letters to nature

short-lived neutron-rich nuclei with exotic properties, for which crystals3 by manipulating individual particles with optical twee-
extensive experimental surveys will not be practicable. The present zers. Direct imaging of these defects veri®ed that their stable
comprehensive data set provides a quantitative test of the physics con®gurations have lower symmetry than the underlying triangu-
incorporated in these models, by focusing on predictions of the lar lattice, as predicted by numerical simulations for a number
boundary between the simple fusion process envisaged by Bohr, and of two-dimensional systems4±7. It was also observed that point
the complex, dynamical process of quasi-®ssion. M defects can dissociate into pairs of well-separated dislocations, a
Received 28 March; accepted 10 July 2001.
topological excitation especially important in two dimensions.
Here we use a similar experimental system to study the dynamics
1. Oganessian, Yu. Ts. et al. Synthesis of nuclei of the superheavy element 114 in reactions induced by
48
Ca. Nature 400, 242±245 (1999).
of mono- and di-vacancies in two-dimensional colloidal crystals.
2. Oganessian, Yu. Ts. et al. Observation of the decay of 292116. Phys. Rev. C 63, 011301R-1±011301R-2 We see evidence that the excitation of point defects into disloca-
(2001). tion pairs enhances the diffusion of di-vacancies. Moreover, the
3. Back, B. B. Complete fusion and quasi®ssion in reactions between heavy ions. Phys. Rev. C 31, 2104± hopping of the defects does not follow a pure random walk, but
2112 (1985).
4. Toke, J. et al. Quasi-®ssionÐthe mass-drift mode in heavy-ion reactions. Nucl. Phys. A 440, 327±365
exhibits surprising memory effects. We expect the results pre-
(1985). sented in this work to be relevant for explaining the dynamics of
5. Blocki, J. P., Feldmeier, H. & Swiatecki, W. J. Dynamical hindrance to compound-nucleus formation other two-dimensional systems.
in heavy-ion reactions. Nucl. Phys. A 459, 145±172 (1986).
Our colloidal crystals were made from negatively charged poly-
6. Hinde, D. J., Leigh, J. R., Newton, J. O., Galster, W. & Sie, S. Fission and evaporation competition in
200
Pb. Nucl. Phys. A 385, 109±132 (1982). styrene-sulphate microspheres (no. 5036, Duke Scienti®c),
7. Hahn, R. L., Toth, K. S., LeBeyec, Y., Lagarde, B. & Guidry, M. W. Reactions with 40Ar and 84Kr leading 0.360 mm in diameter, suspended in highly deionized water. The
to the same compound nucleus, 200Po. Phys. Rev. C 36, 2132±2135 (1987). particles crystallize owing to strong electrostatic interactions. The
8. Sahm, C.-C. et al. Fusion probability of symmetric heavy, nuclear systems determined from
evaporation-residue cross sections. Nucl. Phys. A 441, 316±343 (1985).
experimental details can be found elsewhere2. Brie¯y, a single layer
9. Morton, C. R. et al. Resolution of anomalous ®ssion fragment anisotropies for the 16O + 208Pb of particles is con®ned between two fused-silica substrates separated
reaction. Phys. Rev. C 52, 243±251 (1995). by ,2 mm, resulting in a single-layer colloidal crystal with a lattice
10. Hinde, D. J. et al. Conclusive evidence for the in¯uence of nuclear orientation on quasi®ssion. Phys. constant a < 1.1 mm. The system is quasi-two-dimensional owing
Rev. C 53, 1290±1300 (1996).
11. Bohr, N. Neutron capture and nuclear constitution. Nature 137, 344±348 (1936).
to the suppression of the out-of-plane motion of the particles by
12. Dasgupta, M., Hinde, D. J., Rowley, N. & Stefanini, A. M. Measuring barriers to fusion. Annu. Rev. the negative charge on the silica±water interface. For the conditions
Nucl. Part. Sci. 48, 401±462 (1998). of our experiment, the r.m.s. amplitude of the out-of-plane motion
13. Hagino, K., Rowley, N. & Kruppa, A. T. A program for coupled-channels calculations with all order
is 87 6 5 nm, measured using a Burleigh Instruments TSE-150
couplings for heavy-ion fusion reactions. Comput. Phys. Commun. 123, 143±150 (1999).
14. Lestone, J. P. et al. Pre-scission charged-particle multiplicities following the reactions stage (A.P., unpublished results). As a result, the particles do not
164,167,170
Er + 28Si. Nucl. Phys. A 559, 277±316 (1993). move out of focus (oil-immersion objective lens, Zeiss Plan-Neo-
15. Vergani, P. et al. Complete and incomplete fusion and emission of preequilibrium nucleons in the ¯uar, magni®cation 100 ´, numerical aperture 1.3, depth of focus
interaction of 12C with 197Au below 10 MeV/nucleon. Phys. Rev. C 48, 1815±1827 (1993).
16. Rowley, N., Leigh, J. R., Wei, J. X. & Lindsay, R. Obtaining average angular momenta from fusion
,200 nm).
excitation functions near the Coulomb barrier. Phys. Lett. B 314, 179±184 (1993). Trapping a particle with optical tweezers8 and dragging it away
17. Balantekin, A. B., DeWeerd, A. J. & Kuyucak, S. Relations between fusion cross sections and average from its lattice site creates point defects. We recorded on videotape
angular momenta. Phys. Rev. C 54, 1853±1862 (1996). the dynamics of such single, isolated point defects, using a CCD
18. Ramamurthy, V. S. et al. Entrance-channel dependence of ®ssion-fragment anisotropies: a direct
experimental signature of ®ssion before equilibration. Phys. Rev. Lett. 65, 25±28 (1990).
(charge-coupled device) video camera. Each individual frame (1/
19. Ravi Prasad, G. V. & Ramamurthy, V. S. Light fragment emission during mass asymmetry relaxation in 60 s) was acquired on a PC and processed using a particle-tracking
heavy-ion induced ®ssion. Phys. Rev. C 54, 815±821 (1996). algorithm9. We used a triangulation algorithm to identify the mis-
20. Davies, K. T. R. & Sierk, A. J. Conditional saddle-point con®gurations. Phys. Rev. C 31, 915±922
(1985).
21. MoÈller, P. & Nix, J. R. Potential-energy surfaces for asymmetric heavy-ion reactions. Nucl. Phys. A 281,
354±372 (1977).
22. Abe, Y., Ayik, S., Reinhard, P.-G. & Suraud, E. On stochastic approaches of nuclear dynamics. Phys.
Rep. 275, 49±196 (1996).

Acknowledgements
M.D. acknowledges the support of a QEII Fellowship.

Correspondence and requests for materials should be addressed to D.J.H.

(e-mail: David.Hinde@anu.edu.au).

.................................................................
Diffusion of point defects in
two-dimensional colloidal crystals
Alexandros Pertsinidis & X. S. Ling

Department of Physics, Brown University, Providence, Rhode Island 02912, USA

..............................................................................................................................................
Uniform colloidal microspheres dispersed in a solvent will, under Figure 1 Optical micrographs of point defects. Top row, con®gurations of a mono-
appropriate conditions, self-assemble into ordered crystalline vacancy (from ref. 2): left, three-fold symmetric V3; right, crushed V2a. The 7-fold-
structures1. Using these colloidal crystals as a model system, a coordinated particles are more likely to hop towards the vacant site, resulting in particular
great variety of problems of interest to materials science, physical easy directions for diffusion (indicated with arrows). Bottom row, di-vacancy breaking
chemistry, and condensed-matter physics have been investigated into a dislocation pair. Filled circles denote particles that are 5-fold coordinated (green)
during the past two decades. Recently, it has been demonstrated2 and 7-fold coordinated (purple). Images are separated by roughly 1/6 s. See text for
that point defects can be created in two-dimensional colloidal details.

NATURE | VOL 413 | 13 SEPTEMBER 2001 | www.nature.com © 2001 Macmillan Magazines Ltd 147
letters to nature

Figure 2 Trajectories of defects: left, mono-vacancy; right, di-vacancy. Neighbouring trajectories a few seconds after the defect diffused away from the region shown. The
clumps of dots are separated by roughly half the lattice spacing, and correspond to the positions of the defects are displaced with respect to the lattice, owing to slow, long-
defect being in a con®guration near a lattice point or in a split con®guration between two wavelength, thermally generated deformations.
lattice points. The grey lines show the position of the lattice, superimposed on the

coordinated particles and characterize the con®guration of the a 25

defect. Figure 1 shows examples of microscopic images of the Mono-vacancies
defects as well as the topological analysis of their con®gurations. Di-vacancies
Mean square displacement (a2)

(See also Supplementary Information.) 20

We tracked a number of mono- and di-vacancies for long enough
times (up to 40 s) to obtain reliable measurements of their diffusion
constants. Typical trajectories are shown in Fig. 2. The position of 15
the defect x(t) is de®ned as the average of the positions of the 5-
fold-coordinated particles around the core of the defect. x(t) is
determined with a resolution of ,0.1a. Thus we can identify the 10
defect in a particular lattice site or in between two neighbouring
sites. To measure the diffusion constants, we take the average of
the squared displacements hDx2(dt)i as a function of the time 5
separation dt, running through all the available trajectories. A
linear relationship between hDx2i and dt is obtained for the range
1±5 s (Fig. 3a). The data in this range are weighted by their errors 0
0 1 2 3 4 5
and ®tted to a linear relation with a zero y-intercept, whose slope
b 25
gives the respective diffusion constants, D. The results for mono-
and di-vacancies respectively are Dmono/a2 = 3.27 6 0.03 Hz and "Bare" di-vacancy processes
Ddi /a2 = 3.71 6 0.03 Hz. Dislocation dissociation-
Mean square displacement (a2)

20 recombination
The diffusion constants measured above give the combined
contribution of all possible diffusion pathways. In a previous
study2 it was found that the possible stable con®gurations of
15
mono- and di-vacancies are not just the `well-identi®ed' point
defect con®gurations, but that there are also con®gurations where
the point defects appear as separated dislocation pairs. Here we 10
present evidence that these topological excitations are important in
the diffusion of point defects. We observe that the defects dissociate
into two dislocations that glide from site to site and ®nally 5
recombine at a new position that can be several lattice constants
away from the original one. The individual dislocations glide faster
than the hopping of mono- or di-vacancies. Such a process is a 0
distinct diffusion pathway, giving a separate contribution (Dpair) to 0 0.5 1.0 1.5 2.0 2.5 3.0
the diffusion constant.
Time (s)
We found that mono-vacancies spend considerably less time as a
dislocation pair than do di-vacancies, and the above diffusion Figure 3 Time dependence of mean square displacements. a, Mono- and di-vacancies
mechanism is not important for mono-vacancies. The different (open circles with black errror bars and ®lled circles with red error bars, respectively).
behaviour exhibited by mono- and di-vacancies can be easily The average of hDx2i as function of dt was measured from ,2-minute trajectories. The
distinguished, as illustrated in Fig. 4, where the separation of the red and purple straight lines are linear ®ts for mono- and di-vacancies, respectively,
dislocations comprising the defect is shown as a function of time. for dt in the range of 1±5 s. b, hDx2i separately for `bare' di-vacancies and for
The mono-vacancy is observed primarily to ¯uctuate between di-vacancies excited into dislocation pairs. Fits are for dt values of 1±3 s.

148 © 2001 Macmillan Magazines Ltd NATURE | VOL 413 | 13 SEPTEMBER 2001 | www.nature.com

8 Mono-vacancies
7
6 1D model
5 0.3
r(t)/a
4
3 D2
SD
2 0.1
3
2 V2
SV
0 2 4 6 8 10 12 14
Time (s)
Figure 4 Separation r (t ) of the dislocations comprising a mono-vacancy (top trace)
and a di-vacancy (bottom trace). r(t ) is measured as the standard deviation of positions
of the 5-fold-coordinated particles in the defect core. Dotted horizontal lines
indicate the positions of the minima in the dislocation pair potential, as found in ref. 2.
The blue lines correspond to con®gurations of bound dislocations. These are the split
and crushed con®gurations (labelled SV, SD and V2, D2, respectively) of mono- and
di-vacancies, discussed in detail in ref. 2. The red lines designate con®gurations of two
well-separated dislocations. In the case of di-vacancy, the arrows indicate intervals in
which the defect appears as a repeatedly dissociating and recombining dislocation
pair.
con®gurations in which the dislocations are bound, with very rare
excursions into con®gurations where the dislocations dissociate. In
contrast, the di-vacancy frequently spends long intervals (a few
seconds) as a dislocation pair, with the two dislocations dissociating
and recombining several times during this period.
The combined diffusion constant for di-vacancies is Ddi =
pDpair + (1 - p)Dodi, where p < 0.25 is the fraction of time the di-
vacancies are excited into a dislocation pair, and Dpair and Dodi are the
respective contributions from the dissociation±recombination
mechanism and from the `bare' di-vacancy hopping. In order to
estimate the diffusion constants for the two different mechanisms,
we link into two separate trajectories the intervals spent as a
dislocation pair and as a `bare' di-vacancy. To avoid systematically
picking up hopping events, the shortest interval was 1 s (larger than
the time between jumps). Carrying out the same procedure as for
the original trajectories, we found the two diffusion constants to
be Dpair/a2 = 5.90 6 0.09 Hz and Dodi/a2 = 2.97 6 0.03 Hz (Fig. 3b).
These numbers combine to give the measured value Ddi/a2 =
3.70 Hz.
We now turn our attention to the details of the underlying
stochastic hopping processes. The trajectories of the defects would
be naively expected to be pure random walks on a triangular
latticeÐthat is, when hopping, the system chooses with equal
probability between the six neighbour sites and consecutive events
are uncorrelated. However, a careful analysis reveals a peculiar
feature: unlike a two-dimensional (2D) random walk, the motion
of the defects is often restrained along particular directions. The
existence of easy directions for diffusion is related to the low
symmetry of the stable con®gurations of the relaxed lattice
around the defect. The probability for hopping towards the six
nearest neighbours is not the same for all of them, but is higher
along particular lines of symmetry (Fig. 4). For di-vacancies excited
into a pair of separated dislocations, the easy direction coincides
with the gliding direction of the two dislocations. The easy direction
for diffusion remains unchanged for up to a few seconds, that is,
several consecutive hopping events.
NATURE | VOL 413 | 13 SEPTEMBER 2001 | www.nature.com © 2001 Macmillan Magazines Ltd
letters to nature
0.5
Di-vacancies
2D model
0.4
Return probability
0.2
0.0
0 2 4 6 8 10
Number of steps
Figure 5 Po(n) measured from 40 s of data, or roughly 60 hopping events. The
behaviour observed for both type of defects is more like that predicted for a 1D, rather
than a 2D, random walk. The data show oscillations, with a period of two steps, and
the average trend lies between the curves for 1D and 2D walks.
The existence of easy directions for hopping makes the trajec-
tories of the defects quasi-one-dimensional (quasi-1D). We can
demonstrate this effect by measuring the probability Po(n) for a
defect to return to its original position after it has executed n steps.
We found that the measured Po(n) for small n resembles the
value for random walk on a 1D chain, rather than that for
random walk on a 2D triangular lattice. This is illustrated in
Fig. 5: Po(n) clearly deviates from the expected 2D behaviour,
following the characteristic oscillations of the 1D random walk
with a period of two steps.
Another surprising property of the diffusion of these defects
involves memory effects during hopping. The defects appear to
spend some time around a lattice site, attempting to hop towards
neighbouring sites by changing into a split intermediate con®g-
uration between the two lattice points. Depending on the sym-
metry of the relaxed lattice in the original con®guration (2- or 3-
fold), the defect is seen to attempt hopping towards either two or
three of the six neighbours. The defect typically hops back to the
original site, ®nally getting to the next site only after several
attempts. The back-hopping to forward-hopping ratio is found
to be ,3:1.
We suggest that this memory effect is a result of the slow
relaxation rates of distortions in the over-damped colloidal crystal.
The core region of the defect can rearrange rapidly (,10 Hz), but
the rest of the lattice around the defect cannot follow at the same
rate (for example, distortions with wavelength 10a relax at ,1 Hz).
As a vacancy changes from a con®guration at one site to a split
intermediate con®guration between two sites, the rest of the lattice
does not immediately relax, and seems to `push' the vacancy back to
its initial site.
The results we present here are relevant to a number of other
systems, such as Wigner electron crystals in low-carrier-density
semiconductor heterostructures and Abrikosov vortex lattices,
formed by lines of magnetic ¯ux quanta in type-II superconductors.
In particular, dynamical processes in the vortex lattice should
resemble those we observed in the colloidal crystalÐas the
dynamics of vortices in type II superconductors are also over-
damped. Our observations might also apply to the dynamics of
surfaces of atomic crystals. M
Received 28 March; accepted 21 June 2001.
1. Gast, A. P. & Russel, W. B. Simple ordering in complex ¯uids. Phys. Today 51(12), 24±30 (1998).
2. Pertsinidis, A. & Ling, X. S. Equilibrium con®gurations and energetics of point defects in two-
dimensional colloidal crystals. Phys. Rev. Lett. 87, 098303 (2001).
149
letters to nature
3. Murray, C. A. in Bond-Orientational Order in Condensed Matter Systems (ed. Strandburg, K. J.) Ch. 4 margins are characterized by margin-parallel deep-water zones,
(Springer, New York, 1992).
4. Fisher, D. S., Halperin, B. I. & Morf, R. Defects in the two-dimensional electron solid and implications
apparently representing successive stages of margin evolution,
for melting. Phys. Rev. B 20, 4692±4712 (1979). that is, thinned continental crust dissected by low-angle detachment
5. Cockayne, E. & Elser, V. Energetics of point defects in the two-dimensional Wigner crystal. Phys. Rev. B faults succeeded by exhumed sub-continental mantle, with ocean-
43, 623±629 (1991).
ward-increasing ma®c melt volumes, that merges into oceanic
6. Frey, E., Nelson, D. R. & Fisher, D. S. Interstitials, vacancies, and supersolid order in vortex crystals.
Phys. Rev. B 49, 9723±9745 (1994). crust3,6,13,14. We describe each zone in turn, adducing evidence
7. Jain, S. & Nelson, D. R. Statistical mechanics of vacancy and interstitial strings in hexagonal columnar from margins in both areas, before presenting models to explain
crystals. Phys. Rev. E 61, 1599±1615 (2000). the observations.
8. Ashkin, A., Dziedzic, J. M., Bjorkholm, J. E. & Chu, S. Observation of a single-beam gradient force
optical trap for dielectric particles. Opt. Lett. 11, 288±290 (1986).
The west Iberia margin exhibits tilted fault blocks bounded by
9. Crocker, J. C. & Grier, D. G. Methods of digital video microscopy for colloidal studies. J. Colloid Interf. west-facing normal faults5,14,15. Under the slope and rise larger
Sci. 179, 298±310 (1996). blocks lie 10±20 km apart over continental crust 20±30 km thick.
Supplementary information is available on Nature's World-Wide Web site
Oceanward, wherever crustal blocks are &7 km thick, such faults
(http://www.nature.com) or as paper copy from the London editorial of®ce of Nature. are clearly listric (downward-¯attening and concave upward), even
penetrating the mantle, and closer together; the crust tapers ocean-
Acknowledgements ward to zero thickness within a distance of 50 km (Figs 1d and
We thank S.C. Ying and D.A. Weitz for discussions. This work was supported by the 2)3±5,15,16. Such blocks, where drilled by the Ocean Drilling Program
National Science Foundation, the Petroleum Research Fund, and the Research (ODP), are capped by late Tithonian (,146 Myr ago) shallow-water
Corporation. X.S.L. acknowledges the support of the A.P. Sloan Foundation. sediments and therefore are continental crust7. Thus a marked
Correspondence and requests for materials should be addressed to A.P. change occurs in the mechanical response of the continental litho-
(e-mail: pertsin@barus.physics.brown.edu). sphere to extension once the crust has been thinned to less than
about 7 km.
In the southern Iberia abyssal plain (SIAP) three adjacent ODP
boreholes (Sites 900, 1067 and 1068; Figs 1c and 2), near where the
crust tapers to zero, yielded gabbro, amphibolite, tonalite gneiss,
................................................................. anorthosite (all part of a ,400-m-thick basement layer ¯oored by a
sub-horizontal tectonic crust/mantle contact16) and serpentinized
Evolution of magma-poor peridotite. The basement appears to be capped by a younger low-
angle detachment fault that breaks out 6 km to the east and plunges
continental margins from into basement 14 km to the west10. The layer contains pre-rift
lower crustal rocks, underplated in Late/post-Hercynian time
rifting to sea¯oor spreading (270 6 3 … 6 1j† Myr ago), that later underwent ductile deforma-
tion (at 0.6±0.8 MPa) and cooling before exhumation at the seabed
R. B. Whitmarsh*, G. Manatschal² & T. A. Minshull* (,137 Myr ago)10.
In the Alps, the continental crust of the distal Adria margin
* Southampton Oceanography Centre, European Way, Southampton, SO14 3ZH, appears as fault blocks and as isolated allochthons (masses of rock
UK that have been tectonically moved from their places of origin)
² CGS-EOST, Universite Louis Pasteur, 1 rue Blessig, 67084 Strasbourg, France separated by sub-horizontal detachments from underlying
.............................................................................................................................................. mantle11. This crust is mainly composed of Late to post-Hercynian
The rifting of continents involves faulting (tectonism) and mag- granites intrusive into poly-metamorphic basement. Sporadically,
matism, which re¯ect the strain-rate and temperature dependent underplated meta-gabbros exhumed from a pre-rift lower crustal
processes of solid±state deformation and decompression melting level are observed which intruded the crust±mantle boundary
within the Earth1,2. Most models of this rifting have treated during Late/post-Hercynian time17. Pressure±temperature±time
tectonism and magmatism separately, and few numerical simula- (P±T±t) data indicate that the crust±mantle boundary was rather
tions have attempted to include continental break-up and melt- cool (,550 8C, 0.9±1.0 MPa) at the onset of rifting18. All structures
ing, let alone describe how continental rifting evolves into sea¯oor genetically linked to the ®nal phase of crustal extension, leading to
spreading. Models of this evolution conventionally juxtapose formation of the zone of exhumed continental mantle (ZECM), are
continental and oceanic crust. Here we present observations brittle structures.
that support the existence of a zone of exhumed continental Thus, the distal continental parts of the margins are dissected by
mantle, several tens of kilometres wide, between oceanic and listric detachment faults which either separate continental crustal
continental crust on continental margins where magma-poor blocks from mantle or occasionally penetrate the mantle. Where
rifting has taken place. We present geophysical and geological pre-rift lower crustal rocks are exhumed they frequently show
observations from the west Iberia margin3±7, and geological ma®c compositions and similar pre-rift P±T±t evolutions begin-
mapping of margins of the former Tethys ocean now exposed in ning with crustal underplating followed by isobaric cooling to
the Alps8±13. We use these complementary ®ndings to propose a 550 8C. Syn-rift melt products (contemporaneous with the rifting
conceptual model that focuses on the ®nal stage of continental process) are absent. We estimate that durations of continental
extension and break-up, and the creation of a zone of exhumed extension were, respectively, a few tens of Myr (at ,5 mm yr-1) for
continental mantle that evolves oceanward into sea¯oor spread- the Tethyan margins19 and ,10 Myr (®nal phase, at .3.5 mm yr-1)
ing. We conclude that the evolving stress and thermal ®elds are off Iberia20.
constrained by a rising and narrowing ridge of asthenospheric The ZECM in the SIAP is 40- to 170-km wide with distinctive
mantle, and that magmatism and rates of extension systematically geophysical characteristics3,4. Morphologically, the basement sur-
increase oceanward. face identi®ed on seismic re¯ection pro®les may be divided into
The west Iberia±Newfoundland conjugate margins experienced a two regions of N±S-trending basement ridges and deep (,9 km)
®nal Early Cretaceous phase of rifting and sea¯oor spreading that relatively low-relief basement (Fig. 1a and c); both narrow north-
started ,133 Myr ago7. The Alpine Tethyan margins experienced a ward. Highly serpentinized peridotite was cored at four ODP
®nal Late Triassic/Early Jurassic rift phase presaging the Liguria± sites mostly near the margins of the ZECM7,21,22. Primary-phase
Piemonte ocean where sea¯oor-spreading magmatism began 160± chemistry and clinopyroxene trace-element compositions indicate
165 Myr ago13. Both the west Iberia and Alpine magma-poor heterogeneous mantle depleted by heterogeneous, less than 10%,

150 © 2001 Macmillan Magazines Ltd NATURE | VOL 413 | 13 SEPTEMBER 2001 | www.nature.com