Modified cation-exchange membrane for phosphate recovery in an electrochemically assisted adsorption–desorption process

A novel ion separation methodology using a cation-exchange membrane modified with iron oxide nanoparticles (Fe3O4 NPs) coated with polyhexamethylene guanidine (PHMG) is proposed. The separation is performed in an electrodialysis cell, where firstly phosphate is electro-adsorbed to the PHMG@Fe3O4 NP coating, followed by a desorption step by applying an electric current.

The size distribution and the zeta potential of the formed particles were analyzed, using dynamic light scattering (DLS) technique with a Malvern Zeta-sizer instrument. Table S2 displays the results of this analysis. The coacervates were positively charged for the PHMG:PSS function group molar ratio equal to 1:1 or for the solutions where the number of guanidinium groups was used in excess. As expected, when the PSS used was in excess, charge reversal at the surface of the coacervates occurred. The solution with a PHMG:PSS equal to 2:1 was chosen for further experiments. This was because the formed coacervates formed visible agglomerates (see Fig. 2A in the main text) that were easily to separate by centrifugation. Table S2. Zeta sizer results from coacervate solutions analysis, for optimal ratio.

Ratio (PHMG:PSS)
ζ-potential (mV) Size (nm) The fluctuation in the coacervate size (see Table S2) can be controlled by the careful balance of the charge ratio between functional groups located within polymer structure. When one of the polyelectrolytes is used in excess, the formed particles surface charge is rapidly saturated and their further growth is inhibited by electrostatic repulsion. In turn, particles continue to grow and agglomeration can be triggered close to the isoelectric point. 1

Protocol -phosphate batch adsorption experiments using coacervates
In order to prove the ability of coacervates to adsorb phosphate, five solutions (50 mL each) containing PHMG:PSS at a ratio equal to 2:1 (concentration of PSS was varied from 1.5 mM to 50 mM) were prepared.
The coacervate solutions were stirred for 24 h. Next, the phosphate was added to each solution so that the final molar ratio was 2:1:1 ( PHMG:PSS:PO 4 3 ). The obtained solutions were stirred for 24 h followed by centrifugation at 5000 rpm for 5 min using a Hermle Z 326 K centrifuge. Finally, 10 mL of a supernatant was taken and analyzed by ion chromatography (IC). The results of this analysis can be found in Figure 2B available in the main text of this study.

S4 Fe 3 O 4 NPs modification and batch adsorption experiment
Two separate aqueous solutions of PHMG (2.5 g/L) and Fe 3 O 4 NP suspension (0.5 g/L) were prepared using a ultrasonication bath for 20 min. Next, the pH of both solutions was adjusted to 9. 5  allowing for the particle size distribution, functional group analysis and ζ-potential determinations. Bare Fe 3 O 4 NPs were also analyzed and used as a blank.

S4.1 TGA Analysis
The bare and modified NPs were dried in a Binder oven at 50 °C for 24 h. The TGA analysis was performed in a GA2/SF1100 STARe system from Mettler Toledo. Between 7 and 10 mg of dry NP were exposed to a temperature increase from room temperature up to 800 °C, at a rate of 5°C per min. The analysis was performed in a nitrogen instead of an air environment as the iron (II), present in the magnetite particles, will be oxidized to iron (III). Working under nitrogen also means that the obtained result can be only treated qualitatively (we follow calcination rather than oxidation). The observed weight loss of the organic part informs about the degradation (carbonization) only and not the complete oxidation of the polymer. Figure S1 displays the TGA of modified and unmodified NPs. The presence of the PHMG coating was confirmed by analyzing the differences in weight loss between the two measurements. While the bare NPs are stable throughout the measurement, the modified NPs show a weight loss that we attributed to the polymer degradation. The weight loss of only 3% is not surprising as the PHMG forms a thin layer on the magnetite NPs surface.

S4.2 FTIR analysis
For the FTIR measurements, the NPs were removed from the solution and dried in the way as described for the TGA experiments. Next, a KBr pellet was made using a manual press.

S4.3 ζ-potential measurements of the PHMG@Fe 3 O 4 NPs
The ζ-potential analysis was performed with a Malvern Zeta-Sizer. The dry NPs powder was dispersed in Milli-Q water and sonicated for 10 min. The results are summarized in Table S3. The ζ-potential after the modification is significantly higher as compared with the bare NPs.

S5
Membrane modification The CEM was initially activated by argon plasma treatment (2 min) in a Harrick Plasma PDC-002-CE plasma cleaning equipment, at a high RF level. This step had the aim of generating free radicals, which could potentially react with the CEM surface and increase its charge. 6 Afterward, the membrane was placed in a Millipore filtration holder, simply to provide support for the coating process. Next, 2 mL The NPs count remains the same (see on the SEM images in Fig. 3A and 3B), and the iron and oxygen do not diminish on the corresponding EDS mapping micrographics and the images also look comparable. All combined results confirm the superior stability of the coating.
Furthermore, electrical resistance (ER) was measured in the electrodialysis setup which is described in section S6. A potentiometric measurement was used, by applying a current from 0 to ≈ 0.3 A and measuring the potential difference across the membrane. The measurement was performed in a solution of 0.5 M sodium phosphate (NaH 2 PO 4 ) (pH=5). As it can be observed in Table S4, the ER is reduced in the presence of the coating. A possible explanation for this is that the interface between the anionic coating and the CEM, resembling a bipolar membrane, is responsible for a higher water dissociation rate at a bipolar junction in the presence of externally applied current. This, in turn, can result in the formation of H + or OHions which have higher mobility, and by carrying the current which is being applied, they reduce the resistance. Although this is likely the case, it is just a speculation as the pH at the surface of the membrane is very challenging to be measured.    For the blank experiments, we repeated the protocol described above for the unmodified CEM in the (i) presence and (ii) absence of the applied electric current. For both scenarios, we observed a gradual and similar increase in phosphate concentration in the receiving compartment. This, in turn, suggest that it is physical adsorption of phosphate to the membrane support rather than transmembrane transport. Also, the phosphate concentration in the receiving compartment found during the blank experiment was always lower than that for the modified membrane. Clean and pattern bars correspond to the adsorption and desorption step respectively.