that the difference d = D is called the Dirac operator. Its square D equals the Hodge
Laplacian .
Also, in his QFTbased rewriting the Morse topology, E. Witten [14] considered also the operators:

dt = etf detf , their adjoints : dt = etf detf ,

as well as their Laplacian: t = dt dt + dt dt .

For t = 0, 0 is the Hodge Laplacian, whereas for t , one has the following expansion
X 2h
t = dd + d d + t2 kdf k2 + t [i xk , dxj ],
k,j
xk xj

where (xk )k=1,...,n is an orthonormal frame at the point under consideration. This becomes very large
for t , except at the critical points of f , i.e., where df = 0. Therefore, the eigenvalues of t will
concentrate near the critical points of f for t , and we get an interpolation between De Rham
cohomology and Morse cohomology.

15
 satisfies the following set of rules:
= = d; d = d = dd; =.
A pform is called harmonic iff
= 0 (d = 0, = 0).
Thus, is harmonic in a compact domain D M 6 iff it is both closed and coclosed in D.
Informally, every harmonic form is both closed and coclosed. As a proof, we have:
0 = (, ) = (, d) + (, d) = (, ) + (d, d) .
Since (, ) 0 for any form , (, ) and (d, d) must vanish separately. Thus,
d = 0 and = 0.
p
All harmonic pforms on a smooth manifold M form the vector space H (M ).
Also, given a pform , there is another pform such that the equation
=
is satisfied iff for any harmonic pform we have (, ) = 0.
For example, to translate notions from standard 3D vector calculus, we first identify
scalar functions with 0forms, field intensity vectors with 1forms, flux vectors with 2
forms and scalar densities with 3forms. We then have the following correspondence:
grad d : on 0forms; curl d : on 1forms;
div : on 1forms; div grad : on 0forms;
curl curl grad div : on 1forms.

We remark here that exact and coexact pforms ( = d and = ) are mutually
orthogonal with respect to the L2 inner product (3). The orthogonal complement consists
of forms that are both closed and coclosed: that is, of harmonic forms ( = 0).

2.9 Hodge adjoints and selfadjoints

If is a pform and is a (p + 1)form then we have [9]
(d, ) = (, ) and (, ) = (, d). (5)
This relation is usually interpreted as saying that the two exterior differentials, d and ,
are adjoint (or, dual) to each other. This identity follows from the fact that for the volume
form given by (2) we have d = 0 and thus
Z
d( ) = 0.
M
6
A domain D is compact if every open cover of D has a finite subcover.

16
 Relation (5) also implies that the Hodge Laplacian is selfadjoint (or, selfdual),

(, ) = (, ),

which is obvious as either side is (d, d) + (, ). Since (, ) 0, with (, ) = 0

only when = 0, is a positivedefinite (elliptic) selfadjoint differential operator.

2.10 Hodge decomposition theorem

The celebrated Hodge decomposition theorem (HDT) states that, on a compact orientable
smooth nmanifold M (with n p), any exterior pform can be written as a unique
sum of an exact form, a coexact form, and a harmonic form. More precisely, for any form
p (M ) there are unique forms p1 (M ), p+1 (M ) and a harmonic form
p (M ), such that

any form exact coexact harmonic

HDT : = d + +

For the proof, see [8, 9].

In physics community, the exact form d is called longitudinal, while the coexact
form is called transversal, so that they are mutually orthogonal. Thus any form can
be orthogonally decomposed into a harmonic, a longitudinal and transversal form. For
example, in fluid dynamics, any vector-field v can be decomposed into the sum of two
vector-fields, one of which is divergencefree, and the other is curlfree.
Since is harmonic, d = 0. Also, by Poincare lemma, d(d) = 0. In case is a
closed pform, d = 0, then the term in HDT is absent, so we have the short Hodge
decomposition,
= d + , (6)
thus and differ by d. In topological terminology, and belong to the same
cohomology class [] H p (M ). Now, by the De Rham theorems it follows that if C is
any pcycle, then Z Z
= ,
C C
that is, and have the same periods. More precisely, if is any closed pform, then
there exists a unique harmonic pform with the same periods as those of (see [9, 10]).
The HodgeWeyl theorem [8, 9] states that every De Rham cohomology class has a
p
unique harmonic representative. In other words, the space H (M ) of harmonic pforms
on a smooth manifold M is isomorphic to the De Rham cohomology group (1), or

17
 p p
H (M )
= HDR (M ). That is, the harmonic part of HDT depends only on the global
structure, i.e., the topology of M .
For example, in (2 + 1)D electrodynamics, pform Maxwell equations in the Fourier
domain are written as [15]

dE = iB, dB = 0,
dH = iD + J, dD = Q,

where H is a 0form (magnetizing field), D (electric displacement field), J (electric current

density) and E (electric field) are 1forms, while B (magnetic field) and Q (electric charge
density) are 2forms. From d2 = 0 it follows that the J and the Q satisfy the continuity
equation
dJ = iwQ,

where i = 1 and w is the field frequency. Constitutive equations, which include all
metric information in this framework, are written in terms of Hodge star operators (that
fix an isomorphism between p forms and (2 p) forms in the (2 + 1) case)

D = E, B = H.

Applying HDT to the electric field intensity 1form E, we get [16]

E = d + A + ,

where is a 0form (a scalar field) and A is a 2form; d represents the static field and A
represents the dynamic field, and represents the harmonic field component. If domain
is contractible, is identically zero and we have the short Hodge decomposition,

E = d + A.

3 Hodge Decomposition and Gauge Path Integral

3.1 Feynman Path Integral
The driving engine of quantum field theory is the Feynman path integral. Very briefly,
there are three basic forms of the path integral (see, e.g., [5]):
1. Sumoverhistories, developed in Feynmans version of quantum mechanics (QM)7
[17];
7
Feynmans amplitude is a space-time version of the Schr odingers wave-function , which describes
how the (non-relativistic) quantum state of a physical system changes in space and time, i.e.,
hOuttf in |Intini i = (x, t), (for x [In, Out], t [tini , tf in ]).
In particular, quantum wave-function is a complexvalued function of real space variables x =

18
 2. Sumoverfields, started in Feynmans version of quantum electrodynamics (QED)
[18] and later improved by FadeevPopov [20];
3. Sumovergeometries/topologies in quantum gravity (QG), initiated by S. Hawking
and properly developed in the form of causal dynamical triangulations (see [21]; for a
softer review, see [22]).
In all three versions, Feynmans actionamplitude formalism includes two components:
1. A realvalued, classical, Hamiltons action functional,
Z tf in
S[] := L[] dt,
tini

with the Lagrangian energy function defined over the Lagrangian density L,
Z
L[] = dn x L(, ), ( /x ),

(x1 , x2 , ..., xn ) Rn , which means that its domain is in Rn and its range is in the complex plane, for-
mally (x) : Rn C. For example, the onedimensional stationary plane wave with wave number k is
defined as
(x) = eikx , (for x R),
where the real number k describes the wavelength, = 2/k. In n dimensions, this becomes

(x) = eipx ,

where the momentum vector p = k is the vector of the wave numbers k in natural units (in which ~ =
m = 1).
More generally, quantum wave-function is also time dependent, = (x, t). The timedependent plane
wave is defined by
2
(x, t) = eipxip t/2 . (7)
In general, (x, t) is governed by the Schr
odinger equation [19, 5] (in natural units ~ = m = 0)
1
i (x, t) = (x, t), (8)
t 2
where is the ndimensional Laplacian. The solution of (8) is given by the integral of the timedependent
plane wave (7), Z
1 2
(p)dn p,
(x, t) = n/2
eipxip t/2 0
(2) Rn
which means that (x, t) is the inverse Fourier transform of the function
2

(p, (p),
t) = eip t/2 0

(p) has to be calculated for each initial wave-function. For example, if initial wave-function is
where 0
Gaussian,
x2 1 p2
f (x) = exp(a ), with the Fourier transform f(p) = exp( ).
2 a 2a
2
then (p) =
1
p
exp( 2a ).
0 a

19
while is a common symbol denoting all three things to be summed upon (histories,
fields and geometries). The action functional S[] obeys the Hamiltons least action prin-
ciple, S[] = 0, and gives, using standard variational methods,8 the EulerLagrangian
equations, which define the shortest path, the extreme field, and the geometry of minimal
curvature (and without holes).
8
In Lagrangian field theory, the fundamental quantity is the action
Z tout Z
S[] = L dt = dn x L(, ) ,
tin R4

so that the least action principle, S[] = 0, gives

Z  
L L
0 = dn x + ( )
R4 ( )
Z     
L L L
= dn x + .
R4 ( ) ( )

The last term can be turned into a surface integral over the boundary of the R4 (4D space-time region
of integration). Since the initial and final field configurations are assumed given, = 0 at the temporal
beginning tin and end tout of this region, which implies that the surface term is zero. Factoring out the
from the first two terms, and since the integral must vanish for arbitrary , we arrive at the Euler-lagrange
equation of motion for a field,  
L L
= 0.
( )
If the Lagrangian (density) L contains more fields, there is one such equation for each. The momentum
density (x) of a field, conjugate to (x) is defined as: (x) = L
(x)
.
For example, the standard electromagnetic action
Z
1
S= d4 x F F , where F = A A ,
4 R4
gives the sourceless Maxwells equations:

F = 0, F = 0,

where the field strength tensor F and the Maxwell equations are invariant under the gauge transforma-
tions,
A A + .
The equations of motion of charged particles are given by the Lorentzforce equation,
du
m = eF u ,
d
where e is the charge of the particle and u ( ) its four-velocity as a function of the proper time.

20
 2. A complexvalued, quantum transition amplitude,9
Z
hOuttf in |Intini i := D[] eiS[] , (11)

9
The transition amplitude is closely related to partition function Z, which is a quantity that encodes
the statistical properties of a system in thermodynamic equilibrium. It is a function of temperature and
other parameters, such as the volume enclosing a gas. Other thermodynamic variables of the system,
such as the total energy, free energy, entropy, and pressure, can be expressed in terms of the partition
functionP or its derivatives. In particular, the partition function of a canonical ensemble is defined as a sum
Z() = j eEj , where = 1/(kB T ) is the inverse temperature, where T is an ordinary temperature
and kB is the Boltzmanns constant. However, as the position xi and momentum pi variables of an ith
particle in a system can vary continuously, the set of microstates is actually uncountable. In this case,
some form of coarsegraining procedure must be carried out, which essentially amounts to treating two
mechanical states as the same microstate if the differences in their position and momentum variables are
small enough. The partition function then takes the form of an integral. For instance, the partition
function of a gas consisting of N molecules is proportional to the 6N dimensional phasespace integral,
Z
Z() d3 pi d3 xi exp[H(pi , xi )],
R6N
i
where H = H(pi , x ), (i = 1, ..., N ) is the classical Hamiltonian (total energy) function.
Given a set of random variables Xi taking on values xi , and purely potential Hamiltonian function
H(xi ), the partition function is defined as
X h i
Z() = exp H(xi ) .
xi

The function H is understood to be a real-valued function on the space of states {X1 , X2 , } while is
a real-valued free parameter (conventionally, the inverse temperature). The sum over the xi is understood
to be a sum over all possible values that the random variable Xi may take. Thus, the sum is to be replaced
by an integral when the Xi are continuous, rather than discrete. Thus, one writes
Z h i
Z() = dxi exp H(xi ) ,

for the case of continuously-varying random variables Xi .

Now, the number of variables Xi need not be countable, in which case the set of coordinates {xi }
becomes a field = (x), so the sum is to be replaced by the Euclidean path integral (that is a Wick
rotated Feynman transition amplitude (12) in imaginary time), as
Z
Z() = D[] exp [H()] .

More generally, in quantum field theory, instead of the field Hamiltonian H() we have the action S()
of the theory. Both Euclidean path integral,
Z
Z() = D[] exp [S()] , real path integral in imaginary time, (9)

and Lorentzian one,

Z
Z() = D[] exp [iS()] , complex path integral in real time, (10)

are usually called partition functions. While the Lorentzian path integral (10) represents a quantum-field
theory-generalization of the Schr odinger equation, the Euclidean path integral (9) represents a statistical-
field-theory generalization of the FokkerPlanck equation.

21
where D[] is an appropriate Lebesguetype measure,
N
Y
D[] = lim is , (i = 1, ..., n),
N
s=1

so that we can safely integrate over a continuous spectrum and sum over a discrete
spectrum of our problem domain , of which the absolute square is the realvalued
probability density function,

P := |hOuttf in |Intini ii|2 .

This procedure can be redefined in a mathematically cleaner way if we Wickrotate

the time variable t to imaginary values, t 7 = t, thereby making all integrals real:
Z Z
iS[] W ick
D[] e D[] eS[] . (12)

For example, in non-relativistic quantum mechanics (see Appendix), the propagation

amplitude from xa to xb is given by the configuration path integral 10
Z
iHT
|xa = D[x(t)] eiS[x(t)] ,


U (xa , xb ; T ) = hxb |xa i = xb | e

which satisfies the Schr

odinger equation (in natural units)

2
i (xa , xb ; T ),
U (xa , xb ; T ) = HU where = 1 + V (xb ).
H
T 2 x2b
10
On the other hand, the phasespace path integral (without peculiar constants in the functional measure)
reads !  Z T
YZ  
U (qa , qb ; T ) = D[q(t)]D[p(t)] exp i pi qi H(q, p) dt ,
i 0

where the functions q(t) (space coordinates) are constrained at the endpoints, but the functions p(t)
(canonicallyconjugated momenta) are not. The functional measure is just the product of the standard
integral over phase space at each point in time
Y 1 Z
D[q(t)]D[p(t)] = dq i dpi .
i
2

Applied to a non-relativistic real scalar field (x, t), this path integral becomes
E Z  Z T 
D 1
b (x, t)| eiHT |a (x, t) = D[] exp i L() d4 x , with L() = ( )2 V ().
0 2

22
3.1.1 Functional measure on the space of differential forms
The Hodge inner product (3) leads to a natural (metricdependent) functional measure
D[] on p (M ), which normalizes the Gaussian functional integral
Z
D[] eih|i = 1. (13)

One can use the invariance of (13) to determine how the functional measure transforms
under the Hodge decomposition. Using HDT and its orthogonality with respect to the
inner product (3), it was shown in [23] that

h, i = h, i + hd, di + h, i = h, i + h, di + h, di , (14)

where the following differential/conferential identities were used [24]

hd, di = h, di and h, i = h, di.

Since, for any linear operator O, one has

Z
D[] exp ih|Oi = det1/2 (O),

(13) and (14) imply that

D[] = D[]D[]D[] det1/2 (d)det1/2 (d).

3.1.2 Abelian ChernSimons theory

Recall that the classical action for an Abelian ChernSimons theory,
Z
S= A dA ,
M

is invariant (up to a total divergence) under the gauge transformation:

A 7 A + d. (15)

We wish to compute the partition function for the theory

1
Z
Z := D[A] eiS[A] ,
VG

where VG denotes the volume of the group of gauge transformations in (15), which must
be factored out of the partition function in order to guarantee that the integration is

23
performed only over physically distinct gauge fields. We can handle this by using the
Hodge decomposition to parametrize the potential A in terms of its gauge invariant, and
gauge dependent parts, so that the volume of the group of gauge transformations can be
explicitly factored out, leaving a functional integral over gauge invariant modes only [23].
We now transform the integration variables:

A 7 , , ,

where , , parameterize respectively the exact, coexact, and harmonic parts of the
connection A. Using the Jacobian (14) as well as the following identity on 0forms = d,
we get [23]
1
Z
Z= D[]D[]D[] det1/2 () det1/2 (d) eiS ,
VG
from which it follows that Z
VG = D[], (16)

while the classical action functional becomes, after integrating by parts, using the harmonic
properties of and the nilpotency of the exterior derivative operators, and dropping surface
terms:
S = h, di .
Note that S depends only the coexact (transverse) part of A. Using (16) and integrating
over yields: Z
Z = D[]det1/2 (d) det1/2 () det1/2 (d) .

Also, it was proven in [23] that

3
det(d) = det1/2 ((dd)(d)) = det 2 (d).

As a consequence of Hodge duality we have the identity

det(d) = det(d),

from which it follows that

Z    
Z = D[] det3/4 T(1) det1/2 () det1/2 T(1) .

The operator T(1) is the transverse part of the Hodge Laplacian acting on 1forms:

T(1) := (d)(1) .

Applying identity for the Hodge Laplacian (p) [23]

24




det (p) = det (d)(p) det (d)(p1) ,
we get  
det T(1) = det (1) /det ()

and hence Z
D[] det1/4 (1) det3/4 () .


Z=

The space of harmonic forms (of any order) is a finite set. Hence, the integration over
harmonic forms (3.1.2) is a simple sum.

3.2 Appendix: Path Integral in Quantum Mechanics

The amplitude hqF |eiHT |qI i of a quantum-mechanical system (in natural units ~ = c = 1)
with Hamiltonian H to propagate from a point qI to a point qF in time T is governed by
the unitary operator eiHT , or more completely, by the complex Lorentzian path integral
(in real time): Z RT
hqF |eiHT |qI i = D [q(t)] ei 0
dtL(q,q)

. (17)

For example, a quantum particle in a potential V (q) has the Hamiltonian operator :
H = p2 /2m + V (
q ), with the corresponding Lagrangian: L = 21 mq 2 V (q), so the path
integral reads: Z RT
dt[ 21 mq2 V (q)]
hqF |eiHT |qI i = D [q(t)] ei 0 .

It is somewhat more rigorous to perform a so-called Wick rotation to Euclidean time,

which means substituting t it and rotating the integration contour in the complex
tplane, so that we obtain the real path integral in complex time:
Z R Z RT
iHT 0T dtL(q,q)
1 2
hqF |e |qI i = D [q(t)] e = D [q(t)] e(i/~) 0 dt[ 2 mq V (q)] ,

known as the Euclidean path integral.

One particularly nice feature of the path-integral formalism is that the classical limit
of quantum mechanics can be recovered easily. We simply restore Plancks constant ~ in
(17) Z RT
hqF |e(i/~)HT |qI i = D [q(t)] e(i/~) 0 dtL(q,q)

,

25
and take the ~ 0 limit.
R
Applying the stationary phase method (or steepest descent) , see
(i/~) 0T dtL(qc ,qc ) 11
[25]), we obtain e , where qc (t) is the recovered classical path determined
by solving the Euler-Lagrangian equation: (d/dt)(L/q) (L/q) = 0, with appropriate
boundary conditions.

References
[1] J.E. Marsden, A. Tromba, Vector Calculus (5th ed.), W. Freeman and Company, New
York, (2003)

[2] V. Ivancevic, Symplectic Rotational Geometry in Human Biomechanics, SIAM Rev.

46, 3, 455474, (2004).

[3] V. Ivancevic, T. Ivancevic, Geometrical Dynamics of Complex Systems. Springer,

Dordrecht, (2006).

[4] V. Ivancevic, T. Ivancevic, Applied Differfential Geometry: A Modern Introduction.

World Scientific, Singapore, (2007)

[5] V. Ivancevic, T. Ivancevic, Quantum Leap: From Dirac and Feynman, Across the
Universe, to Human Body and Mind. World Scientific, Singapore, (2008)

[6] R. Abraham, J. Marsden, T. Ratiu, Manifolds, Tensor Analysis and Applications.

Springer, New York, (1988).

[7] Y. Choquet-Bruhat, C. DeWitt-Morete, Analysis, Manifolds and Physics (2nd ed).

North-Holland, Amsterdam, (1982).

[8] C. Voisin, Hodge Theory and Complex Algebraic Geometry I. Cambridge Univ. Press,
Cambridge, (2002).

[9] G. de Rham, Differentiable Manifolds. Springer, Berlin, (1984).

[10] H. Flanders, Differential Forms: with Applications to the Physical Sciences. Acad.
Press, (1963).
11
R +
To do an exponential integral of the form: I = dq e(i/~)f (q) , we often have to resort to the
following steepest-descent approximation [25]. In the limit of ~ small, this integral is dominated by the
minimum of f (q). Expanding: f (q) = f (a) + 21 f (a)(q a)2 + O[(q a)3 ], and applying the Gaussian
integral rule:
Z + r   1
1 2 2 2~
dq e 2 ax = I = e(1/~)f (a) eO(~ ) .
2
, we obtain:
a f (a)

26
[11] C.W. Misner, K.S. Thorne, J.A. Wheeler, Gravitation. W. Freeman and Company,
New York, (1973).

[12] I. Ciufolini, J.A. Wheeler, Gravitation and Inertia, Princeton Series in Physics,
Princeton University Press, Princeton, New Jersey, (1995).

[13] D.K. Wise, p-form electrodynamics on discrete spacetimes. Class. Quantum Grav.
23, 51295176, (2006).

[14] E. Witten, Supersymmetry and Morse theory. J. Diff. Geom., 17, 661692, (1982).

[15] F.L. Teixeira and W.C. Chew, Lattice electromagnetic theory from a topological
viewpoint, J. Math. Phys. 40, 169187, (1999).

[16] B. He, F.L. Teixeira, On the degrees of freedom of lattice electrodynamics. Phys. Let.
A 336, 17, (2005).

[17] R.P. Feynman, Spacetime approach to nonrelativistic quantum mechanics, Rev.

Mod. Phys. 20, 367387, 1948;
R.P. Feynman and A.R. Hibbs, Quantum Mechanics and Path Integrals, McGraw
Hill, New York, 1965.

[18] R.P. Feynman, Spacetime approach to quantum electrodynamics, Phys. Rev. 76,
769789, 1949;
Ibid. Mathematical formulation of the quantum theory of electromagnetic interaction,
Phys. Rev. 80, 440457, 1950.

[19] B. Thaller, Visual Quantum Mechanics, Springer, 1999;

Ibid. Advanced Visual Quantum Mechanics, Springer, 2005.

[20] L.D. Faddeev and V.N. Popov, Feynman diagrams for the YangMills field. Phys.
Lett. B 25, 29, 1967.

[21] J. Ambjorn, R. Loll, Y. Watabiki, W. Westra and S. Zohren, A matrix model for
2D quantum gravity defined by Causal Dynamical Triangulations, Phys.Lett. B 665,
252256, 2008.;
Ibid. Topology change in causal quantum gravity, Proc. JGRG 17, Nagoya, Japan,
December 2007;
Ibid. A string field theory based on Causal Dynamical Triangulations, JHEP 0805,
032, 2008.

[22] R. Loll, The emergence of spacetime or quantum gravity on your desktop, Class.
Quantum Grav. 25, 114006, 2008;
R. Loll, J. Ambjorn and J. Jurkiewicz, The Universe from scratch, Contemp. Phys.

27
 47, 103117, 2006;
J. Ambjorn, J. Jurkiewicz and R. Loll, Reconstructing the Universe, Phys. Rev. D
72, 064014, 2005.

[23] J. Gegenberg, G. Kunstatter, The Partition Function for Topological Field Theories,
Ann. Phys. 231, 270289, 1994.

[24] Y. Choquet-Bruhat, C. DeWitt-Morete, Analysis, Manifolds and Physics (2nd ed).

North-Holland, Amsterdam, 1982.

[25] A. Zee, Quantum Field Theory in a Nutshell, Princeton Univ. Press, 2003.

28