脱合金过程中纳米多孔的演变Evolution of Nanoporosity in Dealloying

14.Schuèring,H.,Stannarius,R.,Tolksdorf,C.&Zentel,R.Liquid crystal elastomer balloons.Macro-

molecules(submitted).

15.Mach,P.,Huang,C.C.&Nguyen,H.T.Dramatic effect of single-atom replacement on the surface

tension of liquid-crystal compounds.Phys.Rev.Lett.80,732±735(1998).

16.Johnson,P.M.,Pankratz,S.,Mach,P.,Nguyen,H.T.&Huang,C.C.Optical reˉectivity and

ellipsometry studies of the Sm-C a*phase.Phys.Rev.Lett.83,4073±4076(1999).

17.Zhao,J.,Zhang,Q.M.,Kim,N.&Shrout,T.Electromechanical properties of relaxor ferroelectric

lead magnesium niobate-lead magnesium titanate ceramics.Jpn J.Appl.Phys34,5658±5663 (1995).

18.Poths,H.,Anderson,G.,Skarp,K.&Zentel,R.Fast electroclinic switching in a ferroelectric LC

Polysiloxane.Adv.Mater.4,792±794(1992).

19.Kremer,F.Electromechanical and/or mechanoelectrical converter.German Patent No.19636909

(1998).

20.Semmler,K.&Finkelmann,H.Mechanical?eld orientation of chiral smectic C-polymer networks.

Macromol.Chem.Phys.196,3197±3205(1995).

21.Lehmann,W.et al.The electromechanical effect in mechanically oriented S C*-elastomers

examined by means of an ultra-stable Michelson interferometer.Ferroelectrics208±209,373±383 (1998).

22.Lehmann,W.et al.Direct and inverse electromechanical effect in ferroelectric liquid crystalline

elastomers.J.Appl.Phys.86,1647±1652(1999).

23.Kremer,F.et al.Piezoelectricity in ferroelectric liquid crystalline elastomers.Polym.Adv.Technol.9,

672±677(1998).

Supplementary information is available on Nature's World-Wide Web site

(https://www.wendangku.net/doc/ed3537766.html,)or as paper copy from the London editorial of?ce of Nature. Acknowledgements

We thank R.Stannarius and D.Neher for discussions,and the``Innovationskolleg Phaènomene an den Miniaturisierungsgrenzen''at the University of Leipzig for support. Correspondence and requests for materials should be addressed to F.K.

(e-mail:kremer@physik.uni-leipzig.de). ................................................................. Evolution of nanoporosity

in dealloying

Jonah Erlebacher*2,Michael J.Aziz*,Alain Karma3,Nikolay Dimitrov§&Karl Sieradzki§

*Division of Engineering and Applied Sciences,Harvard University,

9Oxford Street,Cambridge,Massachusetts02138,USA

3Department of Physics and Center for Interdisciplinary Research on Complex Systems,Northeastern University,360Huntington Avenue,Boston, Massachusetts02115,USA

§Department of Mechanical and Aerospace Engineering and Center for Solid State Sciences,Arizona State University,Tempe,Arizona85287-6106,USA .............................................................................................................................................. Dealloying is a common corrosion process during which an alloy is`parted'by the selective dissolution of the most electro-chemically active of its elements.This process results in the formation of a nanoporous sponge composed almost entirely of the more noble alloy constituents1.Although considerable atten-tion has been devoted to the morphological aspects of the deal-loying process,its underlying physical mechanism has remained unclear2.Here we propose a continuum model that is fully consistent with experiments and theoretical simulations of alloy dissolution,and demonstrate that nanoporosity in metals is due to an intrinsic dynamical pattern formation process.That is, pores form because the more noble atoms are chemically driven to aggregate into two-dimensional clusters by a phase separation process(spinodal decomposition)at the solid±electrolyte inter-face,and the surface area continuously increases owing to etching. Together,these processes evolve porosity with a characteristic length scale predicted by our continuum model.We expect that chemically tailored nanoporous gold made by dealloying Ag-Au

2Present address:Department of Materials Science and Engineering,Johns Hopkins University, Baltimore,Maryland21218,USA.should be suitable for sensor applications,particularly in a biomaterials context.

Selective dissolution has a long history3.For example,the chemical treatment known as depletion gilding selectively dissolves a non-gold element near the surface of a less-expensive alloy such as Au-Cu,leaving behind a surface of pure gold.Early Andean gold-smiths used this technique to enhance the surfaces of their artefacts4. In this century,selective dissolution has been primarily examined in the context of corrosion.It is observed in technologically important alloy systems,notably brasses,stainless steels,and Cu-Al alloys1,5,6. The mechanical properties of a porous overlayer are very different from the bulk alloy on which it sits,leading to brittle crack propagation,stress corrosion cracking,and other undesirable materials failure7.Figure1shows the prototypical dealloyed micro-structure,that of nanoporous gold.Early notions considered porosity as a hidden microstructure revealed by etching,but diffraction experiments showed that no pre-existing length scale exists before acid attack on single-phase alloys8,https://www.wendangku.net/doc/ed3537766.html,ter ideas considered the inˉuence of percolating clusters within the solid solution of the alloy,but models failed to yield behaviour consistent with experiment10,11.

The following argument illustrates the fundamental obstacle to understanding porosity formation during dealloying:consider a silver-gold alloy in an electrolyte under conditions where silver dissolves and gold is inert.Initially,silver will be dissolved from surface sites such as terraces or steps.Gold atoms should accumu-late on the surface and locally block further dissolution.For an alloy containing10%gold,it might be expected that dissolution would stop or be signi?cantly retarded after about10monolayers of the alloy have been dissolved.

A complete model of selective dissolution needs to be multi-scale, involving the kinetics of dissolution,surface diffusion,and mass transport through the bulk of both alloy and electrolyte.Because

b

120 nm

Figure1Scanning electron micrographs of nanoporous gold made by selective dissolution of silver from Ag-Au alloys immersed in nitric acid under free corrosion conditions.a,Cross-section of dealloyed Au32%Ag68%(atom%)thin?lm.b,Plan view of dealloyed Au26%Ag74%(atom%).The porosity is open,and the ligament spacings shown in b are of the order of10nm;spacings as small as5nm have been observed. Measurements of the surface area of nanoporous gold are of the order of order2m2g-1 (refs24,25),comparable to commercial supported catalysts.

mass transport through the bulk of the growing phase (the electro-lyte)is always a stabilizing inˉuence 12,and because mass transport through the bulk of the dissolving phase appears too slow to be signi?cant,we hypothesized that the morphology-determining physical process is con?ned to the interface region between the alloy and the electrolyte.To test this,we developed a kinetic Monte Carlo model to simulate Ag-Au dealloying,including only diffusion of silver and gold and dissolution of silver 13.We found that this model was able to reproduce all relevant experimental trends characteristic of dealloying,both morphological and kinetic.

Figure 2shows a simulated porous structure with ligament widths of 2±5nm.Our simulations were successful in modelling the nanoporous morphology,and also in modelling the dynamic behaviour of the dissolution current versus overpotential.It is a well characterized feature of alloy dissolution that as the over-potential is increased (usually at rates of the order of a few millivolts per second),the dissolution current of ions from the alloy stays at a low level until a bulk-composition-dependent critical potential (V C )is reached,at which point this current rises rapidly 14.Figure 3shows simulated and experimental polarization curves for different alloy compositions.There is clear observation of a composition-dependent V C .To our knowledge,this is the ?rst simulation model to produce such behaviour,suggesting that we have found a minimum set of physical processes to include in any model for

alloy dissolution.

The simulations reveal the following qualitative picture of porosity formation.The process starts with the dissolution of a single silver atom on a ˉat alloy surface of close-packed (111)orientation,leaving behind a terrace vacancy.The atoms coordinat-ing this vacancy have fewer lateral near-neighbours than other silver atoms in the terrace,and are thus more susceptible to dissolution.As a result,the entire terrace is `stripped',leaving behind gold atoms with no lateral coordination (`adatoms').Before the next layer is attacked,these gold adatoms diffuse about and start to agglomerate into islands.As a result,rather than a uniform diffuse layer of gold spread over the surface,the surface is comprised of two distinct kinds of regionsDpure gold clusters that locally passivate the surface,and patches of un-dealloyed material exposed to electrolyte.When silver atoms in these patches dissolve,more gold adatoms are released onto the surface.These adatoms diffuse to the gold clusters left over from dissolution of previous layers,continuing to leave un-dealloyed material exposed to electrolyte.In the early stages,these gold clusters are mounds that are gold-rich at their peaks and that have alloy composition at their bases.These mounds get undercut,increasing the surface area that gold must cover to bring about passivation.Ultimately,this leads to pit formation and porosity.Central to this description is the the coalescence of gold adatoms into stable clusters.The spacing between these `islands'in the initial stages of dissolution is close to the spacing between ligaments in the ?nal porous structure.The physical reason for this coalescence can be understood by considering the gold adatoms to be one compo-nent of a two-component solution of gold and `electrolyte'con?ned to the monolayer-thick interfacial layer sitting on top of un-dealloyed material.We modelled the thermodynamics of the inter-facial layer as a regular solution 15,and found the solubility of gold in electrolyte within the interfacial layer to be of the order of 10-7per site (see Methods).This solubility may be interpreted as the `equilibrium concentration of gold adatoms'on the surface of the alloyDin the absence of etching,it represents a dynamic equilib-rium of adatoms resulting from their two-dimensional evaporation from step edges onto terraces and their subsequent recondensation.In contrast to the equilibrium condition,rapidly stripping a terrace of silver atoms leaves gold adatoms with a local site occupancy fraction equal to that in the bulk,typically 10±40%Dfar above their equilibrated concentration of 10-7per site.Thus,there is an extremely strong driving force for gold adatoms to condense onto nearby gold-rich clusters.In fact,regions of surface with high enough supersaturation of gold adatoms sit `within the spinodal',a special segment of the curve representing free energy f of a spatially uniform layer versus gold concentration C for which ]2f /]C 2,0.Within the spinodal,composition ˉuctuations of in?nitesimal amplitude lead to a lower overall free energy for

the

Figure 2Simulated nanoporous gold.The simulation model was as follows:a bond-breaking model was used for diffusion;atoms with N near neighbours diffused with rate k N =n D exp(-N e /k B T ),where e is a bond energy and n D =1013s -1.Dissolution rates were consistent with the Butler±Volmer (BV)equation in the high-driving-force Tafel regime;the dissolution rate k E,N for a silver atom with N near neighbours was written as k E,N =n E exp[-(N e -f )/k B T ],where n E =104s -1is an attempt frequency determined by the exchange-current density in the BV equation and f is the overpotential.For the ?gure,f =1.75eV,e /k B T =5.51.

0400800

1,2001,6002,000

0.00

0.20

0.40

0.60

40

80120160

2001.25

1.50

1.75

2.00

2.25

2.50C u r r e n t d e n s i t y (μA c m –2)

50%

30%

25%

10%

5%0%

37%

35%33%

30%

28%

20%

10%

5%

E (V, Ag +/Ag)

φ (eV)a

b

Bulk gold fraction

C r i t i c a l p o t e n t i a l (V )

Figure 3Comparison of experimental and simulated current±potential behaviour.a ,Current±potential behaviour for varying Ag-Au alloy compositions (atom%Au)

dealloyed in 0.1M HClO 4+0.1M Ag +(reference electrode 0.1M Ag +/Ag).b ,Simulated

current±potential behaviour of Ag-Au alloys.c ,Comparison of experimental (line)and simulated (triangles)critical potentials;the zero of overpotential has been set equal to the onset of dissolution of pure silver both in simulation and in experiment.

system,and involve atomic diffusion against concentration gradi-ents (the `uphill diffusion'process through which gold condenses onto nearby clusters),that is,the system is inherently unstable and will spontaneously phase-separate.But ˉuctuations of long length scale grow slowly due to the required diffusion times,and short-length-scale ˉuctuations create much energetically unfavourable incipient interface between the phases,inhibiting their growth.Hence,phase separation is manifested most rapidly at an inter-mediate length scale that roughly corresponds to the spacing between the observed gold-rich clusters.This effect is known as spinodal decomposition 16,17.As porosity forms,the decomposition is occurring on a non-ˉat,non-uniform surface with continuously increasing surface area.

In our model,the motion of the alloy±electrolyte interface is fully described mathematically by the ˉux of diffusing adatoms J S ,the velocity of the interface normal to itself v n ,and the concentration accumulation rate ]C/]t ,all of which are interrelated and vary with position along the curve of the interface (for detailed derivations,see Methods).For J S ,we used a model for diffusion during spinodal decomposition known as the Cahn±Hilliard equation 17.The normal velocity depends on C and also on the local curvature k through capillary effects 11.The time evolution of C is uniquely determined by the local mass-conservation condition

]C =]t v n C 02v n k C 2=×J S

1

where C 0is the bulk gold concentration.This condition is analogous to the local conservation of heat or solute appearing in boundary-layer models of solidi?cation 18,with two important differences:(1)the interfacial layer thickness is constant along the interface,and is microscopic,rather than being a spatially varying macroscopic diffusion length,and (2)the surface aggregation process inherent in the Cahn±Hilliard form for J S is essential for porosity formation.Simple (`downhill')surface diffusion (J S =-D S =C )yields an

initially unstable interface that passivates quickly,before well-formed pores have a chance to develop.Cahn±Hilliard diffusion also dominates capillarity-driven surface diffusionDthe effect usually incorporated into interface evolution equations 19.

We performed numerical integration of equation (1)using a relative arc-length parametrization scheme 20,21,and parameters that matched those used in the kinetic Monte Carlo simulations.We observed,as expected,the evolution of gold clusters separated by a characteristic spacing l .An analytic expression for l can be found by a time-dependent linear stability analysis of equation (1)that takes into account the slow increase of gold concentration into the interfacial layer as the instability develops.This effect needs to be included because the spatial period with the largest ampli?cation rate depends sensitively on the gold concentration.Speci?cally,this spatial period decreases sharply as the concentration increases past a threshold concentration for instability that corresponds to the spinodal point ]2f /]C 2=0,and the interface is stable for concen-trations below this threshold.We ?nd a maximally unstable spatial period that scales as l ~(D S /V 0)1/6,where V 0is the velocity of a ˉat alloy surface with no gold accumulated upon it.This prediction is in qualitative agreement with both kinetic Monte Carlo simulations and experiments,both of which show that the characteristic length scale of porosity decreases with increasing driving force.A more elaborate analysis incorporating nonlinear effects,however,remains needed for a detailed quantitative comparison.

There is an analogy between this result,applicable to etching,and two-dimensional island nucleation during submonolayer vapour phase https://www.wendangku.net/doc/ed3537766.html,ly,in the early stages of etching,the dissolution process is analogous to deposition of gold;in both processes,adatoms are added to the surface where they are free to agglomerate into islands.The case of vapour phase deposition has been studied using rate equations that describe an aggregation process where adatoms stick together irreversibly.In these studies it is a well-known result that the island spacing scales as (D S /F )m ,where the deposition rate F is the direct analogue of the surface velocity in etching,and the exponent m depends on details of the aggregation process (see ref.22and references therein).That these results are limited to irreversible aggregation during deposition and our analysis is for reversible aggregation during etching suggests the existence of universal scaling laws for aggregation that do not depend on reversibility or the lack thereof in these two opposite processes.

Our kinetic Monte Carlo simulation elucidates the later stages of morphological evolution,and the mechanism by which three-dimensional porosity evolves.We highlight the features of this process by showing in Fig.4a simulation of an arti?cial pit in an otherwise fully passivated surface.When the pit reaches suf?cient depth,its surface area has increased suf?ciently that a new gold cluster nucleates.When this happens,the pore splits into multiple new pits,each with a smaller surface area than its parent.These `child'pits continue to penetrate into the bulk,increasing their surface area,nucleating new clusters,spawning new pits,and so on,until a full,three-dimensional porous structure evolves,such as those illustrated in Figs 1and 2.M

Methods

In a regular solution,the enthalpy of mixing depends on the bond energies and the entropy of mixing is ideal.The free energy of a regular solution f (C ,T )is written f a c 12c k B T c ln c 12c ln 12c ,where c is the mole fraction of gold (c C -2=3,where -is atomic volume),a =6[E Au±electrolyte -(1/2)(E Au±Au +E electrolyte±electrolyte )],where E x±x are the respective interaction energies between Au and electrolyte,the prefactor 6is the lateral coordination in the two-dimensional hexagonal lattice of the interfacial layer,k B is Boltzmann's constant and T is absolute temperature.For our simulation conditions,realistic timescales and length scales were obtained from the parameters

E Au±Au =-0.285eV (=-e ,the simulation bond energy as described in Fig.2),T =600K,E Au±electrolyte =E electrolyte±electrolyte =0.0eV.With these parameters,a =0.855eV.The free energy has the familiar double-well form 23and a minimum at c <10-7per site,representing the solubility of gold in electrolyte (and vice versa).

The Cahn±Hilliard diffusion equation is J S =-M (C )(]2f /]c 2)=C +2M C w =3C

.

a

b

c d

Figure 4Simulated evolution of an arti?cial pit in Au 10%Ag 90%(atom%),f =1.8eV.

Cross-sections along the (111

?)plane de?ned by the yellow line in a are shown below each plan view.a ,The initial condition is a surface fully passivated with gold except within a circular region (the `arti?cial pit').b ,After 1s,the pit has penetrated a few monolayers into the bulk.We note how there are fewer gold clusters near the side wall than at the centre of the pit.c ,After 10s,a gold cluster has nucleated in the centre of the pit.d ,At 100s,the pit has split into multiple pits;each will continue to propagate into the bulk to form a porous structure like that in Fig.2.

Here,453M(C)is a mobility,w is the so-called gradient energy coef?cient,and the gradients are taken with respect to arc length.The?rst term on the right-hand side describes the chemical effect leading to phase separation within the spinodal;the second term describes the effect that damps short-wavelengthˉuctuations.The mobility is proportional to the surface diffusivity D S and is given by M C =(D S/k B T)c(1-c).The mobility is peaked for c=0.5,and is zero for c=0and c=1(atoms do not diffuse in pure phases because there are no vacancies in our model).The normal velocity is given by v n C V C 12 g-=k B T k ,where g is the surface free energy and V(C)is called the interface response function,equal to the velocity of aˉat surface covered with a concentration C of gold.We?nd in both simulation and experiment that the interface response is?tted well by the functional form V(C)=V0(f)exp(-C/C*),where f is the overpotential and C*is a constant.Experimentally,the gold accumulation can be inferred by integrating the dissolution current versus time at?xed overpotential;it is necessary to use an overpotential that is low enough to ensure that the surface remains planar(that is, porosity does not form)and also to catch the short initial transient rise in current as silver atoms are pulled from the?rst few monolayers.This particular form for the interface response function is quite curious.Naively,one might expect that the local interface velocity would be proportional to the local concentration of silver exposed to the electrolyte,that is,V(C)~(1-c).However,the decaying exponential form suggests that there is an evolving distribution of holes opening and closing within the interfacial region, controlling the accumulation rate.

Physically,the mass conservation condition(equation(1))is the statement that the total number Cb D s of gold atoms in a length D s of interface with lateral width b can change as a result of three distinct effects that correspond to the three terms on the right-hand-side of equation(1):the accumulation of gold atoms into the interfacial layer from the solid being dissolved;the local stretching of the interface(]D s/]t=v n k D s),which can either increase or decrease C depending on whether the solid is concave (k.0)or convex(k,0);and the motion of atoms along the interface driven by the surface diffusionˉux J S.

Received14November2000;accepted10January2001.

1.Pickering,H.W.Characteristic features of alloy polarization curves.Corros.Sci.23,1107±1120

(1983).

2.Forty,A.J.Corrosion micromorphology of noble metal alloys and depletion gilding.Nature282,597±

598(1979).

3.Masing,G.Zur Theorie der Resistenzgrenzen in Mischkristallen.Z.Anorg.Allg.Chem.118,293±308

(1921).

4.Lechtman,H.Pre-Columbian surface metallurgy.Sci.Am.250,56±63(1984).

5.Williams,D.E.,Newman,R.C.,Song,Q.&Kelly,R.G.Passivity breakdown and pitting corrosion of

binary alloys.Nature350,216±219(1991).

6.Newman,R.C.&Sieradzki,K.Metallic Corrosion.Science263,1708±1709(1994).

7.Li,R.&Sieradzki,K.Ductile-brittle transition in random porous Au.Phys.Rev.Lett.68,1168±1171

(1992).

8.Corcoran,S.in Critical Factors in Localized Corrosion III(eds Kelly,R.G.,Frankel,G.S.,Natishan,

P.M.&Newman,R.C.)500±507(Electrochemical Society,Pennington,New Jersey,2000).

9.Pickering,H.W.&Wagner,C.Electrolytic dissolution of binary alloys containing a noble metal.

J.Electrochem.Soc.114,698±706(1967).

10.Sieradzki,K.,Corderman,R.R.,Shukla,K.&Newman,https://www.wendangku.net/doc/ed3537766.html,puter simulations of corrosion:

selective dissolution of binary alloys.Phil.Mag.A59,713±746(1989).

11.Sieradzki,K.Curvature effects in alloy dissolution.J.Electrochem.Soc.140,2868±2872(1993).

12.Wagner,C.Contribution to the theory of electropolishing.J.Electrochem.Soc.101,225±228

(1953).

13.Erlebacher,J.in Dynamics of Crystal Surfaces and Interfaces(eds Duxbury,P.&Pence,T.)24±28

(Plenum,New York,1997).

14.Wagner,K.,Brankovic,S.R.,Dmitrov,N.&Sieradzki,K.Dealloying below the critical potential.

J.Electrochem.Soc.144,3545±3555(1997).

15.Cahn,J.W.&Hilliard,J.E.Free energy of a nonuniform system.I.Interfacial free energy.J.Chem.

Phys.28,258±267(1958).

16.Cahn,J.W.&Hilliard,J.E.Free energy of a nonuniform system.III.Nucleation in a two-component

incompressibleˉuid.J.Chem.Phys.31,688±699(1958).

17.Hilliard,J.E.in Solidi?cation497±560(American Society for Metals,Metals Park,Ohio,1971).

18.Ben-Jacob,E.,Goldenfeld,N.,Langer,J.S.&Schon,G.Dynamics of interfacial pattern formation.

Phys.Rev.Lett.51,1930±1932(1983).

19.Mullins,W.W.Theory of thermal grooving.J.Appl.Phys.28,333±339(1957).

20.Kessler,D.A.,Koplik,J.&Levine,H.Geometrical models of interface evolution.II.Numerical

simulation Phys.Rev.A30,3161±3174(1984).

21.Hou,T.Y.,Lowengrub,J.S.&Shelley,M.J.Removing the stiffness from interfacialˉow with surface

https://www.wendangku.net/doc/ed3537766.html,p.Phys.114,312±338(1994).

22.Barabasi,A.-L.&Stanley,H.E.Fractal Concepts in Surface Growth(Cambridge Univ.Press,1995).

23.Haasen,P.Physical Metallurgy,(Cambridge Univ.Press,1986).

24.Forty,A.J.in Sir Charles Frank:An80th Birthday Tribute(ed.Chamber,R.G.)164±187(Adam Hilger,

Bristol,1991).

25.Tulimieri,D.J.,Yoon,J.&Chan,M.H.W.Ordering of helium mixtures in porous gold.Phys.Rev.Lett.

82,121±124(1999).

Acknowledgements

This work was supported by the US Department of Energy,Basic Energy Sciences. The research of A.K.also bene?ted from computer time allocation at NU-ASCC.

K.S.thanks the AFOSR for support.

Correspondence and requests for materials should be addressed to J.E

(e-mail:Jonah.Erlebacher@https://www.wendangku.net/doc/ed3537766.html,).................................................................. Ice shelves in the Pleistocene Arctic Ocean inferred from glaciogenic deep-sea bedforms

Leonid Polyak*,Margo H.Edwards2,Bernard J.Coakley3

&Martin Jakobsson§k

*Byrd Polar Research Center,Ohio State University,Columbus,Ohio43210,USA 2Hawaii Mapping Research Group,Hawaii Institute of Geophysics and Planetology,University of Hawaii,Honolulu,Hawaii96822,USA

3Department of Geology,Tulane University,New Orleans,Louisiana70118,USA §Department of Geology and Geochemistry,Stockholm University,

10691Stockholm,Sweden .............................................................................................................................................. It has been proposed that during Pleistocene glaciations,an ice cap of1kilometre or greater thickness covered the Arctic Ocean1±3.This notion contrasts with the prevailing view that the Arctic Ocean was covered only by perennial sea ice with scattered icebergs4±6.Detailed mapping of the oceanˉoor is the best means to resolve this issue.Although sea-ˉoor imagery has been used to reconstruct the glacial history of the Antarctic shelf7±9,little data have been collected in the Arctic Ocean because of operational constraints10,11.The use of a geophysical mapping system during the submarine SCICEX expedition in199912provided the oppor-tunity to perform such an investigation over a large portion of the Arctic Ocean.Here we analyse backscatter images and sub-bottom pro?ler records obtained during this expedition from depths as great as1kilometre.These records show multiple bedforms indicative of glacial scouring and moulding of seaˉoor,combined with large-scale erosion of submarine ridge crests.These distinct glaciogenic features demonstrate that immense,Antarctic-type ice shelves up to1kilometre thick and hundreds of kilometres long existed in the Arctic Ocean during Pleistocene glaciations. The central Arctic Ocean contains relatively shallow areas(water depths,1,000m;see Fig.1)on Yermak plateau,Lomonosov ridge and Chukchi borderlandDwhich includes Chukchi plateau,Chuk-chi rise and Northwind ridge.During the SCICEX-99expedition, conducted on the nuclear-powered submarine USS Hawkbill, shallow sea-ˉoor areas were targeted for mapping to detect glacio-genic bedforms.Sea-ˉoor images(collected using a submarine-mounted12-kHz swath bathymetry and sidescan sonar12)from the Chukchi borderland and the Lomonosov ridge show a variety of bedforms,including random or subparallel scours,parallel linea-tions,and transverse ridges.On the records from the chirp sub-bottom pro?ler,these bedforms are associated with planed ridge crests with rough microrelief and obvious angular unconformities cut into the strati?ed sediments.

Randomly oriented furrows,typically,100-m wide and up to 30-m deep,densely cover the shallowest,,400-m-deep portions of seaˉoor on the Chukchi borderland and adjacent continental margin(Fig.2a).Isolated larger scours up to700-m wide and over10-km long occur as deep as500m.Even greater depths, exceeding900m,are attained by closely spaced,subparallel scours on the Lomonosov ridge.Sea-ˉoor scours are known to be formed by the drift of icebergs and pack-ice ridges13.At present,icebergs in the Arctic Ocean have at most50-m draughts14,whereas icebergs off Antarctica and Greenland reach depths of500±550m(refs15,16). The largest depths of gouged seaˉoor,extending to850m,have been reported from the Yermak plateau10,matching the depth of sours on the Lomonosov ridge.

Below the depth range of dense scouring,the seaˉoor exhibits

k Present address:Center for Coastal Mapping,University of New Hampshire,Durham,New Hampshire 03824,USA.