Quantum Mechanics - II

Angular Momentum - II : Symmetry in

Physics
Dipan Kumar Ghosh
UM-DAE Centre for Excellence in Basic Sciences
Kalina, Mumbai 400098
September 16, 2019

1 Introduction
We have been acquainted with the concept of symmetry in classical mechanics and its
relationship with conservation laws. Understanding symmetry helps us in making sense
out of a clutter of experimental data. The relationship between symmetry and conserva-
tion laws was demonstrated by the famous Noether’s theorem, which states that if the
classical Lagrangian of a physical system is left unchanged by a continuous symmetry
transformation, a physical quantity would then become conserved in time. For instance,
homogeneity of space under translation implies that the physical laws must be invariant
under such translation. This gives rise to conservation of linear momentum. Likewise,
isotropy of space would require that the physical laws are invariant under rotation. This
would result in conservation of angular momentum. A somewhat less obvious symmetry is
the connection between time translation and conservation of energy. If the Lagrangian of
the system does not explicitly depend upon time and the potential is velocity independent,
conservation of energy would follow. These symmetries arise due to inherent structure
of space time and are generally referred to as geometrical symmetries. There is a class
of symmetry which has its origin in the nature of forces between interacting bodies. For
instance, there are accidental degeneracies in the spectrum of harmonic oscillator or the

1
hydrogen atom, where the energies are independent of the angular momentum quantum
numbers. The origin of such degeneracies can be traced to a symmetry higher than O(3)
symmetry associated with geometry. The symmetry in such cases, which are relatively
rare, are called dynamical symmetry. In this lecture, we will look into the extension of
these ideas to quantum mechanics. As has been pointed out earlier, there are two corre-
spondence which enables us to map the classical prescriptions to the quantum mechanical
ones. The classical Poisson bracket is replaced in quantum mechanics by the commutator
of the corresponding operators (up to a factor i~) and the canonical transformations of the
phase space is replaced by unitary transformations in the Hilbert space. The conserved
dynamical quantities arising out of the continuous symmetries of the classical Lagrangian,
results, in quantum mechanics from the corresponding symmetry of the Hamiltonian and
yield a constant of motion of a hermitian operator (corresponding to the physical observ-
able).

2 Classification of Symmetry
As in classical mechanics, we may classify transformation as active ( involving, for in-
stance, the physical position of a rigid body on rotation) or passive (which is merely a
transformation of the coordinate axes). In quantum mechanical analogue, we represent
active transformation by physically changing the state of the system by a unitary operator,
| ψi →| ψ 0 i = Û | ψi (1)
Alternatively, we may provide a passive view of transformation by keeping the state
unchanged but transforming the operators corresponding to the physical observables. An
expression for such “rotated operator” is obtained by requiring that the expectation value
of the operator  remains the same in both the pictures, i.e. hψ 0 | Â0 | ψi ≡ hψ |  | ψi,
which gives
hψ | U † Â0 Û | ψi ≡ hψ | Â | ψi (2)
resulting in
Â0 = U ÂU † (3)
We know that the square of the scalar product | hψ | φi |2 gives the probability that a
measurement made on the state | ψi collapses the state to the state | φi. If a transforma-
tion Û takes a system from the state | ψi to | ψ 0 i and a state | φi to the state | φ0 i, then

2
the result of such a measurement would be the same whether the measurement was done
on the unprimed system or on the primed system. This implies
| hψ 0 | φ0 i |2 =| hψ | φi |2 =| hψ | U † U | φi |2
which implies
(
hψ | φi U is unitary
hψ | U † U | φi =
hφ | ψi U is anti − unitary
We define an anti-unitary operator Ũ by the following properties :
hŨ ψ | Ũ φi = hψ | φi (4)
The operator is linear but has different consequence when acting on a complex scalar
Ũ [| ψi+ | φi] = Ũ | ψi + Ũ | φi (5)
∗
Ũ c | ψi = c Ũ | ψi (6)
We will have more to say about such operators when we deal with time reversal and
charge conjugation symmetries.
From a different perspective, symmetries may be continuous or discrete. Continuous
symmetry, as the name suggests is the symmetry which can be built continuously from
identity by means of infinitesimal transformations. As such transformations must contain
the identity element of the group, continuous transformations are given by unitary opera-
tors. We are aware from Noether’s theorem of classical mechanics the intimate connection
between continuous symmetry and conservation of a physical quantity. This is also true
in quantum mechanics; the difference being that instead of a Lagrangian approach to
symmetry, we need to take a Hamiltonian approach. Space translation and rotations are
examples of such continuous transformations. There are symmetries which cannot be
built up by means of infinitesimal transformations starting with identity. Examples of
such discrete transformations are space inversion or reflection.
Consider systems for which an observable A corresponding to the hermitian operator Â
is left invariant under a continuous unitary transformation Û , i.e. U ÂU † = Â. The
transformation being continuous, may be built up from its infinitesimal generator G, i.e.
i
Û = 1 − Ĝ, where  is a real parameter and ~ has been introduced for convenience.
~
Note that
i
Û −1 = 1 + Ĝ = U †
~
3
 i
Since U † = 1 + Ĝ† , unitarity of Û , implies G† = G, i.e. Ĝ is a hermitian operator.
~
Symmetry implies that the system remains invariant under such an operation. Suppose
under the symmetry operation U , a state | ψi →| ψ 0 i and | phii →| φ0 i. If the time
development of the system is governed by Hamiltonian H, then we must have

hφ0 | H | ψ 0 i = hφ | H | ψi

Substituting | ψ 0 i = U | ψi and | φ0 i = U | φi, we get

hU φ | H | U ψi = hφ | H | ψi

which implies
hφ | U † HU | ψi = hφ | H | ψi
Since the above is true fro arbitrary states which are left invariant under the symmetry
operation, we must have
U † HU = H =⇒ [U, H] = 0 (7)
Since U is given by its infinitesimal generator Ĝ defined above, it implies [H, G] = 0.
Ig Ĝ is taken to be independent of time, we get, using Heisenberg’s equation of motion
dG
= [G, H] = 0 which implies that Ĝ is a constant of motion. Thus the symmetry
dt
generated by G leads to conservation of G. Since U commutes with the hamiltonian, it
has simultaneous eigenstates with it. Suppose H | En i = En | En i. Since U is a symmetry
operation, U | En i is also an energy eigenstate with the same energy, as,

H(U | En i) = U H | En i = En (U | En i)

However, the state U | En i need not be identical to the state | En i. In such a situation,
the energy spectrum is said to have degeneracy.

2.1 Translational Symmetry

2.1.1 Space Translation and Conservation of Momentum
Consider eigenkets of the position operator x̂. (Here we consider one dimensional trans-
lation; extension to the three dimensions in straightforward)

x̂ | xi = x | xi (8)

4
Let T (a) be the translational operator, whose action on | xi gives rise to the ket | x + ai.
Here a is a real parameter, independent of x. The effect of the translation operator on
the position operator is given by T (a)x̂T (a)† . Thus we have

T (a)x̂T (a)† | xi = (x + a) | xi

Using the infinite generator of translation, we have

   
ia ia
[ 1 − G x̂ 1 − G ] | xi = (x + a) | xi
~ ~
Expanding and using (8) we get
ia
− (Ĝx̂ − x̂Ĝ) = a =⇒ [Ĝ, x̂] = i~
~
Comparing this with the commutator [x̂, p̂] = i~, we conclude Ĝ = −p̂. Thus the in-
ia
finitesimal generator for translation by a is 1 + p. For finite translation, this gives
~
iap̂
G(a) = exp(i ) (9)
~
Three dimensional translation is given by
i~a · p̂
G(~a) = exp(i )
~
In a similar way one can show that the momentum translation operator is given by
i~p0 · r̂
G(~p0 ) = exp(i )
~
[In the active transformation, the same result can be derived by considering the action of
the translation operator on the wavefunction ψ(x),

T (a)ψ(x) = ψ(x + a)

We may expand ψ(x + a) in a Taylor series

dψ a2 d2 ψ
ψ(x + a) = ψ(x) + a + + ...
dx 2! dx2
dp
Identifying p = −i~ , the result follows.]
dx
5
2.1.2 Time Translation
The time translation is defined in an analogous way1 . Time translation symmetry is
understood in the following way. Time translation invariance means that the laws of
physics are unchanged if all the events are shifted by a common time interval τ , i.e.,
physical laws should not depend upon the origin chosen for time. We may formally define
it by saying that if the state of system at time t is given by | ψ(t)i then, there exists a
solution | ψ 0 (t)i of the quantum mechanical equation of motion which is equal to (up to
an uninteresting multiplicative phase) the previous solution at time t + τ , i.e.

| ψ 0 (t)i = eif (t,τ ) | ψ(t + τ )i

Here τ is a fixed real parameter having the dimension of time by which the two time
intervals are connected. Thus, there should be an operator U (τ ) which, acting on the
state | ψ(t)i would generate the state the state of the system at time t + τ . The form of
U is analogous to the way we defined generator of space translation. We postulate
 
iHτ
U (τ ) = exp − (10)
~

The action of U (τ ) is given by

U (τ ) | ψ(t)i =| ψ(t + τ )i (11)

Setting t = 0 in the above equation, we get

 
iHτ
U (τ ) | ψ(0)i = exp − | ψ(0)i = ψ(τ )
~

As τ is arbitrary, we would then have, for any time t,

 
iHt
| ψ(t)i = exp − | ψ(0)i (12)
~

Note that since τ is a fixed parameter, [U, H] = 0. Starting with the equation for
U | ψi, we have
∂
i~ (U | ψi) = H(U | ψi) = U H | ψi
∂t
1
: Lectures on Quantum Mechanics, Steven Weinberg, Cambridge, 2013

6
which gives
∂
i~ | ψ(t + τ )i = H | ψ(t + τ )i
∂t
Thus the time translated state satisfies the Schrödinger equation.

We have seen that in the Heisenberg picture, the time dependence of an operator Â
is given by    
Ht Ht
AH (t) = exp i ÂS exp −i
~ ~
For the particular case of the operator  being the Hamiltonian Ĥ itself, this would imply
that the form of the Hamiltonian remains the same in both the pictures.
Symmetry principle implies that if an observer sees a state | ψ(t)i evolving as per (12),
the another observer must see the state U | ψ(t)i evolving by the same physical equation,
where
U | ψ(t)i = exp(−iHt/~)U ψ(0) (13)
Equations (12) and (13) will be consistent if we have
   
Ht Ht
exp −i U = U exp −i
~ ~

which requires U to be a linear operator so that

U −1 HU = H

the Hamiltonian must be invariant under the symmetry transformation, For infinitesimal
generator G of U , as we have seen earlier, this gives rise to

[H, G] = 0

The above requires U to be unitary. For anti-unitary operators, because of the presence
of the complex number i, this would give rise to U † HU = −H. This means, for every
energy eigenstate | ψi with eigenvalue E, there would correspond an eigenstate U | ψi
with energy −E. This would make the spectrum of the energy unbounded from below
resulting in an unstable ground state, since we could choose an eigenstate with energy
as large as we desire and there would correspond a state of negative energy of an equal
magnitude.

7
2.1.3 Rotational Symmetry
In our discussion on SO(3) group, we have extensively discussed the subject of rotational
symmetry. We briefly touch upon it again for completeness. Consider rotation about a
fixed axis, say z-axis, by an angle δϕ. Taking δϕ ~ to be a vector of magnitude δϕ and
~ × ~r. Thus the
direction along the axis of rotation, the position vector ~r → ~r0 = ~r + δϕ
position vector changes by δr~ = δϕ~ ×~r. In a similar manner, the momentum vector would
change by δp~ = δϕ
~ × p~.
Let us write the generator of rotation as Ĝ The unitary operator for rotation is then given
~ · Ĝ
δϕ
by Û = exp(−i ), where Ĝ has been shown to be hermitian before. For infinitesimal
~
rotation, we have
~ · Ĝ
δϕ
Û = 1 − i
~
The operators are transformed as  →  = U ÂÛ † . Taking  to be the position operator
0

r̂, we have ! !
~ · Ĝ
δϕ ~ · Ĝ
δϕ
~ × ~r = 1 − i
~r + δϕ ~r 1 + i
~ ~
Taking the i−th component of both sides and using i, j, k as an ordered triad
! !
iX iX
ri + (δϕ)j rk − (δϕ)k rj = 1 − δϕm Gm ri 1 + δϕm Gm
~ m ~ m
" #
i X
(δϕ)j rk − (δϕ)k rj = − δϕm Gm , ri
~ m
i
= − [δϕi Gi + δϕj Gj + δϕk Gk , ri ]
~
Two sides of the above would be consistent, if we require [Gi , rj ] = −i~ijk rk . Compar-
ing this relation with the commutator of the components of angular momentum L ~ with
position, viz., [Li , rj ] = i~ijk rk . Thus we identify G with −L and get
~ · L̂
δϕ
U = exp(i ) (14)
~
For rotation about z− axis, the rotation operator would be written as
δϕ · Lz
U (z, ϕ) = exp(i ) (15)
~
8
and similarly for the axis being along x− or y− directions.
If Uz commutes with the Hamiltonian, we may label the common eigenstate of H and Lz
with E and m. However, for rotationally invariant system (such as the hydrogen atom
problem) , they would not depend on m because Lx and Ly acting on a state of a given
m would give rise to states with m ± 1.

3 Discrete Transformations
3.1 Parity
We had seen that the rotational matrices satisfy RT R = I, which implies detR = ±1. The
subclass of matrices which satisfy detR = +1 are proper rotations which form a subgroup
SO(3) of the full rotation group O(3). The matrices with detR = −1 do not form a
subgroup because (i) it does not have the identity element and (ii) the product of two
elements of such a set has detR = +1 and hence does not form a subgroup. However, the
transformations of space inversions ~r = −~r form an important symmetry transformation
known as Parity. The same is true of momentum operator p~ which changes sign under
parity. These vectors (~r, p~ etc.) which reverse their sign under space inversion are known
as polar vectors or simply as vectors. In contrast, vectors such as angular momentum L ~
does not change sign under parity. They are known as pseudovectors or axial vectors.
Other example f a pseudovector is the magnetic field vector B. ~ One can look at the
expression for Biot Savart’s law to confirm that B ~ does not change sign when the space
coordinates are reversed. One can similarly distinguish between two classes of scalars,
viz. a scalar and a pseudoscalar. Under space inversion, a scalar does not change sign
whereas a pseudoscalar does. An example of a pseudoscalar is the magnetic flux which
being dot product of a pseudovector B ~ is a pseudoscalar.
~ and a polar vector dS
In quantum mechanics, we define a parity operator Π by its action on a state in the
position basis. Thus Π is a unitary operator, whose action on a state in the position bass
is given by
Π | ~ri =| −~ri (16)
This definition, along with its unitarity enables us to obtain some properties of the oper-
ator. Firstly Π2 = I. This is obvious because two application of Π would return the state
back to the original state ~r → −~r → +~r. It also follows from (16) that Π is hermitian.

9
To see this multiply both sides of the equation Π2 = I by Π† to get

Π2 P † = Π†
Π† ΠΠ = Π†

Using the unitarity, we have Π† Π = I which gives Π = Π† . The operator Π turns out
to be an observable (Experiment of Wu on beta decay of Cobalt by observing directional
property of emitted electrons and gamma rays when strong magnetic fields are applied
). Since Π2 = I, the eigenvalues of Π are ±1. Using the definition of a wavefunction
ψ(~r) = hψ | ~ri, we get the action of Π on the position space wavefunction as

Πψ(~r) = ψ(−~r)

ψ(~r) ± ψ(−~r)
The eigenstates of Π corresponding eigen values ±1 may be written as with
2
the plus sign corresponding to what is known as the even parity state, i.e., a state which
does not change sign under parity and the minus sign corresponds to the odd parity state,
which does change sign. For instance, the eigenfunctions of the orbital angular momentum
Ylm (θ, ϕ) has either odd parity or even parity depending on whether the quantum number
l is odd or even,

Ylm (θ, ϕ) → (−1)l Ylm (θ, ϕ)

Let’s look at what happens to the operators under parity? Consider position operator
first under which | ψi → r̂ | ψi. Since | ri is an eigenstate of the position operator with
eigenvalue r (and parallely, | −ri is of −r̂) we may write

hr | Πr̂ | ψi = h−r | r̂ | ψi
= −rh−r | ψi
= −rhr | Π | ψi
= −hr | r̂Π | ψi

Thus Π and r̂ anti-commute. A similar relation may be shown for the momentum operator.
Recall that the translation operator T̂ is given by
~a · p̂
T̂ (~a) = exp(−i )
~

10
The action of T̂ (~a) on the eigenstate of position is given by

T̂ (~a) | ~ai =| ~x + ~ai

Under the action of parity, the operator T̂ (~a) → Π† T̂ (~a)Π. Thus we have

Π† T̂ (~a)Π | ~xi = Π† T̂ (~a) | −~xi

= Π† | −~x + ~ai
= i~∇hr | Π | ψi
= Π† Π | ~x − ~ai
=| ~x − ~ai
= T̂ (−~a) | xi

which shows Π† T̂ (~a)Π = T̂ (−~a). Thus we have, taking infinite generator for translation
   
† i i
Π 1 − ~ · p̂ Π = 1 + ~ · p̂
~ ~

Expanding and keeping terms of order , it follows that Π† p̂Π = −p̂, i.e. p̂ and Π
anti-commute. However, as pointed out earlier, Π commutes with the orbital angular
momentum. To be consistent with the angular momentum commutation relations, we
take Π to commute with Jand ~ ~ Note that x̂ and Jˆ transform similarly under rotation
S.
though they have different behaviour under parity operation.
In general, for a vector operator V , we have ΠV Π† = −V whereas for a psedudovector
ΠÂΠ† = Â. A scalar is invariant both under rotation and parity whereas for a pseu-
doscalar, K we have ΠKΠ† = −K and U (R)KU (R)† = K.
Space inversion symmetry plays an important role in the selection rules of atomic tran-
sitions. Note that for long wavelength atomic transition, the transition rates are propor-
tional to the square of the matrix element of dipole operator ex̂ between initial and final
atomic states. Since the dipole operator has odd parity, this matrix element would be zero
unless the initial and the final atomic states have opposite parity. Recall that we pointed
out earlier that the atomic states have parity proportional to (−1)l . Thus the selection
rule for allowed transition is that the l value must change by ±1. This selection rule is in
addition to the rules imposed by Wigner-Eckart theorem discussed earlier. This implies

11
that no non-degenerate state has a dipole moment because hn | x | ni = 0. However, it
must be mentioned that parity is not conserved in weak interactions.
Examples
(1) Consider the Hamiltonian for one dimensional harmonic oscillator

p2 1
H= + mω 2 x2
2m 2
Recall the definition of ladder operators
r  
mω i
a= x̂ + p̂
2m~ mω
r  
† mω i
a = x̂ − p̂
2m~ mω

with the ground state being defined as a | 0i = 0. Since the Hamiltonian depends on
squares of x and p, it commutes wit the parity operator, [Π, H] = 0. Hence the energy
states of the oscillator have well defined parity. Note further that Πa† Π = −a† . This
gives
Π | n + 1i = πa† | ni = −a† Π | ni
If the eigenvalue of Π in the state | ni is denoted as πn , then

πn+1 | n + 1i = −a† πn | ni = −πn a† | ni = −πn | n + 1i

Thus πn+1 = −πn , which shows that the excited states of the harmonic oscillator alternate
in parity. Since the ground state is even, the parity of the n−th state is given by (−1)n .
(2) Consider hydrogen atom wavefunction. Since the potential energy term depends on
the distance | r | and the kinetic energy depends on p2 , the Hamiltonian is invariant under
parity transformation, meaning thereby the eigenfunctions of hydrogen atom have well
defined parity. This parity depends only on l value and is independent of m. To show this,
we consider the state of maximum m, i.e. m = l. For this case the spherical harmonic
takes a simple form
Yl,l = eilϕ sinl (θ)
Parity transformation implies θ → π − θ, ϕ → ϕ + π. Thus under parity, this state of
maximum l would become
eil(ϕ+π) sinl (π − θ)

12
Since sin(π − θ) = sin θ, and eilπ = (−1)l , we have

ΠYl,l = (−1)l Yl,l

Let us see what happens to the state with m = l − 1 under parity operation. Recall that
the state with m = l −1 is obtained from the state with m = l by application of the ladder
operator L− . Ignoring the multiplicative factor, we have | l, m = l − 1i = L− | l, m = li,
Thus

Π | l, m = l − 1i = ΠL− | l, li = L− Π | l, li = (−1)l L− | l, li = (−1)l | l, l − 1i

i.e., the state still has parity (−1)l . we may repeat the argument for lower values of l and
get the same result.

3.2 Time Reversal

Though the phrase “time reversal” is widely accepted in physics, a more appropriate
nomenclature would be to term this symmetry as reversal of motion. Time reversal itself
would violate the measurement principle as once the state has collapsed, it would not
be possible to precisely turn the sequence of the states which led to the present state.
However, the dynamics of the system when each component of the motion is reversed is
represented by letting t → −t in the equations.
Considering classical motion, Newton’s second law is invariant under t → −t because it
involves second derivative with respect to time. Hence of x(t) is a solution of the equation
x(−t) also is. However, if one includes a damping term in the equation of motion, the
system is not invariant under time reversal as the equation of motion now includes a
first derivative of position with respect to time. Situation is a little more complicated if
one considers electromagnetism. As the equations involves first derivative with respect
to time, it would appear that these equations are not time reversal invariant. However,
one has to take into consideration the fact that when the direction of velocity changes
sign, the direction of current also does so, (we assume that the charge remains invariant)
making the magnetic field B ~ also change sign. This would make, for instance, the term
∂B
remain invariant under time reversal. The electric field depends only on the charge
∂t
distribution and space coordinates and therefore remains invariant under time reversal.
With this prescription, Lorentz force becomes invariant under time reversal.
To see how the phrase “time reversal” becomes appropriate, consider the motion of a

13
particle under the action of a classical Hamiltonian H(~r, p~). Suppose at time t = 0, the
position and the momentum of the particle are given by ~r(0) and p~(0) respectively. Let
the position at time t be ~r(t) and momentum be p~(t). Starting from this instant, suppose
we wish to get back to the initial values of position and momentum after a time −t, as we
cannot change the direction of time at our will, what we do is to change the Hamiltonian
to a form H(~r, −~p) and run the process for a time t, we would then get back to the
original values of position and momentum after a time t in the forward direction. Thus
reversing the direction of momentum had the same effect as it would if we could reverse
the direction of time without having to change the Hamiltonian.
It would appear that all that needs to describe time reversal is to postulate an operator
Θ whose action on all time dependent quantities is to reverse their signs,
Θf (t) = f (−t)
However, this turns out to be unsatisfactory in quantum mechanics. To see this consider
what happens to the Schrödinger equation
∂
i~ ψ(~r, t) = Hψ(~r, t)
∂t
under time reversal. Since H has no explicit time dependence, when t → −t, the equation
would become
∂
−i~ ψ(~r, −t) = Hψ(~r, −t) (17)
∂t
which shows that ψ(~r, −t) is not a solution of the Schrödinger equation. For us to be able
to come to this result, we require that there exists an operator Θ which acting on the
state of the system gives the ”time reversed state”,
| ψi → Θ | ψi
As per our previous description, let us suppose we start from a state at time t = 0 denoted
by | ψ(0)i. Let this state develop for a time t driven by the Hamiltonian H so that at
time t the system is in the state | ψ(t)i, where
| ψ(t)i = e−iHt/~ | ψ(0)i (18)
Suppose we are to apply a time reversal operator Θ on | ψ(t)i and run the system exactly
for the duration of time t so that the state of the time reversed system becomes e−iHt/~ Θ |
ψ(t)i, we would then have arrived at the “time reversed state” Θ | ψ(0i,i.e.
e−iHt/~ Θ | ψ(t)i = Θ | ψ(0)i

14
because the time reversed state in the forward direction of time would retrace its original
path. If in this equation, we substitute (18) for | ψ(t)i, we would get

e−iHt/~ Θe−iHt/~ | ψ(0)i = Θ | ψ(0)i

As this equation must be true for all initial states | ψ(0)i, we get an operator relationship

e−iHt/~ Θe−iHt/~ = Θ (19)

which is equivalent to
e−iHt/~ Θ = Θe+iHt/~ (20)
For infinitesimal time t the above equation gives

−iHΘ = ΘiH

In normal situation, we could cancel the factor i from both sides getting −HΘ = ΘH.
Suppose | Ei is an eigenket of the Hamiltonian with energy E. We then would get

−HΘ | Ei = ΘH | Ei = EΘ | Ei

This implies that the state Θ | Ei is an eigenstate of the Hamiltonian with energy −E.
This, as we have seen in our comment in the last paragraph of the section on time trans-
lation operator (see comment following eqn. (12)), would make the ground state unstable
and is, therefore, an unacceptable situation. It even leads to an absurd situation in the
case of a free particle, suggesting that its time reversed state would have negative energy.
We therefore seek problem seek a resolution of the problem by looking for an alternative.
Recall that we had pointed out that the wavefunction itself is not observable. The prob-
ability, which is an observable, is preserved both by unitary operators and by operators
which we termed anti-unitary. If we assume that in this case our operators Θ is anti-
unitary, which has the following properties (see (4)):

Θ(c1 | αi + c2 | βi) = c∗1 Θ | αi + c∗2 Θ | βi

that is, the operator acts like a linear operator but complex conjugates the scalars which
multiply the states, it has no special action on the basis states themselves. Further, we
do not define action of Θ on a bra, i.e., the operator Θ only works on a ket to

15
its right.Thus, the operator Θ may be thought of as a product of a “complex conjugation
operator” K and a unitary operator U
Θ = UK (21)
where K operating on a scalar, complex conjugates everything to the right of it, giving,
for instance Ki = −iK and Kc | ψi = c∗ K | ψi. It has no effect on a real number.
Returning to the infinitesimal t version of (20), we would then get
[Θ, H] = 0
, which expresses the symmetry property of the Hamiltonian under time reversal. Re-
turning to (17), if we operate both sides of the equation by Θ = U K, remembering that
Θi = −iΘ, and the fact that Θ commutes with H, we get
∂
+i~ Θψ(−t) = HΘψ(−t)
∂t
Define
ψ̄(t) = Θψ(−t),
we get
∂
i~ ψ̄(t) = Hψ̄(t)
∂
which shows ψ̄(t) satisfies Schrödinger equation.
Let us now consider the transformation of the operators under time reversal. Define the
time reversed state corresponding to a state | αi by | α̃i = Θ | αi. Further, let us define
an operator à to be the time reversed operator corresponding to an operator  (we have
suppressed the hat(ˆ) sign on the time reversed form of the operator A to avoid clumsiness
of notation). By definition, we then have, for a hermitian observable Â,
hα̃ | Ã | β̃i = hα | Â | βi (22)
Let us define A† | αi =| γi, so that hγ |= hα | A. Thus, we have,
hα | A | βi = hγ | βi
= hβ̃ | γ̃i using anti − unitarity of the time − reversed states
= hβ̃ | Θ | γi
= hβ̃ | ΘA† | αi
= hβ̃ | ΘA† Θ−1 Θ | αi = hβ̃ | ΘA† Θ−1 | α̃i (23)

16
Note that in deriving the above, we have taken care so that Θ works only on the ket to
its right, never on a bra. If  is hermitian, we have, using A† = A

hα | A | βi = hβ̃ | ΘAΘ−1 | α̃i

An operator  is said to be even or odd under time reversal, if

ΘAΘ−1 = ±A

Taking α = β, it follows that the expectation value of an operator in the time reversed
state | α̃i is the same as in the state | αi if the operator A is even under time reversal
and is negative of the former if it is odd.

hα̃ |  | α̃i = hα | A | αi if  is even

= −hα | A | αi if  is odd

Coming to specific operators, we require expectation value of position operator x̂ to be

even and that of p̂ to be odd. Thus we have

hα | x̂ | αi = hα̃ | x̂ | α̃i

However, we have
hα | x̂ | αi = hα̃ | Θx̂Θ−1 | αi
Comparing the two expressions above, we get

Θx̂Θ−1 = x̂

In a similar way, we can show that since the expectation value of the momentum operator
changes sign under time reversal,
Θp̂Θ−1 = −p̂
It is easy to show that the commutation relation [xi , pj ] = i~δi,j becomes

Θ[xi , pj ]Θ−1 = −i~δi,j

~ −1 = −J.
We can also show that ΘJΘ ~

17
3.3 Charge Conjugation
Charge conjugation symmetry refers to invariance of physical laws when the particles are
replaced by their corresponding antiparticles. All charged particles have their antipar-
ticles, such as the positron corresponding to an electron, antiproton corresponding to a
proton etc. Some neutral particles such as γ and π 0 do not have antiparticles. However, it
is possible for a neutral particle to have an antiparticle. For instance the neutron has its
antiparticle, antineutron which is distinguished from neutron by having an opposite sign
for its magnetic moment. Antineutrino is distinguished from the neutrino by its having
an opposite helicity (component of spin in the direction of its motion).
We define charge conjugation operator C by its action on charge quantum number or on
its magnetic moment, i.e. it converts a particle to its antiparticle. Action of the charge
conjugation operator on the state of a particle with charge -parity state a is given by

Ĉψ(a) = Ca ψ(ã)

where ã is the charge- parity state of the antiparticle. Clearly, a further operation by Ĉ
would return it back to the original charge parity state. Hence, we have

Ca2 = 1 =⇒ Ca = ±1

Ca is called the charge parity. Charge parity is distinct from the charge of the particle. If
an interaction conserves C, then [H, C] = 0. Charge-parity is conserved in strong inter-
action but is violated in weak interaction. Let Q be the charge operator with eigenstate
| ψq i, i.e.
Q | ψq i = q | ψq i
We then have
CQ | ψq i = qC | ψq i = q | ψ−q i
However,
QC | ψq i = Q | ψ−q i = −q | ψ−q i
Thus [Q, C] 6= 0 unless q = 0. This shows that the charge operator is distinct from the
charge parity.
We will not discuss the charge conjugation in any more detail. It is covered in a particle
physics course more rigorously.

18