Analytical Mechanics This page intentionally left blank
Download 10.87 Mb. Pdf ko'rish
|
2 + iε
m∈Z 2 \0 e −|m
1 |−|m
2 | e i(m 1 χ 1 +m 2 χ 2 ) m 1 + m 2 (ω + ε)
. 534 Analytical mechanics: canonical perturbation theory 12.9 Note that for every irrational ω there exists a sequence ε n → 0 such that ω + ε n is rational. Deduce from this fact the divergence of the series expansion of W . 8. Solve Hamilton’s equations for the Hamiltonian of the previous problem. Prove that if ω + ε is rational, J 1 (t) and J 2 (t) are proportional to t. 9. Compute the first order of canonical perturbation theory for the Hamiltonian H = p
1 + ω
2 1 q 2 1 2 3 + p 2 2 + ω 2 2 q 2 2 2 3 + εq
2 1 q 2 2 p 2 1 p 2 2 . Under which hypotheses is this procedure justified? (Solution: H = (ω 1 J 1 ) 3 + (ω 2 J 2 ) 3 + (ε/4)(J 1 J
) 2 and the hypothesis is that k(ω 1 J 1 ) 3 − j(ω 2 J 2 ) 3 = / 0 for all integers j and k with |k| ≤ 2.) 10. Consider a quasi-integrable system with two degrees of freedom described by the Hamiltonian H(q 1 , q 2 , p
1 , p
2 , ε) =
1 2 p 2 1 + q 2 1 + 1 2 p 2 2 + εq 2 1 p 2 2 cos 2 q 2 . (a) Introduce the action-angle variables χ 1 , χ
2 , J
1 , J
2 for the integrable system obtained by setting ε = 0 and express the Hamiltonian H in these variables. (b) Compute the Hamiltonian K(J 1 , J
2 , ε) obtained through the use of first-order canonical perturbation theory, and the frequencies ω 1 (J 1 , J
2 ), ω
2 (J 1 , J 2 ) of the motions. (c) Under which conditions on J 1 , J
2 is this procedure justified? 11. Consider the quasi-integrable system with two degrees of freedom described by the Hamiltonian H(p 1
2 , q
1 , q
2 , ε) =
p 2 1 + p 2 2 2 + 3q 2 1 + q 2 2 2 + εq 2 1 q 2 2 . (a) Introduce the action-angle variables (J 1 , J
2 , χ
1 , χ
2 ) for the integrable system obtained by setting ε = 0 and express the Hamiltonian H in these variables. (b) Compute the generating function W (χ 1 , χ
2 , J
1 , J
2 , ε) of the canonical trans- formation near the identity to new variables χ 1 , χ 2 , J
1 , J
2 , transforming the Hamiltonian H into a new Hamiltonian K which depends (up to terms of order
O(ε 2 )) only on the new action variables. Compute the new Hamiltonian K(J 1 , J 2 , ε) and the frequencies of the corresponding motions. (c) Under which conditions on (J 1 , J 2 ) is this procedure justified? (d) Compute the new Hamiltonian K which depends only on the new action variables up to terms of order O(ε 3
12.9 Additional remarks and bibliographical notes In this chapter we briefly introduced a few perturbation methods for studying the motion of quasi-integrable Hamiltonian systems, and in particular we considered the problem of the existence of (approximate) first integrals of the motion
12.10 Analytical mechanics: canonical perturbation theory 535 (Sections 12.4 and 12.5), of bounded and quasi-periodic motions (Section 12.6), and of the existence of adiabatic invariants (Section 12.7). In the last twenty years, the study of canonical perturbation theory saw a very significant development, justified both by the new theoretical results on the problem of convergence of the series expansions (in particular, the KAM theorem), and because of the appearance of new fields of application (plasma physics, elementary particle accelerators, physical chemistry, dynamics of galaxies, etc.) which complement the classical domain of application of this theory, celestial mechanics. Although this is traditionally considered a ‘difficult’ subject, too difficult to enter the syllabus of an undergraduate course, we thought that it was necessary, from the point of view of general scientific culture, to present, if only briefly, the general lines of the modern theory, skipping many mathematical details. Chapters 5 and 6 of Arnol’d et al. (1983) contain a more detailed exposition, of exceptional clarity, of the material we summarised and of many more results, including a large bibliography. The textbook by Gallavotti (1980) is also a good source for further study, as is the excellent review article by the same author (Gallavotti 1984). The Birkhoff series and their applications are discussed in detail in an article by Moser (1968). See also Moser (1986), especially for the study (here omit- ted) of discrete Hamiltonian dynamical systems which are quasi-integrable (see Arrowsmith and Place (1990) for a short introduction, Moser (1973) and Siegel and Moser (1971) for a more detailed exposition). A very readable proof of the KAM theorem, developing the original argument due to Kolmogorov (1954), is given by Benettin et al. (1984). More recent developments of the KAM theory are discussed in Bost (1986) and Yoccoz (1992). Finally, we must recommend the reading of the vast original work of Poincar´ e on the subject (Poincar´ e 1892, 1893, 1899), which remains, after a century, a constant source of inspiration for research in the field. It is not possible to even hint at the richness of the topics considered, or to illustrate the depth of Poincar´ e’s reasoning. The reader interested in the personality of one of the founders of modern mathematics can find interesting material in Boutroux (1914). 12.10 Additional solved problems Problem 1 Consider a harmonic oscillator with Hamiltonian h(p, q, ε) = p 2 2m + 1 2 mω 2 q 2 + εaq 3 , where ε is a small parameter. Compute, using the perturbation method, the variation in the frequency of the motion to the first significant order in ε. Compare the result obtained with the direct computation of the action and 536 Analytical mechanics: canonical perturbation theory 12.10 of the frequencies of the motions associated with the completely canonically integrable Hamiltonian h(p, q, ε). Solution
The action-angle variables associated with the unperturbed motion (ε = 0) are p =
√ 2mωJ cos χ, q = 2J
sin χ. Substituting in h, we obtain from this the expression for the Hamiltonian H(J, χ, ε): H(J, χ, ε) = ωJ + εa 2J mω
sin 3 χ. We seek a generating function W (J , χ, ε) = J χ+εW (1)
(J , χ)+ε 2 W (2) (J , χ)+
· · · which transforms H to K(J , ε) = ωJ + εK 1 (J ) + ε
2 K 2 (J ) + · · · . Following the procedure described in Section 12.1 we find the equations ω ∂W (1) ∂χ (J , χ) + F (J , χ) = K 1 (J ),
ω ∂W (2) ∂χ (J , χ) + F (2) (J , χ) = K 2 (J ),
where F (J , χ) = a 2J mω
sin 3 χ, F (2)
(J , χ) = ∂F ∂J (J , χ) ∂W (1) ∂χ (J , χ).
Since we are seeking the variation in the frequency of the motions and F (2)
depends only on ∂W (1)
/∂χ, it is not necessary to compute explicitly W (1)
and W (2) and it is sufficient to compute K 1 and K 2 . From the first equation we obtain K 1 (J ) = 1 2π 2π 0 F (J , χ) dχ = 0, and therefore ∂W (1) ∂χ (J , χ) = − F (J , χ) ω , from which it follows that K 2 (J ) = 1 2π 2π 0 F (2) (J , χ) dχ = − 1 2π 2π 0 3a 2 2ω 2 mω 3 (J ) 2 sin 6 χ dχ
= − 3a 2 2ω 2 mω 3 (J ) 2 5 16 = − 15 4 a 2 (J ) 2 m 3 ω 4 . 12.10 Analytical mechanics: canonical perturbation theory 537 Hence the first significant variation in the frequency of the motions happens to the second order in ε and has value ω(J , ε) = ω + ε 2 ∂K
∂J + O(ε 3 ) = ω
− 15 2 a 2 J m 3 ω 4 ε 2 + O(ε
3 ). Let us now see how, thanks to the integrability of the one-dimensional motions, it is possible to arrive at the same conclusion by computing directly the dependence of the energy on the action J of the completely integrable Hamiltonian h(p, q, ε). Indeed, from the equation p 2 2m + 1 2 mω 2 q 2 + εaq 3 = E
one can deduce the relation between the action J associated with the oscillations (near q = p = 0) and the energy E and the parameter ε, through an elliptic integral. Since we are only interested in small values of the parameter ε, we can compute the variation of J and E in the form of an expansion in series of powers of ε. Indeed, we have p = 2mεaf (q), where f (q) =
−q 3 − mω 2 2εa q 2 + E εa = (e 1 − q)(q − e 2 )(q
− e 3 ). (12.147) For small values of ε two roots of the polynomial f (q), which we indicated by e 1
2 , are in a neighbourhood of the points ± 2E/mω 2
place in the interval e 1 ≤ q ≤ e 2 . The third root e 3 is of the order of −mω 2 /2εa. We can therefore expand f (q) =
−e 3 (e 1 − q)(q − e 2 )
− q e 3 = −e 3 (e 1 − q)(q − e 2 ) 1 − q 2e 3 − q 2 8e 3 3 + · · · , from which it follows that J = 1
e 2 e 1 2mεaf (q) dq = √ −2mεae
3 J 0 − 1 2e 3 J 1 − 1 8e 2 3 J 2 + · · · , where J k = 1 π e 2 e 1 q k (e 1 − q)(q − e 2 ) dq,
k = 0, 1, 2, . . . . 538 Analytical mechanics: canonical perturbation theory 12.10 Using the substitution [2q − (e 1 + e 2 )]/(e
1 − e
2 ) = sin ψ which transforms the integration interval [e 1 , e 2 ] into
1 2 π, 3 2 π and explicitly computing we find J 0 = (e 1 − e 2 ) 2 8 , J 1 = (e 1 + e 2 ) 2 2 J 0 , J 2 = J 0 16 [5(e
1 + e
2 ) 2 − 4e 1 e 2 ]. We must now determine e 1 , e
2 , e
3 as functions of ε. Identifying the coefficients in (12.147) and setting θ = εa, we find e 1 + e 2 + e 3 = − mω 2 2θ , (e 1 + e 2 )e 3 = −e 1 e 2 , e 1 e 2 e 3 = E θ . (12.148) Writing e 1 = − 2E/mω 2 + e 1 , e
2 = 2E/mω 2 + e
2 , e
3 = ξ
− mω 2 /2θ and defining η = e 1 + e 2 , ζ = e
1 − e
2 , from equations (12.148) we arrive at the system η + ξ = 0, (12.149)
η ηθ +
mω 2 2 2 = −Eθ, (12.150) η ηθ + mω 2 2 = − 2E mω 2 θ + 2E mω 2 ζθ + η 2 − ζ 2 4 θ. (12.151)
From (12.150) one computes the values taken by η , η , η at θ = 0: η 0
−E mω 2 2 −2 , η 0 = 0, η 0 = −12E 2 mω 2 2 −5 . (12.152)
From (12.151), by differentiating three times, we obtain the values of ζ , ζ : ζ 0 = 0, ζ 0 = − 5 2 E 3/2 mω 2 2 −7/2 . (12.153) We now only need to use that e 1 = 1 2 η 0 θ + 1 2 ζ 0 θ 2 + O(θ
3 ), e 2 = 1 2 η 0 θ − 1 2 ζ 0 θ 2 + O(θ 3 ), ξ = −η 0 θ + O(θ
3 ) to obtain e 1 = − 2E mω 2 − 1 2 E mω 2 2 −2 θ − 5 8 E 3/2 mω 2 2 −7/2 θ 2 + O(θ
3 ), e 2 = 2E mω 2 − 1 2 E mω 2 2 −2 θ +
5 8 E 3/2 mω 2 2 −7/2
θ 2 + O(θ 3 ), e 3 = − mω 2 2θ + E
mω 2 2 −2 θ +
O(θ 3 ). 12.10 Analytical mechanics: canonical perturbation theory 539 Substituting the expansions of e 1 , e
2 , e
3 into the expressions for J 0 , J
1 and J
2 we finally find J = E ω 1 + 15 4 ε 2 a 2 E m 3 ω 6 + · · · .
Inverting these relations, we have E = ωJ
− 15ε
2 a 2 4m 3 ω 4 J 2 + · · · ,
and finally the frequency ω(J ) =
dE dJ = ω − 15 2 ε 2 a 2 m 3 ω 4 J + O(ε 3 ). Problem 2 Consider a harmonic oscillator with two degrees of freedom and Hamiltonian h(p 1
2 , q
1 , q
2 ) =
p 2 1 + p 2 2 2m + 1 2 m(ω
2 1 q 2 1 + ω 2 2 q 2 2 ) + a 30 q 3 1 + a
21 q 2 1 q 2 + a 12 q 1 q 2 2 + a
03 q 3 3 . Introduce the action-angle variables (J 1 , J
2 , χ
1 , χ
2 ) associated with the har- monic part h 0 (p 1 , p 2 , q
1 , q
2 ) =
p 2 1 + p 2 2 2m + 1 2 m(ω
2 1 q 2 1 + ω 2 2 q 2 2 ) of the Hamiltonian h and determine an approximate first integral I in the form I(J
1 , J
2 , χ
1 , χ
2 ) = ω
1 J 1 − ω 2 J 2 + I
(3) (J 1 , J 2 , χ 1 , χ
2 ), where I (3) (J
, J 2 , χ 1 , χ
2 ) =
3 j =0 J (3−j)/2
1 J j/ 2 2 P j (χ 1 , χ 2 ) and P j is a trigonometric polynomial of degree 3 (the adelphic integral of Whittaker, cf. Whittaker (1936, chapter XVI)). Solution
With respect to the action-angle variables p i = 2mω
i J i cos χ i , q i = 2J i mω i sin χ
i , i = 1, 2, the Hamiltonian h becomes H(J,
χ) = ω 1 J 1 + ω
2 J 2 + F (J 1 , J 2 , χ
1 , χ
2 ), F (J 1 , J
2 , χ
1 , χ
2 ) = J
3/2 1 (F 1 sin χ
1 + F
2 sin 3χ
1 ) + J
1 J 1/2 2 [F 3 sin χ 2 + F 4 sin(2χ
1 + χ
2 ) + F 5 sin(2χ
1 − χ
2 )] + J
1/2 1 J 2 [F 6 sin χ 1 + F 7 sin(2χ
2 + χ
1 ) + F Download 10.87 Mb. Do'stlaringiz bilan baham: |
ma'muriyatiga murojaat qiling