articles the role of metal nanoparticles and...

ARTICLES

The role of metal nanoparticles andnanonetworks in alloy degradation

Z. ZENG1*, K. NATESAN1, Z. CAI2 AND S. B. DARLING3

1Nuclear Engineering Division, Argonne National Laboratory, Argonne, Illinois 60439, USA2Advanced Photon Source, Argonne National Laboratory, Argonne, Illinois 60439, USA3Center for Nanoscale Materials, Argonne National Laboratory, Argonne, Illinois 60439, USA*e-mail: [email protected]

Published online: 11 July 2008; doi:10.1038/nmat2227

Oxide scale, which is essential to protect structural alloys from high-temperature degradation such as oxidation, carburization andmetal dusting, is usually considered to consist simply of oxide phases. Here, we report on a nanobeam X-ray and magnetic forcemicroscopy investigation that reveals that the oxide scale actually consists of a mixture of oxide materials and metal nanoparticles. Themetal nanoparticles self-assemble into nanonetworks, forming continuous channels for carbon transport through the oxide scales. Toavoid the formation of these metallic particles in the oxide scale, alloys must develop a scale without spinel phase. We have designed anovel alloy that has been tested in a high-carbon-activity environment. Our results show that the incubation time for carbon transportthrough the oxide scale of the new alloy is more than an order of magnitude longer compared with commercial alloys with similarchromium content.

The United States loses 4% of the gross national productannually owing to alloy corrosion1. Among the corrosion reactions,carburization and metal dusting, a high-temperature corrosion instrongly carburizing gas atmospheres at carbon activities (aC) > 1,are long-standing problems in energy conversion and productionsystems such as ethylene furnaces and hydrogen reformer plants2–5.It is estimated that 137 quadrillion joules of energy is lost annuallyowing to high-temperature corrosion problems2. All alloys aregenerally protected from continued corrosion by oxide scales thatdevelop on the alloy surface at high temperatures. To minimizethe corrosion problem, oxide scales on alloy surfaces need tobe carefully analysed and their composition and phases haveto be optimized to achieve high resistance in various corrosiveenvironments6–8. The diffusion rate of carbon in oxide scales isnegligible9. However, it is often seen that carbon diffuses into alloysand leads to brittleness and even pitting corrosion (a form ofextremely localized corrosion that leads to the creation of smallcraters in the alloy)10–20. Carbon transport through oxide scale istraditionally considered to involve the diffusion of carbon-bearingmolecules such as CO and CO2 through pores or cracks9,21.However, if carbon-bearing molecules can diffuse through thesedefects, oxygen can also diffuse through and react with alloyelements to form a new oxide scale, which will self-heal theoxide and stop the further intrusion of carbon-bearing molecules.According to the diffusion mechanism of oxidation proposed byWagner22, it is possible to form a layer comprising a mixture ofoxides and metal particles for a binary alloy23. However, formationof metal particles in oxide scale has been largely ignored9,21. It is wellknown that carbon can diffuse through nickel and iron metals24.Therefore, if metal particles are present in oxide scale in low pO2

environments, a new path for carbon atom transport is availablethat does not involve oxide defects. Our current study of oxidescales using nanobeam X-ray analysis, magnetic force microscopy(MFM) and scanning electron microscopy (SEM) has led us to

believe that metal nanoparticles are indeed present in the scale.These metal nanoparticles join to form continuous channels forcarbon transfer. This finding may have broad influence on notonly metal dusting and carburization, but also in other researchareas such as alloy development and surface coatings. For example,solid-oxide fuel cells require low resistivity for oxide scales thatdevelop in service on metallic interconnects25. If we can control theformation of clusters of metal particles in oxide scale, we may beable to greatly reduce the area resistance of metallic interconnects,even if the alloy develops an oxide in service. On the other hand, ifwe can prevent the formation of metal particles in oxide scale, thecarburization and metal-dusting problem can be greatly retardedby minimizing the carbon transfer channels.

Areas attacked by metal-dusting corrosion usually formpits10–21. Corrosion pits on the surface of Alloy 321 (composition:Fe-17.3%Cr–10.3%Ni–1.2%Mn–0.3%Si–0.4%Ti–0.04%C) wereobserved after 1,130 h exposure to a metal-dusting gas mixtureconsisting of 53.4%H2–5.7%CO2–18.4%CO–22.5%H2O (gas 1)at 14.3 atm and 593 ◦C with a carbon activity (aC) of 31. Afterexposure, the specimen was cut and a metallographic mount ofthe cross-section was prepared. The cross-section samples weremechanically polished on both sides to a thickness of 70 µm.Care was taken to minimize decomposition and contaminationduring sample preparation. The X-ray nanobeam (beam size of200 nm × 300 nm) was scanned across the cross-section at boththe pit and non-pit areas from the surface to the interior of thealloy. Non-pit areas have not yet been attacked by metal dusting.Surprisingly, strong diffraction (2θ at 42.2◦) from a Fe/Ni metalphase was observed in the oxide scale of non-pit areas (Fig. 1a).The diffraction peak position of the metal particles in the oxidescale is different from the peak position (2θ at 41.1◦) of the alloy,which indicates that the lattice parameters of these embedded metalparticles in the scale are different from the metal particles in thesubstrate alloy. Alloy 321 has an austenitic structure with the 111

nature materials VOL 7 AUGUST 2008 www.nature.com/naturematerials 641

© 2008 Macmillan Publishers Limited. All rights reserved.

mailto:[email protected]

http://www.nature.com/doifinder/10.1038/nmat2227

ARTICLES

42.2°41.1°

At pit

Alloy

Oxide scale

2 (° )θ3540

45

80

40

Non-pit

3438

4248

10

6

22 (°)θ

d (μm)

Oxide scale

Alloy

d (μm)

Oxide scale

Distance (μm)

Coun

t

At 5 μm

At 15 μmAt 25 μm

NiO

NiNiCr2O4

Fe

CrNi

Energy (keV)

–20 0 20 40 60 80 100

Norm

. X-r

ay a

bsor

ptio

n

8.32 8.33 8.34 8.35

Diffraction from Ni/Fe

Diffraction from Alloy 321

Diffraction from Alloy 321

Diffraction from spinel

a b

c d

Figure 1 X-ray data from oxides on Alloy 321 after 1,130 h exposure to a metal-dusting environment. a, XRD of the oxide scale at a non-pit area of Alloy 321. b, XRD ofthe oxide layer at a pit area of Alloy 321. c, Nanobeam X-ray fluorescence of the oxide layer at a pit area of Alloy 321 as a function of scale depth. d, X-ray near-edgeabsorption spectra of various nickel materials and of several locations in the vicinity of the oxide layer at a pit area of Alloy 321. The X-ray nanobeam was scanned across theoxide layer to the metal. The position of 5 µm is at the outer layer of the oxide layer.

diffraction peak at 41.1◦ (d-space at 2.08 A) when an X-ray beamenergy of 8.5 keV is used. However, the metal particles in oxidescale showed a ferritic structure with the 110 diffraction peak at42.2◦ (d-space at 2.03 A).

Both (Fe, Cr) spinel and Cr2O3 phases were observed by X-raydiffraction (XRD) at non-pit areas. There were two strata in theoxide layer at the pit area and only spinel phase was observedin both layers. No diffraction from metal particles or Cr2O3 wasseen in the oxide layer at the pit area (Fig. 1b). However, it issurprising that the nickel content in the inner layer of the scaleis higher than that in the substrate alloy (Fig. 1c). This result wasalso confirmed by energy dispersive X-ray (EDX) analysis, but EDXcould not establish the valence of the nickel in the oxide layer.On the other hand, nanobeam X-ray near-edge absorption analysisshowed that the chemical shift of nickel in most locations of theoxide layer is close to metallic nickel (see Fig. 1d). This indicates

the presence of metallic nickel particles in the oxide layer. These Niparticles are amorphous and were not seen in a conventional XRDscan (Fig. 1b).

EDX analysis was also carried out on the cross-section of Alloy321. The thickness of oxide scale at non-pit areas is only a fewmicrometres, whereas the thickness of oxide layer at pit areas is over50 µm. Bright spots in sizes ranging from several tens of nanometresto 800 nm were observed in the SEM image of oxide scale at thenon-pit area (Fig. 2a). EDX analysis indicated that these brightspots are rich in iron and nickel but contained less oxygen andchromium. There are no such bright spots in the oxide layer atpit areas of Alloy 321, but two distinct strata were observed in theoxide layer. EDX analysis indicated more oxygen and iron in theouter layer than in the inner layer, whereas the concentrations ofchromium and nickel were less in the outer layer than in the innerlayer (Fig. 2a).

642 nature materials VOL 7 AUGUST 2008 www.nature.com/naturematerials


ARTICLES

SEM Oxygen Chromium Iron Nickel

20 µm

1 µm

1 µm

1 µm

a

b c

Figure 2 Electron and scanning probe micrographs of Alloy 321 cross-sections after 1,130 h exposure in a metal-dusting environment. a, EDX mapping of thecross-section at pit and non-pit areas of Alloy 321. Top row: at metal-dusting pit area. Middle row: at non-pit area. Bottom row: enlarged mapping of the bright spots in theimage of the middle row. b, SEM image of oxide scale at a non-pit area of Alloy 321. c, MFM magnetic contrast image of the same location exhibiting some correlationbetween structural and magnetic features.

Atomic force microscopy has been used extensively in corrosionresearch8,26,27. Typically, MFM is used to characterize well-definedfeatures; however, low-contrast MFM imaging can also couplestructural and magnetic information with disordered materials28.The particles in oxide scale at the non-pit area identified asFe- and Ni-rich in the EDX analysis are probably composedof ferromagnetic iron, nickel or Fe/Ni alloy. Whereas themagnetic domains will presumably be isotropically oriented,those domains that have stray fields oriented perpendicularto the sample can produce contrast in an MFM image. Ironchromate, FeCr2O4, conversely, is paramagnetic and chromiumoxide, Cr2O3, is antiferromagnetic at room temperature. Thesesubstances, which are among the materials traditionally identifiedas making up oxide scale, will provide no contribution to MFMcontrast. The XRD and near-edge X-ray absorption analysisindicated that other ferromagnetic or ferrimagnetic oxides didnot form in the low-pO2 environment in this area. Therefore,we scanned through the regions containing bright spots inthe SEM micrographs of the oxide scale at non-pit areausing MFM. Figure 2c shows MFM data exhibiting magnetic

contrast correlating with the regions containing the particles(Fig. 2b), thus indicating ferromagnetism (or ferrimagnetism) inthese locations. Correlation between the MFM image and theSEM image is not perfect owing to such factors as potentialdistortion of nanoparticle magnetic domain orientations by thesample holder and/or the MFM probe and the fact that theMFM can only probe stray fields perpendicular to the sampleplane. However, the observation of magnetic contrast and itscorrelation with bright features from the electron micrograph,when viewed in the context of the XRD data, indicates the presenceof ferromagnetic Fe/Ni particles with ferritic structure in theoxide scale.

Metal nanoparticles were also observed in the oxide scale ofother alloys and in other test environments. Figure 3a shows thecross-section of oxide scales on Alloy 310 (Fe-19.5%Ni–25.5%Cr–1.7%Mn–0.7%Si–0.03%C) after 1,636 h exposure to a carburizinggas consisting of 65.1%H2–4%CO2–30%CO–0.9%H2O (gas 2) at1 atm and 593 ◦C with an aC of 104. Pits were observed afterthe short exposure period, even though the chromium contentin this alloy is high. The density of metal particles in the oxide



ARTICLES

layer in the pit area is higher than that at the non-pit area.An almost continuous channel of metal particles was observedfrom the surface through the oxide layer to the substrate alloy.Because carbon can rapidly diffuse through metals24, this metallicnetwork could enable transport of carbon in a matrix of oxides.In the non-pit regions, the particles have probably not formedcontinuous channels and the orientation of the metal clusters isprobably parallel rather than perpendicular to the alloy surface(Fig. 3a), thereby resulting in less metal-dusting attack. The figurealso shows a schematic representation of possible orientation ofcarbon-transfer channels in the two regions.

Because the electrical conductivity of metallic nickel and ironis 100 million times higher than that of spinel29, the electricalresistivity of the oxide scale at the pit area will be much lower thanthat at the non-pit area if metal clusters connect to form channelsin the pit area. The procedure for detecting low-resistance areas onthe surface of alloys is to deposit metal (such as copper) particlesby an electrochemical method in which copper is deposited onthe alloys by immersing them in CuSO4 solution and applying avoltage across the electrodes11. The areas with low resistance willhave deposits of copper particles because current can pass throughthem easily. Deposited copper particles thus act as an indicator tolocate the carbon transfer channels11. Figure 3b shows the copper-deposited region of a pit on the surface of Alloy 310. Such areas canbe identified under a microscope and probably indicate regions ofactive degradation by metal dusting11.

A metallic network was also observed in the oxide scalein the pit region of Alloy 353MA (composition: Fe–24.4%Cr–34.7%Ni−1.4%Mn−1.3%Si−0.1%Mo−0.18%N−0.06%V−0.05%C)after 1,130 h exposure to a metal-dusting gas mixture consistingof 40.2% H2–0.2%CO2–19.8%CO–0.1%H2O (gas 3) at 1 atm and593 ◦C with an aC of 200 (Fig. 3c). It seems that these networksform along the grain boundary regions (average grain size of700 nm) in the (Fe, Ni)Cr2O4 spinel phase. When the spineldecomposes, the ductile nickel particles deplete from the grain andconcentrate in the grain boundary regions to self-assemble into ametallic network. These metallic networks could act as pathwaysfor the transport of carbon into alloys and lead to metal-dustingcorrosion. The proposed models for carbon diffusion are shownin Fig. 3d. In the earlier model, carbon-containing gases such asCO and CO2 diffuse through pores, cracks or grain boundariesof oxides and the diffusion of carbon atoms through the scaleis not considered. However, in the proposed Model A, carbonatoms diffuse through oxide scale via metallic networks presentin the scale. Model B represents a hybrid between the earliermodel and proposed Model A. Carbon-containing gas moleculesdiffuse through the external oxide scale and deposit carbon atomsonto a metallic network in the scale. Atomic carbon, dissolved inthe metal, diffuses into the substrate alloy through the metallicnetworks in the oxide scale. In the earlier model, oxygen can alsodiffuse into the alloy and form a new oxide scale and self-healthe oxide scale to stop the further diffusion of carbon into alloys.However, in the proposed Models A and B, the metallic networkacts as a filter to allow only carbon atoms to diffuse into alloyswhile preventing the diffusion of oxygen because the diffusionrate of carbon in Ni is much faster than that of oxygen24. Thecarburization process will not be stopped by the self-healing processbecause a new oxide scale cannot form when oxygen cannot diffusein11,24. The continuous ingress of carbon finally leads to metal-dusting corrosion.

The large difference of carbon transport rate at the area witha metallic network from the area without a metallic networkleads to a localized corrosion. The results presented above indicateoccurrence of pitting corrosion when the metal particles jointogether to form a continuous channel through oxide scale for

Pit

Pit Non-pit

Non-pit1 µm

1 µm

Cracks

Current model

COCO2

Model B Model A

Incr

easi

ng p

O 2

AlloyAlloy

Oxide scale

COCO2

Metallic network

C C

a

b

c

d

Figure 3 SEM micrographs of alloys after exposure to a carburizing gas at1 atm and 593 ◦C. a, Cross-sections of pit and non-pit regions of Alloy 310. Thepotential carbon transfer channels are indicated by the red lines. b, EDX mapping ofthe surface of Alloy 310 that was electrochemically deposited by copper. The copperparticles (bright cyan spots) accumulated on the pit region. c, Cross-sections of a pitregion in Alloy 353MA. d, Comparison of the current carbon transport model and ournew models.

carbon transfer, whereas, in the oxide scale of the non-pit area,the metal particles have not formed such a continuous channel.Metallic nanoparticles are randomly distributed in oxide scale, sothe statistical likelihood of forming a continuous channel is low.A consequence of this situation is that the pit density could beless than one per cm2. The density of metallic nanoparticles inoxide scales can be affected by the phase composition of the scale.Cr2O3 cannot be reduced into metallic particles in a metal-dustingenvironment, but (FeNi)Cr2O4 spinel can be. Greater spinelcontent in oxides is expected to lead to increased density of metallic



ARTICLES

2.4 mm

2.4 mm

Alloy 601

ANL1

1.4 mm

1 mm

50 µm

Cr2O3

Spinel

Raman shift (cm–1)

Inte

nsity

601

ANL1

250 µm

400 500 600 700

Figure 4 Three-dimensional profile mapping and Raman spectra of Alloys 601and ANL1 after 12,858 h exposure to the same metal-dusting environment at593 ◦C and 1 atm. A large metal-dusting pit with depth of 490 µm was observed,but the surface of ANL alloy was still smooth.

nanoparticles in oxide scale and, therefore, a greater likelihood toform a continuous channel. Indeed, we have observed that the pitdensity increases with increasing iron content30, because the spinelcontent in oxide scale increases with increasing iron content in thefollowing reaction:

Fe+Cr2O3 +1

2O2 = FeCr2O4.

In the pit area of Alloy 321, Ni2+ was observed at the outer layer ofthe oxide layer at a position close to the surface (Fig. 1d). Although

pure NiCr2O4 is not stable in the reducing environment of thepresent study, a small amount of Ni2+ may enter the Fe2+ sites inFeCr2O4 to form a (Fe,Ni)Cr2O4 solid solution. The maximumconcentration of Ni2+ in the solid solution is 15.2% in gas 1at 14.3 atm and 593 ◦C, calculated from thermochemical data31.However, the calculated value for the maximum concentration ofNi2+ at the oxide–substrate interface, where the oxygen partialpressure is much less than in gas 1, is only 1.2×10−5%. Therefore,the concentration of Ni2+ in the spinel should be high at the gas sideof the oxide scale and it will be low in the oxide on the alloy side ofthe scale.

In the early stage of oxide scale formation, nickel could beoxidized to 2+ at the gas–oxide surface and dope into the solidsolution of (Fe,Ni)Cr2O4. When cations continue to diffuse out,the nickel-containing spinel is buried inside. The pO2 in the innerlayer of the oxide scale decreases with increasing thickness of theoxide scale, and Ni2+ will be reduced to nickel metal when the pO2

drops below the critical value. Such a process leads to the formationof nickel particles in oxide scale. FeCr2O4 spinel itself is not stableeither. Iron ions can also be reduced to metallic particles in theoxide scale. Therefore, formation of the metallic nanoparticles canoccur owing to the reduction of nickel and iron ions in the spinelphase during the growth of the oxide scale.

Nickel oxidation and reduction processes can occur only if ironis present in alloys because pure NiCr2O4 spinel will not formin a reducing environment. Only when FeCr2O4 forms, Ni2+ candope into iron-containing spinel to form a (Fe,Ni)Cr2O4 solidsolution. In the absence of iron, nickel will not be subjected to theoxidation–reduction process. Therefore, the spinel phases in oxidescale can lead to formation of unstable iron-rich regions and actas a solvent for the formation of Ni2+; the Ni2+ will eventuallybe reduced to nickel metal (over a long time), thereby enablingdevelopment of carbon transfer channels.

According to the analysis above, metal dusting would be greatlyretarded if the alloys could develop an oxide scale without theformation of metal nanoparticles or nanonetworks. In the absenceof iron in the alloy, the scale would not contain an iron-containingspinel and NiCr2O4 would be thermodynamically unstable in thereducing environment of reformers. Therefore, it is desirable toprepare alloys that can develop an oxide scale without the spinelphase. We designed and prepared several nickel-base alloys withlow iron content. The chromium content of Alloy ANL1 (74.5% Ni,22% Cr, 2.3%Al, 0.7%Fe, 0.3%Ti, 0.2%Zr, and 0.1%C) is similarto commercial Alloy 601 (61.8%Ni, 21.9%Cr, 1.4%Al, 14.5%Fe,0.3%Ti, 0.1%Nb, and 0.03%C), but the Fe content of the ANL1alloy is only 0.7% compared with 14.5% in Alloy 601.

Raman scattering spectra indicate formation of spinel phaseon the surface of commercial Alloy 601 but not on the surface ofthe low-iron alloy after 12,858 h exposure in gas 2 at 1 atm and593 ◦C with an aC of 104 (Fig. 4). On the other hand, metal-dustingpits were observed on Alloy 601 within 1,000 h exposure. Althoughthe composition of ANL1 is similar to that of Alloy 601, exceptthe low iron content in ANL1, the incubation time for metaldusting in this alloy is much longer than that for Alloy 601.This result indicates that carbon has not penetrated through thenon-spinel oxide scale on ANL1 alloy, whereas its transfer wasevident when the scale contains spinel phase, as in the case ofAlloy 601.

The presence of metallic nanoparticles in the oxide scale onalloys has been validated. These nanoparticles and nanonetworksin the oxide scale provide pathways for the fast transport ofsome elements such as carbon and hydrogen, thereby greatlyaffecting the alloy performance at high temperatures. The resultsenable us to re-evaluate the integrity of oxide scales from thestandpoint of corrosion protection. Formation conditions and



ARTICLES

CCD

Oxide scale

Alloy

Incident beam

2

Scan

ning

path

X–ra

y de

tect

or

θ

Figure 5 Schematic diagram showing the geometry of data acquisition at thesynchrotron beamline.

mechanisms of these metallic nanoparticles and networks shouldbe further studied with the twin goals of controlling their formationfor applications such as reducing the electrical resistance ofoxide scale on metallic interconnects for solid-oxide fuel cellsor improving the bonding between alloys and oxide scales andminimizing their formation to avoid metal dusting, carburizationand hydrogen embrittlement.

METHODS

Nanobeam X-ray analysis was carried out at the 2ID-D beamline of the APS atArgonne National Laboratory. Using a Si(001) double-crystal monochromatorand zone-plate optics, the X-ray nanoprobe in the experimental stationproduces a monochromatic X-ray beam of size 200 nm × 300 nm with aphoton flux of 5×109 photons s−1 with an X-ray energy bandwidth (dE/E)of 0.01%. The focal plane of the zone-plate optics was adjusted so that theminimum spot size was obtained at the surface of the specimen. Figure 5 showsthe scattering and data acquisition geometry for the synchrotron radiationexperiments: the beam is incident on the region of interest on the specimen,and the scattering intensity in a section of reciprocal space is captured by aflat, two-dimensional detector (CCD camera). Samples for the synchrotronradiation experiments were prepared in a similar manner as for the cross-sectionanalysis in scanning electron microscopy. The beam energy was 8.5 keV for theXRD experiment. The position of the oxide–metal interface was determinedby monitoring the Fe-fluorescence counts as the sample was translated acrossthe beam.

MFM measurements were carried out using a vibrationally damped VeecoMultiMode V system with silicon probes coated with CoCr alloy (approximately25 nm tip radius of curvature).

Received 7 February 2008; accepted 17 June 2008; published 11 July 2008.

References1. Ricker, R. E. Cost of corrosion. Science 252, 1232–1232 (1991).2. Brindle, R. & Winkel, D. Energy impacts of corrosion, energy efficiency and renewable energy annual

report and presentation (2005), <http://www1.eere.energy.gov/industry/imf/pdfs/7 corrosion.pdf>.3. Zeng, Z. & Natesan, K. Relationship between the growth of carbon nanofilaments and metal dusting

corrosion. Chem. Mater. 17, 3794–3801 (2005).4. Young, D. J. Metal dusting reaction mechanisms. Mater. Sci. Forum. 522, 15–26 (2006).5. Jakobi, D. & Gommans, R. Typical failures in pyrolysis coils for ethylene cracking. Mater. Corros. 54,

881–886 (2003).6. Yamamoto, Y. et al. Creep-resistant, Al2O3-forming austenitic stainless steels. Science 316,

433–436 (2007).7. Thurmer, K., Williams, E. & Reutt-Robey, J. Autocatalytic oxidation of lead crystallite surfaces.

Science 297, 2033–2035 (2002).8. Stierle, A. et al. X-ray diffraction study of the ultrathin Al2O3 layer on NiAl(110). Science 303,

1652–1656 (2004).9. Wolf, I. & Grabke, H. J. A study on the solubility and distribution of carbon in oxides. Solid State

Commun. 54, 5–10 (1985).10. Hochman, R. F. in Proc. Symp. on Properties of High-Temperature Alloys with Emphasis on

Environmental Effects (eds Foroulis, Z. A. & Pettit, F. S.) 715–732 (The Electrochemical Society,New Jersey, 1977).

11. Zeng, Z. & Natesan, K. Initiation of metal-dusting pits and a method to mitigate metal-dustingcorrosion. Oxid. Met. 66, 1–20 (2006).

12. Baker, B. A., Hartmann, V. W., Shoemaker, L. E., McCoy, S. A. & Rajendran, S. A new nickel-basealloy for metal dusting resistance. Trans. Ind. Inst. Met. 56, 327–333 (2003).

13. Ackermann, H. et al. Metal dusting in low-NOx recirculation burners for fuel oil. Corros. Eng. Sci.Technol. 40, 233–238 (2005).

14. Korkhaus, J. Failure mechanisms and material degradation processes at high temperatures inammonia synthesis. Corros. Eng. Sci. Technol. 40, 204–210 (2005).

15. Albertsen, J. Z., Grong, O., Mathiesen, R. H. & Schmid, B. Metallurgical investigation of metal dustingcorrosion in plant-exposed nickel-based alloy 602CA. Corros. Eng. Sci. Technol. 40, 239–243 (2005).

16. Nishiyama, Y., Otsuka, N. & Kudo, T. Metal dusting behaviour of Cr–Ni steels and Ni-base alloys in asimulated syngas mixture. Corros. Sci. 48, 2064–2083 (2006).

17. Zhang, J., Cole, D. M. I. & Young, D. J. Alloying with copper to reduce metal dusting of nickel. Mater.Corros. 56, 756–764 (2005).

18. Szakalos, P., Pettersson, R. & Hertzman, S. An active corrosion mechanism for metal dusting on 304Lstainless steel. Corros. Sci. 44, 2253–2270 (2002).

19. Schneider, A., Viefhaus, H., Inden, G., Grabke, H. J. & Mullerlorenz, E. M. Influence of H2S on metaldusting. Mater. Corros. 49, 336–339 (1998).

20. Di Gabriele, F., Stott, F. H. & Liu, Z. Effect of experimental conditions on the metal dustingphenomenon in several commercial nickel-base alloys. Mater. Corros. 58, 81–86 (2007).

21. Grabke, H. J. Mechanisms and prevention of corrosion in carbonaceous gases. Mater. Sci. Forum 369,101–108 (2001).

22. Wagner, C. Theory of the tarnishing process. Z. Phys. Chem. B21, 25–41 (1933).23. Maak, F. & Wagner, C. Mindestgehalte von Legierungsbestandteilen fur die Bildung von

Oxydschichten hoher Schutzwirkung gegen Oxydation bei hoheren Temperaturen. Mater. Corros. 12,273–277 (1961).

24. Bergner, D. Diffusion of C, N and O in metals, diffusion in metals and alloys. Proc. Int. Conf. 671,223–240 (1983).

25. Quadakkers, W. J, Piron-Abellan, J., Shemet, V. & Singheiser, L. Metallic interconnectors for solidoxide fuel cells—a review. Mater. High Temp. 20, 115–127 (2003).

26. Renner, F. U. et al. Initial corrosion observed on the atomic scale. Nature 439, 707–710 (2006).27. Schmutz, P. & Frankel, G. S. Corrosion study of AA2024-T3 by scanning Kelvin probe force

microscopy and in situ atomic force microscopy scratching. J. Electrochem. Soc. 145,2295–2306 (1998).

28. Buchanan, K. S., Zhu, X. B., Meldrum, A. & Freeman, M. R. Ultrafast dynamics of a ferromagneticnanocomposite. Nano Lett. 5, 383–387 (2005).

29. Thayer, P. C. The Electrical Properties of Iron–Chromium Spinel, thesis for Degree of Doctor ofPhilosophy, Colorado State Univ., Fort Collins, Colorado (1972).

30. Natesan, K. & Zeng, Z. Argonne National Laboratory Report, ANL-07/30, 68–72, October, (2007).31. Muller, F. & Kleppa, O. J. Thermodynamics of formation of chromite spinels. J. Inorg. Nucl. Chem.

35, 2673–2678 (1973).

AcknowledgementsWe thank D. L. Rink for conducting the metal-dusting experiments, S. Cai for data analysis of thenanobeam XRD and J. Froitzheim for SEM analysis. This work is supported by the US Departmentof Energy, Office of Industrial Technologies. Use of the Advanced Photon Source, the Center forNanoscale Materials and the Electron Microscopy Center for Materials Research were supported bythe US Department of Energy, Office of Science, Office of Basic Energy Sciences, under Contract No.DE-AC02-06CH11357.

Author contributionsZ.Z. and K.N. planned the experiments and analysed the data, Z.C. carried out the APS nanobeamexperiment and S.B.D. carried out MFM measurements and contributed to data analysisand interpretation.

Author informationReprints and permission information is available online at http://npg.nature.com/reprintsandpermissions.Correspondence and requests for materials should be addressed to Z.Z.



http://www1.eere.energy.gov/industry/imf/pdfs/7_corrosion.pdf

http://npg.nature.com/reprintsandpermissions

articles the role of metal nanoparticles and...

Documents