Spectral variability of the uranyl silicates uranophane-α and uranophane-β: polymorphism and luminescence

The luminescence of the uranyl cation UO22+ depends on the local crystalline environment and is sensitive to structural influences. Steady-state photoluminescence emission spectra of the related uranyl silicates uranophane-α, uranophane-β, sklodowskite and haiweeite from various locations are presented and discussed in the light of structure–property relation. The four mineral species were chosen for their close relationships: uranophane-α and uranophane-β are polymorphs and share the underlaying topology with sklodowskite. Haiweeite, with different topology, shares the composing elements Ca, U, Si, O with uranophane, while in sklodowskite Mg replaces Ca. All species show some variability in their spectra, parameterized as a variation of the centroid wavelength. Those variations are linked to defects and structural disorder, relevant in studies of uranyl speciation and migration. We present empiric spectra of the four mineral species with the least influence of structural disorder. As an unexpected feature, a prominent—partly dominating—double peak structure occurs in the case of uranophane-α only, while it is absent in the spectra of the other species. Considering a model of luminescent transitions in the uranyl ion in more detail, this observation is discussed in the light of the polymorphism of uranophane. We show evidence that variable amounts of uranophane-β phase embedded in uranophane-α are possibly at the origin of this spectral signature. Growth of those uranophane-β clusters might be induced by defects in the uranophane-α lattice and further promoted by the polymorphism of uranophane.


Introduction
In many crystalline environments, the uranyl cation UO 2 2+ exhibits a characteristic greenish to yellow photoluminescence with spectroscopic properties mostly related to the molecular character of uranyl luminescence (Frondel 1958;Blasse 1987;Gorobets 2002). The characteristic features of excitation and emission spectra depend on the local environment of the uranyl ion defined by ligands or the crystalline matrix (Blasse 1987;Görller-Walrand et al. 2004;Pierloot and van Besien 2005;Drobot 2015;Višňák and Sobek 2016;Haubitz et al. 2018). With additional ligands in solution, the uranyl ion UO 2 2+ can form more complex ions like e.g. UO 2 OH + , (UO 2 ) 3 (OH) 7− or UO 2 (CO 3 ) 3 4− , depending on environmental conditions (e.g. temperature, pH) and available ligands (Moulin et al. 1998;Mühr-Ebert et al. 2019;Romanchuck et al. 2020). This speciation connects chemically driven alteration processes to environmentally relevant migration processes in specific geological settings. Soluble uranyl ions migrate in the environment and might interfere with human activities, e.g. contaminating water resources or being enriched in the food chain. For remediation of possibly contaminated areas, as well as for environmental monitoring in the context of spent nuclear fuel storage, insight in speciation processes is a key issue. Luminescence spectroscopy proved to be a valuable tool for characterizing uranyl speciation (Meinrath 1997;Moulin et al. 1998;Collins et al. 2011;Drobot et al. 2015). Further methodology has been proposed for environmental monitoring in the context of waste management and remediation of possibly contaminated sites (e. g. DeNeufville et al. 1981;Meinrath et al. 2003;Nelson 2009;Arnold et al. 2011;Frankland et al. 2021 The migration of uranyl ions in a liquid environment is among others influenced by sorption processes and the initiation of crystal growth. Mobile uranium species are transformed into (meta-) stable forms, e.g. as mineral coatings or crystalline components of soil. Local growth of crystals removes uranyl ions from solution, thus limiting migration (Wronkiewicz et al. 1996). However, dust formed subsequently by mechanical erosion is again subject to transport by wind or, in technical systems, by airflow and ventilation. Similar to the case in solution, the luminescence is characteristic for the mineral species involved. Spectra of uranium-bearing minerals have been explored as additional information to identify mineral species (Haberlandt et al. 1950;Gorobets 2002) and as well to monitor environmental issues (Cunnane et al. 1993;Geipel et al. 2000;Wang et al. 2005;Massuyeau et al. 2017). However, with respect to the photoluminescence spectrum of uranophane-α, there are some uncertainties in the literature regarding the assignment of peak positions, as briefly discussed in "Structural elements of uranophane-α, uranophane-β, sklodowskite and haiweeite".
At this point, it is of interest to examine spectral variability and its relation to structure as well as structural disorder. The characteristic photoluminescence spectrum of uranyl in a given mineral species is modified by the structural disorder present in the real crystal. Defects (e.g. substitutions, vacancies or structural distortions) alter local fields and symmetries and thus affect the luminescence of the emitting ions. The corresponding emission spectra are affected by inhomogeneous broadening and redshift (Skinner and Moerner 1996;Lenz 2015). Certain defects can act as luminescence centers themselves and contribute to the emission. The signature of such an additional spectroscopic species adds to the observed emission spectrum. In radioactive minerals, radiation damage due to self-irradiation is an additional source of structural disorder. The extended loss of order leads to a randomization of the ligand field, resulting in a broad and unstructured component in the emission spectra (Lenz 2015;Lenz and Nasdala 2015). Consequently, the spectra measured on an ensemble of crystals of one mineral species exhibit some variability that can be attributed to structural disorder and defects. Photoluminescence spectroscopy addresses different physical entities (ions or groups acting as luminescent centers) as compared to X-ray diffraction (XRD; electron densities) or Raman spectroscopy (polarizable chemical bonds and their vibrational modes). Photoluminescence is sensitive to local influences and allows trends to be detected even at low concentrations. The information obtained by photoluminescence spectroscopy complements that obtained by XRD or Raman. Insights into spectral variability are important for assessing the reliability of spectroscopic screening data, e.g., for environmental monitoring, where the identification of mineral species can be obscured by varying spectral signatures.
In addition, such analysis allows to address structure-property relationships and to better understand the stability of a particular mineral species as well as possible alteration processes.
Here, we investigate the spectral variability of the natural uranyl silicates uranophane-α, uranophane-β, sklodowskite, and haiweeite. Uranyl silicates are relevant in the context of uranium migration, as they represent common and rather stable alteration products (Wronkiewicz et al. 1996). These uranyl silicates form aggregates of similar appearance (see Fig. 1) and are closely related in chemistry and structure ( Table 1).
The main focus is directed towards the polymorphism of uranophane and its influence on spectroscopic properties, with the two other species providing a reference frame. Sklodowskite shares main structural elements with uranophane-α and is chemically closely related to uranophane with Mg replacing Ca. No polymorph is known for sklodowskite. Haiweeite, on the other hand, shares the constituting chemical elements with uranophane, but exhibits a distinctly different structure. The outline of the article is as follows: The structure of the uranyl silicates under focus, their luminescence mechanism and the interplay between structure and luminescence are briefly outlined. Photoluminescence spectra and spectral variability of (in total) more than 80 samples from about 30 different localities have been investigated. The spatial localization of spectral fluctuations is demonstrated by spatially resolved measurements on individual crystals. Based on these findings, we discuss evidence for enhanced spectral variability of uranophane-α as compared to its counterpart uranophane-β and to the other species, sklodowskite and haiweeite. We present empiric base spectra of the four mineral species with the least influence of structural disorder. Results are related to local structural variability and polymorphism, which possibly is at the origin of the observed spectral fluctuations.

Structure and luminescence
Structural elements of uranophane-α, uranophane-β, sklodowskite and haiweeite First, we consider structural elements relevant to spectral variability, for details we refer the reader to the given literature, as referenced below. Figure 2 shows relevant elements of the structures of the four uranyl silicates in focus.

Fig. 2
Similarities and differences in the structures of a uranophane-α (Barinova et al. 2001), b uranophane-β ( Barinova et al. 2003), c sklodowskite (Mokeeva 1964) and d haiweeite (Plášil et al. 2013 with the (UO 7 )-bipyramids connected by edges to form chains. These chains are interconnected by SiO 4 -tetrahedra ( Fig. 2a, b). To highlight the chain-tetrahedra motifs, only (UO 7 )-bipyramids (yellow) and SiO 4 -tetrahedra (blue) are shown in Fig. 2. The free corners of the SiO 4 -tetrahedra point out of the sheet formed by the chains (Burns 2005;Plášil 2018a). This arrangement allows for two possible orientations of tetrahedron corner points as marked in Fig. 2a, b: either out of the plane-denoted up (u), or into the plane-denoted down (d), following the description given in Plášil (2018a). Note, from the point of view of the topology, both orientations are equivalent (Burns 2005;Plášil 2018a). However, the distributions of up-and down-oriented tetrahedra are apparently only favored energetically to result in two polymorphs of uranophane: while in uranophane-α succession of SiO 4 -tetrahedra along a (UO 7 )-chain follows the simple motif ..ud.., the motif is doubled in the case of uranophane-β, giving ..uudd.. (Fig. 2a,b). Based on these characteristic motifs, the structural complexity of uranophane-β was shown to be twice that of uranophane-α (Plášil 2018b;Colmenero et al. 2019). Note that the two characteristic motifs lead to an important difference in symmetry. Focusing on one particular (UO 7 )-unit in uranophane-α, orientations of the SiO 4 -tetrahedra of the two directly adjacent (UO 7 )-units along the chain are identical (both up or both down), opposing the orientation of the SiO 4 -tetrahedron at the particular (UO 7 )-unit in focus (compare Fig. 2a). In the case of uranophane-β, the orientation of those two neighboring SiO 4 -tetrahedra is diametrically opposed (one up, the other down; Fig. 2b). Therefore, we expect differences in the spectra of the two uranophane polymorphs.
Haiweeite-spacegroup Pbcn-on the other hand does not belong to the uranophane topology, even though it shares the existence of (UO 7 )-chains linked to SiO 4 -tetrahedra (McBurney and Murdoch 1959;Burns 2001;Plášil et al. 2013). Yet, in haiweeite these structural elements form a more complex network, where the orientations of the SiO 4 -tetrahedra sharing edges with (UO 7 )-bipyramids remain conserved along the chain (Fig. 2d).

Raman spectroscopy
Raman spectra of the minerals under investigation are shown in Fig. 3 (reference data taken from the RRUFF project, rruff.info; Lafuente et al. 2015). We focus on a spectral range from approximately 700 cm −1 to 1100 cm −1 . Here, the most prominent and discriminating spectral features become visible. As Raman bands are associated with vibrational modes, closer analysis of the spectra yields valuable data for discussing phononic contributions to luminescence spectra.
The dominant peak around 790 cm −1 is associated with stretching vibrations ν 1 of the (linear) UO 2 2+ , while bands in the range of 930 cm −1 to 980 cm −1 are associated with stretching vibrations of the SiO 4 -tetrahedra (Frost et al. 2006a;2006b;2006c;2006d;Wall et al. 2010;Colmenero et al. 2018Colmenero et al. , 2019. As expected from structural similarities, the phonon energies of the UO 2 2+ vibration in the uranyl silicates under investigation are close to each other, because peak shifts are mainly induced by the equatorial ligand field of UO 2 2+ (Pierloot and van Besien 2005). The two polymorphs of uranophane show almost identical phonon energies. In the range 930 cm −1 to 980 cm −1 , Raman spectra of the different species show distinct differences associated to vibrations of the SiO 4 -tetrahedra. In line with their structural equivalence, sklodowskite and uranophane-α exhibit pronounced similarity at this band. In the case of uranophane-β, the band is shifted to lower energies, reflecting its deviating building motif ..uudd… Finally, the significantly different structure of haiweeite results in a distinct Raman signature in that range.

Photoluminescence spectroscopy
Spectroscopic data on uranophane-α, uranophane-β, sklodowskite and haiweeite have been reported under varying conditions on natural crystals as well as on synthetic Fig. 3 Raman spectra of the four minerals in the discriminating range of 700 cm −1 to 1100 cm −1 . Measured data (black) is shown together with reference data (red) taken from the RRUFF project (rruff.info): a uranophane-α, b uranophane-β, c sklodowskite and d haiweeite. The Raman peak around 800 cm −1 is associated with the linear stretching vibrations of (UO 2 ) 2+ and converts into the phononic energy difference of the luminescence spectra. Notably, this energy is the same for uranophane-α and uranophane-β. Bands around 950 cm −1 relate to vibrations of SiO 4 -tetrahedra and are distinctive for the four species material. Despite the broad range of literature in this field, only few data sets are freely accessible. Photoluminescence spectra were already investigated by Haberlandt and colleagues (1950) with further assignment to mineral species by Meixner (1965). Later data refer to spectra obtained as well on naturally grown (Cunnane et al. 1993;Gorobets 2002;Wang et al. 2005;Frankland et al. 2022) as on synthesized material (Lehmann et al. 2008;Kuta et al. 2013). Table 2 provides a compilation of available data together with our own values. We are aware that this listing is not complete, yet tendencies can be seen.
Collecting literature data, some uncertainties were encountered. Remarkably, the emission spectra of uranophane-α given in three reports show evidence of phononic bands consisting of two peaks (double peak structure), notably for natural crystals (Haberlandt et al. 1950;Frankland et al. 2022) as well as for synthetic material (Kuta et al. 2013) . Frankland et al. (2022) discuss this observation and assign it tentatively as part of the uranophane luminescence. Data presented by Wang et al. (2005) is rather assigned to uranophane-β: while only 'uranophane' is specified in the text, uranophane-β is mentioned in the legend of Fig. 2. Further observations by Frankland and colleagues (2022) support such assignment.
The emission spectrum of UO 2 2+ in a well-ordered crystal with only few defects consists of a clear band structure with a succession of (rather sharp) phononic peaks (Cunnane et al. 1993;Lehmann et al. 2008;Drobot 2015;Višňák and Sobek 2016;Haubitz et al. 2018). In the following, we outline the model described separately by Drobot (2015) as well as by Višňák and Sobek (2016). Figure 4 shows a simplified representation of the relevant electronic groundand excited state together with phononic levels (Višňák and   Sobek 2016). Optical relaxations occur from the bottom of the excited state to one of the phononic levels of the ground state, where the phonon energy ω phonon is equivalent to the (UO 2 ) 2+ stretching vibration as determined by Raman spectroscopy (Drobot 2015;Višňák and Sobek 2016;Colmenero et al. 2018). The transition without contribution of a phonon, i.e. from the bottom of the excited state to the bottom of ground state (denoted 0 ′ → 0 , following Višňák and Sobek 2016), defines the zero-phonon line (ZPL), which is the characteristic transition of this system (Drobot 2015;Višňák and Sobek 2016;Haubitz et al. 2018). Thus, starting from the zero-phonon line, the further peaks of the spectrum are determined by multiples of the phonon energy. Note, the more phonons are necessary to balance the transition energy, the less likely this transition will be, so emitted intensity will decrease with phonon number (Višňák and Sobek 2016). Considering the phononic levels of the excited electronic state, (i) their energy separation ω' phonon is equidistant, but will differ from ω phonon of the ground state (due to the fact that the excited state is more extended) and (ii) the occupation follows a Boltzmann distribution (Višňák and Sobek 2016). Thus, there is a small possibility, that hot-band transitions occur, starting from the first phononic level of the excited electronic state (1') to one of the phononic levels of the fundamental electronic state (see Fig. 4;Drobot 2015;Višňák and Sobek 2016). According to the Boltzmann distribution, we can assume transitions from the first phononic level only, which in good approximation will not carry more than a few percent of the intensity (Višňák and Sobek 2016;Haubitz et al. 2018). For the same reason as for the main phononic peak series, we must assume that the hot-band contributions decrease significantly with the number of phonons involved. Thus, the transition from the first phononic level of the excited electronic state to the bottom of the fundamental electronic state, 1 ′ → 0 , is the dominant hot-band. It occurs at higher energy as compared to the zero-phonon line ( 0 ′ → 0 ), but with much lower intensity (Višňák and Sobek 2016;Haubitz et al. 2018). In first approximation, a spectrum composed according to these considerations consists of the sequence of main phononic bands with intensities decreasing with decreasing energies and separated by ω phonon , plus a down-scaled copy of this sequence, shifted by ω′ phonon , representing hot-band contributions (Drobot 2015;Višňák and Sobek 2016;Haubitz et al. 2018).
We follow Drobot (2015) to illustrate the model with a synthetic spectrum as depicted in Fig. 5. We assume Lorentzian peak shapes, the zero-phonon line at 500 nm (2.48 eV), phononic energy ω phonon of 99 meV as given by (UO 2 ) 2+ stretching vibrations and determined by Raman spectroscopy (see Colmenero et al. 2019), an empiric width of the phononic peaks of 25 meV (compare Drobot et al. 2015) and an empiric hot phonon energy ω′ phonon of 60 meV with hot-band intensity of about 5% of the total intensity.
Intensity of the phononic bands follows mainly the sequence 1:1:0.5:0.25:0.1 (compare Višňák and Sobek 2016). The resulting artificial spectrum already covers the main characteristics of generic uranyl emission. In the spectrum, the phononic bands are composed of the transitions from the bottom of the excited electronic state to a phononic level of the ground state plus a small (few %) contribution of hotband transition, which is merely sufficient to show up as a small shoulder (Fig. 5). Here, each phononic band appears as a distinct single maximum, and bands are not doubled.
The parameters used here basically correspond to uranophane-β. Even though the outlined model was developed for uranyl ions in solution interacting with ligands, it represents a surprisingly good approximation in the case of uranyl silicates. Here, we introduced the model to explain spectral features and to lay the foundation for further discussion rather than to provide fitted data.

Samples
Samples from different localities are of the authors' collections, or were kindly provided by various private collectors. Table 3 lists locations and samples. All samples have been examined visually to exclude obvious misidentifications and to check for luminescing contaminations. For a series of samples, energy dispersive X-ray spectroscopy (EDS), Raman analysis and X-ray diffraction (XRD) have been performed to ensure assignment of the minerals. As outlined above, especially Raman analysis was found a useful technique for fast assignments of larger number of samples (Frost et al. 2006a-d;Wall et al. 2010;Colmenero et al. 2018;Colmenero et al. 2019).

Sample characterization
Sample characterization (service purchased from Mineralanalytik Eu, Germany) was performed by Raman, EDX and/or XRD measurements. The Raman system (Horiba XploRa Raman microscope) was mainly used with 532 nm excitation. To reduce influence of fluorescence, 785 nm excitation was used where necessary. Typical spot sizes on the sample were about 2 µm and smaller, ensuring measurement on individual crystallites. As an example, Fig. 3 shows measured Raman spectra (red curves) for each species. They match the corresponding reference data (black curves) well.
For elemental analysis, SEM-EDX was used (Hitachi S3700N together with Bruker Quantax200). Samples were measured at 20 kV acceleration voltage and at a residual pressure of 40 Pa to reduce skirt effect and surface charging. Structural determination was done by powder X-ray diffraction (Rigaku Miniflex benchtop system) in Bragg geometry with Cu K α -source. Additionally, confocal Raman microscopy at 532 nm excitation wavelength was applied on selected samples (alpha300 R; WITec GmbH, Ulm, Germany; equipped with a Nd:YAG laser operated at 3 mW; objective × 20 NA0.75; system established at the Institute of Materials Science, Physics of Surfaces, Technical University of Darmstadt, Germany).

Epiluminescence spectrometer with light sheet excitation
Spectral data were acquired using an epiluminescence spectrometer setup at ambient conditions with optional light sheet excitation, adapted to the investigation of uranyl luminescence in solid state samples. It is designed as a slit spectrometer (25 µm entrance slit) with a 15 × NA0.28 Beck/Ealing reflecting objective for UV applications. A longpass filter (Chroma LP435) blocks the excitation light. The light is collimated (75 mm f/4 achromatic collimator) onto a combined dispersive element (grism, a combination of transmission grating with 600 lines/mm and prism; dismounted material, unknown manufacturer), which results in a rather compact set-up. The spectrum is imaged by a cooled monochrome CCD camera (1392 × 1040 pixel, sensor Sony ICX285AL, camera Starlight Xpress Trius) using a 60 mm f/4 apochromatic objective (Apo-Componon, Schneider Kreuznach).
For excitation of crystal aggregates, a condensed LED (Nichia NVSU233A-D1 with nominal peak wavelength of 365 nm and half width at full maximum of around 8 nm) together with a dichroic beam splitter (Chroma AT440DC) were used. The excitation wavelength was additionally filtered (Schott UG1) to suppress contributions in the visible.
Spatially resolved measurements on individual crystals were achieved by a light sheet excitation scheme. Here, the light source is a continuous wave diode laser with 405 nm central wavelength and nominally 300 mW optical power (Osram LD). The laser is coupled into a 405 nm single mode fiber by passing a field stop and focused by a microscope objective (20 × NA0.5 Olympus HC PL Fluotar). At the fiber output, the beam is first collimated, then expanded and refocused onto the sample (objective: 10 × NA0.3; Olympus HC PL Fluotar). A rotatable polarizer provides intensity control. In order to suppress any possible long wavelength contribution, the excitation is filtered by a bandpass cleanup filter (405 nm, Chroma ZET405/20x). Excitation spot sizes have been determined to be in the range of 3 µm to 5 µm in diameter, depending on the scattering behavior of the sample under investigation, with an estimated extension in depth of about 10 µm. Positioning is fully adjustable in order to line scan an individual crystal. Positioning and focusing were controlled via a second camera. A distance of 25 µm between the measuring positions was selected for the scans.
Wavelength and efficiency calibration of the spectrometer was performed using the solar spectrum and its absorption lines as a reproducible standard. The pixel resolution of the spectrometer is 0.13 nm/pixel at a spectral resolution of 0.94 nm as estimated from the line-broadening of He-emission lines. A dark correction was applied to each raw luminescence spectrum to eliminate thermal effects from the CCD sensor as well as systematic scattering and autofluorescence.

Measurement and data analysis
Through careful examination, the specimens were selected to eliminate any possible influence by intergrown material of different species. Crystals on matrix, detached crystal fragments and individual crystals extracted from aggregates were investigated. No significant influence of the sample type on the measured data was detected.
Aggregates on matrix were directly measured, crystal fragments or individual crystals were prepared onto nonfluorescing double sided tape for spectroscopic measurements. In all cases, background luminescence was excluded by adapting illumination as well as field and direction of view. In addition to different sizes and qualities of crystal fragments, material from different localities showed strongly varying intensities. Exposure times and illumination intensities were adapted to ensure stable measurement conditions especially concerning noise level and sample integrity. Under the given conditions, no common base could be established to quantify intensity. Therefore, only normalized data are presented and discussed.
Spectra were corrected for possible offset and normalized to the maximum of the ZPL. That type of normalization reflects the concept, that the electronic transition into the electronic ground state without contribution of a phonon is the most characteristic and the most conserved. Structural disorder is expected to result in a small redshift of this transition. Note, this normalization conserves spectral shape and peak positions, but not intensities. In case of the variable double peak structure, commonly observed in uranophane-α, the higher peak was chosen.
Peak positions are determined by a second-order fitting algorithm that includes at least 20 raw data points on each side of the peak.
Further characterizing parameters are based on an empirical approach. As measure, the centroid wavelength centroid is introduced, defined according to Eq. (1): where I n denotes the normalized intensity at data point number n with respective wavelength n in the range min (1) centroid � min .. max � = ∑ n I n n ∑ n I n to max . As overall measure the global centroid wavelength global = centroid (485 … 615nm) provides a basic measure for perceived color and respective color shift. The chosen wavelength range is limited to 615 nm to avoid bias by possible contribution of Eu 3+ luminescence.
Local ordering parameters are defined similarly, measuring the contribution of a possible double peak structure. In the case of uranophane (i.e. uranophane-α and uranophane-β), this quantity is given as uranophane = centroid (493. ..500.5nm) . Further local ordering parameters sklodowskite = centroid (503 … 510nm) in the case of sklodowskite and haiweeite = centroid (505.5 … 512nm) for haiweeite. To remain in the scope of this study, further possible parametrizations are not included.
The centroid wavelength calculated in the sense of a "center of mass" includes aspects of the shape of spectral features by weighting wavelength with intensity for each pixel included in the calculation. The values are of the dimension of a wavelength with unit nm. About 50 raw data points (1000 raw data points) contribute to one value of the local centroid wavelengths uranophane , sklodowskite and haiweeite (global centroid wavelength global , respectively).

Spectra of uranophane-α, uranophane-β, sklodowskite and haiweeite
The photoluminescence emission spectra of three uranophane-α crystals (#089/3, Musonoi, DR Congo; #017, Rauris, Austria; #012, Le Limouzat, France; for all of them, identity was confirmed by XRD and EDS) in Fig. 6a together with the photograph of two uranophane-α aggregates (Menzenschwand, Germany) luminescing under ultraviolet excitation (Fig. 6b) give a visual example of variability in spectra and color of uranophane-α. While the phononic sequence in the case of #089-3 is built up by single emission peaks, a pronounced double peak structure is visible in #017. There, the phononic bands exhibit two successive maxima, making the spectrum similar to that presented by Kuta and colleagues (2013)-there, notably, measured on synthetic material. Spectrum #012 is broadened and starts losing structure. In the photograph (Fig. 6b) the visual color of the left aggregate is intense green whereas the aggregate on the right appears almost yellow.
For detailed comparison, emission spectra of the four mineral species from different locations were acquired and analyzed. A full list of the specimens is given in Table 3. Figure 7 shows for each mineral species the respective ensemble of emission spectra. Fig. 6 a Spectra of three samples of uranophane-α, for each of them identity has been confirmed by XRD and EDS. For #089-3 (black; Musonoi, DR Congo), the zero-phonon line (ZPL) is pronounced as compared to #017 (red; Rauris, Austria) and #012 (blue; Le Limouzat, France). While the spectrum #089-3 is composed of single phononic bands with one maximum, those bands appear doubled (with two maxima) in the case of spectrum #017. Spectrum #017 resembles the spectrum given in (Kuta, 2013). In spectrum #012, structure is already partially lost. b This photograph (image width 8 mm) of two uranophane-α aggregates extracted from different samples from Menzenschwand, Germany, exemplifies the emission color range from green (left) to yellow (right). Both samples were imaged together under 365 nm illumination All of the spectra resemble the generic spectrum shown in Fig. 6, in line with the luminescence mechanism outlined. Spectra are normalized to the ZPL, accounting for the characteristic nature of the ZPL. At visual inspection, the most intense sample of each species appeared of similar brightness. However, several inseparable sources lead to differences in brightness: size of the crystallites, efficiency of internal scattering as well as defect and disorder induced quenching affect the measured intensity. Technical issues (e.g. limited available excitation energy) further limit the use of intensities as a parameter in this analysis. Therefore, we omit further discussion of intensity as a useful parameter.
As a general observation, we state that the spectra of the different samples and locations of each species are similar, but to different degrees. The ZPL of uranophane-α shows a redshift up to 5 nm (smaller for the other species) pointing to defects and structural disorder (Fig. 7a), with some spectra apparently losing the characteristic phononic fingerprint. Some of the uranophane-α spectra carry the signature of Eu 3+ , an impurity ion co-excited with the uranyl ions. As further observation, the variation in peak valley contrast of the sklodowskite spectra indicates relative differences in unstructured background emission, while at the same time, the peak widths appear mostly conserved ( Fig. 7c). Overall, the spectral fingerprint is well conserved in the case of sklodowskite. In contrast, peak-broadening together with the unstructured emission background leads to an almost complete loss of spectral features in the case of uranophane-β (Fig. 7b). The spectra of haiweeite appear rather stable (Fig. 7d) with only little broadening.
However, the most striking feature in the spectra is the appearance of a pronounced double peak structure (marked as DPS in Fig. 7a) only in the case of uranophane-α. The double peak structure is best visible at the ZPL. This phononic band splits into two contributions with maxima at about 494 nm and 501 nm. The position of the two maxima is preserved throughout the ensemble of uranophane spectra (no additional intermediate maxima appear), with both peaks affected by the redshift in a similar way. The double peak structure affects all phononic bands in a similar manner, although the effect is masked by a general loss of spectral structure.
The double peak structure does not appear in any of the other ensembles, especially not in the case of the isostructural sklodowskite. In contrast, broadened features or even loss of spectral structure accompanied by a global redshift can be observed for all species.

Individual crystals
Such variations and the appearance of a double peak structure in the case of uranophane-α is not an effect related to the ensemble measured, but an effect already encountered on the level of individual crystals. To address this, individual crystals extracted from aggregates were investigated scanning along the long crystal axis. With spot size below 5 µm diameter and estimated elongation in depth of 10 µm for the exciting 405 nm laser beam, 25 µm spacing was chosen to take spectra. Again, for comparison, crystals of uranophane-α, uranophane-β and sklodowskite have been investigated. For all three samples, identities have been confirmed by EDS and XRD or Raman spectroscopy. Note, that due to generally small crystal dimensions of available uranophane-β, the scan spans only 100 µm while almost 1000 µm was reached on uranophane-α. As a guide to the eye, spectra at equidistant positions are highlighted in red, see respective figure caption. The spectra of all three species show variations at different positions. However, only in the case of uranophane-α (Fig. 8, sample #086, Menzenschwand, Germany), the evolution of a pronounced double peak structure is visible with rather localized changes in spectra, e.g. at positions around 100 µm to 150 µm and 625 µm to 675 µm. At 125 µm the centroid wavelength reaches 496.55 nm, about 0.12 nm more than at the adjacent points, while at 650 µm, the centroid wavelength is 0.15 nm less than at the adjacent points. These values have to be regarded on the scale, where the Large spectral variability with emerging double peak structure is evident in the case of uranophane-α. For each of the species the base spectrum is highlighted in red, details see text. Note, due to normalization to the ZPL, the spectra cannot directly be compared concerning their intensities average value of uranophane of uranophane-α (496.45 nm) and uranophane-β (497.11 nm) are separated by less than 0.7 nm.
For detailed inspection, three spectra are collected in Fig. 9: at 100 µm (#086-4), 625 µm (#086-26) and 650 µm (#086-27). While spectrum 086-4 shows only a small contribution of a double peak structure, it appears rather prominent in the other two cases. The appearance of a double peak structure (labeled DPS in Fig. 9) in the case of trace #086-26, even with dominant peak at around 505 nm, is accompanied by a prominent redshift of the peak positions not only of the ZPL. On the other hand, the leading hot-band transition 1 ′ → 0 hardly changes, neither in position nor in signal contribution. In trace #086-27, immediately taken after #086-26, the double peak structure is again less prominent and the ZPL matches the one of trace #086-4. Note, as data was taken on an individual crystal, those variations are intrinsic to the crystal and not imposed by external influences.
In the cases of uranophane-β ( Fig. 10; sample #088, Nopal mine, Mexico) and sklodowskite ( Fig. 11; #090, Clara mine, Germany), no indication of a possible double peak structure is visible in the spectra taken on individual crystals. This is in line with the observation on the data ensembles (Figs. 7b, c).

Characteristic base spectra
To further analyze these observations, we define the centroid wavelength global evaluated in the range of 485-615 nm (details see "Epiluminescence spectrometer with light sheet excitation") as a global ordering parameter. This quantity is intended to measure global shifts of the spectral fingerprint reflecting to some extent the perceived luminescence color. Lower global correspond to perceived green, while higher values indicate shift to yellow/orange. As a second quantity, the centroid wavelength uranophane , evaluated in the range 493…500.5 nm focuses on the absence or presence of a double peak structure in uranophane. When only a small indication of a double peak structure is present, uranophane will be smaller as compared to the case of a significant contribution. Similarly, local ordering parameters are defined for the other species, as introduced in "Epiluminescence spectrometer with light sheet excitation".  Unfortunately, crystals of uranophane-β were significantly smaller as compared to their counterpart uranophane-α, only 6 spectra could be taken. a The spectra of this series hardly vary in shape, no double peak structure is visible, which is consistent with the overall observation of Fig. 7. To guide the eye, spectra taken at 0 µm and 100 µm are highlighted in red, data offset in the signal scale is 0.15 each. b The variability of the spectra along the crystal is parameterized by the centroid wavelength uranophane = centroid (493 … 500.5nm) (on x-axis) changing with position (y-axis) Before we discuss the variability mentioned above, we aim at providing spectra basically free of additional contributions, e.g. induced by defects. Those spectra are referred to as base spectra. In a crystal with low disorder, only a limited number of nonradiative relaxation pathways will contribute. Therefore, average emitted photon energy will be comparatively high. Hot-bands will be governed by the intrinsic Boltzmann distribution basically lacking alternate mechanisms to promote their contributions.
Thus, base spectra are those with minimal local ordering parameter ( uranophane in the case of uranophane-α and uranophane-β, sklodowskite and haiweeite correspondingly for the other mineral species). Resulting base spectra are highlighted red in Fig. 7 and are collected for better visibility in Fig. 12. To stabilize results, those spectra have been excluded from determination of the base spectra, where no clear peak structure of the phononic transitions was visible. Related peak positions of the resulting base spectra are the ones noted in Table 2. We hope to contribute with these spectra to the clarification of uncertainties arising from the observed double peak structures, e.g. mentioned by Frankland and colleagues (2022).

Spectral variability
Global centroid wavelength global serves for rough quantification of the apparent color shift. For haiweeite, global covers 525-530 nm, for sklodowskite the range is moderately wider (537-547 nm), while for uranophane-α, it spans 520-552 nm and for uranophane-β 524-553 nm. Note, the evaluation range of the global centroid wavelength is limited to exclude contributions of Eu 3+ , thus underestimates the red part of the spectrum.
With the global change of spectra, peak positions are affected as well. We expect the rising influence of further spectral contributions to indicate a rise in competing relaxation pathways, thus additional broadening and redshifting of the spectroscopic structures.
For uranophane-α, peak position of the zero-phonon line is plotted in Fig. 13a against parameter uranophane , used as heuristic order parameter. The peak position shifts from 493 nm to about 498 nm. Along with a peak shift, the centroid wavelength uranophane increases, indicating growing influence of the double peak structure. Beyond uranophane = 496.55nm , the ZPL starts to get drowned in the growing additional peak. In such cases, only this additional peak of the double peak structure is identified, thus giving peak position values around 501 nm. Above uranophane = 496.66nm , the ZPL can no longer be identified as a separated peak, and the additional peak of the double peak structure finally governs the spectral structure leading to peak positions around 501 nm. The ZPL of the uranophane-β base spectrum is 500 nm: uranophane-α spectra at uranophane > 496.66nm apparently mimic spectra of uranophane-β. This is reflected in Fig. 13b, where the The spectra are rather uniform over the whole length of the crystal; no double peak structure is visible. To guide the eye, spectra taken every 100 µm are highlighted in red, data offset in the signal scale is 0.25 each. b The variability of the spectra along the crystal is parameterized by the centroid wavelength sklodowskite = centroid (503 … 510nm) Fig. 12 Normalized base spectra, a uranophane-α (Menzenschwand, Baden-Württemberg, Germany), b uranophane-β (Pasel adit, Radhausberg, Bad Gastein, Austria), c haiweeite (Los Azules mine, Atacama, Chile) and d sklodowskite (Clara mine, Oberwolfach, Germany) ensemble of uranophane-β (red triangles) meets with data points belonging to uranophane-α (black circles). Here, the differences between the parameters, centroid wavelength uranophane and peak position of the ZPL, become visible. While the centroid wavelength is capable to discriminate the two polymorphs (boundary at around 496.9 nm), the ZPL peak positions of uranophane-β are in the range covered by those of uranophane-α. In more detail, the centroid wavelength uranophane of uranophane-α ranges from 496.22 nm to 496.8 nm, while for uranophane-β, it spans 496.97 nm to 497.26 nm. Peak positions of uranophane-α cover 493.1 nm to 502.4 nm and include the range of 499.8 to 500.9 determined for uranophane-β.
Inspecting the same relation for the other mineral species, i.e. ZPL position versus respective local ordering parameters (parameter uranophane , sklodowskite and haiweeite ), reveals that a pronounced peak redshift is a property of uranophane-α only-basically lacking in the other data sets (Fig. 13b).

Spectra and their variability
The measured base spectra of uranophane-α, uranophane-β, sklodowskite and haiweeite as shown in Fig. 8 are expected to be already good estimates for true pure spectra. Both uranophane polymorphs can be distinguished from the closely related sklodowskite as well as from haiweeite on the basis of luminescence spectra. While we are aware that those spectra have been derived from natural material, measured peak positions are well in line with literature values ( Table 2). All four mineral species showed some variability in their spectra. Especially, the overall red shift is a common feature that can be interpreted by an additional defectinduced broadening, possibly related to radiation damage or weathering. Interestingly, the most prominent variability occurs in the case of uranophane-α, especially exhibiting a double peak structure. Our finding of a double peak structure is supported by previous data on natural (Haberlandt et al. 1950;Frankland et al. 2022) as well as on synthetic material (Kuta et al. 2013).
At this point we emphasize, that the prominent double peak structure was observed in the case of uranophane-α only, and for all uranophane-α available to this study, independent of locations. In several cases, the peak of the second spectral species at around 501 nm dominated the spectrum, reducing the ZPL of uranophane-α (at around 495 nm) to barely more than a "shoulder". While the luminescence spectrum in such cases might mimic uranophane-β, Raman and XRD data clearly indicated uranophane-α as the main species. For practical use of luminescence spectroscopy e.g. in field monitoring applications, this observation should be considered to avoid misinterpreting spectra of uranophane-α as uranophane-β. It might be helpful to explicitly address the absence of this "shoulder" for assignment of uranophane-β (provided accordance in the main spectrum).

Spectral reconstruction
To further address the double peak structure observed in uranophane-α spectra, we applied the model outlined in "Photoluminescence spectroscopy". Selected spectra of uranophane-α and uranophane-β are decomposed into respective phononic bands with additional hot-band contributions (Fig. 14). Applying the model, we are aware that in inorganic crystals, localized defects and substitutions might introduce changes to the model. An additional contribution to account for general spectral broadening and redshift, thus reflecting general loss of crystallinity or further alteration, might be necessary (Frankland et al. 2022). However, this analysis is not meant as modelling in a strict sense, but rather intended to illustrate spectral relations.
First, the spectrum of uranophane-β was parameterized in sequences of phononic and hot-peaks using Lorentzian peak shapes. Positions of the phononic peaks were taken Fig. 13 a Peak positions of the zero-phonon line (ZPL) of uranophane-α (on vertical axis) in dependence of the ordering parameter uranophane = centroid (493 … 500.5nm) (horizontal axis). As spectra show increasing double peak contribution with higher values of uranophane the ZPL shifts to longer wavelengths. At around 496.7 nm, the ZPL starts to get drowned in the overall broadened structure, with the emerging second peak dominating (around 501 nm). b Comparison of the variation of ZPL position with the respective ordering parameter centroid for the four species uranophane-α (black circles), uranophane-β (red triangles), sklodowskite (blue squares) and haiweeite (magenta diamonds). The part of the uranophane-α data (black) at around uranophane > 496.66nm and 502 nm peak position represents spectra where the ZPL is no longer discernable. It partially overlaps with the ensemble of uranophane-β (red). Compared to uranophane-α ZPL shift is small in the other cases. Details see text from measured data. Their average energetic separation ω phonon of about 97 meV corresponds rather well to the energy of the UO 2 2+ symmetric stretching vibration as determined by Raman (compared to 99 meV from Colmenero et al. 2019). The sequence of intensities with ratios 0.89:0.93:0.48:0.15:0.03 (compare to Višňák and Sobek 2016) with empiric peak width of 25 meV already delivers the main characteristics of the spectrum (blue components in Fig. 14a). To estimate hot-peak contributions (compare Fig. 5), the obtained phononic sequence is scaled by 0.06 with empiric width of 30 meV and shifted by the estimated hot phonon energy ω′ phonon of 66 meV (magenta components in Fig. 14a). By this scaling, we assure that the hot-band contribution remains limited to several percent. The hot phonon energy estimate is taken from data with the width taken as a fitting parameter. The full composed spectrum (red) matches the measured data (black in Fig. 14a) very well.
In a similar manner the base spectrum of uranophane-α is parameterized (Fig. 14b). Main phononic bands (blue) are characterized by a separation of 98 meV in accordance with Raman data, and intensity ratios (0.74:0.76:0.33:0.1:0.02) are similar to those of uranophane-β. Hot band contributions are shifted by 66 meV.
As the contribution of the hot bands to intensity is limited by the Boltzmann distribution to a few percent (Višňák and Sobek 2016;Haubitz et al. 2018), an additional component is necessary. This becomes evident regarding the relative intensities of bands necessary to reconstruct the uranophane-α spectrum. The leading hot-band transition of highest probability ( 1 ′ → 0 , at around 2.56 eV; Fig. 14b) is comparably small as in the case of uranophane-β. In contrast, the following bands necessary for reconstruction (e. g. 1 ′ → 1 , at around 2.48 eV; Fig. 14b) carry a rather large contribution (about half the amplitude of the related phononic band). Assigning those large contributions to multiphonon hot-bands would contradict thermal limitations-especially as the leading hot-band is of small intensity. Therefore, we conclude that an additional component-a second spectroscopic species-is necessary that shows a phononic structure with a well-defined relation to the phononic bands of uranophane-α.
As an empirical observation, we state that the spectral fingerprint (peak positions, spacing and relative peak intensities) of this second spectroscopic species (necessary to reconstruct uranophane-α spectra) matches well with the base spectrum of uranophane-β. Thus, we propose to directly use the scaled uranophane-β phononic sequence as an additional component representing the second spectroscopic species.
In detail, we reconstruct the uranophane-α spectrum using the main phononic sequence of uranophane-α plus corresponding hot bands (position and spacing from measured data, intensity scaled by 0.06 with respect to phononic peaks) and adding the scaled phononic spectrum of uranophane-β. Here, we use exactly the reconstructed base spectrum of uranophane-β described above (i.e. in the case of #117-6), uniformly scaled by a factor 0.45 (dashed brown bands in Fig. 14b). Again, the composed spectrum (red) matches the measured data rather well (black in Fig. 14b).
In an analogous manner, the spectrum of uranophane-α #065 with prominent double peak structure is reconstructed (Fig. 14c). Notably, the ZPL is redshifted by about 5 meV, separations of the phononic bands (blue in Fig. 14c) are  (Drobot 2015;Višňák and Sobek 2016): a uranophane-β, #117-6, b uranophane-α, #086-4 and c uranophane-α, #065. While the reconstruction of uranophane-β is possible under the assumption of small hot-band contributions, this is not evident for uranophane-α due to a marked double peak structure. However, assuming uranophane-β to contribute spectrally, reconstruction of the measured uranophane-α spectra becomes possible at small hot-band contributions. In cases (b) and (c), hot-band contributions are kept at the same level as in (a), while a significant contribution of uranophane-β spectrum is added. Notably, hot-bands of uranophane-α coincide with main phononic bands of uranophane-β. In all figures, measured data are given in black, main phononic contributions in blue, intrinsic hot-bands in magenta and scaled uranophane-β contributions as dashed brown lines. In each case, the sum spectrum is given in red. Details see text around 95 meV, intensity ratios are 0.38:0.88:0.4:0.12:0.03. Red shift and reduced phononic energy might reflect pronounced presence of defects. Again, following the concept outlined, a small intrinsic hot-band contribution is added (magenta) plus the uniformly scaled (factor 0.72) and shifted (6.5 meV) uranophane-β phononic sequence representing the second spectroscopic species (brown in Fig. 14c). Uniform scaling and the necessity of a uniform shift of the spectrum of the second spectroscopic species shows that the second spectroscopic species is incorporated in the crystal lattice. The reconstructed spectrum (red) approaches the measured data (black in Fig. 14c) rather well, even though some discrepancies become visible. Remarkably, the main phononic contributions are not dominating the double peak structure. Instead, the spectrum seems more related to uranophane-β.

Incorporated clusters of uranophane-β as proposed origin
The pronounced spectral variability of uranophane-α with respect to the other two uranyl silicates haiweeite and sklodowskite points to a pronounced variability on a structural level, as the uranyl luminescence is sensitive to its local environment. We briefly summarize the observations concerning the second spectroscopic species.
Reconstructing spectra of uranophane-α, a second spectroscopic species is necessary to limit hot-band contributions within thermal constraints. This second spectroscopic species is well described by the uniformly scaled phononic sequence of uranophane-β, and it redshifts with the shift of the uranophane-α phononic peaks. Additionally, this signature occurs and varies locally on individual crystals. Thus, the species is incorporated in the uranophane-α crystal. The contribution of a second spectroscopic species was observed to a variable degree in all uranophane-α spectra taken from samples of 16 different locations, and the spectral characteristics of the second spectroscopic species is conserved in the related data (see Fig. 7). Thus, we exclude chemical substitutions as direct source of the spectral contribution, as the double peak structure appears with one systematic signature, only varying in intensity. Next, the spectrum of the second spectroscopic species is well represented by (the phononic sequence of) the base spectrum of uranophane-β. The pronounced and variable double peak structure occurred in the case of uranophane-α only. We emphasize that none of the examined uranophane-α samples showed a spectrum where a contribution of uranophane-β could be excluded. No such double peak structure was visible in the spectra of sklodowskite or haiweeite-for both species no polymorphic forms exist either.
In this light, the hypothesis of strongly varying hot-band contributions as explanation seems unsatisfactory. Such variations should similarly occur in the other mineral species, especially of equivalent uranophane topology, i.e. in sklodowskite. However, this is not observed.
Likely, a possible contribution of a second type of UO 2 2+ luminescent center should apply as well in the case of uranophane-β and sklodowskite, thus not explaining the exclusive behavior of uranophane-α.
Therefore, we suggest that small clusters of a uranophaneβ-like phase are actually present in uranophane-α to a varying degree. As both uranophane polymorphs share chemistry and topology, such defects-essentially exchanging locally the building sequence ..udud.. by ..uudd..-might be stabilized by the polymorphism. Those clusters contribute to the luminescence with their respective spectra. The photon energy corresponds to the photon energy of uranophane-α hot-band luminescence, allowing for energy transfer. As these clusters represent defect structures incorporated in the uranophane-α lattice, the spectral signature shifts with the overall spectral redshift in a disturbed lattice.
Regarding the reverse case of uranophane-α clusters in uranophane-β, no contribution to the spectrum is expected. The bandgap energy of uranophane-α is larger than that of uranophane-β (2.68 eV, Colmenero et al. 2019), related to a higher ZPL energy of uranophane-α as compared to uranophane-β. Additionally, spectral features of uranophane-α do not match hot-bands of uranophane-β hampering reverse energy transfer.
For sklodowskite, no such mechanism is observable, as no polymorphic form exists. Haiweeite on the other hand has a different topology, so the reasoning is not simply transferable.
Our hypothesis of uranophane-β like clusters incorporated in uranophane-α is supported by observations of Wall and colleagues (2010), who report traces of uranophane-β in synthesis of uranophane-α.

Polymorphism and spectral variability: possible relations
Looking at UO 2 2+ in the well-defined environment of sklodowskite, all possible sites are predetermined (neglecting obvious defects like vacancies or impurities). The situation in uranophane is slightly different, as two polymorphs exist. Without obvious defects like vacancies or impurities, a localized transition between the two characteristic structures uranophane-α and uranophane-β might be possible. The uranophane topology is conserved in both cases. We consider the characteristic structural elements of the two uranophane structures: an alternating succession of up-and down-oriented SiO 4 -tetrahedra (Fig. 2). While the characteristic motif of uranophane-α is ..ud.. the equivalent motif of uranophane-β is ..uudd... Thus, at the growth front or at a domain interface, seeds of uranophane-β structure might readily form simply by misorientation or stacking defects of one chain unit, so e.g. ..ududu becomes by addition of one unit..ududuu, where the occurrence of uu might already favor continuation in uranophane-β type phase. Colmenero and colleagues (2019) showed uranophane-α to be the thermodynamically slightly preferred form of uranophane. So, a seed of uranophane-β structure in uranophane-α is not favored and growth will be limited by thermodynamic conditions. Small variations rapidly stop further propagation of this domain. On the other hand, the energy difference appears to be rather low, -12.0 kJmol −1 at zero temperature and pressure according to Colmenero and colleagues (2019). Thus, further variations in environmental conditions might again support reappearance of a uranophane-β domain, its growth eventually again being stopped and so forth. Such a mechanism might produce small, randomly distributed but localized domains of uranophane-β like phases in the general uranophane-α structure. Notably, the mechanism might occur similarly under natural as well as under laboratory conditions of crystal growth. The scenario described here is supported by the observation of typically different morphology of natural uranophane-α (crystals with high aspect ratio) with respect to that of natural uranophane-β (more bulky crystals), indicating differences in growth mechanisms and pointing to a rather fast growth of uranophane-α crystals (Schindler et al. 2004a, b;Schindler et al. 2004b). Additionally, approaches to obtain uranophane-β by synthesis resulted rather in uranophane-α phase (Cesbron et al. 1993).
As potential second pathway, intrinsically induced radiation damage and healing of related defects might play a role (Utsunomiya et al. 2003;Sureda et al. 2011). By radioactive decay, the structure is damaged while the chemical elements basically remain at the site, thus particle numbers can be regarded as approximately conserved. Under the condition, where no particles can be imported, healing of such defects might differ from growth in an open system. Even though this is rather speculative, some support comes from comparison to other systems (Nasdala et al. 2013;Lenz and Nasdala 2015;Lenz et al. 2020).
The situation is substantially different in the cases of sklodowskite and haiweeite. No polymorphic forms are available during growth or restructuring. Thus, a defect site created by a disturbance remains restricted to a point defect, as growth is not supported by a second polymorphic structure, or it vanishes. In consequence, structural variations that affect luminescence emission remain very limited. Thus, spectra are conserved to a higher degree in the case of sklodowskite and haiweeite as compared to uranophane-α.

Conclusion
In this study, the uranyl silicates uranophane-α, uranophane-β, sklodowskite, and haiweeite from more than 30 different localities in total have been investigated by steady-state photoluminescence spectroscopy. By determining spectra with maximum ZPL energy, we identified base spectra of each species representing a relevant approximation of the spectra of the pure mineral phases.
In their pure form, the spectra of the four minerals are clearly measurable and well distinct. For uranophane-β, sklodowskite and haiweeite, only low variability is observed in comparison to spectra of uranophane-α. Spectra of uranophane-α show-to varying degrees-a prominent double peak structure. We propose to attribute those contributions to uranophane-β like domains incorporated into uranophane-α, induced by ordering defects and supported in growth by the polymorphism of uranophane. Such a double peak structure was neither observed in the structurally related sklodowskite, where no polymorph exists, nor in the chemically related but structurally different haiweeite.
Our approach of relating the structure to the spectroscopic properties of uranophane may well be applicable to other systems and provide insights into the growth and decay processes of luminescent uranyl minerals. To progress further and to verify the existence of uranophane-β domains in uranophane-α, microanalytical methods such as confocal microscopy and transmission electron microscopy are highly desired. The use of artificial material, possibly synthesized by different routes in a similar study should allow to address the influence of structural disorder on luminescence induced, for example, by different impurities or growth conditions. However, it should be kept in mind that uranophane-β has not yet been successfully synthesized (Cesbron et al. 1993).
Finally, we like to emphasize that this study was entirely conducted in the spirit of a citizen science project by the authors and is neither linked to, nor funded by, any academic institution or project in industry. We warmly welcome any effort to further continue this work.