A Damped Oscillator as a Hamiltonian System

Kirk T. McDonald
Joseph Henry Laboratories, Princeton University, Princeton, NJ 08544
(June 9, 2015; updated July 1, 2015)

1 Problem
It is generally considered that systems with friction are not part of Hamiltonian dynamics,
but this is not always the case. Show that a (nonrelativistic) damped harmonic oscillator can
be described by a Hamiltonian (and by a Lagrangian), with the implication that Liouville’s
theorem applies here.
Consider motion in coordinate x of a particle of mass m with equation of motion,
mẍ + β ẋ + kx = 0, or ẍ + αẋ + ω 20 x = 0, (1)
where α = β/m and ω 20 = k/m.
Comment on the root-mean square emittance of a “bunch” of noninteracting particles
each of which obeys eq. (1).
Deduce two independent constants of the motion for a single particle.
Hint: Consider first the case of zero spring constant k.

2 Solution
2.1 k=0
When the spring constant k is zero there is no potential energy, so the spirit of Lagrange
and Hamilton is to consider the kinetic energy T = mẋ2 /2. The equation of motion (1) can
be written in the manner of Lagrange as1
d ∂T ∂T
+α = 0. (2)
dt ∂ ẋ ∂ ẋ
This form can be written more compactly as
d ∂T 
= 0, where T  = T eαt . (3)
dt ∂ ẋ
Hence, a Lagrangian for this case is L = T , and the canonical momentum p conjugate to
coordinate x is
∂L
p= = mẋ eαt = pmech eαt , (4)
∂ ẋ
where pmech = mẋ is the ordinary mechanical momentum.
The Hamiltonian for this case is then
mẋ2 αt
H = ẋp − L = L = e . (5)
2
1
See sec. 2a of [1].

1
2.2 Nonzero k
When the spring constant k is nonzero we consider the potential energy V = kx2 /2, and the
equation of motion (1) can be written as

d ∂T ∂T ∂(−V )
+α = . (6)
dt ∂ ẋ ∂ ẋ ∂x
This form can be written more compactly as
d ∂T  ∂(−V  )
= , where T  = T eαt, V  = V eαt . (7)
dt ∂ ẋ ∂x
Hence, a Lagrangian for this case is L = T  − V  . The momentum conjugate to coordinate
x is again
∂L
p= = mẋ eαt = pmech eαt , (8)
∂ ẋ
so the Hamiltonian for this case is2

  αt p2 −αt αt mẋ2 + kx2 αt

H = ẋp − L = T + V = (T + V ) e = e +V e = e = U eαt , (9)
2m 2
where U is the mechanical energy of the system.3

2.3 Liouville
This section is based on sec. 3.8 of [5].
The motion of a system governed by eq. (1) can be written as
 
x(t) = x1 e(α −α)t/2 + x2 e−(α +α)t/2, (10)

where α = α2 − 4ω 20 is possibly imaginary (as may be x1 and x2 ). The canonical momen-
tum (8) is
α − α (α +α)t/2 α + α −(α −α)t/2
p(t) = mx1 e − mx2 e . (11)
2 2
The initial conditions (at time t = 0) are
α − α α + α
x(0) ≡ x0 = x1 + x2, p(0) ≡ p0 = mx1 − mx2 . (12)
2 2
α + α p0 α − α p0
x1 = x0 
+ , x2 = x0 − . (13)
2α mα 2α  mα
2
The Hamiltonian (9) appears to have first been given via the present argument in [2]. Impressively, this
procedure for identifying Hamiltonians can be extended to certain nonlinear examples with dissipation [3].
The Lagrangian L = (T − V )eαt was deduced by Bateman in 1931 [4], top of p. 817, as a special case of his
“dual-Lagrangian”.
3
Clearly, the Lagrangian L = (T − V )eαt and the Hamiltonian H = (T + V )eαt apply to any particle
subject to velocity-dependent damping as well as a force derivable from a potential V .

2
Thus,
⎛ ⎞ ⎛ ⎞⎛ ⎞
x(t) A B x0
⎝ ⎠= ⎝ ⎠⎝ ⎠, (14)
p(t) C D p0
where
 
−αt/2 α + α α t/2 α − α −α t/2
A = e e + e , (15)
2α 2α
α t/2 
−αt/2 e − e−α t/2
B = e , (16)
mα
α2 − α2 α t/2 −αt /2

C = m eαt/2 
e − e , (17)
  4α 
αt/2 α − α α t/2 α + α −α t/2
D = e e + e . (18)
2α 2α
The determinant of the transformation matrix is
Δ = AD − BC
α2 − α2 α t α2 + α2 α2 − α2 −α t α2 − α2 α t −α t

= e + + e − e − 2 + e = 1. (19)
4α2 2α2 4α2 4α2
Thus, the linear (canonical) transformation (14) preserves area in the x-p (phase) space,
which verifies that Liouville’s theorem [6, 7, 8] holds for this Hamiltonian system.

2.4 RMS Emittance

When considering a “bunch” of particles, a practical measure of their extent in phase space
is the root-mean-square emittance, which for a 2-dimensional phase space as in the present
example can be defined as

canonical(t) = x2 (t) p2 (t) − x(t)p(t)2 . (20)
As noted in [9], this rms emittance is invariant under linear canonical transformations, so
 of eq. (20) is a constant of the motion for a “bunch” of (noninteracting) particles each of
which obeys eq. (1).
However, in most applications of the rms emittance concept, people use the mechanical
momentum pmech rather than the canonical momentum p. In the present example, pmech =
e−αtp according to eq. (8), such that the rms mechanical emittance decreases exponentially
with time,

mech (t) = x2 (t) p2mech (t) − x(t)pmech(t)2 = e−αt canonical(t). (21)
The rms mechanical emittance (21) is “cooled” by the damping, whereas the rms canon-
ical emittance (20) is not.4
4
This contrasts with a theorem due to Swann [10] that if a particle of electric charge e is acted upon
only by an electromagnetic field with vector potential A then the volume in (x, p) phase space of a “bunch”
of such particles is the same as that in (x, pmech) phase space, where p = pmech + eA/c.

3
2.5 Relativistic Damped Harmonic Oscillator
In accelerator physics the particles of interest typically have velocities near the speed c of
light in vacuum, so we also give a relativistic version of the preceeding analysis.
If the force on the particle (of rest mass m) can be deduced from a potential V , a
relativistic Hamiltonian is

H(x, pmech ) = Emech + V = c m2c2 + p2mech + V, (22)
 
where pmech = γmẋ, Emech = γmc2 = c m2c2 + p2mech and γ = 1/ 1 − ẋ2 /c2 . Hamilton’s
equations of motion for this system are
dx ∂H pmech
= = = ẋ, (23)
dt ∂pmech Emech
dpmech ∂H ∂V
= − =− = F. (24)
dt ∂x ∂x
If the force includes the velocity-dependent term −β ẋ = −αmẋ, we follow sec. 2.2 to consider
the canonical momentum

p = pmech eαt , (25)

and the Hamiltonian

   

αt αt
H(x, p, t) = e (Emech + V ) = e c m c + pmech + V = eαt c m2c2 + p2 e−2αt + V .(26)
2 2 2

Hamilton’s equations of motion for this system are

dx ∂H p e−αt pmech
= = = = ẋ, (27)
dt ∂p Emech Emech
 
dp d(pmech eαt ) dpmech ∂H ∂V
= = eαt + αpmech = − = −eαt , (28)
dt dt dt ∂x ∂x
The last equation can be rewritten as
dpmech ∂V ∂V
+ αpmech + = γ 3 (ẋ)mẍ + β ẋ + = 0, (29)
dt ∂x ∂x
which is the equation of motion (nonlinear in ẋ) of a relativistic particle subject to velocity-
dependent damping and another force that is derivable from the potential V .
In particular, we see that the relativistic, damped harmonic oscillator is a Hamiltonian
system, and a “bunch” of such (noninteracting) particles obeys Liouville’s theorem.
If we characterize the extent of the “bunch” in phase space by an rms emittance, we must
note that the canonical transformation (x0, p0 ) → (x(t), p(t)) is not linear (unlike the non-
relativistic case), with the implication that the rms canonical emittance (20) actually grows
with time [11]. The rms mechanical emittance (21) is again e−αt times the rms canonical
emittance, and the exponential damping factor is stronger than the emittance growth due to
the nonlinear relativistic time evolution, such that the rms mechanical emittance is “cooled”
with time.

4
2.6 Constants of the Motion and Alternative Hamiltonians
The second-order differential equation (1) has two independent constants of integration,
implying that motion governed by this equation has two independent constants of the motion.
Of course, the physical significance of these constants will be unclear in systems where energy
is not conserved, so this section is largely a mathematical exercise.
If the Hamiltonian of the system were time independent, it would be such a constant
of the motion, so one method of identifying the desired constants of the motion is to seek
alternative Hamiltonians that are independent of time. This was first done for the damped
harmonic oscillator by Havas [12], and next by Leach [13] (and by Lemos [14]5 ), each of
whom displayed one time-independent Hamiltonian (which happen to be “inequivalent”6).
The analysis of Lemos is much simpler than that of Havas, so we first illustrate the
former. Inspection of the Hamiltonian (9) indicates that the change of variables,7

X = x eαt/2, P = p e−αt/2, (30)

leads to the formally time-independent form h(X, P ) = P 2 /2m + kX 2 /2. However, h is

not the Hamiltonian in terms of the coordinates X and P . For that, we need the so-called
generating function Φ of the canonical transformation (30),8 which obeys p = ∂Φ/∂x and
X = ∂Φ/∂P . In the present case, Φ = xP eαt/2, and the transformed Hamiltonian is

∂Φ P2 αXP kX 2
H(X, P ) = H(x, p, t) + = + + , (31)
∂t 2m 2 2
which is independent of time, and hence a constant of motion of the system. We rewrite
H(X, P ) in terms of x and p or ẋ (using eqs. (8) and (30)) as the constant function

p2 e−αt αxp kx2 eαt m 2

F (x, p) = + + = eαt ẋ + αxẋ + ω 20x2 , (32)
2m 2 2 2
which reverts to the mechanical energy when α = 0.

Havas’ approach [12] does not involve a change of variables, but seeks an “integrating
function” f(x, ẋ, t) such that an equation of motion g(x, ẋ, ẍ, t) = 0 can be related to a
Lagrangian L according to9
d ∂L ∂L
fg = − . (33)
dt ∂ ẋ ∂x
Then, if the corresponding Hamiltonian H = ẋp − L is independent of time, where p =
∂L/∂ ẋ, it will be a constant of the motion. For a damped oscillator with g = mẍ + β ẋ + kx,
5
Leach analyzed a damped harmonic oscillator with time-dependent friction and spring constant, and
Lemos gave a simplified analysis analysis for time-independent parameters.
6
Given two “inequivalent”, time-independent Hamiltonians for a systems, one can generate an infinite
set of alternative Hamiltonians, as discussed in [15]. See also [16] and references therein.
7
This change of variables was considered in sec. 11 of [17].
8
See, for example, sec. 45 of [8].
9
Havas’ method differs slightly from that used in secs. 2.1-2 where the nominal Lagrangian L0 = T − V
was multiplied by eαt to give a Lagrangian L = eαt L0 from which the equation of motion can be deduced.

5
Havas found (p. 387 of [12]) that f = (mẋ2 + βxẋ + kx2)−1 , and for 4km > β 2,
 
2mẋ + βx 2m ẋ + βx 1
L =  2
tan−1 
2
− ln(mẋ2 + βxẋ + kx2 ), (34)
x 4km − β x 4km − β 2
 
β 2mẋ + βx 1
H = − 2
tan−1 
2
+ ln(mẋ2 + βxẋ + kx2). (35)
4km − β x 4km − β 2

Other constants of the motion can be found by a clever approach to integrating eq. (1)
[18]. In this, the parameters α and ω 20 are replaced by λ1 and λ2 according to

α ± α2 − 4ω 20
α = λ1 + λ2 , ω 20 = λ1 λ2, λ1,2 = , (36)
2
such that the equation of motion becomes
ẍ + (λ1 + λ2 )ẋ + λ1 λ2x = ẍ + λ1 ẋ + λ2 (ẋ + λ1x) = 0. (37)
This is a first-order differential equation in the variable ẋ + λ1 x, which can be integrated to
give
eλ2 t (ẋ + λ1 x) = const = D1 . (38)
Interchanging λ1 and λ2 we obtain a second constant of the motion,
eλ1 t (ẋ + λ2 x) = const = D2 . (39)
These two constants of the motion can be combined to give other constants of the motion,
such as
m m m
F = D1 D2 = e(λ1 +λ2 )t (ẋ2 + (λ1 + λ2 )xẋ + λ1λ2x2 ) = eαt (ẋ2 + αxẋ + ω 20x2 ), (40)
2 2 2
and
D1λ1 (ẋ + λ1 x)λ1
B = λ2 = . (41)
D2 (ẋ + λ2 x)λ2
The constant F of eq. (40) is the same as that found in eq. (32), while the constant B of
eq. (41) appears to be different from H of eq. (35). The symbol B is used in honor of Bohlin,
who apparently discussed constants of this form in 1908 [18].
Interest in Hamiltonians for the damped oscillator arises mainly in the context of quantum
theory, where there seems to be an ongoing debate as to which of the many “inequivalent”
Hamiltonians is “best”. A review as of 1981 is [19], and an example of recent commentary
is [20].
Another issue illustrated by the present example is that equations of motion for a system
with damping can be deduced from a variational principle in some cases, as perhaps first
noted by Bateman [4]. This issue relates to the origin of conservation laws, as time invariance
of a Hamiltonian (such as eq. (35)) apparently does not necessarily imply that this constant
Hamiltonian is the system energy). That is, Noether’s theorem [21, 22] applies to (classical
and quantum) field theory but not to classical mechanics. The ongoing debate on this theme
is reviewed in [23]; a recent comment is at [24].

6
2.7 Time-Dependent Forces
In the application of the present example to accelerator physics we have tacitly imagined
performing the analysis in the rest frame of a reference particle near the center of a “bunch” of
(noninteracting) particles. In practice, as the “bunch” propagates it will encounter different
external forces at different times, which leads to an interest in a Hamiltonian description for
time-dependent forces.
Studies of an undamped harmonic oscillator with a time-dependent spring constant began
with Lecornu (1895) [25] and Rayleigh (1902) [26] in classical contexts. In a famous discussion
at the 1911 Solvay Conference [27], Lorentz posed the problem of a time-dependent quantum
oscillator, and Einstein argued (briefly) that the ratio of the instantaneous energy U(t) =
(mẋ2 + k(t)x2)/2 to the instantaneous angular frequency ω(t) would be constant (which
notion seems clearer in a quantum view than classically). Einstein’s prescient remark lead to
the development by Ehrenfest (1916) [28] of the concept of adiabatic invariance in systems
with “slow” time dependence.10 Kruskal (1962) [30] suggested a method for analysis when
the time dependence is somewhat more rapid than that for which adiabatic invariance holds,
and Lewis (1967) [31, 32] extended this method to arbitrary time dependence of the spring
constant.11 Leach (1978) [13] extended Lewis’ analysis to include time-dependent damping
proportional to the velocity.
Following Leach [13], we consider a damped harmonic oscillator with time-dependent
forces,
mẍ + β(t)ẋ + k(t)x = 0, ẍ + α(t)ẋ + ω(t)2 x = 0. (42)
Multiplying this differential equation by
 t
f(t) = α(t) dt , (43)
0

it can be rewritten as
 
d ∂ k e f x2
(mẋ ef ) + = 0. (44)
dt ∂x 2
This equation of motion can be deduced from the Lagrangian
mẋ2 − kx2 f
L= e , (45)
2
for which the canonical momentum is
p = mẋ ef , (46)
and the Hamiltonian is12
p2 −f kx2 f ẋ2 + ω 2 x2 f
H = ẋp − L = e + e =m e . (47)
2 2 2
10
Certain mathematical subtleties related to adiabatic invariance in Lorentz’ example are considered to
have been clarified only in 1963 [29].
11
For a perspective on the methods of Kruskal and Lewis, see [33].
12
Similarly, replacing αt by f in eq. (26) provides a Hamiltonian for the relativistic, damped harmonic
oscillator with time-dependent forces.

7
 Thus, the time-dependent damped harmonic oscillator is also a Hamiltonian system, and
a “bunch” of particles that each obey eq. (42) also obey Liouville’s theorem.
If, for example, the damping is constant until time t1, and zero thereafter,
⎧
⎨ eαt (0 < t < t ),
1
f(t) = (48)
⎩ eαt1 (t > t ).
1

When the damping goes to zero at time t1 , the canonical momentum does not revert to the
mechanical momentum mẋ, but to this multiplied by the (large) constant eαt1 . The rms
canonical emittance (20) is ef times the rms mechanical emittance (21) at all times, so again
the former remains constant in time while the latter is exponentially damped.

The literature is much concerned with the invariants of the time-dependent damped
harmonic oscillator, although only one such invariant has been found so far.13 Lewis [31, 32]
deduced an invariant for the undamped, time-dependent harmonic oscillator, which turns out
to have been anticipated in a little-known work by Ermakov in 1880 [34]. The generalization
of this invariant for the case of time-dependent damping was given by Leach [13].14 As might
be expected, the physical significance of these invariants is obscure.

References
[1] P. Caldirola, Forze Non Conservative nella Meccanica Quantistica, Nuovo Cim. 18, 393
(1941), http://physics.princeton.edu/~mcdonald/examples/QM/caldirola_nc_18_393_41.pdf
[2] E. Kanai, On the Quantization of the Dissipative Systems, Prog. Theor. Phys. 3, 440
(1949), http://physics.princeton.edu/~mcdonald/examples/QM/kanai_ptp_3_440_49.pdf
[3] V.K. Chandrasekar et al., A simple and unified approach to identify integrable nonlinear
oscillators and systems, J. Math. Phys. 47, 023508 (2006),
http://physics.princeton.edu/~mcdonald/examples/mechanics/chandrasekar_jmp_47_023508_06.pdf

[4] H. Bateman, On Dissipative Systems and Related Variational Principles, Phys. Rev.
38, 815 (1931), http://physics.princeton.edu/~mcdonald/examples/mechanics/bateman_pr_38_815_31.pdf
[5] G.J. Sussman and J. Wisdom, Structure and Interpretation of Classical Mechanics, 2nd
ed. (MIT Press, 2014),
http://physics.princeton.edu/~mcdonald/examples/mechanics/sussman_mechanics_sec3.8.pdf

[6] J. Liouville, Note sur la Théorie de la Variation des constantes arbitraires, J. Math.
Pures Appl. 3, 342 (1838),
http://physics.princeton.edu/~mcdonald/examples/mechanics/liouville_jmpa_3_342_38.pdf

13
Replacing αt by f in eq. (30) leads to a canonical transformation with generating function Φ = xP ef/2 ,
so the transformed Hamiltonian again has the form (31), but this is no longer independent of time as α = α(t),
and eq. (32) (with f for αt) is not an invariant.
14
Leach’s invariant can be deduced by seeking a (generalized) canonical transformation in the extended
phase space (x, p, t, −U ) such that the Hamiltonian H  (x , p ) is independent of t (and so H  = U  is
constant). See sec. 3.3 of [35].

8
 [7] D.D. Nolte, The Tangled Tale of Phase Space, Phys. Today 63, 4, 32 (Apr. 2010),
http://physics.princeton.edu/~mcdonald/examples/mechanics/nolte_pt_63_4_32_10.pdf

[8] L.D. Landau and E.M. Lifshitz, Mechanics, 3rd ed. (Pergamon, 1976).

[9] E.D. Courant and H.S. Snyder, Theory of the Alternating-Gradient Synchrotron, Ann.
Phys. (NY) 3, 1 (1958),
http://physics.princeton.edu/~mcdonald/examples/accel/courant_ap_3_1_58.pdf

[10] W.F.G. Swann, Application of Liouville’s Theorem to Electron Orbits in the Earth’s
Magnetic Field, Phys. Rev. 44, 224 (1933),
http://physics.princeton.edu/~mcdonald/examples/accel/swann_pr_44_224_33.pdf

[11] K.T. McDonald, Emittance Growth from Weak Relativistic Effects (Mar. 21, 2011),
http://physics.princeton.edu/~mcdonald/examples/growth.pdf

[12] P. Havas, The Range of Application of the Lagrange Formalism, Suppl. Nuovo Cim. 5,
363 (1957), http://physics.princeton.edu/~mcdonald/examples/mechanics/havas_nc_5s_363_57.pdf

[13] P.G.L. Leach, On a Direct Method for the Determination of an Exact Invariant for the
Time-Dependent Harmonic Oscillator, J. Austral. Math. Soc. 209, 97 (1977),
http://physics.princeton.edu/~mcdonald/examples/mechanics/leach_jams_20_97_77.pdf

[14] N.A. Lemos, Canonical approach to the damped harmonic oscillator, Am. J. Phys. 47,
857 (1979), http://physics.princeton.edu/~mcdonald/examples/mechanics/lemos_ajp_47_857_79.pdf

[15] C. Leubner, Inequivalent Lagrangians from Constants of the Motion, Phys. Lett. 86A,
68 (1981), http://physics.princeton.edu/~mcdonald/examples/mechanics/leubner_pl_86a_68_81.pdf

[16] A. Choudhuri et al., Symmetries and conservation laws of the damped harmonic oscil-
lator, Pramana 70, 657 (2008),
http://physics.princeton.edu/~mcdonald/examples/mechanics/choudhuri_pramana_70_657_08.pdf

[17] H. Lamb, The Dynamical Theory of Sound (Arnold, London, 1910),

http://physics.princeton.edu/~mcdonald/examples/mechanics/lamb_sound_10.pdf

[18] E.W. Gettys et al., Bohlin’s and other integrals for the damped harmonic oscillator,
Am. J. Phys. 42, 162 (1981),
http://physics.princeton.edu/~mcdonald/examples/mechanics/gettys_ajp_49_162_81.pdf

[19] H. Dekker, Classical and Quantum Mechanics of the Damped Harmonic Oscillator,
Phys. Rep. 80, 1 (1981),
http://physics.princeton.edu/~mcdonald/examples/QM/dekker_pr_80_1_81.pdf

[20] D. Schuch et al., Interrelations between different canonical descriptions of dissipative

systems, Phys. Scr. 90, 045209 (2015),
http://physics.princeton.edu/~mcdonald/examples/mechanics/schuch_ps_90_045209_15.pdf

9
[21] E. Noether, Invariante Variationenprobleme, Nachr. Kgl. Ges. Wiss. Göttingen, 235
(1918), http://physics.princeton.edu/~mcdonald/examples/mechanics/noether_nkwg_235_18.pdf
English translation: Invariant Variation Problems, Trans. Theory Stat. Phys. 1, 186
(1971), http://physics.princeton.edu/~mcdonald/examples/mechanics/noether_ttsp_3_186_71.pdf

[22] E.L. Hill, Hamilton’s Principle and the Conservation Theorems of Mathematical
Physics, Rev. Mod. Phys. 23, 253 (1951),
http://physics.princeton.edu/~mcdonald/examples/mechanics/hill_rmp_23_253_51.pdf

[23] S.R. Smith, Symmetries and the explanation of conservation laws in the light of the
inverse problem in Lagrangian mechanics, Stud. Hist. Phil. Mod. Phys. Scr. 39, 325
(2008), http://physics.princeton.edu/~mcdonald/examples/mechanics/smith_shpmp_39_325_08.pdf

[24] H Qin, J.W. Burby and R.C. Davidson, Field theory and weak Euler-Lagrange equation
for classical particle-field systems, Phys. Rev. E 90, 032102 (2014),
http://physics.princeton.edu/~mcdonald/examples/mechanics/qin_pre_90_043102_14.pdf

[25] L. Lecornu, Mémoire sur la Pendule de Longeur Variable, Acta Math. 19, 201 (1895),
http://physics.princeton.edu/~mcdonald/examples/mechanics/lecornu_am_19_201_95.pdf

[26] Lord Rayleigh, The Pressure of Vibrations, Phil. Mag. 3, 338 (1902),
http://physics.princeton.edu/~mcdonald/examples/mechanics/rayleigh_pm_3_338_02.pdf

[27] H.A. Lorentz and A. Einstein, Discussion at the 1911 Solvay Conference, La Théorie du
Rayonnement et les Quanta (Gauthier-Villars, Paris, 1912), p. 450,
http://physics.princeton.edu/~mcdonald/examples/QM/solvay_11.pdf

[28] P. Ehrenfest, Adiabatische Invarianten und Quantentheorie, Ann. d. Phys. 356, 327
(1916), http://physics.princeton.edu/~mcdonald/examples/QM/ehrenfest_ap_356_327_16.pdf

[29] J.E. Littlewood, Lorentz’s Pendulum Problem, Ann. Phys. 21, 233 (1963),
http://physics.princeton.edu/~mcdonald/examples/mechanics/littlewood_ap_21_233_63.pdf

[30] M. Kruskal, Asymptotic Theory of Hamiltonian and other Systems with all Solutions
Nearly Periodic, J. Math. Phys. 3, 806 (1962),
http://physics.princeton.edu/~mcdonald/examples/mechanics/kruskal_jmp_3_806_62.pdf

[31] H.R. Lewis, Jr, Classical and Quantum Systems with Time-Dependent Harmonic-
Oscillator-Type Hamiltonians, Phys. Rev. Lett. 18, 510 (1967),
http://physics.princeton.edu/~mcdonald/examples/mechanics/lewis_prl_18_510_67.pdf

[32] H.R. Lewis, Jr, Class of Exact Invariants for Classical and Quantum Time-Dependent
Harmonic Oscillators, J. Math. Phys. 9, 1976 (1968),
http://physics.princeton.edu/~mcdonald/examples/mechanics/lewis_jmp_9_1976_68.pdf

[33] W. Sarlet, Class of Hamiltonians with One Degree-of-Freedom Allowing Application of

Kruskal’s Asymptotic Theory in Closed Form. I & II, Ann. Phys. 92 232,248 (1975),
http://physics.princeton.edu/~mcdonald/examples/mechanics/sarlet_ap_92_232_75.pdf

10
[34] V.P. Ermakov, U. Izvestia Kiev, Ser. III 9 (1880), English translation: Second-order
differential equations. Conditions of complete integrability, Appl. Anal. Discrete Math.
2, 123 (2008), http://physics.princeton.edu/~mcdonald/examples/mechanics/ermakov_aadm_2_123_08.pdf

[35] J. Struckmeier, Hamiltonian dynamics on the symplectic extended phase space for au-
tonomous and non-autonomous systems, J. Phys. Math. A 38, 1257 (2005),
http://physics.princeton.edu/~mcdonald/examples/mechanics/struckmeier_jpa_38_1257_05.pdf

11