Structure evolution and photoluminescence of Lu3(Al,Mg)2(Al,Si)3O12:Ce3+ phosphors: new yellow-color converters for blue LED-driven solid state lighting

(x = 0.5–2.0) series; as a result, the CeO8 polyhedrons were compressed and the emission got red-shifted from green to yellow together with the broadening. The evolution of, the unit cell, the local structural geometry as well as the optical properties of Ce in these garnet creations, in response to the gradual Mg–Si substitution for Al–Al, were studied by combined techniques of structural refinement and luminescence measurements. The new composition Lu2.97Ce0.03Mg0.5Al4Si0.5O12 was comprehensively evaluated regarding its potential application in blue LED-driven solid state white lighting: the maximum emission is at 550 nm under lex = 450 nm; the internal and external quantum efficiencies can reach 85% and 49%, respectively; a 1-phosphor-converted wLED lamp fabricated using the as-prepared phosphor exhibits the luminous efficacy of 105 lm W , the correlated color temperature of 6164 K and the color rendering index (Ra) of 75.6. The new solid solution composition series is open for further optimization to enhance the competence for commercial consideration.


Introduction
White light emitting diodes (wLEDs) exhibit many advantages, such as long lifetime, high luminous efficacy and eco-friendliness over conventional incandescent bulbs or fluorescent lamps. 1 Nowadays, the lighting market undergoes an accelerating replacement of such traditional lighting sources by the wLEDs. Besides, it is also entering into several new application fields, 2 including backlights for liquid-crystal displays and portable electronics, automobile headlights, medical and architecture lighting, plant cultivation, etc.
The dominating technique to achieve white light output in commercial wLEDs is the optical frequency conversion in the phosphor layer: typically, (Y,Gd) 3 (Al,Ga) 5 O 12 :Ce 3+ (YAG:Ce) is dispersed in a transparent polymer matrix, and then, coated on a blue InGaN LED chip. The blue light from the LED is absorbed by the Ce 3+ ions of YAG:Ce via the allowed 4f -5d transition that generates the excited state, which then gives a wide-band yellow emission band through the inverse 5d -4f transition. By combining the blue chip radiation and the yellow phosphor emission, the cool white light is obtained. 3,4 Initially, YAG:Ce was proposed for use in displays in 1967, 5  it was quickly employed for white light generation when the bright blue InGaN diode was invented. 6 Although applying YAG:Ce in wLEDs (patented by Nichia, Japan) is nearly ideal, there is still a strong drive in developing new yellow phosphors that can be considered as alternative candidates for YAG:Ce, so that wLED devices could be produced in an IP (intellectual patent)complication free manner. For example, (Tb, Re) 3 (Al, Ga) 5 O 12 :Ce was developed by Osram to avoid IP-complication. The rapid development of new solid-state lighting devices also motivates the search for new phosphors with specific properties; studies reporting any ideas to discover/develop new phosphor are of high scientific interest. So far, several yellow (oxy)nitride phosphors excitable by blue LEDs, such as Ca-a-Sialon:Eu 2+ , 7 (La,Ca) 3 Si 6 N 11 :Ce 3+ , 8 Sr 0.5 Ba 0.5 Si 2 O 2 N 2 :Eu 2+ , 9 have been reported; but the costly synthesis and the irreversible oxidation degradation of (oxy)nitrides are still the obstacle and weakness. In this paper, we report the development of new yellow oxide phosphor via the solid solution design, the emission feature of which is observed to be readily similar to that of YAG:Ce. The solid solution design starts with Lu 3 Al 5 O 12 :Ce (LuAG:Ce) where a wide-range Mg-Si for Al-Al substitution is performed to create the artificial garnet series of Lu 3 (Al,Mg) 2 -(Al,Si) 3 O 12 :Ce 3+ .
The solid solution design in partial changes of the host compound through cation/anion, 10-12 cation-pair, 13 or chemicalunit substitition 14 has been adopted as a way to tune the spectroscopic properties of Eu 2+ /Ce 3+ which are ligand-sensitive activators in phosphors. 15 In modern photonic technology, doping at, for example, the cation site, is widely used for tuning the structural, linear/nonlinear optical, ferroelectric and other physical properties of functional crystals. [16][17][18] One doping element, as a rule, is applied to doping and the expected solid solution property can be speculated over the solubility range by different empirical relations. When two doping elements with different ion charges and radii are taken, the situation becomes more complicated and the solid solution behavior is less predictable, because, in many cases, the doping ion distribution over the crystallographic sites and the defect formation mechanism are unclear. 19,20 Two doping elements with appropriate ion radii should be selected correlatively to maintain the electroneutrality of the crystal lattice and suppress the point defect generation. Frequently, the search for the optimal solid solution composition is nontrivial as more conditions should be fulfilled in parallel to reach the wide-concentrationrange low-defect solid solution formation. Such observation in tuning these physical properties provides a hint at the photoluminescent solid solution composition design.
In a number of solid solution phosphor cases, desired tuning of a certain optical property has led to undesired negative trends in other significant properties; for example, the color rendering improvement was accompanied by a drop of quantum efficiency. Therefore, when developing new phosphors via the solid solution design, it is necessary to wisely choose the composition design strategy and the starting host. In our practice, several principles have been used to guarantee that the new phosphor creations are of: (1) ease in pure phase formation; (2) less defects generated in the crystal structure; (3) high quantum efficiency and (4) high thermal stability of emission. These rules, based on our experience, consist of (but not limited to) (1) starting with a crystal structure that associates with simple dopant-site geometry; (2) starting from a crystal structure that has a comparatively rigid framework; (3) starting with a phosphor that has the emission already close to a desired emission spectral position; (4) designing the substitution on a low-symmetry site; (5) designing the cation/anion substitution in equal valence and similar radius, or substitution using chemical units with an equal valence and a similar polyhedron size.
Here, LuAG:Ce is chosen as the starting composition to achieve the yellow emission. First, the garnet structure is, at present, the only known oxide host wherein the Ce 3+ dopant can be effectively excited by blue light and emit in the greenorange spectral region with high quantum efficiency and high thermal stability. Reasonably, such advantages can be inherited by the solid solutions derived from it. Second, LuAG:Ce emits a green color, which makes it much easier to reach yellow emission via red-shift. Third, as seen in earlier studies, the garnet structure is comparatively stable in reference to various cation substitutions at the Lu 3+ or Al 3+ crystallographic sites. LuAG, in the cubic garnet structure (Fig. 1a), has a stoichiometric formula of Lu 3 Al(1) 2 Al(2) 3 O 12 or Lu 3 Al(1) 2 [Al(2)O 4 ] 3 , which can be generalized as A 3 B 2 C 3 O 12 or A 3 B 2 (CO 4 ) 3 . Atoms A occupy the 24c positions with point symmetry 222; its coordination polyhedron, a square antiprism, is built with eight coordinating oxygen atoms; atoms B occupy the 16a positions (symmetry of À3) in the oxygen octahedron centers, and atoms C occupy the 24d positions with symmetry 4 in the oxygen tetrahedron centers. The existence of three different oxygen polyhedrons, which considerably differ from each other in geometry and occupy B30% of the unit cell volume, opens a wide opportunity for the substitution design and provides the developed isomorphism. 21 Theoretically, the spectral red-shift of the Ce 3+ emission can be induced by a larger centroid shift, crystal field splitting or Stokes shift. From the compositional viewpoint, a higher lattice covalency such as using less electronegative N 3À to replace O 2À will benefit for such a purpose. As a case, the emission red-shift of Y 3 Al 5 O 12 :Ce 3+ has been previously reported by introducing the Si 4+ -N 3À bond into the lattice. 22 However, such a red-shift induced simultaneously a severe drop in quantum efficiency and thermal emission stability. The O 2À ion in Y 3 Al 5 O 12 occupies the general site (96 h) which is highly multiplied with a lot of operation symmetry; substitution of O by N will generate too many defects and decrease the rigidity of the framework structure. Therefore, introduction of a foreign N ion into the O site is reasonable to increase the host lattice covalency but fails to be a wise choice for the emission red-shift. Alternatively, from the crystal structure viewpoint, compression of CeO 8 may lead to a larger crystal field splitting and the aimed emission red-shift. Potential ways for the crystal structure modification via the solid solution design can be: (1) replacing Lu 3+ at the dodecahedral site with bigger trivalent cations (Fig. 1b); for example, (Y,Lu) 3 Al 5 O 12 :Ce shows the green to greenish-yellow emission shift with increasing Y/Lu ratio; 4,23 (2) replacing Al-Al in tetrahedral/octahedral coordinations with an equal-valence cation pair such as Mg(Ca)-Si (Fig. 1c); for example, the Mg-Si(Ge) incorporation into YAG has been seen to induce the red-shift effect. 24 Compared with the first approach, the second one is more likely to save the structure stability because the tetrahedron (octahedron) expansion can be compensated by a nearby octahedron (tetrahedron) contraction, so that the unit cell volume variation is minimal.
Applying this way where Al(1) is partially replaced by Si, Ge or Ti, SiO 4 , GeO 4 or TiO 4 tetrahedra are formed; simultaneously, Al(2) is replaced by a divalent cation (for charge neutrality) like Ca/Mg, new garnet structural crystals, Lu 2 CaMg 2 (Si,Ge) 3 33,34 have been developed. The Mg-Si cation pair substitution of Al-Al in YAG:Ce, where red-shift of the emission is obtained, seems to be the most interesting variant. 24 Therefore, in LuAG:Ce, the Mg-Si for Al-Al substitution with high probability could result in a stable garnet structure, and meanwhile, the emission can be tuned to be a pure yellow position. Thus, the new solid solution phosphors are designed as Lu 3 (Al 2Àx Mg x )(Al 3Àx Si x )O 12 :Ce 3+ (x = 0.5-2.0) to see the substitution effect. Several critical parameters of the new phosphor for application in the blue LED-driven solid state white lighting including the emission/excitation, quantum efficiency, and thermal emission stability were evaluated. The powder XRD data were collected on an X-ray diffractometer (XRD) (Smartlab, Rigaku, Tokyo, Japan) with CuK a radiation (l = 1.54056 Å), operating at 45 kV and 200 mA, and using the step size of 0.021 with a scan speed of 41 min À1 . Crystal structure refinements employing the Rietveld method were implemented using TOPAS. 35 The morphology of the phosphors was observed by field-emission scanning electron microscopy (FE-SEM, S4800, Hitachi, Tokyo, Japan). Photoluminescence spectra were recorded using a fluorescence spectrophotometer (F-4500, Hitachi, Tokyo, Japan) with a 200 W Xe-lamp as an excitation source. The thermal stability of the emission was investigated in the range of 30-200 1C using a combined setup consisting of a Xe-lamp, a Hamamatsu MPCD-7000 multichannel photodetector and a computer-controlled heater. The method to determine the internal and external quantum efficiency can be found in our previous publications. 36,37 The wLED lamps were fabricated by Everfine Co., Hangzhou, China, using the as-prepared Lu 3 Mg 0.5 Al 3 Si 0.5 O 12 :Ce 3+ ; the phosphor was mixed with blending resin (Dow Corning s OE2140) and then coated onto a blue InGaN LED chip (450 nm). The spatial radiation spectrum was measured using a goniophotometer (LED626, Everfine Co., Hangzhou, China). The optical properties of the wLEDs were evaluated using a spectroradiometer (LHS-1000, Everfine Co., Hangzhou, China).

Results and discussion
Crystal structure evolution The phosphor series show bright green-to-yellow bodycolor under sunlight and the emission undergoes red-shift as per the Mg-Si incorporation. To explore the crystal structure evolution, Rietveld refinement analysis was performed on the powder XRD patterns, as shown in Fig. 2. All samples exhibit patterns similar to pure LuAG, but in the solid solution garnets, the diffraction peaks slightly shift to the lower 2y angles. Almost all peaks of each pattern can be well indexed by a cubic cell (Ia% 3d) with parameters close to those of LuAG. 38 The samples of x = 0 and 0.5 are phase-pure, while the patterns related to x = 1.0, 1.5 and 2.0 contain a small amount (o5 wt%) of Lu 2 SiO 5 . For the refinements, the crystal structure of LuAG was taken as a starting model: the C site (24d) was occupied by Al/Si ions, the B site (16a) was occupied by Al/Mg ions and the A site (24c) was occupied by Lu/Ce with fixed occupancies according to the nominal solid solution formulas. The processes were stable and ended with low R-factors, as provided in Table S1 (ESI †). The atom coordinates and main bond lengths are given in Tables S2  and S3 (ESI †), respectively.
The cell parameters show a linear increase with increasing x (Fig. 3a), which confirms the suggested chemical formulas of these compounds, as the averaged effective ion radius (IR) of the Mg-Si pair is larger than that of Al-Al in respective octahedral and tetrahedral coordination.  Table S3, ESI †) indicates the incorporation of bigger Mg replacing a smaller Al at the B site. Therefore, the proposed structural model is appropriate, and accordingly, the local structure evolution around the CeO 8 polyhedron with x variation can be revealed, as depicted in Fig. 4.
This model for local structure evolution enables us to explain the observed trend of the four increasing Lu/Ce-O bond lengths ( Fig. 5b and Table S3, ESI †) and the other four decreasing Lu/Ce-O bond lengths (Fig. 5c and Table S3, ESI †) with the x increase. The preferable site occupation of the Si/Mg ions leads to the contraction of the (Al/Si)O 4 tetrahedron and enlargement of the (Al/Mg)O 6 octahedron with x increase. Accordingly, the O atom positions get shifts followed by the local structure variation around the Lu/Ce ions (Fig. 5a). Previously, such bond length evolution was assumed in an analogue study on Y 3 (Al 2Àx Mg x )(Al 3Àx Si x )O 12 :Ce 3+ , 24 which gave them a way to explain the emission red-shift, but the structural variations have not been quantified. The bond length    The distortion of the (Lu/Ce)O 8 polyhedron was then evaluated. Generally, three modes, the symmetric/asymmetric stretching (breathing), bending, or twisting, can cause displacement of (Lu/Ce)O 8 . 34,40 In particular, the contribution of the bond length change to the polyhedron distortion can be quantitatively calculated using the equation, 41 D = (1/n) P ((L i À L av )/L av ), where L i is the distance from the central atom to the i-th coordinating atom and L av is the averaged bond length. As seen from Fig. 5d, the D value decreases with increasing x. This is due to the fact that the initially long d(Lu/Ce-O) bond lengths gradually decrease with increasing x (Fig. 5c), while the initially short d(Lu/Ce-O) bond lengths increase with the increase of x (Fig. 5b). As a result, all the (Lu/Ce-O) bond lengths are becoming closer to the averaged value, and the distortion of the (Lu/Ce)O 8 polyhedron decreases if solely considering the effect of the bond length variation. Moreover, it should be noted that not only the bond length change leads to the polyhedron distortion: synchronously with the bond length uniforming, the bond angles also change with increasing x. The two bond angle groups vary in the opposite directions (Fig. 6). This process also distorts the (Lu/Ce)O 8 square antiprism and somewhat changes the local crystal field strength on Ce 3+ .

Micro-morphology
The morphology of the Lu 3 (Al 2Àx Mg x )(Al 3Àx Si x )O 12 :Ce 3+ (x = 0.5) phosphor was observed (Fig. 7). The powder products of the solid state reaction, after post-grinding, contain microcrystalline particles with different sizes (1-8 mm) and morphologies (Fig. 7a). Those relatively big ones mainly consist of aggregated grains with a diameter of 1-2 mm (Fig. 7b), and these grains tend to be of semi-spherical shape due to the intrinsic cubic garnet structure-governed crystallization. To meet the requirement for advanced wLED application, it is necessary to reduce/ eliminate the aggregation by further synthesis optimization of the solid state reaction or employing other synthesis options such as spray-pyrolysis, co-precipitation or sol-gel.
It is suggested that well-dispersion of fine phosphor particles with spherical morphology will have less scattering of incident light from LEDs and benefit for the efficiency performance. Recently, a pressure-stimulated method (1623 K for 5 h under a N 2 -H 2 gas pressure of 1.5 Â 10 7 Pa) 4 was used to synthesize YAG/LuAG phosphors with a faceted particle of low surface defect concentration. This method may be potentially used for the preparation of Lu 3 (Al,Mg) 2 (Al,Si) 3 O 12 :Ce.

Photoluminescence property
The body-color change of the phosphor series suggests the emission red-shift, which can be confirmed by the excitation/ emission spectra (Fig. 8). The x = 0 sample exhibits a wide asymmetric emission band with two peaks at 512 nm and 542 nm under l ex = 450 nm; the well-resolved double-peak emission originates from the transitions from the lowest Ce 3+ 5d excited state to the two 4f ground state levels ( 2 F 5/2 and 2 F 7/2 ) split by the spin-orbit coupling effect. As the Mg-Si pair is incorporated, the maximum emission gradually red-shifts from 542 nm (x = 0) to 571 nm (x = 2.0). The x = 1.5 and 2.0 samples show very similar emission properties, though the maximum excitation peak slightly shifts from 456 nm to 462 nm. It is seen that the x = 0.5 sample possesses the peaking emission wavelength of 550 nm, which is already close to the emission of typical YAG:Ce phosphor. 42 The fwhm (full-width at halfmaximum) of this emission is seen to be 122 nm. Under the identical excitation, the x = 0.5 sample exhibits the highest emission intensity among this phosphor series.  The energy position of the Ce 3+ 4f 6 5d 1 excited state and the 4f 7 -4f 6 5d 1 transition in the inorganic host will be modified by the case-sensitive covalence and polarizability of the Ce-ligand interaction, 43 as the Ce 5d electron is exposed to its surrounding ligands. The introduction of the Mg-Si pair alters the Ce-O bond lengths (four bond lengths increase and four bond lengths decrease), and the as-observed ''sum'' effect of this structural evolution is the shift of the emission transitions towards lower photon energies (Fig. 8). We tried to probe the energy structure of the Ce 5d and 4f levels in Lu 3 (Al,Mg)(Al,Si)O 12 :Ce 3+ by analyzing the excitation and emission spectra using band-deconvolution by Gaussian components, 44 but the Gaussian fitting gave unsuitable results. The reason can be that the Ce 3+ ions occupy the sites with the cubic-like symmetry but with a high distortion degree; besides, the complicated composition would make the as-synthesized samples more likely to contain different kinds of inherent defects and Ce 3+ ions in the host lattice slightly differ from each other regarding the specific local environment (such as bond length, bond-bond angle). 34 Therefore, the PL spectrum of the x = 0 sample clearly consists of two peaks but that of the solid solution derivatives becomes to be of the Gaussian type despite being slightly asymmetric.
Though it is difficult to quantify the Ce 3+ 5d energy level structure in this series, it is evident that the 4f -2 T 2g (higher energy levels of the 5d excited state) excitation band shifts to the shorter wavelengths, while the 4f -2 E g (lower energy levels of the 5d excited state) excitation band shifts to longer wavelengths when x increases from 0 to 1.5 (Fig. 9a). Thus, the energy difference between the 4f 7 ground states and the lowest 4f 6 5d 1 excited state reduces, suggesting a stronger crystal field splitting effect on the Ce 5d orbital energy. Besides, the broadening of the energy transition bands (Table 1 and Fig. 8) indicates a higher diversity of local environments of Ce 3+ ions in the highly Mg-Si doped garnets, which offsets the average coordinating situation identified by the XRD plus the Rietveld refinement.
The 5d level of Ce 3+ splits into two groups of excited states ( 2 T 2g and 2 E g ) when occupying a site with ideal cubic symmetry. If the cubic symmetry is distorted, the 2 T 2g and 2 E g states further split into five sub-levels. With Mg-Si substitution for Al-Al in the garnet phosphors, the structure distortion degree decreases (if the effect of the bond angle variation on the CeO 8 distortion is neglected) while the crystal field splitting of Ce 3+ increases (Fig. 9a). Such an effect leads to the observed 4f -2 E g Fig. 8 Photoluminescence excitation (PLE) and emission (PL) spectra of Lu 3 (Al 2Àx Mg x )(Al 3Àx Si x )O 12 :Ce 3+ (x = 0-2). PL was excited by 450 nm light and PLE was measured for their respective maximum emissions.  excitation transition shifting to lower energies (longer wavelengths) and the red-shift of the 5d -4f emission transition from green to yellow. Based on such transition paths, the energy level scheme for Ce 3+ in Lu 3 (Al,Mg)(Al,Si)O 12 :Ce 3+ is accordingly proposed, as depicted in Fig. 9(b). The Stokes shift is also roughly estimated, i.e., by the energy difference between the maximums of the excitation and emission. The Stokes shift (listed in Table 1) is in the range of 3783-3987 cm À1 , which, together with the increased crystal field splitting, contributes to the emission red-shift of Lu 3 (Al,Mg)(Al,Si)O 12 :Ce compared with LuAG:Ce.
Thermal stability of emission Lu 3 (Al 2Àx Mg x )(Al 3Àx Si x )O 12 :Ce 3+ (x = 0, 0.5) were subjected to the temperature-dependent emission measurements over a test range of 30-200 1C (Fig. 10). Due to the thermal quenching effect, the emission intensity shows a continuous decrease upon the temperature increase. When it turns up to 150 1C, the peak intensity of the x = 0 and 0.5 phosphors becomes 86% and 83% of those recorded at 30 1C. With the Mg-Si introduction, the thermal emission stability decreases, perhaps due to the narrowed band gap of the host and the increased Ce 3+ crystal field splitting, which reduces the energy displacement between the host conduction band and the Ce 5d levels, and the thermal-assisted ionization becomes easier. Anyway, such performance still remains to be relatively good. The maximum emission is nearly constant upon the temperature variation which benefits for providing the stable luminous performance of a wLED lamp.

Quantum efficiency
Besides the desired emission/excitation spectra, quantum efficiency (QE) is another critical criterion for a new phosphor. The QE values under the excitation of 450 nm are provided in Table 2. The IQE values are ranging from 89% to 66%, which show a continuous decrease when a higher Mg-Si quantity is introduced; however, due to the increasing absorption efficiency, the EQE values of the phosphors firstly increase and then slightly decrease, as compared with the initial LuAG:Ce composition. For the x = 0.5 phosphor, the absorption, internal and external quantum efficiencies are 57.9%, 85.1% and 49.3%, respectively.

Luminescence performance of white LEDs
White LED lamps were fabricated by combining a blue LED chip with the Lu 3 Mg 0.5 Al 4 Si 0.5 O 12 :Ce 3+ phosphor. A reference lamp employing a commercial yellow phosphor was also fabricated. As seen from the electroluminescence spectra (Fig. 11), the primary emission bands are centered at 450 nm and    Fig. 12). The samples exhibit improved emission intensity compared with the previous ones. The LuAG-derived x = 0.5 and x = 1.0 compositions are then compared with a commercial LuAG:Ce product. As seen from Fig. 13, under the same excitation by 450 nm blue light, the x = 0.5 and 1.0 compositions show a yellow emission which is much red-shifted and broader compared with the commercial LuAG:Ce. While, the integrated emission intensities of the two compositions are 83.5% and 79.8% of the commercial LuAG, respectively, which suggests a relatively good performance. Thus, (Lu,Ce) 3 (Al 2Àx Mg x )(Al 3Àx Si x )O 12 (0.5 r x r 1.0) is promising to act as a yellow color converter; and further optimization on aspects of synthesis, Ce concentration, crystallinity/defect and particle size/morphology is necessary prior to commercial consideration.