Probing crustal thickness evolution and geodynamic processes in the past from magma records: An integrated approach
Published:January 01, 2017
- PDF LinkChapter PDF
J. Ganne, W.P. Schellart, G. Rosenbaum, X. Feng, V. De Andrade, 2017. "Probing crustal thickness evolution and geodynamic processes in the past from magma records: An integrated approach", The Crust-Mantle and Lithosphere-Asthenosphere Boundaries: Insights from Xenoliths, Orogenic Deep Sections, and Geophysical Studies
Download citation file:
A new and simple integrated approach is proposed for qualitatively unravelling the crustal thickness of fossil magmatic systems based on the chemical and thermal records in amphibole-bearing magmatic rocks. Statistical analyses applied to a large multidimensional amphibole database show that Ti-rich and Si-poor magmatic amphiboles, which formed at high-temperature (T) conditions (>950 °C), were dominantly developed in basaltic to basaltic-andesitic (SiO2-poor, i.e., <55 wt%) magma within relatively thin crust (5–10 km). We find that for crustal thicknesses larger than 10 km, the occurrence of high-T amphiboles and basaltic magma decreases with increasing crustal thickness. This is because of mineral filtering in “mature” deep crustal hot zones that occur at the crust-mantle boundary (Moho). Given that subducting plates exert a direct control on the structural evolution (shortening or extension) of the overriding plates, probing crustal thickness in the past provides first-order information on the geodynamic processes that took place at plate margins. Using this approach, we document the progressive buildup of a thick (>40 km) Jurassic to Cretaceous accretionary belt along the circum-Pacific orogenic belts that bounded the Panthalassa Ocean. The destruction of this thick belt started at ca. 125 Ma and was initially recorded by the thinnest magmatic systems hosting amphibole-bearing magma. Thinning of the circum-Pacific orogenic belts became widespread in the northern regions of western America and in the western Pacific after ca. 75 Ma, possibly in response to oceanic plate segmentation, which triggered slab rollback and overriding plate extension.
Most remnant accretionary belts on Earth are characterized by large exposures of mantle-derived igneous material and only limited occurrences of metamorphic rocks (Cawood et al., 2006). As a result, processes associated with crustal thickening and deformation at plate margins are not fully understood, because their investigations are mainly dependent on classical thermo-barometric tools, such as pseudosections, which are only applicable to metamorphic rocks. In addition, the origin and evolution of the deep crust are normally inferred from exposures of shallower crustal sections, and this results in major uncertainties regarding the pressure (P), temperature (T), and deformation history of accretionary belts.
Accounting for these limitations, various geochemical indices (Leeman, 1983; Plank and Langmuir, 1988) have been proposed as proxies for probing crustal thickness in magmatic arcs (Mantle and Collins, 2008; Zellmer, 2008, Chiaradia, 2014; Chapman et al., 2015; Profeta et al., 2015). Such proxies are based on tens of thousands of analyses of magmatic rocks covering most of Earth’s history, and uncovering patterns and temporal variations using data mining techniques. However, none of these proxies has hitherto been tested against the mineral, thermal, and chemical records of magma, which, together, integrate many of the high-frequency variations and complex details that can exist in magmatic systems. Such an approach is possible through the examination of crystallization conditions of hydrous magmatic minerals, such as amphiboles, which could provide petrological and temperature constraints on the control of crustal thickening on magma evolution (Rutherford and Devine, 2003; Hawthorne et al., 2007; Larocque and Canil, 2010; Ridolfi and Renzulli, 2012; Erdmann et al., 2014; Putirka, 2016).
In this paper, we propose a new integrated approach for qualitatively unravelling the crustal thickness of fossil magmatic systems based on the chemical and thermal records of magma. We refer to “thin” or “thick” magmatic systems based on the crustal thickness (oceanic or continental) on which the arc magmatism was constructed (i.e., we do not refer to the size or volume of the magmatic reservoirs within the crust). We first outline several lines of evidence that arise from petrological analyses and geochemical modeling in magmatic systems hosting amphiboles. We present a multidimensional database of magmatic rocks and minerals, and we explore the sensitivity of different chemical, mineral, and thermal parameters to crustal thickness in young (younger than 10 Ma) igneous rocks. We then test the empirical correlations on a number of ancient subduction margins within the circum-Pacific orogenic belts, for which there are independent constraints on crustal thickness variations through time. In a broader geodynamic context, the new integrated approach sheds light on the growth of a thick (>40 km) Jurassic to Cretaceous accretionary belt along the circum-Pacific orogenic belts and its subsequent destruction.
Crustal thickening in convergent plate margins has profound effects on the chemical evolution of SiO2-poor mantle-derived basalts toward SiO2-rich evolved intermediate to felsic magmas (Miyashiro, 1974; Atherton and Petford, 1993; Tepper et al., 1993; Sisson and Bacon, 1999; Dungan and Davidson, 2004; Mamani et al., 2010; Turner et al., 2012). This evolution is classically explained through the concept of fractional crystallization (DePaolo, 1981) and/or dehydration crustal melting processes (Hildreth and Moorbath, 1988). The efficiency of such processes at middle to shallow crustal levels has been discussed based on geochemical and thermodynamic modeling (Spera and Bohrson, 2001; Glazner, 2007; Clemens et al., 2009; Kratzmann et al., 2010). It has recently been suggested that the geochemical signatures of magmas are preferentially acquired deepest in the crust (Lee et al., 2014) and that early-formed high-P mineral assemblages trapped at the bottom of thickening arcs (Barclay and Carmichael, 2004; Davidson et al., 2007, 2013) could exert a first-order control on magma differentiation (Rutherford and Devine, 2003; Giesting and Filiberto, 2014; Dessimoz et al., 2012; Chiaradia, 2014). The boundary between the lower crust and the upper mantle (Moho) represents a mechanical threshold that favors accumulation of mantle-derived basalt, crystallization, and residual melt segregation, possibly into deep hot zones (Ernst, 1999; Annen and Sparks, 2002; Annen et al., 2006). Other models that attribute the chemical variability of arc magmas to slab or wedge processes are discussed in Turner and Langmuir (2015).
Amphibole in Magma
Magmatic amphiboles are considered to represent either (1) primary minerals formed after relatively low-T melting of a hydrous mantle above a dehydrating subducting slab, or (2) secondary phases formed after crystallization of evolved hydrous silicate melts (Tribuzio et al., 2014), which commonly exsolve from magmas upon crystallization (Coogan et al., 2001; Smith, 2014). Uncertainties surrounding the genesis of magmatic amphiboles have stimulated many experimental studies (Allen and Boettcher, 1983; Sisson and Grove, 1993; Wallace, 2005; Larocque and Canil, 2010), and there is a general agreement that changes in temperature (Ti-and Al-tschermak exchanges: Hammarstrom and Zen, 1986; Hollister et al., 1987; Johnson and Rutherford, 1989; Schmidt, 1992; Anderson and Smith, 1995; Anderson et al., 2008) and magma compositions (Sisson and Grove, 1993; Pichavant et al., 2002) affect the composition of magmatic amphiboles.
The origin of high-Si amphiboles (hence low-Al and low-Ti amphiboles; Putirka, 2016) is classically interpreted as forming in a shallow magmatic reservoir made of evolved (felsic) melts (Grove et al., 2003). This magmatic reservoir may be injected by hot and less-evolved (mafic) melts coming from the deepest magmatic reservoirs, resulting in a mixing of magmas (e.g., Humphreys et al., 2009). Mafic melts will form a new generation of low-Si amphiboles in the shallow magmatic reservoir, destabilizing the early generation of high-Si amphiboles (Erdmann et al., 2014; Kiss et al., 2014). The intrinsic sensitivity of amphibole to temperature and melt composition, rather than pressure (Putirka, 2016), makes the usage of natural amphibole for geobarometry rather controversial, particularly in the absence of careful examination of equilibrium textures (i.e., homogeneous composition or normal zonings; Ague, 1997; Manley and Bacon, 2000; Lindsay et al., 2001; Bachmann et al., 2002; Janoušek et al., 2004; Blundy et al., 2006; Zhang et al., 2006; Rodríguez et al., 2007; Anderson et al., 2008; Needy et al., 2009; Samaniego et al., 2008; Ruprecht and Bachmann, 2010; Turnbull et al., 2010). Nonetheless, the potential of amphibole as a geothermometer is generally accepted (Bachmann and Dungan, 2002; Shane and Smith, 2013). For this reason, we largely based our conclusions on the chemical composition of amphiboles and magma, rather than the P-T estimates.
METHODOLOGY AND APPROACH
What Is a “Mean” in Petrology?
Sampling is the basis of all work in earth science disciplines, even as diverse as they are. In petrology, the sample on which we base our thinking is seen as part of a whole, chosen to represent one or more characteristic properties of that set: but is it really the case? All sampling, even the simplest, is subject to several sources of error related to the sampling technique, the way in which that particular technique is applied, or to the instrument used for sampling. When the volume of the sample is too large, it must be down sampled to only analyze a small portion of the sample. The risk of selecting the most spectacular subsample, distorting the representation of this set, is omnipresent. Each step of down sampling implies an error, and these errors accumulate, altering our perception of an event or a geological process. Geostatistics can help to quantify the magnitude of these errors, but work on quite large data sets often extracts an average value (“mean”). The estimated mean of a variable is only possible if we measured this variable on large numbers of samples. Thus, the quantity and quality of sampling influence that of the “mean” itself. There is no need to develop sophisticated estimation techniques if the database is not reliable. The multiscale processing of these data is a relatively recent approach in earth sciences, arising from the need to take into account all phenomena, in practice coupled, that have acted on (or are present in) a given system. This is the most complete form of modeling of a sequence of various phenomena proceeding at different orders of magnitude, since it integrates, at different scales, all theoretical or empirical knowledge in elementary bricks that need to be assembled.
Recent advances in the domain of igneous petrology (Ghiorso and Sack, 1995; Herzberg and Asimow, 2008; Putirka, 2008) have occurred along with the development of chemical databases of magmatic rocks and minerals, providing access to several tens of thousands rocks and minerals analyses spanning most of Earth’s history. Data mining techniques allow the exploration of these data sets to uncover hidden patterns and trends over time periods of several tens to hundreds of millions of years, matching those of geodynamics (Keller and Schoene, 2012; Tang et al., 2016). As they incorporate a growing number of data, the results of these new global studies are surrounded by a statistical dimension out of necessity, which makes them quite unpopular in the scientific community of petrologists. Yet, these new approaches are necessary because they capture the geological signals at longer wavelengths, smoothing the analytical errors, filling the sampling bias, and averaging values along secular trends. Accordingly, the “petrological mean” computed from these studies has a real meaning because it fully accounts for (and does not mask) the natural variability in petrology that arises from variable temperatures (and depths) of partial melting and crystallization in igneous rocks from widely different geological environments (arcs, cratonic provinces, oceans, etc.). Here, we opted for this integrated approach, and we emphasize that our conclusions can only provide dominant trends rather than strict rules.
Database and Statistical Approach
We used an extensive geochemical data set of continental and oceanic igneous rocks and minerals (amphibole, pyroxene, plagioclase, and olivine) from the Phanerozoic to Archean. Data from the GEOROC database (https://www.georoc.mpch-mainz.gwdg.de/georoc/) were compiled by Ganne et al. (2016), using published and unpublished data, and then filtered to exclude all samples for which the summed oxides yielded totals outside the 95%–101.5% range. Geochronological data associated with these samples were reported as means, with 1-standard-error uncertainty given by the method (e.g., U/Pb geochronology or biostratigraphy). Georeferenced bulk-rock chemistry and mineral analyses (3646 references) were used to derive P and T estimates. The resulting data set includes a total of more than 16 million data points, with up to 130 variables for each of the >55,000 rock samples and 80 variables for each of the >117,000 mineral analyses. Results for global mean intensive (geochemical) and extensive (P, T) values with time are reported with associated 1-standard-error uncertainties of the mean at 50 m.y. intervals. These means were generated by Monte Carlo analysis, using the standard bootstrap resampling approach outlined in the following.
Our approach follows Keller and Schoene (2012), with information from the original source added where necessary. The following procedure was conducted: (1) A subset of data was randomly selected such that the probability of inclusion in the resampled subset was directly proportional to the sample weight; (2) a synthetic data set for each sampled data point was drawn from a Gaussian distribution with a mean equal to the original value of the data point and the standard deviation equal to the estimated 1-standard-error uncertainty for the data point; (3) the resulting data were sorted into 50 m.y. bins, with a mean and variance calculated for each variable; (4) steps 1–3 were repeated 1000 times, where a minimum number of samples (threshold of 3) within each 50 m.y. bin was fixed for steps 1–3; and (5) the total mean and standard error of the mean were calculated for each variable in each bin. Results are reported as means with associated 1-standard-error uncertainty of the mean for 50 m.y. intervals between 0 and 300 Ma.
In contrast to the approach of Keller and Schoene (2012), which attempted to produce a uniform record by reducing the weight of abundantly sampled time periods or terranes, our sampling was not assigned to be inversely dependent on spatial and/or temporal sample density. Similarly, uncertainties surrounding the age of samples were not considered in our statistical approach, as this standard error is not homogeneous and depends on the method of dating used. Uncertainties surrounding ages are less than 50 m.y. for >5% of the magmatic samples, with a mean of ~15 m.y. Uncertainties are thus lower than the results of bootstrap that we reported as means for 50 m.y. intervals, and thus are not likely to change our general conclusions on the chemical evolution of magmas and minerals through geological time. Our Monte Carlo analysis likely results in a more discontinuous geochemical record compared with the statistical approach of Keller and Schoene (2012), which smoothed out stochastic and stepwise transitions; the slopes of any abrupt trends presented in the diagrams are therefore maximum estimates. Overall, we assume that the robustness of our statistical approach is strengthened by the negligible occurrence of sampling gaps for magma and mineral record in the database, more specifically between 0 and 300 Ma.
Thermometry Using Magmatic Amphibole
The database contains compositional analyses of 12,484 amphiboles compiled from the literature (25%), GEOROC (46%), and unpublished data (29%), 50% of which were matched with their bulk-rock chemistry (1438 samples; major and trace elements). Structural formulae for amphiboles were calculated using the Probe-Amph excel program (Tindle and Webb, 1994). Figure 1 presents microprobe analyses from the database. A chemical evolution can be observed between the most Si-depleted amphibole (ferropargasite or ferrotschermakite end members according to the [Na + K]A value), which are characterized by high-Fe contents, and the Si-enriched amphiboles (magnesiohornblende or edenite), which are characterized by elevated Mg contents (Ridolfi and Renzulli, 2012; Putirka, 2016).
In this study, we used a new Al-based coupled thermobarometer model (Ridolfi and Renzulli, 2012; RR2012) built on an extensive synthesis of literature and new experimental data on magmatic amphiboles. Recent studies (Yücel et al., 2013; Kent, 2014; Erdmann et al., 2014) have shown that temperature estimates obtained by the RR2012 method are consistent with the amphibole-plagioclase thermometer (Holland and Blundy, 1994; Blundy and Holland, 1990). However, the confidence of the calculated pressure estimates is less robust (Putirka, 2016). Ideally, each amphibole should be checked for equilibrium textures (i.e., homogeneous composition or normal zonings) before using the RR2012 method (Ridolfi and Renzulli, 2012), because reverse zonings are likely to indicate magma mixing in an environment with a high rheological and viscosity contrast that can lead to overestimated pressure (Tajčmanová, et al., 2014; Moulas et al., 2014). Unfortunately, in the compiled literature (>900 references), information on equilibrium textures is commonly absent.
Chemical data collected on magmatic amphiboles were filtered on the basis of pressure-related “apparent percentage error” (APE) values calculated with the RR2012 thermobarometric method. Following the recommendation of Ridolfi and Renzulli (2012), a maximum of 50 for the APE was set to retain or exclude the data (first filter: moderate confidence level). To secure our estimates, a second filter (high confidence level) was applied, leading to exclusion of amphiboles for which the chemistry did not fit the dedicated range of chemical composition and structural formulae. More than 70% of the calcic-amphiboles in our database were discarded.
Magma as a Proxy for Crustal Thickness
Major-and trace-element signatures of Cenozoic magmas in plate margins show significant along-strike variations that correspond to changes in crustal thickness (Zellmer, 2008; Turner and Langmuir, 2015). Furthermore, it appears that early-formed minerals trapped at the bottom of thickening arcs (Davidson et al., 2007) could exert a first-order control on the magma differentiation trends (tholeiitic vs. calc-alkaline; Miyashiro, 1974). Accounting for this mechanism of mineral fractionation within the crust, various geochemical indices have been proposed as proxies for probing crustal thickness in magmatic arcs, with a representative list given in Chapman et al. (2015). The tholeiitic (Th) index proposed by Chiaradia (2014; more Fe2O3 total-enriched at MgO ~4–6 wt%) is based on the observation that the thinner the magmatic arc, the more mafic and tholeiitic (Th index of ~11–12) is the magma (Chiaradia, 2014). Elevated water content and fO2 in the deep section of thick crusts lead to a more calc-alkaline composition (Th index of ~7–8) through suppression of plagioclase fractionation and promotion of early magnetite and amphibole fractionation (Sisson and Grove, 1993).
Figures 2 and 3 show that crustal thickness exerts a first-order control on the composition of magmas, which become more calc-alkaline and felsic with increasing crustal thickness. We achieved this by compiling a comprehensive database for the geochemistry of magmas from relatively young rocks (younger than 10 Ma) for which the crustal thickness is known (Mooney et al., 1998). While Chiaradia (2014) or Profeta et al. (2015) focused their studies on arc settings, here we looked for more global correlations between crustal thickness and young magma compositions that were likely produced through a variety of magmatic processes and involving different geological settings (i.e., continental arcs, intraplate magmatism, rift, orogenic belts, etc.). For each data point, we plotted the corresponding crustal thickness hosting the magma, thus allowing us to explore relationships between the chemical signature of magma and crustal thickness. We minimized errors induced by a possible thickening of crust following magma emplacement by discarding data older than ca. 10 Ma. The magma series considered along northwestern and southwestern America (Fig. 3) covers the compositional range of basalts to rhyolites and their intrusive equivalents (40–75 wt% SiO2).
Figure 2 shows that the Th index of magma and associated SiO2 (wt%) range around values of ~10 and 55, respectively, for crustal thickness ranging from 25 to 40 km. As a consequence, no good precision can be obtained when probing the crustal thickness of magmatic belts in which the magma evolved in the past within this range of Th index and SiO2 (wt%) composition. Beyond this threshold, the Th index strongly decreases with increasing crustal thickness and SiO2 wt% of the magmatic rocks, but again less confidence can be given to these values because they are not supported by high-density data. This limitation questions the validity of an approach that solely relies on a single magma proxy to depict crustal thickness in the past, as well as the real quantitative significance of values obtained in orogenic belts that reached a crustal thickness exceeding 45 km. As an extension of that conclusion, Figure 3 shows that two major elemental concentrations (SiO2 wt% and Th index) can be used conjointly as crustal thickness proxies and yield consistent results when larger global and regional data sets are averaged out.
Amphibole as a Proxy for Crustal Thickness
Given that magma compositions affect the composition of magmatic amphiboles (Sisson and Grove, 1993; Pichavant et al., 2002), the amphiboles could be potentially used as a complementary proxy for crustal thickness. We again approached this issue by compiling a comprehensive database of the geochemistry of geologically young amphiboles (younger than 10 Ma) and their hosting magmas, together with information on crustal thickness (Mooney et al., 1998). Results for global mean intensive (geochemical) and extensive (P, T) values, and corresponding crustal thicknesses, are reported in Figure 4. Plots of the variability of T and P conditions of amphibole crystallization versus their chemical composition (Ti and Si atoms per formula unit [apfu]) or the chemical composition of the hosting magma (SiO2 [wt%]) are shown in Figures 5 and 6.
Figures 4A and 4B show that the formation of Si-poor (<6 apfu) and Ti-rich (>0.25 apfu) magmatic amphiboles occurs at high-temperature conditions (>950 °C; Figs. 4D, 6A, and 6B). Likewise, such amphiboles dominantly develop in basaltic to basaltic-andesitic composition (SiO2-poor, i.e., <55 wt%) magma emplaced in crustal sections thinner than 10 km (blue point in Fig. 4B). Beyond these values, high-T amphiboles in magma decrease with increasing crustal thickness to average Ti values lower than 0.2 apfu at ~40 km depth, accompanied by a narrower range of magma composition toward andesite to dacite compositions (>60 wt%). Statistical approaches (bootstrapping; Fig. 4B) applied to the large multidimensional amphibole database confirm that with increasing crustal thickness, the abundance of Tirich and Si-poor amphibole decreases (lower temperature and lower pressure; Figs. 4D and 4E) and the magmatic system contains less mafic rock in proportion (Fig. 4C).
Using chemical and thermal records of amphibole and magma as compelling markers, we propose an integrated approach to unravel the crustal thickness evolution of fossil magmatic systems. P estimates are less reliable (see Method section) and were not considered in this study. Figures 2 and 4 provide a simple basis to identify past, overthickened (>40 km) magmatic systems hosting “mature” deep crustal hot zones (Annen et al., 2006), based on the limited Ti-dependent temperature (<900 °C) and Si-dependent pressure (<3 kbar) record of magmatic amphiboles, and the dominant occurrence of andesitic to dacitic magmas (SiO2 > 60 wt%) sharing calc-alkaline affinity (Th index < 10). In addition, identification of thin (>10 km) magmatic systems hosting “immature” deep crustal hot zones relied on the limited Ti-dependent temperature (>950 °C; blue point in Figs. 4D, 5, and 6) and Si-dependent pressure (>5 kbar; Fig. 4E) record of magmatic amphiboles, and the dominant occurrence of basaltic to basaltic-andesitic magmas (SiO2 < 55 wt%; Figs. 4A and 4C) sharing tholeiitic affinity (Th index > 10).
As an extension of that work, we tested the sensitivity of the SiO2 (wt%) content in the magma to crustal thickness variation when amphibole is present (Fig. 7). The results illustrate that amphibole-bearing magmas occur in thinner magmatic systems (~25 km; Fig. 7C) compared to the average magma record (~28 km; Fig. 7C). Accounting for this pattern, we speculated that the progressive thinning of magmatic systems operating along this margin would result in the coalescence of shallow, intermediate, and deep-seated magmatic reservoirs (“deep hot zones”) in the crust into a single reservoir where mantle-derived tholeiitic magmas would not evolve significantly (Fig. 7D). Figure 7A shows that magma containing amphiboles (e.g., amphibole-bearing) yields more felsic compositions compared to those that do not contain amphibole (amphibole-free), regardless of the crustal thickness (e.g., for crustal sections <10 km, amphibolebearing and amphibole-free magma will contain 55 and 50 wt% SiO2, respectively). The corollary is that any change in thickness for magmatic systems will be first recorded by the amphibole-bearing magmas, changing their chemistry from felsic to mafic composition. It is thus expected that they will record the early steps of continental extension, promoting an incipient increase of the mafic magma, and then Ti-rich amphibole in the geological record. At a late stage, further extension of the continental crust will result in the general thinning of magmatic systems (amphibole-bearing and amphibole-free) and a more extensive record of mafic magma in the geological record (Mantle and Collins, 2008; Zellmer, 2008; Chiaradia, 2014; Chapman et al., 2015; Profeta et al., 2015).
Crustal Evolution of the Circum-Pacific Orogenic Belts
Unravelling the paleothickness of magmatic systems provides first-order information on the geodynamic history of subduction zone plate margins, which are known to be subject to transient switches in the style of overriding plate deformation (shortening or extension; e.g., Schellart et al., 2010). Here we focus on the circum-Pacific orogenic belts (Figs. 8 and 9), which constitute the largest, most long-lived, and still operational accretionary orogenic systems in the world (Collins et al., 2011). The Phanerozoic evolution of the circum-Pacific orogenic belts records the assembly and breakup of the last supercontinent, Pangea. Our database in the GSA Data Repository1 reveals an increasing occurrence of tholeiitic magma during the Cenozoic, which contrasts with the calc-alkaline affinity throughout the Jurassic and Cretaceous (Fig. 2D).
In our analysis, samples were filtered according to their geographic position along the circum-Pacific orogenic belts (western and eastern Pacific), and their SiO2 contents (wt%) were plotted against their age of crystallization. A negative correlation (i.e., chemical element decreasing with age) is observed from ca. 175 to 75 Ma in the western Pacific, with the youngest rocks showing the highest SiO2 contents (wt%) and calc-alkaline compositions (Fig. 9C). Beyond this threshold, a positive correlation is observed (i.e., chemical element increasing with age). Interestingly, the peak of SiO2 (>63 wt%) for magmas at ca. 75 Ma (yellow trend in Fig. 9C) is preceded by an increasing occurrence of Ti-rich amphibole in the amphibole-bearing magmatic systems since ca. 125 Ma (black trend in Fig. 9B). In order to better visualize average trends for amphibole, we applied a statistical treatment of data, with large (50 m.y.) and small (10 m.y.) sampling steps (Figs. 9 and 10). Considering the tectonic context, we recognize relationships between crustal thinning of amphibole-magmatic systems and the orogenic events that occurred in the western Pacific region after ca. 75 Ma. This chemical evolution of amphibole-bearing magma toward more mafic (SiO2-poor) composition started at ca. 125 Ma (green trend in Fig. 9C), preceding the onset of continental extension in the SW Pacific (ca. 90–70 Ma; Schellart et al., 2006; Mantle and Collins, 2008) and East Asia (ca. 70–55 Ma; Northrup et al., 1995; Ren et al., 2002; Schellart and Lister, 2005), and accompanied the process of seafloor spreading that took place around ca. 30 Ma around the Japan Sea (Fig. 10; Yoshiyuki et al., 1989).
This chemical trend is equally pronounced in the eastern Pacific. Rocks were arbitrarily filtered according to their actual latitudinal position, allowing us to define two subgroups: (1) North (latitudes north of to 30°N) and Central America (from 30°N to 10°S), and (2) South America (latitudes south of 10°S). A positive correlation was observed in North and Central America from ca. 175 Ma to present (Fig. 11A), with the youngest rocks showing highest SiO2 contents (wt%) and a more calcalkaline composition. This could correspond to a progressive thickening of the Mesozoic coastal belts (Coney et al., 1980; van der Heyden, 1992) from the Nevadan to Sevier-Laramide orogeny and their thinning in the Cenozoic. However, metamorphic studies performed along the Cordilleran belts (Ducea, 2001, and reference therein) argue for a thick crustal section (>40 km) at the time of the Sevier-Laramide orogeny. This hypothesis is supported by the narrow range of amphibole composition (averaged Ti content < 0.2 apfu) characterizing the Pioneer plutonic complex in southwest Montana (Zen, 1985; Hammarstrom and Zen, 1986; Fig. 3). In order to explore the progressive thickening of the Mesozoic coast belts discussed by Chapman et al. (2015) and Profeta et al. (2015), we reused the amphibole database of Hammarstrom (1984) compiled from four Mesozoic batholiths of northwestern America (SW Montana, British Columbia, SE Alaska, Idaho: white stars in Figs. 3 and 12A) and dated from ca. 117 to 70 Ma, respectively (Murphy et al., 2002; Saleeby and Rubin, 2000; Jeffcoat, 2013). Interestingly, no clear evidence of crustal thickening evolution of the amphibole-bearing magmatic systems is recognized for the Cretaceous. The progressive thickening of amphibole-bearing magmatic systems developed earlier, at ca. 200–150 Ma, involving the ca. 200–180 Ma Talkeetna oceanic arc in south-central Alaska (Armstrong, 1988; Greene et al., 2006; black star in Fig. 12A) that formed a crustal section <28 km thick (Hacker et al., 2008) in an oceanic domain (Sigloch and Mihalymuk, 2013) and was later accreted along the circum-Pacific orogenic belts (Coney et al., 1980). This Jurassic evolution was followed by progressive thinning of the amphibole-bearing magmatic systems, from ca. 125 Ma to present day, corresponding to a change from Nevadan to Sevier-Laramide orogeny to Basin and Range extension. A statistical treatment of data from the Sierra Nevada (du Bray, 2007), using a small sampling step (1 m.y.), helped to better visualize the increase of mafic magma record in the last 20 m.y. (Basin and Range extension; Fig. 11C).
The chemical evolution of magma is more linear in South America, with a progressive thickening of the accretionary belts since the Late Cretaceous, following a period of general extension and crustal thinning in the Jurassic and Early Cretaceous (Ramos and Aleman, 2000). This evolution is consistent with the magmatic history of the Central Andean orocline (Haschke et al., 2002; Mamani et al., 2010). The chemical record of magmas in South America (in general) and the Central Andean orocline (in particular) is also marked by an increasing range of composition toward SiO2-rich compositions for the youngest magmas (Fig. 11B). This evolution is consistent with the chemical record of magmatic amphiboles that evolve toward Ti-poor composition (Fig. 12B), but it differs from the Cenozoic Basin and Range record (Fig. 12A). This suggests an evolution under a compressional regime since at least the Cretaceous (Sempere et al., 2002). A significant decrease in the Dy/Yb trace element ratio in post-Oligocene (younger than 30 Ma) magma from the Central Andean orocline, indicative of amphibole sequestration in the lower crust (Davidson et al., 2007), was also observed (Mamani et al., 2010). This suggests major crustal thickening in the mid-Oligocene (Haschke et al., 2002), following an incipient crustal thickening event from ca. 90 to 30 Ma (Fig. 11C).
Implications for Plate-Tectonic Reconstructions
Based on our results, we suggest that accretionary belts in the circum-Pacific orogenic belts were subjected to progressive thickening from the Jurassic to the beginning of the Cretaceous, which involved the emplacement of progressively more differentiated magma (SiO2-rich). This thickening led to the development of “mature” deep hot zones around the Pacific, enabling magma storage and homogenization at depths up to 30–40 km (Annen et al., 2006). A major change took place from the middle of the Cretaceous to the beginning of the Eocene in western Pacific and northeastern America, associated with emplacement of less calc-alkaline and more mafic magmas during the Cenozoic (Figs. 9 and 11). This change is recognized in the chemical record of amphiboles that become Ti-rich, which is indicative of the deepest conditions of crystallization. We interpret this mineral and chemical change in the magma as resulting from the progressive thinning of the Jurassic to Cretaceous accretionary belts, which limited crustal component assimilation and amphibole filtering of the mantle-derived melts. Crustal thinning was likely driven by extensional processes, which occurred throughout the circum-Pacific orogenic belts during the final breakup of Pangea after ca. 125 Ma (Veevers, 2004). These processes were amplified by the segmentation of the Panthalassa oceanic plates during the Eocene, promoting slab rollback and inducing overriding plate extension (Schellart et al., 2010).
The spatial and temporal changes in thickening or thinning of the magmatic belts along the circum-Pacific orogenic belts, as inferred from the chemical and thermal evolution of magmas, are illustrated in Figure 13. McKenzie et al. (2015) recently suggested that most of the Pangea supercontinent was underlain by a contiguous arc of thick lithosphere. Beneath the western convex side of this arc, there was a wide belt of thinner lithosphere underlying what is believed to have been the active margin of Pangea, here named the circum-Pacific orogenic belts, or external (circum-Pacific) orogenies by Collins et al. (2011). In the Early Jurassic Epoch, oceanic crust of the Paleo-Tethys subducted under the Cimmerian plate, closing the ocean from west to east and forming internal orogenies (Collins et al., 2011). Around this period, the South China block collided with the southern margin of the North China block, producing the Qinling-Dabie orogeny (Weislogel et al., 2006), and the Lhasa terrane accreted onto Asia from the south in the Early Cretaceous (Chu et al., 2006). Active subduction and collisional processes on the eastern side of Pangea, coupled with the opening of the North Atlantic Ocean in the Middle Jurassic Epoch, likely produced progressive shortening and thickening of the belts in the western Pacific until ca. 125 Ma. At the same time, a progressive shortening of the magmatic belts took place in the eastern Pacific, thickening the northern segments of the American continent (Nevadan orogeny).
Though Pangea was still broadly intact in the beginning of the Cretaceous, it broke up after ca. 125 Ma as South America, Antarctica, and Australia rifted away from Africa, thus forming the South Atlantic and Indian Oceans. As the Atlantic Ocean widened, the convergent-margin orogenies that had begun during the Jurassic increased in the North American Cordillera, with the Nevadan orogeny, followed by the Sevier and Laramide orogenies. This period of active subduction along the circum-Pacific orogenic belts and active rifting in the core of Pangea was accompanied by a progressive thinning of the amphibole-bearing magmatic systems in the western Pacific and North America belts, whereas shortening and thickening of the belts started in South America after ca. 125 Ma.
Perspectives in Statistical Petrology: Toward an Integrated Approach
While it is increasingly accepted that statistical geochemistry can be established as a viable and powerful approach to understanding global issues in earth sciences, the robustness of geological interpretations derived from this approach remains in doubt when the secular changes are not sufficiently supported by data. A robust bootstrap analysis requires high data density and temporal continuity in the data set to ensure data representativeness (Keller and Schoene, 2012). This criticism potentially applies to this study, as our database illuminates considerable bias of sampling for magmatic amphibole and rocks in the Mesozoic record of the eastern Pacific. Hence, care must be taken when assessing the global significance of the ca. 125 Ma thermal event recorded by amphibole-bearing magmas and its consistency with the existing framework of global tectonics using a comprehensive set of independent data.
Mantle and Global Climate
The disjunction of the African and South American plates after ca. 125 Ma marked a turning point in the evolution of Pangea, which likely impacted the mantle climate evolution (Farrington et al., 2010). Determination of mantle temperature variations below continents at the time of supercontinent destruction is a fundamental issue in the discussion of global geodynamic processes; however, it remains a challenging task (Rey, 2015). Indirect clues come from the thermal signature of magmatic minerals, which provide exceptional insight into the thermal evolution of continental lithosphere through time.
In a companion paper, Ganne et al. (2016) presented the first worldwide compilation and statistical assessment of chemical information on magmatic rocks and minerals preserved in the continental record since 600 Ma, including the circum-Pacific orogenic belts. The aim of their study was to derive a global, but sufficiently representative data set of thermal constraints on magmas to highlight the process of mantle insulation below a supercontinent. Following the methods of Putirka (2008), which were built on a comprehensive review of existing thermobarometer calibrations for magmatic rocks, they compared different magmatic mineral compositions with glass (liquid) compositions using experimentally derived equations to obtain values for the temperature at the time of crystal formation.
The authors observed that the age of the thermal peak for amphiboles (ca. 275 Ma; Fig. 14C), dominantly tapping the hydrous calc-alkaline magmas forming and/or evolving at suprasubduction or crustal level, corresponds to a period of orogenic collapse for the Variscan-type belts suturing the Pangea supercontinent. Furthermore, the thermal peak for olivine (ca. 125 Ma; Fig. 14A), dominantly tapping the deep, anhydrous, and weakly evolved tholeiitic magmas, corresponds to a period of enhanced supercontinent breakup and seafloor spreading (Veevers, 2004). This is consistent with thermal peaks for clinopyroxene and plagioclase, forming in both tholeiitic and calc-alkaline magmas, which occurred at ca. 225 Ma (Fig. 14B). Ganne et al. (2016) proposed that the ~150 m.y. interval observed between thermal peaks of hotter and cooler magmatic minerals (e.g., olivine and amphibole, respectively) could be explained by the progressive thinning and thermal relaxation of the Pangea lithosphere, starting with the collapse of overthickened crustal horizons (i.e., orogenic belts) and ending with those of the lithospheric mantle at plate divergence zones.
Figure 15 summarizes this possible evolution. As the Pangea supercontinent grew in size and locally in thickness (orogenic belts), through arc-continent collision and microcontinent amalgamation, synchronous deformation and crustal reworking, with subsequent migmatite formation, enhanced mixing of magmas in the lower crust (mauve horizons in Fig. 15A). Given the protracted nature of Pangea, the progressive insulation effect due to the supercontinent must have been important, increasing the temperature of the lithospheric and asthenospheric mantle beneath the continental crust. Furthermore, this effect has been shown to be significant in the case of concurrent subduction along the boundaries of supercontinents, preventing lateral thermal mixing of the convective mantle (Lenardic et al., 2011).
During collapse and thinning of the thickened portions of the Pangea supercontinent (Fig. 15B), driven by excess gravitational potential energy and plate motion (divergence), decompression melting of the fertile crust and ambient mantle and extraction of calc-alkaline magmas (yellow trend in Fig. 15C) to more tholeiitic melts (blue and green trends) contributed to the initial thermal relaxation of the upper horizons of the lithosphere (ca. 275 Ma). This then propagated to the lowermost horizons of the lithospheric mantle in less than 150 m.y. (ca. 125 Ma).
After the breakup of the Pangea supercontinent, the subcontinental lithospheric mantle thickened through ongoing cooling and thermal subsidence, first incorporating the layer of moderately depleted mantle and then layers of unmelted fertile mantle (asthenosphere). Thermal cooling of the mantle below continents is hypothesized to have prevailed since 125 Ma, as suggested by the trend of T estimates obtained from primary magmas formed in continental settings (Ganne et al., 2016).
Circum-Pacific Orogenic Belts and Global Climate
The major tectono-magmatic trend captured by the GEOROC data set over the aggregation and dispersal of Pangea gives reasonable confidence that it is representative of exposed magmatic systems through time with minimum sampling bias (Ganne et al., 2016). Accounting for the consistency of mineral information gathered in the database, we now aim to evaluate the reliability of the mid-Cretaceous (ca. 125 Ma) event observed in the amphibole record of the circum-Pacific orogenic belts with respect to other minerals sampled in the same setting (i.e., circum-Pacific orogenic belts), like olivine and pyroxene, and to discuss its significance in the frame of Pangea supercontinent destruction.
We chose to exclude data from the South American regions because it followed a distinct evolution after 125 Ma compared to the rest of the circum-Pacific orogenic belts. The remaining data (>93% of the data set) concern the western Pacific and western regions of North America. A remarkable similarity can be observed between the thermal record of amphiboles sampled along the circum-Pacific orogenic belts, in particular, and those sampled in the continental lithosphere. Both are marked by an increase in temperature after ca. 125 Ma (Figs. 14C and 14F). This evolution differs from the olivine and pyroxene thermal records, which show decreasing temperatures of crystallization since 125 Ma (cf. Figs. 14A and 14D vs. 14B and 14E).
Such a difference in the thermal record of magmatic minerals can be explained by the loci of magma production in the continental lithosphere, including the circum-Pacific orogenic belts. Amphibole considered in this study formed at suprasubduction or crustal levels, dominantly along the circum-Pacific orogenic belts (>40%–50% of the data set; Fig. 8), and any changes in crustal thickness likely resulted in a change of magma and amphibole composition (Fig. 4). Therefore, the temperature increase of magmatic amphibole after ca. 125 Ma likely corresponds to a period of orogenic collapse for the accretionary belts surrounding the Pangea supercontinent (Fig. 13B). Conversely, pyroxene and olivine crystallized or evolved at higher temperatures than amphibole (averaged T < 1000 °C; Figs. 14C and 14F) in less-evolved magmas during their ascent through the mantle or crustal sections of the continental lithosphere. They dominantly record the effect of mantle and magma cooling, rather than a change in crustal thickness during accretionary belt collapse. As a whole, we emphasize that the global cooling and then warming of amphibole-bearing magmas emplaced since ca. 300 Ma along the circum-Pacific orogenic belts have been largely controlled by the disassembly and crustal reworking of the continental masses forming Pangea, related to changes in the temperature (T) and the dynamics of the subcontinental mantle.
Our integrated approach lends confidence to the use of geochemical and mineral correlations to qualitatively determine crustal thickness evolution in orogenic belts. However, quantitative estimates of paleocrustal thickness are far less understood petrologically and are difficult to constrain in orogenic belts that thickened over 45 km. We propose that early steps of continental extension can be distinguished in the geological record using the chemical and thermal signature of magmatic amphibole.
Using this approach, we documented the progressive buildup of a thick (>40 km) Jurassic to Cretaceous accretionary belt along the circum-Pacific orogenic belts that bounded the Panthalassa Ocean. The destruction of this thick belt started at ca. 125 Ma and was first recorded by the thinnest magmatic systems hosting amphibole-bearing magma. Thinning of the circum-Pacific orogenic belts became widespread in the northeastern regions of west America and in the western Pacific after ca. 75 Ma, possibly in response to oceanic plate segmentation and slab segmentation, which triggered slab rollback and overriding plate extension.
The robustness of our conclusions derived from data mining analysis primarily relies on the consistency of data with established constraints on the geodynamic history of the western and eastern Pacific. Accordingly, the major tectono-magmatic trend captured by our data set over the dispersal of Pangea gives reasonable confidence that it is representative of exposed magmatic systems through time with a minimum sampling bias.
Following the pioneering work of Keller and Schoene (2012), our work takes another step at demonstrating the power of data mining and analysis in the earth sciences to provide cutting-edge constraints on controversial problems that could only be answered using a global approach. Given that our results are consistent with the existing framework of global tectonics, we think that the work could serve as a proof of concept.
The project was supported by l’Institut de Recherche pour le Développement (IRD) and National Institute for Earth Sciences and Astronomy–Centre National de la Recherche Scientifique (CNRS-INSU) research funds. Ganne thanks S. Demouy and M. De Saint Blanquat for providing data from North and South America that supported our conclusions. G. Wörner and M.R. Renna are warmly thanked for their stimulating reviews and for the wonderfully spirited exchange of ideas (in sometimes mutual skepticism) surrounding data interpretation. Comments by J. Blundy and P. Ridolfi on an early draft of this paper are gratefully acknowledged. Finally, we wish to thank G. Bianchini and the GSA editorial staff for their excellent editorial work.
Figures & Tables
The Crust-Mantle and Lithosphere-Asthenosphere Boundaries: Insights from Xenoliths, Orogenic Deep Sections, and Geophysical Studies
- amphibole group
- chain silicates
- geologic thermometry
- igneous rocks
- magmatic differentiation
- orogenic belts
- P-T conditions
- plate tectonics
- statistical analysis