Relating gloss loss to topographical features of a PVDF coating.

Semi-gloss commercial poly(vinylidene fluoride) (PVDF) coatings typically have 60[degrees] gloss values between 20 and 50. Gloss is affected by PVDF crystallite structures and by the pigmentation. In this article, we have demonstrated that for some pigmented PVDF coatings, after 10 years of Florida exposure, the principal proximal cause of gloss changes is the formation of micron-scale pits, rather than the emergence of pigment particles at the coating surface. We have used laser scanning confocal microscopy (LSCM) and light scattering to characterize the surface topography and near-surface structure of weathered and unweathered PVDF coatings. Florida-weathered PVDF coatings show only a modest increase in the root mean square (RMS) roughness of the surface, even when noticeable gloss loss has occurred. Changes in gloss can be correlated with surface roughness and other topographical features, including the formation of pits and the emergence of pigments.

Keywords: Surface analysis, photodegradation, fluorinated polymers, pigments, appearance, durability, service life prediction, architectural, laser scanning confocal microscopy, light scattering, gloss measurement.

**********

Gloss and color retention are common performance attributes used in the coating industry to measure the weatherability of paints and coatings, particularly decorative coatings used in architectural, automotive, and other applications. As measures of paint performance, gloss and color can be easily and quantitatively assessed. However, the gloss and color measurement scales cannot, in general, be directly correlated with more fundamental microphysical surface properties of a material. Thus, when a coating begins to degrade in an outdoor environment, it is not possible to link these optical changes to the chemical and physical molecular or microscopic precursors.

We use laser scanning confocal microscopy (LSCM) and other techniques to probe naturally and artificially weathered PVDF architectural paints. Typical PVDF paint binder compositions consist of 70%-80% (by mass) PVDF resin, and 20%-30% of a miscible acrylic co-resin (e.g., a methyl methacrylate-ethyl acrylate copolymer (1)). For pigmentation, the paints generally use inorganic metal oxides. Coating grade PVDF resin is semicrystalline with a melting temperature, [T.sub.m], of about 165[degrees]C. As the coating is baked above [T.sub.m], the resin components form a miscible alloy and, upon cooling, form a film comprised of a complex mixture of crystalline and amorphous regions. The applied paints typically are semigloss in aspect (60[degrees]C gloss values between about 20 and 50), with surface features influenced both by PVDF crystalline structures and by pigmentation. Thus, there are a number of differences between these paints and automotive coating systems that have been examined in many recent service life prediction studies (2): PVDF paints are thermoplastic (but semicrystalline); they are pigmented; and they start with a somewhat rough, semigloss surface.

The motivation for this study is to understand how best to track early changes in a thermoplastic paint as it weathers, which ultimately lead to service life failure through gloss loss. Unlike thermoset clear coats, where it is easy to use vibrational spectroscopy to track photochemical changes leading to catastrophic network failure, observable chemical changes in thermoplastic coatings can be more modest, because of mechanisms like surface erosion which remove degraded material from the surface of the specimen. (3) By using microscopic techniques like LSCM and atomic force microscopy (AFM), however, the physical changes in the coating can be probed directly. As a first step in this quest to be able to predict weatherability from early tests, in this article we compare unweathered coatings with moderately weathered coatings--coatings which are just at the half way point for gloss retention, after about 10-12 years of Florida exposure.

[FIGURE 1 OMITTED]

EXPERIMENTAL (a)

Materials

Paint specimens used in this study were model formulations based on commercial PVDF paint formulations, and contained PVDF and acrylic resin blended in a 70:30 mass ratio. A typical formulation contained 24 g coatings grade PVDF (Akrema Inc.), 25 g Paraloid[R] B-44 acrylic (40% solution in toluene) (Rohm and Haas Company), 16 g weather-able metal oxide pigment, and 35 g isophorone. Pigments used include Shepherd Blue 190 (C.I. Pigment Blue 36, Shepherd Color Company), other mixed metal oxide pigments from the Shepherd Color Company, and TiPure[R] R-960 rutile titanium dioxide (DuPont Corp.). Paints were applied by wire-wound rod to chromated aluminum test panels, and were baked in a high temperature oven for 45 sec to a peak metal temperature of 254[degrees]C, then water quenched. Average dry film thickness was 18-25 [micro]m.

LSCM and gloss measurements were conducted on 14 Florida exposed coatings (exposed and masked areas) and 25 unexposed coatings (control). The results of these measurements (six gloss measurements and six LSCM measurements per location--exposed, masked, and control area) will be discussed later in the text. However, for the purpose of developing a quantitative method to analyze the microscopic weathering of coatings, three coated panels, weathered 10 years in Florida, were selected for more detailed analysis. These panels have been designated as follows:

Panel A: Pigment Shepherd Blue 190, exposed 10 years in Florida.

Panel B (b): Pigment Shepherd Blue 190, exposed 10 years in Florida.

Panel C: Pigment Shepherd Blue 211, exposed 10 years in Florida.

Weathering

Painted test panels were exposed in Miami, South Florida at 45[degrees] south exposure, on open-backed racks. On approximately an annual basis, half of the panel was lightly washed with a sponge, then both the washed and unwashed portions were rated for gloss and color. For this study, microscopy work was performed on the unwashed portions of the panels. In some cases unexposed control specimens, stored in the laboratory from the same batch as that of the exposed specimens, were available for reference.

Laser Scanning Confocal Microscopy

A Zeiss model LSM510 reflection laser scanning confocal microscope (LSCM) was employed to characterize the surface morphology (topographic profile) and to measure the surface roughness of the coatings. LSCM utilizes coherent light and collects light exclusively from the focal plane, while rejecting light out of the focal plane. By using different microscope objectives, the scanned surface area can be varied from 1.84 mm x 1.84 mm down to 61.4 [micro]m x 61.4 [micro]m. The incident laser wavelength was 543 nm. The wavelength, numerical aperture (N.A.) of the objective, and the size of the pinhole dictate the resolution in the thickness or axial (z) direction. (4) By moving the focal plane, single images (optical slices) can be combined to build up a three-dimensional stack of images that can be digitally processed. z-steps were selected to obtain overlapping optical slices (a stack of z-scan images). The z-step size was 0.5 [micro]m using objectives of 5x and 10x, and 0.1 [micro]m using objectives of 20x, 50x, and 150x. LSCM images presented in this article are 2D intensity projections (an image formed by summing the stack of images over the z direction, 512 pixel x 512 pixel), or 3D topographic profiles of the coating surface. The 2D intensity projection images are effectively the sum of all the light scattered by different layers of the coating, as far into the coating as light is able to penetrate. The pixel intensity level represents the total amount of back-scattered light. Darker areas represent regions scattering less light than lighter colored areas.

From the 3D topographic profiles, the root mean square (RMS) surface roughness, [S.sub.q], is calculated using a surface tilt correlation using an automatic plane fit. Plane fit is commonly used to remove tilt from images, a single polynomial fit is calculated for the entire image and then subtracted from the image. The RMS surface roughness is calculated without a numerical filter according to the following formulas:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (1)

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (2)

Here, [N.sub.x], [N.sub.y] ... number of pixels in X- or Y-direction.

Gloss Measurement

Gloss measurements were made using a handheld commercial glossmeter (Minolta, Multi-Gloss model 268). Measurements conform to the ASTM D 523 standard measurement protocol. For architectural coatings, the 60[degrees] gloss measurement is the one that is most often cited in architectural specifications. The reflectance area for 60[degrees] gloss measurements was 9 mm x 18 mm. The collection angle is [+ or -] 0.9[degrees] from the specular angle. All data presented in this report are the average of 36 measurements obtained from each of six different locations (six measurements per location). The estimated uncertainties in the gloss measurement presented in this article are at one standard deviation from the mean value for all 36 measurements.

Light Scattering

To ascertain the specular and nonspecular scattering distribution from the coating and the substrate surfaces, light scattering measurements were performed on both exposed and unexposed sections of the PVDF paint specimens, using a newly constructed light scattering instrument located in the NIST Building and Fire Research Laboratory. The new instrument consists of a laser light source, a five-axis goniometric specimen stage, and a two-dimensional detector mounted in a concentric ring around the specimen stage. The incident laser wave-length was 633 nm, and the beam was polarized and focused on the specimen with a diameter of 1 mm. The specimen rotation stage and the detector ring position determine the incident angle of the beam on the specimen and the viewing angle of the detector. Figure 1 presents the optical geometry, where [[theta].sub.o] and [[theta].sub.s] are the incidence and scattering angles measured with respect to the normal of the specimen. The sign convention is such that [[theta].sub.s] = - [[theta].sub.o] indicates the specular reflection angle. A detailed description of the instrument will be reported elsewhere. (5) In this study, we present the results in terms of the two-dimensional angular distribution of light scattered from a coating surface at incident angle of 60[degrees]. The results were compared to gloss and the LSCM results. The estimated uncertainties in the light scattering intensity data were around 4%.

[FIGURE 2 OMITTED]

RESULTS AND DISCUSSION

Effects of Weathering

Using the techniques previously described, a number of PVDF coatings, exposed in Florida for 10 years, were examined and compared against either control specimens that were prepared at the same time as the specimens that were exposed, or against the masked area of the exposed panels. For the exposed panels, some loss of coating gloss was visible to the naked eye at certain angles; however, as seen in Figure 2, the color change in these coatings was typically small (CIELAB [DELTA]E*<5). The typical gloss retention for masstone coatings was about 50% after 10 years of exposure in Florida. For panel A, the gloss of the masked area was 48, and the gloss of the exposed area was 21.

[FIGURE 3 OMITTED]

Changes in the coating surface topography after exposure were evident at the microscopic level, as shown in the 2D intensity projection images and selected sections of 3D topographic profiles in Figure 3. Figure 3a shows the masked area, while Figure 3b shows the exposed area. In the masked area, the binder (gray background) appears to be homogeneous--the deviations in the height and intensity profiles are small. A few bright spots are also visible in Figure 3a. For the exposed area, the overall intensity of the bright spots appears to be higher than in Figure 3a, while the number of bright spots appears to increase slightly. A detailed discussion of the differences in the scattered intensity distribution between the masked and exposed areas is presented below. However, from a comparison of different coating systems, it is clear that the bright spots represent pigment particles on or near the top surface, which scatter more light than the binder. According to the pigment product literature, the average size of the pigment particles for the coating on panel A is around 0.8 [micro]m. In the LSCM images, the size of the bright spots is much less than 0.8 [micro]m in height as well as in length, presumably because much of the pigment particles remain embedded in the binder. Atomic force microscopy (AFM) measurements were also performed to examine the topographic feature of the bright spots, and gave similar results. At a magnification of 150x, the LSCM RMS roughness was [S.sub.q] = 0.15 [micro]m for the masked area.

[FIGURE 4 OMITTED]

[FIGURE 5 OMITTED]

Unlike the image of the masked area, the 2D intensity projection images of the exposed area also showed many dark regions. The dark regions are associated with depressions or pits in the coating, as illustrated in the bottom topographic files of the selected section in Figure 3b and the height profile of these areas (Figure 4b). Typical sizes for the pits are 1-4 [micro]m in length (x-y directions) and 1-2 [micro]m in depth (z-direction). The surface roughness increases due to the height variation associated with pit formation. An RMS roughness of [S.sub.q] = 0.27 [micro]m was obtained from the exposed area. Figure 4 shows the intensity and height profile from representative topographic features (binder, pit, pigment) in the Figure 3b image. For this particular coating, after 10 years of exposure in Florida, the increase in the surface roughness and the decrease in coating gloss both appear to be dominated by pit formation as opposite to ablation of the coatings and consequential emergence of pigment particles.

Visually comparing the LSCM images for about a dozen weathered PVDF coatings, there generally appears to be an increase in the pigment density at the coating surface after weathering. Such an increase would be consistent with general models of erosion or contraction of pigmented coatings, (6,7) which link gloss loss to the emergence of pigment particles at the coating surface. However, the visual comparisons are based on a subjective assessment of the intensity of the bright spots in the image, and tend to confound areal density and scattering intensity effects. Many of the weathered coatings also show evidence of depressions or pits in the coating, at a 1 [micro]m-4 [micro]m length scale. However, it should be noted that pits or depressions are not observed in all PVDF coatings with this same degree of weathering. In comparing a series of coatings made with different single mixed metal oxide pigments, depressions are evident for some of the coatings, but not for others. The reasons for this difference are not understood. In the current study, we have chosen to conduct most of our analysis on coatings with pits, since this kind of surface feature, at this scale, has received little if any attention to date in published studies of pigmented coatings (a number of recent AFM studies (8,9) have reported smaller depressions or pits in studies of clear coats).

Comparison Between Surface Roughness and Gloss

Since most of the light reflected in gloss measurements is reflected off of the top surface of the specimen, we analyzed a number of weathered and unweathered PVDF paints to determine the degree of correlation between the gloss of the coatings and simple measures of the surface topography such as the root mean square roughness. LSCM and gloss measurements were conducted on 14 exposed coatings (exposed and masked areas), and 25 unexposed coatings (control). The exposed coatings had a variety of different weathered coating morphologies (some with pits, some without). To evaluate the variability of the measurement procedure, multiple measurements were performed on four of these coatings. Comparisons of gloss values and RMS surface roughness are shown in Figure 5 for different LSCM magnification scales: 5x and 150x. The different LSCM magnification levels probe different length scales of the coating surface, as are seen in the abscissa scales of the plots of the RMS roughness in Figure 5. Typical RMS roughness levels are in the micron range using 5x magnification, but are 10 times smaller at 150x magnification.

Figures 5a and 5c show that at the 5x magnification scale, the reproducibility of the RMS roughness measurement is good, and for these different PVDF-based coatings, a reasonably good correlation of gloss and RMS roughness is obtained. (10) ([R.sup.2] = 0.90, where R is the goodness of the linear fit.) Similar results were obtained at 20x magnification. However, at 150x magnification, there was a much lower degree of correlation, and the reproducibility of the RMS measurements was relatively poor, especially for the exposed areas. Biggs et al. have made a similar observation (11) in an AFM study of the weathering of polyester/melamine paint surfaces. They found limited correlation between RMS evolution and gloss loss in the QUV-B experiments, and little correlation between the relative initial RMS evolution rates; when comparing different accelerated weathering techniques. These findings illustrate that while high magnification images are very helpful to visualize physical damage in weathered coatings (e.g., the 150x images of Figure 3, or on an even finer scale, scanning electron microscope (12) or AFM images (13)), the coarser, longer wavelength surface roughness most directly relate to the gloss measurement, and to the visual perception of gloss. This result is in line with theoretical expectations. (14,15)

[FIGURE 6 OMITTED]

[FIGURE 7 OMITTED]

For selected cases, the RMS roughness was also measured on a smaller length scale using an AFM. Table 1 shows the values obtained for panel B, one of the specimens which exhibited pits in the exposed portion of the specimen, over a range of length scales from 10 [micro]m to nearly 2 mm. Figure 6 shows there is an almost linear dependence of the RMS roughness with the length scale of the measurement in a double logarithmic scale, i.e., RMS ~ (length scale)[.sup.f]. This result implies that there is an intrinsic fractal dimension of specimens with a local RMS roughness value reflecting structural characteristics at the corresponding measured length scale. In this case, the scaling factor f = 0.983 [+ or -] 0.078 for the masked area and f = 0.928 [+ or -] 0.039 for the exposed area. Assender et al. (16) and Johnson et al. (17) have observed similar results in other coating systems. Moreover, Johnson et al. (17) have identified the scaling region and roughness exponents for different exposure times for a polyurethane system, and also observed that the scaling factor decreased with increasing exposure time. Although it is too early to make any conclusions, fractal analysis may provide a link between local physical changes in nanoscale to macroscopic measurements, which can be used to predict gloss changes.

[FIGURE 8 OMITTED]

Standardization of LSCM 2D Intensity Projection Images

Since the intensity of the LSCM images is affected by factors such as optical alignment and detector gain, pixel intensity level analyses were conducted for selected images. For the panel A images, Figure 7 shows histograms of the number of pixels in each image at each detector intensity level--from 30 (black) to 255 (white). Note that in this case the lower intensity threshold, around 30, was set by the signal amplitude and detector gain of the LSCM instrument. Each of these histogram curves is the average of six different LSCM images at six different locations on the coating, obtained at the same time with the same detector gain.

The left part of each curve (Figure 7) represents the fraction of the surface image that is black or dark gray (darker than the average), areas which correspond to surface depressions (see topographic scans, Figure 4b). Comparison of the masked and exposed curves shows clearly that these depressions appear after exposure. The right part of each curve represents the fraction of the surface image that is brighter than the average, due to the reflectivity of pigment particles at or near the top surface of the paint. The pigment particles appear as bright spots because they scatter more light than the binder (Figure 4a). The pigment particles near the surface of the binder also have an effect on reflected light. The light scattered from these pigments goes through the layer of binder and is added to the light scattered from the top surface of the binder.

To obtain a more quantitative analysis of the distribution of pits and surface pigments in the 2D projection images, the large central peak in the pixel intensity histograms--due to the principal surface reflection--was modeled as a Gaussian curve. While it is common to use a Gaussian function to model the topographic profile of a rough surface, with the width of the Gaussian distribution being proportional to the RMS of the surface, (18) we are not aware of any special theoretical reason why a Gaussian fit would be the most appropriate in this case, where the scattered intensity is being modeled. Nevertheless, it can be noted from Figure 7a that the gray level intensity distribution for the unweathered (masked) coating fits the Gaussian form well on the low intensity side (where pigment contributions to the scattering are not important). The parameters obtained for the pixel intensity histograms curves from the masked and exposed areas of panel A are the pixel intensity level at maximum, full width at half maximum (FWHM), and the pixel intensity level range at half height. These values are shown in Table 2.

[FIGURE 9 OMITTED]

The Gaussian curve can be considered as the intensity distribution expected for binder surface reflections, without any contribution from pigment particles or pits. The difference between the experimental curve and the Gaussian fit (as shown in the bottom graphs of Figure 7) reflects the contribution of pigments and pit areas in the total 2D projection image. The peaks at intensity level 80-150 show the excess number of higher intensity pixels, from locations where surface pigments are present. The additional peak at intensity level 40-50, for the exposed specimen, shows the excess number of darker pixels in the image, from pit areas. From these curves, the areal contributions of the depressions and the pigments can be calculated directly.

Alternatively, the Gaussian fit information can be used to set threshold intensity levels for particular features of the image, to highlight particular topographic components of the surface, so that their distribution can be analyzed using various image analysis methods. Threshold levels should be set individually for each image, since the position and width of the central curve Gaussian fit depends on various LSCM instrumental factors. Using this approach, we have attempted to model the LSCM images of these coatings using four components: pits--binder--binder with subsurface pigments effect--top-surface pigment. Each of these components has a different light scattering behavior.

In assigning ranges of pixel intensity levels to each component of the coating, the full width at half maximum (FWHM), w, of the Gaussian fit is chosen as a characteristic parameter to describe the distribution (Figure 8). The Gaussian fit region was considered to reflect the scattering contribution of the binder. Then, using the difference curves in the bottom graphs of Figure 7, top surface pigment contributions were defined to be those excess pixels (in the [??] region) with an intensity at least 2.5w greater than the central peak maximum. Excess pixels (in the [??] region) with an intensity level between 0.5w and 2.5w greater than the central peak maximum were assigned to subsurface pigment particles. Excess pixels (in the [??] region) with intensity level less than -0.5w below the maximum were assigned to pits. The results of this analysis, for Florida panels A, B, and C, are shown as a percentage of the total area (Table 3). This analysis shows a significant increase in the area occupied by pits in the exposed coating, and a minimal increase in the density pigment particles on the top surface. For these coatings, the changes in the surface pigment scattering intensity due to exposure may be suggestive of some emergence of pigment particles, but the numerical differences are not statistically significant.

In several standard treatments of gloss loss and chalking for pigmented coatings, gloss loss is chiefly associated with the emergence of pigment particles at the coating surface, after an initial period of largely homogeneous erosion of material from the coating surface. (7,19-21) This "erosion" model does not implicate pits of micron dimensions as the proximal cause of gloss changes in some pigmented coatings. However, the "contraction" model of Colling and Dunderdale, which is based on the idea of coating binder mass loss not only at the surface, but as far into the coating as photons can penetrate, predicts that inhomogeneous distributions of pigment particles, such as pigment aggregates, should lead to the appearance of larger scale surface roughness features after weathering. (6) Those coating surface areas which are lacking in pigment, which should contract the most according to this model, would be predicted to evolve into depressions. This mechanism could be operative in commercial type PVDF paints, especially since they contain an acrylic component that is less inherently weatherable than the PVDF resin, which could photooxidize leading to binder contraction, which is inhomogeneous in the x-y plane. To test this hypothesis, we are planning experiments to compare the degree of dispersion of the pigments in the test coatings.

[FIGURE 10 OMITTED]

[FIGURE 11 OMITTED]

[FIGURE 12 OMITTED]

[FIGURE 13 OMITTED]

The Gaussian fit threshold criteria were used to create enhanced binary (black and white) images of either the pits (dark areas) or the pigments (bright areas) for further digital analysis using Fast Fourier Transform (FFT) and Autocorrelation Function (ACF) techniques. Figure 9 shows the images obtained for the masked and exposed areas of Panel B. Since the evolution of the surface and near-surface pigment distribution is not dramatic, we have chosen to focus most of our attention on the analysis of the pit distribution.

Fast Fourier Transform Analysis of LSCM Images

The freeware ImageJ (22) developed by NIH has been used to process the Fast Fourier Transform (FFT) calculation. Figure 10 shows the result of applying an FFT to the pit and pigment threshold images. The result of the FFT of the image is drawn in a log-log plot. The real size scale is directly given on the x-axis. Figure 10a shows the FFT curve of the pit regions. The curve corresponding to the exposed coating is shifted to the right. This means that the characteristic length of the pits (darker image regions) in the exposed area is larger than the characteristic length of those regions in the masked area. This can easily be seen visually, comparing images [alpha]1 and [beta]1 of Figure 9. However, considering the pigment images, the FFT curves for the masked and exposed area (Figure 10b) are not distinguishable. Thus, the FFT analysis is not appropriate for characterizing the changes in the pigment scattering image--changes which are primarily in intensity rather than spatial distribution (images [alpha]2 and [beta]2 of Figure 9).

Figure 11 shows dark region (pit) images obtained from panels A, B, and C, both in the exposed and masked regions. Figure 12 compares the FFT curves for four of these images. The FFT curves for all the masked images are essentially superimposed. To obtain a better comparison, a characteristic length [L.sub.FFT] is defined as the location of the "knee" in the curve (in a log-log plot). The values of [L.sub.FFT] were estimated and listed in Table 4. The results of the FFT analysis can be summarized as follows:

* Panel A-exposed and B-exposed present the same degradation (same size, same amount of pits) and the FFT of the threshold image gives the same result.

* Panel C-exposed presents smaller pits; the curve is shifted to the left compared to A-exposed and B-exposed, i.e., to smaller characteristic length.

* Panel B-masked is a masked area; it presents small dark features. Curve is shifted to the left compared to C-exposed, i.e., to even smaller length.

As seen in Table 4, the larger the [L.sub.FFT], the lower are the gloss values. This result is consistent with the observation that micron scale structures will be effective in increasing RMS surface roughness and decreasing gloss values in the exposed coatings. From these examples, it can be seen that an FFT analysis of LSCM images can be used to quantitatively describe the emergence of micron-scale pits on the coating surface. However, like the RMS roughness analysis, this method may not be sensitive enough to be used in the early prediction of coating gloss retention.

Autocorrelation Function Analysis Of LSCM Images

Some papers show that the use of autocorrelation functions (ACF) (23) is preferred when studying "random systems." (16) Contrary to FFT, this technique always remains in the spatial domain. For four of the Figure 11 images showing pits (black areas), the two-point correlation functions are determined using the following equation for an MxN pixels image:

S(x, y) = [M-x.summation over (i=1)][N-y.summation over (j=1)] [[I(i, j)I(x + i, y - i)]/[(M - x)(N - y)]] (3)

where I(i,j)=1 if the pixel at location (i,j) is black, and I(i,j)=0 otherwise. As th system studied is isotropic, the formula can be converted to a one-dimensional correlation function, S(r), depending only on the distance, r. (23)

S(r) = [1/[2r + 1]][2r.summation over (l=0)]S(r, [[[pi]l]/[4r]]) (4)

[FIGURE 14 OMITTED]

For comparison, results are presented in terms of the normalized two-point correlation functions where all functions are scaled to a range between 1 and 0, regardless of the value of S(0) in each image. The normalized two-point correlation function is calculated as:

N(r) = [[(S(r) - S(0) x S(0))]/[(S(0) - S(0) x S(0))]] (5)

where S(0) represents the area fraction of the phase of interest. (23) In this case, S(0) is the average area fraction of pits in the image. A graph of the normalized two-point correlation functions N(r) for four of the Figure 11 images is shown in Figure 13.

A characteristic length of the N(r) function, [L.sub.ACF], is defined as the intersection of the x-axis and the tangent to the N(r) curve at x = 0. Results for [L.sub.ACF] and S(0) are shown in Table 5. The results for the autocorrelation function follow the same trends as the FFT analysis, with the higher [L.sub.ACF] value reflecting a larger average size of the dark domains (pits) in the images. In general, for coatings with pits, we expect that the gloss will depend on the RMS roughness, [L.sub.ACF] and S(0).

Preliminary Studies of the Light Scattering Angular Distribution

As the above discussion makes clear, it is possible to use LSCM to accurately characterize the distribution of important topographical components of the coating surface. Moreover, from the theory of gloss, it is expected that for the most part, these components, such as near-surface pigments and surfae pits or depressions, will act as independent diffuse scatterers of the incident light. To fully link gloss changes to the surface morphology changes resulting from weathering, an analysis of the nonspecular diffuse light scattering distribution becomes vital.

[FIGURE 15 OMITTED]

We are now beginning such light scattering studies. Figure 14 shows scattering distributions for panels A, B, and C at the 60[degrees]-60[degrees] scattering configuration (at incident angle, [[theta].sub.o] = 60[degrees], and detection angle, [[theta].sub.s] = -60[degrees]). It may be observed that the angular spread and shape of the intensity distribution is different for the three coatings, with a narrower angular spread for panel B, and for each coating, less near-specular scattered intensity for the lower gloss exposed area.

Using a short exposure time to avoid detector saturation, the total scattered intensity was calculated for the 60[degrees] [+ or -] 0.9[degrees] detection angle range, corresponding to the gloss meter collection angle. The results, shown in Table 6, show a general trend consistent with the gloss measurements, but the two quantities are not strictly proportional. The reasons for this discrepancy might be due to factors such as the difference in incident light source (single wavelength versus continuous spectrum), light absorption effects from the pigment, and differences between the two methods in the collecting geometry and the size of the specimen area.

The off-specular scattering portion--more than [+ or -] 0.9[degrees] off of the specular reflection angle--is not collected in a typical 60[degrees] gloss measurement. The scattering intensity profile in this region can provide information about the roughness at a shorter wavelength scale, including information about the scattering from pigment particles. Figure 15 shows the scattering profiles for panel B over the 60[degrees]-57[degrees] scattering angle range. Again, less near-specular scattered intensity is observed for the lower gloss exposed area of the test panel, but more off-specular diffuse scattering is seen (see the bottom graph in Figure 15). The off-specular scattered intensity from the exposed surface (3.85 x [10.sup.4] counts) is almost twice stronger than that from the masked area (2.13 x [10.sup.4] counts). We are now conducting light scattering experiments to fully measure the off-specular scattering profiles.

CONCLUDING REMARKS

We have investigated the relationship between the gloss loss and the change in surface morphology for PVDF paints weathered for 10 years in Florida using laser scanning confocal microscopy (LSCM). Paints made with different pigment grades showed different degradation characteristics after exposure. Changes in gloss have been correlated with surface roughness and other topographical features, including the formation of micron-scale pits and the distribution of pigment particles at the surface.

For PVDF coatings exhibiting pits, the increase in RMS surface roughness, and the decrease in the coating gloss after exposure, appear to be dominated by the formation of pits, rather than by the emergence of pigment particles at the coating surface. To our knowledge, this is the first time that pits of these dimensions have been reported as the principal proximal cause of gloss changes in pigmented PVDF coatings. Often, theoretical treatments (6,7,21) for gloss loss in pigmented coatings focus instead on the role of emerging pigment particles at the coating surface.

For a relatively large specimen set, a linear relation was found between the 60[degrees] gloss and the RMS roughness measured using the LSCM at a low (5x) magnification (10-1000 [micro]m length scale). At higher magnification (smaller length scales), a similar trend exists, but with a lower degree of statistical correlation. Moreover, a scaling relationship was found between the local RMS roughness value and the corresponding measured length scale. This result implies that each coating surface has an intrinsic fractal surface dimension, implying that fractal analysis may potentially provide a link between physical changes at the nanoscale and macroscale levels, which could be used in service life predictions of gloss retention.

Many of the weathered specimens have prominent micron-scale pits or depressions on the surface. For a number of specimens displaying pits after weathering, the histograms of scattered light intensity for unexposed (masked) and exposed areas of the paints revealed a minimal increase in the density of pigment particles at the top surface due to weathering. Using the intensity histograms, criteria can be established to generate binary images of the pits and the surface pigments. These images were analyzed using the Fast Fourier Transform (FFT) and Autocorrelation Function (ACF) techniques. The techniques generate characteristic lengths ([L.sub.FFT], [L.sub.ACF]) which can be used to quantify weathering-induced changes in the local surface morphology. For both techniques, the characteristic lengths determined from binary images of the pits, especially at high magnification, show a good correlation with macroscopic gloss measurements (the larger the characteristic length of the pit spatial distribution, the lower the gloss values).

To attempt to establish a more quantitative link between the scattering of surface topographical features with the measured coating gloss, we are initiating studies to look at the full angular distribution of scattered light from the coatings, using a new light scattering apparatus developed at NIST. In addition to these studies of the relation of the coating topography to the gloss, mechanistic studies of the coating degradation are underway using the 2-meter-Integrating Sphere (24) developed by NIST, as well as other kinds of accelerated test cabinets.

ACKNOWLEDGMENTS

The author (Jenny Faucheu) gratefully acknowledges funding support from Arkema Inc./Total America Services. The authors also thank Drs. Dale Bentz and Jeffrey Bullard of NIST for insightful discussion and their help in autocorrelation function analyses.

References

(1) Iezzi, R.A., "Fluoropolymer Coatings for Architectural Application," Modern Fluoropolymers, John Wiley & Sons, Chap. 14, 271, 1997.

(2) See for instance Nguyen, T. et al., "Relating Laboratory and Outdoor Exposure of Coatings: II. Effects of Relative Humidity on Photodegradation and the Apparent Quantum Yield of Acrylic-Melamine Coatings," J. COAT, TECHNOL., 74, No. 932, 65 (2002), and a number of articles in Bauer, D.R. and Martin, J.W. (Eds.), Service Life Prediction: Methodology and Metrologies, ACS Symposium, Series 805, 2002.

(3) Wood, K.A., Cypcar, C., and Hedhli, L., "Predicting the Exterior Durability of New Fluoropolymer Coatings," J. Fluorine Chem., 104, 63 (2004).

(4) Corle, T.R. and Kino, G.S., "Confocal Scanning Optical Microscopy and Related Imaging Systems," Academic Press, pp. 37-39, 1996.

(5) Sung, L.P., Garver, J., Embree, E., Dickens, B., and Martin, J.W., "A Novel Light Scattering Instrument for Characterizing Polymeric Coating Materials," NIST Internal Report (2006).

(6) Colling, J.H. and Dunderdale, J., "The Durability of Paint Films Containing Titanium Dioxide--Contraction, Erosion and Clear Layer Theories," Prog. Org. Coat., 9, 47 (1981).

(7) Braun, J.H., "Titanium Dioxide's Contribution to the Durability of Paint Films," Prog. Org. Coat., 15, 249 (1987).

(8) VanLandingham, M.R., Nguyen, T., Byrd, W.E., and Martin, J.W., "On the Use of the Atomic Force Microscope to Monitor Physical Degradation of Polymeric Coating Surfaces," J. COAT. TECHNOL., 73, No. 923, 43 (2001).

(9) Nguyen, T., Gu, X., VanLandingham, M.R., Byrd, W.E., Martin D., Ryntz, R., and Martin, J.W., "Degradation Modes of Polymeric Coatings Exposed to UV," Proc. 7th International Coatings for Plastics Symposium, Troy, MI, June 2004.

(10) Faucheu, J. and Wood, K.A., Proc. 3rd International Symposium on Service Life Prediction of Coatings, Sedona AZ, February 2004.

(11) Biggs, S., Lukey, C.A., Spinks, G.M., and Yau, S.-T., "An Atomic Force Microscopy Study of Weathering of Polyester/Melamine Paint Surfaces," Prog. Org. Coat., 42, 49 (2001).

(12) Wood, K.A., Hedhli L., and Willcox, P.J., "Patterns of Erosion from Acrylic and Fluoropolymer Coatings in Accelerated and Natural Weathering Tests," J. COAT. TECHNOL., 74, No. 924, 63 (2002).

(13) Gu, X., Sung, L., Ho, D.L., Michaels, C.A., Nguyen, D., Jean, Y.C., and Nguyen, T., "Surface and Interface Properties of PVDF/Acrylic Copolymer Blends Before and After UV Exposure," Proc. 80th Annual Meeting of the Federation of Societies for Coatings Technology, Oct. 30-Nov. 1, New Orleans, LA, 2002.

(14) Alexander-Katz, R. and Barrera, R.G., "Surface Correlation Effects on Gloss," J. Polym. Sci. Part B: Polym. Phys., Vol. 36, 1321 (1998).

(15) Whitehouse, D.J., Bowen, D.K., Venkatesh, V.C., Lonardo, P., and Brown, C.A., "Gloss and Surface Topography," Annals of the CIRP, Vol. 43, 541 (1994).

(16) Assender, H., Bliznyuk, V., and Porfyrakis, K., "How Surface Topography Relates to Materials' Properties," Science, Vol. 297, 973 (2002).

(17) Johnson, M.A. and Cote, P.J., "Detrended Fluctuation Analysis of UV Degradation in a Polyurethane Coating," J. COAT. TECHNOL., 75, No. 941, 51 (2003).

(18) Ginneken, B.V., Stavridi, M., and Koenderink, J.J., "Diffuse and Specular Reflectance from Rough Surface," Appl. Opt., 37, 1, p. 130 (1998).

(19) Braun, J.H., "White Pigments," Federation Series on Coatings Technology, FSCT, Blue Bell, PA, 1995.

(20) Falla, N., Surf. Coat. International, 81 (8), 375-380 (1998).

(21) Hunt, F.Y., Galler M.A., and Martin J.W., "Microstructure of Weathered Paint and Its Relation to Gloss Loss: Computer Simulation and Modelling," J. COAT. TECHNOL.; 70, No. 880, 45 (1998).

(22) The NIH freeware ImageJ can be downloaded from http://rsb.info.nih.gov/ij/.

(23) Bentz, D.P., "Quantitative Comparison of Real CEMHYD3D Model Microstructures Using Correlation Functions," in press, Cement and Concrete Research (2006).

(24) The current output at each port of the 2-meter-Integrating SPHERE is about 22 "sun"--Intensity of 1 "sun" is equal to 22 W/[m.sup.2] (ASTM E891) and is rich in the region between 290 nm and 400 nm. For technical information: Chin, J.W., Byrd, E., Embree, N., Martin, J., and Tate, J.D. "Ultraviolet Chambers Based on Integrating Spheres for Use in Artificial Weathering." J. COAT. TECHNOL., 74, No. 929, 39 (2002).

Jenny Faucheu ([dagger]) and Kurt A. Wood -- Arkema Inc.*

Li-Piin Sung and Jonathan W. Martin -- National Institute of Standards and Technology (NIST) ([dagger])

Presented at the 82nd Annual Meeting of the Federation of Societies for Coatings Technology, October 25-27, 2004, in Chicago, IL.

* 900 First Ave., King of Prussia, PA 19406.

([dagger]) 100 Bureau Dr., Stop 8615, Bldg. 226, Room B366, Gaithersburg, MD 20899.

(a) Certain instruments or materials are identified in this paper in order to adequately specify experimental details. In no case does it imply endorsement by NIST or imply that it is necessarily the best product for the experimental procedure.

(b) Panel B has a slightly different PVDF grade than that of Panel A.

Table 1 -- RMS Roughness (Panel B) on Different Length Scales Using AFM
and LSCM

                            AFM                       LSCM

Scan size             10 x 10  30 x 30  61.4 x 61.4  184 x 184  1842 x
  ([micro][m.sup.2])                                              1842
RMS         Masked     0.04     0.06     0.10          0.23        3.28
([micro]m)  Exposed    0.07     0.11     0.27          0.63        5.47

Table 2 -- Parameters Obtained for the Pixel Intensity Histogram Curves
from Masked and Exposed Areas of Panel A (Figure 7, Panel A)

                                            Masked    Exposed

Pixel intensity level at maximum            80        70
Full width at half height (FWHM, w)         24        30
Pixel intensity level range at half height  68 to 92  55 to 85

Table 3 -- Percentage ([+ or -]3%) of the Total Area[.sup.[phi]] Covered
by Each Component from Pixel Intensity Histogram Analysis for Florida
Panels A, B, and C

                                Subsurface  Top-Surface
                  Pits  Binder  Pigment     Pigment

Masked   Panel A   0    92       7.5        0.5
         Panel B   0    89.5    10          0.5
         Panel C   0    87.5    12          0.5
Exposed  Panel A  13    75      10.5        1.5
         Panel B  14    72      11          3
         Panel C   8    82       8.5        1.5

Table 4 -- Characteristic Length ([L.sub.FFT]) from FFT for the Four
Threshold Images (Figure 11) and the Corresponding Gloss Values

                                             Panel A  Panel B  Panel C

Characteristic length [L.sub.FFT],  Masked    1.12     1.06     1.03
  [micro]m ([+ or -]5%)             Exposed   1.39     1.41     1.28
60[degrees] gloss ([+ or -]1)       Masked   48       57       55
                                    Exposed  21       23       30

Table 5 -- Characteristic Length, [L.sub.ACF], and Density of Pits,
S(0), from the Normalized Two-Point Correlation Functions N(r) for the
Four Threshold Images (Figure 11) and the Corresponding RMS Roughness,
and 60[degrees] Gloss Values

                                             Panel A  Panel B  Panel C

S(0) ([+ or -]5%)                   Masked    0.11     0.11     0.10
                                    Exposed   0.23     0.29     0.19
Characteristic length [L.sub.ACF],  Masked    2.35     2.17     2.19
  [micro]m ([+ or -]5%)             Exposed   3.19     3.43     2.71
60[degrees] gloss ([+ or -]1)       Masked   48       57       55
                                    Exposed  21       23       30

Table 6 -- List of 60[degrees] Gloss Values, RMS Roughness Values from
LSCM-150x, and the Corresponding Total Light Intensity Obtained from
Light Scattering Measurement for the Masked and Unmasked Areas of Panels
A, B, and C (a)

                                                    Total Light
                                     RMS, [micro]m  Intensity in the
                                     LSCM 150x      Central Region
                  60[degrees] gloss  ([+ or -]0.05  (x[10.sup.-6] count,
                  ([+ or -]1)        [micro]m)      [+ or -] 4%)

Masked   Panel A      48                 0.15           5.16
         Panel B      55                 0.10           5.34
         Panel C      57                 0.10           5.53
Exposed  Panel A      21                 0.27           4.08
         Panel B      23                 0.27           3.88
         Panel C      30                 0.22           4.17

(a) Here, total light intensity was obtained by integrating the
scattered light intensity (exposure time = 0.03125s) within the angular
range [+ or -]0.9[degrees] from specular angle.