With the arrival of a new era of cloud storage, the Internet of things, and 5G/6G, the demand for improving information transmission capacity in optical communication systems, which is limited by the gain spectra of fiber amplifiers, is even more urgent^[1–3]. To satisfy this need, novel photonic materials with highly efficient and ultrabroadband near-infrared (NIR) emission performance have been pursued for a long time. Traditionally, rare earth (RE)- or transition metal (TM)-doped glasses, glass ceramics, or glass fibers can exhibit higher NIR emission efficiency; however, the narrow gain bandwidth (not exceeding 100 nm) of RE ions^[4–6] and the requirement of suitable crystal field for TM ion-doped glass^[7,8], have resulted in limited success.

Alternatively, Bi-doped glasses and fibers, found by Y. Fujimoto in 1999^[9,10], presenting ultrabroadband NIR luminescence features from 1000 to 1600 nm, have drawn extensive attention, which enables ultrabroadband gain material for full-wavelength amplification covering the entire optical communication window^[11–14]. However, the insufficient efficiency and low gain are the critical defects limiting the application of Bi-based fiber. Therefore, how to improve the NIR luminescence performance of Bi-doped glasses is of great importance to promote the development of the Bi-doped fibers toward highly efficient optical amplifiers and tunable lasers. Over the past two decades, various strategies have been attempted, ranging from modulating the glass networks, optimizing the melting regime (such as melting or annealing temperature, time, and atmosphere), using high-energy radiation or crystallization, co-doping other active NIR centers, to improving the NIR emission intensity and bandwidth of Bi-doped glasses.

To enhance the NIR emission intensity by modulating the glass networks will result in worse thermal properties of Bi-doped glasses (such as soft point and refractive index) at the same time, which will make it complicated to prepare the desired fibers^[15–18]. For optimizing the melting regime, it usually generates little effect when adjusting the melting or annealing temperature and time^[19,20]. However, melting in the H2 atmosphere will greatly enhance the emission intensity of Bi-doped glasses, but it is difficult and dangerous to control the reduction progress due to the rapid reaction at high temperatures^[21]. The use of high-energy radiation or crystallization to enhance the NIR emission intensity of Bi-doped glasses both bring challenges to fabricating fibers^[22,23]. Moreover, co-doping rare earth (RE) or TM active centers is a simple and commonly used method to broaden the NIR emission range to cover the whole optical communication window, but it is also faced with the problem of requiring coupling of multiple pump sources^[24,25]. Hence, it remains challenging to intensify and broaden Bi NIR emission in a proper way.

In this work, we developed a new Bi-doped nitridated germanate glass with tunable ultrabroadband photoemission (850–1700 nm), an enhancement of ∼70 times for the integrated intensity of Bi-related four emission bands, and a high external quantum efficiency (EQE) of ∼50% by regulating the short-to medium-range order of the glass via manipulation of Al2O3, as sketched in Fig. 1. To elucidate how the multiple Bi NIR centers respond to variations of Al2O3 introduced into the glass matrices, a series of microstructure analyses, including Fourier-transform infrared (FTIR) spectroscopy, Raman spectroscopy, Al27 nuclear magnetic resonance (NMR), have been leveraged. Here, the results show that the conversion among AlO6, AlO5, and AlO4 leads to the enhancement and different behaviors of Bi NIR emission. Nevertheless, introducing excessive Al2O3 into the glass will contribute to formation of local excess negative charges, resulting in low-valence Bi being oxidized into high-valence. Thus, a Bi-doped nitridated germanate glass with highly efficient, tunable, and ultrabroadband NIR luminescence, covering the entire optical communication window, can be achieved by reasonable design and manipulation of Al2O3. We believe that this work will contribute to the design and fabrication of Bi-doped glass fiber with highly efficient, tunable, and ultrabroadband NIR emission and provide deep insight into the mechanism of the multiple Bi NIR centers.

Glass samples with molar composition of (87−x)GeO2-xAl2O3-10BaO-3AlN-0.02Bi2O3 (x=0, 3, 6, 8, 9, 10, 11, 12, 13), where x denotes the content of Al2O3 in the glass sample, were prepared by the melting and quenching method. Analytical grade Al2O3, BaCO3, and high-purity GeO2, AlN, and Bi2O3 (99.99%) were used as starting materials. Batches of raw materials weighting 20 g were mixed homogeneously in an agate mortar. Each batch of the mixture was melted in an alumina crucible at a temperature of 1520°C in air for 30 min. Before the glass was formed, a quartz glass rod was used to stir the glass liquid evenly to achieve uniform mixing to a certain degree. Afterwards, the melt was cast onto a stainless-steel plate and pressed by another stainless-steel plate simultaneously so as to speed up the cooling process and avoid crystallization, and eventually form solid-state glass (labeled as xAl, according to the values of x). Finally, the glasses were cut and polished for subsequent optical analysis.

Optical absorption spectra of the samples were measured by using a Perkin Elmer Lambda 900 UV/VIS spectrophotometer covering the spectral range of 300–1200 nm. Static excitation and emission spectra, dynamic emission spectra excited at 460 nm, and luminescence lifetime were recorded by using an Edinburgh FLS920 spectrofluorometer equipped with a continuous-wave 450 W Xe lamp and a microsecond flashlamp (μF900) as excitation sources, respectively. The EQE of the samples was measured by an absolute photoluminescence (PL) quantum yield spectrometer (Quantaurus-QY Plus C13534-11, Hamamatsu Photonics). X-ray diffraction (XRD) patterns were characterized by using an X-ray diffractometer (Rigaku D/max-IIIA) with CuKα1 radiation (1.5405 Å; cathode voltage, 40 kV; current, 40 mA) in the 2θ range of 10°–80°. FTIR profiles were measured in transmission mode with a Bruker Vector 33 spectrometer by dispersing glass powders into KBr pellets. Raman spectra were obtained via a Raman spectrometer (Renishaw inVia) excited by a Nd:YAG laser at 532 nm with a power of 25 mW. A Bruker AVANCE III HD 400 instrument equipped with high-speed magic angle spinning (MAS) probe heads was used to measure the Al27 NMR spectra. If not being specially stated, all measurements were conducted at room temperature.

The NIR PL emission spectra of the glass samples containing different contents of Al2O3 over 0–13% (molar fraction), upon 460 nm excitation, are depicted in Fig. 2(a). It can clearly be observed that all the samples present ultrabroadband emission from 850 to 1700 nm and consist of four emission peaks located at ∼930, ∼1160, ∼1260, and ∼1520nm, respectively. The emission peak at ∼1160nm originates mainly from the typical transition of P31→P30 from Bi+, while the emission band at ∼1260nm is assigned to S43/2→P21/2 of Bi0^[16,26–28], and the origin of newly generated Bi NIR emission centered at ∼930 and ∼1520nm is still controversial^[29,30], the nature of which will be explored and interpreted in the following. Moreover, the intensity of the ultrabroadband NIR emission can dramatically be enhanced (that is, the integrated intensity is enhanced by ∼77 times for 11Al sample compared to that of 0Al sample) by regulating the content of Al2O3, and, the EQE of the glass without Al doping is only 2.52%, while that of 11Al is up to ∼50% [Fig. 2(c)], which has never been reported in Bi-doped NIR glasses^[31], and shows great potential for an ultrabroadband NIR light source. Interestingly, the relative intensity (I1260nm/I1160nm) of emission peaked at 1160 and 1260 nm can also be modulated by tuning the concentration of Al2O3, showing a monotonically decreasing trend with an increase of Al2O3 content, as plotted in Fig. 2(b) and Fig. S1 (see Supplementary Material for more information). Moreover, the emission intensities at ∼930 and ∼1520nm share a similar decreasing trend with that of ∼1260nm, which provides important clues to exploring the origin of the newly generated emission bands at ∼930 and ∼1520nm.

The UV–VIS–NIR absorption feature of Bi-doped nitridated germanate glass with various Al2O3 doping was examined. As presented in Fig. 3(a), the glass free of Al2O3 shows no visible absorption, indicating that there are few activate NIR centers in the sample. Nevertheless, a small amount of alumina leads to fantastic changes of characteristic absorption of Bi NIR centers in the range of 300–1200 nm. By increasing the Al2O3, three obvious absorption peaks located at ∼460, ∼570, and ∼710nm appear, in accordance with the typical transition of S43/2→P21/2 from Bi0, newly generated Bi NIR center, and P30→P32 from Bi+, respectively. Moreover, the absorption intensity increases further as Al2O3 varies from 0 to 11%, and reaches the optimum for the 11Al glass sample, but an opposite phenomenon occurs when further increasing Al2O3 content. The parallel variation between NIR PL and absorption spectra evidences that the number of multiple Bi NIR activation centers can be adjusted by controlling the Al2O3 concentration.

To reveal the deeper relationship between emission peaks and excitation wavelength in more detail, the dependence of emission peaks on excitation wavelength has been studied (Fig. S2; see Supplementary Material), and the contour plot of NIR emission-excitation map from 6Al glass sample was collected as shown in Fig. 3(b). Clearly, the Bi NIR centers show different responses when varying the excitation wavelength from 420 to 750 nm. When monitoring at ∼1260nm (Bi0), a wide absorption band from 454 to 500 nm was observed. Similarly, the centers at ∼930 and 1520 nm can only be excited at around 460 nm. This similar feature gives another clue that the root of the newly generated NIR centers at ∼930 and 1520 nm may be related to ∼1260nm (Bi0). The emission at ∼1160nm (Bi+), however, can simultaneously be excited by blue light (480–520 nm) and red light (∼700nm). The result discussed above further confirms that at least three types of Bi NIR centers coexist in the germanate glass.

To obtain more clues on active Bi NIR centers, the excitation spectra of the four emission bands were collected. As shown in Figs. 3(c)–3(f), the active Bi NIR centers share diverse excitation dependent features, which is consistent with the above discussion. On one hand, all the observed active Bi NIR centers absorb UV–VIS light and can be well excited by blue light, which matches well with commercial blue LEDs and shows great potential for blue LED-excited ultrabroadband NIR light sources. In addition, all the excitation intensities reach the optimum state at x=11, and decrease with further increasing Al2O3 content with respect to the absorption and emission spectra. On the other hand, the four excitation spectra are different in some details. The emission band at 1160 nm shares three excitation bands at ∼340, ∼508, and ∼718nm, respectively. However, the emission band at 1260 nm has two weak and broad excitation bands at ∼346 and ∼736nm and a strong excitation band at ∼502nm. Five excitation bands at ∼336, ∼388, ∼456, ∼570, and ∼682nm with various intensities are observed for the emission band at 930 nm; these complicated excitation features are different from the excitation bands at ∼354, ∼400, and ∼456nm for the emission band at 1520 nm. The complicated features for excitation spectra at 930, 1160, 1260, and 1520 nm discussed above indicate that the newly generated Bi NIR centers at ∼930 and ∼1520nm are quite different from the NIR centers at ∼1160 and ∼1260nm.

PL lifetime is another vital dynamic clue to illustrate the difference among the active Bi NIR centers. The decay curves of multiple active Bi NIR centers from 11Al glass sample are plotted in Figs. 4(a)–4(d). All the decay curves follow a double-exponential decay equation, I(t)=A1exp(−t/τ1)+A2exp(−t/τ2),(1)where I(t) is the intensity at time t; τ1 and τ2 are short and long decay components, respectively; and A1 and A2 are fitting constants. Fitting each curve to the equation produces τ1 and τ2, and the correlation coefficients always lie between 99.94% and 99.96%. The average fluorescence lifetimes τ¯ were calculated by the following formula: τ¯=(A1τ12+A2τ22)/(A1τ1+A2τ2).(2)

The obtained individual fluorescence lifetimes of 11Al glass sample at 930, 1160, 1260, and 1520 nm are 28.74, 482.84, 404.94, and 304.37 µs, respectively. Compared to that of Bi53+ (∼5μs)^[32,33], Bi+ (∼460μs)^[30], and Bi0 (∼400μs)^[34], the great difference in fluorescence lifetime indicates that ∼930 and ∼1520nm emission bands are truly rooted in other types of Bi species. Based on the results discussed above and previous study on the origin of Bi NIR centers, a reasonable conclusion can be drawn that the newly generated Bi NIR centers peaked at ∼930 and ∼1520nm are formed by the interaction between interstitial Bi0 and inherent defects generated by introducing AlN in germanate glass^[35,36].

It is widely accepted that the valence states of Bi show hypersensitization to the variation of the glass local environment around them^[22,37,38]. The details of the variation of the glass microstructure with the substitution of germanium by aluminum can contribute to the valence state change of Bi in the glasses. Hence, the local microstructure of the glass can help to extend our comprehension of the great enhancement and tunability of the Bi NIR emission. As shown in Fig. 5(a), the XRD patterns of xAl (x = 0–13) samples present no distinct diffraction peaks, confirming that the glass samples are in an amorphous state. In addition, the shape of the humps remains almost constant, which rules out the effect of nanocrystal formation on the Bi NIR emission with introduction of Al2O3.

To gain deep insight into the local microstructure feature of the glass, FTIR and Raman spectra of the glass samples are illustrated in Figs. 5(b) and 5(c), respectively. The strongest peak in the Raman spectra located at 446cm−1 is assigned to the symmetric stretching vibration of Ge–O–Ge in [GeO4]6 rings, while the latter peak lies at 524cm−1 can be attributed to the symmetric breathing mode of Ge–O–Ge in [GeO4]3 or [GeO4]4 rings^[39,40]. Although Al2O3 is introduced into the glasses, there is no significant shift of the Raman peaks, indicating that no transformation of [GeO4]n (n = 3, 4, and 6) rings happens. Furthermore, two main absorption bands at 593 and 848cm−1 and a wide absorption at 1036cm−1 are observed in the FTIR spectra. The absorption band at 593cm−1 is assigned to the symmetric stretching mode of Ge–O–Ge in the GeO4 tetrahedra, and the absorption bands at 848 and 1036cm−1 are due to the asymmetric stretching vibrations of Ge–O–Ge and the Ge–O stretching vibrations in the GeO4 tetrahedra^[41,42], respectively. Similarly, no distinct shift of the absorption bands is observed for the glass with various Al2O3, implying that the introduction of Al2O3 has little effect on the overall glass network.

To further probe the glass topology network, in particular of the short- to medium-range order of the glass samples with Al2O3 doping, the normalized Al27 NMR spectra of glass samples are depicted in Fig. 5(d). There are three main types of aluminum coordination polyhedra, including octahedra (≈0ppm), hexahedra (≈25ppm), and tetrahedra (≈50ppm) in the glass matrix^[43,44]. As expected, a vast amount of the alumina in the glasses exits in tetrahedral (AlO4) site, which acts as the network formerly in the glasses. To obtain visual conversion of the coordination polyhedra [AlOm] (m = 4, 5, and 6), Fig. 5(e) clearly presents the percentage of AlO4, AlO5, and AlO6 in each glass sample. With the increase of Al2O3, the percentage of AlO4 increased at the expense of AlO5 and AlO6 polyhedra. The variation of the NIR emission intensity almost coincides with that of AlO4 species, indicating that the AlO4 species play a vital role in affecting the multiple Bi NIR centers.

To the best of our knowledge, AlO4 species usually act as network formers to connect the destroyed glass network, while [AlO5] and [AlO6] species commonly serve as network intermediates and modifiers, respectively. On one hand, the conversion from AlO5 and AlO6 species to AlO4 species will strengthen the glass networks, thus triggering the chain reaction Bi3+→Bi+→Bi0. Meanwhile, the interaction between Bi0 and defects in the glass gets strengthened, which brings about the improvement of the newly generated emission bands at ∼930 and ∼1520nm. On the other hand, the excessive AlO4 bonding with the glass structure will form more Ge–O–Al bonds, resulting in local excess negative charges. In order to maintain charge balance, the low-valence Bi will be oxidized to a high-valence state, resulting in Bi0 being oxidized as Bi+ and Bi3+, which exhibits the ultrabroadband tunable performance of Bi NIR emission and results in the decline of NIR emission intensity. To get deeper insight into this, the visible PL spectra from Bi3+ upon 350 nm excitation can be leveraged to verify it, as shown in Fig. 5(f). The intensity of the characteristic blue light emission from Bi3+ decreases monotonically upon the addition of Al2O3, and then an anomalous phenomenon occurs when the content of Al2O3 exceeding 11%, namely, no red emission from Bi2+ is observed^[45,46], which shows exactly the opposite trend in comparison with that of the NIR emission intensity. The results show clearly that the interaction between the two mechanisms discussed above contributes to the enhancement and ultrabroadband tunability of the multiple Bi NIR emissions, which predominantly depends on the concentration of Al2O3.

In conclusion, we successfully designed and fabricated a novel Bi–N co-doped germanate photonic glass, with highly efficient NIR emission (EQE of ∼50%) and ultrabroadband tunable performance (850–1700 nm), via modulation of the short- to medium-range topological order of the glasses by regulating Al2O3 amounts. As revealed by optical and microstructure characterization, with the introduction of Al2O3, aluminum tetrahedra AlO4 increased at the expense of AlO5 and AlO6 polyhedra, which is favorable for the multiple Bi NIR emissions. Moreover, the parallel variation trend of bands at ∼930, 1520, and ∼1260nm (Bi0) implies that the newly generated NIR center is related to the Bi0 center in the nitridated germanate glasses. Meanwhile, due to the replacement of Al2O3 with GeO2, the formation of local excess negative charges promotes Bi0 to be oxidized as Bi+ and Bi3+, which exhibits the ultrabroadband tunable performance of the multiple Bi NIR emissions. Importantly, this work opens an avenue for enhancing and regulating the multiple Bi NIR emissions and helps us to understand multiple Bi NIR emission behaviors in multicomponent glass, which shows great potential application in ultrabroadband wavelength-tunable glass fiber amplifier in optical communication.