Anda di halaman 1dari 10

Classical Mechanics

Examples (Lagrange Multipliers)


Dipan Kumar Ghosh
Physics Department, Indian Institute of Technology Bombay
Powai, Mumbai 400076
September 3, 2015

1 Introduction
We have seen that the effect of holonomic constraints is to reduce the number of degrees of
freedom. If there are N particles, we expect 3N degrees of freedom if the motion is uncon-
strained. If there are k holonomic constraints, the number of degrees of freedom becomes
3N −k and we need this many independent generalised coordinates to describe the system.
For non-holonomic constraints, however, the number of generalised coordinates required
is more than the number of degrees of freedom. If there are m non-holonomic constraints,
in addition to the k holonomic constraints, the number of generalised cordonates is still
3N − k, i.e. m more than the number of degrees of freedom. These additional quantities
are eliminated from the equations of motion by method of Lagrange Multipliers.

1.1 A mathematical Example:


Find extremal value of f (x, y) = xy subject to the constraint
x2 y 2
h(x, y) = + −1=0
8 2
√ √
This is an equation to al ellipse with major axis 2 2 and minor axis 2. The level
curves for f (x, y) = xy are hyperbola. Suppose the level curve for f (x, y) intersects the
constraint curve h(x, y) at P. If we move along the constraint curve to its right, the value
of xy increases whereas if we move to the left its value decreases. Thus the point P
cannot be a point of minimum or a maximum. Clearly, when the point is an extremum,
the level curve of f and the constraint curve touch each other and they have a common
tangent. Hence at the common point their normals are parallel (or antiparallel) to each
other. Thus the set of points S which satisfies h(x, y) = 0 and ∇f + λ∇h = 0 for some λ
contains the extremal value of f , subject to the given constraint.
D.
c K. Ghosh, IIT Bombay 2

We define
F (x, y, λ) = f (x, y) + λh(x, y)
Setting gradient of F equal to zero implies
   
∂f ∂h ∂f ∂h
+λ x̂ + +λ ŷ + hλ = 0
∂x ∂x ∂y ∂y

which gives us
∂f ∂h
+λ =0
∂x ∂x
∂f ∂h
+λ =0
∂y ∂y
h=0 (1)

For the present problem, we have,

x2 y 2
 
F (x, y, h) = xy − λ + −1
8 2

In this case from (1) we get

λ
y− x=0
4
x − λy = 0
x2 y 2
+ −1=0
8 2
D.
c K. Ghosh, IIT Bombay 3

which gives λ = ±2. Substituting the possible values of λ into the above equations, we
find two maximum values of f = xy = 2 at (1, 2), (−1, −2) and two minimum values
xy = −2 at (1, −2), (−1, 2).

2 Euler Lagrange Equation with constraint


The method of Lagrange multiplier is used in classical mechanics in handling situations
where the number of dynamical variables happens to be more than the number of degrees
of freedom.
R
We had seen that the principle of least action is expressed in the form δ Ldt = 0 from
which we had obtained the Euler Lagrange equation as
XZ    
d ∂L ∂L
dt − δqj = 0
j
dt ∂ q̇j ∂qj

Using the fact that the generalised coordinates are independent, we derived the Euler
Lagrange equation for each pair pf generalised coordinate and velocity
 
d ∂L ∂L
− =0 (2)
dt ∂ q̇j ∂qj
In the presence of non-holonomic constraints, the generalised coordinates are not indepen-
dent as their number is greater than the number of degrees of freedom and consequently
(2) is not valid. In such situations (and also in case care is not taken to reduce the number
of generalised coordinates using the holonomic constraints), the method of undetermined
multiplier is useful. (The method has limitations, for instance, it cannot be used in cases
where the constraints are stated as inequalities)
Suppose there are m number of non-holonomic constraints involving the generalised co-
ordinates in differential form n
X
ar,j dqj + br,j dt = 0 (3)
j=1

ar,j and br,j may depend on j and t. Here r is an index which runs from 1 to m and (3) is
actually m equations, one for each value of r. We can get correct equations motion if the
varied paths are virtual displacements from actual motion in which case the constraint
is nj=1 ar,j dq= 0 because vital displacements take place over constant time. We can then
P

rewrite the principle of least action as


n Z "   m
#
X d ∂L ∂L X
j dt − + λr ar,j δqj = 0
j=1
dt ∂ q̇j ∂qj r=1
Pn Pm
Note that the additional term j=1 r=1 λr ar,j δqj is actually zero and hence we can
put it inside the integral. δqj s are not independent and satisfy (3). Since we have m
D.
c K. Ghosh, IIT Bombay 4

undetermined multipliers λr , we can choose them such that the first m terms, i.e. j = 1
to m is each zero. Suppose we choose λj such that for j = 1, 2, . . . m, the equation to be
satisfied is
  m
d ∂L ∂L X
− + λr ar,j = 0 (4)
dt ∂ q̇j ∂qj r=1
Note that the last term in the above equation is no longer zero as the sum over j is missing
and we have simply redistributed the term which was zero in various ways. With the λr
determined by (4), we are left with
n Z "   m
#
X d ∂L ∂L X
dt − + λr ar,j δqj = 0
j=m+1
dt ∂ q̇j ∂qj r=1

However, now our qj are independent and we have, as a consequence, for j = m + 1, . . . , n


  m
d ∂L ∂L X
− + λr ar,j = 0 (5)
dt ∂ q̇j ∂qj r=1

(4) and (5) allows us to write a single equation for j = 1, 2, . . . n,


  m
d ∂L ∂L X
− + λr ar,j = 0 (6)
dt ∂ q̇j ∂qj r=1
(4) determines the m values of λ and (5) gives us n equations of motion. Define
m
X
Qj = − λr ar,j (7)
r=1

Equation (5) can now be written as for j = 1, 2, . . . n.


 
d ∂L ∂L
− = Qj (8)
dt ∂ q̇j ∂qj

The right hand side of the above is seen to be the generalised force corresponding to the
constraint conditions.

2.1 Example : A sliding mass on a paraboloid of revolution


Consider a particle of mass m moving frictionlessly on a paraboloid of revolution given
by x2 + y 2 = az.
D.
c K. Ghosh, IIT Bombay 5

^z

φ
m ρ^

x
The Lagrangian of the system is
1
L = T − V = m[ρ̇2 + ρ2 ϕ̇2 + ż 2 ] + mgz (9)
2
p
where ρ = x2 + y 2 is the particle from the z axis. The system is holonomic with two
degrees of freedom since the constraint ρ2 = az can be used to eliminate one of the
coordinates.
Let us write the constraint as
2ρδρ − aδz = 0 (10)
Comparing (10) with (3) we identify aρ = 2ρ, aϕ = 0 and az = −a.
As there is only one constraint equation, there is only one Lagrange multiplier λ and
the Euler Lagrange equation is
 
d ∂L ∂L
− = Qj = λaj
dt ∂ q̇j ∂qj

The Euler Lagrange equations for the three coordinates are


 
d ∂L ∂L
− = λ × 2ρ
dt ∂ ρ̇ ∂ρ
 
d ∂L ∂L
− =0
dt ∂ ϕ̇ ∂ϕ
 
d ∂L ∂L
− = −λ × a
dt ∂ ż ∂z

These equations simplify to


D.
c K. Ghosh, IIT Bombay 6

m(ρ̈ − ρϕ̇2 ) = 2λρ


d
m (ρ2 ϕ̇) = 0 =⇒ mρ2 ω0 = L
dt
mz̈ = −mg − λa

We can solve these equations to determine ρ(t), z(t), ϕ(t) and λ.


Consider the case where the mass goes around in a horizontal
√ circe at a height h from
the bottom. z = h. This also implies ρ = constant = ha. We then have (from the last
mg 2g
equation) λ = − . ϕ̇ = constant = ω0 = (from the first equation, using ρ̈ = 0)
a a
mg m 2
λ=− = − ω0
a 2
The force which is responsible for the circular motion is the component N sin ψ of the
dz 2ρ
normal force. tan ψ = = The component of the normal force towards the centre of
dρ a
the circle is

−N sin ψ = −N p ≡ −mρω02
2
4ρ + a 2

The vertical component of N must be equal to mg


a
N cos ψ = N p = mg
4ρ2 + a2

Thus
mg N 1
λ=− = −p = − mω02
a 4ρ2 + a2 2
which is what we expected.

2.2 A hoop rolling down the top of a cylinder without slipping


Consider the case of a hoop rolling down the top of a cylinder without slipping. The
constraint is holonomic till the mass slides off. Let r be the distance from O to C. θ is
the polar coordinate of O and ϕ is the angle of rotation of the hoop about its own axis.
R is the radius of the cylinder.

θ R

V=0
C
D.
c K. Ghosh, IIT Bombay 7

The Lagrangian is
1 1
L = m(ṙ2 + r2 θ̇2 ) + ma2 ϕ̇2 − mgr cos θ
2 2
Constraints are : (i) r = R + a (ii) (R + a)θ̇ = aϕ̇ The first is a holonomic constraint
: f1 = r − (R + a) and the second is a non-holonomic (though integrable) constraint
f2 = −(R + a)θ̇ + aϕ̇ = 0. With the former we associate λ (associate with normal
reaction) while with the latter we associate µ (associated with the tangential force required
for rolling motion). The Euler Lagrange equations (for r, θ and ϕ) are
 
d ∂L ∂L ∂f1 ∂f2
− =λ +µ
dt ∂ q̇j ∂qj ∂qj ∂ q̇j
The equations give

mr̈ − mrθ̇2 + mg cos θ = λ


mr2 θ̈ + 2mrṙθ̇ − mgr sin θ = −µ(R + a)
ma2 ϕ̈ = µa

Hence µ = maϕ̈ = m(R + a)θ̈ Substitute this in the second equation and use ṙ = 0.
g sin θ
θ̈ =
2(R + a)

We multiply the above equation by θ̇ and integrate


g
θ̇2 = (1 − cos θ)
R+a
Use this in the radial equation (put ṙ = 0)

λ = −mrθ̇2 + mg cos θ
g
= −m(R + a) (1 − cos θ) + mg cos θ
R+a
= mg(2 cos θ − 1)

For θ = 0, λ = mg, the normal force. For θ > 60◦ , λ becomes negative and contact
with the cylinder is lost.

2.3 Example : A Rolling hoop on an incline


Consider a hoop rolling down without slipping on an incline. The problem is one of
holonomic constraint and has been solved earlier. However, we will use this problem to
illustrate the method of Lagrange multiplier. In this case there is just one degree of free-
dom and there is just one generalised coordinate, i.e. the angle θ by which the hoop rolls.
The distance moved along the incline is x = Rθ.
D.
c K. Ghosh, IIT Bombay 8

R
x θ

Let us, however, use both x and θ are our generalised coordinates. We then have one
‘holonomic’ constraint
h(x, θ) = x − Rθ (11)
An observation may be made here that (11) is not in the form (3). A holonomic
constraint is expressed in the form f (q1 , q2 , . . . , qn , t) = 0 We can write it in a differential
form n
X ∂f ∂f
δqk + δt = 0 (12)
k=1
∂q k ∂t
Comparing this with (3) we have
∂f
ar,k =
∂qk
∂f
br,t = (13)
∂t
Thus in the present case, where we have only one constraint (11) (we drop the index r)

b=0

ax = (x − Rθ) = 1
∂x

aθ = (x − Rθ) = −R
∂θ
The Lagrangian is then given by
1 1
L = mẋ2 + mR2 θ̇2 − mg(l − x) sin α
2 2
where l is the length of the incline. The Euler-Lagrange equations for the coordinate x
and θ are given by  
d ∂L ∂L
− −λ×1=0
dt ∂ ẋ ∂x
i.e.
mẍ − mg sin α − λ = 0 (14)
and  
d ∂L ∂L
− +λ×R=0
dt ∂ θ̇ ∂θ
D.
c K. Ghosh, IIT Bombay 9

i.e.
mRθ̈ + λR = 0 (15)
Equations (14), (15) along with the equation to the constraint
ẋ − Rθ̇ = 0 (16)
provide the necessary equations for solution of the problem. From (15) and (16) we get
λ = −mẍ. Substituting these in (14) we get
g sin α
ẍ = (17)
2
mg sin α
λ=− (18)
2
The force of constraint corresponding to the motion along x direction is given by
∂h mg sin α mg sin α
|λ |=| − × 1 |= −
∂x 2 2

2.4 Rolling with Spinning – introducing a non-holonomic con-


straint
Let us now allow the hoop to spin about the vertical axis, in addition to rolling. Let us
take the direction down the incline as the x-direction and the direction perpendicular to
it along the plane as the y- direction.

R
θ ϕ

The kinetic energy of the hoop is given by


1 2 1 2 2 1 mR2 2
T = mv + mR θ̇ + ϕ̇
2 2 2 2
1
= mR2 θ̇2 + mR2 ϕ̇2
4
Here we have used the fact that the hoop spins about its diameter and the moment of
inertia about the diameter is mR2 /2. The Lagrangian is
1
L = mR2 θ̇2 + mR2 ϕ̇2 − mg(l − x) sin α
4
D.
c K. Ghosh, IIT Bombay 10

Here y is cyclic and ẏ does not enter into the Hamiltonian. Hence we ignore the equation
of motion for y.
Since the hoop rolls without sliding, the velocity of the rim is v = Rθ̇. When the hoop
has spanned by an angle ϕ, we have

ẋ = Rθ̇ cos ϕ (19)


ẏ = Rθ̇ sin ϕ (20)

These are non-integrable constraints because these cannot be used to connect x with θ
and ϕ. The problem has two degrees of freedom.
Let us freeze the motion and make virtual displacement δθ of the angle θ(a displacement
in tilt angle is of second order). We then have

δx = R cos ϕδθ

As before we can get ax = 1 and aθ = −R cos ϕ. We can now write the Lagrange equations
of motion and get the following equations:

−mg sin α − λ = 0 (21)


mR2 θ̈ + λR cos ϕ = 0 (22)
2
mR ϕ̈ = 0 (23)

These equations determine the motion completely in terms of θ and ϕ. Note that x has
been alienated from the equations of motion.

Anda mungkin juga menyukai