Charge deformation and orbital hybridization: intrinsic mechanisms on tunable chromaticity of Y₃Al₅O₁₂:Ce³⁺ luminescence by doping Gd³⁺ for warm white LEDs

Abstract

The deficiency of Y₃Al₅O₁₂:Ce (YAG:Ce) luminescence in red component can be compensated by doping Gd³⁺, thus lead to it being widely used for packaging warm white light-emitting diode devices. This article presents a systematic study on the photoluminescence properties, crystal structures and electronic band structures of (Y_1−xGd_x)₃Al₅O₁₂: Ce³⁺ using powerful experimental techniques of thermally stimulated luminescence, X-ray diffraction, X-ray absorption near edge structure (XANES), extended X-ray absorption fine structure (EXAFS) and ultraviolet photoelectron spectra (UPS) of the valence band, assisted with theoretical calculations on the band structure, density of states (DOS) and charge deformation density (CDD). A new interpretation from the viewpoint of compression deformation of electron cloud in a rigid structure by combining orbital hybridization with solid-state energy band theory together is put forward to illustrate the intrinsic mechanisms that cause the emission spectral shift, thermal quenching and luminescence intensity decrease of YAG: Ce upon substitution of Y³⁺ by Gd³⁺, which are out of the explanation of the classic configuration coordinate model. The results indicate that in a rigid structure, the charge deformation provides an efficient way to tune chromaticity, but the band gaps and crystal defects must be controlled by comprehensively accounting for luminescence thermal stability and efficiency.

Introduction

The Nobel Prize in Physics 2014 was awarded jointly to three scientists, Isamu Akasaki, Hiroshi Amano and Shuji Nakamura, to honour their significant invention of blue light-emitting diodes (LEDs), which enabled bright and energy-saving light sources¹. White LEDs, a type of solid semiconductor device, exhibiting inherent superiorities over traditional incandescent and fluorescent lamps in terms of high efficiency, low-energy consumption, eco-friendliness without mercury pollution, long lifetime and chip-proof solid-state encapsulation within a small volume, revolutionized human lighting after Edison’s invention^2,3,4. However, almost all white LEDs consist of a blue LED chip combined with one or more luminescent materials, which convert part of the blue light into other wavelengths^2,3,4. Among the commercially available phosphors, the yellow-emitting Ce³⁺ activated yttrium garnet, i.e., Y₃Al₅O₁₂: Ce (YAG: Ce), is the most commonly used one.

Y₃Al₅O₁₂ has a cubic garnet structure, generally expressed with the formula A₃B₂(CO₄)₃, where the Y (A) occupies a dodecahedral site coordinated with eightfold oxygen atoms, two of the five Al (B) atoms occupy octahedral sites and the other three Al (C) occupy tetrahedral sites^4,5,6,7,8. The excellent performance of YAG in terms of high rigidity, low thermal-expansion, high-optical transparency, high threshold for optical damage and chemical inertness makes it applicable for a wide variety of applications. YAG is a member of the large family of garnet-structure compounds, including Andradite, Goldmanite, Grossular, Toturite, Uvarovite, etc. There has been a rich history of the discovery and usage of garnets. The earliest application dates back to the Bronze Age, when they were used as gemstones and abrasives^4,5,6,7,8. The synthetic YAG single crystal was used as a diamond simulant until recently, due to its fairly high hardness and relatively high refractive index.

The scientific research on YAG has been active since 1928 when Menzer first determined the garnet structure had the space group Ia3d^4,5,6,7,8. In 1964 at Bell Laboratories, Geusic demonstrated a type of solid-state laser that adopted YAG as the host⁹. Since then, the YAG: Nd and YAG: Er solid-state lasers have been widely used in medical surgery, optical communications, coherent laser radar, meteorology and industrial manufacturing. The parity-allowed f-d transition of Ce³⁺ in YAG has a very short lifetime and high luminescence efficiency. In 1967, Blasse and Bril firstly reported YAG: Ce as a new phosphor for flying-spot cathode ray tubes¹⁰. Due to high light yield (20000 photons/ MeV) and fast response time, in the 1990s YAG: Ce and the higher density analogue Lu₃Al₅O₁₂: Ce (LuAG) and (Lu,Gd)₃(Al,Ga)₅O₁₂ were proposed as scintillators for X-ray and gamma-ray radiation detection, used in X-ray computed tomographic scanner (CTS) and positron emission tomography (PET) for medical imaging and in scanning electron microscopy (SEM)^{10,11,12,13,14}.

Ce³⁺-activated YAG has been intensively studied with a renewed interest for applications in white LEDs^{4,15,16,17,18}, since it was first introduced into white LEDs by the Nichia Corporation in 1994. Nearly all the high-power white LEDs that are used in street and home lighting are fabricated using YAG: Ce as a phosphor. In addition, the same garnet structure LuAG: Ce was used a green phosphor for white LEDs. Quite sensibly, manufacturers and many scholars pay close attention to YAG: Ce. In addition to efficiency, the chromaticity and thermal stability of the luminescence of the phosphor are crucial to white LED applications because changes in the emission colour and flux during operation as a result of heat released continuously from the p-n junction of LED chips are highly undesirable^4,15,19. This type of white light produced by mixing the yellow from YAG: Ce and the blue (450–460 nm) from InGaN chips, exhibits a cold colour temperature in addition to poor colour rendering ability^{15,16,17,18,19,20,21,22}, due to its deficiency in the red component in its emission spectrum. However, warm white is generally desired for indoor lighting. To decrease the colour temperature, an appropriate amount of Gd³⁺ usually is doped into YAG: Ce, the emission peak of which can be regulated from approximately 543 nm to maximum 575 nm by tailoring the Gd³⁺ concentration until the collapse of crystal lattice^{15,16,17,18,19,20,21,22}. Nevertheless, the luminescence efficiency decreases and the thermal stability of luminescence deteriorate after doping Gd³⁺ into YAG: Ce. However, the understanding of the mechanisms involved in these processes is limited^4,15.

Pan attributed the red shift of YAG: Ce emission upon Gd³⁺ substitution to the difference in ionic radius between Y³⁺ (0.89 Å) and Gd³⁺ (0.94 Å), based on a strong crystal field splitting due to the lattice expansion²³. However, this is not consistent with the opposite behavior that is commonly reported in the case of an expansion of a host lattice, which usually gives rise to a decrease of the crystal field splitting and results in the shift of 5d–4f emission towards short wavelengths²⁴. Robbin observed an interesting phenomenon regarding the luminescence of YAG: Ce; i.e., the characteristic absorption band of YAG: Ce near 460 nm decreases in intensity, whereas the band near 340 nm increases along with a temperature increase²⁵. Chiang and Bachmam et al. provided an explanation about the worsening thermal luminescence quenching of YAG: Ce upon substitution of Y³⁺ by Gd³⁺ using a quantum mechanically based configurational coordinate model, hypothesizing that the garnet structure becomes soft after incorporating Gd³⁺ into the YAG lattice^21,25. Through analysis of atomic displacement parameters (ADPs) from X-ray and neutron scattering and the evaluation of mean-square relative displacement parameters from extended X-ray absorption fine structure (EXAFS), Ram et al. confirmed that YAG has a very rigid crystal lattice with a high Debye temperature (θ_D, an estimate of the temperature at which all vibration modes of a crystal are activated)¹⁵. According to Ram’s result, we can reasonably conclude that the YAG structure should become more rigid before it collapses into orthorhombic YAlO₃ (YAP) when Y³⁺ is substituted by Gd³⁺, because Gd³⁺ has a larger radius in comparison with Y³⁺ and thus it will suffer intensive compression from other neighbouring atoms in the rigid structure.

Figure 1

Band structures, density of state and partial density of state of Y₃Al₅O₁₂, (Y_2.94Ce_0.06)Al₅O₁₂, (Y_0.75Gd_0.25)₃Al₅O₁₂ and (Y_0.25Gd_0.75)₃Al₅O₁₂.

Figure 2

Emission spectra of (Y_1−xGd_x)₃Al₅O₁₂ excited with 460 nm at room temperature.

Figure 3

The relationship of emission peak, integrated intensity of luminescence and crystal lattice constant of (Y_1−xGd_x)₃Al₅O₁₂: Ce as a function of the x value.

Figure 4

Excitation spectra of (Y_1−xGd_x)₃Al₅O₁₂ upon the strongest emission at room temperature.

Figure 5

The crystal field splitting and the energy level diagram for Ce³⁺ in (Y_1−xGd_x)₃Al₅O₁₂ upon substitution of Y³⁺ by Gd³⁺ and its excitation and emission processes, where Δ = 10D_q indicates the crystal splitting energy acting on the 5d_e-doublet state (²E_g) and the 5d_t-triplet state (²T_g) in tetragonal field, Δ₁₂ indicates an additional crystal splitting energy on the 5d_e-doublet state to form the 5d₁ and 5d₂ states, the SS denotes Stokes shift.

The splitting and transition involving the 5d_t states are omitted.

Figure 6

Diffuse reflection spectra of (Y_1−xGd_x)_2.94Al₅O₁₂: 0.06Ce³⁺.

Moreover, there are two facts that cannot be explained with the traditional configuration coordinate diagram. First, the soft structure indicates a large Stokes’ shift. Accordingly, the nonradiative energy loss with electrons relaxed from the 5d excited state to the 4f ground state through the crossover of their potential parabolas should increase with increasing Gd³⁺. However, the overlap of the normalized excitation and emission spectra of (Y_1−xGd_x)₃Al₅O₁₂: Ce³⁺ decreases with an increase of Gd³⁺ content in x value (as seen below in Fig. 7), indicating that the nonradiative transition does not increase. Second, the thermal vibration of the crystal lattice strongly depends on the weight of the atoms and a heavy weight is helpful resisting thermal vibration. To this end, the thermal stability of YAG: Ce should be improved after doping with Gd³⁺ because of its heavier weight compared with Y³⁺. Practically, this result is not observed. Hence, some mechanisms should exist that play a more important role than the thermal vibration of the crystal lattice and are beyond the range of what the configuration coordinate theory can explain.

Figure 7

The emission and excitation spectra of (Y_1−xGd_x)_2.94Al₅O₁₂: 0.06Ce³⁺ with normalized intensity

(a) and the amplified region showing the crossover of the emission and excitation upon different amount of Gd³⁺.

Dorenbos set forth that the ionization of an electron from the excited state to the conduction band is the genuine mechanism of thermal quenching of Eu^2+,26. Ce³⁺ has a similar electronic configuration to Eu²⁺. Moreover, the temperature-dependent lifetime luminescence measurements for low Ce concentrations show that the intrinsic quenching temperature of Ce is higher than 700 K. Meijerink et al. attributed the low quenching temperature of the commercial YAG: Ce phosphor reported previously to thermally activated concentration quenching⁴. With respect to a further mechanism that causes the thermally activated concentration quenching of YAG: Ce, it was explained by energy migration to defects, but the types of defects that quench luminescence were not uncovered in reference⁴. Robbins also pointed out that “the nature of this quenching is not known, but it may involve energy loss to certain persistent but unidentified defect centres”²⁵.

The anti-site defects, the oxidation of Ce³⁺ to Ce⁴⁺ and the ionized defects caused by the X-ray or gamma-ray radiation have been investigated in the single crystal scintillator of YAG: Ce. More than ten traps in YAG: Ce were reported by Zych²⁷. However, the synthesis temperature of the YAG: Ce phosphor is not as high as the growth temperature of the YAG single crystal scintillator. Moreover, the energy of X-ray and gamma-ray radiation used in scintillators is far higher than the 460 nm photons emitted from LED chips and will induce more defects. So, the defects that appear in the scintillator will not necessarily occur in the YAG: Ce phosphor⁴. In this work, crystal defects and trap depths were identified using the thermally stimulated luminescence. The crystal and electronic structure of (Y_1−xGd_x)₃Al₅O₁₂: Ce³⁺ were investigated using a combination of powerful experimental techniques, assisted with theoretical calculations. The results show that band gap and the energy barrier of Ce³⁺ ionization from the 5d₂ state to the conduction band decreases with an increase of Gd³⁺ concentration.

Experimental

The phosphors, (Y_1−xGd_x)₃Al₅O₁₂: Ce³⁺ (x = 0, 0.1, 0.3, 0.5, 0.7, 0.9 and 1.0), were synthesized with a solid-state reaction at 1500 °C for 6 hours in a 25%H₂ + 75%N₂ reduction atmosphere. 2.5% BaF₂ and 2% H₃BO₃ in weight percent were used as fluxes. After reaction at high temperature, the fired products were ground and washed in water several times to remove the residual fluxes. Emission and excitation spectra were collected using a Hitachi F-4600 spectrometer. The thermal stability of the luminescence was examined using the spectrometer in combination with a heating apparatus, operated in an air environment. The phosphor was filled into a bronze sample cell. After heating to an objective temperature and keeping balance for 5 minutes, the emission and excitation spectra were recorded. The heating rate is less than 1 °C/ min and the precision of temperature control is less than 0.1 °C. The diffuse reflection spectra were collected using a UV-VIS-NIR spectrophotometer (Shimadzu, UV-3600). The thermally stimulated luminescence (TSL) was measured using a thermoluminescence dosimeter (Beijing Nuclear Instrument Factory, type FJ427A1). Before collecting data, the powder sample each with weight approximate 0.2 g was pressed into a diameter 10 mm pellet and then the pellet was excited with 365 nm ultraviolet for 30 min to ensure the storage energy of traps have been saturated. The heating rate of TSL measurements was kept at 10 °C/ min. The phases of the phosphors were analyzed with X-ray diffraction (XRD), using the Rigaku D/max-IIIA diffractometer with Cu Ka radiation, operated at 45 kV and 40 mA. The O K-edge and Gd L3-edge X-ray absorption spectroscopy (XAS) was measured on the beamline 20A and 01C1 and the valence band spectra were collected at the beamline 24A of National Synchrotron Radiation Research Centre (NSRRC) in Hsinchu, Taiwan. The electronic band structure (BS), density of states (DOS), partial density of states (PDOS) and charge deformation density (CDD) were calculated, based on density functional theory (DFT). Before calculation, the geometry of the crystal structure was optimized with the generalized gradient approximation (GGA).

Results and Discussion

Theoretical calculations on band structure and density of states

To gain insight into the intrinsic mechanisms of YAG: Ce luminescence upon substitution of Y³⁺ by Gd³⁺, the electronic band structure and density of states were firstly investigated by way of theoretical calculation. Four typical BS, DOS and PDOS of the pure YAG host and YAG doped with Ce³⁺ and a variant amount of Gd³⁺ are presented in Fig. 1, which shows that the band gap of YAG becomes narrow after doping with Ce³⁺ and Gd³⁺. The calculated band gap of the pure YAG host is 6.505 eV, which is consistent with the value of approximately 6.5 eV evaluated by the photoconductivity method²⁸ and the absorption peak at 188 nm determined using spectra method²⁹; and the band gaps of (Y_2.94Ce_0.06)Al₅O₁₂, (Y_0.75Gd_0.25)₃Al₅O₁₂ and (Y_0.25Gd_0.75)₃Al₅O₁₂ are approximately 6.358, 6.206 and 6.071 eV, respectively. The PDOS in Figs 1(b,d) shows that the top edge of the valence band mainly consists of the O 2p orbital and the bottom edge of the conduction band mainly consists of the 3d orbital of Y. However, Figs 1(f,h) indicate that the O 2p orbital makes a great contribution to form the conduction band and the d orbitals (including Y 4d and Gd 5d orbitals) contribute significantly to the valence band after doping with Gd³⁺. This conclusion suggests that the d orbital may hybridize with O 2p orbit intensively after doping with Gd³⁺, further confirmed as follows. The 4f orbital also contributes significantly to the valence band in Figs 1(f,h), but the 4f lies in a deep energy level. In contrast to Figs 1(a,e and g) show that the state density in the conduction band increases intensively as Y³⁺ is replaced with more and more Gd³⁺; moreover, the bandwidth of the valence band enlarges significantly with increasing Gd³⁺. The energy scales of electron spread in the conduction band are approximately 6.5–8.6, 6.2–10.0 and 6.0–12.42 eV, respectively, for x = 0, 0.25 and 0.75 of (Y_1−xGd_x)₃Al₅O₁₂. The expansion of the conduction bandwidth suggests the extensibility of the atomic orbitals, which consist of the band broadening, the reduction of effective mass of the electron and crystal field intensification. Meanwhile, the electrons will have a larger non-localization. The calculations on BS, DOS and PDOS come to the following conclusions: (1) crystal field strength increases, (2) band gap narrows, (3) the effective mass of the electron reduces and (4) the non-localization expands, as consequences of Gd³⁺ doping. These effects will inevitably affect photoluminescence.

Photoluminescence properties of (Y_1−xGd_x)₃Al₅O₁₂: Ce³⁺

Figure 2 presents the emission spectra of (Y_1−xGd_x)_2.94Al₅O₁₂: 0.06Ce³⁺ (x = 0, 0.1, 0.3, 0.5, 0.7 and 0.9) under the excitation of 460 nm at room temperature, which shows that the luminescence intensity decreases and the emission peak red shifts along with increasing Gd³⁺. As the x value increases from 0 to 0.9, the peak of the emission band shifts from approximately 543 to 575 nm, but the shift is negligible as x changes from 0.5 to 0.9, which can be discriminated more clearly by combining with the normalized emission spectra displayed in supplementary Fig.1. The intensity of luminescence decreases continuously with increasing Gd³⁺, but it decreases rapidly as x varies from 0.5 to 0.9. A quantitative assay shows that the luminescence intensity, achieved by integrating from 480 to 750 nm in emission spectra, decreases approximately 5.5% and 23.4% with x ranges from 0 to 0.5 and 0 to 0.9 (as the red line presented below in Fig. 3 indicates), respectively.

The excitation spectra of (Y_1−xGd_x)₃Al₅O₁₂: Ce, obtained by monitoring the strongest emission, are shown in Fig. 4. In addition to a minor neighbourhood maximum at 372 nm, two main excitation band peaks at approximately 460 and 340 nm are observed. The Ce³⁺ has [Xe]4f¹ electronic configuration, the excitation and emission of which originates from the 4f¹ to 5d¹ transition. Because of the shielding by the outer 5s and 5p orbitals, the 4f electron is insensitive to the crystal field. However, the 5d electron interacts strongly with the crystal lattice, resulting in strong phonon coupling and large crystal field effects on the excited states. In the crystal lattice of Y₃Al₅O₁₂, the Ce³⁺ which substitutes for Y³⁺ in the site of D₂ symmetry has 8 nearest neighbour oxygen atoms. The eightfold coordination of Ce³⁺ in the dodecahedral site can be described as a cubic coordination with an additional tetragonal distortion. According to the ligand field theory, the 5d state will split into a lower energy 5d_e-doublet state (²E_g) and a higher 5d_t-triplet state (²T_g) in tetragonal field²⁵. All five possible 5d-state energies of Ce³⁺ should be observed in the excitation spectrum, but all five levels were never unambiguously assigned^29,30. Tanner et al. confirmed the existence of the two excitation bands located at 342 and 467 nm using the room-temperature Xe-lamp excitation spectra and further revealed another band of Ce³⁺ at 225 nm as well as the YAG host band at 188 nm using the synchrotron radiation excitation spectra²⁹. Thus, the two excitation bands at 460 and 340 nm in Fig. 4 could be attributed to the 4f–5d₁ and 4f–5d₂ transitions, respectively.

The 4f-5d transition of Ce³⁺ emission with asymmetric broadband configuration in Fig. 2 consists of doublet sub-emissions from 5d₁ to ²F_2/7 and 5d₁ to ²F_5/2, because the ground state of Ce³⁺ consists of ²F_2/7 and ²F_5/2 sublevels after considering the spin-orbit interaction^19,29. The minor band with a peak at approximately 372 nm (26880 cm⁻¹) in Fig. 4 was reported in the literatures^{21,23,24,25,31,32,33,34,35,36,37,38,39,40} but was not apparent in other studies^{16,18,19,20,25,28,29,30}. Gracia calculated the absorption and luminescence spectra of Ce³⁺ doped YAG using an ab initio embedded cluster approach and concluded that a small peak at 372 nm is not due to Ce³⁺ ions⁴¹. Zeng attributed this band to an F-type colour centre³⁹. Through studying a YAG: Ce single crystal, grown with temperature gradient techniques (TGT), by means of thermal annealing in a H₂ and O₂ atmosphere combining with different doses of gamma irradiation, Dong further demonstrated that it was a type of F⁺-type colour centre³².

Both the intensity of 4f-5d₁ excitation at 460 nm and the excitation of 4f-5d₂ at 340 nm in Fig. 4 decrease with an increase of the x value, which is helpful to illustrate the luminescence decrease in Fig. 2. After normalizing the strong excitation to 1.0, however, the supplementary Fig. 2 shows that the relative excitation intensity of the band at 340 nm increases as x increases from 0 to 0.5 and then decreases as x increases onward to 0.9. This phenomenon puts forward an interesting topic regarding the relative distribution of electrons on different energy levels; this population usually obeys the Fermi-Dirac distribution function (seen below Equation (6)), depending on the energy barrier and temperature.

As x increases from 0 to 0.9, the band of 4f-5d₁ excitation shifts towards longer wavelengths whereas that of 4f-5d₂ shifts towards shorter wavelengths, as shown in Fig. 4. This phenomenon can be seen more clearly in supplementary Fig. 2. The shift in opposite directions of two excitation bands indicates the splitting of the Ce³⁺ 5d state enlarges with increasing Gd³⁺, caused by the intensified crystal field. This conclusion is consistent with the above prediction of the theoretical calculation. Thus, the diagram of Ce³⁺ energy levels and the crystal-field splitting of 5d orbitals in (Y_1−xGd_x)₃Al₅O₁₂ could be described with Fig. 5. With more and more Y³⁺ replaced by Gd³⁺, both the crystal-field splitting and Stokes shift increase with an increase of the x value, as seen from the quantitative values summarized in Table 1.

Table 1 The centroid of the 5d₁ and 5d₂ excited states, crystal field splitting energy and the Stokes shift within the ground state of 4f and the lowest excited state of 5d₁ for Ce³⁺ in (Y_1−xGd_x)₃Al₅O₁₂.

Additionally, Fig. 4 shows that the intensity of the band with a peak at 372 nm decreases step-by-step when with more and more Gd³⁺ is doped into the YAG, suggesting the doped Gd³⁺ is helpful in removing the F⁺ centres.

Diffuse reflection and absorption of (Y_1−xGd_x)₃Al₅O₁₂: Ce³⁺

The intensity of excitation reflects the comprehensive effects of energy absorption and energy transfer in the phosphor. To confirm that the decrease of (Y_1−xGd_x)₃Al₅O₁₂: Ce³⁺ luminescence upon Gd³⁺ doping was not caused by the reduced absorption, the diffuse reflection spectra were measured. As presented in Fig. 6, the two main absorption bands observed in the regions of 410–550 nm and 330–370 nm, respectively, correspond to the two excitation bands of 4f–5d₁ and 4f–5d₂, previously shown in Fig. 4 as well. The spectral shifts of the 4f–5d₁ and 4f-5d₂ absorption bands in two opposite directions were also observed. Differing from the excitation spectra displayed in Fig. 4, both the absorption intensity of 4f–5d₁ and the absorption of 4f–5d₂ increase as Gd³⁺ concentration increases from x = 0.1 to 0.9. In addition, Fig. 6 shows that the relative diffuse reflection intensity (including background) decreases as Gd³⁺ increases from x = 0.1 to 0.9 in the full wavelength range of 300 to 800 nm, except for the absorption of F⁺ centres. With the background intensity of diffuse reflection normalized to 1.0, as shown in supplementary Fig. 3, the conclusion that absorption intensity increases with Gd³⁺ holds true. Therefore, the decrease of (Y_1−xGd_x)₃Al₅O₁₂: Ce³⁺ luminescence upon doping with Gd³⁺ is not caused by the reduced absorption.

It must be noted that the excitation intensity of 4f-5d₁ is far stronger than that of 4f-5d₂ in Fig. 4; however, the relative absorption intensity of 4f-5d₂ is much stronger than that of 4f-5d₁ in Fig. 6. The difference between the absorption area in Fig. 6 and the excitation area in Fig. 4 suggests most of the absorption energy of 4f-5d₂ is lost without emission. This phenomenon indicates that the transition of 4f-5d₂ has a strong ability to absorb incident photons, but only a fraction of them convert to visible light. The energy loss of the 5d₂ state may be caused by the transition from 5d₂ to the conduction band, as discussed below.

Figure 7(a) plots the normalized emission and excitation spectra of (Y_1−xGd_x)_2.94Al₅O₁₂: 0.06Ce³⁺. Partial enlarged detail showing the crossover of the emission and excitation spectra with different amounts of Gd³⁺ is presented in Fig. 7(b), which shows that the crossover shifts towards longer wavelengths as x value increases from 0 to 0.9. According to the coordination-configuration theory, the energy loss of the non-radiative transition from the excited state to the ground state is proportional to the integral of the emission and excitation spectra, i.e.,

where the f_ex(x) and f_em(x) denote the wave functions of the emission and excitation spectra, respectively. However, the overlapping area below the curves of the emission and excitation spectra for each sample decreases with an increase of the x value. Therefore, the nonradiative transition of electrons relaxed from the 5d₁ excited state through the crossover of the potential parabola curves to the 4f ground state is not the main mechanism of energy loss.

Thermal stability luminescence of (Y_1−xGd_x)₃Al₅O₁₂: Ce³⁺

Chromaticity shift and luminescence quenching upon temperature changes are unfavourable for a phosphor applied in white LEDs, thus the thermal stability of luminescence should be carefully examined. The emission and excitation spectra measured at various temperatures of one typical sample with the maximum x = 0.9 for (Y_1−xGd_x)₃Al₅O₁₂: Ce³⁺ is provided in Fig. 8, which shows that excitation and emission intensity decrease rapidly with an increase of temperature. The relative luminescence intensities, achieved by integrating from 480 to 750 nm and with the intensity of each sample luminescence at room temperature normalized to 100%, of samples with x = 0, 0.3, 0.5. 0.7 and 0.9 for (Y_1−xGd_x)₃Al₅O₁₂: Ce³⁺ as function of temperature are presented in supplementary Fig. 4. This shows that the relative intensity decreases more and more along with increasing Gd³⁺. As the temperature increases from room temperature to 125 °C, the relative intensity of the YAG: Ce decreases approximately 4%, but (Y_0.1Gd_0.9)_2.94Al₅O₁₂: 0.06Ce³⁺ luminescence decreases approximately 64%. According to configuration coordinate theory, a heavy weight should benefit in resisting thermal vibration and reducing phonons and accordingly the thermal stability of YAG: Ce luminescence should enhance with Gd³⁺ doping. An evident spectral shift of YAG: Ce emission upon a change in temperature has been observed with and without a small amount of Gd³⁺ doping, as seen from Fig. 5 in reference¹⁹. However, with a large amount of Gd³⁺ doping, as in this work, no spectral shift upon temperature change is observed in the emission spectra of (Y_0.1Gd_0.9)_2.94Al₅O₁₂: 0.06Ce³⁺, as presented in Fig. 8 and the spectra with normalized intensity are displayed in supplementary Fig. 5.

Figure 8

Emission and excitation spectra of (Y_0.1Gd_0.9)_2.94Al₅O₁₂: 0.06Ce³⁺ at various temperatures.

In configuration coordinate theory, the spectrum either red shifts (due to coupling with phonons) or blue shifts (when transiting from a low energy level to a high one by coupling with phonons to give light). Here, no spectral shift but a serious decrease of luminescence occurs, which should be caused by the ionization of electrons from excited states to the conduction band (as discussed below). Moreover, the non-shift of emission suggests that doping Gd³⁺ into YAG: Ce makes the structure more rigid than it was previously. Both the luminescence decreases and the spectral non-shift behaviours upon temperature change are out of the range that the configuration coordinate diagram model can explain. The red shift of (Y_1−xGd_x)₃Al₅O₁₂: Ce³⁺ emission as function of Gd³⁺ concentration at room temperature could be interpreted according to the intensified crystal field, but what mechanism has resulted in the increase of crystal field strength deserves more study. To explore these mysteries, the electronic and crystal structures were examined.

Electronic structure and the mechanism of spectra shift

It is well known that the XANES spectra reflect the unoccupied density of states restricted by the dipole selection. Thus, XANES spectra could give us information on the geometric and electronic structure of the chemical bonding situation and the effective charge density around the X-ray-absorbing atoms. Because the low concentration of Ce³⁺ was out of the sensitivity of the instrument, the XANES spectra monitoring the Gd³⁺ L₃-edge were preferentially measured. The sharp lines presented in Fig. 9(a) were caused by the electron transition from Gd 2p_3/2 to outer unoccupied 5d orbitals, whose absorption intensity decreases with an increase of x value from 0.3 to 0.7 for (Y_1−xGd_x)₃Al₅O₁₂: Ce³⁺. This point is consistent with the decrease of the main peak of the EXAFS spectra in the Fourier transformation R space in Fig. 9(b). The absorption decrease in intensity along with increasing Gd³⁺ indicates that the outer unoccupied 5d orbital is filled with more and more electrons. Figure 9(c) presents the O K-edge XANES spectra of (Y_1−xGd_x)₃Al₅O₁₂: Ce³⁺ (x = 0, 0.3, 0.5 and 0.7), in which the sharp peak of the absorption at 532 eV is attributed to the excitation of O 1s electrons to O 2p states strongly hybridized with Y 3d or Gd 4f states. The sharp peak of the absorption at 532 eV decreases continuously with an increase of Gd³⁺ concentration from x = 0 to x = 0.7, suggesting that the outer unoccupied 2p orbital of O was also filled with electrons. The presence of sharp line absorption peaks of XANES, called the Rydberg states, indicates the atomic orbitals keep well. If the molecular orbital were formed through hybridization, the Rydberg state would not be sharp and the broadband configuration would be present. A possible reason for the unoccupied Gd 5d and O 2p simultaneously filled with electrons is through orbital hybridization to a form molecular orbital.

Figure 9

(a) The Gd L₃-edge XANES, (b) the Fourier transformation EXAFS spectra in R space, (c) the O K-edge XANES spectra and (d) the valence band spectra of (Y_1−xGd_x)₃Al₅O₁₂: Ce³⁺.

Because Gd³⁺ has smaller electronegativity than Y³⁺, Gd³⁺ should have a poorer ability than Y³⁺ to hybridize 5d with O 2p in theory, but it does not practically. Ram’s ^[15] conclusion about the strong rigidity of the YAG structure with small atomic displacement parameters provides us with a useful clue to understanding this phenomenon. Because of the larger radius, Gd³⁺ ions will suffer a compressive resistance from neighbouring atoms when incorporated into the rigid crystal lattice of YAG. In the YAG crystal lattice, Gd³⁺ takes the place of the Y³⁺ site with high symmetry for its eightfold coordination. Moreover, the Gd 5d orbital could spread in a large space in contrast to the Y 4d orbital due to its larger radius. Under the comprehensive effect of above factors, the strong compression effect in a high symmetric site compels the Gd 5d orbital to hybridize with the O 2p orbital.

To support this viewpoint, the existence of strong compression was demonstrated firstly. As for a cubic garnet structure, the spacing d between adjacent (hkl) lattice planes obeys the relationship:

The XRD patterns of (Y_1−xGd_x)₃Al₅O₁₂: Ce³⁺ (x = 0, 0.1, 0.3, 0.5, 0.7, 0.9 and 1.0) are presented in supplementary Fig. 6, in which all diffraction peaks can be indexed to the standard Y₃Al₅O₁₂ (JCPDS: 33-0040) when the value of x is no higher than 0.3. However, the minor phase of GdAlO₃ (JCPDS: 46-0395) is indexed as x = 0.3 to 0.9. The 2θ value of the strongest diffraction peak at the crystal planes (420) decreases with increasing Gd³⁺ from x = 0 to 0.9, as shown on the right side in supplementary Fig. 6, with an amplified range of 33–34°, indicating an increase of the crystal lattice parameter. A linear expansion of the lattice constant as function of Gd³⁺ concentration is found within x ≤ 0.3 in Fig. 3, but it increases nonlinearly when x > 0.3. Such a nonlinear curve is below the line determined with the same slope as x ≤ 0.3, indicating the measured crystal lattice constants of samples doped with Gd³⁺ are lower than the theoretical value predicted with Vegard’s law.

However, the EXAFS data could provide information about the inter-atomic distances and the coordination. The eightfold coordination has been confirmed. Figure 9(b) shows that the first neighbour shell distances in the R space of Fourier transformation EXAFS spectra of (Y_1−xGd_x)₃Al₅O₁₂: Ce³⁺ are approximately 1.8715, 1.9021 and 1.8715 Å for x = 0.3, 0.5 and 0.7, respectively. The deviations of the R values are within experimental error. Therefore, the measured data reveal that there is virtually no increase in bond length with Gd³⁺ doping in the concentration range from x = 0.3 to 0.7. However, the radius of Gd³⁺ is larger than Y³⁺. The result indicates that the crystal lattice after Gd³⁺ doping is smaller than the ideal volume. The shrink of the crystal lattice volume suggests that the doped Gd³⁺ ions receive an intensive squeezing effect from other atoms. When more and more Gd³⁺ is introduced into the lattice, the compressive stress will increase exponentially. Accordingly, the emission wavelength increases nonlinearly, as shown in Fig. 3, as a consequence of the intensified crystal field. The two curves, i.e., luminescence intensity and emission wavelength as function of Gd³⁺ concentrations, are nearly axisymmetric along the vertical axis. The profile of these two curves looks like a butterfly, which can be called the butterfly effect for the compression simultaneously activated on luminescence intensity and emission wavelength.

Finally, a theoretical calculation on CDD provides visual evidence regarding the change of charge distribution after doping Ce³⁺ and Gd³⁺ into YAG. When there is no Ce³⁺ or Gd³⁺ doped in the pure YAG host, as presented in Fig. 10(a), the charge is restricted to a certain space. The blue colour around the Al atoms indicate that they lose electrons and the red colour around the O atoms indicate that they gain electrons. With respect to Fig. 10(a), the intensified red colour around the O atoms in Figs 10(b,c) shows that the charge density around oxygen increases, either doping Ce³⁺ or doping Gd³⁺ into the YAG. Especially for Gd³⁺ doping, the charge density around oxygen increases significantly. Moreover, the dumbbell shape of the p orbitals around O could be discriminated in Figs 10(a,b), as the arrows indicated; however, in Fig. 10(c) they combine into one ellipsoid under the compression effect, which provides a visual picture of the deformation of the electron cloud. Not only does the charge density around oxygen intensify but also the volume of electron cloud expands after doping with Gd³⁺ and Ce³⁺, due to the hybrid orbital. As far as the charge density around Y³⁺ is concerned, the comparison of Figs 10(a,c) shows that the charge density around Y³⁺ increases a little after Gd³⁺ doping. These characteristics indicate that the orbital hybridization between Y and O intensifies with Gd³⁺ doping, consistent with the result displayed in Figs 8(a,b). The electronic structure changes in turn, which can explain the BS and DOS picture presented in Fig. 1. At least, the variation of the valence band structure is observed in this experiment, as shown in Fig. 9(d).

Figure 10

Charge deformation density of Y₃Al₅O₁₂

(a), (Y_0.98Ce_0.02)₃Al₅O₁₂ (b) and (Y_0.75Gd_0.25)₃Al₅O₁₂ (c) calculated based on DFT.

Compared with the space of electron spread in Fig. 10(a), an obvious expansion of the electron cloud around oxygen is observed in Fig. 10(b), which makes the effective Ce-O bond length shorten. According to the relationship between crystal field strength and ionic bond length:

where the parameter D_q presents the crystal field stabilization energy (CFSE), R is the bond length between a central ion and ligand ions, r is the mean size of the central ion and Z is the charge of a central ion. Thus, the shorter bond length implies the stronger crystal field strength. Accordingly, the 5d state will have a larger split after doping with Gd³⁺, which can explain the results in Table 1.

The change of the electron cloud can explain the variation of band structure and the density of states of YAG upon doping with Ce³⁺ and Gd³⁺ as well. Because of the orbital hybridization, electrons can spread in a large space. With the expansion of the electron cloud around O atoms, the band gap of (Y_1−xGd_x)₃Al₅O₁₂ becomes narrow and the density of states becomes dense. The intensified crystal field will reduce the effective mass of the electrons and make them move easily, which will decrease the thermal stability of luminescence. In addition, the expansion of the electron cloud around oxygen makes it easy for electrons to be promoted to excited states. The electronic and crystal structures obtained here provide insight into the photoluminescence properties.

Crystal defects and trap depths

However, the performance of a phosphor, including luminescence efficiency and thermal stability, is closed related to crystal defects. The thermoluminescence which was considered the best tool in identifying crystal defects was used to characterize the YAG: Ce phosphor. Figure 11 shows the TSL of (Y_1−xGd_x)_2.94Al₅O₁₂: 0.06Ce³⁺, in which two emission bands are observed: a weak band in the range of 50–150 °C and a strong one in 125–350 °C. The peak of the weak band shifts from approximately 118 °C to 85 °C as x value increases from 0 to 0.3, accompanied by a decrease in intensity; and then the peak almost disappears as x value increases onwards to 0.5 and 0.7. However, the peak of the strong band shifts continuously with an increase of Gd³⁺ concentration, from approximately 260 °C at x = 0, via 219 °C at x = 0.3 and 197 °C at x = 0.5 and finally reaches 171 °C at x = 0.7, in addition to a continuous decrease of luminescence intensity. The thermally stimulated luminescence shows that the following processes happened: when electrons are excited to the high-energy state, some of them are captured by the defects and then they are thermally released from the defects to the excited state, finally giving light during the transition from a high-level excited state to the ground state, as the blue line depicted in Fig. 12. Previously reports on the persistent luminescence of YAG: Ce also confirms the existence of crystal defects^42,43. The decrease in the intensity of TSL suggests the decrease of the concentration of crystal defects; and the shift of TLS peak from high to low temperature indicates the decrease of the trap depth. Two TSL peaks indicate the existence of two types of defects in the phosphor. The order of kinetics of the glow curves was determined with the peak shape method, by calculating the symmetry (geometrical) factor:

Figure 11

Thermally stimulated luminescence of (Y_1−xGd_x)_2.94Al₅O₁₂: 0.06Ce³⁺.

Figure 12

The relative position of 5d₁ and 5d₂ of Ce³⁺ and the relative depth of trap 1 and trap 2 in the band structure of (Y_1−xGd_x)₃Al₅O₁₂: Ce³⁺.

where ω = Τ₂−Τ₁ denotes the full width of the glow peak at half its maximum height, τ = T_m−T₁ presents the low-temperature half width and δ = T₂ − Τ_m is the high-temperature half width. The values of the geometrical factor suggest that the peaks obey general order kinetics. Thus, the trap depth could be calculated using Chen’s equation⁴⁴:

where k is the Boltzmann constant and T_m is the peak temperature. The calculated data are shown in Table 2. YAG is a high-dielectric material, whose permittivity is approximately 11.7. The high permittivity indicates that a strong polarity (or dipoles) exists in the local circumstance of YAG, which will cause (Y,Gd)₃Al₅O₁₂: Ce³⁺ to have a strong surface adsorption, such as the OH stretching bands observed in the infrared absorption spectra of YAG^{32,33,34,35,36,37}. The surface adsorption will change the surface energy levels and finally affect the luminescence. Thus, we are inclined to attribute the emission in the range of 50–150 °C to the defect of surface adsorption and the other in 125–350 °C to the F⁺ centres caused by O_h^- oxygen vacancies^32,33,34,35.

Table 2 The temperature of maxima thermoluminescence, geometrical factor and trap depth in (Y _1−xGd_x)_2.94Al₅O₁₂: 0.06Ce³⁺.

Mechanisms of luminescence decrease and thermal quenching of YAG: Ce upon doping with Gd³⁺

The population of electrons in different energy levels obeys the Fermi-Dirac distribution:

where T is the absolute temperature, k is Boltzmann’s constant, E_i is the energy of the single-particle state i and μ is the total chemical potential. With more and more Gd³⁺ doped into the YAG, one side the band gap of (Y_1−xGd_x)₃Al₅O₁₂: Ce³⁺ decreases; on the other side, the energy splitting between 5d₁ and 5d₂ expands, which makes the 5d₂ energy level of Ce³⁺ approach the bottom of the conduction band. When the electrons are excited from 4f to 5d₁ state to produce blue light for white LED applications, assuredly certain amount of electrons will be excited to the 5d₂ state according to the Fermi-Dirac distribution. Moreover, the population of electrons in the 5d₂ state will increase with increasing Gd³⁺, due to the reduced energy barrier between 5d₂ and the conduction band. This mechanism can explain the phenomenon of the 4f-5d₂ absorption of Ce³⁺ enhanced evidently in Fig. 6, but the excitation in Fig. 4 does not rise correspondingly upon doping with Gd³⁺. Under the effect of temperature, the electrons are easily ionized from 5d₂ to the conduction band to produce photocurrent. Moreover, the electrons are possibly captured by crystal defects.

The increase of Gd³⁺ concentration reduces the trap depth of crystal defects, which results in the weakly trapped electrons being easily released to the conduction band. Once the excited electrons are captured by crystal defects, they possibly delocalize to the conduction band with the help of temperature. The mechanism of thermal delocalization of electrons from crystal defects to the conduction band can explain the serious thermal luminescence quenching that occurs in Fig. 8, although a spectral shift with increasing temperature is not observed. Therefore, the auto-ionization of electrons from 5d₂ to the conduction band and the thermal delocalization of electrons from crystal defects to the conduction band are the main mechanisms of energy loss involved in (Y_1−xGd_x)₃Al₅O₁₂: Ce³⁺ luminescence. Although the band gaps decrease with increasing Gd³⁺, Fig. 9(d) shows that the position of the valence band relative to Fermi level does not significantly change. Thus, the processes of energy loss and luminescence could be described in Fig. 12.

Conclusion

In summary, a new interpretation from the viewpoint of compression deformation of the electron cloud in a rigid structure by combining hybrid orbital theory with solid-state energy band theory together has been put forward in this work to illustrate the intrinsic mechanisms that cause the emission spectral shift, thermal quenching and decrease in intensity of YAG: Ce luminescence upon substitution of Y³⁺ by Gd³⁺. Due to Gd³⁺ having a larger radius in comparison with Y³⁺, Gd³⁺ is subject to a strongly compressive effect when it takes the place of the eightfold coordinated Y site in the rigid YAG crystal lattice, which compels the Gd 5d and Y 4d orbitals to hybridize with the O 2p orbital to form a molecular orbital. Because of the orbital hybridization, electrons can spread over a large scale of energy levels and the band gap becomes narrow as increasingly more Y³⁺ is replaced by Gd³⁺. The intensified crystal field reduces the effective mass of the electrons and the sub-state of Ce³⁺ 5d orbital splitting expands. Two types of crystal defects, a shallow trap caused by surface adsorption and another deep F⁺ centre, were identified; and the trap depths decrease with increasing Gd³⁺. In addition to the narrow band gap, the energy barriers for electron transition from the 5d₂ sublevel and crystal traps to the conduction band reduce with an increase in Gd³⁺ concentration. Accordingly, the electrons easily auto-ionize or de-localize into conduction band. The auto-ionization of electron from 5d₂ to the conduction band and thermal delocalization of electrons from crystal traps to the conduction band are the main mechanisms of luminescence decrease and thermal quenching of (Y,Gd)AG: Ce upon doping with Gd³⁺. The theoretical calculation of the charge deformation density provides a visualized picture of electron cloud expansion, especially for the expansion of the electron cloud around O atoms which makes the effective Ce-O bond length shorten, further resulting in an increase of crystal field strength and spectral shift. This work provides a solid foundation to tune emission colour in rigid structure, such as in diamond, corundum and high-strength nitrides, using charge deformation. However, the band gaps and crystal defects must be strictly controlled in band gap engineering, in comprehensive consideration of the luminescence thermal stability and efficiency.