Spin-polarized muons in condensed matter physics

arXiv:cond-mat/0207699v1 [cond-mat.str-el] 30 Jul 2002

S. J. Blundell

Oxford University Department of Physics,

Clarendon Laboratory, Parks Road, Oxford OX1 3PU
February 1, 2008

Abstract
A positive muon is a spin-1/2 particle. Beams of muons with all their spins polarized can be prepared and
subsequently implanted in various types of condensed matter. The subsequent precession and relaxation of their
spins can then be used to investigate a variety of static and dynamic effects in a sample and hence to deduce
properties concerning magnetism, superconductivity and molecular dynamics. Though strictly a lepton, and behaving
essentially like a heavy electron, it is convenient to think of a muon as a light proton, and it is often found with
a captured electron in a hydrogen-like atom known as muonium. This article outlines the principles of various
experimental techniques which involve implanted muons and describes some recent applications. The use of muons
in condensed matter physics has shed new light on subjects as diverse as passivation in semiconductors, frustrated
spin systems, vortex lattice melting, and quantum diffusion of light particles.

Originally published in Contemporary Physics 40, 175, (1999)

1 Introduction
Condensed matter physics is concerned with the prop-
erties of solids, liquids and various intermediate states
of matter such as colloids (Chaikin and Lubensky 1995).
From a fundamental viewpoint, all matter is made up of
two main types of constituent particles, quarks and lep-
tons (see Figure 1). The universe is composed of only
these particles in addition to the gauge bosons (such as
photons) which mediate forces between particles. Pro-
tons and neutrons are each made up of three quarks
while electrons are thought to be indivisible and are
members of a class of particle called leptons. Everyday
matter, made up of protons, neutrons and electrons,
is the main concern of condensed matter physics and
therefore the subject arises from particles in the first
column (or ‘generation’) of the standard model (Fig-
ure 1). However, more exotic particles do exist which
are either more unusual leptons or combinations of more Figure 1: Quarks (u=up, d=down, s=strange,
exotic quarks; they are all short-lived and have never c=charm, t=top, b=bottom) and leptons (e=electron,
been found to be relevant in condensed matter physics µ=muon, τ =tau, and their associated neutrinos νe , νµ ,
with one single exception: the muon. ντ . Everyday matter is composed of particles from the
The muon is found in nature; it is the dominant con- first column in this table.
stituent of the cosmic rays arriving at sea-level and
a few will be hitting you each minute as you read
this article. The fact that so many muons make it

1
 Spin-polarized muons in condensed matter physics S. J. Blundell 2

through the atmosphere is a well-known consequence (Xu and Brown 1987). Millikan’s name for the effect,
of time-dilation effects in special relativity (Rossi and cosmic rays, gained universal acceptance, but it was
Hall 1941, Bailey et al. 1977). However, at a number Hess who was awarded the Nobel prize for the discov-
of locations in the world, intense beams of muons are ery in 1936. The major constituent of cosmic rays at
prepared artificially for research in condensed matter ground level turned out to be the muon, but the true
physics. In this article I will try and explain why this identity of the muon remained hidden for some time.
is done and what can be learned by firing this rather Millikan’s group at CalTech began to perform exper-
exotic ‘second-generation’ particle into various types of iments on cosmic rays at ground level by bending the
matter. I will describe how implanted muon studies rays in magnetic and electric fields. This work led to
have shed new light on problems in condensed mat- the discovery of the positron by Carl Anderson in 1932
ter physics, including those concerning semiconductors, (Anderson shared the Nobel prize in 1936 with Hess)
magnetism, superconductors, and quantum diffusion. and then the muon (initially called the mesotron, and
A key difference between this technique and those in- later the mu-meson) was identified in cosmic rays by
volving neutrons and X-rays is that scattering is not in- Seth Neddermeyer and Carl Anderson in 1936. They
volved. Neutron diffraction methods use the change in measured the mass of the muon, which turned out to
energy and/or momentum of a scattered neutron to in- be roughly one-ninth of the proton mass, and agreed
fer details about a particular sample. In contrast muons extremely well with the predictions of the Japanese
are implanted into a sample and reside there for the physicist Hideo Yukawa for a particle to mediate the
rest of their lives. The muons themselves never emerge strong force and hold the nucleons in the nucleus to-
again. It is the positrons into which they decay that are gether. This result can be understood by using Heisen-
released from the sample and yield information about berg’s uncertainty principle ∆E∆t ∼ h̄ for this virtual
the muons from which they came. The experimenter particle. ∆E is the the rest mass energy of Yukawa’s
plays the rôle of a coroner at an inquest; the evidence particle, the energy which has to be “borrowed”, and
of the decay particles given up at death allows one to ∆t is the time for it to cross the nucleus, the time for
infer the nature of the muon’s life. which it has to be “borrowed”; by assuming a speed
∼ c for the particle one arrives at a mass ∼100 MeV,
the mass found for the muon. The muon was therefore
2 The discovery of the muon assumed to be the particle which holds the nucleus to-
gether. It took some technically difficult experiments
Although the discovery of the muon is usually consid-
to show that this assumption was in error and that the
ered as a single event which is credited to Neddermeyer
startling agreement with Yukawa’s predictions was mis-
and Anderson because of their work in 1936, the story is
leading. Marcello Conversi, Ettore Pancini and Oreste
actually more complicated and one might say that the
Piccioni, working in a basement in Rome during and
muon slowly “emerged” over almost a half-century. Fol-
after the Second World War, tried to implant cosmic
lowing the discovery of radioactivity in 1899, it was ob-
ray muons in matter and then measure their lifetime
served that electrometers, instruments capable of mea-
(Conversi et al. 1945, 1947). They found that positive
suring the ionization produced by radioactivity, dis-
muons implanted in anything always live on average for
charged even when there was no obvious radioactive
2.2 µs (the lifetime in vacuum); the lifetime of negative
source. A Jesuit priest called Thomas Wulf noticed in
muons on the other hand depends on the atomic num-
1910 that this effect was more pronounced at the top
ber Z of the material into which it is implanted; they
of the Eiffel tower than at the bottom (Wulf 1910, see
measured 2.2 µs for carbon but found 0.07 µs for lead.
also Xu and Brown 1987). That the rate of discharge
They reasoned that if muons really mediated the strong
of an electrometer was an increasing function of alti-
force they should be gobbled up much more quickly in
tude was demonstrated beyond doubt by Victor Hess
all materials. It appeared that muons interacted only
in an intrepid series of ballooning experiments in the
rather weakly with matter and the only effect that was
period 1911–12 (Hess 1912). He showed that electrom-
observed occurred for negative muons and turned out
eters discharged between three and five times faster at
to be µ− capture:
an altitude of 5000 m than they did at sea level. Such
experiments were later extended to very high altitudes µ− + p → n + νµ .
by the use of remote-controlled electrometers (in which
the experimenter did not actually accompany their ex- A negative muon is attracted by atomic nuclei and since
periment in the balloon); this advance was achieved by its mass is much greater than that of an electron, it
Robert Millikan who was initially sceptical about Hess’ readily displaces an electron from an atom and rapidly
work but became rapidly convinced of its importance drops down to the 1s state. From there it either de-
 Spin-polarized muons in condensed matter physics S. J. Blundell 3

Figure 2: The muon can be considered to be the light-

est member of the family of particles which include the Figure 3: Properties of the electron, muon and proton.
proton, deuteron and triton. The approximate ratio
of their masses are as shown. The muon thus extends 3 Muon production, decay and
the mass range of isotope substitution available to the
experimenter.
implantation
Cosmic rays provide a major source of muons [roughly
one muon arrives vertically on each square centimetre
cays to an electron or undergoes capture as above. The of the earth’s surface every minute (Caso et al. 1998)]
propensity to undergo capture depends on the atomic but, since the 1950s, experiments on muons have needed
number Z (approximately as the fourth power of Z) higher intensities and have required the development
through the Bohr radius of the µ− orbit. of accelerators. High energy proton beams (produced
using synchrotrons or cyclotrons) are fired into a target
Yukawa’s particle turned out to be the pion, discov- (usually graphite) to produce pions via
ered in cosmic rays by Powell’s group in 1947 (Lattes et
al. 1947), which decays into a muon after only 0.026 µs, p + p → π + + p + n,
making it harder to spot in cosmic rays. Thus the muon and the pions decay into muons:
was not a meson (a term now reserved for quark–anti-
quark pairs like the pion) but a lepton, a heavy electron π + → µ+ + νµ ,
with no internal structure. where νµ is a muon-neutrino. The pion decay is a two-
For experiments in condensed matter physics it is body decay and is therefore particularly simple. For
mainly the positive muon which is used. The negative example, consider the pions which are produced at rest
muon µ− , which implants close to an atomic nucleus, is in the laboratory frame. To conserve momentum, the
generally much less sensitive to the phenomena of inter- muon and the neutrino must have equal and opposite
est to condensed matter physicists (magnetism, super- momentum. The pion has zero spin so the muon spin
conductivity, etc) than the site of the implanted positive must be opposite to the neutrino spin. One useful prop-
muon µ+ which sits well away from nuclei in regions of erty of the neutrino is that its spin is aligned antiparallel
large electron density. Condensed matter physics is es- with its momentum (it has negative helicity), and this
sentially the physics of electrons, rather than nuclei, so implies that the muon-spin is similarly aligned. Thus by
that the best place you can put your test particle is in selecting pions which stop in the target (and which are
the electron cloud. In fact even though the muon is a therefore at rest when they decay) one has a means of
lepton (see Figure 1) and therefore essentially a heavy producing a beam of 100% spin-polarized muons. This
electron, for our purpose it is more useful to consider it is the method most commonly used for producing muon
as a light proton (see Figure 2). beams for condensed matter physics research, though
other configurations are in use (Brewer 1994).
Some properties of the electron, muon and proton are The muons are stopped in the specimen of interest
tabulated in Figure 3. The mass of the muon is inter- and decay after a time t with probability proportional
mediate between that of the electron and the proton, to e−t/τµ where τµ = 2.2 µs is the lifetime of the muon.
and thus so are its magnetic moment and gyromagnetic The muon decay is a three body process
ratio. The latter is the constant of proportionality be-
tween angular momentum and magnetic moment. µ+ → e+ + νe + ν̄µ
and so the energy of the positron e+ (which is the only
particle produced in this reaction that we have a sensi-
ble hope of reliably detecting) may vary depending on
 Spin-polarized muons in condensed matter physics S. J. Blundell 4

how momentum is distributed between the three par-

ticles (subject to the constraint that the total vector
momentum will sum to zero, the initial momentum of
the stopped muon). The decay involves the weak in-
teraction and thus has the unusual feature of not con-
serving parity (Garwin et al. 1957). This phenomenon
(which also lies behind the negative helicity of the neu-
trino) leads to a propensity for the emitted positron to
emerge predominantly along the direction of the muon-
spin when it decayed. This can be understood by con-
sidering the mirror image of the muon decay process
(Figure 4).
The angular distribution of emitted positrons is
shown in Figure 5 for the case of the most energeti-
cally emitted positrons. In fact positrons over a range
of energies are emitted so that the net effect is some-
thing not quite as pronounced, but the effect neverthe-
less allows one to follow the polarization of an ensemble
of precessing muons with arbitrary accuracy, providing
one is willing to take data for long enough.
Muons are implanted into the sample with an energy
which is at least 4 MeV. They lose energy very quickly
(in 0.1–1 ns) to a few keV by ionization of atoms and
scattering with electrons. Then the muon begins to un-
dergo a series of successive electron capture and loss
reactions which reduce the energy to a few hundred eV
in about a picosecond. If muonium is ultimately formed
then electron capture ultimately wins and the last few
eV are shed by inelastic collisions between the muo-
nium atom and the host atoms. All of these effects are
very fast so that the muon (or muonium) is thermalized
very rapidly. Moreover the effects are all Coulombic in
origin and do not interact with the muon-spin so that
the muon is thermalized in matter without appreciable
depolarization. This is a crucial feature for muon-spin
rotation experiments. One may be concerned that the
muon may only measure a region of sample which has Figure 4: Top: molecules in a box collide with the
been subjected to radiation damage by the energetic walls. These collisions do not violate parity so that
incoming muon. This does not appear to be a prob- both the process shown on the left, and its mirror im-
lem since there is a threshold energy for vacancy pro- age on the right, could be observed in nature. Bottom:
duction, which means that only the initial part of the the same is not true for the process of muon decay. The
muon path suffers much damage. Beyond this point of direction of the muon-spin is reversed in the mirror so
damage the muon still has sufficient energy to propa- that the positrons are emitted predominantly in a direc-
gate through the sample a further distance thought to tion opposite to that of the muon-spin. The violation
be about 1 µm, leaving it well away from any induced of parity means that in our universe only the process
vacancies (Chappert 1984). on the left-hand side of the diagram is ever observed.

4 Spin precession and relaxation

In a magnetic field B the muon-spin precesses with an-
gular frequency ωµ given by ωµ = γµ B where γµ =
ge/2mµ is the gyromagnetic ratio for the muon. This
is known as Larmor precession. The field-dependent
 Spin-polarized muons in condensed matter physics S. J. Blundell 5

Figure 5: The angular distribution of emitted positrons

with respect to the initial muon-spin direction. The fig-
ure shows the expected distribution for the most ener-
getically emitted positrons.

precession frequencies for the muon, electron and pro-

ton are shown in Figure 6. The highest frequencies are
associated with the lightest particle, the electron, and Figure 6: The Larmor precession frequency f in MHz
the lowest with the proton. For usual laboratory mag- (and the corresponding period τ = 1/f ) for the elec-
netic fields this explains why ESR (electron-spin reso- tron, muon and proton as a function of applied mag-
nance) is typically performed at microwave frequencies netic field B.
while NMR (nuclear magnetic resonance) uses radio-
frequencies. In both of these techniques resonance oc-
curs when the precession frequency matches the reso-
nance frequency. Muon-spin rotation (µSR) is associ-
ated with frequencies intermediate between NMR and
ESR but unlike those resonance techniques, no electro-
magnetic field is necessary since the precessing muon
can be followed directly. [Muon-spin resonance experi-
ments can however be performed (Brewer 1994) but a
discussion is outside the scope of this article.]
A schematic diagram of the experiment is shown in
Figure 7(a). A muon, with its polarization aligned an-
tiparallel to its momentum, is implanted in a sample.
(It is antiparallel because of the way that it was formed,
see above, so the muon enters the sample with its spin
pointing along the direction from which it came.) If the
muon is unlucky enough to decay immediately, then it
will not have time to precess and a positron will be
emitted preferentially into the backward detector. If it
lives a little longer it will have time to precess so that
if it lives for half a revolution the resultant positron
Figure 7: Schematic illustration of a µSR experiment.
will be preferentially emitted into the forward detector.
(a) A spin-polarized beam of muons is implanted in a
Thus the positron beam from an ensemble of precess-
sample S. Following decay, positrons are detected in
ing muons can be likened to the beam of light from a
either a forward detector F or a backward detector B.
lighthouse.
If a transverse magnetic field H is applied to the sample
The time evolution of the number of positrons de- as shown then the muons will precess. (b) The num-
tected in the forward and backward detector is de- ber of positrons detected in the forward and backward
scribed by the functions NF (t) and NB (t) respectively detectors. (c) The asymmetry function.
and these are shown in Figure 7(b). Because the muon
 Spin-polarized muons in condensed matter physics S. J. Blundell 6

This is an example of an experiment with no applied

magnetic field (in fact a small magnetic field was ap-
plied to compensate the effect of the earth’s field). Very
often magnetic fields are applied to the sample either
perpendicular or parallel to the initial muon-spin di-
rection. The perpendicular (or transverse) case causes
the muon to precess in the applied magnetic field and
any dephasing in the observed oscillations is evidence
for either an inhomogeneous internal field distribution
or spin-spin (T2 in the language of NMR) relaxation.
The parallel (or longitudinal) case will be described in
more detail later and does not lead to spin precession,
but spin relaxation. This can be due to inhomogeneous
field distributions or spin-lattice (T1 ) relaxation pro-
Figure 8: Zero-field muon spin rotation frequency in the cesses.
organic ferromagnet p-NPNN (Blundell et al. 1995). Muon experiments can be performed in two differ-
ent ways depending on the time structure of the muon
beam. If the muon beam is continuous (CW, or con-
decay is a radioactive process these two terms sum to an
tinuous wave see Figure 9(a)), then muons arrive at
exponential decay. Thus the time evolution of the muon
the sample intermittently. When the muon enters the
polarization can be obtained by examining the normal-
experiment it must itself be detected to start a clock.
ized difference of these two functions via the asymmetry
When the positron is detected in either the forward or
function A(t), given by
backward detectors, the clock is stopped. If a second
NB (t) − NF (t) muon arrives before the first one has decayed then one
A(t) = , (1) has no way of knowing whether a subsequently emitted
NB (t) + NF (t)
positron came from the first or second muon, so this
and is shown in Figure 7(c). event must be disregarded. Sophisticated high-speed
This experimentally obtained asymmetry function electronics and a low incident muon arrival rate are
has a calculable maximum value, Amax , for a partic- needed. Alternatively one can use an electrostatic de-
ular experimental configuration which depends on the flector triggered by the detectors to ensure no muons
initial beam polarization (usually very close to 1), the enter the experiment until the current implanted muon
intrinsic asymmetry of the weak decay, and the effi- decays.
ciency of the detectors for positrons of different ener- These complications are circumvented with a pulsed
gies, and usually turns out to be around Amax ∼0.25. muon beam (Figure 9(b)). In this case a large number
The function can be normalized to 1, in which case of muons arrive in a very intense pulse so there is no
it expresses the spin autocorrelation function of the need to detect when each muon arrives. The detection
muon, G(t) = A(t)/Amax , which represents the time- of positrons is then made and each event is timed with
dependent spin polarization of the muon. respect to the arrival of the pulse. A typical dataset
A magnetic field does not need to be applied for contains several million detected positrons so that an
the muons to precess if the sample has its own mag- appreciable number of muons (the fraction is given by
netic field. Figure 8 shows muon-spin rotation data for e−20/2.2 ∼0.01%) live for 20 µs or longer. Long-lived
muons implanted into an organic ferromagnet (Blundell muons are difficult to measure with CW beams; the
et al. 1995). This material magnetically orders only at the arrival of the next muon tends to interrupt the first
very low temperatures (TC = 0.67 K) so that exper- muon which has outstayed its welcome! Nevertheless
iments must be carried out in a dilution refrigerator. the long-lived muons can be accurately detected at a
This presents no problems for the muon which passes pulsed source. Unfortunately this method also suffers
through the windows of the cryostat and implants in from a drawback which is that the muon pulse has a
the sample, revealing the internal magnetisation. As finite width, τw , which results in a slight ambiguity in
the sample is warmed, the frequency of oscillations de- all of the timing measurements and leads to an upper
creases as the internal field decreases until it is above limit on precession frequencies which can be measured.
the Curie temperature and no oscillations can be ob- By the standard of other unstable elementary particles,
served, only a weak spin relaxation arising from spin the muon is comparatively long-lived with a lifetime of
fluctuations. τ =2.2 µs. CW muon beams are operated at the Paul
 Spin-polarized muons in condensed matter physics S. J. Blundell 7

defects in a semiconductor. The electronic structures

of analogous hydrogen and muonium centres are ex-
pected to be identical, apart from vibrational effects
reflecting their difference in mass. If muons are im-
planted into a semiconductor like silicon, neutral muo-
nium tends to form at the tetrahedral sites (labelled T
in Figure 11) and can diffuse rapidly between such sites.
The electronic state is isotropic and in a transverse-field
experiment the transitions between energy levels can
Figure 9: Schematic illustration of the two types of be observed as precession frequencies νij equal to the
muon beam (a) continuous wave (CW) and (b) pulsed. splitting between energy levels (Ei − Ej )/h (as shown
in Figure 10). The strength of the hyperfine interac-
tion in semiconductors is usually close to half that of
..
........ the vacuum value (the reduction is due to some ad-
......... ...
................ ....
....
.... .... mixture of the electron spin density with surrounding
........ .... ...
........ .........
. .
....
................ ................... ....
....
atoms). However, a substantial fraction of neutral muo-
... .... ....
......... ........... ...
........ ...........
.
....
. ..
.. ..
. ........ .........................
..
.
....
.... nium is also found in a most unexpected place, wedged
............................................ . .......
.........
......... . ....
....
... into the centre of a stretched Si–Si bond (the site is
......... .. ....
......... .. ........ labelled BC in Figure 11 for ‘bond-centre’, see Patter-
......... ........
.........
.........
......... son 1988 for a review of muonium states in semicon-
......... .......
.........
......... .. .........
.. ....
ductors). This state is extremely immobile, and sur-
......... . ....
................................
...............
....
.
....
....
....
....
prisingly turns out to be the thermodynamically more
............. ....
...........
........... .
....
....
....
stable site. Its hyperfine coupling is much lower than
.......... .... ..
.......... .... .... that of the tetrahedral state, typically less than 10%
.......... .... ....
........ .... .......
.. ..
. ..
of the vacuum value. Furthermore the coupling is very
anisotropic, with axial symmetry about the h111i crys-
tal axis (i.e. along the Si–Si bonds), so that the energy
Figure 10: The Breit-Rabi diagram for isotropic muo- levels behave in a different manner to that indicated
nium with a hyperfine constant A = 4463 MHz appro- in Figure 10. These states have rather interesting dy-
priate for free (vacuum) muonium. Also shown are the namics and can undergo charge and spin exchange pro-
splittings between pairs of levels over a larger range of cesses, cycling rapidly between positive and negative
magnetic field. Levels 1 and 2 cross at a field of ∼16.5 T charge states via interaction with conduction electrons
for free muonium. (Chow et al. 1994). The characterisation of all these
states is important because it is found that atomic hy-
drogen is present in most semiconductors and is able to
Scherrer Institute in Switzerland and at TRIUMF in passivate (i.e. deactivate) the dangling bonds in amor-
Canada. Pulsed muon beams are used at KEK in Japan phous silicon allowing it to show semiconducting prop-
and at ISIS (the spallation source at the Rutherford erties. Hydrogen is inevitably present in all semicon-
Appleton Laboratory) in the UK. ductors, often becoming incorporated during material
production from hydride gases or during etching, but
the low concentration makes direct spectroscopic stud-
5 Muonium ies very difficult. Using muonium (albeit in the ultra-
dilute limit!) as an analogue for hydrogen has therefore
Depending on its chemical environment, the muon can
been a promising method of obtaining a great deal of
thermalize and pick up an electron and form a neutral
spectroscopic information concerning this problem.
atomic state called muonium (abbreviated Mu= µ+ e− )
which is an analogue of atomic hydrogen. In muonium Muonium states can also be formed in many chemical
the electronic spin and the muon-spin are coupled by systems and allow a unique form of radical spectroscopy
a hyperfine interaction which we will initially assume (Roduner 1993). Muonium adds to unsaturated bonds
is isotropic. This leads to two energy levels, a lower to form muonated free radicals. For example, addi-
triplet state and a higher singlet state. In a magnetic tion to benzene (C6 H6 ) leads to the muonated cyclo-
field the triplet levels split and the energy levels move hexadienyl radical (C6 H6 Mu). The advantage here is
as shown in the Breit-Rabi diagram in Figure 10. that one can work with concentrations down to just one
Isolated hydrogen or muonium is one of the simplest muonated radical at a time in an entire sample. In con-
 Spin-polarized muons in condensed matter physics S. J. Blundell 8

further information to be extracted (Pratt 1997).

At very high fields, when the electron and muon spins
are completely decoupled, the initial muon polarization
is preserved. However, at certain values of the magnetic
field a level-crossing of energy levels may occur. In fact,
there is one at very high field in vacuum muonium (see
Figure 10(b)). Interactions between the muon-electron
system and nuclei in the host material cause the pure
Zeeman states to mix near these level-crossings, thereby
avoiding the crossing, and this can lead to a loss of po-
larization. This provides a further technique to extract
hyperfine coupling strengths and also measurements of
Figure 11: The crystal structure of silicon, showing the quadrupole couplings (Cox 1987, Roduner 1993, Brewer
possible muonium sites (T=tetrahedral site, BC=bond- 1994). It turns out that these avoided-level-crossing res-
centre site). onances are also a sensitive probe of dynamics and this
technique has been applied to a number of studies of
molecular motion (Roduner 1993).
An interesting situation occurs in the case of C60 in
which muonium can implant inside the buckyball cage
(this state is called endohedral muonium). The un-
paired electron part of the muonium greatly enhances
the sensitivity to scattering from conduction electrons
so that this state is extremely useful for studying alkali-
fulleride superconductors (MacFarlane et al. 1998). It
is also possible to form a muonium radical by external
addition, essentially muonium attacking the outside of
a buckyball, breaking a double bond and ending up
covalently bonded to a single saturated carbon atom.
This centre is very sensitive to the molecular dynamics
Figure 12: Repolarization of the initial muon polar- of the local environment and has been used to extract
ization. the correlation time for molecular reorientation (Kiefl
et al. 1992).
In metallic samples the muon’s positive charge is
trast ESR detection needs ∼ 1012 radicals in a cavity, screened by conduction electrons which form a cloud
forbidding measurements at high temperatures where around the muon, of size given by a Bohr radius. Thus
the radicals become mobile and terminate by combi- µ+ , rather than muonium, is the appropriate particle to
nation. The technique has been applied to radicals in consider in a metal. (The endohedral muonium found
various environments (Roduner 1993) including those in alkali fulleride superconductors is the only known
absorbed on surfaces (Reid et al. 1990). example of a muonium state in a metal.) In insulators
In all cases the muonium can be studied by mea- and semiconductors screening cannot take place so that
suring precession signals in an applied magnetic field, the muon is often observed in these systems either as
or by using a technique known as repolarization (Fig- muonium or is found to be chemically bound to one of
ure 12). In this latter method a longitudinal magnetic the constituents, particularly to oxygen if it is present.
field is applied to the sample, along the initial muon- Isotropic muonium states are found in many semicon-
spin direction, and as the strength of the magnetic field ducting and insulating systems. The value of the hy-
increases, the muon and electron spins are progressively perfine coupling strength is close to that for vacuum
decoupled from the hyperfine field. For isotropic muo- (free) muonium if the band gap is large. For materials
nium, half of the initial polarization of implanted muons with smaller band gap the hyperfine coupling is lower
is lost because of the hyperfine coupling, but this is re- reflecting the greater delocalization of the electron spin
covered in a sufficiently large applied field (Figure 12), density on to neighbouring atoms (Cox 1987).
allowing an estimate of the strength of the hyperfine
field. If the hyperfine coupling is anisotropic then this
affects the form of the repolarization curve, allowing
 Spin-polarized muons in condensed matter physics S. J. Blundell 9

6 Muons and magnetism

Muons are ideally suited to studying problems in mag-
netism. Implanted muons in magnetically ordered ma-
terials precess in the internal magnetic field and directly
yield signals proportional to that magnetic field. In
this respect the muon behaves as a microscopic mag-
Figure 13: Muon-spin precession with a magnetic field
netometer. The very large magnetic moment of the
B applied at an angle θ.
muon makes it very sensitive to extremely small mag-
netic fields (down to ∼ 10−5 T) and thus is very useful
in studying small moment magnetism. It is also valu- (which couple magnetic moments through the conduc-
able in studying materials where the magnetic order is tion electrons). Depending on the relative strength of
random or of very short range. Since muons stop uni- these effects, magnetic order can be observed at low
formly throughout a sample, each signal appears in the temperature but it is often incommensurate or random
experimental spectrum with a strength proportional to and can be associated with very small static moments.
its volume fraction, and thus the technique is helpful in The sensitivity of the muon to tiny fields and the ab-
cases where samples may be multiphase or incompletely sence of a muon quadrupolar moment (which can com-
ordered. Because no spatial information is obtained (in plicate analogous NMR experiments) has led to muons
contrast to diffraction techniques) single crystal sam- being extensively utilised in this field (Amato 1997).
ples are not essential (though they can be useful in To understand the ability of the muon to study ran-
certain cases) and experiments can often provide in- domness and dynamics in magnetism it is helpful to
formation on the magnetic order of certain materials consider further some aspects of spin precession. If the
where conventional magnetic neutron diffraction can- local magnetic field at the muon-site is at an angle of
not be simply performed. θ to the initial muon spin-direction at the moment of
The most straightforward application of the tech- implantation, the muon-spin will subsequently precess
nique is to ferromagnets and antiferromagnets in which around the end of a cone of semi-angle θ about the mag-
cases the muon is used to follow the temperature de- netic field (Figure 13). The normalized decay positron
pendence of the internal field and to extract critical asymmetry will be given by
exponents (Schenck and Gygax 1995). An example of
this type of experiment has already been presented in G(t) = cos2 θ + sin2 θ cos(γµ Bt). (2)
Figure 8. The internal field at the muon site is due
If the direction of the local magnetic field is entirely
to a sum of the applied field (if used), the dipolar
random then averaging over all directions would yield
and demagnetization fields (which can be calculated
from the magnetization) and the hyperfine field induced 1 2
by the applied field. To extract quantitative informa- G(t) = + cos(γµ Bt). (3)
3 3
tion from µSR experiments it is necessary to know the
muon-site and this can in some cases hinder the search If the strength of the local magnetic field is taken from
for a straightforward interpretation of the data. Usu- a Gaussian distribution of width ∆/γµ centred around
ally there are a small set of possible interstitial sites zero, then a straightforward averaging over this distri-
which the muon can occupy and in favourable circum- bution gives
stances only one will be consistent with the observed 1 2 −∆2 t2 /2
data. Nevertheless the technique has been widely ap- G(t) = + e (1 − ∆2 t2 ), (4)
3 3
plied to magnetic materials (Schenck and Gygax 1995,
Dalmas de Réotier and Yaouanc 1997) and has found a result which was first obtained by Kubo and Toy-
great applicability to the study of heavy fermion sys- abe (Kubo and Toyabe 1967) as an entirely theoreti-
tems (Amato 1997). These latter materials are based on cal exercise. This relaxation function is illustrated in
rare-earth and actinide elements and show strong elec- Figure 14(b). Its origin is indicated schematically in
tronic correlations between localized f moments and Figure 14(a) which shows a number of curves of equa-
conduction electrons. They exhibit a subtle compe- tion 4 for different values of the internal field B. Ini-
tition between the Kondo effect (by which magnetic tially they all do roughly the same thing (i.e. fall from
moments are ‘mopped up’ by the screening spin po- 1 to a minimum value and then increase) but after a
larization of conduction electrons) and the magnetic short time they dephase with respect to each other.
Ruderman-Kittel-Kasuya-Yosida (RKKY) interactions Hence their average, the Kubo and Toyabe relaxation
 Spin-polarized muons in condensed matter physics S. J. Blundell 10

function, would be expected to fall from unity to a min-

imum and then recover to an average value, in this case
to one-third.
This relaxation function is often observed experimen-
tally, for example in experiments on copper at 50 K with
zero applied field (Luke et al. 1991). At this temper-
ature the muon is stationary and precesses in the field
due to the neighbouring nuclear dipoles which are ran-
domly orientated with respect to each other and give
rise to a field distribution each component of which is
Gaussian distributed about zero. (At higher temper-
atures this is not seen because of thermally activated
muon hopping while at lower temperatures an effect
known as quantum diffusion may occur, see section 8.)
If the form of the distribution of internal fields was
different then this would affect the form of the ob-
served muon-spin time evolution. For example, in a
material with a spin-density wave which is incommen-
surate with the crystal lattice, there will be a sinusoidal
modulation of the internal field which the muons will
randomly sample. In this case one finds that the muon-
spin relaxation follows a Bessel function (Major et al.
1986, Amato 1997). Such an effect has been observed
in chromium (Major et al. 1986) and also in an organic
system, (TMTSF)2 PF6 , which exhibits a spin-density
wave ground state (Le et al. 1993).
If there is an almost uniform static internal field in
the sample, but there is a slight variation from site to
site, different muons will precess at slightly different
frequencies and become progressively dephased so that
the oscillations in the data will be damped. If the field
varies a great deal the damping could be so large that
no oscillations can be observed. However this effect
could also be caused by fluctuations either of the in-
ternal field, because of some intrinsic property of the
sample, of the muon’s position, or because of muon dif-
fusion. One method of distinguishing between these
possibilities is to apply a magnetic field in the longitudi-
nal direction, parallel to the initial muon-spin direction.
For example, this modifies the Kubo-Toyabe relaxation
function as shown in Figure 14(c), causing the “ 31 -tail”
to increase since in this case the muons precess in both
Figure 14: (a) The time evolution of the muon-spin
the internal field and the applied field. Since the mag-
polarization for equation 3 with different values of the
netic field BL is applied along the initial muon-spin
magnitude of the local field |B|. (b) The averaging
direction then in the limit of large BL the muon-spin
of terms from (a) yields the Kubo-Toyabe relaxation
is held constant and does not relax from a value near
function (equation 4) with its characteristic dip and
unity. This technique is useful to distinguish the ef-
recovery to a value of 1/3. (c) The effect of a longitu-
fects of inhomogeneous line broadening (a distribution
dinal magnetic field BL . The time is measured in units
of static internal fields as considered above) and fluctu-
of ∆−1 and the longitudinal field values shown are in
ations because the two have very different behaviours
units of ∆/γµ .
in a longitudinal field (Hayano et al. 1979).
The effects of muon hopping on the relaxation are
shown in Figure 15. The different traces are for dif-
 Spin-polarized muons in condensed matter physics S. J. Blundell 11

ders. µSR experiments show no magnetic order down

to very low temperatures. Sr4 Cu6 O10 is a 3-leg lad-
der and µSR finds static magnetic order appearing at
∼52 K. This is consistent with theoretical predictions
of a non-magnetic ground state in even-leg-number sys-
tems (due to the formation of spin-singlets between
pairs of spins on each rung) but a magnetic ground
state in odd-leg-number systems (Kojima et al. 1995).
Similar types of experiments have been performed on
SrCr8 Ga4 O19 , a material which has magnetic moments
lying in an arrangement known as a Kagomé lattice
which is highly frustrated in the sense that there is no
unique ground state. These measurements have demon-
strated the presence of dynamic spin fluctuations which
Figure 15: The relaxation function for a muon hop- persist down to at least 0.1 K (Uemura et al. 1994).
ping at rate ν. After each hop the value of the internal These strong dynamics which remain at very low tem-
field is taken from a Gaussian distribution about zero peratures are characteristic of a spin-liquid state in
with width ∆/γµ . The curve for ν = 0 corresponds to which a small number of unpaired spins migrate in a sea
the zero-field Kubo-Toyabe relaxation function of equa- of singlet pairs. Some compounds with one-dimensional
tion 4. chains of antiferromagnetically coupled spins can show
an effect known as the spin-Peierls transition. Below a
transition temperature the chain dimerises, pairing al-
ferent hopping rates ν. When ν = 0 we recover the
ternate spins into singlets and opening up a gap in the
zero-field Kubo-Toyabe curve of equation 4. For fast
excitation spectrum, drastically changing the dynam-
hopping the relaxation of the muon-spin becomes dom-
ics. µSR studies have been performed on both inorganic
inated by the hopping process and the relaxation is ex-
and organic systems which show this effect (Lappas et
ponential. The relaxation rate goes down as the dy-
al. 1994, Blundell et al. 1997).
namics get faster. This effect is known as motional
narrowing because it is essentially identical to the ef- A celebrated example of a frustrated spin system is a
fect of the same name in NMR spectroscopy. Motional spin glass, such as the dilute alloy spin glass prepared
effects narrow NMR linewidths which are measured in by dissolving small concentrations of Mn in a Cu ma-
the frequency domain; consequently in the time domain trix. In these system dilute magnetic impurities couple
(in which data from µSR experiments are usually con- via an RKKY exchange interaction which alternates in
sidered) this corresponds to a motional widening and a sign as a function of distance. Because the magnetic
reduction in relaxation. impurities are present at random, these materials can-
For slow hopping, very little effect is observed at very not show long range order but have built-in frustration.
short times times but a large sensitivity to weak hop- When cooled one observes a slowing down of spin fluctu-
ping is observed in the 13 -tail which is observed at longer ations and a divergence in the correlation time between
times. This sensitivity to slow dynamics via the be- Mn spin fluctuations at the spin glass temperature. Be-
haviour of the tail of the relaxation function observed low this temperature, a static component of the local
at long times permits a measurement of dynamics over field is observed with muons, corresponding to some de-
a very large range of time scale. gree of spin freezing, with each Mn spin having its own
If a longitudinal magnetic field is applied then it has preferred orientation, although fluctuating around this
a large effect on the muon relaxation if the dynamics (Uemura et al. 1985). Above the spin glass temperature
are weak but much less of an effect if the dynamics are muon-spin relaxation measurements have been used to
fast. Thus by a careful combination of zero-field and follow the spin glass dynamics and to directly extract
longitudinal-field experiments the nature of the internal the form of the autocorrelation function of the spins
field distribution can be extracted. (Campbell et al. 1994, Keren et al. 1996).
Such an analysis has been done for many systems An interesting situation occurs when a muon hops in
in which the issue of the presence of dynamics and/or an antiferromagnetically ordered lattice. With no hop-
magnetic order has profound consequences. A good ex- ping, and if there is one muon site close to each spin in
ample is that of the spin ladder cuprates (Kojima et the system, there will be a single precession frequency.
al. 1995). Sr2 Cu4 O6 has a crystal structure such that Because the sign of the internal field reverses every time
Cu2+ spins are arranged in weakly coupled 2-leg lad- the muon hops, the muon will precess one way and then,
 Spin-polarized muons in condensed matter physics S. J. Blundell 12

after hopping, precess back again. As the hop rate in-

creases the oscillations are progressively destroyed un-
til eventually no relaxation is possible (Keren et al.
1993). This effect has been observed experimentally
in the cuprate compound Ca0.86 Sr0.14 CuO2 (Keren et
al. 1993).

7 Muons and superconductivity

One of the most fruitful areas of recent research with
muons has been in the area of superconductivity. The
last couple of decades have witnessed a renaissance in
this field following the discovery of high-temperature
superconductors, organic superconductors, borocarbide
superconductors, and even superconductors based on
C60 buckyballs. To understand the usefulness of muons,
recall that the two important lengthscales in supercon- Figure 16: The field distribution inside a supercon-
ductors are the penetration depth, λ, which controls the ductor as a function of position and the corresponding
ability of the superconductor to screen magnetic fields, muon-spin relaxation function for three cases: (a) the
and the coherence length, ξ, which controls the length- normal state, (b) the superconducting state, (c) as (b)
scale over which the order parameter can vary without but with a shorter penetration depth.
undue energy cost. If the former is sufficiently√greater
than the latter (the condition is that λ > ξ/ 2) the is related to the penetration depth using
material is a type II superconductor which if cooled
through its transition temperature, Tc , in an applied σ = γµ hB(r) − hB(r)ir i1/2 ≈ 0.0609γµΦ0 /λ2 , (5)
r
magnetic field remains superconducting everywhere ex-
cept in the cores of the superconducting vortices which where B(r) is the field at position r and the averages
usually are arranged in a triangular lattice. Each vor- are taken over all positions. Thus the relaxation rate of
tex is associated with a magnetic flux equal to one flux the observed precession signal can be used to directly
quantum Φ0 = h/2e. The distance between vortices, obtain the magnetic penetration depth. An advantage
d, √is such that the number of vortices per unit area is that data are obtained from the bulk of the supercon-
2/ 3d2 equals the number of flux quanta per unit area ductor, in contrast to techniques involving microwaves
B/Φ0 . Thus d ∝ B −1/2 . In general the vortex lattice which are only sensitive to effects at the surface.
will be incommensurate with the crystal lattice and, ex- This principle has been applied to many different su-
cept at very high magnetic field, the vortex cores will be perconductors to extract both the penetration depth
separated by a much larger distance than the unit-cell and its temperature dependence. This latter quantity
dimensions. Implanted muons will sit at certain crys- is of great interest because it is a measure of the tem-
tallographic sites and thus will randomly sample the perature dependence of the order parameter and can
field distribution of the vortex lattice. yield information concerning the symmetry of super-
In the normal state (T > Tc ) with a transverse field conducting gap and hence the symmetry of the pairing
B, all muons precess with frequency ω = γµ B (Fig- mechanism. For example this approach has revealed
ure 16(a)). In the superconducting state however the unconventional pairing in a sample of the high tem-
muons implanted close to the vortex cores experience perature superconductor YBa2 Cu3 O6.95 (Sonier et al.
a larger magnetic field than those implanted between 1994). It is also possible to extract the vortex-core ra-
vortices. Consequently there is a spread in precession dius from a detailed analysis of the data (Yaouanc et
frequency, resulting in a progressive dephasing of the al. 1997, Sonier et al. 1997).
observed precession signal (Figure 16(b)). The larger A conventional type II superconductor exhibits 3
the penetration depth, the smaller the magnetic field well-defined phases for T < Tc : (1) a Meissner phase
variation and the less pronounced the dephasing (com- for B < Bc1 , (2) a mixed or Shubnikov phase for
pare Figure 16(b) and (c)). It turns out that this idea Bc1 < B < Bc2 (in which the magnetic field enters the
can be quantified (see Aegerter and Lee 1997) and that superconductor in the form of well defined flux lines or
the relaxation rate σ of the observed precession signal vortices arranged in a lattice) and (3) the normal metal-
 Spin-polarized muons in condensed matter physics S. J. Blundell 13

Figure 17: (a) The field distribution p(B) in the vortex

lattice (contours of B shown in (b)).

Figure 18: The Uemura plot, showing the correlation

lic phase for B > Bc2 . In highly anisotropic systems between the superconducting transition temperature Tc
the vortex lattice is no longer a system of rigid rods and the Fermi temperature TF . The exotic supercon-
but should be considered as a system of flexible inter- ductors fall within a common band. Adapted from Ue-
acting lines. A useful picture is that of a weakly cou- mura et al. 1991 and Hillier and Cywinski 1997.
pled stack of quasi-two-dimensional (q2D) “pancake”
vortices, each one confined to a superconducting plane
(Blatter et al. 1995, Clem 1991). The phase diagram is ous anisotropic superconductors using µSR, including
thus substantially altered to take account of field and the high temperature superconductors (Lee et al. 1993,
temperature dependent changes in the vortex lattice Aegerter and Lee 1997) and also in organic supercon-
itself. At low T and low B the stacks resemble conven- ductors (Lee et al. 1997). In both case it is found that
tional vortex lines. Above a characteristic temperature the vortex lattice can be melted with temperature at
Tb , but still below that at which superconductivity is Tb or can cross into a two-dimensional regime at fields
destroyed, the vortex lattice is broken up by thermal above Bcr . Both transitions can be followed by measur-
fluctuations (Clem 1991) (this is called vortex lattice ing the field and temperature dependence of the p(B)
melting). At low T , but this time increasing B, the line shapes.
energetic cost of interlayer deformations of the lattice The London formula for the zero temperature limit of
(local tilting of the lines) is progressively outweighed the penetration depth λ(0) yields in the clean limit (the
by the cost of intralayer deformations within the super- mean free path muchp bigger than the coherence length)
conducting plane (shearing). Above a crossover field the relation: λ(0) ∝ m∗ /ns (0) where m∗ is the effec-
Bcr the vortex lattice enters a more two-dimensional tive mass and ns (0) is the density of superconducting
regime. Thus in anisotropic systems we may expect electrons. This can be combined with measurements of
temperature and field dependent transitions in which the Sommerfeld constant to yield a value for the Fermi
the vortex lattice is destroyed. When muons are im- temperature TF . Thus from muon measurements it is
planted into a superconductor in a field Bapplied one possible to plot a diagram showing the relationship be-
can directly measure the field distribution p(B) which tween the Fermi and critical temperatures for a range
is given by p(B) = hδ(B − B(r))ir and is the proba- of superconductors (Uemura et al. 1991). This picture
bility that a randomly chosen point in the sample has (see Figure 18) has come to be called a Uemura plot
field B ′ (Brandt 1988, Aegerter and Lee 1997). This and shows a clear correlation between Tc and TF for the
is shown in Fig. 17(a) for an ideal vortex line lattice. heavy fermion, organic, fullerene and Chevrel phase su-
The distribution is highly asymmetric, the high field perconductors. The conventional, elemental supercon-
“tail” corresponding to regions of the lattice close to ductors lie away from this correlation and have values
the vortex cores (see Fig. 17(b)). The maximum of of Tc /TF ∼ 10−3 (for the “exotic” superconductors this
the distribution occurs at Bpk , which lies below the value is one or two orders of magnitude larger). This
mean field hBi (see Fig. 17(a)). Such lineshapes have correlation has been interpreted as evidence that the
been observed at low temperatures and fields in vari- exotic superconductors may be close to Bose-Einstein
 Spin-polarized muons in condensed matter physics S. J. Blundell 14

condensation which is expected to occur at a tempera-

ture ∼ TF (Uemura et al. 1991). Whether or not this
speculation is correct, it is expected that this remark-
able correlation should constrain theories to explain the
superconductivity in these various exotic systems and
is possibly suggestive of a common mechanism lying
behind them.

8 Muons as active probes Figure 19: Muonium interaction with trans-

polyacetylene to produce a diamagnetic radical and a
In almost everything we have considered so far we have
mobile neutral soliton.
been tacitly assuming that the muon is a passive probe
and does not disturb its surroundings. If it has been
sensitive to dynamics, we have believed that the muon
In more conventional materials it is the motion of the
takes no part in them itself. This in fact is very of-
muon itself which is of special interest. The dynamics
ten true. However, there are a number of situations
of light atoms such as hydrogen and muonium, or par-
in which the muon plays an active rôle. For example,
ticles such as the proton and muon, are worthwhile to
at high temperature in copper the muon diffuses from
study because they can provide a stringent test for theo-
interstitial site to interstitial site. In this case the ma-
ries on the quantum motion of defects and interstitials
jor component of the observed depolarization is due to
(Storchak and Prokof’ev 1998). The smaller mass of
the muon motion. In semiconductors the muonium can
the muon leads to larger tunnelling matrix elements to
undergo charge and spin exchange with conduction elec-
neighbouring sites and thus enhances the quantum me-
trons and thereby one measures dynamics with which
chanical nature of the motion. Furthermore, because
the muon itself is intimately involved.
muon-muon (or muonium-muonium) interactions can
An extreme case where the muon plays a strongly ac-
be neglected, the intrinsic nature of the dynamics can
tive rôle is found in conducting polymers (Hayes 1995).
be followed without the complications that can be found
Figure 19 shows the reaction between muonium and
in studying the corresponding proton or hydrogen case.
trans-polyacetylene (Nagamine et al. 1984) which pro-
duces a diamagnetic, neutral muon defect and a highly In inorganic materials the muon will usually come
mobile unpaired spin. This soliton diffuses up and down to rest at an interstitial site. The stability of that
the chain but cannot cross the muon defect which acts site will depend on the depth of the potential well.
as a barrier. Every time the soliton briefly revisits the It is of interest to discover whether local diffusion is
muon, the muon-electron hyperfine coupling is turned possible between interstitial sites. Another process is
on and then off, so that successive visits progressively trapping and release from deep potential wells associ-
relax the muon polarization. Measurement of the field ated with imperfections or defects. The muon jump
dependence of this relaxation yields the spectral den- rates are found to be about ten times higher than the
sity function associated with the defect random walk corresponding proton jump rates, consistent with the
and can be used to infer the dimensionality of the soli- lighter muon mass. Hopping is, as expected, assisted by
ton diffusion (Nagamine et al. 1984). This occurs be- phonons and thus rises with temperature, following an
cause the relaxation rate is connected with the noise approximately activated behaviour [the hop rates are
power, J(ωµ ), in the fluctuating magnetic field at the proportional to T −1/2 exp(−Ea /kB T ) where Ea is an
muon Larmor frequency, ωµ , associated with that par- activation energy (Flynn and Stoneham 1972)]. This
ticular magnetic field. Sweeping the magnetic field al- occurs because the muon is initially ‘self-trapped’ by
lows one to extract the frequency distribution of the its own local distortion of the lattice (Figure 20) and
fluctuations. In other polymers, such as polyaniline, a tunnelling transition is only possible if, by the ther-
the muon-induced defect is a negatively charged po- mal fluctuations of the lattice which occur because of
laron. Muons are uniquely sensitive to the motion of phonons, two neighbouring energy levels coincide [the
this defect in undoped materials (Pratt et al. 1997) and coincidence configuration, see Figure 20(b)]. The muon
in contrast to transport studies, which are inevitably can then tunnel through the barrier and becomes self-
dominated by the slowest component of the transport trapped in the next site.
process, can provide information on the intrinsic trans- Lower temperature produces fewer phonons and
port processes governing the mobility of an electronic hence the hop rate falls as the temperature is reduced.
excitation along a chain. However as the temperature is lowered further, a pe-
 Spin-polarized muons in condensed matter physics S. J. Blundell 15

Figure 20: Diffusion process of a muon by phonon-

assisted tunnelling. (a) The muon is stable in an inter-
stitial site and the local distortion leads to self-trapping
so that its zero-point energy level lies a little lower than
the neighbouring site. (b) Thermal fluctuations provide
the opportunity for a coincidence configuration whereby
tunnelling is allowed, leading to (c) a new stable con-
figuration. (Adapted from Cox et al. 1987).

culiar effect is observed: the muon hop rate falls to a

minimum and then begins to rise again. In the very low
temperature regime the phonons appear to be hinder-
ing hopping rather than helping it, as they do at higher
temperatures. The reason is that at low temperatures Figure 21: Muon hop rates as a function of tempera-
coherent tunnelling is possible i.e. the muon is in a band ture for various materials. Copper (Luke et al. 1991),
state. Phonons now are responsible for inelastic scat- aluminium (Hartmann et al. 1988), gallium arsenide
tering which destroys the coherence of this delocalized (Kadono et al. 1990), potassium chloride (Kiefl et al.
state. This coherent effect is known as quantum diffu- 1989), and solid nitrogen (Storchak et al. 1994).
sion (for a review, see Storchak and Prokof’ev 1998).
Experimental data for various materials are shown
in Figure 21 and although there are large differences ence of the gap in the electronic spectrum effectively
in the size of the hop rate and the detailed form of decouples the electron bath from the muon. At low
the temperature dependence, all show an increasing temperature the superconductivity can be removed by
hop rate at high temperature consistent with activated applying a sufficiently large magnetic field to the sam-
behaviour and a decreasing hop rate at low tempera- ple. This dramatically reduces the muon diffusion rate
ture consistent with quantum diffusion. An early the- because the closing of the gap reconnects the muon dif-
oretical treatment of this latter effect predicted that fusion process to the electron bath, introducing drag.
the low temperature hop rate would follow an inverse In insulators there are no conduction electrons to
power law T −α , where the exponent α was large, typ- worry about and in this case the hopping particle is
ically ∼ 9 (Kagan and Klinger 1974). Experiments on a neutral muonium atom, not a charged muon. The
Cu and Al (see Figure 21) showed a more modest be- coherent muon hop rate for KCl (Figure 21) rises much
haviour with α ∼0.6–0.7. However it was shown that more rapidly with decreasing temperature than for met-
this could be explained by considering the effect of the als, and fits to an exponent α = 3.3 (Kiefl et al. 1989).
conduction electrons in a metal, which screen the muon This still does not quite fit with the earlier theory (Ka-
and cannot react fast enough to the diffusing particle gan and Klinger 1974) but agrees with more sophis-
and follow it adiabatically. This produces a net drag ticated treatments (see Storchak and Prokof’ev 1998)
which reduces the particle hop rate and weakens the which take into account the measured phonon spectrum
temperature dependence. A detailed theory of this ef- in KCl measured using neutron scattering.
fect produces agreement with experiment (Kondo 1984, In semiconductors muonium is also formed (see sec-
Yamada 1984, Kagan and Prokof’ev 1986). tion 5) and very similar temperature dependence is
This dominant rôle of the electrons has been demon- found (Figure 21, Kadono et al. 1990) with α ∼ 3 be-
strated by ingenious experiments on aluminium (Karls- low 100 K but the hop rate saturates below ∼10 K due
son et al. 1995). In its superconducting state, the pres- to the presence of disorder. (Below this temperature
 Spin-polarized muons in condensed matter physics S. J. Blundell 16

the coherent tunnelling is dominated by the disorder, References

rather than the phonons, and is therefore temperature
independent.) Data are also shown for solid nitrogen in Aegerter, C. M., and Lee, S. L., 1997, Appl. Mag. Res. 13,
Figure 21. This material shows a very sharp increase in 75.
hop-rate with decreasing temperature (α ∼ 7, Storchak Amato, A., 1997, Rev. Mod. Phys., 69, 1119.
et al. 1994); as the temperature is lowered further it Anderson, C. D., and Neddermeyer, S. H., 1936, Phys. Rev.,
saturates and thereafter begins to decrease. This low 50, 263.
temperature effect is thought to be due to orientational Bailey, J., Borer, K., Combley, F., Drumm, H., Krienen,
ordering of the N2 molecules but is not yet understood F., Lange, F., Picasso, E., von Ruden, W., Farley, F. J.
M., Field, J. H., Flegel, W., and Hattersley, P. M., 1977,
in detail.
Nature, 268, 301.
Blatter, G., Feigel’man, M. V., Geshkenbein, V. B., Larkin,
A. I., and Vinokur, V. M., 1995, Rev. Mod. Phys., 66,
9 Conclusions 1125.
Blundell, S. J., Pattenden, P. A., Pratt, F. L., Valladares,
Muons have found extremely wide application in con-
R. M., Sugano, T., and Hayes, W., 1995, Europhys. Lett.,
densed matter physics, allowing physicists and chemists 31, 573.
to study a variety of problems concerning magnetism, Blundell, S. J., Pratt, F. L., Pattenden, P. A., Kurmoo, M.,
superconductivity, chemical kinetics, diffusion, molec- Chow, K. H., Jestädt, T., and Hayes, W., 1997, J. Phys.:
ular dynamics, and semiconductor physics. The inter- Condens. Matter, 9, 119.
ested reader is referred to a number of excellent spe- Brandt, E. H., 1988, Phys. Rev. B, 37, 2349.
cialist reviews for further details concerning the tech- Brewer, J. H., 1994, Encyclopaedia of Applied Physics 11,
nique and its application. (Schenck 1985, Cox 1987, p23 (VCH, New York).
Brewer 1994, Karlsson 1995, Dalmas de Réotier and Campbell, I. A., Amato, A., Gygax, F. N., Herlach, D.,
Yaouanc 1997). Currently there is also active interest Schenck, A., Cywinski, R., and Kilcoyne, S. H., 1994,
in attempting to utilise muons to catalyse fusion (Pono- Phys. Rev. Lett., 72, 1291.
Caso., C., et al., 1998, The 1998 Review of Particle Physics,
marev 1990).
Eur. Phys. J., C3, 1.
The availability of muons at proton sources in vari- Chaikin, P. M., and Lubensky, T. C., 1995, Principles of
ous places in the world has led to a great deal of ac- Condensed Matter Physics Cambridge University Press
tivity in this field. This is because even though the (1995).
proton source is often designed for some other purpose Chappert, J., 1984, in Muons and Pions in Materials Re-
(the production of neutrons for scattering experiments search ed. Chappert, J. and Grynszpan, R. I., (Elsevier,
or the production of mesons for nuclear and particle 1984).
physics), a muon facility can be “tagged on” for little Chow, K. H., Lichti, R. L., Kiefl, R. F., Dunsiger, S., Estle,
additional cost. The consequent reduction in proton T. L., Hitti, B., Kadono, R., MacFarlane, W. A., Schnei-
intensity for the main purpose of the proton source is der, J. W., Schumann, D., and Shelley, M., 1994, Phys.
only a few per cent. Rev. B, 50, 8918.
Clem, J. R., 1991, Phys. Rev. B, 43, 7837.
One current limitation of muon techniques is that
Conversi, M., Pancini, E., and Piccioni, O., 1945, Phys.
they require bulk samples because the incident muons Rev., 68, 232.
are formed with energy 4 MeV and penetrate a few hun- Conversi, M., Pancini, E., and Piccioni, O., 1947, Phys.
dred micrometres into any sample. Even with the use of Rev., 71, 209.
degraders, surface studies have so far not been possible. Cox, S. F. J., 1987, J. Phys. C, 20, 3187.
Recently however there has been considerable progress Dalmas de Réotier, P., and Yaouanc, A., 1997, J. Phys.:
in the development of so-called ‘slow muon’ beams in Condensed Matter, 9, 9113.
which the energy of the muon beam is reduced down to Flynn, C. P., and Stoneham, A. M., 1972, Phys. Rev. B, 1,
∼1–10 eV. This can be achieved by either moderation 3966.
in thin layers of rare gas solid (Morenzoni et al. 1994) or Garwin, R. L., Lederman, L. M., and Weinrich, M., 1957,
by resonant ionization of thermal muonium (produced Phys. Rev., 105, 1415.
Hartmann, O., Karlsson, E., Wäckelgård, E., Wappling,
from the surface of a hot tungsten foil placed in a pulsed
R., Richter, D., Hempelmann, R., and Niinikoski, T. O.,
proton beam) by a pulsed laser source (Nagamine et al. 1988, Phys. Rev. B, 37, 4425.
1995). The efficiency of these processes is at present Hayano, R. S., Uemura, Y. J., Imazato, J., Nishida, N.,
rather too low to allow routine experiments on thin Yamazaki, T., and Kubo, R., 1979, Phys. Rev. B, 20,
films to take place. Nevertheless as moderator efficien- 850.
cies improve and the necessary technology is developed, Hayes, W., 1995, Phil. Trans. R. Soc. Lond. A, 350, 249.
this promises to be a fruitful area of future research. Hess, V. F., 1912, Phys. Z., 13, 1084.
 Spin-polarized muons in condensed matter physics S. J. Blundell 17

Hillier, A. D., and Cywinski, R., 1997, Appl. Mag. Res. 13, Morenzoni, E., Kottmann, F., Maden, D., Matthias, B.,
95. Meyberg, M., Prokscha, T., Wutzke, T., and Zimmer-
Kadono, R., Kiefl, R. F., Brewer, J. H., Luke, G. M., Pfiz, mann, U., 1994, Phys. Rev. Lett., 72, 2793.
T., Riseman, T. M., and Sternlieb, B. J., 1990, Hyp. Int., Nagamine, K., Ishida, K., Matsuzaki, T., Nishiyama, K.,
64, 635. Kuno, Y., and Yamazaki, T., 1984, Phys. Rev. Lett., 53,
Kagan, Yu., and Klinger, M. I., 1974, J. Phys. C, 7, 2791. 1763.
Kagan, Yu., and Prokofe’v, N. V., 1986, JETP, 63, 1276. Nagamine, K., Miyake, Y., Shimomura, K., Birrer, P.,
Karlsson, E. B., 1995 Solid State Phenomena, Oxford. Iwasaki, M., Strasser, P., and Kuga, T., 1995, Phys. Rev.
Karlsson, E., Wäppling, R., Lidström, S. W., Hartmann, O., Lett., 74, 4811.
Kadono, R., Kiefl, R. F., Hempelmann, R., and Richter, Patterson, B. D., 1988, Rev. Mod. Phys., 60, 69.
D., 1995, Phys. Rev. B, 52, 6417. Ponomarev, L. I., 1990, Contemp. Phys., 31, 219.
Keren, A., Le, L. P., Luke, G. M., Sternlieb, B. J., Wu, Pratt, F. L., 1997, Phil. Mag. Lett., 75, 371.
W. D., Uemura, Y. J., Tajima, S., and Uchida, S., 1993, Pratt, F. L., Blundell, S. J., Hayes, W., Ishida, K.,
Phys. Rev. B, 48, 12926. Nagamine, K., and Monkman, A. P., 1997, Phys. Rev.
Keren, A., Mendels, P., Campbell, I. A., and Lord, J., 1996, Lett., 79, 2855.
Phys. Rev. Lett., 77, 1386. Reid, I. D., Azuma, T., and Roduner, E., 1990, Nature, 345,
Kiefl, R. F., Kadono, R., Brewer, J. H., Luke, G. M., Yen, 328.
H. K., Celio, M., and Ansaldo, E. J., 1989, Phys. Rev. Roduner, E., 1993, Chem. Soc. Rev., 22, 337.
Lett., 53, 90. Rossi, B., and Hall, D. B., 1941, Phys. Rev., 59, 223.
Kiefl, R. F., Schneider, J. W., MacFarlane, A., Chow, K., Schenck, A., and Gygax, F. N., 1995, in Handbook of Mag-
Duty, T. L., Estle, T. L., Hitti, B., Lichti, R. L., Ansaldo, netic Materials 9 edited by Buschow, K. H. J., (Elsevier).
E. J., Schwab, C., Percival, P. W., Wei, G., Wlodek, S., Schenck, A., 1985, Muon Spin Rotation: Principles and Ap-
Kojima, K., Romanow, W. J., McCauley, J. P., Coustel, plications in Solid State Physics (Adam Hilger, Bristol).
N., Fischer, J. E., and Smith, A. B., 1992, Phys. Rev. Sonier, J. E., Kiefl, R. F., Brewer, J. H., Bonn, D. A., Car-
Lett., 68, 2708. olan, J. F., Chow, K. H., Dosanjh, P., Hardy, W. N.,
Kojima, K., Keren, A., Luke, G. M., Nachumi, B., Wu, W. Liang, R., MacFarlane, W. A., Mendels, P., Morris, G.
D., Uemura, Y. J., Azuma, M., and Takano, M., 1995, D., Riseman, T. M., and Schneider, J. W., 1994, Phys.
Phys. Rev. Lett., 74, 2812. Rev. Lett., 72, 744.
Kondo, J., 1984, Physica B+C, 126, 377. Sonier, J. E., Brewer, J. H., Kiefl, R. F., Bonn, D. A., Dun-
Kubo, R., and Toyabe, T., 1967, in Magnetic Resonance and siger, S. R., Hardy, W. N., Liang, R., MacFarlane, W. A.,
Relaxation, ed. Blinc, R., p810, (North-Holland, Amster- Miller, R. I., Riseman, T. M., Noakes, D. R., Stronach,
dam). C. E., and White Jr., M. F., 1997, Phys. Rev. Lett., 79,
Lappas, A., Prassides, K., Amato, A., Feyerherm, R., Gy- 2875.
gax, F. N., and Schenck, A., 1994, Z. Phys. B, 96, 223. Storchak, V. G., Brewer, J. H., and Morris, G. D., 1994,
Lattes, C. M. G., Occhialini, G. P. S., and Powell, C. F., Phys. Lett. A, 193, 199.
1947, Nature, 160, 453, 486. Storchak, V. G., and Prokof’ev, N. V., 1998, Rev. Mod.
Le, L. P., Keren, A., Luke, G. M., Sternlieb, B. J., Wu, Phys., 70, 929.
W. D., Uemura, Y. J., Brewer, J. H., Riseman, T. M., Uemura, Y. J., Yamazaki, T., Harshman, D. R., Senba, M.,
Uspani, R. V., Chiang, L. Y., Kang, W., Chaikin, P. M., and Ansaldo, E. J., 1985, Phys. Rev. B, 31, 546.
Csiba, T., and Grüner, G., 1993, Phys. Rev. B, 48, 7284. Uemura, Y. J., Le, L. P., Luke, G. M., Sternlieb, B. J., Wu,
7284 (1993). W. D., Brewer, J. H., Riseman, T. M., Seaman, C. L.,
Lee, S. L., Zimmermann, P., Keller, H., Warden, M., Savić, Maple, M. B., Ishikawa, M., Hinks, D. G., Jorgensen, J.
I. M., Schauwecker, R., Zech, D., Cubitt, R., Forgan, E. D., Saito, G., and Yamochi, H., 1991, Phys. Rev. Lett.,
M., Kes, P. H., Li, T. W., Menovsky, A. A., and Tar- 66, 2665.
nawski, Z., 1993, Phys. Rev. Lett., 71, 3862. Uemura, Y. J., Keren, A., Kojima, K., Le, L. P., Luke,
Lee, S. L., Pratt, F. L., Blundell, S. J., Aegerter, C. M., G. M., Wu, W. D., Ajiro, Y., Asano, T., Kuriyama, Y.,
Pattenden, P. A., Chow, K. H., Forgan, E. M., Sasaki, Mekata, M., Kikuchi, H., and Kakurai, K., 1994, Phys.
T., Hayes, W., and Keller, H., 1997, Phys. Rev. Lett., Rev. Lett., 73, 3306.
79, 1563. Wulf, T., 1910, Phys. Z., 10, 811.
Luke, G. M., Brewer, J. H., Kreitzman, S. R., Noakes, D. R., Yamada, K., 1984, Prog. Theor. Phys., 72, 195.
Celio, M., Kadono, R., and Ansaldo, E. J., 1991, Phys. Yaouanc, A., Dalmas de Réotier, P., and Brandt, E. H.,
Rev. B, 43, 3284. 1997, Phys. Rev. B, 55, 11107.
MacFarlane, W. A., Kiefl, R. F., Dunsiger, S., Sonier, J. E., Xu, Q., and Brown, L. M., 1987, Amer. J. Phys., 55, 23.
Chakhalian, J., Fischer, J. E., Yildirim, T., and Chow,
K. H., 1998, Phys. Rev. B, 58, 1004.
Major, J., Mundy, J., Schmolz, M., Seeger, A., Döring, K.
P., Fürderer, K., Gladisch, M., Herlach, D., and Majer,
G., 1986, Hyp. Int., 31, 259.