Radiation Measurements 106 (2017) 3e12

Contents lists available at ScienceDirect

Radiation Measurements
journal homepage: www.elsevier.com/locate/radmeas

An overview of recent developments in luminescence models with a

focus on localized transitions
Vasilis Pagonis a, *, Reuven Chen b, Christopher Kulp c, George Kitis d
a
McDaniel College, Physics Department, Westminster, MD 21157, USA
b
Raymond and Beverly Sackler School of Physics and Astronomy, Tel-Aviv University, Tel-Aviv 69978, Israel
c
Department of Astronomy and Physics, Lycoming College, Williamsport, PA 17701, USA
d
Nuclear Physics Laboratory, Aristotle University of Thessaloniki, 54124 Thessaloniki, Greece

h i g h l i g h t s

Overview of recent developments in localized vs delocalized models.

New model for luminescence in quartz applicable for from microseconds to seconds.

Monte Carlo simulations of ground state tunneling in random distribution of defects.

a r t i c l e i n f o a b s t r a c t

Article history: Advances in modeling during the past 20 years have contributed to better understanding of the lumi-
Received 26 September 2016 nescence properties of dosimetric materials. Three types of models have been used extensively in the
Received in revised form literature: delocalized models based on transitions involving the conduction and valence bands, localized
19 December 2016
models usually involving different energy levels of the same trap, and semilocalized models which are
Accepted 9 January 2017
Available online 10 January 2017
based on a combination of localized and delocalized energy levels. The purpose of this paper is to provide
an overview of recent developments in luminescence models, with an emphasis on the importance of
delocalized transitions. Two recent theoretical developments are discussed, namely analytical equations
Keywords:
Localized luminescence models
based on the Lambert W-function which are applicable for delocalized models, and analytical equations
Time resolved luminescence based on tunneling in a random distribution of defects which are applicable for localized models. A new
Feldspar models model for luminescence in quartz is proposed, which is applicable for time scales ranging from micro-
Quartz models seconds to seconds. Recent Monte Carlo simulations of ground state tunneling in a random distribution
of traps and centers are discussed, which are based on a modified version of a previously published
model. Some of the current challenges associated with luminescence signals measured at elevated
temperatures are pointed out, and suggestions are made for future work in this research area.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction The main goals of modeling studies are: to provide a quantitative

description of these dosimetric signals, to develop methods for the
Advances in modeling during the past 20 years have contributed accurate evaluation of parameters characterizing traps and centers,
to better understanding of the luminescence properties of dosi- to search for general models which explain the behavior of pre-
metric materials. In general terms, luminescence signals from conditioned samples, and to help researchers understand the un-
dosimetric materials are characterized by the presence of several derlying luminescence production mechanisms. Three types of
components, originating from different traps or centers. These models have been used extensively in the literature: delocalized
signals vary according to preconditioning of the samples (irradia- models based on transitions involving the conduction and valence
tion dose, prior optical and thermal stimulation, radiation quality). bands, localized models usually involving different energy levels of
the traps/centers, and semilocalized models which are based on a
combination of localized and delocalized energy levels. For a
comprehensive historical summary of these three types of lumi-
* Corresponding author.
nescence models, the reader is referred for example to the recent
E-mail address: [email protected] (V. Pagonis).

http://dx.doi.org/10.1016/j.radmeas.2017.01.001
1350-4487/© 2017 Elsevier Ltd. All rights reserved.
4 V. Pagonis et al. / Radiation Measurements 106 (2017) 3e12

review paper by Horowitz et al. (2015) and the book by Chen and recombination center and the conduction band are described in
Pagonis (2011). detail in Chen and Pagonis (2011). The OTOR rate equations cannot
In terms of the time scales involved in luminescence processes, be solved in closed form. However, by using the simplifying quasi-
one can distinguish two broad types of experiments. In the first equilibrium (QE) conditions it is possible to reduce the system of
category one studies phenomena like thermoluminescence (TL) or equations in the OTOR model into the following single equation
optically stimulated luminescence (OSL) which take place in time known as the general one trap model (GOT):
scales of seconds. In the second category of time-resolved (TR)
experiments, one uses short light pulses to separate the stimulation dn n2
and emission of luminescence in time. TR experiments usually IðtÞ ¼  ¼ pðtÞ ; (1)
dt ðN  nÞR þ n
involve much shorter time scales than TL/OSL, typically of the order
of milliseconds or microseconds. where N (cm3) is the total concentration of electron traps in the
The purpose of this paper is to provide an overview of recent crystal, nc and n (cm3) correspondingly are the instantaneous
developments in luminescence models, with an emphasis on the concentrations of electrons in the conduction band and in the filled
importance of delocalized transitions. The paper is organized as traps, An and Am (cm3s1) are the re-trapping and recombination
follows. Section 2 summarizes two recent theoretical de- rates, and R ¼ An/Am is the retrapping ratio. The quantity p(t) (s1) is
velopments, namely analytical equations based on the Lambert W- the time rate which depends on the stimulation mode used during
function which are applicable for delocalized models (Kitis and the experiments. For the case of isothermal excitation and optically
Vlachos, 2013), and analytical equations based on tunneling in a stimulated luminescence (OSL), p(t) ¼ constant. For the case of
random distribution of defects, and which are applicable for thermoluminescence (TL) the quantity p(t) ¼ sexp(-E/kBT) where E,s
localized models (Kitis and Pagonis, 2013). Section 3 proposes a are the thermal kinetic parameters characterizing the traps, kB is
new model for luminescence in quartz, which is applicable for time the Boltzmann constant and T is the temperature of the sample
scales ranging from microseconds to seconds. Section 4 is a review which is usually increasing linearly with time. The QE conditions
of recent research on two different descriptions of ground state used in deriving the GOT model in Equation (1) are nc << n and jdnc/
tunneling phenomena in a random distribution of defects; a dtj<<jdn/dtj.
macroscopic differential equations approach, and a microscopic Kitis and Vlachos (2013) obtained analytical solutions for
Monte Carlo approach. Finally, section 5 discusses some of the equation (1) in terms of the two real branches of the well-known
current challenges associated with luminescence signals measured transcendental Lambert function W(z). These two branches of the
at elevated temperatures, and suggestions are made for future work Lambert function are shown in Fig. 2. These authors found that the
in this research area. two types of solutions depend on the ratio R ¼ An/Am of the re-
trapping and recombination rates An, Am. For R < 1 the solution of
2. Developments in new analytical equations for localized the OTOR model is based on the first real branch of the Lambert
and delocalized models function, which is defined in the region (1/e)<z<∞ and is denoted
by W[0,z] in Fig. 2. For R > 1 the solution of the OTOR model is based
In this section we summarize two recent theoretical works on on the second real branch of the Lambert which is defined in the
analytical equations which can be used for describing luminescence region (1/e)<z < 0 and is denoted by W[-1,z] in Fig. 2. In its most
signals. general form the Lambert function is defined as the function
satisfying the equation W(z)eW(z) ¼ z for any complex number z. In
2.1. Analytical solutions for the OTOR model using the Lambert mathematical form the solution of equation (1) is:
function
R N
IðtÞ ¼ pðtÞ ; (2)
The OTOR model (one-trap one-recombination center) shown in ð1  RÞ2 W½k; ez  þ W½k; ez 2
Fig. 1 is the simplest model describing luminescence and is based
on delocalized transitions. The differential equations governing the
 Zt
traffic of electrons between the one trapping level, the NR no j1  Rj 1
z¼ ln þ pðtÞdt (3)
no ð1  RÞ NR 1R
to

where k ¼ 0 or k ¼ -1 for the two real branches of W(z)

correspondingly.
The advantages of using the Lambert function as a deconvolu-
tion function for luminescence signals are: (a) Fitting of experi-
mental TL data with the Lambert function produces accurate values
of activation energy E (Sadek et al., 2014) (b) W(z) can be used to fit
unusual shapes of TL glow curves, where the General Order Kinetics
(GOK) and Mixed Order Kinetics (MOK) equations may fail, (c) The
Lambert function W(z) is easy to program, since it is already
available in many commercial software packages like MATLAB,
MINUIT and MATHEMATICA, (d) Analysis of experimental data with
W(z) provides a physical interpretation of the luminescence signals
based on the retrapping ratio R. By contrast, analysis of data using
GOK kinetics is based instead of the empirical kinetic order
parameter b which has no direct physical meaning.
Fig. 1. The delocalized OTOR model containing a single trap and a single recombina-
tion center. Transition 1 represents thermal or optical excitation, transition 2 corre-
This latter case is of major importance, so it will be clarified by
sponds to the luminescence process from the conduction band into the recombination an example. A single TL peak is numerically generated using the
center, and transition 3 represents the retrapping process. OTOR model with input parameters E ¼ 1eV, s ¼ 1010 s1, and
 V. Pagonis et al. / Radiation Measurements 106 (2017) 3e12 5

t ¼ 0 (in m3), b is the general kinetic order parameter, s is an

00

empirical parameter acting as an “effective” frequency factor for

general order kinetics (in s1), E is the activation energy (in eV),
k ¼ Boltzmann constant (in eV. K1), t ¼ time (in s), and T is the
absolute temperature (in K). A linear heating rate b is used to heat
the sample (in K.s1), resulting in the temperature varying as
T ¼ To þ bt, where To is the temperature at time t ¼ 0 (in K).
It is clear that both equations fit both peaks accurately. However,
the fitting results shown in the inset table of Fig. 3 indicate that in
the case of R ¼ 0.25 the Lambert function reproduces exactly the
input values of E and R, and gives a FOM value two orders of
magnitude better than the GOK method which overestimates the E
value by 7%. In the case of R ¼ 1000, again both equations give
excellent fits. However, the Lambert based fit reproduces the E and
R values exactly, whereas the GOK equation gives completely
erroneous results of E ¼ 0.48 eV instead of the expected value
E ¼ 1 eV. This example shows that the differences between the GOK
Fig. 2. The two branches of the Lambert W function, showing the different regions of equation and the Lambert based analytical equations are of a deeply
the z parameter for which they are defined. physical nature, rather than due to mathematical differences be-
tween the two fitting models. The correct values of the kinetic
parameters are only obtained using the physically meaningful
n0 ¼ N ¼ 1010cm3, and for two different values of the retrapping Lambert equation, while the empirical GOK equations fail
ratio R ¼ 0.25 and R ¼ 1000. Fig. 3 shows that the two calculations completely. For further examples of using the Lambert functions,
produce TL glow curves with very different shapes for the two the reader is referred to the recent experiments by Kitis et al.
values of R. These two peaks were fitted using the GOK analytical (2016), who used the Lambert W function to fit isothermal lumi-
equations and also using Equation (2) in the form given by Sadek nescence signals in synthetic luminescence materials.
et al., (2015). The TL peak for R ¼ 0.25 was fitted using the first
branch, and the TL peak for R ¼ 1000 using the second branch of the
Lambert W function. The open circles represent the simulated 2.2. Localized transitions: the tunneling model of Jain et al. (2012)
peaks and the solid lines the fit by the Lambert equation (2) and by
the following well-known GOK equation: A second major recent theoretical development is the model
developed by Jain et al. (2012), which has provided a description of
2 3 b tunneling phenomena in a random distribution of electron-hole

 00 ZT
 b1

00 E 6 s ðb  1Þ E 07 pairs. The model is shown schematically in Fig. 4a. The main as-
IðTÞ ¼ s n0 exp  41 þ exp  0 dT 5
kT b kT sumptions of the model are the presence of a random distribution
To of electron-hole pairs, in which the concentration of holes (accep-
(4) tors) is much larger than the concentration of electrons (donors).
Thermal or optical excitation raises the electrons from the ground
with no representing concentration of trapped electrons at time into the excited state of the system. Tunneling takes place from the

Fig. 3. A single TL peak numerical simulated using the OTOR model for two different values of the retrapping ratio R ¼ 0.25 and 1000, as discussed in the text. The inset table gives
the results of fitting the numerically simulated peak using general order kinetics (GOK) and GOT equations. Excellent fits are obtained with both equations, however the GOK model
fails to reproduce the correct E values in the model.
6 V. Pagonis et al. / Radiation Measurements 106 (2017) 3e12

excited state of the electron trap into the recombination center, and significant temperature lag. For the experiment in Fig. 5a, the
to the nearest neighbors only. Kitis and Pagonis (2013) quantified irradiated aliquot of sample VRS3 was first preheated to a tem-
the semi-analytical model of Jain et al. (2012) by deriving analytical perature of 300  C, followed by measurement of its CW-IRSL signal
expressions for different experimental stimulation modes. The for 2000 s at 50  C. For the prompt isothermal measurements re-
analytical equations for excited state tunneling developed by Kitis ported in Fig. 5b for the Durango apatite, a combination of Pil-
and Pagonis (2013) are: kington HA-3 heat absorbing and Corning 7e59 (320e440 nm)
blue filter were used for light detection. This Durango apatite
h i
nðtÞ ¼ no exp  r0 FðtÞ3 (5) sample is known to exhibiting strong anomalous fading (Sfampa
et al., 2014).

0 1
Zt 3. Developments in comprehensive empirical models for
FðtÞ ¼ ln@1 þ 1:8 Adt 0A
(6) quartz
0
This section is organized as follows. Sections 3.1 and 3.2 sum-
h i marize some of the comprehensive empirical models which have
dn
LðtÞ ¼  ¼ AFðtÞ2 exp½FðtÞexp  r0 ½FðtÞ3 (7) been developed for quartz during the past 10 years. Some of these
dt models are based on localized transitions and these have been used
where A represents the rate of thermal/optical stimulation, r0 is the to describe time resolved experiments in quartz in the microsec-
dimensionless charge density, t is the elapsed time since the onds time scale, while other models are based on delocalized
beginning of experiment, no and n(t) are the initial and instanta- transitions and have been used mostly to describe phenomena in
neous concentration of point defects in the ground state corre- the seconds scale. In Section 3.3 we propose a new model for quartz
spondingly. The parameter A has the same physical meaning for which is based on both localized and delocalized transitions.
different experimental stimulation modes as the parameter p(t) in
Equation (1). 3.1. Empirical models for quartz based on localized transitions- time
The analytical equations (5)e(7) developed by Kitis and Pagonis resolved experiments
(2013) were used recently to describe luminescence signals from a
variety of feldspars and apatites in a quantitative manner For a comprehensive review of quartz models based on localized
(Polymeris et al., 2014; Sfampa et al., 2014; Pagonis et al., 2014a). transitions the reader is referred to the recent review paper by
Fig. 5a shows a typical example of fitting a continuous wave Chithambo et al. (2016). The main purpose of these types of models
infrared stimulated luminescence signal (CW-IRSL) from a ortho- is to describe quantitatively time resolved experiments in quartz in
clase sample with laboratory code VRS3, while Fig. 5b shows an the microseconds time scale. Extensive experimental studies of
example of a fitted isothermal signal from Durango apatite, luminescence lifetimes in quartz (see for example Chithambo et al.,
measured at 220  C. For both types of signals Equations (5)e(7) give 2016), have suggested a general energy scheme consisting of three
excellent fits to the experimental data. The experimental setup, independent radiative luminescence centers denoted by LH, LL and
sample preparation and experimental conditions for the experi- LS, and a non-radiative luminescence center denoted by R. These
ments shown in Fig. 5 were described previously in Polymeris et al. radiative centers are associated with distinct characteristic life-
(2014). Luminescence measurements were carried out using a Riso times denoted by tH, tL and tS correspondingly. Time resolved
TL/OSL reader (model TL/OSL-DA-15), equipped with a90Sr/90Y beta luminescence from sedimentary quartz annealed below 500  C is
particle source, delivering a nominal dose rate of 0.075 Gy/s. A dominated by a single component with a luminescence lifetime tH
9635QA photomultiplier tube was used with a 7.5 mm Hoya U-340 ~42 ms, while luminescence in samples annealed above 500  C is
filter (~340 nm, FWHM ~80 nm), and the IRSL stimulation wave- dominated by a characteristic lifetime of tL~32 ms. Pagonis et al.
length is 875 (±40) nm with the maximum power ~135 mW/cm2. (2010) presented a new kinetic model for time resolved experi-
All necessary heating was performed in a nitrogen atmosphere ments which includes thermal quenching in quartz, and which is
with a low constant heating rate of 2  C/s, in order to avoid based on the Mott-Seitz mechanism. In this model all

Fig. 4. (a) Tunneling phenomena from the excited state are described by Equation (7) in the text. (b) Tunneling from the ground state are described by Equation (14) in the text. (c)
Simultaneous irradiation and tunneling phenomena are described by Equation (15) in the text.
 V. Pagonis et al. / Radiation Measurements 106 (2017) 3e12 7

As the temperature of the sample is increased, electrons are

removed from the excited states according to the Boltzmann factors
and this leads to both a decrease of the intensity of the lumines-
cence signal and to a simultaneous decrease of the apparent
luminescence lifetime tL. The model of Pagonis et al. (2010) suc-
cessfully describes these variations of the luminescence lifetimes
and luminescence intensity when optical stimulation takes place in
temperatures between 20 and 200  C (Chithambo et al., 2016).
Pagonis et al. (2010) showed that their model can also describe
luminescence processes in quartz from a time scale of microsec-
onds (for TR-OSL experiments), up to a time scale of seconds (for TL
and OSL experiments). No modifications are required in the model
in order to transition from one time scale to the other. In later work
Pagonis et al. (2014b) expanded this model, in an effort to include
the experimentally observed lifetimes tH and tL for annealed
sedimentary quartz samples.
While the comprehensive quartz models described in this sec-
tion successfully combine localized and delocalized transitions over
time scales extending several orders of magnitude, the lumines-
cence mechanism in quartz is known to be very complex, and this
has led to the parallel development of several empirical compre-
hensive models which are based on delocalized transitions. One of
these models is described in the next subsection.

3.2. Empirical models for quartz based on delocalized transitions-

TL and OSL experiments

For a recent comparative review of comprehensive quartz

models based on delocalized transitions, the reader is referred to
the recent comprehensive paper by Friedrich et al. (2016). The first
empirical comprehensive quartz model was developed by Bailey
(2001), and it contains 5 electron traps and 4 hole centers. Levels
1e4 in the original Bailey model give rise to TL peaks at ~110  C,
230  C and two peaks around 330  C, and levels 3 and 4 also give
rise to OSL signals. Level 5 is a thermally disconnected deep elec-
tron center, and levels 6e9 are two hole reservoirs, a stable lumi-
nescence center and a non-radiative recombination center
correspondingly. The model by Bailey (2001) was expanded by
Pagonis et al. (2008) to include two additional levels 10 and 11, in
order to simulate the experimentally observed thermally trans-
ferred OSL (TT-OSL) signals.
In an extensive simulation study, Pagonis et al. (2011b) showed
that the model by Pagonis et al. (2008) can simulate a very wide
range of experimental results, as well as various experimental
protocols used for several TL/OSL dating techniques. This study
simulated successfully the following: (a) quantitative description of
TL glow curves and their dose response, (b) OSL bleach curves and
their dose response (c) the predose technique in quartz, (d) pho-
Fig. 5. (a) CW-IRSL experimental data for orthoclase sample VRS3, fitted using Equa-
totransfer phenomena (e) SAR-OSL dating protocol and (f) SAR-TL
tion (7) in the text. (b) Isothermal experimental data for Durango apatite, fitted using
Equation (7) in the text. dating protocols.
The type of delocalized transition model described in this sec-
tion cannot provide a description of TR-OSL experiments in quartz,
recombination transitions are localized within the recombination which take place in the microseconds time range. Therefore there is
center, in contrast to delocalized models of Section 3.2, in which all a need to combine delocalized and localized transitions in a more
charge transitions take place via the conduction and valence bands. comprehensive model, which will cover a wider range of time
One of the main challenges for these localized transition models scales. In the next section we propose such a new comprehensive
is the inclusion of thermal quenching effects, which are well known model for quartz, based on both localized and delocalized
experimentally in quartz and other dosimetric materials (Nikiforov transitions.
et al., 2001; Pagonis et al., 2011a, 2014b). In the quartz model of
Pagonis et al. (2010), thermal quenching arises from the competi- 3.3. A proposed new comprehensive model for quartz
tion between radiative and non-radiative electronic transitions
taking place within the recombination center. Unfortunately, in Fig. 6 shows a proposed new comprehensive model for quartz,
quartz the nature of both the traps and luminescence centers are which maintains the mathematical and physical characteristics of
still unknown or uncertain, and are the subject of current research. both the previous delocalized and localized models discussed in
sections 3.1 and 3.2. Level 1 in the model represents a shallow
8 V. Pagonis et al. / Radiation Measurements 106 (2017) 3e12

electron trapping level, which gives rise to a TL peak at ~110  C with

a heating rate of 5 K/s. Level 2 represents a generic “230  C TL” trap, X5
 X 11

dnc dni dni
while Levels 3 and 4 are usually termed the fast and medium OSL ¼ R  ACB nc ðNL  nL Þ  
dt dt dt
components and they yield TL peaks at ~330  C as well as give rise X  
i¼1 i¼10

to OSL signals. Level 5 is a deep electron center which is considered  nc nj Bj (11)

thermally disconnected, and levels 6 and 7 are thermally unstable, j¼6;7;9
non-radiative recombination centers (“hole reservoirs”). Level 8 is a
thermally stable, radiative recombination center often termed the 5
 X 11
 X
dnj 
dnv dnc X dni dni
“luminescence center” (L) and Level 9 is a thermally stable, non- ¼ þ þ  (12)
radiative recombination center termed a “killer” center (K). As dt dt i¼1
dt i¼10
dt j¼6;7;9
dt
discussed above, levels 10 and 11 were introduced in order to
The instantaneous luminescence IðtÞ resulting from the radia-
simulate the experimentally observed thermally transferred OSL
tive transition shown in the middle of Fig. 6 is defined as:
(TT-OSL) signals.
The important localized transitions in the model are shown in
IðtÞ ¼ AR nL (13)
the middle of Fig. 6, and are denoted by AR and ANR by following the
notation in the localized quartz model of Pagonis et al. (2010). As In these equations Ni are the total concentrations of electron
discussed in section 3.1, thermal quenching arises from the traps or hole centers (cm3), ni are the instantaneous concentra-
competition between these radiative (AR) and non-radiative (ANR) tions of trapped electrons or holes (cm3), si are the frequency
electronic transitions taking place within the recombination center. factors (s1), Ei are the electron trap depths below the conduction
The equations for the traffic of charge carriers in the new proposed band or hole trap depths above the valence band (eV), Ai (i ¼ 1…5,
model are as follows: and i ¼ 10,11) are the conduction band to electron trap transition
probability coefficients (cm3 s1), Aj (j ¼ 6…9) are the valence band
" #   to hole trap transition probability coefficient (cm3 s1), valence
dni Eith E band to trap transition probability coefficients (cm3 s1) and Bj
¼ nc ðNi  ni ÞAi  ni P qoi exp   ni si exp  i ;
dt kB T kB T (j ¼ 6…9) are the conduction band to hole center transition prob-
(8) ability coefficients (cm3 s1). The photo-eviction constant is q0i
(s1) at T ¼ ∞, the thermal assistance energy Ethi (eV) and P (s1) is
the rate of optical stimulation. AR and ANR (s1) are the radiative and
ði ¼ 1; …5 and i ¼ 10; 11Þ non-radiative transition rates necessary to simulate the thermal
quenching process, and their numerical values are taken from
Pagonis et al. (2010).

 It is noted that Equations (8)e(9) describe exclusively delo-
Еj
dnj   k
BT calized transitions involving the conduction and valence bands,
¼ nv Nj  nj Aj  nj sj e  nc nj Bj ; (9) while Equation (10) refers to the luminescence center (level j ¼ 8 in
dt
the model), which is based on localized transitions, and which is the
basis of the new proposed model. A new set of parameters are used
j ¼ 6; 7; 9 in the model to describe localized transitions within the lumines-
cence center. This center is described in terms of electronic states
and electronic transitions, by following the mathematical formalism
dnL in Pagonis et al. (2010). By following the notation of Pagonis et al.
¼ ACB nc ðNL  nL Þ  A8 nv nL  AR nL  nL ANR expðW=kB TÞ
dt (2010), let us denote by nL the concentration of electrons in the
 nL sL exp½EL =kT excited state of the luminescence center, and by NL the corre-
sponding total concentration of these traps in the crystal. The
(10)
quantities nL and NL are then completely analogous to the quanti-
ties n2 and N2 in the model of Pagonis et al. (2010).

Fig. 6. The new proposed model for quartz, which contains both localized and delocalized transitions. The various transitions and the parameters used in the model are described in
the text. This energy scheme describes thermal quenching using the Mott-Seitz mechanism.
 V. Pagonis et al. / Radiation Measurements 106 (2017) 3e12 9

It must be emphasized that the key difference between the time range including thermal quenching, as shown in Fig. 8, and (c)
delocalized model by Pagonis et al. (2008) and the new model is the complex TT-OSL quartz dating protocol discussed in Pagonis
Equation (10), in which the luminescence center is described in et al. (2011b) is shown in Fig. 9. In summary, the new model
terms of electronic states and electronic transitions, instead of in maintains the physical and mathematical properties of the earlier
terms of holes and delocalized transitions. models, by using both their localized and delocalized characteris-
Equation (10) describes mathematically five possible transitions tics. Further testing of the model and comparison with experi-
involving the electrons in the excited state of the luminescence mental data is in progress and will be presented elsewhere.
center, as follows. The first term ACB nc ðNL  nL Þ describes the time
rate of electrons being trapped with a rate coefficient ACB from the
CB into the excited state nL . This term is positive, since it leads to an 4. Models of ground state tunneling in random distributions
increase of the concentration of electrons nL in the excited state of of defects
the recombination center, by the capturing of electrons from the CB.
The second term A8 nv nL describes holes being trapped with a This section presents a general overview of ground state
probability coefficient A8 from the valence band into the excited tunneling phenomena in systems of random defects. Quantum
electronic state. This term is negative, since it leads to a decrease of mechanical tunneling and the associated phenomenon of “anom-
the concentration of electrons nL by the capturing of holes from the alous fading” of luminescence signals are now well established as
VB. The third term AR nL describes electrons undergoing the radia- dominant mechanisms in feldspars and apatites (Pagonis and Kitis,
tive transition from the excited into the ground state of the center, 2015). Two types of tunneling processes have been investigated
and the fourth term nL ANR expðW=kB TÞ describes the non- experimentally and by the development of appropriate models. The
radiative localized transition which competes for electrons from first type of quantum mechanical tunneling is considered to take
the excited state. The last term nL sL eEL =kT describes the thermal place directly from the ground state of the trap, as shown in Fig. 4b.
properties of the electrons in the excited state, using the frequency The second type of quantum mechanical tunneling is considered to
coefficient sL and thermal activation energy EL. It is noted that the take place via the excited state of the system of electron-hole pairs,
above equation (10) for dnL =dt contains both localized and delo- and was discussed briefly in Section 2.2 of this paper and shown in
calized transitions. Fig. 4a. Two possible complementary modeling approaches have
When using the model in Fig. 6 with the original values of the been used in the literature to simulate ground state tunneling in
parameters in Pagonis et al. (2008), it was found that the conduc- random distribution of defects: a macroscopic description using
tion band empties very slowly, on a time scale of ~0.1 s. This value is
in contradiction to several experimental and simulation studies,
which show that the CB probably empties much faster, perhaps
within a few microseconds (Chithambo et al., 2016). This deficiency
of this type of model was discussed in Pagonis et al. (2010, page
909), and is due to the very low values of total concentrations of
electrons used in the empirical quartz models. In order to achieve a
free carrier lifetime of t  1 ms, it was found necessary to increase
the values of all electron and hole concentrations in the model by a
factor of 5x104 , while leaving all other parameters in the model
unchanged. From a physical point of view, when using higher
values of the total concentrations of carriers in the model, all pro-
cesses in the system will take place much faster, but the relative
distribution of electrons and holes in the various traps and centers
remains the same. The final values of all parameters used in the
new comprehensive model of Fig. 6 are shown in Table 1.
The new model shown in Fig. 6 was tested extensively, and it
provides a quantitative description of luminescence phenomena in
quartz from microseconds up to seconds. Specifically, the new
model can describe (a) TR-OSL experiments (in the microseconds
time range) as shown in Fig. 7, (b) TL experiments in the seconds

Table 1
Kinetic parameters for the new proposed model shown in Fig. 4. The various pa-
rameters are discussed in the text.

Ni Ei eV si Ai Bi q0i Eth
i eV
cm3 s1 3 1 s1
cm s

1 7.5e11 0.97 5e12 1e-8 0.75 1

2 5e11 1.55 5e14 1e-8 0 0
3 2e12 1.73 6.5e13 5e-9 6 0.1
4 1.2e13 1.8 1.5e13 5e-10 4.5 0.13
5 2.5e15 2 1e10 1e-10 0 0
6 1.5e13 1.43 5e13 5e-7 5e-9 0 0
7 5e14 1.75 5e14 1e-9 5e-10 0 0
8 1.5e15 5 1e13 1e-10 1e-10 0 0
Fig. 7. Time resolved (TR) experimental data for quartz, in the microseconds scale. (a)
9 6 e16 5 1e13 1e-14 3e-10 0 0
Luminescence lifetime as a function of the stimulation temperature and (b) Total TR-
10 2.5e14 1.65 3.9e13 1e-11 0.01 0.2
OSL area as a function of stimulation temperature. The solid lines represent the best
11 2e14 1.6 5e12 6e-12 0 0
fits obtained from the new proposed model in Fig. 6.
10 V. Pagonis et al. / Radiation Measurements 106 (2017) 3e12

developed the following new analytical equation describing the

effect of anomalous fading on the dose response curves (DRCs) of
naturally irradiated samples, based on the model by Li and Li (2008)
shown in Fig. 4c:
      
Dn Do s 3
Ln ðDn Þ ¼ 1  exp  M exp  r0 ln (15)
D0 DR

where Ln is the luminescence signal, Dn is the natural irradiation

dose (Gy), Do is the characteristic dose (Gy), M is the total number of
traps, DR is the natural irradiation rate (Gy/ka) and s (s1) is the
tunneling frequency.

4.2. Ground state tunneling in random defects: the Monte Carlo

approach

In this subsection we describe a different modeling approach

Fig. 8. Simulated TL quartz data in the seconds scale for different heating rates b,
demonstrating the effect of thermal quenching. These simulated data were also ob-
based on Monte Carlo techniques. There have been several efforts to
tained from the new proposed model in Fig. 6. develop models for luminescence signals in feldspars, based on a
random distribution of donor-acceptor pairs. One of the basic as-
sumptions of these models is that whenever an electron trap is
occupied, the electron will always have the same nearest hole with
which to recombine, i.e. the number density of trapped holes in the
phosphor is assumed to be much larger than the concentration of
trapped electrons. Larsen et al. (2009) presented a numerical
Monte Carlo model that simulated the processes of charge loss,
charge creation and charge recombination in feldspar. In contrast to
the assumptions made by previous analytical models, they
assumed that the number density of electrons and holes are equal
at all times. The focus of their study was to reproduce the experi-
mentally observed values of the well-known g-factor describing
anomalous fading effects. These authors were not able to get reli-
able results for bulk crystals, and obtained good agreement with
experiment only when they assumed that the material consisted of
small nanocrystals, and that charge carriers were allowed to
recombine only within these smaller volumes.
Pagonis and Kulp (2017) presented a different version of the
model by Larsen et al. (2009), in which the number density of ac-
ceptors far exceeds that of donors. The new version of the model
Fig. 9. Simulated TT-OSL protocol dose response, obtained with the model shown in
Fig. 6. was used to simulate the loss of charge due to ground state
tunneling, as well as the charge creation by natural irradiation of
the samples. The results from the model compared well with the
analytical equations (14) and (15) presented in the previous sub-
differential equations, and a microscopic descriptions based on section. The simulations can describe the loss of charge on a wide
Monte Carlo simulations. These two approaches are discussed in variety of time scales, from microseconds to thousands of years. The
the next two subsections. effect of crystal size, charge carrier density, natural irradiation dose
rate and total number of charge carriers were studied in a quanti-
tative manner. Finally, Pagonis and Kulp (2017) examined the
4.1. Ground state tunneling in random defects: the differential possibility of extending the version of the model to describe
equation approach luminescence signals originating in the nearest neighbor hopping
mechanism in feldspars. The results from the model were
The differential equation description of ground state tunneling compared with experimental data from time-resolved infrared
is now well established and provides a macroscopic description of stimulated luminescence (TR-IRSL) in these materials.
the system. Recently Pagonis and Kitis (2015) considered mathe- The Monte Carlo simulations can provide valuable insight into
matical aspects of the two ground state tunneling models by the various factors which affect the luminescence mechanism in
Huntley (2006) and by Li and Li (2008), shown in Fig. 4b and c these materials. Fig. 10 shows an interactive simulation of the effect
correspondingly. The loss of charge in a distribution of random of various parameters on the loss of charge in feldspars (Kulp and
defects is described by the following well-known analytical equa- Pagonis, 2016). The simulation consists of several “buttons” which
tion (Tachiya and Mozumder, 1974; Huntley, 2006): can be activated in any sequence by the user, to investigate the
h i changes taking place in the curve n(t) describing the loss of charge
nðtÞ ¼ no exp  r0 lnð1:8stÞ3 ; (14) due to ground state tunneling. In Fig. 10 these buttons vary the
tunneling frequency s, the potential barrier penetration constant a,
where r0 is the dimensionless charge density describing the system the dimensions of the cube and the number of Monte Carlo runs
and s is the tunneling frequency. Pagonis and Kitis (2015) (number of solids). The solid line in Fig. 10 represents the analytical
 V. Pagonis et al. / Radiation Measurements 106 (2017) 3e12 11

Fig. 10. Interactive demonstrations of ground state tunneling. The user can vary the various parameters in the model and immediately see the effect of each parameter in the loss of
charge over time. Variation of the parameters is achieved by activating the various “buttons” shown here. The solid line represents the analytical equation (14).

equations described by the analytical equation (14). The simula- how this mathematical function can be used to fit experimental TL,
tions were developed by using the command Manipulate in Math- OSL and LM-OSL data based on delocalized transitions within the
ematica. The computer code for the interactive demonstrations in GOT model. The availability of the analytical Lambert solutions for
Fig. 10 are available to all users via the Wolfram Demonstration the GOT model presents a significant mathematical step in the
Project, a web-based open-access collection of interactive scientific development and understanding of basic processes based on
animations (Kulp and Pagonis, 2016). delocalized transitions, and how they may be distinguished
The advantages of using a Monte Carlo method, as opposed to experimentally from localized transitions. For example, recent
the differential approach, are (Pagonis and Kulp, 2017): work has shown that it may be possible for experiments to
distinguish between localized and delocalized processes in a ma-
(a) the method is fast, efficient and avoids numerical in- terial, by analyzing isothermal TL signals assuming different models
tegrations required in the differential equation approach. (Kitis et al., 2016).
(b) it can be used to produce accurate results in cases of low Similarly, in the case of localized tunneling transitions for
stimulation probability, in which it is known that the random distributions of defects, the analytical equations developed
analytical equations of Kitis and Pagonis (2013) fail. by Kitis and Pagonis (2013) based on the model by Jain et al. (2012),
(c) it can be used for both freshly irradiated samples and for represent a new rich area of research which is still being explored
irradiated samples which underwent thermal or optical pre- both experimentally and with modeling work. This work has led to
treatments. a better understanding of ground state tunneling phenomena in
(d) it can also be used to describe time-resolved experiments feldspars, apatites and similar natural and synthetic luminescence
based on Mott hopping processes (Mott and Davis, 1979). materials. However, there is much work to be done in order to
understand the associated phenomena of tunneling from the
5. Conclusions and future challenges in luminescence excited state of the electron-hole pair in a luminescence material.
modeling Closely associated with tunneling from the excited state of the
electron-hole pair are two fundamental areas of research which are
This paper has presented an overview of recent developments in of importance for both luminescence dating and luminescence
luminescence modeling. Section 2 summarized presented recent dosimetry. These are the temperature dependence of the lumi-
theoretical developments based on the Lambert W-function, and nescence signal in feldspars and apatites, and the exact nature of
12 V. Pagonis et al. / Radiation Measurements 106 (2017) 3e12

the various luminescence pathways including the band tail states in Kitis, G., Vlachos, N.D., 2013. General semi-analytical expressions for TL, OSL and
other luminescence stimulation modes derived from the OTOR model using the
these materials. Understanding these various pathways is a key
Lambert W-function. Radiat. Meas. 48, 47e54.
part of improving the various proposed experimental protocols for Kitis, G., Polymeris, G.S., Sfampa, I.K., Prokic, M., Meriç, N., Pagonis, V., 2016. Prompt
dating these materials (postIR-IR protocols, MET-IR protocols etc). isothermal decay of thermoluminescence in MgB4O7:Dy,Na (MBO) and LiB4O7:
Section 3 presented a new model for luminescence in quartz, Cu, in (LBO). Radiat. Meas. 84, 15e25.
Kulp, C., Pagonis, V., 2016. Interactive Monte Carlo Demonstrations of Ground State
which is applicable for time scales ranging from microseconds to Tunneling Phenomena in Solids. Published in the Wolfram Demonstration
seconds. This type of model which combines localized and delo- Project. http://demonstrations.wolfram.com/
calized transitions could also be applicable to the study of lumi- MonteCarloModelOfChargeRecombinationInFeldspars/.
Larsen, A., Greilich, S., Jain, M., Murray, A.S., 2009. Developing a numerical simu-
nescence signals from materials like YPO4 double doped with lation for fading in feldspar. Radiat. Meas. 44, 467.
lanthanides. These materials exhibit temperature dependent Li, B., Li, S.H., 2008. Investigations of the dose-dependent anomalous fading rate of
tunneling phenomena and apparently can be described by a com- feldspar from sediments. J. Phys. D. Appl. Phys. 41, 225502.
Mandowski, A., Bos, A.J.J., 2011. Explanation of anomalous heating rate dependence
bination of localized and delocalized transitions. Often such ma- of thermoluminescence in YPO4:Ce3þ, Sm3þ based on the semi-localized tran-
terials also demonstrate the anomalous heating rate effect, in sition (SLT) model. Radiat. Meas. 46, 1376e1379.
which the area under a TL glow curve increases with heating rate, Mott, N.F., Davis, E.A., 1979. Electronic Processes in Non-crystalline Materials, sec-
ond ed. Clarendon Press, Oxford, pp. 216e218.
instead of the opposite behavior that one sees in materials with Nikiforov, S.V., Milman, I.I., Kortov, V.S., 2001. Thermal and optical ionization of F-
thermal quenching (Mandowski and Bos, 2011). centers in the luminescence mechanism of anion-defective corundum crystals.
Section 4 in this paper presented two different descriptions of Radiat. Meas. 33, 547.
Pagonis, V., Kitis, G., 2015. Mathematical aspects of ground state tunneling models
ground state tunneling phenomena in a random distribution of
in luminescence materials. J. Luminescence 168, 137e144.
defects; a macroscopic differential equations approach, and a Pagonis, V., Kulp, C., 2017. Monte Carlo simulations of tunneling phenomena and
microscopic Monte Carlo approach. Both of these approaches can nearest neighbor hopping mechanism in feldspars. J. Lumin. 81, 114e120.
easily be extended for the case of excited state tunneling. While the Pagonis, V., Wintle, A.G., Chen, R., Wang, X., 2008. A theoretical model for a new
dating protocol for quartz based on thermally transferred OSL (TT-OSL). Radiat.
differential equation approach has been well studied, little work Meas. 43, 704e708.
has been done on the relevant Monte Carlo research for tunneling Pagonis, V., Ankjærgaard, C., Murray, A.S., Jain, M., Chen, R., Lawless, J., Greilich, S.,
materials. This is an area where future research can yield new re- 2010. Modelling the thermal quenching mechanism in quartz based on time-
resolved optically stimulated luminescence. J. Lumin 130, 902e909.
sults, of importance to both fundamental understanding of Pagonis, V., Chen, R., Maddrey, J.W., Sapp, B., 2011a. Simulations of time-resolved
tunneling luminescence and for practical dosimetry applications. photoluminescence experiments in a-Al2O3:C. J. Lumin 131, 1086e1094.
Pagonis, V., Chen, R., Kitis, G., 2011b. On the intrinsic accuracy and precision of
luminescence dating techniques for fired ceramics. J. Archaeol. Sci. 38,
References 1591e1602.
Pagonis, V., Jain, M., Thomsen, K.J., Murray, A.S., 2014a. On the shape of continuous
Bailey, R., 2001. Towards a general kinetic model for optically and thermally stim- wave infrared stimulated luminescence signals from feldspars: a case study.
ulated luminescence of quartz. Radiat. Meas. 33, 17e45. J. Lum 153, 96e103.
Chen, R., Pagonis, V., 2011. Thermally and Optically Stimulated Luminescence: a Pagonis, V., Chithambo, M.L., Chen, R., Chruscin  ska, A., Fasoli, M., Li, S.H., Martini, M.,
Simulation Approach. Chichester: Wiley and Sons. Ramseyer, K., 2014b. Thermal dependence of luminescence lifetimes and radi-
Chithambo, M., Ankjærgaard, C., Pagonis, V., 2016. Time-resolved luminescence oluminescence in quartz. J. Lumin 145, 38e48.
from quartz: an overview of contemporary developments and applications. Polymeris, G.S., Giannoulatou, V., Sfampa, I.K., Tsirliganis, N.C., Kitis, G., 2014. Search
Phys. B 481, 8e18. for stable energy levels in materials exhibiting strong anomalous fading: the
Friedrich, J., Kreutzer, S., Schmidt, C., 2016. Solving ordinary differential equations to case of apatites, 2014. J. Lumin. 153, 245e251.
understand luminescence: ‘RLumModel’, an advanced research tool for simu- Sadek, A.M., Eissa, H.M., Basha, A.M., Kitis, G., 2014. Resolving the limitation of the
lating luminescence in quartz using R. Quat. Geochronol. 35, 88e100. peak fitting and peak shape methods in the determination of the activation
Horowitz, Y.S., Eliyahu, I., Oster, L., 2015. Kinetic simulations of thermolumines- energy of thermoluminescence glow peaks. J. Lumin 146, 418e423.
cence dose response: long overdue confrontation with the effects of ionization Sadek, A.M., Eissa, H.M., Basha, A.M., Carinou, E., Askounis, P., Kitis, G., 2015. The
density. Rad. Prot. Dosim. 1e17. http://dx.doi.org/10.1093/rpd/ncv495. deconvolution of thermoluminescence glow curves using general expressions
Huntley, D.J., 2006. An explanation of the power-law decay of luminescence. J. Phys. derived from one trap-one recombination (OTOR) level model. Appl. Radiat.
Condens. Matter 18, 1359. Isot. 95, 214e221.
Jain, M., Guralnik, B., Andersen, M.T., 2012. Stimulated luminescence emission from Sfampa, I.K., Polymeris, G.S., Tsirliganis, N.C., Pagonis, V., Kitis, G., 2014. Prompt
localized recombination in randomly distributed defects. J. Phys. Condens. isothermal decay of thermoluminescence in an apatite exhibiting strong
Matter 24, 385402 (12pp). anomalous fading, 2014. Nucl. Meth. Instr. B 320, 57e63.
Kitis, G., Pagonis, V., 2013. Analytical solutions for stimulated luminescence emis- Tachiya, M., Mozumder, A., 1974. Decay of trapped electrons by tunnelling to
sion from tunneling recombination in random distributions of defects. J. Lum. scavenger molecules in low temperature glasses. Chem. Phys. Lett. 28, 87e89.
137, 109e115.