Some Notes On "Modern Quantum Mechanics," by Sakurai and Napoletano
Some Notes On "Modern Quantum Mechanics," by Sakurai and Napoletano
and Napoletano
Maarten Golterman,
These notes spell out some details of the arguments in the textbook
Ref. [1]. They do not replace the textbook, so they should be studied
along with the textbook.
then
[A(H) (t), B (H) (t)] = [U (t)† A(S) U (t), U † (t)B (S) U (t)] (1.16)
† (S) † (S) † (S) † (S)
= U (t)A U (t)U (t)B U (t) − U (t)B U (t)U (t)A U (t)
= U † (t) A(S) B (S) − B (S) a(S) U (t)
= U † (t)C (S) U (t) = C (H) (t) ,
2
which is the same commutation rule as Eq. (1.15) with all Heisenberg operators at
the same time t.
1 ρ2 1 ρ2 y
Ax = − B0 a sin ϕ = − B0 2 a 2 , (1.23)
2 ρ 2 x +y
2
1 ρa 1 ρ2 x
Ay = B0 cos ϕ = B0 2 a 2 , (1.24)
2 ρ 2 x +y
Az = 0 , (1.25)
3
hence (check!)
~ =0
B (outside cylinder) . (1.26)
where Rij are the components of a 3 × 3 matrix representing the rotation and ri0 are
the transformed components. Requiring the invariance of the inner product:
in index notation:
0 0 T
r1i r2i = (Ri jr1j )(Rik r2k = r1j Rji Rik r2k = r1j r2j , (2.3)
and using that this has to be true for any pair ~r1 , ~r2 , we find that
RT R = 1 , (2.4)
T
with 1 the 3 × 3 unit matrix. In index notation, we need that Rji Rik = δjk , which
is Eq. (2.4) in index notation. Equation (2.4) implies that R is orthogonal (unitary
and real). This group is denoted as O(3): the group of orthogonal 3 × 3 matrices. If
we also require, as we will do in this chapter, that det(R) = 1, it is the group SO(3)
of “special” orthogonal 3 × 3 matrices.
All elements of SO(3) can be written in the form
ϕ) = eϕi Ti ,
R(~ (2.5)
0 0 0 0 0 1 0 −1 0
Tx = 0 0 −1 , Ty = 0 0 0 , Tz = 1 0 0 ,
0 1 0 −1 0 0 0 0 0
with Tx,y,z the generators of the group. Indeed, SO(3) is a group: it is associative
(because matrix multiplication is associative), there is a unit element R(~ ϕ = 0) = 1,
−ϕi Ti
each element R(~ ϕ) has an inverse R(−~ ϕ) = e (rotate around the same axis by
the opposite angle), and it is closed under multiplication (this means that the product
of two rotations is also a rotation, i.e., that the product of two elements of the group
1
Of course, they also leave the inner product between two momentum vectors, etc., invariant.
4
is an element of the group). Closure is not trivial, and we will return to that below.
First, let us work out an example:
which is a rotation around the z-axis by an angle ϕ. A general rotation with param-
~ is a rotation around an axis with direction n̂ = ϕ̂ over an angle ϕ = |~
eters ϕ ϕ|.
To prove closure, we need the following two ingredients. First, it is straightforward
to check that
(This can be written succinctly as [Ti , Tj ] = ijk Tk .) Second, we need the Baker–
Campbell–Hausdorff formula:
1 1
eA eB = eA+B+ 2 [A,B]+ 12 ([A,[A,B]]+[[A,B],B])+... . (2.8)
There is no closed-form expression, but one can prove that all terms in the exponent on
the right-hand side are “commutators of commutators,” like the ones explicitly shown.
Now, if we apply this to A = ϕi Ti and B = ψi Ti for two rotations R(~ ϕ) = eϕi Ti and
~ = eψi Ti , the exponent on the right-hand side of Eq. (2.8) becomes
R(ψ)
1 ~ i , (2.9)
(ϕi + ψi )Ti + ijk ϕi ψj Tk + . . . = (ϕi + ψi + jki ϕj ψk + . . .)Ti ≡ αi (~
ϕ, ψ)T
2
and the product of R(~ ϕ) and R(ψ) ~ is thus a rotation R(~ α), with α~ a complicated
function of ϕ ~
~ and ψ. We see that if we know the commutation rules of the generators
Ti , we know how elements of the group multiply. In other words, the algebra (2.7)
defines the structure of the group SO(3). Note that, if [A, B] = 0, Eq. (2.8) collapses
to eA eB = eA+B . For rotations, this happens when we consider two rotations around
the same rotation axis. A group for which all generators commute is an abelian group.
An example of an abelian group is the translation group: recall that translations are
generated by the momenta pi , and we have that [pi , pj ] = 0.
In fact, it is sufficient to consider only infinitesimal rotations, i.e., rotations over
an infinitesimal angle ϕ = |~ ϕ|. If we define = ϕ/N with N very large, so that is
very small, we have that
N
ϕ
N
Rz (ϕ) = lim R() = lim 1 + Tz = eϕTz . (2.10)
N →∞ N →∞ N
5
A group like this is called a Lie group, and the corresponding algebra (2.7) a Lie
algebra.
Now, let us see how this all works in quantum mechanics. In the discussion,
note the similarity to the case of translations! The only thing that makes the case
of rotations more complicated (and more interesting!) is that rotations in general do
not commute, while translations always do.
If a system is represented by a ket |αi, the rotated system is represented by some
different ket
|αiR = D(R)|αi , (2.11)
which defines a linear operator D(R) for each rotation R. Note that, of course, D(R)
is represented by a matrix on the Hilbert space (all operators are), but the dimension
of D(R) is equal to the dimension of the Hilbert space, which is not (necessarily)
equal to 3. The operators D(R) have to follow the same rules as the rotations R
themselves; in jargon, the D(R) form a representation of the group SO(3). There is
a homeomorphism R → D(R) such that
In words: if I first rotate a physical system, described by a ket |αi, by R2 , and then I
rotate it again by R1 , the ket describing this system should transform in accordance
with that.
Here is a trivial example: it could be that the physical system represented by |αi
does not change at all under rotations, which means that for such a system, for all R,
D(R) = 1! Clearly, Eq. (2.12) is satisfied, and this one-dimensional representation is
called the trivial representation. Referring to Eq. (3.1.15) of Ref. [1], for such a state
~
we have that J|αi = 0. In other words, such a state has zero angular momentum.
A physical example is the hydrogen atom in the ground state: you may recall that
the ”electron cloud” in the ground state looks spherically symmetric, so that indeed
rotations have no effect.
It is straightforward to work out that in terms of the variables u, v and z, Eq. (4.2)
takes the form
u0 = ueiϕ , (4.4)
z0 = z .
v 0 = ve−iϕ ,
The expression on the right-hand side is just Eq. (4.4) in matrix form, but it is also
what you get if you work out eiJz ϕ/h̄ in matrix form, so the middle expression and the
expression on the right-hand side of Eq. (4.5) coincide! Therefore, in this form, the
rotation around the z-axis is indeed generated by iJz /h̄, so the generators iJz /h̄ and
Tz only differ by a (unitary) basis transformation. This can be proven more generally
(i.e., that a basis transformation exists that maps the Ti into the iJi /h̄), and thus
the two 3-dimensional representations are equivalent. Mathematically, we say there
is only one 3-dimensional representation. Of course, nothing prevents nature from
making use of this representation in different physical applications, such as rotations
in physical 3-dimensional space, but also the existence of a 3-dimensional Hilbert space
corresponding to the angular momentum states of a particle with angular momentum
j = 1. Problem 3.14 of Ref. [1] asks you to work out the complete proof: a basis
transformation exists that takes the set ih̄Tx,y,z into the set Jx,y,z , respectively.
Since rotations only change the polar angles, and not r, we can express the operator
Lx in terms of derivatives with respect to φ and θ, instead of derivatives with respect
to y 0 and z 0 . Assume, therefore, that
!
h̄ ∂ ∂
Lx = A +B . (5.2)
i ∂θ ∂φ
The task is to find A and B. For Lz this was done in Ref. [1], where it was found that
in that case A = 0, and B = 1. Now, because a rotation around the x-axis changes
both θ and φ, both A and B will be non-zero, and more complicated. Start from
!
∂ ∂
1 − δϕA − δϕB hr, θ, φ|αi = hr, θ − δϕA, φ − δϕB|αi , (5.3)
∂θ ∂φ
where we transitioned to write the wave function hx0 , y 0 , z 0 |αi in polar coordinates.
Comparing with Eq. (5.1), and remembering the relation between the cartesian and
polar coordinates, this gives us the equations
From the last of these equations, we find that A = − sin φ, and substituting this into
the first or second equation, we then find that B = − cot θ cos φ (check both these
9
equations!). We thus arrive at
!
h̄ ∂ ∂
Lx = − sin φ − cot θ cos φ , (5.5)
i ∂θ ∂φ
which is Eq. (3.6.11) in position representation. A very similar calculation gives the
expression for Ly in Eq. (3.6.12).
~ = r̂ ∂ + θ̂ 1 ∂ + φ̂ 1 ∂ .
∇ (5.6)
∂r r ∂θ r sin θ ∂φ
With ~r = rr̂, we see that
~ =r ∂ ,
~r · ∇ (5.7)
∂r
because, of course, r̂ · θ̂ = r̂ · φ̂ = 0.
It makes sense to consider this ket, because for a spherically symmetric problem, the
total energy and angular momentum eigen state (in position space) h~x 0 |n, `, mi =
hr, θ, φ|n, `, mi in Eq. (3.6.22) factorizes into two parts, with one only depending on
r, and the other only on the polar angles θ and φ.
6.1. Definitions
Suppose we have a system with two different angular momentum operators, J~1 and J~2 .
Examples are the electron in the hydrogen atom, which has both an orbital angular
momentum J~1 = L ~ and a spin J~2 = S,
~ or a system with two electrons, each with their
own spin, J~1 = S~1 and a spin J~2 = S~2 . Since these two angular momenta operators
are unrelated to each other (i.e., they correspond to different, unrelated, observables
of the system), they commute:
[J1i , J2j ] = 0 , (6.1)
10
for all components i and j of both operators. Formally, the operators J~1 and J~2 act
on different Hilbert spaces. In the example of the hydrogen atom, L ~ acts on the
(angular part) of the spatial wave function, while S ~ acts on the spin of the electron.
To get a description of all properties of the electron, we can combine these two Hilbert
spaces: using a position eigenbasis for the spatial properties of the electron, and the
Sz eigenkets | ± 21 i for the spin of the electron, a combined basis is |~x 0 i ⊗ | ± 21 i, where
hopefully the use of the “direct product” symbol ⊗ will become clear in what follows.
If, instead of position eigenkets |~x 0 i, we only consider orbital angular momentum
eigenkets in position space, we may use the orbital angular momentum eigenkets |`, mi
(here we are ignoring any quantum numbers not related to angular momentum, i.e.,
quantum numbers which do not change under rotations), and a basis for the electron’s
“combined” Hilbert space is
1 1
|`, m; ± i = |`, mi ⊗ | ± i , (6.2)
2 2
where we show two equivalent ways of writing states in this combined Hilbert space.
(We say that the combined Hilbert space is the direct product of the two individual
Hilbert spaces.) In general, for a system with two different angular momenta J~1 and
J~2 , using an eigenbasis for each, the combined eigenbasis of the product Hilbert space
consists of the eigenkets
Since J~1 acts only on the states |j1 , m1 i and J~2 only on the states |j2 , m2 i, the notation
on the right in Eq. (6.3) makes sense if instead of J~1 we write J~1 ⊗ 1, and instead of
J~2 we write 1 ⊗ J~2 , with2
The notation with the ⊗ symbol exhibits the product structure of this Hilbert space
very clearly, but often, in practice, the notation on the left-hand side of Eq. (6.4) is
used. The notation with the ⊗ also makes it very clear that the two angular momenta
commute:
[J~1 ⊗ 1, 1 ⊗ J~2 ] = [J~1 , 1] ⊗ [1, J~2 ] = 0 . (6.5)
With a little less mathematical precision we can also write that [J~1 , J~2 ] = 0, which is
how we think about it in physics, but the notation with the direct product symbol
clarifies the mathematical structure of these operators more clearly. Note that this
is all just notation to give a mathematically precise description of the physics we are
interested in!
2
For the states on the left of this equation I follow the notation of Ref. [1], in which first the j,
and then the m quantum numbers are listed.
11
We can now define a total angular momentum, which acts on the product Hilbert
space,
J~ = J~1 ⊗ 1 + 1 ⊗ J~2 , (6.6)
or, more succinctly, J~ = J~1 + J~2 . First, this is an angular momentum: because
using [J1i , J1j ] = ih̄ijk J1k and [J2i , J2j ] = ih̄ijk J2k , it is straightforward to show
that [Ji , Jj ] = ih̄ijk Jk (show this!). Since the components Ji satisfy the angular
momentum algebra, it is an angular momentum by definition. Of course, it also
makes sense intuitively that the sum of two angular momenta is again an angular
momentum.
The basis |j1 , j2 ; m1 , m2 i = |j1 , m1 i ⊗ |j2 , m2 i is composed of simultaneous eigen-
kets of J~12 , J1z , J~22 and J2z . But, one can, of course, also try to find a basis composed
of eigenkets of J~2 and Jz . However, such eigenkets are not the same as the eigenkets
|j1 , j2 ; m1 , m2 i, because J1z and J2z do not commute with J~2 (because they do not
commute with J~1 · J~2 ; show this!), so we cannot find simultaneous eigenkets of J~2 ,
Jz , J1z and J2z ! In contrast, we can find simultaneous eigenkets of J~2 , Jz , J~12 and J~22 ,
because these operators all commute with each other (again, show this). We can thus
construct a basis for our Hilbert space either using the eigenkets |j1 , j2 ; m1 , m2 i, or
simultaneous eigenkets |j1 , j2 ; j, mi of the commuting set J~2 , Jz , J~12 and J~22 , with
Since these two bases span the same Hilbert space, there should be a unitary basis
transformation relating the two, which is written in Eq. (3.8.33) of Ref. [1]. The
elements of this basis transformation are hj1 , j2 ; m1 , m2 |j1 , j1 ; j, mi, and the task of
Sec. 3.8 is the construction of these elements, the Clebsch–Gordan coefficients, given
two values of j1 and j2 .
An example where the new basis is useful is the case of a spin-orbit coupling L· ~ S,
~
~ ~ ~ ~
where in this example L = J1 and S = J2 . Since
~ = 1 J~2 − 1 L
~ 2 − 1S~ 2 = 1 J~2 − J~2 − J~2 ,
~ ·S
L 1 2 (6.8)
2 2 2 2
~ ·S
the new basis eigenkets |j1 , j2 ; j, mi are eigenstates of L ~ = J~1 · J~2 :
1
J~1 · J~2 |j1 , j2 ; j, mi = h̄2 j(j + 1) − j1 (j1 + 1) − j2 (j2 + 1) |j1 , j2 ; j, mi . (6.9)
2
In contrast, the states |j1 , j2 ; m1 , m2 i are not eigenstates of J~1 · J~2 = J1x J2x + J1y J2y +
J1z J2z . They are eigenstates of the last term in this sum, J1z J2z , but not of the other
two terms (convince yourself!).
12
Suppose we have a basis of simultaneous eigenkets |j1 , j2 ; m1 , m2 i of two angular
momenta, J~1 and J~2 :
In what follows, to avoid cluttered notation, we will use units in which h̄ = 1. Since
Jz = J1z + J2z ,
Jz |j1 , j2 ; m1 , m2 i = (m1 + m2 )|j1 , j2 ; m1 , m2 i , (6.12)
and therefore, that
This state, with j = j1 +j2 is the unique state with the maximal Jz eigenvalue possible,
and thus is the highest m-value component of a multiplet with angular momentum
j = j1 +j2 . The largest possible values of j and m are thus j = j1 +j2 and m = j1 +j2 .
We find the other componets of this multiplet by acting with the lowering operator
J− = J1− + J2− :
1 1
√ J− |j1 , j2 ; j, ji = √ (J1− + J2− )|j1 , j2 ; j, ji (6.14)
2j 2j
1
= √ (J1− + J2− )|j1 , j2 ; j1 , j2 i
2j
1
q q
= √ 2j1 |j1 , j2 ; j1 − 1, j2 i + 2j2 |j1 , j2 ; j1 , j2 − 1i ,
2j
where we used that
q
J± |j, mi = (j ∓ m)(j ± m + 1)|j, m ± 1i . (6.15)
Acting with J− on the state (6.17) will also give a linear combination of the three states
|j1 , j2 ; j1 −2, j2 i, |j1 , j2 ; j1 −1, j2 −1i and |j1 , j2 ; j1 , j2 −2i, but it will be a different linear
combination, orthogonal to the state (6.19); this is the state |j1 , j2 ; j − 1, m = j − 2i.
Excercise: Why? Work this out! But, since these two states are linear combinations
of three states, there is a third independent linear combination of these states, which
is orthogonal to both |j1 , j2 ; j, j − 2i and |j1 , j2 ; j − 1, j − 2i. This new state will be
the maximal m-value component of a multiplet with j = j1 + j2 − 2! It is thus the
state |j1 , j2 ; j − 2, j − 2i. Exercise: Convince yourself ! Of course, the minimal value
to which we can lower m1 is −j1 , and the minimal value to which we can lower m2 is
−j2 ; acting with Ji− on a state with mi = −ji will give zero (i = 1, 2).
We see that m1 or m2 get lowered each time we apply J− = J1− +J2− ; in Eq. (6.19)
to minimally m1 = j1 − 2 or m2 = j2 − 2. Continuing with the tower of states we are
14
obtaining starting with Eq. (6.13), i.e., our first multiplet, we find, ultimately, the
state
|j1 , j2 ; j, −ji = |j1 , j2 ; j, m = −j = −j1 − j2 i = |j1 , j2 ; m1 = −j1 , m2 = −j2 i . (6.20)
One way to see that this is where we have to end up is that we could have started with
this state, and obtain the same tower of states by acting with the raising operator
J+ = J1+ + J2+ .
Collecting what we know thus far, we find J~2 , Jz multiplets with j = j1 + j2 ,
j = j1 + j2 − 1, j = j1 + j2 − 2, etc. Where does this end, i.e., what is the lowest j
value we will find, carrying out this procedure to the end? In constructing our first
(j = j1 + j2 ) multiplet, we started with one state from the collection in Eq. (6.10) (cf.
Eq. (6.13)), we then found a linear combination of two of them (cf. Eq. (6.16)), and
then a linear combination of three of them (cf. Eq. (6.19)). Each time we laddered
down one step we found one more state to show up in the linear combination, and
each time we also found a new multiplet: this is how we found the multiplets with
j = j1 + j2 − 1 and j = j1 + j2 − 2. Now, without loss of generality, take j1 ≥ j2 .
Then, the number of states in these “growing” linear combinations stops growing
once we have encountered the state |j1 , j2 ; m1 = j1 , m2 = −j2 i, because states with
m2 = −j2 −1 do not exist. (Note how the m2 quantum number in the last state in the
linear combinations in Eqs. (6.16) and (6.19) gets lowered by one unit in each step.)
That means that the “latest” new multiplet we found, by systematically carrying out
our procedure is the multiplet with j = j1 − j2 (which is non-negative, because we
chose j1 ≥ j2 ). When m2 = −j2 is reached in our lowering process, no new states
from the collection (6.10) will show up in our linear combinations anymore, so all
states with m1 and m2 such that m = m1 + m2 = j1 − j2 will have been used to
complete multiplets we already were constructing. Therefore, the possible j values
for the total angular momentum J~ = J~1 + J~2 are
j = j1 + j2 , j1 + j2 − 1 , ... , |j1 − j2 | , (6.21)
where we dropped the restriction j1 ≥ j2 . Since the two bases (6.10) and (6.11) have
to consist of an equal number of states, a simple counting argument, given in Ref. [1],
confirms that we found all states in the new basis (6.11). Problem 3.24 of Ref. [1]
asks you to work this out for the case j1 = 1, j2 = 1.
15
so also δij is invariant under rotations, and thus the whole first term is invariant under
rotations. In more technical terms, this part transforms in the j = 0 representation of
the rotation group (i.e., it does not transform at all, just like the angular momentum
eigenstate |j = 0, m = 0i!). Note that δij , a two-index object, does not transform at
all either — such a tensor is called an invariant tensor.
Now let us consider the second term. Applying a rotation to it:
X
Ui Vj − Uj Vi = Ui Vj − Vi Uj → (Rii0 Rjj 0 Ui0 Vj 0 − Rii0 Rjj 0 Vi0 Uj 0 ) (7.2)
i0 j 0
X
= Rii0 Rjj 0 (Ui0 Vj 0 − Vi0 Uj 0 ) .
i0 j 0
We see that the anti-symmetrized product of Ui and Vj transform among each other,
and not into the other two terms in Eq. (3.11.13). In fact, we recognize the compo-
~ × V~ , which has three components, and transforms like a
nents of the cross product U
vector. Thus, the three anti-symmetrized products in the second term of Eq. (3.11.13)
transform in the j = 1 representation of the rotation group.
Likewise, the symmetrized products in the third term of Eq. (3.11.13) also trans-
form among themselves:
X
Ui Vj + Uj Vi = Ui Vj + Vi Uj → (Rii0 Rjj 0 Ui0 Vj 0 + Rii0 Rjj 0 Vi0 Uj 0 ) (7.3)
i0 j 0
X
= Rii0 Rjj 0 (Ui0 Vj 0 + Vi0 Uj 0 ) .
i0 j 0
Now, note that if we set i = j in the third term, and we sum over i, we get
X 1 1~ ~ X
(Ui Vi + Ui Vi ) − U · V δii = Ui Vi = 0 , (7.4)
i 2 3 i
so the third term is traceless, and the “33” component of this combination is thus
equal to minus the sum of the “11” and “22” components. Therefore, there are only
five independent components in the third term of Eq. (3.11.13). They all transform
into each other, in a five dimensional representation of the rotation group; thus, they
transform in the j = 2 representation. These j = 0, j = 1 and j = 2 tensors are
examples of spherical tensors.
The steps leading from Eq. (3.11.18) to Eq. (3.11.22b) in Ref. [1] should be quite
straightforward; note that Eq. (3.11.18) is the same equation as Eq. (3.5.49), but now
16
we consider orbital angular momentum, so we relabel j → `, and instead of taking
the rotation to be R, we take it to be R−1 (which is also a rotation of course, so we
are free to do that).
Note that Eq. (3.11.22a) is a definition, inspired by Eq. (3.11.21): it defines a
spherical tensor Tq(k) by requiring it to transform in the same way as Ykq . Note the
(annoying!) choice of making the label k ↔ ` the superscript on T , even though
it is the subscript on Y , and likewise, the label q ↔ m is a subscript on T , but a
superscript on Y !
Then, since according to Eq. (3.11.4)
D(R) = 1 − i J~ · n̂ , (7.6)
h̄
the signs in Eq. (3.11.23) would be different if you derive it directly from Eq. (3.11.22a):
k
(k)
1 − i J~ · n̂ Tq(k) 1 + i J~ · n̂ = Tq0 hkq 0 | 1 − i J~ · n̂ |kqi .
X
(7.7)
h̄ h̄ q 0 =−k
h̄
This still leads to Eq. (3.11.24), as it should, of course. From there it should also be
straightforward to derive Eq.(3.11.25), following the “directions” given in Ref. [1], if
in Eqs. (3.5.35b) and (3.5.41) you substitute j → k and m → k.
Here, we work out an example. We take j1 = j2 = 21 , so that j can take the values 0
and 1. Taking the lower sign everywhere in Eq. (7.8) for j = 1, m = 0, m1 = 12 and
m2 = 12 yields:
√ √ 1 1 1 1 1 1 √ 1 1 1 1 1 1
* + * + * +
1 1 1 1 1 1
2 ; , , ; 1, 1 = 1 ; − , ; 1, 0 + 1 ; , − ; 1, 0 ,
2 2 2 22 2 2 2 2 22 2 2 2 2 22 2
(7.9)
1 1
while taking the lower sign, j = 1, m = 1, m1 = 2 and m2 = − 2 yields:
√ √ 1 1 3 1 1 1 √ 1 1 1 1 1 1
* + * + * +
1 1 1 1 1 1
2 ; , − , ; 1, 0 = 0 ; , − ; 1, 1 + 1 ; , ; 1, 1 .
2 2 2 22 2 2 2 2 22 2 2 2 2 22 2
(7.10)
Note that, of course, m1 cannot be equal to 3/2, and indeed, that coefficient is
multiplied by zero in Eq. (7.10). Now take
* +
1 1 1 1 1 1
; , ; 1, 1 = 1 . (7.11)
2 2 2 22 2
17
This happens to be the correct value (why?), but since the recursion relation (7.8) is
homogeneous, we have to fix one coefficient to find all the others. Substituting this
in Eq. (7.10), we find * +
1 1 1 1 1 1 1
; , − , ; 1, 0 = √ , (7.12)
22 2 2 2 2 2
and substituting both Eq. (7.11) and Eq. (7.12) into Eq. (7.9) then yields
* +
1 1 1 1 1 1 1
; − , , ; 1, 0 = √ . (7.13)
22 2 2 2 2 2
Note that Eq. (7.11) is equivalent to
+ +
1 1 1 1 1 1
; 1, 1 =
; , , (7.14)
2 2 2 2 2 2
These are just equations (3.8.15a) and (3.8.15b) of Ref. [1], in more “pedantic” no-
tation. It is straightforward to find other Clebsch–Gordan coefficients with j = 1,
by choosing other values of m, m1 and m2 in Eq. (7.8) (note that we always should
choose m = m1 + m2 − 1 for the “upper” equation, and m = m1 + m2 + 1 for the
“lower” equation—why?).
Let us consider also the case j = 0, which is a separate set of recursion relations,
because rotations do not mix different j values. If j = 0 then also m = 0, and the
left-hand side of Eq. (7.8) will now vanish. Choosing m1 = 12 and m2 = 21 in the
“upper” equation thus yields:
√ √ √ 1 1 1 1 1 1
* + * +
1 1 1 1 1 1
0= 1 ; − , ; 0, 0 + 1 ; , − ; 0, 0 , (7.16)
2 2 2 22 2 2 2 2 22 2
and thus * + * +
1 1 1 1 1 1 1 1 1 1 1 1
; − , ; 0, 0 = − ; , − ; 0, 0 . (7.17)
2 2 2 22 2 2 2 2 22 2
We can choose the righ-hand side of this equation to be equal to a constant c, and
that fixes the other j = 0 Clebsch–Gordan coefficient. We thus find that
+ + +!
1 1 1 1 1 1 1 1 1 1
; 0, 0 =c ; ,− − ; − , , (7.18)
2 2 2 2 2 2 22 2 2
√
which is nothing else than Eq. (3.8.15d) of Ref. [1] if we choose c = 1/ 2. Try finding
the remaining equation, Eq. (3.8.15c) from the recursion relations!
because only the term m = n in the sum contributes. Thus, |En i is a parity eigenstate,
with eigenvalue pn . Since the only possible eigenvalues of π are ±1, we thus conclude
that energy eigenstates are parity even (pn = +1), or parity odd (pn = −1). An
example is the one-dimensional harmonic oscillator. Note that non-degeneracy of the
spectrum of H is important!
Consider a hamiltonian
H = H0 + λV , (9.1)
, where we have solved the eigenvalue problem for H0 , i.e., we know the eigenvalues
En(0) and eigenkets |n(0) i, with
We do not know the eigenvalues En and eigenkets |ni of the full hamiltonian, and we
want to approximate them in perturbation theory in λ, which we will assume to be a
small parameter.
We expand
where we already divided by λ. This is a vector equation, with the vectors kets in our
Hilbert space. Therefore, let us consider the components of this vector equation on
19
the basis of normalized energy eigenkets of the unperturbed hamiltonian H0 , {|k (0) i}.
Taking first |k (0) i = |n(0) i, we find, taking the inner product with |n(0) i,
Using that hn(0) |H0 = hn(0) |En(0) two terms cancel between the left- and right-hand
sides, and we find a very simple expression for the first-order energy shift:
where we divided by λ2 . Proceeding as we did before, we first take the inner product
with |n(0) i; using again that hn(0) |H0 = hn(0) |En(0) the terms with |n(2) i cancel again,
and we find
∆(2) (0) (1)
n = hn |V |n i . (9.12)
Here we used that, with choosing cn (λ) = 1, |n(1) i does not have a component along
|n(0) i, i.e., hn(0) |n(1) i = 0 (so we made a clever choice for cn (λ)!). Substituting the sum
in Eq. (9.10), which equals |n(1) i into this, we get our final result for the second-order
energy shift:
1
∆(2) hn(0) |V |k (0) ihk (0) |V |n(0) i .
X
n = (0) (0)
(9.13)
k6=n En − Ek
hk (0) |V |n(1) i + hk (0) |H0 |n(2) i = ∆(1) (0) (1) (0) (0) (2)
n hk |n i + En hk |n i , (9.15)
(0)
and using again that hk (0) |H0 = hn(0) |Ek and hk (0) |n(0) i = 0 we obtain
1
hk (0) |n(2) i = (0) (0)
hk (0) |V |n(1) i − ∆(1) (0) (1)
n hk |n i (9.16)
En − Ek
1 1
hk (0) |V |`(0) ih`(0) |V |n(0) i
X
= (0) (0) (0) (0)
En − Ek `6=n En − E`
1
− (0) (0)
hn(0) |V |n(0) ihk (0) |V |n(0) i
(En − Ek )2
1 X Vk` V`n Vkn Vnn
= (0) (0) (0) (0)
− (0) (0)
.
En − Ek `6=n En − E` (En − Ek )2
Again, we see that the component hn(0) |n(2) i does not follow from our calculation,
but, recall that we chose hn(0) |ni = 1 for the full solution, to all orders in λ.
Let us now address this latter point. We want the state |ni to be normalized,
but the state we constructed actually is not normalized! So, let us define
q
|niN = Zn |ni , (9.17)
for some real number Zn > 0 such √ that |niN is normalized, N hn|niN = 1. Observing
that hn(0) |ni = 1, we see that Zn = hn(0) |niN , so that the probability to find |niN
in the state |n(0) i is equal to Zn .
Using Eq. (9.17) we can calculate Zn from 1 = N hn|niN = Zn hn|ni:
where in the last step we used that |n(1) i and |n(2) i were cleverly chosen to be or-
thogonal to |n(0) i. Substituting |n(1) i into this, we find (prove this!)
|Vkn |2
Zn = 1 − λ2 + O(λ3 ) .
X
(0) (0)
(9.19)
k6=n (En − E k )2
Here we review a of of the technical details of Sec. 5.2 of Ref. [1], considering first
the projectors needed to project on the degenerate subspace D and its “complement”
(which contains all states orthogonal to any of the states inside D), and then a few
calculational details of the example of the linear Stark effect.
21
10.1. Projectors
In Sec. 5.2 of Ref. [1], the projector P0 onto the subspace D spanned by the degenerate
H0 eigenstates is defined as
g
X (0) (0)
P0 = |mi ihmi | , (10.1)
i=1
P1 = 1 − P0 (10.3)
g
X (0) (0) X (0) (0)
= |Ek ihEk | − |mi ihmi |
k i=1
X (0) (0)
= |Ek ihEk | . (10.4)
k∈{m
/ 1 ,...,mg }
We have that
g X
g
(0) (0) (0) (0)
P02 =
X
|mi ihmi |mj ihmj | (10.5)
i=1 j=1
g X g g X
g
X (0) (0) X (0) (0)
= |mi iδij hmj | = |mi ihmi | = P0 ,
i=1 j=1 i=1 j=1
The two matrix elements in Eq. (5.2.19) are complex conjugates of each other (since
the operator V is hermitian), so we need to calculate only one of them:
~
h2s|V |2p, m = 0i = −e|E|h2s|z|2p, m = 0i (10.9)
22
~
= −e|E|h2s|r cos θ|2p, m = 0i
Z ∞ Z 1 Z 2π
~
= −e|E| drr2 d(cos θ) dφ
0 −1
h i0∗ h i
× R20 (r)(Y00 )(θ, φ) r cos θ R21 (r)Y10 (θ, φ)
Z ∞ Z 1
~
= −2πe|E| drr 3 ∗
R20 (r)R21 (r) d(cos θ)(Y00 )∗ (θφ) cos θY10 (θ, φ)
0 −1
Z ∞
1 r r
~
= −e|E| drr 2− 3
√ e−r/a0
3
0 8a0 a0 3a
s 0
Z 1
1 3
× 2π d(cos θ) √ cos2 θ
−1 4π 4π
√
!
~
1 ~ 0,
= −e|E| 3 3a0 × √ = 3e|E|a
3
where a0 is the Bohr radius, we used Eqs. (B.5.7) and (B.6.7) of Ref. [1], and we used
that Z ∞
dx xn e−x = Γ(n + 1) = n! (10.10)
0
after substituting y = r/a0 . Therefore, the matrix (5.2.18) of Ref. [1] is equal to
0 1 0 0
1 0 0 0
P0 V P0 = ~ 0
3e|E|a , (10.11)
0 0 0 0
0 0 0 0
23
The eigenstates for the first two eigenvalues are, respectively,
1 1
√ (|2si + |2p, m = 0i) , √ (|2si − |2p, m = 0i) . (10.15)
2 2
(Check this!) See Eqs. (5.2.20) and (5.2.21), as well as Fig. 5.1 in Ref. [1].
For perturbation theory to work, we want the energy shifts to be small compared
to the differences between energy levels of the unperturbed states. The smallest
energy difference for the n = 2 energy level is that with the n = 3 energy level
e2 e2 5e2
E3 − E2 = − + = . (10.16)
18a0 8a0 72a0
~ 0 5e2 ~ 5e2
We thus want 3e|E|a 72a0
, or |E| 216a20
≈ 2 × 10−19 V/m.
m = −j, −j + 1, . . . , j − 1, j .
| {z }
2j + 1 states
References
[1] J. J. Sakurai and J. Napoletano, Modern Quantum Mechanics, 2nd edition,
Addison-Wesley, 2011
24