1. Introduction

Rare-earth sesquioxides R₂O₃ (where R is a lanthanide, Y or Sc) represent an important class of optical materials. The most interesting form of R₂O₃ compounds is the cubic one (also called C-type, sp. gr. Ia$\bar{3}$, bixbyite structure). Cubic rare-earth sesquioxides such as Y₂O₃ (yttria), Sc₂O₃ (scandia) or Lu₂O₃ (lutetia) are known as excellent host media for doping with trivalent rare-earth ions (RE³⁺) [1]. As host matrices, they feature good thermo-mechanical and thermo-optical properties [2,3] (i.e., high thermal conductivity with weak dependence on the doping level, weak isotropic thermal expansion, small and positive dn/dT coefficient), low phonon energies (for oxide materials), and wide transparency [4]. The RE³⁺ dopant ions substitute for the host-forming R³⁺ metal cations in two types of sites (C₂ and C_3i symmetry) with VI-fold oxygen coordination [5,6]. However, the spectroscopic properties of RE³⁺-doped R₂O₃ compounds are mainly determined by C₂ species (about 3/4 of ions) since for the centrosymmetric C_3i ones (about 1/4 of ions), the electric dipole transitions are not enabled [1]. The RE³⁺ ions in cubic sesquioxides experience strong crystal fields leading to large Stark splitting of their multiplets and, consequently, broad emission bands. The crystal-field strength increases in the series R³⁺ = Y³⁺ → Lu³⁺ → Sc³⁺ according to the variation of the ionic radius, making it possible to alter the spectral properties by changing the host composition [7].

The main drawback of R₂O₃ crystals are their extremely high melting points (e.g., 2425°C for Y₂O₃) complicating the crystal growth. So far different techniques have been used for growing R₂O₃ crystals, such as heat exchanger growth method (HEM), micro-pulling-down (µ-PD), Czochralski (Cz), etc. [8,9]. Still, single-crystals suffer from coloration, Rhenium (Re) impurities (when using Re crucibles), and a gradient of RE³⁺ dopant concentration (especially for Sc₂O₃). During the past decades, the transparent ceramic technology emerged as a competitive approach to the single-crystal growth [10]. It offers: (i) much lower synthesis temperatures (<1800°C); (ii) easier RE³⁺ doping in terms of higher available concentrations and more uniform ion distribution; (iii) well-preserved spectroscopic and thermal properties; (iv) possibility to fabricate mixed compositions (R₁,R₂)₂O₃ with well-controlled R₁/R₂ ratio; and (v) size-scalability. Laser-quality RE³⁺-doped yttria [11,12], lutetia [13,14] and scandia [15] ceramics have been developed. The main challenge for ceramics is reaching high transparency close to the theoretical limit (i.e., weak light scattering owing to residual pores, possible secondary phases at the grain boundaries, etc.).

Substitutional solid solutions A_1-xB_x (0 < x < 1) also called “mixed” materials are attracting attention for tailoring the spectroscopic properties of the dopant RE³⁺ ions [16–19] as, under the condition of different crystal-field strengths associated with the parent compounds A and B, the absorption / emission lines of the dopants may experience strong inhomogeneous broadening. For cubic sesquioxides, isostructural solid-solutions with a general composition (Y_1-x-yLu_xSc_y)₂O₃ exist in the full range of x and y. The same is true also for mixing with some active RE³⁺ ions, as long as their stoichiometric compositions (RE₂O₃) possess the same cubic symmetry, e.g. Yb, Tm, Ho or Er. Although the growth of “mixed” sesquioxide crystals has been reported [19,20], it is much easier to fabricate such compounds via a transparent ceramic technology. Recently, laser ceramics in the lutetia-scandia and lutetia-yttria binary systems were developed [16,18,21–25].

Thulium (Tm³⁺) doped sesquioxides attract attention because of their suitability for efficient lasing around 2 µm according to the ³F₄ → ³H₆ electronic transition [26]. Tm³⁺ ions experience strong crystal-fields in R₂O₃ compounds. The resulting substantial Stark splitting of their multiplets determines broad emission spectra naturally extending above 2 µm [7]. The latter property is of practical importance for broadly tunable and especially mode-locked lasers [27,28] as this helps to avoid the structured water vapor absorption in the atmosphere spectrally located at < 2 µm which is detrimental for achieving femtosecond pulses. Recently, femtosecond mode-locked Tm ceramic lasers based on “mixed” sesquioxides such as Tm:(Lu,Sc)₂O₃ and Tm:(Lu,Y)₂O₃ were reported [28–30].

The present work is devoted to the spectroscopic properties of Tm³⁺ ions in a “mixed” (lutetia – yttria) sesquioxide laser ceramic, with the goal of revealing the effect of compositional disorder on the inhomogeneous broadening of absorption and emission lines, in order to better understand the potential of such materials for generation of few-optical-cycle pulses.

2. Synthesis of transparent ceramics

Commercial powders of rare-earth oxides, Lu₂O₃ (purity: 4N), Y₂O₃ (5N) and Tm₂O₃ (4N), from Jiahua Advanced Material Resources, China, were used as raw materials. They were weighed according to the composition of 3.0 at.% Tm:(Lu_0.5Y_0.5)₂O₃. The mixed powders were ball-milled in ethanol for 24 h with 1.0 at % monoclinic ZrO₂ powder (Shandong Sinocera Functional Material, China) serving as a sintering aid. The milled slurries were dried in an oven at 70°C for 24 h and then sieved through a 100-mesh screen. After that, the powders were dry-pressed into pellets by a 12 mm-diameter mold and cold isostatically pressed at 200 MPa for 5 min. All the green bodies were calcined at 850°C for 4 h to remove the residue organics. After that, the samples were first pre-sintered at 1650°C for 4 h in vacuum under a pressure lower than 1.0 × 10⁻³ Pa, equipped with a tungsten mesh as the heating element. Then, the pre-sintered ceramics were treated by hot isostatic pressing (HIPing) at 1600°C for 3 h in 190 MPa argon pressure to eliminate residual pores. Finally, the HIP-treated samples were annealed at 1200°C for 24 h in a muffle furnace in air to compensate the oxygen loss during the vacuum pre-sintering and the following HIP treatment. Then, the ceramic disks were polished on both sides to laser quality level, Fig. 1. The obtained ceramic disks were transparent and slightly yellow colored due to the Tm³⁺ doping. The calculated Tm³⁺ ion density in the “mixed” ceramic was N_Tm = 8.34 × 10²⁰ at/cm³.

 

Fig. 1. A photograph of annealed and polished 3.0 at.% Tm:LuYO₃ ceramic disk.

Download Full Size | PDF

To observe the microstructure of the ceramic, its polished surface was thermally etched at 1400°C for 2 h in a muffle furnace in air. The thermally etched surfaces of the 3.0 at.% Tm:LuYO₃ ceramics before and after HIPing were characterized using a Scanning Electron Microscope (SEM, TM3000, Hitachi, Japan). The pre-sintered ceramic has an average grain size of 1 µm and it contains sub-µm sized pores localized at the grain boundaries, Fig. 2(a). After HIPing, the mean grain size increases to 2.4 µm. The ceramic exhibits a close-packed microstructure with clean grain boundaries, Fig. 2(b).

 

Fig. 2. SEM images of the thermally etched surface of the 3.0 at.% Tm:LuYO₃ ceramics: (a) a pre-sintered sample; (b) a sample after HIPing.

Download Full Size | PDF

For comparison, two parent ceramics, Tm:Lu₂O₃ and Tm:Y₂O₃, doped with 3 at.% Tm³⁺ (N_Tm = 8.61 × 10²⁰ cm^-3 and 7.85 × 10²⁰ cm^-3, respectively), were fabricated.

3. Results and discussion

3.1 Raman spectra

The room temperature (RT) Raman spectra were measured using a confocal laser microscope (Renishaw inVia) equipped with a × 50 Leica objective and an Ar⁺ ion laser (514 nm). For cubic sesquioxides (sp. gr. Ia$\bar{3}$) possessing a body-centered structure, the factor group analysis predicts the following irreducible representations for the optical and acoustical modes at the center of the Brillouin zone (k = 0): Γ_op = 4A_g + 4E_g + 14F_g + 5A_2u + 5E_u + 16F_u (of which 22 modes (A_g, E_g, and F_g) are Raman-active, 16 modes (F_u) are IR-active, and the rest are silent) and Γ_ac = F_u [31,32].

Figure 3(a) shows the Raman spectra of the Tm³⁺-doped LuYO₃, Y₂O₃ and Lu₂O₃ ceramics. They are typical for cubic (C-type) sesquioxides. The vibrational spectra exhibit two distinct frequency ranges: the relative position and intensities of the modes above 300 cm^-1 are rather similar for different ceramic compositions indicating that they are most probably related to oxygen motions and deformations of the [AO₆] octahedrons [31]. The most intense peak is assigned to F_g + A_g vibrations [31], Fig. 3(b). For the yttria and lutetia ceramics, it is centered at 377.8 and 391.4 cm^-1, respectively, and its linewidth is nearly the same (6.7 and 7.4 cm^-1, respectively). For the “mixed” ceramic, this peak takes an intermediate position (385.4 cm^-1) and it is notably broadened (linewidth: 12.8 cm^-1) confirming the formation of a substitutional solid-solution. A similar tendency is observed for other well assigned modes in the high-frequency range of ∼300 – 600 cm^-1. The peak corresponding to the maximum phonon energy (the F_g + A_g vibrations) is observed at 593 cm^-1 (Tm:Y₂O₃), 603 cm^-1 (Tm:LuYO₃) and 612 cm^-1 (Tm:Lu₂O₃).

 

Fig. 3. Unpolarized Raman spectra of Tm³⁺-doped Lu₂O₃, Y₂O₃ and LuYO₃ ceramics: (a) overview spectra; (b) a close look at the most intense mode (F_g + A_g), λ_exc = 514 nm. Numbers indicate the Raman frequencies in cm^-1.

Download Full Size | PDF

3.2 Absorption spectra and Judd-Ofelt analysis

The transmission / absorption spectra were measured using a spectrophotometer (Lambda 1050, Perkin Elmer).

The Tm:LuYO₃ ceramic exhibited a relatively high linear transmission of 81.5% at 2.2 µm (out of the Tm³⁺ absorption bands), close to the theoretical limit, T₀ = 2n/(n² + 1) = 82.0% (a formula accounting for multiple light reflections, n = 1.918 is the estimated refractive index of LuYO₃ [33]). The absorption spectrum of Tm³⁺ ions in the “mixed” ceramic is shown in Fig. 4. Bands related to transitions from the ground-state (³H₆) to the excited-states ranging from ³F₄ up to ³P_0-2 are observed. The UV absorption edge is observed at ∼250 nm (for undoped Lu₂O₃, the optical bandgap E_g is 5.6 eV [34] or ∼221 nm).

 

Fig. 4. RT absorption spectra of the 3 at.% Tm:LuYO₃ ceramic in the spectral range of (a) 245 – 320 nm, (b) 340 – 500 nm, (c) 640 – 840 nm, (d) 1050 – 2050 nm.

Download Full Size | PDF

The absorption cross-sections σ_abs for the ³H₆ → ³H₄ transition of Tm³⁺ in the “mixed” LuYO₃ ceramic, and in Y₂O₃ and Lu₂O₃, are shown in Fig. 5. This absorption band is suitable for pumping Tm-lasers using commercially available AlGaAs diode lasers emitting around 0.8 µm. For the “mixed” ceramic, the absorption spectrum is broadened as compared to both parent compounds; the maximum σ_abs is 0.33 × 10⁻²⁰cm² at 796.2 nm corresponding to an absorption bandwidth Δλ_abs of ∼21 nm (combining several peaks), compared with σ_abs = 0.37 × 10⁻²⁰cm² at 796.7 nm with Δλ_abs of ∼7 nm for the Tm:Y₂O₃ ceramic.

 

Fig. 5. Absorption cross-sections, σ_abs, for the ³H₆ → ³H₄ transition of Tm³⁺ ions in the LuYO₃, Y₂O₃ and Lu₂O₃ ceramics.

Download Full Size | PDF

The measured absorption spectrum was analyzed using the standard Judd-Ofelt (J-O) theory [35,36]. Eight Tm³⁺ transitions were considered. The set of reduced squared matrix elements U^(k) was taken from [37]. The magnetic-dipole (MD) contributions to transition intensities (for transitions with ΔJ = J – J’ = 0, ±1) were calculated within the Russel-Saunders approximations using the wave functions of the free Tm³⁺ ion. The refractive index of the “mixed” ceramic LuYO₃ was calculated using the dispersion curves of the parent compounds [34]. More details about the J - O analysis can be found elsewhere [38].

For calculating the absorption oscillator strengths, we have used the full Tm³⁺ ions density (N_Tm), although some authors suggest to account only for ions located in C₂ sites [∼(3/4)N_Tm] [39]. However, for a “mixed” ceramic, the actual distribution of dopant ions over the C₂ and C_3i sites may significantly differ from that for the parent material.

Table 1 contains the experimental (f^Σ_exp) and calculated (f^Σ_calc) absorption oscillator strengths. Here, the superscript “Σ” indicates a total value (ED + MD). The root mean square (r.m.s.) deviation between the f^Σ_exp and f^Σ_calc values is δ_rms = 1.202, mainly due to the transitions to thermally coupled levels ³F₂ + ³F₃ and ¹I₆ + ³P₀ + ³P₁. For the lowest-lying excited-state (³F₄), a relatively good agreement is observed. The corresponding J - O parameters are Ω₂ = 2.537, Ω₄ = 1.156 and Ω₆ = 0.939 [10²⁰cm²]. These values agree well with those reported recently for another Tm³⁺-doped “mixed” sesquioxide ceramic with a composition (Lu,Sc)₂O₃, Ω₂ = 2.429, Ω₄ = 1.078 and Ω₆ = 0.653 [10²⁰cm²] [16].

Table 1. Experimental and Calculated Absorption Oscillator Strengths^a for Tm³⁺ Ions in LuYO₃^a

View Table | View all tables in this article

Using the determined J - O parameters, the probabilities (ED + MD) of spontaneous radiative transitions for particular emission channels J → J’ A_Σ^calc(JJ’), the total probabilities of radiative transitions from excited-states A_tot = Σ_J'A_Σ^calc(JJ’), the luminescence branching ratios B(JJ’) = A_Σ^calc(JJ’)/A_tot and the radiative lifetimes τ_rad = 1/A_tot were calculated, cf. Table 2. The mean emission wavelengths 〈λ_em〉 were estimated using the barycenter energies of Tm³⁺ multiplets 〈E〉 from Table 1. For the ³F₄ and ³H₄ states, τ_rad amounts to 4.18 ms and 0.61 ms, respectively.

Table 2. Calculated Emission Probabilities for Tm³⁺ Ions in LuYO₃^a

View Table | View all tables in this article

3.3 Emission (spectra and lifetimes)

The Tm³⁺ luminescence was excited at ∼796 nm using a CW Ti:Sapphire laser and the spectra were measured using an optical spectrum analyzer (OSA, AQ6376, Yokogawa) and a ZrF₄ fiber. The spectral sensitivity of the set-up was calibrated using a 20 W quartz iodine lamp.

The normalized RT emission spectra of Tm³⁺ ions in the LuYO₃, Y₂O₃ and Lu₂O₃ ceramics are shown in Fig. 6(a). The observed emission is related to the ³F₄ → ³H₆ transition. The spectra are very broad spanning from 1.6 up to 2.35 µm. For the “mixed” ceramic, the spectrum exhibits a notable inhomogeneous broadening as compared to those of the parent compounds. The spectral maximum is found at ∼1936nm. For the quasi-three-level ³F₄ → ³H₆ Tm³⁺ laser, emission is expected at the long-wave wing of the luminescence spectrum. For Tm³⁺-doped cubic sesquioxides, a broad peak above 2 µm is observed where maximum laser gain will occur. For the Tm:LuYO₃ ceramic, it is centered at 2058 nm, in between the positions for Tm:Y₂O₃ (2049 nm) and Tm:Lu₂O₃ (2063 nm). The emission bandwidth for this peak also exceeds those for the parent compounds supporting our assumption about the formation of a solid-solution.

 

Fig. 6. (a,b) RT luminescence spectra of Tm:LuYO₃, Tm:Y₂O₃ and Tm:Lu₂O₃ ceramics around 2 µm: (a) overview of the spectra, (b) a close look at the 2100 – 2375 nm range, semi-log scale, grey curve – emission spectrum for the ³H₄ → ³H₅ Tm³⁺ transition in a low-doped (<0.1 at.%) Tm:Lu₂O₃ shown for comparison. λ_exc = 796 nm.

Download Full Size | PDF

A careful examination of the long-wave part of the luminescence spectra of all the studied ceramics indicates that their emission extends up to at least 2.35 µm (further measurement was limited by the sensitivity of our set-up). It is associated with the ³F₄ → ³H₆ transition of Tm³⁺ according to the luminescence decay studies: the luminescence lifetime remains nearly constant when changing the detection wavelength between 1.8 – 2.3 µm. One may argue that such long-wave emission may also partially originate from the ³H₄ → ³H₅ electronic transition [40]. However, no characteristic spectral features of this transition are found in the spectra. To prove this, we have used a low-doped (<0.1 at.%) Tm:Lu₂O₃ crystal exhibiting almost no self-quenching of the ³H₄ lifetime to measure the luminescence spectrum of the ³H₄ → ³H₅ transition, Fig. 6(b). It reveals several well-resolved emission peaks centered at 2282, 2308, 2324 and 2356 nm which are not found in the spectra of the ceramics. In contrast, the long-wave part of the emission spectrum of the ceramics is almost structureless and follows an exponential law at >2.25 µm, as seen from Fig. 6(b) plotted in a semi-log scale.

The long-wave limit for purely electronic transitions ³F₄ → ³H₆ is determined by the crystal-field splitting of the involved multiplets and, in particular, by the energy gap between the lowest sub-level of the ³F₄ state (5643 cm^-1, Y₁) and the highest sub-level of the ³H₆ ground-state (810 cm^-1, Z₁₃, see below) [41]. This yields a value of 2069 nm. All the emissions above this wavelength are multiphonon-assisted (or vibronic) as they are related to the coupling between electrons and host vibrations (phonons) [42].

The very broad peak centered at 2.23 µm (for Tm:LuYO₃) is interpreted as a phonon sideband, see [42,43] for this term. Note that it takes an intermediate position between those for Tm:Y₂O₃ (2.21 µm) and Tm:Lu₂O₃ (2.25 µm) which agrees with the difference in the crystal-field strengths. This indicates that the position of this peak could be linked to those of electronic transitions. Indeed, for the “mixed” ceramic, the energy gap between the prominent electronic emission band centered at 2058nm and the above-described sideband (2236 nm) is ∼387 cm^-1. This well matches the most intense Raman peak of this material corresponding to vibration with an energy hν_ph = 385.4 cm^-1, Fig. 3(b). Thus, one can describe the appearance of the phonon sideband in the Tm³⁺ emission spectrum as transitions to a virtual energy level located slightly above the highest electronic sub-levels of the ³H₆ multiplet [44], i.e., having an energy E(Z₁₃) + hν_ph.

The stimulated-emission (SE) cross-sections, σ_SE, for the ³F₄ → ³H₆ transition of Tm³⁺ ions in the LuYO₃ ceramic were calculated using two methods, namely, (i) the reciprocity method [45], and (ii) the Füchtbauer – Ladenburg (F-L) formula [46]. The σ_SE spectra obtained by both methods were in good agreement with each other, considering the effect of reabsorption on the measured emission spectrum. In the F-L formula, we have used a radiative lifetime of the ³F₄ state τ_rad = 3.85 ± 0.1 ms to fit the two methods which reasonably agrees with that determined using the J - O theory (4.18 ms). In Fig. 7(a), the combined SE cross-section spectrum is shown. The maximum σ_SE is 0.75 × 10⁻²⁰ cm² at 1937 nm and at longer wavelengths where the laser operation is expected, σ_SE = 0.27 × 10⁻²⁰ cm² at 2059 nm.

 

Fig. 7. The ³H₆ ↔ ³F₄ transition of Tm³⁺ ions in the LuYO₃ ceramic: (a) absorption, σ_abs, and stimulated-emission (SE), σ_SE, cross-sections; (b) the gain cross-sections, σ_gain = βσ_SE – (1 – β)σ_abs, for different inversion ratios β = N₂(³F₄)/N_Tm.

Download Full Size | PDF

The ³F₄ → ³H₆ transition of Tm³⁺ ions represents a quasi-three-level laser scheme with reabsorption at the laser wavelength. Thus, the gain cross-sections, σ_gain = βσ_SE – (1 – β)σ_abs, are calculated, where β = N₂(³F₄)/N_Tm is the population inversion ratio. The gain profiles of the “mixed” ceramic are shown in Fig. 7(b). The spectra are smooth and broad extending until 2.35 µm. For small inversion ratios (β < 0.10), two local maxima appear in the spectra, centered at ∼2085 and 2059nm. For β = 0.04, the gain bandwidth (FWHM) is as broad as 75 nm. For higher β > 0.10, the gain maxima experience a blue-shift to ∼1960 and 1938nm. The observed broadband gain properties indicate the high suitability of this ceramic for generation of sub-100 fs pulses.

The existence of gain at long wavelengths well above 2.1 µm due to the multiphonon-assisted transitions is an important prerequisite for generation of ultrashort pulses from mode-locked Tm sesquioxide lasers. Indeed, the emission spectra of such lasers delivering pulses in the sub-100 fs time domain contained spectral components extending up to 2.3 µm [30,47].

The luminescence dynamics was studied using a ns optical parametric oscillator (Horizon, Continuum), a 1/4 m monochomator (Oriel 77200), an InGaAs detector and an 8 GHz digital oscilloscope (DSA70804B, Tektronix). To avoid the effect of reabsorption (radiation trapping) on the measured lifetimes, finely powdered samples were used.

For the ³F₄ Tm³⁺ state, the decay curves are well described by a single-exponential law, Fig. 8(a), yielding τ_lum = 3.470 ms for the Tm:LuYO₃ ceramic. This value is slightly longer compared to the parent compounds, 3.224 ms (Tm:Lu₂O₃) and 2.919 ms (Tm:Y₂O₃). Note that the measured luminescence lifetimes of the ³F₄ state for the studied sesquioxide ceramics are close to those obtained for single-crystals with low Tm³⁺ doping levels (<0.3 at.%), i.e., 3.38 ms (Tm:Lu₂O₃) and 3.54 ms (Tm:Y₂O₃) [7], indicating a relatively weak concentration quenching.

 

Fig. 8. RT luminescence decay curves for Tm³⁺ ions in the LuYO₃, Y₂O₃, and Lu₂O₃ ceramics: (a) decay from the ³F₄ state, λ_exc = 1644 nm, λ_lum = 2000nm; (b) decay from the ³H₄ state, λ_exc = 785 nm, λ_lum = 821 nm. Powdered samples.

Download Full Size | PDF

For the ³H₄ pump level, the decay is clearly not single-exponential, Fig. 8(b), owing to the efficient cross-relaxation (CR) process, ³H₄ + ³H₆ → ³F₄ + ³F₄. For the Tm:LuYO₃ ceramics, the mean luminescence lifetime of the ³H₄ state <τ_lum> is only 24 µs. It is much shorter than the so-called intrinsic lifetime (measured at a very low Tm³⁺ doping level, i.e., unaffected by the CR process), τ_lum,0 = 350 µs for Tm:Lu₂O₃ [7]. Thus, the estimated CR rate, W_CR = (1/τ_lum) – (1/τ_lum,0), is about 3.88 × 10⁴ s^-1 (for 3 at.% Tm³⁺ doping). The CR rate is quadratically proportional to the doping concentration, W_CR = C_CR(N_Tm)² [48], where C_CR = 0.56 × 10⁻³⁷ cm⁶s^-1 is the concentration-independent CR parameter.

Table 3 summarizes the measured luminescence lifetimes.

Table 3. Measured Luminescence Lifetimes of the ³F₄ and ³H₄ Tm³⁺ States in the LuYO₃, Y₂O₃ and Lu₂O₃ Ceramics

View Table | View all tables in this article

3.4 Low-temperature spectroscopy

For low temperature (LT, 12 K) studies, the samples were mounted in an APD DE-202 closed-cycle cryo-cooler equipped with an APD HC 2 Helium vacuum cryo-compressor and a Laceshore 330 temperature controller.

The ³H₆ → ³F₄ (absorption) and ³F₄ → ³H₆ (luminescence) Tm³⁺ transitions were considered giving access to the crystal-field splitting of both multiplets. In Fig. 9, the obtained LT absorption spectra are plotted vs. the photon energy and the LT emission spectra – vs. (E_ZPL – photon energy). The considered transitions are of pure ED nature and thus no signatures of C_3i sites could be observed. For the RE³⁺ ions in C₂ sites, each multiplet ^2S+1L_J with an integer J is split into a total of 2J + 1 sub-levels. A careful examination of the LT spectra of the parent compounds allowed us to determine the full set of Stark sub-levels for both mutiplets, cf. Table 4. Here, we use the empirical notations for the electronic levels proposed by Lupei et al. [49]: ³H₆ = Z_i (i = 1…13) and ³F₄ = Y_j (j = 1…9). The total Stark splitting of the ground-state, ΔE(³H₆), increased in the R = Y → Lu series, from 788 to 828 cm^-1. The peaks corresponding to electronic transitions were relatively narrow and experienced a notable shift between Y₂O₃ and Lu₂O₃ according to the different crystal-field strengths in these compounds. For the most intense electronic line in absorption, designated as Z₀ → Y₇, the peak positions / widths are 6112.8 / 6.1 cm^-1 (Y₂O₃) and 6134.9 / 5.5 cm^-1 (Lu₂O₃), respectively.

 

Fig. 9. Low-temperature (LT, 12 K) spectroscopy of Tm³⁺ ions in the LuYO₃, Y₂O₃, Lu₂O₃ ceramics: (a) absorption spectra, the ³H₆ → ³F₄ transition; (b) luminescence spectra, the ³F₄ → ³H₆ transition, ZPL – zero-phonon-lines, “+” indicate the assigned electronic transitions.

Download Full Size | PDF

Table 4. Experimental Crystal-Field Splitting of the ³H₆ and ³F₄ Tm³⁺ Multiplets in LuYO₃, Y₂O₃, and Lu₂O₃ Ceramics

View Table | View all tables in this article

For the “mixed” ceramic, the LT spectroscopy does not reveal any signatures of two distinct Tm³⁺ species (e.g., the presence of two sets of electronic transitions or any dependence of the luminescence spectrum on the excitation wavelength). Instead, the electronic lines in both the LT absorption and emission spectra are notably broadened and their peak positions take an intermediate place between those for the parent compounds. E.g., the above-mentioned Z₀ → Y₇ Tm³⁺ absorption line shifts to 6123.9 cm^-1 and its width increases to 34.8 cm^-1 (∼5 times that of Tm:Lu₂O₃). This confirms the formation of a substitutional solid-solution (Lu_1-xY_x)₂O₃ with a mixture of the host-forming cations at the atomic level. The compositional disorder originates from the second coordination sphere of Tm³⁺ ions formed by different sets of Lu³⁺ and Y³⁺ cations with different ionic radii. Thus, strictly speaking, one cannot speak about a predominant substitution of the host-forming Y³⁺ or Lu³⁺ cations by the dopant Tm³⁺ ones. Table 3 presents an attempt to assign the Stark sub-levels of the Tm³⁺ ion in the LuYO₃ ceramic.

A similar analysis for Yb³⁺ ions in a mixed (Lu,Sc)₂O₃ sesquioxide crystal was performed recently in [50,51] however not going down to temperatures of about 10 K. The authors have shown a tendency for increasing the crystal-field strengths in R₂O₃ sesquioxides with decreasing the R³⁺ ionic radius (which agrees with our analysis, as R(Y³⁺) = 0.90 Å and R(Lu³⁺) = 0.861 Å for a VI-fold oxygen coordination). They also observed a monotonous shift of the energies of Yb³⁺ Stark sub-levels with the Lu/Sc ratio. One can expect a monotonouis variation of the barycenter multiplet energies of Tm³⁺ ions in (Lu_1-xY_x)₂O₃ solid-solutions with changing the Lu/Y ratio.

4. Conclusion

In the present work, we tried to reveal the effect of a “mixed” host composition on the spectroscopic properties of the dopant Tm³⁺ ions using the cubic sesquioxide system (R₂O₃) and analyzing a lutetia – yttria (LuYO₃) transparent ceramic fabricated by HIPing. Our study evidences the formation of a substitutional sesquioxide solid-solution with a mixture of cations at the atomic level according to the following findings: (i) the dominant peak in the Raman spectrum characteristic to C-type bixbyite structure takes an intermediate position between those for the parent compounds and is notably broadened; (ii) the absorption and emission spectra of the Tm³⁺ ion exhibit significant inhomogeneous broadening; (iii) at 12 K, the absorption / emission peaks corresponding to electronic transitions are notably broadened, their positions follow the variation of the crystal-field strength in the R = Y → Lu series while no evidence of two distinct Tm³⁺ species in C₂ sites with a second coordination sphere predominantly formed by Lu³⁺ or Y³⁺ is observed.

The Tm:LuYO₃ ceramic benefits from inhomogeneously broadened emission related to the ³F₄ → ³H₆ transition naturally extending above 2 µm (the limit set by the total Stark splitting of the Tm³⁺ ground-state, Δ(³H₆) = 810 cm^-1). In addition, it exhibits a long-wave emission observed up to at least 2.35 µm. The analysis of its spectral shape reveals a phonon sideband at 2.23 µm and a part with a nearly exponential dependence at longer wavelengths. The phonon sideband is associated to the coupling of electronic transitions with the most intense Raman mode of the C-type bixbyite structure (centered at 385.4 cm^-1 for LuYO₃) and the exponential part – with multiphonon-assisted (vibronic) processes. All this leads to smooth (structureless) and broad gain spectra of Tm³⁺ ions supporting the generation of ultrashort (sub-100 fs) pulses. The utilization of multiphonon-assisted emission sidebands of Tm³⁺ ions in sesquioxides may be a viable way for further pulse shortening in mode-locked lasers emitting above 2 µm.