Ž .Thin Solid Films 385 2001 132�141
Role of condensates and adsorbates on substrate surface on
fragmentation of impinging molten droplets during thermal spray
Xiangyang Jiang� , Yuepeng Wan, Herbert Herman, Sanjay Sampath
Center for Thermal Spray Research, Department of Materials Science and Engineering, Uni�ersity of New York at Stony Brook, Stony
Brook, NY 11794-2275, USA
Received 16 January 2000; received in revised form 30 November 2000; accepted 14 December 2000
Abstract
We propose that the presence of condensates�adsorbates on low temperature substrate surfaces may be a significant factor
responsible for splat fragmentation of impacting molten droplets. Vaporization and rapid expansion of condensates�adsorbates
upon molten droplet impact cause instability of the spreading droplet. Plasma spraying experiments, using radio frequency
induction processing of ZrO , were designed to test this hypothesis. In order to obtain different levels of condensates�ad-2
Ž .sorbates, steel substrates were heated in a vacuum chamber at 250 torr and allowed to cool under vacuum for different periods
Ž .of time, ranging from 1 to 62 h before splat deposition. For comparison, splats were also produced on ambient 25�C as well as
Ž .on heated substrates 500�C . It was found that splat morphology changed from highly fragmented to a contiguous, disk-like
shape with a decreased level of surface condensates�adsorbates, although the substrate temperature was maintained at ambient
temperature. � 2001 Published by Elsevier Science B.V. All rights reserved.
Keywords: Condensates; Adsorbates; Molten drops; Thermal spray; Fragmentation
1. Introduction
Fragmentation of molten droplets is a commonly
observed phenomenon in many metallurgical processes,
which involve the impact of molten droplets on a
substrate or pre-deposited materials, such as splat
quenching, thermal spray, spray forming, microcasting,
� �etc. 1,2 . A fundamental understanding of fragmenta-
tion mechanisms is not only of scientific interest, but
has technical consequences as well. It is commonly
believed that splat fragmentation is generally detrimen-
� �tal to coating quality 3�6 , affecting not only the splat
� �solidification rate 3 , but also the phase selection,
deposit microstructure development and deposition
efficiency. It is also found that cracks�voids form at
�Corresponding author. Tel.: �1-309-578-6624; fax: �1-309-578-
2953.
Ž .E-mail address: jiang xiangyang@cat.com X. Jiang .�
ZrO splat and substrate interfaces in the presence of2
� �fragmented splats 4 . It is known that improved me-
chanical and physical properties of deposits are
achieved when splat fragmentation is avoided in Mo
� �and ZrO systems 5,6 . A large number of studies have2
shown that contiguous morphology of splats is benefi-
� �cial to the overall quality of coatings 3�5 .
1.1. Sur�ey of the fragmentation mechanism for molten
droplets
A suitable mechanism to explain the fragmentation
of impacting droplets has not been developed. The
fragmentation of non-thermal spray particles, such as
rain or fuel droplets which do not experience a phase
change during deposition, has received considerable
attention in the field of fluid dynamics, where fragmen-
tation is generally attributed to fluid flow instability,
� �such as Rayleigh�Taylor or Weber instabilities 7�9 .
0040-6090�01�$ - see front matter � 2001 Published by Elsevier Science B.V. All rights reserved.
Ž .PII: S 0 0 4 0 - 6 0 9 0 0 1 0 0 7 6 9 - 6
( )X. Jiang et al.�Thin Solid Films 385 2001 132�141 133
With an increase of the Reynolds or Weber numbers,
� �the fragmentation tendency increases 7�9 . However,
in the case of thermal spray droplets, such as molten
metal and ceramic droplets, fragmentation is pre-
sumably still governed by these fluid dynamic parame-
ters, but in a much more complex way due to the strong
thermal interactions between the droplet and the subs-
trate. It was found as early as the 1920s that substrate
temperature has a strong effect on splat morphology.
With an increase in substrate temperature above am-
bient, splat morphology changes gradually from highly
� �fragmented to contiguous 10 . This phenomenon has
� �been widely confirmed 11�13 . Droplet materials in-
clude ZrO , Al O , Al, Ni, Cr, Cu, Mo, etc. and2 2 3
substrate materials such as ceramics, glass and steel
have been used. There is an effective substrate ‘transi-
tion temperature’, below which splat fragmentation oc-
curs and above which the splats are contiguous and
generally well formed. The reported transition temper-
atures for different droplet�substrate combinations fall
into a relatively narrow range. Fukumoto et al. suggests
there is a critical fragmentation�non-fragmentation
transition temperature for each particular droplet�sub-
� �strate combination 12 .
The substrate temperature effect remains an enigma,
with the controlling mechanism still unclear. Most ex-
planations attribute this phenomenon to the solidifica-
� �tion of the impacting droplet 12,14,15 . Differences
among the theories depend only on the details of the
solidification effects. Fukumoto et al. suggests that
fragmentation results from rapid solidification at low
substrate temperatures, in addition to high droplet
� �surface tension 12 . They further argue that the initial
solid layer which forms just after the impingement
� �causes fragmentation through melt jetting 14 . It has
also been postulated that higher substrate tempera-
tures enable the melt to be maintained above a ‘hyper-
cooling temperature’, avoiding the local solidification
and, therefore, the creation of perturbations during
� �spreading 15 .
Other explanations involve wettability theory, in that
improved wettability for a higher substrate temperature
may be responsible for the splat morphology change
� �14,16 . Severe oxidation of the substrate surface is
suspected as a major contributor to fragmentation.
Splat morphology changes from contiguous to frag-
mented when the steel substrates are heated for ex-
� �tended periods or at higher temperatures 16,17 .
1.2. Problems with the a�ailable fragmentation theories for
molten droplets
None of the hypotheses mentioned above can be
used to explain all of the reported experimental results
for different droplet-substrate systems. In our view, it is
unrealistic to apply solidification arguments to systems
as diverse as Mo-on-Mo and ZrO -on-ZrO , which2 2
differ in solidification rate by several orders of magni-
tude, i.e. a 100�C or 200�C rise in substrate tempera-
ture cannot be expected to significantly change the
solidification behavior of a droplet with a melting point
above 2500�C. Also, in some cases, solidification is the
arresting factor in droplet spreading. Thus, splats can-
not remain molten before the spreading is completed
� �18,19 .
Substrate oxidation can not explain the fragmenta-
tion for low substrate temperatures, the trend remain-
ing the same when splats are produced on glass or
� �oxide substrates 16,18 . The reported transition occurs
at approximately 100�400�C without an obvious depen-
�dence on droplet�substrate combination 3,5,6,-
�10,12,14 . These facts strongly suggest that some factor
independent of substrate and droplet materials are
responsible for the morphology transition.
1.3. Proposed hypothesis
We propose a material-independent mechanism,
which is consistent with most experimental observa-
Ž .tions, adsorbed gas�condensation AGC on the sub-
strate surface. During previous experiments, we
observed no deposition of Mo droplets on glass subs-
trates cleaned with alcohol, but found instead that Mo
deposits well on untreated substrates. Sampath and
Herman reported more contiguous Ni splats formed in
a reduced pressure chamber than at atmospheric pres-
� �sure 20 . Furthermore, additional experimental evi-
dence has accumulated, including the substrate tem-
perature effect, which seems to support this concept.
Raising the substrate temperature is expected to effec-
tively eliminate the AGC. It is known that water and
other substances can be adsorbed on clean solid sur-
faces. Desorption tends to occur when the temperature
rises. Adsorption�desorption kinetics and reaction di-
rection is a function of temperature, adsorbate species
and solid surface structure and surface energy. Evap-
oration of the AGC upon the impact of hot droplet
may cause the fragmentation.
A literature survey on the effects of AGC on droplet
deposition has yielded only one relevant paper, Li et al.
address the effects on the splat morphology of organic
� �films brushed on a substrate surface 21 . They observed
a splat morphology transition upon the removing of the
organic film through heating.
The purpose of the present study was to find clear
experimental evidence that confirms the role of AGC
on droplet fragmentation.
2. Experimental design
Experiments were designed to prepare splats on sub-
strates with different AGC levels, while keeping all the
( )X. Jiang et al.�Thin Solid Films 385 2001 132�141134
Ž . Ž .Fig. 1. Chamber pressure a and temperature b experienced by substrates before splats preparation. Gradual chamber pressure increases due
to leaking is recorded after the main vacuum pump is shut down. The pressure for run 4, 3 reached approximately 8 and 10 torr, respectively,
before the refilling of argon and it reached 50 torr for run 2. Pre-heating and spraying was at a pressure of 250 torr. Run 1: no pre-heating; run
2,3 and 4: Pre-heating, cooling down and spray after 62, 17 and 1.2 h, respectively; run 5: pre-heating and spraying.
other conditions constant. By comparing splat mor-
phology, the role of AGC can be inferred. Based on
Ž .this approach, low pressure radio frequency RF
plasma spraying was used to prepare splats in a protec-
tive inert gas environment. Polished steel substrates
Ž .25�50�3 mm were placed in the chamber, which
was then evacuated to less than 0.1 torr. Following this
step, the chamber was refilled with pure argon to 250
torr, which is necessary for the RF plasma to operate.
For each run, the substrate was heated with the plasma
torch for 10 s to approximately 500�C, then the cham-
ber was evacuated to less than 0.1 torr and the subs-
trate was cooled down for a given period of time. The
chamber was then refilled to 250 torr and molten
droplets were deposited on each substrate. For compar-
ison, one splat specimen was prepared on an unheated
substrate and another one was prepared immediately
after the substrate was heated. The substrates were
polished to a roughness of R �0.05 �m and cleaneda
with acetone. The pre-heating and cooling histories of
the substrates are shown in Fig. 1. Substrate tempera-
tures are estimated with a one-dimensional heat con-
duction model to be 25, 25 and 50�C for runs 2 through
4, respectively.
Splats were deposited with a TAFA Model 66 Radio
Frequency Plasma system. The experimental setup is
shown in Fig. 2 and the spraying conditions are listed in
Table 1.
Ž . ŽNorton �235 partially stabilized ZrO PSZ 5.12
.wt.% calcia powder with a mean particle size of 63 �m
was used in this investigation.
Optical microscopy and scanning electronic micros-
Ž .copy SEM were used to observe splat morphology.
Secondary electron, back-scattering and specimen cur-
rent images taken of the as-sprayed splats were used to
evaluate the actual splat contact on the substrate.
During SEM observations, an electrostatic charge builds
up in the insulating PSZ splat, in the good contact
area, the charge leaks to the substrate and forms the
specimen current signal, therefore, bright areas in the
specimen current image represent ‘good’ splat�sub-
� �strate contact areas 22 .
3. Results
Table 2 summarizes the substrate heating histories
and the resulted substrate surface AGC levels, splat
morphologies, etc. The results shown in Table 2 are
discussed in the following sections. It is clear that splat
morphology changes from highly fragmented to con-
tiguous with a decrease in AGC level on the substrate
surface, along with an increase in actual contact area
and crack density.
3.1. Morphologies of splats prepared under different
conditions
It is observed that the splats prepared on the un-
Ž .heated substrate sample 1, shown in Fig. 3a have
highly fragmented shapes which generated a large
quantity of debris. There are many small holes through
each splat with pits and humps located throughout the
splats.
Table 1
Spraying parameter
Plasma gas and Carrier gas Current Voltage Torch�substrate Substrate Substrate
Ž . Ž .flow rate and flow rate A V distance heating time rotating rate
Ž . Ž . Ž . Ž . Ž .slm slm cm S rev.�min
Ar100�He 140 He 4 10 5, 700 26 10 120
( )X. Jiang et al.�Thin Solid Films 385 2001 132�141 135
Fig. 2. Schematic diagram of the experimental setup for splat prepa-
Ž .ration not to scale, vacuum chamber not shown .
Splats prepared on substrates that were pre-heated
and then cooled down to ambient temperature in vac-
uum, show very different morphologies. On a substrate
maintained under vacuum for 62 h, most splats pre-
Ž .pared after the pre-heating sample 2 show a much
more uniform morphology with clear flow patterns and
long projections along the periphery of the splats, Fig.
3b displays highly fragmented, irregular shaped splats
and debris. This suggests that the fragmentation ten-
dency is still very high, but in a much more regular and
uniform way.
ŽOn the substrates allowed to cool for 17 h sample
.3 , the splats have essentially disk-like shapes with a
distinguished rim, humps can be found in the central
Ž . Žarea of some splats Fig. 3c . Only a few splats less
.than 10% with less contiguous shapes were observed.
Cracks can be seen in some splats.
Ž .On the substrates allowed to cool for 1 h sample 4 ,
Žthe splats have very contiguous, disk-like shapes Fig.
.3d . A crack grid pattern can be discerned in the
central part of most splats, but not on the rim. In some
smaller splats, crack grids are well developed through-
out the entire splat, including the rim.
Splats prepared immediately after substrate heating
Ž .sample 5 show a contiguous, disk-like shape with
Ž .well-developed crack grids on all splats Fig. 3e . Com-
pared with the splats in sample 6, the crack grid is
much finer and exists throughout the surface of the
entire splat.
3.2. Effecti�e contact�adhesion of splats on the substrate
It is noted that the splat-substrate adhesion is greatly
enhanced when the substrate is kept in vacuum for
shorter periods of time between pre-heating and splat
deposition. For the splats produced on unheated subs-
Ž .trates sample no. 1 , most splats are loosely attached
to the substrate and many splats are lost while handling
the substrate during optical observation, with only some
debris remaining. Splats produced 62 h after substrate
pre-heating shows some improvement. With the short-
ening of the dwell-time in the chamber, most splats
remain on the substrate. Splats deposited on heated
Ž .substrates sample 5 adhere well. It has been shown
that while splats prepared on cold substrates are loosely
bonded, the splats produced on hot substrates can
� �sustain indentation without delaminating 23 .
Deposits prepared on hot substrates also show
greater adhesion over those prepared on cold sub-
� �strates 16 . This is supported by the observation of
crack patterns in the splats. Cracking is a mean of
stress relief caused by the constraint of the splat to the
substrate during quenching and cooling. The appear-
ance of cracks indicates that the adhesion has im-
proved. Adhesion is greatly enhanced when the time
interval between pre-heating and spraying is shortened.
SEM current images of the splats reveal dramatic
differences in the contact ratio of splats on each of the
different substrates. Splats produced on unheated subs-
Table 2
Summary of substrate temperature history, surface AGC level and splat morphology
Ž .Sample run no. 1 2 3 4 5
Substrate temperature Ambient Ambient�500�C ambient�500�C Ambient�500�C Ambient�500�C
history �ambient �ambient �50�C
Cooling time No heating 62 17 1.2 0
Ž . Ž .after heating h infinite
AGC level Very high Medium Moderate Low Very low
Splat Highly Fragmented Mostly Contiguous Highly
morphology fragmented contiguous contiguous
Contact area Very Low Medium High Very high
low�none
Crack density No No Medium�low High Very high
( )X. Jiang et al.�Thin Solid Films 385 2001 132�141136
Fig. 3. Morphology of typical splat prepared on pre-treated subs-
trates substrate condition.
Ž .trates sample 1, Fig. 4 a2 show virtually no contact
with the substrate, splats from sample 2 show an in-
Ž .creased contact area at the center Fig. 4 b2 , splats
from sample 3 show even larger ‘good’ contact areas
Ž .Fig. 4 c3 . Splats from samples 4 and 5 show much
larger contact area, but the contact pattern changes
from a large patch to an area of mixed good�poor
contact patches. This change is accompanied by in-
creased crack densities and reduced splat lift-over at
the rim. In all cases, the contact at the splat rims is
poor. The back-scattering electron images of the splats
reveal that the dark areas in the secondary electron
images are not indications of holes, but an artifact due
to the electrostatic charging effect.
The contact area pattern change is probably due to
splat cracking. Upon impact and spreading of the
droplet, the contact at the center is good and continu-
ous. This contact exerts a constraint on the splat and
causes cracking. Splat segments lift near the crack
edge, causing the splats to detach from the substrate
and form poor contact areas.
4. Discussion
4.1. Analysis of the substrate surface conditions
Before we can attribute the above experimental re-
sults to AGC on the substrate surface, it is necessary to
first analyze the surface condition, such as surface
oxidation, roughness, temperature and the levels of
substance condensates�adsorbed for different experi-
mental cases.
4.1.1. Surface oxidation� roughness
A thin oxide film is always present on steel exposed
to atmosphere. No significant further oxidation is ex-
pected because substrates were maintained in vacuum
during pre-heating and subsequent cooling. Therefore,
a change in the wettability of a molten droplet on the
steel substrate due to the surface oxidation after sub-
strate pre-heating, is unlikely.
The substrate surface was also inspected with optical
microscopy before and after depositing the splats for
possible changes in surface roughness. No change was
observed.
4.1.2. Substrate temperature
Substrate surface temperatures dropped to ambient
after 1 h of cooling as shown in Fig. 1 for samples 2, 3
and 4. This simple one-dimensional heat conduction
model gives a conservative result by neglecting heat
loss through radiation and convection. Substrate tem-
perature differences, therefore, are not the reason for
the observed splat morphology change under these
conditions.
4.1.3. Adsorption and condensation on the substrate
surface
It can be inferred from the following analysis that
the different levels of AGC were achieved in different
Ž .experimental runs i.e. samples in this investigation.
It is well established that clean surfaces energetically
attract foreign species, which results in adsorption and
condensation of molecules. Condensed volatiles vapor-
ize when the vapor pressure is lower than the satura-
tion vapor pressure at a given temperature. The evap-
oration rate increases with temperature and also with a
decrease in partial pressure of the condensed species.
The most common condensate is H O, which has a2
vapor pressure of 17 mmHg above that of the liquid at
room temperature. In general, the equilibrium volume
of an adsorbate sharply decreases with an increase in
temperature and a decrease in pressure. Multi-layer
adsorption can take place and the adsorbed volume can
be very high and possibly go to infinity as t