You are on page 1of 11

Col!

oids and Surfaces, 58 (1991) 239-249


Elsevier Science Publishers B.V., Amsterdam

239

The influence of droplet volume and contact an& on


liquid surface tension measurements by axisymmetric
drop shape analysis-profile (ADSA-P)
J. Noordmans and H.J. Busscher
Laboratory for Mate%
Technica,
Groningen, The Netherlands

ilniuersity

of Groningen.

Antonius

Deusinglaan

1,9713

AV

(Received 10 October 1990; accepted 11 April 1991)

Abstract
The axisymmetric drop shape analysis-profile (ADSA-P) technique was evaluated with respect
to the influence of contact angle and droplet volume. The system was impleslenked on our standard
contact angle measuring apparatus employing a Vidicon video camera and a 512X 256 pixels
framegrabber. For calibration of the x and y magnification factors a perfectly sphe.kal ball was
used of approximately the same dimensions aE;the liquid droplet under study. Droplets of water,
formamide and hexadecane with volumes varying from 2-200 fl were put on FEP-Teflon and
PMMA. Erroneous surface tensions were found by ADSA-P when too small or too large droplet
volumes were employed. Ideal volumes to be used were dependenton the contact angle and amounted
inoursystemto5-10~lfor8~40,20-50~lfor40c8c90and50-100~for8~90.Statistical
evaluation of results obtained for the same droplet (i.e. water on Pamfilm), digitized 16 times
using two different sets of five calibration frames, indicated that digitization of the liquid profile
is the largest source of variation. Typical standard deviations over 16 images of the same droplet
are 50.9 MJ mm2 in surface tension and k 1.5 in contact angle. However, in summary it is
concluded that ADSA-P is a powerful tool for simultaneou determination of liquid surface tension
and contact angle, even when implemented using a relatively simple image analyser.

INTRODUCTION

The shape of a liquid droplet on a solid surface is de+%rminedby a complicated interplay of the liquid surface tension, volume, density and contact angle.
At equilibrium, an axisymmetric droplet obeys the Eaplace equation
I
AP=Y($+&)

(1)

in which y is the liquid surface tension, Ap is the pressure difference across the
interface and R, and R2 represent the two principal radii of curvature of the
droplet [ I].

01666622/91/$03.50

0 1991 EIsevier Science Publishers D.V. All rights reserved.

240
The surface tension of mercury, for instance, is so high (about 480 mJ rns2 i
that a mercury droplet practically always adopts a perfectly spherical shape.
In contrast, a water droplet: provided it has a sufficiently large volume, tends
to show flattening at the apex as a result of its much lower surface tension
(72-73rnJ mv2).
Several methods have been proposed to use the deviation from sphericity of
a liquid. droplet for determination of the liquid surface tension [2-41.Often,
only the height and width of a droplet are used [2,3
] and the methods cannot
be applied when the contact angle is < 90 O [3 1.
In the past Neumann and co-workers 14-6 ] have developed a method called
axisymmetric drop shape analysis-profile (ADSA-P) to derive the surface tension and contact angle of liquid droplets by fitting the Laplace equation to the
whole circumference of a droplet. In a mathematical description of the method
[a] it was stated that there are no restrictions on the applicability of the
method.
It is the aim of this paper to determine:
(i) the influence of the droplet volume on ADSA-P measurements,
(ii) the influence of the contact angle on the precision of surface tension
measurements by ADSA-P.
To this end, the original ADSA-P program, written in FORTRAN, was purchased from Neumann and co-workers [4-61 rewritten in TURBO PASCAL and
used in conjunction with our standard contact angle measuring apparatus.
Special attention will be given to the overall accuracy that can be achieved in
our system as a guide for interested future users of ADSA-P.
MATERIALSANDMETHODS
Experimental set-up and calibration of the image system
Droplet profiles are digitized using a Vidicon video camera (SONY AVC3250 CES) equipped with a 120 mm lens (Medical Nikkor, Nikon) connected
to a framegrabber (SUPCON EC90, Electronisch Ontwerpburo de Boer, The
Netherlands) providing 512x 256 pixels and 256 grey levels.
The video interface determines the coordinates of itn object by running a 12
bit counter at the beginning of each frameline scanned. AfiGerdetecting an
adjustable threshold, the value of the counter is stored in a buffer. The same
procedure holds for the second threshold on the frameline. Then, the two buffer
values are copied to a memory accessible by the framegrabbers microprocessor. In the meantime, the next counter values are copied to the buffer. The
interface between the framegrabber and the computer running ADSA-P - an
Olivetti M380 20 MHz with mathematical coprocessor -- is made by means of
an ordinary parallel I/O slot card (PI0 12,Metra Byte Corp., Taunton, MA,
USA). Every image thus consists of 256 framelines, each divided into 512 points.

241

By means of this interface, the framegrabbers microprocessor can be requested to give a framedump, i.e. the mean of four successively added frames.
The time for transmitting such a frame to the host computer and for the framegrabber to get ready for the next dump is less than 0.5 s.
Calibration of the x and y magnification factors is done by grabbing an image
of a perfectly spherical, stainless steel ball of approximately the same size as
the droplet to be analysed. After digitizing the dropleA Emage,the stainless steel
ball is positioned at exactly the same spot as the droplet a& its profile digitized. Since the aspect ratio of framegrabber and the Vidicon video cameradoes
not equal unity, the image of the spherical profile of the steel ball is essentiallgelliptical. Horizontal and vertical magnification factors are found by fitting an
ellipse to the data points:

in which x and y are the spatial coordinates, and a the long and b the short
half-axis lengths of the ellipse.
To perform the fit, an error filnction which expresses the sum of the normal
distances between a calculated ellipse and the data points is minimQed by
adjusting the ellipse centre coordinates and the o and b parameters of the ellipse.
The distance of a data point with coordinates (p,q)
to a calculated ellipse
can be written as:

The minima of this function, representing the normal distances, are given by:
(4)
Equation (4) is a quartic equation that can be solved with the Ferrari method
[i] yielding the ellipse centre coordinates (x3) nearest to (p,q).
Minimization with regard to a, b and the ellipse centre coordinates is then
performed using a conjugate gradient method offered by the FRPRMIN function from the MATHPAK-87
package (Precision Plus Software, Toronto, &nada). Subsequently, the ultimate x and y magnification factors, in pm per pixel,
are found by dividing the original stainless steel ball radius by a and b, respectively. The above-described calibration is thus highly objective and observer
independent. Calculation of the x and y magnification factors requires about 5
min CPU time.

242

Axisymmetric drop shape analysis-profile (ALSA-P)


The original FORTRAN
ADSA-P program, purchased from Neumann et al.
[4], was rewritten in TURBO
PASCAL
using gravity, density, the three-phase
li,le, a set of droplet profile coordinates and the x and y magnification factors
as input parameters. The three-phase line, denoting the exact location of the
solid surface, is expressed as a frameline number, determined manually by
moving a line over the droplet image.
The ADSA-P strategy to find the liquid surface tension and contact angle
has been described in detail before [4-61; a short r&urn6 follows for clarity,
however. An objective finction is constructed which expresses the difference
between the digitized c.q. measured droplet profile and a theoretical profile,
calculated on the basis of the Laplace equation, Eqn (1. ). This objective function contains the four adjustable parameters X0, Z,, RD and /? [S] , in which

AmRit
p=

I4

with X0 and 2, the droplet apex coordinates, ROthe radius of curvature in the
apex, dp the density difference across the interface and g the acceleration due
to gravity; p is generally called the shape factor.
The objective function is minimized numerically using the method of incremental loading in conjunction with the Newton-Raphson method. The initial
guess for the fit is a circle segment, i.e. the case of infinitely high sur%ce tension. After that the fit shows the desired degree of convergence by means of
the criteria given in Ref. 141, ADSA-P yields the following output: liquid surface tension, contact angle, droplet height, radius in the apex, volume, contact
area and contact radius. Analysis of one droplet profile, containing 250-300
data points, requires about 15 min CPU time.
Experiments and statistical evaluation
In order to determine the influence of droplet volume and contact angle on
ADSA-P measurements, droplets with volumes varying from 2 to 200 ,cillof
water, formamide and hexadecane were put on poly(methy1 methacryisis)
(PMMA) and FEP-Teflon.
Surface tensions of water (Millipore Q) , formamide (spectrophotometric
grade, Aldrich, Milwaukee, WI, USA) and hexadecane (analytical grade, Merck,
Germany) were determined before ADSA-P measurements by means of a tensiometer (Lauda Autotensiomat, Mijnigshafen, Germany) using a ring at 24 C.
Results were found to coincide with literature values, being 27.5,57.4 and 72.8
mJm - for hexade cane, formamide and water, respectively [ 0,lO ] . Only freshly
tapped Millipore water was employed.
Poly (methyl methacrylate ) (Casterpex, ICI, England) and FEP-Teflon

243

(Fluorplast, Raamsdonkveer, The Netherlands) surfaces were extensively


cleaned by rinsing alternately (twice ) with methanol and distilled water.
Cleaned samples were dried in an incubator at 25 OC before use.
The influence of the digitization and calibration was determined by analysing a 100 ,ul water droplet on ParafilmW, a paraffin-coated polyethylene foil
(American Can Company, Greenwich, CT, USA). The same droplet was
grabbed and digitized 16 times with one minute intervals. Subsequently, these
profiles were analysed with ADSA-P using two sets of sequentially grabbed

1: water

on

FEP-Teflon

.
I

Droplet
ylv

.I

100

50

EmJ-m-23

volume

formamide

2:

200

150

WlI

Qn

FEP-Teflon

- 9
=s

0'

OU

0.

8.

1.

Droplet
ylv

fmJ=m-27

3:

volume

hexadecane

on

200

150

100

50

@.113
FEP-Teflon
I * -

100

0.0

5.0

10.0
Droplet
volume

Fig. 1.Liquid surface tensions yl,.calculated hy ADSA-P


formamide and hexadecane droplets on FEP-Teflon.

15.0
ICI11

20.0

OSa function of droplet volume for water,

244

calibration frames. Each set consisted of five images of a 5000 2 0.5 pm diameter stainless steel ball, grabbed with one minute intervals without repositioning of the calibration ball. The second set of calibration frames was collected
after removing and repositioning the ball every time an image was digitized.
All experiments were done in an enclosed chamber at room temperature
(22-24C).
For statistical evaluation we employed ANOVA with a two-factor model
[ II]. Digitization of the liquid droplet profile was taken as factor A, whereas
calibration was taken as factor B. Subsequently the model estimates var*A
and varAB, expressing the extent to which digitization of the liquid droplet
profile or calibration represents the main source of variance in repeated measureinents. Furthermore, ANOVA yields so-called F values, as a measure of the
variance within repeated digitization (FA) or repeated calibration (Fn ) . Taking a 95% confidence limit, FA and FB are compared with Fomo5,
indicating
whether results of repeated digitization or calibration can be considered
identical.
RESULTS

Figure 1 shows liquid surface tensions yl calculated from droplet profiles on


FEP-Teflon
for water (Fig. 1.1)) formamide (Fig. 1.2) and hexadecane (Fig.
1.3 ) as a function of droplet volume.
%v

ImJ

m-21

1:

water

cn

PMMA

100

50

100

50

Droplet
%V

CmJ -m-21

2:

volume

formamide

on

150

&III

200

PMMA

100

50

OIL
0

, . . . .
50

100
Droplet

volume

Fig. 2. Liquid surface tensions ylycalcuIated by ADSA-P


and formamide on PMMA.

150
Q.111

200

as a function of dropiet vohme for water

-111.

73.4
71.5
7?,4
12.8
72.5
72.8
72.8
73.3
73.7
72.2
72.5
70.9
72.9
72.7
Itlo.

74.0
72.1
73.0
72.5
73.2
73.5
73.4
73.9
74.3
72.8
73.0
71.5
73.5
73.3
9 0.9

73.0
71.2
72.1
72.6

72.3
72.5
72.4
72.9
73.3
71.9
72.1

70.6
12+6

72.4
kO.9

4
5
6
7

8
9
10
11
12
13
i4

15
16

Mean
SD

71.7
73.8
73.5
k 0.9

73.4
73.6
73.6
74.1
74.5
73.0
73.3

74,2
72.3
73.2
73.7

75.0
n
75.

Imqe imperfections prevented convergence of ADSA-P.


F,,=26.5(2.04),varAA=0.82,
Fe=?8.75(2.05) and varB=0.24.
FA=79.8(2.16),varAA=:1.13.F~=48.1(2.64)andvarhB=0.33.

74.2
n
II

3
74.8
U
75.1

73.8
R
73.9

Calibration without repositioning: set i

1
2
3

Droplet
image

70.9
69.5
II
73.1
71.6
a

71.5
k 1.1

73.2
71.8
72.5
71.9
70.9
71.2
72.2
70.8
71.9
7G.9
73.2
71.7
H

i
75.6
74.1
75.6
74.1
73.1
73.3
71.4
73.1!
74.2
73.1
75.4
73.9
II

73.5
f1.3

70.2
a

74.0
72.6
74.1
72.6
71.7
71.9
71.0
71.6
72.7
71.7
73.9
72.4
71.4
a

72.3
k1.1

71.0
fl.1

72.7
71.3
70.4
70.6
70.9
70.3
7l,4
70.4
72.6
7i.l
70.1
70.4
69.0
m

72.8
71.5
70.5
70.8
70.9
70*4
71.6
70*5
72.8
71.3
II
d
69.1
II

72.6
a

4
72.8
II

71.1
t-1.1

Calibration with repositioning: set 2

71.4
69.7
70.5
II

72.2
I
a

70.5
70.9
70.9
71.3
71.7
70.3
70.5
69.1
71.0
70.8
k 0.9

Surface tensions (mJ m-) calculated with ALGA-P from images of a water droplet ilO0 ~1) on Parafilm@, sequentially grabbed and analysed by
two sets of sequentially grabbed calibration frames. VarA and var*B, calculated by ANOVA, indicate up to what extent digitization or calibration
represent the main source of variance. FA and Fn, also resu!ting from ANOVA, indicate whether the means within repeated digitization of droplet
profiles or calibration can be considered identical. The values in parentheses represent the F 0,05values obtained from Ref. [ 111

TABLE 1

98.8
100.5
100.5
99.6
99.7
101.3
-1-1.5

100.4

101.9
100.7
100.4
98.8
100.5
100.5
99.5
99.7
101.3
+,1.5
101.9
100.7
100.4
98.8
100.5
100.5
99.6
99.7
101.3
21.5

102.6
102.8
102.3
II
B

102.8
n

100.5
k1.9
100.6
k1.9

100.5
k1.9

100.5
i- 1.9

100.4
k1.9

96.9
a

96.9
a

101.1
101.6
100.8
101.3
97.6
95.4
LL

102.9
a
n
101.3
I

101.9
lOC.7
xKI.4
98.8
100.6
100.5
99.6
99.7
101.4
k 1.5

99.8

102.7
102.9
102.4
a

99.7
101.3
101.2
102.3
102.2
101.2
101.6
100.9
101.3
97.7
95.4
100.5
96.9
II

101.3
0

99.7
101.3
101.2
102.3
IO2,2
101.2
101.6
1.01,9
101.3
97.7
A
a

101.3
102.7
99.7
101.3
101.2
102.2
102.2

100.5
96.9
a

101.3
101.2
102.3
102.2
101.2
101.6
100.9
101.3
97.7
0

101.3
102.8
a

Calibration with repositioning: set 2

100.5
96.9
II

101.1
102.2
101.1
101.1
101.6
IO&8
101.2
97.6
II

99.6
101.2

101.3
0

102.8
102.7

102.4
102.6
102.8
102.3
99.8
99.7
101.9
100.7
100.4
98.8
100.5
lCO.5
99.5
99.7
101.3
I!I1.5

pImageimperfections prevented convergence of ADSA-P.


bF,=47.63(2.16)+varAA=2.05,Fn=5.77(2.64) andvar*B=O,ll.
cF~=39.05(2.16),varAA=2.66,Fr,=2.31(2.64) andvarAB=0.05.

13
14
15
16
Mean
SD

12

100.7

101.9

99.7

102.3
a

102.4
102.6
102.8
102.3
99.8
99.7

104.5
102.6
102.8

3
4
5
6
7
8
9
10
I1

102.8
II

102.9
II

Calibration without repositioning: set lb

Droplet
image

Contact angles (deg.) calculated with ADSA-P from images of a water droplet (100 ~1) on Parafilm@, sequentially grabbed and analysed by two
sets of sequentially grab&d calibration frames. VarA and varB, calculated by ANOVA,indicate up to what extent digitization or calibration
represent the main source of variance. FAand Fo, also resulting from ANOVA,indicate whether the means within repeated digitization of droplet
profiles or calibration can be considered identical. The values in parentheses represent the F,., values obtained from Ref. [ 111

TABLE 2

247

Figure 2 shows surface tensions ylvcalculated for water (Fig. 2.1) and formamide (Fig. 2.2 ) droplets on PMMA. Hexadecane droplets on PMMA could
not be analysed owing to almost complete spreading.
On all solids a distinct influence of the droplet volume is evident, showing
unrealistically high and scattered surface tension values for small droplet volumes and slightly decreasing surface tension values for larger droplets for water
on FEP-Teflon and PMMA. In all cases, a volume range exists where the calculated surface tensions are more or less constant.
Contrary to the calculated surface tensions, contact angles were independent of the liquid volumes employed. Mean values + SD on FEP-Teflon were
iGG.6 2 8.3, 90.0 -+ 4.4 and 39.0 4 2.6 * for water, formamide and hexadecane,
respectively. Note that the water contact angle on FEP-Teflon
is rather low,
which we attribute to the use of a commercial grade sample. Water and formamide contact angles on PMMA were 74.5 -+ 2.9 and 50.4 4 1.8) respectively.
Tables 1 and 2 show the swrface tensions and contact angles calculated by
ADSA-P for different images of the same water droplet on Parafilm@ using
two sets of calibration frames. Typicai standard deviations over 16 images of
t.he same droplet were 2 0.9 mJ m -* in surface tension and 4 1.5 in contact
angle. Foaos
values for repeated digitization and calibration are an order of magnitude higher than FA and FB calculated by ANOVA for the surface tensions
in Table 1, indicating significant differences within the various sets of data.
Furthermore, since in both calibration sets varA is much larger than varB,
it appears that digitization of the liquid droplet profile is the main source of
variance in liquid surface tension determination by ADSA-P.
Similarly, by comparing F, and FB for the contact angle data in Table 2 with
values it is reasonable to state that contact angles derived
the appropriate Foho5
within different calibration sets are identical, but not so those within repeated
digitizations of liquid droplet profiles. Also, for the contact angles, digitization
of the droplet profile appears to be the main source of variance., i.e.
varA>varB.
DISCUSSION

In this paper we have evaluated the influence of droplet volume and contact
angle on liquid surface tension measurements by ADSA-P. We shall now focus
firstly on factors intrinsically connected to the use of ADSA-P and secondly
on influences and limitations associated with our particular contact angle
measuring appairaks.

InJtuencesconnectedwithA13SA-P
The statement made by Neumann et al. [4 ] and quoted in the introduction
of this paper that ADSA-P had no restriction with regard to its applicability

initially triggered us to evaluate the method critically. This study indeed shows
that, contrary to most other methods [33, ADSA-P works well for contact
angles smaller than 90. However, an experimental restriction, at least in our
system, was the size of the droplet volume used. When the volumes used are
too small, the deviation from sphericity of a droplet is not sufficient and the
surface tensions that result are in general too high. For volumes which are too
large, the droplet will become pancake shaped and ROgoes to infinity, often
yielding a floating point overflow or unrealistic results. Furthermore, droplets with a larger volume have a tendency to be less axisymmetrical than small
droplets.
The effect of volume on ADSA-P measurements is dependent on the contact
angle (see also Figs 1 and 2 ) . Hence we recommend the following ideal volume
ranges for obtaining the most accurate surface tension measurements: 5-10
~1for8c40;20-50~U1for40~~~90and50-100~1for8~90.
Usin? these ideal volume ranges, our data in Figs 1 and 2 yield surface
tensions of 70.9 -t-3.6, 56.4 !I 5.0 and 27.7 + 3.7 mJ mm2 for water, formamide
and hexadecane, respectively. These results are in reasonable agreement with
literature values, although the above standard deviations are larger than those
obtained by tensiometry. However, we note that the relative precision, when
considering only surface tension changes of a given droplet in a given calibration frame, is far better (about 1 mJ mm2).
The standard deviation found with repeated contact angle measurements
(2-4 O) is comparable with those reported for goniometry-based or other techniques, although here again the relative precision in detecting contact angle
changes for one droplet is far better ( 1.5 O).
Influences associated with the particular apparatus used
An additional reason why ADSA-P in our experimental system cannot handle extremely flattened droplets is associated with the principle of our image
analyser. Since a maximum of two points is generated per frameline, the flattened section of the circumference of a droplet is described by only two datapoints. Furthermore, t.he large width of a pancake shaped droplet makes it
necessary to reduce the total magnification of the system, thereby also causing
a reduction in the number of framelines resolved over the droplet height.
The relatively poor quality of the image analyser also becomes obvious from
the fact that the statistical evaluation clearly demonstrated that the major
source of variation was associated with the digitization of the liquid droplet
prcfile and not with, for instance, the calibration.

ADSA-P is a powerful method for the simultaneous determination of liquid


surface tensions and contact angles. Implementation is fairly easy, and even

249

with a relatively simple experimental set-up a reasonable precision can be


obtained.
ADSA-P works equally well for contact angles > 90 and for contact angles
c 90 O.An experimental restriction is posed, however, by the volume to be used;
ideal vo1um.e ranges are dependent on the contact angle in the system under
study.
Xoat
probat;
, the final precision of ADSA-P will benefit. greatly from improvements in both hardware (use of a CCD camera and high-resolution
framegrabber) and software (inclusion of image distortion correction and sophisticated profile edge detection; see also ref. [6 3 ). Adapting our experimental set-up with these modifications is expected to yield a twofold to threefold
improvement in fmal precision.
ACKNOWLEDGEMENTS

The authors are greatly indebted to Mrs M. Schakenraad-Dolfing for the


preparation of this manuscript and to Dr R. Stewart for his kind help with the
statistical evaluation of the data.

REFERENCES
1
2
3
4
5
6
7
8
9

10
11

M.A. Fortes, Phys. Chem. Liq., 9 (1980) 285.


H.M. Princen, in E. Matijevie (Ed. ), Surface and Colloid Science, Vol. 2, Wiley Interscience,
New York, 1969,p. 1084.
J-D. Malcolm and H.M. Paynter, J. Colloid Interface Sci., 83 (1981) 269.
Y. Rotenberg, L. Boruvka and A.W. Neumann, J. Colloid Interface Sci., 93 (1983) 169.
J.F. Boyce, S. Schiirch, Y. Rote&erg and A.W. Neumann, Colloids Surfaces, 9 (1984) 307.
P. Cheng, D. Li, L. Boruvka, Y. Rotenberg aid A.W. Neumann, CoIloids Surfaces, 43 (1990)
151.
L.E. Dickson, New First Course on the Theory of Equations, John Wiley & Sons, New York,
1952, pp. 42-55.
F. Bashforth and J.C. Adams, An Attempt to Test the Theories of Capillary Action, Cambridge University Press and Deighton Bell & Co., Cambridge, 1892.
J-J. Jasper, J. Phys. Chem. Ref. Data, 1 (1972) 841.
C.J. van Oss, L. Ju, M.K. Chaudhury and R.J. Good, J. Colloid Interface Sci., 128 (1989)
313.
J. Neter and W. Wasserman, Applied Linear Statistical ModeIs, Richard D. Xrwin, Inc.,
Homewood, IL, 1974, pp. 569-577.

You might also like