A Direct and Continuous Supercritical Water Process for the Synthesis

Jul 10, 2015 - Continuous Hydrothermal Synthesis of Inorganic Nanoparticles: Applications ... Reaction Conditions Mimicking Deep-Sea Hydrothermal Vent...
0 downloads 0 Views 4MB Size
Article pubs.acs.org/IECR

A Direct and Continuous Supercritical Water Process for the Synthesis of Surface-Functionalized Nanoparticles Robert I. Gruar,† Christopher J. Tighe,† Paul Southern,§,∥ Quentin A. Pankhurst,§,∥ and Jawwad A. Darr*,† †

Christopher Ingold Laboratories, Department of Chemistry, University College London, 20 Gordon Street, London WC1H 0AJ, U.K. ‡ Department of Chemical Engineering, Imperial College London, South Kensington Campus, London SW7 2AZ, U.K. § UCL Healthcare Biomagnetics Laboratory, University College London, 21 Albemarle Street, London W1S 4BS, U.K. ∥ Institute of Biomedical Engineering, University College London, Gower Street, London WC1E 6BT, U.K. S Supporting Information *

ABSTRACT: A new processing methodology is presented for the direct synthesis of surface-functionalized nanoparticles through modification of a single-step continuous supercritical water process. The processing methodology utilizes inexpensive metal salt precursors that form nanoparticles upon mixing the metal salt solution with a supercritical water flow (24 MPa and 450 °C). Surface functionalization is achieved through introducing a supplementary flow of capping agent (citric acid in this example) to the stream of nascent (newly formed) nanoparticles using a novel reactor design. It was found that certain process attributes were key to effective functionalization of the nascent nanoparticle stream, and that high grafting densities of the capping agent were obtained in a relatively narrow process window. We have also used the core design of the reactor to devise and test a scaleup methodology to produce large quantities of surface-functionalized nanoparticles. A method for scaling-up the reactor is described, using a newly developed pilot plant designed to process flow rates 20× greater than the equivalent laboratory-scale process, which yields products at rates of ca. 1 kg/h (effectively semi-industrial-scale production). The method enables large-scale production without recourse to expensive or environmentally damaging reagents and uses water as the only process solvent, a significant advantage over many methods commonly used to produce surface-functionalized nanoparticles. We report the synthesis and characterization of citrate-functionalized Fe3O4 nanoparticles as a model system and present detailed characterization of the materials obtained at both processing scales.



agglomeration.13 However, many of the single-step process suffer experimental limitations such as long reaction times, high production costs (associated with reagents), and low production rates, and they are almost exclusively batch-type processes.1 Although well reported, two-step synthesis methodologies are generally not favored, as the state of particle agglomeration prior to surface functionalization is difficult to control, albeit in some cases it is the only route available to produce dispersions of metal oxide particles due to the high temperatures required for the formation of crystalline phases.14,15 The two-step reactions often use particles produced using more conventional techniques, and a subsequent dispersion and formulation protocol is used to functionalize the surface of the particles. A major drawback of the two-step method is that particle agglomeration is difficult to control and often yields poor dispersion reproducibility.13 Various efforts including ultrasonication, changing pH, using dispersants/ surfactants, and chemical surface modification have been extensively investigated in efforts to improve the dispersion properties.16−21 However, in many cases the poor reproduci-

INTRODUCTION Surface-functionalized nanomaterials can be produced using a number of synthesis procedures, including co-precipitation reactions, microemulsion routes, polyol processes, thermal decomposition reactions, and hydrothermal reactions, all of which have been reviewed in great detail previously.1−6 Many of these processes allow for the addition of capping agents either during particle formation reactions or shortly after the formation of nanoparticles, and correspondingly many materials obtained from these reactions have been studied as colloidal dispersions.5−7 There is growing interest in the development of synthetic methods which produce dispersed nanoparticles, as agglomeration leads to difficulty in targeting specific applications.2−4 There are a number of potential applications of nanoparticles which require high-quality dispersions to fully exploit the unique properties of nanoscale materials. Magnetic nanomaterials are a class of materials which show significant variation in material properties as a function of particle size, dispersion, structure, and method of manufacture and are of specific synthetic interest.8−12 Of the many processing methods used for the synthesis of surface-functionalized nanoparticles, single-step processing methods are favored, as they allow for the direct synthesis of high-quality dispersions. Particles can be grown in the presence of the capping agent, or the capping agent can be added shortly after particle formationone objective being to limit © 2015 American Chemical Society

Received: Revised: Accepted: Published: 7436

May 18, 2015 July 1, 2015 July 10, 2015 July 10, 2015 DOI: 10.1021/acs.iecr.5b01817 Ind. Eng. Chem. Res. 2015, 54, 7436−7451

Article

Industrial & Engineering Chemistry Research

Table 1. Summary of the Reaction Conditions and Characterization Data for Citric Acid-Coated Magnetite Produced Using Continuous Hydrothermal Flow Synthesisa DLS

TEM

ID

Qq (mL min−1)

CA (wt%)

CA:Fe

RT2 (s)

CS (nm)

weight loss (%)

Dh (nm)

PDI

Dh(n)

mean CS (nm)

SD

CA/nm2 b

1.5CA2.5 3.1CA2.0 4.7CA1.7 0.7CA5.2 1.5CA4.2 2.3CA3.4 1.5CA5.2 3.2CA4.2 4.7CA3.4 3.1CA5.2 6.3CA4.2 9.4CA3.4 1.5CA7.9 3.1CA6.3 0.7CA2.2c 1.1CA2.0c 0.7CA2.2c 1.5CA2.2c 2.3CA2.0c 3.1CA2.2c 4.7CA2.0c

10 20 30 10 20 30 10 20 30 10 20 30 10 20 20 30 10 20 30 20 30

1.0 1.0 1.0 0.5 0.5 0.5 1.0 1.0 1.0 2.0 2.0 2.0 1.0 1.0 0.5 0.5 1.0 1.0 1.0 2.0 2.0

1.5 3.1 4.7 0.7 1.5 2.3 1.5 3.2 4.7 3.1 6.3 9.4 1.5 3.1 0.7 1.1 0.7 1.5 2.3 3.1 4.7

2.5 2.0 1.7 5.2 4.2 3.4 5.2 4.2 3.4 5.2 4.2 3.4 7.9 6.3 2.2 2.0 2.2 2.2 2.0 2.2 2.0

12.8 11.8 11.3 13.4 12.8 13.5 11.8 12.9 11.7 12.4 10.9 11.6 12.8 11.5 9.2 8.7 7.4 8.6 7.4 9.2 8.2

4.42 4.22 2.79 7.10 7.37 7.72 10.35 8.64 8.11 8.98 5.05 5.95 7.2 8.23 10.15 8.64 5.52 15.96 22.36 15.76 5.59

79 78 80 100 106 72 74 72 74 73 75 77 77 86 102 95 91 73 75 79 72

0.132 0.19 0.155 0.215 0.178 0.149 0.152 0.193 0.163 0.165 0.21 0.169 0.17 0.189 0.295 0.32 0.178 0.232 0.143 0.183 0.149

43.2 46.4 54.3 23.5 22.1 17.6 17.9 20.6 25.3 13.9 18.5 22.3 26.7 25.0 16.7 18.4 20.2 13.4 16.7 14.6 19.8

13.7 16.6 11.7 11.0 13.4 13.5 15.8 11.2 11.1 14.4 5.64 10.16 8.84 8.94 5.60 8.60 6.70

3.3 3.3 3.1 2.6 4.1 3.5 4.2 2.6 2.6 3.7 1.28 3.00 3.28 3.09 1.20 3.09 2.53

2.08 2.64 2.50 1.96 2.24 2.50 1.64 1.37 1.64 2.44 1.17 1.80 1.00 2.93 2.57 2.78 2.3

The ammonium iron citrate concentration was fixed at 0.066 M, and the heater temperature was set to 450 °C, resulting in a reaction point temperature of ca. 380 °C (Tmix1) at the flow rates used, Qsw = 10 mL min−1 and Qp = 5 mL min−1. Residence times were calculated from the reaction zone length and the velocity of the fluid at the theoretical mixture temperature (i.e., Tmix1 or Tmix2). Samples are identified by the ratio of CA:Fe used in the synthesis and the residence time post citric acid addition; e.g., a sample produced with a CA:Fe ratio of 1.5 and a residence time post CA addition of 2.5 s is identified as 1.5CA2.5. Qq is the flow rate of citric acid solution, RT2 is the residence time between the inlet of the quenching flow and the cooling apparatus, CS is the crystallite size, Dh is the hydrodynamic diameter measured using DLS, PDI is the polydispersity index, and Dh(n) is the number-weighted distribution. bGrafting density of CA (molecules per square nanometer) determined using eq 1 and the crystallite size determined using TEM. cReactions performed at higher volumetric flow rates of preheated water (Qsw = 25 mL min−1) and precursor flow (Qp = 10 mL min−1). a

bility of two-step methods for the synthesis of nanoparticles is far superseded by the reproducibility of single-step methods. In relation to the scale-up of procedures for the synthesis of surface-functionalized nanoparticles, comparatively few methods have been proven to produce surface-functionalized nanoparticles on a sufficiently large scale to allow for broad application or thorough evaluation. That said, several highlighted papers have demonstrated that batch sizes of ca. 40 g can been obtained.2,5 Broadly, the scale-up of surfacefunctionalized nanoparticle synthesis is gaining both academic and industrial interest as the utility of these materials is further explored.22,23 The stability of a nanoparticle dispersion can be controlled by several mechanisms, and the particles may be either charge stabilized (electrostatic) or sterically stabilized (physical separation). Often, the surface functionalization and stabilization strategy employed is application dependent, and is entirely dependent upon the chemical moiety grafted to the particle surface.24 The surface layer of these nanoparticles can be further modified for increased aqueous stability, biocompatibility, biostability, and biorecognition, and a number of chemical modifications are presented in the academic literature.13,24 Hence, there is interest in processing methodologies which can either directly produce functionalized materials or produce materials which can be readily functionalized post-synthesis (i.e., producing materials with a suitable anchoring molecule for further functionalization).24

Hydrothermal routes for the synthesis of surface-functionalized nanoparticles are comparatively poorly described.1,7 Principally, this stems from the oxidation of capping agents under hydrothermal conditions, making effective surface functionalization difficult. The decomposition of organic compounds due to their oxidative potential in such conditions is well known, as in the industrial use of near-critical water in biomass remediation (supercritical water oxidation).7 However, several studies have detailed the synthesis of surface-functionalized inorganic nanoparticles, such as TiO2, ZnO, CeO2, and CoAl2O4, where dicarboxylic acids (C4−C14 acids), oleyl amine, and phosphonated acids have been used to confer stability in aqueous and nonaqueous media, respectively.1,6,25 In these examples, batch hydrothermal reactors were used and required relatively short reaction times of ca. 10 min to minimize oxidation of the capping agent. The authors have extensive experience in developing a laboratory-scale nanoceramics synthesis process known as continuous hydrothermal flow synthesis (CHFS), and have recently published the scale-up of the process to allow for the synthesis of up to 2 kg/h of nanoparticles.15,22,23,26−29 In the CHFS process, nanoparticles form upon mixing a feed of supercritical (or superheated) water with an ambient-temperature stream of aqueous metal salts. This is usually carried out using a particular reactor (mixer) arrangement, resulting in rapid conversion of the metal salts into metal oxides via simultaneous hydrolysis and dehydration reactions.23 In CHFS, 7437

DOI: 10.1021/acs.iecr.5b01817 Ind. Eng. Chem. Res. 2015, 54, 7436−7451

Article

Industrial & Engineering Chemistry Research

Figure 1. Schematic of the three-pump continuous hydrothermal process used for the direct synthesis of surface-functionalized magnetite nanoparticles.

(28 vol%) were supplied by Sigma-Aldrich Chemical Co. (Dorset, UK). Iron(III) citrate powder was first dissolved in water by adding a stoichiometric amount of NH4OH solution, forming a soluble iron citrate complex (ammonium ferric citrate, [NH3Fe(C6H5O7)]. All experiments were conducted using deionized (DI) water (of >10 MΩ resistivity). Materials Synthesis. Laboratory-Scale Process. Table 1 provides a summary of the synthesis conditions used to produce citric acid-coated magnetite nanoparticles using the laboratory-scale CHFS process. A schematic of the process is presented in Figure 1. The core components of the CHFS process were constructed almost exclusively of Swagelok 316 SS components.23 In this study, a flow of DI water was heated to 450 °C at a pressure of 24.1 MPa (i.e., above the critical point of water, TC = 374 °C and PC = 22.1 MPa) by pumping it through a custom-built electrical heater arrangement (2.5 kW). High-performance liquid chromatography (HPLC) pumps P-1 to P-3 (Gilson, SC-type fitted with 25 mL heads) were used to supply reagents and water to the process. The flow rate from each of the pumps was set as detailed in Table 1. Pump P-2 was used exclusively to supply ammonium iron citrate solution (0.066 M), and P-3 was used to pump a citric acid solution of variable concentration and of variable flow rate. In practice, P-3 was constructed of two HPLC pumps working in parallel, each fitted with 25 mL pump heads to allow the flow rate range stated in Table 1 to be used. Supercritical water issuing from P1 was mixed concurrently with the precursors in a Confined Jet Mixer (CJM) as detailed in our previous publications.23,28 The resulting mixture at a temperature of ca. 380 °C, containing nanoparticles, flowed through a pipe to meet a countercurrent flow of a supplementary reagent (citric acid solution) issuing from P-3, as shown in Figure 2, before exiting the process after cooling to ca. 30 °C and passing through a back-pressure regulator (Tescom, 26-1762-62). Hereafter, the outputs of each of the pumps are identified as Qsw (volumetric flow rate of preheated water issuing from P-1), Qp (volumetric flow rate of metal salt precursor issuing from P-2), and Qq (volumetric flow rate of the quenching flow introducing a citric acid solution to the nascent nanoparticle stream issuing from P-3). The reaction point shown in Figure 2 consisted of a 1/16-in. 316SS Swagelok tube swaged into a 1/4-in. 316SS Swagelok Tee-piece using a 1/16-in. bored-through reducer, allowing the tube to

nanoparticles are formed in the reactor (mixer) itself, then cooled via in-line (pipe-in-pipe) cooling, and then passed through a back-pressure regulator, allowing products to be continuously produced and harvested as an aqueous nanoparticle slurry at atmospheric pressure and temperature. Many of the operational aspects of CHFS are detailed in our previous publications.15,22,23,26−29 However, as CHFS is a flow process, this publication details the introduction of supplementary reagents, post particle formation, as an embodiment of a process that has previously received no attention in the academic or patent literature. Herein, we describe the development and optimization of a CHFS process for the direct synthesis of surface-functionalized nanoparticles. The optimization of the process is presented through complete characterization of nanoparticles produced under different processing conditions. We cite the production of citrate-functionalized Fe3O4 nanoparticles as a model system in this publication. We have chosen to use citric acid as a surface functionalization molecule, as the molecule confers electrostatic stabilization to particles and produces stable particle dispersions in aqueous systems, the molecule can be further functionalized through simple coupling chemistry, and, in the process used, electrostatic stabilization provides a convenient method for particle recovery and concentration.24 The structure and morphology of the products have been investigated using X-ray diffraction (XRD), transmission electron microscopy (TEM), high-resolution transmission electron microscopy (HRTEM), dynamic light scattering (DLS), electrophoretic mobility measurements, thermogravimetric analysis (TGA), and magnetometry. We address only the key conditions required to produce stable colloidal dispersions of nanoparticles, and not the development of the reactor or the mixing geometries employed. Many of the details of the latter can be found in our previous publications.15,22,26−28 Synthesis and characterization data from both a laboratory-scale process (nominal output flow rate of ca. 20−60 mL min−1) and a pilot-scale process (nominal output flow rate of ca. 500−1200 mL min−1) are reported.



MATERIALS AND METHODS Materials. Iron(III) citrate ([Fe(C6H5O7], technical grade, >98%), citric acid ([C6H8O7], 99%), and (NH4)OH solution 7438

DOI: 10.1021/acs.iecr.5b01817 Ind. Eng. Chem. Res. 2015, 54, 7436−7451

Article

Industrial & Engineering Chemistry Research

tubing between the inlet of P-3 and the cooler could be used to vary the residence time after the addition of citric acid to the flow of nascent nanoparticles (within the text, the volume of the reactor post addition of Qq is quoted). Pilot-Scale CHFS Process. The pilot-scale process emulates the laboratory-scale process described above and has been designed to process ca. 20 times the flow rate of the laboratoryscale process reported here (up to 1.2 L min−1 of nanoparticleladen slurry); a detailed account of the construction and commissioning of this process can be found in our previous publications.22,23,28 In the pilot-scale CHFS process, a flow of DI water was heated to 450 °C at a pressure of 24.1 MPa by pumping it through a custom-built electrical heater arrangement (24 kW). Chemical dosing pumps (Milton Roy, Primeroyal K) replaced the HPLC pumps used on the laboratory scale, and the flow rate from each of the pumps was set according to the experimental details presented in Table 2.23 In the pilot-scale CHFS process the reaction point was identical to that described in Figure 2. However, the physical dimensions of the reaction point were suitably enlarged. In the pilot-scale CHFS process the reaction point shown in Figure 2 consisted of a 1/4-in. 316SS Swagelok tube swaged into a 3/4in. 316SS Swagelok Tee-piece using a 1/4−3/4-in. boredthrough reducer, allowing the tube to extend into the stream of nascent nanoparticles issuing from a CJM-type reaction point constructed from a 1/4-in. inlet and a 3/4-in. cross piece, similar to that described previously.22,23 In the pilot plant, the exit of the reaction point from which the products of the reaction pass was of fixed length (identified as Tmix2 in Figure 2) to limit the number of experimental variables which needed to be considered in the pilot-scale CHFS process. The total reactor length, defined from the terminus of the preheated water inlet to the inlet, was 40 cm, yielding a residence time of ca. 1.8 s, equivalent to that used in the laboratory-scale process. Details of the product recovery methods can be found in the Supporting Information. In Situ Temperature Measurements. The maximum theoretical temperature, calculated for the outlet of the mixer, prior to the introduction of a quenching flow (Qq), is called the reaction point temperature (Tmix1 in Figure 2) throughout this

Figure 2. Schematic of the modified reaction point geometry used to introduce a citric acid solution into the CHFS system. Tmix1 defines the reaction point temperature for the formation of nanoparticles, and Tmix2 defines the temperature after mixing the products of the reaction with Qq.

extend into the stream of nascent nanoparticles issuing from a CJM-type reaction point similar to that described previously.23 The superheated water flow (issuing from P-1) is introduced into the mixing point through an inner 1/16-in. tube, and the metal salt solution flows upward through the 1/4-in. tube. The exit of the reaction point where the products of the reaction pass was connected to the cooling apparatus through a variable length of 1/4-in. 316SS tubing (identified as Tmix2 in Figure 2). The total reaction zone length, defined from the terminus of the preheated water inlet to the inlet, was 10 cm, yielding a residence time of ca. 1.8 s. Similarly, variation of the length of

Table 2. Summary of the Reaction Condition and Characterization Data for Citric Acid-Coated Magnetite Produced Using a Pilot-Scale Continuous Hydrothermal Flow Synthesis Processa pump flow rates (mL min−1)

DLS

ID

Qsw

Qp

Qq

Tmix2 (°C)

0.6CA1.69 0.8CA1.63 1.0CA1.62 1.2CA1.61 0.7CA1.51 0.9CA1.48 1.1CA1.46 1.5CA1.45

300 300 300 300 400 400 400 400

260 260 260 260 360 360 360 360

206 274 343 411 300 400 500 600

277 257 240 225 268 248 231 216

CA:Fe

RT mix2 (s)

CS (nm)

weight loss (%)

Dh (nm)

PDI

0.6 0.8 1.0 1.2 0.7 0.9 1.1 1.5

1.69 1.63 1.62 1.61 1.51 1.48 1.46 1.45

10.78 13.67 10.94 12.56 13.96 10.79 14.36 15.67

3.28 2.41 2.92 2.57 8.11 7.48 9.80 8.64

96 91 60 70 42 40 60 43.2

0.168 0.120 0.101 0.245 0.132 0.135 0.167 0.131

TEM Dh(n)

mean CS (nm)

SD

CA/ nm2b

34.9 39.7 29.3 28.3 16.7 22.8 19.4 20.3

12.5 13.5 16.6 11.7 15.8 11.6 12.8 11.4

4.2 3.3 3.3 3.1 4.2 3.3 4.5 1.2

0.79 0.67 0.35 0.52 2.64 2.58 2.57 2.02

The ammonium iron citrate concentration was fixed at 0.066 M, and the heater temperature was set to 450 °C, resulting in a reaction point temperature of ca. 360 °C (Tmix1) at the flow rates used. Residence times (RT) were calculated from the reaction zone length and the velocity of the fluid at the theoretical mixture temperature (i.e. Tmix1 or Tmix2). Samples are identified by the ratio of CA:Fe used in the synthesis and the residence time post citric acid addition; e.g., a sample produced with a CA:Fe ratio of 1.5 and a residence time post CA addition of 2.5 s is identified as 1.5CA2.5. Qq is the flow rate of citric acid solution, Tmix2 is the theoretical mixture temperature post citric acid addition, CS is the crystallite size, Dh is the hydrodynamic diameter measured using DLS, PDI is the polydispersity index and Dh(n) is the number weighted hydrodynamic diameter distribution. bGrafting density of CA (molecules per square nanometer) determined using eq 1 and the crystallite size determined using TEM. a

7439

DOI: 10.1021/acs.iecr.5b01817 Ind. Eng. Chem. Res. 2015, 54, 7436−7451

Article

Industrial & Engineering Chemistry Research

total reflectance (ATR) sampling accessory. Spectra were obtained in the wavenumber range from 4000 to 600 cm−1 with 64 scans and a spectral resolution of 4 cm−1. TGA of the samples was performed on a Netzsh STA 449C instrument under a constant flow of air (flow rate = 5 mL min−1); analysis was performed at a heating rate of 5 K min−1 from room temperature to 800 °C. Collection of weight loss and differential scanning calorimetry (DSC) data was simultaneous. Weight loss regions were quantified using Proteus analysis software (Netzsh). An MPMS SQUID-VSM magnetometer (Quantum Design, San Diego, USA), allowing DC magnetization measurements at temperatures between 2 and 400 K in magnetic fields between ±7 T, was used to assess the magnetic properties of the reaction products. All magnetic property measurements were performed on freeze-dried powders, and the mass of the sample was corrected to that determined to be magnetic material from TGA (accounting for the presence of residual organic species or capping agents). Both M(H) curves and zero-field-cooled/fieldcooled (ZFC/FC) measurements were performed on materials prepared in this fashion. The measurements of M(H) curves were performed at 300 K unless otherwise stated, and measurements were performed up to a maximum applied field strength of 5 T. ZFC/FC measurements were performed at constant applied field of 100 Oe. Each sample was measured consecutively from low temperature (5−10 K) to high temperature (300 K). ZFC measurements were cooled in the absence of an applied field, and the magnetization was measured with a temperature sweep rate of 5 K min−1. The magnetization data were recorded point-wise from the lowest measurement temperature. The ferromagnetic volume of particles and a paramagneticlike susceptibility (χ), the latter accounting for the nonsaturating behavior observed in samples, were calculated by fitting the experimental magnetization curve assuming a volume-weighted log-normal ferromagnetic diameter distribution as described Chen et al.31

publication, whereafter the theoretical mixture temperature, determined from overall enthalpy balance, is identified as Tmix2. The specific enthalpies and the temperature at 24.1 MPa were determined from the IAPWS Formulation 1995 using the associated FLUIDCAL software.30 This software was also used to calculate the density and viscosity of water for any given temperature at 24.1 MPa and used to derive the reactor residence times quoted in Tables 1 and 2 (i.e., between the reaction point and inlet of the quenching flow (RT1) or between the inlet of the quenching flow and the cooling apparatus (RT2)). Measurements of temperature within the geometry were allowed through the use of a series of thermocouples located within the flow. The details of thermocouple placement are available in the Supporting Information. A Spectite MF series thermocouple port allowed four simultaneous measurements of temperature using J-type thermocouples (TC Direct). Due to the physical size of the pilot-scale process equipment, we were unable to take temperature measurements from the process, and so we only present data from the laboratory-scale reactor. Characterization. XRD patterns were collected using a Bruker D4 diffractometer (Cu Kα1, λ = 1.540598 Å). A secondary monochromator was fitted to the diffractometer, allowing the use of Cu Kα1 with Fe-based samples. Yttria (Y2O3) was used as a standard for the estimation of instrumental peak broadening. Samples were prepared for TEM by dropping a dilute particle dispersion in DI water (resistivity >15 MΩ) onto carbon film grids (Agar Scientific, UK). A JEOL 1200 transmission electron microscope (120 keV accelerating voltage) was used for generating images of particles, and images were captured using a CCD camera (GATAN). Samples were prepared for HRTEM by dispersing the particles ultrasonically in ethanol (99.9%, Sigma-Aldrich, Dorset, UK) and dropping onto holey carbon film grids (400 mesh, Agar Scientific, UK). A JEOL 4000× transmission electron microscope (350 keV accelerating voltage) was used for generating HRTEM micrographs, and images were captured using a CCD camera (GATAN). The particle size distribution of nanoparticles in suspension, obtained directly from the CHFS process after recovery, was determined using a Malvern Zetasizer Nano (ZEN3600) instrument, calibrated using a size standard (Nanosphere 3200A, 199 ± 6 nm). All measurements were performed in disposable plastic cuvettes with a nominal path length of ca. 1 cm. Each measurement was taken using a backscatter geometry. Each reported measurement is an average of 10−25 runs (each of ca. 10 s duration) reproduced three times. Zeta (ζ)-potential measurements were performed using a Malvern Zetasizer Nano (ZEN3600) instrument. All single ζ-potential measurements were performed on dispersions of nanoparticles from the reactor after cleaning in DI water at pH 7. The dispersions were ultrasonicated for 5 min, and then ca. 2 mL was placed in a disposable ζ-potential cell (Malvern DTS 1060). Each measurement was a composite of 10−20 runs (each lasting ca. 10 s). Calculations of ζ-potential were performed by the software (Malvern DTS Nano). Electrophoretic mobility measurements were taken at constant electrolyte concentrations (5 mM NaCl) using NaOH and HCl as titrants. Electrophoretic mobility measurements were converted to ζpotential using the Smoukowski approximation. Fouriertransform infrared (FTIR) spectra were recorded on a PerkinElmer Spectrum 1 instrument fitted with an attenuated



RESULTS AND DISCUSSION Optimization of the Laboratory-Scale Process. The samples produced in this section are summarized in Table 1 and are identified by the molar ratio of CA:Fe (where CA is citric acid) used in the synthesis and the calculated residence time post addition of citric acid. Initially combinations of flow rates (Qsw, Qp), quench rates (Qq), relative iron to citric acid ratios (CA:Fe) and post capping agent addition residence times were evaluated to assess the effect of each variable on the product obtained from the reaction. The addition of the capping agent (citric acid) after the formation of nanoparticles was used to ensure survival of the capping agent and to segregate the particle growth steps (occurring at ca. 380 °C) from the functionalization steps (occurring at a lower temperature as defined in Table 1). Dissolution of the reaction product was consistently observed when CA:Fe ratios >10, either within the CHFS system or shortly after product collection, which was attributed to the acidity of the particle slurry (ca. pH 1−2) consistent with the known aqueous dissolution of magnetite.32 Powder XRD patterns of the materials produced in each of the reactions identified in Table 1 are shown in the Supporting Information, Figure S1. In each case, the diffraction patterns were in good agreement with those of both magnetite (ICDS 082234) and Maghemite (ICDS 79196), respectively. As 7440

DOI: 10.1021/acs.iecr.5b01817 Ind. Eng. Chem. Res. 2015, 54, 7436−7451

Article

Industrial & Engineering Chemistry Research

Figure 3. TEM images of the products of reactions conducted on the laboratory-scale CHFS process (samples are identified by the ratio of CA:Fe used in the synthesis and the calculated residence time post addition of Qq): (a) 1.5CA4.2, (b) 2.3CA3.4, (c) 1.5CA5.2, (d) 3.2CA4.2, (e) 4.7CA3.4, (f) 3.1CA5.2, (g) 6.3CA4.2, (h) 9.4CA3.4, and (i) 1.5CA7.9. The scale of each image is indicated by the corresponding scale bar.

the particle dispersions, which was dependent upon both the CA:Fe ratio and the residence time post addition of Qq. The products of the reactions which had residence times of >5.2 s yielded materials with the highest degree of dispersion. Figure 4 shows ATR-FTIR spectra obtained for the coated magnetite nanoparticles which showed modes origination from (ν)CO present at 1735 cm−1, (νs)COO− (∼1390 cm−1), (ν)OC−OH (∼1204 cm−1), (νs) OC−OH (∼1429 cm−1), and (νas)COO− (∼1569 cm−1) confirming the presence of citric acid on the surface of the products of all reactions conducted using a flow regime of Qsw = 10 mL min−1 and Qp = 5 mL min−1. The observation of the (νas)COO− stretching mode centered at 1569 cm−1 suggests a proportion of the acid groups present in the sample are coordinated to the particle surface.32 The wavenumber separation between the (νs)COO− (ca. 1390 cm−1) and the (νas)COO− (ca. 1569 cm−1) modes has been used to determine the binding state of the carboxylate group to the metal oxide surface using the wavenumber separation (Δ) between the asymmetric (νas) and symmetric (νs) IR modes.1,34,35 Unidentate complexes exhibit Δ values (ranging from 200 to 320 cm−1) that are much greater than those of

magnetite is a mixed-valence iron oxide and has an inverse spinel structure it is often difficult to differentiate from other magnetic spinel iron oxide phases, hence the phase of these materials will be referred to as magnetite.33 The addition of the capping agent through Qq was shown to have no effect on the absolute phase determined by XRD as shown in Figure S1. Application of the Scherrer equation to the (311) reflection gave estimates of crystallite size in the range of 10−13 nm, and the data for each sample are summarized in Table 1. The crystallite size was shown to be largely invariant of synthesis condition, suggesting that the particle formation steps were independent of both Qp and CA:Fe ratio (until the onset of dissolution CA:Fe >10) and that particle formation occurs on vanishingly small time scales. This is unsurprising given the nucleation-dominated nature of particle formation in CHFS.22,23 TEM images of selected samples are shown in Figure 3. Crystallite size and size distribution data are summarized in Table 1, which yielded values in good agreement with those determined by application of the Scherrer equation to the diffraction data. In Figure 3, it can be seen that processing conditions affected the degree of agglomeration in 7441

DOI: 10.1021/acs.iecr.5b01817 Ind. Eng. Chem. Res. 2015, 54, 7436−7451

Article

Industrial & Engineering Chemistry Research

( 43 πr 3)d( mm ) CA

σ=

Fe

MM CA

/4πr 3

(1)

Here, σ is the grafting density expressed as the number of molecules per square nanometer (CA nm−2), mCA/mFe is the mass ratio of capping agent to magnetite (determined by TGA), MMCA is the molecular weight of the capping agent (192.12 g mol−1), d is the density of magnetite (assumed to be 5.15 g cm−3), and r is the particle’s radius, calculated from the average particle diameter determined by TEM. The number of carboxylate groups available for adsorption per square nanometer (binding area ca. 0.021 nm2) of the nanoparticle surfaces was not used in the calculation of the grafting density due to the number of possible binding configurations citric acid could adopt at the surface of the nanoparticles (as suggested by ATRFTIR).39 Figure 5 shows the variation in CA grafting density as Figure 4. Stacked FTIR spectra of the products from CHFS, showing an expanded region of the complete spectra (range 1000−2000 cm−1) identifying the IR active mode positions. Each sample is identified by the ratio of CA:Fe used in the synthesis and the residence time post addition of Qq.

bidentate complexes (ranging from 140 to 190 cm−1).35 The values of Δ determined for the citric acid-coated magnetite samples varied between 190 and 220 cm−1 as shown in Figure 4 through variation in the (νas)COO− mode position suggesting mixtures of monodentate and bridging coordination with the particle surface consistent with the coordination expected for citric acid-coated nanomaterials.36 TGA was used to quantify the proportion of the sample mass attributable to citrate species obtained from materials produced flow regime of Qsw = 10 mL min−1 and Qp = 5 mL min−1. Figure S2 in the Supporting Information shows the TGA data obtained for samples reported in Table 1 showing variation in the regions of weight loss. The TG curve can typically be divided into two dominant stages of weight loss, the region below 90 °C, over which the mass loss ranged from ca. 5% to 10% (expressed as a percentage of sample mass) can be attributed to the removal of weakly associated and physisorbed water. The second region of weight loss in the 180−400 °C region can be attributed to the decomposition of organic species.37 DSC data obtained for each of the samples showed two characteristic endotherms; the first was attributed to the evaporation of physisorbed water consistent with the first weight loss observed in TGA, and the second endotherm observed can be attributed to the decomposition of organic species (principally citrates) consistent with the second region of weight loss observed in Figure S2. A third small endotherm was also observed at ca. 600 °C which corresponds well with the phase transition temperature of Fe3O4/γFe2O3 → αFe2O3.37 The measured weight loss for each sample is summarized in Table 1. The grafting density of CA on the surface of magnetite nanoparticles was calculated from the composition and surface area data determined from the mean crystallite size measured by TEM (see eq 1). Under this characterization, it is assumed that CA coordination to the particle surface is irreversible, as widely reported for citric acid coordination to nanoparticle surfaces at >300 K, and all binding events occur within the reaction point.38

Figure 5. Variation in citric acid grafting density for magnetite nanoparticles produced under different processing conditions using CHFS.

a function of both CA:Fe ratio used in synthesis plotted against the reactor volume post CA addition, indicative of the residence time post addition of Qq. It can been seen that the reactor volume post addition of the CA feed (Qq at almost any flow rate) was the most influential factor in determining the CA grafting density and was largely invariant of all other processing conditions. The observed surface saturation of ca. 2.0−2.6 CA nm−2 is consistent with the typical saturation density reported in the literature.40,41 The aqueous dispersion characteristics of magnetite nanoparticles produced using the CHFS process were assessed by DLS. Figure 6 shows the variation in hydrodynamic diameter (z-average) as a function of both CA:Fe ratio and the reactor volume post citric acid addition. As shown in Figure 6, the most significant contribution to the reduction in hydrodynamic diameter was the volume of the reactor post addition of citric acid as dispersion quality was not as severely influenced by the ratio of CA:Fe in the reaction. Typically, the measured hydrodynamic diameter and relative polydispersity was reflected by the samples grafting density compared in Table 1. However, samples produced with the longest residence time were the least disperse, suggesting agglomeration of the samples had occurred which could be linked to the lower grafting density discussed earlier or an alteration in the mechanism of particle stabilization due to longer hydrothermal reaction (discussed further later). It is worthwhile to note that 7442

DOI: 10.1021/acs.iecr.5b01817 Ind. Eng. Chem. Res. 2015, 54, 7436−7451

Article

Industrial & Engineering Chemistry Research

Figure 6. Variation in the hydrodynamic diameter (number-weighted) of citrate-functionalized magnetite nanoparticles produced using CHFS. The samples are identified by sample number as described in Table 1.

a function of pH (Figure 7). In all samples, a reduction in the magnitude of the surface charge was observed at pH 6, suggesting suppression of the dissociation of the weakest acid group of citric acid. The reported pKa values of CA as a tricarboxylic acid are pKa1 = 3.13, pKa2 = 4.76, and pKa3 = 6.40. The ζ-potential of samples which showed CA grafting densities of >1 CA nm−2 showed titrations characteristic of the protonation and deprotonation of free acid groups on the surfaces of the particles (samples 1.5CA2.5−4.7CA1.7 and 0.7CA5.2−3.1CA6.3). The magnitude of the ζ-potential followed the trend of grafting density, as samples 0.7CA5.2− 6.3CA4.2 showed the largest magnitude of surface charge. These results suggest the weakest acid group remains free and binding of citric acid to the particle surface occurs through the

the intensity of the distribution in this measurement is biased toward larger particles in the sample as scattering by particles is proportional to r6 (r is particle radius); i.e., particles of ca. 10 nm scatter 105 fewer photons than particles of ca. 100 nm, leading to under-representation of small particles. Quoting the hydrodynamic diameter corrected to particle number, yielded average sizes in each dispersion slightly larger than those determined from diffraction and TEM as summarized in Table 1. It is reasoned that the smaller nanoparticles represented the majority of the sample by mass on the basis of the high citric acid grafting densities calculated for each of the samples and direct observation of particles using TEM (Figure 6). The dispersion stability of iron oxide nanoparticles produced using CHFS were investigated by ζ-potential measurements as 7443

DOI: 10.1021/acs.iecr.5b01817 Ind. Eng. Chem. Res. 2015, 54, 7436−7451

Article

Industrial & Engineering Chemistry Research

attributed to the amphoteric behavior of magnetite, which can develop surface charge in the protonation (Fe−OH + H+ → Fe−OH2+) and deprotonation (Fe−OH → Fe−O− + H+) reactions of Fe−OH sites.36 The IEP (iso-electric point) of magnetite is often quoted at pH 7−8 for both bulk an nanosized variants as evaluated by others.36 In this work, only slight amphoteric behavior (observed as a smooth transition in the ζ-potential) was observed, principally due to the presence of citrates in all samples (Table 1). All samples showing CA grafting densities >1 CA nm−2 yield dispersions which are electrostatically stabilized and characterized as stable, as values of >30 mV are indicative of stable dispersions.36 It was found that particle dispersions which showed citric acid grafting densities >1 CA nm−2 and ζ-potential values in excess of 30 mV formed very stable particle dispersions after recovery. Under this characterization these particle dispersions formed magnetically actuable fluids when dispersed to particle concentrations of 50 mg mL−1, the ferrofluid formed by dispersion of sample 3.1CA5.2 is shown in the Supporting Information, Figure S3, under the influence of a rare-earth magnet. This image provides an indication to the reader of the high-quality particle dispersions formed by this process. Processing at Higher Flow Rates. To further assess the effects of processing conditions on surface functionalization in the modified CHFS process materials were produced under a higher flow regime (Qsw + Qp = 35 mL min−1). Under this flow regime, a different optimization condition was observed, which ultimately yielded materials with physiochemical characteristics similar to those reported earlier within the text. Hence, only the relevant differences are discussed here. XRD patterns of the materials produced in runs 0.7CA2.2− 4.7CA2.0 (as summarized in Table 1) suggest that all products are well crystallized and have a cubic inverse spinel structure known for bulk magnetite and maghemite (ICDS 082234 and ICDS 79196), consistent with the materials produced using a low flow regime. TEM images of the citrate-coated iron oxide nanoparticles produced at a high flow regime are shown in Figure 8. This sample series showed similar degrees of dispersion to that initially observed for iron oxide nanoparticles

Figure 7. Zeta-potential titration of citrate-coated iron oxide nanoparticles produced using the laboratory-scale CHFS process. Samples are identified by the ratio of CA:Fe used in the synthesis and the residence time post addition of Qq.

strongest acid groups consistent with many reports in the literature.38 The liability of the different acid groups of citric acid to decarboxylation as reported by Shock et al. suggest that the acid group with pKa = 4.76 is the most labile and decarboxylates preferentially, possibly providing an explanation for the similarity in surface charge even in samples which showed a reduction in grafting density attributed to thermal decomposition (samples 3.1CA5.2−1.5CA5.9).42−44 The similarity in the magnitude of the ζ-potential measured for samples 1.5CA7.9 and 3.1CA6.3 is consistent with the decarboxylation proposed by Carlsson et al., where the free carboxlic acid group present in the sample is the same as that observed for samples showing a high-citric acid grafting density with an apparent pKa ≈ 6.40, although this could not be readily resolved using ATRFTIR and electrophoretic mobility data alone.37 The surface charge observed for samples showing low CA grafting densities (samples 1.5CAO2.5−4.7CA1.7 as shown in Figure 7) can be

Figure 8. TEM images of citric acid-coated iron oxide nanoparticles produced using a CHFS reactor: (a) 1.1CA2.0, (b) 0.7CA2.2, (c) 1.5CA2.2, (d) 2.3CA2.0, (e) 3.1CA2.2, and (f) 4.7CA2.0. 7444

DOI: 10.1021/acs.iecr.5b01817 Ind. Eng. Chem. Res. 2015, 54, 7436−7451

Article

Industrial & Engineering Chemistry Research produced with high CA grafting densities using a low flow regime (Figure 4). The crystallite size measured for samples 0.7CA2.2−4.7CA2.0 (ca. 5.6−10.4 nm) was smaller than those observed for materials produced in runs 1.5CA2.5−1.5CA7.9 (ca. 11−15 nm). The reduced crystallite size produced in these reactions could be attributed to slight differences in the particle formation reactions occurring when the iron precursor and supercritical water mix. Calculation of the CA grafting density for the iron oxide samples produced in the high flow regime (eq 1) yielded saturation grafting densities of ca. 2.6−2.9 CA nm−2 consistent with the saturation grafting density for samples produced in the low flow regime (TGA data for samples 1.1CA2.0−4.7CA2.0 can be found in the Supporting Information, Figure S5). Figure 9 shows the variation in citric

functionalization was governed by diffusion of CA to the nascent particle surface and was likely linked to mixing between the component streams. In the stated flow regime the mixture issuing from the reaction point (defined by Tmix1 in Figure 1) is a highly turbulent flow with a Reynolds number (Re, see eq 2) between 6.12 × 104 (Qsw + Qp = 15 mL min−1) and 11.23 × 104 (Qsw + Qp = 35 mL min−1).23,26 Mixing of the feed containing nascent nanoparticles and the feed containing citric acid is likely to be complex due to (1) the density difference between the feeds (ca. 280 kg m−3 (Tmix1) and 1002 kg m−3 (Qq), (2) the differences in flow velocity attributed to the difference in density and the mass flow rate, and (3) the heat transfer occurring between Qq and the mixture containing nanoparticles. However, an estimation of the flow type occurring in the system under the different flow conditions suggested the mixture after the addition of citric acid (Qq) was a turbulent flow in the high flow regime (Re = (3.4−4.5) × 103 (Qq = 10− 40 mL min−1) if the mixture is assumed to be at Tmix2, whereas in the low flow regime Re = (1.9−2.6) × 103 (Qq = 10−40 mL min−1) suggesting a transient flow. The Reynolds number of the flow was calculated as follows: ReTmix =

ρT uTmix d i mix

μT

mix

=

4GTmix πd iμT

mix

(2)

where ρTmix, uTmix, and μTmix are respectively the density, velocity, and dynamic viscosity of the mixture of flows at any defied point within the reactor, di is the internal diameter of the confining tube (3.78 mm on the laboratory scale and 22.4 mm on the pilot scale), and G is the mass flow rate of the mixture. This analysis would offer a straightforward interpretation if surface functionalization was linked to mixing at high temperature and thus diffusion of CA to the particle surface as in a turbulent flow diffusion of CA to the particle surface would be enhanced yielding more rapid saturation of the particle surface. To confirm if differences in flow regimes were responsible for the variation in grafting density in situ temperature measurements were used to establish the extent of mixing between the component streams as a function of Qsw + Qp and Qq. The relative placement of thermocouples is represented graphically in the Supporting Information, Figure S7. The temperature profiles measured are presented in Figure S7, for flow regimes of Qsw + Qp = 35 mL min−1 and Qsw + Qp = 15 mL min−1, with variable Qq, as 30 s time-averaged temperature measurements (thermocouples occupying ca. 4% of reactors cross sectional area). It should be noted that significant noise was recorded in the measurements, suggesting the mixing process is not in a steady state. Figure S7 clearly shows that the introduction of a feed post nanoparticle formation results in complex mixing of the two feeds as the theoretical mixture temperature as an average was not met in any flow condition and may be an artifact of poor mixing and the relative position of thermocouples. However, one interesting feature of the data is the degree to which the cold water feed penetrates into the stream of nascent particles and appears related to Qq leading to a reduction in Tmix1. These results suggest a significant density difference between the incoming stream and the nascent stream provides an explanation as the significantly denser Qq falls through the less dense stream of nanoparticles although a degree of heat transfer would be expected to the incoming flow of Qq. To confirm that the discrepancy in temperature did not arise as the result of heat loss a slow response thermocouple (of

Figure 9. Variation in the calculated citric acid grafting density obtained for samples produced using a flow regime of Qsw = 25 mL min−1 + Qp = 10 mL min−1, plotted as a function of CA:Fe ratio used in synthesis. Characterization details are provided in Table 1.

acid grafting density plotted as a function of CA:Fe ratio used in each synthesis. It can be seen from Figure 9 that saturation of the particle surface occurred at a CA:Fe ratio of ca. 1.8, and the grafting density appeared largely proportional to CA:Fe ratio below this value, indicative of a diffusion limit to functionalization in this flow regime. The calculated residence time post synthesis addition was lower (ca. 2.0−2.20 s) than the optimum residence time observed in the low flow regime (ca. 4.42−5.21 s). Achieving similar CA grafting densities at shorter residence times suggests differences in the degree of mixing between Qsw + Qp (containing the nascent nanoparticles) and Qq (citric acid feed), which at a lower flow rate limited the rate of citrate diffusion to the nascent particle surface. All other characterization data for this sample series showed compositionally and physically similar materials at those reported earlier within the text were obtained (data presented in the Supporting Information, Figures S5 and S6). Proposed Mechanism of Surface Functionalization. The density of citric acid was shown to vary with flow regime chosen for synthesis. For samples produced using a low flow rate (Qsw + Qp = 15 mL min−1) estimated residence times post CA addition of ca. 5 s were required to achieve saturation of the particle surface. However, at a higher flow regime (Qsw + Qp = 35 mL min−1) residence times of ca. 2 s were shown to give products with equivalent CA grafting densities when the CA:Fe ratio used in synthesis was >1.8, suggesting the degree of 7445

DOI: 10.1021/acs.iecr.5b01817 Ind. Eng. Chem. Res. 2015, 54, 7436−7451

Article

Industrial & Engineering Chemistry Research

Figure 10. TEM images of the products of reactions conducted on the large-scale CHFS process (samples are identified by the ratio of CA:Fe used in the synthesis and the residence time post addition of Qq): (a) 1.5CA4.2, (b) 2.3CA3.4, (c) 1.5CA5.2, (d) 3.2CA4.2, (e) 4.7CA3.4, (f) 3.1CA5.2, (g) 6.3CA4.2, and (h) 9.4CA3.4. The scale of each image is presented in the corresponding scale bar.

Figure 10. The average particle sizes and distribution were calculated from the measurement of ca. 300 particles and are summarized in Table 2, confirming that crystallite size is largely independent of processing conditions, consistent with the observations made in the laboratory-scale analogue. The crystallite sizes obtained were in good agreement with the crystallite sizes determined from the corresponding diffraction data, as summarized in Table 2. This finding is consistent with one of our previous reports where nanoparticle characteristics were deemed indistinguishable at different production scales if the process was scaled on flow rate alone.22,23 Surface functionalization of the particles produced in the scaled reactions was confirmed using FTIR. Spectra obtained for the coated magnetite particles showed modes originating from (ν)CO present at ca. 1735 cm−1, (νs)COO− (∼1390 cm−1), and (ν)OC−OH (∼1204 cm−1), (νs)OC−OH (∼1429 cm−1), and (νas)COO− (∼1569 cm−1), confirming the presence of citric acid. The apparent wavenumber separation between (νs)COO− (ca. 1390) and (νas)COO− (ca. 1569) varied between 190 and 220 cm−1, suggesting a mixture of monodentate and bidentate coordination to the surface of magnetite (Supporting Information, Figure S9).1,35 The weak and broad nature of the observed modes makes a definitive binding state difficult to confirm using the presented data. However, a tentative approximation from the data suggests a mixture of monodentate and bidentate coordination, consistent with the data presented earlier within the text (Figure 4). To further assess the similarity between the materials produced in the pilot-scale reactions and those reported for the laboratory-scale experiments; DLS, TGA, and electrophoretic mobility measurements were used to study the hydrodynamic diameter, citric acid grafting density and dispersion stability, as a function of the volumetric scale-up. The grafting density of citric acid bound to the surface of magnetite produced using the pilot-scale process was calculated from the composition and surface area (from the mean crystallite size determined using TEM) using eq 1. TGA data collected for the products are shown in the Supporting Information, Figure S10, and the weight loss data are

significantly greater thermal mass) just before the cooling apparatus gave measurements of temperature within 5−10 °C of those expected from enthalpy balance further suggesting the system is not steady state and that some phase separation in the mixture is possible and averages over a longer time period. These measurements suggest the analysis of different flow types based on Tmix2 determined from enthalpy balance may not be locally correct throughout the geometry although averaged it could provide some insight into the mechanisms occurring within the reactor. Although, these temperature measurements coupled to materials characterization data largely support a diffusion limit to grafting density. Investigation of Large-Scale Synthesis. The synthesis of citric acid-coated nanoparticles using a pilot-scale process scaled on a volumetric basis alone was also evaluated to define an operational space in which stable dispersions of particles could be produced. In the large-scale process, the reaction point temperature was reduced slightly by reducing the ratio of Qsw:Qp to prevent temperature and pressure instabilities within the reactor as discussed in our previous publications.22,23 The basis of the scale-up procedure employed here is detailed in our previous publication.23 The experimental details of the reactions conducted using the pilot-scale CHFS process are summarized in Table 2 alongside the characterization data obtained for the samples which were produced at two process output volumes. All materials obtained from the pilot-scale process formed ferrofluids after consolidation, an observation consistent with the materials produced via the laboratory-scale process. Similarly, all diffraction (PXRD) patterns were in good agreement with those of both magnetite (ICDS 082234) and maghemite (ICDS 79196), respectively (Supporting Information, Figure S8). As discussed earlier within the text, magnetite is a mixed-valence iron oxide and it has an inverse spinel structure that is often difficult to differentiate from other magnetic spinel iron oxide phases, as such, we have assumed similar phase behavior in the pilot-scale CHFS process to that reported earlier within the text. TEM images representative of the nanoparticles produced in each synthesis are presented in 7446

DOI: 10.1021/acs.iecr.5b01817 Ind. Eng. Chem. Res. 2015, 54, 7436−7451

Article

Industrial & Engineering Chemistry Research summarized in Table 2. The variation in grafting density as a function of scale-up ratio (Qsw = 400 mL min−1, Qp = 360 mL min−1, 20× on flow vs laboratory scale; Qsw = 300 mL min−1, Qp = 260 mL min−1, 15× on flow vs laboratory scale) is plotted in the Supporting Information, Figure S11, showing an apparent surface saturation of ca. 2.6 CA nm−2, consistent with the saturation density calculated for magnetite produced via the laboratory-scale process and saturation values reported in the literature.18,38 The observations of increasing grafting density as a function of increasing flow rate are assumed to be due to differences in the mixing environment between successive volumetric scales. As initially discussed for the laboratory-scale process it is hypothesized that the increasing grafting density as a function of volumetric scale-up in the pilotscale process could also be linked to mixing and thus citric acid diffusion in the reactor varying as a function of flow rate. The degree of mixing between the citric acid feed (Qq) and the nascent nanoparticle stream (Qsw + Qp) was not experimentally validated using the pilot-scale process due to the size of the process equipment. However, to provide an indication of whether this is a reasonable interpretation of these observations, the flow type occurring within the system was estimated for each condition presented by calculating the Re of the flow eq 2 as a convenient comparison of the mixture. In the stated flow regimes the mixture issuing from the reaction point (defined by Tmix1) could be expected to be a highly turbulent flow with Reynolds numbers ranging from Re = 8.46 × 10 3 (Qsw = 300 mL min−1 + Qp = 260 mL min−1) to Re = 12.65 × 10 3 (Qsw = 400 mL min−1 + Qp = 360 mL min−1). Mixing of the feed containing nascent nanoparticles and the feed containing citric acid is likely to be complex due to the reasons outlined earlier within the text. However, an estimation of the flow type occurring pilot-scale system under the different flow conditions suggested the mixture after the addition of citric acid was a turbulent flow in the highest scale-up (Qsw = 400 mL min−1 + Qp = 360 mL min−1) flow regime, yielding Re = 12.5 to 12.9 × 103 (Qq = 300−600 mL min−1) if the mixture is assumed to be at Tmix2, when Qsw = 300 mL min−1 + Qp = 260 mL min−1, Reynolds numbers ranging from Re = 4.38 × 103 to 7.89 × 103 (Qq= 137−411 mL min−1) were calculated, significantly lower in magnitude than those obtained in the high flow regime, suggesting a transient flow type in the lower flow regimes and a turbulent flow in the high flow regimes. This is consistent with the observations presented in the laboratoryscale process where higher grafting densities of citric acid were obtained in a turbulent flow regime. Again, it is suggested that surface functionalization was linked to mixing at high temperature, and thus diffusion of CA to the particle surface as in a turbulent flow diffusion of CA to the particle surface would likely be enhanced. The stability of the citric acid-coated iron oxide nanoparticles in aqueous dispersions produced via the pilot-scale process were assessed using ζ-potential measurements as shown in Figure 11. Compared to Figure 7 and Figure S6 in the Supporting Information, the variation in the magnitude of the ζ-potential as a function of dispersion pH for samples produced on the pilot scale were similar to those produced via the laboratory-scale process (for materials showing similar grafting densities). The ζ-potential of samples which showed CA grafting densities of >1 CA nm−2 gave characteristic features of the protonation and deprotonation of free acid groups on the surfaces of the particles, yielding values similar to those obtained via the laboratory-scale process. A characteristic

Figure 11. Zeta-potential titrations of citrate-coated magnetite samples produced using the pilot-scale CHFS reactor (measurement standard deviations were typically ca. ±4−6 mV and have been omitted for clarity). Samples are identified as they appear in Table 2.

change in surface charge around pH 6, corresponding well with the pKa of the weakest acid group of citric acid (pKa = 6.13) suggests similar coordination of the acid to the particle surface (as observed for the materials presented earlier), serving to confirm the electrostatic stabilization of the dispersion. Samples produced via the pilot-scale process showing CA grafting densities