Journal of Luminescence 179 (2016) 297–305

Contents lists available at ScienceDirect

Journal of Luminescence

http://d0022-23

n Corrnn Cor

versityFax: þ1

E-mjmckittr

1 Th

journal homepage: www.elsevier.com/locate/jlumin

Full Length Article

An integrated first principles and experimental investigationof the relationship between structural rigidity and quantum efficiencyin phosphors for solid state lighting

Jungmin Ha a,1, Zhenbin Wang b,1, Ekaterina Novitskaya c, Gustavo A. Hirata d,Olivia A. Graeve a,c, Shyue Ping Ong b,n, Joanna McKittrick a,c,nn

a Materials Science and Engineering Program, University of California, San Diego, 9500 Gilman Dr., La Jolla, CA 92093, USAb Department of Nanoengineering, University of California, San Diego, 9500 Gilman Dr., La Jolla, CA 92093, USAc Department of Mechanical and Aerospace Engineering, University of California, San Diego, 9500 Gilman Dr., La Jolla, CA 92093, USAd Center for Nanoscience and Nanotechnology, Ensenada, Mexico

a r t i c l e i n f o

Article history:Received 18 March 2016Received in revised form30 June 2016Accepted 2 July 2016Available online 7 July 2016

Keywords:First-principles calculationDensity functional theoryNear-UV phosphorsEu2þ activationDebye temperatureQuantum efficiency

x.doi.org/10.1016/j.jlumin.2016.07.00613/& 2016 Elsevier B.V. All rights reserved.

esponding author.responding author at: Materials Science anof California, San Diego, 9500 Gilman Dr: 858-534-5698.ail addresses: ongsp@eng.ucsd.edu (S.P. Ong),ick@ucsd.edu (J. McKittrick).ese authors contributed equally to this work

a b s t r a c t

We outline an integrated approach for exploring novel near-UV excited phosphors. To test the hypothesisof whether high host structural rigidity results in phosphors with high quantum efficiency (Φ), wecalculated the Debye temperatures (Θ) for 27 host materials using density functional theory calculations.We identified Eu2þ-activated Ca7Mg(SiO4)4 and CaMg(SiO3)2 as having a relatively high Θ¼601 K and665 K, respectively, and predicted excitation energies of 3.18 eV (337 nm) and 3.29 eV (377 nm),respectively, both of which are in good agreement with the results of photoluminescence spectroscopy.However, the measured Φ for these two phosphors was o 30%, which indicates that Θ alone is not asufficient condition for a high Φ. This work demonstrates the potential of combined first-principlescalculations and experiments in the discovery and design of novel near-UV excited phosphors.

& 2016 Elsevier B.V. All rights reserved.

1. Introduction

White-light sources based on light-emitting diodes (w-LEDs)have attracted intense interest for next generation solid-statelighting technologies due to their longer lifetime, superior effi-ciency, and low operating temperatures compared with traditionalincandescent bulb and fluorescent lamp technologies [1–3]. Thecommon approach to create white-emitting LEDs is to combine ablue-emitting (450 nm) InGaN LED with a yellow-emitting phos-phor [4,5]. The drawbacks of this method are a non-uniform lightoutput and a low color rendering index (CRI) value. An alternative,more attractive approach is to combine a near UV-LED (nUV-LED,370–410 nm) with blue, green, and red phosphors to produce high-efficiency w-LEDs. A high CRI value can be achieved by optimizationof phosphor blends and, ultimately, a higher extraction efficiency

d Engineering Program Uni-., La Jolla, CA 92093, USA.

.

can be obtained due to a smaller current drop at high driving cur-rents, compared to the blue-emitting LEDs [6–8].

Traditional trial-and-error approaches in the discovery of newblue-, green- and red- emitting phosphors for nUV-LED phosphorsare time consuming and costly. Recently, new methodologies, suchas mineral-inspired prototype evolution [9,10] and single-particle-diagnosis [11] have been proposed for the discovery of newphosphors. Although these methods have achieved significantprogress for developing new phosphors, in essence, they still focuson the exploration case-by-case.

An alternative approach is to use easily computable or mea-surable descriptors as a means to screen for various desiredproperties for phosphors, such as an appropriate excitationwavelength (370–410 nm), a high quantum efficiency (4 80%),good thermal quenching behavior (o 20% decrease in emissionintensity from room temperature to 150 °C), and chemical stability(e.g. moisture stability). For example, the narrow-band descriptorcombined with band gap and phase stability was recently pro-posed to discover promising narrow-band red-emitting phosphorsfor high-power LEDs in a high throughput approach [12]. Based onthe analysis of over 1000 rare-earth activated compositions, Dor-enbos [13,14] proposed that the excitation energy, Ea(Re, A) can be

Fig. 1. Schematic diagram of 4f-5d transition in Ce3þ or Eu2þ activated phos-phors. Eg, Eex and E4f-VBM denote the band gap of host material, excitation energy,and 4f–VBM energy gap, respectively.

J. Ha et al. / Journal of Luminescence 179 (2016) 297–305298

predicted from the following empirical relationship:

Ea Re;Að Þ ¼ EF Reð Þ�D Að Þ ð1Þwhere EF(Re) and D(A) are the rare earth (Re) free ion energy andcrystal field depression in host A, respectively. However, D(A) isdifficult to calculate. It is usually tabulated and strongly related tothe crystal structure of the host, therefore it is not easy to use Eq.(1) for new host materials without knowledge of D(A).

More recently, the ab initio calculated Debye temperature (Θ)of host materials was proposed to have a positive relationship withthe experimental photoluminescence (PL) quantum efficiency (Φ).For Ce3þ-activated phosphors, a ΘZ500 K was suggested forΦZ80% [15]. For example, the calculated Θ for Y3Al5O12:Ce3þ

(YAG:Ce3þ) is 726 K with Φ measured to be 90% [15]. These pre-liminary results suggest that host materials with a high Θ mayyield high Φ with an appropriate host band gap. However, therehave been relatively few efforts utilizing a combined first-principles and experimental approach for the discovery of newphosphors.

In this work, 27 host materials were selected based on thecriterion that Ce3þ- or Eu2þ-doping results in an absorptionwavelength in the near-UV light range (370–410 nm). These hostmaterials were found from potential phosphor compositions in theliterature [13,14,16–18]. The Θ of these 27 hosts were calculatedusing density functional theory (DFT) calculations. Two potentialcandidates, Ca7Mg(SiO4)4:Eu2þ and CaMg(SiO3)2:Eu2þ havinghigher Θ were identified, and their excitation energies were cal-culated. Then, powders were synthesized. The (Ca0.96Eu0.04)7Mg(SiO4)4 phosphor was prepared by the sol–gel/Pechini method andthe Ca0.94Eu0.06Mg(SiO3)2 phosphor was prepared by the co-precipitation method. Subsequently, the PL properties were ana-lyzed to compare to the first-principles predictions.

2. Methods

2.1. Computational methods

All first-principles calculations were performed using the pro-jector augmented wave (PAW) method as implemented in theVienna ab initio simulation package (VASP) [19,20]. The Perdew–

Berke–Ernzerhoff (PBE) [21] exchange-correlation functional wasused. More specific calculation parameters can be found in refer-ences [22,23].

The Debye temperature, Θ, was calculated using the quasi-harmonic model given by [15]

Θ¼ ħkB

6π2V12n

� �f ðνÞ

ffiffiffiffiffiBM

rð2Þ

where V, n, f ðvÞ, B and M are the unit cell volume, the number ofatoms in the unit cell, a scaling function in terms of Poisson's ratioν, the bulk modulus and the molar mass, respectively; ħ and kBrefer to the Plank constant and the Boltzmann constant, respec-tively. The elastic moduli were calculated by employing the Voigt–Reuss–Hill (VRH) approximation based on the computed elastictensor [24]. Previous research indicated that an effective HubbardU of 2.5 eV can reproduce the experimental E(VBM-4f) for Eu2þ-activated oxides, [25] and the band width of the Eu2þ 4f bands arenot very sensitive to the U value used [12]. We here adoptedU¼2.5 eV for the evaluation of orbital-projected density of states(DOS) with the PBE functional. The band structures of hosts werecalculated using the modified Becke–Johnson localized-densityfunctional (MBJLDA), which has been shown to yield more accu-rate band gap values for insulators compared to the standardgeneralized gradient approximation (GGA), but at a much lowercomputational cost compared to hybrid density functionals [26].

According to experimental findings on the absolute location oflanthanide (Ce3þ and Eu2þ) energy levels, the 4f and 5d energylevels generally lie in the band gap, with the 5d level closer to theconduction band minimum [27,28]. This PL mechanism(4fn�15d1-4fn transition) for Ce3þ and Eu2þ activated phosphorsis described in Fig. 1 [27,28]. The excitation energy (Eex) is pro-posed from the following equation:

Eex � Eg host;band gapð Þ�E VBM�4fð Þ: ð3Þwhere Eg (host, band gap) is the band gap of the host material andE(VBM–4f) is the energy gap between the valence band maximum(VBM) and 4f level. It is noted that there are some errors in theestimate of the excitation energy on the order of �1 eV due to theneglect of the gap between the ground state Eu 5d level and theCBM. Dorenbos [29] found that the excited Eu 5d level is generallylocated within about 1.0 eV below the conduction band minimum(CBM). Also, the Stokes shift usually lies in the range of 0.64 eV �1.22 eV. Nevertheless, the relative trend is reasonable. The Fermienergy in Eu2þ-activated systems refers to the highest occupied 4forbital and is set to zero in the DOS plot.

2.2. Synthesis of phosphors

2.2.1. ReagentsAll chemicals were used without further purification and

included tetraethyl orthosilicate (TEOS, 99.9%, Sigma Aldrich),nitric acid (69.3%, Fisher Scientific), Eu2O3 (99.99%, Alfa Aesar), Mg(NO3)2.6H2O (98.3%, Fisher Scientific), Ca(NO3)2.4H2O (99.0%,Macron Fine Chemicals), citric acid (C6H8O7.H2O, Macron FineChemicals), ethylene glycol (C2H6OH, Fisher Scientific), poly-ethylene glycol (PEG, C2H4O.nH2O, molecular weight¼20,000 g/mole, Sigma Aldrich), and ammonium hydroxide (28�30%, BDHAristar Plus).

2.2.2. Preparation of (Ca0.96Eu0.04)7Mg(SiO4)4The sol–gel/Pechini process was used to synthesize

(Ca0.96Eu0.04)7Mg(SiO4)4 powders. The 4 at.% Eu2þ concentrationwas chosen because it has been reported to have the highest PLemission intensity [16]. First, tetraethyl orthosilicate (2.23 mL) wasadded to ethanol (20 mL) with several drops of nitric acid andstirred for 30 min. Next, Eu2O3 was added in a dilute nitric acidsolution to form a solution of aqueous Eu(NO3)3. Mg(NO3)2.6H2O,and Ca(NO3)2.4H2O were dissolved in deionized water. After thesesolutions became transparent, silica sol and the Eu(NO3)3 solutionwere poured to the Mg(NO3)2 and Ca(NO3)2 solution with stirring.Next, citric acid (4.2 g) and ethylene glycol (2.23 mL) were addedto the mixed solution (metal:citric acid:ethylene glycol¼1:1:2),which acts as a chelating agent for the metal ions. Polyethyleneglycol (2.5 g) was introduced in the mixture, which was used as acrosslinking agent. After stirring the solution for 30 minutes to

Table 1Calculated Debye temperature (Θ) of 27 identified host for Ce3þ or Eu2þ activation.The compositions in bold have Θ4500 K. The excitation energy (Eex) and emissionenergy (Eem) were selected from [13,14,16–18,37,44].

Host material Eex (nm) Eem (nm) Θ (K)

Mg2SiO4:Ce3þ 373 432 734X1-YSiO2N:Ce3þ – – 695CaMg(SiO3)2:Eu2þ 365 450 665Mg3F3BO3:Ce3þ 399 474 615X2-YSiO2N:Ce3þ 370 405 606Ca7Mg(SiO4)4:Eu2þ 350 505-520 601X2-Y2SiO5:Ce3þ 381 480 512X1-Y2SiO5:Ce3þ 365 430 491SrY2O4:Ce3þ 397 560 465La2Be2O5:Ce3þ 365 445 464β-Sr2SiO4 310 �543 428LuBO3 (vaterite):Ce3þ 365 388 423Ba3Mg(SiO4)2:Eu2þ 410 438 422X2-Lu2SiO5:Ce3þ 376 462 416X1-Sr2SiO4:Eu2þ 390 490 410NaF:Ce3þ 390 472 387CaAl2S4:Ce3þ 396 440 373α-Sr2SiO4 310 573 360Sr2SiCl2O3:Eu2þ 410 490 349SrAl2S4:Ce3þ 397 462 338BaAl2S4:Ce3þ 384 444 331ThO2:Ce3þ 408 – 319Ba2SiO4 360 512 307Ba2YB2O62Cl:Ce3þ 370 466 300Ba2LuB2O6Cl:Ce3þ 370 466 294La3Si2S8I:Ce3þ 370 446 262LaOI:Ce3þ 385 435 161

J. Ha et al. / Journal of Luminescence 179 (2016) 297–305 299

ensure uniformity, precipitation occurred by adding drops ofammonium hydroxide until the pH reached 4. After completingthe preparation, the mixture was continuously stirred at 80 °Covernight to form a gel. The white colored product was preheatedto remove organic materials at 350 °C for 1 hour, and thenannealed at 1100 °C or 1350 °C for 3 h, or 1400 °C for 10 h in air. Allsamples were finally calcined at 1100 °C for 4 h under a slightreducing condition (mixture of 5% H2 and 95% N2).

2.2.3. Preparation of Ca1-xEuxMg(SiO3)2Ca1�xEuxMgSi2O6 powders were synthesized by a co-

precipitation technique that was modified from a previous repor-ted method [30]. TEOS (2.23 mL) was added into ethanol (20 mL)with HNO3 (0.5 mL) and stirred for 30 min to hydrolyze TEOS. Thedesired amount of Eu2O3 (x ¼ 0.02, 0.06, 0.1, and 0.2) was intro-duced in a dilute nitric acid solution to form a solution of aqueousEu(NO3)3. Mg(NO3)2.6H2O and Ca(NO3)2.4H2O were dissolved in30 mL of deionized water. After the europium solution and themagnesium and calcium solution became transparent, the silicasol and europium solution were poured into the magnesium andcalcium solution and the mixture was stirred for 1 h. Ammoniumhydroxide was added dropwise into the solution until the pH was10 to create a white precipitate. The precipitate solution was stir-red for 8 h at room temperature. Next, the precipitate was cen-trifuged and washed with deionized water three times. The pro-ducts obtained by centrifugation were dried at 100°C for 12 h.Then a post-annealing step was performed at 1100 °C for 2 h underair and then at 1100 °C for 4 h under 5% H2 / 95% N2 to changeEu3þ to Eu2þ .

2.2.4. CharacterizationThe crystallite phases and crystallite sizes of the annealed

powders were determined by X-ray diffraction (XRD, Bruker D2Phaser) using CuKα radiation and a step size of 0.014° over the 2θrange of 20–80 degrees. The analysis was performed using theDiffrac.Eva plus program. Lattice parameters and crystallite sizeswere calculated using the approach of Rietveld refinement by theTOPAS 4.2 software (Bruker), which averages the lattice constantsassociated with each XRD peak across the entire spectrum. Thesize and morphology of the particles were analyzed by a fieldemission scanning electron microscope (FESEM, XL30, Philips) at10 keV. Samples were coated with iridium at 85 mA for 10 s beforeFESEM imaging. Absolute quantum efficiency (Φ) measurementswere performed using an integrating sphere system, with sodiumsalicylate (Φ¼44%) as a reference standard. Photoluminescenceemission and excitation spectra were acquired with a fluorescencespectrophotometer (Hitachi F-7000) using λ¼ 350 nm excitationwavelength (pulse¼0.025 sec). This excitation wavelength waschosen as it produced the highest PL emission intensity. Thethermal quenching analysis (25°C � 150 °C) was performed usinga custom designed device that consist of a heater, thermocouple,and the spectrophotometer.

3. Results and discussion

3.1. Screening of rigid structures and predicting the excitation energy

High structural rigidity confines accessible phonon modes thatresult in non-radiative recombination. As mentioned previously,Θwas recently proposed as a descriptor of structural rigidity [15].The higher Θ, the fewer lattice vibration modes can be accessedand thus the more rigid is a crystal structure. The calculated Θ ofthe 27 host candidates are summarized in Table 1. Seven hostcandidates (in bold) have Θ4500 K, implying a strong rigidity ofthe crystal structure and correspondingly, a possible high Φ.

Among these seven candidates, only the three silicate hostmaterials Mg2SiO4, Ca7Mg(SiO4)4, and CaMg(SiO3)2 were con-sidered for further investigation because of their high chemicalstability, good thermal quenching properties [31], and facilesynthesis methodologies. Since, in Mg2SiO4 the radius of Mg2þ

(o0.07 nm) is too small to be replaced by Eu2þ (�0.130 nm),Ca7Mg(SiO4)4:Eu2þ and CaMg(SiO3)2:Eu2þ were selected for fur-ther investigation. In this case, Eu2þ can replace Ca2þ in thecrystal structure due to similar ionic radii (0.118 nm and 0.123 nmfor 9-coordinated and 10-coordinated Ca2þ , respectively;0.130 nm and 0.135 nm for 9-coordinated and 10-coordinatedEu2þ , respectively) [32].

Fig. 2a and b shows the calculated band structures of Ca7Mg(SiO4)4 and CaMg(SiO3)2. Both hosts have direct (Γ- Γ transition)band gaps (Eg¼6.86 eV for Ca7Mg(SiO4)4 and Eg¼7.08 eV for CaMg(SiO3)2). Fig. 3a and b shows the calculated orbital projected DOS.The energies between 4f (located at Fermi energy level of 0) andVBM for Ca7Mg(SiO4)4:Eu2þ and CaMg(SiO3)2:Eu2þ are 3.18 eVand 3.79 eV, respectively. The excitation energies can be estimatedfrom Eq. (3) by subtracting the calculated 4f�VBM gap (3.18 eV forCa7Mg(SiO4)4:Eu2þ and 3.79 eV for CaMg(SiO3)2:Eu2þ) from Eg,which gives Eex¼3.68 eV (337 nm) for Ca7Mg(SiO4)4:Eu2þ and3.29 eV (377 nm) for CaMg(SiO3)2:Eu2þ .

3.2. Crystal structure, X-ray diffraction, and scanning electronmicroscopy of the synthesized phosphors

Fig. 4a and b shows the structures of Ca7Mg(SiO4)4:Eu2þ andCaMg(SiO3)2:Eu2þ drawn by VESTA (Visualization for Electronicand STructural Analysis) [33]. In Fig. 4a, Ca7Mg(SiO4)4 (mineralbredigite) has an orthorhombic structure with a space group Pnn2where Ca2þ has three different crystallographic sites with coor-dination numbers 12, 10, and 9 [34]. As illustrated in Fig. 4b, CaMg(SiO3)2 (mineral diopside) has a monoclinic structure with spacegroup C2/c. The lattice parameters of (Ca0.96Eu0.04)7Mg(SiO4)4 andCa0.94Eu0.06Mg(SiO3)2 were calculated using TOPAS and the

Fig. 3. The orbital-projected density of state (DOS) of Ca7Mg(SiO4)4:Eu2þ (a), andCaMgSi2O6:Eu2þ (b), where the 4f�VMB energy gap is 3.18 eV and 3.79 eV,respectively. The Fermi level is set to 0.

Fig. 2. The band structure of the host material Ca7Mg(SiO4)4 (a) and CaMgSi2O6 (b),where it has a direct Γ - Γ transition band gap of 6.86 eV and 7.08 eV, respectively.

J. Ha et al. / Journal of Luminescence 179 (2016) 297–305300

computed lattice constants are a¼0.6742 nm, b¼1.0887 nm,c¼1.8339 nm for (Ca0.96Eu0.04)7Mg(SiO4)4 and a¼0.9743 nm,b¼0.8879 nm, c¼0.5230 nm for Ca0.94Eu0.06Mg(SiO3)2, which areconsistent with previous reports for Ca7Mg(SiO4)4 [34] and CaMg(SiO3)2 [35].

XRD patterns of the (Ca0.96Eu0.04)7Mg(SiO4)4 powders for dif-ferent calcination times and temperatures are shown in Fig. 5a.The diffraction peaks were indexed by the standard data from PDFcard 00-036-0399 (Ca7Mg(SiO4)4). A pattern for post-synthesisannealing at 1100 °C, shows impurity peaks around 2θ¼27, 29 and30°, which correspond to CaSiO5. However, these peaks disappearafter annealing over 1350 °C. No other diffraction peaks, other thanthose corresponding to (Ca0.96Eu0.04)7Mg(SiO4)4, are observed atboth 1350 °C and 1400 °C. The crystallite sizes were determined bythe TOPAS program and were �54 nm, �67 nm, and �70 nm, forpost-synthesis annealing at 1100 °C for 3 h, 1350 °C for 3 h, and1400 °C for 10 h, respectively. These results are in a good agree-ment with previous studies demonstrating that crystallites growwith an increase of the post-synthesis annealing temperature andtime [36,37].

Fig. 5b shows XRD patterns of the Ca1�xEuxMgSi2O6 (x¼0.02,0.06, 0.1, and 0.2) powders. It is observed that, when the activatorconcentrations are 0.02 and 0.06, the peaks in the XRD patternsare consistent with the standard PDF card 01-070-3482(CaMgSi2O6), indicating the successful synthesis of a pure phaseof Ca1�xEuxMgSi2O6. The crystallite sizes from the TOPAS programwere found to be �16 nm, �13 nm, �17 nm, and �10 nm forx¼0.02, 0.06, 0.1, and 0.2, respectively, indicating that crystallitesizes did not significantly change with x in the current range ofactivator concentrations due to similarity of ionic radii of Ca2þ andEu2þ .

Fig. 6a shows SEM images of the phosphors after post-synthesisannealing at 1350 °C for 3 h. The particles show agglomeration andform aggregates due to the high temperature during post-synthesis annealing. The morphology of the phosphors afterpost-synthesis annealing at 1400 °C for 10 h is shown in Fig. 6b.The particles are more agglomerated than those in Fig. 6a. Fig. 6c–fshows the SEM images of Ca1�xEuxMg(SiO3)2 (x¼0.02, 0.06, 0.1,and 0.2). It should be noted that the morphology of Ca1�xEuxMg(SiO3)2 is not altered when x increases.

3.3. Photoluminescence spectra, color calculator data and quantumefficiency of synthesized phosphors

Fig. 7a shows the PL excitation (PLE, monitored at 511 nm) andPL emission (excitation wavelength, Εex¼350 nm) spectra for(Ca0.06Eu0.04)7Mg(SiO4)4. The PLE spectra show broadbandabsorption in the near UV region from 200 to 450 nm, with amaximum at �350 nm, which arises from the allowed transitionof Eu2þ . The experimental λex (�350 nm) is very similar to thecalculated λex (337 nm), confirming the validity of our approach todiscover new phosphor compositions. The PL emission spectrumconsists of a broad band centered at 511 nm, which is attributed tothe allowed 4f65d1-4f7 transition of Eu2þ . The Φ of(Ca0.96Eu0.04)7Mg(SiO4)4 with the post-synthesis annealing condi-tions of 1100 °C for 3 h under air, 1350 °C for 3 h under air, and1400 °C for 10 h under air, were �19%, � 30%, and 11%, respec-tively. Generally, Φ is increased with increase of annealing tem-perature because of the increase of crystallite sizes [37]. In thiswork, although the crystallite sizes of the powders calcined at1400 °C for 10 h is slightly larger than the powders calcined at1100 °C for 3 h and 1350°C for 3 h, the Φ from the 1400 °C

Fig. 5. X-ray diffraction patterns: (a) (Ca0.96Eu0.04)7Mg(SiO4)4 prepared by the sol–gel/Pechini method with post-synthesis annealing conditions of 1100 °C or 1350 °Cfor 3 h or 1400 °C for 10 h. (b) Ca1�xEuxMgSi2O6 (x¼0.02, 0.06, 0.1, and 0.2) pre-pared by the co-precipitation method with post-synthesis annealing condition of1100 °C for 2 h.

Fig. 4. Unit cell representation of the crystal structure of (a) Ca7Mg(SiO4)4:Eu2þ

(b) CaMgSi2O6:Eu2þ drawn with VESTA [33]. Ca7Mg(SiO4)4 has an orthorhombicstructure with space group Pnn2 with lattice constants a¼0.6742 nm,b¼1.0887 nm, c¼1.8339 nm. CaMg(SiO3)2 has an monoclinic structure with spacegroup C2/c with lattice constants a¼0.9743 nm, b¼0.8879 nm, c¼0.5230 nm. Bothlattice parameters are calculated using TOPAS.

J. Ha et al. / Journal of Luminescence 179 (2016) 297–305 301

powders is significantly smaller than that of the 1100 °C and1350 °C powders. This reduction of Φ can be attributed to aggre-gation of the phosphors particles. As shown in Fig. 6a and b, thepowders calcined at 1400 °C form more aggregated structurecompared to the powders calcined at 1350 °C. Hong et al. [38] andLenggoro et al. [39] reported that highly aggregated phosphorpowders showed a reduction in emission intensity, which is rela-ted to Φ. Aggregated powders can possibly have an influence onthe reabsorption and light scattering, which results in a reductionof Φ [40]. Thus, non-aggregated powders play an important rolefor the obtainment of a high Φ.

Previously reported Φ for (Ca1�xEux)7Mg(SiO4)4, 18% forx¼0.04 (λex¼350 nm) [16] and �23% for x¼0.02 (λex¼400 nm)[17]), synthesized by solid state reaction at 1350 °C for 3 h in N2/H2

at 1100�1400 °C for 4 h in N2/H2, respectively, were similar to thepresent results. However, a high Φ (�64%) for (Ca1�xEux)7Mg(SiO4)4 was previously found for an unusually low activator con-centration, x¼0.001 (λex¼365 nm) [41]. These powders wereformed by a solid-state reaction at 1250 °C for 6 h in N2/H2. Sinceboth the synthesis methodology and post-synthesis calcination,affect the crystallite sizes, the differences between values of Φ forthe present work and for previous reports can be attributed todifferences in synthesis method, post-processing temperature, anddopant concentration.

Fig. 7b shows the PL excitation (PLE, monitored at 458 nm) andPL emission (excitation wavelength, Eex¼350 nm) for CaMg(SiO3)2:Eu2þ . A broad excitation spectrum is observed from200 nm to 400 nm, with a maximum at 350 nm, which is a resultof the allowed transition of Eu2þ . This value is also close to ourcalculated excitation wavelength of 377 nm. The PL emissionshows a spectrum maximum at 458 nm, which is attributed to theparity-allowed 4f 65d1 - 4f7 transition of Eu2þ . Solid lines inFig. 7b refer to the emission monitored at 350 nm for variousconcentrations of Eu2þ (x¼0.02, 0.06, 0.1, and 0.2) in CaMg(SiO3)2.The emission intensity increases until x¼0.06 and then decreasesdue to the concentration quenching effect. The Φ was � 5% whenx¼0.06 for λex¼350 nm, which is the highest quantum efficiencyamong x¼0.02, 0.06, 0.1, and 0.2. This Φ is similar to the pre-viously reported value for this material of �9.1% [42].

Our results indicate that high structural rigidity, as quantifiedby high Θ, is not a sufficient condition for high Φ. Both silicatesstudied in this work have high Θ, but relatively low Φ. For com-pounds that have similar band gaps and chemistries, Θ indeedseems to be correlated with Φ. For example, the calculated Θ’s ofX1–Y2SiO5 (GGA band gap, 4.73 eV) and X2–Y2SiO5 (GGA band

Fig. 6. Scanning electron micrographs of the powders. Ca0.96Eu0.04Mg(SiO4)4 after post-synthesis annealing at: (a) 1350 °C for 3 h and (b) 1400 °C for 10 h. Ca1�xEuxMgSi2O6

with post-synthesis annealing condition of 1100 °C for 2 h: (c) x¼0.02 (d) x¼0.06, (e) x¼0.10 and (f) x¼0.20.

J. Ha et al. / Journal of Luminescence 179 (2016) 297–305302

gap, 4.77 eV) are 491 K and 512 K (Table 1) and the Φ’s are �36%and �45%, respectively, showing that as Θ increases, Φ alsoincreases [43]. However, Φ is also strongly affected by the hostband gap. Furthermore, it has been demonstrated that a host withlow Θ yields high Φ when it is activated with Eu2þ . For example,the computed Θ’s for Ba2SiO4, α-Sr2SiO4 and β-Sr2SiO4 are 307 K,360 K and 428 K, respectively, each of which is smaller than thesuggested 500 K, whereas the experimentally measured Φ forEu2þ-activated Ba2SiO4 and Sr2SiO4 are 4 85% [37,44]. Thesefindings suggest that although calculated Θ may be a startingpoint, caution should be given to the reliability ofΘ as a descriptorof Φ. Therefore, more efforts need to be devoted to explore moreaccurate descriptors to screen for phosphors with high Φ.

For Ca7Mg(SiO4)4:Eu2þ , the color represented by the x, y colorcoordinates (0.25, 0.49) is located in the green region of the dia-gram on the Commission International de I’Eclairage (CIE) dia-gram. According to the National Television System Committee(NTSC) RGB colors, the values of good green-emitting phosphorsare 0.21 and 0.71 [45]. The color coordinates previously reported(0.20, 0.47) were for powders produced by a solid state reactionwith x¼0.001 and λex¼365 nm [41]. The discrepancy in thesecoordinates with the present results can be explained by the dif-ference in x and λex. It was previously shown [41] that the

coordinates were dependent on both λex and x (activator con-centration), attributed to the three different Eu2þ sites. Emissionwavelengths were found to be 455 nm (Eu(1)2þ), 504 nm (Eu(2)2þ), and 540 nm (Eu(3)2þ). Eu2þ will enter in the Eu(1)2þ sitespreferentially for small concentrations of Eu2þ , which willincrease the emission intensity from the Eu(1)2þ sites (455 nm).When the activator concentration increases, Eu2þ preferentiallyoccupies Eu(2)2þ and Eu(3)2þ sites, which will decrease theemission from the Eu(1)2þ sites. It should be noted that thesethree emission spectra overlap and display one broad emissionpeak from 450 nm to 600 nm. Thus, the width of emission spectrawould change, depending on x and λex. Despite the slight differ-ence in values of the color coordinates in this work, the obtainedvalues were in the green color range, indicating that Ca7Mg(SiO4)4:Eu2þ is a potential green-emitting phosphor for use in nUV LEDs.

The x, y color coordinates of Ca0.94Eu0.06Mg(SiO3)2 were 0.14and 0.05, located in the blue region on the CIE diagram and thecolor coordinates of all samples (x ranging from 0.02 to 0.2) aresimilar. These values are similar to those defined by NTSC CIE forblue color (0.14, 0.08), indicating that CaMg(SiO3)2:Eu2þ could beutilized as a blue-emitting phosphor for nUV LEDs.

Fig. 7. (a) Photoluminescence excitation (dashed line monitored at 511 nm) andemission (solid line, λex¼350 nm) spectra of (Ca0.96Eu0.04)7Mg(SiO4). Blue line-s¼post-synthesis annealing at 1350 °C for 3 h and the red lines¼post-synthesisannealing at 1400 °C for 10 h. Both powders were subsequently heated to 1100 °Cfor 4 h in a reducing atmosphere. (b) Photoluminescence excitation (dashed linemonitored at 458 nm) and emission (solid line, λex¼350 nm) spectra ofCa1�xEuxMgSi2O6, x¼0.02, 0.06, 0.1, and 0.2. Post-synthesis annealing conditionwas 1100 °C for 2 h. The inset photographs are the phosphor powders excited with365 nm. Φ¼quantum efficiency (For interpretation of the references to color in thisfigure legend, the reader is referred to the web version of this article.)

Fig. 8. The relationship between the log(I/x) and log(x) of Ca1-xEux(SiO3)2.I¼emission intensity, and x¼0.06, 0.1, and 0.2.

J. Ha et al. / Journal of Luminescence 179 (2016) 297–305 303

3.4. Concentration quenching of CaMg(SiO3)2:Eu2þ

The concentration quenching behavior was analyzed using theemission intensity change as a function of x. Blasse reported thatthe critical transfer distance (Rc) was defined as: [46]:

Rc ¼ 23V

4πxcN

� �1=3

ð4Þ

where xc is the critical dopant concentration when the emissionintensity shows the maximum value, V is the volume of the unitcell, and N is the number of cations in the unit cell. Rc in CaMg(SiO3)2 was calculated to be �1.5 nm by taking the values of V, xc,and N as 0.438 nm3, 0.06, and 4, respectively, from experimentaland analytical evaluations. The interaction type was proposed by

Dexter [47] when x 4 xc [48,49]:

logIx

� �p�θ

3logðxÞ ð5Þ

where I is the emission intensity and θ is a function of electricmultipolar character.

There are three types of concentration quenching mechanismsthat result from electrostatic multipolar interaction: dipole–dipole(d–d), dipole–quadrupole (d–q), and quadrupole–quadrupole (q–q)interactions, corresponding to θ¼6, 8, and 10, respectively. Thetype of interaction can be estimated by plotting (log (I/x) as afunction of (log x). For Ca1-xEuxMg(SiO3)2 with x¼0.06, 0.1, and0.2, the concentration quenching mechanism under 350 nm exci-tation was found to be d–q interaction (θ � 8), as shown in Fig. 8.These findings are in contradiction with the previously reported(d–d) quenching mechanism for the same material for value ofxc¼0.01, with an excitation wavelength of 147 nm [50], possiblydue to the difference in the excitation wavelength that could affectthe emission intensity.

3.5. Thermal quenching process of (Ca0.96Eu0.04)7Mg(SiO4)4 andCa0.94Eu0.06Mg(SiO3)2

The thermal-quenching process of both phosphors was mea-sured as the decrease of emission intensity as a function of tem-perature (from 25 °C to 150 °C). The data was then plotted usingthe following equation to describe thermal quenching of lumi-nescence intensity I(T) with temperature T [29]:

I Tð Þ ¼ I01þΓ0

Γυexp �ΔE

kT

� � ð6Þ

where I0 is the initial intensity at 25 °C, ΔE is the energy barrier forthermal quenching, k is the Boltzmann's constant, Γ0 is theattempt rate for thermal quenching at T¼1, and Γυ is the radia-tive decay rate of the 5d state of Eu2þ . Here we adopted the Γ0 andΓυ with values of 3�1013 Hz and 1.1�106 Hz, respectively, in linewith estimations in the literature [29]. Fig. 9a and b shows the PLemission intensity as a function of temperature from 25 °C to150 °C, demonstrating that there is a significant quenching of theluminescence over this temperature range. Fig. 9c shows a plot of

Fig. 9. The emission intensity (I) of the photoluminescence spectra as a function of temperature for (a) (Ca0.96Eu0.04)7Mg(SiO4)4, (b) Ca0.94Eu0.06Mg(SiO3)2. (c) Plot of therelative intensity as a function of temperature from the data in (a) and (b). ΔE¼activation energy and Γ0 is the attempt rate for thermal quenching at T¼1, and Γυ is theradiative decay rate of the 5d state of Eu2þ .

J. Ha et al. / Journal of Luminescence 179 (2016) 297–305304

the relative intensity as a function of temperature for both phos-phors. Γ0/Γυ is 2.7�107. The values for average ΔE were obtainedin each temperature. ΔE were 0.5670.03 eV for (Ca0.96Eu0.04)7Mg(SiO4)4 and 0.5270.02 eV for Ca0.94Eu0.06Mg(SiO3)2. Note thatthere are many factors that influence the measured intensity thanjust thermal quenching alone (e.g. synthesis conditions, particlecrystallinity). These factors will also affect the measured curve andsubsequently the fitted thermal activation barrier. But we believethe relative trend for both cases is reasonable. These ΔE valueswere in good agreement with the range of calculated values ofsimilar calcium compounds (0.2–0.6 eV) [29]. In general, the largerthe thermal activation energy, the better Φ obtained for Eu2þ-activated phosphors. Although the predicted Θ for (Ca0.96Eu0.04)7Mg(SiO4)4 (601 K) is smaller than Ca0.94Eu0.06Mg(SiO3)2 (665 K),the Φ of (Ca0.96Eu0.04)7Mg(SiO4)4 (30%) is relative higher thanCa0.94Eu0.06Mg(SiO3)2 (5%), which can be attributed to its largerthermal activation energy.

A simplified relationship between the energy barrier and thequenching temperature T0.5 (the temperature at which the emis-sion intensity declined to 50% of the low temperature value) wasdeveloped by Dorenbos [29] from Eq. (6):

ΔE¼ T0:5

680eVð Þ ð7Þ

From this equation, the obtained T0.5 for (Ca0.96Eu0.04)7Mg(SiO4)4and Ca0.94Eu0.06Mg(SiO3)2 are 381 K and 354 K, respectively. Theselow thermal quenching temperatures suggest that Φ’s of(Ca0.96Eu0.04)7Mg(SiO4)4 and Ca0.94Eu0.06Mg(SiO3)2 are severelydegraded, even at room temperature.

Other models of thermal quenching exist. For example, thethermal ionization model [29] suggests that thermal quenchingstability is related to the energy gap between the excited Eu 5dlevel and CBM. However, first principles determination of excitedstates is challenging, though a few recent works have explored thisrelationship using advanced techniques [51]. For high-throughputscreening, a more efficient approximation needs to be developed.

For example, the band gap can sometimes be used as a proxy[12,52]; the larger the band gap of the host, the more likely thatthere is a larger gap between the excited 5d level and the CBM(though the exact relationship depends on the magnitude of theStokes shift). The investigation of these other first principlesdescriptors will be the subject of future work.

4. Conclusions

An integrated approach of exploring new phosphors for near-UV (nUV) excitation by combining experiments and calculations isoutlined. The descriptors of Debye temperature (Θ) and excitationenergy were combined to screen plausible phosphors that arepredicted to have high quantum efficiency (Φ) and can beemployed in near-UV emitting LEDs. From the calculation of 27host materials, Ca7Mg(SiO4)4 and CaMg(SiO3)2 were selectedbecause predicted Θ for Ca7Mg(SiO4)4 and CaMg(SiO3)2 werefound to be 601 K and 665 K, respectively, which are expected tohave high Φ (480%). The predicted and experimental excitationenergies are similar. The measured values of Φ are lower than theexpected values, based on high Θ. The main peak in the photo-luminescence emission spectra under excitation of 350 nm is511 nm for (Ca0.96Eu0.04)7Mg(SiO4)4, with chromaticity coordinatesof (0.25, 0.49), and 458 nm for Ca0.94Eu0.06Mg(SiO3)2 with chro-maticity coordinates of (0.14, 0.05). The concentration self-quenching mechanism for CaMg(SiO3)2:Eu2þ under 350 nm isthe d–q interaction. The thermal quenching temperatures (T0.5) are381 K and 354 K for (Ca0.96Eu0.04)7Mg(SiO4)4 and Ca0.94Eu0.06Mg(SiO3)2, respectively, indicating poor thermal stability. Despitehaving a high Θ, a low Φ is found for the two candidates. Con-versely, other screened hosts (e.g. Ba2SiO4:Eu2þ) have low com-puted Θ, but are known to have a high Φ. Therefore, the use ofhigh Θ as an indicator for high quantum efficiency should betreated with caution when screening host candidates. It is

J. Ha et al. / Journal of Luminescence 179 (2016) 297–305 305

desirable to develop more well-defined criteria for the screeningof host candidates for high Φ phosphors. The approach developedin this work can be leveraged for the discovery of additional nUVexcited phosphors.

Acknowledgments

This work is supported by the United States National ScienceFoundation, Ceramics Program Grant DMR-1411192.

References

[1] S. Nakamura, G. Fasol, The Blue Laser Diode, Springer, Berlin, 1996.[2] N. Narendran, N. Maliyagoda, A. Bierman, R.M. Pysar, M. Overington, Proc. SPIE

Int. Soc. Opt. Eng. 3938 (2000) 240.[3] M.G. Craford, Light Emitting Diode Technology, Kluwer Academic Publishers,

Dordrecht, 1996.[4] H.S. Jang, W.B. Im, D.C. Lee, D.Y. Jeon, S.S. Kim, J. Lumin. 126 (2007) 371.[5] Y.Q. Li, A.C.A. Delsing, G. de With, H.T. Hintzen, Chem. Mater. 17 (2005) 3242.[6] J.K. Sheu, S.J. Chang, C.H. Kuo, Y.K. Su, L.W. Wu, Y.C. Lin, W.C. Lai, J.M. Tsai, G.

C. Chi, R.K. Wu, IEEE Photonics Technol. Lett. 15 (2003) 18.[7] S. Neeraj, N. Kijima, A.K. Cheetham, Chem. Phys. Lett. 387 (2004) 2.[8] S.H.M. Poort, W. Janssen, G. Blasse, J. Alloy. Compd. 260 (1997) 93.[9] Y. Chen, Y. Li, J. Wang, M. Wu, C. Wang, J. Phys. Chem. C 118 (2014) 12494.[10] Z. Xia, Z. Xu, M. Chen, Q. Liu, Dalton Trans. 45 (2016) 11214.[11] T. Takeda, N. Hirosaki, S. Funahshi, R.-J. Xie, Chem. Mater. 27 (2015) 5892.[12] Z. Wang, I.-H. Chu, F. Zhou, S.P. Ong, Chem. Mater. 28 (2016) 4024.[13] P. Dorenbos, J. Lumin. 91 (2000) 155.[14] P. Dorenbos, J. Lumin. 104 (2003) 239.[15] J. Brgoch, S.P. DenBaars, R. Seshadri, J. Phys. Chem. C 177 (2013) 17955.[16] Y. Jia, H. Qiao, Y. Zheng, N. Guo, H. You Mar 14, Phys. Chem. Chem. Phys. 14

(2012) 3537.[17] K.H. Lee, S.H. Park, H.S. Yoon, Y.-I. Kim, H.G. Jang, W.B. Im, Opt. Express 20

(2012) 6248.[18] J. McKittrick, M.E. Hannah, A. Piquette, J.K. Han, J.I. Choi, M. Anc, M. Galvez,

H. Lugauer, J.B. Talbot, K.C. Mishra, ECS J. Solid State Sci. Technol. 2 (2012)R3119.

[19] P.E. Blöchl, Phys. Rev. B 50 (1994) 17953.[20] G. Kresse, Phys. Rev. B 54 (1996) 11169.

[21] J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 77 (1996) 3865.[22] S.P. Ong, S. Cholia, A. Jain, M. Brafman, D. Gunter, G. Ceder, K.A. Persson,

Comput. Mater. Sci. 97 (2015) 209.[23] A. Jain, S.P. Ong, G. Hautier, W. Chen, W.D. Richards, S. Dacek, S. Cholia,

D. Gunter, D. Skinner, G. Ceder, K.A. Persson, APL Mater. 1 (2013) 011002.[24] Z. Deng, Z. Wang, I.-H. Chu, J. Luo, S.P. Ong, J. Electrochem. Soc. 163 (2015) A67.[25] A. Chaudhry, R. Boutchko, G. Zhang, N. Grønbech-Jensen, A. Canning, Phys.

Rev. B 89 (2014) 155105.[26] F. Tran, P. Blaha, Phys. Rev. Lett. 102 (2009) 5.[27] P. Dorenbos, J. Lumin. 122-123 (2007) 315.[28] Z. Zhang, O.M. ten Kate, A. Delsing, E. van der Kolk, P.H.L. Notten, P. Dorenbos,

J. Zhao, H.T. Hintzen, J. Mater. Chem. 22 (2012) 9813.[29] P. Dorenbos, J. Phys.: Condens. Matter 17 (2005) 8103.[30] A.U. Pawar, A.P. Jadhav, U. Pal, B.K. Kim, Y.S. Kang, J. Lumin. 132 (2012) 659.[31] J. Wang, Y. Huang, X. Wang, L. Qin, H.J. Seo, Mater. Res. Bull. 55 (2014) 126.[32] R.D. Shannon, Theor. General. Crystallogr. 32 (1976) 751.[33] K. Momma, F. Izumi, J. Appl. Crystallogr. 44 (2011) 1272.[34] P.B. Moore, T. Araki, Am. Miner. 61 (1976) 74.[35] L. Levien, C.T. Prewitt, Am. Miner. 66 (1981) 315.[36] W.-N. Wang, W. Widiyastuti, T. Ogi, I.W. Lenggoro, K. Okuyama, Chem. Mater.

19 (2007) 1723.[37] J.K. Han, M.E. Hannah, A. Piquette, J.B. Talbot, K.C. Mishra, J. McKittrick, ECS J.

Solid State Sci. Technol. 1 (2012) R98.[38] G.Y. Hong, B.S. Jeon, Y.K. Yoo, J.S. Yoo, J. Electrochem. Soc. 148 (2001) H161.[39] I.W. Lenggoro, Y. Itoh, K. Okuyama, T.O. Kim, J. Mater. Res. 19 (2011) 3534.[40] S. Jun, J. Lee, E. Jang, ACS Nano 7 (2013) 1472.[41] J. Zhang, C. Jiang, Mater. Res. Bull. 60 (2014) 467.[42] A. Pawar, A. Jadhav, C.W. Kim, H.G. Cha, U. Pal, Y.S. Kang, J. Lumin. 157 (2015)

131.[43] D. Cervantes-Vasquez, O.E. Contreras, G.A. Hirata, J. Lumin. 143 (2013) 226.[44] J.K. Han, M.E. Hannah, A. Piquette, G.A. Hirata, J.B. Talbot, K.C. Mishra,

J. McKittrick, J. Lumin. 132 (2012) 106.[45] J.H. Oh, S.J. Yang, Y.-G. Sung, Y.R. Do, Opt. Express 21 (2013) 4539.[46] G. Blasse, J. Solid State Chem. 62 (1986) 207.[47] D.L. Dexter, J. Chem. Phys. 21 (1953) 836.[48] L.G.V. Uitert, J. Electochem. Soc. 114 (1967) 1048.[49] L. Ozawa, P.M. Jaffe, J. Electrochem. Soc. 118 (1971) 1678.[50] W.B. Im, Y.-I. Kim, J.H. Kang, D.Y. Jeon, H.K. Jung, K.Y. Jung, J. Mater. Res. 20

(2005) 2061.[51] S. Poncé, Y. Jia, M. Giantomassi, M. Mikami, X. Gonze, J. Phys. Chem. C 120

(2016) 4040.[52] L. Chen, X. Chen, F. Liu, H. Chen, H. Wang, E. Zhao, Y. Jiang, T.S. Chan, C.

H. Wang, W. Zhang, Y. Wang, S. Chen, Sci. Rep. 5 (2015) 11514.