DYNAMICS OF SUCCESSIVE DROP IMPACTS ON A SOLID SURFACE

MICHAEL MEADEN & EMILY MEISSEN

Abstract. This paper compares the impact and spread velocities of a liquid drop impacting on a solid
surface. It also studies a second, consecutive drop by looking at the maximum spread factors reached by
the first and second drops and the number of secondary droplets (a characterization of splashing) caused by
the impact of the second drop.

1. Introduction

Scientists have studied a liquid drop impacting a solid or liquid surface for over one hundred years.
Worthington [5] pioneered the subject in the early 20th century by investigating drop impacts in a deep
fluid using high speed photography. Research in macro liquid drop impacts on solid surfaces and liquid
films is helpful in understanding the growth and splashing patterns of raindrops, which has important
implications for soil erosion as well as ice accumulation on power lines and aircraft. Research has focused
on a single drop impacting a solid surface or a uniform liquid film (see Yarin [6] for a comprehensive
review); there is little regarding multiple drops impacting successively on a solid surface. In this paper,
proposed scaling laws are experimentally tested and extensions to a second, consecutive impact are explored.
The lamella is defined as the primary part of the liquid film created after impact.
When a single drop impacts a surface, the phenomena of fingering a quasi-periodic
lamella
unevenness of the boundary of the lamella and splashing a breaking-off of such diameter
fingers from the lamella into the air or along the surface (see Figure 1) can occur.
The fingers which break off during splashing are referred to as secondary droplets. Such
phenomena occur when multiple drops impact successively on a surface as well, though
this case is largely unexplored. (a) Fingering, or
unevenness of the
boundary.
d Drop Diameter v Characteristic Velocity secondary
Surface Tension Dynamic Viscosity droplet
Fluid Density g Gravitational Acceleration
Table 1. Relevant parameters.
lamella

dv 2
Weber Number We = Inertia vs. Surface Tension

(b) Splashing, or
dv
Reynolds Number Re = Inertia vs. Viscous Forces breaking off of fingers
from the lamella.
2
gd
Bond Number Bo = Gravitation vs. Surface Tension
Figure 1.
v2 Phenomena
Froude Number Fr = Inertia vs. Gravitation
gd occurring from
Table 2. Non-dimensional groups using parameters from Table 1. impacting liquid
drops.

The common non-dimensional groups for this topic, using the relevant parameters
listed in Table 1, are listed in Table 2. When the Bond number is less than one, as was the case in

Date: December 15, 2011.

1
the experiments presented, gravitational forces are negligible, and the Weber and Reynolds numbers are
considered to be the most important non-dimensional groups [4].

Many liquid drop impact phenomena are related to the Reynolds and Weber numbers, which can be
calculated using either the impact velocity or the spread velocity. Most previous studies have used the
former. Finding the correlation between these two velocities would explain how the relations found depend
on which velocity is used. An investigation of the relationship between the impact velocity of a drop and
the spread velocity of the lamella directly after impact is presented. The maximum spread factor, , defined
as the ratio of the maximum lamella diameter to the drop diameter before impact, is also found for both
the first and second drop impacts. Many scaling relationships, both theoretical and empirical, have been
proposed for this maximum spread factor, and a vast majority of these relate the spread factor to Re and
W e, though many of these relate specifically to certain types of fluids or experimental conditions. The
splashing behavior is characterized by the number of secondary droplets, N , caused by the second impact,
and possible scaling relationships for N is also explored.

For all experiments presented, the impact velocity was varied by adjusting the height from which the
water droplets were released. For water, this gave a range of approximately 100 to 300 for the impact Weber
number; and a range of approximately 5,500 to 10,000 for the impact Reynolds number. For ethanol, this
gave a range of approximately 150 to 600 for the impact Weber number; and a range of approximately 2,500
and 5,000 for the impact Reynolds number.

2. Experimental Methodology

An elevated syringe with a 16 gauge needle (corresponding to an outer

syringe
diameter of 1.62 mm) was used to release successive water drops to or-
thogonally impact a glass surface (see Figure 2). The height from which
drop #2 horizontal the drop was released was measured from the glass surface to the tip of
the syringe. A high-speed camera was used to capture videos at 2000
camera

drop #1
fps. The glass surface was cleaned with distilled water and dried with a
micro-fiber cloth between trials. The data for each height were averaged
glass surface
over four trials.

The relevant parameters for analysis are listed in Table 1. The surface
vertical camera tension, dynamic viscosity, and density of water at room temperature are
known to be 0.07197 N/m, 0.00089 Ns/m2 , and 1000 kg/m3 , respectively.
The respective values for ethanol at room temperature are known to be
Figure 2. 0.02227 N/m, 0.001074 Ns/m2 , and 789 kg/m3 .
Experimental setup. Most ex-
periments used a vertical camera
view. Measuring the impact ve- 2.1. Drop Diameter. To determine the average water drop diameter,
locity required a horizontal cam- twenty drops were sampled out and the mass of the total pool was mea-
era view. sured after each drop. The volume of each drop, V , was deterimed based
on the difference in mass and the known fluid density. p
Assuming a spheri-
cal drop, the drop diameter was calculated using d = 3 6V /. Averaging
over the 20 drops, the average water drop diameter was 0.3778 cm with a standard deviation of 0.0070 cm
less than two percent of the average diameter. Similarly, the average drop diameter for ethanol was found to
be 0.2761 cm with a standard deviation of 0.0026 cm. These values were used when calculating the Weber
number and maximum spread factor.

2.2. Maximum Spread Factor. Recall the definition of the maximum spread factor to be the ratio of the
average diameter of the lamella at the time when this is maximized to the drop diameter (found above). To
find this, a sequence of images was taken with the vertical camera view. Taking the frame in which the size
of the lamella stopped increasing, an edge-detection algorithm (see Appendix) calculated the area, A, of the
2
 (a) original (b) edge() (c) dilate (d) fill/erode

Figure 3. Images produced using the edge-detection algorithm to find the maximum spread
factor. Height of 150mm, drop 1. Maximum lamella diameter: 0.012 m.

p
lamella. The diameter was then calculated as dmax = 2 A/. This process was repeated for the second
drop.

The edge-detection algorithm first used MATLABs edge() routine. In most cases, this did not detect
the full boundary of the drop. The algorithm then dilated this region to merge the boundary and create
a connected component (the lamella). The largest connected component was taken and filled. Then the
boundary was eroded to undo the growth of the boundary caused by the dilation (see Figure 3). Summing
the pixels in the region gave the total area encapsulated. This method worked in all cases for the first drop
in which the lamella had a visible boundary. The second drop usually had regions of the boundary which
were barely visible to the human eye. After manually drawing in these parts of the boundary, the algorithm
successfully determined the diameter.

2.3. Impact Velocity. To determine the impact velocity, a sequence of images was taken from the horizontal
camera view. This view showed the final 5 cm of the fall before the water drop impacted the glass surface.
Tracker [1], an open-source program, was then used to gather data from the sequence. The programs
auto-track feature tracked the position of a selected high-contrast region on the drop. The position data
collected was used to calculate the impact velocity. Figure 4 shows data for a typical trial.

2.4. Spread Velocity. Tracker was used to determine the initial spread velocity of the lamella from the
vertical camera view. Defining the origin to be at the center of the drop impact, the edge of the lamella
along an axis was tracked manually in each frame until the lamella stopped spreading. Figure 5 shows sample
position data for a given trial.

2.5. Splashing Behavior. Using the image sequences from the vertical camera view, the number of secondary
droplets which completely disconnected from the lamella during the second drops impact was counted. This
number characterized the splashing behavior for a given Weber number. Secondary droplets could have
arisen from droplets being jettisoned from the lamella immediately upon impact of the second drop or from
fingers disconnecting from the lamella during spreading. Liquid residue left on the glass after the lamella
receded was not considered a secondary droplet.

3. Experimental Results

3.1. Relating Impact Velocity and Spreading Velocity. Tracking the impacting drop from the horizontal cam-
era view revealed a nearly linear position curve over time, shown in Figure 4. The impact velocity was taken
as the average velocity over the final ten tracked frames. The vertical perspective videos used to measure
the initial spread velocity of the lamella showed only 9-12 frames between when the drop first impacted the
surface and when the lamella stopped spreading. Because the lamella decelerates rapidly, the initial spread
velocity was calculated using only the first two frames of impact.
3
 Figure 4. Example of position data given by Tracker for drop height of 135 mm, trial 1
when tracking a falling water droplet. The coordinate axis was specified such that the origin
was approximately in the center of the drop and it moved along the negative x-axis. The
calculation of the impact velocity used the final ten points.

Figure 5. Example of tracking data given by Tracker for drop height of 135 mm, trial
1 when tracking the rim of the lamella for a water droplet impact. The calculation of the
spread velocity used the first two points.

The measured impact velocities are shown in Figure 6 and the measured spread velocities are shown
in Figure 7. The position of measured impact velocities relative to the predicted values assuming no air
resistance (shown as the theoretical fit in Figure 6) implies that there was both air resistance and a non-zero
initial velocity, giving the ethanol data a higher predicted velocity due to a smaller drop size. The curves
shown in Figures 6 and 7 relate the impact and spread velocities to the square root of the height from which
the drop is released, which we infer from the theoretical fit in Figure 6. If these relations are true, we would
expect that the ratio of the spread velocity to the impact velocity, shown in Figure 8, would be linear.

3.2. Maximum Spread Factor. Using the edge-detection method to find the maximum diameter of the
lamella, a strong positive correlation was found between Weber number and the maximum spread factor
4
 2.5
2.5
(m/s)

2
(m/s)

2
Impact
Impact Velocity
Velocity

1.5
1.5 v = 2gh
v = 2gh
water
water
ethanol
ethanol
theoretical
theoretical
1
60
1 80 100 120 140 160 180 200 220 240 260
60 80 100 120 140 Height
160(mm) 180 200 220 240 260
Height (mm)

Figure 6. Impact Velocity vs. Height. The theoretical curve corresponds to the predicted
impact velocity assuming no air resistance and zero initial velocity.

11
water
ethanol
10 fit water
fit ethanol
Spread Velocity (m/s)

y = 0.466 x1/2 + 2.518

8 R = 0.931

y = 0.260 x1/2 + 3.160

R = 0.885
7

5
60 80 100 120 140 160 180 200 220 240 260
Height (mm)

Figure 7. Spread Velocity vs. Height. The fits relate the spread velocity to the square
root of the height, a relation inferred from the predicted fit in Figure 6
.

for the first and second drops, as shown in Figure 9. The difference between the spread factors of the first
and second drop appeared to be approximately constant for the regime considered. The ethanol and water
data appear to have distinct trends when viewed against W e.

3.3. Splashing Behavior. There was a positive correlation between the number of secondary droplets and
both the Weber number and the Reynolds number. Experimental conditions, such as temperature and
5
 11

10

9
Spread Velociy (m/s)

y = 2.986 x + 3.206
8 R = 0.904

6
y = 1.637 x + 3.299
R = 0.896 water
ethanol
5
fit water
fit ethanol
4
1 1.5 2 2.5
Impact Velocity (m/s)

Figure 8. Spread Velocity vs. Impact Velocity. We expect a linear relationship, as men-
tioned in Results and Discussion.

6.5

5.5

0.8 4.5
10 4

3.5

3
0 100 200 300 400 500 600 700
Spread Factor

10
0.7 y = 1.050 x1/4 + 1.315
R = 0.901

1/4 water, drop 1

0.6
y = 0.546 x + 2.058 water, drop 2
10 R = 0.901
ethanol, drop 1
ethanol, drop 2
fit all drop 1
fit all drop 2

2 3
10 10
We

Figure 9. Maximum spread factor vs. Weber number loglog plot. The smaller plot shows
a non-loglog plot. Both drop 1 and drop 2 exhibit a positive correlation with the Weber
number for ethanol and water data. The water and ethanol data do not coincide well,
implying some other factor may also be used to determine the spread factor.

6
 60
water, day 1
water, day 2
50 water, day 3
Number of Secondary Droplets

ethanol, day 4

40

30

20

10

0
50 55 60 65 70 75 80 85 90 95
3/16 3/8
We Re

Figure 10. Number of secondary droplets vs. W e3/16 Re3/8 . Note that the splashing
thresholds and data do not align well between the ethanol and water data, implying that
there is some other factor which may be used to predict the number of secondary droplets.

humidity which changed between testing days, appears to have shifted the splashing threshold, which made
it difficult to resolve a precise trend. Marmanis and Thoroddsen [3] predict the number of fingers to scale
with W e3/16 Re3/8 . Because secondary droplets result from fingers detaching from the lamella, the number
of secondary droplets should also follow this relationship. Figure 10 shows a plot of the number of secondary
droplets against W e3/16 Re3/8 . This relationship does not give an accurate representation of the number of
secondary droplets since the ethanol data shows a greater number of secondary droplets for smaller values
of We3/16 Re3/8 . Figure 11, however, shows that the number of secondary droplets may instead scale with
(W eRe)3/16 ; when plotted against this, the ethanol and water data better coincide and have similar splashing
thresholds.

4. Discussion

Throughout data acquisition and analysis, many inconsistencies in the data became apparent. These
inconsistencies were most likely the result of variations in experimental conditions, as well as possible human
error, or possible error in the tracking and edge-detection methods. Experimental variations may have
occurred in: drop size, height measurement, angle of impact, initial velocity, room temperature and humidity
(which affects and ), location of second drop impact relative to first, time between impacts, and the shape
of the drop during fall or at impact. While most of these variations were likely very small, they may have
still had an effect on the data, especially near critical cases, such as the splashing threshold. General trends
in the data, however, could still be observed. The following section offers some discussion and theoretical
background related to the results seen above.

4.1. Impact vs. Spread Velocity. Roisman [4] provides a theoretical model of the initial phase of impact of
a liquid drop on a solid surface. Figure 12 shows a schematic of the drop during the initial phase of impact,
assumed to be valid from t = 0, when the droplet first touches the surface, until the crash time tc = d/U0 ,
where d is the initial drop diameter and U0 is the impact velocity. Figure 12 shows three distinct regions
of the liquid during this time. Region 1 is assumed to be unperturbed and behaves as a rigid body moving
down towards the surface with velocity U0 . This causes a high pressure gradient to form in region 2, which
7
 60
water, day 1 5
water, day 2
4
50 water, day 3
3
ethanol, day 4
Number of Secondary Droplets

2
40
1

0
13 13.1 13.2 13.3 13.4 13.5 13.6
30

20

10

0
10 11 12 13 14 15 16 17
(WeRe)3/16

Figure 11. Number of secondary droplets vs. (W eRe)3/16 . Here, the splashing thresholds
of the water and ethanol data align well (given sources of error between trials). The smaller
plot zooms in on the data near the splashing threshold.

forces liquid radially outward into the lamella in region 3, where pressure is assumed to be uniform over the
entire region.

R0
1
2 3 h
D

Figure 12. Schematic of drop impact during the initial phase of impact.

The equations of motion for region 2 dictate the relationship between the impact velocity and the spread
velocity. If inertial terms are neglected, this region can be modeled as a viscous creeping flow between
two approaching discs. Using cylindrical polar coordinates, the Navier-Stokes equations and the continuity
equation for fluid flow give the equations of motion

2 vr p p 1 (rvr ) vz
2
= =0 + =0
z r z r r z
with boundary conditions

vr = vz = 0 for z = 0 vr = 0, vz = U0 for z = h p = p0 for r = R

8
where p is pressure and R is the radius of region 2. Solving this system yields an expression for the radial
velocity in terms of the impact velocity:
3U0
vr = 3 rz(h z).
h
For a given height z and radius r, the relationship between the radial velocity and the impact velocity is
predicted to be linear by Roismans model; for sufficiently small time, the radial velocity at the edge of
the lamella corresponds to the initial spread velocity, implying a linear relationship between the impact
and initial spread velocities. The observed relationship between the initial spread velocity and the impact
velocity shown in Figure 8 appears to support this theoretical prediction of a linear relationship reasonably
well, though the experimental results show that the slope of the linear fit is not a constant for all liquids,
indicating a possible dependence on fluid properties. It is important to note that Roismans model is valid
only when working with sufficiently large impact velocities that are still less than the splashing threshold
for a single drop and only when the ratio of inertial forces to viscous forces is small. While these conditions
are met for the liquids and velocities considered in this paper, this argument cannot be extended to all drop
impact situations in general.

4.2. Maximum Spread Factor. Clanet [2] argues that, for fluids with low viscosities, should scale with
W e1/4 . However, despite the fact that both ethanol and water have low dynamic viscosities, Figure 9 shows
that such a scaling relation does not provide a good fit when considering the water and ethanol data together,
for either the first or the second impact. The ethanol data does not coincide well with the trend predicted
from the water data. This might imply that there is another factor in predicting the spread factor. When
looking at just the water data, as in Figure 13, a scaling relation with W e1/4 may be valid. The difference
between the first and second drops spread factors appears to be constant. An estimate for this difference
can be found using a basic conservation of mass argument (i.e. conservation of volume, given a constant
density).

If the lamella is approximated by a cylinder of diameter dmax and height h, then conservation of volume
would imply
4d3
2hd2max = ,
3
giving a maximum spread factor for the first drop of
r
2
1 = dmax /d = ,
3h
where h = h/d is the non-dimensional lamella thickness. The change in height of the lamella between the
first and second impact can be assumed to be negligible, given that the shape of the lamella is a puddle and
so h is approximately the capillary length. Then conservation of volume gives
4d3
2hd2max = 2 ,
3
and a maximum spread factor for the second drop of
r
4
2 = .
3h
The change in spread factors between the first and second drop impacts is then given by the difference
r r !r
4 2 1
= 2 1 = .
3 3 h
For the experiments that used water, h 0.5 mm and d 3.78 mm, giving 0.93.

Figure 13 shows a test of this prediction for the water data. A curve was fit to the maximum spread factor
data of the second impact, following the We1/4 prediction. Using the same slope, a curve was fit to the data
for of first drop. This ensures a constant difference between the curves, which was given as 0.900. This
agrees well with our predicted .
9
 0.7
10

y =1.0517 x1/4 + 1.2283

R = 0.8623
Spread Factor

0.6
10 y =1.0517 x1/4 + .32812
R = 0.97588

water, drop 1
water, drop 2
fit water drop 1
fit water drop 2
0.5
10
2
10
We

Figure 13. Determining the difference in spread factors between the first and second drops
for water. The second drop data was fit with a scaling of W e1/4 . Using this slope, a curve
was fit to the first drop data.

4.3. Splashing Behavior. Marmanis and Thoroddsen [3] predict the number of fingers to scale with W e3/16 Re3/8 .
It seems reasonable to assume this scaling relationship might also hold for the number of secondary droplets
seen during splashing since these result from disconnected fingers. As Figure 10 shows, this relationship does
not accurately predict the number of secondary droplets when using multiple liquids. We instead propose
that the number may scale with (W eRe)3/16 , as plotted in Figure 11. As shown in the figure, experimental
variations between days (i.e. temperature and humidity) and between trials (i.e. drop size, initial velocity,
etc.) cause a large discrepancy when trying to predict a trend for the number of secondary droplets. When
considering such variations, the splashing thresholds and higher values seem to agree fairly well between
the ethanol data and the water data from different days when using this scaling. Although the splashing
threshold cannot be exactly resolved, it appears to lie in (W eRe)3/16 [13, 13.5].

5. Conclusion

Our data supports a positive, linear correlation between the impact velocity of a drop and the spread
velocity of the lamella after impact, in the range of Weber Numbers between 100 and 600. This suggests
that the impact velocity of a liquid drop can describe behavior such as spreading, splashing, the formation
of fingers, and other such phenomena intuitively related to spread velocity. We have also described the
difference in maximum spread factors between the first and second drop impacts. Finally, we have shown
that the number of secondary droplets does not appear to scale as W e3/16 Re3/8 as one might expect since this
is the proposed scaling of the number of fingers by Marmanis and Thoroddsen. Based on our experimental
data, we have instead proposed the number of secondary droplets for successive drop impacts to scale with
(W eRe)3/16 .
10
 6. Future Directions

We would like to use a larger variety of fluids to gain insight into how the measured quantities vary with
fluid properties. We would also like to find a relationship which predicts the spread velocity based on the
impact velocity and other fluid properties. We would like to explore what could affect the scaling of the
maximum spread factor other than W e. Finally, a theoretical analysis of splashing phenomena might help
offer some insight into how the number of secondary droplets scales with respect to the Reynolds and Weber
numbers for two successive drop impacts.

References

[1] Brown, D. (2011). Tracker: Video analysis and modeling tool. GNU General Public License.
URL [Link] dbrown/tracker/
[2] Clanet, C., Beguin, C., Richard, D., & Quere, D. (2004). Maximal deformation of an impacting drop.
Journal of Fluid Mechanics, 517 , 199208.
[3] Marmanis, H., & Thoroddsen, S. (1996). Scaling of the fingering pattern of an impacting drop. Physics
of Fluids, 8 (6), 13441346.
[4] Roisman, I., Rioboo, R., & Tropea, C. (2002). Normal impact of a liquid drop on a dry surface: model
for spreading and receding. Proceedings of the Royal Society of London, 458 , 14111430.
[5] Worthington, A. M. (1908). A Study of Splashes. Longmans, Green, and Co.
[6] Yarin, A. (2006). Drop impact dynamics: Splashing, spreading, receding, bouncing... Annual Review of
Fluid Mechanics, 38 , 159192.

11
 7. Appendix

Source Code 1 process area.m

function process_radii(fname)
I = imread(fname);
I = edge(I,roberts);
se = strel(disk,8);
I = imdilate(I,se);

% find biggest connected component

cc = bwconncomp(I);
n = size([Link]);
n = n(2);
max = 0; max_el = 0;
for ii=1:n
s = size([Link]{ii});
if( s(1) > max )
max_el = ii;
max = s(1);
end
end

% remove other connected components

for ii=1:n
if(ii ~= max_el)
I([Link]{ii}) = 0;
end
end

% undilate it
I = imfill(I,holes);
I = imerode(I,se);
imshow(I);

% display area in pixels

sum(sum(I2))
end

12