NEWS & VIEWS

If continuous,rapid,essentially defect-free growthof a single (n,m) nanotube species can next be achievedalong a specific,chosen direction, this would go a longway towards reaching the goal of controlled nanotubegrowth over macroscopic length scales.The particular(n,m) nanotube grown by Zhu et al. appears to besemiconducting along its entire length (based on its G-band Raman spectral lineshape).One could imaginethat growth of a unique (n′,m′) metallic nanotubeseveral centimetres in length could likewise be achieved.The authors give us some clues about the growthconditions necessary for achieving ultra-long nanotubegrowth: an alcohol carbon source, fast heating of thesubstrate to prevent catalyst particle growth,and use ofa smooth substrate over which the growing nanotubecan glide,as well as controlled conditions oftemperature,gas pressure and catalyst type6.

Although the work of Zhu and colleagues is animportant advance,much further research anddevelopment will be needed before we can expect to see large-scale synthesis of specified (n,m) nanotubes.The next logical step would be to push forimprovements in the synthesis conditions to furtherincrease the nanotube lengths produced, to improve thereliability of the synthesis process in producing the same

(n,m) nanotube under the same growth conditions,and eventually to increase the control of the synthesisprocess to produce any desired (n,m) nanotube onrequest. In situ nanotube characterization capable ofrapidly assessing the chiral angles and nanotubediameters would be needed for quality control of large-scale synthesis.And the process would need to be scaledup by about 12 orders of magnitude to achieve large-scale (gram) quantities of nanotubes.With all of thesehurdles overcome,a variety of nanotube applicationsbased on the unique quantum properties of carbonnanotubes would at last become feasible .Their highmobilities could then be exploited for semiconductorelectronics,as could their sensitivity to specificadsorbates for biosensors,and their unique mechanicaland thermal conductivity properties could usher in ahost of nano-electromechanical devices and thermalmanagement applications.

References1. Iijima, S. & Ichihashi, T. Nature 363, 603–605 (1993).

2. Bethune, D. S. et al. Nature 363, 605–607 (1993).

3. Saito, R., Dresselhaus, G. & Dresselhaus, M. S. Physical Properties of Carbon

Nanotubes (Imperial College Press, London, 1998).

4. Zhu, Y. et al. Nature Mater. 3, 673–676 (2004).

5. Yakobson, B. I. & Smalley, R. E. Am. Scient. 85, 324–337 (1997).

6. Liu, J., Fan, S. & Dai, H. Mater. Res. Soc. Bull. 29, 244–250 (2004).

666 nature materials | VOL 3 | OCTOBER 2004 | www.nature.com/naturematerials

AMORPHOUS MATERIALSLSFinding order in disorder

TODD C. HUFNAGEL is at the Department ofMaterials Science and Engineering,John Hopkins University,Baltimore,Maryland 21218,USA.

e-mail: hufnagel@jhu.edu

A t first glance,the structures of amorphousmaterials seem rather uninteresting —particularly in comparison with crystalline

materials,which display a wide variety of structures.But advances in characterization techniques describedat a recent symposium* on Order in Disorder: Probing theStructure ofAmorphous Materials,show that thestructures of amorphous materials are much richer thanis commonly appreciated.

The shortest length scale usually used to describethe structure of a material consists of an atom and its

nearest neighbours,out to perhaps two or three atoms distant.All solids and liquids have some structure on this scale,which we call short-range order.For crystalline solids, structural order persists overmuch longer distances (at least tens or hundreds ofatomic distances), such that the atoms occupy sites in aperiodic three-dimensional array.Such materials aresaid to have long-range order and include most metalsand many covalently bonded solids.Non-crystallinesolids, including glasses, lack long-range order and aresaid to be ‘amorphous’(literally,‘without form’) eventhough they can have short-range order that is quitewell defined.

Most of our knowledge of these structures comesfrom the ways in which the atoms in the materialscatter radiation, particularly X-rays and electrons.

Describing the structure of amorphous materials such asmetallic glasses has been a longstanding problem in materialsscience. A new technique called fluctuation microscopy allowsus to see order on length scales that are difficult to study withtraditional scattering techniques.

*Microscopy andMicroanalysis 2004,1–5 August, 2004, Savannah,Georgia, USA.

nv_oct04 9/10/04 10:29 AM Page 666

© 2004 Nature Publishing Group

© 2004 Nature Publishing Group

NEWS & VIEWS

Long-range order gives rise to sharp diffraction peaks,which can be analysed to determine the structure withgreat precision. If there is short-range order but notlong-range order, the structural information that canbe obtained is more limited. Usually, this informationis in the form of a pair correlation function, whichdescribes the probability that a second atom will befound at a given distance from an average atom in the material.

Some amorphous materials,however,can haveorder over length scales longer than those associatedwith short-range order,but not so extensive as toconstitute long-range order.This ‘medium-range order’is difficult to detect using scattering techniques becauseit has a relatively small effect on the pair correlations.The effect of medium-range order on the triplet (threeatom) and pair–pair (four atom) correlations is muchmore significant,but these higher-order correlationscannot be directly discerned from scattering data.Fortunately, fluctuation electron microscopy (Fig. 1) isexquisitely sensitive to these higher-order correlations1.This technique detects the spatial variations in thescattered intensity caused by medium-range order (A.Howie,Cambridge University,UK).

The structure of amorphous semiconductorsillustrates the significance of medium-range order2–4.Scattering data from amorphous silicon andamorphous germanium are consistent with thecontinuous random network model,which iscommonly used to describe the structure of covalentlybonded amorphous solids.But fluctuation electronmicroscopy observations from as-deposited thin filmsreveal the presence of strong medium-range order2–4.To be consistent with both the scattering data and thefluctuation microscopy data,one must invoke thepresence of very small (<3 nm diameter) paracrystals ina disordered matrix.These paracrystals have crystallinetopology,but the atomic positions are highly distortedfrom those of a perfect crystal. In amorphous silicon,the degree of medium-range order increasescontinuously with substrate temperature duringdeposition, implying that there is no abrupt transitionfrom the nominally amorphous structure to one that isclearly polycrystalline5.

Results presented at the symposium and elsewheredemonstrate that medium-range order is alsoimportant in amorphous metallic alloys6,7.For instance,amorphous Al82Sm8 can be prepared by both quenchingfrom the liquid and by solid-state reaction,but the twomethods produce qualitatively different medium-rangeorder (W.G.Stratton,University of Wisconsin,USA).The structural differences correlate with differences incrystallization behaviour,possibly because themedium-range order influences nucleation of thecrystalline phase.

The observation of significant medium-range orderin metallic glasses is surprising because bonding inmetallic alloys is largely non-directional, implying thatthe short-range order is not as well defined as that ofcovalently bonded solids.Models for short-range orderin metallic glasses emphasize the importance of efficientpacking of atoms into clusters8,9.At the symposium(and in this issue10) a new structural model for metallicglasses was proposed in which icosahedral clusters ofatoms are arranged into larger domains that are

nominally ordered,but have significant internal strainsthat prevent the development of long-range order (D.B.Miracle,Air Force Research Laboratory,USA).The predictions of the model agree with pair correlationdata from a variety of amorphous alloys and provide abasis for understanding medium-range order inmetallic glasses.Most impressively, the model allowsprediction of new glass-forming alloy compositions.

Fluctuation microscopy can also be performedusing photons instead of electrons (J.M.Gibson,Argonne National Laboratory,USA).The length scale ofthe medium-range order probed is determined by thevolume of material illuminated,so soft X-rays(E ≈ 2 keV) are of the greatest current interest becausespot sizes of∼50 nm diameter can be produced.The wavelength of these photons (∼1 nm) is appropriate for studying polymers,biological materials,and self-assembled nanostructures.X-rays arecomplementary to electrons for fluctuation microscopybecause their greater penetrating power allowsexamination of thicker specimens,and specimens inaqueous or other environments not amenable toelectron microscopy.Furthermore,resonant scatteringeffects near X-ray absorption edges could be used formeasurements on specific chemical elements.Advances in the focusing of higher-energy X-rays willbroaden the range of applications.

Currently, the most significant challenge isinterpreting fluctuation microscopy data directly interms of real-space structures.Progress here, togetherwith continued advances in experimental techniques, iscertain to lead to new insights into the structure ofdisordered materials.

References1. Voyles, P. M., Gibson, J. M. & Treacy, M. M. J. J. Electron. Microsc. 49, 259–266

(2000).

2. Gibson, J. M. & Treacy, M. M. J. Phys. Rev. Lett. 78, 1074–1077 (1997).

3. Treacy, M. M. J., Gibson, J. M. & Keblinski, P. J. J. Non-Cryst. Solids 231, 99–110

(1998).

4. Voyles, P. M. & Abelson, J. R. Solar Eng. Mater. Solar Cells 78, 85–113 (2003).

5. Voyles, P. M., Gerbi, J. E., Treacy, M. M. J., Gibson, J. M. & Abelson, J. R. Phys.

Rev. Lett. 86, 5514–5517 (2001).

6. Li, J., Gu, X. & Hufnagel, T. C. Microsc. Microanal. 9, 509–515 (2003).

7. Stratton, W. G., Hamann, J., Perepezko, J. H. & Voyles, P. M. Mater. Res. Soc.

Symp. Proc. 806, 275–280 (2004).

8. Egami, T. Mater. Sci. Eng. A 226–228, 261–267 (1997).

9. Miracle, D. B., Sanders, W. S. & Senkov, O. N. Phil. Mag. A. 83, 2409–2428

(2003).

10. Miracle, D. B. Nature Mater. 3, 697–702 (2004).

nature materials | VOL 3 | OCTOBER 2004 | www.nature.com/naturematerials 667

Figure 1 Fluctuation electronmicroscopy.a, In a completelydisordered material, thescattered intensity from allregions will be approximately thesame,and the dark-field image(which is a map of scatteredintensity) will show little point-to-point variation.b, If there areregions of medium-range order,some regions will be orientedsuch that they scatter theincident electrons strongly (right),while other regions are not (left).In this case there will besignificant point-to-pointvariation in the intensity of thedark-field image. (Adapted fromref.1,copyright © JapaneseSociety of Microscopy;http://jmicro.oupjournals.org/.)

Dark-field image

Specimen

Incident electron beam Incident electron beam

a b

nv_oct04 9/10/04 10:29 AM Page 667

© 2004 Nature Publishing Group

© 2004 Nature Publishing Group