Luminescence Intensity Ratio and Principal Component Analysis-Assisted Thermometry in Pr³⁺-Activated Inorganic Hosts

Abstract

Temperature-dependent luminescence of Pr³⁺-doped materials was investigated using both conventional luminescence intensity ratio (LIR) and principal component analysis (PCA)-based thermometry. Three host matrices with distinct structural properties, LiLaP₄O₁₂, YNbO₄, and Y₂O₃, were selected to evaluate the influence of crystal structure on thermometric performance. Temperature-resolved emission spectra recorded over the 103–523 K (−170 to 250 °C) range were analyzed using both approaches, with the first principal component (PC₁) serving as a thermometric parameter in the PCA. The results show that crystal symmetry and site multiplicity strongly influence the temperature-dependent spectral evolution and, consequently, the thermometric response. LiLaP₄O₁₂ exhibits stable and well-defined spectral evolution, resulting in balanced thermometric accuracy and resolution. YNbO₄ shows enhanced sensitivity to temperature variations due to increased spectral complexity and stronger crystal-field effects, leading to improved resolution but increased calibration uncertainty. In contrast, Y₂O₃ exhibits reduced thermometric performance due to overlapping emissions from multiple crystallographically inequivalent sites with distinct thermal responses. Compared to LIR, PCA provides improved thermometric figures of merit, particularly in systems with complex and strongly overlapping emission bands, demonstrating the potential of full-spectrum analysis in luminescence thermometry.

1. Introduction

Rare-earth-doped oxides are widely used in photonic materials due to their chemical stability and characteristic optical properties arising from 4f electronic states [1,2]. Among lanthanide ions, praseodymium (Pr³⁺) is particularly interesting due to its complex electronic structure, which enables both interconfigurational (4f¹5d¹ → 4f²) and intraconfigurational (4f² → 4f²) transitions. As a result, Pr³⁺-activated materials exhibit emission spanning the ultraviolet, visible, and near-infrared spectral regions, making them suitable for applications in lighting, scintillation, sensing, and bioimaging [1,3,4,5,6]. The luminescence properties of Pr³⁺ ions strongly depend on the host lattice. When incorporated into a solid, the energy separation between the 4f¹5d¹ and 4f² configurations is modified by crystal-field effects, covalency, and the local coordination environment [7]. Parameters such as site symmetry, phonon energy, ligand polarizability, and metal–ligand distances determine whether broad 4f–5d emission or narrow 4f–4f transitions dominate. These host-dependent effects have been observed in a wide range of materials, including phosphates, oxides, and fluorides [8,9,10,11]. In addition to radiative transitions, the emission efficiency and spectral shape of Pr³⁺-doped materials are accompanied by several nonradiative processes. Multiphonon relaxation plays a critical role, particularly in hosts with high phonon energies, where excited states such as the higher-lying 5d levels can be efficiently depopulated before radiative emission [7,12]. This behavior follows the energy gap law, according to which nonradiative decay rates increase with decreasing energy separation between electronic states and increasing lattice vibrational energy [13]. Furthermore, energy transfer between neighboring Pr³⁺ ions becomes more significant at higher dopant concentrations, leading to concentration quenching via dipole–dipole interactions [8,12]. Cross-relaxation processes involving intermediate excited states, such as ³P₀ and ¹D₂, can also enhance nonradiative decay channels, thereby further modifying the observed emission spectra [13].

The complexity of Pr³⁺ emission, arising from multiple overlapping transitions originating from different excited states (e.g., ³P₀, ¹D₂, and 4f¹5d¹ levels), presents a challenge for quantitative spectral analysis. Conventional approaches based on selected spectral features, such as intensity ratios or peak positions, are widely used in luminescence thermometry for their simplicity and ease of implementation. However, in spectrally complex systems containing multiple overlapping emission bands, these approaches may not fully utilize the temperature-dependent information distributed across the entire emission profile. Temperature-dependent luminescence forms the basis of optical thermometry, a remote sensing approach that offers high spatial resolution, rapid response, and applicability in environments where conventional non-remote methods are impractical [14,15,16]. In such systems, temperature-induced changes in emission spectra arise from thermally activated population redistribution, nonradiative quenching, and modifications of electron–phonon coupling. These effects are typically manifested across the full emission spectrum rather than in isolated spectral features.

As noted above, the presence of multiple overlapping emission bands complicates the extraction of reliable temperature information and limits the applicability of approaches that rely on isolated spectral features [4,6,13]. This limitation becomes particularly important in oxide hosts, where broad emission bands and pronounced spectral overlap frequently complicate conventional spectral analysis. Therefore, a significant portion of temperature-dependent information contained in the emission spectrum may remain underutilized when only selected spectral features are considered. This is particularly critical for Pr³⁺-activated systems, where the spectral complexity intrinsically favors full-spectrum analysis approaches. Because of these limitations, methods based on analysis of the full emission spectrum have attracted considerable interest. Among them, principal component analysis (PCA) is a powerful multivariate technique that extracts temperature-dependent information from high-dimensional spectral datasets without requiring prior assumptions about specific transitions. By decomposing emission spectra into orthogonal components ranked by explained variance, PCA reduces the dimensionality of the spectral dataset while preserving the dominant temperature-dependent variations [14,15]. This makes PCA particularly suitable for Pr³⁺-activated systems, where multiple overlapping emission bands hinder conventional spectral analysis. In experimental conditions where temperature is the only changing parameter, it is typically the first principal component, defined by its loading vector, that encodes the temperature-dependent spectral evolution. Consequently, the corresponding PC1 score can be used directly as a thermometric parameter [16,17].

Owing to these methodological advantages, the application of PCA has rapidly expanded beyond basic dimensionality reduction to encompass full-spectrum profiling of complex datasets characterized by significant optical band overlap. Outside the domain of thermal sensing, multivariate full-spectrum analysis is well established as a standard methodology in analytical chemistry, chemometrics, and various optical spectroscopies, including Raman, near-infrared (NIR), and UV-Vis absorption, to resolve overlapping electronic transitions, deconvolve spectral shapes, and suppress baseline fluctuations while preserving spectroscopic information [18,19]. The mathematical rigor of decomposing high-dimensional spectral matrices into orthogonal loading vectors renders PCA broadly applicable for extracting subtle variations from complex fluorescent and luminescent systems [20]. This analytical trend has accelerated across modern optical sciences, where researchers extensively employ PCA-driven chemometrics and advanced peak-feature selection to handle complex, highly congested spectroscopic profiles [21,22,23].

The present study focuses on evaluating the advantages of PCA-based thermometry in Pr³⁺-activated materials with complex temperature-dependent spectral evolution, where conventional approaches based on selected spectral features may become limited. Despite these advantages, systematic studies of PCA-based thermometry across different host materials remain limited. Since host-dependent structural and vibrational properties strongly govern the emission behavior of Pr³⁺ ions, it is essential to understand how these factors influence the performance of multivariate thermometric approaches [24,25].

In this work, we investigate temperature-dependent luminescence of Pr³⁺-doped LiLaP₄O₁₂, YNbO₄, and Y₂O₃, three oxide hosts with distinct crystal symmetries and local coordination environments. LiLaP₄O₁₂ is a phosphate-based host characterized by well-defined Pr³⁺ emission features and a high phonon energy (1000–1300 cm⁻¹) due to the phosphate framework, which influences the radiative and non-radiative relaxation processes [25,26]. In contrast, YNbO₄ exhibits lower crystal symmetry and a distinct local coordination environment, leading to pronounced crystal-field splitting and complex emission behavior [27]. Finally, Y₂O₃ is a well-established cubic oxide with moderate-to-low phonon energy and high structural symmetry, commonly used as a reference system in rare-earth luminescence studies [9,28]. By combining these systems, we provide a comparative analysis of how host lattice properties influence both emission characteristics and PCA-based thermometric performance. Temperature-dependent emission spectra are analyzed using both conventional luminescence intensity ratio (LIR) and principal component analysis (PCA) approaches to compare their thermometric performance and evaluate the influence of the host structure on temperature-dependent spectral behavior in Pr³⁺-activated materials.

2. Results and Discussion

2.1. Structural Characterization and Crystal Structure

The phase composition of the synthesized Pr³⁺-doped LiLaP₄O₁₂, YNbO₄, and Y₂O₃ powders was examined by X-ray diffraction, and the corresponding patterns are shown in Figure 1a–c. All diffraction peaks can be indexed to the expected crystal structures of the respective host materials, with no detectable impurity phases within the limits of the XRD technique, confirming the successful formation of single-phase materials. The diffraction pattern of LiLaP₄O₁₂ (Figure 1a) corresponds to a monoclinic structure (space group C2/c) and agrees well with reported data for tetraphosphate compounds. YNbO₄ (Figure 1b) crystallizes in a monoclinic fergusonite-type structure (space group C2/c), with all reflections matching the reference pattern, indicating high phase purity. In contrast, Y₂O₃ (Figure 1c) adopts a cubic bixbyite structure (space group Ia-3), with no evidence of secondary phases. Due to the low dopant concentration (1 mol%), no measurable peak shifts are observed, indicating that incorporation of Pr³⁺ ions does not significantly perturb the host lattice within the resolution of XRD. These structural differences, particularly in symmetry and site multiplicity, are expected to play a key role in determining the luminescence behavior and thermometric performance of the investigated systems. To interpret the differences in luminescence behavior, the crystal structures and local coordination environments of the host materials are taken into account. LiLaP₄O₁₂ consists of interconnected PO₄ tetrahedra and LaO₈ polyhedra, with Li⁺ ions occupying interstitial sites. In this lattice, the La³⁺ ions occupy the special 4e Wyckoff position, which is characterized by a C₂ (2) local site symmetry. The La³⁺ ions are eightfold coordinated (CN = 8), forming distorted oxygen polyhedra. YNbO₄ is composed of NbO₄ tetrahedra and YO₈ polyhedra forming a three-dimensional framework. Similar to the phosphate host, the Y³⁺ ions in YNbO₄ occupy the 4e Wyckoff position with C₂ (2) site symmetry, where they are coordinated by eight oxygen atoms (CN = 8) in a distorted environment characteristic of low-symmetry crystal fields. In contrast, Y₂O₃ has two crystallographically distinct cation sites. In this lattice, all Y³⁺ ions are sixfold coordinated (CN = 6), being distributed between the centrosymmetric 8b position with C_3i ($\overline{3}$) local symmetry and the non-centrosymmetric 24d position with C₂ (2) local symmetry, present in a 1:3 abundance ratio. These sites correspond to distinct configurations of distorted octahedral environments resulting from different body-diagonal oxygen-vacancy arrangements. These differences in coordination number, site symmetry, and lattice connectivity are expected to influence the crystal field splitting significantly and, consequently, the temperature-dependent luminescence behavior of Pr³⁺ ions.

2.2. LIR-Based Thermometric Analysis

For both LIR and PCA thermometric analyses, measurements were performed over the temperature range 103–523 K in 10 K intervals, with 100 emission spectra recorded at each temperature. All spectra were corrected for detector background and subsequently area-normalized to minimize the effects of intensity fluctuations due to instrumental or geometric variations. For each temperature point, the dataset was randomly divided into two equal subsets. One subset (50 spectra per temperature) was used as the training dataset to determine both the LIR values and the PCA principal component loading vectors and values, which were used to construct the temperature-dependent calibration points. The second subset (the remaining 50 spectra) was used as the testing dataset to evaluate the predictive performance of both methods. This approach ensures a statistically robust determination of each method’s figures of merit: accuracy and precision.

The per-temperature averaged emission spectra of all investigated Pr³⁺-doped materials are shown in Figure 2a,d,g. In all systems, increasing temperature leads to a gradual redistribution of spectral intensity from lower-energy to higher-energy spectral regions. Due to the complex nature of the Pr³⁺ emission spectrum, a conventional Boltzmann-type LIR approach based on selected individual transitions was not possible; the LIR analysis was performed using integrated spectral regions. The LIR parameter was defined as the ratio of the integrated intensity at higher energies (I_HI) to that at lower energies (I_LOW), where the threshold wavelength (λ_tr) separating the two spectral regions was determined from the intensity inversion point of the area-normalized, temperature-dependent spectra (Equation (1)). The obtained λ_tr values were 600 nm for LiLaP₄O₁₂:Pr³⁺, 620 nm for YNbO₄:Pr³⁺, and 616 nm for Y₂O₃:Pr³⁺.

Figure 2. Temperature-dependent luminescence and LIR-based thermometric analysis of Pr³⁺-doped (a–c) LiLaP₄O₁₂, (d–f) YNbO₄, and (g–i) Y₂O₃ in the 103–523 K temperature range. (a,d,g) Area-normalized emission spectra showing temperature-induced spectral redistribution. (b,e,h) Temperature dependence of the luminescence intensity ratio (LIR) calibration points. (c,f,i) Thermometric performance expressed as absolute temperature uncertainty (ΔT_LIR) and temperature resolution (δT_LIR).

Figure 2. Temperature-dependent luminescence and LIR-based thermometric analysis of Pr³⁺-doped (a–c) LiLaP₄O₁₂, (d–f) YNbO₄, and (g–i) Y₂O₃ in the 103–523 K temperature range. (a,d,g) Area-normalized emission spectra showing temperature-induced spectral redistribution. (b,e,h) Temperature dependence of the luminescence intensity ratio (LIR) calibration points. (c,f,i) Thermometric performance expressed as absolute temperature uncertainty (ΔT_LIR) and temperature resolution (δT_LIR).

$$LIR\left(T\right)={\displaystyle \frac{I_{HI}\left(T\right)}{I_{LOW}\left(T\right)}}={\displaystyle \frac{{\sum }_{\lambda <{\lambda }_{tr}}I(\lambda ,T)}{{\sum }_{\lambda \ge {\lambda }_{tr}}I(\lambda ,T)}}$$

(1)

The temperature dependences of the averaged LIR values obtained from the training dataset at each stabilized temperature are presented in Figure 2b,e,h. Thermometric accuracy and resolution were evaluated using testing spectra that were not included in the calibration procedure. At each stabilized temperature, the thermometric accuracy was defined as the absolute difference between the mean of the predicted temperature distribution and the stabilized temperature, ΔT(T) = |T_av(T) − T|, while the thermometric resolution (precision) was determined from the standard deviation of the predicted temperature distribution, δT(T) = σ(T). The corresponding temperature-dependent accuracy and resolution plots are shown in Figure 2c,f,i.

LiLaP₄O₁₂:Pr³⁺ exhibits a broad emission band centered around 590 nm arising from overlapping ³P₀,₁ → ³H₆ and ¹D₂ → ³H₄ transitions. With increasing temperature, gradual redistribution of spectral intensity towards higher-energy components is observed, consistent with thermal population redistribution between the ³P₀ and ³P₁ excited states. No detectable spectral shift in the peak maximum is observed, indicating stable crystal-field splitting in this phosphate host. The corresponding LIR analysis (Figure 2b,c) demonstrates high thermometric accuracy and resolution over the investigated temperature range. Compared to LiLaP₄O₁₂:Pr³⁺, YNbO₄:Pr³⁺ exhibits a significantly more structured emission spectrum with multiple resolved Stark components in the 580–670 nm region, originating from transitions involving the ³P₀, ³P₁, and ¹D₂ excited states. The stronger crystal-field splitting characteristic of the low-symmetry fergusonite lattice results in pronounced temperature-dependent spectral variations and progressive redistribution of emission intensity with increasing temperature. The corresponding LIR analysis (Figure 2e,f) shows improved temperature resolution compared with LiLaP₄O₁₂, reflecting the enhanced sensitivity of the spectral profile to temperature-induced changes. Among the investigated systems, Y₂O₃:Pr³⁺ exhibits the most complex emission behavior, characterized by broad, strongly overlapping bands extending from approximately 550 to 710 nm. This complexity originates from the coexistence of two crystallographically inequivalent cation sites (C₂ and S₆ symmetry), each contributing distinct emission transitions with different thermal responses. Increasing temperature leads not only to emission quenching but also to changes in the relative contributions of the two emitting sites. As a result, the corresponding LIR analysis (Figure 2h,i) exhibits significantly reduced thermometric performance compared to the other investigated systems.

2.3. PCA-Based Thermometric Framework

To better extract thermometric information from the full spectral profile, temperature-dependent emission spectra were analyzed using principal component analysis (PCA). PCA is an unsupervised multivariate method that transforms a set of correlated spectral variables into a new set of orthogonal components (principal components, PCs) ranked in descending order according to the amount of explained variance. In this approach, the normalized training spectra were organized into a data matrix M of size n × p. In this matrix, each row represents a single normalized spectrum, and each column corresponds to a specific wavelength channel. The covariance matrix C = MᵀM/(n − 1) is computed and analyzed via eigendecomposition, yielding p orthogonal eigenvectors and their associated eigenvalues. These eigenvectors, called loading vectors (indicated as ${PC}_i^{coeff}$ on Figure 3), define directions in p-dimensional hyperspace. Their corresponding eigenvalue ${\epsilon }_i$ represents the variance of the training set in this direction. The eigenvectors are sorted by descending eigenvalues so that the first component captures the greatest variance. The relative variance in the i-th direction is measured using ${\epsilon }_i/{\sum }_i{\epsilon }_i$ (expressed as a percentage in Figure 3), where ${\sum }_i{\epsilon }_i$ represents the total explained variance. A scalar vector product of any normalized spectrum and the corresponding ${PC}_i^{coeff}$ yields the principal component value ${PC}_i$ for that spectrum, where i = 1 is the most significant one [29].

In all cases presented here, temperature was the sole variable during spectra acquisition. Consequently, temperature-induced spectral changes represent the dominant source of variance in the training dataset. For all investigated phosphors, the first principal component (PC₁) accounts for more than 97.5% of the total spectral variance. This reflects the primary temperature-dependent spectral variations. Because this value exceeds the typical 95% threshold, it indicates that temperature-induced spectral evolution can be effectively described by a single parameter: the value of PC₁ (Figure 3). Conversely, higher-order components (PC₂, PC₃, etc.) account for minor spectral variations caused by alternative influences. These include noise, environmental fluctuations, and other secondary effects beyond temperature changes. While these components can be incorporated through multiparameter analysis, this introduces additional non-temperature sources of variance which may not necessarily improve thermometric performance.

Therefore, each training spectrum was projected onto PC₁, and the corresponding score values were used as a thermometric variable to form the calibration values PC₁(T) (Figure 4a,c,e). Subsequently, a test set was used to reproduce the stabilized temperatures (50 spectra at each stabilized temperature) by linear interpolation between the nearest calibration values. Using the testing dataset, thermometric performance was evaluated based on the accuracy (ΔT_PC1) and resolution (δT_PC1) criteria, defined in the same way as for the LIR analysis. The corresponding temperature-dependent thermometric parameters obtained from the PCA-based prediction are shown in Figure 4b,d,f.

The temperature-dependent spectral changes discussed above are directly reflected in the corresponding PC₁(T) dependences obtained by principal component analysis (Figure 4). In all investigated systems, the first principal component captures the dominant temperature-dependent spectral variance and enables thermometric readout from the full emission spectrum. For LiLaP₄O₁₂:Pr³⁺, the gradual redistribution of spectral intensity with increasing temperature results in a smooth and monotonic PC₁(T) dependence (Figure 4a), enabling reliable temperature calibration. The extracted thermometric parameters are ΔT_PC1 = 0.21 K and δT_PC1 = 0.37 K (Figure 4b), demonstrating high accuracy and resolution over a wide temperature range. This performance can be attributed to the relatively simple emission profile and stable temperature-dependent spectral evolution observed in this phosphate host. YNbO₄:Pr³⁺ exhibits a more structured emission profile and more pronounced temperature-induced spectral variations, which are reflected in the corresponding PC₁(T) dependence (Figure 4c). The obtained dependence is monotonic but slightly non-linear, including a sign change at intermediate temperatures. This effect arises from the mathematical definition of principal components and does not correspond to a physical inversion of emission intensity. Based on the predicted temperatures from the testing dataset, the average values of statistically obtained accuracy and resolution are ΔT_PC1 = 0.12 K and δT_PC1 = 0.20 K, respectively (Figure 4d). The improved temperature accuracy and resolution compared to LiLaP₄O₁₂ reflect the enhanced sensitivity of the emission spectrum to temperature-induced changes. Among the investigated systems, Y₂O₃:Pr³⁺ exhibits the most complex emission behavior due to the coexistence of two crystallographically inequivalent emitting sites with distinct thermal responses. The corresponding PC₁(T) dependence appears quasi-linear across the investigated temperature range (Figure 4e). Despite this apparent simplicity, the thermometric performance is reduced, with ΔT_PC1 = 0.80 K and δT_PC1 = 3.31 K (Figure 4f). The coexistence of multiple emitting centers introduces additional spectral variability that cannot be fully described by a single principal component, leading to increased uncertainty and reduced resolution. Nevertheless, the results demonstrate that PCA-based thermometry remains applicable even in systems with pronounced spectral overlap and complex multi-site emission behavior.

2.4. Comparative Discussion

The results summarized in

Table 1. Summary of thermometric parameters obtained from LIR and PC₁ analysis for Pr³⁺-doped host materials.

	LIR-Based Thermometry		PCA-Based Thermometry
Host Matrix	ΔTLIR (K)	δTLIR (K)	ΔTPC1 (K)	δTPC1 (K)	PC1(T) Behavior
LiLaP4O12:Pr3+	0.24	0.48	0.21	0.37	Non-linear
YNbO4:Pr3+	0.17	0.43	0.12	0.20	Non-linear
Y2O3:Pr3+	1.50	3.61	0.80	3.31	Quasi-linear

show that the thermometric performance of both LIR- and PCA-based approaches strongly depends on the complexity of the temperature-dependent spectral changes. For LiLaP₄O₁₂:Pr³⁺, which exhibits relatively simple and well-defined emission behavior, both methods provide high accuracy and resolution over a broad temperature range. The slightly better performance achieved with PCA suggests that analyzing the full emission spectrum enables more efficient extraction of temperature-dependent information than approaches that rely only on selected spectral features. A different behavior is observed for YNbO₄:Pr³⁺. In this system, the stronger crystal-field splitting produces a more structured emission spectrum and increases the sensitivity of the spectral profile to temperature changes. As a result, both approaches achieve higher resolution than LiLaP₄O₁₂. However, PCA yields noticeably lower ΔT and δT values than the corresponding LIR analysis, indicating that full-spectrum analysis becomes increasingly advantageous as spectral complexity increases. The slightly nonlinear PC₁(T) dependence is consistent with the increased complexity of the temperature-dependent spectral changes. The most challenging case is Y₂O₃:Pr³⁺, in which emissions from two crystallographically inequivalent sites strongly overlap and exhibit distinct quenching dynamics. This leads to reduced thermometric performance for both LIR and PCA approaches. Nevertheless, PCA remains applicable even in this highly complex system and provides significantly higher accuracy (47%) and slightly higher precision than the corresponding LIR analysis. The quasi-linear PC₁(T) dependence most likely reflects the averaged contribution of multiple emission centers with different thermal responses.

Overall, the comparison between LIR and PCA approaches shows that the advantage of PCA becomes more pronounced in systems with structured or strongly overlapping emission bands, where temperature-dependent information is distributed across the entire emission spectrum rather than localized in a few selected transitions. It is also important to note that the LIR approach used here relies on broad integrated spectral regions rather than isolated emission lines. Even under these conditions, PCA-based analysis provides improved thermometric performance, highlighting the advantage of utilizing more complete spectral information in luminescence thermometry. For all investigated materials, thermometric accuracy and resolution tend to deteriorate at elevated temperatures. This behavior is particularly pronounced for Y₂O₃:Pr³⁺ and reflects the increasing influence of thermal quenching and complex temperature-dependent spectral evolution. Consequently, the reliability of temperature read-out progressively decreases as the temperature approaches the upper end of the investigated range.

3. Experimental Part

3.1. Materials and Synthesis

To ensure a direct comparison of host-dependent luminescence behavior and thermometric performance, a fixed Pr³⁺ concentration of 1 mol% was used for all investigated materials. This concentration was selected to minimize concentration-dependent effects while enabling reliable evaluation of the host lattice’s influence on the performance of both LIR- and PCA-based thermometry.

3.1.1. Pr³⁺-Doped LiLaP₄O₁₂ (LiLaP₄O₁₂:1%Pr³⁺)

Pr³⁺-doped LiLaP₄O₁₂ (Li(La_0.99Pr_0.01)P₄O₁₂) was prepared via a coprecipitation method following previously reported procedures [25,30]. The starting precursors were La(NO₃)₃·6H₂O (Alfa Aesar, Ward Hill, MA, USA, 99.99%), (NH₄)₂HPO₄ (Thermo Scientific, Waltham, MA, USA, 99.8%), and Li₂CO₃ (Alfa Aesar, 99.99%). Pr₆O₁₁ (Alfa Aesar, 99.99%) was used as the dopant precursor following the previously reported synthesis procedure, while ascorbic acid (C₆H₈O₆, Merck, Darmstadt, Germany) was added to stabilize Pr³⁺ ions during synthesis. All reagents were dissolved in concentrated HNO₃ (Zorka, Sabac, Serbia, 65%) under constant stirring. The pH of the solution was adjusted to 8 with NH₄OH (Fisher, Polk County, MN, USA, 25%), resulting in a white precipitate. The obtained suspension was dried at 80 °C for several days. The dried precursor was then annealed at 450 °C for 8 h and allowed to cool naturally to room temperature.

3.1.2. Pr³⁺-Doped YNbO₄ (YNbO₄:1%Pr³⁺)

Pr³⁺-doped YNbO₄ (Y_0.99Pr_0.01NbO₄) was synthesized using a two-step mechanical activation-assisted solid-state reaction, as described in the literature [31,32,33]. The starting materials were Y₂O₃ (Alfa Aesar, Ward Hill, MA, USA, 99.9%), Nb₂O₅ (Alfa Aesar, Ward Hill, MA, USA, 99.5%), Pr₂O₃ (Thermo Scientific, Waltham, MA, USA 99.9%), and Na₂SO₄·10H₂O (Alfa Aesar, Ward Hill, MA, USA, 99%), which was used as a flux. Stoichiometric amounts of the precursors corresponding to 1 mol% Pr³⁺ were mixed with 10 mL of absolute ethanol and a few drops of H₂O₂ (Roth, Karlsruhe, Germany, 30%) to prevent oxidation of Pr³⁺. The mixture was mechanically activated in a planetary ball mill (Retsch PM100, Haan, Germany) at 100 rpm for 3 h, then dried at 70 °C. The dried powder was pre-sintered at 800 °C for 2 h, then ball-milled again under the same conditions, dried, and finally annealed at 1200 °C for 2 h. The resulting powder was washed multiple times with deionized water to remove residual flux and dried before further use.

3.1.3. Pr³⁺-Doped Y₂O₃ (Y₂O₃:1%Pr³⁺)

Pr³⁺-doped Y₂O₃ (Y_1.98Pr₀.₀₂O₃) was synthesized by the polymer complex solution (PCS) combustion method [34,35]. Y₂O₃ (Sigma–Aldrich, Burlington, MA, USA, 99.99%) and Pr₂O₃ (Thermo Scientific, 99.9%) were dissolved in concentrated HNO₃ (Superlab, Belgrade, Serbia, 69–71%) at 180 °C under continuous stirring until complete dissolution and evaporation to dryness. A small amount of H₂O₂ (Roth, Karlsruhe, Germany, 30%) was added during dissolution to stabilize Pr³⁺. Polyethylene glycol (PEG 200, Alfa Aesar, Ward Hill, MA, USA) was added as a fuel and complexing agent. The solution was heated at 100–130 °C under stirring until a homogeneous viscous gel formed. The gel was then heated to approximately 800 °C, initiating self-combustion and producing a fine powder. The obtained powder was calcined at 1100 °C for 24 h to improve crystallinity.

3.2. Instruments and Methods

X-ray diffraction (XRD) measurements were carried out using a Rigaku SmartLab diffractometer (Tokyo, Japan) with Cu Kα radiation (λ = 1.5406 Å), operating at 40 kV and 30 mA. Data were collected over a 2θ range of 10–90°. Luminescence measurements were performed using an Ocean FX UV–VIS spectrometer (Ocean Insight, Winter Park, FL, USA) coupled with a 450 nm solid-state laser as the excitation source via a Y-type optical fiber. The spectrometer operates in the 200–1000 nm wavelength range. A 500 nm long-pass filter was used to suppress the excitation signal in the detection branch. Powdered samples were pelletized and placed in a MicroOptik liquid-nitrogen cooling/heating stage equipped with a quartz window, allowing measurements over the temperature range of 80–700 K. After thermal stabilization, 100 spectra were recorded at each temperature.

4. Conclusions

In this work, temperature-dependent luminescence of Pr³⁺-doped LiLaP₄O₁₂, YNbO₄, and Y₂O₃ was analyzed using both luminescence intensity ratio (LIR) and principal component analysis (PCA)-based thermometry. The results show that PCA reliably extracts temperature information from full emission spectra, particularly in systems with complex, strongly overlapping emission bands. LiLaP₄O₁₂:Pr³⁺, characterized by relatively simple and stable spectral evolution, exhibited high thermometric accuracy and resolution for both approaches. In YNbO₄:Pr³⁺, the increased spectral complexity and stronger temperature sensitivity of the emission profile resulted in improved thermometric resolution, with PCA providing noticeably better performance than the corresponding LIR analysis. In contrast, Y₂O₃:Pr³⁺ showed significantly reduced thermometric performance due to overlapping emissions from multiple crystallographically inequivalent sites with different quenching dynamics. Nevertheless, PCA remained applicable even in this structurally complex system and continued to produce improved thermometric parameters compared to the LIR approach. The thermometric performance was found to strongly depend on the complexity of the emission behavior and the structural characteristics of the host material. The obtained results show that PCA-based thermometry becomes particularly advantageous in systems with structured or strongly overlapping emission bands, where temperature-dependent information is distributed across the entire emission spectrum rather than concentrated in a few selected transitions. Compared to the corresponding LIR analysis, PCA generally provides improved thermometric performance, highlighting the potential of multivariate full-spectrum analysis for luminescence thermometry of Pr³⁺-activated materials. Future studies will explore extending the operating temperature range beyond 523 K, assessing thermometric performance under high-temperature conditions where thermal quenching and spectral broadening become increasingly pronounced, and evaluating the long-term thermal stability and temperature-cycling reproducibility of the investigated materials. Further efforts will also focus on quantitative crystal-field analysis and modeling of non-radiative processes to better understand the relationship between host structure and thermometric performance.