Article pubs.acs.org/IC
Combinatorial Screening of Luminescent and Structural Properties in a Ce3+-Doped Ln-Al-Si-O‑N (Ln = Y, La, Gd, Lu) System: The Discovery of a Novel Gd3Al3+xSi3−xO12+xN2−x:Ce3+ Phosphor Woon Bae Park,‡ Satendra Pal Singh,‡ Minseuk Kim, and Kee-Sun Sohn* Faculty of Nanotechnology and Advanced Materials Engineering, Sejong University, Seoul 143-747, Korea S Supporting Information *
ABSTRACT: The discovery of novel phosphors for use in light emitting diodes (LED) has gained in significance because LEDbased solid-state lighting applications now attract a great deal of attention for energy savings and environmental concerns. Recent research trends have centered on the discovery of novel phosphors, not on slight variations of well-known phosphors. In a real sense, novelty goes beyond simple variations or improvements in existing phosphors. A brilliant strategy for the discovery of novel phosphors is to introduce an appropriate activator to existing inorganic compounds. These compounds have structures that are welldefined in crystallographic structure databases, but they have never been considered as a phosphor host. Another strategy is to discover new host compounds with structures that cannot be found in existing databases. We have simultaneously pursued both strategies by employing metaheuristics-assisted combinatorial material search techniques. In the present investigation, we screened a search space consisting of Ln-Al-Si-O-N (Ln = Y, La, Gd, Lu), and thereby we discovered a blue-light-emitting novel phosphor, Gd3Al3+xSi3−xO12+xN2−x:Ce3+, with a monoclinic system in the C2 space groupa potential candidate for UV-LED applications.
1. INTRODUCTION We have recently discovered a series of novel phosphors for use in light emitting diode (LED) applications such as solid-state lighting and flat-panel displays by employing a so-called metaheuristics-assisted combinatorial material search process.1,2 The discovered phosphors exhibit no overlap with any existing phosphors.1−5 Because the issue of novelty has been of great concern relative to the intellectual property (IP) of phosphors in the field,6 the discovery of novel phosphors should be tactfully handled. A brilliant tactic to discover novel phosphors with no IP conflict is to introduce an appropriate activator to known compounds that have never been considered as a phosphor host. The other tactic is to discover new host compounds, the crystallographic structure of which cannot be found in any databases. Novel phosphors belonging to the former case are herein designated as Type III and the latter case are referred to Type IV.1−5 Both types of these phosphors are novel in a real sense, and most commercially available LED phosphors can be categorized into one of these two types, depending upon the history of their discovery.1,2 Either a slight compositional variation or a hybridization of well-known phosphors would give rise to IP complications in the field.6 The discovery of Type IV novel phosphors deserves to be highlighted in both a scientific and a practical sense. The Type III approach is also recommendable because it is more efficient and requires less effort by comparison with the Type IV approach. It is easy to pinpoint a suitable combination between © XXXX American Chemical Society
well-known inorganic compounds and several well-known activators.7 For instance, several well-known, commercially available LED phosphors such as CASN, Beta (or alpha)SIALON, and Sr2Si5N8:Eu2+ are Type III novel phosphors.7a,8−10 Therefore, pursuing Type III novel phosphors has proven to be valuable in the field. In this respect, Kakihana et al.11,12 have recently developed a brilliant strategy (i.e., the mineral-inspired approach based on the solution parallel synthesis (SPS)) and discovered novel Type III phosphors and thereby sorted out possible phosphor hosts from either mineral or inorganic compound databases. Nonetheless, Type IV novelty has been our primary discovery target based on the correct structural determination of unknown inorganic compounds. In reality, the possibility of finding new Type III novel phosphors is limited because many of them have already been exploited by phosphor researchers over the past decade. In addition, it should be noted that there exist only 8230 structure prototypes for inorganic compounds, as determined from the categorization of all entries in the inorganic crystal structure database (ICSD).13,14 The realm for the discovery of Type III novel phosphors narrows every year. As for the discovery of Type IV novel phosphors, however, there remains a great deal of possibility. For instance, Xie et al.15 have recently reported an effective strategy based on a single-particleReceived: November 13, 2014
A
DOI: 10.1021/ic502721h Inorg. Chem. XXXX, XXX, XXX−XXX
Article
Inorganic Chemistry
preliminarily not recognized as a Type IV novel phosphor, but was regarded as a Type III during the screening process because the XRD pattern of it looked very similar to its well-known neighbor, the U-phase [Ln3(Al3+xSi3−x)O12+xN2−x (x ≈ 0.5, Ln = La, Nd)], which has a trigonal structure in the P321 space group.20 However, our precise structural determination process revealed that the crystal structure of Gd3Al3+xSi3−xO12+xN2−x:Ce3+ is slightly different from that of the U-phase. The introduction of Gd at the Ln site was found to induce a phase transformation into one of its subgroup structures. Although the atomic arrangement was not significantly altered by the phase transformation, the unit cell and the space group changed according to the group−subgroup transformation scheme. Therefore, the discovered Gd3Al3+xSi3−xO12+xN2−x:Ce3+ phosphor merits categorization as a Type IV rather than as a Type III.
diagnosis approach. They pinpointed a single luminescent crystallite from a powder sample by employing an optical microscope and glass fiber, and identified their crystal structure by employing a single crystal X-ray diffraction technique. By employing this technology they have discovered several novel phosphors. Although the realization of a single-phase powder sample for practical application is a daunting task, the singleparticle-diagnosis approach should be one of the most promising discovery strategies. Schnick et al.16−18 have published a series of reports on novel nitride phosphors based on AB4X4 (I4/m) and ABC3X4 (P1 or I41/a) structures. They referred to existing NaLi3(SiO4), UCr4C4, and Cs[Na3PbO4] compounds prior to their choice of isotypic nitrides. Although the prototype structure of the discovered phosphors exists in the inorganic compound database, these phosphors can be categorized into Type IV. We have used a special strategy to facilitate the discovery of Type IV and Type III novel phosphors, which is based on a metaheuristics-involved pseudohigh-throughput process.1,2 The final product of this discovery strategy is a practical single-phase powder sample. The use of this discovery strategy was so fruitful that we discovered several Type IV novel phosphors.1,2,4 In this regard, a part of this strategy was also used in the present investigation. We reduced the search space significantly in the present investigation, and therefore, the use of metaheuristics was unnecessary in the first rough screening. Instead, we implemented a simple pseudohigh-throughput process by screening Ce3+-activated lanthanide oxides−aluminum oxide (or nitride)−silicon nitride ternary composition spaces, which is a so-called ternary combinatorial library. The most important criterion, on which the first rough screening was executed, was the issue of whether or not a novel compound was included in samples in the combinatorial library. In this context, the major issue of the first rough screening was not a luminescent property but the phase identification from XRD patterns of various compositions. The luminescence check-up in the first rough screening was ruled out because several well-known phosphors with very good luminescent properties likely exist in the search pool, which could have interfered with our actual motive of finding novel compounds. The first rough screening step was a preliminary process to make it possible to secure a plausible composition range leading to novel phosphors in a real sense. The first rough screening was followed by a particle swarm optimization (PSO)-assisted combinatorial material search (PSOCMS) process.19 PSOCMS is a fine-tuning process used to pinpoint the final choice of a phosphor composition, which could guarantee a nearly single-phase phosphor with no novelty-related complications, thereby leading to no IP conflict in the end. PSO has proven to be a powerful tool in reducing the experimental burdens.1−4 Unless the PSOCMS process had been adopted, the number of samples that we had to synthesize and characterize would have been huge in a given composition search pool. Accordingly, the PSOCMS process is extremely useful, particularly when the search space is larger than the binary system. In the present investigation, by employing the strategy described above, we successfully discovered a novel Type IV phosphor, Gd3Al3+xSi3−xO12+xN2−x:Ce3+, the structure of which was found to be monoclinic in the C2 space group. Because the host structure of the discovered phosphor was unknown, we finally incarnated it as a Type IV phosphor. In fact, the discovered novel phosphor, Gd3Al3+xSi3−xO12+xN2−x:Ce3+, was
2. EXPERIMENTAL PROCEDURES The commercially available starting powder materials, Y2O3 (Kojundo, 99.9% UP), Gd2O3 (Kojundo, 99.9%), La2O3 (Kojundo, 99.99%), Lu2O3 (Kojundo, 99.9%), Al2O3 (Kojundo, 99.99%), AlN (Kojundo, 99.9%), α-Si3N4 (Ube, unreported), and CeO2 (Kojundo, 99.99%), were dispensed into a so-called combi-chem container, a specially designed sample container made of BN (80 × 40 × 20 mm), which involved 18 sample sites that were 8.5 mm in diameter and 16 mm in depth. Preparations such as mixing, grinding, and firing of a large number of samples were executed inside the combi-chem container. The total amount of raw materials at each sample site was about 0.3 g, which produced a sufficient amount of final phosphor powder available for use in any of the conventional characterizations. The exact amounts of the raw materials were weighed and dispensed to the sample sites. The automatic mixing and grinding, which had been normally adopted in our previous combinatorial process,1−4 turned out to be unsatisfactory, so we prepared 36 agate mortars with an inside diameter of 450 mm, and the samples were transferred to these agate mortars and ground manually. Although this manual process required much time, the final quality of the samples was improved significantly. The mixed raw materials were retransferred to the combi-chem container and then fired at 1600 °C for 4 h under a N2 gas flow (500 mL/min) in a sealed tube furnace. Two combi-chem containers (i.e., 36 samples) were fired simultaneously. Each fired sample was ground and subjected to X-ray diffraction (XRD) and photoluminescence (PL) analysis. The emission spectra were monitored either at 400 or 460 nm excitation in a pseudohighthroughput manner using an in-house-fabricated continuous-wave (CW) PL system equipped with a xenon lamp. Finally, discovered novel samples were examined using synchrotron radiation X-ray diffraction (SR-XRD). The SR-XRD measurements of the selected sample were conducted using a 9B high-resolution powder-diffraction beamline at the Pohang Accelerator Laboratory (PAL). The incident synchrotron X-rays were monochromatized to a wavelength of 1.4647 Å by a double-bounce Si(111) monochromator and calibrated with a SRM660a standard sample. The detector arm of the diffractometer had soller slits with an angular resolution of two degrees, a flat Ge{111} crystal analyzer, an antiscatter baffle, and a scintillation detector. Data were collected in the angular 2θ range of 10.0°−130.5° with a step size of 0.01°.
3. RESULTS AND DISCUSSION 3.1. Preliminary Screening. The top row in Figure 1 shows the search space that we adopted for the first screening, which is marked with red dots. Eight ternary composition libraries were separately prepared comprising a total of 288 different samples. Four different lanthanide oxides, Y2O3, La2O3, Gd2O3, and Lu2O3, were adopted as lanthanide sources, and both aluminum oxide (Al2O3) and nitride (AlN) were used B
DOI: 10.1021/ic502721h Inorg. Chem. XXXX, XXX, XXX−XXX
Article
Inorganic Chemistry
superior to any of the computer-aided automatic phasematching processes. A sort of pseudoquantitative phase identification was adopted to obtain a rough estimate of the phase fraction from the phase identification results. This practical phase identification process had nothing to do with the conventional scale factor-related quantitative analysis leading to an accurate evaluation of a weight fraction and a volume fraction, which typically had been adopted either in the Le Bail refinement or in Rietveld refinement.24 It would be impossible and unreasonable to execute such analysis by employing low-quality Laboratory XRD data collected in the combinatorial process. In fact, we used high-resolution synchrotron XRD data of a discovered phosphor to carry out more precise quantitative analysis in the later stages of the present investigation. Prior to precise quantitative analysis of the discovered phosphor, a pseudoquantitative phase analysis seems to be plausible and also very useful. The strategy adopted in the pseudoquantitative phase analysis is as follows. First, we took into account two or three major constituent phases for every sample, and then the ratio between the heights of the strongest peaks of each constituent phase was used as a rough measure of phase fraction. When the strongest peak of a certain constituent phase was hidden because of an overlap with peaks from other phases, then we predicted the height of this hidden peak by referring to other conspicuous peaks of this phase and to the standard data. Although this subjectively defined parameter (i.e., relative peak height parameter) is not a scientifically well-defined and reliable parameter for the precise prediction of a phase fraction, it was utilized only in a practical sense. The use of such a rough measure of the phase fraction was sufficient, because the current stage was used not to pinpoint a specific novel compound but rather to be a preliminary rough screening in order to find a novel phosphor that would certainly be either a single phase of Type III or Type IV. Estimated peak fractions were represented visually using pie graphs located at an appropriate composition site in the combichem library in a systematic manner. As a result, a number of small pie graphs were located in every sample site in the ternary libraries, as shown in Figure 1. The information concerning every constituent compound is summarized in Table 1, wherein the space group and ANX formula (with allowed elements) are also presented along with the typical chemical formula. A nominal composition for each constituent compound should not be equal to the ANX, but it might be a sort of solid solution type as it resides in the combi-chem library because a lanthanide site always welcomes other lanthanides just as Al/ Si sites always tend to be shared by each other as do O/N sites. However, all chemical formulas that are based on shared composition obey the ANX formula. It should be noted that a considerable number of phosphors adopt such a nominal composition with shared Wyckoff sites included in the structure.2 In this respect, the nominal compositions of each constituent compound introduced in Table 1 should be understood. In addition, the peak index that we used for the pie graph calculation is clarified in Table 1, and it designates the strongest peak of each compound’s XRD pattern. In the case of Ln6Si11N20O (A6B11XY20), LnAlO3 (ABX3), and Ln3Si8O4N11 (A3B8X2Y13), three overlapped peak indices were present as a maximum intensity for use in the pie graph calculation. It is for this reason that the inclusion of aluminum produced a solid solution type of compound, which led to site sharing and an
Figure 1. Eight ternary combi-chem libraries consisting of Y2O3, La2O3, Gd2O3, Lu2O3, Al2O3, AlN, and Si3N4. The pie graph located at each site represents a rough measure of the phase fraction, and each color in the pie graph stands for the corresponding compound, as defined in Table 1. The small ternary area marked with red dots at the top is equal to the zoomed-in ternary composition below.
as an aluminum source in the library. The silicon source was Si3N4, and we precluded SiO2 from our ternary composition libraries because of its low melting point. The ternary library consisting of 36 different compositions is located around the silicon nitride corner. The lanthanide side tended to be melted down and the aluminum side was likely to leave starting materials intact as residues even after firing. Therefore, we excluded both the lanthanide oxide and aluminum oxide (or nitride) sides from the search space and constructed the search space as close to the silicon nitride corner as possible. However, getting closer to the silicon nitride corner was not recommended because most of the samples in this area would be apt to turn into various well-known SIALON phases, which have been extensively exploited by phosphor researchers in recent years.21−23 Consequently, it turns out that the choice of the search space in the first rough screening shown in Figure 1 was reasonably constructed in terms of the possibility of finding novel compounds for a phosphor host. In this context, the design of the search space should be the most important step for the discovery of novel phosphors The XRD pattern of every sample was examined to identify the constituent compound. The phase identification was carried out qualitatively by referring to the ICDD database. The phase matching (or identification) was implemented manually based on experience and intuition, even though phase-matching software (PDXL, Rigaku) was used in the initial stage. The manual qualitative phase-matching process proved to be C
DOI: 10.1021/ic502721h Inorg. Chem. XXXX, XXX, XXX−XXX
Article
Inorganic Chemistry Table 1. List of Compounds Residing in the Combi-Chem Librarya
a
Each compound is marked as a distinct color for pie graph representation. Each column of the table exhibits a symbol color, chemical formula, ANX formula, possible lanthanides on the A site, space group, peak angle ranges used for the rough measure of a pie graph, and indexes belonging to the peak.
Figure 2. Quaternary composition search space (Gd2O3−Al2O3−Si3N4−Ce2O3) used for PSOCMS.
of the categorization principle follows. Type I includes phosphors developed long ago for old-fashioned applications that have never included LEDs. Type II is a composition variation that is based in part on well-known LED phosphors, which could be controversial in terms of a novelty claim in the
alteration of the peak index of the maximum intensity according to the Al content. In addition, the classification of every constituent compound should be examined according to the categorization principle suggested by us in our previous reports.1,2 A brief explanation D
DOI: 10.1021/ic502721h Inorg. Chem. XXXX, XXX, XXX−XXX
Article
Inorganic Chemistry
Figure 3. PSOCMS-execution results: instantaneous swarm positions in the quaternary Gd2O3−Al2O3−Si3N4−Ce2O3 composition search space to the fifth swarm. The Ce3+ activator concentration ranged from 0.01 to 0.08 mol. The first column shows the emission spectra of all the samples in the swarm, the second column shows the PSOCMS-execution in the quaternary composition search space, and the third column shows the 29.3° XRD peak of all the samples in the swarm. The inset in the first column shows actual photos and those in the third column shows the pie graphs depicting the rough measure of the phase fraction. The composition data along with the XRD and PL intensity are provided in Table s1 (in the Supporting Information).
field if it was to be used in LED applications. There were no Type I phosphors appearing in our search space in the present investigation. For instance, A3B5X12 phosphors with a garnet structure were categorized into Type II in the present investigation, because garnet phosphors (A3B5X12) have been used in LED applications since 1993.25 Type III and Type IV have already been explained in the Introduction section. The most significant fact impacting the novelty issue is that we pursued Type III and Type IV only in order to avoid controversy and thereby to secure novelty in a real sense. Complete details with respect to the novelty of phosphors have been provided in our previous reports.1,2 There are composition regions in the search space, which exhibit unknown phases with no match to existing compounds in any databases. These unknown phases detected in the combichem library might attract a certain extent of attention because unknown phases would guarantee novelty and lead to the discovery of a Type IV novel phosphor. However, it is unfortunate that all of these unknown phases are mostly
accompanied by a glassy phase, so that their crystallinity is poor. Therefore, we ignored all the unknown phases appearing in the combi-chem library. The ignored unknown phases, along with their symbol color in the pie graph, are presented in Table 1. The most obscured feature of Figure 1 is the presence of a glassy phase. A quantitative analysis of the glassy phase was beyond the scope of the present investigation. It should be noted that a certain amount of a glassy phase is present in almost every sample in the combi-chem library. The worst cases are represented by white color in the pie graph, which denotes a completely melted sample. The intact raw materials, which did not take part in the synthesis process even during the hightemperature firing were also separately categorized and are represented by gray color in the pie graph. Most of the compounds detected in the combi-chem library are considered to be phosphor candidates for use in LED applications. For example, Ln2Si3O3N4 (A2B3X3Y4) has been considered as a host for a Ce3+ activator, which constitutes a blue-light-emitting phosphor.26,27 Ln4Si2O7N2 (A2B4XY8) is E
DOI: 10.1021/ic502721h Inorg. Chem. XXXX, XXX, XXX−XXX
Article
Inorganic Chemistry
Figure 7. Observed (dots), calculated (red line), and difference (blue line) profiles obtained after full-pattern Rietveld refinement of Gd3Al3+xSi3−xO12+xN2−x:Ce3+ using a monoclinic structure in the C2 space group in the 2θ in the range 10 to 80°. Insets depict the Rietveld fit for higher 2θ in the range 80 to 130.5°. The vertical tick marks above the difference profile in the first, second and third line, respectively, from the top denotes the position of Bragg reflections for the Gd 3 Al 3+x Si 3−x O 12+x N 2−x :Ce 3+ phase and impurity phases Si2Al4O4N4 (P63/m space group) and SiAl4O2N4 (R3̅m space group).
Figure 4. Emission and excitation spectra for Gd3Al3+xSi3−xO12+xN2−x:Ce3+.
suitable host structure for Ce3+ activation, and thereby YAG, TAG, and LuAG have been successfully commercialized for use in LED applications.25 The Ln6Si11N20O:Ce3+ phosphor (A6B11XY20) was very recently discovered by the authors as a yellow-light-emitting novel phosphor, and the possibility of its use in LED applications has been examined extensively.28 LnSiO2N (ABXY2) is also known as a blue-light-emitting phosphor after the introduction of a Ce3+ activator.26,27 LnAlO3 (ABX3) has a well-known perovskite structure, and many attempts have been made to convert it into a phosphor by introducing various activators.29 None of these attempts, however, have proven to be a promising candidate for LED applications. Ln3Si8O4N11 (A3B8X2Y13) is also very well-known as a phosphor host via the introduction of a Ce3+ activator, leading to blue-color emitting phosphors, the final target of which was an LED application.27 Ce3+-activated LnSi3N5 (AB3X5) has been of interest as a candidate for LED applications, as it exhibits a blue-colored emission.30 In addition, Al inclusion in LnSi3N5 improves luminescence for LED application.31 Despite such an extensive investigation into the above-described constituent compounds appearing in the combi-chem library, none of them have been successfully commercialized for use in LED applications. What interested us was the presence of a potential Type III candidate in the combi-chem library shown in Figure 1. This phase showed up in the Gd-involved ternary library. The XRD pattern of this phase resembled that of Ln3Al3Si3O12N2 (Ln = La and Nd), with a trigonal structure in the P321 space group, the ANX formula of which is known to be either AB3C5X14 or A2B3C4X14.20,32 It should be noted that the stoichiometry of the Gd-based phase and its categorization into Type III was only tentative because the precise structural determination process (details of which are described in the subsequent section) proved this phase to be of a slightly different structure due to the phase transformation. The Gd-involved novel phase that we discovered in the combi-chem library will be regarded as a novel Type IV phosphor (or compound) and christened as either an “S” phase or an “S” compound. The S compound has
Figure 5. Variation of profile width as a function of 2θ for 200 and 002 reflections of the P321 space group.
Figure 6. Observed (dots), calculated (red line), and difference (blue line) profiles for some selected peak profiles obtained after full-pattern Rietveld refinement of Gd3Al3+xSi3−xO12+xN2−x:Ce3+ using (a) trigonal structure in the P321 space group and (b) monoclinic structure in the C2 space group, along with impurity phases Si2Al4O4N4 (P63/m space group) and SiAl4O2N4 (R3̅m space group). The vertical tick marks in the first and second line, respectively, from the top denotes the position of Bragg reflections for Gd3Al3+xSi3−xO12+xN2−x:Ce3+ phase and impurity phases Si2Al4O4N4 (P63/m space group) or SiAl4O2N4 (R3̅m space group).
known as a blue-light-emitting phosphor that adopts a Ce3+ activator. 26,27 As already mentioned above, Ln 3 Al 5 O 12 (A3B5X12) with a Garnet structure has proven to be a very F
DOI: 10.1021/ic502721h Inorg. Chem. XXXX, XXX, XXX−XXX
Article
Inorganic Chemistry
Table 2. List of the Atomic Coordinates, Isotropic (Uiso) Displacement Parameters (Å2), and Site Occupancy Factor (SOF) Obtained after a Full-Pattern Rietveld Refinement of Gd3Al3+xSi3−xO12+xN2−x:Ce3+ Using Powder SR-XRD Data atom
Wyckoff site
x/a
y/b
z/c
Uiso (Å2)
SOF
Gd1 Gd2 Al1 (Si, Al)2 (Si, Al)3 (Si, Al)4 (O, N)1 (O, N)2 (O, N)3 (O, N)4 (O, N)5 (O, N)6 (O, N)7
2a 4c 2a 4c 2b 4c 4c 4c 4c 4c 4c 4c 4c
0.00000 0.20689 (15) 0.00000 0.8294 (8) 0.00000 0.3778 (7) 0.8327 (18) 0.0717 (16) 0.2849 (16) 0.1456 (14) 0.0758 (13) 0.3949 (16) 0.0463 (13)
0.5711 (18) 0.1925 (18) 0.00000 0.489 (2) 0.223 (2) 0.369 (2) 0.485 (4) 0.365 (4) 0.398 (4) 0.684 (4) 0.142 (3) 0.504 (4) 0.808 (3)
0.00000 0.0007 (5) 0.00000 0.5480 (10) 0.50000 0.502 (3) 0.202 (2) 0.323 (5) 0.297 (5) 0.311 (4) 0.752 (4) 0.762 (4) 0.766 (4)
0.0007 (9) 0.0001 (4) 0.0021 (16) 0.0015 (12) 0.001 (4) 0.004 (2) 0.0038 0.0038 0.0038 0.0038 0.0038 0.0038 0.0038
1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0
Compound name: Gd3Al3+xSi3−xO12+xN2−x. Wavelength (λ): 1.4647 Å. Space group: C2. Space group no.: 5. Z: 2. Lattice parameters: a = 13.67681(15) Å, b = 7.86858(9) Å, c = 4.85157(3) Å. α = γ = 90°, β = 89.8430(9) (Rp = 5.58, Rwp = 7.12, Rexp = 5.45 and χ2 = 1.71).
never before been considered a phosphor host. The Ce3+ activation produced a certain degree of blue-light emission under soft UV excitations. Despite the interesting luminescent property, we focused more on the phase identification because our primary concern in this combi-chem screening centered on the novelty of a phosphor based on the IP availability. In this context, the S phase should be a promising candidate in terms of both luminescence and phosphor novelty. Although La3Al3Si3O12N2 was also detected in the La-involved ternary library, we ignored it because it was inferior to the Gd-based S phase. Accordingly, further efforts were made to improve the luminescent property of only the S phase. The major concern is to constitute a single-phase S compound activated with Ce3+ prior to the improvement in luminescence. As shown in Figure 1, all of the S phases detected in the combi-chem library are not a single phase but involve a considerable amount of impurities. We adopted PSOCMS in order to reduce the impurity content and finally realize a nearly pure S phase activated with Ce3+. More details on the PSOCMS will be discussed in the following subsection. 3.2. PSOCMS Process. PSOCMS has proven to be versatile in the discovery of novel phosphors.1−4 In fact, it has been used for fine-tuning as an auxiliary tool, following a preliminary rough screening, which is implemented by the assistance of a nondominated sorting genetic algorithm (NSGA or NSGA-II) along with the parametrization of the novelty of samples.1,2 However, we did not employ a nondominated sorting genetic algorithm-assisted combinatorial material search (NSGACMS) in the present investigation. Instead, a preliminary screening process based on the ternary combinatorial libraries was implemented in advance of the PSOCMS. The search space adopted in the present investigation was so small that a simple ternary combinatorial library screening was sufficient. PSOCMS proved to be efficient in fine-tuning the discovered materials and thereby in pinpointing the final decision in the reduced search space. The behavior of a swarm in nature is characterized by a social behavior, which is a motivation of PSO. Every individual in the swarm communicates with one another to achieve a common goal. PSO borrowed this concept, and the optimization process in PSO mimics the swarm behavior.19 Every phosphor with a specific composition located in the search space is regarded as an individual in the swarm. First, 18 compositions were
randomly determined, and they moved around the search space by a weighted sum of three velocity vectors. The first vector pointed at the current best composition acquired by the social communication between individuals in the swarm, which means that we had to compare all the phosphor samples in terms of a certain property (i.e., objective function) at a current stage and pinpoint the best one. The second velocity vector was toward the best composition acquired by the individual’s experience, indicating that every individual in the swarm memorized its own best composition in all the past trajectories that it had ever experienced. Finally, the last velocity vector is an inertia term. More details on the fundamentals of PSO are well described in the literature.19 Our optimization problem was relatively small (i.e., no more than a quaternary system where several numbers of iterations would be sufficient to achieve a convergence). It should be noted that the objective function evaluation was based on the actual sample synthesis and characterization in the case of the PSOCMS iteration, which contradicts conventional computation-based PSO iterations. Therefore, it is inevitable that the iteration number must be reduced to a minimal level for economical reasons. In fact, we iterated five swarms only in the PSOCMS process. More details about the computation and experiment for the PSOCMS process are well described in our previous reports.1−4 Among the metaheuristics developed thus far, PSO has proven to work best in a continuous parameter space. In this respect, the chemical composition search space that we adopted for our PSOCMS process was continuous. The search space for PSOCMS was significantly reduced by excluding all the other lanthanides leaving only Gd. The search space for PSOCMS was located in the vicinity of the composition area where we detected the S phase in the preliminary screening. Namely, the Gd2O3−Al2O3−Si3N4 ternary space adopted for the PSOCMS process centered on the Gd3Al3Si3O12N2 stoichiometry, which is the center of gravity of the Gd2O3−Al2O3−Si3N4 ternary diagram. The total area of the ternary space was shifted to the Al2O3 side and slightly enlarged by comparison with the previous ternary space, where the preliminary screening was implemented. Such an alteration of the search space during fine-tuning was based on the fact that the S phase was detected mostly in the left bottom side of the previous search space used for the preliminary screening. By introducing the Ce3+ activator variation on the top of the Gd2O3−Al2O3−Si3N4 ternary space, G
DOI: 10.1021/ic502721h Inorg. Chem. XXXX, XXX, XXX−XXX
Article
Inorganic Chemistry
Figure 8. Crystal structure of Gd3Al3+xSi3−x,O12+xN2−x viewed along (a) [001] and (b) [010] directions. The Al/Si tetrahedra are highlighted in blue and the Al octahedra in green color.
we finally built up a Gd2O3−Al2O3−Si3N4−Ce2O3 quaternary system as a final search space for the PSOCMS process. Figure 2 shows the composition search space adopted for the PSOCMS process, which looks like a prismatic shape. The peak height at 29.3° in the XRD data of the S phase was adopted as a single objective function in the PSOCMS process. If the structure of the S phase were identical to Ln3Al3Si3O12N2 (Ln = La or Nd), this peak index would be (111). However, it might have been controversial to define the discovered phosphor host structure as a well-known Ln3Al3Si3O12N2 (Ln = La or Nd), so we presented only the 2θ angle and not the peak index. In contrast to our previous PSOCMS cases, wherein only the PL intensity had been used as an objective
function,1−4 the XRD peak intensity was introduced as an objective function for the first time in the present investigation. This is analogous to the field trend whereby the phase purity is more important than the luminescent property from the viewpoint of the IP issue in industry. Luminescent properties such as PL efficacy and color chromaticity could be tailored and also improved after the phase identification was first secured. It is obvious that further experimental efforts to obtain a pure single phase of the identified phosphor will give rise to enhancement in the luminescent properties. According to the development history of a majority of commercially available LED phosphors, the effort to secure a novel single-phase compound has always preceded the introduction of an activator H
DOI: 10.1021/ic502721h Inorg. Chem. XXXX, XXX, XXX−XXX
Article
Inorganic Chemistry
An S phase phosphor activated with Ce3+ would not be welcome as the practical choice for blue chip-based LED applications, but it would be suitable for UV chip-based LED applications. The S phase phosphor exhibits a relatively wide emission spectrum compared with other conventional blue phosphors. This phosphor would be favorable for UV LEDbased general lighting applications if improvements were made to increase UV chip efficacy. An exact measurement of internal (or external) quantum efficiency was beyond the scope of the present investigation. The major concern of the present investigation was a determination of the exact structure needed to ensure novelty in a real sense. If conventional powder processing efforts were realized in industry, pure single-phase powder samples could be achieved with a desirable size distribution, and quantum efficiency would improve to a commercial level. This should be an industry goal. 3.4. Structural Properties of the Discovered Phosphor. As discussed in the preceding section, during the preliminary investigation, the powder XRD pattern of Gd3Al3+xSi3−xO12+xN2−x:Ce3+ (with 0 < x
