1. Introduction

Bismuth-doped materials possess intriguing broadband Near-infrared (NIR) luminescence (1100-1700 nm), coinciding well with telecommunication window and show promises for new optical amplifiers and oscillators with outstanding characteristics [1,2]. A number of silicate [1], phosphate [3], borate [4], sulfide [5] and fluoride [6] glasses as well as RbPb₂Cl₅ [7], BaB₂O₄ [8], BaF₂ [9], Ba₂P₂O₇ [10], CsI [11], zeolite [12,13] and sodalite [14] crystals, became NIR luminescent being doped with bismuth. The luminescence of bismuth ions in common oxidation state +3 is well known [15], but it takes place in visible, not NIR, so another less trivial sources of NIR luminescence in bismuth-doped materials should be suggested. Recently, it was demonstrated by synproportionation reaction of metallic Bi with Bi³⁺ in glasses [16] and ionic liquids [17], that subvalent (i.e., lower than usual +3 oxidation state) bismuth species are responsible for NIR luminescence in these media.

Bismuth possesses an interesting ability to form numerous monoatomic and cluster polycation subvalent states: Bi²⁺ [18], Bi⁺ [19–21], Bi₂⁴⁺ [22–24], Bi₃⁺ [25], Bi₅³⁺ [19,21,26], Bi₅⁺, Bi₆²⁺ [27,28], Bi₈²⁺ [29,30], Bi₉⁵⁺ [20,24,31]. All of them can be formed in acidic media via synproportionation of Bi³⁺ with Bi⁰ or by Bi³⁺ reduction.

Usually several bismuth subvalent species exist in the media simultaneously, and their concentrations depend on equilibrium constants of interconversion [19,21], like

That is why several subvalent bismuth species can in principle contribute to NIR luminescence. Until now, only polycation Bi₅³⁺ was definitely shown to possess very broadband NIR luminescence in Bi₅(AlCl₄)₃ crystals [32,33]. Bi⁺ monocation was also suggested as possible source of NIR luminescence [3,4], but in this case evidences are still indirect. So, real contribution of all bismuth subvalent states into NIR luminescence in different bismuth-doped materials is unclear.

Here we try to clarify this problem by studying the NIR luminescence from subvalent bismuth species in chloride glass with composition: AlCl₃-ZnCl₂-BiCl₃. We have investigated this glass system for several reasons:

2. Experimental

Dry AlCl₃ (99.999%) ZnCl₂ (99.999%) BiCl₃ (99.998%) were used for glass preparation. All manipulations with the starting chlorides (weighing and transfer to fused silica cell) were performed in argon-filled glovebox (<2ppm H₂O) due to their extreme hygroscopicity. The proper amounts of chlorides (60% molar ZnCl₂, 38% molar AlCl₃, 2% molar BiCl₃, approximately 4 g batch) were placed into L-shaped fused silica cell (Fig. 1 ). Then, the cell was capped off at the standard-taper joint with special cap, equipped with a vacuum stopcock. It was taken from the glovebox, evacuated, filled with 200 Torr He and sealed at the preformed waist. The fused silica cell also had a little appendix, perpendicular to the plane of L figure. Before the main operations in the glovebox, this appendix was filled with molten zinc metal, which after solidification consolidated with appendix. This zinc metal could not form direct contact with melt, when the chloride composition was melted in the horizontally placed cell with raised appendix (Fig. 1). This isolated metal position allows slow reduction of Bi³⁺ in chloride melt during melting process with convenient control on the concentration of subvalent bismuth species, introduced into resulting glass.

 

Fig. 1 Sealed L-shaped fused silica cell with thin layer of AlCl₃/ZnCl₂/BiCl₃ glass. The glass orange color is due to the presence of subvalent bismuth species (mainly Bi₅³⁺ polycation).

Download Full Size | PDF

The filled cell was heated initially to 473K, then to 573K until moderately colored thermochromic melt was formed. The color and thermochromic properties of our specimens (violet color at 573K and orange at room temperature) was similar to previously reported for chloride melts, containing subvalent bismuth [19]. Upon cooling, the melt not only changes color, but also become viscous, and at this moment it can be spread in a thin layer (approximately 1 mm thick) over the wall of short compartment of L-shaped cell. The walls of this short compartment were flattened and overall dimensions of compartment allow its easy insertion into standard socket of spectrophotometer.

After the final solidification of chloride melt at room temperature it forms a thin transparent glass layer, attached to the wall of the cell. At 77K the glass became very fragile; it was detached from the wall and usually became broken into several pieces. At the room temperature these pieces slowly stick together and to the walls of the cell, indicating, that glass transition of the material occurs near the room temperature.

The optical absorption spectra of AlCl₃/ZnCl₂/BiCl₃ glass were obtained with Shimadzu UV-3600 UV-VIS-NIR spectrophotometer. Photoluminescence spectra were measured using an ARC SpectraPro SP-305 monochromator combined with an ARC ID-441C InGaAs or ID-443 InSb detector for the ranges of 900-1600 nm and 1600-2500 nm respectively. Emission decay curves were obtained with Hamamatsu G8372-01 InGaAs photodiode coupled with a wideband (100 Hz–5 MHz) low-noise preamplifier mounted on the exit slit of the monochromator. Osram HBO 150W Xe-lamp combined with LOMO MDR-12 monochromator was used as variable wavelength light source to obtain luminescence excitation spectra. The measurements were performed at 300 and 77K temperatures. All the obtained spectra were corrected for the spectral response of the system.

3. Results and discussion

The optical absorption spectrum of AlCl₃/ZnCl₂/BiCl₃ glass, prepared, as described above is shown in Fig. 2(a) .

 

Fig. 2 (a) Absorption spectrum of AlCl₃/ZnCl₂/BiCl₃ glass, containing subvalent bismuth species; (b) Photoluminescence spectra of AlCl₃/ZnCl₂/BiCl₃ glass (300K), excited at different wavelengths normalized with respect to the 1320 nm emission.

Download Full Size | PDF

The spectrum contains overlapped bands at 390, 430, 500, 750, 850 nm, corresponding well to the spectra of Bi₅³⁺ polycation in chloride solvents, reported previously [19,21]. In acidic chloride melts with excess Bi³⁺ the polycation Bi₅³⁺ usually exists in equilibrium with only one another subvalent bismuth form: the monocation Bi⁺ (Eq. (1)) [19]. The high temperature and increasing acidity shifts equilibrium to the preferable Bi⁺ formation. In our case the dominant subvalent form is Bi₅³⁺, since the temperature of our specimens during the spectrum recording is low during spectrum recording. At elevated temperatures (300°C) the Bi⁺ concentration in the melt became much higher, changing the color of the melt to violet, but upon cooling to the room temperature the Bi₅³⁺ turn to be dominant. The equilibrium reaction (Eq. (1)) is not frozen in the glass even at ambient conditions due to the low glass transition temperature [34]. Only a very weak absorption line due to Bi⁺ at 660 nm is evident in absorption spectrum at room temperature, whereas the second (and stronger) Bi⁺ absorption line at 580 nm is obscured by overlapping with overwhelming Bi₅³⁺ absorption. From reported molar absorptivities of Bi₅³⁺ and Bi⁺ [21] we can estimate their concentration ratio as [Bi₅³⁺]/[Bi⁺]≈10 at the room temperature. These subvalent Bi₅³⁺ and Bi⁺ cations are formed in AlCl₃/ZnCl₂/BiCl₃ melt by reduction of BiCl₃ with metal Zn:

These reactions proceed slowly in the cell even without direct contact of Zn metal with BiCl₃-containing chloride melt, may be because of non-negligible BiCl₃ vapor pressure at our glass melting temperature. The progress of reaction is seen as melt coloration became deeper.

The NIR luminescence spectra of subvalent bismuth containing AlCl₃/ZnCl₂/BiCl₃ glass at the room temperature are shown in Fig. 2(b). It is evident, that luminescence consists of two bands, peaked at 1080 and 1300 nm, excited differently by 450 and 532 nm radiation. At room temperature no luminescence was observed in the range 1700-2500nm with InSb detector and at the highest sensitivity of the measurement system. After glass cooling to 77K those two bands in luminescence spectrum became stronger and sharper and still another very broad luminescence band, covering interval from 1300 to 2500 nm (maximum at 2000 nm) had appeared (Fig. 3(a) ).

 

Fig. 3 (a) Photoluminescence spectra of AlCl₃/ZnCl₂/BiCl₃ glass, measured at 77 and 300K (excitation wavelength λ_ex = 532 nm). The spectrum at 77K composed from two separate measurements: at short waves (with InGaAs detector), and at long waves (with InSb detector); (b) Photoluminescence excitation spectra, measured for three luminescence bands. Spectra of the first and the second bands (λ_em = 1060 nm and λ_em = 1300nm, respectively) were measured at 300K. Excitation spectrum of the third band (λ_em = 1900 nm) was measured at 77K.

Download Full Size | PDF

The luminescence excitation spectra of all the three bands are shown in Fig. 3(b). All excitation spectra rise at λ<350 nm (nonspecific feature) due to excitation of charge transfer transition of chlorine-coordinated Bi³⁺ with subsequent energy transfer to all NIR-luminescent species. Besides this common UV feature, the excitation spectrum for the first band (measured at 1060 nm emission) consists of two well resolved peaks (at 580 and 660 nm), which are characteristic to Bi⁺ absorption spectrum in chloroaluminate and chlorogallate systems [19,21].

The luminescence excitation spectrum of 1300 nm band (second band) is represented by the single peak at 580 nm and it doesn’t match well both Bi⁺ and Bi₅³⁺ absorption spectra. Yet, this excitation spectrum coincides well with absorption (with single maximum at 580 nm), reported for Bi⁺ in reduced BiCl₃ melts [35]. The presence of Bi⁺ in reduced BiCl₃ melt was demonstrated both by thermodynamic and spectroscopic methods [35–37], although its absorption spectrum was very different from more studied and properly assigned spectrum of Bi⁺ in chloroaluminate and chlorogallate melts [19,21,38]. It was noticed, however, that both thermodynamic and spectroscopic methods could not distinguish between genuine monocation Bi⁺ and its probable complex with arbitrary number of Bi³⁺ from solvent BiCl₃ excess [25]. So, the absorption at 580 nm in reduced BiCl₃ melt, as well as 1300 nm luminescence with excitation peak at 580 nm in our glass can both originate from Bi⁺ complex: Bi⁺·nBi³⁺. There is one known example of such clusters: Bi₂⁴⁺ (n = 1), reported by X-ray crystallography investigations of different systems [22–24], although no study of possible NIR luminescence from these systems has been undertaken until today. Clusters with n>1 unlikely exists due to strong electrostatic repulsion from the large overall charge. That’s why, it is reasonable to consider only dimer Bi₂⁴⁺ as a possible emitter of 1300 nm luminescence band. This hypothesis also agrees well with the observation that in some bismuth-doped materials luminescence intensity is proportional to the square of introduced bismuth concentration [39]. We also should remark, that all bismuth polycations possesses some degree of covalent bonding with surrounding counterions, so, notations like Bi₂⁴⁺ always represent considerable simplification, especially for the chalcogen [22] and trifluoroacetate [23] systems. Generally speaking, terms like Bi₂⁴⁺ should be perceived here in a broad sense - as a designation for dimer of bismuth in +2 valence state.

Finally, the luminescence excitation spectrum for long-wavelength emission (measured at 1900 nm, third band) coincides well with features of Bi₅³⁺ absorption spectrum (peak near 800 nm and steep rising at λ<500 nm). Two peaks, characteristic for Bi⁺ excitation are also present in 1900 nm luminescence excitation spectrum. This fact can be explained by initial Bi⁺ excitation with possible energy transfer from high excited state of Bi⁺ to Bi₅³⁺ polycation. This identification of the long-wavelength emission center as Bi₅³⁺ is in complete agreement with recent luminescence measurements from Bi₅³⁺ in Bi₅(AlCl₄)₃ crystal [33]. In this system very broad luminescence continuum, reaching 3500 nm was observed. Here we report the first observation of broadband Bi₅³⁺ luminescence in glass materials.

The temporal decay of the luminescence in all three bands had been studied at 77K (Fig. 4 ) with an excitation at 532nm. The room temperature data for the temporal decay were too noisy to be correctly approximated.

 

Fig. 4 Low temperature (77K) luminescence decay curves measured at 1350nm(a), 1900nm(b) and 1050nm(c).

Download Full Size | PDF

It was found, that single exponential curve poorly represents all obtained decay data. Decay curve for the emission in the first band (Bi⁺ luminescence, Fig. 4(c)) can be best represented as a sum of three exponents. The fast component has the lifetime (τ₁) <0.2μs, which is beyond the response of the detector and cannot be resolved in our system. The slow component in the first band at 77K has a lifetime (τ₃)~266μs which is similar to Bi⁺ luminescence in chloride crystal RbPb₂Cl₅ [7]. The positions of emission maxima are also coincided well.

Decay data for the second and the third luminescence bands are well approximated by so called “stretched” exponent I ~exp[-(t/τ)^0.5]. One of the possible explanations for such temporal dependence is that dipole-dipole energy transfer quenching is observed in this case [40]. However at this moment we have not enough data to confirm or refute this hypothesis.

Cao et al. [33] demonstrated, that Bi₅³⁺ luminescence decay in Bi₅(AlCl₄)₃ was a single exponent with lifetimes changing from 23 μs (10K) to 6 μs(300K). Not only the lifetime, but also the luminescence intensity diminished at room temperature, indicating increased quenching. In our glass the luminescence of Bi₅³⁺ (third band) is completely quenched at the room temperature, while at 77K its lifetime can be estimated as τ~18μs which is comparable with Cao et al. data.

Steep decreasing of luminescence intensity and lifetimes at warming to the room temperature allows suggesting the multiphonon relaxation as a primary mechanism of Bi₅³⁺ excited state depopulation or possible relaxation via the conical intersection. The fact, that at the room temperature the luminescence from Bi₅³⁺ is completely quenched in the chloride glass, whereas it is still observed in Bi₅(AlCl₄)₃ crystal, argues for multiphonon relaxation, since phonon spectrum in glasses are broadened (compared to crystals), speeding up quenching process. However depletion mechanism of observed luminescent excited levels cannot be clearly determined from our measurements and further investigations are needed.

The most interesting feature of Bi₅³⁺ luminescence is the unprecedented spectral width. Recently, we had calculated (by spin-orbit CI calculation with relativistic pseudopotential) the energies and oscillator strengths for Bi₅³⁺ excited states up to 4 eV [41]. It was shown, that at the ground state equilibrium geometry (trigonal bipyramid, D_3h point symmetry group) the lowest excited state is E” (1.64 eV) with forbidden electrodipole transition to the ground state A₁'. The polycation Bi₅³⁺ in double degenerate state E” can undergo a geometric distortion from D_3h bipyramidal geometry due to Jahn-Teller effect. The possibility of Jahn-Teller effect in Bi₅³⁺ lowest excited state can explain the broad luminescence spectrum and large Stokes shift, since equilibrium geometry of Bi₅³⁺ is very different in ground and excited luminescent states. The additional experimental observations (such as temperature dependence of Bi₅³⁺ luminescence spectrum), analysis of possible distortion of Bi₅³⁺ geometry in the ground and excited states by the crystal field along with the theoretical investigations of Bi₅³⁺ potential energy surfaces should be considered to clarify the possibility of Jahn-Teller effect in Bi₅³⁺ lowest excited state.

4. From chloride to oxide glasses

Let us try to compare the luminescence bands assignment, obtained for the chloride glass, with the situation in oxide glasses, attracting much practical interest. The lack of luminescence from Bi₅³⁺ at room temperature, relatively short luminescence lifetime at 77K and different spectral range (red-shifted, relative to luminescence in oxide glasses) indicate that Bi₅³⁺ hardly contributes to room temperature NIR luminescence in oxide glasses. If we assume, that multiphonon relaxation is the main quenching process for excited Bi₅³⁺ at room temperature, this quenching will be even more severe in the oxide matrices (relative to chlorides) since phonon spectrum spans to higher energies here. On the other hand, demonstration of broadband luminescence from Bi₅³⁺ in chloride glass indicates, that bismuth and other post-transition metal polycations could be attractive active centers for broadband amplifiers and oscillators. Previously, another heavy metal polycation Hg₃²⁺ was shown to be luminescent, although in the visible range [42].

The luminescence in the first band (from Bi⁺) and in the second band (presumably from Bi₂⁴⁺) in our glass resembles the situation in oxide glass. Indeed, it was demonstrated [43], that germanate glasses have two luminescent bands: shortwave (1100 nm) with two-maxima (550, 720 nm) in excitation spectrum, resembling excitation spectrum of Bi⁺ in our glass, and longwave (1300 nm), with excitation profile with one maximum near 550 nm similar to what we assume for Bi₂⁴⁺. Increasing of bismuth concentration in germanate glass diminished luminescence at 1100 nm, but increased luminescence at 1300 nm. It is exactly what we should expect, supposing 1100 nm luminescence originating from Bi⁺ and 1300 nm one from Bi₂⁴⁺, since increased Bi³⁺ concentration shift equilibrium toward Bi₂⁴⁺:

In SiO₂-Al₂O₃ bismuth doped glasses relatively shortwave 1100 nm NIR luminescence is also characterized by two peaks in excitation spectrum (at 705 and 510 nm) [44]. By analogy with chloride glass we can suppose that this emission also originates from Bi⁺ monocation.

4. Conclusions

Lewis acidic chloride glass AlCl₃/ZnCl₂/BiCl₃, doped with subvalent bismuth was prepared. At the room temperature the dominant subvalent bismuth form was Bi₅³⁺ polycation, while Bi⁺ monocation was a minor constituent. Despite this fact, the NIR luminescence at room temperature originates only from Bi⁺ (peak at 1080 nm) and, probably, from its association product with Bi³⁺: dimer Bi₂⁴⁺ (peak at 1300 nm). At 77K broadband luminescence from Bi₅³⁺ emerges, spanning the range from 1300 to 2500 nm. Comparison of the luminescence characteristics allows suggesting, that Bi⁺ and Bi₂⁴⁺ also contribute to luminescence of oxide-based glasses, while Bi₅³⁺ does not. This situation exemplifies one possible problem with bismuth-doped glasses: after the preparation they can incorporate several subvalent bismuth species simultaneously and some of them, being non-luminescent, are either useless or introduce optical losses to active media. To circumvent this problem, the proper preparation conditions should be chosen. For example, to maximize the NIR luminescent Bi⁺ concentration and minimize unnecessary Bi₅³⁺ the glass transition temperature and glass acidity should be as high as possible.