Mechanical property changes and degradation during accelerated weathering of polyester-urethane coatings.

Chemical changes, measured using spectroscopy, and crosslink density, measured by mechanical thermal analysis, were determined during accelerated weathering on a model polyester-urethane coating of known composition. The tensile modulus, measured above the glass transition temperature, and thus the cross-link density, decreased with exposure, as expected from the chemical changes. However, the tensile modulus, measured at room temperature, increased with exposure. Physical aging of the polymer network was found to occur concurrently with photodegradation and accounts for much of the increase in room temperature modulus. Increased hydrogen bonding in the increasingly oxidized polyester-urethane may also contribute to the increase in modulus at room temperature. Both physical and chemical changes must be determined if changes, and rates of change, in performance due to weathering are to be understood.

Keywords: Thermal analysis, photodegradation, polyurethanes, durability, mechanical properties, physical properties, weatherability, aerospace, aircraft

**********

Coatings are used in many applications and environments where high levels of durability are required. Often, the requirements include maintenance of appearance, barrier properties, toughness, etc. during exposure to natural weathering, which includes solar ultraviolet radiation, heat, moisture, and pollutants. In many cases, the lifetime of the coating is determined by a change in appearance, but crucial equipment may rely on the protective properties of the coating. However, understanding the relationship between coating composition and how it changes during exposure and how that is related to macroscopic end-use properties remains a fertile field of research. Many protective attributes of a coating can be related to its mechanical integrity, e.g., chemical and corrosion protection depend crucially on whether a coating remains intact and continuous during service. Cracks or adhesion failures due to in-service stresses immediately define the useful service life of a coating. These macroscopic failures are the accumulations of, or initiated by, the erosion processes that result from chemical changes during degradation.

Understanding how the mechanical integrity of a coating, or any construction material, changes under exposure is very important. However, mechanical measurements, while relating closely to end-use and service life, are not directly linked to the chemical changes in the material and, thus, how the environment affects a polymer. It is necessary to examine the chemical changes, relate those to changes in the polymer network, and include any other physical changes that occur before understanding the important factors that determine the macroscopic mechanical properties of a coating.

There has been a great deal of excellent basic work done in determining the chemical changes that occur during degradation of a particular coating's chemistry. Unfortunately, further progress towards connecting that information with macroscopic properties has often been hindered by examining commercial polymers that are undisclosed complex mixtures of functionality and composition. This work examines model polyester-urethane systems where the composition is completely accessible and measures the changes in crosslink density and other properties with the chemical changes caused by exposure. The use of a known model composition better permits the results of several techniques or hypotheses to be checked and tested. Thus, some progress may be attempted in relating composition to end-use properties and understanding the mechanisms that govern, for example, the mechanical properties.

Changes in mechanical properties, with exposure, were measured in order to ascertain a measure of how coating integrity changed. Since degradation due to irradiation or an aggressive external environment inevitably starts at the outer surface, this study used nanoindentation to measure the surface properties, as well as conventional tensile testing to determine bulk mechanical properties. Fourier transform infrared spectroscopy (FTIR) was used here, as elsewhere, to establish that the model compound degraded in a typical fashion, to quantify the rate of chemical changes, and to measure average relative thickness loss. Similarly, ultraviolet spectroscopy was used as a complement to FTIR for determining chemical changes. Dynamic mechanical thermal analysis (DMTA) provided insight to changes in network structure with exposure via the crosslink density and the glass transition temperature. Differential scanning calorimetry (DSC) provided a method of quantifying changes in polymer relaxation behavior with exposure.

EXPERIMENTAL PROCEDURES

Polyester Synthesis

Trimethylol propane (TMP), neopentyl glycol (NPG), adipic acid (AA), and isophthalic acid (IPA) (all obtained from Aldrich) were used directly from the supplier without further purification. The polyester was synthesized using the same monomer ratios, reaction conditions, and equipment as in previous work. (1) The molecular weight was determined by gel permeation chromatography (GPC) calibrated against polystyrene standards. Hydroxyl functionality was measured using a titration method according to ASTM D 4274-94.

Polyurethane Formulation and Sample Preparation

Hexane diisocyanurate trimer (HDI trimer), Desmodur N3300 (Bayer), was used directly from the supplier without further purification. To ensure all hydroxyls would be reacted, a ratio of 1:1.1 hydroxyl:isocyanate was used. In all formulations a 0.5 wt% catalyst, dibutyltin dilaurate (DBTDL) (Aldrich) was added. Formulations were diluted to 60% solids by weight with a mixture of methyl ethyl ketone (MEK) and methoxy-propane-acetate (MPA). Coatings made from these materials were either spin-coated or cast using a drawdown bar, depending upon purpose. Samples used for mechanical testing were cast on DuPont Tedlar[R], a polyvinyl fluoride film of approximately 50 [micro]m thickness. Tedlar acted as the substrate during weathering and enabled easy removal of a coating from the substrate to obtain free films (of about 30-50 [micro]m dry film thickness) needed for tests. Samples were allowed to fully cure at room temperature for one week prior to accelerated weathering. Coatings had gel times of approximately two hours when cured at ambient conditions. For mechanical testing, samples were cut into rectangular strips of approximately 130 mm long and 30 mm wide, which were weathered prior to cutting standard dumbbell shaped samples (ASTM D 412) for tensile testing. Any remnant film was used for other physical testing.

Exposure Conditions for Accelerated Weathering

All samples were exposed in a Q-SUN 1000[TM] equipped with a xenon arc light source with borosilicate filters, made by Q-Panel Company. The exposure conditions were cycles of four hours light (0.55 W/[m.sup.2] at 340 nm) at 60[degrees]C and four hours dark with water spray at 25[degrees]C according to ASTM D 4587-91. Exposure chambers may not emulate natural exposure closely, but they do provide a controlled and repeatable environment; natural exposure is variable and uncontrolled.

Fourier Transform Infrared Spectroscopy

The infrared absorption spectra were obtained in transmission mode using a Nicholet Magna 850 Fourier transform infrared spectrometer. Substrates were polished silicon infrared transmission windows, 13 mm diameter and 2 mm thick, available from Wilmad Glass. The samples were spun-coat at 4000 rpm for 30 sec at 60% solids. The coating thickness was approximately 5 [+ or -] 0.1 [micro]m to keep the absorbance peak below 1.5 approximately. The spin coater was a model WS-200-4NPP/RV from Laurell Technologies Corporation. Individual samples were secured into custom-made sample holders to avoid any movement and ensure that the samples could be placed with the same orientation to the beam in the FTIR each time.

Fringe spacing was measured using a software program, Peakfit[R] (2.01), using a sine-squared function to match the fringe spacing in the IR spectra. The spacing was recorded for calculating optical path length and then coating thickness.

For further analysis of the IR spectra, the interference fringes were subtracted from the spectra. (2) This technique is reliable as long as there is a sufficient number of fringes observed in the IR spectra. Once the spacing becomes too broad, for example, it was difficult to separate the broad hydroxyl region peaks from the interference fringes.

Ultraviolet Spectroscopy

Ultraviolet spectra were obtained in transmission mode with a Cary 5000 spectrometer from Varian Instruments. Samples were prepared on quartz slides purchased from Chemglass Inc. and spin-coated using the same conditions to achieve a film thickness on the order of 5 [micro]m. In previous work, spectra below 250 nm wavelengths could not be resolved due to the large absorbance of the polyurethane. (3) To achieve useful spectra below 250 nm, additional samples were prepared with a film thickness of 700 nm by diluting the formulation to 18% solids and spin-coating at 4000 rpm.

Dynamic Mechanical Thermal Analysis (DMTA)

Two different experiments were conducted on a Rheometric Scientific DMTA-3E model. One was to obtain the tensile storage modulus (E') and loss tangent (tan [delta]) curves using a conventional temperature ramp. Free films with a sample size of 10 x 5 x (0.03 to 0.05) mm were used. The experiment was carried out from -20[degrees]C to 200[degrees]C, increasing at 5[degrees]C/min at a frequency of 1 Hz and using a strain amplitude of 0.5%. The objectives were to determine the variation in [T.sub.g] of the coating as it weathered and also to obtain the E' values at the rubbery plateau region, around 50[degrees]C greater than the [T.sub.g], to calculate the crosslink density. The maximum in tan [delta] was typically 80[degrees]C, but it was found that at temperatures greater than 100[degrees]C the films became very soft and it was difficult to get reliable results using settings useful for the start of the temperature scan. Instead, a second mode was used, where a constant strain-rate test was done at a single elevated temperature (an isothermal test) giving a stress-strain curve to obtain elastic modulus values. For this test, 130[degrees]C was selected for the temperature since it was well beyond the maximum tan [delta] peak temperature of the coating, even after weeks of weathering, and it is assumed that the material has only rubbery response at this temperature. The sample geometry remained the same as in the previous mode, except the sample compartment temperature was raised to 130[degrees]C and held throughout the experiment. The samples were allowed to equilibrate to this temperature for 10 min, and then a strain rate of 1 x [10.sup.-3] [sec.sup.-1] was applied to the sample. Crosslink density was calculated from the following equation, where [v.sub.e] is the crosslink density. (4,5)

E = 3[v.sub.e]RT (1)

A crosslink functionality of 3 is probably typical of the model polyester-urethane used here, but it was not included in the calculations of crosslink density, although it may be used. (4,6) Crosslink density was expected to change with exposure, which means that the crosslink functionality may change also. Experimentally determined values of crosslink density are often made from measurements at small strains, as here, when the simple equation above is more likely to apply since the network deformation is affine. (4) Equation (1) was also used because it implies the fewest assumptions, and because the contribution of entanglements to the modulus is difficult to quantify. (4,7,8) The values obtained were consistent with what might be expected from the model polymer. (1)

Tensile Property Measurements

It is generally observed that coatings become more brittle upon exposure, so tensile properties at room temperature were measured after each exposure increment. The sample geometry used was the dumbbell shape as given in ASTM D 412-98a, die type C. As elsewhere, sample strips of ~130 mm x 30 mm were exposed on Tedlar substrates and samples were obtained for tensile testing by cutting these strips with the standard die using a press. Tensile testing was done according to ASTM D 882-02. The overall distance between the grips was maintained at 90 mm for the whole dumbbell shape. However, for the calculations the effective load bearing length of 6 mm width (assuming an elastic material) was determined to be 52.3 mm. This equivalent gauge length was calculated by assuming that the load is distributed evenly across the sample cross-section everywhere and then calculating the stress and strain at each position for an elastic material. The effective gauge length is calculated as the length of elastic material of the same width as the main straight (narrow) section of the dumbbell shape that would have the same overall length change for that applied load. An Instron[R] 5542 tensile testing machine was used with a crosshead speed of 5.2 mm/min to give approximately the same strain rate as that used at 130[degrees]C to measure modulus and, thus, crosslink density. Three samples for each cycle of weathering were tested. It was possible to obtain free films up to six weeks of weathering. Beyond this point the films became brittle and could no longer be removed from the Tedlar substrate without significant damage.

Nanoindentation Measurements

The surface layer of a polymer film is the most susceptible to damage under photodegradation conditions. It is very important to follow the mechanical changes that are happening on the surface since those changes may be the limiting factor in the mechanical integrity of the coating. For this investigation, nanoindentation identified whether the surface properties changed in the same way as the bulk mechanical properties. The nanoindentation method is increasingly being used to characterize the surface mechanical properties of many materials.

Nanoindentation simultaneously records the load and the penetration depth during both loading and unloading cycles. The reduced modulus, E*, which is a combination of the properties of the sample and the indenter tip, can be calculated using the following equation (9) from the force-indentation curves (Figure 1).

[dF]/[dh] = 2[h.sub.p]E*[square root of (24.5/[pi])] (2)

In equation (2), dF/dh is the rate of change of indentation force, F, with penetration depth, h, on the linear part of the unloading curve, and [h.sub.p] is the intercept on the indentation axis, as shown in Figure 1. Results were obtained from the experiment by using the procedure proposed by Oliver and Pharr. (10,11)

Young's modulus of the coating can be determined from the following equation,

[1/[E*]] = [(1-[v.sup.2])/E] + [(1-[v'.sup.2])/E'] (3)

Where E* is the "reduced" modulus, E and v the modulus and Poisson's ratio of the polymer, and E' and v' the modulus and the Poisson's ratio of the indenter tip material (diamond).

[FIGURE 1 OMITTED]

[FIGURE 2 OMITTED]

The experiments were conducted using a Dimension 3100[R] atomic force microscope with a Nanoscope IIIa controller (Digital Instruments) and a Berkovich type diamond indenter probe (Digital Probes). The spring constant for the probe cantilever was 177.3 N/m. Force-deflection curves from a standard sapphire substrate were used for calibration of the deflection voltage and found to average 196 nm/V. Two different threshold force values (10 [micro]N and 16 [micro]N) were used to obtain an array of indentations after each weathering cycle of interest. Each small sample of coating still on Tedlar substrate was secured onto a small aluminum panel using a double-sided adhesive tape. The substrate effects were assumed to be negligible. A smooth area (~ 4 [micro]m x 4 [micro]m) was located using normal AFM tapping mode. Rough areas (at [micro]m scale) of the coating were avoided so that indentation profiles could be determined cleanly. Six data sets were obtained for each weathering cycle. For equation (3), the Young's modulus and Poisson's ratio for diamond (12) was taken as 1140 GPa and 0.07, respectively, and the Poisson's ratio of the polymer coating was estimated to be 0.4. (13)

[FIGURE 3 OMITTED]

Physical Aging

Physical aging is a long-term relaxation process in polymers that occurs in its glassy state, below the glass transition temperature, as the macromolecules gradually approach thermodynamic equilibrium. (14) It occurs in the absence of any chemical changes or any aggressive external environment. Physical aging has been found to be an important factor in the room temperature physical properties of many polymers. (15-18) There are a few descriptions of physical aging in coatings, (19-21) but it has been pointed out that the connection between physical aging and long-term environmental degradation is unresolved. (14) Therefore, it is important to identify the extent of physical aging that the coatings undergo in the (accelerated) weathering examined here. Physical aging may be measured by first determining a material property of interest, then "erasing" the sample's thermal history by annealing it, at a temperature higher than its [T.sub.g], then remeasuring the property.

Two methods were used to quantify physical aging. DSC was used in a conventional manner to measure the enthalpy relaxation (13) (or enthalpy recovery) around the glass transition after an exposed sample was annealed at higher temperature. DSC was performed on a TA Instruments[R] Q 1000 series calorimeter, with a sample size of 2-3 mg. Samples were first equilibrated at -20[degrees]C and then measured while increasing the temperature at 5[degrees]C per minute up to 130[degrees]C. They were annealed by holding at 130[degrees]C for 15 min, then cooled down to -20[degrees]C, equilibrated, and then tested again at the same rate as before to 130[degrees]C. The area between the two thermograms is a measure of enthalpy characteristic of aging. (14)

[FIGURE 4 OMITTED]

The second method determined changes in room temperature modulus. Constant strain rate tensile testing was done on a sample size of 10 x 5 x (0.03 to 0.05) mm using the DMTA apparatus. First, the tensile modulus of the exposed (aged) sample was measured at room temperature with a constant strain rate of 1 x [10.sup.-4] [sec.sup.-1] up to a 1% strain. It was important not to break this sample at this juncture, hence the use of a low strain rate and strain limit. After this test, the clamps were slightly loosened to reduce clamping stresses, but the sample was left in the clamps. Next, the heating chamber was used to anneal the sample at 130[degrees]C for 15 min. After cooling to room temperature again, another tensile test was run to characterize the modulus in the annealed sample.

RESULTS AND DISCUSSION

Spectroscopic Determination of Changes To the Crosslinked Polymer

In previous work it was shown that interference fringes could be used to measure changes in the coating optical path length or thickness. (1) Thickness changed approximately linearly over a 10-week period with a 50% decrease in overall thickness for a polyurethane coating that was only 40% crosslinked. The study here used a fully crosslinked formulation of the same polyester-urethane, which proved to be more durable. It required 15 weeks of exposure, with only a ~20% decrease in relative thickness, to reach the same level of oxidation as the 40% crosslinked polyester-urethane at ~50% thickness loss. Figure 2 shows the chemical changes that took place in the fully crosslinked formulation during artificial exposure. There is a 120% increase in the OH-NH absorption, from 3600 to 3100 [cm.sup.-1] (peak area divided by thickness, using the fringe spacing to calculate an analogy to concentration). Calculated in a similar fashion, the carbonyl area, 1625-1825 [cm.sup.-1], shows an increase of ~20%. The results show that the coating is being oxidized, as expected. The CH peak region shows an approximate 30% decrease over the same exposure period.

Similarly, it is possible to quantify the amount of a particular bond remaining, for example, using the isophthalic acid peak at 731 [cm.sup.-1] or urethane linkage peak at 1540 [cm.sup.-1]. There is a 50% decrease in the urethane linkage. There is also a 22% decrease in isophthalic acid peak, indicating that the polyester backbone is dissociating. With the scission of these bonds, it could be expected that the overall crosslink density will decrease. Gerlock (22) observed the same loss in both isophthalate and urethane peaks of related coatings in artificial exposure, but did not witness much change under natural exposure conditions from Florida or Arizona exposure. An explanation was given that the artificial exposure had a higher intensity of short wavelength photons degrading the coatings by a different mechanism. (22)

In our previous study of a partially crosslinked coating, half of the coating was eroded away by the end of the exposure period. (1) In order to remove a fragment of network polymer, at least two bonds must be broken, whereas a dangling chain requires only one chain scission to become a fragment that can be removed from the coating. It was observed that the fully crosslinked system only lost 20% of the original thickness at approximately the same level of oxidation. Thus, the effect of the greater level of crosslinking (and the fewer unreacted dangling chains) in the fully crosslinked coating is readily apparent. Although approximately the same amount of oxidation took place, the overall coating thickness did not decrease as much.

[FIGURE 5 OMITTED]

During the same exposure period, the ultraviolet absorption of the degrading films was measured. Fringes were present in the spectra, so it was possible to measure relative coating thickness changes in the same way, which were in excellent agreement with the thickness changes from the FTIR data. In a previous article it was shown that there was a small absorption tail created (yellowing) above 300 nm into the visible region. (3) Yellowing may be associated with the broadening of a large absorption peak below 250 nm (possibly [pi]-[pi]* transitions) (3,23) or due to photooxidation products from peroxide formation. (24) Here, these transitions were examined by making a set of very thin coatings, ~700 nm thickness, to resolve the peaks below 250 nm (Figure 3a). It can be seen that the IPA [pi]-[pi]* transition does broaden and could be the source for the tail into the visible region of the spectra. (21) The broadening is probably due to the absorption by photooxidation products. The peaks at 280-290 nm wavelength are commonly ascribed (23) to the n-[pi]* transition from the carbonyl on the isophthalic acid portion of the model polyester (Figure 3b). When the peak height is divided by the thickness of the coating, results (average of five replicates) show a 22% decrease in the n-[pi]* transition (Figure 4). This was the same decrease that was seen from the FTIR data. There is good correlation between both spectroscopy techniques.

[FIGURE 6 OMITTED]

[FIGURE 7 OMITTED]

Ultraviolet peak assignments were determined by using the component materials and a model compound to examine the urethane linkage, as seen in Figure 5. Solution concentrations were chosen to give clear spectra and were typically [10.sup.-3] mole/liter. Hexane diisocyanurate was reacted with ethyl alcohol to form a model urethane compound. The ethyl urethane shifted the absorption compared to the HDI trimer, as shown in Figure 5a. In closer examination, a very weak absorption peak evolved at wavelengths between 260-275 nm (Figure 5b). This peak was determined to be the urethane linkage after the spectra was compared to a spectrum of the starting hexane diisocyanurate. These spectra were taken from very dilute solutions of the model urethane in order to distinguish the entire spectra. Generally, coatings would have much larger absorptions due to higher concentrations. Prior to this study, it was assumed that the urethane linkage was being oxidized through hydrolysis and not by the UV photolysis. Now it can be seen that the urethane linkage may undergo photochemical degradation.

[FIGURE 8 OMITTED]

Results in this study are in good accord with photodegradation mechanisms previously reported. (1,25,26) The model polyester had an average of six hydroxyl units per chain length and an average of seven isophthalate groups per chain. Therefore, if the system were fully crosslinked there would be six urethane bonds and seven isophthalate groups in the backbone. After using normalization of the spectroscopic peaks with fringe spacing (coating thickness), half of the urethane bonds and 22% of the isophthalate bonds seem to be broken by the end of the exposure (Figure 2). The chain scission that occurs can account for some of the oxidation of the polymer. Coating thickness does not decrease as fast as chain scission, indicating that most network chains remain attached after a scission or that fragments are too large or entangled to be removed from the coating. At least two bonds need to be broken to remove a segment of coating from a continuous network.

With the increased oxidation and the large increase in the OH/NH area, one would expect an increase in hydrogen bonding. There have been many studies looking at the extent of hydrogen bonding in polyether-urethane elastomers. The simplest measurement is a ratio of hydrogen-bonded carbonyl at 1690 [cm.sup.-1] to free carbonyl at 1725 [cm.sup.-1]. (27,28) In Figure 6, the peak intensities in the FTIR spectra change approximately in step, indicating that, by this measure, there is little variation in the extent of hydrogen bonding of the carbonyl groups. In fact, the peaks broaden more on the free carbonyl peak side in this region, so there is no evidence of hydrogen bonding in the carbonyl vibrations. The NH region can also be used to determine the strength of hydrogen bonding. One problem with using the NH area here is that the polymer is being oxidized and the increasing hydroxyl and NH absorptions overlap considerably in the region from 3000 to 3500 [cm.sup.-1], making it difficult to discern clearly any specific changes (Figure 7). The NH-OH absorption region increases considerably; thus, hydrogen bonding may well be increasing as expected, but it is difficult to identify the degree to which it occurs.

Spectroscopy has shown how the chemical degradation of the polyester-urethane network progresses and indicates that the absorption of bonds associated with crosslinking is reduced as exposure increases. The results of the separate determination of crosslink density are given below.

Crosslink Density

At 130[degrees]C the tensile modulus used in measuring the crosslink density diminishes as the network is degraded. Figure 8 shows the stress-strain curves obtained at 130[degrees]C at a constant strain rate of [10.sup.-3] se[c.sup.-1]. This clearly shows how the modulus at this temperature is decreasing with increased weathering, indicating loss of network integrity. Table 1 shows the crosslink density values obtained by applying the high temperature modulus values to equation (1). In Table 1, [M.sub.c] is the average molecular weight between the crosslinks that was calculated using the following equation, (29)

[M.sub.c] = [rho]/[v.sub.e] (7)

where [v.sub.e] is the calculated crosslink density and [rho] is the coating density measured to be 1.2 g/[cm.sup.3]. As one might expect from the photodegradation indicated by the spectroscopy, the crosslink density diminishes with exposure.

Mechanical Properties at Room Temperature

Before any exposure, the polyurethane coating exhibited a low tensile modulus when tested at room temperature (Figure 9) and there was no breaking of the sample even at 100% strain. The deformation in unweathered samples fully recovered after removal from testing. Thermogravimetric analysis (TGA) of the unweathered coating samples showed a weight loss of ~6.5% within the temperature range of 45[degrees]C-120[degrees]C. This suggests that there was residual solvent in the unweathered coating films that plasticized them and caused them to be very flexible, prior to the first week of exposure. After accelerated exposure for one week, and thereafter, the temperature for weight loss became more centered around 100[degrees]C and was consistently ~2%, which is likely to be a small amount of water absorbed by the coating during exposure cycles.

During exposure, the tensile modulus, measured at room temperature, increased significantly (Table 2) over most of the period of exposure. The strain at break became very low, for example, just 3.5% by the end of the sixth week of exposure. After that, the samples could not be tested since they were too fragile. Reduction in elongation is a common observation with polymers subjected to weathering, (30) where brittle fracture initiates in flaws generated by photooxidation at the coating surface since the surface suffers from earlier degradation than the bulk. Nevertheless, if changes in modulus were caused solely by chemical degradation, one should expect modulus to diminish as crosslink density is reduced by degradation.

[FIGURE 9 OMITTED]

The modulus at the coating surface was determined by nanoindentation. Values obtained are higher than the bulk tensile modulus, particularly initially, but support the trend found for the bulk tensile modulus since values at the surface also increase with exposure. The strain rate in the nanoindentation testing was estimated according to the method proposed by Hochstetter et al. (31), and it was found to be 1-2 [sec.sup.-1], approximately three orders of magnitude higher than that used for tensile testing of the whole films, as described above. This will contribute to the relatively higher modulus values obtained with nanoindentation.

Modulus measured at the surface by nanoindentation is very useful complementary data to the conventional tensile testing, as it will be less dependent on the presence of residual solvent since the solvent would have escaped much more rapidly from the outer layer. Figure 10 shows the combined force-indentation curves for the incremental weathering made with 10 [micro]N threshold force. The surface becomes harder with each cycle of exposure showing lower indentation depth for a given maximum (threshold) force.

Eventually, continued reduction in crosslink density will lead to lower values of modulus and general failure, but for much of the exposure period examined here, the room temperature modulus increases. Table 2 summarizes the data obtained from both conventional tensile testing and nanoindentation. Additional mechanisms must be included to explain this increase in modulus while the crosslink density is diminishing. During exposure, the differences could be explained by preferential oxidation at the surface. Oxidation forms mainly hydroxyl groups, which may result in hydrogen bonding, thus increasing the elastic modulus. Other mechanisms considered here are that physical aging takes place or that the polymer network structure changes with a reduction in the number of elastically inactive chains. A prior section presented the spectroscopic evidence for increased hydrogen bonding. The other two possible mechanisms are discussed below.

[FIGURE 10 OMITTED]

[FIGURE 11 OMITTED]

[FIGURE 12 OMITTED]

Physical Aging

Physical aging has seldom been considered in studies of coating degradation. (14) Aging is often characterized by a relative increase in density (18) of the polymer of ~ [10.sup.-3], but that would be obscured here by the much larger decrease in volume, ~20%, due to degradation, as previously described.

Increases in elastic modulus of polymers are known to occur in polymers subjected to accelerated or natural weathering. (32) Physical aging (without chemical change) of polymers in some other studies has been found to increase the stiffness (both surface and bulk) with incremental aging time. (17,33-35) Another possible method of conceptualizing the modulus increase through aging is, if the difference between room temperature and [T.sub.g] increases, then the room temperature modulus increases, being effectively further below the glass transition.

Physical aging, and the progress towards thermal equilibrium, proceeds more slowly if the ambient temperature is lower, and thus further from the glass transition temperature. If ambient temperature exceeds [T.sub.g], then the polymer network achieves thermal equilibrium. This type of annealing is sometimes referred to as "rejuvenation." Aging starts again once the polymer is returned below its [T.sub.g]. Polymers are often chosen because their use temperature is below the glass transition temperature, so that they are hard and durable, less permeable, etc., but they will then be susceptible to aging. Thus, physical aging may occur in accelerated weathering, depending on the length and temperature of the cycles chosen for a particular investigation. The presence of aging was explored in the polyester-urethanes by using calorimetry, as described in the Experimental Procedures section.

Figure 11 shows the DSC curves for samples tested after incremental exposure periods. Note that the curves are shifted along the y axis for clarity and in each set the lower curve shows the behavior of the exposed sample and the upper curve for that of the same sample after annealing at 130[degrees]C for 15 min. The exothermic direction is up. Data clearly show the characteristic DSC overshoot (13,17,18) of the aged (exposed) samples, showing evidence of physical aging with increased exposure. Curves obtained after annealing do not show this overshoot. Figure 11 shows an apparent increase in [T.sub.g] with exposure, even after the residual solvent is lost after the first week of exposure, but the subsequent annealing process substantially reduces the differences, which must have been largely due to physical aging. A slight increase in [T.sub.g] remains after annealing, which shows that additional mechanisms contribute, but much less than physical aging over this exposure duration.

Figure 12 quantifies the aging by enthalpy recovery (18) (area between the curves) measured from Figure 11 for one set of exposed samples, showing the increasing trend as exposure increased. The glass transition temperature of these polymers changes somewhat with exposure, but is comparable to the temperature during the irradiation period of the exposure, 60[degrees]C. If the exposure temperature remains below the glass transition temperature then the polymer will age throughout, but it will age faster during the hot cycle. If [T.sub.g] is below the temperature of the hot period in the exposure cycle then the polymer will age during the cooler part of the cycle and be, at least partially, "rejuvenated" during the hotter part of the exposure cycle. It is very interesting that this polymer exhibits an accumulation of aging (enthalpy recovery) over the period tested here. One would anticipate that the temperature and periodicity in an exposure cycle would have a significant effect on how the properties of the polymer change. One can expect that the results of testing in accelerated weathering with short intense cycles of heat, irradiation, and moisture will produce very different physical aging than the longer, more varied cycles of natural weathering. This may contribute to the reasons for the imperfect correlation between accelerated and natural weathering.

Figure 13 shows the modulus from the low strain-rate tensile test done at room temperature for samples before and after annealing at 130[degrees]C for 15 min. This experiment also explores the influence of aging. As above, the initial change in the modulus, over the first week of exposure, may be due to the loss of residual solvent molecules from unexposed films. Thereafter, the increase in modulus with exposure requires a different explanation. There is much less variation in the modulus of the exposed materials after annealing at 130[degrees]C, which is consistent with physical aging contributing to the modulus. Again, although much of the variation in room temperature modulus can be erased by "rejuvenation," there is a smaller remnant increase in modulus that indicates that other processes increase the modulus somewhat.

In this work, both the DSC and DMTA gave values of [T.sub.g] that are comparable with the temperature of the hot period of the exposure cycle. Table 3 gives the values from the tan [delta] curves obtained by DMTA as the samples were subjected to increased exposure. The [T.sub.g] (tan [delta] maximum) of the unexposed sample was much lower but increased to higher values with exposure. DSC data obtained for [T.sub.g] also showed an increase (Figure 14), but the values correspond more to the onset temperature of the tan [delta] curve measured in the DMTA than to the peak temperature. There is a somewhat different trend in the DSC measurements of [T.sub.g]. The values for unannealed samples increased more slowly after exposure for the initial weeks. After the second week, the difference in [T.sub.g] (by DSC) between the annealed and unannealed was smaller and could be attributable to moisture from the exposure in the unannealed samples. It is common, and still the subject of research, that different values of [T.sub.g] are obtained in different thermal analysis techniques. (36) Two different methods are cited in the ASTM standards for assigning [T.sub.g] in DMTA analysis: one by using the onset of decrease in the E' curve, ASTM E 1640-99, and the other by the maximum in the E" curve, ASTM D 4065-01. Achorn et al. suggest a method to decrease the difference between the DSC and DMTA values by adopting a new DMTA [T.sub.g] definition. (37-38) A much deeper understanding of the relationship between different measures of the glass transition is necessary, but it is not within the scope of the research done here.

[FIGURE 13 OMITTED]

If one includes the finite width of the glass transition, one can appreciate that there may be components of the polymer network that do not completely relax and come to equilibrium during the hot cycle. Physical aging in this polyester-urethane will take place during the cool period of the exposure and "rejuvenation" may be incomplete during the hot period and aging, thus, may continue as exposure continues. Many polymers absorb a little moisture during exposure and may be plasticized so that their [T.sub.g], and thus their physical aging, will be affected depending on the exposure conditions, which may be another factor in the relationship between accelerated weathering and natural exposure.

In addition to aging, there was some spectroscopic evidence for formation of hydrogen bonding in the sample, especially in the NH-OH functionality, if not around the carbonyl groups. Secondary bond formation, like that of hydrogen bonding, would lead to a more dense and stiffer sample. Some other possibilities on how the network may change, to account for the small increase in modulus and [T.sub.g] that was not eliminated by annealing, are discussed in the following.

If the network degrades in a fashion that leaves a preponderance of stiffer chains and links, then this should affect [T.sub.g] and modulus, not only for the exposed samples but also for the annealed (after exposure) samples. There is only a small trend with exposure time in the values of [T.sub.g] from annealed samples, so the network chains and crosslinks that remain must be largely similar to the starting material. This would be easier to confirm spectroscopically in segmented polyether-urethanes with large composition domains, but is difficult to identify in the more random polyester-urethanes in this investigation.

During network formation of polyurethane coatings, a fraction of the polymer chains may form loops that do not contribute to the elastically active crosslinks, but act more like dangling chains. Dangling chains may also broaden the mechanical tan [delta] peak and may reduce the modulus because they may be more mobile themselves and thus permit neighboring chains to be more mobile. It can be assumed that the gel point of the coating occurs well before the coating is completely reacted, and there is some restriction in forming an idealized structure since these coatings were cured at ambient conditions. Monte Carlo simulations using Dryadd[R] software (Intelligensys Ltd.) calculated that there were approximately 10% dangling chains in the otherwise crosslinked network formed from the composition and conditions used here.

If dangling chains are preferentially removed, or shortened, since it only requires one chain scission, and the remaining network can relax into a conformation that accommodates the loss in material, one might also expect an increase in modulus due to a reduction in elastically inactive chains. This mechanism may contribute to the slight increase in room temperature modulus seen in the exposed and annealed samples. This process deserves more extensive investigation since the network structure during degradation will contain dangling chains generated by photodegradation of previously crosslinked chains and the competing processes will depend on the chemical structure of the polymer.

[FIGURE 14 OMITTED]

Crosslinked polymers have a broad glass transition compared to an equivalent thermoplastic. (39) If the thermal analysis data show a trend of the glass transition width diminishing with exposure, it may be consistent with a reduction in crosslinking. In addition, Dusek has shown that dangling chains increase the width of the glass transition in crosslinked networks, (40,41) so a narrowing of the transition may also indicate a reduction in the number of dangling chains during the degradation, as discussed briefly above. Unfortunately, it proved difficult to identify a trend in the width of the transition in either the DSC or DMTA results for the annealed materials.

CONCLUSIONS

The (NH, OH), CH, and carbonyl regions of infrared spectra show that the coating is being oxidized in the accelerated exposure, as expected, and that crosslinks are being destroyed. It has been confirmed by UV spectroscopy that the urethane linkage does undergo some photolysis, not only degradation through hydrolysis. Changes in thickness, measured by interferometry in both UV and infrared spectrometers, confirm that material is lost, approximately proportional to the length of exposure, and that the loss in thickness was reduced proportionally less than some of the spectroscopic indicators of degradation. Evidently, some of the degradation occurs to material that remains part of the coating until damaged further.

Tensile modulus measured above the glass transition confirmed that crosslink density diminished during exposure. The coatings also became brittle, which was consistent with surface degradation providing cracks and flaws that would initiate failure. However, the tensile modulus of the coatings measured at room temperature increased upon exposure, which is the opposite trend to the values measured above the glass transition. It was also found that [T.sub.g] measured by both DSC and DMTA tended to increase with exposure to the accelerated weathering.

"Enthalpy recovery" measurements using differential scanning calorimetry show evidence of increased polymer physical aging during the exposure duration used here. Annealing the exposed samples, above the glass transition, removed much of the variation in glass transition and room temperature modulus. This is consistent with the influence of physical aging. However, the annealing experiments showed that aging could not account for all of the increase in modulus and that some of the increase seen in glass transition temperature with exposure remained after annealing. Thus, additional explanations are required. Hydrogen bonding due to increased oxidation would contribute to an increase in the room temperature modulus with exposure. Infrared spectroscopy suggests that increased hydrogen bonding may occur due to hydroxyl or amines created in the degradation, but no evidence of hydrogen bonding could be seen in the infrared spectra of carbonyl groups.

Another contribution to the increased modulus at room temperature may be due to a relative reduction in the number of dangling polymer chains in the matrix and, thus, a relative increase in the concentration of elastically effective chains. A dangling chain may be separated from the network by one scission, but a chain that is connected at both ends requires at least two scissions (depending on the functionality of the crosslink junctions) before it can be removed by ablative processes or volatilization. This process requires further study.

In this study, polymer physical aging appears to contribute most to the increase in room temperature modulus as the crosslink density diminishes in this polyester-urethane. Hydrogen bonding and a relative increase in the number of elastically active chains may well contribute in a minor way over the exposure duration here, but are less easy to identify clearly.

ACKNOWLEDGMENTS

The authors would like to thank the U.S. Air Force Office of Scientific Research who funded the research under grant F49620-99-1-0283.

References

(1) Croll, S.G. and Skaja, A.D., "Quantitative Spectroscopy to Determine the Effects of Photogradation on a Model Polyester-Urethane Coating," J. COAT. TECHNOL., 75, No. 945, 85 (2003).

(2) Clark, F.R.S. and Moffatt, D.J., "The Elimination of Interference Fringes from Infrared Spectra," Appl. Spectrosc., 32(6), 547 (1978).

(3) Skaja, A.D. and Croll, S.G., Polym. Degrad. Stab., 79, 123 (2003).

(4) Sperling, L.H., Introduction to Physical Polymer Science, Wiley-Interscience, New York, 3rd Ed., Chapt. 9, 2001.

(5) Hill, L.W., "Structure/Property Relationships of Thermoset Coatings," J. COAT. TECHNOL., 64, No. 808, 31 (1992).

(6) Vandoolaeghe, W.L. and Muller-Nedebock, K.K., J. Phys. A: Math. Gen., 36, 8249 (2003).

(7) Edwards, S.F. and Vilgis, Th., Polymer, 27, p. 483 (1986).

(8) Smedberg A., Hjertberg, T., and Gustafsson, B., Polymer, 45, 4867 (2004).

(9) Fischer-Cripps, A.C., Nanoindentation, Springer, New York, p. 39, 2002.

(10) Oliver, W.C. and Pharr, G.M., J. Mater. Sci., 7(6), 1565 (1992).

(11) Pharr, G.M., Mater. Sci. Eng. A, 253, 151 (1998).

(12) Chowdhury, S., de Barra, E., and de Laugier, M.T., "Hardness Measurement of CVD Diamond Coatings on SiC Substrates," Surf. Coat. Technol., 193 (1-3), 200 (2005).

(13) Bicerano, J., Prediction of Polymer Properties, 2nd Ed., Marcel Dekker, New York, p. 287, 1996.

(14) Perera, D.Y., Prog. Org. Coat., 47, 61 (2003).

(15) Cook, D.C., Mehrabi, M., and Graham H.E., Polymer, 40, 1209 (1999).

(16) Tavares, A.C., Gulmine, J.V., Lepienski, C.M., and Akcelrud, L., Polym. Degrad. Stab., 81, 367 (2003).

(17) Gulmine, J.V., Janissek, P.R., Heise, H.M., and Akcelrud, L., Polym. Degrad. Stab., 79, 385 (2003).

(18) Hutchison J.M., Prog. Polym. Sci., Vol. 20, 703 (1995).

(19) Greidanus, P.J., Proc. XIX FATIPEC Congress, 1, 485 (1988).

(20) Van der Linde, R., Belder, E.G., and Perera, D.Y., Prog. Org. Coat., 40, 215 (2000).

(21) Perera, D.Y., Prog. Org. Coat., 44, 55 (2002).

(22) Gerlock, J.L., Mielewski, D.F., and Bauer, D.R., Polym. Degrad. Stab., 26, 241 (1989).

(23) McKeller, J.F. and Allen, N.S., Photochemistry of Man-Made Polymers, Applied Science Publishing Ltd., Barking, England, p. 5, 1979.

(24) Michaille, S., Arlaun, P., and Lemaire, J., Eur. Polym. J., 29(1), 35 (1993).

(25) Wilheim, C. and Gardette, J.L., Polymer, 39(24), 5973 (1998).

(26) Posada, F. and Gardette, J.L., Polym. Degrad. Stab., 70, 17 (2000).

(27) Lee, H.S., Wang, Y.K., and Hsu, S.L., Macromolecules, 20, 2089 (1987).

(28) Huang, S. and Lai, J., Eur. Polym. J., 33(10), 1563 (1997).

(29) Hill, L.W., Prog. Org. Coat., 31, 235 (1997).

(30) Fechine, G.J.M., Rabello, M.S., Souto Maior, R.M., and Catalani, L.H., Polymer, 45, 2303 (2004).

(31) Hochstetter, G., Jimenez, A., Cano, J.P., and Felder, E., Tribology International, 36, 973 (2003).

(32) Wypych, G., Handbook of Material Weathering, 2nd Ed., ChemTec Publishing, Toronto, Canada, p. 252, 1995.

(33) Signor, A.W., VanLandingham, M.R., and Chin, J.W., Polym. Degrad. Stab., 79, 359 (2003).

(34) Soloukhin, V.A., Brokken-Zijp, J.C.M., VanAsselen, O.L.J., and de With, G., Macromolecules, 36, 7585 (2003).

(35) Ott, H.J., "Einfluss der Thermischen Vorgeschichte auf die Mechanischen Eigenschaften von Amorphen Thermoplasten," Colloid Polym. Sci., 258(9), 995 (1980).

(36) Seyler, R.J., "Opening Discussion," Assignment of the Glass Transition ASTM STP 1249, Seyler, R.J. (Ed.), American Society for Testing and Materials (ASTM) International, West Conshohocken, PA, pp. 88-107, 1994.

(37) Achorn, P.J. and Ferrillo, R.G., J. Appl. Polym. Sci., 54(13), 2033 (1994).

(38) Ferrillo, R.G. and Achorn, P.J., J. Appl. Polym. Sci., 64(1), 191 (1997).

(39) Chartoff, R.P., Weissman, P.T., and Sircar, A, "The Application of Dynamic Mechanical Methods to [T.sub.g] Determination in Polymers: An Overview," Assignment of the Glass Transition ASTM STP 1249, Seyler, R.J. (Ed.), American Society for Testing and Materials (ASTM) International, West Conshohocken, PA, pp. 88-107, 1994.

(40) Dusek, K., Smrckova, M.D., Fedderly, J.J., Lee, G.F., Lee, J.D., and Hartmann, B., Macromol. Chem. Phys., 203, 1936 (2002).

(41) Dusek, K. and Ilavsky, M., Prog. Colloid Polym. Sci., 80, 26 (1989).

Allen Skaja, Dilhan Fernando, and Stuart Croll -- North Dakota State University*

Presented at the 82nd Annual Meeting of the Federation of Societies for Coatings Technology, October 27-29, 2004, in Chicago, IL.

* 1735 NDSU Research Dr., Fargo, ND 58105.

Table 1 -- Tensile Elastic Modulus at 130[degrees]C, Crosslink Density,
and Molecular Weight Between Crosslinks

                                               Molecular
                                               Weight between
Exposure  Modulus           Crosslink Density  Crosslinks,
Period    at 130[degrees]C  x [10.sup.-6]      [M.sub.c]
(Week)    (MPa)             (mol/[cm.sup.3]    (g/mol)

0            10.3              1,020              1,180
1             9.7                975              1,230
2             6.1                680              1,780
4             4.7                490              2,440
6             3.4                350              3,450

Table 2 -- Mechanical Properties at Room Temperature: Nanoindentation
and Tensile Testing

Exposure  Tensile               Tensile   Modulus by
Period    Modulus               Strain @  Nanoindentation
(Week)    (GPa)                 Break, %  (GPA)

0         0.2 [+ or -] 0.06      >100     1.6 [+ or -] 0.27
1         0.97 [+ or -] 0.05       54.0   2.0 [+ or -] 0.15
2         1.5 [+ or -] 0.02         8.0   2.3 [+ or -] 0.35
4         2.3 [+ or -] 0.08         4.5   2.40 [+ or -] 0.3
6         2.2 [+ or -] 0.15         4.5   2.94 [+ or -] 0.50
8         --                    --        3.47 [+ or -] 0.9

Exposure  Max. Indented
Period    Depth (for given peak forces) (nm)
(Week)    ~10 [micro]N   ~16 [micro]N

0             84              112
1             75              104
2             74               91
4             68               88
6             65               84
8             60               75

Table 3 -- Glass Transition Temperature and tan [delta] Curve Half-Width
from DMTA

                                  Width at
Exposure                          Half the Max. Height
Period                            of Tan [delta] Curve
(Week)    [T.sub.g] ([degrees]C)  ([degrees]C)

0                 39.0                   17.4
1                 58.0                   13.5
2                 63.0                    9.1
4                 75.0                   10.3
6                 81.0                   14.3