Low-resistance monovalent-selective cation exchange membranes prepared using molecular layer deposition for energy-efficient ion separations

The desalination of brackish water provides water to tens of millions of people around the world, but current technologies deplete much needed nutrients from the water, which is determinantal to both public health and agriculture. A selective method for brackish water desalination, which retains the needed nutrients, is electrodialysis (ED) using monovalent-selective cation exchange membranes (MVS-CEMs). However, due to the trade-off between membrane selectivity and resistance, most MVS-CEMs demonstrate either high transport resistance or low selectivity, which increase energy consumption and hinder the use of such membranes for brackish water desalination by ED. Here, we introduce a new method for fabrication of MVS-CEMs, using molecular layer deposition (MLD) to coat CEMs with ultrathin, hybrid organic–inorganic, positively charged layers of alucone. Using MLD enabled us to precisely control and minimize the selective layer thickness, while the flexibility and nanoporosity of the alucone prevent cracking and delamination. Under conditions simulating brackish water desalination, the modified CEMs provides monovalent selectivity with negligible added resistance—thereby alleviating the selectivity–resistance trade-off. Addressing the water-energy nexus, MLD-coating enables selective brackish water desalination with minimal increase in energy consumption and opens a new path for tailoring membranes' surface properties.


Selectivity requirements in brackish water electrodialysis
Brackish water varies in salinity and composition according to its source, giving advantage to tunable membranes that could have selectivity fitting specific requirements. As an example, the brackish groundwater of the Israel's Negev region contains approximately 1000 mg/L of Na + and 100 mg/L of Mg 2+ . In addition, drinking water regulations and standards vary, with no clear worldwide standard for permitted or required Na + 1 or Mg 2+ 2 set by the world health organization (WHO). Most freshwater contains Na + in concentrations of <20 mg/L, and water becomes noticeably salty in taste with Na + concentrations above ~200 mg/L. Water is a minor source of Na + in terms of nutrition; diets tend to include too much sodium in western diets, and sometimes lack sodium in nonwestern ones. in the US, tap water was found to contain ~1-400 mg/L Na + . 1 Overall, no single standard value could be chosen. A reasonable target value of <100 ppm sodium has been chosen for demonstration. In Israel, the Ministry of Health has considered adding magnesium to desalinated drinking water in concentrations of 20 mg/L. 3 We performed an estimate on required selectivity based on target values of <100 mg/L sodium and >20 mg/L Mg 2+ : Thus, could be considered sufficient for brackish water to drinking water desalination, in this case. This value could be lower or higher given different initial and target concentrations.

Schematic of the electrodialysis system used in experiments
Figure S1: Schematic illustration of the electrodialysis system, volumes, and flow rates used in the desalination experiments. An applied electric potential on the electrodes leads to the migration of ions. In the diluate compartment, cations permeate through the CEM and anions through the AEM. In the concentrate, channel cations encounter the negatively charged AEM and fail to permeate, and anions are similarly blocked by the CEM, leading to a concentration of ions in the concentrate channel. In experiments with MVS-CEMs, the monovalentselective layer faces the diluate stream. The layer has lower permeability to multivalent ions, reducing their removal rate from the diluate channel.

Thermal stability of the coated membrane
DSC analysis was performed to check the thermal stability range of the PC-SK membrane, using a 2.28 mg sample air dried for several days. The results ( Figure S1) showed glass transition at 85 °C, with possible onset at 70-80 °C.
Based on these results, 65 °C was decided to be the highest safe temperature for performing ALD/MLD without altering the membrane structure. Melting/decomposition onset was at ~190 °C. The change observed in the 30-50 °C range is an artifact due to the initial stabilization of the system.

Development of alucone MLD procedure and growth analysis
Several techniques were used to develop the MLD procedure. In-situ QCM (an example using the procedure used in our experiments is shown in Figure S3A) showed the stepwise growth and gave a measure for the overall growth rate. It also helped to establish the stabilization time after each reagent dose, which was monitored through the change in pressure in the reactor. This allowed the selection of appropriate reagent exposure times (eventually decided to be 21 ms for the TMA and 1 s for the EG) and purge times between subsequent reagent exposures, set to be 60 s. Following deposition, ellipsometry ( Figure S3B) was used to measure the thickness of alucone deposited on Si substrates, and by performing the deposition with a membrane present in the reactor we could see the effect of reagents adsorbed and released from the membrane, leading to an increased growth rate on the Si substrate -3.8 Å/cycle for the alucone with the membrane vs 2.2 Å/cycle without it. Refractive index for the alucone layer was assumed at a constant 1.5, based on ref. 4 A rough, purely geometric analysis ( Figure S3D) shows that a fully extended chain grown by an alucone deposition cycle could have a thickness of up to ~8.4 Å. The fact that the growth rates measured by ellipsometry were lower than that indicates that a monolayer-by-monolayer growth is plausible. As ellipsometry was not feasible on the membranes, the amount of Al on the membranes was quantified by EDS. EDS was performed at the same conditions in all samples (accelerating voltage, current, and working distance), and is assumed to have a similar penetration depth and sampling volume of the electron beam. This depth is estimated to be >1 µm, much higher than the thickness of the deposited layer, meaning that a similar volume of the substrate membrane was sampled each time. We assume the sulfur concentration is homogenous and fixed in the membrane, therefore the Al/S ratio qualitatively indicates the Al concentration in the thin layer at the membrane surface. The results ( Figure S3C) show that the Al/S ratio grows linearly and that the amount of Al added in each cycle is constant in the ALD and MLD procedures.

Spectroscopy
We performed XPS and ATR-FT-IR to better understand the chemical structure of the deposited layer. Primary conclusions are included in the main text; the full spectra and further discussion are presented here.

FTIR.
Through ATR-FT-IR analysis, we find several FT-IR bands in the modified membrane that are not apparent in the pristine membrane ( Figure S6). These bands are centered around ~1598 cm -1 and ~850 cm -1 . The former is attributed to vinyl ether groups, while the latter is a convolution of several bands, including vinyl ether asymmetric vibration and Al-O vibration. Additionally, we found a small shift in the ~1250 cm -1 band and no indication of C=O groups by FT-IR (at 1700-1750 cm -1 ). It is well documented that the alucone layers go through chemical and structural transformation after exposure to ambient conditions, water, or heat. 4  XPS spectra of alucone-coated membranes before and after exposure to water were measured ( Figure S5).
Atomic composition of the coating was 21% Al, 34% C, and 45% before wetting and 16% Al, 45% C, and 39% O after water exposure. The amount of carbon was likely influenced by the presence of adventitious carbon, not support conversion to Al 2 O 3 due to water exposure, as sometimes reported, 6,7 but fits the existence of unreacted methyl groups that reacted with water. The C1s peak of an as-prepared PC-SK membrane coated with 50 cycles of alucone (a thickness that is sufficient to suppress the S2p peak from sulfonate groups of the membrane, as seen in the inset of Figure S5A) shows binding energies corresponding to C-C and C-H bonds (~285 eV), and C-O bonds (a shoulder at ~286.1 eV). After soaking the membrane in water for 2 h, we observed an increase in intensity of the 285 eV peak alongside the emergence of a new peak at ~288.7 eV. We also find small shifts of Al2p and O1s peaks to higher binding energies (0.4 and 0.2 eV, respectively). We attribute the latter changes to a decrease in hydroxide and oxyhydroxide species content (e.g., AlO(OH) 2 or Al(OH) 3 ). S2p peak was observed after coating, signifying that the layer was thick enough to suppress the S2p peak at 168 eV from the coated membrane (which contains sulfonate groups).

Current efficiency
Faradaic efficiency (FE) was calculated for all ED experiments using the equation: The numerator represents the total number of mol-equivalents transferred: percentage transport for each cation (X i ) times the number of moles of the species in the feed solution (n i ) and the valence of that species. The denominator shows the total number of electrons transferred through the system -with t being the desalination time, I the current, and F Faraday's constant (= 96500 C/mol). Results are shown in Figure S6.

Estimation of impact on energy consumption
To quantify the energy saved by using a lower-resistance selective membrane, we use the model described by H. Strathmann's "Assessment of Electrodialysis Water Desalination Process Costs". 9 Energy consumption in commercial scale-ED installations is dominated by the electric energy invested in transferring ions across the ion exchange membranes from the diluate to the concentrate. Energy invested in pumping is mostly relevant at lower salinities and other energy requirements, such as those for electrode reactions, are generally neglected. 8 The specific energy cost for desalinating a unit volume product is proportional to the cell-pair area resistance, which can be expressed as: R being the cell pair area resistance; Δ the cell thickness; and the equivalent concentrations at the feed for the diluate and concentrate, respectively; and the outlet concentrations of the diluate and concentrate, respectively; the equivalent conductivity of the solution (which can be approximated as that of a solution Λ within the operation salinity range without inducing great error) 9 ; and r am and r cm the area resistances of the AEMs and CEMs, respectively. Equation 34 in Strathmann's paper with a recovery ratio of 0.5 (equal diluate and concentrate volumes, as used in this work) reduces to: can be also be expressed in terms of desalination percentage, X: for NaCl was estimated at 25°C using the Debye-Hückel-Onsager equation : 10 Λ Λ = 126.39 -89.14 • #( 6) The ratio of energy consumption between processes using different CEMs can be expressed as: with and being the area resistances of the two CEMs compared. A desalination percentage of 90% was 1 2 chosen. r am was taken as 2 Ω•cm 2 , a typical value for commercial IEMs. 8  resistance of 4.5 Ω•cm 2 ) 11,12,13 predicts alucone-coated membrane will increase the total energy consumption by ~4% compared to non-selective CEMs, whereas state-of-the-art monovalent selective coatings would increase energy consumption by ~40% (for brackish water ED) and up to 44% in higher salinities.