Structural and Optical Characterisation of an Erbium/Ytterbium Doped Hybrid Material Developed via a Nonhydrolytic Sol-Gel Route

2.1. Material

In order to avoid any relaxation due to the OH groups, inherent to the hydrolytic sol-gel route, the goal of the material synthesis consists of obtaining an OH-free condensed organo-silane. To achieve this, the sol-gel synthesis was conducted via a nonhydrolytic sol-gel process, as sketched in Figure 1.

Compared to the hydrolytic routes, which can be processed at ambient temperature, nonhydrolytic processes typically require higher temperature and a catalyst to allow the condensation reaction between precursors of different reactivity. Generally, to condense an alkoxysilane with a chlorosilane, a Lewis acid catalyst, such as zirconium tetrachloride (ZrCl₄) is employed [16].

The sol preparation consisted of mixing a photocurable organically modified silicate, 3-trimethoxypropyltrimethoxysilane (MAPTMS, Assay ~99%) together with silicon tetrachloride (SiCl₄, Assay ~99.99%) in the presence of zirconium tetrachloride (ZrCl₄, Assay ~99.99%). The mixture was refluxed at 120°C for 24 hours before addition of the doping elements ErCl₃ and YbCl₃. The obtained sol was then filtered, through a 0.2 micrometer membrane, to remove any contaminating particle. NIR and neutron scattering spectra were recorded on bulk samples of 2 to 5 millimetre thicknesses. For EPR, the solid gels were crushed and the resulting powders sealed under vacuum into glass tubes.

2.2. Experimental Techniques

The evolution of the siloxane hybrid network has been followed by ²⁹Si-NMR spectroscopy, employing a 400 MHZ Bruker spectrometer. The various oxo bridges formed through either self-condensation or cocondensation reactions between the MAPTMS and the SiCl₄ are easily identified during the synthesis. Spectra were recorded at room temperature from liquid solution. Measurements showed that a 4-second recycle delay time with an 8 μs pulse duration, was sufficient for quantitative measurements. The chemical shifts were referenced with respect to Tetramethylsilane, used as an external reference. The Free Induction Decay processing used a 10 Hz line broadening. Each recorded spectrum is an average of all previously obtained spectra during the instrument acquisition time. 128 scans were accumulated for each spectrum.

NIR is commonly used to identify the absorption of the harmonics and combinations bands of the fundamental vibrations as well as the absorption of rare earth ions [17]. In the present case, this technique was employed to identify the optimum excitation wavelength of Er³⁺ and quantify the effect of the Yb³⁺ codoping effect. NIR spectra were performed using a NIR spectrometer. This optical spectrum analyser included an infrared source, a second-order blocking filter, and a scanning monochromator with both Ge (600–1900 nm) and PbSe (1500 and 3000 nm) detectors. The resolution can be selected between 0.25 and 20 nm. The scanning monochromator used a continuously rotating diffraction grating driven by an electronically controlled DC motor.

EPR spectroscopy characterises the environment of the paramagnetic species (in this case Er³⁺). In this work, this technique was used to investigate the homogeneity of the sample as function of Er³⁺ concentration. EPR spectra were recorded at room temperature employing an ER200D (X band) spectrometer operating at 9 GHz.

Compared to light scattering, which yields information on particles of >100 nm, neutrons scattering allows structural characterisation on a smaller scale, typically from 10 to 500 Å. Neutrons scattering experiments were performed on the small angle neutron scattering spectrometer at the Orphée 14 MW Reactor at the French Atomic Energy (CEA Saclay, France), using a neutron wavelength of 50 Å.

Fluorescence emission measurements were performed at room temperature by optically pumping the samples with a Ti: sapphire laser tuned at 980 nm. The light emitted at 90° from the monolith was analysed spectrally with a Jobin-Yvon U1000 double-monochromator fitted with a grating of 600 grooves/mm. The fluorescence was detected by a North-Coast liquid-nitrogen-cooled germanium detector. The spectral resolution ranged from 1 to 2 nm. The pump signal was mechanically chopped at 80 Hz and the signal from the Ge-detector was preamplified and passed to a lock-in amplifier.

3.1. ²⁹Si-NMR

The ²⁹Si-NMR spectra of the pure MAPTMS and SiCl₄ alkoxides (not shown here) exhibit a single peak at −42.8 and −19.2 ppm, respectively. This is an indication of the high purity of the employed precursors and absence of any hydrolysed species.

Figure 2 shows the ²⁹Si-NMR spectra after 1 hour and 24 hours of reaction. After 1 hour of reaction, in addition to the precursors peaks, two resonances located at −36.5 and −27 ppm are observed, with a respective contribution of 27.0 and 30.3% of the total silicon nuclei, as summarised in Table 1. After 24 hours of reaction, all peaks observed after 1 hour of reaction disappear with the appearance of 3 new peaks, located at −49.1, −58, and −59 ppm, and a large band between −65 and −69 ppm. This band is in fact composed of two main peaks centred at −66 and −67.8 ppm.

The noncondensed silicon nuclei (T₀ groups) are generally observed at chemical shifts lower than –45 ppm [18]. In a similar structure, the progressive condensation provoked an increase of 10 ppm [19, 20]. Based on the literature, it is possible to confirm that the peaks observed in the sample recorded after 1 hour of reaction are not attributable to any condensed silica species. As these peaks are located at chemical shifts located between those of the pure precursors, it is obvious that an exchange of the methoxy and chloride groups between both precursors has occurred, resulting in the formation of a mixture of organochlorosilane as shown in Table 1. Indeed, chloride is well known to be an electro-attractive group by inductive effect, which results in a decrease of the electronic density around the Si nuclei, explaining the displacement of the resonances of the hybrid precursor toward lower chemical shifts.

After 24 hours of reaction, the resonances observed are located in the region of condensed siloxane, previously identified between –50 and –100 ppm [20]. Furthermore, the progressive shift by ~10 ppm suggests a progressive increase of the degree of condensation, as summarised in Table 2. At this stage, it is important to note the absence in both spectra of any band centred at ~–40 ppm confirming the absence of any hydrolysed species. Indeed, in a previous study [20] on a similar material, we have highlighted the resonances of silanol groups (Si-OH) ~–40 ppm. The present result confirms the success of our nonhydrolytic sol-gel synthesis in obtaining OH-free materials and eliminates any further implication of these groups in the degradation of fluorescence via quenching.

Furthermore, examination of these results suggests that the chemical reactions involved in this nonhydrolytic sol-gel synthesis can be divided into two steps.

The first implies the exchange of the chemical groups between both silanes:

3.2. Near-Infrared Spectroscopy

Near infrared (NIR) spectra have been recorded on Er³⁺ doped and Er³⁺/Yb³⁺ codoped samples. Figure 3 shows the NIR absorption of Er³⁺(0.5, 1 and 2% at) doped samples. Compared to the undoped sample [20], in which the absorption bands in this spectral range were ascribed to the first and second overtones of the CH groups, the main difference is the appearance of two bands located at 980 and 1540 nm.

Bands observed at 980 and 1540 nm are assigned to erbium transitions from the fundamental energy level to the excited levels ⁴I_11/2 and ⁴I_13/2, respectively. This is in agreement with the typical erbium absorption [21, 22]. 980 nm has been found to be the most favourable pumping wavelength as erbium-doped optical amplifiers display a quasi-quantum limited 3 dB noise level, while also displaying an improved performance in terms of gain efficiency [23]. In our material, the absorption at 980 nm is particularly weak at around 20% of the absorption measured at 1540 nm, even for the sample containing the highest erbium concentration. To increase this absorption, Yb³⁺ ions can be employed as a codoping agent, as well known to exhibit a strong absorption at 980 nm [24]. When pumped at 980 nm, Yb³⁺ ions are excited to the first energy level (²F_5/2) and then relaxed to the ground level by nonradiative transition. However, in presence of Er³⁺ ions, an energy transfer process from the excited Yb³⁺ ions toward the ²I_13/2 level of the Er³⁺ ions can take place to increase the population of the excited Er³⁺ ions in this level, as sketched in (Figure 4). Indeed, because of the proximity of the ions energy levels, energy transfer is very efficient, as it is quasi-resonant. Since the population of the ⁴I_13/2 is increased by this mechanism, the luminescence lifetime of the Er³⁺ ions on this energy level should also be increased and the resulting fluorescence intensity improved.

Figure 5 shows the NIR absorption spectra of Er³⁺/Yb³⁺ codoped monoliths (6% at Yb³⁺). Compared to Figure 3, two observations are evident. Firstly, the absorption at 980 nm is increased by a factor of around 200. Secondly, the high erbium absorption around 1540 nm is strongly reduced by ~80%, in all samples revealing interactions at the atomic level between both Er³⁺ and Yb³⁺ ions. The only explanation we found to this behaviour involves an up-conversion process from the ⁴I_13/2 to reach the ⁴I_11/2, by absorption of a second photon at 1540 nm. From this excited level, the Er³⁺ ions can either undergo a nonradiative relaxation to the ⁴I_13/2 state followed by a radiative transition to the ground level or process a nonradiative transfer to neighbouring Yb³⁺ ions (²F_5/2 excited level). Comparison of Figures 3 and 5 clearly suggests the involvement of Yb³⁺ ions in the up-conversion process of Er³⁺ ions. However, to our knowledge, no study has previously reported the implication of Yb³⁺ ions in the up-conversion process of the Er³⁺ ions when excited at 1540 nm. This important physical process, highlighted for the first time in this paper, will be further investigated by the authors and further clarification will be proposed in a future study.

3.3. Fluorescence

Figure 6 shows the emission spectra in the near infrared region of Er³⁺/Yb³⁺ codoped bulk samples. Er³⁺ and Yb³⁺ are, respectively, pumped to the ⁴I_11/2 and ²F_5/2 excited levels employing a Ti: sapphire laser tuned to 980 nm. Er³⁺ ions first decay nonradiatively to the ⁴I_13/2 level followed by a final decay to the ground state ⁴I_15/2 by emitting a photon at 1530 nm (Figure 5).

Both spectra show a main photoluminescence peak at 1530 nm and two shoulders around 1500 and 1550 nm with wide tails extending from roughly 1450 to 1650 nm. The peak shape is attributed to the stark splitting of the excited state ⁴I_13/2 and ground state ⁴I_15/2 of Er³⁺ ions. Furthermore, the full width at half maximum is abnormally high at around 30 to 50 nm, which is on average of 30% higher than those measured in pure mineral materials [25, 26]. In a first instance, this can be explained by the amorphous structure of the matrix and the codoping with Yb³⁺ ions. Oppositely to crystalline materials, which can precisely define the position of rare-earth ions within a host matrix, the amorphous nature of our glass-like material favours a random distribution, resulting in a broadening of the luminescence signal.

However, the measured photoluminescence intensity is very low for both samples, which we estimated at 1% of that generally obtained in the case of pure silica matrices. Insofar as the material does not contain any OH groups, quenching by these groups is then eliminated. From these results two hypotheses remain possible: firstly, autoquenching of the fluorescence due to agglomeration of the Er³⁺ ions, secondly, the energy dissipation in the matrix caused by a multiphonon relaxation process. To understand the mechanism responsible for the relatively low level of measured fluorescence, we have studied the Er³⁺ structure within the hybrid material by electron paramagnetic resonance and neutron scattering.

3.4. Electron Paramagnetic Resonance

EPR spectra of samples doped with Er³⁺ at 0.5, 1, and 2% at are shown in Figure 7. A single signal was detected for magnetic fields varying from 0 to 4000 mT that exhibit a constant peak to peak width (~8mT) and a progressive increase of the intensity with the increase of the Er³⁺ ion concentration.

To our knowledge, no literature reported any EPR results on Er³⁺ ions incorporated in either organic polymers or hybrid materials to allow a direct comparison with similar amorphous structures. However, compared to Er³⁺ ions incorporated in crystalline structures [27–29], the EPR signals obtained in our material require a magnetic field of approximately 4 times greater magnitude [30]. This suggests that the Er³⁺ ions are strongly linked to the hybrid matrix.

As the paramagnetic character is preserved, the formation of covalent bonds between Er³⁺ and the matrix is unlikely. However, physical interactions with the strong nucleophilic groups of the matrix can explain the observed behaviour. Indeed, the material contains oxygen atoms in the carboxylic functions that can potentially act as complexing agent by electron donor effect, then reducing the paramagnetic character of the active ions in the material.

Compared to crystalline structures that precisely define the position of the rare earth ion [30], the amorphous structure of the hybrid glass can potentially allow a random distribution of Er³⁺ ions within the matrix. However, according to the EPR data, a single peak was observed indicating that the environment of the different Er³⁺ ions in our material is very similar. However, the EPR technique does not give any information about the concentration effect on the evolution of the Er-Er interatomics distances. This is investigated using neutron scattering characterisation.

3.5. Neutron Scattering

Rare earth agglomerates have been reported with dimensions between a few angstroms to several nanometers [31]. Neutron scattering experiments have been conducted to yield information about any structural change associated with the possible formation of Er³⁺ clusters.

Figure 8 shows the neutron scattering spectra for the undoped and Er³⁺ doped samples with 0.5, 1 and 2% at All spectra exhibit a similar behaviour, with a single band the maximum of which was detected at a wavevector of 1.29.10⁻², which progressively decreases with the increase of the dopant concentration.

The stability of the maximum band position confirms that the size structure of the material is invariant, and excludes the formation of any novel phase, whatever the dopant concentration is. This indicates that fluorescence autoquenching by energy transfer between neighbouring Er³⁺ is not the prevalent relaxation process. Moreover, the decrease of the signal intensity is explained by the increase of the Er³⁺ scattering.

Consequently, the weak fluorescence intensity measured is not due to nonradiative relaxation by either the erbium structure or OH groups. We propose that it is associated with the energy dissipation within the matrix by multiphonon nonradiative relaxation process. In our case, CH groups (ν(C-H) ~2800–3100 cm⁻¹) contained in the organic part of the hybrid, are the only species capable of competing with the radiative emission of Er³⁺ ions around 1550 nm, by a two- or three-phonon relaxation process.