Accelerated weathering parameters for some aromatic engineering thermoplastics

Abstract

Samples of polycarbonate (PC), poly(butylene terephthalate) (PBT), a PC/PBT blend, and poly(styrene-co-acrylonitrile) (SAN), all containing 3% TiO₂ (by mass), were exposed in the NIST (National Institutes of Standards and Technology) SPHERE (Simulated Photodegradation via High Energy Radiant Exposure) to determine the effects of UV intensity (irradiance), temperature, relative humidity (RH), and UV wavelength on yellowing and gloss loss. There was no effect of irradiance, such that the samples obeyed reciprocity and doubling the irradiance doubled the rate of degradation. The activation energy for yellowing was determined to be ≈ 20 kJ/mol for PC, PC/PBT, and SAN and ≈ 16 kJ/mol for PBT. The activation energy for gloss loss was determined to be 9–16 kJ/mol. Thus, a 10 °C increase in temperature results in a 20%–30% increase in degradation rate. There was no consistent effect of RH on PC or PC/PBT yellowing or gloss loss. SAN degraded rapidly under dry conditions but showed little effect for RH > 10%. PBT lost gloss more slowly under dry conditions but displayed no RH effect with yellowing. Shorter wavelength UV had a greater effect on PC/PBT compared to PC or PBT.

1. Introduction

Polymeric materials exposed outdoors undergo chemical and physical changes caused by sunlight, heat, moisture, and/or other environmental factors. It is common to test coatings and plastics at outdoor exposure sites, but it is also common to subject samples to laboratory accelerated weathering. Although this testing process has been used for about 100 years, there is still great frustration at the slow pace of outdoor exposure and the poor predictability of accelerated weathering protocols. A major source of frustration is with the equipment and methods design, which were not meant to accelerate degradation of the materials in common use today [1]. This became apparent to the transportation coating industry several decades ago. An industrial consortium investigated the response of transportation coatings to light sources and moisture, and developed a protocol appropriate for these materials exposed in commercial accelerated weathering chambers [2,3]. It is essential to determine the response of classes of materials to accelerating factors to design predictive accelerated test methods and to interpret the results. It is particularly important to understand effects of all experimental factors to maintain in-service chemical mechanisms.

Sunlight, particularly the ultraviolet (UV) component, is usually the primary factor in outdoor weathering. To achieve acceleration, laboratory UV sources often employ irradiance (intensity or flux) greater than the average outdoor irradiance. Roughly speaking, it can be “high noon” or brighter constantly in a test chamber. The rate of degradation for many materials increases proportionally with irradiance, a property called reciprocity, but some materials do not follow it [4]. Reciprocity must be demonstrated experimentally to properly interpret accelerated weathering results. A second property of light is the spectral power distribution (SPD) of the source, since polymers exhibit different sensitivities to UV wavelengths [5]. Many artificial light sources include UV with wavelengths <295 nm, which is not significantly present at ground level in nature. This problem was largely resolved with newer-generation xenon arc lamps and filters that better match the spectrum of sunlight in the UV [6,7], and knowledge of wavelength effects is no longer critical if state-of-the-art equipment is used. Mathematical models for outdoor weathering still require knowledge of the various wavelength bands and their effect on the degradation rate. The present work investigates reciprocity and wavelength effects for several aromatic engineering thermoplastics.

Temperature is another an important consideration because the temperature in accelerated weathering chambers is usually higher than the temperatures during outdoor exposure [8]. The effective temperature (T_eff) is an irradiance-weighted average temperature, which is about 30 °C for white or clear samples and about 45 °C for black samples at Florida and Arizona exposure sites. In common accelerated weathering tests, T_eff is about 45 °C and 65 °C, respectively. Determining the correction factor using the Arrhenius equation is achievable once the activation energy (E_a) of degradation is determined for the materials. Activation energies will be established in the present work.

Moisture is extremely important for the weathering of transportation coatings [2]. It was shown that the simulation of rain was essential for the prediction of surface appearance of aromatic polymers [9], but little is known about effect of humidity. Humidity effects are also investigated here.

Irradiation and temperature effects on the photodegradation of aromatic polymers were reported previously [10,11]. These experiments are difficult to perform in conventional accelerated weathering chambers and are subject to large errors. Fortunately, a device developed by the National Institutes of Standards and Technology (NIST) in Gaithersburg, Maryland is ideal for this sort of investigation but had not yet been used for these types of materials. The NIST SPHERE (Simulated Photodegradation via High Energy Radiant Exposure) [12,13] can expose many samples simultaneously to individual, carefully-controlled conditions. The polymers studied were bisphenol A polycarbonate (PC), poly(butylene terephthalate) (PBT), a 55:45 PC/PBT blend, and poly(styrene-co-acrylonitrile) (SAN), all pigmented with 3% (by mass) rutile TiO₂ and containing only small amounts of processing stabilizers. These were subjected to a range of irradiance, temperature, relative humidity (RH), and wavelengths in the NIST SPHERE to determine the effects of these factors on yellowing and gloss loss.

2. Experimental

2.1. Samples

Compositions of PC, PBT, a PC/PBT blend, and SAN were made using commercial resins manufactured by SABIC Innovative Plastics. The PC was fully endcapped, and the SAN contained 25% acrylonitrile by mass. The catalyst of the PBT-containing samples was quenched using phosphorous acid, and the PC-containing samples contained <0.2% of an organic phosphite processing stabilizer. Samples were compounded as shown in Table 1, extruded, and injection-molded as high gloss color chips (3.2 mm thick) that were cut down to octagons ≈ 19 mm across to fit the NIST sample holders.

Table 1.

Compositions of samples (percent by mass).

Component	sample designation
	PC	PBT	PC/PBT	SAN
bisphenol-A polycarbonate (PC)	97		53.3
poly(butylene terephthalate) (PBT)		97	43.7
poly(styrene-co-acrylonitrile) (SAN)				97
organic phosphite stabilizer	<0.2	<0.2	<0.2
phosphorus acid P(OH)3		<0.1	<0.1
rutile T1O2	3	3	3	3

2.2. NIST SPHERE exposures

The UV irradiation was performed using the NIST-developed SPHERE in Gaithersburg, Maryland [12–14]. The SPHERE comprises a 2 m diameter integrating sphere illuminated with 6 microwave-driven, electrodeless metal halide lamps similar to D-type UV curing lamps [Heraeus Nobelight]. A WG295 filter [SCHOTT glass] cuts off most of the UV with wavelengths < 295 nm to give a spectral power distribution (SPD) shown in Fig. 1. Most visible and IR radiation is removed by cold mirrors before entering the chamber. The UV filter has transmission similar to borosilicate glass, so the UV cutoff is similar to xenon arc using borosilicate inner and outer filters (boro/boro) as shown in Fig. 1b. Since the SPD is unlike natural sunlight, there is no a priori means to calculate an expected correlation factor between SPHERE exposure and outdoors.

Fig. 1.

a) Spectral power distribution (SPD) of the SPHERE compared to ASTM G177 [15] sunlight; b) SPHERE SPD compared to ASTM G177 sunlight and boro/boro xenon arc [16] on an expanded scale.

Light from the SPHERE is directed by non-focusing cone concentrators into environmental chambers that can precisely control the temperature and RH of a fixed wheel containing 17 samples each 19 mm in diameter as illustrated in Fig. 2. Each sample location can be fitted with individual neutral density or band pass filters to control the irradiance or wavelength of the light. Samples exposed under full spectrum, “100%” irradiance were covered with quartz disks. Irradiance was measured at each location using a spectroradiometer at the end of each experiment. Typical full irradiance through a quartz disk was ≈150W/m² (295 nm–400 nm). Measured irradiance through the neutral density filters is shown in Table S-1 of the Supporting Information. Exposure is reported as MJ/m² over the range of 295 nm–400 nm. The SPHERE does not operate with dark or water spray cycles.

Fig. 2.

NIST SPHERE design and sample wheel.

2.3. Experimental design

The experimental design was accomplished in six runs covering the conditions shown in Table 2. The conditions of 55 °C and 10% RH or 50% RH used the data from the reciprocity and wavelength effect runs. The irradiation values shown in Table 2 are nominal, and actual irradiance for each sample is shown in the Results section. All data not shown in the body of the report are summarized in the Supporting Information.

Table 2.

Experimental design.

Reciprocity, temperature effects
T °C	% RH	Irradiance (%)
35	10	100, 90, 80, 40
55	10	100, 90, 80, 40
70	10	100, 90, 80, 40
RH effects
T °C	% RH	Irradiance (%)
55	0	100, 40
55	10	100, 40*
55	50	100*
55	75	100, 40
Wavelength effects
T °C	%RH	filter maximum (nm)
55	50	full; 306, 326, 354

^*uses data from reciprocity or wavelength effect runs.

2.4. Measurements

Samples were removed periodically, gently washed with deionized water, and measured for 60° Gloss, CIE color (L* a* b*), and yellowness index using a portable spectrophotometer with integrated gloss measurement (spectro-guide sphere gloss meter, BYK-Gardner USA, Columbia, MD). Color shift is expressed as described in ASTM D2244 [17], $\Delta E=\sqrt{{\left(L^{*}-L_0^{*}\right)}^2+{\left(a^{*}-a_0^{*}\right)}^2+{\left(b^{*}-b_0^{*}\right)}^2}$.

All gloss, color, yellow index data were the average of four measurements. Error bars represent one standard deviation from four measurements on the same specimen, and the error bars are smaller than the size of the symbols. Note that measurement uncertainties for different specimens on the same exposure conditions are smaller than 2% according to previous experiments.

3. Results and discussion

The materials in this study all contained 3% by mass coated rutile TiO₂ pigment, which absorbs UV wavelengths <380 nm. Most of the photodegradation of aromatic polymers such as PC is within top 25 μm of the surface due to the strong UV absorption of the degradation products [18]. This loading of TiO₂ absorbs some of the UV in this zone and so reduces the degradation to some degree [19]. However, color and gloss measurements are difficult to interpret for unpigmented materials such as PBT and the PC/PBT blend that scatter light due to crystallites or phase separation. The TiO₂ pigment was incorporated in order to make direct comparisons among the polymers and because many outdoor applications utilize pigmented resins. Restricting degradation to the top 25 μm also minimizes possible limiting effects of diffusion of oxygen or water from the surface.

3.1. Irradiance effects: reciprocity

Samples were exposed to four levels of irradiance using neutral density filters during three runs at 35 °C, 55 °C, and 70 °C all at 10% RH. Typically, the transmitted irradiance for the quartz filter was ≈150W/m² (295 nm–400 nm), which is normalized to 1 in the Figures. Results are plotted as a function of the actual UV energy transmitted by the filters over the range 295 nm–400 nm with the relative irradiance (I_rel) shown in the legends. Representative results at 55 °C are shown in Fig. 3 and Fig. 4. Results at the other temperatures are shown in Figs. S-1 to S-4 of the Supporting Information. Almost all of the color shift for these samples was positive on the b* vector of color space toward the yellow. Data also were acquired as Yellowness Index (YI, ASTM E–313), but gave results identical to Delta E and are not discussed here.

Fig. 3.

Color shift data after exposure at 55 °C, 10% RH and four irradiance levels plotted as a function of radiant exposure. Note that graph b has a different vertical scale. Note that error bars represent one standard deviation of the 4 measurements, and the error bars are smaller than the size of the symbols. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Fig. 4.

Change in 60° gloss after exposure at 55 °C, 10% RH and four irradiance levels plotted as a function of radiant exposure. The errors are smaller than the size of the symbols.

The data for PC, PC/PBT, and SAN in panels a, c, and d, respectively, of the Figures show excellent superposition for both color shift and gloss loss and indicate that these polymers obey reciprocity through this irradiance range. In other words, the amount of degradation depends only on the radiant energy received and not the intensity (irradiance). Reciprocity under xenon arc exposure was shown previously for PC and the PC/PBT blend [10,20]. However, our previous data suggested non-reciprocity for SAN. That previous work used neutral density filters held ≈ 1 cm above the samples in a commercial xenon arc chamber, and light leakage around the edge of the filter could have skewed the results at low irradiance [10].

The color shift data for PBT in Fig. 3b show excellent superposition up to Delta E of ≈3.5 and then divergence with higher irradiance showing higher color shift. This is also evident for the 35 °C exposures shown in Fig. S-1b of the Supporting Information but is less apparent in the 70 °C data in Fig. S-3b. Recently, White et al. showed that poly(ethylene terephthalate) (PET) yellowing exhibits non-reciprocity at high irradiance, with higher irradiance leading to greater yellowing [21]. The authors attributed this to diffusion limited oxidation at high irradiance, allowing accumulation of yellow products generated by more purely photochemical reactions. This could also be the case here with PBT. The apparent reciprocity at 70 °C is consistent, because oxygen diffusion should increase greatly above PBT’s glass transition temperature (T_g) of ≈ 55 °C [22]. PBT gloss loss generally obeys reciprocity. The faster loss at low irradiance at 55 °C is an outlier because no evidence of differences was seen at 35 and 70 °C as shown in Figs. S-2 and S-4.

We conclude that reciprocity is obeyed reasonably well for all four of these polymers, at least in the irradiance range encompassed by this experiment (up to 150 W/m² 295 nm–400 nm). The relationship between irradiance in the SPHERE and outdoors or to xenon arc lamps is dependent on the wavelength range considered because the spectral power distributions are so different. In the range 295 nm–340 nm, which drives most polymer degradation, the SPHERE’s irradiance is 1.12 × the ASTM G177 solar UV spectrum (ASTM G177 uses air mass of 1.05, the sun directly overhead, 0.73 W m⁻² nm⁻¹ at 340 nm). This is comparable to xenon arc irradiance of 0.82 W m⁻² nm⁻¹ at 340 nm. In the UV range 295 nm–400 nm, the SPHERE has 2.71 × the ASTM G177 spectrum because of the inclusion of strong output 350 nm–390 nm from the lamps, which is much less damaging to most polymers.

3.2. Temperature effects: activation energy

The rates of property change from the reciprocity experiments at three temperatures can be used to determine the activation energies of photodegradation. Absolute rates of property change can be difficult to define for noisy, non-linear data typical of polymer degradation. However, relative rates often can be obtained using the shift factor method [23,24]. An example of the data treatment is shown for PC at relative irradiance = 1 in Fig. 5, and shifted data for the other resins and irradiances are shown in Figs S-5 to S-12 of the Supporting Information. The 70 °C data were selected as the reference, and shift factors were applied to the exposure values of the other data sets to superpose all the data onto a master curve by trial and error using an EXCEL™ spreadsheet. If superposition can be achieved, the shift factors are the relative rates of property change with the reference data having a relative rate defined as 1. All the shift factors are shown in Table S-2 of the Supporting Information.

Fig. 5.

Raw data for PC yellowing (a) and gloss loss (c) and after applying shift factors to the 55 °C and 35 °C data to superpose onto the 70 °C data (b and d). Note that error bars represent one standard deviation of the 4 measurements, and the error bars are smaller than the size of the symbols. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

The Arrhenius equation (Eqn. (1)) gives good approximation rates as a function of temperature as long as the extrapolations to other temperatures are not large and the temperatures do not go through a phase transition. (PBT clearly violates the assumption with its T_g of ≈55 °C.) Taking the natural logarithm of both sides gives Eqn. (2) where k is the rate constant, E_a is the activation energy in J mol⁻¹, R is the gas constant 8.314 J mol⁻¹ K⁻¹, T is the temperature in Kelvin, and A is a factor relating to entropy that has no meaning in this analysis. Thus, plotting the natural log of the relative rates vs. 1/T and multiplying the slope by −R gives the activation energy. This is shown for PC color shift and gloss loss in Fig. 6 and for the other polymers in Fig. S-13 of the Supporting Information. The results are summarized in Table 3 with 95% confidence intervals calculated using standard errors of the slope generated by the LINEST function in EXCEL™.

Fig. 6.

Arrhenius plots for PC color shift and gloss loss. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Table 3.

Activation energies and 95% confidence limits (kJ/mol).

	PC	PC/PBT	PBT	SAN
Delta E	21 ± 2	21 ± 3	16 ± 4	19 ± 1
Delta Gloss	13 ± 1	9 ± 1	16 ± 3	16 ± 1

$$k=A\cdot \exp \left(\frac{-E_a}{RT}\right)$$ (1)

$$ln(k)=\frac{-E_a}{R}\left(\frac{1}{T}\right)+ln(A)$$ (2)

Higher temperatures gave higher terminal color shifts for PC (Fig. 5 and Fig. S-5) and SAN (Fig. S-8), which have been observed previously [11]. This could be due to temperature shifting the photo-thermal equilibrium of less and more highly colored isomers resulting from the oxidation of the aromatic rings [25]. PBT (Fig. S-6) shows the opposite effect in which the coolest temperature resulted in higher terminal yellowing, possibly due to oxygen diffusion effects as discussed above. Less difference was observed at lower irradiance. The effects seemed to cancel for the PC/PBT blend, and little consistent trend for terminal color was observed (Fig. S-7). The 35 °C exposures for the blend had slightly less color shift at all irradiances than the 55 °C or 70 °C exposures, which were about the same.

The results are generally consistent with those reported previously using xenon arc exposure [11]. The color shift E_a is 16 kJ/mol to 21 kJ/mol (4kcal/mol to 5kcal/mol), while the E_a for gloss loss appears slightly lower at 9 kJ/mol to 16 kJ/mol (2 kcal/mol to 4 kcal/mol). This gives these polymers a modest temperature dependence, amounting to roughly 25% to 30% rate change for a 10 °C temperature change near ambient temperatures. Terminal color can be higher at higher temperature for several of the polymers. It therefore seems prudent to perform accelerated weathering as close as possible to the effective temperature of the end use application to minimize the correction due to the activation energy and to avoid unrepresentative terminal color shifts.

3.3. Humidity dependence

Samples were exposed to full intensity irradiation (≈ 150 W/m² 295nm–400nm) at 55 °C and at 0%, 10%, 50%, and 75% RH in separate experiments. Results for Delta E are shown in Fig. 7. The only condition that showed any possible deviation is for SAN at 0% RH, which yellowed slightly faster than at higher humidity. The SAN 10% RH yellowing data falls between the 50% and 75% RH data, so there is no evident trend for the higher humidity exposures. The data for gloss loss shown in Fig. 8 are noisier, but no trend is noted for PC, PC/PBT, or SAN. (The 75% RH SAN sample inadvertently was exposed on the low gloss side, and the data have been renormalized to fit on the same axis. The difference with the higher gloss samples is unlikely to be significant.) It should be noted that there is no regular water spray during the SPHERE weathering cycle, so the surfaces are cleaned only during the deionized water rinse before data are taken. Therefore, gloss changes should be considered suggestive but not predictive of outdoor weathering results.

Fig. 7.

Color shift during exposure at 55 °C, full irradiance, and 0–75% relative humidity for PC, PBT, PC/PBT, and SAN. Note that error bars are smaller than the size of the symbols. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Fig. 8.

Change in 60° gloss during exposure at 55 °C, full irradiance, and 0%–75% relative humidity for PC, PBT, PC/PBT, and SAN. Note that error bars are smaller than the size of the symbols.

The PBT sample exhibited more erratic gloss loss. Gloss is lost because oxidized polymeric material is removed from the surface, exposing roughness due to the TiO₂ pigment. PC and styrenic polymers such as SAN undergo oxidation of the aromatic ring and formation of small molecules that easily sublime or are washed away [26]. The aromatic rings of polyesters such as PBT and PET are more resistant to oxidation because of the electron-withdrawing carbonyl groups. Hydrolysis reactions may be important in creating molecules small enough for the esters to erode. It appears that 50% RH and 75% RH give about the same results. Gloss loss under the very dry conditions of 0% and 10% RH was particularly erratic for PBT, indicating that some humidity is highly desirable to produce consistent results.

The materials were also exposed at 55 °C and several RH levels at ≈ 40% irradiance with results shown in Figs S-14 and S-15 of the Supporting Information. Color shift showed no humidity dependence for PC, PBT, and the PC/PBT blend. Gloss loss was more difficult to interpret. PC lost gloss more slowly at 75% RH, but a lag in the color shift data at ≈ 80 MJ/m² of exposure suggests that the specimen may have been incorrectly mounted for one check point that caused a lag in onset of gloss loss as well. The gloss loss rates are identical for all three RH levels after the onset of gloss, so it is likely that the differences are artifacts. The PC/PBT blend showed no consistent trend with humidity while PBT lost gloss much faster at 75% RH than at 0% or 10% RH and ultimately lost more gloss than at 75% RH and full irradiance. This again suggests that hydrolysis plays a role in polyester gloss loss.

SAN showed remarkable dependence on RH at lower irradiance with the results shown in Fig. 9. Dry conditions resulted in faster degradation for both color shift and gloss loss. Even 10% RH moved the rates much closer to the 75% RH results, suggesting that low levels of moisture were sufficient to slow the rate. There is a larger differentiation between the 0% and 75% RH at lower irradiance than at full irradiance, where it was difficult to see an effect at all. If the dry conditions promote dehydration or condensation reactions, lower irradiance would give more time for them to occur, so the effect would be more pronounced. At this point, we can offer no chemical explanation for a humidity effect on styrenic polymers, but it is apparent that some moderate amount of humidity (perhaps 30%–60%) should be included in predictive accelerated weathering protocols. This work does not address the effects of rain or water spray, which is critical for surface appearance of many polymers [9].

Fig. 9.

Color shift and gloss loss for SAN after exposure at 55 °C, ≈ 40% irradiance, and 0%, 10%, or 75% RH. Note that error bars are smaller than the size of the symbols. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

3.4. Wavelength effects

Band pass filters were used to isolate wavelength bands of the SPHERE UV source in the region of 306 nm, 326 nm, and 354 nm with full width at half height of 4 nm, 6 nm, and 27 nm, respectively. Transmission spectra of the filters, SPD of the SPHERE source, and the SPD of UV transmitted by the filters are shown in Fig. S-16 of the Supporting Information. Transmitted radiant energy was calculated by multiplying the transmission spectra of the filters by the measured SPD of the SPHERE at the sample locations and integrating under the curves.

The rate of color shift was very slow for all samples, and no meaningful change in gloss could be measured. The raw color shift data are shown in Fig. S-17 of the Supporting Information, and Fig. 10 shows the data with shift factors applied to superpose the data using 306 nm degradation as the reference. The data superimpose remarkably well for all polymers, but the experiment encompassed only the initial part of the overall degradation process. An initial color jump for PBT was slightly less for the 306 nm-filtered sample, but the subsequent curves were parallel. The shift factors are summarized in Table 4. These shift factors (relative rates) can be interpreted as the amount of degradation caused by 1 MJ/m² of radiant energy in the wavelength range relative to 1 MJ/m² at 306 nm. The relative rates are plotted against wavelength in Fig. 11, which shows that the sensitivity toward color shift increases approximately exponentially with shorter wavelength. The abrupt increase in sensitivity of PC and its blends to wavelengths <295 nm [10,27] is not seen because these wavelengths were not included in the filter experiment, although they are present in the quartz-filtered SPHERE SPD.

Fig. 10.

Color change for samples exposed behind band pass filters with shift factors applied to superpose the data on the 306 nm reference set. Note that error bars are smaller than the size of the symbols. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Table 4.

Rate contributions per unit energy for wavelength bands centered at 326 and 354 nm relative to contribution by the band at 306 nm.

material	306 nm	326 nm	354nm
PC	(1)	0.15	0.0085
PBT	(1)	0.09	0.0034
PC/PBT	(1)	0.05	0.0012
SAN	(1)	0.21	0.013

Fig. 11.

Plot showing log of the rate of color shift per unit energy vs. wavelength normalized to the rate at 306 nm. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

The different responses of the polymers to the three wavelength bands has important consequences to laboratory accelerated weathering. The data show that PBT and the PC/PBT blend are less sensitive to long wavelength UV than PC or SAN. Conversely, it could be said that PBT and the PC/PBT blend are more sensitive to short wavelength UV than PC or SAN. The relative degradation rates of different polymers under artificial UV will not be the same as under natural sunlight if the artificial source has a SPD that differs greatly from sunlight. That is, 1 MJ/m² measured broadly from 300 nm to 400 nm will accelerate the polymers differently if the SPD is concentrated on one end or the other of the UV range. A SPD like the SPHERE’s would accelerate SAN more than it accelerates PC/PBT relative to sunlight because it has more energy in the longer wavelength UV, where SAN is relatively more sensitive. (The excess UV < 300 nm probably has the opposite effect, but that was not determined in this experiment.) In any case, wavelength effects must be understood completely to evaluate dissimilar materials when using sources with SPDs differing from sunlight. This is possible but would require more spectral regions than covered in the present experiment and for the experiments to be carried out to higher degrees of degradation. For this reason, routine evaluation or service life prediction of dissimilar materials is best done using sources with SPDs close to sunlight, such as properly filtered xenon arc [2,3,7]. Any source can be used for research purposes if the wavelength effects are understood.

4. Conclusions

The purpose of this study was to determine the factors that are important for accelerated weathering of several aromatic polymers and to recommend material-appropriate exposure conditions. The data could also be used to model the degradation as a function of exposure conditions, but this was not addressed in the present work. Application of these findings to interpreting xenon arc accelerated weathering data is described in a forthcoming book of conference proceedings [28].

The aromatic thermoplastics PC, PBT, a PC/PBT blend, and SAN exhibit reciprocity for both color shift and gloss loss up to the maximum irradiance level of the NIST SPHERE, approximately 150 W/m² over the range 295 nm–400 nm. PBT had slightly higher terminal yellowing at high irradiance, consistent with the literature for PET [19]. The data show that the irradiance can be increased during accelerated weathering with the degradation increasing proportionally for these polymers. Most current weathering standards call for relatively low irradiance, but the higher irradiance of 0.8Wm² nm−¹ at 340 nm used in the most recent standards [3] seems appropriate for these materials.

Exposures at 35 °C, 55 °C, and 75 °C at four irradiance levels gave activation energies for color shift in the range of 16 kJ/mol to 21 kJ/mol (4 kcal/mol to 5 kcal/mol) for these polymers. The data suggest that these materials can be exposed at the same modestly elevated temperature and be accelerated nearly equally from the temperature effects. Acceleration of 25%–30% per 10 °C temperature increase is expected. The activation energy for gloss loss appeared to be 9 kJ/mol to 16 kJ/mol (2 kcal/mol to 4 kcal/mol). Gloss loss may not be accelerated to quite the same degree as color shift if laboratory test conditions are at higher temperature (12%–25% acceleration per 10 °C). Plateau color shifts after the onset of erosion were higher at higher temperature for PC and SAN. The data suggest accelerated weathering temperatures should be set as close as possible to actual use conditions to reduce errors due to uncertainties and differences in activation energies among samples and failure mechanisms. Effective temperatures [8] rather than worst case temperatures should be determined for applications. For simulating unbacked outdoor exposures, white samples at ≈ 40 °C and black samples at ≈ 55 °C would seem reasonable temperatures, if conditions can be maintained in the weathering equipment at high irradiance, resulting in a 12%–30% acceleration due to thermal effects.

Humidity effects appear to be minor for these polymers in the range of 10%–75% RH, although dry conditions appear to affect the degradation of SAN and PBT. Humidity is controlled in most accelerated weathering equipment and use of moderate RH in the range of 30%–60% seems reasonable for these materials. A critical moisture parameter not addressed in the present work is simulation of rain. Rain and perhaps wind are very important to the surface characteristics of eroding materials such as aromatic thermoplastics [9], and these factors should be addressed in future work.

Wavelength effects are essential to understand if the SPD of the accelerated weathering source differs from the SPD in actual use. The present work shows that the rates of color shift increase exponentially as wavelength decreases over the range of 306 nm–355 nm, but the function is different for each of the four materials studied. The well-known change of rate due to change of mechanism for PC photodegradation [10,24,25] was not studied in the present work, but would be needed for a model of PC and PC/PBT blend degradation using full-spectrum SPHERE data. Protocols that use xenon arc lamps with state-of-the-art filters [3,6] eliminate accounting for differences in SPD when relating accelerated and outdoor data. Protocols using other lamps or filters cannot be expected to accelerate all materials equally, and data would need to be corrected for wavelength effects that may be different for each material.

These conclusions apply only to the tested materials and do not necessarily apply to other materials. Polyolefins might exhibit very different results, for example. Currently, acceleration parameters are being determined for two transparent PC copolymers, a white PC-polyarylate block copolymer, and a white acrylonitrile-co-butadiene-co-styrene (ABS) polymer. Preliminary results show significantly different acceleration parameters for the latter two polymers, and full results will be reported in due course.