1. Introduction
Industrial cleanrooms are specialized environments, designed to maintain high levels of cleanliness based on particle number concentrations, resolved using size, as classified by the International Standardization Organization (ISO) [
1]. In the semiconductor industry, high-level cleanrooms typically utilize laminar airflow combined with High-Efficiency Particulate Air filters (HEPA) to minimize airborne particles [
2]. This approach is used in both fabrication and sub-fabrication areas, with the goal of preventing potential defect formation in the final product. However, with the growing use of engineered nanomaterials (ENM) in production processes, such as in chemical mechanical planarization (CMP) slurries [
3,
4,
5], there is an emerging concern that ENMs may become aerosolized, leading to potential occupational exposure [
6], and release into the environment, especially during maintenance operations [
7].
Observational studies on aerosol concentration and composition are scarce in the scientific literature, as these environments tend to be, by definition, clean so present low concentrations, and only a few studies exist, especially on nanoparticles [
8,
9]. Shepard and Brenner observed a 7.6-fold increase in airborne nanoparticles, ranging from 11.5 to 115.5 nm in size, during a CMP tool set-up process [
10]. Subsequently, Brenner et al. conducted further assessments during CMP maintenance processes and reported a spike in airborne nanoparticle occurrence (~18,000 #/cm
3), while workers vacuumed the dried slurry [
8]. Most of the particles in CMP areas were characterized at a size over 100 nm and composed of silicon, aluminum, and cerium (suggesting material from the CMP slurry) [
8,
11,
12]. Other manufacturing processes like diffusion, wet etching, chemical vapor deposition (CVD), metallization, ion implantation, and dry etching have also been investigated by previous studies [
9,
13,
14]. Maintenance processes were identified as the main contributors to airborne particle release and exposure. Liao et al. reported spikes of 921,500 #/cm
3 and 647,000 #/cm
3 in airborne particle concentration during ion implanter and plasma enhanced chemical vapor deposition (PECVD) tool-maintenance processes, respectively, with all particles nanosized (<60 nm) [
9]. This also suggests the formation and release of incidental, process-generated nanoparticles and raises significant concern in terms of release and worker exposure during maintenance. However, due to the lack of characterization information, it is hard to evaluate the potential exposure risks.
Airborne-particle exposure in the semiconductor industry has been associated with reproductive risks, non-Hodgkin’s lymphoma, leukemia, brain tumors, and breast cancer [
15,
16]. The release of III-V material particles, including indium arsenide (InAs), gallium arsenide (GaAs), and Indium Tin Oxide (ITO) during semiconductor-manufacturing processes could induce cytotoxicity, impact the respiratory system, and damage the liver or kidney [
17,
18,
19]. The National Institute for Occupational Safety and Health (NIOSH) has established recommended exposure limits (RELs), while the Occupational Safety and Health Administration (OSHA) has set permissible exposure limits (PELs) for occupational exposure. The American Conference of Governmental Industrial Hygienists (ACGIH) has set threshold limit values (TLVs). In nearly all cases, the regulations are mass-based and do not differentiate between sizes of airborne particles or consider the unique toxicity considerations for nanoparticles due to their small size, high surface area, and unique physicochemical properties [
20,
21]. To date, only ultrafine titanium dioxide and carbon nanotubes have a size-resolved exposure limit [
22,
23]. However, increasing evidence that highlights the risks associated with nano-sized particle exposure, which may more easily cross biological barriers and translocate in the human body [
24,
25,
26,
27], raises concern. Finally, it is noteworthy that while the international organization for standardization (ISO) has established cleanroom classifications that are based on size-resolved-particle number concentration, as outlined in ISO 14644-1 [
1], nanoparticles (smaller than 100 nm) are not currently considered.
This study aims to quantify the occurrence of nanoparticles in semiconductor-type cleanroom facilities through in situ online observation and filter-based offline characterization methods. We measure the release and resulting exposure to airborne nanoparticles in various semiconductor-manufacturing processes, during routine and preventive maintenance operations. This study provides detailed characterization results, including particle number size distribution, morphology, chemical composition, and elemental mass concentration. By enhancing the characterization of airborne nanoparticles in semiconductor facilities, this study hopes to inform the management of occupational exposure risks associated with airborne nanoparticles.
3. Results and Discussion
3.1. Particle Number Concentration
Table 1 summarizes the observations on airborne particle number concentration during preventive maintenance activities, normal operational processes, and nocturnal background measurements. Particle number concentrations are reported in units of particles per cubic centimeter (#/cm
3), as measured by SMPS using the electrical mobility diameter within the range of 6–220 nm.
The airborne nanoparticle concentration during routine operation were presented as 3 h average values measured on a normal operation day, specifically over 9:00–12:00 and 21:00–24:00, respectively.
Table 1 shows that the nocturnal background particle concentrations vary among different sampling locations. The ASU Nanofab area showed concentrations ranging from 7 to 15 #/cm
3, while the MTW fab area recorded consistent concentration ranging from 2 to 14 #/cm
3. The MTW return air areas exhibited higher particle concentrations, ranging from 2 to 89 #/cm
3.
On a routine operation day, the ASU Nanofab exhibited a significant increase in particle number concentrations for both PECVD and wet etching areas. The concentrations rose from 7 #/cm
3 and 15 #/cm
3 to 313 #/cm
3 and 201 #/cm
3, respectively, during working hours. This increase is associated with heightened occupancy during these hours due to research and training activities. In contrast, MTW did not exhibit this trend, reflecting lower occupancy levels as it operates continuously with a limited operator presence in the cleanroom. However, these conditions change during maintenance operations. In the present study, sampling was performed during preventive maintenance activities, specifically when technicians vent and access the vacuum chambers to clean the chamber surfaces using humidified fabrics before resealing and reestablishing a vacuum. The entire process typically lasts for 30 min, unless unexpected issues arise (
Table 1).
Preventive maintenance significantly affected the occurrence of particles in the environment of some tools. Both PECVD tool areas in Nanofab and in MTW experienced significant airborne-particle increases (618 #/cm
3 and 375 #/cm
3 from 313 #/cm
3 and 22 #/cm
3, respectively). During the MTW PECVD tool-maintenance processes, a spike in particle counts of up to 4120 #/cm
3 was detected, two orders of magnitude higher than during routine operation. This is consistent with a previous report by Liao et al. who observed an even higher spike (647,000 #/cm
3) during PECVD tool maintenance [
9].
Etching tools exhibited different results, based on the specific instruments sampled. Notably during chlorine etching and dry-etching-tool-maintenance processes, airborne nanoparticle concentrations increased with concentrations rising from 16 #/cm
3 and 139 #/cm
3 to 523 #/cm
3 and 260 #/cm
3, respectively, consistent with previous reports by Choi et al. [
13], indicating that maintenance processes can elevate particle concentrations 2–4 times higher than during normal operation processes for dry-etching tools. In contrast, no significant increase in particle number concentration was observed during maintenance processes for fluorine and plasma etching.
Our observational study is the first one to also investigate sputtering tools and plasma asher environments during maintenance procedures. Both sputtering tool areas tested exhibited increased airborne particle concentrations during maintenance (from 10 #/cm3 and 15 #/cm3 to 718 #/cm3 and 227 #/cm3), while concentrations in the vicinity of the plasma asher remained consistently low (from 2 #/cm3 to 7 #/cm3).
3.2. Particle Concentration Temporal Change during Maintenance
Figure 3 displays the evolution over time of particle concentrations during maintenance processes for the PECVD (II) and sputtering tools (I) located in MTW. The heat map highlights the two specific events with the highest maximum spike (
Figure 3a) and highest maintenance average (
Figure 3b) in terms of particle number concentration, the color dots on the heat map indicate the size-resolved-particle number concentration at specific moment. The light-purple background means no particles have been detected during the monitoring processes.
In
Figure 3a the maintenance day on the PECVD tool is shown. The SMPS recording started at 14:13, and the PECVD vacuum chamber was opened at 14:20. A spike in particle concentration (4120 #/cm
3) was detected using the SMPS at the same time as the chamber gate was opened, suggesting the release of particles from inside the vacuum chamber or potentially the generation of new particles upon contact with ambient air [
14]. The strong laminar air flow within the cleanroom effectively removed the particles, resulting in a rapid decrease (4120 #/cm
3 to 225 #/cm
3 in next SMPS scan). Overall, throughout the entire PECVD tool-maintenance processes only a few particle spikes were recorded resulting in an average particle number concentration of 321 #/cm
3, which is significantly higher than the background concentration during routine operation (15 #/cm
3).
Figure 3b shows the particle concentration observations during sputtering-tool (I) maintenance, which presented the highest average concentration of airborne particles (718 #/cm
3) and a peak value of 3440 #/cm
3. In contrast to the observations for the PECVD (II) tool, the airborne nanoparticle concentration during sputtering (I)-tool maintenance remained consistently high throughout the entire period with less-significant peaks. After maintenance, the airborne nanoparticle levels rapidly decreased to ~166 #/cm
3 and eventually back to the background level (8 #/cm
3) at nighttime. The sputtering tool exhibited a higher number of particles, measuring over 100 nm compared to the PECVD tools; additional discussion is provided in the next section. For reference, heat maps for all tool-maintenance activities are available in
Figure S1.
3.3. Particle Size Distribution for Airborne Particles
Based on the heat maps in
Figure 3 and
Figure S1, most particles detected during maintenance events were smaller than 30 nm. To highlight the abundance of small nanoparticles, the total airborne nanoparticle counts were categorized into three size ranges (6–30 nm, 30–100 nm, and 100–220 nm) and summarized in
Figure 4.
In
Figure 4, smaller sized nanoparticles (N
6–30) dominate the observed particles during maintenance, comprising over 80% of particles for each sampled tool. This aligns with previous studies reported by Liao et al. [
9], where the particles in the size range of 9–20 nm was the dominant species released during PECVD tool-maintenance processes. The overall number of nanoparticles observed during maintenance activities did not seem to influence the size distribution. For example, despite the highest average count of sputtering (I) (718 #/cm
3) compared to the sputtering tool (II) (260 #/cm
3), the relative importance of the N
6–30 fraction was similar for both tools at 84% and 88%, respectively. Overall, the PECVD tools, in our study, released more N
6–30 (93%, 97%) compared to etching (82%, 89%) and sputtering tools (84%, 88%), which have relatively higher number concentrations of larger particles (N
30–100 and N
100–220). This finding may be attributed to differences in particle formation mechanisms for each type of manufacturing tool. In the next section, we will investigate this further by looking at the morphology and chemistry of the particles.
3.4. Morphology and Elemental Composition of Collected Airborne Particles within Breathing-Zone Filter Sample
Based on the particle size distribution depicted in
Figure 4, the majority of particles observed during maintenance processes were smaller than 30 nm. STEM analysis of breathing-zone filter samples, as shown in
Figure 5a, confirmed the presence of these small particles; a metallic copper nanoparticle (Oxygen signal not enhanced) with a size around 10 nm was identified in the PECVD sample. Larger particles (30–220 nm) were also observed in this PECVD sample.
Figure 5b shows a 50 nm oxidized silicon particle with trace aluminum mixed in (EDX spectra available in
Figure S2), which may have originated from silica nanoparticles generated during the PECVD operation process for thin film, as indicated using the reaction below.
Furthermore, particles exceeding the measurement range of SMPS (>220 nm) were observed.
Figure 5c illustrates a nickel particle with a size over 1 µm, presumably formed by the agglomeration of smaller nanoparticles and predominantly composed of metallic nickel (91 wt% Ni, EDX spectra available in
Figure S3) [
34].
To confirm the consistency of these findings, breathing-zone samples collected during the maintenance process of the sputtering tool were also analyzed.
Figures S4–S6 display particles smaller than 30 nm, predominantly composed of transition metals such as copper, iron, and nickel. Additionally,
Figure S7 exhibits a larger particle of approximately 100 nm in size, composed of oxidized indium. This particle is process-related, as indium-tin-oxide was employed as the sputtering target. Larger particles exceeding 1 µm in size were also detected, but they did not exhibit agglomeration characteristics (
Figure S8). These particles were primarily composed of process-related elements from the sputtering target, including indium, aluminum, titanium, and zirconium.
In our observations, smaller nanoparticles (<30 nm) predominantly consisted of transition metals. This finding is interesting, considering that transition metals, especially copper, are rarely used in semiconductor products. Further investigation is necessary to determine the sources of these nanoparticles. In contrast, larger nano-particles (>30 nm) are likely to be associated with the manufacturing processes and can be related to process activities. The occasional very large particles (>1 µm), also process related, are significantly contributing to the particle mass. Assessing exposure risk in compliance with current regulations emphasizes the mass concentration over particle number counts.
3.5. Mass Concentrations in Breathing Zone Area
The previous sections highlighted the significance of nanoparticles in terms of number concentrations. However, samples collected on filters in the devices worn near the breathing zone by workers also contained larger particles that fall outside the observation range of the SMPS. These larger particles are of concern for occupational exposure, particularly for metals like indium, as they can contribute to significant mass concentrations. To assess the potential risk, we performed bulk analysis and compared the total mass concentration of breathing-zone samples with established exposure limits set by organizations such as NIOSH, OSHA, ACGIH, and CAL/OSHA. Due to the short duration of our sampling periods, we normalized the mass concentration to an eight-hour time-weighted average (8 h-TWA), which is a standard unit in regulations and allow us to make the comparison to assess whether exposure exceeds the current limit. The equation used for the normalization is shown below, where C
m is the total air mass concentration of specific elements and t
m is the duration of maintenance activity, 480 min represents the eight-hour working time.
Extrapolating the exposure to an eight-hour period may not be unreasonable, considering that workers may move from one tool maintenance to another. However, it’s important to note that this assumption is made. Additionally, we assume all the particles collected in this study are inhalable since most of the particles collected in this study are nanoparticles.
Figure 6 shows the 8 h-TWA mass concentration of elements in the breathing-zone samples of the operators during maintenance. Silicon and aluminum are the most abundant elements, as expected due to the processing of silicon materials in each tool and the ubiquity of aluminum as a tool component material. Aluminum is also used as a target in the sputtering (I) process which resulting in yield the highest concentration (~2 µg/m
3) and at least four times higher than others. Transition metals such as zinc, iron, nickel, copper, and chromium are consistently present in the breathing zone but at lower concentrations. It is noteworthy that these are research and pilot facilities which will change targets and use at times experimental materials, less common in routine production. Indium is only observed in the sample collected from sputtering I tool, which is consistent with the use of an indium-tin-oxide target. These results are consistent with the STEM analysis.
Table S2 summarizes current workplace limits for selecting elements across different regulatory agencies [
35]. These limits tend to be, for most species, in the milligram or hundreds of micrograms range, whereas our results are mostly in the nanogram range (
Figure 6), and hence, they are several orders of magnitude lower than the exposure limits. For example, the highest mass concentration of aluminum in the sputtering (I) sample was ~2 µg/m
3, which is thousands of times lower than the minimum respirable exposure level regulated by ACGIH (1000 µg/m
3). The highest measured nickel concentration in our samples was ~0.3 µg/m
3 from the PECVD tool, which is the closest one to a limit value (15 µg/m
3) but still 50-fold lower [
35]. Although none of the compounds exceeded the exposure limits in our study, it is important to note that current limitations are primarily mass-based, while the toxicity of nanoparticles may differ from bulk material. Additionally, the airborne nanoparticle exposure may vary depending on factors such as the instrument type, working load, maintenance frequency, and engineering control applications. Conducting measurements in different fabs would contribute to a better understanding of overall airborne nanoparticle occurrence in semiconductor-manufacturing environments.
3.6. Study Limitations
This study aimed to investigate the release of nanoparticles and exposure to nanoparticles in semiconductor environments. A variety of tools were tested in two different facilities. This is, of course, a limited set of observations, but they were challenging to overcome for practical reasons. It is clear that there will likely be significant variability between facilities and between tools. The facilities and tools tested here were modern and at least pilot scale tools, but older as well as newer technologies and processes exist. This will lead to different emissions. Finally, even for the same tool, different processes and engineering control implementations could yield different particle formations and exposures, both in terms of nature and amounts. This cannot all be captured in a single study but would require a larger community effort.
4. Conclusions
This study investigated the presence of airborne nanoparticles in semiconductor-manufacturing environments. Through in situ observational studies conducted within two ISO level 5 cleanrooms, we found that maintenance activities led to the release and occurrence of airborne nanoparticles. These nanoparticles’ concentrations increase by one to two orders of magnitude during tool maintenance and typically continued until the end of the maintenance processes. Notable concentration spikes were observed in certain instances, particularly at the opening of vacuum chambers, in a sudden surge in particle number concentration by two or more orders of magnitude. In most cases, a substantial fraction, accounting for over 80% of the observed particles, had a size smaller than 30 nm.
To further characterize the observed particles, filter-based sampling (within worker’s breathing zone) and subsequent instrumental analyses including STEM and ICP-MS were used. STEM imaging demonstrated that the tiniest particles (less than 30 nm) were mainly composed of transition metals such as copper, nickel, and iron, with some particles appearing in metallic form. On the other hand, larger particles (greater than 30 nm) were associated with manufacturing materials. Additionally, a few particles larger than 1 µm were also detected. The study also analyzed the mass concentration of each element present in the filter samples and compared them to the regulations set by government agencies. Encouragingly, none of the elements exceeded the regulatory levels, with concentrations at least 50-fold lower than the limits specified by the regulations.
Although the elemental mass concentration results do not exceed exposure limits, one has to consider that most released particles are less than 30 nm in size, for which we are lacking toxicity studies that address their potentially differential toxicity from larger particles. The results clearly identify a need to further investigate the potential health effects of nanosized particles. In the meantime, it might be appropriate to consider engineering controls during maintenance operations to limit the exposure of operators and the release of nanoparticles to the clean room environment.