Physics 430: Lecture 15

Lagrange’s Equations

Dale E. Gary
NJIT Physics Department
 Problem 6.19 y(x) (x2,y2)
y
 Statement of the problem:
(x1,y1)
 A surface of revolution is generated as follows: Two fixed
points (x1, y1) and (x2, y2) in the x-y plane are joined by a x
curve y = y(x). [Actually, you’ll make life easier if you start
out writing this as x = x(y).] The whole curve is now rotated
about the x axis to generate a surface. Show that the curve
for which the area of the surface is stationary has the form ds

y  yo cosh[( x  xo ) / yo ],
where xo and yo are constants. (This is often called the soap-bubble problem,
since the resulting surface is usually the shape of a soap bubble held by two
coaxial rings of radii y1 and y2.)
 Solution:
 What we have to minimize now is not a line, but rather a surface, dA = y dds.
 We could write this
2 x2
  termswill be non-zero.
  2
A  2 yds
but then both Euler-Lagrange 2 y 1  y ( x ) dx, Instead, follow the hint:
1 x1

2 y2
A 2  yds 2  y 1  x( y ) 2 dy,
1 y1

October 21, 2010

 Problem 6.19, cont’d
 Solution, cont’d:
 We identify f  y 1  x( y ) 2 ,

f d f d yx yx
so 0, and  0  C
x dy x dy 1  x2
1  x 2

 Squaring, we havey 2 x2 C 2 1  x2 .  

C
 Solving for x  .
2 2
y C
 Integrating
C dy du  y
x  dy  C  C arccosh    xo .
y2  C 2 y2
1 u2  1 C
C2
 The statement of the problem wants y(x), so we have to solve for y, to
get
y  yo cosh[( x  xo ) / yo ],
where we identify C = yo.
October 21, 2010
 6.4 More Than Two

Variables
Aside: Maximum and minimum vs. stationary—geodetics on
surfaces.
 When we discussed the shortest path between two points, we
examined the situation in the figure below.
 However, this is not the most general path. Here is one that
cannot be expressed as a function y(x).
 To handle this case, we need to consider a path specified by an
u2
independent variable u along the L path, i.e. duy. = y(u) .

2 x = x(u),2
x ( u )  y (u )
 u y Y ( x)  y ( x) ( x) (wrong)
Now the length of the path is then 1 y2
 In general, the
2 problem is 2
S  f [ x(u ), y (u ), x(u ), y(u ), u ]du. 1
1
y = y(x) (right)
y1 x
 This may seem more difficult than the previous case, x1 x2
but in fact we justx proceed
 x(u )  in the same
(u );  (u1 ) manner
 (u2 ) 0.with
the “wrong” paths y  y (u )   (u );  (u )  (u ) 0. 1 2
October 21, 2010
 More Than Two Variables-2
 The stationary path is the one for which
S S
0 and 0 .
 
 Following the same procedure as before, we end up with two Euler-
Lagrange equations that the function f must satisfy:
f d f f d f
 0 and  0.
x du x  y du y 
 In Chapter 7, we will meet problems with several variables, with time t
as the independent variable. In these cases, there will be two or
more Euler-Lagrange equations to satisfy (for example, equations for
x, y and z).
 One of the great strengths of Lagrangian mechanics is its ability to
deal with cartesian, cylindrical, spherical, and any other coordinate
systems with ease.
 In order to generalize our discussion, we write (x, y, z), (, , z) or (r, , )
as generalized coordinates (q1, q2, …, qN).
October 21, 2010
 The Lagrangian
 In the next chapter, we will use slightly different notation, and refer
to our integrand function
f [q1 , q2 ,, q N , q 1 , q 2 ,, q N , t ],
as the Lagrangian
L L [q1 , q2 , , q N , q 1 , q 2 ,, q N , t ].
 Note that because the independent variable is t, we canq i use ratherqi
than .
 We will then make the action integral stationary
S L [q1 , q2 , , q N , q 1 , q 2 , , q N , t ] dt ,

which requires that L satisfy

L d L L d L L d L
 ;  ; ;  .
q1 dt q 1 q2 dt q 2 q N dt q N

 We will develop this more next time.

October 21, 2010
 Advantages Over Newtonian
Mechanics
 We are now going to use the ideas of the previous lecture to
develop a new formalism for mechanics, called Lagrangian
mechanics, invented by Lagrange (1736-1813).
 There are two important advantages of the Lagrange formalism
over that of Newton.
 First, as we have seen, Lagrange’s equations take the same form
for any coordinate system, so that the method of solution proceeds
in the same way for any problem.
 Second, the Lagrangian approach eliminates the forces of
constraint, which we talked about in Chapter 4. This makes the
Lagrangian formalism easier to solve in constrained problems.
 This chapter is the heart of advanced classical mechanics, but it
introduces some new methods that will take getting used to. Once
you master it, you will find it an extraordinarily powerful way to
solve mechanics problems.
 To help you master it, we will spend the next three lectures on it,
and you will be solving a number of problems. However, there are
many more at the end of the chapter, and you are advised to try a
few more. October 21, 2010
 7.1 Lagrange’s Equations for
Unconstrained Motion
 Recall our general function f, that played such a large role in the
Euler-Lagrange equation f d f
 0.
y dx y
 To make the connection to mechanics, we are now going to show that
a function called the Lagrangian, L = T – U, is the function that, when
used in the Euler-Lagrange equation, leads to the equation of motion
for a particle.
 Here, T is the particle’s1 kinetic
1 energy
T  2 mv  2 mr  2 m( x  y  z ),
2 2 1 2 2 2

and U is the potential energy

U U (r ) U ( x, y, z ).
 Note that L is NOT the total energy, E = T + U. One can ask why the
quantity T – U should give rise to the equation of motion, but there
seems to be no good answer.
 I do note, however, that the problem can be cast in terms of the total
energy, which gives rise to Hamiltonian mechanics (sec. 7.8 and
October 21, 2010
chap. 13).
 Lagrangian
 Obviously, with the dependences of T on the x, y, z velocities, and U
on the x, y, z positions, the Lagrangian depends on both, i.e.
L L [ x, y, z , x , y , z , t ].
 Let’s look at the first two derivatives
L U L T
 Fx , and  mx  p x .
x x x x
 But note that if we differentiate the second equation with respect
to time we get d L d Note: this last equality is
 mx mx  p x Fx , only true in an inertial frame.
dt x dt
 So you can see that the Lagrange equation is manifestly true for a
free particle: L d L
 .
x dt x
 In Cartesian coordinates (so far) in three dimensions, we have:
L d L L d L L d L
 ,  , and  .
x dt x y dt y z dt z
October 21, 2010
 Connection to Euler-
 Lagrange
If you compare the Lagrange equations with the Euler-Lagrange
equation we developed in the previous chapter, you see that they
are identical.
 Since the Euler-Lagrange equation is the solution to the problem of
t2
stationarity of a path integral,
S weLsee
dt that
t1

is stationary for the path followed by the particle.

 This integral has a special name in Physics—it is called the action
integral, and when it is a minimum it is called the principle of least
action, although that is a misnomer in the sense that this could be
a maximum, or even an inflection point.
 The principle is also called Hamilton’s Principle:
The actual path that a particle follows between two points 1 and
2 in a given time interval, t1 to t2, is such that the action integral
t2
S  L dt
t1

is stationary when taken along the actual path.

October 21, 2010
 Generalized Coordinates
 We have now seen that the following three statements are
completely equivalent:
 A particle’s path is determined by Newton’s second law F = ma.
 The path is determined by the three Lagrangian equations (at least in
Cartesian coordinates).
 The path is determined by Hamilton’s Principle.
 Again, the great advantage is that we can prove that Lagrange’s
equations hold for any coordinate system such that, for any value r
= (x, y, z) there is a unique set of generalized coordinates (q1, q2, q3)
and vice versa. L  12 mr 2  U (r )
 Using these, we can write the Lagrangian in
terms of generalized L coordinates
L [q1 , q2 , q3 ,as
q 1 , q 2 , q 3 ].

 L dweLthenhave
In terms of these, L d L L d L
 ,  , and  .
q1 dt q 1 q2 dt q 2 q3 dt q 3

 These are true in any coordinate system, so long as the coordinates

are measured in an inertial frame.
October 21, 2010
 Example 7.1
 For a single particle in two dimensions, in Cartesian coordinates,
under some arbitrary potential energy U(x, y), the Lagrangian is
L  12 m( x 2  y 2 )  U ( x, y ).
 In this case, there are two Lagrange equations
L d L L d L
 ,  .
x dt x y dt y
 The left side of each equation is just
L U L U
 Fx ,  Fy .
x x y y
 The right side of each equation, in turn, is just
d L d d L d
 mx mx,  my my.
dt x dt dt y dt
 Equating these, we have Newton’s second law
Fx mx, Fy my or F ma.

October 21, 2010

 Generalized Force and
 Momentum
Notice that the left side term in Cartesian coordinates gives the
force L U L U
 Fx ,  Fy .
x x y y
 When this is expressed in terms of generalized coordinates (which,
for example, could be angular coordinates) this term is not
necessarily a force component, but it plays a role very like a force,
and indeed is called
L the generalized force.
(ith component of generalized force).
qi

 Likewise, the right side in terms of generalized coordinates plays

the role of a momentum,
L and is called the generalized
momentum. (ith component of generalized momentum).
q i L d L
 ,
qi dt q i
 Thus, the Lagrange equation can be read as

generalized force = rate of change of generalizedOctober

momentum.
21, 2010
 Example 7.2
 Let’s repeat example 7.1, but in polar coordinates. In this case, the
potential energy is U(r, ), and you should be able to write down the
kinetic energy directly without much thought as
T  12 m(r 2  r 2 2 ).
so the Lagrangian is just
L  12 m(r 2  r 2 2 )  U (r ,  ).
 In this case, there are two Lagrange equations
L d L L d L angular momentum mr2.
 ,   .
r dt r  dt 
 Inserting the above Lagrangian into these equations gives
U U d
mr 2 
r
mr,  
 mr 2 2mrr  mr 2.
 dt

 The first of these equations says
centripetal acceleration r2.

Fr m r  r , 2

 The second equation is best interpreted by recalling the gradient in
polar coordinates U 1 U U d torque.
U 
r
rˆ 
r 
2 
φˆ . 

rF 
dt
mr  .
October 21, 2010

N Free Particles
It should be obvious how to extend these ideas to a larger number
of particles. In the case of two particles, at positions r1 and r2, for
example, Newton’s second law says each component of each
F1x  p 1x , F1 y  p 1 y , F1z  p 1z ,
particle obeys the equations
F2 x  p 2 x , F2 y  p 2 y , F2 z  p 2 z .
 In the Lagrangian formalism, we have the equivalent Lagrange
equations L d L L d L L d L
 ,  ,  ,
x1 dt x 1 y1 dt y 1 z1 dt z1
L d L L d L L d L
 ,  ,  .
x2 dt x 2 y2 dt y 2 z 2 dt z 2

 We could likewise
L dwrite
L these
L ind generalized
L L coordinates
d L as
 ,  , ,  .
q1 dt q 1 q2 dt q 2 q6 dt q 6

 An important example we will use repeatedly in Chapter 8 is to

replace r1 and r2 with RCM = (m1r1 + m2r2)/(m1 + m2), and r = r1  r2.
 For N particles, then, there are 3N Lagrange equations.
October 21, 2010
 7.2 Constrained Systems

Example
We have said that the great power of the Lagrange formalism is that
constraint forces disappear from the problem.
 Before getting into the general case, let’s do a familiar problem—
instead of a free particle, let’s consider one that is tied to the
ceiling, i.e. the simple pendulum.
 The pendulum bob moves in both x and y, but it moves under the
constraint 2 2
 x  y .
 A perfectly valid way to proceed might be to eliminate y  the
2
 xy2 . 
variable,
and express everything in terms of x, e.g. replace y with
 However, it is much simpler to express the problem in terms ofh the
natural coordinate .
 The first Ttask is 2towrite2  down the Lagrangian L = T – U in terms of .
2 m  .
2
 12 mv 1
Clearly, for this problem U = mghL d L ). Likewise, the kinetic
  cos
= mgl(1 ,
energy is  dt 
 The relevant Lagrange equation is2  and you tocan
 mg  sin  m   . Equivalent  =basically
I
write down the solution
October 21, 2010