Correlating Scanning Electron Microscopy and Raman Microscopy to Quantify Occupational Exposure to Micro- and Nanoscale Plastics in Textile Manufacturing

Abstract

Airborne micro- and nanoplastic particles (MNPs) are increasingly recognized as a potential occupational exposure hazard, yet substance-specific workplace data remain limited. This study quantified airborne MNP concentrations during polyester microfiber production using a correlative SEM–Raman approach that enabled chemical identification and size-resolved particle characterization. The aerosol mixture at the workplace was dominated by sub-micrometer particles, with PET—handled onsite—representing the main process-related MNP type, and black tire rubber (BTR) forming a substantial background contribution. Across both sampling periods, total MNP particle number concentrations ranged between 6.2 × 10⁵ and 1.2 × 10⁶ particles/m³, indicating consistently high particle counts. In contrast, estimated MNP-related mass concentrations were much lower, with PM₁₀ levels of 12–15 µg/m³ and PM_2.5 levels of 1.3–1.6 µg/m³, remaining well below applicable occupational exposure limits and near or below 8 h-equivalent WHO guideline values. Comparison with earlier workplace and indoor studies suggests that previously reported concentrations were likely underestimated due to sampling strategies with low efficiency for small particles. Moreover, real-time optical measurements substantially underestimated particle number and mass in this study, reflecting their limited suitability for aerosols dominated by small or dark particles. Overall, the data show that workplace MNP exposure at the investigated site is driven primarily by very small particles present in high numbers but low mass. The findings underscore the need for substance-specific, size-resolved analytical approaches to adequately assess airborne MNP exposure and to support future development of MNP-relevant occupational health guidelines.

1. Introduction

Plastics are defined by the International Union of Pure and Applied Chemistry (IUPAC) as polymer materials that can include a large variety of additives, including UV stabilizers, fillers, and dyes, to improve performance and reduce production costs [1]. According to OECD in 2022, the world produces more than 430 million metric tons of plastics per year [2]—generally derived from fossil fuel and gas precursors. The most frequently used types are polystyrene (PS), polyethylene terephthalate (PET), polyurethane (PUR), polypropylene (PP), polycarbonate (PC), polyamide (PA), polyvinyl chloride (PVC), and polyethylene (PE). Ubiquitous in modern society, plastics are employed widely across a multitude of domestic, industrial, and technological applications. With their industrial-scale production, plastics have become indispensable due to their versatility, high performance, and low cost. Today, over 5000 types of plastics are marketed, many containing hundreds of chemicals, posing ecological hazards when released into the environment. This massive and widespread use has raised environmental and health concerns [3,4,5,6,7].

The smaller fractions of plastic litter can be classified as microscale plastics particles (microplastics, MPs, <5 mm) or nanoscale plastics (nanoplastics, NPs, <1 μm). For both, the combined term micro- and nanoplastics (MNPs) is used. MNPs can be distinguished into two general material classes: primary MNPs, intentionally manufactured for products such as cosmetics, and secondary MNPs, formed from the disintegration of larger plastic items like bottles, bags, and textiles [8,9,10]. MNPs contribute to “white pollution”, i.e., the visible accumulation of plastic waste in the environment [11,12]. Fibrous MNPs are primarily released over the lifecycle of synthetic microfibers of textiles, as those can break up into thinner and shorter fibrils by abrasion processes occurring during production, use, laundering, and recycling [13,14]. They are a major source of microplastic pollution [15].

Airborne particle and fiber inhalation can contribute significantly to human MNP exposure, alongside ingestion and dermal contact [16,17]. MNPs can contain and adsorb organic and inorganic contaminants [18], act as substrates for microorganisms, and, after aging, host polymer derivatives that result from oxidation and photolysis. Such contaminants can induce environmental and biological toxicity in initially inert materials [19]. A growing body of experimental and review studies consistently report oxidative stress, inflammation, tissue damage, and particle translocation to organs, suggesting systemic health risks resulting from MNP exposure. However, most data derive from in vitro or high-dose animal and cell studies, limiting direct extrapolation to real-world human exposure. There remains an urgent need for standardized exposure assessments and human-based evidence to clarify the epidemiological relevance of these findings [20,21,22].

A recent literature review stressed that reported MNP release and exposure studies mainly focus on the post-production or consumer phase [23]. A few studies focusing on the production cycle mainly addressed release into the environment, e.g., by wastewater [10,24,25,26,27,28]. MNP release into the air is underexplored, as addressed recently in a review by Rednikin et al. [29]. Only a few studies specifically assessed indoor environments, such as offices and apartments [30,31], during the drying of synthetic clothing [32]. Specifically, occupational diseases pose a critical knowledge gap, even though occupational diseases correlated with inhalation exposure to non-polymer textile fibers, such as cotton, have been diagnosed in workers in the context of polymer textile production. One example is the so-called flock worker’s lung, which developed in workers in the nylon flocking industry [33,34]. For a generalized occupational risk assessment, the needed exposure scenarios cannot currently be accurately described because probabilistic predictive frameworks are missing [35]. Their development would be enabled by a sufficient quantitative database with data for characterization of particle release and measurements of personal exposure. However, there is no recent scientific literature presenting studies addressing inhalative occupational exposure with sufficient analytical power to differentiate the MNP fraction in the general particulate matter [36].

Assessing inhalative exposure to MNPs not only at workplaces but in the atmosphere, in general, is difficult since they occur as part of a complex aerosol, comprising not only locally released particles but also a particle background transported in from the workplace’s premises. A prominent component of the background aerosol is particles stemming from tire wear occurring during urban traffic [37,38,39]. Often, black tire rubber (BTR) particles are labeled to point out their tire-wear-related origin; these particles contribute to the ubiquitous MNP exposure of the environment and humans. BTR particles usually comprise multiple heavy metals, polycyclic aromatic hydrocarbons, and other additives that can be toxic by themselves or as mixtures [40,41]. Their presence in the particle background adds an additional challenge when aiming to identify emission sources since they cannot be easily distinguished from the locally produced ones by means of physical properties. Direct-reading particle concentration instruments and aerosol filter sampling that do not perform airborne particle classification based on morphology and chemical substance only serve to assess the general particle burden. A similar challenge can be found for mixtures of airborne fibers and particles and their agglomerates, for which scanning electron microscopy (SEM)-based measurement strategies have been developed that use morphological properties to recognize, classify, and quantify fiber concentrations at workplaces where fibers of known origin are handled [42]. However, for complex mixtures of particles and fibers of unknown origin and composition, a purely SEM-based strategy will fail. Instead, analytical methods must be included that are capable of identifying particle chemistries. SEM secondary electron imaging allows for fast and very high-resolution imaging. Using SEM-based energy dispersive X-ray spectroscopy (EDS), many inorganic particles can be classified based on their elemental composition. However, MNP-particle identification is difficult since their EDS-derived elemental compositions are less distinctive. Instead, Raman microscopy has gained importance in the field of MNP analysis [43,44,45]. Raman spectroscopy allows for the identification of many chemical groups and molecular structures, even of micro- and nanoscale particles, including polymers and carbon materials [46,47].

The present study introduces the workflow in order to image, localize, chemically characterize, and, finally, identify individual micro- and nanoscale particles on workplace aerosol samples. Morphological information from SEM is correlated to chemical information from Raman microscopy with high spatial resolution, allowing the assessment of the occupational MNP burden in terms of ambient MNP concentration in the workplace air, which would lead to MNP exposure. To demonstrate the power of the analytical procedure, a case study was conducted in which airborne particles were collected at a workstation within a Romanian polyester microfiber textile manufacturing factory on filters and analyzed to classify the composition of the workplace aerosol, including MNP and tire rubber particles, and quantify the respective inhalable and respirable fractions.

2. Methods

2.1. Sampling Site

Aerosol sampling was conducted at two time points (T1 and T2) while workers were at their workstations in the production environment of a Romanian textile production facility. The study focused on cutting and sewing processes, which were expected to generate and emit fine particles and microfibers due to friction and high working speeds. Sampling was synchronized with the production schedule: the first measurement (T1) commenced at 07:00 a.m., and the second (T2) at 03:00 p.m. Only one sample per measurement was collected, as we were aiming to test the analytical approach and perform sophisticated exposure assessment.

2.2. Aerosol Sampling

Two VDI-3492-compatible samplers were used (GSA Messgerätebau GmbH, Ratingen, Germany). They hosted track-etched membrane filters made of PC with a diameter of 25 mm, 10% porosity, and pores of 0.8 µm in diameter, manufactured by APC, Eschborn, Germany. The samplers were positioned at the same height as the machine’s working point and at a distance of 30–50 cm. In contrast to standard 40 nm top and 20 nm bottom sputter-coated air-sampling filters for SEM analysis, these filters were prepared with a 60 nm-thick top and 40 nm bottom gold coating. This effectively suppressed the Raman signal from the PC of the filter to interfere with particle signals, but effectively reduced filter pore sizes to ca. 0.6 µm.

Thus, for particles larger than ca. 0.6 µm, a near 100% collection efficiency can be expected. Collection efficiency is the ratio of particles deposited on the filter membrane’s surface to all incoming particles. As non-impacting smaller particles can become lost through or inside the filter pores, only particles deposited on the membrane surface are microscopically detectable. The size-dependent filter collection efficiency for particles smaller than 0.6 µm is unknown, but most likely decreases to a few percent only for nanoscale particles.

The samplers were connected to Gilian GilAir Plus (Sensidyne, St. Petersburg, FL, USA) personal sampling pumps that were operated at a constant flow. For this study, a flow rate of 2.5–3 L/min was applied, in accordance with the VDI-3492 guideline for the measurement of indoor air pollutants, yielding high collection efficiencies for particles smaller than 10 µm [48]. The ambient conditions during sampling were temperatures of 25–28 °C (77–82.4 °F) and relative humidity of 45–53%.

Filters were weighted prior to aerosol sampling and afterwards, following ISO 15767 [49].

2.3. Fiber Identification

Since all sampling sites were located within a textile production facility, where polyester microfibers were handled, emissions of fibrous polyester fragments were apparent. A fibrous particle is classified as “fiber” based on its geometric properties derived from its 2-dimensional microscopic projection image. Specifically, so-called WHO-fibers were to be identified. They are defined in accordance with the WHO counting convention by a length > 5 µm, a diameter < 3 µm, and a length-to-diameter aspect ratio > 3:1 [50]. Thus, all objects detected on SEM images of the filter surface were characterized morphologically and classified. WHO-fibers were quantified with respect to the analyzed air volume.

2.4. Procedure for Correlative SEM–Raman Microscopy

Filter samples were analyzed using correlative microscopy (CM), combining SEM (SU8230 by Hitachi, Tokyo, Japan) and Raman microscopy (RM, Alpha300 apyron by WITec, Ulm, Germany). Both microscopes were equipped with fully motorized, software-controlled stages. For Raman microscopic imaging, a Zeiss EC Epiplan-Neofluar 100×/0.9NA objective (Zeiss, Oberkochen, Germany) was used. Video camera imaging resulted in an additional magnification factor of 11.6×. Raman spectra were acquired using a laser wavelength of 532 nm and 1 mW laser power, a spectrometer of 300 mm focal length equipped with a grating of 600 lines/mm blazed at 550 nm, and an EMCCD camera (Newton DU970P-BVF-355 by Oxford Instruments GmbH, Wiesbaden, Germany). For each filter sample, 50 SEM micrographs of 5120 × 3840 pixels were acquired with a pixel resolution of 24.8 nm, corresponding to a magnification of 1000×. For secondary electron (SE) imaging, an acceleration voltage of 0.8 kV was used. The image frames were randomly distributed across the filter using custom control and acquisition software that took care of correlating a sample-specific coordinate system to the stage coordinate system. To further increase the spatial correlation of the SEM and Raman analysis data, the random filter pore pattern was used. By matching the pore pattern of the optical microscope image of the filter frame to that in the SEM image, spatial correlation accuracies well below 100 nm were achieved. For Raman analysis, a subset of 10 of the 50 SEM images was selected randomly. These 10 images were sufficient to extrapolate Raman-derived particle identity statistics to the total filter area, as is manifested in the narrow 95% confidence intervals of the results.

SEM acquisition, particle counting rules, particle analysis, and data evaluation followed the procedure described in CEN/TS 18117:2025-04: “Workplace exposure—Detection and characterization of airborne NOAA using electron microscopy—Rules for sampling and analysis” [51].

2.4.1. AI-Assisted Particle Analysis

For particle analysis, custom software (FibreDetect) was used [52]. Using a pre-trained artificial neural network (ANN), binary semantic segmentation of SE images was performed. Groups of pixels classified as “non-filter” by the ANN were allowed to locate image areas that showed collected particles [53]. As the pre-trained ANN’s performance was not satisfactory, resulting in artifacts and unidentified particles, the ANN was re-trained with annotated SE images from BAuA’s training image repository containing indoor and outdoor aerosol samples showing dust particles of various sizes, morphologies, and other distinctive features, as well as similar substrate-related background properties. For the re-trained ANN, the segmentations still required visual validation and minor manual rework, either by deleting artifacts or by manually segmenting unrecognized particles. These corrected image annotations will allow us to improve the ANN’s performance in the future.

2.4.2. Raman Analysis of Particle of Interest

Particles of interest (POI) were selected based on their minimum size for the following two reasons. As mentioned above, for particles smaller than ca. 0.6 µm, the collection efficiency was undetermined, and thus, extrapolated results were highly inaccurate. Also, for smaller particles, weaker Raman signals result. Their detectability, however, also depends on the Raman scattering behavior of the particle’s material. Very small and/or low-Raman active particles can require long signal integration times and/or high laser powers to obtain spectra. Polymer-based particles on dry filters, however, may be prone to laser irradiation damage. Hence, it was decided to use a 2-D-projected SE image area threshold of about 500,000 nm² as a POI-selection criterion, which corresponds to the area of a filter pore of ca. 0.8 µm in diameter.

Reference spectra were acquired on original textile samples from the production site and added to the reference library. On these materials, strong calcium carbonate signals contributed to the spectrum and were mixed with the polymer signals. Calcium carbonates are used as water repellents and are frequently employed as fillers and performance enhancers, particularly in non-woven fabrics, with the objective of enhancing mechanical strength, tear resistance, and durability [54]. Furthermore, it is employed as a sizing agent, and its use extends even to the realm of environmental management, where its ability to decolorize dyes in wastewater is particularly advantageous [55].

Selected filter particles were chemically characterized by RM. Collected spectra were compared to those in the reference library to determine material identities. Spectral acquisition and averaging continued until a good signal-to-noise ratio and a ≥70% match with a database reference spectrum was reached, using the instrument’s analysis software TrueMatch (WITec, Ulm, Germany). In addition, the study-specific reference material library, as well as the commercially available S.T.Japan database, was employed [56]. Identifications were verified visually and refined when necessary.

Black tire rubber particles proved difficult to identify. They showed rather variable spectra, since tire compositions differ depending on the manufacturer and aging state. Tire rubber typically contains ~28% carbon black, 14% natural rubber, 27% synthetic rubber, 14% steel wire, and 16% fillers, fabrics, additives, and sulfur-containing vulcanization chemicals [57]. Carbon black (soot) displays two characteristic Raman peaks: the G-peak at 1582 cm⁻¹ (sp²-hybridized graphitic carbon) and the D-peak at 1345 cm⁻¹ (structural defects, heteroatoms). Signals of natural and synthetic rubber occur mainly in the fingerprint region (400–1800 cm⁻¹) and at ~2900 cm⁻¹ (CH₂-stretch), but in black rubber, characteristic bands are often masked by carbon black. The rubber contribution at 1600 cm⁻¹ can merge with the carbon G-peak. The latter serves to differentiate between black tire rubber and other rubber particles.

2.4.3. Chemical Classification for Substance-Specific Particle Counting

Particles were grouped into different classes based on their chemical composition, as determined by Raman spectroscopy (e.g., PET, PP, mineral, and soot). Particle classes were included in the exposure evaluation when at least three particles of the same chemical substance were detected within the analyzed area. Particles not meeting this criterion were assigned to the “Other” category. Particles that could not be identified based on the matching with Raman reference spectra were classified as “Unknown”. Those particles that were below the area threshold but were, nonetheless, collected on the filter were not analyzed and labeled “Not classified”. These, generally, nanoscale particles were most probably of combustion origin, e.g., soot.

Fibers underwent additional Raman analysis to determine their polymer-type and to identify those that were most likely organic. Also, skin flakes sometimes appeared elongated.

2.4.4. Determination of Particle Number Concentration

The CM analysis procedure resulted in a dataset of particle counts per class together with the projected areas, and with the minimal and maximal Feret diameters of all located particle segments. The procedure used to extrapolate results based on relatively small filter areas represented by 10 SEM images to air concentrations is described in CEN/TS 18117:2025-04 and ref. [42]. From the resulting count statistics, 95% confidence intervals (CI) of the expectation value of particle counts per SEM image were calculated. The average particle number concentration and its CI resulted as a ratio of these values and the product of the specific sampling volume with the evaluated filter area. The specific sampling volume is the ratio of the total sampled air volume and the open filter area. The evaluation area was determined by the evaluated number of SEM images times their area of about 127 × 95 µm².

The dataset from the CM analysis procedure, therefore, yielded CIs and expectation values for number concentrations per particle class, and thus, material-specific concentrations. Fiber counts were handled in the same manner.

Descriptive statistics for the particle sizes gave substance-specific mean particle sizes and their standard deviations. There is no established size-representing metric for particles since the SEM images provide only 2D projections. Projected area diameters, as well as minimal and maximal Feret diameters, are routinely measured, but their interpretation in terms of particle volume is lacking due to missing information about the third dimension. Therefore, particle volumes were approximated geometrically. Particle minimal $d_{f,min}$ and maximal Feret diameters $d_{f,max}$ were used to construct ellipsoids, with volume calculated as follows:

$$V_e={\displaystyle \raisebox{1ex}{$\pi $}\!\left/ \!\raisebox{-1ex}{$6$}\right.}{d_{f,min}}^2{d}_{f,max}.$$

This particle volume was used to determine the diameter of an equivalent sphere as a representative of particle size.

2.4.5. Estimation of Particle Mass Concentrations

The calculated ellipsoid volumes in each particle class bin were factored with the substance-specific material densities, given by the Matweb.com material property online database, to determine particle masses. For “Other”, “Unknown”, and “Not classified” particles, unit density was assumed. The sum of all particle masses in each class bin represents the substance-specific total particle mass.

The ellipsoid volume converted to an equivalent sphere diameter of unit density was used as a proxy for the aerodynamic diameter and allowed us to differentiate inhalable (PM₁₀—particulate matter with sizer smaller than 10 µm) from respirable (PM_2.5) fractions. In each class bin, particles smaller than the threshold diameters of 10 µm and 2.5 µm were selected to determine the fractional particle mass.

In the same manner as the derivation of particle number concentrations, particle mass concentrations were subsequently determined. Although approximate, this method provided a reasonable estimate.

2.5. Real-Time Particle Measurement

A Mini-LAS 11-E Laser Aerosol Spectrometer (GRIMM Aerosol Technik, Muldenstausee, Germany) was used for online monitoring of particle concentrations at the workplaces. This portable optical particle counter uses laser light scattering to detect and measure particles in the size range from 0.25 to 32 µm in 31 size channels. Assuming a standard material density, it reports mass fractions (PM₁₀, PM_2.5) and full-size distributions if required. The device has a sample flow rate of ~1.2 L/min and a mass concentration range of 0.1–10,000 µg/m³. Measurements were conducted for particle diameters of 0.3, 0.5, 1.0, 2.5, 5.0, and 10 µm, with 20 repeated readings per size. In addition, mass concentrations for PM_2.5 and PM₁₀ were recorded 20 times each. Measurement sets took place simultaneously with the aerosol samplings.

3. Results and Discussion

3.1. Aerosol Sampling and Correlative SEM–Raman Microscopy

3.1.1. Filter Weighing

Table 1. Results of filter weighing and sampling conditions.

Sample	Filter Weight [mg]			Air Volume (mL)	Concentration [µg/m3]
	Initial	After	Difference
T1	3.974	4.062	0.088	347,234	253.43
T2	4.225	4.252	0.027	477,804	56.50

shows the results of particle collection on track-etched membrane filters at time T1 (sample T1) and time T2 (sample T2). Filter weights obtained prior to and after sampling are given in Table 1. Mass concentrations were determined by the ratio of the weight difference in filters and the sampled air volume.

3.1.2. SEM Micrographs

Figure 1 shows representative SEM micrographs of the T1 filter together with magnified insets of selected particles. The images show that morphological or visual features alone are insufficient for particle material differentiation. Characterization via Raman spectroscopy was generally mandatory to identify a substance chemically.

3.1.3. Substance-Specific Particle Number Concentrations

Figure 2a,b provides a graphical summary of particle concentrations related to chemical composition for the two PC filters. Detailed numerical results and full chemical compositions are provided in

Table 2. Numerical results of the correlative microscopy procedure for sample T1.

Sample T1	Particle Number Concentration				Particle Mass Concentrations			Size Statistics
Particle Class	LCI [106/m3]	Midpoint [106/m3]	UCI [106/m3]	Density [g/cm3]	Total [µg/m3]	PM10 [µg/m3]	PM2.5 [µg/m3]	Mean [µm]	SD [µm]
Black Tire Rubber	0.52	0.69	0.85	1.4	22.3	5.3	1.0	1.7	1.5
Calcite	0.01	0.04	0.08	2.7	3.2	2.1	0.2	2.7	1.8
Mineral	1.12	1.36	1.59	2.5	66.3	43.1	4.8	2.1	1.3
MNP	0.03	0.08	0.14	1.1	31.6	1.4	0.1	5.2	4.1
Natural	1.26	1.50	1.75	1.0	105.0	52.7	1.9	2.6	2.3
Not Classified	14.56	15.35	16.14	1.0	6.5	4.2	0.9	0.4	0.3
Other	0.25	0.37	0.49	1.0	8.0	5.2	0.7	2.0	1.4
PET	0.08	0.16	0.23	1.4	7.5	4.9	0.2	2.9	1.6
Soot	0.61	0.79	0.97	1.8	2.2	1.4	1.2	1.1	0.4
Unknown	0.36	0.50	0.65	1.0	10.7	7.0	0.5	1.8	1.6
Total	18.78	20.84	22.89		263.3	127.1	11.4

(T1) and

Table 3. Numerical results of the correlative microscopy procedure for sample T2.

Sample T2	Particle Number Concentration				Particle Mass Concentrations			Size Statistics
Particle Class	LCI [106/m3]	Midpoint [106/m3]	UCI [106/m3]	Density [g/cm3]	Total [µg/m3]	PM10 [µg/m3]	PM2.5 [µg/m3]	Mean [µm]	SD [µm]
Black Tire Rubber	0.63	0.77	0.91	1.4	15.1	9.8	1.4	1.6	1.2
Mineral	1.07	1.25	1.44	2.5	125.4	62.4	5.4	2.4	1.7
MNP	0.01	0.05	0.08	1.1	2.2	1.5	0.0	3.1	1.4
Natural	1.29	1.49	1.69	1.0	108.6	21.4	2.2	2.2	2.2
Not Classified	6.39	6.82	7.25	1.0	0.5	0.3	0.3	0.3	0.2
Other	0.05	0.10	0.14	1.0	8.6	0.2	0.1	2.3	2.9
PET	0.08	0.14	0.21	1.4	6.1	3.9	0.2	2.6	1.6
Soot	0.04	0.09	0.13	1.8	0.8	0.5	0.2	1.3	0.9
Unknown	0.13	0.20	0.28	1.0	6.2	4.1	0.4	2.4	1.5
Total	9.68	10.91	12.13		273.5	104.0	10.1

(T2). Both samples were highly heterogeneous, comprising 9 (T1) and 8 (T2) distinct material classes.

The collected samples also contained organic material that could not be clearly identified but showed spectral characteristics of biological material and was, therefore, grouped as “Natural”. The “MNP” class comprises polymers that were not actively handled at the site, including PC, PS, PP, PA, and various non-black rubber types. The “Mineral” class comprises particles of various metal oxides and sulfates, as well as quartz. Apart from PET, calcite was found on the T1 sample, a material supposedly used as an additive to textiles, as described above. Particles of various chemical substances, but occurring in one or two instances, were allocated to the “Other” class, such as particles made from copper, various salts, and pigments, amongst others.

Extrapolated particle number concentrations for each class determined the area of pie chart segments in Figure 2a,b. PET, the material actively handled onsite, was the dominant microplastic particle type in both samples, while black tire rubber contributed substantially to the aerosol background, alongside soot, mineral particles, and natural particles. Numerically, most particles were “Not classified” due to their small size, indicating a strong underestimation of their extrapolated concentration and thus of the estimated total particle number concentration in the workplace atmosphere.

The additional analysis of fibers present on the filters yielded twenty-two WHO-fibers on 50 SEM images of T1, one of them being made from polyethersulfone (PES), and with the others being most probably of natural origin. For time T2, fifteen WHO-fibers were found on 50 SEM filters, with two of them being black tire rubber, one being nitrile rubber, and the rest most probably being from natural sources. These numbers can be converted to fiber concentrations of 626 fibers/m³ for time T1 and 318 fibers/m³ for time T2. Such exposure levels are far below occupational exposure levels that currently exist for asbestos-based WHO-fibers (e.g., 10,000 fibers/m³ in the European Union) [58].

The particle number concentrations of PET particles were found to be in the range of 77 × 10³–234 × 10³ #/m³ for T1 and of 82 × 10³–205 × 10³ #/m³ for T2 (95% confidence interval). These particles most likely stemmed from the textile production process, although mixing with background PET particles could not be excluded. Including the fraction of BTR and other MNPs, which were most likely part of the background aerosol either being transported from the urban environment or produced elsewhere in the factory, the total plastic burden was 625 × 10³–1226 × 10³ #/m³ (T1) and 721 × 10³–1200 × 10³ #/m³ (T2). Thus, possible MNP exposure was fairly similar for the two work shifts.

To the best of the authors’ knowledge, workplace exposure to airborne MNP has been investigated only to a limited extent. Existing studies often did not employ analytical methods with sufficient sensitivity to quantify exposure metrics at a comparable level of detail. Similar campaigns applied light-microscopical means with additional Fourier transform infrared spectroscopy (FTIR) or Raman microscopy as the single analytical method for particle detection and identification of aerosol samples. The former has been applied to assess production-related workplaces in the poly ethylene propylene diene (EPDM) industry, where concentration levels of 360–559 #/m³ have been reported, with a high diversity of plastic types, which were several orders of magnitude lower compared to the results of the present study [59]. The latter has been used in a recent conference paper, whose abstract reported inhalable microplastic concentrations of 1500–2000 #/m³ in non-woven fabric production and plastics recycling facilities [60], and which was also much lower compared to the results presented here. Other indoor environments, such as apartments and cars [61], as well as university dormitories, offices, and dining halls [62], yield similar, much lower number concentrations. Other studies conducted in public indoor environments reported much lower concentrations [63,64]. Most likely, the reported studies applied aerosol-sampling techniques with very low collection efficiency. The conference abstract did not specify which sampling method was used, further limiting interpretability. The EPDM study, for instance, used air filtration with microfiber filters that enabled high flow rates (~100 L/min) but featured much larger pore spaces between fibers, allowing smaller particles to pass through. This is reflected in their reported particle sizes of 20–100 µm. In contrast, the present study predominantly captured particles around 1 µm, including a substantial fraction of nanoplastic particles. The referenced studies conducted in public indoor environments similarly relied on high-flow but coarse-mesh filtration or on sedimentation-based sampling, which preferentially collects larger particles, while smaller particles—likely the majority—remain suspended in the air and are, consequently, underrepresented in the reported concentrations.

This comparison underscores how strongly the technical characteristics of aerosol sampling—such as flow rate, filter material, and mesh size—can influence exposure assessments. In both samples, the contribution of BTR particles to the overall MNP number concentration was 70–87%, which is higher compared to most environment-related proportions that were summarized in a recent review [39]. Indoor or workplace-specific particle number concentrations for BTR particles have not been reported, to the best of the authors’ knowledge.

3.1.4. Estimated Particle Mass Concentrations

Figure 2c,d shows pie charts of the estimated mass contributions. In this metric, “Not classified” particles contributed minimally due to their small mass, whereas classes with few but large particles—such as iron oxide and MNPs—dominated more in comparison to the particle number concentration reflection. Notably, total particle masses were similar for both samples. For time T1, the total mass aligned with gravimetric measurements of the polycarbonate filter (Table 1). This was not true for time T2, indicating unsuccessful filter weighing. More probable for heavier particles, deposited particles can become loose due to vibrations during transport of the sample. In case heavier particles were collected, measuring the deposited mass with a microgram scale would potentially yield false values compared to the initially deposited mass. Whether this happened with the T2 sample could not be assessed afterwards.

Figure 3a,b shows stacked bar charts of the total, PM₁₀, and PM_2.5 particle classes. PM₁₀ concentrations were reduced by ~50% compared to total mass concentrations for both time T1 (127.1 µg/m³) and for time T2 (120.2 µg/m³), whereas PM_2.5 masses were roughly an order of magnitude lower (T1: 11.4 µg/m³, T2: 11.0 µg/m³).

Black tire rubber remained a dominant class across all size fractions, whereas PET particles contributed substantially less to the PM₁₀ fraction compared with total mass. This indicates that PET particles released from onsite handling likely posed limited pulmonary exposure, whereas black tire rubber particles—part of the ubiquitous aerosol background—were the most prominent polymer-based exposure source. Their origin (indoor vs. urban traffic) remained uncertain.

Occupational exposure in workplace settings is regulated by national standards, which are based on 8 h average concentrations and typically allow higher dust levels than general ambient air quality guidelines. In Romania, where the study site was located, the occupational exposure limit for inert respirable dust—defined as particles with aerodynamic diameters below 15 µm—is 4 mg/m³ [65]. This value is about one order of magnitude higher than the total mass concentrations measured in the present study. Currently, no MNP-specific occupational exposure limit values exist worldwide. Based on the measured concentrations, no additional hygienic measures would, therefore, be required to protect workers from inhalable dust exposure.

For comparison, the WHO proposed 24 h average ambient air quality guideline values of 45 µg/m³ for PM₁₀ and 15 µg/m³ for PM_2.5 in 2021. Exceeding these thresholds increases the likelihood of adverse health effects associated with particulate matter exposure. Converted to an 8 h work shift, the equivalent values would be 135 µg/m³ (PM₁₀) and 45 µg/m³ (PM_2.5). In the present study, estimated PM₁₀ concentrations were close to this limit at both sampling times, whereas PM_2.5 concentrations were lower.

Mass concentrations of production-related PET particles were 4.9 µg/m³ (PM₁₀) and 0.2 µg/m³ for T1 and 3.9 µg/m³ (PM₁₀) and 0.2 µg/m³ (PM_2.5) for T2. Including BTR and other MNPs, mass concentrations were 11.5 µg/m³ (PM₁₀) and 1.3 µg/m³ (PM₁₀) for T1 and 15.2 µg/m³ (PM₁₀) and 1.6 µg/m³ (PM₁₀) for T2.

Similarly to the particle number concentrations, reported MNP-specific mass concentrations at occupational sites that could serve for comparison are sparse. The referenced workplace studies reported the abovementioned number concentrations, but conversion of mass concentration or fractionization into PM₁₀ and PM_2.5 levels was not attempted.

For BTR alone, comparable mass concentrations from environmental sampling sites can be found that were often much lower in open environment areas [66,67]. In an indoor parking lot, 0.61–0.73 μg/m³ for PM_2.5 was reported [68], which is similar to the results reported here.

3.2. Real-Time Measurements

Airborne particle concentrations measured by the OPC in each size channel were statistically processed using IBM SPSS Statistics 23. See Supplementary Material for a detailed description of the SPSS-based data processing. Figure 4 shows the mean size distributions for sample T1 (a) and T2 (b), based on 20 measurements.

At time T1, the particle population was dominated by the smallest fraction, with decreasing counts observed for progressively larger particle sizes. OPC-measured PM₁₀ mass concentration was ca. 24.5 µg/m³, and for PM_2.5 it was 9.7 µg/m³. At time T2, particle counts increased overall, with the size distribution still dominated by the smallest particles. PM₁₀ mass concentration was ca. 101 µg/m³, and for PM_2.5 it was 40 µg/m³.

Both OPC-based particle number and mass concentrations are, in general, much smaller compared to the values determined from CM analysis. This is most likely due to a low counting efficiency of the OPC in a mixed dust environment, which was attributed to various types of OPCs in a study by Plitzko et al. (2021) [69]. Particle properties, such as shape, refractive index, and color, are known to influence the intensity of back-scattered light. To relate the detected signal to the number of particles in the sampled air volume, optical instruments rely on internal conversions based on calibration aerosols, typically spherical particles with well-defined concentrations. The more real-world aerosols deviate from these calibration standards, the less accurate the resulting particle number concentrations become. Some instruments apply correction factors to compensate for such deviations, but their validity is highly site-dependent. Black particles, for example, produce weaker scattering signals. Moreover, when converting particle number concentration to mass concentration, instrument software commonly assumes a generalized particle density close to unity. In occupational environments where denser particles, such as mineral dust, contribute substantially, this leads to an underestimation of the actual mass concentration.

Notably, the proportion of the smallest particles (0.3 µm) decreased from 77% to 52%, while medium- and large-sized particles (1–10 µm) increased in both absolute number and relative proportion. However, this was not reflected in an increase in particle concentrations or even particle class fractional cardinalities determined by the CM analysis.

4. Conclusions

This study provides a detailed characterization of airborne micro- and nanoplastic particles in a textile production environment using a high-resolution correlative SEM–Raman approach. The results show that workplace MNP exposure is dominated by very small particles, largely around 1 µm and extending into the nanoplastics range. PET released during onsite handling was the main particle type by number, while black tire rubber (BTR) formed a substantial background contribution from the surrounding environment.

Despite these high particle counts, the corresponding mass concentrations—including PM₁₀ and PM_2.5—were far below national occupational exposure limits and close to, or below, WHO-equivalent guideline values. PET contributed minimally to the respirable mass fractions, indicating that the production process mainly emitted small, light particles. In contrast, BTR contributed meaningfully to both particle number and mass and represented the most prominent non-process-related MNP exposure at the site.

Comparison with previous workplace and indoor studies revealed that earlier reports likely underestimated MNP concentrations due to sampling techniques with low efficiency for small particles. High-flow microfiber filters and sedimentation-based collection preferentially capture larger particles, whereas the abundant sub-micrometer fraction remains airborne. The present sampling and analytical workflow overcomes these limitations and provides a more complete picture of size-resolved and substance-specific MNP exposure.

Real-time optical particle counting substantially underestimated particle number and mass concentrations compared with SEM–Raman results. This reflects the well-known limitations of optical counting in mixed aerosols, where particle shape, refractive index, and color strongly influence detection. These findings confirm that optical instruments alone cannot reliably quantify MNP exposure in complex workplace environments.

Overall, this study shows that (i) airborne MNP exposure in the investigated workplace is dominated by very small particles, (ii) number-based and mass-based exposure metrics can lead to different interpretations, and (iii) accurate assessment of MNP exposure requires high-efficiency sampling combined with chemical identification at the single-particle level. The results highlight the need for standardized, size-resolved MNP measurement methods and provide a basis for future consideration of MNP-specific occupational exposure assessment and potential guideline development.

However, this study served to test the analytical procedure, not to perform a detailed exposure assessment. For a comprehensive exposure assessment of MNPs and other particles, additional samples would be required, e.g., at other locations in the factory or outside to assess the ubiquitous background aerosol. Additionally, aerosol sampling should be applied in the personal breathing zone of workers for a personal exposure assessment, as described by the German statutory accident insurance [70]. In future workplace exposure assessments, combining a comprehensive assessment strategy with the analytical procedure presented here would enable us to produce the quantitative datasets about MNP release at workplaces needed to describe detailed exposure scenarios and link diagnosed adverse health effects to proven occupational exposure to MNPs.