Documentos de Académico
Documentos de Profesional
Documentos de Cultura
Invariant Lagrangians
In recent work, the authors show the mathematical structure behind both the
Euler–Lagrange system and the set of conservation laws, in terms of the
differential invariants of the group action and a moving frame. In this paper,
the authors demonstrate that the knowledge of this structure allows to find the
first integrals of the Euler–Lagrange equations, and subsequently, to solve by
quadratures, variational problems that are invariant under the special Euclidean
groups S E(2) and S E(3).
1. Introduction
plane (classified by Lie [3]). This paper is a continuation of [2]; here, the
new structure of Noether’s conservation laws is used to show that solving
one-dimensional variational problems, which are invariant under S E(2) and
S E(3), reduces to solving the Euler–Lagrange equations for the invariants,
which are partial differential equations for these, followed by quadratures. Note
that these groups are not semisimple and arise in many physical applications.
To make this paper self-contained, we start in Section 2 by giving a
brief overview to the theoretical foundations on the result on the structure
of Noether’s conservation laws: the application of moving frames to actions
on jet spaces that produces a symbolic invariant calculus for differential
invariants and their derivatives, and the use of this symbolic invariant calculus
to obtain the Euler–Lagrange equations for variational problems with a Lie
symmetry group directly in terms of the invariants. For the one-dimensional and
multidimensional symmetric variational problems, it was shown, in [4] and [2],
respectively, that Noether’s conservation laws come from a careful collection
of the boundary
terms obtained during the calculation of the first variation
of L [u] = L[u] dx, where L[u] is a smooth function of the independent
variables x, the dependent variables u, and finitely many derivatives of the u α ;
this calculation of the first variation involved two sets of integration by parts.
Here, we demonstrate a small result which simplifies the computation of these
conservation laws.
To illustrate the concepts presented in this section, we use the group
action of S E(2) on the plane as our pedagogical example. Having obtained
the Euler–Lagrange equation, in terms of the curvature invariant, for
one-dimensional variational problems invariant under S E(2) in Section 2, and
its associated Noether’s conservation laws in their new format, we demonstrate
in Section 3 that finding the solution to such variational problems reduces
down to solving the Euler–Lagrange equation for the invariants as functions of
the independent variable followed by quadratures.
In Section 4, we start by computing the invariantized Euler–Lagrange
equations for one-dimensional variational problems symmetric under S E(3)
and its six conservation laws; note that S E(3) is a six-dimensional Lie group.
Then, using the induced group action on the laws, which we show in explicit
detail, we can simplify the above variational problems; their solutions reduce to
solving Euler–Lagrange equations for the invariants followed by quadratures.
Kogan and Olver [7] for further details. We use the S E(2) action on the plane
as our pedagogical example. We follow, in this paper, the notation used in [2]
and [4].
A smooth group acting on a smooth space induces an action on the set of its
smooth curves and surface elements, as well as an action on their higher order
derivatives in the appropriate jet bundle, known as the prolonged curves and
surfaces. In this paper, the set M on which the group G acts is the set of these
prolonged curves and surfaces.
D ∂ q
∂
Di = = + u αK i α ,
Dxi ∂ xi α=1 K ∂u K
L(z) dx = L(
z) d
x
for all g ∈ G.
Moving Frames and Conservation Laws for Euclidean Invariant Lagrangians 137
O(z)
g
z
Consider a Lie group G acting smoothly on M such that the action is free
and regular. Then, for every z ∈ M, there exists a neighborhood U of z, as
illustrated in Figure 1, such that
- the group orbits have the dimension of the group G and foliate U;
- there is a surface K ⊂ U which intersects the group orbits transversally
at a single point. This surface is called the cross section;
- if O(z) represents the group orbit through z, then the group element
g ∈ G taking z ∈ U to k is unique.
Under these conditions, we can define a right moving frame as the map
ρ : U → G which sends z ∈ U to the unique element ρ(z) ∈ G such that
The element g ∈ G in Figure 1 equals ρ(z). Here, we are using moving frames
as reformulated by Fels and Olver [5, 6], in the context of differential algebra.
To obtain the right moving frame, which sends z to k, we must first define
the cross section K as the locus of the set of equations ψ j (z) = 0, j = 1, . . . , r ,
where r is the dimension of G. Normally, the cross section is chosen so as to
ease the calculations. Then, solving the set of equations
ψ j (
z) = ψ j (g · z) = 0, j = 1, . . . , r, (1)
ψ j (ρ(z) · z) = 0, j = 1, . . . , r.
138 T. M. N. Gonçalves and E. L. Mansfield
x cos θ sin θ x −a
= , (2)
y − sin θ cos θ y−b
where θ , a, and b are constants that parameterize the group action. Here, we
are using the inverse action because it simplifies the calculations. Since x is
not invariant, we reparameterize (x, y(x)) as (x(s), y(s)), where s is invariant
and let g ∈ S E(2) act on (x(s), y(s)) as in (2).
There is an induced action on the derivatives y K , where K is the index
of differentiation with respect to s, called the prolonged action. The induced
action on ys is defined to be
d
y d
y/ ds
ys = g · ys =
=
d
s d
s/ ds
by the chain rule, so the action of (2) on ys is
ys = −xs sin θ + ys cos θ.
Similarly,
d2
y 1 d d
y
yss = g · yss = 2
= = −xss sin θ + yss cos θ. (3)
d(
s) d
s/ ds ds d
s
If we consider M to be the space with coordinates (s, x, y, xs , ys , xss , yss , . . .),
then the action is locally free near the identity of S E(2). Thus, taking the
normalization equations to be x = 0,
y = 0, and ys = 0, we obtain
ys
a = x, b = y, and θ = arctan , (4)
xs
as the frame in parametric form.
Moving Frames and Conservation Laws for Euclidean Invariant Lagrangians 139
EXAMPLE 1 (CONT). Evaluating yss given by (3) at the frame (4) yields
xs yss − ys xss
yss |(a=x,b=y,θ=arctan ( xys )) =
.
s
xs2 + ys2
Thus, evaluating
y K at the frame (4) yields a differential invariant.
Di =
Di |g=ρ(z) = Di .
We know that
∂ α
u = u αK i ,
∂ xi K
although the same cannot be said about its invariantized version and in general
Di I Kα = I Kα i .
where δi j is the Kronecker delta, and Ni j and M Kα j are the correction terms.
The following theorem provides formulas for the correction terms Ni j and
M Kα j , for which we need to define the following notion of infinitesimal of a
prolonged group action.
Let G be a group parameterized by a1 , . . . , ar , where r = dim(G), in a
neighborhood of the identity element. The infinitesimals of the prolonged group
action with respect to these parameters are
j ∂ xj
α ∂ u
α
K
ξi = , φ K ,i = .
∂ai
g=e ∂ai
g=e
Moving Frames and Conservation Laws for Euclidean Invariant Lagrangians 141
and
β
φ Kα ,i (I ) = φ Kα ,i J, I β , I L ,
where the arguments have been invariantized.
THEOREM 3. For a left action on the base space and a right moving frame,
the p × r correction matrix K, which provides the correction terms, is given by
K j = z)|g=ρ(z) = ((Te Rρ )−1 )D j ρ) ,
D j ρ (
where ρ = (ρ1 , . . . , ρr )T is in parameter form and Rρ : G → G is the right
multiplication by ρ. The formulas for the correction terms are
r
r
Ni j = K j ξi (I ), M Kα j = K j φ Kα , (I ),
=1 =1
D
Ti j = I ζj . (8)
Dxi
Moreover, let denote the matrix of infinitesimals with invariantized arguments
∂(g · ζ j )
i j = (I ). (9)
∂ai
g=e
142 T. M. N. Gonçalves and E. L. Mansfield
Then, the correction matrix K providing the correction terms in the process of
invariant differentiation in Theorem 3 can also be given by
K = −TJ (J)−1 .
We will skip the proof of this theorem, as it can be found on page 136 of [4].
EXAMPLE 2. Consider the S E(2) left action on the (s, t, x(s, t), y(s, t))
space
x cos θ − sin θ x a
s = s,
t = t, = + .
y sin θ cos θ y b
Taking the normalization equations to be x = 0,
y = 0, and ys = 0, then the
ψi are ψ1 (x, y, ys ) = x, ψ2 (x, y, ys ) = y, and ψ3 (x, y, ys ) = ys . Thus, the
arguments of the ψi are x, y, and ys , and the invariantized normalization
y
equations are I x = 0, I y = 0, and I1 = 0.
The following table presents the infinitesimals of the prolonged group action:
s t x y xs ys ···
a 0 0 1 0 0 0 ···
b 0 0 0 1 0 0 ···
θ 0 0 −y x −ys xs ···
Hence,
a b θ
⎛ y ⎞
I11
−I x 0 − x⎟
K = ⎜ s⎜ 1 I 1 ⎟.
t⎜
⎝ y
y ⎟
I12 ⎠
−I2x −I2 − x
I1
In any case, software exists to calculate these correction terms [8]. Equations (7)
show that the processes of invariantization and differentiation do not commute.
Considering two generating differential invariants I Jα and I Lα and letting
J K = L M so that I JαK = I LαM , then this implies that
D K I Jα − M JαK = D M I Lα − M Lα M . (10)
These equations are called syzygies or differential identities. Note that even
when differential operators commute, there are still nontrivial syzygies between
invariants, as we see in the following example. For more information on
syzygies, see Section 5 in [4].
syzygy is
y
Dt η = Ds I2x − κηI2 . (11)
y y
Similarly, we have a syzygy between I2 and I11 and the syzygy is
y y ηs y y
Dt I11 = Ds2 I2 − Ds I2 − κ 2 η2 I2 + 2κηDs I2x + κs ηI2x . (12)
η
Syzygies will play a crucial role in the obtention of the invariantized
Euler–Lagrange equations and Noether’s conservation laws.
x x
I22 I122
Dx x
I2x I12
x
I112
Dt
0 I1x x
I11
x
Figure 2. Paths to I12 .
are invariant under some group action and let the κ j , j = 1, . . . , N , denote the
generating differential invariants of that group action. Also, assume that the
action leaves theindependent variables invariant so that the Lagrangians can
be rewritten as L[κ] dx. As mentioned in Remark 1, this can always be
achieved by reparameterization.
To obtain the invariantized Euler–Lagrange equations, Kogan and Olver in
[7] studied invariant calculus of variations from a geometric point of view, we
instead do it from a differential algebra point of view. Thus, to obtain the
Euler–Lagrange equations in terms of the invariants, we proceed in a similar
way as for finding the Euler–Lagrange equations in the original variables
(x, u).
Recall that if x → (x, u) extremizes the functional L [u], then for a small
perturbation of u,
d
0= L [u + εv]
dε
ε=0
q
D ∂L
= Eα (L)v α + α
αv + ··· dx
α=1 i
Dx i ∂u i
after differentiation under the integral sign and integration by parts, where
D |K | ∂
Eα = (−1)|K |
K Dx1k1 ···
k
Dx pp ∂u αK
is the Euler operator with respect to the dependent variables u α . The boundary
terms play an important part in the formulas for Noether’s conservation laws,
in the case where the perturbation is given by the group action.
Moving Frames and Conservation Laws for Euclidean Invariant Lagrangians 145
⎛ ⎞ ⎛ 1⎞
κ1 It
⎜ ⎟ ⎜ ⎟
Dt ⎝ ... ⎠ = H ⎝ ... ⎠ , (13)
q
κN It
[L(κ, κs , κss , . . .) − λ(s)(η − 1)] ds, (14)
where λ(s) is a Lagrange multiplier. This constraint does not reduce the
solution set and it will simplify the calculations.
Since symbolically we know that
d
D
L [x + εv] = L [x],
dε
ε=0 Dt
xt =v
146 T. M. N. Gonçalves and E. L. Mansfield
∂L ∂L ∂L ∂L
= − Ds + Ds 2
− Ds3
+ · · · Dt κ
∂κ ∂κs ∂κss ∂κsss
m−1
∂ L
− λ(s)Dt η + Ds (−1)n Dsn Dsm−1−n Dt κ ds
m=1 n=0
∂κ m
=
Eκ (L) Dt κ − λ(s) Dt η
m−1
∂L
+ Ds (−1) Ds
n n
Dsm−1−n Dt κ ds,
m=1 n=0
∂κm
after differentiating under the integral sign and integrating by parts, where
∂mκ
κm = .
∂s m
Next, we substitute the circled Dt η and Dt κ by their respective syzygies
x
η Ds −κ I2 H1
Dt = y = I (xt ), (15)
κ κs Ds + κ
2 2
I2 H2
where we have already set η = 1; it makes no difference to the final result to
do this at this point. Hence,
(Eκ (L)H2 − λ(s)H1 ) I (xt )
m−1
∂L
+ Ds (−1) Ds
n n
Dsm−1−n Dt κ ds
m=1 n=0
∂κm
(H2∗ (Eκ (L)) − H1∗ (λ(s)))I (xt ) + Ds −λ(s)I2x + Eκ (L)Ds I2
y
=
m−1
∂L
− Ds Eκ (L)I2
y
+ (−1) Ds
n n
Dsm−1−n Dt κ ds, (16)
m=1 n=0
∂κm
Moving Frames and Conservation Laws for Euclidean Invariant Lagrangians 147
after a second set of integration by parts, and where H1∗ and H2∗ are the
respective adjoint operators of H1 and H2 . By definition, we know that I (xt )
contains xt ; hence, from the Fundamental Lemma of Calculus of Variations,
y
the coefficients of I2x and I2 must be zero, that is, they represent, respectively,
the invariantized Euler–Lagrange equations
Ex (L) = κs Eκ (L) + λs ,
E y (L) = Ds2 Eκ (L) + κ 2 Eκ (L) + λ(s)κ.
We use Ex (L) = 0 to eliminate λ(s) instead of E y (L) = 0, as it contains
derivatives of κ of lower order. Since L does not depend on s explicitly, by the
result in page 220 of [4], κs Eκ (L) is a total derivative, specifically
⎛ ⎞
m−1
∂ L
κs Eκ (L) = Ds ⎝ L − (−1) j Dsj κm− j ⎠ ,
m=1 j=0
∂κm
then we obtain
m−1
∂L
λ(s) = −L + (−1) Ds
j j
κm− j , (17)
m=1 j=0
∂κm
where the constant of integration has been absorbed into λ(s) (see Remark
7.1.9. of [4]). Hence, we are left with one invariantized Euler–Lagrange
equation in one unknown
⎛ ⎞
m−1
∂ L
E y (L) = Ds2 Eκ (L) + κ 2 Eκ (L) − κ ⎝L − (−1)m Dsm κm− ⎠
j .
m=1 j=0
∂κm (18)
Equation (18) agrees with the one appearing in Kogan and Olver [7].
We have now covered the theoretical foundations necessary to understand
the structure of Noether’s conservation laws.
where ρ(z)−1 is the right moving frame (4), λ(s) is the Lagrange multiplier
obtained in (17), and c the constant vector.
We note that R(g) is the Adjoint representation of S E(2) on its Lie algebra
se(2). To see how part of the calculations in Section 2.2 can yield the result in
(19), we first show how the Adjoint representation is calculated in the context
we will need.
Consider the S E(2) group action, as in Example 2, with generating
infinitesimal vector fields
∂x , ∂y , −y∂x + x∂ y .
Let g ∈ S E(2) act on
v = α∂x + β∂ y + γ (−y∂x + x∂ y ),
where α, β, and γ are constants. Hence,
g · v = α∂x + β∂y + γ (−y∂x + x∂y )
⎛ ⎞⎛ ⎞
cos θ − sin θ 0 ∂x
⎜ ⎟⎜ ⎟
= (α β γ ) ⎝ sin θ cos θ 0⎠ ⎝ ∂y ⎠.
a sin θ − b cos θ a cos θ + b sin θ 1 −y∂x + x∂ y
Thus, R(g) is the Adjoint representation of S E(2), denoted as Ad(g).
Next, recall the boundary terms, in (16), obtained in the calculation of the
invariantized Euler–Lagrange equations,
−λ(s)I2x + Eκ (L)Ds I2 − (Ds Eκ (L)) I2
y y
m−1
∂L (20)
+ (−1) Ds
n n
Dsm−1−n Dt κ = c,
m=1 n=0
∂κm
y
where c is a constant. Substituting Ds I2 and Dsm−1−n Dt κ for all m in the
above expression by the differential formulas
y y
Ds I2 = I12 − κ I2x ,
y
Dt κ = I112 − 2κ I12
x
,
y y
Ds Dt κ = I1112 − 2κ 2 I12 − 3κ I112
x
− 2κs I12
x
,
..
.
Moving Frames and Conservation Laws for Euclidean Invariant Lagrangians 149
where (Ad(ρ)−1 −1 α α
j∗ ) denotes row j of Ad(ρ) . Substituting the (I2 I12 . . .) in
(21) by (22) for each of the group parameters yields the conservation laws
Ad(ρ)−1 α (I )C α = c
α
150 T. M. N. Gonçalves and E. L. Mansfield
in matrix form. Thus, the vector of invariants υ(I ) in (19) equals the sum of
the products of the matrices of invariantized infinitesimals α (I ) with the
vectors C α ,
⎛ ⎞
−λ(s) − κEκ (L)
x (I )C x + y (I )C y = ⎝ −Ds Eκ (L) ⎠ ,
Eκ (L)
where
x xs xss . . . y ys yss ysss ...
⎛ ⎞ ⎛ ⎞
a 1 0 0 ... a 0 0 0 0 ...
(I ) = ⎜
x
, (I ) = ⎜
y
.
b ⎝0 0 0 . . .⎟
⎠ b⎝ 1 0 0 0 ... ⎟
⎠
θ 0 0 −κ ... θ 0 1 0 −κ 2 ...
Finally, Noether’s conservation laws for one-dimensional Lagrangians
invariant under S E(2) can be written as
⎛ ⎞⎛ ⎞
xs −ys 0 −λ(s) − κEκ (L)
⎜ ⎟⎜ ⎟
⎝ ys xs 0⎠ ⎝ −Ds Eκ (L) ⎠ = c, (23)
κ
x ys − yxs x xs + yys 1 E (L)
where xs2 + ys2 = 1 was used to simplify the conservation laws and
m−1
∂L
λ(s) = −L + (−1) Ds
j j
κm− j .
m=1 j=0
∂κm
The following theorem generalizes what we have just seen for one-dimensional
variational problems that are invariant under S E(2); it states that the conservation
laws from Noether’s first theorem can be written as the divergence of the
product of a moving frame with vectors of invariants.
THEOREM 5. Let L(κ1 , κ2 , . . .) dx be invariant under G × M → M,
where M = J n (X × U ), with generating invariants κ j , for j = 1, . . . , N , and
let g · xi = xi , for i = 1, . . . , p. Introduce a dummy variable t to effect the
variation and then integration by parts yields
∂ α α
L(κ1 , κ2 , . . .) dx = E (L)It + Div(P) dx,
∂t α
where this defines the p-tuple P, whose components are of the form
Pi = I Jαt Ci,J
α
, i = 1, . . . , p,
α,J
Moving Frames and Conservation Laws for Euclidean Invariant Lagrangians 151
where
z i
∂
ζ ji =
∂a j
g=e
where
υ i (I ) = α (I )Ciα .
α
m−1
∂L
(−1) Ds
n n
Dsm−1−n Dt κ
m=1 n=0
∂κm
from the first set of integration by parts do not appear in (23). Following
through the calculations in the proof of Theorem 5 in [2], we show in the
following lemma, in addition, that the first set of boundary terms drops.
Dt I Kα = I Kα t + M Kα t ,
−J(J)−1 φ αK
Dt I Kα = (Tt∗ ) I Kα t , (24)
1
where the components of the vectors B γ are the coefficients of the I γ in (24),
γ γ
we can substitute ( It I j1 t · · ·) on the right-hand side of (26) by (25) for each
group parameter and obtain
Ad(ρ)−1
j∗ γ (I )B γ , j = 1, . . . , r. (27)
γ
which is equivalent to
−J(J)−1 φ αK
Ad(ρ) −1
φ αK = Ad(ρ)−1 −φ αK + φ αK = 0.
1
2 κ 2 κ ∂L
E (L) = Ds E (L) + κ E (L) − κ L −
y
(−1) Ds
j j
κm− j , (28)
m=1 j=0
∂κm
where
m−1
∂L
λ(s) = −L + (−1) j Dsj κm− j , (29)
m=1 j=0
∂κm
and
∂mκ
κm = .
∂s m
Multiplying both sides of the above conservation laws, Ad(ρ(z))−1 υ(I ) = c,
by Ad(ρ(z)) yields the following system of equations
−λ(s) − κEκ (L) = xs c1 + ys c2 , (30)
i.e.,
(λ(s) + κEκ (L))2 + (Ds Eκ (L))2 = c12 + c22 , (33)
with λ(s) as in (29).
Once κ is known, one can verify from the following, that the integration
problem has a straightforward solution. Integrating both sides of Equation (30)
with respect to s yields
xc1 + yc2 = [−λ(s) − κEκ (L)] ds, (34)
and then solving the two linear equations (32) and (34) with respect to x and
y, one obtains
1 κ κ
x(s) = 2 c1 [−λ(s) − κE (L)] ds − c2 E (L) + c2 c3 , (35)
c1 + c22
1 κ κ c22 c3
y(s) = c 2 [−λ(s) − κE (L)] ds + c 1 E (L) + . (36)
c12 + c22 c1
One can easily see that Equation (31) is immediately satisfied as it is the
derivative with respect to s of (32).
Thus, for one-dimensional variational problems invariant under the group
S E(2), to obtain the extremal curves, one only needs to solve (33) for κ and
then use Equations (35) and (36).
In the previous section, we showed that for an invariant Lagrangian under S E(2),
the invariantized Euler–Lagrange equation and its associated conservation
laws could be used to solve for the extremal curves. In this section, we will
proceed analogously and present the solution to one-dimensional variational
problems that are invariant under the following S E(3) group action on the
(x(s), y(s), z(s))-space parameterized by the Euclidean arc length s
x = R−1 (x − a),
x → (37)
where
⎛ ⎞
cos β cos γ cos β sin γ sin β
⎜ ⎟
R−1 = ⎝− sin α sin β cos γ − cos α sin γ − sin α sin β sin γ + cos α cos γ sin α cos β ⎠
− cos α sin β cos γ + sin α sin γ − cos α sin β sin γ − sin α cos γ cos α cos β
(38)
represents the rotation in the three-dimensional space, and a = (a, b, c)T , the
translation vector with α, β, γ , a, b, and c as the constants that parameterize
the group action.
To solve S E(3) invariant variational problems, we need to find the element
g ∈ S E(3) that sends the point (x, y, z) to the origin, the tangent to the curve
at the point (x, y, z) to the x-axis, and the normal to the curve at the point
(x, y, z) to the y-axis; in other words, which sends z = (x, y, z, ys , z s , z ss ) to
the cross section (0, 0, 0, 0, 0, 0). For that we solve the normalization equations
x = 0, y = 0, z = 0, ys = 0, zs = 0, and z ss = 0, and thus obtain the right
moving frame
ys (y s z ss − z s yss ) − x s (z s x ss − x s z ss )
a = x, b = y, c = z, α = tan−1 ,
xs2 + ys2 + z s2 (xs yss − ys xss )
−1 z s −1 ys (39)
β = tan , γ = tan .
xs + ys
2 2 x s
we first rewrite L [x] in terms of the generating invariants of the group action,
which are the Euclidean curvature
xs × xss
κ= ,
xs
3
and torsion
xsss · (xs × xss )
τ= ,
xs × xss
2
and their derivatives with respect to s.
Sinces representstheEuclideanarclength,theconstraintη = xs2 + ys2 + z s2 =
1 must be introduced into the variational problem to fix parameterization.
Hence, the resulting invariantized functional is
[L(κ, τ, κs , τs , κss , τss , . . .) − λ(s)(η − 1)] ds, (40)
where λ(s) is a Lagrange multiplier. As for S E(2), this constraint does not
reduce the solution set and simplifies the computation of the conservation laws.
Next, we introduce a dummy invariant independent variable t and set
x = x(s, t) to effect the variation. The introduction of a new independent
variable results in three new invariants Itα , for α = x, y, z, and a set of syzygies
⎛ ⎞ ⎛ x⎞
η I2
Dt κ = H I 2 ⎠ ,
⎝ ⎠ ⎝ y
(41)
τ z
I2
where the matrix of operators H is
⎛ ⎞
Ds −κ 0
⎜ κs κ 2 − τ 2 + Ds2 −τs − 2τ Ds ⎟
⎝ ⎠,
τs 3τs 2κs τ τ2 τ 2 κs
τs + 2τ Ds Ds κ + κ − κ 2 Ds + 2τκ Ds2 Ds − κ
+ κ − κ Ds − κ 2 Ds + κ Ds
2 1 3
2 κ 2τ 2 τ τs 2τ κs
E (L) = Ds E (L) + Ds E (L) +
y
− 2 Ds Eτ (L) + (κ 2 − τ 2 )Eκ (L)
κ κ κ
+ 2τ κEτ (L) + λ(s)κ,
τ τs τ 2τ τ κ y
(2τ E (L) − λ) I2 +
x
E (L) − Ds E (L) − Ds E (L) I2
κ κ
2τ τ τ2 κs
+ E (L) + E (L) Ds I2 + κEτ (L) − 2τ Eκ (L) − Eτ (L) − 2 Ds Eτ (L)
κ y
κ κ κ
1 1 1
+ Ds2 Eτ (L) I2z − Ds Eτ (L)Ds I2z + Eτ (L)Ds2 I2z = k,
κ κ κ
where k is a constant. Note that by Lemma 1, we can disregard the boundary
terms from the first set of integration by parts.
Using Ex (L) = 0 and the fact that
m−1
∂L
κ τ
κs E (L) + τs E (L) = Ds L − (−1)i Dsi κm−i
m=1 i=0
∂κm
m−1
∂ L
− (−1)i Dsi τm−i
m=1 i=0
∂τm
m−1
∂L
+ (−1)i Dsi τm−i , (42)
m=1 i=0
∂τm
where the constant of integration has been absorbed into the Lagrange
multiplier, and hence we obtain two invariantized Euler–Lagrange equations in
two unknowns.
To calculate the conservation laws associated with the invariantized
Euler–Lagrange equations, we must compute the moving frame Ad(ρ)−1 and
the vector of invariants υ(I ). To compute the former, we proceed as in
Section 2.3: we calculate the Adjoint representation Ad(g) of S E(3) with
respect to its generating infinitesimal vector fields
va = ∂x , vb = ∂ y , vc = ∂ z , vα = y∂z − z∂ y ,
vβ = x∂z − z∂x , vγ = x∂ y − y∂x ,
and evaluate it at the frame (39), which yields
ρ F S 0
Ad(ρ(z))−1 = D Xρ ,
F S Dρ F S D
158 T. M. N. Gonçalves and E. L. Mansfield
where
xss xs × xss
ρ F S = xs
κ κ
As seen in Section 2.3, to obtain the vector of invariants, we must first find
the boundary terms that are linear in the I Kα 2 . To do so, consider the boundary
terms obtained from the calculation of the invariantized Euler–Lagrange
equations
τ τs τ 2τ τ κ y
(2τ E (L) − λ)I2x
+ E (L) − Ds E (L) − Ds E (L) I2
κ κ
2τ τ τ2
+ E (L) + E (L) Ds I2 + κEτ (L) − 2τ Eκ (L) − Eτ (L)
κ y
κ κ
κs 1 1
− 2 Ds Eτ (L) + Ds2 Eτ (L) I2z − Ds Eτ (L)Ds I2z (43)
κ κ κ
1
+ Eτ (L)Ds2 I2z = k.
κ
y
Substituting Ds I2 , Ds I2z , and Ds2 I2z in (43) by the differential formulas
y y
Ds I2 = −κ I2x + I12 + τ I2z ,
y
Ds I2z = −τ I2 + I12
z
,
y y
Ds2 I2z = τ κ I2x − τs I2 − 2τ I12 − τ 2 I2z + I112
z
,
which were obtained from (7), yields that the boundary terms are linear in
the I Kα 2 ,
Moving Frames and Conservation Laws for Euclidean Invariant Lagrangians 159
⎛ τ ⎞
x −Ds Eκ (L) − Ds Eτ (L)
I2 (−κEκ (L) + τ Eτ (L) − λ(s)) + I2 I12 ⎝
y y κ ⎠
κ
E (L)
cx
Cy
⎛ ⎞
1 2 κs τ τ κ
⎜ κ Ds E(L) − κ 2 Ds E (L) + κE (L) − τ E (L)⎟
⎜ ⎟
z z z ⎜ ⎜ 1 τ
⎟
⎟
+ I2 I12 I112 ⎜ − Ds E (L) ⎟ = k.
⎜ κ ⎟
⎜ ⎟
⎝ 1 ⎠
Eτ (L)
κ
Cz
⎛ ⎞
τ Eτ (L) − κEκ (L) − λ(s)
⎜ τ ⎟
⎜ −Ds Eκ (L) − Ds Eτ (L) ⎟
⎜ ⎟
⎜ κ ⎟
⎜1 κs ⎟
⎜ 2 τ τ τ κ ⎟
⎜ D E (L) − D E (L) + κE (L) − τ E (L) ⎟
⎜
υ(I ) = ⎜ κ s
κ 2 s ⎟,
⎟
⎜ τ
E (L) ⎟
⎜ ⎟
⎜ ⎟
⎜ 1 ⎟
⎜ − Ds Eτ (L) ⎟
⎝ κ ⎠
Eκ (L)
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
1 0 0 0 0 0
⎜0⎟ ⎜1 0⎟ ⎜0 0 0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜0⎟ ⎜0 0⎟ ⎜1 0 0⎟
(I ) = ⎜
x ⎟
⎜0⎟ , (I ) = ⎜
y
⎜0
⎟, (I ) = ⎜
z ⎟.
⎜ ⎟ ⎜ 0⎟⎟
⎜0
⎜ 0 κ⎟⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎝0⎠ ⎝0 0⎠ ⎝0 1 0⎠
0 0 1 0 0 0
160 T. M. N. Gonçalves and E. L. Mansfield
⎛ ⎞
τ Eτ (L) − κEκ (L) − λ(s)
⎜ τ ⎟
⎜ −Ds Eκ (L) − Ds Eτ (L) ⎟
⎜ κ ⎟
⎜ ⎟
⎜1 2 τ κs ⎟
|c1 |
z
ss = υ (I ),
(2)
(48)
κ
|c1 |
xs
( yss −
ys x
ss ) = υ (I ),
(3)
(49)
κ
c1 T Dc2
|c1 |(
x
ys −
y xs ) + zs = υ (4) (I ), (50)
|c1 |
162 T. M. N. Gonçalves and E. L. Mansfield
|c1 | c1 T Dc2
xss
( y−
yss
x) − z
ss = υ (I ),
(5)
(51)
κ κ|c1 |
|c1 | c1 T Dc2
( z s x
x( ss −
x
z
s ss ) −
y(
y
z
s ss −
z
y
s ss )) + xs
( yss −
ys x
ss )
κ κ|c1 |
= υ (6) (I ), (52)
where we have used υ ( j) (I ) to denote the jth component of υ(I ). Note that the
variables x,
y, and
z have become our new space coordinates after sending
c to C.
This overdetermined system of equations can be solved more easily
than (44). The two first integrals of the invariantized Euler–Lagrange
equations are obtained as follows. Define B = diag (1, 1, 1, 0, 0, 0), which
satisfies B = Ad(ρ)−T BAd(ρ)−1 . Then, we obtain the first integral of the
Euler–Lagrange equations
υ T (I )Bυ(I ) = CT BC,
which is equivalent to
τ 2
(τ Eτ (L) − κEκ (L) − λ(s))2 + −Ds Eκ (L) − Ds Eτ (L)
κ
2
1 2 τ κs
+ Ds E (L) − 2 Ds Eτ (L) + κEτ (L) − τ Eκ (L) = c12 + c22 + c32 . (53)
κ κ
This result compares to the first integral (33) for the S E(2) invariant
Lagrangians. For the second first integral, we define
0 D
D= , D = diag(1, −1, 1),
D 0
υ T (I )Dυ(I ) = CT DC,
i.e.,
τ 1
(τ Eτ (L) − κEκ (L) − λ(s)) Eτ (L) + −Ds Eκ (L) − Ds Eτ (L) Ds Eτ (L)
κ κ
1 2 τ κs
+ D E (L) − 2 Ds E (L) + κE (L) − τ E (L) Eκ (L)
τ τ κ
κ s κ
(54)
= c1 c4 − c2 c5 + c3 c6 .
Moving Frames and Conservation Laws for Euclidean Invariant Lagrangians 163
We can use the above first integrals to ease the calculation of κ and τ . Once
we have solved for these, we use the system of simplified conservation laws to
solve for x,
y, and
z.
First, solving Equation (47) gives
1
z(s) = υ (1) (I ) ds. (55)
|c1 |
T
Next, multiplying Equation (49) by − c1|c1Dc
|2
2
and adding it to Equation (52) yields
1 2 1 c1 T Dc2 (3)
|c1 | zs Ds ( x) − 1 − z
x · ss Ds (
x ·
x) = κυ (6) (I ) − κ υ (I ),
2 2 |c1 |2
where 12 Ds2 (
x ·
x) − 1 = x · x
ss and 2 Ds (
1
x ·
x) = x · xs . Setting Ds ( x · x) = h(s)
and substituting zs by (47) and z ss by its derivative yields a linear equation for h
!
Ds υ (1) (I ) c1 T Dc2 (3)
Ds h − h = 2κ υ (6)
(I ) − υ (I ) υ (1) (I ) + 2.
υ (1) (I ) |c1 |2
Solving for h, we obtain
!
1 c1 T Dc2 (3)
h(s) = υ (I )
(1)
2κ υ (I ) −
(6)
υ (I ) υ (I ) + 2 ds.
(1)
υ (1) (I ) |c1 |2
Hence,
!
1 c1 T Dc2 (3)
Ds (
x ·
x) = υ (I )
(1)
2κ υ (I ) −
(6)
υ (I ) υ (I ) +2 ds.
(1)
υ (1) (I ) |c1 |2
(56)
To solve Equations (50) and (56), we use the cylindrical coordinates
x(s) = r (s) cos θ (s), y(s) = r (s) sin θ (s), z(s) = z(s).
Starting with Equation (56), we obtain
2
r (s) = h(s) ds − z(s) ,
2
(57)
2
as Ds (
x ·
x) = Ds r (s)2 + z(s) . After applying the change of coordinates,
Equation (50) becomes
1 c1 T Dc2 (1)
r (s) θs =
2
υ (I ) −
(4)
υ (I ) ,
|c1 | |c1 |2
164 T. M. N. Gonçalves and E. L. Mansfield
1 c1 T Dc2 (1)
θ (s) = υ (I ) −
(4)
υ (I ) ds. (58)
r (s)2 |c1 | |c1 |2
To recover x, y, and z, we apply the reverse action to
x,
y, and
z, i.e.,
x → x = R
x + a,
where R is the three-dimensional rotation (45) and a = (a, b, c)T is the
translation vector, with
⎛ ⎞
|c1 |2 cos2 β − c32
α = − tan−1 ⎝ ⎠,
c3
⎛ ⎞
c2 c3 sin β + c1 |c1 |2 cos2 β − c32
γ = tan−1 ⎝ ⎠,
c1 c3 sin β − c2 |c1 |2 cos2 β − c32
the Euler–Lagrange equations in the original variables, and then one has
Equations (55), (57), and (58), which one uses to find
x = r cos θ ,
y = r sin θ ,
and z.
In the following two well-known examples, we can verify that the formulas
derived here provide the expected solutions, giving the obvious check on our
calculations.
5. Conclusion
References
UNIVERSITY OF KENT
(Received 27 June, 2012)