The Impact of XRF Escape Process on the Energy Resolution of a NaI Scintillation Detector for Medical Imaging or Gamma-Ray Spectroscopy

1. Introduction

Robert Hofstadter presented in 1948 the NaI:Tl, a scintillator showing the highest light-output up to this time by initiating the technique of pulse-height scintillation spectroscopy for gamma-rays [1]. A “clear” single-crystal, coupled to a Photo-Multiplier Tube (PMT), gave much higher signals with respect to plastic scintillators, that, in turn, were sometimes used in the early practice. Due to the large preponderance of scattering events, compared to photo-electric ones, a pulse-height spectrum from plastics did not show a real photo-peak, rather a Compton continuum. Consequently the spectrum unfolding resulted in some laboriousness in order to obtain the energy value of primary gamma-rays.

Before 1948, gamma-ray spectroscopy was performed by using experimental methods related to the range of gamma-rays energy.

Above around 500 keV, the value of photon-energy needed to be inferred from the distribution of secondary electrons emitted by atoms during the process of internal conversion, where the energy of the excited nucleus is transferred to the secondary particle to be emitted.

In 1932, Taylor and Mott published their theory of internal conversion of gamma-rays on the atomic orbitals, predicting the dependence of the value of conversion coefficient on both energy and multipolarity of the incident gamma-radiation for different atomic shells [2].

Three years later, Jaeger and Hulme published their theory of internal conversion for gamma-rays above 1.022 MeV with production of pairs electron-positron for certain nuclei and found the dependence of the conversion coefficient on both energy and multipolarity of the radiation [3].

According to these theories an experimental technique was developed for charged particles detection, needing electrons focused by a transverse magnetic field and two Geiger-Müller counters operated in coincidence. The method showed the major limitations of accurate beam focusing which, in tour, needed huge-activity radio-isotopic source (up to 20 GBq) and, therefore, large and heavy shielding [4].

The Reviews of Modern Physics published from 1940 to 1958 five issues of the Tables of (the) Isotopes, under the supervision of G.T. Seaborg, collecting experimental quantities regarding both natural and artificial radio-isotopes [5,6,7,8,9]. After 1948 the amount of contents in the issues has undergone strong growth due to worldwide dissemination of low-cost and portable NaI:Tl instrumentations, well suitable for measurements of low-activity samples.

Edward U. Condon, editor of the last issue, in his preface wrote: “This table is the fifth in a series of compilations which have appeared at approximately equal time intervals whose increasing length supplies a beautiful example of the exponential way in which knowledge grows.” The trend has been reported in Figure 1.

This Introduction ends with an invitation to researchers in this field to investigate the real reasons why NaI still remains widely the most popular scintillator. This work attempts to provide some elements to approach these reasons.

2. Background

In principle, the process of X-ray escape does not occur in the case of a scintillation crystal with infinite size and with the gamma-ray source located inside it, because all the secondary radiation, resulting from the interactions, are absorbed in its volume. This contributes to collect the entire energy of the incoming event, provided that all the scintillation light-photons reach the photocathode (see Ref. [10], Chapter 10).

In real scale, certain cases, in which the atoms emitting the fluorescence X-rays are located in close proximity of an outermost wall of the scintillator, can not contribute to the total event-energy due to the escape of the X-ray itself.

Such events tend to form other peaks on the left of the full-energy one at distances equal to the X-ray energies, being these X-rays characteristic of the scintillator components.

In principle, multiple escape-peaks are created in the spectrum, depending on the number of components of the scintillator, and on the atomic shells (K, L, M, …) involved in the processes.

In practice, for a NaI crystal, it is enough to consider, with a good approximation, the emissions from the iodine because the energy of X-ray K-shell edge of sodium is around 1 keV only, against the 33 keV about of iodine (see Table 1). This, substantially, makes the sodium K-escape peak practically indistinguishable from the full-energy peak.

In general, for all the elements of the Table of Mendeleev, the atomic shells other than the K are characterized by both energies and intensities values strongly lower than the K one [11], which authorizes to disregard them.

The same occurs in the majority of cases for the doping elements added to the scintillator components for optimizing their light-yield.

Finally, since the standard deviation of the average X-ray energy of the K-shells of a given scintillator component is considerably small compared to the typical detector energy resolution, the energy-values of the single-shells of such a component can be replaced, in the calculation of the escape-peak position in the pulse-height spectrum by the respective mean values.

Summarizing, for NaI, the iodine component will produce a specific escape-peak, on the left of the full-energy one, shifted from the last by a quantity equal to the average energy of the K-shells.

Last, it could be helpful to remark that the energy values of K X-rays is strongly increasing with respect to the atomic number Z of the considered individual scintillator component [11]. So, for each component, the higher the value of Z, the larger the shift on the left of the escape-peak from the full-energy one.

Regarding the quantification of iodine X-ray escape from NaI, a former work deducing the correction-factor of the photo-peak area for this process was presented in 1954 by P.Axel [12]. The method supposed that iodine X-rays were only from photoelectric interactions (i.e. the origin from Compton-scattered photons was excluded) and it was based on solid-angle calculations for front or lateral positions of a point source with respect to a right-cylindrical scintillator. The importance of the correction was found dependent on the energy of the radiation, on the dimensions of the crystal, and on the experimental setup. The order of magnitude of the ratio escape-to-total was found in the range from around 0.4 to 0.01 for photon-energy ranging from 33 to 150 keV, respectively.

The same year an experimental work by W.E. Meyerhof and H.I. West jr. substantially confirmed the Axel’s results [13].

3. Materials & Methods

3.1. NaI composition and characteristic fluorescence X-rays

An evaluation of the impact of X-ray escape process on the energy resolution of a NaI scintillation detector is given in present work. For this purpose, we have considered the photon-energy range from 1 keV to 1 MeV, which includes the domain of interest for medical imaging with radionuclides.

The values of the Elemental Composition (ECmp) of the NaI scintillator as well as the fluorescence K-shell edge X-ray energy are reported in Table 1. The ECmp values have been calculated at the NIST 1 website [14], while the X-rays data are from [11].

The characteristic fluorescence X-rays values from the NaI scintillator components and for the two innermost atomic shells are listed by ascending energy in Table 2. The X-ray data have been rearranged from Ref. [11], and the calculated quantities are described in the respective column header. Intensity values have been estimated by using the approximation rules suggested in the same Ref. [11]. As it is well known, the intensity values are usually reported, shell by shell, relative to the most intense line for a given atomic shell. Lines whose energy values is below 1 keV have been omitted.

3.2. NaI Interaction data

The values of the interaction coefficients in the photon energy range from 1 keV to 1 MeV have been calculated at the NIST website by using the XCOM software and the Photon Cross Sections Database [14]. Results are shown in Figure 2.

The trend of the total linear interaction coefficient as a function of the photon energy μ_(E), can be obtained, in the given photon-energy range, by summing the ones of the photoelectric and Compton curves because the other physical processes produce interactions above the threshold of 1.022 MeV. The coherent scatter, that by definition is a not energy-spending physical process, is non added because it does not bring to the excitation of scintillator atoms. This can be written in Eq.(1).

μ_(E) = τ_(E) + σ_(E)

(1)

where: μ_(E) represents the total linear interaction coefficient, while the τ_(E) and σ_(E) stay for the photoelectric and Compton ones, respectively.

By using the μ_(E) data, expressed in cm^-1 with the density value of NaI of 3.67 g/cm³ (Ref. [10], p.235), the corresponding value λ_(E) of the Mean Free Path (MFP_(E)) at the given photon energy, expressed in cm, can be obtained from Eq.(2) (Ref. [10], p.53):

λ_(E) = MFP_(E) = 1 /μ_(E)

(2)

whose trend as a function of the photon energy is reported in Figure 3.

The definition of this quantity, recalled from the same reference, is «the average distance traveled in the absorber before an interaction».

As above mentioned, the events, whose interaction stories within the scintillators include a X-ray photon escaping from the detector, contribute to building a specific escape-peak on the left of the full-energy peak. The distance (in energy units) between the centroids of these two peaks corresponds to the energy of the fluorescence X-ray characteristic of the atom involved in the interaction.

Therefore one can write the Eqs.(3-4) :

E_Escape = E_Gamma – E_XRF,

(3)

where, of course,

E_Gamma≥ E_XRFedge

(2)

where the symbols represent the energy values as follows:

E_Escape of the escaping X-ray,

E_Gamma of the incoming photon,

E_XRF of the X-ray fluorescence from a given atomic shell, and

E_XRFedge of the edge of the given shell.

Due to the substantial similarity of the values of the binding energies within a given atomic shell (see Table 2), in an acceptable approximation for present study, it is satisfactory to assume the E_XRF as the value of the average of the energies weighted by the respective intensity values for the atomic shells. Furthermore, given the difference between the average energy values for different atomic orbitals, it would be enough in the present evaluation to consider the K-orbital only.

The physical process of the X-ray escape is thus a threshold process, that is, it takes place when the value of the energy of the incident photon exceeds or at most equals that of the ionization energy of an electron that is bound to an atom of one of the scintillator components.

The discontinuities of the curves shown in Figure 2 and Figure 3 are characteristic of both sodium and iodine, and are also reported in Table 2.

It should be observed that in the given energy interval (1 keV - 1 MeV) of the incident photon the probability of interaction is distributed in a binary way, with a ratio dependent on the energy, between photo-electric and scattering events. The latter are, in turn, divided between elastic (or coherent, or also Rayleigh) and inelastic scattering (or incoherent, as well Compton) events. The elastic scattering does not imply an exchange of energy between the incident photon and the atom involved in the interaction, but only a deviation of the trajectory of the photon itself from its original direction.

In this regard, see the Figure 2 where the interaction probability trends are reported for the NaI scintillator. In particular, it can be noted that the elastic interaction is (variously) prevalent among the scattering events for photon energy up to 89 keV, while the photoelectric events are (largely) prevalent over the scattering events for energies below 286 keV.

3.3. Organic and Inorganic scintillators

Regarding the process of X-ray escape, the behavior of organic and inorganic scintillators is in practice very different from each other because the formers typically have components with Z = 8 (oxygen) or less, while the latters do not show this limitation in Z.

Consequently, with reference to Eqs.(3-4), for the organic scintillators we have E_Escape ≈ E_Gamma since E_{XRF edge} ≤ 0.533 keV [11], and, experimentally, only a minimally distorted single-peak is observed. In practice, these scintillators can be considered as escape-free.

On the contrary, for the inorganics ones, it is usual to note a broadening of the peak in E_Gamma towards the left, up to its splitting, which depends on the scintillator ECmp and on the value of E_Gamma. This can occur if one or more elements with sufficiently high concentrations and with a high enough Z are among the components of the scintillator. In fact, as the value of the E_{XRF edge} increases with Z at an average rate of about 1 keV per unit of Z [11], its order of magnitude can become greater than the half width at half maximum of the peak in E_Gamma itself.

3.4. Regions of Interest definitions

The evaluation method is based on mathematical analysis tools and, particularly, Gaussians functions for representing the single-peaks and the gamma-ray spectrum obtained by summing individual peaks.

For the calculation purposes, arranged by 2.5 keV steps, two simplifying assumptions have been made:

1) The first one regards the detection efficiency, whose value affects the area of the peaks, that has been assumed constant within each selected Region of Interest (RoI), due to their limited size with respect to centroid. The Table 3, Table 4 and Table 5 report the values of the nuclear, of the atomic emissions, as well as of the iodine-K escape peaks calculated for each of them. The average value and its standard deviation has been added to each Table to quantify the overall dispersion.

2) The second assumption refers to the iodine-K escape energy-value that has been taken equal to the weighted average of the single-shell values with respect to the intensities.

In this regard, the following RoIs have been selected:

- 140 keV from the Tc-99 decay, the most diffuse isotope utilized in the Single Photon Emission Computerized Tomography (SPECT) technique,

- 511 keV from F-18 decay usually used in Positro Emission Tomography (PET) technique and,

- 662 keV from Cs-137 decay, isotope widely in use for calibration in gamma-ray spectroscopy.

As specified in the respective “Table of nuclear and atomic radiations from nuclear decay, in the medical internal radiation dose format from the evaluated nuclear structure data file” 2, the photons included in these RoIs are identified so as to include the emissions whose energy is close to the peak emission which are not distinguishable for energy resolution reasons. Data concerning the gamma emissions within the considered RoIs, are reported in Table 3, Table 4 and Table 5.

Table 3. Tc-99 and the corresponding escape peaks considered in the calculations. The values are reported in descending order. The half-life is of 6.0072 Hours, and the decay modes are β^-3 and IT 4 [15].

(i)	Radiation (i)	Egamma (i) (keV)	Yield (i) (Bq-s)-1
1	γ 2	140.5	0.8890
2	Ce-L, γ 2	137.0	0.0106
3	Ce-K, γ 2	119.0	0.0879
4	γ 2 esc	111.3
5	Ce-L, γ 2 esc	108.2
6	Ce-K, γ 2 esc	90.2
	Average	117.8
	Std.deviation	19.0	16 %

Table 3. Tc-99 and the corresponding escape peaks considered in the calculations. The values are reported in descending order. The half-life is of 6.0072 Hours, and the decay modes are β^-3 and IT 4 [15].

(i)	Radiation (i)	Egamma (i) (keV)	Yield (i) (Bq-s)-1
1	γ 2	140.5	0.8890
2	Ce-L, γ 2	137.0	0.0106
3	Ce-K, γ 2	119.0	0.0879
4	γ 2 esc	111.3
5	Ce-L, γ 2 esc	108.2
6	Ce-K, γ 2 esc	90.2
	Average	117.8
	Std.deviation	19.0	16 %

Table 4. keV RoI. Emissions from 9-F-18, and the corresponding escape peak considered in the calculations. The values are reported in descending order. The half-life is of 109.77 minutes, and the decay mode is ε 5 [15].

(i)	Radiation	Egamma (keV)	Yield (Bq-s)-1
1	γ ±	510.9	1.93
2	γ ± esc	481.7
	Average	496.3
	Semi difference	14.6	3 %

Table 4. keV RoI. Emissions from 9-F-18, and the corresponding escape peak considered in the calculations. The values are reported in descending order. The half-life is of 109.77 minutes, and the decay mode is ε 5 [15].

(i)	Radiation	Egamma (keV)	Yield (Bq-s)-1
1	γ ±	510.9	1.93
2	γ ± esc	481.7
	Average	496.3
	Semi difference	14.6	3 %

Table 5. keV RoI. Emissions from 55-Cs-137, and the corresponding escape peaks considered in the calculations. The values are reported in descending order. The half-life is of 30.04 years, and the decay mode is β ^– [15].

(i)	Radiation (i)	Egamma (i) (keV)	Yield (i) (Bq-s)-1
1	γ 2	661.7	0.8510
2	Ce-L, γ 2	655.7	0.0139
3	γ2 esc	632.5
4	Ce-L γ2 esc	626.5
5	Ce-K, γ 2	624.2	0.0766
6	Ce-K γ2 esc	595.0
	Average	632.6
	Std.deviation	24.1	3.8 %

Table 5. keV RoI. Emissions from 55-Cs-137, and the corresponding escape peaks considered in the calculations. The values are reported in descending order. The half-life is of 30.04 years, and the decay mode is β ^– [15].

(i)	Radiation (i)	Egamma (i) (keV)	Yield (i) (Bq-s)-1
1	γ 2	661.7	0.8510
2	Ce-L, γ 2	655.7	0.0139
3	γ2 esc	632.5
4	Ce-L γ2 esc	626.5
5	Ce-K, γ 2	624.2	0.0766
6	Ce-K γ2 esc	595.0
	Average	632.6
	Std.deviation	24.1	3.8 %

3.5. Peak representation

Each peak, relating to the gamma photons emitted by the given source within the assigned RoI, was calculated by using a Gaussian function of the photon energy as described in Eq.(005) :

Counts_(E) = Yield_(E)× EXP ( - 2× ( (E – E_gamma) / FWHM_(E) ) ² ),

(5)

where: both Yield_(E) and E_gamma are from the Table 3, Table 4 and Table 5, while the Full Width at Half Maximum (FWHM_(E) ) is calculated by using the Eq.(6) describing, for a typical NaI detector, the linear decreasing trend of the natural logarithm of the energy resolution as a function of the natural logarithm of the photon energy expressed in units of m₀c² :

Ln R = a× Ln (E/m₀c² ) + b,

(6)

where: m₀ is the value of the rest mass of the electron, c is the value of the speed of light in the vacuum, a and b are parameters assuming the values of the Eq.(7) which have been obtained from the Ref. [10], Figure 10.17:

a = - 0.4332, and b = - 1.9774

(7)

Once the value of LnR for the given energy has been obtained, the value of R is calculated by using the Eq.(8) :

R = Exp(Ln R) = ΔE / E

(8)

from which one can derive the value of the spread at half height ΔE by using the Eq.(9) for the energy of the considered emission:

FWHM(E) = ΔE = R× E .

(9)

From the total peaks (Figure 4, Figure 5 and Figure 6) one can obtain the values of the spreads at half-height and the peak-positions, giving the corresponding values of energy resolution by using the Eq.(8).

4. Results

The Figure 4, Figure 5 and Figure 6 show the responses of a NaI scintillator for the emissions in the RoIs in Table 3, Table 4 and Table 5.

The calculation results shown so far have been iterated with values of percent of escape varying in the range between 0 (no escape) and 50%, by 10% steps. The results of these evaluations are reported in the Figure 7, Figure 8, Figure 9 and Figure 10.The escape-percent values assumed in these evaluations may appear excessive, but nevertheless a NaI detector can be thought of in both monolithic or pixelated form, whose dimension along the z axis (usually referred to as thickness) can vary in real cases from a few millimeters to a few inches. In the cases of a pixelated structure, a thickness from a few millimeters to a few tens of millimeters, as in the case of PET detection blocks [18]. The ranges of thickness variability therefore includes the so-called thin crystals, the use of which is sometimes preferred in cases of need for optimized energy resolution.

But it is quite clear that the definition of thin crystal can only refer to the definition of MFP, Ref. [10], p.53) which, as illustrated in Figure 3 for the NaI, varies considerably as a function of the photon energy, also showing the discontinuities clearly visible in the Figure 3 itself.

As could have been expected from the observation and the inter-comparison of Table 3, Table 4 and Table 5, among the given RoIs, the one most sensitive to the effects of the shifting of the peaks is that at 140 keV which shows, moreover, the highest dispersion of the average referred to the average itself (16%).

Among the given RoIs, the 511 keV one is the “cleanest” since, in addition to the β ⁺annihilation, it only shows the additional contribution from the X-rays escape (3%).

On the other hand, the case of RoI at 662 keV, while showing a high absolute value of the dispersion of the mean, nevertheless shows the higher mean value which produces a smaller relative dispersion (3.9%).

The graphs shown in the Figure 7, Figure 8 and Figure 9 consistently show an increasing trend of the energy resolution ΔE/E, starting from the respective initial energy-dependent value of each RoI. This is justified, on the one hand, by the widening effect due to the overlapping of the escape peak(s) to the one(s) corresponding to the emission(s) and, on the other hand, by the presence of the other emission(s) present itself in the RoI.

Moving on to the values expressed in percent, the final plot of Figure 10 shows the worsening in energy resolution for the different RoIs, that results in an amplified increase the more the escape percentage of the iodine-K X-ray increases and particularly for the lowest-energy RoI.

Figure 10. NaI: comparison of the calculated percent energy resolution trends for the respective energy RoIs, as a function of the percent of the iodine-K X-ray escape. The RoIs are shown in the legend in the same sequence as the curves.

Figure 10. NaI: comparison of the calculated percent energy resolution trends for the respective energy RoIs, as a function of the percent of the iodine-K X-ray escape. The RoIs are shown in the legend in the same sequence as the curves.