LETTERS

PUBLISHED ONLINE: 23 OCTOBER 2011 | DOI: 10.1038/NMAT3150

Atomic structure of nanoclusters in oxide-dispersion-strengthened steels

A. Hirata1 , T. Fujita1 , Y. R. Wen1 , J. H. Schneibel2 , C. T. Liu2 and M. W. Chen1,3 *
Oxide-dispersion-strengthened steels are the most promising structural materials for next-generation nuclear energy systems because of their excellent resistance to both irradiation damage and high-temperature creep14 . Although it has been known for a decade that the extraordinary mechanical properties of oxide-dispersion-strengthened steels originate from highly stabilized oxide nanoclusters with a size smaller than 5 nm, the structure of these nanoclusters has not been claried and remains as one of the most important scientic issues in nuclear materials research27 . Here we report the atomic-scale characterization of the oxide nanoclusters using state-of-theart Cs-corrected transmission electron microscopy. This study provides compelling evidence that the nanoclusters have a defective NaCl structure with a high lattice coherency with the bcc steel matrix. Plenty of point defects as well as strong structural afnity of nanoclusters with the steel matrix seem to be the most important reasons for the unusual stability of the clusters at high temperatures and in intensive neutron irradiation elds. The safety, reliability, economics, and efficiency of nextgeneration fission and future fusion energy systems will ultimately depend on developing new high-performance structural materials that can provide extended service for at least 60 years under extremely harsh environments where the materials are exposed to high temperatures, large time-varying stresses, chemically reactive surroundings, and intense neutron radiation13 . Oxide dispersion strengthened (ODS) steels have been strenuously developed as a promising structural material for next-generation nuclear energy systems because of their excellent resistance to irradiation damage and high-temperature creep as well as extraordinary structural and chemical stability in extremely harsh environments27 . Small Y- or YTi oxide nanoprecipitates that are uniformly dispersed in the steel matrix with a very high number density are responsible for reducing the creep rates by six orders of magnitude at 650900 C, and contribute to the excellent tensile ductility (RA > 40%) and strength (>2 GPa) of the ODS steels79 . They also present extremely high stability at temperatures as high as 1,400 C (0.91 Tm , where Tm stands for the melting temperature) and in intense neutron irradiation fields. This unusual stability of the oxide clusters cannot be readily explained by thermodynamics and traditional materials theories. To understand such extraordinary mechanical properties owing to the highly stabilized oxide nanoprecipitates, a knowledge of the structure and chemistry of the oxide nanoclusters is necessary in the research field of the ODS steels. The oxide nanoprecipitates in the ODS steels have been characterized mainly using transmission electron microscopy (TEM) by many researchers1018 . These analyses have been performed for relatively large nanoparticles with sizes greater
a b

200 nm

200 nm

Figure 1 | Typical microstructure of the ODS steel. a, BF- and b, HAADF-STEM images obtained from a 14YWT ODS steel.

than 5 nm. The types of crystal structures of the large YTiO nanoparticles have been suggested as Y2 Ti2 O7 (refs 1114) and Y2 TiO5 (ref. 11; Y/Ti 1), and can be found in the published crystal databases19 . Atom probe tomography (APT), on the other hand, reveals the chemical composition of oxide nanoclusters with a size less than 5 nm (refs 20,21). Although the nanoclusters are basically composed of Ti, O and Y, as well as significant amounts of Fe and Cr (refs 2025), the Y/Ti ratio is much smaller than 1, within the range from 0.1 to 0.6, and depends on the material composition and timetemperature history of the material process22 . In any case, the low Y/Ti ratio is not consistent with those of the reported YTiO oxides, such as Y2 Ti2 O7 and Y2 TiO5 , which has been an important chemical feature of the nanoclusters. However, the crystallographic structure of the nanoclusters has been debated for many years and has not been clarified by definite structure characterization4,2027 . As the formation of highly dense oxide nanoclusters is the most effective way to achieve good mechanical properties28 , it is thus vitally important to understand the structural and chemical features of the oxide nanoclusters. Because the nanoclusters are very small (25 nm), embedded in the magnetic bcc-Fe matrix, and may have a coherent relation with the matrix, it is thus extremely difficult to obtain the structural information by conventional TEM, which is limited by a low spatial resolution and the lack of capability for atomic-scale chemical analysis. In this study, we systematically characterized the atomic structure and chemistry of the nanoclusters in an ODS steel using the newly developed state-of-the-art Cs-corrected TEM and scanning TEM (STEM) with ultra-high spatial resolutions of 0.10 nm. The microstructure of the 14YWT-ODS steel (nominal composition: Fe14Cr3W0.4Ti (wt.%) with 0.25 wt.% Y2 O3 ) was first surveyed using bright field-STEM (BF-STEM) and high angle annular dark field (HAADF) STEM techniques. Figure 1a shows the typical microstructure of the ODS steel used in this study. In the

1 WPI

Advanced Institute for Materials Research, Tohoku University, Sendai 980-8577, Japan, 2 Center for Advanced Structural Materials, City University of Hong Kong, Kowloon, Hong Kong, 3 State Key Laboratory of Metal Matrix Composites, School of Materials Science and Engineering, Shanghai Jiao Tong University, Shanghai 200030, China. *e-mail: mwchen@wpi-aimr.tohoku.ac.jp.
922
NATURE MATERIALS | VOL 10 | DECEMBER 2011 | www.nature.com/naturematerials

2011 Macmillan Publishers Limited. All rights reserved

NATURE MATERIALS DOI: 10.1038/NMAT3150

a
Fe-L edge

LETTERS
a
[110]bcc

b
O-K edge

c
Cr-L edge

d
Ti-L edge

1 nm

e
N-K edge

f
HAADFSTEM
Exp. bcc+ NaCl bcc

5 nm
bcc+ NaCl

Figure 2 | EELS chemical mapping of the nanoclusters. a, Fe-L, b, O-K, c, Cr-L, d, Ti-L, and e, N-K EELS elemental maps obtained from a nanoparticle with a size of 23 nm. f, The corresponding HAADF-STEM image.

Exp.

bcc

BF-STEM image, the steel matrix with grain sizes of 100200 nm can be clearly identified. The bright and dark contrast of individual grains comes from the diffraction variations originating from the different crystallographic orientations of each grain. The HAADF-STEM image, on the other hand, mainly shows the mass contrast, where the regions with lower-density and/or including lighter elements (for example oxides in the steel) are imaged with darker contrast and vice versa. From Fig. 1b of the HAADF-STEM image, we can see a large number of dark dots, as indicated by the arrowheads, with sizes ranging from 2 nm to 50 nm, which correspond to Ti-rich oxide nanoprecipitates in the polycrystalline matrix (see Supplementary Fig. S2). Each grain with a bcc structure has homogeneous contrast, although a small residual diffraction contrast is still visible, showing a slight difference in the contrast from grain to grain as well as the bright contrast of a few nanoparticles. Therefore, the HAADF-STEM technique enables us to directly observe the distribution of the nanoprecipitates in the ODS steel. As the most important question on the microstructure of ODS steels is the nature of the small precipitates, we systematically characterized nanoclusters that are smaller than 5 nm and have a composition that features low Y/Ti atomic ratios (0.10.6), as shown in Supplementary Fig. S3. Energy-dispersive X-ray spectroscopy (EDS) analysis confirms that these dark regions correspond to the nanoclusters as the enriched Ti and Y elements can be detected (see Supplementary Fig. S4). The detailed element distribution in the nanoclusters was investigated by electronenergy-loss spectroscopy (EELS) in the STEM mode. Elemental maps were derived from EELS spectrum images, where the EELS spectra were acquired from each pixel (0.3 nm 0.3 nm). In the
NATURE MATERIALS | VOL 10 | DECEMBER 2011 | www.nature.com/naturematerials

Figure 3 | HAADF-STEM characterization of the nanocluster. a, Experimental HAADF-STEM image of a nanocluster. b, Enlarged experimental image from blue box area in a, showing a coherent lattice structure in the interface region. c, Simulated image using a model with a coherent bcc and NaCl structures. d, Simulated image using a model with only the bcc matrix for comparison. Images of lower panels of bd, corresponding to the yellow box area in a, are enlarged views of the upper panels of b,c and d, respectively. The real-space STEM image of the interface region with periodically enhanced atomic columns can be explained well by the coherent bcc/NaCl model, providing straightforward evidence that the nanoclusters have a NaCl structure and are coherent with the bcc steel matrix.

STEMEELS maps (Fig. 2), there are two nanoclusters: one is just at a grain boundary and another is located inside a grain. The enrichments of O, Ti, and N are clearly seen, together with the scarcity of Fe and Cr-rich shells surrounding the Ti(O, N) cores. Furthermore, the EDS analysis shows that the Y element is also enriched in the particle and the Y/Ti ratio is about 0.20 (see Supplementary Fig. S4), consistent with APT measurements21 . The HAADF-STEM image of a nanocluster in a thin region near the TEM specimen edge is shown in Fig. 3a with the electron incidence parallel to the [110]bcc direction. The sample thickness of the thin area was estimated to be 37 nm by means of an EELS log-ratio technique (see Supplementary Fig. S5)29 . As the imaged region is very thin, a nanocluster with dark contrast and distorted lattices can be clearly observed because of weaker interference from the surrounding bcc matrix. The size of the cluster is 23 nm, which is comparable to the nanoclusters detected by APT (refs 20,21). It seems that the central part of the cluster is much more defective whereas the cluster/matrix interface region shows excellent lattice coherence with a particular periodicity composed
923

2011 Macmillan Publishers Limited. All rights reserved

LETTERS
a
Electron

NATURE MATERIALS DOI: 10.1038/NMAT3150

4 nm

Ti(Y, Fe, Cr)O cluster

bcc-Fe

Model 2
4n m

4 nm

Model 1

Model 2

Model 3 [110]bcc

[110]bcc

[001]bcc

Model 1

Model 2

Model 3

Exp.

1 nm

Figure 4 | Structure modelling of the nanocluster from [110]bcc direction. a, 3D external view of a structure model (Model 2) where a 3.0 nm (Ti,Y,Fe,Cr)O nanocluster is embedded in the bcc matrix. bd, The projections of Model 1, Model 2, and Model 3 are depicted, respectively, in b,c, and d, where green, red, blue, purple and orange circles denote Ti, Y, Fe, Cr and O atoms respectively. The corresponding simulated HAADF-STEM images from the three models are shown in the lower panels of bd. e, The experimental HAADF-STEM image. Model 2 is the structure model which is most consistent with the experimental data.

of brighter and darker atomic contrasts arising from the metal (brighter) and oxygen (darker) atomic columns of the nanocluster coherently overlapping the bcc lattice of the matrix (Fig. 3b). The excellent lattice coherency at the interface may be associated with an enrichment of Cr that mediates the lattice mismatch between the bcc Fe and oxide. To understand the atomic structure of the nanoclusters, we constructed three possible structural models in which a nanocluster is embedded in the bcc-Fe matrix (Fig. 4). Based on fast Fourier transform (FFT) analysis of the atomic-resolution STEM images (see Fig. 5, Supplementary Figs S7 and S8), the crystal structure of the nanoclusters is consistent with a NaCl-type TiO structure (the details will be discussed later), which is also confirmed by the periodicity of the atomic columns with enhanced intensity at the interface between the nanocluster and bcc matrix (Fig. 3). Based on APT measurements21 , the chemical composition of the nanocluster was set to be (Ti43.9 Y6.9 Fe3.4 Cr1.1 )O44.7 . Considering vacancies play a key role in stabilizing nanoclusters27,30 , we introduced vacancies at both the O and Ti sites in the TiO cluster model. To be consistent with the experimental images, the concentration of vacancies was varied over a wide range. We found that the best match between the simulated and the experimental images requires 10 at% vacancies. Figure 4a shows a three-dimensional external view of the structure model. The spherical Ti(Y,Fe,Cr)O cluster is positioned at the centre of a bcc matrix box with a dimension of 4 nm 4 nm 4 nm. The orientation relationship between bcc and Ti(Y,Fe,Cr)-O is
924

bcc (002)oxide and [110]bcc [110]oxide based on determined as (110) the FFT analysis of the STEM images (see Fig. 5, Supplementary Figs S7 and S8). In this simulation, the electron incidence is parallel to the [110]bcc direction. The nanocluster in Model 1 (Fig. 4b) has an unrelaxed perfect NaCl structure that is coherent with the bcc matrix. It is apparent that the simulated image is dissimilar to the experimental one as the nanocluster in the experimental image is much darker, accompanied with obvious lattice distortion compared to the simulated one. The model structure was then relaxed using a molecular dynamics (MD) simulation with both many-body and two-body interatomic potentials. During the relaxation, the neighbouring atoms of vacancies move towards the open space, leading to a certain lattice distortion. In particular, the defect density in the central part of the nanocluster seems slightly higher than that at the cluster/matrix interface region. Model 2 (Fig. 4c) is constructed on the basis of the MD relaxed structure. As shown in Fig. 4c, the simulated image based on Model 2 is phenomenologically consistent with the experimental one in contrast variation of the atomic image (Fig. 4e), verifying that the vacancies indicated by ab initio calculations27 and positron-lifetime spectroscopy30 are compatible with our STEM observations. To further verify the reliability of Model 2, we constructed Model 3, in which the relaxed, defective NaCl nanocluster has an incoherent relation with the bcc matrix (Fig. 4d). The lattice constant of the incoherent cluster is 15% larger than that of the coherent one. Interestingly, the simulated image based on Model 3 is also similar
NATURE MATERIALS | VOL 10 | DECEMBER 2011 | www.nature.com/naturematerials

2011 Macmillan Publishers Limited. All rights reserved

NATURE MATERIALS DOI: 10.1038/NMAT3150

a
Exp. Exp. A

LETTERS
b
Model 1 A

bcc matrix

B A

Outer rim part B A

Outer rim part B

Inner part

Inner part

c
Sim. Model 2 A

d
Model 3 A

B bcc matrix A Outer rim part B A

B Outer rim part B

Inner part

Inner part

Figure 5 | Structure characterization of the nanocluster from the [110]bcc direction. FFT patterns of inner and outer rim parts (indicated in the images) obtained from: a, experimental image; b, simulated image from Model 1; c, Model 2; and d, Model 3. FFT patterns from the bcc matrix are also shown. The well-dened FFT patterns of the bcc matrix and NaCl structure can be observed from the coherent interface regions (A, A and A), directly demonstrating that the nanoclusters have a NaCl structure with a high lattice coherency with the bcc steel matrix. The nearly identical FFT patterns taken from central parts of the clusters in B and B indicate that the nanoclusters contain a large number of lattice defects.

to the experimental image, although it seems to be lacking in some details compared with the image from Model 2 and the experimental STEM image. However, the small difference cannot be convincingly observed on the basis of the STEM images alone. To unambiguously distinguish the difference between these STEM images, FFT analysis was performed for the experimental and the simulated HAADF-STEM images (Fig. 5). The outer rim parts of both images provide clear FFT patterns coming from the NaCl structure (also see Supplementary Fig. S7), whereas the inner parts give relatively diffuse FFT patterns owing to a more distorted atomic structure. Moreover, the FFT analysis for both Model 1 and Model 3, which fail to explain the experimental data, provides more rigorous evidence that Model 2 gives the best matching and should be the most dependable structure model of the nanocluster. In Model 2, 111-type spots of NaCl (see Supplementary Fig. S7) are clearly seen in the FFT pattern from the outer rim part, whereas the FFT pattern from the centre part shows a much more diffuse pattern. Furthermore, from the ]bcc direction, an experimental HAADF image is also fairly [111 consistent with Model 2 (see Supplementary Fig. S8). These observations provide compelling evidence that the defective NaCltype Ti(Y,Fe,Cr)-O is the most promising structure model of the oxide nanoclusters in the ODS steel. Vacancies have recently been suggested to be an important factor in stabilizing the oxide nanoclusters27,30 . This feature is consistent with the fact that the NaCl-type TiO structure is capable of accommodating a large number of vacancies31 . Moreover, TiO with a NaCl-type structure is a strongly nonstoichiometric interstitial compound and thus exhibits a very broad compositional range from Ti40 at% O to Ti55 at.% O in the equilibrium TiO phase diagram. Interestingly, the TiN, Ti(O,N), Fe(O,N), YN systems also form a NaCl-type structure19 , indicating that the NaCl-type structure has great chemical flexibility as well as the high vacancy capability to form nanoclusters with multiple constituent elements. On the basis of the present structural and chemical characterization, the unusual stability of the nanoclusters at high temperature and under intense neutron irradiation seems to be
NATURE MATERIALS | VOL 10 | DECEMBER 2011 | www.nature.com/naturematerials

associated with the defective structure and the strong structural affinity with the bcc matrix. The full lattice coherency between the oxide nanoclusters and the steel matrix gives rise to a very low interface energy between the two disparate materials, oxide and metal. The low interface energy, along with the very low solubility of O and Y in bcc Fe, can effectively prevent the coarsening of the oxide precipitates32 . Moreover, the intrinsic defective structure of the nanoclusters can naturally tolerate the radiation-induced damage2 . Therefore, the ODS steels with a large number of defective nanoprecipitates represent a novel material state, which is intrinsically distinct from conventional nano-phase materials, which are usually metastable and can rapidly become coarsened at high temperatures.

Methods
Alloy synthesis. The ODS steel used in this study was prepared by mechanical alloying of the alloy powders, Fe14Cr3W0.4Ti (wt.%), with 0.25 wt. % Y2 O3 powders, followed by canning in an evacuated jacket and hot extrusion at 850 C. The hot-extruded ODS/MA ingot was then annealed for 1 h at 1,000 C as a standard heat treatment for the formation of nanoclusters and nanoparticles. Details of the material processing have been given by Hoelzer et al.33 . This alloy is designated as 14YWT, where 14 indicates the Cr concentration in wt.% and YWT represents the alloying additions of Y2 O3 , W and Ti. TEM characterization. Specimens for TEM characterization were prepared using an electropolishing method with a solution composed of HClO4 , dibutyl-ethanol and ethanol (1:6:2) at 30 C. Microstructure characterization of the ODS steel was performed using a JEM-2100F TEM (JEOL, 200 kV) equipped with double spherical aberration (Cs) correctors for both the probe-forming and image-forming lenses. Elemental mappings were obtained using EELS and EDS. EELS and EDS measurements were carried out using Gatan GIF Tridiem and JEOL JED-2300T, respectively. Quantitative EDS chemical analysis was carried out using theoretical k -factors. High angle annular dark field (HAADF) images, for which the contrast is basically proportional to the square of the atomic number, were acquired using an annular-type STEM detector while BF-STEM images were simultaneously recorded using a STEM BF detector. The collecting angle in this study ranges from 100 to 267 mrad, which is high enough for HAADF-STEM (see Supplementary Fig. S1). Structure modelling and image simulation. The structure models for image simulation were constructed by MD simulation. First we prepared a cubic bcc-Fe
925

2011 Macmillan Publishers Limited. All rights reserved

LETTERS
structure including a spherical NaCl-type TiYFeO structure as an initial structure. Then the structure was annealed at 1,273 K for structure relaxation using many-body generalized embedded-atom potentials for FeFe and FeTi pairs, two-body BornMayerHuggins potentials for TiO, TiTi and OO pairs, and two-body Lennard-Jones potentials for FeY, TiY, OY, YY and FeO pairs. The parameters for Lennard-Jones potentials were determined based on atomic distances in the binary and ternary oxide structures. The annealing time is 23 fs, and further increasing the time does not cause any obvious structure changes. As there were insufficient Cr potentials and Cr does not affect the HAADF contrast much, we replaced some Ti atoms with Cr atoms after the MD simulation to form a (Ti43.9 Y6.9 Fe3.4 Cr1.1 )O44.7 nanocluster for the HAADFSTEM simulations. The HAADFSTEM simulations were performed using the Win HREM software (HREM Research). The algorithm of the code has been verified to be reliable for simulating HAADFSTEM images with a large atomic cell and also for simulating Cs-corrected STEM images34 . In the calculations, the probe convergence angle is 25 mrad and the HAADF detector inner and outer angles are 100 and 267 mrad, respectively.

NATURE MATERIALS DOI: 10.1038/NMAT3150

18. Sakasegawa, H. et al. Correlation between chemical composition and size of very small oxide particles in the MA957 ODS ferritic alloy. J. Nucl. Mater. 384, 115118 (2009). 19. Villars, P. & Cenzual, K. Pearsons crystal data: Crystal structure database for inorganic compounds Release 2008/9 (ASM International). 20. Miller, M. K., Hoelzer, D. T., Kenik, E. A. & Russell, K. F. Stability of ferritic MA/ODS alloys at high temperatures. Intermetallics 13, 387392 (2005). 21. Miller, M. K., Russell, K. F. & Hoelzer, D. T. Characterization of precipitates in MA/ODS ferritic alloys. J. Nucl. Mater. 351, 261268 (2006). 22. Alinger, M. J., Odette, G. R. & Hoelzer, D. T. On the role of alloy composition and processing parameters in nanocluster formation and dispersion strengthening in nanostuctured ferritic alloys. Acta Mater. 57, 392406 (2009). 23. Sakasegawa, H. et al. Stability of non-stoichiometric clusters in the MA957 ODS ferrtic alloy. J. Nucl. Mater. 417, 229232 (2011). 24. Williams, C. A., Marquis, E. A., Cerezo, A. & Smith, G. D. W. Nanoscale characterisation of ODS-Eurofer 97 steel: An atom-probe tomography study. J. Nucl. Mater. 400, 3745 (2010). 25. Brocq, M. et al. Nanoscale characterisation and clustering mechanism in an FeY2 O3 model ODS alloy processed by reactive ball milling and annealing. Acta Mater. 58, 18061814 (2010). 26. Larson, D. J., Maziasz, P. J., Kim, I. S. & Miyahara, K. Three-dimensional atom probe observation of nanoscale titanium-oxygen clustering in an oxide-dispersion-strengthened Fe12Cr3W0.4Ti + Y2 O3 ferritic alloy. Scr. Mater. 44, 359364 (2001). 27. Fu, C. L., Krcmar, M., Painter, G. S. & Chen, X. Q. Vacancy mechanism of high oxygen solubility and nucleation of stable oxygen-enriched clusters in Fe. Phys. Rev. Lett. 99, 225502 (2007). 28. Ohtsuka, S., Ukai, S., Fujiwara, M., Kaito, T. & Narita, T. Improvement of 9Cr-ODS martensitic steel properties by controlling excess oxygen and titanium contents. J. Nucl. Mater. 32933, 372376 (2004). 29. Malis, T., Cheng, S. C. & Egerton, R. F. EELS log-ratio technique for specimen-thickness measurement in the TEM. J. Electron Microsc. Tech. 8, 193200 (1988). 30. Xu, J., Liu, C. T., Miller, M. K. & Chen, H. M. Nanocluster-associated vacancies in nanocluster-strengthened ferritic steel as seen via positron-lifetime spectroscopy. Phys. Rev. B 79, 020204 (2009). 31. Huisman, L. M., Carlsson, A. E., Gelatt, C. D. & Ehrenreich, H. Mechanisms for energetic-vacancy stabilizationTiO and TiC. Phys. Rev. B 22, 9911006 (1980). 32. Zener, C. J. Theory of growth of spherical precipitates from solid solution. J. Appl. Phys. 20, 950953 (1949). 33. Hoelzer, D. T. et al. Influence of particle dispersions on the high-temperature strength of ferritic alloys. J. Nucl. Mater. 367, 166172 (2007). 34. Ishizuka, K. A practical approach for STEM image simulation based on the FFT multislice method. Ultramicroscopy 90, 7183 (2002).

Received 2 June 2011; accepted 19 September 2011; published online 23 October 2011

References
1. Grimes, R. W. & Nuttall, W. J. Generating the option of a two-stage nuclear renaissance. Science 329, 799803 (2010). 2. Odette, G. R., Alinger, M. J. & Wirth, B. D. Recent developments in irradiation-resistant steels. Annu. Rev. Mater. Res. 38, 471503 (2008). 3. Charit, I. & Murty, K. L. Structural materials issues for the next generation fission reactors. J. Oper. Manage. 62, 6774 (2010). 4. Miller, M. K., Fu, C. L., Krcmar, M., Hoelzer, D. T. & Liu, C. T. Vacancies as a constitutive element for the design of nanocluster-strengthened ferritic steels. Front. Mater. Sci. China 3, 914 (2009). 5. Fisher, J. J. Dispersion strengthened ferritic alloy for use in liquid metal fast breeder reactors. US Patent 4,075,010, (1978). 6. Ukai, S. et al. Alloying design of oxide dispersion-strengthened ferritic steel for long-life FBRs core materials. J. Nucl. Mater. 204, 6573 (1993). 7. Ukai, S. & Fujiwara, M. Perspective of ODS alloys application in nuclear environments. J. Nucl. Mater. 307, 749757 (2002). 8. Schneibel, J. H. et al. Development of porosity in an oxide dispersion-strengthened ferritic alloy containing nanoscale oxide particles. Scr. Mater. 57, 10401043 (2007). 9. Klueh, R. L. et al. Tensile and creep properties of an oxide dispersion-strengthened ferritic steel. J. Nucl. Mater. 307, 773777 (2002). 10. Yamashita, S., Akasaka, N. & Ohnuki, S. Nano-oxide particle stability of 9-12Cr grain morphology modified ODS steels under neutron irradiation. J. Nucl. Mater. 32933, 377381 (2004). 11. Okuda, T. & Fujiwara, M. Dispersion behavior of oxide particles in mechanically alloyed ODS Steel. J. Mater. Sci. Lett. 14, 16001603 (1995). 12. Yamashita, S., Ohtsuka, S., Akasaka, N., Ukai, S. & Ohnuki, S. Formation of nanoscale complex oxide particles in mechanically alloyed ferritic steel. Phil. Mag. Lett. 84, 525529 (2004). 13. Klimiankou, M., Lindau, R. & Moslang, A. Energy-filtered TEM imaging and EELS study of ODS particles and argon-filled cavities in ferritic-martensitic steels. Micron 36, 18 (2005). 14. Kishimoto, H., Alinger, M. J., Odette, G. R. & Yamamoto, T. TEM examination of microstructural evolution during processing of 14CrYWTi nanostructured ferritic alloys. J. Nucl. Mater. 32933, 369371 (2004). 15. Klimiankou, M., Lindau, R. & Moslang, A. HRTEM study of yttrium oxide particles in ODS steels for fusion reactor application. J. Cryst. Growth 249, 381387 (2003). 16. Coppola, R., Klimiankou, M., Magnani, M., Moslang, A. & Valli, M. Helium bubble evolution in F82H-modcorrelation between SANS and TEM. J. Nucl. Mater. 32933, 10571061 (2004). 17. Hayashi, T., Sarosi, P. M., Schneibel, J. H. & Mills, M. J. Creep response and deformation processes in nanocluster-strengthened ferritic steels. Acta Mater. 56, 14071416 (2008).

Acknowledgements
This work was sponsored by Global COE for Materials Research and Education, World Premier International (WPI) Research Center Initiative for Atoms, Molecules and Materials, MEXT, Japan. We thank Okunishi of JEOL for his technical assistance and D. T. Hoelzer at ORNL, USA for providing 14YWT samples for this study.

Author contributions
M.W.C. and C.T.L. planned this project. A.H. and M.W.C. designed research, analysed data, constructed models and wrote the paper. A.H. contributed to STEM experiments and image simulation. T.F. contributed to EELS analysis. Y.R.W. contributed to TEM specimen preparation. J.H.S. and C.T.L. contributed to sample preparation. All authors discussed the results and commented on the manuscript.

Additional information
The authors declare no competing financial interests. Supplementary information accompanies this paper on www.nature.com/naturematerials. Reprints and permissions information is available online at http://www.nature.com/reprints. Correspondence and requests for materials should be addressed to M.W.C.

926

NATURE MATERIALS | VOL 10 | DECEMBER 2011 | www.nature.com/naturematerials

2011 Macmillan Publishers Limited. All rights reserved