Luminescence Spectroscopy and Origin of Luminescence Centers in Bi-Doped Materials

Bi-doped compounds recently became the subject of an extensive research due to their possible applications as scintillator and phosphor materials. The oxides co-doped with Bi3+ and trivalent rare-earth ions were proposed as prospective phosphors for white light-emitting diodes and quantum cutting down-converting materials applicable for enhancement of silicon solar cells. Luminescence characteristics of different Bi3+-doped materials were found to be strongly different and ascribed to electronic transitions from the excited levels of a Bi3+ ion to its ground state, charge-transfer transitions, Bi3+ dimers or clusters, radiative decay of Bi3+-related localized or trapped excitons, etc. In this review, we compare the characteristics of the Bi3+-related luminescence in various compounds; discuss the possible origin of the corresponding luminescence centers as well as the processes resulting in their luminescence; consider the phenomenological models proposed to describe the excited-state dynamics of the Bi3+-related centers and determine the structure and parameters of their relaxed excited states; address an influence of different interactions (e.g., spin-orbit, electron-phonon, hyperfine) as well as the Bi3+ ion charge and volume compensating defects on the luminescence characteristics. The Bi-related luminescence arising from lower charge states (namely, Bi2+, Bi+, Bi0) is also reviewed.


Introduction
Luminescence of various Bi 3+ -doped materials (alkali halides; alkaline-earth oxides, sulfates and phosphates; tungstates; garnets; perovskites; silicates; borates; vanadates; niobates, etc.) was systematically investigated starting from the 1960s (see, e.g., review papers [1][2][3][4][5][6][7][8][9][10][11] and references therein). Bi 3+ -doped complex oxides, where a trivalent Bi 3+ ion substitutes for a trivalent rare-earth ion, became the subject of special interest and extensive research due to their possible applications as scintillator and phosphor materials. For instance, Bi 3+ -doped garnets [12][13][14][15][16][17][18][19], oxyorthosilicates [20,21], perovskites [22], borates [23] and phosphates [24] were considered as prospective scintillator materials and the materials for X-ray screens due to the presence of an intense and fast Bi 3+ -related luminescence at room temperature. The materials co-doped with Bi 3+ and trivalent rare-earth ions (Ln 3+ : Dy 3+ , Er 3+ , Yb 3+ , Eu 3+ , Sm 3+ , Ho 3+ , Nd 3+ ) were found to be potentially applicable as spectral converters for solar cells and solid state light sources of a new generation, so called white light-emitting diodes, In addition to rich literature dealing with the Bi 3+ -based luminescence in many kinds of crystalline solids mentioned above, there is an increasing interest in the study of bismuth centers with the valence lower than 3 + for their specific emission characteristics given by radiative transitions within the 6p shell. Namely, the Bi 2+ center was ascribed to red luminescence in SrB 4 O 7 host in 1994 [105] and other examples can be found in the review paper of Sun et al. in 2014 [8]. More recently, several studies appeared ascribing the near infrared luminescence to Bi 0 center, e.g., in Ba 2 P 2 O 7 host [106]. Even the Bi + center was ascribed to new luminescence bands in the visible-near infrared spectral region in Ba 2 B 5 O 9 Cl:Bi. These new bands appear among those associated with the Bi 2+ and Bi 0 centers during the annealing cycles in air and the reduction atmosphere [107].
In this review, we compare the characteristics of the Bi 3+ -related luminescence in different compounds; discuss the possible origin of the corresponding luminescence centers; consider theoretical models of their RES and phenomenological models allowing to describe the excited-state dynamics of the Bi 3+ -related centers of different types and determine the structure and parameters of their RES; address an influence of different interactions (e.g., spin-orbit, electron-phonon, hyperfine) as well as the Bi 3+ ion charge and volume compensating defects on the luminescence characteristics.
In Section 6 we review the state-of-art regarding emission characteristics of bismuth centers with the valence lower than 3 + . Table 1. Emission peak positions (E em ), full widths at half maxima (FWHM), the Stokes shifts (S), positions of the lowest-energy excitation band (E exc ), and decay times (τ SC ) of the slow decay component obtained at 4.2 K for the triplet emission of Bi 3+ centers (denoted in this work the UV emission). The parameters of the triplet relaxed excited state (RES): the spin-orbit splitting energy (D), the probabilities of the radiative decay of the metastable (k 1 ) and emitting (k 2 ) levels of the triplet RES, and the zero-temperature nonradiative transition rate (K) between the metastable and emitting levels. The small Stokes shifts of the UV emission are reported also for NaGdO 2 :Bi (S = 0.6 eV), LiScO 2 :Bi (S = 0.9 eV), NaScO 2 :Bi, YAl 3 B 4 O 12 :Bi, and Cs 2 NaYCl 6 :Bi (S < 0.5 eV) (see [68] and references therein). Table 2. Emission peak positions (E em ), full widths at half maxima (FWHM), the Stokes shifts (S), positions of the lowest-energy excitation band (E exc ) and decay times (τ SC ) of the slow decay component obtained at LHeT for the lower-energy triplet emission of Bi 3+ -doped compounds (denoted in this work the VIS emission). The parameters of the triplet RES: the spin-orbit splitting energy (D), the probabilities of the radiative decay of the metastable (k 1 ) and emitting (k 2 ) levels of the triplet RES, and the zero-temperature nonradiative transition rate (K) between the metastable and emitting levels.

Characteristics of the Ultraviolet Luminescence
The characteristics of the higher-energy (UV) emission in various Bi 3+ -doped materials are similar (Table 1). Let us demonstrate them at an example of the Lu 3 Al 5 O 12 :Bi, Y 3 Al 5 O 12 :Bi, and Lu 2 SiO 5 :Bi single crystalline films investigated in [16][17][18]21].
In Figure 1, the emission spectra of Lu 3 Al 5 O 12 :Bi and Y 3 Al 5 O 12 :Bi are shown. The absorption and excitation spectra of Lu 3 Al 5 O 12 :Bi are presented in Figure 2. At low temperatures (T < 100 K in Lu 3 Al 5 O 12 :Bi and Y 3 Al 5 O 12 :Bi), the UV emission arises from the radiative decay of the lowest-energy metastable level corresponding to the 3 P 0 level of a free Bi 3+ ion. Temperature dependences of the maximum position and FWHM of the UV emission of Y 3 Al 5 O 12 :Bi and Lu 2 SiO 5 :Bi are displayed in Figure 3. As the temperature increases, the UV emission spectrum is shifting to higher energies and becomes broader (see the insets to Figure 1). This effect is caused by the thermally stimulated population of the higher 3 P 1 excited level from the lower 3 P 0 level. Further increase of the temperature results in thermal equilibrium between the 3 P 0 and 3 P 1 levels. As the temperature increases further, a gradual lower-energy shift of the emission band takes place. These processes also appear in the decay kinetics of the UV emission. At 4.2 K, the slow component with the decay time τ SC ≈ 1.1 ms is observed in the decay kinetics of the UV emission of Y 3 Al 5 O 12 :Bi ( Figure 4a). The decay time remains constant up to 100 K and then decreases ( Figure 5). This dependence is characteristic for the radiative transitions from a triplet RES where the lowest-energy (metastable) level has much smaller radiative decay probability as compared to the upper (emitting) level (see also [3,68,80]). Indeed, at T < 100 K, the slow decay component is associated with transitions from the metastable level. As the temperature increases, the decay time shortens exponentially due to thermally stimulated transitions between the metastable and emitting levels and reaches a constant value at the temperatures (around 350 K), where the system achieves thermal equilibrium. At higher temperatures, the decay time decreases due to the luminescence thermal quenching. Analogous τ SC (T) dependences were obtained, e.g., for the triplet emission of Pb 2+ centers in alkali halides [131] and Bi 3+ centers in CaO [85], alkaline-earth sulfides [87], and alkaline-earth fluorides [92]. component is associated with transitions from the metastable level. As the temperature increases, the decay time shortens exponentially due to thermally stimulated transitions between the metastable and emitting levels and reaches a constant value at the temperatures (around 350 K), where the system achieves thermal equilibrium. At higher temperatures, the decay time decreases due to the luminescence thermal quenching. Analogous τSC(T) dependences were obtained, e.g., for the triplet emission of Pb 2+ centers in alkali halides [131] and Bi 3+ centers in CaO [85], alkaline-earth sulfides [87], and alkaline-earth fluorides [92]. In the insets, the ultraviolet emission spectra of Bi 3+ centers (normalized) measured at 80 K (solid line), 150 K (dashed line), and 300 K (dotted line). Based on the data reported in [17,18], presented with the publisher's permission. 4  In the insets, the ultraviolet emission spectra of Bi 3+ centers (normalized) measured at 80 K (solid line), 150 K (dashed line), and 300 K (dotted line). Based on the data reported in [17,18], presented with the publisher's permission. Excitation spectra (normalized) of (b) Lu3Al5O12:Bi and (c) Y3Al5O12:Bi measured at 80 K for different emission spectra regions shown in the legends. Based on the data reported in [16][17][18], presented with the publisher's permission.   Excitation spectra (normalized) of (b) Lu3Al5O12:Bi and (c) Y3Al5O12:Bi measured at 80 K for different emission spectra regions shown in the legends. Based on the data reported in [16][17][18], presented with the publisher's permission.  [17,18,21], presented with the publisher's permission.   [17,18,21], presented with the publisher's permission.       [16][17][18]. Presented with the publisher's permission.    [16][17][18]. Presented with the publisher's permission.

Dynamics of the Triplet Excited State of Bi 3+ Centers
Thermally stimulated transitions between the metastable and emitting minima of the triplet RES and between the excited and ground state of the luminescence centers responsible for the UV emission reveal themselves in the temperature dependences of the luminescence spectra and decay kinetics. The excited states dynamics of the luminescence center responsible for the UV emission are described within the phenomenological model sketched in Figure 6.

Dynamics of the Triplet Excited State of Bi 3+ Centers
Thermally stimulated transitions between the metastable and emitting minima of the triplet RES and between the excited and ground state of the luminescence centers responsible for the UV emission reveal themselves in the temperature dependences of the luminescence spectra and decay kinetics. The excited states dynamics of the luminescence center responsible for the UV emission are described within the phenomenological model sketched in Figure 6. The time evolution of the populations N1, N2 of the excited levels 1 and 2, respectively, can be described by the following rate equations: where k1, k2, k12, k21, and k1 (2)x are radiative transition rates from levels 1,2, non-radiative rates of phonon assisted transitions between the radiative level 2 and metastable level 1 and the quenching The time evolution of the populations N 1 , N 2 of the excited levels 1 and 2, respectively, can be described by the following rate equations: where k 1 , k 2 , k 12 , k 21 , and k 1 (2)x are radiative transition rates from levels 1,2, non-radiative rates of phonon assisted transitions between the radiative level 2 and metastable level 1 and the quenching channel from the level 1(2), respectively. Non-radiative transitions between levels 1,2 can be written as: where K, n, D are the zero-temperature transition rate between the levels 1 and 2, the Bose-Einstein factor, and energy spacing between the levels, respectively. Non-radiative quenching channel is considered in the usual barrier form: with K 1(2)x being a frequency factor and E 1(2)x the height of the barrier. Application of the two-excited-level models on the temperature evolution of the UV luminescence intensity and decay times allowed determination of characteristic parameters of the corresponding triplet RES, e.g., the energy separation (D) between the emitting and metastable levels of the triplet RES, the rates of the radiative (k 1 , k 2 ) and nonradiative (K) transitions from these levels, and activation energy E 1(2)x for the luminescence thermal quenching (for more details, see Refs. [9,13,[16][17][18][20][21][22]70,71]). Some parameters of the triplet RES corresponding to the UV emission of Bi 3+ -doped compounds are shown in Table 1.
As evident from Table 1, the triplet RES responsible for the UV emission is characterized by very large (~10 2 meV) energy distance D between the metastable and emitting levels of the triplet RES which can be explained by extremely large spin-orbit interaction energy characteristic for a free Bi 3+ ion (ξ = 2.102 eV [4]). Therefore, the higher-energy (UV) emission of all the investigated materials can surely be ascribed to the electronic transitions from the triplet RES of Bi 3+ corresponding to the 3 P 1,0 → 1 S 0 transitions of a free Bi 3+ ion.
In some Bi 3+ -doped compounds, a fast (ns) component is observed at low temperatures in the UV luminescence decay. This component is associated with transitions from the emitting level of the triplet RES related to the 3 P 1 level of a free Bi 3+ ion. The electronic transitions between the ground state ( 1 S 0 ) and the 3 P 1 -related excited state are partly allowed due to mixing of the triplet 3 P 1 state with the singlet 1 P 1 state by the spin-orbit interaction. Due to a strong spin-orbit interaction, the probability of the radiative decay of the emitting level (k 2 ) is relatively large (see Table 1). The radiative transitions from the metastable 3 P 0 -related state can occur due to mixing of the 3 P 1 -and 3 P 0 -related states by the vibronic interaction with the non-totally symmetric vibrations or by the hyperfine interaction (see, e.g., [132] and references therein). The only stable Bi isotope 209 Bi has a nuclear spin of I = 9/2. Therefore, in Bi 3+ -doped compounds with a weak vibronic interaction, such as alkali-earth oxides, sulfides, fluorides (see, e.g., [84,86,[88][89][90][91][92] and references therein), ScBO 3 :Bi [78], LuBO 3 :Bi [78], Cs 2 NaYBr 6 :Bi [97], Cs 2 NaLaCl 6 :Bi [97], Cs 2 NaYCl 6 :Bi [98], NaScO 2 :Bi, and YAl 3 B 4 O 12 :Bi (see also [68] and references therein), where the Stokes shift is extremely small (see Table 1) and even a vibronic structure of the emission and excitation spectra is observed at low temperatures, mainly the hyperfine interaction can be expected to be responsible for the radiative decay of the metastable 3 P 0 -related level. The influence of the hyperfine interaction on the probability of the radiative 3 P 0 → 1 S 0 transitions in Bi 3+ -doped alkali-earth oxides was investigated in [133].

Relaxed Excited States Models
Hitherto, two models have been proposed to describe RES of the ns 2 -ion-doped ionic crystals with strongly different electron-phonon and spin-orbit interactions. The systems with a strong spin-orbit interaction and a very weak electron-phonon interaction can be described within the RES model, proposed by Seitz [134], which considers the spin-orbit interaction in RES being dominant. In this model, the excited states of the luminescence center originate from the 3 P 0 , 3 P 1 , 3 P 2 , and 1 P 1 levels of a free ns 2 ion, which are split in the crystal field of the corresponding symmetry. For the degenerate energy levels, the Jahn-Teller effect is taken into account as a perturbation. The totally symmetric 3 P 0 state is not degenerate, therefore it cannot be Jahn-Teller active. The configuration coordinates (q) of the 3 P 0 and 1 S 0 minima in this model should coincide.
The analysis of results obtained in a huge number of works (see, e.g., review papers [131,132,135]) have convincingly confirmed the suggestion of Seitz that the absorption processes in ns 2 -ion-doped compounds can be described in the approximation of a weak crystal field. This means that the spin-orbit interaction must be considered to be dominant in the unrelaxed excited state of the luminescence center. However, this model cannot adequately describe the luminescence characteristics and the relaxed excited state structure of the systems with a strong electron-phonon interaction.
For the systems of this type, a new RES model was proposed by Hizhnyakov [136], and the RES theory was developed in [135,137]. In this theory, the interaction of impurity optical electrons with non-totally symmetric vibrations is considered to be dominant in the relaxed excited state, while the spin-orbit, hyperfine, and other interactions are taken as small perturbations. As a result, the Jahn-Teller minima of different symmetries can be formed on the adiabatic potential energy surface of the singlet ( 1 P) and triplet ( 3 P) excited states. Due to the spin-orbit interaction, each Jahn-Teller minimum of the triplet RES is split into the upper emitting level and the lower metastable level, corresponding to the 3 P 1 and 3 P 0 levels of a free Bi 3+ ion, respectively ( Figure 7, solid lines). The applicability of this model was confirmed by the systematic experimental study of luminescence characteristics of ns 2 -ion-doped alkali halide crystals by the methods of time-resolved polarization spectroscopy in a wide temperature range, down to 0.4 K (see, e.g., [132,138]). It was also shown that in the Tl + , Pb 2+ , Bi 3+ centers with a strong spin-orbit interaction, each metastable minimum of the triplet RES may not lie exactly under the corresponding emitting minimum, like in the Ga + , In + , Ge 2+ , Sn 2+ centers with a weak spin-orbit interaction. Instead, it can be shifted towards smaller coordinate q values with respect to the emitting minimum ( Figure 7). Due to that, the energy barriers for the thermally stimulated transitions between the metastable minima of various orientations can be much smaller than those between various emitting minima (see, e.g., [85,132,139] and references therein). Thus, in the ns 2 -ion-doped compounds characterized by a strong electron-phonon interaction, different models have to be used for the description of the structure and properties of the unrelaxed and relaxed excited states of a luminescence center. Thus, in the ns 2 -ion-doped compounds characterized by a strong electron-phonon interaction, different models have to be used for the description of the structure and properties of the unrelaxed and relaxed excited states of a luminescence center.
Since a free Bi 3+ ion is characterized by the largest spin-orbit interaction energy (ξ) among all ns 2 ions, different models should also be used for the description of the triplet RES responsible for the UV luminescence of Bi 3+ centers in the compounds with an extremely small electron-phonon interaction (e.g., in alkaline-earth oxides, sulfates, fluorides) with respect to the materials with a relatively strong electron-phonon interaction (large FWHM and S), such as alkali halides, oxyorthosilicates, etc. For the description of the latter type systems, the model [135,137] should be considered. However, usually only the model [134] is used in the literature for the description of the UV luminescence of all Bi 3+ -doped materials.
To investigate the applicability of the model [135,137] to the centers with extremely strong spin-orbit interaction, the luminescence characteristics of two Bi 3+ -doped crystals with strongly different electron-phonon interaction (KCl:Bi and CaO:Bi) were compared in [85]. As evident from Figure 8, the characteristics of KCl:Bi and CaO:Bi are different. In the emission spectrum of KCl:Bi at 4.2 K, the broad (FWHM = 0.4 eV) complex band located around 2.5 eV is observed (Figure 8a, curve 1) (see also [140]). The lowest-energy excitation band of this emission is located around 3.8 eV (curve 2), thus, S ≈ 1.3 eV.  In the emission spectrum of CaO:Bi at 4.2 K, the narrow (FWHM = 0.14 eV) strong 3.1 eV and weak 3.3 eV bands are observed ( Figure 8b, curves 1,1'). Their excitation spectra coincide (curves 2,2'). The lowest-energy excitation band is located at 3.5 eV, i.e., S = 0.4 eV. The intensity of the 3.1 eV emission remains constant up to 100 K and then decreases. The reduction of the 3.1 eV emission is accompanied with the 3.3 eV emission enhancement. The intensity redistribution between the 3.1 eV and 3.3 eV emissions around 140 K points to the thermally stimulated transitions between the corresponding levels.
Two fast (17 ns and 27 ns) and two slow (1.38 ms and 2.8 ms) components were observed at 4.2 K in the KCl:Bi emission decay (Figure 9a). Their excitation spectra practically coincide. The emission spectra of two fast decay components are located at 2.54 eV and 2.46 eV, respectively, and the emission spectra of two slow decay components, at 2.84 eV and 2.68 eV, respectively. The decay times of the slow components are constant up to about 60 K and then decrease exponentially. The D value is estimated to be of the order of 10 2 meV.  , and E em = 3.3 eV (curve 2'). Based on the data reported in [85]. Presented with the publisher's permission.
In the emission spectrum of CaO:Bi at 4.2 K, the narrow (FWHM = 0.14 eV) strong 3.1 eV and weak 3.3 eV bands are observed ( Figure 8b, curves 1,1'). Their excitation spectra coincide (curves 2,2'). The lowest-energy excitation band is located at 3.5 eV, i.e., S = 0.4 eV. The intensity of the 3.1 eV emission remains constant up to 100 K and then decreases. The reduction of the 3.1 eV emission is accompanied with the 3.3 eV emission enhancement. The intensity redistribution between the 3.1 eV and 3.3 eV emissions around 140 K points to the thermally stimulated transitions between the corresponding levels.
Two fast (17 ns and 27 ns) and two slow (1.38 ms and 2.8 ms) components were observed at 4.2 K in the KCl:Bi emission decay (Figure 9a). Their excitation spectra practically coincide. The emission spectra of two fast decay components are located at 2.54 eV and 2.46 eV, respectively, and the emission spectra of two slow decay components, at 2.84 eV and 2.68 eV, respectively. The decay times of the slow components are constant up to about 60 K and then decrease exponentially. The D value is estimated to be of the order of 10 2 meV. Results obtained in [85] confirmed the conclusion [84] that the Seitz model [134], considering the spin-orbit interaction in the triplet RES being dominant, should be used for the description of the luminescence characteristics of CaO:Bi. These data also indicate that in each Jahn-Teller minimum of the triplet RES of KCl:Bi, the metastable minimum is located under emitting minimum ( Figure 7). The radiative transitions from these minima result in appearance of the slow and fast decay component, respectively. The fast component is strongly polarized in the <100> direction [96]. The shift of the slow component emission spectra to higher energies with respect to those of the fast components, as well as a very small polarization degree of the slow decay component are caused by the shift of the metastable minima towards a smaller configuration coordinate q values with respect to the emitting minimum as well as the decrease of the energy barriers between the metastable minima of different orientation as compared with the emitting minima (see Figure 7). Analysis of these data allows to conclude that the theoretical model [135,137] is still valid in the case of KCl:Bi, despite the strong spin-orbit interaction.
Comparison of the Stokes shifts and FWHM of emission bands as well as the values of RES parameters presented in Table 1 indicate that the Bi 3+ center in aluminum garnets and oxyorthosilicates can be considered as an intermediate case between Bi 3+ -doped CaO and KCl. These data also allow to conclude that the electron-phonon interaction in oxyorthosilicates and Y4Al2O9:Bi is noticeably stronger than that in garnets. Indeed, the larger probability (k1) of the radiative decay of the triplet RES metastable minima points to a stronger electron-phonon interaction in these compounds. In addition, a smaller value of the spin-orbit splitting energy (D) of the triplet RES of Bi 3+ center in oxyorthosilicates and Y4Al2O9 as compared with aluminum garnets is caused by stronger suppression of the spin-orbit interaction by the electron-phonon interaction (see [135,137]). Thus, the  Figure 9b). These components arise from RES, related to 3 P 0 and 3 P 1 excited levels of a free Bi 3+ ion. Decay time of the slow component decreases exponentially at T > 100 K due to the thermally stimulated 3 P 0 → 3 P 1 transitions with the activation energy of about 152 meV.
Results obtained in [85] confirmed the conclusion [84] that the Seitz model [134], considering the spin-orbit interaction in the triplet RES being dominant, should be used for the description of the luminescence characteristics of CaO:Bi. These data also indicate that in each Jahn-Teller minimum of the triplet RES of KCl:Bi, the metastable minimum is located under emitting minimum ( Figure 7). The radiative transitions from these minima result in appearance of the slow and fast decay component, respectively. The fast component is strongly polarized in the <100> direction [96]. The shift of the slow component emission spectra to higher energies with respect to those of the fast components, as well as a very small polarization degree of the slow decay component are caused by the shift of the metastable minima towards a smaller configuration coordinate q values with respect to the emitting minimum as well as the decrease of the energy barriers between the metastable minima of different orientation as compared with the emitting minima (see Figure 7). Analysis of these data allows to conclude that the theoretical model [135,137] is still valid in the case of KCl:Bi, despite the strong spin-orbit interaction.
Comparison of the Stokes shifts and FWHM of emission bands as well as the values of RES parameters presented in Table 1 indicate that the Bi 3+ center in aluminum garnets and oxyorthosilicates can be considered as an intermediate case between Bi 3+ -doped CaO and KCl. These data also allow to conclude that the electron-phonon interaction in oxyorthosilicates and Y 4 Al 2 O 9 :Bi is noticeably stronger than that in garnets. Indeed, the larger probability (k 1 ) of the radiative decay of the triplet RES metastable minima points to a stronger electron-phonon interaction in these compounds. In addition, a smaller value of the spin-orbit splitting energy (D) of the triplet RES of Bi 3+ center in oxyorthosilicates and Y 4 Al 2 O 9 as compared with aluminum garnets is caused by stronger suppression of the spin-orbit interaction by the electron-phonon interaction (see [135,137]). Thus, the structure and properties of the triplet RES of these materials and the characteristics of their luminescence might be described in terms of the theory [135,137] which considers a strong Jahn-Teller effect to be a dominant interaction in the triplet RES.

Characteristics of the Visible Luminescence
Characteristics of the lower-energy luminescence (the VIS emission) in various Bi 3+ -doped materials are similar and presented in Table 2. Let us demonstrate them at an example of the Lu 3 Al 5 O 12 :Bi and Y 3 Al 5 O 12 :Bi single crystalline films investigated in [16][17][18].
In Lu 3 Al 5 O 12 :Bi, two broad VIS emission bands with large Stokes shifts are located at 2.6 eV and 2.75 eV (Figure 1a). In Y 3 Al 5 O 12 :Bi, analogous bands are observed at 2.63 eV and 2.75 eV (Figure 1b). Comparison of the UV and VIS emission spectra shows that FWHM and S values of the VIS emission are several times larger as compared with those of the UV emission. The lowest-energy excitation band of the VIS emission is always slightly shifted to lower-energies with respect to that of the UV emission ( Figure 2b,c) (compare also Tables 1 and 2). The VIS emission is much more effectively excited in the higher-energy absorption bands as compared with the UV emission. As the temperature increases, the intensity redistribution takes place between the UV emission and the lower-energy VIS emission of Lu 3 Al 5 O 12 :Bi and Y 3 Al 5 O 12 :Bi (see, e.g., Figure 3c).
The decay curves of the VIS emissions in Y 3 Al 5 O 12 :Bi measured at 4.2 K are shown in Figure 4b,c. At 4.2 K, the components with the decay times τ SC ≈ 33 and 38 µs dominate in the decay kinetics of the 2.63 eV and 2.75 eV emissions, respectively. At T < 6 K, the decay times remain constant (Figure 5b,c) which means that the radiative transitions take place from the metastable levels. As the temperature increases, the decay times shorten exponentially due to thermally stimulated transitions between the metastable and emitting levels and reach a constant value at the temperatures (around 100 K), where the system achieves thermal equilibrium. Such temperature dependences are characteristic for the radiative transitions from the triplet RES of a luminescence center (see also [7,9,78,83,93]). At higher temperatures, the decay time decreases due to the luminescence thermal quenching. Analogous τ SC (T) dependences were obtained, e.g., for the triplet emission of Ga + -and In + -doped alkali halides (see, e.g., [132]) due to a small spin-orbit interaction energy characteristic for free Ga + and In + ions (ξ ≈ 0.2-0.3 eV, see, e.g., [4]). Indeed, D = 0.33-0.67 meV was obtained in [132] for Ga + centers and D = 2.15-3.04 meV, for In + centers.
In Lu 3 Al 5 O 12 :Bi, as well as in some other Bi 3+ -doped materials (e.g., Lu 2 SiO 5 :Bi, Gd 3 Ga 5 O 12 :Bi), the participation of the singlet exciton state in the VIS luminescence decay kinetics is also evident. This is caused by the fact that in case of excitons, the singlet state is located close to the triplet state ( Figure 10b). The probability (k 3 ) of its radiative decay is found to be 10 8 -5 × 10 9 s −1 and the energy distance between the singlet and triplet states E = 75-150 meV (see [9,16,21,74]).
for Ga + centers and D = 2.15 -3.04 meV, for In + centers.
In Lu3Al5O12:Bi, as well as in some other Bi 3+ -doped materials (e.g., Lu2SiO5:Bi, Gd3Ga5O12:Bi), the participation of the singlet exciton state in the VIS luminescence decay kinetics is also evident. This is caused by the fact that in case of excitons, the singlet state is located close to the triplet state ( Figure  10b). The probability (k3) of its radiative decay is found to be 10 8 -5 × 10 9 s −1 and the energy distance between the singlet and triplet states E = 75-150 meV (see [9,16,21,74]).

Dynamics of the Bi 3+ -Related Exciton-like States
Thermally stimulated transitions between the metastable and emitting minima of the triplet RES, triplet and singlet excited states, and excited and ground states of the luminescence center responsible for the VIS emission reveal themselves in the temperature dependences of the VIS emission spectra and decay kinetics. The phenomenological models are proposed to describe the excited-state dynamics of these centers. Application of the two-or three-excited-level models on the temperature evolution of the VIS luminescence decay times allows determination of characteristic parameters of the corresponding RES (the energy separations between the excited states and the rates of the radiative and non-radiative transitions from these states). The excited states dynamics of the luminescence center responsible for the VIS emission is described within the phenomenological model sketched in Figure 10.
In case of the two-excited-level model (Figure 10a), the time evolution of the populations N 1 , N 2 of the excited levels 1 and 2, respectively, can be described by the rate equations given by Equation (1) with consideration of Equations (2) and (3). In case of the three-excited-level model (Figure 10b), the time evolution of the populations N 1 , N 2 , N 3 of the excited levels 1,2 and 3, respectively, can be described by the following rate equations: where analogous parameters have the same meaning as in (1). In addition, k 3 is the radiative transition rate from the singlet level 3 and k 3(2)1 , k 1(2)3 , are non-radiative rates of transitions between the levels 3(2) and 1: E is the energy distance between the triplet and singlet levels. Since D<<E we consider The values of some parameters are presented in Table 2. A strong difference between the structure and parameters of RES responsible for the UV and VIS emissions clearly indicates their different origin.

On the Origin of the Excited States Responsible for the VIS Luminescence
In most of the studies, a possible origin of the unrelaxed excited states, responsible for the lowest-energy excitation band of the VIS emission, was mainly considered for various Bi 3+ -doped compounds (see, e.g., [7,64]). For example, the lowest energy excitation band of the VIS emission of YNbO 4 :Bi was ascribed to the Bi 3+ (6s 2 ) → Nb 5+ (d 0 ) MMCT (see also [141]). It should be noted that the consideration of absorption and emission bands of some Bi 3+ -doped complex oxides as an electron transfer between Bi 3+ and host lattice transition metal ions was proposed about 50 years ago [1] (see also [6,141,142]).
In [65], an empirical model, proposed earlier for Pr 3+ -and Tb 3+ -doped d 0 closed-shell transition metal compounds, was applied to predict the energy position of the MMCT bands in various Bi 3+ -doped closed-shell d 0 transition metal (M n+ ) complex oxides. The dependence of the energy positions of the Bi 3+ -related absorption bands E abs on the ratio between the optical electronegativities X opt (M n+ ) of the d 0 metal cations and the shortest Bi 3+ -M n+ interatomic distances was found to be linear. It was described by the following empirical equation: It was concluded that any Bi 3+ -related absorption band satisfying this equation is of the MMCT origin.
This model was developed further in [7] where the structural characteristics of the host lattice, anion relaxation resulting from Bi 3+ doping, and electronegativities and coordination numbers of the Bi 3+ and M n+ ions in the compounds were taken into account. For the metals with the coordination number CN' (M n+ ) = 4, the Equation (1) was modified to: where d corr is the shortest distance between Bi 3+ and M n+ ions corrected to account for the effect of anion relaxation due to Bi doping. For the metals with the coordination numbers CN' (M n+ ) > 4, the following equation was proposed: MMCT(Bi 3+ , cm −1 ) = 55,000 -45,500 {X CN'>4 (M n+ )/d corr }.
In [7,64,65], the energies of the 1 S 0 → 3 P 1 transitions of a Bi 3+ ion and the Bi 3+ → M n+ MMCT transitions were calculated for many Bi 3+ -doped compounds. In a few cases, e.g., in YVO 4 :Bi, these energies were obtained to be very close (3.779 eV and 3.778 eV, respectively [65]). As both these energies are close to the energy of the lowest-energy excitation band of the VIS emission in YVO 4 :Bi (3.78 eV), in principle, both the 1 S 0 → 3 P 1 transitions of a Bi 3+ ion and the Bi 3+ → V 5+ MMCT transitions could be considered as responsible for the lowest-energy Bi 3+ -related excitation band in YVO 4 :Bi. In [1,3], this band was ascribed to MMCT and in [76], to the 1 S 0 → 3 P 1 transitions. However, in most of the considered materials, the position of the lowest-energy excitation band of the VIS emission and the MMCT energy (see, e.g., [10,65]) are significantly different. For example, in Lu 3 Al 5 O 12 :Bi, the MMCT energy is 5.95 eV [10] while the lowest excitation band of the 2.6 eV emission is located at 4.6 eV. The same is true for Y 3 Al 5 O 12 :Bi. In YPO 4 :Bi, the MMCT energy (7.3 eV [10]) is also much higher as compared with the position (≈5.5 eV, Table 2) of the lowest excitation band.
In Bi 3+ -doped compounds, the UV emission is concluded to arise from the radiative decay of the triplet RES of a single Bi 3+ ion, and the lowest-energy excitation band of this emission corresponds to the 1 S 0 → 3 P 1 transitions of Bi 3+ . According to [7,64], the MMCT occurs from the ground 1 S 0 state of a Bi 3+ ion to the bottom of the conduction band (CB) formed by the energy levels of d 0 or d 10 host lattice ions, and the energy of the MMCT is defined as the energy difference between the 1 S 0 and MMCT states ( Figure 11). Therefore, in general, the lowest-energy excitation bands of the UV and VIS emissions should not coincide as these bands arise due to processes of different origin. However, in all the Bi 3+ -doped compounds, where both the UV and VIS emissions are present, their lowest-energy excitation bands are close (see, e.g., Figure 2b,c and compare the E exc values in Tables 1 and 2). In some cases, they practically coincide (see, e.g., [24,120]), and the intensity redistribution is observed between the UV and VIS emissions (see, e.g., [9,17,18,93,114,115,120] and Figure 3c). For example, the lowest excitation bands of the UV and VIS emissions completely coincide in La 2 Zr 2 O 7 :Bi [120]. This indicates that the VIS emission of La 2 Zr 2 O 7 :Bi, ascribed in [120] to the impurity trapped exciton, is excited in the 1 S 0 → 3 P 1 absorption band of Bi 3+ (4.32 eV) [7,64,65] even at 4.2 K, despite the fact that the lowest calculated MMCT energy in this material is 4.88 eV [65] or 5.06 eV [10]. In our opinion, the intensity redistribution between the UV and VIS emissions observed in this work can be caused by the thermally stimulated release of an electron from the 3 P level of Bi 3+ into CB. From the I(T) dependence in the T < 50 K temperature range, the activation energy of this process can be estimated as ≈ 2 meV. This value can correspond to the energy distance between the triplet RES of Bi 3+ and CB. The reverse intensity redistribution observed in [120] at higher temperature can be caused by thermally stimulated transitions from the localized exciton state to the 3 P 1 state of Bi 3+ over the energy barrier of about 27 meV.  According to [7], the UV emission appears when the 3 P1 level of Bi 3+ is located below CB. In this case, the energy of the MMCT should always be higher as compared with the energy of the 1 S0 → 3 P1 transition ( Figure 11). However, in most of the investigated cases, the excitation band of the VIS emission is slightly shifted to lower energies as compared to that of the UV emission (see, e.g., [14,15,19,24,64,78,[93][94][95][96]104,109,110,119,121,140,141] and Figure 2b,c).
Thus, the data considered above allow us to suggest that the electron transitions, corresponding to the 1 S0 → 3 P1 transition of a free Bi 3+ ion, are most probably responsible for the lowest excitation band of both the UV and VIS emissions. The MMCT transitions as well as the 1 S0 → 3 P2 and 1 S0 → 1 P1 transitions of a Bi 3+ ion, resulting in the electron release into CB, can be responsible for the higherenergy excitation bands of the VIS luminescence.
Detailed investigations of the VIS emission by the time-resolved spectroscopy methods in a wide temperature range (down to 0.4 K) and determination of the RES parameters carried out in [7,9, Figure 11. Schematic presentation of energy levels in Bi 3+ -doped compounds proposed in [7].
According to [7], the UV emission appears when the 3 P 1 level of Bi 3+ is located below CB. In this case, the energy of the MMCT should always be higher as compared with the energy of the 1 S 0 → 3 P 1 transition ( Figure 11). However, in most of the investigated cases, the excitation band of the VIS emission is slightly shifted to lower energies as compared to that of the UV emission (see, e.g., [14,15,19,24,64,78,[93][94][95][96]104,109,110,119,121,140,141] and Figure 2b,c).
Thus, the data considered above allow us to suggest that the electron transitions, corresponding to the 1 S 0 → 3 P 1 transition of a free Bi 3+ ion, are most probably responsible for the lowest excitation band of both the UV and VIS emissions. The MMCT transitions as well as the 1 S 0 → 3 P 2 and 1 S 0 → 1 P 1 transitions of a Bi 3+ ion, resulting in the electron release into CB, can be responsible for the higher-energy excitation bands of the VIS luminescence.
Detailed investigations of the VIS emission by the time-resolved spectroscopy methods in a wide temperature range (down to 0.4 K) and determination of the RES parameters carried out in [7,9,13,[16][17][18]21,[72][73][74][75][76][77][78]83,93,95,99,156] allow us to make a justified conclusion on the exciton-like origin of the emission in the considered Bi 3+ -doped compounds. The VIS emission was interpreted as the luminescence of an exciton localized around a Bi 3+ -related center. Indeed, since a free Bi 3+ ion has the largest spin-orbit interaction energy among all the other ns 2 ions (ξ = 2.102 eV [4]), a very small energy distance (D = 0.34-5.7 meV, see Table 2) between the metastable and emitting levels of the triplet RES responsible for the VIS emission can only be explained by the exciton-like origin of this emission. In addition to the small spin-orbit interaction energy, a strong exciton-phonon interaction is also characteristic for excitons which explains the large Stokes shift and FWHM of the VIS emission.
The structure and parameters of the triplet RES should be similar for all exciton-like emissions in the same host material. This was clearly demonstrated on the example of caesium iodides in [160,161] where similar RES parameters were obtained for the self-trapped excitons and for the excitons localized around various intrinsic and impurity defects. Therefore, in Table 3, available data on the exciton-like luminescence in some undoped materials are collected to compare them with corresponding characteristics of the Bi 3+ -related VIS emission.   Unfortunately, for most of these materials, the RES parameters were not determined since for that task investigations of the luminescence decay kinetics at temperatures well below 4.2 K are needed. For example, in PbWO 4 the slow component decay time (τ SC ) reaches its maximum value only at T < 0.6 K [178][179][180].
Comparison of the data presented in Table 3 with those in Table 2 demonstrates that the exciton-like emissions in an undoped material and the lower-energy (VIS) emission of the same host material, Bi 3+ -doped have very close values of D, FWHM, and S. This is an additional confirmation of the exciton-like origin of the broad lower-energy emission bands in these compounds (see also [69,83]).
A drastic difference in the D values (up to two orders of magnitude), appearing in the decay kinetics of the luminescence arising from the triplet RES of an impurity ion with respect to that arising from the triplet state of an exciton localized around the impurity ion, was also clearly demonstrated for Tl + - [188,189] and Pb 2+ -doped [190][191][192] caesium halides. In these compounds both types of the impurity-related emission bands mentioned above were observed in a single system. The Stokes shifts and FWHM related to the two types of the emission bands were found to be considerably different as well.
Let us consider possible mechanisms of the processes resulting in the appearance of the exciton-like luminescence in Bi 3+ -doped compounds.

On possible Mechanisms of Processes Responsible for the VIS Luminescence
We suggest that the VIS emission appears under excitation of a Bi 3+ ion with the energy (hν exc ) which allows an electron delocalization from the excited state of Bi 3+ into CB and its subsequent immediate recombination with the hole remained at the Bi 3+ ion (the Bi 4+ hole center). For that, the lowest-energy relaxed excited state of Bi 3+ (responsible for the UV emission) should be located inside CB (see Figure 7). As a result of the electron-hole recombination around the Bi 3+ ion, a localized exciton (ex 0 Bi 3+ ) is created. The radiative decay of the lowest-energy triplet localized exciton state results in the appearance of the Bi 3+ -related exciton-like emission (hν em ) in case the corresponding RES is located below the bottom of CB: In the considered case, only the VIS emission can appear. In case the relaxed excited states of both the Bi 3+ ion and the localized exciton ex 0 Bi 3+ are located below or close to the bottom of CB, both the UV and VIS emissions can appear. The UV/VIS emission intensity ratio depends not only on the RES position with respect to the bottom of CB but also on the rate of vibronic relaxation in the Bi 3+ excited state and the probability of an electron delocalization from this state. No VIS emission can appear in case the lowest-energy Bi 3+ -related level (corresponding to the 3 P 1 level of a free Bi 3+ ion) is located well below the bottom of CB.
Similar mechanism of the appearance of the VIS emission was proposed in [155] (see also [120]), where this emission was ascribed to the impurity-bound exciton recombination. It was suggested that the emitting level of InBO 3 :Bi is situated either close to or inside CB of the host lattice, so that after excitation the luminescence center can get ionized. According to [155], this results in formation of an impurity-trapped exciton, with the hole located at the luminescence center and the electron located in its neighborhood.
It should be noted that the above-mentioned Bi 3+ -related center can be not only a single Bi 3+ ion, but also a dimer {Bi 3+ -Bi 3+ } or a Bi 3+ ion located close to a crystal lattice defect d ({Bi 3+ -d} center). In this case, the localized excitons of the type of ex 0 {Bi 3+ -Bi 3+ } and ex 0 {Bi 3+ -d} can also be created and their radiative decay can be accompanied by the VIS emission. This means that several overlapping emission bands of exciton-like origin can appear in Bi 3+ -doped compounds resulting in a complex structure of the VIS emission band.
The luminescence of the localized exciton of the type of ex 0 {Bi 3+ -d} can appear with the highest probability in case the Bi 3+ ion substitutes for a divalent or monovalent host lattice ion where the excess charge of Bi 3+ should be compensated by some defect (d). Probably, this is the case of Sr 3 Ga 4 O 9 :Bi [58], where two emission bands could arise not only from single Bi 3+ ions, substituting for Sr 2+ ions in different lattice sites, but also from the localized excitons of the type of ex 0 Bi 3+ and ex 0 {Bi 3+ -d}. In the materials of this type, an electron transfer from the valence band to the Bi 3+ ion, resulting in the formation of a stable Bi 2+ center, is also possible. In more detail, this process was considered for PbWO 4 :Bi in Ref. [73] (see Appendix A). In this case, besides the radiative electron-hole recombination, resulting in the appearance of the Bi 3+ -related exciton-like luminescence, the electron Bi 2+ centers and the self-trapped holes can also be optically created. These centers were indeed detected in EPR [73,193].

On the Dependence of the UV/VIS Emission Intensity Ratio on the Band-Gap and Band-Edge Energy
According to Figure 7, the UV/VIS emission intensity ratio should strongly depend on the position of the lowest-energy RES of Bi 3+ with respect to the CB edge. Therefore, it could also depend on the band gap energy E g of the host material and increase with increasing E g [154]. The best materials to investigate these dependences could be multicomponent garnets where both the CB edge energy and E g can be changed by variation of their composition (see, e.g., [194][195][196][197][198][199]). Let us consider some examples.
The UV/VIS emission intensity ratio was found to be much larger in Y 3 Al 5 O 12 :Bi (E g ≈ 7.7 eV [10,154]) as compared to Lu 3 Al 5 O 12 :Bi (E g ≈ 7.9 eV [10,154]) (compare Figure 1a and b). This could mean that the triplet RES of Bi 3+ is located closer to the bottom of CB in Lu 3 Al 5 O 12 as compared to Y 3 Al 5 O 12 , despite the larger band gap in Lu 3 Al 5 O 12 .
In Ref. [200], the effect of Ga 3+ doping on the photoluminescence properties of Y 3 Al 5-x Ga x O 12 :Bi was studied. It was shown that the incorporation of the Ga 3+ ions results in a strong reduction of the band gap edge (by 1 eV from Y 3 Al 5 O 12 :Bi to Y 3 Ga 5 O 12 :Bi). In [195][196][197][198][199], it was shown that the increasing Ga content also results in a drastic decrease of the CB bottom energy in different Ce 3+ -doped multicomponent garnets. The presence of two-UV and VIS-emission bands in Y 3 Al 5 O 12 :Bi with the close lowest-energy excitation bands indicates that the triplet RES of Bi 3+ should be located close to the bottom of CB of Y 3 Al 5 O 12 [7]. In this case, the reduction of E g and the CB bottom energy with the increasing Ga content should result in the disappearance of the UV emission in Y 3 Ga 5 O 12 :Bi. Indeed, no UV emission was observed in [12]. However, according to [13,14,16,17], both in Y 3 Al 5 O 12 :Bi (E g ≈ 7.7 eV [10,154]) and in Y 3 Ga 5 O 12 :Bi (E g ≈ 6.6 eV [10]), the UV emission is much stronger than the VIS emission (compare Figure 12a and b). In [110], the UV/VIS emission intensity ratio is about 2.
The reason of such strong difference in the experimental data [12], [13,14], and [110] is not clear.
In Ce 3+ -doped multicomponent garnets [196], the influence of Gd on the E g value and the position of the lowest-energy 5d 1 excitation band of Ce 3+ with respect to CB was found to be much weaker [194] and dependent on the Ga and Gd content [196,199]. In general, the 5d 1 -CB energy distance slightly increases with the increasing Gd content. In case the same dependence is valid for the Bi 3+ -doped gallates, the luminescence spectra of Y 3 Ga 5 O 12 :Bi and Gd 3 Ga 5 O 12 :Bi should be similar (like in [12]). However, according to [13,14,74,75,110], in Bi 3+ -doped (Gd,Y) 3 Ga 5 O 12 multicomponent garnets, the UV/VIS emission intensity ratio decreases drastically with the increasing Gd content (compare Figure 12b and c). The absence of the UV emission in Gd 3 Ga 5 O 12 :Bi [12,74,75] and the strongly dominating UV emission in Y 3 Ga 5 O 12 :Bi [13,14] could indicate that the RES of Bi 3+ is located inside CB in Gd 3 Ga 5 O 12 :Bi and well below the bottom of CB, in Y 3 Ga 5 O 12 :Bi. However, the E g values (≈ 6.6 eV and ≈ 6.4 eV [10]) in Y 3 Ga 5 O 12 :Bi and Gd 3 Ga 5 O 12 :Bi are close.
The data reported above indicate that the UV/VIS emission intensity ratio is not determined only by the band gap energy. Most probably, it depends much more on the probability ratio of the vibronic relaxation in the triplet excited state of Bi 3+ center and the Bi 3+ ionization followed by the formation of the exciton-like state. In some Gd-based compounds, the UV emission of Bi 3+ centers can be absent due to an effective Bi 3+ -Gd 3+ energy transfer. In addition, in case the VIS emission arises from ex 0 {Bi 3+ -Bi 3+ } or ex 0 {Bi 3+ -d}, its intensity should also depend on the Bi 3+ content and on the concentration of defects (d) in the investigated sample. The data reported above indicate that the UV/VIS emission intensity ratio is not determined only by the band gap energy. Most probably, it depends much more on the probability ratio of the vibronic relaxation in the triplet excited state of Bi 3+ center and the Bi 3+ ionization followed by the formation of the exciton-like state. In some Gd-based compounds, the UV emission of Bi 3+ centers can be absent due to an effective Bi 3+ -Gd 3+ energy transfer. In addition, in case the VIS emission arises from ex 0 {Bi 3+ -Bi 3+ } or ex 0 {Bi 3+ -d}, its intensity should also depend on the Bi 3+ content and on the concentration of defects (d) in the investigated sample. Figure 12. Emission spectra of (a) Y3Al5O12:Bi [17], (b) Y3Ga5O12:Bi [13], and (c) Gd3Ga5O12:Bi [74]. T=80 K. Based on the data reported in [13,17,74], presented with the publisher permission.   [13], and (c) Gd 3 Ga 5 O 12 :Bi [74]. T = 80 K. Based on the data reported in [13,17,74], presented with the publisher permission.

Ultraviolet Luminescence
The dimer impurity centers in alkali halides were found to appear only in highly doped crystals, and their concentration was usually much smaller as compared with the concentration of single impurity centers. The absorption coefficient in the single center absorption band shows a linear dependence on the impurity content. In the dimer-related absorption bands, the absorption coefficient quadratically increases with the increasing impurity concentration. However, for the luminescence intensity, the same concentration dependence can be obtained only under excitation in the absorption spectrum region, where the optical density (OD) does not exceed 0.2, since only at OD < 0.2, a linear dependence holds between the number of the absorbed and radiated quanta (see, e.g., [204] and references therein). Therefore, at OD < 0.2, the linear concentration dependence should be observed for the luminescence intensity of single impurity centers and quadratic, for the impurity dimers. At higher optical density (e.g., OD ≈ 0.5 -1.0), the luminescence intensity dependence on the impurity concentration becomes sublinear for single impurity centers and superlinear for dimers. When OD > 2, the luminescence intensity becomes practically independent of the impurity concentration or even decreases due to the reabsorption, concentration quenching, or energy transfer processes.
The Stokes shift an FWHM of the dimer-related emission bands in alkali halides were found to be close or even smaller as compared with emission bands of the corresponding single centers [204]. The low-temperature luminescence decay kinetics was also found to be similar in the single and dimer Tl + centers [206].
Two types of dimer Tl + centers were detected in alkali halides, the centers of the D 2h symmetry and the centers of the D 4h symmetry (see, e.g., [203][204][205][206][207]). In some systems, the centers of both types coexist (see, e.g., [205]). In the case of the D 2h -type centers, two close Tl + ions can strongly perturb each other and can be considered as a quasimolecule (Tl + ) 2 consisting of two Tl + ions. The electron states of the quasimolecule can be considered as molecular orbitals constructed from the electron states of the two Tl + ions. In the case of the D 4h -type centers, two Tl + ions are separated by an anion (see, e.g., [206]). Such a {Tl + -anion -Tl + } dimer center can be considered as a single Tl + ion perturbed by the field of the second Tl + ion. The absorption and emission bands of the dimer centers of both types arise from electronic transitions between the energy levels of these centers (for more details, see [203,204,207] and references therein).
According to [7,65,77,158], the Bi 3+ -Bi 3+ distances in the investigated materials are about 3−4 Å. Therefore, the same considerations could also be applied in case of dimer Bi 3+ centers in more complicated materials. The analysis of literature data allows to suggest that the UV luminescence of such {Bi 3+ -Bi 3+ } dimer centers was also observed in complex oxides, namely, in Y 3 Al 5 O 12 :Bi [201] and Lu 2 SiO 5 :Bi [9,21].
In [201], the spectra of Y 3 Al 5 O 12 :Bi with two different Bi 3+ concentrations (0.13 and 0.27 at.%) were compared at room temperature (RT). Under excitation in the region of the lowest-energy (4.54 eV) absorption band of Bi 3+ , the presence of two types of Bi 3+ -related centers was revealed from the dependence of the UV emission band position on the excitation energy. The higher energy (4.045 eV) emission band was ascribed to single Bi 3+ centers. The 3.995 eV emission band, slightly shifted to lower energies with respect to the former, was ascribed to Bi 3+ pairs since it can be better distinguished with the increasing Bi 3+ concentration. The decay kinetics is not too different for these UV emissions. It is slower for the 3.995 eV emission as compared to the 4.045 eV emission. The energy transfer between the single and dimer centers was suggested due to the overlap of the emission band of the single Bi 3+ centers with the absorption band of dimers. However, no dependence of the UV emission band position on the Bi 3+ content (varying from 0.07 to 0.18 at.%) was noticed in Lu 3 Al 5 O 12 :Bi [19].
In [21], two UV emission bands located at 3.45 eV and 3.30 eV and having similar characteristics (Table 5) were observed in Lu 2 SiO 5 :Bi with a large Bi content (2.24 at.%). The steady-state emission spectrum of Lu 2 SiO 5 :Bi is presented in Figure 13a, however, the presence of two emission bands becomes evident in the time-resolved emission spectra shown in Figure 13b. Comparison of the shapes of their excitation spectra ( Figure 14) indicates that the center with the 3.30 eV emission competes with the dominating 3.45 eV emission center in the excitation light absorption process. The excitation spectrum of the 3.30 eV emission is shifted to lower energies with respect to that of the 3.45 eV emission and distorted due to a strong absorption arising from the main Bi 3+ centers. Such behavior is characteristic for the dimer centers of the D 4h symmetry in alkali halides (see, e.g., [204][205][206]). The decay kinetics of both these emissions is alike (Figure 15). From the temperature dependence of the decay times, the parameters of the corresponding RES were calculated and found to be similar (see Table 5).         The obtained data indicate that both the strong 3.45 eV emission and weak 3.30 eV emission arise from the triplet RES of Bi 3+ -related centers. It is not excluded that the 3.30 eV emission of Lu2SiO5:Bi arises from dimer {Bi 3+ -Bi 3+ } centers. It should be noted that in [21], the 3.30 eV emission was ascribed to the Bi 3+ ions located in the Lu2 lattice sites (Bi2 centers). If so, the Bi2 centers should exist in Y2SiO5:Bi as well. However, no related emission was observed in [20] in the Y2SiO5:Bi sample with much smaller Bi 3+ content (0.04 at.% Bi). This fact could confirm our suggestion on the dimer origin of the 3.30 eV emission in Lu2SiO5:Bi. Unfortunately, the concentration dependences of the luminescence intensity were not studied in [20,21] due to the absence of samples with various Bi 3+ content.
The search for the UV emission of {Bi 3+ -Bi 3+ } dimers should be carried out by the comparison of the emission spectra and decay kinetics under excitation in different regions of the lowest-energy (A) absorption band, as well as by the detailed study of the concentration dependence of the UV emission intensity.

Visible Luminescence
As mentioned above, in many works devoted to the investigation of the luminescence characteristics of the Bi 3+ -doped compounds, the lower-energy broad emission bands, usually located in the visible spectral range and having the characteristics (see Table 4) which are strongly different from the characteristics of the UV emission of the single Bi 3+ centers (Table 1), were ascribed to the {Bi 3+ -Bi 3+ } dimers or Bi 3+ clusters (see, e.g., [158] and references therein). This conclusion was often based only on the spectra measured at room temperature, without precise measurements of the Bi 3+ concentration dependence of the luminescence intensity. A strong difference in the luminescence characteristics (in particular, the values of S, FWHM, and D) of the single Bi 3+ centers and the centers ascribed to {Bi 3+ -Bi 3+ } dimers was not explained. For example, in [58,97,99,104,110,112,121], the lowerenergy emission of Y3Ga5O12:Bi [110]  The obtained data indicate that both the strong 3.45 eV emission and weak 3.30 eV emission arise from the triplet RES of Bi 3+ -related centers. It is not excluded that the 3.30 eV emission of Lu 2 SiO 5 :Bi arises from dimer {Bi 3+ -Bi 3+ } centers. It should be noted that in [21], the 3.30 eV emission was ascribed to the Bi 3+ ions located in the Lu2 lattice sites (Bi2 centers). If so, the Bi2 centers should exist in Y 2 SiO 5 :Bi as well. However, no related emission was observed in [20] in the Y 2 SiO 5 :Bi sample with much smaller Bi 3+ content (0.04 at.% Bi). This fact could confirm our suggestion on the dimer origin of the 3.30 eV emission in Lu 2 SiO 5 :Bi. Unfortunately, the concentration dependences of the luminescence intensity were not studied in [20,21] due to the absence of samples with various Bi 3+ content.
The search for the UV emission of {Bi 3+ -Bi 3+ } dimers should be carried out by the comparison of the emission spectra and decay kinetics under excitation in different regions of the lowest-energy (A) absorption band, as well as by the detailed study of the concentration dependence of the UV emission intensity.

Visible Luminescence
As mentioned above, in many works devoted to the investigation of the luminescence characteristics of the Bi 3+ -doped compounds, the lower-energy broad emission bands, usually located in the visible spectral range and having the characteristics (see Table 4) which are strongly different from the characteristics of the UV emission of the single Bi 3+ centers (Table 1), were ascribed to the {Bi 3+ -Bi 3+ } dimers or Bi 3+ clusters (see, e.g., [158] and references therein). This conclusion was often based only on the spectra measured at room temperature, without precise measurements of the Bi 3+ concentration dependence of the luminescence intensity. A strong difference in the luminescence characteristics (in particular, the values of S, FWHM, and D) of the single Bi 3+ centers and the centers ascribed to {Bi 3+ -Bi 3+ } dimers was not explained. For example, in [58,97,99,104,110,112,121], the lower-energy emission of Y 3 Ga 5 O 12 :Bi [110], (Y,Gd) 2 O 2 SO 4 :Bi and Li 6 (Y,Gd)(BO 3 ) 3 :Bi [121], LaB 3 O 6 :Bi [104], La 2 O 3 :Bi [112], Cs 2 NaYBr 6 :Bi and Cs 2 NaLaCl 6 :Bi [97], YOCl:Bi and GdOCl:Bi [99], Sr 3 Ga 4 O 9 :Bi [58] was attributed to Bi 3+ pairs or clusters based solely on their large FWHM and S.
In [14,15,19,24,58,78,95,102,109,113,119,202], the lower-energy emission band of Y 3 [102], and Lu 2 SiO 5 :Bi [202] was attributed to Bi 3+ pairs or clusters due to an increase of the VIS/UV emission intensity ratio with the increasing Bi 3+ content. However, this effect cannot be used for the confirmation of the dimer-related origin of the VIS emission.
Indeed, the number of single Bi 3+ centers responsible for the UV emission has to increase linearly with the increasing Bi 3+ concentration. However, the emission spectra are usually measured under excitation in the absorption band maximum where OD is too large. In this case, the emission intensity can be practically independent of the impurity concentration due to the saturation effect (see, e.g., [14,19,119]). In many works even the decrease of the UV emission intensity of single Bi 3+ centers with the increasing Bi 3+ content was observed (see, e.g., [15,24,113]). This effect can be caused by various processes, such as increasing reabsorption or concentration quenching of the UV emission (in case of the small Stokes shift where the emission and absorption bands of single Bi 3+ centers are overlapped), energy transfer to some other centers (e.g., to Gd 3+ ions in Gd-containing materials [15,75,76,78,104,111,208] or to other Bi 3+ -related centers, see, e.g., [97,104,201]). For example, the absence of the Bi 3+ emission in GdP 3 O 9 :Bi [111] and GdB 3 O 6 :Bi [104,208] was explained just by an effective Bi 3+ → Gd 3+ energy transfer as the Bi 3+ emission band overlaps the Gd 3+ 8 S → 6 P absorption lines. In LaB 3 O 6 :Bi, the Bi 3+ emission is very weak due to the energy transfer between the centers responsible for the UV and VIS emissions [104]. The concentration quenching of the UV emission was reported, e.g., in [111] for ScP 3 O 9 :Bi and in [78] for LnBO 3 :Bi (Ln=Sc, Lu). It takes place in case the Bi 3+ concentration exceeds the critical value needed for energy migration among the Bi 3+ ions, due to that, the excitation energy can be transferred to quenching centers.
Unlike for the UV emission, for the VIS emission, the probability of its reabsorption, concentration quenching, and energy transfer is negligible due to its large Stokes shift and the absence of absorption bands in visible spectral region. Therefore, the VIS emission intensity always increases with the increasing Bi 3+ content, which could explain the increasing VIS/UV ratio. However, the sublinear dependence of the emission intensity on the Bi 3+ concentration is usually reported (see, e.g., [15,24,58,113,119]). For example, as the Bi 3+ content in YPO 4 :Bi increases 40 times, the lower-energy (3.81 eV) emission intensity increases only about 4 times [113]. In our opinion, such concentration dependence does not allow to conclude that the 3.81 eV emission arises from the {Bi 3+ -Bi 3+ } pairs.
The lower-energy broad emission bands of Bi 3+ -doped compounds were considered as arising from Bi 3+ pairs also in [10,103,113,154,158,159]. A new mechanism for the {Bi 3+ -Bi 3+ } emission was proposed in [103] and further investigated in [10,154,159]. It was suggested that in a pair of neighboring Bi 3+ ions, an electron transfer from the excited state of one Bi 3+ ion to a neighboring Bi 3+ ion is possible. The electronic transitions within Bi 3+ pairs were associated to an intervalence charge transfer (IVCT) of the type Bi 3+ (6s 2 ),Bi 3+ (6s 2 ) → Bi 4+ (6s 1 )Bi 2+ (6s 2 p 1 ). The IVCT is only possible when the ground state of Bi 2+ is located below the first excited state of Bi 3+ . In [103] was shown that this is the case of Li 2 BaP 2 O 7 :Bi. In [159], the vacuum referred binding energies of the electron in the ground state of Bi 2+ and in the lowest-energy excited state of Bi 3+ were compared for 15 compounds. It was found that the ground level of Bi 2+ is always located below the 3 P 1 level of Bi 3+ . This means that the excitation of one Bi 3+ ion in the {Bi 3+ -Bi 3+ } pair can always result in the electron transfer toward another Bi 3+ ion of the pair and formation of the {Bi 4+ -Bi 2+ } pair. The electron back transfer in the {Bi 4+ -Bi 2+ } pair was suggested to result in the appearance of the initial {Bi 3+ -Bi 3+ } pair in the ground state. It was concluded that this process can be radiative resulting in the broad emission band in the visible region, which is the case, e.g., of LaBO 3 [78] , La 2 O 3 [112], and YOCl [99]. In Li 2 BaP 2 O 7 :Bi, this process was assumed to be nonradiative and, as a result, the emission ascribed to dimers was absent [103]. Since also no emission of single Bi 3+ centers was observed in this compound even at the lowest temperatures, it was concluded in [103,159] that IVCT between two neighboring Bi 3+ ions can be responsible for the quenching of the triplet luminescence of single Bi 3+ centers as well.
However, the absence of the UV luminescence of single Bi 3+ centers in Li 2 BaP 2 O 7 :Bi (as well as in many other Bi 3+ -doped compounds, see, e.g., [12,58,[72][73][74][75][76][77]99,102,104,109]) is most probably caused by the location of the lowest-energy relaxed excited level of Bi 3+ inside CB. The same can also be true for the VIS emission ascribed in literature to Bi 3+ dimers as well as for any other emission. In our opinion, the luminescence quenching in Bi 3+ -doped compounds considered in [103,159] can be caused by the location of the corresponding relaxed excited states inside the conduction band.
It should also be noted that in the Bi 3+ -Bi 3+ IVCT model, the perturbation of energy levels of one Bi 3+ ion by another closely located Bi 3+ ion as well as the possibility of the molecular bond formation between the two close Bi 3+ ions were not taken into account. Under the lowest-energy excitation, the 1 S 0 → 3 P 1 transitions of the single Bi 3+ ion were considered despite the presence of a closely located second Bi 3+ ion. However, as the perturbation of a Bi 3+ ion by another Bi 3+ ion in the {Bi 3+ -Bi 3+ } pair was considered to be negligible, it is not clear why the electron recombination with almost unperturbed Bi 4+ ion results in the broad VIS emission with a large Stokes shift, but not in the slightly perturbed UV emission of a single Bi 3+ center. As the considered IVCT process takes place in the dimer {Bi 3+ -Bi 3+ } center, it is also not clear how it can explain the quenching of the UV emission of another, single Bi 3+ center. In some works, the energy distances D between the emitting and metastable levels of the triplet RES were determined for the centers responsible for both the UV and VIS emissions of the same compound. From comparison of the data in Tables 1 and 4 it is evident that the values of D differ by up to two orders of magnitude. However, it was not explained how the formation of the {Bi 3+ -Bi 3+ } pair can result in such strong reduction of the spin-orbit splitting energy of the triplet RES of Bi 3+ (e.g., from 55 meV to 5.7 meV in LaBO 3 :Bi, see [78]).
In [158], an empirical equation was proposed to estimate of the Bi 3+ -Bi 3+ IVCT energy, similar to that proposed in [7] for the calculation of the MMCT energy. However, for most of the Bi 3+ -doped compounds considered in [158], the experimental position of the lowest-energy excitation band of the VIS emission ascribed in the literature to {Bi 3+ -Bi 3+ } centers (see E exc values in Table 4) markedly differs from the calculated IVCT energy. Only in LaZr 2 O 7 :Bi (where, however, the VIS emission was ascribed to the impurity trapped exciton [120]), these energies were found to be close (4.27 eV [120,159] and 4.22 eV [158], respectively).
Thus, in our opinion, the presence of the luminescent {Bi 3+ -Bi 3+ } pairs or clusters in Bi 3+ -doped compounds is not confirmed by experimental data. The data of [21,201] allow only to suggest that the lower-energy UV emission bands reported in these papers could arise from the {Bi 3+ -Bi 3+ } dimers. The broad visible Bi 3+ -related emission bands with the large Stokes shifts presented in Tables 2  and 4 are all of an exciton-like origin. These bands can arise from the excitons localized around different Bi 3+ -related centers, including also the {Bi 3+ -Bi 3+ } dimers, i.e., from ex 0 Bi 3+ , ex 0 {Bi 3+ -Bi 3+ }, ex 0 {Bi 3+ -d).
In our opinion, in case the IVCT between two close Bi 3+ ions can really take place, the electron-hole recombination in the optically created {Bi 4+ -Bi 2+ } pair could result in the formation of an exciton localized around the {Bi 3+ -Bi 3+ } pair. The radiative decay of ex 0 {Bi 3+ -Bi 3+ } should result in the appearance of the broad emission band with the large Stokes shift characteristic for an exciton-like emission.
The luminescence origin of Bi 3+ -doped compounds can be confirmed only by the study of the luminescence decay kinetics in a wide temperature range which allows to determine the parameters of the corresponding RES. Only the dependence of the number of luminescence centers on the concentration of Bi 3+ in the crystal can indicate, whether the single Bi 3+ ions or the {Bi 3+ -Bi 3+ } pairs are responsible for the investigated emission. As an example, in Figures 16 and 17, the dependences of the luminescence intensity on the Bi 3+ concentration inside the investigated samples are presented for the exciton-like VIS emission of Gd 3 Ga 5 O 12 :Bi [74,75] and the Bi 3+ -doped vanadates [76] and niobates [77]. These dependences were measured under excitation in the absorption band region where the optical density is surely small (OD < 0.5) (see Figure 16a). The superlinear dependence of the emission intensity on the Bi 3+ content was found only for the VIS emission of the Bi 3+ -doped niobates [77] (Figure 17b). This allowed us to ascribe this band to the exciton localized around a dimer {Bi 3+ -Bi 3+ } center: ex 0 {Bi 3+ -Bi 3+ }. In other cases, the dependence was linear or sublinear (Figures 16b  and 17a). This indicates that the lower-energy exciton-like emission is connected with a single Bi 3+ ion associated with a lattice defect (ex 0 {Bi 3+ -d}).  Figure 16a for some selected excitation energies, on the Bi 3+ content. See also [76]. Presented with the publisher's permission. It should be noted that the sublinear dependence of luminescence intensity on the impurity concentration, reaching the saturation (for example, similar to that presented in Figure 16b, empty squares), usually obtained under excitation in the absorption band maximum, is often interpreted as the luminescence concentration quenching. For example, the Bi 3+ -related emission intensity was found to increase with the increasing Bi content only up to ≈0.5 at.% in GdNbO4:Bi [125,157] and up to 0.5 at.% [153], 1 at.% [101], or 1.5 -2 at.% [125,150] in YNbO4:Bi. At higher Bi content, it reaches the maximum and then decreases. However, such dependence could appear due to too large optical density in the chosen excitation region. As evident from Figure 16b and Figure 17a,b, no  Figure 16a for some selected excitation energies, on the Bi 3+ content. See also [76]. Presented with the publisher's permission.  Figure 16a for some selected excitation energies, on the Bi 3+ content. See also [76]. Presented with the publisher's permission. It should be noted that the sublinear dependence of luminescence intensity on the impurity concentration, reaching the saturation (for example, similar to that presented in Figure 16b, empty squares), usually obtained under excitation in the absorption band maximum, is often interpreted as the luminescence concentration quenching. For example, the Bi 3+ -related emission intensity was found to increase with the increasing Bi content only up to ≈0.5 at.% in GdNbO4:Bi [125,157] and up to 0.5 at.% [153], 1 at.% [101], or 1.5 -2 at.% [125,150] in YNbO4:Bi. At higher Bi content, it reaches the maximum and then decreases. However, such dependence could appear due to too large optical density in the chosen excitation region. As evident from Figure 16b and Figure 17a,b, no It should be noted that the sublinear dependence of luminescence intensity on the impurity concentration, reaching the saturation (for example, similar to that presented in Figure 16b, empty squares), usually obtained under excitation in the absorption band maximum, is often interpreted as the luminescence concentration quenching. For example, the Bi 3+ -related emission intensity was found to increase with the increasing Bi content only up to ≈0.5 at.% in GdNbO 4 :Bi [125,157] and up to 0.5 at.% [153], 1 at.% [101], or 1.5-2 at.% [125,150] in YNbO 4 :Bi. At higher Bi content, it reaches the maximum and then decreases. However, such dependence could appear due to too large optical density in the chosen excitation region. As evident from Figures 16b and 17a,b, no concentration quenching is observed in the considered Bi 3+ -doped compounds at least up to 6 at.% of Bi 3+ .
In many cases, the lowest excitation band of the VIS emission is shifted to lower energies with respect to the lowest excitation band of the UV emission of a Bi 3+ center (see, e.g., [14,15,19,24,64,78,[93][94][95][96]104,109,110,112,119,121,140,141] and compare E exc values in Table 1, Table 2, and Table 4). It is not excluded that in this case the lower-energy part of the excitation band arises from {Bi 3+ -Bi 3+ } dimers. Under excitation in this {Bi 3+ -Bi 3+ }-related band, the VIS emission of ex 0 {Bi 3+ -Bi 3+ } can appear which is overlapped with the ex 0 Bi 3+ emission. Indeed, in many cases, the VIS emission band is found to be complex (see, e.g., [9,[16][17][18]21,74,75,77,95,102,128]). This means that it can consist of emission bands of the excitons localized around different Bi 3+ -related centers.
The mechanism of the appearance of the visible {Bi 3+ -Bi 3+ }-related exciton-like emission could be the following. Like in the case of a single Bi 3+ center (see equation (11)), the lowest-energy RES of a {Bi 3+ -Bi 3+ } dimer can be located inside CB. Under excitation in the lowest-energy absorption band of the {Bi 3+ -Bi 3+ } center, the electron from the {Bi 3+ -Bi 3+ } center can be optically released into CB and then recombine with the hole remained at the dimer center. The electron-hole recombination results in the formation of ex 0 {Bi 3+ -Bi 3+ } whose radiative decay results in the appearance of the {Bi 3+ -Bi 3+ }-related exciton-like emission:
The third excitation maximum in Figure 18 at shorter wavelength below 300 nm was ascribed to 2 P 1/2 → 2 S 1/2 (6s 2 7s) allowed transition [105,213]. However, later on, based on theoretical calculations determining the intensity of transitions resulting from the admixture of parity-allowed 6s → 6p and 6p → 6d transitions [219], the 300 nm excitation maximum was attributed rather to a mixed state of 6s6p 2 , 6s 2 6d, and 6s 2 6p configurations. ground state and P3/2(1) and P3/2(2) excited states [105]. It gives rise to the emission transition P3/2(1) → 2 P1/2 and two excitation maxima arising from 2 P1/2 → 2 P3/2(1), 2 P3/2(2) ones, see Figure 18. The third excitation maximum in Figure 18 at shorter wavelength below 300 nm was ascribed to 2 P1/2 → 2 S1/2 (6s 2 7s) allowed transition [105,213]. However, later on, based on theoretical calculations determining the intensity of transitions resulting from the admixture of parity-allowed 6s → 6p and 6p → 6d transitions [219], the 300 nm excitation maximum was attributed rather to a mixed state of 6s6p 2 , 6s 2 6d, and 6s 2 6p configurations.  Vibrational structure at the high energy side of Bi 2+ emission in SrB 4 O 7 at 4.2 K observed already in [105] was subjected to theoretical calculations and ascribed to two totally symmetric off-center vibrations of Bi 2+ [221]. The emission transition is parity forbidden (p-p), but this selection rule is relaxed by admixture of emitting level with higher lying 2 S 1/2 term and observed decay times are about 10 µs in SrB 4 O 7 host [105]. In [103] the energy levels of Bi 2+ were situated within the band gap of Li 2 BaP 2 O 7 and it turns out that the emitting 2 P 3/2 (1) state is about 1 eV below the bottom of CB providing sufficiently high energy barrier against the thermal ionization quenching. Furthermore, high thermal stability of Bi 2+ emission in SrB 4 O 7 (reported in 20-300 K) [222] and Sr 2 P 2 O 7 (reported up to 500 K) [223] also points to very limited thermal quenching to the ground state within the temperature range studied.
Given the dominant p-character of the ground and lowest excited levels which are included in the radiative deexcitation of Bi 2+ , such a center becomes sensitive to the surrounding crystal field defined by the chemical composition and structure of the host. Currently available studies show the spread in the Bi 2+ emission peak positions from about 550 nm in CaF 2 [218] to about 716 nm in Ba 2 P 2 O 7 [106]. In case of stable solid solutions, expected in (Ca,Sr)F 2 [218], (Ca,Sr)BPO 5 [213], or (Ca,Sr)SO 4 [224] the peak position and FWHM could even be finely tuned. In case when Bi 2+ shows the emission peak within 580-630 nm and its thermal stability is sufficient, application of such a phosphor for white LED sources was considered, e.g., in [220,222]. Occurrence of a broad excitation 2 P 1/2 → 2 P 3/2 (2) band within 380-480 nm enables to use the blue or near UV LEDs as excitation sources. Nevertheless, due to partially forbidden character of this transition, its oscillator strength might be too low for practical application. It has been partially improved e.g., by admixing Ca into SrB 4 O 7 host [222].
The discovery of a new broad ((FWHM 150 nm) emission peak at 1150 nm with the decay time of 650 µs in the Bi-doped SiO 2 glass in 1999 [225] provided a new hot topic widely explored by many laboratories. Since amorphous hosts are out of the scope of this review, the reader is directed to a nice review on this subject published by Sun et al. in 2014 [8]. The charge state of bismuth and composition of related emission center in amorphous matrices were a subject of debate for a decade, briefly reviewed in [106]. In that work, the Bi 0 center in Ba 2 P 2 O 7 crystalline powder host was ascribed to a broad emission band in the near infrared region peaking at 1100 nm, having the FWHM of 140 nm, and decay time above 600 µs, see Figure 19.
laboratories. Since amorphous hosts are out of the scope of this review, the reader is directed to a nice review on this subject published by Sun et al. in 2014 [8]. The charge state of bismuth and composition of related emission center in amorphous matrices were a subject of debate for a decade, briefly reviewed in [106]. In that work, the Bi 0 center in Ba2P2O7 crystalline powder host was ascribed to a broad emission band in the near infrared region peaking at 1100 nm, having the FWHM of 140 nm, and decay time above 600 µs, see Figure 19. This emission center can be formed and removed reversibly by annealing the sample at 1100 °C in the CO atmosphere or air, respectively. Its counter-appearance with the Bi 2+ center in such annealing cycles proves bismuth with a valence lower than 2 + to be a responsible center. Based on analyses of the literature data and available lattice sites it has been attributed to radiative transition between to 2 D3/2 → 4 S3/2 levels of Bi 0 . In continuation of the search for such near infrared luminescence bismuth-based centers the sintered ceramic of Ba2B5O9Cl:Bi was prepared and treated under air or Figure 19. Near infrared emission spectra of Ba 2 P 2 O 7 :Bi excited at 586 nm, 723 nm (dotted lines: Gaussian peak fits), 838 nm, and 924 nm, respectively, and dependence of NIR emission intensity on nominal bismuth concentration (inset). Reprinted from [106] with permission.
This emission center can be formed and removed reversibly by annealing the sample at 1100 • C in the CO atmosphere or air, respectively. Its counter-appearance with the Bi 2+ center in such annealing cycles proves bismuth with a valence lower than 2 + to be a responsible center. Based on analyses of the literature data and available lattice sites it has been attributed to radiative transition between to 2 D 3/2 → 4 S 3/2 levels of Bi 0 . In continuation of the search for such near infrared luminescence bismuth-based centers the sintered ceramic of Ba 2 B 5 O 9 Cl:Bi was prepared and treated under air or 95%N 2 /5%H 2 atmospheres at 850 • C [107]. It follows from the set of excitation and emission spectra collected that there are at least two emission centers in the bismuth doped compound, which correspond to the emissions at 1030 nm and 1061 nm. Two different emission centers are due to the substitution of bismuth for Ba in two different sites, Ba(2) and Ba(1) in the Ba 2 B 5 O 9 Cl structure. The presence of two centers is also reflected in the decay curves of the emissions at 1030 nm and 1061 nm with the decay times of 30.2 µs and 35.9 µs, respectively, as illustrated in Figure 20.  When treating Ba2B5O9Cl:Bi alternatively in air and N2/H2 atmosphere, the above described near infrared (1030+1061 nm) and red 650 nm emission centers, the latter being Bi 2+ , can be removed and restored reversibly. With prolongation of treatment in N2/H2 the near infrared emission increases monotonically at the expense of that of 650 nm. With the same reasoning it proves bismuth with the valence lower than 2 + to be a center responsible for the infrared emission. By analysis of its excitation spectra peak positions it turns out that they cannot be ascribed to Bi + [226]. Therefore, the most probable assignment becomes again Bi 0 despite the decay time more than one order of magnitude shorter in the near infrared emission of Ba2B5O9Cl:Bi compared to the above described Ba2P2O7:Bi. Shorter decay time is explained by the enhanced electron-phonon interaction which promotes admixing of 4 S3/2 with higher 4 P, 2 P, or 2 D energy states. Consequently, it decreases the forbidden When treating Ba 2 B 5 O 9 Cl:Bi alternatively in air and N 2 /H 2 atmosphere, the above described near infrared (1030+1061 nm) and red 650 nm emission centers, the latter being Bi 2+ , can be removed and restored reversibly. With prolongation of treatment in N 2 /H 2 the near infrared emission increases monotonically at the expense of that of 650 nm. With the same reasoning it proves bismuth with the valence lower than 2 + to be a center responsible for the infrared emission. By analysis of its excitation spectra peak positions it turns out that they cannot be ascribed to Bi + [226]. Therefore, the most probable assignment becomes again Bi 0 despite the decay time more than one order of magnitude shorter in the near infrared emission of Ba 2 B 5 O 9 Cl:Bi compared to the above described Ba 2 P 2 O 7 :Bi. Shorter decay time is explained by the enhanced electron-phonon interaction which promotes admixing of 4 S 3/2 with higher 4 P, 2 P, or 2 D energy states. Consequently, it decreases the forbidden character of radiative transition 2 D 3/2 → 4 S 3/2 of Bi 0 center.
Interestingly, the emission picture in Ba 2 B 5 O 9 Cl:Bi becomes even richer, if the changes in emission pattern are monitored more systematically during the annealing cycle [227].

Conclusions
In Bi 3+ -doped compounds, the Bi 3+ -related luminescence of at least three types can be observed: 1. The UV emission band of single Bi 3+ centers with the relatively small FWHM and S and the ms-decay time at 4.2 K. As the temperature increases, the decay time remains constant up to 40−100 K owing to a large (~10 2 meV) energy distance D between the emitting and metastable levels of the lowest-energy triplet RES of Bi 3+ corresponding to the excited 3 P1 and 3 P0 levels of a free Bi 3+ ion. Similar characteristics should also be observed for a Bi 3+ ion perturbed by a crystal lattice defect (a {Bi 3+ -d} center). The concentration dependence of the UV emission intensity should be linear.
2. The UV emission of dimer {Bi 3+ -Bi 3+ } centers with the emission and excitation spectra slightly shifted to lower energies with respect to the spectral bands of the single Bi 3+ centers, showing the

Conclusions
In Bi 3+ -doped compounds, the Bi 3+ -related luminescence of at least three types can be observed: 1. The UV emission band of single Bi 3+ centers with the relatively small FWHM and S and the ms-decay time at 4.2 K. As the temperature increases, the decay time remains constant up to 40−100 K owing to a large (~10 2 meV) energy distance D between the emitting and metastable levels of the lowest-energy triplet RES of Bi 3+ corresponding to the excited 3 P 1 and 3 P 0 levels of a free Bi 3+ ion. Similar characteristics should also be observed for a Bi 3+ ion perturbed by a crystal lattice defect (a {Bi 3+ -d} center). The concentration dependence of the UV emission intensity should be linear.
2. The UV emission of dimer {Bi 3+ -Bi 3+ } centers with the emission and excitation spectra slightly shifted to lower energies with respect to the spectral bands of the single Bi 3+ centers, showing the superlinear dependence of the intensity on the Bi 3+ concentration.
3. The lower-energy (usually VIS) emission bands of an exciton-like origin with the large FWHM and Stokes shift and with the temperature dependence of the decay time characteristic for the triplet RES with a very small (~1 meV) energy distance D between the emitting and metastable levels. These bands arise from the excitons localized around different Bi 3+ -related defects: ex 0 Bi 3+ , ex 0 {Bi 3+ -d}, and ex 0 {Bi 3+ -Bi 3+ }. In the latter case, the VIS emission intensity shows the superlinear dependence on the Bi 3+ concentration.
The appearance of the exciton-like emission under excitation in the lowest-energy Bi 3+ -related absorption band can be explained in the following way: Under this excitation, the electron transition takes place from the ground state of the Bi 3+ , {Bi 3+ -d} or {Bi 3+ -Bi 3+ } center into the triplet excited state of this center located inside CB. As a result, the electron in CB and the hole remained at the Bi 3+ , {Bi 3+ -d}, or {Bi 3+ -Bi 3+ } center are optically created. An immediate electron-hole recombination results in the formation of an exciton localized around the Bi 3+ , {Bi 3+ -d}, or {Bi 3+ -Bi 3+ } centers. The radiative decay of these localized excitons results in the appearance of the exciton-like emission bands related to the Bi 3+ , {Bi 3+ -d}, or {Bi 3+ -Bi 3+ } centers, respectively. The exciton-like origin of these emission bands is evident from the structure and parameters of the corresponding relaxed excited state (especially, from a very small values of the spin-orbit splitting energy D of the triplet RES) as well as from the large Stokes shift and FWHM of the emission band characteristic just for the luminescence of an exciton-like origin.
Luminescence spectra of some Bi 3+ -doped compounds contain the overlapping emission bands of excitons localized around different Bi 3+ -related centers. This explains the dependence of the intensity and position of the complex VIS emission band on the concentration of Bi 3+ ions and crystal structure defects (d) in the investigated material.
The luminescence quenching observed in some Bi 3+ -doped compounds can be caused by the location of the corresponding relaxed excited states inside the conduction band.
From the comparison of the FWHM and S values of the UV emission bands and the parameters of the triplet RES of Bi 3+ centers (Table 1), we conclude that the electron-phonon interaction of Bi 3+ with its nearest surroundings noticeably increases in the following sequence of oxides: CaO:Bi → Y 3 Al 5 O 12 :Bi → YAlO 3 :Bi → Y 2 SiO 5 :Bi → Y 4 Al 2 O 9 :Bi → La 2 SO 6 :Bi. In the same sequence of oxides, the spin-orbit splitting energy (D) decreases due to suppression of the spin-orbit interaction by the increasing electron-phonon interaction.
The analysis of the experimental data on luminescence characteristics of Bi 3+ -doped materials reported above allows to conclude that for the correct interpretation of an emission band origin, the following investigations have to be carried out: 1. The measurement of temperature dependence of the investigated luminescence decay time in a wide temperature range (down to 4.2 K and lower temperatures) and the determination of the RES parameters, especially the D value. These measurements allow to separate the triplet emission of single (Bi 3+ , {Bi 3+ -d}), and dimer ({Bi 3+ -Bi 3+ }) centers from the triplet Bi 3+ -related luminescence of an exciton-like origin.
2. The measurement of the dependence of the investigated luminescence intensity on the concentration of Bi 3+ under excitation in the absorption spectrum region where the optical density does not exceed 0.5. Only at these conditions, the luminescence intensity is proportional to the absorption intensity and, consequently, to the number of the corresponding luminescence centers. In case of absence of the re-absorption, concentration quenching and energy transfer from Bi 3+ ions, the number of single Bi 3+ centers should increase linearly with the increasing concentration of Bi 3+ ions, while the superlinear (quadratic) concentration dependence is characteristic for the number of dimer Bi 3+ related centers (both the {Bi 3+ -Bi 3+ } centers and the ex 0 {Bi 3+ -Bi 3+ }). These measurements allow to separate the single Bi 3+ -and the dimer {Bi 3+ -Bi 3+ }-related luminescence centers.
Regarding the bismuth centers with the valence lower than 3 + , the Bi 2+ emission in the red spectral region was convincingly evidenced in a number of crystalline hosts. Pronounced treatment of some oxide or oxyhalide hosts with large divalent site in reduction atmosphere can decrease further the charge state of bismuth to zero and such a Bi 0 center provides a broad infrared emission around 1000-1200 nm. As an intermediate state during the reduction atmosphere-air annealing cycles also the Bi + center was proposed in Ba 2 B 5 O 9 Cl host to explain emission bands at 660 nm, 790 nm and 970 nm, but these assignments still need further verification in another host(s).
Besides the above-mentioned detailed investigations of Bi 3+ -doped compounds in a wide temperature range by the methods of the steady-state and time-resolved luminescence spectroscopy, the application of other investigation methods (e.g., electron paramagnetic resonance, nuclear magnetic resonance, thermally stimulated luminescence, etc.) could also be very useful for clarification of the geometrical structure of the Bi 3+ -related centers and determination of their excited states location with respect to the conduction band. In this respect, it is worth mentioning the recently published positioning of Bi 3+ and Bi 2+ ground states in YPO 4 [228]. Further progress in the experimental research of these materials could surely be complemented by theoretical models of the Bi 3+ -related excited states, including into consideration the energy levels of Bi 3+ ions and the host states and allowing to determine the energy level structure of Bi 3+ -related luminescence centers. These studies as well as the electronic band structure calculations could allow to predict the luminescence characteristics of new Bi 3+ -doped compounds suitable for different applications. This is also valid for Bi ions of lower valence states. The knowledge of the valence state of bismuth ions as well as the host influence on the valence state and structure of Bi-related centers and on the luminescence origin in various Bi-activated materials is needed to understand the physical processes taking place in these materials. This could help to create novel materials for practical applications. In more detail, possible future directions for the search, fabrication, research, and development of Bi-activated photonic materials suitable for different applications are considered in the review paper [8].
Author Contributions: All the authors (A.K., E.M., M.N., S.Z., and Y.Z.) participated in the acquisition, analysis, and interpretation of the experimental data on luminescence of Bi 3+ -doped compounds, discussion and improvement of the manuscript and approved the submitted version. Y.Z. collected and analyzed the data presented in Section 1. The manuscript was written by M.N. (Section 1, Section 6, Section 7), S.Z. (Section 1, Section 2, Section 3, Section 4, Section 5, Section 7, and Appendix A), and E.M. (Sections 2.2 and 3.2). E.M. elaborated phenomenological models for description of excited-state dynamics and calculated the parameters of the excited states of Bi 3+ -related luminescence centers. The figures for Section 2, Section 3, Section 4, Section 5 were prepared by A.K. All authors have read and agreed to the published version of the manuscript.

Conflicts of Interest:
The authors declare no conflict of interest.

Appendix A.
Appendix A. 1

. Bi 3+ -Doped Alkali Halides
Luminescence characteristics of KCl:Bi were studied in [2,85,94,96,140,206,[229][230][231]. In [85,94], different data were obtained on the emission and excitation bands positions, luminescence polarization characteristics, and decay kinetics (Table 1) indicating the possibility of formation of Bi 3+ centers of different structure (e.g., Bi 3+ ions located close to one or two cation vacancies, impurity anions like O 2-, S 2-, etc.). This effect is caused by a small solubility of trivalent ns 2 ions as well as their twofold excess charge whose compensation can take place in different ways, depending on the conditions of the single crystal synthesis and quenching, the purification degree of the host salt, etc.
In [94], two emission bands were observed at 15 K, peaking at 3. In [85], the complex emission band located around 2.5 eV was observed in the emission spectrum of KCl:Bi (Figure 8a). Two fast (17 ns and 27 ns) and two slow (1.38 ms and 2.8 ms) components were observed at 4.2 K in the emission decay (Figure 9a). Their excitation spectra practically coincide. The emission spectra of two fast decay components are shifted to lower energy with respect to the spectra of two slow components. The analysis of the temperature dependences of the decay times allows to conclude that the emission of Bi 3+ centers arises from the triplet RES and that the theoretical model [135,137] is still valid in the case of KCl:Bi, despite the strong spin-orbit interaction.
As an example, the characteristics of triplet luminescence of CaO:Bi are presented in Figures 8b  and 9b. At 4.2 K, the strong slow 3.1 eV emission and weak fast 3.3 eV emission were observed, arising from the radiative decay of two triplet RES, corresponding to the 3 P 0 and 3 P 1 excited levels of a Bi 3+ , respectively [85]. It was concluded that the model [134], considering the spin-orbit interaction being dominant in the triplet RES, should be used for the description of the luminescence characteristics of the systems of this type.
In addition to the triplet emission, a new 4.15 eV emission was found in [85] and ascribed to the electron transitions from the singlet excited state of a Bi 3+ ion related to the 1 P 1 excited state of a free Bi ion. At the temperatures around 30 K, the intensity redistribution takes place between the 4.15 eV and 3.1 eV emissions indicating thermal transitions between the singlet and triplet excited levels of the same luminescence center. No such emission was observed in other Bi 3+ -doped compounds. of the emission intensity was not measured). In [10], the visible emission was ascribed to the CT transitions and in [154], it was interpreted in terms of IVCT in Bi 3+ pairs. It is not excluded that the broad VIS emission band of GdAlO 3 :Bi consists of the emission bands of excitons localized around single and dimer Bi 3+ centers. This could explain the dependence of its position and shape on the Bi 3+ content.
Appendix A.6. Bi 3+ -Doped Oxyorthosilicates For the first time, the luminescence of Lu 2 SiO 5 :Bi and Y 2 SiO 5 :Bi was detected in [20,21]. The characteristics of Bi 3+ -doped Lu 2 SiO 5 and Y 2 SiO 5 were found to be similar. In the emission spectrum of Lu 2 SiO 5 :Bi, a strong UV band (3.455 eV), much weaker 3.3 eV band, and two visible bands (2.2 eV and 2.3 eV) are observed at 4.2 K. In Y 2 SiO 5 :Bi, only the 3.56 eV band can surely be ascribed to Bi 3+ centers. The low-temperature luminescence decay kinetics clearly indicates that the UV emission of both Lu 2 SiO 5 :Bi and Y 2 SiO 5 :Bi arises from the triplet RES of Bi 3+ ions. The light sum of the VIS emission of Bi 3+ -doped silicates is in at least two orders of magnitude smaller as compared with the light sum of the UV emission of the main Bi 3+ centers. Temperature dependences of decay times of both VIS emissions of Lu 2 SiO 5 :Bi are similar to those observed for the VIS emission of other Bi 3+ -doped compounds and allow to connect them with the Bi 3+ -related localized excitons.
The luminescence of PbWO 4 :Bi crystals was studied in detail in [73]. The 2.2 eV emission was ascribed to the excitons localized around the Bi 3+ -related centers of two types. Three possibilities were considered for the interpretation of the lowest absorption (excitation) band of PbWO 4 :Bi: (i) This band arises from the electron transitions from the ground state to the lowest-energy triplet excited state of Bi 3+ , corresponding to the 1 S 0 → 3 P 1 transitions of a free Bi 3+ ion. As the emission from the triplet RES of Bi 3+ is absent, the 3 P 1 -related level of Bi 3+ should be located inside CB, and the electron transitions result in the ionization of Bi 3+ and release of electrons into CB. Similarly to the undoped PbWO 4 crystal, the subsequent fast electron-hole recombination followed by vibronic relaxation results in the formation of the (WO 4 ) 2--type exciton localized around the Bi 3+ ion. The optically released electrons can also be trapped at different traps (e.g., around oxygen vacancies V O or at Bi 3+ and Bi 5+ ions), producing stable paramagnetic electron {Pb + -V O }, Bi 2+ , and Bi 4+ centers.
(ii) Similar absorption band observed in PbWO 4 :Pr [243] was explained by the Pr 3+ /W 6+ → Pr 4+ /W 5+ electron-transfer transitions proposed in [142]. The Bi 3+ → W 6+ charge-transfer transitions in Y 2 WO 6 :Bi were suggested to occur in [1]. In tungstates, the transfer of an electron from the ground state of a Bi 3+ ion to the nearest host W 6+ ion: Bi 3+ (6s 2 )/W 6+ (5d 0 ) → Bi 4+ (6s 1 )/W 5+ (5d 1 ) was concluded to result in the formation of the hole Bi 4+ and electron W 5+ centers. Their subsequent fast recombination and the following relaxation results in the creation of the exciton localized around the Bi 3+ ion and its subsequent radiative decay. With the use of the equation presented in [7], the MMCT energy in PbWO 4 :Bi was estimated to be ≈4.2 eV. This value is close to the suggested position of the Bi 3+ -related absorption band maximum (≈4 eV).
(iii) A trivalent Bi 3+ ion in the PbWO 4 crystal lattice, substituting for a divalent Pb 2+ ion, has an excess positive charge. By analogy with some other materials (see, e.g., [244]), it is not excluded that under photoexcitation of PbWO 4 :Bi, the photostimulated electron transfer from the valence band (VB) to the ground state of a single Bi 3+ ion can also take place, resulting in creation of an electron Bi 2+ center and a mobile hole in VB. The hole can be trapped by oxygen ions located close to the Bi 3+ -related center and/or lead vacancy V Pb . The subsequent fast recombination in close pairs of the optically created electron and hole centers is suggested to result in the creation of excitons localized around Bi 3+ -related centers. The presence of two types of excitons in PbWO 4 :Bi, probably, ex 0 Bi 3+ and ex 0 {Bi 3+ -V Pb } with strongly different thermal stabilities is suggested. The radiative decay of these excitons is accompanied with the 2.2 eV emission. Similar processes were proposed to explain the appearance of the localized exciton emission and creation of the impurity-related electron centers (Tl 0 , Pb + ) and self-trapped holes (V K centers) in the UV-irradiated Tl + -and Pb 2+ -doped caesium halides (see, e.g., [245] and references therein).
The data obtained by the TSL and EPR methods on the origin of stable electron and hole centers created under photoexcitation of PbWO 4 :Bi crystals in the lowest-energy absorption band [73,193] indicate that, in principle, all the above-mentioned processes can take place. As both the single Bi 3+ ions and the Bi 3+ ions associated with lead vacancies can exist in PbWO 4 , the process (ii) seems to be more preferable for {Bi 3+ -V Pb } centers with a negative excess charge, but the process (iii), for single Bi 3+ centers with a positive excess charge.
It should be noted that the same processes can also take place in other materials where a Bi 3+ ion is substituting for mono-or divalent host lattice ion.

Appendix A.8. Bi 3+ -Doped Phosphates
In Bi 3+ -doped LiLaP 4 O 12 glasses and powders, two emission bands were registered at 4.2 K. The 3.02 eV emission was observed under the 5.28 eV excitation while the 2.78 eV emission was observed under the 4.95 eV excitation. In [102], these bands were ascribed to single Bi 3+ ions and Bi 3+ dimers, respectively. In [246], both these bands were ascribed to single Bi 3+ ions. To explain this fact, the presence of two crystallographic positions for a Bi 3+ ion in the phosphate host was suggested. However, a later work [247] on the single crystal X-ray diffraction excluded such possibility.
In [72], the LiLaP 4 O 12 :Bi with different Bi contents and the undoped LiLaP 4 O 12 were investigated. Only the 2.95 eV emission of LiLaP 4 O 12 :Bi was shown to arise from Bi 3+ -related centers and ascribed to an exciton localized around a single Bi 3+ ion. The characteristics of this emission are similar to the characteristics of the 2.8 eV emission of undoped LiLaP 4 O 12 . The absence of the UV emission was explained by the location of the lowest-energy excited state of Bi 3+ inside CB. Owing to a close position of the lowest excitation band of the 2.95 eV emission (5.4 eV) to the estimated (according to [7]) value of the Bi 3+ → La 3+ charge transfer energy (≈ 5.5 eV), it was suggested that under the 5.4 eV excitation, an electron transfer from the ground state of a Bi 3+ ion to the nearest host La 3+ ion, Bi 3+ (6s 2 )/La 3+ (5d 0 ) → Bi 4+ (6s 1 )/La 2+ (5d 1 ), takes place. The subsequent fast electron-hole recombination at the Bi 3+ ion and the following relaxation result in the formation of the exciton localized around Bi 3+ whose radiative decay is accompanied with the 2.95 eV emission.
The same interpretation can be given to both broad visible emission bands of LiLaP 4 O 12 :Bi located at 2.78 eV (S = 2.17 eV) and 3.02 eV (S = 2.25 eV) and ascribed to the triplet emission of Bi 3+ [102]. Based on the concentration dependence reported in [102], the 2.78 eV emission can be assumed to arise from an exciton localized around a Bi 3+ dimer.
In YPO 4 :Bi, LuPO 4 :Bi, two emission bands, an intense narrow band at about 5.12−5.17 eV and a weak broad band at about 3.7−3.8 eV, were observed under excitation around 5.37−5.45 eV. The higher-energy band was ascribed to isolated Bi 3+ ions [7,24,113,128,154,248] while the lower-energy band, to Bi 3+ dimers [24,113,128], MMCT transitions [7,128], or to the electron transfer in the Bi 2+ -Bi 4+ pairs [154]. In our opinion, the lower-energy broad 3.7−3.8 eV emission band with the large Stokes shift (S = 1.75 -1.79 eV) can arise from the excitons localized around Bi 3+ -related centers.
From the temperature dependence of the decay time of the broad (FWHM = 0.6 eV) 2.75 eV emission band of LaPO 4 :Bi with the large Stokes shift (S = 2.4 eV) measured in [83], the value of D = 2.05 meV was obtained. Such small D value allows to make a reliable conclusion that the 2.75 eV emission of LaPO 4 :Bi arises from an exciton localize around a Bi 3+ ion.
In [76], the conclusion on the exciton-like origin of both the intrinsic blue emission and the Bi 3+ -related yellow emission in Bi 3+ -doped rare-earth orthovanadates was made based on the analysis of temperature dependences of their decay times. It was suggested that the Bi 3+ -related absorption (excitation) band around 3.73-3.79 eV arises from the electronic transition from the ground 1 S 0 state to the excited 3 P 1 state of Bi 3+ located inside CB. The excitation in this band was suggested to result in an electron release from the 3 P 1 level into CB and its subsequent immediate recombination with the hole remained at the Bi 3+ ion. As a result, an exciton localized around the single Bi 3+ ion is created whose radiative decay is accompanied with the 2.12-2.19 eV emission. No other VIS emissions, which could arise, e.g., from the excitons localized around dimer {Bi 3+ -Bi 3+ } centers, and no UV emission, which could arise from the triplet excited state of Bi 3+ ions, were found in YVO 4 :Bi, LuVO 4 :Bi, and GdVO 4 :Bi.
Appendix A.10. Bi 3+ -Doped Lanthanide Niobates Luminescence of Bi 3+ -doped niobates was studied mainly at 295 K. Not only the interpretation of experimental results but also positions of the emission and excitation bands reported in different papers were different. For YNbO 4 :Bi, the MMCT origin of the spectral bands was suggested in [1,7,64,65,157]. The lowest-energy absorption band was connected with the Bi 3+ (6s 2 ) → Nb 5+ (3d 0 ) transitions and the emission band, with the Nb 4+ (3d 1 ) → Bi 4+ (6s 1 ) transitions. However, in [149][150][151][152][153], the excitation and emission bands of YNbO 4 :Bi were ascribed to the electron transitions between the ground state ( 1 S 0 ) and the triplet excited state ( 3 P 1 ) of a single Bi 3+ ion. In [254], the absorption of YNbO 4 :Bi was ascribed to the charge transfer transitions from O 2-2p to the excited 6p levels of Bi 3+ . The energies of the 1 S 0 → 3 P 1 transitions (3.91 eV) and the MMCT transitions (3.91 eV) calculated in [65] were found to be close to the position of the lowest-energy excitation band maximum in YNbO 4 :Bi and GdNbO 4 :Bi. This means that, in principle, the appearance of both these types of transitions is possible in the absorption (and luminescence) spectra of YNbO 4 :Bi.
In [77], two Bi 3+ -related emission bands were observed in YNbO 4 :Bi. Based on detailed and systematic investigation of the intrinsic and Bi 3+ -related luminescence in YNbO 4 :Bi powders with different Bi content, carried out by the methods of the steady-state and time-resolved luminescence spectroscopy in the 4.2-500 K temperature range, the conclusion on their exciton-like origin was made. The most intense emission band peaking at about 2.53 eV and excited around 4.09 eV was ascribed to the radiative decay of an exciton localized around a single Bi 3+ ion. The weaker 2.41 eV emission with the superlinear dependence of intensity on the impurity concentration was ascribed to an exciton localized around a dimer Bi 3+ center. No ultraviolet emission arising from the 3 P 1,0 → 1 S 0 transition of a Bi 3+ ion was found. This fact as well as the exciton-like origin of the Bi 3+ -related emission bands indicate that the triplet excited level of Bi 3+ ion is located inside CB of YNbO 4 .
Appendix A.11. Bi 3+ -Doped Borates In [1,78,95], the luminescence characteristics of Bi 3+ -doped lanthanide orthoborates LnBO 3 (Ln: Sc, Y, La, Gd, Lu) of different structure and coordination numbers were investigated. In ScBO 3 :Bi, the appearance of the vibronic structure of the emission and excitation spectra of Bi 3+ centers at low temperatures and the small Stokes shift (0.22 eV) indicate a very small electron-phonon interaction. The energy distance between the emitting and metastable minima of the triplet RES is large (D = 120 meV) indicating a strong spin-orbit interaction. For the UV (3.46 eV) emission of LaBO 3 :Bi, which also arises from Bi 3+ centers, the Stokes shift is much larger (1.16 eV), the decay time at 4.2 K is several times shorter, and the D value is about two times smaller (55 meV). These features are caused by much stronger electron-phonon interaction with respect to the spin-orbit interaction in this compound as well as an increase of the probability of the radiative decay of the metastable minima of the triplet RES. Thus, the variation of luminescence characteristics in the sequence of the orthoborates ScBO 3 → LuBO 3 → YBO 3 → LaBO 3 is caused by the increasing electron-phonon interaction. The broad VIS emission of LaBO 3 :Bi (2.69 eV) with the large Stokes shift (1.84 eV) ascribed in [78] to Bi 3+ pairs or clusters arises most probably from the Bi 3+ -related localized exciton. Indeed, as follows from the temperature dependence of its decay time, the D value for the corresponding triplet RES is in an order of magnitude smaller (5.7 meV) as compared to that of the UV emission (55 meV) and is characteristic for the exciton-like states.
Similar broad emission band of InBO 3 :Bi with the large Stokes shift (1.41 eV) was ascribed to the impurity-bound exciton [155]. It was suggested that the lowest excited level of Bi 3+ is located inside or close to CB of InBO 3 . The excitation of Bi 3+ results in the release of an electron into CB with the subsequent recombination with the hole on the luminescence center and formation of an impurity-trapped exciton.