Passivation of lanthanide surface sites in sub-10 nm NaYF₄:Eu³⁺ nanocrystals

Recently, various groups have reported the preparation of sodium yttrium fluorides (NaYF₄) nanocrystals (NCs) in either α or β phase (Wang et al. 2010a). Yan’s group initially reported the first use of trifluoroacetate salts as precursors for one step synthesis of NaYF₄ NCs (Mai et al. 2006). Since yttrium can be simply substituted by other lanthanide ions, the material has received considerable attention as a nanocrystalline matrix for doping by lanthanide ions. The additional advantages of the β-NaYF₄ matrix arise from a wide energy bandgap (~8 eV) (Chong et al. 2007) and a very low phonon energy (~360 cm⁻¹ ≈ 45 meV) (Li et al. 2007; Haase and Schäfer 2011), as compared to oxide NCs (Güdel and Pollnau 2000). Thus, the photoluminescence quantum yield (QY) in this material is high, since non-radiative relaxation of excited electrons diminishes. The reduced non-radiative relaxation rate is also responsible for the intense emission from upper excited states of lanthanide ions in a low concentration (Li et al. 2009). The same reasons also account for an efficient up-conversion emission for co-doped NaYF₄:Yb³⁺, Er³⁺ NCs (Li et al. 2007). Nowadays, NCs of the β phase NaYF₄ doped with various lanthanide ions are well-known phosphors which have found use in a variety of applications, especially as efficient markers in biological systems (Chatterjee et al. 2010).

In the search for the best capping ligands for fluoride NCs, Shan et al. proposed the use of trioctylphosphine oxide (TOPO) as a coordinating solvent in the reaction from trifluoroacetate precursors (Shan et al. 2007). Since TOPO has a higher boiling point than the previously used solvents (oleylamine, oleic acid, or octadecene) (Mai et al. 2006; Shan et al. 2007; Boyer et al. 2006), the resulting particles were more monodispersed, crystallized in the pure hexagonal phase and have found to be small enough (~10 nm) to be used as biological labels.

Small size of NCs is a requirement in bioapplications but the large number of surface atoms could have several implications. In the case of NaYF₄, the fraction of surface Y³⁺ reaches 17 % for 8.7 nm sized NCs. For doped NCs, the percentage of doping ions on the surface can even be higher as dopants are often pushed out from the NC core (Podhorodecki et al. 2009). This leads to a decrease in QY, because of an increase in the efficiency of non-radiative relaxation by interaction of the optically active ions with surface defects as well as with organic ligands (Heer et al. 2004). Yi et al. ascribed this interaction to a significant reduction of lanthanide (Ln) emission efficiency in NCs compared to the bulk matrix (at least one order of magnitude) (Yi and Chow 2006b; Mialon et al. 2010).

In order to preserve the high PL QY, Yi et al. initially reported the synthesis of a core/shell structure composed of NaYF₄:Yb³⁺, Er³⁺/NaYF₄ and NaYF₄:Yb³⁺, Tm³⁺/NaYF₄, and showed reduced interactions of lanthanides with surface defects, ligands, and solvent (Yi and Chow 2006a). As a result, the enhancement of PL intensity increased up to 29 times when a shell was deposited on up-converting NaYF₄:Yb³⁺, Tm³⁺ core NCs. Even bigger increase, up to 450 times, was reported by Wang et al. for shell covered small NCs (10 nm) (Wang et al. 2010b). Additionally, more complex shell compositions (e.g., NaYF₄:Yb³⁺,Er³⁺/NaGdF₄:Yb³⁺) were also proposed to further increase the observed PL intensity (Guo et al. 2010). However up till now, only the PL intensity was investigated in the above-mentioned studies. Moreover, the work only considered up-conversion PL and did not take advantage of crystal phase and/or oxygen-sensitive Eu³⁺ ions.

In order to correctly interpret an increase in PL intensity of colloidal NCs, its concentration has to be evaluated precisely. There is often lack of information on how the NCs concentration was calculated. To the best of our knowledge, none of the studies related to Ln doped NCs show the absorption spectra, due to difficulties associated with weak absorbing NCs:Ln³⁺. Evaluation of NCs concentration based on NCs mass is even more problematic due to differences in mass of core and core/shell NC as well as different mass contribution from surface passivated ligands. Thus, the calculation of lanthanide doped NCs concentration is a nontrivial task, leading to a significant experimental error. Therefore, the impact of the shell formation only on PL intensity can over-estimate the obtained results.

The lack of structural information brings additional uncertainty to the discussion on the origin of enhanced PL intensity. A crucial current limitation is that transmission electron microscopy (TEM) methods are generally unable to differentiate between the core and the shell structure of NaYF₄/NaLnF₄ NCs due to the similarity of the lattice parameters resulting in small contrast in TEM images (Abel et al. 2011; Liu et al. 2010; Cheng et al. 2010). On the other hand, techniques such as Electron Energy Loss Spectroscopy (EELS), Energy-Dispersive X-ray Spectroscopy (EDS), and X-ray photoelectron spectroscopy (XPS) provide presence and distribution of ions inside NCs. Veggel and co-workers recently carried out detailed XPS studies to determine the formation of lanthanide based core/shell particles (Abel et al. 2009). In another study, the authors have used EELS and EDS techniques and concluded that a complete and uniform shell growth does not occur on all of the NCs and an improvement in the synthesis of core/shell structure is required (Abel et al. 2011). Despite the encouraging results presented by Veggel’s group, more widely available and user-friendly techniques and protocols are required to distinguish in a routine manner the formation of shell around NC core when doped with lanthanide ions.

In this article, we have carried out detailed investigations of the optical properties of β-NaYF₄:Eu³⁺ NCs and in particular Eu³⁺ ions, which are an excellent probe of a local crystal field. The europium doped NCs were passivated by NaYF₄ shell and a broad range of optical spectroscopy experiments were conducted to distinguish between the emission properties of europium in the core and intended core/shell samples (Fig. 1). Based on the obtained results, we are able to point out the differences and propose a simple yet reliable confirmation of the surface passivation.

The optical properties were investigated to examine the influence of Eu³⁺ located on the surface on the excitation and emission of NCs smaller than 10 nm in diameter. We started our investigations with the absorption spectra measurements which are often ignored in the lanthanide doped fluoride NCs. The NaYF₄ crystals exhibit a large energy gap (~8 eV), thus absorption edge is not expected in UV–Vis range. Moreover, due to the very small value of Ln absorption cross section, the absorption bands related to f–f transitions are strongly limited. (Carlos et al. 2009). However, we were able to observe the ⁷F₀‐⁵L₆ transition for the high concentrations of NaYF₄:Eu³⁺/NaYF₄ NCs (inset of Fig. 4) which confirms a relaxation of selection rules making f‐f transitions slightly probable when Eu³⁺ occupy non-centrosymmetric crystal sites. In the inset of Fig. 4, we can notice that the band of ⁷F₀‐⁵L₆ transition is settled on another wide absorption band. In order to determine its origin, the absorption spectrum was recorded in a wider spectral range. Fig. 4 shows the absorption spectra of core and core/shell NCs capped by TOPO and TOPO itself. It is clearly visible that the absorption of core/shell NCs is determined by absorption of TOPO ligands. The spectrum of core only has additional absorption band at around 250 nm. This band is significantly wider and much more intensive in comparison with f‐f transitions bands. We attribute this band to a transition between the ground and the charge-transfer state of the Eu–O bond (CT). We presented the detailed discussion about parameters influencing the exact CT position in our previous work (Banski et al. 2010). Moreover, the observed position of absorption band is in a good agreement with the literature values equal to 240–255 nm for Y₂O₃:Eu³⁺ (Shang et al. 2011).

On the XRD spectra there were no diffraction peaks corresponding to residual Ln₂O₃ phase. Moreover, the CT band totally disappeared for surface passivated sample, what suggests that all Eu–O bonds are efficiently broken. It confirms that the CT band is due to interaction of Eu³⁺ on the NCs surface and oxygen from the P=O group of TOPO surface ligands. The absorption spectra can be used in a simple way to prove the passivation of surface sites in Eu³⁺ doped fluoride NCs.

The photoluminescence excitation (PLE) experiment is much more sensitive to the optical properties for weakly absorbing materials. Hence, the PLE spectra corresponding to the direct excitation of the Eu³⁺ ions in NaYF₄:Eu³⁺ and NaYF₄:Eu³⁺/NaYF₄ NCs were recorded.

The PLE spectra presented in Fig. 5 (PL intensity monitored at 616 nm) clearly show that direct ⁷F₀–⁵L₆ transition at 395 nm wavelength causes the most intensive Eu³⁺ excitation in both samples. Other excitation bands associated with f–f transitions are also present. At longer wavelengths, the transitions from ⁷F₀ to ⁵D₁, ⁵D₂ and ⁵D₃ states are present at 527, 464 and 416 nm, respectively. At higher energy, the excitations through the following transitions ⁷F₀–⁵G_j, ⁷F₀–⁵D₄, ⁷F₀–⁵H_j, and ⁷F₀–⁵F_j are possible at 373–387, 360 and 275–325 nm, respectively (Gao et al. 2009). However, all of these excitation bands are ~3–20 times less intensive. The full width at half maximum (FWHM) of ⁷F₀–⁵L₆ in β-NaYF₄:Eu³⁺ NCs is only 4.7 nm, which reduces to 3.8 nm after the shell formation. The small FWHM confirms the high quality of nanocrystal structure.

The PLE spectra of β-NaYF₄:Eu³⁺ and β-NaYF₄:Eu³⁺/β-NaYF₄ samples are quite similar. Due to the low symmetry of the crystal field of β-NaYF₄:Eu³⁺ NCs, the ⁵H_J and ⁵F_J terms are split into separate energy levels: 285, 297, 307 and 317 nm. After the shell formation, the excitation band at 307 nm disappears, which can be associated with the reduction of Eu³⁺ on the surface site.

In the PL spectra given in Fig. 5, many well-resolved emission bands corresponding to the electronic and magnetic dipole transitions from ⁵D₁ and ⁵D₀ states to ⁷F_j manifolds can be distinguished. The PL peak positions of all the observed bands are consistent for both samples. The green emission from upper excited state (⁵D₁) is possible due to the limited phonon assisted relaxation rate in the high quality β-NaYF₄ matrix. It indicates that the emission from ⁵D₁ and ⁵D₀ states is competitive.

The surface passivation causes an increase in the intensity ratio of the electric dipole transitions from the lowest and upper excited state (⁵D₁–⁷F₃/⁵D₀–⁷F₂). Two possible phenomena could cause this observation. First, when β-NaYF₄ shell was formed, some of Eu³⁺ ions may diffuse from the core to the shell (Dong and van Veggel 2008). The separation between the ions increases and the probability of cross-relaxation processes diminishes. As only a thin layer of shell is applied (~1.5 nm from XRD, ~1.9 nm form TEM), the diffused Eu³⁺ ions would, in fact, again occupy the surface sites. Hence another mechanism takes place in this case. Excited Eu³⁺ ions located on the surface sites interact with higher energy surface phonons and/or ligand molecules, and the nonradiative relaxation from ⁵D₁ state is more prominent due to multi-phonon transitions. A thin β-NaYF₄ shell successfully passivates the surface active sites and separates Eu³⁺ ions from TOPO and solvent. As a result, the green emission from ⁵D₁ of Eu³⁺ state of core/shell NCs increased, which is an important evidence for the shell formation. Moreover, this explanation is consistent with conclusions derived from absorption measurements.

The characteristic FWHM parameters of the magnetic and electric dipole transitions presented in the PL spectra (Fig. 5) are reduced by 1 nm when the shell is formed (5.86–4.87 nm and 5.63–4.65 nm for ⁵D₀–⁷F₁ and ⁵D₀–⁷F₂ transitions, respectively). This observation is yet another indication that the number of Eu³⁺ occupied sites was reduced by the surface passivation.

In order to closely look at the emission properties, the PL spectra of ⁵D₀–⁷F₁ and ⁵D₀–⁷F₂ transitions were recorded with 0.3 nm resolution (Fig. 6). Crystal field theory predicts that for ⁵D₀–⁷F₂ transition of Eu³⁺ in a C _3h site, we should observed only one emission peak (Ju et al. 2009). However, for the nanocrystalline matrix the symmetry of Eu³⁺ sites could be reduced to C ₃ or even C _S symmetry with the number of observed PL lines increasing to five (Ju et al. 2009). The PL spectra of both samples are presented in Fig. 6. They are composed of four emission bands, and they look similar, at a first glance. However, the difference between β-NaYF₄:Eu³⁺ and β-NaYF₄:Eu³⁺/β-NaYF₄ NCs is well-resolved. The ⁵D₀–⁷F₁ as well as the ⁵D₀–⁷F₂ transition in the core sample contains wide emission band cantered at 590.3 and 613.3 nm, respectively. These bands diminished in the core shell/sample. We ascribed their origin to Eu³⁺ ions on the NC surface, which become passivated by β-NaYF₄ shell. The other narrow PL peaks did not change after shell formation, so they can be related to the well-defined internal sites of β-NaYF₄ matrix.

Complementary results were also obtained from PL decay experiments (Fig. 7). The PL decay time of the upper ⁵D₁ state is driven by cross-relaxation between ions and phonon related to non-radiative processes. Considering that shell formation influences the relaxation related to ions interactions with surface phonons and/or ligand molecules, an elongation of PL lifetime is expected. We fitted PL decays of ⁵D₁ state with stretched exponential functions.

where τ is the PL lifetime, β can be interpreted as a disorder parameter, and I ₀ is a constant. A slight elongation of the PL decay time from 1,325 to 1,623 μs was observed for core and core/shell samples, respectively (Table 1). Moreover, the β parameter, which estimates the uniformity of Eu³⁺ ions environment in NCs increased from 0.835 to 0.912 after surface passivation. This further confirms our expectation according to homogeneous shell formation during surface passivation processes. The elongation of decay and rise times of PL from ⁵D₀ state are other characteristic observations.

Recently, the cation exchange was suggested to form a gradient alloy structure, when the core/shell fluoride nanocrystals were expected (Dong et al. 2011). The authors concluded that this mechanism can significantly deplete the Eu³⁺ concentration in a NC core when undoped shell is intended to be deposited. As a result, any changes in the optical properties due to a reduction of the ion–ion interaction are incorrectly interpreted as a proof of shell formation.

In order to exclude the cation exchange as a possible mechanism for improvement in the optical properties of our core/shell structure, we evoke our results of PL decay times of NaYF₄ NCs doped with 2 % of Eu³⁺. The doping level was chosen to be 2.5 times lower than the original concentration, because the volume of core/shell NCs increases two times compared to core ones. For 2 % sample, in a comparison with 5 % core sample, the τ_PL decay time will increase due to bigger inter ionic distances, which are supposed to reduce the efficiency of Eu³⁺–Eu³⁺ cross relaxation. The determined value of PL decay times (τ_PL) are 7,782 and 2,682 μs for emission from ⁵D₀ and ⁵D₁, respectively. Both of them significantly exceed the values obtained for core as well core/shell sample (Table 1). These results indicate that during shell formation the cation diffuse was not sufficient to separate Eu³⁺ ions efficiently, thus the observed changes of PL spectra have to arise directly from surface passiation.

Finally, the photoluminescence quantum yield (PL QY), defined as QY = τ _PL /τ _R *100 %, was calculated to assess the improvement in optical properties of NaYF₄:Eu³⁺ NCs after the shell formation. The modified equation (eq. 2) originally given by Verhoeven et al. was used to calculate radiative decay time from ⁵D₀ state (τ _R) (Werts et al. 2002).

ε is the dielectric constant of NaYF₄ matrix and cyclohexane equal 2.477 and 2.023, respectively (Lage et al. 2005), A _MD,0 is the spontaneous emission probability for ⁵D₀‐⁷F₁ transition in a vacuum (14.65 s⁻¹), and (I_TOT/I_MD) is the ratio of the total Eu³⁺ emission spectrum to the area of the ⁵D₀‐⁷F₁ band (Vela et al. 2008). The calculated values of radiative lifetime are τ _R  = 10.87 ms (NaYF₄:Eu³⁺) and τ _R  = 11.39 ms (NaYF₄:Eu³⁺NaYF₄). Taking values of τ _PL from PL decay experiments, the PL QY was calculated to be 57 % for the core and 64 % for the core/shell. This observation implies that the shell formation results in an increase in the efficiency of Eu³⁺ luminescence.

In our investigations of Eu³⁺ doped NaYF₄ NCs, we observe a significant absorption band in the UV range. We attribute this to the Eu–O charge-transfer state, which arises from the interaction between oxygen (from trioctylphosphine oxide) and Eu³⁺ at the nanocrystal surface. For NCs smaller than 10 nm, due to their high surface to volume ratio, charge transfer becomes an important absorption mechanism. We have shown that the surface of NaYF₄:Eu³⁺ NCs can be successfully passivated by a thin layer of undoped matrix. This was confirmed by the absence of CT band in the absorption spectra. Moreover, the surface passivation modifies crystal field influence on the significant fraction of Eu³⁺ ions. This was observed as changes in PL and PLE characteristic features, ED/MD ratio and peaks broadenings, which are discussed in details. The change in the emission lifetime is determined as well, which is used to calculate an increase in the PL emission efficiency up to a value of 64 % for surface passivated NaYF₄:Eu³⁺/NaYF₄ NCs.