1

PHYS 705: Classical Mechanics

Derivation of Lagrange Equations
from D’Alembert’s Principle
 2

D’Alembert’s Principle

Following a similar argument for the virtual displacement to be consistent

with constraints, i.e,  f   r  0 (no virtual work for f )
i
i i i

We can write down,   F  p    r  0 when we include dynamics,

i
i
(a)
i i

This is the D’Alembert’s Principle.

Again, since the coordinates (and the virtual variations) are not necessary
independent. This does not implies, Fi  (a)
 p i   0 .

We now need to look into changing variables to a set of independent

generalized coordinates so that we can write  ?
j
j
  q j  0 and set
the independent coefficients ? j  0 in the sum to zero.
 3

Derivation of Lagrange Equations

Break  Fi
i
(a)
 p i    ri  0 into two pieces:

1. F
i
i
(a)
  ri (1)

Assume that we have a set of n=3N-K independent generalized coordinates

q j and the coordinate transformation,
r is the position vector in
ri  ri  q1 , q2 , , qn , t 
Cartesian Coordinates
From chain rule, we have

ri
 ri    qj (note:
ri
 t  0 since it is a virtual disp)
j q j t

(Index convention: i goes over # particles and j over generalized coords)

 4

Derivation of Lagrange Equations

This links the variations in ri to qj, substituting it into expression (1), we have,

 ( a ) ri   ( a ) ri

F i
(a)
  ri    Fi 
 q j
 q j      Fi 
  qj
q j 
i i j    i
j 

Defining ri
Q j   Fi( a )  as the “generalized forces”
i q j

We can then write,

 i   ri   Q j  q j
F
i
(a)

j
(1')

(Note: Qj needs not have the dimensions of force but Q j  q j must have
dimensions of work.)
 5

Derivation of Lagrange Equations

i :
Now, we look at the second piece involving p

2.  p   r
i
i i (2) (don’t forget the “-” sign in the original Eq)

  mi
ri   ri (mass is assumed to be constant)
i

 ri 
  mi
ri     qj 
 q 
i  j j 
 ri 
   mi ri    q j (2a )
i

j  q j 
 6

Derivation of Lagrange Equations

Let, go backward a bit. Consider the following time derivative:

d ri  d  ri  ri Note: This is the term


 i i
m r    mi ri     m 
r
i i 
dt  q j  dt  q j  q j from previous page

Rearranging, the term (from the previous page) can be written as,

ri d ri  d  ri  dri


mi ri    mi vi    mi v i    (2b) where v i 
q j dt  q j  dt  q j  dt

Now, consider the blue and red terms in detail,

 7

Derivation of Lagrange Equations

ri
blue term:
q j
Since we have ri  ri  q1 , q2 , , qn , t  , applying chain rule, we have

dri r r
vi    i qk  i
dt k qk t
Taking the partial of above expression with respect to q j, we have

v i ri
 (note: ri does not depend on q )
q j q j j

red term: d  ri    dri  v i

    dt   q (switching derivative order)
dt  q j  q j   j
Is it ok? Check …
 8

explicit check ri  ri  q1 , q2 , , qn , t 

d  ri    ri    ri 

LHS:       qk   
dt  q j  k qk  q j  t  q j 

  dri    ri ri 

  
q  
q j  dt  q j  k qk
RHS: k
t 
  ri    ri   qk 
   qk   t   note

 0 
k q j qk q q
 j    
j

  ri    ri 
   qk   
k qk  q j   t 
 jq

Check! The two terms are the same.

 9

Derivation of Lagrange Equations

ri d r  d  ri 
Putting these two terms back into Eq. (2b): mi
ri    mi v i  i   m v   
q j dt 
i i
q j  dt  q j 

ri d v i  v i

mi ri    mi vi    mi v i 
q j dt  q j  q j

With this, we finally have the following for expression (2):

 ri 
i i i 
p   r  
 i i
j 
m r 
q j
  q j
i 
d  v i  v i 
    mi v i    mi v i   qj (2c)

j  dt  q j q j 
i
 
(reminder: i sums over # particles and j sums over generalized coords)
 10

Derivation of Lagrange Equations

We are almost there but not quite done yet. Consider taking the qj derivative of
the Kinetic Energy,

T   1 2   1 

   2 i i  q   2 i i i 
m v  m v  v
q j q j  i  j  i 
1  v i   vi 
  mi  v i      vi 
2 i  q j   q j


v i
  mi v i 
i q j
Similarly, we can do the same manipulations on T wrt to q j ,
T   1 2 v i
  i i  i i i q
q j q j  i 2
m v  m v 
j
 11

Derivation of Lagrange Equations

Substituting these two expressions into Eq. (2c), we have:

d  v i  v i 
i p i   ri  j  dt  i mi vi  q   i mi vi  q   q j
  j  j 

 d  T  T 
      qj

j  dt  q  j  q j 
 
Finally, reconstructing the two terms in the D’Alembert’s Principle, we have:
 
 i F (a )
 
p i    ri  0 
 i 

  d  T  T 
j Q j   dt  q     q j  0
   j  q j 
 12

Derivation of Euler-Lagrange Equations

Now, since all the  q j are assumed to be independent variations, the
individual bracketed terms in the sum must vanish independently,

d  T  T
    Qj (3)
dt  q j  q j

There are 3N-K of these differential equations for 3N-K qj and the solution of these
equations gives the equations of motion in terms of the generalized coords
without explicitly needing to know the constraint forces.

Also, note the advantage of this equation as a set of scalar equations (with T)
instead of the original 2nd law which is a vector equation in terms of forces.
 13

E-L Equation for Conservative Forces

Case 1: Fi( a ) derivable from a scalar potential

Fi( a )  iU  r1 , r2 , , rN , t  (note: U not depend on velocities)

ri ri
Qj   F i
(a)
   iU 
i q j i q j
  
    i  ˆ  ˆ 
j ˆk  U    x ˆi  y ˆj  z kˆ  
   i i i 
i   xi yi zi  q j 
 U xi U yi U zi 
     

i  xi q j yi q j zi q j 
U
Qj  
q j
 14

E-L Equation for Conservative Forces

Putting this expression into the RHS of Eq. (3), we have,

d  T  T U
    Qj  
dt  q j  q j q j

d  T   T  U 
   0
dt  q j  q j

Notice that since U does not depends on the generalized velocity q j , we

are free to subtract U from T in the first term,

d   T  U    T  U 
   0
dt  q j  q j
 15

E-L Equation for Conservative Forces

We now define the Lagrangian function L = T – U and the desired
Euler-Lagrange’s Equation is:

d  L  L
   0
dt  q j  q j

Note: there is no unique choice of L which gives a particular set of

equations of motion. Given G(q, t) being a differentiable function of
the generalized coordinate, then
dG
L '  q, q , t   L  q, q , t  
dt
is a different Lagrangian but will result in the same EOM.
 16

E-L Equation for Velocity Dependent Potentials

Case 2: U is velocity-dependent, i.e., U (q j , q j , t )
U d  U 
In this case, we redefine the generalized force as, Q j     
q j dt  q j 
d  T  T
Now, substitute this Q j into     Q j , we then have,

dt  q j  q j

d  T  T U d  U 
     
 q 
dt  q j  q j q j dt  j 
 17

E-L Equation for Velocity Dependent Potentials

Combing terms using L = T – U, we again have the same Lagrange’s Equation,

d   T  U    T  U 
   0
dt  q j  q j

d  L  L
   0
dt  q j  q j

This is the case that applies to EM forces on moving charges q with velocity v,

U  q   A  v  where  is the scalar potential

1 2
And, L mv  q   A  v  and A is the vector potential
2
 18

E-L Equation for General Forces

Case 3 (General): Applied forces CANNOT be derived from a potential

One can still write down the Lagrange’s Equation in general as,

d  L  L
    Qj
dt  q j  q j
Here,

- L contains the potential from conservative forces as before and

- Qj represents the forces not arising from a conservative potential
 19

E-L Equation for Dissipative Forces

Example (dissipative friction):

F f    k x vx ,  k y v y ,  k z v z 

For this case, one can define the Rayleigh’s dissipation function:

1
 
2 i

k xi vxi 2  k yi v yi 2  k zi vzi 2 
Then, the friction force for the ith particle can be written as,

    
F f ,i   , ,    v ,i 
 vx v yi vzi
 i 
 20

E-L Equation for Dissipative Forces

Plugging this into the component of the generalized force for the
force of friction, we can get,
ri 
Q j   F f ,i    
i q j q j

ri vi
To see this, plug in our earlier relation :  , we have
q j q j

v i
Q j   F f ,i 
i q j

v i 
   v ,i  
i q j q j
 21

E-L Equation for Dissipative Forces

Then, the Lagrange’ Equation for the case with dissipation becomes,

d  L  L 
     Qj
dt  q j  q j 
q j

- Both scalar function L and  must be specified to get EOM.

- L will contain the potential derivable from all conservative forces as
previously.
 22

Simple Applications of the Lagrangian Formulation

A particle moving under an applied force F in Cartesian Coordinates :

In 3D, r = (x, y, z) and there will be three diff eqs for the EOMs.

1
The Lagrangian is given by, L  T  m  x 2  y 2  z 2 
2

Then, the x-equation is given by,

d  L  L (Note: Qx   Fi 
ri
 Fx )
    Fx x
dt  x  x i
This gives,
similarly for y & z
d
 mx   0  Fx m
rF
dt
 23

Simple Applications of the Lagrangian Formulation

Let redo the calculation in Cylindrical Coords with the same applied force F:

The coordinates are: r = (r, q, z) z

1
From before, we have T  m  x  y  z 
2 2 2

2 r y
Tranformation:  x, y , z    r ,q , z 
r
x  r cos q  x   r sin q q  r cos q q

y  r sin q   y  r cos q q  r sin q x

zz  z  z

Expressing T in Cylindrical Coords:

m 2 2 2
T  (r q sin q  2rrq sin q cos q  r 2 cos 2 q
2
 r 2q 2 cos 2 q  2rrq cos q sin q  r 2 sin 2 q  z 2 )
 24

Simple Applications of the Lagrangian Formulation

Combining and canceling like-color terms, we have

m 2 2 2
T (r  r q  z 2 ) (*)
2
It is constructive to consider the following alternative way to get to this expression,

Let try to express the speed in T in cylindrical coordinates,

z Start with the position vector r,

zzˆ r  r rˆ  z zˆ
r y Taking the time derivative,

dr d rˆ Note: the directional

rrˆ v  r  rrˆ  z zˆ vectors rˆ change in
dt dt time as the particle
x
moves
 25

Simple Applications of the Lagrangian Formulation

To examine on how these directional vectors changes, consider the following
infinitesimal change,

r (t ') Notice that,

y
drˆ  ˆ
qθ
r (t ) d rˆ  dq θˆ  dt
 
d θˆ   dq rˆ  d θˆ
 qrˆ
drˆ
dθˆ θ̂ rˆ '
dt
θˆ ' r̂
dq t '  t  dt x

dr drˆ
Back to the v vector,  r  rrˆ  z zˆ  rqθˆ  rrˆ  z zˆ
v
dt dt
   m m
v  rq  r 2  z 2 and T  v 2  ( r 2  r 2q 2  z 2 )
2
2
So,
2 2
 26

Simple Applications of the Lagrangian Formulation

Now, let calculate the generalized force in cylindrical coordinates,

r:
r r
Qr  F  Since, r  r rˆ  z zˆ , we have  rˆ
r r

So, Qr  F  rˆ  Fr

q:
r r rˆ q θˆ
Qq  F  r r  r θˆ (recall previous page)
q q q q

 
So, Qq  F  rθˆ  rFq (this looks like torque)
z:
r
Qz  F   F  zˆ  Fz
z
 27

Simple Applications of the Lagrangian Formulation

m 2 2 2
The EOM is then given by: Recall, T  ( r  r q  z 2 )
2
d  T  T
r:    Fr  mr  mq 2 r  Fr
dt  r  r

q : d  T   T  rFq 
d
 
mr 2q  rFq (this is
dL
 N)
dt  q  q dt dt

mr 2q  m2rrq  rFq  

m rq  2rq  Fq

d  T  T d
z:    Fz   mz   Fz  mz  Fz
dt  z  z dt
 28

Simple Applications of the Lagrangian Formulation

Putting the components of F together,

F  Fr rˆ  Fq θˆ  Fz zˆ

F  m    
r  q 2 r rˆ  m rq  2rq θˆ  mzzˆ (*)
Is that the same F  mr that we have gotten previously in Cartesian Coords?

Check: v  r  rqθˆ  rrˆ  z zˆ


r
dv d  ˆ
dt dt
=  
rq θ  rrˆ  z zˆ


rr
dt
 
d ˆ  ˆ drˆ
q θ  rq θ  r  
dt
rrˆ  
z zˆ

  dθˆ ˆ   drˆ

 r  q  q θ   rq θ  r  
ˆ rrˆ  
z zˆ
 dt  dt
 29

Simple Applications of the Lagrangian Formulation

Using the directional vectors relations that we had earlier,

d θˆ drˆ  ˆ
 qrˆ , qθ
dt dt

     
r  r q qrˆ  qθˆ  rqθˆ  r qθˆ  
 rrˆ  
z zˆ
Collecting all terms in the same direction,

    
r  rq 2 rˆ  rq 2rq θˆ  
r   z zˆ

So the EOM in (*) on the previous page is indeed F  mr , i.e.,

   
r  q 2 r rˆ  rq  2rq θˆ  
F  m   z zˆ   m
r