Z. Phys.

D Atoms, Molecules and Clusters 20, 293-296 (1991)

Atoms.Molecules
zo,.=.,
~JrPhy~k D and Clusters

© Springer-Verlag 1991

Spontaneous coalescence in ultrafine metal particle aggregates

Y.K. Huang, A.A. Menovsky, and F.R. de Boer
Natuurkundig Laboratorium, University of Amsterdam, Valckenierstraat 65, 1018 XE Amsterdam, The Netherlands

Received 10 September 1990

Abstract. In the preparation of ultrafine metal particles below 10 nm. These materials consist of two components:
(Cu, Ti, Ni or A1) by opposed-targets dc sputtering, a A crystalline component formed by all atoms located on
substrate cooled by liquid helium flow was used, on which the lattice of the crystallites and an interfaciat component
the metal clusters and argon atoms were condensed simul- comprising all atoms situated at the grain boundaries
taneously, forming a solid layer of the mixture. By raising between the crystallites. The atomic structure of an inter-
the temperature of the substrate and the argon pressure in face is known to depend on the relative orientation of the
the chamber, the solid argon melts and evaporates, re- adjacent crystals and on the grain-boundary inclination. If
maining the metal-particle aggregates. After complete the crystatlites are oriented at random, the atomic struc-
evaporation of the argon, a phenomenon was observed in tures (e.g. the interatomic spacings) of different grain
which the metal-particle aggregates shrank and suddenly boundaries are different from each other. A nanocrystal-
changed their colour from black to bulk-metal colour, line solid with a crystallite size of 5 nm contains typically
followed by an explosion-like noise. We offer an explana- about 1019 interfaces per cm 3. The interracial component
tion for this phenomenon by the mechanism of spontane- is then the sum of the 1019 interfacial structures. If the
ous coalescence due to the large surface energy of the interatomic spacings in all grain boundaries are different,
ultrafine particles. From our observation and analysis, we the average of 1019 different boundaries results in no
conclude that a certain particle size exists, below which preferred interatomic spacings except for the one pre-
the spontaneous coalescence may occur in a very fast way, vented by interatomic penetration. Hence, the interfacial
leading to melting of the particles. The fast process of component is proposed to represent a solid-state structure
coalescence, melting and cooling introduces large internal without long- or short-range order [2, 3]. Due to these
stress which splits the sintered particle, giving the ex- features different physical and chemical properties are
plosion-like noise. This phenomenon may imply a size expected, which open the way for applications. Theretore
limitation in forming nanocrystalline solid materials of these materials have caught increasing attention in recent
pure metals. years.
Owing to the intensive research in the last few decades
PACS: 75.60.J in the fields of microclusters, catalysis and magnetic fluids
a lot of information is available on preparation methods,
physical and chemical properties and structure stabilities
of individual small particles, free or supported on a sub-
strate. New problems arise when the small particles come
1. Introduction into direct contact with each other during generation and
compaction. In this paper we will describe and discuss an
The fundamental types of solids can be distinguished observation concerning this problem.
according to their characteristic ranges of structural or-
der: crystals with long-range order and non-equilibrium
systems (glasses or other amorphous solids) with short- 2. Experimental
range order only. However, there is evidence for the
existence of a third solid-state structure in nanocrystalline We have constructed an opposed-targets dc sputtering
solid [1], characterized by the absence of both long- and equipment with a small copper block as a substrate, which
short-range order. can be cooled to 30-40 K by liquid-helium flow (Fig. 1).
Nanocrystalline solids can be formed by compacting The target materials used are copper, titanium, nickel or
pre-generated randomly oriented crystallites with size aluminum. The sputtering conditions are: background
294

I " TARGETS
j ~
COPPERBLOCK
ASSUBSTRATE K
#
>o 84K ~ / ~ aroundt70
GLASSCOLLECTOR

0 PUM Time[sec]
Fig, 1. Sketch of the experimental c o n f i g u r a t i o n Fig. 2. The temperature change during the process. At point "a",
the liquid argon was evaporated completely. The particle aggreg-
ates began to shrink, change colour and split. The thermocouple
attached to a copper foil on the bottom of the glass collector
vacuum 2 x 10 - 7 mbar, sputtering argon pressure 2 to detected some of the processes, showing several vertical steps. The
8 x 10-3 mbar, voltage 500-1000 V, current 300-600 mA. most pronounced one, appearing around 170K, is probably due to
the process of a particle aggregate situated on the copper foil
During the sputtering process, metal clusters and argon
atoms condens simultaneously on the substrate, forming a
layer of the solid-argon matrix in which the metal clusters
are embedded. When the solid layer is about 3-5 mm
thick, the sputtering is terminated. Both, the temperature temperature, at which the temperature increase is most
of the substrate and the Ar pressure in the chamber are clearly observed is around 170 K (for the copper sample).
increased to the triple point of argon (T=83.8K, Attempts to measure the temperature dependence of
p=688 mbar). The solid Ar melts and falls into a collector the resistance of the floc have failed because of the disturb-
together with the metal clusters. At the same time, the ing evaporation of the liquid argon and the quickness of
metal clusters begin to settle as black floc and the liquid the process.
argon becomes clear and transparent.
Upon reduction of the Ar pressure in the chamber
and/or by thermal conduction through the collector, the
3. Discussion
liquid Ar evaporates continuously. The temperature of the
collector remains in the range 84-87 K. No change in
In previous experiments in which we have sputtered cop-
color or morphology of the floc can be seen by the naked
per onto the liquid-nitrogen cooled substrate, the phe-
eye.
nomenon described above was not observed. Under argon
After all liquid argon is evaporated, the temperature of
atmosphere, these Cu particles were stable even at room
the collector begins to rise and the floc begins to shrink.
temperature. By x-ray diffraction it was determined that
Suddenly, some floc, beginning at the edges, turns from
the average particle size was around 70 A. In view of this,
the black color into the bulk-metal color, followed by a
we believe that the coalescence phenomenon described in
tiny stream of light white smoke and a soft explosion-like
Sect. 2 is closely connected to the size of metal particles.
noise. This phenomenon occurs discontinuously until all
Below a critical size the phenomenon can occur. Unfortu-
floc has vanished. Some material with bulk-metal color
nately, due to the instability we are not able to determine
and irregular shape remains in the collector. By x-ray
the size of the black floc at this moment.
diffraction analysis, this material turns out to beoPUre
From thermodynamic point of view, the high surface
metal with an average grain size of a few hundred Angs-
fraction in the small particles can provide an explanation
trom. Sometimes droplet (in the case of Cu) and networks
of the phenomenon of spontaneous coalescence. Consider
shape (in the case of Ti) can be found under optical
n spherical particles with initial radius ro and initial
microscope. The titanium sample is still very reactive,
temperature To which coalesce to form one spherical
which could burn spontaneously in air at room temper-
particle with radius r. If the total volume change in this
ature.
process is zero, the relation between ro, r and n is given by
In order to follow the temperature change during the
transition from the black to the bulk-metal color, we have r = nl/3ro . (1)
attached a copper-constantan thermocouple to a copper
The energy change is only due to the change in surface
foil on the bottom of a glass collector. The recorded
energy:
temperature curve is shown in Fig. 2. The result clearly
indicates that the process is an exothermal one. The A H ~urf = ( 4 n r o e n - 4 n r 2) ~' = 4 n o ~ n ( 1 - n - 1/3)roe (2)
 295

where ?Jis the specific surface energy of the material of the where T~ is the melting point of the bulk material, B is a
particle. If all the energy released is absorbed by the final constant depending on the surface and/or the interfacial
particle because of the poor heat contact with the sur- (solid-liquid) energy, the density of the material and the
rounding and the fastness of the process, the temperature heat of fnsion. Substituting relation (8) into (7), we find
of the final particle will increase to T. The enthalpy change
is equal to r o <_ [3y(1 - - n -1/3) + pcpTmB]/p[L + cp(T~ - To)]
(9)
AH T°-*T = m i cpdT = 4n
- f r3onpcp(T_ To) (3)
This implies that the melting process folloveing the coales-
To
cence of the small particles can occur even for particles
if the system does not undergo any phase transition. Here, with larger initial size.
m is the total mass, p is the density and cp is the specific From the dynamic point of view, the mechanism of
heat, in this energy consideration assumed to be temper- this process may be explained by Smotuchowski's model
ature independent. By equalizing the released energy [8] in which surface diffusion by random motion of the
AH surf and the enthalpy change AH T°~r, we obtain a entire particles is followed by collisions and coalescence.
relationship between the temperature change Nichols and Mullins [9] have considered the coalescence
AT = T - - To and the initial particle size ro: of two equal spherical particles and have calculated the
relaxation times for three distinct stages of the process: the
A T = 3~'(1 - n-1/3)/p%ro (4)
formation of a "neck" between particles, the disappear-
Taking Cu as an example, using the bulk parameters ance of the "neck" with the formation of single elongated
7 = 1.56Jm -2, cp=27.98Jmol 1K-1, p = 0 . 1 4 1 x 106 particles and the transformation of the new particle to a
m o l m 3, we get spherical one. The total time for all three stages is:
AT ~ 12000(1 -- n- 1/3)/ro, (5) z = 0.890 r~/A. (10)
where r o is expressed in the unit A. Apparently, when n Here, r0 is the radius of the spherical particle before
is sufficiently large, the particles are readily small enough coalescence and A is given by the expression
to reach high temperature (e.g. if r o < 10A then
A T > 1200 K). A =D~NoYof~2/kT, (11)
The enthalpy AH ruse needed to melt the particle if it
reaches the melting temperature is given by in which Ds is the diffusion coefficient for atomic migra-
tion at the surface of the particle, No is the average density
AH fuse = mL = 4~zr3npL/3, (6) of adsorption sites at the surface, 7o is the average surface
where L is the heat of fusion per unit mass. The energy energy and f~ is the atomic volume. From (10), we can see
requirement for the particle to melt is AH surf > AH ~r° ~Tm that the relaxation time strongly depends on the particle
+ AH fu~¢. From (2), (3) and (6) it follows size. If the particle size is large, the process will take very
long time, which means that, in practice, the process does
r o _< 3~(1 -- n-~/3)/p[L + cp(Tm-- To)J, (7) not happen. For very small particles, however, the process
where Tm represents the melting temperature. Again tak- will take place unmeasurably fast.
ing Cu with Tm = 1358K and L = 0.205 x 103 kJkg -~ as In our observation, the explosion-like process seems to
be the splitting of the particles formed by the coalescence
an example, we find that for Too= 87 K an initial particle
size r o < 6.8 A (diameter _< 14 A) is required in order to into smaller ones, which is similar to the observation by
reach the melting temperature. Ozin and Huber [4] have Ruckenstein and Malhotra [10] in platinum crystallites
shown that, when silver is dispersed to a very low concen- supported on alumina films. One of the possible explana-
tration in argon, very small Ag, clusters (with n < 10) are tions for the occurrence of the process, may be that it is
formed which could be identified by spectroscopic tech- due to the high internal stresses introduced by the fast
niques. On the other hand, for a high concentration of Ag cooling, tt might also be due to the expansion of trapped
in the argon matrix, colloidal particles with a diameter of argon.
the order of 10 A are formed, showing the broad absorp-
tion band typical for small metallic particles. In our
experimental configuration, the condition are similar to 4. C o n c l u s i o n
the later situation. Therefore it is very likely that we may
obtain the particles with the initial size required to reach From our observation and analysis, we conclude that
the melting temperature. small particles below a certain size may coalesce spontan-
From literature, there appears to be some evidence eously. The enthalpy necessary for the melting of the
that for small particles the melting temperature, T~,, may particles is provided by the large surface energy release in
depend on the size, ro, of the particle. Several models (e.g. the coalescence process. This process may imply a size
[5 7]) have been put forward to describe this dependence. limitation in the formation of nanocrystatfine solids from
In general, the expressions proposed to describe this pure metal particles. On the other hand, the phenomenon
phenomenon have the form of spontaneous coalescence may be used in the pre-
paration of some novel alloys which require additional
T~ = T,,(1 -- B/ro) , (8) energy in the formation.
296

For a full understanding of the observed phenomenon References

additional experiments are needed. Although the model of
spontaneous coalescence explains the main features, other 1. Birringer, R., Gleiter, H., Klein, H.-P., Marquardt, P.: Phys.
explanations cannot be excluded. At present, we are con- Lett. 102A, 365 (1984)
structing a low-temperature (T ~ 77 K) pressurizing unit, 2. Schaefer, H.E., Wurschum, R., Birringer, R., Gleiter, H.: J. Less-
in order to press the particles which are kept at liquid Common Met. 140, 16t (1988)
3. Zhu, X., Birringer, R., Herr, U., Gleiter, H.: Phys. Rev. B 35, 9085
nitrogen temperature into pitlets of which the physical (1987)
properties will be investigated. 4. Ozin, G.A, Huber, H.: Inorg. Chem. 17, 155 (1978)
5. Hanzen, K.J.: Z. Phys. 157, 523 (1960)
6. Buffat, Ph., Borel, J-P.: Phys. Rev. A 13, 2287 (1976)
7. Hill, T.L.: Thermodynamics of small systems New York: W.A.
This work is part of the research program of the "Stichting voor Benjamin 1964
Fundamenteel Onderzoek der Materie (FOM)', which is financially 8. Smoluchowski, M.V.: Z. Phys. Chem. 92, 129 (1918)
supported by the "Nederlandse organisatie voor Wetenschappelijk 9. Nichols, F.A., Multins, W.W.: Trans. AIME 223, 1840 (1965);
Onderzoek (NWO)". Mr. H. Gelders is kindly acknowledged for his Nichols, F.A.: J. Appl. Phys. 37, 2805 (1966)
technical assistance. 10. Ruckenstein, E., Malhotra, M.L.: J. Catal. 41, 303 (1976)