J. Phys. Chem.

B 2000, 104, 1153-1175 1153

FEATURE ARTICLE

Transmission Electron Microscopy of Shape-Controlled Nanocrystals and Their Assemblies

Z. L. Wang*
School of Materials Science and Engineering, Georgia Institute of Technology, Atlanta, Georgia 30332-0245
ReceiVed: October 7, 1999; In Final Form: NoVember 18, 1999

The physical and chemical properties of nanophase materials rely on their crystal and surface structures.
Transmission electron microscopy (TEM) is a powerful and unique technique for structure characterization.
The most important application of TEM is the atomic-resolution real-space imaging of nanoparticles. This
article introduces the fundamentals of TEM and its applications in structural determination of shape-controlled
nanocrystals and their assemblies. By forming a nanometer size electron probe, TEM is unique in identifying
and quantifying the chemical and electronic structure of individual nanocrystals. Electron energy-loss
spectroscopy analysis of the solid-state effects and mapping the valence states are even more attractive. In
situ TEM is demonstrated for characterizing and measuring the thermodynamic, electric, and mechanical
properties of individual nanostructures, from which the structure-property relationship can be registered
with a specific nanoparticle/structure.

1. Structure of Nanocrystals and Physical Chemistry solid structural transformations from the low-temperature
The unique chemical and physical properties of nanocrystals optimal structures (such as truncated octahedron and truncated
are determined not only by the large portion of surface atoms decahedron) to icosahedral structures, which are believed to be
but also by the crystallographic structures of the particle surface. the intrinsic thermodynamic precursors to the melting transition.8
The former is determined by the size of the particles, and the Crystal structures can be determined by X-ray and neutron
latter relies on the particle shape. A large amount of research diffraction, while transmission electron microscopy (TEM) is
has been carried out to investigate the effects of particle size indispensable for characterization of nanocrystal materials,
on their properties. One typical example is that the melting particularly when particle shape is important. Although scanning
temperature of nanocrystals strongly depends on the crystal size tunneling microscopy and atomic force microscopy can provide
and is substantially lower than the bulk melting temperature. atomic-resolution images of large crystal surfaces, they are
The melting point of Au particles of core size 2.5 nm is 40% unlikely to clearly resolve the atomic lattices of nanoparticles
lower than that of bulk gold.1 Similar behavior has been because of the surface coating and the wobbling of the
observed for sodium clusters2 and CdS nanocrystals.3 The other nanocrystals under the scanning tip. TEM is likely to be very
interesting property is the quantum confinement effect in small powerful for revealing the atom distributions on nanocrystal
size metallic and semiconductor quantum dots.4,5 The shift of surfaces even when they are passivated with polymers. Todays
electron energy levels as a function of particle size gives rise TEM is a versatile tool that provides not only atomic-resolution
of emission of photons with unique wavelengths, potentially lattice images but also chemical information at a spatial
useful for optoelectronics. resolution of 1 nm or better, allowing direct identification the
The particle shapes are closely related to the crystallographic chemistry of a single nanocrystal.9-12 With a finely focused
surfaces that enclose the particles. The {111}, {100}, and electron probe, the structural characteristics of a single nano-
possibly {110} surfaces of face-centered cubic structured metal particle can be fully characterized. TEM is unique for character-
particles, for example, are different not only in the surface atom izing the in situ structural evolution of nanocrystals resulting
densities but also the electronic structure, bonding, and possibly from annealing, electric field, or mechanical stress, such as
chemical reactivities. The sublimation activation energy of Au imaging a single carbon nanotube when a mechanical or
{110} faces, for example, has been found to be significantly electrical measurement is being carried out in situ.13
lower than that of {100} and {111}, resulting in the sublimation Much of the research in the literature has been focused on
of Au atoms from the surface at temperatures as low as 220 the size effect. We have been interested in the property of shape-
C, much lower than the melting point of bulk gold of 1063 controlled nanocrystals, aiming at exploring the unique catalysis
C.6 The tetrahedral Pt nanocrystals enclosed mainly by {111} properties of different crystal faces.14 The research involves three
facets have been found to be more stable than the {100} critical steps: synthesis of shape-controlled nanocrystals, struc-
enclosed cubic Pt nanocrystals simply because of the difference tural characterization, and property characterization. TEM has
in surface energies.7 Molecular dynamics calculation reveals that played a major role in characterizing the shape-controlled
the melting process of Au clusters is punctuated by solid-to- nanocrystals. As a review of the research field, this article
focuses on the fundamentals of TEM and its applications in
* Corresponding author. E-mail: [email protected]. characterizing the structure of shape-controlled nanocrystals. The
10.1021/jp993593c CCC: $19.00 2000 American Chemical Society
Published on Web 01/11/2000
1154 J. Phys. Chem. B, Vol. 104, No. 6, 2000 Feature Article

Figure 1. Abbes theory of image formation in a one-lens transmission

electron microscope. This theory is for a general optical system in TEM.

physical basis of atomic scale imaging using TEM will be

Figure 2. Schematic diagrams showing the phase shift introduced by
illustrated first. Then, the applications of TEM for determination (a) spherical aberration and (b) objective lens defocus on the electrons
of the particle shape and surface structures will be given. In scattered off the central optical axis. This nonlinear information transfer
situ structural characterization by TEM will be illustrated. The characteristics of the TEM gives rise the complexity in interpretation
analysis of nanocrystal self-assembled arrays and the role played of TEM images. (c) The relationship between the diffraction beams in
by particle shapes in the assembly will be demonstrated. Finally, reciprocal space and the interference lattice fringes in real space (image).
the investigation of nanocrystal chemical and electronic proper- The phases of the reflected Bragg beams are modulated by the lens,
thus, the image contrast strongly depends on the defocus.
ties by electron energy-loss spectroscopy will be reviewed.
2. High-Resolution Transmission Electron Microscopy probably the most precise microanalysis technique in TEM.
A modern TEM is composed of an illumination system, a EELS analyzes the intensity distribution of the transmitted
specimen stage, an objective lens system, the magnification electrons as a function of their energy loss. It provides not only
system, the data recording system(s), and the chemical analysis the chemical information on the specimen but also its electronic
system. The electron gun is the heart of the illumination system, structure. A complementary application of the diffraction,
which typically uses LaB6 thermionic emission source or a field imaging, and spectroscopy techniques available in a TEM is
emission source. The LaB6 gun gives a high illumination current, likely to give a more precise and reliable determination of the
but the current density and the beam coherence are not as high crystal structure.
as those of a field emission source. Field emission source is 2.1. Image Formation. As a start, we first illustrate the image
unique for performing high coherence lattice imaging, electron formation process in an TEM. For easy illustration, a TEM is
holography, and high spatial resolution microanalysis. The simplified into a single lens microscope, as given in Figure 1,
illumination system also includes the condenser lenses that are in which only a single objective lens is considered for imaging
vitally important for forming a fine electron probe. The specimen and the intermediate lenses and projection lenses are omitted.
stage is key to carrying out structure analysis, because it can This is because the resolution of the TEM is mainly determined
be used to perform in situ observations of phenomena induced by the objective lens. The entrance surface of a thin foil
by annealing, electric field, or mechanical stress, giving the specimen is illuminated by a parallel or nearly parallel electron
possibility of characterizing the physical properties of individual beam. The electron beam is diffracted by the lattices of the
nanostructures. The objective lens is the heart of a TEM which crystal, forming the Bragg beams which are propagating along
determines the limit of image resolution. The magnification different directions. The electron-specimen interaction results
system consists of intermediate lenses and projection lenses, in phase and amplitude changes in the electron wave that are
and it gives a magnification up to 1.5 million. The data recording determined by quantum mechanical diffraction theory. For a
system tends to be digital with the use of a charge coupled thin specimen and high-energy electrons, the transmitted wave
device (CCD), allowing quantitative data processing and function (x,y) at the exit face of the specimen can be assumed
quantification. Finally, the chemical analysis system is the to be composed of a forward-scattered wave. The exit wave
energy-dispersive X-ray spectroscopy (EDS) and electron energy (x,y) contains the full structural information on the specimen.
loss spectroscopy (EELS), both can be used complementary to Unfortunately, this wave will be transmitted nonlinearly by the
quantify the chemical composition of the specimen. EDS relies optic system.
on the counting of X-rays emitted from the beam-illuminated The nonnear-axis propagation of the electrons through the
specimen region as a function of the photon energy, and it is objective lens is the main source of nonlinear information
Feature Article J. Phys. Chem. B, Vol. 104, No. 6, 2000 1155

Figure 3. Theoretically simulated images for a decahedral Au particle at various orientations and at focuses of (A) f ) 42 nm and (B) f ) 70
nm, illustrating the contrast reversal in the two cases (Courtesy of Drs. Ascencio and M. Jose-Yacaman).

transfer in TEM. The diffracted beams will be focused in the of the electrons scattered to an angle owing to the increased
back focal plane, where an objective aperture could be applied. path length (Figure 2b). From the Huygens principle, the
An ideal thin lens brings the parallel transmitted waves to a spherical wave emitted from point O to point C and D will
focus on the axis in the back focal plane. Waves leaving the have a path length difference of ) f(1 - cos )/cos
specimen in the same direction (or angle with the optic axis) f2/2. The correspondence phase shift increases as the scat-
are brought together at a point on the back focal plane, forming tering angle increases.
a diffraction pattern. The electrons scattered to angle experi- A high-resolution TEM image of a crystalline specimen is
ence a phase shift introduced by the spherical aberration and formed by the interference of the Bragg reflected beams (Figure
the defocus of the lens, and this phase shift is a function of the 2c). Since the phase of each Bragg beam is perturbed by the
scattering angle. The phase shift due to spherical aberration is phase shift induced by spherical and defocus, the object
caused by a change in focal length as a function of the electron information will be transmitted nonlinearly. Mathematically, the
scattering angle, and the phase shift owing to defocus is caused wave observed in the front focal plane is the inverse Fourier
by the spherical characteristics of the emitted wave in free space transform (FT-1) of (u), and the observed image intensity is
(e.g., the Huygens principle). The electron wave in real space
is related to that in reciprocal space by a Fourier transform for I(x,y) ) |FT-1[(u)]|2 ) |(x,y) X tobj(x,y,f)|2 (3)
the case of Fraunhofer diffraction. If (u) is the Fourier
transform (FT) of the wave (r) at the exit face of the specimen, where X indicates a convolution calculation of (x,y), tobj(x,y,f)
the diffraction amplitude function at the back focal plane is the inverse Fourier transform of the phase function
(diffraction plane) is exp[i(u)]. The convolution of the lens transfer function
introduces the nonlinear information transfer characteristics of
(u) ) (u) exp[i(u)] (1) the objective lens, leading to complexity in image interpretation.
where u is the reciprocal space vector that is related to the Equation 3 is the Abbes imaging theory which correlates the
scattering angle by u ) 2 sin /, and (u) is the phase function object wave function (x,y) with the experimentally observed
introduced by the objective lens, which is determined by the image intensity distribution I(x,y). It is easier to work from
spherical aberration coefficient Cs of the objective lens and the (x,y) to I(x,y), but not vice versa.
lens defocus f, 2.2. Contrast Mechanisms. Images in TEM are usually
dominated by three types of contrast. The first is diffraction
contrast, which is produced due to the perturbation of local
(u) ) Cs3u4 - fu2 (2) strain/defects/dislocation on the intensities of the Bragg reflected
2
beams.15 This contrast emphasizes the effect of defects on the
where is the electron wavelength. The aberration and defocus amplitude of the transmitted wave, and it is also called amplitude
of the lens is to modulate the phases of the Bragg beams contrast. If the image is formed by selecting the central
distributed in reciprocal space. The phase shift introduced by transmitted wave only, the image contrast is most sensitive to
spherical aberration is due to the variation in focal length as a defect structures and the image resolution is limited to 2 nm.
function of the electron scattering angle, resulting in a path For nanocrystals, most of the grains are defect-free in volume,
length difference (Figure 2a). The ray OA scattered to higher while a high density of defects are localized at the surface or
angle will be focused at a shorter focal length in comparison to grain boundary, diffraction contrast can be useful for capturing
the OB ray scattered to a lower angle; thus, an imaging strain distribution in nanocrystals of sizes larger than 15 nm.
broadening will be observed in the back focal plane. The phase Second, phase contrast is produced by the phase modulation
factor introduced by defocus represents the relative phase shift of the incident electron wave when it transmits through the
1156 J. Phys. Chem. B, Vol. 104, No. 6, 2000 Feature Article

be revealed using the images recorded at least from two

independent orientations.
Electron transmission through a thin crystal can be character-
ized by a phase modulation function, which is known to be the
phase object approximation (POA). This approximation assumes
that the electron wave is modulated only in phase but not in
amplitude (e.g., phase contrast). Therefore, the electron wave
at the exit face of a thin crystal is approximated to be

(x,y) ) exp[iVp(x,y)] (6)

If the incident beam travels along a low-index zone axis, the

Figure 4. Optic ray diagrams for performing (a) bright-field, (b) dark- variation of Vp(x,y) across atom rows is a sharp varying function
field, and (c) high-resolution imaging in an TEM. because an atom can be approximated by a narrow potential
well and its width is on the order of 0.2-0.3 . This sharp
crystal potential. This type of contrast is sensitive to the atom phase variation is the basis of phase contrast, the fundamental
distribution in the specimen. To illustrate the physics of phase of atomic-resolution imaging in TEM.
contrast, we consider the modulation of a crystal potential to The last one is mass thickness or atomic number contrast.
the electron wavelength. From the de Brogli relation, the Atoms with different atomic numbers exhibit different powers
wavelength of an electron is related to its momentum, p, by of scattering. If the image is formed by collecting the electrons
scattered to high angles, the image contrast would be sensitive
h to the average atomic number along the beam direction. This
) (4)
p type of imaging is most frequently performed in scanning
transmission electron microscopy.
When the electron goes through a crystal potential field, its 2.3. The White and Dark Spots in the Image, Which
kinetic energy is perturbed by the variation of the potential field, Represent Atoms? In a high-resolution TEM (HRTEM) image,
resulting in a phase shift with respect to the electron wave that one usually wonders if the atoms are dark or bright. To answer
travels in a space free of potential field. For a specimen of this question, one must examine the imaging conditions. For
thickness, d, the phase shift is the clarity of the following discussion, the weak scattering object
approximation (WPOA) is made. If the specimen is so thin that
Vp(x,y) ) 0d dz V(x,y,z) (5) the projected potential satisfies |Vp| , 1, the phase grating
function in eq 6 is approximated by
where ) /U0, U0 is the acceleration voltage, and Vp(x,y)
the thickness-projected potential of the crystal. Equation 5 (x,y) 1 + iVp(x,y) (7)
clearly shows that the phase contrast image is the result of the
thickness-projected image along the beam direction (z-axis) of From the imaging theory presented in eq 3, and ignoring the 2
a three-dimensional object. The 3-D shape of the crystal can term, the image intensity is calculated by

Figure 5. (a) Geometrical shapes of cubooctahedral nanocrystals as a function of the ratio, R, of the growth rate along the 100 to that of the
111. (b) Evolution in shapes of a series of (111) based nanoparticles as the ratio of {111} to {100} increases. The beginning particle is bounded
by three {100} facets and a (111) base, while the final one is a {111} bounded tetrahedron. (c) Geometrical shapes of multiply twinned decahedral
and icosahedral particles.
Feature Article J. Phys. Chem. B, Vol. 104, No. 6, 2000 1157

Figure 6. HRTEM images of cubic Pt nanocrystals oriented along (a) [001] and (b) [110], showing surface steps/ledges and the thermodynamically
inequilibrium shapes (The Pt nanocrystals were supplied by Prof. M. A. El-Sayeds group).

Figure 7. HRTEM images of Pt nanocrystals (a) with a truncated octahedral shape and oriented along [110] and (b) with a octahedral shape and
oriented along [110] and [001]. The inset in (a) is a model of the particle shape (The Pt nanocrystals were supplied by Prof. M. A. El-Sayeds
group).

I(x,y) 1 - 2Vp(x,y) X ts(x,y,f) (8) Bragg reflected beams. Any phase modulation introduced by
the lens would result in contrast variation in the observed image.
where ts(x,y,f) ) Im[tobj(x,y,f)]. The second term in eq 8 is The weak scattering object approximation fails in most practical
the interference result of the central transmitted beam with the cases, but it is the easiest and clearest way to illustrate the
1158 J. Phys. Chem. B, Vol. 104, No. 6, 2000 Feature Article

Figure 8. HRTEM images of truncated tetrahedral Pt nanocrystals oriented along [110]. The surface steps and ledges at the truncated corners are
clearly resolved (The Pt nanocrystals were supplied by Prof. M. A. El-Sayeds group).

Figure 9. TEM images of decahedral Au nanocrystals when the incident electron beam is parallel or nearly parallel to the 5-fold symmetry axis
(The Au nanocrystals were supplied by S. Link).

physics involved in HRTEM imaging. Equation 8 clearly shows the information provided by HRTEM. The dynamic diffraction
that an TEM image is a two-dimensional projection of a three- process is to solve the Schrodinger equation under given
dimensional object. The atom rows can be resolved clearly if boundary conditions. There are several approaches for perform-
their orientations are parallel to the incident beam. One must ing dynamic calculations.17 For a finite crystal containing defects
keep this in mind in interpretation of TEM images. and surfaces, the multislice theory18 is most adequate for
The optimum shape of the lens transfer function tobj is a numerical calculations. This theory has the unique advantage
narrow Gaussian-like function with a small tail. The spherical of fully accounting for the defects, surfaces, particle shapes,
aberration coefficient Cs of a TEM is fixed, the only variable is and interfaces in image calculation. Shown in Figure 3 are
the objective lens defocus f. Under the Scherzer defocus,16 simulated images of a decahedral Au particle in different
f ) (4/3Cs)1/2, ts(x, y,f) is approximated to be a negative orientations under two defocus conditions.19 At the Scherzer
Gaussian-like function with a small oscillation; thus, under the defocus (Figure 3a), the atoms are dark, while the contrast is
WPOA, the image is directly related to the two-dimensional reversed at defocus f ) 70 nm (Figure 3b). The contrast
thickness-projected potential of the crystal and the atoms show reversal is introduced mainly due to the phases introduced by
dark contrast. This is the basis of structure analysis using the defocus to the beams that contribute to the image. On the
HRTEM. On the other hand, the form of the ts(x,y,f) function other hand, the particle shape is apparent only if the decahedron
depends on the defocus of the objective lens; contrast reversal is oriented along its 5-fold symmetry axis. This may give some
is possible simply by changing the defocus of the objective lens. experimental difficulties of estimating the population of deca-
Therefore, the occurrence of dark atoms or bright atoms in the hedra in a specimen.
image is determined largely by the defocus value. 2.4. The Bright-Field, Dark-Field, and High-Resolution
For a general case, image simulation is required to quantify Lattice Imaging. There are two basic modes of TEM operation,
Feature Article J. Phys. Chem. B, Vol. 104, No. 6, 2000 1159

namely, the bright-field mode, where the (000) transmitted beam

contributes to the image, and the dark-field imaging mode in
which the (000) beam is excluded. The size of the objective
aperture in the bright-field mode directly determines the
information to be emphasized in the final image. When the size
is chosen so small so as to exclude all of the diffracted beams,
one has the configuration normally used for low-resolution
imaging of defects, called diffraction contrast. In this case, a
crystalline specimen is oriented to give a particular diffracted
beam, or a systematic row of reflections and the image is
sensitive to the differences in specimen thickness, distortion of
crystal lattices due to defects, strain, and bending. High-
resolution imaging is usually performed in bright-field mode
by including a few Bragg diffracted beams within the objective
aperture. When a crystal is oriented along one of its low-index
zone axes (e.g., [110] or [100]), strong Bragg reflected beams
are present. The lattice images are the result of interference
among the Bragg reflected beams and the central transmitted
beam, forming the phase contrast image.
Parts a and b of Figure 4 compare the ray diagrams for
forming the bright-field and dark-field images. The dark-field
mode involves the tilt of the incident beam by a pair of
deflection coils to bring the diffracted beam, to be selected for
imaging, parallel to the optic axis for reducing the spherical
aberration effect. Recording of high-resolution images is the
same as in the case of bright-field imaging except with a larger
size objective aperture (Figure 4c). The most important factors
in HRTEM are the astigmatism of the objective lens and the
adjustment of the lens defocus. The latter critically affects the
image contrast.

3. Structures of Individual Nanocrystals

3.1. Shapes of Polyhedral Nanocrystals. Surface energies
associated with different crystallographic planes are usually
different, and a general sequence may hold, {111) < {100) <
{110). For a spherical single-crystalline particle, its surface must
contain high-index crystallography planes, which possibly result
in a higher surface energy. Facets tend to form on the particle
surface to increase the portion of the low-index planes.
Therefore, for particles smaller than 10-20 nm, the surface is
a polyhedron. Figure 5a shows a group of cubooctahedral shapes
as a function of the ratio, R, of the growth rate in the 100 to Figure 10. (a) A high-resolution TEM image showing the presence
that of the 111. The longest direction in a cube is the 111 of twinning in Au nanoparticles. (b) A dark-field image recorded using
diagonal, the longest direction in the octahedron is the 100 a small section of the {111} and {200} reflection rings displaying the
diagonal, and the longest direction in the cubooctahedron (R ) distribution of twinned particles in the specimen.
0.87) is the 110 direction. The particles with 0.87 < R < 1.73 is face-centered cubic. The surface of the particle may have
have the {100} and {111} facets, which are named the truncated some steps and ledges particularly at the regions near the corners
octahedral (TO). The other group of particles has a fixed (111) of the cube. To precisely image the defects and facets on the
base with exposed {111} and {100} facets (Figure 5b). An cubic particles, a particle oriented along [110] is given (Figure
increase in the area ratio of {111} to {100} results in the 6b). This is the optimum orientation for imaging cubic structured
evolution of particle shapes from a triangle-based pyramid to a materials. In additional to the {111} and {100} facets, {110}
tetrahedron. facets are also observed, although they are considered as the
If the particle is oriented along a low-index zone axis, the surfaces with higher surface energy. The {110} facets are rather
distribution of atoms on the surface can be imaged in profile, rough, and the {111} facets are present. These higher energy
and the surface structure is directly seen with the full resolution structural features are present because the particles were
power of a TEM.20 This is a powerful technique for direct prepared at room temperature.
imaging the projected shapes of nanoparticles, particularly when An octahedron has eight {111} facets; four {111} facets are
the particle size is small. Figure 6a gives a profile HRTEM edge-on if viewed along [110]. If the particle is a truncated
image of cubic Pt nanocrystals oriented along [001]. The Pt octahedron, six {100} facets are created by cutting the corners
particles were prepared by reduction of K2PtCl4 by hydrogen of the octahedron, two of which are edge-on while viewing along
gas in the presence of polymer.14,21,22 The particle is bound by [110]. Parts a and b of Figure 7 show the HRTEM images of
{100} facets, and there is no defect in the particle volume. The [110] oriented truncated octahedron and octahedral Pt particles,
distances between the adjacent lattice fringes are the interplanar respectively. A variation in the area ratio of {100} to {111}
distance of Pt {200), which is 0.196 nm, and the bulk structure results in a slight difference in particle shapes.
1160 J. Phys. Chem. B, Vol. 104, No. 6, 2000 Feature Article

Figure 11. HRTEM images of Pt nanocrystals oriented along [110], showing the reconstructed {100} surfaces and surface relaxation (indicated
by arrowheads, where the interlayer distance is slightly larger than that between the layers inside the volume).
A tetrahedral particle is defined by four {111} faces, and it constituted of two decahedra. Twinning is the mechanism for
usually gives a triangular shape in HRTEM.23 Figure 8 gives forming these particles. The multiple twinning is the favorable
two HRTEM images of truncated tetrahedral particles oriented structural configuration when the particles are small,29 possibly
along [110]. Two {111} facets and one {001} facet (at the top because of the smaller surface and volume energies.
of the image as a result of truncation) are images edge-on. There The presence of stacking faults and twins in the nanocrystals
are some atom-high surface steps on the {111} surfaces and may strongly affect their optical properties. The fraction of
the corners. These atomic rough structures are likely important twinned particles can be reasonably evaluated using dark-field
for enhancing the catalysis activities of the nanocrystals. imaging, which is formed by selecting a small segment of the
3.2. Twinned and Multiply Twinned Nanocrystals. Twin- {111} and {200} reflected rings. The particles with orientations
ning is one of the most popular planar defects in nanocrystals, that can produce strong Bragg reflections falling into the angular
and it is frequently observed for face-centered cubic (fcc) selection range of the objective aperture in TEM show bright
structured metallic nanocrystals. Twining is the result of two contrast, provided the particles are single crystalline. For the
subgrains sharing a common crystallographic plane; thus, the twinned particles, however, part of the particles may show bright
structure of one subgrain is the mirror reflection of the other contrast and the other parts are darker, providing an easy
by the twin plane. The fcc structured metallic nanocrystals technique for identifying the presence of twins and their
usually have {111} twins. The two most typical examples of distribution among the particles. Figure 10 gives such an
multiply twinned particles (MTP) are decahedron and icosahe- example. The particles having twin structure are partially
dron.24,25 Starting from an fcc structured tetrahedron, a deca- diffracting the electrons into the angular range selected by the
hedron is assembled from five tetrahedrals sharing an edge objective aperture and the remaining parts show darker contrast.
(Figure 5c). If the observation direction is along the 5-fold axis The high magnification image clearly shows the twinned
and in an ideal situation, each tetrahedron shares an angle of structure and the strain induced by the twin structure (Figure
70.5, five of them can only occupy a total of 352.6, leaving 10a), and the low magnification dark-field image unambiguously
a 7.4 gap. Therefore, strain must be induced in the particle to illustrates the distribution of twins (Figure 10b).
fill the gap.26 An icosahedron is assembled using 20 tetrahedra 3.3. Surface Defects. Simply speaking, surface atoms have
via sharing an apex (Figure 5d). The icosahedral and decahedral less bonds in comparison to the atoms in the bulk because of
particles are the most extensively studied twinned nanocrys- the loss in nearest neighbors. The surface atoms tend to find
tals.27,28 The easiest orientation for identifying the MTPs is along new equilibrium positions to balance the forces, resulting in
the 5-fold symmetry axis. surface reconstruction and defects. Figure 11 gives a group of
Figure 9 shows two MTP Au particles, where the 5-fold axis images recorded from Pt nanocrystals. The {111} faces in parts
is clearly seen. The particle given in Figure 9a is a single a, c, and d of Figure 11 are relatively flat, while the {100} types
decahedral particle, while the one given in Figure 9b is a particle of faces show a large deviation from the perfect lattice structure,
Feature Article J. Phys. Chem. B, Vol. 104, No. 6, 2000 1161

suggesting the occurrence of a nonordered surface reconstruc-

tion. Surface relaxation is clearly seen in Figure 11b in the layers
indicated by arrowheads. Nanocrystals have larger surface
energy and the surface defects are believed to be abundant. The
surface atoms are likely to have dangling bonds, resulting in
higher surface reactivity. The {110} surface is rarely seen
possibly because of higher surface energy. The Au {110} surface
in nanoparticles has been found to be reconstructed so that the
surface is composed of corners bound by {100} mini-facets,
e.g., the Au atoms are missing every other row parallel to the
[110] direction.30
3.4 Imaging of Ultrasmall Particles. High-resolution TEM
image of crystalline specimens is the result of electron wave
interference among diffracted Bragg peaks. If the number of
atoms in a particle is so small that no significant diffraction
will be generated, high-resolution TEM is unlikely to reveal
the intrinsic structure of the cluster. Therefore, the question is
how small can a particle be before HRTEM becomes ineffective
in identifying its shape? To answer this question, we have tried
to image the Pt atom clusters that were nucleated at the
beginning of the growth. A lower magnification image of the
particles dispersed on a carbon substrate shows the coexistence
of large size as well as small size particles (Figure 12a), where
the smaller particles are dominated by tetrahedra and truncated
octahedra, while the large particles are dominated by cubic. This
shape transformation from tetrahedral-to-octahedral-to-cubic is
the result of the kinetically controlled growth.31 Small size
clusters of truncated octahedra can be imaged (Figure 12b). The
number of atoms presented in these particles may be 225 and
79 provided the particle shapes are perfect.
3.5. Faceted Shape of Rodlike Nanocrystals. Nanocrystals
can have a wide range of shapes. The gold nanorods recently
synthesized by a reverse miscell technique are rather unique.32
The rod is bounded by facets and it is a thermodynamically
unstable crystal growth. Almost all of the Au nanorods are single
crystalline and contain no twins or dislocations.6 When the liquid
droplet containing the suspended Au nanorods was dispersed
onto a flat carbon substrate, the nanorods were preferentially
aligned and oriented along [110] perpendicular to the substrate
(Figure 13a). The single Au nanorods also prefer to orient along
Figure 12. HRTEM image of Pt nanocrystals whose shapes were
[110] (Figure 13b). This is possible only if the nanorods have dominated by tetrahedra when they are small, but a transformation to
{110} facets. A dark-field TEM image recorded using a {111} cubic shape occurred as they grow larger. (b) The possible smallest
reflected beam when the rod is oriented nearly parallel to the nanocrystals whose shape still can be identified by HRTEM. The
[110] direction gives the thickness fringes (Figure 13c) owing numbers represents the possible of numbers of atoms in the particles
to thickness variation across the specimen. The intervals between if they are perfect truncated octahedra without point defects.
the fringes would have an equal distance if the nanorods were
bound only by four {100} facets. The finite width of the band provided by TEM. For particles larger than 5 nm, the beam
at the center and its uniformity in intensity indicate that the effect is quite small.
thickness at the center region is nearly constant with a width of
5 nm, strongly supporting the presence of {110} facets. The 4. Structures of Self-Assembled Nanocrystal Arrays
axial growth direction is [001], and its cross section is given in Size- and even shape-controlled nanoparticles behave like
Figure 13d. molecules that can be taken as building blocks for constructing
3.6. Electron Beam Effect. In HRTEM, nanocrystals are two- and three-dimensional cluster self-assembled superlattice
illuminated by an intensive electron beam; thus, radiation structures.35-39 Nanocrystal arrays are characterized by unprec-
damage is inevitable. Radiation damage can be due to ionization edented size uniformity and translation and even orientation
of the atoms, breaking bonds, local heating, and knock-on order. The nanocrystals are passivated by a monolayer of long-
displacements. Which of these takes the dominant role depends chain organic molecules, such as SR, where R ) n-CnH2n+1, n
on the material and the accelerating voltage, the electron dose, ) 4, 6, 8, 12 ..., called thiolates, that serve not only as the
and the dose rate of the microscope. For particles smaller than protection for the particles to avoid direct contact between the
5 nm, the electron beam can result in structural transformation particles with a consequence of coalescing but also the
and/or rotation of nanocrystals,33,34 and these effects are interparticle bonding. By changing the length of the molecular
enhanced as the particle size decreases. As we know that chains, quantum transitions and insulator to conductor transition
radiation damage is inevitable in TEM, one must learn how to could be introduced, resulting in tunable electronic, optical, and
identify the true structure of the nanocrystal from the information transport properties.40
1162 J. Phys. Chem. B, Vol. 104, No. 6, 2000 Feature Article

Figure 13. (a) Bright-field TEM image of Au nanorods and the corresponding electron diffraction pattern recorded from the first three rods
aligned on the left, proving the unified [110] orientation of the rods. (b) Bright-field TEM image of individually dispersed Au nanorods, showing
the [110] orientation perpendicular to the substrate. (c) Dark-field TEM image of an Au rod oriented nearly parallel to [110]. These images
unambiguously suggest the formation of {110} facets on the nanorods. (d) The cross section of an Au rod on the substrate.

4.1. Particle Shape in Self-Assembly. If the nanocrystals facets are imaged edge-on. The unit cell of the NCS is also
can be taken as the building blocks, their 3-D assembly is oriented along [110]s of fcc, where the subscript refers to the
unavoidably affected by the particle shape.41,36 TEM is unique supercrystal. Therefore, the orientational relationship between
in determination of the relationship between particle orientations the Ag particles and the nanocrystal lattice is [110] || [110]s
and the assembling crystallography of the ordered array. Figure and [001] || [11h0]s. This registered orientational order can be
14a is an TEM image recorded from an Ag nanocrystal used to determined the orientations of the nanocrystals in the
superlattice (NCS) deposited on a carbon substrate, from which unit cell of the superlattice, as shown in Figure 14c. The
the orientation of the particles and the arrays can be identified. nanocrystals are packed into a face-centered cubic lattice, and
The Ag nanocrystals have a truncated octahedral shape (Figure the nanocrystals are arranged following a principle of face-to-
14b), and they are oriented along the [110] of the Ag atomic face [10]. The image recorded from the same area at a slightly
lattice in the image, along which four {111} and two {100} large defocus condition gives the channels (the white dots in
Feature Article J. Phys. Chem. B, Vol. 104, No. 6, 2000 1163

Figure 14. (a) TEM image of a face-centered cubic packed Ag nanocrystals whose shape is dominated by truncated octahedral (TO) (see the inset
HRTEM image of a single Ag nanocrystal). The image displays the orientational relationship between the nanocrystals and the assembled superlattice.
(b) A TEM image recorded from the same specimen area under slightly defocused condition. The image in (a) is recorded at near focus condition,
and it is sensitive to the particle shape, and the image in (b) is most sensitive to the phases caused by a variation in the projected mass thickness.
(c) A structural model of the TO particle and the [110]s projection of the unit cell as indicated in (a), where the gray pasted particles are located
at different z heights. The channels enclosed by the bundled thiolates are presented.
1164 J. Phys. Chem. B, Vol. 104, No. 6, 2000 Feature Article

Figure 15. (a) TEM image of a monolayer self-assembly of tetrahedral Ag nanocrystals. (b and c) Two possible assemblies of tetrahedra with
short-range orientational order. A small-angle electron diffraction pattern from the monolayer assembly, reflecting the mirror symmetry (not 3-fold)
(The Ag nanocrystals were supplied by R. L. Whetten and S. A. Harfenist).

Figure 14b) enclosed by the bundling distributed thiolates nanocrystals.44 The bright-field TEM images (parts a and c of
molecules on the surface of the nanocrystals. This model has Figure 16) illustrate that the particles have triangular-like
been proved by the energy-filtered chemical imaging technique projected shapes. The projected triangles may correspond to
in TEM.42 either tetrahedral or platelet-like structure. If the latter were the
The self-assembly of nanocrystals is strongly affected by the case, the dark-field image recorded using the diffracted electrons
shape of the nanocrystals. The tetrahedral shaped nanocrystals would show a uniform contrast across the particle because there
have sharp corners and their assembly may have long-range is no variation in the projected thickness across the particle. In
translational symmetry but the orientational order is limited only contrast, the dark-field image (Figure 16b) shows bright spots
to a short range.43 Figure 15a gives the monolayer self-assembly at the center of some particles, indicating the tetrahedral shape
of tetrahedral Ag nanocrystals. Two fundamental principles are because the dark-field image is sensitive to the projected density
usually observed in particle self-assembling. The most dense of the atoms under the first-order scattering approximation. On
packing is the favorable configuration from the energy point of the other hand, the bright-field image given in Figure 16c cannot
view. The passivation molecules are likely to form interdigitative uniquely define the particle shape.
bonds that link the nanocrystals. Two possible monolayer Dark-field imaging can also be useful for identifying the
assembly models satisfying the two requirements are given in number of stacking layers in particle self-assembling.45 Figure
parts b and c of Figure 15. The only difference between the 17 is such an example, which was recorded by positioning a
two models is that half of the particles face down, while the small size objective aperture on the optic axis while the incident
projected structures of the two are identical. These models do beam is scanned conically to average over the particle orienta-
not support the 3-fold symmetry of the packing due to the shape tions. This type of image is called conical scan dark-field
of the particles, in agreement with the mirror symmetry imaging and it is most sensitive to the local projected mass
presented by the electron diffraction pattern (the inset in Figure thickness, and the diffraction effect is minimized. The monolayer
15a). The regions indicated by parenthesis in Figure 15a agree assembly preserves very well the translational symmetry. The
well with the models. second layer introduces a minor short-range disorder, while the
4.2 Dark-Field Imaging in Identification of Particle Shape. third layer is largely disordered. The reason for the increased
Particle shapes can be identified by high-resolution TEM images, disordered structure is likely due to the introduction of larger
but in some cases the dark-field imaging is indispensable for size particles, resulting in the creation of interstitial and
confirming the result. Figure 16 shows such an example of CoO vacancy point defects in the arrays. The contrast of this type of
Feature Article J. Phys. Chem. B, Vol. 104, No. 6, 2000 1165

Figure 16. (a, c) Bright-field and (b) dark-field TEM images of monolayer self-assembly of tetrahedral CoO nanocrystals of 4 nm in size,
displaying short-range translational ordering but no orientational order. An electron diffraction pattern from the nanocrystals is shown in the inset,
exhibiting broadened rings.

image is proportional to the specimen thickness for relatively assembled array may not prove the tetrahedral shape of the
thin specimens, allowing for a direct identification of the number nanocrystals. Figure 18a shows a bright-field TEM image
of stacking layers from the image. recorded from the ordered array of Ag nanocrystals. The image
We have shown in Figure 16c that tetrahedral nanocrystals clearly shows some triangular-shape contrast, in which the
gives triangular projections in the TEM image. However, a triangles are as if representing the nanocrystals (in dark contrast).
triangular pattern observed in the bright-field image in a self- A conical scan dark-field image recorded from the same region
1166 J. Phys. Chem. B, Vol. 104, No. 6, 2000 Feature Article

high does the annealing temperature need to be to remove the

capping polymer while the particle shape is still preserved?;
(2) to what temperature is the particle shape still stable?; (3)
how high is the melting point of the Pt particles? In situ TEM
observation of the particle shape transformation and melting
can provide answers to these questions.
Figure 19 shows a series of TEM images recorded from the
same region when the specimen temperature was increased from
25 to 610 C. These images were selected from a group of
images to present the most significant changes in the particle
shapes. For easy notation, particles are labeled as groups to trace
their shape transformation behavior. Most of the particle shapes
showed no significant change when the specimen temperature
was below 350 C (parts a-c of Figure 19). Truncated cubic
and tetrahedral particles were formed when the temperature
arrived at 410 C (Figure 19d). The corners and edges of the
particles were disappearing because the local atoms have higher
energy. The tetrahedral particles could still be identified even
when the temperature reached 500 C (Figure 19e), while the
Figure 17. Dark-field TEM image recorded by scanning the incident cubic particles became spherical when the temperature was
beam conically, showing the monolayer, double-layer, and triple-layer above 500 C. This indicates that the tetrahedral particles are
self-assembly of Ag nanocrystals (The Ag nanocrystals were supplied more stable than cubic ones possibly because the {111} surfaces
by S. A. Harfenist). have lower surface energy than the {100}. A feature observed
in Figure 19 is that the tetrahedral shape of a particle indicated
of the specimen, in which the particles show bright contrast, by an arrowhead was preserved when the specimen temperature
proves that the particles have a nearly spherical shape (Figure was as high as 610 C. This is possibly due to the contact of
18b) because the bright contrast area in this image usually the apexes of the particle with the adjacent particles so that the
represents the objects with a larger mass thickness; the triangles interparticle diffusion can still sustain the shape of the apexes.
are in darker contrast corresponding to the cavities among the Our results indicate that the surface capping polymer is removed
particles. Thus, the particles cannot be tetrahedra. To explain by annealing the specimen to a temperature of 180-250 C,
this imaging phenomenon, a hexagonal close packed A-B- while the particle shape shows no change up to 350 C. In
A-A-B stacking sequence is created for an odd number of the temperature range 350-450 C, a small truncation occurs
layers (Figure 18c). The stacked layers create a channel of close in the particle shape but no major shape transformation. The
to a triangular shape if the projected mass is considered, which particle shape experiences a dramatic transformation into a
corresponds to the darker contrast triangle observed in the dark- spherical-like shape when the temperature is higher than 500
field image. Image simulations have confirmed this type of C (the melting point of bulk Pt is 1773 C).
interpretation.46 Alternatively, an examination of the dispersed The thermodynamic properties of nanocrystals depend strongly
particles can also provide additional proof about the particle on not only the crystal size but also their shapes. The
shape. thermodynamic properties of Au nanorods is an example.6 No
structural change is observed at temperatures below 215 C.
5. In Situ Temperature-Induced Structural Evolution of However, some tiny Au clusters appeared on the substrate as
Nanocrystals the temperature arrived at 230 C (Figure 20c). The clusters
In situ study of the temperature-induced phase transformation grew rapidly as temperature increased to 350 C, above which
and structural and chemical evolution of nanocrystals is the clusters shrank in size (parts f-h of Figure 20). The Au
important for understanding the structure and structural stability clusters are believed to be formed by the atoms sublimated from
of nanomaterials. TEM is an ideal approach for conducting these the nanorod surfaces due to the presence of the high energy
types of experiments, in which a specimen can be cooled to {110} faces.6 Since the supply of the atoms to the cluster is
the liquid nitrogen or liquid helium temperatures or heated to proportional to its contact length with the substrate (proportional
1000 C. The in situ process can be recorded at TV rate for to the particle radius r) and the sublimation is proportional to
exhibiting the time and temperature-dependent phenomena. the surface area of the cluster (r2), the clusters grew faster when
The large percentage of surface atoms in nanocrystals is the they are small because of the dominant r term. The shrinkage
origin of their unique properties. The melting of Pt particles42 of the clusters at higher temperatures is due to the increased
and the surface sublimation of Au nanorods6 are taken as sublimation rate. An interesting phenomenon observed in Figure
examples here to illustrate the unusual thermodynamic properties 20 is that the nanorod indicated by an arrowhead shrank as
of nanocrystals. Platinum nanoparticles with a high percentage temperature went up, but it regrew at sufficiently high temper-
of cubic-, tetrahedral-, and octahedral-like shapes, respectively, ature. This reversed atom diffusion is likely to be associated
have been synthesized by changing the ratio of the concentration with its interaction with the neighboring rods.
of polymer capping material (polyacrylate) to that of Pt2+ being To reveal the unique surface processes associated with Au
reduced by H2 from K2PtCl4 at room temperature.14,47 The nanorods, the rods were photoisomerized to form gold nano-
polymer acts not only as the passivative protection layer of the spheres by the influences of femtosecond laser pulses.48,49 Both
nanocrystals but also as the key factor in controlling their shapes. the nanospheres and nanorods have the same average volume
To determine the shape-dependent catalytic behavior of Pt and almost equally exposed surface areas (with consideration
nanocrystals, the passivation layer on the nanocrystal must be of the contact of the particle with the substrate); thus, the size
removed. Several questions remain to be considered: (1) how effect of the two geometrical shapes should not make a
Feature Article J. Phys. Chem. B, Vol. 104, No. 6, 2000 1167

Figure 18. (a) Bright-field and (b) conical-scan dark-field TEM images recorded from same region of a multilayer self-assembly of Ag nanocrystals.
(c) A model for interpretation of the image contrast.
significant difference. After being heated to as high as 650 C, addition to the {100} and {111} facets present in the spheres.
no clusters were formed near the gold nanospheres. This clearly The presence of a significant percentage of gold atoms on the
shows that the newly formed Au particles are unique charac- {110} surface suggests that the nanorod is relatively thermo-
teristics of Au nanorods. Using HRTEM, it has found that the dynamically unstable. The {111} and {100} faceted Au spheres
nanorods have the unstable {110} facets (see Figure 13), in have lower surface energy. Thus, their atoms may not sublime
1168 J. Phys. Chem. B, Vol. 104, No. 6, 2000 Feature Article

Figure 19. A series of TEM images recorded in situ from Pt nanocrystals dispersed on a carbon substrate and being heated to different temperatures,
exhibiting the shape transformation as well as surface melting phenomenon of the particles at temperatures much lower than the melting point of
the bulk (The Pt nanocrystals were supplied by J. Petroski and M. A. El-Sayed).
at relatively lower temperatures. Therefore, the sublimation of In situ TEM is also a powerful approach to examine the
Au atoms from the nanorods at lower temperatures is due to structural evolution of nanocrystal self-assembled arrays. The
the shape of the nanorods, which has the unstable {110} ordered assembly of Ag nanocrystals of 5 nm in core size
surfaces. preserves the order to a temperature as high as 500 C, much
Feature Article J. Phys. Chem. B, Vol. 104, No. 6, 2000 1169

Figure 20. TEM images of the Au nanorods recorded in situ at different temperatures, illustrating the nucleation and growth of Au clusters on the
substrate. The Au clusters are the condensation of the sublimated Au atoms from the nanorods (The Au nanocrystals were supplied by M. Mohamedi
and M. A. El-Sayed).

higher than the expected temperature.50 This is likely due to their physical properties could be quite diverse. An essential
the strengthened binding force owing to the formation of the task in nanoscience is the property associated characterization
interdigitative molecular bonds of the adsorbed molecules on of an individual nanostructure with well-defined atomic struc-
the nanocrystal surfaces. Monolayer self-assembly has been ture. But as limited by the size of the nanostructures, new
found to exhibit the highest thermal stability,51 possibly because methods and methodologies must be developed to manipulate
of the strong effect from the carbon substrate. and quantify the properties of individual nanostructures. It is
important to directly view the nanostructure while a testing or
6. Measuring the Properties of Individual Nanostructure measurement is being carried out.
by In Situ TEM We have recently applied TEM as a means of nanomeasure-
There are four key steps in the development of nanoscience ment in characterizing the mechanical strength of a single carbon
and nanotechnology: materials preparation, property charac- nanotube.52 Carbon nanotube has a nanofiber-like structure
terization, device fabrication, and system integration. Because consisting of concentric cylindrical graphitic sheets, with
of the high size and structure selectivity of the nanomaterials, diameters of 5-50 nm and lengths of 1-20 m. A TEM
1170 J. Phys. Chem. B, Vol. 104, No. 6, 2000 Feature Article

Figure 22. A small particle attached at the end of a carbon nanotube

at (a) stationary and (b) the first harmonic resonance ( ) 0.968 MHz).
The mass of the particle was determined to be 22 fg (in collaboration
Figure 21. Alternating field-induced mechanical resonance. A selected with P. Poncharal and W. A. de Heer).
carbon nanotube at (a) stationary and (b) the first harmonic resonance
( ) 1.21 MHz) (in collaboration with P. Poncharal and W. A. de 7. Chemical, Electronic, and Bonding Structures of
Heer). Nanocrystals

specimen holder was specially built for applying a voltage across 7.1. Chemical Microanalysis by EDS and EELS. HRTEM
the nanotube and its counter electrode. Each single nanotube is a powerful technique for identifying the crystal structure of
can be clearly observed in the TEM. Thus, the measurements nanophase materials, but it may not be sensitive enough to detect
can be done on a specific nanotube whose microstructure is a small change in atomic number. To precisely determine the
determined by transmission electron imaging and diffraction. content of a particular element in a specimen, energy dispersed
If an alternating voltage is applied on the nanotube with an X-ray spectroscopy and electron energy loss spectroscopy can
ability to tune the frequency of the applied voltage, resonance be very useful. Under the impact of an incident electron, the
can be induced when the frequency of the applied voltage equals electrons bounded to the atoms may be excited either to a free
the natural frequency of the nanotube (Figure 21). Thus, the electron state or to an unoccupied energy level with a higher
bending modulus of nanotubes can be calculated from the energy. The quantum transitions associated with these excitations
resonance frequency. This is likely to be the most precise will emit photons (e.g., X-rays) and Auger electrons. These
method for characterizing the mechanical properties of a single inelastic scattering signals are the fingerprints of the elements
fiberlike structure, providing a simple approach toward nano- that can provide quantitative chemical and electronic structural
mechanics. information. Two analytical techniques are most commonly used
Similar to a cantilevel pendulum and using the carbon in chemical microanalysis in TEM, EDS,54 and EELS.12 The
nanotube with a calibrated bending modulus, the mass of a tiny size of the incident electron probe can be as small as 2-3 nm,
particle attached to the end of the nanotube can be determined giving the feasibility of acquiring EDS and EELS spectra from
from the vibration frequency. For the case shown in Figure 22, individual nanocrystals. The data can be used for quantitatively
the mass of the particle is determined to be 22 ( 6 fg (1 fg ) for determining the chemical compositions of nanoparticles.
10-15 g). This is the newly discovered nanobalance, the most EDS is mainly sensitive to heavy elements due to the fluores-
sensitive and smallest balance in the world.52 We anticipate that cence effect arising from Auger emission for light elements,
this nanobalance will have applications in measuring the mass while EELS is most adequate for light elements. Both EDS and
of large biomolecules and biomedical particles, such as virus, EELS can be applied complementary to determine the chemical
possibly leading to a nanomass spectrometer. Using the in situ composition. The detection limit for EDS can be as high as
TEM specimen holder we built, the ballistic quantum conduc- 1%, and that for EELS can be as high as 5% depending on
tance in defect-free carbon nanotubes53 has been repeated, and elements.
the characteristics of electron field emission from individual 7.2. Near Edge Fine Structure and Electronic Structure.
carbon nanotube is under investigation. EELS is not only a powerful tool for microanalysis but can
Feature Article J. Phys. Chem. B, Vol. 104, No. 6, 2000 1171

Figure 24. Fractions of the Mn2+ and Mn3+ ions in MnFe2O4 measured
by quantitatively fitting the experimental EELS spectra with the standard
spectra of the oxides containing Mn2+ and Mn3+ ions. Five repeated
Figure 23. A comparison of EELS spectra of Co-L2,3 ionization edges measurements at 25 C are shown and give consistent result.
acquired from Co3O4 and CoO standard specimens and the synthesized
nanocrystals, proving that the valence state of Co is 2+ in the newly Analysis of the valence states of Mn has also been carried
prepared nanocrystals. out for MnFe2O4 spaniel structured nanocrystals. The AB2O4
also provide electronic structural information. In the ionization type of spaniel structure has two types of cation lattice sites: a
edge observed in EELS, fine peaks present near the threshold tetrahedral site A2+ formed by four nearest-neighbor oxygen
represent the density of states in the conduction band. The anions, and an octahedral B3+ site formed by six oxygen anions.
intensities and energy position of the peaks are sensitive to the In MnFe2O4, the percentage of the A sites occupied by Fe
solid state structure of the material. The L ionization edges of specifies the degree of valence inversion. For a general case,
2+
transition metal and rare earth elements, for example, usually the ionic structure of MnFe2O4 is written as (Mn1-x Fe2+
x )
3+ 3+
display sharp peaks at the near edge region, which are known (Fe1-yMny )O4, in which the A and B sites can be occupied by
as white lines. For transition metals with unoccupied 3d states, either Mn or Fe. An experimental measurement of the valence
the transition of an electron from 2p state to 3d levels leads to conversion of Mn in this material can provide concrete informa-
the formation of white lines. The L3 and L2 lines are the tion on the distribution of Fe in the A and B sites, possibly
transitions from 2p3/2 to 3d3/23d5/2 and from 2p1/2 to 3d3/2, leading to a better understanding on its magnetic property.62
respectively, and their intensities are related to the unoccupied Shown in Figure 24 is the EELS measured fractions of the Mn2+
states in the 3d bands. and Mn3+ ions in the MnFe2O4 specimen as a function of the
Numerous EELS experiments have shown that a change in in situ specimen temperature in TEM. The fraction was
valence state of cations introduces a dramatic change in the ratio calculated by fitting the experimentally observed L3 and L2
of the white lines, leading to the possibility of identifying the EELS spectra by a linear summation of the spectra acquired
occupation number of 3d orbital using EELS. The most from MnO and Mn2O3, and the coefficients for the linear
successful examples are Co and Mn oxides.55-59 The oxidation combination give the percentages of the Mn ions of different
states of Fe can also be determined, but the sensitivity is much valence states in the material. It is clear that the fractions of
lower.60-62 The oxidation states of Ce and Pr have been Mn2+ and Mn3+ ions at room temperature is 0.5:0.5, while a
determined in an orthophosphate material in which the con- complete conversion into divalent Mn occurs at 600 C. These
stituents of Ce and Pr are on the order of 100 ppm.63 data explicitly illustrate the evolution in the valence state of
The determination of the crystal structure of nanoparticles is the Mn ions, leading to the temperature-dependent magnetic
a challenge particularly when the particles are smaller than 5 properties of MnFe2O4.
nm. The intensity maxima observed in the X-ray or electron 7.3. Chemical Mapping and Valence State Mapping. High
diffraction patterns of such small particles are broadened due spatial resolution mapping on the elemental distribution has vital
to the crystal shape factor (see the diffraction pattern from the importance in materials analysis. This can be carried in the
CoO nanocrystal inserted in Figure 16), greatly reducing the scanning mode by collecting EDS/EELS signal at each scanning
accuracy of structure refinement. On the other hand, the quality position of the beam.63,64 A detailed data processing provides a
of the high-resolution TEM images of the particles is degraded distribution map of the elements. The spatial resolution is mostly
because of the strong effect from the substrate. This difficulty determined by the size of the electron probe. In TEM, images
arises in the study of CoO nanocrystals whose shape is can be formed by the electrons that have lost a specific energy
dominated by tetrahedral of sizes smaller than 5 nm, but EELS in correspondence to an energy threshold of the atomic inner
can help to solve this problem. Figure 23 gives the EELS spectra shell. This is the energy-filtered transmission electron micros-
acquired from the as-synthesized Co-O nanocrystals, and the copy (EF-TEM).67,68 With consideration of the fact that the inner
bulk Co3O4 and CoO specimens. The spectrum of the nano- shell excitation of an atom occurs only when the impact
particles fits very well with that of CoO, while that of Co3O4 is parameter of an incident electron is less than 0.05 nm, this very
distinctly different, unambiguously proving the CoO structure localized scattering characteristic is the origin of mapping the
of the nanocrystals.44 distribution of the element that gives rise to the inner shell
1172 J. Phys. Chem. B, Vol. 104, No. 6, 2000 Feature Article

Figure 25. (a) Bright-field and (b) Si-L edge energy filtered TEM images of a TiO2 particle coated with silica.

excitation. This technique has been applied to map the distribu- states (Figure 26e), where the diffraction contrast disappears.
tion of thiolates in the self-assembled Ag nanocrystals.42 The The region with lower oxidation state (Co2+) shows stronger
result proves that the adsorbed molecules are distributed on the contrast, and the ones with high oxidation states show weaker
surfaces of the nanocrystals and they form interdigitative contrast. The O/Co image was calculated from the images
bundling that holds the nanocrystal together. recorded from the O-K edge and the L3 + L2 white lines for an
Figure 25a shows a bright-field TEM image recorded from a energy window width of ) 24 eV. Although the energy-
TiO2 specimen coated with 1 at. % silica and embedded in filtered O-K edge image exhibits some diffraction contrast and
epoxy. The amorphous layer of thickness 4 nm coated on the the thickness effect, the O/Co compositional ratio image greatly
particle surface cannot be easily claimed as silica from this reduces the effect. The high intensity region in the O/Co image
image. Using the Si-L edge, the energy-filtered image confirms indicates the relative high local concentration in oxygen (e.g.,
that the layer is indeed silica (Figure 25b). The width of the higher Co oxidation states), the low intensity region contains
silica layer in the Si-L edge filtered image is significantly wider relatively less oxygen (e.g., lower Co valence state), entirely
than that in the bright field image. This is due to a deterioration consistent with the information provided by the L3/L2 image.
in image resolution by chromatic aberration because of the 20 A spatial resolution of 2 nm has been achieved. This is
eV wide energy window. remarkable in comparison to any existing techniques.
We have illustrated the sensitivity of near edge fine structures
to the valence states of transition metal elements. Energy-filtered 8. Holography Imaging of Charged and Magnetic
imaging by selecting the electrons with an energy loss in Nanocrystals
correspondence to the near edge structure contains the spatial
distribution of the valence states across the region. A typical HRTEM is powerful in resolving the atomic scale micro-
example is the mapping of SP2 and SP3 bonds in diamond-like structure of nanocrystals, but it is insensitive to the electrostatic
materials.69,70 Using the images acquired from the white lines, charging or magnetic domain of a nanocrystal because the
an experimental approach has been demonstrated to map the electron phase caused by the electromagnetic phenomena is
valence state distributions of Mn and Co using the energy- lost in the recording of the image. The phase image can
filtered TEM.69 Figure 23 clearly shows the strong dependence provide the distribution of electrostatic field around a charged
of the relative intensities of the Co-L3 and Co-L2 white lines particle as well as the thickness-projected potential image of a
on the valence state of cobalt. In EF-TEM, if an energy window nanocrystal. The former can be used to extract the charge
of 12 eV in width is positioned at the L3 and the L2 white lines, distribution in the particle,72 and the latter is useful to determine
the energy filtered images using the Co-L3 and Co-L2 lines can whether the particle contains a hard core or an empty cavity.73
be obtained after subtraction of background. The ratio L3/L2 Mapping the electron phase around a magnetic particle can
image gives the distribution of valence states across the reveal the magnetic domain structure in a nanocrystal.74,75 The
specimen.71 mapping of electron phase is based on an electron holography
Figure 26 shows a group of energy-filtered TEM images from technique.
a triple point in the CoO-Co3O4 specimen. The energy-filtered The principle of electron holography was first introduced by
images using the L3 and L2 lines and the post-L2 line region Gabor.76 Holography is based on the interference and diffraction
(parts a-c of Figure 26) show a distinct difference in contrast properties of waves, thereby producing a true image of an object
distribution due to a difference in the relative white line (including amplitude and phase) without any distortion of the
intensities. From these three images, the L3/L2 is calculated after lenses. The development of high-brightness, high-coherence
subtracting the contribution from the continuous energy loss electron sources has made it possible to obtain holograms using
region that is due to single atom scattering, and the image clearly electron waves in TEM.77-79 The key application of electron
displays the distribution of cobalt oxides having different valence holography is studying magnetic materials. A key step in
Feature Article J. Phys. Chem. B, Vol. 104, No. 6, 2000 1173

Figure 26. A group of images recorded from the same specimen region using signals of (a) the Bright-field, (b) the Co-L3 edge, and (c) the Co-L2
edge. (d) The processed L3/L2 image, displaying the distribution of the valence states. (e) Oxygen map acquired by the O-K edge. (f) The atomic
concentration ratio image of O/Co. The continuous background contributed from the single atom scattering has been removed from the displayed
Co-L3 and Co-L2 images. The O/Co image is normalized in reference to the standard composition of CoO for the low portion of the image in order
to eliminate the strong influence on the ionization cross section from the white lines. Each raw image was acquired with an energy window width
of ) 12 eV except for O-K at ) 24 eV.
observation of magnetic flux lines is to create a magnetic-field- can be achieved either by switching off the objective lens or
free zone in which the specimen to be examined is placed. This by using a Lorentz objective lens. In the former, the experiments
1174 J. Phys. Chem. B, Vol. 104, No. 6, 2000 Feature Article

can be performed using a conventional TEM equipped with an (17) Wang, Z. L. Elastic and Inelastic Scattering in Electron Diffraction
FEG at the expense of image magnification, which is usually and Imaging; Plenum Press: New York, 1995; Chapters 2-5.
(18) Cowley, J. M.; Moodie, A. F. Acta Crystallogr. 1957, 10, 609.
reduced by 50-80 times. The image resolution is also greatly (19) Ascencio, J. A.; Gutierrez-Wing, C.; Espinosa, M. E.; Marin, M.;
reduced due to the aberration effect of the intermediate lenses. Tehuacanero, S.; Zorrilla, C.; Jose-Yacaman, M. Surf. Sci. 1998, 396, 349.
Thus, the technique can only be applied to image magnetic (20) Smith D. J.; Marks, L. D. Philos. Mag. A 1981, 44, 735.
(21) Rampino, L. D.; Nord, F. F. J. Am. Chem. Soc. 1942, 63, 2745.
domains of large sizes. (22) Henglein, A.; Ershov, B. G.; Marlow, M. J. Phys. Chem. 1995,
99, 14129.
9. Summary (23) Wang, Z. L.; Ahmadi, T. S.; El-Sayed, M. A. Surf. Sci. 1996, 380,
302.
Controlling the shape of nanocrystals is the-state-of-the-art (24) Ino, S. J. Phys. Soc. Jpn. 1966, 21, 346.
in nanophase materials research. The shape-controlled nano- (25) Allpress, J. G.; Sanders, J. V. Surf. Sci. 1967, 7, 1.
crystals have many prospects that are likely to impact the fields (26) Marks, L. D. Rep. Prog. Phys. 1994, 57, 603 and references therein.
(27) Yang, C. Y. J. Cryst. Growth 1979, 47, 274; 1979, 47, 283.
of catalysis, self-assembly, and nanodevices. An important (28) Buffat, P.-A.; Flueli, M.; Spycher, R.; Stadelmann, P.; Borel, J.-P.
technique for characterizing the shape-controlled nanocrystals Faraday Discuss. 1991, 92, 173.
is transmission electron microscopy. The objective of this article (29) Cleveland, C. L.; Landman, U.; Schaaff, T. G.; Shafigullin, M. N.;
Stephens, P. W.; Whetten, R. L. Phys. ReV. Lett. 1997, 79, 1873.
is to introduce TEM techniques for characterization of the shape- (30) Ikarashi, N.; Kobayashi, K.; Koike, H.; Hasegawa, H.; Yagi, K.
controlled nanocrystals. The most important application of TEM Ultramicroscopy 1988, 26, 195.
is to define the shape of nanocrystals. This type of imaging (31) Petroski, J. M.; Wang, Z. L.; Green, T. C.; El-Sayed, M. A. J.
capability is well received by the physical chemistry community. Phys. Chem. B 1998, 10, 3316.
(32) Yu, Y.-Y.; Chang, S.-S.; Lee, C.-L.; Wang, C. R. J. Phys. Chem.
In additional to high-resolution lattice imaging, TEM is B 1997, 101, 6661.
powerful in identifying and quantifying the chemical and (33) Iijima, S.; Ichihashi, T. Phys. ReV. Lett. 1986, 56, 616.
electronic structure of nanomaterials. High spatial resolution (34) Smith, D. J.; Petford-Long, A. K.; Wallenberg, L. R.; Bovin, J.-O.
Science 1986, 233, 872.
analysis is vitally important for solving many of the practical (35) Whetten, R. L.; Khoury, J. T.; Alvarez, M. M.; Murthy, S.; Vezmar,
problems of nanomaterials. Spectroscopy analysis of the solid I.; Wang, Z. L.; Cleveland, C. C.; Luedtke, W. D.; Landman, U. AdV. Mater.
state effects and the valence states mapping are new directions 1996, 8, 428.
of quantitative microscopy. Further exploration in the fine (36) Wang, Z. L. AdV. Mater. 1998, 10, 13.
(37) Motte, L.; Billoudet, F.; Lacaze, E.; Douin, J.; Pileni, M.-P. J. Phys.
structures in EELS is likely to reveal rich information about Chem B 1997, 101, 138.
bonding in nanocrystals. In situ TEM is anticipated to be (38) Andres, R. P.; Bein, T.; Dorogi, M.; Feng, S.; Henderson, J. I.;
important for characterizing and measuring the properties of Kubiak, C. P.; Mahoney, W.; Osifchin, R. G.; Reifenberger, R. Science
1996, 272, 1323.
individual nanostructures, from which the structure-property (39) Murray, C. B.; Kagan, C. R.; Bawendi, M. G. Science 1995, 270,
relationship can be clearly registered to a specific nanoparticle/ 1335.
structure. This is a new growing field of nanomeasurements. (40) Collier, C. P.; Saykally, R. J.; Shiang, J. J.; Henrichs, S. E.; Heath,
J. R. Science 1997, 277, 1978.
(41) Harfenist, S. A.; Wang, Z. L.; Alvarez, M. M.; Vezmar, I.; Whetten,
Acknowledgment. Thanks to my collaborators, M. A. El- R. L. J. Phys. Chem. 1996, 100, 13904.
Sayed, W. A. de Heer, R. L. Whetten, S. A. Harfenist, Jin Song (42) Wang, Z. L.; Harfenist, S. A.; Whetten, R. L.; Bentley, J.; Evans,
Yin, Z. J. Zhang, P. Poncharal, T. Ahmadi, M. Mohamed, S. N. D. J. Phys. Chem. B 1998, 102, 3068.
Link, Janet Petroski, and T. C. Green, for their contributions to (43) Wang, Z. L.; Harfenist, S. A.; Vezmar, I.; Whetten, R. L.; Bentley,
J.; Evans, N. D. AdV. Mater. 1998, 10, 808.
the work covered in this review article. This research was (44) Yin, J. S.; Wang, Z. L. Phys. ReV. Lett. 1997, 79, 2570.
sponsored by NSF Grants CHE-9727633 and DMR-9733160. (45) Harfenist, S. A.; Wang, Z. L.; Alvarez, M. M.; Vezmar, I.; Whetten,
R. L. AdV. Mater. 1997, 9, 817.
References and Notes (46) Bentzon, M. D.; Tholen, A. Ultramicroscopy 1991, 38, 105.
(47) Ahmadi, T. S.; Wang, Z. L.; Green, T. C.; Henglein, A.; El-Sayed,
(1) Buffat, Ph.; Borel, J.-P. Phys. ReV. A 1976, 13, 2287. M. A. Chem. Mater. 1996, 8, 1161.
(2) Schmidt, M.; Kusche, R.; von Issendorff, B.; Haberland, H. Nature (48) Link, S.; Burda, C.; Mohamed, M. B.; El-Sayed, M. A. J. Phys.
1998, 393, 238. Chem. B 1999, 103, 1165.
(3) Goldstein, A. N.; Echer, C. M.; Alivisatos, A. P. Science 1992, (49) Mohamed, M. B.; Ismail, K. Z.; Link, S.; El-Sayed, M. A. J. Phys.
256, 1425. Chem. B 1999, 102, 9370
(4) de Heer, W. A. ReV. Mod. Phys. 1993, 65, 611. (50) Harfenist, S. A.; Wang, Z. L. J. Phys. Chem. B 1999, 103, 4342.
(5) Alivisatos, A. P. Science 1996, 271, 933. (51) Yin, J. S.; Wang, Z. L. J. Phys. Chem. 1997, 101, 8979.
(6) Mohamed, M.; Wang, Z. L.; El-Sayed, M. A. J. Phys. Chem. B, in (52) Poncharal, P.; Wang, Z. L.; Ugarte, D.; de Heer, W. A. Science
press. 1999, 283, 1513.
(7) Wang, Z. L.; Petroski, J.; Green, T.; El-Sayed, M. A. J. Phys. Chem. (53) Frank, S.; Poncharal, P.; Wang, Z. L.; de Heer, W. A. Science 1998,
B 1998, 102, 6145. 280, 1744.
(8) Cleveland, C.; Luedtke, W. D.; Landman, U. Phys. ReV. Lett. 1998, (54) Goldstein, J. I.; Newbury, D. E.; Echlin, P.; Joy, D. C.; Romig, A.
81, 2036. D.; Lyman, C. E.; Fiori, C.; Lifshin, E. Scanning Electron Microscopy and
(9) Wang, Z. L., Ed. Characterization of Nanophase Materials; Wiley- X-ray Microanalysis, a text for biologists, materials scientists and geologists;
VCH: New York, 2000. Plenum Press: New York, 1992.
(10) Williams, D. B.; Carter, C. B. Transmission Electron Microscopy; (55) Kurata, H.; Colliex, C. Phys. ReV. B 1993, 48, 2102.
Plenum Press: New York, 1996. (56) Mansot, J. L.; Leone, P.; Euzen, P.; Palvadeau, P. Microsc.
(11) Wang, Z. L.; Kang, Z. C. Functional and Smart Materials - Microanal. Microstruct. 1994, 5, 79.
Structural EVolution and Structure Analysis; Plenum Press: New York, (57) Wang, Z. L.; Yin, J. S.; Zhang, J. Z.; Mo, W. D. J. Phys. Chem. B
1998; Chapter 6. 1997, 101, 6793.
(12) Egerton, R. F. Electron Energy-Loss Spectroscopy in the Electron (58) Wang, Z. L.; Yin, J. S.; Jiang, Y. D.; Zhang, J. Appl. Phys. Lett.
Microscope, 2nd ed.; Plenum Press: New York, 1996. 1997, 70, 3362.
(13) Frank, S.; Poncharal, P.; Wang, Z. L.; de Heer, W. A. Science 1998, (59) Wang, Z. L.; Yin, J. S. Philos. Mag. B 1998, 77, 49.
280, 1744. Poncharal, P.; Frank, S.; Wang, Z. L.; de Heer, W. A. Proceeding (60) Colliex, C.; Manoubi, T.; Ortiz, C. Phys. ReV. B 1991, 44, 11402.
of the International Conference on Small Clusters, in press. (61) Lloyd, S. J.; Botton, G. A.; Stobbs, W. M. J. Microsc. 1995, 180,
(14) Ahmadi, T. S.; Wang, Z. L.; Green, T. C.; Henglein, A.; El-Sayed, 288.
M. A. Science 1996, 28, 1924. (62) Yuan, J.; Gu, E.; Gester, M.; Bland, J. A. C.; Brown, L. M. J.
(15) Hirsch, P. B.; Howie, A.; Nicholson, R. B.; Pashley, D. W.; Whelan, Appl. Phys. 1994, 75, 6501.
M. J. Electron Microscopy of Thin Crystals; Roberts E. Krieger Publishing (63) Fortner, J. A.; Buck, E. C. Appl. Phys. Lett. 1996, 68, 3817.
Company: New York, 1977. (64) Zhang, Z. J.; Wang, Z. L.; Chakoumakos, B. C.; Yin, J. S. J. Am.
(16) Scherzer, O. J. Appl. Phys. 1949, 20, 20. Chem. Soc. 1998, 120, 1800.
Feature Article J. Phys. Chem. B, Vol. 104, No. 6, 2000 1175

(65) Craven, A. J.; Colliex, C. J. Microsc. Spectrosc. Electron 1977, 2, (73) Datye, A. K.; Kalakkad, D. S.; Voikl, E.; Allard, L. F. In Electron
511. Holography; Tonomura, A., Allard, L. F., Pozzi, G., Joy, D. C., Ono, Y.
(66) Leapman, R. D.; Hunt, J. A. Microscopy: The Key Research Tool; A., Eds.; Elsevier Science: New York, 1995; p 199.
The Electron Microscopy Society of America, 1992; Vol. 22, p 39. (74) Mankos, M.; Cowley, J. M.; Scheinfein, M. R. Mater. Res. Soc.
(67) Reimer, L., Ed. Energy Filtering Transmission Electron Microscopy; Bull. 1995, 20, 45.
Springer Series in Optical Science; Springer-Verlag: New York, 1995. (75) Hirayama, T.; Chen, J.; Ru, Q.; Ishizuka, K.; Tanji, T.; Tonomura,
(68) Wang, Z. L.; Shapiro, A. J. Ultramicroscopy 1995, 60, 115. A. In Electron Holography; Tonomura, A., Allard, L. F., Pozzi, G., Joy,
(69) Mayer, J.; Plitzko, J. M. J. Microsc. 1996, 183, 2. D. C., Ono, Y. A., Eds.; Elsevier Science: New York, 1995; p 145.
(70) Batson, P. E.; Browning N. D.; Muller, D. A. MSA Bull. 1994, 24,
(76) Gabor, D. Proc. R. Soc. London A 1949, 197, 454.
371.
(71) Wang, Z. L.; Bentley, J.; Evans, N. D. J. Phys. Chem. B 1999, (77) Tonomura, A. Electron Holography; Springer-Verlag: New York,
103, 751. 1993.
(72) Frost, B. G.; Allard, L. F.; Volkl, E.; Joy, D. C. In Electron (78) Lichte, H. AdV. Opt. Electron Microsc. 1991, 12, 25.
Holography; Tonomura, A., Allard, L. F., Pozzi, G., Joy, D. C., Ono, Y. (79) Tonomura, A., Allard, L. F., Pozzi, G., Joy, D. C., Ono, Y. A.,
A., Eds.; Elsevier Science: New York, 1995; p 169. Eds.; Electron Holography; Elsevier Science: New York, 1995.