1. Introduction

The wonder metal bismuth exhibits wonder spectroscopy properties particularly when introduced into glasses and crystals [1–30]. For instance, when bismuth is controlled to “+2,” it can absorb ultraviolet and blue lights and emit intensive red lights. The Bi²⁺ doped crystals can then be used to improve color rendering index of commercial white light LEDs [8–11]. As bismuth goes lower than “+2,” it can emit lights broadly distributed in the spectral range from 1 to 1.6μm at the same time with considerable absorptions in visible and near infrared range [1–7,12–26]. This endows the doped glasses a unique ability to provide high optical gain in the bandgap where traditional rare earth based devices cannot reach [2–7,27,28]. The intriguing spectral properties of such materials tempt scientists in materials and optics communities to perform intense researches on this topic. It is because of this tremendous input throughout the world that rapid progress has been enabled and witnessed within only two years from first lasering to efficient all fiber optical amplifiers and lasers [27,28]. Now, bismuth doped broadband amplifiers have been recognized as next generation devices after erbium doped fiber amplifier. However, despite of sequence success in fabrication of device and exploration of various potential laser materials, the nature of the optically active NIR emission centers remains unclear [17–25].

This seemingly elemental problem is complicated partly by the diversity of valence states of monoatomic bismuth ion and the lapsable transformation between these states, and also by the strong proneness of bismuth to oligomerize or copolymerize into unusual homo- or hetero-nuclear clusters. The fascinating cluster species are often with novel unexpected chemical and optical behaviors, bonding and reactivity. For instance, neutral bismuth clusters can emit NIR light [14]. Strikingly, Bi₅³⁺ was found to even luminesce in superbroadly spectral range from about 1 to 4μm [29]. None of any rare earth and transition metal doped optical materials could be on a par with it (see below, Fig. 2(b) ). Nevertheless, yet to date, the studies carried out on the fluorescent properties of bismuth polycation groups have been very limited especially in the range of NMIR [29–40].

 

Fig. 2 (a) Raman spectrum of Bi₅(GaCl₄)₃-A-36h. Vibration modes of Bi₅³⁺ are labeled in the figure. Inset is the polycation group of Bi₅³⁺; (b) Emission (curves 1-3) and excitation (curves 4-5) spectra of Bi₅(GaCl₄)₃-A-36h at room temperature. Curve 1 is upon 470nm excitation, curve 2 is the emission spectrum calibrated to the detector spectral response, and both of them were recorded with liquid N₂ cooled NIR photomultiplier (Hamamatsu R5509-72). Curve 3 was collected with a PbSe detector under the excitation of 808nm laser diode. Curves 4 and 5 were measured by monitoring the emissions at 1400 and 1500nm, respectively. Curve 4 is shifted vertically for clarity. For reference, curve 6 an emission spectrum of Er³⁺ doped fluorophosphate glass is listed upon 808nm excitation and the FWHM is much smaller than Bi₅(GaCl₄)₃.

Download Full Size | PDF

Basic motivation for this work is to study the optical properties of bismuth polycations. It will in one hand help better understanding the origin of NIR fluorescence of bismuth doped materials; on the other hand, it will possibly prompt the finding of optical materials in new spectral regime for new laser sources.

When measuring Raman spectroscopy of Bi₅³⁺ in (1 – n – butyl – 3 – methylimidazonlium) Cl/AlCl₃ ionic liquid ([BMIM]Cl/AlCl₃) with 1064 nm laser excitation, Ruck et al observed strong luminescence disturbance over the spectral range of 1075-1852 nm [32]. This hints possible excited electronic states of Bi₅³⁺ lying around 1064nm. However, none of absorptions have been reported beyond 1000nm [29–39]. Sun et al followed Ruck's approach and synthesized Bi₅(AlCl₄)₃ where bismuth exists in an entity of Bi₅³⁺, and reported a broad fluorescence band spanning the range of 900-1600 nm with a FWHM >500 nm [14,15]. This agrees well with the finding by Ruck et al [32]. Very recently, we synthesized Bi₅(AlCl₄)₃ via a molten salt route with NaAlCl₄ as solvent and excess Lewis strong acid AlCl₃ as stabilizer, and found it fluoresce in NIR-MIR at room temperature [29]. In the final products, however, Bi₅(AlCl₄)₃ is miscellaneous with NaAlCl₄ phase, and it seems physically impractical to separate them from each other within this framework. Subsequently Romanov et al found similar NMIR fluorescence of Bi₅³⁺ in chloride glass at 77K, and it is thermal quenched at room temperature [40]. In the glass Bi₅³⁺ coexists with Bi⁺ and Bi₂⁴⁺, which probably is due to the violent and indiscriminate reactivity of acidic molten salts as reaction medium [40].

Selection of Bi₅(GaCl₄)₃ is prompted by these concerns:

To elucidate these queries, in this work, we will first synthesize the pure phase of Bi₅(GaCl₄)₃ with benzene as solvent and mild oxidizer such as BiCl₃ or GaCl₃ for bismuth metal at room temperature, determine the crystal structure of Bi₅(GaCl₄)₃ with the classic Rietveld refining process, analyze the spectroscopic properties of Bi₅³⁺ in Bi₅(GaCl₄)₃ by collecting Raman spectrum, excitation spectra, and site-selectively excited emission spectra, time resolved NIR emission spectra as well as absorption spectrum beyond 1000nm, and study the thermal resistance to laser irradiation in the case of uncooling. The compound exhibits extraordinarily broad luminescence in near to mid infrared spectral region even at room temperature.

2. Experimental section

2.1 Materials and sample synthesis

Anhydrous GaCl₃ (Alfa, 99.999%), BiCl₃ (Aladin, 99.9%), bismuth metal powder (Aladin, 99.999%) and benzene (Sigma-Aldrich, AR) were used as received. All the operations were carried out in a glovebox under dry nitrogen. When molar ratio of GaCl₃ to benzene X(GaCl₃, benz)<0.3 and the solution is saturated with bismuth, it tends to form into two liquid phases. When X(GaCl₃, benz) = 0.3 and molar ratio of bismuth to GaCl₃ X(Bi, GaCl₃) is 0.2, solution will produce a homogenous liquid phase [42]. In view of this, we kept X(GaCl₃, benz) = 0.3 constant throughout the study. Though BiCl₃ or GaCl₃ has been suggested as suitable oxidizing species to prepare Bi₅(GaCl₄)₃ using benzene as reaction medium, none of reaction rate has been evaluated before when the two oxidizers are used respectively [37,42]. Partially for this, two parallel experiments were designed according to the works by Ulvenlund et al [37,42]. Experiment (A) with BiCl₃ as oxidizer selected the molar composition of 4.00Bi:1.00BiCl₃: 8.00GaCl₃:26.67 Benzene. 178.3mg GaCl₃ was first dissolved in 0.3ml benzene, 105.8mg Bi was then added right after 39.9mg BiCl₃. Reaction time was counted from zero to 96 hours. The mixture was rocked to improve the homogeneity. Experiment (B) with GaCl₃ as oxidizer employed 1.00Bi: 5.00GaCl₃: 16.67Benzene. 178.3mg GaCl₃ and 42.3mg Bi were consequently dissolved in 0.3ml benzene, and further admixed by rocking. Reaction time changed from 0 to 120 hours. The reactions were all confined in evacuated and flame sealed Pyrex tubes. Both experiments were at room temperature. For convenience, samples were coded as Bi₅(GaCl₄)₃-A(or B)-reaction time. For instance, Bi₅(GaCl₄)₃-A-36h means that the sample was synthesized by Experiment (A) with reaction time of 36h. The solutions were deep reddish brown before separation as inset shown in Fig. 1(b).

2.2 Measurements and characterization

Because the solutions are too deep, any structures will be smoothed by the strong absorption. However, when the glass tubes are shaken, some reddish crystals will adhere on the inner side of the tube wall. By letting light through the wall thinner covered with the crystals, the absorption spectrum was measured with a Perkin Elmer Lambda-900 UV-VIS-NIR spectrometer. Before other measurements, the solution was centrifuged, and red powders were separated and washed with cyclohexane, and part of them was transferred and sealed in the same type of tube. The residual solution shows weak luminescence when compared to the powder. Static and dynamic excitation and emission spectra as well as emission lifetime were recorded with a high-resolution spectrofluorometer Edingburgh FLS 920 with a liquid nitrogen cooled NIR photomultiplier (Hamamatsu R5509-72). Due to limitation of the detector in the spectrometer, only time resolved NIR emission spectra could be achieved in the range from 900 to 1650nm. Photoluminescence between 1000 and 3500 nm was observed with a PbSe detector on a Horiba Jobin Yvon Triax 320 fluorometer, using an 808 nm laser diode as pump source. X-ray diffraction patterns of the samples were recorded with a Rigaku D/max-IIIA X-ray diffractometer (40kV, 1.2°min⁻¹, 40 mA, Cu-Kα1, λ = 1.5405Å). Raman spectra were measured on a Horiba Jobin Yvon LabRAM Aramis spectrometer equipped with a 633 nm He-Ne laser. Signal was collected vertical to laser beam.

3. Results and discussion

3.1 Structural identification

When refining the crystal structure of Bi₅(GaCl₄)₃ in R-3c, Lindsjö et al found a poorer result with significantly larger values of R factors [41]. The situation could be improved as the structure was evaluated with R3c space group. And goodness of fit (GOF) could be, thus, produced as 1.215. This led to the obvious change of cation geometry of Bi₅³⁺. The two apical bismuth atoms Bi(2) and Bi(2)’ of the entity, which are crystallographically equivalent in R-3c, became quite different. This yielded two different bond lengths 2.97 and 3.02Å between the equatorial atoms and the apical atoms. However, they could not find any experimental proof to confirm the symmetry. Ulvenlund et al [37] believed that it crystallized in space group R-3c and satisfactory residual values (R_p = 3.9%; R_wp = 5.2%; R_B = 2.3% and GOF = 3.0) could be obtained when the strongest reflection (012) around 2θ = 10.47° was excluded during the refining.

To settle the divergence we prepared the samples with experiments A and B. The structures of the samples are rather analogous and, thus, only XRD pattern of sample Bi₅(GaCl₄)₃-A-36h is shown exemplarily in Fig. 1(b). For the structural refining (with FullProf Suite), angular range is restricted to 9.8≤2θ≤42 and it includes the reflection (012). The start of the refining is based on the crystallographic data of Bi₅(AlCl₄)₃. And stop condition of convergence is set as all parameter shifts are lower than 0.1 of the corresponding estimated standard deviation. The refinement converges with residual values of R_p = 7.47%, R_wp = 14.6%, R_exp = 11.0%, GOF = 1.76 and R_B = 3.95%. The refining results are listed in Fig. 1(b) also with the difference profile between observed and calculated values. This comparison clearly suggests that Bi₅(GaCl₄)₃ belongs to the same space group R-3c as Bi₅(AlCl₄)₃, which consists with Ulvenlund et al. The analysis produces structural parameters shown in Table 1 , based on which Fig. 1(a) the unit cell of the compound is drawn, and bond distances and angles are calculated and listed in Table 2 along with those of Bi₅(AlCl₄)₃.

Table 1. Wyckoff positions (wp) and coordinates for Bi₅(GaCl₄)₃

View Table | View all tables in this article

Table 2. A comparison of unit cell parameters (Å), distances (Å) and angles (deg) between Bi₅(GaCl₄)₃ and Bi₅(AlCl₄)₃

View Table | View all tables in this article

Isostructural with Bi₅(AlCl₄)₃, Bi₅(GaCl₄)₃ comprises two types of bismuth atoms Bi(1) (Wyckoff position 18e) and Bi(2) (12c), one type of Ga atoms and two types of chlorine atoms Cl(1) and Cl(2) (see Table 1 and Fig. 1(a)). Occupation ratio is 3:2 for Bi(1) and Bi(2). These bismuth atoms form into a single type of Bi₅³⁺ polyhedron, as Ulvenlund et al had noted before composed of three equatorial Bi(1) and two equivalent apices Bi(2). Two Cl(1) and two Cl(2) ions coordinate one Ga ion and form into the tetrahedron group of GaCl₄^-. In each unit cell, there are on average six Bi₅³⁺ and eighteen GaCl₄^- polyhedra. The Bi₅³⁺ cluster preserves an ideal symmetry of D_3h for which geometry the representation of Raman and Infrared active normal vibrations can be Γ_vib(D_3h) = 2A₁’ + A₂” + 2E’ + E”. Among the modes, A₁’, E’ and E” are Raman active [38]. A 633nm He-Ne laser rather than former 1064nm was selected as pump light for collecting Raman since the compound has less absorption at the wavelength as will be shown in the excitation spectrum. A good spectrum of Raman was therefore achieved and depicted as Fig. 2(a). Typical vibrations of trigonal bipyramid Bi₅³⁺ indeed appear at 54, 62, 99, 120, 124, 138, 151cm⁻¹, which can be attributed to ν₅(E’), ν₄(E’), ν₆(E”), ν₂(A₁’), ν₂(A₁’), ν₁(A₁’) and ν₁(A₁’) of Bi₅³⁺ [29,32–35,38], respectively. This presents clear evidence for the symmetry of Bi₅³⁺ polycation. Rest Raman peaks at 87, 111 and 166cm⁻¹ are due to lattice modes and ν₂(E) of GaCl₄^- [32,38].

For Bi-Bi and Bi-Cl bonds, the distances are slightly longer in Bi₅(GaCl₄)₃ as compared to Bi₅(AlCl₄)₃. The distance between two bismuth apexes is elongated to 4.696(2)Å in the direction of c-axis. This elongation expands the angles Bi(2)-Bi(1)-Bi(2)’ and Bi(1)-Bi(1)’-Bi(2), and simultaneously suppresses Bi(1)-Bi(2)-Bi(1)’ (see Table 2). Cell parameters of Bi₅(GaCl₄)₃ are refined as a = 11.886(0) Å and c = 30.136(0) Å, slightly larger than those of Bi₅(AlCl₄)₃ which are a = 11.86 Å and c = 30.10 Å reported by Krebs et al [39], a = 11.8712 Å and c = 30.1203 Å by Ruck et al [32], and a = 11.8698 Å and c = 30.1133 Å by Cao et al [29]. These, we think, are due to the substitution of Ga³⁺ (Ionic radius is 0.47 Å for four coordination.) for Al³⁺ (Also for four coordinated ions, radius is 0.39 Å.) ions.

When Ulvenlund et al measured liquid X-ray scattering spectrum of a solution containing Bi₅(GaCl₄)₃, they found [42] that the reduced radial distribution function of the sample reveals the structure of the building blocks Bi₅³⁺ and GaCl₄^-. Peaks at 2.24, 3.13, 3.58 and 4.70 Å can be found in Fig. 7 of [42] and they are due to the average distances of Ga-Cl, Bi-Bi (axial-equatorial and equatorial-equatorial) and Bi-Cl, and the distance of Bi(2)-Bi(2)’. They agree with our results (2.242, 3.142, 3.584, 4.696 Å) better than those (2.249, 3.170, 3.479, 4.649 Å) by Ulvenlund et al [37]. This probably is because our refining is more reasonable since it does not abandon the important reflection peaks such as the strongest (012).

3.2 Near to mid infrared (NMIR) luminescence from Bi₅(GaCl₄)₃

When exposed to the light of 470 nm, broad near infrared luminescence was observed in the range of 900 to 1650nm from all examined samples. The emission spectrum of representative sample Bi₅(GaCl₄)₃-A-36h is depicted as curve 1 of Fig. 2(b), and it is similar to Bi₅(AlCl₄)₃ as reported in references [29,33,34]. The full width at half maximum (FWHM) is 555nm, comparable to 570nm and 510nm of Bi₅(AlCl₄)₃ reported by Cao et al and Sun et al [29,33,34]. The sudden falloff around 1380nm and in 1500-1650nm is due to the spectral response drop of the NIR photomultiplier in the areas. After correction of curve 1 over the detector sensitivity curve 2 in Fig. 2(b) is resulted and the tail of the spectrum rises steeply in the wavelengths longer than 1380nm. This indicates a possible luminescence peak lying beyond 1650nm. To confirm it a broader band PbSe detector was summoned. Subsequent measurement testifies the presence of the band as clearly shown in curve 3 of Fig. 2(b). It peaks at ~1835nm with FWHM of 800nm and persists to ~3500nm. The emission peak of Bi₅(GaCl₄)₃ is located at longer wavelength than ~1700nm of Bi₅(AlCl₄)₃ [29]. This probably correlates with the slight change of geometry of Bi₅³⁺ in Bi₅(GaCl₄)₃. As discussed above, the entity is elongated in the c-axis when compared to Bi₅(AlCl₄)₃.

Curve 1 in Fig. 3(a) is the absorption spectrum of the representative sample Bi₅(GaCl₄)₃-A-36h. Absorptions can be found at 430, 580, 854, 1137(sharp), respectively. An additional peak around ~1100nm can be traced though it is superimposed by the sharp peak at 1137nm. To assign these accurately, absorption spectrum of benzene (curve 2 of Fig. 3(a)) is listed as references. Peak at 1137 is from second overtone stretching of C-H in benzene (see curves 1 and 2 of Fig. 3(a)). Intense absorption blurs fine structures in the wavelengths shorter than 400nm, the excitation spectra, however, can differentiate these. Thus, the excitation spectra of the compound of Bi₅(GaCl₄)₃ were recorded upon infrared emissions and illustrated as curves 4 and 5 in Fig. 2(b). The excitation spectrum shows no clear dependence on emission wavelength varying in the range of 1000 to 1600nm. This is possible due to the existence of only one type of Bi₅³⁺ emission centers in the compound as aforementioned (see Fig. 1(a)). Curves 4 and 5 are similar, and they are visually composed of two strong peaks at 473 and 766nm and two shoulders at 362 and 430nm. The peaks at 473 and 766nm are both broad and they contain the 854nm peak in curve 1 of Fig. 3(a) and other unresolved peaks which have been observed in the compound of Bi₅(AlCl₄)₃. Due to instrumentation limit, excitation spectrum cannot be achieved in the range of wavelengths longer than 900nm. In all, by combining absorption and excitation spectra, it can be concluded that absorptions at 362, 430, 473, 580, 766, 854 and ~1100nm are from Bi₅³⁺. On the basis of the calculation by Corbett et al [36] on the entity of Bi₅³⁺, the former six can be assigned to e’→e’(3)(E’), e’(1)→a₂’(E’), e’(1)→a₁’(E’), a₁’→e’(2)(E’), e’(1) →e’(2)(E’) and e”→e’(2)(E’), respectively. The last one was observed for the first time and, however, it cannot be assigned properly at the moment since the complete energy level diagram of Bi₅³⁺ is absent for beyond 900nm [36].

 

Fig. 3 (a) Absorption spectra of (1) Bi₅(GaCl₄)₃-A-36h and (2) benzene, I₀ and I are the incident and transmitted radiation through the sample; (b) Decay curve of the emission at 1365nm from Bi₅(GaCl₄)₃-A-36h excited by 470nm at room temperature. Red circles are fitted results by an equation I = 442.596exp^(-t/8.67) with χ² = 1.086.

Download Full Size | PDF

Excitation into either of these absorption bands does not lead to the shape change of the emission spectrum. This proves in the other way that they all belong to the same emission center. The situation is similar to Bi₅(AlCl₄)₃ [29]. Figure 3(b) illustrates exemplarily the temporal decay of the emission at 1365nm upon 470nm excitation. The curve fits well the single exponential decay equation with χ² = 1.086, and the fitting produces lifetime of 8.67μs, comparable to 5.96μs of Bi₅(AlCl₄)₃ at room temperature [29,33,34]_. Fig. 4(a) depicts the time resolved emission spectra of Bi₅(GaCl₄)₃-A-36h excited by 470nm at room temperature. With increasing delay time, intensities of the emission monotonically decrease while the shape of the spectra does not change. This again evidences one type of infrared emission centers in the compound.

 

Fig. 4 (a) Time resolved emission spectra of Bi₅(GaCl₄)₃-A-36h excited by 470nm at room temperature. Delay times are marked beside the curves; (b) Dependence of the intensity of infrared emission@1835nm of Bi₅(GaCl₄)₃-A-36h on pump power of 808nm laser diode.

Download Full Size | PDF

For potential practical applications such as laser gain medium, excited state absorption and appropriate pumping schemes are essential issues to be considered. Here, for instance, we examined the 808nm pump plan, since 808 nm diode laser is low cost and more importantly it is commercially accessible. The dependence of infrared luminescence on the pump power can to some extent reflect the information on the issues. When the pump power increases from 400mW to 1100mW, the emission intensity increases linearly (see Fig. 4(b)), so in this regime, no severe excited state absorption occurs. Nevertheless, when the laser power rises higher than 1100mW, emission intensity starts to fall slightly, what we think perhaps is due to thermal dissociation of the compound of Bi₅(GaCl₄)₃, since its melting points is 321°C, five degree lower than the structural analogue Bi₅(AlCl₄)₃ [37]_. Similar phenomenon appeared in Bi₅(AlCl₄)₃ [29]. In view of this, initializing laser operation with the crystal needs efficient cooling especially when the excitation power exceeds the W-regime.

3.3 Effect of oxidizing species on reaction rate of formation of Bi₅(GaCl₄)₃

Ulvenlund et al found with GaCl₃ as the oxidant that introduction of bismuth metal to a solution of GaCl₃ and benzene could induce immediately a strongly colored liquid [35]. Rather, here, we observed different results. Since Bi₅(GaCl₄)₃ can luminesce we can then use the luminescence signal as a formation indicator of the crystal. After reaction for 60 hours the solution becomes reddish brown and red precipitates appear at the bottom of the tube and typical infrared luminescence of Bi₅(GaCl₄)₃ can be detected as shown in Fig. 5(a) . As the reaction continues, the luminescence becomes stronger. As the reaction time prolongs from 96 to 120 hours, the increase of the luminescence is fairly slight. Contrarily, with BiCl₃ as the oxidizing species for bismuth metal, we found the crystal starts to come into being after reacting for 4hours as depicted in Fig. 5. Comparing the two experiments with different oxidizing reagents, we can find out that the rate of the reaction using BiCl₃ oxidant is much faster than that of GaCl₃ especially at the very beginning period.

 

Fig. 5 (a) Dependence of fluorescence intensity of Bi₅(GaCl₄)₃-A (-●-) and Bi₅(GaCl₄)₃-B (-●-) on reaction time; (b) Evolution of emission spectrum of Bi₅(GaCl₄)₃-A with reaction time. Curve 1: 2h; curve 2: 4h; curve 3: 8h; curve 4: 24h; curve 5: 36h; curve 6: 72h; curve 7: 96h.

Download Full Size | PDF

4. Conclusion

Clearly, the compound of Bi₅(GaCl₄)₃ can be synthesized by redox reactions with BiCl₃ or GaCl₃ as the mild oxidizing species for bismuth metal in benzene. The rate of the former reaction is much speedier than the latter in the beginning phase. Addition of bismuth metal into GaCl₃ + benzene solution did not initialize instant reaction, which is different from what Ulvenlund et al found. Rietveld refining process has identified that Bi₅(GaCl₄)₃ belongs to trigonal space group R-3c with a larger lattice cell than structural analog Bi₅(AlCl₄)₃. This is due to the substitution of Ga³⁺ for Al³⁺, and it can elongate the spacing between two bismuth apexes of Bi₅³⁺. The change of geometry of the entity doesn’t make the two apical bismuth atoms unequivalent as reported by Kloo et al; it also doesn’t destroy the ideal symmetry of Bi₅³⁺, that is, D_3h as further verified by Raman spectrum of the compound. But it results in the redshift of emission peak as compared to Bi₅(AlCl₄)₃. The closo-detahedra of Bi₅³⁺ exhibit extraordinary NMIR luminescence peaking at ~1835nm with the FWHM of 800nm and the lifetime of 8.67μs at room temperature. Consistent with crystallographic data, excitation, emission, dynamic decay and time resolved infrared emission spectra all prove only one type of emission centers in the crystal, that is Bi₅³⁺, which is intrinsically differing from those in bismuth doped glasses, as discussed in [29,40]. Except typical absorptions of Bi₅³⁺ reported previously, a new absorption was found at ~1100nm. Before thermally decomposing, Bi₅³⁺ doesn’t appear severe excited state absorption. This work opens a new direction to manipulate the fluorescence of polycationic species such as Bi₅³⁺ by controlling geometry of the clusters. Guided by it and choosing similar entities to Bi₅³⁺ as activators, we can in the future develop new glass systems, where they can be stabilized or prevail over other bismuth species. And this will become the solid foundation to develop practical devices working in the desired spectral range of NMIR, the frequencies where traditional rare earth based schemes cannot fulfill.