Nanoscale Infrared Spectroscopy Reveals Nanoplastics at 5000 m Depth in the South Atlantic Ocean


 Millimeter- and micrometer-sized plastics are well-documented in aquatic ecosystems, but little is known about sub-micrometer particles because conventional analytical techniques lack sufficient spatial resolution or the spectroscopic means to unambiguously identify individual nanometer-sized plastic particles. We combined the spatial resolution of atomic force microscopy with chemical information from infrared spectroscopy to detect, identify, and count nanoplastics down to 20 nm in diameter in samples from different depths in the South Atlantic Ocean. We present evidence for the presence of polyethylene terephthalate (PET) nanoplastics in different states of degradation at 5000 m. Using lab-based ageing of PET, we demonstrate that nanoplastics can form even without light or interaction with the plastisphere, and that macroscopic PET items are a plausible source of PET nanoplastics in the ocean.

instead their movement follows a random walk 34 . This may lead to widespread transport of nanoplastics by ocean currents 35 , but experimental evidence is lacking.
Here we collected samples in the South Atlantic (-32.171° S, 6.287° E) in January 2020 (Fig. 1a), close to the South Atlantic Subtropical Gyre proposed to have high micro-and nanoplastic concentrations 36 . We analyzed samples from the surface (5 m), the pycnocline (60 m), and close to the ocean bottom (5170 m). Sampling, sample handling, and sample processing details with contamination prevention measures can be found in the Methods section. PiFM analysis was performed by spin coating a small volume (10 µL) of sample taken from the total sample volume of 1 L collected in glass bottles at each depth. Three technical replicates consisted of multiple scans of the wafer surface after spin coating. Assuming that the particles were equally distributed on the wafer surface after spin coating and knowing the total wafer surface area scanned during each measurement, we calculated a rough estimate of the particle concentration in the spin coated volume (see reference 25 and the Methods section for further information). No pre-processing of the seawater was performed to remove larger material. However, as PiFM provides spectroscopic information on the detected particles, it is possible to separate nanoplastics from other particles such as salt crystals 25 or non-plastic micro bers.
Figure 1b-f and Supplementary section S2 provide unambiguous images and chemical validation for the presence of PET nanoplastic particles in the South Atlantic. We only detected PET nanoparticles close to the ocean bottom and not at any other depth. We identi ed one Nylon and three cellulose micro bers at the surface and one Nylon micro ber, three polystyrene (PS), and four polyethylene (PE) nanoparticles at the pycnocline. From the sum of the 5179 m-replicates, we found 12 PET nanoparticles and one PET micro ber. We also identi ed two PS particles at this depth. Because our sample size was low, our concentration estimates have large standard deviations and are not statistically signi cant (Supplementary section S2.5). However, given that we found nanometer-sized objects in just a fraction of 250 µL sampled from a volume as large as the ocean, our ndings are noteworthy. All other samples (treated identically, including blanks) did not contain any PET nanoplastics, effectively ruling out contamination. We therefore chose PET nanoplastics which showed signi cant occurrence to focus on further.
Since we found PET nanoplastics exclusively close to the ocean oor, we hypothesize that sunken macroand micro-PET might act as a local source for nanoplastic formation. At these depths PET litter is not exposed to sunlight and thus, barring biotic degradation, hydrolysis must be the major form of degradation, i.e. bond breakage via water addition 27,[37][38][39][40] . To investigate this hypothesis, namely, that PET generates nanoplastics via hydrolysis in salt water even in the absence of light, we performed controlled laboratory degradation experiments (Fig. 2), by submerging pristine PET (granular microplastics of 2-4 mm) in a 35 g/L NaCl solution with a pH of ~7, i.e., less basic than seawater (see also the Methods section). The sample was kept in the dark at room temperature and water was sampled after 1, 2, 4, and 8 weeks. The results of IR imaging at the 1725 cm -1 band, which is the characteristic C=O stretching vibration for PET, show that PiFM can unambiguously detect, identify, and quantify PET nanoplastics at all stages of the aging experiment ( Fig. 2 and Supplementary section S3), con rming that PET microplastic indeed generates detectable amounts of PET nanoplastic particles even after 1 week (Supplementary Table 4). Signi cantly larger numbers of particles were detected per unit sample volume than in the seawater samples, however, as PiFM is not a high-throughput method and the number of detected PET nanoplastic particles was still low, the calculated particle concentrations remain estimates and trends in concentrations versus time suffer from large uncertainties. However, we managed to measure about 100 particles for each sample, which allowed us to establish particle size distributions (PSDs) (Fig. 2bc). The PSDs show a shift towards smaller particle sizes with time that could indicate that fragmentation of nanoplastics continues as time progresses. Larger PET particles were atter than smaller ones, thus deviated from a sphere in aspect ratio. This has potentially important implications for toxicology studies 25 (Fig. 2d) or reference materials that employ nanoscale polymer beads in laboratory studies 18,41 .
We characterized a smaller subset of particles of each age group in more detail to investigate potential chemical differences, recording the full IR point spectra to establish a spectroscopic ngerprint for each particle. The multiple-particle averaged IR spectra suggest a clear correlation of the ratio of the  Table 3). Based on the experiment (the microplastic remained submerged), one would expect a growing standard deviation for this ratio with time as the sample after 8 weeks should contain particles of 'age' 0 to 8 weeks if particle formation proceeds linearly. This was not the case and seems in agreement with particle concentration estimates (Supplementary Table 4) that do not suggest a simple linear rate of formation. However, as mentioned above, the number of particles measured for each age group was not statistically su cient to unambiguously con rm these trends.
From a chemical perspective, the IR spectra indicated partial degradation of functional groups in PET nanoplastics 42 (see also Supplementary sections S1.1 and S1.2). During the aging process both the C=O stretching peak ~1725 cm -1 and the wagging vibration modes of the ethylene glycol segment ~1360 cm -1 gradually decreased in intensity, while the C=O stretching peak of terephthalic acid ~1740 cm -1 increased in relative intensity ( Fig. 2a and Supplementary section S3.1). To con rm hydrolysis is the main degradation mechanism producing PET nanoplastics, we aged PET in a basic environment at pH=12 for 4 weeks (Supplementary section S3.2) to see if increased OH -42 concentrations enhanced degradation.
Our results con rmed this: the ratio of the 1740/1725 cm -1 bands increased signi cantly compared to the experiment at pH=7 (Fig. 2a). The particle concentration estimates were not fully conclusive: one sample at pH 12 (although the statistically more signi cant one) showed much higher particle numbers, while the other sample, where only a very small wafer surface area was measured, did not (Fig. 2c). We propose that the aging experiments performed at pH=7 may even underestimate the degradation rate in seawater (pH ~8, room temperature). Interestingly, the PSD after 4 weeks was very similar for both experiments, suggesting a more complex interplay between rate of formation and particle morphology.
With our experimental evidence for the formation of nanoplastics from microplastics in salt water in the dark, we compared the full PiFM IR point spectra of individual PET nanoplastic particles from the South Atlantic water samples (Fig. 3) with those obtained for lab-generated PET nanoplastics. The spectral differences we observed between ocean samples were very similar to the ones recorded for the labgenerated PET nanoplastics with different aging times ( Fig. 2a and Supplementary section S3.1). Based on the correlation, we established above between the increase of the band ratio 1740/1725 cm -1 with ageing time, we therefore suggest that the PET nanoplastics we identi ed in the ocean are in different stages of degradation. Assuming more or less constant conditions for hydrolysis of PET at the ocean bottom, we speculate that particle 2 in Fig. 3ab is 'younger' than particle 3 in Fig. 3cd based on their respective spectra (Fig. 3g).
However we interpret this observation with caution since PET degradation is a complex process 42 and in order to establish a more quantitative age classi cation of individual PET nanoplastic particles, we would need to assess a larger spectral range, and considering other factors such as molecular orientation effects (see, for example, Supplementary section S3.4 reporting experiments we performed with polarized light). Furthermore, as we found PET nanoparticles only in water samples collected close to the ocean bottom, future lab-based ageing experiments need to consider the lower temperature (< 4°C) and higher pressure experienced at these depths 43 . Nevertheless, the results we report here show that such age classi cation may be possible, introducing new ways for tracing the origin and fate of marine nanoplastics. As nanoplastics below a certain size behave distinctly, knowledge of their age alongside that of ocean currents transporting them enables both their point of origin, as well as potential sinks for PET nanoplastic particles, completing existing models for transport of macro-and microplastics 35 .

Conclusion
In conclusion, we show unambiguous evidence for the presence of PET nanoplastics in the South Atlantic Ocean at 5179 m depth. We did not detect any PET nanoplastics at the ocean surface or in the pycnocline (60 m) but did nd PS nanoplastics both in the pycnocline and close to the ocean bottom and PE nanoplastics only in the pycnocline layer. At the ocean surface, we only detected cellulose (e.g. cotton) and Nylon micro bers; the latter was also found in the pycnocline. The method we used to count nanoplastics can unambiguously identify and speciate nanoplastic particles down to ~20 nm in size but it is a low-throughput method and particle concentration estimates show large standard deviations. The comparison of the spectroscopic ngerprints of individual PET nanoplastic particles found close to the ocean oor with data obtained via controlled ageing of PET provided a PET age distribution and con rmed hydrolysis as a formation mechanism. Interestingly, we also found two PS particles at this depth. PS macro-and microplastics may sink to the ocean oor if their density is high enough, for example in the case of crystal-grade PS (~1.05 g/cm 3 ) 32 . Alternatively, lower density PS (< 1 g/cm 3 ) or foam expanded PS (~0.05 g/cm 3 ) will not sink 32 and in that case, transport of nanoplastics via ocean currents could play a role in the 3D distribution of nanoplastics in the oceans as well. Finally, it is important to note that most nanoplastics we detected (see Supplementary section S2) were aggregated with non-plastic material. This complicates collection via ltration since, as most of these particles, although signi cantly smaller than 500 nm, would be missed with pre-lters with pore sizes of 500 nm or greater. Likewise, the true concentrations of nanoplastics would be largely underestimated. Competing interests The authors declare no competing interests.

Additional information
Supplementary information is available for this paper at …

Reprints and permissions information is available at …
Correspondence and requests for materials should be addressed to F.M. or B.M.W.

Methods
Sample collection, handling, and preparation. Samples were collected during research cruise 64PE448 onboard the RV PELAGIA, from the South Atlantic at -32.171° S and 6.287° E. At this location water was collected from different depths using a Rosette sampler equipped with a CTD (Conductivity, Temperature and Depth) sensing system (Seabird SBE21) 1 . Before deploying the Rosette system, the bottles were thoroughly cleaned to avoid contamination: 1L glass sample bottles were thoroughly rinsed using ultrapure (UPW) water before they were lled straight from the Rosette system, to reduce contamination. The UPW used for cleaning was also checked for impurities using the same measurement procedure as for ocean water samples, but no contamination was found. Besides, the UPW was used to make blanks that were treated in the exact same way as the samples from the ocean. No nanoplastics were found in the blanks.
We sampled from the surface (5 m depth), the pycnocline at 60 m depth and the bottom layer at 5179 m depth. For Photo-induced force microscopy (PiFM) analysis 10 µL of sample was drop casted without any pretreatment using a Finnpipette (Thermo Fischer Scienti c; polypropylene tips) and spin coated onto 7x7 mm SiO 2 /Si(100) wafers (sample carriers) that had been cleaned as described previously. Spin coating was performed at 3000 rpm for 30 s using an Ossila spin coater operated in a fume hood.
Avoiding contamination. During transit, the samples were kept in closed glass bottles. All sample preparation steps were performed in the air-controlled environment of a laminar owhood. Cotton clothing was worn at all times, as cotton bers can easily be distinguished from nanoplastics with PiFM. All labware used, such as beakers, vials, and syringes, was made of glass. The Finnpipette tips were made of polypropylene (PP) (Thermo Fischer Scienti c). Even though we did not detect any PP, this polymer could easily be distinguished from polyethylene terephthalate (PET) in the photo-induced force microscopy (PiFM) measurements. Auxiliary equipment, such as tweezers and spatulas were made of stainless steel. Equipment was always rinsed 3 times with UPW, before usage. UPW was also measured as a reference, and we did not nd any nanoplastics. To make sure the SiO 2 /Si(100) wafers (sample carriers) were pristine before usage, they were ultrasonicated in a H 2  This instrument is equipped with a quantum cascade laser (QCL) unit and IR spectra were recorded in the 775-1950 cm -1 range. Atomic force microscopy (AFM) topography images were recorded in non-contact mode together with PiFM maps and IR point spectra (60 accumulations of 500 ms with 1 cm -1 spectral resolution). Additionally, hyperspectral maps, with every pixel being a full IR spectrum, were recorded. Typical image sizes were 1-400 μm 2 with a spatial resolution of 10-15 nm, dependent on the size of the AFM tip, and a probing depth of ~30 nm. The data were analyzed with the VistaScan 1.7, SurfaceWorks 2.4 software, Origin 9, and MATLAB.
PiFM machine calibration using varying sizes of polystyrene (PS) Nanospheres (Thermo Fischer Scienti c) can be found in reference 2 .
Particle size distributions (PSD) and particle concentration estimates of nanoplastics from PiFM measurements have been performed as detailed in reference 2 and summarized in Supplementary section S3.3.
PiFM IR reference spectra were recorded for PET, PS, and polyethylene (PE). For this, granular macroplastic pieces (PET) and beads (PS, PE) in the size range of 2-4 mm (Sigma-Aldrich) were used.
The plastic pieces were cut using a stainless-steel knife (Stanley) in order to create a smooth surface suitable for PiFM measurements. For all polymers, 3 hyperspectral images were recorded (1-4 μm 2 ) and the average of these 3 images was used as reference PiFM IR spectrum.
Lab-generated polyethylene terephthalate (PET) nanoplastics. The nanoplastics were prepared by aging PET (granular macroplastic pieces in the size range of 2-4 mm) in NaCl solutions. 280 mg NaCl was dissolved in 8 mL deionized water and 160 mg of PET (Sigma-Aldrich) was added to the solution (pH=7). For samples studied at pH=12, NaOH pellets (Alfa Aesar) were added to increase the pH to a value of 12.
The pH was measured with a TitraLab AT1000 Series (HACH) pH meter. After aging for 1, 2, 4, and 8 weeks, always 10 µL of solution were drop casted with a Finnpipette (Thermo Fischer Scienti c; polypropylene (PP) tips) and spin coated onto ~7x7 mm SiO 2 /Si(100) wafers at 3000 rpm for 30 s (Ossila spin coater operated in a fume hood).
Attenuated Total Re ection -Infrared spectroscopy. Attenuated total re ection-infrared spectroscopy (ATR-IR) reference spectra of bulk PET, PS, and PE (Sigma-Aldrich) were recorded using a PerkinElmer FT-IR spectrometer 'Frontier'. The spectra were recorded at room temperature in the 400-4000 cm -1 range (32 accumulations of 500 ms with 2 cm -1 spectral resolution).
The band assignments of reference IR spectra are displayed in Supplementary Fig. 1

Data availability
The data that support the ndings of this study are available within the paper and its Supplementary Information, and all data are available from the authors on reasonable request.

Figure 1
Studying individual nanoplastics in the South Atlantic Ocean with photo-induced force microscopy (PiFM). a, Sampling location in the South Atlantic Ocean close to Cape Basin. b, Infrared spectra of a polyethylene terephthalate (PET) bulk reference, obtained with attenuated total re ectance -infrared (ATR-IR) spectroscopy and PiFM, and of a PET nanoplastic particle (1) found at 5179 m depth. c, Topography map of -PET nanoplastic 1 and d, corresponding IR map recorded at the characteristic C=O stretching vibration of PET at 1725 cm-1. The location of the PiFM IR point spectrum from b, is indicated in the topography map by the label '1'. e, 3D topography map of the particles from c, indicating how the nanoplastic particle is half covered by another non-plastic particle. f, Depth pro le of the sampling location and the types of nanoplastics detected at every depth.

Figure 2
Studying individual polyethylene terephthalate (PET) nanoplastics in the laboratory. a, PiFM IR spectra of PET nanoplastics created after 1, 2, 4, and 8 weeks of microplastic degradation in salt water at pH=7, as well as after 4 weeks in salt water at pH=12. The PiFM IR spectrum of bulk PET (gray line) was added as a reference. The colored regions in the spectra indicate the C=O stretching peak (yellow), the wagging vibration modes of the ethylene glycol segment (green), the C-O stretching vibrations (blue), and the vibrations from the terephthalate group (magenta). b, Particle size distributions (PSDs) of PET nanoplastics after 1, 4, and 8 weeks of microplastic degradation in salt water at pH=7, as well as after 4 weeks in salt water at pH=12. c, PET nanoplastic concentrations over time during the degradation experiments as estimated per measurement. The size of the bubbles scales with the number of particles found per measurement. The pH=7 sample is indicated in red and the pH=12 sample in green. d, Height in Z dimension of all PET nanoplastic particles (red) and polystyrene (PS) size standards 25 (black) plotted versus the particles' XY diameter.