Step 1: Calculation of the band gap of the host crystal. Step 2: Assessment of the position and intensity of transitions from the ¹S₀ state of Pr³⁺. Step 3: Determine the lowest energy 4f5d state for the longwave boundary of far-UVC emission. Step 4: Analysis of the situation in which the transitions from the 4f5d state of Pr³⁺ dominate the emissions as well as the shortwave boundary of far-UVC emission.

Next, we pay attention to the ¹S₀ state of Pr³⁺, which may generate far-UVC emissions through the ¹S₀ → ³H₄ (≈215 nm) and ¹S₀ → ³H₅ (≈225 nm) transitions (Fig. 1, step 2). It should be noted, however, that the emission intensities of these intra-configurational transitions are very weak due to their parity-forbidden nature. Another downside is that the total transition branching ratios of ¹S₀ → ³H_4,5 are usually less than 1%. More than 60% ~ 80% of the excited ¹S₀ state terminates in the lower-lying ¹I₆ state, peaking at ≈400 nm^49,50. This explains why the ¹S₀ → ³H₄ (≈215 nm) and ¹S₀ → ³H₅ (≈225 nm) transitions are hardly observed in most reports. Even if they are observed under optimal conditions, the emission intensity is too weak^51,52.

The above discussions suggest that the ¹S₀ state is unfavorable to far-UVC emissions. Therefore, it is more advisable to expect far-UVC emissions from the 4f5d state. The longwave boundary of far-UVC is at 230 nm, corresponding to ≈43478 cm⁻¹. Accordingly, the lower limit of the 4f5d state should be above 43478 cm⁻¹ (Fig. 1, step 3). Under this premise constraint, the energy position of the 4f5d state should be as close as possible to the center wavelength (≈46500 cm⁻¹) of far-UVC, ensuring that most emissions fall within the far-UVC range.

Notably, an over-high energy of the 4f5d should also be avoided. Otherwise, the excited electrons at the 4f5d state will be non-radiatively relaxed to the lower ¹S₀ state, followed by the emissions from the ¹S₀ state (Fig. 1, step 4). In SrAl₁₂O₁₉:Pr³⁺, for example, the lowest 4f5d state and ¹S₀ state were ≈47600 cm⁻¹ and ≈46500 cm⁻¹, respectively, with an energy difference of 1100 cm^−1 53. In this compound, only weak transitions from the ¹S₀ state were observed over the far-UVC region at 220 nm. In contrast, the lowest 4f5d state in LaB₃O₆:Pr³⁺ is higher than the ¹S₀ state by ≈900 cm⁻¹, and the 5d → 4f transitions dominated the emission spectrum in the far-UVC range due to the Boltzmann distribution between the ¹S₀ state and the upper 4f5d state⁵⁴. Combining the above two situations, we suggest that the position of the 4f5d state should not be higher than that of the ¹S₀ state by 1000 cm⁻¹.

Summarizing all the information above, we can draw the following semi-empirical conclusion: the 4f5d state of Pr³⁺ should be in the range of 43478 to 47500 cm⁻¹. Considering that the position of the ¹S₀ state may slightly change with the alteration of the host (<500 cm⁻¹), the upper energy limit of the 4f5d state of Pr³⁺ can be appropriately relaxed to ≈48000 cm⁻¹. Additionally, it is important to note that the host material where Pr³⁺ ions are embedded needs to have good far-UVC transmission. This clue points to fluorides that typically possess a wide band gap and good ultraviolet transmittance. Based on previous reports, the 4f → 5d transition of Pr³⁺ in SrF₂ had an absorption peak at ≈46600 cm^−1 55. Such energy value well satisfies the above guidelines, thus making it possible to achieve far-UVC ML. Therefore, SrF₂:Pr³⁺ was chosen as a representative for examination in the subsequent sections.

Material synthesis and characterization

A series of SrF₂:x%Pr³⁺ (x = 0.1, 0.2, 0.5, 0.7, and 1) phosphors were synthesized by a high-temperature solid-state technique. The X-ray diffraction (XRD) patterns of these samples, mainly comprising the diffraction peaks of (1 1 1), (2 0 0), (2 2 0), (3 1 1), (4 0 0), and (3 3 1) planes, matched well with cubic phase SrF₂ crystals (PDF#88-2294) (Fig. 2a). It suggested that SrF₂:Pr³⁺ crystalized in the single center-symmetric structure with the Fm-3m space group (Fig. 2b). In this structure, Sr ions occupied an octahedral coordination environment, where each Sr ion was surrounded by eight F ions. Meanwhile, Pr³⁺ ions entered the host lattice by substituting the Sr²⁺ sites (Supplementary Table 2), which was confirmed by the Rietveld refinement analysis (Fig. 2c and Supplementary Table 3).

Fig. 2: Characterization of SrF₂:Pr³⁺.

a X-ray diffraction patterns of SrF₂:x%Pr³⁺ (x = 0.1, 0.2, 0.5, 0.7, and 1). b Crystal structure of SrF₂:Pr³⁺. c Rietveld structure refinements of SrF₂:0.5%Pr³⁺. d Scanning electron microscopy and elemental mapping images SrF₂:0.5%Pr³⁺ (scale bar: 50 μm). e X-ray photoelectron spectroscopy profile of SrF₂:0.5%Pr³⁺. f Selected area electron diffraction image of SrF₂:0.5%Pr³⁺ (scale bar: 10 nm⁻¹). g High-resolution transmission electron microscopy image of SrF₂:0.5%Pr³⁺ (scale bar: 2 nm). Source data are provided as a Source Data file.

The scanning electron microscopy (SEM) image revealed an irregular morphology of the as-synthesized SrF₂:Pr³⁺ in the micrometer length scale, with Sr, F and Pr elements distributed uniformly at the single particle level (Fig. 2d and Supplementary Fig. 2). The energy dispersive spectroscopy (EDS) results proved that the element ratios of Sr, F, and Pr were as expected (Supplementary Figs. 3–7). The X-ray photoelectron spectroscopy (XPS) results further validated the presence of constituent elements (Fig. 2e). The XPS peaks centered approximately at 966 and 946 eV were not observed, indicating the absence of tetravalent Pr. The peaks at the binding energy of 931.9 and 935.3 eV were ascribed to trivalent Pr. We also measured the Raman shift of SrF₂:0.5%Pr³⁺ (Supplementary Fig. 8). A prominent band appeared at approximately 285 cm^-1, attributed to the first-order F_2g mode of SrF₂⁵⁶. The selected area electron diffraction (SAED) exhibited sharp diffraction spots with a periodic arrangement, demonstrating good crystallinity (Fig. 2f). The high-resolution transmission electron microscopy (HR-TEM) images further disclosed the good crystallinity of our samples, which show well-defined (2 0 0) and (1 1 1) lattice fringes with observed d-spacings of 0.290 and 0.335 nm, respectively (Fig. 2g and Supplementary Fig. 9).

Multi-stimulated far-UVC luminescence

We designed a set of experimental procedures to systematically investigate the multi-stimulated luminescence, as schematically shown in Fig. 3a. Figure 3b depicts the emission spectra of SrF₂:Pr³⁺ under different conditions. As can be seen, the sample can be directly excited by mechanical action through the self-recoverable ML mechanism (Stage I). After being exposed to X-ray irradiation, bright persistent luminescence (PersL) was detected (Stage II), indicating the occurrence of charge carrier trapping. Following PersL, we further detected trap-controlled ML by application of force (Stage III). The trap-controlled ML was about 90-fold stronger than the self-recoverable ML, and the intensity fell back to the original level by thermal cleaning of the trap states (Stage IV). Notably, all the spectra are dominated by the characteristic emission peaks of Pr³⁺ ions, suggesting that Pr³⁺ ions consistently acted as luminescent centers during the whole process (Fig. 3b). Specifically, the visible emission bands are attributed to the ³P₀ → ³H₄ (488 nm), ³P₁ → ³H₅ (524 nm), ³P₀ → ³H₆ (606 nm) /¹D₂ → ³H₄ (606 nm) and ³P₀ → ³F₂ (640 nm) transitions of Pr³⁺. The UVC emission bands are assigned as the 4f5d → ³H₄ (222 nm), 4f5d → ³H₅ (233 nm) and 4f5d → ³H₆ (259 nm) transitions of Pr³⁺. Note that the broad band peaking at 300 nm came from the self-trapped exciton of the host (Supplementary Fig. 10)^57,58,59. The presence of the ML line at 222 nm proves the reliability of our guidelines for designing far-UVC luminescence.、

Fig. 3: Optical properties of SrF₂:Pr³⁺.

a Schematic of experimental procedures for mechanoluminescence (ML) and persistent luminescence (PersL) measurement. b Comparison of spectra of SrF₂:0.5%Pr³⁺ under different conditions. Stage I: obtained before X-ray irradiation. Stage II: obtained immediately after stoppage of 5-min X-ray irradiation. Stage III: obtained upon force stimulation at 5 N after PersL. Stage IV: obtained after thermal bleach. c Repeatability results of trap-controlled ML. d PersL and ML intensity of SrF₂:Pr³⁺ as a function of Pr³⁺ content. e Photographs of visible and UVC PersL of SrF₂:0.5%Pr³⁺, showing progressive changes over time following a 5-min X-ray irradiation. f Photographs of visible and UVC ML of SrF₂:0.5%Pr³⁺@polydimethylsiloxane films stretched under different magnitudes of force. g ML spectra of SrF₂ singly doped with different lanthanides (0.5%mol). Source data are provided as a Source Data file.

We next examined the repeatability of ML. Both the self-recoverable and trap-controlled ML decreased after several cycles of continuous operation (Fig. 3c and Supplementary Fig. 11). This phenomenon was also previously observed by several groups^{13,15,22,25,43}. By checking the XRD results of SrF₂:0.5%Pr³⁺ after grinding, we found no noticeable change in the crystal structure (Supplementary Fig. 12). The comparison of SEM images of SrF₂:Pr³⁺ before and after grinding suggested that the samples became progressively finer (Supplementary Fig. 13). As the surface area and morphology of samples had been modified, it was reasonable to assume that the ML property, including both self-recoverable and trap-controlled ML, were affected to some extent⁶⁰.

We optimized the content of Pr³⁺ in SrF₂ with emphasis on the 4f5d → ³H_4,5,6 and ³P₀ → ³H₄ transitions. The transitions from the 4f5d and ³P₀ states exhibit opposite trends with doping concentration (Fig. 3d and Supplementary Fig. 14). Such discrepancy is explained by the possible cross-relaxation-induced concentration quenching mechanism (Supplementary Fig. 15). With gradually increasing the content of Pr³⁺, the 4f5d → ³H_4,5,6 transition showed a monotonic decrease trend, as the [4f5d → ³P_0,1]/[¹D₂ ← ³H₄] cross-relaxation process was harmful to the 4f5d state. By contrast, the ³P₀ → ³H₄ transition increased first over the range of 0.1 < x < 0.5 because the [4f5d → ³P_0,1]/[¹D₂ ← ³H₄] cross-relaxation would populate the ³P₀ state, together with the increment of luminescent centers. When the content of Pr³⁺ exceeded 0.5%, the ³P₀ → ³H₄ transition experienced a continuous decline, attributed to the predominance of [³P_0,1 → ¹G₄]/[¹G₄ ← ³H₄] cross-relaxation⁵⁹.

Under a doping concentration of 0.5%, the SrF₂:Pr³⁺ exhibited pronounced and balanced visible and UVC PersL after the stoppage of X-ray irradiation (Fig. 3e and Supplementary Fig. 16). It was easy to distinguish the UVC PersL signal from the background noise after several hours. Even after one day, the visible PersL was observed clearly. Over the whole PersL process, the positions of emission spectra remained unchanged (Supplementary Fig. 17). After the release of room temperature PersL, SrF₂:Pr³⁺ continued to emit in the visible and UVC regions by mechanical stimulation (Fig. 3f and Supplementary Movie 1). With increasing the force, both the UVC and visible ML were enhanced. Notably, the ML intensity of ³P₀ → ³H₄ (488 nm) and 4f5d → ³H_4,5,6 transitions displayed different responses to the external force (Supplementary Figs. 18 and 19), pushing us to establish a ratiometric method for stress sensing^61,62. The ML intensity ratio between the ³P₀ → ³H₄ and 4f5d → ³H_4,5,6 transitions presented a monotonic decrease by raising the force, with the relative sensitivity varying from 3% N⁻¹ to 7% N⁻¹ over the 5–25 N range (Supplementary Fig. 20).

We also prepared a series of SrF₂ samples doped with various lanthanides (Ce³⁺, Nd³⁺, Sm³⁺, Eu^2+,3+, Gd³⁺, Tb³⁺, Dy³⁺, Ho³⁺, Er³⁺, Tm³⁺, and Yb³⁺) to expand the spectral range. The XRD and XPS results proved the successful preparation of target materials (Supplementary Figs. 21 and 22). All these samples emitted self-recoverable ML (Supplementary Figs. 23 and 24). After being pre-charged by X-ray, these samples emitted tunable PersL due to characteristic transitions in relevant lanthanide dopants (Supplementary Figs. 25–27). After the PersL faded out in one and a half days, all samples presented trap-controlled ML, nearly covering the whole spectral range from 200 to 1700 nm (Fig. 3g and Supplementary Figs. 28–31).

Mechanistic investigation of multi-stimulated emission

The self-recoverable ML was previously observed in some asymmetric materials like CaZnOS:Mn²⁺ and ZnS:Mn²⁺ due to piezoelectricity^28,63. Accordingly, we conducted the piezoresponse force microscopy (PFM) tests to check the piezoelectric effect. Pure SrF₂ did not show any piezoelectric properties due to the centrosymmetric nature (Supplementary Fig. 32). Nevertheless, piezoelectric hysteresis and butterfly loops appeared for SrF₂:Pr³⁺ during the PFM measurements, indicating that the introduction of Pr³⁺ disrupted the local symmetry of SrF₂ and thereby endowed SrF₂:Pr³⁺ with local piezoelectric properties (Supplementary Fig. 33). The PFM results verify the arise of piezoelectric potential in the ML process.

It should be noted that, in addition to the piezoelectric effect, triboelectrification has recently been considered as another key mechanism for ML. A notable feature of triboelectrification-induced ML is that the ML intensity exhibits a strong dependence on the triboelectricity of the embedding substrates (Supplementary Table 4). Accordingly, we assessed the ML intensity of SrF₂:Pr³⁺ particles embedded in a series of polymer substrates (Fig. 4a). The results show that SrF₂:Pr³⁺@polydimethylsiloxane (PDMS) presented the strongest ML as PDMS has the best negative triboelectricity than ER (epoxy resin), PU (polyurethane) and SG (silicon gel), consistent with the previously reported results (Supplementary Table 4). The above observations indicate that the triboelectric effect is also responsible for the ML from SrF₂:Pr³⁺. The conclusion was further corroborated by green-emitting SrF₂:Tb³⁺ (Fig. 4a).

Fig. 4: Mechanistic study of SrF₂:Pr³⁺.

a Photographs of SrF₂:0.5%Pr³⁺ powder and SrF₂:0.5%Pr³⁺ embedded in different polymers, including epoxy resin (ER), polydimethylsiloxane (PDMS), silicon gel (SG) and polyurethane (PU), under mechanical action. b Persistent luminescence (PersL) decay curve of SrF₂:0.5%Pr³⁺ after stoppage of 5-min X-ray irradiation. Inset: enlarged PersL signal upon stimulation of external force. c Photographs of SrF₂:0.5%Pr³⁺ upon single- and multiple-sliding during the PersL process. d Thermoluminescence (TL) curves of SrF₂:0.5%Pr³⁺ before and after grinding for 20 min. e Mechanoluminescence (ML) spectra of SrF₂:0.5%Pr³⁺ after different X-ray exposure dose. f Normalized TL and ML intensities of SrF₂:0.5%Pr³⁺ as a function of X-ray irradiation dose. Source data are provided as a Source Data file.

To gain insight into the trap-controlled ML mechanism after X-ray charging, we measured the PL (λ_ex = 440 nm, λ_em = 488 nm) and ML decay curves of SrF₂:0.5%Pr³⁺ (Supplementary Fig. 34). The results reveal that the duration time of ML is three orders of magnitude larger than that of PL. The observation supports the trap-controlled nature of the ML behavior, which involves the release and migration of charge carriers, thus taking a significantly longer time than PL (Supplementary Fig. 35). In line with the mechanically induced release of trap carriers, a transient enhancement of emission was detected at the PersL stage upon stimulation of external force (Fig. 4b). After repeated force stimulation, a notable dimming in PersL occurred at the location where the force was applied along with a significant decrease in the TL intensity (Figs. 4c, d). At the same time, the ML intensity constantly decreased. These results indicated that the PersL and ML originated from the same trapped charge carriers that can be released by both heat and force.

To shed more light on the multi-stimulated luminescence, the trapped charge carriers were adjusted by regulating the X-ray dose. With the increase of charging time, the ML intensified gradually (Fig. 4e). The observation is as anticipated due to the increased population in trap states, as confirmed by TL measurement (Supplementary Fig. 36). After performing the integral calculation, we clearly demonstrated the positive correlation of ML and TL intensities (Fig. 4f). As PersL is also closely related to TL, it is not difficult to understand that the stronger the PersL, the stronger the ML typically is (Supplementary Figs. 37 and 38). These observations provided additional evidence that the PersL and ML originated from the same trapped charge carriers. According to the TL spectra, we determined continuously distributed traps (Supplementary Figs. 39 and 40). The XRD pattern of SrF₂:Pr³⁺ was essentially unchanged upon exposure to X-ray (Supplementary Fig. 41), suggesting that the traps of SrF₂:Pr³⁺ were not due to the macro-structural change. We thus ascribe the trap states to intrinsic defects resulting from aliovalent doping. Our theoretical calculations (Supplementary Fig. 42 and Table 5 and Note 1) suggest that the substitution of two Pr³⁺ ions for two Sr²⁺ sites most likely occurred, resulting in one consequent Sr²⁺ vacancy. Moreover, fluorine vacancies and interstitials were expected to be generated in the fluorides under X-ray irradiation according to previous reports. Therefore, there were abundant defects in SrF₂:Pr³⁺ to trap charge carriers^64,65,66.

Far-UVC emission for solar-blind imaging

We designed a quick response (QR) code made of SrF₂:0.5%Pr³⁺ and put it on the surface of high-valued goods. During X-ray inspections, the QR code was charged. After inspections, the QR code emitted both UVC and visible PersL that could be recorded separately by solar-blind and common cameras (Fig. 5a), providing a dual-layer anti-counterfeiting verification. Furthermore, even if the QR code’s PersL diminished over time, the QR code could still be captured by a typical smartphone under mechanical stimulation (Supplementary Fig. 43). By incorporating such a unique QR code on product packaging, counterfeiting and illegal transportation can be prevented to a great extent.

Fig. 5: Optical imaging through far-UVC emission of SrF₂:Pr³⁺.

a Customs inspection. I and II show the photographs of SrF₂:Pr³⁺ coated on the surface of goods after X-ray inspections, captured by collecting the UVC and visible emissions, respectively. b Solar-blind marking. The right panel presents the images of different complex scenarios that contain the SrF₂:Pr³⁺ tablets, captured by using the visible or UVC+visible channel of a solar blind camera. c Structural health monitoring. The right panel shows the photographs of SrF₂:Pr³⁺@polydimethylsiloxane and ZnS:Cu²⁺@polydimethylsiloxane films, as well as SrF₂:Pr³⁺@photopolymer resin and ZnS:Cu²⁺@photopolymer resin under force stimulation by using the visible or UVC+visible channel of s solar blind camera.

Another important application for the UVC emission of SrF₂:Pr³⁺ is high contrast marking. We prepared a series of SrF₂:Pr³⁺ tablets with a diameter of 13 mm (Supplementary Fig. 44). As the samples are visually small and white, they are almost imperceptible to the naked eyes when placed in complex backgrounds such as in bamboo forests, holly bushes or snow. However, the positions of SrF₂:Pr³⁺ tablets can be clearly identified by using a solar blind camera to capture the UVC PersL signal (Fig. 5b). We further constructed an array by mixing SrF₂:Pr³⁺@PDMS with SrF₂@PDMS (Supplementary Fig. 45). After X-ray irradiation, the pattern ‘Pr’ was captured using the combined UVC and visible channel of solar blind camera. This method enables advanced marking techniques.

Health monitoring of engineering structures such as bridges and buildings is important for preventing structural failures that can lead to major disasters and economic losses. SrF₂:Pr³⁺ holds important value for remote and real-time structural monitoring. We simulated two scenarios to illustrate the superiority of UVC photons. One involved directly pulling the SrF₂:Pr³⁺@PDMS film coated on the surface of a real bridge (Fig. 5c-I), while the other involved compressing the SrF₂:Pr³⁺@PDMS film coated on the surface of a 3D-printed bridge model (Fig. 5c-II). When external force was applied to the SrF₂:Pr³⁺@PDMS film, the UVC emission was clearly captured in high contrast even in a bright environment. In this way, structural damage to the building, such as cracks, could be promptly collected and warned to prevent accidents and ensure safety. As a control experiment, we also employed the well-established ZnS:Cu²⁺ with visible emission to replace SrF₂:Pr³⁺ for the above demonstrations. Expectedly, no green ML could be observed due to the strong interference of background light, including intense outdoor sunlight and indoor lighting sources, although super bright ML was observed in dark environments (Supplementary Fig. 46). Considering the remarkable anti-interference capability of UVC photons, our samples are expected to unlock more applications such as emergency rescue (Supplementary Fig. 47). As the multi-stimulated luminescence can be substantially tuned across the full spectrum by doping different lanthanides, these materials also hold significant promise for information security and other advanced applications (Supplementary Figs. 48–51 and Note 2).