Chemicals
2-(2-Butoxyethoxy) ethyl acetate, decamethyl tetrasiloxane, di(propylene glycol) methyl ether (mixture of isomers), 2-ethylhexyl acrylate, hexamethyl disiloxane, methanol, 2-(2-methoxypropoxy)-1-propanol, 1H,1H,2H-perfluoro-1-decene, 1H,1H,2H,2H-perfluorodecyl acrylate, 1H,1H,2H,2H-perfluoro-1-octanol, toluene, trimethyl silanol were obtained from Sigma Aldrich (Brøndby, Denmark). 1H,1H,2H-perfluoro-1-tetradecene was obtained from Synquest Laboratories, Alachua, FL. All chemicals were of the highest available purity (98-99.9%).
Waterproofing product
The label of the product indicated it was water-based. According to the user’s instructions it would make wooden surfaces water and dirt resistant and explain how it should be applied. The product was advertised for surface treatment of hardwood furniture (for both indoor and outdoor use). The label did not contain symbols referring to any health or environmental hazards, but the product contained the following phrases (translated from Dutch/French): ‘Keep out of range of children; prevent direct contact with skin and eyes; use only in a well-ventilated area; store in closed bottle; do not induce vomiting if swallowed and consult a physician and show the package and label; we do not accept liability in the case of wrong use; attention: store frost-free’. The label did not provide further information about the composition, e.g. fluorinated compounds or nanoparticles. The manufacturer provided information about the formulation of the product to the authors (Table 1). On the label, instructions were given to rub the product on the surface using a cloth. However, in this case the product was sprayed by use of a spray gun.
Analysis of the product showed presence of nano-sized spheres with a median diameter of ca. 72 nm, with strong hydrophobic properties and with a slight positive charge (zeta potential of +31.3 mV, see Additional file 1: Figure S2-S5). These particles formed a stable suspension in water with a negligible tendency to form clusters. The particles were chemically characterized as water-free solid organic silica cores, with a soft shell consisting of tri-block-copolymers containing perfluoroalkyl acrylate, in addition to polyethylene oxide and polypropylene oxide, and presumably end-capped with ethylene oxide.
Reconstruction of spray incident
In order to obtain information about the emission of volatile organic compounds (VOC) and aerosols during and after application of the product, spray tests were carried out under different test conditions in two different test chambers (Experiments 1 and 2). Experiment 1 was carried out at 6-fold the concentration of the incident; Experiment 2 at 46-fold higher concentration for the evaluation of larger particles and collection of particle samples for further analysis (not presented here). In both experiments, the spray gun (Eminex type E31 EHT M 01, Eminent, Oss, NL) was equipped with similar nozzles and operated at the same spray pressure (2.0 Bar) as the one originally used in the wood workshop on the day of the incident.
Experiment 1
In this experiment, 47 g of product was sprayed in a uniform layer on a 0.6 m2 untreated plywood surface placed inside a climate steel chamber (h x l x w = 2.29 × 3.46 × 2.56 m) with an ante-chamber (2.72 m3) as inner entrance. The experiment was carried at a ventilation, temperature and relative humidity of 0.08 ± 0.03 h−1, 22 ± 2 °C and 45 ± 5%, respectively. Immediately after the start of the experiment, three mixing fans were turned on for 60 s in order to ensure a homogeneous distribution of aerosols and VOC. The fans were placed in three corners on the floor and at 5 cm from the chamber wall. VOC were sampled through a 10 mm stainless steel sampling manifold placed at a height of 1.0 m from the floor and ca. 1.2 m from the spray position; air was sampled at 5 cm from the inner chamber wall. During the first 5 h, samples were taken in duplicate at 10-30 min intervals, starting the first sampling event 1 min after the spray application was initiated. Additional samples were taken after 23 and 25.5 h, following the start of the spray application, respectively. The time of sampling is given as the midpoint between start and end of each sampling period. VOC data are reported as mean of duplicates, corrected for chamber background air and rounded to the nearest integer. VOC were sampled on clean Tenax TA (60-80 mesh) adsorbent tubes (200 mg) with a sampling time of 10 min at 100 mL/min, using calibrated pumps (Gillian Gilair 5, Sensidyne, US). The Tenax TA tubes were analyzed on a Perkin Elmer Turbo Matrix 350 thermal desorber (TD) coupled to a Bruker SCION TQ GC-MS system (Bruker Daltonics, Bremen, DE). Tube desorption was carried out at 275 °C for 20 min and the low and high temperatures of the cryo trap were −20 °C and 280 °C, respectively. Separation was performed on a 30 m GC column with 0.25 mm internal diameter and 0.25 μm film thickness (type VF-5MS, Agilent Technologies, Santa Clara, US). The oven program was as follows: 50 °C for 4 min, ramp 1: 4 °C/min to 120 °C, ramp 2: 50 °C/min to 250 °C, hold for 2 min. Helium was used as carrier gas at an inlet pressure of 0.97 bar (1.5 mL/min). The mass spectrometer was operated in SIM/scan mode using either electron ionization or chemical ionization with methane (5.0) as ionization gas. Argon of ultrahigh purity (99.999%) was used for collision induced dissociation (CID) experiments. Valves, transfer lines and ion source were kept at 270 °C. Six-point calibration was applied (r2 > 0.99) using authentic standards in methanol. Identification of observed VOC was based on retention time and mass spectra of authentic standards, when applicable, in addition to library search [29], chemical ionization and CID.
Number size distribution measurements were conducted using a TSI Model 3091 Fast Mobility Particle Sizer (TSI, Shoreview, NM). The instrument was operated at 1-s time resolution in a measurement range of 5.6-560 nm. An optical particle spectrometer Grimm 1.109 (Grimm Aerosoltechnik, Ainring, Germany) was used to measure the number size distribution from 0.25 to 32 μm at 6 s time resolution (count distribution mode). Total number concentration was integrated from the number size distribution. Spherical SOA particles with a density of 1.0 g/ml was assumed for mass calculations. In addition, a density of 2.6 g/ml was used for comparison of mass results with experiment 2 measurements where an environmental mode was used (in the environmental mode a default density of 2.6 g/ml was also used to calculate mass concentration).
Experiment 2
In this experiment, ca. 0.7 l of the product was sprayed ‘mid air’ inside a 40 m3 paint booth (h x l x w = 4.0 × 4.0 × 2.5 m) during a period of 10 min. During the last 2 min of spraying, the spray booth was turned off to allow the particles to remain in the booth and to describe the time-resolved post-spraying changes in particle size. Aerosol mass size distribution spectra in the size range from 0.25 to 32 μm were measured using a Grimm 1.109 operating at a time resolution of 6 min. The instrument was set to occupational mode (PM-10, PM-2.5, PM-1.0 in μg/m3). A nephelometer (type IV Hazdust, Environmental Devices, Plaistow, NH, USA), equipped with a cyclone pre-seperator, was used for collection of the thoracic fraction (equivalent to PM-10) at a 10-s time resolution.
Filter samples were collected twice; the first period of 0-50 min after the start of application and a second period of 1-16 h after application. Inhalable dust was collected using an IOM sampler and a calibrated personal air sampling pump (A.P. Buck, Orlando, Florida, USA) at 2.0 L/min. A similar set-up was used to collect respirable particles by a Cassella cyclone (Cassella Measurement, Bedford, UK) at 1.9 L/min. Thoracic dust and respirable dust fractions were collected using PM-10 and PM-2.5 Harvard impactors, respectively, at 10.0 L/min. The filter samples were collected 2.0 m from the spray application at height of 1.2 m from the floor. All filters used were Teflon membrane filters (Sartorius, Göttingen, Germany).
Modelling
A two compartment model was constructed using Simulink/MatLab. This model was used to estimate the time pattern of VOC in the wood workshop and the mail sorting and distribution centre.
The air concentrations of VOC were modelled using the measurement data from experiment 1. The modelling was carried out by use of the data from total glycol ether which was by far the most abundant of measured VOCs. The initial concentration of the glycol ethers at the start of the spray application in the wood workshop was calculated to be 46 mg/m3, using a standard spray scenario of the exposure model ConsExpo 5.0 (www.rivm.nl/en/Topics/C/ConsExpo). The inter-compartment flow (Qi) in the open air connection with the dimensions of 20 × 0.1 m (2.0 m2) was estimated to be 7200 m3/h for an air speed of 1 m/s (1 m/s × 2.0 m2 = 2 m3/s = 7200 m3/h). We then estimated the concentration in the wood workshop over time by adopting an algorithm used by Vernez et al. [30]:
$$ {C}_1=\int \frac{1}{V_1}\ \left({Qi}^{\ast }{C}_2-{Qi}^{\ast }{C}_1-{Q}^{\ast }{C}_1\right) $$
(1)
where C1 is the concentration in the wood workshop (initial concentration 46 mg/m3),
C2 is the concentration in the mail sorting centre (initial concentration 0 mg/m3),
V1 is the volume of the wood workshop (2600 m3),
Q is the air exchange rate with outdoor air (0.08 h−1, 1.0 h−1 and 2.5 h−1),
Q
i
is the inter-compartment flow (7200 m3/h).
A second equation was used to calculate the concentrations over time in the mail sorting centre:
$$ {C}_2=\int \frac{1}{V_2}\ \left({Qi}^{\ast }{C}_1-{Qi}^{\ast }{C}_2-{Q}^{\ast }{C}_2\right) $$
(2)
where C1 is the concentration in the wood workshop (initial concentration 46 mg/m3),
V2 is the volume of the mail sorting centre (2400 m3),
C
2
, Q and Qi are the same as in eq. (1).
The initial concentration in the mail sorting centre was assumed to be zero. As the complaints of the mail workers started at 8:00 am the next morning, we calculated the concentration at 15 h after the start of the spray event. We modelled three different air exchange rates. The first value of 0.08 h−1 is the air exchange rate of the measuring chamber and corresponds to a more or less airtight building. The other two values represent more realistic values to describe the type of naturally ventilated industrial building that is usually not very air tight, a low estimate of 1.0 h−1, corresponding to a low natural ventilation in a situation with low wind speed and closed doors and a high value of 2.5 h−1 to a moderately ventilated building with doors opened and at moderate wind velocity conditions. The real situation was probably in the range between these two conditions.