Unusual thermal stability of nano-structured ferritic alloys

Journal of Alloys and Compounds 529 (2012) 96–101Contents lists available at SciVerse ScienceDirectJournal of Alloys andCompoundsj o u r n a l h o m e p a g e :w w w.e l s e v i e r.c o m /l o c a t e /j a l l c omUnusual thermal stability of nano-structured ferritic alloysX.L.Wang a ,∗,C.T.Liu b ,c ,U.Keiderling d ,A.D.Stoica a ,L.Yang a ,ler b ,C.L.Fu b ,D.Ma a ,K.An aaNeutron Scattering Science Division,Oak Ridge National Laboratory,Oak Ridge,TN 37831,USAbMaterials Science and Technology Division,Oak Ridge National Laboratory,Oak Ridge,TN 37831,USA cDepartment of System Engineering and Engineering Management,City University of Hong Kong,83Tat Chee Avenue,Kowloon Tong,Hong Kong dHelmholtz Center Berlin for Materials and Energy,Glienicker Strasse 100,D-14109Berlin,Germanya r t i c l ei n f oArticle history:Received 18October 2011Received in revised form 23February 2012Accepted 24February 2012Available online 7 March 2012Keywords:NanostructureSmall angle neutron scattering Neutron diffractionHigh temperature deformation Atom probe tomography (APT)a b s t r a c tA scientific question vitally important to the materials community is whether there exist “self-assembled”nanoclusters that are thermodynamically stable at elevated ing in situ neutron scat-tering,we have characterized the structure and thermal stability of a nano-structured ferritic alloy.Nanometer sized nanoclusters were found to persist up to ∼1400◦C,providing direct evidence of a thermodynamically stable alloying state for the nanoclusters.High-temperature neutron diffraction measurements show a stable ferritic matrix,with little evidence of recrystallization or grain growth at temperatures up to 1300◦C.This result suggests that thermally stable nanoclusters and the oxygen-vacancy interaction limit the diffusion of Fe atoms and hence the mobility of grain boundaries,stabilizing the microstructure of the ferritic matrix at high temperatures.© 2012 Published by Elsevier B.V.1.IntroductionEmergent behaviors are found in many complex material sys-tems,where a small adjustment in chemical composition can cause a dramatic change in properties.Excellent mechanical properties are obtained,for example,by mechanical alloying of ferritic steel with only minor addition of Ti and Y 2O 3(≤0.5wt.%)[1–13].At temperatures between 650and 900◦C,the creep rate of Fe–14Cr–3W +0.25Y 2O 3+0.4Ti (wt.%)(hereafter designated as 14YWT)ferritic alloy is six orders of magnitude lower as compared to the similar cast materials [5].This level of improvement repre-sents a major breakthrough in the design of ferritic alloys for high temperature structural applications.Similarly,significant improve-ment in creep resistance (by two orders of magnitude)was reported by introducing carbonitride precipitates with minor addition of N and C [14].It has been suggested that the unusually low creep rate in these micro-alloyed steel is related to the nanoclusters that formed dur-ing processing.Here,by nanoclusters,we mean fine scale particles with less well defined boundaries.Indeed,transmission electron microscopy (TEM)carried out on 14YWT demonstrated high-density stable nanoclusters with diameters of 2–5nm,and atom probe tomography indicates that these nanoclusters are distinctly enriched with Ti,O and Y,as compared to the matrix composition∗Corresponding author.Tel.:+18655749164;fax:+18655746080.E-mail address:wangxl@ (X.L.Wang).[15].Similarly,atom probe tomography studies on a related 12YWT (with 12wt.%Cr)revealed stable clusters with a diameter of ∼4nm [3].Small angle neutron scattering (SANS)is another powerful probe to characterize nanoscale structures [12,13,16–19].For 14YWT,Alinger et al.[13]reported a particle size of 2–3nm,but it is clear from the SANS intensity data [12]that more than one particle size exists.A detailed comparison between TEM measurements and SANS analysis was given for MA957[11],a Mo containing alloy with a nomination composition of Fe–14Cr–0.9Ti–0.3Mo–0.25Y 2O 3.From SANS data,a particle size distribution was estimated with a peak around 4nm.On the other hand,TEM work revealed larger particles with a particle size distribution centered around 10–20nm.The combination of SANS and TEM suggests that two particle sizes exist in MA957.Unlike most engineered or artificially created nano-materials which are meta-stable in nature,the self-assembled clusters in the 14YWT alloy appear to be extremely thermally stable.Microscopy and atom probe tomography analyses on quenched 14YWT samples after extensive heat treatments at 1380◦C still show high-density nanoclusters,without significant coarsening [15].There are two schools of thoughts as to why these nanoclusters are still present in the quenched samples.One possibility is that these nanoclusters remain during annealing at high temperatures.It is also conceiv-able,however,that the nanoclusters dissolve at high temperatures and re-precipitate during quenching,so long as the quenching rate is not exceedingly high.In situ observation is the only way to0925-8388/$–see front matter © 2012 Published by Elsevier B.V.doi:10.1016/j.jallcom.2012.02.143X.L.Wang et al./Journal of Alloys and Compounds529 (2012) 96–10197Fig.1.High-temperature furnace and sample mount used in the in situ small angle scattering experiments.unambiguously determine the exact mechanism.However,in situ scattering experiments are difficult.This is not just because of the limitation in scattering intensity;rather,it is the availability of a reliable vacuum furnace capable of reaching1300◦C that posed the greatest challenge.In this paper,we present,from in situ neutron scattering experiments,the evidence of thermally stable nanoclus-ters and show their connection to the improved creep properties.2.Experimental details2.1.Sample preparationTwo types of samples were investigated:14YWT with the composi-tion Fe–14Cr–3W+0.25Y2O3+0.4Ti(wt.%)and FNAN-1with the composition Fe–12.3Cr–3W+0.42Ti(wt.%).The former is a known nanostructured steel,whereas the latter is a base alloy containing no nanoclusters and used as a reference sam-ple.The14YWT alloy was fabricated by mechanical alloying followed by extrusion, whereas the FNAN-1alloy ingot was prepared by arc melting and drop casting into a copper mold.The14YWT powder was canned in mild steel,soaked,and then extruded at850◦C.There was no further consolidation(i.e.,no rolling or cold work-ing).In FNAN-1,the oxides werefloated to the ingot top as slag during casting without any residual Y retained in the alloy.All samples were given a standard heat treatment of1h at1000◦C in vacuum.2.2.In situ neutron scattering experimentsThe SANS experiment was carried out at Berlin Neutron Scattering Center, Helmholtz Center Berlin for Materials and Energy,using the V4instrument.The incident neutron wavelength was6˚A.The samples were cut into1-mm thick discs with a diameter of approximately8mm.An AS Scientific furnace,shown in Fig.1, with a maximum temperature of∼1600◦C,was used for the high-temperature mea-surements.Plate like specimens were mounted on a special sample holder,between two ceramic baffles,and heldfixed using a ceramic set screw.Heating was achieved by radiation with surround Nb heaters.SANS data were collected every∼25◦during heating to1400◦C,and at selected temperatures during cool down to room tem-perature.The temperature profile is plotted in Fig.2(a).For each temperature,two detector settings were used,with detectors at1and4m from the sample position, respectively,in order to cover a Q-range of0.01–0.3˚A−1.The total data collection time was10min for each temperature.The heating experiment was repeated with-out a sample for each detector setting,to determine the background scattering which also varied with the temperature.The high-temperature diffraction experiment was conducted at Intense Pulsed Neutron Source,Argonne National Laboratory,using the GPPD powder diffractome-ter.Heating was done with the“Miller”furnace.Cylindrical specimen was mounted on top of a ceramic pedestal which was then lowered into the furnace.After a rather fast temperature rise up to1000◦C with a heating rate of0.1◦C/s,both samples were subjected to consecutive stepwise annealing treatments covering the temperature range of1000–1300◦C with3h per50◦C step,during which time neutron diffraction data were collected.Additional high-temperature measurements were conducted using the new VULCAN diffractometer[20,21]at the Spallation Neutron Source[22]. In situ neutron diffraction was collected during continuous heating up to1300◦C.Fig.2.(a)Temperature profile for in situ SANS measurements.SANS data were collected every∼25◦during heating to1400◦C and at selected temperatures during cool down.At each temperature,the data collection was10min.(b)Differential scattering cross section per unit sample volume for alloy14YWT(with nanoclusters) at room temperature(blue triangles)and1412◦C(red solid circles).For comparison, room temperature SANS data for alloy FNAN-1containing no nanoclusters are also presented.(For interpretation of the references to color in this Fig.legend,the reader is referred to the web version of the article.)3.Experimental results3.1.Nanoscale particle size distributionStrong SANS intensity(over four decades)was observed in alloy14YWT,as shown in Fig.2(b).The experimental data were reduced to the differential scattering cross section per unit sample volume,d /d˝,following a standard data reduction procedure[23].For comparison,the room temperature data for the reference sam-ple FNAN-1are also plotted.These results confirm that,as in Refs.[11–13],the SANS intensity originates from scattering by nano-scale structure features formed during processing.At1400◦C,the low-Q portion of the SANS data moved towards lower Q, while the high-Q portion remained the essentially same or even increased slightly.Plotting SANS data in log–log scales revealed multiple length scales in14YWT alloys.As shown in the inset of Fig.3,three regimes in Q can be identified:0.01–0.03, 0.03–0.1,and0.1–0.3˚A−1,corresponding to three characteristic particle sizes.How-ever,modeling the SANS data with three particle size distribution functions will result in too many parameters that cannot be determined reliably.As the large par-ticles are not of interest in this study,only data for Q>0.03˚A−1corresponding to smaller particles were analyzed.The SANS data were therefore modeled with a bi-modal log-normal particle size distribution:f(D)=i=1,2N∗iD·s i√exp−1ln(D)−ln(D i)i2,(1)98X.L.Wang et al./Journal of Alloys and Compounds 529 (2012) 96–1013020100.0000.0050.0100.015P a r t i c l e S i z e D i s t r i b u t i o n F u n c t i o n [f (D )*V ]Diameter, D (nm)Fig.3.Bi-modal particle size distribution obtained by fitting the SANS data at1300◦C.The vertical axis is the volume fraction of the particle size distribution (weighted by scattering contrast),f (D )·V ,where V =( /6)D 3.The inset shows quality of the fit along with contributions by each particle size distribution.where D i is the centroid and s i is the standard deviation for each particle size dis-tribution.N ∗i is the number of particles N i weighted by the scattering contrast i ,i.e.,N ∗i =N i 2i .Satisfactory fits were obtained with this model,and an example of the particle size distribution is given in Fig.3.The maxima in the particle size distribution correspond to particles of diameters of 2nm (i =1)and 12nm (i =2),respectively.The contributions by each particle size distribution are also plottedin the inset of Fig.3.TEM experiments show that the larger particles (∼12nm)are either TiO 2or Y 2Ti 2O 7located mainly at grain boundaries.The 2-nm particles have been observed by atom probe tomography which showed that they are clusters composed of Y,Ti,and O,located both within the grains and at grain boundaries [15].An example of atom probe tomography study is shown in Fig.4for a 14YWT sample quenched after annealing for 1h at 1400◦C.Several nanometer sized Y–Ti–O clusters can be identified.Evidence for small,nanome-ter sized clusters was also reported by Hayashi et al.[24]using energy-filtered TEM.Note that SANS intensity contains scattering from both structural and magnetic heterogeneities.For a ferromagnetic material,the magnetic scattering comes from magnetic domain.Loffler et al.[17]showed that for nanostructured materials,the magnetic correlation length or domain size is at least 2–3times the grain size.For 14YWT,the typical grain size is ∼0.2–0.4␮m [4],so the magnetic domain is ∼0.5␮m.SANS from such length scale features typically show up at Q below 0.01˚A−1,very different from the Q region of our SANS data,0.01–0.3˚A−1.The presence of magnetic scattering intensity at low-Q does not affect our analysis,especially for the smaller 2-nm Y–Ti–O nanoclusters.10-410-310-210-1N *1 (a r b . u n i t )Temperature (C)Fig.5.Temperature dependence of N ∗1(see Eq.(1))for the 2-nm nanoclusters in 14YWT nanostructured ferritic alloy.The blue triangle data point was obtained aftercooling to room temperature.Reliable N ∗1couldnot be determined near the Curie temperature (T c ∼700◦C),so only data for up to 450◦C and above 1300◦C are pre-sented.(For interpretation of the references to color in this Fig.legend,the reader is referred to the web version of the article.)3.2.Thermal stability of the nanoclustersAs noted in Fig.2,the low-Q part of the SANS curve shifts slightly towards lower Q values at 1400◦C,suggesting limited growth of the 12-nm particles.Indeed,thefitted N ∗2decreases slightly,indicating a small degree of coarsening of the larger particles.This is consistent with (ex situ)TEM studies of annealed samples.However,our interest is in the 2-nm nanoclusters and their temperature dependence.In order to obtain stable fits for data sets at all temperatures,D 1and s 1were held fixed.Atom probe tomography experiments on samples annealed at 1000◦C showed noevidence of cluster growth.In Fig.5,we plot N ∗1,the number of the 2-nm clusters weighted by the scattering contrast 1,as a function of temperature.Significant magnetic critical scattering was observed due to the fluctuation of Fe magnetic moments across a Curie temperature of T c ∼700◦C.Here we need to emphasize that neither the 12-nm particles (TiO 2and Y 2Ti 2O 7)nor the 2-nm clusters (Y–Ti–O)are magnetic.In principle,the contribution by non-magnetic nanoclusters could be separated from the magnetic critical scattering,but this would require applying a magnetic field at elevated temperatures.Such an experimental environ-ment does not exist at present.Without applying a magnetic field,it is impossible to determine unambiguously the SANS intensity by the nanoclusters.For this reason,data close to T c cannot be discussed in a meaningful way and must be excluded.To determine the upper temperature limit of the magnetic critical scattering,a refer-ence alloy FNAN-1containing no nanoclusters was measured with a similar heatingFig.4.Atom maps produced by atom probe tomography on a 14YWT sample quenched from 1h annealing at 1400◦C.The size of the selected volume (i.e.,the box)is 20nm ×50nm,with a thickness of 5nm into the page.A grain-boundary is intersected,as indicated by the increased concentration of Cr atoms.Clustering of Y–Ti–O is evident.Note that the cluster is also slightly enriched in Cr.X.L.Wang et al./Journal of Alloys and Compounds 529 (2012) 96–101991400120010008006004002000.00.10.20.30.4I n t e n s i t y (a u )Temperature (C)Fig.6.SANS intensity for reference FNAN-1alloy without nanoclusters as function of temperature.The SANS intensity turned into a constant at temperatures above 1100◦C,marking the upper bound for magnetic critical scattering.profile.Above T c ,the SANS intensity at high Q for FNAN-1decreased monotonously with increasing temperature and turned into a constant for temperatures above ∼1100◦C.This is illustrated in Fig.6.Apparently,the magnetic critical scattering in the reference sample vanished at 1100◦C.Further evidence of this upper limit comes from the temperature dependence of the magnetic correlation length in pure iron.It has been established that as T c is approached from above,the magnetic critical scattering is characterized by a correlation length that grows exponentially with the inverse of temperature [25].Indeed,as Fig.7shows,the logarithm of the magnetic correlation length determined by Gersch et al.[26]for pure iron varies linearly withtemperature.At 850◦C,the magnetic correlation length is 3.2˚A.Extrapolating these data to 1000◦C shows a correlation length of less than 0.1˚A.From the analyses for pure iron and the reference alloy FNAN-1,we can safely conclude that the magneticcritical scattering is negligible above ∼1300◦C.Therefore in Fig.5,only N ∗1data for temperatures up to 450◦C and above 1300◦C are presented.It can be seen that N ∗1is essentially constant up to 450◦C.A dashed line is drawn as a guide to the parable values of N ∗1were found up to ∼1400◦C.Thus,despite the strong thermal fluctuations at this extreme temperature,a great num-ber of 2-nm nanoclusters still exist.It follows,therefore,that the 2-nm nanoclusters formed in 14YWT alloy are stable.Notice the blue triangle data point after cooling down to room temperature,further confirming the thermal stability of the nanoclus-ters.Limited coarsening can be seen,as the data point after cool down is slightly below that before heating.3.3.High temperature diffusionUnlike conventional ferric steel,14YWT alloys are thermally stable even at tem-peratures up to 1300◦C (0.87T m ,the melting temperature).This is illustrated in Fig.8(a)and (b),where the in situ neutron diffraction patterns for 14YWT and the nanoclusters-free FNAN-1are compared.After two cycles of heating,the diffraction pattern for 14YWT is virtually unchanged after 3h of annealing at 1300◦C.110010009008007000.1110100C o r r e l a t i o n L e n g t h (A )T (C)Fig.7.Magnetic correlation length determined for pure iron as a function of temper-ature.The data were taken from Ref.[26].The magnetic correlation length increases exponentially as the Cure temperature isapproached.Fig.8.(a)Temperature profile and diffraction data for 14YWT measured in situ at GPPD.The experiment covered two thermal cycles.Each step at high temperatures is approximately 3h.The vertical axis is the data file number.The left panel shows the temperature profile as a function of the data file number,whereas the right panel is a two-dimensional plot showing the evolution of neutron diffraction patterns with the data file number or temperature.(b)Temperature profile and diffraction data for FNAN-1measured in situ at GPPD.Each step at high temperatures is approximately 3h.The designation of the axes in (b)is the same as in (a).(c)Comparison of neu-tron diffraction data for 14YWT and FNAN-1collected in situ at 1300◦C.For GPPD,each data was sum over 3h.Data for VULCAN at the Spallation Neutron Source was collected for 4min in a vacuum furnace after 1h anneal at 1300◦C.Note that GPPD and VULCAN have different resolutions,so the peak widths are different.100X.L.Wang et al./Journal of Alloys and Compounds 529 (2012) 96–101Fig.9.Room temperature diffraction patterns of 14YWT (a)and FNAN-1(b)after step-wise heating (annealing)to 1300◦C.The horizontal axis is lattice spacing.The vertical axes designate the angle of the individual detectors.These detectors are grouped into detector banks.For details of the detector layout on GPPD,see Ref.[27].The nominal angle of each detector bank is given on the right side of the two-dimensional plot.Diffraction data from different detectors are stacked vertically (separated by the dark horizontal bands),with their diffraction peaks lined up in lattice spacing.With the furnace in place,only two of the detector banks (centered around ±90◦)were accessible due to the limited size of the furnace windows.The 14YWT alloy (a)is stable as evidenced from the powder diffraction lines.FNAN-1alloy (b)without nanoclusters experienced massive recrystallization and grain growth during heating.On the other hand,dramatic changes are seen for the diffraction peaks from FNAN-1.Most of the strong peaks in FNAN-1(e.g.,(200))are significantly weak-ened upon heating to 1150◦C,while (211)peak first decrease and then intensifies.In Fig.8(c),the diffraction patterns collected at 1300◦C are compared.While the diffraction pattern for nanocluster-free FNAN-1becomes very diffusive at high tem-peratures,alloy 14YWT retains the clean body-centered-cubic diffraction pattern up to a maximum annealing temperature of ∼1300◦C.The change of diffraction inten-sity with temperature,as shown in Fig.8(b)indicates that FNAN-1recrystallized upon heating and annealing at 1300◦C,see Section 4below for more details.Addi-tional data for 14YWT from the new VULCAN instrument at the Spallation Neutron Source are also included in Fig.8(c).Again,sharp diffraction peaks are identified,indicating that the sample is stable.Note that the VULCAN and GPPD have different resolutions,so the diffraction peak widths are different.Here,the fine resolution with VULCAN is evident.With the furnace in place,only a small range of reciprocal lattice space is covered due to the limited size of the furnace window.To confirm that the intensity change in FNAN-1is indeed due to recrystallization,we conducted additional measurements on GPPD on samples cooled to room temperature,without the furnace.Before presenting the full data set collected over all detectors,it is instructive to describe briefly how a time-of-flight diffractometer such as GPPD works.The detec-tor coverage for GPPD has been given in detail elsewhere [27],with the scattering angle ranging from 30◦to 150◦.For a time-of-flight diffractometer,each detector can collect a whole diffraction pattern,using an incident beam over a large wavelength range.Data from different detectors can be time-focused,i.e.,lined up with the same d-spacing.For powder samples (of random grains),the data from different detectors can be summed together to produce a single diffraction pattern.For materials with preferred grain orientations,data from different detectors no longer look the same.Analysis of the difference yields information on the texture of the sample.In Fig.9(a)and (b),diffraction data collected over all detectors on sam-ples annealed at 1300◦C are presented.It can be seen that recrystallization or grain growth in 14YWT is strongly suppressed,as its diffraction pattern remainspowder-like.In fact,the diffraction patterns recorded at room temperature before and after annealing are almost identical,demonstrating full recovery of the microstructure.On the other hand,FNAN-1alloy containing no nanoclusters expe-rienced massive recrystallization.Its diffraction pattern,after cooling back to room temperature,is spotty,with bright single crystal spots aligned along the weak poly-crystal lines,as shown in Fig.9(b).The contrasting images of Fig.9(a)and (b)show that in nanostructured steel,the grain boundary mobility is significantly slower.Since the grain boundary mobility is directly related to bulk diffusion,this result means that the diffusion of Fe atoms is slow,even at temperatures up to 1300◦C.4.DiscussionThe remarkable stability of Y–Ti–O nanoclusters is the result of strong chemical bonding that holds Y,Ti,and O together.First-principle calculations offered some clues as to why this is hap-pening [28].A new alloying state forms,the calculations show,with a binding energy between the stable states of TiO 2and Y 2Ti 2O 7.The formation of the new alloying state is facilitated by mechanical alloying,which created high density vacancies.Although the solu-bility of O in the body-center-cubic Fe lattice is extremely limited,O atoms binds strongly with vacancy,with a binding energy as high as −1.5eV.This greatly increases the solubility of O in the pres-ence of pre-existing vacancies.The presence of Y is essential for creating stable nanoclusters states,but the Y concentration must also be carefully controlled.Without Y,the stable alloying state is TiO 2oxide phase,with a heat of formation of −4.3eV per oxygen atom [28].In the presence of Y,the cluster state becomes compet-itive,and a small addition of Y actually promotes the formation of nanoclusters instead of TiO 2.With too much Y,however,the Y 2Ti 2O 7oxide phase,with a heat of formation of −5.6eV,is likely to precipitate.Thus,the condition for stable Y–Ti–O nanoclusters requires a well-dispersed Y concentration with more Ti than Y in the nanoclusters.Now that the nanoclusters are thermodynamically stable,and many of them remain at extreme temperatures up to 1400◦C,why are they thermally stable exhibiting minimal growth?To answer this question,we must consider both the thermodynamic driving force and growth kinetics.For normal particle growth,the pri-mary driving force is minimization of the interfacial energy [29].However,first-principle calculations suggest that the new metallic cluster is coherent with the Fe lattice.The neutron diffraction pat-terns in 14YWT remained body-centered-cubic like;no peaks from other phases were observed.These data support the predication by first-principle calculations.If the clusters are indeed coherent,the driving force for normal particle growth,due to high incoher-ent interfacial energy,no longer exists.The kinetics is governed by short-range and long-range transport of atoms that make up the cluster [30].The former involves the transfer of atoms across the particle boundary or interface,whereas the latter is by dif-fusion in which atoms are transported over distances of several atomic spacing.In Y–Ti–O nanoclusters,the atoms are held tightly by strong chemical bonding.Further growth requires concerted or collective migration of Y,Ti,and O-vacancy pairs.This requirement severely constraints the mobility of the desired chemical species.In order for an oxygen atom to move,for example,a vacancy must be created first.The energy required to break an O-vacancy pair is 1.5eV,far beyond the energy from thermal fluctuation,which is only ∼150meV at 1400◦C.Even if some chemical species are moved close to a cluster by long-range diffusion,the delicate chemistry requirement (e.g.,for Y as discussed earlier)can still prevent them to attach to the interface of the clusters limiting the growth.In nanostructured steel with minor N and C addition,the improvement in creep properties is attributed to a mechanism of grain boundary pinning by carbonitride precipitates [14].Hayashi et al.proposed a similar mechanism for the dramatically improved creep resistance in the present 14YWT alloy [24].Their analysis,X.L.Wang et al./Journal of Alloys and Compounds529 (2012) 96–101101based on microscopy and creep test results,suggests that the pin-ning effect of segregation and nanoclusters formation might be strong enough to prevent grain-boundary migration and grain-boundary sliding,even at test temperatures up to800◦C.There appears to be microscopic evidence from the high-temperature neutron diffraction experiments that supports the above proposed mechanism.As shown in Fig.9(a),recrystallization in14YWT alloys is strongly suppressed.This means that com-pared to cluster-free alloy FNAN-1,the grain-boundary mobility and hence the diffusion of Fe atoms in nanostructured steel is signif-icantly reduced.Indeed,O atoms in the lattice or grain boundaries serve as a sink for vacancies due to their strong binding energy.This will severely limit the diffusion of Fe atoms.Grain-boundary pin-ning by the coherent nanoclusters also contributes to the sluggish diffusion kinetics of Fe atoms.Notice that these data were collected after annealing at1300◦C.Even at this temperature,the mobility of Fe atoms is sufficiently slow to prevent massive recrystallization to occur.5.Concluding remarksIn summary,we have carried out a systematic investigation of the structure and thermal stability of a nanostructured ferritic alloy with in situ neutron scattering experiments.From the experimental results,we conclude the following:1.Nanoscale particles in the14YWT nanostructured ferritic alloyhave a bi-modal particle size distribution,with maxima around 2and12nm,respectively.2.The2-nm clusters survived up to1400◦C,suggesting thatthey are thermodynamically stable.The unusual stability of the nanoclusters is attributable to the unique chemistry,as thefirst-principle calculations predicted.3.The nanoclusters in14YWT are thermally stable,with littlegrowth even at high temperatures.The reduced cluster growth is attributed to the coherent nature of the cluster state and the chemistry requirement(e.g.,vacancies).4.The microstructure of the Fe matrix is stable,as evidenced fromthe lack of recrystallization at temperatures up to1300◦C.This result indicates that the mobility of grain boundaries or bulk diffusion is strongly suppressed.The creep rate of14YWT nanostructured ferritic alloy is six orders of magnitude slower,which,along with its incredible resis-tance against radiation damage[31,32],offers a great potential for use in advanced energy conversion systems.While the exact mechanisms of the low creep rate are not yet fully understood,the unusual thermal stability of the nanoclusters and the microstruc-ture of the Fe matrix must be playing a major role.The results of this study are expected to have a broad implication in understand-ing the emergent behavior of complex alloys and in the synthesis of new-generation nanostructured materials.Note added in proofAfter the submission of this manuscript,we became aware of the publication by Hirata et al.[33]which proved that the nan-oclusters discussed in this paper are indeed coherent with the bcc steel matrix.AcknowledgementsResearch sponsored by the Materials Sciences and Engineering Division,Office of Basic Energy Sciences,US Department of Energy. Atom probe tomography(MKM)was supported by ORNL’s Shared Research Equipment(SHaRE)User Facility,which is sponsored by the Office of Basic Energy Sciences,US Department of Energy.The Spallation Neutron Source was operated with support from the Sci-entific User Facilities Division,Office of Basic Energy Sciences,US Department of Energy.We thank Dr.D.T.Hoelzer for providing the 14YWT samples for this study.References[1]ai,T.Nishida,T.Okuda,T.Yoshitake,J.Nucl.Sci.Technol.35(1998)294–300.[2]rson,P.J.Maziasz,I.S.Kim,K.Miyahara,Scripta Mater.44(2001)359–364.[3]ler,E.A.Kenik,K.F.Russell,L.Heatherly,D.T.Hoelzer,P.J.Maziasz,Mater.Sci.Eng.A:Struct.Mater.Properties Microstruct.Process.353(2003) 140–145.[4]I.S.Kim,B.Y.Choi,Met.Mater.Int.8(2002)265–270.[5]R.L.Klueh,P.J.Maziasz,I.S.Kim,L.Heatherly,D.T.Hoelzer,N.Hashimoto,E.A.Kenik,K.Miyahara,J.Nucl.Mater.307(2002)773–777.[6]D.T.Hoelzer,J.Bentley,M.A.Sokolov,ler,G.R.Odette,M.J.Alinger,J.Nucl.Mater.367(2007)166–172.[7]I.S.Kim,B.Y.Choi,C.Y.Kang,T.Okuda,P.J.Maziasz,K.Miyahara,ISIJ Int.43(2003)1640–1646.[8]H.Kishimoto,M.J.Alinger,G.R.Odette,T.Yamamoto,J.Nucl.Mater.329(2004)369–371.[9]J.H.Schneibel,M.Heilmaier,W.Blum,G.Hasemann,T.Shanmugasundaram,Acta Mater.59(2011)1300–1308.[10]J.H.Schneibel,C.T.Liu,ler,ls,P.Sarosi,M.Heilmaier,D.Sturm,Scripta Mater.61(2009)793–796.[11]M.J.Alinger,Materials,University of California,Santa Barbara,2004.[12]M.J.Alinger,G.R.Odette,D.T.Hoelzer,J.Nucl.Mater.329(2004)382–386.[13]M.J.Alinger,G.R.Odette,D.T.Hoelzer,Acta Mater.57(2009)392–406.[14]M.Taneike,F.Abe,K.Sawada,Nature424(2003)294–296.[15]ler,C.L.Fu,M.Krcmar,D.T.Hoelzer,C.T.Liu,Front.Mater.Sci.China3(2009)9–14.[16]F.Bergner,A.Ulbricht,H.Hein,M.Kammel,J.Phys.Condensed Matter20(2008).[17]J.F.Loffler,H.B.Braun,W.Wagner,Phys.Rev.Lett.85(2000)1990–1993.[18]M.Ohnuma,J.Suzuki,S.Ohtsuka,S.W.Kim,T.Kaito,M.Inoue,H.Kitazawa,ActaMater.57(2009)5571–5581.[19]L.Yang,ler,X.L.Wang,C.T.Liu,A.D.Stoica,D.Ma,J.Almer,D.Shi,Adv.Mater.21(2009)305–308.[20]X.L.Wang,T.M.Holden,G.Q.Rennich,A.D.Stoica,P.K.Liaw,H.Choo,C.R.Hub-bard,Phys.B:Condensed Matter385(2006)673–675.[21]X.L.Wang,T.M.Holden,A.D.Stoica,K.An,H.D.Skorpenske,A.B.Jones,G.Q.Rennich,E.B.Iverson,Mater.Sci.Forum652(2010)105–110.[22]T.E.Mason,D.Abernathy,I.Anderson,J.Ankner,T.Egami,G.Ehlers,A.Ekkebus,G.Granroth,M.Hagen,K.Herwig,J.Hodges,C.Hoffmann,C.Horak,L.Horton,F.Klose,rese,A.Mesecar,D.Myles,J.Neuefeind,M.Ohl,C.Tulk,X.L.Wang,J.Zhao,Physica B385–386(2006)955–960.[23]U.Keiderling,Appl.Phys.A74(2002)S1455–S1457.[24]T.Hayashi,P.M.Sarosi,J.H.Schneibel,ls,Acta Mater.56(2008)1407–1416.[25]M.F.Collins,Magnetic Critical Scattering,Oxford University Press,Inc.,UK,1989.[26]H.A.Gersch,C.G.Shull,M.K.Wilkinson,Phys.Rev.103(1956)525–534.[27]Y.D.Wang,H.B.Tian,A.D.Stoica,X.L.Wang,P.K.Liaw,J.W.Richardson,Nat.Mater.2(2003)101–106.[28]C.L.Fu,M.Krcmar,G.S.Painter,X.Q.Chen,Phys.Rev.Lett.99(2007).[29]G.Gottstein,Physical Foundations of Materials Science,Springer,Berlin,2004.[30]R.D.Doherty,in:R.W.Cahn,P.Hason(Eds.),Physical Metallurgy,Elsevier Sci-ence B.V.,The Netherlands,1998.[31]G.R.Odette,M.J.Alinger,B.D.Wirth,Annu.Rev.Mater.Res.38(2008)471–503.[32]A.G.Certain,K.G.Field,T.R.Allen,ler,J.Bentley,J.T.Busby,J.Nucl.Mater.407(2010)2–9.[33]A.Hirata,T.Fujita,Y.R.Wen,J.H.Schneibel,C.T.Liu,M.W.Chen,Nat.Mater.10(2011)922–926.。