Nitrogen amended graphene catalyses fast reduction of vinyl chloride by nano zerovalent iron

Vinyl chloride (VC) is a dominant carcinogenic residual in many aged chlorinated solvent plumes, and it remains a huge challenge to clean it up. Zerovalent iron (ZVI) is an effective reductant for many chlorinated compounds but shows low VC removal efficiency at field scale. Amendment of ZVI with a carbonaceous material may be used to both preconcentrate VC and facilitate redox reactions. In this study, nitrogen-doped graphene (NG) produced by a simple co-pyrolysis method using urea as nitrogen (N) source, was tested as a catalyst for VC reduction by nanoscale ZVI (nZVI). The extent of VC reduction to ethylene in the presence of 2 g/L of nZVI was less than 1% after 3 days, and barely improved with the addition of 4 g/L of graphene. In contrast, with amendment of nZVI with NG produced at pyrolysis temperature (PT) of 950 ◦ C, the VC reduction extent increased more than 10-fold to 69%. The reactivity increased with NG PT increasing from 400 ◦ C to an optimum at 950 ◦ C, and it increased linearly with NG loadings. Interestingly, N dosage had little effect on reactivity if NG was produced at PT of 950 ◦ C, while a positive correlation was observed for NG produced at PT of 600 ◦ C. XPS and Raman analyses revealed that for NG produced at lower PT ( < 800 ◦ C) mainly the content of pyridine-N-oxide (PNO) groups correlates with reactivity, while for NG produced at higher PT up to 950 ◦ C, reactivity correlates mainly with N induced structural defects in graphene. The results of quenching and hydrogen yield experiments indicated that NG promote reduction of VC by storage of atomic hydrogen, thus increasing its availability for VC reduction, while likely also enabling electron transfer from nZVI to VC. Overall, these findings demonstrate effective chemical reduction of VC by a nZVI-NG composite, and they give insights into the effects of N doping on redox reactivity and hydrogen storage potential of carbonaceous materials.


Introduction
Vinyl chloride (VC), a known carcinogen, is found in many contaminated soils and groundwaters due to past uncontrolled releases to the environment but also because it forms and accumulates during microbial reductive dechlorination of per-and tri-chloroethylene (PCE, TCE) widely used as solvents in industries such as wood processing, metal degreasing and dry cleaning (Entwistle et al., 2019).While TCE and PCE are relatively easily degraded via abiotic reductive dechlorination processes, VC is more resistant and persists under anoxic soil conditions (Dolinová et al., 2017;Han et al., 2012).However, despite VC contamination being widespread, relatively little research has been done to develop effective chemical reductants and processes for in-situ clean-up of VC contaminated soils and groundwater.
A few laboratory studies have tested iron based reductants, such as zerovalent iron (ZVI) and soil iron minerals i.e., iron sulfide (FeS), green rust (GR) and magnetite (Fe 3 O 4 ) (Table 1).These showed that VC degradation by iron minerals proceeds very slowly, taking several weeks to months for measurable degradation to be observed.In contrast, VC degradation by ZVI materials can be significantly faster, particularly if amended with a catalyst, such as Pd (Table 1).For example, Elsner et al. (2008) showed that 930 µM VC were degraded by 25 g/L nanoscale ZVI (nZVI) within 4 days, while in the presence 5 g/L Pd-nZVI, 320 µM VC were degraded already after 1.5 h.Noteworthy that nZVI reactivities with VC can also vary strongly with the chosen nZVI synthesis method, exemplified by the study of Elsner et al. (2008), where no reactivity was observed with VC when nZVI was synthesized by the H 2 method, but reasonable reactivity with nZVI synthesized with the borohydride method (Table 1).
The capacity of nZVI to degrade VC under field conditions seems however, magnitudes lower compared to the laboratory studies (Wei et al., 2010(Wei et al., , 2012;;Yang et al., 2018).This is because nZVI is non-selective and reacts much faster with other chlorinated solvent (CS) co-contaminants (e.g., TCE, PCE, 1,1-dichloroethylene, dichloroethane), as well as other oxidants (e.g., trace oxygen, nitrate), due to their higher electrochemical reduction potentials compared with VC.Also side reactions such as Fe • corrosion (i.e., anaerobic oxidation) leads to depletion of nZVI reducing capacity with aging time and the formation of passivating iron oxide shells (Bruton et al., 2015;Guan et al., 2015).Recently developed nZVI modifications such as sulfidation and nitriding (Fe x N) have shown great potential to enhance nZVI selectivity and reactivity towards chlorinated solvents by enhancing electrical conductivity and surface hydrophobicity of nZVI, while protecting the Fe • core from fast anaerobic oxidation (Bhattacharjee and Ghoshal, 2018;Brumovský et al., 2023Brumovský et al., , 2022;;Gong et al., 2021;Islam et al., 2020;Mo et al., 2022).However, given VC is generally reduced via hydrogenolysis, thus requiring access to atomic hydrogen, nZVI sulfidation, which suppresses anaerobic oxidation (i.e., atomic hydrogen formation), has shown insignificant enhancement of VC dechlorination compared to nZVI (Mo et al., 2022).
Another nZVI modification that has gained increasing attention, is the amendment of nZVI with a carbonaceous material (CM).This is because CM materials enable fast sorption of CS (Dong et al., 2017;Su et al., 2013), thus rapid removal and pre-concentration of CS from solution.In some instances, the presence of CM also induced higher rates of CE reduction and product formation, which suggests that certain CMs also act as redox catalysts.For example, Guan et al. (2020) observed an up to 16.6-fold increase in TCE degradation efficiency by mixing nZVI with carbon fibers.Similarly, Vogel et al. (2019) reported that the addition of a fibrous activated carbon to mZVI led to a 5000-fold increase in PCE degradation rate constant, compared to the system with mZVI only.Such enhanced efficiency in the presence of CM is explained by CM mediating electron transfer via its redox active surface functional groups (e.g., phenolic, quinone, and hydroquinone groups), heteroatoms (e.g., N, S, P, B) and graphitic structure (Dorner et al., 2022;Wang et al., 2019;Yuan et al., 2022b;Zhao et al., 2022).Furthermore, it has been proposed that certain CMs can shuttle and stabilize atomic hydrogen generated on ZVI surfaces (Kopinke et al., 2016(Kopinke et al., , 2020;;Li et al., 2023), also referred to as the hydrogen spill-over mechanism.This implies the possibility of achieving higher nZVI selectivity, i.e., lower anaerobic oxidation, potentially to a similar degree as for nZVI modified by sulfidation and nitriding, while also guaranteeing rapid CS removal and degradation from the contaminated water.However, to our knowledge there are currently no studies that investigated the effects of CM on the reduction of VC with nZVI.
Thus, this study aimed to test VC sorption and degradation kinetics in the presence of nZVI and graphene nanoplatelets.Furthermore, a simple co-pyrolysis method was used to synthesize nitrogen doped graphene materials based on recent studies showing enhanced catalytic performance of CM modified with nitrogen sources (Ma et al., 2022).It is hypothesized that graphene doped with nitrogen will exhibit a higher abundance of N-functional surface groups and structural defects, which both have been postulated to enhance electron shuttling (Gao et al., 2022;Li et al., 2022) as well as improve storage of atomic hydrogen (Singla and Jaggi, 2021), and thus will enhance VC degradation rates by nZVI.It is also argued that pyrolysis temperature and N-dosage will be key parameters affecting CM catalytic properties and hence VC degradation kinetics by nZVI.The specific study objectives were as follows: 1) Preparation of N-doped graphene (NG) as a function of pyrolysis temperature and N (urea) dosage followed by determination of VC sorption capacity and degradation kinetics in the presence of nZVI, and 2) Characterization of NG structure and composition and their correlations with VC reactivity to gain mechanistic insight on VC degradation pathway.Finding an effective and selective reductant system for VC degradation is clearly of high importance, and the results presented below show promise for moving towards an efficient and cost-efficient solution for VC groundwater and soil remediation.

Materials
Detailed information of all used chemicals can be found in Text S1.Deoxygenated ultrapure (DU) water (PURELAB Chorus, ELGA) was used throughout the experiments and prepared by bubbling with Ar (purity greater than 99.9%) for 3 h before transfer to an anaerobic glove box (Coy Laboratory, 3% H 2 /97% N 2 ) for overnight equilibration.Synthesis of nZVI and VC reactivity experiments were done inside the anaerobic glove box unless stated otherwise.

Synthesis of nZVI
nZVI was synthesized using the borohydride method (Chen et al., 2012) as described in Text S2.Purity of the synthesized nZVI material was verified by X-ray diffraction (Fig. S1).The nZVI product has been shown to consist of spherical particles with sizes between 20 and 80 nm, that form chain-like aggregates (Chen et al., 2012;Fang et al., 2010Fang et al., , 2011)).

Production of N-doped graphene
Different N-doped graphene (NG) materials were produced by varying pyrolysis temperature (PT) and urea dosage as outlined in Table 2, but using the same basic pyrolysis steps.Specifically, urea (0 -8 g) was first ground to a fine powder using an agate mortar and then transferred to a 100 mL Coors porcelain crucible.Graphene nanoplatelets (4 g) were added and mixed well with the urea powder using a spatula.The crucible (closed with a lid) was then transferred to a muffle furnace (LT3/11, Nabertherm) and pyrolysed at the desired temperature using a heating rate of 150 K/h, a holding time at the target PT of 1 h and a N 2 flow rate of 400 L/h.After cooling to room temperature, the NG was ground using an agate mortar and stored in 100 mL PTFE tube until further use.For comparison, untreated graphene (G) was also pyrolysed at 950 • C, labelled here forth as G950.

VC reactivity experiments
VC reactivity experiments were set up in 10 mL headspace vials.Each vial was added 0.02 g NG, 0.01 g nZVI, 5 mL DU water, and 7 µL VC stock (2000 µg/mL in methanol) and then quickly sealed with a crimp cap and PTFE/silicone septum.This yielded final concentrations of 2 g/L nZVI, 4 g/L NG, and 44.8 µmol/L VC.The sealed vials were moved outside the chamber to shake at 300 rpm for up to 8 days at room temperature.At regular time intervals, duplicate vials were sacrificed to determine VC and product concentrations using a gas chromatograph equipped with a flame-ion detector/electron-capture detector and coupled with a headspace autosampler (HS-GC-FID/ECD, Thermo Scientific).Three controls, with i) nZVI and VC, ii) NG and VC, and iii) VC only were set-up in parallel and in duplicate.
VC reduction extent, RE, was determined from the quantity of measured products as follows: where n products(t) and n VC(0) represent the amount of product at time t and the initially added amount of VC, respectively (both in moles).Note that the subscript added to RE indicates the time of measurement.For example, RE 68h refers to RE measured after 68 h of reaction.Additionally, the time dependent increase in products was also fitted with a pseudo-first-order kinetic model (Eq.( 2)): where k is the first-order rate constant for VC reduction.Note that VC degradation led to the formation of mainly ethylene and some ethane, present in both the headspace and aqueous phase, so n products(t) was calculated as follows: where ethylene and ethane headspace concentrations, c g (µmol/L), were determined by HS-GC-ECD/FID and signal comparison to a standard gas samples of known concentrations (40.1 µM), and the ethylene and ethane aqueous concentrations, c a (µmol/L), were calculated by Eq. ( 4): using Henry's law constant, K H , of 0.12 and 0.0465 for ethylene and ethane (Ai et al., 2019), respectively.Note that total VC concentration in reaction vials were determined by signal comparison to a calibration prepared from VC stock solutions of known concentrations prepared with the same solution: gas volume as for the reaction vials.

NG characterization
NG structure was probed by powder X-ray diffraction (XRD) using a PANalytical X'Pert Pro-MPD instrument with Co-Kα radiation (Kα = 1.789Å, 40 kV, 40 mA).Powders were measured at 2θ from 5 • to 90 • , with a step size of 0.02 • and 10 s step − 1 , and patterns analysed with Jade 6.5 and HighScore.Selected XRD peaks were fitted with Gaussian curves to determine the full width at half maximum (FWHM).The Brunauer-Emmett-Teller (BET) specific surface area (SSA) of NG materials was determined via N 2 sorption using a Micromeritics Gemini VII 2390.Samples were first degassed at 200 • C for 6 h under N 2 flow.Carbon black (004-16,833-00, standard, SSA of 21.52 ± 0.75 m 2 /g) was used as a reference material.The C and N contents were determined using an Elementar Vario Macro Cube instrument.X-ray photoelectron spectroscopy (XPS, Nexsa G2, Thermo Scientific) applying monochromatic Al-K radiation (300 W) was used to determine NG surface elemental composition and speciation.The XPS data were analysed with CasaXPS according to procedures in Ai et al. (2020) and Fairley et al. (2021).This included calibration of C1s peak to 284.8 eV, Shirley background subtraction and peak fitting with Gaussian (70%) -Lorentzian (30%) distribution.NG structural properties were further examined by Raman using a WITec alpha 300R confocal Raman microscope (WITec GmbH) equipped with a 1X objective and running at 100% effect of a 532 nm Ar laser (~60 mW before the objective).Typically, 1 mL of DU water was added to 0.1 g of NG inside a vial and sonicated for 15 min.Then, the mixture was placed on a glass slide and covered with a cover slip.Each spectrum was obtained as an average of 100 scans with 0.1 s exposure.Raman results were processed using MATLAB to determine intensity ratio of the D (~1350 cm − 1 ) and G (~1580 cm − 1 ) bands.Principal Components Analysis (PCA) and Multiple linear regression (MLR) (in OriginPro 2020 software) were used for exploring possible relationships between VC reduction extent (RE 68h ) and NG properties.

Catalytic reactivity of graphene
In reactions with nZVI only, VC removal was barely visible over the monitored 68 h period (~3 days), and only minor amounts of ethylene formed (RE 68h < 1%, Fig. 1a).Similarly, only little ethylene formed in reactions with nZVI and untreated graphene (G) (RE 68h = 2.1%), demonstrating that nZVI on its own or amended with graphene had little effect on VC reduction.Interestingly, when graphene was pyrolyzed at 950 • C (G950) VC reduction increased about 5 times (RE 68h = 11.6%)compared with pristine graphene.In G and G950 amended experiments significant VC removal was observed within the first 2 h, but without corresponding formation of ethylene (Fig. 1a).This was due to VC sorption to the graphene materials, with slightly higher adsorption to G than for G950 (Fig. S2a).

Effect of different NG loadings
The addition of nitrogen-doped graphene pyrolysed at 950 • C (NG950) to reactors with nZVI and VC had a dramatic effect on the extent of VC reduction to ethylene (Fig. 1b), compared to experiments with non-urea amended graphene, G and G950 (Fig. 1a).Furthermore, when the NG950 loading was increased from 1 to 6 g/L, RE 68h increased linearly from 32 to 73%, demonstrating that VC reduction is facilitated by the NG950 and proportional to its concentration (Fig. 1c).With increase loading of NG950 the corresponding pseudo first-order rate constant (k obs ) increased 4.7 times from 0.0159 to 0.0739 h − 1 (Table S1).As observed for the G and G950 systems, the fast initial drop in VC concentration was attributed to sorption to NG950, but now with less sorption than to the undoped G. Similar to k obs , the extent of sorption correlated linearly with the amount of added NG950, as demonstrated by separate sorption experiments (Fig. S2b).

Effect of different NG pyrolysis temperatures (PTs)
The pyrolysis temperature is known as one of the main variables affecting the redox catalytic properties of carbonaceous materials (Pan et al., 2021;Yuan et al., 2022a).Thus, NGs produced at five different PTs ranging from 400 to 1100 • C were tested over 192 h (~8 days).VC degradation rates and extents were highest with NGs produced at PT of 600, 800 and 950 • C, yielding k obs values of 0.0257, 0.0266 and 0.0465 h − 1 , respectively, and RE 192h values > 70% (Fig. 2b).In comparison, for NG produced at PT 400 • C, RE 192h was much lower (13.6%), while at PT 1100 • C, RE 192h was 61.7%.Note that k obs values were not determined for NG400 and NG1100, but they were likely also much smaller given their low RE 68h of only 4.4% and 21.4%, respectively.
In terms of VC sorption, NGs exhibited increased sorption, from 10.3 to 41.8% of the initial added VC, when PT was increased from 400 to 1100 • C (Fig. S3).Thus, a high extent of sorption, as observed for NG1100, does not necessarily correlate with a high VC catalytic reactivity.Similarly, NG600 which exhibited a similar high VC catalytic reactivity as NG800 and NG950, showed a ~50% lower sorption extent than the high temperature NGs (Fig. S3).

Effect of urea dosage for graphene N doping
The amount of nitrogen co-pyrolysed with the carbonaceous substrate is another variable known to affect its redox catalytic properties (Wan et al., 2020).For NG600 materials produced with 2, 4 and 8 g of urea (per 4 g graphene corresponding to urea: graphene mass ratios of 0.5, 1.0 and 2.0, respectively), the RE 68h of VC increased from 32.8% to 61.0% and 67.3%, respectively (Fig. 3).In contrast, for NG950 materials, an increase in urea: graphene ratio from 0.25 to 2.0 had no systematic effect on VC reduction, with RE 24h and RE 68h values around 38% and 68%, respectively (Fig. 3).
The extent of VC sorption by the NG600 and NG950 materials produced with different urea: graphene ratios were somewhat similar, varying from 19.9 to 27.8% for NG600 materials and from 20.7 to 24.9% for NG950 materials (Fig. S4), with no strong dependencies on applied mass ratios.Also, there was no clear correlation between VC sorption and reduction extent of these materials (Fig. S4).

NG characterization and its relation to reactivity
The extent and rates of VC reduction in reactions with nZVI and different NGs are clearly affected by the applied NG loading and the PT used to produce the NG.Furthermore, at lower PT of 600 • C, the chosen urea: graphene mass ratio (i.e., urea dosage) also affected the NG catalytic activity, while this was not the case at a higher PT of 950 • C. No systematic correlation was observed between VC degradation extent/ rates and VC sorption (Figs.S2-S4).Also, for untreated graphene, which exhibited the highest VC sorption (Fig. S2a), no significant catalytic activity (i.e., product formation) was observed over 3 days (Fig. 1a).This suggests that the sorption of NG towards VC likely played a minor role in VC reduction.This also agrees with previous studies on similar reaction systems (Ai et al., 2021).Note that NG specific surface area correlated positively with NG sorption (Fig. S5), thus clearly affecting total VC removal from solution by NG.However, in terms of VC reduction, there is a poor correlation with NG specific surface area (Fig. S5).Instead, it is hypothesized that the NG properties that affect VC reduction are rather to be found in the graphene structure and composition, such as the presence of N-heteroatoms and N and C surface functional groups as they are key for electron and/or atomic hydrogen mediation between nZVI and VC based on previous nZVI research with CMs (Liu et al., 2022;Qu et al., 2022).

Compositional changes with different PT and urea dosage
Elemental analyses showed that the N content in NG materials markedly decreased from 11.1 to 0.78% with increasing PT from 400 to 800 • C, while the C content increased (Table S2).Notably, all NGs produced at PT of 950 and 1100 • C, contained no N, independent of applied urea dosage.The absence of N in NG950 and NG1100 materials was also confirmed by XPS (Fig. S6).
In terms of surface functional groups, XPS showed that C in all NGs eV).The C-C/C = C group is dominant in all NGs, which is attributed to sp2-hybridized C in graphite-type structures.Only small changes in C speciation were observed with increasing PT (Fig. 4a).Specifically, quinoid (>C = O) groups gradually decreased from 14.2 to 6.6%, while C = C/C-C groups increased from 49.9 to 54.4% when PT increased from 400 to 800 • C. For NGs with PT of 950 and 1100 • C, C speciation was very similar independent of urea dosage (Fig. S7).The relative C speciation in NG600 materials produced with varying urea dosages were very similar, and only slight decreases in total C abundance were noted for NG600 materials with increasing urea dosage (Fig. S8).Overall, there was no noticeable correlation between applied urea dosage and surface C speciation and abundance at the two tested PTs, 600 and 950 • C. As discussed above, only NGs produced at PT ≤ 800 • C had measurable N contents.For these NGs, XPS N 1 s spectra showed contributions from pyridinic N, pyrrolic N, graphitic N and pyridine-Noxide (PNO), and their abundance generally decreased with PT increasing from 400 to 800 • C.An exception is PNO, which had higher abundance in NG600 compared to NG400 and NG800 (Fig. 4b).For NG600 materials produced with varying urea dosages, a clear increase in N surface species abundance was observed from 1.50 to 6.43% with increasing urea: graphene ratio from 0.5 to 2.0 (Fig. 4c), but urea dosage did not seem to change their relative abundance in any systematic way.
To sum up, total N content and N surface species in NG materials varied greatly with PT and also with applied urea dosage for NG600 materials, while variations in C content and speciation were less obvious.While N content and VC reactivity correlated positively for NG600 materials produced with varying urea dosage, this was not observed for NG materials produced at higher PT (Fig. S9).Notably, NG950 materials (independent of applied urea dosages) showed the highest VC reactivities but they exhibited no N surface groups or Nheteroatoms and postulated redox active surface C groups such as quinoid groups were also lower compared to other NGs.

Structural changes with different PT and urea dosage
In terms of graphitic structure, XRD patterns of NGs produced at different PT all exhibited a clear diffraction peak at about 30.9 • 2θ representing the (002) reflection of graphite (PDF#008-0415), and minor peaks at 49 • , 52 • and 64.5 • 2θ, representing graphite (100), ( 101) and (004) reflections.Overall, no significant variations in XRD peak intensity and FWHM were observed between the different NG materials (Fig. S10).
NG graphitic structure was further probed with Raman spectroscopy,  where NGs exhibited two prominent bands at ~1350 and ~1580 cm − 1 , corresponding to the D-and G-band of graphene, respectively.The intensity ratio of these two bands (I D /I G ) is commonly used to estimate the defect density, with an increase in I D /I G ratio reflecting an increase in defect density (Ai et al., 2021).Several measurements (up to 24) were made per sample to delineate data variability, with I D /I G standard errors < 0.05 (Table S3).There are many possible reasons for higher defect density, such as impurities, dopants, vacancies, edge defects, and low graphitization (Hu et al., 2015;Terrones et al., 2012).For NGs varying in PT, the I D /I G ratio was around 0.63-0.64 for both NG400 and NG600, then considerably higher at 0.76 and 0.81 for NG800 and NG950 materials, and a little lower again at 0.74 for NG1100 (Fig. 5b).No significant variations in I D /I G ratio were observed as a function of urea dosage, with all tested NG600 materials exhibiting I D /I G values between 0.64 and 0.66, and all NG950 materials exhibiting I D /I G values between 0.81 and 0.83 (Table S3).For comparison, untreated and pyrolysed graphene, G and G950, exhibited I D /I G values of 0.62 and 0.73, respectively.Overall, these data indicated a clear difference in defect density between NG600 and NG950 materials, independent of applied urea dosage.In summary, NGs produced at higher PT (e.g., NG950 materials) seemed to have more defects (i.e., higher I D /I G ratios), contained no N species, exhibited high catalytic reactivity for VC reduction, and varying urea dosages had no significant impact on these properties.In comparison, NGs produced at lower PT (NG600 materials) seemed to exhibit fewer defects than G, G950 and NG950 materials (i.e., lowest I D /I G ratios), and catalytic reactivity correlated positively with N content/species, i.e., urea dosage.Overall, this showed that N amendment greatly enhanced the catalytic performance of graphene, but the NG properties that contribute to VC reduction are clearly different at different PT, i.e., 600 and 950 • C, as probed here.

Relationship between structural defects, PNO and dechlorination
Urea is fully decomposed when heated to 400 • C and higher.During pyrolysis, gasses/byproducts form, mainly NH 3 , CO 2 , cyanuric acid and melamine, which then react with the carbonaceous matrix to form N surface functional groups and N-substituted graphene.However, the process involved is complex and not yet fully understood.There are currently two main hypotheses (Wan et al., 2020): (1) the formed NH 3 combines with oxygen-containing functional groups of the carbonaceous matrix, which leads to N retention and gradual conversion to various N bonding configurations, and (2) carbon nitride evolves at 300 • C and forms various aminated moieties on the surface of the carbonaceous matrix.Once the temperature rises above 400 • C, these N-based moieties enter the carbonaceous matrix lattice and form various N bonding configurations.
Most NGs were produced using the same urea: graphene ratio (2: 1) and the same heating rate (150 K/h), thus they exhibited the same urea induced N structural and compositional changes of graphene when reaching 400 • C during initial heating (lowest common PT amongst all NGs).Moreover, these properties were likely not too dissimilar from what was measured for the NG400 material (Fig. 4a), although that material was kept at 400 • C for an additional hour.The NG400 material is thus taken here as a reference material to evaluate structural changes induced by higher PT.
Comparison of NG600 to NG400 properties indicated that with the increase in PT, a large fraction of the "just" formed N surface functional groups (at lower PT) were lost again.Also, with the apparent decrease in the relative content of pyrrolic N, the relative content of PNO and graphitic N gradually increased (Fig. S11), which also corroborates previous studies (Kasera et al., 2022;Leng et al., 2020;Oh et al., 2018).Overall, no significant differences were observed in I D /I G ratios between NG400 and NG600 materials, suggesting that loss and changes in N functional groups with PT increasing from 400 to 600 • C mostly occurred directly at graphene particle surfaces, with no significant effect on graphene structure.In terms of VC catalytic reactivity, NG600 led to an almost 10-fold higher RE 68h compared to NG400.Given that PNO groups have been pointed out as key sites for electron shuttling in previous studies on iron-CM composites for CE degradation (Ai et al., 2020), the higher catalytic reactivity of NG600 may be explained by the  C with varying urea dosages (i.e., urea: graphene ratios of 0.5, 1.0 and 2.0).relatively higher amount of PNO functional groups on NG600 surfaces.This could also explain the observed increase in VC RE for NG600 materials with increasing urea:graphene ratio (Fig. 3), as RE 68h correlated well with the relative amount of PNO groups (R 2 =0.99, p<0.05), while such a linear trend was less obvious for the other N-functional groups (Fig. S12).
Comparison of NG950 to NG400 showed that all N functional groups that formed during initial heating to 950 • C were lost eventually as no N species could be detected with elemental and XPS analyses.However, the I D /I G ratio of NG950 was considerably higher compared to NG400, suggesting higher defect density.This could be explained by N-moieties entering the graphene structure when heating above 600 • C, creating defects.Concomitantly however, with increasing PT, more and more N functional groups were also lost from the graphene structure, leaving vacancies i.e., defects, behind.It is well accepted that defects in the graphene structure drastically enhance electron transfer rates and act as active sites for H* storage (Singla and Jaggi, 2021;Wan et al., 2021;Yuan et al., 2022b;Zhao et al., 2022).Thus, this could explain the more than 15.7-fold higher RE 68h observed for NG950 materials compared to NG400.Also, given all NG950 materials had similar defect densities, independent of urea dosage, it may not be surprising that they all exhibited similar VC catalytic activities.
The relevance of PNO groups and structural defects for the VC catalytic reactivity of NG600 and NG950 materials, respectively was further indicated by PCA analysis where RE 68h values were set as observation and NG physicochemical properties as indicators (Fig. S13).The first two principal components (PC), PC1 and PC2, accounted for 91.6% of the variation in all indicators, and it showed that the reactivity of NG600 materials were more related to PNO, while reactivity of NG950 and also NG1100 materials were more correlated with defects.The above observations further suggest that both PNO (%) and defects (I D /I G ) are two critical properties for NG reactivity (RE 68h ,%).The factor loadings of PC1 further showed that the loadings of multiple factors in PC1 are in close proximity, i.e., PC1 may be affected by the collective effect of these factors.Thus, a multiple linear regression was fitted between RE 68h and the two variables PNO and defects, and the following correlation was obtained: This yielded an excellent fit giving further support to the key role of PNO surface groups and/or structural defects on NG materials for catalysis of VC reduction by nZVI.Notably, when the non-urea amended samples, G and G950, were included in this regression analysis, the goodness of fit (R 2 ) decreased significantly to 0.63, which emphasizes the importance of N-doping.

Dechlorination mechanism
The dehalogenation of CS by nZVI is argued to occur via direct electron transfer from Fe 0 to CS and/or involving atomic hydrogen (abbreviated as H*), that forms during Fe 0 anaerobic oxidation (Eq.( 6)).
The presence of CM can stimulate CS dehalogenation either by shuttling electrons from nZVI to CS adsorbed to the CM (Dorner et al., 2022;Gao et al., 2015;Kopinke et al., 2016;Wang et al., 2019), or by acting as a storage for H* spilled over from nZVI surface (Li et al., 2023).There are potentially two rate limiting steps here: electron and/or H*-transfer between nZVI and NG and/or between NG and VC.Earlier, it was shown that an increase in NG loading leads to a linear increase in VC reduction rate (Fig. 1c).In addition, separately performed experiments with varying nZVI loadings (from 0.4 g/L to 3.0 g/L) showed no effect on VC reduction extent and rate (Fig. S14).These findings suggest that the amount of active sites for VC reduction on NG are limited, while interactions (i.e., electron-and/or H*-transfer) between nZVI and NG do not seem to limit the reaction.In terms of the importance of electronand/or H*-transfer for VC reduction, one would assume hydrogenolysis (via reaction with H*) to play the major role in VC reduction, because ethylene seemed to be the only product that formed in VC reduction experiments with nZVI and NG materials tested here.A way to test this hypothesis is by adding a H* scavenger to the reaction with VC, nZVI and NG (Deng et al., 2021;Lan et al., 2016;Mao et al., 2019;Mezyk et al., 2004).This was done here for one of the experimental conditions (VC, nZVI and NG950-2.0) by addition of tert-butanol (TBA).The results showed that with the addition of 50 mM and 200 mM TBA, rate constants for ethylene formation decreased by 57 and 64%, respectively (Fig. S15).Similarly, RE 68h values were also lower, at 89.5% (50 mM TBA) and 71.1% (200 mM TBA) compared to 97.7% with no added TBA.These results support the hypothesis that H* is involved in VC reduction, but it is also surprising that the reaction could not be fully inhibited given the large excess of added TBA.The role of H* was further tested by examining the amount of hydrogen (H 2 ) produced in the different systems (Text S3).
H 2 production was around 75 µmol after 8 days in reactions with only nZVI and with nZVI and NG400, while in reactions between nZVI and NG600, NG800, NG950 and NG1100, H 2 formation was about 13.5% lower, at ~65 µmol (Fig. 6a).These H 2 yields suggest nZVI corrosion of approximately 4.2 and 3.6 mg Fe 0 (i.e., 42 and 36% of added nZVI), respectively.H 2 production further seemed to depend on NG loadings, as indicated by the gradual decrease in produced H 2 from  75 to 70 and 65 µmol with the addition of 0, 2 and 4 g/L NG950 (Fig. 6b).Notably, no VC was added to these reactions.
Overall, these results suggest that in the presence of NG materials that exhibited high RE 192h (Fig. 2a), H 2 formation after 192 h was generally lower.A lower H 2 formation could signify less overall H 2 production, for example by NG blocking sites for anaerobic oxidation on nZVI surfaces, i.e., creating a physical barrier.However, given no significant decrease in H 2 formation was observed in the reaction between nZVI and NG400, this seemed less likely.A lower H 2 formation could also signify enhanced H 2 sorption by NG materials and/or enhanced H* storage by NG materials, thereby inhibiting H* recombination.The sorption of NG materials towards H 2 was also tested, but no significant differences were observed between different NGs (Fig. S16), indicating that sorption cannot explain the reduced H 2 formation in nZVI reactions with NGs produced at PT ≥ 600 • C (Fig. 6a).It is experimentally challenging to assess the capacity for NG materials to store H*, i.e., inhibit H* recombination.However, some indications can be gained from theoretical studies.These have shown that the incorporation of heteroatoms in the graphene structure, e.g., N, can greatly enhance its H* storage capacity (Lee et al., 2013;Ma et al., 2014).Similarly, defects in the graphene structure (i.e., Stone-Wales, single vacancy, double vacancy and edge) have also been shown to enhance H* storage (Singla and Jaggi, 2021).Given that NG materials produced at PT ≥ 600 • C exhibit N-substitution and/or defects in the graphene structure as showcased by XPS and Raman analyses, this could explain their lower H 2 formation in reactions with nZVI (Fig. 6a).For NG400, the PT may have been too low to create N substitutions within the graphene lattice, and the surface N functional groups may not have been sufficient to increase H 2 storage capacity.
Given all observations, the following mechanisms of N-amended graphene catalysis for the reduction of VC by nZVI were drawn (Fig. 7).In this study, the co-pyrolysis of graphene with urea led to the creation of active sites for VC reduction by nZVI.The data here suggest that PNO groups and defect sites are key active sites on NG to catalyze VC by nZVI but their relative importance depends on the PT used to produce NG.It is further argued that NG accelerates the reduction of VC by storage of H* produced and spilled over from attached nZVI surfaces, thereby limiting H* recombination.It cannot be excluded that VC reduction also occurred via direct electron transfer from nZVI via NG, given that some VC was still reduced despite excess addition of H* scavenger TBA.Indeed, separate experiments performed with nZVI, NG950 and TCE showed an initial, fast evolution of acetylene (Fig. S17), which is presumably observed when TCE is reduced via direct electron transfer (Ai et al., 2020;Yang et al., 2019).Notably, no TCE reduction is seen in  reactions with nZVI only (i.e., no added NG950) over 3 days, verifying the low reactivity of nZVI with TCE observed in previous studies (He et al., 2018;Schöftner et al., 2015) and the high catalytic activity of NG towards VC but also towards TCE degradation.

Conclusion
This study shows that N-amended graphene can catalyze almost full reduction of VC by nZVI in 3 days, while no reactivity was seen in system with only nZVI.The PT during NG preparation is crucial for its catalytic activity, with higher performance observed at higher PTs of 600 to 950 • C. At these PTs, PNO groups and/or structural defects are generated on G during N doping, and their abundance correlate best with the observed VC reduction rate and extent, corroborated by PCA and multiple linear regression analyses.Thus, it is suggested here that PNO groups are key active sites for VC reduction on NG when produced at lower PT (<800 • C), while N-induced structural defects catalyze VC reduction on NGs at PT ≥ 800 • C, where these sites become more abundant and N functional groups disappear.The storage of hydrogen and the transfer of electrons may be the role played by these active sites.
Overall, this study demonstrates that the co-application of a N-doped carbonaceous material with nZVI could potentially be an effective strategy to degrade VC pollution in soils and groundwater.Notably, many aspects in regards to remediation strategy (e.g., via injection, soil mixing, installing of reactive wall), transport and retention behavior, and reactivity and longevity in groundwater matrices would need to be assessed next.Furthermore, it is envisaged a more sustainable carbon catalyst could be developed, by use of biographene or biochar materials, and taking from this study learnings of N-doping.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 1 .
Fig. 1.Kinetics of VC removal (filled symbols) and product formation (empty symbols) in reactions with (a) nZVI only and nZVI with untreated and 950 • C pyrolysed graphene (G and G950), and (b) nZVI with varying NG950 loadings (1-6 g/L).VC control (VC + water only) shows insignificant VC removal by volatilization and/or sorption to headspace vials.Error bars represent standard deviations of duplicates.Line segments added as a guide to the eye.The dashed lines in (b) represent pseudo first order kinetic fits to product data (Eq.(2)) with k obs and R 2 values given in TableS1.Experimental conditions: 2 g/L nZVI, 44.8 µM VC.(c) Linear relationship between rate constant, k obs (derived in b), and NG950 loading.

Fig. 2 .
Fig. 2. (a) VC reduction extents after 68 and 192 h in reactions with nZVI and NGs produced at PTs ranging from 400 to 1100 • C. (b) Kinetics of product formation in VC reactions with nZVI and NGs produced at 600, 800 and 950 • C. The dashed lines represent pseudo first-order kinetic fits to the data (values in TableS1).Experimental conditions: 2 g/L nZVI, 4 g/L NG, 44.8 µM VC.The error bars represent the standard deviation of duplicate data.

Fig. 3 .
Fig. 3. VC reduction extents in nZVI reactions with NG600 and NG950 materials produced with varying urea: graphene mass ratios from 0.25 to 2. Experimental conditions: 2 g/L nZVI, 4 g/L NG, 44.8 µM VC.The error bars represent the standard deviation of duplicates.

Fig. 4 .
Fig. 4. Changes in NG surface composition as determined by XPS analyses (in atomic%).(a) Distribution of C species in NG materials produced at varying PT, and abundance of N surface species in NG materials produced at b) varying PT and c) PT of 600 • C with varying urea dosages (i.e., urea: graphene ratios of 0.5, 1.0 and 2.0).

Fig. 5 .
Fig. 5. Comparisons of trends in I D /I G ratio and VC RE 68h for NG materials produced at (a) varying PT and (b) with different urea: graphene ratios for NG600 and NG950 materials.
Q.Ouyang et al.

Fig. 6 .Fig. 7 .
Fig. 6.H 2 formation in nZVI (2 g/L) reactions with a) NG materials produced at different PTs after 192 h (control is nZVI only system), and b) NG950 at loadings of 0, 2 and 4 g/L after 24, 68 and 192 h.No VC was added to these reactors.Error bars represent the standard deviation of duplicates.
Q.Ouyang et al.

Table 1
Previous laboratory studies on VC chemical reduction.
a Reported pseudo-first-order rate constants were normalised to applied reactant loading (as g/L Fe).b Percentage of the initial VC concentration that has been degraded.cProducedby reduction of an Fe II solution with NaBH 4 .dManufacturedby reduction of goethite and hematite particles with H 2 at 200− 600 • C.Q.Ouyang et al.