1. Introduction

Infrared luminescent materials active in new spectral region of 1150-1500nm are of crucial importance to develop new fiber lasers and optical amplifiers, which could find potential applications such as in medicine, astrophysics, or next generation optical communication system [1–5 ]. Bismuth doped glasses were recently recognized as such a promising candidate which could luminesce in an extremely broad spectral range, that is, from 1000 to 1700nm, or even to 2000nm [6–12 ]. Shortly after this, the potency in fiber laser and amplifier was confirmed, and broadly tunable fiber laser was initially demonstrated from 1150nm to 1300nm, and soon extended to 1550nm [13–15 ]. These works have turned the researches on bismuth doped materials and devices into a very hot area in consequent years.

Nevertheless, one of open problems in the area is the nature of bismuth near infrared (NIR) luminescence. It remains unclear despite of the subsequent successes in bismuth fiber device. The problem comes from the multiple valence states (from positive to negative) and species (from single ion to cluster ion) of bismuth, and the easy conversion between them. It is very challenging to stabilize the element to a specific state and to identify the state with traditional techniques such as X-ray photoelectron spectrum, electron energy loss spectroscopy etc. For design of efficient bismuth doped glass in the future, it is essential to establish the origin of bismuth NIR emission. For this, many efforts have been made in experimental and theoretical aspects [16–27 ]. For instance, Sokolov et al had performed quantum chemical calculations on spectroscopic properties of Bi₂ ^-, Bi₂ ^2-, Bi⁺, Bi₂ ⁺, Bi⁺…V_Cl ^-, Bi…≡Si- Si≡ complex (formed by interstitial Bi atoms and glass intrinsic defects) etc [24,25 ]. Romanov et al had tried to prepare Bi⁺ doped single crystals of chlorides, such as CsCdCl₃ [26,27 ].

Our previous work reported the Bi⁰ luminescence from Ba₂B₅O₉Cl: Bi with a lifetime of ~30μs and the conversion of Bi²⁺ to Bi⁰. The luminescence peaks at 1030 and 1061nm due to substitution of Bi⁰ for two different types of barium sites in the compound, and it only appears after the doped sample was treated in a reducing atmosphere and it disappears once the sample was in turn sintered in an oxidizing atmosphere [21]. For the sample treated in the reducing atmosphere, will the valence of bismuth change directly from + 2 to 0 or via + 1 to 0? If the reduction reaction happened via + 1, we should be able to detect the luminescence from Bi⁺. Motivated by this, we doped trivalent bismuth into the sample first in air, and then reduced it in situ by hydrogen flow. We noticed that as the treatment time is prolonged gradually, Bi²⁺ and Bi⁰ luminescences reach maximum as the times are 0.5 and 2.5 hours, respectively [21]. In between, we observed a luminescence spanning from 600 to 1300nm with a lifetime longer than 1ms becomes strongest when the treating time is 1 hour. Comparison to the previous reports confirms it is from Bi⁺, and very different from either Bi²⁺ or Bi⁰ in luminescence wavelength and lifetime. We also observed the visible Bi⁺ emission at 660nm for the first time due to the transition from ³P₂ to ³P₀.

2. Experimental procedure

All samples of Ba₂B₅O₉Cl: Bi were prepared by solid state reaction at high temperature. Analytic reagents BaCO₃, H₃BO₃, BaCl₂.2H₂O and Bi₂O₃ were selected as raw materials. The samples were weighed according to Ba_{2(1- x%)}B₅O₉Cl: 2x%Bi (x = 0, 0.1, 0.3, 0.5, 1.0, 1.5, 2.0, 2.5, 3.0, 5.0) and mixed, and sintered similar to reference [21]. The samples were sintered in air or in 95%N₂/5%H₂ for different times at 850°C to control the valence state of bismuth. Boiling temperature of Bi₂O₃ is 1890°C while the samples were synthesized at temperature not higher than 850°C. So, the bismuth loss due to volatility will be negligible, and the nominal content should be close to the bismuth content residual in the samples after sintered.

X-ray diffraction (XRD) patterns of the samples were recorded with a Rigaku D/max-IIIA X-ray diffractometer (40kV, 1.2° min−1, 40mA, Cu-Kα1, λ = 1.5405Å). Static excitation and emission spectra and dynamic emission decay spectra were measured with a high resolution spectrofluorometer Edinburgh Instruments FLS 920 equipped with a red sensitive single photon counting photomultiplier (Hamamatsu R928P) in Peltier air-cooled house for ultraviolet to visible range and a liquid nitrogen cooled photomultiplier (Hamamatsu R5509-72) for NIR range. A microsecond pulsed xenon flashlamp μF900 with an average power of 60W applicable to a lifetime of 1μs to 10s was used to measure the decay curves. A 450 W ozone-free xenon lamp was used for steady-state measurements. For the low temperature measurements, the sample was housed in a closed cycle helium cryostat. All the measurements were performed at room temperature except otherwise specified.

3. Results and discussion

3.1 Effect of bismuth doping on crystal structure

We examined the XRD patterns of all the samples which were prepared in air or N₂/H₂, and they are all pure phase of Ba₂B₅O₉Cl. Rietveld refining results for the doped samples prepared in air reveal that the unit cell shrinks with the increase of bismuth content [21]. This is a sign that smaller Bi³⁺ ions have successfully substituted for Ba²⁺ ions. The samples were submitted for consequent treatment in the flow of nitrogen and hydrogen in different times to precipitate in situ different valences of bismuth species inside the compound.

3.2 Conversion of Bi³⁺ to Bi²⁺by in situ reduction process

The Bi doped samples prepared in air show only very weak blue luminescence from Bi³⁺ at room temperature. This is due to thermal quenching. As the temperature is lowered to 10K, the blue emission can be enhanced more than 6 times. Once the samples were annealed in hydrogen, red luminescence was detected immediately, and it was intensified as the annealing time increased when annealed at 850°C. It became strongest as the time prolonged to 0.5h, and the luminescence was weakened as the dwell time was longer than 0.5h. As the time was 2.5h, the intensity decreased to 60% of the sample which was annealed for 0.5h. The red luminescence is peaked at 655nm as Fig. 1(a) illustrates, and it depends tightly on the concentration of bismuth. The strongest luminescence was found at Bi% = 0.5%. As displayed in Fig. 1(b), the excitation spectrum of the emission at 655nm comprises three excitation bands located at 273, 440 and 624nm, respectively. Luminescence decay curves were collected and Fig. 1(c) illustrates one of examples, which was monitored at the emission at 655nm upon an excitation into 273nm. The decay follows well the single exponential decay equation and fitting to it produces a lifetime of 8.00μs. The lifetime shows clear dependence on temperature, and it reduces from 13.47μs @ 10K to 8.00μs @ 300K, as shown in Fig. 1(d). As we changed the excitation wavelengths to 440 or 624nm, the emission peak does not move and the lifetime is always around 10μs even at 10K. Surprisingly, the lifetime does not show the dependence on the Bi concentration. Comparing the results above to the reports on Bi²⁺ doped compounds confirms the red emission at 655nm is from Bi²⁺ ions in the compound [21]. This implies that the conversion takes place from Bi³⁺ to Bi²⁺ after in situ reduction process. The excitations at 273, 440 and 624 nm correspond to ²P_1/2 → ²S_1/2, ²P_1/2 →²P_3/2(2), and ²P_1/2 → ²P_3/2(1) respectively and the emission at 655nm is due to ²P_3/2(1) → ²P_1/2.

 

Fig. 1 (a) Emission spectra (λ_ex = 273nm) of Ba_2(1-x%)B₅O₉Cl: 2x%Bi (x = 0.1, 0.3, 0.5, 0.7, 1.0, 1.5, 2.0, 3.0, 5.0) treated in N₂/H₂ for 0.5h, and inset demonstrates in situ reduction; Orange balls are for barium atoms, yellow for boron, deep blue for oxygen, green for chlorine, and blue for H₂ atmosphere; (b) excitation spectra (λ_em = 655nm) of Ba_1.99B₅O₉Cl: 1%Bi treated in N₂/H₂ for 0.5h. Visible excitations were enlarged for clarity; (c) decay curves (λ_em = 655nm, λ_ex = 273nm) of Ba_1.99B₅O₉Cl: 1%Bi (0.5h treatment in N₂/H₂) at different temperatures as indicated and (d) the dependence of the emission lifetimes on temperature (green ball) and bismuth concentration (black ball).

Download Full Size | PDF

3.3 Conversion of Bi²⁺→ Bi⁺→Bi⁰ by in situ reduction process

As the annealing time continues prolonging from 0.5h to 2.5h, Bi⁰ luminescence at 1055nm is intensified gradually at expense of Bi²⁺ emission and it reaches maximum at the dwell time of 2.5h. This reflects the valence change of bismuth from + 2 to 0 in the in situ reduction process. Beside Bi⁰ and Bi²⁺ luminescences, we found a very broad emission spectrum starting from 600 to 850nm upon 330nm excitation, as Fig. 2(a) shows, and it comprises two peaks at 660 and 790nm. Since the shape of the peak at 790nm implies there should be a tail in the longer wavelength, we detected the emission spectrum with another NIR photomultiplier.

 

Fig. 2 (a) Emission spectra (λ_ex = 330nm) of Ba_1.99B₅O₉Cl: 1%Bi treated in N₂/H₂ for different time; (b) emission spectra (λ_ex = 330nm) of Ba_2(1-x%)B₅O₉Cl: 2x%Bi (x = 0.1, 0.5, 1.0, 1.5, 5.0) treated in N₂/H₂ for 1h detected by visible and NIR photomultipliers; The spectra were rescaled and combined; (c) excitation spectra (λ_em = 790nm, λ_em = 970nm) of Ba_1.99B₅O₉Cl: 1%Bi treated in N₂/H₂ for 1h and 0h; (d) decay curves of Ba_1.99B₅O₉Cl: 1%Bi treated in H₂/N₂ for 1h.

Download Full Size | PDF

Unexpectedly, a new emission peak was found centering at 970nm beside the 790nm emission, as shown in Fig. 2(b). The corresponding excitation spectrum of the emission at 790nm is depicted in Fig. 2(c), and it is made up of two new strong peaks at 330nm and 606nm along with a weak peak at 478nm from Bi⁰. The excitation spectrum of the 970nm emission comprises two strong peaks at ~320nm and ~670nm. This is different from the case of the 790nm emission. Perhaps, the excitation peak at 606nm has immerged into the tail of the excitation peak at 670nm since the 670nm peak is very broad. The excitation peaks of Bi⁰ become slightly stronger as compared to that of the emission at 790nm. This should be due to the more overlap between the 970nm emission and the Bi⁰ luminescence. Figure 2(d) illustrates the decay curves of the emissions at 660nm, 790nm and 970nm upon the excitation of 330nm. The decay of the 790nm emission complies well with the single exponential decay equation, and the lifetime is 1.151ms. The decays of the emissions at 660nm and 970nm are different from the 790nm emission. They experience a fast decay rightly followed by a slow decay, which should be due to the overlaps to Bi²⁺ and Bi⁰ emissions. The fast decay lies within 200μs and it is contributed mainly by Bi²⁺ and Bi⁰ species. The slow decays show the intrinsic lifetimes of the emissions at 660nm and 970nm to be 1.149 and 1.192ms, respectively. So the lifetimes of the emissions at 660nm, 790nm and 970nm are in same order of millisecond irrespective of upon which excitation wavelength it is, as displayed in Fig. 2(d), and they are very different from either Bi²⁺ or Bi⁰ species. So, what's the origin of these emissions?

The evolution of the emission along with the annealing time in hydrogen may give us the inspiration on it. Look back on Fig. 2(a), what we find is that the air sample does not show any luminescence in the spectral range from 600 to 850nm upon 330nm excitation while the annealed samples all exhibit the broad luminescence. As the annealing time increases, the emission at 660nm decreases all the way down, and the emission at 790nm increases first and it reaches the maximum as the time extends from 0.5h to 1.0h, afterwards, it is weakened rapidly. Since the time when the intensity maximum of the emission at 790nm is 1.0h, and it is right in between 0.5h for Bi²⁺ and 2.5h for Bi⁰, we believe the species of bismuth which is responsible for the long lived luminescence is most probably Bi⁺. This is because Bi⁺ is the only intermediate state between Bi²⁺ and Bi⁰, and also because it is very difficult to precipitate bismuth metallic colloid or polycations inside the compound due to the limit space of barium sites available to bismuth species in Ba₂B₅O₉Cl.

In molten salts of AlCl₃-NaCl, Bjerrum et al found five absorptions at 308, 333, 585, 658, 694 and 901 nm from Bi⁺, and they assigned 308 and 333 nm to the transition of ³P₀→¹D₂, 585, 658 and 694 nm to ³P₀→³P₂, and 901nm to ³P₀→³P₁, respectively [28]. We did not observe all similar peaks. But as we compared our results to theirs, it can easily come to the conclusion that the excitation peak at 330 nm may be due to ³P₀→¹D₂ and the peaks at 606 and 670nm are to ³P₀→³P₂, and the emissions at 660nm and 790nm may be to ³P₂→³P₀, and the emission at 970nm is to ³P₁→³P₀. The lifetime of the emissions at 660nm and 790nm are almost same, perhaps due to the same origin of transitions from ³P₂ which is split by crystal field. It, however, is slightly shorter than the lifetime of the emission at 970nm as shown in Fig. 2(d). This is because the transition of ³P₁→³P₀ is slightly more forbidden. Romanov et al found the absorptions at 601nm and 660nm and an emission at 950nm with a lifetime of 400μs in KMgCl₃: Bi⁺, and the absorptions at 596nm and 644nm and an emission at 980nm with a lifetime of 525μs in KAlCl₄: Bi⁺ [26], and the absorptions at 610nm and 650nm and an emission at 980nm with a lifetime longer than 300μs in CsCdCl₃: Bi⁺ [27]. These results are similar to this work and confirm the assignments. In view of radius match, Bi⁺ ions once formed will substitute barium sites, which are coordinated by seven oxygen and two chlorine atoms and located in the center of channels formed by [B₅O₉]_∞ chains and [BO₃] units. They will experience stronger crystal field due to the enhanced interaction between Bi⁺, BO₃ and BO₄ units in the host as compared to the free state, and their absorptions and emission, therefore, lie at longer wavelengths than free Bi⁺ ions. The decay of Bi⁺ luminescence follows well the single exponential decay equation at different temperatures between 10 to 300K. The lifetime of the 790nm emission upon 330nm excitation depends tightly on temperature. It is 2.071, 2.041, 1.944, 1.754, 1.476ms at 10, 100, 150, 200, 250K, respectively.

4. Conclusions

In all, the valence conversion happens from Bi²⁺ via Bi⁺ to Bi⁰ in Ba₂B₅O₉Cl: Bi by the in situ reduction process. The intermediate species shows extraordinarily broad luminescence from 600 to 1200nm with the lifetime longer than 1ms, due to the cascade transitions from ³P₂ and ³P₁ to ³P₀, and it is rather different from Bi⁰ and Bi²⁺ in the compound. The lifetimes of Bi⁰, Bi⁺ and Bi²⁺ show similar dependence on temperature in the range from 10 to 300K. This work deepens our understanding on the intriguing nature of bismuth luminescence. It should be helpful for design of efficient bismuth laser materials in future.