Coal Feed-Dependent Variation in Fly Ash Chemistry in a Single Pulverized-Combustion Unit

: Four suites of ﬂy ash, all generated at the same power plant, were selected for the study of the distribution of rare earth elements (REE). The ﬂy ashes represented two runs of single-seam/single-mine coals and two runs of run-of-mine coals representing several coal seams from several mines. Plots of the upper continental crust-normalized REE, other parameters derived from the normalization, and the principal components analysis of the derived REE parameters (including the sum of the lanthanides plus yttrium and the ratio of the light to heavy REE) all demonstrated that the relatively rare earth-rich Fire Clay coal-derived ﬂy ashes have a different REE distribution, with a greater concentration of REE with a relative dominance of the heavy REE, than the other ﬂy ashes. Particularly with the Fire Clay coal-derived ﬂy ashes, there is a systematic partitioning of the overall amount and distribution of the REE in the passage from the mechanical ﬂy ash collection through to the last row of the electrostatic precipitator hoppers.


Introduction
The concentration of the lanthanide-also known as the rare earth elements (REE)-and other critical elements during coal combustion has made fly ash a potential target for the extraction of metals.Whether through the incidental use of fly ash produced in the routine utility combustion of coal [1][2][3][4][5] or of fly ash produced for the secondary or even express purpose of recovering critical elements [6-9], the latter being a novel addition to the typical powergeneration uses of the region's coals, fly ash has some advantages compared to coal in terms of the removal of most of the carbon, the fine size of the material, and the potential availability of decades of fly ash production in landfills at or near the power plants.
The chemistry of the feed coal influences the chemistry of the fly ash.In consideration of Meij's [10] Venn diagram of element partitioning from the feed coal to the ash products, particularly for the low volatility trace elements that tend to concentrate in the fly ash and bottom ash, it is noted that the concentrations of the REE do not vary substantially from the feed coal ash to the combustion fly ash.Some partitioning may occur because of the redistribution of REE-bearing minerals [11].In this study, we discuss the differences in rare earth elements inherent in the differences between the coal sources fed into a single 100-MW boiler with an unchanging ash-collection system through the years of our studies.

Sample Sources
All fly ash samples were collected from the series of ash collection units for boiler unit 1 of Kentucky power plant I (letters were assigned to Kentucky power plants to anonymize

Notes on the Comparison of ICP Methods
One suite of series 4 samples (94012-94014, 94017-94019) was originally run on an inductively coupled plasma-optical emission spectroscopy (ICP-OES) instrument at the CAER (samples were digested following the ASTM D6357-11 [22] digestion method, which utilizes heating the sample with a combination of nitric, hydrochloric, and hydrofluoric acids).The method was modified to include an additional nitric acid step at the end to handle any potential solid residue that might be present, and a sample weight of 0.1 g was utilized.The certified reference material NIST 1633b was utilized as the primary method control sample).The samples were re-examined with ICP-MS for this study.In consideration of the known issues with the comparisons between ICP-OES and ICP-MS, studies have been made of the results from the two methods [3,23] and between ASTM D6357-11 22 (mixed acid digestion with heat); ASTM D4503 [23] (lithium borate fusion); and the ASTM D6357-11 [22] method, and with the addition of boric acid to neutralize the HF [24].For these six fly ashes, except for five fly ashes for the Nd analysis and one Minerals 2022, 12, 1071 3 of 14 ash for the La analysis, the comparison of the analyses of the light REE vs. the cluster of the heavier REE (including Pr and Sm) gave the impression of a reasonable correlation between the two techniques (Figure 1A).A closer examination of the results for Pr, Sm, and the heavy REE illustrated the uncertainties in the comparison of those elements (Figure 1B).Given the problems with high Tm analyses on the CAER's ICP-OES, the non-detection of Ho, and the general wider range of ICP-OES analyses than the corresponding ICP-MS analyses, with Ce being an exception (Figure 1A), ICP-OES is generally considered to be a less reliable technique than ICP-MS.
method control sample).The samples were re-examined with ICP-MS for this study.In consideration of the known issues with the comparisons between ICP-OES and ICP-MS, studies have been made of the results from the two methods [3,23] and between ASTM D6357-11 22 (mixed acid digestion with heat); ASTM D4503 [23] (lithium borate fusion); and the ASTM D6357-11 [22] method, and with the addition of boric acid to neutralize the HF [24].For these six fly ashes, except for five fly ashes for the Nd analysis and one ash for the La analysis, the comparison of the analyses of the light REE vs. the cluster of the heavier REE (including Pr and Sm) gave the impression of a reasonable correlation between the two techniques (Figure 1A).A closer examination of the results for Pr, Sm, and the heavy REE illustrated the uncertainties in the comparison of those elements (Figure 1B).Given the problems with high Tm analyses on the CAER's ICP-OES, the non-detection of Ho, and the general wider range of ICP-OES analyses than the corresponding ICP-MS analyses, with Ce being an exception (Figure 1A), ICP-OES is generally considered to be a less reliable technique than ICP-MS.In this study, we used REE to describe the lanthanide elements, REY for REE + Y, and REYSc for REY + Sc.The light REE (LREE) are defined as La through Sm and the heavy REE (HREE) are defined as Eu through Lu [25,26].Following the normalization of REE abundances to crustal averages (indicated by the suffix "N") [27], the normalized distribution can be divided into L-type (light type; La N /Lu N > 1); M-type (medium type; La N /Sm N < 1, Gd N /Lu N > 1); and H-type (heavy type; La N /Lu N < 1) enrichment patterns [28].Ratios based on the upper continental crust (UCC) corrections after Taylor and McLennan [27] are used to decouple Ce, Eu, and Gd from the other REE in the distribution patterns [20,[29][30][31]: Ce N /Ce N * = Ce N /(0.5LaN + 0.5Pr N ) (2)

Basic Element Trends
The chemistry of the four series of fly ashes is shown in Tables 1-6.Among the non-REE, the general trend for an increase in volatile elements towards the last rows of the electrostatic precipitator (ESP) array, a function of both the decreasing particle size and the cooler flue gas temperatures in the back rows, has been noted by Sakulpitakphon et al. [12], Mardon and Hower [32], Hower et al. [33], and Hood et al. [34], among others.In these samples, As, V, Mo, Zn, Cu, Ge, Ga, and Pb generally exhibit an increase in concentration toward the last ESP rows.Selenium also increases toward the third-row ESP in the series 3 fly ashes.Mercury concentration also increases towards the cooler end of the ash-collection system, but Hg capture is complicated by Hg's dependence upon the amount and form of carbon for efficient capture [35,36].

Principal Components Analysis on REY and Selected Major Oxides and Minor Elements
A principal components analysis (PCA; JMP ® Pro 16.0.0,SAS Institute, Cary, NC, USA) was implemented to further understand the distributions of the major oxides and minor elements.The elements and combinations of selected elements and oxides were REY, LREE/HREE, and K 2 O/(SiO 2 + Al 2 O 3 ) as an indicator of clay minerals, and Zr and TiO 2 /(TiO 2 + Al 2 O 3 ) as indicators of detrital minerals.The results, shown in Figure 2 with details on the statistics on the PCA tab in Table S1, demonstrate that the Zr and REY axes are in close proximity to each other, opposite the K 2 O/(SiO 2 + Al 2 O 3 ) axis, and orthogonal to the opposing LREE/HREE and TiO 2 /(TiO 2 + Al 2 O 3 ) axes.The first principal component, with nearly co-equal contributions from Zr, REY, and, in the opposite direction, K 2 O/(SiO 2 + Al 2 O 3 ), accounts for 62.51% of the variation.The first three principal components account for 96.80% of the variation.All of the eigenvectors make sense geologically.While not always specifically for REE or in the context of PCA, the nature of geochemical associations in eastern Kentucky and other coals has been discussed elsewhere [17,20,26,30,32,[37][38][39].For example, Y is an accessory element in zircons; therefore, Zr and REY are related.The nature of the REY associations with clays is different from the association in detrital minerals.Moreover, clays can act as a diluent of the REY-bearing clastic minerals, and TiO 2 /(TiO 2 + Al 2 O 3 ) indicates a strictly clastic source in contrast to the broader array of sources encompassed by the LREE/HREE.

Principal Components Analysis on REY and Selected Major Oxides and Minor Elements
A principal components analysis (PCA; JMP ® Pro 16.0.0,SAS Institute, Cary, NC, USA) was implemented to further understand the distributions of the major oxides and minor elements.The elements and combinations of selected elements and oxides were REY, LREE/HREE, and K2O/(SiO2 + Al2O3) as an indicator of clay minerals, and Zr and TiO2/(TiO2 + Al2O3) as indicators of detrital minerals.The results, shown in Figure 2 with details on the statistics on the PCA tab in Table S1, demonstrate that the Zr and REY axes are in close proximity to each other, opposite the K2O/(SiO2 + Al2O3) axis, and orthogonal to the opposing LREE/HREE and TiO2/(TiO2 + Al2O3) axes.The first principal component, with nearly co-equal contributions from Zr, REY, and, in the opposite direction, K2O/(SiO2 + Al2O3), accounts for 62.51% of the variation.The first three principal components account for 96.80% of the variation.All of the eigenvectors make sense geologically.While not always specifically for REE or in the context of PCA, the nature of geochemical associations in eastern Kentucky and other coals has been discussed elsewhere [17,20,26,30,32,[37][38][39].For example, Y is an accessory element in zircons; therefore, Zr and REY are related.The nature of the REY associations with clays is different from the association in detrital minerals.Moreover, clays can act as a diluent of the REY-bearing clastic minerals, and TiO2/(TiO2 + Al2O3) indicates a strictly clastic source in contrast to the broader array of sources encompassed by the LREE/HREE.S1.
The PCA analysis, as shown in Figure 2, also demonstrates that the series-3 Fire Clayderived ashes are distinct from the other series, a function of their higher concentrations of Zr and REY than the other three series of fly ashes.Further, the mechanical ashes occupy a distinct area from the ESP ashes with the ESP rows showing a sequential distribution (Figure 2 inset).In the first case, the differentiation is a function of the higher  S1.
The PCA analysis, as shown in Figure 2, also demonstrates that the series-3 Fire Clayderived ashes are distinct from the other series, a function of their higher concentrations of Zr and REY than the other three series of fly ashes.Further, the mechanical ashes occupy a distinct area from the ESP ashes with the ESP rows showing a sequential distribution (Figure 2 inset).In the first case, the differentiation is a function of the higher LREE/HREE and lower TiO 2 /(TiO 2 + Al 2 O 3 ) in the mechanical ashes vs. the ESP ashes.A similar differentiation drives the partition between the ESP rows.While this seems to contradict the inference of subtle, if any, REY variations between ashes from the same source [11], the differentiation in the amount and nature of the mineral inclusions seems to be sufficient to segregate the ashes.Of course, Liu et al. [11] focused on REY distributions, not a wider spectrum of major oxides and minor elements.They also examined the size fractions of a few single series-1 ashes with less emphasis on the discrete nature of the ashes from individual mechanical and ESP hoppers, as in this study.The nature of the ashes with respect to REY is further discussed below.

Lanthanide Elements
Apart from two fly ashes in the Manchester coal-derived series 1, all of the fly ashes have more than 390-ppm REE.The series-3 Fire Clay coal-derived ashes are all in the 604-to 638-ppm REE range.The series-3 Fire Clay-derived mechanical-collection fly ashes are the only ones in the study to show an L-type (normalized La > Lu) distribution ("dist."column in Table 6).The LREE/HREE distributions differ, with the mechanical ashes from the latter series having an LREE/HREE from 6.80 to 6.93, while the ashes from the 1st-to 3rd-row ESPs steadily decrease from 5.92-5.95 to 5.81 to 5.59-5.60.The mechanical ashes from series 2 and 4 also have higher LREE/HREE than the ESP ashes in those series.The overall order mimics the trend seen for the series-3 Fire Clay principal components (Figure 2), demonstrating the influence of the LREE/HREE on the trends and emphasizing the importance of REE distributions as a tool in understanding their behavior in combustion systems.
The upper continental crust-normalized REE distribution after Taylor and McLennan (1985) (Figure 3; Table 5) is cluttered.In consideration of the two single-source-coal-feed sets (Figure 4), it is evident that the trends seen in the PCA plots are also clearly seen in the normalized data.While both the ESP and mechanical ashes from the Fire Clay coal-derived series show negative Eu and positive Gd anomalies, the ESP ashes have a higher Tb through Lu distribution.Note that, while some of the lithotypes contributing to the Manchester coal-derived ash have high ash-basis REE contents, the ash contents of those lithotypes are low, therefore, they are not major contributors to the overall REE concentration of the fly ash [37,38].[11], the differentiation in the amount and nature of the mineral inclusions seems to be sufficient to segregate the ashes.Of course, Liu et al. [11] focused on REY distributions, not a wider spectrum of major oxides and minor elements.They also examined the size fractions of a few single series-1 ashes with less emphasis on the discrete nature of the ashes from individual mechanical and ESP hoppers, as in this study.The nature of the ashes with respect to REY is further discussed below.

Lanthanide Elements
Apart from two fly ashes in the Manchester coal-derived series 1, all of the fly ashes have more than 390-ppm REE.The series-3 Fire Clay coal-derived ashes are all in the 604to 638-ppm REE range.The series-3 Fire Clay-derived mechanical-collection fly ashes are the only ones in the study to show an L-type (normalized La > Lu) distribution ("dist."column in Table 1 (Rare earth-related parameters)).The LREE/HREE distributions differ, with the mechanical ashes from the latter series having an LREE/HREE from 6.80 to 6.93, while the ashes from the 1st-to 3rd-row ESPs steadily decrease from 5.92-5.95 to 5.81 to 5.59-5.60.The mechanical ashes from series 2 and 4 also have higher LREE/HREE than the ESP ashes in those series.The overall order mimics the trend seen for the series-3 Fire Clay principal components (Figure 2), demonstrating the influence of the LREE/HREE on the trends and emphasizing the importance of REE distributions as a tool in understanding their behavior in combustion systems.
The upper continental crust-normalized REE distribution after Taylor and McLennan (1985) (Figure 3; Table 1) is cluttered.In consideration of the two single-source-coal-feed sets (Figure 4), it is evident that the trends seen in the PCA plots are also clearly seen in the normalized data.While both the ESP and mechanical ashes from the Fire Clay coalderived series show negative Eu and positive Gd anomalies, the ESP ashes have a higher Tb through Lu distribution.Note that, while some of the lithotypes contributing to the Manchester coal-derived ash have high ash-basis REE contents, the ash contents of those lithotypes are low, therefore, they are not major contributors to the overall REE concentration of the fly ash [37,38].The distribution of the decoupled Ce N /Ce N *, Eu N /Eu N *, and Gd N /Gd N * distributions (Figures 5-7) all show the segregation of the Fire Clay-derived mechanical and ESP fly ashes both from each other and, particularly, for the Gd N /Gd N * vs. Eu N /Eu N * (Figure 6) and Gd N /Gd N * vs. Ce N /Ce N * (Figure 7) distributions from the other three series.The latter trends are largely driven by the high Gd N /Gd N * in the Fire Clay-derived ashes, an indicator of their high HREE concentration, compared to the other fly ashes.Series 2 is separated from the Manchester-coal-derived fly ashes and from the series 4 fly ashes in the Ce N /Ce N * vs. Eu N /Eu N * (Figure 5) and Gd N /Gd N * vs. Eu N /Eu N * (Figure 6) distributions, owing to their low Eu N /Eu N *.

Discussion
By focusing on a single 100-MW unit at a power plant, we were able to focus on variations in fly ash chemistry, as they are influenced by variations in the feed coal.From the PCA of several parameters, plots of the decoupled Ce N /Ce N *, Eu N /Eu N *, and Gd N /Gd N * distributions, and spider plots of the upper continental crust-normalized REE values, we observed that the Fire Clay coal-derived fly ashes were distinctive in composition compared to the other three series.(Although, note that the spider-plot contrast is only evident in the comparison of the Fire Clay-and Manchester-derived fly ashes).Largely, this contrast is a function of the increased presence of REY-bearing minerals in the Fire Clay coal [26,32,38,39] compared to other coals in the region.The other coals are not necessarily depleted of REY.For example, the Manchester coal in the mine supplying the coal for Sakulpitakphon et al.'s [12] study of the series-1 fly ash has several benches with >1600-ppm REY (ash basis) [38].Nevertheless, none of them have the lateral continuity of the Fire Clay coal.

Discussion
By focusing on a single 100-MW unit at a power plant, we were able to focus on variations in fly ash chemistry, as they are influenced by variations in the feed coal.From the PCA of several parameters, plots of the decoupled CeN/CeN*, EuN/EuN*, and GdN/GdN* distributions, and spider plots of the upper continental crust-normalized REE values, we observed that the Fire Clay coal-derived fly ashes were distinctive in composition compared to the other three series.(Although, note that the spider-plot contrast is only evident in

Discussion
By focusing on a single 100-MW unit at a power plant, we were able to focus on variations in fly ash chemistry, as they are influenced by variations in the feed coal.From the PCA of several parameters, plots of the decoupled CeN/CeN*, EuN/EuN*, and GdN/GdN* distributions, and spider plots of the upper continental crust-normalized REE values, we observed that the Fire Clay coal-derived fly ashes were distinctive in composition compared to the other three series.(Although, note that the spider-plot contrast is only evident in Using the Fire Clay-derived fly ashes as an example, it was noted above that the LREE/HREE ratio decreases from the mechanical rows through to the third-ESP row.The PCA analysis demonstrated a subtle partitioning between the three ESP rows.This is confirmed for Ce N /Ce N * vs. Eu N /Eu N * (Figure 5) and Gd N /Gd N * vs. Eu N /Eu N * (Figure 6), owing to the Eu N /Eu N * increases from the first to the third ESP rows.Further, the Eu N /Eu N * for the mechanical fly ash is lower than for the ESP ashes.As with Liu et al.'s [11] examination of REY partitioning in fly ashes from the Series 1 fly ashes point towards the variations in LREE/HREE and the decoupled element distributions being a function of (1) petrographic variations between the samples from the individual collection times; (2) partitioning of the REY-bearing minerals; (3) variations in the chemistry of certain minerals.

Summary
The chemistry of four series of fly ashes generated in the same boiler but with different feed coal-either single-mine/single-seam coals in two cases or run-of-mine blends of coals from multiple mines and multiple seams-was examined with special attention to the concentration and distribution of the rare earth elements.
1/The principal components analysis demonstrated that series 3, the fly ashes derived from the combustion of the Fire Clay coal, were (1) distinctly partitioned from the other fly ash series; (2) internally divided between the mechanical hoppers and the three rows of ESP hoppers; (3) the ESP rows showed a subtle separation.The separation of the series-3 Fire Clay-derived ashes from the other series was driven by the higher Zr and REY in the series 3 ashes and the partitioning between the mechanical and ESP fly ashes, and among the ESP ashes is a function of the variations in the LREE/HREE and the TiO 2 /(TiO 2 + Al 2 O 3 ) ratios.
2/Upper continental crust normalization [27], particularly the plot of just the two single-mine/single-seam-coal-derived fly ashes, demonstrated the differentiation of (1) the Manchester coal-derived ash from the Fire Clay coal-derived ash; (2) in the REE abundances and distributions between ESP rows (Manchester coal source, series 1 fly ashes), and between the mechanical and ESP rows (Fire Clay coal source, series 3).
3/The decoupled Ce N /Ce N *, Eu N /Eu N *, and Gd N /Gd N * distributions are complicated due to the Ba interferences with Eu.Nevertheless, the Ce N /Ce N * vs. Eu N /Eu N * and, in particular, the Gd N /Gd N * vs. Eu N /Eu N * distributions confirm the distinction between the Fire Clay coal-derived ashes and the remainder of the fly ashes, and the Fire Clay coal-derived mechanical and ESP ashes from each other.Both trends are also evident in the plot of Gd N /Gd N * vs. Ce N /Ce N *, although the differentiation between the ESP rows is not as distinct as in the Ce N /Ce N * vs. Eu N /Eu N * and the Gd N /Gd N * vs. Eu N /Eu N * plots.
4/The chemistry of the feed coal, while not specifically addressed here, is an obvious factor in the concentration and distribution of REE in the fly ash.Specifically, the Fire Clay coal-derived series 3 fly ash has distinctly different distribution patterns than the other single-seam or coal blend series.In addition, there is evidence that the distribution of REE progressively changes from the mechanical ash collection system through the ESP rows.While this might seem to differ from the conclusions of Liu et al. [11], it is noted that this study focused on sized fractions of single fly ashes, not the differentiation associated with the temperature and particle size gradients in the passage from the mechanical hoppers to the third row of the ESP array.

Figure 1 .
Figure 1.(A) Comparison of ICP-OES and ICM-MS analyses for rare earth element contents in samples 94012-94019 (series 4).(B) A subset of the data, within blue outline in panel (A), focused on the lower abundance (and heavier) REE.Note that La, Ce, and Nd do not appear in this portion of the graph.

Figure 1 .
Figure 1.(A) Comparison of ICP-OES and ICM-MS analyses for rare earth element contents in samples 94012-94019 (series 4).(B) A subset of the data, within blue outline in panel (A), focused on the lower abundance (and heavier) REE.Note that La, Ce, and Nd do not appear in this portion of the graph.2.5.Notes on Rare Earth Nomenclature, Normalization, and the Expressions of Normalized Data

Figure 2 .
Figure 2. Plot of principal component analysis axes (right) and individual points (left).The inset shows the detail of the Fire Clay source (93953-93960) series 3 ESP data.The detailed key to the symbols is on TableS1.

Figure 2 .
Figure 2. Plot of principal component analysis axes (right) and individual points (left).The inset shows the detail of the Fire Clay source (93953-93960) series 3 ESP data.The detailed key to the symbols is on TableS1.

Figure 3 .
Figure 3. Upper continental crust-normalized REE distribution after Taylor and McLennan [27].All of the fly ash series are shown on this graph.

Figure 3 .
Figure 3. Upper continental crust-normalized REE distribution after Taylor and McLennan [27].All of the fly ash series are shown on this graph.

Figure 4 .
Figure 4. Upper continental crust-normalized REE distribution after Taylor and McLennan [27] for the Fire Clay-and Manchester-coal-derived fly ashes.The distribution of the decoupled CeN/CeN*, EuN/EuN*, and GdN/GdN* distributions (Figures 5-7) all show the segregation of the Fire Clay-derived mechanical and ESP fly ashes both from each other and, particularly, for the GdN/GdN* vs. EuN/EuN* (Figure 6) and GdN/GdN* vs. CeN/CeN* (Figure 7) distributions from the other three series.The latter trends are largely driven by the high GdN/GdN* in the Fire Clay-derived ashes, an indicator of their high HREE concentration, compared to the other fly ashes.Series 2 is separated from the Manchester-coal-derived fly ashes and from the series 4 fly ashes in the CeN/CeN* vs. EuN/EuN* (Figure 5) and GdN/GdN* vs. EuN/EuN* (Figure 6) distributions, owing to their low EuN/EuN*.

Figure 4 .
Figure 4. Upper continental crust-normalized REE distribution after Taylor and McLennan [27] for the Fire Clay-and Manchester-coal-derived fly ashes.

Figure 4 .
Figure 4. Upper continental crust-normalized REE distribution after Taylor and McLennan [27] for the Fire Clay-and Manchester-coal-derived fly ashes.The distribution of the decoupled CeN/CeN*, EuN/EuN*, and GdN/GdN* distributions (Figures5-7) all show the segregation of the Fire Clay-derived mechanical and ESP fly ashes both from each other and, particularly, for the GdN/GdN* vs. EuN/EuN* (Figure6) and GdN/GdN* vs. CeN/CeN* (Figure7) distributions from the other three series.The latter trends are largely driven by the high GdN/GdN* in the Fire Clay-derived ashes, an indicator of their high HREE concentration, compared to the other fly ashes.Series 2 is separated from the Manchester-coal-derived fly ashes and from the series 4 fly ashes in the CeN/CeN* vs. EuN/EuN* (Figure5) and GdN/GdN* vs. EuN/EuN* (Figure6) distributions, owing to their low EuN/EuN*.

Figure 5 .
Figure 5. EuN/EuN* vs. CeN/CeN* for all of the samples.Figure 5. Eu N /Eu N * vs. Ce N /Ce N * for all of the samples.

Figure 5 .
Figure 5. EuN/EuN* vs. CeN/CeN* for all of the samples.Figure 5. Eu N /Eu N * vs. Ce N /Ce N * for all of the samples.

Figure 6 . 14 Figure 6 .
Figure 6.Eu N /Eu N * vs. Gd N /Gd N * for all of the samples.

Figure 7 .
Figure 7. Ce N /Ce N * vs. Gd N /Gd N * for all of the samples.

Author Contributions:
Conceptualization, J.C.H. and H.H.-K.; Data curation, J.C.H. and R.K.T.; Formal analysis, J.C.H., J.G.G., S.D.H., T.D.M., H.H.-K.and R.K.T.; Funding acquisition, J.C.H., J.G.G. and H.H.-K.; Investigation, J.C.H., J.G.G., H.H.-K.and R.K.T.; Methodology, J.C.H., H.H.-K.and R.K.T.; Project administration, J.G.G. and H.H.-K.; Resources, J.G.G.; Writing-original draft, J.C.H.; Writing-review and editing, J.C.H., J.G.G., S.D.H., T.D.M., H.H.-K.and R.K.T.All authors have read and agreed to the published version of the manuscript.Funding: The current study is based upon work supported by the Department of Energy, Office of Fossil Energy and Carbon Management under Award Number DE-FE0032054.The 1999 and 2012 collections were funded by grants from the Commonwealth of Kentucky to the Center for Applied Energy Research.The collection and the original analyses of the 2014 and 2016 samples sets were completed as part of U.S. Department of Energy contract DE-FE0026952 and National

Table 3 .
Minor elements (ash basis with the exception of Se and Hg on the whole-sample basis).dl-detection limit.

Table 5 .
Upper continental crust-normalized rare earth elements.
N /Eu N * Ce N /Ce N * Gd N /Gd N * dist.