SCINTILLATION PROPERTIES OF SELECTED OXIDE MONOCRYSTALS ACTIVATED WITH Ce AND Pr

Andrzej J. Wojtowicz1,[*]), Winicjusz Drozdowski1), and Dariusz Wisniewski1)

Jean-Luc Lefaucheur2), Zbigniew Galazka2), and Zhenhui Gou2)

Tadeusz Lukasiewicz3), and Jaroslaw Kisielewski3)

1) Institute of Physics, N. Copernicus University, ul. Grudziadzka 5, 87-100 Torun, Poland

2) Photonic Materials, Ltd, Strathclyde Business Park, Bellshill, ML4 3BF, Scotland

3) Institute of Electronic Materials Technology, ul. Wolczynska 133, 01-919 Warszawa, Poland

ABSTRACT

In the last 10-15 years there has been a significant effort toward development of new, more efficient and faster materials for detection of ionizing radiation. A growing demand for better scintillator crystals for detection of 511 keV gamma particles has been due mostly to recent advances in modern imaging systems employing positron emitting radionuclides for medical diagnostics in neurology, oncology and cardiology. While older imaging systems were almost exclusively based on BGO and NaI:Tl crystals the new systems, e.g. ECAT Accel, developed by Siemens/CTI, are based on recently discovered and developed LSO (Lu2SiO5:Ce, Ce-activated lutetium oxyorthosilicate) crystals. Interestingly, despite very good properties of LSO, there still is a strong drive toward development of new scintillator crystals that would show even better performance and characteristics.

In this presentation we shall review spectroscopic and scintillator characterization of new complex oxide crystals, namely LSO, LYSO, YAG, LuAP (LuAlO3, lutetium aluminate perovskite) and LuYAP activated with Ce and Pr. The LSO:Ce crystals have been grown by CTI Inc (USA), LYSO:Ce, LuAP:Ce and LuYAP:Ce crystals have been grown by Photonic Materials Ltd, Scotland (PML is the only company providing large LuAP:Ce crystals on a commercial scale), while YAG:Pr and LuAP:Pr crystals have been grown by Institute of Electronic Materials Technology (Poland). All these crystals have been characterized at Institute of Physics, N. Copernicus University (Poland). We will review and compare results of measurements of radioluminescence, VUV spectroscopy, scintillation light yields, scintillation time profiles and low temperature thermoluminescence performed on these crystals. We will demonstrate that all experiments clearly indicate that there is a significant room for improvement of LuAP, LuYAP and YAG. While both Ce-activated LSO and LYSO perform very well, we also note that LuYAP:Ce, LuAP:Ce and YAG:Pr offer some advantages and, after a likely improvement of some parameters, may also present a viable and desired alternative in applications that require high counting rates or better time resolution. Unfortunately, LuAP:Pr, although the fastest among all the materials studied, may be seriously limited in its achievable light yield by inherent physical processes that are responsible for nonradiative quenching of scintillation light in this material.

PACS codes: 29.40.Mc; 72.20.Jv; 78.55.Hx

Key words: scintillators, LSO, LYSO, LuAP, LuYAP, YAG, Ce, Pr, VUV spectroscopy, thermoluminescence

1. Introduction

Although Ce-activated lutetium oxyorthosilicate (LSO, Lu2SiO5:Ce) with its superior scintillator characteristics has been clearly established as a leading scintillator material in the area of medical imaging [1][1] many researchers believe that it is possible to develop a material that would display even better scintillation properties. It is now fairly well established that scintillation light in LSO is produced at two different sites (Ce1 and Ce2) one of which (Ce2) is strongly quenched at room temperature [2][3][2,3]. The dominant Ce1 site emission (unquenched at ambient temperatures) decays with the time constant of about 35ns and any faster components in the scintillation time profiles (STP) of LSO are due to the quenched emission from the Ce2 site. Clearly part of the energy deposited in the LSO host by ionizing radiation is lost to nonradiative processes introduced by inefficient Ce2 sites.

It is likely that in the mixed crystal, (Lu1-x,Yx)2SiO5 (LYSO), in which some of Lu cations are replaced by Y cations, the distribution of energy between the two sites may change and, consequently, the scintillation light yield of LYSO may increase as well. On the other hand disorder typical of the mixed crystal may be responsible for additional losses due e.g. to additional traps, scintillation light scattering and so on.

Another interesting possibility to extend the pool of available dense oxide scintillators offers another Lu-based scintillator material, LuAlO3:Ce (LuAP), an isostructural analog of the well known laser and scintillator material YAlO3 (YAP). Scintillation properties of Ce-activated LuAP, recommended earlier by a number of groups, in a pure perovskite monocrystal phase were first studied and reported by Lempicki and coworkers at the 1994 IEEE Nuclear Science Symposium [4][4]. Importantly, LuAP features only one Ce site and the radiative lifetime of the excited Ce3+ ion in this site and, consequently, the decay time of the dominant scintillation component, is only about 17ns, the shortest scintillation decay time in any known Ce-activated scintillator material. This is mostly due to relatively short emission wavelength (the Ce emissions in LuAP and YAP peak at 360 nm, in LSO at about 420 nm).

Since the two isostructural materials, LuAP and YAP, are closely related, their spectroscopic properties upon Ce-activation, such as emission wavelengths and radiative lifetimes, are nearly the same. This is also true of the mixed material, Lu1-xYxAlO3 (LuYAP, typically at 30% of Y), which is easier to grow than LuAP. Yet the scintillation properties of LuAP and LuYAP differ which comes as no surprise since the scintillation properties of LuAP and YAP differ as well.

Shorter decay times can be obtained only from the Pr- and Nd-activated materials since the 4fn-15d - 4fn emissions from these ions are characterized by even shorter wavelengths. The Pr and Nd emissions typically peak at about 250 and 190 nm, hence, from the wavelength dependence of the electric dipole spontaneous transition rate [5][5], the estimated radiative lifetimes are 8 and 5ns for Pr and Nd, respectively. A scintillator material that would possess such a short scintillation decay time would be of major interest in a number of applications that require high counting rates or a high timing resolution, such as PET and time-of-flight PET.

Although transition energies between Nd3+d and f levels usually fit fluoride bandgaps (typically 10eV), in oxide scintillators, which have lower bandgaps (6-8eV), the only viable alternative, with only a few notable exceptions (e.g. Nd-activated phosphates, [6][6]), to Ce-activation, is Pr-activation.

In this paper we shall review results of a number of R&D projects aimed at development of Ce-activated LYSO, LuAP and LuYAP scintillator crystals at Photonic Materials Ltd, Bellshill, Scotland (PML), and development of Ce- and Pr-activated LuAP, YAP and YAG scintillator crystals at Institute of Electronic Materials Technology, Warsaw, Poland (ITME), in collaboration with N. Copernicus University, Torun, Poland (UMK).

2. Crystals and experimental set-ups

All the samples were cut as pixels (2x2x10 mm) or plates (5x5x1mm) from larger boules grown by Czochralski method at Photonic Materials Ltd, Bellshill, Scotland (PML) and at Institute of Electronic Materials Technology, Warsaw, Poland (ITME).

PML is the only company growing and providing large LuAP:Ce crystals on a commercial scale (diameter up to 50mm, length of a cylindrical body up to 120mm). More information on growth of LYSO, LuAP and LuYAP by PML has already been presented (LYSO, [7][8][7,], LuAP and LuYAP, [8]).

At ITME, the LuAP, YAP and YAG single crystals doped with Ce and Pr ions were grown with use of the oxypuller 05-03 (Cyberstar France). The system consisted of the Ir crucible, 50 mm in diameter, embedded in zirconia grog, the passive iridium afterheater and alumina insulation around a crucible and the afterheater. Raw materials Y2O3, Al2O3 and dopants CeO2 and Pr6O11 (4N) were used. The growth process, carried on in nitrogen atmosphere, was driven by a computer program controlling the weight of a growing crystal. The rate of growth was kept constant at 1mm/h and the rotation rate was 12-15rpm. The crystals measured 20mm in diameter and 50-60 mm length.

The samples were of high optical quality, clear and displayed no visible color and no inclusions.

The radioluminescence spectra were measured using a modified Dron X-ray set-up with an X-ray tube employing a copper anode operated at 35kV, and the Advanced Research System closed-cycle He cooler equipped with a heater driven by a temperature controller (Lakeshore). The light emitted by the sample was collected by quartz optics and passed through the 0.5m SpectraPro Acton monochromator with the 1200 g/mm grating working in the 1-th order.

The luminescence spectra, luminescence excitation spectra and emission time profiles under pulsed VUV and UV synchrotron excitation in the wavelength range 50-335nm were measured using experimental facilities of Superlumi station at Hasylab in Hamburg, Germany [9][9].

Scintillation light yield (LY) was measured at Institute of Physics, N. Copernicus University in Torun, Poland using a typical set-up consisting of radioactive source (Na22, 511 and 1274 keV), a photomultiplier tube (Hamamatsu R2059), Canberra modules, and BGO (Bi4Ge3O12) standard crystal (2x2x10 mm pixel). Since the current drawn from the photocathode of the photomultiplier tube (PMT) is, at the time of scintillation, quite large (one to six thousand photoelectrons in 10-100ns) it is not surprising that the standard method of measuring the relative scintillation light yield by comparing photopeak positions in the energy spectra of a given scintillator material and, say, a BGO standard crystal, is likely to provide results with unacceptably large systematic errors. We have developed a method that allows to extract a correct value of the relative scintillation light yield from a number of spectra measured for different amplifier gains (to exclude the effect of the offset in energy spectra) and for a number of PMT voltages (to correct distortions introduced by a PMT caused, most likely, by space charge in the PMT itself). The method will be described in details later [10][10].

The scintillation time profiles were measured by a classical synchronous photon counting method [11][11] using a Time-to-Pulse Height Converter and other relevant Canberra modules.

Thermoluminescence glow curves were measured using a modified method originally proposed by Bartram et al [12][12] and a Dron based X-ray set-up used for radioluminescence. The temperature controller was programmed to run a closed-cycle He cooler and the heater to accomodate an irradiation step and the TL run. During the irradiation step the temperature of the sample was kept constant at about 10K while the sample was irradiated by an X-ray tube for some time (usually 10 minutes). The light emitted by the sample was collected by quartz optics and passed through a 0.5m SpectraPro Acton monochromator with the 1200 g/mm grating working in the 0-th order producing, at the PMT, a steady state radioluminescence signal which was then sent to computer and recorded (a flat part of the curve). After irradiation the X-ray tube was switched off and the TL run started. During the TL run the temperature of the sample was raised by controller with constant heating rate at about 9K/min, up to 300K. Consequently, the emission released from the sample by heating was recorded by the same set-up and under the same geometrical conditions as in the irradiation step. This method allows to directly measure the fraction of the total energy deposited in the sample that is intercepted by traps.

3. Experimental results and discussion

3.1. Radioluminescence

Unlike photoluminescence spectra which may vary strongly with the excitation wavelength, radioluminescence spectra are determined by a dominant host-to-ion energy transfer mechanism and as such constitute a first basic step in characterization of any scintillator material. Since radioluminescence spectra of most of the materials reported in this paper have already been published, we have chosen to present here only two cases, namely LuAP:Pr and YAG:Pr. As we will demonstrate shortly, the spectra of these two materials show some interesting untrivial differences that are likely to be reflected in their, significantly different, scintillation performance.

In Figs. 1 and 2 we present steady state, X-ray excited radioluminescence spectra of five pixel samples (2x2x10 mm) of LuAP:Pr (0.003, 0.04 and 0.08 mol%Pr), and YAG:Pr (0.1 and 0.3 mol%Pr) measured at 15K (LuAP) and at room temperature (RT, YAG), respectively. The spectra show “host” emissions at the shorter wavelength side of the d-f emission bands that are strongly suppressed for higher Pr concentrations, as observed in Ce-activated materials. The Pr3+d-f emission bands peak at about 245nm in LuAP and at 320nm in YAG which reflects a higher crystal field in YAG pushing down the lower emitting d-level.

Since in YAG Pr3+5d emission lifetimes and intensities at RT are only marginally affected by nonradiative d-f transitions (see [13][13]) and there is no direct energy transfer from host to f-levels, the contribution of f-f emissions to the spectra should be low, as observed in Fig. 2. We also note that the relative intensities of f-f transition lines in two YAG samples (0.1 and 0.3 mol%Pr) are almost the same. The lines are slightly broader in the 0.3mol%Pr sample which is not unusual and may explain slightly lower intensities in this sample by means of energy migration between Pr ions leading to concentration quenching of the f-f emission. However the effect is certainly marginal and the final conclusion is that there is practically no interaction between Pr ions in YAG.

Since the energy gap between d and f Pr3+ levels in LuAP is larger one would expect even less nonradiative d-f relaxation and, consequently, even smaller contribution from f-f emission lines to the total steady state radioluminescence spectrum. As shown in Fig. 1, this is consistent with experimental observations only for the lowest Pr concentration (0.003 mol%Pr). For higher Pr concentrations we clearly observe higher relative intensities of f-f emission lines. Although the effect is hard to quantify (there possibly still is energy migration between Pr ions in the 3P0,1,2 state) it is, nevertheless, obvious that there must be a mechanism by which, in LuAP, but not in YAG, energy is transferred from d levels to f levels for higher Pr concentrations. We note that the d-f transition energy in LuAP, corresponding roughly to 250nm wavelength, is enough to excite the neighboring Pr ion to the 3P0 state assuming that the final state of a given Pr ion is also 3P0. The same is not true of the d-f transition in YAG:Pr as its transition energy corresponds to emission wavelength of 320 nm which is not short enough to have the two interacting Pr ions even in the lower 1D2 final state (the emission from 1D2 peaks at about 620 nm).

The deexcitation process we propose can be described shortly as:

which is a reversed well known process of energy up-conversion. Interestingly such a process of the two-step excitation of d levels has been demonstrated in Pr-doped YAP [14][14].

3.2. VUV spectroscopy

Selective VUV (vacuum ultraviolet) excitation is a powerful tool to study large bandgap scintillation materials as it allows to identify all the different channels of host-to-ion energy transfer that can potentially contribute to scintillation. At the VUV station Superlumi at Hasylab, Hamburg, Germany, it is possible to simultaneously measure at least three different spectra under the wide range of exciting wavelengths between 50 and 335 nm. In each spectral run we have chosen to measure a time-integrated spectrum and two different time-gated spectra. In the time-integrated spectrum the signal was accumulated during the time between two consecutive synchrotron pulses (separated by 192ns) while the time-gated spectra were measured within a 40ns time window triggered by a synchrotron pulse with almost no delay (2-3 ns to reduce stray light) or 150ns delay.

In Figs3 and 4 we present selected time resolved excitation spectra of different emissions from the pixel (2x2x10mm) sample of LYSO grown at PML. The spectra shown in Fig.3 were measured with the emission wavelength set at 400nm to select the Ce3+ emission from the Ce1 site. These are the time-gated spectra with a zero delay gate (“fast” emission, thick solid line) and 150ns delay (“slow” emission, thin solid line). The spectra shown in Fig.4 were measured with emission wavelengths set at 480nm favoring the emission from the Ce2 site (thick solid line) and 320nm corresponding to the so-called “host” emission, characteristic of the undoped crystal but still present to some extent in activated crystals (thin solid line). Both spectra shown in Fig. 4 are time-integrated with the signal accumulated within the 192ns time window between the consecutive synchrotron pulses.

Clearly the excitation spectra of the Ce1 emission strongly change with the delay of the time gate. The “slow” emission, represented by a “delayed” spectrum, is excited more efficiently in a narrow wavelength range around the bandgap energy (160 to 180nm) while excitation efficiency at the wavelengths corresponding to VUV (50 to 150nm) and Ce3+f-d absorption bands (240 to 335nm) is clearly lower than in the “zero-delay” spectrum representing the “fast” emission. Nevertheless, some slow Ce3+ emission produced (mostly at Ce1 sites) under the VUV excitation (compare also Fig.7) strongly suggests that there are some slow energy transfer channels from the host to the Ce3+ ions most likely related to traps that intercept and then release charge carriers with some, trap dependent delay. Delayed recombination of carriers released from traps is likely to produce slower components in VUV (or X-, or gamma-ray) excited pulses of emission light.

The structure in VUV below 140nm reflects the spectral characteristics of the primary monochromator equipped with the Al grating, and does not correspond to any real physical processes. Nevertheless the high signal at these wavelengths is indicative of strong sensitivity of a given emission to the VUV excitation. Since the VUV photons provide over the bandgap excitation of the host material generating free electron-hole pairs, it also points out that Ce-ions must be efficient recombination centers.

Note that because of the upper wavelength limit (335nm) imposed by the detection monochromator the lowest energy f-d absorption band in the Ce1 excitation spectrum is missing and there are only two bands at 260 and 286nm corresponding to the higher energy d-levels split by a lower symmetry crystal field component.

The time-integrated excitation spectrum measured at 320nm emission wavelength corresponding to the “host” emission band and shown in Fig.4, shows no contribution from the Ce3+f-d bands. A steep increase at wavelengths corresponding to the energy bandgap of the material and, consequently, the large Stokes shift between emission and excitation wavelengths, are typical of emissions due to decay of self- or defect-trapped excitons (“host” emissions).

Finally the time-integrated excitation spectrum measured at 480nm shows the structure that was previously assigned to the Ce2 center [2,3]. Note in particular that the 325nm peak is very well suited to maximize the contribution of the Ce2 emission since the Ce1 excitation spectrum at this wavelength shows a local minimum.