Ultraviolet and blue cathodoluminescence from cubic Y2O3 and Y2O3:Eu3+ generated in a transmission electron microscope

Herein we describe the investigation of cubic spherical submicron particles of non-doped Y2O3 and Y2O3 doped with Eu 3+ in a transmission electron microscope (TEM) equipped with a spectrometer to detect cathodoluminescence from individual particles. Each submicron particle was made up of nanometre sized crystals. We found that these crystals showed a broad emission band at 353 nm upon bombardment with 200 keV or 80 keV electrons. Upon increasing the Eu concentration from 0 to 2 mol% this UV/blue emission was gradually quenched: at Eu concentrations 42 mol% no UV/blue emission was detected, only the well-known cathodoluminescence (CL) spectrum of Y2O3:Eu 3+ could be recorded. This UV/blue emission has been attributed to the intrinsic luminescence of Y2O3 caused by self-trapped excitons. We found that the UV/blue luminescence was strongly temperature dependent and that the trap depth of the self-trapped excitons was 0.14 eV. The ratios of the spectral radiances of D1 FJ and D0 FJ (J = 0, 1. . .6) Eu 3+ transitions in the CL-TEM spectra of Y2O3:Eu 3+ at low Eu concentrations was about a factor of 10 larger than those recorded at 15 keV. This phenomenon has been explained by absorption of the intrinsic luminescence of Y2O3 by Eu .


Introduction
Yttrium oxide (Y 2 O 3 :Eu 3+ ) is a well-known phosphor from its application as the red emitting phosphor in fluorescent lamps and projection cathode ray tubes (CRTs). [1][2][3] Due to the industrial applications and the attractive properties of Y 2 O 3 as host lattice, Y 2 O 3 doped with Eu 3+ and other rare earth ions has been well studied and documented. The expected application of Y 2 O 3 :Eu 3+ in low-voltage field emission displays has stimulated intensive research in the synthesis and characterisation of nanosized phosphors, recently reviewed by Li and Lin. 4 In our laboratory we have studied the enhancement of cathodoluminescence by applying double layers of ZnO:Zn and Y 2 O 3 :Eu 3+ . 5 Non-doped Y 2 O 3 also yields photoluminescence when excited with UV radiation. Although this has been known for more than 50 years, 6 the luminescence of non-doped Y 2 O 3 has not been studied in much detail until recently. The reason for this recent interest is the possible application of non-doped Y 2 O 3 and Y 2 O 3 doped with Eu 3+ , Tb 3+ , Nd 3+ or Tm 3+ as a scintillation material. [7][8][9][10][11][12][13][14][15] When non-doped Y 2 O 3 is excited by a-particles, 7 UV-radiation (207 nm) 8 or X-rays, [9][10][11][12][13] a broad luminescence band is observed between 340 nm and 500 nm. Table 1 summarizes the wavelength at maximum (l max ) of this broad luminescence band recorded in the literature. The values for l max listed in Table 1 vary substantially. This erratic behaviour of the luminescence of Y 2 O 3 elicited the following statement from Fukabori et al. 7 ''Light yields of Y 2 O 3 ceramics are different from specimen to specimen. Nature of this phenomenon is not clear yet''. Konrad et al. 8 found that the size of nanoparticles plays an important role and might explain a variation of about 30 nm in l max ; however, this does not explain the variation shown in Table 1. Fukabori et al. 7 found a relation between the scintillation light output and crystal size in their Y 2 O 3 samples: the larger the crystallites, the higher the light output. a-particles 7 410 X-ray 9 364 X-ray 10 500 X-ray 11 385 X-ray 13 Wood and Hayes 10 reported that the 364 nm emission band is particularly strong upon X-ray excitation at 1.6 K; Tanner et al. 12 also found that the 344 nm band in Y 2 O 3 :Eu 3+ with either 0.1 mol% or 1 mol% Eu 3+ is strongly temperature dependent.
Besides the position of l max there is another interesting characteristic of the intrinsic luminescence of Y 2 O 3 : the quenching of this luminescence upon doping with rare earth ions. This was first noticed by Wickersheim and Lefever 6 and later by Jacobsohn et al. 9 and Tanner et al. 12 14 it may be concluded that Tb 3+ is more effectively quenching the intrinsic Y 2 O 3 luminescence than Eu 3+ . Ato et al. 18 reported a strong thermo-luminescence peak of Y 2 O 3 :Eu 3+ upon irradiation with g-rays from a Co 60 source at low and high dose rates. They did not suggest at that time that Eu 3+ was being reduced by g-rays: this mechanism was suggested nine years later by Ozawa, 19 a co-author of Ato. 18 The intrinsic luminescence of Y 2 O 3 has been explained by three different mechanisms: (1) oxygen vacancies, 9 (2) selftrapped excitons (STE) 7,8,10,12,14,20,21 and (3) ligand-to-metal charge-transfer involving the empty 3d orbitals of the Y 3+ ion. 13 The latter two mechanisms are more plausible, because luminescence due to STEs 14 or charge-transfer generates broad bands, while Hayes et al. 14 also concluded that the STE-mechanism is likely from their measurements of optical detection of magnetic resonance (ODMR). No explanation is apparent in the literature for the quenching of the intrinsic luminescence of Y 2 O 3 upon doping with rare earth ions.
The current density in a projection CRT at normal operating conditions has a maximum at about 0.25 A m À2 . There have been no reports of finding blue CL from Y 2 O 3 :Eu 3+ in such CRTs. In a scanning electron microscope the beam current at the specimen is normally between 0.01 nA and 1 nA, dependent on the settings of the condenser lenses, 22 and the spot size is often of the order of 1.5 nm, leading to current densities between 6 Â 10 6 A cm À2 and 6 Â 10 8 A cm À2 , which are 7 to 9 orders of magnitude larger than in a CRT. If the electron transmission at 200 keV is about 99.99% in 200 nm crystals of Y 2 O 3 , then the effective current density is still 3 to 5 orders of magnitude higher. This difference in effective current density could explain why blue emission was never observed before or previously reported. Another reason is the quenching of this emission at Eu 3+ concentrations greater than 2 mol% (which is typical for CRTs and fluorescent lamps red phosphors). In our laboratories we are currently involved in cathodoluminescence (CL) and photoluminescence (PL) studies of nanosized and submicron Y 2 O 3 doped with Eu 3+ ,Tb 3+ and other rare earth ions. [23][24][25][26] Yttrium oxide is a stable compound and when doped with Eu 3+ , nobody (to our knowledge) has observed broad emission bands between 350 and 500 nm as mentioned above. It was therefore a surprise when investigating Y 2 O 3 :Eu 3+ samples in our transmission electron microscope (TEM) equipped with an optical spectrometer that we observed a broad emission band at 350 nm. Preliminary results of this observation have been presented at the 22nd International Display Workshops in Japan. 27 Herein we describe further measurements and present an analysis of the intrinsic emission of non-doped Y 2 O 3 and Y 2 O 3 :Eu 3+ in the electron beam of a TEM.

Materials and synthesis
Yttrium oxide (99.99%, Ampere Industrie, France) and europium(III) oxide (99.99%, Neo Performance Materials, UK) were used to prepare the europium-doped yttrium nitrate stock solutions. Urea, nitric acid, oxalic acid and isopropanol (IPA) were supplied by (Fisher Scientific, UK). All chemicals were used as received.
The synthesis of sub-micrometre spherical Y 2 O 3 :Eu 3+ precursor particles via a homogeneous precipitation route utilising the hydrothermal decomposition of urea method, followed by annealing the precursor particles at 980 1C resulting in cubic Y 2 O 3 :Eu 3+ has been described extensively in our earlier work. [23][24][25][26] The concentration of Eu 3+ in Y 2 O 3 was adjusted to 0.1, 0.5, 1, 1.5, 4, 20, 60 and 100 mol%. For comparison reasons three samples of micrometre sized Y 2 O 3 :Eu 3+ precursor particles were prepared via an oxalate precipitation route. 1,3 The first sample was made by co-precipitation of Eu 3+ -oxalate (2%) and Y 3+ -oxalate. The second and third samples were made by separately precipitating Y 3+oxalate and Eu 3+ -oxalate (6% and 2%) and then slurry-mixing (SM) these precipitates before annealing. These samples were then annealed for four hours at 980 1C in air.

Transmission electron microscope
The submicron spherical Y 2 O 3 and Y 2 O 3 :Eu 3+ samples were investigated with a TEM (2100F, JEOL, Japan) equipped with a Schottky-type field emission gun. When operated in scanning mode (STEM), the spot size of the e-beam at the specimen was adjusted to 0.2 nm or 1.5 nm. Initial work demonstrated the need to reduce the X-rays in the column generated from the condenser lens aperture, which were found to significantly contribute to disperse excitation of phosphor samples. These X-rays excited the phosphor and caused the emission of visible light when the electron beam was not on the sample, leading to unwanted interference and a loss of resolution. To reduce this X-ray excitation of the sample, a hard X-ray aperture was inserted into the column, which reduced the background noise in CL imaging and spectroscopy modes considerably. The TEM was equipped with a Vulcant CL detector, Gatan, USA, for imaging and spectroscopic purposes. This system used a Czerny-Turner spectrometer with back-illuminated CCD and a grating with 1200 lines per mm (blazed at 500 nm) for collection of CL emission spectra. Light was collected from the sample using a mirror above and below the sample, which enabled a solid angle of about 5 sr, which is almost half of a sphere. This high solid angle made light collection highly efficient and enabled the collection of CL at low intensity. Unfortunately, the cooled detector of this spectrometer did not allow the recording of spectra at l o 380 nm. In the subsequent sections spectra recorded with this spectrometer will be represented at l 4 400 nm. Spectra between 200 and 400 nm were recorded with the Black Comet spectrometer of StellarNet Inc. (USA) for undoped Y 2 O 3 . This spectrometer had an uncooled detector and the spectra were therefore much noisier.
By collecting the visible light with the Vulcan system simultaneously with JEOL's high-angle annular dark-field (HAADF) detector, it was possible to observe the visible light that was emitted from the particles. A small cryostat connected to the sample holder enabled cooling of the samples in the TEM down to 102 K (À171 1C); adjustment of the sample temperature anywhere between 102 K and 303 K could be made. A Gatan electron energy loss spectrometer (EELS) was used to map the position of europium ions at the surface of the nanocrystals.

X-ray powder diffraction
The crystalline phases of the products were determined by X-ray powder diffraction (XRPD) using a Bruker D8 Advance X-ray powder diffractometer fitted with a nickel-filtered copper source and a LynxEyet silicon strip detector. Data were recorded from 5 to 100 2y degrees at 25 1C. The diffractometer was previously calibrated using an aluminium oxide line position standard from Bruker and a LaB 6 NIST SRM 660a line profile standard. Diffractograms were collected using the annealed powders in a conventional holder. The emission of the nickel filtered Cu source and hence the instrumental line broadening was determined by fitting the NIST standard using Bruker Topas version 3. Phases in the combusted products were identified from the XRD patterns by peak search matching using the ICCD PDF-2 data files.

Results
Transmission electron microscope analysis Fig. 1a is a TEM image of the urea-precipitated cubic spherical submicron Y 2 O 3 :Eu 3+ phosphor particles with diameters between 200 nm to 300 nm. Fig. 1b presents a TEM image of a single particle, which is composed of a number of tessellated nanocrystals from 40 nm to 80 nm.
In Fig. 1c the oxalate-precipitated cubic micrometre sized Y 2 O 3 :Eu 3+ phosphor particles are presented. These have an irregular morphology with a size range between 1 and 3 micrometres.

X-ray diffraction
X-ray diffraction was only used for the analysis of the oxalateprecipitated samples of Y 2 O 3 :Eu 3+ , because urea-precipitated samples of Y 2 O 3 and Y 2 O 3 :Eu 3+ after annealing at 980 1C for 4 hours are known to consist of the cubic phase for 100%. [23][24][25][26] Fig. 2 shows the diffractogram for oxalate-precipitated Y 2 O 3 :Eu 3+ . This diagram proved that this material after annealing for 4 hours at 980 1C also consisted for 499% of the cubic phase.  (1), which is represented by a Gaussian profile (3) corrected for the background (2). The Gaussian profile has been fitted to the spectrum with a least squares algorithm using Microsoft's Excel solver.

UV/blue emission band
By averaging the values for l max and the full width at half maximum (FWHM) from spectra recorded at various temperatures, we determined that l max = 353 nm (28 300 cm À1 ) and FWHM = 5735 cm À1 . This value of l max is identical to the value published by Fukabori et al. 7 and close to the values published by Konrad et al., 8 Wood and Hayes 10 and Tanner et al. 12 Fukabori et al. found that in some of their samples the emission extended much further in the visible region. This has not been confirmed in our measurements, neither in those of the other workers. [8][9][10]12,13 Nevertheless, the broad UV emission band represented in Fig. 4 extends substantially into the visible part of electromagnetic radiation. At 405 nm the spectral radiance is almost 5 times smaller than at 353 nm; however, for comparing the intrinsic Y 2 O 3 luminescence with the Eu 3+ emission lines for the doped samples, we used the Gatan spectrometer with the cutoff at about 390 nm, basically because of the detector limitation of the StellarNet spectrometer. Fig. 4 shows how we were able to turn the red-emitting phosphor Y 2 O 3 :Eu 3+ into a UV/blue-emitting phosphor in the TEM: the radiance of the UV/blue luminescence is dwarfing the (area) radiance of the well-known emission lines of Eu 3+ in Y 2 O 3 at 0.1 mol% Eu 3+ . Important conditions for observing the broad UV luminescence are: low temperature of the sample: À171 1C, large spot size of the e-beam of 1.5 nm, high beam current by maximising the condenser lens aperture and a specimen thickness of at least 100 nm. The energy of the electron beam does not seem paramount: we observed the UV/blue luminescence both at 200 keV at 80 keV. In Fig. 4A the sharp emission lines of Eu 3+ are not present: whereas the strongest transition in Fig. 4B-D is that due to the 5 D 0 -7 F 2 Eu 3+ transition at 611 nm. The vertical scales of the graphs in Fig. 4 cannot be compared due to the different integration times during spectra recording and different specimen thicknesses. It can be seen that the spectral radiance (normalised to the spectral radiance at 611 nm) of the blue emission at 400 nm decreases strongly when the concentration of Eu 3+ is increased from 0.1 to 1 mol%. At Eu 3+ concentrations Z2 mol% in urea-precipitated Y 2 O 3 :Eu 3+ we could not detect any UV/blue emission at 400 nm. Beside the strong UV/blue emission at 400 nm, two very weak long wavelength bands can be observed (inset of Fig. 4A), one at about 675 nm and the other at about 770 nm. Electron bombardment of the carbon-coated Cu-grid without Y 2 O 3 :Eu 3+ particles did not show any CL; hence, interference from the Cu-grid holder material can be excluded.
The luminescence spectrum between 400 nm and 500 nm in Fig. 4A (and the weak long wavelength bands) did not noticeably change its shape upon reducing the energy of the electron beam from 200 keV to 80 keV. This therefore excludes Cherenkov radiation being the origin of the observed emission bands at   353 nm, 675 nm and 770 nm. 28 At 400 nm the refractive index of Y 2 O 3 is 1.98: to calculate the refractive index of Y 2 O 3 as a function of l, use was made of the dispersion formula given by Nigara. 29 The condition for Cherenkov radiation in a medium with a refractive index of 1.98 is satisfied when the electron velocity 41.52 Â 10 8 m s À1 , which starts at an electron energy of 80 keV. So, at 80 keV there cannot be Cherenkov radiation in Y 2 O 3 at wavelengths 4B405 nm. From this straightforward calculation it can be concluded that any light at wavelengths 4405 nm that is emitted upon bombarding non-doped, transparent Y 2 O 3 crystals with electrons at 80 keV cannot be ascribed to Cherenkov radiation. Since the position and shape of the blue emission did not change upon increasing the electron energy to 200 keV, we can reasonably exclude Cherenkov radiation for 200 keV electrons as well. Additional evidence for this conclusion is the strong temperature dependence of the UV/blue and red emission, as we shall discuss in the following paragraphs. Cherenkov radiation is virtually not affected by temperature as long as the refractive index and the density of the specimen do not change substantially. In ref. 27 it was supposed that the blue emission of Y 2 O 3 :Eu 3+ at 400 nm was caused by reduction of Eu 3+ to Eu 2+ . From Fig. 4A it must be concluded that this is not the case and that the blue emission in all graphs of Fig. 4 is intrinsic luminescence of Y 2 O 3 . In other words, it is concluded that the UV/blue luminescence observed in the TEM is identical to the blue/UV emission from undoped Y 2 O 3 and Y 2 O 3 doped with small amounts of rare earth ions upon excitation by other sources of ionizing radiation. [7][8][9][10][11][12][13][14][15][16] In order to quantify the quenching of the intrinsic luminescence of Y 2 O 3 upon increasing the Eu 3+ concentration we define a quenching factor Z: where SR 405 is the spectral radiation of the UV/blue emission at 405 nm and SR 611 is the spectral radiance of the 5 D 0 -7 F 2 Eu 3+ transition at 611 nm. For non-doped Y 2 O 3 Z = 1 and for Eu 3+ concentrations 42 mol%, Z = 0. Fig. 4B-D refer to Y 2 O 3 doped with 0.1, 0.5 and 1 mol% Eu 3+ . Apart from the diminishing intrinsic luminescence of the host material there is another interesting phenomenon, viz. the changing ratios between the spectral radiance of the strongest Eu 3+ transition at 611 nm and the spectral radiances of Eu 3+ transitions at l o 611 nm. Furthermore, these latter transitions are about two orders of magnitude stronger than the corresponding transitions in the photoluminescence (PL) spectrum of Y 2 O 3 :Eu 3+ as represented in Ozawa's book on page 165. 3 We shall discuss these phenomena together with the intrinsic luminescence of Y 2 O 3 in the next section. Fig. 5 is a HAADF image of urea-precipitated Y 2 O 3 :Eu 3+ (0.1 mol% Eu 3+ ) at À168 1C and 200 keV. The image illustrates the effect of positioning of the e-beam on Z. The square denoted by ''Spatial Drift'' indicates the image that was used to compensate the thermal drift of the sample holder during cooling and warming up. This feature enabled a stable position of the electron beam on the nanocrystal during recording of the spectra, which required more than 2 minutes in some cases. The sites SI.1 to SI.5 in Fig. 5 refer to spots with different particle thickness: SI.1 is a very small particle, whereas SI.2 and SI.3 refer to spots with two particles on top of each other. This is reflected in the rather high value for the Eu 3+ emission. In the other urea-precipitated samples of Y 2 O 3 :Eu 3+ containing larger Eu 3+ concentrations, the variation of the quenching factor Z was less than a factor of two. Since some Y 2 O 3 :Eu 3+ particles are partially hollow, it is impossible to determine a relation between R 405 and R 611 and specimen thickness from Fig. 5. It should be kept in mind that the radiance R b of the Y 2 O 3 emission band, defined as where bÀ and b+ are the integration limits, is outstripping the radiance of the 5 D 0 -7 F 2 Eu 3+ transition at 611 nm at low Eu 3+ concentrations, because the UV emission band is much broader than the 611 nm peak. Fig. 6 shows the effect of temperature on the spectral radiance of the intrinsic emission of undoped Y 2 O 3 : the lower the temperature, the stronger the UV/blue luminescence. The drift corrector facility, indicated in Fig. 5, guaranteed that the spectra shown in Fig. 6 were recorded at the same spot. Since the time for recording the spectra (integration time) was also kept constant, the spectra for different temperatures can be compared directly: i.e. there is no effect of thickness. The intrinsic blue luminescence in Y 2 O 3 :Eu 3+ also increased by more than a factor of 10 upon decreasing the temperature from 31 1C to À171 1C.
The two long wavelength bands shown in the inset of Fig. 6 are much weaker than the UV/blue emission and are also stronger at À172 1C than at 30 1C. It can be seen that the temperature behaviour of the band at 675 nm deviates from that of the other band at 770 nm and the UV/blue band. Upon excitation of Y 2 O 3 or Y 2 O 3 :Eu 3+ by X-rays or a-particles no emission was observed from these bands by other workers. 7,[9][10][11][13][14][15][16] These bands also disappeared when the Eu 3+ concentration was increased. The origin of these two bands is unknown. Fig. 7 is an Arrhenius plot of the spectral radiance (SR) measured at 353 nm with the StellarNet spectrometer (data points 1) and SRs measured at 405 nm with the Gatan  2 and 3). The data points of (1) refer to undoped Y 2 O 3 and 200 keV beam energy, the data points of (2) refer also to undoped Y 2 O 3 but at 80 keV and the data points of (3) refer to Y 2 O 3 :Eu 3+ with 0.1 mol% Eu 3+ and 200 keV. The drift corrector facility enabled us to stay at the same point during recording of the spectra and after changing the temperature. In some cases a tiny shift was applied to deal with beam degradation. However, the measured effect of beam degradation on spectral radiance was o2%, which was smaller than the noise level in the spectra, especially in the StellarNet spectra, as shown in Fig. 3. The curves in Fig. 7 have been fitted to the data points with eqn (4) that describes the temperaturedependent effect of STEs. This will be discussed in the next section. The trap depth that describes the experimental data for non-doped Y 2 O 3 is 0.13 eV and it is 0. 16 Fig. 7. In view of the spread in the data we consider that the trap depth of the STEs in undoped and doped Y 2 O 3 is identical and has a value of 0.14 eV. In Fig. 8 we have summarized the quenching factor Z (for the blue band) as a function of Eu 3+ concentration for all samples investigated in the TEM. There are three types of samples collected in Fig. 8, viz. urea-precipitated (without a number), oxalate co-precipitated (no. 1) and SM-oxalate precipitated (no. 2). As mentioned above, blue luminescence could not be observed for the urea-precipitated samples at Eu 3+ concentrations 41.5 mol%, whereas the SM-oxalate samples showed very large blue luminescence for 2 and 6 mol% Eu 3+ . The spread in Z was about a factor of 3 in the SM-oxalate samples; in the urea-precipitated samples it was less. Fig. 8 illustrates the different behaviour of the SM-oxalateprecipitated and urea-precipitated samples, whereas the oxalate co-precipitated sample (point 1 in Fig. 8) with molecular mixing of the Eu 3+ and Y 3+ ions indicates that this sample behaves as the urea-precipitated samples. The deviating behaviour of the SM-oxalate-precipitated sample is explained by its inhomogeneity, in which Eu 3+ -rich areas in the crystals are alternating with areas with very low Eu 3+ concentration. The penetration depth of 200 keV electrons in crystalline Y 2 O 3 is about 75 mm, 5,30 which is much larger than the size of the oxalate-precipitated crystals. The Eu 3+ rich areas in the crystal hit by the e-beam do not contribute to the build-up of the blue band because of concentration quenching, whereas the areas with very low Eu 3+ concentration take care of the strong blue band signal. This hypothesis was confirmed by the very low Eu 3+ signal in the EELS (not shown) at various spots. From this consideration it can be concluded that the SM-oxalate precipitated sample should be inserted at a much lower concentration in Fig. 8. Since the effective Eu 3+ concentration is unknown, the best we could do is plotting the result at the as-made Eu 3+ concentration.
Finally we would like to mention that we also observed blue luminescence in the TEM when bombarding undoped monoclinic Y 2 O 3 crystals with 200 keV electrons at 1.5 nm spot size. The monoclinic material was unstable under these electron bombardment conditions, decomposing partly into the more stable cubic form, and therefore we do not reproduce any results of these measurements in this article.   Other points refer to urea-precipitated samples. The abscissa is quasilogarithmic, because the point with 0 mol% Eu 3+ dope has been indicated.

Discussion of blue luminescence in Y 2 O 3 and transition ratios
When an electron beam of 200 keV hits an Y 2 O 3 crystal it produces besides defects, backscattered and secondary electrons, X-rays and holes by inelastic scattering processes. Electrons and holes can combine to form free excitons and STEs. Herein we adopt the model that STEs are responsible for the blue luminescence in undoped Y 2 O 3 and Y 2 O 3 :Eu 3+ as indicated in the literature. 7,8,10,12,14,20,21 According to Mikhailik and Kraus 31 and Blasse 32 the radiance B of a scintillator upon excitation can be written as: where a is a proportionality constant, N e-h is the number of electron-hole pairs that is generated by the electron beam inside the crystal, S is the probability of transferring energy from an e-h pair to a luminescent centre, e.g. a Eu 3+ ion, and Q is the luminescence quantum efficiency. Undoped Y 2 O 3 is a so-called self-activated or intrinsic scintillator, 31 in which S is 1 by definition. The intrinsic luminescence in non-doped Y 2 O 3 is considered to come from O-2p levels and Y-3d levels after an STE has combined with a luminescent centre in Y 2 O 3 . 11 The quantum efficiency Q in eqn (3) can be described by Mott's equation on thermal quenching: 31 where P and C are constants to be fitted, E A is the activation or trap energy, k is Boltzmann's constant and T is the absolute temperature. In Fig. 8 (3), which is 0 for undoped Y 2 O 3 and has a maximum value, albeit o 1, for a dopant concentration of 2 mol% and larger.
As mentioned above, Fig. 4 shows that the spectral radiance of Eu 3+ transitions at l o 580 nm decreases in the sequence Fig. 4B-D. We have plotted this behaviour for some transitions in Fig. 9. Fig. 9 represents the ratios R 1 and R 2 , which are the ratios of the spectral radiances 5 D 1 -7 F 1 (533 nm) and 5 D 0 -7 F 2 (611 nm) and that of 5 D 0 -7 F 4 (713 nm) and 5 D 0 -7 F 2 (611 nm) respectively. It can be seen that R 1 decreases almost by a factor of 10 upon increasing the dopant concentration from 0.1 mol% to 4%, whereas R 2 is constant, both at cryogenic and room temperature. The latter ratio is constant, because it refers to 5 D 0 transitions. The transitions of the 5 D 2 and 5 D 3 states to the 7 F J manifold show a similar behaviour as the 5 D 1 -7 F 1 transition; however, due to the lower spectral radiances the error bar in the graphs is larger and therefore these will not be reproduced here. The partial concentration quenching of the 5 D 1 transition of the Eu 3+ ion represented in Fig. 9 is well known and it has been described by Blasse and Grabmaier 2 and Klaassen et al. 33 Blasse and Grabmaier explained the behaviour of R 1 in terms of the following cross relaxation in Eu 3+ : This cross relaxation is facilitated by the ten times faster decay rate from 5 D 1 levels as that from 5 D 0 levels as indicated by Klaassen et al. 33 These latter authors described the partial quenching of 5 D J (for J 4 0) transitions in terms of radiative decay time and the rate of non-radiative transfer processes. However, neither cross relaxation nor rate constants explain why R 1 is so large in Fig. 9 at low Eu 3+ concentrations. For a qualitative explanation we shall make use of the energy diagram depicted in Fig. 10. Most levels indicated in Fig. 10 have been reported by Wen et al., 34 only the lowest energy level of the manifolds is indicated. The energy level of the intrinsic Y 2 O 3 emission band has been centred at 28 300 cm À1 above the ground level 7 F 0 of Eu 3+ , based on the spectrum depicted in Fig. 4. The level of the STE Y 2 O 3 emission band overlaps with several Eu 3+ levels; hence, it seems obvious that energy can radiationless be transferred from an STE in Y 2 O 3 to the 5 D 4 level of Eu 3+ , which is only slightly lower. This process, which takes place at Eu 3+ concentrations 42 mol%, is indicated by arrow 4. From the 5 D 4 level energy will radiationless trickle down to the 5 D 3 , 5 D 2 , 5 D 1 and 5 D 0 levels, indicated by arrows 7. As mentioned above, the decay times of transitions from 5 D J levels with J 4 0 are about 10 times shorter than these from 5 D 0 , 33 which enhances efficient energy transfer and hence rather fast quenching of the Y 2 O 3 emission upon increasing the Eu 3+ concentration. 2,35 In terms of the probability S in eqn (3) it means that energy from an e-h pair is transferred to a Eu 3+ ion.
At low Eu 3+ concentrations o2 mol% the strong intrinsic emission at 28 300 cm À1 may be absorbed by Eu 3+ ions in the lattice: this process is indicated by arrows (5) and (6) in Fig. 10. It has been indicated in this figure that the Y 2 O 3 UV radiation is Fig. 9 Ratio of spectral radiance between 5 D 1 -7 F 1 (533 nm) and 5 D 0 -7 F 2 (611 nm): R 1 and ratio of spectral radiance between 5 D 0 -7 F 4 (713 nm) and 5 D 0 -7 F 2 (611 nm): R 2 . For R 1 the measurements at À170 1C (cold) and room temperature (RT) gave different results.
absorbed from the ground level 7 F 0 ; hence, the 5 D 4 Eu 3+ level will be populated more densely. As mentioned above, from 5 D 4 energy will trickle down to the lower 5 D J levels. The process indicated by arrows (5) and (6) will stop when the intrinsic luminescence of Y 2 O 3 has quenched at large Eu 3+ concentration. The latter process explains the high value of R 1 in Fig. 9 and similarly the rather strong 5 D 1 -7 F J and 5 D 2 -7 F J transitions in Fig. 4B-D. The foregoing consideration also explains why the ratio R 1 in Fig. 9 at low temperature is higher than at room temperature: the process indicated by the arrows (5) and (6) is more dominant at low temperature because of the much stronger UV luminescence of Y 2 O 3 .

Conclusions
In the preceding sections we have described the cathodoluminescence of undoped Y 2 O 3 and Y 2 O 3 :Eu 3+ crystals in the TEM by high-energy electron bombardment. At low temperatures we observed a broad emission band at about 353 nm, which has been ascribed to the migration of excitons to luminescence centres in Y 2 O 3 . We found that the UV/blue luminescence is strongly temperature dependent and that at concentrations 42 mol% Eu 3+ no blue light could be detected.
The temperature behaviour of the intrinsic luminescence of Y 2 O 3 has been explained with a model for the self-trapped excitons. The depth of these traps was found to be 0.14 eV. The concentration dependence of the UV/blue luminescence has been explained by the good overlap between the level of the blue Y 2 O 3 band and the 5 D 4 level of the Eu 3+ ion. The strong radiance of 5 D 1 -7 F J and 5 D 2 -7 F J Eu 3+ transitions in the spectra excited at 200 keV in Y 2 O 3 :Eu 3+ with low Eu 3+ concentration has been explained by absorption of the intrinsic Y 2 O 3 radiation by Eu 3+ ions. Although we have presented a detailed explanation of the concentration and temperature behaviour of the UV/blue emission, we have also generated new questions. The most important is about the nature of the red emission bands at 675 nm and 770 nm.
Based on the results described in this article we would like to make a suggestion to other scientists working with electron microscopes in the field of biomedical imaging with phosphors. This fast growing technology has recently been reviewed by Gai et al. 36 Since undoped Y 2 O 3 or Y 2 O 3 :Eu 3+ with low Eu 3+ concentration will manifest the strong intrinsic luminescence of Y 2 O 3 in the e-beam of a TEM (and likely also in a SEM) at low temperatures, we think that this work opens new perspectives for labelling (biological) samples with nanocrystals of Y 2 O 3 :Eu 3+ .  1) and (2) refer to the 5 D 0 -7 F 2 , 5 D 1 -7 F 1 and 5 D 0 -7 F 4 Eu 3+ transitions respectively, while (4) indicates the radiationless energy transfer from Y 2 O 3 to Eu 3+ . For arrows (5), (6) and (7) we refer to text.