Broadband infrared LEDs based on europium-to-terbium charge transfer luminescence

Efficient broadband infrared (IR) light-emitting diodes (LEDs) are needed for emerging applications that exploit near-IR spectroscopy, ranging from hand-held electronics to medicine. Here we report broadband IR luminescence, cooperatively originating from Eu2+ and Tb3+ dopants in CaS. This peculiar emission overlaps with the red Eu2+ emission, ranges up to 1200 nm (full-width-at-half-maximum of 195 nm) and is efficiently excited with visible light. Experimental evidence for metal-to-metal charge transfer (MMCT) luminescence is collected, comprising data from luminescence spectroscopy, microscopy and X-ray spectroscopy. State-of-the-art multiconfigurational ab initio calculations attribute the IR emission to the radiative decay of a metastable MMCT state of a Eu2+-Tb3+ pair. The calculations explain why no MMCT emission is found in the similar compound SrS:Eu,Tb and are used to anticipate how to fine-tune the characteristics of the MMCT luminescence. Finally, a near-IR LED for versatile spectroscopic use is manufactured based on the MMCT emission.

X-ray spectroscopy X-ray spectroscopy was performed at the ID26 beamline at the European Synchrotron Radiation Facility (ESRF) 1 . High Energy Resolution Fluorescence Detected X-ray Absorption Near Edge Structure (HERFD-XANES) spectra were recorded on the Eu L 3 edge (6.9769 keV) and on the Tb L 3 edge (7.5140 keV). only a small bandwidth near the maximum of a characteristic X-ray emission line is integrated, yielding an appreciable improvement in energy resolution 2 . For this, an X-ray emission spectrometer based on Rowland geometry was used. The spectrometer employs an array of four spherically-bent crystal analyzers to select by means of the Bragg's law only the X-rays of the desired energy. Ge(333) crystals were aligned to the Eu L α1 line (5.846 keV), or Ge(330) crystals to the Tb Lα 1 line (6.2728 keV). The incident energy was selected with a Si(311) double crystal monochromator. Pressed pellets of CaS:Eu,Tb, diluted in boron nitride (BN) in a 1:10 mass ratio were used as sample. X-ray induced valence changes or degradation was excluded during the measurement.
Supplementary Figure 2 shows the resulting HERFD-XANES spectra for the CaS:Eu 0.03 Tb 0.03 sample, compared to the reference compounds EuS (EuII), EuF 3 (EuIII), TbF 3 (TbIII) and Tb 4 O 7 (TbIII and TbIV). From these measurements, it is clear that the no detectable amount of Eu 3+ is present in the CaS:Eu 0.03 Tb 0.03 sample. The same is true for Tb 4+ . This indicates that, within the experimental detection limits, the prepared CaS:Eu,Tb powders contain solely Eu 2+ and Tb 3+ .

Dopant distribution
Scanning electron microscopy (SEM), complemented with Energy-dispersive X-ray Spectroscopy is applied to verify the dopant distribution inside the prepared phosphors. For this, a Hitachi S-3400N scanning electron microscope, equipped with a Thermo Scientic Noran System 7 energy-dispersive X-ray detector is used.
Supplementary Figure 3 shows a SEM-EDX scan on a 125 × 90 µm 2 area of the CaS:1%Eu,1%Tb powder. EDX maps for Eu and Tb indicate that the dopants are well distributed in the powder, apart from some small µm sized grains with higher Eu or Tb concentrations that can be found here and there. Both dopants mix appreciably as shown in Supplementary Figure 3(d) where simultaneous detection leads to a yellow color in the map. Finally, the dopant concentration was locally quantied by Phi-Rho-Z (φ(ρz)) quantication 3 , and collected in histograms (see Supplementary Figure 3(e)) 4 . It is found that the doping concentrations amount to 0.014 ± 0.008 and 0.014 ± 0.009 for Eu and Tb, respectively. The relatively narrow distributions, quantied by the abovementioned standard deviations, demonstrate the overall

IR emission vs. precursors / charge compensation
To exclude that the observed broadband IR emission is a spurious eect, induced by the used precursors, several dierent syntheses were performed using dierent precursors for the lanthanides. Supplementary Figure 4 shows the emission spectra of dierent CaS:1%Eu,1%Tb powders. This illustrates that the IR emission is indeed always present, safely excluding that it is induced as a consequence of the used precursors.
Although the presence of the IR emission is not affected by the used precursor, small variations in the ratio of the red and IR emission intensities can however be seen, indicating that the precursors have some eect on the distributions and incorporation of the dopants, possibly due to dierent charge compensating mechanisms, the ux eect of some precursors, notably uorides 710 , or other hidden variables that are only of secondary importance.
To acquire more information about the role of charge compensators, especially for the trivalent Tb 3+ ion, sodium (Na + ) was selected as codopant. It was found that the addition of Na + strongly increases the IR:red ratio (see Supplementary Figure 4), however at the cost of a strongly decreased overall quantum eciency (see further). This could be expected as it is known that Na + will stabilize Eu 3+ as well 11 , which is undesired because Eu 2+ -Eu 3+ pairs will be generated in addition to Eu 2+ -Tb 3+ pairs. The former pairs will negatively impact the quantum eciency because of intervalence charge transfer (IVCT) quenching 12,13 . As a further consequence of the Na + addition, fewer isolated Eu 2+ centers will exist, explaining the reduced red component in the spectrum. It is hence concluded that the addition of Na + charge compensators negatively aects the luminescent properties of the Eu,Tb co-doped phosphor because it stabilizes Eu 3+ . It has however no direct impact on the emission Supplementary Figure 4. Photoluminescence emission spectra of Eu 2+ and Tb 3+ codoped CaS phosphors, prepared with a solid state reaction, using dierent precursors for Eu and Tb.
spectrum of the IR-emitting Eu 2+ -Tb 3+ centers. As the broadband IR luminescence is the main topic of this article, the role of precursors or intrinisic or extrinsic charge compensation will not be elaborated further.

Additional Photoluminescence spectra
High doping concentration Supplementary Figure 5 displays the emission spectra for CaS powders with a xed Eu concentration of 3% and an increasing Tb concentration. It is found that the IR:red ratio is maximal for CaS:3%Eu,3%Tb and decreases again for higher Tb concentrations. This result conrms the SEM-CL mapping ( Figure 4) where it was found that the IR:red ratio is limited to the one for CaS:3%Eu,3%Tb. For higher doping concentrations, it is likely that defect clusters are formed containing multiple lanthanide dopants in a small region of the crystal, probably in addition to other defects. These complex structures strongly aect the local crystal structure, as well as the associated excited state landscape. It is hence not surprising that the radiative MMCT decay channel gets quenched for these high concentrations.
It should also be noted that the red emission is always found in addition to the IR emission. This equilibrium between red and IR emissions implies that the IR emission is presumably not (only) the result of an energy transfer from Eu 2+ to another emitting center because it would then be expected that the Eu 2+ emission would vanish completely if the concentrations are stretched suciently 14,15 .
Supplementary Figure 5. Photoluminescence emission spectra of Eu 2+ and Tb 3+ codoped CaS (top) and SrS (bottom) phosphors. The spectra were collected upon 285 nm excitation. All spectra were measured at room temperature.
Low temperature Supplementary Figure 6 displays the photoluminescence spectra of CaS:1%Eu,1%Tb measured at 10 K. Overall, the same spectral features and assignments apply as at room temperature (RT, see main text). In contrast to the RT spectra, here some phonon ne structure can be resolved. The zero-phonon line is indicated by the black arrow.
Supplementary Figure 6. Photoluminescence emission and excitation spectra of CaS:1%Eu,1%Tb, measured at 10 K. The emission spectrum was measured upon 490 nm excitation, the excitation spectra were measured for 670 nm (solid line) and 850 nm (dashed line) emission.

Diuse reectance spectra
The diuse reectance spectra of several CaS:Eu,Tb powders, shown in Supplementary Figure 7, were obtained with a Perkin Elmer Lambda 1050 UV-Vis-NIR spectrophotometer, equipped with an integrating sphere. The sample was kept in a powder sample holder and covered by a quartz window during the measurement.
The absorption band in the range 650 nm to 450 nm is due to the numerous Eu 2+ 4f 7 → 4f 6 5dt 2g transitions. Below 400 nm, fundamental absorption of the CaS host is visible, overlapping with the broad Eu-to-Tb MMCT absorption band. For more information on spectral assignments, see main text and Ref. 16.
As evidenced from the diuse reectance spectra, the prepared powders suer from signicant greying, showing reectance values of 35-40% in the long wavelength region where no Eu 2+ related absorptions are expected. Furthermore, the Eu 2+ 4f 7 − 4f 6 5dt 1 2g absorption strength in the 500 nm region does not straightforwardly scale with the Eu 2+ concentration. In part, this can be explained by the dierent greying for the dierent powders. Furthermore, the crystallization, dopant incorporation and hence absorption strength of inorganic phosphors, notably suldes 7,8 , is strongly aected by uxing eects during the solid state synthesis, caused by, among others, uoride precursors for the lanthanide dopants 9,10 . These observations indicate that there is margin for improving the eciency of the prepared sulde phosphors, possibly by carefully selecting uxing agents. Photoluminescence quantum eciency Photoluminescence quantum eciencies (QE) were measured inside an integrating sphere (152 mm, spectralon coated), using an LED (λ max = 470 nm) for excitation. The reection of the LED and the red and IR luminescence were collected by an optical ber, analyzed by an Acton SP2300 monochromator and recorded by a ProEM 1600 EMCCD camera (both Princeton Instruments). The detection setup was properly calibrated for its spectral response 17 .
Supplementary Figure 8 shows the resulting internal QE's, i.e. the ratio of the number of emitted (red+IR) photons to the number of absorbed (blue) photons as a function of the Eu concentration:

Luminescence decay
Luminescence decay proles were measured using a pulsed Optical Parametric Oscillator (OPO) laser (Ekspla NT342B), set at 470 nm, as excitation source in combination with an Andor intensied CCD. The resulting proles for the red and IR emission of CaS:Eu 0.01 Tb 0.01 are shown in Supplementary Figure 9. These proles were t using the sum of three exponentials, The obtained parameters are given in Supplementary Table 1. The short decay component, τ 1 , of the red Eu 2+ 4f 6 5d 1 → 4f 7 and IR MMCT emissions are of the same order of magnitude, around 500 ns. This corresponds to the expected radiative decay rate of Eu 2+ in CaS, which is relatively fast for Eu 2+ due to the high refractive index of CaS 18 .
Furthermore, similar for both emission bands, τ 2 and τ 3 correspond to delayed luminescence with characteristic times of the order of 4 and 35 µs, respectively. This indicates that presumably some trapping and detrapping is involved. This observation is in correspondence with the TL peak that shows up around and above room temperature in the thermal quenching measurement (see Figure  5). The trapping/detrapping kinetics are not further investigated here as they fall out of the scope of the MMCT luminescence that we report here.  In order to construct the diabatic congurational coordinate diagram ( Figure 5), the congurational coordinate diagrams along the fully symmetric breathing mode for the isolated Eu 2+ , Eu 3+ , Tb 3+ and Tb 2+ defects are needed. This is calculated with the ab initio multicongurational SA-RASSCF/MS-RASPT2/RASSI-SO approach, as described in the Methods section and in more detail in Ref. 16 . The detailed analysis of the electronic structure of Eu 2+ and Eu 3+ doped CaS and SrS can be found in Ref. 16 and its supplementary info. An analogous analysis for Tb 3+ and Tb 2+ doped CaS and SrS can be found here.
The results of the spin-free (SA-RASSCF/MS-RASPT2) calculations can be found in Supplementary Tables 2-3 and in Supplementary Figures 10-13. The results after spin-orbit coupling was included at RASSI-SO level and can be found in Supplementary Tables 6-7 and in Supplementary Figures 14-15.
The excited state landscape of the Tb 3+ doped suldes corresponds to the well-known structure; the 7 F J (J = 6 − 0) multiplet covers the lowest 6000 cm −1 after which a gap is found up to the 5 D 4 levels around 20000 cm −1 . Another small gap of 5000 cm −1 is then found before the 5 D 3 level, the second emitting level of isolated Tb 3+ ions. Above this level, a very dense set of spin-quintets is found. In our calculations, the levels originating from the 5 L and 5 G terms are calculated in addition to the 7 F and 5 D terms.
In contrast to Tb 3+ , divalent terbium, Tb 2+ , is a more exotic ion, which is likely not stable in CaS nor SrS. However, in order to construct the MMCT excited states of a Eu 2+ -Tb 3+ pair, the computation of the ground and excited states of Tb 2+ in CaS and SrS is needed. Our results show that the 4f 8 5d 1 conguration forms the ground state for the Tb 2+ ion in CaS. A similar picture is found for SrS at SA-RASSCF level (see Supplementary  Fig. 11), while the 4f 9 states undergo a signicant shift with respect to the 4f 8 5d 1 states upon including the nonstatic correlation (MS-RASPT2, see Supplementary Figure 13), leading to a near-degeneracy for the lowest 4f 9 and the lowest 4f 8 5d 1 levels. Even though it is expected that the 4f 8 5d 1 − 4f 9 energy dierence decreases when going from CaS to SrS from extrapolating spectroscopic information from Eu 2+ to other divalent lanthanides such as Tb 2+ based on empirical models 19,20 , the large decrease of the 4f 8 5d 1 − 4f 9 gap at MS-RASPT2 level is surprising. In any case, the value of the 4f 8 5d 1 −4f 9 gap does not aect any conclusions concerning the MMCT emission reported here, as this is determined predominantly by the Tb 3+ electron anity, breathing mode vibrational frequencies and lanthanide-ligand bond lengths (see main text).