Nanophase Transition Metal Oxides Show Large Thermodynamically Driven Shifts in Oxidation-Reduction Equilibria

DOI: 10.1126/science.1195875 , 199 (2010); 330Science

et al.Alexandra Navrotsky,Oxidation-Reduction EquilibriaThermodynamically Driven Shifts in Nanophase Transition Metal Oxides Show Large

This copy is for your personal, non-commercial use only.

. clicking herecolleagues, clients, or customers by , you can order high-quality copies for yourIf you wish to distribute this article to others

. herefollowing the guidelines can be obtained byPermission to republish or repurpose articles or portions of articles

(this information is current as of October 7, 2010 ):The following resources related to this article are available online at www.sciencemag.org

http://www.sciencemag.org/cgi/content/full/330/6001/199version of this article at:

including high-resolution figures, can be found in the onlineUpdated information and services,

http://www.sciencemag.org/cgi/content/full/330/6001/199/DC1 can be found at: Supporting Online Material

http://www.sciencemag.org/cgi/content/full/330/6001/199#otherarticles, 7 of which can be accessed for free: cites 23 articlesThis article

http://www.sciencemag.org/cgi/collection/geochem_physGeochemistry, Geophysics

http://www.sciencemag.org/cgi/collection/chemistryChemistry

: subject collectionsThis article appears in the following

registered trademark of AAAS. is aScience2010 by the American Association for the Advancement of Science; all rights reserved. The title

CopyrightAmerican Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by theScience

on

Oct

ober

7, 2

010

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

Nanophase Transition Metal OxidesShow Large Thermodynamically DrivenShifts in Oxidation-Reduction EquilibriaAlexandra Navrotsky,* Chengcheng Ma, Kristina Lilova, Nancy Birkner

Knowing the thermodynamic stability of transition metal oxide nanoparticles is important forunderstanding and controlling their role in a variety of industrial and environmental systems.Using calorimetric data on surface energies for cobalt, iron, manganese, and nickel oxide systems,we show that surface energy strongly influences their redox equilibria and phase stability.Spinels (M3O4) commonly have lower surface energies than metals (M), rocksalt oxides (MO),and trivalent oxides (M2O3) of the same metal; thus, the contraction of the stability field ofthe divalent oxide and expansion of the spinel field appear to be general phenomena. Usingtabulated thermodynamic data for bulk phases to calculate redox phase equilibria at thenanoscale can lead to errors of several orders of magnitude in oxygen fugacity and of 100 to 200kelvin in temperature.

Differences in surface energies alter rela-tive free energies of polymorphs (mate-rials with different crystal structures but

the same composition), which cause size-driventhermodynamic crossovers in phase stability atthe nanoscale (1–4). Because oxyhydroxides gen-erally have smaller surface energies than oxides,dehydration reactions (2FeOOH→ Fe2O3 +H2O)shift to higher temperatures by as much as 100 Kat the nanoscale (4). These shifts in phase sta-bility change what materials form under givenconditions (pressure, temperature, or humidity) andaffect physical properties and chemical reactivity.Here, we document similarly large thermodynam-ic shifts in the positions of oxidation-reduction(redox) equilibria in oxygen fugacity-temperaturespace for nanoscale transition metal oxides (5).

The Co-O system was chosen for detailedstudy because both Co(II) oxide (rocksalt CoO)and Co(II, III) oxide (Co3O4 spinel) can be pre-pared and characterized as both bulk and nano-scale materials, and the surface energy of CoOhas already been measured (6). There is recentinterest in the Co-O system for water splitting(7, 8), but it also serves as a useful model forother transition metal-oxygen systems, such asFe-O,Mn-O, andNi-O. Using calorimetric meth-ods (5), we determined the surface energies forhydrated and anhydrous surfaces of CoO andCo3O4. To complement the Co-O data, we alsoexamined geochemically and technologically rel-evant metal oxides, includingNiO, Fe3O4,Mn2O3,and Mn3O4 (5).

According to the surface energies in Table 1,calculated changes in enthalpy occur relative to

the bulk phasewith increasing surface area (Fig. 1).The surface energy of oxides is based on calori-metric data, whereas that of anhydrous Co metalis based on computation (9), and that of thehydrated metal surface is approximated as 75%that of the anhydrous surface. This is roughly theratio for oxides, but metal surfaces probably hy-

drate less strongly than oxides, so the actualdifference may be less for metals. Thus, thedifference in surface energy for hydrated surfacesbetween metal and oxide may be even greaterthan this estimate, which would make the trueshifts in redox equilibria even greater. A steeperslope in Fig. 1 indicates a larger surface energyand a greater destabilization for small particles.CoO is thus more destabilized than either Co orCo3O4 for a given surface area. The Fe-O systembehaves similarly (Fig. 1).

A number of spinel phases have lower surfaceenergies than other oxides of the same metal(Table 1). Previous work showed that the lowersurface energy of the defect spinel forms ofAl2O3 and Fe2O3 relative to the corundum formcauses a crossover in phase stability at the nano-scale (1, 10). Data on MgAl2O4 (11) and Co3O4,Fe3O4, and Mn3O4 also confirm low surfaceenergies for these spinels (Table 1). Thus, itseems probable that other transition metal spinels(such as CoFe2O4, NiFe2O4, CoMn2O4, and per-haps even the lithium-bearing spinels used inbatteries) have low surface energies as comparedwith those of more oxidized and more reducedphases, and the size-driven changes in redoxequilibria may be general phenomena.

The observed differences in surface energiesshift redox free energies by 10 to 30 kJ/mol,oxygen fugacity by several orders of magnitude,

REPORTS

Peter A. Rock Thermochemistry Laboratory and Nanomaterialsin the Environment, Agriculture, and Technology OrganizedResearch Unit, University of California at Davis, Davis, CA95616, USA.

*To whom correspondence should be addressed. E-mail:[email protected]

0 5000 10000 150000

20

40

60CoOCoCoO

CoOCoCoO

A

C D

B

Surface area (m2/mol)

Exc

ess

enth

alpy

(kJ

/mol

)

Surface area (m2/mol)

Surface area (m2/mol) Surface area (m2/mol)

Exc

ess

enth

alpy

(kJ

/mol

)

Exc

ess

enth

alpy

(kJ

/mol

)

Exc

ess

enth

alpy

(kJ

/mol

)

00

20

40

60

0 2000 4000

10

0

20

0 2000 4000

10

0

20

5000 10000 15000

FeOFeα - FeO

FeOFeα - FeO

FeO FeO

NiONiOMnO

1.331.33

1.33

1.331.33

1.5 1.5

Fig. 1. Enthalpy relative to bulk phase and liquid water (excess enthalpy in kJ/mol, caused by increase insurface area), plotted versus surface area for (A) Co-O phases with anhydrous surfaces, (B) Co-O phaseswith hydrated surfaces, (C) Fe-O phases with anhydrous surfaces, and (D) Fe-O phases with hydratedsurfaces. Oxide data (Co3O4, Fe3O4, Mn3O4, and Fe2O3) in table S1 are converted to enthalpy per mole ofCoO1.33, FeO1.33, MnO1.33, and FeO1.5 along with Co and Fe to maintain 1 mol metal stoichiometry forclear comparison of all phases.

www.sciencemag.org SCIENCE VOL 330 8 OCTOBER 2010 199

on

Oct

ober

7, 2

010

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

or temperature by 100 to 200 K for 10-nm-diameter particles. A general formulation of theeffect of particle size on chemical equilibriaamong solid phases is that systems favor phaseassemblages of lower surface energy to counter-act the high surface areas that are inherent tonanoparticles.

The calculated Co-O phase diagram for 10-nmnanoparticles confirms a much diminished sta-bility field for the divalent oxide (Fig. 2). Thisnarrowing is consistent with several qualitativeobservations made during the synthesis andhandling of transition metal oxide nanoparticles.Decomposition ofCoCO3 in vacuumusually yields

Co3O4 unless the reaction occurs at high temper-ature, in which the divalent oxide coarsens (12).Size-driven oxidation by CO2 also occurs whendecomposing MnCO3 and FeCO3 (13). At a rel-atively high CO/CO2 ratiowith little oxygen supply,nano-Co3O4—which we argue to be thermody-namically stabilized at the nanoscale under cat-alytic conditions—makes an excellent catalystfor low-temperature CO oxidation (14). Indeed,understanding nanoscale redox shifts in thesesystems may hold a key to understanding anddesigning efficient catalysts, including cobalt-based nanocluster catalysts for the splitting ofwater (7, 8). We observed that CoO nanoparticles

smaller than 8 nm, when dropped into water,evolved a gas that was shown to be hydrogen byits flammability. Such spontaneous oxidation ofCo2+ and reduction of H2O is not seen for largerparticles. We also observed that 13-nm Mn2O3

particles, when heated in air at 975 K for 4 hours,are partially reducedwith the appearance of smallamounts of Mn3O4 (5), whereas 38-nm particlesare not reduced. Such reduction is thermody-namically spontaneous in air for coarse materialonly above 1138 K according to standard ther-modynamic data (15). These observations pro-vide further evidence for shifts in the free energiesof redox reactions at the nanoscale.

Using the values for surface energies in Table1 and thermodynamic data for the bulk Fe-Osystem (5), the phase equilibria involving metalFe, wustite Fe0.947O, magnetite Fe3O4, and hem-atite a-Fe2O3 were computed (Fig. 2). The redoxchemistry is altered drastically at the nanoscale,with the elimination of Fe0.947O as a stable phase,substituted by direct equilibrium between Fe andFe3O4. Fe0.947O has a low-temperature stabilitylimit of 850 K relative to Fe and Fe3O4 in thebulk (15). At the nanoscale, this temperature great-ly increases. The calculations imply that Fe0.947Onanoparticles are unlikely to form below 100 nmin size because the lowest temperature at whichthey would be stable with respect to Fe3O4 + Fe,~1000 K, is high enough that coarsening wouldbe rapid. Particles below ~16 nm in diameterwould not be stable up to the melting point ofbulk Fe0.947O.

The interpretation of the conditions of forma-tion of Fe-O phase assemblages in technological,geological, and planetary environments must bedone in the context of the effect of surface energyon the position of the redox boundaries. Wustitenanoparticles of a few nanometers in diameterproduced by high-energy ball milling (16) spon-taneously oxidize, starting at the surface, to Fe3O4.This creates a core-shell structure with exchangeanisotropy that is critical for hard disk read-headand sensor applications (17, 18). We argue thatthe surface thermodynamics drives the formationof such technologically useful multiphase nano-composite structures.

Redox and dissolution reactions of iron andmanganese oxides provide energy sources formicrobial communities (19). For example, Fe(II)–rich mineral dissolution is usually accompaniedby oxidation, which supports Fe-oxidizing micro-organisms (20). Shifts in thermodynamics at thenanoscale will change the free energy of suchreactions. It is conceivable that organisms tailorparticle size to optimize both the kinetics and thefree energy change available to them. It is alsopossible that nanoparticle redox reactions, withenergetics different from the bulk, may haveplayed a role in chemical processes leading toabiotic organic synthesis and the origin of life onthe early Earth.

The smaller range of stability of divalent tran-sition metal oxides at the nanoscale also implieseasier reduction to metal in vacuum. Small par-

Table 1. Surface energies for transition metal oxides and related systems.

Phase Surface energy(anhydrous surface) (J/m2)

Surface energy(hydrated surface) (J/m2)

Co (hcp) 2.22 T 0.30 (9) 1.66 T 0.23*Fe (bcc) 2.12 T 0.29 (9) 1.59 T 0.22*Fe0.947O (rocksalt) [3.6]† [2.8]†CoO (rocksalt) 3.57 T 0.30 2.82 T 0.20 (6)NiO‡ (rocksalt) 3.5 T 0.5 2.4 T 0.4Co3O4 (spinel) 1.96 T 0.05 0.92 T 0.04Fe3O4‡ (spinel) 1.44 T 0.30 0.79 T 0.28Mn3O4‡ (spinel) — 0.5 T 0.4g-Fe2O3 (spinel) 0.71 T 0.13 (10) 0.57 T 0.10 (10)g-Al2O3 (spinel) 1.53 T 0.40 (25) 0.7 T 0.4 (25)MgAl2O4 (spinel) 1.8 T 0.3§ 0.9 T 0.3§Fe2O3 (hematite) 1.9 T 0.3 (26) 0.75 T 0.16 (27)Mn2O3‡ (bixybite) — 1.66 T 0.21*Energy of hydrated metal surface is assumed to be 0.75 that of the anhydrous surface, as discussed in the text. †Thesimilarity of values for CoO and NiO suggests that it is reasonable to assign a similar surface energy to wustite, Fe0.947O.Changes in the oxygen nonstoichiometry of wustite at the nanoscale are not considered. Neglecting such variations does notsignificantly change the positions of the redox equilibria. ‡For calorimetric data, see table S1. §Calculated usingdata from previous study (11).

300 900

-60

-40

-20

0

bulk

Co

Co3O4

Fe3O4 Fe3O4 Fe3O4

α – Fe2O3α – Fe2O3α – Fe2O3

Co3O4 Co3O4

600 300 900

-60

-40

-20

0

Co

10 nm

600 300 900

-60

-40

-20

0

Co

10 nm

600

800 1000 1200

-20

-10 bulk

Fe

Fe0.947

O

800 1000 1200

-20

-10 10 nm

Fe

800 1000 1200

-20

-10 10 nm

Fe

A B C

D E F

Temperature (K) Temperature (K) Temperature (K)

Temperature (K) Temperature (K) Temperature (K)

log[

P (

O2/

1 at

m)]

log[

P (

O2/

1 at

m)]

log[

P (

O2/

1 at

m)]

log[

P (

O2/

1 at

m)]

log[

P (

O2/

1 at

m)]

log[

P (

O2/

1 at

m)]

CoOCoOCoO

Fig. 2. Calculated phase diagrams for bulk and 10-nm-diameter spherical particles. (A) Co-O (bulk). (B)Co-O (anhydrous 10-nm particles). (C) Co-O (hydrated 10-nm particles). (D) Fe-O (bulk). (E) Fe-O(anhydrous 10-nm particles). (F) Fe-O (hydrated 10-nm particles). The temperature scale is in 1/T, in orderto make a linear plot, although the labeled temperatures are in T (kelvin). There is no stability field for 10-nmwustite Fe0.947O.

8 OCTOBER 2010 VOL 330 SCIENCE www.sciencemag.org200

REPORTS

on

Oct

ober

7, 2

010

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

ticles of hydrated NiO and NiFe2O4 form Nimetal when examined by means of transmissionelectronmicroscopy (21, 22). There are reports ofunusual magnetism in semiconductors (ZnO andTiO2) doped with cobalt (23, 24). The crystallo-graphic position, clustering, phase separation,and oxidation state of the dopant are critical tounderstanding the origin of the reported ferro-magnetism. Changes in the Co-O phase diagramat the nanoscale present additional variables,particularly because many measurements of spec-troscopic, structural, and physical properties aredone under vacuum conditions that may produceCo metal more readily in nanoscale systems.

References and Notes1. J. M. McHale, A. Auroux, A. J. Perrotta, A. Navrotsky,

Science 277, 788 (1997).2. M. R. Ranade et al., Proc. Natl. Acad. Sci. U.S.A. 99

(suppl. 2), 6476 (2002).3. A. Navrotsky, J. Chem. Thermodyn. 39, 2 (2007).4. A. Navrotsky, L. Mazeina, J. Majzlan, Science 319,

1635 (2008).5. Materials and methods are available as supporting

material on Science Online.

6. L. Wang et al., Chem. Mater. 16, 5394 (2004).7. M. W. Kanan, D. G. Nocera, Science 321, 1072 (2008).8. J. Feng, H. Frei, Angew. Chem. 121, 1873 (2009).9. W. R. Tyson, W. A. Miller, Surf. Sci. 62, 267 (1977).10. O. Bomati-Miguel, L. Mazeina, A. Navrotsky,

S. Veintemillas-Verdaguer, Chem. Mater. 20, 591 (2008).11. J. M. McHale, A. Navrotsky, R. J. Kirkpatrick,

Chem. Mater. 10, 1083 (1998).12. D. Mehandjiev, E. Nikolova-Zhecheva, Thermochim. Acta

37, 145 (1980).13. K. H. Stern, High Temperature Properties and Thermal

Decomposition of Inorganic Salts with Oxyanions(CRC Press, Boca Raton, FL, 2000), pp. 17–20.

14. X. Xie, Y. Li, Z. Q. Liu, M. Haruta, W. Shen, Nature 458,746 (2009).

15. R. A. Robie, B. S. Hemingway, Thermodynamic Propertiesof Minerals and Related Substances at 298.15 K and1 Bar Pressure and at Higher Temperatures(U.S. Geological Survey Bulletin 2131, 1995),pp. 190–193, pp. 199–202.

16. M. Mozaffari, M. Gheisari, M. Niyaifar, J. Amighian,J. Magn. Magn. Mater. 321, 2981 (2009).

17. C. Chen, R. Chiang, H. Lai, C. Lin, J. Phys. Chem. C 114,4258 (2010).

18. G. C. Papaefthymiou, Nano Today 4, 438 (2009).19. K. Kashefi, D. R. Lovley, Science 301, 934 (2003).20. S. A. Welch, J. F. Banfield, Geochim. Cosmochim. Acta

66, 213 (2002).

21. M. P. Harmer, R. K. Mishra, G. Thomas, J. Am. Ceram.Soc. 66, 44 (1983).

22. P. K. Davies, I. D. R. Mackinnon, J. Am. Ceram. Soc. 69,124 (1986).

23. T. Dietl, H. Ohno, F. Matsukura, J. Cibert, D. Ferrand,Science 287, 1019 (2000).

24. K. Ueda, H. Tabata, T. Kawai, Appl. Phys. Lett. 79,988 (2001).

25. R. Castro, S. V. Ushakov, L. Gengembre, D. Gouvêa,A. Navrotsky, Chem. Mater. 18, 1867 (2006).

26. J. Majzlan et al., Am. Mineral. 88, 846 (2003).27. J. Majzlan, thesis, University of California at Davis

(2003).28. The work received support from the U.S. Department of

Energy, grants DE-FGO2-05ER1S667 (Co-O and Ni-O)and DE-FGO2-07ER14749 (Fe-O and Mn-O). We thankH. Ma, B. Olsen, B. Woodfield, W. Casey, and Y. Du fortechnical assistance and valuable discussion.

Supporting Online Materialwww.sciencemag.org/cgi/content/full/330/6001/199/DC1Materials and MethodsFigs. S1 and S2Tables S1 to S3References

30 July 2010; accepted 31 August 201010.1126/science.1195875

Tracking Hydrocarbon PlumeTransport and Biodegradation atDeepwater HorizonRichard Camilli,1* Christopher M. Reddy,2 Dana R. Yoerger,1 Benjamin A. S. Van Mooy,2

Michael V. Jakuba,3 James C. Kinsey,1 Cameron P. McIntyre,2 Sean P. Sylva,2 James V. Maloney4

The Deepwater Horizon blowout is the largest offshore oil spill in history. We present results from asubsurface hydrocarbon survey using an autonomous underwater vehicle and a ship-cabledsampler. Our findings indicate the presence of a continuous plume of oil, more than 35 kilometersin length, at approximately 1100 meters depth that persisted for months without substantialbiodegradation. Samples collected from within the plume reveal monoaromatic petroleumhydrocarbon concentrations in excess of 50 micrograms per liter. These data indicate thatmonoaromatic input to this plume was at least 5500 kilograms per day, which is more thandouble the total source rate of all natural seeps of the monoaromatic petroleum hydrocarbons inthe northern Gulf of Mexico. Dissolved oxygen concentrations suggest that microbial respirationrates within the plume were not appreciably more than 1 micromolar oxygen per day.

The Deepwater Horizon blowout at theMC252 Macondo well site released morethan 4 million barrels (636 million liters)

of oil into the Gulf of Mexico (1). Its scale andsource depth, at 1500 m below the sea surface,represent a relatively uninvestigated category ofoil spill. The mechanisms of plume formationare complex due to many factors, including theinterplay of gas and oil in multiphase flow, pref-

erential solubility of each oil constituent, and po-tential gas hydrate formation (2). Consequently,deep-water oil spills are difficult to model, andplume dynamics remain challenging to predict(2–4). Many deep-water models include the Gulfof Mexico in their spill scenarios (4–6).

We initially observed a subsurface layer of oilbetween 1030- and 1300-m depth during a U.S.Coast Guard–authorized flow-assessment effortat the well site in late May and early June 2010(fig. S1). To further characterize any resultantplume stemming from the Deepwater Horizonblowout, we performed a 10-day subsurfacesampling effort, including three long-range sur-veys from 19 to 28 June 2010 using the NationalDeep Submergence Facility’s autonomous under-water vehicle (AUV) Sentry (fig. S7) and a cable-lowered sample-collection rosette (fig. S2), each

equipped with a TETHYS in situ membrane inletmass spectrometer (7, 8). Sentry was chosen forthese operations based on this vehicle class’sdemonstrated utility in characterizing deep-oceanhydrothermal vents (9) and cold seeps (10).Sampling made use of an iterative approach ofin situ sensing and automated data analysis toidentify select petroleum hydrocarbons and anyassociated oxygen anomalies. The three Sentrysurveys, all conducted between 23 and 27 June2010 at depths in excess of 1000 m, operatedfor 64 hours to cover a linear distance of 235 km.During these deployments, Sentry’s mass spec-trometer recorded more than 3500 discrete sam-ple measurements, simultaneously tracking 10independent chemical parameters in real time.Another 2300 samplemeasurementswere recordedby mass spectrometry during rosette profiling.These mass spectrometers have previously beenused for analyzing naturally occurring oil seepsoff the coast of California and the Gulf ofMexico(11, 12), tracking subsurface oil leaks from blow-out preventers damaged by hurricanes in the GulfofMexico (13), andmapping deep-ocean hydratesin real time (10).

Mass spectrometric and fluorescence data,recorded during vertical profiling with the ship’ssampling rosette approximately 4 km from theleak source, confirmed the presence of a largeplume at ~1000- to 1200-mdepth, aswell as amorediffuse plume existing between 50- and 500-mdepth (Fig. 1). We operationally define a plume asa discrete spatial interval with hydrocarbon signalsor signal surrogates (i.e., colored, dissolvedorganic matter or aromatic hydrocarbon fluores-cence) more than two standard deviations abovethe root mean square baseline variability.

Mass spectra indicate a heterogeneous hydro-carbon mixture changing in composition as afunction of depth (Fig. 2); for example, ion peaks

1Applied Ocean Physics and Engineering Department, WoodsHole Oceanographic Institution, Woods Hole, MA 02543, USA.2Department of Marine Chemistry and Geochemistry, WoodsHole Oceanographic Institution, Woods Hole, MA 02543, USA.3Australian Centre for Field Robotics, University of Sydney,Sydney, New South Wales 2006, Australia. 4Monitor Instru-ments Company, Cheswick, PA 15024, USA.

*To whom correspondence should be addressed. E-mail:[email protected]

www.sciencemag.org SCIENCE VOL 330 8 OCTOBER 2010 201

REPORTS

on

Oct

ober

7, 2

010

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

www.sciencemag.org/cgi/content/full/330/6001/199/DC1

Supporting Online Material for

Nanophase Transition Metal Oxides Show Large Thermodynamically Driven Shifts in Oxidation-Reduction Equilibria

Alexandra Navrotsky,* Chengcheng Ma, Kristina Lilova, Nancy Birkner

*To whom correspondence should be addressed. E-mail: [email protected]

Published 8 October 2010, Science 330, 199 (2010)

DOI: 10.1126/science.1195875

This PDF file includes:

Materials and Methods

Figs. S1 and S2

Tables S1 to S3

References

Supporting Online Material for:

Nanophase Transition Metal Oxides Show Large Thermodynamically Driven Shifts

in Oxidation-Reduction Equilibria

Alexandra Navrotsky*, Chengcheng Ma, Kristina Lilova, and Nancy Birkner Peter A. Rock Thermochemistry Laboratory and NEAT ORU, University of California at Davis, Davis CA 95616, USA. *To whom correspondence should be addressed. Email: [email protected]

This WORD file includes: Materials and Methods Figs. S1, S2 Tables. S1, S2, S3 References

1

Materials and Methods

1.1 Sample preparation and characterization.

Nanocrystalline Co3O4 was prepared by calcination of cobalt hydroxide carbonate (S1).

CoO data are reported from our previous study (S2). NiO, Fe3O4, Mn2O3 and Mn3O4

samples were newly synthesized. All samples showed X-ray diffraction patterns

appropriate to a single phase, with peak broadening consistent with the small particle

size.

Co3O4 nanoparticle synthesis. 10.00 g of cobalt hexahydrate was mixed with 5.43 g of

ammonium bicarbonate (molar ratio 1:2), and ground for one hour. After repeated

washing with deionized water, the precursor was dried at 323 K overnight and then

calcined in air for half an hour. Various holding temperatures (623 K, 673 K, 723 K, 823

K, and 973 K) were applied to give a wide range of particle sizes.

Characterization. All phases were confirmed by powder X-Ray diffraction using a

Bruker diffractometer (Cu Ka radiation) operated at 45 kV and 40 mA. The XRD patterns

were collected with a 0.02o step size and 10 seconds dwell time, and analyzed by Jade

software (version 6.11, 2002; Materials Data Inc., Livermore, CA) to evaluate the phase

and size of the nanoparticles. Specific surface areas were measured by N2 adsorption at

77 K using a five-point Brunauer–Emmett–Teller (BET) technique on the analysis port of

a Micromeritics ASAP 2020 in the P / P0 range 0.05 to 0.3. Prior to analysis, the samples

were made into 20 mg pellets and degassed at 473 K for 3 h. After synthesis the sample

2

was placed in a small open vial and transferred into the calorimetry lab and stored for 1

month under constant temperature (298 K) and humidity (50%) to equilibrate the

hydration state. H2O content was determined from the weight difference before and after

annealing the sample at 1123 K for 3 hours. A microbalance was used to maximize the

weight loss measurement accuracy (to the microgram level).

1.2 Calorimetry

High temperature oxide melt solution calorimetry and water adsorption calorimetry were

performed using techniques described previously (S3, S4). Very briefly, the experiments

and data analysis were as follows. Hydrated nanoparticles were reacted in a molten oxide

solvent at 973 K, dissolving the oxide and liberating water vapor. The contribution of the

H2O was subtracted from the experimental enthalpy, and the slope of the line relating the

corrected heat of solution to surface area provides the surface enthalpy. When the

enthalpy associated with the H2O is taken as that of pure water for the process: H2O

(liquid, 298 K) => H2O (gas, 973 K), the calculated surface enthalpy represents the

enthalpy of the hydrated surface. When the enthalpy of H2O includes the exothermic

adsorption enthalpy as well, the calculated enthalpy is that of the anhydrous surface (S4).

We note that the surface enthalpy is essentially the same as the surface energy and is a

good representation of the surface free energy because both the volume change and

entropy change on forming nanoparticles from bulk material is small (S5).

Oxide melt solution calorimetry. For Co3O4, Fe3O4, Mn3O4, and Mn2O3, sample pellets

(~5 mg) were dropped into sodium molybdate (3Na2O4-MoO3) melt (20 g) at 973 K with

3

oxygen flushing through the calorimeter at 30 mL / min and also bubbling though the

solvent at 5 mL / min. NiO sample pellets (~5 mg) were dropped into lead borate (2PbO-

B2O3) melt (30 g) at 973 K with air flushing through the calorimeter at 30 mL / min and

also bubbling though the solvent at 5 mL / min. Flushing and bubbling maintains

oxidizing conditions, removes the evolving moisture, and aids dissolution. The custom

built Calvet twin calorimeter and techniques have been described previously (S3).

Water adsorption calorimetry. H2O content of metal oxide nanoparticles can be as high

as 10 wt % and its contribution to the drop solution enthalpy was measured by water

adsorption calorimetry at room temperature using a Calvet microcalorimeter, coupled

with a Micromeritics ASAP 2020 analysis system as described previously (S4). Sample

pellets were placed in one side of a forked silica gas tube and degassed under a static

vacuum (<10-3 Pa) at elevated temperatures to remove most of the water without

coarsening the sample. After BET measurement of the surface area of the sample and the

free space of the tube, the system was re-evacuated. Then, a series of precisely controlled

small doses of gaseous water were released into the system at room temperature until P /

P0 reached 0.35. The adsorption heat of each dose generated an exothermic calorimetric

peak. The simultaneous record of the amount of adsorbed water and the adsorption

enthalpy provided a higher resolution measurement of differential heat of adsorption as a

function of surface coverage.

2.1 Surface Enthalpy Derivation:

Co3O4: Thermochemical Cycle for Water Correction

4

(1) Co3O4 (nano, 298 K) Co3O4 (dissolved in 3Na2O4-MoO3, 973 K)

(2) Co3O4·nH2O (nano, 298 K) Co3O4 (dissolved in 3Na2O4-MoO3, 973 K) + nH2O

(gas, 973 K)

(3) H2O (gas, 973 K) H2O (gas, 298 K)

(4) Co3O4 (nano, 298 K) + nH2O (gas, 298 K) Co3O4·nH2O (nano, 298 K)

(1) = (2) + n(3) + (4)

∆H1 = ∆H2 + n∆H3 + ∆H4

∆H2 = ∆Hdrop solution

∆H3 = -25.0 ± 0.1 kJ / mol

∆H4 (hydrated surface) = ∆Hwater condensation (298K) = -44.0 ± 0.1 kJ / mol

∆H4 (anhydrous surface) = ∆Hwater adsorption (298K)

∆H1 (hydrated surface) = ∆Hdrop solution + n*(-25.0 ± 0.1) + n*(-44.0 ± 0.1)

∆H1 (anhydrous surface) = ∆Hdrop solution + n*(-25.0 ± 0.1) + n*∆Hwater adsorption

∆H1 represents the enthalpy of oxide melt drop solution for pure Co3O4 nanoparticles.

∆H2 is the original drop solution enthalpy for hydrated Co3O4•nH2O nanoparticles. ∆H4 is

the water adsorption enthalpy we measure. For calculating the enthalpy of hydrated

surfaces, water adsorption enthalpy is simply -44.0 kJ / mol, which is the heat of water

vapor condensation at 298.15 K. For calculating the energy of the anhydrous surface,

adsorption experiments give an integral heat of adsorption of -93.1 kJ / mol for chemi-

5

absorbed water. Analogous cycles were used for NiO, Fe3O4, Mn2O3, and Mn3O4 data,

for which the drop solution enthalpy in molten oxide melt had been measured.

CoO: Thermochemical Cycle for Water Correction

(1) CoO (nano, 298 K) CoO (dissolved in 5 N HCl, 298 K)

(2) CoO• nH2O (nano, 298 K) CoO (dissolved in 5 N HCl, 298 K) + nH2O (liquid, 298

K)

(3) H2O (liquid, 298 K) H2O (gas, 298 K)

(4) CoO (nano, 298 K) + nH2O (gas, 298 K) CoO•nH2O (nano, 298 K)

(1) = (2) + n(3) + (4)

∆H1 = ∆H2 + n∆H3 + ∆H4

∆H2 = ∆Hacid solution

∆H3 = 44.0 ± 0.1 kJ / mol

∆H4 (hydrated surface) = ∆Hwater condensation (298 K) = -44.0 ± 0.1 kJ / mol

∆H4 (anhydrous surface) = ∆Hwater adsorption (298 K)

∆H1 (hydrated surface) = ∆Hacid solution + n*(44.0 ± 0.1) + n*(-44.0 ± 0.1) = ∆Hacid solution

∆H1 (anhydrous surface) = ∆Hacid solution + n*(44.0 ± 0.1) + n*∆Hwater adsorption

∆H1 represents the enthalpy of acid solution for pure CoO nanoparticles. ∆H2 is the acid

solution enthalpy for hydrated CoO•nH2O nanoparticles. ∆H4 is the water adsorption

enthalpy we measure. For hydrated surface, ∆H1 = ∆H2 = ∆Hacid solution, no water

6

correction is needed, while as for anhydrous surface, enthalpy corrections are made by

water adsorption calorimetry measurements.

2.2 Phase Diagram Calculations

Co - O phase diagrams

Reactions:

I. Co / CoO: Co + 0.5 O2 CoO,

∆Gr (I) = ∆Gf(CoO),

∆Gr (I) = -R * T * lnK = 0.5RT * ln(PI(O2) / P0) = 0.5 * 2.303RT * log(PI(O2) / P0),

log (PI(O2) / P0) = ∆Gr(I) / (0.5 * 2.303RT).

II. CoO / Co3O4: 3CoO + 0.5 O2 Co3O4,

∆Gr(II) = ∆Gf(Co3O4) - 3 * ∆Gf(CoO),

∆Gr(II) = -R * T * lnK = 0.5RT * ln(PII(O2) / P0) = 0.5 * 2.303RT * log(PII(O2) / P0),

log (PII(O2) / P0) = ∆Gr(II) / (0.5 * 2.303RT).

Calculations of shifts in free energies for nanoparticles:

III. Co (nano) + 0.5 O2 CoO(nano),

∆Gr (III) = ∆Gf (CoO) + ∆Hsurface (CoO) - ∆Hsurface (Co),

∆Gr (III) = -R * T * lnK=0.5RT * ln(PIII(O2) / P0) = 0.5 * 2.303RT * log(PIII(O2) / P0),

log (PIII(O2) / P0) = ∆Gr(III) / (0.5 * 2.303RT).

7

In each case ∆Hsurface is the total enthalpy of the surface, that is, the surface energy times

the surface area per mole of material corresponding to the given particle size. Details of

surface energy calculations are found in Table S2.

IV. 3CoO(nano) + 0.5 O2 Co3O4(nano),

∆Gr (IV) = ∆Gf (Co3O4) - 3 * ∆Gf (CoO) + ∆Hsurface (Co3O4) - 3∆Hsurface (CoO),

∆Gr (IV) = -R * T * lnK = 0.5RT * ln(PIV(O2) / P0) = 0.5 * 2.303RT * log(PIV(O2) / P0),

log (PIV(O2) / P0) = ∆Gr(IV) / (0.5 * 2.303RT).

The log P / T curves were plotted by Origin Software, OriginLab Corporation, Version 8,

2009. All standard Gibbs free energies of formation are taken from Thermodynamic

Properties of Elements and Oxides, L.B. Pankratz, 1982 (S6) and listed in Table S3.

More recent tabulations have essentially identical data.

Fe-O phase diagrams

Similarly the Fe-O phase diagram can be calculated by using the data from reference S3.

The low temperature limit of the wustite nanoparticle stability field is calculated using

the following reaction:

(a). 4Fe0.947O(nano) Fe3O4 (nano) + 0.788Fe (nano)

∆Gr (a) = ∆Gf (Fe3O4) - 4∆Gf (Fe0.947O) + ∆Hsurface (Fe3O4) + 0.788∆Hsurface (Fe) -

4∆Hsurface (Fe0.947O)

At T = 1000 K, ∆Gr (a) = 0 when all particles have a diameter of 100 nm. This implies

that nanowustite of 100 nm diameter would be unstable relative to nanomagnetite and

8

iron of the same size below 1000 K. At T = 1650 K where bulk wustite would melt, ∆Gr

(a) = 0 for a diameter of 16 nm, implying no stability field for wustite for particles below

this diameter. These calculations are illustrative of the large effect of surface energies on

the wustite – iron - magnetite equilibrium.

Details of surface energy calculations are found in Table S2.

9

0 2 4 6 8 10 12 14 16 18 20 22

-300-300

-250

-200-200

-150

-100-100

-50

00

ΔH

ads

(kJ

/ mol

)

Coverage (molecules of H2O per nm2)

differential enthalpy integral enthalpy

ΔΗbulk water condensation A

0 2 4 6 8 10 12 14 16 18 20 22

-300-300

-250

-200-200

-150

-100-100

-50

00ΔΗbulk water condensation

ΔH

ads

(kJ

/ mol

)

Coverage (molecules of H2O per nm2)

differential enthalpy integral enthalpy

B

Fig. S1. Differential and integral heats of water adsorption ∆Hads on Co3O4 (A) and CoO

(B) as a function of surface coverage.

10

Fig. S2. XRD patterns for (A) starting material of 13 nm Mn2O3 before annealing in air

and (B) a product of 98% Mn2O3 and 2% Mn3O4 after heating in air at 975 K for 4 hours.

(major Mn3O4 peaks are marked with * )

11

Sample / XRD size

(nm)

BET surface

area (m2/mol)

BET size (nm)

Measured enthalpy for oxide•nH2O

*†

(kJ / mol)

Enthalpy correction of nH2O

Corrected enthalpy of pure oxide

Using liquid

water as reference

state (kJ / mol)

Using water

adsorption calorimetry

data (kJ / mol)

For hydrated surface

(kJ / mol)

For anhydrous

surface‡ (kJ / mol)

Co3O4•1.450H2O (7 nm)

37985.4 (6 nm) 316.30±1.51 100.07 139.58 216.23±2.05 176.72±2.46

Co3O4•1.197H2O (9 nm)

29695.9 (8 nm) 306.46±1.40 82.59 113.47 223.87±1.68 192.99±2.06

Co3O4•0.761H2O (12 nm)

21552.6 (11 nm) 284.85±1.50 52.50 74.91 232.35±1.65 209.94±1.92

Co3O4•0.546H2O (16 nm)

14673.6 (16 nm) 274.92±1.54 37.65 52.91 237.27±1.68 222.01±1.89

Co3O4•0.137H2O (52 nm)

4445.9 (53 nm) 257.33±1.45 9.48 14.11 247.85±1.48 243.22±1.55

Co3O4•0.035H2O (>1000 nm)

491.9 (>1000) 253.16±0.86 2.42 2.93 250.74±0.87 250.23±0.88

CoO1.043•0.094H2O (7 nm)

8920.7 (8 nm) -112.27±2.80 20.89 28.47 -133.16±3.32 -140.74±3.93

CoO1.047•0.061H2O (13 nm)

5745.1 (12 nm) -104.01±2.53 20.09 24.95 -124.1±3.02 -128.96±3.48

CoO1.014•0.019H2O (17 nm)

4451.7 (16 nm) -111.18±3.08 5.82 8.01 -117.0±3.25 -119.19±3.77

CoO1.010•0.008H2O (21 nm)

3061.7 (23 nm) -108.87±2.74 3.97 5.98 -112.84±2.81 -114.85±3.33

CoO (>1000 nm) NA -106.84±0.12 0 0 -106.84±0.12 -106.84±0.12 NiO•0.019H2O (43 nm)

1562 (XRD) 70.09±0.76 1.35 2.93 68.74±0.76 67.16±0.91

NiO (>1000 nm) NA 72.51±0.56 0 0 72.51±0.56 72.51±0.56 Fe2.80O4•0.70H2O (15 nm)

13685 (18 nm)

111.19 ± 1.54 48.30 57.20 62.89 ± 1.54 53.99±2.01

Fe2.80O4 (>1000 nm) NA 73.64±3.52 0 0 73.64±3.52 73.64±3.52 Mn3O4•0.606H2O (38 nm)

4431.2 (52 nm)

186.22 ± 1.13 41.81 NA 144.41± 1.25 NA

Mn3O4•0.458H2O (>1000 nm) NA 178.28±0.86 31.60 NA 146.68±0.91 NA Mn2O3•0.089H2O (30 nm)

5741.1 (33 nm)

150.27 ± 0.53 6.11 NA 144.16 ± 0.54 NA

Mn2O3•0.085H2O (49 nm)

3501.7 (51 nm) 152.25 ± 0.76 5.87 NA 146.38 ± 0.77 NA

Mn2O3 (>1000 nm) NA 154.34±1.13 0 0 154.34±1.13 NA

Table S1. Oxide melt solution calorimetric data and correction for H2O.

12

*In figure 1, both enthalpy and surface area of metal oxide (MO, M2O3, M3O4) are

divided by 1, 2, 3, respectively, for comparison on one mole metal basis (MO, MO1.5,

MO1.33).

†Enthalpy of CoO•nH2O is from acid solution calorimetry (S2).

‡Enthalpy of anhydrous CoO surface is obtained after water adsorption corrections

measured in this study.

Sample

Density (g / cm3, 298K)

Molar mass (g / mol)

Particle diameter

(nm)

Surface area * (m2 / mol)

Free energy shift arising from surface energy term†

(kJ / mol)

Free energy shift arising from surface energy term†

(kJ / mol) Co 8.90 58.93 10.00 3972.8 8.82 6.59 CoO 6.44 74.93 10.00 6981.1 24.92 19.69 Co3O4 6.11 240.80 10.00 23646.5 46.35 21.75 Fe 7.87 55.85 10.00 4255.4 9.02 6.77 Fe0.947O 5.75 68.89 10.00 7194.3 25.68 20.29 Fe3O4 5.17 231.53 10.00 26870.4 38.69 21.23 α-Fe2O3 5.24 159.69 10.00 18285.1 34.74 13.71 γ-Fe2O3 5.23 159.69 10.00 18320.1 13.00 10.44 Table S2. Calculation of energy shifts for phase equilibrium.

* Molar surface area is given by: molar mass / density / spherical volume * spherical

surface area

† Total free energy shift arising from surface energy term (kJ / mol) is given by: molar

surface area (m2 / mol) * surface energy (J / m2) / 1000.

13

Sample Temperature range A B C D E

CoO 298.15 – 700 K -56.861 -2.201 0.751 - 89.15 32.313 CoO 700 – 1394 K -63.393 -10.687 4.389 704.05 93.072 Co3O4 298.15 – 700 K -221.903 -2.001 -1.552 233.8 110.378 Co3O4 700 – 1394 K -241.499 -27.459 9.362 2613.4 292.653 Fe0.947O 298.15 – 1043 K -66.616 - 8.417 5.083 136.895 71.822 Fe0.947O 1043 – 1185 K -44.304 33.233 - 14.531 28.7 - 218.476Fe0.947O 1185 – 1652 K -66.666 - 2.995 0.329 85.047 39.147 Fe3O4 298.15 – 850 K -271.279 - 5.266 - 7.717 256.9 125.401 Fe3O4 850 – 1043 K -228.046 1.882 10.553 - 9925 24.887 Fe3O4 1043 – 1185 K -157.364 133.825 - 51.580 - 10267.75 - 894.747Fe3O4 1185 – 1652 K -228.205 19.06 - 4.507 - 10089.25 - 78.624 α-Fe2O3 298.15 – 960 K -203.228 - 13.067 3.263 390.95 155.628 α-Fe2O3 960 – 1043 K -204.088 - 23.637 12.897 194.600 220.071 α-Fe2O3 1043 – 1185 K -156.966 64.325 - 28.525 - 33.9 - 393.018α-Fe2O3 1185 – 1652 K -204.194 - 12.185 2.857 85.1 151.063

Table S3. Calculation of free energies (kJ / mol) at temperature T (K): ∆Gf (T) = ( A + B

/ 1000 * T * ln(T) + C / 1000000 * T * T + D / T + E / 1000 * T ) / 0.23901, from

reference S6.

References S1. S. Liu, J. Boerio-Goates, B. F. Woodfield, “Preparation of Uniform Nanoparticles of

Ultra-High Purity Metal Oxides, Mixed Metal Oxides, Metals, and Metal Alloys”,

Provisional Patent Application, U.S.A., (2006).

S2. L. Wang, et al., Chem. Mater. 16, 5394 (2004).

S3. A. Navrotsky, Phys. Chem. Miner. 24, 222 (1997).

S4. S. V. Ushakov, A. Navrotsky, Appl. Phys. Lett. 87, 164103 (2005).

S5. J. Boerio-Goates, et al., Nano. Lett. 6, 750 (2006).

14

15

S6. L.B. Pankratz, Thermodynamic Properties of Elements and Oxides, (U. S. Bureau of

Mines Bulletin 672, 1982) pp.115-116, pp.155-161.