The raw geochemical data (Supporting Information Material S3) were studied to identify the element concentration patterns char-acteristic for Petit‐Chasseur ceramics. Most of the analyzed pot-sherds have SiO2content exceeding ~50 wt%, Al2O3~15 wt%, and CaO below 5 wt%, except for samples PC03 and PC04 (CaO ~22 wt
%, taken from the same FN vessel). The amounts of Fe2O3, MgO, and TiO2vary considerably (~2–10, ~0.5–4, and 0.3–2 wt%, respectively), whereas K2O and Na2O contents range between 0.5 and 5 wt% and 0.1 and 2.7 wt%. Minor amount of P2O5(<0.7 wt%) may be attributed to postdepositional alteration (Freestone et al., 1985; Schwedt et al.,2004). Considering a myriad of inclusions in analyzed ceramics consisting quartz, feldspar, and rock fragments (fabrics 1–7, 8, and 9;
Table2), the SiO2–Al2O3–Na2O–K2O versus CaO (Figure8a) and SiO2–Al2O3–Na2O–K2O versus Fe2O3–MgO–TiO2(Figure8b) plots were utilized to discriminate between studied samples. The outliers are marked by either higher CaO or Fe2O3and TiO2, which is cor-related with their respective petrographic groups impoverished in felsic component (Tables2and 3; Supporting Information Materi-alS3). Rare earth elements (REE) were found to highly discriminate the Petit‐Chasseur ceramics (Figure 8c). They range from 108 to 343 ppm, forming the two arbitrary defined groups: low‐REE and high‐REE ceramics (Figure8d).
The PCA based on major elements only shows the first three components to account for 80.2% of the total variance (40.7%, 25.3%, and 14.2%, respectively; Figure9a,b). The FN and some of the BB samples are impoverished in the felsic component (Figure8a,b) and clearly differ from the main projection field (Figure9a,b). This is due to their petrographic composition, featured by the calcite, the epidote‐rich granite, or the glaucophane schist present in the cera-mic paste (Table2). On the contrary, EBA potsherds and the re-maining BB samples displayed different fabrics, and are thus enriched in felsic component (Figure8a,b; Supporting Information Material3). Consequently, they cluster into the same projection area (Figure9a,b).
The information provided above accentuates the importance of inclusions, which effectively control the total geochemical budget of analyzed potsherds. This is attributed to a high abundance of aplastic inclusions (Figure4and Table2) and their relative chemical fertility TABLE5(Continued) SamplePC63PC63PC63PC32PC32PC32PC73PC73PC73PC73 MineralK‐ poor mica (H)
Muscovite(M)Illite(IL)Vermiculitizedmica(I)Vermiculitizedmica(I)Muscovite(M)Al‐rich stilpnomelane(J) Al‐rich stilpnomelane(J) Muscovite (smectitization) (K)
Muscovite (smectitization)(K) Spectrum454451462487493500159171166170 MnO‐‐‐‐‐‐‐‐‐‐ Na2O0.40.50.3‐‐0.5‐0.3‐‐ SO‐‐‐‐‐‐0.3‐‐‐ P2O5‐‐‐‐‐‐0.4‐‐‐ Total100100.1100100.1100100100100100100
with regard to an Si–Al‐based matrix. This becomes even more ap-parent when PCA is applied to a subset of samples for which the lithological assemblage is known (Figure9c,dand Table2). There, the first three components account for 80.9% of the total variance (40.9%, 24.0%, and 16%, respectively; Figure 9c,d). As previously noted, samples of the calcite group (fabric 9; Table2) and those rich in Fe‐bearing glaucophane schist and epidote‐rich granite (fabrics 8 and 5; Table2) are outliers. The oscillations in the amount of fine‐ grained granite with calcite infills define the patchy distribution of the pottery of the granite and quartz–feldspar gneiss groups (fabrics 2 and 6, respectively; Table2). Both of these groups are marked by petrographic heterogeneities in terms of the type and number of inclusions (fabric 2 var. 1 and fabric 6 var. 1; Table2), which, in turn, controls their whole‐rock geochemistry and scattered distribution in PCA biplots. The projection areas of other fabrics seem to be more homogenous (Figure 9c). The position of the weathered granite group (fabric 4; Table 2) is defined by CaO enrichment due to a secondary calcite infill of igneous aplastic inclusions. Projections of
the amphibole gneiss‐rich pottery (fabric 7; Table 2) are largely defined by a uniform abundance of their inclusions consisting of plagioclase, amphibole–pyroxene pseudomorphs, and epidote (Figure5). Finally, the projection of fine‐grained granite pottery with Fe oxide (fabric 7; Table2) along the first principal component has been defined by the steady increase of its Fe content (Figure9c,d).
The Petit‐Chasseur pottery serves as a good example of how dif-ferent aplastic inclusions control the overall geochemical makeup of ceramic vessels. This adds to the importance of temper and natural inclusions when it comes to pottery classification and reconstruction of the procurement of raw material (Brunelli et al., 2013; Day et al.,2011; Salanova et al.,2016).
The REE‐based PCA (84.6% of the total variation for the first three components) pointed out a lack of correlation between the inclusion types (i.e., fabrics) and REE abundances (Figure 10a,b).
This means that the trace element budget of analyzed pottery is essentially controlled by the clay matrix substrate (e.g., Michard,1989). This holds true for REE in particular, which tends to F I G U R E 8 Bivariate plots discriminating the Petit‐Chasseur ceramics: (a) SiO2–Al2O3–Na2O–K2O versus CaO; (b) SiO2–Al2O3–Na2O–K2O versus Fe2O3–MgO–TiO2; (c) SiO2–Al2O3–Na2O–K2O–CaO–Fe2O3–MgO–TiO2versus REE; (d) LREE versus HREE. BB, Bell Breaker; EBA, Early Bronze Age; FN, Final Neolithic; HREE, heavy rare earth element; LREE, light rare earth element; SEM‐EDS, scanning electron
microscopy–energy‐dispersive spectroscopy [Color figure can be viewed atwileyonlinelibrary.com]
be allocated to reactive surfaces of the clay minerals (e.g., Andersson et al.,2004;Šegvićet al.,2014). There is a relatively recent experimental study showing that temper effects on the total budget of trace elements in pottery are negligible (Munita et al.,2008). Moreover, it has also been demonstrated that even the firing process does not significantly alter the initial clay composition (Gutsuz et al.,2017). The PCA biplot with indicated sample pro-jections whose clay matrix had been investigated by SEM‐EDS of-fers further inferences (Figure10cand Table4). The matrices made of mica are intrinsically enriched in both LREE and HREE (Figure 10d; high‐REE group, Figure 8d), whereas the opposite holds true for those which are illite‐based (Figure 10d; low‐REE group, Figure8d). This is attributed to a higher charge of micaceous
phases that promote a retention of REE (e.g., Andersson et al.,2004), whereas lower charges and an ongoing smectitization of illite reduce its capability to attract REE (Honty et al.,2008;
Zanoni & Šegvić,2019). The behavior of high‐field strength ele-ments (HFSE) is analogous to that of REE (85.9% of the total var-iance for the first three components; Figure 10e,f); mica‐based matrix is richer in HFSE as compared with the one consisting of illite. The only exceptions are samples PC05 and PC73, which are rich in TiO2, though this does not stem from the matrix but rather makes up part of rock inclusions (Table2).
The normalization to the NASC allows for an inspection of ele-mental concentrations with respect to the standard (Figure 11).
Other than the major elements (Al2O3, Fe2O3, CaO, Na2O, K2O, and F I G U R E 9 Statistical treatments of geochemical data of analyzed pottery: (a) and (b) principal component analysis (PCA) performed on the entire ceramic data set considering major and minor element content—principal component biplot; (c) and (d) PCA performed only on the 42 samples for which the lithological assemblage is known, major and minor element content—principal component biplot [Color figure can be viewed atwileyonlinelibrary.com]
F I G U R E 10 (See caption on next page)
TiO2) and group‐specific elemental anomalies (e.g., Sc, Ba, Sr) con-trolled by aplastic inclusions and carbonate (Figure11c), the rest looks quite instructive. Thus, the trace element normalization curves look quite similar (concentrations/trends) regardless of the group, which is another line of evidence in favor of the clay matrix
controlling the total trace element budget. However, a relatively pronounced compositional spread in NASC diagrams (Figure11a,b) and a scattering in the PCA biplot (Figure10a) reflect a range of plausible sources from where the very similar clay (mica vs. illite) resources were procured.
F I G U R E 10 Statistical treatments of geochemical data of the 42 samples for which the lithological assemblage is known: (a) and (b) principal component analysis (PCA) performed considering only rare earth elements and trace element budget—principal component biplot; (c) location on the factor plane (a) of the samples for which the scanning electron microscopy–energy‐dispersive spectroscopy (SEM‐EDS) analysis has been executed; (d) scatterplot of light rare earth elements (LREE) versus heavy rare earth elements (HREE); (e, f) PCA performed considering only high‐field strength element budget—principal component biplot [Color figure can be viewed atwileyonlinelibrary.com]
F I G U R E 11 Multielement plots after North American Shale Composite (NASC) normalization: (a) granite‐rich pottery combined; (b) gneiss‐rich pottery combined;
(c) calcite fabric pottery combined