A spectral library for laser-induced fluorescence analysis as a tool for rare earth element identification

With the recurring interest on rare-earth elements (REE), laser-induced fluorescence (LiF) may provide a powerful tool for their rapid and accurate identification at different stages along their value chain. Applications to natural materials such as rocks could complement the spectroscopy-based toolkit for innovative, non-invasive exploration technologies. However, the diagnostic assignment of detected emission lines to individual REE remains challenging, because of the 5 complex composition of natural rocks in which they can be found. The resulting mixed spectra and the large amount of data generated demand for automated approaches of data evaluation, especially in mapping applications such as drill core scanning. LiF reference data provide the solution for robust REE identification, yet they usually remain in the form of tables of published emission lines. We show that a complete reference spectra library could open manifold options for innovative automated analysis. 10 We present a library of high-resolution LiF reference spectra using the Smithsonian rare-earth phosphate standards for electron microprobe analysis. We employ three standard laser wavelengths (325 nm, 442 nm, 532 nm) to record representative spectra in the UV-visible to near-infrared spectral range (340 1080 nm). Excitation at all three laser wavelengths yielded characteristic spectra with distinct REE-related emission lines for EuPO4, TbPO4, DyPO4 and YbPO4. In the other samples, the high-energy excitation at 325 nm caused unspecific, broadband defect emissions. Here, lower en15 ergy laser excitation showed successful for suppressing non-REE-related emission. At 442 nm excitation, REE-reference spectra depict the diagnostic emission lines of PrPO4, SmPO4 and ErPO4. For NdPO4 and HoPO4 most efficient excitation was achieved with 532 nm. Our results emphasise on the possibility of selective REE excitation by changing the excitation wavelength according to the suitable conditions for individual REEs. Our reference spectra provide a database for transparent and reproducible evaluation of REE-bearing rocks. The LiF spectral library is available at zenodo.org and 20 the registered DOI: 10.5281/zenodo.4054606 (Fuchs et al., 2020). It gives access to traceable data for manifold further studies on comparison of emission line positions, emission line intensity ratios and splitting into emission line sub-levels or can be used as reference or training data for automated approaches of component assignment. 1 https://doi.org/10.5194/essd-2020-296 O pe n A cc es s Earth System Science Data D icu ssio n s Preprint. Discussion started: 26 October 2020 c © Author(s) 2020. CC BY 4.0 License.

. Overview of rare earth element (REE) phosphate samples from the Smithsonian National Museum of Natural History used in this study for the LiF spectral library (based on Jarosewich and Boatner (1991); Donovan et al. (2002Donovan et al. ( , 2003

Data processing
The raw spectra were corrected for the spectral sensitivity of the detection unit with in-house calibration data for the respective technical setup. Consecutive data processing and visualisation were done using the R environment (R Core Team, 2014). Spectra over the full detection range of 340 -1080 nm required two separate measurements (long-pass 120 filter of e.g. 334 nm and 550 nm) to suppress the second order signals at double wavelength. For merging both, a mergesegment was defined with the starting point set to 560  difference between both spectra. Within the merge-segment, both spectra are then merged using the intensity difference between both spectra multiplied by a weighting factor that gradually changes from 0 to 1. At the beginning of the mergesegment, the difference multiplied by 0 is added to the first spectrum (measurements with 334 nm long-pass filter covering 125 the short wavelength range), and then added after multiplying by the gradually increasing weighting factor until the end of the segment to become congruent with the second spectrum (measurements with 550 nm long-pass filter covering the longer wavelength range). Finally, the partial spectra were joined using the first spectrum at wavelengths shorter than 560 nm, the calculated new merge-segment of the overlapping range and the upper spectrum of longer wavelength above the calculated merge-segment. 130 Spectra obtained from the set of embedded phosphate standards showed an intense broad-band emission attributed to the epoxy resin, in particular because of partly small and translucent sample crystals that are not perfectly exposed at the resin surface. We measured the epoxy resin at three locations next to REE crystals. The epoxy resin spectra show reproducible luminescence indicating material homogeneity and accordingly, allow for subtraction from the REE sample spectra. Minor variations can be attributed to signal noise, which was removed by calculating a mean spectrum from 135 the three epoxy resin spectra. We then subtracted the epoxy resin-related emission from the REE sample spectra. This operation compares to the removal of a non-sample related background continuum, allowing for an unbiased detection of remaining REE-related emission lines as well as of absorption features.
Peak detection was challenged by clustered emission lines and by distinguishing small REE emission peaks from noise artefacts. Therefore, merged spectra were smoothed in a running mean interval of 1.2 nm to reduce the noise, while 140 keeping the spectral sampling interval. The smoothing interval was carefully chosen to retain emissions of low intensity and details of clustered narrow emission lines. In the case of emission bands with clusters of many narrow emission lines, an automated peak detection was complicated and resulted typically only in the detection of the most prominent peak.
To identify also the individual emission lines within emission bands, we calculated a background continuum based on the spline smoothed running minimum (degrees of freedom 50, bin size 1.2 nm). After removing this background continuum, 145 we detected all peaks above a threshold of typically two standard deviations of the entire spectrum.

Sample homogeneity
Recorded LiF spectra represent a single measurement spot, where the signal to noise ratio was best. We checked for spatial variations and representativity of the selected LiF measurement spot using hyperspectral reflectance imaging. The images from an FX10 camera (Spectral Imaging Ltd.) with a spatial resolution of 0.17 mm depict the sample grains with 5 150 to 25 pixels (pure sample pixels), each containing the reflectance spectrum in a range of 400 nm -1000 nm. All spectra of the measured REE sample grain were analysed for features deviating from expected REE absorptions. The reflectance data was analysed using the R package 'hyperSpec' (Beleites and Sergo, 2018) and compared to the reflectance spectral library of the USGS (Kokaly et al., 2017).
The obtained reflectance spectra of pure sample pixels were consistent with respect to absorption feature position 155 (Fig.3). Variations in signal intensities have minor influences on relative absorption depth (standard deviation below 5 %).
The reference spectral library for standard material from USGS (Kokaly et al., 2017) comprises only REE oxides and in some cases chlorides, but no phosphates, but comparison of absorption position confirmed agreement with observed absorption features on our REE samples. None of the pure sample spectra contained additional absorption features and thus, reflectance spectra attest the spatial homogeneity and good sample quality.

REE reference spectra at 325 nm laser excitation
With a laser excitation at 325 nm the Smithsonian REE phosphate samples yielded representative spectra for Ce 3+ , Nd 3+ , Eu 3+ , Tb 3+ , and Dy 3+ (Fig.4). Prominent emission lines with high signal to noise ratios allow for an unequivocal assignment of transitions, and represent the diagnostic features needed for a LiF-based REE identification. The 165 high spectral resolution (spectral sampling 0.13 nm) of the measurement setup enables to separate multiple lines within emission clusters associated to splitting of transitions into sub-levels (Stark level splitting) as observed particularly in the spectra of Nd 3+ , Tb 3+ and Dy 3+ (Fig.4). Table 2, therefore, differentiates between most intense emission lines (main) and additionally detected less intense emission lines of the same emission band (minor).
For CePO 4 , the excitation results in a broadband emission between 374 nm and 540 nm (here given as FWHM: full 170 width at half maximum without prior background removal as this is not required for such broadband emission) and a maximum at 420 nm (Fig.4). The very broad emission is difficult to interpret based on our results and goes beyond the scope of this study. In general, it overlaps with the characteristic Ce 3+ emission band around 360 nm that is associated to a 5d -4f transition (e.g., Reisfeld et al., 1996;Gaft et al., 2005;Shalapska et al., 2014). Several studies report broadened   Figure 3. Spatial homogeneity evaluation with hyperspectral reflectance images from a SPECIM FX10 camera with a spatial resolution of 0.17 mm, example DyPO4. A: For REE spectra extraction from all spectra of a measured scene (image), the mean of three band ratios was calculated to reduce influences from noise and ensure accurate identification of sample spectra (pixels containing spectra from the sample grain, in the text referred to as pure sample pixels). Bands were selected according to position of diagnostic absorption features (BR = band ratio representing the ratio of signal intensities at two given wavelengths). B: A mean spectrum was calculated from all pure sample pixels. The standard deviation is given as an indication of the variability between spectra of pure sample pixels.
The resulting REE absorption spectrum (mean spectrum) was compared to reference data from the USGS spectral library (Kokaly et al., 2017).
LuPO 4 and YPO 4 (grey lines in Fig.4, both samples from same set of Smithsonian REE phosphate standards, cf. Tab.1).
The spectrum of EuPO 4 (Fig.4) shows distinct emission lines at 588 nm, 593 nm, 612 nm, 651 nm and 699 nm, repre-190 senting five transitions from 5 D 0 to 7 F i (i = 0, 1, 2, 3, 4, Gaft et al. 2005;Friis et al. 2010;Shalapska et al. 2014). The spectrum indicates splitting into sub-level emissions. Triplets of emission lines are recorded for the transitions to 7 F 2 and 7 F 4 with additional prominent lines at 622 nm and 685 nm, and a line of minor intensity at 616 nm and 689 nm, respectively. We observe two additional weak emission lines at 812 nm and 832 nm. These rarely described emissions relate to the 7 F 6 transition (Binnemans, 2015). The emission pattern corresponds to Eu 3+ observed in apatite (Reisfeld et al., 195 1996), where the ratio of emission intensities for 5 D 0 to 7 F 2 and 5 D 0 to 7 F 1 is low. However, the detailed spectroscopic sample characterisation can be found in Sharma et al. (2019). The emission lines in the TbPO 4 spectrum represent five prominent transitions (from 5 D 4 to 7 F i , i = 6, 5, 4, 3, 2) and two weak transitions (from 5 D 4 to 7 F 1 and 7 F 0 , (Nazarov et al., 2009;Shalapska et al., 2014;Fang et al., 2017). The transitions are evident as clusters of multiple emission lines with maximum intensities at 489 nm, 544 nm, 587 nm, 587 nm, 200 620 nm, 645 nm, 679 nm and 703 nm (Fig.4). Reisfeld et al. (1996) and Gaft et al. (2005) reported emission lines below 440 nm for the phosphate minerals apatite and monazite. Those transitions from 5 D 3 to 7 F i (i = 6, 5, 4) are not evident in our sample spectrum.
The DyPO 4 spectrum displays narrow emission lines with two prominent bands at 485 nm, 574 nm and two weak emission bands at 662 nm and 749 nm. The clustered emission lines correspond to the transitions 4 F 9/2 to 6 H i (i = 15/2, 205 13/2, 11/2, 9/2, Reisfeld et al. 1996;Gaft et al. 2005;Friis et al. 2010). Our instrumental setup allows for a detection of separate sub-level emission lines corresponding to effects of the phosphate crystal field. Absorption spectroscopy suggests sufficient Dy 3+ excitation when using a 325 nm laser, although higher efficiency can potentially be achieved using slightly longer wavelength laser (e.g. 352 nm or 366 nm, Friis et al. 2010). The broad, unspecific emission band causes strong masking of most REE-diagnostic emission lines <650 nm. In case of REE with diagnostic emission lines >650 nm, the broad-band emission is less problematic because observed patterns in the long wavelength region may be sufficient for the REE identification (e.g. in the case of ErPO 4 , spectrum of the SG: 220 single grain specimen in Fig. 5). Such 325 nm-excited spectra depicting both, absorptions and emissions in the same spectrum. The REE-related absorption features are recorded at wavelength <800 nm and diagnostic emission lines in the spectral region >650 nm. To suppress the broad-band emission and unravel the REE emission lines of interest, longer wavelength laser excitation proved successful.   Figure 5. REE spectra illustrating diagnostic absorption features within prominent broad-band emission at wavelengths <800 nm and REE emission lines for those >650 nm. Violet and pink PL spectra were measured using laser excitation at 325 nm for both sample specimen (D: samples embedded in disc, SG: single grain samples, grey shaded area represents epoxy resin mean spectrum expected for D samples without absorption features, emissions >650 nm are better visible in SG samples due to the lower intensity of the broadband luminescence). Black reflectance spectra depict reference data from the USGS library (Kokaly et al., 2017) 12 https://doi.org/10.5194/essd-2020-296 as well as Gaft et al. (2005), who also attributed the emission around 500 nm to the same origin. However, the observed positions lie at slightly (around 10 nm) higher emission wavelength that agree better to positions found by e.g. Friis et al. 240 (2010) and Runowski et al. (2019), and potentially result from a superposition of different transitions originating to certain portions from 3 P 0 , 3 P 1 and 1 D 2 . Nevertheless, the origin of this rather large shift of 10 nm is currently unknown and will require in-depth investigation beyond the scope of this study.

REE reference spectra at 442 nm laser excitation
The 442 nm laser-induced SmPO 4 spectrum depicts prominent emission line clusters at 560 nm, 597 nm, 643 nm and a cluster of minor intensity at 702 nm (Fig. 6). The first three emission lines correspond to Sm 3+ transitions from 4 G 5/2 245 to 6 H i (i = 5/2, 7/2, 9/2, see Tab Laser excitation of ErPO 4 at 442 nm unravels an emission at 550 nm, besides the emission at 855 nm and the prominent 260 cluster between 977 nm and 1004 nm (see Fig. 6), which were already detected at 325 nm excitation (see above). This first line represents an important diagnostic emission in the visible range (e.g., Reisfeld et al., 1996;Gaft et al., 1998;Friis et al., 2010). Gaft et al. (2005) assign respective emission lines to transitions from 4 S 3/2 to 4 I i (i = 15/2, 9/2) (first two lines) and 4 I 11/2 to 4 I 15/2 (third line) (see Tab  about 653 nm to Er 3+ that is only weakly present in our reference ErPO 4 spectrum (see Fig. 6) and identify the transition 265 from 4 F 9/2 to 4 I 15/2 to be responsible for this emission.

REE reference spectra at 532 nm laser excitation
For HoPO 4 , the diagnostic emission line around 540 nm (e.g., Gaft et al., 2005;Qin et al., 2011;Pandey and Swart, 2016) was masked by broad-band luminescence with Ho 3+ -related absorptions (e.g., Turner et al., 2014;Kokaly et al., 2017), when using 325 nm and 442 nm laser excitation (see above). The 532 nm excited spectrum captures only emissions 270 above 550 nm, but successfully suppressed the masking broad-band luminescence below 700 nm. Recorded emission lines cluster at 662 nm, 754 nm and 982 nm (see Fig. 7 and Tab. 4). The first, relatively weak emission line corresponds to a 5 F 5 to 5 I 8 transition (e.g., Friis et al., 2010;Qin et al., 2011;Pandey and Swart, 2016), while the second, more intense emission cluster results from a 5 F 4 / 5 S 2 to 5 I 7 transition (e.g., Qin et al., 2011;Pandey and Swart, 2016). The third and most prominent emission cluster in the near-infrared wavelength range resembles observations by Yu et al. (2012), who 275 detect a transition from 5 F 5 to 5 I 7 with emission lines at 965 nm. A combination with transition processes assigned to  Depending on the embedding material of measured NdPO 4 grains, the laser excitation with 325 nm may cause a 280 broad-band luminescence below 700 nm, while the use of the 442 nm laser revealed inefficient Nd 3+ excitation. Instead, the 532 nm excitation proved efficient and revealed two major emission line clusters, one between 860 nm and 910 nm (maximum at 908 nm) and the other one from 1040 nm to 1075 nm (maximum at 1059 nm) (see Fig. 7). The two emission clusters agree to the Nd 3+ -related lines resulting from a 4 F 3/2 to 4 I 9/2 and a 4 F 3/2 to 4 I 11/2 transition, respectively, and resemble common observations in phosphate minerals at 532 nm laser excitation (Tab. 4) (e.g., Reisfeld et al., 1996; 285 Gaft et al., 2005;Czaja et al., 2012;Shalapska et al., 2014). Reported emission lines below 800 nm by Gaft et al. (2005) are not present in any of our reference NdPO 4 spectra. Shalapska et al. (2014) identified additional emissions to be too weak, of which, however, the one to 4 I 15/2 at 846 nm (cf. Gaft et al. (2005)) may be also present in our reference spectrum contributing to the broad cluster around 908 nm.
The 532 nm laser excitation of YbPO 4 revealed a relatively weak Yb 3+ emission without the strong broad-band emission 290 at shorter wavelength than 700 nm observed in several of our REE PO 4 samples at 325 nm excitation (see Sect. 3.2).
The diagnostic Yb 3+ emission appears as a relatively broad band with a maximum at 1011 nm and a tentative shoulder at 1023 nm (Fig. 7). Results correspond to the main emission line reported by Gaft et al. (2005) and Czaja et al. (2012) and reflect a ( 2 F 5/2 , 4 F 5/2 ) to 2 F 7/2 recombination process (Tab. 4). Potentially overlapping emissions from splitting into Stark levels (Gaft et al., 1998(Gaft et al., , 2005Czaja et al., 2012)   The efficiency of excitation can be evaluated based on absorption spectroscopy, where prominent absorption features indicate energy uptake by the crystal. Figure 8 depicts the three laser wavelengths (325 nm, 442 nm, 532 nm) compared to our own absorption spectra using the same REE phosphate samples from the Smithsonian library measured with the portable field-spectrometer Spectral Evolution PSR 3500 (Fig. 8, orange line). Complementary, Figure 8 shows reflectance spectra from synthetic REE phosphates presented by Ropp (1969)  suitable conditions. Blue laser excitation at 442 nm is most efficient for Pr 3+ , matches also for Sm 3+ and Er 3+ , and overlaps with an absorption shoulder in Dy 3+ and Ho 3+ . The 532 nm laser excitation showed successful in particular for Nd 3+ and also Ho 3+ . The good excitation conditions with green laser light for Nd 3+ agree to other studies on excitation wavelength (e.g. Czaja et al., 2012;Lenz, 2015), while excitation at 325 nm was concluded to be inefficient. Fig. 8 suggests that a 325 nm laser is also able to excite Nd according to the coincidence with the edge range of an absorption feature. 320 Nevertheless, the real configuration of the crystal lattice apparently influences intensity ratios of sub-level emissions, and shifting emission band positions (Lenz et al., 2013). For Yb 3+ , it seems surprising that 532 nm as well as 325 nm excitation were successful, because both excitation wavelengths lie at longer wavelengths than the absorption at < 250 nm shown in Ropp (1969) (cf. Fig. 8). However, for example Liang et al. (2016) or Chakraborty et al. (2016) found also efficient absorption of excitation energy between 300 nm and 400 nm to charge transfer states that emit at wavelengths 325 longer than 900 nm. Another possible explanation may be the sensitisation of Yb 3+ by traces of other REE detected in the studied material by NAA (Donovan et al., 2002). The trace contaminations result from the reference material's production procedure, where the chemical similarity of REEs limits the complete separation of REE in the samples. Gaft et al. (2005) report Yb 3+ excitation to be strongly dependent on Nd, but NAA results from Donovan et al. (2002)  Although ultra-violet laser wavelengths (such as 325 nm used in our study) seem to be preferred in analyses of REE 335 emission spectra, the advantages of a selected use of excitation wavelengths were already recognised in many other studies, especially for Pr and Sm (blue, e.g. 442 nm) and for Nd and Ho (green, e.g. 532 nm) (e.g. Reisfeld et al., 1996;Gaft et al., 2005;Friis et al., 2010;. Our results agree to those investigations and further highlight the specific drawbacks of non-suitable excitation conditions. For example, we observed in several of our REE-PO 4 samples that non-selective or high-energy excitation (in our case especially the 325 nm laser) seem to cause a broad-band lumi-340 nescence with predominant absorption features that mask potential diagnostic REE emissions at wavelengths <800 nm (see chapter 3.2). This outlines the importance of appropriate laser wavelength selection, which in turn influences the spectral detection range. Fig. 9 shows how 442 nm and 532 nm lasers can unravel REE emission lines (cf. HoPO 4 ), but also illustrates how improved REE-signal to noise ratios may be at the cost of reduced detection ranges, which may cut-off those emission lines (cf. ErPO 4 in Fig. 9) at shorter wavelengths than the laser and long-pass filter wavelength, 345 respectively.
To give an overview, we summarise our results with benefits and drawbacks for REE identification dependent on the selected laser wavelength used in this study in Figure 10. Our results emphasise the potential of selective REE excitation for technical implementation or REE differentiation and the need for careful interpretation of selectively excited LiF spectra.
Such a selective approach may deliver the needed tools in analysis of material with unknown, mixed REE contents such 350 as in applications to natural rock.

Application to natural rock samples
Applications of the LiF spectral library for automated REE identifications have to deal with effects from overlaps, masking or energy transfer when several REE are present in the studied material. The presence of multiple REE may influence relative emission intensities and also the highly variable crystal field in natural samples may influence exact emission 355 line positions. Mixed spectra have to be expected especially from LiF of natural rocks because of the REE's similar geochemical properties and corresponding collective occurrence. Several REE seem to compete particularly due to a concentration of emission bands in the spectral range between 500 nm and 700 nm. As a result, the lines of certain REE are more likely hidden by the stronger luminescence of other REE. Therefore, an automated REE identification benefits from several emission bands available for peak matching and needs to account for suitable excitation conditions and 360 effects from co-occurrence.
Sm 3+ , for example, shows prominent peaks at 600 nm (major) and 650 nm to be most relevant for identification, while the peak around 560 nm overlaps with the more prominent Tb 3+ emission around 545 nm (Guan et al., 2016). Similarly, the same emission lines (600 nm and 650 nm) persist, when Dy 3+ contributes to the measured spectrum and masks the lower 560 nm emission line (Gaft et al., 2005). Pr 3+ is difficult to detect because its emissions are hidden by the lines 365 18 https://doi.org/10.5194/essd-2020-296    of Sm 3+ (600 -650 nm), Dy 3+ (470 -490 nm) and Nd 3+ (870 -900 nm) (Gaft et al., 2005). Blue lasers being suitable for both, Pr and Sm, complicate their differentiation, while for Dy 3+ and Nd 3+ selective excitation with UV (e.g., 325 nm) or green (e.g., 532 nm), respectively, provides an option. Tm 3+ was not recorded with our system but is reported to be typically masked by Tb 3+ (Gaft et al., 2005). None of the employed laser wavelengths were successful in revealing any diagnostic emission lines of Tm 3+ , while for both, the embedded and the single grain specimen, the significant 370 broad-band luminescence in the spectral range <700 nm dominated the spectra irrespective of 325 nm, 442 nm or 532 nm excitation. Therefore, our library cannot provide a Tm-spectrum for further applications to identify Tm in mixed spectra of natural samples. Nevertheless, we detected in all resulting spectra consistent absorption features at 355 -365 nm, a doublet at 465 -495 nm (most significant) and another smaller doublet 660 -710 nm (see above Fig. 5, bottom) that agree to diagnostic absorption features of reference samples from the USGS library (Kokaly et al., 2017). Other difficulties can 375 arise from energy transfer effects between REE, when several REE are present as it needs to be expected in natural materials such as rocks, or from cross-relaxation phenomena for high rare-earth concentrations. Extensive research has been done, of which one simplified example are suppressed Ho 3+ emissions in the presence of Dy 3+ , which can result in sensitisation of the Dy 3+ characteristic emissions at 480 nm and 580 nm (Friis et al., 2010;Lenz, 2015) 20 https://doi.org/10.5194/essd-2020-296  An application example is given in Figure 11 for a xenotime sample from Novo Horizonte. The LiF reference spectra 380 (black lines) allow for precise REE assignment to major emission peaks in the detected mixed spectra that were obtained by 325 nm (violet line), 442 nm (blue line) and 532 nm (green line) laser excitation. Especially Dy 3+ , Sm 3+ and Nd 3+ can be matched to the most prominent emission bands. The identified REE Dy 3+ , Sm 3+ and Nd 3+ (cf. Fig. 11 emissions compared to weaker Eu 3+ emissions (EMPA: 0.12 wt.%) at partly overlapping positions. Here, the co-existence of both REE (Sm 3+ and Eu 3+ ) seems to affect the observed emission line pattern and relative intensities shown by, for 395 example, an intense emission peak at around 600 nm (with an orange question mark between the diagnostic Sm 3+ and Eu 3+ emission lines in Figure 11). The Eu 3+ emission lines may also be identified in the 325 nm excited spectrum (cf. results section, where all three laser wavelength showed efficient for Eu 3+ excitation), but are in the case of this sample much less intense compared to the 442 nm excited spectrum, which may be attributed to complex interactions, e.g. with REE that are more efficiently excited with a 442 nm laser. Another Eu 3+ sub-level at 685 nm is, in contrast, not apparent 400 next to the intense Sm 3+ emission cluster around 700 nm. A change in relative emission intensities is further evident in the 532 nm excited xenotime spectrum at 872 nm. The only candidate responsible for this significant peak could be Nd 3+ that has a broad emission cluster in the same range but with opposite intensity ratios. The excitation at 532 nm suggests also efficient Ho 3+ stimulation (EMPA: 1.67 wt.%), which is a potential candidate for explaining the 750 nm emission overlapping with a Dy 3+ position, as well as for the 1000 nm cluster, where it is possibly mixed with Yb 3+ and Er 3+ . The 405 spectroscopic analysis of the presented phosphate rock sample from Novo Horizonte shows that the new LiF spectral library for REE phosphates enables to identify all EMPA-detected REE, despite Tm, down to a concentration of 0.08 wt%.
The observations confirm the valuable nature of reference spectra for natural sample evaluation and REE assignment.
But the application example highlights at the same time that care needs to be taken with respect to complex effects between REE and crystal lattice and between the, in natural minerals typically jointly occurring REE. Resulting mecha-410 nisms of energy transfer might influence relative peak intensities and complex effects from the crystal field may cause shifts of exact peak positions. Differences in relative peak intensities between assigned REE also underline the limits for quantitative conclusions. Compared to independent EMPA analysis, the Er 3+ related emissions remain weak relative to e.g. Sm 3+ and Nd 3+ during blue und green laser excitation despite its higher concentration of 5.26 wt.% versus 0.70 wt.% and 0.08 wt.%, respectively. Dedicated studies are required to draw conclusions on if and how a relative quantification 415 may be possible under which circumstances, e.g. in same host rocks from the same parent formation system.