The POM (FM090^®) raw material (~3 mm elliptical shaped pellets) used in this research was supplied by Formosa Plastics Corporation. It has a density of 1.41 g cm^-3, a melting point of 165°C as measured by Formosa Plastics Corporation and is widely used in the automotive and consumer electronics industry due to its thermal stability and mechanical strength. American Chemical Society (ACS) grade chemicals and reagents were used in the experiments; while artificial seawater was prepared following the protocols set up by the Marine Biological Laboratory, Woods Hole, MA (http://comm.archive.mbl.edu/BiologicalBulletin/COMPENDIUM/CompTab3.html), with NaCl 423.00 g, KCl 9.00g, CaCl₂ 9.27g, MgCl₂ 22.94g, MgSO₄ 25.50g, * NaHCO₃ 2.14g (added last), deionized (DI) water from a Milli-Q system (Millipore, Billerica, MA) to 1L and 200 ppm NaN₃ (antibacterial agent) to prevent biological effects.

The sample code names, and weathering conditions are listed in Table 1. POM-Su simulated POM weathering in a sunlit terrestrial environment (Su); while POM-SWSu simulated POM weathering in marine environments (SW) under solar exposure and POM-SWSuN3 with azide (N3) to reduce microbiological effects. These were exposed on the rooftop of the Gushan precinct building of the Kaohsiung City Police Department, Taiwan (22°37’36.0”N 120°16’41.5”E). In this research, POM-Su and POM-SWSuN3 were exposed to sunshine with radiance 338–718 (MJ m^-2) for ~8255 h at 10.2–35.6°C, measured by the Central Weather Bureau, Taiwan.

POM-SWN3 simulated seawater-only weathering. The morphology and chemical properties of POM following photo- and thermal-oxidation were monitored with samples: POM-O50, POM-O100. According to the literature, the theoretical maximum possible ground surface temperature is between 90 and 100°C, and the maximum natural ground surface temperature, is 93.9°C. Therefore we examined thermal degradation of POM pellets in a channel drying oven, (OV-452) set at 50°C (O50) and 100°C (O100), for 42 and 36 months (Kubecka, 2001; Mildrexler et al., 2011), designated POM-O50 and POM-O100. POM-U and POM-SWUNa were exposed for 36 months at 25 ± 2°C to radiation below 350 nm under GL20SE ultraviolet B lamps (Sankyo Denki Co., Japan), dry and in artificial seawater. The weathering methods mentioned above were previously used in studies of the degradation of polyvinyl chloride and polypropylene (Tang et al., 2018; Tang et al., 2019).

The morphology characteristics of weathered POM samples were observed by environmental scanning electron microscopy (JEOL JSM 6380 SEM), designating three randomly chosen plastic pellets. Image J was used to analyze SEM images, in a supervised mode, to determine fragment area and measures of diameter. The pellets were weighed, and following exposure the remnant material was sieved using a range of sieves, but most frequently those of mesh size 50 (0.297 mm) and 80 (0.177 mm). The dry weight of pellets was measured with a Shimadzu electronic analytical balance ATX224.

Fourier transform infrared spectroscopy (FTIR) was used to determine the original and new functional groups on the surface of three randomly sampled virgin and weathered pellets. The FTIR spectra over the range from 4000 cm^-1 to 600 cm^-1 were obtained using a Nicolet 6700 FT-IR system integrated with a Smart Endurance single bounce diamond attenuated total reflectance (ATR) accessory (Nicolet Instrument Corp., Madison, WI) and Thermo Electron Corporation OMNIC software.

We often used medians and quartiles: Q1 and Q3 analysis were used to represent central tendency and dispersion as the data were not normally distributed. Correlation at small sample sizes with non-integer values and frequent ties adopted the Kendall rank correlation coefficient, with test statistic τ (Wessa.net). It was used in preference to Spearman’s test (test statistic ρ), because it met our needs and as Gilpin (1993) notes, Kendall τ “approaches a normal distribution more rapidly than ρ, as … sample size, increases; and τ is also more tractable mathematically, particularly when ties are present”. The Theil-Sen slope was used as a nonparametric representation of the linear slope, using the Single Case Research Calculator. The Kruskal-Wallis test, a non-parametric equivalent of ANOVA used Vassarstats Net, with test statistic H. Throughout we have accepted test statistics where they met a 95% confidence level, and p values are listed in the text.

Scanning electron micrographs at a magnification of 30-120x reveal a range of different degradation patterns (Figure 1 and electronic Supplementary Figures SI-1, SI-2). A new, spherical virgin pellet (V) is shown in Figure 1A, which is around 3 mm in diameter and weighs about 18 mg. The micrograph reveals characteristic striations that represent valleys of varying depth left during manufacture, typically some 600-900 µm in length and separated by 100-200 µm. The particle changes dramatically after sunlight irradiation. Over 1.5 years, particles become smaller (typically ~ 2.5 mm), losing material, and as shown in Figure 1B. The surface is indented with deep cracks that form knobby, yet angular features as “islands”, some 100 µm across. The size suggests that the cracks have formed at the scale of the original striations. Higher magnification (Figure 1C) shows finer level cracking with a spacing of about 15-30 µm. After three years the cracks penetrate even more deeply into the pellet (Figures 1D, E) and the size has decreased further (~1.27 mm) and by 3.5 years (Figure 1F), the pellets are so degraded they begin to fragment and separate into flat sided spherical sectors with dimensions up to several hundred microns across. Degradation under UV reveals fine cracks (Figure 1G) after 1.5 years that run across the original striations and are separated by about 200 µm; leading after further exposure to a crazed surface to the pellet (Figure 1H). In other samples larger cracks align with striations, but damage under UV is limited to the surface as noted by Wu et al. (2011), who observed that chain scission of POM under UV irradiation was constrained within 100 µm of the surface. Photo degradation resulted in the cracks forming on the surface gradually extending to the interior, and sunlight resulted in stronger fragmentation than UV-B, due to a control by oxygen diffusion or by penetration of UV light, and can be limited in the bulk (Masry et al., 2021). In the case where a plastic is exposed to the same external conditions, the surface degradation should increase with the UV or sunlight irradiation exposure time, and similar results were found in polyvinyl chloride and polypropylene (Tang et al., 2018; Tang et al., 2019; Masry et al., 2021).

Pellet exposure to seawater with sodium azide under sunlight (Figures 1I, J), also seawater and sunlight exposure (Figures 1K, L) and UV exposed seawater with sodium azide (Figures 1M, N) show little surface change, the seawater and perhaps biofilms, offering some degree of protection as has been noted with polyvinyl chloride and polypropylene (Tang et al., 2018; Tang et al., 2019; Masry et al., 2021). Furthermore, the slower degradation rate in water may also be due to air having higher oxygen content, UV transmittance and higher temperature as compared with water (Pimentel et al., 2005; Andrady, 2011; Cai et al., 2018; Tang et al., 2018; Biber et al., 2019; Ranjan and Goel, 2019; Tang et al., 2019; Masry et al., 2021). High temperature exposures (Figures 1O, P) show no change until the pellets melt and reveal an amorphous appearance.

The size change of pellets over time under various conditions is shown in Figure 2A. Sunlight irradiation causes an obvious reduction in the pellet diameter which amounts to about -570 µm years^-1, determined from the Theil-Sen slope. The Kendall rank coefficient (τ=-0.91; n=18) shows the decrease to be significant (p<.0001). UV irradiation and UV irradiation of the particles in seawater in the presence of azide also shows a decrease in pellet diameter -147 and -95 µm years^-1, again significant from Kendall’s test (τ=-0.7; p<.0002 and τ=-0.5; p=~.003). There is a slight decrease in diameter from the samples placed in an oven: 34 µm years^-1 (τ=-0.35; p=~.05). The other exposures, plotted as black dots, revealed no significant change in pellet diameter (p>.15). The degraded plastic remains white, although under solar irradiation in the seawater, where the azide biocide was absent.

The pellets clearly fragment especially under solar illumination, so it is hardly surprising that they lose mass as they age (Figure 2B) in addition to the shrinkage alluded to earlier. POM is well known for experiencing tens of percent mass loss during thermal ageing (Fayolle et al., 2008; Li et al., 2019) and when under stress (Parrington, 2002). As expected from the measurements of particle diameter the pellets under sunlight irradiation rapidly lose mass. Over the first year and a half the particles shed some 11 mg of the mass, i.e. almost 60%. So much is lost that it takes on an exponential form, and unsurprisingly Kendall’s test reveals high significance for the changes (τ=-0.95; p<.0001; n=18). After three years the particles are around a milligram, just 10% of their original mass. Decreasing mass is also evident from particles under UV irradiation which lose about 2.24 mg a^-1 (τ=-0.83; p<.0001), but the other degradation processes show little convincing change, although there may have been some hints of slow (0.57 mg a^-1), though statistically significant (τ=-0.81; p<.0001) mass loss from the sample exposed in seawater under sunlight irradiation in the presence of azide. This would have reduced any biofilm formation (samples POM_SwSuNa). However, many of the pellets are the same size and mass, though a few are highly degraded.

The fragmentation of a pellet under sunlight irradiation for more than three years yields pieces some hundred microns across (Figure 1F). These derive from cracks that penetrate radially inward, developing over the first 1.5 a to ~225 µm although nine measurements from the five exposure times show little subsequent change (Kruskal-Wallis H=3.5; dof=4; p~.48), with the median for the 45 measurements of crack penetration 224; Q_{1 =} 212; Q_{3 =} 263 µm. The fragments seem to take on characteristic dimensions, set perhaps by the original striations on the pellet, so sieves of 297 and 177 µm proved convenient to separate the degraded plastic. This size was noted by Conkle et al. (2018) who argue that an important portion of marine plastic <300 µm in diameter was often neglected in surveys. The mass fractions in the three size classes (pellets >297 µm, fragments 297-177 µm and fines <177 µm), over time are shown as a ternary plot in Figure 2C for sunlit conditions. The triangle represents only a small part of the complete ternary diagram, to make the growing amount of finer material more obvious. It is evident under sunlight illumination that the larger fragments grow at first, but there are increasing amounts of fines <177 µm. The other degradation modes reveal much less fragmentation to finer material, the UV exposure being the most pronounced. After three years ~0.13% are fragments and 0.91% are fines, with 99% of the mass remaining as the pellets.

As the pellets degrade, they develop features at the surface that resemble broccoli florets (Figure 3A) that extend to a depth of 100 µm or more. Some 80% of visible light across the wavelength range 400-2000 nm is transmitted through a 15 µm POM film (Whittet et al., 1976), so sunlight penetration into the particles is likely, allowing these to break away as fragments shown in Figure 3B. We determined the areas of these using Image J for the sunlight exposures, examining almost 350 fragments ranging from 0.5-325x10^-9 m^-2. This was converted to a dimension representative of an idealized square by taking the square root of the area. The Kruskal-Wallis test suggested little difference in the size of the fragments from various exposure times in sunlight (H=4.4; df=4; p~.35). In contrast to sunlight, UV radiation which interacts with the polymer is likely to be absorbed near the surface of the pellet (Wu et al., 2011), hence the concentration of degradation features at the surface (Figure 1H).

Photo degradation can lead to cracks on the surface which propagate deeper into the plastic and cause the plastic to split into two or more fragments. Fragments sizes varied overall from 30 nm to 1000 μm, and the particle size distribution showed an increase in particle concentration with decreasing particle size (Masry et al., 2021). The fragmentation process described here is more extensive than parallel studies of the weathering of polyvinylchloride or polypropylene pellets, and the fragmentation rate and generated particle sizes depends on the polymer type, shape, composition and external conditions (Tang et al., 2018; Tang et al., 2019; Masry et al., 2021). POM degradation would occur more rapidly in sunlit water to produce fragments where ~95% are likely to be <300 µm (Figure 3C). These smaller sizes are important as they are easily ingested by fish. Particles >100 µm are almost five times more likely to be ingested by fish than larger particles of >500 µm (Jovanović, 2017). The degraded POM fragments are also in the size encountered in natural turbidite systems, where 90% of the particles are <214 µm (Pohl et al., 2020), so POM fragments are likely to become associated with turbidity currents in sediments. On the other hand, as fragment size decreases, ingestion by a wide diversity of organisms is favored (Masry et al., 2021).

The infrared spectra (Figure 4A) show that the main changes under weathering relate to the OH bond at 3398 cm^-1. The changes in OH absorbance as a function of age is shown in Figure 4B. Under sunlight irradiation we see an increase in the OH absorption until the particle begins to fragment after 3.5 years. In the case of UV irradiation after a rapid increase in OH intensity in the first year it stabilizes at more modest values, so although OH seems to increase it appears to be lost as weathering proceeds to its final stages. These changes are mirrored in the physical alteration to the pellets and illustrated in the reduction of their diameter as shown in Figure 4C. Links between observable change and the spectra were noted in earlier work especially during the first year of PVC and PP exposure (Tang et al., 2018; Tang et al., 2019). As a result, the formation of oxidation products leads to modifications of the infrared spectrum of the polymer. Identification and quantification of the oxidation products are essential for understanding of the mechanism of degradation and to determine the weathering extent of plastic debris (Masry et al., 2021). The main photodegradation products of polyoxymethylene: H2CO, CO, HCOOH, O2, CH3OH, CH3OCHO, CH3OCH3OCH3 and C3H6O3 (trioxane) were identified (Cottin et al., 2000). These products may be released to the marine environment with POM weathering.

Solar and UVB degradation resulted in severe eroding behavior and reflected by cracking and the formation of hydroxyl functional groups on the surface of eroded polyoxymethylene (POM) pellets, an emerging plastic pollutant in runoff. These are new scientific findings, and the altered morphology and chemical properties could explain the affinity of plastic marine debris for trace metals, persistent organic pollutants, and microbes, as well as their tendency for biofouling.