Full text of DOI:10.1557/JMR.2002.0444

Precipitation from a reactive silicate on MgO
S.V. Yanina and C. Barry Cartera)
Department of Chemical Engineering and Materials Science, University of Minnesota,
421 Washington Avenue SE, Minneapolis, Minnesota 55455
(Received 5 November 2001; accepted 28 August 2002)
For this paper, high-temperature interactions between a reactive silicate liquid and the
(001) surface of a MgO single crystal were studied. The paper discusses the influence
of the morphology of the MgO surface on both the dewetting of a silicate liquid and
the mechanism of precipitation of excess MgO out of this silicate liquid. Alternative
pathways were considered for MgO precipitation; it may occur by nucleation and
growth of plateaus on the MgO surface or by MgO absorption at surface steps. On flat
MgO(001) surfaces, precipitation of MgO from the continuous layer of silicate liquid
led to the formation of plateaulike precipitates. Precipitation from disconnected silicate
droplets onto stepped MgO(001) surfaces resulted in the growth of ridges at steps on
the surface.
I. INTRODUCTION
II. BACKGROUND
Interactions between intergranular phases and crystal-
line grains frequently define the properties of liquid-
phase sintered ceramics.1 The complexity of
multicomponent polycrystalline starting materials usu-
ally allows numerous processes to occur simultaneously
at grain boundaries during sintering of a ceramic product,
which considerably complicates studies of such proc-
esses. A typical approach taken in ceramics research is to
identify these processes so they can be recreated sepa-
rately under controlled conditions in model systems with
fixed geometries and known chemical composition. Bi-
crystals2 or single-crystal surfaces coated by thin films
whose composition corresponds to—or is closely related
to—typical intergranular phase compositions3,4 provide
such model systems.
When ceramic materials are processed in the presence
of a liquid phase, the integrity of interfaces developing in
the sintered product depends directly on the mechanism
of interactions between the ceramic grains and the reac-
tive liquid phases present at the grain boundaries.5 The
present work illustrates directly how the morphology of
the surface of a ceramic grain can affect crystallization
of material from a liquid intergranular phase. The system
chosen for this study is the (001) surface of single-crystal
MgO in contact with a common constituent of the inter-
granular phase in MgO-based ceramics,6,7 namely, cal-
cium magnesium silicate (monticellite, CaMgSiO4).
At elevated temperatures, the processes dominating
the transfer of material at the CaMgSiO4/Mg(001) inter-
face include melting and crystallization of the monticel-
lite,8 wetting and dewetting of the MgO(001) surface by
a silicate liquid,8 and dissolution and precipitation of
MgO in and out of the silicate phase.9 According to
the CaO–MgO–SiO2 phase diagram,8,10,11 crystalline
CaMgSiO4 may melt incongruently at 1765 6 K, re-
sulting in the precipitation of solid MgO and the forma-
tion of a liquid silicate phase. It has been suggested that
this process may be accompanied by the precipitation of
merwinite (Ca3Mg[SiO4]2),12 and that the merwinite then
melts incongruently at 1773 K to give solid MgO and
liquid.11 In the present study, for which annealing tem-
peratures in the range of 1920 50 K were used, there
was no evidence for the formation of merwinite at any
stage of the process. In the present study, annealing tem-
peratures in the range 1920 50 K were used, so any
possible effect of formation or decomposition of mer-
winite on the morphology of the CaMgSiO4/MgO in-
terface would be minimized.
According to the phase diagram, the monticellite sili-
cate liquid continues to dissolve MgO up to approxi-
mately 1955 7 K,8,10 the liquidus temperature. At this
temperature, the liquid attains the nominal chemical
composition of CaMgSiO4.8 If such liquid is annealed
above approximately 1955 7 K by itself, i.e., in the
absence of additional quantities of MgO, it will behave as
a one-component system as T is increased.8,10 If, how-
ever, the CaMgSiO4 liquid is maintained in contact with
solid MgO, such a system will follow the liquidus on the
CaO–MgO–SiO2 phase diagram, and on heating above
a)Address all correspondence to this author.
e-mail: carter@cems.umn.edu
3056
J. Mater. Res., Vol. 17, No. 12, Dec 2002
© 2002 Materials Research Society

S.V. Yanina et al.: Precipitation from a reactive silicate on MgO
approximately 1955 7 K, the process of dissolution of
which are clearly visible under the silicate film, are ori-
ented along the [001] direction of the MgO. Monticellite
melts at approximately 1765 6 K;8,10 it is therefore ex-
pected that the small “particles” in the deposited film
may correspond to both crystalline and glassy regions.
As discussed above, when the CaMgSiO4/MgO(001)
MgO into the silicate phase will continue;8,10,11 this is
the usual peritectic reaction.
On cooling CaMgSiO4/MgO from approximately
1955 7 K to approximately 1765 6 K,10,11 precipita-
tion of MgO out of the silicate liquid occurs. At approxi-
mately 1765 6 K, the reaction between the silicate
phase and MgO leads to the formation of monticellite
crystals.10,11 At lower temperatures, monticellite exists
in equilibrium with solid MgO. If the CaMgSiO4/MgO
system is rapidly quenched from a temperature above
approximately 1765 6 K, the crystallization of monti-
cellite may not occur, and the silicate liquid will be
oversaturated with MgO.10 It is expected that if monti-
cellite crystallization does not take place on cooling the
CaMgSiO4/MgO system, precipitation of MgO out of
the supercooled liquid phase may still continue at tem-
peratures below approximately 1765 6 K. The present
paper specifically discusses the mechanism of the crys-
tallization of MgO from the overcooled silicate phase in
the CaMgSiO4/MgO system as it relates to the changing
morphology of the MgO(001) surface.
system is annealed at temperatures above 1771 K,8,10
a
layer of liquid phase forms on the MgO(001) surface.
When rapidly quenched to room temperature, the silicate
liquid does not crystallize, but, instead, dewets18 the
MgO(001) surface by forming dome-shaped droplets.9,19
When the quenching rate is too high, the silicate liquid
may not complete the dewetting before vitrification oc-
curs. Figure 2 shows the image of the same surface area
as that in Fig. 1 after the specimen was annealed at
1963 K and rapidly (in ഛ10 s time) quenched to room
temperature. In Fig. 2, the silicate liquid did not form
dewet droplets but instead broke into the irregularly
shaped pools of glass. It is proposed that the liquid pools
were formed in the initial stages of dewetting when the
continuous layer of the silicate phase melted and frag-
mented. The shape and position of the liquid areas on the
MgO(001) surface do not correspond to the shape and
positions of the grains in the original monticellite film
(compare Figs. 1 and 2), which indicates that the silicate
liquid spread across the MgO(001) surface in the course
of the 60-s annealing.
III. EXPERIMENTAL
Films of monticellite composition were deposited onto
single-crystal MgO substrates of (001) orientation by
pulsed-laser deposition. The MgO surfaces were pre-
pared by cleaving prior to deposition. Surfaces of clean
MgO, before and after annealing, have been discussed
elsewhere.13–16 The films were grown in 12 mtorr oxy-
gen. During the film growth, the substrates were main-
tained at 520–770 K. The average thickness of the films
varied from about 50 to approximately 200 nm.
Samples 2 × 2 × 1 mm were annealed in air in Lind-
bergTM (Lindberg, Watertown, WI) furnaces at 1870–
1970 k in closed single-crystal MgO crucibles. Upon
completion of the heat treatment, the specimens were
either quenched or slowly cooled at room temperature.
Heat-treatment times varied from 60 to 720 s. Quenching
and slow cooling took about 10 and approximately 300 s,
respectively.
Surface imaging was performed by scanning probe
microscopy in air in the contact mode using a Nanoscope
III (Digital Instruments, Santa Barbara, CA) with Si3N4
V-shaped cantilevers (Digital Instruments), which have a
nominal spring constant of 0.12 M/m.
IV. RESULTS
FIG. 1. SPM image, 20 × 20 ␮m height mode, of a thin (ഛ100 nm)
CaMgSiO4 film deposited onto a cleaved (001) MgO surface. The
cleavage steps on the MgO surface, which are seen as vertical lines in
the image, are aligned along the [100] direction of the crystal. The
average film shows approximately 50 nm granularity. The black to
white height scale is 20 nm.
Figure 1 shows a film of CaMgSiO4, with thickness ഛ
100 nm, deposited on the MgO(001) substrate. The
appearance of the continuous film resembles a dense
packing of small (approximately 50 nm diameter) par-
ticles. The cleavage steps on the substrate surface,17
J. Mater. Res., Vol. 17, No. 12, Dec 2002
3057

S.V. Yanina et al.: Precipitation from a reactive silicate on MgO
Figure 3 shows the flat CaMgSiO4-coated surface of
located adjacent to a precipitate, in contact with both the
upper and lower precipitate layers. Because such droplets
are dome-shaped, their positioning is unlikely to be
caused by incomplete dewetting of the MgO(001) sur-
face by the silicate phase.
MgO precipitation can also occur from silicate drop-
lets located at steps on the MgO(001) surface. Observa-
tions show that for droplets attached to a surface step,
MgO precipitation occurs predominantly onto the
step, rather than onto the terrace. Thus, as demonstrated
in Fig. 5, silicate droplets in contact with a large cleavage
step on the MgO(001) surface do not produce large MgO
plateaus.
In some instances, the precipitation of MgO may pro-
ceed both at the base and at the side of the dewet droplet.
As seen in Fig. 6, precipitates formed from the silicate
droplets on quenching can have a complex morphology
after annealing for 60 s at 1923 K. The lower layers of
the precipitated material surrounding each droplet in
Fig. 6 appear as plateaus. The narrow surface ridges of
MgO, which connect the silicate droplets to the steps on
the MgO(001) surface, appear to lie above these plateaus.
This observation indicates that the plateaus formed be-
fore the ridges. All the ridges shown in Fig. 6 point from
the droplets to the neighboring upper steps.
MgO(001) which was quenched at a slower rate (in about
30 s time) after the 360-s annealing at 1923 K. Precipi-
tation of MgO from the silicate phase resulted in the
formation of concentric plateaus positioned underneath
the dewet droplets. The circular precipitates in Fig. 3 are
themselves found on large irregular plateaus. It is pro-
posed that these plateaus were created during the earliest
stages of the dewetting process. The texturing of the
MgO surface between the droplets is attributed to
the formation of small MgO precipitates, which formed
prior to the completion of the dewetting transition.
If, instead of rapid quenching, the system is allowed to
cool more slowly (300 s) to room temperature, the shape
of the precipitate plateaus becomes rectangular, with
sides faceted close to the [100] and [010] directions of
the MgO crystal (see Fig. 4). The surface of the slowly
cooled sample is smoother than that of an otherwise simi-
larly treated quenched specimen (compare Figs. 3 and 4),
which indicates that, on slow cooling, the silicate liquid
tends to dewet the MgO(001) surface before extensive
MgO precipitation occurs.
Crystallization of MgO from the spherically symmet-
ric silicate droplets frequently yields irregularly shaped
precipitates. Thus, in Fig. 4, many of the droplets are
FIG. 2. SPM image, 20 × 20 ␮m height mode, after annealing the
CaMgSiO4-coated (001) MgO surface at 1963 K for 60 s and then
quenching in ഛ10 s to room temperature. The image shows the same
surface area as that in Fig. 1. The silicate film, which melted on heat-
ing, partially dewet the MgO(001) surface on cooling, forming irregu-
larly shaped fragments. The average size of the fragments is
approximately 500 nm. Large cleavage steps on MgO(001) surface are
seen under the silicate film; the smaller cleavage steps are not now
visible. The black to white height scale is 10 nm.
FIG. 3. SPM image, 5 × 5 ␮m height mode, after annealing the
CaMgSiO4-coated (001) MgO surface at 1923 K for 360 s and
quenched in approximately 30 s to room temperature. The silicate
liquid dewet the MgO(001) surface on quenching and formed hemi-
spherical droplets. The droplets sit atop concentric layers of MgO
precipitates. Low broad precipitate plateaus of irregular shape underlie
some of the droplets. The surface of the MgO substrate between the
droplets is decorated with small MgO precipitates. The black to white
height scale is 15 nm.
3058
J. Mater. Res., Vol. 17, No. 12, Dec 2002

S.V. Yanina et al.: Precipitation from a reactive silicate on MgO
The formation of whiskers was also observed when the
It is important to be sure that the precipitated materials
are MgO and not monticellite. Both the plateaus and the
ridges in question exhibit a terrace-and-ledge morphol-
ogy. The height of the ledges decorating the plateaus and
the ridges is equal to the height of the ledges observed
CaMgSiO4/MgO(001) system was repeatedly annealed.
Figure 7 shows the same area of surface as in Fig. 2; the
image was recorded after the specimen was subjected to
an additional 60 s annealing/quenching cycle at 1963 K.
The silicate phase dewetted the MgO surface by forming
droplets connected to the steps on the MgO(001) surface
by narrow ridges. These ridges, which are on average
approximately 500–1000 nm long and approximately
0.8–2.5 nm high, have flat tops and faceted sides. The
average thickness of the ridges is the same as that of steps
on the MgO(001) surface (approximately 0.2–0.4 nm).
Those in Fig. 7 connect silicate droplets to the upper
surface steps. However, the droplets in Fig. 7 do not
appear to have formed MgO plateaus.
As demonstrated in Fig. 8, the ridges may be as long as
about 1.5 ␮m when the CaMgSiO4/MgO(001) system is
maintained under the oscillatory heat-treatment condi-
tions (1920 50 K with the oscillation period of approxi-
mately 60 s) for an extended period of time (12 min).
Like those shown in Figs. 6 and 7, these ridges also con-
nect the droplets of the dewet silicate phase to the upward
steps on the MgO(001) surface. However, unlike those in
Figs. 6 and 7, many of the ridges seen in Fig. 8 curve.
on the surface of MgO(001) substrate (usually of about
1
1 Å, or integer multiples of 1 Å, which is close to ⁄
4
of
the lattice parameter of MgO structure, for which
b ס 4.2 Å). Similarly, the terraces on the plateaus and
the ridges have orientation identical to that of the (001)
plane of the MgO(001) substrate. It is unlikely that mon-
ticellite crystals [space group Ppnm, a ס 4.82 Å,
b ס 11.11 Å, c ס 6.38 Å)] could form the plateaus and
ridges, which have ledges and terraces with height
and orientation identical to those of terraces and ledges
on the MgO(001) surface.
The ridges (such as those seen in Figs. 5–8) appear to
have grown directly out of the bunched ledges on the
surface of the substrate, with no boundaries or trenches
separating them from one another. If the ridges were
monticellite, well-defined grain boundaries would sepa-
rate them from the substrate. No such boundaries were
seen in the images. The explanation for this is that the
ridges formed from MgO grown out of the droplets onto
FIG. 5. SPM image, 5 × 5 ␮m deflection mode, after annealing the
CaMgSiO4-coated (001) MgO surface at 1923 K for 360 s and cooling
in 300 s to room temperature. The direction of scanning was from left
to right. Dark and white lines correspond to down and up surface steps,
respectively. Flat surface areas appear gray. A large cleavage step on
the surface of the MgO substrate, which decomposed into a train of
smaller surface ledges on annealing, is seen as group of dark lines
extending from the top to the bottom of the image. The steps bow out
between the droplets. The bowing of the lower ledges may be due to
etching by the silicate liquid during heating. The black to white height
scale is 50 nm.
FIG. 4. SPM image, 10 × 10 ␮m deflection mode, after annealing the
CaMgSiO4-coated (001) MgO surface at 1923 K for 360 s and cooling
in approximately 300 s to room temperature. The direction of scanning
was from left to right. In the image, the dark and bright lines indicate
the down and up surface steps, respectively. The flat surface areas
appear gray. The dewet droplets sit atop the flat tiered layers of MgO
precipitates. The precipitate sides are oriented close to the [100] and
the [010] directions of the MgO substrate. The remnants of two large
cleavage steps on the MgO surface extend from the top to the bottom
of the image. The black to white height scale is 50 nm.
J. Mater. Res., Vol. 17, No. 12, Dec 2002
3059

S.V. Yanina et al.: Precipitation from a reactive silicate on MgO
the ledges on the MgO(001) surface. Furthermore, x-ray
from the silicate liquid. Growth of this island stops when
its edges are no longer in contact with the silicate phase.
This may happen either when the growing precipitate
edge reaches the air/silicate/MgO triple junction or when
the moving triple junction reaches the precipitate
edge; the droplet becomes smaller by a combination of
MgO precipitation and silicate vaporization. In Figs. 3
and 4, the lower precipitate plateaus are larger than the
upper ones, which indicates that the upper plateaus result
from subsequent nucleation events (Fig. 9). Some of the
plateaus in Figs. 3 and 4 are not associated with silicate
droplets. Such plateaus could have formed under the con-
tinuous layer of silicate liquid, or the silicate could have
already completely evaporated.
When the droplet sits on a flat MgO(001) terrace, the
silicate/MgO interface is flat. When the droplet is in
contact with a surface step (Fig. 10) precipitation of
MgO can occur on both the terrace and the step. How-
ever, as demonstrated in Figs. 5 and 7, precipitation of
MgO from such droplets occurs preferentially at the step
by the growth of ridges parallel to the (001) plane of the
MgO substrate.
diffraction analysis performed on several crushed speci-
mens did not reveal the presence of crystalline phases
other than MgO.
V. DISCUSSION
Based on these observations of the morphology of the
MgO precipitates, two major routes can be identified for
the crystallization of MgO out of the silicate liquid in the
CaMgSiO4/Mg(001) system. A MgO plateau can form at
the silicate/MgO(001) interface under the glass drop-
let. The plateau morphology is shown in Figs. 3 and 4.
Alternatively, ridges can develop where the droplet
touches a step on the MgO(001) surface (Figs. 5 and 7).
The mechanism of the MgO precipitation may change
during a single annealing/quenching cycle from plateau
growth to ridge growth (Fig. 6).
It is proposed that the formation of plateaus (Figs. 3
and 4) starts with the nucleation of a MgO island at the
silicate/MgO(001) interface under the droplet. Once
nucleated, the MgO island grows laterally [parallel to the
(001) surface of the MgO substrate]. The well-defined
tiered morphology of the precipitates in Figs. 3 and 4
suggests that only one island grows under each droplet.
The island of MgO grows primarily by precipitation
Each ridge in Fig. 5 is connected to a droplet; the
contracting droplets at steps accommodate the volume
reduction predominantly through the continuous dewet-
FIG. 6. SPM image, 5 × 5 ␮m deflection mode, after annealing the
CaMgSiO4-coated (001) MgO surface at 1923 K for 60 s and quench-
ing. The direction of scanning was from left to right. Dark and white
lines correspond to down and up surface steps, respectively. Flat sur-
face areas appear gray. The droplets sit atop flat circular plateaus of
MgO precipitates. The droplets are connected to the steps on the sur-
face of MgO substrate by ridges, which lie above the round precipitate
plateaus and connect the droplets to particular higher steps. The black
to white height scale is 10 nm.
FIG. 7. SPM image, 20 × 20 ␮m deflection mode, after annealing the
CaMgSiO4-coated MgO surface at 1963 K for two separate 60 s peri-
ods. The direction of scanning was from left to right. Dark and white
lines correspond to down and up surface steps, respectively. Flat sur-
face areas appear gray. This figure shows the same surface area as
Figs. 1 and 2. The droplets are connected to the steps on the surface of
MgO substrate by ridges. The ridges are oriented along the steepest
descent direction on the stepped MgO(001) surface. The black to white
height scale is 20 nm.
3060
J. Mater. Res., Vol. 17, No. 12, Dec 2002

S.V. Yanina et al.: Precipitation from a reactive silicate on MgO
ting of the lower terraces. Consider a narrow flat (010)
of MgO from the continuous liquid layer may remove
steps and thus smooth the MgO(001) surface. It is sug-
gested that in the case of mixed plateau/ridge MgO pre-
cipitation, the melting of the monticellite film produced
disconnected fragments rather than a continuous layer of
silicate liquid on the MgO(001) surface. On initial
quenching, the liquid precipitates plateau under these
fragments. On further cooling, when the liquid dewets
the substrate surface, the droplets attach to surface steps,
which are numerous on the MgO(001) surface (Fig. 6).
These droplets contact the upper steps located at the
edges of the plateau. Any further MgO precipitation from
step facet descending to a (001) flat terrace. If the effect
of gravitation on the droplet shape is neglected,20,21 the
droplet will make approximately a 12° contact angle9
with the (010) and the (001) planes simultaneously. As
shown in Fig. 10, such a droplet will have a larger sur-
face area than the droplet that wets the step and the lower
surface terrace. Therefore, the droplets tend to remain
attached to steps.
The formation of plateau precipitates occurs on the flat
areas of the MgO(001) surface. To avoid the nucleation
of these plateaus, the droplets must stay in contact with
the steps on MgO(001) surface (see Figs. 5 and 7). This
is unlikely for droplets located on the large flat platforms
of MgO precipitates, shown in Figs. 3 and 4. However, if
the amount of silicate liquid formed on heating is not
sufficient to produce a continuous layer, on cooling, the
droplets may contact the surface steps and grow ridges.
The development of the two precipitate morphologies
during one annealing treatment, such as that shown in
Fig. 6, requires that, upon creating plateaulike MgO pre-
cipitates, the silicate droplets remain in contact with sur-
face steps at the perimeter of the plateaus. Precipitation
FIG. 8. SPM image, 10 × 10 ␮m deflection mode, after annealing the
CaMgSiO4-coated MgO(001) surface for 720 s, cycling from 1870 to
1970 K with a period of approximately 30 s. Dark and white lines
correspond to down and up surface steps, respectively. Flat surface
areas appear gray. The droplets are connected to the curved ledges on
the MgO(001) surface by ridges. Some of the ridges are curved and
follow local changes in the direction of the steepest descent on the
stepped MgO(001) surface. The hillocks and wells on the MgO(001)
surface are expected to have formed during the precipitation and dis-
solution of MgO while the surface was covered with the silicate liquid
in the initial stages of annealing. The black to white height scale is
20 nm.
FIG. 9. Schematic of the formation of a MgO plateau under a droplet
of silicate liquid on a flat MgO(001) surface. (a) The domed droplet is
formed. (b) The MgO precipitate nucleates at the silicate/MgO inter-
face at the droplet base. (c) The MgO precipitate reaches the triple
air/silicate/MgO boundary at the droplet circumference. (d) The con-
tracting droplet dewets the newly formed precipitate.
J. Mater. Res., Vol. 17, No. 12, Dec 2002
3061

S.V. Yanina et al.: Precipitation from a reactive silicate on MgO
the droplets in contact with steps proceeds by ridge
growth. The presence of some reverse bowing of the
surface steps in Fig. 6 may indicate that the silicate liquid
in the fragments etched the surface of MgO substrate
during annealing.
As illustrated in Fig. 11, the developing ridge distorts
the domed shape of the droplet.18 As in the formation of
a precipitate plateau, the development of the ridge also
decreases the volume of the silicate phase. To restore the
wetting contact angle and to reduce the surface area,
the droplet at the ridge dewets the step but remains
attached to the tip of the ridge (Fig. 11). This suggestion
is supported by the observed widening of the ridges near
the steps and away from the droplets; this widening in-
dicates that the droplets connected to these segments
were originally larger in size. As long as the ridges con-
tinue to develop, the droplets are expected to continu-
ously dewet them by moving closer to the growing ridge
tips (Fig. 10).
The driving force for the droplet rearrangement, which
leads to the formation of the ridges, is the reduction of
the surface area of the droplet and the restoration of the
force balance at the air/silicate/MgO triple junction.
The dewetting droplets adopt shapes and positions that
will optimize the balance of forces along their circum-
ferences.20,21 The droplets attached to the ridges in
Figs. 6–8, which are only approximately 100–500 nm in
FIG. 11. Schematic of the formation of a ridge on a stepped
MgO(001) surface. In each figure, the upper schematic represents the
cross section, while the lower schematic shows the plan view. (a) A
domed droplet just touches the step. (b) MgO precipitates out of the
silicate liquid onto the step. The growing ridge distorts the surface of
the droplet. (c) The ridge grows and forces the droplet to move.
FIG. 10. Schematic of the dewetting at a surface step. (a) The droplet
dewets both the step and the lower terrace and makes the same contact
angle with both surfaces; the surface area of the droplet is large. (b)
The droplet dewets only the lower terrace minimizing the surface area
for the same volume of liquid.
3062
J. Mater. Res., Vol. 17, No. 12, Dec 2002

S.V. Yanina et al.: Precipitation from a reactive silicate on MgO
diameter, are smaller in size and have a higher surface-
surface area after two consecutive annealing/quenching
cycles (Fig. 7) show that no large precipitate plateaus
were formed during the two annealing cycles, which in-
dicates that the quenching prevented MgO precipitation
from the continuous layer of silicate liquid, while, as
Fig. 2 shows, still causing it to fragment. The layer of
silicate phase formed on melting is thin because the
deposited layer was thin (Fig. 1). As a result, on comple-
tion of the second annealing/quenching cycle, the frag-
ments of silicate layer dewet the MgO(001) surface and
form small droplets, which are nearly all about 100–
500 nm in diameter (Fig. 7). The inter-step distance in
this area of the MgO(001) surface (approximately 100–
500 nm) is smaller than the average size of the liquid
regions formed on the first quenching of the system (ap-
proximately 500–1000 nm; Fig. 2). The majority of the
dewet droplets created on the second quenching thus
contact surface steps. The formation of ridges seen in
Fig. 7 occurred in a manner similar to that of ridge
growth from droplets in Fig. 6. However, due to the rapid
first quench, the droplets in Fig. 7 did not form the pla-
teaus first.
In Fig. 7, some of the larger droplets attached to sur-
face steps did not create ridges. Steps could bow around
such droplets during the second heating cycle when re-
gions of the silicate liquid, which formed on the first
quench, wet the terraces on MgO(001) surface. The
spreading silicate liquid could then etch the MgO, start-
ing at step edges, and cause the surface steps to bow. On
the second quench, when the silicate liquid dewet the
MgO(001) surface, the large droplets in Fig. 7 would not
create ridges but instead remained over-saturated with
MgO (as suggested by their larger than average size).
This is in contrast to the droplets attached to the
ridges, which also etched the MgO(001) surface on
the second annealing, but which, on quenching, were
driven away from the etched sites by the growing ridges.
Thus, in Fig. 7, the ridges frequently connect the droplets
to large clean surface areas bound by the bunched bow-
ing steps; such sites probably indicate the original droplet
locations.
to-volume ratio than the ridge-less droplets observed in
the CaMgSiO4/MgO(001) system.9,19 The smaller drop-
lets will be more susceptible to changes in the force
balance at the triple junctions. In particular, changes in-
duced by the growing ridges exert a force on the attached
droplets. For a small droplet, the direction of this force
coincides with that of the surface tension at the site of
contact with the ridge and may exceed the surface tension
elsewhere at the droplet circumference. The growing
ridge may either cut the stationary droplet into two parts,
which is unfavorable as two new silicate/MgO interfaces
are created, or force the droplet to move to a new position
to reduce the force imbalance.
Droplet motion occurs at elevated temperatures when
the silicate liquid is sufficiently viscous. For small drop-
lets, the rate of ridge growth depends primarily on the
rate of absorption of MgO from the silicate phase and
should be less dependent on the rate at which the MgO in
the liquid is supplied to the moving interface. Therefore,
small droplets at steps are expected to create long ridges,
while large, less-mobile silicate droplets may not do so.
The ridges, which predominantly grow in the (001)
surface plane, force the droplets to move along the sur-
face in the downward direction; i.e., the droplets descend
the surface steps, since climbing steps requires the drop-
lets to perform work against gravity. Therefore, the di-
rection of alignment of ridges may not necessarily be
directly related to the crystallography of the MgO(001)
surface. In Figs. 6–8, surface steps descend in the same
direction, so all ridges appear to be crystallographically
aligned.
For the ridgelike precipitates to grow, two conditions
must be met. The surface of the MgO(001) substrate
should be stepped, and precipitation should occur pre-
dominantly from isolated droplets, rather than a continu-
ous liquid layer. The second condition is met when there
is insufficient liquid to form a continuous film on the
MgO(001) surface. Alternatively, to fragment a continu-
ous liquid layer without causing a considerable MgO
precipitation, the CaMgSiO4/MgO(001) system must be
rapidly quenched, as shown in Fig. 2. However, at low
temperatures, both the dewetting and MgO precipita-
tion are slow. To promote the growth of ridges, it is
necessary to subject the system to an additional heat
treatment. The repetitive annealing/quenching of the
stepped CaMgSiO4-coated MgO(001) surface thus ap-
pears to be a requirement to produce the growth of ridge-
like step nodes on MgO(001) surface.
The mechanism of the formation of ridges in Fig. 8
and Fig. 7 differ mainly in that those in Fig. 8 were
grown in the system, which was not brought to room
temperature during the annealing cycles. The longer heat
treatment of the sample shown in Fig. 8 resulted in the
formation of the longest ridges. The longer ridges
in Fig. 8 follow the trajectory of motion of the silicate
droplets across the more topographically complex area of
the MgO(001) surface and produced curved ridges.
The precipitation of MgO in this study is similar to the
process of liquid-phase epitaxy since the newly grown
MgO has the same orientation as the substrate. Since it
precipitates from a liquid of a different composition,
the process is also similar to flux growth of single
As demonstrated in Fig. 2, the first annealing/
quenching of the system results in the formation of
the disconnected regions of the silicate liquid on the
MgO(001) surface. The comparison between the im-
age of the MgO(001) surface under the as-deposited
CaMgSiO4 film (Fig. 1) and the image of the same
J. Mater. Res., Vol. 17, No. 12, Dec 2002
3063

S.V. Yanina et al.: Precipitation from a reactive silicate on MgO
REFERENCES
crystals. The ridges are themselves similar to single-
crystal whiskers, which grow from a liquid in the vapor–
liquid–solid process.22 Here, the ridges have a lower total
surface energy since one side of the “whisker” is not
required. This is very similar to the process proposed for
the growth of indium tin oxide whiskers along the sur-
face of a substrate.23 Finally, in this material system the
dewetting process does not occur because of the change
in composition of the glass due to heating above the
melting point;24 rather, it occurs because of the depend-
ence of surface energy on temperature.
1. D.R. Clarke, in Surfaces and Interfaces of Ceramic Materials,
edited by L-C. Dufour, C. Monty, and G. Petot-Ervas (Kluwer
Academic Publishers, Dordrecht, The Netherlands, 1989), pp. 57.
2. N. Ravishankar, H. Schmalzried, and C.B. Carter, Microsc.
Microanal. 5(Suppl. 2), 812 (1999).
3. S. Ramamurthy, H. Schmalzried, and C.B. Carter, Philos. Mag. A
80, 2651 (2000).
4. S. Ramamurthy, Ph.D. Thesis, University of Minnesota, Minne-
apolis, MN (1996).
5. R. Raj and F.F. Lange, Acta Metall. 29, 1993 (1981).
6. G.B. Rigby and H.M. Richardson, Trans. Brit. Ceram. Soc. 46,
313 (1947).
7. D.H. Hubble and N.B. Dodge, J. Am. Ceram. Soc. 43, 343 (1960).
8. E.M. Levin, C.R. Robbins, and H.F. McMurdie, in Phase Dia-
grams for Ceramists, edited by M.K. Reser (Am. Ceram. Soc.,
Columbus, OH, 1964), p. 210, Fig. 598.
VI. CONCLUSIONS
At elevated temperatures, the interface between the MgO
crystal and the silicate phase in the CaMgSiO4/MgO(001)
system is shaped by several processes which, depending on
the heat treatment, may occur simultaneously. The wetting
of the MgO(001) surface by a silicate liquid formed on
incongruent melting of crystalline CaMgSiO4 is accompa-
nied by the dissolution of MgO into the liquid on heating;
MgO precipitates out of the liquid on cooling. The dewet-
ting of MgO(001) by a silicate liquid on cooling is accom-
panied by MgO precipitation.
9. S.V. Yanina, M.T. Johnson, Z. Mao, and C.B. Carter, Microscopy
and Microanalysis, 4(Suppl 2) 590, (1998).
10. E.M. Levin, C.R. Robbins, and H.F. McMurdie, in Phase
Diagrams for Ceramists, edited by M.K. Reser (Am. Ceram. Soc.,
Columbus, OH, 1964). p. 224. Fig. 4551.
11. H-Y. Yang, Am. Mineral. 58, 343 (1973).
12. R.W. Ricker and E.F. Osborn, J. Am. Ceram. Soc. 37(3), 133
(1953).
13. S.L. King, S. McKernan, and C.B. Carter, in Evolution of Sur-
face and Thin-Film Microstructure, edited by H.A. Atwater,
E.H. Chason, M.L. Grabow, and M.G. Lagally (Mater. Res. Soc.
Symp. Proc. 280, Pittsburgh, PA. 1992), p. 661.
The MgO precipitation may occur either through the
nucleation and growth of plateaus from droplets on
MgO(001) surfaces or by deposition of MgO from drop-
lets in contact with steps on the MgO(001) surface. The
former process occurs on flat MgO(001) surfaces when a
continuous layer of silicate liquid is formed. The second
mechanism leads to the development of ridges on the
surface. Ridge growth occurs on stepped MgO(001) sur-
faces during precipitation from the disconnected regions
of the silicate liquid. The intermediate plateau/ridge mor-
phology is expected when MgO precipitation occurs out
of the disconnected liquid regions onto the stepped
MgO(001) surface.
14. S.L. King and C.B. Carter, in Evolution of Surface and Thin-
Film Microstructure, edited by H.A. Atwater, E.H. Chason,
M.L. Grabow, and M.G. Lagally (Mater. Res. Soc. Symp. Proc.
280, Pittsburgh, PA, 1992), p. 157.
15. S.L. King and C.B. Carter, 51st Annual Meeting of Microscopy
Society of America, edited by G.W. Bailey and C.L. Rieder (San
Francisco Press, San Francisco, CA, 1993), pp. 1134–1135.
16. S.L. King, J.R. Heffelfinger, M.P. Mallamaci, S. McKernan, and
C.B. Carter, in Interface Control of Electrical, Chemical, and Me-
chanical Properties, edited by S.P. Murarka, K. Rose, T. Ohmi,
and T. Seidel (Mater. Res. Soc. Symp. Proc. 318, Pittsburgh, PA,
1994), p. 407.
17. K. Reichert and B. Lampert, Vacuum 30, 383 (1980).
18. T. Young, Philos. Trans. Roy. Soc. London 95, 65 (1805).
19. S.V. Yanina and C.B. Carter, in Interfacial Engineering for Opti-
mized Properties II, edited by C.B. Carter, E.L. Hall, S.R. Nutt,
and C.L. Briant (Mater. Res. Soc. Symp. Proc. 586, Warrendale,
PA, 1999) pp. 87–91.
20. Y. Liu and R.M. German, Acta Mater. 44, 1657 (1996).
21. P. Dimitrakopoulos and J.J.L. Higdon, J. Fluid Mech. 395, 181
(1999).
22. R.S. Wagner and W.C. Ellis, Trans. Met. Soc. AIME 233, 1053
(1965).
23. H. Yumoto, T. Sako, Y. Gotoh, K. Nishiyama, and T. Kaneko,
J. Cryst. Growth 203, 136 (1999).
24. P.G. de Gennes, Europhys. Lett. 39, 407 (1997).
ACKNOWLEDGMENTS
S.V.Y. is in the Chemical Physics Program at the Uni-
versity of Minnesota. This research has been supported
by the United States Department of Energy under Grant
Nos. DE-FG02-92ER45465 and DE-FG-02-01ER45883.
The scanning probe microscope used is part of the IT Char-
acterization Facility at the University of Minnesota.
3064
J. Mater. Res., Vol. 17, No. 12, Dec 2002