Effect of Synthesis Conditions on the Photoluminescent Properties of Si-Substituted CaYAlO₄:Eu: Sources of Experimental Errors in Solid-State Synthesis

Abstract

To improve the luminescent efficiency of and to design the color spectrum of phosphors, the comprehensive understanding of the correlation between physical parameters and luminescent properties is imperative, necessitating systematic experimental studies. However, unintentional variations across individually prepared samples impede the thorough investigation of the correlation. In this study, we investigate the possible sources of unintentional variation in the photoluminescence properties of phosphors during sample preparation using a solid-state reaction, explicitly focusing on the ball milling process. Based on the quantitative features of the photoluminescent properties and their associated statistical errors, we explore the impact of unintentional variation alongside intended systematic variation, highlighting its potential to obscure meaningful trends.

1. Introduction

Phosphor-converted white-light-emitting diodes (pc-wLEDs) have emerged as the next-generation lighting technology, promising enhanced efficiency and longevity compared to traditional lighting sources [1,2,3]. The most common method for WLED production involves utilizing blue LEDs as the primary light source and incorporating yellow phosphors to achieve white light emission. While this approach offers the advantage of simplicity in obtaining white light, it has several issues, including a high color temperature [4,5]. To address the issues in the conventional method, alternative approaches have been explored, including the integration of blue, green, and red phosphors [6,7,8] or the development of single-component white phosphors containing multiple luminescent activators [9,10]. To improve luminescence efficiency as well as to achieve precise control over emission characteristics, the thorough understanding of the correlation between physical parameters and the luminescent properties of a phosphor is essential.

Beyond simply examining emission or excitation wavelengths, it is required to scrutinize subtle changes in both photoluminescent excitation and emission spectra, such as the peak intensity ratio of the charge transfer band (CTB) to f-f transition, Judd–Ofelt parameters, and asymmetry ratios (electric dipole transition to magnetic dipole transition) [11,12,13,14,15]. An intriguing aspect stems from the inconsistency observed in the literature regarding the ratio of CTB to f-f transition in PLE spectra of Eu-doped oxide phosphors [16]. This inconsistency, lacking a clear explanation, hints at unintentional variation inherent in the synthesis process, which likely affects the ratio of CTB to f-f transition. Oxygen vacancy is one of the most common and inevitable defects in oxides; thus, oxygen vacancy has been attributed to this ratio, yet a comprehensive understanding of its origin remains elusive [17,18]. A thorough grasp of the factors correlated with an experimental feature and a keen awareness of statistical error, such as standard deviation, are essential for drawing accurate conclusions.

In this study, we discuss the potential sources of unintentional change in the photoluminescence properties of phosphors prepared via solid-state reaction. By investigating the change in the structural and photoluminescent properties of Si-substituted CaYAlO₄:Eu phosphors under different sample preparation conditions, we aim to identify and address unintentional variations that may arise during the preparation process. Our findings highlight the importance of carefully designing experiments to avoid unintentional variation, thus contributing to a rigorous investigation for a comprehensive understanding of phosphors.

2. Results and Discussion

2.1. Grinding

We first discuss tolerance, defined as the permissible error, for analyzing the ratio of CTB to f-f transition. Figure 1a shows the photoluminescence excitation (PLE) spectra of two CYASO:Eu samples prepared individually (by two different individuals). The PLE spectra are normalized to the highest peak intensity in each spectrum (here, the intensity of the f-f transition at 395 nm) to exclude the error due to mass variation between measurements. Both samples exhibit different CTB intensities, consistent across repeated measurements (standard deviation ≈ 0.01). Typically, in sample preparation using a solid-state reaction, the grinding process is performed after heating to eliminate lumps and ensure homogeneity. We initially hypothesized that differences in grinding conditions (e.g., force, duration, concentration) by different individuals might account for the observed discrepancy. To check this, we further performed grinding processes for 10 min and 20 min for the same individual. The ratio slightly increases continuously as the grinding process is executed further, likely due to the reduction in oxygen vacancies on the particle surface during the grinding process [19,20]. Although the ratios slightly increase, the gap between the ratios of the two different samples almost persists (Figure 1b). This result suggests that unintentional variation can arise at any stage during sample preparation using a solid-state reaction. The difference observed in the two different samples is considered as the tolerance (~0.1) in analyzing the ratio of CTB to f-f transition, provided the grinding processes are consistently executed. Furthermore, we are not able to see a significant difference in the Judd–Ofelt parameters and asymmetric ratios (i.e., R-factor) across the samples (

Table 1. Optical transition strength parameters and asymmetric ratios (R-factors) of two CYASO:Eu samples prepared under the same conditions nominally.

Samples	Grinding Time (min)	Ω2 (10−20 cm2)	Ω4 (10−20 cm2)	R-Factor
As prepared #1	5	3.484	2.467	2.284
	10	3.448	2.367	2.260
	20	3.435	2.374	2.251
As prepared #2	5	3.498	2.466	2.294
	10	3.374	2.269	2.212
	20	3.428	2.352	2.249

), indicating that the local environment of Eu ions is insensitive to the grinding process (see Figure S1 and Table S1 in Supplementary Information for details). The ratio of CTB to f-f transition seems much more sensitive to the sample preparation than optical transition strength parameters. Hereinafter, we investigate the possible sources causing the change in PL(E) spectra while considering the tolerance discussed here.

2.2. Ball Milling

The sample preparation using a solid-state reaction may include ball milling. The ball milling process is used for grinding and mixing materials and often to reduce the particle size. We attempted to reduce the particle size and induce surface defects in Al₂O₃, one of the starting materials, to see their effects on the photoluminescent response. We used a planetary ball mill machine with a Teflon container (i.e., milling jar) for the ball milling process. Initially, we used zirconia (ZrO₂) balls, which are commonly used. Compared to the pristine Al₂O₃ powder, the widths of the XRD peaks become broader after a ball milling process for 3 h (Figure 2a), indicating reduced particle size. However, additional peaks corresponding to the ZrO₂ phase emerge. The Mors hardness values of Al₂O₃ (9.0) and ZrO₂ (6.5) suggest possible scraps from the ZrO₂ ball. The impurity phase, such as Fe or Ni, due to the ball milling process with stainless steel balls has been investigated [21]. However, detecting insignificant contamination from ceramic balls might be very challenging, especially after the entire solid-state reaction process. Hence, careful consideration should be given when selecting materials for ball milling processes.

Using Al₂O₃ balls could resolve this issue. However, attention must also be paid to scraps that peel off from the container. While the ball milling process with Al₂O₃ balls excludes the ZrO₂ contamination, an additional peak at ≈22.93° persists irrespective of the kind of balls (Figure 2a). This peak is unrelated to ZrO₂ or both α- and γ-Al₂O₃ phases. Heat treatment at 800 °C eliminates this peak (Figure 2b), implying that the phase might be organic scraps from the Teflon container. Since solid-state reaction includes a high-temperature heating process, one may think that the organic scraps are not an issue. However, it can create a significant issue related to off-stoichiometry if the ball milling process is used for precursors individually.

As previously mentioned, we aimed to reduce the particle size and induce surface defects in Al₂O₃, one of the starting materials, to examine its influence on the photoluminescent properties of the final compound, CYASO:Eu. We performed the ball milling process on Al₂O₃ for 1 to 10 h and checked the change in the width of XRD peaks (Figure 3a). The peak width becomes broader after 1 h of ball milling, indicative of the reduced particle size. However, additional milling does not significantly alter the peak width. Using the Scherrer formula, we estimated the particle sizes from the XRD peak width. The particle size reduces from 40.0 nm to 13.5 nm after 1 h of ball milling (the inset of Figure 3a).

We noticed that the total mass of powders increases after the ball milling compared to the mass of initially loaded Al₂O₃ powder, and the total mass was proportional to the milling time. We attribute the increased mass after the ball milling to the peeled-off scraps from the Teflon container. A problem may arise when we take and use some portion of the ball-milled powder for subsequent processes, and if we need to mix powders by considering the stoichiometry of a final compound, we are not able to determine the mass of ball-milled powder for mixing because chemical formula or mixture ratio of ball-milled powder is unknown. Such an issue becomes evident in the subsequent experimental results. While the overall mass of the powder increased after milling, we assumed that we did not recognize the mass increase, and we took only the required mass of Al₂O₃ for the stoichiometric CYASO synthesis. As milling time increases, so does mass; hence, taking the equal masses of the precursors prepared under different milling times causes adding different moles of a specific element, potentially leading to defective samples. In other words, although one intended to systematically prepare samples using different Al₂O₃ subjected to different milling times, the samples would be prepared with different Al contents. Figure 3b shows the XRD patterns of the CYASO:Eu prepared using the Al₂O₃ powders subjected to different milling times. No significant difference is observed for the CYASO phase, but the peak intensity of the secondary phase, Y₂O₃, slightly increases as the ball milling time increases. The formation of secondary phases might seem correlated with the particle size of Al₂O₃ or simply ball milling time. The PL(E) spectra of the CYASO:Eu samples are represented in Figure 4a. The overall intensity of both PLE and PL spectra decreases, and the ratio of CTB to f-f transition increases as the ball milling time increases (Figure 4b), while no meaningful differences are observed for the optical transition strength parameters and R-factors within the tolerance (

Table 2. Optical transition strength parameters and asymmetric ratios (R-factors) of the Si-substituted CaYAlO₄:Eu (CYASO:Eu) samples prepared with the Al₂O₃ powders subjected to the different ball milling times.

Ball Milling Time	Ω2 (10−20 cm2)	Ω4 (10−20 cm2)	R-Factor
10 h	3.474	2.531	2.295
5 h	3.483	2.550	2.300
3 h	3.493	2.614	2.307
1 h	3.463	2.531	2.287
Pristine	3.498	2.466	2.294

). The photoluminescent properties might also seem correlated with the ball milling time. This is wrong—the observed changes in both structural and optical measurements are due to off-stoichiometry. This will be evident in the next section. Many sources cause contamination in the ball milling process, and they can lead to misleading or incorrect interpretations.

2.3. Precursors

To see the effect of different Al precursors on the photoluminescent properties of CYASO:Eu while avoiding potential off-stoichiometry issues, we used entire powders after a specific procedure to manipulate the precursors. We initially determined the masses of the starting materials considering the stoichiometry and performed the ball milling process on Al₂O₃ powder. In the ball milling, the overall mass of the powder increases because of the organic scraps from the container, while the mole number of Al remains consistent. To avoid any loss of the ball-milled powders, we did not collect the powders from the container, and we put the other starting materials into the container and performed the mixing procedure. We applied a similar strategy for preparing Al precursor from the solution of Al(NO₃)₃·9H₂O (aluminum nitrate nonahydrate); the Al precursor obtained by this method was expected to have very small grain and fine particle size. The mass of Al(NO₃)₃·9H₂O powder was first determined considering the stoichiometry, and the powder was dissolved in distilled water. The solution was poured into a milling jar and dried using an oven. The Al precursor prepared by drying the solution was mixed with the other starting materials. The XRD patterns of the Al precursors used in this study are shown in Figure 5a. The pristine Al(NO₃)₃·9H₂O powder was also used as one of the Al precursors for comparison. It can be seen that the ball-milled Al₂O₃ powder exhibits broader XRD peak widths, indicating reduced particle size, and the powder obtained by drying the Al(NO₃)₃·9H₂O solution does not show any diffraction peaks, as expected for amorphous or very small grain powders.

The XRD patterns of the CYASO:Eu prepared with the different Al precursors are presented in Figure 5b. First, we note the very similar diffraction patterns regardless of the ball milling of Al₂O₃ powder. This result attests that the previous results observed in Figure 3b and Figure 4 are indeed due to off-stoichiometry. Notably, the samples prepared with Al(NO₃)₃·9H₂O show the presence of an additional secondary phase, Ca₂Al₂SiO₇, compared to the sample prepared with the Al₂O₃ powders. No significant difference is observed for the CYAO phase.

The PL(E) spectra of CYASO:Eu prepared with the different Al precursors are presented in Figure 6a. The overall intensities seem dependent on the precursors, but the quantum efficiencies of the samples were found to be similar; we thus attribute the difference to the mass difference used for each measurement. The difference in the ratio of CTB to f-f transition is insignificant considering the tolerance as discussed above (Figure 6b). The optical transition strength parameters and R-factors of the CYASO:Eu samples are listed in

Table 3. Optical transition strength parameters and asymmetric ratios (R-factors) of the Si-substituted CaYAlO₄:Eu (CYASO:Eu) samples prepared with the different Al precursors.

Precursors		Ω2 (10−20 cm2)	Ω4 (10−20 cm2)	R-Factor
Al2O3	Commercial	3.498	2.466	2.294
	Ball-milled	3.502	2.415	2.298
Al(NO3)3·9H2O	Commercial	3.472	2.414	2.282
	Solution	3.355	2.121	2.225

. Except for the sample prepared with the Al(NO₃)₃·9H₂O solution, all samples exhibit very similar optical strength parameters and R-factors. Ω₂ parameter depends on the local crystal environment of rare earth ion sites; Ω₄ and Ω₆ are related to the viscosity and rigidity of a host matrix. The CYASO:Eu sample prepared with the Al(NO₃)₃·9H₂O solution exhibits the considerable formation of secondary phases compared to the other samples. The formation of secondary phases may affect both stoichiometry and defect levels of a desired compound, and/or directly host the luminescent activator with a different local structure and environment. Further study is required to elucidate the effect of precursors on the luminescent properties. This study warns of possible misleading due to unintentional variation atop intended systematic variation. We need to pay attention when designing sample preparation protocols involving systematic variation to minimize unintentional variation across individual sample preparations. Combinatorial sample preparation with the slogan “many at a time” serves as a valuable approach not only for conducting high-throughput experiments but also for facilitating systematic studies while excluding unintended variation [22,23,24,25,26].

3. Experimental Methods

Sample preparation: Samples with a composition of Ca_1.15Y_0.8Al_0.8Si_0.2O₄:Eu_0.05 (CYAGO:Eu) were prepared through a solid-state reaction method using CaCO₃ (99.95%, Alfa Aesar, USA), Y₂O₃ (99.99%, Sigma Aldrich, St. Louis, MO, USA), SiO₂ (99.998%, Acros Organics, Waltham, MA, USA), and Eu₂O₃ (99.99%, Alfa Aesar, Ward Hill, MA, USA); as Al precursors, Al₂O₃ (99.95%, Sigma Aldrich, St. Louis, MO, USA) and Al(NO₃)₃·9H₂O (99.99%, Sigma Aldrich, St. Louis, MO, USA) were used (see the main text for details). The mixtures of the starting materials were subjected to a planetary ball milling process using ZrO₂ or Al₂O₃ balls for 3 h and subsequently heated at 1400 °C using a box furnace. After the synthesis process, the obtained powders were ground in an agate mortar with a pestle for 10 min to remove lumps and ensure homogeneity.

Characterizations: X-ray diffraction (XRD) measurements were carried out using an Ultima IV (Rigaku, Tokyo, Japan) with Cu Kα₁ = 1.5406 Å. The diffraction patterns were collected in the 2θ range of 10–80° with a step size of 0.02°. Photoluminescence (PL) and PL excitation (PLE) measurements were performed using a Photon Technology International (PTI) spectrophotometer equipped with a 60 W Xe-arc lamp.

4. Conclusions

When preparing phosphor samples using a solid-state reaction, maintaining consistency across multiple sample preparations for systematic studies can be challenging due to the numerous steps involved. Inconsistent grinding processes can lead to unintentional changes in photoluminescence spectra, likely due to the oxidation of particle surfaces. Additionally, the possible contamination from ball and container scraps during the ball milling process may result in mass change and significant off-stoichiometry issues if the process is carried out for precursors individually. These issues can be addressed by implementing carefully designed sample preparation protocols, which enable systematic investigation, like the present study on the effect of different Al precursors. This study highlights the potential for unintentional variation to obscure systematic variation and emphasizes the importance of meticulous sample preparation, awareness of potential sources of error, and the determination of statistical error bars in systematic studies.