TFY4250/FY2045 Tillegg 12 - Magnetic moments.

Spin

TILLEGG 12

12 Magnetic moments. Spin

The hypothesis about the electron spin (intrinsic angular momentum) was published already in 1925 by Uhlenbeck and Goudsmit, almost at the same time as
the discovery by Heisenberg and Schrodinger of quantum mechanics and Paulis
formulation of the exclusion principle. The main experimental clues leading to
the spin hypothesis were
the fine structure of optical spectra
the Zeeman effect
the SternGerlach experiment
All these effects involve the behaviour of a particle with spin and/or orbital
angular momentum in a magnetic field.
In section 12.1 in this Tillegg, we begin by considering the magnetic moment due to the orbital motion of a particle, both classically and quantummechanically. We then review the experiment of Stern and Gerlach, which clearly
reveals the intrinsic angular momentum (spin) of the electron. We also give a
survey of the spins of other particles.
In section 12.2 we establish a formalism for spin 12 , based on the general discussion
of angular momenta in Tillegg 11.

12.1 Magnetic moments connected with orbital angular

momentum and spin
12.1.a Classical magnetic moment
(Hemmer p 178, 1.5 in B&J)
First a little bit of classical electrodynamics.

TFY4250/FY2045 Tillegg 12 - Magnetic moments. Spin

In electromagnetic theory (see e.g. D.J. Griffiths, Introduction to Electrodynamics, chapter

6) one learns that an infinitesimal current loop, placed in a magnetic field, experiences a
torque
=B
(T12.1)
and a force
F = (B).

(T12.2)

Here is the magnetic (dipole) moment of the infinitesimal current loop. For a planar
loop encircling an area A this magnetic moment is
IA,
= I An
is the unit vector perpendicular to the loop plane.
where I is the current and n
The same effect is observed when a magnetic needle is placed in a megnetic field.

The needle experiences a torque

() = B sin =

(B cos ) = (B).

(T12.3)

Note that this relation is analogous to Fx (x) = V /x. Equations (T12.2) and (T12.3)
show that the magnetic moment placed in the magnetic field B corresponds to the interaction
energy
E = B.

(T12.4)

This potential energy is by choice equal to zero when E is perpendicular to B, and

is maximal (and positive) when and B are antiparallell. The energy has its lowest value
when is parallell to B; this is the orientation preferred by the needle.

Another example: The figure shows a particle with mass m and charge q (< 0) which is
kept moving in a classical circular orbit by a central field V (r). This constitutes a current

TFY4250/FY2045 Tillegg 12 - Magnetic moments. Spin

loop. With the radius r, velocity v and revolution frequency = v/(2r) the current is
I = q. This results in a magnetic moment
|| = I A = q

v
r2 = 12 qrv = 21 q |r v|.
2r

Thus the magnetic moment due to the motion of the charge is proportional to the orbital
angular momentum L = m r v :
L =

q
L.
2m

(T12.5)

The ratio q/2m between these two quantities is known as the gyromagnetic ratio.
Let us add that for this kind of classical motion the angular momentum L and the
magnetic moment L are not constants of motion when B differs from zeo. According to
Newtons 2. law and equation (T12.1) we have that
q
dL
= = L B =
B L L L.
dt
2m

(T12.6)

From this equation it follows that L (and hence L ) precess. The precession frequency,
L

q
B,
2m

(T12.7)

is known as the Larmor frquency.

With these classical considerations in mind, let us see what quantum mechanics has to say
about these things.

12.1.b Magnetic moment due to orbital motion quantum-mechanical

treatment
From the classical relation (T12.5) we see that the quantum-mechanical operator corresponding to the observable L (the magnetic moment due to the orbital motion) must be
bL =

q b
q
b
L=
rp
2m
2m

(T12.8)

TFY4250/FY2045 Tillegg 12 - Magnetic moments. Spin

for a particle with mass m and charge q. As an example we may consider the doughnut
state
211 = (64a50 )1/2 r er/2a0 sin ei
b with eigenvalue h
for the hydrogen atom. This is an eigenstate of L
. According to (T12.8)
z
it is then also an eigenstate of
e b
b L )z =
(
Lz ,
2me
with eigenvalue e
h/(2me ). (In this example we neglect the difference between the electron
mass and the reduced mass.)

It is instructive to cosider this example more closely. The probability distribution

resembles a doughnut and is of course time independent, as is the case for all
stationary states,
1
r2 er/a0 sin2 .
211 =
64a50
Inside this doughnut there is a probability current. This can be calculated from
the formula for the probability-current density:
"

j = <e
;
im

r
=e

1
1

+e
+e
.
r
r
r sin

Here we observe that the components of the gradient in the r- and directions
give imaginary contributions to the expression inside the brackets, and hence
no contribution to j. Thus the probability in the doughnut is flowing in the
direction. We find that
h
/me
1
h

211
.
j=e
r er/a0 sin = e
5
64a0
me r sin
Defining a velocity by the relation j = 211 v, we can calculate the local velocity
of the probability current:
v=

|j|
h

1
a0
=
= c
.
211
me r sin
r sin

Here, r sin the distance from the z axis is of course typically of the order of
the Bohr radius a0 , so that the velocity is of the order of c, as could be expected.
Note, however, that this velocity is variable; the probability distribution is not
rotating around the z axis as a solid doughnut.
The probability d3 r in the volume element d3 r corresponds to a momentum
contribution d3 r mv = mjd3 r and an electric-current contribution ejd3 r.
Since h p i is real, the former also follows from
hpi =

d3 r = m
i

"

Z
h

<e
d3 r = mj d3 r.
im

The expectation values of Lz and (L )z can be calculated in a similar manner,

)z = r sin , we have
and the results are as expected: Because (r e
h Lz i211 =
=

211

r
i
3

rme j sin d r =

211 d3 r =
zZ

(r me j)z d3 r

h
211 d3 r = h
,

TFY4250/FY2045 Tillegg 12 - Magnetic moments. Spin

and
h (L )z i211 =

1
2

e
hZ
eh
[r (ej)]z d r =
211 d3 r =
, q.e.d.
2me
2me
3

This example illustrates the connection between the orbital angular momentum
and the magnetic moment of a moving particle.
As illustrated by this example, the operator relation (T12.8) implies that the magnetic
moment L due to the orbital motion is quantized in the same way as the orbital angular
momentum. This means that the size and one of the components can have sharp values
simultaneously. For an electron (with q = e) this implies that the size can take the values
|L | =

q
e
|L| = B l(l + 1) ;
2me

l = 0, 1, 2, ,

(T12.9)

while for example the z component can take the values

(L )z =

e
Lz = mB ;
2me

m = 0, 1, 2, , l.

(T12.10)

Here the quantity

e
h
(1 Bohr magneton)
(T12.11)
2me
is the natural unit for the magnetic moment for the elctron, just as h
is the natural unit for
5
angular momenta. [1 B = 5.788 10 eV/T(esla).]
We note also that the quantization of Lz and (L )z corresponds to a so-called space
quantization, that is, quantized values of the angle between the vector L and the z axis.
B

Note also that because L and L are not classical constants of motion in a magnetic
field, they also are not quantum-mechanical constants of motion. With this statement
we mean that the expectation values h L i and h L i are not constant. These vector-valued
expectation values will in fact precess in the same manner as the classical quantities.

12.1.c The SternGerlach experiment and the electron spin

(Section 4.4 in Griffiths, 1.5 in B&J.)
In a homogeneous magnetic field the magnetic moment experiences a torque = B;
the force F = (B) then is equal to zero, and the magnetic moment only precesses
around B, as discussed above. In 1921 Stern and Gerlachs had the following idea:

TFY4250/FY2045 Tillegg 12 - Magnetic moments. Spin

Let us try to measure the magnetic moment of an atom by sending a beam of these atoms
(collimated by a slit) through an inhomogeneous magnetic field in order to measure the
deflection. If the beam is positioned such that it passes through a vertical field, it should be
deflected vertically (in the z direction). We can show this by assuming an inhomogeneous
field given by
B = ax
ex (B0 + az)
ez ,
where a = Bz /z is a suitable constant. You can easily check that this field is divergence
free (B = 0), as is the case for all magnetic fields. The force becomes
x z e
z ).
F = (B) = (x ax z (B0 + az)) = a(x e

(T12.12)

Here we must now remember that precesses rapidly around the B field, that is, about
the z axis. The x-component x therefore averages to zero and gives no deflection in the
x direction. Thus we get a deflection in the positive or negative z direction, depending on
the force Fz = az . The measurement of the deflection therefore is a measurement of z .
(Note that z is constant during the precession.)
In 1921 (before quantum mechanics) Stern and Gerlach expected that the directions of
the magnetic moments || of the atoms entering the magnetic field should be randomly
distributed, corresponding to a continuous variation of z between || and +||. They
were hoping to measure the maximal deflection up and down (corresponding to = 0 and
= ), making it possible to calculate ||, using the velocity of the atoms, the atomic mass,
the length of the magnet, and the parameter a.
After 1925, quantum mechanics has taught us that a measurement of z must give one
of its eigenvalues (according to the measurement postulate). Since the direction of and
hence z are quantized according to (T12.10), we must therefore expect to observe a quantized deflection. If the total angular momentum of the atom is integral (given by an
integral angular-momentum quantum number l), we should then expect to find 2l + 1 discrete deflections, that is, an odd number of pictures of the slit on the screen in the figure
above.
The experiment, however, showed something else: Stern and Gerlach (1922) used a gas
of silver atoms (Z = 47) heated in a furnace. The result was two distinct pictures of the slit
on the screen.

TFY4250/FY2045 Tillegg 12 - Magnetic moments. Spin

A similar experiment with hydrogen atoms, conducted by Phipps and Taylor in 1927, also
resulted in two pictures of the slit.
Spin and magnetic moment of the electron
(8.3.3 in Hemmer)
The explanation was found by Uhlenbeck and Goudsmit in 1925, and is most easily
understood in the hydrogen case: Even with l = 0 the electron in the hydrogen atom has
a magnetic moment S , which causes the deflection of the orbit of each atom. This magnetic
moment is connected with an intrinsic angular momentum of the electron, the so-called spin
angular momentum S, simply called the spin. As all other angular momenta, the spin
can be characterized by
q an angular-momentum quantum number which is usually denoted
by s, so that |S| = h
s(s + 1), and such that the z component can take the values
Sz = ms h
,

where ms = s, s + 1, s + 2, , +s.

This is analogous to m = Lz /
h taking the values l, l + 1, l + 2, , +l for a
given orbital angular-momentum quantum number l. From the general discussion of angular
momenta in Tillegg 11, it follows that a spin quantum number must in general take one of
the values s = 0, 21 , 1, 23 , 2, . The number of ms values is 2s + 1, in analogy with the
2l + 1 values of Lz .
The fact that we observe two discrete deflections in the SternGerlach experiment then
leads to the interpretation that in this case 2s + 1 is equal to 2; that is, the spin quantum
number s of the electron is equal to 12 . Thus
the electron
has spin one half, as we use
q
q
to express it. This corresponds to |S| = h
s(s + 1) = h
3/4 = 0.866 h
. The two possible
values of the magnetic quantum number of the electron spin then are ms = 21 , corre . These two spin states are commonly denoted by spin up and
sponding to Sz = 12 h
spin down.

Experiments show that the intrinsic magnetic moment connected with the spin is
S = ge

e
S.
2me

(T12.13)

TFY4250/FY2045 Tillegg 12 - Magnetic moments. Spin

Here the factors in front of S is the product of the gyromagnetic ratio we found for the
orbital motion, and ge which is a dimensionless factor. Very accurate measurements show
that this factor is
ge = 2 1.001159652187 (4)

(gyromagnetic factor of the electron),

(T12.14)

where the uncertainty (4) is in the last digit.

To sum up, we may state that the two discrete deflections observed in the SG-experiment
are due to the fact that there are only two possible values of z , Sz and Fz :
ms = + 21 :

Sz = + 12 h
,

ms = 12 :

(S )z = 21 ge

Sz = 21 h
,

e
h
= 12 ge B ,
2me

(S )z = + 21 ge B ,

Fz = az = 21 ge aB ,

Fz = az = 21 ge aB .

With a > 0 we see that the upper beam emerging from the SG magnet corresponds to
measuring Sz = + 12 h
(spin up), while the lower beam corresponds to Sz = 21 h
(spin
1
down).

12.1.d Spins of other particles

(8.3 in Hemmer)
The proton and the neutron
What about the proton, which is the nucleus of the hydrogen atom doesnt also this
particle have spin 21 and a magnetic moment? The answer is yes, but it turns out that that
the magnetic moment of the proton is much smaller than that of the electron;
e
p = 5.59
Sp
(gp = 5.59).
(T12.15)
2mp
Thus, for the proton the natural unit for the magnetic moment is
N
1

e
h
= 3.1524515 108 eV/T(esla)
2mp

(1 nuclear magneton).

(T12.16)

With hydrogen atoms or silver atoms from a furnace, the SG experiment gives a fifty-fifty distributions
of atoms in the upper and lower beams; the probabilities of measuring spin up or spin down are both equal to
1
2 . (This holds also if the SG magnet is turned so that another component of the spin is measured.) One then
says that the beam emerging from the furnace is unpolarized. On the other hand, after the passage through
the inhomogeneous field, the atoms on the upper beam are all left in the state spin up, they constitute
an ensemble of completely spin-polarized atoms. Thus a SternGerlach device, characterized essentially by
the direction of the inhomogeneous field, can be used both to measure a spin component and to prepare an
ensemble in a well-defined spin state.

TFY4250/FY2045 Tillegg 12 - Magnetic moments. Spin

We note that the gyromagnetic factor gp = 5.59 for the proton deviates quite much from the
factor ge 2 for the electron. This has to do with the fact that the proton is a composite
particle, consisting of two u-quarks with charge q = 2e/3 and one d-quark with q = e/3.
The neutron is also a composite particle, consisting of one u-quark and two d-quarks, and
then it is perhaps not so surprising that it has a non-zero magnetic moment,
n = 3.83

e
Sp
2mp

(gn = 3.83),

(T12.17)

even if it is a neutral particle.

It is important to note that p and n are a factor 1000 smaller than e . Therefore
the behaviour of atoms in a magnetic field is largely determined by the electrons. 2
Why do we measure only two discrete deflections also for the silver atoms? The answer
lies in the fact that 46 out of the 47 silver electrons are in a quantum-mechanical state in
which both the total orbital momentum and the total spin are equal to zero. Both the orbital
and the spin angular momenta then are determined by electron number 47. This electron
is in an s-state, with l = 0. Thus the total magnetic moment of the atom simply is the
intrinsic magnetic moment of electron no 47 (as for a free electron). (Here we may neglect a
possible magnetic moment of the nucleus, because this will anyhow be about a factor 1000
smaller than that of the electron.)
Spins of other particles
The electron and its antiparticle, the positron, are only two of several so-called leptons
(particles that do not interact strongly), which all have spin 21 . Among these are the muons
(with mass m = 105.66 MeV/c2 ) and (m = 1777 MeV/c2 ), which in a way are
heavier relatives of the electron. Thus the g factor of the muon is also very close to 2:
= g
g = 2 1.0011659160 (6)

e
S ,
2m

(gyromagnetic factor of the muon).

(T12.18)
(T12.19)

Related to these charged leptons there is a set of almost massless neutral leptons which
interact only through the weak force. These are the electron neutrino e and the electron
anti-neutrino e . Related to the muon there is a corresponding pair ( and ), and
related to the tau lepton there is and . All these neutrinos are spin-one-half particles.
The photon is somewhat special, because the component of the spin along the direction
of motion can only have the two values h, corresponding to spin 1. This is due to the fact
that the photon is massless. Related to the photon are the massive vector bosons, W
(MW 80.4 MeV/c2 ) and Z 0 (MZ 91.2 MeV/c2 ), which have spin 1, and which mediate
the elektro-weak interactions together with the photon. Spin 1 is also found in the gluons,
which mediate the strong forces. The graviton, the carrier of the gravitational force, is
believed to have spin 2.
All the remaining particle species are so-called hadrons (which all interact strongly).
These can be divided into two groups, one of which is the baryons (p, n, , , ), which
2

There are exceptions: In nuclear magnetic resonance (NMR) it is an important point that the coupling
term B between the magnetic field and the magnetic moments of the nuclei are so small that they
correspond to radio-frequency energies.

TFY4250/FY2045 Tillegg 12 - Magnetic moments. Spin

10

are all three-quark systems. Because the quarks have spin 21 , it follows from the rules for the
addition of angular momenta (which we shall treat later) that the baryons can only have
half-integral spins (1/2, 3/2, etc). The other group consists of the mesons (, K, , , ),
which are quark-antiquark systems, and which therefore can only have integral spins (e.g. 0
for the mesons and 1 for the mesons).

12.2 Formalism for spin

1
2

(Hemmer 8.3, Griffiths 4.4, B&J 6.78)

What the spin is not
Uhlenbeck and Goudsmit based their spin hypothesis (in 1925) on the classical notion of a
rotating electron, with a certain mass and charge distribution. Modern scattering experiments have shown, however, that the size of the electron, if it differs at all from zero, must be
smaller than 1018 m. It has also turned out to be impossible to construct a classical model
with a mass and charge distribution that reproduces the spin and the magnetic moment of
the electron. Thus the electron behaves as a point particle, and we have to state that the
spin and the magnetic moment (with ge 2) of this particle can not be understood as the
result of any kind of material rotation which can be pictured classically and which can
be described in terms of a wave equation and a wave function. The latter is only possible
for the orbital angular momentum, for which l can take only integer values, while the spin
quantum number s can also take half-integral values, depending on which particle we are
looking at. Note that for a given particle species s is completely fixed.
q Thus for the electron
b
the intrinsic angular momentum has no choice, it has to be |S| = h
3/4, in contrast to the
orbital angular momentum which can vary, even if it is quantized. Again we see that the
spin does not behave as we would expect for an ordinary rotational motion.

12.2.a Ladder of ket vectors for spin

1
2

Even if we do not quite understand what the spin is, we have a perfectly applicable theoretical model in the abstract ket-vector formalism for angular momenta which was developed
in Tillegg 11, in the sense that we can apply it to processes involving the spin and accurately
predict the outcome of experiments.
b = S,
b and with
The starting point is (T11.38) and (T11.39), which for the spin (with J
1
j = s = 2 ) take the form
b 2 | 1 , mi =
S
2

3 2
h

| 12 , mi,

Sbz | 12 , mi = h
m | 21 , mi ;

m = 12 ,

The two eigenvectors are

| 21 , + 12 i |+ 21 i |+i |i

(spin up)

and
| 12 , 21 i | 21 i |i |i

(spin down).

(T12.20)

TFY4250/FY2045 Tillegg 12 - Magnetic moments. Spin

11

Note that we may choose the labels as we wish. Thus we may use e.g. only m or only the sign
of m; without stating explicitly that s = 21 . These two vectors form a basis of orthogonal
and normalized vectors; they are the only eigenvectors of the Hermitian operator Sbz . Thus
the state space of this half-integral spin is a two-dimensional vector space, and is spanned
by a ladder with only two rungs:

The ladder operators Sb+ and Sb take us up and down in the ladder: According to (T11.52)
we have (with j = s = 12 )
q

Sb |mi = h
( 12 m)( 32 m) |m 1i,

m = 12 ,

or explicitly:
Sb+ |+ 21 i = 0 ;

Sb+ | 12 i = h
|+ 12 i ;

Sb | 21 i = 0 ;

Sb |+ 21 i = h
| 12 i .

(T12.21)

From these formulae we see that

Sb+2 | 21 i = 0

and

Sb2 | 12 i = 0.

(T12.22)

The last two formulae are particular for spin 21 and are of course due to the fact that we
have a ladder with only two rungs.
Note that as long as we consider only the spin, that is, do not take into account the other
degrees of freedom for the particle, the two vectors |+ 12 i and | 12 i are a complete set; the
spin space is a two-dimensional sub-space of the Hilbert space. The completeness relation
of this set is (cf (T10.29))
X

|mihm| = |+ 12 ih+ 21 | + | 12 ih 21 | = 11.

(T12.23)

1
m=
2

Applying this unit operator we may expand an arbitrary vector |i in the two-dimensional
spin space:
|i =

X
m

|mihm|i =

X
m

a+ |+ 12 i + a | 21 i.

hm|i |mi = h+ 12 |i |+ 12 i + h 12 |i | 21 i

TFY4250/FY2045 Tillegg 12 - Magnetic moments. Spin

12

The normalization condition for the vector |i is



a+ h+ 21 | + a h 12 |

= |a+ |2 + |a |2 = 1.
(T12.24)
1
Note that the coefficient (or projection) a+ = h+ 2 |i is the probability amplitude of
measuring spin up (Sz = 12 h
) and leaving the spin in the state |+ 21 i. In the same manner,
a = h 12 |i is the amplitude of measuring spin down.
h|i h| |i =

a+ |+ 12 i + a | 12 i

12.2.b Matrix formulation. The Pauli matrices

(8.3 in Hemmer, 6.8 in B&J)
Ket- and bra-vectors are represented by column and row vectors
It is customary, both for spin 12 and higher angualar momenta, to use a matrix representation
of vectors and operators, with the 2j + 1 vectors |j, mi (where m goes from j to j) as
basis. In the spin- 21 case the general vector |i is then represented by the column matrix
!

h+ 12 |i
h 21 |i

a+
a

(T12.25)

which we denote simply by , and which is called a spinor. The basis vector|+ 12 i is represented by 3
!
!
1
h+ 12 |+ 12 i
=
+
(T12.26)
0
h 12 |+ 12 i
and the spin-down vector | 21 i is represented by
h+ 12 | 12 i
h 12 | 12 i

0
1

(T12.27)

These are known as the Pauli spinors, after Pauli, who invented the theory for spin
1925. We note that these spinors are orthonormal,

1
2

in

0  = 0  .
The adjoint of the ket vector |i = a+ |+ 21 i + a | 12 i, which is (|i) = h|, is represented by the adjoint matrix (transpose and complex conjugate), that is, by the row
matrix
= (a+ a ).
The normalization condition may then be written as
= (a+ a )

a+
a

= |a+ |2 + |a |2 = 1,

Note: In the literature it is not uncommon to use an equality sign between the vector |i and the column
matrix or spinor :
a
|i = ( + ).
a
The first equality sign should be read as represented by.

TFY4250/FY2045 Tillegg 12 - Magnetic moments. Spin

13

Operators are represented by 2 2-matrices

b 2 and S
b as a basis, these operators must be repSince we are using the eigenvectors of S
z
resented by diagonal matrices; cf the discussion of matrix mechanics in Tillegg 10 (section
10.4): From (T12.20) we find that
b 2 |mi = 3 h
2 m0 m
hm0 |S
4

and

hm0 |Sbz |mi = h

m m0 m ,

and

1
h

or
3 2

S = h
4
2

1 0
0 1

3 2
11
h
4

Sz =

1
0
0 1

z .
12 h

(T12.28)
!
++ +
Note that the matrix elements are numbered acording to the scheme
, and
+
that 11 here stands for the unit matrix. Note also that the diagonal elements are identical
to the eigenvalues, and hence to the possible results for measurements of S2 and Sz ; a
measurement of Sz can only give + 21 h
or 12 h
, no matter which state the spin is prior to the
measurement.
To find the matrices representing the remaining operators we take (T12.21) as the starting
point. Projecting these relations on |+ 12 i and | 12 i we get for example
(S+ )++ = h+ 12 |Sb+ |+ 21 i = 0

(S+ )+ = h+ 12 |Sb+ | 12 i = h
.

and

In this manner we find that the ladder operators Sb+ = Sbx + iSby and Sb = Sbx iSby are
represented by the matrices

S+ =

1
h

0 2
0 0

and

S =

1
h

0 0
2 0

(T12.29)

By adding and subtracting these we find at last the matrices for Sbx and Sby :

Sx =

1
h

0 1
1 0

1
h
x
2

and

Sy =

1
h

0 i
i
0

12 h
y .

(T12.30)

Here we observe that the matrices Sx and Sy become non-diagonal, which was of course to be
expected, since Sbx , Sby and Sbz do not commute, and since we have chosen the eigenvectors
of Sbz as our basis. To gain some experience with these matrices you should now check the
eigenvalue equations
Sz = 12 h
,
(T12.31)
which correspond to (T12.20), and also the relations
S+ + = 0,
which are (T12.21) on matrix form.

S+ = h
+ ,

etc,

TFY4250/FY2045 Tillegg 12 - Magnetic moments. Spin

14

A small exercise: Check that these matrices satisfy [Sx , Sy ] = ihSz , that is,
the angular-momentum algebra.
Another exercise: Show that the matrix squares Sx2 , Sy2 and Sz2 are equal.
What does this tell us about the possible eigenvalues of Sx and Sy , and hence
about possible measurement results?
Solution: If you square the matrices (T12.30), you will find that
Sx2

Sy2

h
2
= 11 = Sz2 .
4

(T12.32)

This implies that the eigenvalues of Sx and Sy are 21 h

, just as for Sz . Thus the possible
measured values for the x- and y-components of S are the same as for Sz . This should of
course be expected from symmetry considerations; we are free to choose the orientation of
our coordinate system, and then there can be no difference between the possible measured
values for Sx and Sz .
Pauli matrices. Rules of calculation
To simplify the notation it is customary to use the dimensionless matrices

x =

0 1
1 0

y =

0 i
i
0

and

z =

1
0
0 1

(T12.33)

which are called the Pauli matrices. The matrix representation of the spin operator may
then be written as
x x + e
y y + e
z z .
S = 21 h
;
e
The eigenvalue equation (T12.20) takes the form
z + = 1 + ,

z = 1 .

(T12.34)

For future use we include some rules of calculation for the Pauli matrices. The angularmomentum algebra [Sx , Sy ] = i
hSz takes the form

which can be written as

[x , y ] = 2iz , etc,

(T12.35)

i j j i = 2i ijk k .

(T12.36)

We use Einsteins summation convention: When a latin index as e.g. k occurs twice in a term, it
means that this index is summed over, from k = 1 to k = 3. Example:
ak bk

3
X

ak bk .

k=1

(If an index occurs more than twice in a term we have made a mistake).

TFY4250/FY2045 Tillegg 12 - Magnetic moments. Spin

15

Here we use Einsteins summatiuon convention. The Levi-Civita tensor is defined by

ijk =

1
1

for ijk = 123, 231, 312,

for ijk = 132, 321, 213,
otherwise.

(T12.37)

By direct calculation (or from (T12.32)) we see that

x2

y2

z2

1 0
0 1

11.

(T12.38)

It is also straightforward to show that the matrices anticommute;

x y + y x = 0, etc.
See the relations (T12.22) which imply that Sx and Sy satisfy the equation
Sx2 Sy2 + i(Sx Sy + Sy Sx ) = 0.
These relations, which can be collected in the formula
i j + j i = 2ij 11,

(T12.39)

are special for spin 21 .

By combining (T12.36) and (T12.39) we find that
i j = ij 11 + i ijk k .

(T12.40)

From this we can derive the following relation which holds for arbitrary vectors a and b:
(a)(b) = i ai j bj = ij ai bj 11 + i ijk ai bj k ,
that is,
(a)(b) = ab 11 + i (a b).

(T12.41)

(where n
is a unit vector):
In particular we have for a = b = n

(
n)(
n) = n
n 11 = 11.

(T12.42)

Note that the relations (T12.38) are special cases of this formula, and that it corresponds to
h
2
(S
n) =
4
2

1 0
0 1

h
2
= 11.
4

(T12.43)

This formula is a generalization of (T12.32), and had to be expected considering the symmetry argument above; no matter which component S
n of the spin we choose to measure,
the result must be 21 h
or 12 h
.

TFY4250/FY2045 Tillegg 12 - Magnetic moments. Spin

16

12.2.c The spin direction

(B&J p 311)
Even if the formalism above was based on abstract Hilbert vectors (without a wavefunction representation), we can use it to calculate all the relevant physical quantities which
can be measured experimentally. This means that we have a well-functioning physical theory
for the spin. Thus we can calculate e.g. the expectation value h Sz i with the following recipe
(which we may as usual call the sandwich or burger recipe):
h Sz i = h|Sbz |i = Sz .

(T12.44)

A small exercise: Check the last step above, using

h|Sbz |i = h| 11 Sbz 11 |i =

h|mihm|Sbz |nihn|i.

mn

[Hint: Set hm|i = am , so that h|mi = am .]

1
0

As an example, we may choose the state + =

, which gives

+ = 12 h
,
h Sz i = + Sz + = + 21 h
h Sx i =

h Sy i =

Thus, for the state + =

1
0

Sx + = (1,

0) 21 h

0 1
1 0

Sy + = (1,

0) 21 h

0 i
i
0

1
0

= 0,
1
0

= 0.

we find that
z .
h i = 12 h
e
h S i = 12 h

z z
This is a good reason to call this state spin up. The direction of h S i, which is h i = e
in this case, may be called the spin direction. Therefore we also use the notation
+ =

1
0

z ,

0
1

z ,

(T12.45)

where the direction h i is used as an index.

spin up: z
h i = z

spin down: z
h i = z.

h Sz i becomes equal to 12
h because + is an eigenspinor with sharp Sz = 12 h. That h Sx i and h Sy i

become equal to zero is not strange. It would be strange if h S i were to point in another direction than z
after the preparation of the state + .
5

TFY4250/FY2045 Tillegg 12 - Magnetic moments. Spin

17

These states are illustrated in the figure, which is equivalent to that used for the orbital
angular
q momentum in Tillegg 5. This picture is of limited value. It gives the correct |S|
, and also Sx2 + Sy2 = S2 Sz2 = 12 h
2 . The picture may also
(= h
3/4 ) and Sz = 12 h
(barely) remind us that the expectation values of Sx and Sy are equal to zero for both the
two states ,
h Sx i = h Sy i = 0.
The drawback of this illustration is that it can mislead us to believe that the direction of S
is an observable, which it is not, because the componentsqare not compatible; they can not
have sharp values simultaneously. The angle arccos( 12 / 34 ) = 54.7 or 125.3 between S
and the z axis thus is only another way to illustrate the
q fact that the two possible values
1
of Sz are 2 h
, compared with the fixed value |S| = h
34 . This is the same type of space
quantization that was found for the orbital angular momentum in Tillegg 5. When the
particle is sent through the SternGerlach magnet in section 12.1.c, the spin has to choose
one of these two states (no matter what state it is in before the measurement). This
choice determines whether the force on the magnetic moment acts upwards or downwards,
and decides whether the particle goes into the upper or lower beam.
However, a SternGerlach magnet can also be rotated. Suppose that we rotate it 90
degrees so that it measures Sx instead of Sz . What kind of states do we then get for the
ensembles in the upper and lower beam? According to the measurement postulate the
answer is that the spin must then choose between the eigenstates x of Sx .
a
b

A small exercise: Set =

Sx

and solve the eigenvalue problem

a
b

=e

a
b

where e is the eigenvalue.

(the same as for Sz ) and that the eigenstates
The solution is that the eigenvalues are 12 h
and
with spin directions x
x are respectively
1
x =
2

1
1

and

1
1

1
=
2

1
i

1
=
2

(T12.46)

(T12.47)

In the same manner one finds that

1
y =
2

1
i

and

with the same eigenvalues. In these spinors we have chosen the phases in such a way that
the upper components are real and positive.
General spin- 21 states
(Hemmer p 178, B&J p 309)
The physical rotation of the SternGerlach magnet by 90 degrees may be called an active
rotation. Consider now instead a so-called passive rotation, where the SternGerlach
device keeps the original physical orientation, while we imagine that the coordinate system

TFY4250/FY2045 Tillegg 12 - Magnetic moments. Spin

18

is rotated instead. The up direction of the device (the original z direction) in the new
coordinate system then becomes a unit vector
x
sin cos + y
sin sin + z cos ,
n

(T12.48)

characterized by the angles and :

In this coordinate system, the spin directions in the upper/lower beam are h i =
n; the
spin directions are physical and measurable and remain fixed together with the device.
Viewed from the new coordinate system we are now measuring the spin component S
n.
The two possible measured values are of course unchanged, and so are the physical states, but
the matrix representations of these states are changed, because the new coordinate system
corresponds to a change of basis; the Pauli spinors
1
0

+ =

and

0
1

now correspond to spin up and spin down related to the new z direction.

These matrix representations, which correspond to spin up and spin down in the n
direction, are determined by the eigenvalue equations
n
S
n n = 12 h

n n = n .

It is sufficient to solve the equation

n n = n . We set n =

(T12.49)
a
b

. This spinor

satisfies the eigenvalue equation

(
n 11) n = (x nx + y ny + z nz 11)
=

cos 1 sin ei
sin ei cos 1

a
b
a
b

= 0.

(T12.50)

As you can easily check, both the upper and the lower components of this equation are
satisfied when
(cos 1)a + sin ei b = 0,
that is, for
b
1 cos i sin 12 ei
=
e =
.
a
sin
cos 21

(T12.51)

This complex ratio between the lower and upper components is in fact all we need to know
about the spinor. Admittedly, the two complex numbers a and b contain four real parameters, but the normalization condition |a|2 + |b|2 = 1 reduces the number to three, and one

TFY4250/FY2045 Tillegg 12 - Magnetic moments. Spin

19

parameter may be allowed to be free, in the form of a common phase factor which is of no
physical significance. Explicitly we have from the normalization condition
 2

b
sin2 12
1
1
 
=1+  =1+
,
1 =
2
2


|a|
a
cos 2
cos2 12

that is,
a = ei cos 21 ,
where the phase can be chosen freely. Hence
b
b = a = ei sin 12 ei ,
a
so that
cos 12
sin 12 ei

n = e

(T12.52)

where the arbitrary phase is unimportant.

A small exercise: Show that the other state n (with spin down compared to
) may be written as
the direction n

n = e

sin 12
cos 12 ei

=e

i0

sin 12 ei
cos 21

(T12.53)

There are several ways to do this: You may (i) change the sign of the eigenvalue
in (T12.49) and (T12.50) and proceed as above, or (ii) use that n must be
orthogonal to n . A third method is (iii) to notice that this state has spin up
0 =
in the direction n
n, which corresponds to the angles 0 = and
0 = + .
Some other exercises:
a) Show that for a general normalized spinor =

a
b

we have

h x i = = <e(2a b),
h y i = =m(2a b),

(T12.54)

h z i = |a|2 |b|2 ,
giving the spin direction
<e(2a b) + y
=m(2a b) + z (|a|2 |b|2 ).
h i = x

(T12.55)

TFY4250/FY2045 Tillegg 12 - Magnetic moments. Spin

20

b) Show that the length of the real vector h i is equal to 1. [Hint: [<e(a b)]2 +
[=m(a b)]2 = |a b|2 = |a|2 |b|2 .]
c) Show that the direction of h i is determined uniquely by the complex ratio
b/c. [Hint: Take a factor |a|2 = a a outside.]
for the state n , by inserting a and b from (T12.52).
d) Check that h i = n
e) Find h S i for the state y using the formula (T12.54).
!

a
As a conclusion of the discussions above, we can state that an arbitrary spinor =
b
corresponds to a well-defined spin direction h i, and that is an eigenspinor of the h i
component of S:
.
h i S = 21 h
Note that both h i and are uniquely determined by the complex ratio b/a (modulo a
phase factor for ).

12.2.d Precession in homogeneous magnetic field

(8.3.5 in Hemmer, 4.3 in Griffiths)
A homogeneous magnetic field B = B
ez will not split a particle beam the way a Stern
b
Gerlach magnet does, but it will still affect the spin state, via the interaction term B.
If we disregard the orbital motion of the electron and all other interactions, that is, if we
consider only the behaviour of the spin, then the Hamiltonian reduces to this interaction
term:


ge e
c = B
b = B
S S = Sz ,
(T12.56)
H
2me
where
ge eB
z .

e
2me
We start by noting that since the Hamiltonian is proportional to Sz , the spin eigenfunctions z of Sz now become energy eigenstates, with the energies
E = 21 h
.

TFY4250/FY2045 Tillegg 12 - Magnetic moments. Spin

ground state z

21

1. excited state z

The state z has h i parallell to the B field, and therefore is the ground state, with
the energy E = 21 h
. The first (and only) excited state has spin up and therefore h i
antiparallell to the B field.
Each of these energy eigenstates corresponds to a stationary state,
(t) = eiE t/h z = eit/2 z .

(T12.57)

In these stationary states nothing happens, so h i stays parallell or antiparallell to B the

whole time.
On the other hand, if we consider the superposition
(t) = a0 + (t) + b0 (t) =

a0 eit/2
b0 eit/2

a
b

(T12.58)

then this is a non-stationary state where things happen. If we assume for simplicity that
a0 b0 is real, we have
a b = a0 b0 eit .
Inserting into (T12.55) we then find that
x (2a0 b0 ) cos t + e
y (2a0 b0 ) sin t + e
z (|a0 |2 |b0 |2 ) .
h S i = 12 h
h i = 12 h
e
h

(T12.59)

Here we see that h Sz i is constant (independent of t), while

d
h Sx i = h Sy i
dt

and

d
h Sy i = h Sx i .
dt

(T12.60)

This shows that h S i (and hence also h i) precess around the z axis with angular frequency
:
d
d
hSi = hSi ,
hi = hi.
(T12.61)
dt
dt
Note that this is quite analogous to the classical precession that we found for the magnetic
moment on page 3 [see equation (T12.6)].

A small challenge: Use the formula ((4.19) in Hemmer) for the time development of expectation values,
d
iD c b E
hF i =
[H, F ] ,
dt
h

to derive the formulae (T12.60).