Radiation Protection Dosimetry Vol. 106, No. 1, pp.

33–43 (2003)
Published by Nuclear Technology Publishing
 2003 Crown coypright

THE RESPONSE OF LiF THERMOLUMINESCENCE

DOSEMETERS TO PHOTON BEAMS IN THE ENERGY RANGE
FROM 30 kV X RAYS TO 60Co GAMMA RAYS
S. D. Davis(1,2), C. K. Ross(1), P. N. Mobit(3), L. Van der Zwan(1), W. J. Chase(4) and K. R. Shortt(5)
(1)
Ionizing Radiation Standards, Institute for National Measurement Standards
National Research Council, Ottawa, ON K1A 0R6, Canada
(2)
Medical Physics Unit, McGill University Health Centre, Montreal General Hospital
1650 Cedar Ave, Montreal, QC H3G 1A4, Canada
(3)
Department of Oncology, University of Calgary and Department of Medical Physics
Tom Baker Cancer Centre, 1131-29 Street NW, Calgary, AB, T2N 4N2, Canada
(4)
Ontario Power Generation, 1549 Victoria Street East, Whitby, ON L1N 9E3, Canada
(5)
Dosimetry and Medical Radiation Physics Section, International Atomic Energy Agency

Downloaded from rpd.oxfordjournals.org by guest on February 1, 2011

P. O. Box 200, Wagramer Strasse 5, A-1400 Vienna, Austria

Received March 28 2003, amended May 27 2003, accepted June 6 2003

Abstract — The energy response of standard (TLD-100) and high-sensitivity (TLD-100H) LiF thermoluminescence dosemeters
(TLDs) has been studied for photon beams with mean energies from about 25 keV to 1100 keV. Canadian primary standards for
air kerma were used to establish the air kerma rates for each of the photon beams. TLDs were mounted in a PMMA holder and
the air kerma response was measured as a function of energy. The EGSnrc Monte Carlo code was used to model the TLD holder
and calculate the absorbed dose to the TLD chip per unit air kerma for each beam. The measured and calculated results were
combined to obtain the intrinsic dose response of the TLD chip. Broadly, our results are consistent with existing data, which
show a marked difference in the energy dependence of the two materials. However, the precision of our measurements (standard
uncertainty of about 0.6%) has permitted the identification of features that have not been noted before. In particular, the energy
dependence of the two materials is quite different in the important energy region delimited by 137Cs and 60Co gamma rays.

INTRODUCTION filters (5), either to provide build-up, or to provide spec-

tral information. Any attempt to use calculational tech-
Because of their small size, high sensitivity and ease
niques such as Monte Carlo codes to model the response
of use, thermoluminescence dosemeters (TLDs) are
of these composite structures requires knowledge of the
widely used for radiation dosimetry (1–3). The most com-
energy response of the TLD element.
monly used TLD material is based on crystalline LiF,
Bilski et al (6) suggest that the properties of LiF:
which has been doped with small quantities of Mg and
Mg,Cu,P produced by different manufacturers may not
Ti, and is denoted by LiF:Mg,Ti. More recently, a high
be the same. The energy response of the formulation of
sensitivity form of LiF has become available, and is
LiF:Mg,Cu,P denoted by MCP-N has been studied (6–8)
denoted by LiF:Mg,Cu,P, because the dopants are now
extensively by a group at the Institute of Nuclear Phys-
Mg, Cu and P.
ics (INP) in Kraków, Poland. MCP-N was developed at
The characteristics of both LiF:Mg,Ti and
INP so that the manufacturing process could be care-
LiF:Mg,Cu,P are summarised by McKeever et al (3) and
fully controlled. Olko et al (7) report measurements of
they provide extensive references to the original litera-
the energy response of MCP-N for photon energies from
ture describing these materials. The present work is con-
about 12 to 1200 keV, and provide a microdosimetric
cerned with the energy dependence of the response of
interpretation of the observed variation.
these materials to ionising radiation. Accurate knowl-
In addition to MCP-N, LiF:Mg,Cu,P is available as
edge of the energy response is important for fundamen-
GR-200 from Beijing Radiation Detector Works in
tal reasons, because any detailed model of thermolumi-
China and as TLD-100H from Thermo Electron RM&P
nescence must be able to explain the observed energy
(formerly Harshaw TLD) in the USA. Sáez-Vergara
dependence. The energy response is also of practical
et al (9) have studied the energy dependence of the vari-
importance (4) because TLDs are often used in radiation
ous formulations of LiF:Mg,Ti and LiF:Mg,Cu,P. Their
fields that have a broad spectrum of energies. In the
TLD chips were mounted in either PMMA or PVC hol-
areas of radiation protection and environmental moni-
ders, and the results of their measurements are summar-
toring, TLD chips are often mounted behind various
ised in Figures 1 and 2. These figures show that the
holder has a major effect on the air kerma response.
They also indicate that the response does not depend
Contact author E-mail: carl.ross@nrc.ca strongly on the details of the manufacturing process.

33
 S. D. DAVIS, C. K. ROSS, P. N. MOBIT, L. VAN DER ZWAN, W. J. CHASE and K. R. SHORTT
In previous work Shortt et al (10) measured the air
kerma response of TLD-100 to 137Cs and 60Co gamma
rays. They showed that the holder could have a signifi-
cant effect on the response, but also noted that the
intrinsic response was about 2.5% greater for 137Cs than
for 60Co. Although the radiation absorption character-
istics of LiF would not predict such a difference, it was
consistent with a more extensive study of the energy
dependence of LiF:Mg,Ti carried out by Tochilin
1.6
1.4
Relative air kerma response
LiF:Mg,Ti
1.2
1.0
0.8
10 100
Mean photon energy (keV)
Figure 1. Graphical summary of the measured data tabulated
by Sáez-Vergara et al (9) for various formulations of LiF:Mg,Ti
and LiF:Mg,Cu,P. The data were obtained by a Spanish sec-
ondary calibration laboratory, INTE-UPC, and by the Spanish
national laboratory for ionising radiation, CIEMAT. In this
case the TLD chips were mounted in a PMMA holder. The
standard uncertainty on each datum point ranges from 0.05%
to 2%, although it is generally about 1%.
1.4
1.2
Relative air kerma response
1.0
0.8
0.6
10 100
Mean photon energy (keV)
Figure 2. Same as Figure 1 except that the TLD chips were
mounted in a PVC holder.
et al (11). The results of Sáez-Vergara et al (9) (Figures 1
and 2) are also consistent with the results of Shortt et al
for LiF:Mg,Ti. However, Sáez-Vergara et al point out
that the change in response of LiF:Mg,Cu,P from 137Cs
to 60Co is different from that of LiF:Mg,Ti.
In this paper only the LiF formulations commercially
available from Thermo Electron RM&P are considered.
They are denoted by TLD-100 and TLD-100H for
LiF:Mg,Ti and LiF:Mg,Cu,P, respectively. The results
of measurements on the energy dependence of the air
kerma response of TLD-100 and TLD-100H for photons
in the energy range from 25 keV to 1100 keV are
reported. Monte Carlo calculations have been used to
model the TLD holder and to determine the dose to the
TLD chip per unit air kerma. Thus, it was possible to
extract the energy dependence of the dose response of
Downloaded from rpd.oxfordjournals.org by guest on February 1, 2011
the bare TLD chip. The observed structure is correlated
with the calculated electron spectrum and our results are
compared with earlier data, including the results
obtained with MCP-N.
MATERIALS AND METHODS
The TLD-100 and TLD-100H chips have slightly dif-
LiF:Mg,Cu,P
ferent formats. Both are 0.38 mm thick, but the TLD-
100 chips have a square cross section, 3.2 mm on a side,
1000 whereas the TLD-100H chips have a circular cross sec-
tion, 3.6 mm in diameter. TLD-100 has a density of
2.64 g cm⫺3 and the dopants Mg and Ti are present at
concentrations of 200 and 10 ppm by weight, respect-
ively. The density of TLD-100H is 2.48 g cm⫺3 and the
dopant concentrations are 3000, 2000 and 40 ppm by
weight for P, Mg and Cu, respectively*. Both TLD-100
and TLD-100H contain the naturally occurring isotopic
ratio of 6Li to 7Li. Our protocol for using TLD-100 is
described in detail by Janovský and Ross (12). Briefly,
the chips are first annealed at 400ºC for 1 h, followed
by a second anneal at 100ºC for 2 h. Each time the chips
are removed from the oven, they are rapidly cooled on
a large brass block. After irradiation, but before readout,
the TLDs are annealed for 12 min at 100ºC, to clear the
LiF:Mg,Ti low energy traps. The TLD light output is read using a
Victoreen 2800M reader. The response of the photomul-
tiplier used in the reader has its maximum in the blue–
green region of the spectrum, and a blue glass filter is
installed between the planchet and the entrance window
of the photomultiplier. The readout cycle consists of a
LiF:Mg,Cu,P preheat to 100ºC which takes about 2 s, followed by
heating at 5ºC s⫺1 to 240ºC. The temperature is held at
240ºC for 15 s, giving a total read time of 45 s. Nitrogen
gas, at a flow rate of 16 cm3 s⫺1 is used to remove oxy-
gen from the sample compartment. The planchet is
1000
* The composition and density data for TLD-100 and TLD-
100H were kindly provided by K. J. Velbeck of Thermo Elec-
tron RM&P.
34
 ENERGY RESPONSE OF LiF TLDs
allowed to cool to 40ºC before the chip is removed from
the reader.
The TLD-100H chips are annealed for 15 min at
240ºC, followed by rapid cooling. After irradiation but
before readout, they are annealed for 12 min at 130ºC.
For these short anneal times, there is some question as
to exactly how long the chips were at the specified tem-
perature. The approach to equilibrium was measured
after the ovens had been opened long enough to insert
the TLD holders. It is estimated that, in both cases, the
chips were at the specified temperature for about 10
min. The readout cycle is the same as that used for the
TLD-100 chips.
The calibration procedure is the same for both the
TLD-100 and TLD-100H chips (12). The chips are
irradiated in groups of 25 using 60Co gamma rays in a
chips that are placed symmetrically about the central
axis. In addition to supporting the chips, the holder is
also designed to provide charged particle equilibrium at
the position of the TLD. For the 60Co and 137Cs
irradiations, the thickness of the PMMA overlying the
TLDs was 6.2 mm, while for the X ray irradiations the
thickness was reduced to 1.6 mm.
The 60Co and 137Cs irradiation facilities at the
National Research Council (NRC) have been described
in some detail by Shortt et al (10). The 60Co air kerma
has been established using a graphite cavity chamber as
a primary standard (13). The 137Cs air kerma was obtained
using commercial ionisation chambers calibrated against
the 60Co primary standard. The estimated change in the
response of the chambers between 60Co and 137Cs was
obtained using Monte Carlo calculations (14). The stan-
dard uncertainty on either the 60Co or 137Cs air kerma

Downloaded from rpd.oxfordjournals.org by guest on February 1, 2011

field where the air kerma rate is accurately known. After
readout, the individual calibration factor, defined as the
ratio of the delivered air kerma to the TLD reading, is
determined for each of the 25 chips. The TLDs are
annealed again, and divided into groups of five. One
group is irradiated in a 60Co reference field to determine
if there has been any overall change in sensitivity. If so,
this change is applied to the readings of the remaining
four groups, which are irradiated in unknown fields. The
work of Janovský and Ross (12) and of Shortt et al (10)
indicated that this procedure led to results where the
standard uncertainty on the mean of a set of five read-
ings was typically about 0.5%.
The holder used to irradiate the TLDs is shown in
Figure 3. It is constructed from PMMA, and holds five
TLD cavity
is approximately 0.5%.
The 60Co irradiations to establish the individual chip
sensitivities and the 60Co reference irradiations were
carried out using an irradiator manufactured by Atomic
Energy of Canada Ltd and identified as the Theratron
Junior. The irradiations were done at a point 3 m from
the source where the field had a square cross section,
80 cm on a side and the air kerma rate was approxi-
mately 2.4 ⫻ 10⫺2 mGy s⫺1.
The X ray air kerma is established with a standard
uncertainty of approximately 0.5% using free-air cham-
bers. One chamber is used to cover the energy range up
to 60 kV, while the second can be used up to 300 kV (15).
Filter assemblies are available to generate the ISO series
of narrow spectrum beams (16). Table 1 lists the X ray
beams that were used and gives their measured half-
value layer (HVL). The TLDs were irradiated at 1 m
from the X ray source, where the field has a circular
cross section, 15 cm in diameter.

Table 1. Characteristics of the X ray beams used for

measurements and calculations. Each beam is intended to
satisfy the requirements of the ISO narrow spectrum
series (16). The second column is the measured HVL of the
NRC beam, while the third column is the HVL reported
for the GSF beams (22). The final column is the mean photon
energy calculated using the GSF spectra.

Identifier HVL Mean energy

(keV)
NRC beams GSF beams
5 cm
N-30 1.20 Al 1.12 Al 24.2
Figure 3. Basic design of the PMMA holder used for the N-40 0.083 Cu 0.081 Cu 32.5
irradiations. The holder is 32 mm in diameter and has five cyl- N-60 0.23 Cu 0.23 Cu 47.3
indrical cavities, each large enough to accommodate one TLD N-80 0.58 Cu 0.57 Cu 64.5
chip. The TLD cavities are placed symmetrically with respect N-100 1.10 Cu 1.09 Cu 82.6
to the flat faces of the holder, but different wall thicknesses N-150 2.37 Cu 2.29 Cu 117
were used, depending on the beam energy. For the X ray N-200 4.00 Cu 3.91 Cu 164
irradiations, the wall thickness was 1.6 mm, while for the 137Cs N-250 5.19 Cu 5.11 Cu 207
and 60Co irradiations the thickness was 6.2 mm.

35
 S. D. DAVIS, C. K. ROSS, P. N. MOBIT, L. VAN DER ZWAN, W. J. CHASE and K. R. SHORTT
The Monte Carlo calculations were carried out using
the EGSnrc code (17). The user code DOSRZnrc was
used to calculate the absorbed dose to the TLD chip per
unit fluence, while the electron spectrum in the chip was
obtained using FLURZnrc. These codes require the
geometry to have cylindrical symmetry so a simplifi-
cation of the actual holder geometry was necessary.
Most of the simulations were carried out using a single
chip located at the centre of a cylindrical PMMA phan-
tom. However, some tests were also done with a ring
of TLD material at the same position as the individual
chips in the actual holder.
The material data sets for EGSnrc are prepared by a
pre-processor called PEGS4. The default photon cross
sections available to PEGS4 are those of Storm and
Israel (18). Revised photoelectric cross sections have been
compiled by Berger et al (19), and they provide a pro-
gram called XCOM to extract the cross section data.
The XCOM cross sections differ by up to 5% from those
of Storm and Israel for photoelectric interactions below
200 keV. Hobeila and Seuntjens (20) have implemented
the XCOM photoelectric cross sections in EGSnrc and
they provided us with the data file to be used with
PEGS4. We have carried out calculations with both data
sets to determine the sensitivity of our results to the
photoelectric cross sections. The PEGS4 data sets were
generated for photon and electron kinetic energies in the
range from 1 to 2000 keV.
The simulations using DOSRZnrc report the absorbed
dose per unit fluence at the front face of the holder, so
a separate set of simulations was needed to relate the
fluence to the air kerma. The air kerma per unit fluence
was calculated using the technique described by Borg
et al (21). The absorbed dose to a 2 ␮m thick air slab with
a 10 cm radius was calculated using DOSRZnrc for each
of the incident photon beams. The cut-off energy for
electron transport was set very high (2000 keV) so that
electrons deposited their energy locally. With no elec-
tron transport, the absorbed dose is equivalent to the
kerma. This approach to determining the air kerma has
the advantage that a consistent set of photon cross sec-
tions are used for both the kerma and dose calculations.
For the air kerma calculations, the photon cut-off was
set to 1 keV. For the calculations of the absorbed dose
to the TLD chip, the photon and electron energy cut-
offs were 1 keV and 10 keV, respectively. This meant
that, for any interaction that resulted in a photon below
1 keV or an electron with kinetic energy below 10 keV,
its energy was deposited locally. All of the improved
electron transport options in EGSnrc were turned on for
the simulations, as well as all the low energy photon
and electron options. The same transport options were
used with FLURZnrc to calculate the electron spectrum
except that the electron cut-off was lowered from
10 keV to 1 keV. The energy width of the scoring bins
for the electron spectra was modified depending on the
incident photon energy, and the number of bins varied
from 30 to 80.
36
Seelentag et al (22) from the Gesellschaft für Strahlen-
und Umweltforschung (GSF) have reported extensive
measurements of X ray spectra. Tabulated data from
Seelentag et al for the photon spectra of the ISO series
of narrow-spectrum X ray beams have been used here
as source inputs for the simulations. Table 1 gives the
HVL and mean photon energy for the GSF beams. The
137
Cs and 60Co spectra were taken from the EGSnrc
distribution (23).
Most of the calculations were carried out on a
733 MHz computer. The air kerma calculations required
about 30 min to achieve an uncertainty of 0.01% using
1 ⫻ 108 histories. The calculation of the dose to the chip
required from 5 ⫻ 108 to 2.2 ⫻ 109 histories, depending
on the photon energy and phantom thickness, to achieve
an uncertainty of less than 0.1%. The computing time
Downloaded from rpd.oxfordjournals.org by guest on February 1, 2011
ranged from 3 to 70 h. The FLURZnrc calculations used
1 ⫻ 107 histories and required from 0.5 to 7 h of com-
puting time.
RESULTS
The air kerma response is defined here as the ratio of
the measured TLD reading to the value of the air kerma
for the field in which the chip was irradiated. Similarly,
the dose response is defined as the ratio of the measured
TLD reading to the absorbed dose delivered to the chip.
The absorbed dose to the chip is denoted by DTLD and
is approximately equivalent to the absorbed dose to pure
LiF, although the Monte Carlo model includes the
dopants.
Figure 4 shows typical glow curves for TLD-100 and
TLD-100H. The TLD-100 chip was irradiated in an air
700
600
TLD-100H
500
Reader output (nA)
400
300
200 TLD-100
TLD-100H
100
0
10 100 200 300 400 500
Time (ds)
Figure 4. Typical glow curves obtained for TLD-100 and TLD-
100H. The air kerma values were 20 mGy and 1.5 mGy for
TLD-100 and TLD-100H, respectively. The TLD-100 glow
curve has been scaled using a factor of 2.57 to have the same
area as the TLD-100H curve. The TLD-100H material is
approximately 34 times more sensitive than TLD-100.
 ENERGY RESPONSE OF LiF TLDs
kerma field approximately 13 times greater than that
used for the TLD-100H chip. In addition, the TLD-100
glow curve was multiplied by a factor of about 2.6 so
that the areas under the curves in Figure 4 are the same.
Thus, for our TLD protocol, we find that TLD-100H is
approximately 34 times more sensitive than TLD-100.
The TLD reading for each chip was obtained by inte-
grating between channels 42 and 430, where one chan-
nel corresponds to 0.1 s.
The background signal was obtained by comparing
the reading of the empty planchet with that of a chip
that had recently been annealed. It was found that the
background with a chip present was typically about 10%
greater than the empty planchet reading. Thus, a back-
ground reading of 1.1 times the empty planchet value
was subtracted from each TLD reading and amounted
would give the same reader output for both the test
beam and the 60Co reference beam were compared. To
do so, each data set for the test beam was fitted to a
straight line, as illustrated in Figure 5. The result is an
equation of the form
R = (sKa + i)Ka, (1)
where R is the reader output, Ka is the air kerma and s
and i are the slope and intercept, respectively. Equation
1 was used to find the value of Ka for the test beam that
would give the same reader output, R, obtained using
the 60Co reference beam. The relative air kerma
response was then obtained as the ratio of the air kerma
delivered by the Co irradiation to that for the test beam.
The maximum effect of ignoring the non-linearity is
about 0.6% and occurs for the N-150 beam.

Downloaded from rpd.oxfordjournals.org by guest on February 1, 2011

to a correction of from 0.1% to 2.3%, depending on the
dose level and TLD type.
Figure 5 shows typical results obtained using a set of
25 TLDs. One group of five chips was irradiated in a
60
Co reference beam to establish the overall set sensi-
tivity. The remaining chips were divided into groups of
five, and irradiated in a test beam at a point in space
where the air kerma was known. Each group was
irradiated for a different time so that a range of air
kerma values was covered. Figure 5 suggests that the
response versus air kerma is slightly non-linear, and we
have observed a similar trend with all the other data
sets. We suspect that this non-linearity is an artifact of
the reader, but have not demonstrated it conclusively.
In order that the non-linearity, regardless of its ori-
The measured air kerma response is tabulated for
each beam quality in Table 2 and shown graphically as
a function of beam energy in Figure 6. The response
for each energy has been normalised to the measured
response for 60Co gamma rays. In order to determine
the TLD dose response as a function of energy it is
necessary to relate the air kerma to the absorbed dose
delivered to the chip. We first consider a simple model
in which a LiF chip is embedded in a uniform phantom
of LiF, thick enough to provide charged particle equilib-
rium at the position of the chip. The incident photon
beam is taken to be parallel and monoenergetic and we
assume that photon attenuation and scatter can be neg-
lected. Then the absorbed dose to the TLD, DTLD, is
given by

冉 冊
gins, does not bias our results the air kerma values that
␮en
DTLD = ⌿ , (2)
1.02 ␳ LiF

where ⌿ is the photon energy fluence and (␮en/␳)LiF is

1.00
Relative air kerma response

Table 2. Measured air kerma response for TLD-100 and

TLD-100H for several beam qualities. The response is
0.98 reported relative to the measured response to 60Co gamma
rays. The Type A standard uncertainty in the last digit is
given in parentheses.
0.96
Identifier Relative air kerma response

TLD-100 TLD-100H
0.94

N-30 1.445(7) 0.930(7)

0.5 1.0 1.5 2.0
N-40 1.464(8) 1.004(6)
Air kerma (mGy)
N-60 1.387(6) 0.971(6)
Figure 5. Typical results obtained using a set of 25 high-sensi- N-80 1.272(6) 0.846(5)
tivity TLDs. Five chips (crosses) out of the set of 25 were N-100 1.211(7) 0.778(4)
irradiated with 60Co ␥-rays to establish the overall sensitivity N-150 1.163(6) 0.775(5)
of the set. The remaining four groups (circles) were irradiated N-200 – 0.825(6)
using N-60 X rays. The irradiation time for each group was N-250 1.110(6) –
137
adjusted so that the air kerma varied from about 0.5 to 2 mGy. Cs 1.023(5) 0.972(6)
60
The results are normalised to the mean 60Co response. The Co 1.000 1.000
dashed line is a linear fit to the N-60 data.

37
 S. D. DAVIS, C. K. ROSS, P. N. MOBIT, L. VAN DER ZWAN, W. J. CHASE and K. R. SHORTT
the mass energy absorption coefficient for LiF. For the Because g is never bigger than 0.003 in this energy
same energy fluence, ⌿, the air kerma, Ka, is given by range, it was ignored in the calculations.

冉 冊
Detailed Monte Carlo simulations to estimate the
␮en 1
Ka = ⌿ , (3) absorbed dose per unit air kerma have also been carried
␳ air (1 ⫺ g) out. Because the Monte Carlo user code requires cylin-
where (␮en/␳)air is the mass energy absorption coef- drical symmetry, the holder geometry was approximated
ficient for air and g is the fraction of the incident energy as a single TLD at the centre of a cylindrical PMMA
that is lost to radiative processes. Data for the mass phantom. In order to test the sensitivity of the model to
energy absorption coefficients were obtained from Hub- the phantom radius, DTLD/Ka as a function of the radius
bell and Seltzer (24), and DTLD/Ka is plotted in Figure 7. was calculated and the results are shown in Figure 8.
Because of scattered photons, the response increases as
the holder radius increases. The effect becomes larger
TLD-100 as the photon energy decreases, reaching a maximum at
1.4 TLD-100H about 60 kV. These results provide no clear-cut choice
Relative air kerma response

for a holder radius to approximate the true geometry.

However, the radius cannot be greater than 16 mm,

Downloaded from rpd.oxfordjournals.org by guest on February 1, 2011

1.2
which is the radius of the holder used for the measure-
ments, and it should not be less than 8 mm, which is
the minimum distance from a TLD chip to the edge of
the holder. The maximum change in response for radii
1.0 from 8 to 16 mm is about 1.7%.
As a second approximation, a holder with the same
radius as the actual holder (16 mm) but with a ring of
0.8 TLD material at 9 mm from the centre was considered.
This model maintains the cylindrical symmetry required
10 100 1000
by the EGSnrc user codes and the holder radius is cor-
Mean photon energy (keV)
rect. However, the individual TLDs are distorted into a
uniform ring of TLD material. For all the photon beams,
Figure 6. Measured air kerma response of TLD-100 and TLD- the results from the ring model corresponded well with
100H as a function of photon energy. Each datum point is those for the chip-centred model with a holder radius of
based on at least one set of measurements similar to that shown 13 mm (Figure 8). All subsequent calculations are based
in Figure 5. The Type A standard uncertainty on each point is on the chip-centred model with a radius of 13 mm.
typically 0.6%.
1.30 1.025

Ring geometry
1.25 Simple model 1.020
MC, TLD-100
1.20 MC, TLD-100H
Relative response

1.015
1.15
DTLD /Ka

1.10 1.010

1.05
1.005
1.00 60
Co
0.95 1.000 N-150
N-60
0.90
0.995
10 100 1000 6 8 10 12 14 16 18 20
Energy (keV) Holder radius (mm)

Figure 7. Calculated air kerma response as a function of photon
energy. The pluses were obtained by considering a LiF TLD
to be embedded in a uniform phantom of LiF. The effects of
photon attenuation and scattering within the phantom were neg-
lected. The results of the EGSnrc Monte Carlo calculations are
shown by the circles and crosses for TLD-100 and TLD-100H,
respectively. The statistical uncertainty on each point calcu-
lated using the Monte Carlo technique is about 0.1%.
Figure 8. Effect of the holder radius on the calculated value of
DTLD/Ka, for a single TLD chip mounted at the centre of the
holder. The results are given relative to the value for a phantom
with a radius of 8 mm. The results obtained for a ring approxi-
mation to the distribution of TLD material are also shown. The
ring results are plotted to show that a single chip at the centre
of a holder with a radius of 13 mm gives similar results for all
beam qualities.

38
 ENERGY RESPONSE OF LiF TLDs
The calculations of DTLD/Ka were repeated for both
sets of photoelectric cross sections. Although differ-
ences of up to 3% were observed in the values of DTLD
and Ka per unit incident fluence, the maximum effect
on the ratio, DTLD/Ka, was 0.6%. As expected, the ratio
is rather insensitive to the details of the photoelectric
cross section. The Monte Carlo calculations also depend
on knowledge of the X ray spectra. Because spectral
measurements for the NRC beams were not available
the spectra measured at the GSF (22) were used. Spectral
differences will be most important for the low energy
beams because the photon cross sections are a strong
function of energy below about 80 keV. In order to test
the sensitivity of DTLD/Ka to the X ray spectrum, the N-
30 result was re-calculated using the spectrum measured
at the Physikalish-Technische Bundesanstalt (PTB) (25).
the chip will not affect the measured energy depen-
dence. However, the dose due to low energy X rays will
show more variation over the thickness of the chip than
60
Co gamma rays. In order to estimate the effect of light
absorption on our measured response we used the mass
energy absorption coefficients for LiF to estimate the
dose distribution within the chip for N-30 X rays and
60
Co gamma rays. By combining the light attenuation
and dose distribution data, it is estimated that the correc-
tion to be applied to the N-30 response relative to that
for 60Co is only about 0.2% and will be even smaller
for higher energy X ray beams. Because of the small
size of the correction, and in the absence of data specific
to TLD-100 and TLD-100H, our results have not been
corrected for light attenuation within the chip. Olko
et al (27) have carried out a similar calculation for MCP-

Downloaded from rpd.oxfordjournals.org by guest on February 1, 2011

The GSF and PTB spectra give values of 1.12 and 1.16
for the HVL (mm of Al), respectively. The difference
in DTLD/Ka for the two spectra is 0.4%, suggesting that
our results are not unduly sensitive to knowledge of the
X ray spectra.
The measured TLD signal arises from light that
escapes from within the chip. Because the TLD material
is not perfectly transparent, the portion of the TLD clos-
est to the photomultiplier will contribute more to the
signal than portions at greater depth. Ipe et al (26) have
measured the light attenuation in the Polish formulations
of LiF:Mg,Ti and LiF:Mg,Cu,P which are denoted by
MTS-N and MCP-N, respectively. They find that the
light attenuation is similar in the two materials and we
assume that the same would apply for TLD-100 and
TLD-100H. As long as the dose distribution within the
chip does not depend on energy, light absorption within
N and show the results graphically for a wide range of
photon energies.
The EGSnrc results for DTLD/Ka are tabulated in
Table 3 and shown graphically in Figure 7. By combin-
ing the measured air kerma response from Figure 6 with
the Monte Carlo results for DTLD/Ka from Figure 7, the
TLD dose response as a function of energy can be
obtained. The results are tabulated in Table 3 and shown
graphically in Figure 9 for both TLD-100 and TLD-
100H. These data represent the TLD reading per unit
absorbed dose to the chip, normalised to the value for
60
Co gamma rays.
DISCUSSION
Both TLD-100 and TLD-100H are thermoluminesc-
ent materials based on LiF. However, Figure 9 shows

Table 3. The results calculated using EGSnrc for the dose

to the chip per unit air kerma are given in columns 2 and 1.1
3 for TLD-100 and TLD-100H, respectively. The values of
DTLD/Ka have a statistical uncertainty of about 0.1%. The
measured values of the air kerma response from Table 2 1.0
Relative dose response

were combined with the calculated values of DTLD/Ka to

obtain the dose response of the TLD chip and the results
are given in columns 4 and 5. The Type A standard uncer- 0.9
tainty in the last digit is given in parentheses.

Identifier DTLD/Ka Relative dose response 0.8

TLD-100 TLD-100H TLD-100 TLD-100H

0.7 TLD-100
TLD-100H
N-30 1.208 1.242 1.100(5) 0.689(5)
N-40 1.253 1.285 1.075(6) 0.718(5) 0.6
N-60 1.211 1.238 1.053(5) 0.722(5) 10 100 1000
N-80 1.117 1.133 1.047(5) 0.687(4)
Mean photon energy (keV)
N-100 1.042 1.051 1.069(6) 0.681(4)
N-150 0.983 0.986 1.089(6) 0.723(5) Figure 9. Results from the present work for the dose response
N-200 0.958 0.959 – 0.790(6) as a function of energy for TLD-100 and TLD-100H. Each
N-250 0.951 0.952 1.073(6) – datum point was obtained by dividing the measured air kerma
137
Cs 0.924 0.924 1.018(6) 0.968(6) response from Figure 6 with the value of DTLD/Ka calculated
60
Co 0.920 0.919 1.000 1.000 using the EGSnrc Monte Carlo code. The Type A standard
uncertainty on each point is typically 0.6%.

39
 S. D. DAVIS, C. K. ROSS, P. N. MOBIT, L. VAN DER ZWAN, W. J. CHASE and K. R. SHORTT
that the dose response as a function of energy of the
two materials is very different. The dose response of
TLD-100H decreases dramatically as the photon energy
decreases while that of TLD-100 shows a slight increase
in the same energy region. Both response curves exhibit
additional structure as the photon energy decreases
below about 100 keV. In earlier work (10) we had shown
that the dose response of TLD-100 was greater for 137Cs
than for 60Co. The opposite effect for TLD-100H is
found, that is, the dose response is less for 137Cs than
for 60Co. Sáez-Vergara et al (9) already noted that the
change in response between 137Cs and 60Co was differ-
ent for LiF:Mg,Ti and LiF:Mg,Cu,P. However, their
data does not clearly show the 3.3 (⫾0.5)% change that
we measure for TLD-100H.
Several measurements of the energy response of
LiF:Mg,Ti have been reported in the literature. Budd
et al (28) reported measurements in the energy range from
30 to 148 keV and compared their results to earlier
work. Although there is considerable scatter within the
data, the general trend is consistent. That is, the meas-
ured response to low energy X rays relative to 60Co
gamma rays is greater than expected based on the
energy deposited in the material. One of the most exten-
sive sets of measurements were carried out by Tochilin
et al (11) and the smooth curve they drew through their
data is reproduced in Figure 10. The typical uncertainty
on each of their datum points was about 5% so they
were not able to detect the structure we observe between
20 and 100 keV.
Olko et al (7) have measured the dose response as a
function of energy for the formulation of LiF:Mg,Cu,P
identified as MCP-N and their results are shown in
1.2
Olko et al.(29) (prediction)
1.1
Relative dose response
NRC
1.0
Tochilin et al.(11)
0.9
Olko et al.(7)
0.8
NRC
0.7
0.6
10 100
Mean photon energy (keV)
Figure 10. Comparison of the dose response as a function of
energy for TLD-100 and TLD-100H with the results obtained
by Tochilin et al (11) for LiF:Mg,Ti and the results of Olko
et al (7) for the formulation of LiF:Mg,Cu,P identified as MCP-
N. The dotted curve near the top of the figure is a model predic-
tion developed by Olko et al (29) for LiF:Mg,Ti.
Figure 10. There is fair agreement between our results
for TLD-100H and those for MCP-N, indicating that
these two formulations of high sensitivity LiF have
similar characteristics.
Olko et al (7) have developed a microdosimetric model
to help explain the dose response curve of LiF:Mg,Cu,P.
They point out that the structure below 100 keV arises
from the change in the electron spectrum because the
photoelectric effect becomes more dominant as the pho-
ton energy decreases. In order to demonstrate this point,
EGSnrc was used to calculate the mean electron energy
for all electrons set in motion with energy greater than
1 keV. The mean electron energy as a function of the
photon energy is shown in Figure 11. Starting from
60
Co, the mean electron energy decreases as the photon
energy decreases. However, near 100 keV, the mean
Downloaded from rpd.oxfordjournals.org by guest on February 1, 2011
energy no longer continues to decrease, and it increases
slightly until about 60 keV. Below 60 keV the mean
electron energy again begins to decrease as the photon
energy decreases. This suggests that the structure in the
dose response curves for both TLD-100 and TLD-100H
is roughly correlated with the change in the mean elec-
tron energy with photon energy.
Olko et al (7) have calculated the mean lineal energy,
yF, for a target diameter of 40 nm for each of the photon
beams. They show, in Figure 4 of their paper, that the
measured dose response of LiF:Mg,Cu,P is well corre-
lated with yF. EGSnrc does not permit the calculation
of microdosimetric quantities, but it can be used to cal-
culate the restricted electron stopping power, (L/␳)TLD,
which is also an indicator of local energy deposition.
The electron stopping power for each of the photon
15
Stopping power (MeV cm2 g–1)
Mean electron energy (keV)
100 10
5
10 0
10 100 1000
1000 Mean photon energy (keV)
Figure 11. Results obtained using EGSnrc for the mean elec-
tron energy (diamonds) and the mean restricted collisional stop-
ping power (circles) as a function of the mean photon energy.
The energy cut-off used for calculating the restricted stopping
power was 1 keV. The results are the mean values for TLD-
100 and TLD-100H, although the differences between the two
materials are generally less than 1%.
40
 ENERGY RESPONSE OF LiF TLDs
beams has been calculated, using an electron energy cut- CONCLUSIONS
off of 1 keV. Because the values for TLD-100 and TLD-
A high precision comparison of the dose response
100H do not differ by more than 1%, the two have been
versus energy of TLD-100 and TLD-100H has been
averaged. The results are tabulated in Table 4 and
carried out. Both these LiF-based dosemeters are widely
shown graphically in Figure 11. It can be seen that the
used and are available from the same supplier. By carry-
restricted stopping power shows a structure below
ing out Monte Carlo calculations of the absorbed dose
100 keV that is reminiscent of the structure of the dose
to the chip per unit incident fluence, it was possible to
response curves. In order to test the correlation between
eliminate the effects of the holder on the measured
the dose response and the restricted stopping power, the
response. Our results are in general agreement with
dose response versus the stopping power has been plot-
other work, indicating that the specific formulation of
ted in Figure 12. Although the stopping power is a
the TLD material does not play a major role in affecting
rough predictor of the dose response, it does not perform
its energy response.
as well as yF.
The dose response of TLD-100H shows a large vari-
In a later paper, Olko et al (29) studied the energy
ation with energy and must be accounted for if the TLDs
response of LiF:Mg,Ti. The formulation they used is
are to be calibrated at one beam quality but used at
identified as MTS-N and was produced at their institute
another. Even between 60Co and 137Cs we see a change

Downloaded from rpd.oxfordjournals.org by guest on February 1, 2011

(INP). Their experimental results show considerable
of 3.3%. This is similar in size, but in the opposite direc-
scatter but are consistent with our data. They also
tion, to the effect reported earlier by Shortt et al (10) for
developed a microdosimetric model to help explain their
TLD-100. For TLD-100, the change in response with
results. Their model depends on a target diameter which
energy is considerably smaller than for TLD-100H.
they take as a free parameter and varies from 8 to
In addition to their importance for radiation protec-
40 nm. Their model results for a target diameter of
tion dosimetry both TLD-100 and TLD-100H dosemet-
40 nm have been reproduced in Figure 10. Their model
ers are used for brachytherapy dosimetry. For this appli-
shows a structure similar to that measured here,
cation, high precision is required but there remains some
although the fit is not perfect. Furthermore, the model
uncertainty regarding the energy dependence of the
indicates that the response continues to increase as the
dosemeters (30–32). The mean photon energies of the most
photon energy decreases. This prediction is consistent
commonly used brachytherapy seeds (33) range from
with our measurements but in disagreement with the
22 keV to 380 keV. Our measurements show that the
work of Tochilin et al. However, for target diameters
variation of the dose response with energy will lead to
smaller than 20 nm, the calculated response tends to
corrections of up to 10% for TLD-100 and up to 47%
decrease below about 25 keV. Additional measurements
for TLD-100H in this energy range, assuming the dose-
below N-30 are needed to determine how the response
meters have been calibrated using 60Co gamma rays.
changes in this low energy region.

ACKNOWLEDGEMENTS
Table 4. Quantities calculated using the EGSnrc Monte One of the authors (S. D. Davis) is grateful for the fin-
Carlo code. The meaning of each symbol is as follows: E␥, ancial support provided by Ontario Power Generation.
mean photon energy of the incident beam; Ee, mean elec-
tron energy; (L/␳)TLD, mean restricted collisional stopping
power. Ee includes all electrons with an energy greater than 1.1
the cut-off of 1 keV and is the mean of the values for TLD-
100 and TLD-100H. The stopping power was calculated
with a cut-off of 1 keV and is the mean of the values for 1.0
Relative dose response

TLD-100 and TLD-100H.

0.9
Identifier E␥ (keV) Ee (keV) (L/␳)TLD
(MeV cm2 g⫺1)
0.8
N-30 24.2 14.4 16.3
N-40 32.5 19.3 13.4
N-60 47.3 26.9 11.7 0.7 TLD-100
N-80 64.5 33.0 12.1 TLD-100H
N-100 82.6 33.7 13.3
N-150 117 31.1 12.5 0.6
N-200 164 36.2 9.23 0 5 10 15
N-250 207 47.9 7.09 2 –1
137 Stopping power (MeV cm g )
Cs 613 206 2.55
60
Co 1055 421 1.78 Figure 12. The dose response of TLD-100 and TLD-100H as
a function of the mean restricted collisional stopping power.

41
 S. D. DAVIS, C. K. ROSS, P. N. MOBIT, L. VAN DER ZWAN, W. J. CHASE and K. R. SHORTT
Another (P. N. Mobit) would like to acknowledge the sup- chington in constructing TLD holders. Stewart Walker,
port provided by the Alberta Heritage Foundation for Feridoun Faravash and Hong Shen helped with electronic
Medical Research. The support of the technical staff of the issues, computer systems and X ray irradiations, respect-
Ionizing Radiation Standards Group was critical in leading ively. They are grateful to Dr Jan Seuntjens and Fadi Hob-
to the timely completion of this work. The authors would eila of McGill University for providing them with revised
like to acknowledge, in particular, the work of David Mar- cross section data for low energy X rays.

REFERENCES
1. Attix, F. H. Introduction to Radiological Physics and Radiation Dosimetry (New York: Wiley) (1986).
2. Horowitz, Y. S. (Ed.) Thermoluminescence and Thermoluminescent Dosimetry, Volumes I, II and III. (Boca Raton, Florida:
CRC Press) (1984).
3. McKeever, S. W. S., Moscovitch, M. and Townsend, P. D. Thermoluminescence Dosimetry Materials: Properties and Uses
(Ashford, England: Nuclear Technology Publishing) (1995).
4. Pradhan, A. S. Photon energy response of luminescence dosemeters and its impact on assessment of Hp(10) and Hp(0.07)
in mixed fields of varying energies of photons and beta radiation. Radiat. Prot. Dosim. 101, 173–178 (2002).

Downloaded from rpd.oxfordjournals.org by guest on February 1, 2011

5. Perry, O. R., Moscovitch, M., Velbeck, K. J. and Rotunda, J. E. LiF:Mg,Cu,P based environmental dosemeter and dose
calculation algorithm. Radiat. Prot. Dosim. 85, 273–281 (1999).
6. Bilski, P., Olko, P., Burgkhardt, B., Piesch, E. and Waligórski, M. P. R. Thermoluminescence efficiency of LiF:Mg,Cu,P
(MCP-N) detectors to photons, beta electrons, alpha particles and thermal neutrons. Radiat. Prot. Dosim. 55, 31–38 (1994).
7. Olko, P., Bilski, P., Ryba, E. and Niewiadomski, T. Microdosimetric interpretation of the anomalous photon energy response
of ultra-sensitive LiF:Mg,Cu,P TL dosemeters. Radiat. Prot. Dosim. 47, 31–35 (1993).
8. Olko, P. Microdosimetric interpretation of thermoluminescence efficiency of LiF:Mg,Cu,P (MCP-N) detectors for weakly
and densely ionising radiations. Radiat. Prot. Dosim. 65, 151–158 (1996).
9. Sáez-Vergara, J. C., Romero, A. M., Ginjaume, M., Ortega, X. and Miralles, H. Photon energy response matrix for environ-
mental monitoring systems based on LiF:Mg,Ti and hypersensitive phosphors (LiF:Mg,Cu,P and ␣-Al2O3:C). Radiat. Prot.
Dosim. 85, 207–211 (1999).
10. Shortt, K. R., Ross, C. K. and Janovský, I. The Response of LiF TLDs to 137Cs and 60Co ␥ rays. Radiat. Prot. Dosim. 69,
257–266 (1997).
11. Tochilin, E., Goldstein, N. and Lyman, J. T. The quality and LET dependence of three thermoluminescent dosimeters and
their potential use as secondary standards. In: Proc. Second Int. Conf. on Luminescence Dosimetry (Washington, DC: US
Atomic Energy Commission) (1968).
12. Janovský, I. and Ross, C. K. The IRS thermoluminescent dosimetry system. NRC Report PIRS-0369 (Ottawa, Canada: National
Research Council) (1993).
13. Shortt, K. R. and Ross, C. K. The Canadian 60Co exposure standard. NRC Report PIRS-0052 (Ottawa: National Research
Council) (1986).
14. Shortt, K. R., Bielajew, A. F., Rogers, D. W. O. and Ross, C. K. Calculation of the ratio of Nx for cesium and cobalt. NRC
Report PIRS-0142 (Ottawa: National Research Council) (1988).
15. Henry, W. H. and Garrett, C. The Canadian standard free-air chamber for medium quality X-rays. Can. J. Phys. 38, 1677–
1689 (1960).
16. ISO. X and gamma reference radiation for calibrating dosemeters and doserate meters and for determining their response as
a function of photon energy — Part 1: Radiation characteristics and production methods. ISO Report 4037–1:1996 (Geneva:
International Organization for Standardization) (1996).
17. Kawrakow, I. Accurate condensed history Monte Carlo simulation of electron transport. I. EGSnrc, the new EGS4 version.
Med. Phys. 27, 485–498 (2000).
18. Storm, E. and Israel, H. I. Photon cross sections from 1 keV to 100 MeV for elements Z = 1 to Z = 100. Atomic Data and
Nuclear Data Tables 7, 565–681 (1970).
19. Berger, M. J., Hubbell, J. H., Seltzer, S. M., Coursey, J. S. and Zucker, D. S. XCOM: photon cross section database (Version
1.2). http://physics.nist.gov/xcom (Gaithersburg, MD: National Institute of Standards and Technology) (1999).
20. Hobeila, F. and Seuntjens, J. Implementation of NIST XCOM-based photoelectric cross-sections in EGSnrc (Abstract). Med.
Phys. 29, 1353 (2002).
21. Borg, J., Kawrakow, I., Rogers, D. W. O. and Seuntjens, J. P. Monte Carlo study of correction factors for Spencer–Attix
cavity theory at photon energies at or above 100 keV. Med. Phys. 27, 1804–1813 (2000).
22. Seelentag, W. W., Panzer, W., Drexler, G., Platz, L. and Santner, F. A catalogue of spectra used for the calibration of
dosemeters. GSF Report 560 (München: Gesellschaft für Strahlen und Umweltforschung) (1979).
23. Rogers, D. W. O. and Kawrakow, I. Distribution area for EGSnrc: a Unix/Linux-based system for Monte Carlo simulation
of electron and photon transport. www.irs.inms.nrc.ca/inms/irs/irs.html (2003).

42
 ENERGY RESPONSE OF LiF TLDs
24. Hubbell, J. H. and Seltzer, S. M. Tables of X-ray mass attenuation coefficients and mass energy-absorption coefficients
(Version 1.03, Online). http://physics.nist.gov/xaamdi (Gaithersburg, MD: National Institute of Standards and Technology)
(1997).
25. Ankerhold, U. Catalogue of X-ray spectra and their characteristic data — ISO and DIN radiation qualities, therapy and
diagnostic radiation qualities, unfiltered X-ray spectra. PTB Report: PTB-Dos-34 (Braunschweig: Physikalisch-Technische
Bundesanstalt) (2000).
26. Ipe, N. E., Fassò, A., Kase, K. R., Kaur, R., Bilski, P. and Olko, P. Characterisation of the low energy X-ray response
of Polish TLDs to synchrotron radiation and the determination of some TLD quantities. Radiat. Prot. Dosim. 84, 169–
173 (1999).
27. Olko, P., Bilski, P., Budzanowski, M., Waligórski, M. P. R., Fasso, A. and Ipe, N. Modelling of the thermoluminescence
response of LiF:Mg,Cu,P (MCP-N) detectors after doses of low energy photons. Radiat. Prot. Dosim. 84, 103–108 (1999).
28. Budd, T., Marshall, M., Peaple, H. J. and Douglas, J. A. The low- and high-temperature response of lithium fluoride dosemet-
ers to X-rays. Phys. Med. Biol. 24, 71–80 (1979).
29. Olko, P., Bilski, P. and Kim, J.-L. Microdosimetric interpretation of the photon energy response of LiF:Mg,Ti detectors.
Radiat. Prot. Dosim. 100, 119–122 (2002).
30. Das, R. K., Li, Z., Perera, H. and Williamson, J. F. Accuracy of Monte Carlo photon transport simulation in characterizing
brachytherapy dosimeter energy-response artifacts. Phys. Med. Biol. 41, 995–1006 (1996).

Downloaded from rpd.oxfordjournals.org by guest on February 1, 2011

31. Pradhan, A. S. and Quast, U. In-phantom response of LiF TLD-100 for dosimetry of 192Ir HDR source. Med. Phys. 27,
1025–1029 (2000).
32. Hood, C., Duggan, L., Bazley, S., Denham, J., Budzanowski, M. and Kron, T. LiF:Mg,Cu,P ‘pin worms’: miniature detectors
for brachytherapy dosimetry. Radiat. Prot. Dosim. 101, 407–410 (2002).
33. Coursey, B. M., Goodman, L. J., Hoppes, D. D., Loevinger, R., McLaughlin, W. L. and Soares, C. G. The needs for
brachytherapy source calibrations in the United States. Nucl. Instrum. Methods. Phys. Res. A. 312, 246–250 (1992).

43
Downloaded from rpd.oxfordjournals.org by guest on February 1, 2011

44