Cordula Braun*a,
Liuda Mereacrea,
Zheng Chenb,
Adam Slabonc,
David Vincentd,
Xavier Rocquefelted and
Jean-François Halete
aKarlsruhe Institute of Technology (KIT), Institute for Applied Materials (IAM), Herrmann-von-Helmholtz-Platz 1, D-76344 Eggenstein-Leopoldshafen, Germany. E-mail: Cordula.Braun@kit.edu
bInstitute of Inorganic Chemistry, RWTH Aachen University, Landoltweg 1, D-52056 Aachen, Germany
cChair of Inorganic Chemistry, University of Wuppertal, Gaussstr. 20, 42119 Wuppertal, Germany
dUniv. Rennes – CNRS, Institut des Sciences Chimiques de Rennes, UMR 6226, 35042 Rennes, France
eCNRS – Saint-Gobain – NIMS, IRL 3629, Laboratory for Innovative Key Materials and Structures (LINK), National Institute for Materials Science (NIMS), 1-1 Namiki, Tsukuba 305-0044, Japan
First published on 10th November 2022
Phosphor-converted white light emitting diodes (pc-LEDs) are efficient light sources for applications in lighting and electronic devices. Nitrides, with their wide-ranging applicability due to their intriguing structural diversity, and their auspicious chemical and physical properties, represent an essential component in industrial and materials applications. Here, we present the successful adsorption of Eu and Tb at the grain boundaries of bulk β-Si3N4 and β-Ge3N4 by a successful combustion synthesis. The adsorption of europium and terbium, and the synergic combination of both, resulted in intriguing luminescence properties of all compounds (red, green, orange and yellow). In particular, the fact that one host can deliver different colours renders Eu,Tb-β-M3N4 (M = Si, Ge) a prospective chief component for future light emitting diodes (LEDs). For the elucidation of the electronic properties and structure of β-Si3N4 and β-Ge3N4, Mott–Schottky (MS) measurements and density functional theory (DFT) computations were conducted for the bare and RE adsorbed samples.
Doped gallium nitride (GaN) has been the benchmark in the past decades in this domain and has indeed driven the LED revolution in lighting and displays as a key material. The concept of down-conversion of a GaN-based blue LED, being awarded the Nobel prize in 2014 in physics,3–5 offers the possibility to provide efficient generation of monochromatic, high-colour purity light resulting in a highly efficient warm-white all-nitride phosphor-converted light emitting diode (pc-LED). The combination of the lower energy consumption, high light quality, preservation of colour point stability and long lifetime is one of the key benefits of pc-LEDs, having the potential to reduce global energy consumption in the lighting sector substantially.
Silicon nitride (Si3N4) high performance ceramics are used in numerous applications because of their superior mechanical properties.6,7 The wide-ranging applicability of nitrides and their related (oxo)nitridosilicate family can be ascribed to their significantly extended structural varieties as well as their auspicious chemical and physical properties (very high chemical and thermal stability, very high quantum efficiency, very low thermal quenching).
In particular, Eu2+-doped (oxo)nitridosilicates and SiAlONs have been amply studied as important host lattices for phosphor-converted light-emitting diodes (pc-LEDs).8–25
For instance, M2Si5N8:Eu2+ (red-orange, 2-5-8 phosphor) and MSi2O2N2:Eu2+ (yellow-green, 1-2-2-2 phosphor) (M = Ca, Sr, Ba) were significant discoveries in this field. Several Mg-nitridosilicates and nitridoaluminates were also developed as next generation high efficient red emitting phosphor materials with superior luminescence properties.26–30
The next step in phosphor development is the investigation of novel host materials for narrow-band emitting phosphors to enhance luminous efficacy and improving therefore the quality of light emitting diodes (LEDs) for diverse applications upon doping.
Silicon nitride (Si3N4) materials have been found to meet these requirements due to its rigid lattices with highly covalent network and high thermal stability.31,32 However, looking at the literature indicates that there are only a few investigations concerning rare-earth doped silicon nitride materials. In most cases only Eu- and Tb-doped α-Si3N4 thin films and nanowires illustrating broad band emission are concerned.33–41 Here we use the expression doping explicitly as this is given in the references.
There are many works in the literature regarding the influence of rare-earth oxides additives often defining the morphology of β-Si3N4 crystallites growing in a multiphase ceramic, thereby affecting its microstructure and mechanical toughness of the ceramic.42–49 Densification additives (e.g., rare-earth (RE) oxides) play an important role in the fabrication of silicon nitride-based ceramics. Indeed, the affinity of the RE cations to segregate and adsorb on the prismatic planes of the hexagonal grains in β-Si3N4, exhibiting very anisotropic shapes, is used to develop microstructural features (e.g., the initiation of the α,β-transformation and the formation of elongated reinforcing grains) required for particular applications.43
Phosphors are usually doped with Eu2+ and only one activator ion is used. Therefore, one host shows one colour. The fact that nitride bulk material can be doped with several rare-earth activator ions, even at the same time, is absolutely new resulting in the fact that one host can deliver different colours.
As well novel is the fact that a pure nitride host material shows green luminescence,31,32 till now this could only be realized by oxynitrides and oxonitridosilicates.18,50–53
Just recently we demonstrated that the so called “yellow gap” could be closed for bulk GaN via co-doping with europium and terbium.32 Co-doping really opens up a multitude of degrees of freedom to customize and adapt a luminescent material to specific needs.
Herein, we experimentally and theoretically examine the adsorption of Eu and Tb at the grain boundaries of β-Si3N4 and β-Ge3N4 and discuss their resulting intriguing luminescence properties.
At this point we would like to point out that we explicitly talk here of adsorption of the RE elements and not of doping. In the nitridosilicate Ca2Si5N8:Eu2+ the RE cations go the Ca site of the host structure, which is named doping. But there is no analogous cation site in β-Si3N4 and β-Ge3N4.
As described in detail in literature43,44,49 the RE cations go to the grain boundaries in β-Si3N4. Therefore we choose the term Eu adsorbed-Si3N4 and not doped as it is used for e.g. Eu-doped nitridosilicates. For a detailed discussion concerning this topic we would like to refer to lit.43,44,49 and the section below, where a detailed explanation of the RE adsorption sites is given.
(For β-Si3N4 and β-Ge3N4 adsorbed with europium and/or terbium we would like to introduce the notation Eu,Tb-β-M3N4.) The results indicate that Eu,Tb-β-Si3N4 (ref. 31) and as well as Eu,Tb-β-Ge3N4 (ref. 31) should be considered as prospective chief components for highly efficient warm-white all-nitride phosphor-converted light emitting diode (pc-LED).
Fig. 1 (a) X-ray powder diffraction patterns of pure β-Si3N4 (blue), adsorbed Eu-β-Si3N4 (red) and the bar graph of β-Si3N4 ICSD [33-1160] (bright blue), (λ = 0.709026 Å), (b) structure of α-Si3N4 view along [001], (c) cavity for a possible insertion of a RE cation in α-Si3N4, (d) structure of β-Si3N4 view along [001], (e) six-ring representation of β-Si3N4, (f, g and h) structure of β-Si3N4 view along [001] with three possible RE adsorption sites A, B, and M.44 (Si atoms are depicted in yellow, N atoms in blue, the RE adsorption sites A, B, and M in orange, red and light blue). |
For a comparison of the powder diffraction patterns of Eu-β-Si3N4 and Tb-β-Si3N4 see Fig. S1,† for Tb-β-Ge3N4 and Eu,Tb-β-Ge3N4, see Fig. S3.† SEM EDX measurements of β-Si3N4 and β-Ge3N4 confirmed the atomic ratio of M:N (M = Si, Ge) of 3:4 and europium and terbium contents of 3–5% were found (see Fig. S4†).
Understanding the influence of selective dopant additions and the role of interfacial interactions is central to the design of novel high-performance Si3N4 ceramics by offering the potential for customizing the materials properties. Rare-earth cations are often located within regions of the oxynitride glassy phase of triple-junction pockets, disordered amorphous nanometer scale intergranular films (IGF) and at the glass or IGF/β-Si3N4 grain interfaces.44
According to the literature44,49 there are three independent stable RE adsorption sites per surface unit cell along the N-terminated prismatic planes of β-Si3N4 (see Fig. 1f–h). A and B are stable RE equilibrium sites, while calculations pretended the M site to be theoretically unstable, although observed experimentally in the case of La adsorption.44 Stereochemical bonding factors are found to determine the adsorption site preferences contrary to ionic size effects, and the strength of the rare-earth interface bonding is defined inter alia by the electronic structure of the nitride surface. Shibata et al.49 showed that these RE adsorption sites have higher binding energies than Si, partially reside on cation sites normally only available for Si and that the RE–N bonds are longer than comparable Si–N bonds. While in β-Si3N4 an adsorption of the RE ions at the grain boundaries is possible (see Fig. 1f–h), in α-Si3N4 an insertion of cations into the structure (see Fig. 1c) takes place. In the α-Si3N4 structure there are two caves (for the bigger one see Fig. 1c), which can be occupied when charge is balanced by cation and/or anion substitutions (e.g., Si by Al and N by O). This additional insertion of an RE ion is stabilizing the α-Si3N4, which is otherwise metastable.54 And indeed, no insertion of cations is realized into the six rings in β-Si3N4 (see Fig. 1e).
An important point here is that normally the RE ion of luminescent phosphors is inserted during the main synthesis and not afterwards. In general, doping of nitridosilicates is performed with Eu2+ and only one activator ion is used for one luminescent host. Blending the colours within one host and colour tuning by mixing different coloured luminescent hosts should open up tremendous opportunities for highly efficient phosphors. Within the scope of this work, β-Si3N4 and β-Ge3N4 were adsorbed with Eu and Tb activator ions each individually and both simultaneously.
In the CIE 1931 diagram (see Fig. 2e) the chromaticity coordinate positions of Eu- and Tb-β-Si3N4 and β-Ge3N4 are indicated, proving that a colour tuning of one host with different activator ions and their combination is feasible.
Fig. 2 Excitation (black) and emission spectra (coloured) of (a) Eu-β-Si3N4 (red, λem = 612/614 nm), (c) Tb-β-Si3N4 (green λem = 544 nm) and of (b) Eu-β-Ge3N4 (red, λem = 613 nm), (d) Tb-β-Ge3N4 (green λem = 544 nm) and (f) Eu- and Tb-β-Ge3N4 (orange λem = 544/614 nm), (e) CIE 1931 diagram of β-Si3N4 (♦), β-Ge3N4 (▼) and GaN ()31,32 adsorbed/doped with Eu, Tb and Eu and Tb. |
Taking into account the basics of colour mixing, it is clear that the combination of green and red leads to orange (see Fig. 2f). This effect could be proven in the meantime for other doped nitrides31 e.g., GaN31,32 (see Fig. 2e), and carbodiimides as well.55
Fig. 2a shows that in the emission spectrum of Eu-β-Si3N4, the transitions 5D0 → 7F0 and 7F3 seem to be more intense than usually, and the latter is even more pronounced than the 5D0 → 7F4 transition, which is not common. This means a strong J-mixing and a strong crystal-field perturbation might occur in this matrix. The peak corresponding to the 5D0 → 7F0 transition is also broad, indicating the location of Eu3+ ions on several sites in the host structure (see Fig. 2a).
Fig. 2f–c proves that for Eu,Tb-Ge3N4 two different activator ions can be adsorbed in one host showing the typical bands of Eu3+ as well as the ones of Tb3+ in one spectrum resulting in a saturated orange body colour. However, this is not only a superposition of the Eu3+- and the Tb3+-spectrum of β-Ge3N4 because peak form, intensity and wavelength of this orange spectrum differ clearly in comparison to the spectra of the single doped materials. Therefore, this orange colour is only possible by mixing the ions at an atomic scale and cannot be realized simply by a mixture of particles of the red and the green adsorbed Ge3N4.
To get further evidence powder samples of Eu-Ge3N4 and Tb-Ge3N4 (same molar ratio Eu:Tb as in Eu,Tb-Ge3N4) were mixed in a mortar. This definitely did not result in an orange luminescent β-Ge3N4 sample but a yellow luminescent powder instead (see Fig. S5c and d†). Indeed, compared to a superposition of the Eu- and Tb-spectra, all peaks of the Eu,Tb-β-Ge3N4 spectrum slightly changed in energy and intensity with the peaks resulting from Tb3+ significantly stronger and tuning the colour to the yellow region (CIE coordinates x, y = 0.456, 0.499). This clearly evidenced that the orange body colour (CIE values x, y = 0.559, 0.438) can only be obtained by an atomic scale mixing and that we are able to tune the colour resulting finally in red, green, orange and yellow luminescence for Eu- and Tb-β-Ge3N4. As the mixing at the atomic scale for Eu,Tb-β-Si3N4 was not successful, the combination of the powder samples of equal parts of Eu-β-Si3N4 and Tb-β-Si3N4 was tested. Here we got a completely different result and the CIE diagram shows a saturated orange colour with the coordinates x, y = 0.570, 0.417. (see Fig. S5a and b†). Here we find nearly the same CIE values as the ones of the amber emitting 2-5-8 nitridosilicate phosphor (Ba,Sr)2Si5N8:Eu2+ (x, y = 0.579, 0.416) being considered as an important breakthrough for bridging the “yellow gap”.25
For Eu,Tb-β-Ge3N4 the FWHM pointed out the same value as those of the single doped materials which are in the range of 11–12 nm. This holds true for Tb-β-Si3N4 as well. A much more narrow band is observed for Eu-β-Si3N4 with a FWHM of 2–4 nm. Interestingly, these values are quite smaller than the line widths (FWHM) of the emission spectra of the very narrow-band nitridosilicate phosphors which range between 35–50 nm2 or those of the nitride-based LEDs which vary typically between 20–35 nm.60 The quantum yields were in the range of 5–16%. Tb-β-Si3N4 was 16% and the other compounds around 5%.
To elucidate if the luminescence does not result from the respective rare-earth (RE) compounds and if the Eu and Tb ions really have been adsorbed at the grain boundaries of β-Si3N4 and β-Ge3N4 a comparison of the luminescence spectra of Eu-Si3N4 and Ge3N4 and EuCl3·x6H2O is shown in Fig. S6 and S7.† A comparison between Tb-β-Si3N4 and TbOCl61 is also shown in Fig. S15.† Optical and luminescence spectra are highly sensitive to structural deformation of the nearest environment of RE ions. Therefore, it becomes clearly evident here that the emission spectrum has changed due to the adsorption of the Eu cations into the β-Si3N4 and β-Ge3N4 structures.
A very detailed comparison of our results with those of several Eu- and Tb-Si3N4 thin films and nanowires and possible side products published in the literature was carried out (see ESI Fig. S8–S15† and corresponding section). This indicated that if some luminescence spectra may seem very similar overall at first sight, it turns out that definitely some changes are observed.33–41
The β-Si3N4 models were generated in such a way to reproduce the interface observed in high-angle annular dark-field STEM (HAADF-STEM) images of RE doped Si3N4 by Ziegler et al.,63 which evidenced a grain orientated along the [0001] zone axis with the prismatic plane of Si3N4 facing the amorphous intergranular phase. In particular, the rare earth elements have been observed sitting on two atomic sites at the interface, labeled A and B, which are, respectively, small and large open hexagonal rings Fig. 3.
Note that the creation of the interface led to dangling bonds, which were passivated using both O2− and N3− ions, in order to respect the electroneutrality of the unit cell. For both Eu and Tb ions on sites A and B, atomic coordinates were relaxed leading systematically to a higher stability when the rare earth resides on site A (highly coordinated) than on site B, with an energy difference of 745 and 851 meV per rare earth for Eu3+ and Tb3+, respectively.
For the more stable interface, i.e., with the rare earth located on site A, simulations of the Si L23 edge were performed for silicon atoms close to the interface in order to probe both their location and chemical environment. Indeed, precise electron energy loss spectroscopy (EELS) measurements were reported for RE doped Si3N4 ceramics,63 allowing us to compare our model with experimental data. Such a simulation required to properly treat the electron–hole interaction (excitonic effects) using Bethe–Salpeter equations for instance. Here, due to the size of the system, we chose to consider this electron–hole Coulomb interaction as a static screening using the core–hole approximation. More precisely, half an electron was removed from a core orbital, i.e., the Slater's transition state.64 For the doped Tb-β-Si3N4 model, the core-hole was introduced in the 2p1/2 and 2p3/2 states of the excited silicon atoms. The simulation is thus a summation of two spectra resulting from two static calculations with, respectively, half a core-hole in 2p1/2 and 2p3/2 states of the probed silicon atoms. In order to validate such a static screening approximation, the related bulk system Si2N2O was simulated and compared to experiments63 (see Fig. 4a). A good agreement is observed, with a small discrepancy around 104 eV both in terms of peak position and intensity. Such an agreement confirms that the Slater's transition state allows to properly describe the Si L23 edge of silicon atoms surrounded by nitrogen and oxygen atoms.
Fig. 4 Simulated Si L23 EELS data (red and blue lines) compared to measurements (open circles) from ref. 63 of (a) Si2N2O and (b) the adsorbed Tb-β-Si3N4 model with Tb ions on A site. The simulations have been done using the Slater's transition state. For the interface, two simulations have been done for one silicon atom surrounded by N3− ions only, and one silicon atom with O2− ions in its first coordination sphere. |
Fig. 4b shows two simulations corresponding to Si L23 edge of two silicon atoms nearby the interface with Tb ions on site A. One silicon atom, labeled Si1 (Fig. 4b) is surrounded only by nitrogen atoms while the other one, labeled Si2 (Fig. 4b) is environed by oxygen and nitrogen atoms. Note that the experimental spectrum (open circles) related to site A is characterized by a double peak (102 and 103.5 eV), with a first peak assigned to Si–N bonds and the second one to Si–O bonds. Interestingly, our simulations evidenced a peak at 103 eV for Si1 and 104.1 eV for Si2, confirming the previous interpretation and validating the model used to mimic locally the interface in the grain boundary region.
DFT + U + SOC calculations were then carried out on the adsorbed Tb-β-Si3N4 model considering the previously optimized atomic structures. Such calculations were indeed not trivial and could be difficult to converge when the ground state is close in energy to other magnetic states. The convergence of these calculations was achieved only for the Tb compounds. Spin (orbital) moments of 6.03 (1.45) and 6.04 (2.73) μB were computed for Tb ions on sites A and B, respectively, leading to total magnetic moments of 7.48 and 8.77 μB, respectively. These values are somewhat smaller than the theoretical value of 9 μB expected for a free Tb3+ ion. This is due to the crystalline field of the surrounding ligands which is stronger for Tb ions on site A than on site B. On site A, the most energetically preferred site of Tb is positioned in an octahedral environment, while on site B, Tb is sitting in a too large site, which creates a less intense crystalline field. This explains the weaker moment value, in particular the orbital moment of the former with respect to the latter.
Fig. 5 shows the densities of states (DOS) obtained with Tb3+ atoms on A and B sites computed at the DFT + U + SOC level of theory. The valence band (VB), from −12 to 0 eV, is mainly based on N 2p states interacting with Si(3p) states. The O 2p states of the interface appears in the VB from −10 to −2 eV. The conduction band, starting at 2 eV, is mainly composed of Si 3p states interacting with N 2p states. The band gap is only 2 eV in both cases. For comparison, the DFT estimated band gap of β-Si3N4 was 4.2 eV. The decrease in the band gap value is indeed the consequence to additional interactions involving the Tb 5d states found both at the top of the VB and the bottom of the conduction band (CB) (see Fig. 5).
Such a band gap reduction was already discussed by Huang et al.40 for Y-doped Si3N4, due to Y 4d states. These reduced band gaps are directly responsible for the luminescent properties of the adsorbed RE-β-Si3N4 (RE = Eu, Tb) materials.
Especially, the fact that one host can deliver different colours renders Eu,Tb-Si3N4 and Eu,Tb-β-Ge3N4 as prospective chief components for future light emitting diodes (LEDs). Not only several colours could indeed be realized by an atomic scale mixing, but the colour could also be tuned by mixing the adsorbed hosts enlarging the colour range with red, green, orange and yellow luminescence for the extremely narrow-band Eu- and Tb-M3N4 materials (M = Si, Ge). It could be shown that the realization of an amber emitting phosphor for both, Si3N4 and Ge3N4, is possible.
This work also studied the types and the flat band edges of β-Si3N4 and β-Ge3N4 before and after adsorption with RE materials were also studied in this work. A detailed physical characterization by MS analysis of EIS revealed that a significant shift of the flat band edge was observed caused by the adsorbing with Eu and Tb ions. This can be seen as a complementary strategy to modify the band edge of materials. Moreover, this could be applicable to photoelectrodes used to adapt and optimize (photo)electrochemical performances.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d2ra04663f |
This journal is © The Royal Society of Chemistry 2022 |