Physics 215 Winter 2018

Implications of Time-Reversal Symmetry in Quantum Mechanics

1. The time reversal operator is antiunitary

In quantum mechanics, the time reversal operator Θ acting on a state produces a state
that evolves backwards in time. That is, if we consider the time evolution of a state under
the assumption that the Hamiltonian is time-independent,
|Ψ(t)i = e−iHt/~ |Ψ(0)i , (1)
then
Θ |Ψ(−t)i = e−iHt/~ Θ |Ψ(0)i . (2)
Letting t → −t in eq. (1) and comparing with eq. (2) yields
e−iHt/~ Θ = ΘeiHt/~ .
Taking t infinitesimal yields
−iHΘ = ΘiH . (3)
Suppose that Θ were a linear operator. Then, one could cancel the factors of i in eq. (2) to
obtain −HΘ = ΘH. Acting on an energy eigenstate |En i, we would conclude that
HΘ |En i = −ΘH |En i = −En Θ |En i .
which implies that if |En i is a state of energy En then Θ |En i is a state of energy −En . This
result would then imply the absence of a ground state, since one could generate a state of
arbitrarily negative energy by choosing a state with arbitrary large positive energy.
We can avoid this dilemma by declaring Θ to be an antilinear operator. We first note two
important properties of antilinear operators. Given an antilinear operator Θ and two states
|Ψi and |Φi, then

∗
hΦ| Θ |Ψi = hΦ| Θ |Ψi . (4)
Second, for any complex constant c and antilinear operator Θ,
Θc |Ψi = c∗ Θ |Ψi . (5)
Since Θ is a symmetry operator, we also demand that it should preserve the absolute value
of any inner product. That is, Θ is an antiunitary operator, which in addition to satisfying
eqs. (4) and (5), also satisfies,
hΘΨ | ΘΦi = hΨ | Φi∗ = hΦ | Ψi . (6)
Applying eq. (5) to eq. (3) then yields
 
H, Θ = 0 . (7)
Consequently, |En i and Θ |En i have the same energy (as expected for a symmetry operator Θ),
and the existence of a ground state is preserved.

1
 The complex conjugated one-component wave function satisfies the Schrodinger equation
with t → −t. It follows that
Θ = UK , (8)
where K is the complex conjugate operator and U is an arbitrary phase. When applied to a
(2j + 1)-component wave function that describes a particle of spin j, U is a (2j + 1) × (2j + 1)
unitary matrix. The complex conjugate operator is antiunitary. In particular, in light of
eq. (6),
hKΨ | KΦi = hΨ | Φi∗ = hΦ | Ψi .
Hence, the time-reversal operator Θ defined by eq. (8) is antiunitary as expected.1

2. The time reversal operator acting on states of definite angular momentum

Since the angular momentum changes sign under time reversal, the quantum mechanical
angular momentum operator J~ must satisfy
Θ J~ Θ−1 = −J~ . (9)
We shall use this property to obtain an explicit form for the unitary operator U that appears
in eq. (8) when acting on the |jmi basis,
By definition, any vector operator V ~ satisfies
X
U † [R(n̂, θ)]Vi U[R(n̂, θ)] = Rij Vj , (10)
j

where U[R(n̂, θ)] = exp(−iθn̂· J~/~) is the unitary operator that rotates states of the Hilbert
space. We choose V ~ = J~ and consider a rotation parameterized by n̂ = ŷ and θ = π. The
corresponding 3 × 3 rotation matrix R is given by [cf. eq. (20) of the class handout entitled
Three Dimensional Rotation Matrices]:
 
−1 0 0
R(ŷ, π) =  0 1 0 .
0 0 −1
It follows from eq. (10) that
exp(iπJy /~)Jx exp(−iπJy /~) = −Jx , (11)
exp(iπJy /~)Jy exp(−iπJy /~) = Jy , (12)
exp(iπJy /~)Jz exp(−iπJy /~) = −Jz . (13)
In class, we evaluated the matrix elements of the angular momentum operators Ji with
respect to the |j mi–basis. In particular, we found that the matrix elements of Jx and Jz are
real and those of Jy are pure imaginary. In light of eq. (5), it follows that
K −1 Jx K = Jx , K −1 Jy K = −Jy , K −1 Jz K = Jz . (14)
1
For further details, see Section 4.4 of Sakurai and Napolitano.

2
Thus, if we multiply eqs. (11)–(13) on the left by η −1 K −1 and on the right by ηK and then
make use of eq. (14), it then follows that eq. (9) holds, where

Θ = η exp(−iπJy /~)K , (15)

is an antiunitary operator and and η is a complex phase that can be chosen by convention to
be unity.
We can confirm eq. (15) in more detail by examining the effect of operating the time reversal
2
operator Θ on a simultaneous eigenstate of J~ and Jz , denoted by |j mi. It is convenient to
rewrite eq. (9) by multiplying from the right by Θ−1 and the left by Θ, which yields

Θ−1 J~ Θ = −J~ . (16)

It follows that
2 2
Θ−1 J~ Θ = Θ−1 J~ Θ · Θ−1 J~ Θ = J~ . (17)
Using eq. (16),
Θ−1 Jz Θ |j mi = −Jz |j mi = −~m |j mi .
Multiplying both sides by Θ yields
   
Jz Θ |j mi = −~m Θ |j mi . (18)

Likewise, using eq. (17),

2 2
Θ−1 J~ Θ |j mi = J~ |j mi = ~j(j + 1) |j mi .

Multiplying both sides by Θ yields

2
J~ Θ |j mi = ~2 j(j + 1) Θ |j mi .
  
(19)

The results of eqs. (18) and (19) imply that

Θ |j mi = c |j , −mi , (20)

where c is some complex number. Using eq. (6) with |Ψi = |φi = |j mi, eq. (20) implies that
|c|2 = 1. Thus, c is a complex phase, which we can write as c = eiδ . In principle, δ can depend
on j and m, so that
Θ |j mi = eiδ(j,m) |j , −mi . (21)
To determine the functional form of δ(j, m), we examine the x and y components of eq. (16).
It is more convenient to write J± = Jx ± iJy and use eq. (16) to obtain

Θ−1 J± Θ = −J∓ , (22)

where the extra sign in eq. (22) arises as a result of eq. (5). In addition, recall that
 1/2
J± |j mi = ~ (j ∓ m)(j ± m + 1) |j, m ± 1i . (23)

3
Using eqs. (22) and (23),
1/2
Θ−1 J± Θ |j mi = −J∓ |j mi = −~ (j ± m)(j ∓ m + 1)

|j , m ∓ 1i .

Multiplying both sides of the above equation by Θ and using eq. (21),
1/2 iδ(j,m∓1)
eiδ(j,m) J± |j , −mi = −~ (j ± m)(j ∓ m + 1)

e |j , −m ± 1i . (24)

We again use eq. (23) to write the left hand side of eq. (24) as
1/2
eiδ(j,m) ~ (j ± m)(j ∓ m + 1)

|j , −m ± 1i .

Inserting this result into eq. (24) then yields eiδ(j,m) = −eiδ(j,m∓1) . Given the value of δ(j, j),
one can obtain δ(j, m) for m = −j , −j + 1 , . . . , j − 1 , j,

eiδ(j,j−n) = (−1)n eiδ(j,j) , for n = 0, 1, 2, . . . , 2j .

Thus, writing n = j − m, it follows that

eiδ(j,m) = (−1)j−m eiδ(j,j) , for m = −j , −j + 1 , . . . , j − 1 , j .

Hence, we conclude that

Θ |j mi = eiδ(j,j) (−1)j−m |j , −mi ,
where eiδ(j,j) is an irrelevant j-dependent (but m-independent) phase that can be set to 1 by
convention.2 Henceforth, we shall write

Θ |j mi = (−1)j−m |j , −mi . (25)

Finally, we can compare the result of eq. (25) with eq. (13) of the class handout entitled,
Properties of the Wigner d-matrices,

exp(−iπJy /~) |j mi = (−1)j−m |j , −mi . (26)

2
Since the complex conjugation operator K commutes3 with J± , Jz and J~ , it follows that
K |j mi = η |j mi, where η is a phase that could depend on j but is independent of m. By
convention, we can choose η = 1, in which case K |j mi = |j mi. Consequently,

exp(−iπJy /~)K |j mi = (−1)j−m |j , −mi , (27)

and we can identify the antiunitary operator Θ = exp(−iπJy /~)K, since they have the exact
same matrix elements in the |j mi–basis [cf. eqs. (25) and (27)].
2
In some cases other choices for eiδ(j,j) can be more convenient. For example, if eiδ(j,j) = (−1)j then
Θ |j mi = (−1)2j−m |j , −mi = (−1)2(j−m) (−1)m |j , −mi = (−1)m |j , mi, where we have used the fact that
j − m is an integer. This convention is convenient when applied to orbital angular momentum, since the
spherical harmonics satisfy Yℓm (θ, φ)∗ = (−1)m Yℓ,−m (θ, φ), which implies that the effect of time reversal on
Yℓm (θ, φ) ≡ hθ, φ | ℓ mi is equivalent to complex conjugation with no additional phase factors.
3
This result follows from eq. (14). In particular, K −1 J± K = J± since J± = Jx ± iJy are real matrices in
the |j mi–basis.

4
3. The square of the time reversal operator

Starting from eq. (25), we apply Θ twice to obtain

Θ2 |j mi = Θ Θ |j mi = Θ(−1)j−m |j , −mi
= (−1)j−m Θ |j , −mi = (−1)j−m (−1)j+m |j mi = (−1)2j |j mi .

Note that the step Θ(−1)j−m = (−1)j−m Θ is valid because (−1)j−m is a real number for both
integral and half-integral values of j, since in either case j − m is an integer.4 Hence, we
conclude that
Θ2 |jmi = (−1)2j |jmi . (28)
One can also derive eq. (28) by directly employing eq. (15). First, we note that the
antiunitary complex conjugation operator satisfies,

K = K † = K −1 ,

which yields K 2 = I, where I is the identity operator. Moreover, in light of eq. (5), one
can write Kz = z ∗ K, where z is any complex number. Consequently, K commutes with
exp(−iπJy /~) since iJy is a real matrix with respect to the |j mi basis. Starting with eq. (15),
it then follows that

Θ2 = Θ Θ = η exp(−iπJy /~)Kη exp(−iπJy /~)K

= ηη ∗ exp(−iπJy /~)K 2 exp(−iπJy /~) = exp(−2iπJy /~) ,

after noting that ηη ∗ = 1 for any complex phase. Finally,

Θ2 |j mi = exp(−2iπJy /~) |j mi = exp(−iπJy /~) exp(−iπJy /~) |j mi

= (−1)j−m exp(−iπJy /~) |j , −mi = (−1)j−m (−1)j+m |j mi = (−1)2j |j mi .

That is, we have recovered eq. (28).

Note that eq. (28) is equivalent to the relation,
(j)
dm′ m (2π) = (−1)2j δmm′ , (29)

which is derived in the class handout, Properties of the Wigner d-matrices. Eq. (29) implies
that for bosonic systems (with integer values of j), a rotation by 2π is equivalent to the identity
operator, whereas for fermionic systems (with half-odd-integer values of j), a rotation by 2π
is equivalent to the negative of the identity operator (which implies that one must rotate by
4π to recover the initial fermionic system).
4
Had we adopted the phase convention of footnote 2, where Θ |j mi = (−1)m |j , −mi = i2m |j , −mi, then

Θ2 |j mi = Θ Θ |j mi = Θ i2m |j , −mi = (−i)2m Θ |j , −mi = (−i)2m i−2m |j mi

= (−i)4m |j mi = (−1)−2m |j mi = (−1)2j (−1)−2(j+m) |j mi = (−1)2j |j mi ,

after noting the relation Θz = z ∗ Θ for any complex number z, and using the fact that j + m is an integer in
the final step. The end result coincides with that of eq. (28), independently of the phase convention.

5
4. Even and odd irreducible tensor operators with respect to time reversal

An irreducible tensor operator is defined to be even or odd under time reversal if,
(k)
ΘTq(k) Θ−1 = ±(−1)q T−q , (30)

where even [odd] corresponds to the plus [minus] sign, respectively.

It is more convenient in the analysis that follows to rewrite eq. (30) as
(k)
Tq(k) = ±(−1)q Θ−1 T−q Θ . (31)

Consider the matrix element of this operator equation with respect to the states |α j mi. Since
Θ is antiunitary (so that Θ−1 = Θ† ),
  h  i∗
(k) (k)
hα′ j m′ | Θ−1 T−q Θ |α j mi = hα′ j m′ | Θ† T−q Θ |α j mi ,

after employing eq. (4). Hence, using eq. (31),

h  i∗
(k)
±(−1)q hα′ j ′ m′ | Tq(k) |α j mi = hα′ j m′ | Θ† T−q Θ |α j mi ,

If the theory is time-reversal invariant, then the Hamiltonian H commutes with the time
reversal operator Θ. As a result, the energy eigenstate |En i and the state Θ |En i have the same
energy eigenvalue, as previously noted below eq. (7). Assuming that En is a non-degenerate
state,5 then Θ |En i = eiβ |En i for some complex phase eiβ . More generally, suppose that
2
the maximal set of simultaneous diagonalizable operators is {H, J~ , Jz }, and perhaps an
additional set of operators A. The corresponding simultaneous eigenvalues shall be denoted
collectively by |α j mi, where the α represent the energy eigenvalue and eigenvalues of the
operators A if present. For a time-reversal invariant theory H is even under time reversal
since Θ−1 HΘ = H. Likewise, we shall assume that the operators A are also even under time
reversal so that Θ−1 AΘ = A. In this case, Θ |α j mi does not change the quantum numbers α.
In particular, eq. (25) takes on the more general form,6

Θ |α j mi = (−1)j−m |α j , −mi ,
′
which implies that hα′ j ′ m′ | Θ† = (−1)j−m hα′ j , −m′ |. It then follows that
(k)
±(−1)q hα′ j ′ m′ | Tq(k) |α j mi = (−1)j−m (−1)j−m hα′ j , −m′ | T−q |α j , −mi∗ .
′
(32)

Applying the Wigner-Eckart theorem,

hα′ j m| Tq(k) |α j mi = hjk ; mq | jk ; jm′ ihα′ jkT (k) kα ji ,

5
Even in the case of degeneracies, one can form appropriate linear combinations within the degenerate
subspace such that each energy eigenstate is a simultaneous eigenstate of Θ.
6
In a theory where the time-reversal symmetry is violated, we could only conclude that
X
Θ |α j mi = (−1)j−m cα′ |α′ j , −mi ,
α′

for some suitable coefficients cα′ .

6
to eq. (32), we obtain:
′
hjk ; mq | jk ; jm′ ihα′ jkT (k) kα ji = ±(−1)q+2j−m−m hjk ; −m , −q | jk ; j , −m′ ihα′ jkT (k) kα ji∗ .

Since the Clebsch-Gordan coefficients vanish unless m′ = q + m, it follows that

′
(−1)q+2j−m−m = (−1)2(j−m) = 1 , (33)

since j − m is an integer. Hence,

hjk ; mq | lk ; jm′ ihα′ jkT (k) kα ji = ±hjk ; −m , −q | jk ; j , −m′ ihα′ jkT (k) kα ji∗ . (34)

Finally, if we employ the following symmetry relation of the Clebsch-Gordan coefficients in

eq. (34),7
hjk ; −m , −q | jk ; j , −m′ i = (−1)k hjk ; mq | jk ; jm′ i ,
then it follows that for a time-reversal invariant theory,

hα′ jkT (k) kα ji = ±(−1)k hα′ jkT (k) kα ji∗ . (35)

That is, the reduced matrix elements of a time-reversal even or odd irreducible tensor operator
are either purely real or purely imaginary depending on the sign of ±(−1)k . Eq. (35) is an
important result, which will be used in the next section of these notes.

5. A nonzero electric dipole moment of the neutron implies that time reversal
symmetry is broken

The electric dipole operator is ρ ~ In this section, we shall prove that if a neutron
~ = eX.
is observed to have a nonzero electric dipole moment, then both the parity and time reversal
symmetries are separately violated.
Denote the parity operator by P. Then,
~ P = −X
P −1 X ~, (36)

~ → −X
since X ~ under an inversion of the coordinate system. The parity operator satisfies,

P = P −1 = P † ,

so that P 2 = I. If parity is conserved, then P commutes with the Hamiltonian H. Hence, the
neutron would be an eigenstate of parity. Since P 2 = I, the possible eigenvalues of P are ±1.
Denoting the neutron state by |ni, we must have P |ni = ± |ni. (Conventionally, the positive
sign is chosen, although the following argument does not depend on this choice.) It follows
that
~ |ni = e hn| P −1 XP
e hn| X ~ |ni = −e hn| X
~ |ni = 0 ,
7
See, e.g. eq. (15.2.11) of R. Shankar, Principles of Quantum Mechanics, 2nd edition (Springer Science, New
York, NY, 1994). Note that we are employing the notation of Sakurai and Napolitano for the Clebsch-Gordan
coefficients.

7
where P |ni = ± |ni has been employed at the first step and eq. (36) was used in the second
step. Hence, if parity is conserved,
~ |ni = 0 .
hn| ρ |ni = e hn| X

That is, a non-zero electric dipole moment indicates a violation of the parity symmetry of the
theory.
Next, consider a time-reversal invariant theory. Note that the operator ρ = eX ~ is even
under time-reversal since
ΘX~ Θ−1 = X ~.
Thus, we can apply eq. (35) with k = 1, which yields

hα jkρkα ji = −hα jkρkα ji∗ . (37)

However, ρ is an hermitian operator, which implies that its diagonal elements are real. Since
the Clebsch-Gordan coefficients are real (by convention), the Wigner-Eckart theorem implies
that hα jkρkα ji is also real. In light of eq. (37), it follows that

hα jkρkα ji = 0 . (38)

Since the neutron has spin- 12 , it follows that the neutron (in its rest frame) is an eigenstate of
2
J~ with j = 21 . Using the Wigner-Eckart theorem, eq. (38) implies that

hn| ρ |ni ∝ hα jkρkα ji = 0 .

Hence, a non-zero static electric dipole moment indicates a violation of the time-reversal
symmetry of the theory.

6. Final remarks

Both parity and time-reversal are good symmetries of quantum electrodynamics and ap-
pear to also be symmetries of the strong interactions. The weak interactions maximally violate
parity symmetry, but time-reversal symmetry is only very weakly violated by the weak in-
teractions. Indeed, the Standard Model of particle physics does predict the existence of an
electric dipole moment of the neutron, but the predicted value is many orders of magnitude
below the currently observed limits. Thus, if a neutron electric dipole moment were to be
measured and shown to be nonzero, it would be strongly suggestive of the existence of new
physics beyond the Standard Model of particle physics.
It is a curious fact that the strong interactions appear to be parity and time-reversal in-
variant. It turns out that there is a dimensionless parameter of the theory of of the strong
interactions (called quantum chromodynamics), denoted by θ, which if non-zero would be a
signal of a violation of both parity and time-reversal symmetries. The absence of an observed
electric dipole moment for the neutron implies that |θ| < 10−10 . You might think that θ = 0
exactly, but since the parity and time-reversal symmetries are violated by the weak interac-
tions, there is no justification for setting θ = 0 by hand. It is currently a great mystery as to

8
why the θ parameter of the strong interactions is so small (this is called the strong CP problem
since in quantum field theory, a violation of time reversal symmetry implies a simultaneous
violation of the charge conjugation and parity symmetries). Clever solutions to the strong
CP problem lead to new physics beyond the Standard Model. One such solution requires the
existence of a new extremely light spin-0 particle called the axion. Dedicated experiments
searching for axions are currently running. So far, no evidence for the existence of axions has
been found.