Major-element geochemistry of pelites

Pelites (shales and mudstones) are arguably the most important rock type for interpreting metamorphism. Their significance derives from their widespread occurrence and the range of mineral assemblages they develop at different conditions of pressure and temperature. We compiled a global database of 5729 major-element whole-rock analyses of pelites from different metamorphic grades (shales to granulite-facies paragneisses) to (1) determine an average composition, (2) examine the range and variability in their composition, and (3) assess if there is evidence for grade-related geochemical changes. Median values are given instead of average values to eliminate the effect of extremes. The median worldwide pelite is as follows (anhydrous, values in wt%): SiO 2 = 64.13, TiO 2 = 0.91, Al 2 O 3 = 19.63, FeO total = 6.85, MnO = 0.08, MgO = 2.41, CaO = 0.65, Na 2 O = 1.38, and K 2 O = 3.95. The median X Mg = MgO/(MgO + FeO total ) in moles is 0.39. The median X Fe3 + = 2 × Fe 2 O 3 / (2 × Fe 2 O 3 + FeO) in moles was measured in 1964 samples and is 0.23. On an Al 2 O 3 -FeO- MgO (AFM) diagram, the median worldwide pelite plots within a strong clustering of analyses between X Mgproj = projected molar MgO/(MgO + FeO total ) = 0.30–0.55 (median = 0.42) and A Ms = molar [Al 2 O 3 – (3 × K 2 O)]/[Al 2 O 3 – (3 × K 2 O) + FeO total + MgO] = 0.0–0.4 (median = 0.19). Pelites show a continuous decrease in volatile content with increasing metamorphic grade and a decrease in X Fe3 + from the diagenetic to biotite zone. Lower median SiO 2 values and higher median Al 2 O 3 and A Ms values in the porphyroblast and subsolidus sillimanite or K-feldspar zones, as well as higher median MnO values in the garnet zone, may reflect sampling bias or metasomatism.


INTRODUCTION
Metapelites are metamorphosed clay-rich sedimentary rocks (e.g., shales and mudstones).They are widespread in the rock record, and ever since the pioneering work of George Barrow (1893), they have been useful as indicators of metamorphic grade because their mineral assemblages are sensitive to changing pressure and temperature (P-T) conditions (Hietanen, 1967;Pattison and Tracy, 1991;Spear, 1993).Several compilations of the geochemical compositions of metapelites have been assembled and their values compared to those of unmetamorphosed shales (Shaw, 1956;Ague, 1991).Numerous studies have calculated phase diagrams for an "average" pelite composition, as a means to infer representative P-T conditions of commonly occurring metapelitic mineral assemblages, and to compare P-T conditions in different metamorphic belts around the world (Mahar et al., 1997;Tinkham et al., 2001;Caddick and Thompson, 2008;White et al., 2014;Pattison and Spear, 2018;Forshaw and Pattison, 2021).However, these average pelite compositions are rarely the same, making strict comparisons difficult.Thus, there remains uncertainty over what truly is an average pelite, and how much compositional variability pelites show.
A related topic of debate is the degree to which elements are redistributed during the devolatilization that accompanies prograde metamorphism (Ague, 1991;Stepanov, 2021).While mass transfer of major elements can be significant in domains of high fluid flux (e.g., veins, skarns, and sometimes along lithologic contacts), mass transfer away from these domains is thought to be broadly negligible (Ague, 2011(Ague, , 2017)).Despite this, small variations in several of the most important major elements (e.g., Al, K, Mg, Fe 2+ , and Fe 3+ ) can lead to different mineral compositions, proportions, and, in turn, assemblages (Spear, 1993).Therefore, it is important to assess the degree of major-element mobility during prograde metamorphism, since this underpins our assumption that a single bulk composition can be used for phase diagram calculations across a range of metamorphic grades.
We present a new compilation of published pelitic whole-rock analyses ranging in metamorphic grade from shales to granulite-facies gneisses.It is ∼20 times larger than any previous compilation.We used this database to document the range in chemical compositions of pelites, provide an average pelite composition, and assess whether there are any significant compositional changes as a function of metamorphic grade.

NEW COMPILATION OF LITERATURE DATA
Several previous studies compiled bulk compositional data from the literature in order to assess major-element compositions of pelites and possible changes with metamorphic grade (Lapadu-Hargues, 1945;Shaw, 1956;Ague, 1991).Since these studies, the number of published pelitic whole-rock analyses in the literature has increased considerably, and many unpublished theses containing analyses are now available online.Using these, we constructed a new database of 5729 analyses from 364 studies categorized into 103 regions.Only individual rock samples that had been crushed and analyzed using bulk techniques such as wet chemistry or X-ray fluorescence (XRF) were included.Samples from domains of high fluid flux (e.g., described as selvedge or coming from vein margins) were excluded from the database because they have been subject to localized mass transfer.The following major rock-forming components were analyzed in all samples: SiO 2 , TiO 2 , Al 2 O 3 , FeO, MnO, MgO, CaO, Na 2 O, and K 2 O.Where reported by the authors of the previous studies, we also included determinations of P 2 O 5 , Fe 2 O 3 , and loss-on-ignition (LOI).For wet chemical analyses, we combined H 2 O, CO 2 , and SO 3 together as an estimate for LOI (e.g., volatile content), as was done by Ague (1991).The Supplemental Material1 describes in detail the methodology used to exclude nonpelitic analyses; categorize the samples according to metamorphic grade; and portray the data in Al 2 O 3 -FeO-MgO (AFM) diagrams involving compositional projection.A complete catalog listing all analyses, including the sample name, metamorphic zone, literature reference, and whole-rock data in weight percent, is available in the Supplemental Material.
Due to the large number of analyses in the database, we were able to divide the samples into finer-scale metamorphic grade-related categories than previous studies (cf.Shaw, 1956;Ague, 1991) while maintaining >250 analyses in each category (Fig. 1).Nine zones were defined: diagenetic, anchizone, chlorite, biotite, garnet, porphyroblast, subsolidus sillimanite or K-feldspar, migmatite, and garnetcordierite.The diagenetic zone refers to samples described as unmetamorphosed shales and mudstones, while the anchizone and chlorite zones refer to low-grade shales and/or slates classified using illite or chlorite crystallinity (Merriman and Frey, 1999).Higher-grade zones were based on the appearance of index minerals.The porphyroblast zone refers to muscovite-bearing, sillimanite-absent mineral assemblages containing combinations of cordierite, staurolite, andalusite, and with or without garnet.Although the stability of these mineral assemblages varies with pressure, they typically develop in a relatively narrow range of metamorphic grade (approximately equal to metamorphic temperature), upgrade of the consumption of chlorite and downgrade of the development of sillimanite or K-feldspar (Fig. 1).The subsolidus sillimanite or K-feldspar zone includes muscovite + sillimanite-bearing mineral assemblages and, at low pressure, subsolidus K-feldsparbearing mineral assemblages, which develop at different pressures but at similar metamorphic grade.The migmatite zone includes mineral assemblages in suprasolidus rocks that contain combinations of garnet, cordierite, sillimanite, and K-feldspar but lack coexisting garnet and cordierite.In total, 308 samples could not be classified into one of these zones.

TREATMENT OF DATA Projections
Projection of bulk compositions into an AFM diagram (Fig. 2A; Thompson, 1957) permits visualization of the variation in the three main compositional variables of pelites, namely, Al, Fe, and Mg.It also allows the compositional effects of varying modal proportions of quartz and plagioclase feldspar in the original sedimentary protolith to be filtered out.Whole-rock analyses comprising 9-12 components (see above) were reduced to the six-component KFMASH (K 2 O-FeO-MgO-Al 2 O 3 -SiO 2 -H 2 O) system using projections from apatite, ilmenite, albite, and anorthite to remove P 2 O 5 , TiO 2 , Na 2 O, and CaO, respectively.All iron was assumed to be FeO (ferrous), and MnO was omitted.To plot the KFMASH analyses in an AFM diagram, further projection from quartz (SiO 2 ), hydrous fluid (H 2 O), and either muscovite or K-feldspar (K 2 O) was required.Given that the majority of samples in the database were muscovite-bearing samples, the AFM diagram in Figure 2A shows all sam-ples projected from muscovite (A Ms ). Figure S2 includes diagrams projected from K-feldspar, as well as diagrams separating lower-grade, muscovite-bearing samples from higher-grade, K-feldspar-bearing samples.

Statistical Analysis
Geoscientists commonly use the mean and standard deviation to describe the compositional variation in major-element geochemical data sets.However, of the various methods used to assess "average" values, the mean and standard deviation consistently perform the worst because they are the most susceptible to outliers (Rock, 1988;Woronow and Love, 1990).Therefore, we used the median and the interquartile range as the most representative measures of the average and the spread of the whole database, as well as subsets of data.When assessing possible grade-related variations, the mean is additionally shown for comparison (Fig. 3).Compositional data are inherently multivariate, have a constant sum, and in turn are constrained by the closure problem (Rollinson and Pease, 2021).To ensure that correlations observed in our grade-related assessment did not result from the interdependence of weight percent oxide values (Chayes, 1960), we transformed compositional data using log 10 compositional mass ratios (Aitchison, 1982).Drawbacks of this approach are that a choice must be made for the conserved element in the ratios, and that log ratios are sometimes difficult to interpret.We found that trends between log ratio and weight percent oxide graphs were similar (Fig. S3), and thus we only show weight percent oxide values in Figure 3.

COMPOSITIONAL VARIABILITY AFM Diagrams
Of the 5729 analyses, 409 plotted at anomalously low or high A Ms values (>1.0 or <−0.4; ), and those are not shown in Figure 2A.Most analyses plotted between X Mg proj = 0.30-0.55[ X Mg proj = projected molar MgO/(MgO + FeO total )] and A Ms = 0.0-0.4,with our median worldwide pelite plotting at X Mg proj = 0.42 and A Ms = 0.19 (Fig. 2A; Table 1).There was a strong clustering midway between the AFM plotting positions of garnet and chlorite in the AFM diagram and no clear distinction between high-and low-Al pelites (cf.Spear, 1993, his figure 10-3; see Fig. 2A

Ferric Iron
There were 1964 analyses in the database for which FeO was measured using titration, permitting an estimate of the whole-rock ferrous/ferric iron ratio.Figure 2B shows that most analyses plotted between X Mg * = 0.30-0.55[X Mg * = molar MgO/(MgO + FeO)] and

Regional Analysis
In order to determine if it is justified to use the above median pelite composition to compare P-T conditions in different metamorphic belts, we calculated median compositions for 11 regions and/or orogens in the database for which there were more than 100 analyses.Most of these regional median values clustered together, with compositions similar to the worldwide median pelite (Fig. 2; Table 1).Exceptions included the Moine (Scotland) and Sanbagawa (Japan) regions, which had lower A Ms values, the Buller (New region, which had higher X Mg , and the Alpine (Austria, Italy, and Switzerland) region, which had higher X Fe3+ .

Metamorphic Grade
Previous authors have reported a decrease in both volatile content and X Fe3+ with increasing metamorphic grade (Shaw, 1956;Ague, 1991).We found a consistent decrease in LOI across all nine zones, and a decrease in X Fe3+ from the diagenetic to biotite zone (Fig. 3).In order to assess grade-related variations in the other major elements, analyses were normalized to 100% on a volatile-free basis with all iron converted to FeO.With a few exceptions, the major elements showed little consistent variation with metamorphic grade.Pelites from the porphyroblast and subsolidus sillimanite or K-feldspar zones showed lower median SiO 2 contents and higher median TiO 2 and Al 2 O 3 contents compared to higher-and lower-grade zones.Median MnO contents of the garnet zone were elevated compared to other zones.Median K 2 O contents were relatively constant across metamorphic grade.Samples from the diagenetic, anchizone, chlorite, porphyroblast, and subsolidus sillimanite or K-feldspar zones gave higher median A Ms values than those of the biotite, garnet, migmatite, and garnet-cordierite zones (Fig. 3B).

DISCUSSION
While pelitic compositions occur over a range of AFM values, there is a strong clustering of analyses at X Mg proj ∼0.4 and A Ms ∼0.2, with no separation between high-and low-Al pelites (Fig. 2A).AFM values for most regions cluster together (Fig. 2; Table 1), implying that it is justified to compare the inferred P-T conditions of commonly occurring metapelitic mineral assemblages between different orogens.An unexpected result of our study is the relatively high median whole-rock X Fe3+ of 0.23, given that X Fe3+ in oxides and silicates in metapelites other than hematite, magnetite, and muscovite are lower than this value, and muscovite has low absolute Fe 3+ (Forshaw and Pattison, 2021).This question requires further study.
Previous studies differ concerning whether there is significant bulk compositional change in major elements as a function of metamorphic grade: Some authors have argued for progressive change in bulk composition with increasing grade (Lapadu-Hargues, 1945;Ague, 1991Ague, , 1997)), while others have argued that metamorphism is essentially isochemical apart from the loss of volatiles and reduction of iron (Shaw, 1956;Vidale, 1974;Atherton and Brotherton, 1982;Walther et al., 1995;Moss et al., 1996;Stepanov, 2021).While we found a continuous decrease in volatile content across the full range of metamorphic grades, median X Fe3+ only decreased markedly from the diagenetic to biotite zones and remained relatively constant at higher grades (Fig. 3).After accounting for these differences by normalizing analyses to 100% on a volatile-free basis with all iron converted to FeO, there were still distinct differences in concentrations of elements for certain metamorphic zones.Examples include lower median SiO 2 and higher median Al 2 O 3 and A Ms for the porphyroblast and subsolidus sillimanite or K-feldspar zones, and higher median MnO in the garnet zone.
Ague (2011,2017) demonstrated that mass transfer is an important process in domains of high fluid flux and more minor away from these domains.While our database excluded samples described as or interpreted to have been affected by metasomatism, the low SiO 2 values in the porphyroblast and subsolidus sillimanite or K-feldspar zones could result in part from the increase in silica solubility with metamorphic grade (Manning, 1994).Alternatively, differences may be explained by sampling bias, in which especially aluminous layers containing abundant, petrologically significant porphyroblasts may have been preferentially sampled by petrologists in the field (e.g., Walther et al., 1995;Stepanov, 2021).The preferential sampling by metamorphic petrologists is perhaps best exemplified by the following quote: "In col-lecting and sectioning pelitic rocks, minimum variance specimens were emphasized, so that samples were biased in favor of rocks containing several of the phases staurolite, garnet, sillimanite, andalusite, and cordierite" (Holdaway et al., 1982, p. 574).
Despite the above considerations, the size of our database and the fact that it incorporated a wide range of metamorphic grades and geographic locations mean that analyses were not skewed toward a single metamorphic zone.Therefore, phase diagram modeling using the median worldwide pelite presented in this paper (Table 1) should allow broad comparison of the P-T conditions of different pelitic mineral assemblages and of different metamorphic terrains.
Figure 2. (A) 5320 analyses plotted on Al 2 O 3 -FeO-MgO (AFM) diagram after projection from muscovite and other phases (+).(B) Plot of X Fe3+ versus X Mg * for 1964 analyses for which FeO was measured using titration.Median worldwide pelite composition (star) and median compositions from different regions and/or orogens are shown.Database analyses were colored using the "batlow" color scheme (Crameri, 2021) according to probability density estimates (PDE).Mineral abbreviations are after Warr (2021).

Figure 3 .
Figure 3. Changes in composition as function of metamorphic grade.(A) Elemental oxides and loss-on-ignition (LOI) in weight percent.(B) A Ms , X Mg , X Fe3+ , and X Mg * .For each category, box shows median and interquartile range, white circle represents the mean value, and error bar denotes the standard error (2σ).See Figure 1 for zone abbreviations.
This work was supported by the Natural Sciences and Engineering Research Council of Canada (Discovery grant 037233 to D.R.M. Pattison) and the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation program (grant 850530).We acknowledge the work of the many petrologists who obtained the whole-rock chemical analyses that were used in this study.Jay Ague, Brendan Dyck, and Frank Spear are thanked for insightful reviews that helped to improve the manuscript.

TABLE 1
All iron is shown as FeO total with volatiles (loss-on-ignition [LOI], H 2 O, CO 2 , and SO 3 ) removed and values renormalized to 100%.X Mg , X Fe3+ , X Mg * , A Ms , and X Mg proj are defined in the text; n-number of analyses.Number in parentheses is the number of analyses with measured FeO and Fe 2 O 3 .See the Supplemental Material (see text footnote 1) for locations of each region.