Characterizing Phosphorus Speciation of Chesapeake Bay Sediments

Dec 3, 2014 - *Phone: +86(25)-836-86042; e-mail: [email protected]., *Phone: ... Additionally, solution 31P NMR results show that the sediments fro...
5 downloads 0 Views 3MB Size
Download PDF
Article pubs.acs.org/est

Characterizing Phosphorus Speciation of Chesapeake Bay Sediments Using Chemical Extraction, 31P NMR, and X‑ray Absorption Fine Structure Spectroscopy Wei Li,*,†,‡ Sunendra R. Joshi,† Guangjin Hou,§ David J. Burdige,∥ Donald L. Sparks,† and Deb P. Jaisi*,† †

Department of Plant and Soil Sciences, University of Delaware, Newark, Delaware 19716, United States Key Laboratory of Surficial Geochemistry, Ministry of Education, School of Earth Sciences and Engineering, Nanjing University, Nanjing 210093, People’s Republic of China § Department of Chemistry and Biochemistry, University of Delaware, Newark, Delaware 19716, United States ∥ Department of Ocean, Earth, and Atmospheric Sciences, Old Dominion University, Norfolk, Virginia 23529, United States ‡

S Supporting Information *

ABSTRACT: Nutrient contamination has been one of the lingering issues in the Chesapeake Bay because the bay restoration is complicated by temporally and seasonally variable nutrient sources and complex interaction between imported and regenerated nutrients. Differential reactivity of sedimentary phosphorus (P) pools in response to imposed biogeochemical conditions can record past sediment history and therefore a detailed sediment P speciation may provide information on P cycling particularly the stability of a P pool and the formation of one pool at the expense of another. This study examined sediment P speciation from three sites in the Chesapeake Bay: (i) a North site in the upstream bay, (ii) a middle site in the central bay dominated by seasonally hypoxic bottom water, and (iii) a South site at the bay-ocean boundary using a combination of sequential P extraction (SEDEX) and spectroscopic techniques, including 31P NMR, P X-ray absorption near edge structure spectroscopy (XANES), and Fe extended Xray absorption fine structure (EXAFS). Results from sequential P extraction reveal that sediment P is composed predominantly of ferric Fe-bound P and authigenic P, which was further confirmed by solid-state 31P NMR, XANES, and EXAFS analyses. Additionally, solution 31P NMR results show that the sediments from the middle site contain high amounts of organic P such as monoesters and diesters, compared to the other two sites, but that these compounds rapidly decrease with sediment depth indicating remineralized P could have precipitated as authigenic P. Fe EXAFS enabled to identify the changes in Fe mineral composition and P sinks in response to imposed redox condition in the middle site sediments. The presence of lepidocrocite, vermiculite, and Fe smectite in the middle site sediments indicates that some ferric Fe minerals can still be present along with pyrite and vivianite, and that ferric Fe-bound P pool can be a major P sink in anoxic sediments. These results provide improved insights into sediment P dynamics, particularly the rapid remineralization of organic P and the stability of Fe minerals and the ferric Fe-bound P pool in anoxic sediments in the Chesapeake Bay.

■

INTRODUCTION The Chesapeake Bay is one of the largest and most biologically productive estuaries in the United States.1−3 Elevated nutrient loading in the bay during the last 100 years has led to extensive eutrophication on the surface water and hypoxia in the bottom water.4−6 The long-term (1985−2010) monthly mean ratio of dissolved inorganic nitrogen to phosphorus (DIN: DIP) in the Bay is higher than the Redfield ratio (i.e., 16 for N/P) in all seasons except midsummer7 suggesting that phosphorus (P) is the limiting nutrient in all seasons and until the onset of phytoplankton blooms in the Bay, unless nutrient concentrations are higher than biological demand. Understanding the sources, cycling, and transformation of P is not straightforward, compared to many other nutrients, particularly because of (i) a low concentration of readily bioavailable (dissolved) P © 2014 American Chemical Society

compared to colloidal/particulate and sediment P; (ii) active but variable transformations of organic and inorganic P; (iii) co-occurring biotic and abiotic reactions that sorb/desorb, dissolve/precipitate, and cycle P at various temporal and spatial scales; and (iv) organisms’ variable strategies for P uptake and cycling at different concentrations and compositions of organic and inorganic P compounds. These complexities not only hamper the identification of sources and processes of P in these forms, but also restrict the formulation of effective guidelines for nutrient management plans in any ecosystem. Therefore, a Received: Revised: Accepted: Published: 203

September 22, 2014 December 2, 2014 December 3, 2014 December 3, 2014 dx.doi.org/10.1021/es504648d | Environ. Sci. Technol. 2015, 49, 203−211

Environmental Science & Technology

Article

ferric Fe-bound P; (ii) identify whether ferric Fe-bound P could be a major P sink in anoxic sediments, and (iii) identify how bottom water hypoxia/anoxia impacts sediment P speciation particularly the sources of P for authigenic P precipitation. Our results provide important insights into the interplay of P partitioning, remobilization, and precipitation in the sediment column in response to bottom water hypoxia and sediment anoxia.

systematic approach aimed at identifying the composition and stability of P species and P sinks in the sediment could be the first step to better understand P dynamics and biogeochemical cycling in the bay. In the Chesapeake Bay, sediments play a dual role of P deposition and removal.8,9 On the one hand, sediments are steadily deposited and have stored large amount of P in the sediment.8−10 On the other hand, strong summer stratification produces bottom water hypoxic conditions which induces sulfate reduction, and the H2S thus produced promotes reductive dissolution of Fe oxides and release of P associated with these oxides to bottom waters.11 This process generates a fresh pool of inorganic P, which is estimated to be ∼25−30% to even >100% of the burial P flux.10−13 Because of the differential reactivity of sedimentary P pools in response to redox conditions, diagenesis, and biological cycling, P speciation in the sediment column varies accordingly. A detailed analysis of P speciation provides information on P cycling, particularly the formation of one pool at the expense of another and the stability of a particular P pool. This information enables a better understanding of P regeneration and reintroduction into the water column and thus how P cycling is impacted due to hypoxic bottom water conditions. Existing knowledge on P speciation in marine sediments relies heavily on the sequential extraction method (SEDEX).14 The SEDEX method separates sediment P into five operationally defined pools and has largely improved the understanding of marine P biogeochemistry particularly the formation of authigenic P and source-sink switching in sediments. However, it suffers from intrinsic limitations owning to artifacts associated with reagent selectivity and inability to differentiate specific solid-state P forms (e.g., amorphous versus crystalline mineral P phases). Additionally, quantification of organic P species is not possible because “labile” organic P is partially removed in different extraction steps and unaccounted for before residual/ recalcitrant organic P is extracted in the last step in the SEDEX protocol. In recent years, advanced spectroscopic techniques such as X-ray absorption near edge structure spectroscopy (XANES) and nuclear magnetic resonance (NMR) spectroscopy have been increasingly applied to complement the SEDEX and other extraction methods.15,16 Both P K-edge XANES and solid-state 31P NMR spectroscopy are nondestructive, elementspecific techniques that can probe the local molecular bonding environment around P atoms17−24 and thereby can be used to determine P species. Solid-state NMR can readily distinguish Ca-bound P and Al-bound P, as the chemical shifts for these two phases are well separated.24 Furthermore, P K-edge XANES complements solid-state NMR spectroscopy. For example, ferric Fe-bound P is invisible to NMR spectroscopy due to the paramagnetic effect but can be detected in XANES. However, both techniques are not sensitive enough to organic P species. But solution 31P NMR spectroscopy can be used to identify the composition of organic P compounds.25−31 In summary, individual techniques have certain limitations, but a combination of complementary techniques better constrains the P speciation results and has been used successfully in different environments.16,17,31 In this study, we combined the chemical sequential extraction method with in situ spectroscopic techniques to identify and constrain sediment P speciation in three selected sites in the Chesapeake Bay. The major objectives of our research are to (i) apply a suites of P specific techniques to characterize molecularlevel P speciation in sediments and constrain the identity of

■

EXPERIMENTAL SECTION Sampling Sites and Sediment Characterization. The three sampling sites were located on the western slope of the Chesapeake Bay central channel (Supporting Information (SI) Figure S1). Physical and biogeochemical characteristics of these sites are summarized in SI Table S1.4,32,33 We ran synchrotron based X-ray diffraction on sediment at beamline X7B at the National Synchrotron Light Source, Brookhaven National Laboratory (Upton, NY). The two-dimensional XRD patterns were calibrated with lanthanum hexaboride (LaB6, NIST 660a) and integrated to generate XRD profile of intensity versus 2θ with Fit2d.34 Elemental analysis and X-ray diffraction results are included in SI Table S2 and Figure S2. Sequential Extraction of P Pools. The P speciation in the Chesapeake Bay sediments was determined using the sequential extraction method (SEDEX).14 This method allows for differentiation and quantification of five major P pools: (i) exchangeable or loosely sorbed P, (ii) easily reducible or reactive ferric Fe-bound P, (iii) authigenic P carbonate fluoroapatite, biogenic apatite, and CaCO3-bound P, (iv) detrital P, and (v) residual P. The sequential extraction was performed using 0.20 g sediment at a constant solid/solution ratio of 1:100 for all samples. The P concentration extracted from each fractionation was analyzed using the standard molybdate blue method35 or ICP-OES (for CDB extracted P due to color interference14 in molybdate blue method). We also concentrated P in CDB extracted solution by using MagIC (magnesium-induced coprecipitation) method and quantified colorimetically. The butanol extraction method for quantifying CDB extracted P, originally purposed by Ruttenberg,14 was not adopted due to its poor precision among different samples.14 Solution 31P NMR Analysis of Organic P Species. Sediments were extracted using the method developed by Cade-Menun36 for solution 31P NMR analysis. In brief, 2.0 g of sediment was mixed with 30 mL of a NaOH/EDTA solution (i.e., 0.25 mol L−1 NaOH and 0.05 mol L−1 EDTA), and shaken for 16 h at 25 °C. After separating the supernatant by centrifugation for 30 min at 10 000g, it was further filtered through a 0.2 μm syringe membrane filter (Waterman). Subsequently the supernatant was frozen at −80 °C and lyophilized, which yielded approximately 600 mg of lyophilized sample. For solution 31P NMR analysis, 600 mg of lyophilized material was dissolved in 2.5 mL of a 1 M NaOH solution, but only 0.6 mL solution plus 0.1 mL D2O was pipetted into a 5 mm diameter NMR tube for one analysis. Liquid-state 31P NMR spectra were obtained on a Bruker AV400 NMR spectrometer (9.4 T) equipped with a cryogenic quadruple nucleus probe (QNP) (i.e., 1H, 13C, 19F, and 31P), at the operating frequency of 161.8 MHz for 31P. The NMR parameters were 30° pulse (i.e., 3.07 μs pulse width), 0.68 s acquisition time, 2.0 s pulse delay, 25 °C, and ∼3600 scans. The spectra were processed using a commercial NMR software package (MestRenova) and plotted with a line broadening of 1 204

dx.doi.org/10.1021/es504648d | Environ. Sci. Technol. 2015, 49, 203−211

Environmental Science & Technology

Article

Hz. The 31P chemical shifts (δP‑31) are reported relative to an external 85% H3PO4 solution at 0 ppm. 31 P Solid-State NMR Spectroscopy. Solid-state 31P{1H} single pulse (SP) magic-angle-spinning (MAS) NMR spectra for several sediment samples were collected on a 500 MHz Bruker AVIII solid-state spectrometer (11.7 T), at operating frequency of 202.5 MHz for 31P, and 500.1 MHz for 1H. A Bruker 4.0 mm HX double resonance probe was used for all solid-state NMR measurements. Powder sediment was packed into a rotor and spun at the MAS frequency of 10 kHz. The 31P 30° pulse length was 0.8 μs, and the relaxation delay was 30 s. 1 H SPINAL-64 heteronuclear decoupling37 with field strength of 90−100 kHz was used during the acquisition period. More than ∼3000 scans were collected for each sample depending on signal sensitivity. The 31P chemical shifts (δP‑31) are reported relative to an external 85% H3PO4 solution at δP‑31 = 0 ppm. Phosphorus K-Edge XANES Spectroscopy. All sediment samples and reference compounds were analyzed by XANES at beamline X15B at the National Synchrotron Light Source (NSLS), Brookhaven National Laboratory (Upton, NY). The electron beam energy was 2.5 GeV at the maximum beam current of 300 mA. The XANES spectra were collected in the fluorescence mode in a He chamber at room temperature using a Germanium multiple element detector. A finely ground, homogenized dry sediment powder was spread as a thin film in a custom-built sample holder, an envelope made from polyethylene membrane (with zero P background). The incident X-ray energy was calibrated against a standard (e.g., fluorapatite). Spectra were obtained by averaging 25−40 scans to achieve a good signal-to-noise ratio. Data processing was performed using the software package ATHENA.38 Iron K-Edge EXAFS Spectroscopy. Extended X-ray absorption fine structure (EXAFS) spectroscopy data were collected for sediments from the site M and several Fe reference minerals at beamline X11A at the NSLS and beamline 14W at the Shanghai Synchrotron Radiation Facility. The sediment sample was mounted in a thin custom-built plastic sample holder covered with Kapton tape and placed at 45° to the incident X-ray beam. This allows the EXAFS data to be collected in both fluorescence and transmission mode using a Lytle detector positioned 90° to the beam and the ionization chamber parallel to the beam, respectively. A 3 mm-thin Fe foil (K-edge 7712 eV) was used for energy calibration. A pair of Si (111) crystals was used as the monochromator, with one crystal detfe spuned by 40% for harmonic rejection. The EXAFS spectra for each sample were averaged after energy calibration and proper background subtraction. The χ(k) function was Fourier transformed using k3 weighting. The contribution of different Fe species to total Fe fitting was calculated by linear combination fitting (LCF) in the k3 weighted χ-space (3−12 Å−1) using the software package ATHENA.38

Figure 1. Sediment P pools from North (N), middle (M), and South (S) sites of the Bay. The sediments were sequentially extracted by using 1 mol L−1 MgCl2 (pH 8), citrate-dithionite-bicarbonate (CDB) solution (pH 8.4), 1 mol L−1 Na-acetate (pH 4), and 1 mol L−1 HCl.

Ferric Fe-bound P constitutes the largest fraction of P reservoir in the sediment, with depth average concentrations ranging from 7.39−11.77 μmol g−1. In general, the concentration of the ferric Fe-bound P pool rapidly decreases with depth in the upper 0−5 cm in all sites as expected (e.g., ref 39). However, this P pool shows a marked increase in concentration below 5 cm in the North and South sites, but a small and gradual decrease in the middle site. Please note that there was a methodological difference for P concentration measurement for ferric Fe-bound P (CDB extracted P was measured by ICPOES as well colorimetrically after concentrating P by using the MagIC method). Ferric Fe-bound P concentration measured by using the MagIC method was ∼30−50% of less than that measured directly by the ICP-OES method. Overestimated P in the ICP-OES method is most likely due to the fact that it is total P (inorganic plus organic P extracted by CDB) including unknown plasma interference by CDB reagents and P loss during MagIC concentration, if any. The highest concentration of ferric Fe-bound P pool in all sites, and particularly at the middle (M) site that often becomes euxinic with substantial hydrogen sulfide40,41 is rather unexpected because it is generally assumed that the majority of Fe oxides undergoes reductive dissolution and releases P in hypoxic/anoxic sediment columns. Nonetheless no major dissolution of Fe oxides was reported in a permanently stratified lake with dissolved O2 3%) among the three sites (SI Table S1). The lower organic P content in site S sediments could be due to mixed redox (oxic-anoxic) conditions and heavy bioturbation and bioirrigation.4,33 On average, SEDEX inorganic P and solution 31P NMR account for ∼75% of the total sediment P indicating that ∼25% P is residual, and EDTA/NaOH reagents extracted 24−28% of total P (SI Table S3). 31 P Solid-State NMR Spectroscopy. Solid-state 31P NMR allows further direct characterization of the solid-state P speciation in sediments by analyzing the chemical shifts of P nuclei in samples. For example, all calcium phosphate minerals yield chemical shifts at −2 to 3 ppm, whereas crystalline aluminum phosphates usually yield chemical shifts from −10 to −30 ppm. The marked negative chemical shifts of aluminum phosphates result from the shielding effect of the Al to P nucleus via a P−O−Al covalent bond. For instance, brazilianite [NaAl3(OH)4(PO4)2] yields a peak at δP = −10.2 ppm, wavellite [Al3(OH)3(PO4)2] at −11 ppm, metavarisite [AlPO4· 2H2O] at −13.2 ppm and variscite [AlPO4·2H2O] at −19 ppm.46 Figure 3 shows the 31P solid state NMR results for selected sediments from the three sites studied in the Chesapeake Bay. Due to the relatively low P content in the sediments, the signal-to-noise ratios for all spectra are relatively low even after unusually long scan times (e.g., 5 days for N12− 15). Nevertheless, a dominant peak at 2.6 ppm was clearly observed for all samples analyzed. Based on the chemical shift, this peak can be assigned to an apatite-group mineral. However, it is not possible to further identify specific apatite mineral such as hydroxyl apatite (HAP) or fluorapatite (FAP), or carbonate fluorapatite (CFP), because of the extremely close chemical shifts among apatite minerals. Because carbonate fluorapatite

Authigenic P constitutes the second largest fraction in the sediment, with depth average concentration varying from 2.07 to 4.96 μmol g−1 (15.81 to 28.70% of all extracted P) (Figure 1). As expected, the middle (M) site has the highest concentration of this P pool and the North (N) site the lowest. Overall, concentrations of authigenic and ferric Febound P form approximate mirror-image profiles particularly in the upper 0−10 cm. Such profiles along with the stoichiometry of inorganic constituents in the porewater particularly sulfate, ammonia, and reactive P, are classically used as evidence of the sink switching as P from one pool is transformed into another. While it is generally assumed that P released from the reductive dissolution of Fe-oxides could precipitate as authigenic P, this assumption may not necessarily be true because the sequential extraction protocol does not account for labile organic P that is extracted in all SEDEX steps except for the final (ashing) step as residual/recalcitrant organic P. It means labile organic P hydrolysis at or below sediment-water interface and inorganic P produced is unaccounted for in the SEDEX method. However, the extent of organic P extracted in a specific step, including reagent-driven hydrolysis ultimately depends on the amount and composition of organic P. Given that a significant amount of organic P (e.g., extracted by NaOH-EDTA) is mineralized at a shallow depth (Figure 2a), it is likely that inorganic P released from remineralization of organic P is precipitated as authigenic P. Solution 31P NMR Spectroscopy. Solution 31P NMR can distinguish different P species based on the chemical shift because the shift in frequency is based on the chemical environment of a particular P species. In the series of spectra

Figure 2. 31P solution state NMR spectra for Chesapeake Bay sediments extracted with 0.25 mol L−1 NaOH/0.05 mol L−1 EDTA. Sediments were collected from sites M (a), N (b) and S (c) at different depths. The label “N0−1” refers to the sediment collected at 0−1 cm of N site. The 31P NMR chemical shifts suggest the presence of orthophosphate (6 ppm), phosphate monoesters (4−5 ppm), phospholipids (1 to 2.5 ppm), DNA (−1 ppm), and pyrophosphate (−4 to −5 ppm). 206

dx.doi.org/10.1021/es504648d | Environ. Sci. Technol. 2015, 49, 203−211

Environmental Science & Technology

Article

organic P) in the water column/sediment can precipitate via chemical reaction with the saline water in the Chesapeake Bay which contains a large amount of dissolved calcium.49 According to thermodynamic calculation, for bottom water pH (pH 7.5) and [Ca] concentration (∼5 mmol L−1) at the site M, the minimum concentration for dissolved phosphate needed to precipitate Ca−P is ∼0.1 μmol L−1. This value is much lower than the dissolved phosphate concentration in the sediment pore water (i.e., >250 μmol L−1 for site M). This means that the precipitation of authigenic apatite is energetically possible through numerous phosphate intermediates or precursors. But the higher dissolved P concentration could be due to the other ions and compounds present in the porewater that inhibit apatite precipitation. The second pathway is the transformation of other minerals, such as carbonates, to authigenic apatite. In this case, dissolved phosphate in the pore water is adsorbed by calcium carbonates, and the anion sites are partially (or even fully) replaced to produce amorphous calcium phosphate (phosphatization),50 a precursor of apatite. Irrespective of the pathway of formation, authigenic P is the most stable P sink in all three sites in the Chesapeake Bay. P XANES. P XANES data were collected on P containing reference compounds and selected marine sediments (Figure 4a,b). P XANES spectra for mineral standards exhibited distinct

Figure 3. 31P solid state NMR spectra for six Chesapeake Bay sediments collected at two depths from three sites. Peak at 2.6 ppm is indicative of apatite (Ca−P mineral). Spectra were acquired at the single-pulse magic angle spinning (MAS) condition at a spinning rate of 10 kHz, with a π/6 pulse and a pulse delay of 10 s. Approximately 3000−14 000 scans were accumulated. Sediment labels are same as in Figure 2

[Ca5(PO4,CO3,OH)3F] is the most dominant apatite in the marine environment,48 authigenic apatite in the sediemnts is most likely a carbonate fluorapatite. Identification of Fe−P minerals such as vivianite and strengite is not possible by using 31 P NMR because measurement of the chemical shifts of Febearing compounds has long been a challenge in NMR analysis. Very recently, chemical shift of strengite was successfully measured to be ∼15 880 ppm,47 an extremely large shift far beyond the regular observation range. Based on the sequential extraction and NMR results, it is relevant to discuss the origin and form of authigenic P. In marine sediments, there are three major sedimentary P pools: (i) P associated with organic matter; (ii) P associated with ferric Fe-minerals; and (iii) P in authigenic apatite. Our sequential extraction data show that a significant amount of authigenic P is present in site M sediments (Figure 1). On the other hand, detrital P is almost negligible. This means that the sinking particulate P in the water column does not contain much detrital apatite and that the apatite found in the sediments is almost exclusively authigenic by origin. The authigenic apatite could be formed by two possible pathways. First, dissolved phosphate (the source of which could be reductive dissolution of ferric Fe minerals or remineralization of

Figure 4. P K-edge X-ray absorption near edge spectra (XANES) for several phosphate minerals (a) and three Chesapeake Bay sediments (b). Sediment labels are same as in Figure 2

spectral features in the pre-edge, white-line peak or postedge region, which allowed differentiation of solid state P species.51−53 Specifically, apatite minerals, which include apatite, carbonate apatite, and fluorapatite, have two postedge resonances between 2161 and 2182 eV, with peaks at 2164 and 2170 eV. A closer inspection indicates that carbonate apatite and fluorapatite have a shoulder peak at 2155 eV, whereas other apatite minerals do not, which is probably due to different degrees of crystalinity.38 The P K-edge XANES spectra of two aluminum phosphates, wavellite [Al3(OH) 3(PO4)2] and 207

dx.doi.org/10.1021/es504648d | Environ. Sci. Technol. 2015, 49, 203−211

Environmental Science & Technology

Article

variscite [AlPO4·2H2O], have similar trends, in particular, a close white-line peak (2154 eV) and a postedge peak at 2160 eV, but differences are still clear in the broad features between 2165 and 2180 eV, which reflects slight differences in Al:P stoichiometery between similar minerals such as variscite and wavelite. Two iron phosphates, strengite [FePO4·2(H2O)] and vivianite [Fe3(PO4)2·8(H2O)], have similar postedge features, but strengite has a unique pre-edge feature at 2150 eV due to the presence of Fe(III) in the mineral. The pre-edge feature is not present in the Fe(II) phosphate minerals such as vivianite, unless they are partially oxidized.51,53 Low P content in the sediment generates relatively poor signal-to-noise ratios even after averaging 30−48 scans (which took almost 24 h on the beamline). All three sediments yield white-line peaks at 2153−2154 eV. For the N0−1 sample, two clear postedge peaks were observed at 2163 and 2170 eV, consistent with those for apatite minerals. This finding agrees well with the results from solid state NMR analysis. But for M0−1 and S0−1 samples, the postedge features are not as clearly resolved as in N0−1, which yielded broad features between 2160 and 2180 eV. As discussed above, apatite minerals yield a signal in this range, and Fe phosphate such as strengite and vivianite also show features at 2164 and 2170 eV. As a result, it is hard to exclude the presence of iron phosphate from P XANES requiring Fe EXAFS to identify the presence of iron phosphate minerals (see below). Fe Speciation in the Sediments. Since P chemistry is often coupled with Fe in marine sediments, knowledge on Fe mineralogy provides information useful to understand transformations of these minerals in response to hypoxia and other environmental conditions as well as their roles in retaining or releasing P from the sediment column.8 Fe K-edge EXAFS54,55 for site M sediments at five depths (Figure 5) are, in general, similar especially at low K range. This suggests that these sediments may contain very similar types of iron containing minerals. To fit K-edge EXAFS spectra from sediments, we used a wide range of naturally occurring Fe-containing minerals that are likely present in the marine sedimentary environment, including ferric hydroxides [ferrihydrite, goethite (αFeOOH), and lepedocrocite (β-FeOOH)], iron phosphate [strengite (FePO4·2(H2O)) and vivianite (Fe3(PO4)2·8(H2O))], iron sulfide [pyrite (FeS2) and mackinawite (FeS)], iron-bearing clays (Fe-vermiculite and Fe-smectite), and ironbased layered double hydroxide (nikischerite and green rust). Linear combination fitting (Figure 5) suggests that five common Fe minerals (Fe-vermiculite, pyrite, nekischerite, vivianite, and lepidocrocite) are present in these sediments but at different proportions. For example, Fe(III)-vermiculite is the major Fe mineral and accounts for ∼33−51% of the total Fe. Pyrite (independently confirmed by XRD, SI Figure S2) is another important mineral and accounts for ∼21% of the Fe in the deep sediments (M12−15 and M30−32) and ∼10.4% in the upper 0−1 cm sediment (M0−1). This difference is consistent with the previous result 12 that shows the concentration of pyrite increases with depth. Prolonged exposure to severe anoxia in the deeper sediments results in the reductive dissolution of Fe(III) minerals which generates dissolved Fe2+ that in sulfidic environment induces the precipitation of pyrite with reduced sulfur species.56,57 The vivianite concentration increased from 7.6% to 18.2% with increasing sediment depth. The presence of vivianite and siderite has been reported before in the Chesapeake Bay58 and

Figure 5. Linear combination fitting results for bulk Fe EXAFS spectra of the Chesapeake Bay sediments (a−e) collected at different depths at the site M, and k3 weighted bulk EXAFS spectra for reference Fe compounds (f−m). Black solid lines represent the k3 weighted χspectra and the red dashed lines represent the best fits obtained using linear least-squares fitting. Sediment labels are same as in Figure 2

some other sedimentary environments42 particularly where reductive dissolution of Fe oxides occurs. The high porewater phosphate concentration at the site M (>250 μmol L−1) suggests that the presence of at least 0.2 μmol L−1 [Fe2+] could lead to the solution being supersaturated with respect to vivianite based on a thermodynamic calculation. The dissolved Fe concentration is quite high (>250 μmol L−1)59 at 0−2 cm depth in this site but rapidly decreases and becomes insignificant below 3 cm, indicating formation of vivianite and iron sulfide minerals likely occurs within the 0−3 cm depth. Comparison of Sequential Extraction and Spectroscopic Techniques. In order to gain a comprehensive understanding of the P species buried in the Chesapeake Bay sediments, we employed four methods namely, (i) sequential chemical extraction, (ii) solution 31P NMR, (iii) solid state 31P NMR, and (iv) P K-edge XANES. Here, we discuss their merits and limitations. The sequential chemical extraction is a traditional technique to characterize sediment P speciation. Although it suffers from the limitation as an indirect method, we found it is still a useful tool and allowed for comparisons of P pools among different sites and depths. As described above citrate-dithionite-bicarbonate (CDB) extraction could represent more P pools than those P species classically presumed.45 It is yet unclear what fraction of P in vivianite and other reduced Fe 208

dx.doi.org/10.1021/es504648d | Environ. Sci. Technol. 2015, 49, 203−211

Environmental Science & Technology

Article

scientific basis for a more accurate assessment of P source and sink and the long-term efficacy of current management strategies for minimizing P pollution in the bay.

minerals (occluded and/or precipitated) is extracted by CDB but the ferric-Fe-bound P pool includes those in Fe(III) minerals and partly Fe(II) minerals as well (at least vivianite). However, a direct characterization of the Fe speciation by Fe EXAFS in the Chesapeake Bay sediments shows the presence of abundant Fe(III)-bearing minerals in the sediment column (Figure 5) indicating that some of these minerals are recalcitrant in anoxic sediments. While this is not expected, it is, however, not surprising given that a strong reductant such CDB [with standard oxidation potential (Eo) of 1.12 V] has been found to incompletely dissolve goethite and hematite.60 The extent of Fe(III) reduction by CDB in ferruginous phyllosilicates is low (varies from 11 to 100%).61 This wide range of Fe(III) reduction is due to the contrasting crystal chemical environment of Fe(III) sites as well as layer charge of the overall unit cell structure.61−63 The Fe EXAFS results indicated the presence of vivianite at all depths of the site M sediments, with the ratio of vivianite to total Fe increasing with sediment depth. The redox condition and porewater chemistry (SI Table S1) is supportive of precipitation of Fe(II)-bearing minerals such as vivianite and pyrite and the presence of these minerals has been reported from this site.64 Overall, ferric Febound P pool can still be the largest P pool in anoxic and sulfidic sediments in the Chesapeake Bay similar to that in the Baltic Sea.45 Spectroscopic techniques are a good addition to sequential extraction techniques because they provide direct identification of important P phases. Each spectroscopic technique, however, has better specificity or sensitivity to certain P species than to others. For example, solid-state 31P NMR spectroscopy and P K-edge XANES provide evidence for the presence of apatite, and solution 31P NMR identifies the composition of organic P. Spectroscopic techniques are designed for structural analysis and normally the quantitative detection limits are not reported. To analyze samples with low concentration, either enrichment of the element/phase of interest and/or increasing the number of scans is required. In this study, most samples were analyzed by either solid-state 31P NMR or XANES which took nearly 24 h or as long as 5 days. Based on our study and a survey of literature, spectroscopic analysis on samples with P concentration of 1000 ppm or higher would yield good results and samples with