Article Cite This: Environ. Sci. Technol. 2018, 52, 14208−14215
pubs.acs.org/est
Predicting Indoor Emissions of Cyclic Volatile Methylsiloxanes from the Use of Personal Care Products by University Students Tao Yang,† Jianyin Xiong,*,†,‡ Xiaochen Tang,§ and Pawel K. Misztal‡,∥ †
School of Mechanical Engineering, Beijing Institute of Technology, Beijing 100081, China Department of Environmental Science, Policy and Management, University of California, Berkeley, California 94720, United States § Indoor Environment Group, Energy Technologies Area, Lawrence Berkeley National Laboratory, Berkeley, California 94720, United States ∥ Centre for Ecology & Hydrology, Edinburgh, Midlothian EH26 0QB, U.K. ‡
Environ. Sci. Technol. 2018.52:14208-14215. Downloaded from pubs.acs.org by TULANE UNIV on 01/11/19. For personal use only.
S Supporting Information *
ABSTRACT: Characterization of indoor emissions of cyclic volatile methylsiloxanes (cVMS) due to the use of personal care products is important for elucidating indoor air composition and associated health risks. This manuscript describes a mass transfer model to characterize the emission behaviors of decamethylcyclopentasiloxane (D5, the most abundant indoor cVMS) from skin lipids. A C-history method is introduced to determine the key parameters in the model, i.e., the initial concentration and diffusion coefficient of D5 inside the skin lipids. Experiments were conducted in a university classroom to examine the D5 emission behaviors by using a proton-transfer-reaction time-of-flight mass spectrometer (PTR-TOF-MS). Data from the first class session of two typical days was applied to obtain the key parameters, which were subsequently used for predicting D5 concentrations in other class sessions. Good agreement between model predictions and experiments demonstrates the effectiveness of the model and parameter determination method. With the model, we found that the reuse of personal care products has a significant impact on the D5 emissions. In addition, the time-dependent emission rate and remaining amount of D5 inside the skin can also be calculated. These results indicate a fast decay pattern during the initial emission period, which is consistent with prior experimental studies.
■
wastewater.12,13 Therefore, characterization of cVMS due to the frequent use of personal care products in the indoor environment is needed, to understand its transport and fate. A rapid progress can be achieved thanks to the development of time-resolved analytical tools that strengthen our ability on the characterization of cVMS emissions. Previous studies have generally focused on investigating the occurrence and concentration level of cVMS in typical personal care products or in the indoor environment. The concentrations or contents of cVMS in personal care products may differ due to variations in countries and types of personal care products and compounds. For collected 76 personal care products in U.S. and Japan, the concentrations of D4, D5, and D6 were measured to be 0.35−9380 μg/g, 0.39−81800 μg/g and 0.33−43100 μg/g.3 By testing 252 cosmetic products in Canada, the results for the above three cVMS were reported to be 10−11000 μg/g, 20−683000 μg/g, and 10−97700 μg/g.14
INTRODUCTION The emission of cyclic volatile methylsiloxanes (cVMS) from the use of personal care products is regarded as an important contributor influencing indoor air quality.1,2 cVMS are defined as volatile, low-viscosity, silicone fluids containing different numbers of “Si−O” bonds making up the chain. cVMS are organic ingredients or solvents used in a wide spectrum of personal care products due to their low surface tension, high thermal stability, and smooth texture. These products include antiperspirants, skin and hair formulations, sunscreen creams, and cosmetics.3−5 Three widely used cVMS are octamethylcyclotetrasiloxane (D4), decamethylcyclopentasiloxane (D5), and dodecamethylcyclohexasiloxane (D6).2 These cVMS are currently of particular concern and being considered for regulation, because they are potentially persistent, bioaccumulative, and toxic6,7 and have been shown to be associated with connective tissue disorders, adverse immunologic responses, and fatal liver and lung damage.3,8−11 cVMS are ubiquitous in the indoor environment, since they emit from the original sources and then partition into the indoor air due to their high Henry’s law constants. In addition to their presence in indoor atmosphere, cVMS are also detected in indoor dust, water, and © 2018 American Chemical Society
Received: Revised: Accepted: Published: 14208
January 23, 2018 May 29, 2018 June 8, 2018 June 8, 2018 DOI: 10.1021/acs.est.8b00443 Environ. Sci. Technol. 2018, 52, 14208−14215
Article
Environmental Science & Technology Lu et al.4 examined 158 personal care products in China, and the concentration ranges of D4, D5, and D6 were determined to be 0−72.9 μg/g, 0−1110 μg/g, and 0−367 μg/g. These results indicate that the total cVMS (sum of D4, D5, and D6) can account for as much as 80% of the total content of some personal care products (e.g., cosmetics, hair care products), meaning that they are an important source of cVMS in the indoor environment. Up to now, there are limited reports of measured indoor gas phase cVMS concentrations, and most of them lack the time-resolved data to assess cVMS dynamics. Tests from 91 samples in different sites in UK and Italy indicated that the D4, D5, and D6 concentrations in indoor air were in the range of 19−270 μg/m3, 2.4−440 μg/m3, and 0.47−79 μg/m3.15 By collecting 3857 samples in Canadian residential indoor air, Zhu et al.16 reported that the representative levels of indoor D4 and D5 concentrations were within 5.45−7.94 μg/m3 and 35.47−47.18 μg/m3. In Chicago, the indoor concentrations of the three cVMS were measured to be 0.03−0.5 μg/m3, 0.97−56 μg/m3, and 0−2.8 μg/m.317 Tang et al.2 measured the cVMS levels from engineering students in a classroom in California, and the average D4, D5, and D6 concentrations were 3.9 μg/m3, 99 μg/m3, and 2.0 μg/m3 in a typical day. For most of these previous experimental studies, D5 is recognized as being the dominant cVMS, both in personal care products and in the indoor air, even comprising over 90% of total cVMS in some studies.2,17 For this reason, the research presented here mainly looks at the characterization of D5 emissions. At present, some modeling work is being done to predict D5 emissions. The transport modeling of D5 concentrations in the outdoor atmosphere is well-developed, and good agreement has been found between the model predictions and experiments, implying that the sources, transport paths, and sinks of D5 in the outdoor atmosphere are well-understood.18−20 However, very little research has been done to systematically model D5 emissions in the indoor environment, especially from human skin after the use of personal care products. Some limited investigations aimed at understanding dermal absorption of D5 in the skin have been carried out by introducing a compartmental model.21 However, the model did not consider the internal diffusion process. In addition, subsequent experimental dermal absorption studies5,22 demonstrated that only less than 0.2% of the applied D5 was absorbed through the skin, and it was reported that the majority of the applied D5 would volatilize from the skin surface. These results imply that the dominant pathway of D5 to the indoor environment is from skin emissions after the use of personal care products, and the main exposure pathway of D5 is probably via inhalation. The lack of a model for characterizing D5 emissions from the skin surface prevents a deep understanding of D5 occurrence in indoor air, as well as limiting the ability to assess exposure. In addition to model development, the characteristic parameters (initial concentration, diffusion coefficient) in the model need to be determined as a prerequisite for prediction. To our knowledge, no attempt has been made to measure the diffusion coefficient of D5 in the skin. The objectives of this research are therefore to (a) describe a model of D5 emission from an occupant’s skin due to the use of personal care products and determine the key parameters in the model; (b) test the model efficacy by comparing predictions with observations of D5 concentrations in a university classroom; and (c) perform a dimensionless analysis
to show how the model could be generalized to other cVMS emissions.
■
METHODS Model Development. When personal care products (e.g., antiperspirants, sunscreen creams) are applied, the ingredients are initially assumed to be present in a separate layer on top of the skin lipids. Here we make a further assumption that this separate layer will partition into the skin lipids. Emissions of the dominant cVMS, D5, from the skin lipids of the human body are modeled as a physical process. Since the skin lipids form a thin layer, it is assumed that the diffusion of D5 inside the skin lipids can be represented as one-dimensional. The skin is assumed to form an impermeable barrier with regard to D5 penetration. Such an assumption is consistent with earlier dermal absorption studies as mentioned in the previous section.5,22 We also make an estimation to verify this. According to the multipathway exposure model,23 the depletion of chemicals in the skin lipids is controlled by two rate constants, i.e., the product to air rate constant (kp,a) and the product to skin (dermal intake) rate constant (kp,s). For D5 emission from skin lipids, based on the formula for kp,a and kp,s,23 and the physical-chemical property data from EPI Suite, it is calculated that the ratio of kp,a to kp,s for D5 is about 27 (see details in Section S1 of the Supporting Information), meaning that most of D5 are volatilized from the skin lipids surface and dermal intake can be neglected. In addition, adsorption or absorption of D5 to interior surfaces is assumed to be negligible. A further assumption is that the D5 level indoors is well mixed. With these assumptions, the emission process can be represented as shown in the schematic in Figure 1. In this figure, hm represents the external mass transfer coefficient.
Figure 1. Schematic of D5 emission from skin lipids.
The governing equation describing the transient diffusion through the skin lipids, together with the boundary conditions and the initial condition, is described by eqs 1−4 ∂Cm ∂ 2C = Dm 2m ∂t ∂x
(1)
∂Cm = 0, ∂x
(2)
−Dm
x=0
∂Cm iC y = hmjjj m − Ca zzz, ∂x kK {
Cm(x , t ) = C0 ,
t = 0,
x=L 0≤x≤L
(3) (4)
where Cm is the concentration of D5 in the skin lipids, μg/m3 or ppb; Ca is the concentration of D5 indoors, μg/m3 or ppb; Dm is the diffusion coefficient of D5 in the skin lipids, m2/s; L is the thickness of the skin lipid layer, m; K is the skin/air partition coefficient, dimensionless; and C0 is the initial 14209
DOI: 10.1021/acs.est.8b00443 Environ. Sci. Technol. 2018, 52, 14208−14215
Article
Environmental Science & Technology concentration of D5 in the skin lipids (i.e., at the time the emitters enter the indoor space), μg/m3 or ppb. Equation 1 stands for the one-dimensional diffusion process of D5 inside the skin lipids, while eqs 2 and 3 stand for two kinds of boundary conditions (no mass flux and convective mass flux). The mass balance of D5 indoors is given as V
dCa ∂C = QCin − ADm m |x = L − QCa dt ∂x
It should be noted that the present model assumes that all emitters have the same initial D5 concentration C0. This is certainly a rough estimate, but it provides a substantial simplification for the model development. In some extreme cases, if C0 differs greatly for different emitters, the mass balance eq 5 should be improved accordingly V
(5)
where V is the volume of the indoor space, m3; A is the total area of the skin lipid layer emitting D5, m2, which can be calculated by Ne × Ap (Ne is the number of emitters; Ap is the area of the skin lipid layer of one emitter including face and hand; the value of Ap is taken as 0.095 m2 from ref 5, which is in accordance with standard clinical applications for personal care products); Cin is the inlet D5 concentration via ventilation system from outdoors, μg/m3 or ppb; and Q is the ventilation rate, m3/s. It should be noted that eqs 1−5 are widely used for building material emissions.24,25 Here we apply these equations to characterize D5 emissions from skin lipids. We assume here that D5 is absent from the indoor air prior to the arrival of occupants: Ca(t ) = 0,
t=0
∞
Ei(t ) = A pDm ∑
A3 ·qn sin qn − A4 ·cos qn Gn
n=1
−2 2 qn t
e−DmL
[Kβ + (α − 3qn2)KBim−1 + α − qn2]qn sin qn; A1 = Cin; A2 = C0; A3 = β; A4 = 0; α = QL2/DmV; β = AL/V; Bim = hmL/Dm; and qn are the positive roots of Kβ + (α − qn2)KBim−1
(n = 1,2, ...) (8)
For real occupant emission tests, the volume of the environment is often very large (room or full-scale chamber), and the convective mass transfer coefficient is very small (natural convection in many scenarios). Under this condition, the term Kβ will be much smaller than αK/Bim (generally by 2−4 orders of magnitude smaller) in eq 8 and thus can be ignored. Equation 8 can then be simplified: qn tan qn =
Bim K
(n = 1,2, ...)
Ca(t ) = Cin +
∑ n=1
2βC0 (α − qn2)(1 + K /Bim + qn2K 2/Bim2)
e
∫0
t
−2 2 qn (t − τ )
e−DmL
ln[Ca(t ) − Cin ] = SL ·t + INT
(9)
(13)
with
Such simplifications can significantly reduce the calculation complexity and increase the accuracy of finding the root of a trigonometric equation. Finding the root is the key point of applying the analytical solution to predict the D5 concentration or emission rate. With eq 9 and the condition Kβ ≪ αK/Bim, the analytical solution (eq 7) can be further reduced to ∞
ÉÑ Ñ KdCa(τ )ÑÑÑÑ ÑÑÖ
L(qn2 + Bim2 /K 2 + Bim /K )
Subsequently, the indoor D5 concentration can be obtained by combining eqs 11 and 12, which is called the semianalytical solution. For the sum of infinite series in eq 10 or 12, about 15 terms are used for the calculation in the present study, which can maintain high calculation accuracy. The analytical solution (eq 10) establishes the explicit relationship between the indoor gas phase concentration and emission time. According to the concept of the C-history method widely used in building material emission scenarios,28−31 this explicit relationship is a prerequisite for rapidly and accurately determining the key parameters in the model. This is one of the main merits of the analytical model. In this study, the following approach is used: (1) the analytical solution is used to determine the key parameters (this will be introduced in the following section) and (2) the key parameters are then substituted into the semianalytical solution to predict the D5 emission characteristics. Method for Determining the Key Parameters. To calculate the indoor D5 concentration, the three key parameters (C0, Dm, K) for the model need to first be determined. The key parameters C0 and Dm can be determined using a ventilated chamber C-history method31 based on the analytical solution. The principle behind this method is briefly introduced here. By examining the cVMS emissions in detail, it is found that during the midterm emission period (i.e., mass transfer Fourier number, Fom, is larger than 0.125), the following equation can be derived based on the analytical model
w h e r e G n = [Kβ + (α − qn2)KBim−1 + 2]qn2 c o s q n +
α − qn2
2(qn2 + Bim2 /K 2) sin 2 qn
(12)
(7)
qn tan qn =
(11)
i
ÄÅ −2 2 Å × ÅÅÅÅC0e−DmL qn t + ÅÅÇ n=1
By virtue of Laplace transforms, eqs 1−6 can be solved to determine the indoor D5 concentration in this form26 ∞
∑ Ei − QCa
where Ei stands for the emission rate for emitter i. Under this condition, the equations can be solved by the separation of variables method,27 where the emission rate for emitter i can be represented as (the indoor D5 concentration prior to the arrival of occupants is set to be zero):
(6)
Ca(t ) = A1 + 2A 2 · ∑
dCa = QCin + dt
SL = −DmL−2q12
ÄÅ ÉÑ ÅÅ ÑÑ 2βC0 ÅÅ ÑÑ Ñ INT = lnÅÅ ÅÅ (α − q 2)(1 + K /Bi + q 2K 2/Bi 2 ) ÑÑÑ ÅÅÇ m m Ñ 1 1 ÑÖ
(14)
(15)
where q1 is the first positive root of eq 9, which is in the range of 0−π/2. Equations 13−15 indicate that the logarithm of Ca(t)-Cin is in a linear relationship with the emission time t, and the slope
−DmL−2qn2t
(10) 14210
DOI: 10.1021/acs.est.8b00443 Environ. Sci. Technol. 2018, 52, 14208−14215
Article
Environmental Science & Technology (designated by SL) and intercept (designated by INT) are functions of the two key parameters, C0 and Dm. Therefore, if we perform linear curve fitting with the experimental data of Ca(t)-Cin against t (Fom = Dmt/L2 > 0.125), the SL and INT can be obtained, following which C0 and Dm can be determined by solving two equations with these two key parameters. When applying the C-history method to obtain the two key parameters, an initial value of K is preassumed and taken as known. The rationality of such treatment is demonstrated by the sensitivity analysis in the following section, which shows that the midterm or long-term emission behaviors are insensitive to the preassumed K (K only influences the shortterm emission behaviors), meaning that the measured C0 and Dm based on midterm emission data should be also insensitive to the preassumed K. For some emission scenarios, once the C0 and Dm have been obtained, these values together with the short-term emission data can be further used to obtain an updated K.
log K = log Kow + log(HRT )
(16)
where Kow is the octanol−water partition coefficient; H is the Henry constant; R is the gas constant; and T is the temperature. The use of eq 16 is based on that octanol represents a good surrogate for the liquid-like material described in the present study. Based on the physical-chemical property data from EPI Suite, the K for D5 at 305 K (32 °C) is calculated to be 3.27 × 104. The thickness of the skin lipid is an important parameter for D5 emission. Some previous studies investigated the thickness of the skin lipids and found variations in different persons and different regions of the body.35,36 The determined thickness of skin lipids is generally in the range of 0.5−1.3 μm.37 To simplify the analysis, the thickness of skin lipids in the present study is taken as uniformly 1 μm. The calculation of hm is based on the Chilton-Colburn analogy for heat and mass transfer.38,39 The environmental conditions and dimensions of the classroom and skin lipids are summarized in Table 1. Other information such as the recorded data of inlet D5 concentrations via the ventilation system from outdoors for different class sessions is given in SI Figure S1.
■
EXPERIMENTAL SECTION Generally, there are two approaches to experimentally examine the emission characteristics of D5 from skin lipids. One is to perform the experiment in a controlled environmental chamber with occupants, and the other is to conduct the experiment in the field. We used the latter approach in this study. The field campaign was conducted in a university classroom in California. The volume of the classroom was about 670 m3. The air-exchange rate (N) during the daytime 8:00−20:45 was evaluated through CO2 tracer decay method (utilizing deliberate releases) in a prior study for the same classroom.32 It was determined to be 5 ± 0.5 h−1 (replacing ventilation) and was adopted in the present study. The real-time occupant (student) number in the classroom was recorded for several days. The D5 concentrations in the classroom were continuously monitored by an online sampling instrument proton-transfer-reaction time-of-flight mass spectrometer (PTR-TOF-MS) (IONICON Analytik GmbH). This instrument uses H3O+ as the primary reagent to softly ionize volatile organic compounds across the mass spectrum for mass-tocharge ratio (m/z) of 20.00−500.00 Th. PTR-TOF-MS enables fast response time measurements (a full mass spectrum in seconds or minutes) and has high sensitivity (tens of parts per trillion or ppt in a second) and high mass resolution.2,33 Since the supply air (outdoor air) could contain a certain amount of D5, a three-way valve was used to switch the sampling of PTR-TOF-MS between the supply air and the classroom air every 5 min. To reduce possible effects from valve switching and sampling tube interactions, the data for the first 2 min after every switching were discarded. The gas phase D5 concentration was then averaged based on the data in the remaining 3 min, to get relatively smooth results.
Table 1. Environmental Conditions and Dimensions of the Classroom and Skin Lipids parameters
Nov 6
Nov 13
temperature (8:10−11:50 AM, °C) relative humidity (8:10−11:50 AM) V (m3) N (1/h) Q (m3/h) L (μm) Ap (m2) hm (m/s)
23.3 ± 0.6 (46.2 ± 1.8)% 670a 5 ± 0.5 3350 ± 335 1 0.095 9 × 10−4
22.8 ± 0.2 (62.6 ± 3.5)%
a
The values of this parameter and the following parameters in this table are the same for the two tested days.
Figure 2 gives the linear curve fitting results using eq 13. Based on the SL and INT, the C0 and Dm determined for class
■
Figure 2. Linear curving fitting results by virtue of eq 13 for class session 1 on Nov 6 and Nov 13.
RESULTS AND DISCUSSION Determination of the Key Parameters. To check that the model provides reasonable results, the key parameters in the model are first determined. The experimental data from the classroom tests are used to achieve this goal. The key parameters can be extracted based on data from the first class session observed on Nov 6 and Nov 13, 2014 (class session 1:8:10−9:40 AM). For D5 emission from skin lipids, the initial K can be approximately predicted by the following correlation34
session 1 on Nov 6 (student number: 24) are C0 = (8.80 ± 0.62) × 1010 μg/m3, Dm = (1.46 ± 0.08) × 10−16 m2/s, while the values for class session 1 on Nov 13 (student number: 26) are C0 = (1.93 ± 0.16) × 1011 μg/m3, Dm = (1.35 ± 0.07) × 10−16 m2/s. The standard deviations for C0 and Dm are calculated based on the method in ref 28. These results indicate that the relative standard deviation for C0 and Dm is 14211
DOI: 10.1021/acs.est.8b00443 Environ. Sci. Technol. 2018, 52, 14208−14215
Article
Environmental Science & Technology less than 10%, which is relatively small. In performing curve fitting with eq 13, the time t should belong to the midterm emission period (Fom > 0.125 or t > 0.125L2/Dm). Since Dm is unknown before the experiment, we can check whether or not the selected time t meets the condition after determining Dm. By validation, the selected time of the field campaign for curve fitting meets the requirement (t > 0.125L2/Dm). By comparing the determined key parameters on different test days, we find that Dm is very close, while C0 differs by about 1-fold. The environmental conditions in Table 1 indicate that the temperature on Nov 6 and Nov 13 is almost the same, while the relative humidity changed a great deal (It was a rainy day on Nov 13.). The above results imply that the relative humidity can significantly influence C0 while having a negligible effect on Dm. This conclusion is consistent with prior studies which examined the emissions of other types of VOCs.40,41 Moreover, detailed analysis based on tests on 2 other days (Nov 5 and Nov 12) demonstrates that the average emission rate in the morning differs less than 20% if the environmental conditions are similar (temperature difference
