SPA5304 Physical Dynamics Lecture 10

David Vegh

29 January 2019

1 Lagrangian mechanics

1.1 Example: The planar pendulum

The purpose of this section is to:

(i) see the power of the Lagrangian approach

(ii) demonstrate that the variational principle works for constrained systems

1.1.1 Newtonian approach

p~˙ = m~g + N
~

Components:


mẍ = −N sin φ
mÿ = mg − N cos φ
where N ≡ |N~ |.
~ is a constraint force. It is perpendicular to the motion ⇒ does no work ⇒ energy is conserved.
N
Relation between the Cartesian coordinates (x, y) and φ:

ẍ = l(+ cos φ φ̈ − sin φ φ̇2 )

  
x = l sin φ ẋ = l(cos φ)φ̇
⇒ ⇒
y = l cos φ ẏ = −l(sin φ)φ̇ ÿ = l(− sin φ φ̈ − cos φ φ̇2 )

Plug these into the equation of motion:

ml(cos φ φ̈ − sin φ φ̇2 ) = −N sin φ



ml(− sin φ φ̈ − cos φ φ̇2 ) = mg − N cos φ

1
Multiply the first equation by cos φ and the second equation by sin φ, then subtract the two equations from
each other to eliminate N . We get
g
φ̈ = − sin φ
l
N can be expressed
ml(cos φ φ̈ − sin φ φ̇2 )
N= = mg cos φ + mlφ̇2
− sin φ | {z } | {z }
counter gravity centripetal force

~ . We also needed the EOMs for x and y to derive an EOM for φ.

We had to introduce a constraint force N

1.1.2 Lagrangian approach

There is one (holonomic) constraint: r = l. Thus, there is 1 degree of freedom. We can use φ as the
generalised coordinate.  
x = l sin φ ẋ = l(cos φ)φ̇
⇒
y = l cos φ ẏ = −l(sin φ)φ̇

• Kinetic energy
Plug the above expressions for ẋ and ẏ in the Cartesian formula:
1 1
T = m(ẋ2 + ẏ 2 ) = ml2 φ̇2
2 2
Alternatively, we can start with the kinetic energy expressed in polar coordinates, and then set r = l, ṙ = 0:
1 1
T = m(ṙ2 + r2 φ̇2 ) = ml2 φ̇2
2 2

• Potential energy

V = −mgy = −mgl cos φ (the y axis is pointing downwards!)

• Lagrangian

1 2 2
L=T −V = ml φ̇ + mgl cos φ
2

The momentum conjugate to φ:

∂L
pφ = = ml2 φ̇
∂ φ̇
The Euler-Lagrange equation
d ∂L ∂L
=
dt ∂ φ̇ ∂φ
gives
ml2 φ̈ = −mgl sin φ

g
φ̈ = − sin φ
l

2
 • This derivation is much simpler and more straightforward than the old Newtonian method:

~ never appeared.
(i) N

(ii) Directly got the equation of motion for φ.

• The energy E = T + V is conserved. (Check it)

• pφ is angular momentum

~ = ~r × m~r˙
L
p
~ = l · m ẋ2 + ẏ 2 = l · ml|φ̇| = |pφ |
|L|
Therefore
~
|pφ | = |L|

3
2 Constraints
• Holonomic constraints
Constraints that can be written in the form of

f1 (~r1 , . . . , ~rN , t) = 0 

f2 (~r1 , . . . , ~rN , t) = 0 

.. h equations → reduce the #DoF from 3N to 3N − h ≡ n
. 



fh (~r1 , . . . , ~rN , t) = 0

The constraint equations can be solved in therms of n independent variables

~q = (q1 , . . . , qn )

These are the generalised coordinates.


 ~r1 = ~r1 (q1 , . . . , qn )

..
 .

~rN = ~rN (q1 , . . . , qn )

• Non-holonomic constraints
These are all other types of constraints which are not holonomic.

4
2.1 Examples for holonomic constraints
2.1.1 The planar pendulum

The constraint equation is

f = x2 + y 2 − l 2 = 0

This can be solved using the generalised coordinate φ as

x = l sin φ y = l cos φ

2.1.2 Rigid bodies

|~ri − ~rj | − cij = 0

2.1.3 A disk rolling without slipping in one dimension

No slipping means dx = Rdφ, or if we divide both sides by dt,

ẋ = Rφ̇ (1)
This can be integrated to give
x = Rφ + const.
which is holonomic. We can use either x or φ as the generalised coordinate.

2.2 Examples for non-holonomic constraints

2.2.1 A disk rolling without slipping in higher dimensions

In higher dimensions, the analog of (1) cannot be integrated and the constraint is therefore non-holonomic
(see Goldstein et al. p15-16).

2.2.2 Other examples

5
2.3 Variational calculus with constraints

Z t2
S= dt L(~r, ~r˙ )
t1

Find the extremum as before. Now the variation is restricted to be in M :

Z t2 Z t2   Z t2  
∂L ∂L ∂L d ∂L
0 = δM S = δM dt L(~r, ~r˙ ) = dt δM ~r + δM ~r˙ = dt − δM ~r
t1 t1 ∂~r ∂~r˙ t1 ∂~r dt ∂~r˙

This is the same result as in the unconstrained case, except we have δM ~r instead of δ~r. From this result
we cannot conclude that the Euler-Lagrange equation inside the parentheses vanishes, because δM ~r is not
completely arbitrary: it is constrained to lie in the tangent space of M .

So the above result does not put any constraint on the normal component and we get the equation
d ∂L ∂L ~
− =N
dt ∂~r˙ ∂~r
~ is a constraint force perependicular to M .
where N

6
2.4 Procedure for solving problems with holonomic constraints
(i) Determine the configuration manifold (e.g. x2 + y 2 = l2 ) and introduce generalised coordinates on it:

{qi }, i = 1, 2, . . . , n where n is the #DoF

(ii) Re-express the kinetic energy T in terms of ~q and ~q˙.

If the constraints are time-independent and ~ri = ~ri (~q), then T is quadratic in ~q˙

X1
T = aij (~q)q˙i q˙j
i,j
2

E.g. planar pendulum: q = φ and T = 1

2 ml2 φ̇2
|{z}
a(φ)

(iii) Construct the Lagrangian

L = T − V (~q)

and solve the Euler-Lagrange equations

d ∂L ∂L
= , i = 1, 2, . . . , n
dt ∂ q̇i ∂qi