imaging and analysis of nanowires - the lieber...

17
Imaging and Analysis of Nanowires DAVID C. BELL, 1,2 * YUE WU, 3 CARL J. BARRELET, 3 SILVIJA GRADEC ˇ AK, 3 JIE XIANG, 3 BRIAN P. TIMKO, 3 AND CHARLES M. LIEBER 2,3 1 Center for Imaging and Mesoscale Structures, Harvard University, Cambridge, Massachusetts 02138 2 Division of Engineering and Applied Sciences, Harvard University, Cambridge, Massachusetts 02138 3 Department of Chemistry and Chemical Biology, Harvard University, Cambridge, Massachusetts 02138 KEY WORDS HRTEM; STEM; microanalysis; nanotechnology ABSTRACT We used vapor-liquid-solid (VLS) methods to synthesize discrete single-element semiconductor nanowires and multicomposition nanowire heterostructures, and then characterized their structure and composition using high-resolution electron microscopy (HRTEM) and analytical electron microscopy techniques. Imaging nanowires requires the modification of the established HRTEM imaging procedures for bulk material to take into consideration the effects of finite nanowire width and thickness. We show that high-resolution atomic structure images of nanowires less than 6 nm in thickness have lattice “streaking” due to the finite crystal lattice in two dimensions of the nanowire structure. Diffraction pattern analysis of nanowires must also consider the effects of a finite structure producing a large reciprocal space function, and we demonstrate that the classically forbidden 1/3 {422} reflections are present in the [111] zone axis orientation of silicon nanowires due to the finite thickness and lattice plane edge effects that allow incomplete diffracted beam cancellation. If the operating conditions are not carefully considered, we found that HRTEM image delocalization becomes apparent when employing a field emission transmission electron microscope (TEM) to image nanowires and such effects have been shown to produce images of the silicon lattice structure outside of the nanowire itself. We show that pseudo low-dose imaging methods are effective in reducing nanowire structure degradation caused by electron beam irradi- ation. We also show that scanning TEM (STEM) with energy dispersive X-ray microanalysis (EDS) is critical in the examination of multicomponent nanowire heterostructures. Microsc. Res. Tech. 64: 373–389, 2004. © 2004 Wiley-Liss, Inc. INTRODUCTION Richard Feynman’s 1959 lecture entitled “There’s Plenty of Room at the Bottom” is a source of inspiration for nanotechnology. During this lecture, Feynman fore- saw the importance and application of electron micros- copy. Looking back at this lecture, the prediction was uncanny, given that the imaging of nanostructures us- ing electron microscopy has played a critical role in the development of nanoscience, not to mention acting as a catalyst for further developments in electron micro- scope instrumentation. A beautiful example of nanoscale imaging was the discovery of the carbon nanotube using HRTEM by Sumio Iijima (1991). Nanoscale objects are by definition small (having di- mensions in the nanometer scale), their physical prop- erties being different from the corresponding bulk ma- terial; similarly, the imaging of these objects presents the need for new microscopy techniques. One-dimen- sional nanostructures such as nanowires have stimu- lated great interest due to their importance in funda- mental scientific research and their potential new tech- nology applications (Lieber, 2003). Studies to date have shown that these attractive nanowire building blocks can be assembled into nano-electronic and photonic devices including field-effect transistors (Cui and Lieber, 2001), logic gates (Huang et al., 2001b), decod- ers (Zhong et al., 2003), sensors (Cui and Lieber, 2001), light-emitting diodes (Duan et al., 2001), photodiodes (Wang et al., 2001), and laser diodes (Duan et al., 2001). The emergence of nanotechnology and the devel- opment of these fascinating applications rely on the rational control over the electronic and optical proper- ties of nanowires and the chemical synthesis (including the adjustment of the morphology, size, and material composition in various nanowire structures), all of which can only be observed by electron microscopy due to the finite size of nanowires. TEM and STEM imaging and analysis are critical for the future developments of nanowire-related structures and devices. This article describes the electron microscope techniques that we have developed to enable the characterization of nanowires both structurally and compositionally and to discuss the challenges and difficulties involved. From our experience, the characterization of nanow- ires takes place in two stages. The first stage focuses on the synthesis and development of the desired nanowire building block, with electron microscopy providing a feedback to rationally grow the desired nanowire. Com- positional information of the nanowires can be ob- tained using energy dispersive X-ray analysis (EDS). General nanowire structural information, such as how *Correspondence to: David C. Bell, Center for Imaging and Mesoscale Struc- tures, Harvard University, 7 Divinity Ave., Cambridge MA 02138. E-mail: [email protected] Received 2 July 2004; accepted in revised form 27 August 2004 DOI 10.1002/jemt.20093 Published online in Wiley InterScience (www.interscience.wiley.com). MICROSCOPY RESEARCH AND TECHNIQUE 64:373–389 (2004) © 2004 WILEY-LISS, INC.

Upload: buithien

Post on 08-Aug-2018

218 views

Category:

Documents


0 download

TRANSCRIPT

Imaging and Analysis of NanowiresDAVID C. BELL,1,2* YUE WU,3 CARL J. BARRELET,3 SILVIJA GRADECAK,3 JIE XIANG,3 BRIAN P. TIMKO,3

AND CHARLES M. LIEBER2,3

1Center for Imaging and Mesoscale Structures, Harvard University, Cambridge, Massachusetts 021382Division of Engineering and Applied Sciences, Harvard University, Cambridge, Massachusetts 021383Department of Chemistry and Chemical Biology, Harvard University, Cambridge, Massachusetts 02138

KEY WORDS HRTEM; STEM; microanalysis; nanotechnology

ABSTRACT We used vapor-liquid-solid (VLS) methods to synthesize discrete single-elementsemiconductor nanowires and multicomposition nanowire heterostructures, and then characterizedtheir structure and composition using high-resolution electron microscopy (HRTEM) and analyticalelectron microscopy techniques. Imaging nanowires requires the modification of the establishedHRTEM imaging procedures for bulk material to take into consideration the effects of finitenanowire width and thickness. We show that high-resolution atomic structure images of nanowiresless than 6 nm in thickness have lattice “streaking” due to the finite crystal lattice in twodimensions of the nanowire structure. Diffraction pattern analysis of nanowires must also considerthe effects of a finite structure producing a large reciprocal space function, and we demonstrate thatthe classically forbidden 1/3 {422} reflections are present in the [111] zone axis orientation of siliconnanowires due to the finite thickness and lattice plane edge effects that allow incomplete diffractedbeam cancellation. If the operating conditions are not carefully considered, we found that HRTEMimage delocalization becomes apparent when employing a field emission transmission electronmicroscope (TEM) to image nanowires and such effects have been shown to produce images of thesilicon lattice structure outside of the nanowire itself. We show that pseudo low-dose imagingmethods are effective in reducing nanowire structure degradation caused by electron beam irradi-ation. We also show that scanning TEM (STEM) with energy dispersive X-ray microanalysis (EDS)is critical in the examination of multicomponent nanowire heterostructures. Microsc. Res. Tech. 64:373–389, 2004. © 2004 Wiley-Liss, Inc.

INTRODUCTIONRichard Feynman’s 1959 lecture entitled “There’s

Plenty of Room at the Bottom” is a source of inspirationfor nanotechnology. During this lecture, Feynman fore-saw the importance and application of electron micros-copy. Looking back at this lecture, the prediction wasuncanny, given that the imaging of nanostructures us-ing electron microscopy has played a critical role in thedevelopment of nanoscience, not to mention acting as acatalyst for further developments in electron micro-scope instrumentation. A beautiful example ofnanoscale imaging was the discovery of the carbonnanotube using HRTEM by Sumio Iijima (1991).Nanoscale objects are by definition small (having di-mensions in the nanometer scale), their physical prop-erties being different from the corresponding bulk ma-terial; similarly, the imaging of these objects presentsthe need for new microscopy techniques. One-dimen-sional nanostructures such as nanowires have stimu-lated great interest due to their importance in funda-mental scientific research and their potential new tech-nology applications (Lieber, 2003). Studies to date haveshown that these attractive nanowire building blockscan be assembled into nano-electronic and photonicdevices including field-effect transistors (Cui andLieber, 2001), logic gates (Huang et al., 2001b), decod-ers (Zhong et al., 2003), sensors (Cui and Lieber, 2001),light-emitting diodes (Duan et al., 2001), photodiodes(Wang et al., 2001), and laser diodes (Duan et al.,

2001). The emergence of nanotechnology and the devel-opment of these fascinating applications rely on therational control over the electronic and optical proper-ties of nanowires and the chemical synthesis (includingthe adjustment of the morphology, size, and materialcomposition in various nanowire structures), all ofwhich can only be observed by electron microscopy dueto the finite size of nanowires. TEM and STEM imagingand analysis are critical for the future developments ofnanowire-related structures and devices. This articledescribes the electron microscope techniques that wehave developed to enable the characterization ofnanowires both structurally and compositionally and todiscuss the challenges and difficulties involved.

From our experience, the characterization of nanow-ires takes place in two stages. The first stage focuses onthe synthesis and development of the desired nanowirebuilding block, with electron microscopy providing afeedback to rationally grow the desired nanowire. Com-positional information of the nanowires can be ob-tained using energy dispersive X-ray analysis (EDS).General nanowire structural information, such as how

*Correspondence to: David C. Bell, Center for Imaging and Mesoscale Struc-tures, Harvard University, 7 Divinity Ave., Cambridge MA 02138.E-mail: [email protected]

Received 2 July 2004; accepted in revised form 27 August 2004DOI 10.1002/jemt.20093Published online in Wiley InterScience (www.interscience.wiley.com).

MICROSCOPY RESEARCH AND TECHNIQUE 64:373–389 (2004)

© 2004 WILEY-LISS, INC.

thick the nanowires are and if the nanowires are singlecrystal or amorphous, can readily be determined usingTEM. Both TEM and EDS characterization provide thenecessary inputs to tune the synthesis parameters toproduce the desired building block. The second stagestarts once the synthesis of the desired building blockhas been achieved. At this point, more advanced tech-niques are used to investigate the structure and prop-erties of the nanowire. A thorough characterizationincludes the determination of the growth direction,cross-section, surface morphology, dislocations, andstacking faults. Determination of the nanowire growthdirection and cross-section are important to further ourunderstanding of the surface energetics and the mech-anisms of the crystal formation using a catalyst-basedsynthesis process. Moreover, they provide an under-standing of the material properties that affect the elec-tronic or optical response of a nanowire fabricateddevice.

The imaging and analysis of nanowires with electronmicroscopy also involves an understanding of the lim-itations of standard imaging and analysis techniquesand how they can be applied to nanoscale objects. Thetechniques that we describe herein differ from thestandard imaging and analysis techniques involvedwhen characterizing bulk materials. It is possible toform incorrect structure determinations if the presenceof forbidden reflections in the diffraction patterns ofsilicon nanowires are not correctly taken into account,and the description of these reflections are detailed inthis article. Other considerations, such as beam dam-age, can affect the HRTEM images, and also changethe compositional information obtained using EDS.Nanowires are finite objects and their imaging andanalysis requires a new approach to the methodology ofdiffraction studies, HRTEM, and analytical microanal-ysis.

We organized the article into four broad topics:Nanowire Synthesis; Experimental Techniques; Dis-cussion; and Conclusion. In the section entitled Nano-wire Synthesis, we discuss the principles of synthesisof single composition nanowires, axial heterostruc-tures, and core-shell nanowires. We then move througha discussion of our experimental techniques, detailingnanowire sample preparation; HRTEM imaging ofnanowires; image artifacts such as image delocaliza-tion and incorrect lattice spacings; nanowire HRTEMimage simulations; diffraction studies, demonstratingthe large reciprocal space function of the nanowire; andmicroanalysis of nanowires. This is followed by a dis-cussion of HRTEM imaging and the possible effect ofcrystal faceting, anomalous diffraction from nanowires,and microanalysis, and finally we present a summaryof our conclusions.

NANOWIRE SYNTHESIS:THE MAIN CONCEPTS

Most nanowire syntheses we have undertaken arebased on the VLS growth mechanism as summarizedby Lieber (2003). In this mechanism, a nanocluster isused as a catalyst and exposed to carefully designedprecursor vapor (V). The phase diagram of the catalyst(S) and the decomposed precursor determines the tem-perature and pressure required to form a liquid alloydroplet (L) from the nanocluster. As more of the pre-

cursor incorporates, the nanocluster supersaturatesand initiates a crystal nucleation process, which re-sults in an axial growth of the solid nanowire (S). Adirect observation of VLS nanowire growth has beenreported using TEM by Wu et al. (2001). By choosingmonodisperse nanoclusters as catalysts, the size of theliquid alloy droplets can be confined and yield nano-wires with narrow diameter distribution and a closecorrelation between the nanowire diameter and initialnanocluster size as reported previously (Gudiksen,2000; Cui et al., 2001; Wu et al., 2004b). One of themain examples is the synthesis of silicon nanowiresusing differing sizes of monodisperse gold (Au) nano-particles as catalysts and silane (SiH4) as the vapor-phase precursor as shown in Figure 1a. A eutecticdroplet of Au and Si stays at one end and serves as acatalyst particle (Fig. 2a), setting a size limit fornanowire diameter. TEM studies (Gudiksen andLieber, 2000) have shown that by this means siliconnanowires with a narrow diameter distribution can besynthesized ranging from several nanometers to tens ofnanometers (Fig. 2b,c). Using similar synthesis tech-niques, nanowires of differing elemental compositioncan be obtained, such as silicon, germanium, galliumnitride, or cadmium sulfide, although depending on therequired end product the synthesis method can varyfrom chemical vapor deposition (CVD) to a laser abla-tion technique. Nanowire axial heterostructures arenanowires that have a modulation in composition alongtheir lengths, as shown in Figure 1b. In the planar

Fig. 1. Illustrations of nanowire synthesis process. a: Vapor-liq-uid-solid (VLS) mechanism for nanowire growth. b: Synthesis methodto form nanowire axial heterostructure. c: Synthesis method fornanowire core-shell heterostructure formation.

374 D.C. BELL ET AL.

semiconductor industry, compositionally modulatedthin film structures have been developed for electronicsand photonics applications. By changing the composi-tion along the nanowire, new opportunities for band-gap engineering and dopant modulation are possible atthe nanoscale. The synthesis of nanowire axial hetero-structure involves changing the types of precursorsduring the nanowire growth. For example, during sili-con nanowire CVD growth, by switching the dopantprecursor from diborane (B2H6) to phosphine (PH3)while keeping the silane (SiH4) precursor flowing, acomplementary doping along a single silicon nanowirecan be made with a p-n junction in a single nanowire.By similar approaches, GaAs/GaP, InAs/InP, and Si/SiGe nanowire axial heterostructures can also be pre-pared for photonic and electronic applications.

A third type of complex nanowire structure is calleda core-shell nanowire heterostructure, which has amodulation in elemental composition along the nanow-ire radius. The changing of the precursors during thenanowire growth is similar to the axial heterostruc-tures synthesis; however, the preference of synthesiz-ing a core-shell structure over an axial heterostructurecan be achieved by shifting the growth temperatureand pressure. This relates to employing a much higherhomogeneous precursor decomposition rate on thenanowire surface than the axial growth rate as shownin Figure 1c. By changing the different reactants anddopants, it is possible to produce multiple shell struc-tures of differing compositions. Many examples withcore-shell compositions such as Ge/Si, Si/Ge, and p-Si/SiO2 core-shell structures have been demonstrated.

One of the purposes of these synthesis techniques isto produce nanowire-based functional devices. Thescope of such devices ranges from basic electronic de-vices like field-effect transistors (FETs) (Cui et al.,2001), logic gates (Huang et al., 2001a,b), decoders(Zhong et al., 2003), and sensors (Cui et al., 2001) tooptoelectronic devices like light-emission diodes (Duanet al., 2001), photodiodes (Wang et al., 2001), and laserdiodes (Duan et al., 2001). In these devices, either asingle nanowire with a source-drain contact or crossednanowires with a source-drain contact on each of thewires is used. For example, in the case of nanowireFETs, a single silicon nanowire can be used as a build-ing block, the nanowire FET can be switched on and offby either the back gate of a heavily doped silicon waferor a top gate of lithographically defined metal linecrossing the nanowire. This kind of nanowire FET hasbecome the basis of other nanoelectronic devices in-cluding logic gates, decoders, sensors, and memories.More complex devices, such as nanoscale optoelectronicdevices, employ two crossed nanowires with one thep-type silicon nanowire and the other a direct band-gapcompound semiconductor nanowire, such as a CdS orGaN nanowire. In this way, taking advantage of abottom-up approach of using building blocks to assem-

Fig. 2. HRTEM images of silicon nanowires. a: Silicon nanowireshowing attachment of gold catalyst particle on the end. b: Thin�5 nm silicon nanowire showing lack of contrast along edges, butatomic structure information indicated along the length of the wire. c:Silicon nanowire �17 nm wide; micrograph shows clear atomic struc-ture detail.

375IMAGING AND ANALYSIS OF NANOWIRES

ble functional nanoscale devices, the desire is to inte-grate as many functions as possible into the fundamen-tal building block components.

EXPERIMENTAL TECHNIQUESHRTEM Imaging of Nanowires

HRTEM of nanowires and nanowire devices followsfrom high-resolution imaging methods such as thosedescribed by Williams and Carter (1996) and Spence(1981), with adaptations of imaging methods requireddue to the finite nature of the samples. Since some ofour nanowires can be thinner than 6 nm, imagingbecomes problematic due to the lack of total atom den-sity (dependent, of course, on the atomic number of thematerial involved), and hence lack of scattering ele-ments to produce a suitable phase contrast image. Ad-justment of the objective lens defocus to obtain opti-mum imaging conditions and hence sufficient contrastto record the micrograph is problematic, and objectivelens defocus values which deviate from an instrumen-tal optimum can lead to other imaging issues that wedetail below. Tilting nanowires to a required zone-axisneeded for HRTEM imaging is also difficult due to thelarge reciprocal space function from the thin diameterof the wire. Apart from imaging nanowires dispersedon a TEM grid, we have previously demonstrated thatHRTEM imaging of a nanowire cross section can beachieved by embedding nanowires into epoxy and mic-rotoming the nanowire/epoxy solid mixture as de-scribed in Wu et al. (2004b).

Experimentally, the nanowires that have been syn-thesized as previously described are mechanically re-moved from the growth substrate and dispersed into2-propanol, which is then droppered onto a TEM gridwith lacey or holey carbon support film. Lacey carbonis suitable when the nanowires are longer than10 microns. For shorter nanowires, holey carbon or aplain support film is often preferred to increase surfacearea for securing nanowires for observation. Anothermethod for removal of the nanowires from the sub-strates involves sonication of the substrate in ethanoland then dispersion onto the lacey carbon TEM grids.HRTEM imaging of our synthesized nanowires wasmade using a JEOL 2010F TEM operating at 200 kV.Prior to insertion, the TEM sample and holder rod wereprepumped with a Fischione 1020 pumping station for20 minutes to ensure complete removal of remainingsolvents or water contamination. A thin silicon nanow-ire of order 5 nm in diameter can amorphize in lessthan a minute under a 200 kV electron beam; thus, thethinner the wire the faster the damage, as Figure 5a,bindicates for a ZnS nanowire. We therefore developedan imaging protocol based on low-dose microscopy tech-niques in order to minimize the effects of beam damageon the crystalline structure of the sample. In ourmethod, the focusing and crystal zone-axis orientationare performed on an adjacent region of the sample. Inthe case of a nanowire, the most logical position forthese alignments are further along the wire lengthfrom the area of interest; the sample is then moved tothe area of interest and the HRTEM micrographsquickly recorded.

Figure 2a shows a single silicon nanowire with a goldcatalyst particle still attached to one end, the other endof the wire that is not shown is anchored to a holey

carbon support film. Figure 2b presents an HRTEMimage of a thin �5 nm silicon nanowire indicatingsome of the problems of nanowire image interpretation.In this image the core region of the wire appears crys-talline, while the edges appear amorphous; it is unclearfrom this image if the wire edges are actually amor-phous or if the structure of the edges only appearsamorphous due to the lack of atomic scattering neededto produce a phase contrast image. Figure 2c shows asilicon nanowire; this wire is 17 nm wide and the imageshows clear atomic structure detail, similar to whatwould be observed from a bulk single crystal siliconsample.

Other nanowire systems have been examined withHRTEM. Figure 3a shows a TEM image of an NiSi/Sinanowire axial heterostructure; in this image the me-tallic nickel silicide section of the nanowire appearsclearly dark with a well-defined junction to the siliconsection. Figure 3b shows the corresponding HRTEMimage of the NiSi/Si interface; from such an image thedegree of lattice misfit can be calculated following con-ventional TEM methods as applied to semiconductormaterials. This method of image analysis has beenpreviously shown to be possible for the case of nanow-ires by Lauhon et al. (2002) and Gudiksen (2002). Wealso imaged core-shell nanowire structures with HR-TEM. Figure 4a,b shows a core-shell heterostructure;the core is a 20-nm n-doped silicon nanowire grownfrom a gold catalyst via the VLS mechanism as de-scribed previously. The shell is intrinsic Ge that wasdeposited by CVD. HRTEM allows for a detailed exam-ination of the interface between the Si and Ge (Fig. 4c)and, combined with selected area diffraction patterns,these images allow the degree of lattice misfit along theboundary to be determined, an important factor for

Fig. 3. NiSi/Si nanowire axial heterostructure. a: TEM image ofthe nanowire device clearly showing the junction between NiSi andNi. b: HTREM image of the atomically sharp metal-semiconductorjunction of NiSi/Si. c: EDS linescan taken along the length of thenanowire showing the junction as a function of characteristic Ni-KX-ray line.

376 D.C. BELL ET AL.

relating nanoscale structure to electronic device func-tion.

Imaging nanowires using a thermal field emissionmicroscope, such as the JEOL 2010F used in thisstudy, can also lead directly to “artifacts” in the HR-TEM nanowire images. Should the microscope operat-ing conditions not be properly considered, for instance,if the chosen objective lens underfocus value is signif-icantly away from an optimum “close to” Scherzer de-focus value, the micrograph can show image delocal-ization, where the lattice image is displaced from theapparent sample (Otten and Coene, 1993). Althoughthis effect is not limited to the examination of nanow-ires, it becomes readily apparent that even smallamounts of delocalization will be easily noticeable inthe recorded micrograph due to the finite width of thewire. Without correctly considering the setting of theobjective lens underfocus value, lattice fringes are no-ticeably formed outside the borders of the nanowireitself and usable phase contrast is lost from the imageof the wire. We have demonstrated this effect as shownin Figure 6, when imaging a silicon nanowire and ad-justing the objective lens defocus to 200 nm below themicroscopes Scherzer value. Since for imaging nano-scale objects, phase contrast is small initially, this ef-fect needs to be aggressively minimized. From Lichte(1991) the optimum value of underfocus to minimizeimage delocalization should be:

�fopt � � MCS�2umax2

Fig. 4. TEM images of Ge/Si core-shell nanowire. a: The Si corewhich is a 20 nm n-doped nanowire grown from a gold catalyst via theVLS mechanism. b: TEM image of the nanowire with clearly definedshell structure. c: HRTEM image of the shell which is intrinsic Gethat was deposited by CVD.

Fig. 5. HRTEM images of ZnS nanowire. a: HRTEM image ofnanowire edge showing atomic structure information. b: HRTEMimage of nanowire edge after several minutes of 200 kV electron beamirradiation.

377IMAGING AND ANALYSIS OF NANOWIRES

where M is an operational value actually determinedby the cut-off value chosen for u (spatial frequency) andhas a value between 0.75 and 1. For our JEOL 2010F,which has a spherical aberration value (Cs) of 1.0 mm,and operating at 200 kV, a typical range of underfocusvalues for the objective lens should be between –70 nmto –50 nm, to minimize the effects of delocalization (atsuitable spatial frequencies corresponding to d110 of Siof 0.34 nm), which fortunately encompasses the ex-tended Scherzer defocus value (for our microscope it is–61 nm). As stressed by Williams and Carter (1996),image delocalization can only be minimized in a fieldemission TEM, but not eliminated.

Other HRTEM nanowire imaging “artifacts,” such asincorrect “lattice spacings” caused by the large recipro-cal space function, can be avoided. Using the break-down of the weak phase object (WPO) approximation toindicate correct alignment of the zone-axis, it is possi-ble to know when the spacings in the HRTEM imagecan be reasonably trusted. Malm and O’Keefe (1997)described this effect when imaging nanocrystalline pal-ladium particles by conducting image simulations atvarious tilts and defocus values. In this article, theauthors argue that a nanoscale crystal oriented withina few milliradians of the zone-axis will show white dotscentered on black dots, indicating that the WPO hasbroken down and that the image is close to the zone-axis, and hence the measured lattice spacings of theHRTEM or “structure image” will be near the correctvalues. Tilting away from the zone-axis can producewhat appears to be an atomic structure image far awayfrom the zone-axis that has apparent spacings that candeviate by up to 10%. We have experimentally observedthese effects when imaging silicon nanowires that aretilted away from a defined zone-axis. With the micro-scope magnification previously being calibrated with abulk crystal sample, Figure 7a shows a silicon nanow-ire tilted � 1° away from the [111] zone-axis orienta-tion and shows fine lattice spacings that when mea-

sured from the micrograph indicate a distance in the0.21–0.24 nm range (interestingly dependent on fringeorientation), the corresponding value for d220 �0.19 nm. Figure 7b shows the same wire tilted to zone-axis orientation with the lattice spacing measured as0.19 � 0.01 nm, as expected. It should be noted that itis possible to measure lattice spacings differing frompublished values due to chemical or structural changes(perhaps due to beam irradiation) or strains present inthe material, so the experimental procedure should be

Fig. 6. HRTEM image of a 10 nm silicon nanowire showing imagedelocalization.

Fig. 7. HRTEM images of a silicon nanowire. a: Aligned along the[111] zone axis. b: Tilted slightly �1° away form the [111] zone axisand having “deceptive lattice spacings.”

378 D.C. BELL ET AL.

repeated several times to confirm the reproducibility ofthe results.

Nanowire HRTEM Image SimulationThe finite diameter of the wires, especially wires

which are �6 nm, produce “nanoscale imaging” effectsthat need to be considered when recording HRTEMmicrographs and either confirmed and understood ordismissed when interpreting the HRTEM structure im-ages. Complete HRTEM “structure image” understand-ing must include the corresponding multislice imagesimulations for the particular thickness and defocusvalues employed. We have found that the HRTEM im-age simulations of very thin nanowires must be basedon a modified multislice approach, which needs to in-clude the effects of the reducing number of atoms to-wards the edge of the wire, such as previously dis-cussed above and outlined for nanomaterials by Nihoulet al. (1998), and perhaps more importantly, must in-clude consideration for the large reciprocal space func-tion which corresponds to the wire’s thin diameter. Thesymmetry of the wire allows the calculation to employa modified two-dimensional shape function; our peri-odic object being modeled is now finite in two dimen-sions. The Fourier transform for a nanowire orientedalong the x direction becomes:

V(x, y, z) � V(y)�FT [V�u�]

where we have assumed that z is the direction of theincident electrons, and u is the associated vector inFourier space corresponding to the real space vector y.Modifying the multislice algorithm directly is possible,but a more useful approach to the finite size problem isdefining a modified crystal unit cell that produces thesame effect and can be achieved easily. In this ap-proach, the unit cell is a modified crystal unit cell thatencompasses the width of the wire and slightly beyondit into “free space” (Fig. 8) and then propagating thatcell along the required wire length before inputtinginto the multislice algorithm of choice. In our case, wecreated a small algorithm that required defining thewire width and desired length, elemental components,and base unit cell structure; the output from this is adata file that can be read into the multislice algorithm,in our case the Electron Microscopy Image Simulation(EMS) for which a Java-based version is available on-line (Stadelmann, 1995) or the original EMS program(Stadelmann, 1987).

HRTEM image simulations of nanowires are shownin Figures 9 and 10; the numbers on the image simu-lation inset reflect defocus/thickness values for the sim-ulation. The thickness values were chosen to corre-spond closely to the wire diameter. Unlike bulk orinfinite crystal image simulation, where the crystalthickness is estimated from the simulation, nanowiresimulation is actually more defined, since the thicknessof the wire is within several nm of the diameter. Thepossible thickness range is reduced and the objectivelens defocus value is obtained from the microscope;hence, image simulation parameters are well defined.Simulation of the CdS nanowire (Fig. 9) was performedby assuming an infinite crystal; this nanowire is rela-tively thick and the atomic structure image indicated

well-defined 2D fringes. Correspondingly, we discov-ered that image simulation of silicon nanowires thatare thicker than 6 nm can be performed under theassumption of an infinite crystal (Fig. 10a,b). If, how-ever, the silicon nanowire is 6 nm or less, then thestructure image recorded shows a pseudo 1D latticefringe effect or “streaking” as a result of the largereciprocal space function (Fig. 10c,d). In this case themodified multislice unit cell algorithm (Fig. 8) wasused to produce the matching simulations as indicated.It is of course possible to have 1D lattice fringes in anyHRTEM image; however, a corresponding image sim-

Fig. 8. Illustration of the new unit-cell definition for input into anelectron multislice simulation program.

Fig. 9. HRTEM image of CdS nanowire; the inserts are multislicesimulations with the indicated values of objective lens defocus/crystalthickness in nanometers.

379IMAGING AND ANALYSIS OF NANOWIRES

ulation at the same defocus and thickness values of athin film sample shows only a well-defined 2D atomicstructure image.

Diffraction Studies of Silicon NanowiresBecause the reciprocal space function of the nano-

wire is large corresponding to the “thinness” of thenanowires, the interception of the diffracted beamswith the Ewald sphere can occur at angles significantlyaway from a zone axis, and still produce a diffractionpattern that at first appears to be zone axis-oriented. Aslightly converged incident electron beam will producein diffraction space rods rather than the typical con-vergent beam disks, as shown in Figure 11, which is asilicon nanowire that has been initially tilted to a [111]zone-axis orientation. Indexing of this zone-axis orien-tation is presented in Figure 15b; note that this patternshows reflections present at the 1

3{4�22} positions that

should be classically forbidden. A possible explanationfor the presence of these reflections is presented in theDiscussion section.

The large reciprocal space function of the siliconnanowire has been experimentally demonstrated byperforming a sequence of tilting experiments along thenanowire’s radial and longitudinal axis. For this tiltseries we used a Fischione Instruments advanced to-mography holder that has a tilt limit of �75° on ourmicroscope.

In the initial sequence of experiments we foundthat tilting longitudinally (along the length of thewire) produces changes in the diffraction patternthat match the equivalent tilting of a bulk or equiv-alently infinite sample and the diffraction patternchanges rapidly as the Ewald sphere rotates, indi-cating well-defined changes of the reflections at allorders of Bragg diffraction, as illustrated by the se-ries of images in Figure 11.

By performing the second tilting experiment orthog-onally to the longitudinal axis, i.e., tilting radiallyaround the wire axis, the nanowire diffraction patternindicates a slowly changing pattern when compared tothe above for the equivalent tilt angle, as indicated inFigure 12. This is due to the large reciprocal spacefunction—in this direction the diffracted beams inter-sect the Ewald sphere and produce significant intensityat larger angles than the corresponding bulk samplediffraction pattern. Figure 12 indicates that it is notuntil the tilt is greater than 20° away from the initial[111] zone-axis orientation that the {220} reflectionsbecome noticeably absent, further tilting to 50° com-pletes the change of all orders of diffracted beams. Thisalso demonstrates that the orientation of the nanowireto a precise zone-axis in reciprocal space becomes prob-lematic: higher-order reflections must necessarily beused to correctly ascertain zone-axis alignment. Figure12 shows the higher-order reflection effects; at 20° tilt

Fig. 10. HRTEM images of silicon nanowires of varying thicknessesand inserts are the corresponding multislice simulations. a,b: Simu-lations of thick nanowires can be modeled as for the bulk materials.c,d: Simulations of thin wires less than 6 nm are modeled using themodified unit cell approach and show “streaking” that matches theHRTEM image. The indicated values of objective lens defocus/crystalthickness are in nanometers.

380 D.C. BELL ET AL.

and greater from the initial zone axis the (220) reflec-tions do not appear to be symmetrical. If we were toshow a wider field of view we would still observe thiseffect at low tilt angles, but notably only with very highorder Bragg reflections. Note that tilting 30° degreesoff zone-axis still shows the presence of the forbidden13{4�22} reflections, which could contribute to an incor-

rect zone-axis alignment needed for HRTEM studies insilicon nanomaterials along the [111].

The above experiments indicate that the higher-or-der reflections being produced from the larger Braggscattering angles are correspondingly more sensitive tochanges in sample tilt, and should be considered toascertain a correct zone-axis alignment.

Microanalysis of NanowiresComplex nanowire systems require characterization

by microanalysis techniques so as to provide a feedback

Fig. 11. Diffraction patterns at indicated sample tilt angle, taken from a silicon nanowire (starting at[111] zone-axis orientation) tilted along the nanowire longitudinal axis.

381IMAGING AND ANALYSIS OF NANOWIRES

on the synthesis process and to gain insight into thechemical composition of the resultant nanowire struc-tures. Typically, microanalysis is performed using aTEM or STEM fitted with a EDS system; with such asystem, characteristic EDS spectra, EDS linescans,and elemental or spectral maps can be obtained. Com-plex nanowire devices such as core-shell structures arebest examined using STEM and EDS with elementalline profiles acquired along or across the nanowire andspectrum images or maps of the overall structure. EDSspectrum imaging in particular is a valuable techniqueto provide understanding of the resultant synthesisprocesses and nanowire structures, which is furtherdetailed in this article.

For our experiments we employed a Vacuum Gener-ators HB-603 dedicated STEM or a JEOL 2010F oper-ating in STEM mode. Spectrum imaging obtained of anaxial nanowire heterostructure consisting of zinc sul-fide (ZnS) and zinc selenide (ZnSe) shows a uniformdistribution of zinc along the whole wire (Fig. 13b),while the elements selenium (Fig. 13c) and sulfur(Fig. 13d) are concentrated at the opposite side of thejunction. Figure 13a shows the corresponding bright-field image of the nanowire. In this particular wire weshow that the sulfur and selenium diffused towards

each other in the heterostructure and made a moder-ated structural transition, rather than similar abrupttransitions, as in the case of an Ni/NiSi nanowire. Wehave also found that STEM can be used to analyze thechemical composition distribution in a core-shellnanowire structure such as the type shown by thebrightfield image of an Si/CdS core-shell nanowire(Fig. 14a). For the Si/CdS core-shell nanowire struc-ture, the EDS spectrum (Fig. 14g) and spectrum im-ages were obtained using a JEOL 2010F operating inSTEM mode. These images clearly show that sulfur(Fig. 14c) and cadmium (Fig. 14d) concentrated in theshell region, while silicon (Fig. 14b) was concentratedonly in the core region.

Compared with the case of axial heterostructures, itis more complicated to determine the interface betweendifferent compositions due to the overcoating of shell.To extract this type of information, we employed EDSline profiles across the wire diameter and the line pro-files were then modeled by assuming a Gaussian elec-tron probe profile. In Figure 14e,f, the Cd-L (red) andSi-K (blue) line profiles, markers were the measuredvalues and solid lines were calculated profiles for theSi/CdS core-shell. Figure 14f shows the best fit andincludes a small concentration of Si in the shell. This

Fig. 12. Diffraction patterns at indicated sample tilt angle, taken from a silicon nanowire (starting at[111] zone-axis orientation) tilted along the nanowire radial axis, indicating lack of significant low-orderbeam change until the sample has been tilted 20° from zone-axis.

382 D.C. BELL ET AL.

experiment demonstrates that Gaussian fitting proce-dures are readily suitable for analysis of nanowire sys-tems in which the core and shell are made from distinctmaterials.

DISCUSSIONHRTEM Imaging

The HRTEM imaging and analysis of discrete single-element nanowires and more complex nanowire hetero-structures presents the researcher with distinct chal-lenges. As always, the interpretation of atomicstructure images obtained by HRTEM requires consid-eration of the physics involved in imaging and, in thecase of imaging nanowires, the effects of a “finite” crys-tal size (and hence not a completely periodic structure).Also, because of the large reciprocal space function (dueto the diameter of the wires), HRTEM imaging requiresthe careful consideration of possible artifacts and dif-

fraction effects before conclusions as to the structure ofthe wire can be made. We believe that for completeunderstanding a model of the nanowire or nanowireheterostructure should be constructed based on thediffraction, HRTEM structure images, and image sim-ulations, such as for our nanowires (Wu et al., 2004a,b).

Determination of the nanowire growth direction isone element required to construct a model of thenanowire structure and to help provide insight intothe dynamics of nanowire growth during the synthe-sis process. Determination of the crystal growth di-rection follows from Williams and Carter (1996); fordetermination of the growth direction of nanowireswe recorded and indexed the nanowire selected areadiffraction pattern from a planar wire tilted to azone-axis and compared the diffraction pattern ori-entation relative to the nanowire image. The latticespacing in the HRTEM structure image can also be

Fig. 13. Images of a ZnS/ZnSe nanowire obtained using scanning transmission electron microscopy(STEM). a: Darkfield STEM image of a ZnS/ZnSe nanowire. b: Elemental mapping of the zinc X-rayintensity compositing along the nanowire. c: Elemental mapping of the selenium compositing along thewire. d: Elemental mapping of the sulfur compositing along the nanowire.

383IMAGING AND ANALYSIS OF NANOWIRES

Fig. 14.

384 D.C. BELL ET AL.

used to assist in determining growth direction. (Itshould be noted that for some TEMs it is first neces-sary to perform a calibration using a crystal such asMgO to determine image rotation relative to thediffraction pattern.) The growth direction of thenanowire will lie orthogonal to the observed crystal

zone-axis and be in the direction of the Bragg reflec-tion that corresponds to the nanowire length. Forexample, as in Wu et al. (2004b), if a silicon nanowireis imaged along the [111] zone axis and the nanowiredirection corresponds to the (110) direction, then weknow that the nanowire growth occurs along the

Fig. 14. Scanning transmission electron microscope (STEM) im-ages including EDS spectrum imaging from a CdS/Si core-shell struc-ture. a: Brightfield image of the nanowire. b: Extracted Si-K X-rayintensity image. c: extracted Cd-L X-ray intensity. d: Extracted S-KX-ray intensity. e: Cd-L (red) and Si-K (blue) line profiles; markers aremeasured values and solid lines are calculated profiles for Si/CdS

core-shell structure in which the calculations do not assume Si in theshell. f: Same as e; however, the best fit includes small concentrationof Si in the shell. g: EDS spectrum of overall CdS nanowire; note thelow total counts. The characteristic Cu X-ray peaks are due to scat-tering from the TEM support grid.

Fig. 15. Anomalous diffractionin silicon nanowires along the[111] zone axis. a: HRTEM imageof a silicon nanowire with [111]zone axis. b: The correspondingdiffraction pattern with indexedreflections. c: Hexagonal unit cellfor an fcc structure. The a’ and b’are in the plane of the (111) (cubicnotation) of the fcc crystal andc’ is normal to the a’b’ plane.c � 7 �6b � �6a. d: Recipro-cal lattice points along the [001]axis of the hcp unit cell and [111]of the fcc unit cell. e: Projection of[111] zone fcc crystal reciprocallattices onto the zero order zoneplane. Circles: zero-order; trian-gles: first-order; squares: second-order Laue zone.

385IMAGING AND ANALYSIS OF NANOWIRES

{110} family of lattice planes. As an additional checkthe same nanowire can also be tilted axially to an-other zone-axis orientation and the growth directedagain determined.

We also discovered, by employing the modelingmethods as described above and confirmed by prepar-ing and imaging nanowire cross-sections, that mostnanowires have surfaces that are faceted (Wu et al.,2004a). This faceting of the wire surface most likelyexplains why the typical HRTEM structure images ofthin �20 nm nanowires show a central core area thatappears in images as distinctly crystalline, whereasthe edges can appear amorphous or at least with dif-fering phase contrast. Further study is needed to con-firm this hypothesis and develop HRTEM image simu-lation methods that take nanowire faceting into ac-count for thin nanowires.

Anomalous Diffraction From NanowiresFor certain silicon nanowire orientations, the diffrac-

tion patterns obtained differ from a corresponding bulksilicon pattern with the same zone-axis orientation. Asmentioned above, the diffraction pattern from a thinsilicon nanowire (Fig. 15a) taken along the [111] zoneaxis showed additional reflections a short reciprocaldistance from the transmitted beam (Fig. 15b). Suchreflections were first observed for silicon nanowires byGudiksen (2002) and have been observed in othernanoscale materials such as gallium oxide nanoribbonsby Dai et al. (2002). Other studies of cubic materialsimaged with TEM along the [111] zone axis also showlow intensity reflections that can be indexed to the13{4�22} family of lattice planes such as Xiao and Daykin

(1994). The 13{4�22} family of planes are forbidden in a

corresponding bulk crystal, for which the lattice can beconsidered infinite. Selected area electron diffractionpatterns of silicon nanowire samples of up to �60 nmin diameter shows the same pattern, with bright {220}spots and six smaller anomalous spots with weakerintensity inside the {220}. The spots that correspondwith the type 1

3{4�22} and have also been seen in [111]

diffraction from fcc films such as gold or silicon (Morriset al., 1968; Gibson et al., 1989).

We propose that there are two compounding reasonswhy these classically forbidden reflections appear soprominently in our recorded nanowire diffraction pat-terns. The first reason is that because the reciprocalspace function of the nanowire along the beam direc-tion is large, due to the thinness of the wire, the inter-ception of Ewald sphere by the diffracted beams in-creases the intensity transfer of any individual dif-fracted beam. The second reason is due to the nature ofthe nanowire object itself: physically, edge effects be-come important and noticeable as compared to diffrac-tion from a bulk or effectively infinite thin film crystallattice.

Our proposed explanation of this effect is as follows:silicon has a diamond structure, which consists of twosets of fcc lattices of Si offset by 1

4lattice spacing along

the [111] direction. An fcc unit cell has four latticepoints and an easy analysis of the diffraction structurefactor Fhkl gives only reflections from all even or all odd(hkl) lattice planes that have nonzero intensity. Insilicon, each basis contains two identical silicon atomsat (0,0,0) and (1

4,14,14), therefore allowing only reflections

if h,k,l are all odd or even and hkl is a multipleof 4. The silicon nanowire diffraction pattern shown inFigure 15b shows a set of six diffraction spots whichappear inside the {220}, while the {220} is the smallestset of spots kinematically allowed in [111] zone axis ofa bulk fcc crystal, as discussed above.

One proposed origin of these anomalous diffractionspots can be explained by unequal numbers of the A, B,and C hexagonal planes perpendicular to [111] direc-tion in an fcc crystal. It can be better understood if wechoose a hexagonal hcp unit cell instead of a cubic onefor an fcc crystal as in Figure 15d (Cherns, 1974). Insuch an hcp unit cell, each cell contains three scatter-ing centers, at (0,0,0), (1

3,13,13), and (2

3,23,23), which cor-

responds to the three successive layers ABC. Thereciprocal lattice from a [001] or c’ axis in the hcpnotation can be viewed in Figure 15c. Figure 15c sug-gests that the 1

3{4�22} reflection seen in silicon nanow-

ires could originate from the (100) and (010) reflectionin the hexagonal plane.

The structure factor for the hcp unit cell depicted inFigure 15d can be easily calculated to be:

Fhkl � f ��� �1 � exp �2 i �h � k � l� / 3�

� exp �4 i �h � k � l� / 3��

which describes that only hkl�3n reflections willhave nonzero intensity. In Figure 15b it appears thatthe (100) and (010) are again forbidden. Importantlyfor a nanowire with small diameter and near sphericalcross-sectional shape, the number of ABC layers isfinite and it would be expected that there will be in-complete unit cells with only A or AB planes. If weconsider one of the forbidden 1

3{4�22} type reflection

and sum up all the unit cell column along the c’ direc-tion (normal to the plane as shown in Fig. 15d), thenthe column structure factor would be:

F � f ����1

3n

exp [2 in]

where n is the number of unit cells along c’ and does nothave to be an integer. The sum from 1 to 3n is over allABC planes in the column. Therefore, it is obvious thatif n is an integer, then all unit cells are complete and Fwill be zero but it is nonzero for n � m � 1

3, and

it explains the observation of forbidden diffractionspots in silicon nanowires. In the language of the mul-tislicing method, if we choose a slice thickness to be1/3c’ instead of c’, then each successive three slices hasa translational offset of (1

3,13,13) with respect to each

other. By using a Fourier transform, this offset leads to120° phase change in phase grating through each of thethree slices. Thus, a total of 3m slices will generate onlyan identity transformation and therefore no anomalousspots, while 3m � 1 slices will lead to finite intensity atthe anomalous spots similar to diffraction on a singleclosed-packed atom layer.

An alternative approach to understand the anoma-lous diffraction spots in silicon nanowires is to view theforbidden in-zone reflection spots as higher-order Lauezone reflections (Cherns, 1974; Lynch, 1971) where the

386 D.C. BELL ET AL.

sz (interaction distance) at the forbidden spots is solarge that it is equal to 1/c’. Figure 15e illustrates therelative position of the projection of the first and secondLaue zone reflections onto the zero-order zone plane. Itshows that the observed finite intensity in the forbid-den zero-order zone spots can be assigned to the high-er-order Laue zone diffraction of {111}1 and {002}2.

The existence of higher-order Laue diffractions asanomalous diffraction spots has been illustrated by anexperiment on gallium oxide nanobelt structures,which are also a family of high aspect-ratio quasi-1Dnanostructures, having a thickness of only 10–60 nm(Dai et al., 2002). Typical diffraction patterns takenfrom a single Ga2O3 nanobelt along [101] shows thereflections with strong intensity are determined to bezero-order Laue zone reflections. However, a set ofsmaller, weaker reflection spots can also be observedwhich do not belong to any of the allowed zero-orderspots, but rather, can be explained by first-order Lauereflection. The appearance of this high-order Laue zonereflection is because the nanobelt sample is so thin thatthe elongated “reciprocal rod” is long enough to beintersected by the Ewald sphere. A further reason con-tributing to this observation is that the lattice spacingd(100) is very large in gallium oxide, so the (100) spot inreciprocal space is very close to the origin, since [100]*is almost parallel to the zone axis [101], with only 13°difference; the first-order Laue zone in reciprocal spacecontaining (100) is very close in spacing with zero-orderLaue zone, both of these effects contribute to the ease ofobserving the higher order Laue reflection in the dif-fraction pattern.

Nanowire MicroanalysisAs previously mentioned, microanalysis of synthe-

sized nanowires and nanowire structures is importantfor several reasons: to understand the nature of thewires themselves, to provide a feedback mechanism tofurther enhance the development of nanowire synthe-sis, and to examine modes of synthesis failure. Themost frequently used form of microanalysis is energydispersive X-ray spectroscopy (EDS); because of theultrasmall nature of the nanowires the analysis is com-monly carried out in a TEM or STEM. Basic qualitativecompositional analysis using a TEM is achieved byfocusing a probe on the sample in imaging mode andexamining the EDS spectrum. The advantage of thismethod of operation is that imaging and analysis canbe performed at the same time or at least during thesame microscope session. Nanowire EDS analysis isbest conducted with either a dedicated STEM or aTEM/STEM operating in STEM mode. A detailed un-derstanding of the synthesized nanowire or nanowiredevice requires: knowledge of what the wire is made of,analysis of whether the synthesis components thatformed the wire ended up in the expected locations andin the correct quantities, and finally a determination ofthe sufficiency of the quality of the finished wire for theintended device purpose. For example, in the case of ananowire core-shell structure it is important to knowthat the shell structure is uniform and complete. Mi-croanalysis is also essential to determine levels of re-action by-products and the identification of possiblecontamination in the nanowire or device.

One of the required features in a semiconductor het-erostructure is abrupt compositional or structuraltransition between different semiconductor regions. Inthe mature field of planar semiconductors, electronmicroscopy techniques have been developed that allowinvestigation of strains, chemical, and structural com-position on a nanoscale. For planar semiconductorstructures, where the thickness of TEM thin foils doesnot vary significantly over the region of interest, high-resolution TEM (Rosenauer, 2003) and electron ener-gy-loss spectroscopy (EELS) (Leifer et al., 2000) tech-niques are suitable for determination of compositionand strains or identification of point defects (Gradecaket al., 2002). However, these methods are not readilyapplicable for study of nanostructures in which thechemical composition changes simultaneously with thesample thickness. Similar problems are faced whenchemically sensitive TEM techniques like dark-field orweak-beam imaging are used.

In a STEM, a finely focused electron beam (typicallywith a diameter of less than 5 nm) scans across thenanowire sample. Transmitted, secondary, back-scat-tered, and diffracted electrons as well as the character-istic X-ray spectrum can be observed. By evaluation ofthe EDS spectrum, the information of the distributionof different composition in complex nanowire structurecan be retrieved by either extracting an EDS line pro-file first by doing a line scan along or across the nano-wires or to construct spectrum images of different ele-ments (also known as elemental mapping) by collectingX-ray signals over a defined area.

Experimental considerations are critically importantfor successful nanowire analysis; in the case of EDSanalysis, with a fine �3 nm electron probe it is moreprudent to limit the probe current to reduce damage ofthe sample and collect for longer duration than to max-imize probe current, as is the normal procedure. Beamdamage caused by the electron probe on a typicalnanowire sample can easily cut into the wire and ac-tually break the wire, so test runs of differing probecurrents and size are needed to optimize detector countrate but leave the wires intact. Generally requiredmicroscope conditions for microanalysis follow fromGarratt-Reed and Bell (2003); issues to avoid includeshadowing effects and to understand the EDS geome-try considerations to ensure that the sample presents aplanar view to the location of the EDS detector in themicroscope. If performing line scans or spectrum imag-ing the nanowire must be aligned to present a uniformaspect ratio to the detector along all points of thescanned area, failure to understand the geometry ofthe analysis results in acquiring data where the effectsof the sample aspect ratio change can be misinter-preted as compositional changes along the wire lengthor around the radius.

Collected EDS spectra are analyzed by identifyingthe characteristic X-ray peaks, subtracting the EDSbackground, and performing a peak quantification(commercial software packages are available suitablefor large dataset analysis, but special care should betaken during this procedure to avoid quantificationartifacts). In the first approximation for nanowireswith relatively small thickness, when absorption ef-fects are neglected, intensity of an elemental peak canbe considered as proportional to the volume containing

387IMAGING AND ANALYSIS OF NANOWIRES

the specific element and excited by an electron beam.The general morphology and approximate cross-sectionhence have to be determined with imaging methodsprior to EDS analysis to allow correct modeling of thesample thickness, and therefore EDS intensity, alongthe line profile. In this sense, double tilt stages withlarge tilt angles and/or cross-section TEM samples areof great help. Quantitative line profiles can be readilymodeled by assuming a Gaussian electron probe pro-file. From these measurements thickness of the coreand shells in the core-shell nanowires can be calculated(for noncylindrical nanowires the line profile woulddepend on the sample orientation toward an electronbeam). Reabsorption effects should be taken into ac-count when interpreting quantitative results and esti-mating the concentration of more complex structures,in which one element might be present in both core andthe shell.

CONCLUSIONSElectron microscopy will continue to play an impor-

tant role in the further development of nanotechnologyand nanomaterials. Further enhancements of electronmicroscope instrumentation as well as the correspond-ing refinements of electron microscopy methods andtechniques will lead to greater understanding of mate-rial fundamentals of nanoscale devices and synthesismethods. Imaging and analysis of nanowire-basednanotechnology follows from traditional high-resolu-tion and analytical electron microscopy methods; how-ever, we have shown the need for modifications to theprocedures because of the nanoscale dimensions of thesamples involved.

Nanowire syntheses are based on the vapor-liquid-solid (VLS) growth mechanism in which a nanoclusteris used as a catalyst, and then exposed to a vaporprecursor that initiates through alloying a crystal nu-cleation process to grow the nanowire. More complexnanowire devices are possible with modification of thesynthesis methods and compositionally modified het-erostructures can be formed either along the wire,termed a nanowire axial heterostructure, or radially asa core-shell heterostructure.

Techniques for HRTEM of nanowires and nanowire-based devices must correspondingly account for thefinite size of the sample, and the effects on the crystalstructure due to electron beam irradiation. Knowledgeof the crystal structures involves a corresponding un-derstanding of the effects that the large reciprocalspace function of the wire have on the analysis. Tiltingof the nanowires to a zone-axis alignment for HRTEMrequires the use of high-order reflections for correctorientation. Forbidden reflections are also possiblewhen recording diffraction patterns from the nanoscaleobjects due to crystal thickness and composition. HR-TEM multislice image simulation routines need modi-fication to incorporate the effects of finite samplewidth, as well as the possibility of “incorrect latticespacings” being present in the HRTEM structure im-age.

Scanning electron microscopy techniques have beenshown to be critical to the microanalysis of nanowiresand nanowire-based devices, with spectral imagingnecessary to quantify and successfully fabricate nano-devices or understand modes of synthesis failure. The

future of nanotechnology development is fundamen-tally dependent on the techniques of high-resolutionelectron microscopy and microanalysis.

ACKNOWLEDGMENTSThe authors thank Anthony J. Garratt-Reed at the

Center for Materials Science and Engineering at MITfor assistance and the use of the Vacuum GeneratorsHB-603 STEM.

REFERENCESBarrelet CJ, Wu Y, Bell DC, Lieber CM. 2003. Synthesis of CdS and

ZnS nanowires using single-source molecular precursors. J AmChem Soc 125:11498–11499.

Bell DC, Wu Y, Barrelet CJ, Lieber CM. 2003. Imaging nanotechnol-ogy. Microsc Microanal 9(suppl 2):284–285.

Cherns D. 1974. Direct resolution of surface atomic steps by trans-mission electron microscopy. Philos Mag 30:549.

Cui Y, Lieber CM. 2001. Functional nanoscale electronic devices as-sembled using silicon building blocks. Science 291:851–853.

Cui Y, Lauhon LJ, Gudiksen MS, Wang J, Lieber CM. 2001. Diame-ter-controlled synthesis of single-crystal silicon nanowires. ApplPhys Lett 78:2214–2216.

Dai ZR, Pan ZW, Wang ZL. 2002. Gallium oxide nanoribbons andnanosheets. J Phys Chem B 106:902–904.

Duan X, Huang Y, Cui Y, Wang J, Lieber CM. 2001. Indium phos-phide nanowires as building blocks for nanoscale electronic andoptoelectronc devices. Nature 409:66–69.

Garratt-Reed AJ, Bell DC. 2003. Energy-dispersive X-ray analysis inthe electron microscope. London: BIOS Scientific (US distributor,Springer).

Gibson JM, Lanzerotti MY, Elser V. 1989. Plan-view transmissionelectron diffraction measurement of roughness at buried Si/SiO2interfaces. Appl Phys Lett 55:1394–1396.

Gradecak S, Wagner V, Ilegems M, Riemann T, Christen J, Stadel-mann P. 2002. Microscopic evidence of point defect incorporation inlaterally overgrown GaN. Appl Phys Lett 80:2866–2868.

Gudiksen MS. 2002. Semiconductor nanowires and nanowireheterostructures: development of complex building blocks for nano-technology. Ph.D. Thesis, Harvard University, Cambridge, MA.

Gudiksen MS, Lieber CM. 2000. Diameter-selective synthesis of semi-conductor nanowires. J Am Chem Soc 122:8801–8802.

Hu J, Odom TW, Lieber CM. 1999. Chemistry and physics in onedimension: synthesis and properties of nanowires and nanotubes.Acc Chem Res 32:435–445.

Huang MH, Mao S, Feick H, Yan H, Wu Y, Kind H, Weber E, RussoR, Yang P. 2001a. Room-temperature ultraviolet nanowire nanola-sers. Science 292:1897–1899.

Huang Y, Duan X, Cui Y, Lauhon LJ, Kim K, Lieber CM. 2001b. Logicgates and computation from assembled nanowire building blocks.Science 294:1313–1317.

Iijima S. 1991. Helical microtubules of graphitic carbon. Nature 354:56–58.

Lauhon LJ, Gudiksen MS, Wang D, Lieber CM. 2002. Epitaxial core-shell and core-multishell nanowire heterostructures. Nature 420:57–61.

Leifer K, Buffat PA, Stadelmann PA, Kapon E. 2000. Theoretical andexperimental limits of the analysis of III/V semiconductors usingEELS. Micron 311:411–427.

Lichte H. 1991. Optimum focus for taking electron holographs. Ultra-microscopy 38:13–22.

Lieber CM. 2003. Nanoscale science and technology: building a bigfuture from small things. MRS Bull 28:486–491.

Lynch DF. 1971. Out of zone effects in dynamic electron diffractionintensities from gold. Acta Cryst A27:399.

Malm J-O, O’Keefe MA. 1997. Deceptive “lattice spacings” in high-resolution micrographs of metal nanoparticles. Ultramicroscopy 68:13–23.

Morris RH, Bottoms WR, Peacock RG. 1968. Growth and defect struc-ture of lamellar gold microcrystals. J Appl Phys 39:3016–3021.

Nihoul G, Sack-Kongehl H, Urban J. 1998. Electron microscopy struc-tural characterization of nano-materials: image simulation and im-age processing. Cryst Res Technol 33:1025–1037.

Otten MT, Coene WMJ. 1993. High-resolution imaging on a fieldemission TEM. Ultramicroscopy 48:77.

Rosenauer A. 2003. Transmission electron microscopy of semiconduc-tor nanostructures. Heidelberg: Springer.

388 D.C. BELL ET AL.

Spence JCH. 1981. Experimental high resolution electron microscopy.Oxford: Oxford University Press.

Stadelmann PA. 1987. EMS—a software package for electron-diffrac-tion analysis and HREM image simulation in materials science.Ultramicroscopy 21:131–145.

Stadelmann PA. 1995. Electron microscopy image simulation on-line Java version, http://cimewww.epfl.ch/people/stadelmann/jemsWebSite/jems.html

Wang J, Gudiksen MS, Duan X, Cui Y, Lieber CM. 2001. Highlypolarized photoluminescence and photodetection from single in-dium phosphide nanowire. Science 293:1455–1457.

Williams DB, Carter CB. 1996. Transmission electron microscopy.New York: Plenum Press.

Wu Y, Yang P. 2001. Direct observation of vapor-liquid-solid nanowiregrowth. J Am Chem Soc 123:3165–3166.

Wu Y, Xiang J, Yang C, Lu W, Lieber CM. 2004a. Single-crystalnanowires and metal/semiconductor nanowire heterostructures.Nature 430:61–65.

Wu Y, Cui Y, Huynh L, Barrelet CJ, Bell DC, Lieber CM. 2004b.Controlled growth and structures of molecular-scale silicon nano-wires. Nano Lett 4:433–436.

Xiao HZ, Daykin AC. 1994. Extra diffractions caused by staking faultsin cubic crystals. Ultramicroscopy 53:325–331.

Zhong Z, Wang D, Cui Y, Bockrath MW, Lieber CM. 2003. Nanowirecrossbar arrays as address decoders for integrated nanosystems.Science 302:1377–1379.

389IMAGING AND ANALYSIS OF NANOWIRES