Comparison of Nanoparticle Exposure Levels Based on Facility Type — Small-Scale Laboratories , Large-Scale Manufacturing Workplaces , and Unintended Nanoparticle-Emitting Workplaces

The aims of this study were to investigate the concentrations and characteristics of nanoparticle exposure at various workplaces. We compared the concentration and characteristics of nanoparticles at nine workplaces of three types; i.e., small laboratories (LAB), large-scale engineered nanoparticle manufacturing workplaces (ENP), and unintended nanoparticle-emitting workplaces (UNP), using real-time monitoring devices including scanning mobility particle sizers (SMPS), condensation particle counters (CPC), surface area monitors (SAM), and gravimetric sampling. ANOVA and Scheffe’s post hoc tests were performed to compare the concentration based on the type of workplace. The concentrations at UNPs were higher than those at other types of workplace for all measured metrics followed by (in order) ENP manufacturing workplaces and LAB (p < 0.01). Geometric means and geometric standard deviations of LAB, ENP, and UNP for total number concentration measured using SMPS were 8,458 (1.41), 19,612 (2.18), and 84,172 (2.80) particles cm, respectively. For CPC, the concentrations were 6,143 (1.45), 11,955 (2.42), and 38,886 (2.61) particles cm, respectively. The surface area concentrations were 32.79 (1.46), 93.68 (2.60), and 358.41 (2.74) μm cm, respectively. The characteristics of exposure and size distributions differed among the workplaces. Some tasks or processes at LAB exhibited higher concentrations than those at ENP or UNP workplaces, and LAB showed the lowest concentration. In conclusion, we observed different exposure characteristics at LAB, ENP, and UNP suggesting that different risk management strategies are required.


INTRODUCTION
Three types of nanoparticle source are classified as: naturally occurring (e.g., volcano ash and forest fires), unintended emission of nanoparticles (UNP) (e.g., welding, smelting, and diesel exhaust), and engineered nanoparticles (ENP) (Oberdörster et al., 2005).Recently, the application of engineered nanoparticles has increased rapidly in various industries.Accordingly, the number of workers involved with nanotechnology is increasing, and is estimated to be approximately 2 million worldwide in 2023 (Schulte et al., 2008).
Based on the probable health hazards of nanoparticles and exposure concerns via inhalation and skin absorption, risk management including traditional control strategies in industrial hygiene; i.e., elimination, substitution, isolation, engineering control, administrative control, and personal equipment, have been reported.Moreover, exposure levels may differ throughout the life cycle of nanoparticles due to handling size; i.e., research, development, and production/ manufacturing (Schulte et al., 2008).To implement appropriate strategies, it is important to examine the exposure level.However, data on nanoparticle exposure assessment at laboratories and ENP manufacturing workplaces are insufficient.
Workers in workplaces such as ENP manufacturing, UNP workplaces, and laboratories (LAB) are exposed to nanoparticles.For example, the exposure levels of workers at ENP manufacturing workplaces varied according to the task performed; e.g., handling, use, research, and development.
Levels of UNPs such as welding fumes and diesel exhaust were reported to be high (Zimmer, 2002;Ham et al., 2012;Lehnert et al., 2012).The exposure levels of university students and researchers in the laboratory varied according to the process type and amount of nanoparticles used during experiments (Demou et al., 2009;Lehnert et al., 2012).
To our knowledge, no study has compared simultaneously nanoparticle exposure at three workplace types: small-scale laboratories (LAB), engineered nanoparticle manufacturing workplaces (ENP), and unintended nanoparticle-emitting workplaces (UNP).It is important to compare the nanoparticle concentration and characteristics based on workplace type to ensure that the appropriate exposure assessments and management strategies are applied.Researchers in a laboratory could be exposed irregularly due to intermittent and short handling durations during experiments and the development process.At UNP workplaces, workers might be exposed continually during the production process due to repetition of the same task over a long period.Nanoparticle exposure to workers at ENP workplaces may be variable according to the type and amount of nanoparticle, and handling tasks and processes performed.Thus, comparison of the exposure level at LAB, ENP, and UNP is important to implement appropriate control strategies.The aims of this study were to investigate the concentrations and characteristics of nanoparticle exposure at LAB, ENP, and UNP workplaces.

Sampling Site Information
We classified three types of workplace: laboratory (LAB), engineered nanoparticle (ENP) workplace, and UNP-emitting workplaces (Table 1).Measurements at each type of workplace were performed at three locations.Therefore, a total of nine workplaces were investigated.Three laboratories at the university were investigated.LAB-A was an earth environment laboratory, and the primary nanoparticle was Al 2 O 3 .Two workers performed experiments of transfer to the crucible, transfer from the crucible to a vial, and weighing.LAB-B was involved with development of new materials, with the primary nanoparticles used being Fe 2 O 3 and TiO 2 .Primary experiments were weighing, sonication, and reaction.Seven workers performed the experiments.LAB-C dealt with graphene for space aviation.DIP-coating processes to fabricate graphene were the primary experiments performed; together with spraying the base of the DIP coater for cleaning by five workers.A general ventilation system and a fume hood were installed in all laboratories.
Three ENP manufacturing workplaces participated in this study.ENP-D manufactured Ti and Zn powder for cosmetic sunscreen.Reaction, dehydration, mixing, drying, and bagging were the main processes at ENP-D.The reaction was performed at 120°C and 3 atm, and dehydration was applied at 60°C.There was a general ventilation system without local exhaust ventilation (LEV).The production rate at ENP-D was 10 tons for TiO 2 and 50 tons for ZnO per year.ENP-E dealt with metallic nanopowders such as copper, nickel, and silver.Nanopowders were produced using the high-voltage pulsed-wire evaporation (PWE) method, and the main products were copper-nickel alloy and nickel nanopowders for use as additives in automobile engines.The main processes were collecting and sieving.Manufacturing equipment for PWE was isolated in a cabinet equipped with a LEV system.Production rates for Ni and Cu-Ni alloy were 100 kg per year.During the manufacturing process, the glass door was closed.Amorphous silica was manufactured at ENP-F and all processes were automated.Amorphous silica was used for abrasive materials that were applied in the chemical mechanical polishing (CMP) process in the semiconductor industry.The main process was packaging of a 10-kg bag.A total of 9,000 tons of amorphous silica was manufactured annually at ENP-F.A general ventilation system and LEV were installed during the bagging process.Forklifts were used at ENP-D and F.
Three UNP-emitting workplaces were sampled.UNP-G manufactured heat exchangers and steel structures, such as the H-beam; arc welding and SUS welding were the main processes and a total of 100 workers performed welding.The main product of UNP-H was bodyframes for the back hoe and forklift.Welding and grinding were the primary processes at UNP-H.Sampling was performed at the arc welding process on the first day and at the SUS welding process on the second day.A total of 30 workers were active.UNP-I manufactured automobile engine parts, which involved smelting and welding.Sampling was performed during the smelting process on the first day and during the welding process on the second day.There were two shifts with 15 workers per shift, and three welders worked on one shift.A general ventilation system but no LEV was installed at the UNP workplaces.Forklifts were operated at all UNP workplaces.

Sampling and Analysis
To determine the distribution of particle sizes, an SMPS (Nanoscan, Model 3910, TSI Inc., USA) with a detectable size range from 10 to 420 nm and a concentration range of 0 to 10 6 particles cm -3 was used.The inlet flow rate was 0.75 L min -1 and the sample flow rate was 0.25 L min -1 .A flow check was performed in the laboratory prior to taking measurements.The sampling time of the SMPS was one minute per averaging time, and particle sizes from small to large were measured using 13 sequence channels.A cyclone was used to remove the larger particles.The particles were collected from the inlet and passed through the aerosol neutralizer using a unipolar charger.Particles were separated using a mobility diameter with a radial differential mobility analyzer (RDMA) before being counted using an isopropanolbased CPC.An external isopropyl reservoir was used.
A CPC (P-Trak Model 8525, TSI Inc., USA) was used to measure the number concentration of particles of 20 to 1,000-nm diameter with a 0.1 L min -1 sample flow rate.The capable concentration range was 0-500,000 particles cm -3 .Isopropyl alcohol (Sigma Aldrich, USA) was used for particle condensation, which increases the particle size to enable optical detection.One sampling averaging time was  set as 1 min.The zero calibration was performed using a HEPA filter before sampling.
To measure the surface area concentration, a surface area monitor (SAM) (AeroTrak Model 9000, TSI Inc., USA) was used.One averaging time was set as 1 min.The measuring particle size range was 10-1,000 nm and the aerosol concentration ranged from 1-10,000 µm 2 cm -3 for the alveolar deposition method.A cyclone was installed on the inlet to prevent the entry of particles larger than 1 µm.The flow rate of the aerosol sample branch was 1.5 L min -1 .Zero calibration of the electrometer was performed before sampling.The sampled particles were mixed with ions in the chamber of the device and the particles were charged.The charged particles were passed along the electrometer, and the charges of the ions were measured and converted to surface area metrics.
Integrated sampling using filter media was performed for gravimetric analysis.Sampler (2 L min -1 , Escort ELF, MSA, USA) with an open-faced three-piece cassette was used to capture airborne nanoparticles.Sampling was performed during a full working shift.Filters were pre-and postequilibrated before weighing in an environmentally controlled weighing room that was maintained at a temperature of 20°C ± 1°C and a relative humidity of 50% ± 5%.After sampling, the cassettes were sealed tightly using silicon tape and transported in a clean box.Weighing was performed using a microbalance.A transmission electron microscope (TEM) (JEM-3010, JEOL, Japan) grid (Q225-CR1, 200 mesh copper, EMS, USA) was used to analyze particle sizes and morphologies.
The height of the inlets ranged between 1.2 and 1.5 m for all measuring devices.In this study, researchers observed and recorded the TAD for all tasks during the sampling time, except during off-duty periods.

Statistical Analysis
Descriptive statistics was performed to show the concentration level.Analysis of variance (ANOVA) tests were performed to compare both workplace types and individual workplaces of the same type.Scheffe's post hoc analysis was performed to identify significant differences between means.Data analysis was performed using SPSS (version 20.0, IBM, USA).

RESULTS
Table 1 summarizes the general characteristics of workplaces investigated based on the type of workplace, emitted or source of nanoparticles, ventilation type, processes or tasks, size of workplaces, number of workers, and other possible sources.Two LABs were dealing with metal and one LAB handled graphene.Two ENPs manufactured metal nanoparticles and one ENP manufactured fumed silica.All UNP workplaces performed welding, and UNP-I also undertook smelting processes.
All LABs had general ventilation (GV) and a fume hood.For ENP workplaces, ENP-D had GV.Local exhaust ventilation (LEV) and isolated cabinets for facilities were installed at ENP-E.GV and LEV were installed at the ENP-F workplace.Only GV was installed at the UNP workplaces in this study.
There were two, seven, and five workers at the LAB workplaces.Six to twelve workers were engaged at the ENP workplaces.One hundred, thirty, and three welders were employed at the UNP-G, UNP-H, and UNP-I workplaces, respectively.At the UNP-I workplace a total of 30 workers participated in two shifts, each of which comprised 15 workers.
Table 2 shows descriptive statistics of workplaces measured using SMPS, CPC, SAM, and gravimetric sampling.In addition, it shows the concentration levels between groups (LAB, ENP, and UNP) and homogeneous subsets between workplaces.
The geometric mean for LAB was lower concentration than those of ENP and UNP workplaces for all metrics.UNP workplaces were highest concentration among the workplaces (Table 2).Table 2 shows the differences in nanoparticle exposure levels among LAB, ENP, and UNP workplaces.The total number concentrations, number concentrations below 100 nm, CPC, and SAM differed significantly among LAB, ENP, and UNP workplaces (p < 0.01).
Homogenous workplaces in terms of exposure level were observed in Table 3.For example, LAB-A, LAB-B, LAB-C, ENP-E, and ENP-F showed similar low exposure levels in terms of the total number concentration, particles greater than 100 nm in diameter, and SAM (p < 0.01).The concentrations of particles less than or equal to 100 nm in diameter were similar among LAB-A, LAB-B, and LAB-C, and ENP-E and ENP-F (p < 0.01).LAB-A, LAB-B, and ENP-E were categorized as homogeneous in terms of number concentration measured based on CPC (p < 0.01).
Exposure levels within UNP workplaces varied markedly compared to the LAB and ENP workplaces.GSD of UNP ranged from 2.61 to 3.23, while it was less than 1.49 in LAB workplaces and 2.00-2.60 in ENP workplaces, regardless of the exposure metric.In addition, UNP workplaces were classified as non-homogenous with other workplaces (Table 3).For example, the exposure level at each UNP workplace (A, B, and C) was non-homogenous in terms of the total number concentration and surface area (SAM).
The total number concentrations by task or process for each workplace as determined using SMPS are shown in Fig. 1 for LAB (a), ENP (b), and UNP (c).The concentrations varied according to the task or process.In general, tasks such as welding or smelting in UNP workplaces emitted a higher concentration of nanoparticles compared to the LAB and ENP workplaces.Although the average concentration in LAB was lower than those at ENP and UNP workplaces (Table 1), some tasks-such as sonication and reactions at LAB-B, spraying, and DIP-coating at LAB-C-resulted in high levels of exposure.Fume hoods had been installed and were in operation at all LABs but all experiments we sampled were performed at the table outside of the fume hoods.Background (BG) concentration at each workplace was selected to compare processes and tasks with this BG concentration during measurement.For BG concentration determination, low and stable concentration which is not relevant to the ENP particles should be considered.In this study, off-duty time was selected at the all LABs and ENP-F while, lunch times were selected at ENP-D, ENP-E, UNP-G, and UNP-I.Compared to the BG, processes and tasks were higher.In addition, BG levels differed among the workplaces.Unlike LAB and ENP workplaces, the concentrations during a break or lunch at UNP workplaces were relatively high because tasks such as welding or smelting emitted a high concentration of particles, and GV in the absence of LEV did not result in exhausting of particles.
Fig. 2 shows the nanoparticle concentration based on size distribution at the workplaces.Figs.2(a), 2(b), and 2(c) show the concentrations at LAB.The majority of concentrations at LAB did not exceed 4,000 particles cm -3 for each size channel.However, sonication and reaction processes at LAB-B were associated with concentrations of over 6,000 particles cm -3 of ~100-nm diameter (Fig. 2(b)).Size distributions for ENP workplaces are shown in Figs.2(d), 2(e), and 2(f).The dehydration process at ENP-D emitted over 6,000 particles cm -3 of 20 to 50-nm diameter.In addition, the reaction and mixing process at ENP-D emitted slightly over 4,000 particles cm -3 (Fig. 2(d)).
We next examined UNP workplaces that exhibited emissions of over 10,000 particles cm -3 during major processes such as welding and smelting (excluding SUS welding) at UNP-G (Fig. 2(g)).Size distributions and number concentrations varied widely among workplaces.The majority of processes-such as transferring from a crucible to a vial at LAB-A, dehydration process for ENP-D, and the smelting process at UNP-I-showed a bimodal distribution (Fig. 2).
Fig. 3 shows the geometric mean of total number concentration (particle cm -3 ) measured using SMPS, together with the percentage of particles of less and greater than 100 nm diameter.The percentage of nanoparticles less than or equal to 100-nm diameter was over 65% at all workplaces, regardless of type of workplace.
Fig. 4 shows TEM images of particles collected, including the primary particle size, shape, and agglomeration of particles.Manufactured primary particles at ENP-D were about 100 nm (a) in diameter.The primary particle size at ENP-F was ~7-40 nm, and those particles were agglomerated (b).At UNP-G, chain-like particles were detected (c), (d).

DISCUSSION
This study compared the concentrations and characteristics of nanoparticles at nine workplaces categorized as three types (LAB, ENP, and UNP) to examine nanoparticle exposure based on real-time data and gravimetric sampling.A major challenge in comparing nano-related workplaces of various sizes is implementation of control strategies (Schulte et al., 2008).No previous reports have compared three types of workplace simultaneously.
The exposure levels at LAB differed from those at ENP and UNP workplaces.Although the concentration levels at LAB was low (Table 2), exposure levels for some tasks could be high (Fig. 1).For example, two tasks emitted high concentrations of nanoparticles; namely, sonication and reaction (Fig. 1(a), Fig. 2(b)) at LAB-B.These showed similar concentration levels to those at dehydration processes ENP-D and were higher than those of mixing, drying, and bagging (Fig. 1(b), Fig. 2(d)).Characteristics of LAB were: (1) intermittent performance of experiments, (2) the possibility of exposure to high nanoparticle concentrations in a short time period, (3) the existence of many unknown risks, and (4) difficulty in determining the exposure history due to the lack of an exposure assessment record.Engineered nanoparticle workplaces (ENP) showed higher concentrations than LABs and lower concentrations than UNP workplaces.The packaging process showed the highest nanoparticle concentration at ENP-F.Unintended nanoparticle release occurred at ENP-D and F because a forklift was used to transport pallets of product to the truck or warehouse.UNP workplaces showed the highest concentration for all measured metrics, which differed significantly compared to those at LAB and ENP.Welding is a well-known source of nanoparticles, and welding fumes could lead to adverse health effects (Yoon et al., 2003;Donaldson et al., 2005).Smelting showed the highest nanoparticle concentration (CPC: 50,826 particles cm -3 ) in this study.In previous reports, aluminum smelting processes to manufacture battery parts showed a mean concentration from 70,000 to 144,000 particles cm -3 as determined using the same CPC device as in our study.The study reported higher concentrations than our results (Table 2) (Debia et al., 2012).Exposure assessment and control strategy based on time-weighted average (TWA) approaches may not be appropriate, especially at LAB, because it may not reflect variation between tasks.Therefore, task-based exposure assessment strategies may be a feasible method of evaluating nanoparticle exposure (Peters et al., 2008;Cena and Peters, 2011;Ham et al., 2012).
According to the hierarchy of controls for general industrial hygiene (Halperin, 1996) and nanoparticle risk management (Schulte et al., 2008), elimination, substitution, isolation, engineering controls, administrative controls, and personal protective equipment are required for control measures.However, tasks were not performed in the fume hood.At LAB-B, sonication, reactions, and weighing processes were performed outside of the fume hood.This might have increased the nanoparticle concentration in the air (Fig. 1(a), Fig. 2(b)).In addition, transfer of nanoparticles from a crucible to a vial, to a crucible, and weighing were performed outside of the fume hood.This might have increased the concentration, as shown in Fig. 1(a) and Fig. 2(a).In a previous report, it was shown that laboratory nanoparticle experiments should be performed inside the fume hood using the smallest quantity possible and the least energetic handling (Tsai et al., 2009).Thus, it is important to handle nanoparticles in a well-controlled environment such as an LEV or fume hood.At the ENP-E workplaces, an isolation cabinet equipped with an LEV system to exhaust the emitted nanoparticles from the facility during the collecting process.In addition, there was an LEV system was installed for the sieving process (Fig. 1(b)).Isolation and engineering control measures, such as LEV systems, could mitigate exposure to nanoparticles.At the ENP-F workplace an LEV system was involved in the packing process.At the ENP-F workplace all processes were automated, with the exception of the packaging process.Therefore, there are fewer possibilities for exposure to nanoparticles during regular processes, excluding the packaging process.No control measures were installed at the ENP-D workplace because general ventilation, which opens the entrance door of factory building, was present.The nanoparticle concentration was higher for the reaction, dehydration, and mixing processes than the other processes.There was no LEV at UNP workplaces, and welding and smelting processes generated higher nanoparticle concentrations than those at the LAB and ENP workplaces.Welding fumes generated by stainless steel welding, which contains hexavalent chromium, is known to cause lung cancer (Yoon et al., 2003), and manganese is a cause of manganese-induced parkinsonism (Olanow, 2004).
In addition, respiratory symptoms such as asthma and lung function decline have been reported in the smelting industries (Søyseth et al., 2013).In a previous report, efficient use of LEV was shown to lead to a significant reduction in exposure to welding fumes (Ashby, 2002;Lehnert et al., 2012).In addition, portable LEV is a feasible control method for welding because welding process is not always performed at fixed point (Meeker et al., 2007).At the UNP-I workplace a lid with a ventilation system was installed for smelting processes, but its purpose was to exhaust heat from the factory and it does not have sufficient capacity to capture the nanoparticles emitted during the smelting process.Therefore, this process emitted a high concentration of nanoparticles.Thus, we suggest that an LEV system should be installed at UNP workplaces.Engineering controls such as facilities, process design changes, and ventilation could be effective for reducing nanoparticle exposure.Control measures, such as isolation of the process or worker, should be applied during the initial design prior to construction of the production facilities, because modifying the facilities thereafter is difficult.LEV systems are effective for capturing airborne nanoparticles.According to Schulte et al. (2008), 1-300-nm particles and fine dusts can be captured by diffusion using the LEV system.However, microscale particles are moved by inertia, which results in them crossing streamlines of moving air and so avoiding capture.
Overall, LEV systems could be appropriate control measures to reduce nanoparticle concentrations.
We applied SMPS to investigate the size distribution of collected particles.SAM, CPC, and gravimetric sampling are incapable of obtaining information on size distribution.Size distribution based on the number concentration was varied depending on the situation.A bimodal size distribution according to nanoparticle concentration was observed during this investigation (Fig. 2), similar to previous reports (Zimmer et al., 2002;Dasch and D'Arcy, 2008;Pfefferkorn et al., 2010).As shown in Figs.2(g), 2(h), and 2(i), welding processes showed a bimodal nanoparticle size distribution.Two peaks occurred at 5,366 and 18,730 particles cm -3 at 27 and 116 nm, respectively, at UNP-G (Fig. 2(g)).Smelting processes also showed a bimodal nanoparticle size distribution at the UNP-I workplace (Fig. 2(i)).The concentrations at 15 and 37 nm were 21,698 and 27,398 particles cm -3 , respectively, during the smelting process at the UNP-I workplace.In contrast, the nanoparticle size distribution was unimodal during the welding process at the UNP-H workplace (Fig. 2(h)).A unimodal size distribution was also reported for gas metal arc welding processes (Zhang et al., 2013).In addition, bimodal size distributions were detected during some tasks at LAB and ENP.The transfer of nanoparticles from a crucible to a vial at LAB-A, sonication at LAB-B and spraying at LAB-C showed a bimodal distribution.In addition, the majority of processes and tasks showed bimodal distributions.
The contents of particles less than or equal to 100-nm diameter was higher than that of those greater than 100 nm (Fig. 3).Over 65% of particles were of less than 100-nm diameter at all workplaces.A similar finding (60.7%) was reported during gas metal arc welding (Zhang et al., 2013).Nano-sized particles have a large surface area.However, this characteristic of airborne nanoparticles could change upon their agglomeration after emission into the air.Although terms such as agglomeration and aggregation are defined to differentiate the strength of adherence (ISO, 2008), measuring or differentiating nanoparticles existing as individual particles, agglomerations, or aggregates in the air remains challenging.Concentrations measured using SMPS and CPC in ENP-E and ENP-F were similar, but the surface area concentration and mass concentration were higher at the ENP-F workplace (Table 2), while the percentage of particles less than 100 nm was greater at the ENP-E workplace.This is partly due to the difference in adherence.As shown in Fig. 4, particles at (a) ENP-E (ca. 100 nm) showed larger primary particle sizes than those at (b) ENP-F (7 to 40 nm).Based on the number concentration, surface area concentration, and SEM images, we confirmed that particles at ENP-F were present as agglomerates rather than individually.During the welding process, particles in chain-like formations were detected; Figs.4(c) and 4(d), as have been reported previously (Zimmer and Biswas, 2001;Antonini et al., 2007).However, there are still controversial issues for exposure assessment strategy as well as toxicology studies (Brouwer et al., 2012).Toxicology studies have reported that nanoparticles are potentially hazardous compared with the same mass of larger particles of the same chemical composition and generally more biologically active; however, this remains controversial (Wittmaack, 2007).Thus, mass-based exposure limits should be reevaluated to improve worker protection.Lower exposure limits may need to be established for nanoparticles using a mass-based metric; alternatively, exposure limits may need to be re-defined using the particle number or surface area which better explains the dose-response relationships for particles of the same chemical composition (Oberdörster et al., 2007;Schulte et al., 2008).Thus, mass, number, and surface area concentration measurements should be performed prior to setting of the occupational exposure limit.
Background (BG) concentration measurements are important for exposure assessment to distinguish particles emitted by the process itself from those with other sources.We measured background concentration using the nearfield method in the same place at different times (i.e., off-duty, during lunch, or continuously stable time during sampling duration) (Fig. 1).The majority of BG concentrations were lower compared to other tasks performed at workplaces because the BG was selected as constantly stable.However, the BG concentration of UNP workplaces was higher than that at other workplaces, except at lunch time during stainless steel welding processes at UNP-I because the production rate was lower than that during arc welding.In some cases, lunch-time levels were not lowest and so could not be used as a background.During non-working periods, such as offduty time or lunch, no air circulation occurred because there was only general ventilation through windows and doors.Therefore, many factors affect background concentrations; e.g., task, time, stability, other possible sources, and measured concentrations.
There is no single instrument that could characterize nanoparticles perfectly.Therefore, several instruments have been used to measure nanoparticles (Peters et al., 2008;Ham et al., 2012;Methner et al., 2012).The SMPS could characterize the size distribution of nanoparticles as well as concentration with several channels but it is expensive and heavy.The CPC is portable and cheaper than SMPS but it needs to refill the isopropyl alcohol (IPA) regularly (8 hours in manual but shorter than 8 hours) and could not measure size distribution.The SAM could measure lung deposited surface area concentration but it is heavy and measures integrated size range (10-1,000 nm).All these instruments are not for personal monitoring but for area sampling.The point of view for industrial hygiene, portable and inexpensive measurement devices are required.
ANOVA revealed a limitation in the auto-correlation, and we did not consider auto-correlation in this study.The real-time monitored data are highly auto-correlated between samples.ARIMA model could be used to account for the effects of autocorrelation for real-time data (Brouwer et al., 2012;Ham et al., 2012).

CONCLUSIONS
We compared the concentration and investigated the characteristics of nanoparticles at nine workplaces categorized into three types.UNP workplaces generated the highest concentration for all measured metrics.The nanoparticle concentration at LAB workplaces was lower than those at other workplaces.In addition, the characteristics of exposure and size distributions differed according to workplace.Some tasks or processes at LAB showed higher concentrations than at ENP or UNP workplaces, despite the short exposure period.The LEV system may be appropriate for reducing exposure to nanoparticles at ENP workplaces.UNP workplaces emit high levels of nanoparticles, and exposure to workers should be reduced.Therefore, we found that the characteristics of exposure differ at LAB, ENP, and UNP workplaces, and so different risk management strategies are required.

Fig. 1 .
Fig. 1.Boxplot of total number concentrations based on the type of workplace.(a) LABs, (b) ENPs, and (c) UNPs.BG is the background concentration for each workplace and was selected as continuous and stable time to compare with processes and tasks with the lowest concentrations during measurement.Upper and lower whiskers represent the 5th and 95th percentiles, respectively.The boxes show the 25th and 75th percentiles.The median is indicated by the solid line inside the box.

Fig. 3 .Fig. 4 .
Fig. 3. Nanoparticle (less than or equal to 100 nm) fraction (given as a percentage of total number concentration) and concentration as measured by SMPS for all workplaces.

Table 1 .
General characteristics of workplaces.

Table 2 .
Descriptive statistics of workplaces measured using SMPS, CPC, SAM, and gravimetric sampling.

Table 3 .
Homogeneous subset by post hoc test for workplaces.