2.1 Brachiopod Sample Selection

The δ¹¹B of well-preserved, carefully sampled brachiopods has been shown to reflect the δ¹¹B of seawater and ambient pH, similar to other marine calcifiers (Jurikova et al., 2019; Lemarchand et al., 2002; Penman et al., 2013). For this study, we used well-characterized brachiopods from Ethan Grossman's collection at Texas A&M University (TAMU) and from the Yale Peabody Museum (YPM) collection. These brachiopods (Spirifer, Composita, and Dictyoclostus) were collected from low-latitude Carboniferous and Permian deposits in the U.S. Mid-continent (USM), Ural Mountains (UM), and West Spitsbergen (WSP; Grossman et al., 1991, 2008; Mii et al., 1997, 2001). USM brachiopods represent Carboniferous-Permian epicontinental seas and interior basin environments. They were collected from shales, limestones, and calcareous sandstones in the Kincaid, Fayetteville, Hale, Bloyd, and Tradewater Formations in Nebraska, Oklahoma, Arkansas, and Illinois, as well as the Admire, Council Grove, Chase, Cisco, and Wichita Groups in Nebraska, Kansas, Oklahoma, and Texas. UM brachiopods are predominantly Carboniferous (mid-Serpukhovian through early Permian). Finally, WSP brachiopods represent early Permian epicontinental seas and were collected from the Kapp Starostin formation in the westernmost island of the Norwegian Svalbard archipelago.

The samples used for this study all passed initial screening tests based on petrographic and cathodoluminescence (CL) microscopy (Figure 1; Grossman & Mii, 1996; Grossman et al., 1991, 1993, 2008; Mii et al., 2001). More recently, samples from the TAMU collection underwent clumped isotope analyses, and some of them (specifically UM samples) recorded anomalously high temperatures (Henkes et al., 2018), suggesting reordering at temperatures greater than 100°C. Additional work by electron backscatter diffraction demonstrated a range of preservation quality even in samples that appear well preserved (Pérez-Huerta & Laiginhas, 2018). Thus, there is clear evidence for solid-state alteration even if shells appear chemically and texturally preserved. This is important because boron is very sensitive to diagenesis and is typically lost with fluid-rock interactions, generally becoming isotopically lighter (Gaillardet & Allègre, 1995; Stewart et al., 2015). To screen for these potentially altered samples, elemental analyses (Li, Mg, Mn, Fe, and Sr) of solid brachiopods via laser ablation as well as dissolved sample solutions were performed using an Agilent 7500cx quadrupole inductively-coupled plasma mass spectrometry (ICP-MS). The elemental analysis results as well as δ¹⁸O values for samples previously analyzed by Grossman et al. (1991, 1993, 2008); Grossman and Mii (1996) and Mii et al. (2001) are listed in the supporting information (SI) associated with this article.

2.2 Boron Separation and Analysis via ICP-MS

Powdered samples were collected from nonluminescent areas (low Mn/Fe) in the secondary layers of brachiopod shells and dissolved in 2% nitric acid. Once samples were fully dissolved, the carbonate solution was increased to an alkaline pH by adding high-purity ammonia to ensure retention on the ion exchange resin. Columns were created by fitting a 1,000 μl pipet tip with a polyurethane frit and filling the tip with ~50 μl Amberlite 743 boron-specific ion exchange resin that was crushed and sieved to 63–120 μm. The flow rate for the columns was controlled by connecting the pipet tips to clean peripump tubing attached to a peristaltic pump. Before loading samples, the resin was cleaned with 3 ml 2% nitric acid and equilibrated to the sample pH (pH = 9–10) with a solution of Milli-Q water and high-purity ammonia at a flow rate of 5 rpm. Samples were then loaded onto the columns at a rate of 0.5–3 rpm. Columns were washed with 1.5 ml Milli-Q water and high-purity ammonia solution at 0.5–3 rpm before eluting the boron in 1.2 ml 2% nitric acid at 0.5–3 rpm. The load and wash steps were collected as boron-free sample solutions for trace-element analysis and archived for additional isotope work in the future.

Most boron isotopic compositions and concentrations were determined using a Nu Plasma II MC-ICPMS at the FIRST@StonyBrook isotope lab. We used a Glass Expansion Tracey PFA44 spray chamber with helix, a Glass Expansion 100 microliter/minute nebulizer with a flow rate of 30–35 psi, Nu Instruments light element sampling cones, and collectors H7 (¹¹B) and L6 (¹⁰B) in low-resolution mode. Each sample was bracketed with 50 ppb or 25 ppb (matching concentrations of diluted samples) NIST SRM 951 boric acid standard with a 2% nitric acid background run between all standards and samples. Washes with Milli-Q deionized water were completed before and after every analysis (90 s after standard and sample measurement; 60 s after background nitric measurements). Background was subtracted from samples and standards by taking the average signal of the bracketing nitric backgrounds and subtracting from the signal of the sample or standard, and the δ¹¹B was calculated based on the average of the bracketing standards after background correction. All samples were run multiple times with an in-house coral standard (SB Coral; δ¹¹B = 23.6 ± 0.3‰, [B] = 51.8 ppm, n = 35; comparison with international coral standard JCp-1 shown in Table 1), and the reported results (SI) were separated into two categories: reliable data with 2σ better than 1.5‰, a sample/blank ratio of 30 or more, and SB Coral δ¹¹B = 23.6 ± 0.5‰; and less reliable data with 2σ better than 3‰, a sample/blank ratio of 10 or more, and SB Coral δ¹¹B = 23.6 ± 1.5‰. Overall, we completed >200 analyses on Carboniferous and Permian brachiopods, though only 58 analyses met our data selection criteria. The reliable data are the only data shown in the figures within this article, but the results of all analyses are shown in the SI. All sample δ¹¹B values obtained via MC-ICPMS were corrected by the offset (±0.5‰) of SB Coral from 23.6‰ in each batch of analyses.

For samples analyzed by NTIMS at LDEO, all samples were bleached overnight in NaOCl to remove organic material and subsequently rinsed and ultrasonicated 10 times in quartz-distilled water then dissolved in quartz-distilled 2N hydrochloric acid. The 1 μl aliquots of sample solution (~1 ng of B) were loaded on degassed zone refined rhenium filaments with 1 μl of boron free seawater. Boron isotope ratios were measured in oxide form, acquiring masses 43 (¹¹B¹⁶O₂) and 42 (¹⁰B¹⁶O₂) using the NTIMS techniques of Hemming and Hanson (1992). To avoid possible biases resulting from in-run instrumental fractionation, analyses were accepted only if the signal was stable or rising, the measured ratio (43/42) varied less than 1‰ throughout the course of the run, and at least three replicates gave concordant 43/42 ratios to within <1‰. To test the consistency of δ¹¹B values between MC-ICPMS and NTIMS analyses, we ran some samples on both instruments (Table 2). When samples were taken from the same areas of the brachiopod shell, the variation in δ¹¹B values from NTIMS and MC-ICPMS was very small (±0.2‰). However, when samples were taken from different areas of the brachiopod shell, δ¹¹B values differed by up to 2‰, confirming the findings of Penman et al. (2013), which showed an intrashell δ¹¹B variability in modern brachiopods of up to 3‰.

2.3 Principle Component Analysis of Brachiopods

In addition to the abovementioned screening techniques, we also performed principle component analysis (PCA) for all brachiopods used in this study to check for any variables that may bias the δ¹¹B results (i.e., to determine if the δ¹¹B values of our samples are due to sample locations or genera rather than seawater chemistry). To be sure that these analyses were complete, we also included the samples from Joachimski et al. (2005), which are plotted in Figure 4 along with samples from this study. Along with their δ¹¹B, boron concentrations, and ages, each sample was assigned numbers from 1 to 13 to represent the sample's genus and numbers from 1 to 4 to represent the location from which the sample came. The results of these analyses are shown below (Figure 2).

Based on the results of the PCA, we see that there is no correlation between the location of a brachiopod sample and any other variable. Unsurprisingly, there is a significant correlation between the geologic age of the brachiopods and the boron concentration (Figure 2a). Additionally, there seems to be a slight correlation between brachiopod genus and δ¹¹B, though when looking at the second and third principle components, genus is also correlated with the geologic age and boron concentration of the brachiopod (Figure 2b). This correlation is likely the result of different brachiopod genera occurring at different times in the geologic record.

3.1 Results

Based on the δ¹¹B of the samples that passed screening tests (58 of 173), we can group brachiopods into two categories: the Pennsylvanian to earliest Permian (320–295 Ma) and the early-to-middle Permian (295–270 Ma). The older brachiopod δ¹¹B values average ~15‰, while the younger brachiopods average much lower δ¹¹B values around 10‰ (Figure 3 and Table 3). Based on the average δ¹¹B of brachiopods from this study and Joachimski et al. (2005), we can calculate possible Carboniferous-Permian seawater δ¹¹B values using the following equation (Equation 1; Foster & Rae, 2016):

(1)

where δ¹¹B_SW is the δ¹¹B of seawater δ¹¹B_Carb is the δ¹¹B of carbonates (in this case, brachiopods), α_B is the boron isotopic fractionation factor, K_B is the dissociation constant, pH is the pH of seawater, and ε_B = (α_B − 1) × 1,000. For the seawater δ¹¹B reconstructions in this study, we used the mean surface seawater pH values of a “Neritan” ocean modeled by Ridgwell (2005) as well as constant modern-day values (pH = 8.2), a pK_B of 8.6 (Dickson, 1990; T = 25°C; salinity = 35 psu), and boron isotopic fractionation factors from Lecuyer et al. (2002); α_B = 1.020023) and Klochko et al. (2006); α_B = 1.019368; Figure 3).

3.2 Modeled Boron Isotopic Budget of Seawater

Boron isotopes have been shown to have great utility in the Cenozoic for estimating changes in atmospheric pCO₂ due to a pH-controlled isotope fractionation. A major hurdle to taking this further into the geologic past is the need to know the δ¹¹B of ancient seawater to calculate its pH. For the Cenozoic, foraminifera that inhabit known depths in the ocean can be used to consider gradients, allowing the estimation of seawater δ¹¹B (Greenop et al., 2017; Pearson, 1999). However, beyond the time when these taxa are recognized, there are few ways to definitively determine the δ¹¹B of past seawater. Several studies have addressed this issue by modeling the boron isotopic ratio of seawater through time based on estimates of past boron fluxes (i.e., Joachimski et al., 2005; Lemarchand et al., 2002; Simon et al., 2006). These estimates are based on modern boron fluxes and their isotopic ratios as well as assumptions about the controlling mechanisms behind those fluxes throughout geologic time (Table 4).

Using two mass-balance equations for the boron content and boron isotopic composition of seawater (Equations 2 and 3; Lemarchand et al., 2002), we modeled Carboniferous-Permian seawater δ¹¹B in an attempt to fit our data-based seawater δ¹¹B curve (Figure 4d). For this model, we used scaled versions of the flux rates and the boron isotopic compositions seen in Table 4 along with the following equations from Lemarchand et al. (2002; Equations 2 and 3):

(2)

(3)

where B_SW is the boron content of seawater, t is time in years, φ_in is total boron flux into the ocean, φ_out is total boron flux out of the ocean, R_SW is the ¹¹B/¹⁰B of seawater, and R_in and R_out are the weighted averages of the ¹¹B/¹⁰B for the fluxes into and out of the ocean. We initialized our model at 320 Ma with modern-day boron flux rates and isotopic compositions. The data-based seawater curve seen in Figure 4 is a locally weighted scatterplot smoothing (LOESS) regression and 95% confidence interval of seawater δ¹¹B calculated from our brachiopod data using the Lecuyer et al. (2002) brachiopod boron fractionation factor and seawater pH from Ridgwell (2005); for comparison, the seawater δ¹¹B calculated using the inorganic calcite boron fractionation factor of Klochko et al. (2006) can be seen in Figure 3. While there may be unknown vital effects affecting the relationship between the δ¹¹B of Paleozoic brachiopods and the δ¹¹B of Paleozoic seawater, we have chosen to use the modern brachiopod-based boron fractionation factor of Lecuyer et al. (2002) for our seawater δ¹¹B reconstruction. We then used scaling factors for the boron fluxes throughout the Carboniferous and Permian based on Wold and Hay (1990; riverine/weathering flux; Figure 4a), Gaffin (1987; hydrothermal, accretionary prisms, and low-temperature alteration fluxes; Figure 4b), and Walker et al. (2002; carbonate flux; Figure 4c). The δ¹¹B of the fluxes were kept constant at their modern values with the exception of flux out from carbonates, which was based on the average δ¹¹B values of the brachiopods used in this study.

3.3 Comparison to Other Seawater Isotopes

While the modeled seawater δ¹¹B mirrors the general trend seen in the seawater δ¹¹B estimated from our brachiopod data, it fails to capture the ~4‰ drop across the Asselian-Sakmarian transition. To better understand what the seawater δ¹¹B model may be missing and why it does not match the δ¹¹B curve suggested by our brachiopod data, we compared our δ¹¹B record to trends in other seawater isotope systems during this time (Figure 5). The strontium isotopic composition of seawater (⁸⁷Sr/⁸⁶Sr) based on brachiopod measurements shows a trend that is similar to our δ¹¹B curve, with a high plateau during the Pennsylvanian followed by a steep decline across the boundary and into the Permian (Figure 5b; McArthur et al., 2012; Veizer et al., 1999). The brachiopod-based calcium isotopic composition (δ^44/40Ca) of seawater appears to show an overall decrease across the Carboniferous-Permian transition and continuing further into the Permian (Figure 5; Blättler et al., 2012). Unfortunately, because of the gap in δ^44/40Ca data from ~283 to ~265 Ma, we do not know if this decrease was gradual or steep. Overall, all three seawater isotope proxies decrease as the Carboniferous transitions into the early Permian, suggesting a common controlling mechanism.

The most significant source of boron to the oceans is riverine discharge (86% of boron input), while seawater boron outflux is dominated by low-temperature oceanic crust alteration (59% of boron output) and, to a lesser extent, adsorption onto clay minerals (28% of boron output; Table 4; Lemarchand et al., 2002). The strontium isotopic composition of seawater is controlled by the balance between riverine (granitic) and hydrothermal (basaltic) inputs to the oceans (Antonelli et al., 2017). Similarly, the major controls on the calcium isotopic composition of seawater are the balance between riverine and hydrothermal inputs and outflux via carbonate precipitation and oceanic crust alteration (Table 5; Fantle & Tipper, 2014; Farkaš et al., 2007; Griffith et al., 2008). The dominant mineralogy of seawater (i.e., calcite vs. aragonite seas) is also thought to play a significant role in the calcium isotopic composition of seawater (Blättler et al., 2012), but there is no such switch during the period of interest in this study. Based on these, the major factors likely controlling all three seawater isotope systems during the early Permian climate transition are riverine and/or hydrothermal fluxes.

The Carboniferous-Permian decline in seawater ⁸⁷Sr/⁸⁶Sr has traditionally been interpreted as the result of two major environmental changes: (1) an initial decrease in continental weathering due to the extreme aridity associated with the assembly of Pangaea and (2) a continued decrease caused by increased hydrothermal activity associated with the opening of the Neotethys and heightened volcanism (Korte et al., 2006; Martin & Macdougall, 1995). With a decrease in continental weathering at the Carboniferous-Permian boundary, we would expect an increase in seawater δ¹¹B and δ^44/40Ca as the major source of “lighter” (relative to seawater) isotopes are removed (Tables 4 and 5; Farkaš et al., 2007; Griffith et al., 2008; Lemarchand et al., 2002). Because both the boron and calcium seawater isotopic compositions show decreases in the Permian, decreased continental weathering does not appear to be a common direct controlling factor for all three isotopic systems examined here. Similarly, an increase in hydrothermal activity in the early Permian is also unlikely to be directly controlling all three isotope systems as this would also result in an increase in seawater δ¹¹B and δ^44/40Ca by intensifying a major removal mechanism for light isotopes of boron and calcium (Tables 4 and 5; Farkaš et al., 2007; Griffith et al., 2008; Lemarchand et al., 2002).

Another possible interpretation of the changes in seawater isotopic composition in the early Permian is a change in the isotopic composition of riverine discharge. Traditionally, continental weathering materials transported by rivers are assumed to have an overall granitic composition (Taylor & McLennan, 1985). However, this may not have been the case throughout the entire Phanerozoic. In fact, a significant amount of weathering material delivered to the oceans may come from “basaltic rivers” and volcanic islands causing the weathering flux to have a more mafic composition (Allègre et al., 2010; Antonelli et al., 2017; Godderis et al., 2009; Kump et al., 2000). A more mafic riverine input due to increased weathering of basalts would certainly fit the decline seen in seawater ⁸⁷Sr/⁸⁶Sr across the Carboniferous-Permian boundary; however, it cannot explain the decrease in seawater δ¹¹B and δ^44/40Ca during the early Permian as the isotopic composition of the source rocks weathered and transported by rivers has been found to have little effect on the boron and calcium isotopic compositions of the riverine input to seawater (Rose et al., 2000; Tipper et al., 2006).

3.4 Permian Weathering and the Carbon Cycle

Overall, there seems to be no singular mechanism controlling the boron, strontium, and calcium isotopic compositions of seawater during the Carboniferous-Permian transition. Instead, there were likely multiple connected controls in effect. One such scenario would be the carbon cycle changes resulting from a decrease in continental weathering. Goddéris et al. (2017) suggest that the assembly of Pangaea combined with the lowering of the Hercynian Mountains in the early Permian led to a significant decrease in continental weathering due to extreme aridity and the creation of a thick, weather-resistant soil layer on the continent. Along with this reduction in continental weathering, there was also subsequent rise in atmospheric CO₂. Diminished continental weathering explains the drop we see in seawater strontium isotopes, while the decreases in boron and calcium isotopes fit with a drop in seawater pH resulting from the increased atmospheric CO₂. As the pH of seawater declines, we would see an increase in the dissolution of marine carbonates. These dissolved carbonates would, in turn, release light boron and calcium isotopes back into seawater as well as diminish one of their removal mechanisms. Both of these consequences would lead to a decrease in seawater δ¹¹B and δ^44/40Ca. However, though carbonates are the largest remover of calcium from seawater (Table 5; Farkaš et al., 2007; Griffith et al., 2008), they only make up 13% of the seawater boron outflux (Smith et al., 1995; Spivack & Edmond, 1987; Vengosh et al., 1991), meaning there must be an additional boron removal mechanism affected by the decrease in seawater pH.

To better constrain the extent to which boron removal from seawater would need to decrease in the early Permian to reproduce the seawater δ¹¹B curve suggested by our data, we modeled the boron outflux fraction values needed to fit the seawater δ¹¹B curve based on our brachiopod data (Figure 6). The results of this calculation imply that boron outflux from seawater began to decrease at the beginning of the Sakmarian before reaching its lowest value (a 29.2% reduction) by the early Artinskian. The dissolution of marine carbonates alone would only result in a maximum boron removal reduction of 13%, so we need to examine the chemistry behind how boron is removed from the seawater system.

All known boron fluxes out of the ocean are isotopically light compared to seawater (Joachimski et al., 2005; Lemarchand et al., 2002; Simon et al., 2006). This is due to the preferential removal of ¹⁰B in the form of borate sorption onto clays and coprecipitation in carbonates. Since the availability of borate in seawater is pH dependent (Hemming & Hanson, 1992), the decrease in seawater pH caused by reduced weathering in the early Permian would lead not only to the dissolution of carbonates releasing light boron back into the system but also to a depletion of borate in solution throughout the ocean. At the modern seawater pH of 8.2, the proportion of aqueous boron species is approximately 70% boric acid (B (OH)₃) and 30% borate (B (OH)₄⁻; Figure 7). Based on the modeling study of Ridgwell (2005), the seawater pH was 8.05 during the Bashkirian at 320 Ma and dropped to 7.85 by the Roadian at 270 Ma. According to the seawater boron speciation graph (Figure 7; Hemming & Hanson, 1992), the availability of borate drops by 23% from modern seawater pH to Bashkirian seawater pH, while the change from Bashkirian to Rodian seawater pH shows a borate availability decrease of 31%. Because boron removal from seawater via coprecipitation in carbonates and adsorption onto clays relies almost exclusively on the availability of borate, a decrease in the borate composition of seawater due to lowered pH would directly and significantly impact the amount of boron outflux from seawater.