Analytical Mechanics This page intentionally left blank
Download 10.87 Mb. Pdf ko'rish
|
n (s) t (s) x 1 Fig. 1.6 D efinition 1.4 The unit vector t(s) = dx(s)
ds (1.9)
is called the unit tangent vector to the curve. D efinition 1.5 At any point at which d 2 x/ds
2 = / 0 it is possible to define the unit vector n(s) =
1 k(s)
d 2 x ds 2 , (1.10) called the principal unit normal vector (Fig. 1.6), where k(s) = |d 2
2 | is the
curvature of the plane curve. R(s) = 1/k(s) is the radius of curvature. It easily follows from the definition that straight lines have zero curvature (hence their radius of curvature is infinite) and that the circle of radius R has curvature 1/R. Remark 1.4 Given a point on the curve, it follows from the definition that n(s) lies in the half-plane bounded by the tangent t(s) and containing the curve in a neigh- bourhood of the given point. The orientation of t(s) is determined by the positive orientation of the curve. Remark 1.5 Consider a point of unit mass, constrained to move along the curve with a time dependence given by s = s(t). We shall see that in this case the curvature determines the strength of the constraining reaction at each point. The radius of curvature has an interesting geometric interpretation. Consider the family of circles that are tangent to the curve at a point P . Then the circle
1.3 Geometric and kinematic foundations of Lagrangian mechanics 9 c
x(s 0 ) Fig. 1.7 that best approximates the curve in a neighbourhood of P has radius equal to the radius of curvature at the point P . Indeed, choosing a circle of radius r and centred in a point c = (c 1 , c
2 ) lying on the normal line to the curve at a point x(s 0
by the function g(s) =
|x(s) − c| − r, with s a variable in a neighbourhood of s 0 . Since
g (s 0 ) = 1 r (x(s 0 ) − c) · t(s 0 ) = 0,
g (s 0 ) = 1 r (1 − kr), it follows that g(s) is an infinitesimal of order greater than (s −s 0
2 if g (s
0 ) = 0,
and hence if c − x(s
0 ) = R(s
0 )n(s
0 ). D efinition 1.6 The circle tangent to the given curve, with radius equal to the radius of curvature and centre belonging to the half-plane containing the unit vector n is called the osculating circle. Considering a generic parametrisation x = x(t), one obtains the following relations: ˙x(t) = v(t) = ˙st (1.11) and
¨ x(t) = a(t) = ¨ st + ˙s
R n, (1.12) which implies for the curvature k(t) =
1 |v(t)|
2 a(t)
− v(t)
· a(t) |v(t)|
2 v(t) .
(1.13) 10 Geometric and kinematic foundations of Lagrangian mechanics 1.3 The vectors v, a are also called the velocity and acceleration, respectively; this refers to their kinematic interpretation, when the parameter t represents time and the function s = s(t) expresses the time dependence of the point moving along the curve. We remark that, if the curvature is non-zero, and ˙s = / 0, then the normal component of the acceleration ˙s 2 /R is positive. We leave it as an exercise to verify that the curvature of the graph x 2 = f (x 1 ) is given by k(x 1 ) = |f (x 1 ) | [1 + f
2 (x 1 )] 3/2
, (1.14)
while, if the curve is expressed in polar coordinates and r = r(ϕ), then the curvature is given by k(ϕ) = |2r
2 (ϕ)
− r(ϕ)r (ϕ) + r 2 (ϕ) | [r 2 (ϕ) + r 2 (ϕ)] 3/2 . (1.15) Example 1.9 Consider an ellipse x 1
x 2 (t) = b sin t. In this case, the natural parameter s cannot be expressed in terms of t through elementary functions (indeed, s(t) is given by an elliptic integral). The velocity and acceleration are: v(t) = (
−a sin t, b cos t) = ˙st, a(t) = (−a cos t, −b sin t) = ¨st + ˙s 2 R n, and using equation (1.13) it is easy to derive the expression for the curvature. Note that v(t) · a(t) = ˙s¨s = / 0 because the parametrisation is not the natural one. T heorem 1.1 (Frenet) Let s → x(s) = (x 1 (s), x
2 (s)) be a plane curve of class at least C 3 , parametrised with respect to the natural parameter s. Then dt ds = k(s)n, dn ds = −k(s)t.
(1.16) Proof
The first formula is simply equation (1.10). The second can be trivially derived from d ds
· n) = 0, d ds (n · t) = 0. 1.3 Geometric and kinematic foundations of Lagrangian mechanics 11 We end the analysis of plane curves by remarking that the curvature function k(s) completely defines the curve up to plane congruences. Namely, ignoring the trivial case of zero curvature, we have the following. T heorem 1.2 Given a regular function k : (a, b) → R such that k(s) > 0 for every s ∈ (a, b), there exists a unique plane regular curve, defined up to translations and rotations, such that k(s) is its curvature, and s its natural parameter. Proof
The proof of this theorem depends on Frenet’s formulae and on the existence and uniqueness theorem for solutions of ordinary differential equations. Indeed, from (1.16) it follows that d 2 t ds 2 − k (s)
k(s) dt ds + k 2 (s)t = 0; (1.17) after integration this yields t = dx/ds, up to a constant vector (i.e. a rotation of the curve). One subsequent integration yields x(s) up to a second constant vector (i.e. a translation of the curve). Remark 1.6 Uniqueness can only be guaranteed if the curvature is not zero. As a counterexample, consider the two curves of class C 2 (Fig. 1.8) x(t) = (t, t 3 );
(t, t 3 ), if t < 0, (t,
−t 3 ), if t ≥ 0.
These curves are evidently distinct for t > 0, but their curvatures are equal for every t and vanish for t = 0. x 2
1 (t, t 3 ) (t, t 3 ) (t, – t 3 ) Fig. 1.8 12 Geometric and kinematic foundations of Lagrangian mechanics 1.4 1.4
Curves in R 3 We have already remarked how it is possible to define regular curves in R 3 in analogy with (1.4): such curves are maps x : (a, b) → R 3 of class C 2 , with ˙x = / 0. Consider now a curve t → x(t) = (x 1 (t), x 2 (t), x
3 (t))
∈ R 3 ; the equation defining the natural parameter is ds dt = ˙ x 2 1 + ˙ x 2 2 + ˙ x 2 3 . Suppose that the curve is parametrised through the natural parameter s. As for the case of a plane curve, we can introduce the unit tangent vector t, the unit normal vector n, and the curvature k(s) according to Definitions 1.4 and 1.5. However, contrary to the plane case, these quantities are not sufficient to fully characterise a curve in three-dimensional space. D efinition 1.7 The unit vector b = t × n
(1.18) is called a binormal unit vector. The triple of vectors (t, n, b) is orthonormal. In the case of a plane curve, it is easy to verify that db/ds = 0, and hence that the binormal unit vector is constant and points in the direction orthogonal to the plane containing the curve. Hence the derivative db/ds quantifies how far the curve is from being a plane curve. To be more precise, consider a point x(s 0 )
to the curve at x(s 0 ). The equation of the plane of the pencil with unit normal vector ν is (x − x(s 0 )) · ν = 0. The distance from such a plane of a point x(s) on the curve is given (up to sign) by
g(s) = [x(s) − x(s
0 )] · ν, and hence g (s
0 ) = t(s
0 ) · ν = 0; in addition, g (s
0 ) = k(s
0 )n(s
0 ) · ν. It follows that if n(s 0 ) is defined (i.e. if k(s 0 ) =
/ 0), there exists a unique plane such that g (s 0 ) = 0; this plane is the one whose normal vector is precisely the unit vector b(s 0 ). 1.4 Geometric and kinematic foundations of Lagrangian mechanics 13
osculating plane Fig. 1.9 D efinition 1.8 The plane normal to b(s 0 ) is called the osculating plane to the curve at the point x(s 0 ) (Fig. 1.9). Hence the osculating plane has parametric equation y = x(s
0 ) + λt(s
0 ) + µk(s
0 )n(s
0 ). (1.19) In the case of curves in space as well, we have the following. T heorem 1.3 (Frenet) Let s → x(s) = (x 1 (s), x
2 (s), x
3 (s)) be a curve in R 3 endowed with the natural parametrisation. Then the following equations hold: dt ds = +k(s)n(s), dn ds = −k(s)t(s) −χ(s)b(s), db ds = +χ(s)n(s), (1.20) where χ(s) is called the torsion (or second curvature) of the curve. The proof of Frenet’s theorem is based on the following lemma, of interest in its own right. L emma 1.1 Let A : (t 1 , t
2 ) → O(l) be a function of class C 1 , taking values in the group of orthogonal matrices l × l, such that A(t 0 ) = 1. Then ˙ A(t
0 ) is a
skew-symmetric matrix. Proof
By differentiation of the orthogonality relation A T (t)A(t) = 1 for all t ∈ (t 1
2 ), if B(t) = dA/dt (t), one obtains B T
T (t)B(t) = 0. 14 Geometric and kinematic foundations of Lagrangian mechanics 1.4 Evaluating this relation at t = t 0 , we obtain B T
0 ) =
−B(t 0 ). Proof of Theorem 1.3 Apply Lemma 1.1 to the matrix A(s − s), transforming the orthonormal triple (t(s), n(s), b(s)) to the orthonormal triple (t(s ), n(s ), b(s )). Evidently A(s − s) is orthogonal and A(0) = 1. Hence its derivative at the point s = s is a skew- symmetric matrix; equations (1.20) follow if we observe that, by definition dt/ds = k(s)n, while χ(s) is defined as the other non-zero element of the matrix A (0). The third of equations (1.20) implies that the osculating plane tends to rotate around the tangent line with velocity equal to the torsion χ(s). The second of equations (1.20) shows what causes variation in n: under the effect of curvature, the normal vector tends to rotate in the osculating plane, while under the effect of torsion it tends to follow the rotation of the osculating plane. Moreover, if χ(s) =
/ 0, the curve crosses the osculating plane. This follows from the fact that d 3 x ds 3 = d 2 t ds 2 = dk ds n − k
2 t − kχb; hence for s s 0 one has x(s) −x(s
0 ) (s −s 0 )t + 1 2 ·(s−s 0 ) 2 kn + 1 6 (s −s 0 ) 3 (k n − k 2 t − kχb), and thus (x(s) − x(s 0 ))
− 1 6 kχ(s − s
0 ) 3 . Example 1.10 Consider the cylindrical circular helix x 1 = R cos ϕ, x 2 = R sin ϕ, x 3 = λϕ. If the origin of the arcs is at A (Fig. 1.10), we have s(ϕ) = √ R
+ λ 2 ϕ; hence t =
dx dϕ dϕ ds = 1 √ R 2 + λ 2 ( −R sin ϕ, R cos ϕ, λ), dt ds = dt dϕ dϕ ds = − R R 2 + λ
2 (cos ϕ, sin ϕ, 0), from which it follows that n = (
− cos ϕ, − sin ϕ, 0), k(s) = R R 2 + λ
2 , and finally b = 1 √ R 2 + λ 2 (λ sin ϕ, −λ cos ϕ, R). It is easy to compute that db ds
− λ R 2 + λ
2 n,
1.5 Geometric and kinematic foundations of Lagrangian mechanics 15
3
2
1
n A w Fig. 1.10 yielding for the torsion χ =
− λ R 2 + λ
2 . Curvature and torsion are the only two geometric invariants of a curve in space. Namely we have the following. T heorem 1.4 Let k(s) > 0 and χ(s) be two given regular functions. There exists a unique curve in space, up to congruences (rotations and translations), which has s as natural parameter, and k and χ as curvature and torsion, respectively. The proof is similar to the proof of Theorem 1.2 and is based on the fact that t(s) solves the differential equation d 2
ds 2 − k k dt ds + k
2 t + χt
× dt ds = 0. (1.21)
1.5 Vector fields and integral curves In complete analogy with (1.4), a regular curve in R l is a map x : (a, b) → R l of class C 1 such that ˙x = / 0. In this section we shall investigate the relation between curves and vector fields.
16 Geometric and kinematic foundations of Lagrangian mechanics 1.6 D
l . A vector field X on U is a regular function X : U → R
l (e.g. of class C ∞
∈ U a vector X(x) of R l , which is said to be applied at the point x. Example 1.11 To every regular function f : U → R one can associate the gradient vector field X(x) =
∇f(x) = ∂f ∂x 1 (x), . . . , ∂f ∂x
(x) . The gradient vector field is orthogonal to the level sets of f . D efinition 1.10 A curve x : (a, b) → R l is called an integral curve of a vector field X : U → R
l if for all t ∈ (a, b) the following conditions hold: (a) x(t)
∈ U; (b) ˙x(t) = X(x(t)). Example 1.12 Consider the vector field in R 2 defined by X(x 1 , x
2 ) = (x
2 , −x 1 ). The integral curve of the field passing through (x 1 (0), x 2 (0)) at t = 0 is given by x 1
1 (0) cos t + x 2 (0) sin t, x 2 (t) = −x 1 (0) sin t + x 2 (0) cos t. Note that, if (x 1 (0), x 2 (0)) = (0, 0), the integral curve is degenerate at the point (0, 0). This is possible because at the point (0, 0) the vector field vanishes, i.e. it has a singular point. It evidently follows from Definition 1.10 that the existence and uniqueness theorem for ordinary differential equations ensures the existence of a unique integral curve of a vector field passing through a given point. The question of the continuation of solutions of differential equations (hence of the existence of a maximal integral curve) yields the following definition. D efinition 1.11 A vector field is called complete if for every point x the maximal integral curve (cf. Appendix 1) passing through x is defined over all of R. Example 1.13 The vector field given in Example 1.12 is complete. The field X : R → R,
X(x) = 1 + x 2 is not complete. When not otherwise stated, we shall implicitly assume that the vector fields considered are complete. 1.6 Surfaces
The study of the local properties of plane curves, which we considered in the first three sections of this chapter, is rather simple: one invariant—curvature (as 1.6 Geometric and kinematic foundations of Lagrangian mechanics 17 a function of arc length)—is sufficient to characterise the curve. Matters are not much more complicated in the case of curves in R 3 . The essential reason for this is that the intrinsic geometry of curves is ‘trivial’, in the sense that for all curves there exists a natural parametrisation, i.e. a map x(s) from an interval (a, b) of R to the curve, such that the distance between any two points x(s 1 ) and x(s 2 ) of the curve, measured along the curve, is equal to |s 2 − s 1 |. Hence the metric (i.e. the notion of distance) defined by means of the arc length coincides with that of R. The situation is much more complicated for the case of surfaces in R 3 . We shall see that the intrinsic geometry of surfaces is non-trivial due to the fact that, in general, there is no isometry property between surfaces and subsets of R 2
metric using just one scalar function. In analogy with the definition of a curve in the plane (as the level set of a function of two variables), surfaces in R 3 can be obtained by considering the level sets of a function F : U → R (for simplicity, we assume that this function is of class C ∞ , but it would be sufficient for the function to be of class C 2 ), where U is an open subset of R 3 . The surface S is hence defined by S = {(x
1 , x
2 , x
3 ) ∈ U|F (x 1 , x
2 , x
3 ) = 0
}, (1.22)
assuming that such a set is non-empty. D efinition 1.12 A point (x 1 , x
2 , x
3 ) of the surface F (x 1 , x
2 , x
3 ) = 0 is called non-singular if the gradient of F computed at the point is non-vanishing: ∇F (x
1 , x
2 , x
3 ) =
/ 0. (1.23)
A surface S whose points are all non-singular points is called regular. By the implicit function theorem, if P is non-singular, in a neighbourhood of P the surface can be written as the graph of a function. For example, if (∂F/∂x
3 ) P = / 0 there exists a regular function f : U → R (where U is an open neighbourhood of the projection of P onto the (x 1 , x
2 ) plane) such that S = graph (f ) = {(x
1 , x
2 , x
3 ) ∈ R 3 |(x
1 , x
2 ) ∈ U, x 3 = f (x
1 , x
2 ) }. (1.24) In addition, from F (x 1 , x
2 , x
3 ) = 0 it follows that ∂F ∂x
dx 1 + ∂F ∂x 2 dx 2 + ∂F ∂x 3 dx 3 = 0; hence from F (x 1 , x 2 , f (x
1 , x
2 )) = 0 it follows that ∂f ∂x
= − ∂F/∂x 1 ∂F/∂x
3 , ∂f ∂x 2 = − ∂F/∂x
2 ∂F/∂x
3 .
18 Geometric and kinematic foundations of Lagrangian mechanics 1.6 The analogous analysis can be performed if (∂F/∂x 2 ) P = / 0, or (∂F/∂x 1 )
= / 0.
Equation (1.24) highlights the fact that the points of a regular surface are, at least locally, in bijective and continuous correspondence with an open subset of R 2
It is an easy observation that at a non-singular point x 0 there exists the tangent plane, whose equation is (x − x 0 ) · ∇F = 0. More generally, it is possible to consider a parametric representation of the form x : U → R 3
2 : S = x(U ) = {(x 1 , x 2 , x
3 ) ∈ R 3 |there exist (u, v) ∈ U, (x 1 , x
2 , x
3 ) = x(u, v) }. (1.25)
Note that the graph of (1.24) is a particular case of the expression (1.25), in which the parametrisation is given by x(u, v) = (u, v, f (u, v)). It is always possible to transform (1.24) into (1.25) by the change of variables on the open set U of R 2 , x
1 = x
1 (u, v), x 2 = x
2 (u, v), provided the invertibility condition det [∂(x 1 , x 2 )/∂(u, v)] = / 0 holds. The latter condition expresses the fact that the coordinate lines u = constant and v = constant in the (x 1 , x 2 ) plane are not tangent to each other (Fig. 1.11). It follows that Definition 1.12 is equivalent to the following.
3
2
1
v = constant u = constant u = constant Fig. 1.11 1.6 Geometric and kinematic foundations of Lagrangian mechanics 19 D
point P is called non-singular if rank
⎛ ⎜ ⎝ ∂x 1 ∂u ∂x 2 ∂u ∂x 3 ∂u ∂x 1 ∂v ∂x 2 ∂v ∂x 3 ∂v ⎞ ⎟ ⎠ P = 2.
(1.26) Equation (1.26) is equivalent to requiring that the vectors x u , x
v are linearly independent. Example 1.14 The sphere of radius R > 0 is a regular surface; it is the level set of F (x
1 , x
2 , x
3 ) = x
2 1 + x 2 2 + x 2 3 − R 2 . A parametrisation of the sphere is given by x(u, v) = R(cos v sin u, sin v sin u, cos u), where (u, v) ∈ [0, π] × [0, 2π]. Here v is also called the longitude, and u the colatitude, as it is equal to π/2 minus the latitude (Fig. 1.12). This parametrisation of the sphere is regular everywhere except at the two poles (0, 0, ±1). The sphere of radius 1 is usually denoted S 2 . x 3
2
1
u v Fig. 1.12 20 Geometric and kinematic foundations of Lagrangian mechanics 1.6 Example 1.15 The ellipsoid is a regular surface; it is the level set of F (x
1 , x
2 , x
3 ) =
x 2 1 a 2 + x 2 2 b 2 + x 2 3 c 2 − 1, where a > b > c > 0 are the semi-axes of the ellipsoid. A parametrisation is given by
x(u, v) = (a cos v sin u, b sin v sin u, c cos u), with (u, v) ∈ [0, π] × [0, 2π]. Note that this parametrisation is not regular at the points (0, 0, ±c); however at these points the surface is regular. Example 1.16 The one-sheeted hyperboloid, level set S = F −1 (0) of F (x 1 , x 2 , x
3 ) =
x 2 1 a 2 + x 2 2 b 2 − x 2 3 c 2 − 1, or the two-sheeted hyperboloid with F (x
1 , x
2 , x
3 ) =
− x 2 1 a 2 − x 2 2 b 2 + x 2 3 c 2 − 1, are regular surfaces. A parametric representation is given, respectively, by x(u, v) = (a cos v cosh u, b sin v cosh u, c sinh u), and
x(u, v) = (a cos v sinh u, b sin v sinh u, c cosh u), where (u, v) ∈ R × [0, 2π]. Example 1.17 A particularly interesting class of surfaces is given by the surfaces of revolution; these surfaces are obtained by rotating, e.g. around the x 3 -axis, a curve (implicitly defined) in the (x 1 , x 3 ) plane. If f (x 1 , x
3 ) = 0 is the implicit representation of the curve, the surface of revolution corresponds to the level set of the function F (x
1 , x
2 , x
3 ) = f (
x 2 1 + x 2 2 , x 3 ) = 0. Among the previous examples, we have already encountered surfaces of revolution, e.g. the ellipsoids (if two of the semi-axes are equal) or the hyperboloids (if a = b). A parametric representation of the surfaces of revolution is given by x(u, v) = (u cos v, u sin v, f (u)), if the generating curve has equation x 3 = f (x 1 ).
1.6 Geometric and kinematic foundations of Lagrangian mechanics 21 Example 1.18 The elliptic paraboloid is the graph of x 3 = x 2 1 a 2 + x 2 2 b 2 , a > b > 0, (x 1
2 ) ∈ R 2 , while the hyperbolic paraboloid is the graph of x 3 = x 2 1 a 2 − x 2 2 b 2 , a > b > 0, (x 1 , x 2 ) ∈ R 2 . Remark 1.7 In analogy with the definition of surfaces in R 3 one can introduce (hyper)surfaces in R l , as: (1) level sets of functions from (subsets of) R l into R; (2) graphs of functions defined in an open subset of R l −1 and taking values in R; (3) through a parametric representation, with l − 1 parameters x(u 1 , . . . , u l −1 ). In this section we will focus primarily on studying surfaces in R 3 , while in the next section we shall define the notion of a differentiable manifold, of which surfaces and hypersurfaces are special cases. Let F : U → R be a C ∞ function, U an open subset of R 3 , and denote by S the surface S = F −1 (0). It is important to remark that, in general, it is not possible to find a natural parametrisation that is globally non-singular for the whole of a regular surface. Example 1.19 The bidimensional torus T 2 is the surface of revolution around the x 3 -axis
obtained from the circle in the (x 1 , x 3 ) plane, given by the equation x 2
+ (x 1 − a) 2 = b
2 , thus with centre x 1 = a, x
3 = 0 and radius b, such that 0 < b < a. Hence its implicit equation is F (x
1 , x
2 , x
3 ) = x
2 3 + ( x 2 1 + x 2 2 − a) 2 − b 2 = 0.
It is easy to verify that a parametrisation of T 2 is given by x 1 = cos v(a + b cos u), x 2 = sin v(a + b cos u), x 3 = b sin u, where (u, v) ∈ [0, 2π] × [0, 2π] (Fig. 1.13). The torus T 2 is a regular surface. Indeed, ∇F (x
1 , x
2 , x
3 ) =
2x 1 − 4ax 1 x 2 1 + x 2 2 , 2x 2 − 4ax 2 x 2 1 + x
2 2 , 2x 3 = / 0 on T 2 ,
22 Geometric and kinematic foundations of Lagrangian mechanics 1.6
3
2
1
b u v Fig. 1.13 and correspondingly ∂(x
1 , x
2 , x
3 ) ∂(u, v) = −b sin u cos v −b sin u sin v b cos u
−(a + b cos u) sin v (a + b cos u) cos v 0 has rank 2 on T 2 . Example 1.20 The sphere S 2 x 2 1 + x 2 2 + x 2 3 − 1 = 0 is a regular surface; the parametrisation x(u, v) = (cos v cos u, sin v cos u, sin u) in non-singular everywhere except at the points u = ±π/2 (corresponding to the north pole x = (0, 0, 1) and the south pole x = (0, 0, −1) of the sphere) where the parametrisation is singular (this is intuitively evident by observing that the parallels degenerate to a point at the poles, and hence that the longitude is not defined at these points). However, the parametrisation x(u, v) = (sin u, cos v cos u, sin v cos u) is regular at the poles, while it is singular at x = ( ±1, 0, 0). The stereographic projection from one of the poles of the sphere (cf. Example 1.29) is an example of a parametrisation that is regular over the whole sphere minus one point. There is no global regular parametrisation of the whole sphere.
1.6 Geometric and kinematic foundations of Lagrangian mechanics 23 Example 1.21 The cone x 2 1 a 2 + x 2 2 b 2 − x 2 3 c 2 = 0 is not a regular surface: the origin x 1 = x
2 = x
3 = 0 belongs to the cone but it is a singular point. Excluding this point, the surface becomes regular (but it is no longer connected), and x(u, v) = (au cos v, bu sin v, cu) is a global non-singular parametrisation. Consider a surface S = F −1 (0), and a regular point P ∈ S. At such a point it is possible to define the tangent space T P S to the surface S at the point P . D efinition 1.14 A vector w ∈ R 3 at the point P is said to be tangent to the surface S at the point P , or w ∈ T
P S (tangent space to the surface at the point P ) if and only if there exists a curve x(t) on the surface, i.e. such that F (x
1 (t), x
2 (t), x
3 (t)) = 0 for all t, passing through the point P for some time t 0 ,
0 ) = P , with velocity ˙x(t 0 ) = w.
In the expression for the tangent vector at a point x(u 0 , v 0 ) ˙x = x u ˙ u + x v ˙v (1.27) we can consider ˙ u, ˙v as real parameters, in the sense that, given two numbers α, β, it is always possible to find two functions u(t), v(t) such that u(t 0 ) = u 0 , v(t 0 ) = v
0 , ˙
u(t 0 ) = α, ˙v(t 0 ) = β. Hence we can identify T p S with the vector space, of dimension 2, generated by the vectors x u , x v (Fig. 1.14). ∇F ∇F v = constant u = constant x u x v Fig. 1.14 24 Geometric and kinematic foundations of Lagrangian mechanics 1.6 D
every point P of the surface, a vector X(P ) ∈ R
3 applied at the point P . The field X is called a tangent field if X(P ) ∈ T
P S for every P ∈ S; the field is a normal field if X(P ) ∈ (T P
⊥ for every point P ∈ S. Remark 1.8 Since a vector field tangent to S is expressed by X = X 1 (u, v)x u + X
2 (u, v)x
v , the equations of its integral curves are ˙ u = X 1 (u, v), ˙v = X 2 (u, v) and the curves lie on S. T heorem 1.5 Let P be a non-singular point of the surface F (x 0 ) = 0. Then the tangent space to the surface at P coincides with the orthogonal space to the gradient of F at P : T P
∇F (P )) ⊥ . (1.28) Proof
Differentiating the expression F (x(u, v)) = 0 we obtain ∇F · x
u = ∇F · x v = 0.
Hence ∇F is orthogonal to every vector of T p S. Conversely, if w is ortho- gonal to ∇F at P ∈ S, it must necessarily belong to the plane generated by x u
v . D efinition 1.16 A connected surface S is said to be oriented when a unitary normal vector field is uniquely assigned on the surface. Remark 1.9 The regular surfaces we have defined (as level sets S = F −1 (0)) are always ori- entable, with two possible orientations corresponding to the two unitary normal vector fields n 1
∇F (P ) |∇F (P )| , n
(P ) = − ∇F (P ) |∇F (P )| . (1.29) However, it is possible in general to extend the definition of surface to also admit non-orientable cases, such as the M¨ obius strip (Fig. 1.15). For applications in mechanics, it is very important to be able to endow the surface with a distance or metric, inherited from the natural immersion in three-dimensional Euclidean space. To this end, one can use the notion of length of a curve in space, using the same definition as for curves lying on a surface. If S = F −1 (0) is a regular surface, x = x(u, v) is a parametric representation for it, and t → (u(t), v(t)), t ∈ (a, b) is a curve on S, the length of the curve is given by (cf. (1.5)) l =
b a dx(u(t), v(t)) dt dt =
b a (x u ˙ u + x v ˙v)
· (x u ˙ u + x v ˙v) dt. (1.30) 1.6 Geometric and kinematic foundations of Lagrangian mechanics 25
Fig. 1.15 M¨obius strip. If we define E(u, v) = x u · x
u = ∂x 1 ∂u 2 + ∂x 2 ∂u 2 + ∂x 3 ∂u 2 , F (u, v) = x u · x
v = ∂x 1 ∂u ∂x 1 ∂v + ∂x 2 ∂u ∂x 2 ∂v + ∂x 3 ∂u ∂x 3 ∂v , G(u, v) = x v · x
v = ∂x 1 ∂v 2 + ∂x 2 ∂v 2 + ∂x 3 ∂v 2 , (1.31) equation (1.30) can be rewritten as l =
b a E(u(t), v(t)) ˙ u 2 + 2F (u(t), v(t)) ˙ u ˙v + G(u(t), v(t)) ˙v 2 dt. (1.32) Setting
(ds) 2 = dx · dx, (1.33)
we obtain for (ds) 2 the expression (ds) 2 = E(u, v)(du) 2 + 2F (u, v)(du)(dv) + G(u, v)(dv) 2 .
D efinition 1.17 The quadratic form (1.34) is called the first fundamental form of the surface.
26 Geometric and kinematic foundations of Lagrangian mechanics 1.6 This form fixes the metric on the surface, as it makes it possible to compute lengths. Remark 1.10 The expression (1.34) represents a positive definite quadratic form: this means E > 0 and EG − F 2
u , x
v (linearly independent) is exactly √ EG
2 . Example 1.22 Consider the sphere S 2 of radius 1, parametrised by x = (cos v sin u, sin v sin u, cos u). Then
x u = (cos u cos v, cos u sin v, − sin u), x v = ( − sin u sin v, sin u cos v, 0), and hence E = 1, F = 0,
G = sin 2 u, from which it follows that (ds)
2 = (du)
2 + sin
2 u(dv)
2 . For example, the length of a parallel at colatitude u 0 is given by l = 2π
˙ u 2 + (sin u 0 ) 2 ˙v 2 dt = 2π sin u 0 , since the curve has parametric equations u = u 0 , v = t. Having defined the first fundamental form, it is possible to compute not only the lengths of curves lying on the surface, but also the angle ϕ between two intersecting curves: if their parametric representation is u = u
1 (t), v = v 1 (t)
and u = u
2 (t), v = v 2 (t)
(1.35) and the intersection point is denoted by P , corresponding to the value t = t 0 ,
w 1 = ˙ u 1 (t 0 )x u (u 1 (t 0 ), v
1 (t 0 )) + ˙v 1 (t 0 )x v (u 1 (t 0 ), v
1 (t 0 )), w 2 = ˙ u 2 (t 0 )x u (u 2 (t 0 ), v 2 (t 0 )) + ˙v 2 (t 0 )x v (u 2 (t 0 ), v
2 (t 0 )) are both tangent to the surface at the point P . The angle between the two vectors is given by cos ϕ =
w 1 · w 2 |w 1 ||w 2 | = E ˙
u 1 ˙ u 2 + F ( ˙ u 1 ˙v 2 + ˙v
1 ˙ u 2 ) + G ˙v
1 ˙v 2 E ˙ u 2 1 + 2F ˙
u 1 ˙v 1 + G ˙v
2 1 E ˙ u 2 2 + 2F ˙ u 2 ˙v 2 + G ˙v 2 2 . (1.36) 1.6 Geometric and kinematic foundations of Lagrangian mechanics 27 Remark 1.11 The parametrisation of a surface is called orthogonal if F = 0: (ds)
2 = E(u, v)(du) 2 + G(u, v)(dv) 2 . In this case the curves x(u, v 0 ), x(u
0 , v) on the surface, obtained by fixing one of the two parameters, are mutually orthogonal. If in addition E = G = g(u, v), and hence (ds) 2
2 + (dv)
2 ), the parametrisation is called conformal, since the angle in (1.36) between the two curves on the surface is equal to the angle between the two curves (1.35) in the (u, v) plane. It can be proved (cf. Dubrovin et al. 1991a,b) that given a regular surface, there always exist orthogonal as well as conformal coordinates. Moreover, the first fundamental form allows one to compute the area of the surface. Consider the tangent parallelogram defined by the vectors x u ∆ u and x v ∆ v. The total area of this parallelogram is given by |x u
u × x
v ∆ v | = |x u × x v | |
∆ u ∆ v | = EG − F 2 |
u ∆ v |. The area of the part S D of the surface corresponding to the parameters (u, v) varying within a bounded domain D is area (S
D ) =
D EG − F 2 du dv.
(1.37) A very important feature of the first fundamental form of a surface is how it behaves under coordinate transformations. T heorem 1.6 The first fundamental form is a covariant tensor of rank 2 (cf. Appendix 4). Proof
Let (u , v ) be a new parametrisation of the surface. From the identities du =
∂u ∂u du + ∂u ∂v dv , dv = ∂v ∂u du + ∂v ∂v dv it follows immediately that (ds) 2
E F F G du dv = (du dv )J T E F F G J du dv , (1.38)
where J =
⎛ ⎜ ⎝ ∂u ∂u ∂u ∂v ∂v ∂u ∂v ∂v ⎞ ⎟ ⎠ ,
E F F G = J
T E F F G J, and E , F , G are expressed in terms of the new parameters. 28 Geometric and kinematic foundations of Lagrangian mechanics 1.6 Among all the possible curves on a surface, the class of geodesics deserves special attention. Indeed, we shall see that geodesic curves play a very important role in mechanics. Let S be a regular surface, and x(u, v) its parametric representation. Consider a curve on the surface parametrised with respect to the natural parameter s: s → (u(s), v(s)) → x(u(s), v(s)). (1.39) The unit vector t tangent to the curve is given by t(s) = dx
(s) = u (s)x u (u(s), v(s)) + v (s)x v (u(s), v(s)) ∈ T x(u(s),v(s)) S and the normal unit vector n is given by n(s) = 1
d 2 x ds 2 = 1 k(s)
(u x u + (u ) 2 x uu + 2u v x uv + (v ) 2 x vv + v x v ), (1.40) where k(s) is the curvature, x uu = ∂ 2 x ∂u 2 , x uv = ∂ 2 x ∂u∂v
, x vv = ∂ 2 x ∂v 2 . D efinition 1.18 The curve (1.39) is called a geodesic if at every point of the curve the unit vector n normal to the curve belongs to the space normal to the surface, i.e. if n(s) ∈ (T
x(u(s),v(s)) S) ⊥ (1.41) for all s, and hence if and only if n(s) · x
u (u(s), v(s)) = 0, n(s) · x
v (u(s), v(s)) = 0. (1.42) Remark 1.12 Given a curve with an arbitrary parametrisation, denoting by s = s(t) the time dependence, its acceleration a is given by the expression (1.12), and the condition for this curve to be a geodesic consists in this case of imposing the condition that the acceleration be orthogonal to the surface. The condition for a curve in the Euclidean space R 3 to be a geodesic is satisfied by straight lines, for which d 2 x/ds 2 = 0.
Example 1.23 It is easy to convince oneself that the maximal circles are geodesics on the sphere, while on a cylinder with circular normal section, the geodesics are the generating lines and helices (cf. Example 1.10), including the ones that degenerate to circles.
1.6 Geometric and kinematic foundations of Lagrangian mechanics 29 From equations (1.40) and (1.42) it is easy to derive a system of ordinary differential equations which the geodesics must satisfy: (u x
u + (u )
2 x uu + 2u v x uv + (v ) 2 x vv + v x v ) · x u = 0, (u x u + (u ) 2 x uu + 2u v x uv + (v ) 2 x vv + v x v ) · x v = 0. (1.43) Recall that E = x u · x
u , F = x
u · x
v and G = x v · x
v , and note that ∂E ∂u
uu · x
u , ∂E ∂v = 2x
uv · x
u , ∂F ∂u = x
uv · x
u + x
uu · x
v , ∂F ∂v = x
uv · x
v + x
vv · x
u , ∂G ∂u = 2x
uv · x
v , ∂G ∂v = 2x
vv · x
v ; hence equations (1.43) become Eu + F v + 1 2 ∂E ∂u (u ) 2 + ∂E ∂v u v +
∂F ∂v − 1 2 ∂G ∂u (v )
2 = 0,
F u + Gv + 1 2 ∂G ∂v (v ) 2 + ∂G ∂u u v +
∂F ∂u − 1 2 ∂E ∂v (u )
2 = 0.
(1.44) Denoting the matrix representing the first fundamental form by (g ij
E F F G , (1.45) and its inverse by (g kl ) = 1 EG − F 2 G −F −F E , (1.46)
we can introduce the so-called Christoffel symbols Γ k ij = 1 2 2 l =1 g kl ∂g lj ∂u i + ∂g il ∂u j − ∂g ij ∂u l , (1.47) where u
1 = u, u
2 = v. Using Christoffel symbols, one finds that the system of differential equations (1.44) for the geodesics can be written in the form d 2 u k ds 2 + 2 i,j =1 Γ k ij du i ds du j ds = 0, k = 1, 2. (1.48)
Example 1.24 For a cylinder with generic section x 1 = f
1 (v), x
2 = f
2 (v), x
3 = u and (f 1 )
+ (f 2 ) 2 = 1, one obtains E = G = 1, F = 0 and equations (1.44) yield u = v = 0, i.e. u = as + b, v = cs + d, with a, b, c, d arbitrary constants. When c = 0 one 30 Geometric and kinematic foundations of Lagrangian mechanics 1.6 obtains the generating lines; a = 0 yields the normal sections; in all other cases v − d = c/a (u − b), and hence one finds helices. Since du/ds = a, the geodesics intersect the generating lines at a constant angle. Example 1.25 The first fundamental form of a surface of revolution with the parametrisation x = (u cos v, u sin v, f (u)) can be written as (ds) 2
2 ](du)
2 + u
2 (dv)
2 , (1.49) and hence the Christoffel symbols have the values Γ 1 11 = f (u)f (u) 1 + (f (u)) 2 , Γ 1 22 = − u 1 + (f (u)) 2 , Γ 2 12 = Γ 2 21 = 1 u , while Γ 1 12 = Γ 1 21 = Γ 2 11 = Γ 2 22 = 0. The geodesic equation (1.48) on the surface is thus equivalent to the system d 2
ds 2 + f (u)f (u) 1 + (f (u)) 2 du
2 − u 1 + (f (u)) 2 dv ds 2 = 0, d 2 v ds 2 + 2 u du ds dv ds = 0. (1.50)
The second of equations (1.50) can be rewritten as 1 u 2 d ds u 2 dv ds = 0,
from which it follows that there exists a constant c ∈ R such that for every s u 2
ds = c,
(1.51) and hence, if c = / 0, ds =
1 c u 2 dv.
Substituting the latter expression into the first fundamental form (1.49) one obtains the relation u 4
2 = c
2 [1 + (f (u)) 2 ](du)
2 + c
2 u 2 (dv) 2 ; (1.52) this leads to the elimination of ds and one can hence consider v as a function of u. The geodesics on a surface of revolution thus have the implicit form v − v 0 = ±c u u 0 1 + (f (ξ)) 2 ξ ξ 2 − c 2 dξ.
(1.53) If c = 0, from equation (1.51) it follows that u 2 (dv/ds) = 0, i.e. that v is constant: the meridians are geodesic curves. On the other hand, the parallels 1.6 Geometric and kinematic foundations of Lagrangian mechanics 31 (the curves corresponding to u = constant) are geodesics only if u 1 + (f (u)) 2 dv
2 = 0,
d 2 v ds 2 = 0, i.e. only if dv/ds is in turn constant, and if dx 3 /du = f (u) = ∞, which implies that along the given parallel, the planes tangent to the surface envelop a cylinder whose generator lines are parallel to the x 3 -axis. The relation (1.51) has an interesting consequence. Let α be the angle between the geodesic (u(s), v(s)) at s = s
0 and the meridian v = v(s 0 ) (Fig. 1.16). It is immediate to verify that u(s 0 ) dv ds (s 0 ) = sin α, x 3 (m) v f (u) v O ( p): parallel, ( m): meridian, ( g): geodesic ( p) (g) a Fig. 1.16 Geodesics on a surface of revolution. 32 Geometric and kinematic foundations of Lagrangian mechanics 1.6 since the unit vector tangent to the parallel is simply ( − sin v, cos v, 0); hence substituting in the expression (1.51) we obtain Clairaut’s theorem: u(s) sin α(s) = c. (1.54)
Hence the geodesic must lie in the region u(s) ≥ |c|.
In the case of a surface of revolution, with a cusp at infinity, i.e. such that lim
u →0 f (u) = ∞ (Fig. 1.17), every geodesic, after attaining the minimum value of u allowed by equation (1.54), reverses the motion (along the x 3 -axis) and comes back into the region corresponding to values of u satisfying |u| > |c|. Fig. 1.17 Reversal of geodesics on a surface of revolution.
1.7 Geometric and kinematic foundations of Lagrangian mechanics 33 It is possible to prove that geodesics on a surface of revolution which are neither meridians nor closed curves are dense in the region u ≥ |c|.
1.7 Differentiable Riemannian manifolds Let x be a point in Euclidean n-dimensional space R n , and let f 1 , . . . , f m be m
regular real-valued functions defined on the same connected open subset A ⊂ R
n . Just as the level set of a real function of three real variables identifies a surface in Euclidean three-dimensional space, the level sets of any of the functions f j identify a (hyper)surface in R n . With the requirement that x lies in the intersection (supposed non-empty) of the level sets of all the functions f j , one identifies a submanifold of R n . In analogy with the notion of a regular surface introduced in the previous section, as a surface endowed with a tangent plane to all of its points, we can introduce the notion of a regular submanifold of R n by
(and a normal space). The dimension of the submanifold is then defined as the dimension of its tangent space. These sketchy introductory remarks justify the following definition. D efinition 1.19 Let A be an open connected subset of R n , n > 1, and f : A → R
−l , 1
≤ l < n, a map of class C k , k ≥ 2. The zero level set V = {x ∈ A|f(x) = 0} of f , assumed non-empty, is called a regular submanifold of R n of class
C k and of dimension l if the Jacobian matrix of the map f is of maximal rank (hence if its rank is equal to n − l) at every point of V . Remark 1.13 Evidently the condition that the Jacobian matrix of f = (f 1 , . . . , f n −l ) be of rank n − l at every point of V is equivalent to requiring that the gradient vec- tors ∇
f 1 , . . . , ∇ x f n −l be an (n − l)-tuple of vectors in R n which are linearly independent on V . Consider as an example the case shown in Fig. 1.18, for which n = 3, l = 1, f = (f 1
2 ), where
f 1 (x 1 , x
2 , x
3 ) = x
3 − x
2 1 + x 2 2 , f 2 (x 1 , x
2 , x
3 ) = x
2 1 + x 2 2 + x 2 3 − 1. The set V is a circle. Note that the vectors ∇f 1 = −x 1 x 2 1 + x 2 2 , −x 2 x 2 1 + x 2 2 , 1 , ∇f 2 = 2(x
1 , x
2 , x
3 ) are linearly independent on V . This definition includes in particular plane regular curves (n = 2, l = 1), regular curves in R 3 (n = 3, l = 1), considered as the intersection of two non-tangential surfaces, and regular surfaces in R 3 (n = 3, l = 2). 34 Geometric and kinematic foundations of Lagrangian mechanics 1.7
3
1
2 ∇F 2 ∇F 1 1
Fig. 1.18 D efinition 1.20 The tangent space T P V to a regular submanifold V at the point P is the l-dimensional vector space of the velocities ˙x(t 0 ) along the curves x(t) belonging to V (hence such that f (x(t)) = 0 for every t) and passing through P for t = t 0 .
It is easy to verify that T P V coincides with the vector space generated by the vectors which are orthogonal to the gradients ∇ x f 1 (P ), . . . , ∇ x f n −l (P ) (cf. Theorem 1.5). The latter will be called a basis of the normal space to V in P . Having chosen a local parametrisation x = x(u 1 , . . . , u l ) of V , whose existence is guaranteed by the implicit function theorem, the tangent space at a point P of V has as a basis the vectors x u 1
u l , where x u i = ∂x ∂u i (1.55)
and derivatives are computed at the point P . Example 1.26 The sphere S l of unit radius is the regular submanifold of R l +1 defined by f (x 1 , . . . , x l +1 ) = x 2 1 + · · · + x 2 l +1 − 1 = 0.
The tangent space at one of its points P , with coordinates (x 1 , . . . , x l +1 ), is the hyperplane of R l +1 described by the equation x · x = 0. 1.7 Geometric and kinematic foundations of Lagrangian mechanics 35 Example 1.27 The group of real n × n matrices A with unit determinant, denoted by SL(n, R), is a regular submanifold of R n 2 of dimension n 2 − 1, defined by the equation det(A) = 1. Its tangent space at the point corresponding to the identity matrix can be identified with the space of n × n matrices of zero trace. Indeed, if A(t) is any curve in SL(n, R) passing through the identity at t = 0, and thus such that A(0) = 1, we have that 0 = d
det A(t) | t =0 = Tr ˙
A(0). Indeed, if we set X = ˙ A(0) we have that det A(t) = det(1 + tX) + O (t) =
1 + t Tr X + O (t).
Example 1.28 The group of real orthogonal n × n matrices A, denoted by O(n), is a regular submanifold of R n 2
− 1)/2 defined by the system of equations AA T = 1. Its tangent space at the point corresponding to the identity matrix can be identified with the vector space of n ×n skew-symmetric matrices (cf. Lemma 1.1). The connected component of O(n) containing the identity matrix coincides with the group SO(n) of orthogonal matrices of unit determinant. We now turn to the problem of parametrising regular submanifolds. We have already remarked that for surfaces in R 3 it is not possible in general to give a global parametric representation. For example, the sphere S 2 is a regular submanifold of R 3 , but the parametrisation given by the spherical coordinates x 1 = (sin u 1 cos u
2 , sin u
1 sin u
2 , cos u
1 ) is singular at the points (0, 0, 1) and (0, 0, −1). A regular parametrisation at those points is given instead by x 2 = (cos u 1 , sin u
1 cos u
2 , sin u
1 sin u
2 ), which however is singular at (1, 0, 0) and ( −1, 0, 0). Hence there exist two regular injective maps x 1 , x 2 defined on R = (0, π) × [0, 2π) such that S 2 = x
1 (R)
∪ x 2 (R). Moreover, if we consider the intersection W = x 1 (R) ∩ x 2 (R) = S 2 \{(0, 0, 1), (0, 0, −1), (1, 0, 0), (−1, 0, 0)}, the preimages x −1
(W ) = R \{(π/2, 0), (π/2, π)} and x −1 2
\{(π/2, π/2), (π/2, 3π/2)} are set in one-to-one correspondence by the map x −1 2
1 , which expresses u 1 , u
2 as functions of u 1 , u
2 , and by its inverse x −1 1
2 . In summary, these are the properties of any ‘good’ parametrisation of a regular submanifold. We can now consider the problem of parametric representation in a more general context, by referring to a set M which is not necessarily endowed with a metric structure, as in the case of regular submanifolds of R n . D efinition 1.21 A differentiable manifold of dimension l and class C k consists
of a non-empty set M and of a family of injective maps x α : U α ⊂ R
l → M, with 36 Geometric and kinematic foundations of Lagrangian mechanics 1.7
Download 10.87 Mb. Do'stlaringiz bilan baham: |
ma'muriyatiga murojaat qiling