PHYSICS (Part 1)

Department of Physics
I.I.T. Madras

Ayan Mukhopadhyay
Contents

1 What is classical mechanics? 1

1.1 The theoretical framework called Newtonian mechanics . . . . . . . 1
1.2 Basic mathematical vocabulary of classical mechanics . . . . . . . . 2
1.2.1 The trajectory . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2.2 Function of many variables and the partial derivative . . . . 3
1.2.3 More about the gradient . . . . . . . . . . . . . . . . . . . . . 11
1.2.4 The line integral . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.3 Principles of mechanics and the laws of motion . . . . . . . . . . . . 20
1.3.1 Determinism, simplicity and the cluster decomposition principle 20
1.3.2 There is no special place in space for the laws of motion . . . 24
1.3.3 There are no special directions in space for the laws of motion 27
1.3.4 There is no special moment of time for the laws of motion . . 33
1.3.5 Galilean invariance and the inertial frames . . . . . . . . . . . 37

2 Motion in one dimension 39

2.1 How to obtain the phase portrait of all possible motions . . . . . . . 39
2.1.1 Types of equilibria . . . . . . . . . . . . . . . . . . . . . . . . 39
2.1.2 Detour: Equilibria in two and higher dimensions . . . . . . . 42
2.1.3 How energy determines motion in one dimension . . . . . . . 43
2.2 Phase space and phase trajectories . . . . . . . . . . . . . . . . . . . 44
2.3 Some properties of phase trajectories . . . . . . . . . . . . . . . . . . 45
2.4 Motion in one dimension . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.5 States of equilibrium in one-dimensional motion . . . . . . . . . . . . 48
2.6 Equilibrium in two- and three-dimensional motion . . . . . . . . . . 49

i
Chapter 1

What is classical mechanics?

1.1 The theoretical framework called Newtonian

mechanics
One of the basic goals of physics is to understand the fundamental laws. These
fundamental laws, which are expected to be simple and elegant, should explain all
or a wide range of phenomena around us quantitatively. Experiments should be
designed and observations should be made using refined instruments so that we
can verify the predictions of a theory based on these laws, and set bounds to the
degree of precision to which our fundamental laws hold, or show that they are only
approximate and fail beyond a certain regime. Experiments based on more precise
observations have overthrown many otherwise successful laws like the Newtonian
law of gravitation.
However, at a deeper level, we seek for a theoretical framework, even before
looking for laws. Newtonian mechanics gives such a theoretical framework through
which we can deduce the laws of motion of all kinds of matter at a wide range of
scales, from particles of dust to rigid and deformable objects, planets, comets and
even gases and fluids. The framework just requires the description of forces in each
specific instance. For the solar system, we need to know how the force of gravitation
works between two bodies. On the other hand, in the case of molecules of a gas, we
will need to know the intermolecular forces.
Today we know that the theoretical framework of Newtonian mechanics is only
approximate, and should be replaced by Einsteinian (relativistic) mechanics espe-
cially when the velocities of the objects are close to the speed of light, or by quantum
mechanics especially when the measurement probes very short distance scales and
time scales, or by the framework of quantum field theory especially when all of these,
large velocities and short length/time scales are relevant. We also know that our
deeper theoretical framework of quantum field theory allows us to state that there
are four fundamental forces, namely gravity, electromagnetic force, and weak and
strong nuclear forces, and all other forces such as the elastic force, inter-molecular
force and friction are consequences of these fundamental forces. We also have very
good reason to believe that our current theoretical framework is incomplete. Nev-
ertheless, we do have a more unified understanding of phenomena today than what
Newton could have had, as we can in principle also derive how di↵erent types of
forces arise from the fundamental ones, and even how the fundamental particles
gain their masses.
In this course, we will study Newtonian mechanics, and how it can be used
to understand phenomena such as the motion of planets and the flow of fluids,
and also their utility in constructing stable machines. However, we will also go

1
2 Physics I

much deeper and understand that Newtonian mechanics actually is a consequence

of simple and elegant principles which are much more fundamental in the sense that
these remain either unchanged, or are replaced by something better when we go to
the Einsteinian framework. As for example, we will see that Newton’s third law
should be understood as a consequence of the principle that there is no distinguished
point in space, or in other words, the laws of motion take the same form irrespective
of where we choose the origin of our coordinates that label the location of particles.
This principle is called the translation invariance of laws of motion and remains
true in relativistic and quantum mechanics, although Newton’s third law need not
hold, as for instance in the case of the Lorentz force.
The deeper understanding of Newtonian framework is not only required to see
why we need more fundamental framework like relativistic mechanics, but also is
the key for technological innovation. The simple and elegant principles behind
Newtonian framework tell us a lot already about the motion of particles, planets,
satellites, rockets and machines without actually solving the equations of motion
explicitly, and helps us understand and push the limits of possibilities. This is not
only limited to just understanding motion. Many principles that we will learn here
also apply, as for instance, to even how machine learning works!
Objectives of this course: Our objectives are essentially as follows.

1. Understand how Newtonian mechanics is structured by deeper principles of

symmetry,

2. Understand and visualize possibilities and forms of motion of particles, as for

instance why the motion of planets are periodic, why small displacement from
stable equilibrium can be mapped to a harmonic oscillator, how can chaos
arise, etc., and

3. Understand how to apply Newtonian mechanics to continuous media like flu-

ids.

All of these will require us to develop the mathematical language through which we
can describe and compute motion. We will have to do this along this course.

1.2 Basic mathematical vocabulary of classical me-

chanics
1.2.1 The trajectory
In the language of mathematics, the motion of a particle, is a trajectory. The idea
of a Cartesian coordinate system is already familiar. We can label any point in
space with a set of three values (x, y, z) with the origin being (0, 0, 0). We can thus
define the coordinate vector
~r = x î + y ĵ + z k̂ (1.1)

to label a point in space.

The location of a particle in space can thus be specified by a coordinate vector
~r at any given time, based on the values of x, y and z coordinates labelling its
position. However, as time changes, the particle moves. Therefore, the coordinates
labelling its position, should be functions of time, i.e. to describe the motion of the
particle we need to know these three functions, namely x(t), y(t) and z(t), and thus

~r(t) = x(t) î + y(t) ĵ + z(t) k̂. (1.2)

Ayan Mukhopadhyay 3

The triplet of functions, which is a time-dependent vector, describes a trajectory,

which is mathematically, a map from the real line (in which time t takes its values)
to three dimensional space R3 where the particle is moving.
Then it is straightforward to define the velocity and acceleration vectors, which
are

~v (t) = ẋ(t) î + ẏ(t) ĵ + ż(t) k̂, (1.3)

~a(t) = ẍ(t) î + ÿ(t) ĵ + z̈(t) k̂, (1.4)

respectively, with the dot denoting di↵erentiation w.r.t. time. We will also denote
the velocity as ~r˙ (t) and the acceleration as ~r¨(t).
Let us take three examples of trajectories. The first example is

~r(t) = cos(!t) î + sin(!t) ĵ + z0 k̂ (1.5)

with ! and z0 constants. Here, obviously,

x(t) = cos(!t), y(t) = sin(!t), z(t) = z0 . (1.6)

The second example is

t2 t20 2tt0
~r(t) = 2 î + 2 ĵ + z0 k̂ (1.7)
2
t + t0 t + t20
with t0 and z0 constants. Here, obviously,
t2 t20 2tt0
x(t) = , y(t) = , z(t) = z0 . (1.8)
t2 + t20 t2 + t20
Note both these examples correspond to moving on the same path, namely

x2 + y 2 = 1, z = z0 , (1.9)

i.e. a unit circle on the plane z = z0 . However, the trajectories corresponding to the
two motions are di↵erent. At a given moment of time, the velocity and acceleration
vectors are di↵erent.
The first trajectory (1.5) is periodic, since the particle returns to the same point
in space after fixed intervals of time. What is the time period? In case of the second
trajectory (1.7), the particle manages to go only once around the unit circle as time
runs from infinite past to infinite future. Check it out for yourself. Is the motion
clockwise or anti-clockwise?
A third example is

~r(t) = cos(!t) î + sin(!t) ĵ + v0 t k̂. (1.10)

with ! and v0 constants. This corresponds to moving on a right handed spiral.

Can you write down an expression for a trajectory corresponding to motion on a
left-handed spiral?

1.2.2 Function of many variables and the partial derivative

We will see that Newtonian mechanics of a fixed number of particles can be specified
by a single function called the potential. It is a function of many variables, which
we will denote as V . In the case of a single particle, this will be V (x, y, z) or more
compactly V (~r). An example is:
q̃ q̃
V (~r) = p = , (1.11)
x2 + y 2 + z 2 |~r|
4 Physics I

with q̃ constant. This function maps each coordinate vector ~r denoting points in R3
space to a point on the real line R (in the above case to the half line R+ ) in which
it takes a value. Note, in the case of trajectory, the map was opposite, namely from
R to R3 .
When we have many particles, say two we will need a function of six variables,
i.e. a potential function V (~r1 , ~r2 ) which assigns a value on the real line for each
choice of the position of the first particle ~r1 = x1 î + y1 ĵ + z1 k̂ and that of the
second particle ~r2 = x2 î + y2 ĵ + z2 k̂. An example is:
K K
V (~r1 , ~r2 ) = p = , (1.12)
(x1 x2 )2 + (y1 y2 )2 + (z1 z2 ) 2 |~r1 r~2 |

with K constant. Mathematically, this is a map from R6 to R. It should then be

obvious that in case of N particles, the potential will be a function V (~r1 , ~r2 , · · · , ~rN )
which is a map from R3N to R.
You can recognize that (1.12) is the Coulomb potential with K = q1 q2 /(4⇡✏0 ) or
the gravitational potential. We can have a situation where the electric permeability
of the medium is a function of time. In that case the potential is also a function of
time, i.e.

K(t) K(t)
V (~r1 , ~r2 , t) = p = . (1.13)
(x1 x2 )2 + (y1 y2 )2 + (z1 z2 ) 2 |~r1 r~2 |

This is an example of explicit time dependence. If the two particles start moving,
then obviously the potential becomes time-dependent through ~r1 (t) and ~r2 (t) on
the trajectories of the particles. This time-dependence is however implicit.
There are many ways to visualize a potential function. Let’s consider the simple
case of
1 1
V (~r) = K x2 + y 2 + z 2 = Kr2 , (1.14)
2 2
with K constant, and r = |~r|. To visualize we can plot the function with y and z
held fixed. This reduces the function of three variables to that of a single variable.
An example of such a plot is Fig. 1.1 in which V is plotted for K = 1 as a function
of x with y and z set to zero. Alternatively, we can plot V as a function of x and y
for each fixed z. An example is the 3D plot in Fig. 1.2 where z = 0. Yet another
alternative is the contour plot in Fig. 1.3.
An important concept is that of equipotential surfaces, where the potential V
takes constant values, i.e. surfaces where

V (x, y, z) = V0 , (1.15)

with V0 arbitrary constants in the range of V . The contour plot, such as Fig. 1.3,
gives a concrete way to visualize sections of equipotential surfaces by freezing one
coordinate to a fixed value, in which case these are actually equipotential lines on
a plane where the other two coordinates can take arbitrary values.
Let us consider motion in 2D and hence a two-dimensional potential. An inter-
esting example is ⇣y⌘
V = K arctan , (1.16)
x
with K a constant. This function is ill-defined at the origin. To visualize this, we
can simply see the contour plot in Fig. 1.4. Since the equipotential lines meet at
the origin, the function is multi-valued at the origin – the value will depend on the
path with which we approach it. We will explore some interesting consequences of
such a potential. For practical purposes, we will consider that the space on which
the particle moves in the two dimensional space R2 with the origin removed. The
Ayan Mukhopadhyay 5

12

10

x
-4 -2 2 4

Figure 1.1: The plot of the potential V given by (1.14) as a function of x for
y = z = 0. Here K = 1.

Figure 1.2: A 3D plot of the potential V given by (1.14) as a function of x and y

for z = 0. Here K = 1.
6 Physics I

22.5
2
20.0

17.5

15.0

0 12.5

10.0

7.5

5.0
-2
2.5

-4

-4 -2 0 2 4

Figure 1.3: A contour plot of the potential V given by (1.14) as a function of x and
y for z = 0. Here K = 1. Note the sections of equipotential surfaces are visible.

origin is a singular point where the law of motion is ill-defined. Furthermore we

should also restrict the range of V , such as from ⇡/2 to ⇡/2 as shown in Fig. 1.4,
which leads to discontinuity along the y-axis (see the figure). We will study this
case in more detail later.
We would now introduce the concept of a partial derivative. The partial deriva-
tive measures the rate of change of a function with respect to one of the variables,
with other variables held fixed. Thus for arbitrary point (x0 , y0 , z0 ) we can define
the partial derivative w.r.t. x as follows

@V (x0 , y0 , z0 ) V (x0 + x, y0 , z0 ) V (x0 , y0 , z0 )

= V,x (x0 , y0 , z0 ) = lim . (1.17)
@x x!0 x

provided the limit exists. Similarly, we can define the partial y-derivative and the
partial z-derivative, etc. As for example, let

V (x, y) = x2 y 2 . (1.18)

Then you can readily find from the definitions that

V,x = 2xy 2 , V,y = 2x2 y. (1.19)

Consider for simplicity a function V (x, y) of two variables, and also V,x and
V,y which are continuous for all points in a given region (say a disc) D. With the
functions V,x and V,y in D, we can then define the second order partial derivatives
Ayan Mukhopadhyay 7

1.5
2
1.0

0.5

0 0

-0.5

-1.0
-2
-1.5

-4

-4 -2 0 2 4

Figure 1.4: A contour plot of the potential V given by (1.16) as a function of x and
y. Here K = 1. Note the equipotential lines pinch o↵ at the origin.

as follows:

@2V V,x (x0 + x, y0 ) V,x (x0 , y0 )

V,xx (x0 , y0 ) = = lim ,
@x2 x!0 x
@2V V,x (x0 , y0 + y) V,x (x0 , y0 )
V,xy (x0 , y0 , ) = = lim ,
@y@x y!0 y
@2V V,y (x0 + x, y0 ) V,y (x0 , y0 )
V,yx (x0 , y0 , ) = = lim ,
@x@y x!0 x
@2V V,y (x0 , y0 + y) V,y (x0 , y0 )
V,yy (x0 , y0 , ) = = lim , (1.20)
@y 2 y!0 y

if these limits exist. For V given by (1.18), we obtain from (1.19) and the above
definitions that

V,xx = 2y 2 , V,xy = V,yx = 4xy, V,yy = 2x2 . (1.21)

We note that V,xy = V,yx . In fact, this is generic due to Schwarz’s theorem (also
called Clairaut’s theorem) which states that if V,xy and V,yx are continuous on a
disc D containing the point (a, b), then V,xy (a, b) = V,yx (b, a). This extends to the
case when V is a function of more than two variables. In this course, we will not
need to consider more than two variables, however for a solid foundation it is very
necessary to study real analysis of function of many variables.
Furthermore, the Taylor series expansion extends to function of many variables.
Consider a point (x0 , y0 , z0 ) at which V is continuous, and its first and second deriva-
tives exist. Then for a sufficiently small neighborhood, i.e. a ball with a sufficiently
small radius ✏ centered around (x0 , y0 , z0 ), the function V can be approximated as
8 Physics I

follows:

V (x, y, z) = V (x0 , y0 , z0 ) + (x x0 )V,x (x0 , y0 , z0 ) + (y y0 )V,y (x0 , y0 , z0 )

+(z z0 )V,z (x0 , y0 , z0 )
1 1
+ (x x0 )2 V,xx (x0 , y0 , z0 ) + (y y0 )2 V,yy (x0 , y0 , z0 )
2 2
1 2
+ (z z0 ) V,zz (x0 , y0 , z0 )
2
+(x x0 )(y y0 )V,xy (x0 , y0 , z0 ) + (y y0 )(z z0 )V,yz (x0 , y0 , z0 )
+(z z0 )(x x0 )V,zx (x0 , y0 , z0 )
+O(✏3 ). (1.22)

The proof of this theorem and its extension with higher order partial derivatives
(see below) should be learnt in a maths course on real analysis.1
It is instructive to verify the above theorem explicitly for the function V given
by (1.18). (In this example, there is no z dependence, so V,z , V,xz , V,yz and V,zz
vanish.) Let us change variables from (x, y) to (✏1 , ✏2 ) as follows

x = x0 + ✏ 1 , y = y0 + ✏ 2 . (1.23)

Then, for (1.18), we get via binomial expansions that

V (x, y) = (x0 + ✏1 )2 (y0 + ✏2 )2

= x20 y02 + 2x0 y02 ✏1 + 2x20 y0 ✏2
+y02 ✏21 + 4x0 y0 ✏1 ✏2 + x20 ✏22
+O(✏3 ) (1.24)

with O(✏3 ) denoting terms which are cubic or higher order in ✏1 and ✏2 . Substituting
(1.19) and (1.21) above we find that

V (x, y) = x20 y02 + ✏1 V,x (x0 , y0 ) + ✏2 V,y (x0 , y0 )

+V,xx (x0 , y0 )✏21 + 4V,xy (x0 , y0 )✏1 ✏2 + V,yy (x0 , y0 )✏22
+O(✏3 ) (1.25)

Finally, substituting ✏1 = x x0 nd ✏2 = y y0 above we do verify the Taylor

expansion (1.22).
Let us try to make our notations more compact. To do this, we introduce the
gradient operator :
~ = î @ + ĵ @ + k̂ @ .
r (1.26)
@x @y @z
This operator acts on a function and converts it to a vector field, i.e.
~ (x, y, z) := rV
V (x, y, z) ! U ~ (x, y, z)

with

Ux (x, y, z) = V,x (x, y, z), Uy (x, y, z) = V,y (x, y, z), Uz (x, y, z) = V,z (x, y, z).

A vector field is essentially an assignment of a vector to every point in space, i.e.

a map from R3 to R3 (or more generally a map from Rn to Rn in n-dimensional
1 In rigorous mathematical analysis, we have two infinitesimals actually. One is which sets
the error tolerance limit of an approximation, and the other is ✏ which tells us where and how
should be truncate a series. For the Taylor series truncated at an order, we will need to show
that for a given there is a sufficiently small ball ✏ where it provides the desired approximation.
Understanding the proof is thus also essential for practical applications.
Ayan Mukhopadhyay 9

space). A physical example of a vector field is a force field. Note that a vector
field need not be a gradient of a function, although the gradient of a function is a
~ (x, y, z) is indeed a gradient of a
vector field. It is easy to see when a vector field U
function, then we should have

Ux,y = Uy,x , Uy,z = Uz,y , Uz,x = Ux,z (1.27)

at all points in space. This simply follows from V,xy = V,yx , etc. Later we will
express the above in a more compact way, through another derivative operator
called the curl, and say that the above condition defines an irrotational vector field.
In fact, the converse statement is also true. If a vector field U ~ is irrotational
at a point P , i.e. satisfies (1.27) at P , then in a sufficiently small ball around P ,
there exists a function V such that U ~ = rV~ .
It is easy to see that (1.22) can be written in a more compact way as follows:

V (~r) = V (~r0 ) + (~r ~ (~r0 ) + O(✏2 ),

~r0 ) · rV (1.28)

where second order derivatives have been omitted, ~r0 is the coordinate vector cor-
responding to the chosen point (x0 , y0 , z0 ), and · denotes the vector dot product,
namely A~·B ~ = Ax Bx + Ay By + Az Bz . Also ✏ is the (very small) size of the ball
centered at r~0 where this approximation is good.
To write the second order terms, we will use the so-called index notation. We
will denote Ai for the components of a vector A, ~ with i = 1, 2, 3 and V1 = Vx ,
V2 = Vy and V3 = Vz . In this notation,
3
X
~·B
A ~ = Ai B i (1.29)
i=1

In n-dimensional space,
n
X
~·B
A ~ = Ai B i . (1.30)
i=1

~ can be written as @i , with

Furthermore, the components of the gradient operator r
@ @ @
@1 = , @2 = , @3 = . (1.31)
@x @y @z

so that the components of rV ~ are @i V . Clearly we can write the second derivatives
similarly in the form V,ij . With these definitions, the Taylor series expansion (1.22)
take the form

V (~r) = V (~r0 ) + (~r ~ (~r0 )

~r0 ) · rV
3 3
1 XX
+ (r r0 )i (r r0 )j V,ij (~r0 ) + O(✏3 ). (1.32)
2 i=1 j=1

Note that v,xy automatically appears twice above. We can further simplify the
notation by using what is called the Einstein summation convention introduced by
Einstein in which it is assumed that any index i repeated twice is summed over.
It is important that no index is repeated more than twice. With this convection
A~ ·B
~ = Ai Bi where we have omitted the summation symbol which is implicit. With
these notations, we can write
1
V (~r) = V (~r0 ) + (r r0 )i @i V (~r0 ) + + (r r0 )i (r r0 )j V,ij (~r0 )
2
+O(✏3 ). (1.33)
10 Physics I

It turns out that the full Taylor expansion of a function in its domain of analyticity
(just the domain of validity which is typically a ball of finite (non-vanishing) radius
centered at ~r0 ) is given by
1
X 1
V (~r) = V (~r0 ) + (r r0 )i1 · · · (r r0 )ik V,i1 ···ik (~r0 ). (1.34)
k!
k=1

Quite remarkably compact and this is true for n-dimensional space also (with each
index taking n possible integer values from 1 to n). Furthermore, the order of
derivatives in mixed partial derivatives (at any order) does not matter, i.e. V,xxy =
V,xyx = Vyxx , etc. (You can try to show this yourself as a consequence of Schwarz’s
theorem.) Exercise: Verify (1.34) for V (x, y) = cos x sin y to all orders in the
Taylor expansion about the origin. To do this you will only need the
Taylor’s theorem for a function of a single variable.
Finally, partial derivatives also satisfy the chain rule, i.e.
@ @f (x, y) @K(f (x, y), g(x, y))
K(f (x, y), g(x, y)) =
@x @x @f (x, y)
@g(x, y) @K(f (x, y), g(x, y))
+ ,
@x @g(x, y)
@ @f (x, y) @K(f (x, y), g(x, y))
K(f (x, y), g(x, y)) =
@y @y @f (x, y)
@g(x, y) @K(f (x, y), g(x, y))
+ , (1.35)
@y @g(x, y)
which generalizes to functions of more variables. It is easy to verify this for the
function (1.18). As for instance, by choosing,
K(u, v) = uv, f (x, y) = x2 , g(x, y) = y 2 , V (x, y) = K(f (x, y), g(x, y))
and applying the chain rule, we can readily reproduce (1.19). We can also do this
by choosing
K(u, v) = u2 , f (x, y) = xy, g(x, y) = x2 y2 , V (x, y) = K(f (x, y), g(x, y)).
and applying the chain rule.

Supplemental Material: Here we give you a heuristic understanding of why

partial derivatives commute, i.e. V,xy = V,yx in the domain they are both continuous
functions. To see this let us consider a function of a single variable f (x). The first
derivative is at x = x0 is de
f (x0 + ✏) f (x0 )
f 0 (x0 ) = lim , (1.36)
✏!0 ✏
and the second derivative is
f 0 (x0 + ✏) f 0 (x0 )
f 00 (x0 ) = lim
✏!0 ✏
f (x0 + 2✏) 2f (x0 + ✏) + f (x0 )
= lim , (1.37)
✏!0 ✏2
In the second line above, we have used the definition of the first derivative.
Similarly, in the context of the mixed partial derivative,
V,x (x0 , y0 + ✏) V,x (x0 , y0 )
V,xy (x0 , y0 ) = lim (1.38)
✏!0 ✏
V (x0 + ✏, y0 + ✏) V (x0 , y0 + ✏) V (x0 + ✏, y0 ) + V (x0 , y0 )
= lim ,
✏!0 ✏2
Ayan Mukhopadhyay 11

where in the second line we have used the definition of V,x , namely

V (x0 + ✏, y0 + ✏) V (x0 , y0 + ✏)
V,x (x0 , y0 + ✏) = lim ,
✏!0 ✏
and
V (x0 + ✏, y0 ) V (x0 , y0 )
V,x (x0 , y0 ) = lim .
✏!0 ✏
Similarly, we can find that
V,y (x0 + ✏, y0 ) V,y (x0 , y0 )
V,yx (x0 , y0 ) = lim (1.39)
✏!0 ✏
V (x0 + ✏, y0 + ✏) V (x0 + ✏, y0 ) V (x0 , y0 + ✏) + V (x0 , y0 )
= lim .
✏!0 ✏2
Comparing (1.38) and (1.39), we readily see that V,xy = V,yx .

1.2.3 More about the gradient

The most fundamental vector field in Newtonian mechanics is the force field which
in almost all forces of fundamental origin is

F~ (~r) = ~ (~r).
rV (1.40)

This implies that the force field is directed along the direction of steepest descent of
the function V . To see this, we need to first connect the gradient to the convective
derivative and the directional derivative.
Let us consider a trajectory of a particle: ~r(t) = x(t) î + y(t) ĵ + z(t) k̂. Consider
the potential function V (~r, t) which is explicitly time-dependent. We can then
ask, what is the rate of change of the function V (~r, t) that the particle measures
along its trajectory. The rate of change has two origins, one from the explicit
time-dependence of V , and another from the implicit dependence on time through
x(t), y(t) and z(t). In fact, along the trajectory, V is function of time alone, if we
account for both these implicit and explicit time dependences. So we can readily
apply the chain rule for a function of a single variable and obtain
dV (~r(t), t) @V (~r(t), t) dx(t) @V (~r(t), t) dy(t) @V (~r(t), t)
= + +
dt @t dt @x dt @y
dz(t) @V (~r(t), t)
+ . (1.41)
dt @z
We readily note that the above can be written compactly in the form
✓ ◆ ✓ ◆
dV (~r(t), t) @ d~r(t) ~ @ ~ V (~r(t), t).
= + · r V (~r(t), t) = + ~v (t) · r (1.42)
dt @t dt @t
Thus, we can define the convective derivative,
@ ~
D⌘ + ~v (t) · r (1.43)
@t
which measures the rate of change of a function along a trajectory ~r(t).
The notion of the directional derivative follows from the Taylor expansion (1.28).
Consider a function V (~r, t) at a point P with coordinate vector ~r0 at a fixed moment
of time. Then it follows from (1.28) that for a spatial displacement r along a
direction n̂, we should obtain
V (~r0 + r n̂, t) V (~r0 , t) ~ (~r0 , t).
lim = (n̂ · r)V (1.44)
r!0 r
12 Physics I

To see this, simply use ~r = ~r0 + r n̂ in (1.28). Thus

~
n̂ · r

is the directional derivative measuring the rate of change of a function due to spatial
displacement of the point of measurement along a unit direction n̂.
~ be oriented towards the unit direction ˆl, so that rV
Let rV ~ = |rV ~ |ˆl. The
directional derivative is then
(n̂ · ˆl)|rV
~ |.

Obviously the maximum magnitude of this derivative is |rV ~ | and it occurs when
ˆ ˆ ˆ
n̂ = l or when n̂ = l. Therefore, l, i.e. the direction along rV ~ , is the direction
along which the function V increases (ascends) most rapidly, and ˆl, i.e. the
direction along rV ~ , is the direction in which it decreases (descends) most rapidly.
This readily implies that the force field F is oriented along the direction of steepest
descent of V . An interesting physical consequence is that if a particle has zero
velocity at a given point, then the force directs its motion along the direction of
steepest descent of V .
There is yet another geometric interpretation of the gradient which has impor-
tant physical consequences. Consider an equipotential surface (or an equipotential
line in case of two dimensions) for a potential V (~r) which has not explicit time
dependence. Clearly, along any trajectory which lies on the surface, the value of
V (~r) cannot change. The tangential directions to a surface at a given point P is
simply the set of directions which can be tangential to arbitrary curves lying on
the surface and passing through P . It follows then that for any direction t̂ which is
tangential to the equipotential surface of V , the directional derivative of V should
~ ) = 0. Since t̂ is an arbitrary tangential direction, rV
vanish, i.e. (t̂ · rV ~ should be
2
normal to the equipotential surface.
Let us therefore consider the example of the two-dimensional potential V given
by (1.18) for which
F~ = rV ~ = 2xy 2 î 2x2 y ĵ. (1.45)
A possible trajectory that follows the equipotential line and passes through P whose
coordinate vector is ~r0 = x0 î + y0 ĵ will follow the curve x2 y 2 = x20 y02 . An example
is
x2 y 2
x(t) = t, y(t) = 0 0 . (1.46)
t
The tangent to this curve is t̂ = v̂(t) = ~v (t)/|~v (t)|, i.e.

1 dx(t) dy(t)
t̂ = r⇣ ⌘2 ⇣ ⌘2 dt î + dt ĵ
dx(t)
dt + dy(t)
dt
x 0 y0
1 2
= q î q t ĵ
x2 y 2 x20 y02
1 + 0t4 0 1+ t4
2
x x 0 y0
= p î p (1.47)
x4 + x20 y02 x4 + x20 y02

In the last step above we have substituted x(t) = t from (1.46). Note that the
parametrization t has disappeared. Of course, if we specify a curve in space (of
arbitrary dimensions), its unit tangential direction is unique at a given point up to
2 Normally the normal to a surface given by V (~ r) = constant is defined by the direction along
which V increases (steepest). Therefore, rV ~ is indeed the normal to the surface (and not the
~ is anti-parallel to the normal.
anti-parallel direction), and F
Ayan Mukhopadhyay 13

300 300
4 250 250
200 200
150 150
100 100
50 50

-2

50 50
100 100
150 150
200 200
-4 250 250
300 300

-4 -2 0 2 4

Figure 1.5: The econtour plot showing the quipotential lines of the two-dimensional
potential (1.18).

an overall sign, and therefore should not depend on the explicit parametrization.
So the tangential direction is not dependent on the trajectory (how we are moving
on the path) but by the path (here the line along which the potential is a constant).
Finally, we can verify from (1.45) and (1.47) that
t̂ · F~ = 0 (1.48)
at the chosen point P where x = x0 and y = y0 . Since this point has been chosen
arbitrarily, we have actually verified for this example that the force field F~ is indeed
anti-parallel to the normal to the equipotential lines. In Fig. 1.5 we have plotted
the equipotential lines of the two-dimensional potential (1.18). In Fig. 1.6, we
have plotted the force field F~ given by (1.45). For a sample of points, we have
attached a vector which points along F~ and the vectors (which have fixed length in
the plot) are colored according to their magnitudes (the lighter the color, the higher
the magnitude). You can readily see by comparing Fig. 1.5 and 1.6 that indeed
the force field point towards the directions of steepest descent of V . In a more
colloquial way, Fig. 1.6 is a plot of the lines of force corresponding to potential
given by (1.18).
You may have already noticed a peculiarity. The lines of force become confused
in their directions near the origin. Evidently, from (1.45), the lines of force point
towards ⇣ ⌘
ˆl = p 1 y î x ĵ .
x2 + y 2
Clearly, this direction is ill-defined at the origin. Nevertheless, F~ is continuous at
the origin, meaning both its components are continuous functions, since F~ vanishes
at the origin.
However, the origin is a singular point for the lines of force corresponding to
the two-dimensional potential given by (1.16). The corresponding force field (for
14 Physics I

-2

-4

-4 -2 0 2 4

Figure 1.6: The force field F~ corresponding to the two-dimensional potential (1.18).
Comparing with Fig. 1.5, we note that the lines of force are directed towrds the
direction of steepest descent of the potential (1.18).

-2

-4

-4 -2 0 2 4

Figure 1.7: The force field F~ corresponding to the two-dimensional potential (1.18).
Ayan Mukhopadhyay 15

K = 1) is
y x
F~ = î ĵ, (1.49)
x2 + y 2 x2 + y 2
and is plotted in Fig. 1.7. The lines of force circulates around the origin, where the
force is ill-defined.

1.2.4 The line integral

The line integral appears in physics ubiquitously. The first appearance is in the
form of work.
A line integral of a function on a curve is defined as follows. Consider a function
f (~r) and a curve or a trajectory ~r( ) parametrized by a real parameter (this could
be the time) which lies in the interval 2 [ 1 , 2 ], If ~r( 1 ) = ~r1 and ~r( 2 ) = ~r2 , this
curve (trajectory) starts and ends at ~r1 and ~r2 , respectively. This trajectory needs
to be piecewise smooth, i.e. dx( )/d , dy( )/d and dz( )/d should be piecewise
continuous functions (and of course x( ), y( ) and z( ) should be continuous for
the entire interval of from 1 to 2 ). The line integral of the function f (~r) along
this curve (trajectory) is then
Z 2

I= d f (~r( )). (1.50)

1

The line integral of a vector field F~ (~r) is defined as the line integral of the
function
d~r( )
P ( ) = F~ (~r( )) · , (1.51)
d
which is defined on the trajectory ~r( ). Explicitly,
Z Z
2 2
d~r( )
I= d P( ) = d F~ (~r( )) · (1.52)
1 1
d

Denoting
d~r( )
d~r = d ,
d
we can also write Z
I= F~ (~r) · d~r (1.53)
C

where it is understood that the infinitesimal vector d~r denotes infinitesimal dis-
placement along the trajectory C specified by the function ~r( ). The symbol C is
often used to represent a trajectory abstractly.
For the integration of a time-dependent force field on a trajectory ~r(t) we need
to also take into account an explicit time-dependence. The work done by the force
field F~ (~r(t), t) on the trajectory ~r(t) between the time interval from t1 to t2 is
Z t2 Z t2
d~r(t)
W = dt F~ (~r(t), t) · = dt F~ (~r(t), t) · ~v (t). (1.54)
t1 dt t1

Clearly, Z
W = F~ (~r) · d~r. (1.55)
C

As an example, let the force field to be

F~ (~r) = (x y)î + (x + y)ĵ. (1.56)

16 Physics I

Figure 1.8: The trajectory (1.57) is shown as C1 and the trajectory (1.58) is shown
as C2 above. Note C2 is a union of two pieces, namely one which is on the x-axis
and the other which is on the y-axis.

This force field cannot be written in the form F~ = rV

~ because the condition (1.27)
is not satisfied. Particularly, Fx,y = 1 and Fy,x = 1, and therefore Fx,y 6= Fy,x .
Let’s consider two trajectories which start at the origin O at t = 0 and ends on
P corresponding to (1, 1, 0) at t = 1. The first trajectory is given by

x(t) = t, y(t) = t, z(t) = 0. (1.57)

This trajectory corresponds to motion at constant velocity on the diagonal from O

to P on the xy-plane. The second trajectory is given by
(
2t for 0  t  12 ,
x(t) = ,
1 for 12 < t  1
(
0 for 0  t  12 ,
y(t) = ,
2t 1 for 12 < t  1
z(t) = 0. (1.58)

Here the motion from O to P is first at uniform velocity along the x-axis from t = 0
to t = 1/2, and then at uniform speed along y-axis from t = 1/2 to t = 1. The two
trajectories (1.57) and (1.58) are shown in Fig. 1.8.
On the first trajectory,
~v (t) = î + ĵ,
and therefore

F~ (~r(t)) · ~v (t) = (x(t) y(t)) ⇥ 1 + (x(t) + y(t)) ⇥ 1 = (t t) ⇥ 1 + (t + t) ⇥ 1 = 2t.

Therefore, on this trajectory, the work done from t = 0 to t = 1 is

Z 1
W = dt 2t = 1. (1.59)
0
Ayan Mukhopadhyay 17

On the second trajectory,

(
2î for 0  t  12 ,
~v (t) = ,
2ĵ for 12 < t < 1

and therefore
(
(x(t) y(t)) ⇥ 2 = (2t 0) ⇥ 2 = 4t for 0  t  21 ,
F~ (~r(t)) · ~v (t) =
(x(t) + y(t)) ⇥ 2 = (1 + 2t 1) ⇥ 2 = 4t for 12 < t < 1.

Therefore, on this trajectory, the work done from t = 0 to t = 1 is

Z 1/2 Z 1
W = dt 4t + dt 4t = 2t2 |10 = 2. (1.60)
0 1/2

Therefore twice more work is performed along the second trajectory.

Let us now consider an irrotational force field F~ which is of the form F~ (~r) =
~
rV (~r), and which has no explicit dependence of time. In this case, the work done
by such a force field on a trajectory ~r(t) corresponding to motion from ~r(t1 ) = ~r1
at time t = t1 to ~r(t2 ) = ~r2 at time t = t2 is
Z t2 Z t2
W = ~ (~r(t)) · d~r(t) =
dt rV dt
dV (~r(t))
t1 dt t1 dt
= (V (~r(t2 )) V (~r(t1 ))) = V (~r1 ) + V (~r2 ). (1.61)

Above, we have used explicitly the chain rule in (1.42) and that V has no explicit
time-dependence, i.e. @V /@t = 0. Therefore, if a force field is both irrotational
and time independent, then the work done by such a force field on a trajectory is
simply the minus of the di↵erence in the potential energies at the two end points.
For reasons to be clear later, a force field F~ (~r) which is both irrotational and time
independent (so that there exists a V (~r) with no time dependence such that F~ (~r) =
~ (~r)) is called a conservative force field.
rV
The work done by a conservative vector field on any trajectory thus depends
only on the end points, and is independent of the path which connects the two end
points. As for instance, let us take
⇣ ⌘
F~ (~r) = (x y) î ĵ . (1.62)

Note Fx,y = Fy,x = 1 and it is time-independent. Therefore, there must exist a

V (~r) such that F~ (~r) = rV
~ (~r). To find this V , we can do the indefinite integral,
Z Z
1
V (~r) = dx V,x + f (y, z) = dx Fx + f (y, z) = (x y)2 + f (y, z)
2
The above ensures that V,x = Fx = (x y) while f (y, z) appears as an integration
constant (in the sense that f is independent of x). Furthermore, V,y = Fy = x y
and V,z = Fz = 0 implies that

@y f (y, z) = @z f (y, z) = 0,

i.e. f is independent of y and z. Therefore,

f (y, z) = C0 ,

with C0 constant, i.e.

1
V (~r) = (x y)2 + C0 .
2
18 Physics I

Thus we explicitly determine V for the force field (1.62) uniquely up to an additive
constant. One can explicitly check that the work done by the force field (1.62) on
the trajectories (1.57) and (1.58), which are W1 and W2 respectively are the same,
and furthermore
W1 = W2 = V (~r1 ) + V (~r2 ) = 0 (1.63)
since both trajectories begin and end at the origin and (1, 1, 0), respectively.
A closed curve or a trajectory is one in which the end point is identical to the
starting point, i.e. ~r(t1 ) = ~r1 = ~r(t2 ) = ~r2 . Clearly, the work done by a conservative
force field on such a closed trajectory should vanish because

W = V (~r1 ) + V (~r2 ) = V (~r1 ) + V (~r1 ) = 0,

as ~r1 = ~r2 . This is often written as

I
W = F~ (~r) · d~r = 0 (1.64)
C
H
if F (~r) is a conservative force field. The sign implies integration on a closed path,
i.e. the trajectory C is closed.
There is a delicate subtlety here which has profound physical consequences. Let
us consider the force field (1.49). This is, in fact, a conservative force field since
Fx,y = Fy,x . In fact, the corresponding potential function is (1.16) with K = 1.
Let’s try to compute the work done by this force field on the two-dimensional
trajectory
x(t) = cos t, y(t) = sin t, 0  t  2⇡ (1.65)
Clearly, this is a closed trajectory starting and ending at the point (1, 0). On this
trajectory, the force field (1.49) takes the form

F~ = sin t î cos t ĵ. (1.66)

Since,
~v (t) = sin t î + cos t ĵ,
on the trajectory
F~ (t) · ~v (t) = 1. (1.67)
Therefore, the work done by the force (1.49) on the closed trajectory (1.58) is
Z 2⇡
W = dt ( 1) = 2⇡. (1.68)
0

This is a rather big surprise! We have just shown that work done by a conservative
force field on a closed trajectory should vanish. Here, F~ is a conservative force field,
however the work done by it on the closed trajectory (1.65) is not zero!
What goes wrong? To see this, we first note that we can use the variables r and
✓ (what we will call later as polar coordinates) which are defined as

x = r cos ✓, y = r sin ✓. (1.69)

Clearly, 0  ✓ < 2⇡. In these coordinates, the potential V corresponding to the

force field (1.49) given by (1.16) with K = 1 takes the form

V = ✓. (1.70)

The reason for choosing the polar coordinate system is that we get a well behaved
potential everywhere except for the origin (where the behavior was bad anyway)
Ayan Mukhopadhyay 19

Figure 1.9: The work done by the force (1.49) on various possible closed trajectories:
(i) the trajectory C1 crosses the positive x-axis from below once, and the work done
on it should be 2⇡ as computed in (1.68), (i) the trajectory C2 has two loops
crossing the positive x-axis from below twice and from above once, so the work
done on it is also 2⇡, (iii) the trajectory C3 does not cross the positive x-axis and
the trajectory C4 crosses the positive x-axis from below and from above once each;
the work done in both cases is zero. Note that C1 and C2 both encircle the origin
once counterclockwise, while C3 and C4 do not encircle the origin.

once we also recognize that ✓ is periodic. Technically, the function V is a map from
R2 to a circle S 1 and not an interval.
The point is that ✓ increases by 2⇡ as we cross the positive x-axis from below,
while it decreases by 2⇡ as we cross the positive x-axis from above. Take any closed
trajectory C which goes around the origin counterclockwise. For such a trajectory,
the value of ✓ increases by 2⇡ as it crosses the x-axis from below only once (as for
C1 in 1.9), or an odd number of times with the number of crossing from below larger
than the number of crossings from above by one (as for C2 in 1.9). Since the work
done is minus the change in the potential energy (= ✓), the work done on such a
closed trajectory is
✓ = 2⇡
exactly as we computed in (1.68) for the trajectory (1.65).
However, if we consider a closed trajectory which does not encircle the origin,
then the work done on this trajectory is zero since it should not cross the x-axis at
all (as for C3 in 1.9), or cross it from below as many times as it does from above
(as for C4 in 1.9). The more general statements, which follows from this line of
reasoning, are as follows
1. The work done by the force (1.49) on a closed trajectory which goes fully
around the origin counterclockwise n number of times and clockwise m number
of times, is 2⇡(n m).
2. The work done by the force (1.49) on any closed trajectory which does not
enclose the origin is zero.
Exercise: Verify the above by choosing appropriate trajectories and do-
ing the line integral with the force field (1.16) explicitly for these trajec-
tories.
20 Physics I

The lesson we learn from the above example is that the work done by a conser-
vative force field with lines of force tangential to a two-dimensional plane vanishes
on a closed trajectory provided the trajectory does not enclose any singular point
where the potential is ill-defined. In the example above, V had a singular point at
the origin as ✓ is not well-defined at the origin. The feature of the above example
that the work done on a closed trajectory does not depend on any detail except
for its topological property of how many times (counterclockwise minus clockwise) it
goes around the origin can be understood in a more general mathematical concept
called the homotopy.3 This is however beyond the present course. However, this is
essential to understand under what circumstances a force field can be locally con-
servative but not globally, and can do non-vanishing work on systems which follow
closed trajectories. This also has profound applications in quantum mechanics. I
hope that you will learn more about this in the future.

1.3 Principles of mechanics and the laws of motion

1.3.1 Determinism, simplicity and the cluster decomposition
principle
The central tenet of mechanics is the principle of determinism. Suppose the universe
has N particles. The principle of determinism states that if we know the positions
and velocities of all these particles at a given moment of time, then we can know
their positions and velocities at any time in the future, and also at any time in the
past. Essentially, the present has a unique future and a unique past.
Let’s see how a mathematical formulation of such a statement can work. Let’s
choose Cartesian coordinates in space with an arbitrary but fixed origin O. Let’s
also label time t such that the present is t = 0. The coordinates of the N particles
can be labelled are ~ri (t) = xi (t) î+yi (t) ĵ +zi (t) k̂. with i = 1, 2, · · · , N labelling the
particles. The velocities of these particles are ~vi (t) = ~r˙i (t) = ẋi (t) î+ ẏi (t) ĵ + żi (t) k̂.
Our task then is then to predict or postdict (or rather infer) ~ri (t) and ~vi (t) for t > 0
and t < 0, respectively given the knowledge of ~ri (0) and ~v0 (0).
For this to work, we will need a set of rules of the following form:

1 ~
~r¨i = Fi (~rj , ~r˙j , t), (1.71)
mi

for i = 1, 2, · · · , N . Note indices are not summed above. (In what follows we
will not invoke the Einstein summation convention unless explicitly mentioned.)
Equivalently,
mi~ai = F~i (~rj , ~r˙j , t), (1.72)

On the right hand side of this equation are the force fields F~i which are functions
of the positions and velocities of all particles only at the present time. These force
fields are essentially a set of 3 functions, namely Fix , Fiy and Fiz for each particle
i. Each of these three functions is a function of 6N + 1 variables (the positions and
velocities of all particles and time). So in total the set of rules involve 3N functions
(for N force fields) each of which are functions of 6N + 1 variables. Quite a mess!
We need to simplify. However, before that, let’s look into the left hand side. We
have parameters mi , which are the masses, and which can in principle be di↵erent
3 Homotopy is a property of a function which cannot change under smooth deformations. As

for instance, on a closed trajectory the function V (✓) becomes a map from a circle S 1 to another
circle S̃ 1 . How many times (counterclockwise - clockwise) S 1 wraps around S̃ 1 is a property of
such a map which cannot change under smooth deformations.
Ayan Mukhopadhyay 21

for each particle. Both the left and right sides of these equations have the same
physical dimensions for each particle.4
Let’s see why determinism works if the laws of motion take the form (1.71). First
and foremost, in order to know the future, we only need to know a finite amount of
information about the present, namely the positions and velocities of all particles
right now, and nothing about the past. Similarly to infer the the past, we only need
to know a finite amount of information about the present, namely the positions
and velocities of all particles right now, and nothing about what will happen in the
future. Secondly, the present has a unique future and a unique past. To see this,
let’s simplify to just one particle. Our law is then just
m~r¨ = F~ (~r, ~r˙, t). (1.73)
We can rewrite this second order di↵erential equation as a pair of first order di↵er-
ential equations:
p~
~r˙ = , p~˙ = F~ (~r, p~, t). (1.74)
m
The first equation is simply a definition of momentum. In any case, we know ~r(0)
and ~r˙ (0) at present time t = 0. The first equation implies that we also know p~(0) at
present time t = 0 from ~r˙ (0). Let’s now consider a moment t = t infinitesimally
in the future. In this case, (1.74) implies that
p~(0)
~r( t) = ~r(0) + t, p~( t) = p~(0) + F~ (~r(0), p~(0), t = 0) t, (1.75)
m
implying that we uniquely know ~r( t) and p~( t) (and thus also ~r˙ ( t)). We can
then iterate to compute ~r(2 t) and p~(2 t) (and thus also ~r˙ (2 t)), and so on.
Suppose we want to predict the position and velocity at time t = t0 , then we will
divide the time interval from present t = 0 to t = t0 into K small intervals t,
i.e. t = t0 /K, and we can use the algorithm above to compute ~r(i t) and ~r˙ (i t)
for i = 1, 2, · · · , K. One can show rigorously that if the force fields are sufficiently
smooth functions of the positions and velocities, then we can take the limit K ! 1
in which we obtain the actual functions ~r(t) and ~r˙ (t) for 0 < t  t0 . A sufficiently
large K usually gives a very good approximate prediction. One can run the same
argument back in the past with t < 0. It is very important to understand this
argument more rigorously, however at present we obtain a heuristic understanding
of how the present leads to unique future and why it has a unique past. This
argument can be readily generalized to N particles.
Determinism is not enough. We want our laws of motion to have underlying
simplicity. So clearly, it is too much to set laws of motion by specifying N force fields.
In reality, we have 1023 particles in just a mole of gas, and we will need to specify 1023
force fields to even state (not even solve) how these particles move. Experiments
suggest that all fundamental forces can be captured by a single function, namely a
potential function
V(N ) (~r1 , ~r2 , · · · , ~rN ),
so that the force fields can be obtained by computing the gradients, i.e. the force
field for the particles are:
F~1 = ~ 1 V(N ) ,
r F~2 = ~ 2 V(N ) ,
r ··· , , F~N = ~ N V(N ) ,
r
with
~ 1 = î @ + ĵ @ + k̂ @ ,
r
@x1 @y1 @z1
4 The masses can also depend on time but in the way we have written these equations, the change
in mass can be absorbed into the definition of the force fields through factors like mi (t)/mi (0)
since mi (t) = (mi (t)/mi (0)) ⇥ mi (0) – in this case mi (0) would be the constant masses.
22 Physics I

where partial derivatives are taken with respect to the coordinates of the first
particle only, etc. Furthermore, the potential function V(N ) has no explicit time-
dependence. As a consequence, we have the first crucial symmetry, namely the laws
of motion take the same form at any moment of time, i.e. they are immutable with
time. We will come back to this a bit later.
It is still a bit too much because V can in principle be a function of 3N variables,
namely the coordinates of the N particles. Every time we add a particle we will
need to figure out how to define the potential function. There is another natural
expectation. Suppose, we have a cluster of particles confined in a region of space
and all other particles are far away. It will be natural if we can compute the motion
of particles of this cluster of particles without knowing anything about the other
particles which are far away. Indeed when we want to compute what is happening
inside the solar system, we do not need to follow all stars in the Milky way or in other
galaxies. Also for predicting tides on Earth, we do not need to know much about
what is happening in Mars. We can call this the cluster decomposition principle,
which states that if all particles fall into two clusters, the first say with particles i
with 1  i  M and the second one with particles j such that M + 1  j  N , then
in the limit when all the particles in the first cluster are very far from the second
cluster (compared to the average distances of particles within each cluster), then
we should have

V(N ) (~r1 , ~r2 , · · · , ~rN ) ⇡ V(M ) (~r1 , ~r2 , · · · , ~rM ) + V(N rM +1 , ~rM +2 , · · ·
M ) (~ , ~rN ).

It then follows that the force fields in the first cluster can be obtained from V(M )
and the force fields in the second one can be obtained from V(N M ) , leading to
a decoupling of the two clusters. The force fields experienced by the particles in
each cluster is independent of the positions and velocities of the particles in the
other cluster when the distance between particles in these two clusters are much
larger than the distances between particles within each cluster. Note this readily
generalizes to multiple clusters simply by further splitting each cluster into two,
and using induction.
As a special case, if the first particle is far away from all other particles (com-
pared to the mutual distances between the other particles), then

V(N ) (~r1 , ~r2 , · · · , ~rN ) ⇡ V(1) (~r1 ) + V(N r2 , ~r3 , · · ·

1) (~ , ~rN ).

The potential V(1) then accounts for the self-force of the particle on itself. We will
later show that the simple principle of spatial translation symmetry forbids self
forcing, so that the one-body potential V(1) should vanish (or be an inconsequential
constant that can be set to zero by using the freedom that V is defined up to a
constant). So if the first particle is far away from all the others (compared to the
mutual distances between the other particles), then

V(N ) (~r1 , ~r2 , · · · , ~rN ) ⇡ V(N r2 , ~r3 , · · ·

1) (~ , ~rN ).

This implies that the first particle is free, meaning that it experiences no force field.
Similarly if both the first and second particles are far from each other, and both
are also far from the rest, then

V(N ) (~r1 , ~r2 , · · · , ~rN ) ⇡ V(N r3 , · · ·

2) (~ , ~rN ).

This implies that the two particles are free – both do not experience any force field.
Continuing this way, we obtain that in the limit where all particles are infinitely far
from each other, we should obtain that

V(N ) (~r1 , ~r2 , · · · , ~rN ) ⇡ 0,

Ayan Mukhopadhyay 23

i.e. all particles are free.

We can achieve all these properties of V(N ) very elegantly if V(N ) is a sum of two-
body terms V(2) which is further determined by a single function of six variables.
Explicitly,
N
X N
X
V(N ) (~r1 , · · · , ~rN ) = V(2)ij (~ri , ~rj ), with
i=1 j=1,j6=i

V(2) (~ri , ~rj ) = Kqi qj v(~ri , ~rj ). (1.76)

Note in the first equation above, the summations run over pairs of distinct particles
(hence we have the condition i 6= j) with V(2)ij denoting the pairwise potential
function. We note in the second equation above that the two body potential is
determined by a single function v of six variables, namely the coordinates of the
pair of particles. The dependence of V(2)ij on the chosen pair of particles, comes
only through the parameters qi and qj which we will call charges. The charge qi of
the i-th particle is an intrinsic property of the particle just like its mass. K is a
constant which should be there on dimensional grounds.
To satisfy the cluster decomposition principle, we further need that

lim v(~ri , ~rj ) = 0. (1.77)

|~ rj |!1
ri ~

One can readily check that this simple requirement on v is enough to ensure the
cluster decomposition principle for an arbitrary number of particles.
Thus we have achieved the following
1. Simplicity: The laws of motion of N particles governed by a fundamental
force can be described by a single function that determines the two-body force,
and the charge and mass (that specify the intrinsic property) of each particle.
2. Cluster Decomposition Property: When particles can be grouped into
clusters such that the distance between particles in two di↵erent clusters is
very large compared to distance between particles within each cluster, the
forces acting on particles in a cluster depends only on the positions of the
particles within that cluster.
All fundamental forces in nature which have a macroscopic classical limit, espe-
cially gravitation and the electric force satisfy these properties. (In fact the cluster
decomposition principle is a fundamental tenet of even quantum field theory!) For
each fundamental force, one needs a corresponding charge. If the charge of the par-
ticle (object) under a fundamental force is zero, e.g. the case of the neutron for the
electric charge, then it does not experience this force (the neutron does not expe-
rience the electric force).5 If a particle participates in multiple fundamental forces,
then it should have a corresponding charge for each such force, and each such force
is specified by a function v(~r1 , ~r2 ).
The gravitational force is special. The charge for the gravitational force is pro-
portional to the mass itself! By choosing K appropriately, we can thus say that
the inertial mass (which appears in the definition of the momentum) and the grav-
itational mass (which gives the strength of the gravitational force) are exactly the
5 The neutron has a magnetic dipole moment and it does experience the magnetic force though.

The magnetic force is also fundamental, but its origin is relativistic, so we are omitting it in our
present discussion. Also charged particles like the electron and muon, in addition to their electric
charges, have intrinsic quantum angular momentum, called the spin, which also implies an intrinsic
magnetic dipole moment as well. We will ignore the spin and the intrinsic magnetic dipole moment
in this present course. To see why fundamental particles should have spin, we really need both
quantum mechanics and relativistic physics, i.e. quantum field theory.
24 Physics I

same for any particle. Experimental tests (to be discussed later in this course) have
confirmed the equality of the inertial and gravitational masses to a very high degree
of precision. You can easily check right away that if the inertial and gravitational
masses are identical, then the acceleration of the particle in a gravitational force
field will be independent of its mass. Einstein realized the significance of this prop-
erty of gravitation and used it to reformulate gravity in a way which is consistent
with relativistic physics, and this is arguably the most beautiful classical theory
ever constructed by any human. This is however outside of the scope of this present
course.
Finally, we can write down the laws of motion corresponding to a fundamental
force. For a system of N particles, these take the form.
N
X
mi~ai = ~ i V(N ) (~r1 , . . . , ~rN ) =
r qi ~ i v(~ri , ~rj ),
qj r i = 1, · · · , N. (1.78)
j=1,j6=i

It is easy to see that the force on i-th particle F~i can be split into a sum of forces
F~ij (with j 6= i) denoting the force on the i-th particle exerted by the j-th particle:
N
X
mi~ai = F~i , F~i = F~ij , F~ij = ~ i v(~ri , ~rj ).
qi qj r (1.79)
j=1,j6=i

As an example, take three particles interacting via the electric force. Then, after
writing K in the standard form 1/(4⇡✏0 ), we get
✓ ◆
1 q1 q2 q2 q3 q3 q1
V(3) = + + . (1.80)
4⇡✏0 |~r1 ~r2 | |~r2 ~r3 | |~r3 ~r1 |

Exercise: Verify that (1.78) implies that

✓ ◆
1 ~r1 ~r2 ~r1 ~r3
m1~a1 = q1 q2 3
+ q1 q3 ,
4⇡✏0 |~r1 ~r2 | |~r1 ~r3 |3
✓ ◆
1 ~r2 ~r3 ~r2 ~r1
m2~a2 = q2 q3 3
+ q1 q2 ,
4⇡✏0 |~r2 ~r3 | |~r2 ~r1 |3
✓ ◆
1 ~r3 ~r1 ~r3 ~r2
m3~a3 = q3 q1 3
+ q3 q2 . (1.81)
4⇡✏0 |~r3 ~r1 | |~r3 ~r2 |3

Although a fundamental force is simple in the sense that we can express it

through a single function determining the two-body force, other forces which re-
sult from fundamental forces need not have this feature. As for instance, there
is evidence that nuclear forces cannot be understood only via two-body interac-
tions of nucleons – we need to account for three-body and four-body interactions
in the potential which cannot be written as a sum of two-body terms. However,
nuclear forces are a result of more fundamental strong nuclear force (also called the
color force) between quarks and gluons which are the elementary constituents of
the nucleons. Similar considerations can arise in molecular clusters also.
In what follows, we will show that principles of symmetry further simplify the
two-body force.

1.3.2 There is no special place in space for the laws of motion

We have shown above that Newton’s second law F~i = m~ai is a consequence of
determinism, that we need to know only the positions and velocities of all particles
only at the present moment to know the unique future and past. We have also shown
Ayan Mukhopadhyay 25

that simplicity and cluster decomposition property tells us that we can describe the
forces in terms of a single function which governs the two-body interactions, and
the intrinsic charges of the particles.
The laws of motion for fundamental forces/interactions are dramatically further
simplified by principles of symmetry. The first such principle is that the fundamental
laws take the same form irrespective of where we choose the origin of our coordinate
system, i.e. there is no special place in space.
Suppose there is only one particle in the Universe. Its motion can be described
in terms of a single function V(1) (~r). We need to specify V(1) (~r) so that we get a law
of motion for this particle. Let O be the origin of one coordinate system, and O0
be the origin of another coordinate system with O0 being at ~r0 = x0 î + y0 ĵ + z0 k̂ in
the coordinate system with origin at O. If ~r and ~r0 be the coordinates of a particle
in the coordinate systems with origins at O and O0 respectively, then clearly

~r0 = ~r ~r0 . (1.82)

If a particle has a trajectory ~r(t) in the first (unprimed) coordinate system, then
the trajectory in the primed coordinate system is

~r0 (t) = ~r(t) ~r0 . (1.83)

It follows that
~v 0 (t) = ~v (t), ~a0 (t) = ~a(t), (1.84)
so that the particle has the same velocity and accelerations in both coordinate
systems. However, if I use the second coordinate system to describe motion of the
0
particle, then I must use a di↵erent function V(1) so that
0
V(1) (~r0 ) = V(1) (~r) = V(1) (~r0 ~r0 ) (1.85)

If we now forget the reference (unprimed) coordinate system, I should use a new
potential function Ṽ (~r) which is related to the earlier potential function as Ṽ (~r) =
V (~r ~r0 ). Let’s see this concretely in a simple example by considering one-dimensional
motion along x-axis for simplicity. Here we have x0 = x x0 and the particle tra-
jectories in the two coordinate systems are related by x0 (t) = x(t) x0 . If in
the unprimed coordinate system, V (x) = (K/2)x2 , then we get mẍ = Kx =
K(x0 x0 ) = mẍ0 . To obtain the same trajectory in the second coordinate sys-
tem, we should then use the potential Ṽ (x) = (K/2)(x x0 )2 (where we have
removed the prime over x).
The problem is that the choice of potential, and hence the form of the force and
thus the law of motion will then depend on where we are choosing our coordinate
system unless
V (~r) = V (~r + ~r0 ) (1.86)
at all ~r for any arbitrary but fixed ~r0 . This is just saying that

V (~r) = c (1.87)

with c constant. In other words, spatial translation invariance of the potential for a
single particle implies that it is just a constant. This implies that the force vanishes.
A particle in free space far away from all other particles experiences no force. It
cannot force itself. This is just Newton’s first law, the law of inertia, which we now
realize to be a consequence of spatial translation symmetry.
It is useful to see this in another way. Using Taylor expansion we can can argue
that if all the components of ~r0 are small, then
~ (~r).
V (~r + ~r0 ) ⇡ V (~r) + ~r0 · rV (1.88)
26 Physics I

Since ~r0 is an arbitrary infinitesimal displacement, (1.86) implies that

~ (~r) = 0,
rV (1.89)

at any value of ~r (as (1.86) is required for all points ~r). So the force vanishes. Since,
F~ = 0, we have for one free particle in space:

d
m~a = (m~v ) = p~˙ = 0. (1.90)
dt
The vanishing of force implies conservation of momentum.
Let’s see how this generalizes to N particles interacting with a potential V(N ) (~r1 , · · · , ~rN ).
Repeating the above arguments for N particles, we can readily see that the laws of
motion will be invariant under spatial translations i↵

V(N ) (~r1 + ~r0 , ~r2 + ~r0 · · · , ~rN + ~r0 ) = V(N ) (~r1 , ~r2 , · · · , ~rN ) (1.91)

for any any displacement ~r0 of the origin. Considering ~r0 infinitesimal, we have

V(N ) (~r1 + ~r0 , ~r2 + ~r0 · · · , ~rN + ~r0 ) ⇡ V(N ) (~r1 , ~r2 , · · · , ~rN )
+ ~r0 · (r~1+r ~2+·+r ~ N )V(N ) (~r1 , ~r2 , · · · , ~rN ) (1.92)

The above is obtained by considering the Taylor expansion of 3N variables about an

arbitrary point in R3N corresponding to the coordinates of the N particles. Since
~r0 is an arbitrary infinitesimal vector, spatial translation invariance requires that

~ 1 +r
(r ~ 2 +·+r
~ N )V(N ) (~r1 , ~r2 , · · · , ~rN ) = 0, i.e. F~1 + F~2 + · · · + F~N = 0. (1.93)

Since
d
F~1 + · · · + F~N = m1 a~1 + · · · + mN a~N = p1 + · · · + p~N ) ,
(~ (1.94)
dt
we obtain that the total momentum of the N particles
N
X
P~ = p~i (1.95)
i=1

should be conserved, i.e.

˙
P~ = 0 (1.96)
Therefore, spatial translation invariance implies that the total momentum of a sys-
tem of particles is conserved.
Note that above we have not used that V(N ) can be further split into two body
terms V(2) (~ri , ~rj ). Translation invariance of V(2) (~ri , ~rj ) implies that

~i+r
(r ~ j )V(2) (~ri , ~rj ) = 0. (1.97)

However, the force on the i-th particle exerted by the j-th particle is

F~ij = ~ i V(2) (~ri , ~rj )

r

.Therefore, translation invariance implies that

F~ij = F~ji . (1.98)

Ayan Mukhopadhyay 27

Thus Newton’s third law is just a consequence of spatial translation symmetry, and
the requirement of simplicity that the N -body potential can be split into two body
terms! The cancellation of the total force (1.93) should occur pairwise as follows
N
X N
X N
X X
F~i = F~ij = (F~ij + F~ji ) = 0 (1.99)
i=1 i=1 j=1,j6=i <ij>

P
with <ij> denoting summing over pairs of distinct particles and not distinguishing
ij from ji.
To conclude our exploration of consequences of spatial translational invariance,
let’s see how this constrains the two-body potential V(2) . To see this let’s go to the
case of one-dimensional motion. Here, (1.97) reduces to
✓ ◆
@ @
+ V(2) (xi , xj ) = 0. (1.100)
@xi @xj

Let’s change variables from xi , xj to xc , xr :

xi + xj xi xj
xc = , xr = (1.101)
2 2
Then by applying chain rule it is easy to see that (1.100) implies that

@
V(2) (xc , xr ) = 0, (1.102)
@xc
i.e.
V(2) (x1 , x2 ) = V(2) (xr ) = V(2) (x1 x2 ). (1.103)
Generalizing this argument to higher dimensions we readily obtain that (1.97) im-
plies that

V(2) (~r1 , ~r2 ) = V(2) (x1 , y1 , z1 , x2 , y2 , z2 )

= V(2) (x1 x2 , y1 y2 , z1 z2 ) = V(2) (~r1 ~r2 ). (1.104)

Another way to see this is simply that ~r1 ~r2 is invariant under the transformations
~r1 ! ~r1 + ~r0 and ~r1 ! ~r2 + ~r0 implied by translation, but ~r1 + ~r2 is not. So spatial
translation invariance implies that V(2) is actually a function of three variables,
namely ~r1 ~r2 , and not six variables (as it should be independent of ~r1 + ~r2 ).
The profound connection between symmetry of physical theories and their con-
servation laws was first made by Emmy Noether (1882-1935), one of the greatest
creative mathematicians in history. Her contribution has significantly changed our
thinking of how we search for the fundamental laws of Universe.

1.3.3 There are no special directions in space for the laws of

motion
Consider two particles interacting via the potential

V (~r1 , r~2 ) = k(x1 x2 ) 2 (1.105)

with k a non-vanishing constant. This potential is invariant under spatial transla-

tions, implying total momentum is conserved. However, it requires us to distinguish
a special direction in space, namely the x-axis. The principle of isotropy states that
the fundamental laws of nature do not prefer any distinguished direction in space.
28 Physics I

Let’s see how this work in two dimensions. Let’s choose our origin at O. Suppose
n̂ is a preferred direction. If n̂ · î = cos ✓, then

n̂ = cos ✓ î + sin ✓ĵ. (1.106)

The direction orthogonal to n̂ is

m̂ = sin ✓ î + cos ✓ĵ. (1.107)

If n̂ is nothing special, then we can orient our Cartesian coordinates x and y along
these new orthogonal directions n̂ and m̂, i.e. use the coordinates x0 and y 0 defined
via

x0 = x cos ✓ + y sin ✓,
0
y = x sin ✓ + y cos ✓, (1.108)

and yet we should find

V(N ) (x1 , y1 , · · · , xN , yN ) = V(N ) (x01 , y10 , · · · , x0N , yN

0
). (1.109)

What does this physically imply? To see this we use the same trick as in the
case of spatial translation invariance, we consider ✓ to be an infinitesimal angle.
Then
cos ✓ ⇡ 1 + O(✓2 ), sin ✓ ⇡ ✓ + O(✓2 ), (1.110)

and therefore,

x0 ⇡ x + y ✓ + O(✓2 ),
0
y ⇡ y x ✓ + O(✓2 ). (1.111)

The implication of (1.109) is then

V(N ) (x1 + ✓y1 , y1 ✓x1 , · · · , xN + ✓yN , yN ✓xN ) =

2
V(N ) (x1 , y1 , · · · , xN , yN ) + O(✓ ). (1.112)

However, applying the chain rule, and Taylor expanding in ✓ we obtain that

V(N ) (x1 + ✓y1 , y1 ✓x1 , · · · , xN + ✓yN , yN ✓xN ) = V(N ) (x1 , y1 , · · · , xN , yN )

X N ✓ ◆
@ @
+✓ xk yk V(N ) (x1 , y1 , · · · , xN , yN )
@yk @xk
k=1
2
+ O(✓ ). (1.113)

Therefore, for (1.112) to hold, we need the term linear in ✓ to vanish. Consequently,

N ✓
X ◆
@ @
xk yk V(N ) x1 , y1 , · · · , xN , yN ) = 0,
@yk @xk
k=1
N
X
i.e. (xk Fky yk Fkx ) = 0, (1.114)
k=1

where Fkx and Fky denote the x-th and y-th components of the total force acting
on the k-th particle.
Ayan Mukhopadhyay 29

We note that in two dimensions, we get6

N
X N
X N
X
(xk Fky yk Fkx ) = ~rk ⇥ F~k = ~rk ⇥ p~˙ k = 0. (1.115)
k=1 k=1 k=1

Furthermore,

N
! N N N
d X X X X
~rk ⇥ p~k = ~vk ⇥ ⇥~
pk + ~rk ⇥ p~˙ k = ~rk ⇥ p~˙ k (1.116)
dt
k=1 k=1 k=1 k=1

where we have used ~vk ⇥ p~k = 0 (which follows from ~vk = p~k /mk ). Therefore, we
obtain from (1.115) that
N
!
d X
~rk ⇥ p~k = 0, (1.117)
dt
k=1

~ defined by
implying that the total angular momentum L

N
X
~ =
L ~rk ⇥ p~k (1.118)
k=1

is conserved. Therefore, in two dimensions, the absence of any distinguished direc-

tion in space, amounts to total angular momentum conservation. Note that here
we have not used the simplicity condition that the V(N ) splits into sum over two
body terms. However, note that we have chosen a specific origin O about isotropy
(equivalence of all directions) hold. So it is important that we consider the angular
momentum about that point.
We also note that the transformation (1.108) is just rotation in the xy-plane
about the z-axis. Therefore, the conservation of L ~ is a consequence of rotational
invariance about the z-axis. More generally, rotational invariance about an axis in
space is equivalent to the statement that there is no distinguished direction in the
plane orthogonal to the axis.
Suppose you do an infinitesimal rotation about n̂ axis by a small angle ✓. Then
the generalization of (1.111) is that

~r0 = ~r ✓ n̂ ⇥ ~r + O(✓2 ). (1.119)

We will prove the above statement later in this course. You can readily check that
when n̂ = ẑ, then we get back (1.111). Then rotational invariance about the axis n̂
implies that
V(N ) (~r10 , · · · ~rN
0
) = V(N ) (~r1 , · · · ~rN ) + O(✓2 ) (1.120)

while Taylor expanding in ✓ and applying the chian rule gives

V(N ) (~r10 , · · · ~rN

0
) = V(N ) (~r1 , · · · ~rN )
N
X
✓ ~ i V(N ) (~r1 , · · · ~rN ) + O(✓2 ).
(n̂ ⇥ ~ri ) · r (1.121)
i=1

6 Note if we think of the two dimensional vectors A ~ and B

~ as three dimensional vectors A~ =
~ = Bx î + Bj ĵ, respectively, then A
Ax î + Aj ĵ and B ~⇥B ~ = (Ax By Ay Bx )ẑ. However, in two
dimensions there is ẑ component (it is out of the plane). The two dimensional cross product A~ ⇥B~
is actually not a vector but a pseudoscalar, a rotationally invariant quantity which switches sign
if instead of right-handed coordinate system, we us a left-handed one. See more below.
30 Physics I

Vanishing of the term linear in ✓ yields

N
X
~ i V(N ) (~r1 , · · · ~rN ) = 0,
(n̂ ⇥ ~ri ) · r
i=1
N
X N
X
i.e. n̂ · ~ i )V(N ) (~r1 , · · · ~rN ) = 0,
(~ri ⇥ r i.e. n̂ · (~ri ⇥ F~i ) = 0 (1.122)
i=1 i=1

In the second line above, we have used the vector identity A ~ · (B

~ ⇥ C)
~ =B
~ · (C
~ ⇥ A).
~
Using the manipulations shown before, we can readily see that
N N
!
X d X
(~ri ⇥ F~i ) = ~˙
(~ri ⇥ p~i ) = L (1.123)
i=1
dt i=1

~ is the total three dimensional momentum. Therefore, (1.122) implies that

where L
⇣ ⌘
~˙ = d n̂ · L
n̂ · L ~ = 0, (1.124)
dt
since n̂ is a fixed unit vector. We therefore conclude that rotational invariance
about the axis n̂ implies that L~ · n̂ is conserved.
In case of three dimensions, the absence of any distinguished direction requires
more. The potential V(N ) should be invariant under rotations about an arbitrary
axis n̂. This would then imply that all components of the total angular momentum
should be conserved. Therefore, even in three dimensions, full rotational invariance
(about a point) implies the conservation of the total angular momentum (about
that point).
Note that in our proof of the statement that full rotational invariance implies the
conservation of the total angular momentum, we have not used spatial translation
invariance. It is possible to have rotational invariance about a point without spatial
translation invariance, and thus conservation of total angular momentum about
that point with violation of total momentum conservation. To see this you can just
consider the potential
V(2) (~r1 , ~r2 ) = K|~r1 + ~r2 |2 , (1.125)
with K constant. Exercise: Write down the laws of motion for the two
particles interacting via the above potential, and verify that total mo-
mentum is not conserved but the total angular momentum about the
origin is conserved.
However, if spatial translation invariance and rotational invariance about a point
both are valid, then angular momentum about any arbitrary point is conserved. To
see this, we use that if we translate our origin, then

~r ! ~r + ~r0 (1.126)

but p~ is unchanged. Therefore,

N
X N
X N
X
~ =
L ~0 =
~rk ⇥ p~k ! L ~ + ~r0 ⇥
(~rk + ~r0 ) ⇥ p~k = L ~ + ~r0 ⇥ P~ . (1.127)
p~k = L
k=1 k=1 k=1

Since both L~ and P~ are conserved, L

~ 0 should be conserved too. Therefore, if both
spatial translation invariance and rotational invariance about a point hold, then
angular momentum about any arbitrary point is conserved.
How does rotational invariance constrain the two-body potential? To see this,
we will need to find combinations of two vectors ~r1 and ~r2 . In the context of
Ayan Mukhopadhyay 31

translational invariance the combination ~r1 ~r2 was invariant under translations,
so V(2) was a function of ~r1 ~r2 only.
We will show later that rotations leave the norm of any vector invariant, and
also the dot product of any two vectors invariant. It is easy to see this in two
dimensions where rotation by ✓ implies that
A0x = Ax cos ✓ + Ay sin ✓, A0y = Ax sin ✓ + Ay cos ✓,
Bx0 = Bx cos ✓ + By sin ✓, By0 = Bx sin ✓ + By cos ✓ (1.128)
Then it is easy to verify that
~0 · B
A ~ 0 = A0x Bx0 + A0y By0 = Ax Bx + Ay By = A
~ · B.
~ (1.129)
The generalization to three dimensional rotations will be proved later in this course.
In any case, one can form three independent rotationally combinations of ~r1 and ~r2
which are7
~r1 · ~r1 , ~r1 · ~r2 , ~r2 · ~r2 . (1.130)
Therefore rotational invariance implies,
V(2) (~ri , ~rj ) = V(2) (~ri · ~ri , ~ri · ~rj , ~rj · ~rj ), (1.131)
i.e. V(2) is a function of three variables instead of six variables. Therefore, applying
the chain rule and using the above result, we find that
V(2) (~ri , ~rj ) @V(2) (~ri , ~rj )
F~ij = ~ i @V (2) (~ri , ~rj ) =
r 2~ri ~rj ,
@(~ri · ~ri ) @(~ri · ~rj )
@V(2) (~ri , ~rj ) @V(2) (~ri , ~rj )
F~ji = ~ j V(2) (~ri , ~rj ) =
r 2~rj ~ri (1.132)
@(~rj · ~rj ) @(~ri · ~rj )
Note rotational invariance does not guarantee validity of Newton’s third law, i.e.
Fij 6= F~ij generically if we have only rotational invariance about a point in space.
You can check the above identities and also Fij 6= F~ij for the rotationally invariant
but translationally non-invariant two-body potential (1.125).
One can again check that rotational invariance gives conservation of angular
momentum as follows:
N
X N
X N
X X
~˙ =
L (~ri ⇥ F~i ) = (~ri ⇥ F~ij ) = (~ri ⇥ F~ij + ~rj ⇥ F~ji )
i=1 i=1 j=i,j6=i <ij>
X✓ @V(2) @V(2)(~ri ,~rj ) (~ri , ~rj )
◆
= ~ri ⇥ ~rj + ~rj ⇥ ~ri
<ij>
@(~ri · ~rj ) @(~ri · ~rj )
X @V(2) (~ri , ~rj )
= (~ri ⇥ ~rj + ~rj ⇥ ~ri ) = 0. (1.133)
<ij>
@(~ri · ~rj )

In the second line, we have used (1.132) and ~ri ⇥ ~ri = ~rj ⇥ ~rj = 0, and finally
~ri ⇥ ~rj = ~rj ⇥ ~ri in the third line. Of course, this result was already expected, but
now we also see that the cancellation of the torques occurs pairwise, i.e. for each
pair of particles, when V(N ) can be written as a sum of two-body terms.
When we have both translation and rotation invariance, we will need to find a
combination of vecri · ~ri , ~ri · ~rj and ~rj · ~rj which depends only on ~ri ~rj and this
is simply the combination
p
|~r1 ~rj | = ~ri · ~ri 2~ri · ~rj + ~ri · ~rj . (1.134)
7 In two dimensions, one can check that A ~⇥B ~ = Ax By Ay Bx is also rotationally invariant.
However, if we reflect the y-axis on the x-axis and thus create a left-handed coordinate system,
~ ⇥ B.
then the y-components of all vectors switch sign and so does A ~
32 Physics I

So, when both spatial translational invariance and rotational invariance about a
point hold, we should have

V(2) (~ri , ~rj ) = V(2) (|~ri ~rj |) (1.135)

Therefore, V(2) is a function of a single variable, namely |~ri ~rj |.

We thus come to a remarkable conclusion. All fundamental forces can be de-
scribed by a function of a single variable, namely the interparticle distance, and the
charges and masses of the particles.
Applying chain rule and (1.135), we get

~ri ~rj @V(2) (~ri , ~rj )

F~ij = ~ i V(2) (~ri , ~rj ) =
r = F~ji (1.136)
|~ri ~rj | @|~ri ~rj |

Here we note that not only we have Newton’s third law but also the result that the
two-body force is always in the direction of the vector connecting the positions of
the two particles.
Once again let’s check how we obtain angular momentum conservation.

N
X N
X N
X X
~˙ =
L (~ri ⇥ F~i ) = (~ri ⇥ F~ij ) = (~ri ⇥ F~ij + ~rj ⇥ F~ji )
i=1 i=1 j=i,j6=i <ij>
X X✓ ~ri ~rj @V(2) (~ri , ~rj )
◆
= ((~ri ~rj ) ⇥ F~ji ) = (~ri ~rj ) ⇥ = 0 (1.137)
<ij> <ij>
|~ri ~rj | @|~ri ~rj |

In the last line, we have used F~ij = F~ji for the first equality, then (1.136) for the
second equality, and finally (~ri ~rj ) ⇥ (~ri ~rj ) = 0. Of course, we are overdoing
it because we already showed angular momentum conservation, and cancellation of
torques pairwise using only rotational invariance. One point of the above exercise
is that angular momentum conservation is usually proved in textbooks in this way.
So we want to alert you that the above proof assumes both translation and rota-
tional invariance (implicitly). The result, as demonstrated earlier, follows just from
rotational invariance and not assuming Newton’s third law. Secondly, the above
manipulations can also be used to see a simple and useful result shown below.
Let’s see what happens if N particles experience an external potential of the
form
N
X
ext
V(N r1 , · · · , ~rN ) =
) (~ V ext (~ri ) (1.138)
i=1

This could be due to the presence of a N + 1-th particle which is extremely heavy so
that its motion is una↵ected by those of the N particles and which is at rest. In this
case, each particle will experience an additional force F~iext = r ~ i V ext (~ri ). Let’s
assume both translational and rotational invariance, so that Fij = F~ji / ~ri ~rj
~
holds. In this case,
0 1
N
X N
X N
X
˙
P~ = F~i = @F~iext + F~ij A
i=1 i=1 j=1,j6=i
N
X X N
X
= F~iext + (F~ij + F~ji ) = Fiext , (1.139)
i=1 <ij> i=1
Ayan Mukhopadhyay 33

where all manipulations above should already be familiar. Also.

N
X N
X N
X N
X
~˙ =
L (~ri ⇥ F~i ) = ~ri ⇥ F~iext + (~ri ⇥ F~ij )
i=1 i=1 i=1 j=i,j6=i
N
X X N
X
= ~ri ⇥ F~iext + ((~ri ~rj ) ⇥ F~ji ) = ~ri ⇥ F~iext (1.140)
i=1 <ij> i=1

where we have used the same manipulations as in (1.137). Thus when we have both
spatial translation invariance and rotational invariance, we obtain that

XN X N X N
˙ ~˙ =
P~ = F~iext , L ~ri ⇥ F~iext = ~⌧iext , (1.141)
i=1 i=1 i=1

with ⌧iext = ~ri ⇥ F~iext denoting the external torque on the i-th particle. The net
result is that the rate of change of total momentum is just the sum of the external
forces on each of the particles, and the rate of change of total angular momentum is
simply the sum of the external torques acting on each of the particles. This justifies
all the force diagrams that you have used to compute motions of bodies in high
school.
Now we use the definition of center of mass which is located at
PN PN
mi~ri i=1 mi~ri
~rCM = Pi=1
N
= (1.142)
i=1 m i M
PN
with M = i=1 mi being the total mass of the system. We readily see that
PN
mi~r˙i P~
~r˙CM = Pi=1
N
= , i.e. P~ = M ~r˙CM (1.143)
i=1 mi
M

It follows that

X N N
˙ 1 X ~ ext
P~ = M ~r¨CM = F~iext , i.e. ~r¨CM = F (1.144)
i=1
M i=1 i

One can use more complicated manipulations for a system of particles forming a
rigid body and obtain more interesting expressions, but we will come to these later
in this course.

1.3.4 There is no special moment of time for the laws of

motion
The fundamental laws of motion are immutable with time simply if the masses and
charges of all particles do not change in time, and the potential V(N ) do not have
any explicit time dependence.
Is anything conserved if this is the case? Consider N trajectories ~ri (t) for the
N particles, with i = 1, · · · , N . Then
N
X
d ~ i V(N ) (~r1 , · · · , ~rN , t)
V(N ) (~r1 , · · · , ~rN , t) = ~vi · r
dt i=1
@V(N ) (~r1 , · · · , ~rN , t)
+ . (1.145)
@t
34 Physics I

Since these trajectories ~ri (t) physically can result from following the laws of motion,
we have r~ i V(N ) = mi~ai , and therefore

N
X
d
V(N ) (~r1 , · · · , ~rN , t) = mi~vi · ~ai
dt i=1
@V(N ) (~r1 , · · · , ~rN , t)
+ . (1.146)
@t
However,
d ~vi · ~vi
~vi · ~ai = ,
dt 2
and therefore
N ✓ ◆
d d X 1
V(N ) (~r1 , · · · , ~rN , t) = mi~vi · ~vi
dt dt i=1 2
@V(N ) (~r1 , · · · , ~rN , t)
+ . (1.147)
@t
We can rewrite the above in a more compact form by defining the energy:
N ✓
X ◆
1
E(N ) = mi~vi · ~vi + V(n) (~r1 , · · · , ~rN , t),
i=1
2
N
X p~i · p~i
= + V(N ) (~r1 , · · · , ~rN , t) (1.148)
i=1
2mi

Then (1.147) implies that (recall p~i = mi~vi )

@V(N ) (~r1 , · · · , ~rN , t)
Ė(N ) = (1.149)
@t
Therefore, the total energy is conserved by the laws of motion if and only if the
N -body potential has no explicit time dependence.
It is convenient to define the kinetic energy of the i-th particle as
1 p~i · p~i
Ti = mi~vi · ~vi = (1.150)
2 2mi
and the total kinetic energy of the N -body system as
N
X
T = Ti . (1.151)
i=1

Then,
E(N ) = T + V(N ) (~r1 , · · · , ~rN , t). (1.152)
The total energy of the N -body system is simply the sum of the total kinetic energy
and the potential energy.
Note that in the above derivation, we have not used the criterion that V(N )
should be a sum of two-body terms. We also have not assumed spatial translational
or rotational symmetry. So conservation of energy can hold although momentum
conservation and angular momentum conservation can be violated.
Let’s see the more general result (1.149) in action. Consider one-dimensional
motion of a particle along z-axis under the influence of the potential

V (z, t) = mg(t)z (1.153)

Ayan Mukhopadhyay 35

which can result from a time-dependent g. Then the equation of motion is

mz̈ = mg(t). (1.154)

The energy is
1
E= mż 2 + mg(t)z. (1.155)
2
One can check that

Ė = mż z̈ + mg(t)ż + mz ġ(t)

= ż(mz̈ + mg(t)) + mz ġ(t) = mz ġ(t)
@V (z, t)
= . (1.156)
@t
In the second line above, we have used the equation of motion (1.154). Thus we
have verified (1.149).
For many practical applications, it is useful to consider a more general situation
where each particle experiences an external force field. Consider a single particle
again. Then ✓ ◆
~ d 1
m~a · ~v = F · ~v , i.e. m~v · ~v = F~ · ~v . (1.157)
dt 2
Integrating with time we readily obtain that
Z t2
T (t2 ) T (t1 ) = dt F~ · ~v = W (1.158)
t1

Therefore, the change in kinetic energy is equal to the work done. This is known
as the work-energy theorem. Only when F~ = rV ~ and V has no explicit time-
dependence, we get
W = V (~r(t2 )) + V (~r(t1 )) (1.159)
implying
T (t1 ) + V (~r(t1 )) = T (t2 ) + V (~r(t2 )), (1.160)
i.e. energy conservation. This is why F~ = rV ~ is called a conservative force
field when V has no explicit time-dependence. In reality, F~ can be split into two
parts, one part is conservative and another is not and can arise from friction as for
instance. In this case,

F~ = F~c + F~nc , with F~c = ~ (~r).

rV (1.161)

In this case, it is easy to see that

Z t2 Z t2
W = ~
dt F · ~v = V (~r(t2 )) + V (~r(t1 )) + dt F~nc · ~v (1.162)
t1 t1

Therefore, substituting this into (1.158) we get

Z t2
E(t2 ) E(t1 ) = dt F~nc · ~v . (1.163)
t1

Therefore the change in energy of the system is simply the work done by the non-
conservative (frictional) force.
Di↵erentiating (1.163) (w.r.t. t2 ) or otherwise we can also derive that

Ė = F~nc · ~v (1.164)
36 Physics I

If this non-conservative force is proportional to the velocity

F~nc = ~v ,  > 0,  constant, (1.165)
then clearly
F~nc · ~v = ~v · ~v < 0. (1.166)
This implies that W < 0 and from (1.163) we see that
E(t2 ) E(t1 ) = W < 0. (1.167)
Therefore the energy of the system should always decrease. Note for the above result
to hold, V (~r) giving the conservative force itself has no explicit time-dependence.
The above generalizes to a system of N particles mutually interacting with a
potential V(N ) without any explicit time dependence. You can readily show this
yourself. If each particle experiences a non-conservative (frictional) force F~i,nc ,
then we can show that
N Z t2
X N
X
E(N ) (t2 ) E(N ) (t1 ) = dt F~i,nc · ~vi = Wi,nc = Wnc (1.168)
i=1 t1 i=1

where Wi,nc is the work done on i-th particle by the external non-conservative force
and Wnc is the total work done on the system of N particles by the non-conservative
forces.
It is also easy to see that the di↵erential version of the above result is
N
X
Ė(N ) = F~i,nc · ~vi (1.169)
i=1

However, now we have a puzzle. Under time translations, t ! t0 = t + t0 . We

have d/dt = d/dt0 , and therefore ~v 0 = ~v and ~a0 = ~a. So an equation of motion of
the kind
m~a = ~ (~r)
rV ~v (1.170)
is actually time-translation invariant since V (~r) has no explicit time-dependence.
However, the energy is not conserved as evident from (1.163). Well, we have already
shown that in addition to time-translation invariance, we also need absence of non-
conservative forces. However, is there a symmetry which rules out an equation of
the above type? Indeed such a symmetry exists. It is time-reversal symmetry in
which t ! t. which changes the direction of flow of time from future to past.
Under this operation d/dt = d/dt0 , and therefore ~v 0 = ~v and ~a0 = ~a. Since
the velocity reverses sign but acceleration does not, (1.170) is not invariant under
time-reversal symmetry.
Practically, we can observe time-reversal by recording a motion and running the
movie backwards. If the law of motion is (1.170), then energy always decreases in
the future. When the movie runs backwards, the energy always increases.
All fundamental laws of physics are time-reversal invariant. How come then
force like friction which break time-reversal symmetry arise in real life? The answer
to this profound question lies in the law of probabilities, and it was unanswered
fundamentally by the great Austrian physicist Ludwig Boltzmann. This is however
beyond the present course.
To summarize, after invoking spatial translation symmetry, time translation
symmetry and rotational symmetry, we obtain that the N -body potential for any
fundamental interaction should take the form of a sum of two-body terms:
X
V(N ) (~r1 , · · · , ~rN ) = qi qj v(|~ri ~rj |), (1.171)
<ij>
Ayan Mukhopadhyay 37

and thus given by the intrinsic charges of the particles, and a single function (v)
of a single variable, namely the inter-particle distance. It turns out, for reasons we
will mention later in this course, this function V (x) is typically 1/x, as in the case
of the electric and gravitational force.

1.3.5 Galilean invariance and the inertial frames

We have not yet answered one fundamental question: how do we physically con-
struct the coordinate systems in which the equations of motion hold? In particular,
where should the origin of such a coordinate system be? It is true that translation
invariance tells us that we can put the origin anywahere, but practically we cannot
label any point in empty space.
Galileo’s insight leads to a very profound answer: we can have the origin at
the location of any free particle, or the center of mass of any free body which
does not experience any external force, as for example, by virtue of being very
far away from all other particles. (In reality, of course, the frame will be only
approximately inertial.) We will call this particle an observer – although it is not
actually observing anything. Note that rotational invariance tells us we can orient
our coordinate system however we like.
The principle of Galilean invariance states that the fundamental laws of physics
should take the same form in all inertial frames. Therefore, one can check their
validity in any inertial frame. An inertial observer does not experience any force, so
its acceleration vanishes in any inertial frame. This statement should be understood
as saying that any inertial observer should observe another inertial observer moving
at a constant velocity. Mathematically, this implies that the two inertial frames
should be related by the following Galilean transformation:

~r0 = ~r + ~r0 + ~v t. (1.172)

Setting ~r = 0, we get the origin of the first inertial frame. The origin of the
second inertial frame (attached to the second observer) then follows the trajectory
~r0 (t) = ~r0 + ~v t.
For an arbitrary particle, the trajectory ~r(t) in the first inertial frame maps to
the trajectory
~r0 (t) = ~r(t) + ~r0 + ~v t. (1.173)
in the second inertial frame. Clearly,

~v 0 (t) = ~v (t) + ~v . (1.174)

and
~a0 (t) = ~a(t). (1.175)
This implies that acceleration is absolute, meaning that all inertial frames measure
the same acceleration. So the statement that the acceleration of an inertial frame
vanishes is self-consistent, as it is valid in all inertial frames.
The left hand side of the laws of motion involving the acceleration is automati-
cally Galilean invariant. What about the force? If we have N particles, then under
Galilean transformations:
~ri0 (t) = ~ri (t) + ~r0 + ~v t. (1.176)
Therefore, for Galilean invariance we will need

V(N ) (~r1 , · · · , ~rN ) = V(N ) (~r1 + ~r0 + ~v t, · · · , ~rN + ~r0 + ~v t) (1.177)

This relation is apparently non-trivial because the left hand side has no explicit
time-dependence. However, it is easy to see that if the potential is translation
38 Physics I

invariant, then it is automatically Galilean invariant too. Any combination of the

positions, such as ~r1 ~r2 which is translation invariant is also automatically Galilean
invariant. Can you provide a general proof? (Hint: use Taylor expansion in ~v , etc.
and check.) You can also check that the forces are also Galilean invariant.
Since translation invariance of Newtonian laws automatically implies Galilean
invariance, the later does not imply any new conservation law other than that of
the total momentum.
The conceptual significance of Galilean invariance is that it tells us about how
can we physically deduce the fundamental laws of motion. In practice, no experi-
mental setup is inertial. Galilean invariance also tells us about the origin of pseudo-
forces in non-inertial frames which we should then take into account systematically
while verifying a physical theory. We will discuss this later.