Superparamagnetic Maghemite Nanorods: Analysis by Coupling Field

SAXS curves were recorded in a 1 s time interval. ... A more detailed analysis of the scattering curves revealed that the particles were cylindrical i...
1 downloads 0 Views 1MB Size
Download PDF
Anal. Chem. 2008, 80, 5905–5911

Superparamagnetic Maghemite Nanorods: Analysis by Coupling Field-Flow Fractionation and Small-Angle X-ray Scattering Andreas F. Thu ¨ nemann,* Jenny Kegel, Jo ¨ rg Polte, and Franziska Emmerling BAM Federal Institute for Materials Research and Testing, Richard-Willsta¨tter-Strasse 11, 12489 Berlin, Germany We report on the online coupling of asymmetrical flow field-flow fractionation (A4F) with small-angle X-ray scattering (SAXS) for the detection of nanoparticles. The A4F was used to fractionate superparamagnetic maghemite nanoparticles, which were prepared continuously with a micromixer. The outlet of the A4F was directly coupled to a flow capillary of a SAXSess instrument (Kratky type of camera). SAXS curves were recorded in a 1 s time interval. This was possible by using intense synchrotron radiation. The radii of gyration of the nanoparticles, as determined from Guinier plots, increased from 2 to 6 nm with increasing fractionation time of the A4F. A more detailed analysis of the scattering curves revealed that the particles were cylindrical in shape (nanorods), which we attributed to the micromixing preparation technique. The radii of the nanorods increased only slightly from 1.2 to 1.7 nm with increasing fractionation time, while the lengths increased strongly from 7.0 to 30.0 nm. The volume distribution of the nanorods was determined and described by Schultz-Zimm and log-normal distributions. Nanorod volumes increased from 45 to 263 nm3, corresponding to molar masses of 140 × 103 to 820 × 103 g mol-1. We propose A4F-SAXS coupling as a new method for analysis of nanoparticles of complex composition in solution. It allows precise online determination of the particle’s shape and size distributions. This method can be applied to mixtures of nanoparticles of arbitrary shapes and sizes (1-100 nm). Moreover, the total time needed for fractionation and online SAXS data recording is usually only 20 min. Nanoparticles are of tremendous importance in many fields of nanotechnology but also in the classical field of colloid and polymer science. Application ranges from solar cells1 to target specific drug delivery,2 gene therapy,3,4 and cancer diagnosis.5,6 * To whom correspondence should be addressed. E-mail: andreas. [email protected]. (1) Kondo, Y.; Yoshikawa, H.; Awaga, K.; Murayama, M.; Mori, T.; Sunada, K.; Bandow, S.; Iijima, S. Langmuir 2008, 24 (2), 547–550. (2) Solaro, R. J. Polym. Sci., Part A: Polym. Chem. 2008, 46 (1), 1–11. (3) Weiss, S. I.; Sieverling, N.; Niclasen, M.; Maucksch, C.; Thunemann, A. F.; Mohwald, H.; Reinhardt, D.; Rosenecker, J.; Rudolph, C. Biomaterials 2006, 27 (10), 2302–2312. (4) Rudolph, C.; Sieverling, N.; Schillinger, U.; Lesina, E.; Plank, C.; Thunemann, A. F.; Schonberger, H.; Rosenecker, J. Biomaterials 2007, 28 (10), 1900–1911. 10.1021/ac8004814 CCC: $40.75  2008 American Chemical Society Published on Web 07/09/2008

Along with their benefits, serious concerns also arise about safety risks of nanoparticles to human health and the environment.7–10 Consequently, implementation of sustainable and reliable analytical methods in nanoparticle characterizations is struggling to keep pace with developments. Fundamental knowledge of nanoparticle size, shape, number, composition, and structure is of the utmost importance as there is no doubt that these quantities determine efficiency in technical applications as well as risk potential. It is often necessary for analytical purposes to calibrate methods with reference materials, which are rare in nanotechnology. Synthesis, protection, functionalization, and development of applications of superparamagnetic nanoparticles are currently the object of a tremendous amount of research.11 However, fine control of the properties of superparamagnetic particles remains challenging, especially, for example, in complex biological systems as reviewed by Majewski and Thierry.12 Standard superparamagnetic nanoparticles are currently not available. Coprecipitation is a convenient method to synthesize superparamagnetic iron oxide nanoparticles in the form of magnetite or maghemite from aqueous Fe2+/Fe3+ salt solutions by the addition of a base.13 Size, shape, and composition of the nanoparticles depend on a number of different parameters such as pH values, ionic strength of the media, the Fe2+/Fe3+ ratio and concentrations, and type of salts used (chlorides, sulfates, nitrates).11 Once the synthetic conditions are fixed for the coprecipitation synthesis, the quality of the superparamagnetic nanoparticles is reproducible. Unfortunately, these conditions are often very unique to the experimental conditions used. An up-scaling (5) Ferrari, M. Nat. Rev. Cancer 2005, 5 (3), 161–171. (6) Jain, P. K.; El-Sayed, I. H.; El-Sayed, M. A. Nano Today 2007, 2 (1), 18– 29. (7) Tsuji, J. S.; Maynard, A. D.; Howard, P. C.; James, J. T.; Lam, C. W.; Warheit, D. B.; Santamaria, A. B. Toxicol. Sci. 2006, 89 (1), 42–50. (8) Maynard, A. D.; Aitken, R. J.; Butz, T.; Colvin, V.; Donaldson, K.; Oberdorster, G.; Philbert, M. A.; Ryan, J.; Seaton, A.; Stone, V.; Tinkle, S. S.; Tran, L.; Walker, N. J.; Warheit, D. B. Nature 2006, 444 (7117), 267–269. (9) Maynard, A. D. Ann. Occup. Hyg. 2007, 51 (1), 1–12. (10) Balbus, J. M.; Maynard, A. D.; Colvin, V. L.; Castranova, V.; Daston, G. P.; Denison, R. A.; Dreher, K. L.; Goering, P. L.; Goldberg, A. M.; Kulinowski, K. M.; Monteiro-Riviere, N. A.; Oberdorster, G.; Omenn, G. S.; Pinkerton, K. E.; Ramos, K. S.; Rest, K. M.; Sass, J. B.; Silbergeld, E. K.; Wong, B. A. Environ. Health Perspect. 2007, 115 (11), 1654–1659. (11) Lu, A. H.; Salabas, E. L.; Schuth, F. Angew. Chem., Int. Ed. 2007, 46 (8), 1222–1244. (12) Majewski, P.; Thierry, B. Crit. Rev. Solid State Mater. Sci. 2007, 32 (34), 203–215. (13) Bee, A.; Massart, R.; Neveu, S. J. Magnet. Magnet. Mater. 1995, 149 (12), 6–9.

Analytical Chemistry, Vol. 80, No. 15, August 1, 2008

5905

of these syntheses in batch processes is very difficult. In addition, a systematic size and/or shape variation of the nanoparticles is almost impossible. A recent review on different preparation methods is given by Schu¨th et al.11 In our earlier work we used a layer-by-layer technique in a batch process to produce protectively coated superparamagnetic maghemite nanoparticles, which were tested as an MRT contrast agent.14 However, the size distribution of these nanoparticles was not very narrow. The use of a micromixer allows continuous preparation of nanoparticles by precipitation reactions: For details of the micromixing technique see refs 15 and 16. The technique will be used in this work for the production of superparamagnetic maghemite (see experimental section). Our long-term aim is to develop a synthesis for superparamagnetic maghemite nanoparticles that have the potential to be useful as nanoparticle standard materials. However, for a systematic optimization of the desired nanoparticles, it is necessary to have a fast and precise analytical method for particle size and shape determination. Two main techniques are currently dominating particle size determination: electron microscopy and dynamic light scattering (DLS). Scanning and transmission electron microscopy are intensively utilized but inherent drawbacks are random sampling instead of overall sample analysis, time-consuming sample preparation procedures, and the danger of producing measurement artifacts. Stabilizers of nanoparticles can, for example, cover the nanoparticles completely.17 Furthermore, samples cannot be analyzed in situ with electron microscopy. DLS is also extensively used because it is a noninvasive and nondestructive fast method. It produces results on the particle sizes within minutes. The major obstacle to attaining reliable results is the underlying principle of DLS, since it measures an effective z average of the diffusion coefficient. Sizes derived from this diffusion coefficient are influenced by the presence of dust and agglomerated fractions in the sample. Furthermore, a number of assumptions inherent in the DLS analysis make it impossible to determine particle size distributions precisely. Therefore, a fractionation of the sample is necessary if detailed information on size distribution is of interest. The situation is similar for small-angle X-ray scattering. SAXS is a state-of-the-art method for the determination of nanoparticle size, shape, number, and even the internal structure of nanoparticles in solution without the need of preparation steps.18–21 However, SAXS data interpretation is often ambiguous if the particle sizes are very polydisperse and/or differ in their shape. This is due to the fact that SAXS averages the scattering pattern of all the particles in the sample. In short, SAXS is normally very accurate as long as all particles are of similar size and shape. Therefore, fractionation of the particle mixtures before SAXS is (14) Thunemann, A. F.; Schutt, D.; Kaufner, L.; Pison, U.; Mohwald, H. Langmuir 2006, 22 (5), 2351–2357. (15) Ehrfeld, W.; Golbig, K.; Hessel, V.; Lowe, H.; Richter, T. Ind. Eng. Chem. Res. 1999, 38 (3), 1075–1082. (16) Haverkamp, V.; Ehrfeld, W.; Gebauer, K.; Hessel, V.; Lowe, H.; Richter, T.; Wille, C. Fresenius J. Anal. Chem. 1999, 364 (7), 617–624. (17) Kreuter, J. Int. J. Pharm. 1983, 14 (1), 43–58. (18) Glatter, O.; Kratky, O., Small Angle X-ray Scattering; Academic Press: London, U.K., 1982. (19) Fritz, G.; Glatter, O. J. Phys.: Condens. Matter 2006, 18 (36), S2403–S2419. (20) Konarev, P. V.; Petoukhov, M. V.; Volkov, V. V.; Svergun, D. I. J. Appl. Crystallogr. 2006, 39, 277–286. (21) Svergun, D. I. J. Appl. Crystallogr. 2007, 40, S10–S17.

5906

Analytical Chemistry, Vol. 80, No. 15, August 1, 2008

necessary for unambiguous results. Asymmetrical flow field-flow fractionation (A4F) is a suitable method for gentle fractionation (low sheer forces compared to GPC and HPLC) before determination of the sizes, which can be determined, for example, with multiangle light scattering (MALS). An online A4F-MALS system has been demonstrated for the first time by Kulicke et al.22 for polystyrene nanoparticles and coilshaped dextrans. Later they investigated the accuracy using polystyrene latex standards.23 Fraunhofer et al. applied an online A4F-MALS system for the analysis of gelatin nanoparticle drug carriers.24 An A4F-MALS was also successfully used for the characterization of chitosan25 and polyorganosiloxane nanoparticles.26 To the best of our knowledge, SAXS has not yet been combined with field flow fractionation. The absence of studies on the coupling of field-flow fractionation with SAXS is most probably due to the fact that the scattering of visible light by nanoparticles is several orders of magnitude higher than that of X-rays. A priori, one may expect that the SAXS signal of the nanoparticles after field-flow fractionation is too low to be detected. Nevertheless, we believe that this problem can be circumvented when a highly intense X-ray source in the form of a synchrotron is used in combination with a SAXS system optimized for nanoparticle detection. The aim of this work is to show that an A4F-SAXS coupling allows fast determination of the shape and size distribution of very small anisotropic nanoparticles. MATERIALS AND METHODS Materials. Ammonium hydroxide, ferric chloride hexahydrate (FeCl3 · 6H2O), and ferrous chloride tetrahydrate (FeCl2 · 4H2O) were obtained from Fluka, Germany. Carboxydextran (CDx) was purchased from Meijto Sangyo Co. Ltd., Japan. NovaChem Surfactant 100 was obtained from Postnova Analytics GmbH, Germany. Nanoparticle Preparation. The maghemite nanoparticles were prepared by adapting the precipitation method of Bee13 using a micromixer with two influxes (IMM GmbH, Mainz, Germany). The first stream was an aqueous solution of a mixture of Fe(II)Cl2 4H2O and Fe(III)Cl3 6H2O with a 2:1 molar ratio of Fe(III) to Fe(II) and a total iron amount of 0.13 mol L-1 (7.1 g L-1). Carboxydextran was also added (weight ratio of iron salts to carboxydextran 1:3). The second stream was a 25% aqueous ammonium hydroxide solution. Streams one and two were mixed in a volume ratio of 19 to 1. Methods Asymmetrical Flow Field-Flow Fractionation (A4F). The A4F was from Postnova Analytics GmbH (Germany) and consisted of an AF 2000 focus system (PN 5200 sample injector, PN 7505 inline degaser, PN 1122 tip and focus pump). An inline solvent filter (100 nm, regenerated cellulose, Postnova) was placed between the tip and focus pump and the channel to reduce the background signal. The channel thickness was 500 µm, and the ultra filtration membrane was a regenerated cellulose membrane (22) Roessner, D.; Kulicke, W. M. J. Chromatogr., A 1994, 687 (2), 249–258. (23) Thielking, H.; Roessner, D.; Kulicke, W. M. Anal. Chem. 1995, 67 (18), 3229–3233. (24) Fraunhofer, W.; Winter, G.; Coester, C. Anal. Chem. 2004, 76 (7), 1909– 1920. (25) Mao, S.; Augsten, C.; Mader, K.; Kissel, T. J. Pharm. Biomed. Anal. 2007, 45 (5), 736–741. (26) Jungmann, N.; Schmidt, M.; Maskos, M. Macromolecules 2001, 34 (23), 8347–8353.

with a cutoff of 10 × 103 g mol-1. The detector flow rate was 0.5 mL min-1 for all A4F experiments, and the cross-flow rate were controlled via AF2000 software (Postnova Analytics). The crossflow decreased linearly with time, starting with a cross-flow of 2.5 mL min-1 and decreasing at a rate of 2.5/70 mL min-2. A sample of nanoparticles with a concentration of 0.71% (w/v) of iron (1.0% (w/v) of maghemite) and a volume of 0.02 mL was injected into the channel for each experiment. The outlet of the A4F was coupled directly to the UV-detector (detection at 300 nm) and the flow cell of the SAXS machine. The fractionated stream reached the UV detector about 2 min before it passed through the flow cell of the SAXS machine. Small-Angle X-ray Scattering (SAXS). SAXS measurements were performed at the BAMline at BESSY II (Berlin, Germany) with a Kratky-type instrument (SAXSess from Anton Paar, Austria). The SAXSess has a low sample-to-detector distance which is suitable for investigation of low scattering intensities. The measured intensity was corrected by subtracting the intensity from a capillary filled with pure water. The scattering vector is defined in terms of the scattering angle θ and the wavelength λ of the radiation (λ ) 0.154 nm): thus q ) 4π/λ sin(θ/2) . Desmearing the SAXS curves was performed with the SAXS-Quant software (version 2.0) from Anton Paar. Dynamic Light Scattering (DLS). The DLS measurements were carried out using a Malvern Instruments particle sizer (Zetasizer Nano ZS, Malvern Instruments, U.K.) equipped with a He-Ne laser (λ ) 632.8 nm). The scattering data were recorded at 298 K in a backscattering modus at a scattering angle of 2θ ) 173°, which corresponded to a scattering vector of q ) 4πn/λ sin θ (0.02636 nm-1). The aqueous sample solutions were placed into a square 10 × 10 mm disposable polystyrene cuvette. Prior to measurement, the sample was filtered with a 0.45 µm Millipore syringe filter to clear it of dust particles. The hydrodynamic radius, Rh, (of a hydrodynamic equivalent sphere) was obtained from the diffusion coefficient using the Stokes-Einstein relation Rh ) kT/ (6πηD) with the viscosity of water being η ) 0.9387 mPa s at a temperature of 296 K. Transmission Electron Microscopy (TEM). Standard TEM measurements were carried out with a Phillips EM 420, equipped with an ORCA-ER camera (1024 × 1024) and run with AMT Image Capture Engine v5.42.540a (120 kV). Samples were dried on carbon filmed TEM grids.

Figure 1. Sketch of the experimental setup of the coupling of A4F (top) and SAXS (bottom). Particle fractionation results from a crossflow which is superpositioned to a parabolic flow profile. Particles leave the outlet of the A4F in the order of small, medium sized, and large particles. The stream of the fractionated particles is first detected by UV absorption at 300 nm and 140 s later in a flow cell by SAXS (detection time 1 s).

nately, we had no information on the shape of the nanoparticles and only a vague estimate of their size distribution by the DLS. Therefore, we performed a more detailed analysis by coupling A4F particle fractionation with SAXS shape analysis. The nanoparticle analysis by A4F-SAXS coupling was based on the following strategy: the fractionation process of the A4F provides nanoparticles with a narrow polydispersity ideal for SAXS analysis. This is done online with the help of a flow cell and intense synchrotron radiation. A sketch of the experimental setup is shown in Figure 1. An inherent property of the A4F is the separation of nanoparticles according to their diffusion coefficient. The principle of A4F has been described in a number of reviews.28 Briefly, the nanoparticles are injected and fractionated in a separation channel. The parabolic flow profile of a laminar flow transports them along the channel toward the outlet and separation is caused by a second flow profile perpendicular to the laminar flow. This cross-flow leaves the channel through an ultrafiltration membrane covering one side of the separation channel (cf. upper left sketch in Figure 1). Smaller particles are eluded prior to the larger ones due to the difference in their diffusion coefficient. As an example, the retention time, tr, of gelatin nanoparticles24 is given by

RESULTS AND DISCUSSION The maghemite nanoparticles were prepared with the micromixer by precipitation of a 2:1 molar ratio of Fe(III) and Fe(II) ions to superparamagnetic magnetite nanoparticles which oxidized within several days to maghemite in the presence of air as had already been proved by Mo¨ssbauer spectroscopy.14 Carboxydextran was used as a polymeric stabilizer, which has been proved to provide long-term stability and biocompatibility of similar particles, for example, in medical applications.27 The intensity weighted hydrodynamic radius as determined by DLS was Rh ) 9.8 nm with a polydispersity index PDI ) 0.153. Volume averaged hydrodynamic radius was 6.3 nm. This result is similar to the sizes and PDIs observed for batch preparation procedures.13 Unfortu-

where t0 is the retention time of the solute, vc the cross-flow rate, V0 the volume of the channel, w the channel thickness, and D the diffusion coefficient of the separated nanoparticles. Equation (1) must in general be considered as a rough estimation for the determination of the retention time especially if the particles are not spherical. In addition the flow conditions in the separation channel are more complicated in practice than in theory. For the current work, it is sufficient to assume that tris approximately inversely proportional to D.

(27) Lawaczeck, R.; Menzel, M.; Pietsch, H. Appl. Organomet. Chem. 2004, 18 (10), 506–513.

(28) Fraunhofer, W.; Winter, G. Eur. J. Pharm. Biopharm. 2004, 58 (2), 369– 383.

tr )

t0vcw2 , 6DV0

Analytical Chemistry, Vol. 80, No. 15, August 1, 2008

(1)

5907

Figure 2. Guinier fits of the SAXS data for t ) 444 s (Rg ) 2.42 nm, O), 644 (Rg ) 3.39 nm, 4), and 778 (Rg ) 4.13 nm, 0).

Figure 3. Radii of gyration (O) and SAXS intensity (0, from extrapolation to q ) 0 using the Guinier approximation) as a function of the fractionation time. For clarity, only every tenth data point is plotted.

We observed particle scattering in a fractionation time interval from 330 to 1130 s. The scattering intensity from nanoparticles in a diluted solution can be analyzed at low q-values by the Guinier approximation. Guinier29 has shown that the scattering of nanoparticles follows an exponential decay where the steepness of the decay is determined by the nanoparticles’ radius of gyration, Rg. Normally, the approximation is used in its linear form 1 ln[I(q)] ) ln[I(0)] - Rg2q2 3

(2)

to determine Rg and the intensity at q ) 0. The ln [I(q)] of an experimental curve is plotted as a function of q2 for this purpose. If a straight line can be fitted to the data, the slope is equal to 1 /3Rg2 and the section with the ordinate is ln [I(0)]. Note that the Guinier approximation is valid only for small qRg. In practice, reliable fits can be obtained by using a q-range such that qRg < 1.3.30 Guinier fits presented in this work use a fitting range such that qRg e 1.2, as shown for example in Figure 2 for t ) 444, 644, and 778 s with Rg values of 2.42, 3.39, and 4.13 nm, respectively. An overview of the radii of gyration and I(0) as a function of the fractionation time is given in Figure 3. In this case, it can be seen that the first detectable particle scattering occurs for the smallest particles at t ) 330 s (Rg ) 1.97 nm), the maximum intensity is for medium sized particles at 770 s (Rg ) 3.96 nm), and the largest particles are detected at 1130 s (Rg ) 6.11 nm). (29) Guinier, A., X-Ray Diffraction in Crystals, Imperfect Crystals, and Amorphous Bodies; Dover Publications, Inc.: New York, 1994. (30) Svergun, D. I.; Koch, M. H. J. Rep. Prog. Phys. 2003, 66 (10), 1735–1782.

5908

Analytical Chemistry, Vol. 80, No. 15, August 1, 2008

Figure 4. SAXS curves of fractions of nanoparticles observed at fractionation times of 444 s, lower red curve), and 777 s (upper gray curve). The straight solid lines indicate the cylinder typical q-1 scaling of the intensity in the intermediate q-region. Fit curves from the cylinder model give R ) 1.22 nm and L ) 9.87 nm for the sample at t ) 444 s (dashed curve) and R ) 1.52 nm and L ) 16.43 nm for the sample at t ) 777 s (dashed curve). The recording time for both curves was 1 s.

Nanoparticle Morphology. The Guinier approximation is easy to handle for large sets of SAXS curves but gives no direct information on the shape of the nanoparticles. A reasonable method for shape determination is the direct fitting of model curves to experimental scattering curves. A detailed review is given by Pedersen.31 The scattering intensity from noninteracting individual particles is determined by their form factor P(q). In our case, the particles are already diluted when injected into the A4F (iron content 7.1 g/L) and further diluted by a factor of about 100 (iron content of 0.07 g/L) during fractionation. The scattering intensity scales at q-1 in the range of 0.5 < q < 1.0 nm-1, as indicated by a straight line in the double logarithmic plots shown in Figure 4. This scaling is typical for cylindrical or rodlike nanoparticles. Nevertheless, we tentatively tried to fit the scattering curves to model curves of monodisperse and polydisperse spheres,31 which clearly failed (not shown). Therefore, as a next step we used the cylinder model, which was shown to be suitable by the q-1 scaling of the intensity. The scattering intensity of a cylinder with radius R and length L is given as31

k(F - Fs)2 I(q) ) V

∫

π⁄2

0

[

1 2J1(qR sin R) sin 2 qL cos R qR sin R 1 qL cos R 2

(

)

]

2

sin R dR (3)

where k is the scaling factor, F and Fs the electron densities of the cylinder and the solvent, respectively, and V the volume. In this case the electron density of the cylinder is assumed to be constant. The cylinder model fits all measured scattering curves very well, as shown for example in Figure 4 for two curves from samples at fractionation times of 444 and 777 s. Note that the data quality is outstanding considering the data recording time of 1 s. The resultant radii and lengths as a function of fractionation time are summarized in Figure 5 (error bars are smaller than symbol sizes). It can be seen that the radii of the cylinders vary between (31) Pedersen, J. S. Adv. Colloid Interface Sci. 1997, 70, 171–210.

Figure 5. The radii (4) and lengths (O) of the cylindrical nanoparticles as a function of the fractionation time (left and right axis, respectively). The increase of the radii with time is approximated as a linear (dashed line) and the length as a quadratic function (solid line).

1.2 and 1.7 nm. The slight increase of the radii can be approximated by a linear function as R(t) ) 1.19 nm + (4.3 × 10-4nm s-1)(t) (dashed line in Figure 5). In contrast to the radii, the lengths of the cylinders are seen to increase strongly with fractionation time. The smallest lengths are 7 nm at t ) 333 s and the largest 30 nm at t ) 1133 s. The increase of the cylinder lengths can be approximated by a quadratic equation as L(t) ) 5.55 nm + (1.78 × 10-3nm s-1)(t) + (1.61 × 10-5nm s-2)(t2) (solid line in Figure 5). It is surprising that the shape of the nanoparticles is very anisotropic with a maximum length to radius ratio of 20. This strong anisotropy obviously results from the micromixer preparation technique. Earlier studies have shown that nanocrystals of BaSO4 and boehmite with narrow size distributions can be produced with microfuidic reactors.32 CdSe nanocrystals have been produced with size distributions comparable to those for conventional macroscale synthesis.33 The size distributions of nanoparticles generated with a micromixer have also been calculated theoretically34 and compared with experimental results.35 The conditions to predict sizes and size distributions are not fully understood. Nevertheless, it is clear that supersaturation is the key parameter for the resulting nanoparticle. However, we are aware of no study showing the formation of strongly anisotropic nanoparticles when a micromixer is used as observed here for maghemite. The laminar flow in the micromixer, which is directed from the precipitation zone to its outlet, is probably the reason for the cylindrical geometry of the nanoparticles. Particle Volume and Molar Mass Distribution. Of great interest is the size distribution when nanoparticles are fractionated. This knowledge is very helpful in judging the quality of the nanoparticles and systematically improving the preparation procedure. The most convenient detection for A4F is UV, which is a fast and cheap online detection method. An example is shown here in Figure 6 for a fractogram using UV absorption detection at 300 nm. The same fractogram displays the SAXS intensity derived from Guinier plots (0) and scale factors from fits of eq (32) Ying, Y.; Chen, G. W.; Zhao, Y. C.; Li, S. L.; Yuan, Q. Chem. Eng. J. 2008, 135 (3), 209–215. (33) Chan, E. M.; Mathies, R. A.; Alivisatos, A. P. Nano Lett. 2003, 3 (2), 199– 201. (34) Gradl, J.; Schwarzer, H. C.; Schwertfirm, F.; Manhart, M.; Peukert, W. Chem. Eng. Process. 2006, 45 (10), 908–916. (35) Schwarzer, H. C.; Peukert, W. AIChE J. 2004, 50 (12), 3234–3247.

Figure 6. A4F-SAXS analysis of superparamagnetic nanoparticles, detection via UV at 300 nm (solid line), SAXS intensity from Guinier plots (0), and the scale factors of the cylinder fits (4). The maximum UV signal is at t1 ) 610 s and the maximum SAXS intensity at t2 ) 750 s. Maxima are normalized to 1.

Figure 7. Frequency distribution of the nanorods (0) as a function of their volumes (lower x-axis) and masses (upper x-axis), respectively. The lines are fit cures according to a Schultz-Zimm distribution (solid line) and a log-normal distribution (dotted line).

(3) (4). As can be seen, the shapes of UV absorption and SAXS intensity as a function of time are very similar. The main difference is a time delay between the UV and SAXS curves with a maximum at t1 ) 610 s and t2 ) 750 s, respectively. The reason for the difference between t2 and t1 is the time it takes the stream of particles to move from the UV detector (first detection system) to the SAXS capillary (second detection system, cf. Figure 1). Knowledge of SAXS intensities and nanorod radii and lengths allows us to calculate the volume distribution of the nanorods. In addition, the molar mass distribution of the nanorods can be estimated by multiplying the nanorods’ volumes with the bulk density of maghemite (4.89 g cm-3) and the Avogadro constant. The results for volume and molar mass distribution are shown together in Figure 7 (lower and upper x-axis, respectively). It can be seen that the smallest nanorods have a volume of 45 nm3 that contain about 1650 iron atoms. This corresponds to a molar mass of M ) 1.4 × 105 g mol-1. The maximum is at 97 nm3 (containing 3560 iron atoms, M ) 3.0 × 105 g mol-1), and the largest nanorods have a volume of 263 nm3 containing 9640 iron atoms (M ) 8.2 × 105 g mol-1). It should be mentioned that the molecular weights and number of iron atoms in a nanorod are upper estimates. The density of maghemite in nanorods must be assumed to be lower than the Analytical Chemistry, Vol. 80, No. 15, August 1, 2008

5909

Figure 8. TEM figures of the iron oxide nanoparticles. Scale bars from left to right are 100 and 20 nm.

density in bulk, mainly because of the high number of atoms near the particles’ surfaces. More precise values of the molecular weights can be determined, at least in principle, as shown by Orthaber and Glatter,36 but this is beyond the scope of the current work. TheparticlevolumedistributionhasbeenfittedbyaSchulz-Zimm distribution

h2(V) )

[

(z + 1)z+1Vz V exp -(z + 1) z+1 〈V〉 〈V〉 Γ(z + 1)

]

(4)

for a quantitative description. In this case, we have the mean nanorod volume 〈V〉 and the relative standard deviation σ ) (z + 1)-1/2. For comparison we also used a log-normal distribution

h1(V) )

A

√2πσV

[

exp -

ln(V ⁄ V0)2 2σ2

]

(5)

where V0 is the mean volume and σ the standard deviation. The fitted curves are shown in Figure 7 for the Schultz-Zimm (solid line) and log-normal distribution (dotted line). The mean nanoparticle volume is 99 ± 3 nm3 with a width of σ ) 0.50 for the Schulz-Zimm distribution. Values for the log-normal distributions are V0 ) 128 ± 5 nm3 and σ ) 0.55. Both distributions are suitable here for a description of the volume distribution as the reduced χ2of the fits for Schultz-Zimm and log-normal that are very similar (0.0197 and 0.0172). The 〈V〉 value is significantly lower than V0, which is a consequence of the extended tail of the log-normal distribution for large volumes. Tentatively we assume 〈V〉 to be more realistic because of the smaller tailing of the Schulz-Zimm distribution. The width of both distributions, as described by σ, is very similar for both distributions. Comparing SAXS and TEM. The samples were investigated for comparison by transmission electron microscopy. Surprisingly we cannot detect cylinders in the images as must be expected from our SAXS investigations. Instead of cylinders we always find monodisperse spherical objects with radii in the range of 10-15 (36) Orthaber, D.; Bergmann, A.; Glatter, O. J. Appl. Crystallogr. 2000, 33, 218– 225.

5910

Analytical Chemistry, Vol. 80, No. 15, August 1, 2008

Figure 9. SAXS of nonfractionated nanoparticles (solid line) which are identical to the particles investigated by TEM (see Figure 8). The SAXS simulation of spherical nanoparticles is from particles with a Schulz-Zimm distribution with a mean radius of 12 nm and polydispersity of 0.3 (dashed line). The best fit curve is from cylindrical nanoparticles with a length of 29 nm (dotted line). The cylinder radii are 1.3 nm, and the polydispersity is 0.3.

nm as shown in Figure 8 (shown are particles at different magnifications). At first sight, this might be surprising but it is well-known that the presence of surface active compound produce often very strong artifacts in TEM pictures.17 In our case, the large amount of stabilizing carboxydextran may be reasonable for the disagreement between SAXS and TEM. Tentatively we interpret the presence of spherical particles in the TEM pictures as an artifact from the standard TEM specimen preparation of the nanoparticles. We assume that the forces of minimizing the surface tension during drying of the nanoparticles on the TEM grid are responsible for the formation of spherical objects. The structure of the particles in solution is different from there structure in the dried state. To make this more clear, we simulated the SAXS curve of spherical nanoparticles as found with TEM and compared them with the SAXS measurement of the same sample. It is obvious from Figure 9 that the simulated SAXS curve from spheres with a mean radius of 12 nm and a polydispersity of 0.3 (dashed line) is in strong disagreement with the measured curve (solid line). We also tried to fit the data with spheres of different sizes and polydispersities, but no reasonable parameter combination was found that results in a good fit. It should be mentioned that at least in principle any SAXS curve can be constructed as a superposition of spherical objects,18 but we can conclude that the

measured SAXS cannot be a result from spherical nanoparticles seen in the TEM figures. Instead of that, a simple cylinder model with cylinder length of 29 nm and a radius of 1.3 nm with a polydispersity of the radii of 0.3 fits the measured SAXS relatively well. On the basis of the SAXS of the nonfractionated sample alone, there may be doubt about the structure of the particles. However, the results from the A4F-SAXS coupling make it very unlikely that the particles are spherical as indicated by TEM. It should be mentioned that extensive investigations are necessary in the future to compare in detail results from A4F-SAXS and TEM. It has to be taken into account that TEM produces a number of artifacts including the contrast transfer function and envelope functions. In addition, cryo electron microscopy should be performed for purposes of single particle reconstruction, but here the signal-tonoise-ratios are very low. CONCLUSIONS We have shown that the coupling of asymmetric flow fieldflow fractionation with small-angle X-ray scattering as a combined instrument (A4F-SAXS) is possible for the characterization of nanoparticles. This allows a detailed but also time-saving analysis of complex mixtures of nanoparticles of sizes in the range of 1-100

nm. The micromixing preparation technique enables continuous production of strongly anisotropic nanorods of maghemite with a maximum aspect ratio (length-to-radius) of 20. The radii of the nanorods are relatively monodisperse, while the length distribution is very broad. Preparation of the nanorods for the desired monodisperse size distribution must be significantly improved. We believe that A4F-SAXS coupling is a very helpful analytical instrument for this purpose which complements, and/or substitutes for the dynamic light scattering and electron microscopy analysis. ACKNOWLEDGMENT The financial support of the Federal Institute for Materials Science and Testing is gratefully acknowledged. We also thank M. Klinger for drawing the sketch of A4F-SAXS coupling in Figure 1, S. Rolf and H. Schnablegger for their help in SAXS at BESSY, S. Weidner for help in A4F, and Stephan Wolf for TEM.

Received for review March 7, 2008. Accepted June 4, 2008. AC8004814

Analytical Chemistry, Vol. 80, No. 15, August 1, 2008

5911