Quantifying the Chemical Composition and Real-Time Mass Loading of Nanoplastic Particles in the Atmosphere Using Aerosol Mass Spectrometry

Plastic debris, including nanoplastic particles (NPPs), has emerged as an important global environmental issue due to its detrimental effects on human health, ecosystems, and climate. Atmospheric processes play an important role in the transportation and fate of plastic particles in the environment. In this study, a high-resolution time-of-flight aerosol mass spectrometer (HR-ToF-AMS) was employed to establish the first online approach for identification and quantification of airborne submicrometer polystyrene (PS) NPPs from laboratory-generated and ambient aerosols. The fragmentation ion C8H8+ is identified as the major tracer ion for PS nanoplastic particles, achieving an 1-h detection limit of 4.96 ng/m3. Ambient PS NPPs measured at an urban location in Texas are quantified to be 30 ± 20 ng/m3 by applying the AMS data with a constrained positive matrix factorization (PMF) method using the multilinear engine (ME-2). Careful analysis of ambient data reveals that atmospheric PS NPPs were enhanced as air mass passed through a waste incinerator plant, suggesting that incineration of waste may serve as a source of ambient NPPs. The online quantification of NPPs achieved through this study can significantly improve our understanding of the source, transport, fate, and climate effects of atmospheric NPPs to mitigate this emerging global environmental issue.


Aerosol Mass Spectrometry and Potential Aerosol Mass reactor operation conditions
The Aerosol Mass Spectrometry (AMS) is the main instrument for collecting the mass spectra of the nanoplastic particles in this study.Due to the aerodynamic lens and the chopper, the AMS samples particles 10 million times more efficient than sampling gases. 1 Other statistical tools, including the PMF, are used to further aid the data analysis of complicated particle mass spectra collected from the AMS and described in Section S2 below.
The Potential Aerosol Mass (PAM) reactor is a horizontal 13 L copper cylindrical chamber (46 cm long × 22 cm inner diameter) operated in continuous flow mode, and could be used to generate Secondary Organic Aerosols (SOA) from Volatile Organic Compounds (VOCs) with either the ozonolysis or photooxidation pathways. 2,3In this study, the SOA particles are formed in the absence of any seed particles through OH radical or ozone oxidation.
Regarding OH radical oxidation, a 1 liter-per-minute (LPM) purified compressed air passing through the deionized water was introduced into the PAM for the humidification.Another 1.5 LPM air flow was introduced into the homebuilt ozone generator and then entered the PAM.
The VOC precursors were introduced into the PAM via the syringe pump injection into a threeneck round-bottom flask, which was carried out by a 1 LPM compressed air flow to the PAM.
The OH radicals were generated via the photolysis of ozone with UV lamps in the PAM.In addition, 1 LPM flow of ozone was introduced into the PAM and the ozonolysis was conducted under the dry condition.The UV lights in the PAM were turned off and the ozonolysis was the predominated reaction under dark condition to form low volatile oxidation products, which then undergo self-nucleation to form SOA particles.

Positive matrix factorization (PMF) working principle
PMF was first introduced in 1994 by Paatero and Tapper, 4 and has been a useful tool for resolving the time series data for mass spectrometer.The fundamental principle in PMF analysis is that the measured data matrix of mass spectrum could be expressed by the combination of selected number of factors based on mass conservation. 5The organic data matrix org with the dimensions  ×  is modelled according to Eq. S1: where P is the number of factors in the solution,  !,# is the signal of ion fragment j at time step i in the organic data matrix,  !$ is the concentration of factor p at the given time step i,  $# is the fraction of ion fragment j in the mass spectrum for the particular factor p, and  !# is the residual not included by the solution for ion fragment j at time step i.With a given factor number, the PMF solution provides a minimum summation of the weighted squared residuals Q as indicated by Eq.S2 below: where  !# is the estimated errors corresponding to the  !,# .As the Q value also depends on the dimensions of the data matrix and the number of factors chosen, normalizing the Q value with the expected Q value (Qexp), which is the degree of freedom of the solution, leads to a useful diagnostic.The absolute Q/Qexp value could be influenced by the unknown modelled uncertainties, wrongly chosen number of factors, etc., and consequently, monitoring how the Q/Qexp varies among different iterations could assist in assessing the optimal solution. 6s with signal-to-noise ratio (S/N) lower than 0.2 were removed, and the ions with S/N between 0.2 and 2 were downweighed by increasing the estimated error values. 5,7The PMF solutions with one to ten factors are evaluated in this study.To further explore how rotation and the random starting point would influence the uncertainty of the solutions, different FPeak (from −1 to +1, step: 0.1), and SEED values (from 1 to 10, step: 1) were adapted, as described by Zhang et al. 8

Detailed analysis procedures of ME-2
As mentioned in the main text section 2.3, ME-2 allows for solution of S1 while using a prior information about component mass spectra or time series.In this work we utilize a reference input profile for the PS mass spectrum with a scalar a value that determines the extent to which the derived factor profile could vary from the input spectrum profile. 6,9The a value spans from 0 to 1 with an increment of 0.1, while the lower the a value, the more rigorous the constraint is.For example, if a = 0.1, for the profile of the extracted factor, the intensity of all the ions could vary as much as ± 10% compared with the input reference spectrum.The ME-2 with the a value approach could provide a more complete exploration of the rotational ambiguity of the solution space. 6

The relative ionization efficiency (RIE) of polystyrene nanoplastic particles
To develop the calibration curve of pure polystyrene (PS) nanoplastic particles (NPP), a mixing condensation particle counter (CPC, Model 1720, Brechtel) was deployed together with AMS.The aerosols generated are split into two lines for AMS and the CPC, and the detected ion rate in Hz (I) from AMS could be calibrated with the CPC measurements. 10,11The AMS ion signal was compared with the input number of molecules derived from the CPC particle number concentration and an assumed collection efficiency of 1 to obtain the ionization efficiency.
With ammonium nitrate as the standard, the relative ionization efficiency (RIE) for PS particles can be derived by comparing the ionization efficiency of the PS and ammonium nitrate.
The detailed equation to calculate RIE is described in Eq.S1, Eq.S2, and Eq.S3 below: 10 where the Cs is the mass concentration of the PS particles in this study, Q is the volumetric sample flow rate in cm filtering period and the time series of tracer ions during the filtering period, which are 16 and 14 times lower than ambient-derived concentrations, respectively.

The comparison of PS NPPs quantification with the AMS and pyrolysis-GC/MS
To further validate the quantification of PS NPPs with the AMS, the pyrolysis GC-MS analysis was also conducted following the established method. 12PS NPPs standard were collected by the AMS and onto a glass fiber filter (Cytiva, 0.7 μm particle retention) simultaneously with 2 liter per minute flow rate for 77 minutes.
The pyrolysis GC-MS method was calibrated with the mass of PS NPP standards ranging from 0.5 to 5 μg with the method established in previous study. 12Briefly, the standards were first injected in the pyrolysis chamber.Pyrolysis was performed at 700 °C for 1 min, and thermodesorption was performed at 300 °C for 1 min.Gas phase pyrolysis products were injected directly into a Shimadzu GCMS-QP2020NX (Shimadzu Corporation, Kyoto, Japan) equipped with an RTX-1 capillarity column (30 m, 0.25 mm i.d., 25 μm film thickness).Column flow was 1.1 mL/min with helium as the carrier gas at a split ratio of 5.The injection port temperature was set at 310 °C.GC oven started at 50 °C for 2 min and ramped to 180 °C at 15 °C/min, increased to 310 °C at 5 °C /min, and hold at 310 °C for 44 mins.Mass spectrometry scanned from 50 to 600 m/z range with ion source temperature at 200 °C and interface temperature at 280 °C.As shown below in Figure S8, the peak (~5 min) on the total ion chromogram (TIC) shows mass pattern matching to Styrene, a monomer of PS. 13 The instrument and membrane filter blanks do not show this PS peak on their chromogram.The areas of the PS peak (~5 min) on the TIC showed a linear relationship with the spiking masses, (R 2 > 0.999).
The linear relationship indicates the feasibility of quantification for PS MPs in test samples.
Two replicated sets of 16 circular pieces (1.9 mm diameter) with a total area of 0.454 cm 2 were cropped from the sample filter.Given that the effective filter area is 7.55 cm 2

The second ambient sampling
The time series data for organic aerosols during the ambient2 sampling period is shown in Figure S14.Applying similar analysis procedures to ambient1, the best ME-2 solution for ambient2 sampling period has 3 factors, and they are identified as constrained PS factor, moreoxidized OOA (MO-OOA), and hydrocarbon like organic aerosols (HOA).Again, the MO-OOA factor has a higher f44 and O:C, and the HOA factor contains ions that have been identified as markers of fresh fossil fuel combustion, including CnH2n-1 + and CnH2n+1 + . 14, 15

ME-2 analysis of synthetic data matrix
To further validate the ability of ME-2 to quantify the concentration of ambient PS NPPs, the organic data matrix was modified before reapplying the ME-2 analysis.First, all the ME-2 derived factors were summed up except the extracted PS factor.The ME-2 analysis with the synthetic organic data matrix as input did not extract any meaningful PS factor, as the 10minute moving average concentration has always been lower than the detection limit (12 ng/m 3 for 10 minutes).Second, the PS section in the original organic matrix was substituted with a fraction of the PS factor extracted by the original ME-2 analysis.50%, 10%, 5%, and 2% of the original PS factor were added to the organic matrix generated by summing all factors except the original PS factor, and the ME-2 analysis was reapplied.The averaged PS concentration derived from the ME-2 analysis with the synthetic data matrix during the intentional injection period was 51%, 9%, 4%, and 2% of the original PS factor derived with the actual ambient data matrix with their absolute mass loading in the table S1.The additional ME-2 analysis successfully extracted a PS factor of around 16 ng/m 3 , which is lower than the averaged concentration derived by the ME-2 for ambient sampling (30 ± 20 ng/m 3 ).Such results further validate the quantification of ambient PS NPPs with AMS and ME-2 analysis.

The concentration of airborne microplastic particles from literatures
In previous studies, the concentration of atmospheric microplastic particles was reported in various units across different geographical areas.Herein, this section illustrates the conversion of the previously reported concentrations into ng/m 3 , which is the unit of measurement utilized in the present study, to provide a uniform basis for comparison.The majority of prior research quantified the microplastics in terms of number concentration, i.e.
count per cubic meter. 16,17,18,19By assuming a sphere for the collected microplastics, with the density and the midpoint of the reported size, the mass concentration of microplastics are derived from the number concentration.Another study reported the concentration of microplastics as a percentage of the total particles collected, 20 and considering our sampling site, a total organic concentration of 6 μg/m 3 , which is the averaged concentration measured during the sampling, is adopted for the derivation in the present study.

Concentration of PS nanoplastic particles derived from the tracer ions
As described in section 3.4, to further validate the feasibility of using tracer ions to derive the abundance of PS nanoplastic particles, the concentrations from ME-2 and tracer ions are assessed.The concentration is calculated based on the normalized intensity of the tracer ion among the mass spectrum of the pure PS particle standard with the Eq.S4 below: where [PS] is the concentration of PS nanoplastic particles in μg/m 3 , [C8H8 + ] is the concentration of C8H8 + in μg/m 3 , and 0.06 is the normalized intensity of C8H8 + for the mass spectrum of PS NPP standard.

S24
Table S1.The ratio of the PS factor added to the synthetic organic data matrix for ME-2 analysis to the total PS factor derived from the original ME-2 analysis, the ratio of the PS factor derived from the corresponding ME-2 analysis to the total PS factor derived from the original ME-2 analysis, and the averaged mass loading of the PS factor derived from the corresponding ME-2 analysis during the intentional injection period.
3 s -1 , NA is Avogadro number, and ∑  +,! is the summation S5 of the detected ion rate in Hz for all fragment ions.IEs and MWs are the ionization efficiency and molecular weight of the PS particles, respectively, and the  27 ' and  27 ' represent the ionization efficiency and molecular weight of nitrate, respectively.The molecular weight of polystyrene cancels out by combining Eqns.(S1) and (S2) to derive Eqns.(S3).To further test the reproducibility of the decomposition products of PS NPPs, the correlation values of the five mass spectra of PS NPPs collected during the five points calibration was calculated and plotted in Figure S4 below, and it shows nearly identical mass spectra.Besides, during the calibration experiment, the background blank samples were tested.Below in Figure S5, S6, S7 show the AMS mass spectra collected during the nitrate calibration, , the mass of PS in each set of 16 circular pieces was projected to the mass of PS in the whole sampling filter by multiplying 16.63 (the ratio of the sample area to the measured area).Each set of 16 circular filter pieces was analyzed on a pyrolysis probe (CDS 6200, CDS Analytical, LLC.Oxford, PA, USA) with the same pyrolysis and GC/MS methods from above.The mass of PS NPPs in the 16 pieces of filter circles was estimated to be 0.57µg, which corresponding to 9.45 µg of PS collected on the filter and 61.36 µg/m 3 in the air sample measured by Pyrolysis-GC/MS.The mass concentration during the sampling time quantified by the AMS was 60.78 µg/m 3 .The difference of PS concentration quantified by the AMS and the Pyrolysis-GC/MS is shown to be 2%, further validating an accurate quantification of PS NPPs with the AMS.

Figure S1 :
Figure S1:The normalized number and mass distribution of self-nucleated SOAs generated through PAM.

Figure S3 .
Figure S3.Schematic of ambient sampling setup.The excess air pump 1 drew 5 LPM of air

Figure S4 . 3 Figure S6 .Figure S7 .
Figure S4.The Pearson's correlation heatmap of the five mass spectra of PS NPPs collected

Figure S14 .(
Figure S14.The time series of total organic concentration during ambient1 (A) and ambient2 217

Table S2 .
The averaged mass concentration and standard deviation of PS NPPs during specific period and their corresponding p value from t-test.The 4-6 pm indicate the background period, and the rest are the color-coded period in Figure4A.44´10 -3 1.54´10 -76 2.34´10 -79 8.66´10 -98