You are on page 1of 22

DEPARTMENT OF PHYSICS

INDIAN INSTITUTE OF TECHNOLOGY, MADRAS

PH1010 Physics I Unit 3: 2013 edition

One-dimensional dynamics, Periodic Motion,


Equilibria and Phase trajectories
Contents: Phase space, 1-d potentials and integrating equations of motion,
periodic motion and time period, states of equilibrium in one-dimensional motion,
phase trajectories in simple 1-d potentials, including the simple pendulum and
the double-well.

1 Phase space and phase trajectories


Newton’s Second Law (the “equation of motion”) for a particle of mass m is

d2 r d 2 xi
m = F (r) or, in component form, m = Fi (r). (1)
dt2 dt2
If the values of the six dynamical variables (x1 , x2 , x3 , ẋ1 , ẋ2 , ẋ3 ), are given at
some initial instant of time t0 , and the force F (r) is known, Eq. (1) can be
solved (in principle), to find the values of these variables at any later instant
of time t > t0 . We say that the instantaneous state of the particle is specified
by the values of these six variables, namely, three coordinates and three velocity
components. Why is the state of the particle given by six dynamical variables
rather than just the three coordinates? Because Eq. (1) are three, in general
coupled, second-order differential equations in time. Therefore we need to
specify the initial values of the coordinates (x1 , x2 , x3 ) of the particle as well as
its velocity components (ẋ1 , ẋ2 , ẋ3 ), in order to obtain a unique solution at all
subsequent instants of time.
Instead of using the velocity components one may use the momentum com-
ponents. Note that the equations of motion Eq. (1) can be equivalently written
as
dr p dp
= , = F (r). (2)
dt m dt
The differential equations are now first-order but there are now six in general
coupled differential equations. Integrating each of these gives the six arbitrary
constants that can be associated with the initial conditions of three initial position
components and three initial momentum components. Writing given higher-order
differential equations in terms of first order differential equations by introducing
other dynamical variables (such as momentum above) is a general and useful

1
technique for further analysis. Also note that although we have indicated F (r)
as the force it can at this be dependent time or velocity/momentum. For ex-
ample the case of the forced and dampled harmonic oscillator requires such a
generalization.
Using the momentum instead of velocity may seem a decorative move, but its
significance cannot be overestimated and leads to the view of classical mechanics,
in the absence of dissipation, as an incompressible flow in an abstract but
physically crucial space which we now introduce. It is very convenient to imagine
that the six dynamical variables (r, p) or more explicitly (x1 , x2 , x3 , p1 , p2 , p3 )
constitute a space called the phase space of the system. The instantaneous
state of the system is then given by a point in this phase space. As time
progresses, the variables change in accordance with Eq. (2). The initial point in
phase space traces out a path called the phase trajectory of the particle.
We emphasise strongly once again that the initial values of both xi and pi
must be specified independently, in order to obtain a unique solution to the
equation of motion. Only after solving Eq. (2) with the specified initial values,
do we get xi (t) and pi (t) as a function of t. Different initial conditions lead, in
general, to different phase trajectories. The complete set of phase trajectories of
the particle is called its phase portrait.
All these ideas can be carried over to more complicated systems. For instance,
a system consisting of N particles requires 3N coordinate components to specify
the locations of all the particles. Together with the 3N corresponding velocities,
or momenta, there are 6N dynamical variables. Each phase trajectory of the
system is a one-dimensional curve in a 6N -dimensional phase space.
A rigid body requires only 6 independent coordinates to specify its location
in space: 3 to specify the location of its centre of mass, and 3 to specify its
orientation with respect to any fixed set of coordinate axes. The phase space of a
rigid body is therefore 12-dimensional, no matter how many particles it is made
up of.

2 Some properties of phase trajectories


If the force F does not explicitly depend on time, for example we do not force
the particle externally or change the parameters in time, the system is said to
be autonomous. It can still have dissipation, think of a dissipative harmonic
oscillator for an example.

2
p

D
A

Figure 1: Uniqueness of solutions forbids self intersection of phase space trajec-


tories. Simlarly two different trajectories also cannot intersect.

Some important general properties of phase trajectories are as follows:

• For an autonomous system a phase trajectory cannot intersect


itself, nor can two distinct phase trajectories (of a given system)
intersect each other. Suppose such an intersection did occur, as shown
in Fig. (1): the arrows indicate how the state of the system evolves as
time progresses. The system goes from state A, say, to B, C, D, . . .; the
trajectory intersects itself at B. But we can choose to begin with B as
the initial state. Subsequent states can then lie either on the segment
BC or the segment BD, as seen from the directions of the arrows on these
segments. But this is impossible, because the subsequent evolution must be
uniquely determined by the equation of motion. Hence a phase trajectory
cannot intersect itself.

• A closed phase trajectory implies periodic motion: While a phase


trajectory cannot intersect itself, it may return to its starting point, thus
forming a closed curve. The particle thus returns to its initial state after
a certain amount of time, and this is repeated over and over again. In
other words, the motion is periodic. The time period is equal to the

3
time taken to traverse the closed curve once. We will use this fact in what
follows.
The general nature of phase trajectories even for a single particle in 6 dimen-
sional phase space can be very complex, with a combination of orbits that are
periodic and those that are non-periodic, with the latter sometimes dominating.
The non-periodic phase trajectories can be so complicated and appear to be so
irregular that they are sometimes called “chaotic”, a topic that is “out of syl-
labus”, but which you can learn about in more advanced and specialized courses.
We now turn to the simplest case when the phase space is two-dimensional and
where there can be non-periodic motion, but no chaos!

3 Motion in one dimension


The motion of a particle restricted to move in one-dimension, say the x-axis, or
constrained to lie on some specified curve like a circle, is easily studied. Although
it is relatively simple, it illustrates several general features, and this helps us
understand more complicated kinds of motion.
Let the one-dimensional coordinate be called x, and let the force be only a
function of x. In particular it does not contain time t or speed ẋ. Newton’s
second law reads:
d2 x
m 2 = F (x). (3)
dt
Simple harmonic motion is a special case with F (x) = −k x where k > 0 is
constant. Another example is the simple pendulum with x = θ and F (θ) =
−(mg/l) sin θ. While this is a second order differential equation, it can integrated
once in the following way. Multiply both sides by the speed ẋ to get

dx d2 x dx
m 2
= F (x), (4)
dt dt dt
and realize that 2
dx d2 x

1d dx
= . (5)
dt dt2 2 dt dt
It would be nice if the RHS of Eq. (4) too can be written as a derivative of time.
Towards this end define a function V (x) such that

dV (x)
F (x) = − . (6)
dx
Note that this can always be done for any function F (x) as V (x) is nothing but
(the negative) its integral. Therefore
d dx d dx
− V (x) = − V (x) = F (x). (7)
dt dt dx dt

4
Using Eq. (5) and Eq. (7) in Eq. (4) yeilds
 2 !
d 1 dx
m + V (x) = 0. (8)
dt 2 dt

Thus the quantity within the brackets must be a constant, independent of time,
but dependent on initial conditions. Denoting this as E, we get
 2
1 dx
m + V (x) = E. (9)
2 dt
If the initial speed is x˙0 and initial position is x0 then
1
E = mx˙0 2 + V (x0 ). (10)
2
Although the position and speed with change with time, the combination in
Eq. (9) will remain the same. The physical interpretation of the E is that it is
the total energy, while of course the first term in the kinetic energy and V (x) is
the potential energy. The somewhat odd negative sign in the definition of V (x)
is to ensure that the total energy is defined as the sum of these two quantities.
The simple harmonic oscillator has the potential function V (x) = kx2 /2, while
the simple pendulum has the potential
Z Z
V (θ) = − F (θ)dθ = (mg/l) sin θdθ = −(mg/l) cos θ + C = (mg/l)(1 − cos θ).
(11)
Here C is a constant of integration chosen so that at θ = 0, when the pendulum
is at the bottom of its circular path, the potential energy is zero. Notice that
as we want to treat uniformly different problems with the equation in Eq. (net-
wononedim), we have taken F (θ) = −(mg/l) sin θ which differs from the force
acting on the pendulum which is lF (θ) and the potential V (θ) differs from the
potential energy by the factor l2 , that is the actual potential energy is l2 V (θ).
We will see what happens in higher dimensions later, but suffices to mention
here that the existence of a potential energy function places much stricter con-
ditions on the force and these are called conservative forces. Also note that if
the force is a also a function of time, say F (x, t), while it is possible to define a
time dependent potential energy V (x, t) such that
∂V (x, t)
F (x, t) = − ,
∂x
the sum of the kinetic and this potential energy is not a constant: the time
varying potential energy renders the total energy a variable. For example, the
forced harmonic oscillator with the force

F (x, t) = −k x + F0 cos ωt (12)

5
admits the potential energy
1
V (x, t) = kx2 − xF0 cos ωt. (13)
2
With general potentials and time dependent forces, motion in one-dimension
could itself be complex, for example a forced pendulum could become “chaotic”.
We therefore concentrate on the case when the force is not explicity dependent
on time, the autonomous case. The conservation of energy, which we have shown,
is actually sufficient to “solve” the problem of such one-dimensional motion in
the following sense. The phase space in this case has just two dimensions. The
equation Eq. (9) can be written using momentum instead of speed as
1 2
p + V (x) = E. (14)
2m
This represents a curve in the (x, p) phase plane, which is nothing but the phase
trajectory. Of course, to find x(t) explicitly as a function of t, we do have to solve
the equation of motion.

4 States of equilibrium in one-dimensional motion


The simplest trajectory in space or in phase space is one that does not move
at all! Recall that the total energy of a particle in one-dimensional motion is
given by Eq. (14). The force on the particle is given by F (x) = −dV (x)/dx .
The force vanishes at the extrema of V (x), so that dp/dt (or dẋ/dt) is zero at
these points. If dx/dt is also zero, then we have a situation in which both the
dynamical variables x and ẋ are stationary: we have a state of equilibrium.
If the initial conditions are such that the particle starts in such a state, it will
remain in that state for all time.
Alternatively look at the pair of first order differential equations, the equations
of motion:
dx p dp
= , = F (x). (15)
dt m dt
If x∗ is such that F (x∗ ) = 0, and if x(t = 0) = x∗ and p(t = 0) = p∗ = 0,
then for all time t we have x(t) = x∗ and p(t) = 0. Note that this x(t), p(t)
is a solution of the equations of motion (as dx/dt = dx∗ /dt = 0 = p(t) and
dp(t)/dt = 0 = F (x(t)) = F (x∗ )), which satisfies the initial conditions stated.
There can be only one solution with a given initial condition (“uniqueness”) and
therefore this is the solution.
In the case of the simple harmonic oscillator, the only extremum of V (x) =
(1/2)kx2 is a minimum at x∗ = 0. Therefore (x∗ = 0, p∗ = 0) is the sole state
of equilibrium for this system. In the case of the simple pendulum the potential
function V (θ) = (mg/l)(1 − cos θ) has extremum at three points θ1∗ = 0, θ2∗ = ±π.

6
The equilibrium point at (θ∗ = π, θ̇∗ = 0) corresponds to the pendulum being
at rest at the bottom, while the two equilibirum points at (θ∗ = π, θ̇∗ = 0) and
(θ∗ = −π, θ̇∗ = 0) are really one and the same and correspond to the pendulum
being at rest in the inverted position.
It is important to note that equilibrium states refer to states or points
in phase space, and not merely in coordinate space. For example, in the
case of linear harmonic oscillator, if x = 0 but ẋ 6= 0, we do not have a state of
equilibrium – rather, we merely have an instant at which the oscillator is passing
through the centre of oscillation. In all the systems we shall consider, however,
the total energy will be of the form in Eq. (9), or equivalently the equations of
motion will be of the form Eq. (15). Hence states of equilibrium automatically
correspond to points in phase space at which

1. all velocities (or momenta) are zero, together with the condition that

2. all force components are zero.

We shall therefore occasionally omit stating the former condition explicitly when
referring to states of equilibrium, but this is to be implicitly understood.
If the state of equilibrium corresponds to a minimum of V (x), say at some
point x = x∗ , it is easy to see that the force −dV /dx in the immediate neigh-
bourhood of this point, on both sides of it, is always directed towards this point.
When displaced slightly from x = x∗ , the particle will tend to return to this point.
(Hence the term, “restoring” force.) We then have a state of stable equilib-
rium at x = x∗ , p = p∗ = 0. The point x∗ = 0, p∗ = 0 (i.e., the origin) in the
phase plane of the simple harmonic oscillator is an example of such a state.
If the state of equilibrium x = x∗ corresponds to a maximum of V (x), it
is again easy to see that the force −dU/dx in the immediate neighbourhood of
x = x∗ is directed away from this point on either side of it. An infinitesimal
displacement from x = x∗ therefore causes the particle to move away from this
point. Therefore x = x∗ , p = p∗ = 0 is a state of unstable equilibrium.
The “inverted parabola” potential V (x) = −(1/2)kx2 (where k > 0 as before)
provides a simple example of this situation. No periodic motion is possible in
such a potential. In this case the origin in the phase plane (x = 0, p = 0) is a
state of unstable equilibrium.
Another important example is the simple pendulum where it follows that
(θ∗ = 0, p∗ = 0) is a stable equilibrium point, while (θ∗ = π, p∗ = 0) or (θ∗ =
−π, p∗ = 0) is an unstable equilibrium point.
Now consider a general potential V (x), with a simple extremum (either a
minimum or a maximum) at the point x = x∗ . Expanding V (x) in a Taylor
series about x = x∗ we have, correct to second order in (x − x∗ ),

V 00 (x∗ )
V (x) = V (x∗ ) + (x − x∗ ) V 0 (x∗ ) + (x − x∗ )2 + ··· (16)
2!

7
VHxL

x
-3 -2 -1 1 2 3
-1

-2

-3

-4

Figure 2: An example of a “double-well potential” V (x) = −2x2 + x4 /4 is shown


along with approximating parabolic potentials about the three extrema. The
latter are drawn with thinner lines.

But x = x∗ is an extremum of V (x), so that V 0 (x∗ ) = 0. Therefore


1
V (x) ≈ V (x∗ ) + V 00 (x∗ ) (x − x∗ )2 (17)
2
in the neighbourhood of x = x∗ . Now V (x∗ ) is just a constant, and so does
not contribute to the force on the particle. Equation (17) then shows that the
potential is effectively parabolic in the vicinity of an extremum: it is
like that of a simple harmonic oscillator if x = x∗ is a minimum (V 00 (x∗ ) > 0), or
like an inverted parabola if x = x∗ is a maximum (V 00 (x∗ ) < 0). Figure 3 shows a
“double well” potential with a maximum in the middle and two minima on either
side of it. The thinner lines indicate the parabolic or harmonic potentials that
approximate V (x) at its three extrema.
In the case of the simple pendulum the harmonic oscillator approximation at
the stable point is (V 00 (θ∗ )/2)(θ − θ∗p
)2 = (mg/2l)θ2 , as V 00 (θ) = (mg/l) cos θ.
Thus the angular frequency is ω = g/l, well known small-angle result. The
resultant equation of motion is
d2 θ
m = −V 0 (θ) = −(mg/l)θ, (18)
dt2
and the general solution is given by

θ(t) = Aeiωt + A∗ e−iωt (19)

On the other hand at the unstable point the approximate potential is

V (θ) = (2mg/l) − (mg/2l)(θ − π)2 . (20)

8
VHxL

20

10

x
-3 -2 -1 1 2 3

-10

-20

Figure 3: The cubic potential V (x) = x3 with a point of inflection at the origin.
Dynamically a stationary particle at the origin is unstable.

The resultant equation of motion is


d2 θ
m 2
= −V 0 (θ) = (mg/l)(θ − π). (21)
dt
which is solved by putting φ = θ − π so that φ satifies the homgeneous equation
d2 φ/dt2 = gφ/l, and therefore the solution is
√ √
φ = θ − π = Ae g/lt + Be− g/lt . (22)
Note that this solution unlike the stable solution cannot remain valid for long
as the deviation from π can be too large and make invalid the Taylor expansion
used in the first place. One exception is the solution with A = 0, which is such
that t → ∞ the solution is bounded and the pendulum approches ever closer the
unstable equilibrium point as φ → ∞, and hence θ → π. This corresponds to
a pendulum with total energy 2mgl (show this!) that approaches the unstable
equilibrium upright position, but indeed takes an infinite amount of time
doing so. We will say more about this later.
If at an equilibrium point V 00 (x∗ ) = 0, we cannot in general say if the equi-
librium is stable or unstable, further examination is needed. For example if
V (x) = x4 it is clear that x∗ = 0 is a stable equilibrium as this is a minimum
point, but for V (x) = −x4 the origin is an unstable equilibrium. In both cases
V 00 (x∗ ) = V 00 (0) = 0. If there is a continuous set of equilibrium points there
could be marginal stability, or otherwise called neutral or indifferent equilibrium.
A trivial case of this is free motion, when the equilibirum is neutral for all equil-
brium points of the form (x∗ , p∗ = 0) for any x∗ .
Consider, the potential V (x) = x3 . The equilibrium point x∗ = 0 is an
inflection point: we have V 0 (0) = 0, but V 00 (0) = 0 as well. A sketch of the

9
VHxL
6 ¥ ¥

4
L
3

x
1 2 3 4

Figure 4: A one-dimensional infinity deep potential well of width L is illustrated.

potential (Fig. ( 4)) shows immediately that the force is directed towards x = 0
at points to the right of it, but away from x = 0 at points to its left. Hence x = 0
behaves like a minimum of the potential for small displacements to its right, but
like a maximum for small displacements to its left. Stable equilibrium, however,
is obtained only if there is a restoring force for a small displacement in
each and every direction. It is thus clear that inflection points can only lead
to unstable equilibrium states in the one-dimensional case.
Apart from free motion, neutral equilibrium is possible in one-dimensional
motion only if we allow non-smooth potentials. One common model used in
physics for many situtaions is that of a particle that is free to move in some
region, but when it hits the boundary of the region it experiences an infinite
potential that it is unable to penetrate. One dimensional “wells”, studied in
first courses on quantum mechanics is of this kind with the particle say bouncing
freely between two walls, see Fig. (5). In this case all states with p = 0 are states
of neutral equilbrium. Higher dimensional generalizations of such potentials are
also very useful and well studied, for example in two dimensions, you may think
of a billiard table.

5 Integrating the equation of motion and peri-


odic orbits
The energy equation Eq. (9) enables us to solve easily one-dimensional problems
with potential energies that are not explicitly dependent on time. This equation
can used to solve for the momentum as
p
p = ± 2m(E − V (x)). (23)

10
Using p = m dx/dt leads to

dx p
m = ± 2m(E − V (x)). (24)
dt
The two signs of momentum or velocity imply that for a give position and en-
ergy there are two possible directions of motion with the same magntitude of
momentum.
The ± sign above may seem more like a minor irritant than an important
feature. Not only is periodic motion possible because of this, it is also responsible
for any type of non-monotonic behavior in the coordinate x(t) or momentum p(t)
as a function of time t. Indeed if only one sign say + sign is chosen, then dx/dt ≥ 0
for all time and x(t) can only increase with time, and if the sign is always −, it
can only decrease all the time: but with the signs switching at the turning points,
x(t) can be non-monotonic and lead to periodic orbits as well.
Say in some time interval p > 0, and the particle goes from x0 at time t0 to
x1 at time t1 . Then we have as a consequence of the above that
Z t1 Z x1 r
dx p m dx
m = 2m(E − V (x)), or dt = p . (25)
dt t0 x0 2 (E − V (x))

This in principle solves the problem as it tells us how to find the position at a later
time from a knowledge of the initial position. Note that the initial momentum
is implicitly specified through the energy. The integral may not be solvable in
terms of known standard mathematical functions, but the problem is considered
solved or “integrated”.
When V (x) = E, the momentum vanishes and the particle “turns” and these
turning points are important. If there are two solutions to the equation V (x) = E,
then these two turning points are such that the particle performs oscillations
between these at energy E. These oscillations are not in general “harmonic”,
but are periodic oscillations nevertheless. See the figure where the turning points
are marked as A and B for some trajectory whose energy is indicated by the
horizontal line. Let the coordinates here be xA and xB . Then the time to go
from A to B is Z xB r
m dx
TA→B = p . (26)
xA 2 (E − V (x))
Note that the time to go from B back to A, the “return journey” is
Z xA r
m −dx
TB→A = p = TA→B . (27)
xB 2 (E − V (x))

The momentum is negative on this part and we take the negative sign in Eq. (23).
Thus the time to go on the return journey, actually any segment of it, is equal

11
VHxL
4

A
B
3

x
-2 -1 1 2

Figure 6: Example of a potential energy graph with the horizontal line indicating
a specific energy for which the turning points are marked A and B.

to that of the forward. Thus the time period of the periodic motion is
Z xB r
m dx
T =2 p . (28)
xA 2 (E − V (x))

Example 1: The good old simple harmonic oscillator is given by the


2 2
potential
p V (x) = kx /2 and
p the turning points are given by kx /2 = E or
xA = − 2E/k and xB = 2E/k. Therefore the time period is
Z √2E/k r r Z 1 p r
m dx m 2/k du m
T =2 √ p =2 √ = 2π . (29)
− 2E/k 2 (E − kx /2)
2 2 −1 1 − u 2 k

Here the substitution u2 = kx2 /2E is used.


Example 2: The quartic anharmonic oscillator provides a non-trivial
example. Let the potential energy in this case be V (x) = λ x4 /4, where λ is a
positive constant. The motion is periodic, but not simple harmonic: the force
on the particle is F (x) = −dV /dx = −λ x3 , which is a nonlinear function
of x. The explicit solution for x(t) cannot be written in terms of elementary
periodic functions like sines or cosines. And yet the time period corresponding
to motion with any total energy E > 0 is easily found. The turning points are at
xB,A = ±(4E/λ)1/4 . The time period is therefore given by
Z xB √ r  2 1/4 Z 1
2m xB dx0
Z
mdx dx m
T =2 p =4 p =4 √ .
xA 2(E − λx4 /4) λ 0 x4B − x4 λE 0 1 − x04
(30)

12
The integral in the final equation is just a numerical constant(≈ 1.31). The point
to note is the dependence of T on the total energy. It decreases with increasing
energy, or equivalently amplitude xB .
Example 3: The (not so simple?) simple pendulum: While the small angle
approximation leads to simple harmonic motions, large amplitude oscillatons are
far from simple harmonic. Recall that the equation of motion is

d2 θ g
+ sin θ = 0,
dt2 l
and the energy equation is
 2
1 2 dθ
ml + mgl(1 − cos θ) = mgl(1 − cos θ0 ), (31)
2 dt

where θ0 is the maximum angle to which the pendulum rises, in other words it is
the amplitude. Thus the time period is
s Z s Z
2l θ0 dθ 2l θ0 dθ
T = √ =2 √ . (32)
g −θ0 cos θ − cos θ0 g 0 cos θ − cos θ0

It is left as an exercise for you to check that in the small angle approximation
cos θ ≈ 1 − θ2 /2, the above leads to the well-known result for the time period
of small oscillations. As cos θ is large for small amplitude oscillations, we would
like to go to some quantity that is small for small oscillations. Hence use cos A =
1 − 2 sin2 (A/2) for both the cosines to get
s Z θ0
l 1 dθ
T =2 s . (33)
g sin(θ0 /2) 0 sin2 (θ/2)
1−
sin2 (θ0 /2)

Now change variable according to sin x = sin(θ/2)/ sin(θ0 /2). Note that RHS is
less than or equal to 1 in modulus and hence x is real. The time period of the
simple pendulum can then be written as
s Z
π/2
l dx
T =4 p , (34)
g 0 1 − κ2 sin2 x

where κ = sin(θ0 /2) is the only parameter and depends on the amplitude. It is
easy to derive the small angle approximation from this by simply setting κ = 0.
However this simplified form for the time period is also exactly evaluated as an
elliptic integral of the first kind. This is not an “elementary function” but
is a well-studied and important mathematical function which arose historically

13
T
T0
8

0 Θ0
0.0 0.5 1.0 1.5 2.0 2.5 3.0

Figure 7: The scaled ptime period of a simple pendulum (T /T0 , where T is the
period, and T0 = 2π l/g) as a function of the amplitude θ0 . It starts from
1, which is the small angle approximation and increases goes to infinity as the
amplitude approaches π.

from the problem of determining the perimeter of an ellipse (hence the “elliptic”).
In fact the definition of the complete elliptic integral K(κ) is
Z π/2
dx
K(κ) = p , (35)
0 1 − κ2 sin2 x
and hence the time period of the simple pendulum is given exactly by
s
l
T =4 K (sin(θ0 /2)) . (36)
g

Of course this maybe considered to be of little use, as we may not know the
properties of K(κ). Fortunately it is well tabulated and used. Consider the
dimensionless ratio T /T0 where T0 is the small oscillation period:
r
T T g 2
= = K (sin(θ0 /2)) . (37)
T0 2π l π

Fig. (7) shows T /T0 as a function of the initial angle when 0 ≤ θ0 ≤ π. The
horizontal line at 1 is the small angle approximation. It is quite clear that for
larger angles the period substantially deviates from the small angle approximation
and increases to infinity as the amplitude tends to π. Note that in contrast

14
to the quartic oscillator whose period decreased with amplitude or energy, here
the period diverges all the way to infinity!
This is due to the existence of the unstable equilibrium point at θ0 = π. Indeed
we have seen earlier that motion near this unstable equilibrium point will have
solutions that are either exponentially increasing or decreasing with time. But we
can do better: we can solve exactly the equation of motion for a pendulum that
has the same energy as the equilibrium point, which is 2mgl. Actually there are
three distinct trajectories having this same energy: one that is exactly at
the equilibrium point, whose initial energy is all potential. This does not change,
it is alwatys at θ = π and p = 0. Other two are “dynamic” ones, which at
some time is at the bottom, θ = 0, with all the energy 2mgl being kinetic. The
distinction between these two dynamic orbits is the sense of rotation, one is going
clockwise, while the other anticlockwise. That these are really two and not one
should become clear soon. These dynamic trajectories can be exactly solved for
as follows.
The orbits which have just enough energy to reach the top (E = 2mgl), will
be such that  2
1 2 dθ
ml + mgl(1 − cos θ) = 2mgl, (38)
2 dt
or  2
1 2 dθ
ml = mgl(1 + cos θ) = 2mgl cos2 (θ/2). (39)
2 dt
Taking the positive squareroot corresponds to the pendulum moving upwards in
the anticlockwise sense and gives
r
dθ g
=2 cos(θ/2). (40)
dt l
This can happily be integrated, and let us take the time t = 0 at the bottom of
the orbit, that is θ(0) = 0, then (fill in the details)

θ(t) = 4 tan−1 (et g/l ) − π. (41)
Indeed this satisfies the initial condition, but more importantly it shows that as
t → ∞, θ(t) → π, and for all other times the angle is less than π. Thus the
orbit asymptotically approaches the equilibrium point. The fact that it
does not reach it in a finite time is also a consequence of the no intersection rule
we have mentioned earlier: The equilibrium point (π, 0) is itself an orbit and
cannot be intersected by another. The third orbit with this energy corresponds
to taking the negative squareroot and leads to the solution

θ(t) = 4 tan−1 (e−t g/l ) − π, (42)
which approaches −π asymptotically. Thus we have solved for all the three types
of motion for energy E = 2mgl and seen that the dynamic orbits take an infinite

15
time to reach the equilibrium point. It is now reasonable that for orbits that have
slightly smaller energy than 2mgl there will be oscillations but with very large
periods.
While we have explicitly integrated the large amplitude orbit that corresponds
to E = 2mgl, we cannot do so for any other energy exactly using ele-
mentary functions. Also notice that at this special energy there are really no
oscillations. This singular orbit separates two types of motion: those with energy
less than 2mgl, which oscillate as usual (referred to sometimes as ”librations”),
and those with energy greater than 2mgl, which rotate fully in either clockwise
or anticlockwise directions (referred to as ”rotations”). Thus the dynamic orbits
with E = 2mgl are called separatrices.
For energies that are different from 2mgl the exact solutions are in terms of
Jacobi elliptic functions which are fascinating doubly periodic functions of a
complex variable in general, but is outside the scope of this course. Suffice to say
that the Jacobi elliptic functions generalize both the trignometric functions as well
as the hyperbolic functions, and appear as solutions to a very broad number of
nonlinear differential equations including that of the quartic oscillator mentioned
previously! Thus we see that the simple pendulum has in it a rich source of a
lot of new behaviours, when compared to the simple harmonic oscillator, and
indeed any perturbations or modifications of the pendulum often render it very
”nonsimple”.
These results helps illustrate a very important fact:

• Except for the simple harmonic motion, the time period of pe-
riodic motion generally depends on the total energy (or, equiva-
lently, on the amplitude of oscillation).

• The independence of T on the amplitude of oscillation is what is so special


about simple harmonic motion.

6 Phase space trajectories for simple systems


Phase space analysis is a powerful and useful way of visualizing the dynamics.
We will now study a series of simple, but important, examples.

6.1 Free motion


In this case V (x) = 0 and F (x) = 0. Note that the potential energy function is
defined up to a constant, which we take to be zero here. The energy conservation
gives
1 2 √
p = E, or p = ± 2mE. (43)
2m

16
The phase trajectories for a given energy E are horizontal straight lines with
p = constant. There are two such lines for every energy E, with positive and
negative momenta. The physical interpretations are evident. These phase space

H0, 0L x

Figure 8: A few phase space trajectories for the case of free motion. The thick
line at p = 0 is filled with neutrally stable equilibrium points. The sense of flow
in time is indicated by the arrows.

traectories are solutions to the equations:


p dp
ẋ = , = 0. (44)
m dt
Thus we have that p(t) = p(0). Each of the lines in Fig. (8) depicts one such
solution. The different lines correspond to different initial conditions (x(0), p(0)).
The arrows on the lines indicate the direction in which the phase space variable
changes as time goes by. Thus when p(0) > 0, the flow is to the right, as from
the differential equation for x, it is clear that dx/dt > 0 and hence x increases
with time. On the other hand the flow below the p = 0 line flows to the left: the
position keeps decreasing. Thus time reamains in this figure only through these
arrows. Equilibrium or fixed points are all the points on the p = 0 axis. Note
that actually the entire line is a set of equilibrium points. Also it should be clear

17
that these equilibrium points are neutral in character. Shifting a particle with
zero momentum from “here” to “there” renders it still sitting in the new position.

6.2 Particle in a well


This example confines an otherwise free particle in one-dimension between two
impenetrable walls. This is commonly studied in quantum mechanics as a ”par-
ticle in a infinitely deep well”. The phase space trajectory corresponds now to
straight lines confined within the well boundaries, which are say separated by L.
More importantly a single trajectory has both positive and negative momentum
segments, see Fig. (9). Thus√ if the energy is E, the particle moves to the right
with the momentum p = 2mE, gets reflected at√ the right wall and suddenly
appears, in phase space at the momentum p = − 2mE due to an elastic colli-
sion, goes to the left and hits the other wall where its momentum is once again
reversed and becomes the original. This cycle repeats and indeed the motion is
periodic with period
r
2L 2Lm 2Lm 2m
T = = =√ =L . (45)
v p 2mE E

Hence unlike simple harmonic motion, the period is dependent on the energy and
is proportion to E −1/2 and decreases with energy. The interval with p = 0, 0 ≤
x ≤ L is an interval of neutral fixed points.
3 3
p p
2 2
p= 2mE p= 2mE
1 1

x x
0 0
0.5 1.0 1.5 2.0 2.5 3.0 0.5 1.0 1.5 2.0 2.5 3.0

-1 -1

p=- 2mE p=- 2mE


-2 -2

L
-3 -3

Figure 9: A phase space trajectory of a particle with energy E trapped in a


well 0 ≤ x ≤ L. A single trajectory is a periodic orbit with a discontinuous
momentum jump at the walls. The figure on right illustrates the same, but the
area enclosed by the orbit is shaded. The sense of flow in time is indicated by
the arrows.

While phase space and the “conjugate” variables (x, p) form the gateway from
classical to modern physics, such as quantum mechanics, statistical mechanics
and so on, it is perhaps easy to point out here one simple application without

18
going into details. Consider the simple phase space trajectory of a paticle in well
shown the right Fig. (9).
Let A be the area enclosed by the phase space trajectory, the area of the
shaded rectangle whose width√is the length of the well
√ or the distance between the
walls, L and the breadth is 2 2mE. Thus A = 2L 2mE. Quantum mechanics
asserts that such areas cannot take any possible value, but must be an integer
multiple of h, Planck’s constant. This leads to
√ n2 h2
2L 2mE = nh, or E = , n = 1, 2, · · · . (46)
8mL2
Exactly the same result is obtained from Schrodinger’s equation, that the energy
is quantized as above. Historically Bohr first used such a scheme to derive levels
of the hydrogen atom, more than 10 years before Schrodinger developed equations
that go by his name. Bohr was motivated to consider the area in phase space as
it has a classical dynamics significance as an action that does not change under
small perturbations of the system. While we will study a slightly different action
later on, the exact details of the connection between Schrodinger equation and
the Bohr approach is outside the scope of this course. Note that dimensionally
Planck constant is same as that of area of phase space.

6.3 The simple harmonic oscillator


This is an important and fundamental example. The potential energy of the
oscillator is V (x) = Kx2 /2 with K > 0. E is now the sum of two squares, so
that the physically possible values of the total energy are restricted to E ≥ 0.
For any E > 0, it is clear that
p the physically accessible region of x is given by
x− ≤ x ≤ x+ , where x± = ± 2E/K. The two points x± are the turning points
– the velocity of the particle vanishes at these point. Each phase trajectory is
given by
p2 Kx2 p2 x2
+ = E, or + = 1. (47)
2m 2 2Em (2E/K)
This is of the form of an p ellipse in the x −√ p phase space as this satisfies
x2 /a2 + p2 /b2 = 1, with a = 2E/K and b = 2mE. Every E > 0 specifies
a particular ellipse. The phase portrait is a family of such ellipses in the phase
plane, see Fig. (10). The initial conditions determine the particular ellipse on
which the system lies, as well as the precise point on it at which the motion starts;
but the sense or direction in which the ellipse is traversed as time progresses is not
arbitrary – it is clockwise, as indicated by the arrows. This follows immediately
from the equations of motion:
dx p dp
= , = −Kx, (48)
dt m dt

19
as looking at the first quadrant where x, p > 0 we see that dx/dt > 0 and
dp/dt < 0, thus the position increases, while the momentum decreases and hence
the sense of motion must be clockwise. This is consistent with the other three
quadrants as well, as you may now check. All phase trajectories in this example

4 p

x
-4 -2 2 4

-2

-4

Figure 10: Some phase space trajectories of a harmonic oscillator. These are
ellipses and the center is the stable fixed point. The sense of flow in time is
indicated by the arrows.

are closed and hence represent periodic motion. The point (x = 0, p = 0) is the
only equilibrium or fixed point and is stable. The period has already been
calculated above and is independent of energy.
The area enclosed by a phase space orbitp is the√area of an ellipse with half
axes
p a and b and is therefore A = π a b = π 2E/K 2mE = 2π E/ω where ω =
K/m is the angular frequency. Thus Bohr’s rule applied here gives 2πE/ω =
nh or En = nh̄ω, where n is an integer, En denotes corresponding energy and
h̄ = h/(2π). This is very close to the exact quantum energies given by En =
(n + 1/2)h̄ω. The missing “zero-point” energy given by h̄ω/2 is considered to
be a quantum effect that is not captured by this approximation. Classically we
emphasize that all the energies are allowed and each phase space trajectory is an
ellipse for the harmonic oscillator.

20
6.4 The simple pendulum phase space

1.0
Rotations Librations

0.5

-Π Π
-3 -2 -1 1 2 3

-0.5

-1.0

Figure 11: A few phase space trajectories of the simple pendulum. The two ends
at θ = ±π are to be identified with each other. There is a stable fixed point at
(00) and an unstable one at (π, 0). The dynamic orbits with energy 2mgl are
marked by thicker lines. Trajectories insides these are oscillations or “librations”
and have energy < 2mgl, while those outside have energy > 2mgl and undergo
rotations.

Consider again the plane pendulum, which we recall is a mass m suspended


from a rigid massless rod of length l and is allowed to swing about one end
without any friction and in a vertical plane. The phase space of the pendulum
is properly described by the pair of angle and angular momentum. The angle θ,
the deviation from the downward vertical say takes values in the range [−π, π],
with θ = 0 being the position of the lowest swing. Then the torque about the
point of support is mgl sin θ in the clockwise sense. The angular momentum, pθ
is equal to ml2 dθ/dt and therefore we have the following equations:
dθ pθ dpθ
= 2
, = −mgl sin θ. (49)
dt ml dt
Note that this is again a coupled set of first order differential equations and the
phase space can be the space (θ, pθ ). As θ is an angle this space is actually a
cylinder rather than a plane.
The energy equation now reads
p2θ
+ mgl(1 − cos θ) = E. (50)
2ml2
The phase space trajectories are simply curves that satisfy this equation and are
different depending on the value of the energy E, see Fig. (11). As discussed

21
earlier, when E < 2mgl there are oscillations or “librations”. For small ampli-
tudes when cos θ ≈ 1 − θ2 /2 the energy equation is approximately an ellipse, but
strictly speaking it is not, and with increasing amplitude it most certainly is not.
The phase space curves have no special geometric names.

Unstable fixed
point
0.5

Anticlockwise swing

-3 -2 -1 1 2 3

-Π Π

Clockwise swing
-0.5

Figure 12: Just the three orbits with energy E = 2mgl are shown. The first are
the fixed points at (±π, 0) which is just one point and corresponds to teh pen-
dulum being upright and still. Another the upper curve has positive momentum
and corresponds to a pendulum that swings anticlockwise and takes an infinite
time to reach the fixed point. The third is the lower curve and corresponds to a
clockwise swing. Notice that these orbits seem to intersect, but do not as they
never reach the fixed point. Eq. (51) gives the form of these curves.

For example the orbits with exactly the energy E = 2mgl are given by the
curves p
pθ = ±2ml gl cos(θ/2) (51)
The positive sign corresponds to the anticlockwise swing of the pendulum that
asymptotically reaches the unstable equilibrium point at (π, 0), while the negative
sign to the clockwise swing of the same nature. Fig. (12) shows and discusses
these in detail.

22

You might also like