Analytical Mechanics This page intentionally left blank
Download 10.87 Mb. Pdf ko'rish
|
2 , vanishing on S and with a very large constant k. The Lagrangian corresponding to this system is given by ˆ L(x, ˙x) = m 2 ( ˙ x 2 1 + ˙ x 2 2 + ˙
x 2 3 ) − V (x
1 , x
2 , x
3 ) − 1 2 k(f (x 1 , x
2 , x
3 )) 2 . (A5.4)
Let x(t, k) be a solution of the Lagrange equations associated with (A5.4) with initial conditions x(0, k) = x 0 ∈ S, ˙x(0, k) = ˙x 0 ∈ T x 0 S, (A5.5) independent of k. If x = x(q 1 , q
2 ) is a local parametrisation of S in a neighbour- hood of x 0 , let (q 1 (0), q
2 (0), ˙
q 1 (0), ˙ q 2 (0)) be the local coordinates corresponding to the initial conditions (A5.5). To compare with (A5.4) we then consider the Lagrangian of the system subject to the constraint, given by L(q, ˙q) = m 2 (E(q 1 , q 2 ) ˙
q 2 1 + 2F (q 1 , q 2 ) ˙
q 1 ˙ q 2 + G(q 1 , q
2 ) ˙
q 2 2 ) − V (x(q
1 , q
2 )) (A5.6) (where E, F , G are the coefficients of the first fundamental form associated with S) and the solution q 1 (t), q
2 (t) of the Lagrange equations associated with (A5.6) and corresponding to initial conditions (q 1 (0), q 2 (0), ˙
q 1 (0), ˙ q 2 (0)). The Lagrangian (A5.6) is obtained from (A5.4) by imposing the restrictions x ∈ S, ˙x ∈ T x S and
introducing local coordinates. Under the above assumptions the following important result holds. T heorem A5.1 (Rubin, Hungar, Takens) Fix T > 0. Then for every t ∈ [0, T ] the limit ˆ x(t) = lim k →∞ x(t, k) (A5.7) exists. The function t → ˆx(t) is of class C 2 , ˆ x(t) belongs to S for every t ∈ [0, T ] and hence can be written using the local parametrisation of S as ˆ x(t) = x(q 1 (t), q
2 (t)).
(A5.8) Here (q
1 (t), q
2 (t)) is a solution of the Lagrange equations associated with the Lagrangian (A5.6) and satisfying the initial conditions (q 1 (0), q 2 (0), ˙
q 1 (0), ˙ q 2 (0)) associated with (A5.5) by the parametrisation. We shall not give the proof of this theorem (see Rubin and Hungar 1957, Takens 1970, or Gallavotti 1980), which guarantees that the limiting procedure used to build a physical model of a system subject to a constraint converges to the ideal notion discussed in Chapters 1, 2, and 4. We remark, however, that a fundamental step in the proof is the following observation. If e denotes the total mechanical energy of the system with Lagrangian (A5.4), by the theorem of conservation of energy x(t, k) cannot have a distance larger than e/ √ k
from S, as k → ∞.
A5 Physical realisation of constraints 731 The problem of the physical realisation of constraints is also of considerable interest in the study of classical statistical mechanics. In particular, it is at the centre of several research projects on the so-called problem of the ‘ultraviolet catastrophe’ predicted by classical statistical mechanics (see Born 1927), and hence the conflict between the predictions of the equipartition theorem and of some experimental evidence relating to the specific heat of polyatomic gases, or to the form of the spectrum of black body radiation. 1 This problem deserves a more detailed discussion, which however cannot find space in this brief introduction. 1 In a remarkable series of recent papers, Benettin, Galgani and Giorgilli have studied in great detail the consequences of the modern canonical theory of perturbations for this problem, and deduced interesting physical outcomes: see the original papers cited in the bibliography (1985, 1987a, 1987b, 1989).
APPENDIX 6: KEPLER’S PROBLEM, LINEAR OSCILLATORS AND GEODESIC FLOWS In this appendix we briefly study the connections between some seemingly unrelated dynamical systems. (1) Kepler’s problem (cf. Section 5.2): a point particle of unit mass moves in the plane of coordinates (x, y) under the action of the central field with potential energy V (r) = −k/r, with fixed k > 0. (2) Linear oscillators: a point particle of unit mass moves in the (x, y) plane under the action of a central field with potential energy V (r) = 1 2
2 = 1 2 α(x
2 + y
2 ), α > 0 fixed. (3a) Geodesic flow on the sphere (cf. Section 1.6): a point particle of unit mass moves freely on the sphere of radius R immersed in three-dimensional Euclidean space. (3b) Geodesic flow on the Poincar´ e disc (cf. Section 1.7): a point particle of unit mass moves freely on the Riemannian manifold (D, ds 2 R
{w = x + iy
∈ C| |w| < 1} and ds 2 R = 4R
2 dw dw
(1 − |w|
2 ) 2 = 4R 2 (dx) 2 + (dy)
2 (1 − x 2 − y
2 ) 2 . (A6.1)
Remark A6.1 The Gaussian curvature (see Appendix 3) of the sphere of radius R is equal to 1/R 2
e disc (D, ds 2 R ) is −1/R
2 . The dynamical systems (1) and (2) are related by the theorem of Bertrand (Theorem 5.3). This is not by chance; in fact, by an appropriate coordinate and time transformation, it is possible to transform one into the other (see Levi-Civita 1920). T
z = x + iy ∈ C,
(A6.2) the change of (space and time) variables z = w
2 , w ∈ C, dt dτ = |w|
2 = 2
|z| (A6.3)
734 Kepler’s problem, linear oscillators and geodesic flows A6 transforms the equations of motion of the Kepler problem ¨ z =
−k z |z| 3 , (A6.4) for a fixed value of the energy E: E =
1 2 | ˙z| 2 − k |z| , (A6.5) into the equations of a linear oscillator: w − 2Ew = 0, (A6.6) where w = d 2 w/dτ
2 . Proof We check first of all that equation (A6.4) coincides with the equation of motion for Kepler’s problem. Separating real and imaginary parts in (A6.4) we find ¨ x =
− k r 2 x r , ¨ y = − k r 2 y r , (A6.7)
and similarly we check that equation (A6.6) coincides with the equation of a linear oscillator (set w = ξ + iη, α = −2E) and that (A6.5) is the energy of Kepler’s problem. From equations (A6.3) it follows that ˙ z = dz dt = dz dw dw dτ dτ dt = w |w| 2 w (A6.8) and |z| = |w| 2 /2; hence in the new variables the energy is expressed as E = 1 2 |w| 2 |w | 2 − 2k |w| 2 , and multiplying by |w|
2 , it follows that 1 2
2 = E
|w| 2 + 2k. (A6.9) Equation (A6.6) is obtained by computing ¨ z =
1 |w|
2 d dτ ww |w|
2 = 1 |w| 2 d dτ w w , substituting this into (A6.4) and taking into account (A6.9). A6 Kepler’s problem, linear oscillators and geodesic flows 735 Remark A6.2 An analogous theorem holds for the three-dimensional Kepler’s problem (the transformation then goes under the name of Kustaanheimo–Stiefel–Schiefele; see Schiefele and Stiefel 1971). Remark A6.3 It is interesting—and applicable to the numerical study of orbits—to note that equations (A6.6) are regular at |w| = |z| = 0. At the same point, equations (A6.4) are singular. Remark A6.4 Theorem A6.1 relates Keplerian motions with fixed energy E and solutions of the linear oscillator with α = −2E (then with the harmonic oscillator of frequency ω 2
−2E, if E < 0, while for E > 0 the oscillator if hyperbolic). The relation between problem (1) (Kepler) (and, thanks to Theorem A6.1, linear oscillators) and the geodesic flows (3a) and (3b) is the object of the following theorem (see Moser 1970 and Alekseev 1981). T heorem A6.2 The phase flow of the plane Kepler problem for a fixed value E of the energy is equivalent, up to a time parametrisation, to the geodesic flow on a surface with constant Gaussian curvature −2E/k 2
After the addition of a point, this surface is isometric to the sphere of radius R = k/
√ −2E if E < 0, to the Euclidean plane if E = 0 or to the Poincar´e disc (D, ds 2
), with R = k/ √ 2E if E > 0. Remark A6.5 The meaning of the ‘equivalence’ of two phase flows will be made clear in the course of the proof. Remark A6.6 An analogous theorem holds for the Kepler problem in three-dimensional space.
Proof of Theorem A6.2 Let q = (x, y), p = ˙q. The Hamiltonian of the plane Kepler problem is H(p, q) = 1 2 |p| 2 − k |q|
. (A6.10)
Consider a fixed value of the energy E and the three-dimensional manifold of constant energy M E
{(p, q) ∈ R 2 × R 2 \{(0, 0)}|H(p, q) = E}. (A6.11) Introduce the transformation dt dτ = |q| k . (A6.12) 736 Kepler’s problem, linear oscillators and geodesic flows A6 We want to show that on the manifold M E , Hamilton’s equations for (A6.10) become dp
dτ = − ∂ H ∂q i , dq i dτ = ∂ H ∂p i , (A6.13) where i = 1, 2, H(p, q) = |q| k
− E) = |q|
2k ( |p| 2 − 2E) − 1 (A6.14) and the manifold M E coincides with M E
{(p, q) ∈ R 2 × R 2 \{(0, 0)}|H(p, q) = 0}. (A6.15) Indeed on M E we have ∂ H ∂p i = |q| k ∂H ∂p i = dt dτ dq i dt = dq i dτ , − ∂ H ∂q i = − 1 k ∂ ∂q i |q| [H(p, q) − E] M E − |q|
k ∂H ∂q i = dt dτ ˙ p i = dp i dτ . In addition, note that M E = M E and the following identity holds: |q| 2k ( |p| 2 − 2E) = 1. (A6.16) This identity, applied to Hamilton’s equations (A6.13), yields dp i
= − |q| 2k ( |p| 2 − 2E)
∂ ∂q i |q| 2k ( |p| 2 − 2E) = − ∂ H ∂q i , dq i dτ = |q| 2k ( |p| 2 − 2E) ∂ ∂p i |q| 2k ( |p| 2 − 2E) = ∂ H ∂p i , (A6.17) where we set H(p, q) = |q| 2
|p| 2 − 2E) 2 8k 2 (A6.18) and the manifolds M E =
E coincide with M E
(p, q) ∈ R
2 × R
2 \{(0, 0)}| |p| 2 > 2E,
H(p, q) = 1 2 (A6.19) (the requirement that |p| 2
H 1/2
). A6 Kepler’s problem, linear oscillators and geodesic flows 737 From the first of equations (A6.17) we deduce q as a function of p and dp/dτ : indeed, taking into account (A6.18) we find dp dτ = −∇ q H = −q ( |p| 2 − 2E)
2 4k 2 , (A6.20)
from which it follows that q =
− 4k 2 ( |p|
2 − 2E)
2 dp dτ . (A6.21)
Differentiating the expression (A6.21) we have dq dτ = 8k 2 ( |p|
2 − 2E)
3 2 p · dp dτ dp dτ − 4k 2 ( |p| 2 − 2E) 2 d 2 p dτ 2 , (A6.22)
to be compared with dq dτ = ∇ p H = |q|
2 2k 2 ( |p|
2 − 2E)p =
8k 2 ( |p| 2 − 2E) 3 dp dτ 2 2 p (A6.23) (where we used (A6.21) in the last equality). Equating (A6.22) and (A6.23) we find 8k
( |p|
2 − 2E)
3 2 p · dp dτ dp dτ − 8k 2 ( |p| 2 − 2E) 3 dp dτ 2 p =
4k 2 ( |p| 2 − 2E) 2 d 2 p dτ 2 , and hence the second-order equation d 2
dτ 2 = 2 |p|
2 − 2E
2 p · dp dτ dp dτ − p
dp dτ 2 . (A6.24)
From (A6.20), (A6.16) it also follows that dp dτ = ( |p| 2 − 2E)
2 4k 2 |q| = |p|
2 − 2E
2k . (A6.25) Consider the projection π :
M E → R 2 , (p, q) → π(p, q) = p. It is immediate to check that: (a) if E < 0, π( M E ) = R 2 ; (b) if E = 0, π( M E ) = R 2 \{(0, 0)}; (c) if E > 0, π( M E ) = R 2 \D √ 2E = {(p 1 , p 2 ) ∈ R 2 |p 2 1 + p
2 2 > 2E }. 738 Kepler’s problem, linear oscillators and geodesic flows A6 Let A
E = π(
M E ) and introduce the Riemannian metric (ds E ) 2 = 4k 2 ( |p| 2 − 2E)
2 [(dp
1 ) 2 + (dp 2 ) 2 ]. (A6.26) We now check that equation (A6.24) is the geodesic equation on the two- dimensional Riemannian manifold (A E , (ds
E ) 2 ), and that τ is the arc length parameter along the geodesic. By (A6.26) the components of the metric tensor are g
(E) = 4k 2 δ ij ( |p| 2 − 2E) 2 , g ij (E) =
( |p|
2 − 2E)
2 4k 2 δ ij , (A6.27) and hence it is immediate to compute the Christoffel symbols (see (1.69)) Γ i
(E) = 1 2 2 n =1 g ni (E) ∂g jn (E) ∂p k + ∂g kn (E) ∂p j − ∂g jk (E) ∂p n = 1 2 2 n =1 ( |p| 2 − 2E) 2 4k 2 δ ni − 8k 2 ( |p| 2 − 2E) 3 [2p
k δ jn + 2p j δ kn − 2p
n δ jk ] = − 2 ( |p| 2 − 2E)
2 n =1 δ ni [p k δ jn + p j δ kn − p
n δ jk ] = − 2 ( |p| 2 − 2E)
(δ ij p k + δ
ik p j − p i δ jk ). (A6.28) The geodesic equation (see (1.68)) on (A E , (ds E ) 2 ) can then be written as d 2 p i ds 2 = − 2 j,k
=1 Γ i jk (E)
dp j ds dp k ds = 2 ( |p| 2 − 2E) 2 j,k
=1 (δ ij p k + δ ik p j − δ jk p i ) dp j ds dp k ds = 2 ( |p| 2 − 2E)
2 p · dp ds dp i ds − p i dp ds 2 , (A6.29) which coincides with (A6.24) by setting s = τ . Finally, we consider the map ψ : M
→ T A E , (p, q) → (p, u), (A6.30) where T A E is the tangent bundle of A E and
u = dp dτ . (A6.31)
A6 Kepler’s problem, linear oscillators and geodesic flows 739 Since from (A6.25), (A6.26) it follows that |u| ds E = 2k ( |p| 2 − 2E) dp dτ = 1, (A6.32) and the projection π : M E
E is onto, the map ψ is a diffeomorphism from M E
E , the bundle of tangent vectors of unit length: U A E
{(p, u) ∈ T A Download 10.87 Mb. Do'stlaringiz bilan baham: |
ma'muriyatiga murojaat qiling