1. Introduction

Rapidly developing global demand for fast information transfer and ultra-high data capacity networks calls for novel types of superbroadband optical amplifiers. NIR emitting laser materials that can be processed into low-loss optical fiber are thought to offer a pathway towards this goal. In this context, glasses and glass ceramics doped with Ni²⁺, Cr⁴⁺ and Bismuth-ions of unknown valence, respectively, are presently receiving significant attention [1–6]. Over the last decade, especially Bi-doped glasses have seen the transition from original discovery [3] to the first demonstration of continuous wave lasing and efficient all-fiber optical amplifiers [7,8]. The optical properties of this class of materials offer the potential to replace classical fiber amplifiers: their emission spectrum covers the whole spectral range in which silica telecommunication fibers exhibit their lowest optical loss [5,6]. Only recently, a quantum efficiency of up to 1.0 ± 0.05 has been reported [9]. However, while reports on Bi-activated NIR emission now cover almost all classes of inorganic glasses, only little is known on NIR-emitting Bi-doped crystalline materials [3–6,10,11]. So far, broadband emission has been observed from Bi-doped crystalline SrB₄O₇ [10], RbPb₂Cl₅ [11], 2MgO·2Al₂O₃·5SiO₂, BaF₂, α-BaB₂O₄, and zeolites [12–16]. While a crystalline environment may be favorable for the optoelectronic properties of the emission center (quantum yield, excitation cross section, quenching effects and lifetime), studying crystalline matrices may also allow new insights into the nature of the active center: as of today, the seemingly simple question which Bi species is responsible for NIR photoluminescence remains highly controversial. Single consensus is that NIR luminescence is totally different from what can be observed from Bi³⁺ and Bi²⁺-doped materials. Going further, arguments diverge between either Bismuth in lower valence states or Bi⁵⁺ [3–7,10–18]. Experimental evidence particular for or against Bi⁵⁺ often appears counterintuitive. For instance, Bi⁵⁺ was initially proposed as the most probable NIR emission center [3] and its existence in NIR emitting glasses was recently verified by X-ray photoelectron spectroscopy (XPS) and analyses of the extended X-ray absorption fine structure (EXAFS) [19,20]. Razdobreev et al. performed optically detected magnetic resonance spectroscopy (ODMR) on Bi-doped silica glass during excitation (808 nm, 1.8 K) and suggested the radiative e-h recombination between filled electronic configurations of Bi⁵⁺O_n^2- molecules as the origin of NIR luminescence [21]. At the same time, however, these experimental findings are contradicted by other observations [4,5,18]: (1) at elevated temperature, Bi₂O₃, conventionally used as source of the bismuth dopant, readily decomposes into BiO or even Bi atoms; (2) stabilization of Bi⁵⁺ appears to require elevated amounts of alkali or alkaline earth species (e.g. NaBiO₃ or KBiO₃), which are not present in most NIR-active Bi-doped glasses; (3) the known Bi⁵⁺-containing compounds become extremely unstable at temperature higher than 300 °C [22,23]; (4) Bi NIR emitting glasses can be prepared in reducing atmosphere [6] or vacuum [24]; (5) bismuth nanoparticles are present in NIR-emitting bismuthate glasses (although these nanoparticles are not the emission centers) [18]; (6) NIR-emission can be extinguished completely by adding oxidizing reagents (e.g. Sb₂O₅ [18] or CeO₂ [25]) to the glass; and (7) it was found in several compounds that NIR emission centers can coexist with Bi²⁺ [10,13,14]. In the framework of Duffy’s [26] concept of optical basicity, conditions that favour lower valence states of polyvalent constituents may be achieved in glass melts with low basicity. Vice versa, increasing basicity (obtained by, e.g., the addition of alkali or alkaline earth species to borate or silicate glasses) favours the formation of Bi⁵⁺. However, only glasses with low optical basicity seem to provide NIR luminescence [27–30].

These arguments suggest that NIR emission is related to the presence of low-valence bismuth species [4–8,10–18, 27-25]. Beyond that, however, the question as to which of the various possible low-valence Bi species is (or are) the actual emission center(s) remains intensively debated also on this side of the general discussion. While atomic spectral data suggest Bi⁰ as the active center [18], also Bi⁺ [27], Bi₂ [17^,31,32], Bi₂^- [12,17], Bi₂^2- [12] and point defects [33] have been proposed, based on electron spin resonance spectroscopy (ESR), the quadratic dependence of center absorption on bismuth concentration and quantum chemical calculations, respectively. In the present study, a different strategy is followed to approach the nature of the emission center. Compared to corresponding glasses, crystals exhibit a much more defined lattice configuration and, usually, a much lower free volume. In glasses on the other hand, network connectivity and, hence, ligand field strength can easily and to a large extent be altered by introducing modifier ions. This property enables glass networks to accommodate relatively large clusters or even particles (e.g. Au, Ag or Bi [34,18,35]), what is difficult to achieve in crystalline matrices. In crystals, the size of a dopant should rather be comparable to the size of either a network constituent (or vacancy) or an interstitial network site. For example, in the case of SrB₄O₇:Bi, NIR-emission could be observed from the polycrystalline material, but not from single crystals [10,15], suggesting that emission centers precipitate at grain boundaries. Here, M₂P₂O₇ (M = Ca, Sr, Ba) are chosen as the host crystal. It is demonstrated that Bi³⁺ and Bi²⁺ can be doped into all three lattices, but that NIR emission can be obtained only from Ba₂P₂O₇. For this specific material, analyses provide evidence for Bi⁰ as NIR emission center.

2. Experimental

Doped and un-doped samples of M₂P₂O₇ (M = Ca²⁺, Sr²⁺ and Ba²⁺) were prepared via solid state reaction. For that, analytical grade reagents CaCO₃, SrCO₃, BaCO₃, NH₄H₂PO₄ and Bi₂O₃ were used as raw materials. Individual batches of 20 g were weighed according to nominal compositions M_2(1-x)P₂O₇: 2xBi (x = 0, 0.001, 0.003, 0.005, 0.01, 0.02, 0.03, 0.05) and mixed thoroughly. NH₄H₂PO₄ was added in excess of 3% to compensate for volatilization losses (in the following, all percentages are given in mol.%). To prevent batch foaming, heating and cooling rates were adjusted to between 80 K/h and 110 K/h. The complete reaction procedure comprised the following steps: (1) preheating of all batches at 500 °C for 6 h in air, using somewhat bigger alumina crucibles; (2) secondary grinding of the pre-reacted samples to improve homogeneity; (3) sintering at 1100°C, again in alumina crucibles. Step (3) was conducted in four different ways (A-D), respectively. (A) eight crucibles, containing samples of Ba_2(1-x)P₂O₇: 2xBi, were put onto an alumina plate in a bigger alumina container. The container was covered with an alumina plate. A layer of ~2 cm of powdered graphite was put on the bottom of this container. At the reaction temperature of 1100 °C, this leads to an atmosphere where practically all oxygen and CO₂ are converted to CO. Sintering was then conducted for 24 h. (B) another eight samples of Ba_2(1-x)P₂O₇: 2xBi were first sintered for 24 h at 1100 °C in air to form solid cylinders. After this procedure, they were taken out of the crucible and put into an alumina container in a way similar to (A). Annealing in CO atmosphere was performed for 1 h at 1100 °C. (C) one sample of Ba_1.994P₂O₇: 0.3%Bi was sintered in the same way as (B), but without the secondary treatment in CO atmosphere. (D) a final sample of Ba_1.994P₂O₇: 0.3%Bi was sintered according to (B), but conducting a third annealing step (1h, 1100 °C), again in air. In parallel, Ca_1.994P₂O₇: 0.3%Bi and Sr_1.994P₂O₇: 0.3%Bi were prepared according to (A). Sample nomenclature is MXXY, whereby “M” represents the employed alkaline earth species (Ca, Sr, Ba), “XX” the concentration of Bi in mol.% (01, 03, 05, 10, 20, 30, 50) and “Y” the employed sintering procedure (A-D). Samples that were prepared via paths (A) and (B) are further referred to as A-type M₂P₂O₇:Bi and B-type M₂P₂O₇:Bi, respectively.

Crystal structures of obtained samples were examined by X-ray diffractometry (XRD, Siemens Kristalloflex D500, 30 kV/30 mA, Cu K_α, λ = 1.5405 Å, scan rate of 1 °/min). UV-Visible dynamic and static optical emission and excitation spectra, and fluorescent lifetimes of Bi³⁺ and Bi²⁺ were recorded with a high-resolution photoluminescence spectrometer (Horiba Jobin Yvon Fluorolog-3), using a static Xe lamp (450 W) and a Xe flashlamp (75 W) as excitation sources, respectively. NIR emission and uncorrected excitation spectra were obtained with the setup described in Ref. [36]. For measuring the excitation spectra of NIR emissions at 1.1 and 1.15µm, an emission monochromator mounted with a grating of 300 g/mm blazing at 1250 nm and an excitation monochromator with a grating of 600 g/mm blazing at 750 nm were used. The grating of the emission monochromator was adjusted to exactly disperse emission at either 1.1 or 1.15µm, respectively. The grating of the excitation monochromator was then gradually turned to separate excitation light (500 nm-1050 nm) from the Xe light source. At the same time, variation of the emission intensity was recorded as a function of excitation wavelength. Corresponding lifetime of the NIR emission center was recorded with the help of an 80 Hz optical chopper and an oscilloscope with a bandwidth of 1.5 GHz at a sampling rate of 8 × 10⁹s⁻¹, and averaged over 1024 individual measurements. The experimental error of lifetime measurements is ± 1%.

3. Results and discussion

3.1 Crystal structure

XRD patterns of M_2(1-x)P₂O₇: 2xBi are consistent with JCPDS cards 71-2123, 75-1490, 83-990 for M = Ca²⁺, Sr²⁺ and Ba²⁺, respectively. In a first consideration, this confirms that synthesized samples are of pure phase, tetragonal space group P4₁ for Ca_2(1-x)P₂O₇: 2xBi, orthorhombic space group P_nma for Sr_2(1-x)P₂O₇: 2xBi and hexagonal space group P$\stackrel{\_}{6}$2m for Ba_2(1-x)P₂O₇: 2xBi, respectively. Figure 1 shows the XRD pattern of sample Ba03A and the result of Rietveld refinement (fullprof suite [37], ) to space group P$\stackrel{\_}{6}$2m. The calculation produces lattice parameters a = 9.423Å, c = 7.081Å and the unit cell volume V = 544ų with the accuracy R_B = 0.086 and R_F = 0.053. This agrees well with reported reference data, i.e. a = 9.415Å, c = 7.078Å and V = 543ų [38]. Noteworthy for the following considerations, in Ba₂P₂O₇, there are two types of barium sites, Ba(1) and Ba(2), two types of phosphorus sites, P(1) and P(2), and four types of oxygen sites, O(12), O(22), O(11) and O(21) (inset of Fig. 1). Ba(1) is surrounded by two O(12) and eight O(22). The bond length ranges from 2.545 to 2.977 Å (average: 2.767 Å). Surrounding Ba(2) are four O(12), four O(22) and four O(21) with bond lengths between 2.744 and 3.141 Å (average distance 2.910 Å). Ba(1) and Ba(2) sites have the symmetry of m2m (depicted in the insets A and B of Fig. 1). In one unit cell, layers Ba(1) and Ba(2) type are stacked along the c-axis in I-II-I order (inset C of Fig. 1).

 

Fig. 1 XRD pattern of Ba03A and corresponding Rietveld refining results. The inset shows the coordination environment of a Ba(1)- (inset A) and Ba (2)-site (inset B). Inset (C) schematically shows stacking of the two Ba-layers in a unit cell. P and O atoms are omitted for clarity.

Download Full Size | PDF

3.2 Photoluminescence from A-type M₂P₂O₇:Bi

Ba₂P₂O₇:Bi samples that were prepared via path (A) (i.e., sintering for 24 h in CO atmosphere) exhibit strong NIR photoemission, centering at a wavelength of ~1.1 µm (for an excitation wavelength of 723 nm, Fig. 2A ). The full width at half maximum (FWHM) of this emission peak is 147 nm, narrower than for typical Bi-doped silicate and germanate glasses, but wider than for crystalline α-BaB₂O₄:Bi (108nm [14], ) and comparable to the emission bandwidth of Bi-doped zeolites (152 nm [16], ). The peak position corresponds to observations in other bismuth doped crystals [11,13,14] and remains unchanged with changing dopant concentration. On the other hand, emission intensity is strongly dependent on nominal bismuth concentration (inset of Fig. 2A). It exhibits a maximum at a dopant concentration of ~1 mol.%.

 

Fig. 2 A: NIR emission spectra of Ba10A excited at 586nm, 723 nm (dotted lines: Gaussian peak fits), 838nm and 924nm, respectively, and dependence of NIR emission intensity on nominal bismuth concentration (inset). B: Uncorrected excitation spectra of Ba10A for emission at 1100 and 1150 nm, respectively.

Download Full Size | PDF

Lifetime of NIR emission was measured at room temperature, using a laser diode (980 nm, 200 mW) as excitation source. Temporal decay follows a single exponential decay equation with a lifetime of 634 μs. The uncorrected excitation spectrum (monitoring emission at 1.1 µm) is shown in Fig. 2B. It comprises at least eight individual peaks, at 586, 700, 723, 838, 900, 924, 955 and 997 nm (compared to only four structureless absorption bands that can be observed in typical Bi-doped glasses at ~500, 700, 800 and ~1000 nm [18]). While absorption at 723 nm is strongest, switching the excitation wavelength between either of the eight peaks does not affect the position of the emission peak (Fig. 2A). This finding is similar to observations of RbPb₂Cl₅:Bi [11], but differs from what was reported on BaF₂:Bi. In the latter case, the ratio of the emission bands at 1070 and 1500 nm was reported to change when changing the excitation wavelength from 500 to 590 nm [13]. Nevertheless, large similarity in emission, excitation and lifetime of Ba₂P₂O₇:Bi as compared to other Bi-doped crystals allows to assume that in Ba₂P₂O₇:Bi, the emission center is of the same nature as it is in other NIR-emitting Bi-doped crystals and glasses. In contrast to glassy hosts, the presence of much sharper excitation peaks is attributed to the fact that Bi-dopants are incorporated on strongly localized lattice sites. The appearance of additional absorption bands is related to the possible presence of Bi on more than one lattice site (see section 3.1 and the following paragraphs).

A more careful examination of the emission spectrum in Fig. 2A reveals an asymmetric shape with a tail extending towards the IR-side. The emission peak is best fit by two Gaussian functions, centering at 1.1 and 1.15 µm. FWHM of these individual emission peaks are 142 nm and 265 nm, respectively. The excitation spectrum that corresponds to the 1.15 µm emission is shown in Fig. 2B. While here, too, eight absorption peaks can be detected, a change occurs in the ratio between the excitation bands 924 nm and 723 nm: their intensities become comparable. Also the emission lifetime of the 1.15 µm band was found to differ from the 1.1 µm band, i.e. 656 μs as compared to 634 µs. These differences further indicate that two different centers contribute to NIR emission from Ba₂P₂O₇:Bi. In sharp contrast, no NIR emission could be detected from Ca03A and Sr03A at excitation wavelength between 500 and 900 nm.

3.3 Photoluminescence from B-type M₂P₂O₇:Bi

After sintering in air for 24 h (first step of synthesis path (B)), all Ba₂P₂O₇:Bi samples exhibit typical Bi³⁺ UV photoluminescence. When subsequently annealing for another 1 h in CO atmosphere (second step of synthesis path (B)), intense red photoluminescence (peaking at 716 nm) can be detected together with emission in the NIR spectral range. Corresponding emission and excitation spectra are shown in Fig. 3A (sample Ba10B). Emission at 716 nm exhibits a FWHM of 67 nm. Monitoring this emission peak, the excitation spectrum consists of bands at ~286, 384 and 616 nm respectively. Shape and position of the emission peak are independent on excitation wavelength. The decay curve of the 716 nm emission can be fit to a single exponential equation (correlation coefficient of 99.06%), resulting in a lifetime of 14.18 μs (Ba10B, excitation at 286 nm). Both emission intensity and emission lifetime depend on bismuth concentration: maximum intensity was observed for a critical concentration of 1 mol.% (Fig. 3B). Differing from this observation, lifetime was found to decrease monotonically with increasing concentration (Fig. 3B).

 

Fig. 3 A: Emission (λ_ex = 286, 388 and 618 nm) and excitation spectra (λ_em = 716 and 733 nm) of Bi²⁺ in Ba10B; and excitation spectra (λ_em = 716 nm) of samples Ba01B, Ba30B and Ba50B (curves I, II and III respectively). Insets show a zoom at the spectral regions 370-400nm (left) and 600-630nm (right). B: Lifetime and relative fluorescence intensity of B-type Ba_2(1-x)P₂O₇: 2xBi (x = 0.001, 0.003, 0.005, 0.01, 0.02, 0.03, 0.05).

Download Full Size | PDF

Photoluminescence from Ba10B exhibits strong similarity to that of Bi²⁺-doped compounds [39,40]. Excitation bands at 286, 384, and 618 nm and emission at 716 nm can readily be assigned to the electronic transitions ²P_1/2 → ²S_1/2, ²P_1/2 → ²P_3/2 (2), ²P_1/2 → ²P_3/2 (1) and ²P_3/2 (1) →²P_1/2, respectively. Noteworthy, strong changes in the position of the emission peak in different host materials (588nm→660nm→716nm for Bi²⁺-doped SrB₄O₇, SrB₆O₁₀ and Ba₂P₂O₇, respectively) imply a strong dependence of the optoelectronic properties of Bi²⁺ on ligand field strength [39,40].

In order to examine the distribution of Bi²⁺-species in the Ba₂P₂O₇ host, dynamic emission spectra were recorded (excitation at 286 nm). Data are shown in Fig. 4 . For increasing temporal delay, a gradual change can be observed in the shape of the emission peak. Starting at 733 nm, the position of the emission peak starts to blue-shift after a delay of 3 μs. After 6 μs, the maximum of the peak has moved to 716 nm. For longer delay times, the position remains fixed. This observation shows the presence of at least two emission sites in Bi²⁺-doped Ba₂P₂O₇ and agrees with crystallographic data: Bi²⁺-ions are assumed to occupy both of the two Ba²⁺-sites. A slight difference can be detected in the excitation spectra when monitoring emission at 716 nm and 733 nm, respectively (Fig. 3): for 733 nm emission, the two excitation peaks lie at 381 and 619 nm while for emission at 716 nm, they are shifted to 384 and 616 nm. This reflects differences in the Ba-O bond length for the two different Ba-sites (Ba(1)-O is somewhat shorter than Ba(2)-O, see 3.1). When Bi²⁺ is incorporated on a Ba(1) site, it experiences a slightly stronger crystal field as compared to the Ba(2) site. As a result, splitting of the two lowest excited states (²P_3/2 (1) and ²P_3/2 (2)) increases slightly for Bi²⁺ on Ba(1) (splitting energies of 10092 cm⁻¹ and 9808 cm⁻¹ are derived from the excitation spectra for Bi²⁺ on Ba(1) and on Ba(2), respectively). In the following, the two sites will be referred to as Bi²⁺ (1) and Bi²⁺ (2). We attribute emission at 716 nm and 733 nm to Bi²⁺ (2) and Bi²⁺ (1), respectively. Since the Ba(2) site offers a slightly larger free volume (inset of Fig. 1), Bi²⁺ may preferably occupy this site. This leads to dominance of the 716 nm emission in Fig. 3A.

 

Fig. 4 Time resolved emission (λ_ex = 286 nm) spectra of Ba10B (labels: delay time).

Download Full Size | PDF

Comparing static and dynamic emission curves (Figs. 3A and 4) reveals that at the initial stage, the peak 733 nm is dominant. This can be interpreted as a result of energy transfer from Bi²⁺ (2) to Bi²⁺ (1). Since the excitation spectra of 733nm and 1100nm emissions overlap partly with the emission at 716nm (see Figs. 5 and 6 ), it is assumed that emission at 716nm can partly be re-absorbed, contributing to the emission at 733 nm. This energy migration is fast and dominates particularly the early stage (< 6 μs, Fig. 4).

 

Fig. 5 Emission and excitation spectra of Ca03A (1: λ_ex = 465nm; 3: λ_em = 653nm) and Sr03A (2: λ_ex = 450nm; 4: λ_em = 702nm).

Download Full Size | PDF

 

Fig. 6 Emission spectra of Ba03A, Ba03B, Ba03C and Ba03D under 723 nm excitation. The curve of Ba03D is shifted vertically for clarity.

Download Full Size | PDF

Normalizing the intensity of the excitation spectra of Ba01B to Ba50B to the height of the peak at 286 nm reveals a gradual change in the ratio of excitation intensities at 618 nm and 286 nm (R_618nm/286nm). R_618nm/286nm is 0.338 for a Bi-concentration of [Bi] = 0.1%, increases to 0.800 for [Bi] = 1.0%, and again decreases to 0.423 for [Bi] = 3.0%. The reason for this can presently not be explained.

The critical distance R_c, the mean separation between the nearest NIR emission centers at the critical concentration x_c, can be estimated from [41]

If the transfer process occurs via an electric dipole-dipole interaction, R_c can also be estimated from excitation and emission spectra [41],

Here, f is the oscillator strength of the ²P_1/2 → ²P_3/2 transition in Bi²⁺, ∫f(E)F(E)dE represents the overlap of normalized emission and excitation spectra and E is the energy of maximum spectral overlap. For E = 1.91 eV and ∫f(E)F(E)dE = 0.045eV⁻¹, combining Eq. (1) and 2 produces a value of f = 0.03, comparable to the oscillator strength of the 4f⁷ → 4f⁶5d transition in Eu²⁺ [42]. Since in Bi-doped Ba₂P₂O₇, there are coexisting Bi³⁺, Bi²⁺ and NIR emission centers, the actual value of x_c is lower than the nominal critical concentration of 1.0% and, hence, R_c is somewhat longer than 25.9 Å. As a result, also f should be slightly larger than 0.03. This means that in the BaP₂O₇ matrix, the in principle parity forbidden transition ²P_1/2 → ²P_3/2 (Bi²⁺) is strongly allowed.

In a similar way, emission from Bi²⁺ can be observed in Ca₂P₂O₇:Bi and Sr₂P₂O₇:Bi (Fig. 5), peaking at 653 nm and 702 nm with FWHM of 68 nm and 80 nm, respectively. The absorption properties of all three components enable the use of blue LEDs (InGaN) for excitation, suggesting further examination with respect to potential use in white-light emitting devices (WLEDs) [39,40].

3.4 NIR-emission from Ba₂P₂O₇:Bi

To assess the nature of NIR emission from Bi-doped Ba₂P₂O₇, the various synthesis procedures A-D have to be considered. After a single-step sintering in air for 24h / 1100 °C (path C), only Bi³⁺-emission centers (active in the UV spectral range) can be observed (e.g. sample Ba03C). Neither red (typical for Bi²⁺) nor NIR photoluminescence occur. Emission from Bi³⁺, however, is strongly dependent on excitation wavelength: exciting at 290 nm produces a dominant peak at ~335 nm with a shoulder emission at ~356 nm. Switching the excitation wavelength to 310 nm results in dominance of the 356 nm emission peaks. Correspondingly, the excitation spectra peak at 290 and 310 nm when emissions at 335 and 356 nm, respectively, are monitored. This observation suggests that Bi³⁺ ions may occupy both of the two Ba-sites, Ba(1) and Ba(2) (Fig. 1). The presence of two kinds of Bi³⁺ emission centers is further confirmed by lifetime analyses: lifetime for the emission at 335 nm is 5.6 μs (290 nm excitation) while that of the 356 nm emission is 8.5 μs (310 nm excitation). These excitation and emission bands can be assigned to transitions between ground state ¹S₀ and excited state ³P₁ in Bi³⁺ [10].

Bi²⁺-related emission from B-type Ba₂P₂O₇:Bi (two-step sintering with second step 1 h in CO, 1100 °C) has been discussed in the previous chapter. In these samples, also broad NIR emission can be observed (sample Ba03B in Fig. 6). NIR emission strongly intensifies if the treatment time in CO is prolonged (sample Ba03A in Fig. 6; NIR-emission from A-type Ba₂P₂O₇:Bi is discussed in paragraph 3.2). When sample Ba03B is annealed for a third time, but in air (path D), both Bi²⁺-related and NIR-photoluminescence can be completely erased (sample Ba03D in Fig. 6). These observations lead to the conclusion that Bi-based NIR emission centers can be created only in reducing atmosphere in Ba₂P₂O₇. The related oxidation-reduction mechanism is reversible. The fact that Bi⁵⁺ cannot be stabilized at elevated temperature in CO atmosphere makes it rather unlikely that this species is the emission center. Annealing the sample in air appears to transform all Bismuth into its trivalent form. Further treating in CO results in partly and reversible reduction of Bi³⁺ to Bi²⁺ and Bi⁰: Bi³⁺ $\stackrel{CO,1100°C}{\to }$ Bi²⁺, Bi⁰ $\stackrel{air,1100°C}{\to }$Bi³⁺.

As discussed previously [18], atomic spectral data of Bi⁰ show best consistence with spectroscopic data of Bi-doped glasses and crystals.

The presence of NIR photoemission in Ba₂P₂O₇:Bi, but not in Ca₂P₂O₇:Bi and Sr₂P₂O₇:Bi can be interpreted as follows: Conventionally, dopant ions are supposed to efficiently substitute a specific lattice atom of a crystalline compound only if the difference between ionic radii of the dopant and the lattice species should be as close as possible. This maximal allowed difference appears to be ~30% [43]. Then, ionic radii of the potential Bismuth-species and their comparison to the available network sites must be considered. Relative ionic radius differences D_r between possible dopant species (Bi³⁺, Bi⁰, Bi dimer ions) and potential substituted ions (Ba²⁺, Sr²⁺, and Ca²⁺) in M₂P₂O₇:Bi (M = Ca, Sr, Ba) can be calculated from

In Ca₂P₂O₇, there are four types of calcium sites coordinated by 7, 7, 8, and 9 oxygen ions, respectively [46]. In Sr₂P₂O₇, there are two types of 9-coordinated strontium sites [47]. The two Ba-sites in Ba₂P₂O₇ (CN = 10 and 12, respectively) have already been discussed. Radius data of Ca²⁺, Sr²⁺ and Ba²⁺ with these coordination numbers is adopted from Ref. 45. Derived values of D_r are given in Table 1 . They allow the following conclusions: (1) Bi³⁺ can substitute alkaline earth ions in either of the three lattices (M²⁺ in M₂P₂O₇). As a confirmation, when sintered in air, all three materials exhibit Bi³⁺-related photoluminescence. (2) According to the size mismatch, none of the three lattices can accommodate dimer ions. (3) Bi⁰ species can be accommodated in the Ba₂P₂O₇ lattice, but not in Ca₂P₂O₇. Accommodation in Sr₂P₂O₇ appears borderline. Correspondingly, NIR emission is observed only from Ba₂P₂O₇, but not from Sr₂P₂O₇ or Ca₂P₂O₇.

Table 1. Relative difference in ionic radii (D_r, %) between matrix cations and various Bi species (data on effective ionic radii taken from Refs.44, 45).

View Table

For judging whether a dopant may actually occupy a certain lattice site, besides radius matching, also the charge match (or mismatch, respectively) has to be taken into account. If Bi³⁺ substitutes M²⁺, the charge mismatch is + 1. If the charge can be compensated appropriately, ion size remains the dominant factor in deciding whether or not substitution is possible. However, for larger charge mismatches (e.g. + 2 as in the case of Bi⁰ ↔ M²⁺), if accompanied by a moderate radius mismatch, charge may become the dominant destabilizing factor. This is why we think that Sr₂P₂O₇ does not accommodate Bi⁰ and, hence, does not exhibit NIR emission.

Commonly, Bi-doped glasses exhibit absorption bands at ~500, 700, 800 and 1000 nm. Emission is usually located within 1.1 - 1.65 μm [3–8,27–32,48]. However, additional absorption bands were recently reported by Hughes et al. (~1180 nm, Bi-doped chalcogenide [49]) and Dvoirin et al. (1.4μm of bismuth in silica fiber [50], ). For the case of chalcogenide glass, emission was found at 2.0 and 2.6 μm [49]. Generally, spectroscopic data appear very complex. They may origin from different Bi-based centers, some of which can exist in an amorphous matrix but not in crystals. For example, Hughes et al. assigned absorption and emission to Bi₂^2- dimers [⁴⁹]. On the other side, regarding this example, Bi₂^2- cannot be stabilized on Ba-sites in Ba₂P₂O₇ (see above discussion).

Thermal treatment of different Bi-doped silica glasses in hydrogen atmosphere [1,29,51] has been shown to promote precipitation of Bi nanoparticles, whereas in agreement with another study of Bi-particle containing glasses [18], no NIR emission could be observed in these studies. However, this observation does in fact not rule-out Bi⁰ or Bi-clusters as the active NIR emission centers.

Therefore, we attribute the absorption and NIR emission of polycrystalline Ba₂P₂O₇: Bi to Bi⁰ species. Actually, laser oscillation of Bi⁰ atoms has been observed by Zhu and Liu [52] at 1171.1 nm as early as 1982. Chou and Cool reported pulsed laser action of Bi⁰ at 5.326μm [53]. Laser operation was also realized in gaseous Bi₂, but in the wavelength range of 590-790nm [54]. In KCl:Bi⁰, Goovaerts et al. observed the Bi⁰ absorption at 629 and 725 nm. These bands were assigned to transitions between ground state ⁴S_3/2 and excited states ²D_3/2 and ²D_5/2 [55].

Presently, for reasons of calibration, the absorption bands that were found here cannot be assigned to individual transitions. However, relying on Ref. 18, the groups of bands at 838, 900, 924, 955 and 997 nm peaks, 700 and 723 nm peaks, and the individual band at 586 nm are tentatively assigned to transitions in two types of Bi⁰ centers between a doubly degenerated ground state ⁴S_3/2 and the degenerated excited levels ²D_3/2, ²D_5/2 and ²P_1/2, respectively. NIR emission is then attributed to ²D_3/2→⁴S_3/2.

4. Conclusion

In summary, we have reported on a novel type of NIR-luminescent Bi-doped crystal, Ba₂P₂O₇. The luminescence peaks around ~1.1 µm with a FWHM of ~140 nm and a lifetime of > 600 μs from two different lattice sites. The absorption peaks cover the spectral range from 550 nm to 1050 nm. NIR emission centers can be generated and removed reversibly by treating the polycrystalline material in CO atmosphere or air, respectively, at 1100 °C. This reveals directly lower valence bismuth as NIR emission centers. Based on previous study and analyses of available lattice sites, 1.1µm emission of Ba₂P₂O₇:Bi is attributed to Bi⁰. For reasons of charge and radius mismatch, this species cannot replace Ca²⁺ and Sr²⁺ in Ca₂P₂O₇ and Sr₂P₂O₇, and accordingly, none of them exhibits NIR luminescence. In Ba₂P₂O₇:Bi, two types of each Bi³⁺, Bi²⁺ and Bi⁰ may be accommodated on Ba(1) and Ba(2) sites, respectively. For the case of Bi²⁺, the critical distance for energy transfer is 25.9 Å. M₂P₂O₇:Bi²⁺ (M = Ca, Sr and Ba) exhibit broad blue absorption and intense red emission.