Investigation of the Coupled Effects of Molecular Weight and Charge-Transfer Interactions on the Optical and Photochemical Properties of Dissolved Organic Matter

Abstract

We studied the formation of photochemically produced reactive intermediates (RI) from dissolved organic matter (DOM). Specifically, we focused on the effects of variable molecular weight and chemical reduction on the optical properties of DOM (absorbance and fluorescence) and the formation of singlet oxygen (¹O₂), DOM triplet excited states (³DOM*), and the hydroxyl radical (^•OH). The data are largely evaluated in terms of a charge-transfer (CT) model, but deficiencies in the model to explain the data are pointed out when evident. A total of two sets of samples were studied that were subjected to different treatments; the first set included secondary-treated wastewaters and a wastewater-impacted stream, and the second was a DOM isolate. Treatments included size fractionation and chemical reduction using sodium borohydride. Taken as a whole, the results demonstrate that decreasing molecular weight and borohydride reduction work in opposition regarding quantum efficiencies for ¹O₂ and ³DOM* production but in concert for fluorescence and ^•OH production. The optical and photochemical data provide evidence for a limited role of CT interactions occurring in lower-molecular-weight DOM molecules. In addition, the data suggest that the observed optical and photochemical properties of DOM are a result of multiple populations of chromophores and that their relative contribution is changed by molecular-weight fractionation and borohydride reduction.

Graphical Abstract

INTRODUCTION

The fate and transport of organic contaminants in aquatic systems is an important research topic in the field of environmental engineering and chemistry. Contaminant removal and transformation mechanisms in these environments include sorption, biotransformation, volatilization, and many abiotic chemical reactions.¹ One reactive pathway for contaminant degradation is through direct or sensitized photolysis.^1,2 Direct photolysis of a contaminant occurs due to the absorption of light by that chemical, which may induce its degradation. Sensitized degradation (commonly known as indirect photolysis) of a contaminant can occur through the production of reactive intermediates (RI) following absorption of light by chemical sensitizers. Multiple chemical compounds can act as sensitizers in aquatic systems, such as nitrate (NO₃⁻), nitrite (NO₂⁻), and dissolved organic matter (DOM).¹

DOM has been recognized as a significant absorber of light and sensitizer, leading to RI formation in aquatic systems as well as some atmospheric systems.^3–5 Absorption of light by DOM produces a variety of RI, such as singlet oxygen (¹O₂), DOM triplet excited states (³DOM*), hydrogen peroxide (H₂O₂), and the hydroxyl radical (^•OH).^1,6–15 These species can oxidize or reduce organic contaminants,^16–18 inactivate pathogens,^19,20 and are involved in carbon cycling in aquatic systems.²¹ Although many studies have examined the photochemical formation of RI from DOM, there are still unanswered questions regarding the underlying mechanisms. Some of these mechanisms and knowledge gaps are reviewed below.

Reactions 1–3 illustrate the formation of ³DOM* and ¹O₂.

$$\text{DOM}+\text{h}\nu \to {}^1\text{D}\text{OM*}$$ (1)

$${}^1\text{D}\text{OM*}\stackrel{k_{\text{isc}}}{\to }{}^3\text{D}\text{OM*}$$ (2)

$${}^3\text{D}\text{OM*}+{\text{O}}_2\stackrel{Y}{\to }{}^1\text{O}{}_2+\text{DOM}$$ (3)

where k_isc is the rate of intersystem crossing, and Y is the yield of ¹O₂ from quenching of ³DOM* by O₂. Although the energy difference between ¹O₂ and its triplet ground state is only 94 kJ mol⁻¹, only approximately 30–50% of the ³DOM* pool is thought to contribute to ¹O₂ formation.²² It was shown that ~50% of ³DOM* derived from commercial and isolated humic substances, as well as natural water samples, had energies of at least 250 kJ mol⁻¹.²³

The formation of ^•OH from DOM photolysis is less-understood because typically used probes can be oxidized by other, unknown photooxidants. Indeed, the combined use of arene probe compounds and methane has shown that DOM photolysis produces a combination of free ^•OH and so-called low-energy hydroxylators.^24,25 The exact identity of these latter species are unknown, but two possibilities are water–quinone exciplexes²⁶ or DOM radicals.²² Furthermore, the sources of free ^•OH in this system are not known. In both DOM isolates and natural waters, some free ^•OH is attributable to the photo-Fenton reaction, but this cannot account for all of it.^24,27,28 Recent work suggests that hydroxy aromatic acids may be important sources of free ^•OH from DOM photolysis.²⁹

Various models have been proposed to explain the photophysical and photochemical properties of DOM. One model that has received much attention in recent years is that of charge-transfer (CT) interactions, recently reviewed by Sharpless and Blough (2014).²² In this model, electron-rich donors (D groups) and electron-poor acceptors (A groups), which are thought to be in close proximity within a DOM molecule, can interact in the ground or excited state via the partial or full transfer of an electron (from D to A) to form a CT state, (D⁺/A⁻). D and A groups are thought to result from the oxidation of lignin precursors, and model compounds for typical D and A groups are hydroxylated and alkoxylated aromatics and aromatic ketones and aldehydes or quinones, respectively. It is thought that the presence of CT states provides: (i) lower-energy electronic transitions (excitation into CT states), thereby explaining the long-wavelength absorption of DOM;^30,31 (ii) a deactivation pathway for local singlet excited states (¹DOM* → DOM^•+/•−),^30–32 thereby explaining long-wavelength fluorescence at short excitation wavelengths; and (iii) a relaxation pathway for local triplet excited states (³DOM* → DOM^•+/•−).³³ A total of two different notations will be used for CT complexes in this paper: one is DOM^•+/•−, and the other is (D⁺/A⁻). Additional details regarding CT are given in the Supporting Information; also see references 22, 34, and 35 for information on these two notations, respectively. Although additional studies have indicated other physical bases for some of these processes,³⁶ there has yet to be a comprehensive model for DOM photophysics and photochemistry proposed other than CT interactions. For example, in contrast to the CT model, it could be assumed that DOM’s photochemical and photophysical behavior results from a superposition of individual chromophores, where the behavior of the whole is the average of its components.

Another important variable influencing DOM photophysics and photochemistry is the molecular-weight distribution of the DOM molecules.^{12,15,25,37–39} DOM of lower molecular weight exhibits increased spectral slopes (S),^13,32,40,41 E2/E3 values (Abs₂₅₀/Abs₃₆₅),^12,32,40,41 and increased fluorescence, ¹O₂, ³DOM*, and ^•OH quantum yields.^12,13,15,42 S and E2/E3 are thus thought of as surrogates for molecular weight and (indirectly) the prevalence of CT interactions. Although the relationship between molecular weight and S or E2/E3 has been documented for some time, the specific physical bases have not been firmly established. On the basis of a negative correlation between S and average gel-permeation chromatography retention time, Boyle et al. (2009) suggested that higher-molecular-weight DOM has, on average, a greater number and variety of CT interactions that are possible within the larger-size ensemble.³² While conceptually possible, this idea is difficult to verify and assumes a consistent functional-group composition between molecular-weight fractions, which may not be the case.^37,43 Indeed, the relationships between DOM optical properties, photochemistry, and molecular weight are not completely understood. For example, it is unknown whether size fractionation, either by natural (e.g., biological) or intentioned (e.g., SEC or ultrafiltration) means produces samples of identical chemical composition to that of the unfractionated sample. This has been noted in previous studies,^12,15 and thus an important knowledge gap remains.

Therefore, we have studied the relationship between DOM molecular weight and its optical and photochemical properties for both whole water samples (secondary-treated wastewaters and a wastewater-impacted stream) as well as an isolated terrestrial fulvic acid from the Suwannee River. A solar simulator was used to measure “environmental” quantum yields (for simplicity, the term “quantum yield”, Φ, will be used). Chemical (base modification and coagulation with alum) and physical (size fractionation using an ultrafiltration membrane) treatments were used to examine the effect of molecular weight. Aliquots of the molecular-weight fractions obtained by ultrafiltration were chemically reduced with borohydride to decouple the effects of molecular weight and carbon oxidation state. Base modification, coagulation, and ultrafiltration are shown below to lower the average molecular weight of each DOM sample. This relates to the CT model because lower-molecular-weight DOM molecules are hypothesized to be less-able (probabilistically) to form CT states. Reduction with borohydride affects DOM photophysics via a distinctly different mechanism. Borohydride reduction reduces carbonyl groups to alcohols (R₂C=O → R₂HCOH), which decreases the number of A groups within DOM molecules. As a result, the CT model would again predict that borohydride-treated DOM is less-able to form CT states.

Some of the data are interpreted in light of the CT model, but deficiencies in the ability of this model to explain the data are discussed. In addition to providing new insight into the photophysical and photochemical processes of DOM, the data presented here will be useful for wastewater-treatment systems employing lime softening (mimicked by base modification) or coagulation in assessing the photochemical activity of discharged effluents.

MATERIALS AND METHODS

Samples.

Suwannee River Fulvic Acid (SRFA, 1S101F) was purchased from the International Humic Substances Society. A total of three secondary-treated wastewaters were used: Boulder Wastewater (BWW); Orange County Water District (OCWD), as produced by the Orange County Sanitation District and used by OCWD as influent to an advanced wastewater recycling facility; and Longmont (LM) wastewater. An additional sample was collected from Boulder Creek (BC) at 75th and Jay St. (Boulder, CO). Additional sample details are provided in the Supporting Information. Samples were collected in precombusted (550 °C for 4 h in a muffle furnace) amber glass bottles, filtered with precombusted and rinsed (1 L Milli-Q water, 18 MΩ-cm) 0.7 μm GF/F filters, and stored at 4 °C. Water-quality data are presented in Table S1. Sample pH adjustment, when indicated, was performed with concentrated phosphoric acid or sodium hydroxide.

Experimental Matrix.

It should be made explicit at this point that all treatments were not performed on every sample because of the inherent differences in the samples (isolates versus more complex natural and anthropogenic sources). A total of two wastewaters (BWW and OCWD) and the wastewater-impacted stream (BC) were subjected to base modification, one wastewater (LM) was subjected to coagulation with alum, and SRFA was subjected to both ultrafiltration and reduction with borohydride. There are a few reasons for this. First, we demonstrate that base modification, coagulation with alum, and ultrafiltration all show results characteristic of high-molecular-weight removal. Base modification and coagulation with alum both resulted in the coagulation of organic matter, and ultrafiltration is a well-documented size-fraction technique. Because the effects following each treatment were similar, we did not perform all size-fractionation procedures on each sample (e.g., coagulation of OCWD or SRFA). We chose to reduce only SRFA molecular-weight fractions with borohydride, and not wastewater samples, due to the complex background matrix of inorganic constituents in wastewater. This choice seems justified on the basis that most of the papers reporting CT interactions in DOM use SRFA as an exemplar^{30–32,44,45} and that the results for SRFA appear to be general across DOM of different types and origins.^31,35

Base Modification and Coagulation Procedure.

A total of two chemical size-fractionation procedures were used. The first was a base-modification process⁴⁶ in which the pH of a 1 L aliquot was quickly (less than 1 min) adjusted to 11.0 ± 0.2 and stirred for 1 h at a constant rate, which resulted in formation of a gray precipitate. After 1 h, the solution was filtered, and the solution pH was adjusted to 7.2. Nonbase-modified samples were also adjusted to pH 7.2 to remove the effect of pH on measured optical and photochemical properties. As shown in Table S1, this process resulted in significant coagulation of organic matter for wastewater samples (~13% and 10% reduction in DOC for BWW and OCWD, respectively) but less for the wastewater-impacted river water (~3% reduction in DOC). The minimal change in DOC, optical properties, and SEC chromatograms for the BC sample that underwent minimal coagulation is taken as evidence that the structure of the organic matter was unchanged after adjustment to pH 11.

Coagulation of LM wastewater with aluminum sulfate hexadecahydrate (alum) (Al₂(SO₄)₃·16H₂O) was performed using a jar tester (Phipps & Bird) in 1 L volumes at doses of 0, 30, 60, 90, and 120 mg alum L⁻¹. Following coagulant injection, samples were rapidly mixed (290 rpm for 1 min), followed by two flocculation phases (10 min at 55 rpm and 10 min at 20 rpm) and a sedimentation period (30 to 60 min with no mixing). The supernatant was filtered through 1.5 and 0.7 μm GF/F filters in series and adjusted to pH 7.2. Chemical analyses, fluorescence measurements, and photochemical experiments were performed within 1 week of coagulation.

Borohydride-Reduction Procedure.

SRFA was dissolved in 10 mM phosphate buffer at a concentration of 80 mg L⁻¹ (43.2 mg_C L⁻¹) by stirring overnight. After filtration, this solution was fractionated by passing the 80 mg L⁻¹ solution through a 5-kDa membrane (Millipore; Billerica, MA) to obtain a <5K (permeate) and >5K (retentate) fraction. Aliquots presaturated with N₂ (30 min.) were reduced with sodium borohydride (30 mg sodium borohydride per mg DOM) under N₂. Sodium borohydride was added over a period of 5 to 10 min, and the reaction mixture was stirred vigorously for 3 h. The mixture was resaturated with breathing-quality air (28% O₂/72% N₂, Airgas) after dilution with buffer to ~10 mg_C L⁻¹, and the solution pH was adjusted to 3 and then readjusted to 7.2. The long reaction time as well as the absence of hydrogen (H₂) evolution after adjustment to acidic pH was taken as evidence of no residual borohydride. Absorbance spectra of the reduced and native samples were monitored throughout the course of study (Figure S1) and did not change.

Analytical Methods.

The dissolved organic carbon (DOC) content of filtered samples was measured after acidification to pH < 2 with phosphoric acid with a TOC-V_SCH (Shimadzu Corp.) analyzer using a nonpurgeable organic carbon method. The accuracy and precision of 1.4 and 3.0 mg_C L⁻¹ potassium hydrogen phthalate standards were within 5% throughout the course of this study. Nitrate (NO₃⁻) and nitrite (NO₂⁻) were analyzed by the Laboratory for Environmental and Geological Sciences (LEGS) at CU Boulder using ion chromatography. Total iron was measured by LEGS using atomic-emission spectroscopy (AES) with a reported detection limit of 0.002 ppm. Absorbance was measured in triplicate with a Cary Bio 100 (Agilent Technologies; Santa Clara, CA) from 200 to 800 nm in a 1 or 5 cm quartz cuvette and baseline-corrected to deionized water. Fluorescence was measured in triplicate on a Fluoromax-4 (Horiba) with excitation wavelengths ranging from 240 to 550 nm in 10 nm increments and emission scans collected from 300 to 700 nm in 2 nm increments.⁴⁷ The bandpass for both excitation and emission monochromators was 5 nm, and the integration time was 0.25 s. Raw fluorescence data were corrected using the method of Murphy et al. (2010),⁴⁸ and quantum yields were calculated by using quinine sulfate as a reference standard following the method of Cawley et al. (2014).⁴⁷ In many cases, samples were diluted for fluorescence measurements such that the maximum emission intensity was less than 2 × 10⁶ counts per second to avoid detector saturation. Size-exclusion chromatography (SEC) was performed using previously described methods.⁴⁹

Photochemical Measurements.

Data shown in figures and tables represents an average of at least duplicate irradiations, and error bars represent two standard deviations (s). Samples were irradiated using an Oriel 94041A Solar Simulator (Newport Corp.) with an AM 1.5 filter in clear, borosilicate glass vials laid flat in a water-jacketed Petri dish that was cooled with a chiller to 20 ± 2 °C. Lamp spectra were measured with an Ocean Optics USB 2000 spectroradiometer. Probe compounds used for the detection of RI have been described previously, and details are provided in the Supporting Information. A lamp spectrum and description of the polychromatic quantum yield calculation is also described in Figure S2. Polychromatic quantum yields are denoted from this point on as Φ_f, Φ_1O2, and Φ_OH for fluorescence ¹O₂, and ^•OH, respectively and f _TMP represents the TMP degradation efficiency (at [TMP]₀ = 5 μM).

RESULTS

DOM Size Characterization.

SEC data is plotted in Figures S3 to S6. The data generally show results consistent with a decrease in apparent molecular weight upon base modification, coagulation, and ultrafiltration (average retention times given in Table S2). Coagulation and ultrafiltration have previously been shown to remove high-molecular-weight molecules from DOM.^50,51 It is important to note that E2/E3 values, which have a negative correlation with molecular size, increased as a result of each of these treatments, which corroborates the SEC measurements.

A critical assumption in the use of borohydride is that neither borohydride nor OH⁻ ions from the high pH it causes induces reactions involving esters (e.g., hydrolysis), which would result in lower-molecular-weight compounds. SEC was used to test for this, and the data for each fraction pre- and postreduction are shown in Figure S6. Although a peak at a retention time of ~45 min appears in all reduced samples, there is essentially no change in average retention time, and thus borohydride reduction can be considered, at a first approximation, as a means to examine the effect of oxidation state independent of molecular weight (see the Supporting Information for additional discussion).

Optical Changes following Physicochemical Treatments.

Base modification, coagulation, and ultrafiltration all caused changes in absorption that have been attributed to removal of high-molecular-weight DOM (see E2/E3 values in Table 1). Figure 1 shows the molar absorptivity (units of M_C⁻¹ cm⁻¹) and fraction molar absorptivity remaining for samples subjected to each of these treatments (see also Figure S7 and S8). All samples exhibited a preferential decrease in molar absorptivity with increasing wavelength into the visible region of the spectrum, as does borohydride-treated SRFA (Figure 2 and ref 31). The ratio of reduced to native molar absorptivity (Figure 2) decreases systematically for each molecular-weight fraction to ~450 nm, with varying behavior at longer wavelengths. Molar absorptivity is used because of varying carbon concentrations across samples.

Table 1.

Apparent Quantum Yields ±2s for All Samples in This Study^a

			fluorescence		reactive intermediates
sample	treatment	E2/E3	λ max	Φf,max	Φ1O2	fTMP (M−1)	ΦOH × 105
BC	none	5.56	380	0.0570 ± 0.0024	0.0365 ± 0.015	88.9 ± 7.8	0.98 ± 0.05
BC	base modification	5.61	380	0.0579 ± 0.0044	0.0382 ± 0.001	81.4 ± 1.2	1.00 ± 0.50
BWW	none	5.25	370	0.0211 ± 0.0012	0.0277 ± 0.005	37.1 ± 3.8	3.97 ± 0.74
BWW	base modification	5.79	380	0.0270 ± 0.0032	0.0358 ± 0.007	69.2 ± 2.6	4.81 ± 0.74
OCWD	none	5.01	360	0.0251 ± 0.0018	0.0261 ± 0.001	118.9 ± 0.2	26.28 ± 2.07
OCWD	base modification	5.14	370	0.0295 ± 0.0012	0.0371 ± 0.005	136.4 ± 37.5	35.37 ± 0.74
LM	none	4.59	360	0.0285 ± 0.0012	0.0236 ± 0.003	23.0 ± 2.7	6.13 ± 0.95
LM	30 mg alum/L	4.88	370	0.0330 ± 0.0016	0.0278 ± 0.002	28.7 ± 2.8	7.02 ± 0.45
LM	60 mg alum/L	5.21	370	0.0396 ± 0.0018	0.0300 ± 0.004	49.0 ± 4.2	8.25 ± 0.81
LM	90 mg alum/L	5.37	370	0.0428 ± 0.0022	0.0352 ± 0.004	66.3 ± 1.8	9.22 ± 0.66
LM	120 mg alum/L	5.51	370	0.0456 ± 0.0022	0.0363 ± 0.001	79.1 ± 2.7	10.85 ± 0.49
SRFA < 5K	fractionation	5.46	390	0.0176 ± 0.001	0.0298 ± 0.004	40.7 ± 4.7	1.71 ± 0.50
SRFA < 5K	fractionation and reduction	7.37	390	0.0321 ± 0.0022	0.0256 ± 0.000	30.1 ± 0.2	2.85 ± 0.74
SRFA	none	4.73	390	0.0095 ± 0.0004	0.0181 ± 0.001	25.7 ± 1.2	1.58 ± 0.07
SRFA	reduction	6.36	390	0.0204 ± 0.001	0.0164 ± 0.000	25.0 ± 0.7	2.02 ± 0.43
SRFA > 5K	fractionation	4.57	410	0.0071 ± 0.0006	0.0154 ± 0.001	21.5 ± 2.5	1.10 ± 0.37
fractionation and reduction	fractionation and reduction	6.33	340	0.0160 ± 0.0014	0.0152 ± 0.001	17.1 ± 1.6	1.91 ± 0.45

^aΦ_f measured from 280–500 nm. Φ_1O2, f _TMP, and Φ_OH are polychromatic (290–400 nm) quantum yields.

Figure 1.

Carbon-based molar absorptivities plotted from 254 to 600 nm for wastewater samples and size-fractionated SRFA. (Right) Fraction of the remaining absorptivity based on data in left panels. Absorbance scans collected with a 1 cm path length. [DOC] ≈ 5–10 mg L⁻¹.

Figure 2.

Carbon-based molar absorptivities for SRFA plotted from 254 to 600 nm for SRFA < 5K, unfractionated, and >5K before and after reduction with sodium borohydride. (Right) Fractional absorptivity remaining based on data in left panels. Dashed line represents local f_a,min and λ_min as described in text (f_a,min, λ_min): SRFA > 5K (0.389, 452 nm); SRFA (0.485, 432 nm); and SRFA < 5K (0.620, 442 nm). Absorbance scans collected with a 5 cm path length. [DOC] ≈ 10 mg L⁻¹.

Φ_f as a function of excitation wavelength for LM wastewater and SRFA is shown in Figures S9 and 3, respectively. Maximum Φ_f values and their corresponding excitation wavelengths (λ_max) are listed in Table 1. Similar to previous studies, Φ_f was increased by the removal of high-molecular-weight DOM by coagulation⁴² and ultrafiltration^13,52 as well as by reduction with borohydride.³¹ λ_max did not change for SRFA after reduction with borohydride (except by +70 nm for the >5K fraction). Reduction resulted in a significant increase in Φ_f for all samples. A greater relative difference is observed with increasing molecular weight, but a greater absolute difference is seen with decreasing molecular weight (see Table S3). Finally, the maximum Φ_f for SRFA is in good agreement with the data reported in Ma et al. (2010).³¹

Figure 3.

Apparent fluorescence quantum yields for native and borohydride-reduced SRFA molecular-weight fractions. [DOC] ≈ 4 mg L⁻¹.

There are clear differences between wastewater and SRFA samples regarding the Φ_f–λ_max relationship. Figures S9 and 3 show that Φ_f varies more strongly with the excitation wavelength for LM wastewater than for SRFA (regardless of molecular-weight fraction or carbon oxidation state). This is not just a scaling issue; the ratio of max to min Φ_f is consistently ~3 for LM wastewater and less than ~2 for SRFA. Also, λ_max is ~20 nm shorter for LM wastewater than SRFA. Whole-water EfOM samples (and their corresponding fluorescence properties) are much more heterogeneous than the SRFA isolate.⁵³ In previous work, fluorescence properties of EfOM have been explained in part by individual fluorophores as opposed to the CT model.¹³

Changes in RI Quantum Yields following Physicochemical Treatments.

Φ_1O2, f _TMP, and Φ_OH were measured for each of the samples described in this study (Table 1).

The Φ_OH values here represent a combination of free ^•OH and low-energy hydroxylators because benzene may also be oxidized by the latter species.²⁵ Table 1 shows RI quantum yields for BWW, OCWD, and BC pre- and post-base modification and demonstrates an increase in Φ_1O2, f _TMP, and Φ_OH for BWW and OCWD but not BC. Figure 4 shows RI quantum yields for coagulated LM wastewater plotted versus coagulant dose (note that greater coagulant doses resulted in increased E2/E3). Φ_1O2, f _TMP, and Φ_OH increased systematically as a result of all water-treatment processes. These results are consistent with previous reports, which demonstrate similar changes in RI quantum yields with changes in DOM molecular weight obtained by different processes (e.g., gel electrophoresis and ultrafiltration) as well as changes in E2/E3 for non-(size) fractionated samples and DOM isolates.^13,15,37,54

Figure 4.

Apparent quantum yields for ¹O₂, ^•OH, fluorescence, and TMP degradation efficiencies for LM wastewater pre- and postcoagulation with alum. Data points are the averages of duplicate runs, and error bars represent two standard deviations.

In an attempt to decouple the effect of molecular weight and carbon oxidation state on RI quantum yields, we measured Φ_1O2, f _TMP, and Φ_OH for SRFA molecular-weight fractions before and after reduction with sodium borohydride (Table 1). Previous studies have examined the effect of borohydride reduction on Φ_1O2 and f _TMP for DOM isolates that have not undergone size fractionation.^35,44,45 Similar to Sharpless (2012), we saw no significant change in Φ_1O2 for bulk SRFA.³⁵ However, there was a significant (one-tailed t-test, α = 5%) decrease for SRFA < 5K. This result was paralleled by f _TMP values, except that there was also a significant difference for the >5K fraction. A significant (one-tailed t-test, α = 5%) increase in Φ_OH was observed for each molecular-weight fraction after borohydride reduction.

DISCUSSION

The following discussion seeks to interpret the observed results in light of the different models (i.e., superposition of chromophores, CT) that could be used when studying the photophysics and photochemistry of DOM.

Optical Changes following Physicochemical Treatments.

The presence and preferential relative removal of long-wavelength absorption in this study and others seems to provide evidence for the CT model. This is because the types of chromophores necessary for these low-energy transitions (characteristic of very conjugated π systems) are likely nonexistent or exist at low concentrations within DOM. As a result, the transitions are assigned to excitation into CT states, which would also have high extinction coefficients.^22,30,31 This preferential removal of long-wavelength absorption was observed for wastewater samples undergoing base modification and coagulation (Figure 1) and SRFA following ultrafiltration and borohydride reduction (Figure 2). In the CT model, a reasonable quantitative measure for the proportion of CT interactions is the minimum value of fractional absorptivity remaining following borohydride reduction, f_a,min, which can be defined as eq 4:

$$f_{a,\text{min}}={\frac{{\epsilon }_{\text{red}}}{{\epsilon }_{\text{nat}}}|}_{{\lambda }_{\text{min}}}$$ (4)

where λ_min is defined as eq 5:

$${\frac{d\left({\epsilon }_{\text{red}}/{\epsilon }_{\text{nat}}\right)}{d\lambda }|}_{{\lambda }_{\text{min}}}=0$$ (5)

A lesser f_a,min indicates more CT removal, and a greater f_a,min indicates less CT removal. Because of the varying behavior at λ > 450 nm, a local f_a,min was calculated. As shown in Figure 2, the effect of reduction was more pronounced in higher-molecular-weight fractions, exhibited by a decreasing f_a,min with increasing molecular weight. Absorbance data were collected with a 5 cm path length for this analysis, and the absorbance was always greater than the quantitation limit (average of water blank +10s ≃ 0.0013) at wavelengths shorter than 700 nm. The reduction procedure was repeated on a different size-fractionated SRFA solution (Figure S8) with similar qualitative results. Other potential metrics to quantify CT loss are provided as Table S3 and Figure S11.

DOM optical properties do not generally depend on concentration, so CT contacts are assumed to be intra-molecular;⁵⁵ however, the ability of DOM molecules to form these associations based on known DOM molecular weights has not been discussed. An analysis based on SRFA molecular weight and ¹³C NMR estimation of functional groups is presented in the Supporting Information. This analysis indicates that for multiple CT contacts to exist in SRFA (on the basis of the number of A groups), esters and carboxylic acids must be able to act as A groups. The prevailing CT model states that only ketone or aldehyde carbonyl groups are acceptors.²² It is thus noteworthy that borohydride is usually not able to reduce esters and carboxylic acids.

The percent change in maximum Φ_f for reduced SRFA increased with increasing molecular-weight fraction. This result is consistent with the idea that there are fewer CT interactions in the <5K fraction because ¹DOM* inactivation into CT bands would be in competition with fluorescence. However, one issue that should be noted is that Φ_f values for each fraction are relatively constant with excitation wavelength, especially for excitation wavelengths >350 nm, before and after borohydride reduction (see Figure 3 in contrast to ref 31). It would thus seem that DOM fluorescence is obeying Kasha’s rule (Φ_f independent of excitation wavelength), in contrast to that described by the CT model.³⁰ Although excitation into CT bands and subsequent charge-recombination-induced luminescence can occur at these longer excitation wavelengths, it is unlikely that the process would occur with similar efficiency as fluorescence at shorter wavelengths. This is a subject for future research.

As noted above, the explanation put forth by Boyle et al. (2009)³² that fewer and less-varied CT interactions are possible in low-molecular-weight DOM is conceptually sound but lacks experimental support. The f_a,min values and percent increase in maximum Φ_f (see Table S3 for values) following borohydride reduction of the SRFA molecular-weight fractions reported here strongly support this conclusion.

Changes in RI Quantum Yields following Physicochemical Treatments.

The trend of increasing Φ_1O2, f _TMP, and Φ_OH with decreasing molecular weight is consistent with other reports in the literature.^{12,13,15,34,37} This study shows for the first time that these effects occur following two commonly used water-treatment processes: base modification (to mimic lime softening) and coagulation. On the basis of the consistent increase in SEC retention time and increase in E2/E3 values following these treatments, it is thought that the physical basis for the increase in RI quantum yields observed here is strongly linked to decreases in molecular weight. However, we note that in the case of coagulation with alum, the potential effect of Al-DOM complexes on the photophysics and photochemistry cannot be ruled out (although the residual Al concentrations were <0.03 mg/L; see Table S1). Additional discussion is provided in the Supporting Information.

The experiments involving reduction of SRFA size fractions represent an attempt to decouple the effects of molecular weight and CT interactions on DOM’s optical and photochemical properties. There are two important results from photochemical studies of these samples: (i) that Φ_1O2 decreased following borohydride reduction only for SRFA < 5K, and (ii) that Φ_OH increased following borohydride reduction for all molecular weight fractions. Under the CT model, D and A groups can form CT interactions in either the ground or excited electronic state, which would lessen the ability of these moieties to act independently. Similar to the model of Blough and Sharpless,^30,34,35,56 we propose that three distinct pools of chromophores best describe the photophysics and photochemistry of DOM: D groups, A groups, and CT groups. D and A groups can exist in close proximity within DOM molecules but have the ability to absorb light and lead to (productive) photochemical reactions. CT groups are those coupled D and A groups that have either ground- or excited-state CT character, and their excitation does not lead to photochemical reactions that generate RI. CT character is defined as the contribution of the no-CT-state and CT-state wave functions to the ground- and excited-state complexes (see the Supporting Information for additional discussion).⁵⁷

Considering the results for Φ_1O2, fewer potential CT interactions as the result of decreased molecular weight or reduction would decrease the rate of reactions 6 and 7

$${}^1\left(\text{D}+\text{A}\right)+\text{h}\nu \to {}^1\left({\text{D}}^{+}/{\text{A}}^{-}\right)\text{ direct excitation into CT states}$$ (6)

$${}^1\left(\text{D*}+\text{A}\right)\to {}^1\left({\text{D}}^{+}/{\text{A}}^{-}\right)\text{ decay into CT states}$$ (7)

and increase the rate of reactions 8–10.

$${}^1\left(\text{D}+\text{A}\right)+\text{h}\nu \to {}^1\left(\text{D*}+\text{A}\right)\text{ excitation of local donors}$$ (8)

$${}^1\left(\text{D*}+\text{A}\right)\to {}^1\left(\text{D}+\text{A*}\right)\text{ energy transfer between D and A}$$ (9)

$${}^1\left(\text{D}+\text{A*}\right)\to {}^3\left(\text{D}+\text{A*}\right)+{\text{O}}_2\to {}^1{\text{O}}_2{\text{ intersystem crossing and quenching by O}}_2$$ (10)

A possible explanation for the decrease in Φ_1O2 and f _TMP for the <5K fraction after reduction is derived from the result that lower-molecular-weight DOM has, on average, a fewer number of CT interactions. With this in mind, a reasonable hypothesis is that any nonreduced A groups are less conformationally able to couple to D groups either in the ground or excited state, decreasing the proportion of light absorbed by CT groups in the reduced <5K sample. Another possibility, though also speculative, is that reduction of ketones and aldehydes to alcohols via borohydride increases the ability of ¹O₂ to diffuse from the hydrophobic DOM core, which would result in a greater observed Φ_1O2 (as measured by the aqueous-phase probe compound FFA).³⁶ This change in three-dimensional configuration could occur, for example, due to increased hydrogen bonding between water and newly formed hydroxyl groups.

It is interesting that borohydride reduction increases Φ_OH regardless of molecular-weight fraction. That the behavior of ^•OH is distinct from that of ¹O₂ and ³DOM* following borohydride reduction is similar to behavior observed for the quenching of these RI by halides.¹⁴ In this instance, Φ_OH increased with increased halide concentrations, whereas Φ_1O2 and f _TMP decreased, indicating different precursors. In addition, the consistent increase in Φ_OH suggests that reduction of O₂ or H₂O₂ by CT states (DOM^•+/•− + O₂/H₂O₂ → ^•OH) is an insignificant source of ^•OH.²² A likely reason for this observation is that hydroxyaromatic acids (potential D groups) are significant sources of free ^•OH, as was recently reported.²⁹ Reduction of carbonyl-containing A groups would allow D groups, which are not reducible by borohydride, to be excited into local excited states as opposed to CT bands, increasing the rate of ¹(*D + A) formation. Furthermore, the correlation between Φ_f and Φ_OH (Figure 5) suggests that either ¹DOM* is a precursor to ^•OH or that its deactivation into CT states is in competition with decay into ^•OH precursors (for example, DOM radicals (reactions 11–14)).

Figure 5.

Correlation between ^•OH and maximum fluorescence quantum yields. Data are a combination of SRFA molecular-weight fractions before and after reduction with borohydride and LM wastewater at each coagulation dose. See Table 1 for maximum fluorescence quantum yields and the wavelengths at which they occur.

$${}^1\text{DOM*}\to {\text{R}}^{•}+{\text{DOM}}^{•}$$ (11)

$${}^1\text{DOM*}{\to }^3\text{DOM*}\to {\text{R}}^{•}+{\text{DOM}}^{•}$$ (12)

$${\text{R}}^{•}\to {\to }^{•}\text{OH}$$ (13)

$${\text{DOM}}^{•}\to {\to }^{•}\text{OH}$$ (14)

Excitation into CT bands would also compete with photoionization and subsequent reaction of DOM^•+ with water, a potential (although speculative) source of ^•OH (reactions 15–17).

$$\text{DOM}+\text{h}\nu \to {\text{DOM}}^{+•}+{\text{e}}^{-}$$ (15)

$${\text{DOM}}^{+•}+{\text{H}}_2\text{O}\to {\left[\text{DOM}-{\text{OH}}_2\right]}^{+•}$$ (16)

$${\left[\text{DOM}-{\text{OH}}_2\right]}^{+•}\to \text{DOM}+{\text{H}}^{+}{+}^{•}\text{OH}$$ (17)

Reduction of A groups by borohydride would decrease the rate of reaction 6 and thus increase the rate of reactions 15–17. ¹DOM* is not directly responsible for ^•OH production because its lifetime is too short.⁵⁸ Therefore, it seems that a dominant control on Φ_OH is the ability of D groups to act as independent chromophores. Additional evidence for involvement of DOM radicals and radical cations comes from the fact that apparent activation energies for ^•OH production from DOM photolysis are on the order of ~15 to 30 kJ mol⁻¹,²⁵ which are typical of radical reactions.⁵⁹ Importantly, the results presented in Figure 4 and Table 1 demonstrate that decreasing molecular weight and chemical reduction work in concert to produce greater Φ_OH.

ENVIRONMENTAL IMPLICATIONS

The results presented here have significant consequences on the photochemical processes occurring in natural and engineered aquatic systems. This is clearly demonstrated by the increase in Φ_1O2, f _TMP, and Φ_OH after the application of two commonly used water-treatment technologies (base modification to mimic lime softening and coagulation). Receiving water bodies heavily impacted by wastewater undergoing these treatment processes could expect greater RI steady-state concentrations and potentially increased contaminant removal, although other factors such as changes in light absorption and scavenging of RI by DOM would need to be considered.

Another finding of this research is that the correlations between RI quantum yields and E2/E3 (Figure S12) may be more sample-specific than previously demonstrated.^12,34,56,60 Other studies have reached this conclusion. In particular, Sharpless et al. (2014) observed different dependencies (slopes) between Φ_1O2 and E2/E3 for photo-oxidized humic substances of varying origin⁵⁶ compared to previous reports.^12,34,60 An earlier study by Peterson et al. (2012) showed that this same relationship for water samples collected over a multiyear time frame from Lake Superior was weaker than that observed by Dalrymple et al. (2010)³⁴ (again, in terms of the linear regression slope).⁶⁰ One contrasting point, however, is that Mostafa and Rosario-Ortiz (2013)¹² observed a relationship between Φ_1O2 and E2/E3 for wastewater samples between that of Peterson et al. and Dalrymple et al. (slope of 0.72 versus 0.61 and 0.87). The correlations between Φ_1O2 and E2/E3 observed in this study as well as those just mentioned are summarized in Table S5.

Considering the current study, although physical and chemical treatments applied to these samples caused systematic changes in fluorescence and RI quantum yields that correlated in expected ways with E2/E3, we found that these correlations did not hold across all samples and all treatments, as exemplified by Figure S12 and Table S5. Although E2/E3 may be a useful surrogate for assessing DOM molecular weight and predicting RI quantum yields, this relationship is highly specific on the sample type and treatment. An additional aspect relating to our study in the context of water-treatment applications is that coagulated or base-modified wastewater and borohydride-treated SRFA are not best-characterized as natural samples. Of all of the RI shown in Figure S12, Φ_1O2 correlates the best with E2/E3. Finally, linear regression of Φ_1O2 to E2/E3 for the samples in our study, besides reduced SRFA (which could be considered the least-natural sample) yields a relationship of Φ_1O2 (%) = 1.59 × E2/E3 − 5.29 (R² = 0.733, slope p < 0.0001, and intercept p < 0.01).