Effects of irradiation conditions on the weathering of engineering thermoplastics

Effects of irradiation conditions on the weathering of engineering thermoplastics

Polymer Degradation and Stability 93 (2008) 1597–1606 Contents lists available at ScienceDirect Polymer Degradation and Stability journal homepage: ...

710KB Sizes 15 Downloads 75 Views

Polymer Degradation and Stability 93 (2008) 1597–1606

Contents lists available at ScienceDirect

Polymer Degradation and Stability journal homepage: www.elsevier.com/locate/polydegstab

Effects of irradiation conditions on the weathering of engineering thermoplastics J.E. Pickett*, D.A. Gibson, M.M. Gardner GE Global Research, 1 Research Circle, Niskayuna, NY 12309, USA

a r t i c l e i n f o

a b s t r a c t

Article history: Received 3 January 2008 Received in revised form 18 February 2008 Accepted 28 February 2008 Available online 6 March 2008

The nature of the light source and its intensity can be important variables in accelerated weathering of aromatic engineering thermoplastics. Activation spectra show that BPA polycarbonate (PC) and its blends are very sensitive to UV with wavelengths <300 nm, as is well-known in the literature, but other resins gave unremarkable results. Xenon arc weathering experiments performed under identical conditions, but with different filter combinations did not show consistent rate enhancements. For 48 samples across a range of aromatic engineering thermoplastics, conditions using borosilicate inner and outer filters were 1.7 harsher than the CIRA/soda lime filter combination. However, the range was 1.0–2.5 and the standard deviation was approximately 0.35 making the correlation 1.7  0.7 at 95% confidence level for any given sample. The quartz/borosilicate combination used in SAE J1960 was 2.3 harsher than CIRA/ soda lime conditions, but the standard deviation was 1.1 making the correlation 2.3  2.2 at the 95% confidence level for any given sample. The effects of irradiance level and the dark cycle were determined in order to establish the legitimacy of accelerated testing methods. Linear increases in degradation with increased irradiance were observed for PC, poly(butylene terephthalate), and blends of PC with other polymers. Some non-linearity was found for styrene acrylonitrile copolymer (SAN), and extreme nonlinearity was found for ABS. No effect was found from a light/dark cycle other than the rate reductions expected from the lower dose rate. Thus, for accelerated weathering of engineering thermoplastics, the best possible match for sunlight is required, but increasing the intensity and decreasing or eliminating the dark period are permissible for most aromatic thermoplastic resins. Ó 2008 Elsevier Ltd. All rights reserved.

Keywords: Accelerated weathering Thermoplastics UV irradiance Polycarbonate Activation spectrum Reciprocity

1. Introduction Radiation in the ultraviolet (UV) and visible is the primary factor in outdoor polymer weathering. Effects from heat, moisture, pollutants, mechanical stresses, and biological attack can come into play, but usually the weathering process begins with a photochemical event. All laboratory accelerated weathering techniques employ artificial light sources of some sort that imperfectly duplicate the spectral power distribution of natural sunlight. The light intensity often is increased and the dark periods are decreased to gain acceleration over natural exposure. However, the legitimacy of these methods has not been established for most materials. We have been examining the effects of the various environmental factors on the weathering of engineering thermoplastics for the past several years and have reported a summary of our results [1].

Accelerated weathering testing is done for a variety of reasons, but most tries to predict what will happen under some sort of natural use conditions. Often, the benchmark is outdoor exposure near Miami, Florida. Industry has been using Miami sites to test paint and other materials since the 1930s, so this area has become a de facto standard. For this reason, we use Florida sunlight as the standard for comparison to artificial sources. The spectral power distributions of the common artificial UV sources are shown in Figs. 1–4 along with a reasonable representation of Florida sunlight [2]. Xenon arcs can be filtered using a variety of materials; common filters are quartz, type S borosilicate, soda lime glass, and CIRA (quartz with an IRreflecting coating). Xenon lamp power distributions can be further modified using additional filters in a lantern surrounding the lamp.

Abbreviations: PC, BPA polycarbonate; PBT, poly(butylene terephthalate); SAN, styrene acrylonitrile copolymer; ABS, acrylonitrile/butadiene/styrene copolymer; ASA, acrylonitrile/styrene/butyl acrylate copolymer. Analogous to ABS; IM, acrylic impact modifier; HALS, hindered amine light stabilizer. Usually 1% SanduvorÒ 3058; UVA, UV absorber. Usually 0.6% CyasorbÒ 5411. * Corresponding author. Tel.: þ1 518 387 6629; fax: þ1 518 387 7403. E-mail address: [email protected] (J.E. Pickett). 0141-3910/$ – see front matter Ó 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.polymdegradstab.2008.02.009

1598

J.E. Pickett et al. / Polymer Degradation and Stability 93 (2008) 1597–1606

1

2.5

0.8

Energy (W/m2/nm)

Energy (W/m2/nm)

2

1.5

1

Florida, normalized Q/B B/B CIRA/SL

0.5

0 250

350

550

450

Florida, normalized UVB-313 UVA-340

0.6

0.4

0.2

0 260

750

650

280

300

320

340

360

380

400

Wavelength (nm)

Wavelength (nm) Fig. 1. Spectral power distribution of xenon arc lamps equipped with various filter combinations: quartz inner/borosilicate outer (Q/B), borosilicate inner and outer (B/B), IR-reflecting quartz inner/soda lime outer (CIRA/SL). Data from Atlas [2].

Fig. 3. Spectral power distribution of fluorescent lamps used in QUV and UVCON testers. FS-40 lamps have output about 0.7 less intense than UVB-313 but otherwise similar distributions. Data from Atlas [2].

The importance of the spectral distribution has been widely discussed in the literature [3–6] but has been widely ignored in practice. In particular, the automotive industry in the United States often specifies using quartz/borosilicate filtered xenon arc conditions that employ unnaturally short wavelengths of UV in its SAE J1960 and J1885 standards (Fig. 2). The Japanese automakers continue to rely on carbon arc exposures (Fig. 4) that also have an unnatural wavelength spectral distribution with excess short wavelength UV. Fluorescent testing using FS-40 or UVB-313 lamps is still performed despite advice to the contrary [7], and while fluorescent UVA-340 lamps have no unnaturally short wavelength UV, they have very little output in the visible that might affect color shifts. A second critical variable is the intensity of the light. Some Japanese testers operate at very high irradiance with up to 15 the UV intensity of summer noon in Miami or Phoenix [8]. Operating continuously, these testers could apply 1 year’s worth of UV in 100 hdan acceleration of approximately 90 over natural exposure. Xenon arc testers operate at much lower intensity, usually about the equivalent of summer noon or less giving acceleration factors of approximately 3–10 Miami or Phoenix depending on the settings. Acceleration is achieved by using noon-like sunlight nearly continuously. None of these intensities are likely to cause

much trouble with non-linear photochemical effects that can occur with laser photolysis such as excited state saturation or multiple photon excitation except in rare cases of very long lived excited states. However, photodegradation is a very complex process with many diffusion and thermally controlled steps between the absorption of the photon and the formation of the final products. It is easy to imagine that these processes can become bottlenecks if the primary photoreactions are accelerated too much. During natural weathering, non-photochemical reactions in the degradation pathway may proceed during periods of lower light intensity or at night and keep pace with the photochemical steps. Thus, peak light intensity may not be as important as overall dose rate measured over a longer period of time. A critical consideration is whether or not a material obeys reciprocity, that is, whether a doubling of the light intensity results in a doubling of the degradation rate. There has been surprisingly little attention to the effects of light intensity or dose rate on polymer weathering. Martin et al. have published a comprehensive review of the literature and found very few examples of actual polymer degradation studies in which effects of light intensity were studied [9]. Some materials obeyed reciprocity and others did not. A third consideration is the inclusion of a dark period in the weathering protocol. No chemistry should occur in the dark that

1 1

0.8 0.6

Energy (W/m2/nm)

Energy (W/m2/nm)

0.8

Florida, normalized Q/B B/B CIRA/SL

0.4

0.2

Florida, normalized Sunshine Enclosed

0.6

0.4

0.2 0 260

280

300

320

340

360

380

400

Wavelength (nm) Fig. 2. UV spectral power distribution of xenon arc lamps equipped with various filter combinations: quartz inner/borosilicate outer (Q/B), borosilicate inner and outer (B/B), IR-reflecting quartz inner/soda lime outer (CIRA/SL). Data from Atlas [2].

0 260

280

300

320

340

360

380

400

Wavelength (nm) Fig. 4. Spectral power distribution of carbon arc lamps. Data from Q-Panel [11].

J.E. Pickett et al. / Polymer Degradation and Stability 93 (2008) 1597–1606

does not also occur in the light, so there should be no chemical reason to include a dark portion. One expects a dark period might allow diffusion or other thermally driven reactions to ‘‘catch up’’ with products formed by photochemical reactions so that the effect should be comparable to simply running continuously at some lower irradiance. Dark periods might be advised if temperature cycling or condensation is an important part of the test protocol. However, the pattern of light and dark periods has been little studied. The goal of our work has been to develop highly predictive accelerated weathering conditions for engineering thermoplastics [10]. We discuss the effects of spectral power distribution, lamp filter combination, intensity, and light/dark cycle in this paper. We will report on thermal and moisture effects in future papers. 2. Experimental 2.1. Exposure conditions Xenon arc exposures were carried out in Atlas xenon arc Weather-ometerÒ under the conditions shown in Table 1. Some exposures were performed in a Q-Panel QUVÒ using UVA-340 lamps at 50  C. The samples received 5 min of a misting water spray once per week.

Table 2 Acceleration factors relative to CIRA/soda lime xenon arc B/B xenon

1 4 5 6 7 8 9 10 11 12 13 14 15 16 18 19 20

2.2. Instrumentation

21

Color measurements were made on a Macbeth ColorEyeÒ 7000A spectrophotometer in reflectance mode using the small sample aperture. Color readings were with illuminant D65, 10 observer, L*, a*, b* scale. Yellowness Index (YI) is according to ASTM D1925, and total color shift DE ¼ (DL*2 þ Da*2 þ Db*2)½. Gloss readings were taken using a BYK-Gardner micro TRI gloss meter at 60 .

22

2.3. Activation spectra Activation spectra were obtained by the sharp cut-off method described by Searle [5] while exposed in an Atlas Ci35a xenon arc Weather-ometer under condition A of Table 1. Details of the procedure were previously reported [11]. 2.4. Sample arrays The compositions shown in Table 2 were extruded and injection molded into plaques approximately 3 mm thick. These were cut into rectangles 0.500  1.2500 (1.3  3.2 cm) and attached to 600  2.700 Table 1 Specifications for xenon arc exposure A

B

C

D

Instrument Inner filter Outer filter Wavelength cut-off (nm) Irradiance (W/m2 nm at 340 nm)

Ci35a Type S boro Type S boro w290

Ci4000 Type S boro Type S boro w290

Ci4000 Type S boro Type S boro w290

Ci4000 CIRA Soda lime w300

0.77

0.75

0.75

0.75

Air temp ( C) Black panel temp ( C) Relative humidity (%)

45 70

43 63

35 55

35 55

50

30

30

30

160 5 15

102

Continuous

Continuous

20 min Weekly

20 min Weekly

Light (min) Dark/no spray (min) Dark/misting spray (min) Light/misting spray (min) External vigorous spray

None

18 None

1599

23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48

PC 0.6% carbon black ASA/PC blend 1 black ASA/PC blend 2 black PC/PBT blend 1 black PC/PBT blend 2 black PC/PBT blend 3 black PC/ABS blend 1 black PC/ABS blend 2 black ABS 1 black ABS 2 black PC/PBT blend 4 black PC/PBT blend 5 black PC/BPA polyarylate blend gray 1 PC/BPA polyarylate blend gray 2 Resorcinol polyarylate–PC copolymer black Resorcinol polyarylate–PC copolymer green Resorcinol polyarylate–PC copolymer blue Resorcinol polyarylate–PC copolymer red Resorcinol polyarylate–PC copolymer 2% TiO2 PBT 3% TiO2 PBT 3% TiO2 þ UVA PBT 3% TiO2 þ UVA þ HALS PBT 0.6% carbon black PC 3% TiO2 PC 3% TiO2 þ UVA PC 3% TiO2 þ UVA þ HALS PC 0.6% carbon black PC/PBT blend 3% TiO2 PC/PBT 3% TiO2 þ HALS PC/PBT 0.6 carbon black PC/PBT/acrylic IM blend 3% TiO2 PC/PBT/IM 3% TiO2 þ UVA PC/PBT/IM 3% TiO2 þ UVA þ HALS PC/PBT/IM 0.6% carbon black SAN 3% TiO2 SAN 3% TiO2/HALS ABS 3% TiO2 ABS 3% TiO2/HALS PC/ABS blend 3% TiO2 PBT 3% TiO2 þ UVA þ HALS ASA 3% TiO2 ASA 3% TiO2 þ HALS PC/ASA/PMMA 3% TiO2 PC/ASA/PMMA 3% TiO2/HALS PC/0.6 carbon black Mean Standard deviation

Fluorescent UV-A (kJ/m2 per hour)

DGloss

DYI

DGloss

DYI

1.7 1.7 1.6 2.5 2.0 2.3 1.0 1.2 1.0 1.0 1.6 1.6 1.5

na na na na na na na na na na na na na

1.6 1.6 1.7 1.7 1.9 1.5 1.0 1.4 2.5 2.3 1.5 2.0

na na na na na na na na na na na na

1.4

na

1.3

na

1.0

na

1.0

na

1.0

na

1.2

na

1.0

na

1.2

na

1.0

na

1.3

1.1

1.0

2.0 2.0 2.0 2.0 2.0 2.0 2.0 1.8 2.0 2.0 2.0 2.0 2.0 2.0 1.8 1.5 1.5 1.5 1.5 1.6 2.0 1.5 1.8 1.8 2.0 2.0

1.9 1.7 2.0 na 2.0 2.2 2.3 na 1.8 1.8 na 1.7 1.9 2.0 na 1.5 1.5 1.4 1.5 1.7 2.0 1.6 2.1 2.0 2.5 na

2.5

5.5

2.0 2.0 2.1

na 2.5

2.0 2.0

na 5.0

1.6

na

1.7 0.37

1.8 0.32

2.5 2.3 2.1 1.9

5.0 5.0 4.0

1.6 1.6 1.5

na

1.7 0.47

4.5 1.10

Italics ¼ imperfect superposition; bold italics ¼ very poor superposition, na ¼ not applicable.

(13.2  6.9 cm) aluminum panels using silicone RTV adhesive with two columns of 12 samples on each array. Since the samples varied in thickness, they were arrayed face down within a frame on a glass plate, adhesive was liberally applied to the panel, and the panel was pressed onto the back of the samples and allowed to cure. Thus, samples 1–12 and 13–24 occupied two columns of one array and samples 25–36 and 37–48 occupied two columns of another, and the faces of the arrays were flat. Replicate samples were located strategically on the arrays. The arrays were mounted backwards in standard Atlas sample holders, held in place with a spring clip, and attached to the inside of the sample rack of the Weather-ometer.

1600

J.E. Pickett et al. / Polymer Degradation and Stability 93 (2008) 1597–1606

The sample size is just large enough to take color readings using the Macbeth spectrometer’s small aperture and to take 60 gloss readings using the BykGarder micro TRI gloss meter.

All on

2.5. Irradiance study Samples made from formulations 27, 23, 31, 38, 40, and 42 of Table 2 were cut into 0.500  0.7500 (1.3  1.9 cm) rectangles and mounted on two arrays as shown in Fig. 5 using silicone RTV adhesive. Each array was attached to a larger backing panel with a screw, and a neutral density filter was held approximately 1 cm over each sample region using bolts and nuts as shown in Fig. 5. The filters were 200 (5.1 cm) square quartz, and quartz neutral density filters with nominal absorbances of 0.3, 0.4, 0.5, 0.7, and 1.0. The measured transmission at 340 nm is shown in Table 6. The finished arrays were mounted in the Weather-ometer for exposure using the conditions of Table 1, method B.

6:6 1:1 Sample under disk

2.6. Light/dark cycle experiment

50% NDF

A small hole was drilled into a 600 (15.2 cm) diameter quartz disk, a drilled, circular 50% neutral density filter was placed at the center, and a mask cut from opaque, black polycarbonate film was attached to the front using silicone RTV adhesive as shown in Fig. 6. A battery-powered quartz clock motor was attached to a drilled steel panel, and the quartz disk was attached to the hour hand drive of the clock approximately 1 cm above the surface of the panel. Samples were cut into strips 2.500  0.500 (6.4  1.3 cm), marked into five 0.500  0.500 (1.3  1.3 cm) zones, and self-adhesive magnetic strips were attached to the backs. The samples were arrayed in a star pattern beneath the quartz disk as shown for one sample in Fig. 6. The zones received direct radiation, continuous radiation through quartz, 6 h light/6 h dark, 1 h light/1 h dark, or continuous radiation at 50% intensity. Unfortunately, the neutral density filter used was not stable upon prolonged exposure to xenon arc, and this portion of the data was spoiled. The finished assembly was mounted in a Weather-ometer for exposure using the conditions of Table 1, method B or in a QUV for exposure to UVA-340 lamps at 45  C and 0% relative humidity with no spray. 3. Results and discussion 3.1. Effects of spectral power distribution An activation spectrum is the most common representation of the effect of UV wavelength on polymer degradation. It is a qualitative or semi-quantitative measure of the response of a material to bands of radiation from a particular light source, usually xenon arc

Fig. 6. Masking pattern for rotating disk experiment.

or sunlight, and can determined by exposing samples behind a series of sharp cut-off filters [5]. The difference in rate (or amount) of degradation for samples behind different filters is due to the difference in the light passed by those filters. In this way, the effect of various virtual bands can be calculated. Trubiroha et al. have described various methods for determining spectral sensitivities of polymers as well as their limitations [12]. Aromatic polymers usually have large absorption maxima at wavelengths <300 nm that tail into the wavelength range >300 nm. This approximately exponential decrease in absorption multiplied by the approximately exponential increase in xenon arc light intensity with wavelength should result in the activation spectrum going to zero <300 nm and >400 nm with a maximum somewhere around 310 nm–340 nm if there is no change in quantum yield or degradation mechanism with wavelength. Because the activation spectrum is dependant on the light source it can look very different when determined under different conditions. For example, the activation spectra for yellowing of a PC/ABS blend determined under boro/boro and quartz/boro conditions little resemblance [11]. We have determined the activation spectra for yellowing of a variety of aromatic engineering polymers when exposed to boro/boro xenon arc as shown in Figs. 7–10. The abscissa values represent the center point of the difference in transmission of the filters [11]. The borosilicate filters transmit measurable radiation to at least 285 nm, and the small amount of energy at these shortest wavelengths is included in the band centered at 280 nm. 0.50

4

2

5

3

6

1

4

2

5

3

6

1

4

2

5

3

6

later rate

bolts, nuts and washers neutral density filter samples

backing

0.40

Rate contribution (b*/1000 kJ/nm)

1

0.30

0.20

0.10

0.00 280 Fig. 5. Left: panel bearing three sample arrays. Middle: array with three neutral density filters held above sample areas. Right: side view showing neutral density filters held between nuts and washers on screws.

initial rate

300

320

340

360

380

400

420

440

Wavelength (nm) Fig. 7. Activation spectrum for PC-containing 2% TiO2 exposed to boro/boro xenon arc.

J.E. Pickett et al. / Polymer Degradation and Stability 93 (2008) 1597–1606

0.50

0.30 PBT

0.40

PC/PBT

Rate contribution (b*/1000 kJ/nm)

Rate contribution (b*/1000 kJ/nm)

1601

0.20

0.10

SAN initial rates PC/ABS

0.30

0.20

0.10

0.00

0.00 280

300

320

340

360

380

400

420

280

440

300

320

340

360

380

400

420

440

Wavelength (nm)

Wavelength (nm) Fig. 8. Activation spectrum for PBT with 2% TiO2 and a 55/45 PC/PBT blend containing 3% TiO2 exposed to boro/boro xenon arc.

Fig. 10. Activation spectrum for SAN and a 67/33 PC/ABS blend containing 2% TiO2 exposed to boro/boro xenon arc.

Examination of these activation spectra reveals that polycarbonate and its blends are very sensitive to wavelengths shorter than 300 nm, a fact widely known in the literature (see Ref. [5]). The yellowing of PC has distinct initiation and growth aspects that have different activation spectra [13]. In addition, ABS yellowing shows sensitivity extending into the visible range >400 nm as has been reported in the literature [14]. From this information one expects that UV sources with UV <300 nm or no light >400 nm would present problems for accelerated weathering if one wishes to test and compare a variety of materials under the same test conditions.

rates under boro/boro conditions relative to CIRA/soda lime is shown in Table 2. On average, the boro/boro conditions are about 1.7 harsher than the same measured exposure under CIRA/soda lime conditions. However, the shift factors range from 1.0 to 2.5 depending on the nature of the resin and the color. The standard deviation for the rate increase is 0.37 for gloss loss (22% of the mean) and 0.32 for yellowing (18% of the mean). One notes a higher relative rate for polycarbonate-containing resins than for materials such as SAN and ABS. Polyarylates showed no particular sensitivity to the extra short

a

We subjected a series of 48 samples to xenon arc weathering using boro/boro or CIRA/soda lime filters, keeping all the other conditions constant (Table 1, C and D). The shorter wavelength UV radiation under boro/boro conditions generally caused faster weathering even when the same amount of exposure measured at 340 nm is applied. The rate enhancement can be determined by plotting both sets of data on the same graph and applying a shift factor to the abscissa (X axis) of the boro/boro data until the data superpose [15,16]. The shift factor is the relative rate. An example is shown in Fig. 11. In this case, a shift factor of 2.0 applied to the boro/ boro data causes the onset of gloss loss to superpose on the CIRA/ soda lime data and implies that the boro/boro rate is 2.0 higher than the CIRA/soda lime rate. However, the data do not superpose perfectly suggesting a change in mechanism. A summary of the

0

CIRA/SL boro/boro

-20

Delta Gloss

3.2. Comparison of light sources in testing

-40

-60

-80

-100 0

1000

2000

3000

4000

Exposure (kJ/m2/nm)

b

0

0.15

CIRA/SL Shift =1 boro/boro Shift = 2.0

Delta YI

Rate contribution (b*/1000 kJ/nm)

-20 ABS 0.10

-40

-60 0.05 -80

-100 0.00 280

0 300

320

340

360

380

400

420

440

1000

2000

3000

4000

Exposure x shift (kJ/m2/nm)

Wavelength (nm) Fig. 9. Activation spectrum for ABS containing 2% TiO2 exposed to boro/boro xenon arc.

Fig. 11. a. Raw data for gloss loss of white PC exposed under CIRA/soda lime and boro/ boro xenon arc conditions. b. Data from (a) after applying a shift factor to the boro/boro abscissa to best superpose the data. In this case, the superposition is not perfect.

1602

J.E. Pickett et al. / Polymer Degradation and Stability 93 (2008) 1597–1606

wavelength light. Interestingly, the black ABS samples (#11, 12) had a considerably lower shift factors for gloss loss than the white ABS samples (#40, 41). Some of these samples were also included in experiments run under fluorescent UV-A conditions (UVA-340 lamps) at approximately the same temperature. The number of kJ/m2 nm at 340 nm equal to 1 h of fluorescent UV-A exposure was determined by the superposition method, and those values are also shown in Table 2. The shapes of the curves were different in many cases leading to poor superposition of the data. The relative rates for gloss loss and yellowing are considerably different. This undoubtedly is due to the lack of visible light in the fluorescent tester that allows accumulation of the yellow products (see Fig. 3). This is illustrated for white polycarbonate in Fig. 12. Data for boro/boro xenon arc, fluorescent UV-A, and outdoor Miami, Florida exposures are superposed on CIRA/soda lime xenon arc data using shift factors of 2.0 for boro/boro xenon arc, 2.5 kJ/m2 per hour of fluorescent UV-A, and 2700 kJ/m2 per year in Florida. The Florida and CIRA/soda lime data can be made to superpose nearly perfectly. The boro/boro data also superpose well, but the shape of the curve is slightly different. The fluorescent UV-A data have an initial shape much like the B/B xenon but reach a plateau with higher yellowness. This is due to the lack of light with wavelengths >400 nm from the UVA-340 lamps. Such visible light causes bleaching of some of the yellow material formed by the shorter wavelength UV. For this reason, color shifts obtained from exposure to UVA-340 lamps cannot be trusted to be representative of outdoor weathering. Gloss loss data from UVA-340 exposure can be somewhat more predictive than yellowing data, but lack good superposition in many cases. The plateau yellowing values for all of the white samples are shown in Table 3. We have acquired data on a set of 120 samples representing many different aromatic engineering resins in black, white, and some in colors in validation experiments for experimental conditions [10], ASTM G 155 Cycle 1, and SAE J1960. These were run with filter combinations of CIRA/soda lime, boro/boro and quartz/boro, respectively. However, temperature and water spray conditions varied among the tests so this was not a clean comparison of just light sources. Summaries for color shift (DE ) and gloss loss broken out by resin class are shown in Tables 4 and 5. Both black and white are represented in each class. ‘‘Other’’ included ASA, ABS, a cycloaliphatic polyester, a resorcinol polyarylate, and PBT. There is some consistency among the results. In general, the ASTM G 155 Cycle 1 conditions (boro/boro) are little less than 2 faster than CIRA/soda lime when compared at equal exposures measured at 340 nm. The SAE J1960 conditions (quartz/boro) are about 2.5 faster than CIRA/soda lime and 1.4 faster than G 155.

Composition

UVA-340

Boro/boro ‘‘C’’

CIRA/SL ‘‘D’’

Miami, FL

PC SAN PBT PC/PBT ABS ABS/HALS PC/ABS

36 45 11 26 48 40 41

22 30 6 15 25 18 23

25 36 6 16 30 25 27

25 30 2 15 26 21 29

One major exception is the yellowing of polycarbonate. PC yellows dramatically faster under J1960 because of the abundance of short wavelength UV radiation. Note that all of the standard deviations shown in Tables 4 and 5 are large. There is a very large range of acceleration factors even among each class of resin due to differences in pigment and the specific resin included in ‘‘other’’. This greatly compromises the predictive value of comparing results among different tests. In other words, SAE J1960 may be 1.4 harsher than ASTM G 155 on average, but the 95% confidence error bars are 1.0 for color shift and 2.0 for gloss loss for any randomly chosen sample. The results from one test simply do not map onto another with any confidence. 3.3. Effects of irradiance or dose rate

20

Most accelerated weathering tests attempt to apply photons at a faster rate than natural exposure by increasing the intensity and/ or reducing the time of darkness. Data supplied by Atlas [2] indicate that the average irradiance in Miami is about 0.35 W/m2 nm at 340 nm during the time that the sun is shining while the maximum noon summer irradiance is about 0.74 W/m2 nm at 340 nm. This is approximately the practical range of most xenon arc exposures. A fundamental difference between the laboratory exposures and Florida is that the artificial tests are ‘‘high noon’’ all the time during the light cycle. The radiation occurs normal to the surface and usually does not change intensity. The time of darkness invariably is compressed or eliminated altogether in order to achieve additional acceleration. In general, the overall long-term dose rate is 3–8 the rate in Miami. An important question is whether the degradation rates increase in direct proportion to the dose rate. The rates of photochemical processes as a function of light intensity usually can be fit to the Schwarzchild law shown as Eq. (1) where k is the rate of reaction, A is a proportionality constant, I is intensity, and p is an experimentally derived number. The value of p can be obtained as the slope of the line plotted as log(k) vs. log(I ). When p < 1, the rate increases less than expected from increased light intensity. In the event that p ¼ 1, Eq. (1) reduces to a simple law of linear reciprocity, the Bunsen–Roscoe law (Eq. (2)) where B is a proportionality constant. The light intensities used in conventional accelerated weathering testers are well below the level that would cause a significant amount of two-photon processes in most cases. Non-linearities would be expected if diffusion of oxygen into the sample, additive migration, or other thermally controlled reactions become rate limiting.

10

Table 4 Acceleration factors of one test condition relative to others for color shift (DE )

40

CIRA/SL Shift =1 boro/boro Shift = 2.1 QUV-A Shift = 2.3 Florida (2700 kJ/year)

30

Delta YI

Table 3 Terminal DYI for samples containing 3% TiO2 under different exposure conditions

Test conditions

0 0

1000

2000

3000

4000

5000

Exposure x shift (kJ/m2/nm) Fig. 12. Data for weathering of white PC superposed onto CIRA/soda lime xenon arc data using the shift factors shown in the legend.

G155 over CIRA/SL J1960 over CIRA/SL J1960 over G 155

Resin class PC (n ¼ 10)

PC/PBT (n ¼ 34)

PC/other (n ¼ 25)

Other (n ¼ 51)

Overall (n ¼ 120)

2.4 (0.8) 4.3 (2.2) 1.8 (0.5)

1.9 (0.6) 2.1 (0.8) 1.2 (0.3)

1.8 (0.8) 3.2 (1.3) 1.6 (0.6)

1.4 (0.4) 1.7 (0.7) 1.3 (0.4)

1.7 (0.6) 2.3 (1.1) 1.4 (0.5)

Total of 120 samples; standard deviations in parentheses.

J.E. Pickett et al. / Polymer Degradation and Stability 93 (2008) 1597–1606

1603

Table 5 Acceleration factors of one test condition relative to others for gloss loss

Table 6 Shift factors for PC

Test conditions

Filter absorbance (nominal)

Filter transmission (measured)

Irradiance (W/m2 nm at 340 nm)

Shift factor (relative rate)

Quartz 0.3 0.4 0.5 0.7 1.0

0.93 0.65 0.42 0.42 0.30 0.11

0.70 0.49 0.32 0.32 0.23 0.08

(1) 0.67 0.45 0.45 0.33 0.16

Resin class PC/PBT (n ¼ 34)

PC/other (n ¼ 25)

Other (n ¼ 51)

Overall (n ¼ 120)

2.4 (0.8) 2.5 (0.7) 1.1 (0.2)

1.8 (0.6) 2.4 (1.0) 1.3 (0.4)

2.1 (0.8) 2.9 (1.0) 1.3 (0.2)

1.8 (0.6) 2.5 (0.9) 1.3 (0.4)

1.9 (0.7) 2.5 (0.9) 1.3 (1.0)

Total of 120 samples; standard deviations in parentheses.

k [ AI p

(1)

k [ BI

(2)

Little has been published in the literature on the reciprocity of engineering thermoplastics [9]. One study has been published by Jorgensen [17], who examined PVC and polycarbonate at intensities up to 50 natural sunlight using a solar concentrator. His results showed p values of 0.67 and 1.1 for PVC and UV-stabilized PC, respectively. Thus, polycarbonate apparently has excellent reciprocity over a broad range. We exposed white (2% coated, rutile TiO2) samples of PC, PBT, PC/PBT blend, SAN, ABS, and PC/ABS blend behind a series of neutral density filters under boro/boro xenon arc conditions. The relative rates of yellowing were determined by superposition of the curves onto the data for exposure behind quartz. This is illustrated in Fig. 13 for white PC. The X axis of the raw data in Fig. 13a was multiplied by shift factors until the best superposition was obtained (Fig. 13b, Table 6.) These shift factors were then plotted against the intensity calculated from measured light transmission of the filters multiplied by the unfiltered intensity to test the reciprocity law shown in Eq. (2) and plotted in Fig. 13c. The data were also plotted on a log–log plot to determine the p in Eq. (1) as shown in Fig. 13d.

a

20

b PC-Q PC-.3 PC-.4 PC-.5 PC-.7 PC-1.0

15

Delta YI

In this case, excellent linearity is observed up to the maximum irradiance of 0.75 W/m2 nm at 340 nm and there is no suggestion that use of the exponential form improves the fit to the data. The data for the other samples were treated in the same way with the results summarized in Figs. 14 and 15 and in Table 7. In each case the data superposed very well, although the SAN and ABS samples had more noise than the other materials. Good linearity with slopes near 1 is observed for the plots of relative rate of yellowing against irradiance for PC, PBT, PC/PBT, and PC/ABS. Any lack of strict reciprocity (Eq. (2)) is very subtle, and there is little to differentiate linear from exponential through the intensity range tested. In other words, p y 1 within experimental error. On the other hand, SAN appears to be non-linear with Eq. (1) giving a better fit to the data when p ¼ 0.63. ABS yellowing (Fig. 14) is clearly non-linear and gives an excellent fit with Eq. (1) using p ¼ 0.34. Hardcastle has reported a p value of 0.65 for polystyrene as measured using focused sunlight [18], and recent data presented by Scott showed a similar value for polystyrene in xenon arc exposure [19]. Clearly, styrenic polymers behave quite differently from polycarbonate and PBT. We did not study gloss loss in this experiment because the samples were not subject to water spray. Yellowing may be peculiar because of the complex photochemistry and dark reactions that

10

10 PC-Q PC-.3 PC-.4 PC-.5 PC-.7 PC-1.0

8

Delta YI

G155 over CIRA/SL J1960 over CIRA/SL J1960 over G 155

PC (n ¼ 10)

6 4

5 2 0

0 0

1000

2000

0

3000

200

Relative Rate

PC linear

0.8 0.6 0.4

y = 1.36x + 0.03 R2 = 1.00

0.2 0.0 0

0.2

0.4

0.6

Intensity (W/m2 at 340 nm)

0.8

d

0.2

log (Relative Rate)

c 1.0

400

600

Exposure x shift (kJ/m2/nm)

Exposure (kJ/m2/nm at 340 nm)

0.0

PC log-log

-0.2 -0.4 y = 0.83x + 0.09 R2 = 0.99

-0.6 -0.8 -1.0 -1.2

-1

-0.8

-0.6

-0.4

-0.2

0

Log (Intensity) (W/m2 at 340 nm)

Fig. 13. a. Raw data for white PC exposed behind a series of neutral density filters. The nominal absorbances of the filters (0.3, 0.4, etc.) are shown in the legend; Q ¼ quartz. b. Data from (a) replotted using the shift factors shown in Table 6. c. Relative rates (shift factors) plotted as a linear function of transmitted light intensity. d. Relative rates (shift factors) and transmitted light intensity as a log–log plot.

1604

J.E. Pickett et al. / Polymer Degradation and Stability 93 (2008) 1597–1606 Table 7 Relative rates of yellowing for polymers behind neutral density filters, calculated values for p from Eq. (1) and B from Eq. (2), and their R2 values

1

Absorbance Transmission Irradiance Relative rates (nominal) (measured) (W/m2 nm PC PBT PC/PBT SAN at 340 nm)

Relative Rate

0.8

0.6

Quartz 0.3 0.4 0.5 0.7 1.0

PC PBT 0.4

PC/PBT SAN

0.2

ABS PC/ABS

0 0

0.2

0.4

0.6

0.8

Irradiance (W/m2 at 340 nm) Fig. 14. Summary of relative yellowing rates as a linear function of light intensity. Slopes ¼ B in Table 7.

form and destroy yellow products. Loss of gloss and mechanical properties could very well follow different kinetics. However, we can conclude that most of these resins show good reciprocity, and that the light intensity can be increased to at least 0.75 W/m2 nm at 340 nm without loss of predictive value, except for the yellowing of ABS resin and perhaps SAN.

3.4. Light/dark cycles The final variable having to do with light is the dark cycle. Many test protocols call for the inclusion of a dark cycle. For example, we have included it in some coating protocols in order to put the coating through the stress of a thermal cycle during cool-down. However, there seems little reason to include a dark period on photochemical grounds. The only reason would be if there were reaction steps that occur thermally and are inhibited by light, but this seems unlikely. The effect of a dark cycle should be the same as continuous irradiance at a lower light intensity. We tested this hypothesis using the apparatus described in the Section 2 and preformed a preliminary experiment irradiating the samples in a QUV using UVA-340 lamps at 45  C and 0% relative humidity. The samples were exposed behind a rotating quartz disk bearing a mask that periodically blocked the light. The light cycles were continuous, 6 h light/6 h dark, and 1 h light/1 h dark. A central region was covered with a 50% transmitting neutral density filter. The absolute yellowing rates are not meaningful under these conditions, but the central 50% neutral density filter was stable, so 0.2

log(Relative Rate)

0.70 0.49 0.32 0.32 0.23 0.08

p (Eq. (1)) R2 B (Eq. (2)) B (normalized)a R2

PC/ABS

1 0.67 0.45 0.45 0.33 0.16

1 0.58 0.40 0.40 0.35 0.20

1 0.60 0.41 0.44 0.30 0.14

1 0.70 0.50 0.54 0.45 0.24

1 0.83 0.70 0.75 0.70 0.46

1 0.63 0.43 0.47 0.35 0.20

0.85 1.00 1.36 0.95 1.00

0.70 0.96 1.27 0.89 0.98

0.90 0.99 1.37 0.96 0.99

0.64 0.99 1.20b 0.84b 1.00

0.35 0.99 0.80b 0.56b 0.97

0.73 0.98 1.28 0.90 0.99

No value for zero irradiance was used in the calculation of p or B. a With rate at 0.75 W/m2 irradiance set to 0.75. B should be 1.00 for perfect reciprocity. b Least squares fit does not pass through origin.

the relative rates, determined using the superposition method, could be compared. The results are shown in Table 8. In no case could the rate under light/dark cycles be considered significantly different from the rate under the 50% neutral density filter. In other words, cycling the light on and off gave the same rate as continuous, lower intensity light. In a second experiment, the assembly was exposed in the xenon arc Weather-ometer under conditions described in Table 1, method B. The neutral density filter was found to have bleached to greater transmission over the course of this experiment, so those data were not included in the analysis. Instead, the expected rates were calculated from Fig. 14 for an irradiance of 0.375 W/m2 at 340 nm. The results are shown in Table 9. The rates for PC, PBT, PC/PBT, and PC/ABS were 0.52–0.62 relative to the continuously exposed samples. The expected rate was 0.50 for these samples. The mask was about 5 mm above the sample surface, and undoubtedly there was some leakage of light, so the observed rates probably are not significantly different from the expected rate. There was no significant difference between samples exposed for 6 h or 1 h light and dark cycles. Because of their non-linear responses to light intensity, the cycled SAN and ABS samples are expected to have rates 0.60 and 0.78 relative to the continuously exposed areas. The observed rates are approximately 0.6 and 0.8, respectively. We conclude that the rate of yellowing is not affected significantly by a dark cycle except as expected by the lower dose rate. 4. Conclusions

0

By far, the most important aspect of accelerated weathering is getting the spectral distribution of the light as close to natural as possible. The inclusion of UV wavelengths shorter than 300 nm can cause test results to become very unpredictable. While it is true that shorter wavelength UV can cause faster degradation, the

-0.2 -0.4

PC PBT

-0.6

Table 8 Rates of yellowing for samples exposed to 6 h light/6 h dark, 1 h light/1 h dark, or continuously under a 50% neutral density filter (NDF) relative to continuous UVA340 irradiation

PC/PBT SAN

-0.8

ABS PC/ABS

-1 -1.2

0.93 0.65 0.42 0.42 0.30 0.11

ABS

-1

-0.8

-0.6

-0.4

-0.2

0

log(Irradiance) (W/m2 at 340 nm) Fig. 15. Summary of relative yellowing rates as a log–log plot according to Eq. (1). Slopes ¼ p in Table 7.

Material

Continuous

6:6

1:1

50% NDF

PC PBT PC/PBT SAN ABS PC/ABS

(1) (1) (1) (1) (1) (1)

0.70 0.60 0.60 0.70 0.75 0.65

0.65 0.50 0.55 0.60 0.70 0.60

0.65 0.50 0.55 0.60 0.70 0.60

J.E. Pickett et al. / Polymer Degradation and Stability 93 (2008) 1597–1606 Table 9 Rates of yellowing for samples exposed to 6 h light/6 h dark or 1 h light/1 h dark relative to continuous xenon arc irradiation Material

Continuous

6:6

1:1

Expecteda

PC PBT PC/PBT SAN ABS ABS/HALS PC/ABS PC/ABS/HALS

(1) (1) (1) (1) (1) (1) (1) (1)

0.58 0.60 0.52 0.58 0.75 0.85 0.62 0.60

0.58 0.55 0.55 0.60 0.78 0.80 0.58 0.60

0.50 0.50 0.50 0.60 0.78 0.78 0.50 0.50

a

Calculated from Fig. 14.

acceleration factor has such wide variability across resin types, pigments, and properties that the acceleration is non-predictive. In some cases, such as for polycarbonate, UV < 300 nm changes the mechanism so that the results cannot be made to superpose on outdoor results or on more realistic artificial light sources. This is a serious problem because it means that the acceleration factor cannot be unambiguously defined, and without a reliable acceleration factor, lifetime prediction is not possible. Bauer has also noted the issue of unnaturally short wavelength UV affecting acceleration factors for acrylic and aromatic polyester coatings under boro/boro xenon arc conditions [20]. Visible light also is an important part of the mix. It is not simply possible to devise a general test protocol predictive of color shifts unless both the UV and visible components are present. This is because the visible light can do photochemistry on yellow products, and the absence of such light allows unnatural accumulation of these products. Fluorescent UV sources, mercury lamps, and other line sources cannot be used to predict either the rate or magnitude of color shifts with any degree of certainty, and we have seen that the acceleration factors for color shift and gloss loss can be very different. The light must be right. Use of xenon arc irradiances as high as 0.75 W/m2 nm at 340 nm (giving a dose rate of 64.8 kJ/m2/day at 340 nm, about 8 the average Miami daily dose) seems justified for most engineering thermoplastics with the exception of ABS and perhaps other styrenics. ABS yellowing shows a very non-linear response to dose rate making its acceleration problematic. However, ABS is well-known to undergo rapid yellowing outdoors, so it should not present many surprises. Inevitably, attempts will be made to expose at yet higher irradiances, and checks will have to be performed to ensure continued linear response to dose rate for the other materials. Reduction or elimination of the dark cycle seems similarly justified. Given a linear response to dose rate, it is difficult to propose a mechanism that would require a dark period, and the experimental data do not support the need for one for aromatic engineering thermoplastics. Dark periods might be justified to induce thermal cycling or to increase wet times for other materials. These conclusions probably extend to many coatings, but cannot be assumed valid for aliphatic polymers. Certainly, proper spectral distribution of the light is desirable, but a number of factors could cause non-linear response of degradation rate to irradiance. The degraded layer of aromatic polymers usually is quite thin, so oxygen diffusion is unlikely to become a limiting factor. This may also be the case for many real polyolefin formulations, but it certainly is not the case for unpigmented polyolefins containing no UV absorber; limiting oxygen diffusion during accelerated weathering is a well-documented phenomenon [21,22]. In addition, aromatic polymers usually have high glass transition temperatures, so additives and reaction products are frozen in place and do not diffuse. By contrast, additive migration occurs readily through the rubbery amorphous phase of polyolefins. Migration from the bulk can maintain some steady state concentration near the surface

1605

during outdoor weathering while additives can become depleted at the surface during accelerated weathering as the rate of photochemistry is increased faster than the rate of diffusion. For all these reasons, polyolefin weathering could present reciprocity problems. There is some documentation on intensity effects for polyolefins in the literature [23], but we are aware of no complete studies. In sum, the spectral distribution of the light source is critical while high irradiance and lack of dark periods are not for the weathering of most aromatic engineering thermoplastics. Xenon arc with CIRA inner and soda lime outer filters is the generally available artificial source with the best overall match to natural sunlight, although improved filters are under development at Atlas. We have shown that this combination can give excellent correlation to Florida across a wide variety of engineering thermoplastics [10], and will provide more details in a future report. Acknowledgements We thank Randy Carter for donating many samples used in this project and to SABIC Innovative Plastics for permission to publish this work. References [1] Pickett JE, Gardner MM. Effect of environmental variables on the weathering of engineering thermoplastics. Polym Prepr 2001;42(1):423–6. [2] Courtesy of Kurt Scott, Atlas Material Testing Technology, LLC, 4114 North Ravenswood Avenue, Chicago, IL 60613. [3] Torikai A. Wavelength sensitivity of photodegradation of polymers. In: Hamid SH, editor. Handbook of polymer degradation. 2nd ed. Marcel Dekker; 2000. [4] Searle ND. Effect of light source emission on durability testing. In: Ketola WD, Grossman D, editors. Accelerated and outdoor testing of organic materials. ASTM STP; 1994. p. 52–67. 1202. [5] Searle ND. Activation spectra of polymers and their application to stabilization and stability testing. In: Hamid SH, editor. Handbook of polymer degradation. 2nd ed. Marcel Dekker; 2000. [6] Andrady AL. Wavelength sensitivity in polymer photodegradation. Adv Polym Sci 1997;128:47–94. [7] Standard practice for operating fluorescent light apparatus for UV exposure of nonmetallic materials. ASTM G 154-97. [8] Available from: ; ; . [9] Martin JW, Chin JW, Nguyen T. Reciprocity law experiments in polymeric degradation: a critical review. Prog Org Coating 2003;47:292–311. [10] Pickett JE. Highly predictive accelerated weathering of engineering thermoplastics. In: Martin JW, Ryntz RA, Dickie RA, editors. Service life prediction: challenging the status quo. Federation of Societies for Coatings Technology; 2005. p. 93–106. [11] Pickett JE, Barren JP, Oliver RJ. Effect of accelerated exposure conditions on the photodegradation of BPA polycarbonate/ABS blends. Angew Makromol Chem 1997;247:1–18. [12] Trubiroha P, Geburtig A, Wachtendorf V. Determination of the spectral response of polymers. In: Martin JW, Ryntz RA, Dickie RA, editors. Service life prediction: challenging the status quo. Federation of Societies for Coatings Technology; 2005. p. 241–52. [13] Pickett JE. Kinetics of polycarbonate photoyellowing: an initiation/spreading model. Polym Mater Sci Eng 2000;83:141–2. [14] Searle ND, Maecker NL, Crewdson LFE. Wavelength sensitivity of acrylonitrile– butadiene–styrene. J Polym Sci Polym Chem 1987;27:1341. [15] Simms JA. Acceleration shift factor and its use in evaluating weathering data. J Coating Tech 1987;59(748):45. [16] Gillen KT, Clough RL. Time–temperature–dose rate superposition: a methodology for extrapolating accelerated radiation aging data to low dose rate conditions. Polym Degrad Stab 1989;24:137. [17] Jorgensen G, Bingham C, King D, Lewandowski A, Netter J, Terwilliger K, et al. Use of uniformly distributed concentrated sunlight for highly accelerated testing of coatings. In: Martin JW, Bauer DR, editors. Service life prediction, methodology and metrologies. ACS Symposium Series 802. American Chemical Society; 2002. p. 115. [18] Hardcastle HK. A new approach to characterizing reciprocity. In: Martin JW, Ryntz RA, Dickie RA, editors. Service life prediction: challenging the status quo. Federation of Societies for Coatings Technology; 2005. p. 217–26. [19] Scott KP. A new approach to characterizing weathering reciprocity in xenon arc weathering devices. Fourth International Symposium on Service Life Prediction, Key Largo (FL), Dec 3–8, 2006.

1606

J.E. Pickett et al. / Polymer Degradation and Stability 93 (2008) 1597–1606

[20] Bauer DR. Interpreting weathering acceleration factors for automotive coatings using exposure models. Polym Degrad Stab 2000;69:307–16. [21] Audouin L, Langlois V, Verdu J, de Bruijn JCM. Role of oxygen diffusion in polymer ageing: kinetic and mechanical aspects. J Mater Sci 1994; 29:569.

[22] Gardette JL. Heterogeneous photooxidation of solid polymers. Angew Makromol Chem 1995;232:85. [23] Philippart JL, Sinturel C, Arnaud R, Gardette JL. Influence of exposure parameters on the mechanism of photooxidation of polypropylene. Polym Degrad Stab 1999;64:213–25.