Study of bremsstrahlung radiation coming from a transmission type x-ray generator

and its application on EDXRF technique

Soumya Chatterjee1 , Sumana Ghosh1 and D. Mitra1
1
Department of Physics, University of Kalyani, Kalyani, West Bengal-741235, India.

Email:[email protected]

Abstract:

We have demonstrated the use of the bremsstrahlung radiation from a small, portable, transmission
type x-ray generator that can be used in energy dispersive x-ray fluorescence technique as a tool to
perform non-destructive elemental analysis of solid samples employed in inter-disciplinary science
research. As our knowledge goes, till date, this is the first attempt to generate the bremsstrahlung
spectrum theoretically from a portable transmission type x-ray generator and compared it with
the actual observation. Theoretically generated bremsstrahlung spectra are also found to be in
good agreement with the experimentally observed spectra obtained with various operating anode
voltages of the x-ray generator. A computer program has been developed utilizing a few atomic
parameters to obtain the elemental concentrations in the sample by a single run using the whole
bremsstrahlung. The knowledge of incoming x-ray flux, geometry of experimental arrangements
are not required in this technique. To validate the technique, we have taken two samples: an
Indian one rupee coin of the year 2000 and a NIST made brass sample, whose compositions are
well known. We have used these samples to expose to the bremsstrahlung radiation, produced by
the operating anode voltages 20, 25, and 30 KV. The relative concentrations of different elements
are determined, which is in good agreement with the earlier results.

Keywords: Elemental analysis, Bremsstrahlung, EDXRF.

I. INTRODUCTION ple use fundamental parameter method (FPM) (Wilson

et al. 1959) to determine the elemental concentration us-
ing EDXRF technique. In this method either one can use
Elemental analysis using Energy Dispersive X-ray Flu- mono-chromatic primary radiation coming directly from
orescence (EDXRF) technique (Mandal et al. 2002) has a radioactive nucleus (e.g. 109 Cd and 241 Am), or by us-
several advantages over other techniques like Particle In- ing nearly monochromatic fluorescence radiation coming
duced x-ray Emission (PIXE), Neutron Activation Anal- from a secondary target due to the incidence of polychro-
ysis (NAA) etc. A wide range of elements can be detected matic primary radiation, produced by an x-ray generator.
using EDXRF technique, even one can detect and quan- In the latter case weighted average energy of the emit-
tify several elements, simultaneously. Moreover, this is a ted Kα and Kβ radiation of the secondary target can be
non-destructive, fast, accurate, and environment friendly treated as mono-energetic radiation which will be con-
technique. To determine the elemental concentration in sidered as the exciting radiation for the sample. Charac-
a specimen with the EDXRF technique, in early days, teristic x-rays coming from different elements present in
people used to draw calibration curves (Rendle 1996) the sample can be detected using x-ray detector and from
with several standard reference samples. With the ad- the intensities of the characteristic lines one can calculate
vancement of computer facilities people started using the elemental concentrations with the help of detector ef-
the alpha-coefficient method (Tertian and Claisse 1982) ficiency and few atomic parameters. No standard sample
where exhaustive regression analysis is required. In this is required in this method. In case of low wattage, small
procedure there is a need to know the ratio (Ri ) of in- size x-ray generators, emitting x-ray flux is very small
tensity of a particular element coming from the sample and the intensity of the near mono-chromatic secondary
to that from a standard. On the other hand, knowl- radiation is even smaller which in turn will take longer
edge of the values of the incident flux, experimental ge- time for accumulating appreciable statistics under differ-
ometry, detector efficiency as well as the photo-electric ent characteristic x-ray lines in the spectrum. So, one
cross-sections are not required in this method. But for can take the whole bremsstrahlung coming from the x-
the specimen with a large number of elements, one has ray generator for exciting the sample instead of taking the
to make several standards for each and every element in fluorescence radiation coming from the secondary target
the sample to determine the (Ri ), which is very time con- to reduce the experimental time. But due to the pres-
suming as well as costlier. Accurate knowledge of the val- ence of a large band energy range in bremsstrahlung it
ues of different atomic parameters (e.g. photo-ionisation is not an easy task at all. In this paper we would like to
cross sections, fluorescence yields, Coster-Kronig tran- demonstrate how one can use the whole bremsstrahlung
sitions, radiative widths, etc.) and the availability of radiation, coming from a low power, transmission type
high speed computer with large memory, now a days peo-
 2

portable (Ferrero et al. 2002, Gigante and Cesareo 1998) number of thin slices so that each slice contributes to
x-ray generator (Amptek), to find the elemental concen- energy loss ∆E, selected by continuous slowing down ap-
trations of the samples using the principles of Fundamen- proximation assumption. Thick target bremsstrahlung
tal Parameter Method (FPM). As per our knowledge, no emission per unit path length per unit solid angle in the
attempt has ever been made for EDXRF technique by energy interval k to k + dk, is
using ‘white spectrum’(Pantenburg et al. 1992) of the Z E0
bremsstrahlung radiation directly coming from a trans- dE
IE0 ,k = IE0 ,k ,E ( ) (1)
mission type x-ray generator. E>k −dE/dx

Where IE0 ,k ,E is for thin target bremsstrahlung emis-

II. EXPERIMENT sion per unit length per unit solid angle, reviewed by
Koch and Motz (Koch and Motz 1959) in the energy in-
Here we have used an EDXRF system with a small terval k to k + dk. dE/dx is the mean electron energy
transmission type portable x-ray generator of Amptek loss per unit path length. From equation (1), thick tar-
(Mini-X), having maximum operating power of 4 W with get distributions are calculated which is comparable with
maximum attainable high voltage of 50 KV. This x-ray the Kramers’ semi classical non relativistic calculations
generator is made with a metal-ceramic type anode made (Storm 1972). Further they have considered the electron
of silver (Ag) backed with Al2 O3 . While doing the ex- back scattering losses by modifying the expression (1) as
periment, we first record bremsstrahlung spectra directly Z E0
for three different anode voltages of 20 KV, 25 KV and dE
IE0 ,k = IE0 ,k ,E ( )(1 − ηϵE0 ,k,E ) (2)
30 KV, coming from x-ray tube by placing x-ray detec- E>k −dE/dx
tor in front of it at a distance of 12 cm in air. To clean
the required energy region, which is comparatively at the Fraction of the incident back-scattered electrons coming
low energy side, we placed two absorbers of tungsten (W ) from a thick target and total number of electrons aver-
and aluminium (Al) having thicknesses of 1 mil (=0.0254 aged over all angles back-scattered with energies between
mm) and 10 mil, respectively in front of the x-ray genera- E to E0 are denoted as η and ϵE0 ,k,E respectively. A
tor. The intrinsic efficiency of the detector is supplied by modified semi-empirical formula of Storm (Storm 1972),
the maufacturer at different energies. The sample was is used to fit experimental bremsstrahlung is given as,
placed in front of the x-ray generator at a distance of
12 cm to irradiate it with the bremsstrahlung radiation 11 Z(E0 − k)(1 − e−3k/Ek )
IE0 ,k = × f E0 , k
in such a way that the entrance and exit angles of x- 4π 1
(k/E0 ) 3 (1 − e−E0 /Ek ) (3)
rays were fixed at 45◦ with respect to the target plane.
[erg/s/mA/keV /sr]
The Amptek Si pin detector (Amptek-X-123 Complete
X-Ray Spectrometer) was utilized to detect the x-rays.
where IE0 ,k , Z, E0 are for bremsstrahlung emission in-
The x-ray detector exhibited a resolution of 160 eV at 5.9
tensity, target atomic number and incident electron en-
keV. Two samples were used in this study: sample-1 : an
ergy in keV respectively. k and Ek denote the emitted
Indian one rupee coin, minted in 2000, and sample-2 was
photon energy and K-shell ionization potential in keV
a piece of Brass sourced from NIST (National Institute
of anode element, respectively. It can be seen that the
of Standards and Technology).
Kramer-Kulenkampff-Dyson (KKD) model agrees well
with experimental values in the range 12-300 keV. In
the present set-up, transmission type x-ray tube is used.
III. ANALYSIS
In this scenario, the effective penetration length is the
width of the effective thickness of the target. To the best
For convenience we split our work into two parts. of our knowledge no attempt has been made to gener-
Firstly, the experimental bremsstrahlung spectra is fitted ate bremsstrahlung x-rays theoretically from a transmis-
with the appropriate theoretical model for transmission sion type x-ray generator. For transmission type metal-
type x-ray tube. Secondly, by using these bremsstrahlung ceramic x-ray tube with silver anode, the photon atten-
x-rays as an exciter of the sample, the constituent el- uation correction factor fE0 , k is given by
ements and their relative concentrations present in the
samples were estimated.
fE0 ,k = exp[−µk x] (4)
A. Understanding the Bremsstrahlung Spectrum where µk is the total attenuation coefficient of a photon
at energy k in silver anode. Effective thickness of silver
The bremsstrahlung spectrum for thick targets can be anode can be calculated by using the following expression
found from thin target distribution by calculating elec-
tron energy loss with target thickness by collisions with E0 2 − k 2
atomic electrons. Target thickness is divided into a large x= (5)
ρC
 3

where the density of silver is denoted by ρ. The above-

mentioned equations (4) and (5) are the modified ex-
pressions for transmission type x-ray tube. “C” is
the Thompson-Whiddington constant which can be ex-
tracted by fitting the data given in the paper of Birch
and Marshall (Birch and Marshall 1979) for all the an-
ode voltages.
To compare the bremsstrahlung emission intensity as
given by equation (3) with the experimental spectrum,
we converted it to the photon number in the energy gap
(∆k) at photon energy k. This energy gap is exactly the
same as we used during the calibration of the x-ray de-
tector in the present experiment. The same calibration
throughout the experiment for all the operating anode
voltages is maintained. For conversion of bremsstrahlung
emission intensity to photon number (Goel and Shanker
1996), we have used the relation as follows

Nk 1.602 × 103 [erg]

IE0 ,k = (6)
NE ∆k (Ω/4π) ϵ [smAkeV sr]

where Nk denotes the total photon number in the en- FIG. 1: The figure depicts three different
ergy gap (∆k). NE , Ω, ϵ are the total charge (nC) at bremsstrahlung spectra for various operating anode
the target, solid angle subtended by the x-ray detector voltages of the transmission type x-ray generator. The
at target and the detector efficiency at photon energy k experimental spectrum is represented by black dotted
which is supplied by the manufacturer of the detector re- points, the theoretically generated spectrum without
spectively. considering the detector efficiency is shown as a blue
Now this photon has to cross the ceramic (Al2 O3 ) part dashed line, and the red line corresponds to the
of the metal-ceramic anode. As the thickness of Al2 O3 is theoretically generated spectrum with the inclusion of
not supplied by the manufacturer, we normalised the the- the x-ray detector’s effective efficiency.
oretical spectra with different thicknesses through several
trials in order to match the experimental one and used
the same thickness of Al2 O3 (nearly 0.25 cm), for all the the characteristic lines are very small with respect to the
operating anode voltages. Two absorbers of aluminium continuous one. We have done this experiment for var-
and tungsten of thicknesses 1 mil and 10 mil respectively ious operating anode voltages of 20 KV, 25 KV, 30 KV
are placed to suppress the low energy side of the spec- and for each case the experimental spectrum is perfectly
trum. Moreover, the x-ray tube has a beryllium window matched with the theoretical one. As the K-shell binding
in front of it. To consider the total absorption of x-rays energy of Silver (Ag) is 25.514 keV, so among our three
before coming out from the tube the following order is operating voltages we can see the characteristic lines for
considered as i = 1 to 4 for Al2 O3 , Be, Al and W , only the tube voltage of 30 KV. For concentration cal-
respectively. We take all required mass absorption coeffi- culation we have used the bremsstrahlung profile (blue
cient values from XCOM (Berger 1987). Bremsstrahlung dashed line), which is actually coming out from the x-
emission intensity that emerges out by passing through ray generator.
all the aborbers is given by

P B. Concentration Calculation
Nkf inal = Nk e i (−µk )i (x)i
(7)
Now while considering the elemental analysis by mono-
Fig. 1 shows the bremsstrahlung spectrum coming out chromatic x-rays, fluorescent x-ray line intensity of an
from x-ray generator without taking the detector effi- element (Ii ) (Birch and Marshall 1979) can be written as
ciency contribution and the spectrum after considering
the detector efficiency along with the actual experimen-
tal values for three different operating anode voltages (20 I0 Ω
Ii = [σi ωi fi ] Ai ϵi Ci (1 + Hi )
KV, 25 KV, 30 KV). We have also generated the char- 4πsinθ1
acteristic lines of silver K x-rays due to the silver an- = S [σi ωi fi ] Ai ϵi Ci (1 + Hi )
ode with the help of reference (Ebel 1999, Pella et al. = SVi Ci (8)
1985). However, contributions due to the characteristic
lines, used as an exciter are negligible in comparison to Where I0 , Ω, θ1 are for incident flux, solid angle sub-
the continuous bremsstrahlung spectra as the area under tended by the detector with respect to the target and
 4

FIG. 2: A typical spectrum obtained due to the bremsstrahlung x-ray irradiation, from a transmission type x-ray
generator operating at 25 KV anode voltage, of an Indian one rupee coin [2000][2(A)]. A typical spectrum of Brass
Sample [NIST] obtained due to the bremsstrahlung x-ray irradiation, from a transmission type x-ray generator
operating at 25 KV operating anode voltage[2(B)].

entrance angle, respectively. σi and ωi denote photo- by

ionization cross section of the ith element for primary
 µs 0
radiation and fluorescence yield. fi is the fraction of radi- Cj [σ0 ]j [ωk fk ]j [σj ]i
 
0 sinθ1
ation of the considered x-ray line and Ai is the absorption Hi = × [sinθ1 /µs ]ln +1
2[σ0 ]i µs j
correction not only for the primary x-rays but also for the  µs 0 
fluorescent x-rays inside the target. ϵi is the detector ef- 0 sinθ2
ficiency for fluorescent radiation, Hi is the inter-element +[sinθ2 /µs ]ln +1
µs j
enhancement factor and Ci is for relative concentration (14)
for ith element of the sample. We calculate Ii by using This enhancement factor denotes the effect of excitation
the area under the Kα line of each sample. The other of the ith element due to the j th element. [σj ]i is for
terms can be explained as photo-ionization cross-section of the ith element due to
the j th element whereas [σ0 ]i and [σ0 ]j are the same for
ith and j th element due to the primary radiation. This
factor depends on the matrix element inside the target.
I0 Ω For our target this factor contributes a very little amount,
S= (9)
4πsinθ1 much less than 1% to the concentration calculation. This
1 − exp(−[µ0s /sinθ1 + µis /sinθ2 ]t) contribution is so small that we neglect the contribution
Ai = (10) coming from all other characteristic lines of element “j”
[µ0s /sinθ1 + µis /sinθ2 ]
that can excite atoms of element “i” in the relevant shell.
µ0s = ΣCi µ0 ; (11) Now from equation (8) we can write
µis = ΣCi µi ; (12)
Ii
Vi = [σi ωi fi ]Ai ϵi (1 + Hi ) (13) Ci = (15)
SVi
We have assumed that all the detectable elements will
add up to 100%. To fulfill this condition we must use
where θ2 is for exit angle, “t” denotes the thickness of X
the sample, µ0s and µis denote the mass absorption coef- Ci = 1 (16)
ficients of the sample for primary and fluorescent radia-
tions, respectively. Inter-element enhancement factor Hi So, by using the set of equations (9) to (16), assuming
for infinitely thick sample can be easily calculated by an- an initial arbitrary concentration of each element we can
alytically developed expression of Sparks (1975) as given find the corresponding modified concentrations, which is
 5

Results of an Indian one rupee coin(2000) made of IV. RESULTS AND DISCUSSIONS
Cr and Fe
Earlier Results [22]
Anode
Our Average using Fig. 2 shows the characteristic x-ray spectrum of a coin
Voltage
Results (in wt%) Value (in wt%) mono-chromatic x-rays
(in KV)
(in wt%)
and brass sample, obtained due to the irradiation of the
Cr 17.1 ± 1.5 samples with bremsstrahlung radiation emitted directly
20
Fe 82.9 ± 4.2 Cr 17.1 ± 1.4 Cr 16.8 ± 1.2 from the x-ray tube operating at a 25 KV anode voltage.
Cr 17.2 ± 1.4 Fe 82.9 ± 4.2 Fe 83.2 ± 5.8 By using the above mentioned procedure the concentra-
25
Fe 82.8 ± 4.1
Cr 16.9 ±1.4
tions of each elements of two samples have been cal-
30 culated and the results are shown in Table. I and Ta-
Fe 83.1 ± 4.3
ble. II along with the previously published results [(Man-
TABLE I: Concentrations of the compositions of an dal et al. 2003) for Indian one rupee coin] and quoted
Indian one rupee coin (minted in 2000) NIST Brass Sample compositions. Both the results agree
well within the error bar. From equation 15, it seems
that the error in the concentration comes from only the
Results of Brass Sample (NIST) terms Ii and Vi . The term S cancels out while normal-
Anode Voltage Our Results Average Value Quoted
(KV) (in wt %) (in wt%) (in wt %)
izing the concentrations. Errors are almost negligible for
Fe 0.20 ± 0.02 the elements with higher concentrations, and written as
Ni 0.10 ± 0.02 follows,
Cu 62.5 ± 4.3
20 Fe 0.20 ± 0.02 Fe 0.088 r
Zn 35.3 ± 4.3 ∆Ci ∆Ii 2 ∆Vi 2
Ni 0.10 ± 0.02 Ni 0.07 ( )= ( ) +( ) (18)
Sn Not detected Ci Ii Vi
Cu 62.8 ± 4.3 Cu 61.33
Pb 1.90 ± 0.2
Zn 35.1 ± 4.3 Zn 35.31
Fe 0.20 ± 0.02 The error in Vi is propagated through the errors in (i)
Sn Not detected Sn 0.43
Ni 0.10 ± 0.02
Cu 62.9 ± 4.3
Pb 1.8 ± 0.2 Pb 2.77 photo-ionization cross section of k-shell (∼2%) and (ii)
25 atomic parameters (ω) of k-shell (∼3%), (iii) mass atten-
Zn 35.1 ± 4.3
Sn Not detected uation coefficients (∼2%).
Pb 1.7 ± 0.2 s
Fe 0.20 ± 0.02 ∆Vi ∆σi 2 ∆ωi 2 ∆fi 2 ∆ϵi 2 ∆Ai 2
Ni 0.10 ± 0.02 ( )= ( ) +( ) +( ) +( ) +( )
Cu 63.1 ± 4.3 Vi σi ωi fi ϵi Ai
30
Zn 34.8 ± 4.3 (19)
Sn Not detected Neglecting the error in Hi due to its small contribution,
Pb 1.8 ± 0.2
the error in Ci turns out to be within 3-8%. Overall
errors in the concentration for each element is listed in
TABLE II: Concentrations of the compositions of Brass
table I and II.
Sample [NIST]
Tin was not detected in the peresent measurement due to
its low concentration and the present experimental sce-
nario. As we know that the K-shell binding energy of tin
used as assumed concentrations for the next iteration. is around 29 keV, so for the tube voltage of 20 and 25
After each iteration, concentration of each elements (Ci ) KV its K-shell will not be excited, resulting no emission
is normalized by of K x-rays. Also for the case of L x-rays, ionzation cross
section will be too small for these energies. For the case
Ci of 30 KV operating voltage, due to the very low intensity
Ci = P ′
(17) under the Kα line of tin, it is submerged within the scat-
i′ Ci tered bremsstrahlung profile and cannot be detected.
For the case of Iron and Nickel, large differences are
The above mentioned procedure is applicable only when noticed between present and the quoted values due to
mono-energetic x-rays are used as an exciter. But here the low concentrations of these two elements. Moreover
the whole bremsstrahlung coming from a transmission Nickel Kα peak is situated on the rising shoulder of the
type x-ray tube were used to irradiate the sample. In high intense Cupper line. So the uncertainties of these
order to use the above mentioned procedure, the whole two elements are much larger compared to the other de-
bremsstrahlung spectrum is divided into a large number tected elements present in the sample.
of energy slices so that each energy slice can be treated Thus it can be concluded that one can easily find the con-
as a source of mono-energetic x-ray beam. The width centrations of each element of the sample by irradiating
of this energy slice is the same as the energy per unit them with the bremsstrahlung radiation directly coming
channel in the present experiment. We have calculated from a portable transmission type x-ray tube without
the concentrations (Ci ) for all the energy slices separately using any secondary source as well as any standard ref-
and add them up and normalize to get the modified one erences.
for the next iteration if needed. The main advantages of this type of setup are (i) a very
 6

low powered portable x-ray tube can be used, (ii) it can Ghosh: Review & editing, Preparation of tables. D. Mi-
be operated by using a stable power source like a car bat- tra: Conceptualization, Formal analysis, Methodology,
tery, (iii) peltier cool small x-ray detector and (iv) with Validation, Supervision, Writing – review & editing.
a laptop.

V. ACKNOWLEDGEMENTS

The authors would like to acknowledge DST-FIST for

funding the procurement of the instruments and Univer-
sity of Kalyani for providing necessary infrastructural fa-
cility.

VII. DATA AVAILABILITY STATEMENT

VI. AUTHOR CONTRIBUTION STATEMENT

Soumya Chatterjee: Methodology, Calculation, Anal-

ysis, Preparation of tables and figures, Writing. Sumana This manuscript has no associated data.

Berger, M. (1987). ” xcom: Photon cross section database,” Mandal, A., Sarkar, M., and Bhattacharya, D. (2002). A sim-
web version 1.2. http://physics. nist. gov/xcom. ple edxrf technique to analyse alloys. The European Physical
Birch, R. and Marshall, M. (1979). Computation of Journal-Applied Physics, 17(1):81–84.
bremsstrahlung x-ray spectra and comparison with spectra Pantenburg, F., Beier, T., Hennrich, F., and Mommsen, H.
measured with a ge (li) detector. Physics in Medicine & Bi- (1992). The fundamental parameter method applied to x-ray
ology, 24(3):505. fluorescence analysis with synchrotron radiation. Nuclear In-
Ebel, H. (1999). X-ray tube spectra. X-Ray Spectrometry: struments and Methods in Physics Research Section B: Beam
An International Journal, 28(4):255–266. Interactions with Materials and Atoms, 68(1-4):125–132.
Ferrero, J., Roldán, C., Juanes, D., Rollano, E., and Morera, Pella, P., Feng, L., and Small, J. (1985). An analytical algo-
C. (2002). Analysis of pigments from spanish works of art rithm for calculation of spectral distributions of x-ray tubes
using a portable edxrf spectrometer. X-Ray Spectrometry: for quantitative x-ray fluorescence analysis. X-Ray Spectrom-
An International Journal, 31(6):441–447. etry, 14(3):125–135.
Gigante, G. and Cesareo, R. (1998). Non-destructive analysis Rendle, D. (1996). , r. jenkins, rw gould, d. gedcke (eds.)
of ancient metal alloys by in situ edxrf transportable equip- quantitative x-ray spectrometry, dekker, amsterdam (1995),
ment. Radiation Physics and Chemistry, 51(4-6):689–700. p. xi+ 484, isbn: 0-8247-9554-7.
Goel, S. and Shanker, R. (1996). Experimental thick-target Sparks, C. (1975). Quantitative x-ray fluorescent analysis
bremsstrahlung spectra from collisions of 7.0-kev electrons using fundamental parameters. Advances in X-ray Analysis,
with ag and au. Physical Review A, 54(3):2056. 19:19–52.
Koch, H. and Motz, J. (1959). Bremsstrahlung cross-section Storm, E. (1972). Calculated bremsstrahlung spectra from
formulas and related data. Reviews of modern physics, thick tungsten targets. Physical Review A, 5(6):2328.
31(4):920. Tertian, R. and Claisse, F. (1982). Principles of quantitative
Mandal, A., Santra, S., Mitra, D., Sarkar, M., and Bhat- X-ray fluorescence analysis. Heyden.
tacharya, D. (2003). Edxrf as a routine tool for numismatic Wilson, C. L., Wilson, D. W., and Svehla, G. (1959). Compre-
studies. Current Science, 85(2):134–135. hensive analytical chemistry, volume 37. Elsevier Publishing
Company.