Analytical Mechanics This page intentionally left blank
Download 10.87 Mb. Pdf ko'rish
|
x
a (U a )
b (U b )
b x a x a –1 (W ) x b –1 ˚ x a x b –1 (W ) x a –1 ˚ x b U a ⊂ Rl U b ⊂ Rl W M Fig. 1.19 U α
∈ A = / ∅ such that: (a) α ∈A x α (U α ) = M ;
(b) for any α and β in A, if x
α (U α ) ∩ x
β (U β ) = W = / ∅ the sets x −1 α (W ) and x −1 β (W ) are open subsets of R l and the maps x −1 β ◦ x α and x
−1 α ◦ x β (inverses of each other) are differentiable maps of class C k . The pair (U α , x
α ) (or the map x α ) is called a local parametrisation or a chart of M , while a family {(U
α , x
α ) } α ∈A with the properties listed in the definition is called a differentiable structure on M or an atlas of M (Fig. 1.19). In the example of the sphere in R 3 ,
The set A may have only one element if the representation of M is global. Evidently the Euclidean space R l endowed with the differential structure induced by the identity map is a differentiable manifold of dimension l. Example 1.29 Consider the l-dimensional sphere S l = {(x
1 , . . . , x l , x
l +1 ) ∈ R l +1 |x 2 1 + · · · + x 2 l
= 1 } with the atlas given by the stereographic projections π 1 : S
l \{N} → R
l and
π 2 : S l \{S} → R
l from the north pole N = (0, . . . , 0, 1) and from the south pole 1.7 Geometric and kinematic foundations of Lagrangian mechanics 37 S = (0, . . . , 0, −1), respectively: π 1 (x 1 , . . . , x l , x
l +1 ) = x 1 1 − x l +1 , . . . , x l 1 − x
l +1 , π 2 (x 1 , . . . , x l , x
l +1 ) = x 1 1 + x l +1 , . . . , x l 1 + x l +1 . It is immediate to verify that the parametrisations (R l , π −1 1 ), (R l , π
−1 2 ) define the structure of a differentiable manifold. Comparing this with the definition of a regular submanifold of R n , we note that the common feature of both definitions is the existence of local regular parametrisations (i.e. parametrisations without singular points). Indeed, we have the following. T heorem 1.7 Every regular l-dimensional submanifold V of R n is a differenti- able manifold. Proof
It follows from the implicit function theorem that to every point p of V one can associate an open neighbourhood A ⊂ R n
l , an open neighbour- hood U of u and a differentiable, invertible map x p : U → V such that x p (u) = p and x p (U ) = V ∩ A, and hence a local parametrisation of V (Fig. 1.20). Consider now the pairs (U p , x
p ) as p varies in V ; clearly the conditions of Definition 1.21 are satisfied, and thus {(U
p , x
p ) } p ∈V is an atlas for V . V u x A p = x(u) x(U ) U Fig. 1.20 38 Geometric and kinematic foundations of Lagrangian mechanics 1.7 Remark 1.15 The definition of a differentiable manifold naturally yields a topological space structure: we will say that a subset A of M is open if x −1 α
∩ x α (U α )) is an
open subset of R l for every α ∈ A. Hence a subset K of M is compact if every covering of K with open sets A has a finite subcovering. The manifold M is connected if for any two points P 1 , P 2 ∈ M there exists a finite sequence of charts {(U
j , x
j ) } j =1,...,N
such that P 1 ∈ x 1 (U 1 ), P N ∈ x N (U N ), the open sets U j are connected and U j ∩ U
j +1 = / ∅ for every j = 1, . . . , N − 1. Remark 1.16 With the topology induced by the differentiable structure, the manifold M is separable (i.e. every pair of points m 1 , m 2 in M has two open disjoint neigh- bourhoods A 1 and A 2 , m
1 ∈ A
1 and m
2 ∈ A
2 ) and the topology has a countable base (there is no loss of generality in assuming that A is countable). D efinition 1.22 A differentiable manifold M is orientable if it admits a dif- ferentiable structure {(U
α , x
α ) } α ∈A such that for every pair α, β ∈ A with x α (U α ) ∩ x β (U β ) =
/ ∅ the Jacobian of the change of coordinates x −1 α
β is positive. Otherwise the manifold is called non-orientable. D efinition 1.23 Let M 1 and M
2 be two differentiable manifolds of dimension l and m, respectively. A map g : M 1 → M 2 is differentiable at a point p ∈ M 1
⊂ R m → M 2 with y(V ) g(p), there exists a parametrisation x : U ⊂ R l → M 1 with x(U ) p, such that g(x(U ))
⊂ y(V ) and the function y −1 ◦ g ◦ x : U ⊂ R l → V ⊂ R m (1.56)
is differentiable in x −1 (p) (Fig. 1.21). The map g is differentiable in an open subset of M 1 if it is differentiable at every point of the subset. x –1 ( p) g( p) y –1 ˚ g ˚ x y –1 ˚ g ˚ x(U )
U g p x(U ) y(V ) M 1
1 )
2 Fig. 1.21 1.7 Geometric and kinematic foundations of Lagrangian mechanics 39 Note that by choosing M 2 = R this defines the notion of a differentiable map (in an obvious way we can also define the notion of a map of class C k or C ∞ ) from M to R. If we denote by f = (f 1 , . . . , f m ) the map (1.56), we have v i = f
i (u 1 , . . . , u l ), i = 1, . . . , m, where f i are differentiable functions. D efinition 1.24 A curve on a manifold M is a differentiable map γ : (a, b) → M.
If (U, x) is a local parametrisation of M in a neighbourhood of a point p = x(0), we can express a curve γ : ( −ε, ε) → M using the parametrisation (x −1 ◦ γ)(t) = (u 1 (t), . . . , u l (t))
∈ U. (1.57)
In spite of the fact that M has no metric structure, we can define at every point p of the curve the velocity vector through the l-tuple ( ˙ u 1
u l ). It is then natural to consider the velocity vectors corresponding to the l-tuples (1, 0, . . . , 0), (0, 1, . . . , 0), . . . , (0, 0, . . . , 1). We denote these vectors by the symbols ∂ ∂u
, . . . , ∂ ∂u l ; the generic velocity vector is expressed in the form of a linear combination ˙x = l i =1 ˙ u i ∂x ∂u i , (1.58) exactly as in the case of a regular l-dimensional submanifold. It is now easy to show that for p ∈ M and v ∈ T p M , it is possible to find a curve γ : ( −ε, ε) → M such that γ(0) = p and ˙γ(0) = v. Indeed, it is enough to consider the decomposition v =
l i =1 v i ∂x ∂u i (0) for some local parametrisation (U, x), and to construct a map µ : ( −ε, ε) → U such that its components u i (t) have derivatives u i (0) = v
i . The composite map x ◦ µ hence defines the required function γ (Fig. 1.22). D efinition 1.25 The tangent space T p M to a differentiable manifold M at a point p is the space of vectors tangent to the curves on M passing through p. The notion of a tangent space allows us to define the differential of a dif- ferentiable map g between two differentiable manifolds M 1 , M 2 . Given a point p ∈ M
1 , we define a linear map between T p M
and T g (p) M 2 . Consider a curve γ : ( − , ) → M 1 , such that γ(0) = p and ˙γ(0) = v, the given element of T p M 1 . The map g defines a curve on M 2 through β = g ◦ γ. It is natural to associate with v
∈ T p M 1 the vector w = ˙ β(0) ∈ T
g (p)
M 2 . 40 Geometric and kinematic foundations of Lagrangian mechanics 1.7
0
g p M U p = x(0) = g(0) x –1
dg dt (x –1 ˚ γ) (t) 0 Fig. 1.22 The construction of the vector w is easy after remarking that, if the curve γ(t) on M 1 possesses the local parametrisation (u 1 (t), . . . , u l (t)), then the curve β(t) on M 2 has the parametrisation (v 1 (t), . . . , v m (t)), where v i = f
i (u 1 . . . . , u l ), i = 1, . . . , m (cf. (1.56)). Hence if the vector v = ˙γ(0) is characterised with respect to the basis ∂ ∂u
, . . . , ∂ ∂u l by having components ( ˙ u 1
u l (0)), the vector w = ˙ β(0) with respect to the basis
∂ ∂v 1 , . . . , ∂ ∂v m has components ( ˙v 1 (0), . . . , ˙v m (0)), where ˙v i
l j =1 ∂f i ∂u j (u 1 (0), . . . , u l (0)) ˙ u j (0). We can thus give the following definition. D efinition 1.26 Let g : M 1 → M
2 be a differentiable map between the differ- entiable manifolds M 1 , M 2 of dimension l, m, respectively. The linear map which with every v ∈ T
p M 1 , defined by v = ˙γ(0), associates w ∈ T
g (p)
M 2 , defined by w = ˙ β(0), with β = g ◦ γ, is the differential dg p : T p M 1 → T g (p) M 2 . We showed that the map dg p acts on the components of the vectors in T p M 1 as the row-by-column product with the Jacobian matrix ∂(f 1 , . . . , f m )/∂(u
1 , . . . , u l ).
1.7 Geometric and kinematic foundations of Lagrangian mechanics 41 This happens in particular when the map is the change of parametrisation on a manifold (the Jacobian is in this case a square matrix). D efinition 1.27 Let M 1 and M
2 be two differentiable manifolds, both of dimen- sion l. A map g : M 1 → M 2 is a diffeomorphism if it is differentiable, bijective and its inverse g −1 is differentiable; g is a local diffeomorphism at p ∈ M 1 if there exist two neighbourhoods, A of p and B of g(p), such that g : A → B is a
diffeomorphism. Applying the theorem of local invertibility, it is not difficult to prove the following. T heorem 1.8 Let g : M 1 → M
2 be a differentiable map, and let p ∈ M 1
that dg p : T p M 1 → T g (p) M 2 is an isomorphism. Then g is a local diffeomorphism. Given a differentiable manifold M of dimension , the set of its tangent spaces T p M when p varies inside M has a natural structure as a differentiable manifold. Indeed, if {(U α
α ) } α ∈A is an atlas for M and we indicate by (u (α) 1 , . . . , u (α) ) the
local coordinates of U α , at every point of U α the vectors e (α) i
(α) i when i = 1, . . . , are a basis for the tangent space of M , and every tangent vector v ∈ T
p M can be written as v = i
v (α)
i ∂ ∂u (α) i p . D efinition 1.28 We call the tangent bundle of M, denoted by T M, the differentiable manifold of dimension 2 : T M =
p ∈M {p} × T p M (1.59) with the differentiable structure {(U
α × R , y
α ) } α ∈A , where y α (u (α) , v (α)
) = (x α (u (α)
), v (α)
), with u (α)
∈ U α being the vector of local coordinates in U a and v
(α) is a vector in the tangent space at a point x α (u
). The manifold M is called the base space of the tangent bundle. The map π : T M → M which associates with every point (p, v) ∈ T M the point p itself (at which v is tangent to M : v ∈ T
p M ) is called the projection onto the base. Clearly T p M = π −1 (p), (1.60) and T
p M is also called the fibre corresponding to the point p of the tangent bundle. The notion of a tangent bundle of a manifold is important as it allows one to extend to manifolds the notions of a vector field and a differential equation. D efinition 1.29 A (tangent) vector field on M is a map X : M → T M which associates with every point p ∈ M a vector v p ∈ T
p M in a differentiable way, i.e. it is a differentiable map X such that π(X(p)) = p, ∀p ∈ M.
42 Geometric and kinematic foundations of Lagrangian mechanics 1.7 For a given vector field, the integral curves are the curves γ : (a, b) → M such that ˙γ(t) = X(γ(t)). (1.61)
It is now natural to consider the problem of integrating differential equations on a manifold. Recalling equation (1.58), equation (1.61) can be written as a system of first- order differential equations: namely, if X is given in the form X(p) = i
α i (u 1 , . . . , u ) ∂x ∂u
, with p = x(u), then equation (1.61) is simply ˙ u
(t) = α i (u 1 (t), . . . , u (t)), i = 1, . . . , . Example 1.30 Let M be the unit sphere; consider the parametrisation x = (sin u 1 cos u
2 , sin u
1 sin u
2 , cos u
1 ), with the tangent vectors ∂x ∂u 1 = (cos u 1 cos u 2 , cos u
1 sin u
2 , − sin u 1 ), ∂x ∂u 2 = ( − sin u 1 sin u 2 , sin u
1 cos u
2 , 0).
A vector field tangent over M takes the form α 1 (u 1 , u 2 ) ∂x ∂u 1 + α 2 (u 1 , u 2 ) ∂x ∂u 2 . For example, if α 1 = constant, α 2 = constant the integral curves are given by u 1
1 t + u
(0) 1 , u 2 (t) = α
2 t + u
(0) 2 . We now extend the fundamental notion of a metric to differentiable manifolds. D efinition 1.30 A Riemannian metric on a differentiable manifold M of dimen- sion is a symmetric, positive definite bilinear form ( , ) p defined in the tangent space T p M , which has differentiable dependence on p. A differentiable manifold with a given Riemannian metric is called a Riemannian manifold. Example 1.31 The first fundamental form (1.34) is a Riemannian metric for any regular surface in R 3 .
1.7 Geometric and kinematic foundations of Lagrangian mechanics 43 Let x : U → M be a local parametrisation in p ∈ M with local coordinates (u 1 , . . . , u ). We saw that at every point q ∈ x(U), q = x(u 1 , . . . , u ), the vectors e i (q) = ∂ ∂u i q , i = 1, . . . , , are a basis for T q M . If ( , ) p is a Riemannian metric on M the functions g ij
1 , . . . , u ) = (e i (q), e
j (q))
q (1.62)
are differentiable in U for every i, j = 1, . . . , . Evidently g ij = g ji and if
(u 1 , . . . , u ) is a new local parametrisation, compatible with the former one, setting g ij = (e i (q), e
j (q))
q we have
g ij = m,n =1 J mi g mn J nj , (1.63) where J
mi = ∂u
m /∂u
i . Hence a Riemannian metric defines a symmetric covariant tensor of order 2 on the manifold (cf. Appendix 4). In analogy with the case of surfaces, we write (ds) 2
i,j =1 g ij (u 1 , . . . , u ) du i du j . (1.64) It is possible to prove that every differentiable manifold can be endowed with a Riemannian metric. Using this metric, one can define—in analogy with equation (1.32)—the notion of the length of a curve over M and of the arc length parameter s. We can also say that the metric tensor g ij (u) defines the scalar product in T p M and hence the norm of a vector in T p M . In particular, on the curve (u 1
l (s)) written with respect to the natural parametrisation, the tangent vector has unit norm. Example 1.32 The Lobaˇ cevskij half-plane is the Riemannian manifold given by {(x 1
2 ) ∈ R 2 |x 2 > 0
} with the usual differentiable structures (H is an open set of R 2 ) and the metric (ds)
2 = (dx 1 ) 2 + (dx 2 ) 2 x 2 2 , i.e. g 11 = g
22 = 1/x
2 2 , g 12 = g
21 = 0. A curve γ : (a, b) → H, γ(t) = (x 1 (t), x 2 (t))
has length = b a 1 x 2 (t)
˙ x 2 1 (t) + ˙
x 2 2 (t) dt. 44 Geometric and kinematic foundations of Lagrangian mechanics 1.7 For example, if γ(t) = (c, t) we have = b a dt t = log b a . D efinition 1.31 Let M and N be two Riemannian manifolds. A diffeomorphism g : M → N is an isometry if (v 1 , v 2 ) p = (dg p (v 1 ), dg
p (v 2 )) g (p) (1.65) for every p ∈ M and v 1 , v 2 ∈ T
p M . If N = M , g is called an isometry of M . It is not difficult to prove that the isometries of a Riemannian manifold form a group, denoted Isom(M ). Example 1.33 Let M = R be endowed with the Euclidean metric. The isometry group of R contains translations, rotations and reflections. Example 1.34 Consider the sphere S as immersed in R +1 , with the Riemannian metric induced by the Euclidean structure of R +1 . It is not difficult to prove that Isom(S ) = O( + 1), the group of ( + 1) × ( + 1) orthogonal matrices. Example 1.35 Consider the Lobaˇ cevskij plane H. Setting z = x 1 + ix 2 (where i = √ −1) the
mappings w =
az + b cz + d
, (1.66)
with a, b, c, d ∈ R, ad − bc = 1, are isometries of H. Indeed, (ds) 2
(dx 1 ) 2 + (dx
2 ) 2 x 2 2 = −4 dz dz (z − z)
2 . To prove that (1.66) is an isometry, we compute 4 dw dw
(w − w)
2 = 4
dw dz dw dz dz dz
az +b cz +d − az +b cz +d 2 . (1.67) Immediately one can verify that dw dz = 1 (cz + d) 2 , dw dz = 1 (cz + d) 2 , and that az + b
cz + d − az + b cz + d = z − z (cz + d)(cz + d) .
1.7 Geometric and kinematic foundations of Lagrangian mechanics 45 Substituting these relations into (1.67) yields 4 dw dw
(w − w)
2 = 4
dz dz (cz + d)
2 (cz + d)
2 (cz + d)
2 (cz + d)
2 (z − z) 2 = 4
dz dz (z − z) 2 . Among all curves on a Riemannian manifold M we now consider the particular case of the geodesics. D efinition 1.32 Given a local parametrisation (u 1 , . . . , u ) of M , and denoting by s the natural parameter along the curve, a geodesic s → (u
1 (s), . . . , u (s)) is a solution of the system of equations d 2 u k ds 2 + i,j =1 Γ k ij du i ds du j ds = 0,
k = 1, . . . , , (1.68)
where the Christoffel symbols Γ k ij are given by Γ k
= 1 2 n =1 g kn ∂g ni ∂u j + ∂g nj ∂u i − ∂g ij ∂u n (1.69) and (g
kn ) is the matrix inverse to (g ij ), which defined the metric (1.64). We shall consider in Chapter 9 the geometric interpretation of these equations, which are obviously an extension of equations (1.47), (1.48). Example 1.36 The Christoffel symbols corresponding to the Riemannian metric of the Lobaˇ cevskij half-plane are Γ 1 12 = Γ 1 21 = − 1 x 2 , Γ 2 11 = 1 x 2 , Γ 2 22 = − 1 x 2 , while Γ 1 11 = Γ 1 22 = Γ 2 12 = Γ 2 21 = 0. The geodesic equations are then given by the system d
x 1 ds 2 − 2 x 2 dx 1 ds dx 2 ds = 0, d 2 x 2 ds 2 + 1 x 2 dx 1 ds 2 − 1 x 2 dx 2 ds 2 = 0.
The first equation can be written as x 2 2 d ds 1 x 2 2 dx 1 ds = 0;
it follows that there exists a constant c ∈ R such that dx 1
= cx 2 2 . If c = 0 it follows that x 1 = constant, and hence vertical lines are geodesics. 46 Geometric and kinematic foundations of Lagrangian mechanics 1.8 Otherwise, substituting d ds = cx 2 2 d dx 1 into the second geodesic equation yields x 2 d 2 x 2 dx 2 1 + dx 2 dx 1 2 + 1 = 0. The general integral of this equation is given by x 2 =
2 − (x
1 − A)
2 , and
hence the geodesics corresponding to the values of c = 0 are semicircles with the centre on the x 1 -axis (i.e. on ∂H). Remark 1.17 Geodesics are invariant under any isometry of a Riemannian manifold. Indeed, thanks to (1.65) the Christoffel symbols (1.69) do not change. More generally, if g : M
→ N is an isometry, the geodesics on N are the images, through the isometry g, of geodesics on M and vice versa (cf. Problem 13.29). 1.8 Actions of groups and tori One way of constructing a differentiable manifold M from another manifold M is to consider the quotient of M with respect to an equivalence relation. This situation occurs frequently in mechanics. D efinition 1.33 A group G acts (to the left) on a differentiable manifold M if there exists a map ϕ : G × M → M such that: (a) for every g ∈ G the map ϕ g : M
→ M, ϕ g (p) = ϕ(g, p), where p ∈ M, is a diffeomorphism; (b) if e denotes the unit element in G, ϕ e = identity; (c) for any choice of g 1 , g 2 ∈ G, ϕ
g 1 g 2 = ϕ
g 1 ϕ g 2 . The action of G on M is free if for every p ∈ M the unit element e ∈ G is the only element of G such that ϕ e (p) = p. The action is discontinuous if every point p ∈ M has a neighbourhood A ⊂ M such that A ∩ ϕ g (A) =
∅ for every g ∈ G, g =
/ e. The action of a group on a manifold determines an equivalence relation on the manifold. D efinition 1.34 Two points p 1 , p
2 ∈ M are equivalent (denoted p 1 ∼ p
2 ) if and
only if there exists an element g ∈ G such that p 2 = ϕ
g (p 1 ). Two points of the manifold are equivalent if they belong to the same orbit Gp = {ϕ
(p) |g ∈ G}. The orbits of the points of M under the action of the group G are the equivalence classes [p] = Gp = {p ∈ M|p ∼ p}. 1.8 Geometric and kinematic foundations of Lagrangian mechanics 47 The quotient space M /G = {[p]|p ∈ M}, (1.70) with respect to the equivalence relation introduced, is a topological space, with the topology induced by the requirement that the projection π : M
→ M/G, π(p) = [p] (1.71)
is continuous and open (hence the open subsets of M /G are the projections of the open subsets of M ). It is not difficult to prove (cf. Do Carmo 1979) the following. T heorem 1.9 Let M be a differentiable manifold and let ϕ : G × M → M be the free discontinuous action of a group G on M . The quotient M = M /G is a differentiable manifold and the projection π : M → M is a local diffeomorphism. Proof
A local parametrisation of M /G is obtained by considering the restrictions of the local parametrisations x : U → M to open neighbourhoods U ⊂ R l of x −1 (p), where p ∈ x(U), such that x(U) ∩ ϕ g (x(U )) = ∅ for every g ∈ G, g =
/ e. We can then define the atlas of M /G through the charts (U, x), where x = π
◦ x : U → M/G (notice that, by the choice of U, π| x(U)
is injective). We leave it as a problem for the reader to verify that these charts define an atlas. Example 1.37 The group 2πZ acts on R 2 as a group of translations: ϕ k (x 1 , x 2 ) = (x 1 + + 2πk, x 2 ). The action is free and discontinuous, and the quotient is diffeomorphic to the cylinder S 1 × R. Example 1.38 The group (2πZ) l (whose elements are the vectors of R l of the form 2πm, where m ∈ Z
l ) acts on R l as the translation group: ϕ(x) = x + 2πm. It is easy to verify that the action is free and discontinuous, and that the quotient R l /(2πZ) l is a compact and connected differentiable manifold of dimension l called the l-dimensional torus T l . Its elements are the equivalence classes [x] of l-tuples of real numbers x = (x 1 , . . . , x l ) with respect to the equivalence relation x ∼ y ⇔ x − y ∈ (2πZ) l , and hence if and only if (x j − y
j )/2π is an integer for every j = 1, . . . , l. A geometric representation of T l is obtained by considering the cube of side 2π in R l , identifying opposites sides (Fig. 1.23). An alternative way to construct a manifold is to start from two manifolds M 1 and M 2 (of dimension l 1 and l
2 , respectively) and consider their Cartesian product, endowed with the product topology. T heorem 1.10 The Cartesian product M 1 × M
2 is a differentiable manifold of dimension l 1 + l 2 called the product manifold of M 1 and M
2 .
48 Geometric and kinematic foundations of Lagrangian mechanics 1.8 10p 8p 6p 4p 2p –2p –4p –2p 2p 4p 6p 8p 10p 12p
Download 10.87 Mb. Do'stlaringiz bilan baham: |
ma'muriyatiga murojaat qiling