1 Introduction

Lighting accounts for about 19 % of electricity worldwide, and new environmentally friendly and energy efficient lighting devices are strongly required. It is widely accepted that solid state lighting (SSL) will replace the old technologies that have been phased out (incandescent lamps) or are problematic from an environmental point of view (compact fluorescent lamps) [1]. The invention of bright-blue LEDs (light emitting diodes) by Nakamura and others in the mid-1990s [2, 3] has led to the development of devices capable to produce white light in a reliable and efficient way, based on inorganic phosphors [4]. Presently, pcLED (phosphor converted light emitting diode) lamps are characterized by a number of important advantages compared to older generations of lamps, as they contain no highly toxic mercury [5], turn on instantly, are mechanically robust, have a longer life expectancy (up to 25,000–30,000 h) and especially have a higher energetic efficiency. In principle, white light could be obtained by combining the three fundamental colors (red, green, and blue, RGB) generated by three distinct LEDs. This method employs multiple LED chips and has a major disadvantage due to the high production costs. For this reason, it is currently preferred to use only one LED combined with suitable phosphors.

Presently, two families of LEDs are easily available on the market: (a) blue emitting LEDs (around 450–480 nm) and (b) near ultraviolet (NUV) emitting LEDs (around 380–410 nm), as described specifically in Sect. 2. The two methods currently offer the production of devices with comparable brightness.

Inorganic phosphors normally consist of a crystalline host material doped with a relatively small amount of rare earth or transition metal ions, called activators, which convert light from shorter wavelength to longer wavelength (Stokes shift) via electronic transitions. In the case of LED lighting, the short wavelength excitation is provided by a blue LED or a NUV LED, and therefore, radiation in this spectral range must be strongly absorbed by the phosphor(s).

This contribution presents a description and survey, taken from the recent literature, of some important phosphor materials that find current or perspective applications in the generation of white light using LED excitation. It is intended to be complementary to the many excellent review articles that have appeared in the literature (e.g., [413], to cite only a few).

The chapter is organized as follows. After the introduction, a section will deal with generalities of pcLEDs, followed by a description and discussion of advanced techniques of structural and dynamical characterization of inorganic phosphor materials. A selection of important LED phosphor materials will be then presented, separated on the basis of the various emission colors (Fig. 1). We choose the following wavelength ranges: blue (435–500 nm), green (520–565 nm), yellow (565–590), and red (625–740 nm) [14]; of course this is a very qualitative and approximate description, chosen for the sake of simplicity. The phosphor selection is based on our personal choice and the coverage of this huge research field is not at all intended to be exhaustive. We shall try to give an idea of some of the materials and the concepts used in this rapidly evolving research area. Conclusions and perspectives will close the chapter.

Fig. 1
figure 1

Approximate color regions on the Commission Internationale de l’Éclairage (CIE) 1931 chromaticity diagram, adapted from [15]

2 Phosphor Converted White Light Emitting Diodes

Two of the main types of pcWLEDs are based on (1) a NUV LED chip coated with a mixture of RGB phosphors (Fig. 2a) and (2) a blue LED chip coated with a yellow phosphor (Fig. 2c), with their typical emission spectra shown in Fig. 2b and d, respectively. For the pcWLED in Fig. 2c, the yellow phosphor can be replaced by a mixture of green and red phosphors to have a similar emission profile as depicted in Fig. 2b. Two parameters are useful to define the brightness and the color emitted by a pcWLED, the luminous efficacy (LE) and the color rendering index (CRI). The LE expresses how well a light source produces visible light, whereas the CRI measures quantitatively the ability of a light source to reproduce the colors of various objects faithfully in comparison with an ideal or natural light source. Figure 2e shows the coordinates of emitted visible light from NUV- and blue-LED based pcWLEDs in the CIE 1931 chromaticity diagram, where the correlated color temperature (CCT) is indicated along the blackbody locus. The higher the CCT is, the colder human beings perceive the light. For comparison, incandescent lamps, as reference sources based on blackbody radiation, have a CRI of 100. A high value of CRI is required for high quality and comfortable lighting.

Fig. 2
figure 2

Schematic depiction of pcWLEDs based on a a NUV-LED and a mixture of RGB phosphors, and c a blue LED combined with a yellow phosphor, and their corresponding emission spectra in (b) and (d). e CIE 1931 chromaticity diagram indicating the CIE coordinates converted from the emission spectra in (b) and (d). f A schematic comparison of device efficiency with respect to CCT between NUV- and blue-LED based pcWLEDs, adapted from [6]

A comparison between the two types of pcWLEDs (Fig. 2e) shows that the white light emission from the blue-LED based pcWLEDs containing a yellow phosphor has a higher CCT (7002 K) than that of the NUV-based one (4521 K), where the CCT values are calculated in terms of the chromaticity coordinates shown in Fig. 2e, based on the cubic approximation proposed by McCamy [16]. The CCT values here reveal that the light emitted from NUV-based pcWLEDs is perceived as warmer light than the one from blue-based pcWLEDs. Additionally, the NUV-based LEDs perform with higher efficiency for warmer white light emissions (Fig. 2f). On the downside, they are more difficult to fabricate and hence of higher cost. We also point out that pcWLEDs emitting warm white light can be obtained via exciting green and red phosphors by blue LEDs, with having high efficiency due to the small difference between excitation and emission wavelengths. For the latter reason, many investigations focus on improving the technology based on blue-LEDs, notably with the specific aim of lowering their CCT as well as increasing their CRI. Strategies to improve these properties include for example the addition of highly efficient red components made of InP quantum dots [6]. A compilation of luminescence data for various inorganic phosphors used in pcWLED devices is shown in Table 1. The CIE 1931 coordinates of some selected phosphors for NUV- and blue-LED based pcWLEDs are indicated in Fig. 3.

Table 1 Phosphors (red: 625–740 nm, green: 520–565 nm, blue: 435–500 nm) for NUV-based pcWLEDs
Fig. 3
figure 3

The CIE 1931 diagram with coordinates of different phosphors typically used in a NUV-LED based and b blue-LED based pcWLEDs, respectively

3 Advanced Structural and Dynamical Characterization of Inorganic Phosphor Materials

3.1 Energetics of the Activator Ions and the Effect and Importance of the Structure and Dynamics of the Host

Although the activator ions, most often lanthanide ions (Ce3+–Yb3+, denoted as Ln3+, and Eu2+ and Yb2+) have intrinsic characteristics that contribute to the optical properties of phosphors, the electronic energy levels of an activator ion in a crystal differ greatly from those of a free ion. The following discussion refers specifically to activator ions showing 5d–4f luminescence, such as Ce3+ and Eu2+.

The energy separation of the energy levels can give rise to emission of light from UV across visible wavelengths, depending on the properties of the host crystal. In fact, the local geometry around the activator ion has a remarkable importance in determining the spectroscopic behavior of lanthanide ions incorporated in a host matrix. In particular, the lowering of the energy gap between the 5d and 4f levels, which is commonly referred to as the redshift, is dictated by two major effects; the centroid shift and the ligand field splitting. Figure 4a shows these effects for Ce3+ embedded in yttrium aluminum garnet, Y3Al5O12 (YAG), as an example; the total redshift is labeled D(A) where A refers to the host crystal [4]. As can be seen, both of these processes greatly affect the 5d electrons of the activator ion, whereas the well-shielded 4f electrons are not strongly affected. The position of the 5d levels can indeed change by several eV from one compound to another.

Fig. 4
figure 4

a Energy diagram of Ce3+ in the free-ion state and when incorporated in a host; the figure is adapted from [39]. b Schematic configurational coordinate diagram for Ce3+ in YAG. Notice that non-radiative relaxations may occur when the 5d electrons are thermally populated to the crossover point of the 5d 1 and 4f parabolae

The first effect, centroid shift, refers to a lowering in the energy of the 5d levels of the activator ion due to a decrease in the interelectron repulsion [4]. This effect depends on the polarizability of the surrounding anion (oxygen) ligands and on the covalency of the chemical bonds between the ligands and the activator ion [4, 38]. The centroid shift increases with increasing anion polarizability and decreasing average electronegativity of the host cations. As the degree of covalency between the activator ion and surrounding anions increases, the electrons of the metal ion are partly delocalized on the ligands and this decreases their interelectronic repulsion, therefore, decreasing the energy of the excited levels of the metal ion. This feature is called the nephelauxetic effect (Greek for cloud-expanding).

The second effect, ligand field splitting, refers to the difference in energy between the highest and lowest 5d levels, referred to as yDq, where y relates to the type of cubic coordination geometry of the activator ion and is again an effect of the host crystal. Its magnitude depends on the bond lengths from the activator ions to the coordinating oxygen ligands, the molecular overlap or degree of covalency between the activator ion and its ligands, the coordination environment, and the symmetry of the activator-ion sites [4, 40]. Dq, which is proportional to the splitting energy in high symmetry (cubic) environments can be estimated on the basis of a point-charge model according to the following expression:

$$ {\text{Dq}} = \frac{{Ze^{2} r^{4} }}{{6R^{5} }}, $$

where Z is the charge or valence of the anion, e is the charge of an electron, r is the radius of the d wave function, and R is the bond length between the activator ion and its ligand [4, 40]. Comparing different coordination geometries with the same R, Rogers and Dorenbos [41] determined that the ligand field splitting tends to be largest for octahedral coordination, followed by cubal coordination. The smallest ligand field splitting was found for tetrahedral and cuboctahedral coordinations [41].

Furthermore, whereas in many oxide hosts the activator ions occupy well-defined substitutional sites, predictable on the basis of simple arguments based on the charge and the size of the dopant and the substituted ions, this is not the case for all materials. A notable example of this behavior is the non-linear host LiNbO3 [42], which is of great interest for its ability to generate the second harmonic radiation of the light emitted from Ln3+ ions and for which the Ln3+ ions have a different charge with respect to the host cations (Li+ and Nb5+) and also a different ionic radius (Li+ and Nb5+ are smaller than the Ln3+ ions [43]). It is clear that in such cases the local structure around the activator ions is not easily predictable and the simultaneous occupation of multiple substitutional sites cannot be excluded. In this perspective, the nature of the sites occupied by Ln3+ ions has been widely discussed and in the literature many contrasting opinions have been reported. This situation occurs for many classes of host matrices of technological interest and in particular for garnet type host lattices [4446], such as YAG, which are described in more detail in Sect. 4.1. Moreover, it is well known that garnet crystals (in particular YAG) are subject to significant amounts of native defects that appear during crystal growth depending on the temperature and crystallization procedure [47, 48]. Among these intrinsic defects, antisite defects are dominant. They appear in YAG when some Y and Al atoms exchange positions. The antisite defects are known to act as electron traps and to affect the structure, luminescence, and other properties of doped and undoped YAG. Moreover, in garnets not containing a large trivalent cation, such as Ca3Sc2Si3O12, the localization of the trivalent dopant is not generally trivial [32]. It can be concluded that it is important to get more insight on this subject.

Additionally, the energetics of the activator ions is determined by the Stokes shift, which refers to the difference in energy between the band maxima of the absorption and emission spectra. This can be understood by the use of a configurational coordinate diagram, in which the potential energy curves of the activator ion in the ground and excited state are plotted against a so-called configurational coordinate that represents the normal coordinate of a vibrational mode localized at the optical center, see Fig. 4b. The potential energy curves of the electronic levels have a parabolic shape in the likely case of a harmonic behavior. The horizontal lines within the parabolae depict the vibrational states relative to each electronic state. According to the Franck–Condon principle, the electronic transitions (absorption and emission) are vertical in nature and (at low temperature) occur from the lowest vibrational level. In effect, this means that once an electron is excited to the 5d level, the surrounding lattice first relaxes to the lowest vibrational level in the electronic excited state before the electron is de-excited to the electronic ground state. Since the Stokes shift requires the interaction with phonons, this property is also dictated by the host crystal.

The phonons also play a key role in a phenomenon known as thermal quenching, which describes the pronounced reduction in emission intensity, or quantum efficiency, observed at elevated temperatures, typically a few hundred degree centigrade. Since LED chips produce heat, which may easily get transferred to the phosphor coating, this is of pivotal importance with respect to the energy efficiency of pcLED devices. On a fundamental level, thermal quenching relates to non-radiative relaxation mechanisms of the excited-state electron, other than concentration quenching that relates to the activation of energy processes involving the excited states of the Ln3+ ion and depopulating in a non-radiative way the light emitting levels. In the case of Ce3+ dopants, the two primary mechanisms are thought to be thermal promotion of the 5d electron to the conduction band of the host crystal, and 5d–4f crossover, of which the latter is illustrated in Fig. 4b. Regardless of the mechanism, phonons are needed to bring the excited activator ion to a point at which non-radiative processes can happen.

3.2 Advanced Structural and Dynamical Characterization

As can be understood from above, it can be concluded that a detailed insight into the composition-structure-dynamics relationships in phosphors is crucial for the understanding of their properties and hence for the development of new, next-generation, better performing, materials. This requires the combination of several, complementary, techniques, which enable precise structural and dynamical analysis of substituent ions at small concentrations of approximately 1 mol%. Examples of two powerful techniques for exactly this purpose are neutron and X-ray total (Bragg + diffuse) scattering, and X-ray absorption fine structure (XAFS) spectroscopy. Compared to “conventional” diffraction, i.e., where only the Bragg peaks are considered, the diffuse scattering between the Bragg peaks relates to the local structural details (e.g., symmetry, bond angles and distances) of the investigated materials, and can be analyzed by e.g., reverse Monte-Carlo or pair-distribution function (PDF) analysis [49]. Similar results can be obtained from XAFS spectroscopy, but here the local surrounding, such as the distortive nature, bond distances and angles, and covalency of the chemical bonds, around specific Ln3+ ions are highlighted. Complementary information about the local structure can be obtained for example from vibrational (Raman and infrared) spectroscopy, and inelastic neutron scattering techniques. All together, this allows obtaining information about the real local structure of the sites occupied by the activator ions at a much deeper level than is possible by “conventional” diffraction measurements and simple crystallographic considerations based on ionic radii and charges.

3.3 Case Studies

There have been numerous studies aimed to understand the composition-structure-dynamics relationships of phosphors, many of them focusing on the correlation between emission color to the local or long-range structures of the material. Examples include the study by Robertson et al. [50], who reported on the relation between emission spectra and cation size in Ce3+-doped YAG, the study by Pan et al. [51], who reported on the correlation between the red-shift of emission to the long-range crystalline structure of the same material, and the study by Furman et al. [52], who showed that the quantum efficiency of YAG:Ce3+ is related to the degree of crystallinity. Other, more recent examples include the study by Ghigna et al. [53] and George et al. [54], who examined the local structure of the Ce3+ ion using extended XAFS (EXAFS) and found that the nearest eight Ce–O distances are expanded by about 3 % compared to the average Y–O distances in Y2.97Ce0.03Al5O12, and Gracia et al. [55] and Muñoz-García and Seijo [56], who showed that the substitution of Ce3+ for Y3+ in YAG results in a small expansion of the coordination environment around the substituent ion. Further examples include the study by Wu et al. [57], who reported on a relationship between Ga-induced lattice expansion in Y3Al5-x Ga x O12 (0 ≤ x ≤ 5) and a corresponding blue-shift of both the excitation and emission spectra. Local structural analysis specifically suggested that the blue-shift effect can be related to a decrease in the so-called distortion parameter d 88/d 81, which is defined in Fig. 5a. The value of d 88/d 81 relates specifically to the degree of tetragonal distortion of the nearly cubic-coordinated Ce3+ ions in YAG. Thus, a smaller d 88/d 81 has the effect of making the Ce3+ local structural environment more cubic-like, which in turn weakens the ligand fields with tetragonal symmetry and hence the splitting energy \( \Delta_{12} \) as denoted in Fig. 4a. Consequently, the lowest lying 5d state shifts to a higher energy and a blue-shift effect for the 4f–5d transitions is observed. The same phenomenon is observed also in other materials, such as Y3Sc2Ga3-x Al x O12:Ce3+ [58], and is also supported by theoretical simulations [59]. On the contrary, a (counteracting) red-shift for YAG:Ce3+ substituted with Mg2+ and Si4+ to replace Al3+, i.e., Y3Al5-2x (Mg,Si) x O12:Ce3+ has been observed [60]. Although the introduction of Mg2+ and Si4+ ions leads to an elongation of the Ce–O bonds, which should result in a blue-shift, the doping also causes an increase of the Ce–O covalency and furthermore increases the distortion of the CeO8 13− dodecahedra, which in effect lead to an over-compensating red-shift. Furthermore, Tien et al. [61], Chiang et al. [62] and Shao et al. [63] investigated the effect of doping on the Y3+ site in YAG:Ce3+. The interatomic M–O distance, where M is the dopant on the Y3+ site, is found to go as Gd–O > Tb–O > Y–O > Lu–O [64]. A key observation was the observation of a red-shift when the lattice expands and a corresponding blue-shift when it contracts. A general conclusion can be made here that the local structure around Ce3+ is more distorted as larger cations replace Y3+ in YAG:Ce3+, and the local lattice distortion intensifies the ligand fields which strengthen the red-shift effect. The collected data from previously measured luminescence spectra of Ce3+-doped (Gd,Y,Lu)3(Al,Ga)5O12 by Dorenbos [39] indicate that the blue-/red-shift effect is mainly attributed to the splitting energy \( \Delta_{12} \) instead of Dq, cf. Fig. 4(a). The photoluminescence (PL) and PL excitation (PLE) spectra of a series of Ce3+ and Li+-doped garnet phosphors have been investigated by Kalaji et al. [65], and the results for the blue-/red-shift effect originating from different lattice size, bond length and local lattice distortions, are consistent with the conclusions found for (Gd,Tb,Y,Lu)3(Al,Ga)5O12:Ce3+ [64]. In more detail, Wu et al. [57] and Seijo et al. [66] found that the red-shift effect is related to static as well as vibrationally induced tetragonal distortions around the Ce3+ ions. The vibrationally induced distortions refer specifically to symmetric compression and bending motion of Ce–O bonds at the dodecahedral sites (of D2 symmetry). These vibrational modes are depicted in Fig. 5b, c.

Fig. 5
figure 5

a Schematic illustration of the local structure of YAG, with the d 88 and d 81 distances indicated. The d 88 distance is referred to as the length of the O–O edge shared with another dodecahedron, whereas the d 81 distance is the distance shared with one tetrahedron, as defined in Ref. [57]. b, c Schematic illustration of the local, dynamical lattice distortions at the Y site in YAG, adapted from [66]. In (b) the symmetric compression is shown, whilst in (c) the symmetric bending vibration is shown

4 Yellow Phosphors

Yellow emitting phosphors can be used for SSL when they are excited by a NUV (λ ≈ 380–410 nm) or blue (λ ≈ 450–480 nm) source. Emission bands from yellow phosphors are centered in the region of 565–590 nm. Since yellow light can be obtained by combining green and red light, the roles of green and red phosphors in pcWLEDs made of a NUV LED chip and RGB phosphors can be substituted by yellow phosphors. Therefore, yellow phosphors are preferably employed with a blue LED chip to generate white light if a high CRI is not needed. Whereas, when a high CRI is required, RGB phosphors coated on a NUV LED chip or RG (red and green) ones with a blue LED pump are favorable. The activator ions for yellow radiations are primarily Ce3+ and Eu2+. As discussed below in Sects. 5 and 6, the luminescence of Ce3+ and Eu2+ derives from 5d–4f transitions and is red-shifted from the blue/green to the yellow region because 5d states shift to lower energy levels due to stronger interactions between activator ions and host materials. Since 5d–4f transitions are allowed electric dipole transitions, the emission from Ce3+ and Eu2+ is relatively bright and such strong transitions are ideal for pcWLED applications.

Examples of host matrices for the generation of yellow light are oxynitrides, silicates, and aluminates. The most important class of hosts of technological interest is based on garnet type lattices such as YAG, which when Y3+ is substituted with a few mol% of Ce3+ ions, is a luminescent material that has nearly ideal optical characteristics for the generation of white light and has been widely commercially used in pcWLEDs. In the following sections, we describe in detail the general characteristics of structural, dynamical, and optical properties of garnet type host lattices, with a specific focus on YAG:Ce3+, in order to provide a basis for the understanding of this technologically important material.

4.1 Garnet Type Host Lattices and YAG:Ce3+

The garnet structure was firstly reported by Menzer [67]—according to Geller [68]. It has been found that the garnet structure exists in a number of minerals, particularly in oxide minerals. For oxide garnets, the chemical formula can be generally expressed as A 3 B 2(XO4)3, where A, B, X stand for cations, and O is an oxygen anion. Different oxide garnets have been used as host lattices for phosphors, such as silicates (e.g., Ca3Sc2Si3O12), aluminates (e.g., (Gd,Tb,Y,Lu)3Al5O12) and more complex compositions such as Y3Mg2AlSi2O12 and Ca2GdZr2Al3O12, to mention a few. As seen in Table 1, the optical properties of identical activator ions accommodated in different garnet structured host lattices vary substantially. Hence, it is critical to explore the local crystal structure of the host lattice with and without activator ions.

Regarding the canonical compound, YAG, this can be described in terms of a body centered cubic unit cell containing 160 atoms (80 atoms in the primitive cell) in the \( Ia\overline{3} d \) space group, where the primitive cell consists of four molecules of \( {\text{Y}}_{ 3} {\text{Al}}^{\text{VI}}_{ 2} \left( {{\text{Al}}^{\text{IV}} {\text{O}}_{ 4} } \right)_{ 3} \). The superscripts VI and IV refer to octahedral and tetrahedral coordination, respectively. The Y3+ ions thus occupy the 12(c) Wyckoff position (W.P.) and are dodecahedrally coordinated to eight O2− ions, leading to a D2 site symmetry, whereas the Al ions occupy the sites 8(a) and 12(d) and the oxygen ions occupy the site 48(h). It follows that the YAG structure may be viewed as a 3D network of AlVIO6 9− octahedra and AlIVO4 5− tetrahedra, and YO8 13− dodecahedra with shared O2− ions at the corners (Fig. 6a).

Fig. 6
figure 6

a Schematic picture of the YAG unit cell containing tetrahedral AlO4 5−, octahedral AlO6 9−, and dodecahedral YO8 13− moieties, with the W.P.s for the different types of ions indicated in the figure. b Local view of the structure around the Y3+ ions in YAG, where the Ce–O bonds of the first two shells are indicated. The figure is adapted from [69]. c Figure over the Ce3+ induced lattice expansion in YAG:Ce3+, adapted from [54]. The lattice expansion obeys Vegard’s law [54, 70]

Since in YAG:Ce3+ the activator ions replace Y3+ ions on the 12(c) positions, it is especially important to investigate the local structural properties at and around these sites. In a local view, each Y3+ (or Ce3+) ion is surrounded by two nearest AlIVO4 5− tetrahedra and four nearest AlVIO6 9− octahedral (Fig. 6b). Several previous works have focused specifically on the difference in local structure with and without Ce3+ doping. The introduction of Ce3+ dopants leads to a lattice expansion, see Fig. 6c, (about 3 % for the first and second shell Ce–O distances, cf. Fig. 6b), but significantly less than expected from the 12 % increased size in ionic radius (Ce3+: 101 pm and Y3+: 90 pm). This means that the Ce3+ ions are “squeezed” into the lattice, or in other words that the Ce3+ ion local environment is relatively compressed, which is strongly related to the material’s optical properties. Table 2 shows the interatomic distances between Y3+/Ce3+ and the 1st–6th shell ions denoted by Z, as have been measured by using EXAFS and XRD. Figure 6c shows the Ce3+ ion concentration dependence of the lattice parameter obtained from XRD.

Table 2 Interatomic Ce–Z distances for YAG:Ce3+

Further details into the local structural properties of YAG and YAG:Ce3+ can be obtained from the atomic displacement parameters (ADPs), or so called B factors, which relate to the degree of thermal motions and/or static disorder in the materials. Table 3 shows the B factors for YAG and YAG:Ce3+ as extracted from neutron diffractions [54]. A general observation can be made here that the B factors of all constituent atoms (Y/Ce, Al, and O) increase as the Ce3+ ions are incorporated in the YAG matrix. The increase of the B factors is further seen as a decrease of the Debye temperatures (θ D) of these atoms, which in turn gives a larger accessibility of vibrations, particularly low-frequency vibrational modes [54]. Since the lowest-frequency vibrations in YAG:Ce3+ are specifically related to localized vibrations around the Ce3+ ions, these local vibrational modes are believed to play a key role in luminescent properties, especially at elevated temperatures (This will be further elucidated in Sects. 4.3, 4.4).

Table 3 Crystallographic data of YAG and YAG:Ce3+

4.2 4f–5d Transitions in YAG:Ce3+

Generally, the energy levels of the activator ions in phosphors determine their excitation and emission properties and are directly related to the 4f and 5d energy levels. Among the RE ions, optical transitions can occur between 4f and 5d levels. For parity allowed 4f–5d transitions, the excitation and emission spectra may be substantially affected by the host crystal, because the 5d orbitals are strongly affected by the surrounding lattice, leading to large variation in the emission/excitation wavelength depending on the choice of crystal. For the specific case of YAG:Ce3+, the energy of 4f–5d 1 transitions is around 2.7 eV, i.e., in the visible range, which may be compared to the 4f–5d transitions for a free Ce3+ (6.12 eV), which is in the UV range [39], cf. Fig. 7a. More specifically, the nephelauxetic effect is for YAG:Ce3+ particularly strong since the orbitals of Ce3+ ions and of nearest eight-coordinated O2− ions overlap extensively, leading to significant Ce–O covalency, which results in a large centroid shift ɛ c . Apart from the nephelauxetic effect, ligand fields also have a significant impact on the 5d electronic states of Ce3+. If Ce3+ ions are surrounded by adjacent oxygens with a cubic configuration, the centroid shifted 5d states are split into triplet 2T2g states with higher energy and doublet 2Eg states with lower energy. The splitting energy between 2T2g and 2Eg states for 5d orbitals of Ce3+ is Δ = 8/9(10Dq) [41] for cubal symmetry of CeO8 13−, where Dq can be approximately estimated by using Eq. (1). However, Ce3+ ions are not located in a perfectly cubic environment in the YAG host lattice but located at sites affected by tetragonal deformations. Such locally environments create additional ligand-field components which lead to further energy splitting of the 2T2g states and the 2Eg states, thus all five 5d states of Ce3+ are then non-degenerate. For YAG:Ce3+, the combined influence of the nephelauxetic effect and ligand-field splitting gives rise to a large redshift, D(A), which makes some of the 4f–5d transitions of Ce3+ observable in the range of visible light. Apparently, the electronic states of Ce3+ are significantly altered as the ion is doped in crystals. The effect of local structural distortions on the optical properties of YAG:Ce3+ has been investigated for example by Seijo et al. [71].

Fig. 7
figure 7

a A schematic energy level diagram for Ce3+ showing the effect of the centroid shift and ligand field splitting as Ce3+ ions are embedded in YAG, adapted from [39, 72]. b The energy level scheme of YAG:Ce3+ illustrating the relative energy states of Ce3+ with respect to the conduction and valence bands of YAG. EPI and Eg are photoionization and band-gap energy respectively, adapted from [73]

The optical transitions within energy levels of Ce3+ are the major processes for photon emission; however, the energetic position of these levels with respect to the conduction band (CB) and the valence band (VB) of the host lattice can also play an important role, particularly for the luminescence thermal stability of inorganic phosphors. To construct the energy level scheme of YAG:Ce3+ as depicted in Fig. 7b, Ueda et al. [73] have investigated the activation energy for thermally stimulated ionization processes for 5d electrons of Ce3+ to be promoted to the CB of YAG. The energy level scheme of YAG:Ce3+ reveals that the first two 5d states of Ce3+ are situated between the CB and VB of YAG, and it also implies that the 4f–5d transitions of Ce3+ will effectively proceed when an excitation source (hυ < 3.8 eV) is provided and the thermal energy is lower than the activation energy required for the thermally stimulated ionization. For instance, when YAG:Ce3+ is excited by a 340 nm (≈3.64 eV) NUV LED the 5d 2 state is populated, and then relaxes to the 5d 1 state via coupling to phonons, giving rise to a Stokes shift \( \Delta S \) and a radiative emission process to emit yellow light, as illustrated in Fig. 7. The excitation bands 4f–5d 1 (λ max ≈ 460 nm) and 4f–5d 2 (λ max ≈ 340 nm) are observed in the PLE spectra and a broad emission band (λ max ≈ 565 nm) is present in the PL spectrum, see Fig. 8a. A red-shift of the emission band with increasing Ce3+ concentration can be observed. Although the higher Ce3+ leads to an elongation of Ce–O bond distances, which weakens the ligand fields and should induce a blue-shift, the probability of energy transfer from excited Ce3+ ions to neighboring Ce3+ ions of lower energy. This leads to a counteracting and predominating red-shift [54]. A comparison with the low-temperature spectrum for one selected composition, YAG:0.0033 %Ce3+, reveals splitting of the 4f states into 2 F 5/2 and 2 F 7/2 states due to spin–orbit coupling [4], see Fig. 8b.

Fig. 8
figure 8

a Room temperature PLE and PL spectra of YAG:Ce3+ as a function of Ce3+ concentration. b Low temperature PL spectrum of YAG:0.033 % Ce3+. Both figures are adapted from [74]

4.3 Thermal Quenching in YAG:Ce3+

The quenching temperature (T q ) of YAG:Ce3+, i.e., the temperature at which the emission intensity or emission decay time is reduced to half of the low-temperature value, can be as high as about 600 K if the Ce3+ concentration is sufficiently low, see Fig. 9, which is higher than for most other phosphors (cf. Table 1). The reduced quenching temperature with increasing Ce3+ concentration (Fig. 9b) has been related to thermally activated concentration quenching [74], but has also been explained in terms of θ D [54]. More specifically, it has been found that the increasing concentration of Ce3+ in YAG leads to a systematic decreasing in the θ D of YAG:Ce3+. Such a decreased θ D means that for a fixed temperature more phonons are activated, which in turn increases the probability for non-radiative processes (via 5d–4f crossover according to Fig. 4b) to happen. But, additionally, thermal ionization of 5d electrons to the CB of the YAG garnet crystal may also be at play. For this purpose, Ueda et al. [73, 75] and Dorenbos [39] constructed the VRBE diagrams of various garnet crystals, notably YAG:Ce3+, Y3Al2Ga3O12:Ce3+ (YAGG:Ce3+), and Y3Al5-x Ga x O12:Ce3+ (x = 0–5). A key result is that the substitution with Ga3+ lowers the CB level and also reduces the splitting energy, \( \Delta_{12} \), between the 5d 1 and 5d 2 states (see Fig. 10a). This leads to a concomitant lowering of activation energy for 5d 1–CB transitions, i.e., of \( \Delta E_{1} \) in Fig. 10b, which in turn leads to a lower T q [76]. In addition, it has been shown that the thermally stimulated ionization of 5d electrons occurs not only from the 5d 1 state, but may also occur from defect states related to the presence of intrinsic defects such as oxygen vacancies, bound excitons, or other types of “impurities”, cf. Fig. 10b. Since the presence of such defects is inevitable, the investigations of their nature and specifically their relation to the materials’ optical properties have become increasingly important. Examples of recent works in this context include studies by Stanek et al. [77] and Muñoz-García et al. [78], who have predicted different defect structures in garnet phosphors and the effects of antisite defects on the electronic configurations of Ce3+ for YAG:Ce3+.

Fig. 9
figure 9

a Emission intensity and b luminescence lifetime of YAG:Ce3+ as a function of temperature, adapted from [74]

Fig. 10
figure 10

The VRBE diagrams of a Y3Al5-x Ga x O12:Ce3+ and b YAGG:Ce3+, adapted from Dorenbos [39] and Ueda et al. [73], respectively. In (a), Ec and Ev denote the energy levels for the CB and VB of the host lattice (YAGG). In (b), Eg, EPI, \( \Delta E_{1} \) , and \( \Delta E_{2} \) are the band-gap energy, photoionization energy, and activation energies for 5d1–CB and defect–CB transitions, respectively

Apart from the pronounced reduction in emission intensity at elevated temperatures, the temperature increase of YAG:Ce3+ also leads to an important red-shift of the emission wavelength (Fig. 11a). Generally, the energy difference between the excited and ground states at a given temperature can be estimated by the Varshni equation,

Fig. 11
figure 11

a Temperature dependence of the emission spectra for YAG:Ce3+, adapted from [6]. b Configurational coordinate diagrams of YAG:Ce3+, TAG:Ce3+, and GAG:Ce3+, indicating the correlation of radiative emissions, Stokes shift, thermally excited Stokes shift, and non-radiative relaxations. The figure is adapted from [62]

$$ E (T ) = E_{0} - \frac{{\alpha T^{2} }}{{T{ + }\beta }}, $$

where T is the operating temperature, E 0 is the energy difference between the ground and excited state at 0 K, and \( \alpha {\text{and }}\beta \) are fitting constants [79]. Although blue-shifted emissions are observed for most phosphors and may be expected since while the lattice thermally expands and as an effect ligand fields are reduced, the converse effect observed for YAG:Ce3+ is most likely attributed to its highly rigid structure [54]. Chiang et al. [62] have explained that rigid structures have a larger energy barrier for thermal red-shifted emissions to occur prior to non-radiative relaxations as illustrated in Fig. 11b. Since the rigidity of the host lattice (Y,Gd,Tb)3Al5O12, abbreviated by YAG, GAG, and TAG, respectively, goes as YAG > TAG > GAG, YAG:Ce3+, this would explain the greatest capacity for thermal red-shifted emissions among these Ce3+-doped garnet phosphors. An alternative explanation invoked in the literature for other hosts involves Jahn–Teller couplings [80].

4.4 Lattice Vibrations in YAG:Ce3+

As understood from above, the vibrational dynamics of inorganic phosphors is crucial in determining their optical properties. The processes involving both optical and vibrational transitions are known as vibronic transitions [81]. For the specific case of YAG:Ce3+, the presence of vibronic transitions is evident from the presence of phonon side bands of the low temperature luminescence spectra [72] as shown in Fig. 12. The phonon side bands are specifically located in the range 131–211 cm−1, which, for simplicity, has been averaged to 200 cm−1 [74]. The broadening of the excitation/absorption and emission bands with increasing temperature is attributed to complex electron–phonon coupling which may be pictured via the overlapping of wavefunctions of vibronic transitions in the configurational coordinate diagram, see Fig. 4b. The strength of the electron–phonon coupling can be evaluated by the Huang-Rhys parameter, S, which can be determined from the following relationship: Stokes shift = 2Sħω, where ħω is the energy of the relevant phonon [82, 83]. For YAG:Ce3+, S ≈ 6, which is classified in terms of strong electron–phonon coupling [84]. Strong electron–phonon coupling implies a large offset \( \Delta R \) in the nuclear coordinate between the minima of the excited and ground parabolic potential curves (see Fig. 4b), thus increasing the probability for non-radiative processes. However, other phonons rather than those of 131–211 cm−1 could possibly result in non-radiative relaxation via electron–phonon coupling. For example, some phonons of higher vibrational frequency than 200 cm−1 were also found by Robbins [72] to be involved in vibronic transitions in the NUV absorption spectrum of YAG:Ce3+; however, these are not observed in the luminescence spectrum, probably due to a smearing-out effect caused by multi-phonon couplings in the electronic 4f–5d transitions.

Fig. 12
figure 12

Low temperature (T = 4.2 K) luminescence spectra of YAG:Ce3+ showing phonon side bands fine structure in both the absorption and emission spectra, adapted from [72]

Of particular interest is to understand which particular phonons, and local structures that are of importance in electron–phonon coupling processes. The phonons of YAG can be represented in terms of their irreducible representation:

$$ \varGamma = 5 {\text{A}}_{{ 1 {\text{u}}}} + 3 {\text{A}}_{{ 1 {\text{g}}}} + 5 {\text{A}}_{{ 2 {\text{u}}}} + 5 {\text{A}}_{{ 2 {\text{g}}}} + 10{\text{E}}_{\text{u}} + 8 {\text{E}}_{\text{g}} + 1 4 {\text{T}}_{{ 1 {\text{g}}}} + 1 8 {\text{T}}_{{ 1 {\text{u}}}} + 1 4 {\text{T}}_{{ 2 {\text{g}}}} + 1 6 {\text{T}}_{{ 2 {\text{u}}}} . $$

The 25 modes having symmetries A1g, Eg and T2g are Raman active, while 17 of the modes having T1u symmetry are IR active (the 18th T1u is acoustic). Here, the Eg modes are doubly degenerated and the T2g, T1u are triply degenerated. In Fig. 13a the IR and Raman spectra of YAG are shown. As can be seen, there are 17 distinguishable bands in the IR spectrum and 25 bands in the Raman spectrum, plotted over the frequency range 100–900 cm−1, which is in agreement with the group-theoretically predicted number of modes. Here, the lower-frequency part of the spectra (<200 cm−1) is related to vibrational modes involving the heavy Y ions, whereas the higher-frequency part (200–900 cm−1) relates primarily to vibrational modes of the lighter O and Al atoms. This is seen in Fig. 13b, which shows the one phonon density of states (DOS) of the rare-earth aluminate garnets (RE3Al5O12) [85]. To assign each band to specific vibrational motions, researchers have used different methods [86, 87], and both YAG and its RE-doped variants have been studied in detail. Drawing in complementary local structural information, which generally indicates only small differences in the structure between these different materials, notably related to slight lattice expansion/distortion upon the Ce3+ doping with no change of the overall (average) structural symmetry  [54]. It follows that the vibrational spectra of YAG:Ce3+ can be expected to be similar to that of YAG, meaning that the vibrational properties of the undoped compound may be used as a very good starting point for the understanding of the vibrational dynamics in these technologically important materials. Further research along these lines may be very rewarding for elucidating the mechanistic aspects of luminescence in YAG type phosphors, which would be central for understanding inorganic phosphors generally.

Fig. 13
figure 13

a IR and Raman spectra of YAG as measured at 300 K, adapted from [88] (b) One phonon DOS of RE3Al5O12 (Y3Al5O12), adapted from [85]

4.5 Other Yellow Phosphors

Although the performance of YAG:Ce3+ as a yellow phosphor coupled with a blue LED chip is almost ideal for LED lamps with a high CCT, and this has slightly hampered the research of new improved materials, a variety of other classes of inorganic, yellow-emitting phosphors have been developed and are now attracting increased attention. Examples are materials based on oxynitrides, such as Ca-α-SiAlON:Eu2+ and Li-α-SiAlON:Eu2+ [37, 89]. The CIE coordinates of Ca-α-SiAlON:Eu2+ (0.491, 0.497) and Li-α-SiAlON:Eu2+ (0.48, 0.51) and YAG:Ce3+ (0.41, 0.57) reveal that these two oxynitride phosphors are more suitable to produce warm white light in pcWLEDs than YAG:Ce3+. Furthermore, pcWLEDs using Ca-α-SiAlON:Eu2+ have better chromaticity stability, i.e., small variations of the CIE coordinates, in the temperature range of 25–200 °C than based on YAG:Ce3+ [37]. However, the absorption and external quantum efficiency of Li-α-SiAlON:Eu2+ are about 70 and 40 %, respectively, which are lower than for YAG:Ce3+, 90 and 75 %, respectively [90]. Other examples include phosphors based on silicates, such as Sr3SiO5:Eu2+ [91], which provide a warmer white light compared to YAG:Ce3+, when combined with an identical blue source. As comparing to YAG:x %Ce3+ generally for different Ce3+ concentrations (e.g., x = 0.033–3.33 [74]), Sr3SiO5:Eu2+ also shows better thermal stability of the emission intensity as the temperature increases from 25 to 250 °C [91]; however, the quantum efficiency (68 %) and CRI (64) are still worse than for YAG:Ce3+. Another silicate phosphor is Li2SrSiO4:Eu2+ [92]. PLE measurements of Li2SrSiO4:Eu2+ show that the excitation spectrum is relatively broad, covering both the NUV (380–410 nm) and blue (450–480 nm) ranges, whereas the emission band is centered at 562 nm and a strong orange-yellow light is typically perceived. When combined with a 455 nm blue LED, it generates white light located at the CIE coordinates (0.3346, 0.3401), corresponding to a CCT of 5406 K. Li2SrSiO4 doped with 0.005 mol% Eu2+ shows the strongest emission intensity compared to the other Eu2+ doping concentrations, whereas x mol% (x = 0.0025–0.05) and the LE is comparable to that of YAG:Ce3+.

Further strategies to develop new, better performing phosphors include the tuning of the optical properties of YAG:Ce3+ by co-doping with a 2nd or even a 3rd dopant atom, on either the Y3+ and/or the Al3+ site(s). For example, the influences on optical properties resulting from cation substitutions have been discussed previously for (Y,Tb,Gd,Lu)3(Al,Ga)5O12:Ce3+ [57, 6165] and Y3Al5–2x (Mg,Si) x O12:Ce3+ [36]. According to Katelnikovas et al. [36], the pcWLEDs comprising Y3Mg2AlSi2O12:Ce3+ (YMASG:Ce3+) and a 483 nm blue LED emit warm white light characterized by the coordinates (0.4338, 0.4030) in the CIE 1931 chromaticity diagram. The emitted white light corresponds to a CCT of 3000 K and a CRI of 75. The peak position of the emission band is at around 600 nm as excited by a 440 nm blue LED. A strong red-shift (40–50 nm) takes place when Mg2+ and Si4+ substitute Al3+ in YAG:Ce3+, and the observed red shift is attributed to increased covalency and Stokes shift. The LE of the pcWLEDs containing YMASG:Ce3+ is about 81 % of that of the pcWLEDs coated with YAG:Ce3+. The annealing temperature (1400–1550 °C) and Ce3+ concentrations (0.5–2 %) for YMASG:Ce3+ synthesized via sol–gel combustion methods have shown about 10 and 20 % impacts on the luminescence decay time, respectively.

5 Blue Phosphors

Blue emitting phosphors can be useful for SSL if they can be excited by a NUV LED, i.e., they convert radiation in the range 380–410 nm to light centered in the blue region (420–500 nm, including part of the violet range). It has been pointed out that it is presumably impossible to develop an efficient blue emitting phosphor that can be excited at λ ≥ 410 nm due to Stokes shift requirements [12]. This constraint limits possible applications to materials showing suitable (and possibly allowed) absorption and emission bands in those two spectral regions. In practice, the only two activator ions that have attracted interest for the development of blue phosphors for WLEDs (White Light Emitting Diodes) are Eu2+ and Ce3+. This section will be mainly devoted to these ions.

Both activators give rise to luminescence bands due to 5d–4f transitions (4f 65d 1 → 4f 7 and 5d 1 → 4f 1, respectively) with radiative decay times of the order of 1 µs for the former [93], and some tens of nanoseconds for the latter [94]. The 5d–4f transitions are allowed through the electric dipole mechanism and, therefore, are generally very strong. Their quantum efficiency at room temperature (RT) is in principle high, apart from the cases in which photoionization occurs, i.e., escape of an electron to the conduction band [95, 96].

Among the inorganic hosts developed for blue LED phosphors, prominent are oxide-based materials, such as phosphates, silicates, aluminates; on the other hand, additional host compositions have been proposed and developed successfully, such as nitrides, oxynitrides, and others. These two host categories will be reviewed separately.

5.1 Blue Phosphors—Oxide Hosts: Phosphates, Silicates, Aluminates

An important class of blue phosphors for pcLEDs is the phosphate family, which couples a relatively easy preparation to a great structural flexibility. Moreover, it has been shown that Eu2+ gives rise in many cases to blue emission in these hosts [97]. Possible explanations are a low centroid shift of the 4f 65d 1 configuration, a weak ligand field and/or a small Stokes shift. The same consideration applies to the case of Ce3+ [98101].

Please note that in the rest of this chapter the doping level will always be given in mol%, which will be abbreviated as % for simplicity.

The apatite family of phosphate crystalline materials has attracted attention as a host for Ce3+ and Eu2+. Shang et al. [101] have reported on the synthesis, luminescence, and energy transfer properties of the apatite Ca8La2(PO4)6O2 activated with Ce3+ and/or Eu2+. Samples were made using a Pechini-type sol–gel method. Both dopants show blue photoluminescence emission upon UV excitation, and efficient energy transfer from Ce3+ to Eu2+ has been demonstrated. The best results are obtained for a sample containing 4 % of Ce3+ and 2 % of Eu2+, showing emission of high color purity and a PLE spectrum spanning the whole UV region (Fig. 14). A similar host, doped with Eu2+, but containing chlorine, Ca2PO4Cl with the spodiosite crystal structure, was investigated by Chiu et al. [25]. Phosphors containing various amounts of dopants were prepared by solid state reaction; the sample containing 11 % of Eu2+ showed intense blue emission peaking at 454 nm upon excitation at 400 nm, with internal and external quantum efficiencies of 85 and 61 %, respectively. These phosphors are considered good candidates for pcLEDs.

Fig. 14
figure 14

The CIE chromaticity diagram for the Ca8La2(PO4)6O2:0.04Ce3+ (point 1), and Ca8La2(PO4)6O2:0.05Eu2+ (point 2) samples, adapted from [101]

Double orthophosphates containing an alkali and an alkaline earth cation have also been widely studied. KBaPO4:Eu2+ has been considered by Im et al. [102] who have reported on its synthesis, crystal structure, and luminescence spectroscopy. The material emits at 420 nm, which is almost in the violet range, with excitation extending over the whole UV region. When the surface is coated with a SiO2 layer, the phosphor shows an excellent moisture resistance. The phosphor γ-KCaPO4:Eu2+ has been studied by Yim et al. [24] who have reported on the synthesis and optical spectroscopy. The material can be excited in a wide spectral range in the UV and visible (200–450 nm) (Fig. 15), and the emission peaks at 473 nm, resulting in a whitish blue luminescence with CIE coordinates (0.199, 0.319) for the sample doped with 3.2 % Eu2+. Another example of double orthophosphate is LiCaPO4:Eu2+. Zhang et al. [103] have reported on the structure-properties relationships of this phosphor, and shown that when the dopant concentration is optimized to 3 %, the peak luminescence is at 470 nm with excitation at 395 nm, the CIE coordinates are (0.119, 0.155, greenish-blue) and the quantum efficiency is 52 %.

Fig. 15
figure 15

Excitation (a) and emission spectra (b) of KCa1-x PO4:xEu2+ (x = 0.032) phosphor. The inset shows the emission spectra of the different excitation peaks, adapted from [24]

Diphosphates are another class of phosphate hosts that has been explored for blue phosphors. β-Ca2P2O7, in which 5 % Eu2+ substitutes for Ca2+, has been prepared by combustion synthesis by Ta and Chen [104]. Excitation at 387 or shorter wavelengths originates emission peaking at 421 nm (more violet than blue) but extending to about 475 nm in the blue. A similar host is the boro-diphosphate KBaBP2O8 having a tetragonal structure. Sun et al. have reported on the material doped with Eu2+ [105], showing that upon excitation in a broad spectral region extending from 250 to 420 nm, emission peaking at 445 nm is observed. This emission is very broad and extends up to 600 nm. The ideal Eu2+ concentration appears to be 3 %, since higher concentration samples suffer from concentration quenching.

Another important class of blue phosphors is the silicate family, characterized by excellent chemical stability and again large composition variability, although the synthetic procedures might require higher temperatures than phosphates. Liu et al. [106] have reported on the phosphor Li2Sr1−3x/2Ce x SiO4 (x = 0.0025–0.07). The material was prepared by high temperature solid state synthesis as a single phase and it was found that Ce3+ is located in the Sr2+ site. The phosphor produces broad blue emission peaking at 442 nm with a broad excitation profile covering the spectral region from 200 to 400 nm. The thermal quenching is weak and the internal and external quantum efficiencies are 81 and 66 %, respectively, upon 358 nm excitation. The phosphor is promising for SSL applications.

The doped disilicate Sr2MgSi2O7:Eu2+ (well known in the field of persistent luminescence [107]) has been investigated as blue phosphor for pcLEDs by Zhang et al. [27]. The material was prepared by solid state reaction as a single phase and showed broad and asymmetric emission centered at 470 nm, upon excitation at the peak located just below 400 nm. The optimum Eu2+ concentration was found to be 6 % and the CIE coordinates (measured in a blue LED fabricated ad hoc) were (0.142, 0.228). (Ca1−x ,Eu x )MgSi2y O6+δ (x ranging from 0.001 to 0.06, y from 1.0 to 2.0) has been considered as a blue phosphor by Jung et al. [108]. The samples were produced by spray pyrolysis from suitable precursors. The highest emission intensity around 450 nm was found for y above 1.0, i.e., in the presence of a silicon excess. In particular, for x = 0.04 and y = 1.3, the emission intensity showed an increase of about 150 % compared to the stoichiometric sample with x = 0.01 and y = 1.0.

The novel fluoro-silicate phosphor Ca5.45Li3.55(SiO4)3O0.45F1.55 doped with Ce3+ has been developed by Zhou et al. [109]. The material was prepared by high temperature solid state reaction yielding a single phase in which the Ce3+ ions substitute for Ca2+. The luminescence spectra of the sample containing 3 % Ce3+ show a broad asymmetrical emission band peaking around 470 nm (Fig. 16), which is excited in a broad wavelength range with maxima at about 256 and 371 nm. The RT decay times of the emission are close to 60 ns, a value typical for Ce3+ emitting in the visible region. For the optimal dopant concentration of 3 %, the color coordinates are (0.154, 0.188), and the internal quantum efficiency is 82 %. The phosphor shows high color purity and good thermal stability.

Fig. 16
figure 16

Excitation and emission spectra of Ca5.45Li3.55(SiO4)3O0.45F1.55. The inset shows the comparison of the emission spectra of Ce3+-doped and the undoped host, adapted from [109]

An interesting approach was proposed by Park et al. [110] in order to identify silicate phosphors for WLEDs. They employed combinatorial chemistry, developing quaternary and ternary combinatorial libraries to synthesize, process, and screen silicate materials. This combinatorial chemistry consisted of solution-based combinatorial synthesis and characterization, and swift scanning of luminance. The authors found several candidates not only for blue, but also green and red, efficient emission upon NUV excitation.

Aluminates are another important phosphor family. BaMgAl10O17:Eu2+ (BAM:Eu2+) is a commercial phosphor emitting blue light in fluorescent lamps. Wang et al. [111] have studied its spectroscopic behavior as a phosphor for pcLEDs. The samples were prepared by solid state reaction in the presence of different amounts of H3BO3. Excitation spectra cover the range 280–400 nm and the emission is broad, peaking at about 440 nm and extending to 520 nm. For the sample prepared with a 6 % excess of boric acid the CIE coordinates are (0.151, 0.058). BAM:Eu2+ appears to be suitable for the use in SSL. Cui et al. have studied (Mg,Sr)Al2O4:Eu2+ as a blue phosphor for WLEDs [112]. Various materials were prepared with different amounts of Mg and Sr and of the dopant. The materials were characterized by XRD and electron microscopy. MgAl2O4:Eu2+ shows blue luminescence in the range 400–600 nm peaking around 460 nm, with excitation extending in the UV up to 400 nm. The addition of Sr2+ increases the phosphor emission intensity up to 6 %. The material Sr3Al2O6 doped with Ce3+ and alkali ions has been synthesized by solid state reaction and its luminescence has been investigated by Li et al. [113]. For all samples the excitation spectrum shows a strong peak at 395 nm, and excitation at this wavelength produces a very broad band peaking at about 460 nm and extending up to 600 nm. The addition of Li+, Na+, K+ or Rb+, slightly changes the profile of the excitation bands, but does not affect the emission profile. However, the spectral intensity appears to increase with the alkali ion doping, and the highest increase is obtained with the addition of 4 % Rb+. The decay times of the Ce3+ luminescence in the presence of alkali ions is rather short around 20 ns.

5.2 Blue Phosphors—Other Hosts

In this subsection other hosts will be discussed, not belonging to the category of common oxide hosts such as phosphates, silicates, and aluminates. The main focus will be on nitrides and oxynitrides, but also other hosts will be considered. Also, another activator ion (Ce4+) will be briefly discussed.

Nitride and oxynitride phosphors have become important in the field of SSL due to their high efficiency, stability, and strong absorption in the NUV and blue spectral regions [8]. As an example, the phosphor BaSi7N10 doped with Eu2+ has been reported by Li et al. [114]. The material can be produced by sol–gel, followed by direct gas reduction and nitridation. The luminescence spectra are composed of a strong green–blue emission band centered at about 482–500 nm (Fig. 17), with corresponding excitation profile in the region 250–400 nm with a maximum around 300 nm. The external QE reaches 52 % upon excitation at 300 nm, for the sample doped with 20 % Eu2+. The coordinates of the phosphors containing various concentrations of Eu2+ fall in the cyan region of the CIE diagram. The strong emission, high QE and low thermal quenching show that these phosphors are suitable for pcLEDs applications. Tang et al. have developed a blue emitting oxynitride phosphor of formula BaSi3Al3O4N5:Eu2+ [115], synthesized at high temperature in flowing N2 atmosphere. The phosphor shows a blue emission peak around 470 nm and a broad excitation spectrum in the NUV. The optimum concentration of the dopant is 15 % replacing Ba2+. At this doping level, the decay time of the emission is just above 800 ns. The internal and external quantum efficiencies of the luminescence are 85 and 79 %, with excitation at 305 nm. The phosphor appears to be superior to commercial BAM:Eu2+ because it has a much higher converted efficiency than the latter material under the emission spectrum of a NUV LED chip. The oxynitride material β-SiAlON activated with Ce3+ was investigated by Ryu et al. [116]. The hexagonal compound was prepared with various amounts of the dopant and suitably characterized using XRD and other techniques. The phosphors emit blue light with a maximum at about 470–490 nm and an excitation peak at 410 nm. The optimum dopant concentration appears to be 2 % (replacing Si). It was found that the thermal quenching is significantly smaller than for YAG:Ce3+, confirming the good properties of this material as a blue phosphor.

Fig. 17
figure 17

Temperature dependence of emission intensity of BaSi7N10:Eu2+ (20 mol%) phosphors excited by 300 nm radiation, adapted from [114]

Mikami and Kijima have studied in depth how covalency affects the position of the 5d levels of rare earth ions in nitride and oxynitride phosphors [117]. Using theoretical arguments, they conclude that the idea of covalency does not seem to be fully reliable in tailoring the position of 5d levels of Ce3+ and Eu2+ in nitrides and oxynitrides. They propose that quantitative dielectric properties reflecting spectroscopic polarizability should be used in addition to covalency. These ideas have been applied to a series of nitride and oxynitride phosphors such as LaSi3N5:Ce3+ (blue), La3Si6N11:Ce3+ (yellow), BaSi2O2N2:Eu2+ (bluish green), Ba3Si6O9N4:Eu2+ (bluish green), and Ba3Si6O12N2:Eu2+ (green).

The oxychalcogenide phosphor CaLaGa3S6O:Ce3+ was developed by Yu et al. in order to obtain a suitable material for the absorption of light emitted by InGaN chips [118]. The phosphor was prepared by using a solid state reaction starting from appropriate sulfide and oxide precursors. Luminescence spectra are composed of a double peaked emission feature with maxima at 442 and 478 nm, and excitation is dominated by a peak at 398 nm. Maximum intensity is obtained for the sample containing 10 % Ce3+; for this sample the decay time of the Ce3+ emission at RT is close to 16 ns. The color coordinates of the light emitted by a suitably fabricated pcLED are (0.147, 0.089) and fall in the blue region. We note that the experimental decay time of the blue emission is quite short, and this could indicate that the quantum efficiency is not very high.

The chloroborate compounds M2B5O9Cl (M = Sr, Ca) were proposed as hosts for Eu2+ by Guo et al. [119]. These materials already show possible applications as storage phosphors for X-Ray imaging and were then considered as blue phosphors for SSL. For both phosphors, the optimal concentration of Eu2+ was found to be 8 %. In the case of M = Sr, the luminescence excitation spectrum features a broad band between 250 and 410 nm, and the emission one shows a strong peak centered at 425 nm. For M = Ca, the excitation profile is more or less similar, but the emission peak red-shifts to 452 nm. As a consequence, the CIE coordinates are different, but in both cases in the blue region [(0.162, 0.015) for M = Sr and (0.143, 0.048) for M = Ca)].

A different approach for the development of a blue phosphor has been used by Luo et al. [120], based on Eu2+ doped glass ceramics containing BaF2 nanocrystals. The authors prepared an oxyfluoride glass containing SiO2 under reducing conditions, and then heated samples at various temperatures to obtain glass ceramics. These materials showed luminescence spectra characterized by blue emission centered at 450 nm that could be excited between 370 and 390 nm. It can be inferred that Eu2+ is formed from the Eu3+ precursor and that it is preferentially located in the BaF2 nanocrystals.

In order to improve the performance of the well-known Sr2CeO4 phosphor upon excitation in the NUV, Hsu et al. have proposed the mixed oxide material Sr2Ce1-x Sn x O4 (0 > x ≥ 0.07) as a blue phosphor [121]. In these materials the optical transitions are based on a ligand-to-metal charge transfer (CT) from O2− to Ce4+ [122, 123]. It was found that the addition of Sn4+ modifies the excitation profile, so that the low energy feature around 345 nm becomes dominant for x = 0.07. Upon excitation in this feature, the intensity of the broad emission band centered around 480 nm is highest when x = 0.07. The presence of the tin dopant makes the phosphor more suitable for NUV excitation.

6 Green Phosphors

Green emitting phosphors can be employed for SSL either upon excitation of a NUV or a blue LED so that they convert radiation in the ranges 380–410 or 450–480 nm to light centered in the green region (520–565 nm). In this case, the arsenal of activator ions that can be used for the development of green phosphors for WLEDs is larger than for the blue ones, as Eu2+ and Ce3+ are supplemented by Tb3+ and Mn2+. The core of this section will be devoted to these ions.

As presented above, the two former ions give rise to luminescence bands due to 5d–4f transitions. On the other hand, the latter two ions emit through 4f–4f and 3d–3d transitions, respectively. These intraconfigurational transitions are parity-forbidden through the electric dipole mechanism and are, therefore, weak. This can create problems in the phosphor performance, as the available excitation transitions are in general also forbidden and weak. For this reason, in many cases excitation of Tb3+ and Mn2+ in these materials occurs through an energy transfer scheme in which Ce3+ or Eu2+ are used as sensitizers for the activator ions (or the other way around).

As in the section above, common oxide-based hosts, such as phosphates, silicates, and aluminates, and other host compositions will be reviewed separately.

6.1 Green Phosphors—Oxide Hosts: Phosphates, Silicates

Also, in the case of green phosphors for pcLED, the phosphate family has proved to be important, as the considerations mentioned above about preparation and structural flexibility still apply. The eulytite material Sr3Gd(PO4)3 co-doped with Ce3+ and Eu2+ has been investigated by Sun et al. [124]. Single phase samples containing various amounts of the dopants were prepared by solid state reaction in a reducing atmosphere. Energy transfer was found to occur from Ce3+ to Eu2+, with an efficiency of about 95 % for the highest concentration of Eu2+ (2 %). The emission profile of Eu2+ peaks at 518 nm in the green spectral region. The authors point out that in the co-doped samples the excitation of the Eu2+ luminescence (spanning the whole UV region) is greatly enhanced due to the energy transfer process mentioned above, without increasing the Eu2+ concentration and therefore avoiding the onset of concentration quenching.

The double phosphate NaCaPO4 doped with Tb3+ has been proposed as a green phosphor for NUV LEDs by Ratnam et al. [20]. Samples containing 1–7 % Tb3+ were obtained by solid state reaction and characterized by XRD and electron microscopy. The optimal concentration for Tb3+ emission was found to be 5 %. At this doping level, blue and violet luminescence from 5 D 3 is quenched by cross relaxation processes [125] and the emission spectrum is dominated by the 5 D 4 → 7 F 5 band in the green at 547 nm. The excitation spectrum of this feature is composed of stronger 4f–4f bands peaking around 370 nm, accompanied by weaker 4f–5d bands (presumably spin-forbidden) at lower wavelength. Upon excitation in the NUV, the CIE coordinates for this Tb3+ concentration are about (0.27, 0.58), which falls in the green region. In principle this phosphor could be useful for pcLEDs, but the excitation bands are probably too weak for practical purposes. Excitation of the green Tb3+ emission can be enhanced through energy transfer from a suitable donor ion, such as Eu2+. These phenomena have been studied in the double phosphates Ca10K(PO4)7 [21] and LiSrPO4 [22] co-doped with the sensitizer Eu2+ and the activator Tb3+. Both phosphors were prepared by solid state reaction yielding single phase materials. The Eu2+–Tb3+ transfer allows exciting the 5 D 4 Tb3+ emission around 540 nm efficiently in the UV, in the region where NUV LEDs operate. The energy transfer is not complete (the efficiency is about 66 % for LiSrPO4), so that the broad luminescence of Eu2+ is still present in the spectra in the blue region. pcLEDs devices were built by coupling a 398 nm NUV LED with the phosphors in which the concentration of the dopants had been optimized. The color coordinates were found to be (0.283, 0.383) for Ca10K(PO4)7 and (0.294, 0.428) for LiSrPO4; both points appear to be located in the green part of the CIE diagram, indicating that these materials are suitable as green phosphors.

The borophosphate material Sr6BP5O20:Eu2+ was prepared by Zhang et al. [126] and used as a bluish-green phosphor in the fabrication of a WLED. The sample containing 9 % of Eu2+ was characterized by XRD. The emission spectrum of Eu2+ has a peak at 475 nm and extends up to 650 nm, whilst the excitation spectrum shows maxima at 290 and 365 nm, with a shoulder at 390 nm. The decay time of the emission is in the µs range, typical for this dopant. The color coordinates of the phosphor are (0.174, 0.312) located in the bluish-green region. A mixed silicophosphate material Ca5(PO4)2SiO 4 doped with Eu2+ and having the silico-carnotite structure has been proposed by Roh et al. [127] as a green emitting phosphor. Luminescence from this material can be excited from 220 to above 400 nm, and is composed of a broad band peaking at 530 nm. This feature is asymmetric, reflecting the multi-site nature of the host. The optimum concentration for Eu2+ was found to be 5 %; with this composition, a pcLED was fabricated by coupling with a NUV LED emitting at 380 nm. The yellow-green emission spectrum of the device was stable as a function of the DC current.

As it is well known, Eu2+ in many cases gives rise to green emission in silicate hosts. Also Ce3+ in silicate garnet hosts has been found to produce green emission (see below). For this reason, silicate compounds are important hosts for green emitting phosphors for pcLEDs. Shimomura et al. [32] reported in 2007 on the luminescence and crystal structure of Ca3Sc2Si3O12:Ce3+ prepared by solid state reaction. The material has a cubic garnet structure and XAFS data evidenced that the Ce3+ dopant enters the structure in the Ca2+ dodecahedral site. The emission spectrum shows an asymmetric band peaking at 505 nm, with a corresponding excitation profile peaking at 455 and extending to higher wavelengths; this is very suitable for excitation using a blue LED. The color coordinates obtained upon excitation at 455 nm are (0.30, 0.59), which is in the green range. The temperature quenching is not significantly efficient. This material appears to be a useful phosphor. An alternative synthesis of this phosphor (based on the sol–gel technique) has been described by Liu et al. [128].

Ba2MgSi2O7:Eu2+ has been investigated by Zhang et al. [129] as a green phosphor for pcLEDs. The material was prepared by solid state reaction and suitably characterized. The excitation spectrum is very broad and covers all the region ranging from 250 to 450 nm. Excitation at 395 gives rise to a broad emission feature peaking at 505 nm and extending to almost 650 nm. In the case of a 7 % doping level for Eu2+, the QE was found to be 92 %, and the temperature quenching appears to be very weak. This green phosphor is efficient and suitable for excitation with NUV chips. The double orthosilicate SrBaSiO4:Eu2+ was also considered by Zhang et al. [33]. In this case, the excitation is less broad, but spans the NUV region, peaking around 430 nm. Excitation in this range originates a relatively sharp emission in the green (full width at half maximum of 80 nm) with a peak around 512 nm. The emission intensity is highest for a Eu2+ concentration of 2 %, and for this sample the internal QE was found to be 78 %. The quenching temperature T q , at which the emission intensity is reduced to 50 %, is 170 °C. Xia et al. [130] have reported on the phosphor Na3Y1–x Sc x Si3O9:Eu2+. This material can be prepared by high temperature solid state reaction in a reducing atmosphere. Samples with x ranging from 0.1 to 0.9 were synthesized and fully characterized. The best composition in terms of spectroscopic properties was found to be the one with x = 0.7, and 3 % Eu2+ (substituting for Na+). In this case, the excitation is very broad in the UV with several features up to 430 nm. The corresponding emission spectrum upon excitation at 365 nm is characterized by broad emission at about 500 nm. The decay times of the emission amount to several hundreds of ns, depending on the Eu2+ concentration. Thermal quenching reduces the RT intensity to 63 % at 100 °C. The CIE coordinates upon excitation at 365 nm (0.164, 0.549) show that the phosphor emits green light with high color purity.

The luminescence of the aluminosilicate MAl2Si2O8 (M = Sr, Ba) codoped with Eu2+ and Mn2+ has been investigated by Ye et al. [131]. Materials containing various amounts of the dopant ions were prepared by solid state reaction under reducing atmosphere. It was found that the optimal concentrations for luminescence are 1 % Eu2+ and 10 % Mn2+. Excitation in the UV yields Mn2+ emission in the green-yellow (peaks at 564 and 518 nm for M = Sr and Ba, respectively). However, the Eu2+–Mn2+ energy transfer is weakly efficient (10 % for M = Sr and 5 % for M = Ba) due to the limited spectral overlap between the Eu2+ emission and the Mn2+ absorption.

6.2 Green Phosphors—Other Hosts

The double oxide CaSc2O4 doped with Ce3+ has been studied by Shimomura et al. [132]. The material doped with 1 % of the dopant was prepared by solid state reaction above 1400 °C in a reducing atmosphere. The luminescence peak is located at 515 nm with a shoulder to longer wavelength, as typical of Ce3+ (Fig. 18). The excitation spectrum has a maximum at 450 nm, which is suitable for excitation with a blue InGaN LED. The quantum efficiency is high (90 % of the value observed for Ca3Sc2Si3O12:Ce3+ [32]) and the thermal stability is good. The effects of the introduction of other cations, such Mg2+, Sr2+, Lu3+, were examined; they are reflected in shifts of the emission peak and changes in the peak intensity.

Fig. 18
figure 18

Photoluminescence and photoluminescence excitation spectra of CaSc2O4:Ce3+ (1 mol%), adapted from [32]

The borate material LiBaBO3:Tb3+ has been investigated by Li et al. [133]. Phosphors containing various amounts of Tb3+ were obtained by solid state reaction at 700 °C. Only emission from the 5 D 4 level is observed, presumably due to fast multiphonon relaxation from the upper level 5 D 3 induced by high frequency vibrations of the borate host. The optimal concentration is 3 % of Tb3+. Emission peaks at 544 nm, whilst the excitation profile comprises two structured bands peaking at 242–277 and 368–381 nm. This latter band is due to 4f–4f transitions and, therefore, is relatively weak. The CIE coordinates of the emission are in the green region at (0.249, 0.578). The addition of charge compensating alkali ions appears to increase the emission intensity from 5 D 4. SrMoO4 doped with various amounts of Tb3+ and Na+ (for charge compensation) has been proposed as a yellowish green phosphor by Li et al. [134]. Emission is strong from 5 D 4 at 548 nm. However, excitation maxima are located at 488 or 375 nm, corresponding to 4f–4f transitions from the 7 F 6 ground state to 5 D 4 and 5 D 3, respectively. Despite the efficient emission, the excitation probability is weak and it is doubtful that these phosphors can find practical applications for pcLEDs.

The oxyfluoride Ca2Al3O6F doped with Eu2+ (0.5–20 %) has been investigated by Xia et al. [19]. The material can be synthesized by high temperature solid state reaction under reducing atmosphere; it was thoroughly characterized from a structural and morphological point of view. The luminescence spectra demonstrate a very broad excitation band spanning the range 250–450 nm, and emission peaking at 502 nm with a FWHM of 70 nm. The emission intensity reaches a maximum for the 5 % sample, whilst the decay times decrease from 0.68 to 0.25 µs in the 0.5–20 % concentration range. The emission intensity decreases to 39 % of the RT value at 150 °C. The measured internal QE of this green phosphor is 34 % upon excitation at 400 nm, and the CIE coordinates are (0.171, 0.470) (Fig. 19).

Fig. 19
figure 19

CIE coordinates and digital photo under 365 nm UV lamp of the Ca1.95Al3O6F:0.05Eu2+ phosphor, adapted from [19]

The authors conclude that the efficiency can be improved by optimizing the particle size, size distribution, morphology, and crystalline defects. Xia and Liu have also reported on the same host Ca2Al3O6F co-doped with Ce3+ and Tb3+ [135]. Luminescence spectra and decay curves of these materials (containing 8 % of Ce3+ and variable amounts of Tb3+) clearly show the presence of very efficient energy transfer from the former ion to the latter. Green Tb3+ emission from 5 D 4 (542 nm) is obtained upon excitation of the allowed 4f–5d transitions of the Ce3+ sensitizer in the NUV region (320–400 nm), with maximum intensity for 5 % Tb3+. Emission peaking at 413 nm from the Ce3+ ion becomes rapidly weaker when the Tb3+ concentration is increased. The energy transfer efficiency is very high, being almost 92 % for the sample containing 5 % Tb3+ (Fig. 20). The CIE coordinates for this composition are (0.287, 0.455), indicating that this material is a potential green emitting candidate phosphor for NUV LEDs. Moreover, the emitted color can be tuned from blue to blue-greenish, green, and greenish-yellow by changing the concentration of the two dopants.

Fig. 20
figure 20

Dependence of the Ce3+ emission, Tb3+ emission, and energy transfer efficiency of Ce3+−Tb3+ on Tb3+ doping concentration for Ca2Al3O6F:0.08Ce3+,yTb3+ phosphors, adapted from [135]

Setlur et al. [136] have proposed oxyfluoride and fluoride phosphors for energy-efficient, high-color rendering LED lamps. Among them, it is worth mentioning the oxyfluoride (Sr,Ca)3(Al,Si)O4(F,O):Ce3+ yellow-green emitting materials. These phosphors are made by solid state reaction at high temperature in a reducing atmosphere using SrF2 as a flux. The luminescence spectra are dominated by a strong emission around 540–550 nm; excitation shows two broad peaks in the range 250–500 nm. The QE at RT (excitation at 450 nm) appears to be comparable to commercial Ce3+ doped garnet phosphors (>80 %). The thermal quenching of the compositions (Sr0.98Na0.01Ce0.01)3AlO4F and (Sr0.595Ca0.4Ce0.005)3Al0.6Si0.4O4.415F0.585 is weak (<15 % loss at 150 °C) (Fig. 21).

Fig. 21
figure 21

Integrated intensity vs. temperature (λ ex = 405 nm) for (Sr0.98Na0.01Ce0.01)3AlO4F, (Sr0.595Ca0.4Ce0.005)3Al0.6Si0.4O4.415F0.585 and (Sr0.98Na0.01Ce0.01)3Al0.25Si0.75O4.75F0.25. The lines are only guides to the eye, adapted from [136]

Oxynitrides and nitrides form an important family of green phosphors for SSL, with interesting optical properties [8]. In particular, SiAlONs are hosts that have been widely investigated. As an example, β-SiAlON doped with Eu2+ (with a composition Eu x Si6-z Al z O z N8-z , with x = 0.018 and z = 0.23) has been considered by Ryu et al. [137]. Samples were obtained by gas-pressured solid state reaction. This material emits in the green region at 539 nm, and shows excitation peaks at 300, 358, and 407 nm, superimposed to a very broad band. Thermal quenching is weak, as at 150 °C the emission intensity is reduced to 85 % of that measured at RT. The results show that β-SiAlON:Eu2+ is a promising green phosphor for pcLED applications. Another example is given by the green emitting material α-SiAlON doped with Yb2+ [138]. In this case, the doping ion shows luminescence spectra due to 4f 14↔ 4f 135d 1 interconfigurational transitions [139]. The oxynitride material is characterized by a single broad band centered at 549 nm (Fig. 22). The low energy spectral position of this transition is attributed to a large ligand field splitting of the 4f 135d 1 configuration and to a large nephelauxetic effect induced by the nitrogen coordination of Yb2+. Excitation spectra show several broad peaks in the UV and blue region, in particular one located at 445 nm, which makes this phosphor suitable for blue LED excitation. Concentration quenching is efficient, so that the optimal Yb2+ concentration seems to be relatively low. The CIE coordinates of α-SiAlON:Yb2+ are (0.323, 0.601). This phosphor is potentially useful to create white light when coupled with a red phosphor and a blue LED chip.

Fig. 22
figure 22

a Excitation and b emission spectra of the Yb2+-doped Ca-α-SiAlON. The excitation spectrum was monitored at λ em = 549 nm and the emission spectrum was recorded at λ ex = 342, 300, and 445 nm. Both figures are adapted from [138]

A green emitting phosphor of composition SrSi2O2N2 codoped with Eu2+ and Mn2+ has been developed by Song et al. [140]. Materials containing 2 % Eu2+ and various amounts of Mn2+ have been synthesized via a solid state reaction. XRD showed that the main product is the desired SrSi2O2N2 phase, but some traces of impurities are present. The materials only containing Eu2+ show a broad emission band centered at 530 nm in the green, with excitation spanning the UV–visible region from 250 to almost 500 nm. It was found that addition of Mn2+ leads to energy transfer from this ion to Eu2+, increasing the excitation pathways in the NUV, the optimum content of Mn2+ appears to be 3 %. Tang et al. have investigated the phosphor Ba3Si6O12N2:Eu2+ and reported its synthesis, characterization, and theoretical simulation [141]. The excitation spectra of the doped materials are characterized by a very broad band in the 200–500 nm range, and the emission is located in the green at 527 nm. The maximum intensity is obtained for relatively high Eu2+ concentration, and therefore, it appears that concentration quenching is not effective. The excitation matches well with blue and NUV LEDs. The CIE coordinates are compatible with high green color purity. Theoretical simulations attempt an explanation of the high absorption intensity of this phosphor in the NUV and blue region. The novel oxynitride material (Ba,Sr)Y2Si2Al2O2N5:Eu2+ has been studied by Liu et al. [142]. Phosphors containing various proportions of Ba2+ and Sr2+, and of the dopant Eu2+ have been prepared by sintering in reducing atmosphere and suitably characterized by XRD. The luminescence spectra of BaY2Si2Al2O2N5:Eu2+ are dominated by a broad emission band that shifts from 510 to 533 nm when the dopant concentration is increased. The excitation is broad with features at 291, 354, and 426 nm. In this case, the optimal Eu2+ concentration is 8 %, and the corresponding QE is 32.5 % upon 400 nm excitation. Addition of Sr2+ leads to a red shift and to an increase of the emission intensity. The CIE coordinates could be tuned from (0.292, 0.577) to (0.417, 0.516). The thermal quenching behavior of BaY2Si2Al2O2N5 doped with 8 % Eu2+ is similar to the one for the commercial phosphor (Ba,Sr)2SiO4:Eu2+.

The sulfide phosphor Sr1-x Ca x Ga2S4:Eu2+ (with doping in the region of 2 %) has been considered by Do et al. [143]. Samples were prepared by NaBr-aided solid state reaction in flowing H2S. Excitation at 455 nm yields a broad symmetric emission band in the range 500–625 nm, whose maximum is red shifted from 535 to 555 with increasing x (moving from SrGa2S4 to CaGa2S4). This red shift is interpreted on the basis of a stronger ligand field around Eu2+ when Sr2+ is replaced by Ca2+. The excitation spectra are extremely broad in the NUV and the blue regions and are compatible with excitation using NUV, violet, and blue LED chips. The emission efficiency appears to decrease only slightly with increasing x, i.e., when the emission color shifts from green to greenish-yellow.

7 Red Phosphors

As for the green phosphors, red emitting materials can be employed for SSL both upon excitation of a NUV or a blue LED, and therefore, they must be able to convert efficiently radiation in the ranges 380–410 or 450–480 nm to light centered in the red region (625–740 nm). This is not trivial to do in an efficient way, as in many hosts the most common activator ions for pcLEDs phosphors (Ce3+ and Eu2+) do not emit strongly in the red region. In order to achieve efficient red emission based on 5d–4f excitation and emission transitions, it is necessary to rely on host materials in which (1) the centroid of the 4f n−15d 1 configuration is shifted to low energy (nephelauxetic effect, due to increased covalency) and/or (2) the ligand field is strong, lowering the energy of the lowest (emitting) level belonging to the 4f n−15d 1 configuration. This occurs only in few hosts, as shown below.

The alternative is the use of activator ions that emit in the red region, exploiting intraconfigurational transitions. Two such examples are Eu3+ and Mn4+, that provide luminescence through 4f–4f and 3d–3d transitions, respectively. As mentioned above, these intraconfigurational transitions are parity-forbidden through the electric dipole mechanism and are, therefore, weak, and this is not ideal for phosphor performance. These issues will be discussed below.

As in the section above, conventional oxide-based hosts, such as phosphates, silicates, and borates, will be discussed in Sect. 7.1. Molybdates, tungstates, niobates, and tantalates, which are hosts commonly proposed for Eu3+, will be considered in Sect. 7.2. All other host compositions (e.g., nitrides and oxynitrides) will be finally reviewed in Sect. 7.3.

7.1 Red Phosphors—Oxide Hosts: Phosphates, Silicates, Borates

It is well known that the Eu2+ ion in phosphate hosts shows emission bands located in the NUV, blue, and green regions. However, Srivastava et al. [144] have reported on a family of materials, the pyrophosphates Cs2CaP2O7 and Cs2SrP2O7 doped with 1 % Eu2+, which is characterized by relatively sharp emission bands centered at 601 and 567 nm, respectively, i.e., in the orange and yellow spectral regions. The excitation spectra of these bands in both materials show two features extending from 300 to 450 nm, and from 200 to 300 nm, respectively (Fig. 23). The former band exhibits the “staircase” profile which derives from the transition from the 4f 7 ground state to the multiplets (J = 0–6) belonging to the 4f 6[7 F J]5d 1 excited configuration [145], and that is typical of Eu2+. The decay time of the luminescence is 1.2–1.3 µs, characteristic of Eu2+, and the Stokes shift exceeds 8000 cm−1. This is much larger than the most frequent value for Eu2+ emission in solids (1350 cm−1), and explains the location in the yellow–orange region of the Eu2+ emission. After a thorough discussion, the authors conclude that this is an extreme case of Eu2+ interconfigurational transition. This conclusion is reflected in the lack of Eu2+ doped phosphate materials emitting in the red spectral region and suitable as red phosphors.

Fig. 23
figure 23

The excitation (solid line; λ em = 600 nm) and the emission (dotted line; λ ex = 360 nm) spectra of Cs2CaP2O7:Eu2+ at T = 80 K, adapted from [144]

A different approach was used by Peng et al. [146] who proposed the borophosphate materials MBPO5 (M = Ba, Sr, Ca) doped with the Bi2+ ion as novel orange and red phosphors for pcLEDs. In this case, the optical features of Bi2+ (valence shell 6s 26p 1) are transitions of the 6p–6p type that gain intensity due to mixing with higher even parity states [147]. The materials can be obtained via solid state reaction with a doping level of 0.5 % under air atmosphere, and were characterized by XRD and vibrational spectroscopy. Emission spectra present strong bands peaking at 650–660 nm with shoulders at higher wavelengths. Broad excitation bands occur in the NUV blue (350–450 nm) with weaker bands at higher wavelength, close to 600 nm, and at lower wavelength, close to 370 nm. The emission transition is assigned to 2 P 3/2 → 2 P 1/2. For M = Ba, the decay time of the emission is 22 µs. The strong absorption in the blue indicates that these materials could be suitable red phosphors for SSL.

The crystal chemistry and the luminescence of the Lu2CaMg(Si,Ge)3O12 materials doped with Ce3+ has been investigated by Ye et al. [28]. Samples containing 4–6 % (nominal) of the dopant ions were prepared by solid state reaction under reducing atmosphere at 1300–1450 °C. The composition regions for phase stability in this system have been established. It was found that in these materials the Ce3+ ion gives rise to luminescence that is significantly located more clearly in the red than for typical Al-based garnet phosphors. For instance, (Lu0.94Ce0.06)2CaMg2Si2.2Ge0.8O12 gives rise to a broad band peaking at about 600 nm upon excitation at 470 nm. For this composition the thermal quenching is slightly more efficient than for YAG:Ce3+. These phosphors are in general characterized by a high QE.

A different approach to the development of red phosphors has been proposed by Carrasco et al. [148] who have investigated materials with composition Ca3Tb2–x Eu x Si3O12 (x = 0–2) having the silico-carnotite structure. In this case, the 4f–4f emission of Eu3+ occurs from the 5 D 0 excited level and is composed of a dominant sharp peak located at 612 nm. This efficient emission is normally difficult to excite due the small absorption cross section of the 4f–4f transition of the dopant. However, in the compositions with x in the range 0.02–0.10, the 5 D 0 luminescence can be efficiently sensitized by the Tb3+ ion, which is present with a concentration 20–100 times higher than the activator. In fact, excitation at 377 nm, in the 5 D 3 level of Tb3+, gives rise to fast migration in the Tb3+ subset followed to energy transfer processes populating the 5 D 0 level of Eu3+. The overall transfer efficiency reaches 94 % in the sample with x = 0.10. So, a NUV chip can be used to excite red emission efficiently. This concept was proposed by Bettinelli et al. [149] for the eulytite material Sr3Tb0.90Eu0.10(PO4)3 (where the transfer efficiency is 93 %) and by Xia et al. [150] for the chloroborate Ba2Tb(BO3)2Cl:Eu (where both Eu3+ and Eu2+ coexist, and the transfer efficiency is 72 %). In the case of the Ca3Tb2-xEuxSi3O12 silico-carnotite with x = 0.10, the CIE coordinates upon 377 nm excitation are (0.633, 0.355) (Fig. 24), in the red region, and the Eu3+ emission decay time is only weakly temperature dependent.

Fig. 24
figure 24

CIE diagram coordinates of Ca3Tb2Si3O12 (CaTbSi) and Ca3Tb2-x Eu x Si3O12 excited at 377 nm and Ca3Eu2Si3O12 (CaEuSi) excited at 393 nm, and a typical picture of the phosphors when excited at 365 nm, adapted from [148]

7.2 Red Phosphors—Oxide Hosts: Molybdates, Tungstates, Niobates, and Tantalates

A wealth of papers have appeared in the literature dealing with oxyanion compounds of transition metal ions (in particular molybdates, tungstates, niobates, and tantalates) containing Eu3+ ions [151167]. These studies involve the luminescence from the 5D0 level of this ion, having a sharp peak located at about 610–620, i.e., in the pure red region (see above) (Fig. 25). In fact, Y2O3 and YVO4 doped with Eu3+ have been employed as red phosphors for a long time [168, 169]. Moreover, it has been shown that the optimum red phosphor for a warm-white pcLED having simultaneously a high luminous efficacy and a high CRI should have a narrow emission band between 615 and 655 nm [3, 170]. These conditions are perfectly met by the 5 D 0 → 7 F 2 band. NUV and blue chips in principle can excite the Eu3+, having absorption bands peaking around 395 and 465 nm, but the absorption cross sections of these bands is quite low, due to the fact that they are due to 4f–4f transitions (7 F 0 → 5 L 6 and 7 F 0 → 5 D 2, respectively).

Fig. 25
figure 25

Temperature dependent emission spectra (λ ex = 310 nm) of Li3Ba2La3(MoO4)8:Eu3+ samples (a) doped with 20 % Eu3+ and its (b) estimation of T q (TQ1/2), temperature at which the intensity decreases to 50 %; (c) doped with 100 % Eu3+ and its (d) estimation of Tq (TQ1/2). All figures are adapted from [151]

Intuitively, phosphor performance should increase with increasing concentration of Eu3+, as this should lead to an increase in the population of the emitting level and, therefore, to higher emission intensity. However, this is not the case, as the increased concentration of the activator ions usually results in a decrease of the emission QE. This is due to the activation of energy transfer processes involving the excited states of Eu3+ and depopulating in a non-radiative way the light emitting levels. This behavior is known as concentration quenching and is normally limited to a few percent of the activator ion concentration in a given host in order to obtain the optimum QE and emission intensity [171]. On the other hand, energy transfer processes between Eu3+ ions strongly depend on the interionic distance [172]. If these ions are kept a long way apart one from another by large, bulky anions, high QEs can be achieved even for fully concentrated hosts. Good results have been obtained for concentrated molybdates [151] in which concentration quenching is prevented by the presence of large anions between Eu3+ centers. Another approach is based on the presence of efficient energy transfer from the ligand-to-metal CT states of the oxyanion (molybdate, tungstate, etc.) to the excited states of Eu3+. This can increase the excitation efficiency of the Eu3+ ion, but can be useful only when the electric dipole allowed CT absorption bands are located in spectral ranges where NUV or blue chips emit [152]. As mentioned above, molybdate, tungstate, niobate, and tantalate hosts activated with Eu3+ have been widely investigated [151167], also exploting co-doping with Sm3+, but only in a limited number of cases the two strategies described above (high Eu3+ concentration and CT → Eu3+ energy transfer) have been successfully exploited. For this reason, not all the phosphors that have been suggested are really fully suitable for pcLEDs applications.

7.3 Red Phosphors—Other Hosts

Oxyfluorides, and especially fluorides, have been proposed as materials for red phosphors applications. Huang et al. [17] have investigated the Ca2RF4PO4:Eu3+ (R = Gd, Y) new oxyfluoride phosphors. Samples were prepared by solid state reaction and characterized by XRD. The excitation band of the sharp main emission line at 611 nm consists of a broad band around 280 nm of CT nature, and sharp lines comprising the 7 F 0 → 5 L 6 at 394 nm and the 7 F 0 → 5 D 2 4f–4f transitions. The Eu3+ concentration can be increased up to 40 % with relatively weak concentration quenching. The reported QE values are (presumably for 5 % Eu3+ doping) 64 % for R = Gd and 48 % for R = Y upon excitation at 394 nm, and the thermal quenching is weak up to 150 °C.

Setlur et al. [136] have proposed the development of a red phosphor for pcLEDs based on the ion Mn4+ as a dopant in octahedral coordination. This ion has a 3d 3 configuration (like Cr3+) and in octahedral sites it has a 4 A 2 → 4 T 2 spin allowed, parity forbidden transition that matches excitation in the violet or blue spectral regions. The emission transition is 2 E → 4 A 2 with a sharp profile and peaking around 632 nm in fluoride hosts [173] (Fig. 26). Because of the strong and broad absorption/excitation bands in blue and sharp emission peaks in red for K2SiF6:Mn4+, Lv et al. [174] have suggested that K2SiF6:Mn4+ is fairly suitable to be used in the pcWLEDs made of yellow phosphor YAG:Ce3+ and blue LEDs to obtain a warm white light emission (CCT = 3900 K) with a high CRI (89.9) and a high efficiency (LE = 116 lm/W). Another fluoride red phosphor K2TiF6:Mn4+ also shows a bright and warm white emission with a high CRI as employed with YAG:Ce3+ and blue LEDs [175]. The phosphors K2TiF6:Mn4+ and K2SiF6:Mn4+ show QE values similar to commercial garnets doped with Ce3+, and their thermal quenching at 150 °C is less than 5 %. They could suffer from saturation based quenching due the long decay time of 2 E → 4 A 2, but this issue can be virtually eliminated by using a remote phosphor configuration reducing the incident radiation flux [176]. The thermal stability for K2TiF6:Mn4+ (activation energy ΔE = 0.34 eV) has shown to be better than nitride compounds (ΔE ≈ 0.25 eV) based on thermal quenching of emission intensity [175]. Apart from the temperature effect, the relative emission intensity of K2TiF6:Mn4+ and K2SiF6:Mn4+ is shown to depend on the initial KMnO4 concentration, where KMnO4 is used as the source of Mn4+ dopants when Mn7+ in KMnO4 is reduced by H2O2 in HF solution. The decrease of emission intensity because of too high concentrations of KMnO4 is attributed to the concentration quenching effect [175] or the formation of the MnO2 phase [174]. These materials are very interesting phosphors. The investigation of this class of luminescent materials has been extended by the research group based at Gunma University in Japan, which has considered and investigated the hosts above, and several additional ones (Na2SnF6, Cs2SnF6, Cs2SiF6, Cs2GeF6, KNaSiF6) doped with Mn4+ [177180]. All these phosphors emit in the red region, and their synthesis, characterization and spectral properties have been reported. These ideas have been extended to oxide hosts (e.g., SrGe3.495Mn0.005Si0.5O9 [181]).

Fig. 26
figure 26

a Excitation (λ em = 632 nm) and emission spectra (λ ex = 450 nm) for K2TiF6:Mn4+ (solid line) and K2SiF6:Mn4+ (dotted line). b Relative intensity vs. temperature (λ ex = 450 nm) for K2TiF6:Mn4+. The line in panel (b) is a guide to the eye. Both figures are adapted from [136]

Le Toquin and Cheetham [182] have proposed for the first time a Ce3+ doped nitride phosphor emitting in the red spectral region. The material has composition CaSiN2 doped with 3 % Ce3+ and was synthesized at high temperature under flowing N2. The emission profile shows a broad band extending from 550 to 700 nm with maximum at 625 nm. The corresponding excitation peaks around 535 nm and spans the region between 425 and 575 nm. Upon laser excitation at 515 nm an external QE of about 40 % is obtained. Substitution of Ca with Mg or Sr, and of Si with Al, allows shifting the emission and excitation maxima, confirming the potential of these phosphors in combination with a blue LED. The nitride SrAlSi4N7 has been proposed as a host for the lanthanide ions Ce3+ and Yb2+ by Zhang et al. [183]. SrAlSi4N7 codoped with Ce3+ and Li+ (for charge compensation) is characterized by a broad emission band in the range 450–700 nm, peaking at about 500 nm, with a complex excitation spectra dominated by broad features around 275 and 425 nm. On the other hand, SrAlSi4N7:Yb2+ gives rise to a single peak at 600 nm with excitation in the NUV-blue region between 325 and 500 nm. Concentration quenching has been observed and the optimal concentrations for the two phosphors are around 1 %. Ce3+ retains 80 % of the emission intensity when temperature is increased from 30 to 200 °C, whilst Yb2+ decreases to 40 % from 50 to 150 °C. The two phosphors emit yellow–red and red light, respectively, and could be useful for pcLED applications.

Eu2+ doped phosphors in the nitride system Ca3N2-AlN-Si3N4 have been investigated by Uheda et al. [18]. The authors found the new red phosphor CaAlSiN3:Eu2+, that can be synthesized at high temperature under N2 pressure. This material can be effectively excited using blue GaN and NUV InGaN LED chips and exhibits a broad emission band peaking around 650 nm. The optimum concentration of the dopant is 1.6 % and the CIE color coordinates of the emission are (0.667, 0.327) (for 0.8 % Eu2+). Wang el al. [184] have found that the strongest emission intensity is from CaAlSiN3:x %Eu2+ (x = 1.5) and further increasing Eu2+ content gradually reduce emission intensity due to energy transfer between activator ions. A red shift effect was observed as increasing Eu2+ content from 0.05 to 2.0 %. This phenomenon was elucidated by the site-to-site energy transfer, i.e., there are more chances to have energy transfer to lower energy sites at the local environments of Eu2+. To solve the red-deficiency problem for some types of pcWLEDs, Lin et al. [185] have fabricated a warm pcWLED by introducing a red CaAlSiN3:Eu2+ phosphor layer onto a yellow YAG:Ce3+ layer coupling to a InGaN LED chip. When Eu2+ concentrations vary from 0 to 50 wt%, the emitting white light become warmer and high color saturated (CCT: from 5426 to 2763 K, CRI: from 67.3 to 80.2); however, the LE decreases from 113.3 to 78.7 lm/W. As for thermal stability of CaAlSiN3:Eu2+, Piao et al. [186] indicated that the thermal quenching is inefficient as CaAlSiN3:Eu2+ shows 83 % of its initial (RT) emission efficiency at 150 °C as compared to 60 % of RT emission efficiency for YAG:Ce3+ [186]. Furthermore, they have demonstrated that Ca0.98Eu0.02AlSiN3 is chemically stable at high temperature. More specially, Ca0.98Eu0.02AlSiN3 is not oxidized until 400 °C which is rather higher than the operating temperature for pcWLEDs. Another interesting Eu2+-doped red nitridosilicate phosphor, Sr2Si5N8:Eu2+, has been reported by Xie et al. [187]. They have synthesized this phosphor by utilizing a solid-state reaction, which is a rather simple, cheap and high-yield method. Although the synthesized phosphor are complex phosphors of Sr2Si5N8:Eu2+ (~64 wt%) and Sr2SiO4:Eu2+ (~36 wt%), Sr2SiO4:Eu2+ has only 0.04 % of the emission intensity of Sr2Si5N8:Eu2+, implying that the luminescence from Sr2SiO4:Eu2+ can be ignored. Additionally, the complex phosphors have shown comparable luminescence properties with respect to Sr2Si5N8:Eu2+ prepared by traditional methods. The excitation spectrum has revealed a strong and broad band in the range of 200–600 nm. The emission band is centered at 618 nm under 450 nm excitation. The emission becomes more red shifted as increasing Eu2+ concentration although the excitation spectrum is not affected significantly. Its internal and external quantum efficiency were measured to be around 75–80 and 64 %, respectively. As observed for CaAlSiN3:Eu2+, the emission intensity of Sr2Si5N8:Eu2+ is about 86 % of that at RT as increasing temperature to 150 °C, indicating promising thermal stability in lighting brightness. Thus, CaAlSiN3:Eu2+ and Sr2Si5N8:Eu2+ can be suitably used for SSL applications, particularly for lowering CCT and increasing CRI.

A different approach was followed by Liu et al. [30] who studied a β-SiAlON oxynitride host activated with Pr3+. In this case, excitation occurs at 460 nm (corresponding to the emission of a blue chip) and is due to the 4f–4f transition 3 H 4 → 3 P 2. The emission spectrum for the sample of composition Si5.9Al0.1O0.1N7.9:Pr0.016 is dominated by several sharp bands in the region 600–660 nm, attributed to the 4f–4f transitions originating from the levels 1 D 2 (to 3 H 4) and 3 P 0 (to 3 H 6 and 3 F 2). At temperatures above RT, emission bands from 3 P 0 become weaker, and from 1 D 2 stronger, due to 3 P 01 D 2 non-radiative relaxation. This causes the integrated emission intensity to increase up to 423 K, and then slightly decrease up to 573 K. However, the colour coordinates of the emission only shift from (0.688, 0.312) at 298 K to (0.685, 0.315) at 573 K. The high thermal stability of β-SiAlON:Pr3+ makes it an alternative for red emitting pcLEDs phosphors, despite the relatively low absorption cross section for excitation.

The chalcogenide phosphors Ca1-x Sr x (S y Se1-y ):Eu2+ have been considered by Nazarov and Yoon [188]. Upon excitation at 460 nm, these phosphors emit in the red, orange, or yellow spectral ranges depending on the x and y values. The authors have proposed a simple empirical model to choose the composition, in terms of x and y values, that can provide emission spectra peaking at a selected wavelength. In the case of red phosphors, emitting between 600 and 650 nm, the sum x + y must be in the range 0 ≤  x + y <  1.25. Kuo et al. [29] have proposed the oxysulfide CaZnOS doped with Eu2+ as a candidate red phosphor working upon blue excitation. Single phase Ca1–x Eu x ZnOS materials can be synthesized by solid state reaction at high temperature under a reducing atmosphere. The composition with x = 0.04 was found to be optimal in terms of emission intensity; it is characterized by a broad excitation feature extending from 400 to 600 nm and the emission peaks at about 650 nm. The Ca0.96Eu0.04ZnOS phosphor was found to have CIE coordinates of (0.69, 0.31) and a QE of 35.5 %. The thermal stability of the emission is superior to the one of SrS:Eu2+ in the temperature range 50–125 °C.

8 Conclusions

In this chapter dedicated to inorganic phosphors and their characteristics for SSL, our goal was to give to the readers an idea of the basics, the scope and the diversity of this research field. We have tried to point out the current main avenues of research and to indicate the most important topics, especially the ones still needing major advancements or breakthroughs. In our opinion, it is important to highlight that the study of inorganic materials for pcLEDs is a truly interdisciplinary field that requires competence and creative research in many different areas of investigation, such as luminescence and optical spectroscopy, inorganic chemistry, solid state physics, and materials science. In particular, we have given a snapshot of current research and demonstrated the important role of detailed local structural and dynamical investigations of phosphors such as YAG:Ce3+. Although YAG and other garnet-type host lattices continue to be considered very promising materials, it should be noted that a variety of other classes of phosphors have been developed and are now attracting increased attention. Examples of such are silicates [189], nitrides [190192], and oxy-nitrides [193, 194], to name a few.

With a view to the future, it is clear that there exists great scope for further structural and dynamical investigations in order to increase the current understanding of these phosphors. In particular, we foresee an increasing use of PDF analysis and reverse Monte-Carlo modeling of neutron and X-ray total scattering data for the investigation (and re-investigation) of local structural details, such as bond distances and angles, and nature of the activator ion sites, for example, of both well-known and new phosphors. Furthermore, we will follow with great interest the upgrade and methodological developments of instruments and techniques in a wide context. As an example, the development of new neutron and X-ray sources and free electron laser (FEL) facilities, will soon allow for investigations that have previously not been within reach.

Similarly, the use and development of computational techniques are equally important. Computational modeling of materials is particularly useful prior to experiments to predict the outcome and thus optimize the use of experiment beam time, as well as to guide future experiments. Likewise, experimental data can be used to test and benchmark computational methods and results.

Another interesting perspective for the future comes from the observation that an efficient way to discover and develop new phosphors for pcLEDs derives from the modification of the local structure and luminescence tuning of existing phosphor materials. This concept consists basically on strategies such as control of the doping level, cationic substitution, anionic substitution, cationic–anionic substitution, crystal-site engineering, and mixing of nanophases, and is discussed at length in an important recent review on Ce3+ and Eu2+ doped phosphors [195]. These strategies were successfully employed in the tailoring of novel phosphors for wLEDs, as shown in recent papers [196198]. This approach seems to be promising for significant advancements in the development of new and improved luminescent materials.

We would like to conclude this last section with a caveat. Phosphors for LEDs are a very important and fast developing field, and this should always be taken into account. In some cases, fast development may lead to fast publication, and it may happen that in some articles the interpretation of the experimental results on the basis of solid and reliable models may not be entirely correct. For this reason, we point out that in a few of the multitude of papers cited above the interpretation of the experimental results is not flawless, and must be taken with some due care, especially when energy transfer and relaxation mechanisms are discussed in detail. We have decided to make reference to these papers anyway, due to the presentation of highly interesting experimental data.