Luminescence properties of europium titanate thin films grown by atomic layer deposition

Luminescent europium titanium oxides (EuxTiyOz) have been deposited as thin films by atomic layer deposition (ALD) in the temperature range 225 to 375 C with control of the stoichiometry from pure TiO2 to pure Eu2O3, and their luminescent properties have been characterized. The current work demonstrates the feasibility of tailoring the luminescent properties of these films by varying ALD deposition parameters. The highest luminescence efficiency was observed for the amorphous phase while it decreases steadily with increasing annealing temperature up to 700 C due to the formation of crystalline Eu2Ti2O7. The limiting factor for luminescence at low Eu contents is the low concentration of emitting Eu atoms, while the luminescence at high Eu contents is limited by a low UV absorption due to the low Ti content in these samples. The precursors used for the Ti–O deposition were TiCl4 and H2O, while Eu(thd)3 (thd 1⁄4 2,2,6,6-tetramethyl-3,5-heptanedione) and ozone were used for the deposition of Eu–O. The films have been characterized by spectroscopic ellipsometry, photoluminescence, X-ray diffraction, X-ray fluorescence and atomic force microscopy. The depositions gave smooth (RMS roughness < 0.4 nm) and amorphous films for 30% and more Eu–O cycles (9.5 cationic% Eu), while 10 and 20% gave rough and partly crystalline films. The refractive index, extinction coefficient and luminescence efficiency are shown to vary continuously with Eu content for 30% Eu–O cycles or more, while the shape of the luminescence spectrum is independent of Eu content in the same range. For the 10 and 20% Eu–O depositions, we observed a gradual change in the luminescence spectra, coinciding with an increase of the proportion of anatase phase with decreasing Eu content.


Introduction
Luminescent materials nd many applications, including phosphors for lighting and TV screens, lasers, markers for security tags and in biology, optical ampliers and for energy applications like solar cells.Silicon based solar cells 1,2 as well as thin lm silicon cells 3 have been shown to benet from thin lm phosphor materials.Models show that in particular CdTe based solar cells would benet from such lms. 4For luminescent materials, the crystal structure, texture, purity, defects, distribution of atoms and in particular the site symmetry of the luminescent center affects the luminescence efficiency of the material.In general, large single crystals are preferable to polycrystalline samples in order to reduce the surface defects and scattering, while thin lm processes is helpful to tailor the texture and distribution of dopants.
Atomic layer deposition offers unique possibilities in designing new materials and thin lms.The low temperature layer-by-layer growth allows nanostructuring and next-neighbor distribution of atoms not easily achievable by other methods.The low deposition temperatures oen results in growth of amorphous materials.This is fortunate as it enables investigation of the correlations between the luminescence efficiency and the crystallinity, or lack thereof.Europium titanate was chosen as a model material in this work due to its luminescent properties and because it could be obtained as thin coatings by combining the known deposition processes of Eu 2 O 3 and TiO 2 using the Eu(thd) 3 /O 3 (ref.5 and 6) and TiCl 4 -H 2 O 7 precursor pairs.Europium titanate has previously been synthesized by several techniques and giving rise to several types of microstructures.The synthesis routes includes hydrothermal, [8][9][10] solgel, [11][12][13][14][15][16][17][18][19][20] high temperature ring, 21 ame pyrolysis 22 and electrospinning. 23,24The different types of microstructures includes hollow spheres, 8 thin lms, 12,16,17,19,20 porous lms, 13,18 nanoparticles, 9,15,22,25 nanobers, 23,24 nanotubes 10 and powders. 11,21uropium titanium oxides represents a diverse and fairly well known system which make them suitable for investigating the effects that layered ALD-type growth has on their luminescent properties.
The current work focuses on the application of ALD for deposition of luminescent materials.First, we explore the luminescence properties of thin lms of europium titanium oxides as a function of their europium content, ranging from pure TiO 2 to pure Eu 2 O 3 .Secondly, we describe how annealing changes the crystallinity of as-deposited lms and thereby the luminescence of the samples.Lastly, we investigate the correlation between the deposition temperature and the composition, crystallinity and luminescence properties of the lms.

Experimental
The thin lms were deposited in a F-120 Sat-type ALD-reactor (ASM Microchemistry Ltd) using a combination of the processes for deposition of Eu 2 O 3 , 5 and TiO 2 . 7Eu 2 O 3 was deposited using Eu(thd) 3 [europium tris(2,2,6,6-tetramethyl-3,5-heptanedionate), Strem > 99.9%] and ozone, and TiO 2 was deposited using TiCl 4 (Aldrich > 99.9%) and water.The ozone was produced from 99.8% O 2 (AGA) in a BMT 803N Ozone Generator.Nitrogen was used as carrier and purge gas, and was separated from air in a nitrogen generator (Schmidlin UHPN3001 N 2 purier, >99.999%N 2 + Ar purity).The sublimation temperature used for Eu(thd) 3 was 145 C, and the pulse durations for the deposition of Eu 2 O 3 layers were 1.5/1.5/4/1.5 s for Eu(thd) 3 /purge/O 3 /purge.For TiO 2 , pulse durations were 0.8/1/1/1 s for TiCl 4 /purge/H 2 O/ purge.These pulse and purge times have through prior experience proven to be sufficient for proper surface saturation and purging for the reactor used, in agreement with prior work. 5The substrates used for all depositions were cut from polished Si(100) wafers into pieces of 3 Â 3 cm 2 .One of these substrates was used per deposition.In addition, 2 strips of 0.5 Â 4 cm 2 were cut from the same type of wafer and positioned perpendicular to the gas ow and about 8 cm apart by the inlet and the exhaust of the deposition chamber to monitor the gradients of the growth.The native oxide on the silicon was not removed.All substrates were cleaned in ethanol and dried using pressurized air.The substrates were further pre-cleaned by an in situ 30 minute pulsing sequence of ozone and nitrogen at the deposition temperature to remove any organic residues prior to the deposition.To aid comparison, the total lm thicknesses of the samples were kept at ca. 80 nm, based on the growth rates deduced for Eu 2 O 3 , TiO 2 , and samples with 1 : 1 Eu : Ti pulsing ratio.The total number of cation cycles used per deposition is summarized in Table 1.For brevity, a sample consisting of 20% Eu 2 O 3 and 80% TiO 2 cycles will be called 20Eu.All lms were deposited at 300 C, while lms of 50Eu were additionally deposited at 225 and 375 C, keeping the number of cycles the same.
The crystallinity and phase content of the deposited lms were evaluated by means of a Siemens D5000 X-ray diffractometer using CuK a1 (nm) radiation from a Ge(111) monochromator.The chemical stoichiometry was measured by X-ray uorescence on a Phillips PW2400 Spectrometer and interpreted with the Uniquant analysis soware.Annealing was conducted in a rapid thermal processing oven (RTP) in vacuum for 10 minutes at a given temperature using a heating rate of 25 C s À1 .Film thickness and refractive index were determined with a spectroscopic ellipsometer (J. A. Woollam VASE) in the 250-1000 nm range.The Tauc-Lorentz formalism 26,27 was used to model and parameterize the ellipsometry experimental data.These thicknesses were used to determine the growth rate per cation cycle (GPC).Surface morphology was characterized with a Park XE70 atomic force microscope (AFM) in tapping mode.Photoluminescence in the lms were excited by a 325 nm CdHe laser and detected by an OceanOptics USB4000 photospectrometer in the 350-1000 nm range.Diffuse reectance measurements were done with a Shimadzu UV-3600 photospectrometer with an integrating sphere.

Results
Fig. 1 shows the growth rates of the deposited lms for samples with different pulsed stoichiometries.The variation in growth rate show local minima at 30Eu and 90Eu and a local maximum at 50Eu.The lms were rather uniform with negligible gradients, except for 10Eu and 20Eu.Thus the measured growth rate for the inlet and exhaust monitor strips are shown for 10Eu and 20Eu, while the other runs showed no signicant variations in the lm thickness.It can be seen from Fig. 1 that the growth rate at the gas inlet side was substantially lower than that at the exhaust side for 10Eu and 20Eu.The three data points for 10Eu and 20Eu are from three separate depositions.The growth rate of pure TiO 2 was 73.9 pm per cycle which is consistent with the values reported by Aarik et al. 7 for ALD under similar conditions.
The dependency of the pulsing composition on the cationic composition of the deposited lms was analyzed by X-ray uorescence and showed a smooth variation as function of the  The concentration of chlorine in the lm was also measured to be mostly less that 0.5 at% except for 90Eu were the Cl concentration was nearly 2 at%.The crystallinity of the as-deposited lms depended on the applied pulsing composition, as seen in Fig. 3.The X-ray diffractogram for 100Eu shows a broad feature around 30 that may stem from monoclinic Eu 2 O 3 , 28 similar to what we found in our previous work on Ln 2 O 3 depositions. 5For the range 30Eu to 90Eu, the depositions resulted in amorphous lms.The TiO 2 depositions (0Eu) gave polycrystalline anatase.For the range 10Eu to 20Eu, the growth characteristics depended on the position in the reaction chamber.The samples were consistently thinner at the gas inlet side than at the exhaust side.From the XRD data it is seen that the exhaust side contains more anatase than the inlet side for both 10Eu and 20Eu depositions.The expanded views of the (011) anatase reection show that the peak is shied from 25.4 for pure anatase to 25.3 for the Eu doped samples.
AFM images of the 10 and 20Eu samples are shown in Fig. 4. At the exhaust side, both samples consist of a dense layer of roughly 100 nm size crystallites.At the gas inlet, the crystallites are also ca. 100 nm in size, but the crystallite density is notably lower, in particular for the 20Eu sample where the crystallites are separated by smooth regions.The RMS roughness for the 3 Â 3 mm 2 area is 4.3 nm for 0Eu (not shown), 11 and 10 nm for 10Eu inlet and exhaust respectively, and 3.2 and 6.8 nm for 20Eu at the inlet and exhaust respectively.The smooth region between the crystallites has a RMS roughness of 0.1-0.3nm, which also applies for the 30-90Eu samples.The 100Eu sample has a RMS roughness of 0.6 nm.
The PL spectra of the 10 to 90Eu depositions are presented in Fig. 5.The spectra are normalized to the most intense peak and the 0Eu spectrum is in addition multiplied by 0.5 for graphical reasons.The spectrum consists of a number of distinct lines.For Eu 3+ they range from 580 to 710 nm, and are associated with the transitions from the excited state 5 D 0 to 7 F J (J ¼ 0, 1, 2, 3 and 4) levels as indicated in the gure.The shape of the PL spectra is similar for the 30Eu to 90Eu depositions, in addition to the gas inlet side of the 20Eu depositions.There is a gradual change in the shape of the spectra going from the exhaust side of 20Eu to 10Eu inlet to 10Eu exhaust.This is particularly visible for the hypersensitive 5 D 0 / 7 F 2 transition.
The total PL intensity in the 550-750 nm range (400-900 nm used for 0Eu) varies with the europium content as shown in Fig. 6.
In order to shed light on the luminescent properties of the samples, the extinction coefficient and refractive index was modeled from ellipsometry data, Fig. 7.The stapled line in Fig. 7 show the wavelength of the CdHe laser used for PL   excitation.The refractive index (n) and extinction coefficient (k) at 325 nm as a function of Eu 2 O 3 cycle ratio is given in Fig. 8.The extinction coefficient at the laser wavelength decreases with increasing Eu concentration down to zero for pure Eu 2 O 3 , while n decreases from 3.5 for TiO 2 to 2.1 for Eu 2 O 3 .Variations in optical scattering between the samples can also cause differences in the measured emission intensities.Diffuse reectance measurements at zero degree incidence showed that the scattering at both the excitation and emission wavelengths were similar for all the samples and below 1%, with the exception of the 10Eu and 20Eu samples which showed a diffuse reectance of 2 and 1.2% respectively at 325 nm.
The 50Eu depositions were in addition carried out at 225 C and 375 C, keeping a constant number of 965 supercycles (1 Eu 2 O 3 + 1 TiO 2 cycle), consistent with Table 1.These depositions also resulted in amorphous lms similar to the depositions at 300 C, as shown in Table 2.However, there are notable differences.The growth rate rst increases and reaches a plateau at 300 C, while the Eu concentration and total PL intensity increases and k decreases for increasing deposition temperatures.The 50Eu samples obtained at the three deposition temperatures were annealed in order to investigate the luminescence properties as function of lm crystallinity.The lms remain amorphous up to an annealing temperature of 800 C where rutile starts to form, as can be seen in Fig. 9. Aer 900 C, the (222)-reection of Eu 2 Ti 2 O 7 appeared, while aer 1000 C the sample is best described as a polycrystalline mixture of Eu 2 Ti 2 O 7 and rutile.
The total PL intensity decreased when 50Eu samples were annealed, Fig. 10.For annealing temperatures of 700 C and above, the samples show very low luminescence.The exception is the sample deposited at 375 C which show a higher luminescence when annealed at 800 C than at 700 C. The PL spectra for the samples deposited at 375 C are shown in Fig. 11.The spectra are normalized to the highest peak.It is further seen from the gure that although the PL intensity decreases with annealing, the shape of the spectra remain similar for annealing temperatures up to 700 C, whereaer it changed in shape.

Discussion
By varying the overall composition of the deposited lms between TiO 2 and Eu 2 O 3 , several different phases were formed.From low to high Eu content, these phases were: anatase, europium doped anatase, amorphous Eu x Ti y O z , and poorly crystalline Eu 2 O 3 .The results further show that upon annealing the 50Eu samples at 800 C and higher, rutile and crystalline Eu 2 Ti 2 O 7 is formed from the amorphous oxide.In this work, XRD and PL have been used to distinguish these phases.
The total emission intensity of our samples change with different Eu 3+ concentration, deposition and annealing temperature.We have tried to design the samples to be optically similar to keep the main differences within k(l), Eu 3+ concentration, and their local symmetry.All samples were approximately of the same thickness and with a surface which is smooth and similar, as measured by AFM.In addition, diffuse reectance measurements show that the scattering in the samples is rather low.
The Eu 3+ PL is sensitive to the coordination and local environment around the cation site based on the fact that the    splittings of the 2S+1 L J multiplets and the transition probabilities between them are symmetry dependent.Usually the transition 5 D 0 / 7 F 2 is the strongest.It is hypersensitive to site symmetry and of electric-dipole nature.For comparison, the 5 D 0 / 7 F 1 transition is mostly of magnetic-dipole nature and insensitive to site symmetry. 29The current discussion is divided into three parts: anatase containing 0Eu to 20Eu samples, amorphous 30Eu to 100Eu samples, and annealed 50Eu samples.
0Eu to 20Eu: anatase and Eu x Ti y O z The 10 and 20Eu samples showed a notable difference in growth positioned at the inlet and exhaust side of the reaction chamber (Fig. 1).The lms grown on the monitor strips at the inlet were thinner than those at the exhaust.The entire substrates appeared either thin or thick rather than exhibiting a smooth gradient through the chamber.This growth pattern was reproducible.The thinner lms have a lower Ti content (Fig. 2).The AFM images (Fig. 4) show that the size of the crystallites is similar for all the samples.However, there are differences with respect to the density of crystallites, indicating that the nucleation of crystallites in the amorphous matrix is a limiting process.
The PL spectra in Fig. 5 suggests that these lms consist of two distinct phases as they can be regarded as the superposition of two characteristic PL spectra; the spectra of 80Eu having broad peaks from the 7 F J manifolds, and the sharp lines of the inlet strip of 10Eu.The superposition and gradual growth of one and decay of the other phase can be observed most clearly for the 7 F 2 peak.The PL from the inlet strip of 10Eu also has a shape that is consistent with Eu 3+ doped anatase. 13,18The PL from 80Eu has smeared out peaks which one can imagine arise from the continuous slight variation in the coordination environment for a cation in an amorphous oxide where the bond angles are varying.Very similarly shaped PL spectra have also been observed before and been described as typical for a Eu 3+ doped disordered system. 10Leroy et al. 13 observed very similar spectra for PL from Eu 3+ doped mesoporous titania with the same gradual spectrum change accompanying the change from anatase to amorphous phase.
Comparing the AFM images, XRD data, and PL spectra, there is a clear correlation between the amount of crystallites, the XRD peak intensities, and ratio between the two overlapping PL spectra.This correlation only applies for samples with less than 30% Eu 2 O 3 cycles.Any correlation between the extinction coefficient k and ratio between the anatase and amorphous phases is less clear.For example, at the PL excitation wavelength of 325 nm the extinction coefficient does not increase monotonically with decreasing %Eu below 30Eu.This lack of correlation for k can also have to do with difficulties of adopting a reliable optical model for the ellipsometry data used to arrive at k for this range of %Eu and wavelength.Since the size of the crystallites is close to the wavelength of the light in the UV part of the spectrum, a lot of scattering is expected for these samples.Such scattering is not trivial to model and could result in some uncertainty in k in the UV range of these samples.
Scattering in the UV range does not affect the modeling of thicknesses as this is done by modeling the VIS and NIR range ellipsometry data.
By comparing the (011) reection of pure and doped anatase in Fig. 3, the peak shi from 25.4 to 25.3 indicates a slight unit cell expansion on incorporation of the larger Eu 3+ ions.The shi is identical for both 10Eu and 20Eu, indicating that the anatase solid solution phase is likely to be saturated with Eu 3+ in all these samples.Thus, for low Eu contents, simultaneous growth of amorphous Eu x Ti y O z and Eu-saturated anatase may occur.This competition between the growths of the two phases may explain the difference in growth rate between the samples positioned at the inlet and exhaust side in the reaction chamber.
The depositions in the 30Eu to 100Eu range all resulted in amorphous and very at lms with a RMS roughness of approximately 0.3 nm, with the exception of pure Eu 2 O 3 (100Eu) which show sign of poor crystallinity (Fig. 3) and a slightly higher roughness of 0.6 nm.The normalized PL spectra in Fig. 5 indicate that the coordination of the Eu 3+ is similar for all of these samples.On the other hand, the emission intensity varies strongly with the Eu content as shown in Fig. 6.At low Eu content, the luminescence intensity increases with the Eu content.The limiting factor for the emission in this Eu range is likely the energy transfer from TiO 2 to Eu 3+ .Increasing the Eu content increases the likelihood of an Eu 3+ being available to accept the energy transfer thus increasing the emission intensity.The strongest emission is seen for the 50Eu sample which corresponds to 17.7 cat.%Eu.At higher Eu contents, the luminescence intensity decreases.
Although it would be tempting to assume that the decrease in luminescence intensity for >50Eu samples is caused by concentration quenching, we do not believe this is the case.At least part of the reason for the decrease in emission intensity is that k (325 nm) decreases, as seen in Fig. 7.This leads to a reduction in the absorption of the 325 nm laser in the thin lms.Ovenstone et al. 14 found that for amorphous Eu x Ti y O z powders, the emission intensity increased continuously for x up to 0.5 (compared to our maximum at x ¼ 0.177), which was the highest Eu content in their investigation.This supports our explanation about the lack of laser absorption being the limiting factor for our thin lms with high Eu contents, as UV absorption will not be a limiting factor in bulk powders.If thicker lms had been grown, the emission maximum would likely be at higher Eu contents as more of the laser would be absorbed.For pure Eu 2 O 3 , the emission was not detectable with our set-up and only a very weak red spot was visible by eye, as the absorption at 325 nm is close to zero.

Effects of the deposition temperature and annealing on Eu x Ti y O z
The effects resulting from different deposition temperatures are summarized in Table 2.The temperature dependence of the TiCl 4 -H 2 O and Eu(thd) 3 /O 3 ALD growth processes may explain these results.It is known that the growth rate of TiO 2 decreases 7 while that of Eu 2 O 3 increases 5 with temperature.Thus, by keeping the cycle sequence the same, one would expect a relative increase in the Eu content with increasing deposition temperature, which is consistent with the XRF results.Also, the extinction coefficient decreases with the relative decrease in the Ti content at higher temperatures.Still, the change in emission intensity cannot be fully explained by the growth rates.As the number of cycles was constant for these samples, the lower growth rate at 225 C resulted in a thinner lm, but not sufficiently thin to explain why the emission intensity should be reduced to just 1/3 of that of the 300 C deposited sample.Possibly, the higher Cl content in the sample grown at 225 C could indicate that the surface reactions during the ALD growth is more incomplete at this temperature than at higher temperatures.
Since the 50Eu samples showed the highest emission intensity, we chose to anneal these samples to investigate how changes in the amorphous nature would affect the luminescence properties of these lms.Fig. 9 shows that the lms remained amorphous for annealing temperatures up to 800 C, whereaer rutile started to crystallize.The emission intensity decreases with increasing annealing temperature, as seen in Fig. 10, and remains low for all annealing temperatures above 700 C.This is similar to what Ovenstone et al. 11 found for amorphous europium titanate powders.However, their longer annealing times gave a lower quenching temperature.The only exception currently observed is the deposition at 375 C, for which the emission intensity is higher aer annealing at 800 C annealing than at 700 C. The origin of this is unclear.Whereas the XRD analysis could not reveal any crystalline phases for these samples, the PL spectra in Fig. 11 indicates that the coordination environment of the Eu 3+ ions changes.However, at this point we do not have sufficient experimental data to indicate what kind of coordination or what kind of crystal phase this spectrum corresponds to.A change in PL spectrum aer annealing at 800 C was also observed for the depositions at 225 and 300 C, but these samples did not experience an increase in emission intensity and the luminescence intensity was very low.

Conclusion
By using the precursor pairs TiCl 4 -H 2 O and Eu(thd) 3 /O 3 in ALD growth processes it is possible to obtain controlled Ti-Eu oxide materials by varying the cycle ratio, pulse duration and annealing temperatures.The observed phases are anatase, europium doped anatase, amorphous Eu x Ti y O z , crystalline Eu 2 Ti 2 O 7 and poorly crystalline Eu 2 O 3 .The highest luminescence emission intensity was found for amorphous samples grown with alternating Eu 2 O 3 and TiO 2 cycles.The limiting factor for luminescence at low Eu contents is the low concentration of emitting Eu 3+ atoms, while the luminescence at high Eu contents is limited by a low UV absorption due to the low Ti 4+ contents in these samples.
From a general point of view it is interesting to note that large crystals are in fact not always preferable.We were indeed expecting better luminescence efficiency aer annealing due to formation of larger crystals, but rather observed that the amorphous phase gave the highest luminescence efficiency.From a practical point of view, this can be rather advantageous.Large single crystalline lms are usually only obtainable through epitaxial growth on single crystal substrates, while amorphous lms can be achieved through ALD-growth on most types of substrates.

Fig. 1
Fig. 1 Growth rate per cation cycle (GPC) as a function of the percent of Eu 2 O 3 cycles.Black diamonds indicate smooth depositions, while red and blue symbols indicate depositions with gradients, measured the inlet and exhaust of the reaction chamber, respectively.

Fig. 2
Fig. 2 Cationic composition of the deposited films as function of % Eu 2 O 3 cycles, as measured by XRF.The analyzed chlorine content is also shown (right scale).

Fig. 3 X
Fig. 3 X-ray diffractograms for 0, 10, 20, 30 and 100% Eu 2 O 3 cycles.The observed reflections are from the anatase polymorph of TiO 2 , while the sharp reflection at 33 is from the Si (001) substrate.The left side show an expanded view of the (011) anatase reflection.

Fig. 4
Fig.4AFM pictures of 10Eu and 20Eu films at the inlet and exhaust side.Note the differences in the z scale.

Fig. 5
Fig.5Normalized emission PL spectra of 0 to 90Eu samples, excited by a 325 nm HeCd laser.For the 10 and 20Eu samples, the inlet and exhaust sides are named "inl."and "exh."respectively.Some detector artifacts have been removed from the data and the 0Eu data have been multiplied by 0.5.

Fig. 7
Fig. 7 Extinction coefficient from modeling VASE data, showing that k increases with decreasing Eu concentration.The 10Eu inlet sample was difficult to model in the UV range, probably due to its large surface roughness compared to the other samples.

Fig. 8
Fig.8Refractive index and extinction coefficient at 325 nm for the europium titanium oxide depositions.

Fig. 9
Fig.9XRD data for a 50Eu films after annealing.The sharp reflection at 33 and the large reflection at 69 are from the Si(001) substrate.

Fig. 11
Fig. 11 Normalized emission PL spectra of the 50Eu samples deposited at T dep ¼ 375 C and the indicated temperatures.

Table 2
Summary of 50Eu depositions at 225, 300 and 375 C