Professional Documents
Culture Documents
Applied Differential
Equations of
Mathematical Physics
This page intentionally left blank
Lecture Notes in
Applied Differential
Equations of
Mathematical Physics
Luiz C L Botelho
Federal University, Brazil
World Scientific
NEW JERSEY • LONDON • SINGAPORE • BEIJING • SHANGHAI • HONG KONG • TA I P E I • CHENNAI
This page intentionally left blank
Published by
World Scientific Publishing Co. Pte. Ltd.
5 Toh Tuck Link, Singapore 596224
USA office: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601
UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE
For photocopying of material in this volume, please pay a copying fee through the Copyright
Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to
photocopy is not required from the publisher.
ISBN-13 978-981-281-457-9
ISBN-10 981-281-457-4
Printed in Singapore.
Foreword
Contents
Foreword . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
1.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2. The Laplace Differential Operator
and the Poisson–Dirichlet Potential Problem . . . . . . . . 1
1.3. The Dirichlet Problem in Connected Planar Regions:
A Conformal Transformation Method for Green Functions
in String Theory . . . . . . . . . . . . . . . . . . . . . . . . 13
1.4. Hilbert Spaces Methods in the Poisson Problem . . . . . . . 18
1.5. The Abstract Formulation of the Poisson Problem . . . . . 21
1.6. Potential Theory for the Wave Equation in R3 —
Kirchhoff Potentials (Spherical Means) . . . . . . . . . . . . 23
1.7. The Dirichlet Problem for the Diffusion Equation —
Seminar Exercises . . . . . . . . . . . . . . . . . . . . . . . 27
1.8. The Potential Theory in Distributional Spaces —
The Gelfand–Chilov Method . . . . . . . . . . . . . . . . . 28
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
Appendix A. Light Deflection on de-Sitter Space . . . . . . . . . 31
A.1. The Light Deflection . . . . . . . . . . . . . . . . . . . . . . 31
A.2. The Trajectory Motion Equations . . . . . . . . . . . . . . 34
A.3. On the Topology of the Euclidean Space–Time . . . . . . . 36
ix
August 6, 2008 15:46 9in x 6in B-640 fm
Contents xi
Contents xiii
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
August 6, 2008 15:46 9in x 6in B-640 ch01
Chapter 1
1.1. Introduction
∆(N ) : C 2 (Ω → C(Ω)
n
∂2
f (x) → (∆f )(x) = f (x1 , . . . , xn ). (1.1)
i=1
∂x2i
We now have the following simple result about the kernel of the above
linear transformation in R3 (holding true for any RN , with N ≥ 2).
1
August 6, 2008 15:46 9in x 6in B-640 ch01
Lemma 1.1. The kernel of the Laplacean operator as defined by Eq. (1.1)
has infinite dimension in R3 .
2 ¯
Proof. Let f¯ ∈ C 2 (C × R) such that ∂∂2 w f
(w, t) exists and the function
1 2 3
defined below (x = x, x = y, x = z — the usual classical notation in
potential theory in the physical space R3 )
As a second step to the proof of Theorem 1.1, we point out that the
sequence of functions
1
∇ · U (δ) = d3r f (r )∇ · G(δ) (|r − r |) (1.16)
4π Ω
August 6, 2008 15:46 9in x 6in B-640 ch01
3 1
I=+ d r f (r )∇r . (1.17)
Ω 4π|r − r |
2
|r − r |2 (x − x )
lim d3r ∇rx − +
= 0. (1.18)
δ→0 Bδ (
r) δ 2 |r − r |3
1
d3r f (r ) ∇r
Ω |r − r |
1
=− d3r f (r ) ∇r
Ω |r − r |
1
= −O f (r )
∠r ) dS(r )
cos(N
∂Ω |r − r |
1
+ d3r ∇r · f (r ) . (1.19)
Ω |r − r |
Here cos(N ∠r ) is the cosine between the normal field of the globally ori-
ented surface ∂Ω and the point r, argument of the function U (r ). At this
point we are using the fact that f ∈ C 1 (Ω̄). However, Eq. (1.19) has a rig-
orous meaning for f ∈ L2 (Ω̄) (left as an exercise to our diligent reader!).
It is important to remark that at this point of our exposition all these
topological–geometrical constraints imposed on the surface ∂Ω by means
of Gauss theorem must be imposed as complementary hypothesis on the
geometrical nature of the Poisson problem.
By deriving a second time the left-hand side of Eq. (1.19) and taking
into account that in the surface integral we always have r = r (we do
not evaluate the solution U (r ) at the boundary points) and obviously
∇r f (r ) ≡ 0, we arrive at the following result (without bothering ourselves
August 6, 2008 15:46 9in x 6in B-640 ch01
G(δ) |r, r )
1
if |r − r | > δ,
|r − r |N −2
= N
δ 2−N 2 N −2 |r − r |
2− − if |r − r | ≤ δ.
(N − 2)wN N 2 δ
(1.24)
This case can be seen as coming from the general case Eq. (1.23) after
considering the (distributional) limit
Γ 22 1 r |
r− 1
lim √ e (2−N )g|
=− g|r − r |. (1.26)
N →2 2( π)2 N − 2 2π
0
U · ∆U +|∇U | (r )d r = O (U · ∇N U )(r )dΓ(r ) = 0.
2 3
(1.28)
Ω ∂Ω
As a consequence,
|∇U |2 (r )d3r = 0 ⇔ U (r ) ≡ 0 in Ω. (1.29)
Ω
Let
1
G(r, r ) = H(r, r ) + , (1.30)
4π|r − r |
where H(r, r ) is a harmonic function in Ω in relation to the variable r ,
satisfying the following condition:
1
H(r, r )|r ∈∂Ω = − . (1.31)
4π|r − r | r ∈∂Ω
Then we have the following explicit representation:
Theorem 1.2. We have the integral representation for the Dirichlet
problem
− 1
U (r ) = 4π O g(
r )∇N G(
r ,
r ) dS(
r )
(−1/2π if r ∈ ∂Ω − exercise) ∂Ω
1 1
=− O g(r )∇N H(r, r ) + dS(
r
) . (1.32)
4π ∂Ω 4π|r − r |
Proof. Let us consider the regularized form for the Green function as
proposed by the Poincaré ansatz
G(δ) (r, r ) = H (δ) (r, r ) + U (δ) (r − r ). (1.33)
Here
∆r H (δ) (r, r ) = 0,
(1.34)
H (δ) (r, r ) = −U (δ) (r − r ),
r ∈∂Ω
with
1
, if |r − r | > δ,
4π|
r − r |
U (r − r ) =
(δ)
(1.35)
1 1 |r − r |2
3− , if |r − r | ≤ δ.
4π 2δ δ2
By applying the second Green identity to the pair of functions U (r )
and G(δ) (r, r ), we obtain the relation
U (r ) ∆r U (δ) (r, r ) d3r
Ω
= O U (r ) ∇N G(δ) (r, r )dS(r ). (1.36)
∂Ω
August 6, 2008 15:46 9in x 6in B-640 ch01
1
−g(r )∇N ∇N dS(r ). (1.39c)
|r − r |
August 6, 2008 15:46 9in x 6in B-640 ch01
The same result holds true for the case of the Newmann problem written
below:
∆ U (r ) = 0, for r ∈ Ω,
(1.44a)
∇N U (r ) = g(r ) O g(r )dS(r ) ≡ 0 ,
∂Ω
August 6, 2008 15:46 9in x 6in B-640 ch01
with
U (r ) = + O g(r )G(r, r )dS(r ), (1.44b)
∂Ω
where
βn (r )βn (r )
G(r, r ) = , (1.44c)
n
λn
Another point worth remarking is that Eqs. (1.38), (1.39a) and (1.39c)
have potentialities for problems involving weak perturbations of the domain
boundary ∂Ω(0) → ∂Ω(0) + ε ∂ Ω(1) , since we have for instance the
explicit perturbation result for the integrand of Eq. (1.38) for a given
parametrization of the perturbed surface
(0)
= W1 (u, v)ê1 + W2 (u, v)ê2 + W3 (u, v)ê3 = W
W + εW (1) (u, v)
(u, v) = (W + εW ) × (W + εW ) = N
(0) (1) (0) (1)
N (0) + εN (1) + · · ·
|(W
(0) + εW (1) ) × (W
(0) + εW
(1) )| 3 R
·(0) εW
1 1 (r − W (1) )
= + + ···
|r − W
(0) − εW
(1) | |r − W
(0) | |r − W (0) |3
) = Φ(0) (r, W (0) ) + ε∇Φ(1) (r, W (0) ) · ∇W (1) + · · · .
Φ(r, W
(1.47d)
and so on,
Area(∂Ω)
sup |Un (r ) − U (r )| ≤ ||gn − g||C 0 (∂Ω) → 0. (1.47f)
r ∈Ḡ
4πδ
August 6, 2008 15:46 9in x 6in B-640 ch01
Here the function v(r ) is the harmonic function related to the boundary
data
∆ v(r ) = 0,
3 f ( r ) (1.50)
v(r )∂Ω = − d r + g(
r )
.
Ω |r − r | r ∈∂Ω
F
W ⊂ R2 −→ Ω ⊂ R2
(1.51)
(ξ, η) (x(ξ, η), y(ξ, η))
∆x,y Ū (x, y) = 0 (x, y) ∈ Ω,
(b) (1.55)
Ū (x, y)
= f¯(x, y)Γ .
Γ2 2
Then,
one can expect that fn (z) converges into H(Ω) to a function f (z) ∈ H(Ω)
carrying the conformal transformation of W into B1 (0). (Details of our sug-
gested proof are left to our mathematically oriented readers again.)
At this point we intend to present a Hilbert space approach to construct
explicitly such canonical conformal mapping of the domain W into the disc
B1 (0).
Let H = {F (W, C), f : W → C, f is a holomorphic function in W }.
We introduce the following Hilbert space inner product in this space of
holomorphic (analytical) functions in W :
1
f, gc = dz ∧ dz̄ f (z)g(z). (1.57)
2i W
We have the following straightforward result for two elements of the
Hilbert space (H, , c ) obtained by a simple application of the Green
theorem in the plane for any domain W̄ ⊂ W :
1
g(z)h̄(z) dz = (g(z) h (z))dz ∧ dz̄. (1.58)
∂ W̄ =Γ1 2i W̄
If one considers h(z) = z −z0 and W̄ = Br (z0 ) in Eq. (1.58), one obtains
that
g(z) r2
g(z)(z − z0 )dz = r 2
= g(z0 ). (1.59)
|z−z0 |=r |z−z0 |=r z − z0 2πi
In other words,
1
2π
|g(z0 )| ≤ 2 g(z) dxdy
r 2 |z−z0 |≤r
1
2
1
2
2π 1
≤ |g 2 (z)| dxdy dxdy
r2 2 |z−z0 |≤r |z−z0 |≤r
π √
≤ π · r ||g||2H(W )
r2
π 3/2
≤ ||g||2H(W ) ≤ c||g||2W (D) . (1.60)
r
By the Riesz theorem for representations of linear functionals in
Hilbert spaces, the bounded (so continuous) linear evaluation functional on
(H(D), , c ) = HD has the following representation:
1
g(z0 ) = dz ∧ dz̄ R(z, z0 )g(z) (1.61)
2i D
August 6, 2008 15:46 9in x 6in B-640 ch01
and
∞
H
R(z, z0 ) = ek (z)ek (z0 ), (1.62)
k=1
Let Cr = {(f (z))(f (z)) = r < 1} be the curve in the complex plane.
For this special contour, we have that (see Eq. (1.58))
1 g(z) 1
dz = g(z)f¯(z) dz
2πi Cr f (z) 2πir2 Cr
1
= 2 g(z)f (z) dxdy. (1.64)
πr (W \f −1 (Br (0))
Finally let us devise some useful formulae for the Dirichlet problem in
planar “smooth” regions, useful in string path integral theory.
Let U0 f −1 : B1 (0) → R be the associated harmonic function on the
disc B1 (0) related to the Dirichlet problem in the region W . Note that by
construction f (z0 ) = 0 and f (z0 ) > 0 for a given z0 ∈ W . As a consequence
of the mean value for harmonic function, we have that
2π
1
U (z0 ) = (U0 f −1 )(0) = (U0 f −1 )(cos θ, sin θ)dθ
2π 0
1 f (z)
= U (z) z0 dz. (1.67)
2πi ∂W fz0 (z)
From Eq. (1.67) we can write explicitly the Green functions G(z, z0 ) =
fz 0 (z) fz0 (z) for the given canonical regions. For instance, fz0 (z) = R(z −
z0 ) (R2 − z̄0 z) applies BR (0) conformally into B1 (0), leading to the Poisson
formulae in the circle of radius R
2π
1 R2 − r 2
U (r, θ) = dφ U (Reiφ ) . (1.68)
2π 0 R2 + r2 − 2Rr cos(φ − θ)
The function fz0 (z) = (z−z0 ) (z+z0 ) applies conformally the right-half-
+
plane = {z = x + iy, x ≥ 0, −∞ < y < ∞} into B1 (0) and fz 0 (z0 ) = 0.
As a consequence, one has the integral representation for this region
x +∞ U (it) dt
U (x, y) = · (1.69)
π −∞ (t − y)2 + x2
The function fz0 (z) = (z 2 −z02 ) (z 2 −(z̄0 )2 ) solves the Dirichlet problem
in the first quadrant plane, leading to the integral representation below:
∞
4xy t U (t) dt
U (x, y) =
π 0 t4 − 2t2 (x2 − y 2 ) + (x2 − y )2
∞
t U (it) dt
+ . (1.70a)
0 t4 + 2t2 (x2 − y 2 ) + (x2 − y 2 )
In general
fz0 (z ) − fz0 (z)
G(z, z ) = (1.70b)
1 − fz0 (z) fz0 (z )
with f (z0 ) = 0 is the explicit Green function associated with the Dirichlet
problem in the planar region W .
August 6, 2008 15:46 9in x 6in B-640 ch01
Just for completeness let us write the Green function of the Dirichlet
2
problem in the SR in polar coordinates
1
G((ρ, θ), (ρ , θ )) = − log[ρ2 + ρ2 − 2ρρ cos(θ − θ )
4π
ρ2 ρ2
+ log R2 + − 2ρρ cos(θ − θ ) . (1.70c)
R2
Here
∆ρ,θ G((ρ, θ), (ρ , θ ) = 0,
1
G((R, θ), (R, θ )) = + g|r − r | . (1.70d)
2π r =Reiθ
r =Reiθ
Let us pass now to the formulation of the Poisson and Dirichlet problems in
the Hilbert spaces L2 (Ω), with Ω̄ a compact set of RN . This formulation is
known in the mathematical literature as the Hilbert space weak formulation
of the problem.
Let H01 (Ω) be the completion
of the vector space Cc1 (Ω) in relation to the
inner product f, gH 1 = d4 x ∇f ∇g; the weak formulation of Poisson
Ω
problem reads as of
U, ϕH 1 = dN x ∇U (x) · (∇ϕ)(x)
Ω
= dN x(f · ϕ̄)(x) = f, ϕL2 ∀ ϕ ∈ H01 (Ω), (1.71)
Ω
where one must determine a U (x) ∈ H01 (Ω) for a given f ∈ L2 (Ω). That
such (unique) U (x) exists is readily seen from the fact that the functional
linear below is bounded in H01 (Ω):
Lf : H01 (Ω → C)
ϕ→ dN x(f · ϕ̄)(x) = Lf (ϕ) (1.72)
Ω
as a consequence of the Friedrichs inequality
N −1
d x|U | (x) ≤ diam(Ω̄)
N 2
dN x|∇U |2 (x) . (1.73a)
Ω
Ω
||U ||2L2 (Ω) ≤ C(Ω) ||U ||2H 1 (Ω) (1.73b)
0
August 6, 2008 15:46 9in x 6in B-640 ch01
H01 (Ω)
U (x) = Lf (en )en (x)
n
= dN y f¯(y)en (y) ēn (x)
n Ω
= d y N ¯
ēn (x)en (y) f (y) = dN y G(x, y)f¯(y). (1.74)
Ω n Ω
to the space C 2 (Ω) ∩ C 1 (Ω̄), which will lead to the strong solution of the
Poisson problem in Ω
N
∂ ∂
− aij (x) U (x) + V (x)U (x) = f (x) x ∈ Ω. (1.77)
i,j=1
∂xj ∂xi
Here f (x) ∈ L2 (Ω) and F (z) are real Lipschitzian functions (i.e. ∀(z1 , z2 ) ∈
R × R ⇒ |F (z1 ) − F (z2 )| ≤ C|z1 − z2 | for a uniform bound c).
In order to show the existence and uniqueness through a construction
technique, we rewrite Eq. (1.78) into the form of an integral equation (see
Theorem 1.1)
1 F (U (x )) + V (x )U (x ) − f (x ) 3
U (x) = − d x = (T U )(x).
4π |x − x |
(1.79)
One can see easily that T defines an application with domain L2 (Ω) and
range in L2 (Ω). Besides, T is a contraction application in L2 (Ω) for small
domains, as we can see from the following estimate:
Un = T (Un+1 ) (1.81)
Proof. Let AUλ1 = λ1 Uλ1 and AUλ2 = λ2 Uλ2 (λ1 = λ2 ). Thus we have the
identity
A U = f, (1.87)
Let us exemplify the above result for the Poisson problem in bounded
open sets in RN with a Riemannian structure {gab (xa ); a, b = 1, . . . , n}
(manifolds charts).
Thus we consider the following covariant Hilbert space associated with
the given metric structures
L2g (Ω) = closure of f ∈ Cc (Ω) | , ≡ dN x g(x)f (x)f (x) ,
W̄
(1.88a)
√
1
H0,g (Ω) = closure of f ∈ C01 (Ω) | , = dN x g g ab ∂a f ∂b f ,
W̄
(1.88b)
p
H0,g (Ω) = closure of f ∈ C0p (Ω) | , H 0
√
= dN x g(∇a1 . . . ∇a(p/2) f )(x)
W̄
× {g a1 ,a(p/2+1) . . . g a(p/2) ap }(x)(∇a(p/2+1) . . . ∇ap f )(x). (1.88c)
1
In the covariant Hilbert space H0,g (Ω), we consider the functional
(positive-definite) associated to the Laplace–Beltrami operator ∆g =
√
− √1g ∂a ( g δ ab ∂b ) acting on real functions
√ √
F∆q (ϕ) = dN x g ϕ(−∆g )ϕ (x) − dN x g f (x)ϕ(x). (1.89)
W̄ W̄
August 6, 2008 15:46 9in x 6in B-640 ch01
First, let us see that the integral operator T has as its domain C 1 (Ω) ⊂
L (Ω) and as its range the functional space C 2 (Ω). However, it can be
2
3 3
Here, Sct (r ) = ∂{Bct (r )} is the spherical surface centered at the point
r = (x, y, z) with radius |r − r| = ct. The spherical parameter α, β, and γ
are defined as usual as
α = cos ϕ sin θ,
β = sin ϕ sin θ,
(1.95)
γ = cos θ,
dΩ = sin θdθ · dϕ.
U (1) (r, 0, [f ]) = 0,
∂ (1) t
U (r, 0, [f ]) = f (x, y, z) × lim dΩ = f (r ),
∂t t→0+ 4π 3 (
Sct r) (1.97)
U (2) (r, 0, [g]) = g(r ),
∂
U2 (r, 0, [g]) = 0 (exercise).
∂t
The main differentiability property of the above-written wave potentials
is the following:
2π π
(1) t
∆r U (r, t, [f ]) = dϕ sin θdθ∆r f (x + αct, y + βct, z + γct) .
4π 0 0
(1.98)
∂U (1) (r, t, [f ])
∂t
U (1) (r, t, [f ]) t ∂ ∂ ∂
= + cα + cβ + cγ f (x̄, ȳ, z̄)d2 Ω
t 4π 3 (
Sct r) ∂ x̄ ∂ ȳ ∂ z̄
August 6, 2008 15:46 9in x 6in B-640 ch01
n·∇
U (1) (r, t, [f ]) 1 ∂ ∂ ∂
= + α +β +γ f (x̄, ȳ, z̄)dA
t 4πct 3 (
Sct r) ∂ x̄ ∂ ȳ ∂ z̄
(Gauss theorem)
U (1) (r, t, [f ]) 1
= + (∆ f )(x̄, ȳ, z̄)dx̄ dȳ dz̄
t 4πct 3 (
Bct r)
ct
U (1) (r, t, [f ]) 1 2
= + dρ d A(∆f )(x̄, ȳ, z̄) (1.99)
t 4πct 0 3 (
Sct r)
since
dx̄ dȳ dz̄ = dρ d2 A,
(1.100)
0 ≤ ρ ≤ ct.
Since we have the usual Leibnitz rule for derivatives inside integrals
ct
d
f (x) dx = f (ct) · c. (1.101)
dt 0
We have the result for evaluation of the second-time derivative of Eq. (1.94):
Theorem 1.7. The solution for the Cauchy (initial values) problem for
the wave equation with “smooth” C 2 (Ω) data (including an external forcing
F (r, t), continuous in the t-variable), is given by the wave potentials
2
∂ U (r, t) 1
= 2 ∆r U (r, t) + F (r, t),
∂ t 2 c
U (
r , 0) = f (
r ), r ∈ Ω, (1.104)
Ut (r, 0) = g(r ), r ∈ Ω,
E(t) = E(0)
= lim −c O 2
U · ∆U (r, 0) + lim O Ut · ∇N U dS = 0.
R→∞ BR (0) 3 (0)
SR
(1.109)
Let us consider the following Dirichlet problem for the diffusion equation in
a given domain Ω ⊂ R3 with a nontrivial time-dependence on the boundary
(an interior problem):
1 ∂
∆ U (r, t) = 2 U(r,t) ,
a ∂t
U (r, 0) = f (r ) ∈ C ∞ (Ω), (1.110)
U (r, t)∂Ω = g(r, t)∂Ω ∈ C ∞ (∂Ω).
It is clear that the formal solution written above satisfies the boundary
condition in Eq. (1.110) if one can determine the density function Φ(r, t)
from the integral equation coming from the imposition of the boundary
condition as given by Eq. (1.110) (exercise):
U (r, t)∂Ω = g(r, t)∂Ω = U (2) (r, t)∂Ω
1 t
Π(r , t) − 4a2r(t−ζ)
2
+ −Φ(r, t)∂Ω + dζ O e · r
8πa2 0 ∂Ω (t − ζ)2
× cos(N ∠r )dS(r ) . (1.113)
One can show that for small densities Φ(r, t) → λΦ(r, t) with λ 1 ,
one can solve Eq. (1.113) by means of the fixed point theorem of Banach
(exercise) and yielding the solution as a power series in λ (the size of the
area of ∂Ω).
Seminar Exercises
aα Dxα U (x) = δ (N ) (x). (1.114)
|α|≤m
At this point one can consider the following analytical regularized asso-
ciated problem:
|r · r |λ
(L U )(r ) = aα Drα v (λ) (r ) = n−1 · (1.117)
|α|≤m
ωn π 2 Γ λ+n 2
Due to the linearity of Eq. (1.116), we should search for a solution of this
analytical regularized equation in the linear-superposing form of a spherical
mean (Radon transforms)
(λ)
U (r ) = dΩ(r ) v (λ) (r, r ). (1.118)
(n)
S1 (0)
d |ξ|λ
L ωa v λ (ξ) = n−1 · (1.119)
dξ ωn π 2 Γ λ+n 2
d
L ωa G(ξ, ξ ) = δ (1) (ξ − ξ ). (1.121)
dξ
August 6, 2008 15:46 9in x 6in B-640 ch01
2α
d d2 |ξ|λ
L ωa = − ((ω1 )2 + (ω2 )2 + (ω3 )2 ) 2 v λ (ξ) = ·
dω d ξ ω3 π Γ λ+n2
(1.122)
References
The most general covariant second-order equation for the gravitation field
generated by a given (covariant) enegy–matter distribution on the space–
time is given by the famous Einstein field equation with a cosmological
constant Λ with dimension (length)−2 (Ref. 5), namely,
1
Rµν (g) − gµν R + Λgµν (x) = 8πGTµν (x), (A.1)
2
where x belongs to a space–time local chart.
It is well known that studies on the light deflection by a gravitational
field generated by a massive point-particle with a pure time like geodesic
trajectory (a rest particle “sun” for a three-dimensional spatial space–time
section observer!) is always carried out by considering Λ ≡ 0.6,7
Our purpose in this appendix is to understand the light deflection
phenomena in the presence of a nonvanishing cosmological term in Einstein
equation (A.1), at least on a formal mathematical level of solving trajectory
motion equations.
Let us, thus, look for a static spherically symmetric solution of Eq. (A.1)
in the standard isotropic form6,7
B 1 B A B 1 B
Rtt = − + + − = −ΛB, (A.5)
2A 4 A A B r A
B 1 B A B 1 A
Rrr ≡ − + − = ΛA, (A.6)
2B 4 B A B r A
r A B 1
Rθθ = −1 + − + + = Λr2 , (A.7)
2A A B A
or equivalently
α
A(r) = , (A.10)
B(r)
αΛr2 β
B(r) = +α+ , (A.11)
3 r
with β denoting another integration constant.
In the literature situation,6,7 one always considers the case Λ = 0 in
a pure classical mathematical vacuum situation context, the so-called de-
Sitter vacuum pure gravity. However, in our case it becomes physical to
consider that our solution depends analytically on the cosmological con-
stant. In other words, if the parameter Λ → 0 in our solution, it must
converge to the usual Schwarzschild solution with a mass singularity at the
origin r = 0, that is our boundary condition hypothesis imposed on our
solution.
August 6, 2008 15:46 9in x 6in B-640 ch01
−1
Λr2 2M G
− +1− (dr)2 − r2 [(dθ)2 + (sin2 θ)(dφ)2 ].
3 r
(A.12)
At this point let us comment that for the space–time region exterior to
the spatial sphere r > ( 3mG
Λ )
(1/3)
, the field gravitation approximation leads
to the antigravity (a repulsion gravity force) if Λ < 0; So, explain from
this Einstein Gravitation theory of ours the famous “Huble accelerating
Universe expansion”.
In what follows we are going to consider a nonvanishing Λ < 0 and study
the path of a light ray on such negative cosmological constant Einstein
manifold Eq. (A.12).
We have the following null-geodesic equation for light propagating in
θ = π/2 plane (Einstein hypothesis) for light propagation in the presence
of the sun (Sec. A.2 for the related formulae):
2 2
1 dr dφ
0 = B(r) − − r2 . (A.13)
B(r) dt dt
dφ B(r)J
= , (A.14)
dt r2
where J is an integration constant.
After substituting Eq. (A.14) into Eq. (A.13) we have the following
differential equation for the light trajectory as a function of the deflection
angle φ
2
dr dφ J 2 B 3 (r)
+ − B 2 (r) = 0, (A.15)
dφ dt r2
dr
By supposing a deflection point rm where dt = 0 and, thus, J =
rm / B(rm ), we get the deflection angle
rm 1
dr rm dU
∆1 φ = B(rm ) B(r) (1/2)
= 2 − U 2 ) − 2M G(U 3 − U 3 )](1/2)
,
∞ r2 [ r 2 − r2 ] 0 [(Um m
m
(A.17)
which is exactly the one given in the pure (Λ = 0) Schwarzschild famous
case. However, if one supposes that there is no deflection (a continuous
monotone trajectory r = r(φ)!), the total deflection angle now depends on
the cosmological constant and is given formally by the expression below:
rm
1 1
∆2 φ = dr ' + = ∆1 φ. (A.18)
3−ΛJ 2
∞ r2 − 12 + 2mG
r 3 ( 3J 2 )
r r3 1 + 2MG−r
The body trajectory (t(p), r(p), θ(p), ϕ(p)) in the presence of the gravita-
tional field generated by the metric Eqs. (A.2)–(A.10) is described by the
following geodesic equations:
d2 t B dr dt
+ = 0, (A.19)
d2 p B dp dp
A B
2 2 2 2
d2 r dr r dθ r sin2 θ dφ dt
+ − − + = 0,
d2 p 2A dp A dp A dp 2A dp
(A.20)
2 2
d θ 2 dθ dr dφ
+ − sin θ · cos θ = 0, (A.21)
d2 p r dr dp dp
d2 φ 2 dφ dr dφ dθ
+ + 2cotg(θ) = 0. (A.22)
d2 p r dp dp dp dp
August 6, 2008 15:46 9in x 6in B-640 ch01
d dφ
n + nr2 + 2 n sin θ = 0, (A.24)
dp dp
dφ 2
r (p) sin2 (θ(p)) = J, (A.25)
dp
B B
2 2
d2 r dr dθ J 2B
− − rB − + = 0, (A.26)
d2 p 2B dp dp r3 2B
d2 θ 2 dθ dr cos θ J 2
+ − = 0. (A.27)
d2 p r dr dp sin3 θ r4
B B
2
d2 r dr J 2B
2
− − 3
+ =0 (A.28)
d p 2B dp r 2B
2 3
dφ r 2
By writing r as a function of φ and using Eq. (A.25) dt B = J! , we
get our final trajectory equation
(1/2)
dr 1 B BE
= ±r 2
− 2 − 2 , (A.30)
dφ J2 r J
with Tij being the matter energy tensor and e⊥ p the basis two-plane
orthogonal to ep , one can in principle write the sectional curvatures
K(ep ∧eq ) in terms of the quadratic energy–momentum sectional curvatures
Tij (ep , eq ) at least for “short-time” cylindrical geometrodynamical space–
time configurations as expected in a Quantum theory of gravitation (see
August 6, 2008 15:46 9in x 6in B-640 ch01
Chapter 2
H = H0 + V, (2.1)
or equivalently
39
August 6, 2008 15:46 9in x 6in B-640 ch02
We now show the existence of the wave operators equation (2.5) for a
certain class of self-adjoint operators (H, H0 ).
d
(W (t)f ) = (eitH (H − H0 )e−itH0 )f. (2.8)
dt
August 6, 2008 15:46 9in x 6in B-640 ch02
(exp(−it(−∆))ψa )(x) = 3n
(xj − aj )e
4(1+it)
. (2.12)
(1 + it) 2
j=1
|(exp(−it(−∆))ψa )(x)|
N
3N −1−δ
−N |x − a|−( 2 −1)+δ x − a 2 |(x − a)|2
≤2 2
1 + it exp − .
|1 + it|1+δ 4(1 + t2 |
(2.13)
August 6, 2008 15:46 9in x 6in B-640 ch02
1
As a consequence and by choosing 0 < δ < 2, we have the combined
estimate involving the scattering potential
N
N |V (x)|x − a|−( 2 −1)+δ |
|(V exp(−it(−∆))ψa )(x)| ≤ Ca 2− 2 , (2.14)
|1 + it|1+δ
3N
−1−δ −Ω2
where t > 0 and Ca = supΩ∈R (2Ω) 2 e L∞ (RN ) < ∞, here
1
Ω = |(x − a)|/(1 + t2 ) 2 .
If we choose δ < min( 12 , ), we have
NV
(V e−it(−∆) ψa )(x)L2 (RN ) ≤ , (2.15)
|1 + it|1+δ
which obviously leads to the searched integrability result
∞ ∞
dt
V e−it(−∆) ψa L2 (RN ) ≤ NV 1+δ
<∞ (2.16)
δ+ δ + |(1 + it)|
H0
z }| {
Let H = (− 12 ∆) +V be a self-adjoint operator defined by an Enss potential
with M = 1 (see Eq. (2.13)). Then we have the result5 :
Enss Lemma. Let f ∈ Hc (H) be a state belonging to the continuity sub-
space of the Enss Hamiltonian. Then there is a sequence of times tn such
that limn→∞ tn = +∞ (tn ≤ tn+1 , ∀n ∈ Z) and
n
lim dt|F (|x| < n)e−iH(t+tn ) (H + i1)f L2 = 0. (2.21b)
n→∞ −n
Let us now define the Enss’ scattering states for a given state f ∈ Hc (H)
since
14n
0 = lim dRhz (R) ≥ lim nhz (13n). (2.28)
n→∞ 13n n→∞
where EH ([a, b]) denotes the spectral projection of the full Enss
Hamiltonian H, associated with the spectral interval [a, b]. (Note that
Hc (H) ⊂ R+ since (H + z1)−1 is a compact operator — see Appendix C.)
Let us consider φ̃(x) ∈ C ∞ (R) such that φ(x) = 1 for x ∈ [73a2 ,
(b − a)2 /4] and φ(x) ≡ 0 for r ∈ [72a2 , (b − a)2 /4]. We now consider the
“physical free-energy packet” of scattering states
1
φn = ϕ̃ − ∆ f¯n . (2.31)
2
At this point we can see that the family of functions {F0 (Aj )}j∈I is a
unity decomposition if {Aj }j∈J is a disjoint decomposition of RN
F0 (Aj ) = S(y)dN y = S̃(0) = 1. (2.34)
j∈I
Another important point to call the readers’ attention is the fact that
the quantum localized states F0 (A)φn all have their momenta in the Ball
with radius b. This result is easily obtained through the result that for
m(A) < ∞, we have
and
lim F0 (B12n (fn )) = 0. (2.38b)
n→∞
j∈I L2
Let us finally consider the physical Enss localized scattering states by its
classical motion direction. Let hj (p) ∈ C0∞ (RN ) be the functions satisfying
the conditions
hj (p) = 0 for p · êj ≤ −2a,
hj (p) + hj (−p) = 1 for |p| < b (2.39)
(j)
with êj denoting the vector in the direction of the cone’s axis C12n as
previously stated.
The Enss OUT and IN states are explicitly given by5
(j)
φn (j)OUT = F0 (C12n )hj (p)φn , (2.40a)
(j)
φn (j)IN = F0 (C12n )hj (−p)φn . (2.40b)
lim |ψn , ψn |
n→∞
IN OUT
= lim ψn , ψn − φn (j) + φn (j)
n→∞
j∈I j∈I
IN OUT
− φn (j) + φn (j)
j∈I j∈I
IN OUT
≤ lim ψn , ψn − φn (j) + φn (j)
n→∞
j∈I j∈I
IN OUT
+ lim ψn , φn (j) + lim ψn , φn (j)
n→∞ n→∞
j∈I j∈I
(2.43a)
≤ lim ψn ψn − φn (j)IN + φn (j)OUT
n→∞
j∈I j∈I
(2.43b)
August 6, 2008 15:46 9in x 6in B-640 ch02
IN
+ lim ψn , φn (j) (2.43c)
n→∞
j∈I
+ OUT
+ lim ψn , W φn (j) . (2.43d)
n→∞
j∈I
References
Appendix A
IN
Theorem A.1. Let us consider the Enss scatterbug states φn (j)OUT . Then
we have the result
∞
lim dtF (|x| ≤ n + at)e−iH0 t (H0 + i1)φn (j)OUT = 0, (A.1)
n→∞ 0
∞
lim dtF (|x| ≤ n − at)e−iH0 t (H0 + i1)φn (j)IN = 0. (A.2)
n→∞ 0
Just in order to simplify the estimates which follows and choose the axis
(k)
x1 parallel to fj under consideration. Let us thus consider
(j)
Fm = (H0 + i1)F0 (C12n ∩ {2n + m ≤ x1 ≤ 2n + m + 1}). (A.5)
August 6, 2008 15:46 9in x 6in B-640 ch02
Obviously
∞
(j)
(H0 + i1)F0 (C12n ) = Fm . (A.6)
m=0
is in the interval a ≤ p1 ≤ b.
Let us now pass to the estimates. First, the estimate equation (A.3)
reduces to the one-dimensional case as a consequence of the dependence of
solely the coordinate x1 into our sliced up states equation (A.7).
We have thus reduced our analysis to the one-dimensional case
∞ 2
!
−i ∂ ∂ t
dtF (x1 ≤ n + at)e 2x
1 φ(x1 , x̂) 2 (A.8)
0 L
L2
2 # iy12 %
∂ iy12
2
= (1 + t)t − i + 2a e ib21 /2t
−1+ + 2t
φ(y1 , x̂)
∂y1 2t 2 2
L
(1 + t)t 2 ∂
≤ |P1∗ (|y1 |)|(φ(y1 , x̂)) + |P2∗ (|y1 |)| φ(y1 , x̂)
t3 ∂y1
2
∂
+ |P3 (|y1 |)| 2 φ(y1 , x̂)
∗
2, (A.12)
∂y1 L
where P∗ (|y1 |), = 1, 2, 3 are polynomials satisfying the relations (x1 ≥ ã)
'
P3∗ (|y1 |) = |y1 |6 48,
4a|y1 |6 2|y1 |5
P2∗ (|y1 |) = + ,
48 8
1 1
∗
P1 (|y1 |) ≥ P |y1 |, ≥ P |y1 |, .
(A.13)
ã t
We have this for x1 ≤ at the estimate
2 # %
iy12 (iy12 ) 1
(1 + t)t2 − i ∂ + 2a e iy12 /2t
− 1 + + · φ(y , x̂)
∂y1 2t (2t) 2
1
a ∗ ∂
≤ +1 ∗
P1 (|y1 |)φ(y1 , x̂) + P2 (|y1 |) ∂y1 φ(y1 , x̂)
ã
"
∂ 2 φ(y1 , x̂)
= a P1∗ ∂
∂
+ P3∗ (|y1 |) + P ∗
ip1
∂ 2 y1 ã ∂p1 2
∂p1
&
∂
+ P3∗ (−p1 )2 F [φ](p1 , p̂) 2 ≤A (A.14)
∂p1 L
1 1
≤ · (A.17)
(−x1 + 2at)2 (v + at)2
We have
(j)
F (x1 < n + at)e−iH0 t (H0 + i1)ξk (p)φn (x)
∞
−iH0 t (j)
= F (x1 < n + at)e Fn ξk (p)φn (x)
m=0
∞
(j)
≤ F (x1 + n + at)e−iH0 t Fn ξk (p)φn (x)
m=0
∞
1
≤A .
m=0
|(1 + t)(m + n + at)2 |
Appendix B
Let us show the following technical result used in the proof of the Enss
theorem.
IN
Lemma B.1. Let W± be the Enss wave operators and φn (j)OUT considered
in Sec. 2.2, then we have the result ( = L2 ).
Proof. Since
t
(W (t) − W (0)f ≤
dζ(V e −iH0 ζ
f )
(B.3)
0
we have that
∞
+
(W − 1)φn (j) ≤ OUT
dζV e −iH0 ζ
φn (j)OUT
0
∞
−1 −iHo ζ OUT
≤ V (H0 + i1) op dζF (|x| ≤ n + aζ)e (H0 + i1)φn (j)
0
∞
−1
+ dζV (H0 + i1) F (|x| ≥ n + aζ)op
0
The first term in Eq. (B.4) goes to zero by the long estimate given in
Appendix A.
The second term can be estimated as follows (e−iH0 ζ (· · · ) ≤ (· · · )!)
∞
(H0 + i1)φn (j)OUT h(n + aζ)dζ
0
∞
≤ (H0 + i1)φn (j)OUT h(R)dR . (B.5)
n
August 6, 2008 15:46 9in x 6in B-640 ch02
Since
since F [φn (j)OUT ](p) has support in |p| ≤ b and obviously due to the fact
that h(R) ∈ L1 (R), we have
∞
lim dR h(R) = 0. (B.7)
n→∞ n
We now give a proof of the following additional result in the Enss proof of
asymptotic completeness presented in Sec. 2.3
|φn (j)IN , ψn | ≤ |(W − φn (j)IN , e−iHtn ψ + (I − W − )φn (j)IN , e−iHtn ψ|
(B.9)
IN − IN
≤ |W− e iH0 tn
φn (j) , ψ| + (I − W )φn (j) ψ. (B.10)
As a consequence
since
Appendix C
Chapter 3
3.1. Introduction
58
August 6, 2008 15:46 9in x 6in B-640 ch03
Note that Eq. (3.4) is well defined since we assume that the range of A is
dense on H (0) . LA is a symmetric operator on H (1) since
and due to the fact that Range(A ± i) is dense on H (0) and since A is self-
adjoint it leads to ϕ1 = 0 and π1 = 0. In the same way one can prove that
n− (L̄A ) = 0.
Collecting all the above exposed results, one has the following theorem:
∂ 2 ϕ(t)
= −(Aϕ)(t) (3.10)
∂ 2t
with the initial date
ϕ(0) = U0 , (3.11)
ϕt (0) = v0 . (3.12)
It is worth noting that one has always a weak solution of Eqs. (3.1)
and (3.2) with a nonzero external term
t
ϕ itL̄A U0 i(t−s)L̄A 0
(t) = e + ds e , (3.13)
π −iϕ0 0 −if (s)
≤ sup {(A0, 0)(0) + (F (ϕ1 (s)) − F (ϕ2 (s)), F (ϕ1 (s) − F (ϕ2 (s)))(0) }
0≤s≤T
In this section we generalize the Hilbert space method of Sec. 3.2 to the
case of a wave equation in the presence of damping terms, a very important
problem on statistical continuum mechanics.4
We, thus, consider the abstract problem in the notation of Sec. 3.2:
∂2U ∂U
+ B + (AU )(t) = 0, (3.23)
∂2t ∂t
U (0) = U0 ∈ D(A), (3.24)
Ut (0) = v0 ∈ H (0) , (3.25)
d2 U dU
= −(AU )(x, t) − ν(x) (x, t) + f (x, t) (3.32)
d2 t dt
U (x, 0) = f (x) ∈ H 2m (Ω) ∩ H0m (Ω) (3.33)
2
Ut (x, 0) = g(x) ∈ L (Ω). (3.34)
0 −I
 = (3.35)
A B
August 6, 2008 15:46 9in x 6in B-640 ch03
where γ(Ω) is the universal constant on the Poincaré relation for function
ϕ(x) on H 2 (Ω)
0 2 31
0 17C
B 6
5A
+∞ @iε4
− E(x) 2 E (x)
p + ρ0 (ip) 0 f˜
= lim dpeipx e ρ0
(p)
ε→0+ −∞
g̃
(3.44)
August 6, 2008 15:46 9in x 6in B-640 ch03
with f˜(p) and g̃(p) denoting the Fourier transforms of the functions f (x)
and g(x), respectively, and E (x)p means the scalar product ∇E(x)p for
short.
It is worth calling to the reader’s attention that
! "
0 1 f˜(p)
exp iε
− E(x)
ρ0 p2 + Eρ(x)
0
ip 0 g̃(p)
12 ˜
E(x) 2 E (x) f (p)
= cos ε − p + ip
ρ0 ρ0 g̃(p)
1
2 E (x) 2
−i sin ε − E(x)
ρ0 p + ρ0 ip iεg̃(p)
# (1/2) .
2 E (x)
− E(x) 2 E (x) ε − E(x)
ρ0 p + ρ0 ip f˜(p)
ρ0 p + ρ0 ip
(3.45)
As a consequence of the above formulae we have our discretized path-
integral representation for the initial-value Green function of Eqs. (3.36)–
(3.38), i.e. by writing the evolution operator as an integral operator-
Feynman Propagator
2 3
0 17
7
6
6
5
it4 ∂ E(x) ∂ 0 f
∂x ρ0 ∂x
e
g
+∞
G11 (x, x , t) G12 (x, x , t) f (x )
= dx (3.46)
−∞ G21 (x, x , t) G22 (x, x , t) g(x )
In this section we give a proof for the existence of somewhat reduced wave-
scattering operators associated to the following wave dynamics pair (H̃, H̃0 )
on the H (1) = H 2 (Ω) ⊕ L2 (Ω) (with Ω = RN , N ≥ 3)7 :
0 1
H̃0 = E0
, (3.48)
ρ0 ∆ 0
0 1
H̃ = E(x) . (3.49)
∇· ρ0 ∇ 0
Let us remark that the existence of the strong limit of Eq. (3.50) is
a direct consequence of the validity of the famous Cook–Kuroda–Jauch
integral criterion6 applied to the pair (H̃, H̃0 ) in a dense subset of the
domain of H̃. This integral criterion means that there exists a δ > 0, such
that the following integral is finite for any (f, g) belonging to a dense set of
H (1) , i.e.
T
f
dt (H̃ − H̃0 )e iH̃0 t
< ∞. (3.51)
δ g H (1)
or
T
dt((+∇(Ẽ(x)∇)f (x, t), +∇(Ẽ(x)∇)f (x, t)))L2 (RN ) < ∞, (3.53)
δ
August 6, 2008 15:46 9in x 6in B-640 ch03
where the pair of functions f (x, t) and g(x, t) satisfy the free wave equation
d2 f (x, t) E0
= ∆ f (x, t), (3.54)
d2 t ρ
f (x, 0) = f ∧ (x), (3.55)
∧
ft (x, 0) = g (x), (3.56)
and
d
g(x, t) = f (x, t). (3.57)
dt
Let us choose the initial dates f ∧ (x) and g ∧ (x) as the Fourier trans-
forms of L2 (RN ) functions polynomial-exponentially bounded ( ∈ Z+ and
Λ ∈ R+ ):
∧
f (x) = dN keikx f˜Λ (k); |f˜Λ (k)| ≤ M e−Λ|k| |k| , (3.58)
RN
∧
g (x) = dn keikx g̃Λ (k); |g̃Λ (k)| ≤ M e−Λ|k| |k| . (3.59)
RN
g̃Λ (|k|)
f˜(k, t) = f˜Λ (|k|) cos(|k|t) + sin(|k|t), (3.60)
|k|
∂
g̃(k, t) = (f˜(k, t)). (3.61)
∂t
At this point we observe the estimates for any δ > 0:
T
dt|(Ẽ∆f + ∇Ẽ · ∇f, Ẽ∆f + ∇Ẽ · ∇f )L2 (RN ) (t)|
δ
T
≤ dt sup |Ẽ(x)|2 (k 4 f˜(k, t), f˜(k, t))L2 (RN )
δ x∈RN
By using the explicit representations Eqs. (3.60) and (3.61) and the
simple estimate below for α > 0 with t ∈ [δ, T ) and δ > 0 (with T < ∞)
∞
I(t) = d|k||k|α+ cos2 (|k|t)e−Λ|k|
0
∞
1 ΛU M̄
≤ × dU U α+ e− T ≤ ∈ L1 ([δ, ∞)), (3.63)
tα+1+ 0 t1+α
one can see that by just choosing δ > 0, the existence of the reduced
(T < ∞) wave operators of our pair Eqs. (3.48)–(3.49) is a direct con-
sequence of the application of the Cook–Kuroda–Jauch criterion,6 since
the set of functions polynomial-exponentially bounded is dense on C0 (RN )
(continuous functions vanishing at the infinite), and so, dense in L2 (RN ).7
The complete wave operator (with T → +∞) can be obtained by means
of Zorn theorem applied to the chain of reduced wave operators {W± (T )}
ordered by inclusion, i.e. T1 < T2 ⇒ W± (T1 ) ⊂ W± (T2 ).
∂ 2 U (x, t)
= ∆U(x, t) + ([∇ × h](x, t) × B), (3.64)
∂2t
∂ h(x, t) ∂U
β = ∆h(x, t) + β ∇ × (x, t) × B . (3.65)
∂t ∂t
August 6, 2008 15:46 9in x 6in B-640 ch03
and the (physical) Dirichlet-type boundary conditions with n(x) being the
normal of the boundary of the medium Ω9 :
Let us follow the previous studies of Sec. 3.1 to consider the contraction
semigroup defined by the following essentially self-adjoint operators in the
H-energy space:
namely,
0 i 0
L0 = i∆ 0 0 (3.68)
0 0 −∆
and
0 0 0
V = 0 0 [∇ × ( )] × B . (3.69)
0 −(∇ × [( ) × B]) 0
It is worth to call the reader’s attention that we have proved the essential
self-adjointness of the operator V by using explicitly the boundary condi-
tions on the relationships below:
||(U , ∂t U , h)||H̃Ω
≡ dx(|∇U |2 + |∂t U|2 + |h|2 )(x, t)
Ω
[∇ × (hn )] × B = λn πn , (3.74)
−∇ × [(πn ) × B] = λn hn , (3.75)
which, by its turn, leads to the usual spectral problem for the Laplacean
with the usual Dirichlet boundary conditions on Ω:
λ2n
∆π2n = − π2n , (3.76)
B
As a result of Eqs. (3.75) and (3.76), one finally gets the exponential
bound for the total magneto-elastic energy equation (3.73):
V1 = B1 − V2 , (3.83)
2V2 = −B1 ± (B12 − 4B2 ) 1/2
. (3.84)
We are going to show next that there exists a value of the exponential
weight parameter k defining the above-considered space such that the appli-
cation T̂ written below is a contraction between the space of Hilbert-valued
continuous functions (Eq. (3.85)), i.e.
In order to show this result, let us point out the usual exponential
bound for the free Klein–Gordon abstract propagator (operator norm) (see
Eq. (A.23)):
So, T has a fixed point on Ck ([0, ∞), H (1) ) if one makes the choice of
the value k such that (C(k, µ)) < 1, namely,
1
K> (||L||Lq (R+ ) )p + µ . (3.90)
p
A very important result obtained as a consequence of the above “fixed-
point theorem” analysis is that the sequence of iteratives
(n) (n) U0
U (t) = (T̂ ) (3.91)
V0
converges in Ck ([0, ∞), H (1) ) to the solution of Eq. (3.81) with the error
estimate
References
we have used the fact that the operator −V + iL0 is the generator of a
contraction semigroup on this energy Hilbert space H̃ 1 (Ω) = (H 1 (Ω))3 ⊕
(L2 (Ω))3 ⊕ (L2 (Ω))3 in order to write Eq. (A.1) in a mathematically
rigorous way.
Let us give a proof of this mathematical result by means of a direct
application of the Hille–Yoshida theorem9 to the operator (−V + iL0 ).
August 6, 2008 15:46 9in x 6in B-640 ch03
V x, xH̃ 1 (Ω) = V x, xH̃ 1 (Ω) = iL0 x, xH̃ 1 (Ω) = iL0 x, xH̃ 1 (Ω) , (A.6)
or equivalently
Note that we have used the fact that V is a positive operator on H̃ 1 (Ω).
As a consequence of the above-exposed results we have that −V + iL0
is a generator of a contractive semigroup on the space (1 − PKer(V )∩Ker(L0 ) )
H̃ 1 (Ω), where PKer(V )∩Ker(L0 ) is the orthogonal projection on the Kernel
subspaces of the self-adjoint operators V and L0 .
As a final remark to be made in this Appendix A, let us call the physicist-
oriented reader’s attention to the following (somewhat formal) abstract
Lemmas, alternatives to the Banach space methods used in Ref. 9.
(U, π, h); (U , π , h )H̃ = (D∗ DU, U )H1 + (π, π)H1 + (h, h)H2 , (A.13)
if we have the constraints below for the operators on the Hilbert component
spaces:
C = C∗ on H2
∗ ∗
A D D=B on H1 (A.14)
Range(A) ⊂ Dom(D).
and
(U, π, h); L0 (U , π , h )H̃ = (DU, DAπ )H1 + (π, BU )H1 + (h, Ch )H2 .
(A.16)
August 6, 2008 15:46 9in x 6in B-640 ch03
On the basis of Eqs. (A.15) and (A.16) one can see that L0 is a closed
symmetric operator. Besides, if there is a (U0 , π0 , h0 ) ∈ H̃ such that for
every (U, π, h) we have the orthogonality relation
then
As a result
V1∗ = V2 . (A.23)
Proof. One can use similar arguments of the proof of Lemma A.1 to arrive
at such a result.
Let us now take into account the case of the spectrum of L0 contained
on the negative real line (−∞, 0] and that one associated with V is by
its turn, contained only on the upper-bounded negative infinite interval
(−∞, −c] (with c > 0); then we have by a straightforward application of the
August 6, 2008 15:46 9in x 6in B-640 ch03
One can see that it satisfies the conditions of the above lemmas with the
operator D = ∇ and with the Hilbert space
and c = inf spec(−∆) and H 2 (Ω) ∩ H01 (Ω) (see Sec. 3.5), and thus, leading
to the expected total energy exponential decay showed in Sec. 3.5.
I : Ω → R∞
(B.1)
ω → {Xα (ω)}α∈Λ .
the compactification) is the space of all sequences with only a finite number
of nonzero entries as it is necessary to consider in the case of the famous
extension theorem of Kolmogorov. It turns out that one can consider the
support of the probability measure as the set of all linear functionals on
E, the so-called algebraic dual of E. This result is a direct consequence
of considering the set of random variables given by the family {e∗α }α∈Λ .
For applications one is naturally lead to consider the probabilistic object
called characteristic functional associated with the measure dν (∞) (f ) when
F (Λ, R) is identified with the vector space of all algebraic linear functionals
E alg of a given vector space E as pointed out above, namely,
Z[j] = dν (∞) (f ) exp(if (j)), (B.3)
E alg
0
where j are elements of E written as j = α∈Λfin xα eα with {eα }α∈Λ
denoting a given Hammel (vectorial) basis of E.
One can show that if there is a subspace H of E with a norm || ||H
coming from an inner product (, )H such that
dν (∞) (f ) · ||f ||2H < ∞ (B.4)
E∩H
one can show the support of the measure of exactly the Hilbert space
(H, (, )H ) — the famous Minlos theorem.
Another important theorem (the famous Wienner theorem) is that when
we have the following condition for the family of random variables with the
index set Λ being a real interval of the form Λ = [a, b]:
dν (∞) (f )|Xt+h (f ) − Xt (f )|p ≤ c|h|1+ε , (B.5)
E
Chapter 4
4.1. Introduction
84
August 6, 2008 15:46 9in x 6in B-640 ch04
∂U (x, t)
= (−AU )(x, t) + ∆(F (U (x, t))) + f (x, t) (4.1)
∂t
Proof. The existence proof will be given for a general A as stated below
Eq. (4.3). Let {ϕi (x)} be the spectral eigenfunctions associated with the
operator A. Note that each ϕi (x) ∈ H 2 (Ω) ∩ H01 (Ω) and this set is complete
on L2 (Ω) as a result of the Gelfand generalized spectral theorem applied
for A1 since H 2 (Ω) ∩ H01 (Ω) is compactly immersed in L2 (Ω).
Consider the following system of nonlinear ordinary differential equa-
tions associated with Eqs. (4.1)–(4.3) — the well-known Galerkin system:1
∂U (n) (x, t)
, ϕj (x) + (AU (n) (x, t), ϕj (x))L2 (Ω)
∂t 2
L (Ω)
n
U (n) (x, 0) = (g, ϕi )L2 (Ω) · ϕi (x), (4.5)
i=1
n
(n)
U (n) (x, t) = Ui (t)ϕi (x), (4.6)
i=1
n
f (n) (x, t) = (f (x, t), ϕi (x))L2 (Ω) ϕi (x), (4.7)
i=1
and (, )L2 (Ω) is the usual inner product on L2 (Ω). Note that
n
U (n) (x, t)∂Ω = Ui (t)(ϕi (x))∂Ω = 0.
(n)
(4.8)
i=1
1 d
U (n) 2L2 (Ω) + (AU (n) , U (n) )L2 (Ω)
2 dt
(n)
+ dxF (U (n) )(∇U (n) ∇U ) = (f, U (n) )L2 (Ω) . (4.12)
Ω
(AU (n) , U (n) )L2 (Ω) ≥ γ(Ω)U (n) 2L2 (Ω) . (4.13)
August 6, 2008 15:46 9in x 6in B-640 ch04
1 d
(U (n) 2L2 (Ω) ) + γ(Ω)U (n) 2L2 (Ω) + (F (U (n) ))(1/2) ∇U (n) 2L2 (Ω)
2 dt
1 2 1 (n) 2
≤ pf (x, t)L2 (Ω) + U L2 (Ω) . (4.14)
2 p
At this point we observe that the square root of the function F (x)
makes sense since it is always positive: F (x) > 0.
1
By choosing Eq. (4.14) p big enough such that ap ≡ γ(Ω) − 2p > 0, we
have that
1 d 1
(U (n) 2L2 (Ω) ) + ap U (n) 2L2 (Ω) ≤ pf 2L2(Ω) (t), (4.15)
2 dt 2
or by the Gronwall lemma, that there is a uniform constant M [even for
T = +∞ if f ∈ L2 ([0, ∞], L2 (Ω))] such that
(4.16)
with
E{gi gk } = Kik = dxdyK(x, y)ϕi (x)ϕk (y). (4.24)
Ω×Ω
U (x, t, [g])
n1
,n2 ,n3
iπ jπ kπ
= n lim di,k,j (t, [g ]) sin x sin y sin z ,
1 →∞
n2 →∞
L L L
i,j,k=1
n3 →∞
(4.25)
where the set of absolutely continuous function {di,j,k (t, [g0 ])} on [0, T ] sat-
isfies the Galerkin set of ordinary differential equations with “random”
Gaussian initial conditions
iπ jπ kπ
di,j,k (0, [g ]) = g(x), sin x sin y sin z = g .
L L L L2 (Ω)
(4.26)
Let us now comment on the uniqueness of solution problem for the
nonlinear initial-value diffusion equations (4.1)–(4.3).
In the case under study, the uniqueness comes from the following tech-
nical lemma.
Lemma 4.1. If Ū(1) and Ū(2) in L∞ ([0, T ] × L2 (Ω)) are two functions
satisfying the weak relationship, for any v ∈ C0∞ ([0, T ], H 2 (Ω) ∩ H01 (Ω))
T
∂v
dt Ū(1) − Ū(2) , − + (Ū(1) − Ū(2) , Av)L2 (Ω)
0 ∂t L2 (Ω)
− (F (Ū(1) ) − F (Ū(2) ); ∆v)L2 (Ω) ≡ 0, (4.27a)
The proof of Eq. (4.27) is a direct consequence of the fact that F (x) is
a Lipschitz function satisfying an inequality of the form
|F (Ū(1) ) − F (Ū(2) )|(x, t) ≤ sup F (x) |Ū(1) − Ū(2) |(x, t) (4.27c)
−∞<x<+∞
(for a technical proof see Ref. 5). However in the case of A = −∆ (minus
Laplacian), as stated in our theorem, the identity Eq. (4.27a) means that for
the functions of the form v(x, t) = e−λt vλ (x) ∈ C0∞ ([0, T ], H 2 (Ω) ∩ H01 (Ω))
with ∆vλ (x) = −λvλ (x), the following identity must hold true:
T
dt(F (Ū(1) ) − F (Ū(2) ), vλ )L2 (Ω) ≡ 0, (4.28)
0
which means that F (Ū(1) ) = F (Ū(2) ) a.e., and thus, Ū(1) = Ū(2) a.e., since
F (x) satisfies the lower-bound estimate
|F (x) − F (y)| ≥ inf |F (x)| |x − y|.
−∞<x<∞
where V is the velocity of gas, ρ is the density of the gas, and P is the
pressure, together with Darcy’s law and the isothermic equation of state,
namely,
P = cργ (4.30)
= −k∇P
V
Since it is showed that for finite volume open, convex with C ∞ -boundary
domain Ω as in our study, the function ρ(x, t) which satisfies Eq. (4.31)
August 6, 2008 15:46 9in x 6in B-640 ch04
+ (F (U̇ (n) ))(1/2) ∇U̇ (n) 2L2 (Ω) ≤ ν(AU (n) , U (n) )
1 1
+ pf 2L2(Ω) + U̇ (n) 2L2 (Ω) . (4.37)
2 p
1
If one chooses here the integer p such that 2p = ν, we have the simple
bound
1 d (n) 2 1
U̇ L2 (Ω) + (AU (n) , U (n) )L2 (Ω) ≤ f 2L2(Ω) . (4.38)
2 dt 4ν
As a consequence of Eq. (4.38), there is a constant M such that the
uniform bounds hold true (even if for the case T = +∞ for the case of
f ∈ L2 ([0, ∞), L2 (Ω))):
sup ess0≤t≤T U̇ (n) 2L2 (Ω) ≤ M, (4.39)
on L∞ ([0, T ] × L2 (Ω)):
weak-star lim (U (n) (x, t)) = Ū (x, t), (4.41)
n→∞
∂ (n)
weak-star lim U (x, t) = P̄ (x, t). (4.42)
n→∞ ∂t
We thus have that the relationship below hold true for any test function
v(x, t) ∈ Cc∞ ([0, T ] × H 2 (Ω) ∩ H01 (Ω)) obviously satisfying the relations
v(x, 0) = v(x, T ) = ∆v(x, 0) = ∆v(x, T ) = vt (x, 0) = vt (x, T ) = vtt (x, 0) =
vtt (x, T ) ≡ 0 as a consequence of applying the Aubin–Lion theorem in a
similar way as it was used on Eq. (4.17):
T
d2 v
dt Ū , 2 + (Ū , Av)L2 (Ω)
0 d t L2 (Ω)
dv
+ ν Ū , − − (F (P̄ ), ∆v)L2 (Ω)
dt L2 (Ω)
T
= dt(f, v)L2 (Ω) , (4.43)
0
(AU (n) , U (n) )L2 (Ω) (t) ≥ γ(Ω)U (n) 2L2 (Ω) (t), (4.48)
one can see straightforwardly from Eq. (4.47) by choosing 2p > ν1 and the
previous bounds Eq. (4.39) that there is a positive constant B such that
T
dsU̇ (n) (s)2L2 (Ω) ≤ B = M T, (4.49)
0
Let us start the physicist-oriented section of our note by writing the abstract
scalar parabolic equations (4.1)–(4.3) in the integral (weak) form
t
U (t) = e−At U (0) + ds e−(t−s)A ∆F (U (s)), (4.50)
0
where
dxK(x, x) < ∞. (4.52)
Ω
August 6, 2008 15:46 9in x 6in B-640 ch04
t
× δ (F ) U (t) − e−At U0 + ds e−(t−s)A ∆F (U (s))
0
∞
× exp i dt j(t), U (t) L2 (Ω) , (4.54)
0
where formally
1 −1
dµ[U0 ] = (dU0 (x)) × exp − dxdyU0 (x)K (x, y)U0 (y) .
2 Ω
x∈Ω
(4.55)
By using the well-known Fourier functional integral representation for
the delta-functional inside the identity equation (4.54)6 (see this old idea
in A. S. Monin and A. M. Yaglom, Statistical Fluid Mechanics, Vol. 2 (MIT
Press, Cambridge, 1971),6 one can rewrite Eq. (4.54) in the following form:
1
Z[j(t)] = (dU (x, t)) × dλ(t)
Z(0) L∞ ([0,T ],L2 (Ω)) L2 ([0,T ],L2 (Ω))
! t "
t
−(t−s)A
× exp i λ(t), U (t) + ds e ∆F (U )(s)
0 0 L2 (Ω)
1 t # $
× exp − ds λ(s)((e−As K −1 e+As )λ)(s) L2 (Ω) . (4.56)
2 0
August 6, 2008 15:46 9in x 6in B-640 ch04
T
1 T
× exp − ds ds
2 0 0
!
∂U
× + AU − ∆F (U ) (s);
∂s
"
∂U
× + AU − ∆F (U ) (s )
∂s L2 (Ω)
T
× exp i ds j(s), U (s) L2 (Ω) . (4.57)
0
(4.66)
August 6, 2008 15:46 9in x 6in B-640 ch04
with K(x, y) defining an integral operator of the trace class on L2 (RD ), one
can easily make a further simplification by integrating out exactly the V (x)-
Gaussian functional integral and producing the effective quartic interaction
term as written below:
0 V λ
dµ[V ] exp i [λ, U ],
L2 (RD ) V 0 U L2 (RD )
∞ ∞
1
= exp − dx dt dy ds
2 RD 0 RD 0
× (λ(x, t)U (x, t))K(x, y)(λ(y, s)U (y, s)) . (4.67)
Let us call the reader’s attention that at this point one can straightfor-
wardly implement the usual Feynman–Wild–Martin–Siggia–Rosen diagra-
matics with the free propagator given explicitly in the momentum space
by6
1
[∂t − D0 (−∆)α ]−1 ↔ . (4.68)
iw − D0 (|k|2α )
It is important to remark that all the above analyses are still somewhat
formal at this point of our study since it is based strongly on the hypothesis
that the operator A as given by Eq. (4.64) is a C0 -generator of a contractive
semigroup on L2 (RD ), which straightforwardly leads to the existence and
uniqueness of global solution for Eq. (4.59). Let us give a rigorous proof of
ours of such a self-adjointness result, which by its turn will provide a strong
connection between the parameter α of the Laplacean power, related to the
anomalous diffusion coefficient, and the underline dimension D of the space
where the anomalous diffusion is taking place.
Let us first recall the famous Kato–Rellich theorem on self-adjointness
perturbative of self-adjoint operators.
In order to apply this powerful theorem to our general case under study,
let us recall that the fractional power of the Laplacean operator, (−∆)α ,
has as an operator domain the (Sobolev) space of all square-integrable
functions ϕ(x) such that its Fourier transform ϕ̃(k) satisfies the condition
|k|α ϕ̃(k) ∈ L2 (RD ) and as a core domain the Schwartz space S(RD ).8
As a consequence, we have the straightforward estimate for ϕ ∈ S(RD )
as
1
ϕ̃L2 (RD ) ≤ (k 2α + 1)ϕ̃L2 (RD )
1 + k 2α L2 (RD )
We thus have
ϕ̃r L1 (RD ) = dkrβ ϕ̃(rk)
RD
= rβ−D ϕ̃L1 (RD ) (4.73)
and
together with
or equivalently
/ D
ϕ̃L1 (RD ) ≤ C(α, D) r 2 −2α K 2α ϕ̃L2 (RD )
D
0
+ r 2 ϕ̃L2 (RD ) . (4.77)
in order to obtain the validity of the bound a < 1 on Eq. (4.69) and, thus,
concluding that (−∆)α +V is an essentially self-adjoint operator on S(RD ).
So, its closure on L2 (RD ) produces the operator used in the above-exposed
semigroup C0 -construction for V (x) being a positive function almost every-
where. This result of ours is a substantial generalization of Theorem X.15
presented in Ref. 8. Note that the V (x) sample positivity is always the
physical case of random porosity in a Porous medium. In such a case, the
random potential is of the exponential form V (x) = V0 (exp{ga(x)}), where
g is a positive small parameter and V0 is a background porosity term. The
effective cylindrical measure on the random porosity parameters is written
as (see Eq. (4.66))
∼ 1 ga(x) −1
dµ[V ] = D [a(x)] exp −
F
dx dye K (x, y)e ga(y)
.
g1 2 RD RD
(4.80)
References
Appendix A
1 1
T 1 (Λ) dUn 1
+ 1
dt 1∆ F (Un (t)) · 1
0 dt 1 L2 (Ω)
T
2
≤0+0+a ∆F (Un (t))L2 (Ω) dt
0
1 1
1 T 1 dUn 12
+ dt 1 1
1 dt 1 2 . (A.5)
a 0 L (Ω)
In other words,
T 1 1
1 1 dUn 12
1− dt 1 1
1 dt 1 2
a 0 L (Ω)
T
≤ caΛ2 · Un (t)2L2 (Ω)
0
2
≤ CaΛ T M, (A.6)
where c = supa≤x≤b |F (x)|, and M is the bound given by Eq. (4.16).
August 6, 2008 15:46 9in x 6in B-640 ch04
Appendix B
or equivalently
ε ε
1 1
dt Um (x, t), ϕi (x) L2 (Ω) → dt U (x, t), ϕi (x) L2 (Ω) . (B.2)
ε 0 ε 0
August 6, 2008 15:46 9in x 6in B-640 ch04
result, which by its turn, means that the function U (x, t) obtained by means
of our compacity technique satisfies the initial-condition Eq. (4.2) in the
L2 (Ω)-mean sense.
Appendix C
∂U (x, t)
= P̄ (x, t) a.e. in ([0, T ] × Ω. (C.4)
∂t
I : Ω → R∞
(D.1)
ω → {Xα (ω)}α∈Λ .
August 6, 2008 15:46 9in x 6in B-640 ch04
one can show that the support of the measure of exactly the Hilbert space
(H, (, )H ) is the famous Minlos theorem.
Another important theorem (the famous Wiener theorem) is when we
have the following condition for the family of random variables with the
index set Λ being a real interval of the form Λ = [a, b]
dν (∞) (f )|Xt+h (f ) − Xt (f )|p ≤ c|h|1+ε (D.5)
E
Chapter 5
5.1. Introduction
where φ(m) (x, it) is the supposed self-adjoint Minkowski quantum field ana-
lytically continued to imaginary time and j(x) = j(x, it) is a set of func-
tions belonging to a given topological vector space of functions denoted
by E which topology is not specified yet and will be called the Schwinger
classical field source space. It is important to remark that {φm (x, it)} is a
commuting algebra of self-adjoints operators as Symanzik has pointed out.7
In order to write Eq. (5.1) as an integral over the space E alg of all linear
functionals on the Schwinger source space E (the so-called algebraic dual
of E), we take the following procedure, different from the usual abstract
approach (as given — for instance — in the proof of IV-11-2), by making
the hypothesis that the restriction of the Schwinger generating functional
equation (5.1) to any finite-dimensional RN of E is the Fourier transform
of a positive continuous function, namely
N
N
Z Cα j̄α (x) = exp i Cα Pα g̃(P1 , . . . , PN )dP1 , . . . , dPN .
α=1 RN α=1
(5.2)
Eq. (5.2) and h(j(x)) is the action of the given linear (algebric) functional
(belonging to E alg ) on the element j(x) ∈ E.
At this point, we relate the mathematically non-rigorous physicist point
of view to the Kolmogorov measure dµ(h) (Eq. (5.3)) over the algebraic
linear functions on the Schwinger source space. It is formally given by the
famous Feynman formulae when one identifies the action of h on E by
means of an “integral” average
h(j) = dxD j(x)h(x). (5.4)
RD
dµ(h) = dh(x) exp{−S(h(x))}, (5.5)
x∈RD
where S is the classical action of the classical field theory under quanti-
zation, but with the necessary coupling constant renormalizations needed
to make the associated quantum field theory well-defined.
Let us outline these proposed steps on a λφ4 -field theory on R4 .
At first we will introduce the massive free field theory generating func-
tional directly into the infinite volume space R4 .
1
Z[j(x)] = exp − d4 xd4 x j(x)((−∆)α + m2 )−1 (x, x )j(x ) , (5.6)
2
eik(x−x )
((−∆)α + m2 )−1 (x, x ) = d4 k , (5.7)
k 2α + m2
LΛf g(Pλs1 , . . . , PλsN ) = Q g(Pλs1 , . . . , PλsN )
Rλ
λ∈Λf
1
× exp − (λ2α + m2 )(Pλ )2
2
λ∈Λf
× d(Pλ π(λ2α + m2 )) . (5.8)
λ∈Λf
1 (jλ )2
= exp − . (5.10)
2 λ2α + m2
λ∈Λf
1 (α) 1 (α)
V (α) (φ) = λR φ4 + (Zφ (λR , M ) − 1)φ((−∆)α )φ − [(m2 Zφ (λR , m) − 1
2 2
(α)
− (δm2 )(α) (λR )]φ2 − [Zφ (λR , m)(δ (α) λ)(λR , m)]φ4 . (5.11)
Here, the renormalization constants are given in the usual analytical finite-
part regularization form for a λφ4 -field theory. It is still an open problem
in the mathematical-physics of quantum fields to prove the integrability
in some distributional space of the cut-off removing α → 1 limit of the
interaction lagrangean exp(−V (α) (φ)) (see Sec. 5.4 for the analysis of this
cut-off removing on space functions).
Let us now analyze the measure support of quantum field theories gener-
ating functional equation (5.3).
For higher dimensional space-time, the only available result in this
direction is the case that we have a Hilbert structure on E.4−6
At this point, we introduce some definitions. Let ϕ : Z+ → R be an
increasing fixed function (including the case ϕ(∞) = ∞). Let E be denoted
by H and H Z be the subspace of H alg = (Πλ∈A=[0,1] Rλ ) (with A being the
index set of a Hamel basis of H), formed by all sequences {xλ }λ∈A ∈ H alg
with coordinates different from zero at most a countable number
and
1
N
lim (xλσ(µ) )2 <∞
N →∞ ϕ(N ) n=1
Here , (0) is the inner product of H and {ep }1≤p≤N is a vectorial basis of
the subspace H N with dimension N .
Suppose that Z[j(x)] is a continuous function on D(T ) = (D(T ), , (1) ),
where , (1) is a new inner product defined by the operator T (j, j̄(1) =
T j, T j̄(0) ). We have, thus, that the support of the cylindrical measure
equation (5.3) is the measurable set HeZ .
Proof. Following closely Refs. 1, Theorems 2.2 and 4, let us consider the
following representation for the characteristic function of the measurable
set HeZ ⊂ H alg
·XHeZ ({xλ }λ∈A )
1α 2
N
1 2
N
= lim lim exp − xλl = 1 if lim xλl < ∞
α→0 N →∞ 2ϕ(N ) N →∞ ϕ(N )
=1 =1
0 otherwise
(5.14)
But
α 2
N
µ(HeZ ) = lim lim dµ(h) exp − xλ
α→0 N →∞ H alg 2ϕ(N )
=1
1
= lim lim 2πα N/2 dj1 , . . . , djN
α→0 N →∞ ( ϕ(N )) RN
N
1 ϕ(N )
× exp − j2 Z̃(j1 , . . . , JN ), (5.16)
2 α
=1
August 6, 2008 15:46 9in x 6in B-640 ch05
where
N
Z̃(j1 , . . . , jN ) = dµ({xλ }) exp i xλ j
πRλ
λ∈A =1
N
= dµ(h) exp i xλ j . (5.17)
H alg =1
Now due to the continuity and positivity of Z[j] in D(T ); we have that
for any ε > 0 → ∃δ such that the inequality below is true since we have
that: Z(j1 , . . . , jN ) ≥ 1 − ε − δ22 (j, j)(1)
1 ϕ(N ) 2
N
1
N/2 dj1 · · · djN exp − j Z̃(j1 , . . . , jN )
2πα RN 2 α
=1
ϕ(N )
2
N
1
≥1−ε− 2 dj1 · · · djN
δ 2πα N/2 RN
(m,n)=1 ϕ(N )
N
1 ϕ(N ) 2 (1)
× exp − j jm jn em , en
2 α
=1
N
2 α
=1−ε− 2 δmn T en , T em (0)
δ ϕ(N )
(m,n)=1
2 α 2
≥1−ε− ϕ(N ) ≥ 1 − ε − α. (5.18)
δ2 ϕ(N ) δ2
2
1 ≥ µ(HeZ ) ≥ 1 − ε − lim α = 1 − ε. (5.19)
δ 2 α→0
where
N
2
= (xn )n∈Z , xn ∈ R (xn )2 nm < ∞ (5.21)
m
n=0
We, thus, consider E = S(RD ) in Eq. (9.3) and Z[j(x)] = Z[{jn }n∈Z ] is
"∞ "∞
continuous on n n . Since Z[{jn }n∈Z ] ∈ C( n=0 2n , R) we have that for
any fixed integer p, Z[{jn }n∈Z ] is continuous on the Hilbert space 2p which,
by its turn, may be considered as the domain of the following operator:
2
Tp : pc 0 → 0
for some positive integer B and {ei } being the canonical orthonormal basis
of l02 . By an application of our theorem for each fixed p; we get that the
support of measure is given by the union of weighted spaces
∞
! ∞
!
supp dµ(h) = ( 2p )∗ = 2
−p = S (RD ). (5.25)
p=0 p=0
At this point we can suggest, without a proof of straightforward (non-
topological) generalization of the Minlos theorem.
August 6, 2008 15:46 9in x 6in B-640 ch05
Theorem 5.2. Let {Tβ }β∈C be a family of operators satisfying the hypothe-
#
sis of Theorem 5.3. Let us consider the Locally Convex space β∈C Dom(Tβ )
(supposed non-empty) with the family of norms ψ β = Tβ ψ, Tβ ψ1/2 .
If the Functional Fourier transform is continuous on this locally convex
space, the support of the Kolmogoroff measure equation (5.3) is given by the
#
following subset of [ β∈C Dom(Tβ )]alg , namely
!
supp dµ(h) = Hϕ2 β , (5.26)
β∈C
where ϕβ are the functions given by Theorem 5.1. This general theorem will
not be applied in what follows.
and j(x) is, thus, the Schwinger source for the φ(x) field taken as a basic
dynamical variable.9
Since Eq. (5.27) is continuous in L2 (R2 ) with the inner product defined
2
by the trace class operator ((−∆)2 + eπ (−∆))−1 , we conclude on the basis
of Theorem 5.1 that the associated Kolmogoroff measure in Eq. (5.3) has
its support in L2 (R2 ) with the usual inner product. As a consequence,
the quantum observable algebra will be given by the functional space
L1 (L2 (R2 ), dµ(h)) and usual orthonormal finite-dimensional approxima-
tions in Hilbert spaces may be used safely, i.e. if one considers the basis
$∞
expansion h(x) = n=1 hn en (x) with en (x) denoting the eigenfunctions of
the operator in Eq. (10.27) we get the result
∞
!
L1 (RN , dµ(h1 , . . . , hN )) = L1 (L2 (R2 ), dµ(h)). (5.29)
n=1
August 6, 2008 15:46 9in x 6in B-640 ch05
the associated measure support will now be the Schwartz space S (R2 )
2
since the operator (−∆ + eπ )−1 is an application of S(R2 ) to S (R2 ). As a
consequence, it will be very cumbersome to use Hilbert finite-dimensional
approximations8 as in Eq. (5.29).
An alternative to approximate tempered distributions is the use of its
Hermite expansion in S (R) distributional space associated with the eigen-
#
functions of the Harmonic-oscillator V (x) ∈ L∞ (R) L2 (R) potential per-
tubation (see Ref. 3 for details with V (x) ≡ 0).
d2
− 2 + x2 + V (x) Hn (x) = λn Hn (x). (5.31)
dx
If A is an operator of trace class on (L2 (Ω), dν) we have, thus, the validity
of the usual eigenvalue functional representation
Z[{jn }µ∈Z ]
∞ ∞
∞
1
= d(c λ ) exp − λ c2 X2 ({cn }n∈Z ) exp i c j
2
=1 =1 =1
A−1 σ = λ σ ,
(5.33)
j = j, σ
August 6, 2008 15:46 9in x 6in B-640 ch05
It is instructive to point out the usual Hermite functional basis (see 5.4,
Ref. 5) as a complete set in L2 (E alg , dµ(h)), only if the Gaussian Kol-
mogoroff measure dµ(h) is of the class above studied.
A criticism to the usual framework to construct Euclidean field theories
is that it is very cumbersome to analyze the infinite volume limit from the
Schwinger generating functional defined originally on compact space times.
In two dimensions the use of the result that the massive scalar field theory
generating functional
1
exp − d2 x d2 yj(x)(−∆ + m2 )−1 (x, y)j(y) (5.35)
2 R2 R2
with j(x) ∈ S(R2 ) is given by the limit of finite volume Dirichlet field
theories
L
1 L 0 T 1
lim exp − dx dx dy 0
L→∞ 2 −L −T −L
T →∞
T
1 0 1 2 −1 1 1 0 0 0 1
× dy j(x , x )(−∆D + m ) (x , y , x , y )j(y , y )
−T
(5.36)
may be considered, in our opinion, as the similar claim made that is possible
from a mathematical point of view to deduce the Fourier transforms from
Fourier series, a very, difficult mathematical task (see Appendix B).
Let us comment on the functional integral associated with Feynman
propagation of fields configurations used in geometrodynamical theories in
the scalar case
where
Let us call to attention that there is still a formal definition of the above
Feynman path propagator for fields equation (5.37) which at large time
T → +∞ gives formally the quantum field functional integral equation
(5.5) associated with the Schwinger generating functional.
We thus consider the functional domain for Eq. (5.37) as composed of
field configurations which has a classical piece added with another fluctu-
ating component to be functionally integrated out, namely
σ(x, t) = σCL (x, t) + σq (x, t). (5.41)
Here, the classical field configuration problem (added with all zero modes
of the free theory) defined by the kinetic term £
d2
− 2 + £ σ CL (x, t) = j(x, t), (5.42)
dt
with
σ CL (x, −T ) = β1 (x); σ CL (x, T ) = β2 (x), (5.43)
namely,
−1
d2
σCL (x, t) = − 2 + £ j(x, t) + (all projection on zero modes of £).
dt
(5.44)
As a consequence of the decomposition equation (5.43), the formal geomet-
rical propagator with an external source is given by
G[β1 (x), β2 (x), T, [j]] = D[σ(x, t)]
σ(x,−T )=β1 (x)
σ(x,+T )=β2 (x)
1 T
d2
× exp − dtd xσ(x, t) − 2 + £ σ(x, t)
ν
2 −T dt
T
× exp i dt dν xj(x, t)σ(x, t) (5.45)
−T
∂
G[β1 (x), β2 (x), T, [j]]
∂T
( )
δ2
= dν x + 2 − |∇β2 (x)|2 + j(x, T ) G[β1 (x), β2 (x), T, [j]]
Ω δ β2 (x)
(5.48)
(0) 1
Zα,ε IR
[j] = exp − j, (−∆)−α ε jL2 (R2 )
2
= d(0)
α,ε µ[ϕ] exp(ij, ϕL2 (R2 ) ). (5.50b)
(0)
is an integrable function on L1 (L2 (R2 ), dα,εIR µ[ϕ]) and leads to a well-
defined ultraviolet path-integral in the limit of α → 1.
The proof is based on the following estimates:
Since almost everywhere we have the pointwise limit
exp −gbare (α) d2 xV (ϕ(x))
N
(−1)n (gbare (α))n
× lim dk1 · · · dkn Ṽ (k1 ) · · · Ṽ (kn )
N →∞ n! R
n=0
× dx1 · · · dxn eik1 ϕ(x1 ) · · · eikn ϕ(xn ) (5.53)
R2
with
ZεαIR [gbare ] = d(0)
α,εIR µ[ϕ] exp −gbare (α) d2 xV (ϕ(x)) (5.54b)
(0)
we have, thus, the more suitable form after realizing d2 ki and dα,εIR µ[ϕ]
integrals, respectively
(gbare (α))n n
∞
|ZεαIR =0 [gbare ]| ≤ ||Ṽ ||L∞ (R)
n=0
n!
× dx1 · · · dxn det− 2 [Gα (xi , xj )] 1≤i≤N .
1
(N )
(5.55)
1≤j≤N
August 6, 2008 15:46 9in x 6in B-640 ch05
(N )
Here, [Gα (xi , xj )] 1≤i≤N denotes the N × N symmetric matrix with (i, j)
1≤j≤N
entry given by the Green-function of the α-Laplacean (without the infrared
cut-off here! and the needed normalization factors!).
Γ(1 − α)
Gα (xi , xj ) = |xi − xj |2(1−α) . (5.56)
Γ(α)
At this point, we call the readers’ attention that we have the formulae on
the asymptotic behavior for α → 1.
1
(N − 1)(−1)N − 2
lim det − 12 (N )
[Gα (xi , xj )] ∼ (1 − α) N/2
× .
α→1 π N/2
α>1
(5.57)
After substituting Eq. (5.57) into Eq. (5.55) and taking into account the
hypothesis of the compact support of the nonlinear V (x) (for instance:
supp V (x) ⊂ [0, 1]), one obtains the finite bound for any value grem > 0,
and producing a proof for the convergence of the perturbative expansion in
terms of the renormalized coupling constant.
∞
n n
( Ṽ L∞ (R) ) gren (1n )
lim |ZεαIR =0 [gbare (α)]| ≤ 1 √ (1 − α)n/2
α→1
n=0
n! (1 − α) 2 n
L = (−∆)α + m2 (5.59)
In this case, let us consider the double functional integral with functional
domain L2 (Rν )
(0) (0)
Z[j, k] = dG β[v(x)] d(−∆)α +v+m2 µ[ϕ]
× exp i dν x(j(x)ϕ(x) + k(x)v(x)) , (5.62)
Note that if one introduces from the begining a bare mass parameters m2bare
depending on the parameters α, but such that it always satisfies Eq. (5.60)
one should obtain again Eq. (5.64) as a well-defined measure on L2 (Rν ).
Of course, that the usual pure Laplacean limit of α → 1 in Eq. (5.59), will
need a renormalization of this mass parameters (limα→1 m2bare (α) = +∞!)
as much as its done in the previous example.
Let us continue our examples by showing again the usefulness of
the precise determination of the functional-distributional structure of the
domain of the functional integrals in order to construct rigorously these
path-integrals without complicated limit procedures.
August 6, 2008 15:46 9in x 6in B-640 ch05
Here, ϕ ∈ W2n (Ω) is the Sobolev space of order n which is the functional
domain of the cylindrical Fourier transform measure of the generating func-
tional Z (0) [j], a continuous bilinear positive form on W2−n (Ω) (the topo-
logical dual of W2n (Ω)).
By a straightforward application of the well-known Sobolev immersion
theorem, we have that for the case of
ν
n−k > (5.73)
2
including k a real number of the functional Sobolev space W2n (Ω) is con-
tained in the continuously fractional differentiable space of functions C k (Ω).
As a consequence, the domain of the Bosonic functional integral can be
further reduced to C k (Ω) in the situation of Eq. (5.73)
(0)
Z (0) [j] = d(2n) µ[ϕ] exp(ij, ϕL2 (Ω) ). (5.74)
C k (Ω)
That is our new result generalizing the Wiener theorem on Brownian paths
in the case of n = 1 , k = 12 , and ν = 1.
August 6, 2008 15:46 9in x 6in B-640 ch05
with the non-linearity F (x) denoting a lower bounded real function (γ > 0)
F (x) ≥ −γ (5.76)
1
Z (0) [j] = exp − j, exp{−α∆}jL2 (Ω) (5.78)
2
since any given class trace operator can always be considered as the
composition of two Hilbert–Schmidt, each one defined by a function on
L2 (M × M, dν ⊗ dν).
Here, the cylindrical measure dA µ(v), defined already on the vector
space of the linear forms of E, with the topology of pontual convergence —
the so-called algebraic dual of E — has its support concentrated on the
Hilbert spaces H, through the isomorphism of H and its dual H by means
of the Riesz theorem.
This result can be understood more easily, if one represents the given
Hilbert space H as a square-integrable space of measurable functions on a
complete measure space (M, dν)L2 (M, dν). In this case, the class trace pos-
itive definite operator is represented by an integral operator with a positive-
definite Kenel K(x, y). Note that it is worth to re-write Eq. (5.79) in the
Feynman path-integral notation as
1
exp − dν(x)dν(y)f (x)K(x, s)f (s)
2 M×M
1
= dϕ(x)
Z(0) L2 (M,dν)
x∈M
& '
1
× exp − dν(x)dν(y)ϕ(x)K −1 (x, y)ϕ(y)
2 M×M
× exp dν(x)f (x)ϕ(x) . (5.80)
M
August 6, 2008 15:46 9in x 6in B-640 ch05
1
1 |ϕ|ϕn |2
N N
1
2
N
× exp − exp i ϕn v̄n .
2 n=0 λn 2πλn
n=0 n=0
(5.84)
The above cited theorem for the support characterization of Gaussian
path-integrals can be generalized to the highly non-trivial case of a non-
linear functional Z(v) on E, satisfying the following conditions:
(a) Z(0) = 1,
N
(b) Z(vj − vk )zj z̄k ≥ 0, for any {zi }1≤i≤N : zi ∈ N
j,k
if
/
lim vn , Avn = 0 A ∈ (H) (5.87)
n→∞ 1
where
dµ(ϕ) = 1. (5.89)
H
1
Z(j) = exp − j, A−1 j (5.90)
2
with σ(A) denoting the spectrum of A (a subset of R+ !) and dE(λ) are the
spectral projections associated with the spectral representation of A.
In this case one has the result for the path-integral weight as
1 1
exp − λϕ, dE(λ)ϕ = exp − ϕ, Aϕ
2 σ(A) 2
1
= lim sup exp − ϕ, λdE(λ)ϕ . (5.92)
2
λ∈σFin (A)
Here σFin (A) denotes all subsets with a finite number of elements of σ(A).
August 6, 2008 15:46 9in x 6in B-640 ch05
1 2
1 1/ε
dt −tλ
= lim+ exp + e
ε→0 2 ε t
λ∈σFin (A)
1/ε
1 dt
= lim exp + T r(e−tA ) . (5.96)
ε→0+ 2 ε t
In the case of the finitude of the right-hand side of Eq. (5.94) (which
means that e−tA is a class trace operator and the finitude up the proper-
time parameter t), one can proceed as in the appendix to define mathemat-
ically the Gaussian path-integral.
Let us now consider the proper-time (Cauchy principal value sense)
integration process as indicated by Eq. (5.95), for the case of N an even
space-time dimensionality
1/ε 2 1/ε 2
dt e−tm e−tm
I(m2 , ε) = = dt N +2 (5.97)
ε t tN/2 ε t 2
Theorem 5.3. Let (H, , 1 ) be a separable Hilbert space with the inner
product , 1 . Let (H0 , , 0 ) be a subspace of H, so there is a trace class
operator T : H → H, such that the inner product , 0 is given explicitly
by g, h1 = g, T h0 = T 1/2 g, T 1/2 h0 . Let us, thus, consider a positive
1
definite functional Z(j) ∈ C((H, , 1 ), R) [if jn −→ j, then Z(jn ) → Z(j)
on R+ ]. We obtain that the Bochner path-integral representation of Z(j)
is given by a measure supported at these linear functionals, such that their
restrictions in the subspace H0 are continuous by the norm induced by the
“trace-class” inner product , 1 .
∞
2p
= 0, with the inner product (xn ), (yn )sp = n xn ȳn . (5.99)
n=1
and
∞
!
N
S (R ) = s−p . (5.101)
p≥1
and
∞
1 π2
= . (5.103)
n=1
n2 6
As a consequence
π2
||(xn )||sp ≤ ||(xn )||sp+1 (5.104)
6
If one has an arbitrary continuous positive-definite functional in S(RN ),
necessarily its measure support will always be on the topological dual of
s , for each p. As a consequence, its support will be on the union set
#p+1
∞
p=1 s−p (since s−p ⊂ s−(p+1) ) and thus it will be the whole distribution
space S (RN ).
Similar results hold true in other distributional spaces.
The application of the above-mentioned result in Gaussian path-
integrals is always made with the use of the famous result of the kernel
theorem of Schwartz–Gelfand.
August 6, 2008 15:46 9in x 6in B-640 ch05
where T (j) denotes the action of the distribution T on the test function
j ∈ S(RN ).
At this point of our exposition let us show how to produce a fundamental
solution for a given differential operator P (D) with constant coefficients,
namely
P (D) = ap D p . (5.107)
|ρ|≤m
or equivalently
where t P (D) is the transpose operator through the duality of S(RN ) and
S (RN ).
August 6, 2008 15:46 9in x 6in B-640 ch05
with (p) depending on the order of Hp and the order of the differential
operator t P (D). (For instance in S(R): (d/dx)Hn (x) = 2nHn−1 (x);
d2 d
2
Hn (x) = 2n Hn−1 (x) = 4n(n − 1)Hn−2 (x), etc.,
d x dx
one obtains the recurrence equations for the searched coefficients Ep in
Eq. (5.110)
ϕn Mnq Eq = ϕn Hn (0) (5.113)
n |q|≤(n) n
2
for any (ϕn ) ∈ or equivalently
Mnq Eq = Hn (0) (5.114)
|q|≤(n)
(5.119a)
1
holds mathematically true since the locally integrable function 8π |x|2 n|x|
is a fundamental solution of the differential operator ∆ : S (R ) → S(R4 )
2 4
P −α : L2 (RN ) → H+α
(5.122)
ϕ → (P −α ϕ)
P α : H+α → L2 (RN )
(5.123)
ϕ → (P α ϕ)
since
and
(T −1 )∗ Hα = [(P −α T0 P α )−1 ]∗ (P α (S(RN )))
= [(P α )∗ (T0−1 )∗ (P −α )∗ ]P α (S(RN ))
= P α T0−1 (L2 (RN )). (5.128)
2
, +2 m N) : S (R ) → S(R
In the important case of L = (−∆ N N
)
∗ 2 −2β ∗
and T0 T0 = (−∆ + m ) ∈ 1 (L (R )) since Tr(T 0 T 0 ) =
2 N Γ( N )Γ(2β− N )
1
2(m2 )β
( m1 ) 2 2 Γ(β) 2 ∞ for β N4 with the choice P = (−∆ + m2 ), we
can see that the support of the measure in the path-integral representation
of the Euclidean measure field in RN may be taken as the measurable subset
below
with a function f (x) ∈ L2 (RN ) and any given ε > 0, even if originally
all fields configurations entering into the path-integral were elements of
the Schwartz tempered distribution spaces S (RN ) certainly very “rough”
mathematical objects to characterize from a rigorous geometrical point of
view.
We have, thus, made a further reduction of the functional domain of the
free massive Euclidean scalar field of S (RN ) to the measurable subset as
given in Eq. (5.130) denoted by W (RN )
1
exp − [(−∆ + m2 )−1 j](j)
2
= d(−∆+m2 ) µ(ϕ)eiϕ(j)
S (RN )
2 N
) 4 +ε−1 f
= d(−∆+m2 ) µ̃(f )eif,(−∆+m L2 (RN ) . (5.131)
W (RN )⊂S (RN )
References
10. E. Nelson, Regular probability measures on function space, Ann. Math. 69(2),
630–643 (1959); V. Rivasseau, From Perturbative to Constructive Renormal-
ization (Princeton University, 1991).
11. L. C. L. Botelho, Il Nuovo Cimento 117B, 37 (2002); Il Nuovo Cimento
117B, 331 (2002); J. Phys. A: Math. Gen. 34, 131–137 (2001); Mod.
Phys. Lett. 173, Vol. 13/14, 733 (2003); Il Nuovo Cimento 117B, 15
(2002); Ya. G. Sinai, Topics in Ergodic Theory (Princeton University Press,
1994).
12. R. Teman, Infinite-Dimensional Systems in Mechanics and Physics, Vol. 68
(Springer-Verlag, 1988); J. M. Mourão, T. Thiemann and J. M. Velhinho, J.
Math. Phys. 40(5), 2337 (1995).
13. A. M. Polyakov, Phys. Lett. 103B, 207 (1981); L. C. L. Botelho, J. Math.
Phys. 30, 2160 (1989).
14. L. C. L. Botelho, Phys. Rev. 49D, 1975 (1994).
15. A. M. Polyakov, Phys. Lett. 103B, 211 (1981).
16. L. Brink and J. Schwarz, Nucl. Phys. B121, 285 (1977).
17. L. C. L Botelho, Phys. Rev. D33, 1195 (1986).
18. B. Durhuus, Quontum Theorem of String, Nordita Lectures (1982).
19. C. G. Callan Jr., S. Coleman, J. Wess and B. Zumine, Structure of Phe-
nomenological Lagrangians — II, Phys. Rev. 177, 2247 (1969).
20. L. C. L. Boteho, Path integral Bosonization for a non-renormalizable
axial four-dimensional. Fermion Model, Phys. Rev. D39(10), 3051–3054
(1989).
Appendix A
with
LΛF [ĝ(Pλ1 , . . . , PλN )] = Q
ĝ(Pλ1 , . . . , PλN )(dPλ1 · · · dPλN ).
λ∈ΛF (Rw )λ
(A.4)
Here, ĝ(Pλ1 , . . . , PλN ) still denotes the unique extension of Eq. (A.2) to the
Alexandrov compactification Rw .
Now, we remark that the above family of linear continuous functionals
have the following properties:
(2) If the index set ΛF , contains ΛF the restriction of the associated linear
functional ΛF , to the space C( λ∈Λf (Rw )λ , R) coincides with LΛF .
August 6, 2008 15:46 9in x 6in B-640 ch05
where the union is taken over all family of subsets of finite elements of the
index set A.
As a consequence of the Remark 2 and Eq. (A.6) there is a unique
extension of the family of linear functionals {LΛF } to the whole space
C( λ∈A (Rw )λ , R) and denoted by L∞ . The Riesz Markov theorem gives
us a unique measure dµ̄(h) on λ∈A (Rw )λ representing the action of this
functional on C( λ∈A (Rw )λ , R).
We have, thus, the following functional integral representation for the
function f (g):
f (g) = Q exp(ih̄(g))dµ(h̄). (A.7)
( w λ
λ∈A (R ) )
$N
Or equivalently (since h̄(g) = i=1 pi ai for some {pi }i∈N < ∞), we have
the result
f (g) = Q (exp ih(g))dµ(h) (A.8)
( λ∈A Rλ )
For the proof of the theorem under this more general mathematical
condition, we will need two lemmas and some definitions.
with
Proof. Let f (x) ∈ C̃0 (X, R) and gµ ∈ C(X, R), the (Uryhson) functions
c
associated with the closed disjoints sets K̄n and (Kn+1 ). Now it is straight-
forward to see that (f · gn )(x) ∈ Ci (X, R) and converges uniformly to f (x)
by definition (A.11).
At this point, we consider a linear positive continuous functional
L on C̃0 (X, R). Since the restriction of L to each subspace C(Kn , R)
satisfy the conditions of the Riesz–Markov theorem, there is a unique
measure µ(n) on Kn containing the Borelians on Kn and representing
this linear functional restriction. We now use the hypothesis equation
(A.10) to have a well-defined measure on a σ-algebra containing the
Borelians of X
µ̄(A) = lim sup µ(n) A Kn . (A.12)
λ
Unfortunately, the non-countable product space λ∈A R is not a
normal topological space (the famous Stone counter example) and we
cannot, thus, apply the above lemma to our vectorial case equation (A.8).
However, we can overcome the use of the Stone–Weierstrass theorem in the
proof of the Bochner–Martin–Kolmogorov theorem by considering directly
a certain functional space instead of that given by Eq. (A.6).
Thus, we define the following space of infinite-dimensional functions
vanishing at finite
def
!
C0 (R∞ , R) ≡ C0 Rλ , R = C̃0 Rλ , R , (A.14)
λ∈A ΛF ⊂A λ∈λF
Let us show explicitly by one example of ours of the quite complex behavior
of cylindrical measures on infinite-dimensional spaces R∞ .
Firstly, we consider the family of Gaussian measures on R∞ =
{(xn )1≤n≤∞ , xn ∈ R} with σn ∈ 2 .
N
x2
(∞) 1 − 2σn2
d µ({xn }) = lim sup dxn √ e n . (B.1)
N
n=1
σn π
R∞ ε→0+
1 n
2
1
= lim lim sup (1 + 2ε α2n σn2 )− 2 (B.3)
ε→0+ 0≤≤n =1
Note that
n
1 1
(1 + 2ε α2n σn2 )− 2 ≤ $n . (B.4)
1+ =1 α2n σn2
=1
As a consequence, one can exchange the order of the limits in Eq. (B.3)
and arriving at the result
1 n 2
1
(∞)
µ(E(αn ) ) = lim sup lim (1 + 2ε α2n σn2 )− 2
0≤≤n ε→0+
=1
So we can conclude, on the basis of Eq. (B.5) that the support of the
measure equation (B.1) is the set of E(αn ) for any possible sequence
{αn } ∈ 2 . Let us show that (E(αn ) )C ∩ E(βn ) = {φ}, so that these sets
are not coincident.
Let the sequences be
σn = n−σ ,
αn = nσ−1 , (B.6)
βn = nσ−λ ,
βn2 σn2 = n−2λ < ∞. (B.8)
We have that
2
(x̄n )2 α2n = n−2(σ−1)−ε · nn (σ−1)
= n−ε
and
(x̄n )2 βn2 = n−2(σ−1)−ε · n2(σ−λ) = n2−ε−2λ
where the subscript f reminds us the functional nature of the objects under
study and so its trace.
It is thus expected that the definition equation (C.1) has divergents
counter terms as ε → 0+ , since exp(−tA) is a class trace operator only
for t ≥ ε. Asymptotic expressions at the short-time limit t → 0+ are well-
known in mathematical literature. However, this information is not useful in
the first sight of Eq. (C.1) since one should know TrF (e−tA ) for all t-values
in [ε, ∞).
A useful remark on the exact evaluation is in the case where the operator
A is of the form A = B + m2 1 and one is mainly interested in the effective
August 6, 2008 15:46 9in x 6in B-640 ch05
(C.2)
1
(σ) 2
+ dσ lim TrF [ge−ε(A ) ] . (C.3b)
0 ε→0+
The proof of Eqs. (C.3a) and (C.3b) is based on the validity of the
differential equation in relation to the σ-parameter
d * 2 2
+
[log detF (A(σ))2 ] = lim+ 2 TrF (f e−ε(A(σ)) ) + Trε (ge−ε(A(σ)) )
dσ ε→0
(C.4)
which can be seen from the obvious calculations written down in the above
equation
Here, the Chiral Phase W [ψ] in Eq. (C.6) takes value in SU (N ) for instance.
One can see that
7 1
( D(σ))2 = −(∂µ + iGµ (σ))2 1 − [γµ , γν ] Fµν (−iGµ (σ)) (C.7a)
4
with the Gauge field
γµ Gµ = γµ (W −1 ∂µ W ). (C.7b)
Hence, the asymptotics of the operator (C.7a) are easily evaluated (see
Chap. 1, Appendix E)
1 g(iεµν γ5 )
= lim+ Tr 1+ Fµν (−iGµ (σ)) , (C.8a)
ε→0 4πε 2
where we have used the Seeley expansions below for the square of the Dirac
operator in the presence of a non-Abelian connection
∂ ∂
Aϕ = (−∆) − V1 (ξ1 , ξ2 ) − V2 (ξ1 , ξ2 ) − V0 ((ξ1 , ξ2 ), (C.8c)
∂ξ1 ∂ξ2
8 9
lim TrCc∞ (R2 ) (e−tA )
t→0+
1 1 ∂ ∂ 1 1
= + − V1 + V0 − (V 2 + V22 ) − V0
4πt 8π ∂ξ1 ∂ξ2 16π 1 4π
× (ξ1 , ξ2 ) + O(t), (C.8d)
7 7
( ∂ − ig Gµ )2 = (−∂ 2 )ξ + (2igGµ ∂m u)ξ
& '
igσ µν 2 2
+ ig(∂µ Gµ ) + Fµν (G) + g Gµ , (C.8e)
2 ξ
August 6, 2008 15:46 9in x 6in B-640 ch05
detF (D(1))
log
detF (∂)
& 1 '
i 2 a µν
= d x TrSU(N ) ϕa (x)λ dσEµν F (−iGµ (σ)) .
2π 0
(C.9)
A more invariant expression for Eq. (C.9) can be seen by considering the
decomposition of the “SU (N ) gauge field” Gµ (σ) in terms of its vectorial
and axial components:
Fµν (−iGµ (σ)) = {(iDαv (σ)Aα (σ))E µν − [Aµ (σ), Aν (σ)] + Fµν (Vµ (σ))}.
(C.12)
Here
Note that Aµ (σ) and Vµ (σ) are not independent fields since the chiral
phase W (σ) satisfies the integrability condition
Fµν (W ∂µ W ) ≡ 0 (C.14)
or equivalently
After substituting Eqs. (C.12)–(C.16) in Eq. (C.9), one can obtain the
results
1
i
=
d2
x Tr λa
φ (α) dσ(2iDµ (σ)Aµ (σ)
SU(N ) a
2π
0
−[Aµ (σ),Aν (σ)]
: ;< =
E ) − [Aµ (σ), Aν (σ)] ,
+ µν ( F µν (Vα (σ)) (C.17)
i
d2 x TrSU(N )
=
detF (D(1)) 2π
log 1
detF (∂)
× λ φa (x)
a
dσ(2iDµ (σ)Aµ (σ))
0
i
+ d2 x TrSU(N )
2π
1
× a
λ φa (x) dσ(−2[A µ (σ), A ν (σ)])
0
= I1 (φ) + I2 (φ). (C.18)
Let us now show that the term I1 (φ) is a mass term for the physical gauge
field Aµ (σ = 1) = Aµ .
First, we observe the result
Lµ (σ)
: ;< =
TrDirac ⊗ TrSU(N ) γ 5 Aµ (σ) dσ
d
(W ∂µ W )(σ)
(C.21)
Since we have the identity as a consequence of the fact that [σγ5 φa (x)λa ,
γ5 φb (x)λb ] = σ(φa (x)φb (x)[λa , λb ] ≡ 0;
∂
W −1 (σ) W (σ) = γ5 φa (x)λa , (C.22)
∂σ
we can see the appearance of a term of the form of a Wess–Zumino–Novikov
topological functional for the Chiral group SU (N ), namely
i ∂
I2 (φ) = exp − d x TrColor⊗Dirac Eµν γ5 W −1 (σ) W (σ)
2
2π ∂σ
× [γ5 W −1 (σ)∂µ W (σ), γ5 W −1 (σ)∂ν W (σ)]
Finally, let us call our readers attention that the Dirac operator in the
presence of a non-Abelian SU (N ) gauge field Aµ (x) = Aaµ (x)λa , can always
be re-written in the “chiral phase” in the so-called Roskies gauge fixing
Here
* Rx +
> [φ] = eiγs φea (x) ≡ PDirac PSU(N ) eiγs −∞ dξ (εµν Gν)(x) .
µ
W (C.26)
Since
> )∂µ (W
(γµ Gµ )(x) = +γµ (W > )−1 (x). (C.27)
(C.28)
since
* +
> )W
γµ (δGµ ) = γµ ∂µ (δ W > −1 + ∂µ W
> (−W > )W
> −1 (δ W > −1 )
>W
= γµ (∂µ − [Gµ , ])(δ W > −1 )
and
>W
[γµ , δ W > −1 ] = δ W
>{γµ , W
> −1 } + {γµ , δ W
> }W
> −1 ≡ 0 (C.29)
August 6, 2008 15:46 9in x 6in B-640 ch05
1
Here, detF,adj φφ∗ 2 is the functional Dirac operator in the presence of the
gauge field and in the adjoint SU (N )-representation. Its explicit evaluations
are left to our readers.
The full gauge-invariant expression for the Fermion determinant is con-
jectured to be given on explicit integration of the gauge parameters con-
sidered now as dynamical variables in the gauge-fixed result. For instance,
in the Abelian case and in the gauge fixed Roskies gauge equation (C.6)
result, we have the Schwinger result
and thus, we have the exactly soluble expression for the (Q.E.D.)2 -
generating functional (a non-gauge invariant object!)
1
Z[Jµ , η, η̄] = DF [φ]DF [χ]DF (χ̄]
2
& '
1 e2
× exp − d2 x φ − ∂ 4 + ∂ 2 φ + Eµν (∂ν Jµ )φ (x)
2 π
& 7'
1 0 i∂ χ
× exp − d2 x(χ, χ̄) 7
2 i∂ 0 χ̄
& −igγ φ '
s
2 e 0 η
× exp − d x(χ, χ̄)
0 e−igγs φ η̄
(C.37)
where
−iγµ Gµ (σ) = exp (σγ 5 φa (x)λa )(i∂) exp (σγ 5 φa (x)λa ) (C.38)
Aµ (σ) = ∆−1
γs φa λa {sin h(∆γs φa λa ) ◦ ∂µ (γ5 φ λa )},
a
(C.39b)
Vµ (σ) = ∆−1
γs φa λa {(1 − cos h(∆γs φa λa )) ◦ ∂µ (γ5 φ λa )}
a
(C.39c)
∆X ◦ Y = [X, Y ] (C.39d)
(n)
and ∆X denoting its n-power.
For a complete quantum field theoretic analysis of the above formulae in
an Abelian (theoretical) axial model we refer our work to L. C. L. Botelho.20
Finally and just for completeness and pedagogical purposes, let us
deduce the formal short-time expansion associated with the second-order
positive differential elliptic operator in Rν used in the previous cited
reference
Its evolution kernel k(x, y, t) = x| − exp(−t L) |y satisfies the heat-kernel
equation
∂
K(x, y, t) = −Lx K(x, y, t), (C.41)
∂t
August 6, 2008 15:46 9in x 6in B-640 ch05
H1 (x, x) = −V (x),
1
H2 (x, x) = [−∂ 2 V + aµ ∂µ V + V 2 ](x). (C.47)
2
with
Aj (x, ξ, λ) = aj (x)(ξ1α1 )(ξ2α2 ) 0 ≤ j < 2, (D.5)
|α|=α1 +α2 =j
A2 (x, ξ, λ) = −λ1q×q + aj (x)(ξ1α1 )(ξ2α2 ) j = 2. (D.6)
|α|=α1 +α2 =j
1
∞
[(Dξα (σ(A − λ1))(x, ξ)Dxα [C−2−j (x, ξ, λ)] = 1. (D.10)
a!
|α|≤2
j=0
1
C−2−j (x, pξ , (p(λ ) 2 )2 ) = p−(2+j) C−2−j (x, ξ , λ ). (D.12)
After using Eq. (D.11) in Eq. (D.10) and for each integer j, comparing the
resultant power series in the variable 1/p(1 = 1 + 0( p1 ) + · · · + 0( p1 )n + · · · ),
one can get the result as
8 9
>
> >
>
< X =
1
0 = a2 (x, ξ)C−2−j (x, ξ) + Dξα ak (x, ξ) ((iDxα C−2− (x, ξ, λ)))
>
> α! >
>
: <j ;
k−|α|−2−=−j
(D.14)
with all the coefficients in Cc∞ (R2 , Rq ), one obtains the following results
after calculations:
and
∞ 2
1 1
=
j=0
2π 2πi
& −∞ '
× d(−is)eist d2 xd2 ξC−2−j (x, ξ, −is)
+∞ R2 ×R2
∞
& +∞ '
1 1 2 2 is −is
= d ξ d x e C−2−j (x, ξ, )ds
j=0
t (2π)3 R2 R2 −∞ t
∞
1 1 2+j 1
= d2 ξ d2 xeis t( 2 )
C−2−j (x, t 2 ξ, −is)
j=0
t (2π)3 R2 R2
∞
+∞
(j−2)
−3 2 2
= t 2 (2π) d ξ d x C−2−j (x, ξ, −is) .
j=0 R2 R2 −∞
(D.18)
August 6, 2008 15:46 9in x 6in B-640 ch05
& '
1 (A1 )2 (A2 )2
+ d2 x − Tr + + A0 + 0(t). (D.19)
4 g11 g22
By applying the above formula for the Polyakov’s covariant path-integrals,
let us first introduce the R2 complex structure
z = x1 + x2 , z̄ = x1 − ix2 , (D.20)
∂ ∂ ∂ ∂ ∂ ∂
= −i , = +i . (D.21)
∂z ∂x1 ∂x2 ∂ z̄ ∂x1 ∂x2
For each integer j, let us define two Hilbert spaces Hj and H̄j as follows:
(a) Hj is defined as the vector space of all complex functions f (z, z̄) =
f1 (x, y) + if2 (x, y) with the following tensorial behavior under the
action of a conformal tranformation
z = z(w), (D.22)
−j
∂w
f (z, z̄) = f˜(w, w̄). (D.23)
‘∂z
Let us introduce into Hj the following inner product
(q, f )Hj = dz z̄(ρ(z, z̄)j+1 ḡ(z, z̄)f (z, z̄)), (D.24)
R2
(b) H̄j is defined same as above exposed with the following tensor
−j
∂w
f (z, z̄) = f˜(w, w̄). (D.25)
∂z
At this point, we can verify that the above written inner products are
conformal invariant
j+1
∂z ∂ z̄ ∂w 2
(g, f )Hj = dwdw̄ ρ̃(w, w̄)
∂w ∂w ∂z
R2
1 −j 2 1 −j 2
∂ ∂w
× w f˜(w, w̄) (g̃(w, w̄))
∂z ∂z
= dwdw̄(ρ̃(w, w̄))j+1 f˜(w, w̄)(g̃(w, w̄)). (D.26)
R2
(a) Lj = Hj → H̄−(j+1)
f → (ρ(z, z̄))j (∂z̄ + Az̄ )f,
(D.27)
(b) L̄j = H̄j → H(j+1)
f → (ρ(z, z̄))j (∂z + Az )f
together with the adjoint operators (Lj ϕ, f )H̄−(j+1) = ϕ, L∗j f Hj , namely:
Lj = L∗j Lj = −L̄−(j+1) Lj : Hj → Hj ,
(D.29)
L̄ = (L̄j )∗ L̄j : −L−(j+1) L̄j : H̄−(j+1) → H̄−(j+1) .
(D.36)
August 6, 2008 15:46 9in x 6in B-640 ch05
since
δ( g det Lj )
& '
1 ϕ(x) (1 + 3j)
= −(j + 1) 2
d xδϕ(x) e − ∆ϕ(x)
2πε 12π
R2 ε→0+
& '
1 ϕ(x) (2 + 3j)
+j 2
d xδϕ(x) e + ∆ϕ(x) . (D.41)
2πε 12π
R2 ε→0+
Chapter 6
The words set and function are not as simple as they may seem.
They are potent words. They are like seeds, which are prim-
itive in appearance but have the capacity for vast and intricate
developments — G.F. Simmons.
Let us begin our proof by introducing the smallest ring containing the
compact subsets of X.
∗ Author’s original results.
169
August 6, 2008 15:46 9in x 6in B-640 ch06
here α∆β denotes the topological closure of the difference set α∆β = (α −
β) ∪ (β − α) = (α ∩ β ) ∪ (β ∩ α ). On this coset algebra of sets A/∼, denoted
by EBaire (X), we introduce a metrical structure by means of the metric set
function
Now the assertive expressed by Eq. (6.1) is a simple result of the defi-
nition of integration
)=L
L(f ) = L(f lim Sn (f ) = lim L(Sn (f ))
n→∞ n→∞
n
n2
j − 1
= lim n
µL (En,j ) + nµL (Fn )
n→∞
j=1
2
def
= lim fn (x)dL µ(x) ≡ f (x)dL µ(x), (6.8)
n→∞ X X
Proof. Let us firstly consider the Fourier transform of the function fΩ (p).
Namely
n
1
φL (x) = √ dN p · eiP ·x fΩ (p). (6.15)
2π F N
(6.16)
which means that the inversion Fourier transform theorem holds true
N
1
L((x1 )1 · · · (xn )n ) = √ dN xe−ipx {(x1 )1 · · · (xn )n φL (x)}|p=0
2π RN
N
1
= √ dN x(x1 )1 · · · (xn ) φL (x) . (6.17)
2π RN
August 6, 2008 15:46 9in x 6in B-640 ch06
continuous applications.
A topological basis for the origin of C ∞ (Ω) is formed by all those convex
and barreleds sets U ⊂ C ∞ (Ω) such that U ∩ C ∞ (Kn ) is always a neigh-
borhood of the origin in C ∞ (Kn ). The main result and reason for intro-
ducing such inductive topology in C ∞ (Ω) is that it leads to the fundamental
∞
result that Cind (Ω) is a sequentially complete topological vector space.
At this point, it is worth to call to the readers’ attention that the usual
∞
nondistributional topological definition of C ∞ (Ω) as m=0 C m (Ω) (always
August 6, 2008 15:46 9in x 6in B-640 ch06
where
1 N
∂ (Pn +···+Pn )
(Dp ϕ)(x) = Pn1 PN
ϕ(x1 , . . . , xN ). (6.22)
∂x1 · · · ∂xNn
Proof. Let L ∈ D (Ω), but with compact support Ks ⊂ Ω. By the inductive
limit topology (exercise 1), there exists a constant cs > 0, and an integer
ms > 0, such that for any ϕ ∈ D(Ω), we have the estimate
|L(ϕ)| ≤ cs sup sup |Dp ϕ(x)| . (6.23)
x∈Ks |p|≤ms
Note that the triple (C, Ks , m) is not unique. We now consider the
∂ p1
following elliptic operator L = ( ∂x 1
) · · · ( ∂x∂ n )pn (p = p1 + · · · + pn ). Since
August 6, 2008 15:46 9in x 6in B-640 ch06
f ∈ C(Ks ) : L(L−1
p f ) ≤ cs · sup |f (x)| (6.24)
x∈Ks
D : D(Ω) → D(Ω).
By the duality equation (6.26) mentioned above, one has a natural deri-
vative application in D (Ω)
def def
(−DL) ≡ (t D)(L)(f ) ≡ L(Df ), (6.29)
D (Ω)
besides being always a continuous operation in D (Ω) if Ln −→ ⇔
D (Ω) t
t
DLn −→ DL.
At this point, we call our reader to show that the sequence of functions
fn (x) = n1 sin(nx) seen as kernels of distributions in D (R) obviously con-
verges to the zero distribution in D (R):
1
lim dx sin nx ϕ(x) = 0. (6.30)
n→∞ R n
Namely
D (R)
d sin nx) D (R)
= cos(nx) −→ 0. (6.32)
dx n
∞
+ (Dps ϕ)(x)dνrsing (x) , (6.33)
s=0 Ω
August 6, 2008 15:46 9in x 6in B-640 ch06
where hs (x) ∈ L1 (Ω, dN x), and dVnsing (x) is a singular measure (in relation
to the Lebesgue measure dN x in Ω) with support at points (Dirac delta
functions) and on sets of Lebesgue zero measure
∞
dVssing (x) = a,s δ(x − x,s ) + dVs(continuous singular) (x). (6.34)
=0
We have, thus, the Kakutani theorem and the equivalence of the above
measures.
Theorem 6.5 (Kakutani). The two measures Eqs. (6.43a) and (6.43b)
are mutually equivalent or singular. In the first case, we have the criterion
that
λ2 − α2
n n
< ∞, (6.44)
n
λ n αn
On the basis of Eqs. (6.45) and (6.46), one can show the Wienner result
about translation invariant of Gaussian measures. Let Th : H → H be the
translation operator in H. Let us consider the translated Gauss measure
∞ '
%
− 12 λn [ϕ,ϕn +h,ϕm ]2
dA−1 µ(Th ϕ) = dA−1 µ(ϕ + h) = dϕ|ϕn e .
n=1
(6.47)
First, let us rewrite the finite action set of path integrated configurations
(= Dom(A)) in the following form (for ε > 0):
At this point, one can see that the usual Schwinger procedure to deduce
functional equations for the quantum field generating functional does not
make sense in the Euclidean framework of path-integrals since the measure
is not translational invariant. Namely, in the usual Feynmann notation
δ ( − 1 ϕ,Aϕ −V (ϕ) )
DF [ϕ] e 2 e = 0. (6.51)
H δϕ
As one can see from the above exposed result that the Minlos theorem is
a powerth “tool” in the functional integration theory in infinite dimension.
In this context, we have the converse of the Minlos theorem in Hilbert
spaces, the so called Sazonov theorem. Let us give a simple proof of this
result.
N N
(en , en )Sz = dµ(ϕ)ϕ, en ϕ, en
n=1 n=1 H
≤ dµ(ϕ)||ϕ||2 < ∞. (6.55)
H
|| ||
It is obvious to see that if one has a sequence σn ∈ H and σn −→ 0,
then we have that
(σn , σn )Sz = dµ(ϕ)(ϕ, σn )2 → 0. (6.56)
H
By using the proper time definition for evaluating the functional deter-
minant in Eq. (6.59)
1 d2
x|A|x = − δ(x − x2 ) + gV (x, x ) = B + gC. (6.62)
2 d2 x
d2 d2
[B, C] = 0 ⇔ V (x, x ) = V (x, x ). (6.63)
d2 x d2 x
C 2 (L2 (Ω), dA−1 µ(ϕ)). Let us start by considering the estimate of the
Newton differential quotient
# #
# ϕ(x + h) − ϕ(x) ##
dA−1 µ(ϕ) ## lim #
L2 (Ω) h→0 h
1/2
[φ(x + h)φ(x + h) + φ(x)φ(x) − 2φ(x + h)φ(x)]
≤ dA−1 µ(ϕ)
L2 (Ω) h2
1 −1
≤ [A (x + h, x + h) + A−1 (x, x) − (A(x + h, x) + A(x, x + h))]1/2 ,
h
(6.65)
Here, χBR (0) (x) is the characteristic function of the ball of radius R in
H (BR (0) = {x ∈ H | x, x ≤ R), u(x, t) ∈ C 2 (H, R) ∩ L2 (H, dQ u)
and- D2 the second-order
. (operator) Frechet derivative in H (for instance
2 iP,xH
Tr QD (e ) = −(p, QpH )eip,xH ).
August 6, 2008 15:46 9in x 6in B-640 ch06
one can rewrite Eq. (6.66) into a functional integral form where it is possible
now to consider a rigorous mathematical meaning for weak solutions in
L2 (H, dQ µ)
dA µ(x)BDu, BDg = dA µ(x)(f (x)χBR (x))g(x), (6.68)
H H
,+
where B ∈ 2 (H), and BB ∗ = Q is the usual decomposition of a trace
class (positive-definite) operator in terms of Hilbert–Schmidt root square
operator B.
Since Q is positive-definite, there exists a positive constant c such that
for any x ∈ H, we have the lower bound
We now show that the real linear function associated to our class of cut-
off datum f (x)χBR (x) as expressed by the right-hand side of Eq. (6.132) is
continuous in the topology associated to Eq. (6.70). Since the above space is
separable, one can verify the continuity through a sequential argument. Let
us thus consider a sequence of elements {gn } in H 1 (L2 (H, dQ µ)) converging
to a given ḡ ∈ H 1 (L2 (H, DQ µ)).
We have the estimate below
# #
# #
# dQ µ(x)(f (x)χBR (x))(gn − g)(x)#
# #
H
1/2
1/2
≤ ||f ||L2 (H,dQ µ) dQ µ(x)(gn − gn )2 (x)χBR (x) . (6.71)
H
August 6, 2008 15:46 9in x 6in B-640 ch06
Namely
Lf (g) = dQ µ(x)BDg, BDu(x). (6.75)
H
H = L2 ([0, 2π])
−1
d2 2
Q(σ, σ ) = − 2 + m (σ, σ ), (6.78a)
dx
Functional Laplacean equation
2π 2π
δ δ
dσ dσ Q(σ, σ
) ψ[f (σ)]
0 0 δf (σ ) δf (σ)
= FR [f (σ)]χ||f ||H 1 ≤R (f (σ)). (6.78b)
0
2
g
× exp − d4 x(ψ̄b γ 5 γ µ (λA )bc ψc )2 (x)
2 Ω
4
× exp −i d x(ψ̄a ηa + η̄a ψa )(x) . (6.80)
Ω
Z[ηa , η̄a ]
N%
2
−N N% 2
−N % 3
1
= D[ψa (x)]D[ψ̄a (x)] D[Aaµ (x)]
Z(0, 0) a=1 a=1 µ=0
1 4 0
∂ + igγ5 A ψa
× exp − d x(ψa , ψ̄a ) (x)
2 Ω )∗
( ∂ + igγ5 A 0 ψ̄a
1 4 4
× exp − d x(Aµ Aµ )(x) exp −i
a a
d x(ψ̄a ηa + η̄a ψa )(x) .
2 Ω Ω
(6.81)
At this point, we present our idea to bosonize (solve) exactly the above
fermion path-integral. The main point is to use the long-time ago suggestion
that at the strong coupling gbare → 0+ and at a large number of colors
(the t’Hooft limit2 ), one should expect a great reduction of the (continuum)
vector dynamical degrees of freedom to a manifold of constant gauge fields
living on the infinite-dimensional Lie algebra of SU (∞). In this approximate
leading t’ Hooft limit of large number of colors, we can evaluate exactly the
fermion path-integral by noting the Dirac kinetic operator in the presence
of constant SU (N ) gauge fields written in the following suitable form
1 4 0 V (ϕ) ∂V (ϕ) ψa
exp − d x(ψa ψ̄a ) (x) ,
2 Ω V (ϕ)∗ ∂ ∗ V ∗ (ϕ) 0 ψ̄a
(6.82)
with the chiral SU (N ) valued phase defined by the constant gauge field
configuration
After implementing the variable change equations (6.7) and (6.8), the
fermion sector of the generating functional takes to a form where the
independent Euclidean fermion fields are decoupled from the interacting–
intermediating non-Abelian constant vector field, Aaµ , namely
2
N% −N
1
Z[ηa η̄a ] = D[χa (x)]D[χ̄a (x)]
Z(0, 0) a=1
2
+∞ N% −N
1
× d[Aaµ ] × exp − (vol(Ω))(Aaµ )2
−∞ a=1
2
× det+1
[( ∂ + igγ5 A
F )( ∂ + igγ5 A )∗ ]
1 0 ∂ χa
× exp − d4 x(χa , χ̄a ) ∗ (x)
2 Ω ∂ 0 χ̄a
4 −gγ5 ϕ(x) −gγ5 ϕ(x)
× exp −i d x(χ̄a e ηa + η̄a e χa )(x) . (6.87)
Ω
(σ) = ( ∂ + ig A
D (σ) ) = exp{−gσγ5 ϕ(x)}( ∂) exp{−gσγ5 ϕ(x)}. (6.88)
August 6, 2008 15:46 9in x 6in B-640 ch06
detF [( ∂ + ig A )∗ ]
)( ∂ + ig A
= 1, (6.91)
detF [( ∂)( ∂)∗ ]
2
+∞ N% −N
1 2
× d[Aaµ ] exp
+ vol(Ω) TrSU(N ) (Aµ )
−∞ a=1
2
1 4 0 ∂ χa
× exp − d x(χa , χ̄a ) (x)
2 Ω ∂∗ 0 χ̄a
4 −gγ5 (Aa λa )xµ −gγ5 (Aa λa )xµ
× exp −i d x(χ̄a e µ ηa + η̄a e µ χa )(x) .
Ω
(6.92)
with Cxy a planar closed contour containing the points x and y and pos-
sessing an area S given roughly by the factor S = (x − y)2 .
The notation (0) means that the Fermionic average is defined solely
by the fermion-free action as given in the decoupled form (Eq. (6.18)).
Let us pass to the important step of evaluating the Wilson phase factor
average (Eq. (6.20)) on the coupling regime of gbare → 0+ and formally at
the limit of t’Hooft of large number of colors N → ∞. As the first step to
implement such evaluation, let us consider our loop Cxy as a closed contour
lying on the plane µ = 0, ν = 1 bounding the planar region S.
We now observe that the ordered phase-factor for constant gauge fields
can be exactly evaluated by means of a triangularization of the planar
region S, i.e.
M
S= ∆(i)
µν . (6.95)
l=1
(i)
Here, each counter-clock oriented triangle µν is adjacent to next one
(i) (i+1)
µν ∩ µν common side with the opposite orientations.
At this point, we note that
( −g R (i) Aα ·dxα )
P e ∆µν ∼ (1) (2) (3)
= e−gAα ·α · e−gAα ·α · e−gAα ·α , (6.96)
(i)
where { α }1=1,2,3 are the triangle sides satisfying the (vector) identity
(1) (2) (3)
α + α + α ≡ 0.
Since we have that
( H ) %n ( −g R (i) Aα dxα )
P e−g (x) Aα dxα = lim P e ∆µν
, (6.97)
n→∞
i=1
and by using the Campbel Hausdorff formulae to sum up the product limit
(Eq. (6.23)) at g 2 → 0 limit with X and Y denoting general elements of
the SU (N ) — Lie algebra:
1
ex · ey = eX+Y + 2 [X,Y ] + 0(g 2 ), (6.98)
August 6, 2008 15:46 9in x 6in B-640 ch06
one arrives at the non-Abelian stokes theorem for constant gauge fields
( R ) ( RR 01(s)
)
−g C Aα dxα
P e xy = P e−g S F01 dσ
( 2
)
= P e+(g) [A0 ,A1 ]·S . (6.99)
(g 2 s)2 1
∼ exp + (TrSU(N ) [A0 , A1 ])2 + O . (6.100)
2 N
Let us now substitute Eq. (6.26) into Eq. (6.20) and taking into account
the natural two-dimensional degrees of freedom reduction on the average
Eq. (6.20)
2
+∞ N% −N
1
G((x − y)) = d[Aa1 ]d[Aa0 ]
G̃(0) −∞ a=1
1
× exp + vol (Ω)[TrSU(N ) (A20 + A21 )]
2
(g 2 s)2
× exp + (TrSU(N ) [A0 , A1 ])2 , (6.101)
2
By using the well-known result (see Ref. 5 — pp. 307, 323, Eq. (3))
∞ 4 4
3 γ γ γ
exp(−β 2 x4 − 2γ 2 x2 )dx = 2− 2 e 2β2 K 14 2
, (6.104)
0 β 2β
−1 N 2 −N
+ (vol(Ω))2
„ « √
32 g2√SN 2 (vol(Ω))2 N 2
π
×e N 2 2 K 14 1 .
16g 4 N 2 S 2 vol(Ω) 2
2· 2
(6.105)
Let us now give a (formal) argument for the theory’s triviality at infinite
volume vol(Ω) → ∞. Let us, first define the infinite-volume theory’s limit
by means of the following limit
vol(Ω) = S 2 , (6.106)
So, we can see that for big N , there exists a fast decay of Eq. (6.19)
without any bound on the power decay. However, in the usual L.S.Z.
framework for quantum fields, it would be expected a nondecay of such
factor as in the two-dimensional case (see Eq. (6.33) for vol(Ω) = S),
meaning physically that one can observe fermionic scattering free states at
August 6, 2008 15:46 9in x 6in B-640 ch06
(N 2 n2 a8 )
N 2 −N
2 na2 2
!
n→∞ N 32.
ḡ∞
√ N 2 (n2 a8 )
G(na)N 1 ∼ e 2
K 41
ḡ
∞
2 · na2 2 )2 n2 a4
16(g∞
N 2 −N
1
∼ . (6.109)
na4
2
N% −N 3
% R
1
d4 x(Aa a
Z(0, 0) = D[Aaµ (x)]e− 2 Ω µ Aµ )(x)
a=1 µ=0
)∗ ]
)( ∂ + igγ5 A
g detF [( ∂ + igγ5 A
/
= PSU(N ) · P Dirac exp − g Aµ (Xβ (σ)dXµ (σ)
Cxy Cxy
/
i α β
+ [γ , γ ] Fαβ (Xβ (σ))ds . (6.111)
2 Cxy
The sum over the closed loops Cxy with fixed end-point xµ is given by
the proper-time bosonic path-integral
$ '
∞
dT 1 T
= d4 xµ DF [X(σ)] exp − Ẋ 2 (σ)dσ) .
0 T χµ (0)=xµ =χµ (T ) 2 0
Cxy
(6.112)
one can easily see that the average over the Aµ (x) fields is straightforward
and produced as a result of the following self-avoiding loop action
2 / 3
PSU(N ) exp −g Aµ (Xβ (σ))dXµ (σ)
Cxx Aµ
−N
N% 2 2
−N
N
= DF [θa (σ)]DF [θa∗ (σ)] θa (0)θa∗ (T )
a=1 a=1
2
N −N
i T
d ∗
d
× exp dσ θa (σ) θa (σ) + θa∗ (σ) θa (σ)
2 0 a=1
dσ dσ
g2 T T
× exp dσ dσ [(θb (λa )bc θc∗ )(σ)(θb (λa )bc θc∗ )(σ )]
2 0 0
× δ (D) (Xµ (σ) − Xµ (σ ))dXµ (σ)dXµ (σ ) . (6.117)
August 6, 2008 15:46 9in x 6in B-640 ch06
which in turn leads to the thirring model’s triviality for space-time RD with
D ≥ 4.
References
16. K. Fujikawa, Phys. Rev. D 21, 2848 (1980); R. Roskies and F. Schaposnik,
ibid. 23, 518 (1981).
17. L. Bothelho, Phys. Rev. D 31, 1503 (1985); A. Das and C. R. Hagen, ibid.
32, 2024 (1985).
18. L. Botelho, Phys. Rev. D 33, 1195 (2986).
19. O. Alvarez, Nucl. Phys. B238, 61 (1984).
20. V. Romanov and A. Schwartz, Teor. Mat. Fiz. 41, 190 (1979).
21. K. Osterwalder and R. Schrader, Helv. Phys. Acata 46, 277
(1973).
22. P. B. Gilkey, in Proceeding of Symposium on Pure Mathematics, Stanford,
California, 1973, eds. S. S. Chern and R. Osserman (American Mathematical
Society, Poidence, RI, 1975), Vol. 127, Pt. II, p. 265.
23. B. Schroer, J. Math. Phys. 5, 1361 (1964); High Energy Physics and Ele-
mentary Particles (IAEC, Vienna, 1965).
24. K. Bardakci and B. Schroer, J. Math. Phys. 7, 16 (1966).
25. A. Jaffe, Phys. Rev. 158, 1454 (1966).
∂µ (ψ̄γ 5 γµ ψ) = 0, ∂µ (ψ̄γ µ ψ) = 0.
with
1 1
2 (ψ, ψ̄, φ, Aµ ) = −iψ̄γµ (∂µ − igAµ )ψ + (∂µ φ)(∂µ φ) + m2 φ2 .
L (A.5)
2 2
Now we can proceed as in the case of gauge theories9,10 in order to
decouple the fermion fields from the auxiliary vector field Aµ (x) by making
the chiral change of variables
In terms of the new fermionic field χ(x), χ̄(x) and taking into account the
Jacobian (A.8), the generating functional reads
Z[J, n, n̄] = D[χ]D[χ̄]D[φ]
1
× exp − d2 x[−iχ̄γµ ∂µ χ + (1 − g 2 /π)(∂µ φ)(∂µ φ)
2
1
+ m2 φ2 + Jφ + n̄eigγ5 φ χ + χ̄eigγ5 φ η](x) . (A.9)
2
August 6, 2008 15:46 9in x 6in B-640 ch06
Γαβ (x, y, z)
1 (xµ − yµ ) ig
= (γµ γ5 )αβ
2π |x − y|2 2π(1 − g 2 /π)1/2
m m
× K0 |x − z| − K 0 |y − z|
(1 − g 2 /π)1/2 (1 − g 2 /π)1/2
g2 m
× exp K 0 |x − y| . (A.12)
2π(1 − g 2 /π) (1 − g 2 /π)1/2
1
Zφ = ,
(1 − g 2 /π)1/2
1 (xµ − yµ )
(ψα (x)ψ̄β (y))|x−y|→0 ∼
(R)
= (γµ )αβ · (A.15)
2π |x − y|2
The ultraviolet behavior (A.14) implies that the fermion field carries an
2
anomalous dimension γψ = g(R) /4π. We remark that the model displays
the appearance of the axial anomaly as a consequence of the presence of
the massive field π.8,12 For instance, in terms of the bare field φ, we have
g g m2
∂µ (ψ̄γµ γ5 ψ) = φ = φ. (A.16)
π π 1 − g 2 /π
This fact is related to the fermion mass independence of the anomalous part
of the divergence of chiral current ∂µ (ψ̄γ µ γ 5 ψ) in fermion models inter-
acting with gauge fields. Proceeding as above we obtain the new effective
(bare) Lagrangian
L
(χ, χ̄, φ)(x)
2igγ5 ψ 1 2 1 2 2
= −iχ̄γµ ∂µ χ + µχ̄e χ + (1 − g /π)(∂µ φ)(∂µ φ) + m φ (x),
2 2
(A.18)
where now the fermion fields χ(x), χ̄(x) possess an interacting term
µχ̄[cos(2gφ) + iγ5 sin(2gφ)]χ which contributes to the phenomenon of for-
mation of fractionally fermionic solutions.15
As we have shown, chiral changes in path-integrals provide a quick,
mathematically and conceptually simple way to solve and analyze the model
studied in this appendix.
Note added. We could like to make some clarifying remarks on the analysis
implemented above. First, in order to implement the regularization rule
in Eqs. (A.3)–(A.11), we consider the associated perturbative power series
in the coupling constant g dimensionally regularized which automatically
takes into account the Wick normal-order operation in the correlation
functions (Eqs. (A.9)–(A.11)). Second, we observe that the model should
be redefined in the region g 2 > π since its associated Gell–Mann–Low
function has a nontrivial zero for g 2 = π (the model contains a tachyonic
excitation).12
B.1. Introduction
The study of two-dimensional fermion models in the framework of chiral
anomalous path-integral has been shown to be a powerful nonperturbative
technique to analyze the two-dimensional bosonization phenomenon.
It is the purpose of this appendix to implement this nonperturbative
technique to solve exactly a nontrivial and nonrenormalizable
four-dimensional axial fermion model which generalizes for four dimensions
the two-dimensional model studied in the previous appendix.
August 6, 2008 15:46 9in x 6in B-640 ch06
−g 4 g2
V (φ) = 4
φ(∂µ φ)2 (−∂ 1 φ) + 2 (−∂ 2 φ)(−∂ 2 φ). (B.2)
12π 4π
The presence of the above φ potential is necessary to afford the exact
solubility of the model as we will show later (Eq. (B.18)).
The Hermitian γ matrices we are using satisfy the (Euclidean) relations
The Lagrangian L 1 (ψ, ψ̄, φ) is invariant under the global Abelian and
chiral Abelian groups
On the other hand, the fermion measure D[ψ]D[ψ̄], defined by the eigen-
vectors of the Dirac operator D[φ] D[φ]∗ , is not invariant under the chiral
change and yields a nontrivial Jacobian, as we can see from the relationship
4
D[ψ]D[ψ̄] exp − d x(ψ̄ D[φ]ψ)(x)
and
By substituting Eqs. (B.14) and (B.15) into Eq. (B.12), we obtain finally
the result for the above-mentioned Jacobian:
From Eqs. (B.17) and (B.1), we can see that the (bare) coupling constant
2
g gets an additive (ultraviolet) renormalization. Besides the Jacobian term
cancels with the chosen potential V (φ) in Eq. (B.2).
As a consequence of all these results, we have the following expression
for Z[J, η, η̄] with the fermions decoupled:
1
Z[J, η, η̄] = D(φ]D(χ]D(χ̄]J[φ](x)
Z[0, 0, 0]
1 1 1 2
× exp − d4 x χ̄(γµ ∂µ )χ + x̄(iγµ ∂µ )χ + gR (∂µ φ)2
2 2 2
+ η̄ exp(igR γ5 φ)χ + χ̄ exp(igR γ5 φ)η . (B.19)
This expression is the main result of this appendix and should be com-
pared with the two-dimensional analogous generating functional analyzed in
Appendix A, Eq. (A.9). Now, we can see that the quantum model given by
Eq. (6.6), although being nonrenormalizable by usual power counting and
Feynman-diagrammatic analysis, it still has nontrivial and exactly soluble
Green’s functions. For instance, the two-point fermion correlation function
is easily evaluated and produces the result
However the path-integral study (Eq. (B.19)) shows that the operator
solution of Eq. (B.21) (in terms of free (normal-ordered) fields) is given by
B.3. Complement
We now briefly calculate the asymptotic term of the second-order positive
differential elliptic operator:
For this study, let us consider the more general second-order elliptic four-
dimensional differential operator (non-necessarily) Hermitian in relation to
the usual normal in L2 (RD ) (Ref. 22):
∂
K(x, y.ζ) = −Lx K(x, y; ζ),
∂ζ
lim K(x, y; ζ) = δ (D) (x − y). (B.26)
η→0+
H0 (x, x) = 14×4 ,
H2 (x) = −V (x),
H2 (x, x) = 2[−∂x2 V (x) + aµ (x)∂µ V (x) + V 2 (x)]. (B.29)
Now substituting the value aµ (x) = [−ζ ḡ∂µ φ]14×4 and V (x) =
[(−ḡζγ5 ∂x2 φ + ∂µ φ]14×4 into Eq. (A.6), we obtain the result (Eqs. (B.14)
and (B.15)) quoted in the main text of this appendix.
August 6, 2008 15:46 9in x 6in B-640 ch07
Chapter 7
7.1. Introduction
∂U (x, t)
= (+∆U )(x, t) + ∆(∧) F (U (x, t)) + f (x, t), (7.1)
∂t
209
August 6, 2008 15:46 9in x 6in B-640 ch07
n
f (n) (x, t) = (f (x, t), ϕi (x))L2 (Ω) ϕi (x), (7.5)
i=1
After substituting Eqs. (7.5) and (7.6) in Eq. (7.1), one gets the weak
form of the nonlinear diffusion equation in the finite-dimension approxi-
mation as a mathematical well-defined systems of ordinary nonlinear differ-
ential equations, as a result of an application of the Peano existence-solution
theorem.
∂U (n) (x, t)
, ϕj (x) + (−∆U (n) (x, t), ϕj (x))L2 (Ω)
∂t 2
L (Ω)
= (∇(∧) · [(F (U (n) (x, t))∇(∧) U (n) (x, t)], ϕj (x))L2 (Ω)
1 d
U (n) 2L2 (Ω) + (−∆U (n) , U (n) )L2 (Ω)
2 dt
+ d3 x(F (U (n) )(∇(∧) U (n) · ∇(∧) U (n) )(x, t) = (f, U (n) )L2 (Ω) . (7.8)
Ω
This result in turn yields a prior estimate for any positive integer p:
1 d 1
(U (n) 2L2 (Ω) ) + γ(Ω)U (n) 2L2 (Ω) + (F (U (n) )) 2 (∇U (n) )2L2 (Ω)
2 dt
1 1 (n) 2
≤ pf (x, t)L2 (Ω) + U L2 (Ω) .
2
(7.9)
2 p
U (n) 2H 1 (Ω) = (−∆U (n) , U (n) )L2 (Ω) ≥ γ(Ω)U (n) ||2L2 (Ω) . (7.10)
T T
1 1
≤ p ||f ||2L2 (Ω) dt + U (n) 2L2 (Ω) dt < M̄ < ∞,
2 0 2p 0
(7.11)
(−∆U (n) , U (n) )L2 (Ω) ≥ C̄0 (U (n) , U (n) )H01 (Ω) . (7.12)
Note that U (n) (0)2 ≤ 2g(x)2L2 (Ω) (see Eq. (7.8)), and {U (n) (T )2L2 (Ω) }
is a bounded set of real positive numbers.
As a consequence of a prior estimate of Eq. (7.11), one obtains that the
previous sequence of functions {U (n) } ∈ L∞ ([0, T ], H01(Ω) ∩ H 2 (Ω)) forms
a bounded set on the vector-valued Hilbert space L2 ([0, T ], H01 (Ω)) either.
Finally, one still has another a prior estimate after multiplying the
Galerkin system (Eq. (7.7)) by the time-derivatives U̇ (n) , namely
T 2
dUn (t)
dt ≤ Real(∆Un (T ), Un (T ))
0 dt L2 (Ω)
T
dUn
− (Un (0), Un (0)) + (∧)
dt ∆ F (Un (t)),
0 dt L2 (Ω)
T 2
T 2
1 1 dUn
≤ p ∆ (∧)
F (Un (t)) dt + dt .
2 0 L2 (Ω) 2p 0 dt L2 (Ω)
(7.13)
By noting that
2
T sup {F (x)}
∆(∧) F (Un (t))2L2 (Ω) dt ≤ ∆(∧) 2op
0 x ∈ [−∞, ∞]
T
× dtUn (t)2L2 (Ω) < ∞, (7.14)
0
dt } is
one obtains as a further result that the set of the derivatives { dU n
2 2 2 −1
bounded in L ([0, T ], L (Ω)) (so in L ([0, T ], H (Ω)).
August 6, 2008 15:46 9in x 6in B-640 ch07
Lemma 7.1. If Ū(1) and Ū(2) in L∞ ([0, T ] × L2 (Ω)) are two functions
satisfying the weak relationship below
T
∂v
dt Ū(1) − Ū(2) , − + (Ū(1) − Ū(2) , +∆v)L2 (Ω)
0 ∂t L2 (Ω)
× (F (Ū(1) − F (Ū(2) ); + ∆v)L2 (Ω) ≡ 0, (7.17)
August 6, 2008 15:46 9in x 6in B-640 ch07
then Ū(1) = Ū(2) a.e in L∞ ([0, T ] × L2(Ω)). The proof of Eq. (7.17) is easily
obtained by considering the family of test functions on Eq. (7.16) of the
following form vn (x, t) = g(ε) (t)e+αn t ϕn (x) with −∆ϕn (x) = αn ϕn (x) and
g(t) = 1 for (ε, T − ε) with ε > 0 arbitrary. We can see that it reduces to
the obvious identity (αn > 0).
T −ε
dt exp(αn t)(F (Ū(1) ) − F (Ū(2) ), ϕn )L2 (Ω) ≡ 0, (7.18)
ε
which means that F (Ū(1) ) = F (Ū(2) ) a.e on (0, T )×Ω since ε is an arbitrary
number. We have thus Ū(1) = Ū(2) a.e, as F (x) satisfies the lower bound
estimate by our hypothesis on the kind of nonlinearity considered in our
nonlinear diffusion equation (7.1).
inf(F (x))
|F (x) − F (y)| ≥ |x − y|. (7.19)
−∞ < x < +∞
T
× exp i dt D
d xU (x, t)J(x, t) , (7.21c)
0 Ω
August 6, 2008 15:46 9in x 6in B-640 ch07
T
1
Z[J(x, t)] = DF [U ] exp − dt dD x
2λ 0 Ω
× (∂t U − ∆U − ∆(∧) (F (U ))2 (x, t))
T
× exp i dt dD xU (x, t)J(x, t) . (7.21d)
0 Ω
Let us now present some mathematical results for the diffusion problem
in Hilbert spaces formed by square-integrable functions L2 (Ω),5 with the
domain Ω denoting a compact set of RD .
The diffusion equation in the infinite-dimensional space L2 (Ω) is given
by the following functional differential equation (see Ref. 5 for mathematical
notation).
∂ψ[f (x); t] 1
= TrL2 (Ω) ([QDf2 ψ[f (x, t)]) (7.23a)
∂t 2
Qϕn = λn ϕn , (7.24)
The L2 (Ω)-infinite-dimensional diffusion equation takes the usual form:
Ψ fn ϕn , t = ψ (∞) [(fn ), t], (7.25a)
n
Ω fn ϕn = Ω(∞) [(fn )], (7.25b)
n
∂ψ (∞) [(fn ), t]
= [(λn ∆fn )ψ (∞) [(fn ), t]], (7.25c)
∂t n
Here, the integral operator Kernel of the trace class operator is explicitly
given by
Q(x, x ) = (λn ϕn (x)ϕn (x )). (7.26c)
n
At this point it is worth to call to the readers’ attention that dtQ µ and
dQ µ Gaussian measures are singular to each other by a direct application of
Kakutani theorem for Gaussian infinite-dimensional measures for any time
t > 0:
dtQ µ[g(x)]/dQ µ[g(x)] = +∞. (7.31)
Let us apply the above results for the physical diffusion of polymer
rings (closed strings) described by periodic loops X(σ) ∈ RD , 0 ≤ σ ≤
A = X(σ) with a nonlocal diffusion coefficient Q(σ, σ ) (such that
A,AX(σ+A)
0
dσ 0 dσ Q(σ, σ ) = Tr[Q] < ∞). The functional governing equation in
loop space (formed by polymer rings) is given by
A A
∂ψ (ε) [X(σ); A] δ2
dσ Qij (σ, σ )
(ε)
= dσ ψ (ε) [X(σ), A],
∂A 0 0 δ Xi (σ)δ Xj (σ)
(7.32a)
August 6, 2008 15:46 9in x 6in B-640 ch07
A A
λ
ψ (ε)
[X(σ); 0] = exp − dσ dσ Xi (σ)Mij (σ, σ )Xj (σ ) .
2 0 0
(7.32b)
(7.34)
Here, the kernel operator R(λ ) satisfies the integral equation (accessible
for numerical analysis)
(7.37)
It is worth to call to the readers’ attention that if A ∈ 1 and B is
a bounded operator — so A · B is a class trace operator. The functional
determinant det[1 + AB] is a well-defined object as a direct result of the
obvious estimate, the result which was used to arrive at Eq. (7.37).
N
N
lim (1 + λn ) ≤ exp λn = exp (Tr AB).
N →∞
n=0 n=0
1
≤ DG(f (x), g(x))L2 (Ω) + dσ(1 − σ)(D2 G[f (x)
L2 (Ω) 0
+ σg(x)]g(x), g(x)L2 (Ω) dtQ µ[g(x)]
1
≤ 0 + C̄ dσ(1 − σ) g(x)2L2 (Ω) dtQ µ[g(x)]
0 L2 (Ω)
≤ C̄ g(x)2L2 (Ω) dtQ µ[g(x)]
L2 (Ω)
∂ 1
ψ[f (x), t] = TrL2 (Ω) [QDf2 ψ[f (x), t]] − V [f (x)]ψ[f (x), t] (7.39)
dt 2
with
T
×Ω X(σ)dσ + X(0)
0
T T
× exp − dσV X(σ )dσ + X(0) , (7.41)
0 0
References
Un L2 (Ω) ≤ εUn H01 (Ω) + C(ε)Un 2H −1 (Ω) . (A.1)
T
+ 2εC(ε) dt(Un − Um H01 (Ω) × Un − Um H −1 (Ω) )
0
0
T T
≤ ε2 dtUn − Um 2H 1 (Ω) + C(ε)2 dtUn − Um 2H −1 (Ω)
0
0 0
T
1
2 T
1
2
T
1
2
1
≤ 2ε2 M + 2εC(ε)M 2 dtUn − Um 2H −1 (Ω)
0
T
+ (C(ε)2 ) dtUn − Um 2H −1 (Ω) . (A.2)
0
At this point, we use the Arzela–Ascoli theorem to see that {Un (x, t)} is
a compact set on the space C([0, T ], H −1 (Ω)) since we have the set equi-
continuity:
t
Un (t) − Um (s)H −1 (Ω) ≤ Un (τ )H −1 (Ω) dτ
0
s
T
1
2
(1− 12 )
≤ |t − s| Un (τ )2H −1 (Ω) dτ
0
1
≤ M̄ |t − s| 2 . (A.3)
T
1
2
T
1
2
L2 L2
1
= dQ µ(p)F (p) · ei p,x L2 e− 2 t p,Qp L2 . (B.2)
L2
∞
∂2 P∞
QD2 ei p,x L2 = λ ei( n=1 pn xn ) = − ( p, QpL2 ) ei p,x L2 .
∂ 2 x
=1
(B.3)
August 6, 2008 15:46 9in x 6in B-640 ch07
1
2 1
2
≤ dQ µ(p)|F (p)| 2
dQ µ(p)| p, QpL2 (Ω) | 2
L2 (Ω) L2 (Ω)
Now, it is a simple application to verify that Eq. (B.2) satisfies the dif-
fusion equation in L2 (Ω) (or in any other separable Hilbert space). Namely
∂U (x, t) 1
= dQ µ(p)F (p)ei p,x L2 − p, QpL2 (Ω)
∂t L2 (Ω) 2
t
× e− 2 , p,Qp L2 (Ω) , (B.5)
t
2
TrL2 (Ω) [QD U (x, t)] = dQ µ(p)F (p){D2 ei p,x
}e− 2 p,Qp L2 (Ω)
L2 (Ω)
t
= dQ µ(p)F (p){− p, QpL2 }ei p,x
e− 2 p,Qp L2 (Ω) ,
L2 (Ω)
(B.6)
with
i p,x − 2t p,Qp L2 (Ω)
U (x, 0) = dQ µ(p)F (p)e lim e = f (x). (B.7)
L2 (Ω) t→0+
August 6, 2008 15:46 9in x 6in B-640 ch08
Chapter 8
8.1. Introduction
227
August 6, 2008 15:46 9in x 6in B-640 ch08
Proof. In order to show the validity of the above ergodic limit, let us
rewrite Eq. (8.1) in terms of the spectral resolution of L, namely
T
1
I= |(ψ̃, exp(itL)ψ)|2 dt
T 0
T +∞ +∞
1 +itλ −itµ
= e dλ (ψ̃, Ej (λ)ψ) e dµ (ψ̃, Ej (µ)ψ) .
T 0 −∞ −∞
(8.2)
Additionally, f (λ) is uniform continuous on the closed interval [−λ̃, λ̃], and
f (λ) cannot make variations greater than 2ε on (−∞, −λ̃] and [+λ̃, ∞),
besides being a monotonic function on R. These arguments show the
uniform continuity of f (λ) on whole line R. Hence, we have that for a given
ε > 0, there exists a δ > 0 such that
ε
|(ψ, Ej (λ )ψ) − (ψ, Ej (λ )ψ)| ≤ , (8.11)
ψ 2
for
ε
≤ ψ 2
2
≤ ε. (8.14)
ψ
Let us note that for each ε > 0, there exists a set Dδ = {(λ, µ) ∈ R2 ;
|λ − µ| < δ} which contains the line λ = µ and has measure less than
ε in relation to the measure dλ (Ej (λ)ψ, ψ) ⊗ dµ (Ej (µ)ψ, ψ) as a result of
Eq. (8.14). This shows our claim that I = 0 in our special case.
In the general case of ψ̃ = ψ, we remark that solely the orthogonal
component on the continuity sub-space Hc (L) has a nonvanishing inner
product with exp(itL)ψ. By using now the polarization formulae, we reduce
this case to the first analyzed result of I = 0.
At this point, we arrive at the complete RAGE theorem.2
RAGE Theorem. Let K be a compact operator on (H, (, )). We have thus
the validity of the Ergodic limit
1 T
lim K exp(itL)ψ 2H dt = 0, (8.15)
T →∞ T 0
with ψ ∈ Hc (L).
August 6, 2008 15:46 9in x 6in B-640 ch08
We leave the details of the proof of this result for the reader, since any
compact operator is the norm operator limit of finite-dimension operators
and one only needs to show that
N
1 T
lim cn (eitL ψ, en )gn dt = 0, (8.16)
T →∞ T 0
n=0 H
One of the most important statement in physics is the famous zeroth law
of thermodynamics: “any system approaches an equilibrium state.” In the
classical mechanics frameworks, one begins with the formal elements of the
theory. Namely, the phase-space R6N associated to a system of N -classical
particles and the set of Hamilton equations
∂H ∂H
ṗi = − ; q̇i = , (8.17)
∂qi ∂pi
where H(q, p) is the energy function.
The above cited thermodynamical equilibrium principle becomes the
mathematical statement that for each compact support continuous func-
tions Cc (R6N ), the famous ergodic limit should hold true.3
1 T
lim f (q(t); p(t))dt = η(f ), (8.18)
T →∞ T 0
3N
∂f ∂H ∂f ∂H
(Lf )(p, q) = {f, H}(q, p) = − (q, p). (8.20)
i=1
∂qi ∂pi ∂pi ∂qi
The basic result we are using to show the validity of the Ergodic limit
equation (8.18) is the famous RAGEs theorem exposed in Sec. 8.1.
or
T T 12
1 1
β, Ut ψdt ≤ 2
(β, Ut ψ) dt (8.24)
T T 0
0
η(f ) = d3N qd3N p(f (p, q))hη(H) (p, q), (8.26)
R6N
3N
∂hη ∂H ∂H ∂hη
{hη(H) , L} = − = 0, (8.27)
i=1
∂qi ∂pi ∂qi ∂pi
exp{−βH(q, p)}
hη(H) (q, p) = , (8.28)
d3N qd3N p exp{−βH(q, p)}
T
1
lim dt(Ut ψ)(q, p)hη(H) (q, p) = hη(H) , Pker(−iL̄) (ψ), (8.29)
T →∞ T 0
that is
qi (0) = qi ,
pi (0) = pi ,
N
1 1 2a (qi+1 − q1 )2 g
= exp − + (qi )2k
Zn
(0) 2 i=1
N a2 2k
Zn(0) = d(inv) µ(q1 , . . . , qN ). (8.35b)
RN
August 6, 2008 15:46 9in x 6in B-640 ch08
where the wave field U (x, t) ∈ L∞ ([0, ∞), L2k−1 ((−a, a))) satisfies in a gen-
eralized weak sense the nonlinear wave equation below (see Appendix A):
∂ 2 U (x, t) ∂ 2 U (x, t)
2
= + g(U (x, t))2k−1 , (8.37)
∂ t ∂2x
U (x, 0) = U (0) (x) ∈ H 1 ((−a, a)), (8.38)
Ut (x, 0) = 0, (8.39)
dWiener µ[U (0) (x)] denotes the Wiener measure over the unidimensional
Brownian field end-points trajectories {U (0) (x)| with U (0) (−a) = U (0) (a) =
0}, and F (U (0) (x)) denotes any Wiener-integrable functional.
Note that C α (−a, a) denotes the Banach space of the α-Holder con-
tinuous function (for α ≤ 12 ) which is obviously the support of the Wiener
measure dWienerµ[U (0) (x)].
It is worth to remark the normalization factor Z of the nonlinear Wiener
measure implicitly used on Eq. (8.36).
a
1 g
dWiener µ[U (σ)] exp − dx(U (σ))2k = 1. (8.41)
Z C α (−a,a) 2k −a
Note there exists a constant C0 (Ω) > 0(C0 (Ω) → 0 if vol (Ω) → ∞) such
that the Garding coerciviness condition holds true on D(A)
∂ 2 U (x, t)
= (AU )(x, t) + G (U (x, t)), (8.53)
∂2t
U (x, 0) = U (0) (x) ∈ H 2m (Ω) ∩ H0m (Ω), (8.54)
Note that the Green function of the operator A belongs to the trace class
operators on H 2m (Ω), which results on Eq. (8.57) through the application
of the Minlo’s theorem.7
August 6, 2008 15:46 9in x 6in B-640 ch08
belongs to L1 (L2 (Ω), dA µ(φ(x))) due to the Lipschitz property of the non-
linearity G(z), i.e. there exist constants c+ and c− , such that
ϕ∗ (x)|∂Ω = 0. (8.66)
August 6, 2008 15:46 9in x 6in B-640 ch08
Note that the stationary strong solution ϕ∗ (x) is not a real atractor in the
sense of the theory of dynamical systems, since we can only say that there
is a specific sequence of growing {tn }n∈Z , with tn → ∞, such that at the
leading asymptotic limit β → ∞ we have the time-infinite limit for every
solution of Eqs. (8.65) and (8.66):
lim F (ϕ(x, tn ))
tn →∞
2
∗ 1δ F ∗
= F (ϕ (x)) + (ϕ (x))
β δ2ϕ
2
d δ G ∗
× lg det 1 + A−1 (ϕ ) + α + O(β −2 ) (8.68)
dα δ2 ϕ
α=0
2
1 δ F ∗
= F (ϕ∗ (x)) + (ϕ (x))
β δ2 ϕ
2 −1
−1 δ G ∗
× TrH 2m (Ω) 1+A (ϕ ) + O(β −2 )
δ2 ϕ
∗ 1 δ2 F ∗
= F (ϕ (x)) + (ϕ (x))
β δ2 ϕ
∞ 2 n
δ G
× (−1)n TrH 2m (Ω) A−1 2 (ϕ∗ ) + O(β −2 ) (8.69)
n=0
δ ϕ
In this case, one should consider that the continuum version of Eq. (8.70)
leads to (strongly) continuos functionals on H 2m (Ω) ∩ H0m (Ω):
The z → 0− limit of the integral of Eq. (8.75) (since Real (z) ⊂ ρ(A) ⊂
(−∞, 0)) can be evaluated by means of saddle point techniques applied to
Laplace’s transforms. We have the following result
∞
lim ((z + A)−1 f ) = lim e−zt (e−tA f ) dt
z→0− z→0− 0
−tA
= lim (e f ) = Pker(A) (f ). (8.76)
t→∞
Let us apply the above theorem to the Lengevin equation. Let us con-
sider the Fokker–Planck equation associated to the following Langevin
equation:
dq i ∂V j
(t) = − (q (t)) + η i (t), (8.77)
dt ∂qi
where {η i (t)} denotes a white-noise stochastic time process representing the
thermal coupling of our mechanical system with a thermal reservoir at tem-
perature T . Its two-point function is given by the “fluctuation–dissipation”
theorem
η i (t)η j (t ) = kT δ(t − t ). (8.78)
The Fokker–Planck equation associated to Eq. (8.74) has the following
explicit form
∂P i i
(q , q̄ ; t) = +(kT )∆qi P (q i , q̄ i ; t) + ∇qi (∇qi V · P (q i , q̄ i ; t)) (8.79)
∂t
lim P (q i , q̄ i ; t) = δ (N ) (q i − q̄ i ). (8.80)
t→0+
For the general Langevin equation in the complete phase space {qi , pi }
as in the bulk of this note, one should reobtain the complete Boltzmann
statistical weight as the equilibrium ergodic probability distribution
N
1 1 1 i 2
P (q̄ , p̄ ) = exp −
eq i i
(p̄ ) + V (q̄) , (8.86)
Z kT 2
l=1
Let us apply the above exposed result for the following nonlinear (poly-
nomial) stochastic diffusion equation on a one-dimensional domain Ω =
(−a, a) with all positive coefficients {λj } on the nonlinear polynomial term
1 d2 U (x, t)
2k−1
∂U (x, t)
=+ − λj U j
(x, t) + η(x, t), (8.88)
∂t 2 d2 x j=0
j=odd
(8.91)
In this technical section, we give an argument for the global existence and
uniqueness solution of the Hamiltonian motion equations associated first to
Eq. (8.32) — Sec. 8.3 and by secondly to Eq. (8.53) — Sec. 8.3. Related to
the two-dimensional case, let us equivalently show the weak existence and
uniqueness of the associated continuum nonlinear polynomial wave equation
in the domain (−a, a) × R+ .
∂ 2 U (x, t) ∂ 2 U (x, t)
− + g(U (x, t))2k−1 = 0, (8.94)
∂2t ∂ 2x
U (−a, t) = U (a, t) = 0,
(8.95)
U (x, 0) = U0 (x) ∈ H 1 ([−a, a]),
with
1 ˙ 2 2 1 2k
ϕ(t) = Ūn (t) L2 + γ0 (a) Ūn H1 + Ūn L2k . (8.100)
2 2k
At this point, we observe that for any p > 1 (with Ãi constants) and
T < ∞, we have the relationship
T T
p
p p
Ūn H 1 (−a,a) dt ≤ T (A1 ) 2 ⇔ Ūn L2 (−a,a) ≤ Ã1 , (8.107)
0 0
T T
p
Ū˙ n p
L2 (−a,a) dt ≤ T (A2 ) 2 ⇔ Ū˙ p
L2k (−a,a)
dt ≤ Ã2 , (8.108)
0 0
Un − Ū Lq (−a,a) ≤ Un − Ū 1−θ
L2 (−a,a) Un − Ū θ
L2k (−a,a) , (8.112)
with 0 ≤ θ ≤ 1
1 1−θ θ
= + , (8.113)
q 2 2k
d2
C ∞ ((0, T ), L2 (−a, a)) lim (Ūn , v)L2 (−a,a)
n→∞ d2 t
d2
+ − 2 Ūn , v + g(Ūn2k−1 , v)L2 (−a,a)
d x L2 (−a,a)
d2 d2
= 2 (Ū , v)L (−a,a) + − 2 Ū , v
2 + g(Ū 2k−1 , v)L2 (−a,a) = 0,
d t d x L2 (−a,a)
(8.114)
W (0) = 0, (8.117)
Wt (0) = 0. (8.118)
with the set of couplings {Cj } belonging to a positive set of real number.
In this case, we have that the general solution belongs a priori to the space
%
of functions j=1 L∞ ((0, T ), L2k−j (−a, a)).
j<2k
References
Associated to the Banach space C(X), there exists its dual M(X) which
is the set of random measures on X with norm given by its total variation
µ = |µ|(X), where µ ∈ M(X).
It is easy to see that if β(X) is the Stone-chech compactification of X,
we have the isometric immersion M(X) → M(β(X)).
We have, thus, the important theorem of Prokhorov on the accumulation
points of sets of random measures.7
Theorem (Prokhorov). If a family of measures {µα } on M(X) is such
that it satisfies the following conditions
(a) Bounded uniformly
(b) Uniformly regular — For any given ε > 0, there exists a compact set
K(ε) ⊂ X such that
f (x)dµα (x) ≤ ε f ∞ ,
X
and note the strict positivity of the integrand (the path-integral weight)
on Eq. (8.35) and its boundedness by the Gaussian free term since
g( 2a
N )
N (N
exp{− 21 (2k) l=1 (qi ) } < 1 for {qi } ∈
2k
l=1 (Ṙ)i .
August 6, 2008 15:46 9in x 6in B-640 ch08
We note either the chain property of the invariant measure set ensuring
us the measure uniqueness limit on Eq. (B.4) when applied to Eq. (8.35) —
Sec. 8.4, namely:
|f (q1 , . . . , qN −1 )|d(inv) µ(q1 , . . . , qN )
RN
≥ |f (q1 , . . . , qN −1 )|d(inv) µ(q1 , . . . , qN −1 ). (A.6)
RN −1
In the simple case of N -particles with unity mass and a positive two-
body interaction with a potential V (|xi − xj |2 ), such that the set in Rn ,
{xi | V (xi ) < E} is a bounded set for every E > 0. Our condition Eq. (B.1a)
takes a form involving only the configurations variables ({xp }p=1,...,N ; xp ∈
R3 ). Namely,
with
N
Mks ({xp }) = [V (|xi − xj |2 )4(xi − xj )2 (δis − δjs )(δik − δjk )
i<j
the (global) solutions Eq. (B.2) of the given mechanical system. Note that
H(xi (t), pi (t)) = E
P : HE → HE ,
(B.4)
(xi (t0 ), pi (t0 )) → (xi (T ), pi (T )),
Chapter 9
9.1. Introduction
254
August 6, 2008 15:46 9in x 6in B-640 ch09
+∞
X(t) = U (t)X(0) = eiwt d(E(w)X(0)) = eiHt (X(0)), (9.1)
−∞
T
1
L(Y (λ)) = lim dtE{Y (λ)X(t, λ)}. (9.2)
T →∞ 2T −T
August 6, 2008 15:46 9in x 6in B-640 ch09
T
1
= lim E{Y e−iHt X̄}dt
T →∞ 2T −T
= dµ(λ)Y (λ)dE(0)X(0, λ)
Ω
For the auto-correlation process function, we still have the result for the
translated time ζ fixed (the lag time) as a direct consequence of Eq. (9.1)
or the process’ stationarity property
T
1
lim dtE{X(t)X(t + ζ)} = E{X(0)X(ζ)}. (9.6)
T →∞ 2T −T
one has that the random variables as given by Eqs. (9.7) and (9.8) converge
at T → ∞ to the left-hand side of Eqs. (9.4)–(9.6) and producing thus
an Ergodic theorem on the equality of ensemble-probability average of the
wide sense stationary process {X(t), −∞ < t < ∞} and any of its single-
sample {X̄(t), −∞ < t < +∞} time average
T
1
lim dtf (X̄(t)) = E{PKer(H) (f (X0 ))} = E{f (X(t))},
T →∞ 2T −T
(9.10)
T
1
lim dtX̄(t)X̄(t + ζ) = E{X(0)X(ζ)}
T →∞ 2T −T
= E{X(t)X(t + ζ)}
= RXX (ζ). (9.11)
As much as in the usual proof of the Shanon results Eq. (9.15), we introduce
Eq. (9.16) into Eq. (9.14) and by using the Lebesgue convergence theorem,
August 6, 2008 15:46 9in x 6in B-640 ch09
N
1 ti
exp inarc sin · f (ti ) , (9.18)
i=1
2π A
with the spectral process possessing the canonical form (in the quadratic
mean sense)
+∞
π/2
X(w) = Jn (wA) einθ d(Xt=A sin θ ) . (9.20)
n=−∞ −π/2
Let us expose the usefulness of the sampling result Eqs. (9.17) and (9.18)
for the very important practical engineering problem of estimate the self-
correlation function of a time-finite observed sampling of an Ergodic process
August 6, 2008 15:46 9in x 6in B-640 ch09
+∞
1
=√ dw|F (w)|2 eiwζ . (9.22)
2π −∞
with
+∞
1
Rf f (ζ) = √ dw eiwζ Sf f (w). (9.24)
2π −∞
(9.26)
where
1
In,m (t) = 2−m−n (−it)−1−m−n
Γ(m + 1)
∞
Γ(1 + n + m + ) 1
× F (−, −n − , m + 1, 1) − 2 .
!Γ(n + + 1) 4t
=0
(9.28)
√
with, here, the explicitly expressions Tn (t) = 2−n [(t +
√ t2 − 1)n + (t −
t2 − 1)n ].
∞
cos(wt)Jn (w)dw
0
1
(−1)k √ Tn (t), for 0 < t < 1 and n = 2k,
1 − t2
0, for 0 < 1 < t,
= ∞ 4k (9.30)
cos(wt) Jn−1 (w)dw,
w
0 ∞
− cos(wt)Jn−2 (w)dw, for n = 2k + 1,
0
∞
sin(wt)Jn (w)dw
0
1
(−1)k √ Tn (t), for 0 < t < 1 and n = 2k + 1,
− t2
1
0, for 0 < 1 < t,
= ∞
sin(wt)
2(2k + 1) Jn−1 (w)
∞ 0 w
− sin(wt)Jn−2 (w)dw, for n = 2k + 2.
0
(9.31)
∂ 4 W (x, t) ∂ 2 W (x, t)
B + m = P (x, t) + (p1 − p2 )(x, d, t), (9.34)
∂4x ∂2t
with B denoting the bending rigidity, m the mass per unit length of the
beam, and p(x, z, t) the (supersonic) boundary-layer pressure fluctuation of
the exterior medium.
The searched induced pressure p2 (x, z, t) in the fluid interior medium
(0 ≤ z ≤ d) is governed by
∂ 2 p2 ∂2 ∂2
(x, z, t) − (v2 )2 + p2 (x, z, t) = 0, (9.35)
∂2t ∂2x ∂2z
∂p2 (x, d, t) ∂2
= −ρ2 2 W (x, t). (9.36)
∂z ∂ t
The solution of Eqs. (9.32) and (9.33) is straightforward obtained in the
Fourier domain
v1 (w + U∞ k)2
p̃1 (k, z, w) = −ρ1 2 w̃(k, w)
v1 k 2 − (w + U∞ k)2
1
× exp − × v12 k 2 − (w + U∞ k)2 (z − d), (9.37)
v1
1 ∂ k p̃2 (k, w, z)
+ ···+ (z − d)k + · · · . (9.39)
k! ∂kz
The second z-derivative of the interior pressure (and the higher ones!)
is easily obtained recursively from the wave equation (9.35) (k ≥ 0, k ∈ Z+ )
and Eq. (9.40)
∂ 2+k
2+k
p̃2 (k, z, w)
∂ z z=d
1 ∂k
= × (−w2 + (v2 )2 k 2 ) p̃ (k, z, w) . (9.41)
(v2 )2 k
∂ z
2
z=d
References
Lett. B 12, 301 (1998); L. C. L. Botelho et al., Mod. Phys. Lett. B 13, 317
(1999), J. M. F. Bassalo, Nuovo Cimento B 110 (1994).
16. L. C. L. Botelho, Phys. Rev. E 49R, 1003 (1994).
17. L. C. L. Botelho, Mod. Phys. Lett. B 13, 363 (1999).
18. B. Simon, Functional Integration and Quantum Physics (Academic,
New York, 1979).
19. B. Simon and M. Reed, Methods of Modern Mathematical Physics, Vol. 2
(Academic, New York, 1979, ).
20. B. Simon and M. Reed, Methods of Modern Mathematical Physics, Vol. 3
(Academic, New York, 1979).
21. T. Kato, Analysis Workshop (Univ. of California at Berkeley, USA, 1976).
22. Any standard classical book on the subject as for instance
E. C. Titchmarsh, The Theory of Functions, 2nd ed. (Oxford University
Press, 1968); H. M. Edwards, Riemann’s Zeta Function, Pure and Applied
Mathematics (Academic Press, New York, 1974); A. Ivic, The Riemann Zeta-
Function (Wiley, New York, 1986); A. A. Karatsuba and S. M. Doramin, The
Riemann Zeta-Function (Gruyter Expositions in Math, Berlin, 1992).
Hardy Theorem. Let f ∈ L1 ([0, 2π]) and its complex Fourier coefficients
satisfying an estimate as
1
|bn | ≤ O . (A.1)
n
Then, the usual partial sums Sn (f )(x) and σn (f )(x) (Fejer partial sum
of order n) converge to the sum limit.
Theorem A.1. Let f (x) ∈ L1 ([0, 2π]) and thus usual Fourier coefficients
obeying the estimates
ãn
|an | ≤ , (A.2)
n
b̃n
|bn | ≤ , (A.3)
n
where
n
(ãk )2 ≤ Ma (n)2−ε , (A.4)
k=0
n
(b̃k )2 ≤ Mb (n)2−ε , (A.5)
k=0
We have as well as
√
2π
||Sn (f )(x) − f (x)||L1 ≤ (MA + MB )n−ε . (A.7)
π
L1
As a consequence Sn (f ) −→ f (x), so it exists a basic result that there
exists a subsequence Snh (f )(x) converges almost everywhere to f (x) in
[0, 2π].
Theorem A.2. Let f (x) ∈ L2 ([0, 2π]) and suppose that the estimate below
holds true
∞
||f − Sn (f )(x)||2L2 = |b |2 ≤ M · nε−2 , (A.8)
≥n
for M and ε denoting positive real numbers (0 ≤ ε < 2). Then, there
exists a subsequence Snk (f )(x) which converges uniformly to f (x) almost
everywhere.
August 6, 2008 15:46 9in x 6in B-640 ch09
Proof. Let
12
∞
√
an = ||f − Sn (f )||L1 ≤ 2π |b |2 (A.9)
≥n
and
gn (x) = Sn (f )(x) + an .
L1
By the Luzin theorem and the obvious result that an → 0 Sn (f ) −→
f , we have that there is a subsequence Snk (f )(x) almost everywhere con-
verging to f (x) uniformly.
(B) Ut (x, 0) = g(x) with g (2) (x) ∈ L2 ([0, ]) and g(x) being an absolutely
continuous with g (2) (x) piecewise continuous in [0, ] (g() = g(−) =
g (1) () = g (1) (−) = g (2) (−) = g (2) () = 0).
Theorem A.4. Let f (x) ∈ L1loc (R) and f (x) periodic of period T . Let its
formal Fourier complex coefficients cn satisfy the polynomial bound
1 T /2
2πin
|cn | = f (x)e− T x dx ≤ M (n)L ,
T −T /2
On the light of this distributional sense, one can see that it is always
possible
to write a solution of the string linear wave equation in the space
C Dperiodic (R), (0, ∞) . (Exercise for our diligent readers.) For any given
datum f (x) and g(x) in Dperiodic (R).
d dyn
− p(x) + g(x)yn (x) = λn yn (x),
dx dx
(A.17)
a11 yn (a) + a12 yn (a) = 0,
a21 yn (b) + a22 yn (b) = 0,
Here
b
cn = dxf (x)yn (x).
a
B.1. Introduction
Theorem B.1. Let {λn } be a sequence of positive real numbers such that
! 1
λn = +∞. Then, the span of the set {1, f , f , . . . , f
λ1 λ2 λn
, . . .} is the
whole C(∆, R).
then also
(f )n dµ = 0, (n = 1, 2, 3, . . .), (B.2)
∆
August 6, 2008 15:46 9in x 6in B-640 ch09
and (eiθ ) = 0 for eiθ ∈ T . Then, for any Müntz–Szasz sequence |λn |, we
have the result
+∞ +∞
1
iθ
= γn eiλn θ with |γn | < ∞. (B.5)
(e ) −∞ n=−∞
Proof. The element f (θ) = eiθ satisfies the conditions of Theorem B.1.
In the case of the Müntz–Szasz sequence {λn } be the prime numbers
sequence or that one made up by their logarithms, the result may have
interesting applications in the theory of Zeta functions.13
In the case of the existence of a polynomial subalgebra generated by
a finite number of generators {f1 , . . . , fp } we can use the Hahn–Banach
theorem for complex homomorphism of our Banach algebra through the
Gleason, Kahane, Zelazuo theorem (Ref. 12, Theorem 10.9).
Finally in the Banach algebra of the bounded continuous functions
vanishing at the infinity in a locally compact space X, it is straight-
forward to see that if f (x) is an arbitrary continuous function satisfying
August 6, 2008 15:46 9in x 6in B-640 ch09
the condition f (x) ≥ 1 and such that the polynomial algebra generated
by the bounded function h = f exp(−f ) is dense on C0 (X), then the set
{1, hλ1 , hλ2 , . . . , hλn , . . . } remains dense in C0 (X) the only difference with
the standard proof presented in Theorem B.1 is related to the estimate on
the boundedness of the function h(z) on H + , which now takes the following
form:
sup |h(z)| ≤ sup |f e−f |z (m)dµ(m)
z∈H + z∈H +
≤ µ(X) × sup |f |x · e−x|f |(m)
0<x<∞
f (m)∈[1,∞)
x(gf −f )
≤ µ(X) × sup |e |(m)
0<x<∞
f (m)∈[1,∞)
≤ µ(X) × sup |e−x | = µ(X) < ∞.
0<x<∞
C.1. Introduction
Let us, thus, write a path-integral for the associated Schrödinger equa-
tions (C.4)–(C.7) by considering Ḡ(s; (t, t )) in the operator form (the
Feynman–Dirac propagator):
If we substitute Eq. (C.11) into Eq. (C.10) and take the Feynman limit
of N → ∞, we will obtain the following path-integral representation after
evaluating the wi -Gaussian integrals of the representation equation (C.8):
) *
% i 2
2
dt(σ)
Ḡ(s; t, t )) = dt(σ) exp dσ
0<σs
2 0 dσ
t(0)=t :t(s)=t
s
2
× exp i [w0 (1 + g(t(σ)))]dσ . (C.12)
0
The averaged out Eq. (C.7) is thus given straightforwardly by the fol-
lowing Feynman polaron-like path-integral:
∞
2
G(t, t , [g]) g = −i ds(ei(w0 ) s ) DF [t(σ)]
0 t(0)=t :t(s)=t
) *
2
i s dt
× exp dσ
2 0 dσ
s s
4
× exp −w0 dσ dσ K(t(σ); t(σ )) . (C.13)
0 0
s1 s1
× exp −w04 dσ dσ K(t1 (σ) · t1 (σ ))
0 0
s1 s2
+ dσ dσ (K(t1 (σ) · t2 (σ )) + K(t2 (σ) · t1 (σ )))
0 0
s1 s2
+ dσ dσ K(t2 (σ) · t2 (σ )) . (C.14)
0 0
Similar N -iterated path-integral expressions hold true for the N -point
correlation function x(t1 , [g]) · · · x(tN , [g]) g . Explicit and approximate
evaluations of the path-integral equations (C.13) and (C.14) follow pro-
cedures similar to those used in the usual contexts of physics statistics,
quantum mechanics, and random wave propagation (last reference of
Ref. 14).
Let us show such exact integral representation for Eq. (C.13) in the case
of the practical case of a slowly varying (even function) kernel of the form:
1/2
0 2 K(0)
K(t) ∼ K(0) − |t| |t| + = L,
2 0 (C.15)
K(t) ∼ 0 |t| L.
In this case, we have the following exact result for the path-integral in
Eq. (C.13):
G(t, t , [g]) g
∞ ) *
−s(−iw02 ) −(u40 K(0))s2 1/2 w02 (− 2 )1/2 S 3/2
= dse e (2πis)
0 sin[s3/2 w02 (− 20 )1/2 ]
") " #* #
1/2 1/2
iw02 1/2 0 0
× exp s − cot w02 − s3/2 (t − t )2 .
2 2 2
(C.16)
Another useful formula is that related to the “mean-field” averaged
path-integral when the kernel K(t, t ) has a Fourier transform of the general
form:
+∞
1
K(t, t ) = √ dp · eip(t−t ) K̃(p). (C.17)
2π −∞
The envisaged integral representation for Eq. (C.13) is, thus, given by
∞ +∞
isu20 w04
G(t, t , [g]) g = −i dse exp − √ K̃(p) × M (p, s, t, t ) ,
0 2π −∞
(C.18)
August 6, 2008 15:46 9in x 6in B-640 ch09
where
s s 2
i s
F dt
M (p, s, t, k ) = dσ dσ D [t(σ)] exp
0 0 t(0)=t 2 0 dσ
t(s)=t
× exp[ip(t(σ) − t(σ ))] . (C.19)
where we have used the fact that the Jacobian associated with the functional
change g(t) → y(t) is unity:
d δg(t)
detF + 2y = = 1. (C.28)
dt δy(t)
2
∞
γ dy
Z[j(t)] = D [y(t)] exp −
F
dt + y 2 (t) + wo2
2(w0 )4 0 dt
∞
× exp i dty(t)j(t) , (C.30)
0
August 6, 2008 15:46 9in x 6in B-640 ch09
and 0 < a < 1 and b arbitrary, one can apply the Kato–Rellich theorem19
d2 2 2
to be sure that the domain of the differential operator − dt 2 + w0 + w0 g(t) is
2
(at least) contained on the domain of − dt d
2 which in turn has a purely con-
2
tinuous spectrum on L (R). As a consequence of the above-exposed remarks
one does not have bound states on the Schrödinger operator spectrum of
Eq. (C.4) for each realization of g(t) on the above-cited functional class. As
a consequence, we have that the evolution operator
d2
exp is − 2 + w02 (1 + g(t)) , (C.33)
dt
2
d
is a unitary operator on Dom(− dt 2 ) and Ḡ(s, t, t ) in turn is expected to
d2
have the same behavior of exp(is[− dt2 ]) (asymptotic completeness20 ) at
s → ∞, which in turn vanishes at s → ∞, making our condition Eq. (C.6)
highly reasonable from a mathematical physicist’s point of view.
At this point, we remark that the different problem of determining the
quantum propagator of a particle under the influence of a harmonic potential
V (x) = 12 w02 x2 and a stochastic potential g(x) may be possible to handle
August 6, 2008 15:46 9in x 6in B-640 ch09
2 2
in our framework. Note that the “únperturbed” Hamiltonian − 2mdx h d
2 +
1 2 2
2 mw 0 x has a pure point spectrum (bound states) and its perturbation by
2
a g(x) ∈ L (R) potential does not alter the spectrum behavior. However,
one can proceed in a mathematical physicist’s (formal) way as exposed in
our letter and see that Eq. (C.3) adapted to this case is still (formally)
correct. Namely
t ) *
% i dx(σ)
2
G(t, (x, x )) g =
dx(σ) exp m dσ
0<σ<t
2 0 dσ
x(0)=x :x(t)=x
t
i 2 2
× exp mw0 dσ[x(σ)]
2 0
t t
4
× exp −w0 dσ dσ K(x(σ); x(σ)) . (C.34)
0 0
Finally, we want to point out that in the general case where k(t, t ) is not
defined on an operator of the trace class, the realizations g(t) will be dis-
tributional objects. Unfortunately in this case there is no rigorous spectral
perturbation mathematics for differential operators acting on nuclear spaces
(s (R), D (R)), etc, which is the natural mathematical setting to under-
stand Eq. (C.1) of this letter.
In this complementary paragraph, we complete our study by considering
the problem of a harmonic oscillator in the presence of a damping term
(0 < t < ∞)
2
d d 2
+ v + w0 (1 + g(t)) x(t) = F (t). (C.35)
dt2 dt
In order to map the above differential equation in the analysis pre-
sented in Sec. C.2, we implement in Eq. (C.35) the following time-variable
change:16
m v
ζ = (1 − e−t m u ),
v (C.36)
y(ζ) = x(t).
We obtain, thus, the following pure harmonic oscillator differential
equation without the damping term in place of the original equation (C.35),
namely
2
d 2
+ w0 [1 + g̃(ζ)] y(ζ) = F̃ (ζ). (C.37)
dζ 2
August 6, 2008 15:46 9in x 6in B-640 ch09
Here
m v
g̃(ζ) = g − lg 1 − ζ , (C.38)
v m
m v
− v lg 1 − m (ζ)
F̃ (ζ) = F v 2 , (C.39)
1− m ζ
and the correlator stochastic frequency
" ) *#
m (1 − ( m
v
)ζ)
K(ζ, ζ ) = K((ζ, ζ )) = K lg v . (C.40)
v 1− m ζ
From here on the analysis goes as in the bulk of this appendix under
the condition that the range associated with this new time variable is the
finite interval [0, m
v ].
here Bn are the Bernoulli’s numbers and C is any contour in the complex
plane, coming from positive infinity and encircling the origin once in the
positive direction.
August 6, 2008 15:46 9in x 6in B-640 ch09
ζ(z) πz
= 2z · π z−1 · sin · Γ(1 − z). (D.2)
ζ(1 − z) 2
Conjecture D.1. In each horizontal line of the complex plane of the form
Imaginary (z) = y = b = constant, the zeta function ζ(z) possess at most a
unique zero.
Proof. Let us consider a given nontrivial zero z̄ = x̄ + iȳ on the open strip
0 < x < 1, −∞ < y < +∞. It is a direct consequence of the Schwartz’s
reflection principle since ζ(x) is a real function in 0 < x < 1 that (z̄)∗ = x̄−
iȳ is another nontrivial zero of the Riemann function on the above pointed
out open strip. The basic point of our proof is to show that 1 − (z̄)∗ =
(1 − x̄) + iȳ is another zero of ζ(z) in the same horizontal line Im(z) = ȳ.
This result turns out that (1 − (z̄)∗ ) = z̄ on the basis of the validity of our
conjecture. As a consequence, we obtain straightforwardly that 1 − x̄ = x̄.
In others words, Real (z̄) = 12 .
August 6, 2008 15:46 9in x 6in B-640 ch09
1 d
ζ(x + ib) = exp ib (ζ(x)Γ(x) . (D.3)
Γ(x) dx
As a consequence, one can easily see that if ζ (x + ib) (the zeta function
derivative on the horizontal line Im(z) = b = constant) does not possesses
zeros between two consecutive zeros of the zeta function ζ(x + ib) in 0 <
x < 1, then the Rieman’s conjecture is proved.
∞
ζ (n) (t)
ζ(x̄) = 0 = (x̄ − t)n > 0, (D.4)
n=1
n!
here x̄ is the supposed different zero of ζ(x) on the open interval (0, 1).
Finally, let us point out the following formula of ours, related to the
zeros of the Riemann zeta function on the complex plane C.
August 6, 2008 15:46 9in x 6in B-640 ch09
Lemma D.1. Let {zn }n=0,1,2,... denote the zeros of the Riemann’s zeta
function ζ(z). We have thus the following result, for any integer p > 0.
∞
1
−
n=1 (zn − zk )
p+1
n=k
2π
1 1 ζ (z)
= lim dθ + (zk + eiθ ) · (zk − z)p+1 e−ipθ .
z→zk 2π 0 z−1 ζ(z)
(D.5)
Proof. This result can be seen from the fact that h(z) = (z − 1)ζ(z) is
an integral function and from any integral function of order k, with zeros
z̄1 , z̄2 , . . . , (h(0) = 0)
∞
h (z)
p
d 1
lim (z − z̄k ) p+1
= −p! . (D.6)
z→zk dz h(z) (z̄n − z̄k )p+1
n=1
n=k
1 i(i − w)
h lg = h(g(w)) = h̄(w). (D.8)
πi i+w
where η(x) is the number of zeros of the zeta function on the region
g −1 (|w| < x), contained on the strip 0 < Real(z) < 1.
Proof. Since h(z) is an analytic function in the strip 0 < Real(z) < 1, we
are going to prove the general result for a region Ω conformally equivalent
to a disc of radius 1, including its boundary correspondence, namely
g : Ω → D(0, 1),
(D.10)
g(∂Ω) = ∂D(0, 1),
Since η(x) coincides with the number of zeros of the function f (z) in the
region g −1 (D(0, 1)), we obtain the Jensen’s formula in Ω in the general case
r
η(x) 1 g (z)
dx = lg |l(z)| · dz − lg|f (0)|. (D.12)
0 x 2πi ∂Ω g(z)
Equation (D.8) comes from the fact that the strip 0 < Real(z) < 1 is
conformally mapped — with the boundary correspondence — on the unit
1
circle {w|j | |w| ≤ 1} by the function z = πi lg( i(i−w)
i+h ).
August 6, 2008 15:46 9in x 6in B-640 ch10
Chapter 10
10.1. Introduction
287
August 6, 2008 15:46 9in x 6in B-640 ch10
condition
Tr
∂ ∂ i
vi − ν∆x vi + vk v = Fi , (10.1a)
∂τ ∂xk
vi (x, 0) = ϕi (x), (10.1b)
ϕi1 (x1 )ϕi2 (x2 ) = λδ (3) (x1 − x2 )δi1 i2 . (10.1c)
Our task, now, is to compute the ϕ-averge of the N -point fluid velocity
field equation (10.1a) for an arbitrary space-time points, by means of a
functional integral representation for the characteristic functional of the
random fluid velocity fields Z[Ji (x, ξ)]; namely3
The functional measure dµ[Vi ] in Eq. (10.3b) is defined over the func-
tional space M of all possible realizations of the random fluid motion defined
by Eq. (10.1). An explicit (formal) expression for the above functional
measure should be given by the product of the usual Feynman measure
weighted by a certain functional S[Vi ] to be determined,
with
∞
Ai [v] = Vi (x, ξ) − ds dyOijk (x − y, ξ − s)(Vj Vk )(y, s)
0 R3
∞
− ds dyH(1) (x − y), (ξ − s)Fi (y, s) (10.6a)
0 R3
Bi [ϕ] = dyH(0) (x − y, ξ)ϕi (y). (10.6b)
R3
Let us now introduce the following functional representation for the gener-
ating functional Z[Ji (x, ξ)]4,5
Z[Ji (x, ξ)] = DF [V i ]δ (F ) (V i − Ṽ i [ϕ])ϕ
∞
× exp − dx dξ(V i · J i )(x, ξ) , (10.8)
R3 0
Tr
∂ ∂ i Tr
× − ν∆x v + vk
i
v + Fi (x, ξ)
∂ξ ∂xk
Tr
∂ ∂ i
× − ν∆x v + vk
i
v + Fi (x, ξ ).
Tr
(10.20)
∂ξ ∂xk
We obtain, thus, our proposed functional integral representation for
Eqs. (10.1a)–(10.1c)
∞
Z[Ji (x, ξ)] = D [v ] exp(−S[v ]) exp −
F i i
dx dξ(Ji · Vi )(x, ξ) .
R3 0
(10.21)
The above written functional integral is the main result of this section.
A perturbative analysis for Eq. (10.21) may be implemented by using
the free propagator Eq. (10.16) in the context of a background field decom-
position Vi = V̄i + βViq where V̄i satisfies the non-random Navier–Stokes
equation
Tr
∂ ∂
V̄i = ν∆x V̄ − V̄k
i
V̄ i
+ F Tr , (10.22a)
∂ξ ∂xk
with
V̄ i (x, 0) ≡ 0, (10.22b)
with β being a coupling constant (β 1). It is worth remarking that the
cross term
∞
dx dξ(∂i v i ∆x vi )(x, ξ) (10.23a)
R3 0
i
vanishes in S[v ] as a result of the boundary condition
vi (x, 0) = vi (x, ∞) = 0 (10.23b)
for the pure random diffusion free propagator (Eq. (10.16)).
Finally, we point out that our proposed functional integral
equation (10.21) differs from that proposed in Ref. 7.
(r)
= λ2 (δ δ jj − δ j δ j )∂j ∂j (∆−1
r δ
(3)
(r − r ) × δ(t − t ))
(r) (r )
= λ2 δ δ (3) (r − r )δ(t − t ) − λ2 ∂ ∂ (∆−1
r δ(r − r ))δ(t − t ).
(10.28)
(10.30a)
August 6, 2008 15:46 9in x 6in B-640 ch10
equation (10.27) in order to see the connection with strings (random sur-
faces) (see Appendix A for the relevance of theses non-local objects for
advection phenomema)
exp i v((σ), t)d(σ)
v
2
1 +∞ 3 +∞
∂v
≡ DF [v(r, t)] exp − d r dt + νλ2 v (r, t)
2 −∞ 0 ∂t
× exp i v((σ), t)d(σ) . (10.34)
Since the flux is of a Beltrami type in our soluble model equation (10.29),
we propose to rewrite the circulation phase factor as a sum over all surfaces
bounding the fixed loop by making use of Stokes theorem and by taking
into account again the Beltrami condition, i.e.
+∞
H
v·d 1 +∞ 3 ∂
ei c = DF [v(r, t)] exp − d r dt v − 2 + νλ2 v
2 −∞ 0 ∂t
× exp iλ v(x, t) · dA(x) . (10.35)
S S
with t, t ∈ [0, ∞] and θ(0) = 1/2 in this initial value problem, we can easily
evaluate the average equation (10.35) and producing a strongly coupled area
dependent functional for the spatial vortex phase factor in our proposed
turbulent flux regime
H
W [, v] ≡ ei v(,t)·d
λ sin(λ|x − y|)
= exp − dA(y) dA(y) . (10.37)
ν S |x − y|
{S}
Just for completeness of our study and in order to generalize the Beltrami
flux turbulence analysis represented in the main text, for the physical case
August 6, 2008 15:46 9in x 6in B-640 ch10
of the complete wave vector range 0 ≤ |k| < ∞ in our turbulent path-
integral soluble model studies, we propose to consider a kind of generalized
Beltrami condition to overcome this possible drawback of our turbulence
modeling, namely:
rot v(r, t) = λ(r)v(r, t), (10.38)
where λ(r) is a positive function varying in the space and to be determined
from a phenomenological point of view. Note that the Fourier transformed
(wave vector) condition now takes the general form
|k| · |ṽ(k, t)| = d3 p|λ̃(p − q)| · |ṽ(p, t)| (10.39)
R3
which, by its turn, leads to the full range scale 0 < |k| < ∞ for the eddies
hydrodynamical motions under study. By supposing that the “vortical”
stirring equation (10.28) is a pure white noise process with strength D,
Ωext ext 3
(r, t)Ω (r, t ) = D · δ δ (r − r )δ(t − t ). (10.40)
It is a straightforward deduction by following our procedures as exposed
in the text to arrive at an analogous Gaussian path-integral for the gener-
alized Beltrami random hydrodynamical defined by Eq. (10.38). The gen-
eralized effective motion equation is given, in this new situation, by
∂ ∆λ ν ∂λ(r) ∂
−ν (r) − + νλ2 (r) δ ik
∂t λ λ(r) ∂xe ∂xe
∂λ(r)
+ ν εijk vk (r, t) = Ωext
i (r, t). (10.41)
∂xj
The Gaussian path-integral, thus, is exactly written as
3 +∞ ∞
3
Z[ji (r, t)] = F
D [vi (r, t)]exp i d r i
dt(j vi )(r, t)
i=1 ∞ 0
+∞ ∞
1
× exp − d3 r dtvk (r, t)(M ∗ki · M is )vs (r, t) .
2D −∞ 0
(10.42)
Here, the differential operators entering in the kinetic term of the tur-
bulent path-integral are
∂ ν ∂λ(r) ∂ ν
M ∗ki = − + + · ∆λr
∂t λ(r) ∂xe ∂xe λ(r)
∆λ(r) ∂λ(r)
−ν + νλ2 (r) δ ki + νεkji (10.43)
λ(r) ∂xj
August 6, 2008 15:46 9in x 6in B-640 ch10
and
∂ ν ∂λ(r) ∂ 2 ∆λ(r) ∂λ(r)
M is
= + − e
+ νλ (r) − ν δ is + νεijs .
∂t λ(r) ∂xe ∂x λ(r) ∂xj
(10.44)
where we have replaced the usual fluid viscosity term νd2 v(x, t)/dx2 by
the pure damping term −νλ2 v(x, t) (the reader should compare our pro-
posed Brownian-like flux with that of the Navier–Stokes–Beltrami studied
in Sec. 10.3).
One of the most important observable objects in fluid turbulence is the
fixed velocities measurements at the grid points (x ) and at a common
observation time t
N
δ(v(x ; t) − v ) (10.48)
=1
The Hopf wave equation associated with our model equation (10.48) is
given in a closed form by applying straightforwardly the methods of Ref. 10
∂
−i ψ((x1 , . . . , xN ); (p1 , . . . , pN ); t)
∂t
N
N
∂ 1 ∂ ∂
= p − (νλ2 )p + (k(x − x )p p
∂p p ∂x ∂p
=1 =1, =1
(a)
Let us remark that in the physical grid on R3 = {xk ; a = 1, 2, 3; k =
1, . . . , N }, Eq. (10.51) naturally reads
∂ (a) (a) (a) (a)
−i ((x , . . . , xN ); (p1 , . . . , pN ); t)
∂t 1
3
N
(a) ∂ 1 ∂ 2 (a) ∂
× p (a) (a) (a)
− νλ p (a)
=1 a=1 ∂p p ∂x ∂p
(a) (a) (a)
+ ψ((x1 , . . . , xN ); (p1 , . . . , pN ); t)
N
(a) (a) (a) (b)
+ Kab (x − x )p p
=1, =1
Hereafter as said earlier we will present our study of Eq. (10.53) for the one-
dimensional case equation (10.51). By introducing the mixed coordinates
defined by the transformation law.
pj + xj = uj ; pj − xj = vj . (10.54)
The turbulent wave equation (10. 51) takes the more invariant form similar
to a many-particle Schrödinger equation in quantum mechanics.
∂
−i (ψ(u1 , . . . , uN ); (v1 , . . . vN ); t)
∂t
N
1 ∂2 ∂2 2 ∂ ∂
= − − −
4 ∂ 2 u ∂ 2 v u + v ∂u ∂v
=1
νλ2 ∂ ∂
− (u + v ) + ψ((u1 , . . . , uN ); (v1 , . . . , vN ); t)
2 ∂u ∂v
1
N
u − u + (v − v )
+ (u + v )(u + v )K
4 1
2
=1; =1
and
N
u + v u − v
ψ((u1 , . . . , uN ); (v1 , . . . vN ); 0) = exp i g
2 2
=1
(10.56)
The above written closed partial differential equation is the basic result of
this section. At this point we can implement perturbative calculations for
our turbulent wave equation by considering a physical slowly varying (even
function) correlation function of the form
1/2
κ0 2 K(0)
K(x) ≈ K(0) − x ; |x| ≡ L,
2 κ0
(10.57)
0; |x| L,
which by its turn leads to the harmonic and quartic anharmonic potential
which is written as
N
K(0)(u u + v v + u v + v u )
=1; =1
1
− κ0 (u + v )(u + v [u2 + u2 + v2 + v2 − 2u u − 2v v ]) .
8
(10.58)
In the important case of the single fluid velocity average, our turbulent wave
equation takes the following form, after making an analytic continuation
v → iv; namely
∂
i ψ(u, v; t) = (L0 + L1 )ψ(u, v; t), (10.59)
∂t
with the initial condition
+ u + iv u − iv
ψ(u, v; t → 0 ) = exp i g . (10.60)
2 2
Here the kinetic and perturbation terms are
1 ∂2 ∂2 k(0) 2
L0 = − + 2 + (u − v 2 ),
4 ∂u2 ∂v 4
(10.61)
2 ∂ 1 ∂ 2 ∂ 1 ∂
L1 = − − νλ (u + iv) + .
u + iv ∂u i ∂v ∂u i ∂v
August 6, 2008 15:46 9in x 6in B-640 ch10
N
νλ2 (I) () ()
(I) () () ()
− (u + iv ) ipI + qI − exp ipI (uI − uI−1 )
2
=1
N
()
× exp iqI (vI − vI−1
−1
) , (10.64)
=1
and evaluate the (pI , qI )-momenta functional integrals (see Ref. 11 for a
detailed exposition), we get our searched trajectory path-integral represen-
tation for the Green-function of Eqs. (10.56)–(10.65) in the free case k ≡ 0
N
d () 2
−2 Ū (σ) − − (νλ2 )(Ū (σ) + iV̄ (σ))
∂σ Ū (σ) + iV̄ (σ)
=1
N 2
2
+ − − (νλ2 )(Ū (σ) + iV̄ (σ))
=1
Ū (σ) + i V̄ (σ)
2 N
i t ∂ ∂
× exp dσ V̄ (σ) − 2 V̄ (σ)
4 0 ∂σ ∂σ
=1
2
2·i
× − + i(νλ2 )(Ū (σ) + iV̄ (σ))
U (σ) + iV̄ (σ)
N
2i 2 ()
× − + i(νλ )(Ū (σ) + iVI (σ)) . (10.66)
=1
Ū (σ) + iV̄ (σ)
References
T x̂ = x̂. (A.2)
(b) For any x̄ ∈ X, the sequence (xn ) defined by the recurrrence relationship
xn+1 = T n x̄ (A.3)
Ln−1
d(xn , x̂) ≤ d(x̄, T x̄). (A.4)
1−L
Namely,
t
U (t) = U0 + F (s, U (s))ds = F̂ (U )(t). (A.9)
0
Let us now choose the global time evolution T in such a way that
or
K(R) − R
T ≤ . (A.12b)
||F (0)||E + 2K(R)L(KR )
We have the following additional estimate:
t
d(F̂ U, F̂ V ) = sup ds||F (s, U (s)) − F (s, V (s))||E
0≤t≤T 0
If one chooses the global time satisfying again the bound for any δ > 1.
1
T ≤ (A.14)
L(KR )(1 + δ)
we have a unique solution of Eq. (A.7) in E(T,R) . For instance, if F (0) = 0
and K(R) = R + b, the compatibility of Eqs. (A.12b) and (A.14) means the
existence of b > 0 such that b < 2R/(δ − 1).
Related to the material exposed in this chapter, let us apply the
Banach–Picard fix point theorem to a Burger equation with a viscosity
time-dependent parameter and zero initial conditions in R × R+
t +∞
1
U (x, t) = dτ dy
ν 0 −∞
∂ (x − y )2 1 def
× − exp − U (y , τ )2 · τ (− 2 +β) ≡ F (u)
∂y 4τ
(A.15)
E(T,R) = {U (·, t) ∈ C([0, T ], R); sup (||U (·, t)||L∞ (R) ) ≤ R}. (A.16)
0≤t≤T
August 6, 2008 15:46 9in x 6in B-640 ch10
R2 c 1 1 ν(2β + 1)
≤ T β+ 2 < R ⇔ T β+ 2 < (A.17)
ν(β + 12 ) 2cR
Let us further make a choice of the global time T in such a way that F̂
is a contractive application in E(T,R) . In order to show such a result, let us
consider the estimate below
. /
sup ||(T U )(x, t) − (T V )(x, t)||L∞ (R)
0≤t≤T
t +∞ e−|x−y|2 · 2|x − y| · ζ β
1
≤ sup dζ
ν 0≤t≤T 0 −∞ 4·ζ 1/2 ·ζ
We have thus the existence and unicity for the Burger equation in the
intergral form (A.15) for global times satisfying Eq. (A.19).
By the Sobolev theorem which means that the embeeded below is con-
tinuous (with Ω ⊆ Rν denoting a smooth domain), one can further reduce
the measure support to the Hölder α continuous function in Ω if 2m− ν2 > α.
Namely, we have a easy proof of the famous Wiener theorem on sample
continuity of certain path-integrals in Sobolev spaces
Chapter 11
where α = (α1 , α2 ) are non-negative integers associated with the basic self-
adjoint differential operators
α1 α2
1 ∂ 1 ∂
Dxα = , (11.2)
i ∂x1 i ∂x2
with Aα (x) ∈ Cc∞ (R2 )q×q .
∗ Author’s original results.
312
August 6, 2008 15:46 9in x 6in B-640 ch11
and for j = 2,
A2 (x, ξ, λ) = −λ11 q×q + Aα (x)(ξ1 )α1 (ξ2 )α2 . (11.5b)
|α|=α1 +α2 =2
1
(Af )(x) = d2 ξeixξ [σ(A)(x, ξ)]F (ξ). (11.6b)
(2π)2 R2
Here
α! = α1 ! · α2 !. (11.12)
we have that
or in an equivalent way
1
∞
[(Dξα σ(A − λ11)(x, ξ))iDξα (C−2−j (x, ξ))] = 1. (11.16)
α!
|α|≤2
j=0
and
1
C−2−j (x, tξ , ((tλ ) 2 )2 ) = t−(2+j) C−2−j (x, ξ , λ ). (11.18)
After substituting Eq. (11.18) into Eq. (11.16) and by comparing the
resulting power series in the scale factor 1/t, one obtains the famous Seeley
recurrence equations which for t → 1, give us explicitly expressions of all
Seeley coefficients {C(−2+j) (x, ξ)} of the Green function of the operator A
(note that they are elements of Mq×q (C))
where {g11 (x), g22 (x); x ∈ R2 } are positive definite functions in Cc∞ (R2 )q×q ,
we can evaluate exactly the Seeley relationship (Eq. (11.20)).
August 6, 2008 15:46 9in x 6in B-640 ch11
∞ 2 −∞
1 1
= d(is)eist d2 xd3 xC−2−j (x, ξ, −is)
j=0
2π 2πi +∞ R2 ×R2
∞
+∞
1 1 2 2 is −is
= d ξ d x e C−2−j x, ξ, ds
j=0
t (2π)3 R2 R2 −∞ t
∞
1 1 +∞
2+ξ
=− d2 xd2 ξ eis (t)( 2 ) C−2−j (x, t1/2 ξ, −is)
t (2π)3 R2 ×R2 −∞
j=0
∞
+∞
(ξ−2)
= t 2 (2π)3 d2 ξ d2 x dsC−2−j (x, ξ, −is) . (11.23)
j=0 R2 R2 −∞
August 6, 2008 15:46 9in x 6in B-640 ch11
After substituting Eq. (11.22) (together with the term C−4 (x, ξ, λ) not
written here), we have the Seeley short-time expansion for the Heat–Kernel
evolution operator associated with our Laplacean A given by Eq. (11.21):
For each integer j, let us introduce Hilbert spaces H1 and H̄j defined as5
vectorial set of all complex valued functions which under the
action of a complex coordinate transformation z = z(w) of W ,
Hj =
has the tensorial like transformation law
f (z, z̄) = (∂w/∂z)−j f˜(w, w̄).
(11.26)
where ρ(z, z̄) denotes a real continuous function in W and of the prove-
nance of a conformal metric structure in a given Riemann surface with local
chart W
As an exercise for our readers we have that the above inner products are
invariant under the action of the conformal (complex) coordinate transfor-
mations of the open set W .
Let us now introduce the following self-adjoint operators in L2 (W̄ ):
Lj : Hj −→ H̄−(j+1)
(11.30)
f −→ ρj ∂z̄ f
L∗j : H̄−(j+1) −→ H( j
(11.31)
f −→ ρ−(j+1) ∂z f.
Lj : L∗j Lj : Hj −→ Hj , (11.32a)
L∗j : Lj L+
j : H̄−(j+1) −→ H̄−(j+1) . (11.32b)
By using now the Seeley asymptotic expansion in Cc∞ (R2 ), we have the
following t → 0+ expansions for the heat kernels associated with the Dirac
operator Eq. (11.33)5 :
dz ∧ dz̄ ρ(z, z̄) (1 + 3j)
= − (∆gρ)(z, z̄) + O(t) (11.34a)
W 2i 2πt 12π
lim TrCc∞ (R2 )2×2 (exp(−tL∗j Lj ))
t→0+
dz ∧ dz̄ ρ(z, z̄) (2 + 3j)
= + (∆gρ)(z, z̄) + O(t). (11.34b)
W 2i 2πt 12π
Now the famous topological heat kernel index can be obtained easily
through the Atiyah–Singer definition:
Here, the curvature of the Riemann surface M with metric ds2 = ρ(z, z̄)dz
∧ dz̄ is given by
1
R(z, z̄) = ∆gρ (z, z̄), (11.36)
ρ
References
∂
hµν = (R0 − R)hµν , (A.1)
∂t
The matrices γ µ = êµa γa are the (Euclidean) Dirac matrices associated with
the metric Riemann surface structure ĥµν = êµa êνa . Note that ĥµν can always
be written in the canonical conformal form (ζ ∈ M)
1 µν
ĥµν (ζ) = h̃ (ζ), (A.3)
ρ(ζ)
εµν
Aµ = − √ ∂µ φ + AH
µ (A.4)
h
with
∇µ (Aµ − AH
µ ) ≡ 0. (A.5)
It is useful to remark that the effects of the nontrivial topology of the genus
of Riemann surface M reflects itself in the Hodge harmonic term AH µ of the
U (1)-Connection through the Abelian differential forms (and its respective
complex conjugates)7
g
AH
µ = 2h (p αµ + g βµ ) , (A.6a)
=1
1 dz ∧ dz̄ zz̄
= [h̃ (∂z ϕ∂z̄ ϕ)(z, z̄)]
π M 2i
1 dz ∧ dz̄ zz̄
+ h̃ [(gρ)t (gρ)zz̄ ]. (A.8)
12π M 2i
Note that this metric variation is evaluated for the normalized Ricci flux
below
∂
gρ ≡ [gρ]t = R0 − R, (A.9a)
∂t
2π(2 − g)
R0 = , (A.9b)
area(M)
1
R = − ∂zz̄ (gρ). (A.9c)
ρ
As a consequence we have the positivity of Eq. (A.8):
∂ det D(A, ĥ)
g ≥ 0. (A.10)
∂t area(M) det(D(A = 0, h̃))
At this point we remark that Eq. (A.10) vanishes solely for all those metric
of constant curvature, which at the asymptotic limit t → ∞ leads us to the
usual result (R → R0 )
√ t→∞
(R0 − R)R · h dz ∧ dz̄ −→ 0,
M
Index
323
August 6, 2008 15:47 9in x 6in B-640 index