AN INTRODUCTION TO LAGRANGE EQUATIONS

Professor J. Kim Vandiver

1.0 INTRODUCTION

This paper is intended as a minimal introduction to the application of Lagrange equations

to the task of finding the equations of motion of a system of rigid bodies. A much more
thorough and rigorous treatment is given in the text “Fundamentals of Applied
Dynamics” by Prof. James H. Williams, Jr., published in 1996 by John Wiley and Sons.
In particular the treatment given here to the evaluation of non-conservative external
forces is simple and pragmatic. It will suffice for most of the applications at the level of
an undergraduate course in engineering dynamics. Deeper understanding requires more
work and the reader is referred to the Williams text. In writing this summary I have
borrowed liberally from the Williams text and also from a previous subject handout,
prepared by Prof. Sanjay Sarma of the Mechanical Engineering Department of MIT.

2.0 ENERGY METHODS IN STATICS

Principle of virtual work: Energy methods were in use to solve statics problems before
discovering them for dynamics. One example is the principal of virtual work. If a
structure such as the one shown in the diagram below is in static equilibrium then from
the principle of virtual work we can deduce the static equilibrium equation by computing
the virtual work done during an infinitesimal deflection. In this case the uniform bar,
shown in Figure 1, rotates about the hinge at A.

It is acted upon by its weight, Mg, applied at its center of gravity at B, a spring force, Fs,
applied at its point of attachment at C, and by the workless constraint reaction force, Ry,
applied at A. Assume the static equilibrium position is horizontal and assume a small
virtual angular displacement, δθ . The principal of virtual work tells us that the work
done by the all of the forces in the course of the virtual displacement is zero. Putting this
into equation form yields directly the relationship between the applied external
gravitational force and the internal spring force as shown below.

1
Figure 1. Hinged bar with a small virtual displacement

δ W= [R y • 0 − Mgδ y1 + Fsδ y2 ]= 0
L L
δ y1
Noting that= δθ and= δ y 2 Lδθ , then [−Mg + Fs= L]δθ 0
2 2
Since δϕ ≠ 0, then the part inside the brackets must equal zero.
L (2.1)
[−Mg + Fs L] = 0, the condition for static equilibrium.
2
Solving for the unknown spring force yields
Mg
Fspring =
2

An important feature to note is that Ry, the reaction force at the hinge does not appear in
the final expression, because it does no work during the virtual displacement. This is a
benefit that carries over to dynamical systems when the concept of virtual work is
extended to application of Lagrange equations to problems in dynamics. The equations
of motion can be found without having to solve for workless constraint forces.

Principle of minimum potential energy: Another related concept came into use for the
analysis of elastic structures for which it was possible to evaluate the total potential
energy of the system, including strain energy stored in structural elements and strain
energy due to gravitational loads. This can be illustrated by considering a simple weight
hanging on a spring, as shown in Figure 2. This is a system in which only one
coordinate, x is needed to describe the vertical deflection of the system.

2
Figure 2. Mass spring system.

As shown in the diagram, the origin is taken to be at the position where the spring force is
zero. The deflection, x, is positive downwards. The total potential energy of the system
for a positive deflection of the mass may be expressed as:
1
V= − ∫ F dx = Kx 2 − Mgx (2.2)
2
The principle of minimum potential energy states that “Of all the states which satisfy the
boundary conditions(constraints), the correct state is that which makes the total potential
energy minimum.” Mathematically this can be expressed as:

∂V
=kx − Mg =0
∂x (2.3)
⇒ x= Mg / k

Thus, from knowing the total potential energy of the system, and by applying this
principle, one may directly obtain the static equilibrium equation as well as the
deflection, x, as a function of the applied load, Mg.

There is a very useful conceptual link between the principle of virtual work and the
principle of minimum potential energy. If we imagine a small virtual deflection of the
system about its static equilibrium position then we can modify the equations obtained
above to define the variation in the potential energy in terms of the virtual work, which
results from a small virtual displacement, δ x , as shown below. The two expressions on
the left hand side represent the variation in the potential energy whereas the expression
on the right hand side shows the virtual work. Both are equal to zero because this is a
small variation about a static equilibrium position for a system with no non-conservative
forces acting on it.

3
 ∂V
δV =δ x =
(kx − Mg )δ x =
0
∂x (2.4)

∂V
In fact the term δ x is in the form of the potential energy term which appears in the
∂x
Lagrange equations defined in the remainder of this document.

3.0 AN INTRODUCTION TO ENERGY METHODS IN DYNAMICS

Newton published the Principia in 1687, in which he laid down his three laws of motion.
101 years later, in 1788, Lagrange developed a way to derive the equations of motion by
beginning with work and energy rather than force and momentum. One way to think
about the leap from the statics problems described above and a dynamics problem is to
consider what textbooks refer to as the work energy theorem, which may be stated as:

(T2 − T1 ) + (V2 − V1 ) = nc
W1→2 (3.1)

This states that the change in kinetic energy, T, and potential energy, V, as a dynamic
system moves from state 1 to 2 is equal to the non-conservative work that is done by the
external non-conservative forces applied to the system, as it moves from position 1 to 2.
At all times in the movement of this system, the above statement must be true, if the
system is in dynamic equilibrium. Another way of saying this is to say that the system
would satisfy the equations of motion at all times.

Now imagine that at an instant in time, there is an infinitesimal variation in the position
of the system from its natural dynamic equilibrium position. Just as is the case for the
principal of virtual work for systems in static equilibrium, the net variation in the kinetic
and potential energy and the work of the non-conservative forces would be zero. The
variation in energy or work in each term is not zero, but the net sum is zero. Stated in
equation form:

δ T + δ V − δ W nc =
0
Or if one prefers: (3.2)
δ T + δV =
δW nc

The second expression says that the change in the sum of the kinetic and potential
energies of the system must equal the work done by the external non-conservative forces.
Lagrange discovered a way to express this for multiple degree of freedom systems.
Before jumping directly to the equations, it is essential to carefully explain how one
determines the correct number of necessary coordinates and degrees of freedom for any
dynamical system that is to be modeled using the Lagrangian approach. There are even
some dynamical systems for which Lagrange equations may not be used. One needs to
know how to identify them.

4
4.0 GENERALIZED COORDINATES AND DEGREES OF FREEDOM

Generalized coordinates: The generalized coordinates of a mechanical system are the

minimum group of parameters which can completely and unambiguously define the
configuration of that system. They are called generalized because they are not restricted
to being Cartesian coordinates and are not even required to be measured from an inertial
reference frame. However, they are used to express the kinetic and potential energies of
the rigid bodies which make up the dynamical system. The kinetic and potential energy
must be computed with respect to an inertial reference frame. Therefore the generalized
coordinates must be able to express the velocity and displacement of rigid bodies with
respect to an inertial frame of reference.

Holonomic systems: A holonomic system is one in which the number of independent

generalized coordinates required to describe the motion of the system equals the number
of degrees of freedom. We study only holonomic systems in this class. An example of a
non-holonomic system is given in lecture and others may be found in the text
“Fundamentals of Applied Dynamics” by Prof. James H. Williams, Jr. An excerpt from
that text, which describes Lagrangian dynamics in detail is available on the course Stellar
website. You should study that after reading this introduction.

Degrees of freedom: The number of degrees of freedom of a system is the number of

coordinates that can be independently varied in a small displacement. Said another way,
the number of degrees of freedom is the same as the number of independent directions in
which a system can move from any given initial configuration. Consider a system
consisting of N rigid bodies in 2D space. Each rigid body has 3 degrees of freedom: two
translational and one rotational. The N-body system has 3N degrees of freedom. Now
let’s say that there are k kinematic constraints. Then the system has =d 3 N − k degrees
of freedom. For example a single uniform stick confined to a vertical, x-y plane has 3
degrees of freedom, x, y, and rotation, ϑ about the z axis. It has no additional constraints.
If we connect one end of the stick to a pivot, it now has two constraints, one in each of
the x and y directions. It is still free to rotate. Therefore d=3-2=1 dof. The motion of the
pinned stick may be completely specified by one coordinate, the angle of rotation.

Completeness and Independence of generalized coordinates: In choosing which

coordinates to use one must obey two requirements.
1. A set of generalized coordinates is complete, if it is capable of locating all parts of
the system at all times.
2. A set of generalized coordinates is independent if, when all but one of the
generalized coordinates are fixed, there remains a continuous range of values for
that one coordinate.
There you have it. To find the equations of motion of a dynamical system using
Lagrange equations, one must first determine the number of degrees of freedom, ‘d’, and
then choose a set of ‘d’ generalized coordinates, which make up a complete and
independent set.

5
5.0 THE LAGRANGE EQUATIONS

For a j-dof (degree-of-freedom) system with generalized coordinates q j , it is possible to

formulate the Lagrangian L= T − V where T is the kinetic energy and V is the potential
energy. The Lagrangian is a function of generalized coordinates q j and generalized
velocities q j :
L = L ( q1 ,...q j ...qd , q1 ,...q j ...qd ) (4.1)
where d is the number of degrees of freedom. The Lagrange Equations are then:

d  ∂L
 ∂L
 −
 =
Qj (4.2)
dt  ∂q j
 ∂q j

where Q j are the external generalized forces. Since j goes from 1 to d, Lagrange gives us
d equations of motion the same number as the degrees of freedom for the system.

The left hand side of Equation 4.2 is a function of only T and V, the potential energy and
kinetic energy and is straightforward to evaluate once the analyst has found T & V. The
right hand side is where most difficulties arise. The right hand side requires that one take
the real non-conservative forces that one would specify on a free body diagram and re-
express them in terms of the component of the real forces that act in the direction of each
generalized deflection. This is the process of finding the generalized forces.

6.0 FINDING THE GENERALIZED FORCES

Let’s consider first just one external real force, applied at some point on the body. An
example, borrowed from Williams is shown in the figure below. It is a hinged rigid link,
which is acted upon by an external force, F(t) which always points down. This system
has one degree of freedom and therefore the angle, θ , that the link makes with the
vertical has been specified as a complete and independent set of generalized coordinates.
q= j q=
1 θ
δ q1 = δθ , a small virtual variation in θ
For this example,
and Q =1δθ F=δ r the work done by the external force F
during the variation δθ .
The magnitude of the generalized force, Q1 , associated with the virtual deflection of the
system, δθ , may be obtained by finding the virtual work done by the real force F(t) in
the course of the virtual displacement.

In equation form, the work done is the dot product of the applied external force, F(t), and
the virtual deflection in the direction of the generalized coordinate which is going
through the variation δθ . In this case the generalized coordinate is θ , and the deflection
component in the direction of F(t) is l sin(θ )δθ , as shown in part c of the figure below.
The virtual work done during this small variation is given by

6
 δ W=
nc
1δθ
Q= F (t )l sin(θ )δθ
Therefore (4.3)
Q1 = F (t )l sin(θ )
In this example, the actual displacement at the location of and in the direction of the force
F(t) is given by l sin(θ )δθ , as shown in the diagram above.

For many problems the dot product of the external force and the deflection component in
the direction of each generalized displacement may be deduced by common sense as was
done above. Several examples are given in the chapter by Williams and more will be
done in lecture.

In the above example the system had only one degree of freedom and therefore there was
only one virtual displacement of the single coordinate θ . In general there will be many
coordinates and many applied external forces. We need to broaden the approach to
account for this.

At every point of application of a real external force, Fi, to the system work will be done
if that point moves. Some external forces like those produced by gravity or lossless
springs are usually included in the potential energy expression. We will assume so here.
The ones remaining are called non-conservative forces because they may add or extract
energy from the system. The technique used to evaluate these non-conservative
generalized forces is to assume the system undergoes a small virtual deflection for which
we compute the increment or variation in work that results, as expressed in the following

7
equation, which says that the total variation in the work of the non-conservative forces is
given by the sum of the dot products of the external forces with the total virtual deflection
at the location of each force, due to the combined virtual deflections of each coordinate.
N d
nc
=δW ∑= Fi δ ri (δ q ,δ q ...δ d )
1 2
∑ Q jδ q j (4.4)
=i 1 =j 1

The term on the right of equation 4.4 says that the same total variation in the work done
by the non-conservative forces must equal the sum of the generalized forces, Qj,
multiplied the virtual displacement δ q j associated with that generalized force. In this
course we will do problems for which we can separate out the individual terms by
applying geometry and simple vectors. We will in general consider the virtual work done
by considering the variation of one coordinate at a time and from that deduce the
generalized force associated with that motion.

A more rigorous method exists which is described in a brief appendix at the end of this
paper.

In summary, some key points to remember regarding generalized forces are:

• Forces of constraint which do no work can be ignored because they will always be
perpendicular to allowable motions, also known as admissible motions. Examples
include the internal forces in a rigid body, the forces of reaction in friction-less
sliding, and so on.
• Forces which derive from a potential function like gravity or a spring can be
considered in potential energy, V. They too can be ignored when computing
generalized forces.
• Forces which are neither included in the computation of potential energy nor are
workless forces of constraint must be expressed as non-conservative generalized
forces in the Lagrange formulation. Examples include dissipative forces from
dashpots, external excitations. If one mistakenly accounts for a potential force as
a non-conservative force, it will work out just fine in the end, as long as you don’
also account for it in the potential energy expression.

7.0 USING LAGRANGE’S EQUATIONS

Now we can return to the Lagrange equations as given in Equation 4.2.

d  ∂L
 ∂L
  − =
Qj (4.5)
dt  ∂q j
 ∂q j

Mathematicians like the elegance of this expression, but it obscures some of the physics.
Lagrange was proud of the fact that his 1788 treatise did not contain a single diagram.

8
Let’s reconnect it to the physics. Recall that the Lagrangian, L is defined as L=T-V.
Substitution in the above equation yields:

d  ∂ (T − V )  ∂ (T − V ) d  ∂ (T )  d  ∂(V )  ∂ (T ) ∂(V )
 − =  −  − + = Qj a.
dt  ∂q j  ∂q j dt  ∂q j  dt  ∂q j  ∂q j ∂q j

d  ∂(V ) 
Noting that   = 0 for common mechanical systems,
dt  ∂q j 

d  ∂(T )  ∂(T ) ∂ (V )
⇒  − + = Qj b.
dt  ∂q j  ∂q j ∂q j
Multiplying by δ q j ⇒ (4.6)
 d  ∂(T )  ∂(T )  ∂ (V )
   − δ qj + δ qj =
Q jδ q j c.
 dt  ∂q j  ∂q j  ∂q j
This is of the form
δ T + δV =
δ W nc d.

In common mechanical systems, potential energy accounts for gravitational potential

energy or the potential energy stored in springs. The potential energy for such systems is
d  ∂(V ) 
never a function of velocity, and hence the term   vanishes in the above
dt  ∂q j 
∂(T )
equations. For simple mechanical systems the term − is often zero. The form of
∂q j
Lagrange equations expressed above in terms of T and V and not L is therefore easier to
use and requires less wasteful bookkeeping. It may be used for all mechanical problems
of the kind done in this course. When one should be cautious with this simplification is
when there are potential forces involving charged particles and/or electric currents and
d  ∂(V ) 
magnetic fields. Then the assumption that   vanishes is not valid.
dt  ∂q j 

The steps in computing the equations of motion using Lagrange’s method are below.

Start with the LHS of Equation 4.2 or the more direct 4.6b:

1. Identify the generalized coordinates. Make sure that you have just as many as there
are degrees-of-freedom and confirm that the system is holonomic.

2. Compute the kinetic energy T as a function of q j ' s and q j ' s .

9
3. Compute the potential energy V as a function of q j ' s . Clearly designate the forces
which you will call out as potential and forces and those which you will call out as
non-conservative external forces.
d  ∂T  ∂T ∂V
4. Compute separately   ,- and for each generalized coordinate and
dt  ∂q j  ∂q j ∂q j
from the sum of the three pieces you will have the LHS of the equation of motion
associated with each generalized coordinate j, as given in equation 4.6b. The LHS by
itself represents the equations of motion of the system without the action of external
non-conservative forces. It is useful by itself in understanding the system.

Now the RHS of Equation 4.2 and 4.6b:

1. Identify all non-conservative forces. They may be many or few and have
nothing to do with the number of degrees of freedom in the problem.

2. Number them as i = 1, 2,...n . Call the forces F1 , F2 ,...Fn .

3. For each j, compute the virtual work associated with the virtual deflection
of each generalized coordinate. Q jδ q j = ∑ i Fi δ ri (δ q j ) where each δ ri is
a function of but not necessarily equal to δ q j as was true in the first
example which considered the motion of the hinged link. The RHS of this
expression will be of the form [?]δ q j , where the term in [ ] will involve
the real forces, Fi , and geometry. This term [ ] = Qj the generalized force
you are seeking to put into the RHS of equation 4.6b or equation 4.2.

4. For more difficult problems employ the explicit computation prescribed

N  ∂r 
δ q j ∑ Fi •  i  δ q j which is described in the
by the equation: Q j=
 ∂q 
i=1  j
appendix.

5. Now equate the LHS to RHS of equation 4.6b for each j. You will finish
with d equations of motion, one for each degree of freedom.

The final step is to put this to practice doing lots of examples. Examples are to be found
in lecture, in the course problem sets and in the text by Williams.

8.0 ADDVANTAGES AND DISADVANTAGES OF LAGRANGE

Advantages:
1. The Lagrange Approach automatically yields as many equations as there are degrees
of freedom.

10
2. The Lagrange equations naturally use the generalized coordinates of the system. By
contrast, Newton’s Equations are essentially Cartesian. You end up having to convert
everything into Cartesian components of acceleration and Cartesian components of
forces to use Newton’s Equation. Lagrange bypasses that conversion.

3. The Lagrange approach naturally eliminates forces of constraint. You could do the
same with the direct (Newtonian) approach, but your ability to minimize the number
of variables depends very much on your skill; Lagrange takes care of it for you
automatically because the generalized forces only include force components in
directions of admissible motion.

Disadvantages:

1. If you make mistakes in formulation of T, V and Wnc you may never notice. The
physics of forces and free body diagrams is easily overlooked. One should always
evaluate each of the equations obtained by the Lagrange method and question
whether or not it makes physical sense. Do the expected accelerations occur? For
example for systems which rotate are there the expected centripetal and/or coriolis
accelerations.

2. Computation of the generalized forces can be as much or more work in the

Lagrangian method as in the direct method.

APPENDIX: A KINEMATICS APPROACH TO COMPUTING GENERALIZED

FORCES

In order to use the Lagrange method one must find the Generalized forces, Qj. What one
must identify first are the external forces, Fi, and then must find a way to express the
arbitrary virtual displacement δ ri in terms of the variations of the generalized
displacements, δqj .
The generalized force Q j is defined below:

N  ∂ri 
=
Qj ∑ F •  ∂q
i  (4.5)
i =1  j

where Fi is the vector force at point i and ri is the position vector to point i, measured
with respect to an inertial frame of reference. The index j corresponds to each of the
generalized coordinates, which must be evaluated one at a time. The summation over i
accounts for the fact that there may be more than one real force, Fi , which does work
when the virtual displacement δ q j takes place. It is the total of all these contributions
that makes up the generalized force Qj. The total virtual work associated with the virtual
displacement δ q j is the sum of the virtual work done by each of the real forces Fi

11
as a result of the virtual displacement δ q j .
N  ∂ri 
δW
= j
nc
δ qj
Q j= ∑ F •  ∂q
i δq
 j
(4.6)
i =1  j

At the location of Fi a small variation of δ q j results in a deflection at i in the j

direction, δ ri j , given by:
∂ri
δ ri j = δ qj (4.7)
∂q j
δ ri j is never in the direction of a kinematic constraint, because no virtual displacement is
allowed in the direction of a constraint. In general, for a d-degree-of-freedom system
with generalized coordinates q1 ,..., qd , the total admissible motion at a point i with
position vector ri is given by a summation over the contributions from the small virtual
deflection of every generalized coordinate that affects the motion at position i, as
specified below.

d ∂ri j
δ ri = ∑ δ qj (4.8)
j=1 ∂q j

12
Example 1
This example considers a single non-conservative force acting on a one degree of
freedom system.

Consider the generalized forces in this system of two masses. Although there are two
masses, only one variable is needed to completely fix the configuration, which implies
that there is a single degree of freedom. There are two possibilities: θ or x, as defined
by the figure. Now we can address the generalized forces. Only one non-conservative
is present that we need to account for: the dashpot.
If we have chosen to use θ to describe the system, we need to find the amount of
virtual work that the non-conservative force will do for a virtual displacement in
θ. In order to find this, we first need to express the point of the system where the
non-conservative force is acting, O r B in terms of θ: O r B = 2l cos (θ) Î. Intuitively
obtaining δ O r B from a small deflection δθ is not very clear, so we can use equation
4.7 from above:
∂ O rB
δO rB = δθ
∂θ
= (−2l sin (θ)) Îδθ

With this relation, we can then look at the generalized force:

δW nc = F · δr
= −cẋÎ · (−2l sin (θ)) Îδθ
= 2clẋ sin (θ) δθ
= −4cl2 sin2 (θ) θ̇δθ
= Qθ δθ

From this, we see that the generalized force is equal to −4cl2 sin2 (θ) θ̇, where we had
to find ẋ in terms of θ̇.

13
If we had instead chosen to use x to describe our system, the calculation for the
corresponding generalized force would have been more straightforward (at the expense
of more complications elsewhere in finding the EoM). In this case, O r B = xÎ, so we
can see that a virtual displacement δx will immediately create a displacement in the
point where the non-conservative force is acting by the same amount δxÎ. Therefore:

δW nc = F · δr
= −cẋÎ · δxÎ
= = −cẋδx
= Qx δx

So we find that the generalized force associated with the dashpot in x is equal to −cẋ.

Example 2
In this example, we consider a single non-conservative force acting on a more compli­
cated two degree of freedom problem.

This problem has two degrees of freedom, which we will take to correspond to the
θ and x generalized coordinates defined in the figure. The non-conservative force
F (t) acts on the center of mass of the ball in the Î direction. In order to find the
generalized forces (corresponding to each generalized coordinate), we can still use the
two approaches discussed earlier.
In the first approach, we consider the virtual deflection of the point of action of the
force and how it is affected by each general coordinate’s independent virtual deflection.

14
Consider a small deflection δθ first. Clearly from the figure, this causes a rotation
about point O, and since the length of the pendulum for a (temporarily) fixed value
ˆ where θˆ
of x is l0 +x, the displacement of the point of contact due to δθ is (lo + x) θδθ,
is a unit vector perpendicular to the long axis of the pendulum. If we now hold fixed
the angle θ and allow a small displacement δx, then the point of contact displaces by
δxr̂, where r̂ is parallel to the long axis of the pendulum. Combining these results,
we find:

δ O r C = δxˆ ˆ
r + (lo + x) θδθ
= δx sin (θ) Î − cos (θ) Ĵ + (lo + x) δθ cos (θ) Î + sin (θ) Ĵ

F · δ O r C = F Î · δx sin (θ) Î − cos (θ) Ĵ + (lo + x) δθ cos (θ) Î + sin (θ) Ĵ

= F sin (θ) δx + F (lo + x) cos (θ) δθ
= Qx δx + Qθ δθ

So we find that the general force associated with the x general coordinate is F sin (θ)
and the one associated with the θ general coordinate is F (lo + x) cos (θ). Note that
the correct units (Newtons for the linear coordinate and Newton-meters for the an­
gular coordinate) fell out directly from solving for the generalized forces.
In the second approach, we can find the generalized forces more directly, but first
need an expression for the position where the non-conservative force acts.

O
r C = (lo + x) r̂ = (lo + x) sin (θ) Î − cos (θ) Ĵ

From equation 4.6 above,w we can find that the generalized force for the generalized
coordinate x is given by N ∂r i
i=1 F i · ∂x , where N is the number of non-conservative
forces (one in this case). A similar expression finds the generalized force for the
generalized coordinate θ. Evaluating these, we find:

∂ O rC
Qx = F ·
∂x
= F Î · sin (θ) Î − cos (θ) Ĵ
= F sin (θ)
∂ O rC
Qθ = F·
∂θ
= F Î · (lo + x) cos (θ) Î + sin (θ) Ĵ
= F (lo + x) cos (θ)

These results match the earlier results, as expected. However, this approach required
writing out an explicit expression for the point of application for each non-conservative

15
force in terms of all the generalized forces. This means that while there are multiple
valid ways of expressing the position of the application of the force, only some of
these expressions will be useful in this approach. For example, writing the position
vector in terms of the r̂ unit vector only does not explicitly depend on the angle θ
since that information is contained in the direction of r̂ unit vector.

16
MIT OpenCourseWare
http://ocw.mit.edu

2.003SC / 1.053J Engineering Dynamics

Fall 2011

For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.