Content uploaded by Klemens Zink
Author content
All content in this area was uploaded by Klemens Zink on Jun 06, 2019
Content may be subject to copyright.
A TLD-based ten channel system for the spectrometry
of bremsstrahlung generated by laser-matter interaction
Felix Horst
a,b,
n
, Georg Fehrenbacher
a
, Torsten Radon
a
, Ekaterina Kozlova
a
, Olga Rosmej
a
,
Damian Czarnecki
b
, Oliver Schrenk
b
, Joachim Breckow
b
, Klemens Zink
b,c
a
GSI Helmholtzzentrum für Schwerionenforschung GmbH, Planckstr. 1, 64291 Darmstadt, Germany
b
Institut für Medizinische Physik und Strahlenschutz, Technische Hochschule Mittelhessen, Wiesenstr. 14, 35390 Gießen, Germany
c
Universitätsklinikum Gießen und Marburg GmbH, Baldingerstr., 35043 Marburg, Germany
article info
Article history:
Received 11 August 2014
Received in revised form
9 January 2015
Accepted 3 February 2015
Available online 12 February 2015
Keywords:
Laser acceleration
X-ray spectrometry
Thermoluminescence dosimetry
Pulsed radiation
abstract
This work presents a thermoluminescence dosimetry based method for the measurement of brems-
strahlung spectra in the energy range from 30 keV to 100 MeV, resolved in ten different energy intervals
and for the photon ambient dosimetry in ultrashort pulsed radiation fields as e.g. generated during
operation of the PHELIX laser at the GSI Helmholtzzentrum für Schwerionenforschung. The method is a
routine-oriented development by application of a multi-filter technique. The data analysis takes around
1 h. The spectral information is obtained by the unfolding of the response of ten thermoluminescence
dosimeters with absorbers of different materials and thicknesses arranged as a stack each with a
different response function to photon radiation. These response functions were simulated by the use of
the Monte Carlo code FLUKA. An algorithm was developed to unfold bremsstrahlung spectra from the
readings of the ten dosimeters. The method has been validated by measurements at a clinical electron
linear accelerator (6 MV and 18 MV bremsstrahlung). First measurements at the PHELIX laser system
were carried out in December 2013 and January 2014. Spectra with photon energies up to 10 MeV and
mean energies up to 420 keV were observed at laser-intensities around 10
19
W=cm
2
on a titanium foil
target. The measurement results imply that the steel walls of the target chamber might be an additional
bright x-ray source.
&2015 Elsevier B.V. All rights reserved.
1. Introduction
Today high power laser systems may reach peak intensities up to
10
22
W=cm
2
onthetargetatsub-picosecond pulse durations. The
resulting relativistic laser-plasma interaction leads to the directed
acceleration of electrons and ions up to GeV energies [1,2]. Especially
the accelerated electrons generate bremsstrahlung in the target and
in the surrounding materials, e.g. the steel walls of the target
chamber. These effects lead to ultrashort x-ray pulses in the same
time scale as the laser pulse during the shots of high power lasers.
These x-ray pulses determine the need for radiation protection
during the operation of such laser systems. As a result of the sub-
picosecond time scale of the laser pulses, an active measurement of
x-ray spectra under these conditions is quite difficult to imple-
ment. For that reason passive ionizing radiation detectors, such as
thermoluminescence dosimeters (TLDs) or image plates are usually
applied for the diagnostics in high intensity laser experiments.
This work presents a TLD-based method for the measurement of
bremsstrahlung spectra in ultrashort-pulsed radiation fields. Ten
TLD cards (Harshaw TLD-700H) are placed into a stack of absorbers,
made of various materials and thicknesses, surrounded by a
shielding. The response functions of the ten TLD's to parallel
monoenergetic photon and electron radiation were simulated by
the use of the Monte Carlo code FLUKA [3,4]. The photon response
was verified by irradiation of the prototype with radioactive sources
(
137
Cs and
60
Co). The different gradients and thresholds of these
response functions allow the reconstruction (unfolding) of photon
spectra from the readings of the ten TLD's. An algorithm for the
purpose of unfolding bremsstrahlung spectra in the range of 30 keV
to 100 MeV, resolved in 10 different energy bins, was developed
(written in SCILAB [5]). The method is a further development of a
work from Behrens et al. [6] and has been validated at a clinical
electron-linac (Elekta Synergy: 6 and 18 MV bremsstrahlung) by
comparing spectra measured by the developed method with
spectra obtained by detailed Monte Carlo simulations (Monte Carlo
code: EGSnrc [7]) of the linac. A prototype has been built and first
applied at PHELIX (Petawatt High Energy Laser for Heavy Ion
EXperiments) at the GSI Helmholtzzentrum für Schwerionen-
forschung in Darmstadt, Germany during beamtimes in December
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/nima
Nuclear Instruments and Methods in
Physics Research A
http://dx.doi.org/10.1016/j.nima.2015.02.010
0168-9002/&2015 Elsevier B.V. All rights reserved.
n
Corresponding author. Present address: Institut für Medizinische Physik und
Strahlenschutz, Technische Hochschule Mittelhessen, Wiesenstr. 14, 35390 Gießen,
Germany. Tel.: þ49 641 3092642.
E-mail address: felix.ernst.horst@kmub.thm.de (F. Horst).
Nuclear Instruments and Methods in Physics Research A 782 (2015) 69–76
2013 [8] and January 2014. This paper presents the method of
measurement, the principle of the unfolding-algorithm, the com-
parison of the results from the spectrometry measurements at the
electron linac with the reference spectra from Monte Carlo simula-
tions and the results from the measurements at PHELIX.
2. Methods and materials
Harshaw TLD-700 H: In this work Harshaw TLD-700 H cards
each with four LiF-crystals are used for the measurement of the
spectral distribution of the x-rays. The tolerance of this material
against high dose rates is experimentally verified for dose rates up
to 10
9
Gy=s¼1 mGy=ps [9].
The statistical uncertainty of the readings of these TLD cards
relative to the mean dose as a function of the mean dose has been
experimentally determined by measurements at a
137
Cs source with
varied durations of the irradiation of 25 TLD cards. This dependence
is shown in Fig. 1. It shows that doses of at least 10
μ
Gy are needed
for reliable readings (uncertainty of single LiF-crystal below 40%).
Principles of the measurement method: The presented spectro-
metry method is based on the idea, that the spectral information
of a photon radiation field can be obtained by some dosimeters
being specifically attenuated by absorbers of different materials
and thicknesses. The readings of those dosimeters are then
processed by a so called unfolding-algorithm which calculates
the photon spectrum to the dose values measured by the dosi-
meters approximately by the use of the simulated response
functions of the single dosimeters.
The dose which is measured by a dosimeter (channel) in a radi-
ation field with the spectral fluence Φ
E
ðEÞis given by
D
i
¼Z
1
0
Φ
E
ðEÞR
i
ðEÞdEð1Þ
with the dose Don dosimeter-channel i, the energy Eand the
energy dependent response R(E) of dosimeter-channel i.
The approximate dose values for 10 energy intervals with
different interval widths ΔEcan be calculated by
D
calc
i
X
10
j¼1
Φ
j
E
R
i;j
ΔE
j
ð2Þ
with the average response R
i;j
of channel iover the energy interval
ΔE
j
. In this work the approximation that R
i;j
is equal to R
i
for
monoenergetic radiation of the center E
j
of the energy interval ΔE
j
(average fluence
Φ
E
j
) is assumed.
The spectrum cannot be derived from the dose values by an
analytical calculation (inverse problem). This calculation has to be
performed iteratively by an unfolding-algorithm. The task of the
unfolding-algorithm is to find a spectral fluence configuration
Φ
E
j
according to E. (2) as precise as possible. The response matrix R
i;j
is
unknown, it can be obtained by e.g. Monte Carlo simulations.
The X
2
-value is a measure for the quality of the approximation
of the calculated dose values D
i
calc
to the measured dose values
D
i
meas
. The X
2
-value is calculated by
X
2
¼1
10 X
10
i¼1
ðD
meas
i
D
calc
i
Þ
2
ðΔD
meas
i
Þ
2
ð3Þ
The closer the X
2
-value is on 1, the better is the approximation.
The uncertainty ΔD
meas
i
of the readings D
i
meas
of the dosimeters
depends on the mean dose. The relative standard deviation can be
obtained from the fit function in Fig. 1. The statistical uncertainty
from the single LiF crystal is divided by ffiffiffi
n
p¼ffiffiffi
4
p(standard
deviation of the mean) when it is averaged over the four single
readings from the TLD cards (Harshaw TLD-700H).
TLD-spectrometer: The spectrometer prototype presented in
this work has a cylindrical shape with a default incident direction
of the radiation (see Fig. 2). It is equipped with 10 Harshaw TLD-
700H cards. The spectrometer is designed for an energy range
from 30 keV to 100 MeV, resolved in 10 different energy bins. The
widths of the ten energy bins are:
ΔE
j
¼f20 keV;50 keV;150 keV;250 keV;500 keV;
1;5 MeV;2;5 MeV;5 MeV;40 MeV;50 MeVg
with the interval centers:
E
j
¼f40 keV;75 keV;175 keV;375 keV;750 keV;
1;75 MeV;3;75 MeV;7;5 MeV;30 MeV;75 MeVg
The response of the spectrometer to monoenergetic photon
and electron radiation in the energy range 30 keV–100 MeV has
been simulated by use of the Monte Carlo code FLUKA [3,4]. The
transport thresholds within FLUKA were 10 keV for photons and
521 keV for electrons (including rest mass m
e
). The geometry and
material composition were modeled as precise as possible. This
Monte Carlo model is shown in Fig. 2, it has been drawn with the
FLUKA user interface Flair [10].
Fig. 3 shows the response of the spectrometer to photon radia-
tion. Fig. 4 shows the response to electron radiation. Both were
obtained by FLUKA simulations. The response functions obtained by
these simulations describe the response to parallel radiation. If the
radiation is emitted from a point source and the dimension of the
spectrometer is not negligible compared to the distance from the
source to the spectrometer's surface a correction of the TLD readings
is needed. The conversion from the dose in a divergent radiation field
D
i
div
to the dose in a parallel radiation field D
i
par
as needed for
unfolding can be done by application of the inverse-square law
D
par
i
¼D
div
i
x
1
þx
2
x
1
2
ð4Þ
with the distance x
1
from the source to the spectrometer's
surface and the distance x
2
from the spectrometer's surface to the
TLD channel i.
The fact that the TLD's respond not only to photon radiation but
also to electron radiation makes it necessary that the dose coming
from electron radiation has to be detected and subtracted by the
unfolding-algorithm if the measurement has been carried out in a
mixed radiation field.
The photon response was verified by irradiating the spectrometer
prototype with
γ
-radiation from radioactive sources (
137
Cs and
60
Co)
and comparing the readings from the ten TLD's with the predictions
from FLUKA simulations for the photon energies of the two used
radionuclides (shown in Fig. 5).Thedivergenceofthephotonsemitted
from the sources was taken into account for the simulations. The
FLUKA prediction agreed with the measurements for all the ten TLD's
and both nuclides within 720%(uncertainty of the source activities).
Fig. 6 shows the relative response distribution inside the
spectrometer (value of the single TLD normalized to the sum of
all ten TLD's).
Fig. 1. Relative standard deviation σ=Dof the readings from the single LiF-crystals
in the Harshaw TLD-700 H cards as a function of the mean applicated dose D.
F. Horst et al. / Nuclear Instruments and Methods in Physics Research A 782 (2015) 69–7670
Unfolding-algorithm: The spectrum of laser-generated brems-
strahlung in the high energy range can approximately be
described by a Boltzmann distribution [11]:
Φ
E
ðEÞ¼Φ
0
e
E=kT
ð5Þ
In this work, the Boltzmann distribution was modified to obtain
a more realistic spectrum with an additional build-up curve shape
in the low-energy range:
Φ
E
ðEÞ¼ðEXÞΦ
0
E
e
E=kT
Ph
;Φ
E
AR
þ
ð6Þ
For ten energy bins ΔE
j
with mean energies E
j
it follows
Φ
j
E
¼ðE
j
XÞΦ
0
E
e
E
j
=kT
Ph
;Φ
E
AR
þ
ð7Þ
In a first step the unfolding algorithm (shown in Fig. 7)fits a fluence
distribution, a vector of ten values
Φ
E
j
, to the measured dose distribution
inside the spectrometer by variation of the three parameters (
Φ
E
0
,T
Ph
and
X)inEq.(7) using random numbers. They are varied within the following
ranges: Φ
0
E
40; T
Ph
¼10
7
... 10
18
KandX¼0:2...0:2. The three
parameters are varied independently, one after the next. After every
variation of one of the parameters the ten dose values D
i
calc
are calculated
by Eq. (2). From these ten calculated dose values, the measured dose
values and their uncertainties calculated by the fitfunctioninFig. 1,the
X
2
-value is calculated by Eq. (3).IftheX
2
-value comes closer to 1 after
the variation the new parameter will be maintained, else it will be
discarded. This procedure is repeated in a loop until no further reduction
of X
2
has been performed for 30 iterations. Because of its stochastic
nature this method of unfolding is called Monte Carlo unfolding [12].
Before starting the unfolding-procedure the ten dose readings and
the information if high energy electrons may have been present in the
radiation field in a relevant quantity to affect the dose readings are
queried by the algorithm. When a measurement was carried out
behind thick steel walls or similar shielding an electron contamination
can be negated. The dose contributions from direct electron radiation
are considered by an increase of the uncertainty of the measured dose
values from the five TLD's in the front by the factor 3 (experience
value) before the first step of the algorithm if an electron contamina-
tion is assumed. With this increased uncertainty these dose values
contribute less to the X
2
-value (Eq. (3)). After the first step the
readings from the first five TLD's are substituted by the calculated dose
values as an estimate for the pure photon caused dose. With this new
corrected dose distribution the first step is repeated.
In a last step the fluence configuration is further optimized; the
ten fluence values from the analytical approximation of Eq. (7) are
varied individually also using random numbers and a difference
from the fit result up to 30% for the six lower energy values and up
to 70% (experience values) for the four higher energy values is
allowed. The optimization is checked by calculating the X
2
-value
(Eq. (3))justlikeinthefirst step. The loop breaks after 5000
iterations without a further reduction of X
2
.Thislaststepenables
the unfolding of spectra, which are not well fitted by Eq. (7) for
example because of the common cold and hot electron component
in laser experiments. The algorithm calculates the ambient dose
(H
n
(10)) from the fluence spectrum and the conversion coefficients
from [13]. To estimate the average uncertainty of the ten calculated
spectral fluence values the algorithm calculates the average devia-
tion from the calculated dose values and the (corrected) measured
dose values. The X
2
-minimization process takes around 15 min.
The whole unfolding-procedure has been developed and eval-
uated by means of FLUKA simulations. Several dose distributions
Fig. 2. Schematic view of the TLD-spectrometer: Ten TLD-Cards (Harshaw TLD-700 H) are placed between absorbers of different materials and thicknesses inside a shielding
from lower to higher Z materials (sand-steel-lead) with a collimator window in the front. The incident x-rays penetrate the TLD's. The absorbers cause different response of
every TLD, which can be used as information about the spectrum of the incident x-rays.
Fig. 3. Simulated response functions of the ten TLD-channels in the spectrometer
to photon radiation: The TLD's in the front respond to lower photon energies than
the TLD's in the rear. These different energy thresholds deliver the spectral
information. The different gradients of the response functions provide additional
spectral information in the high energy range.
Fig. 4. Simulated response functions of the ten TLD-channels in the spectrometer to
electron radiation: The steep increase in response with energy visible for the first five TLD's
is due to the direct interaction of the electrons with the LiF crystals. For lower energies and
deeper-lying TLD's only bremsstrahlung generated by the primary electrons is detected.
F. Horst et al. / Nuclear Instruments and Methods in Physics Research A 782 (2015) 69–76 71
inside the spectrometer in bremsstrahlung fields generated by
high energy electrons striking high Z targets were simulated and
the photon spectra unfolded from these dose distributions were
compared to directly scored photon spectra. Such simulations
were carried out for several electrons energies and target thick-
nesses. The sources were set to a widespread shape so that the
divergence did not affect the dose distributions. By the simulation
of very thin targets which could be passed by the primary
electrons the procedure for the elimination of the electron caused
dose on the front channels could also be checked.
3. Results
Validation at Elekta Synergy Linac: For the purpose of validating the
unfolding-procedure the prototype has been irradiated at a clinical
electron linear accelerator (Elekta Synergy) with bremsstrahlung
generated by electrons with energies of 6 MeV and 18 MeV. The
output of the accelerator is expressed in monitor units (MU). The
calibration of the monitor ionization chamber is done in a way, that
100 MU cause an absorbed dose of 1 Gy at the maximum of the depth
dose curve in a water phantom with a field size of 10 10 cm
2
on the
water surface in 1 m distance from the x-ray source. The readings of
the ten TLD's for both electron energies are shown in Fig. 8.
The TLD readings were entered into the unfolding-algorithm.
The photon spectra obtained from the algorithm could then be
compared with spectra from detailed Monte Carlo simulations
(simulated with EGSnrc with transport thresholds of 10 keV for
photons and 521 keV for electrons) of the accelerator, shown in
Fig. 9. By an additional simulation of the calibration situation with
the water phantom the relative results from the Monte Carlo
simulations normalized per primary electron could be converted
into absolute values normalized per MU so that an absolute
comparison is possible. The agreement between the Monte Carlo
simulated spectra and those resulting from the measured values is
quite good (difference of factor 3 and less).
Measurements at PHELIX in December 2013: First measurements at
PHELIX were carried out during a beamtime in December 2013 with
the main objective to study material heating driven by TNSA (target
normalsheetacceleration)protons.Titaniumfoiltargetsof5
μm
thickness were irradiated by the laser with the following parameters:
Fig. 5. Comparison of the dose distribution in the spectrometer obtained by
measurements at radioactive sources (
137
Cs and
60
Co) and by Monte Carlo
simulations: The error bars of the measured dose values represent the statistical
uncertainty due to the dose values (see Fig. 1). The simulation results are given in
dose per primary particle. They were multiplied by the corresponding irradiation
time and source activity. Therefore the error bars of the simulated dose values
consist of the simulation's statistical uncertainty ( o71%) as well as the uncer-
tainty of the source activity ( 720%).
Fig. 6. Simulated response distribution in the spectrometer to monoenergetic
parallel photon beams: The different curve progression for the ten individual
energies enables the unfolding-algorithm to separate between the variable photon
energies from a given response distribution.
Fig. 7. Flow diagram of the unfolding-algorithm. The electron caused dose is
detected and removed by an enhancement of the uncertainties ΔD
meas
i
of the first
five dose values if required. The optimization step (second loop) varies the ten
fluence values Φ
E
j
directly using random numbers in a defined range of deviation
from the fit from the first loop (detailed description in text).
F. Horst et al. / Nuclear Instruments and Methods in Physics Research A 782 (2015) 69–7672
Pulse energy E
P
¼60 J130 J; Pulse duration Δt¼400 fs6 ps;
Focus ΔA¼25 μm25 μm; Angle between beam axis and target
surface φ¼701; Temporal contrast o10
6
.
At PHELIX the energy is measured in front of the pulse
compressor. Around 50% of that energy is deposited on the focal
spot. The following nominal laser intensities Iwere calculated by
I¼E
P
ΔtΔA0:5sin ðφÞð8Þ
The spectrometer was placed outside the target chamber in
80 cm distance from the target with an exchangeable window in
between (see Fig. 10). For comparison of the measured ambient
dose with the ambient dose calculated from the unfolded spectrum
an ambient dosimeter (H
n
ð10Þ
meas
in Fig. 10), a TLD-700 H card in a
polyethylene scattering body (specification in [15]), was placed next
to the spectrometer outside the target chamber. The measurements
were done over two or three shots because of the need for a
cumulated absorbed dose of at least 10 μGy even on the last TLD.
The dose readings had to be corrected for the influence of the beam
divergence (Eq. (4)) before unfolding. The laser focal spot on the
target as the main source of high energy electrons was considered
as the effective bremsstrahlung point source. That assumption
seems justified due to the strongly forward directed bremsstrahlung
emission by relativistic electrons [16]. Because of that effect the
bremsstrahlung photons produced in material around the primary
target can roughly be traced back to the laser focal spot.
The nominal laser intensities as calculated by Eq. (8) can be consid-
ered as the actual intensities because the preplasma expansion was
assumed to be small (some 10 μm according to [17])comparedtothe
Rayleigh length of the used parabola (Rayleigh length: 300 μm).
Also a significant enhancement of the intensity due to self-focusing of
the laser pulse in such a short preplasma is not to be expected.
Fig. 11 shows the corrected dose distributions normalized per laser
shot with three different windows (10 mm steel, 20 mm steel and
1 mm plastic) between the spectrometer and the target. The dose
distribution for the 10 mm steel window is the average over 3 shots
(intensity between 1:210
19
and 1:310
19
W=cm
2
). The dose
distribution for the 20 mm steel window is the average over 2 shots
Fig. 9. X-ray spectra obtained by the unfolding-procedure for an Elekta Synergy clinical linear accelerator with 6 MV and 18 MV bremsstrahlung spectra in comparison to
reference spectra from Monte Carlo simulations of the linac. The photon fluence is normalized per monitor unit (MU).
Fig. 8. Dose distribution inside the spectrometer for measurements at an Elekta
Synergy clinical linear accelerator with two different bremsstrahlung spectra
corrected for the influence of the beam divergence. The dose values are normalized
per monitor unit (MU). 20 monitor units were applied for each measurement. The
spectrometer's position was in 1 m distance from the x-ray source.
Fig. 10. Experimental setup at the PHELIX system in December 2013 (detailed
description in text).
Fig. 11. Dose distribution inside the spectrometer for measurements at PHELIX in
December 2013 corrected for the influence of the beam divergence with three
different windows (10 mm steel, 20 mm steel, and 1 mm plastic) between the
spectrometer and the titanium foil target. The distance between the spectrometer
and the target was 80 cm. The dose values are normalized per shot.
F. Horst et al. / Nuclear Instruments and Methods in Physics Research A 782 (2015) 69–76 73
(1:310
19
and 1:510
19
W=cm
2
). In the case of the measurement
behind the 1 mm plastic window the dose was averaged over 2 shots
with 1:410
18
and 7:510
18
W=cm
2
intensity. The plastic window
wassothinthatitmustbeassumedthatprimaryelectronsmayreach
the TLD's. Therefore the electron correction was activated before
unfolding. The electron caused dose was detected and subtracted as
shown in Fig. 11.Fig. 12 shows the unfolded photon spectra, also
normalized per shot. Table 1 shows the average laser intensity, the
corresponding target chamber window, the mean electron energy E
El
as calculated from the laser intensity by ponderomotive scaling [14],
the mean photon energy E
FLUKA
Ph
as generated by a Boltzmann
distribution of electrons (Eq. (5))withkT¼E
El
in the chamber
window (simulated with FLUKA), the mean photon energy E
Unf
Ph
calculated from the unfolded spectrum, the measured ambient dose
H
n
ð10Þ
meas
, the ambient dose calculated from the unfolded spectrum
H
n
ð10Þ
Unf
as well as the integral photon yield.
As expected the spectra measured behind the steel windows are
harder than the spectrum measured behind the thin plastic window.
The steel windows filtered a large amount of the photons below
100 keV while they could pass the thin plastic window. The mean
photon energy E
Unf
Ph
increases for higher laser intensities as predicted
by ponderomotive scaling (E
FLUKA
Ph
) though there are moderate
differences between prediction and measurement (20–30%). The
ambient dose values H
n
ð10Þ
calc
calculated from the unfolded spectra
fit the measured ambient dose values from the dosimeter next to the
spectrometer (H
n
ð10Þ
meas
) quite good. The diameter of the plastic
window was too small (5 cm) to put an ambient dosimeter next to
the spectrometer, so there is no measurement and H
n
ð10Þ
meas
value
available. Concerning the measurement behind the 10 mm steel
window the measured ambient dose is higher (around factor 2) than
the value calculated from the spectrum. This is plausible due to the
response function of the used ambient dosimeter (shown in [15]). It
is not absolutely identical with the idealized H
n
(10) conversion
function the algorithm used for calculation (only in their operating
range: 10 keV to 1 MeV). Furthermore the ambient dosimeters are
designed for the use in the middle of a radiation field, the
surrounding material could cause additional dose amounts by
scattering effects and x-ray fluorescence. The confidence intervals
given for the H
n
ð10Þ
meas
values do not contain such systematic
uncertainties and do only cover one standard deviation.
ThedosedistributioninFig. 11 showsalargedoseamountonthe
first TLD at the measurements with the plastic window probably
caused directly by electrons. This dose has been detected and
subtracted by the algorithm. Regarding the measurements with the
steel windows there is no comparable step in the dose distribution.
As expected the integral photon yield increases with the laser
intensity. A noticeable fact is that the photon yield was about a factor
2 higher behind the 20 mm steel window compared to the 10 mm steel
window for similar laser intensities(forbothmeasurementsresulting
from 110 710 J in 600 fs). However an enhanced bremsstrahlung
production due to the thicker steel window is implausible because that
would require several electrons with energies higher than 15 MeV (CSDA
range in iron: 10 mm) which is not in accordance to the ponder-
omotive scaling prediction at 1:410
19
W=cm
2
.Thelargerphotonyield
from shot to shot at similar laser intensity and pulse energy might have
itsorigininsomeotherreasonsuchasslightly different target properties
which cannot be determined by the measurement data presented here.
However the observation of electrons penetrating through the 1 mm
plastic window (see Fig. 11) suggests that the steel material around the
primary target might be an additional bright source of bremsstrahlung.
Measurements at PHELIX in January 2014: The spectrometer was
applied for single shot measurements at PHELIX during a beamtime
in January 2014 which had the objective to generate collimated high
energetic TSE (target surface electron) beams from a copper target.
The TSE regime requires a preplasma formed by a prepulse [18].
A magnet electron spectrometer with a 5 mm thick lead collimator
in the front and image plates inside was placed into the expected
beam path. The presented TLD-spectrometer was focused on the lead
collimator from out of the target chamber because besides the primary
copper target the lead was expected to be a bright source of
bremsstrahlung due to the electron beamloss (see Fig. 13).
The laser shot that produced the measurement data shown in
Figs. 14 and 15 had a nominal intensity of 4:510
18
W=cm
2
(pulse energy: 120 J, pulse duration: 476 fs, focus: 20 μm25 μm,
incidence angle: 101from beam axis to target surface) with a prepulse
(intensity ratio: 5 10
6
,timedelay:2:8 ns). The angle of the
spectrometer to the expected electron beam axis was 251.The
distance between the spectrometer and the secondary target was
30 cm, the distance between the spectrometer and the primary target
was 60 cm. As described above the laser focal spot on the primary
target was assumed as the effective point source of the photons for
Table 1
Measurement data from beamtime in December 2013 (see Fig. 10) and theoretical predictions.
Intensity (W/cm
2
) Chamber window E
El
[14] /MeV E
FLUKA
Ph
/ MeV E
Unf
Ph
/MeV H
n
ð10Þ
meas
/μSv/shot H
n
ð10Þ
Unf
/μSv/shot Photon Yield / 1/(sr shot)
1:310
19
10 mm steel 1.14 0.52 0.42 180720 100730 1:25 10
11
1:410
19
20 mm steel 1.21 0.55 0.40 22877 220 770 2:87 10
11
4:510
18
1 mm plastic 0.55 0.20 0.26 / 90730 6:28 10
10
Fig. 12. X-ray spectra obtained by the presented method at PHELIX in
December 2013.
Fig. 13. Experimental setup at the PHELIX system in January 2014 (detailed
description in text).
F. Horst et al. / Nuclear Instruments and Methods in Physics Research A 782 (2015) 69–7674
the correction of the divergence. The door of the target chamber,
25 mm aluminium, was ahead of the spectrometer. The mean energy
of the spectrum shown in Fig. 15 is 498 keV. The corresponding
ambient dose for that shot was 80720 μSv. Due to the strong
prepulse and the resulting plasma expansion the actual laser intensity
was expected to differ substantially from the nominal intensity as
calculated by Eq. (8). Therefore a comparison of the mean photon
energy from the unfolded spectrum to the theoretical prediction of
the hot electron distribution by [14] seems not reasonable here.
4. Discussion
Validation at Elekta Synergy Linac: The accordance of the spectral
distributions measured by the presented method at the linear
accelerator with the reference spectra from Monte Carlo simulations
proves the applicability of the method. There are only moderate
deviations (factor 3 and less).
Simulations showed that the few primary electrons which pass
the filters inside the radiation head (electron contamination) as
well as the secondary electron build-up in material ahead of the
spectrometer are negligible. The actual response of the spectro-
meter to high energetic photons depends on the distance of the
spectrometer from the x-ray source because of the divergence of
the x-ray beam. This has not been taken into account for the
simulation of the response functions. In the case of the measure-
ments at the Elekta linac the distance from the x-ray source to the
spectrometer's surface is well known. Therefore the TLD readings
could be corrected (Eq. (4)) before unfolding.
Measurements at PHELIX in December 2013: The x-ray spectra
measured at PHELIX in December 2013 are confirmed by the
comparison of the ambient dose values calculated from the spectra
with the ambient dose measured beside the spectrometer. The
tendency of an increasing mean electron energy respectively photon
energy with an increasing laser intensity as theoretically predicted
(ponderomotive scaling) is described correctly by the measured
spectra. The moderate deviations of 20–30% between the predicted
mean photon energies and measured mean photon energies could be
caused by the fact that the ponderomotive scaling considers only one
of various acceleration mechanisms which occur in laser-matter
interaction. Furthermore simply averaging over quite different inten-
sities such as 1:410
18
and 7:510
18
W=cm
2
as done for the
1 mm plastic window measurement and calculating the mean
electron energy for the average intensity might be slightly inaccurate.
Moreover the laser focal spot was assumed as the effective point
source of the bremsstrahlung photons for the divergence correction
which is also just a rough approximation.
In theory the laser-accelerated electrons pass the thin foil
target and generate bremsstrahlung in the surrounding material.
The electrons reaching the first TLD channel in the spectrometer
through the plastic window (shown in Fig. 11) confirm that. Thus
the target chamber wall should be taken into account as an x-ray
source additional to the primary target e.g. for the shielding design
around the experimental area of such laser systems as PHELIX.
Measurements at PHELIX in January 2014: The results show that
a single shot x-ray spectrometry measurement is possible with
acceptable accuracy by the presented TLD-based method at
PHELIX or comparable high energy high power lasers. Due to the
short time between measurement and result of around 1 h the
data analysis before the next laser shot can appear (current
repetition rate at PHELIX: 90 min). The high mean photon energy
of 498 keV seems reasonable due to the thick attenuation between
primary target and spectrometer.
5. Conclusion
The presented spectrometry method for high energy photons in
ultrashort pulsed radiation fields has been validated at a clinical
electron linac (6 MV and 18 MV bremsstrahlung). There is a good
agreement between the bremsstrahlung spectra obtained from the
TLD readings by a developed unfolding algorithm and the reference
spectra from Monte Carlo simulations. Spectra with photon energies
up to 10 MeV and mean energies up to 498 keV were obtained at
PHELIX. The actual x-ray source is of great interest for e.g. shielding
design around the target chamber. As discussed above the spectro-
metry measurement results presented here imply that the material of
the target chamber wall might be a bright x-ray source additional to
the primary target. The ambient dose calculated from the spectra
measured at PHELIX (up to 220 770 μSv=shot in 80 cm distance
from the target at laser intensities of around 10
19
W=cm
2
)werein
good agreement with experimental results from conventional ambient
dosimeters. The mean photon energies calculated from the spectra
were in acceptable agreement with theoretical predictions. Further-
more the feasibility of a TLD-based single shot spectrometry measure-
ment at PHELIX was proven.
In [19] two bremsstrahlung spectra observed at the Vulcan
Laser located at the Rutherford Appleton Laboratory under similar
conditions as in our measurements (10
19
W=cm
2
) are presented.
They are qualitatively comparable (in shape and dimension) to the
spectra presented here (Fig. 12).
Photon spectrometry methods based on attenuation as presented
here deliver good spectral information from keV energies up to some
MeV. Therefore the spectrometry method presented here is not really
suitable to resolve the x-ray content above 10 MeV which is
indicated by the two very large energy bins from 10 MeV to
100 MeV. For photon energies of 7 MeV and above methods based
on nuclear activation as described in [20,21] are well suited especially
because they are entirely insensitive to electron radiation.
Fig. 14. Dose distribution inside the spectrometer after a single shot measurement
at PHELIX in January 2014 corrected for the influence of the beam divergence. The
spectrometer was aligned on the back side of the secondary lead target, which was
expected to be hit by laser accelerated electrons. The nominal laser intensity on the
primary copper target was about 4:510
18
W=cm
2
.
Fig. 15. X-ray spectrum obtained by the presented method for a single shot at
PHELIX in January 2014.
F. Horst et al. / Nuclear Instruments and Methods in Physics Research A 782 (2015) 69–76 75
Concerning the presented method the presence of electrons can
disturb low-energy photon spectrometry and should be handled
with great care. Finally, it should be pointed out that spectrometry
methods using a fit procedure as presented here require general
information about the shape of the spectrum which is not
always given.
References
[1] M. Borghesi, Nuclear Instruments and Methods in Physics Research Section A
740 (2014) 6. http://dx.doi.org/10.1016/j.nima.2013.11.098.
[2] C. Gahn, G.D. Tsakiris, A. Pukhov, J. Meyer-ter-Vehn, G. Pretzler, P. Thirolf,
D. Habs, K.J. Witte, Physical Review Letters 83 (23) (1999) 4772. http://dx.doi.
org/10.1103/PhysRevLett.83.4772.
[3] A. Ferrari, P.R. Sala, A. Fasso, J. Ranft, FLUKA: a multi-particle transport code,
CERN-2005-10.
[4] G. Battistoni, F. Cerutti, A. Fassò, A. Ferrari, S. Muraro, J. Ranft, S. Roesler
P.R. Sala, AIP Conference Proceeding 896 (2007) 31.
[5] SCILAB website, 〈http://www.scilab.org〉, version: 1.9.2013.
[6] R. Behrens, H. Schwoerer, S. Düsterer, P. Ambrosi, G. Pretzler, S. Karsch,
R. Sauerbrey, Review of Scientific Instruments 74 (2003) 961. http://dx.doi.
org/10.1063/1.1532831.
[7] I. Kawrakow, D.W.O. Rogers, The EGSnrc Code System: Monte Carlo Simulation
of Electron and Photon Transport, NRCC Report PIRS-701, 2003.
[8] F. Horst, G. Fehrenbacher, T. Radon, E. Kozlova, K. Zink, J. Breckow, Radiation
protection related x‐ray spectrometry at PHELIX, GSI Scientific Report 2013
(2014) p. 200. http://dx.doi.org/10.15120/GR-2014-1-PNI-PP-19.
[9] S.G. Gorbics, F.H. Attix, K. Kerris, Health Physics 25 (1973) 499.
[10] Flair website, 〈http://www.fluka.org/flair〉, version: 22.11.2013.
[11] S. Hasegawa, R. Takashima, M. Todoriki, S. Kikkawa, K. Soda, K. Takano, Y. Oishi,
T. Nayuki, T. Fujii, K. Nemoto, Review of Scientific Instruments 82 (2011)
033301. http://dx.doi.org/10.1063/1.3553496.
[12] R. Sanna, K. O'Brien, Nuclear Instruments and Methods 91 (4) (1971) 573. http:
//dx.doi.org/10.1016/0 029-554X(71)90680-X.
[13] M. Pelliccioni, Radiation Protection Dosimetry 88 (4) (2000) 279.
[14] S.C. Wilks, W.L. Kruer, M. Tabak, A.B. Langdon, Physical Review Letters 69 (9)
(1992) 1383. http://dx.doi.org/10.1103/PhysRevLett.69.1383.
[15] S. Grosam, J.G. Festag, G. Fehrenbacher, K. Vogt, Dose measurements at the
pre-accelerator section of the GSI Unilac, in: Proceedings of the HPS 2008
Midyear Meeting, 20 08, pp. 111–119 .
[16] H.A. Bethe, L.C. Maximon, Physical Review 93 (4) (1954) 768. http://dx.doi.org/
10.1103/PhysRev.93.768.
[17] F. Wagner, S. Bedacht, A. Ortner, M. Roth, A. Tauschwitz, B. Zielbauer,
V. Bagnoud, Optics Express 22 (24) (2014) 29505. http://dx.doi.org/10.1364/
OE.22.029505.
[18] L. Willingale, A.G.R. Thomas, P.M. Nilson, H. Chen, J. Cobble, R.S. Craxton,
A. Maksimchuk, P.A. Norreys, T.C. Sangster, R.H.H. Scott, C. Stoeckl, C. Zulick,
K. Krushelnick, New Journal of Physics 15 (2013) 025023. http://dx.doi.org/
10.1088/1367-2630/15/2/025023.
[19] P.A. Norreys, M. Santala, E. Clark, M. Zepf, I. Watts, F.N. Beg, K. Krushelnick,
M. Tatarakis, A.E. Dangor, X. Fang, P. Graham, T. McCanny, R.P. Singhal, K.W.
D. Ledingham, A. Creswell, D.C.W. Sanderson, J. Magill, A. Machacek, J.S. Wark,
R. Allott, B. Kennedy, D. Neely, Physics of Plasmas 6 (5) (1999) 2150. http://dx.
doi.org/10.1063/1.873466.
[20] T.E. Cowan, A.W. Hunt, T.W. Phillips, S.C. Wilks, M.D. Perry, C. Brown,
W. Fountain, S. Hatchett, J. Johnson, M.H. Key, T. Parnell, D.M. Pennington
R.A. Snavely, Y. Takahashi, Physical Review Letters 84 (5) (2000) 903. http:
//dx.doi.org/10.1103/PhysRevLett.84.903.
[21] M.M. Günther, K. Sonnabend, E. Brambrink, K. Vogt, V. Bagnoud, K. Harres,
M. Roth, Physics of Plasmas 18 (8) (2011) 083102. http://dx.doi.org/10.1063/
1.3613923.
F. Horst et al. / Nuclear Instruments and Methods in Physics Research A 782 (2015) 69–7676
