You are on page 1of 13

6006

Langmuir 2005, 21, 6006-6018

Controlled Clustering and Enhanced Stability of Polymer-Coated Magnetic Nanoparticles
Andre Ditsch, Paul E. Laibinis,† Daniel I. C. Wang, and T. Alan Hatton*
Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139 Received November 30, 2004. In Final Form: March 23, 2005
The clustering and stability of magnetic nanoparticles coated with random copolymers of acrylic acid, styrenesulfonic acid, and vinylsulfonic acid has been studied. Clusters larger than 50 nm are formed when the coatings are made using too low or too high molecular weight polymers or using insufficient amounts of polymer. Low-molecular-weight polymers result in thin coatings that do not sufficiently screen van der Waals attractive forces, while high-molecular-weight polymers bridge between particles, and insufficient polymer results in bare patches on the magnetite surface. The stability of the resulting clusters is poor, but when an insufficient polymer is used as primary coating, and a secondary polymer is added to coat remaining bare magnetite, the clusters are stable in high salt concentrations (>5 M NaCl), while retaining the necessary cluster size for efficient magnetic recovery. The magnetite cores were characterized by TEM and vibrating sample magnetometry, while the clusters were characterized by dynamic light scattering. The clustering and stability are interpreted in terms of the particle-particle interaction forces, and the optimal polymer size can be predicted well on the basis of these forces and the solution structure and hydrophobicity of the polymer. The size of aggregates formed by limited polymer can be predicted with a diffusion-limited colloidal aggregation model modified with a sticking probability based on fractional coating of the magnetite cores.

Introduction Colloidally dispersed magnetic nanoparticles show considerable promise for a wide range of applications as sealants,1 damping agents,2 drug-delivery vehicles,3 contrast agents in magnetic resonance imaging,4 and separation aids.5-7 In many cases, these colloidal dispersions, or magnetic fluids, consist of magnetite (Fe3O4) nanoparticles, typically ∼10 nm in size, coated with surfactants or polymers6-15 both to stabilize the particles in suspension and to provide favorable surface properties tailored for specific applications of interest. The small size of the stabilized particles results in dispersions that remain suspended indefinitely in gravitational and moderate
* Author to whom correspondence should be addressed. E-mail: tahatton@mit.edu. Tel: 617-253-4588. Fax: 617-253-8723. † Department of Chemical Engineering, Rice University, Houston, Texas, 77005.
(1) Rosensweig, R. E. Chem. Eng. Prog. 1989, 85, 53-61. (2) Raj, K.; Moskowitz, R. J. Magn. Magn. Mater. 1990, 85, 233245. (3) Lubbe, A. S.; Bergemann, C.; Brock, J.; McClure, D. G. J. Magn. Magn. Mater. 1999, 194, 149-155. (4) Kawaguchi, T.; Yoshino, A.; Hasegawa, M.; Hanaichi, T.; Maruno, S.; Adachi, N. J. Mater. Sci.-Mater. Med. 2002, 13, 113-117. (5) Safarik, I.; Safarikova, M. J. Chromatogr. B 1999, 722, 33-53. (6) Moeser, G. D.; Roach, K. A.; Green, W. H.; Laibinis, P. E.; Hatton, T. A. Ind. Eng. Chem. Res. 2002, 41, 4739-4749. (7) Bucak, S.; Jones, D. A.; Laibinis, P. E.; Hatton, T. A. Biotechnol. Prog. 2003, 19, 477-484. (8) Jones, F.; Colfen, H.; Antonietti, M. Colloid Polym. Sci. 2000, 278, 491-501. (9) Mendenhall, G. D.; Geng, Y. P.; Hwang, J. J. Colloid Interface Sci. 1996, 184, 519-526. (10) Rosensweig, R. E. Ferrohydrodynamics; Cambridge University Press: London, 1985. (11) Shen, L. F.; Laibinis, P. E.; Hatton, T. A. J. Magn. Magn. Mater. 1999, 194, 37-44. (12) Shen, L. F.; Laibinis, P. E.; Hatton, T. A. Langmuir 1999, 15, 447-453. (13) Shen, L. Ph.D. Thesis, MIT, Cambridge, MA, 2000. (14) Shen, L. F.; Stachowiak, A.; Fateen, S. E. K.; Laibinis, P. E.; Hatton, T. A. Langmuir 2001, 17, 288-299. (15) Wooding, A.; Kilner, M.; Lambrick, D. B. J. Colloid Interface Sci. 1992, 149, 98-104.

magnetic fields.10 Large surface areas per unit volume make the particles ideally suited for use in adsorptive separations6,7 since their capacity for targeted solutes is considerably greater than the capacity of commercial resins. The surface area is available without internal pores, and thus, separations are not limited by pore diffusion and can, in principle, be performed much more quickly than with standard porous materials. In separation processes, the magnetic properties of the loaded nanoparticles can be exploited in their recovery from process streams by using high-gradient magnetic separation (HGMS) technology.6,7 This process relies on the fact that the force acting on a magnetic particle in a magnetic field depends on the particle size and the magnetic field gradient according to

F ) -µ0VM H
where V is the volume of the magnetic nanoparticle and M is its magnetization in a given field, H. An HGMS system typically consists of a column packed with a bed of magnetically susceptible wires, on the order of 50 µm in diameter, placed between the poles of an electromagnet. The wires dehomogenize the applied magnetic field to establish the large field gradients near the wire surfaces required for capturing the particles. HGMS has been examined for the capture of magnetic nanoparticles6,7,16-18 where it has been demonstrated, both experimentally and theoretically, that individual nanoparticles cannot be captured effectively by HGMS because diffusional and drag forces can overcome the forces of magnetic attraction to the wires, and the particles are swept through the column relatively unimpeded. Small clusters of magnetic
(16) Moeser, G. D.; Roach, K. A.; Green, W. H.; Laibinis, P. E.; Hatton, T. A. AIChE J. 2004, 50, 2835-2848. (17) Fletcher, D. IEEE Trans. Magn. 1991, 27, 3655-3677. (18) Gerber, R.; Takayasu, M.; Friedlaender, F. J. IEEE Trans. Magn. 1983, 19, 2115-2117.

10.1021/la047057+ CCC: $30.25 © 2005 American Chemical Society Published on Web 05/18/2005

The clustering and stability of the nanoparticles were analyzed on the basis of calculated energies of interaction of the particles. The deoxygenation was achieved by bubbling (22) Bokias.. with 0. It is the purpose of this paper to demonstrate both that clusters of controlled size can be synthesized readily and that the required stability can be achieved through posttreatment with no further growth of these clusters. Polymer Synthesis. sodium chloride.. ammonium hydroxide (29. D.028 mol) was dissolved in Milli-Q water. . it is highly unlikely that the resulting clusters will be stable. A. acrylic acid (99%).. which are suitable for adsorptive ionexchange purification of proteins.: Lindenmann.5 mmol (typically 2.20 the particles do not cluster and are difficult to capture. Cationic polymers were made following the same procedure. The magnetic nanoparticles are coated with random copolymers of acrylic acid. G. Laibinis. Hourdet.7% in water). C. Academic Press: Orlando. then many applications requiring capture of these clusters and their reuse in adverse environments can be realized. (21) Ditsch.15 The synthesis of stable clusters of these particles is a difficult prospect since the primary particles must be unstable to form the initial clusters and the clusters formed are therefore themselves on the threshold of stability and are prone to further aggregation with small changes in solution conditions.Polymer-Coated Magnetic Nanoparticles Langmuir. When particles are formed with coatings that provide robust stabilization. vinylsulfonic acid. (20) Napper.22 The reaction was performed as follows. P. The magnetic nanoparticles were produced by chemical co-precipitation. a common situation is when attractive van der Waals interactions initially overcome electrostatic repulsion forces to allow cluster formation but where the electrostatic energy barrier to coagulation increases as clusters grow so that beyond a finite cluster size no further growth occurs. Macromol. such as those that repel each other sterically and provide thermodynamic stability. and the mixture was placed in a sealed glass vial and heated to 80 °C for 3 h. E. magnetic nanoparticles will tend to aggregate to reduce their surface energy.86 g of iron(II) chloride tetrahydrate were added to 40 mL of deoxygenated water.9. We have explored the clustering of nanoparticles by three methods: (i) using a low-molecular-weight polymer that provides a coating of insufficient thickness to provide steric stabilization against the van der Waals interactions. These ions greatly change the strength and range of electrostatic interactions. Vol. (ii) using a high-molecular-weight polymer to bridge between particles. Thus. Acetone. 1998. particularly for biological applications. we outline models for the formation of clusters and for determining the optimum molecular weight of a polymer coating. Wang. Synthetic Procedure. No. potassium persulfate (99% ACS reagent). The result is a random copolymer in which hydrophobicity (styrene sulfonic acid). Monomer (0. as shown in Figure 1. (ar-vinylbenzyl)trimethylammonium chloride (99%). When a polymer is either too large or too small. however.. and hydrophobicity of the coatings. (19) Bagchi. 4-styrenesulfonic acid sodium salt hydrate (technical). D. Molecular weight was controlled via chain transfer with sodium metabisulfate.6 mmol) of sodium metabisulfate were added. as shown in Figure 2. and vinyl sulfonic acid. and 3-acrylamidopropyltrimethylammonium chloride (75 wt% in water) were obtained from Sigma-Aldrich and used as received.21 If we can control the stability of clusters formed under incipiently unstable conditions. The final volume was adjusted to 22 mL. 1983. and 0. 2. the ionic strength can vary over a wide range and multivalent ions may be present. attachment density (acrylic acid). if the ionic strength of the medium is increased. J. 1974. Chem. 21. S. T. submitted for publication. leading to instability and coagulation of poorly coated nanoparticles at moderate ionic strengths. the instability that causes clustering could not be removed easily and highly stable clusters could not be synthesized. typically a surfactant or polymer. and the many polymer coatings that have been examined for magnetic nanoparticles vary from simple homopolymers to graft and block copolymers6.007 mol of acrylic acid and 0. be stabilized by use of a suitable surface coating. A. 86-99. In a typical procedure. vinylsulfonic acid sodium salt (technical.14 However. be removed efficiently from process streams using HGMS. ferrous chloride tetrahydrate (99%). Experimental Section Materials. Due to their small size and large surface area. 47. however. and (iii) using a moderatemolecular-weight polymer in a limited amount to coat the particles partially. and acrylic acid with potassium persulfate as the initiator. as described by Bokias et al. and sodium metabisulfite as a chain-transfer agent.35 g of iron(III) chloride hexahydrate and 0. Hatton.16. and from this analysis. We show that all three methods result in clusters large enough for capture in an HGMS column. 199. and sodium metabisulfate were obtained from Mallinckrodt and used as received. In separation processes. and molecular weight can be independently tuned. A. 2005 6007 particles on the order of 50 nm or larger can. I. The instability that causes clustering with a limited amount of low-molecular-weight polymer was easily removed by using a polymer that coats well in limited amounts as the first coating followed by the subsequent addition of more polymer to stabilize the clusters formed during the first step.8. allowing a systematic study of the effects of the coating properties on the resulting particle clusters to be performed. styrene sulfonic acid. For example. A wide range of random co-polymers of varying composition and molecular weight was prepared by aqueous free-radical polymerization of styrenesulfonic acid. This simple polymer system also allows independent tuning of the molecular weight. D. Phys.19. Nanoparticle dispersions can.37 mmol of potassium persulfate and 0-10. it is desirable to synthesize clusters of the particles to provide the size needed for the particle recovery but without compromising the adsorptive capacity nor the stability of the suspension. The monomers are mixed in aqueous solution with potassium persulfate as a free-radical initiator.. 25% in water). Figure 1. attachment density. 13. Polymer synthesis. Unless there is a method for removing the initial instability that permits cluster formation.6. P.021 mol of either 3-acrylamidopropyltrimethylammonium chloride or vinylbenzyltrimethylammonium chloride used as monomers.. Ferric chloride hexahydrate (97%). H. Colloid Interface Sci. these clusters aggregate further to form permanent precipitates that cannot be resuspended. Polymeric Stabilization of Colloidal Dispersions. P. Durand. Particle Synthesis.6. 1387-1392.

13. Intensity-averaged size distributions were converted to volume-averaged and number-averaged size distributions for further analysis. Magnetic fluids were diluted to 0. Stability Determination. John Wiley and Sons: New York. Molecular weights were determined with a Brookhaven BI-200SM light scattering system using the provided Zimm plot software at angles from 30° to 150°. indicating the precipitation of magnetite. Vol. 1991. They were then mixed again and centrifuged for 15 min at 4000 rpm in an Eppendorf 5810R centrifuge. 2005 Ditsch et al.0 M. this condition was always satisfied since κa > 100. Zeta Potential Measurement. The zeta potentials of particle suspensions were measured on a Brookhaven ZetaPals Zeta Potential Analyzer. 3. A mixture of 5 mL of 28% ammonium hydroxide and various amounts of polymer dissolved in 5 mL of water. . at increments of approximately 1 M. or if the particle size distribution indicated that particles greater than 0. from 0.1 M NaCl to just above the highest stable concentration. typically 4 mmol on a monomer basis. Refractive indexes were determined for each polymer composition and concentration to enable analysis of the scattering data to give the weight-averaged molecular weight.22 and 0. 3rd Edition ed. Particle suspensions were diluted to 0. The magnetic fluid was then precipitated with 50-100 mL of acetone. Dynamic Light Scattering. 1997. P. was added to ensure complete coating of the exposed particle surfaces. Rajagopalan. G. Quoted particle sizes are volume averages and are the average of four measurements. typically 7 mmol of monomer units.45 µm were present. typically 2-10 mg/mL in 1 M NaCl to limit interparticle interactions. If evidence of particles larger than the filter pore size was observed. and the experiment was then repeated in increments of 0. Vibrating Sample Magnetometry (VSM). Transmission electron microscopy (TEM) experiments were performed on a JEOL 200CX (200 kV) instrument. mixed well with a vortex mixer. (23) Odein. at least four different molecular weights were obtained at varying chain transfer amounts and fit to eq 1: η ζ ) µe (2) where η and are the viscosity and dielectric constant of the dispersion medium. For each monomer composition studied.025 wt% in solutions of various sodium chloride concentrations. Samples were prepared by evaporating dilute suspensions on a carbon-coated film. (24) Hiemenz. The reaction was allowed to proceed for 15 more minutes (30 min total) before cooling to room temperature. Principles of Colloid and Surface Chemistry. CS is the chain transfer coefficient. with particles larger than 70 nm. Molecular Weight Determination. The electrophoretic mobility (µe) of the particles measured over 25 electrode cycles was converted to the zeta potential (ζ) using the Smoluchowski equation: Figure 2. The quoted zeta potential is an average of five measurements.25 where κ is the inverse Debye length and a is the particle radius. either by visually inspecting the filter. Principles of Polymerization. Equation 2 is only valid when κa . additional polymer. The autocorrelation function was fit with an exponential fitting software program to extract the diffusion coefficient.24. respectively. Oxford University Press: New York. Dynamic light scattering (DLS) experiments were performed with the Brookhaven BI-200SM light scattering system at a measurement angle of 90°. [S] 1 1 ) + CS XW X0 [M] (1) where XW is the weight averaged degree of polymerization. Magnetic fluid synthesis. The resulting Fe3+ and Fe2+ concentrations were 0. Electron Microscopy Measurements. X0 is the degree of polymerization when no chain transfer is added. Characterization of Polymers and Magnetite Nanoparticles In this section. and resuspended in 30 mL of Milli-Q water and sonicated for 30 s with a Branson sonifier 450 at an output of 40%. No.45 µm syringe filter to remove dust.6008 Langmuir.. [S] is the molar concentration of chain transfer agent. 1998. resulting in the required 2:1 ratio for magnetite (Fe3O4) production. and the mixture was then heated to 80 °C. Marcel Dekker: New York. R. magnetically decanted. 21. The resulting particles are clustered and not completely coated. The mixture immediately turned black. The samples were filtered through a 0. The median size and polydispersity of the magnetite particles were determined by visually measuring the major and minor axes of 150 particles and taking their averages. respectively. Most samples were filtered with a 0. and [M] is the total molar concentration of the monomer. The highest concentration that resulted in no particle sedimentation was reported as the critical coagulation concentration.1 M NaCl.1 to 5. Introduction to Modern Colloid Science.1 M NaCl prior to measurement. C. A second polymer addition provides complete coating. and the Stokes-Einstein equation was used to convert the diffusion coefficient to the hydrodynamic diameter. was added to the three necked flask. J.. The highest concentration at which no particle coagulation occurred was noted. Vol. The polymers were dissolved at five concentrations. we provide primary characterization of the polymers and the isolated magnetic nanoparticles.005 wt% Fe3O4 with 0.11 M. The magnetic nanoparticles are produced by chemical co-precipitation of iron salts in an aqueous solution of a limited amount of SSA/VSA/AA copolymer. A detailed discussion of the characteristics of the nanoparticle clusters and their stability is given in subsequent sections.22 µm syringe filter to eliminate dust. nitrogen through the solution in a vigorously stirred 100 mL three-necked flask. As all samples were run in 0. Two milliliters of the sample were loaded into the electrode cell. After 15 min. The nitrogen bubbling was ceased. greatly stabilizing the clusters without affecting cluster size. The magnetic field was varied from -1 to 1 and back to -1 T in steps of 0. the measurement was repeated several times without filtering to obtain the true size distribution. VSM measurements were performed on 30 µL samples of magnetic fluid using an ADE 880 VSM instrument. 1. (25) Hunter. and left at room temperature for a minimum of 24 h.05 T at room temperature to obtain the magnetization of each sample as a function of the applied field. R.

11 Figure 4. The hydrodynamic diameter of the polymer in 1 M NaCl solution. Even with large aggregates. Phys. No.. Muller. these micrographs may capture the actual solution structure of the clusters. D. TEM Size Distributions. closely resembling a random coil. Polymers with a wide range of monomer compositions and molecular weights were synthesized.0 M NaCl. R. was used to obtain the magnetic characteristics of the particles and to provide a second measurement of the core size distribution. only the magnetite cores can be seen in the micrograph. A vibrating sample magnetometer (VSM). The molecular weight for a polymer composed of heavier monomers. there is further particle aggregation during sample preparation. The size distribution of magnetic nanoparticles formed by chemical precipitation can often be fit by a log-normal size distribution6. the polymer size scales with XW to the 0. Chem. the median diameter is 7. F. Polymer hydrodynamic diameter vs number of monomer units in a chain. When a higher concentration is used (0. All co-polymer compositions (closed markers) agree well with published data for poly(acrylic acid) by Reith et al. The distribution is well described by a log-normal fit with an average size of 7. Chantrell et al. Due to the low contrast of the polymers relative to that of the carbon substrate. the particles showed no remanence when the field was removed. Wiegand.55 power and the polymer conformation is approximately that of a random coil. Muller-Plathe. polymers are compared on the basis of XW and not on molecular weight.. For the particles shown in Figure 4a-c. the size exponent is 0. such as styrene sulfonic acid.26. which assumes a log-normal size distribution of noninteracting particles. TEM cannot generally be used to ascertain the sizes of the clusters in suspension.12 wt% in (c)).26 Thus.06 wt% in (b)). obtained using DLS. indicating that the individual cores in the clusters were noninteracting. of a superparamagnetic suspension can be used to determine the size distribution of the particles within the suspension. Indeed. the cluster size observed on the TEM grid was constant. Charles.26 (open markers). VSM Characterization of Magnetic Properties.12 wt% in Figure 4c is most likely an artifact of the deposition process. the cluster size of 70 nm agrees well with DLS measurements.5 nm with a polydispersity of 0. while the more extensive clustering seen at 0. in subsequent analyses. 13. However. DLS was used for this purpose. J. 975-977. (d) Size distribution of the magnetite cores in (a)(c). The clusters observed in TEM measurements may simply be artifacts of the preparation method since the particles must be deposited on a grid and may be aggregated differently than when in solution. will be proportionally higher for the same XW than one composed of lighter monomers.med))2 2σc2 ] (3) where Rc. All data were taken in 1. Popplewell. 1978. 2005 6009 Figure 3.55. the end-to-end distance and radius of gyration are approximated by the known expressions for random coil polymers. consistent with the TEM data. The analysis. since the cluster sizes shown in Figure 4a and b for measurements made at low concentrations (<0. At these high ionic strengths. p(Rc) ) 1 2πσcRc exp [ -1 (ln(Rc/Rc..5 and polydispersity of 0. W. therefore. Throughout this paper. which may imply that no additional clustering or artifacts were introduced during the preparation of the TEM sample. 2002.med is the median core radius and σc is the core polydispersity. .Polymer-Coated Magnetic Nanoparticles Langmuir.26 as shown in Figure 3. The polymer system was chosen to allow independent tuning of hydrophobicity and attachment density through selection of the relative monomer fractions and of the molecular weight through the chain transfer agent concentration. in agreement with published data for poly(acrylic acid). Vol.27 showed that the field strength dependence of the magnetization. 14. B. S. Only the magnetic cores are visible. IEEE Trans. Magn. At high ionic strength. J. yields expressions for the median diameter (Dp) and standard deviation (σ):27 Dp ) σ) ( [ ]) ( ( )) χi 18kT πMd 3 MdH′ 3χiH′ 1 ln 3 Md 1/2 1/2 1/3 (4) (5) where χi is the initial magnetic susceptibility on a volume (27) Chantrell. 21. correlated well with the total number of repeat units in the backbone of the polymer (XW) for all compositions synthesized. (a)-(c) TEM micrographs of the magnetic nanoparticles at various concentrations. DLS measurements. such as vinyl sulfonic acid or acrylic acid.26 as shown in Figure 4d. 116. 9100-9106. below a certain overall particle concentration. Polymer Characterization. W.03 wt% in (a) and 0.06 wt%) agree well with (26) Reith. This superparamagnetism is important in applications using magnetic separation devices so that the clusters can easily be eluted from an HGMS column when the applied magnetic field is removed.. M.. S. At low concentrations (0.

7 Thus. For magnetic nanoparticles to be useful for charge-based separations.1 M. which usually have a saturation magnetization of 50-70 emu/g. The zeta potential of the particles shows strongly negatively charged surfaces for all particles tested over a wide range of pH. this nonmagnetic layer results in the magnetic diameter being significantly smaller than the TEM diameter. as they are stable to centrifugation at much larger sizes. The core magnetization of our particles is lower than the saturation magnetization of bulk Fe3O4. As the fraction of SSA in the polymer increases. Clustering of Nanoparticles The VSM and TEM results indicate that the magnetite cores produced by our synthesis method are typical of those made by chemical co-precipitation.e. and form much more viscous magnetic fluids at the same magnetite content. while lowmolecular-weight polymers yield dense clusters of magnetite nanoparticles. even at low pH. the zeta potential is less negative at low pH due to the protonation of acrylic acid. The salt concentration was relatively high. With further increases in molecular weight. is the volume fraction of particles. Clusters were synthesized with excess polymer to evaluate the effect of molecular weight (from 2000 to as high as 300 000. When the acrylic acid content is higher. and a leveling out of the curves at higher pH. given by the slope of the magnetization curve at H ) 0. The zeta potential indicates strongly negative charges. those not chemisorbed on the particle surfaces) of the poly(acrylic acid) with increasing pH below its pKa. these clusters are not useful for separation processes. It has been shown previously16-18. the relatively low magnitude of the Zeta potential (-30 mV) is due to the highly compressed double layer. Md is the saturation magnetization of bulk magnetite on a volume basis.12 but is typical of that for Fe3O4 nanoparticles. The clustering falls into three distinct regimes. and thus also its hydrophobicity. the size of the clusters is large. the resulting clusters are not sufficiently stable for many applications in which they are exposed to highionic-strength environments.4 nm and σ ) 0. Xmin increases since more hydrophobic coatings provide poorer stabilization and thus require a thicker coating to prevent aggregation. with small clusters exhibiting the greater stability. As the molecular weight increases. Figure 5. The cluster stability very closely mirrors the cluster size. these stable clusters are too small for efficient HGMS capture. the clusters decrease in size until a minimum is reached at Xmin. at 0. If these particles are to be useful for separations. as shown in Figure 6a. they must be recoverable. This phenomenon . the cluster size increases rapidly and the clusters approach a micrometer in size. it is desirable that they be strongly charged over a wide pH range. at this point. 2005 Ditsch et al.. the polymer begins to bridge between the particles and the clusters again begin to form. as shown in Figure 6b. attributed partially to enhanced steric stabilization due to the thicker polymer coatings on the primary particles. The median size obtained from this analysis was Dcore ) 6. while clusters large enough for capture are not stable to high salt concentrations. Thus. basis. varying molecular weight of a single coating is not sufficient.6. No. both for affinity for the targeted solutes and for stabilization of the nanoparticles.1 M NaCl. 13. Vol. consistent with the deprotonation of free carboxyl groups (i. Effect of Primary Polymer Molecular Weight on Cluster Size and Stability. there is a steady drop with pH in the magnitude of the surface charge below a pH of 5. The zeta potential of the same particles at low ionic strength is around -50mV at neutral pH. and thus. as was seen in our analysis. We will show in the following sections that several methods can be used to create clusters sufficiently large for capture. as shown in Figure 5. up to 800 nm as compared to 150 nm for clusters formed with small polymers.6010 Langmuir. where the charged particles can be used for protein separations. the clusters can be stabilized by later addition of a second polymer. Zeta Potential Results. The saturation magnetization of the fluids in the limit as H f ∞ can be determined from the intercept at 1/H ) 0 of a plot of M versus 1/H.12 The lower magnetization is attributed to formation of a nonmagnetic layer owing to disruption of the magnetic moments of atoms on the surface of the magnetite core. The zeta potential of particles as a function of pH for two acrylic acid compositions. and H′ is obtained from the M ) 0 intercept of a graph of M versus 1/H at high applied fields. did not significantly affect the ionic strength and to evaluate the surface charge at ionic strengths similar to those found in fermentation broths. For particles with 50% acrylic acid groups. This indicates that the large aggregates consist mainly of an extended network of hydrated polymer bridges. even in 0. possibly because all the carboxyl groups are chelated with the surface Fe atoms on the particles. with relatively low magnetite content. the particles are dispersed individually and do not participate in cluster formation. None of the clusters formed by using a single polymer coating was both stable and recoverable. an average value of 63 emu/g Fe3O4 was determined for the particles. with the maximum stability occurring with polymers slightly larger than Xmin. When the polymer molecular weight is small. The intrinsic stability of the nanoparticle clusters shown in Figure 6a was assessed by determining the critical NaCl concentration at which coagulation occurs. For particles with 25% acrylic acid groups. resulting in extremely stable clusters of any desired size. on the order of 200 nm.9. The clusters must also be stable to further aggregation and sedimentation for practical use. 87 emu/g. Once the polymer becomes sufficiently large that bridging predominates. and another method of clustering must be utilized. Effect of Polymer Composition on Xmin. XW from 15 to 2500) on clustering. However. When a good coating polymer that results in single nanoparticles is used in a limited amount during synthesis to form clusters.35. 21. Stability of Nanoparticle Clusters Formed With Excess Polymer.21 that such nanoparticles are too small to be recovered efficiently from the dispersion medium. These large bridged clusters have a significantly different morphology from that of the large clusters formed with small polymers.10 Usually. there is very little dependence of the zeta potential on pH. to ensure that the concentration of protons. but unless the clustering and stabilizing steps are decoupled. which makes clustering of the nanoparticles essential. The larger clusters formed at very low and very high molecular weights are not stable. Clusters formed near Xmin are stable in 5 M NaCl (the highest concentration tested).

(b) Critical coagulation concentration of particles in (a). A model based on this mechanism is developed later. where individual particles are stabilized under excess polymer conditions. (a) Cluster size as a function of polymer size (number of monomer units or Xw) in the polymer backbone for polymers of various styrene sulfonic acid content.e. stable particle suspensions are not formed. the latter as reflected in the SSA content of the polymer. we synthesized copolymers of acrylic acid (25mol%) with the cationic monomers vinylbenzyltrimethylammonium chloride (VBTMAC) and acrylamidopropyltrimethylammonium chloride (AATMAC). as shown in Figure 7b. 75% vinylbenzyltrimethylammonium chloride (VBTMAC) and 75% 3-acrylamidopropyltrimethylammonium chloride (AAPTMAC) with 25% acrylic acid. the data are represented approximately by a single curve. is examined in more detail below when we analyze the particle-particle interaction energies on the basis of simple theoretical models of the clustering processes. similar to the anionic particles in (a). presumably due to the larger polymer size and increased steric repulsion to coreto-core aggregation with incomplete coating. in the polymer. The results shown above were obtained for excess polymer added to the solution and depended only on polymer molecular weight and hydrophobicity. indicating that the observed clustering behavior is due to the hydrophobicity of the coating. clustering of these particles can be induced by limiting the amount of polymer used or by limiting the fraction of attachment groups. for copolymers of acrylic acid and VSA only). further indicating that the observed behavior is similar for all polymers tested..5.e. Vol. 21. the cluster size is a function of the ratio of the concentration of acrylic acid (chelating groups) to the concentration of surface Fe2+ and Fe3+ (chelated ions) and is independent of the fraction of acrylic acid relative to SSA and VSA on each polymer molecule. (c) Cluster size vs Xw for cationic polymer coatings. When the primary polymer molecular weight is near that corresponding to Xmin.as shown in Figure 6c. Secondary bridging particles are bridged by later polymer addition. The polymer size that results in the minimum cluster size. For polymers with some hydrophobic character. the cluster size starts at a smaller value.Polymer-Coated Magnetic Nanoparticles Langmuir. both strategies lead to incomplete coverage of the particle surfaces and hence some instability in the particle suspension to allow clustering to occur. Since more than the stoichiometric amount of polymer is required to stabilize individual particles. the cluster size decreases rapidly with increasing attachment group density for molar ratios less than 1. the relative ratios of which do not affect the clustering behavior under these excess polymer conditions for acrylic acid fractions over the range of 20-75%. Effect of Attachment Group Density. Xmin increases as the coating becomes more hydrophobic. When the results for all particle clusters are plotted in reduced form with D/Dmin as a function of X/Xmin to account for differences in Xmin. reaching the single particle size with 2. all particles fall on approximately the same curve. The remaining fraction of the polymer consisted of acrylic acid and vinyl sulfonic acid. As was observed above with variations in the polymer molecular weight. as shown by the dotted lines connecting the maximum stability to the cluster size in (a). as shown in Figure 6d. When no hydrophobic monomer is present (i. the latter observation is consistent with published reports9. The effect of Xw on cluster size for the cationic particles is similar to that for the anionic particles. i. The size of the clusters coated with more hydrophobic coatings decreases more slowly with addition of polymer. 13. VBTMAC is more hydrophobic and results in a higher Xmin. further indicating the generality of the observed phenomena. as shown in Figure 7a. The maximum stability of the particles occurs above Xmin and below where significant bridging has occurred.5 acrylic acid groups/surface iron ion. The stability of the clusters formed with limited amounts of polymer mirrors the particle size..0 and reaches the single coated particle size at a ratio of about 1. Within the range of 20-75% acrylic acid. To verify that the observed effect was due to hydrophobicity. (d) Cluster size relative to minimum size as a function of Xw relative to Xmin. increases as the coating becomes more hydrophobic. 2005 6011 Figure 6. acrylic acid. (when the fraction of acrylic acid is less than 20% or more than 75%. only single . the cause of clustering is not an insufficient number of attachment groups but rather the rate at which the particles are coated (which increases as the concentration of polymer increases) relative to the rate of core-to-core aggregation driven by attractive van der Waals forces. When the differences in Xmin are accounted for.15 that PAA itself is incapable of stabilizing particle suspensions). Xmin. No.

. 21. The importance of these results is that they show that we can decouple the fast clustering step and the coating step needed for stabilization of these clusters. At a molar ratio of 1:1 (i. (b) Stability in NaCl of clusters in (a). first. In the absence of the secondary polymer. These clusters are not. the clusters are not stable even to small changes in ionic strength. Second. The gray line is the predicted size for polymers with SSA (Xw ) 200). as shown in Figure 8b. Effect of Secondary Polymer Addition. . Thus. increasing molar ratio) and are completely stabilized at a total molar ratio of about 2:5.e. presumably due to bridging interactions. nanoparticles are stable when a single polymer coating is used in limited amounts. The lower curve is for initial clusters formed with the primary polymer only. which is much slower. Stabilization of Large Clusters by Secondary Polymer. there are vacant sites on the particle surfaces available for the attachment of the secondary polymer. and we can now see why multiple coatings are of value in the preparation of useful particles. would prevent this cluster formation.e. the clusters are stabilized but the size of the clusters remains the same or increases slightly. where the critical coagulation salt (NaCl) concentration is shown as a function of the total molar ratio of the acrylic acid (primary and secondary) to surface iron. that the initial clustering process is completed within 15 min. No. For particles to be useful in a wide range of applications. The cluster size is essentially unchanged by the addition of the secondary polymer when its molecular weight is small (<15 kDa. while all clusters larger than 50 nm aggregate further in high ionic strength environments. the cluster size increases. When the molecular weight of the polymer coating is varied. the cluster size must be larger than 50 nm and the resulting clusters must be stable. 15 min into the particle synthesis. and thus. which is easily changed by further polymer addition. closed symbols for polymers with no SSA.. otherwise the added low-molecular-weight polymer would occupy sites that would normally be responsible for partial instability and growth of clusters and. as the upturn in the curve in Figure 8a occurs near the XW indicated for the primary polymer. hence. However. the resulting clusters are not stable. Bridging occurs for all particles. the Requirement for Decoupling Clustering and Stabilization Steps. again showing the stabilization of individual particles at higher carboxyl/surface iron molar ratios seen in Figure 7a. Since the nature of the polymer cannot be easily changed after synthesis. (a) Cluster size and model fits for clusters formed with limited polymer near Xmin. These results indicate. The effect of the addition of low-molecular-weight (5 kDa) secondary polymer on the stability of 74 nm clusters is shown in Figure 9. the instability of the clusters formed with limited polymer is not due to the nature of the coating. the particles are not widely useful for separations. therefore.6012 Langmuir. Vol. However. poly(acrylic acid). 13.8kT. particularly suitable for use in separation processes performed in high-ionic-strength media. for large added secondary polymers. even those synthesized with a large ratio of acrylic acid to iron. We show below that this instability problem can be overcome through the addition of a secondary polymer to complete the coating of the surface sites left vacant because of insufficient primary polymer addition. For this bridging to occur. bridging can occur over long time frames. only individual particles could be synthesized that were stable at high salt concentrations. before the addition of the secondary polymer). the instability will also be difficult if not impossible to remove and clusters made with polymer far from Xmin will not be useful. 2005 Ditsch et al. the degree of polymerization of the secondary polymer should be greater than that of the primary polymer to ensure that the coatings on both particles to be bridged can be spanned by the secondary polymer. The open symbols are for coatings with 25-75% SSA. but when larger polymers are added. the clusters stabilized with two polymer additions are as stable as individual particles formed with an excess of polymer and retain the required cluster size for efficient capture. indicating that even when the individual particles are well-stabilized by the primary polymer. even when a large excess of polymer is added. The upper curve shows clearly that the addition of a high-molecular-weight (100 kDa) secondary polymer leads to the formation of clusters by bridging. the factors that affect clustering and stability have been outlined. However. which is difficult to change. Only the smallest clusters are stable. but these particles were too small to be captured by HGMS. i. when polymer is added after cluster formation.e.. XW < 200). In Figure 8a. The solid black line is a fit for polymers with no SSA (Xw ) 22) with Ea of 4. we show the effect on cluster size of the addition of a secondary polymer. but rather due to the amount of the coating. These clusters are all stable in pure water due to strong surface charges and the resulting interparticle electrostatic repulsion but not to moderate changes in ionic strength where charge stabilizing mechanisms break down. In the prior sections. The primary particles were synthesized with a 1:1 molar ratio of carboxyl groups to surface iron atoms. certainly after 15 min into synthesis. clusters are formed for XW substantially smaller or larger than Xmin. This is indeed the case. The stability improves with increasing amounts of secondary polymer added to the suspension (i. it would lead to a reduction in cluster size. as will be outlined below. making these clusters widely useful for separations. Figure 7. Cluster Stability and Requirement for Secondary Polymer We have shown above that there are many coatings that allow formation of clusters that are sufficiently large (>50 nm) to be captured by HGMS. allowing stable clusters of any desired size to be synthesized. for which some initial clustering occurs to give aggregates of 74 nm in diameter.

and the clusters thus formed are extremely stable. 21. Minimal changes in particle size are seen until the secondary polymer becomes larger than the primary polymer (primary Xw ) 85) above which bridging occurs.A1/2)2H m c s (d . but rather due to the amount of the coating. enhancing the effect of the electrostatic interaction contributions. Without any change in the materials used. steric interactions due to interpenetrating polymer coatings. With this information.Polymer-Coated Magnetic Nanoparticles Langmuir. which is difficult to change. we provide a method to predict the optimal size polymer for stability (Xmin) when excess polymer is used. With this insight. in this case. M. electrostatic interactions.1) + + 2δ s (A1/2 . 1. Finally.2δ . as well as that bridging can occur much later in synthesis than clustering.14 Thus. the magnetic dipole-dipole interactions are less than 1kT. In contrast to when more primary polymer is added.A1/2)2H . The main attractive energy for the coated magnetic nanoparticles is due to the van der Waals interactions. 1961. it is primarily the nature of the polymer coating that determines particle-particle interactions by modulating the van der Waals interactions between the cores. (b) Cluster sizes obtained with 100 kDa PAA as secondary polymer for polymers with various ratios of acrylic acid to surface iron. No.1 + 2(A1/2 .A1/2)(A1/2 c p m c c d s-δd+δ . Analysis of Clustering and StabilitysPrediction of Optimal Polymer The clustering behavior outlined in the previous section will now be examined by quantifying the particle interaction energies at the synthesis conditions. with critical coagulation concentrations in NaCl in excess of 5 M. we will outline a method based on DLVO theory that will allow prediction of Xmin from the solution structure and hydrophobicity of the polymer. with many below 0. When a large excess of primary polymer is added. 2005 6013 Figure 8. The first step in predicting the optimal polymer coating is to find the optimal polymer size when a large excess of polymer is used. The clustering behavior of the particles in the absence of bridging is dominated by the interparticle interactions. the utility of the particles can be greatly improved by a small change in synthesis method. and providing steric repulsion when the polymer layers begin to overlap. but the resulting clusters are also unstable. The total number of attachment groups required to form stable clusters is similar for particles formed with a large excess of primary polymer and for those where stabilization is by secondary polymer shown here. the instability is not due to the nature of the coating. However. clusters of controlled size are formed. we introduce methods to predict the size of clusters that result when a polymer coating is used that is too small (Xw < Xmin) or too large (Xw > Xmin). In this section. Figure 9. With all the models.14 For a core-shell structure. The mechanism for growth of these clusters indicates the need for a secondary coating to complete stabilization. J. and we find that it is advantageous to use multiple coatings to get clusters that are stable to a wide range of solution conditions. Stability of particles formed with secondary polymer added. indicating that bare surfaces are still present but to a lesser degree. allowing stable clusters to be made. the cluster size is not reduced when additional secondary polymer is added. This result verifies both that the particles are incompletely coated. and magnetic dipole-dipole interactions. it follows directly that the addition of a secondary polymer at a time after the clusters are formed should stabilize the clusters. Prediction of Xmin (Excess Polymer). with contributions from both the magnetite cores and the polymer coatings. If polymer with size of approximately Xmin is used in limited amounts. and will thus be ignored in this analysis. In the absence of an applied field. 13. A1/2)H (6) p d d ( ) ( ) where (28) Vold. we postulate a mechanism for cluster growth in the presence of limited polymer. allowing us to predict how much polymer must be present at nucleation to get a desired final cluster size. while clusters formed by other methods all have critical coagulation concentrations in NaCl below 1 M. All these interactions can be estimated as a function of the separation distance of the particles. the van der Waals interaction energy is given by28 -12uvdw ) (A1/2 . the cluster size still increases but to a lesser extent than when less primary polymer is added. For polymer-coated magnetic particles. Particle Interaction Energies. which can easily be changed by adding more polymer. 16. allowing bridges to form. Then.1 M. . Colloid Sci. The mechanisms used in predicting the size of clusters with nonoptimal polymer coatings provide explanations for the relative instability of these clusters. the primary forces are the long-range van der Waals interactions. without significantly changing the cluster size. (a) Particle size vs size of poly(acrylic acid) secondary polymer for aggregates starting at 74 nm. J. This is in fact the case. we are able to predict both the size and relative stability of clusters resulting from different polymer coatings. Vol.

7 M. M. For polymers with hydrophobic domains. when δ is limited to a maximum of 7. due mainly to increased electrostatic repulsion with larger particle diameter. and the iron oxide core diameter is d. the Hamaker constant of the polymer is not the same as that of the aqueous medium. such as the aromatic rings on styrenesulfonic acid. J. Adv. increasing the layer thickness results in increased stability. however.8 × 10-21 J) and because the increasing size increases electrostatic repulsion. and Am represent the Hamaker constants for the particles. E. each AAMTMAC repeat unit was assumed to be one-half of an SSA unit from the propyl group. Overbeek. 13. TH.5 nm. ψ∞ ) -50 mV.poly consisting of a binary mixture of n styrene molecules (number of SSA repeat units in the backbone) in a volume of water equal to the volume of the sphere. The Hamaker constant is estimated by assuming that the polymer is a sphere with radius Rg. C. Examples of the interaction energy profiles calculated using eqs 6 and 9 for polymer layers of different thickness. The energy of interaction between two charged spheres can be estimated using29. the fit fails at higher hydrophobicities (dashed line). S. Theory of the Stabilization of Lyophobic Colloids. the steric interactions of the coatings are no longer important and the problem reduces to a standard DLVO balance between van der Waals and electrostatic interactions.31 In the example given in (31) Vold. 2005 Ditsch et al. J. and because the increasing size leads to increased electrostatic repulsion. are shown in Figure 10b. 1983 . Figure 10. J. near that estimated for 75% SSA. 1948. Vol. so that only the interactions of the magnetite cores will contribute to the attractive force between the particles and the interaction energy will reduce to s -12uvdw ) (A1/2 .. When no coating is present.8 × 10-21 J). In this case. both because the polymer has a lower Hamaker constant than magnetite (4. is the dielectric constant. and medium. 47. respectively. as shown see Figure 10a. there is little barrier to aggregation. κ is the inverse Debye length. the polymer coating thickness is δ. H(x. In this case. Thus. If no δmax is used to account for curvature of the particles. The Hamaker constants for the polymer were estimated by assuming that the polymer is a binary mixture of water and polystyrene. Colloid and Interface Chemistry. For the cationic polymers. AddisonWesley: London. and VBTMAC was assumed to be seven-sixths of an SSA from the extra methyl group. No. the Hamaker constants for the coating and the medium are essentially the same. When there are hydrophobic domains in the polymer. W. G. Ionic strength ) 0. (30) Verwey. but also due to reduced van der Waals interactions because of the lower Hamaker constant of the polymer relative to that of the magnetite core. J. (b) Interaction energy for attracting polymers vs δ (constant Hamaker constant A ) 2.1 c p d ( ) (8) The particles are coated by polyelectrolytes and thus will be electrostatically repulsed. Radke. the Hamaker constant for the polymer layer is different from those for both the cores and the suspension medium. E. there is little barrier to aggregation. D. The layer thickness required for stabilization of the individual nanoparticles against agglomeration is that for which the energy barrier is at least 15kT. the Hamaker constants were evaluated on the basis of the number of hydrophobic carbons in the cationic repeat unit relative to those in SSA. the activation energy to aggregation increases. and 0 is the permittivity of free space. Layer Thickness Required for Stability.3 × 10-22 J).5 nm. When no coating is present. As the layer thickness increases. Elsevier: New York. The layer thickness is assumed to be 2Rg.3 × (29) Reiner. Vold. whichever is smaller. coatings.2δ. (c) Comparison of Xmin for several compositions of styrenesulfonic acid (closed markers) and for the SSA fraction that would result in the same hydrophobicity for cationic polymers (open markers) with the Xmin required for an energy barrier of 15kT with Hamaker constant and δ of that polymer estimated as shown by the solid line. having an effective Hamaker constant of A ) 2. R. The particle core surface-to-surface distance is s.. both because the polymer has a lower Hamaker constant than magnetite (4.6014 Langmuir. ψ∞ is the surface potential of an isolated particle (approximately equal to the zeta potential of the particle). the activation energy for aggregation increases. where the total volume is the volume in a sphere with a radius equal to the radius of gyration of the polymer and the number of styrene molecules equal to the number of styrene sulfonic acid units in a single polymer. assuming a constant interfacial potential of ψ∞ ) -50mV. When the attractive forces dominate before the coatings can touch. δ.poly or 7. with the balance water. Colloid Interface Sci. As the layer thickness increases. 59-147. the van der Waals attractive force becomes strong when the polymer coatings come in close contact..y) ) y y + 2 + x + 2xy + x x + 2xy + x + y x2 + 2xy + x 2 ln 2 (7) x + 2xy + x + y 2 and Ap. 10-22 J. 1993. (a) Schematic of method used to estimate layer thickness (δ) and the Hamaker constant for the coating.A1/2)2H . 21. For hydrophilic polymers.30 ue ) 2π ∞ 2 0a(ψ ) 2 exp(-κh) 2 + (h/a) (9) where h ) s . Ac.

we can estimate the values of Xmin that will give energy barriers of 15kT for different polymer compositions.θ) e-Ea/kTMP dt (15) where RM is the radius of the aggregate core. The maximum stability of the particles occurred with molecular weights higher than the minimum required for stabilization of individual particles during synthesis. Since the mechanism indicates core-to-core contact. if the coating is not complete. . at the pH of synthesis (around 13-14). particularly for hydrophobic polymers since. When the polymer is larger than Xmin. Thus. The rate of particle aggregation in such a system will be24. 21. η is the viscosity.0 for diffusion-limited colloidal aggregation. Adv. while no aggregation occurs if either surface is completely coated. for practical applications in which the ionic strength is higher than that used in the particle synthesis. 2005 6015 Figure 10b.poly)2(DP + DM)(1 . the required layer thickness for stability is about 8 nm. the resulting probability is PM ) (1 . Colloid Interface Sci. and DM and DP are the particle and polymer diffusion coefficients.32 We therefore assume that when RG > 1/2DP.25 dM 4kTM2 )PM dt 2η ( ) (11) where M is the total number of particles in a given volume. the polymer size that yields individually coated single nanoparticles that are stable against aggregation.Polymer-Coated Magnetic Nanoparticles Langmuir. either because it is already covered with polymer or because it is blocked by other particle cores in the cluster. then θ can be calculated from the rate of free polymer adsorption: np dθp dP )dt dt nmM0 ( )( ) (13) where P is the number of polymer molecules in a given volume. With the preceding equations. No. If we assume that the particle has nm binding sites. 13. (33) Adachi. this layer thickness will be less for more hydrophilic polymers with Hamaker constants nearer that of water.32 For thicker layers. M. Vol. owing to smaller van der Waals interactions. and the fractional coating is equal to the fraction of binding sites blocked. for polymer adsorption. while these particles may be stable at the synthesis conditions of 0. 56. is M0/M. that each polymer can block nP sites. the core of the cluster will have a size of DH ) DP ( ) M0 M 1/DF + 4RG.20. the curvature of the particle will limit the maximum effective thickness of the coating since the coating far away from the surface will tend to be diffuse and provide little stability. These estimates agree well with the experimentally determined values of Xmin for a range of polymer compositions expressed in terms of the SSA content or. the number of particles in a given volume can be calculated. The fractional particle surface inaccessible to polymer includes both the binding sites blocked by other cores and those already occupied by the polymer. of aggregation less than unity. Prediction of Cluster Size with Limited Polymer (X ) Xmin). and can be approximated by θ ) θP + 1 M 13 M0 ( ) (14) The rate of polymer disappearance will be related to the diffusion-limited rate of collisions between magnetite cores and polymer molecules by33 dP ) -4π(RM + RG. If we ignore the kinetics of nucleation and assume that the cores are initially 7. for the cationic polymers.. With the above assumptions on the composition dependence of the Hamaker constants and on the effective layer thicknesses. the barrier for aggregation of uncoated particles is low and the aggregation would be expected to be limited only by diffusion and the fractional free surface available to accommodate more polymer adsorption. δmax ) 7. owing to an increased screening of the interparticle electrostatic interactions. the maximum effective layer thickness is simply the diameter of the particle. which may be useful as a starting point for evaluating the potential efficacy of other polymers for use in stabilizing nanoparticle suspensions.5 nm. δ should be larger than that required for particle stabilization under synthesis conditions. expected to be around 2. respectively. All that is needed to calculate the cluster size is to note that the aggregation number. However. i.5 nm uncoated magnetite spheres. For this purpose. Effect of Molecular Weight on δ and Determination of Xmin. and decreases with increasing ionic strength. M0 is the number of magnetite cores initially in the control volume. the dependence of δ on the polymer molecular weight must be known.θ)2 (12) where θ is the fractional particle surface inaccessible to the polymer. Y. The interaction energy profiles calculated above can be used to estimate the value of Xmin. and PM is the probability of a collision resulting in aggregation. The layer thickness typically varies as twice the radius of gyration. the equivalent SSA content. δ = 2RG. Raphael. The height of the energy barrier is a strong function of the ionic strength of the medium. both magnetite and the polymer are negatively charged and there will be an activation energy barrier.. however. we are able to predict the molecular weight required for stabilization on the basis of only the composition of the polymer and the Hamaker constants of the components. based on the degree of coating of the particles.7 M. i. Thus. modified with a probability (32) Aubouy. 1-31. Macromolecules 1998. RG. and θP is the fraction of sites occupied by adsorbed polymer. E.. On the basis of the particle-particle interactions of well-coated particles.poly (16) where DF is the fractal dimension of the cluster. 4357-4363. At the same ionic strength. it is expected that coated surfaces will not aggregate.e. for relatively thin polymer layers. Since magnetite has a pI of about 6. If a collision of two bare surfaces results in aggregation. where nP is the total number of carboxylic acid groups on the backbone of the polymer. 31. or the number of cores in an average cluster. a thicker coating is required for the particles to be stable at 5 M. and report these results in Figure 10c. however. we postulate a model for the clustering of particles in the presence of insufficient polymer.Poly is the radius of gyration of the polymer. Ea. we can analyze the rate of aggregation as diffusion-limited colloidal aggregation (DLCA). 1995.e. This simple result indicates that the clustering below Xmin is due to an insufficient layer thickness and contributions from the polymer layers to the overall van der Waals interactions between the particles. but small enough that bridging is limited.

. secondary polymers that can access the uncoated sites on the particle surfaces. In the previous sections. an optimal molecular weight polymer. No. T. 21. and are stable over a wide range of ionic strength and pH. we provide models to predict the cluster size both below and above Xmin. 2005 Ditsch et al. the size of a primary particle. The resulting clusters are not stable but can be stabilized with the addition of smaller. An example of the model predictions for the three different regimes is given in Figure 11a.. In this section.. The three relevant sizes are the size that results in an energy barrier of 15kT. Vol. This method should be useful generally in producing magnetic nanoparticle clusters for a wide range of uses. An important result of the simulations is that the cluster size is predicted to be constant after a few milliseconds. The size of the clusters can be estimated with DLCA kinetics modified with a sticking probability based on fractional coating. A. where it is shown clearly that the system jumps from one regime to the next as the polymer molecular weight increases.8kT. either larger or smaller than Xmin. when larger than 50 nm can be captured in HGMS. The two parameters that are required to make the desired size and stability are Xmin and the amount of polymer in the first coating. The models presented above were developed to predict the optimal polymer size (Xmin) and the amount of polymer at Xmin that must be present at nucleation to get the desired cluster size. have also been predicted. near Xmin. this result indicates that the cluster size is set well before the addition of the second polymer at 15 min. The single adjustable parameter in the curve fitting was the activation energy for the polymer-core interaction. With this information. I. both data sets are fit semiquantitatively with only θ as a global fit parameter. A. which was found to be 4. Complete Clustering Model for Excess Polymer Addition. particularly considering the simplicity of the models used to describe these systems. C. D. Although the models are significant oversimplifications of the actual clustering behavior.34 Additionally. The resulting particles can be made to a desired size. low-molecular-weight polymer for stabilization. the cluster sizes obtained using nonoptimal polymers. both below and above Xmin. as shown by the bold line. Following the procedures outlined earlier we are able to predict the size of clusters formed with polymers smaller than Xmin. Although of less importance. These equations were solved for different polymer combinations and fit to the experimental data shown in Figure 7a. but the relative instability of the clusters formed by such methods can be explained.: Yin. Conclusions We have introduced a new class of magnetic nanoparticles with controllable cluster size and extreme stability in high-ionic-strength media. P. (See Appendix for details of these calculations) The size of the coating can be approximated as twice the radius of gyration of the polymer. providing further verification of the need for decoupling of the coating and stabilizing steps. The optimum molecular weight (Xmin) can be estimated using standard DLVO analysis with an estimated polymer Hamaker constant based on the monomers used and the size of the polymer in solution. Both of these approaches have relied on minimal fitted constants and should be useful for many polymers. not only can the cluster size and morphology resulting from a coating of arbitrary polymer be explained and predicted. When a larger polymer molecule coats the surfaces. The fit of the model is within experimental error for all but the largest clusters formed with the larger. the model assumption that clustering occurs until θ is unity will likely break down since a partially coated surface will limit aggregation due to steric effects. Appendix Effect of Nonoptimal Polymer on Cluster Size (Excess Polymer). Wang. more hydrophobic polymers. In the Appendix. and (iii) bridging occurs above Xmin. 13. (b) Predictions for 0% and 75% SSA. The largest of the three sizes will be the actual size. The (34) Ditsch. but clusters of a wide range of hydrophobicity and of both positive and negative charge could easily be made by decoupling the clustering and stabilization steps. (ii) individual particles are stabilized by polymers. was used in limited amounts to get the desired cluster size followed by addition of a secondary hydrophilic. allowing rational design of magnetic nanoparticle clusters for specific applications. with only one adjustable parameter. J. E. in preparation Figure 11. Laibinis. Along with the clustering model for limited polymer at Xmin and these models for clustering away from Xmin. and the size obtained when bridging occurs. The use and optimization of these particles for use in purification of recombinant proteins from fermentation broth is discussed elsewhere. we derive models for the three specific cases when excess polymer is added and (i) clustering occurs below Xmin. Even if the time progression is off by several orders of magnitude. P.6016 Langmuir. Both of the mechanisms developed indicate that the stability of clusters made with polymers larger or smaller than Xmin will be strongly dependent on ionic strength. Cluster Size Prediction below Xmin. Clusters of useful size and stability could not be synthesized with the use of a single polymer coating of any composition or amount. Hatton. we developed models to predict the optimal polymer size (Xmin) and the amount of polymer at Xmin that must be present at nucleation to get a desired cluster size. (a) Example of the predicted sizes below and above Xmin. To obtain particles that are both stable and can be captured using HGMS. the cluster sizes for a wide range of polymer coatings can be predicted with only Ea and θ as fit parameters. ment is remarkable. The agree- . The general method and equations should be applicable to a wide range of polymers. The model predictions are compared with experimental data in Figure 11b for two polymers with different SSA content. it is also of interest to understand the effects of excess polymer concentration at all values of X..

while the secondary coating will tend to have a more extended structure and will dominate the outer layer of the coating. we can calculate the cluster size required for the barrier to reach 15kT. In the previous sections. and the depth of penetration normalized to the primary particle size is Dpen ) Dpoly . In such a case. Several expressions exist for the size of clusters formed by bridging. Rev. while the core-to-core model predicts that coating will occur over a time scale of milliseconds. The mechanism developed here is strongly dependent on the ionic strength of the medium. the clusters formed by small polymers (below Xmin) and limited polymer have been described. No. La Mer. 2003. 2005 6017 interaction energy between fractal clusters and individual nanoparticles has been analyzed by by Shen et al.. . 62.38 However. the polymer adsorption has no significant energy barrier and coagulation occurs on a time scale similar to DLCA. bridging occurs after coating is nearly complete since the ratio of polymer to colloid is much higher than in the typically modeled case and is characterized by a sizable activation energy for adsorption. Vavylonis. W. we have assumed that bridging stops when the probability of a bridge forming during a collision is consistent with an energy barrier of 15kT. Trans. We apply their analysis to the case of our particles with thin polymer coatings (XW < Xmin) The variation of the electrostatic interaction energy with interparticle distance. Vol. Thus. Dpoly is the polymer diameter. H. we can approximate the van der Waals interactions by treating the particles and clusters as uniform spheres of constant effective Hamaker constant. both a thick coating and a reduced Hamaker constant can be combined in a single particle. This penetration is illustrated schematically in Figure 12a.. we can also predict the cluster size due to bridging.14 who assumed that the clusters could be treated as equivalent spheres with constant surface potential and averaged Hamaker constant based on the fractal dimensions of the cluster. 2Rp Rp ) (A-2) and where we assume that we can approximate the van der Waals interactions through a Hamaker constant averaged over the particle cluster according to Aeff ) (ApAcl)1/2 ) fAp (A-3) where f is the volume fraction of particles in the cluster. Healy. Faraday Soc. Colloid Sci.33.. according to 12uvdw ) AeffH ( s Rcl . K. δ is the thickness of the inner stabilizing layer. Substituting the new expressions for the interaction energy of a cluster and primary particle and using the expressions for δ and the Hamaker constant from the earlier discussions. The probability consists of two parts: (i) that a polymer attached to a primary particle can contact a magnetite core in a cluster and (ii) that the polymer-magnetite collision will result in adsorption. which accounts for the finite cluster sizes formed by primary particles coated with polymers smaller than Xmin. Fuerstenau. With the insights of the kinetics of polymer adsorption from the limited polymer clustering section. Aeff. D.36 lowering the coating Hamaker constant. As in the clustering model for polymers smaller than Xmin.. T. 13. When the cluster size becomes sufficiently large. the existing models are designed to explain the behavior when large polymers (on the order of 106 Da) of an opposite charge to a colloid adsorb and cause flocculation. (38) Smellie.5). 1958.δ ) Dp 2Rg. D. the electrostatic interaction energy increases. T. The effective Hamaker constant for the interaction of the particle with the cluster depends on the particle density within the cluster. as evidenced by a later transition to bridging. 90. Colloid Sci. The larger size increases the electrostatic repulsion of the cluster for a single particle.Polymer-Coated Magnetic Nanoparticles Langmuir. W. R. J.. Lett. The much slower rate of bridging relative to coating in our system is verified experimentally by the observation that bridging can be induced readily by the addition of a secondary polymer 15 min into synthesis when a sufficiently large polymer is used. Similarly. R. while the open structure reduces the effective Hamaker constant. 1964. the amount of polymer in the system is much less than the amount of colloid. Phys. Thus. La Mer. 13. None of these conditions is true in the case examined here. The probability that a collision of a particle and a fractal aggregate will result in the extended polymer reaching a magnetite core is equal to the number of layers that the polymer tail can penetrate times the volume fraction of magnetite in that layer.(2Rg. The hydrophobic primary coating results in a thicker coating. the activation energy barrier to adding another particle becomes prohibitively high to permit further growth. V. The importance of the layer thickness and hydrophobicity on the stability of the particles is evident in the extreme stability of clusters formed with a hydrophobic primary coating and a hydrophilic second(35) Hogg.35 by ue ) 4π 0 ( RpRcl (ψ∞)2 ln(1 + exp(-κs)) Rp + Rcl ) (A-1) where Rcl is the radius of the cluster and Rp is that of the primary particle. since bridging is most efficient with partial coating (θ ) 0. (37) Healy. V. We then find the cluster size by predicting the probability that a bridge will form during a collision. J. which can be estimated in terms of the fractal dimension through the equation f ) (Rp/Rcl)3-DF. 21.poly(Xmin ) + κ-1) (A-4) Dp where Dpen is the depth of polymer penetration. indicating that coagulation in a high-ionicstrength medium is to be expected and that ultra stable clusters cannot be formed by simply using a polymer smaller than Xmin. 859.poly(XW) . such as in wastewater treatment. and in particular on the particle density near the cluster surface. which were in good agreement with dynamic light scattering. B. W. small angle neutron scattering. If we assume that the polymer size remains the same as in solution the length of the tails will be 2Rg. K. Cluster Size from Bridging (X > Xmin). s. a new model for bridging which accounts for the size of bridged clusters formed with nearly completely coated particles needs to be formulated to explain the observed bridging behavior. can be approximated by a simplified form of the equations given by Hogg et al. which we have approximated as the layer thickness (36) O’Shaughnessy..37.poly. 1966. and cryo-TEM studies. ary coating. 19. 323. 1638. As Rcl increases. They assumed that clusters grow until the energy barrier for aggregation exceeds 15kT and were able to estimate the cluster sizes of surfactant-bilayer-coated nanoparticles. as the void spaces are filled with solvent and do not contribute to the particle/cluster interactions.

0 and that the primary particle size is the diameter of the magnetite core plus 2Rg. If we assume a fractal dimension of 2. This probability equals All parameters in eq A-8 have been estimated above with the exception of θ.. κ-1.CSME) for their assistance in VSM measurements. i.poly < δ is desired. The observation that even when a large excess of polymer is used there is still incomplete coating explains the lack of stability of bridged particles. and Sunil Jain and Mariam Kandil for their assistance in particle synthesis. and for both anionic and cationic polymers. θ is expected to be a value near to.θ) e15-Ea/kT ) 1/(3-DF) (A-8) at Xmin plus the electrical double layer thickness. The exact value of θ is not easily estimated. CSME) and Caroline Ross (MIT. that the penetrating polymer tail reaches a magnetite core is equal to the volume fraction of particles at the outer surface times the volume fraction of magnetite in a primary particle or Pmag ) DF Dcl 3 Dp ( ) DF-3 φpDpen ) DF Dcl 3 Dp ( ) ( ) DF-3 Dm 3 Dpen Dp (A-5) where Dcl is the cluster diameter. LA047057+ . particularly with polymers bound to multiple sites.δ)(1 .5 nm. although sufficient polymer has been added to coat the entire magnetite surface.8kT. The probability. Vikram Sivakumar (MIT. and Dp is the diameter of a primary particle.6018 Langmuir. 13. The probability of a polymer attaching to a core in the aggregate increases as the polymer size increases. Cluster diameters for particles formed with all polymers having polymerization degree greater than Xmin are shown as a function of the polymer penetration depth in Figure 12b together with the predicted behavior for three different values of θ. (b) Size data used to fit θ. but slightly less than. corresponding to a layer thickness of 7. it would be unfavorable entropically for all sites to be bound. Vol. are correlated well by eq A-8 when the value of θ is 0. To ensure longterm stability.θ) e-Ea/kT (A-6) We assume that the activation energy for bridging is the same as in the clustering model in the previous section. Acknowledgment. and decreases as the aggregate grows. the mechanism indicated seems to be general and useful for the range of polymers studied here. 2005 Ditsch et al. No. (A-7) Dcl ) 1 ((3)D φ (D F p poly . The remaining component of the bridging probability is the probability of adsorption after the polymer has come into contact with a magnetite core. The x axis is made up of all the terms in eq A-8 that differ from polymer to polymer. unity since.poly up to a maximum size of 22.. φp is the volume fraction of magnetite in a primary cluster. Thus. the data for all composition ratios. 4. so that bridging cannot occur. We would like to thank Yuki Yanagisawa for his help in TEM measurements.. and it is used as a global fit parameter on the assumption that it is the same for all polymer compositions.5 nm. allowing a single fit to all data. 21. i.98. particularly in high-ionic-strength media. that of a binding site being open times the activation energy of adsorption or PAds ) (1 . Clearly. Combining the probabilities derived above and setting them equal to the total probability equivalent to an energy barrier of 15kT. the use of polymers where 2Rg. (a) Schematic of bridging between a fractal aggregate and a single particle.e. This work was supported by the DuPont MIT Alliance.e. Pmag. then we can calculate the cluster diameter directly. letting PAdsPmag ) e-15 we derive Figure 12. and Dm is the diameter of the magnetite core. with the relevant length scales indicated.