Stabilization of aqueous nanoscale zerovalent iron dispersions by anionic polyelectrolytes: adsorbed anionic polyelectrolyte layer properties and their effect on aggregation and sedimentation

Abstract

Nanoscale zerovalent iron (NZVI) particles are 5–40 nm sized Fe⁰/Fe-oxide particles that rapidly transform many environmental contaminants to benign products and are a promising in situ remediation agent. Rapid aggregation and limited mobility in water-saturated porous media limits the ability to deliver NZVI dispersions in the subsurface. This study prepares stable NZVI dispersions through physisorption of commercially available anionic polyelectrolytes, characterizes the adsorbed polymer layer, and correlates the polymer coating properties with the ability to prevent rapid aggregation and sedimentation of NZVI dispersions. Poly(styrene sulfonate) with molecular weights of 70 k and 1,000 k g/mol (PSS70K and PSS1M), carboxymethyl cellulose with molecular weights of 90 k and 700 k g/mol (CMC90K and CMC700K), and polyaspartate with molecular weights of 2.5 k and 10 k g/mol (PAP2.5K and 10K) were compared. Particle size distributions were determined by dynamic light scattering during aggregation. The order of effectiveness to prevent rapid aggregation and stabilize the dispersions was PSS70K(83%) > ≈PAP10K(82%) > PAP2.5K(72%) > CMC700K(52%), where stability is defined operationally as the volume percent of particles that do not aggregate after 1 h. CMC90K and PSS1M could not stabilize RNIP relative to bare RNIP. A similar trend was observed for their ability to prevent sedimentation, with 40, 34, 32, 20, and 5 wt%, of the PSS70K, PAP10K, PAP2.5K, CMC700K, and CMC90K modified NZVI remaining suspended after 7 h of quiescent settling, respectively. The stable fractions with respect to both aggregation and sedimentation correlate well with the adsorbed polyelectrolyte mass and thickness of the adsorbed polyelectrolyte layers as determined by Oshima’s soft particle theory. A fraction of the particles cannot be stabilized by any modifier and rapidly agglomerates to micron sized aggregates, as is also observed for unmodified NZVI. This non-dispersible fraction is attributed to strong magnetic attractions among the larger particles present in the polydisperse NZVI slurry, as the magnetic attractive forces increase as r⁶.

Appendix

Effects of particle polydispersity on Ohshima’s soft particle analysis

Equation 4 consists of three terms: the first term is a weighted average of the Donnan potential (ψ _DON ) and the surface potential (ψ ₀ ), the second term is the ratio of the electric force acting on the layer fixed charges (EZeN) and the frictional force (γu), and the third term is due to the zeta potential (ζ) of the bare particle (Ohshima 1995a). The first term has the correction factor f(d/a) that accounts for alteration of an applied electric field acting on the polyelectrolyte layer due to the presence of a particle core (Ohshima 1994). This term is a function of the radius of bare particles a and the layer thickness (Eq. 7). For polydisperse colloidal particles with a range of radii a _min  − a _max and an average radius a _ave , and assuming that the layer thickness d is the same for all particles in this size range, using a _ave as a representative of the entire particle population in Eq. 4 may lead to error in estimating the layer parameters because each a in the population of a _min  − a _max has a different degree of applied electric field alteration and, thus, a different f(d/a).

Further analysis of f(d/a) can be done to show that there are two obvious cases where particle polydispersity does not affect the sensitivity of the calculated adsorbed polyelectrolyte layer properties. Figure A1a shows f(d/a) as a function of d/a. f(d/a) consists of three regions. Regions 1 and 3 represent the cases of a very thin (d « a) and a very thick (d » a) adsorbed polyelectrolyte layer, respectively, with respect to radius of the particle. The f(d/a) of both regions become constant, 1 and 2/3, respectively, with respect to d/a. In region 3, d is much larger than a; the applied electric field acting on the polyelectrolyte layer is not significantly altered by the presence of the core a. Therefore, f(d/a) becomes a constant, 2/3, which is similar to the case of a spherical polyelectrolyte without a core (Ohshima 1994). In contrast, in region 1, d is much smaller than a; the applied electric field acting on the polyelectrolyte layer becomes so distorted that it can have only its tangential component near the particle core (Ohshima 1995a). The value of this distorted field is about 3/2 times larger than the undisturbed field in the absence of the particle core, i.e. f(d/a) in region 1 is ∼3/2 times of f(d/a) in region 3 or ∼1. For this reason, in region 1 and 3, f(d/a) becomes a constant, and Eq. 4 becomes independent of a. Consequently, particle polydispersity does not affect the sensitivity of the calculated adsorbed polyelectrolyte layer properties using Ohshima’s soft particle theory if the properties of soft polydisperse particles fall into those two extreme regions. However, in region 2, the intermediate case, f(d/a) varies from 1 to 2/3, and the calculated layer properties may be sensitive to the particle polydispersity.

Fig. A1

(a) f(d/a) defined in Eq. 7 corresponding to d/a determined for PSS1M, PAP2.5K, and CMC90K modified RNIP. (b) |Δu _e | as a function of ionic strengths for CMC90K modified RNIP

Figure A2 illustrates the particle size distribution of bare RNIP determined from transmission electron microscope (TEM) images. d/a values calculated from the RNIP particle size distribution and the calculated d for polyelectrolyte-modified RNIP in this study reveal that PSS70K-, PSS1M-, CMC700K-, PAP2.5K-, and PAP10K-modified RNIP all fall into region 3 (Fig. A1a shows only PSS1M- and PAP2.5K-modified RNIP). Therefore, RNIP polydispersity should not affect the calculated layer properties of these polyelectrolyte modified RNIP. In contrast, CMC90K-modified RNIP lies in region 2 and the calculated layer properties using Ohshima’s soft particle theory may be affected by the RNIP polydispersity.

Fig. A2

Particle size distribution of RNIP obtained form TEM images

We theoretically evaluated (1) if RNIP polydispersity affects the appropriateness of using a _ave in Eq. 4 to represent the particle population for CMC90K-modified RNIP, and (2) if particle polydispersity affects the sensitivity of the calculated adsorbed CMC90K layer properties. To evaluate the former, the difference between u _e calculated using only a _ave (u _e-ave ) and the average of u _e calculated from u _e determined for each particle diameter in the entire particle size distribution (u ^ave _e-poly ) was determined for the layer properties for CMC90K-modified RNIP (Table 3). The distribution of u _e-poly is essentially a normal distribution with a mean of u ^ave _e-poly (Eq. A1, Fig.  A3a). The difference between u _e-ave and u ^ave _e-poly was attributed to the effect of particle polydispersity (Fig.  A1b).

$$ \% number(u_{e} ,u_{{e - ave}} ,\sigma _{{ue}} ) = \frac{1} {{\sigma _{{ue}} {\sqrt {2\pi } }}}\exp {\left( { - \frac{{(u_{e} - u^{{ave}}_{{e - poly}} )^{2} }} {{2\sigma _{{ue}} ^{2} }}} \right)} $$

(A1)

Fig. A3

(a) Effect of particle polydispersity on the difference between u _e-ave and u ^ave _e-poly (Δ u _e ) for CMC90K-modified RNIP in 1 mM NaCl. The filled circles represent u _e-poly theoretically generated by from the layer characteristics of CMC90K (Table 3) and particle size distribution of RNIP (Fig. A2). The curve is the fit of the data according to Eq. A1. (b) Measured u _e and u ^ave _e-poly as a function of ionic strengths and their least-squares fit of Eq. 4

To determine the sensitivity of the calculated layer properties, d _poly ,1/λ _poly , and N _poly to polydispersity, u ^ave _e-poly values at various ionic strengths (1 to 61 mM) were fit using Eqs. 4–8 and a least-squares fitting protocol. These calculated properties, which account for polydispersity, are then compared to the properties determined from ue-ave alone (Table 3). The difference between these two sets of parameters (Δd, Δ1/λ, and ΔN) is attributed to particle polydispersity effect. As shown Fig. A3a, particle polydispersity does indeed affect u ^ave _e-poly of CMC90K-modified RNIP because their d/a are in region 2. However, the effect is small and the best fit curve for both cases are the same (Fig.  A3b), yielding the same layer properties. Thus, the calculated layer properties for CMC90K-modified RNIP are indeed insensitive to polydispersity, i.e. Δd, Δ1/λ, and ΔN = 0.