CHARACTERIZATION OF A NEW TITANIUM DIOXIDE-POLYMER
COMPOSITE MATERIAL FOR ELECTRONIC PACKAGING APPLICATIONS
B. Dantal a, M.A. Zimmerman b and A. Saigal c
a
Graduate Student, [email protected]
b
Adjunct Associate Professor, [email protected]
c
Professor, [email protected]
Department of Mechanical Engineering
Tufts University, Medford, MA 02155
United States of America
ABSTRACT
Titanium dioxide (TiO2) is a versatile filler material used in polymer based applications. The most common application is in the
form of white pigment. However, titanium dioxide brings more to the polymer industry than just white, bright opacity. Titanium
dioxide is a photo-responsive material - its value is in its interaction with light. Injection molding processes are used to
manufacture liquid crystal polymer / Titania optical lids. TiO2-polymer composite materials are being developed for new
applications such as Liquid Crystal Displays (LCD), Light Emitting Diodes (LED) etc. The refractivity of the composite material
is a function of the dispersion of the titanium dioxide particles. Experimental methods to characterize dispersion of titanium
dioxide particles include Scanning Electron Microscopy (SEM). In this study, results from microscopy experiments, which yield
information about the particle distribution on the surface of the composite material are used in the image analysis process. The
image analysis approach has been developed using Matlab software. Results obtained with this method are helpful in
analyzing regions with single particles and regions with agglomerations. These two regions are important for determining the
percentage reflectivity of the material. Although there are many types of software available in the market to do similar tasks,
understanding of the features used to achieve the final goal is essential before the application of particular software. A Matlab
image analysis approach is developed based on the sieving technique. This paper focuses on detailed characterization of
Titania particles on the surface of the polymer. It is based on the application of the Scanning Electron Microscopy technique
together with the implementation of the image analysis procedure to compute dimensional and morphological parameters that
are useful for characterization and classification of Titania particles.
Introduction
Metal-oxide particles (or pigments), such as TiO2, serve many functions in polymeric materials. Traditionally, they have been
used as pigments to enhance the appearance and improve the durability of polymeric products, and usually they have been
considered to be inert. Titanium dioxide is the most important white pigment used in the plastics industry, offering a higher
refractive index than other white pigments and good chemical stability. Titanium dioxide is non-hazardous and has good
dispersability and thermal stability. Unlike colored pigments that provide opacity by absorbing visible light, titanium dioxide and
other white pigments provide opacity by scattering light. If there is a sufficient amount of pigment in the system, all light striking
the surface, except for the small amount absorbed by the polymer or pigment, will be scattered outward, and the system will
appear opaque and white. Light scattering is accomplished by refraction and diffraction of light as it passes through or near
pigment particles [1].
Experimental Details
The SEM has a large depth of field, which allows a large sample size to be in focus at a given time. The SEM also produces
images of high resolution, which means that closely spaced features can be examined at a high magnification. In order to view
non-conductive samples such as liquid crystal polymer composites, samples are covered with a thin layer of a conductive
material using a sputter coater. SEM –JEOL JSM-840, a 14 kV instrument, was used to obtain images. The samples were first
coated with gold to obtain the maximum efficiency during the SEM images acquisition procedure. These images were then
characterized with an image analysis software tool, developed using Matlab® programming language, for the dimensional and
morphological analysis. Samples are prepared by cutting optical lids into the squares of length 12.5 mm. Resolution along both
x- and y-axes is 4 dots per mm. Image width is 25 µm and length is 19 µm.
Image Analysis
The Matlab image analysis approach is developed based on the sieving technique. In the sieving technique, disks of known
pixel radius are passed through the image in increasing order and the intensity of the remaining field of the image is then
calculated. Minima and maxima of the intensity are calculated by taking the derivative of the intensity curve, which is helpful in
determining the particles of interest.
Figure 1 outlines some of the important steps in Matlab image processing, which include contrast enhancement, intensity plot,
background elimination, segmentation, connected regions, region properties and statistical data collection.
Raw Image
[SEM image]
Contrast Enhancement
[Contrast-limited adaptive histogram equalization (CLAHE)]
Intensity Plot
[Minima and Maxima calculations)
Background Elimination
[STREL-by creating structural element]
Segmentation
[Foreground and Background separation]
Connected Regions
[Labels the connected regions in binary Image]
Region properties
[Measures a set of properties for each labeled region]
Statistical Data collection
Figure 1. Major steps involved in image analysis
Contrast Enhancement
Contrast of the intensity image is enhanced by transforming the values using contrast-limited adaptive histogram equalization
(CLAHE) subroutine. CLAHE operates on small regions in the image, called tiles, rather than the entire image. The contrast of
each tile is enhanced, so that the histogram of the output region approximately matches the histogram specified by the
'Distribution' parameter. The neighboring tiles are then combined together using bilinear interpolation to eliminate artificially
induced boundaries [2].
Intensity Plot
Once the image contrast is adjusted, the intensity plot is then obtained by using the “sieving technique”. In this technique disks
of increasing diameter screens through the image and at each time the intensity of the remaining field is calculated. This plot
provides information about the maximum disk size that can pass through the image in order to get the remaining intensity of
the image equal to almost zero. In other words, if we filter the image with the disk size where the remaining intensity is zero
then there will be nothing left in the image. This data helps to get rid of background in following steps of image processing.
Taking the derivative of the intensity plot gives the minima, which means when we filter the image with that disk size a
maximum number of regions of interests are covered [3]. Figure 2 shows the plot of intensity vs. opening radius while Fig. 3
shows the maxima and minima obtained by taking derivative of the intensity plot. The minimum occurs at a radius of 6 pixels.
pixel value sum of opened objects (intensity)
2.5
x 10
7
Sum of pixel values in opened im age as a function of radius
2
1.5
1
0.5
0
0
5
10
15
20
25
30
35
radius of opening (pixels)
40
45
50
Figure 2. Sum of remaining intensity vs. radius of opening (Pixels)
Sum of pixel values in TiO2 as a function of radius
0
x 10
5
Granulom etry (S iz e Dis tribution) of TiO2
-2
-4
-6
-8
-10
-12
-14
-16
-18
0 2 4
6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46
radius of TiO2s (pix els )
Figure 3. Sum of remaining intensity vs. radius of opening pixel. After differentiating the intensity plot,
it shows the minima at 6 pixel radius.
Background Elimination
It is important in image analysis to eliminate the unwanted background noise in the image. ‘Strel’ command in Matlab creates
the morphological structuring element which is a flat, disk-shaped structuring element. These flat disks help to remove
unwanted background in image [3-5].
Figure 4 shows the SEM image of a TiO2-polymer composite while Figure 5 shows the image after Matlab image processing.
These figures show that the blurry part in the SEM image is successfully removed and areas with particles are highlighted.
This is beneficial as compared to the normal image in order to measure particle characteristics.
Segmentation
In the analysis of the objects in images it is essential that we can distinguish between the objects of interest and "the rest."
This latter group is also referred to as the background. The techniques that are used to find the objects of interest are usually
referred to as segmentation techniques, i.e. segmenting the foreground from the background.
Figure 4. Scanning Electron Micrograph of a TiO2-polymer composite
Figure 5. Micrograph after image processing using Matlab
Two of the most common techniques used are thresholding and edge finding. In this study we consider thresholding only. The
‘graythresh’ function in Matlab uses Otsu's method, which chooses the threshold to minimize the intra class variance of the
black and white pixels. The image is then converted to a binary image. The output binary image BW has values of 1 (white) for
all pixels in the input image with luminance greater than a given level and 0 (black) for all other pixels [6].
Connected Regions
The function “bwlabel” takes a binary image and calculates which groups of white pixels are connected to each other. The
output matrix is called a label matrix. It has the same size as BW and contains nonnegative integers. Each positive integer
value corresponds to a particular object [7].
To determine the number of particles in the image, the ‘bwlabel’ function is used. This function labels all of the connected
components in the binary image BW and returns the number of objects it finds in the image in the output value. The accuracy
of results depends on a number of factors, including:
• The size of the objects
• The accuracy of approximated background
• Whether you set the connectivity parameter to 4 or 8
Region Properties
This function in Matlab measures the properties of the image region. Statistical data about the regions in the image can be
gathered by using ‘stats’ command. Along with ‘stats’ command, the user can pass the parameters like area, equivalent
diameter, centroid etc. to obtain the required information about the region of interest in the image.
g
a
b
d
c
e
Figure 6. Essential steps in image processing
a) Raw Image b) Contrast Enhancement c) Background Removal d) Gray Threshold e) Connected Regions
Particle Size
Describing a 3D particle is often a more complex matter than it first appears. For simplicity it is convenient to describe particle
size as one single number. However, unless the particle is a perfect sphere (which is rare in ‘real-world’ samples) there are
many ways to describe the size of a particle. One of the basic challenges of particle size analysis is to describe a 3dimensional object with one number only.
An image analysis system captures a 2-dimensional image of the 3D particle and calculates various particle sizes and particle
shape parameters from this 2D image. One of the principle diameters calculated is CE (Circle Equivalent) diameter which is
the diameter of a circle with the same area as the 2D image of the particle. Of course different shaped particles will have an
influence on this CE diameter but, importantly, it is a single number that gets larger or smaller as the particle does and it is
objective and repeatable [8].
Figure 7. Calculation of Circle Equivalent (CE) diameter [8]
Circle equivalent diameter is calculated using Eq. (1):
CE (Circle _ Equivalent _ Daimeter ) =
4 * Area
(1)
π
Results and Discussion
Table 1 compares the data obtained from the raw image without any image processing and after image processing. When the
image is not processed, there are several factors that play an important role in the final analysis. One of the important factors
is calculating noise in the image because noise (i.e. blur in the image) can be detected as a separate particle. It is very
important to define a threshold in the image between particles and background otherwise additional area is considered as part
of the particle where in reality it may be noise surrounding the particle. Table 1 shows that the number of particles and
maximum diameter acquired before image processing are much higher as compared to the values obtained after image
processing. This explains the importance of the noise detection and defining threshold in the image.
Table 1. Particle distribution before and after image processing
Before Image Processing
After Image Processing
Number of Particles
412
209
Area Occupied by Particles
24.51 %
15.50 %
Maximum Diameter (µm)
11
2.7
Agglomeration
Can’t determine
4.21 %
Noise in the Image
Can’t determine
0.17 %
Figure 8 shows the particle size distribution when the image is analyzed without image processing. The numbers of particles
obtained are 412 and the area fraction is approximately 24 percent. Material manufacturing data indicates that particles in the
range of 0.1 to 1.2 µm are blended in the composite matrix. Maximum particle diameter detected is around 11 µm.
Size Distribution
200
180
Number of Particles= 412
Area occupied by Particles= 24.51 %
160
140
Number of Particles
120
100
80
60
40
20
0
1
2
3
4
5
6
7
8
9
10
11
12
Particle Diameter (µm)
Figure 8. Particle size distribution obtained before Matlab image processing
Figure 9 shows the particle size distribution of the image analyzed after the image processing technique. The numbers of
particles obtained are 209 and the area fraction is approximately 15 percent. Maximum particle diameter estimated is 2.7 µm.
Size Distribution
30
Number of Particles
25
Number of Particles= 209
Area occupied by Particles= 15.50 %
20
15
10
5
0
0.5
1
1.5
2
2.5
3
Particle Diameter (µm)
Figure 9. Particle size distribution obtained after Matlab image processing
Figure 10 shows the pie chart for the area percentage occupied by different parameters in the image after image processing.
Image processing techniques help to determine the noise in the image, which reduces the blur and helps to calculate the
actual number of particles. Regions with single particles and regions with agglomerations play different roles in reflectivity of
the composite material. For the most effective light scattering, the pigment diameter should be slightly less than one-half the
wavelength of light to be scattered. Light scattering imparted by diffraction is affected by particle spacing and average pigment
particle size. If the particles are too large or too closely spaced (agglomerated), little diffraction takes place which reduces the
reflectivity of the medium [1]. As such, regions with single particles refract light more effectively as compared to regions with
agglomerated particles because agglomerated particles have diameters larger than the wavelength of incident light.
Noise in the Image
0.17%
Single Particles
11.12%
Agglomerate
4.21%
Approximately 15% of the
area is occupied by
Titanium Dioxide particles
Liquid Crystal
Polymer
84.50%
Figure 10. Area occupied by different parameters in Liquid Crystal Polymer-Titania composite
Conclusions
The proposed approach shows a new method for the characterization of titanium dioxide particles both for dimensional and
morphological parameter estimation. The possibility to characterize the titanium dioxide particles on the surface of the injection
molded part using a SEM image coupled with the application of the developed image analysis tool provides important
additional information. Results obtained with this method are helpful in analyzing regions with single particle and regions with
agglomeration. These two regions are important for determining the percentage reflectivity of the material.
References
1.
2.
3.
4.
5.
6.
7.
8.
Polymer Light and Science of TiO2 Particle, DuPont Literature, (2006).
Matlab Image Processing Tool Box User Guide, Version 7, (2006).
Adams, R., "Radial Decomposition of Discs and Spheres," Computer Vision, Graphics, and Image Processing: Graphical
Models and Image Processing, 55(5), 325-332, September 1993.
van den Boomgard, R. and van Balen, R., "Methods for Fast Morphological Image Transforms Using Bitmapped Images,"
Computer Vision, Graphics, and Image Processing: Graphical Models and Image Processing, 54(3), 252-254, May 1992.
Jones, R. and Soille, P., "Periodic lines: Definition, Cascades, and Application to Granulometrie," Pattern Recognition
Letters, 17, 1057-1063 (1996).
Otsu, N., "A Threshold Selection Method from Gray-Level Histograms," IEEE Transactions on Systems, Man, and
Cybernetics, 9(1), 62-66 (1979).
Haralick, R. M. and Shapiro, L.G., Computer and Robot Vision, Volume I, Addison-Wesley, 28-48 (1992).
“Particle Size and Shape Measurement Using image Analysis”, Malvern Instrument Application Notes, MRK664-01,
(2006).