Real-Space Imaging of Nucleation and Growth in Colloidal Crystallization

Author(s): U. Gasser, Eric R. Weeks, Andrew Schofield, P. N. Pusey and D. A. Weitz

Source: Science , Apr. 13, 2001, New Series, Vol. 292, No. 5515 (Apr. 13, 2001), pp. 258-
262
Published by: American Association for the Advancement of Science

Stable URL: https://www.jstor.org/stable/3082728

REFERENCES
Linked references are available on JSTOR for this article:
https://www.jstor.org/stable/3082728?seq=1&cid=pdf-
reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

American Association for the Advancement of Science is collaborating with JSTOR to digitize,
preserve and extend access to Science

This content downloaded from

223.178.81.239 on Thu, 21 Mar 2024 17:24:15 +00:00
All use subject to https://about.jstor.org/terms
 REPORTS

References and Notes of

ofcorrelation
correlationfunctions
functions
yield coefficient
yield ofcoefficient
determi- of determi- where
where F is Ftheisgamma
the function,
gammadefined
function,
as F(n) =defined as F(n) =
oe-x xn-1 dx.
1. C. A. Angell, K. L. Ngai, G. B. McKenna, P. F. McMillan,nation
nation R2 R2
values
values
greatergreater
than 0.99 in
than
most0.99
cases. in most cases.
Standard
Standarderrors
errors
for thefor
fitting
theparameters
fittingtend
parameters
to be tend to be 36. The orientational transient can be calculated from
S. W. Martin, J. Appl. Phys. 88, 3113 (2000).
2. M. D. Ediger, Annu. Rev. Phys. Chem. 51, 99 (2000).extremely
extremely smallsmall
(1 part (1
in 105)
partbutinprovide
105) little
but provide little the polarized fluorescence data with
3. H. Sillescu, J. Non-Cryst. Solids 243, 81 (1999). meaningful
meaningful basis basis
for comparison
for comparison
because of thebecause
large of the large
covariance
covariance of 13of
and13 T. A
anduseful
T. measure
A useful of this
measure
error of this error
4. M. D. Ediger, C. A. Angell, S. R. Nagel, J. Phys. Chem.
determines
determines that that
a 1% variation
a 1% variation
in 1 or a 5% in
variation
1 or a 5% variation
0(t)= tan L[ (]
100, 13200 (1996).
in
inr ris is
sufficient
sufficient
to reducetothereduce
R2 valuethe
fromR20.99
value
to from 0.99 to Molecules were determined to have exchanged environ-
5. J. T. Fourkas, D. Kivelson, U. Mohanty, K. A. Nelson,
0.98.
0.98.TheTherotational
rotational
correlationcorrelation
time Tc is defined
timeasTc is defined as ments when the average angle jump changes by more
Eds., Supercooled Liquids: Advances and Novel Appli-
cations, vol. 676 of ACS Symposium Series (American the
theintegral
integral
fromfrom
zero to zero
infinitytoofinfinity
the correlation
of the correlation than 2 SD from the previous average angle jump.
Chemical Society, Washington, DC, 1997). function,
function, but for
buta stretched
for a stretched
exponential function,
exponential function, 37. C. Bennemann, C. Donati, J. Baschnagel, S. C. Glotzer,
6. G. Harrison, The Dynamic Properties of Supercooled this
thisintegral
integral
can becan
conveniently
be conveniently
written in terms
written
of in terms of Nature 399, 246 (1999).
the
theT T andand
13 parameters
13 parametersas as 38. D.A.V.B. is a Research Corporation Cottrell Scholar
Liquids (Academic Press, New York, 1976).
7. T. Gleim, W. Kob, K. Binder, Phys. Rev. Lett. 81, 4404 and a Camille and Henry Dreyfus Foundation New
(1998). Faculty Awardee. This work is supported by grants
8. D. B. Hall, A. Dhinojwala, J. M. Torkelson, Phys. Rev. from the Research Corporation.
Lett. 79, 103 (1997). JC
JCC(tdt
C(tdt
= KWW
= KWW
P KWWP KWW
9. G. Adam, J. H. Gibbs, . Chem. Phys 43, 139 (1965).
0
10 October 2000; accepted 28 February 2001
10. H. Wendt, R. Richert, Phys. Rev. E. 61, 1722 (2000).
11. E. R. Weeks, J. C. Crocker, A. C. Levitt, A. Schofield,
D. A. Weitz, Science 287, 627 (2000).
12. C.-Y. Wang, M. D. Ediger,J. Phys. Chem. B 104, 1724
(2000).
13. I. Chang, H. Sillescu, J. Phys. Chem. B 101, 8794 Real-Space
Real-Space Imaging
Imagingof
of
(1997).
14. K. L. Ngai, J. Phys. Chem. B 103, 10684 (1999).
15. R. Richert, j. Phys. Chem. B 101, 6323 (1997).
Nucleation and Growth in
16. E. V. Russell, N. E. Israeloff, Nature 408, 695 (2000).
17. W. E. Moerner, Science 265, 46 (1994).
18. P. F. Barbara, W. E. Moerner, Acc. Chem. Res. 29, 561
(1996).
Colloidal Crystallization
19. H. P. Lu, X. S. Xie, Nature 385, 143 (1997). U. Gasser,'* Eric R. Weeks,1t Andrew Schofield,z P. N. Pusey,2
20. S. Nie, D. T. Chiu, R. N. Zare, Science 266, 1018
D. A. Weitz'
(1994).
21. D. A. Vanden Bout et al., Science 277, 1074 (1997).
22. R. M. Dickson, A. B. Cubitt, R. Y. Tsien, W. E. Moerner, Crystallization of concentrated colloidal suspensions was studied in real space
Nature 388, 355 (1997). with laser scanning confocal microscopy. Direct imaging in three dimensions
23. T. Basche, S. Kummer, C. Braeuchle, Nature 373, 132
(1995). allowed identification and observation of both nucleation and growth of crys-
24. The single-molecule environment is probe depen- talline regions, providing an experimental measure of properties of the nucle-
dent, with its size determined by the dye used. ating crystallites. By following their evolution, we identified critical nuclei,
25. J. J. Macklin, J. K. Trautman, T. D. Harris, L. E. Brus,
Science 272, 255 (1996).
determined nucleation rates, and measured the average surface tension of the
26. The sample is held atop a closed loop piezo scanning crystal-liquid interface. The structure of the nuclei was the same as the bulk
stage and illuminated with the second harmonic of a solid phase, random hexagonal close-packed, and their average shape was rather
Nd:yttrium-aluminum-garnet laser through a 1.25 nu- nonspherical, with rough rather than faceted surfaces.
merical aperture (NA) oil immersion objective in the epi
configuration. By scanning the sample in the focus of a
laser beam, we can observe the fluorescence from in- The study of the structure, growth, and prop-
the surface energy and bulk energy is reflect-
dividual molecules. The fluorescence signal from the erties of crystals is one of the most important
ed in the free energy for a spherical crystallite
molecules is passed through a dichroic beam splitter to
remove excitation light and split into two orthogonal
areas of solid state physics (1). Although a
4'Ta
polarizations, which are imaged onto the front of two great deal is known about the behavior of AG = 4'rr2y - r3AJJpl (1)
3
single-photon-counting avalanche photodiodes. For bulk crystals, considerably less is known
temperature-dependent data, the sample was held at about their earliest stage, when they nucleate
where
where r is therradius,
is they is the
radius,
free energyyof is the free energy of
constant temperature in an Oxford Microstat-N cryo-
stat, and the fluorescence was excited and collected
and grow from a liquid. Homogeneous nucle-the
thecrystal-liquid
crystal-liquid
interface per unit
interface
area, A!L per unit area, A!L
through a 0.6-NA long working distance air objective. ation occurs when small crystalline regions
isisthethe
difference
difference
between the liquid
between
and solid the liquid and solid
All other aspects of the experiment remain unchanged. chemical
chemical
form from structural fluctuations in a liquid potentials,
potentials,
and n is the numberandden- n is the number den-
27. The Tg of the polymer was measured by differential
cooled below its freezing point. The growth
sity
sity of particles
of particles
in the crystallite
in(2).the
The size
crystallite (2). The size
scanning calorimetry.
28. T. Ha, T. A. Laurence, D. S. Chemla, S. Weiss, J. Phys. ofof
of these regions depends on a competitionthethe
criticalcritical
nucleus is rc =nucleus
2'y/(ALn), is rc = 2'y/(ALn),
Chem. B. 103, 6839 (1999). between a decrease in bulk energy, which
corresponding
corresponding to the maximumto ofthe
AG (Eq.maximum of AG (Eq.
29. R. M. Dickson, A. P. Bartko, J. Phys. Chem. B. 103,
1).
1).
favors growth, and an increase in surfaceDespite
Despite
its crucialits
role, very
crucial
little is role, very little is
3053 (1999).
30. R. M. Dickson, D. J. Norris, Y.-L. Tzeng, W. E. Moerner, energy, which favors shrinkage. The smallest
known
known about this
about
nucleation
this
process,
nucleation
primar- process, primar-
Science 274, 966 (1996). crystals continually form by fluctuations ily
ilybecause
but because
of the difficulty
of the of directly
difficulty
ob- of directly ob-
31. Measurement of the full three-dimensional orientation
then typically shrink away because of serving
the the nuclei
serving thein real
nuclei
space. in real space.
(27, 28) is not required because the dynamics are
isotropic in three dimensions and thus revealed fully in
high surface energy. Growth becomes ener- No
Noexperiment
experiment
has ever directly
has measured
ever directly measured
the size of the critical nucleus. It cannot be
the fluctuations of the dichroism. Additionally, the use getically favorable only when the crystallites
of a high NA objective (NA = 1.25) leads to a slight reach a critical size. The competition between determined theoretically-even the surface
contamination of the dichroism signal due to the out-
tension is unknown. Furthermore, an addi-
of-plane component of the fluorescence that manifests
itself as an additional background but does not adverse- tional important yet unresolved question in-
1Department of Physics and Division of Engineering
ly affect the measured dynamics (28). volves the internal structure of critical nuclei,
and Applied Sciences, Harvard University, Cambridge,
32. J. A. Veerman, M. F. Garcia-Parajo, L. Kuipers, N. F. MA 02138, USA. 2Department of Physics and Astron-
because nucleation need not occur via the
van Hulst, Phys. Rev. Lett. 83, 2155 (1999). omy, University of Edinburgh, Edinburgh, Scotland
33. T. Ha, J. Glass, T. Enderle, D. S. Chemla, S. Weiss, Phys. EH9 3JZ, UK.
stable bulk phase (3). General arguments
Rev. Lett. 80, 2093 (1998). based on symmetry suggest that nucleation
*To whom correspondence should be addressed. E-
34. G. E. P. Box, G. M. Jenkins, Time Series Analysis Forecast-
mail: gasser@deas.harvard.edu
may proceed by a body-centered cubic (bcc)
ing and Control (Holden-Day, San Francisco, 1976).
35. For all correlation functions, the first point containing tPresent address: Physics Department, Emory structure as an intermediate phase, with the
Univer-
the noise is not shown and not included in the fit. Fits sity, Atlanta, GA 30322, USA. final equilibrium solid phase having a differ-

258 13 APRIL 2001 VOL 292 SCIENCE www.sciencemag.org

This content downloaded from

223.178.81.239 on Thu, 21 Mar 2024 17:24:15 +00:00
All use subject to https://about.jstor.org/terms
 REPORTS

ent symmetry (4). Computer simulations

manner, but the of a boundaries of the phasethe sample
dia- before observation. Heterogeneous
Lennard-Jones system found gram critical nuclei
are shifted nucleation
to slightly lower ) values (11). at the walls of the sample chamber
with a face-centered cubic (fcc) core(6-8,
Colloidal and13-16)a and protein (17) sys-by fusing colloidal particles of
was suppressed
predominantly bcc surface layertems(5).have Howev-
been valuable for studying crystal
a smaller size onto the walls before adding the
er, light-scattering measurements frombut
nucleation, hard these studies did not directly
colloidal samples. Further, we focused at least 8
spheres (6-8), which provide visualize the critical nuclei and thus were un-
an average and typically 15 particle diameters from the
over crystalline regions of allable
sizes, suggest
to directly observe their properties. walls to ensure that we studied bulk homoge-
that nucleation occurs through We the report
randomon three-dimensional real- neous nucleation.
hexagonal close-packed (rhcp) space phase, a ran-
imaging of the nucleation and growth We observed the growth of crystallites
dom mixture of hexagonal close-packed of nearly hard-sphere colloidal crystals. We by extracting the particle centers from the
(hcp) and fcc-like stackings of planes
observed with of the smallest nuclei, raw images with an accuracy of -0.04 pxm
the formation
hexagonal order, and this is which supported
shrink far bymore frequently than they (23). From these particle positions, crystal-
recent computer simulations (9). grow; impor-their size distribution, line regions were identified with an algo-
by measuring
The
tant questions about the structure we were
and ablesize
to directly
of determine the sur- rithm that finds ordered regions, indepen-
the nucleating crystallites can face best
tensionbe of are-
colloidal crystal. We were dent of any particular crystal structure (5,
solved by direct experimental also able to determine
observation inthe size of the critical 24). It relies on the assumptions that the
real space. nuclei by observing when the probability to neighbors of a particle in a crystal lattice
The small length and time scales that char- grow exceeds the probability to shrink. We are arranged in a particular orientation
acterize homogeneous crystal nucleation have found that the structure of critical nuclei is about the particle and that this orientation
to date precluded this direct real-space observa- rhcp, the same as that of the bulk. The critical of neighbors is the same for nearby parti-
tion of critical nuclei in atomic or molecular nuclei have rough surfaces, and their average cles. Quantitatively, this is achieved by
systems. By contrast, however, colloidal sys- shape is ellipsoidal. calculating local bond-order parameters
tems can be studied directly, because their larg- We used poly(methyl methacrylate) spheres (25) that are based on spherical harmonics
er size and concomitant slower time scale sterically stabilized with poly-12-hydroxy- Y,m and are a measure of the orientation of
makes them much more experimentally acces-stearic acid (18, 19), with a radius a = 1.26 Ixm the neighbors around a particle. Two adja-
sible. Colloidal particles can serve as good
and a polydispersity of <5%. They were sus-cent particles with similar orientation of
pended in a mixture of decahydronaphthalene
models for atomic or molecular materials; they
show an analogous phase diagram (10), and the and cyclohexylbromide, which closely matches
volume exclusion that results in crystallization
both the refractive index and the density of the M

particles. To make the particles visible with

of colloidal particles also plays a dominant role 100 101 102 103
_ _ _______ _ _ ___?___ _ _____?

in traditional crystallization (2). Concentrated

fluorescence microscopy, we dyed them with 0.5-
rhodamine, which gives the particles a small
suspensions of hard-sphere colloids crystallize
because of entropy (11): The free volume for charge, resulting in freezing at 4f = 0.38 and
ordered spheres (maximum packing volume
melting at ?m = 0.42 (20). We used a fast laser
scanning confocal microscope, which allowed
fraction () = 0.74) is greater than that of disor-
l

(p
dered spheres (maximum packing 4 - 0.64),
us to measure the three-dimensional arrange-
ment of colloidal particles in suspension (21,
and thus, the entropy is higher in the crystalline
state. The phase behavior is determined solely
22). We observed a 58 ,Im by 55 pxm by 20 txm A

-0. ..b * .. .
volume, containing -4000 particles. The total
by the volume fraction (), with freezing occur- 2 10
<r>la
ring at f4)S = 0.494 and melting occurring volume
at of our samples was -50 dLI, and they
4HS = 0.545 (HS, hard sphere) (12). Weakly contained a short piece of paramagnetic wire, Fig. 2. The difference pg -Ps of the probabilities
which we agitated with a magnet to shear melt with which a crystallite is growing or shrinking
charged colloids behave in essentially the same
plotted against the normalized average crystal-
Fig. 1. Four snapshotsA lite radius (r)/a (bottom) and the number of
during crystallization particles M in the crystallite (top) for ( = 0.47.
The size of the critical nucleus lies in the range
of a sample with () =
0.45. The red spheres in which Pg - P goes from negative to positive
values.
are drawn to scale and
represent particles
that were identified as
Table 1. The experimentally measured bond-order
crystal-like. The parti-
cles in the metastable histograms for the crystal regions were least
liquid state are shown squares fit with the histograms for crystals and
by the blue spheres, liquid. For three samples, the fraction of each
reduced in size for structure type found by the fit is given, with the
sum of all the fractions constrained to be 1. The
clarity. (A) Time t =
20 min after shear I values in parentheses are the uncertainty of the
melting, (B) t = 43 last digit. The sum of the squared differences
between the measurement and the fit, A2, is a
min, (C) t = 66 min,
and (D) t = 89 min. measure for the quality of the fit, with smaller
values indicating a better fit. The fit for the ) =
0.45 sample is shown in Fig. 4.

) A2 bcc fcc hcp Liquid

0.49 0.021 0.00(3) 0.10(3) 0.32(6) 0.58(14)

0.45 0.006 0.00(1) 0.58(7) 0.20(3) 0.22(6)
0.43 0.006 0.00(3) 0.34(5) 0.25(4) 0.41(10)

www.sciencemag.org SCIENCE VOL 292 13 APRIL 2001 259

This content downloaded from

223.178.81.239 on Thu, 21 Mar 2024 17:24:15 +00:00
All use subject to https://about.jstor.org/terms

ticles were present.
their neighbors have similar bond-order
rameters and are therefore saidearly-time
to besnapshots
by a crystal-like "bond" (26). Even
where
we represent crystal-like particles as
disordered liquid, crystal-likered spheres are
bonds
uncommon; thus, only particles spheres,
with
showneight
or more crystal-like bonds improve
being crystal-like. All particles
tallite with perfect fcc, hcp,
structure are correctly recognized,
only an insignificant number od
the liquid are found to be crystal-like.
At the beginning of an experiment,
ples started in the metastable
but because of random structural fluctua-
tions, subcritical nuclei of crystal-like par-
Fig. 3. A snapshot of a A * B
crystallite of postcritical left front
size in a sample with 4 = f t
0.47 is shown from three
different directions (A
through C). The
spheres represent crystal-
like particles and are *
drawn to scale; the 243
extra blue particles share ?? *
at least one crystal-like ? .
"bond" to a red particle
but are not identified as C D
crystal-like and are re- top cut
duced in size for clarity.
(D) A cut with a thickness L
of three particle layers
through the crystallite, il-
lustrating the hexagonal
structure of the layers.
Blue, red, and green
spheres represent parti-
cles in the different layers
(front to rear). This cut
was taken from the re-
gion that is indicated by
the green boxes in (A) and
(B). The particle diameter in (D) is reduced in order to improve the visibility of the second Fig.
layers.
Fig. 4. (A) q4, q6, and A
w6 bond-order parame-
ter histograms for fcc
(blue curves), hcp (red
curves), bcc (black
curves), and liquid
(purple curves). (B) The
measured bond-order
histograms (black plots)
of a sample with 4) =
0.45 are shown togeth-
er with a least squares
fit (blue curve) using B
the bond-order histo-
grams from (A). The re-
sults of this and other
structure fits are sum-
marized in Table 1.
260 13 APRIL 2001 VOL 292 SCIENCE www.sciencemag.org
223.178.81.239 on Thu, 21 Mar 2024 17:24:15 +00:00
REPORTS
pa-This is shown in two or shrink (27). Because pg = Ps at the
joined(Fig. 1, A and B), critical size, we plot the difference pg - Ps
in the
as a function of crystallite radius and par-
and liquid-like
not particles as blue ticle number M in Fig. 2 for a sample with
with a reduced diameter to 4) = 0.47. We found an abrupt change from
visibility. as
are defined Typically, these sub- negative to positive values of g - Ps (28),
criticalinnuclei contained no more than 20
a crys- allowing us to identify the critical size,
rhcp,
particles andor which is 60 < M < 160, in good agreement
bcc to reduce their surface
shrank
energy. After
whereas with recent computer simulations (9). This
a strongly 4)-dependent peri-
ofofparticles
time, criticalinnuclei formed and rap-
corresponds to rc - 6.2a, assuming a spher-
ical nucleus. The volume fraction of the
idly grew into large postcritical crystallites
(Fig. 1, C andsam- D). By following the timenuclei is larger than the 4) value of the
evolution
liquidofstate, fluid; above coexistence, the difference is
many crystallites, we deter-
mined the size dependence of the probabil- A4) = 0.012 + 0.003, independent of ),
ities pg and Ps with which crystallites growwhere A() increases slightly for M > 100.
We can understand this A4) value as result-
ing from the higher osmotic pressure exert-
ed by the fluid on the nuclei (16), whereas
in the coexistence regime, A4) must reflect
the evolution of 4) to the higher value,
ultimately attained by the crystallites,
206 red * - . where A() = (m, - (tf. The nucleation rate
densities were slower than 5 mm-3 s- for
4) < 0.45, as well as for 4) > 0.53. Values
of the order of 10 mm-3 s-' were found for
0.45 < ) < 0.53. However, for 0.47 < 4 <
0.53, the average size of the nuclei began to
grow immediately after shear melting; thus,
there was little time for the sample to equil-
ibrate after shear melting, and we were not
able to observe the formation process of
critical nuclei entirely. Our nucleation rate
densities are of the same magnitude as
values obtained by small-angle light scat-
tering from hard spheres (29, 30) but are
two orders of magnitude larger than those
obtained from Bragg scattering (31).
The direct imaging afforded by confocal
microscopy enabled us to determine the
structure and shape of individual nuclei. In
and 3,
third
A through C, we show a crystallite
that is slightly larger than the critical size.
Again, the red spheres represent crystal-
like particles, and the blue ones depict par-
ticles in the liquid state that are on the
i \ <. . ' .
103 . . . v
0.04 -
2: . \ * 0.40 0.45 0.50 0.55.
2lo,it 0.03-
101 X
\ * - *
loi ~\. . .
10? \ * ** '* *
0 200 400 600 800
A (pm2)
Fig. 5. The number of nuclei N(A) (circles) is
shown as a function of the nucleus surface area
A approximated by an ellipsoid (4 = 0.445).
From the fit (line) with the function N(A) =
constant-exp[-Ay/(kBT)], the surface tension
y = 0.026kBT/a2 is determined. The values of -y
for all samples are given in the inset as a
W6 function of 4.
This content downloaded from
All use subject to https://about.jstor.org/terms
 REPORTS

ellipsoid
surface of the crystallite and haveaxesat was
leastnever higher than 0.65 4.? S. Alexander, J. P. McTague, Phys. Rev. Lett. 41, 702
(1978).
one crystal-like bond with a red
0.15 for particle.
nuclei of all sizes. This suggests that
5. P. R. ten Wolde, M. J. Ruiz-Montero, D. Frenkel,
The cores of the crystalline thenuclei are
effects of the nonspherical shape could beJ. Chem. Phys. 104, 9932 (1996).
formed from hexagonal layers,included in any future
as shown by modification of the6. P. N. Pusey, W. van Megen, Nature 320, 340 (1986).
7. P. N. Pusey et al., Phys. Rev. Lett. 63, 2753 (1989).
the cut through the center of theory.
this crystallite
Furthermore, we found that the radius
8. J. Zhu et al., Nature 387, 883 (1997).
(Fig. 3D), and their surfaces ofare more
gyration dis-
rg of crystallites was related to the
9. S. Auer, D. Frenkel, Nature 409, 1020 (2001).
ordered. The interface between theofhighly
number particles M within each crystallite
10. P. N. Pusey, in Liquids, Freezing and the Glass Transi-
ordered cores and the amorphous rgf with the fractal dimension tion, D. Leveesque, J. P. Hansen, J. Zinn-Justin, Eds.
as M(rg) ocliquid
(Elsevier, Amsterdam, 1991), pp. 763-942.
df = 2.35in
exterior is not sharp; this decrease + 0.15 for all values of (; the
order 11. A. K. Sood, in Solid State Physics, H. Ehrenbach, D.
fractal
from the interior to the exterior is behavior
in agree-presumably reflects the Turnbull, Eds. (Academic Press, Boston, 1991), vol. 45,
ment with computer simulations ofofatheir
roughness Len- surfaces. pp. 1-73.
The interfacial tension between the 12. B. J. Alder, W. G. Hoover, D. A. Young, J. Chem. Phys.
nard-Jones system (5) and density function- 49, 3688 (1968).
al calculations for hard spheres (32).
crystal and fluid phases is a key parameter
13. K. Schatzel, B. J. Ackerson, Phys. Rev. Lett. 68, 337
To determine the crystal structure the nucleation process, yet(1992).
of the
that controls y is
14. P. Bartlett, R. H. Ottewill, P. N. Pusey, Phys. Rev. Lett.
ordered cores, we used the rotational in-
difficult to calculate (1, 2) and to measure
68, 3801 (1992).
variants q4(i), q6(i), and w6(i) (33), whichbut with our data,15.we
experimentally, Y. He,can
B. J. Ackerson, W. van Megen, S. M. Underwood,
are quantitative measures for the local
directly measure or-y by examining the statis- K. Schatzel, Phys. Rev. E 54, 5286 (1996).
der around particle i, and distinguished tics of the smallest be- nuclei. For r < 16. rc,B. J. Ackerson, K. Schatzel, Phys. Rev. E 52, 6448
the
(1995).
tween fcc, hcp, bcc, and liquid surface (5, term
9, 24). in Eq. 1 dominates the free
17. S.-T. Yau, P. G. Vekilov, Nature 406, 494 (2000).
Because of thermal fluctuations, energy even ho-crystallites, and consequent-
of the 18. L. Anti et al., Colloids Surf. 17, 67 (1986).
19. S.-E. Phan et al., Phys. Rev. E 54, 6633 (1996).
mogeneous crystals have a distribution ly, we expect the ofnumber of crystallites to
20. We determined the particle radius by measuring the
these parameters. Nevertheless, be N(A) theseoc exp[-Ay/(kBT)],
dis- where A is the
volume per particle at random close packing (,rcp =
tributions are distinct for surface every area, crystal which we approximate by anrhcp (crhcp = 0.74) with the microscope.
0.64) and
structure, as shown by the calculations ellipsoid. In Fig. (34)5, we show that thisWepro- obtained the same particle radius from these two
points in phase space; therefore, we used this radius
in Fig. 4A. These are compared portionality
to measure- does indeed hold, allowing us ) of all other samples from the mea-
to calculate
ments of experimental crystalline nuclei,to obtain values for the surface tension. As sured volume per particle.
21. A. van Blaaderen, P. Wiltzius, Science 270, 1177
shown by a black curve in Fig. shown 4B. in the The
inset, /y w 0.027kBT/a2 and
(1995).
blue curve is a least squaresmay fitdecrease
with slightly
the with increasing22.( E. R. Weeks, J. C. Crocker, A. C. Levitt, A. Schofield,
values.
distributions from Fig. 4A, and the fit pa- This value of y is in reasonable D. A. Weitz, Science 287, 627 (2000).
rameters are summarized in Table 1. Dur- agreement with density functional calcula- 23. J. C. Crocker, D. G. Grier,J. Colloid Interface Sci. 179,
298 (1996).
ing the nucleation process, bcc ordertions is for hard spheres and Lennard-Jones 24. P. J. Steinhardt, D. R. Nelson, M. Ronchetti, Phys. Rev.
completely absent; in particular, as shown
systems (32). By contrast, an older densityB 28, 784 (1983).
in Fig. 4, the q4 distribution for bccfunctional
is study (35) and computer simula- 25. The normalized order parameter q, (i) of particle i has
(21 + 1) complex components
markedly different from the measuredtions dis- of hard spheres (9, 36) gave results B(i)
approximately
tribution. Instead, the main contributions to four times as large. Our
qtm(i) = NB(i) a()
measurement of a low value of y is consis-
the structure of the experimental crystalline j=l
nuclei are due to fcc, hcp, and liquid. tent
The with the observed rough surfaces of thewhere N is a normalization factor such that
mixture of fcc and hcp is the signature of
crystallites; this may reflect the effects of ,mlm(i)qtm(i) = 1, B(i) is the number of neighbors
rhcp structure and is consistent withthe thesofter potential due to the weak chargesof particle i, Pr is the unit vector pointing from
particle i to its jth neighbor, and Ylm is a spherical
stacked hexagonal planes in Fig. 3D. The of our particles. Approximating the criticalharmonic function. Two particles are defined as
liquid contribution in the fit is mainlynucleus
due as an ellipsoid, with M, 110, we neighbors if their separation is smaller than a cutoff
to the surface of the crystallites. Theseobtain Ac = 880 Lxm2, AL - 0.13kBT, andvalue of 2.83a, chosen to be close to the minimum
between the first and second peaks of the pair
AG(AC) 7.4kBT.
results are in agreement with those expect-
correlation function g(r) for our samples. The "bond"
ed for hard spheres, which also crystallize The direct observation of the structure, between two neighboring particles i and j is recog-
in the rhcp structure (7), confirming dynamics,
that and evolution of small crystalline nized as crystal-like if the complex inner product
the critical nuclei are formed only from nuclei
the allowed us to characterize the key r,mqm(i)mtm(j) exceeds a threshold value of 0.5. As
in other studies (5, 9), we used the I = 6 order
bulk solid phase, rather than forming from
processes of nucleation and growth of colloi- parameter.
a different metastable solid phase. dal crystals and to quantitatively determine 26. In this context, "bond" denotes the direction that

theisimportant parameters that control this aconnects

In classical nucleation theory (2), it a particle with one of its neighbors and not
bond in the chemical sense.
assumed that the nuclei have a spherical
process. These results also provide crucial27. Whether a nucleus was fully visible or not was de-
shape due to surface tension. To the best of
experimental input to help guide any future cided by looking for particles in the nucleus that were
our knowledge, this assumption has never refinements of theories for nucleation and closer than 0.8a to an edge of the observed volume.
Nuclei that did not meet this criterion were not used
been experimentally verified. As showngrowth by of colloidal crystals. Moreover, the to determine the probabilities pg and Ps. A typical
the example in Fig. 3, our data indicate ability
that to obtain three-dimensional movies of measurement contained between 1500 and 2500
the crystallite surfaces were quite rough, the
sug-dynamics of colloidal solids will allow subcritical and critical nuclei that were fully visible
and could be tracked for some time.
gesting that the surface tension is low. the
Con-direct observation and study of other im- 28. Larger nuclei were not observed as often as smaller
sistent with this, the crystallite shapesportant
ap- phenomena, such as defect motion, ones, because they have a higher free energy and,
effects of impurities, and crystallization of consequently, do not form as often. Thus, the bin size
peared rather nonspherical. To confirm this,
on the average crystallite radius (r) axis of Fig. 2 is
colloidal alloys.
we approximated each nucleus with an ellip-
increased logarithmically, which also reflects the ac-
soid and found that the average nucleus is curacy with which the critical size can be identified.
best described by an ellipsoid whose major References and Notes 29. Nucleation density rates obtained with particles of
axes have the ratio 1:1.8:3.2. Nuclei contain- 1. T. Palberg,J. Phys. Condens. Matter 11, R323 (1999). different sizes can be compared when expressed in
2. K. F. Kelton, in Solid State Physics, H. Ehrenbach, D. dimensionless form using the intrinsic length and
ing >50 particles were found to be somewhatTurnbull, Eds. (Academic Press, Boston, 1991), vol. 45,time scales. For this comparison, Do/(2a)-5 is used as
less anisotropic than smaller ones, but onpp. 75-177. the unit of the nucleation rate density, where Do 0
average, the ratio of the shortest and longest
3. W. Ostwald, Z. Phys. Chem. 22, 289 (1897). 0.085 JiMm2 s- is the free-particle diffusion constant.

www.sciencemag.org SCIENCE VOL 292 13 APRIL 2001 261

This content downloaded from

223.178.81.239 on Thu, 21 Mar 2024 17:24:15 +00:00
All use subject to https://about.jstor.org/terms
 REPORTS

30.
30.K. Schatzel,
K. Schatzel,
B. J. Ackerson,
B. Phys.
J. Ackerson,
Rev. E 48, 3766 Phys. Rev.
w,(i) aE
w,(i)is is a 48,
third-order3766
third-orderrotational
rotational
invariant defined by defined by calculated
invariant calculatedin
inthe
thesame
samewayway as as
forforparticle
particle
positions
positions
(1993). from
from aa measurement.
measurement. The
Theliquid
liquid
histogram
histogram was was
taken
taken
31. J. L. Harland, W. van Megen, Phys. Rev. E 55 3054 w(i) = m,,m2,m3E (m1 m2 m/3 from
from observations
observationsshortly
shortly after
after
shear
shear
melting,
melting,
when
when
(1997). mn + m2 + m3 = O Ml lm l0 the
the sample
samplewaswasininthe
the
liquid
liquidstate.
state.
32. R. Ohnesorge, H. Lbwen, H. Wagner, Phys. Rev. E 50, 35.
35. D.
D. W.
W.Marr,
Marr,A.A.P.P.
Gast,
Gast,Phys.
Phys.Rev.
Rev.E 47,
E 47,
12121212
(1993).
(1993).
4801 (1994). x q,m,(i)qm,,2(i)qtm3(i) 36.
36. R.
R. L.
L. Davidchack,
Davidchack,B.B.
B. B.
Laird,
Laird,
Phys.
Phys.
Rev.
Rev.
Lett.
Lett.
85, 4751
85, 4751
33. q,(i) is a second-order rotational invariant of the where (:::) stands for a Wigner 3j symbol. (2000).
(21 + 1)-component bond-order parameter q,(i) for 34. The histograms for fcc, bcc, and hcp order were 37. This work was supported by NSF (grant
particle i and is defined by obtained from computer-generated crystal data. Ran- DMR-9971432) and NASA (grant NAG3-2284). U.G.
dom fluctuations were added to the particle positions also acknowledges support by the Swiss National
4,T in such a way that the width of the first peak of g(r) Science Foundation.
ql(i) = 21 +-m= I- Jq,,(i) for the computer-generated data matched the exper-
imental value. The bond-order histograms were then 20 December 2000; accepted 28 February 2001

Capturing
Capturinga aPhotoexcited
Photoexcited
catalysis,
catalysis, and
and the
thephotoinduced
photoinducedredox
redoxreactions.
reactions.
According
According toto previous
previousstudies
studies(20-22)
(20-22)inina a
weakly
weakly coordinating
coordinatingorornoncoordinating
noncoordinating sol-
sol-
Molecular
MolecularStructure
StructureThrough
Through square
square planar
planar geometry
geometryofofNiTPP
NiTPPadopt
adopta sin-
a sin-
vent,
vent, the
the 3d8
3d8 electrons
electronsof
ofthe
theNi(II)
Ni(II)ion
ioninin
the
the

Time-Domain
Time-DomainX-ray
X-rayAbsorption
Absorption glet
glet spin
trons
spin state
trons in
state (3dz2)2,
in the
(3dz2)2,SO,
the octahedral
SO,whereas
whereasthese
octahedralgeometry
theseelec-
elec-
geometryofofNiTPP-L2
NiTPP-L2
Fine Structure prefer
prefer the
the triplet
tripletspin
spinstate
state3(3d,2_2,
3(3d,2_2,3dz2),
3dz2),To.
To.
This
This difference
difference reflects
reflectsthe
theresponse
responseofof
the
the

Lin X. Chen,1* Wighard J. H. Jager,1 Guy Jennings,2 3dx2y2and

3dx2y2and 3dz2
3dz2 orbitals
orbitalsininthe
thesquare
squareplanar
planar
and
and
octahedral
octahedral ligand
ligandfields.
fields.Starting
Startingwith
withNiTPP-
NiTPP-
David J. Gosztola,l Anneli Munkholm,lt Jan P. Hessler'
L2, aa laser
laser pump
pump pulse
pulseinduces
inducesananelectronic
electronic
transition
transition from
from the
thetriplet
tripletground
groundstate,
state,
T(,T(,
toto
anan
The determination of the structure of transient molecules, such as photoexcited excited
excited triplet
triplet state,
state,T*T*(Fig.
(Fig.1B).
1B).This
Thisstate
state
decays
decays and
states, in disordered media (such as in solution) usually requires methods with high and undergoes
undergoesintersystem
intersystemcrossing
crossingtoto
anan
temporal resolution. The transient molecular structure of a reaction intermediateexcited
excited singlet
singlet state,
state,S*.
S*.This
Thisexcited
excitedstate
state
ejects
ejects
produced by photoexcitation of NiTPP-L2 (NiTPP, nickeltetraphenylporphyrin;twotwo
L, ligands
ligands LL to
to produce
producethe thesquare
squareplanar
planar
geometry
geometry (20-22).
piperidine) in solution was determined by x-ray absorption fine structure (XAFS) (20-22).If
Ifthe
thesolvent
solventwere
werenoncoor-
noncoor-
dinating
dinating or
data obtained on a 14-nanosecond time scale from a third-generation synchrotron or weakly
weaklycoordinating,
coordinating,this thisreaction
reaction
source. The XAFS measurements confirm that photoexcitation leads to the rapid sequence
sequence would
would bebecompleted.
completed.However,
However,inin a a
strongly
strongly coordinating
removal of both axial ligands to produce a transient square-planar intermediate, coordinatingsolvent,
solvent,S,S,isisunstable
unstableand
and
the system
NiTPP, with a lifetime of 28 nanoseconds. The transient structure of the photo- system will
will return
returnto toTo,
To,bybyrecombining
recombining
with
with two
dissociated intermediate is nearly identical to that of the ground state NiTPP, two piperidine
piperidineligands
ligandstotoform
formananoctahe-
octahe-
suggesting that the intermediate adopts the same structure as the ground statedrally
drally
in coordinated
coordinatedNiTPP-L2
NiTPP-L2(Fig.
(Fig.1).1).The
Thetran-
tran-
sient structures involved in this recombination
a noncoordinating solvent before it recombines with two ligands to form the more
stable octahedrally coordinated NiTPP-L2. are not known. A penta-coordinated NiTPP-L
molecule could be involved in the recombina-
Photoexcited states of molecules are often stud- Here
Here we
wepresent
presenttime-domain
time-domain laser
laser
pump/
pump/tion, the square planar NiTPP could return to T(
ied with "pump-probe" methods, in which a x-ray x-ray probe
probeXAFS
XAFSstudies
studies
using
using
x-ray
x-ray
pulses
pulses
via a concerted axial chelation with two piperi-
short optical pulse excites the ground state, and from
from aa third-generation
third-generationsynchrotron
synchrotron source.
source.
dine molecules, or an intermediate square pyra-
a second probe pulse interrogates the excited (The(The experiments
experimentswerewereconducted
conducted at aatwig-
a wig-
mid NiTPP-L structure could be important. Be-
state. For structural studies, substantial progress gler
gler beamline
beamlineatatthe
theBasic
Basic
Energy
Energy
Sciences
Sciences
cause it is difficult to identify a unique optical
has been made in generating ultrashort x-ray Synchrotron
SynchrotronResearch
Research
Center
Center
at Argonne
at Argonne
Na- Na-
absorption spectrum for each possible interme-
pulses (1-4) that have been used to obtain tional Laboratory's Advanced Photon diate involved in photodissociation, transient
time-resolved x-ray diffraction information Source.) Although the duration of the x-ray optical absorption spectra of NiTPP in coordi-
with picosecond resolution (5-7). Because pulses (100 ps at full width at half-maximum)nating solvent alone are not sufficient to identify
many light-driven processes occur in disordered is too long to follow atomic displacements the intermediates.
media, techniques that do not rely on the long- due to the photoexcitation, it is suitable for The excited-state dynamics of NiTPP-L2
range order of a system must be developed. The capturing intermediate structures with sub- have been studied by ultrafast optical transient
approach of laser pump/x-ray probe x-ray ab- nanosecond resolution. We are able to take absorption and Raman spectroscopy (20-25).
sorption fine structure (XAFS) was proposed "snapshots" of molecules when the popula- The intersystem crossing from T* to S * and the
earlier (8-10), and transient molecular struc- tion of the laser-generated transient speciessubsequent dissociation to SO occur in less than
tures, generated in light-induced processes, reaches a maximum; in this case, the species a few hundred picoseconds (22). In a strongly
were captured without using the temporal res- is a reaction intermediate with a 28-ns life- coordinating solvent at room temperature, this
olution of the source (11-19). state of NiTPP has a lifetime of 28 ns, because
time that results from the photodissociation
of NiTPP-L2 (NiTPP, nickeltetraphenylpor- it must react with two piperidine molecules to
phyrin; L, piperidine) in solution (Fig. 1A).
reform NiTPP-L2. We can therefore use x-ray
'Chemistry Division and 2Materials Science Division, The photoinduced processes of hemelike
pulses from a synchrotron source to determine
Argonne National Laboratory, Argonne, IL 60439,
the local structure of these intermediates and
porphyrin derivatives are examples of how,
USA.
through electron transfer, the coordination ge-
compare them with the structure of ground state
*To whom correspondence should be addressed. E-
mail: lchen@anl.gov
NiTPP. In addition, it is possible to determine
ometry of a metal ion is altered by the electronic
tPresent address: LumiLeds Lighting, 370 West structure of its macrocycle ligands. These mol- whether an intermediate species with only one
Trimble Road, San Jose, CA 95131, USA. ecules can serve as models for oxygen transport,axial ligand is present (22). The experiment was

262 13 APRIL 2001 VOL 292 SCIENCE www.sciencemag.org

This content downloaded from

223.178.81.239 on Thu, 21 Mar 2024 17:24:15 +00:00
All use subject to https://about.jstor.org/terms