1. High sensitivity thermoluminescence dosimetry
A.J.J. Bos *
Interfaculty Reactor Institute, Delft University of Technology, Mekelweg 15, 2629 JB Delft, The Netherlands
Received 23 February 2001; received in revised form 30 April 2001
Abstract
This paper reviews the physics of the phenomenon of thermoluminescence (TL) related to dosimetric applications.
Basic concepts are given using the simple model of one trapĀ±one recombination centre. General characteristics of
thermoluminescence dosimetry (TLD) materials are reviewed. Two high sensitivity TL materials are discussed in detail
namely LiF:Mg, Cu, P and a-Al2O3:C. What is understood and what knowledge is still lacking of the TL mechanism in
both materials is indicated. Field measurements show that in spite of incomplete understanding of the TL mechanism,
both materials can be used to measure very low doses in a reliable way. Ć 2001 Elsevier Science B.V. All rights reserved.
PACS: 78.60.K; 87.56; 87.53.B
Keywords: Thermoluminescence dosimetry; Sensitivity; TL materials; LiF:Mg, Cu, P; Al2O3:C
1. Introduction
The phenomenon thermoluminescence (TL) has
been known for a long time. The Ā®rst application
of this phenomenon for dosimetric purposes was
from Daniel et al. [1]. Since then much research
has been carried out for a better understanding
and improvement of the material characteristics as
well as to develop new TL materials. Nowadays,
thermoluminescence dosimetry (TLD) is a well-
established dosimetric technique with applications
in areas such as personnel, environmental and
clinical dosimetry.
TLD is based on materials which (after expo-
sure to ionising radiation) emit light while they are
heated. It is believed that the impurities in the TL
material give rise to localised energy levels within
the forbidden energy band gap and that these are
crucial to the TL process. As a means of detecting
the presence of these defect levels, the sensitivity of
TL is unrivalled. Townsend and Kelly [2] estimate
that the technique is capable of detecting as few as
109
defects levels in a specimen. To put this num-
ber into perspective one should realise that de-
tectable chemical `purity' in a sample is six orders
of magnitude higher. The high sensitivity, on the
one hand, allows the determination of very low
radiation doses. Several examples of it will be
given later in this paper. On the other hand, it
hampers us in investigation into the relation be-
tween the luminescence and the defects involved in
Nuclear Instruments and Methods in Physics Research B 184 (2001) 3Ā±28
www.elsevier.com/locate/nimb
*
Tel.: +31-15-278-4705; fax: +31-15-278-6422.
E-mail address: bos@iri.tudelft.nl (A.J.J. Bos).
0168-583X/01/$ - see front matter Ć 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 1 6 8 - 5 8 3 X ( 0 1 ) 0 0 7 1 7 - 0
2. this process. It also causes the TLD properties to
be highly sensitive to many parameters.
The purpose of this review is to give insight into
the physics of the phenomenon of TL with the
dosimetric application in the back of one's mind.
First, the basic concepts of TL will be given. Next,
general characteristics of TLD materials are re-
viewed. Finally, we concentrate on two high sen-
sitivity TL materials notably LiF:Mg, Cu, P and
Al2O3:C which have received a lot of attention
during the last decade and show excellent dosi-
metric properties.
The treatment is such that insight is given into
the factors which inĀÆuence the TL intensity and
thus the value of the absorbed dose that has to be
determined. It is meant to give the dosimetrist the
necessary background information as to why the
TLD procedures are as they are.
Much of the contents of this review can be
found in text books on TL and its application such
as those of [3,4]. Text books speciĀ®c on TL ma-
terials are published by Vij [5] and McKeever et al.
[6]. With special attention the author mentions a
recent excellent text book on the theory of TL and
related phenomena by Chen and McKeever [7]
which has been consulted and cited many times in
this review.
2. Basic concepts of thermoluminescence
TL is a luminescence phenomenon of an insu-
lator or semiconductor which can be observed
when the solid is thermally stimulated. TL should
not be confused with the light spontaneously
emitted from a substance when it is heated to in-
candescence. At higher temperatures (say in excess
of 200Ā°C) a solid emits (infra) red radiation of
which the intensity increases with increasing tem-
perature. This is thermal or black body radiation.
TL, however, is the thermally stimulated emission
of light following the previous absorption of en-
ergy from radiation. From this description the
three essential ingredients necessary for the pro-
duction of TL can be deduced. Firstly, the material
must be an insulator or a semiconductor Ā± metals
do not exhibit luminescent properties. Secondly,
the material must have at some time absorbed
energy during exposure to ionising radiation.
Thirdly, the luminescence emission is triggered by
heating the material [4]. A thermoluminescent
material is thus a material that during exposure to
ionising radiation absorbs some energy which is
stored. The stored energy is released in the form of
visible light when the material is heated. Note that
TL does not refer to thermal excitation, but to
stimulation of luminescence in a sample which was
excited in a diā¬erent way. This means that a TL
material cannot emit light again by simply cooling
the sample and reheating it another time. It should
Ā®rst be re-exposed to ionising radiation before it
produces light again. The storage capacity of a TL
material makes it in principle suitable for dosi-
metric applications.
2.1. The one trapĀ±one centre model
An explanation of the observed TL properties
can be obtained from the energy band theory of
solids. In an ideal crystalline semiconductor or
insulator most of the electrons reside in the valence
band. The next highest band that the electrons can
occupy is the conduction band, separated from the
valence band by the so-called forbidden band gap.
The energy diā¬erence between the delocalised
bands is Eg. However, whenever structural defects
occur in a crystal, or if there are impurities within
the lattice, there is a possibility for electrons to
possess energies which are forbidden in the perfect
crystal. In a simple TL model two levels are as-
sumed, one situated below the bottom of the
conduction band and the other situated above the
top of the valence band (see Fig. 1). The highest
level indicated by T is situated above the equilib-
rium Fermi level (Ef ) and thus empty in the equi-
librium state, i.e. before the exposure to radiation
and the creation of electrons and holes. It is
therefore a potential electron trap. The other level
(indicated by R) is a potential hole trap and can
function as a recombination centre. The absorp-
tion of radiant energy with hm > Eg results in
ionisation of valence electrons, producing ener-
getic electrons and holes which will, after ther-
malization, produce free electrons in the
conduction band and free holes in the valence
band (transition a). The free charge carriers re-
4 A.J.J. Bos / Nucl. Instr. and Meth. in Phys. Res. B 184 (2001) 3Ā±28
3. combine with each other or become trapped. In
the case of direct recombination an amount of
energy will be released which may excite a lumi-
nescent centre (which may coincide with the re-
combination centre). The luminescent centre
relaxes (returns to the ground state) under the
emission of light. The phenomenon of direct
(<10Ć8
s) recombination of free electrons and
holes under emission of light is called radiolumi-
nescence. However, in semiconductors and insu-
lators a certain percentage of the charge carriers is
trapped: the electrons at T and the holes at R
(transition b). The probability per unit time of
release of an electron from the trap is assumed to
be described by the Arrhenius equation,
p Ė s exp
&
Ć
E
kT
'
; ā¦1ā
where p is the probability per unit time. The
terms s is called the frequency factor or attempt-
to-escape factor. In the simple model s is con-
sidered as a constant (not temperature dependent)
with a value in the order of the lattice vibration
frequency, namely 1012
Ā±1014
sĆ1
. E is called the
trap depth or activation energy, the energy nee-
ded to release an electron from the trap into the
conduction band (see Fig. 1). The other symbols
have their usual meaning: k Ė Boltzmann's con-
stant Ė 8:617 Ć 10Ć5
eV/K, and T the absolute
temperature. If the trap depth E ) kT0, with T0
the temperature at irradiation, then any electron
that becomes trapped will remain so for a long
period of time, so that even after exposure to the
radiation there will exist a substantial population
of trapped electrons. Furthermore, because the
free electrons and holes are created and annihi-
lated in pairs, there must be an equal population
of trapped holes at level R. Because the normal
equilibrium Fermi level Ef is situated below level
T and above level R, these populations of trap-
ped electrons and holes represent a non-equilib-
rium state. The reaction path for return to
equilibrium is always open, but because the per-
turbation from equilibrium (during exposure to
ionising radiation) was performed at low tem-
perature (compared to E=k), the relaxation rate as
determined by Eq. (1) is slow. Thus, the non-
equilibrium state is metastable and will exist for
an indeĀ®nite period, governed by the rate pa-
rameters E and s.
The return to equilibrium can be speeded up by
raising the temperature of the TL material above
T0. This will increase the probability of detrapping
and the electrons will now be released from the
trap into the conduction band. The charge carrier
migrates through the conduction band of the
crystal until it undergoes recombination at re-
combination centre R. In the simple model this
recombination centre is a luminescent centre where
the recombination of the electron and hole leaves
the centre in one of the higher excited states. Re-
turn to the ground state is coupled with the
emission of light quanta, i.e. TL. The intensity of
TL Iā¦tā in photons per second at any time t during
heating is proportional to the rate of recombina-
tion of holes and electrons at R. If m ā¦mĆ3
ā is the
concentration of holes trapped at R the TL in-
tensity can be written as
Iā¦tā Ė Ć
dm
dt
: ā¦2ā
Fig. 1. Energy band model showing the electronic transitions in
a TL material according to a simple two-level model: (a) gen-
eration of electrons and holes; (b) electron and hole trapping;
(c) electron release due to thermal stimulation; (d) recombina-
tion. Solid circles are electrons, open circles are holes. Level T is
a electron trap, level R is a recombination centre, Ef is Fermi
level.
A.J.J. Bos / Nucl. Instr. and Meth. in Phys. Res. B 184 (2001) 3Ā±28 5
4. Here we assume that each recombination produces
a photon and that all produced photons are de-
tected. The rate of recombination will be propor-
tional to the concentration of free electrons in
the conduction band nc and the concentration of
holes m,
Iā¦tā Ė Ć
dm
dt
Ė ncmA; ā¦3aā
with the constant A the recombination probability
expressed in units of volume per unit time which is
assumed to be independent of the temperature.
The rate of change of the concentration of trapped
electrons n is equal to the rate of thermal release
minus the rate of retrapping,
Ć
dn
dt
Ė np Ć ncā¦N Ć nā Ar; ā¦3bā
with N the concentration of electron traps and Ar
the probability of retrapping ā¦m3
=sā . Likewise the
rate concentration of free electrons is equal to the
rate of thermal release minus the rate of retrapping
and the rate of recombination,
dnc
dt
Ė np Ć ncā¦N Ć nā Ar Ć ncmA: ā¦3cā
Eqs. (3a)Ā±(3c) described the charge carrier traĀc in
the case of release of a trapped electron from a
single-electron trap and recombination in a single
centre. For TL produced by the release of holes the
rate equations are similar to Eqs. (3a)Ā±(3c). These
equations form the basis of many analyses of TL
phenomena. There is no general analytical solu-
tion. To develop an analytical expression some
simplifying assumptions must be made. An im-
portant assumption is that at any time
dnc
dt
(
dn
dt
;
dnc
dt
(
dm
dt
: ā¦4ā
This assumption is called by Chen and McKeever
[7] the quasiequilibrium assumption since it re-
quires that the free electron concentration in the
conduction band is quasistationary. The trapped
electrons and holes are produced in pairs during
the irradiation. Charge neutrality dictates there-
fore
nc ā” n Ė m; ā¦5ā
which for nc % 0 means that n % m and
Iā¦tā Ė Ć
dm
dt
% Ć
dn
dt
: ā¦6ā
Since dnc=dt % 0 one gets from Eqs. (3a) and (3b):
Iā¦tā Ė
mAns exp Ć E
kT
Ć Ć
ā¦N Ć nā Ar ā” mA
: ā¦7ā
2.1.1. First-order kinetics
Even Eq. (7) cannot be solved analytically
without additional simplifying assumptions. Ran-
dall and Wilkins [8,9] assumed negligible retrap-
ping during the heating stage, i.e. they assumed
mA ) ā¦N Ć nā Ar. Under this assumption Eq. (7)
can be written as
Iā¦tā Ė Ć
dn
dt
Ė sn exp
Ć
E
kT
'
: ā¦8ā
This diā¬erential equation describes the charge
transport in the lattice as a Ā®rst-order process and
the glow peaks calculated from this equation are
called Ā®rst-order glow peaks. Solving the diā¬eren-
tial equation (8) yields
ā¦tā Ė Ć
dn
dt
Ė n0s exp
Ć
E
kT
'
Ć exp
Ć s
Ā t
0
exp
Ć
E
kTā¦tHā
'
dtH
'
; ā¦9ā
where n0 is the total number of trapped electrons
at time t Ė 0. Usually the temperature is raised as
a linear function of time according to
Tā¦tā Ė T0 ā” bt; ā¦10ā
with b the constant heating rate and T0 the tem-
perature at t Ė 0. This gives for the intensity as
function of temperature
Iā¦Tā Ė Ć
1
b
dn
dt
Ė n0
s
b
exp
Ć
E
kT
'
Ć exp
Ć
s
b
Ā T
T0
exp
Ć
E
kTH
'
dTH
'
: ā¦11ā
This is the well-known RandallĀ±Wilkins Ā®rst-order
expression of a single glow peak. The peak has a
6 A.J.J. Bos / Nucl. Instr. and Meth. in Phys. Res. B 184 (2001) 3Ā±28
5. characteristic asymmetric shape being wider on the
low temperature side than on the high temperature
side. On the low temperature side, i.e. in the initial
rise of the glow peak, the intensity is dominated by
the Ā®rst exponential (expā¦ĆE=kTā ). Thus, if I is
plotted as function of 1=T, a straight line is ex-
pected in the initial rise temperature range, with
the slope of ĆE=k, from which the activation en-
ergy E is readily found.
The properties of the RandallĀ±Wilkins equation
are illustrated in Fig. 2. In Fig. 2(a) it is shown
how Iā¦Tā varies if n0 varies from n0 Ė 0:25 mĆ3
till
n0 Ė 2 mĆ3
while E Ė 1 eV, s Ė 1:0 Ć 1012
sĆ1
and
b Ė 1 K=s are kept constant. It can be noted that
the temperature at the peak maximum, Tm, stays
Ā®xed. This is a characteristic of all Ā®rst-order TL
curves. The condition for the maximum can be
found by setting dI=dt Ė 0 (or, somewhat easier
from d ln Iā¦Tā =dt Ė 0). From this condition one
gets
bE
kT2
m
Ė s exp
Ć
E
kTm
'
: ā¦12ā
In this equation n0 does not appear which shows
that Tm does not depend on n0. From Fig. 2(a) it
further can be seen that not only the peak height at
the maximum but each point of the curve is pro-
portional to n0. In the application in dosimetry n0
Fig. 2. Properties of the RĀ±W Ā®rst-order TL equation, showing: (a) variation with n0; the concentration of trapped charge carriers after
irradiation; (b) the variation with E, the activation energy; (c) the variation with s, the escape frequency; (d) the variation with b, the
heating rate. Parameter values: n0 Ė 1 mĆ3
; E Ė 1 eV; s Ė 1 Ć 1012
sĆ1
; b Ė 1 K=s of which one parameter is varied while the others
are kept constant.
A.J.J. Bos / Nucl. Instr. and Meth. in Phys. Res. B 184 (2001) 3Ā±28 7
6. is the parameter of paramount importance since
this parameter is proportional to the absorbed
dose. It is simple to see that the area under the
glow peak is equal to n0 since
Ā I
0
Iā¦tā dt Ė Ć
Ā I
0
dn
dt
dt Ė Ć
Ā nI
n0
dn Ė n0 Ć nI
ā¦13ā
and nI is zero for t 3 I.
In Fig. 2(b) the activation energy E has been
varied from 0.8 to 1.2 eV. As E increases the
peak shifts to higher temperatures with a de-
crease in the height and an increase in the
width keeping the area (i.e. n0) constant. Similar
changes can be noticed as s is varied (see
Fig. 2(c)) but now in the opposite way: as s
increases the peak shifts to lower temperatures
with an increase of the height and a decrease in
width. In Fig. 2(d) the heating rate has been
varied. As b increases the peak shifts to higher
temperatures while the height decreases and the
width increases just as in the case of decreasing
s. This can be expected since s and b appear as
a ratio s=b in Eq. (11). It is worthwhile to note
that of the four parameters the activation en-
ergy E and the frequency factor s are the main
physical parameters. They are called the trap-
ping parameters and are Ā®xed by the properties
of the trapping centre. The other two parame-
ters can be chosen by the experimenter by
choosing a certain dose ā¦n0ā and by read-out of
the signal at a certain heating rate b. Investi-
gation of a new TL material will therefore start
with studying the glow peak behaviour under
variation of the absorbed dose and the heating
rate.
The evaluation of Eq. (11) is hampered by the
fact that the integral on the right-hand side is not
elementary in the case of linear heating. Chen [10]
has shown how the integral can be approximated
by asymptotic series. In practical applications it is
convenient to describe the glow peak in terms of
parameters which are easy to derive experimen-
tally, namely the intensity of peak at the maximum
Im and the temperature at the maximum Tm. Kitis
et al. [11] have shown that Eq. (11) can be quite
accurate approximated by
Iā¦Tā Ė Im exp 1 ā”
E
kT
T Ć Tm
Tm
Ć
T2
T2
m
Ć exp
E
kT
T Ć Tm
Tm
'
ā¦1 Ć Dā Ć Dm
!
; ā¦14ā
with D Ė 2kT=E and Dm Ė 2kTm=E. Recently
Pagonis et al. [12] have shown that a Weibull
distribution function also accurately describes the
Ā®rst-order TL curve. These expressions may be
convenient for peak Ā®tting purposes.
2.1.2. Second-order kinetics
Garlick and Gibson [13] considered the
possibility that retrapping dominates, i.e.
mA ( ā¦N Ć nā Ar. Further they assume that the
trap is far from saturation, i.e. N ) n and n Ė m.
With these assumptions, Eq. (7) becomes
Iā¦tā Ė Ć
dn
dt
Ė s
A
NAr
n2
exp Ć
E
kT
'
: ā¦15ā
We see that now dn=dt is proportional to n2
which
means a second-order reaction. With the addi-
tional assumption of equal probabilities of re-
combination and retrapping, A Ė Ar, integration
of Eq. (15) gives
Iā¦Tā Ė
n2
0
N
s
b
exp
Ć
E
kT
'
Ć 1 ā”
n0s
Nb
Ā T
T0
exp
Ć
E
kTH
'
dTH
!Ć2
: ā¦16ā
This is the GarlickĀ±Gibson TL equation for sec-
ond-order kinetics. The main feature of this curve
is that it is nearly symmetric, with the high tem-
perature half of the curve slightly broader than the
low temperature half. This can be understood
from the consideration of the fact that in a second-
order reaction signiĀ®cant concentrations of re-
leased electrons are retrapped before they recom-
bine, in this way giving rise to a delay in the
luminescence emission and spreading out of the
emission over a wider temperature range. The
initial concentration n0 appears here not merely as
a multiplicative constant as in the Ā®rst-order case,
so that its variation at diā¬erent dose levels change
the shape of the whole curve. This is illustrated in
Fig. 3(a). It is seen that Tm decreases as n0 in-
8 A.J.J. Bos / Nucl. Instr. and Meth. in Phys. Res. B 184 (2001) 3Ā±28
7. creases. It can be derived [14] that the temperature
shift can be approximated by
T1 Ć T2 % T1T2
k
E
ln f ; ā¦17ā
where T1 is the temperature of maximum intensity
at a certain dose and T2 the temperature of
maximum intensity at f times higher dose. With
the parameter values of Fig. 3(a) the shift is 25 K.
When E Ė 1 eV, T1 Ė 400 K and the absorbed
dose is increased by a factor 1000, which is easy
to realise experimentally, a temperature shift of
77 K can be expected. From Eq. (17) it follows
further that for a given increase of the dose the
shallower the trap, i.e., the smaller E, the larger
the peak shift. Fig. 3(b) illustrates the variation in
size and position of a second-order peak as
function of E, in Fig. 3(c) as function of s=N, and
in Fig. 3(d) as function of the heating rate. The
area under the curve is, as in the case of Ā®rst-
order kinetics, proportional to the initial con-
centration n0 but the peak height is no longer
directly proportional to the peak area, although
the deviation is small.
Note that, similarly to the Ā®rst-order case, the
term dominating the temperature dependence in
the initial rise is expā¦ĆE=kT ā . So the `initial rise
method' for the determination of the trap depth
can be applied here as well.
Also for second-order kinetics the glow peak
shape, Eq. (16), can be approximated with a
function written in terms of maximum peak in-
tensity Im and the maximum peak temperature Tm
[11],
Fig. 3. Properties of the GarlickĀ±Gibson second-order TL equation, showing: (a) variation with n0, the concentration of trapped charge
carriers after irradiation; (b) the variation with E, the activation energy; (c) the variation with s=N; (d) the variation with b, the heating
rate. Parameter values: n0 Ė 1 mĆ3
; E Ė 1 eV; s=N Ė 1 Ć 1012
sĆ1
m3
; b Ė 1 K=s of which one parameter is varied while the others are
kept constant.
A.J.J. Bos / Nucl. Instr. and Meth. in Phys. Res. B 184 (2001) 3Ā±28 9
8. Iā¦T ā Ė 4Im exp
E
kT
T ĆTm
Tm
Ć
T 2
T2
m
ā¦1ĆDā exp
E
kT
T ĆTm
Tm
'
ā”1ā”Dm
!Ć2
ā¦18ā
with D and Dm the same meaning as in Eq. (14).
2.1.3. General-order kinetics
The Ā®rst- and second-order forms of the TL
equation have been derived with the use of speciĀ®c,
simplifying assumptions. However, when these
simplifying assumptions do not hold, the TL peak
will Ā®t neither Ā®rst- nor the second-order kinetics.
May and Partridge [15] used for this case an em-
pirical expression for general-order TL kinetics,
namely
Iā¦tā Ė Ć
dn
dt
Ė nb
sH
exp
Ć
E
kT
'
; ā¦19ā
where sH
has the dimension of m3ā¦bĆ1ā
sĆ1
and b is
deĀ®ned as the general-order parameter and is not
necessarily 1 or 2. Integration of Eq. (19) for b TĖ 1
yields
Iā¦T ā Ė
sHH
b
n0 exp
Ć
E
kT
'
Ć 1ā”ā¦bĆ1ā
sHH
b
Ā T
T0
exp
Ć
E
kTH
'
dTH
!Ćb=bĆ1
;
ā¦20ā
where now sHH
Ė sH
nbĆ1
0 with unit sĆ1
. Eq. (20) in-
cludes the second-order case (b Ė 2) and reduces
to Eq. (11) when b 3 1. It should be noted that
according to Eq. (19) the dimension of sH
should be
m3ā¦bĆ1ā
sĆ1
that means that the dimension changes
with the order b which makes it diĀcult to inter-
pret physically. Still, the general-order case is
useful since intermediate cases can be dealt with
and it smoothly goes to Ā®rst- and second-orders
when b 3 1 and b 3 2, respectively (see Fig. 4).
2.2. Advanced models
The one trapĀ±one centre model shows all the
characteristics of the phenomenon TL and ex-
plains the behaviour of the glow peak shape under
variation of the dose and heating rate. However,
there is no existing TL material known that ac-
curately is described by the simple model. This
does not mean that the simple model has no
meaning. On the contrary, it can help us in the
interpretation of many features which can be
considered as variations of the one trapĀ±one centre
model. There is no room to discuss all the ad-
vanced (more realistic) models in detail. The
reader is referred to the text book of Chen and
McKeever [7] for a deeper and quantitative treat-
ment. Here, only some models are very brieĀÆy
mentioned in order to get some idea about the
complexity of the phenomenon in a real TL
material.
In general, a real TL material will show more
than one single electron trap. Not all the traps will
be active in the temperature range in which the
specimen is heated. A thermally disconnected trap
is one which can be Ā®lled with electrons during
irradiation but which has a trap depth which is
much greater than the active trap such that when
the specimen is heated only electrons trapped in
the active trap (AT) and the shallow trap (ST) (see
Fig. 5(a)) are freed. Electrons trapped in the dee-
per levels are unaā¬ected and thus this deep elec-
tron trap (indicted in Fig. 5(a) with DET) is said to
be thermally disconnected. But its existence has a
Fig. 4. Comparison of Ā®rst-order (b Ė 1), second-order (b Ė 2)
and intermediate-order (b Ė 1:3 and 1.6) TL peaks, with E Ė 1
eV, s Ė 1 Ć 1012
sĆ1
; n0 Ė N Ė 1 mĆ3
and b Ė 1 K=s (from [7]).
10 A.J.J. Bos / Nucl. Instr. and Meth. in Phys. Res. B 184 (2001) 3Ā±28
9. bearing on the trapping Ā®lling and eventually on
the shape of the glow peak [16].
In Section 2.1 it was assumed that the trapped
electrons are released during heating while the
trapped holes are stable in the recombination
centre. A description in which the holes are re-
leased and recombine at a centre where the elec-
trons are stable during heating is mathematically
identical. However, the situation will change if
both electrons and holes are released from their
traps at the same time at the same temperature
interval and the holes are being thermally released
from the same centres as are acting as recombi-
nation sites for the thermally released electrons
and vice versa (see Fig. 5(b)). In this case Eq. (2) is
no longer valid. New diā¬erential equations should
be drafted. Analysis of this complicated kinetic
model reveals a glow TL glow curve which retains
the simple RandallĀ±Wilkins (Eq. (11)) or GarlickĀ±
Gibson (Eq. (16)) shape, depending upon the
chosen values of the parameters. However, the E
and s values used in Eqs. (11) and (16) in order to
obtain a Ā®t on this complicated kinetic model need
further interpretation.
Another process which might happen is a re-
combination without a transition of the electron
into the conduction band (Fig. 5(c)). Here the
electron is thermally stimulated into an excited
state from which a transition into the recombina-
tion centre is allowed. This means that the trap has
to be in the proximity of a centre. The transition
probability may strongly depend on the distance
between the two centres. Under certain assump-
tions an expression for the TL intensity can be
derived [17] which has the same form as Eq. (11)
but with s replaced by a quantity related to the
probability for recombination. This means that
these localised transitions are governed by Ā®rst-
order kinetics.
Finally, we will mention the possibility that the
defect which has trapped the electron is not stable
but is involved in a reaction with another defect
(Fig. 5(d)). The result may be that at low tem-
perature the trap depth is changing while the
trapped electron concentration is stable. At higher
temperatures electrons are involved in two pro-
cesses: the escape to the conduction band and the
defect reaction. Piters and Bos [18] have defect
reactions incorporated into the rate equations and
glow curves simulated. It appears that the simu-
lated glow curves can be very well Ā®tted by
Eq. (11). It is clear that (again) the Ā®tting param-
eters do not have the simple meaning of trap depth
and escape frequency.
2.3. Some examples
The TLD material that has been studied most
extensively is LiF:Mg, Ti. This material is widely
used in personnel dosimetry and available on the
market under trade names like TLD-100, MTS
and DTG. A glow curve is shown in Fig. 6. The
measured curve has been analysed with a super-
position of four Ā®rst-order RandallĀ±Wilkins
(RĀ±W) equations (Eq. (11)). Also shown in the
Ā®gure is the residue, i.e. the diā¬erence between the
experimental data and the sum of the four com-
ponents. The almost random nature of the residue,
centred on a value of zero for temperatures less
than 480 K, is evidence of the excellence of this
particular Ā®t. It is tempting to draw conclusions
from this excellent Ā®t about the applied model.
Fig. 5. Advanced models describing the thermally stimulated
release of trapped charged carriers including: (a) a shallow trap
(ST), a deep electron trap (DET), and a active trap (AT); (b)
two active traps and two recombination centres; (c) localised
transitions; (d) defect interaction (trapping centre interacts with
another defect).
A.J.J. Bos / Nucl. Instr. and Meth. in Phys. Res. B 184 (2001) 3Ā±28 11
10. However, one should realise that a good Ā®t does
not mean as a matter of course that the applied
RandallĀ±Wilkins model correctly describes the TL
process in this material. There are (at least) three
reasons why the simple two level model with no
retrapping does not fully describe the process in
this material. Firstly, if the process behaves as the
RĀ±W model then one should expect that the E and
s values of each glow peak correctly describes the
thermal fading. A good TLD material is able to
store the information (trapped charge carriers)
without loss. However, there is a probability, given
by Eq. (1), that the electrons escape before the
read-out and thus the peak intensity will decrease
in course of time at a rate governed by the tem-
perature at which the sample is stored after the
irradiation and by the rate parameters E and s.
The signal is said to have faded. The half-life for
thermal fading is deĀ®ned as
t1=2 Ė
ln 2
p
Ė
ln 2
s expā¦ĆE=kTā
: ā¦21ā
The Ā®t of Eq. (11) on the measured glow curve, as
shown in Fig. 6, yields E and s values which can be
used to calculate the half-life of each glow peak.
On the other hand, t1=2 can be measured. Bos and
Piters [19] compared the measured and calculated
half-life of peak 2 at room temperature in 10 dif-
ferent production batches of TLD-100. They cal-
culated t1=2;cal Ė 19 Ć 3 days but the measured
value was a factor 24 shorter: t1=2;meas: Ė 18:6 Ć 0:9
h (see also Fig. 7). So, another process or processes
than just thermal fading may play a role in this
material.
The second reason that the TL mechanism in
TLD-100 is more complex than described by the
RĀ±W model is the behaviour of the glow curve
shape under diā¬erent experimental conditions. In
varying the absorbed dose the same peak pa-
rameters are indeed found [20] and the peak po-
sitions don't shift as expected from Ā®rst-order
kinetics. However, when the heating rate is varied
both the peak area and the activation energy (i.e.
the Ā®tting parameter E) are not constant [21]. The
remarkable thing is that at higher heating rates
Eq. (11) still is able to produce a perfect Ā®t but
with other values of the trapping parameters. The
E value (Ā®tting parameter) of the main dosimetric
peak, peak 5, for example, varies from 1.80 till
2.05 eV as the heating rate increases from 0.12 till
6.0 K/s.
Another indication that other processes are
active in TLD-100 is shown in Fig. 7. It shows the
glow curve of a sample read-out after diā¬erent
storage times (at room temperature in dark). Peak
2 has almost faded away after 14 days while peak 5
is stable and show no signiĀ®cant change in peak
area after 28 days. The behaviour of peak 4 is
striking. The Ā®tted peak height increases but the
width decreases such that the peak area keeps
constant. This suggests a process as indicated in
Fig. 5(d). The defect that constitute the trap un-
dergoes a change such that the captured electrons
(peak area) do not change but the activation en-
ergy increases (a higher E value means a smaller
peak width, see Fig. 2(b)). It is remarkable that the
RĀ±W model provides us with a glow peak shape
which perfectly Ā®ts the measured glow curve in all
cases. It means that one has to be very careful to
draw conclusions from a good Ā®t.
Fig. 6. Glow curve from LiF:Mg, Ti (TLD-100) following 45
mGy of 60
Co gamma irradiation at room temperature and read-
out at 0.24 K/s. The glow curve has been analysed into its
component peak using a superposition of peaks, each described
by Eq. (11) with E, s and peak height as Ā®tting parameters. The
solid lines are Ā®ts to the data and the dots are the measured data
points. The top Ā®gure shows the diā¬erence between the Ā®tted
and the measured values. Glow peak number 1 has already
faded (from [21]).
12 A.J.J. Bos / Nucl. Instr. and Meth. in Phys. Res. B 184 (2001) 3Ā±28
11. Finally, an example of TL in Ce3ā”
-doped
Lu2SiO5 material. This material is not used for
dosimetric purposes but as sensor material in
radiation detectors since it shows a large scintil-
lation light yield. TL has been studied to get a
better understanding of the scintillation proper-
ties [22]. It is mentioned here since at least some
glow peaks behave as expected from the simple
model. Fig. 8 shows glow curves recorded at
three diā¬erent heating rates. The shift to the peak
Fig. 7. Glow curves from LiF:Mg, Ti (TLD-100) following 60 mGy of 60
Co gamma irradiation at room temperature after a 1 h 400Ā°C
anneal with fast ā¦500Ā°C= minā cooling. The sample has been read-out at 6Ā°C/s immediately after irradiation (left Ā®gure), after 14 days
(middle Ā®gure) and after 28 days (right Ā®gure) of storage at room temperature in dark. The measured curves are Ā®tted with a su-
perposition of curves described by Eq. (11). The top Ā®gure shows the diā¬erence between the Ā®tted and the measured values.
Fig. 8. Glow curves of Lu2SiO5:Ce3ā”
after exposure with UV radiation of a Hg lamp recorded at three indicated heating rates (from
[22]).
A.J.J. Bos / Nucl. Instr. and Meth. in Phys. Res. B 184 (2001) 3Ā±28 13
12. maxima as predicted by the RĀ±W model is clearly
seen. Results of the glow curve analysis with a
superposition of RĀ±W equations are shown in
Table 1. It is seen that within the experimental
errors there is a good agreement between the E
and ln s values for peaks 1, 2 and 3. For peak 4
this does not hold. This latter peak shows, in
contrast with peaks 1, 2 and 3, a non-Ā®rst-order
behaviour if the dose is varied, i.e. the peak po-
sition of the Ā®rst three peaks stays Ā®xed but for
peak 4 the peak maximum shifts to lower tem-
peratures as the dose increases (see Fig. 9). The
trapping parameters of peak 1 were also derived
by studying the decay time s by observing the TL
signal while the sample was hold at a Ā®xed tem-
perature T. The experiment was repeated several
times at other Ā®xed temperatures. From the Ar-
rhenius plot (ln s versus 1=T) the trapping pa-
rameters were derived yielding E Ė 1:01 Ć 0:03 eV
and lnā°s ā¦sĆ1
ā Å Ė 30:8 Ć 0:9. This is in excellent
agreement with the values in Table 1 for peak 1.
Since, for peak 1, four independent experiments
(read-outs at three diā¬erent heating rates and one
isothermal decay) show the same trapping pa-
rameters it is believed that the Ā®tting parameters
E and s indeed reĀÆect the activation energy and
the escape frequency. The numerical values were
able to predict the correct fading rate. Recent
research [23] on TL measurements on cerium
doped M2SiO5 (M Ė Lu, Y, Yb, Er) shows a
common glow peak corresponding to an intrinsic
trapping site whose formation is independent of
the cerium doping and related to the host metalĀ±
ionĀ±ligand conĀ®guration. It should be noted that
such a clear RĀ±W behaviour of a glow peak is
rather exceptional.
Table 1
Activation energy E and logarithm of the frequency factor s derived with glow curve analysis using the RĀ±W equation on TL glow
curves of a Lu2SiO5:Ce3ā”
crystal recorded at three heating rates b (from [22])
E (eV) ln [s ā¦sĆ1
ā ]
Peak no. b Ė 0:24 K=s b Ė 1:0 K=s b Ė 6:0 K=s b Ė 0:24 K=s b Ė 1:0 K=s b Ė 6:0 K=s
1 0.957 0.980 0.981 29.0 29.6 29.3
2 1.174 1.171 1.173 30.6 30.5 30.5
3 1.200 1.258 1.299 25.6 27.1 28.4
4 1.1Ā±1.5 20.2Ā±26.7
Fig. 9. Glow curves of Lu2SiO5:Ce3ā”
recorded at a heating rate of 6 K/s after exposure to X-rays for exposure times of 5, 10, 20, 30, 40
and 180 s. The inset shows the temperature range 570Ā±630 K magniĀ®ed to show the shift of the peak maximum of peak 4 (from [22]).
14 A.J.J. Bos / Nucl. Instr. and Meth. in Phys. Res. B 184 (2001) 3Ā±28
13. 3. General characteristics of TLD materials
The phenomena of TL can be observed in many
materials. However, only a few materials show the
properties required for dosimetry. These require-
ments depend on the dosimetric application. TLDs
are applied in various Ā®elds each with its own de-
mands and constraints (see Table 2). In this section
some main characteristics of TL materials are gi-
ven. In Section 4 we discuss the properties of two
high sensitivity TL materials in more detail.
3.1. Sensitivity
The sensitivity of a particular TLD material is
deĀ®ned as the TL signal (peak height or TL in-
tensity integrated over an certain temperature re-
gion) per unit absorbed dose and per unit mass.
One should realise that the sensitivity, as deĀ®ned
before, depends upon the read-out system used in
the measurement. This includes many parameters
such as heating rate and light detection system, etc.
Especially the eĀciency of the light detection sys-
tem (including the geometrical light collection ef-
Ā®ciency, optical Ā®lters and detector eĀciency) are
diĀcult to measure in an absolute sense. There-
fore, one normally deĀ®nes a relative sensitivity by
comparing the TL signal from the material of in-
terest with the TL signal from LiF(TLD-100),
since this latter material is widely used. Apart from
the TL reader equipment properties and the read-
out conditions the TL signal will, of course, de-
pend on the TL material itself. Again many factors
hereby inĀÆuence the TL signal Ā± for example, the
physical form (mono- or poly-crystalline, powder,
sintered, thin chips, thick chips) and the applied
annealing treatment. Finally the sensitivity de-
pends on the type and energy of the ionising ra-
diation. In general low LET radiation (photons,
beta particles) show a higher TL signal than high
LET radiation (alpha particles, neutrons). The
gamma radiation of the decay of 60
Co (Eaverage Ė
1:25 MeV) is widely used as reference radiation.
3.2. Dose response curve
The TL signal is a function ā¦F ā of the absorbed
dose D. Ideally F ā¦Dā shows a linear dose response
over a wide dose range at least over the range of
interest of the application. However, most mate-
rials used in practical dosimetry show a variety of
non-linear eā¬ects. A pattern that is found fre-
quently as the dose is increased is Ā®rst a linear
response, then a supralinear and Ā®nally during the
approach to saturation a sublinear response. The
normalised dose response function (also called the
supralinearity index) f ā¦Dā is deĀ®ned as [6]
f ā¦Dā Ė
F ā¦Dā =D
F ā¦D1ā =D1
; ā¦22ā
where F ā¦Dā is the dose response at a dose D and
D1 is a low dose at which the dose response is
linear. The ideal TLD material has f ā¦Dā Ė 1 in a
wide dose range. If f ā¦Dā 1 the response is called
supralinear, if f ā¦Dā 1 the response is called
sublinear. These features are illustrated in Fig. 10
adopted for McKeever et al. [6] and are discussed
in depth in this issue by Horowitz [24] and Horo-
witz et al. [25], for gamma rays and heavy charged
particles, respectively.
Table 2
Dosimetric requirements in some major application areas
Application area Dose range (Gy) Uncertainty, 1 S.D. (%) Tissue equivalencya
Personnel 10Ć5
Ā±5 Ć 10Ć1
)30, +50 +
Environmental 10Ć6
Ā±10Ć2
Ć30 )
Clinical
Radiotherapy 10Ć1
Ā±102
Ć3:5 ++
Diagnostic radiology 10Ć6
Ā±10 Ć3:5 +
Radiation processingb
101
Ā±106
Ć15 )
a
The more + the more required.
b
Involves sterilization, food processing, material testing, etc.
A.J.J. Bos / Nucl. Instr. and Meth. in Phys. Res. B 184 (2001) 3Ā±28 15
14. 3.3. Energy response
In many TLD applications the main purpose is
to determine the dose absorbed in human tissue
(see Table 2). For this reason it is desirable that the
TLD has an energy response equal or at least
proportional to that of human tissue. Suppose a
TL material is exposed to photons with a certain
ĀÆuence U and energy E. Under charge particle
equilibrium the absorbed dose in that TL material
DTL can then be written as
DTL Ė UEā¦len=qā TL; ā¦23ā
with ā¦len=qā TL the mass energy absorption coeĀ-
cient of the TL material. If the TL material is re-
placed by tissue, i.e. exposed to photons with the
same U and E then for the absorbed dose in tissue
a similar equation can be written. From both
equations it follows
DTL
Dtissue
Ė
ā¦len=qā TL
ā¦len=qā tissue
; ā¦24ā
with ā¦len=qā tissue the mass energy absorption coef-
Ā®cient of tissue. In Fig. 11 this ratio has been
plotted for several TL materials as function of the
energy. It is seen that above 200 keV the ratio is
constant for all materials shown. In the 10Ā±200
keV region, however, there are deviations indi-
cating how much the measured dose deviates from
the dose to be determined. In the mentioned en-
ergy region the photoelectric eā¬ect is dominant.
The photoelectric component of the mass energy
absorption coeĀcient of a certain element varies
approximately as Z3
Ć Z4
. In order to characterise
the energy response with one number instead of
with a whole curve the concept of eā¬ective atomic
number Zeff has been introduced deĀ®ned as
Zeff Ė
ĀĀĀĀĀĀĀĀĀĀĀĀĀĀĀĀĀĖ
i
aiZm
i
m
r
; ā¦25ā
with ai the fractional electron content of element i
with atomic number Zi. The value of m will be
typically range from 3 to 4, with 3.5 a reasonable
value [26]. Eā¬ective atomic numbers for a number
of common materials calculated with Eq. (25) with
m Ė 3:5 are given in Table 3. Tissue as deĀ®ned in
the ICRU sphere has Zeff Ė 7:35. TL materials
with a Zeff value the same or near this number are
called tissue equivalent. Correspondingly fat and
bone equivalent materials can be indicated. It can
be seen from Table 3 and Fig. 1 that BeO and
Li2B4O7 are almost tissue equivalent. It should be
noted that in the literature various deĀ®nitions of
Zeff are used. The composition of tissue is also not
always the same. So the numbers for Zeff found in
Table 3 can diā¬er a little from those in the litera-
Fig. 11. Photon energy response for a few indicated TLD
materials. The reference material is soft tissue with Zeff Ė 7:35.
Fig. 10. Examples of dose response curves of three TLD
materials. (A) The TL response for the 100Ā°C peak in SiO2
(supralinearity over the entire dose range shown). (B) The
well-known linearĀ±supralinearĀ±sublinear behaviour of peak 5
of LiF:Mg, Ti(TLD-100). (C) The dose response of
CaF2:Mn(TLD-400) with a weak supralinearity. A linear
response is shown by the dashed line (from [6]).
16 A.J.J. Bos / Nucl. Instr. and Meth. in Phys. Res. B 184 (2001) 3Ā±28
15. ture. In practical situations the photon energy re-
sponse is given by the relative energy response
deĀ®ned as the ratio of the photon response at a
certain low energy (often 30 keV) and the response
to 1.25 MeV photons from a 60
Co source.
3.4. Annealing behaviour
One of the attractive points of TLD is the
possibility to reuse the TL material many times.
To be sure that in the reuse the TL material has
precisely the same properties as before a thermal
annealing procedure is required. This pre-irradia-
tion annealing consists of holding the sample at an
elevated temperature for some time followed by a
cooling to room temperature at a certain cooling
rate a. Sometimes this is followed by a low tem-
perature anneal (see Fig. 12). The annealing pro-
cedure serves several purposes. At Ā®rst it empties
all the traps as far as this has not happened during
the read-out, i.e. it resets the TL signal to zero.
Secondly, it re-establish the thermodynamic defect
equilibrium which existed in the material before
irradiation and read-out. During irradiation and
the read-out this equilibrium is disturbed and the
reaction is pushed back in the opposite direction
by thermal annealing. Finally, it resets the occu-
pancy of the deep, thermally disconnected, traps.
They are not always present but, if so, they inĀÆu-
ence the TL sensitivity of a given peak since they
act as competitors [24]. In this context the impor-
tance of the cooling rate should be mentioned [6].
Defects, which act as trapping centres, may cluster
and even precipitate at low temperatures while
they dissociate and dissolve into the material at
high temperature. Rapid cooling can freeze in the
Fig. 12. Overview of the various stages of annealing, storage
and read-out of a typical TLD material, where a is the cooling
rate following pre-irradiation annealing, and b is the heating
rate during TL read-out (from [21]).
Table 3
Eā¬ective atomic number of common materials calculated with Eq. (25) with m Ė 3:5 arranged according increasing Zeff
Material Composition element: mass fraction in % Zeff
Polyethylene ā¦C2H4ā n H: 14.37, C: 85.63 5.53
Fat H: 11.95, C: 63,72, N: 08; O: 23.23, Na: 0.05, P: 0.02,
S: 0.07, Cl: 0.12, K: 0.03, Ca: 0.01
6.38
PMMA ā¦C5H8O2ā n H: 8.05, C: 59.99, O: 31.96 6.56
BeO Be: 36.0, O: 64.0 7.21
Li2B4O7 Li: 8.21, B: 25.57, O: 66.22 7.32
Tissue
ICRU-sphere H: 10.1, C: 11.1, N: 2.6, O: 76.2 7.35
ICRU-striated H: 10.20, C: 12.3, N: 3.5, O: 72.90; Na: 0.08, Mg: 0.02, P: 0.02, S: 0.5, K: 0.3 7.63
ICRP-skeletal H: 10.06, C: 10.78, N: 2.77, O: 75.48, Na: 0.075, Mg: 0.019,
P: 0.18, S: 0.24, Cl: 0.079, K: 0.3, Ca: 0.003, Fe: 0.004, Zn: 0.005
7.65
Water H2O H: 11.19, O: 88.81 7.51
Air C: 0.0124, N: 75.53, O: 23.18, Ar: 1.28 7.77
LiF Li: 26.75, F: 73.25 8.31
Al2O3 Al: 47.08, O: 52.93 11.28
SiO2 O: 53.3, Si: 46.7 11.75
Compact bone H: 4.72, C: 14.43, N: 4.2, O: 44.61, Mg: 0.22, P: 10.5, S: 0.32, Ca: 20.99, Zn: 0.01 13.59
CaSO4 O: 47.0, S: 23.6, Ca: 29.4 15.62
CaF2 Ca: 51.33, F: 48.67 16.90
A.J.J. Bos / Nucl. Instr. and Meth. in Phys. Res. B 184 (2001) 3Ā±28 17
16. high temperature equilibrium of defects whereas a
slow cooling rate can result in a variable degree of
defect aggregation. In LiF(TLD-100) it has been
shown that the cooling rate eā¬ects the TL sensi-
tivity and the measured trapping parameters [21].
Sometimes a post-irradiation anneal is applied.
The purpose is to remove low-temperature glow
peaks which often show a strong fading. The
various stages in the process of preparation and
use of a TLD material are illustrated schematically
in Fig. 12. One should realise that all steps inĀÆu-
ence the TL signal in some way. Therefore, strict
and, above all, reproducible procedures should be
applied in order to get a high precision in the dose
determination.
3.5. Detection limits
In most cases one uses the phrase `detection
limit' to mean the lowest level of detection. How-
ever, it also can be used to indicate the highest
detectable level. For TL materials the highest TL
signal is determined by the dose for which all the
trapping centres present in the sample are Ā®lled. It
is diĀcult to indicate a single dose value as the
highest detectable since the dose response curve
shows a sublinear behaviour during its approach
to saturation (see Fig. 10). The end-point of the
linear range may be considered as the highest limit
of detection although higher doses in the sublinear
range also can be determined by careful calibra-
tion.
The lower limit of detection is important in low
dose measurements where the signal of an irradi-
ated TLD is almost the same as the signal of the
background. It is deĀ®ned as the smallest absorbed
dose that according to an analytical process can be
detected at a speciĀ®ed conĀ®dence level. It is fre-
quently expressed as two standard deviations from
the signal from an unexposed dosemeter. Note,
that this deĀ®nition covers not only the TL material
and read-out system but also the analytical process
(anneal procedure, algorithm and analysis rou-
tine). If one likes to compare the lower limit of
detection of diā¬erent TL materials the TL reader
and analytical process should be kept the same as
much as possible. The simple deĀ®nition of two
standard deviations from the signal from an un-
exposed dosemeter can be used for comparison of
the lower limit of detection of diā¬erent TL mate-
rials, e.g. as relative measure. If one likes to de-
termine the absolute value of the lowest detectable
dose in a background at certain dose level the
simple deĀ®nition is not an appropriate speciĀ®ca-
tion since it does not take into account the statis-
tical uncertainty of the background dose. Hirning
[27] has applied the statistical theory of detection
and determination to TLD and derived an ex-
pression for the lower limit of detection LD in case
of n background dosimeters and m irradiated
dosimeters,
LD Ė
2ā¦tnsb ā” t2
ms2
lKbā
1 Ć t2
ms2
l
; ā¦26ā
where tn and tm are Student t factors for sample
sizes of n and m dosimeters at the conĀ®dence level
required; sb is the sample standard deviation of n
background dosimeters; sl is the relative standard
deviation of the m irradiated dosimeters and Kb is
the average measured background air kerma free
in air. In many cases (but not all) the second term
in both the numerator and the denominator are
small compared to the Ā®rst term so that LD ap-
proaches to 2tnsb which is, for a high number of
dosemeters, close to the earlier mentioned two
standard deviations.
3.6. Fading
Fading is the unintentional loss of the TL-sig-
nal. It leads to an underestimation of the absorbed
dose. Fading may be due to several causes. Ther-
mal fading originates from the fact that even at
room temperature there is a certain probability
that charge carriers escape from their trapping
centres. The half-life of a glow peak in case of Ā®rst-
order kinetics has already been given by Eq. (21).
For a useful TL signal t1=2 must be several times
longer than the time between the beginning of the
exposure and the read-out. The Commission of the
European Communities requires a fading of less
than 5% over the monitoring period at 25Ā°C.
Fading may be also caused by optical stimulation,
for example, by sunlight. Some high sensitively TL
materials such as CaF2 and CaSO4 are extremely
18 A.J.J. Bos / Nucl. Instr. and Meth. in Phys. Res. B 184 (2001) 3Ā±28
17. light sensitive and the fading is enhanced consid-
erably. In general high sensitivity materials should
be handled, used and stored in opaque containers
to prevent fading from light exposure. Other types
of fading which are not temperature dependent are
caused by quantum mechanical tunneling of the
trapped charge to the recombination site and lo-
calised transitions which do not take place via the
delocalised bands.
3.7. Physical forms [28]
The physical form of TL materials are generally
loose powders or solid pieces. The solid form may
be composed entirely of single crystals or poly-
crystalline extrusions or as a homogeneous com-
posite of the phosphor and some binding material
such as PFTE. The powder dosimeter has some
special advantages, e.g. good ĀÆexibility of the
dosimeter size and shape. The grain size aā¬ects the
sensitivity and inĀÆuences its handling in dispens-
ers. A good compromise is to use powders with
grain sizes between 75 and 200 lm. Hot-pressed
forms of TL detectors are produced by compres-
sion of polycrystalline material at elevated tem-
perature. Another solid form is obtained by
extrusion of the fused polycrystalline material
through a shaped die, and by cutting and polishing
the resultant extrusion. A variety of geometries
and sizes are available: discs, square chips, rods
and Ā®lms. Sintered dosimeters are prepared by
high temperature sintering of pressed phosphor
powder. LiF, BeO, Li2B4O7, Al2O3 and CaSO4
dosimeters may be produced in this way but
Table 4
A checklist for reporting of thermoluminescence dosimetry measurements adopted [29]
Area of
attention
Relevant informationa
Detectors Ć
Material (including a potential matrix such as TeĀÆon)
Doping (if available with concentrations)
Manufacturer
Ć
Physical form and dimensions
Selection for measurements if any (e.g., discard discoloured chips or only select detectors with reproducibility
better than Ć3%)
Ć
Casing or wrapping of the detectors (during measurement and calibration)
Annealing Ć
Oven (or reader) type and manufacturer
Annealing tray material
Temperature proĀ®le, maximum temperature
Heating/cooling rate in Ā°C/min Ā± reproducibility thereof
Read-out Ć
Reader type, model and manufacturer
Spectral sensitivity of the light collection system
Manual or automated read-out
Read-out temperature proĀ®le (ramp or step)
Use of inert gas (e.g., nitrogen)
Ć
What data were collected (e.g., photomultiplier charge between TLD temperature of 160Ā°C and 240Ā°C)
Evaluation What data were evaluated (if diā¬erent from data collected)
What (if any) conversion coeĀcients used to convert the measured physical quantity to the reported quantity
Ć
Regions of interest Ā± range of integration (if used)
Glow peak analysis (if any, including the mathematical model used)
Ć
Mode of calibration (individually calibrated detectors?)
Ć
Number and type of reference/standard exposure
Ć
Radiation quality in which the calibration was performed
Ć
Any corrections used in the calibration (e.g., energy dependence, supralinearity, angular dependence)
General Ć
Typical reproducibility achieved at what dose level (1 S.D.)
Comments on handling and storage, if any
Other information as relevant to the material in use
a
Information indicated by an asterisk is considered essential.
A.J.J. Bos / Nucl. Instr. and Meth. in Phys. Res. B 184 (2001) 3Ā±28 19
18. physical fragility limits their use, particularly in
large and thin form.
3.8. Factors to be controlled
From the foregoing it can be concluded that the
measured TL signal ultimately depends on many
factors. A small uncertainty in the value of the
evaluated absorbed dose will only be achieved if all
these factors are controlled. This implies Ā®xed and,
above all, reproducible procedures. It means also
that comparisons of TLD measurements of the
same TL material from diā¬erent groups is not
straightforward since diā¬erent procedures were
probably used when handling and evaluating the
TLD materials. Some TLD experts [29] have for-
mulated a checklist for reporting of TLD mea-
surements (see Table 4). It can be used as a
reminder of which factors should be controlled.
Although the list has been devised for clinical TLD
measurements it can be used fruitfully for other
dosimetric areas as well.
4. Two high sensitivity TL materials
Since the introduction of LiF:Mg, Ti (TLD-
100) many new TL materials with much higher
sensitivity have been introduced. Several of them
show a high fading rate, a poor energy response or
suā¬er from other unwanted properties. However,
two TL materials have stood out as promising and
attracted a lot of attention. They will discussed
here in some detail.
4.1. Lithium ĀÆuoride LiF:Mg, Cu, P
4.1.1. Introduction
Nakajima et al. [30] were the Ā®rst to describe
the properties of LiF crystals doped with Mg, Cu
and P impurities. This TL material combines two
attractive properties, namely a high sensitivity and
a good tissue equivalency. As a result, LiF:Mg,
Cu, P has generated extensive interest in the radi-
ation dosimetry community as an ultrasensitive
TLD material. The sensitivity of this material is
more than 20 times higher as compared to
LiF:Mg, Ti (TLD-100). However, it was reported
to lose its sensitivity after only one use [31,32] and
to show a high residual signal [33]. (The residual
signal is deĀ®ned as the intensity of the signal
during second read-out and is usually expressed
into a percentage of the signal from the Ā®rst read-
out.) Much research has been carried out to im-
prove the properties and this has led to a material
with a high, stable sensitivity, an low residual
signal, an almost ĀÆat energy response, a low fading
rate and a linear dose response. LiF:Mg, P, Cu is
sold commercially by the Solid Dosimetric and
Detector Laboratory in Beijing, China as GR-200,
by Nemoto in Japan as NTL-500, by the Institute
of Nuclear Physics in Poland as MCP-N, and by
Saint-Gobain Crystals and Detectors (formerly
Bicron-NE and previous to that Harshaw) in the
USA as TLD-100H, TLD600H and TLD700H.
One has to realise that the material has passed
through a gradual evolution (and is still under
development) so that the material sold today is not
the same as that sold 10 years ago.
4.1.2. TL characteristics
A glow curve typical for LiF:Mg, P, Cu is
shown in Fig. 13 adopted from McKeever et al.
[34]. Peak 4 is the main dosimetric peak. When
recorded at a heating rate of 2.1Ā°C/s the peak
maximum of the main peak is found at 225Ā°C. It
Fig. 13. Glow curve from LiF:Mg, Cu, P (Mg: 0.02 M%; Cuā”
:
0.004 M%; P: 3.0 M%) separated into its components peaks.
Thermal annealing procedure: 10 min at 240Ā°C followed by a
cooling in air. The heating rate was 2:1Ā°C=s and the absorbed
dose 0.4 Gy from a 60
Co source (from [34]).
20 A.J.J. Bos / Nucl. Instr. and Meth. in Phys. Res. B 184 (2001) 3Ā±28
19. has been believed for a long time that the sensi-
tivity will be lost, or at least reduced, if the sample
is heated beyond the 240Ā°C. Peak 5 and the other
high temperature peaks, have, therefore, attracted
a lot of attention since these peaks are not read-
out fully and the remaining charge carriers in those
traps may cause an accumulation of a residual
signal which eā¬ects the detection limit. The resid-
ual signal varies from batch-to-batch and depends
on the annealing procedure, notably the cooling
rate [35], but a value of 5Ā±10% seems typical [33].
Results of Oster et al. [36] indicate that a short
duration above 240Ā°C as during TL read-out can
signiĀ®cantly reduce the residual signal without
aā¬ecting the sensitivity. Recent research [37] has
shown that GR-200 can be used at heating rates as
high as 15Ā°C/s to a maximum of 270Ā° for 12 s with
a residual signal of 2% under these conditions.
Meanwhile the Beijing group [38] have developed
an improved preparation procedure by which the
residual signal was reduced to 0.4% while the
sensitivity was 41 times as high as of TLD-100.
This has been achieved by sintering at a tempera-
ture of 720Ā°C and using an increasing Cu con-
centration.
Attempts to Ā®t the glow curve of this material
using Ā®rst-order expressions have proved more
diĀcult than for LiF:Mg, Ti because of the ex-
treme overlap of the various peaks, particularly in
the region of the main peak (see Fig. 13). For the
peaks at 504 and 498 K a large activation energy
(E 2 eV) is found. As in TLD-100, the trapping
parameters as found by glow curve analysis are
inĀÆuenced by the annealing procedure [39]. Fur-
thermore, Alves et al. [40] have shown that the
dominant processes inĀÆuencing the stability of the
main glow peaks of GR-200 are processes aā¬ecting
the trapping centres and not the trapped charges.
All these observations indicate that, like in TLD-
100, defect interactions occur which inĀÆuence the
TL mechanism to a large extent.
The dose response of LiF:Mg, Cu, P is linearĀ±
sublinear rather than linearĀ±supralinearĀ±sublinear.
The linearity range extends from 1 lGy to 10 Gy.
The lack of supralinearity is a particular advantage
in radiotherapy applications where the dose levels
are typically of the order of several grays. A lower
detection limit for a TLD system based on
LiF:Mg, Cu, P(GR-200A) of 3.5 lSv in a back-
ground of 78 nSv/h, calculated using Eq. (26), has
been determined by Vismara and Furetta [41].
The high sensitivity of this material has already
been mentioned several times. Relative sensitivities
(compared to TLD-100) found in the literature
vary from 10 to 65. Factors inĀÆuencing the re-
ported sensitivity are: the manufacturer, the an-
nealing procedure and the spectral sensitivity of
the light detection system including the optical
Ā®lters. It should be realised that the mentioned
factors refer to both LiF: Mg, Cu, P and LiF:Mg,
Ti. A sensitivity of 25 times higher as compared to
TLD-100 can be considered as a weighted average.
The photon energy response of LiF: Mg, Cu, P
diā¬ers from LiF:Mg, Ti (see Fig. 14). Instead of an
overresponse it shows in the energy range from 60
to 250 keV an underresponse. This response is
explained as a microdosimetric ionisation density
eā¬ect [43]. This eā¬ect occurs at microdosimetric
volumes along the tracks of the secondary elec-
trons and causes a local saturation which reveals
itself in a lower eĀciency.
4.1.3. TL mechanism
The study of the TL mechanism is complicated
by the fact that all three dopants play a role. A
complete description of the TL mechanism should
be able to explain the high eĀciency, the role of
Fig. 14. Air kerma relative photon energy response of LiF:Mg,
Ti (TLD-100) and LiF:Mg, Cu, P (GR-200A) compared with
the expected response Sā¦Eā air. Responses are normalised to 662
keV photons from a 137
Cs source (from [42]).
A.J.J. Bos / Nucl. Instr. and Meth. in Phys. Res. B 184 (2001) 3Ā±28 21
20. each dopant, the observation that the sensitivity is
strongly reduced if the sample is heated above
270Ā°C and the eā¬ects of the thermal treatments
and read-out procedures on the glow curve. It is
questionable whether such a complete description
will ever be found. From the studies carried out so
far only some trends become visible. Bilski et al.
[44] conducted a systematic study to determine the
inĀÆuence of the diā¬erent impurities on the TL
properties of LiF(MCP). They conclude:
1. The height of peak 4 depends on the concentra-
tion of Cu and Mg, showing a clear maximum.
2. The intensity of the high temperature peaks in-
creases with Mg and decreases with increasing
Cu concentration. This role of Cu is similar as
found by Tang et al. [38].
3. The dependence of the height of peak 4 on the
concentration of P behaves like a step function,
increasing above a certain threshold value.
McKeever [45] concluded from an examination of
the emission spectra that the recombination site is
related to P, which was found to be essential for
the high intensity emission, whereas the trapping
sites all appear to be related to Mg. The role of Cu
has been investigated by Chen and Stoebe [46] who
used EXAFS studies to show that a change in the
valence of copper ion from Cuā”
to Cu2ā”
occurs
upon annealing above 240Ā°C, a change that also
occurs at radiation damage levels above 104
Gy.
This result indicated that the presence of Cuā”
ions
may be essential for the high TL sensitivity of
LiF:Mg, Cu, P, and that a change in the copper
valence to Cu2ā”
, a process shown to be only par-
tially reversible, reduces the TL sensitivity. Meij-
vogel and Bos [47] studied in detail the emission
from GR-200 and MCP-N and found that the
main emission band consists of two, overlapping
Gaussian shaped bands, one at 384 nm dominating
peak 4 and one at 350 nm dominating peak 5.
Information in the literature seem to indicate that
Cuā”
-related defects may be involved as lumines-
cent centres, particularly for the TL emission at
380 nm while the 350 nm emission originates from
Cu2ā”
.
New insight into the role of the phosphorous
impurity has been found from an electron energy
loss spectroscopy study [48]. This study indicates
the presence of previously undetected inclusion
particles. As-received pellets contain small con-
centrations of inclusions analysed to contain Li, F
and O. After the heat treatment at 320Ā°C (causing
a 95% loss in TL sensitivity) the inclusions are
found to contain principally Li, P and O. The in-
clusions may be the results of phosphorous pre-
cipitation. The loss in TL sensitivity seems to be a
result of a complex set of defect reactions involv-
ing the LiF matrix and the Mg, Cu and P dopants,
along with the oxygen impurities. Very recently,
Tang et al. [49] showed that the change in glow
curve structure and the loss of TL sensitivity, as a
consequence of annealing between 260Ā°C and
400Ā°C, can be recovered fully by annealing at
720Ā°C for 30 min in a nitrogen atmosphere fol-
lowed by the standard anneal of 10 min at 240Ā°C.
This seems to indicate that the copper valence
change is fully reversible and also that the clus-
tering and precipitation process can be fully re-
versed. The full recovery of the loss of the TL
sensitivity is a very interesting observation for the
practical dosimetry (reuse of dosemeters which
were unintentionally heated too high) and is a
point of departure for further research into the TL
mechanism of this material.
4.1.4. Applications
LiF:Mg, Cu, P has found application in the
Ā®elds of personnel, environmental and clinical
dosimetry. An example of each Ā®eld will be given.
Because of its sensitivity and near ĀÆat energy
response LiF:Mg, Cu, P is very well suited for dose
levels typical to personnel dosimetry. Many radi-
ation facilities are now in the process of switching
over part of their external dosimetry programme
to this material (e.g. the US Navy, Los Alamos
National Laboratory, Ontario Hydro and AECL,
Canada). Bicron (Harshaw) has recently intro-
duced a personnel dosimetry system based on
LiF:Mg, Cu, P [50]. Tests, designed to determine
the usefulness for practical personnel dosimetry
were carried out and results compared with the
international standard IEC 1066. Requirements
with respect to batch homogeneity, repeatability,
linearity, detection threshold, self-irradiation were
amply met. Those for the residual signal and light
induced fading after some adaptations. These re-
sults clearly demonstrate the potential of LiF:Mg,
22 A.J.J. Bos / Nucl. Instr. and Meth. in Phys. Res. B 184 (2001) 3Ā±28
21. Cu, P to replace LiF:Mg, Ti in large scale per-
sonnel dosimetry programmes.
The practical aspects on the implementation of
LiF:Mg, Cu, P in routine environmental moni-
toring programmes has been described by Saez
Vegara [51]. Under Ā®eld conditions a lower limit of
detection of 0.7 lGy was derived. This corre-
sponds to 5Ā±8 h exposure at usual environmental
dose rates and means that environmental mea-
surements can be carried out on a daily basis
allowing the rapid assessment of slight increases
in background due to man-made sources or
incidents.
Saez Vegara [51] also illustrated another feature
of LiF:Mg, Cu, P, namely its capability to distin-
guish two dose levels which are very close. This so-
called dose resolution for short exposure periods
has been studied by measuring the weekly envi-
ronmental dose ā¦% 105 nGy=hā with LiF:Mg, Cu,
P detectors for 10 months in two outdoor locations
at CIEMAT: ESMARALDA ā¦% 130 nGy=hā and
a building rooftop ā¦% 105 nGy=hā . The results
obtained for 30 consecutive weeks are shown in
Fig. 15, which clearly shows the capability of the
phosphor to distinguish environmental doses
which are similar. Fig. 15 also shows a reasonable
agreement with the data from the high pressure
ionisation chambers.
LiF:Mg, Cu, P is also implemented in medical
practice. Ginjaume et al. [52] have shown that the
energy response for several high energy photon (6
and 18 MV) and electron beams (6Ā±18 MeV) used
in radiotherapy is within Ć2:5% with respect to
60
Co. Together with other characteristics such as
lack of supralinearity makes LiF:Mg, Cu, P useful
for radiation therapy dosimetry. The high sensi-
tivity makes LiF:Mg, Cu, P also applicable for low
dose applications. Duggan et al. [53] have dem-
onstrated the applicability of LiF;Mg,Cu,P for
measuring extremely low doses in medical imaging
such as neonate X-ray examinations. A skin en-
trance dose of only 35 Ć 10 lGy could be assessed
with Ć20% (1 S.D.) reproducibility, the same order
of magnitude as the reproducibility of the expo-
sure.
In conclusion, in all Ā®elds LiF:Mg, Cu, P is
being accepted as a reliable dosimetric material.
Drawbacks reported in previous works, such as
poor reproducibility and a high residual signal,
have now been overcome.
4.2. Aluminium oxide Al2O3:C
4.2.1. Introduction
Single crystal a-Al2O3 is commonly known as
sapphire or corundum, with the last name being
used generically for the type of structure. The
material is used as substrates for microelectronic
devices, as laser host material (e.g. doped with
Cr3ā”
i.e. ruby) and as a high strength window
material. It has already been known as TL mate-
rial for a long time [54]. Interest in the material has
increased considerably since the improvement of
the dosimetric properties by intentional inclusions
of oxygen vacancies into its structure. Such a
modiĀ®cation of the oxide is possible by annealing
and melting under strongly reducing conditions in
the presence of graphite [55Ā±57]. The resulting,
ultra-high sensitivity TL material is referred to as
a-Al2O3:C. The material was Ā®rst developed and
produced at the Urals Polytechnical Institute in
the form of single crystals. Nowadays the material
is produced in diā¬erent forms: single crystals,
powders and thin layers on substrates and in a
Fig. 15. Weekly environmental air kerma rates at two locations
(j: ESMERALDA and : building rooftop), as measured
with LiF:Mg, Cu, P detectors. Error bars indicate 95% conĀ®-
dence level. Solid lines indicate the results obtained using the
corresponding high pressure ionisation chamber (HIPC) at each
location (from [51]).
A.J.J. Bos / Nucl. Instr. and Meth. in Phys. Res. B 184 (2001) 3Ā±28 23
22. number of diā¬erent laboratories. The material has
attracted even more interest since it became clear
that the information stored during exposure to
ionising radiation can also be read-out by optical
stimulation. In this paper we limit ourselves to a
review of the TL properties. The optically stimu-
lated luminescence properties are reviewed else-
where in this issue [58].
4.2.2. TL characteristics
A glow curve typical of a-Al2O3:C is shown in
Fig. 16. In the reproduced temperature range it
shows a single glow peak with a maximum at 453
K for a heating rate of 1 K=s. The glow curve
shape depends upon many factors among which
impurities, growth conditions, absorbed dose and
heating rate. At lower temperatures (in the range
from )20 to 50Ā°C two other glow peaks are pre-
sent [59] while at 570 K a deep trap has been ob-
served [60]. These traps may act as competitors to
the main dosimetric trap. The competition aā¬or-
ded by the shallow (low temperature) traps give
the TL signal a slight irradiation temperature de-
pendence.
The TL dose response is linearĀ±supralinearĀ±
sublinear (see Fig. 17). The sensitivity is 40Ā±60
times more as compared to LiF:Mg, Ti (at a
heating rate of 4Ā°C=s). It shows a very low zero
dose (read-out value of an unirradiated dosemeter)
which makes it possible to detect very low dose
levels down to 300 nGy [63]. The eā¬ective atomic
number of pure Al2O3 is 11.3. The carbon content
(100Ā±5000 ppm) lowers Zeff but it still results in an
photon sensitivity which is at 30 keV 2.9 times
higher than at 1.25 MeV. The relative photon en-
ergy response between 50 and 150 keV is reduced
to 0.9 by using a Ā®lter of 0.2 mm thick lead and is
nearly ĀÆat [3]. Akselrod et al. [57] report low fad-
ing during storage in the dark (less than 5% per
year). However, Musk [62] observed stronger
fading after one month (11%) and after three
months 21%.
A curious observation in a-Al2O3:C is a strong
heating rate dependence of the TL intensity
[60,63]. As the heating rate increases not only does
the peak maximum shift to higher temperatures as
expected from the kinetics but the intensity falls oā¬
strongly on the high temperature side of the peak.
With an increase in the heating rate from 1 to
10Ā°C/s the total light output decreases four times
while from the kinetics of TL production the total
light sum is expected to be independent of the
heating rate (see Figs. 2(d) and 3(d)). Akselrod
et al. [64] have demonstrated that this decrease of
Fig. 16. Glow curve from a-Al2O3:C (Victoreen) recorded at 1
K/s after a reader anneal followed by irradiation with a 90
Sr/90
Y
beta source (30 mGy). A Hoya U-430 Ā®lter was used for the TL
output.
Fig. 17. The TL response as a function of gamma dose for
a-Al2O3:C showing a linearĀ±supralinearĀ±sublinear behaviour
(from [56]).
24 A.J.J. Bos / Nucl. Instr. and Meth. in Phys. Res. B 184 (2001) 3Ā±28
23. TL intensity with increasing heating rate is the
result of thermal quenching of the luminescence of
the emission centres. Further they give a plausible
explanation that the heating rate dependence itself
alters with the dose (i.e. with the degree of trap
Ā®lling) as a natural result of the kinetics of the TL
process.
In studying the glow curve of a-Al2O3:C for
increasing doses Agersnap Larsen et al. [65]
showed that the main TL glow peak shifts to
lower temperatures as the dose increases, but not
in a monotonic way, as would be expected, for
example, for non-Ā®rst-order kinetics (see Fig.
3(a)). This behaviour is consistent with an over-
lap of several Ā®rst-order TL peaks. The shift of
the peak maximum can be explained by the dif-
ferential growth of the individual Ā®rst-order peak
components. In measuring the photoconductivity
Whitley and McKeever [66] were able to dem-
onstrate the existence of a wide array of trapping
states. The dose dependence of TL could only be
explained on basis of a distribution of thermal
trap depths. From annealing experiments they
showed that the TL in the 100Ā±250Ā°C range
originates from a trap depths energy distribution
in the 1.7Ā±2.5 eV range.
A disadvantage of a-Al2O3:C as a dosimetric
material is the occurrence of light-induced eā¬ects.
This manifests itself in two diā¬erent ways: light-
induced TL, i.e. an increase in TL intensity and
light-induced fading, i.e. an decrease of the TL
intensity. These eā¬ects have been studied by
Walker et al. [67]. They explained the increase of
the TL intensity by optical transfer of charge from
deep traps into the dosimetric traps and the de-
crease of the TL intensity (fading) by optical
emptying of the dosimetric traps. Both eā¬ects have
their own wavelength dependence. Under red light
(580Ā±600 nm) stimulation no signiĀ®cant fading of
the TL signal have been observed. Moreover, at
long wavelengths (k J 620 nm) excitation out of
the deep traps does not occur. However, at short
wavelengths both eā¬ects can be considerable (at
450 nm the combined eā¬ect resulted in a particular
case in a 80% loss of the TL signal) and should be
taken into account in the application as dosemeter.
It should be noted, however, that the sensitivity of
this material can be turned to an advantage using
the material as a very sensitive optically stimulated
luminescence dosemeter.
4.2.3. TL mechanism
The synthesis of a-Al2O3:C under strongly re-
ducing conditions produces oxygen vacancy cen-
tres. Occupancy of such a centre by two electrons
gives rise to a neutral F centre, whereas occupancy
by one electron forms a positively charged Fā”
centre. The formation of Fā”
centres requires the
presence of negatively charged compensators
which can be realised in this material by substitu-
tion of Al3ā”
by a C2ā”
impurity. Optical absorption
of unirradiated a-Al2O3:C reveals the presence of
both F and Fā”
centres [57]. An important obser-
vation is that the TL intensity from a-Al2O3:C
appears strongly correlated with the Fā”
centre
concentration [68]. Furthermore, the TL emission
spectra shows two emission bands near 180Ā°C.
One broad band peaking near 420 nm due to re-
laxation of F centres and another near 326 nm due
to relaxation of excited Fā”
centres. The 420 nm
emission is believed to be caused by the recombi-
nation of an electron with an Fā”
centre according
to
Fā”
ā” e 3 FĆ
3 F ā” ht ā¦420 nmā ; ā¦27ā
where FĆ
means an excited F centre which decays
from a 3
P state to the 1
S ground state with the
emission of a photon at 420 nm. Likewise the 326
nm emission is believed to be caused by the re-
combination of a hole with an F centre according
to
F ā” h 3 Fā”Ć
3 Fā”
ā” ht ā¦326 nmā ; ā¦28ā
where the Fā”Ć
means an exited Fā”
centre which
decays via a 1B 3 1A transition. So it seems that
there is reasonably explanation which luminescent
centres are involved. However, both the identity of
the electron trap and the hole trap are presently
unknown. Moreover, an obvious question, raised
by McKeever et al. [68], is why the two emission
bands occur at the same temperature (i.e. the same
TL peak). It requires both electrons and holes to
be released at the same temperature. One possible
mechanism, mentioned by the authors, may be the
energy transfer from the excited F centres to the
A.J.J. Bos / Nucl. Instr. and Meth. in Phys. Res. B 184 (2001) 3Ā±28 25
24. Fā”
centres. This implies that there are no holes
released in the 180Ā°C temperature region. This
explanation is in agreement with results of Whitley
and McKeever [67] that the traps causing the TL
of the main dosimetric peak are indeed electron
traps.
In conclusion it can be said that some of the TL
mechanism in a-Al2O3:C is clear but many ques-
tions remain to be answered.
4.2.4. Applications
Like LiF:Mg, Cu, P material, Al2O3:C has
found application as thermoluminescence dose-
meter (TLD) material in the environmental dosi-
metry. In the 11th international intercomparison
of environmental dosimetry 16 dosemeters based
on LiF:Mg, Cu, P and eleven based on Al2O3:C
joined in the project, compared to eight and three
respectively for the 10th intercomparison [69]. This
shows the increased use of both materials in en-
vironmental dosimetry. Current activities seem to
aim at application of a-Al2O3:C as an optically
stimulated luminescence dosemeter (OLSD).
Nevertheless because of its ultra high sensitivity
the material has rendered services as a TL material
in low dose environmental monitoring. Moscov-
itch et al. [61], for example, have demonstrated
that with background subtraction and by elimi-
nating any possibility of the material being ex-
posed to light, dose measurements down to 300
nGy are possible. Al2O3:C has also been used
successfully applied in short-term measurements of
the natural background radiation [70]. Duggan et
al. [53] report on the application of this material in
medical imaging. They compared the performance
with LiF:Mg, Cu, P in neonatal X-ray dosimetry.
The performance of Al2O3:C surpassed LiF:Mg,
Cu, P with regard to a lower measurement limit
and improved reproducibility in the lGy range.
However, at diagnostic energies the undesirable
high energy photon response would warrant
careful calibration.
5. Concluding remarks
TLD are widely used in personnel, environ-
mental and clinical dosimetry. Despite of its fa-
miliarity and its broad application TLD is often
regarded somewhat as a `black art'. Some users
achieve astonishing results with great accuracy,
while for others all attempts seem to fail. In this
review the background is given from which the
various aspects which inĀÆuence the TL signal can
be understood. It makes the reader not a specialist
but makes him/her at least aware on which factors
one should take care of and gives warning of some
potential pitfalls.
The basic concepts of TL have been presented
using the well-known one trapĀ±one recombination
centre model. This simple model can explain, at
least qualitatively, all the fundamental features of
the luminescence production. The model teaches
that the presence of traps are essential for the TL
production. Thus (ideal) materials where no traps
are present, do not show TL. The absorption of
radiant energy is needed to Ā®ll the traps and
heating is necessary to release the charge carriers
from the traps. The model gives insight into how
radiant energy is stored. The model explains fur-
ther that glow peaks at lower temperatures, cor-
responding to shallower traps, are more sensitive
to fading. It explains also the characteristic asym-
metric glow peak which is often actually measured.
The simple model, however, is phenomenolog-
ical of nature. The model does not say anything
about the nature of the trapping and recombina-
tion centres. In real materials it appears not easy to
identify these centres. Sometimes, as for rare earth
elements, the emission wavelength is very charac-
teristic which makes identiĀ®cation of the lumines-
cent centres unequivocal but in many cases
considerable eā¬ort is required to reveal the struc-
ture and nature of the centres. Most centres appear
not to be highly localised point defects. Moreover,
truly isolated trapping and recombination centres
are an exception rather than a rule. It implies that
in real TL material one is faced with complex de-
fect structures. In heating the material the defects
may be not stable which makes the process even
more complicated. This explains why each step in
the procedure of the determination of the dose
(annealing, irradiation, read-out) should be very
carefully controlled.
Two high sensitivity TL materials, notably
LiF:Mg, Cu, P and a-Al2O3:C, have been dis-
26 A.J.J. Bos / Nucl. Instr. and Meth. in Phys. Res. B 184 (2001) 3Ā±28
25. cussed in more detail. These two materials cur-
rently mark the vanguard of TLD research. Ex-
tensive study, up to now, has revealed a lot of data
which have been summarised. Looking back it
must be concluded that a complete understanding
of the TL mechanism in both materials is lacking.
On the other hand, there is suĀcient knowledge to
understand the behaviour in general terms. From
the Ā®eld applications of these materials it appears
that they can be used to measure very low doses in
a reliable way.
References
[1] F. Daniels, C.A. Boyd, D.F. Saunders, Science 117 (1953)
343.
[2] P.D. Townsend, J.C. Kelly, Colour Centres and Imperfec-
tions in Insulators and Semiconductors, Sussex University
Press, London, 1993, p. 70.
[3] Y.S. Horowitz (Ed.), Thermoluminescenct and Thermolu-
minescent Dosimetry, Vols. 1Ā±3, CRC Press, Boca Raton,
FL, 1984.
[4] S.W.S. McKeever, Thermoluminescence of Solids, Cam-
bridge University Press, Cambridge, 1985.
[5] D.R. Vij (Ed.), Thermoluminescent Materials, PTR Pren-
tice-Hall, New Jersey, 1993.
[6] S.W.S. McKeever, M. Moscovitch, P.D. Townsend, Ther-
moluminescence Dosimetry Materials: Propeties and Uses,
Nuclear Technology Publishing, Kent, 1995.
[7] R. Chen, S.W.S McKeever, Theory of Thermolumines-
cence and Related Phenomena, World ScientiĀ®c, Singa-
pore, 1997.
[8] J.T. Randall, M.H.F. Wilkins, Proc. R. Soc. London Ser.
A 184 (1945) 366.
[9] J.T. Randall, M.H.F. Wilkins, Proc. R. Soc. London Ser.
A 184 (1945) 390.
[10] R. Chen, in: Y.S. Horowitz (Ed.), Thermoluminescent and
Thermoluminescent Dosimetry, Vol. 1, CRC Press, Boca
Raton, FL, 1984, p. 81.
[11] G. Kitis, J.M. Gomez-Ros, J.W.N. Tuyn, J. Phys. D 31
(1998) 2636Ā±2641.
[12] R.V. Pagonis, S.M. Mian, G. Kitis, Radiat. Prot. Dosim.
93 (2001) 11.
[13] G.F.J. Garlick, A.F. Gibson, Proc. Phys. Soc. 60 (1948)
574.
[14] A.J.J. Bos, J.B. Dielhof, Radiat. Prot. Dosim. 36 (1991)
231.
[15] C.E. May, J.A. Partridge, J. Chem. Phys. 40 (1964) 1401.
[16] P.J. Kelly, M.J. Laubitz, P. Braunlich, Phys. Rev. B 4
(1971) 1960.
[17] Chen, A.A. Winer, J. Appl. Phys. 41 (1970) 5227.
[18] T.M. Piters, A.J.J. Bos, J. Phys. D 26 (1993) 2255.
[19] A.J.J. Bos, T.M. Piters, Radiat. Prot. Dosim. 47 (1993) 41.
[20] Y.S. Horowitz, M. Moscovitch, Nucl. Instr. and Meth. A
243 (1986) 207.
[21] A.J.J. Bos, R.N.M. Vijverberg, T.M. Piters, S.W.S.
McKeever, J. Phys. D 25 (1992) 1249.
[22] P. Dorenbos, C.W.E. van Eijk, A.J.J. Bos, C.L. Melcher, J.
Phys.: Condens. Mater. 6 (1994) 4167.
[23] D.W. Cooke, B.L. Bennett, K.J. McClellan, J.M. Roper,
M.T. Whittaker, J. Lumin. 92 (2001) 83.
[24] Y.S. Horowitz, Nucl. Instr. and Meth. B 184 (2001) 68.
[25] Y.S. Horowitz, O. Avila, M. Rodriguez-Villafuerte, Nucl.
Instr. and Meth. B 184 (2001) 85.
[26] R.L. Kathren, in: K.R. Kase, B.E. Bjarngard, F.H. Attix
(Eds.), The Dosimetry of Ionizing Radiation, Vol. II,
Academic Press, Orlando, 1987, p. 303.
[27] C.R. Hirning, Health Phys. 62 (3) (1992) 223.
[28] G. Paic, Ionizing Radiation: Protection and Dosimetry,
CRC Press, Boca Raton, 1988.
[29] T. Kron, L. De Werd, P. Mobit, J. Muniz, A. Pradhan, M.
Toivonen, M. Waligorski, Phys. Med. Biol. 44 (1999) L15.
[30] T. Nakajima, Y. Maruyama, T. Matsuzawa, A. Koyano,
Nucl. Instr. and Meth. 157 (1978) 155.
[31] D.K. Wu, F.Y. Sun, H.C. Dai, Health Phys. 46 (1984)
1063.
[32] L.A. Dewerd, J.R. Cameron, D.K. Wu, I.J. Das, Radiat.
Prot. Dosim. 6 (1984) 350.
[33] A. Horowitz, Y.S. Horowitz, Radiat. Prot. Dosim. 40
(1992) 265.
[34] J. McKeever, F.D. Walker, S.W.S. McKeever, Nucl.
Tracks Radiat. Meas. 21 (1993) 179.
[35] T.M. Piters, A.J.J. Bos, Radiat. Prot. Dosim. 33 (1990) 91.
[36] L. Oster, Y.S. Horowitz, A. Horowitz, Radiat. Prot.
Dosim. 49 (1993) 407.
[37] G. Ben-Amar, B. Ben-Shacher, L. Oster, Y. Horowitz, A.
Horowitz, Radiat. Prot. Dosim. 84 (1999) 235.
[38] K. Tang, G. Cai, W. Shen, L. Min, H. Zhu, Radiat. Prot.
Dosim. 84 (1999) 227.
[39] T.-C. Chen, T.G. Stoebe, Radiat. Meas. 29 (1998) 39.
[40] J.G. Alves, J.L. Muiz, A. Delgado, Radiat. Prot. Dosim.
78 (1998) 107.
[41] L. Vismara, C. Furetta, Radiat. Prot. Dosim. 85 (1999)
267.
[42] J.C. Saez Vergara, EULEP-Eurados protocol for X-ray
dosimetry in radiobiology, EUR 19606 EN, 2000.
[43] P. Olko, P. Bilski, E. Ryba, T. Niewiadomski, Radiat.
Prot. Dosim. 47 (1993) 31.
[44] P. Bilski, M. Budzanowski, P. Olko, Radiat. Prot. Dosim.
69 (1997) 187.
[45] S.W.S. McKeever, J. Appl. Phys. D 24 (1991) 998.
[46] T.-C. Chen, T.G. Stoebe, Radiat. Prot. Dosim. 78 (1998)
101.
[47] K. Meijvogel, A.J.J. Bos, Radiat. Meas. 24 (1995) 239.
[48] T.-C. Chen, M. Qian, T.G. Stoebe, J. Phys.: Condens.
Matter 11 (1999) L341.
[49] K. Tang, H. Zhu, W. Shen, Radiat. Prot. Dosim. 90 (2000)
449.
[50] M. Moscovitch, Radiat. Prot. Dosim. 85 (1999) 49.
[51] J.C. Saez Vergara, Radiat. Prot. Dosim. 85 (1999) 237.
A.J.J. Bos / Nucl. Instr. and Meth. in Phys. Res. B 184 (2001) 3Ā±28 27
26. [52] M. Ginjaume, X. Ortega, M.A. Dutch, N. Jornet, A.
Sanchez-Reyes, Radiat. Prot. Dosim. 85 (1999) 389.
[53] L. Duggan, C. Sathiakumar, H. Warren-Forward, M.
Symonds, P. McConnell, T. Smith, T. Kron, Radiat. Prot.
Dosim. 85 (1999) 425.
[54] G.P. Summers, Radiat. Prot. Dosim. 8 (1984) 69.
[55] M.S. Akselrod, V.S. Kortov, D.J. Kravetsky, V.I. Gotlib,
Radiat. Prot. Dosim. 32 (1990) 15.
[56] M.S. Akselrod, V.S. Kortov, D.J. Kravetsky, V.I. Gotlib,
Radiat. Prot. Dosim. 33 (1990) 119.
[57] M.S. Akselrod, V.S. Kortov, E.A. Gorelova, Radiat. Prot.
Dosim. 47 (1993) 159.
[58] S.W.S. McKeever, Nucl. Instr. and Meth. B 184 (2001) 29.
[59] M.S. Akselrod, A.C. Lucas, J.C. Polf, S.W.S. McKeever,
Radiat. Meas. 29 (1998) 391.
[60] V.S. Kortov, I.I. Milman, V.I. Kirpa, J. Lesz, Radiat. Prot.
Dosim. 55 (1994) 279.
[61] M. Moscovitch, M.N. Kaufman, J.E. Rodgers, A. Niroo-
mand-Rad, Radiat. Prot. Dosim. 47 (1993) 173.
[62] J.H. Musk, Radiat. Prot. Dosim. 47 (1993) 247.
[63] G. Kitis, J.G. Papadopoulos, S. Charalambous, J.W.N.
Tuyn, Dosim. 55 (1994) 183.
[64] M.S. Akselrod, N. Agersnap Larsen, V. Whitley, S.W.S.
McKeever, J. Appl. Phys. 84 (1998) 3364.
[65] N. Agersnap Larsen, L. BĆ¹tter-Jensen, S.W.S. McKeever,
Radiat. Prot. Dosim. 84 (1999) 87.
[66] V.H. Whitley, S.W.S. McKeever, J. Appl. Phys. 87 (2000)
249.
[67] F.D. Walker, L.E. Colyott, N. Agersnap Larsen, S.W.S.
McKeever, Radiat. Meas. 26 (1996) 711.
[68] S.W.S. McKeever, M.S. Akselrod, L.E. Colyott, N.
Agersnap Larsen, J.C. Polf, V. Whitley, Radiat. Prot.
Dosim. 84 (1999) 163.
[69] G. Klemic et al., Radiat. Prot. Dosim. 85 (1999) 201.
[70] M. Budzanowski, P. Bilski, L. BĆ¹tter-Jensen, A. Delgado,
P. Olko, J.C. Saez Vergera, M.P.R. Waligorski, Radiat.
Prot. Dosim. 66 (1996) 157.
28 A.J.J. Bos / Nucl. Instr. and Meth. in Phys. Res. B 184 (2001) 3Ā±28