1. Introduction
Although all living organisms require trace amounts of heavy metals such as cobalt, copper, iron, and manganese, non-essential heavy metals released in large quantities such as cadmium, chromium, mercury, lead, arsenic, and antimony are of great environmental concern [
1]. Exposure to such metals can cause such serious health effects as reduced growth and development, cancer, organ damage, nervous system damage, and in extreme cases, death [
2]. Many heavy metal pollutants are present in wastewaters that are discharged from metallurgical/metal manufacturing, electroplating and mining operations, printing, dye and paint, pulp and paper, textiles, and petrochemical operations, and from the manufacture of batteries and chemicals [
3,
4]. The growth of these industries results in direct or indirect discharge of larger amounts of heavy metals into the environment where they can accumulate in lethal quantities, particularly in developing countries that lack proper waste management protocols [
3,
4,
5].
Substantial efforts have been undertaken to develop strategies to mitigate the amount of metal released to the environment by waste disposal and wastewater discharge. Many conventional wastewater treatment processes such as chemical precipitation, adsorption and ion exchange, and electrochemical deposition have been used to remove and recover heavy metals from impaired waters [
2]. Additionally, newer approaches including membrane separation and electrodialysis have been developed to increase the level of metals removed from wastewaters [
2,
4].
Ion-exchange processes have been used widely to separate and recover specific metal ion impurities from wastewaters via the use of packed beds of ion-exchange resins. The most common cation exchangers in use are strongly acidic resin beads with sulfonic acid groups and weakly acidic resin beads with carboxylic acid groups. Separation occurs through the exchange of hydrogen ions (at pH < pKa) or cations such as sodium ions (at pH > pKa) on the resin with the metal cations in solution. However, the ion exchange resin process is limited in part by pressure drop across the resin bed during flow, which can be exacerbated by media deformation and non-uniform packing [
6,
7]. In several studies undertaken to correlate the pressure drop across packed beds of ion-exchange resin with flow rate, it was determined that the pressure drop must be known to design an ion exchange process effectively [
8,
9,
10]. It was further determined that backwashing was required to minimize the pressure drop caused by resin fines and suspended solids [
10].
The ion exchange resin process is also limited by mass transfer of ions to binding sites. Although it is possible to decrease the characteristic time for diffusion through the use of smaller porous beads, the pressure drop will increase accordingly [
11]. To understand diffusion limitations in packed-bed resin processes, studies have been undertaken to elucidate the effect of flow rate (i.e., residence time) on the ion-exchange capacity. In a study on the use of a Purolite C100-MB cation exchange resin bed to remove copper(II) from aqueous solution, Hamdaoui [
12] found that the removal depends upon the flow rate of the system. He also determined that the flow rate dependence is characterized by the need for a long residence time to achieve a high enough ion exchange capacity by providing more time for interaction between the metal and the resin [
12]. In a similar experiment, Amberlite IRC-718 was used to elucidate the binding capacity of the resin for zinc(II). A higher capacity was measured at the lower flow rate, but reducing the flow rate also reduced productivity [
13].
Membrane columns are an alternative to packed resin columns. A common configuration is to pack the column with a short stack of porous membranes with a large diameter to avoid high pressure drops [
11]. Feed solution passes through the membrane bed, and since binding ligands are attached to the pore surfaces of the porous membrane, the diffusional path length of the target molecules (or ions in this case) to the functional groups is short [
14]. Therefore, the transport of target solutes to the binding sites is controlled mainly by pressure-driven convective flow through the membrane [
14,
15]. As long as the residence time from convective flow is longer than the characteristic time for ion binding (i.e., Damköhler number, Da > 1), binding capacities are expected to be independent of flow rate. Membrane sheets can be cut to size and pre-packed into commercial filtration cells of various diameters. Scale up of membrane columns is linear and easy to do, unlike resin-packed beds [
14,
15].
Wada et al. studied copper chelation under flow conditions using a functional polymer grafted to porous polyethylene sheets as an alternative to conventional sorbents in bead form [
14]. It was determined that porous sheets modified with iminodiacetate (dicarboxylate) chelating groups had a three-fold improved copper dynamic binding capacity at the same flow velocity than the DIAION
®CR 11 chelating beads with the same functional groups [
14]. Additionally, compared to conventional resin-based ion-exchange media, high-surface area ion-exchange membranes coated with polymer ligands exhibited a higher binding capacity and higher volumetric throughput [
4,
16]. For instance, it was reported that Amberlite IRC 748 ion exchange resins bind Cu(II) at flow rates of 10 bed volumes (BV)/h and a productivity of 0.13 mg Cu recovered/g resin/min (at an initial concentration of 317 mg/L), while functionalized nylon membranes could achieve flow rates up to 400 BV/h for Cu(II) and productivity of 2 mg Cu recovered/g membrane/min (at an initial concentration of 100 mg/L) [
17,
18].
This paper reports findings on the performance of nanofiber-based ion-exchange membranes for the rapid recovery of heavy metals from impaired waters. Membranes were prepared according to previous work [
19]. The goals were to understand the role of flow rate on membrane dynamic binding capacity and productivity, and to develop a strategy to regenerate the membranes for metal ion recovery and membrane reuse. To attain these goals, we prepared macroporous cellulose nanofiber membranes by electrospinning and modified them by grafting poly(acrylic acid) (PAA) and poly(itaconic acid) (PIA) onto the nanofibers. The dynamic binding capacity of cadmium on the modified membranes was evaluated at different flow rates. Cadmium was selected for the study because it is a toxic metal of great toxicological concern, in that it is persistent and cannot be broken down into less toxic substances in the environment [
20,
21]. This metal is of such a concern that most countries have their own pollution control department to limit cadmium in all industrial effluents prior to disposal [
21]. The role of the polymer coating type on both the dynamic binding capacities and the ion-exchange kinetics was analyzed, as was the polymer swelling behavior in the presence and absence of cadmium.
2. Experimental Section
2.1. Materials
The following chemicals and solvents were purchased from Sigma-Aldrich (St. Louis, MO, USA) and used as received: Alumina (an inhibitor remover), azobisisobutyronitrile (AIBN, 98 wt %), cadmium nitrate tetrahydrate ((Cd(NO3)2·4H2O), ≥99 wt %), cellulose acetate (CA, average Mn = 30,000 Da by GPC), diethyl ether (≥99%), ethylenediaminetetraacetic acid (EDTA, 99.4 wt %, powder), glycidyl methacrylate (GMA, 97 wt %), itaconic acid (IA, ≥99 wt %), poly(acrylic acid) (PAA, average Mw = 250,000 Da, 35 wt % in water). The following chemicals and solvents were purchased from Fisher Scientific (Waltham, MA, USA) and used as received: acetone (histological grade), 2-butanol (2-BuOH, 99%), dioxane (≥99%), nitric acid (90%, aq). Dimethylacetamide (DMAc, 99 wt %) and sodium hydroxide (NaOH, pellets, 97 wt %) were purchased from Alfa Aesar (Ward Hill, MA, USA). Chloroform (HPLC grade) was purchased from Honeywell (Morristown, NJ, USA).
2.2. Methods
2.2.1. Preparation of Poly(glycidyl Methacrylate) (PGMA)
PGMA was prepared by radical polymerization of GMA using AIBN as initiator. GMA was dehibited by passing it through a column packed with alumina. The polymerization was conducted by adding dehibited GMA (10 g) and AIBN (1 g) in acetone (15 mL) at 40 °C under a nitrogen atmosphere and stirring with a magnetic bar. PGMA (Mw = 89,500 Da by gel permeation chromatography) obtained after a 2 h polymerization was purified by multiple precipitations (4 times) from acetone by the addition of diethyl ether. The PGMA was dried in a vacuum oven under 50 kPa overnight at room temperature.
2.2.2. Preparation of Regenerated Cellulose Nanofiber Membrane Support
Regenerated cellulose nanofiber membranes were prepared according to previously established procedures [
19]. Concisely, cellulose acetate in DMAc/acetone solvent was electrospun for 4 h using a voltage of 12.5 kV and a polymer solution flow rate of 0.3 mL/h. The relative humidity of the system was maintained at 65% to avoid bead formation on the fibers. The membranes were sintered thermally under an applied pressure and converted to regenerated cellulose nanofiber membranes by treatment with dilute NaOH solution. PGMA was coated on the regenerated cellulose nanofiber membrane from a solution in chloroform, and annealed under a vacuum. PGMA-coated regenerated cellulose nanofiber membranes (RC-PGMA) had an average thickness of 72 ± 13 µm.
2.2.3. Preparation of Nanofiber Ion-Exchange Membranes
Again using a previously established procedure [
19], PAA was precipitated by mixing the PAA solution with acetone. The mixture was centrifuged and the clear PAA gel was separated from the solution by decanting and was dried completely.
PIA was prepared by the radical polymerization of IA using AIBN as an initiator. Specifically, IA (10 g) in dioxane (40 mL) was added into an Erlenmeyer flask (150 mL). The quantity of the initiator AIBN was 1.6% by weight (0.16 g) in relation to the monomer. It was added to the solution containing IA in dioxane. The mixture was heated for 48 h at 60 °C, and acetone was added to the mixture in a separatory funnel to precipitate PIA and was left overnight in the hood. The product PIA (Mw = 7000 Da by intrinsic viscosity measurements) was separated from the monomer by vacuum filtration, washed three times with 50 mL of acetone, and dried overnight at 50 °C and 50 kPa in a vacuum oven.
PAA and PIA were dissolved in deionized water (DI water) to prepare 7 wt % polymer solutions, which were then sonicated for either 1 h or until they formed a homogenous solution. Each polymer solution was passed through a cellulose acetate sterile syringe filter (with a pore size 0.45 µm) purchased from VWR International (Radnor, PA, USA) to remove large size aggregates before use. Each RC-PGMA membrane was submerged in 4 mL of the filtered polymer solution for 5 min, removed from the solution, and annealed at 70 °C for 2 h. Thereafter, the membranes were washed three times (5 min per wash step) via immersion in 30 mL of water to remove any non-grafted polymer.
2.3. Membrane Morphology
The uniformity and quality of the fibers were examined by scanning electron microscopy (SEM model: S4800, Hitachi High Technologies America, Inc., Schaumburg, IL, USA). Representative 0.5 cm2 samples of the membranes were attached with carbon tape to aluminum stabs prior to the SEM measurements. The SEM measurements were performed at an accelerating voltage of 5.0 kV, current of 2 mA, and magnifications of 3000–12,000×. Average fiber diameters and fiber diameter distributions were determined from SEM images. Moreover, changes in the morphology of fibers due to sintering, hydrolysis, and modification of polymers were monitored by analysis of SEM images.
2.4. Performance Properties of Nanofiber Membranes
2.4.1. Dynamic Binding Capacity and Regeneration
The PAA- and PIA-modified regenerated cellulose nanofiber membranes (RC-PGMA-PAA and RC-PGMA-PIA) were neutralized via immersion in a sodium hydroxide solution (0.05 M) for 1 h, cut into 45 mm diameter sections using a circular die, weighed, and placed into a 300 mL ultrafiltration stirred cell (Sterlitech Corporation, HP 4750 Stirred Cell, Kent, WA, USA). A cadmium solution with a concentration of 10 mg/L in DI water was prepared from Cd(NO3)2·4H2O and added to the ultrafiltration cell connected to an air cylinder. The permeate samples were collected every 5 min until fully loaded with Cd using feed pressures of 20.7, 34.5, and 48.3 kPa. Each permeate sample was weighed to determine mass flow rate, and Cd concentrations were measured to prepare breakthrough curves for evaluating the dynamic binding capacities. A 0.5 M aqueous solution of EDTA was prepared and pH adjusted to 6 for use as the regeneration reagent. The EDTA solution was passed through the membrane in the ultrafiltration cell and permeate samples were collected, weighed, and analyzed for Cd concentrations to prepare the regeneration curve. After regeneration to recover the Cd, sodium hydroxide solution (0.05 M, 10 mL) was passed through the membrane to ensure that the membrane was neutralized to the sodium carboxylate form. Finally, the membrane was rinsed by passing 20 mL of DI water through it before testing the next cycle. Five cycles of Cd binding and regeneration by EDTA were done at an applied pressure of 20.7 kPa. A single cycle was performed for each of the two other applied pressures.
Inductively coupled plasma optical emission spectroscopy (ICP-OES, Optima 5300 DV, Perkin Elmer, Waltham, MA, USA) was used to measure Cd(II) concentrations in permeate samples, with the peak area for each ion measured five times for each sample and calibration solution.
2.4.2. Static Cadmium Binding Kinetic Study
Batch ion-exchange measurements were used to evaluate cadmium binding kinetics and to determine the static (i.e., equilibrium) binding capacities. Cadmium solutions (3 mL) with an initial concentration of 20 mg/L were placed in 20 mL vials. Next, PAA- and PIA-modified cellulose nanofiber membranes were cut into 10 mg pieces and placed into each vial. Experiments were conducted for contact times from 1 to 24 h in a shaker bath (22 °C, 100 rpm). ICP-OES was used to measure Cd(II) concentrations.
2.4.3. Polymer Characterization in Solution
Dynamic light scattering (DLS, Model: Dawn Heleos-II, Wyatt Technology, Santa Barbara, CA, USA) was used to determine the hydrodynamic radius (Rh) of PAA and PIA in DI water before and after Cd(II) loading. Cd(II) solutions with concentrations from 490 to 6250 mg/L in DI water (20 mL) were prepared and adjusted to pH 7 by NaOH solution. PAA (0.15 g) was dissolved in each Cd(II) sample and pH was again adjusted to pH 7 by NaOH. Separately, PAA and PIA (0.15 g) were dissolved in DI water (20 mL) and neutralized by NaOH to prepare controls. All solutions were analyzed to determine Rh of the polymer under different conditions. The light scattering analyses were performed every 3 s over the course of 5 min for each sample, to determine values with uncertainties representing one standard deviation.