Application of canonical Hamiltonian formulation to nonlinear light-envelope propagations

0.17 MB·

by  Guo Liang

.
Application of canonical Hamiltonian formulation to nonlinear
light-envelope propagations
Guo Liang and Qi Guo
∗
Laboratory of Nanophotonic Functional Materials and Devices,
4
South China Normal University, Guangzhou 510631,China
1
0
(Dated: January 7, 2014)
2

n
Abstract
a
J

4 We first point out it is conditional to apply the variational approach to the nonlocal nonlinear

] Schro¨dinger equation (NNLSE), that is, the response function must be an even function. Different
S
P
fromthevariationalapproach, thecanonical Hamiltonian formulation forthefirst-orderdifferential
.
n
i system are used to deal with the problems of the nonlinear light-envelope propagations. The
l
n
[ Hamiltonian of the system modeled by the NNLSE is obtained, which can be expressed as the sum

1
of thegeneralized kinetic energy and the generalized potential. Thesolitons correspondto extreme
v
4
points of the generalized potential. The stabilities of solitons in both local and nonlocal nonlinear
1
8
0 media are also investigated by the analysis of the generalized potential. They are stable when the
.
1
potential has minimum, and unstable otherwise.
0
4
1
: PACS numbers: 42.65.Tg;42.65.Jx;42.70.Nq
v
i
X
r
a
∗
Electronic address: [email protected]
1
I. INTRODUCTION
TheHamiltonianviewpoint provides aframeworkfortheoretical extensions inmanyareas
of physics [1–5]. In classical mechanics it forms the basis for further developments, such as
Hamilton-Jacobi theory, perturbation approaches and chaos. The canonical equations of
Hamilton in classical mechanics are of the form
∂H ∂H
q˙ = , p˙ = , (i = 1, ,n), (1)
i i
∂p − ∂q ···
i i
where q and p are said to be the generalized coordinate and the generalized momentum,
i i
q˙ = dq /dt, p˙ = dp /dt, and H is the Hamiltonian. There are many situations in which
i i i i
the Hamiltonian is equal to the sum of the generalized kinetic energy T and the generalized
potential V. The conditions are that the generalized potential is not a function of general-
ized velocities, and the generalized kinetic energy is a homogeneous quadratic function of
generalized velocities. Once the generalized potential V, which is under the framework of
the Hamiltonian system, is obtained, the problem of the small oscillations of a system about
positions of equilibrium can be easily dealt with. For a conservative mechanical system its
equilibrium state canbeobtained by ∂V = 0.The generalized potential hasanextremum
(cid:16)∂qi(cid:17)0
at the equilibrium configuration of the system, which is marked with the subscript 0. The
equilibrium is stable when the extremum of the potential V is the minimum, and unstable
otherwise.
Up to now, to our knowledge, the canonical equations of Hamilton appearing in all the
literatures are of the form (1) except for our recent work, where it is pointed out that the
canonical equations of Hamilton (1) are only valid for the second-order differential system
(the system described by the second-order partial differential equation about the evolution
coordinate) but not valid for the first-order differential system (the system described by
the first-order partial differential equation about the evolution coordinate). The nonlin-
ear Schro¨dinger equation (NLSE) is the first-order differential system, which is a universal
nonlinear model that describes many nonlinear physical systems and can be applied to hy-
drodynamics [6], nonlinear optics [7], nonlinear acoustics [8], Bose-Einstein condensates [9],
andsoon. TheworkaboutthecanonicalequationsofHamiltonforthefirst-orderdifferential
system will be introduced briefly here.
The approximate analytical solutions of the NLSE can be obtained by the variational
approach [7, 10–12], where the light-envelope is treated as the classical particle traveling in
2
an equivalent potential, whose minimum corresponds to the soliton. In this paper we use
the canonical equations of Hamilton to deal with the nonlinear light-envelope propagations.
Such an approach is different from the the variational approach, which will be illustrated in
the paper. We can divide the Hamiltonian of the system into the generalized kinetic energy
and the generalized potential, the extreme point of which corresponds to the soliton. But in
some other literatures [13–16], solitons are regarded as the extrema of the Hamiltonian of
the system. Such a treatment has some problems, which will be illustrated in the paper. To
determine the stabilities of the soliton, we can determine whether the generalized potential
has a minimum. Solitons are stable when the generalized potential has a minimum, but
unstable otherwise. In fact, the similar expression, the Hamiltonian expressed as the sum
of the generalized kinetic energy and the generalized potential, appeared in Ref.[17], but
the elaboration of the systematic theoretical principle was absent, which is often of great
importance.
The paper is organized as follows. In Sec. II, we briefly introduce the model, the nonlocal
nonlinear Schro¨dinger equation (NNLSE). The restriction on the the response function in
the the variational approach is discussed in Sec. III, we point out the variational approach
can be used to find the approximately analytical solution of the NNLSE if and only if
the response function is an even function. We use the canonical equations to deal with the
nonlinear light-envelope propagationsin the paper, but theconventional canonical equations
of Hamilton are not valid for NNLSE, so in Sec. IV we will briefly introduce the canonical
equations of Hamilton valid for the first-order differential system, which will be reported
elsewhere in detail. The application of the canonical equations of Hamilton introduced in
Sec. IV to the NNLSE is shown in Sec. V. By use of the canonical equations of Hamilton,
we can find the soliton solutions of the NNLSE, and can analyze the stability characteristics
of the solitons. The difference between the variational approach and the approach employed
in the paper is discussed in Sec. VI. Sec. VII gives the summary.
3
II. MODEL
Thepropagationofthelight-envelope inthenonlocalcubicnonlinear media ismodeledby
the nonlocal nonlinear Schro¨dinger equation (NNLSE) in the dimensionless system [18–21]
∂ϕ
i +∆ ϕ+ϕ ∞ R(r r) ϕ(r,z) 2dDr = 0, (2)
′ ′ ′
∂z ⊥ Z − | |
−∞
where ϕ(r,z) is the complex amplitude envelop, z is the longitudinal coordinate, r and r are
′
the D-dimensional transverse coordinate vectors. dDr is a D-dimensional volume element
′
at r, ∆ is the D-dimensional transverse Laplacian operator, and R is normalized response
′
⊥
functionofthemediasuch that ∞ R(r′)dDr′ = 1. Forasingularresponse, i.e., R(r) = δ(r),
R−∞
Eq.(2) simplifies to the NLSE
∂ϕ
i +∆ ϕ+ ϕ 2ϕ = 0. (3)
∂z ⊥ | |
When D = 1, NNLSE (2) can describe the propagations of both optical beams [18–20]
and pulses [21]. Particularly, it was predicted very recently [21] that strongly nonlocal
temporal solitons can exist in the model (2). The second term of (2) models the diffraction
for the optical beam, and the group velocity dispersion (GVD) [22] for the optical pulse.
The nonlinear term of (2) describes the self focusing of the optical beam [23] and optical
pulse [10]. Generally speaking, when D = 2, NNLSE (2) only describes the propagations of
optical beams. The propagation of a pulsed optical beam can be described by the NNLSE,
and a optical bullet [24] can be obtained when D = 3. For D > 3, the NNLSE (2) is
just a phenomenological model, the counterpart of which can not be found in physics. The
response functionR canbesymmetric fortheoptical beam, but isasymmetric fortheoptical
pulse due to the causality [25].
III. DISCUSSION ABOUT THE VARIATIONAL APPROACH FOR THE NON-
LOCAL NONLINEAR SCHRO¨DINGER EQUATION
To find the approximately analytical solution of the NNLSE, the variational approach is
widely used [26]. The reason that the variational approach can be applied to the NNLSE is
that the NNLSE can be viewed as the Euler-Lagrange equation
∂ ∂l ∂ ∂l ∂l
+ = 0, (4)
∂z∂ ∂ϕ∗ ∂x∂ ∂ϕ∗ − ∂ϕ
∗
∂z ∂x
(cid:0) (cid:1) (cid:0) (cid:1)
4
where l is the Lagrangian density. Replacing ϕ with ϕ, the complex-conjugate equation of
∗
the NNLSE can be obtained from the Euler-Lagrange equation (4). It is easy to calculate
the first two terms of Eq. (4), but is some difficult to calculate the last term because of
the convolution between the response function and the intensity of the optical beam for the
NNLSE. In the following we will take the NNLSE (2) with D = 1 as an example to discuss
the condition under which the NNLSE (2) is equivalent to the Euler-Lagrange equation (4).
The Lagrangian density of the NNLSE (2) is [26]
2
i ∂ϕ ∂ϕ ∂ϕ 1
l = ϕ ϕ ∗ + ϕ(x,z) 2∆n, (5)
∗
2 (cid:18) ∂z − ∂z (cid:19)−(cid:12)∂x(cid:12) 2 | |
(cid:12) (cid:12)
(cid:12) (cid:12)
where ∆n = ∞ R(r r′) ϕ(r′,z) 2dDr′. In(cid:12)serti(cid:12)ng the Lagrangian density (5) into the
− | |
R−∞
Euler-Lagrange equation (4), the first two terms of (4) can be easily obtained as
∂ ∂l ∂ ∂l ∂ϕ ∂2ϕ
+ = i + . (6)
∂x∂ ∂ϕ∗ ∂z ∂ ∂ϕ∗ ∂z ∂x2
∂x ∂z
(cid:0) (cid:1) (cid:0) (cid:1)
To calculate the last term, we first construct a functional by integrating the last term of the
Lagrangian density (5) as
1 1
F(ϕ,ϕ ) = ∞ ∆n(x) ϕ(x) 2dx = ∞ ∞ R(x x) ϕ(x) 2 ϕ(x) 2dx. (7)
∗ ′ ′ ′
2 Z | | 2 Z Z − | | | |
−∞ −∞ −∞
The variation of the functional F(ϕ,ϕ ) can be obtained by definition as
∗
∂
δF(ϕ,ϕ ) = F(ϕ+εδϕ,ϕ +εδϕ )
∗ ∗ ∗ ε 0
∂ε | →
1
= ∞ ∞ R(x x) ϕ(x) 2[ϕ(x)δϕ (x)+ϕ (x)δϕ(x)]dxdx
′ ′ ∗ ′ ∗ ′ ′ ′
2 Z Z − | |
−∞ −∞
1
∞
+ ∆n[ϕ(x)δϕ (x)+ϕ (x)δϕ(x)]dx. (8)
∗ ∗
2 Z
−∞
If the response function is a even function, i.e., R(x) = R( x), then
−
we can obtain that ∞ ∞ R(x x′) ϕ(x) 2[ϕ(x′)δϕ∗(x′)+ϕ∗(x′)δϕ(x′)]dx′dx =
− | |
R−∞R−∞
∞ ∆n[ϕ(x)δϕ∗(x)+ϕ∗(x)δϕ(x)]dx. Then the variation of the functional F(ϕ,ϕ∗) is sim-
R−∞
plified to
∞ ∞
δF(ϕ,ϕ ) = ∆nϕ(x)δϕ (x)dx+ ∆nϕ (x)δϕ(x)dx. (9)
∗ ∗ ∗
Z Z
−∞ −∞
Because the variation of the functional F(ϕ,ϕ ) can be also expressed as
∗
∂ 1 ∂ 1
δF(ϕ,ϕ ) = ∞ ∆n(x) ϕ(x) 2 δϕ(x)dx+ ∞ ∆n(x) ϕ(x) 2 δϕ (x)dx.
∗ ∗
Z ∂ϕ (cid:20)2 | | (cid:21) Z ∂ϕ (cid:20)2 | | (cid:21)
∗
−∞ −∞ (10)
5
Comparing Eq.(9) and (10), we obtain
∂ 1
∆n(x) ϕ(x) 2 = ∆nϕ(x), (11)
∂ϕ (cid:20)2 | | (cid:21)
∗
∂ 1
∆n(x) ϕ(x) 2 = ∆nϕ (x). (12)
∗
∂ϕ (cid:20)2 | | (cid:21)
Then the NNLSE (2) can be obtained from the Euler-Lagrange equation (4) by combining
Eq.(6) and Eq.(11), its complex-conjugate equation can be obtained by combining Eq.(6)
and Eq.(12).
Consequently, it is conditional to apply the variational approach to the NNLSE, that is,
the response function must be an even function. When the response function is not an even
function, the variational approach will do not work any longer.
IV. CANONICAL EQUATIONS OF HAMILTON FOR THE FIRST-ORDER DIF-
FERENTIAL SYSTEM
We will use the canonical equations of Hamilton to deal with the nonlinear light-envelope
propagations. However the canonical equations of Hamilton appearing in all the literatures,
except for our recent work, are only valid for the second-order differential system but not
valid for the first-order differential system while the NNLSE (2) is the first-order differential
system. Therefore, it is necessary to briefly introduce the canonical equations of Hamilton
for the first-order differential system first.
For the first-order differential system of the continuous systems, the Lagrangian density
must be the linear function of the generalized velocities, and expressed as
N
l = R (q )q˙ +Q(q ,q ), (13)
s s s s s,x
Xs=1
where R is not the function of a set of q . Consequently, the generalized momentum p ,
s s,x s
which is obtained by the definition p = ∂l/∂q˙ as
s s
p = R (q ),(s = 1, ,N) (14)
s s s
···
is only a function of q . There are 2N variables, q and p , in Eqs. (14). The number of Eqs.
s s s
(14) is N, which also means there exist N constraints between q and p . So the degree of
s s
freedom of the system given by Eqs. (14) is N. Without loss of generality, we take q , ,q
1 ν
···
6
and p , ,p as the independent variables, where ν +µ = N. The remaining generalized
1 µ
···
coordinates and generalized momenta can be expressed with these independent variables as
q = q (q , ,q ,p , ,p )(α = ν + 1, ,N), and p = p (q , ,q ,p , ,p )(β =
α α 1 ν 1 µ β β 1 ν 1 µ
··· ··· ··· ··· ···
µ + 1, ,N). The Hamiltonian density h for the continuous system is obtained by the
···
Legendre transformation as h = N q˙ p l, where the Hamiltonian density h is a function
s=1 s s−
P
of ν generalized coordinates, q , ,q , and µ generalized momenta, p , ,p . We can
1 ν 1 µ
··· ···
obtain N canonical equations of Hamilton
N N
δh ∂p ∂q ∂ ∂h ∂f
s s α
= q˙ p˙ + , (15)
s s
δq (cid:18) ∂q − ∂q (cid:19) ∂x∂q ∂q
λ Xs=1 λ λ αX=ν+1 α,x λ
N N
δh ∂p ∂q ∂ ∂h ∂f
s s α
= q˙ p˙ + (16)
s s
δp (cid:18) ∂p − ∂p (cid:19) ∂x∂q ∂p
η Xs=1 η η αX=ν+1 α,x η
(λ = 1, ,ν, η = 1, ,µ, and ν+µ = N). The canonical equations of Hamilton (15) and
··· ···
(16) can be easily extended to the discrete system, which can be expressed as
N
∂H ∂p ∂q
s s
= q˙ p˙ , (17)
s s
∂q (cid:18) ∂q − ∂q (cid:19)
λ Xs=1 λ λ
N
∂H ∂p ∂q
s s
= q˙ p˙ , (18)
s s
∂p (cid:18) ∂p − ∂p (cid:19)
η Xs=1 η η
where λ = 1, ,ν, η = 1, ,µ, and ν +µ = N.
··· ···
V. APPLICATION IN NONLINEAR LIGHT-ENVELOPE PROPAGATIONS
Before the application of the canonical equations of Hamilton (17) and (18), we should
firstly calculate the Hamiltonian by the Legendre transformation reading H = N q˙ p
s=1 s s −
P
L, where the Lagrangian L can be obtained as L = ∞ ldDr. The Lagrangian L is a
R−∞
functionofgeneralizedcoordinates, ϕ,ϕ andgeneralizedvelocities, ϕ˙,ϕ˙ . Itisclear thatthe
∗ ∗
Lagrangian is not an explicit function of z, so the Hamiltonian of the system is conservative.
Now, we assume the light-envelop has a given form, ϕ = ϕ(q , ,q ), where q , ,q are
1 n 1 n
··· ···
the parameters changing with z. It can be regarded as the variables transformation, with
which we transform the coordinate system expressed by the set of generalized coordinate ϕ
to the one expressed by another set of generalized coordinates q , ,q .
1 n
···
Here we assume the material response is the Gaussian function R(r) =
(√πw1m)D exp(cid:16)−w|rm2|2(cid:17), and the trial solution has the form, ϕ(r,z) =
7
q (z)exp r2 exp[iq (z)r2 +iq (z)], where q ,q are the amplitude and phase of
A h−qw2(z)i c θ A θ
the complex amplitude of the light-envelope, respectively, q is the width of the light-
w
envelope, q is the phase-front curvature, and they all vary with the propagation distance
c
z. We obtain the Lagrangian
L = 2 2 DπD/2q2q 2+D(w2 +q2) D/2
− − − A w− m w −
[ 2q2q2+D +2D/2(w2 +q2)D/2(4D
− A w m w
+4Dq2q4 +Dq4q˙ +4q2q˙ )], (19)
c w w c w θ
which is a function of generalized coordinates, q ,q ,q and generalized velocities, q˙ ,q˙ , but
A w c c θ
not an explicit function of z. The generalized momenta can be obtained
p = p = 0, (20)
A w
pc = −2−2−D2DπD/2qA2qw2+D, (21)
π D/2
p = q2qD. (22)
θ − 2 A w
(cid:16) (cid:17)
The Hamiltonian of the system then can be determined by Legendre transformation
H = 2 1 DπD/2q2q 2+D(w2 +q2) D/2[ q2q2+D
− − A w− m w − − A w
+21+D2D(w2 +q2)D/2(1+q2q4)], (23)
m w c w
and can be proved to be a constant, i.e. H˙ = 0.
Therearefourgeneralizedcoordinatesandfourgeneralizedmomentainthefourequations
(20)(21)(22). So the degree of freedom of the set of equations (20)(21)(22) is four. Without
loss of generality, we take q ,q ,p and p as the independent variables. From Eqs.(21)(22),
c θ c θ
the generalized coordinates q ,q can be expressed by generalized momenta p and p as
A w c θ
q = ( p )1/2[Dp /(2πp )]D/4,q = [4p /(Dp )]1/2, inserting which into the Hamiltonian
A θ θ c w c θ
−
(23), we have
D2p2 +16p2q2 1 4p
H = θ c c π D/2( c +w2 ) D/2. (24)
− 4p − 2 − Dp m −
c θ
By use of the canonical equations of Hamilton (17) and (18), where µ = ν = 2 and n = 4,
8
we can obtain the following four equations
q˙ = D2p2θ 4q2 + Dπ−D/2p2θ(D4ppcθ +wm2 )−D/2, (25)
c 4p2 − c 4p +Dp w2
c c θ m
(4+D)π D/2p p (4pc +w2 ) D/2
q˙ = − c θ Dpθ m −
θ − 4p +Dp w2
c θ m
D2pθ Dπ−D/2p2θwm2 (D4ppcθ +wm2 )−D/2, (26)
− 2p − 4p +Dp w2
c c θ m
p˙ = 8p q , (27)
c c c
p˙ = 0. (28)
θ
Because q is a cyclic coordinate, the corresponding generalized momentum p is a con-
θ θ
stant, which can be confirmed by Eq.(28). In fact, this also represents that the power of
the light-envelope, P0 = ∞ |ϕ|2dDr = qA2( π/2qw)D, is conservative. From this we can
R−∞ p
obtain
q2 = P ( π/2q ) D. (29)
A 0 w −
p
Taking the derivative with respect to z on both sides of Eq.(21), then comparing it with
Eq.(27), we can obtain with the aid of Eq.(29)
q˙
w
q = . (30)
c
4q
w
TheninsertingEq.(30)intotheHamiltonian(23)withtheaidofEq.(29), wehaveH = T+V,
where
1
T = DP q˙2, (31)
16 0 w
DP 1
V = 0 π D/2P2 w2 +q 2 D/2. (32)
q 2 − 2 − 0 m w −
w (cid:0) (cid:1)
For the system, there are only one independent generalized coordinate q and one indepen-
w
dent generalized velocity q˙ , which can be proved in the Appendix. When the Hamiltonian
w
is expressed with independent variables, it is indeed the total energy expressed as the sum
of the generalized kinetic energy and the generalized potential.
The problem associated with the NNSLE is a problem of small oscillations from the
Hamiltonian point of view. The soliton corresponds to the extremum point of the gener-
alized potential. But in some literatures [13–16], solitons were regarded as the extrema of
the Hamiltonian. Such a treatment has some problem, because in those literatures the trial
9
solution has a changeless profile (solitonic profile), the system expressed with the solitonic
profile is the static system. The kinetic energy of the static system is zero, and the Hamil-
tonian is equal to the potential of the static system. In this connection, the extremum of
the Hamiltonian is the extremum of the generalized potential of the static system. But
when the system deviates from the equilibrium, the extremum of the Hamiltonian is not the
extremum of the generalized potential.
From ∂V/∂q = 0, we have
w
− q332 +8π−D/2P0qw wm2 +qw2 −1−D2 = 0. (33)
w (cid:0) (cid:1)
From Eq.(33) we can easily obtain the critical power, with which the light-envelope will
propagate with a changeless shape. Here we take the notation P to denote the critical
c
power instead of P , then we obtain
0
4πD/2(w2 +q2)1+D2
P = m w . (34)
c
q4
w
When P = P , we can obtain that q˙ = q = 0, which implies that the wavefront of the
0 c c c
soliton solution is a plane. The propagation constant is q˙ = [(4 D)q2 +4w2 ]/q4.
θ − w m w
Then we elucidate the stability characteristics of the soliton by means of the analysis
of the generalized potential V. Performing the second-order derivative of the generalized
potential V with respect to q , then inserting the critical power into it, we obtain
w
∂2V 64 2+D
Υ = 2 , (35)
≡ ∂q2 (cid:12) q4 (cid:20) − 2(1+σ2)(cid:21)
w (cid:12)P0=Pc w
(cid:12)
(cid:12)
where σ = w /q is the degree of nonlocality. The larger is σ, the stronger is the degree
m w
of nonlocality. When Υ > 0, the generalized potential has a minimum, and the soliton is
stable. From Eq.(35) we can obtain the criterion of the stability of solitons, that is
1
σ2 > (D 2), (36)
4 −
which is, in fact, consistent with the Vakhitov-Kolokolov (VK) criterion [27] with the aid of
the results of Ref.[28], and is proved briefly in [29].
10