Quantitative 3D evolution of colloidal nanoparticle oxidation in solution

See allHide authors and affiliations

Science  21 Apr 2017:
Vol. 356, Issue 6335, pp. 303-307
DOI: 10.1126/science.aaf6792

Watching nanomaterials transform in time

Real-time analysis of chemical transformations of nanoparticles is usually done with electron microscopy of a few particles. One limitation is interference by the electron beam. Sun et al. monitored the oxidation of iron nanoparticles in solution by using small- and wide-angle x-ray scattering and molecular dynamics simulations (see the Perspective by Cadavid and Cabot). These methods revealed the formation of voids within the nanoparticles, diffusion of material into and out of the nanoparticles, and ultimately the coalescence of the voids.

Science, this issue p. 303; see also p. 245


Real-time tracking of the three-dimensional (3D) evolution of colloidal nanoparticles in solution is essential for understanding complex mechanisms involved in nanoparticle growth and transformation. We used time-resolved small-angle and wide-angle x-ray scattering simultaneously to monitor oxidation of highly uniform colloidal iron nanoparticles, enabling the reconstruction of intermediate 3D morphologies of the nanoparticles with a spatial resolution of ~5 angstroms. The in situ observations, combined with large-scale reactive molecular dynamics simulations, reveal the details of the transformation from solid metal nanoparticles to hollow metal oxide nanoshells via a nanoscale Kirkendall process—for example, coalescence of voids as they grow and reversal of mass diffusion direction depending on crystallinity. Our results highlight the complex interplay between defect chemistry and defect dynamics in determining nanoparticle transformation and formation.

Metal oxidation has been extensively studied because of its broad relevance in areas such as catalysis and corrosion, which can dramatically influence the properties and stability of materials, as well as for the fabrication of functional metal oxides. For example, forming iron oxide (FexOy) nanoparticles (NPs) through the oxidation of iron NPs is an important reaction that bears on many technological areas, such as clean fuels (1), catalysis (2), and electrochemical energy storage (3, 4). The Fe oxidation state, geometry, crystallinity, and composition of the NPs strongly depend on the oxidation process and play important roles in determining their application performance. Therefore, real-time monitoring of the oxidation reaction—in particular, oxidation of colloidal NPs in solution—is critical to track the detailed evolution of different parameters of NPs upon oxidation and to synthesize oxide NPs that deliver appropriate performance. In situ transmission electron microscopy (TEM) has been developed to monitor the growth (58), assembly (9), oxidation (10), and ripening (11) of nanocrystals in thin liquid TEM cells. However, it is still fundamentally difficult to study amorphous NPs in large-volume solutions under ambient environmental and complex reaction conditions and to obtain three-dimensional (3D) structures with high spatial resolution (12, 13). On the other hand, various in situ x-ray techniques have been developed recently to probe nanoparticle growth and transformation in liquid environments in real time by taking advantage of the high penetration power of synchrotron x-rays (1416).

We synthesized colloidal Fe NPs through thermal decomposition of iron pentacarbonyl [Fe(CO)5] in 1-octadecene containing oleylamine as a stabilizer under N2 (17). Exposing the dispersion of Fe NPs to the ambient environment oxidized surface Fe atoms to quickly form an FexOy layer with a thickness of 2 to 3 nm that prevented continuous oxidation of the Fe NPs. The Fe core–FexOy shell NPs (Fig. 1, A and B) were amorphous and highly uniform in size and morphology (fig. S1). Further oxidizing the NPs converted them to core-void-shell particles (Fig. 1, C and D) and shells with completely hollow interiors (Fig. 1, E and F) through a nanoscale Kirkendall effect at elevated temperatures (100° to 180°C; Fig. 1H) (18). The initially formed thin FexOy layer was necessary to provide a matrix in which the diffusion of iron and oxygen could be differentiated to drive the nanoscale Kirkendall process. The unbalanced diffusion of iron and oxygen in the solid FexOy layer with outward iron diffusion faster than inward oxygen diffusion led to the formation of iron oxide nanoparticles with hollow interiors. The amorphous hollow nanoshells crystallized under incubation at even higher temperatures. For example, FexOy nanoshells that formed after annealing at 220°C under N2 exhibited the crystalline lattices of Fe3O4 (Fig. 1G). The high uniformity of size and morphology was maintained during the multiple-step oxidation process (figs. S2 to S5). Taking advantage of the high quality of the colloidal Fe NPs, we applied the in situ small-angle x-ray scattering (SAXS) technique—which is sensitive to NP size, morphology, and electron density—to precisely track the oxidation process of the colloidal Fe-FexOy NPs in real time.

Fig. 1 TEM characterization of products formed from oxidation of iron NPs.

(A to G) TEM images of NPs at different oxidation stages. TEM imaging cannot provide the details of vacancies between the solid cores and shells [(C) and (D)]. Imaging the NPs with 3D TEM tomography is impossible because of the instability of amorphous Fe-FexOy NPs under long-time exposure to the electron beam. (H) Plot of the programmed temperature and reaction atmosphere as a function of reaction time. Temperature was increased by 10°C/min in the sloping portions of the plot. The samples presented in the TEM images were collected at times indicated by the arrows. The scale bar in (A) is 10 nm and applies to (C) and (E); the scale bar in (B) is 2 nm and applies to (D), (F), and (G).

The SAXS pattern of these as-synthesized NPs exhibits at least five well-defined Bessel oscillation peaks and valleys (Fig. 2, bottom black curve), confirming their spherical shape and uniform size (16). The position of the first valley determines the maximum lateral dimension of the solid NPs: The lower the corresponding value of q (the norm of the scattering vector, q; i.e., q = |q|) is, the larger the NPs are. Once the NPs are oxidized at elevated temperatures, the position of the first valley shifts to the lower q, indicating an increase in NP size. The enlargement of NPs implies that Fe atoms diffuse outward to react with oxygen and form iron oxides. The SAXS patterns of NPs were continuously recorded (Fig. 2 and figs. S6 to S9) at an interval of 30 s over the course of the oxidation. As the reaction time increases, the shift of the first valley position becomes slower, corresponding to a minor increase in the particle size. The five distinct SAXS peaks remain throughout the oxidation reaction—although their positions, widths, and intensities change gradually—indicating that the NPs remain highly uniform. The fourth and fifth peaks disappear at 10 min, and they recover again at longer times (highlighted with yellow shading, Fig. 2). The temporary disappearance of these two peaks corresponds to the initial formation of voids and vacancies, leading to nonuniform and asymmetric mass distribution in the NPs. Reconstruction (including coalescence, growth, and diffusion) of these voids and vacancies redistributes the material in NPs to form a more symmetric geometry, resulting in the recurrence of the high-order peaks.

Fig. 2 Time-resolved SAXS patterns.

SAXS patterns recorded from the colloidal Fe-FexOy NPs shown in Fig. 1A at different reaction times over the course of oxidation. For visual clarity, the SAXS patterns are plotted at an interval of 5 min and with arbitrary vertical offsets. The black curves indicate the times at which the reaction conditions were changed (see Fig. 1H). The yellow shading emphasizes the disappearance of the fourth and fifth peaks, which corresponds to the intermediate process associated with the initial formation of voids in the NPs. The wavelength of the x-ray beam was 1.033 Å.

For objects with high homogeneity in size and conformation, 3D molecular envelopes can be achieved from their SAXS pattern by using ab initio structure programs, such as DAMMIN and DAMMIF (19). Because SAXS data have degeneracy, instead of trying to solve the structure directly from the SAXS data, these programs search a molecular envelope with a SAXS profile matching the experimental data. The search is performed in a spherical space with a diameter determined by the maximum lateral dimension of the objects calculated from an experimental SAXS pattern (i.e., the corresponding pair-distance distribution function; fig. S10). Filling the spherical space with small solid beads (with radii of 5 Å in this study) produces a SAXS pattern that matches the experimental one. The high structural homogeneity of the Fe-FexOy NPs and the well-defined multiple Bessel oscillations in the SAXS patterns make it possible to calculate the 3D molecular envelopes of the NPs with DAMMIF (version 1.1.2) (19). Thirty runs of calculation were performed for each set of data, and the resulting bead models were further averaged using DAMAVER (version 5.0) (20) to produce the consensus, final bead model structures that represent the 3D geometries of the colloidal NPs at different reaction times.

Figure 3 highlights the 3D structural evolution of the NPs. At the early oxidation stage, the solid geometry of the NPs is preserved, whereas their size expands because of the outward diffusion of Fe atoms (Fig. 3A; 6.5 min). As the oxidation continues, the outward diffusion of more material leads to the formation of voids at the Fe/FexOy interface (13 min). These voids are well separated, and their sizes increase independently as the reaction continues (Fig. 3B and movie S1). The isolation of individual voids does not completely consume the Fe/FexOy interface, and the rest of it allows the continuous outward diffusion of Fe atoms to enlarge the voids. When individual voids in a NP are large enough, adjacent voids fuse together (190 min; Fig. 3C and movie S2), coalescing into a single void with a waxing crescent moon shape (195 min; movie S3). This asymmetric shape of the void ensures that the Fe atoms in the center of the NP are still able to diffuse outward through the solid portion (highlighted by the arrow in Fig. 3A), further enlarging the void. The crescent tips of the void merge to form a waxing gibbous moon shape, as long as the central Fe atoms completely diffuse out (275 min; movie S4). Incubation of the NPs at higher temperatures for longer times transforms the hollow interiors into a full moon shape (305 and 355 min; Fig. 3D and movie S5). The 3D geometries of NPs at different reaction times thus reveal the evolution process of the hollow interiors over the course of oxidation. These results are distinct from the previously proposed model, in which the internal material diffuses out through some thin filaments in single core-void-shell NPs (18, 21).

Fig. 3 3D structures of colloidal Fe-FexOy composite NPs over the course of the oxidation process.

(A to D) Structures reconstructed from SAXS patterns by using ab initio structural modeling and (E) typical snapshots from large-scale reactive MD simulations. The red arrow in (A) indicates the position of the Fe/FexOy interface. The cross-sectional geometries of the NPs at 135 and 190 min are shown projected from different orientations in (B) and (C), respectively. The former highlights the separation of individual voids (v1 to v4), whereas the latter emphasizes the interconnection of two voids (vv1 and vv3). The concentric cross-sectional geometry of the hollow nanoshell formed at 305 min is shown in (D). The scale bar in (A) is 5 nm and applies to (B) to (D). The scale bar in (E) is 5 nm. The structural models are presented as mesh models generated from PyMol (version 0.99) to clearly show the voids in the NPs.

We performed large-scale molecular dynamics (MD) simulations of the oxidation of Fe NPs with a diameter of 10 nm (supplementary methods and fig. S11). Simulation results depict the temporal transformation of a Fe NP to a hollow oxide shell over 150 ns (Fig. 3E). Initial oxidation of the Fe NP is rapid, leading to the formation of a thin oxide shell (time t < 1 ns). In the presence of the metal/oxide interface, continuous oxidation generates a large number of cationic vacancies (t ≈ 1 to 2 ns). The oxide growth proceeds via inward movement of oxygen anions and outward movement of iron cations across the metal/oxide interface. Analysis of the MD trajectories at various stages of oxide growth (fig. S12) suggests that the outward Fe diffusion is consistently much larger than the inward O diffusion. This difference in directional diffusion mobility between Fe and O leads to the formation of vacancies, which subsequently coalesce and form voids at the Fe/FexOy interface. Consistent with the experimental observations, these voids coalesce during oxidation so that the particle eventually becomes hollow. The simulations also reveal the asymmetric nature of the growing voids, which results in a high degree of openness and facilitates the outward Fe migration, in good agreement with the experiments.

Figure 4 shows the quantitative analysis of oxidation kinetics, as deduced from in situ experimental data and simulations. The extrapolated x-ray scattering intensity when q is equal to zero [I(q=0)] in a SAXS pattern depends on the electron density of the NP material. Comparing the value of I(q=0) for products formed at different reaction times provides the amount of oxygen deposited to the NPs during the reaction (supplementary methods). The amount of oxygen uptake increases, following a diffusion-limited behavior (i.e., Log3P1 function; figs. S13 and S14), at 100° and 140°C (Fig. 4A, black symbols). The maximum lateral dimension of the NPs follows a similar function as it increases (Fig. 4A, red symbols, and figs. S15 and S16). The diffusion rate of Fe atoms that is proportional to the rate of oxygen uptake strongly depends on the reaction temperature and the area of the Fe/FexOy interface. The average Fe/FexOy interfacial area in an individual NP is estimated from its calculated 3D configuration (supplementary methods). When the temperature ramps from 30° to 100°C at a rate of 10°C/min, the uptake rate of oxygen increases rapidly (Fig. 4B, squares). In contrast, the Fe/FexOy interfacial area remains essentially constant (Fig. 4B, circles), indicating the absence of voids at the Fe/FexOy interface. Once the Fe/FexOy interfacial area starts to decrease because of the formation of voids, the rate of oxygen uptake also decreases simultaneously, even though the temperature remains at 100°C. Both the Fe/FexOy interfacial area and the oxygen uptake rate quickly decrease within 15 min (Fig. 4B), indicating a high dependence of the oxidation rate on the Fe/FexOy interfacial area.

Fig. 4 Quantitative analysis of oxygen uptake kinetics.

(A) Dependence of the (relative) amount of oxygen taken up by the Fe-FexOy NPs (black squares) and the maximum lateral dimension of the NPs (red circles) on the reaction time. The programmed reaction condition is plotted to help correlate the jumps in reaction kinetics and the changes in reaction conditions. (B) Plot of the rate of oxygen uptake (black squares) and the Fe/FexOy interfacial area (red circles) in a single NP as a function of reaction time in the early oxidation stage. (C) Oxygen uptake curve obtained from reactive MD simulations of oxidizing a 10-nm Fe NP. The red line is the simulated curve, whereas the dashed black line is the logarithmic fit to the simulated data. The inset is a cross-sectional snapshot of the oxidized Fe-FexOy NP, highlighting the distribution of Fe (green) and O (red) atoms. (D) Simulated oxygen uptake rate (black squares) and Fe/FexOy interfacial area (red circles) in a NP as a function of oxidation time.

The oxidation kinetics obtained from the MD simulations (Fig. 4C, red curve) indicate a quick initial formation of oxide shell, followed by slower growth. The logarithmic growth kinetics (Fig. 4C, dashed black curve) correspond to the microscopic mechanism of oxide growth when iron movement results from extended defects, such as voids in this case (22). The simulations also give the rate of oxygen uptake and the corresponding variation of the Fe/FexOy interfacial area (Fig. 4D). At the initial oxidation stage (t < 1 ns), the rate of oxygen uptake quickly increases because oxygen can easily access the surface Fe atoms to react and form a continuous oxide shell. At longer times, the oxygen uptake rate decreases because of void formation at the Fe/FexOy interface, which leads to a drop in its area. These results are consistent with the experimental measurements (Fig. 4B).

Experimentally, when the reaction temperature increases to 180°C, at which amorphous FexOy can crystallize, the diffusion-limited behavior observed at 100° and 140°C no longer exists (Fig. 4A and fig. S17). The amount of oxygen uptake increases abruptly at 269 min, whereas the maximum lateral dimension of the NPs does not undergo an apparent change, indicating that the oxygen uptake mechanism in this short period is different from the oxidation at 100° and 140°C. The real-time wide-angle x-ray scattering (WAXS) patterns show that the amorphous FexOy nanoshells start to crystallize at 265 min, coincident with a jump in the WAXS signal (fig. S18). The coincident increases in oxygen uptake and WAXS signals at 265 to 270 min indicate that converting amorphous FexOy nanoshells to crystalline ones enhances the oxygen uptake rate. The oxygen uptake quickly reaches a maximum at 273 min, whereas the size of the nanoshells shows negligible change (Fig. 4A). Such dependence on reaction time implies that increasing crystallinity of the nanoshells prompts a fast inward oxygen diffusion in the crystalline iron oxide. Therefore, the diffusion direction of materials involved in the oxidation of colloidal metal NPs can be tuned by controlling the crystallinity of the NPs.

Once the oxygen uptake reaches the maximum, further crystallizing the nanoshells does not increase the amount of oxygen uptake, indicating that the oxidation process stalls during the deep crystallization process. The maximum lateral dimension of the nanoshells slightly fluctuates as a result of the material reorganization associated with the crystallization process. When the reaction atmosphere switches from air to N2, annealing the colloidal FexOy nanoshells releases oxygen to reach a stable state, at which the crystallinity of the nanoshells ceases to change, culminating in well-defined Fe3O4 nanocrystals (fig. S18, C and D). The loss of oxygen might be attributed to desorption of the oxygen adsorbed on the NP surface and crystalline grain boundaries in the environment absent of molecular oxygen. The results indicate that the ratio of oxygen to iron in the iron oxide nanoshells can be tuned by controlling the reaction environment.

We have used in situ SAXS and WAXS to track, with high fidelity (fig. S19) and a spatial resolution of ~5 Å, the full 3D geometrical evolution of hollow interiors in the transformation of amorphous Fe core–FexOy shell solid NPs into crystalline Fe3O4 hollow nanoshells. Large-scale reactive MD simulations corroborate the experimental observations and elucidate the underlying atomistic mechanism associated with the geometrical transformation. The compositional evolution of the NPs over the course of oxidation highlights that material diffusion and the stoichiometric ratio of oxygen to iron can be tuned to control doping concentrations of anions and crystalline defects in oxide NPs. More generally, we have shown that in situ synchrotron x-ray scattering techniques combined with ab initio structural modeling enable quantitative reconstruction of the 3D geometry of colloidal NPs in reactive solutions and offer great potential for addressing many fundamental questions in materials science and chemistry.

Supplementary Materials

Materials and Methods

Figs. S1 to S19

Movies S1 to S5

References (2339)

References and Notes

  1. Acknowledgments: Y.S. thanks Temple University for startup support. The SAXS and WAXS measurements were performed at Beamline 12ID-B of the Advanced Photon Source, which is a U.S. Department of Energy (DOE) Office of Science User Facility operated for the DOE Office of Science by Argonne National Laboratory under contract no. DE-AC02-06CH11357. Use of the Center for Nanoscale Materials, a DOE Office of Science User Facility, was supported by the DOE Office of Science, Office of Basic Energy Sciences, under contract no. DE-AC02-06CH11357. This research used resources of the National Energy Research Scientific Computing Center, which is supported by the DOE Office of Science under contract no. DE-AC02-05CH11231. An award of computer time was provided by the Innovative and Novel Computational Impact on Theory and Experiment (INCITE) program. This research used resources of the Argonne Leadership Computing Facility at Argonne National Laboratory, which is supported by the DOE Office of Science under contract no. DE-AC02-06CH11357. All data are reported in the main text and supplementary materials. Y.S. initiated and led the project. X.Z., Y.S., and S.P. carried out the experiments. X.Z. and Y.S. analyzed the experimental data. S.K.R.S.S. directed the MD simulations. B.N., G.K., and S.K.R.S.S. carried out and analyzed the MD simulations. Y.S. wrote the manuscript with input from all the authors.
View Abstract

Navigate This Article