Simplified speciation and atmospheric volatile organic compound emission rates from non‐aerosol personal care products

Abstract Volatile organic compounds (VOCs) emitted from personal care products (PCPs) can affect indoor air quality and outdoor air quality when ventilated. In this paper, we determine a set of simplified VOC species profiles and emission rates for a range of non‐aerosol PCPs. These have been constructed from individual vapor analysis from 36 products available in the UK, using equilibrium headspace analysis with selected‐ion flow‐tube mass spectrometry (SIFT‐MS). A simplified speciation profile is created based on the observations, comprising four alcohols, two cyclic volatile siloxanes, and monoterpenes (grouped as limonene). Estimates are made for individual unit‐of‐activity VOC emissions for dose‐usage of shampoos, shower gel, conditioner, liquid foundation, and moisturizer. We use these values as inputs to the INdoor air Detailed Chemical Model (INDCM) and compare results against real‐world case‐study experimental data. Activity‐based emissions are then scaled based on plausible usage patterns to estimate the potential scale of annual per‐person emissions for each product type (eg, 2 g limonene person−1 yr−1 from shower gels). Annual emissions from non‐aerosol PCPs for the UK are then calculated (decamethylcyclopentasiloxane 0.25 ktonne yr−1 and limonene 0.15 ktonne yr−1) and these compared with the UK National Atmospheric Emissions Inventory estimates for non‐aerosol cosmetics and toiletries.

secondary organic aerosols when oxidized over several generations and in the presence of co-pollutants such as NO x . The ability of SOA to scatter and absorb solar and terrestrial radiation, influence cloud formation, and participate in atmospheric chemical reactions means they play a significant role at scales beyond that of urban and regional air pollution. 2 Additionally, as VOCs are a precursor to ozone and a sub-component of PM 2.5 , they contribute to poor air quality and related health effects such as pulmonary inflammation and respiratory illness. 3 From the 1970s onwards, global regulation and policy has focused primarily on reducing VOC emissions from sources such as the extraction and distribution of fossil fuels, combustion and leakage of fuels from road transport, natural gas networks, landfills, and coal-fired power stations. 4 Recently, as VOC emissions from fossil fuels and the transport sector have declined, the relative importance of other VOCs sources has increased. 5 Historically, aims to regulate indoor VOCs tend to focus on building materials, and with particular attention toward compounds such as formaldehyde, benzene, and toluene. Less thought has been paid to the VOCs emitted from the use of PCPs (personal care products) [6][7][8][9][10][11][12][13][14][15] and HCPs (household cleaning products) [16][17][18][19][20][21][22][23] which, along with other domestic emissions of VOCs, [24][25][26] are now known to be a substantial contributor to overall VOC emissions. 4 Within this study, PCPs refer to cosmetic and hygiene products available to the public for personal use. PCPs are often split into two broad classifications for the purposes of VOC emissions reporting, described as non-aerosol and aerosol, and it is non-aerosol products that are reported here. The non-aerosol class is potentially a smaller collective source of VOCs than aerosols, since the product matrix is often aqueous, whereas in the case of aerosol-based PCPs, it is typically a hydrocarbon blend based around butane. Ethanol or oil-based perfumes would be examples of PCPs based on hydrocarbons, although we do not test any of these in this study.
The mixture of VOCs emitted from sources such as gasoline evaporation is highly complex, but the detailed speciation of that source is reasonably constant and has been well-characterized over time (see eg, Europe Environment Agency, emission inventory guidebook 2016 27 ). Such mixtures are represented in some emissions inventories by an often complex speciation of VOCs, for example in the UK National Atmospheric Emissions Inventory. 28 Air pollution models typically have a more simplified speciation, through combining (lumping) different VOCs into a smaller sub-group of surrogate compounds, normally simple hydrocarbons, that are then explicitly treated subsequent oxidation mechanisms (see an overview of the topic in Carter, 2015 29 ).
The situation is less well developed for consumer products, since each has a unique, generally proprietary, formulation and a substantial diversity in both speciation and emissions rates exists. To add to the complexity, many of the VOCs used in consumer products are high molecular weight and produce a range of multifunctional species when oxidized, some of which may be more harmful to health than the VOCs contained in the original product. 30 For instance, the Master Chemical Mechanism, which is a near explicit mechanism developed to represent the degradation of VOCs in the atmosphere, 31 needs 1244 reactions and 712 species to represent all of the reactions needed to go from limonene to the final oxidation products of water and carbon dioxide. This complexity means that representing their chemistry in models for indoor air chemistry is extremely challenging.
The ability to predict VOC emissions (both in terms of speciation and in absolute amounts) is needed however for management of indoor air quality, and to quantify the effects that domestic releases of VOCs have on outside air once ventilated. Nearly 90% of human exposure to VOCs is now believed to come from this kind of diffuse and largely unregulated set of sources that are within individual or household control, which includes consumer products, 7 as well as other domestic sources such as glues, paints, sealants and other building products and materials. Other VOC sources in the home include natural gas leakage, pesticides, cooking, and combustion of wood, coal, and candles. 32,33 To understand our overall exposure to air pollution, it is vital to quantify the different sources of pollution both outdoors and indoors. In developed countries, we spend 80%-90% of our time indoors and so our exposure to air pollutants, whether generated indoors or outdoors, will happen in the indoor environment. The use of PCPs is likely to represent a fraction of our overall exposure to pollution, but to date there has been little information available on how the use of an individual product could contribute to the emissions of VOCs, or the secondary products that can then be formed through subsequent chemical reactions. This knowledge requires detailed emissions measurements with sufficient speciation of the often complex formulations to understand the ongoing chemistry.
The estimation of VOC emissions rates from non-aerosol PCPs is potentially a lengthy and time-consuming process. Quantifying VOC content and emissions from PCPs using traditional methods

Practical Implications
• Emissions of VOCs from the domestic sector, including personal care products, are highly uncertain, yet make up an increasing fraction of total VOC emissions in developed economies.
• The quantitative estimates of VOCs emitted from a range of personal care products provided here show that this information can constrain models of indoor air chemistry, particularly to make estimates of indoor concentrations of pollutants for which measurements are largely absent.
• Scaled estimated emissions provide a better guide to the contributions made by this source sector to personal emissions of VOCs than currently available and can be used to better understand personal exposure according to typical activities.
such as headspace GC-MS relies on the ability to predict the liquid-gas partitioning of any given VOC, something that is virtually impossible to do given unknown formulations. Establishing whether an equilibrium has been reached between sample and the atmosphere above, it is difficult to achieve under realistic conditions with GC-MS since the measurement frequency is rather slow, perhaps one measurement every 30 minutes. In a complex matrix where Henry's Law conditions likely do not apply, and where surface tension effects may be significant, a static headspace established over minutes to hours may not necessarily reflect VOC outgassing under more realistic non-saturated dynamic conditions. The availability of fast responding on-line mass spectrometry methods makes this a more tractable task in terms of tracking equilibration and VOC exchange, albeit with a penalty of less capability to speciate isomers and generally greater uncertainties in quantitative determinations. With on-line methods such as proton-transfer reaction mass spectrometry (PTR-MS) and selected-ion flow-tube mass spectrometry (SIFT-MS), the emission rate from a PCP sample can be tracked over minutes to hours using a dynamic flow of diluent gas over the sample and the temporal profile of concentrations then used to estimate the likely VOC emission rate and general VOC. The major advantage of using this method is that it has sufficient sensitivity for a direct analysis of a diluted dynamic headspace, avoiding the need for a pre-concentration/thermal desorption step, and an equilibrium headspace concentration is typically determined in a few minutes. A limitation however of the method is that, like all online and direct inlet mass spectrometry methods, there is a more limited ability to differentiate between isobaric compounds, a notable issue if resolution between specific isomers (eg, monoterpenes or monoaromatics) is important. There are some advantages in terms of calibration using online MS, in that some reasonable first order estimate can be made of the concentrations of unknown VOCs in an unknown mixture, and without a primary standard available. But on-line methods are inevitably less accurate than GC-MS, if primary calibration mixtures for individual VOCs are available.
In this paper, the aim is to produce simplified emission profiles with a grouped speciation that are suitable for chemical models of indoor air and that can provide a guide to the scale of potential personal emissions of VOCs from this class of products. In turn, these values are then scaled upwards to place national emissions of VOCs from PCPs in context to other sources.  The samples were drawn into the SIFT-MS at atmospheric pressure from the dynamic headspace of the stainless steel vessel at a flow rate of 25 mL min −1 , with the inlet to the vessel connected to a VOC-free supply of N 2 gas. Before and after each PCP sample, an experimental nitrogen blank was carried out which was subsequently subtracted from each sample, although these VOC concentrations were typically very much smaller than the measured amounts, typically < 5%). For all the samples tested here, an equilibrium concentration of VOCs was established in the exiting gas, proportional to the amount of material under test and the VOC content. Over the temperatures and timescales of the testing, which are similar in nature to products in use, each sample acts as an approximately constant emission source of VOCs, and that emission rate is not appreciably changed through VOC depletion in the raw product. Over much longer timescales (hours to days) and/or higher test temperatures, then it is possible to drive off VOCs such that the emission rate declines until ultimately the VOCs are exhausted and emissions fall close to zero. For PCP use, we assume that VOC content in the mixture is not a limiting factor since both time and temperature fall within bounds of a few minutes and no more than 40°C. With that assumption, the amount of VOC released is then proportional to the amount of product

| Data analysis
Measured product ions were normalized (for both blank and samples) by dividing the identified product ion intensities by the sum of their reagent and their respective water cluster ion intensities.   Table S1. On a small number of occasions where samples contained major VOC ions in the SIFT-MS that could not be directly identified or attributed to a given VOC class, like monoterpenes, we used a confirmatory GC-MS (Agilent 6890-5973) analysis to provide us with further information in toward an identification.

| Atmospheric model
The  The limonene measured in this study represents the sum of all monoterpenes. For the purposes of modeling, we treat this mechanistically as limonene, but denote it our results as limonene* in recognition that our model is not predicting for limonene exclusively.
Although there are differences in chemistry between different monoterpenes in terms of rate coefficients for reaction with OH, O 3 and NO 3 and also yields of radical production, it is the most ubiquitous and abundant monoterpene measured indoors 46 and so this simplification seems reasonable for the purpose of this study.

| Estimation of emission rates
The SIFT-MS is used to measure the time-dependant concentra-  It is possible that VOCs also escape to the air at some later stage, for example from waste-water, but we do not attempt to account for this in the scale-up calculations. For leave-on products such as moisturizers and liquid foundation (which remain on the skin, not washed off) more time is potentially available for VOCs to evaporate to air compared to wash-off products. Here, longer "in-use" scenarios are probably appropriate, but these must have some upper bound since the amount of VOC in the product is finite. We chose to express the individual VOC emissions as a mass released per unit time per gram of product and then, in a later section, apply

F I G U R E 2 Visualization of VOC emissions from 36 different PCPs based on H 3 O + ionization.
Data from each PCP sample is normalized to the maximum product ion intensity in that sample. Fragment ions are removed. *LF-liquid foundation, **Con-conditioner F I G U R E 3 A, Summation of total VOC product ion peak intensities for each PCP tested and B, median emission intensity for each product class an in-use period to each product. For example, one scenario is that a shower gel unit of activity may comprise a 4 g PCP sample in use for 30 seconds. Such an approach has to assume that as for the laboratory equilibrium determinations, over the actual periods of PCP activity/usage, the VOC liquid phase concentrations are not a limiting factor for VOC transfer to the gas phase, but rather the limitation is the mass transfer of VOC out of the product as a vapor. Table 1 shows the calculated emission factors from the simplified emission profiles as a function of time and mass of product at 25 ℃.
Since the range of total VOC emissions found in each product class is highly variable, for the subsequent calculations we report the median emissions of each VOC within each of the PCP classes.
The values in Table 1  There is limited literature guidance on typical in-use scenarios, so we must use our own best-estimates of a plausible range. The range of these scenarios (meaning amount of product used and time-scale for use) is such that this in turn creates a wide range of potential VOC emissions, something that could only be narrowed if more precise information on PCP in-use activity was available to us.
For our estimates, shampoo usage is assumed to be proportional to that of conditioner. Moisturizer is the most difficult product class to estimate, as many products fall into this category and are used in a variety of ways, both in terms of amount and frequency (eg, a small amount of eye cream is used daily compared to multiple hand cream applications), and it therefore has the largest estimated range of inuse emissions.

| Annual estimates of emissions of VOC from non-aerosol PCPs
The laboratory measured emissions factors are combined with the range of activity scenarios in Table 2 Table 2. Table S3 provides the summary of emissions for each product type and for the seven VOCs in the simplified VOC profile. We show this data in graphical format in Figure 4 for each of the products and for each of the seven VOCs within the simplified profile.
The seven species are selected to represent a simplified speciation based on data from Figure 2.

| Comparisons against emission inventory estimates
The  Table S3 providing ultimately a very broad range of potential emissions. Nonetheless, it is potentially useful to place those bottom-up estimates of emissions against the emissions currently included for this source class within the UK NAEI. Table 3 shows the activity and frequency scenarios then scaled for the UK as a whole, but with the application of some de-ratings to reflect TA B L E 1 Estimated product emission factors at 25°C for each non-aerosol PCP type using a simplified VOC emission profile that not all of the population will be users of each of those product types. We apply a reduction factor of 0.8 to shampoo and shower gel, 0.4 to conditioner, 0.25 to moisturizer, and 0.2 to foundation.
For comparison, we then extract from the 2017 UK NAEI the VOC emissions estimated under the EEA/EMEP Guidebook categorization of "Solvent Use," sub-class "Non-aerosol Products -Cosmetics and Toiletries," NFRCode: 2D3a and Source Code: 256.
The most immediate observation to be drawn from Table 3

| Model simulations
The emission rates from Table 1 were used to explore ambient concentrations that could arise following a representative use of PCPs within a shower. The median values were used for each of the seven VOCs/VOC classes, and the activity was assumed to be as follows. The shower commenced at 07:30 h, with the first 2 minutes spent using shampoo, followed by 2 minutes using conditioner and a further 3 minutes using shower gel. It was then assumed that there was a 3-minute pause to dry off, followed by 2 minutes spent applying moisturizer. The model was then used to explore the mixing ratios that could arise following the shower. Figure 5 shows the concentrations of the primary emissions based on Table 1 and focusing on the period from 07:00 to 08:30 hours.  There is evidence that many people do not use their bathroom fans when showering and certainly not to the extent that ventilation rates would be as high as 9 h −1 . 41 In order to test model sensitivity to this factor, the model runs were repeated using a ventilation rate  Figure 6 at the lower ventilation rate.

| Comparisons against a proof of concept reallife study
The activity assumptions used in Table 2 were assessed during a real-life shower study. Product classes of facewash, followed by F I G U R E 4 Range of potential VOC emissions from various non-aerosol personal care products on an annualized basis covering three activity and frequency scenarios outlined in Table 2 Compound  Table S4). To support the assumption that VOC emission will change linearly based on the amount of product used, a single participant showered three times, using

| CON CLUS IONS
Online mass spectrometry methods have provided a straightforward method to screen for VOC composition and emission amount in a range of different VOC-containing non-aerosol personal care products. While every product has a unique composition, simplified profiles could be reported using seven common VOCs found in most of the samples screened (four alcohols and two siloxanes, and the lumped value for limonene to represent all monoterpenes). Overall, we find that amounts of individual VOCs released vary considerably between products, but are in the range of a few milligrams to a few grams of each VOC from each product per person per year. Shower gels and liquid foundation were found to have the highest rates of VOC emissions, dominated by limonene (representing all monoterpenes) for the former and D5 cVMS for the latter.