Petrogenesis of siliciclastic sediments and sedimentary rocks explored in three-dimensional Al₂O₃–CaO^*+Na₂O–K₂O–FeO+MgO (A–CN–K–FM) compositional space

In certain circumstances, such as when source rocks are in contact with, or adjacent to, derived sedimentary material direct assessment of the source is possible. This provides the greatest capacity to pair source and sediment compositions and document the processes that drive compositional departures. As established in Nesbitt and Young (1984, 1989) and Fedo et al. (1995) quartzo-feldspathic igneous rocks (and metamorphosed equivalents) ranging in composition from granite to tonalite have CIA values of approximately 50, which means that regardless of starting composition fresh rocks have the same starting value from which to compare subsequent alteration from weathering through diagenesis. However, in 3D A–CN–K–FM space, substantial whole-rock compositional variation controlled by the FM component becomes apparent. These variations are reflected in the MIA changes in igneous source rocks that are not captured by the CIA (see Babechuk and Fedo, in press).

Figure 4 illustrates a range of igneous rocks of differing compositions (points 1–5) similar to the range used in Fedo et al. (1995). In both the perspective (Fig. 4a) and nadir (Fig. 4b) views, the increasing amount of ferromagnesian mineral components from granite to diorite forms a trend towards the FM pole. As expected, basalt (point 1) plots closest to the FM pole. In both views, gray lines connect the FM pole and the data point (red-filled circles) in the tetrahedron to their projected positions along or near the feldspar join (open circles; especially in Fig. 4b) on the back side of the A–CN–K face. This sample trend towards FM in the tetrahedron somewhat resembles the one apparent in the A–CNK–FM ternary plot whereby igneous rock fall close to a line joining the composition of feldspars (halfway along the A–CNK join) to the FM pole. What is clear is that rocks with increasingly abundant ferromagnesian minerals show the greatest loss of compositional fidelity on the A–CN–K triangle. By contrast, igneous sources that are the most felsic (point 5) have the least amount of ferromagnesian phases such that the data in the tetrahedron plot close to the A–CN–K face.

A core aspect of the CIA (Nesbitt and Young 1982) and MIA (Babechuk et al. 2014) concepts and derivative ternary diagrams has been their application to deciphering weathering/paleoweathering and associated climate/paleoclimate conditions. A well-characterized, Holocene, deeply weathered profile is developed on top of the Toorongo granodiorite in southeastern Australia (Nesbitt and Markovics 1997). Fifteen samples originating at an unweathered corestone and going outward to material filling joints show CIA values that range from 50 to 95, indicative of incipient-to-advanced chemical weathering.

All data from the Toorongo weathering profile are plotted in Fig. 6, with five specific points (1–5) noted to show how composition varies from fresh to weathered. In A–CN–K–FM tetrahedral space, it is clear the unweathered granodiorite (point 1, filled blue circle) has a substantial FM component, lying well within the tetrahedron. Projecting this point (open blue circle) onto the A–CN–K face reveals that the CIA has the expected value of 50 (Nesbitt and Young 1984; Fedo et al. 1995) and sits on the feldspar join. Points 2–4 within the tetrahedron represent samples increasingly moving away from the unweathered source along a systematic linear trend directed mainly away from the CN pole that shows increasing residual accumulation of aluminum (formation of kaolinite) with the loss of calcium and sodium (Nesbitt and Markovics 1997). The pathway of samples between points 2–4 also shows a minor shift away from the FM pole, representing a very minor component of Mg loss alongside Ca and Na. However, most Mg is retained in vermiculite (similar to K in K-feldspar and illite) during these early weathering stages (Fig. 1a; Nesbitt and Markovics 1997). At the transition between points 4 and 5, approximately corresponding to the stage of near-complete Ca and Na loss, there is a shift in the sample trend to one of a steeper slope upwards towards the A apex. This shift is dominated by a vector away from the FM pole (Mg loss) but also has a minor shift away from the K pole.

With all of the data points projected (gray lines) onto the A–CN–K face (open red circles), the samples form the expected linear array (noted by red arrows, Fig. 5) from plagioclase hydrolysis during early to intermediate weathering stages, as well as the inflection towards the A pole after calcium and sodium exhaustion and subsequent potassium loss from alkali feldspar and illite hydrolysis. However, the A–CN–K projection neglects that the advanced weathering stage also includes the accumulation of Al from Mg loss (due to more advanced weathering of early formed vermiculite) in addition to the minor component of K loss (loss from illite) as two contributing pathways to kaolinite formation in addition to that formed after plagioclase. In contrast, the tetrahedral plot shows a singular compositional space whereby all of the major labile components removed during incipient to advanced weathering can be evaluated collectively, better revealing the advanced secondary mineral transformations that contribute to Al accumulation in the profile.

As pointed out in Nesbitt and Markovics (1997), a strict interpretation of major-element trends as arising from in situ mineral transformation at the advanced stages of weathering (near point 5) is not warranted due to the translocation of some highly aluminous clay minerals. This would result in some of the trends in A–CN–K–FM tetrahedral space (and A–CN–K space) possibly arising from what could be considered a mixing line between more advanced (kaolinite) and less advanced (vermiculite/illite) clays. Such an inference is also identified in the A–CN–K–FM tetrahedron in combination with mass-balance considerations and knowledge of Fe behavior. The fresh Toorongo granodiorite has an MIA_(R) value of 35 (A value in the tetrahedron). If iron retention during oxidative weathering is either assumed or verified (as done in Nesbitt and Markovics 1997), then a maximum achievable MIA_(R) value via only the processes of element leaching can be calculated as the Al₂O₃/(Al₂O₃+FeO_(T)) ratio (see Babechuk and Fedo, in press). This maximum value, which is 69 for the Toorongo granodiorite (see reference plane at A = 69 in Fig. 5a), would reflect an advanced weathering substrate that is a mineralogical mixture of Al-oxides and Fe-oxides after complete Ca, Na, Mg, and K loss. The most highly weathered sample in the Toorongo profile (point 5) plots above this maximum at an MIA_(R) value of 82, which can only be explained with an additional process of Al accumulation beyond what can be residually enriched from leaching of other elements from the parent rock. Such values (higher than the aforementioned maximum) could also be explained by Fe loss, but evidence for this being the dominant factor was absent for the Toorongo profile (Nesbitt and Markovics 1997). As such, a further advantage of the A–CN–K–FM tetrahedral plotting approach is revealed whereby the higher plotting positions are best explained by combined physical and chemical processes. Further quantitative analysis of the relative role of each process is beyond the scope of this contribution, but we note that the requirement for physical kaolinite addition in the most weathered samples is not possible in the A–CN–K plot, which can achieve a maximum value of 100 via physical addition of kaolinite or in situ loss of CNK from other minerals. Consequently, this also exposes the caution of needing to evaluate other characteristics of a weathering profile (e.g., Al/Ti ratios alone, Fe speciation data) in combination with mineralogical and physical observations for effective use of all chemical weathering ternary or tetrahedral plots.

In summary, the use of the A–CN–K–FM tetrahedral plot allows compositional analysis that includes all main igneous silicate to secondary mineral transformations and can aid with using major-element data to support physical observations of mineral translocation. This is possible in a way that is not achievable with the A–CN–K with or without a companion A–CNK–FM plot due to the inclusion of all major-element components that contribute chemically to Al accumulation and Fe that generally follows an independent pathway to Fe-oxides in the same compositional space. Nevertheless, like in the A–CN–K ternary diagram, if guided by geological context, the major-element geochemistry in the A–CN–K–FM tetrahedral diagram still follows a systematic pathway indicative of a predictable set of circumstances resulting from hydrolysis (Nesbitt et al. 1997).

A Holocene weathering profile developed on the Baynton basalt is exposed in southeastern Australia (Nesbitt and Wilson 1992). This profile, in combination with other well-characterized basaltic profiles (e.g., Babechuk et al. 2014), provides context to evaluate the effects of chemical weathering on mafic bedrock. While less common for Earth, such context is particularly important for interpreting sediment (Ehlmann et al. 2017) and sedimentary rock (Mangold et al. 2019) data from Mars, whose crust is overwhelmingly basaltic in composition (McSween 2015).

Nine samples (sample set A1–A9) of the Baynton profile were analyzed by Nesbitt and Wilson (1992) from the center of an unweathered corestone outward to advanced weathered saprolite with CIA values ranging from 39 to 95. The data are indicative of progressive weathering from incipient to advanced stages. Assessing chemical weathering effects using CIA and ternary diagrams for mafic samples functions the same as with more felsic igneous sources, but with an understanding that mafic sources initially may be strongly depleted in both sodium and potassium in the source, which are critical in calculating CIA. Consequently, CIA values for many mafic weathering profiles are calculated with an understanding that calcium and aluminum primarily control the value. In contrast to behavior in felsic rocks, Mg is a more significant labile component for mafic rocks, which was a major justification for expanding the elements in the CIA to formulate the MIA (Babechuk et al. 2014; Babechuk and Fedo, in press). The MIA_(R) values that project into the A–CN–K–FM tetrahedron range from ∼20 to 22 in the fresh Baynton basalt to 43 in the most weathered sample.

The nine samples for the Baynton profile are plotted in A–CN–K–FM tetrahedral space (Fig. 6), where unweathered basalt (point 1) is shown with a filled blue circle. Two points are identified that highlight compositional changes from the intermediate (point 2) advanced (point 3) parts of the profile. As expected for compositions with a considerable amount of ferromagnesian components, the source and profile samples plot deep within the tetrahedron near the FM pole (Fig. 6), corresponding to lower MIA_(R) values than for that of the Toorongo granodiorite (Fig. 5). Similar to the Toorongo profile, the Baynton profile samples define a systematic path away from the source rock that is largely controlled by calcium (and lesser sodium) loss accompanied by residual aluminum accumulation through the formation of clays during hydrolysis. The loss of FM component across the lower part of the weathering profile, where calcium and sodium loss dominates (between points 1 and 2), is generally minimal apart from the first samples away from the source that show a greater deflection away from FM in the tetrahedron (below point 2, Fig. 6a). This deflection has also been noted in A–CNK–FM ternary space (Nesbitt and Wilson 1992) and was interpreted by McGlynn et al. (2012) to represent an initial loss of olivine followed by the domination of clay production at the expense of aluminous silicates like pyroxene and plagioclase. Rather than a general expectation in basaltic profiles, this early FM component loss is likely to be a site-specific (i.e., mineralogically and/or texturally controlled) feature as it is not recognized in other basaltic weathering profiles (e.g., Babechuk et al. 2014). Between points 2 and 3 (sample A9 in Nesbitt and Wilson 1992), there is a transition in the sample trend in the A–CN–K–FM tetrahedron that corresponds to enhanced Mg loss after near-complete Ca and Na loss.

Projection of the samples from source (blue) through profile (red) samples to the A–CN–K face (red open circles) produces the expected chemical weathering trend near the A–CN join ending in the very high CIA value of 95 (point 3). However, by working only with the A–CN–K plot and neglecting the FM component, it is possible to reach the conclusion that the profile is much more chemically weathered and that there has been much more residual accumulation of kaolinite than is realistic. While smectite does drop in abundance relative to kaolinite in the most weathered samples (Nesbitt and Wilson 1992), the tetrahedron better reveals the important role of the FM component, which in this case is controlled by both the development of Fe-oxides and the retention of Mg in smectites.

Following a similar exercise as for the Toorongo granodiorite, the maximum MIA_(R) achievable by leaching of labile Ca, Na, Mg, and K in the Baynton profile is ∼50 (see reference plane at A = 50 in Fig. 6a). The final plotting position of point 3 at an MIA_(R) of 43 is consistent with the following observations: (1) all of the products being an in situ chemical residuum (i.e., no physical clay mineral translocation), (2) the retention of iron in Fe-oxides during weathering, and (3) the appreciable amount of Mg remaining in smectites (Nesbitt and Wilson 1992). The latter minerals contribute to the mineral-chemical budget controlling Al in addition to kaolinite, but their influence on whole-rock compositions is only exposed if the FM component is included. If Fe behaves conservatively and concentrates in secondary minerals in parallel with Al, the whole-rock composition of the final weathered product evolves towards a mixture of both Al and Fe phases, but the CIA and A–CN–K plot captures only the aluminous secondary component. Use of the mafics ternary diagram (A–CNK–FM; Fig. 1c), and additional more newly developed ternary diagrams with a FM component (A–CNKM–F, AF–CNK–M; Babechuk et al. 2014), helps to resolve the Fe-redox and Mg behaviour in combination with the Ca, Na, and K from the CIA and A–CN–K diagram. However, tetrahedral compositional space provides the advantage of retaining more correspondence of individual poles to specific processes during mafic rock weathering and post-depositional alteration, which becomes particularly important when examining for later diagenetic/metasomatic effects in lithified paleosols (e.g., Babechuk and Fedo, in press).

In many circumstances, sedimentary rocks have been detached from their source for several potential reasons, including (1) the sedimentary succession has been later structurally severed from the basement terrane (e.g., Yan et al. 2003), a common feature in fold-and-thrust belts; (2) the source region may have been thousands of kilometers distant and never really connected to the depositional basin (e.g., Rainbird et al. 1992; Muhlbauer et al. 2017); or (3) the source has been deeply eroded (peneplain) and subsequently buried by younger strata (Japsen et al. 2016). The end result is that the sedimentary rocks of interest may have no obvious connection to their source region, leaving that to be reconstructed from the sedimentary rocks themselves.

Except in certain environments (e.g., glacial), siliciclastic sediment is typically drawn from weathering profiles rather than directly from bedrock sources prior to entering the transport system. Nesbitt and Young (1984) and Nesbitt et al. (1997) demonstrated that a systematic pattern from fresh bedrock to the top of the profile occurs in A–CN–K and A–CNK–FM space (Figs. 1b, 1c). A consequence of this relationship is that sediments, particularly clay through fine sand, drawn from such a profile may lie along this trend so that a line fit through the data will project back to the feldspar join at the position of bulk source composition in A–CN–K space (e.g., Schoenborn et al. 2011).

A global survey detailing the suspended-load sediment geochemistry from almost 20 large rivers on Earth shows a remarkable pattern similar to what would be expected if having been sourced from a complete weathering profile (Fig. 1 in McLennan 1993). CIA values for the rivers range from 50 to 95. Figure 7 shows these data in the A–CN–K–FM tetrahedron (closed red circles) and projections of the data to the A–CN–K face (red open circles), including a fit through the data (red arrow) that projects back to the feldspar join. The projected position for granodiorite (open blue circle) is plotted as a proxy for the average upper continental crust. Given that the rivers sampled in this compilation scatter across latitudes and climates, there is little surprise that the data appear to represent all degrees of chemical weathering. Furthermore, on the A–CN–K face, the intersection of the approximate fit line through the data and granodiorite reinforces the concept of a present-day upper crust is dominated by granodiorite on average. However, there is some scatter on the A–CN–K face with data on both sides of the line projecting to granodiorite, particularly at higher CIA values. Viewing the river data within the A–CN–K–FM tetrahedron (Fig. 7), more compositional scatter in the data becomes particularly apparent, with many river examples plotting much closer to the FM pole than would be expected for weathering of a granodiorite. Such positions likely require the involvement of non-weathering processes that concentrate ferromagnesian phases and result in whole-rock compositions that can affect the determined CIA value independent of chemical weathering.

We highlight the positions of four rivers (point 2 = Columbia, point 3 = Indus, point 4 = Nile, point 5 = Niger) to illustrate the complexity in connecting increasing CIA and prevailing climate conditions. Across this span of rivers, the CIA ranges from a low of 57 (Columbia) to a high of 95 (Niger), with the Indus (66) and Nile (76) sitting in between. The difference in CIA between the Columbia (∼45°–52° N) and Niger (∼4°–16° N) rivers makes sense given their prevailing boreal-warm humid and equatorial-arid climates (Beck et al. 2018) at higher and lower latitudes, respectively. At ∼31 wt.%, the Al₂O₃ in the Niger River is nearly double that of the Columbia (∼17 wt.%) consistent with intense chemical weathering in the Niger watershed.

Both the Indus and Nile rivers (points 3 and 4, Fig. 7) have CIA values indicative of more chemical weathering than the Columbia (point 2), yet in terms of their Al₂O₃ all are very similar (17–19 wt.%); however, in terms of FM components they are comparatively much different, which shows up in their positions in A–CN–K–FM tetrahedral space (Fig. 7). Unlike the Niger River (point 5), which has the highest CIA value in this assessment (CIA = 95) and approximately lies along a predicted granodiorite weathering trend (light blue dashed arrow), the Nile River (point 4) has a higher CIA value, but plots deep within the tetrahedron (enriched in FM) in a position far removed from the reference granodiorite weathering trend in this tetrahedral space. Thus, the relationship between source rock, extent of total element loss during weathering, residual aluminous clay abundance, and climate is not as clear.

We explore three possibilities to explain why the data scatter away from the granodiorite weathering trend on the A–CN–K triangle. First, if the addition of FM component occurred in the form of adding chlorite or vermiculite (ferromagnesian clays with no CNK component), which projects to the A pole on the A–CN–K triangle (CIA = 100), then it can ultimately lead to an increase in CIA along a mixing trajectory towards the A pole independent of weathering extent. Second, concentrating biotite can produce a trend towards FM in the A–CN–K–FM tetrahedron that would project as a trend away from a granodiorite weathering line on the A–CN–K triangle and towards the A–K join. The result would be generating scatter to the right of a granodiorite weathering trend (see Fig. 3). Third, adding a non-aluminous ferromagnesian phase (e.g., Fe-oxide), which is highly likely in fluvial sediment, would add more FM component and produce trends in tetrahedral space that do not correspond to a change in CIA because it would be equivalent to a mixing line between the A–CN–K face and FM pole. The tetrahedral plotting space provides the ability to investigate these possible hydrodynamic effects in a manner that can potentially relate climate/weathering information to the amount of pure ferromagnesian mineral availability/mixing.

Plotting in A–CN–K–FM tetrahedral space illustrates that there is much less overall geochemical coherence between the world river sediments when compared to their positions on the A–CN–K ternary diagram. Examining trends in the A–CN–K plot alone cannot differentiate between the three possible causes related to the FM component. The tetrahedral plot provides a solution that better relates all major element components together. Although a connection to a generalized “granodioritic” source is consistent with the totality of the sediment data, sufficient geochemical variability exists to suggest that exposed crust includes mafic rocks as is expected and that sedimentary processes, such as mineral sorting and concentration during transport along with any and minerals precipitated in the transport environment, considerably increase data scatter within the tetrahedron. The purpose of this exercise is not to explicitly identify the specific processes in the major rivers generating the data scatter. Instead, what is revealed is that each individual river and samples from within each river should be evaluated independently, and using more textural (e.g., grain size) and elemental/mineralogical components in the sedimentary system, to isolate controls that cannot be identified completely using 1D values or 2D ternary diagrams.

Sediments derived from felsic-to-intermediate continental bedrock and recycled sedimentary sources are being transported >60 km along the Genoa River to the Tasman Sea in an area experiencing advanced chemical weathering in southeastern Australia (Nesbitt et al. 1996). Eight samples of mud and nine samples of sand were collected from the headwaters area down to the Mallacoota estuary (Nesbitt et al. 1996), providing a comprehensive collection of material along the course of the river. That mud samples have a higher CIA and plot closer to the A pole than corresponding sands in the A–CN–K ternary plot (Figs. 8a, 8b, open pink and purple circles), led Nesbitt et al. (1996) to conclude that mud samples contain more secondary aluminous phases relative to sand, an expected conclusion with sub-sand-sized clay minerals being concentrated as suspended load during transport.

Mud and sand data from the Genoa River are plotted as filled pink circles (mud) and filled purple circles (sand) in the A–CN–K–FM tetrahedron in Fig. 8. Out of the total plotted data, four specific samples, two sand (points 1, 2) and two mud (points 3, 4), have been highlighted to illustrate the importance of assessing the data with the A–CN–K–FM tetrahedron. Sand samples have a minimal compositional range in terms of Al₂O₃ from about 6.5–9 wt.% and virtually identical abundances of CaO, Na₂O, and K₂O, indicating there should be minimal variation in the degree of chemical weathering, and consistent with the inferred loss of Ca-plagioclase controlling changes in sand composition (Nesbitt et al. 1996). The lack of increasing aluminous components is particularly evident in the perspective view of the tetrahedron (Fig. 8a) where point 1 represents the point closest to the A–CN–K face and point 2 follows a linear path at about the same A value towards the inside of the tetrahedron. The linear trend is also clearly seen in the nadir view (Fig. 8b). Mud samples, bracketed by points 3 and 4, plot more as a group as seen in the nadir view (Fig. 8b), but like with the sands have relatively consistent Al₂O₃, CaO, Na₂O, and K₂O abundances, suggesting that weathering intensity should not vary much within muds nor be significantly different from sand because these components do not differ significantly across the data set. However, as seen on the A–CN–K face, where A = CIA value, sands range from A = 53 to 62 and muds range from 63 to 71, consistent with the notion of clay formation during moderate chemical weathering followed by sorting into sand and mud fractions.

Linking the data in the tetrahedron with the corresponding data for only the A–CN–K face shows how the apparent discrepancy between samples possessing relatively consistent Al₂O₃, CaO, Na₂O, and K₂O abundances and yet different CIA values may lead to potentially overestimated inferences of source weathering. One aspect that changes dramatically within the sand and mud samples, and collectively for all samples, is the progressively increasing FM component, which was observed on the A–CNK–FM ternary diagram as a horizontal line (Nesbitt et al. 1996). Sample points 1 and 2 represent the minimum and maximum CIA values for sand; sample points 3 and 4 represent the minimum and maximum CIA values for mud on the A–CN–K face. The importance of the A–CN–K–FM tetrahedral plot in further evaluating this trend is apparent through the horizontal trend in the tetrahedron (especially apparent in perspective view, Fig. 8a) being a significant departure from an igneous rock weathering trend that climbs on a slope towards the A apex (e.g., Fig. 5). The horizontal trend instead likely indicates that there is a strong control from the physical mixing of an aluminous FM component that increases in abundance from sand to mud. Alternatively, the A component value could be matched by the mixing of a non-aluminous ferromagnesian component (e.g., Fe-oxides) with more aluminous clays, which would have the net effect of lowering the A component (MIA_(R)) in the tetrahedron. Mineralogically, kaolinite is documented in these sediments (Reinson 1973; Nesbitt et al. 1996), which indicates the role of chemical weathering as an influence increasing the CIA. However, vermiculite–smectite still dominates the mineralogical budget (secondary minerals after biotite that have retained most of their Fe and Mg budget) in the highest CIA samples. These aluminous ferromagnesian minerals, which have lost some FM component relative to original igneous FM minerals (amphibole, biotite), can increase the Al slightly (and thus the CIA), but when examining only A–CN–K trends the role of kaolinite accumulation and chemical weathering could be overestimated. In other words, the mafic component weathered from the source granite (amphibole, biotite) preferentially produces clay-sized particles and (or) mafic mineral-derived clays concentrate preferentially via sorting in the finer fraction of the sediment relative to the feldspar and quartz in the sand fraction.

This consequence also raises the capacity to resolve between steady- and non-steady-state weathering (Nesbitt et al. 1997). Samples that record non-steady-state weathering are derived from multiple levels in weathering profiles and so show compositional variation that mimics the path for a weathering profile (e.g., Fig. 5). In the Genoa River example, data form a line on the A–CN–K face consistent with non-steady-state weathering, but this line does not make it apparent that some of the aluminum accumulation may be the result of mixing in minimally weathered aluminous ferromagnesian phases rather than sampling different parts of a weathering profile (Figs. 1a, 5).

Evaluating this concept further, we plot data from bulk and sieved fractions of Genoa River sample GR5 in expanded perspective (Fig. 8c upper) and nadir (Fig. 8c lower) views. The sieve fractions represent detritus that is coarser and finer than 63 µm, which represents the cutoff between silt and sand. Compositions for data plotted in the A–CN–K–FM tetrahedron are filled pink circles, while compositions plotted on the A–CN–K ternary diagram are in open pink circles; the bulk GR5 sample composition is represented by the larger circle and sits between the two sieved fractions. The data on the A–CN–K face show that the finer sieve fraction has a much higher A value (CIA = 78) relative to the sand fraction (CIA = 58), suggestive of more highly weathered clay phases accumulating in the finest fraction and being compositionally differentiated because of grain-size sorting (Nesbitt et al. 1996). However, the positions of the sieve fraction within the tetrahedron clearly show that while there is some increase in the A component in the <63 µm fraction, there is also a considerable increase in the FM component (particularly evident in Fig. 8c upper) relative to the >63 µm fraction. Plotting the data in the tetrahedron demonstrates that the much higher CIA in the <63 µm fraction corresponds also to increasing FM, which is likely explained by a combination of preferential accumulation of Fe-oxide (hematite documented in samples) and aluminous ferromagnesian minerals (vermiculite–smectite). In the case of the Genoa River, concentrating ferromagnesian phases in the finer grain sizes plays a vital role in compositional differentiation, which then carries significant consequences for interpreting weathering conditions using CIA. These relationships are best visualized and interpreted using the A–CN–K–FM tetrahedron.

Holocene sediments located within the ∼8 km² Guys Bight basin on Baffin Island, Nunavut, Canada, were derived from crustal sources (consistent with a bulk granodiorite composition) that are presently undergoing extreme comminution as a result of glaciation (Nesbitt and Young 1996). After emanating from glacial termini, sediments enter a fluvial-dominated transport system that is capable of naturally sorting detritus into separate gravel, sand, and mud accumulations. Given the polar climate and overall compositional immaturity of the sediment, Nesbitt and Young (1996) concluded that chemical weathering is at a minimum in this basin and specifically chose this location to study the effects of abrasion and sorting in the absence of significant chemical weathering.

Fifty samples representing six grain-size fractions (silt = purple, mud = blue, fine sand = light blue, medium sand = teal, coarse sand = orange, gravel = red) are plotted in the A–CN–K–FM tetrahedron (filled circles) and projected on to the A–CN–K face (open circles, Fig. 9). Two sample points (1, 2) are highlighted as reference guides. Finer grain sizes are shown in cooler colors; coarser grain sizes are shown in warmer colors. On the A–CN–K face, all data regardless of grain size plot as a tight, overlapping, data cluster (Fig. 9a) with A (CIA) values just above 50, and a slight spread along the feldspar join (Fig. 9b). The simplest interpretation of the data on the A–CN–K face is that the sediments have approximately the same composition and that chemical weathering is minimal. Furthermore, despite tremendous mechanical breakdown, there is little noticeable compositional change.

Including the ferromagnesian component to make the A–CN–K–FM tetrahedron, the data plot along a linear array where coarser grained deposits sit closer to the A–CN–K face and where finer grained deposits show an increasing FM component, as also previously identified in the A–CNK–FM ternary diagram (Figs. 9a, 9b; Nesbitt and Young 1996). There appears to be an inflection in the data separating two sub-parallel trends defined by grain size: (1) silt-fine sand and (2) medium sand-gravel (Figs. 9a, 9b). Figure 9c shows part of the tetrahedron from the nadir view with only the finer grain sizes in the upper plot and coarser grain sizes in the lower plot. A line passing through the finer grained samples (gray arrow) points to biotite in the tetrahedron, strongly suggesting that it concentrates in the finer fractions (Fig. 9c upper) as Nesbitt and Young (1996) and Young and Nesbitt (1998) previously identified. A line passing through the coarser grained data (Fig. 9c lower) defines a different trend more focused on the different proportions of mixed plagioclase and alkali feldspar. Dashed gray lines in both images represent the orientation of the second data set as a reference guide. Details in sediment composition as a result of sorting are clearly evident in the A–CN–K–FM tetrahedron and contrast with a pure chemical weathering trend from granodiorite (Fig. 5) and the mixed chemical weathering plus sorting example from the Genoa River (Fig. 8). In the Guys Bight basin example, the data scatter in a line almost parallel to the projection lines from the FM pole or the composition of biotite solely because of sorting, essentially forming mixing lines with biotite and (or) Fe-oxides as an endmember. In the absence of chemical weathering, data essentially cluster as a single group on the A–CN–K face with a CIA value near fresh bedrock. Working in the A–CN–K–FM tetrahedron thus provides a clear way to identify both chemical weathering (if any exists) and mixing effects associated with specific mineral phases (e.g., biotite) by deflections towards a range of minerals with variable C–N–K–FM proportions.

Sediment derived from a sample of crushed, unweathered, Kilauea basalt was studied by Fedo et al. (2015) as a way to test mineralogic and geochemical variation across grain sizes as a proxy for size sorting. Although mafic point sources and derived sediment are not overly common on Earth, they do occur in a number of places, such as hot spot island chains, in large igneous provinces, and as part of Archean greenstone belt terranes. By contrast, studies of sediments and sedimentary rocks on Mars almost uniquely use basalt as source given its widespread nature as the principle composition of the crust (McSween 2015). As a result, investigating sediment composition trends from mafic sources is essential to understanding the range of compositional starting points.

A sample of original olivine phyric basalt (blue filled and open circles) and 13 samples of synthetic sediment (made by crushing; orange and red filled and open circles) are plotted in A–CN–K–FM tetrahedral space (Fig. 10). On the A–CN–K face, all the data plot as a very tight cluster, where sediments expectedly have the same A (CIA) value given the complete absence of chemical weathering (Figs. 10a, 10b). However, in the tetrahedron, there is a considerable amount of compositional variation that essentially lies on the blue line from the FM pole, through the bedrock point (blue-filled circle), and to the bedrock position on the A–CN–K face (blue open circle). This very coherent linear array results from the concentration of olivine, which plots at the FM pole, in the finer grain-size fractions (Fedo et al. 2015). Essentially, the composition of each grain-size fraction lies on a mixing line between the FM pole where olivine and iron oxides plot and whole-rock composition. In this example, as with the Guys Bight basin where chemical weathering is essentially absent, the data scatter does not result in a change of the CIA.

Closer inspection of the data array reveals that, similar to the naturally sorted deposits in the Guys Bight basin, grain size controls where compositions plot inside the tetrahedron. Inset plots show sieve fractions less than 1 mm as red-filled circles and fractions coarser than 1 mm as orange-filled circles (Figs. 10c, 10d). Coarser than 1 mm, sediment samples remain dominated by rock fragments that contain the essential mineralogy of the host bedrock with data making a cluster near the bedrock composition. Finer than 1 mm, sediment samples scatter much more significantly as individual mineral phases (phenocrysts of different compositions) break out of the host and reproportion elements leaving only glass and very small phenocrysts in the residual rock fragments (Fedo et al. 2015). We note these data only show the sieve fractions and not sorting driven by mineral density (e.g., Fig. 2C), which would drive further mineral segregation and spread in the data as recognized in Martian sediment (Siebach et al. 2017). Consequently, even in this simple example that explores the composition of sediment derived from crushing an olivine phyric basalt, enough variation exists to produce trends similar to naturally segregated sediment.

Rainbird et al. (1990) studied the geochemistry of an approximately 12 m thick paleosol developed on top of the Archean Ville Marie granite, which is exposed near the community of Ville Marie, Québec, Canada (Rainbird et al. 1990). In this location, a hematitic breccia followed by quartzites of the Paleoproterozoic Lorrain Formation (Cobalt Group, Huronian Supergroup) overlie the paleosol, which by nature of its age provides data about weathering conditions at the onset of the “Great Oxidation Event” (Gumsley et al. 2017) coupled with the important effects of diagenesis after burial by the Lorrain Formation and subsequent strata.

Twenty-three samples inclusive of unweathered bedrock, variably weathered saprolite, and overlying brecciated material show the extensive weathering and diagenesis preserved in the profile (Rainbird et al. 1990). Here we plot the 19 samples on the A–CN–K–FM tetrahedron (Fig. 11) inclusive of all zones except the brecciated material overlying the saprolite. An example of unweathered bedrock (point 1, blue-filled circle) shows the location of the granite, as well as its location (open blue circle) on the A–CN–K face following the projection from the FM pole (blue line). A true granite composition can be inferred from the plagioclase to alkali feldspar ratio on the feldspar join (Figs. 11a, 11b) and the near absence of FM components (Fig. 11a). One example from the top of the paleosol is highlighted for reference (point 2). Data for the paleosol samples projected onto the A–CN–K face (open red circles) describe a trend (solid red arrow) that deviates from the expected trend (red dashed line, also see Fig. 1b) for chemical weathering of a bedrock (e.g., Figs. 1a, 6) with the composition of the Ville Marie granite.

Projections (gray lines) of all the data from within the tetrahedron to the A-CN-K face are shown in the expanded plot of the nadir view (Fig. 11c). The deviation between the actual (solid red arrow) and expected (dashed red arrow) locations of data on the A–CN–K face is more clearly observed in the expanded view (Fig. 11c) and results from the diagenetic addition of potassium after burial (Rainbird et al. 1990; Fedo et al. 1995), which may occur at any time after burial, even at times tens to hundreds of millions of years after the known time of weathering (e.g., Macfarlane and Holland 1991; Roscoe et al. 1992). The expanded plot (Fig. 11c) also shows that there is a substantial variation in FM components in the tetrahedron (red solid circles) that is lost when data are projected on the A–CN–K face.

In minimally weathered samples (close to fresh bedrock), FM components are depleted and enriched relative to fresh bedrock, both compositional changes that could be related to Fe mobility within the profile during weathering or later burial (Rainbird et al. 1990). Expanding the plot to focus on data from only the upper part of the profile (Fig. 11d; see inset in Fig. 11c for location) in the tetrahedron shows a path from lower (left side) to the top (right side) reveals a “loop” trend in successively higher samples in the profile. These samples plot in positions more complicated than would be expected from simple CN loss followed by K addition. Specifically, in the uppermost weathered samples (Fig. 11d), there is a minor but clear loss of FM component in the samples with the greatest K addition (e.g., point 2). This trend away from FM is also apparent in the A–CNK–FM ternary plot (Rainbird et al. 1990), but the tandem effect of K addition in the samples showing FM loss is not readily apparent in the ternary configuration where K is combined with CN. In contrast, the tetrahedron retains K on its own such that its more apparent how the FM loss and shift of samples downwards towards the K apex both contribute to the paleosol composition. The addition of K is a diagenetic/metasomatic effect that is well documented as a process producing deviations from predicted weathering trend arrays in the A–CN–K ternary diagram (see Fedo et al. 1995; Nesbitt and Young 1989). The advantage of using the tetrahedron is in the better identification of weathering trends associated with calcium, sodium, and magnesium loss prior to potassium addition by keeping K as its own component (unlike the A–CNK–FM ternary). Similar effects of K addition are apparent in other paleosols of the Huronian Supergroup, such as the Cooper Lake paleosol developed on a mafic source (Babechuk et al. 2019) and the use of tetrahedral plots to expose these combined weathering-diagenetic/metasomatic effects in mafic paleosols is further explored in Babechuk and Fedo (in press).

Mesoarchean (3 Ga) shales from the Buhwa Greenstone Belt, Zimbabwe, accumulated on the margin of the Paleoarchean Zimbabwe Archean Craton (Fedo and Eriksson 1996). Modeling of trace and rare earth elements led Fedo et al. (1996) to propose a mixed source dominated by felsic plutonic rocks (tonalite, granite) and lesser basalt, all common elements of granite-greenstone terranes. The stratigraphic succession in the variably metamorphosed Buhwa Greenstone Belt (Fedo et al. 1995) consists of interstratified quartz arenite and shale that transitions basinward to shale interbedded with banded iron formation (Fedo and Eriksson 1996).

Here we plot data from 17 shale (and phyllite) samples in the A–CN–K–FM tetrahedron (pink-filled circles), highlighting three samples (points 1–3) for reference. Projection lines (gray) from the FM pole show their locations on the A–CN–K face (pink open circles, Fig. 12). We also estimate the position for the mixed felsic-mafic source based on the modeling Fedo et al. (1996) in both A–CN–K–FM tetrahedral (dashed, filled, light-blue circle) and A-CN-K ternary diagram (dashed, open, light-blue circle) compositional spaces.

The projected shale data on the A–CN–K face all lie around the location of idealized muscovite (black open box), which is used as a proxy for idealized illite (Fig. 12). Points 2 (CIA = 72) and 3 (CIA = 79) bracket the maximum and minimum CIA values on the A–CN–K face, although both are the least aluminous of all the samples at ∼11 wt.% Al₂O₃ (note their positions in Fig. 12a). Based on the trace- and rare earth-element geochemistry (Fedo et al. 1996), the source was interpreted as dominated by tonalite that underwent extreme chemical weathering. Dashed blue arrow “a” shows a general expected weathering trend in the A–CN–K–FM tetrahedron (see Fig. 5) for a starting point approximating tonalite; dashed blue arrow “b” represents the same weathering trend but plotted on only the A–CN–K face. With plagioclase hydrolysis producing abundant kaolinite, not illite, as the major weathering end product (Nesbitt and Young 1984, 1989; Banfield and Eggleton 1990), both arrows “a” and “b” head towards kaolinite in Fig. 12.

“Present day” compositions for shale samples from the Buhwa Greenstone Belt lie on the line connecting the A and K poles on the A–CN–K face astride the composition for idealized illite (black open box, Fig. 12). Given that illite is not the primary weathering product of plagioclase, the present shale compositions were interpreted to be the result of diagenetic potassium addition to kaolinite to form illite (Fedo et al. 1996) at a time determined to be hundreds of millions of years younger than the time of deposition (Krogstad et al. 2004). Potassium enrichment attributed to this pathway has also been also been recognized in rocks from Earth (Yang et al. 2022) and Mars (Mangold et al. 2019) with mafic sources that initially have low abundances of potassium. The solid blue arrow “c” on the A–CN–K face traces a path from the estimated source position to the idealized illite position (open black box) in the middle of the shale data guided by the path of progressive potassium addition in the Ville Marie paleosol (cf., Figs. 11a–11c).

Plotting these data in the A–CN–K–FM tetrahedral helps reveal several important compositional features with significant implications for inferred mixing/sorting and degree of weathering. All but one shale sample (point 1) form a data array in the A–CN–K–FM tetrahedron that is approximately between the positions of the reference (not idealized) illite (Ilt; which projects to a CIA of ∼72) and chlorite (which projects to a CIA of ∼100) compositions (Fig. 12). Thus, while most of the samples cluster near the composition of idealized illite on the A–CN–K face, they have a significant compositional trend towards the FM pole (deep into the tetrahedron). The process of potassium addition to kaolinite, as implied when interpreting the data on the A–CN–K face, oversimplifies the likely diagenetic reactions. A direct path from kaolinite to the reference illite composition (Ilt, Fig. 12, black arrow) would also require addition of iron and magnesium either through diagenetic addition and (or) mixing with clays of chloritic composition for example.

Because the modeled source (Fedo et al. 1996) carries a mafic–ultramafic component, a more likely explanation of the shale data is that the samples never essentially reached the level of 100% kaolinite (plotting at the A pole, where both CIA and MIA_(R) = 100), but rather consisted of a mixture of kaolinite, other clays (e.g., nontronite, vermiculite, montmorillonite), and iron oxides (of different generations) from mafic phases in the tonalite and alteration of basaltic sources. Regardless, the amount of K₂O in the shale samples (average ∼6 wt.%) requires potassium addition via diagenesis. The mafic clays would be currently represented by chlorite and intermediate mixed-layer clay compositions between chlorite and illite. The interbedding of some shale samples with banded iron formation also raises the potential that data array lies on a mixing line between illite and the FM pole, which could result from iron addition at the time of deposition or from variable diagenetic addition of iron sourced from adjacent lithologies. However, apart from points 1 and 2, the three samples that plot deep in the tetrahedron lie on a mixing line between illite and chlorite (purple dashed line Fig. 12; originally kaolinite and smectites, respectively). Such a compositional mixture carries significance for the CIA values and their relationship to climate and extent of weathering. First, because chlorite plots with a CIA of 100 on the A–CN–K face, it becomes apparent that the composition of sample point 3 (highest CIA), which plots deepest in the tetrahedron, is best explained as a sample with the greatest mafic component in a mixture rather than the most weathered sample with a greater felsic component. Second, the original clay mineral array prior to metasomatism is likely to have retained a significant FM component despite both Fe and Mg being leachable from mafic materials under anoxic conditions. This would also be consistent with a lower extent of overall leaching from chemical weathering than would be assumed from the A–CN–K plot alone that does not accurately budget the fate of mafic minerals.

In summary, although plotting the shale data in the A–CN–K–FM tetrahedron does not fully resolve the exact mechanisms for generating the shale compositions, only considering their CIA values or positions on an A–CN–K triangle loses considerable compositional information needed to address their petrogenetic history. Further, as with the Ville Marie paleosol example, the use of a tetrahedron relative to the A–CNK–FM ternary diagram provides the ability to better assess the combined effects of provenance, sorting, and diagenesis by retaining K as a separate pole.