Journal of Contaminant Hydrology 166 (2014) 52–63

Contents lists available at ScienceDirect

Journal of Contaminant Hydrology journal homepage: www.elsevier.com/locate/jconhyd

Nitroglycerin degradation mediated by soil organic carbon under aerobic conditions Geneviève Bordeleau a,⁎, Richard Martel a, Abraham N'Valoua Bamba b, Jean-François Blais a, Guy Ampleman c, Sonia Thiboutot c a

Institut National de la Recherche Scientifique, Centre Eau, Terre et Environnement (INRS-ETE), 490 de la Couronne, Quebec City, QC, Canada, G1K 9A9 Université Laval, Département de Géographie, Faculté de foresterie, de géographie et de géomatique, Pavillon Abitibi-Price, 2405, rue de la Terrasse, Local 3137, Quebec City, QC, Canada, G1V 0A6 c Defence Research and Development Canada – Valcartier, 2459 Pie-XI Blvd. North, Quebec City, QC, Canada, G3J 1X5 b

a r t i c l e

i n f o

Article history: Received 25 November 2013 Received in revised form 20 June 2014 Accepted 27 June 2014 Available online 6 July 2014 Keywords: Nitroglycerin Propellant Military training range Natural attenuation Soil organic matter

a b s t r a c t The presence of nitroglycerin (NG) has been reported in shallow soils and pore water of several military training ranges. In this context, NG concentrations can be reduced through various natural attenuation processes, but these have not been thoroughly documented. This study aimed at investigating the role of soil organic matter (SOM) in the natural attenuation of NG, under aerobic conditions typical of shallow soils. The role of SOM in NG degradation has already been documented under anoxic conditions, and was attributed to SOM-mediated electron transfer involving different reducing agents. However, unsaturated soils are usually well-oxygenated, and it was not clear whether SOM could participate in NG degradation under these conditions. Our results from batch- and column-type experiments clearly demonstrate that in presence of dissolved organic matter (DOM) leached from a natural soil, partial NG degradation can be achieved. In presence of particulate organic matter (POM) from the same soil, complete NG degradation was achieved. Furthermore, POM caused rapid sorption of NG, which should result in NG retention in the organic matter-rich shallow horizons of the soil profile, thus promoting degradation. Based on degradation products, the reaction pathway appears to be reductive, in spite of the aerobic conditions. The relatively rapid reaction rates suggest that this process could significantly participate in the natural attenuation of NG, both on military training ranges and in contaminated soil at production facilities. © 2014 Elsevier B.V. All rights reserved.

1. Introduction

− Abbreviations: NG, nitroglycerin; NO− 2 , nitrite; NO3 , nitrate; DNG, dinitroglycerin; MNG, mononitroglycerin; POM, particulate organic matter; DOM, dissolved organic matter; SOM, soil organic matter (comprises POM and DOM); NDOM, natural dissolved organic matter ⁎ Corresponding author. Tel.: +1 418 654 2530x4485. E-mail addresses: [email protected] (G. Bordeleau), [email protected] (R. Martel), [email protected] (A.N. Bamba), [email protected] (J.-F. Blais), [email protected] (G. Ampleman), [email protected] (S. Thiboutot).

http://dx.doi.org/10.1016/j.jconhyd.2014.06.012 0169-7722/© 2014 Elsevier B.V. All rights reserved.

Nitroglycerin (NG) has often been detected in soils on military training ranges, as a result of incomplete combustion of the propellant used for training activities such as firing of anti-tank rockets, artillery, machine gun grenades, and small arms (Walsh et al., 2012). Once deposited on the ground, the NG located within propellant residues can partially dissolve in infiltration water and migrate downward in the soil profile, eventually reaching the water table, and potentially discharging to surface water bodies. The presence of NG in groundwater and surface water is of concern, as prolonged

G. Bordeleau et al. / Journal of Contaminant Hydrology 166 (2014) 52–63

exposure to this compound may cause severe and diverse health effects on humans (Bhaumik et al., 1997), as well as toxicity to microorganisms and wildlife (Burton et al., 1993). To date, the presence of NG has been very sparsely documented in groundwater on military training ranges. However, at the firing position of an active anti-tank range in Eastern Canada, NG and its common degradation products (dinitroglycerin (1,2- and 1,3-DNG), mononitroglycerin (1- and 2-MNG), nitrate (NO− 3 )) have been detected in water from the unsaturated zone (pore water), at depths up to 5.0 m below ground surface (Bordeleau et al., 2012a). The presence of these degradation products is an indication that natural attenuation of NG is occurring. Such natural attenuation can be the result of various processes, which have not been fully characterized until now. The two main degradation processes that have been documented for NG are biodegradation (Bernstein et al., 2010; Clausen et al., 2011; Husserl et al., 2010; Jenkins et al., 2003; Pennington and Brannon, 2002) and photolysis (Bedford et al., 1996; Bordeleau et al., 2013; Pennington et al., 2001). Both processes proceed through sequential denitration of the NG molecule, releasing a nitrite (NO− 2 ) ion at each step, which can then oxidize to NO− 3 in presence of oxygen. However, in several biodegradation studies, the released NO− 2 was consumed by the microorganisms and was therefore not found in solution (Accashian et al., 2000; Ducrocq et al., 1989; White et al., 1996). Likewise, photolysis of solid NG-bearing propellant particles located at the soil surface was shown to produce − relatively little NO− 2 or NO3 available for dissolution in infiltration water (Bordeleau et al., 2013). Therefore, these two processes may not satisfactorily explain the high NO-3 concentrations observed at the anti-tank firing position, which underlines the necessity to better characterize the fate of NG in the subsurface. One potential process that has been very sparsely documented is the degradation of energetic materials mediated by the presence of organic carbon. Indeed, organic carbon was traditionally only considered to cause sorption of energetic materials (Clausen et al., 2011), but in the last decade, it was also recognized to mediate the reduction of organic contaminants by promoting electron transfer (Tratnyek et al., 2001). Such reaction has been documented in a few studies for energetic materials, such as NG, trinitrotoluene (TNT), and 1,3,5-hexahydro-1,3,5-trinitro-1,3,5-triazine (RDX). However, these studies focused on degradation in anoxic conditions, which provide favourable conditions for reduction reactions. Such conditions are common in lacustrine or marine sediments, in saturated soils with high organic matter content, or they may be induced in ex-situ remediation strategies. On training ranges, however, most of the NG is located at the soil surface and in the first few centimeters of the soil profile, where soils are unsaturated and conditions are aerobic. One study has shown that unsaturated soils could support degradation of TNT in presence of organic carbon (Singh et al., 2008). Degradation of NG mediated by soil organic matter (SOM) was briefly discussed in another study, but the authors mentioned that aerobic conditions resulted in very slow degradation compared to anaerobic conditions, and they did not explore this reaction further (Saad et al.,

53

2010). Aerobic SOM-related NG degradation may therefore not be suitable for active remediation strategies, where process efficiency is sought, however in a natural attenuation context where most degradation reactions are slow, it could be a relevant process. The goal of this study is therefore to determine whether the presence of SOM (in the form of particulate organic matter (POM) and/or dissolved organic matter (DOM)) may significantly participate in the natural attenuation of NG, and the concomitant production of NO− 3 under aerobic conditions. More specifically, the objectives were to: 1) verify whether POM causes sorption and/or degradation of NG; 2) distinguish the roles of POM and DOM in NG degradation; and 3) investigate the degradation mechanism involved. These issues were first investigated through a series of batch-type experiments, where concentrations of NG and its common degradation products were monitored over time. Experimental conditions were chosen to favour NG degradation, i.e. there was permanent contact and stirring of NG solution with soil having a relatively high (up to 4.5%) organic carbon content. The results obtained from batch experiments were then verified in column-type experiments, where conditions were less ideal but more representative of several of the contaminated zones on military ranges, i.e. infiltration of precipitation water into shallow, well-oxygenated contaminated soils with moderate organic carbon content.

2. Experimentals 2.1. Chemicals and reagents For batch-type experiments, spiking of treatments with NG was achieved using a concentrated (220 mg/L) NG stock solution. The stock solution was prepared by stirring doublebase propellant grains (Powder C propellant: 65% NC, 34% NG, and 1% ethyl centralite) in distilled water for 7 days, which causes the dissolution of some of the NG contained in the propellant. The solution was then filtered on a 0.45 μm nylon membrane, and kept at 4 °C in an amber glass bottle. Spiking of experimental treatments with NO− 2 was achieved using a stock solution (5.2 mg N-NO− 2 /L) prepared from ACS-grade NaNO2 (EMD Millipore, Billerica, MA). For column-type experiments, fragments of AKB 204 propellant (61% nitrocellulose, 37.5% NG, 1.5% ethyl centralite) were used. These fragments were picked from bulk propellant residues collected during the live firing of 84-mm CarlGustav anti-tank ammunition (Bordeleau et al., 2012a). All instruments that had to be in contact with the experimental soils and solutions were cleaned using soapy water, distilled water, and ACS-grade methanol (MeOH; BDH Merck Ltd., Poole Dorset, UK). For removal of NG by solidphase extraction in aqueous samples, HPLC-grade acetonitrile (EMD Millipore, Billerica, MA) was used to condition the extraction cartridges. For chemical analysis of NG, DNGs and MNGs, OmniSolv LC-MS grade MeOH (EMD Millipore, Billerica, MA) and HPLC-grade acetonitrile were used. Finally, samples for NH+ 4 analyses were acidified using ACS-grade H2SO4 (Sigma-Aldrich, St. Louis, MO), while samples for metal analyses were acidified using trace metal-grade HNO3 (Thermo Fisher Scientific, Waltham, MA).

54

G. Bordeleau et al. / Journal of Contaminant Hydrology 166 (2014) 52–63

2.2. Preparation of soils and solutions The sand and the two organic soils used in the batch- and column-type experiments were collected in a sandpit and at the edge of a forest, respectively; both locations were near an anti-tank range, and neither contained energetic materials. The total organic carbon (TOC) contents are 2.8% and 4.5% for the two organic soils, and b 0.1% for the sand. The bulk sand and soils were dried at room temperature, and sub-samples were then sieved manually (mesh size of 0.063, 0.125, 0.25, 0.5, 1 and 2 mm) in order to compute a grain size curve that allows calculating the d50, which represents the diameter of particles at the median of the grain size distribution (see supplementary Fig. 1 for grain size curves). The finer fractions (b0.063 mm) were subjected to laser grain size analysis, allowing the determination of silt (0.002–0.063 mm) and clay (b0.002 mm) contents. The sand and soils used in abiotic batch and column treatments were sterilized by autoclaving at 121 °C and 15 psi for 3 h; this sterilization procedure was repeated before starting the experiments, as described in their respective section. Some of the experiments were also conducted on DOM, which is a complex mixture of several molecules and is usually defined as the organic matter that passes through a 0.45-μm filter (Weng et al., 2002). For these experiments, an aqueous solution (henceforth called natural dissolved organic matter (NDOM) solution) was prepared by mixing 1.5 kg of the 4.5%-TOC organic soil, with 6 L of distilled water in a plastic container. The container was stirred manually several times over a period of 24 h. The solution was then passed through a 0.45 μm nylon filter cartridge (VWR, Radnor, PA). A small sample was collected for the analysis of dissolved organic carbon (DOC) and metals. Finally, some of the experiments were conducted using a solution of commercially-available humic acids of natural origin and molecular weight range between 2000 and 500 000, consisting of polysaccharides, proteins, simple phenols and chelated metal ions (Sigma-Aldrich, St.Louis, MO). To achieve a DOC concentration comparable to that of the soil

water solution, 575 mg of humic acids were dissolved in 1 L of distilled water. The solution was shaken manually several times over a period of 24 h, after which it was filtered on a 0.45 μm nylon membrane. 2.3. Description of treatments A total of 13 batch and six (6) column experiments were carried out (Table 1). The first five batch treatments (B1, B2, B3, B4-A/B, B5) consisted in slurries containing particles of either sand or organic soil. The first set of treatments (B1, B2, B3), which lasted 20 min, aimed at verifying whether NG is rapidly removed from solution when put in contact with different amounts of SOM. The next treatment (B4-A and B4-B, which are duplicates of the same treatment) aimed at verifying whether NG is removed from solution through adsorption only, or whether it is also degraded, when in contact with sufficient amounts of SOM. This was achieved by measuring the concentrations of common NG degradation − + products (DNG, MNG, NO− 2 , NO3 , NH4 , glycerol) over a period of 15 days. Treatment B5 is a blank control (no NG) which serves to correct for the natural background concentrations of − + ubiquitous compounds (NO− 2 , NO3 , NH4 , glycerol) measured in treatments B4-A/B. The following four batch treatments (B6, B7, B8, B9) did not contain any sediment particles. They aimed at verifying whether NG degradation occurs in presence of DOM, but in absence of POM. Treatment B6 contains the NDOM solution, spiked with NG. Treatments B7 and B8 contain the same amount of NG as treatment B6, but the NDOM solution was diluted by a factor of 2 and 4, respectively. The objective is to determine the effect of DOM concentration on NG degradation. Treatment B9 serves as a control to correct the natural background concentrations of ubiquitous compounds. The purpose of the next two treatments (B10 and B11) was to verify whether the type of DOM influences the degradation of NG. Both treatments contain the solution of commercially-available humic acids. Treatment B10 was spiked with NG, while treatment B11 serves as a blank

Table 1 Description of the different experimental treatments. Treatment

Description

Duration

Analyses

B1 B2 B3 B4-A/B B5 B6 B7 B8 B9 B10 B11 B12 B13 C1 C2 C3 C4 C5 C6

Sand (b0.1% TOC) + NG solution (10.5 mg/L) Organic soil (2.9% TOC) + NG solution (10.5 mg/L) Organic soil (4.5% TOC) + NG solution (10.5 mg/L) Organic soil (4.5% TOC) + NG solution (10.5 mg/L) Organic soil (4.5% TOC) + distilled water (blank control) NDOM solution + NG (10.5 mg/L) NDOM solution (diluted by 2) + NG (10.5 mg/L) NDOM solution (diluted by 4) + NG (10.5 mg/L) NDOM solution + distilled water (blank control) Commercial humic acid solution + NG (10.5 mg/L) Commercial humic acid solution + distilled water (blank control) NDOM solution + NO− 2 (5.2 mg/L) Organic soil (4.5% TOC) + NO− 2 (5.2 mg/L) Organic soil (2.8% TOC), non-sterilized, + AKB 204 propellant fragments Organic soil (2.8% TOC), non-sterilized Organic soil (2.8% TOC), sterilized, + AKB 204 propellant fragments Organic soil (2.8% TOC), sterilized Sand, sterilized, + AKB 204 propellant fragments Sand, sterilized

20 min 20 min 20 min 15 days 15 days 15 days 15 days 15 days 15 days 15 days 15 days 15 days 15 days 16 weeks 16 weeks 16 weeks 16 weeks 16 weeks 16 weeks

− NG, DNG, MNG, NO− 2 , NO3 − NG, DNG, MNG, NO− 2 , NO3 NG, DNG, MNG, NO2−, NO3− − + NG, DNG, MNG, glycerol, NO− 2 , NO3 , NH4 − + NO− 2 , NO3 , NH4 , glycerol, metals − − NG, DNG, MNG, NH+ 4 , glycerol, NO2 , NO3 NG, DNG NG, DNG − + NO− 2 , NO3 , NH4 , metals NG, DNG, MNG, NH4+, glycerol, NO2−, NO3− − + NO− 2 , NO3 , NH4 , metals − NO− 2 , NO3 − NO− 2 , NO3 − NG, DNG, MNG, NO− 2 , NO3 − NO− 2 , NO3 − NG, DNG, MNG, NO− 2 , NO3 NO2−, NO3− − NG, DNG, MNG, NO− 2 , NO3 − NO− 2 , NO3

G. Bordeleau et al. / Journal of Contaminant Hydrology 166 (2014) 52–63

control. Finally, the last two treatments (B12 and B13) served − to determine whether NO− 2 would rapidly oxidize to NO3 in presence of DOM (B12) and/or POM (B13). For the column experiments, the first four treatments (C1–C4) contained the 2.8%-TOC soil. Columns C1 and C3 contained propellant fragments spread onto non-sterilized and sterilized soil, respectively, while columns C2 and C4 contained no propellant and served as blank controls for the non-sterilized and sterilized soil, respectively. The last two columns contained sterilized sand and either propellant (C5), or no propellant for a blank control (C6). 2.4. Batch experimental procedures Batch treatments B1, B2 and B3 were carried out in 50-mL polypropylene tubes (Sarstedt, Nümbrecht, Germany). A preliminary test was conducted to ensure that polypropylene would not adsorb any significant amount of NG over the experimental period. Then, in each tube, 5 g of dry sterilized sand or soil was placed, along with 20 mL of a 10.5 mg NG/L solution. The tubes were placed in a hand-shaker for either 1, 10, or 20 min. They were then centrifuged for 5 min. The supernatant was collected for the analysis of NG, DNGs, − MNGs, and NO− 2 /NO3 . The sand or soil remaining in the tubes was dried in darkness, and used for analysis of NG and DNGs. MNGs were not analyzed, due to analytical interferences in sediment samples. All other batch experiments took place over a total period of 15 days. They were carried out in glass bottles closed with a Teflon-coated lid. The experimental bottles were prepared with their respective contents as described in Table 1, but were not spiked with NG right away. Instead, they were all sterilized by autoclaving at 121 °C and 15 psi for 3 h. They were then kept at 5 °C. The next day, they were spiked with the NG stock solution, for a final NG concentration of 10.5 mg/L. They were then placed in a wooden box kept at 5 °C, which prevented possible degradation due to light or to microorganisms that would have survived the sterilization process. The box was placed on a rotary shaker operated at 100 rpm. Aqueous samples were collected from the bottles after 2 h, then 3.3, 6, 9, 12 and 15 days, for the analysis of NG, − + DNGs, MNGs, NO− 2 /NO3 , NH4 and glycerol. The pH was also measured using an Orion 9206BN probe connected to an Accumet Excel XL50 Dual Channel pH/Ion/Conductivity meter (Thermo Fisher Scientific, Waltham, MA). At the end of the experiments, soils in bottles B4-A and B4-B were dried and analyzed for NG and DNGs. 2.5. Column experimental procedures The experimental columns (117 cm height, 4.2 cm diameter) were made of either Teflon® (columns C1, C3 and C5), or polyethylene teraphtalate (PETG; blank controls C2, C4, C6), and were protected from light to avoid photolysis. The bottom was closed with a Teflon®-lined rubber cap with a small hole in the middle. Six 15-cm long fibreglass wicks were inserted through the hole, to allow water to drip out of the column. A 1-L amber glass jar was placed below each column to collect effluent water. The columns, wicks and jars were thoroughly cleaned with distilled water and MeOH. For column treatments C3 to C6, the sediments were sterilized by

55

autoclaving at 121 °C and 15 psi for 3 hours, and repeating this cycle twice, before being placed in the empty columns. The sediments were then placed in each column in successive 2-cm increments, up to a height of 106 cm. Between increments, sediments were compacted by 12 successive hits with a 0.19-kg weight dropped from a height of 9.3 cm, to create a compaction level representative of field condition (Martel and Gélinas, 1996). The surface was then combed to ensure good hydraulic contact between layers. When filling was completed, a volume of 50 mL of ultrapure water was poured daily for 7 days at the top of the columns, to ensure moist, unsaturated conditions. Then, a 3-cm layer of sterilized sand was added at the top of each column. In columns C1, C3 and C5, 314 mg of AKB 204 propellant fragments (corresponding to 117.7 mg NG) were placed as evenly as possible on top of the upper sand layer. Finally, an additional 1-cm layer of sterilized sand was added in each column. The sand on top of the propellant fragments ensured that the fragments would not float and adhere to the upper part of the column walls upon watering. The 3-cm layer of sand below the fragments ensured that water would percolate at the same rate through the contaminant source zone of all columns, so that the contact time between the fragments and the incoming water would be similar in the well-drained sand columns and the organic-matter rich columns. The experiments were then started, and 50 mL of ultrapure water were poured manually into each column three times a week for a total period of 16 weeks. Aluminum foil was placed on top of each column to limit evaporation. The clear material of the column walls allowed verifying visually that conditions in a column would not become saturated due to clogging. Unsaturated conditions were also verified at the end of the experiments by comparing the water content in the columns with the total porosity (see Supplementary Table 2). The effluent from each column was collected daily and placed at 5 °C; samples were then pooled in two-week time steps, filtered on a 0.45 μm membrane, and frozen until analysis. Aqueous samples were analyzed for NG, − DNGs, MNGs, NO− 2 and NO3 . At the end of the experiments, the sediments in columns C1, C3 and C4 were sampled in 5-cm layers between 0 and 20 cm below the surface, and then in 10-cm layers between 20 and 110 cm. The water content was obtained from the difference between the mass of the moist sediments at the end of the experiments, and the mass measured after drying completely at room temperature for 7 days. For the analysis of NG and DNGs in the source zone (0–5 cm), the whole samples were analyzed, in order to overcome the high heterogeneity due to the presence of propellant fragments. For the other layers, the soil was thoroughly mixed and a 10-g composite sub-sample was collected. In this case, the heterogeneity is expected to be low, because no fragments are present and NG transited in dissolved form. 2.6. Sample preparation and chemical analyses All aqueous samples from treatments containing sediments were filtered on a 0.45 μm nylon membrane. Analysis of NG, DNGs and MNGs was done by high performance liquid chromatography (HPLC), according to a modified version of USEPA method 8330B, described in detail in Martel et al.

56

G. Bordeleau et al. / Journal of Contaminant Hydrology 166 (2014) 52–63

(2010) and Bordeleau et al. (2012b). For sediment samples, the dry sediment was mixed with acetonitrile with a ratio of 1:2 w/w. A vortex was applied for one minute, followed by a sonication period of 2 h in an ultrasonic bath at 18 °C, in darkness. After sonication, the samples were left to settle for 30 min. Two mL of the acetonitrile were collected from the vial, and were filtered on a 0.45 μm membrane. The rest of the procedure is the same as for aqueous samples. Detection limits in water samples were 0.1 mg/L for NG, 0.05 mg/L for DNGs and 0.015 mg/L for MNGs. In sediment samples, the detection limit was 0.3 mg/kg for NG, and 0.4 mg/kg for DNGs. − For analysis of NO− 2 , NO3 and glycerol, samples from treatments containing NG were passed through Sep-Pak® Vac Porapak RDX cartridges prior to analysis, in order to remove NG from solution. This was necessary because the presence of NG would otherwise interfere with the analysis − (Bordeleau et al., 2012b). NO− 3 and NO2 concentrations were analyzed by ion chromatography, as described in Bordeleau et al. (2012b); the detection limit was 0.05 mg/L NO− 3 and 0.01 mg/L NO− 2 . Glycerol was analyzed by HPLC with an IC DX500 system from Dionex (Sunnyvale, CA) equipped with a GP 50 Gradient pump, an AS 50 Autosampler, a LC30 Chromatography oven, and an ED50 electrochemical detector. The analytical column was a Dionex CarboPac MA1 (4 × 250 mm) and a CarboPac MA1 Guard column (4 × 50 mm) eluted with 480 mM sodium hydroxide aqueous solution (MilliQ Waters) at 0.4 mL/min (10 μL injection volume). Analysis of NH+ 4 was done by automated colorimetry using a QuikChem FIA + 8000 series automated ion analyzer (Lachat, Loveland, CO), according to the Lachat QuikChem® Method 31-107-06-2-B (Prokopy, 2001). The detection limit was 0.004 mg/L NH+ 4 . Dissolved organic carbon (DOC) analyses were done using a TOC analyzer. Samples were first acidified with HCl and degassed for 7 min in order to remove dissolved inorganic carbon and volatile organic carbon. They were then heated at 720 °C with a platinum catalyst, resulting in the combustion of non-purgeable organic carbon compounds and their conversion CO2, which was then detected by a Shimadzu (Kyoto, Japan) TOC-VCPH analyzer. The detection limit was 0.05 mg/L. Dissolved metals were determined by inductively coupled plasma atomic emission spectroscopy (ICP-AES; Varian (Palo Alto, CA) model Vista AX CCD) following EPA method 200.7 (USEPA, 2001). The following metals were targeted: Al, As, B, Ba, Be, Ca, Cd, Co, Cr, Cu, Fe, K, Mg, Mn, Mo, Na, Ni, P, Pb, Sb, Sc, Se, Si, Sn, Sr, Ti, Tl, U, V, W, Zn, Zr, as well as sulfur (S). Detection limits varied between 0.00012 à 0.012 mg/L. 3. Results 3.1. Batch experiments: pH, DOC, metals In all batch treatments from the 15-day experiments, the pH remained stable throughout the experiments. Values were lowest in treatments containing the organic (4.5%-TOC) soil (pH = 4.4–4.6), followed by treatments with the NDOM solution (pH = 6.7–7.3), and finally treatments with the humic acid solution (pH = 7.7–8.4). In the starting solutions, DOC concentrations were 134 mg/L (NDOM solution) and

104 mg/L (commercial humic acid solution). Naturally, in the treatments containing the organic soil (B4-A/B), DOC concentrations gradually increased; at the end of the experiments, they reached 1225 mg/L. Metals and sulfur were also measured in the different solutions (see Supplementary Table 1 for detailed results). In the diluted NDOM solutions, concentrations were calculated based on the concentrations measured in the non-diluted NDOM solution. Most of the detected elements were present in all of the solutions. These include Al, B, Ba, Ca, Co, Fe, K, Mg, Mn, Na, S, Si, Sr, Ti, W and Zn. In terms of concentration, the main difference is the concentration of K, which is one to two orders of magnitude lower in the humic acid solution than in any of the natural soil water solutions. There are also some metals that were detected in small amounts in the NDOM, but not in the commercial humic acid solution. These include Cr, Cu, Mo, Ni, P, Pb, Sc, V, and Zr. As with DOC, concentrations of metals in the treatments containing the organic soil particles increased during the experiment. 3.2. Batch experiments: nitroglycerin and its degradation products Dissolved NG concentrations were monitored in the 20-min experiments (B1, B2 and B3). With the sand (treatment B1), no significant change in NG concentration was detected in solution (Fig. 1A). In contrast, NG concentrations decreased in presence of the 2.8%-TOC soil (treatment B2; 17% decrease within 20 min), and even more so in presence of the 4.5%-TOC soil (treatment B3; 29% decrease within 20 min). None of the usual degradation products were detected in solution. At the end of the experiments, NG and its degradation products were undetected in the dried sand or soils. The rate of NG disappearance from solution (k1) was calculated according to 1st-order kinetics, with correlation coefficients of 0.96 and 0.90 in treatments B2 and B3, respectively (Table 2). To make the rate constant independent of the amount of TOC, a 2nd-order rate constant (k2) was also calculated. Uncertainties were calculated at the 95% confidence level using student's t-distribution. Dissolved NG concentrations were also monitored in the 15-day treatments containing both DOM and POM from the 4.5%-TOC soil (treatments B4-A/B). NG concentrations decreased rapidly, and were below detection limit from t = 3.3 days onward (Fig. 1B). Complete disappearance of dissolved NG in treatments B4-A/B occurred too fast to allow a reliable computation of the rate constant, due to the limited number of sampling points. However, contrary to treatments B1-B3, in treatment B4-A/B degradation products gradually appeared in solution. The main N-bearing compounds are 1,2-DNG and NO–3. NO− 2 remained undetected throughout the experiments, while 1,3-DNG was detected during the first half of the experiments, but at concentrations below the quantification limit. Throughout the experiments, concentrations of NH+ 4 were similar to those in the blank control (treatment B5; results not shown), suggesting that NH+ 4 is not a product of NG degradation under these conditions. Glycerol concentrations in treatments B4-A/B were generally higher than in control treatment B5. However, the presence of important concentrations of DOM in the aqueous samples caused significant analytical interference,

G. Bordeleau et al. / Journal of Contaminant Hydrology 166 (2014) 52–63

57

the reaction rate (k1), correlation coefficient (R2) and half-life (t1/2) were calculated according to 1st-order kinetics (Table 2). Then, individual second-order rate constant (k2) were calculated, as well as an average of all individual k2. Degradation products were monitored in the treatment containing the non-diluted NDOM solution (B6). The main detected N-bearing compounds are 1,3-DNG and NO− 2 . Concentrations of 1,3-DNG increased throughout the experiment, while 1,2-DNG started to be detected in the second half of the experiment, at concentrations below the quanti+ fication limit until the last sampling point. NO− 3 and NH4 concentrations remained similar to the natural background concentrations observed in control treatment B9. Finally, in this treatment, the amount of DOM was not high enough to cause interference in glycerol analyses, which allowed its quantification; however, glycerol remained undetected throughout the experiment. Finally, in treatments B4-A/B and B6, the stoichiometry was verified by computing a nitrogen (N) mass balance. Values are expressed in per cent (%) of the total amount of N initially present in NG (Fig. 2). The balance is calculated based on concentrations in solution only; indeed, NG and DNGs were measured in soils of treatments B4-A/B at the end of the experiments, but neither of these compounds was detected. In these duplicate treatments, a sharp decrease in the N mass balance is observed at the beginning of the experiment, after which the total amount of N increases slightly. In treatment B6, the total mass of N decreased slightly during the experiment, but much less than in treatments B4-A/B (Fig. 2). The last two treatments (B12 and B13) aimed at assessing the stability of NO− in presence of DOM and POM, 2 respectively. In the NDOM solution, the mass of NO− 2 decreased by 5.6% over 15 days, and NO− 3 did not appear in solution. In presence of POM, NO− 2 concentrations drastically decreased before the first sampling point (the concentration at t = 0 is based on the NO− 2 concentration in the stock solution, before it was put in contact with the soil). This rapid decrease could be due to a lower analytical sensitivity towards NO− 2 due to the organic matter in the aqueous samples. Nonetheless, following this, NO− 2 concentrations kept decreasing, and were accompanied by a corresponding increase in NO− 3 concentrations (Fig. 2), suggesting that the − NO− 2 was being oxidized to NO3 .

Fig. 1. Nitroglycerin concentrations in short-term (A) and long-term (B) batch experiments B1-B6, and B7-B8. Error bars are contained within the size of the data points.

such that the high uncertainty on glycerol measurements only allowed semi-quantitative observations. Dissolved NG concentrations were also measured in the treatments containing DOM, but no POM. In the commercial humic acid solution (treatment B10), NG concentrations did not decrease over the experimental period (results not shown). In contrast, in treatments containing various concentrations of the NDOM solution (B6–B8), NG degradation was observed. At the end of the experiments, 20% of the initial NG had been degraded in the non-diluted NDOM solution, against 15% in the solution diluted by 2, and 8% in the solution diluted by 4 (Fig. 1B). For these three treatments,

3.3. Column experiments: transport and degradation of NG − Concentrations of NG, DNGs, MNGs, NO− 2 and NO3 were measured in the effluent of columns C1, C3 and C5 over the

Table 2 Kinetic parameters (rate constants (k1, k2), correlation coefficient (R2), half-life (t1/2)) for NG disappearance in batch experiments. Treatment

B2 B3 B6 B7 B8 a b

TOC or DOC content

2.8% TOC 4.5% TOC 134 mg/L DOC 67 mg/L DOC 33.5 mg/L DOC

1st-order kinetic parameters

2nd-order kinetic parameters

k1 (sec−1)a

R2

t1/2 (days)

Individual k2b

Global k2

5.8E-05 (±3.6E-05) 9.0E-05 (±9.4E-05) 1.7E-07 (±1.0E-07) 1.2E-07 (±5.1E-08) 6.6E-08 (±3.5E-08)

0.96 0.90 0.78 0.87 0.81

0.14 (±0.09) 0.09 (±0.09) 46 (±27) 67 (±29) 122 (±65)

2.1E-03 (±1.4E-03) 2.0E-03 (±2.1E-03) 1.3E-03 (±7.9E-04) 1.8E-03 (±8.2E-04) 2.0E-03 (±1.1E-03)

2.0E-03 (±1.7E-03)

Units of k2 for solutions are: L · sec−1 · kg−1 (where kg−1 refers to the mass of DOC). Units of k2 for soils are: kg · sec−1 · kg−1 (where kg refers to soil mass, and kg−1 refers to TOC mass).

1.7E-03 (±9.1E-04)

58

G. Bordeleau et al. / Journal of Contaminant Hydrology 166 (2014) 52–63

Fig. 2. Nitrogen (N) mass balance in batch treatments B4-A/B, B6, B12 and B13. The total N mass (thick black line) is the sum of all N-bearing compounds detected in a given treatment, such as NG (diamonds), DNGs (triangles), NO-3 (circles), and NO-2 (squares).

16-week experimental period. NG, 1,2-DNG and 1,3-DNG were only detected in the effluent of column C5 (sterilized sand + propellant). Concentrations increased until 260 mm of infiltration, where they reached 53 mg/L NG, 4.8 mg/L 1,3-DNG, and 5.3 mg/L 1,2-DNG (Fig. 3). MNGs and NO− 2 were not detected in the effluent of any column, while NO− 3 was detected in all columns. In the blank controls (C2, C4, C6), concentrations remained below 1 mg/L N-NO− 3 . Concentrations in samples from the control columns were subtracted from the concentrations measured in samples from the corresponding experimental columns, in order to compute only the NO− 3 that is related to the degradation of NG. The resulting maximum concentrations in the columns containing propellant (C1, C3 and C5) were 0.62, 13.1, and 2.43 mg/L, respectively (Fig. 3). The peaks in NO− 3 concentrations in all columns coincide with the peak in NG and DNGs in column C5. The experiments were stopped after approximately 700 mm of infiltration, although NG leaching was not completely over (NG concentration was 2.7 mg/L in the last sample of C5). At the end of the experiments, NG concentrations were measured in all sediment layers from C1, C3 and C5. NG was detected in the source zone (0–5 cm layer) of each of these columns, as well as in the 5–10 cm layer of columns C1 and C3. No NG was detected in any other layer. A nitrogen mass balance calculation was done by computing the total amount of N recovered in the form of NG, DNGs, and NO− 3 measured in water and sediment samples (Fig. 4). The N contained in the nitrocellulose matrix of the propellant was not considered, as nitrocellulose is stable and non watersoluble, and is not expected to have degraded over the course of the experiment (Bordeleau et al., 2014). In all columns, a large fraction (28–62%) of the initial N was not recovered, similarly to the batch-type experiments. The highest N

Fig. 3. Concentrations of NG, DNG and N-NO− 3 in water samples from column experiments.

recovery in water and sediment samples was observed in column C5, followed by C3, and finally C1. 4. Discussion 4.1. Effects of different SOM components in batch experiments The results from batch experiments demonstrate that NG concentrations decrease when put in contact with POM and/or NDOM. The disappearance of NG from solution, and the fact that it was not detected in any of the batch soils at the end of the experiments, could be explained by three factors, namely: 1) irreversible sorption onto soil particles, 2) adsorption or complexation with POM which causes or mediates degradation, and 3) complexation with DOM which causes or mediates degradation. These hypotheses are discussed in the next sub-sections. 4.1.1. Sorption versus degradation related to POM The results from 20-min batch experiments (B1 to B3) clearly demonstrate that NG is rapidly adsorbed onto POM. Indeed, NG concentrations decreased in solutions which were left in contact with the two organic soils, but not in solutions left in contact with sand. Furthermore, the absence of degradation products in these treatments demonstrates

G. Bordeleau et al. / Journal of Contaminant Hydrology 166 (2014) 52–63

59

Fig. 4. Nitrogen (N) mass balance calculations for column experiments. Quantities are expressed in terms of percentage (%) of the initial N from NG in propellant fragments recovered within the different N-compounds in water and sediment samples.

that NG was adsorbed, but not degraded, over the short experimental period. These results are consistent with the very few studies having examined sorption of NG onto soils (Clausen et al., 2010; Speitel et al., 2002). The conclusions from these studies are that sorption of NG occurs, that it may be related to the organic carton content and/or cation exchange capacity of soils, and that some retardation and/or irreversible sorption is possible. However, contrary to sorption of nitroaromatics which has been comparatively well documented, to our knowledge no mechanism has been proposed in the literature for the sorption of NG onto organic soil particles. While identifying specific mechanisms is beyond the scope of this paper, NG sorption appears to be an important process that can participate in NG natural attenuation, and represents an important avenue for future research. In contrast to 20-min batch experiments, results from the 15-day batch treatments containing SOM in the form of POM and NDOM (B4-A/B) demonstrate that not only is NG sorbed onto soil particles, but at least part of it is eventually degraded. Indeed, the total N mass balance in these treatments rapidly decreased by nearly 80%, without the appearance of degradation products, which can be explained by rapid sorption of NG onto POM. Following this, 1,2-DNG and NO− 3 began to appear in solution. Concentrations of 1,2-DNG increased and then decreased, possibly being further degraded to MNGs and glycerol. Unfortunately, analytical interferences due to the high DOM content in these treatments prevented the quantification of MNGs and glycerol. However, the observed increase in NO− 3 concentrations, which lasted until the end of the experiment, is consistent with DNG degradation. It is therefore clear that the presence of SOM causes the adsorption of NG, and either causes its degradation directly, or facilitates its degradation by other factors. 4.1.2. NG degradation: role of POM versus DOM To determine whether degradation is caused by POM and/ or by DOM, the results from batch treatments B4-A/B and B6 are compared. First of all, with the NDOM solution (treatment B6), NG decay was accompanied by increasing

concentrations of DNGs and NO− 2 , which confirms that some degradation can be caused by DOM. While 1,3-DNG was detected throughout the whole experiment, 1,2-DNG started to be detected only after 6 days, and concentrations remained below the quantification limit. The same pattern was observed in treatments B7 and B8, which contained lower concentrations of NDOM. This does not correspond to random removal of a nitro group, which would yield 1,2-DNG 67% of the time, and 1,3-DNG 33% of the time. Instead, it indicates a regioselectivity towards 1,3-DNG, which can be explained by the lower stability of the secondary (central) NO2 group compared to the primary (terminal) NO2 groups. Indeed, the vibration orbital of the secondary group is hindered by the orbitals of the primary groups, so that its departure is energetically favorable. If further degradation occurred, 1-MNG and then glycerol could be produced. The absence of these compounds, together with increasing concentrations of DNGs, suggests that degradation did not proceed further in these conditions. In presence of POM (treatments B4-A/B), degradation also occurred, as shown by the appearance of degradation products. In this case, at least part of the degradation must be due to the DOM in solution (as seen from results from treatments B6-B8), but there may also be a contribution of POM. To distinguish these two components, it is not possible to compare the rate of NG disappearance in treatments B4-A/B and treatment B6, because of the rapid adsorption onto soil particles in treatments B4-A/B. It is not possible either to compare the rate of appearance of degradation products, because DOM concentrations in treatments B4-A/B gradually increased during the experiments, due to constant contact with the soil. Instead, the distinct role of POM in NG degradation is outlined by differences in the distribution of degradation products. First of all, with the NDOM solutions (treatments B6-B8), 1,3-DNG was produced preferentially over 1,2-DNG, while the opposite was observed in presence of POM (treatments B4-A/B). Second, DNG concentrations eventually decreased in presence of POM, indicating that the reaction can proceed further than with DOM only. The higher concentrations of 1,2-DNG compared to 1,3-DNG in treatments B4-A/B

60

G. Bordeleau et al. / Journal of Contaminant Hydrology 166 (2014) 52–63

should not be the result of produced 1,3-DNG decaying faster than 1,2-DNG, as the latter is energetically less stable than the former. This was confirmed by Oh et al. (2004), who found that 1,2-DNG is degraded faster than 1,3-DNG in presence of elemental iron. Instead, the dominance of 1,2-DNG could be explained by steric hindrance, which may limit the binding of POM molecules, which are comparatively larger than DOM molecules, to the central NO2 group of NG. Hence, it may be easier for the actives sites of POM to bind to the terminal NO2 groups of NG, resulting in the preferential production of 1,2-DNG. − A third difference resides in the different NO− 2 /NO3 concentrations produced by NG degradation. With the NDOM solution (treatment B6), for the equivalent of one mole of NG that was degraded, 0.68 mole of NO− 2 was produced. This is consistent with the hypothesis that degradation did not proceed past the DNG stage, in which − case the number of moles of NO− 2 /NO3 that are produced per mole of NG degraded should not exceed 1. In contrast, in presence of POM (treatments B4-A/B), the NO2 groups released from NG were recovered in the form of NO− 3 , most − likely as a result of conversion of NO− 2 to NO3 in the solution. This hypothesis is supported by results from treatment B13 − (POM + NO− 2 solution), where NO2 was rapidly oxidized to NO− , and where an important part of the initial N was not 3 recovered. The same phenomenon of incomplete recovery is expected to occur in treatments B4-A/B, so the amount of NO− recovered in these bottles is most certainly an 3 underestimation of the true amount that was released through NG degradation. Even with this factor in mind, the number of moles of NO− 3 recovered per mole of NG degraded reaches between 0.80 (treatment B4-B) and 1.22 (treatment B4-A). This confirms that in presence of POM, NG degradation can proceed past the DNG stage. Unfortunately, analytical interferences due to the presence of DOM prevented the detection of MNGs and the precise quantification of glycerol. Nonetheless, glycerol was detected (semi-quantitatively) in treatments B4-A/B, and was not detected in treatment B6, which confirms once again that degradation past the DNG stage only occurred in presence of POM. 4.1.3. Effect of organic carbon concentration on sorption and degradation The effect of organic carbon concentration was studied for the processes of NG adsorption onto POM, and degradation by DOM. With the POM left in contact with dissolved NG for up to 20 min (treatments B2–B3), disappearance of NG from solution is attributed to sorption rather than degradation. The similarity of the 2nd-order rate constants calculated for NG sorption on the two organic soils (Table 2) suggests that sorption is directly proportional to the organic carbon content of soil, within the range of TOC concentrations tested in this study. For these experiments, the high uncertainty associated with k2 values does not reflect large inconsistencies between samples or treatments, but is rather a consequence of having only three sampling points per treatment. The same exercise was repeated for NG degradation by NDOM. The 2nd-order rate constants (k2) computed for reactions in batch treatments B6, B7 and B8 vary by as much as 25% from the mean value, which is a relatively large

difference (Table 2). However, there was a relatively large uncertainty associated with the values of k1 themselves, which suggests that the amount or type of available reactive sites of DOM may have changed during the experiments. This uncertainty is then necessarily reflected in the values of k2. Nonetheless, the reaction appears to be 2nd-order (1st-order with respect to NG, and 1st-order with respect to DOC). This means that like NG sorption onto POM, the rate of NG degradation by DOM is proportional to the amount of organic carbon in solution. 4.1.4. Effect of DOM type on NG degradation While the results discussed above have shown that the concentrations of POM or DOM directly influence the sorption and/or degradation of NG, the type of organic matter also appears to be important. Indeed, in the treatment containing DOM in the form of commercial humic acids, no degradation of NG was observed, despite the fact that the DOC concentration was within the range of concentrations found in the treatments containing the NDOM solution (treatments B6–B8). This difference cannot be explained by a difference in pH conditions, as the pH in both types of solutions was comparable. One factor that could theoretically explain this difference is the formation of active complexes between DOM and metal ions in solution, which in turn can play a role in the degradation of contaminants (Kim and Strathmann, 2007). However, among the 33 cations analyzed, only potassium (K+) had a significantly lower concentration in the humic acid solution than in the most dilute NDOM solution. While K+ is known to adsorb onto mineral and organic surfaces (Wang and Huang, 2001), it is an alkali metal and does not act as a reducer. Alternatively, the difference in NG reactivity towards different types of DOM could be explained by the type of functional groups contained in DOM. As an example, a study of the degradation of NG by hydrogen sulfide mediated by the presence of black carbons outlined significant differences in the rate of reaction when using different types of black carbons, namely carbon nanotubes, sheet graphite, activated carbon, and diesel soot (Xu et al., 2010). The authors attributed the variation in reactivity to differences in conductivity, functional groups, and possibly unidentified reactive sites of these materials. Therefore, at the moment, a difference in the types of organic matter functional groups is the most probable factor explaining the lack of NG degradation in the commercial humic acid solution. 4.2. Fate of NG in experimental columns Results from the batch experiments showed that under ideal conditions, NG may be degraded within a few days when left in contact with SOM. However, field conditions on training ranges are usually less ideal, in that the contact between infiltration water and contaminated soil may be shorter and less homogenous, and the organic carbon content of soil may be lower. The column experiments served to confirm whether similar sorption and degradation would be observed under conditions that are more likely to be encountered at contaminated field sites. The NG concentrations in surface sediments of columns C1, C3 and C5 at the end of the experiment indicate that

G. Bordeleau et al. / Journal of Contaminant Hydrology 166 (2014) 52–63

between 59 and 68% of the initial NG leached out of the propellant particles after 700 mm of infiltration. This amount of NG therefore transited the columns, and was either adsorbed, degraded, or recovered intact in the effluent. In the organic soil columns C1 and C3, at the end of the experiment NG was present in the soil layer underlying the source zone (5–10 cm) at very low concentrations (0.4– 0.5 mg/kg), and was not detected in lower layers. The high proportion of missing N (Fig. 4) in the mass balance of these columns, combined with the absence of NG in soil layers below the source zone, suggests that NG may have been irreversibly sorbed onto soil particles, either as intact NG molecules, or as partially-degraded NG by-products. In column C3 (sterilized organic soil), degradation clearly occurred, as shown by the high NO− 3 concentrations in the effluent, which account for 21.1% of the final N balance. While NO− 3 is the major degradation product recovered from this column, an important portion of the N balance (46.5%) remains missing. A possible explanation is that after the removal of one or more nitro group(s) from the NG molecule, the generated alcohol group(s) may react with a carboxylic acid group of the organic matter and as a result, the newly formed ester may explain the disappearance of the DNG, MNG or glycerol. Another potential explanation could be a possible consumption of NO− 3 and other N-bearing compounds by microorganisms that were not killed during the sterilization process. Consumption of NO− 3 by microorganisms present in this soil is clearly shown in column C1 (non-sterilized organic soil), where the effluent NO− 3 concentrations amount to only 1.2% of the total N balance. The lower NO− 3 concentrations in the effluent of C1 can fully account for the difference in total N mass recovered between C1 and C3, indicating that bacterial NO− 3 consumption is responsible for the poorer mass balance in the non-sterilized column. Consumption of NO− 3 by bacteria able to degrade NG has often been reported (Accashian et al., 2000; Ducrocq et al., 1989; White et al., 1996). This was notably the case with bacteria isolated from soil of a former anti-tank firing point located within 5 km of the present study site (Bordeleau et al., 2013). In contrast, in the column containing sterilized sand (C5), signs of NG degradation were much more limited. In the effluent of this column, NO− 3 concentrations account for 3.9% of the N mass balance, while DNGs account for 3.4%. However, most of the NG was recovered intact in the effluent. The limited degradation in the sand could be directly related to the absence of organic carbon, but also to the shorter water residence time in this well-drained column. Indeed, water percolated more rapidly in the sand than in the organic soil, and the measured water content was four times lower in the sand column than in the organic soil columns. This difference is not due to the grain size distributions, as the d50 and uniformity coefficient (d60/ d10) are similar for both soils, but is rather due to the organic matter content. The abundant organic matter in columns C1–C4 made it difficult to wet the soil at the beginning, but once wetting was achieved, water retention remained higher in these columns over the course of the experiments. The absence of organic carbon in C5 could therefore explain the lower NG degradation, both directly and indirectly.

61

4.3. Degradation mechanism The results from batch and column experiments may be used to infer a degradation pathway for SOM-mediated NG degradation. While the pathway has not specifically been investigated in our experiments, the range of observed degradation products may be compared to the degradation products reported for the two degradation pathways documented in the literature, namely the reductive and the hydrolytic pathways. Naturally, reactions with organic matter are complex, and other types of reactions, such as esterification of the hydroxyl groups of NG degradation products with carboxylic groups of the organic matter, may be possible. Because there is still little understanding of the role of organic matter and its functional groups in contaminant abiotic degradation, the following description will be limited to the two main NG degradation pathways. The reductive pathway is the most frequent of the two, and has been documented namely for abiotic transformation of NG using reducing agents (Oh et al., 2004; Xu et al., 2010), as well as biodegradation under aerobic (Accashian et al., 1998; Bhaumik et al., 1997; Marshall and White, 2001; Meng et al., 1995) or anaerobic (Christodoulatos et al., 1995, 1997) conditions. The pathway proceeds through denitration of the NG molecule, yielding sequentially DNGs, MNGs, and finally glycerol. At each step, one NO− 2 ion is released, and may − further be reduced to NH+ 4 or oxidized to NO3 . In contrast, the hydrolytic pathway results from a nucleophilic substitution reaction, such as with alkaline hydrolysis, where hydroxide ions (OH−) act as the nucleophile. Alkaline hydrolysis has been documented for NG (Halasz et al., 2010); the reaction causes the release of the two primary − functional groups as NO− 2 , and the secondary group as NO3 . Instead of the usual degradation products (DNGs, MNGs, glycerol), a series of more complex molecules are produced, including notably formic acid, glyoxylic acid, and oxalic acid (Halasz et al., 2010). In our experiments, the occurrence of the reductive pathway appears more likely than the hydrolytic pathway. Indeed, the neutral to acidic pH values in the NDOM solutions, the presence of DNGs in the treatments containing POM and NDOM, the presence of glycerol in the treatments containing POM, and the release of the central functional − group as NO− 2 (instead of NO3 ) in the treatment containing NDOM, are all factors pointing towards the reductive pathway. The fact that the pathway proceeded to completion only when POM was present suggests that the reduction of DNGs, which are energetically more stable than NG, requires the polarization of the DNG molecule caused by its adsorption onto POM. Once polarized, further DNG reduction by a reducing agent, such as Fe or S, could occur. This mechanism would be analogous to the one described by Xu et al. (2010) for black carbon-mediated degradation of NG by sulfides. Although our experiments were done under aerobic conditions which are comparatively less favorable to reduction, an interesting parallel can be made with the aforementioned study (Xu et al., 2010), which was conducted under anaerobic, reducing conditions typical of marine sediments. Their experiments were conducted on solutions containing sulfides, in presence or absence of solid sheet graphite as a model black carbon. Although redox conditions

62

G. Bordeleau et al. / Journal of Contaminant Hydrology 166 (2014) 52–63

were different, the results of Xu et al. (2010) are strikingly similar to our results. First of all, the authors demonstrated that reactions both in the solid phase (sheet graphite and sulfides) and in the aqueous phase (sulfides alone) contribute to NG decay in a batch reactor system. They also showed that the 1st-order NG decay rates were linearly proportional to sulfide concentration. Furthermore, in presence of both sheet graphite and sulfides: 1) DNGs were formed and then decayed, 2) glycerol was detected as an end-product, and 3) the amount of NO− 2 produced was much more important than in treatments containing sulfides but no graphite. In their experiments with sulfides alone: 1) NG was degraded less rapidly, 2) DNGs were formed but did not decay further, and 3) concentrations of 1,3-DNG were higher than 1,2-DNG. While these are all common points between their results and ours, there is one notable difference, which is that in presence of sulfides and sheet graphite, strong regioselectivity favoured the presence of 1,3-DNG over 1,2-DNG, in a ratio of 3:1. This is in opposition with our results with POM, where 1,2-DNG was the favoured isomer. However, increased steric hindrance related to active sites on the POM versus sheet graphite could explain this difference. 4.4. Environmental relevance According to the results from batch and column experiments, sorption as well as degradation mediated by SOM could be important processes participating in the natural attenuation of NG on military training ranges contaminated with propellant residues. However, documented field studies, or even laboratory studies using range soils, are still very scarce. In a few published studies where TOC content of soil is reported, it appears that SOM could have played a role in both NG sorption and degradation, even though the authors usually did not specifically investigate this factor. In one of the reported degradation studies, training range soils containing between 1.1 and 2.0% TOC were spiked with a NG solution (Jenkins et al., 2003). The authors reported the complete disappearance of NG in soils within less than a day, except in the soil with the lowest TOC content. The authors did not monitor NG degradation products, and did not seek abiotic conditions, but it appears that NG disappearance could be related to both biodegradation and degradation related to SOM. In another study, Bellavance-Godin (2009) − observed complete NG degradation and significant NO− 2 /NO3 production in a large column that contained 58 cm of sand (TOC b 0.1%) overlaid with a 2-cm layer of NG-contaminated organic soil (6.8% TOC) from a training range. In contrast, the author observed incomplete NG degradation and lower NO− 2 / NO− 3 concentrations in a similar column where the top soil layer contained 1.2% TOC. Once again, conditions in the columns were not sterile, but the differences in the results could be related to the organic carbon content of the top soil layer. Finally, at an anti-tank firing position that has been extensively studied by our research group (Bordeleau et al., 2012a), the presence of NG at the soil surface has been documented, as well as the occurrence natural attenuation accompanied by the production of DNGs, MNGs, and high NO− 3 concentrations, in water from the unsaturated zone at all sampled depths between 0.1 and 5.0 m below ground

surface. Photolysis and biodegradation could not fully account for the observed NO− 3 production, so this site is a good example of the possible contribution of SOM to NG degradation. The site has been described in detail in Bordeleau et al. (2012a), and various soil properties collected specifically for the present paper (TOC, clay content, water content, d50, bacterial density) are reported in Supplementary Table 2 and Supplementary Figs. 1 and 2. Briefly, the top 30 cm of the soil profile contain between 3.0 and 5.8% TOC, while deeper sediments contain less than 0.5% TOC. Naturally, the upper soil layer also contains significant quantities of − microorganisms, which could consume the NO− 2 /NO3 released through NG biotic or abiotic degradation, as shown by the results from experimental column C1. However, the presence of high concentrations of NO− 3 in pore water down to 5.0 m could be explained by either: 1) a local microorganism population that does not consume the released NO− 2 / NO− 3 , or 2) the relatively rapid migration of infiltration water from the uppermost layer, where NG degradation occurs, into deeper units where fewer bacteria are present, thereby preventing complete consumption of the produced NO− 3 . In any case, it appears very likely that SOM-mediated degradation participates in NG natural attenuation at this site. NG sorption onto organic soil particles may also potentially participate in natural attenuation, but this process was not studied specifically. 5. Conclusion The role of SOM in NG sorption and degradation was evaluated through batch- and column-type experiments. Results from the batch experiments clearly demonstrate that while POM causes sorption of NG, both POM and NDOM can also mediate the degradation of NG and the concomitant − release of NO− 2 /NO3 . It appears that the type of functional groups could be a factor determining whether the reaction can occur, as shown by the absence of degradation in presence of a commercial humic acid solution serving as the source of DOM. Column experiments were then conducted in order to better represent usual field conditions, as most of the contamination on training ranges is found in the welloxygenated, unsaturated soils at or near the surface. Under these conditions, sorption of NG onto POM, as well as SOM-related NG degradation, have been confirmed. Noteworthy, even if the TOC content of training range soils decreases rapidly with depth, sorption of NG onto POM will cause its retention within the upper soil layers, where the higher TOC content also favours degradation. Therefore, the results from this study indicate that SOM could significantly participate in the natural attenuation of NG on contaminated training ranges. Eventually, a better understanding of the specific factors which allow this reaction to occur could allow using organic matter as an amendment in the surface soil or as a reactive barrier on training ranges, especially at firing positions. This could represent a cost-effective, ecological way of ensuring that NG does not migrate to the water table. Acknowledgements The authors wish to thank the Director Land Environment of the Canadian Ministry of Defence for funding the project,

G. Bordeleau et al. / Journal of Contaminant Hydrology 166 (2014) 52–63

as well as Richard Lévesque, Dr. Peter G. Campell, and Dr. Claude Fortin from INRS, for their precious help and useful advice. Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.jconhyd.2014.06.012. References Accashian, J.V., Vinopal, R.T., Kim, B.J., Smets, B.F., 1998. Aerobic growth on nitroglycerin as the sole carbon, nitrogen, and energy source by a mixed bacterial culture. Appl. Environ. Microbiol. 64 (9), 3300–3304. Accashian, J.V., Smets, B.F., Kim, B.J., 2000. Aerobic biodegradation of nitroglycerin in a sequencing batch reactor. Water Environ. Res. 72 (4), 499–506. Bedford, C.D., Carpenter, P.S., Nadler, M.P., 1996. Solid-state photodecomposition of energetic nitramines (RDX and HMX). Report NAWCWPNS TP 8271. Naval Air Warfare Center Weapons Division, China Lake, CA. Bellavance-Godin, A., 2009. Devenir environnemental des résidus de propulsif aux positions de tir anti-char à travers la zone non saturée(M.Sc. Thesis) INRS-ETE, Quebec City (131 pp.). Bernstein, A., et al., 2010. Quantifying RDX biodegradation in groundwater using δ15N isotope analysis. J. Contam. Hydrol. 111 (1–4), 25–35. Bhaumik, S., Christodoulatos, C., Korfiatis, G.P., Brodman, B.W., 1997. Aerobic and anaerobic biodegradation of nitroglycerin in batch and packed bed bioreactors. Water Sci. Technol. 36 (2–3), 139–146. Bordeleau, G., Martel, R., Ampleman, G., Thiboutot, S., Poulin, I., 2012a. The fate and transport of nitroglycerin in the unsaturated zone at active and legacy anti-tank firing positions. J. Contam. Hydrol. 142–143, 11–21. Bordeleau, G., et al., 2012b. Overestimation of nitrate concentrations in water samples due to the presence of nitroglycerin or hexahydro-1,3,5trinitro-1,3,5-triazine. J. Chromatogr. A 1252, 130–135. Bordeleau, G., Martel, R., Ampleman, G., Thiboutot, S., 2013. Photolysis of RDX and nitroglycerin in the context of military training ranges. Chemosphere 93 (1), 14–19. Bordeleau, G., Martel, R., Drouin, M., Ampleman, G., Thiboutot, S., 2014. Biodegradation of nitroglycerin from propellant residues on military training ranges. J. Environ. Qual. 43 (2), 441–449. Burton, D.T., Turley, S.D., Peters, G.T., 1993. Toxicity of nitroguanidine, nitroglycerin, hexahydro-1,3,5-trinitro-1,3,5-triazine (RDX), and 2,4,6trinitrotoluene (TNT) to selected freshwater aquatic organisms. Technical report #AD A267467. US Army Medical Research and Development Command, Fort Detrick, Frederick, MD. Christodoulatos, C., Pal, N., Bhaumik, S., 1995. Anaerobic biodegradation of nitroglycerin by digester sludge. In: Sengupta, A.K. (Ed.), Twenty-Seventh Mid-Atlantic Industrial Waste Conference. Technomic Publishing Co. Christodoulatos, C., Bhaumik, S., Brodman, B.W., 1997. Anaerobic biodegradation of nitroglycerin. Water Res. 31 (6), 1462–1470. Clausen, J., et al., 2010. Adsorption/desorption measurements of nitroglycerin and dinitrotoluene in Camp Edwards, Massachussets soils. ERDCCRREL TR-10-1US Army Corps of Engineers, Cold Regions Research and Engineering Laboratory, Hanover, NH. Clausen, J., Scott, C., Osgerby, I., 2011. Fate of nitroglycerin and dinitrotoluene in soils at small arms training ranges. Soil Sediment Contam. 20, 649–671. Ducrocq, C., Servy, C., Lenfant, M., 1989. Bioconversion of glyceryl trinitrate into mononitrates by Geotrichum candidum. FEMS Microbiol. Lett. 65 (1–2), 219–222. Halasz, A., Thiboutot, S., Ampleman, G., Hawari, J., 2010. Microwave-assisted hydrolysis of nitroglycerin (NG) under mild alkaline conditions: New insight into the degradation pathway. Chemosphere 79 (2), 228–232.

63

Husserl, J., Spain, J.C., Hughes, J.B., 2010. Growth of Arthrobacter sp. strain JBH1 on nitroglycerin as the sole source of carbon and nitrogen. Appl. Environ. Microbiol. 76 (5), 1689–1691. Jenkins, T.F., Bartolini, C., Ranney, T.A., 2003. Stability of CL-20, TNAZ, HMX, RDX, NG, and PETN in moist, unsaturated soil. ERDC/CRREL TR-03-7US Army Corps of Engineers, Cold Regions Research and Engineering Laboratory, Hanover, NH. Kim, D., Strathmann, T.J., 2007. Role of organically complexed iron(II) species in the reductive transformation of RDX in anoxic environments. Environ. Sci. Technol. 41 (4), 1257–1264. Marshall, S.J., White, G.F., 2001. Complete denitration of nitroglycerin by bacteria isolated from a washwater soakaway. Appl. Environ. Microbiol. 67 (6), 2622–2626. Martel, R., Gélinas, P.J., 1996. Surfactant solutions developed for NAPL recovery in contaminated aquifers. Ground Water 34 (1), 143–154. Martel, R., Bellavance-Godin, A., Lévesque, R., Côté, S., 2010. Determination of nitroglycerin and its degradation products by solid-phase extraction and LC-UV. Chromatographia 71 (3–4), 285–289. Meng, M., et al., 1995. Denitration of glycerol trinitrate by resting cells and cell extracts of Bacillus thuringiensis/cereus and Enterobacter agglomerans. Appl. Environ. Microbiol. 61 (7), 2548–2553. Oh, S.Y., Cha, D.K., Kim, B.J., Chiu, P.C., 2004. Reduction of nitroglycerin with elemental iron: Pathway, kinetics, and mechanisms. Environ. Sci. Technol. 38 (13), 3723–3730. Pennington, J.C., Brannon, J.M., 2002. Environmental fate of explosives. Thermochim. Acta 384, 163–172. Pennington, J.C., et al., 2001. Monitored natural attenuation of explosives. Soil Sediment Contam. 10 (1), 45–70. Prokopy, W.R., 2001. Determination of ammonia by flow injection analysis. Lachat Instruments, Milwaykee, WI. Saad, R., Thiboutot, S., Ampleman, G., Dashan, W., Hawari, J., 2010. Degradation of trinitroglycerin (TNG) using zero-valent iron nanoparticles/nanosilica SBA-15 composite (ZVINs/SBA-15). Chemosphere 81, 853–858. Singh, N., Hennecke, D., Hoerner, J., Koerdel, W., Schaeffer, A., 2008. Mobility and degradation of trinitrotoluene/metabolites in soil columns: Effect of soil organic carbon content. J. Environ. Sci. Health A Tox. Hazard. Subst. Environ. Eng. 43 (7), 682–693. Speitel, G.E., Yamamoto, H., Autenrieth, R.L., McDonald, T., 2002. Laboratory fate and transport studies of high explosives at the Massachusetts Military Reservation. Final report. University of Texas at Austin and Texas A&M University. Tratnyek, P.G., Scherer, M.M., Deng, B., Hu, S., 2001. Effects of natural organic matter, anthropogenic surfactants, and model quinones on the reduction of contaminants by zero-valent iron. Water Res. 35 (18), 4435–4443. USEPA, 2001. Determination of metals and trace elements in water and wastes by inductively coupled plasma-atomic emission spectometry (Washington, DC). Walsh, M.E., et al., 2012. Accumulation of propellant residues in surface soils of military training range firing points. Int. J. Energetic Mater. Chem. Propuls. 10 (5), 421–435. Wang, F.L., Huang, P.M., 2001. Effects of organic matter on the rate of potassium adsorption by soils. Can. J. Soil Sci. 81 (3 SPEC. ISS.), 325–330. Weng, L., Temminghoff, E.J.M., Lofts, S., Tipping, E., van Riemsdijk, W.H., 2002. Complexation with dissolved organic matter and solubility control of heavy metals in a sandy soil. Environ. Sci. Technol. 36, 4804–4810. White, G.F., Snape, J.R., Nicklin, S., 1996. Biodegradation of glycerol trinitrate and pentaerythritol tetranitrate by Agrobacterium radiobacter. Appl. Environ. Microbiol. 62 (2), 637–642. Xu, W., Dana, K.E., Mitch, W.A., 2010. Black carbon-mediated destruction of nitroglycerin and RDX by hydrogen sulfide. Environ. Sci. Technol. 44 (16), 6409–6415.