William A.
Maza
*,
James A.
Ridenour
,
Brian L.
Chaloux
,
Albert
Epshteyn
and
Jeffrey C.
Owrutsky
Chemistry Division, US Naval Research Laboratory, Washington, D.C., 20375, USA. E-mail: william.maza@nrl.navy.mil
First published on 10th October 2023
The initial reduction dynamics of perfluoroalkyl substances (PFASs) by the hydrated electron, eaq−, is a topic of great interest and importance due to the pervasive environmental presence of PFAS in soil and waters and the need to remediate the contamination. Understanding how PFAS behaves in water, including the potential effect of aggregation on the apparent PFAS reduction rate constant (kPFAS) is, therefore, of paramount importance. In this publication we re-examine the proposition that submicellar aggregation decreases the apparent kPFAS for sodium perfluorocarboxylates of varying chain lengths (NaPFxA, x = number of carbons in the PFAS backbone, ranging from 4 to 8) using transient absorption spectroscopy. We compare the dynamics for eaq− quenching by NaPFxA in aqueous solutions of ferrocyanide, Fe(CN)64−, and sulfite, SO32−. The results demonstrate that the apparent rate constant depends on the choice of eaq− precursor. We demonstrate that the ionic strength of the solution and the counterion of the PFxA salt both affect the measured rate constant of PFAS reduction by eaq−. The results presented here help to better understand PFAS degradation by advanced reduction processes.
Environmental significanceAdvanced reduction processes (ARP) have proven to be quite effective at the destruction of PFAS and other contaminants in tainted waters. However, the processes underlying the reduction of PFAS by hydrated electrons (formed by the UV-excitation of appropriate ions capable of carrying out ARP, like sulfite) are still poorly understood; particularly, the early processes related to the initial reduction of PFAS. Until now, studies utilizing fast spectroscopic techniques like nanosecond transient absorption spectroscopy to elucidate the dynamics between the hydrated electron and PFAS have used ferrocyanide instead of sulfite as the electron donor for a variety of reasons with the assumption that the electron precursor has little effect on the reaction between the hydrated electron and PFAS. In this study we demonstrate that this assumption may not hold true and explore ionic strength effects, as well as counterion effects, to try to pinpoint the factors that make the fast reaction occurring in solutions of ferrocyanide and sulfite so vastly different. We argue that perfluoroalkyl carboxylates (PFxA) tend to aggregate in water at concentrations well below the critical micelle concentration. Furthermore, the concentration of sulfite needed to effectively carry out ARP for practical degradation of contaminated waters shifts the PFxA equilibrium towards aggregation. The implication is that the [PFxA] used in typical degradation studies found in the literature are primarily in some aggregated state. Therefore, the degradation models proposed thus far require revision to include the aggregation of PFxA and the role of aggregation in modulating reactivity with hydrated electrons. |
UV-irradiation of solutions containing sulfite (SO32−) under alkaline conditions (pH > 8) efficiently degrades PFAS by producing hydrated electrons, eaq−, that are photodetached from SO32−.6–8 This process has become the prototype of so-called ‘Advanced Reduction Processes’ (ARP) for high energy reductive decontamination of tainted waters. The large standard reduction potential (−2.9 V)9 of the eaq− is nearly iso-energetic to the potentiometric biases required for the defluorination reaction of many perfluoroalkanes (e.g. and ).10–14 However, most of the work reporting on the fast reaction dynamics of the reduction of PF8A by eaq− has been done in solutions of ferrocyanide (Fe(CN)64−) rather than SO32−.15–18
Fe(CN)64− is a well-studied precursor for generating solvated electrons via photodetachment.19–23 It is an attractive alternative to SO32− for modelling the behaviour of eaq− quenching by PFAS, particularly in the absence of Brønsted acids. Unlike SO32− (pKb = 6.8), Fe(CN)64− is aprotic at pH > 4.3 and has a 100-fold larger molar extinction coefficient (ε) at the 254 nm excitation wavelength (SO32−ε254nm ∼ 50 M−1 cm−1; Fe(CN)64−ε254nm ∼ 5000 M−1 cm−1).24–28 Ferrocyanide also has a high photodetachment quantum yield, even in the near UV.23 In particular, the quantum yield of eaq− formation (Φe) from Fe(CN)64− approaches unity as the excitation wavelength decreases (SO32−Φe ∼ 0.1 at 254 nm excitation, pH 8; Fe(CN)64−Φe ∼ 0.65 at 254 nm excitation, pH 8).28
SO32− and Fe(CN)64− are not unique in their ability to undergo photoinduced ionization to produce eaq−. Other photochemically active anions also produce eaq− and have been used in ARP for the degradation of PFAS. Examples include ethylenediaminetetraacetate (EDTA4−),29 ionic indole derivatives,30 as well as iodide (I−) solutions that produce eaq− in both the absence and presence of SO32−.31–34
We previously reported that the apparent rate constant measured for eaq− quenching by PF8A can be underestimated at high concentrations (above ∼1 mM) due to surfactant aggregation18 including micellization above ∼30 mM. Aggregation of PF8A results in a decrease of the monomeric quencher concentration and appears to reduce the apparent eaq− quenching rate constants.18
As mentioned above, the majority of transient studies probing PFAS reduction dynamics by the eaq− have been carried out in solutions of Fe(CN)64− due to the reasons already outlined. This assumes that the initial reduction of PFAS by eaq− is independent of the source of the eaq−. However, this assumption omits other factors that could potentially modulate eaq− reactivity with PFAS. Due to the differences in the photophysical properties of the various electron donors and the solution conditions required to optimize the degradation of contaminants by ARP, it is imperative to characterize the reduction dynamics of PFAS in each case. In this report, we challenge the supposition that the dynamics of ARP are independent of the eaq− donor, find clear differences between processes carried out in solution of Fe(CN)64−versus SO32−, and present evidence that suggest that these differences are due to differences in the aggregation state of the PFAS induced by the respective solution conditions. The evidence presented here suggests that these differences are, in part, due to ionic strength effects that affect not only the apparent quenching rate constants resulting from well-known screening affects, but also promote submicellar aggregation of PFAS at low concentrations. In addition to the transient absorption studies used to characterize the quenching rate constants, concentration-dependent UV-visible absorption spectra were measured and provide corroborating evidence of aggregation.
The diffusion coefficient for eaq− is reported to be 2.5 × 10−5 cm2 s−1 (r = 1.5 Å),40 and determined by us for PF8A to be 4 × 10−6 cm2 s−1 (r = 4 Å) using 19F DOSY NMR (see ESI, Fig. S1 and S2†). We, therefore, chose a = 3 to correct the observed decay rates (Fig. S3†).
(1) |
(2) |
(3) |
In Fe(CN)64−, the eaq− decay rate depends linearly on [NaPF8A] for [NaPF8A] < 1 mM with a kNaPF8A of ∼3.3 × 108 M−1 s−1. However, the eaq− decay rate becomes nonlinear for [NaPF8A] > 1 mM. These results are consistent with our previous results,18 which we attributed to the submicellar aggregation of NaPF8A. We will note that others have reported on the propensity of PF8A to form premicellar aggregates. For example, Sarmiento, et al.44,45 invoked this process to explain changes in the partial molar volume in aqueous solutions of LiPF8A at submicellar concentrations.
Szajdzinska-Pietek and Gebicki46 reported radiolysis data that also exhibited two distinct linear regimes in the dependence of the eaq− decay rate as a function of the concentration of ammonium perfluorooctanoate, NH4PF8A, in water at [NH4PF8A] between 5 mM and 40 mM. In their work, they reported a discontinuity that occurs near the critical micelle concentration (∼30 mM).
We previously developed a kinetic model for the eaq− decay rate dependence on [NaPF8A]18 that included eaq− quenching terms by both NaPF8A monomers according to reaction 4 and aggregates assuming that the only configuration of the aggregates are dimers according to reaction 5.
(4) |
(5) |
The fits reported were restricted to [NaPF8A] < 2 mM to isolate the dynamic concentration range in which monomers and dimers are presumed to dominate the population of NaPF8A. Although adequate fits of the data < 1 mM NaPF8A were obtained using the model, it was stressed that the PF8A aggregation is likely complex and involves a wide range of aggregation states. We, therefore, will refrain from invoking the dimerization model here and, instead, report kNaPF8A based on linear fits over regions in which the concentration dependence is nearly linear (e.g. [NaPF8A] < 1 mM, 1 mM < [NaPF8A] < 10 mM, and [NaPF8A] > 10 mM). From this point on, [NaPF8A] refers to the total concentration of NaPF8A rather than distinguishing between monomers and aggregates; as a result, the reported kNaPF8A are likely lower than the actual values for the bimolecular rate constants.
A linear fit of the data in 40 mM borate at [NaPF8A] < 1 mM yields a kNaPF8A = (2.0 ± 0.9) × 108 M−1 s−1, which is approximately a factor of 3 smaller, but on the same order of magnitude as the result previously reported at concentrations below 200 μM in neat water, (7.1 ± 0.6) × 108 M−1 s−1.17 The reaction conditions here differ from those of our previous reports due to the presence of borate, which buffered the sample solutions at pH 10. Repeating the experiment where aliquots of NaPF8A were added to a solution of Fe(CN)64− in neat water (Fig. S3B†) yields a kNaPF8A = (5.5 ± 1.6) × 108 M−1 s−1 at [NaPF8A] < 1 mM, which is in excellent agreement with our previous results.17,18 The reason for the near 3-fold decrease in kNaPF8A in borate is not readily apparent and outside the scope of this study.
Linear fits at 1 mM < [NaPF8A] < 10 mM yield a kNaPF8A of (5.9 ± 2.0) × 107 M−1 s−1, which is in agreement with previous results reported by us18 and by Szajdzinska-Pietek and Gebicki46 (e.g., 5.1 × 107 M−1 s−1) for [NH4PF8A] < 30 mM, but still about five times greater than the (1.7 ± 0.5) × 107 M−1 s−1 value reported by Huang, et al.16 Linear fits of the data at [NaPF8A] > 10 mM yield a kNaPF8A of (2.4 ± 0.7) × 107 M−1 s−1, which is within experimental error of the aforementioned results reported by Huang et al.16 and in good agreement with the 1.3 × 107 M−1 s−1 value reported by Szajdzinska-Pietek and Gebicki46 for [NH4PF8A] > 30 mM, above its critical micelle concentration, CMC.
The reduction of PFxA by eaq− is believed to proceed by one of three mechanisms: the associative (reaction 6), concerted dissociative (reaction 7), or stepwise dissociative mechanism (reaction 8). Due to experimental limitations, the dominant mechanism driving PFAS reduction and defluorination has yet to be positively identified. As such, a good deal of theoretical work has been dedicated to disentangling these fast initial reactions.12,47–51
RCF2COO− + eaq− → RCF2(COO−)− | (6) |
RCF2COO− + eaq− → RC˙FCOO− + F− | (7) |
RCF2COO− + eaq− → [R(CF2)−COO−] | (8a) |
[R(CF2)−COO−] → RC˙FCOO− + F− | (8b) |
Daily and Minakata48 used density functional theory to calculate the one electron reduction potentials corresponding to reactions 6–8 for PF4A, PF6A, and PF8A among other organic compounds. From the theoretical reduction potentials the authors then calculated the predicted rate constants for reduction and found that for PF8A the associative and concerted dissociative mechanisms are expected to proceed with a rate constant of 5.8 × 108 M−1 s−1 and 1.7 × 107 M−1 s−1, respectively. These predicted values are in excellent agreement with our results in neat aqueous solutions of Fe(CN)64− at [NaPF8A] < 1 mM and >10 mM, respectively, and in very good agreement with solutions of Fe(CN)64− in borax (Table 1) in the same concentration ranges.
40 μM Fe(CN)64− | k NaPFxA (M−1 s−1) | Literature | Ref. | ||
---|---|---|---|---|---|
[NaPFxA] < 1 mM | 1 mM < [NaPFxA] < 10 mM | [NaPFxA] > 10 mM | [NaPFxA] > 5 mM | ||
NaPF8A | (2.0 ± 0.9) × 108 | (5.9 ± 2.0) × 107 | (2.4 ± 0.7) × 107 | (1.7 ± 0.5) × 107 | 16 |
NaPF7A | (7.7 ± 2.0) × 107 | (2.5 ± 0.4) × 107 | (1.9 ± 0.4) × 107 | ||
NaPF6A | (9.8 ± 2.0) × 107 | (3.3 ± 0.3) × 107 | (2.4 ± 0.3) × 107 | ||
NaPF5A | (6.7 ± 1.6) × 107 | (1.5 ± 0.3) × 107 | (1.3 ± 0.1) × 107 | ||
NaPF4A | (1.0 ± 0.2) × 107 | (3.7 ± 0.2) × 107 | (2.5 ± 0.3) × 107 | (7.1 ± 0.3) × 106, (1.3 ± 0.1) × 107 | 15 and 16 |
10 mM SO32− | k NaPFxA (M−1 s−1) |
---|---|
NaPF8A | (5.6 ± 0.2) × 107 |
NaPF7A | (2.0 ± 0.1) × 107 |
NaPF6A | (2.9 ± 0.2) × 107 |
NaPF5A | (1.0 ± 0.1) × 107 |
NaPF4A | (2.0 ± 0.2) × 107 |
We note that a recent study has disputed the nonlinearity observed by us for eaq− quenching in solutions of Fe(CN)64− at PFAS concentrations < 1 mM.15 In order to verify that our observed trend is not due to impurities in our commercially obtained K4Fe(CN)6, we performed the same experiment where aliquots of NaPF8A were incrementally added to aqueous solutions (in the absence of borate) of K4Fe(CN)6 obtained from two different suppliers and three different degrees of purity. We found the trend and results, particularly at [NaPF8A] < 1 mM, to be virtually indistinguishable between the three data sets (Fig. S3B†).
Interestingly, the interactions between eaq− and NaPF8A are quite different in 10 mM SO32− solutions compared to Fe(CN)64− over similar [NaPF8A]. Upon addition of NaPF8A to 10 mM SO32−, negligible eaq− lifetime quenching was observed below 200 μM. At [NaPF8A] > 200 μM, however, the eaq− decay rate increases linearly with [NaPF8A] up to 10 mM and has a kNaPF8A of (5.6 ± 0.2) × 107 M−1 s−1, which correlates well with the values obtained here for the kNaPF8A in 1 mM < [NaPF8A] < 10 mM solutions of Fe(CN)64−. A nonlinear dependence of the eaq− decay rate on [NaPF8A] is also observed in solutions of 1 mM KI (Fig. S4†) in neat water; however, the dependence of the eaq− decay rate on [NaPF8A] becomes increasing linear when the concentration of KI is increased. The kNaPF8A between 1 mM < [NaPF8A] < 10 mM was also observed to decrease from 5.9 × 107 M−1 s−1 in 1 mM KI to 3.5 × 107 M−1 s−1 in 10 mM KI. Moreover, a significant difference in the kNaPF8A obtained between 1 mM < [NaPF8A] < 10 mM was noted when comparing results in 10 mM KI and 80 μM indole to 40 μM Fe(CN)64− and 10 mM SO32−.
There is less of an effect from adding NaCl to the 10 mM SO32− because it has a higher initial ionic strength than that of the 40 μM Fe(CN)64− so the former is less susceptible to additional increases in the ionic strength. Increasing the NaCl concentration promotes NaPF8A aggregation but the effect is reduced in sulfite solution because the ionic strength is already high. This interpretation is corroborated by the trend observed in varying concentrations of KI with NaPF8A (Fig. S4†) in which the non-linearity of the data decreases with increasing [KI] and, as a result, the solution ionic strength (vide supra). The results obtained at high [NaCl] lend further support to this proposal. As noted earlier, the critical micelle concentration is known to decrease with increasing ionic strength.55,56 Analogously, we believe that the interactions between PFAS monomers increase with increasing ionic strength shifting the equilibrium from monomers to aggregates.
Fig. 4 Modified Stern–Volmer plots of the eaq− lifetime dependence on the [NaPF8A] (black) and [NH4PF8A] (red) in 40 mM borate solutions of (A) 40 μM Fe(CN)64− and (B) 10 mM SO32−. |
Fig. 5 Modified Stern–Volmer plots of the eaq− lifetime dependence on the [NaPFxA] in 40 mM borate solutions of (A) 40 μM Fe(CN)64− and (B) 10 mM SO32−. |
As a demonstration that surfactant absorption can be used to identify aggregation, steady-state spectra were measured for aqueous solutions with increasing amounts of NaPFxA up to 40 mM. At low concentrations of NaPF8A (<100 μM), the spectrum obtained is relatively featureless at wavelengths > 200 nm (Fig. 6). A shoulder centered at ∼210 nm emerges at ∼100 μM NaPF8A and increases in prominence with increasing [NaPF8A]. At [NaPF8A] > 2 mM, another shoulder is observed centered at ∼255 nm; at much higher concentrations another feature is observed at ∼288 nm. A small deviation from linearity in the steady-state absorption is observed between 5 mM and 40 mM (as shown in the inset of Fig. 6). The first derivative plot of the absorbance at 230 nm with respect to the [NaPF8A] displays an inflection point between 30 mM and 38 mM; this is consistent with the reported CMC values between 33 mM and 36 mM for NaPF8A.60–66 When adding NH4PF8A to neat water up to 10 mM, we observed only a feature at 210 nm up to 10 mM.
Fig. 6 (A) Steady-state absorbance of aqueous solutions (in the absence of borate, Fe(CN)64−, and SO32−) at concentrations of NaPF8A increasing from 40 μM to 40 mM. (B) First derivative of the steady-state absorbance at 230 nm as a function of the concentration of NaPF8A between 5 mM and 40 mM. The dotted red line is the CMC value reported for NaPF8A in ref. 60. (C) Steady-state absorbance at 230 nm as a function of the concentration of NaPF8A (black) and NH4PF8A (red) up to 10 mM; the inset shows the first-derivative of the same data between 0 mM and 2 mM. (D) Steady-state absorbance at 230 nm as a function of the concentration of NaPF8A (black), NaPF7A (red), NaPF6A (blue), NaPF5A (green), and NaPF4A (purple) up to 10 mM; the inset shows the first-derivative of the same data between 0 mM and 2 mM. |
Results similar to those for NaPF8A were observed for other NaPFxAs up to 10 mM. In all cases, a prominent inflection point in the steady-state absorbance data of aqueous NaPFxA solutions was observed at [NaPFxA] < 0.5 mM, well below the respective CMCs, which we attribute to the formation of pre-micellar aggregates (vide infra).18,44,45 It is interesting that this inflection feature occurs at nearly the same concentration for each NaPFxA when the CMCs are very much chain length dependent and vary in magnitude by a considerable amount suggesting that premicellar aggregation is less influenced by the chain length and dominated by some other property.60,66 It is also noteworthy that, although absent in the transient absorption data in solutions of SO32−, deviations from linearity in the observed eaq− decay rates as a function of [NaPFxA] in solutions of Fe(CN)64− occur at nearly the same concentrations lending more weight to the argument that the trends observed in the transient absorption data in solutions of Fe(CN)64− are due to aggregation effects. We believe that the latter is not coincidental and that the physical process inducing the nonlinearity in the steady-state absorbance is the same process that underly the trend in the transient absorption data in solutions of Fe(CN)64−; namely, submicellar aggregation of NaPFxA.
We found that the kNaPF8A for eaq− quenching by NaPFxA is not constant over the quencher concentration range for ferrocyanide. Additional studies were carried out to demonstrate the aggregation of NaPFxA, even below the CMC. Ionic strength is known to not only influence the measured rate constants for eaq− due to electrostatic screening, which we accounted for, but it can also influence aggregation, as shown here. The results of other measurements further corroborate our interpretation that aggregation explains the anomalous, concentration-dependent rate constant values. The eaq− decay rate was also found to vary when the PFxA counterion was changed in a manner consistent with aggregation. Steady-state UV-visible spectra of NaPFxA without ferrocyanide or sulfite (or borate) deviate significantly from the Beer–Lambert law well below their CMC, which is yet another indication of submicellar aggregation of PFxA. The results here reinforce our previous interpretation that PFxA exhibit strong aggregation tendencies, which should be taken into account to get accurate rate constants when measuring solvated electron quenching kinetics.
Footnote |
† Electronic supplementary information (ESI) available. CCDC 2279639 contains the supplementary crystallographic data for the crystal structure of NaPF8A in Fig. S7 in the supplemental. For ESI and crystallographic data in CIF or other electronic format see DOI: https://doi.org/10.1039/d3va00223c |
This journal is © The Royal Society of Chemistry 2023 |