1. Introduction

In the past decades, the trivalent praseodymium Pr³⁺ offering interesting perspectives as a multifunctional dopant has been intensively investigated. Wide applications have been found in holographic image recording [1], laser emitting [2], fluorescent tubes and scintillators [3,4]. Particular interest was given to the efficient characteristic red emission, with which Pr³⁺ could be applied to white-LEDs for high color rendering index on account of the common deficiency of red component [5]. Prominent red luminescence of high color purity via atypical quenching of blue-green emission had been observed under UV excitation in some Pr³⁺ doped materials [6,7], demonstrating the potential application in displaying and lighting. However, the red luminescence of Pr³⁺ is commonly unstable and inefficient at high temperature [8]. The thermal quenching behavior and low quenching temperature of red emission would prevent the application in high-power optical devices [9,10]. To interpret the blue-green emission quenching and the temperature dependent red emission, the Pr³⁺-trapped exciton [11] and the metal-to-metal intervalence charge transfer (IVCT) model [12] presenting a low-lying charge transfer state of Pr⁴⁺-M^(n-1)+ (Mⁿ⁺: d⁰ electron configured transition metal ions) had already been proposed, which was consistent with the natural tendency of Pr³⁺ to a tetravalent electronic configuration.

Lithium niobate (LiNbO₃) [13] is one of the most widely used multi-functional crystals due to the excellent optical properties and outstanding performance in integrated optics. An overview of the spectroscopic properties in Pr³⁺ doped LiNbO₃ (Pr:LiNbO₃) had been given by Lorenzo [14] in 1995. Except for applications in holographic recording and compact laser devices, efficient red-emitting piezoluminescence was achieved in Pr:LiNbO₃ [15] and the pressure effect on the luminescence properties had been investigated [11] recently. However, to our knowledge, despite the importance of thermal quenching, the temperature dependence of red luminescence in Pr³⁺ doped LiNbO₃ is less reported.

The traditional dopant Mg²⁺ is commonly introduced in congruent LiNbO₃ (CLN) at threshold (4.6mol% [16]) or exceeding concentration for improving the optical performance, including suppressing photo-damage and enhancing the fluorescence intensity [17,18]. For improving the widespread applications of LiNbO₃, 5mol% Mg²⁺ was introduced in the Pr³⁺ doped CLN in our previous work [19]; Long persistent luminescence was achieved in CLN and the essential role of Mg²⁺ incorporation had been demonstrated. However, the research on the influence of Mg²⁺ codoping on red luminescence in Pr:CLN is still insufficient. More importantly, the effect of Mg²⁺ codoping on thermal quenching behavior is worth investigating.

In this paper, based on the detailed spectroscopic characterization of Pr:CLN and Pr:Mg:CLN, the effect of Mg²⁺ codoping on the optical properties of Pr³⁺ in CLN was systematically analyzed. An in-depth analysis of the luminescence mechanism on the frame of the IVCT model was performed based on absorption and luminescence spectra, with which the enhanced red emissions via quenching of blue-green emission could be accounted for. To investigate the difference between the thermal quenching processes, the temperature dependent red emissions and lifetimes related to ¹D₂→³H₄ in Pr:CLN and Pr:Mg:CLN were studied in detail. The higher quenching temperature of red luminescence in Pr:Mg:CLN would be explained with the configuration coordinate diagram.

2. Experimental

The single crystals Pr:CLN and Pr:Mg:CLN of good quality, as shown in the insets of Fig. 1, were grown along the c axis by the Czochralski method. High-purity raw materials (>99.99%) were adopted here. The Pr³⁺ doping contents were intended to be 1mol% in both crystals, while the Mg²⁺ doping content was 5mol% in Pr:Mg:CLN, slightly higher than the threshold concentration. Firstly, the raw materials were mixed well and calcined at 1150°C for the solid-state synthesis of doped CLN polycrystals. Secondly, the polycrystals were melt in a platinum crucible; after seeding the doped CLN single crystals were grown using a homemade automatic control Czochralski puller at a rotating rate of 15rpm and a pulling rate of 1.2mm/h. Finally, the as-grown crystals were cooled to the room temperature at an annealing rate of 10–30°C/h. Wafers oriented perpendicular to and parallel to c axis, labeled as σ-polarized and π-polarized, were cut from the central part of the Pr:CLN and Pr:Mg:CLN crystals. The wafers were referred as 1Pr0Mg-σ, 1Pr0Mg-π, 1Pr5Mg-σ and 1Pr5Mg-π respectively hereinafter.

 

Fig. 1 Raman spectra of Pr³⁺ doped and Pr³⁺-Mg²⁺ codoped congruent LiNbO₃ crystals (Pr:CLN and Pr:Mg:CLN), with the photographs displaying the appearance of the as-grown single crystals (a) Pr:CLN and (b) Pr:Mg:CLN in the insets.

Download Full Size | PDF

The concentrations of dopants Pr³⁺ and Mg²⁺ in as-grown crystals were estimated by Inductively Coupled Plasma Optical Emission Spectrometry (ICP-OES) with VARIAN-750 type. The Raman spectra under 514.5nm excitation were recorded using a Laser Micro-Raman Spectrometer (Renishaw inVia). The transmission absorption spectra of the wafers were recorded using a UV-Vis-NIR Spectrophotometer (Lambda 950) at room temperature.

The fluorescence excitation spectra, emission spectra and decay curves were recorded by a Photoluminescence Spectrometer (EDINBURGH FLSP920) equipped with a xenon lamp and a microsecond Xe-flash lamp (μF900). The optical signals at the visual (near-infrared) range were detected using a cooled HAMAMATSU RP-928 (R5509-72) photomultiplier tube. For the measurements at temperature from 20K to 430K, an ARS closed cycle He refrigerator was used.

3. Results and discussion

3.1 Raman spectra and phonon vibration

The congruent LiNbO₃ bulk crystal has a trigonal crystal structure. The concentrations of Pr³⁺ in Pr:CLN and Pr:Mg:CLN were determined to be 1.06 mol% and 1.09 mol%, while the concentration of Mg²⁺ in Pr:Mg:CLN was determined to be 5.21 mol% by the ICP-OES; The effective segregation coefficients were 1.06, 1.09 and 1.04 respectively. The consistent contents of Pr³⁺ and Mg²⁺ in the single crystals and melts verified the successful introduction of the dopants. According to the lithium-vacancy model [20], large amounts of Li-vacancy ${\text{V}}_{\text{Li}}^{-}$and Nb-antisite ${\text{Nb}}_{\text{Li}}^{4+}$ distributed in Pr:CLN. In the Pr:Mg:CLN crystal, with the exceeding doping of Mg²⁺, ${\text{Nb}}_{\text{Li}}^{4+}$ was completely eliminated by ${\text{Mg}}_{\text{Li}}^{+}$ and excess Mg²⁺ started to replace both normal Li⁺ and Nb⁵⁺ forming ${\text{Mg}}_{\text{Li}}^{+}$ and ${\text{Mg}}_{\text{Nb}}^{3-}$ [21]. Pr³⁺ would locate at the normal Li-site and Nb-site since no obvious change could be observed in the Raman tests.

Raman scattering measurements were carried out and the lattice vibrations (phonons) were analyzed. The vibrational modes in Pr:CLN and Pr:Mg:CLN marked in Fig. 1 were consistent with the data of LiNbO₃ in previous study [22,23]. The positions of Raman lines appeared to be little affected by exceeding Mg²⁺-doping. Broadening of Raman lines could be observed in Pr:Mg:CLN especially at 871.68cm⁻¹ (E(LO9)), resulting from the formation of new defects ${\text{Mg}}_{\text{Li}}^{+}$ and ${\text{Mg}}_{\text{Nb}}^{3-}$ and the increased disorder. In both crystals, the energy of the most energetic phonon was 871.68cm⁻¹. So, about 4 phonons of this type were needed to bridge the gap (about 3592cm⁻¹) between ³P₀ and ¹D₂ of Pr³⁺, while the gap (about 6619cm⁻¹) between ¹D₂ and ¹G₄ would take about 8 phonons, indicating that the multi-phonon relaxation between ¹D₂ and ¹G₄ was relatively inefficient.

3.2 Absorption spectra

The transmission absorption spectra of the wafers 1Pr0Mg-σ, 1Pr0Mg-π, 1Pr5Mg-σ and 1Pr5Mg-π at wavelength range from 350nm to 2500nm were presented in Fig. 2.

 

Fig. 2 The polarized transmittance absorption spectra of Pr:CLN and Pr:Mg:CLN wafers, referred as 1Pr0Mg-σ, 1Pr0Mg-π, 1Pr5Mg-σ and 1Pr5Mg-π, at the wavelength range from 350nm to 2500nm. In the insets: the infrared absorption spectra at around 3500cm⁻¹ (2857nm).

Download Full Size | PDF

The spectra consisted of absorption bands corresponding to the transitions of Pr³⁺ from ground level ³H₄ to excited levels ³P₀, ³P₁, ³P₂, ¹D₂, ¹G₄, ³F₄, ³F₃ and ³F₂. There was an apparent decrease in the absorption coefficient after 5mol% Mg²⁺-doping, revealing that the efficiencies of the 4f-4f absorption transitions decreased in Pr:Mg:CLN. This could be partially explained with the less polarizable chemical environment [24,25] in Pr:Mg:CLN when all intrinsic defects were eliminated. According to the equations of Judd-Ofelt theory, there would be a decrease in the efficiencies of both the 4f-4f absorption (η_ap) and emission (η_em) transitions in Pr:Mg:CLN.

The infrared absorption spectra at around 3500cm⁻¹ (2857nm) of Pr:CLN and Pr:Mg:CLN were presented in the insets of Fig. 2. Evident violet shift of the absorption peak related to the OH^- absorption could be observed after Mg²⁺ codoping.

3.3 Luminescence spectra and mechanism analysis

The fluorescence excitation spectra of the four wafer samples monitored at 618nm were measured and depicted in Fig. 3 (a) and (d). Except for the excitation peaks centered at 462nm 489nm and 501nm associated with the excited levels ³P₂ ³P₁ and ³P₀ (³P_2-0), broad excitation bands at around 360nm were observed in all wafers. The 360nm excitation bands were very close to the absorption edge, which was determined to be 349.6nm (325.10nm) in Pr:CLN (Pr:Mg:CLN) based on absorption spectra, and located in the strong fundamental absorption region of CLN; It was hard to identify them in the absorption spectra. The wide excitation bands could not be assigned to any 4f-4f transition and would be attributed to the transition from ground level ³H₄ to the IVCT state according to the IVCT model. After doping with 5mol% Mg²⁺, no evident changes were found on the ³P_2-0 excitation bands. However, the IVCT excitation bands extended to shorter wavelength from 300nm to 250nm in 1Pr5Mg-σ and 1Pr5Mg-π, resulting in changes of spectra in appearance. The extension was supposed to be related to a slight elevation of the IVCT state after exceeding Mg²⁺ doping.

 

Fig. 3 The (a) σ-polarized (d) π-polarized fluorescence excitation spectra measured by monitoring 618nm emissions in Pr:CLN and Pr:Mg:CLN. The Gaussian decomposition of the excitation band from 250nm to 450nm in (b) 1Pr0Mg-σ (c) 1Pr5Mg-σ (e) 1Pr0Mg-π (f) 1Pr5Mg-π, with the central wavelengths of Gauss peaks Peak0 Peak1 and Peak2 marked beside.

Download Full Size | PDF

For further insight of the excited levels related to the broad excitation band, the excitation bands from 250nm to 450nm in1Pr0Mg-σ 1Pr5Mg-σ 1Pr0Mg-π and 1Pr5Mg-π were decomposed into Gauss peaks, as presented in Fig. 3 (b) (c) (e) and (f). An extra Gauss peak Peak0 centered at around 280nm was observed only in Pr:Mg:CLN. Peak0 might be attributed to a transition related to the new defect levels introduced by Mg²⁺ incorporation, which had been demonstrated to be essential for the occurrence of long persistence luminescence in Pr:Mg:CLN [19]. Besides, the Peak1 and Peak2 could be found in all 4 wafers and might be attributed to the transition from ³H₄ to IVCT level. With the violet shift of Peak1 and Peak2, we confirmed a slight elevation of the IVCT state in Pr:Mg:CLN.

Higher intensities of 360nm excitation bands were also detected in Pr:Mg:CLN. Accordingly, the main red emissions (¹D₂→³H₄ of Pr³⁺) under 360nm excitation were stronger in 1Pr5Mg-σ and 1Pr5Mg-π respectively compared to those in Pr0Mg-σ and 1Pr0Mg-π, as presented in Fig. 4 (a) and (b). Since the emission spectra were all measured under the same circumstances, higher luminescence efficiency P_Lum related to the transition ¹D₂→³H₄ under 360nm excitation was achieved after exceeding Mg²⁺ doping.

 

Fig. 4 The (a) σ-polarized (b) π-polarized fluorescence emission spectra recorded under 360nm excitation. The (c) σ-polarized (d) π-polarized fluorescence emission spectra recorded under 463nm excitation.

Download Full Size | PDF

In principle, the luminescence efficiency P_lum should be positively associated with the feeding probability via IVCT (P_IVCT) and the 4f-4f transition efficiency η_em, expressed as: $P_{Lum}\propto P_{IVCT}\times {\eta }_{em}$. To confirm the decrease of η_em in Pr:Mg:CLN, fluorescence emission spectra under 463nm excitation were measured and depicted in Fig. 4(c) and (d). When excited with 463nm incident ray, Pr³⁺ could only be excited to 4f levels ³P_2-0. The excited Pr³⁺ in ³P_2-1 would relax to the nearby lower level ³P₀ promptly via multi-phonon relaxation and then relax to ¹D₂ via a nonradiative process, followed by efficient red emissions at 618nm. This route was less related to the IVCT state. The emission efficiency and intensity under 463nm excitation were mainly associated with η_em. The slightly reduced 618nm emission intensities in 1Pr5Mg-σ and 1Pr5Mg-π compared to those in 1Pr0Mg-σ and 1Pr0Mg-π indicated a lower 4f-4f transition efficiency η_em when doped with 5mol% Mg²⁺. This observation coincided with the changes in absorption spectra in Section 3.2. With the increasing P_Lum and decreasing η_em, we concluded that the feeding probability to ¹D₂ via IVCT P_IVCT increased with exceeding Mg²⁺ doping.

An configuration coordinate diagram which could offer great convenience for the explanation of the IVCT-based luminescent mechanism in Pr:CLN and Pr:Mg:CLN was constructed and depicted in Fig. 5. The configurational coordinate diagram was based on the breathing model of optical center (Pr³⁺ and its ligands ions) [26]. The parabolas here represented the Pr³⁺-Nb⁵⁺ IVCT state and the different energy states of the optical center, including Pr³⁺ in different 4f levels and the vibration energy of the ligands. The positions of all parabolas were determined based on the data from luminescence and absorption spectra.

 

Fig. 5 The configurational coordinate diagram of Pr:CLN and Pr:Mg:CLN, depicting the light-emitting mechanism based on the intervalence charge transfer. IVCT_i (i = a, b) represented the Pr³⁺–Nb⁵⁺ intervalence charge transfer state; E_actv,i represented the quenching activation energy; ΔR_i represented the coordinate offset of IVCT state; Dashed arrows ① and ② signified the quenching of ³P₀ emissions under ① 360nm excitation and ② 463nm excitation.

Download Full Size | PDF

Two mechanisms for red emissions were presented in Fig. 5. Under 360nm excitation, charge transfer occurred from Pr³⁺ to 4d⁰ orbit of Nb⁵⁺ and Pr³⁺ was excited into IVCT state, followed by de-excitation directly to ¹D₂ bypassing the levels ³P_2-0. Under 463nm excitation, Pr³⁺ reached the excited levels ³P_2-0 and then de-excited to ¹D₂ via the crossovers between the IVCT state and levels ³P_2-0 or ¹D₂. The energy barriers were low and the IVCT state could work as an effective quenching channel for ³P_2-0 in both crystals. So in all our samples, the characteristic blue-green emissions at around 500nm related to transition ³P₀→³H₄ were almost completely quenched. Only 618nm and 715nm emissions from excited level ¹D₂ could be found in Fig. 4.

3.4 Temperature quenching of red emissions and lifetimes related to ¹D₂→³H₄

The unpolarized fluorescence emission spectra of Pr:CLN and Pr:Mg:CLN under 360nm excitation were recorded at the temperature ranging from 20K to 430K, as presented in Fig. 6 (a) and (b). As temperature increased, the normalized intensity I of the red emission weakened slowly at first but then faster and faster. The decrease of intensity ratio I/I_0K was also pronounced at higher temperatures as depicted in Fig. 6 (c). The inflexions of the ratio curves were found at around 250K, followed by a sharp drop and finally quenching of red emissions. Noteworthily, compared to Pr:CLN, the intensity ratio in Pr:Mg:CLN decreased less quickly with smaller curvature.

 

Fig. 6 The normalized temperature-dependent fluorescence emission spectra at temperatures ranging from 20K to 430K under 360nm excitation in (a) Pr:CLN and (b) Pr:Mg:CLN. (c) The temperature dependences of emission intensity ratio I/I_0K related to the transition ¹D₂→³H₄ and the fitting curves acquired using formula (1). The temperature-dependent fluorescence decay curves from excited level ¹D₂ measured under 360nm excitation in (d) Pr:CLN and (e) Pr:Mg:CLN. (f) The temperature dependences of the radiant lifetime and the fitting curves acquired using formula (1).

Download Full Size | PDF

The quenching at higher temperatures was caused by the thermally activated energy transfer from IVCT state to ground level ³H₄ of Pr³⁺. With temperature rising, the electrons on the bottom of IVCT state could more easily migrate to its crossover with ³H₄ and then relax there, bypassing and depopulating ¹D₂.

Further analysis on the difference between the temperature dependence of fluorescence in Pr:CLN and that in Pr:Mg:CLN could be carried out according to the theory put forward by Struck [8,27]. The temperature-dependent red emission data could be reliably fitted by the formula:

Where I_0K represents the emission intensity at 0K; A is a pre-exponential constant; k_B is the Boltzmann constant; The quenching activation energy E_actv, representing the vertical distance from the minimum of IVCT state to its crossover with the quenching level in the coordinate diagram [28], is correlated to the probability of the temperature-dependent nonradiative relaxation [26]. The fitting curves were also presented in Fig. 6 (c). The indicative value of E_actv was determined to be 2420.07cm⁻¹ in Pr:CLN and 4134.97cm⁻¹ in Pr:Mg:CLN. Despite only eight sampling temperatures and consequently relative low accuracy of E_actv, we could draw a qualitative conclusion that the red emissions tended to be quenched more difficultly due to the higher energy barrier in Pr:Mg:CLN.

The decay curves related to the ¹D₂→³H₄ transition from 20K to 430K were measured under 360nm excitation and presented in Fig. 6 (d) and (e). All the decay curves exhibited a single exponential behavior. The values of lifetime were determined and plotted in Fig. 6 (f). The lifetime values of Pr:CLN at low temperatures were consistent with the data reported in [14]. With exceed Mg²⁺ doping, the lifetimes in Pr:Mg:CLN at the temperature from 20K to 430K were significantly increased. The increased lifetime of red emission after Mg²⁺-codoping agreed with the increasing radiative lifetime of excited level ¹D₂, which would result from the lower transition probabilities as discussed in Section 3.2 and 3.3.

The temperature dependences of the lifetime and those of emission intensity related to ¹D₂→³H₄ were interrelated. As temperatures rose, the reduction of lifetime became more and more apparent, following the same tendency of intensity variation. Similarly, the E_actv could be also calculated from the lifetime data. The fitting curves were presented in Fig. 6 (f). The values of E_actv were 2486.79cm⁻¹ and 4519.52cm⁻¹ for Pr:CLN and Pr:Mg:CLN respectively. Moreover, according to the fitting results, it was obvious that the more accurate way to determine E_actv was to measure the lifetimes rather than emission intensities; the latter could be more easily influenced by environmental disturbances and absorption strength change with temperature.

The reason for higher E_actv in Pr:Mg:CLN was illustrated in Fig. 5. Firstly, the extension to shorter wavelength of the IVCT-related excitation band mentioned above vertically elevated the corresponding IVCT parabola and increased E_actv. Secondly, the smaller coordinate offset of IVCT state (ΔR), which represented the distance between ligand and central ion Pr³⁺, was also responsible. When doping with exceeding Mg²⁺ ions, excess Mg²⁺ would partially locate at normal Nb-site according to the Li-vacancy model. Due to the larger diameter of Mg²⁺ than Nb⁵⁺, the substitution would enlarge the lattice environment in the vicinity of ${\text{Mg}}_{\text{Nb}}^{3-}$. Supposing that there was also ${\text{Pr}}_{\text{Li}}^{2+}$ in the vicinity, it was reasonable to infer that the lattice environment of ${\text{Pr}}_{\text{Li}}^{2+}$ and the nearest Nb⁵⁺ shrank. With smaller ΔR, the IVCT parabola of Pr:Mg:CLN shifted right, inducing a further increase of E_actv.

As a practical matter, the heating effect caused by the strong absorption of excitation light would raise the temperature of crystals. Consequently, the relative low quenching temperature of 310K (36.85°C), at which precipitous quenching for red-emission intensities and lifetimes in Pr:CLN started to occur, would limit the applications in luminescent/laser devices. It was noteworthy that the higher quenching temperature of 370K (96.85°C) was achieved in Pr:Mg:CLN, which was of practical value in luminescent or laser applications. Besides, the temperature dependences of Pr³⁺ doped CLNs from 275K to 400K could be tuned with Mg²⁺ dopant. The temperature sensitivities offered interesting prospects for use in temperature sensors and deserved a further investigation.

4. Conclusion

The effect of exceeding Mg²⁺ doping on the optical properties of Pr³⁺ in CLN has been investigated based on the absorption, excitation and emission spectra. Blue-green emissions were quenched completely in CLN due to the depopulation of ³P₀ via IVCT state. The exceeding Mg²⁺ doping was found to reduce the red emission excited at 463nm indicating the decrease of 4f-4f transition probability, but enhance the one excited at 360nm due to the larger population of ¹D₂ via IVCT. The temperature-dependent red emissions and lifetimes under 360nm excitation were measured. The lifetimes from 20K to 430K related to ¹D₂→³H₄ were significantly increased by the Mg²⁺ incorporation. Thermal quenching for red light emission and lifetime in Pr:CLN and Pr:Mg:CLN was demonstrated to be caused by thermally activated energy transfer from IVCT state to ³H₄ of Pr³⁺. The quenching temperature was found to be raised by the exceeding Mg²⁺ doping, which was meaningful in actual luminescent and laser application. The quenching activation energy was calculated and higher energy barrier for quenching in Pr:Mg:CLN was interpreted considering the displacement of the IVCT state in configurational coordinate diagram. Besides, the temperature sensitivities and the tunability by Mg²⁺ doping in Pr³⁺ doped CLN were regarded to be of practical value in sensor applications. This work is considered to be promotive for understanding the luminescence mechanism based on IVCT process in Pr³⁺ doped materials.