Isolation of dissolved organic matter from aqueous solution by precipitation with FeCl3: mechanisms and significance in environmental perspectives

Ferric ions can bind strongly with dissolved organic matter (DOM), including humic acids (HA), fulvic acids (FA), and protein-like substances, whereas isolation of Fe-DOM precipitates (Fe-DOMP) and their biochemical characteristics remain unclear. In this work FeCl3 was used to isolate DOM components from various sources, including river, lake, soil, cow dung, and standard tryptophan and tyrosine, through precipitation at pH 7.5–8.5. The Fe-DOMP contribute to total DOM by approximately 38.6–93.8% of FA, 76.2% of HA and 25.0–30.4% of tryptophan and tyrosine, whilst fluorescence spectra allowed to monitor/discriminate the various DOM fractions in the samples. The relative intensity of the main infrared peaks such as 3406‒3383 cm−1 (aromatic OH), 1689‒1635 cm−1 (‒COOH), 1523–1504 cm−1 (amide) and 1176–1033 cm−1 (‒S=O) show either to decline or disappear in Fe‒DOMP. These results suggest the occurrence of Fe bonds with various functional groups of DOM, indicating the formation of π–d electron bonding systems of different strengths in Fe‒DOMP. The novel method used for isolation of Fe-DOMP shows promising in opening a new frontier both at laboratory and industrial purposes. Furthermore, results obtained may provide a better understanding of metal–organic complexes involved in the regulation of the long-term stabilization/sequestration of DOM in soils and waters.

In particular, Reiller et al. showed that HA of high molecular weight can preferentially adsorb to hematite 34 , whilst Newcomb et al. measured directly the binding between organic ligands with soil minerals 40 . However, the precipitation mechanisms of soil HS by Fe 3+ by forming a π-d electron bonding system between the HS functional groups and d-orbitals of Fe remain unclear. Secondly, Fe 3+ shows to sink HS in aquatic environments which typically occurs by trapping terrestrially derived DOM by Fe 3+20 , flocculation and coprecipitation [19][20][21][22]41 . Similarly, Fe 3+ can form complexes with protein of extracellular polymeric substances (EPS) and siderophores of prokaryotes [23][24][25][26]42 , and may act as a key factor for their subsequent sinking in aquatic environments. However, the mechanisms by which functional groups of various DOM components, including amino acids (tryptophan and tyrosine), can effectively form complexes with d-orbitals of Fe 3+ in terms of bonding formation mechanism is still unclear to date. Thirdly, FA is commonly isolated from aqueous solutions by the well-accepted method that uses organic-based XAD-8 resins [43][44][45] . However, a problem with XAD-8 resins is the possible incorporation of organic matter from the adsorbent which demands a careful resin cleanup 43,46 , the loss of a substantial portion of FA as shown by the 23-58% of dissolved organic carbon (DOC) recovered by precipitation and adsorption onto XAD-8 resin in clear water and 50-90% in colored surface waters 43,47 , and recovering of only the hydrophobic fraction of FA 43,48 . Furthermore, PPL (styrene divinyl benzene polymer) is also reported to be able to extract an average 43% of deep sea DOC and up to 65% of freshwater DOC 49 . Therefore, a novel extraction technique based on the use of Fe 3+ ions appears very relevant to isolate diverse forms of FA from aqueous solutions as Fe-FA precipitates (Fe-FA P ). In particular, FA isolated in a pure form as Fe-FA P may provide different individual fractions of FA free of impurities, which cannot be isolated using XAD-8 resins. In this regards, Fe-FA P are obtained based on the principle that supersaturation governs the appearance of precipitates 50,51 and alkaline conditions may lead to form large aggregates that can settle under the influence of gravity 52 . Therefore, a novel method based on the use of FeCl 3 is developed in this work aiming to: (i) isolate FA from diverse environmental aqueous media, including rivers, lakes, soils and cow dungs, as Fe-FA P ; (ii) characterize them by FTIR and fluorescence excitation-emission (EEM) spectroscopy coupled with parallel factor (PARA-FAC) modelling to assess their molecular nature and adsorption/complexation behavior, and propose tentative mechanisms for their formation; (iii) isolate HA, FA and protein-like substances from HA in alkaline extracts of soil and characterize the corresponding Fe 3+ precipitates and remaining supernatants obtained by a series of successive precipitation steps by fluorescence EEM spectroscopy, and ascertain the precipitation of HA by FeCl 3 (Fe-HA P ); (iv) characterize the pH-8.5 re-dissolved Fe-HA P isolated from the extracted HA sample in order to ascertain the occurrence of coprecipitation by HA; (v) isolate and characterize low molecular weight DOM, i.e., tryptophan and tyrosine, from their aqueous solutions to determine their Fe-DOM P ; and (vi) propose the optimal conditions for the application of FeCl 3 to precipitate DOM fractions from various aqueous solutions and discuss their relevance at environmental and industrial scales.

Materials and methods
Samples. The DOM fractions were isolated from the Baigu river and Jingye lake waters, a forest soil, and original and irradiated cow dung samples. The 38.5-km-long Baigou River features clear and unpolluted waters originating from the foothills of the Taihang Mountains in the northwest of Laiyuan County, Hebei Province, China. The sampling site was located in the upper part of the river, approximately 150-m from the middle section of the North Juma River. The Jingye lake is situated in the Weijin Road campus of Tianjin University, Tianjin, China, and its DOM was comprehensively characterized in earlier studies that showed its origin was mostly from photosynthetically-derived primary productivity 6 . The forest soil was collected under the Panshan deciduous forest, Tianjin city of north China and details on its sampling site, vegetation covers and major physico-chemical characteristics were reported previously 2 .
The cow dung samples were collected from a farmhouse located in a village in the Jixian district of Tianjin city. Besides the original sample, a cow dung sample irradiated for 1 day under natural sunlight was used in order to ascertain how does the DOM in cow dung transform under sunlight conditions. Raw cow dung samples without any treatment were shown to be primarily composed of high amounts of tyrosine 9 , with an increasing humified fraction during its full-scale vermicomposting 8 . The reason to study also an irradiated sample is based on the fact that during cattle grazing on the grass-field in day-time, cow dung is subject to irradiation by natural sunlight and mixing up with soil and rainwater. Thus, the study of one-day sunlight-irradiated cow dung is expected to provide useful information on the possible alteration of organic matter, and especially its FA fraction, in these conditions. Thus, the cow dung samples were dried in an oven at 60 °C, ground into fine particles, and then stored at − 20 °C until further processing.
Furthermore, two standard compounds, i.e. tryptophan and tyrosine purchased from BBI Life Sciences (Shanghai) and Shanghai Macklin Biochemical Co., Ltd, respectively, which are commonly detected in surface waters, rainwaters, glaciers, clouds and aerosols 5,6,15,[53][54][55][56][57] , were used. www.nature.com/scientificreports/ Extraction of the liquid phase from forest soil and cow dung samples. The liquid phase was extracted from the forest soil using a number of subsequent steps (Fig. S1). In the first step, the ground and 0.2-mm-sieved soil sample was added with ultrapure water (18.2 MΩ·cm, Mill-Q, Millipore) at a soil/water ratio of 1:10, vortexed for 1 min in closed 500-mL brown bottles and then shaken for 6 h at 25 °C. The mixture was centrifuged for 20 min at 4000 rpm using a Thermo Fisher Scientific SORVALL ST 16 centrifuge to remove suspended solids. The supernatant solution was then filtered through a 0.45-µm membrane filter (GF/F type, Shanghai Xin Ya Purification Equipment Co. Ltd, China), whereas the remaining solid residue was extracted again with fresh ultrapure water for 1 h and the procedure described above applied again to obtain the supernatant solution that was mixed with the previous one and stored in a freezer at − 20 °C until further processing. This solution represents the soil water extract (W e ). The soil residue from water extraction was then subjected to alkaline extraction under N 2 with a 0.1 M NaOH solution at a soil residue/alkaline solution ratio of 1:10 by shaking for 3 h at 25 °C. The mixture was then centrifuged as described above and the supernatant solution filtered through a 0.45-µm membrane filter (polytetrafluoroethylene membrane, PTFE, Shanghai Xin Ya Purification Equipment Co. Ltd, China). The remaining solid residue was extracted again with a fresh alkaline solution for 3 h and the procedure described above applied again to obtain the supernatant solution that was mixed with the previous one and stored in a freezer at − 20 °C until further processing. This solution represents the soil alkaline extract (A e ).
The HA fraction of the forest soil was obtained by acidifying an aliquot of A e at pH 2 under N 2 at 4 °C. Many studies have shown that alkaline conditions (A e ) may alter the molecular compositions of soil HA in the presence of O 2 during its extraction from soil 45,58 , thus IHSS recommends to conduct the alkaline extraction in the presence of N 2 gas to reduce any alterations 45 . After 24 h, the precipitated HA was centrifuged and freeze-dried. Cow dung samples were processed using the same procedure applied to the soil and described above. In this case, the precipitated HA fraction was redissolved in NaOH, adjusted to pH 8.5 and stored in a freezer at − 20 °C until further processing.

Precipitation of DOM fractions by FeCl 3 .
The Fe-FA precipitates (Fe-FA P ) were obtained from the corresponding aqueous solutions by precipitation with FeCl 3 according to the procedure outlined in the flow diagram in Fig. S2. Solid ferric chloride hexahydrate, analytical grade, 99.0% purity (Chemart Chemical Technology Co., LTD., AR, Tianjin, China) has been used in this study First, the surface water samples from Baigou River and Jingye Lake were filtered through a 0.45-μm GF-F type filter and the filtrates were acidified at pH 2 with HCl, kept for 24 h in a refrigerator at 4 °C, and then centrifuged for 20 min at 4000 rpm to remove HA. The supernatants were adjusted to pH 7.5-8.5 in 500 mL brown glass bottles, and then analyzed by fluorescence EEM spectroscopy to measure the initial fluorescence intensity of FA. Then, the FeCl 3 solution (30 g L −1 ) was slowly added drop by drop to the supernatant solutions with simultaneous addition of NaOH to keep the pH at 7.5 ~ 8.5 and stirring gently using a magnetic stirrer. The pH values of 7.5-8.5 were chosen for three specific reasons: (i) Fe-DOM P can form large aggregates under alkaline conditions 52 ; (ii) the pH of the solution is decreasing constantly due to addition of Fe 3+ , thus it is needed to adjust the pH by NaOH continuously to maintain the chosen pH range; and (iii), this pH range is optimal for fluorescence measurements. The FeCl 3 addition was stopped upon the appearance of Fe-FA P , but stirring was maintained for a further 5-10 min until the pH reached a constant value. The mixtures were then centrifuged for 20 min at 4000 rpm to separate the solid Fe-FA P , whereas the remaining supernatants were then adjusted to pH 7.5-8.5 and their fluorescence EEM spectra measured to monitor the approximate extent of FA precipitation in the first phase by comparing the actual intensity of FA peak M to the corresponding initial ones (at 285-315/373-416 nm). Peak M of FA is denoted as marine humic-like peak, and typically appears at shorter excitation-emission wavelengths (~ 280-320/370-420 nm) in comparison with peak C (~ 320-410/420-520 nm), whilst peak A occurs at shorter excitation and relatively longer emission wavelengths, and largely differ for HA (~ 270-280/440-520 nm), FA (~ 230-270/370-440 nm) and protein like substances (~ 220-230/385-420 nm) of both allochthonous and autochthonous origin 1,2,7,59,60 . About 3.0 mL of FeCl 3 were needed for approximately 900 mL of supernatant to start the precipitation of Fe-FA P . The procedure was then repeated several times to precipitate subsequently Fe-FA P fractions until the intensity of peak M of FA in the remaining supernatants reached a minimum. The solid Fe-FA P fractions were then kept under inert N 2 in a freezer at − 20 °C until further processing.
The Fe-HA precipitates (Fe-HA P ) were obtained using the same isolation procedure used for Fe-FA P (Fig. S2). Briefly, a 60-mg aliquot of soil forest HA was re-dissolved in 600 mL of 0.1 M NaOH and then adjusted to pH 8.5. The FeCl 3 solution (30 g L −1 ) was slowly added drop by drop to a 400-mL aliquot of the HA solution, with simultaneous addition of NaOH to maintain the pH at 7.5 ~ 8.5, until the appearance of Fe-HA P . Approximately 1-mL of FeCl 3 solution was needed to complete the precipitation. The Fe-HA P fractions were isolated by centrifugation and the EEM fluorescence spectra of the corresponding remaining supernatants were measured to estimate the amounts of Fe-HA P fractions. Then, the same procedure described above (Fig. S2) was applied to obtain the Fe-FA P fractions from the remaining supernatants of HA. The dissolved organic carbon (DOC) content in the final remaining supernatants was measured to evaluate the amount of Fe-HA P and Fe-FA P fractions.
Similarly, a 40 mg aliquot of each standard tryptophan and tyrosine was dissolved in 400 mL aqueous solution at pH 8.5. The Fe-Tryptophan and Fe-Tyrosine precipitates (Fe-Tryptophan P and Fe-Tyrosine P , respectively) were obtained by adding the FeCl 3 solution to a 380-mL aliquot of each standard and using the same procedure described above for Fe-FA P and Fe-HA P (Fig. S2). Approximately 1 mL of FeCl 3 solution was necessary to complete the precipitation of both compounds. In both cases, five subsequent precipitation steps were performed in order to obtain a higher amount of precipitate. The EEM fluorescence spectra of the remaining supernatants obtained after each step were measured to estimate the amount of Fe-Tryptophan P and Fe-Tyrosine P fractions on the basis of the corresponding fluorescence intensities. The DOC content in the final remaining supernatants www.nature.com/scientificreports/ was measured to evaluate the amount of Fe-Tryptophan P and Fe-Tyrosine P . Noteworthy, the identical amounts of FeCl 3 solution required in the first and subsequent steps of its drop by drop addition to samples to isolate Fe-DOM P indicates that there is no excess amount of free Fe 3+ in the remaining supernatant solutions. However, the Fe 3+ concentration of the samples has not been measured in this study, which will focus for further study.
Analytical protocols. The elemental C, H, N and S composition was measured using an elemental analyzer (Elemental Vario E.L. III, Germany) by placing an aliquot of each dried, ground and homogenized sample into a clean, carbon-free, pre-combusted tin boat placed on an autosampler rack assembly loaded onto the elemental analyzer. Sulfanilamide was used as a standard after every ten measurements. The O % was calculated by difference as O% = 100 − (C + H + N + S)%. The P% has not been measured in this study due to its low content 17,47 . The dissolved organic carbon (DOC) content was measured in triplicate for each sample using a combustion total organic carbon (TOC) auto-sampler analyzer (OI Analytical Aurora, Model 1030 W + 1088, USA). In this study, the blank value was below 0.15 mg L −1 (precision 15%), which was < 8.7% of the DOC concentration measured in all samples studied except two low DOC samples (11.6-15.2%). Fourier transform infrared (FTIR) spectra were recorded using an IRAffinity-1S spectrometer (Shimadzu, Japan) that included a high-energy ceramic light source, a temperature-controlled, high-sensitivity (deuterated l-alanine triglycine sulfate: DLATGS) detector and a high-throughput optical element, with optimization of the electrical and optical systems. The IRAffinity-1S instrument achieves the highest signal/noise (SN) ratio in its class. A mixture of 2 mg of each dehydrated and freeze-dried sample was mixed homogeneously with 200 mg of dried KBr and pelletized by pressing under reduced pressure. Then the FTIR spectra of the pellets were recorded over the range 4000-400 cm −1 by averaging 30 scans at a 4 cm −1 resolution.
Fluorescence EEM spectra were measured by a fluorescence spectrophotometer (F-7000, Hitachi, Japan) using a procedure described previously 2,6 . Ultrapure (18.2 MΩ.cm) MQ water was used as the blank and measured every ten samples to check the performance of the instrument and ensure data quality. A 4-μg L −1 quinine sulfate (QS) solution in 0.01 mol L −1 H 2 SO 4 was used to achieve fluorescence normalization and the fluorescence intensity of each sample was calibrated using the intensity of the QS (1 μg L −1 = 1 QS unit, QSU) peak at Ex/ Em = 350/450 nm 2 . To avoid inner-filter effects and fluorescence quenching, each sample solution was diluted on the basis of the DOC concentration measured initially before EEM measurements 61 . Furthermore, the fluorescence intensity of each peak was rechecked and corrected using the common absorbance-based approach 62 . The Rayleigh and Raman peaks and the ultrapure water blank spectrum were subtracted from each experimental EEM spectrum using a home-made Excel program 2,6,63 . Preprocessed EEM data were then processed by the parallel factor (PARAFAC) model using the N-way toolbox for MATLAB 64 as described elsewhere 3 . To avoid mixing of fluorescent components of different samples that could produce artifacts 63,65 , PARAFAC analysis was performed individually on each sample and on their remaining supernatant solution. Finally, non-negative constraints were applied to the PARAFAC model. The detailed procedure used for the PARAFAC analysis of EEM spectra was described previously 6,63,65 . Ethical approval. This manuscript does not involve to the researches regarding human participants, and also animals.

Results and discussion
Fe-FA P from river, lake, soil and cow dung samples. The Fe-FA P obtained from the various environmental samples, i.e., Baigu river, Jingye lake, water-extracted (W e ) and alkaline-extracted (A e ) of the forest soil, and water extracts of original and irradiated cow dung, amounted respectively to 68.1%, 38.6%, 72.9%, 82.1%, 92.2% and 93.8% of FA (Table S1). The FA was monitored after each extraction step by measuring the fluorescence intensity of peak M of the remaining supernatants (Table S2). Peak M maxima varied in the ranges: 295-310/394-416 nm and 300-310/392-412 nm, respectively for soil W e and A e ; 315/403-410 nm and 315/400-411 nm, respectively for the first and the second precipitation from river water; 285-305/369-396 nm for lake water; and 315/402-415 nm and 312-315/402-415 nm, respectively for raw and irradiated cow dung (Table S2). Original EEM spectra (left-side) and their fluorescent components (right-side) identified by the EEM-PARA-FAC model in all individual aqueous samples and their RS after each repeated precipitation of Fe-FA P are shown in Fig. 1 and described in Table S3, and are very similar to those described in our earlier studies 2,6,7,9,57,63,65 . Furthermore, the loading figures corresponding to the fluorescent components identified by the EEM-PARAFAC model are presented in Figs. S3 and S4.
The FA precipitation as Fe-FA P can be associated with two important biogeochemical phenomena: (i) the major fractions of terrestrial and autochthonous FA in riverine systems can form Fe-FA P in coastal seawaters or estuarine environments commonly characterized by alkaline conditions 66 , which may sink a major fraction of FA in seawater environments as co-precipitates and/or flocculates 19,20,67,68 ; and (ii) FA in soil can form Fe-FA P in a complex state, Thus, FA in different environments such as streams, rivers, ponds, lakes, reservoirs, oceans, estuaries, ground waters, pore waters, rainwaters, clouds, snow, glaciers, soils, sediments and atmospheric aerosols can be easily isolated by precipitation with Fe 3+ ions, which would allow their molecular-level analysis that can provide relevant information on their biogeochemical functions and processes in the corresponding environment/ecosystem. Noteworthy, Fe-FA P isolated from various environmental samples represent the fraction of bulk DOM that associates with Fe 3+ in the corresponding samples and thus cannot be directly compared with FA in pure forms.
Soil Fe-HA P . The PARAFAC EEM images of soil HA and their corresponding peak maxima featured four fluorescent components, including HA-1, HA-2, FA and protein-like substances (Fig. 2, Table S3), which sug-  www.nature.com/scientificreports/ sample were precipitated and separated in the first step (Fig. 2), which suggested that two distinct types of HA would exist in the soil matrix. In the first extraction, 76.2% of total DOM was isolated as HA, whereas 41.0% of total DOM was isolated as FA by the following six extractions (Table S1). These results suggested that the main responsible for the formation of Fe-HA P would be complexation by Fe 3+ followed by co-precipitation. To confirm the occurrence of the co-precipitation of HA with protein-like substances, an aliquot of the Fe-HA P sample was redissolved at pH 12.0 and analyzed by EEM fluorescence spectroscopy. Two fluorescent components from the HA-2-like and protein-like substances appear (Fig. 3), which suggest that co-precipitation of these two fractions www.nature.com/scientificreports/ occurred. Simultaneously, HA-1 would form strong complexes in Fe-HA P that were not dissolved at pH 12.0, the evidence of which is provided from FTIR spectra that show the disappearance in Fe-HA P of all IR peaks present in the original HA spectra (Fig. 4e). Similar co-precipitation processes were reported previously 20,35,36,38 . Only two fluorescent components, i.e. FA-like and protein-like substances, were identified by EEM-PARAFAC analysis in all seven remaining supernatants after each subsequent precipitation step of Fe-FA P . In particular, FA in the first four remaining supernatants showed EEM peaks maxima (305-320/440-449 nm for peak M and 250-270/440-449 nm for peak A) that were relatively longer than the three subsequent supernatants (300-305/431-440 and 240-245/431-440 nm, respectively for peak M and peak A) (Fig. 2). These results suggested that the relatively longer excitation-emission wavelength peaks of FA would imply its initially high complexation capacity for Fe 3+ involving high functionality molecules, whereas the shorter excitation-emission wavelength peaks would be due to molecules featuring a lower functionality. This interpretation is confirmed by earlier studies showing a decrease of the binding capacity of photodegraded FA with respect to original FA 69,70 , and by the well-known blue-shift to shorter excitation-emission wavelengths with respect to the original ones usually caused by photodegradation 71,72 . This interpretation is further supported by the first extracted HA-1 and HA-2 that show excitation-emission peaks longer than the subsequent seven remaining supernatants, which would suggest that, the isolation of Fe-HA P and Fe-FA P from the various environmental samples, no artifact effects would occur on the EEM measurements by the DOM remained in the samples. Differently, peak A and peak T UV of protein-like substances showed shorter wavelength maxima that varied slightly from the first to the seventh remaining supernatant, which might be due to their pH variation between 8.41 and 8.56, as it has been shown for protein-like peaks of extracellular polymeric substances 73 . In conclusion, the preferential formation of stable Fe-HA P and then Fe-FA P would confirm the stable sequestration of organic carbon in organic-mineral complexes in the soil matrix 10,13,14,36,74,75 . Fe-Tryptophan P and Fe-Tyrosine P . Approximately 25.0% and 30.4% of total DOC was precipitated by FeCl 3 as Fe-Tryptophan p and Fe-Tyrosine p from their corresponding water solutions. Both before and after repeated precipitations the EEM spectra of Fe-Tryptophan p and Fe-Tyrosine p featured the two expected proteinlike substances peaks T and T UV (Table S3). Tryptophan and tyrosine are composed of known molecular structures rich in amino carboxylic and hydroxyl functional groups that are able to complex Fe 3+ ions. These results suggest that low molecular weight organic compounds such as tryptophan and tyrosine could form strong complexes with Fe 3+ via COOH and OH functional groups of amino acids 76,77 , as it has been shown to occur in surface waters for proteins of extracellular polymeric substances and siderophore complexes of prokaryotes [23][24][25][26]42 . The occurrence of such complexation is also evidenced by the FTIR spectra discussed in section "FTIR spectra". (Table 1) shows that the C% in all Fe-FA P samples is quite low. In particular, the % of Fe-FA P from river lower than that from lake and soil W e , which are all lower than those from soil A e and original and irradiated cow dung samples that are the highest of all Fe-FA P samples. The low C%, especially in some samples, might arise from the low molecular weight (LMW) FA of aqueous samples. The relatively longer Ex/Em of peak M suggests that high molecular weight FA would precipitate initially as Fe-FA P (Fig. 4), whereas the relatively LMW FA would be involved in the last precipitation steps. Obviously, the O% in all Fe-FA P samples is relatively high and, as a consequence, the C/O ratios are generally quite low, with the lowest values for Fe-FA P from river, lake and soil W e , and values ranging from 0.6 to 1.5 for the remaining samples.

Elemental composition. The elemental composition of Fe-DOM P fractions
Also the H% in Fe-FA P from the river is relatively lower than those in the Fe-FA P from lake and soil W e , which are, however, higher than those in samples from soil A e and original and irradiated cow dung samples. Such differences may be due to the different molecular composition of FA (Baluha et al., Stenson et al., Stubbins et al.) isolated using FeCl 3 from various sources in each extraction process. Differently from the C%, the H% in Fe-HA P , Fe-Tryptophan P and Fe-Tyrosine P are higher than those in the corresponding HA, Tryptophan and Tyrosine. Also the atomic C/H ratios are very low for Fe-FA P from river, lake and soil W e , and lower than that from soil A e . The C/H ratios of Fe-FA P from cow dung and soil HA are higher than those of Fe-FA P from irradiated cow dung and Fe-HA P , whereas the C/H ratios of Fe-Tryptophan P and Fe-Tyrosine P are much lower than those of the corresponding Tryptophan and Tyrosine. www.nature.com/scientificreports/ The N% features the lowest values in Fe-FA P from river, lake and soil W e and the highest ones in the original HA and Fe-FA P from cow dung, whereas in Fe-FA P from irradiated cow dung sample, Fe-HA P , Fe-Tryptophan P and Fe-Tyrosine P the N% is lower than that in the corresponding unreacted sample. The high N% in the original cow dung sample may be attributed to the high content of Tyrosine 65 that declines upon sunlight exposure of cow dung. The C/N ratio is relatively low in Fe-FA P from soil A e , and much higher in irradiated cow dung, Fe-HA P , Fe-Tryptophan P and Fe-Tyrosine P than in the corresponding not-irradiated cow dung, soil HA, Tryptophan and Tyrosine. Finally, the S% results ≤ 1.0% in most Fe-FA P samples and only in the Fe 3+ precipitate from irradiated cow dung sample reaches the maximum value of 1.0%. As expected from the corresponding S%, the C/S ratios of Fe-FA P from cow dung, soil A e , and river and lake waters are very high, and those of the remaining samples quite low.
Thus, the Fe-FA P fractions isolated in each step feature a different elemental composition, which is supported by the blue-shift of FA peaks in the RS after each precipitation step (Fig. 2), and confirms previous FTICR-MS studies [78][79][80] . Further studies on the elemental composition of Fe-FA P isolated in each step will be necessary to achieve a better understanding of FA components and their biogeochemical transformation in specific ecosystems. FTIR spectra. The most relevant infrared absorption features and their differences (Fig. 4)   www.nature.com/scientificreports/ and lake waters, soil W e and A e , and both cow dung samples may be ascribed to aromatic C-H stretching in aromatic ring structures 80 . (ii) A band of strong/medium relative intensity at 3400-3450 cm −1 appears in the remaining supernatants of Fe precipitates from river and lake waters, soil W e and A e , original and irradiated cow dung samples and original soil HA, the relative intensity of which decreases in the corresponding Fe-FA P and Fe-HA P samples. These results would suggest that aromatic OH in FA could effectively donate electrons to d-orbitals of Fe 3+ , thereby forming a strong π-d electron bonding system in Fe-O-FA, which, in turn, causes IR absorption to decline or completely disappear. Furthermore, the relatively higher intensity of this band in Fe-FA P from soil A e would suggest that soil FA is richer in electron donating functional groups compared to FA from lake and river waters and cow dung samples 17,65,[81][82][83] , and that many functional groups in soil FA would be unable to donate electrons to the d-orbital of Fe 3+ possibly due to intermolecular interactions among them [84][85][86] . (iii) The relative intensity of the peaks in the range 1685-1662 cm −1 , assigned to aromatic C=C stretching and/or COOH groups, in the remaining supernatants from river and lake waters, soil W e and A e and original and irradiated cow dung samples feature a is slightly lower than that of the corresponding Fe-FA P . These results would suggest that these groups could donate electrons to d-orbitals of Fe 3+ forming a strong π-d electron bonding system Fe-OOC-FA. (iv) The peaks at 1523-1500 nm −1 , assigned to the amide II band, are relatively more intense in Fe-FA P than in the corresponding remaining supernatants from river waters, soil W e , and original and irradiated cow dung samples, whereas the opposite occurs for lake and soil A e . These results would suggest that the donation of non-bonding electrons from amide II N to d-orbitals of Fe would form either a strong or a weak bonding system in Fe ← :N-FA P . (v) A weak peak around 1400 cm −1 , assigned to C=N stretching of amide I, appears in Fe-FA P from soil W e and original and irradiated cow dung samples, which would imply, similarly to amide II, the formation of a weak π-d electron bonding system between non-bonding electrons of amide I N and d-orbitals of Fe 82,87,88 . (vi) A weak peak around 1150 cm −1 , assigned to S=O and C-O-S stretching of S-containing functional groups, is present in the remaining supernatants from river and lake waters, soil A e and original and irradiated cow dung samples, but it disappears in the corresponding Fe-FA P . These results would suggest that S-containing functional groups can donate electrons to d-orbitals of Fe 3+ , forming a π-d electron bonding system in Fe-S-FA P . (vii) A weak band at 1100-1000 cm -1 , attributed to C-O stretching of polysaccharides and polysaccharidelike components, is present in Fe-FA P from soil W e and A e , irradiated cow dung samples, original HA and the first Fe-HA P , but not in the other samples, which would suggest a partial involvement of also these structures in Fe 3+ complexation.
In general, all FTIR peaks appearing in the spectra of original soil HA precipitated at pH 2 and the first Fe-HA P sample almost disappear in the subsequent Fe-HA P samples, which would imply that all HA functional groups have formed complexes with Fe 3+ . Also for tryptophan and tyrosine all FTIR peaks are strongly reduced in intensity or disappear in Fe-Tryptophan P and Fe-Tyrosine P , which confirms that the functional groups of both compounds are capable of forming strong complexes with Fe 3+ .
In conclusion, FTIR results discussed above suggest that either strong or weak π-d electron bonding systems can be formed in Fe-FA P , Fe-HA P , and Fe-Tryptophan P and Fe-Tyrosine P by electron donation from the functional groups of the corresponding original samples.
Mechanisms for Fe-DOM P formation and isolation. The conceptual model of the processes of formation and isolation of Fe-DOM P (Fig. 5) illustrates how Fe-DOM P complexes are constantly formed from DOM components and Fe 3+ ions in aqueous, slightly alkaline conditions (pH 7.5-8.5). Specifically, DOM functional groups, such as carboxylate (-COO − ), hydroxylate (-O -) and thiol (-SH) groups, rapidly deprotonate in these conditions, thus determining the formation of Fe-DOM P involving strong or weak π-d electron bonding systems by donation of electrons from deprotonated electron-rich functional groups of DOM to the unpaired d-orbitals of Fe 3+31, 89 . Moreover, the π-d electron bonding system in Fe-DOM would be stabilized by the lower Table 1. Elemental composition (C, O, N, H, S) (moisture and ash free) and atomic ratios (C/N, C/S, C/H and C/O) of Fe-fulvic acid precipitates (Fe-FA P ) and Fe-humic acid precipitates (Fe-HA P ) from river, lake, soil, original and irradiated cow dung samples, Fe-tryptophan precipitates (Fe-TR P ) and Fe-tyrosine precipitates (Fe-TY P ). www.nature.com/scientificreports/ energy of the high spin state due to its greater electron-nucleus attraction 90 , which would yield a strong binding of Fe 3+ to HS and DOM, so that Fe 3+ results widely distributed in the larger molecular size fractions 91 . The π-d electron bonding system occurring between Fe 3+ and DOM 92 can be outlined in the following two reactions: Finally, the different electron donation capacity/efficiency of the various DOM functional groups to unpaired d-orbitals of Fe 3+ would be the determining factor for the formation of either strong or weak complexation in the various Fe-DOM P samples. Notably, the anionic forms (-O − or -COO − ) of HS functionalities (phenolic OH and carboxylic acid-COOH, respectively) would be favored in donating π-electrons to Fe 3+ d-orbitals due the high availability of Fe 3+ before reaching the saturation state during its addition in the precipitation process. The achieved Fe-DOM P might be comparable with both Fe-(oxy)hydroxides (FeOOH) in either dissolved or colloidal state in water 31,52,[93][94][95] , and Fe-(oxy)hydroxide minerals and organo-mineral complexes in soils and sediments 10,12,13,[96][97][98][99][100][101] . In particular, Fe-DOM P would include, besides organo-Fe(OH) 3 complexes, also organo-Fe(III) oxyhydroxides and organo-Fe(III) minerals 102 . Moreover, Fe-DOM P might undergo microbial reduction either biologically or by conversion to organo-Fe(II) oxyhydroxides under diverse environmental conditions [103][104][105][106][107] . Finally, Fe(II) oxyhydroxide (FeOOH) forms large aggregates in alkaline conditions at pH > 7 using Fe(II)-salts in the absence of DOM/HS 52 , which is very similar to the formation of Fe-DOM P using Fe(III)salts in the presence of DOM/HS in this work. Therefore, Fe-DOM P /Fe(III) oxyhydroxides and Fe(II) oxyhydroxides would be formed depending on the occurrence of Fe(III) or Fe(II) in the diverse environments, including soils, sediments and surface waters 107-113 . Environmental and practical implications. The precipitation of Fe-DOM P in slightly alkaline conditions would suggest that DOM fractions can exist in a strong binding state with Fe 3+ in slightly alkaline surface waters, soil and other environmental systems such as rainwater, groundwater, glaciers, aerosols, etc. Such conditions are crucial for the fate of DOM and Fe-DOM P , i.e., their solubility, speciation, reactivity, degradation, redox www.nature.com/scientificreports/ reactions, bioavailability and transport 19,21,29,31,114 . Differently, DOM in surface waters having a neutral to acidic pH would show less probability to be complexed by metal ions such as Fe 3+ . Noteworthy, the observation that approximately 3.0 mL of FeCl 3 (30 g L −1 ) solution is needed to ~ 900 mL of a water sample to start the precipitation of Fe-FA P implies that up to ~ 2.99 mL Fe-DOM complexes remain in a soluble state in surface waters, and that dissolved iron is removed only at concentrations greater than its apparent solubility 115 . This situation is very similar to that of Fe-(oxy)hydroxides in freshwaters and seawaters either in the dissolved or colloidal state 31,52,[93][94][95]116,117 . Moreover, Fe 3+ in oxygenated seawater at pH ~ 8 is thermodynamically stable in the form of Fe(III) hydroxide complexes, which have a high possibility to be scavenged by sinking particulate matter and are in equilibrium with Fe(III) oxyhydroxide particulates characterized by low solubility 118 . Furthermore, Fe 3+ -DOM complexes are known to be the primary responsible of the occurrence of photo-Fenton reactions in DOM degradation under sunlit surface waters 31,95,114 , where Fe oxyhydroxide are involved in reactions involving O 2 ·− and ligand-to-metal charge transfer 95 . Furthermore, various siderophores enable the acquisition of Fe from its minerals (e.g. oxides, hydroxides, goethite) and organic complexes under various environmental conditions [119][120][121][122][123] . Siderophores are relatively low molecular weight organic substances that can bind up to 99% of Fe dissolved in seawater 124 and may be compared with Fe-Tryptophan P and Fe-Tyrosine P . In particular, siderophore-bound Fe is utilized by marine bacteria and eukaryotic phytoplankton as a sink in seawaters [23][24][25][26]42,[125][126][127][128] . Furthermore, the primary factor that affects the up to 99% of Fe 3+ to remain in a complex state with DOM in seawaters 32,33 can be ascribed to their common slightly alkaline conditions (average pH = 8.07 ± 0.02) 66 . Finally, these phenomena would lead to a long-term preservation of marine DOM that can be partly transferred from surface to deep waters where it can be sequestered over thousand-year timescales 129 .
On the other hand, the Fe 3+ capacity to form insoluble complexes with soil DOM fractions provides evidence of sequestration and sedimentation of DOM, and its subsequent long-term stabilization as previously reported [10][11][12][13][14]74,75 . The results of this work show that Fe-HA P are rapidly precipitated from the forest soil HA due to its macromolecular size, and then Fe-FA P are formed from the remaining soil HA solution due to smaller size of FA (Fig. 2). This would imply that Fe-HA P /Fe-FA P , similarly to Fe-(oxy)hydroxide minerals and organomineral complexes, can stabilize SOM in soils and sediments 10,12,13,74,[96][97][98][99][100][101]112 . The existence of strong complexes of HA/FA in Fe-DOM P is further supported by a recent study showing that Fe (oxy)hydroxides generally exert a stronger impact on SOC storage and stabilization than aluminous clays 112 . These results also suggest that Fe-(oxy)hydroxide minerals and organo-mineral complexes consist of strong complexes formed by donation of electrons from functional groups to d-orbitals of Fe 3+ which result in strong π-d electron bonding systems the existence of which was not completely clarified earlier 10,12,13,99,130,131 . The new results of this work are thus expected to provide new insights and a better understanding of soil Fe minerals and organo-mineral complexes, as well as their behaviors not only in soils/sediments, but also in surface waters where they are introduced from soils 115,116,132 . In particular, the Fe precipitation technique might be applied efficiently to isolate specific DOM/ HS components from specific Fe minerals and clay mineral after their dissolution in alkaline or acidic media and then assess its biogeochemical state in the minerals.
In conclusion, the Fe precipitation method allows a very high recovery of HA and FA, approximately 76.2% by a single precipitation from soil HA and 38.6-93.8% from the environmental samples. Therefore, the Fe precipitation method would represent an effective/useful procedure for extracting/isolating humic substances rather than the entire DOM from aqueous media in the laboratory or for industrial purposes. The novel approach proposed and developed in this work appears robust and promising in opening a new frontier for the isolation of even tiny DOM amounts from any environmental sample, also considering that FeCl 3 is easily available, very cheap and free of organic contaminants and can replace the currently used organic-based resins.

Data availability
All data and materials have been provided in this manuscript and also supplementary material. www.nature.com/scientificreports/