Antonia
Praetorius
*a,
Rickard
Arvidsson
b,
Sverker
Molander
b and
Martin
Scheringer
a
aInstitute for Chemical and Bioengineering, ETH Zurich, Wolgang-Pauli-Strasse 10, 8093 Zurich, Switzerland. E-mail: antonia.praetorius@chem.ethz.ch; Fax: +41-44-632 1189; Tel: +41-44-632 6098
bEnvironmental Systems Analysis, Chalmers University of Technology, Rännvägen 6, 412 96 Gothenburg, Sweden. E-mail: rickard.arvidsson@chalmers.se; Fax: +46 31 772 21 72; Tel: +46 31 772 21 61
First published on 7th December 2012
Exposure assessment of engineered nanoparticles (ENPs) is a challenging task mainly due to the novel properties of these new materials and the complexity caused by a wide range of particle characteristics, ENP-containing products and possible environmental interactions. We here present a research agenda in which we propose to face the complexity associated with ENP exposure assessment through informed and systematic simplifications. Exposure modelling is presented as a method for addressing complexity by identifying processes dominant for the fate of ENPs in the environment and enabling an iterative learning process by studying different emission and fate scenarios. Furthermore, the use of models is important to highlight most pressing research needs. For this reason, we also strongly encourage improved communication and collaboration between modellers and experimental scientists. Feedback between modellers and experimental scientists is crucial in order to understand the big picture of ENP exposure assessment and to establish common research strategies. Through joint research efforts and projects, the field of ENP exposure assessment can greatly improve and significantly contribute to a comprehensive and systematic risk assessment of ENPs.
Environmental impactIn this article we present a research agenda that will help overcome the complexity associated with aquatic exposure assessment of anthropogenic nanoparticles. Risk assessment of nanoparticles is still in its infancy, but with the increasing number of nanoparticle-containing products entering the market it is important to develop strategies for assessing the environmental exposure to these new materials in spite of the enormous complexity of the task. Our work contributes to this aim by suggesting methods for informed simplification based on a combination of modelling and experimental approaches. Goals of informed simplification are, for example, that the environmentally most relevant forms of a nanomaterial and the processes that govern the material's environmental fate are identified and further investigated with highest priority. We specifically encourage collaboration and feedback between modellers and experimental scientists, which is essential to gain a well-founded understanding of nanoparticle exposure. |
The sum of these difficulties may seem to indicate that assessing the exposure to ENPs is a task made almost impossible by immense complexity. This complexity is much due to the increased complexity of ENPs as compared to “ordinary” chemicals. It is also due to the complexity of the natural environment that provides a constant challenge for risk assessment.10 Some authors have expressed resignation in front of this complexity, turning away from risk-based methods.3 However, we believe that there is a way to overcome these challenges. In this article, we present a research agenda for exposure assessment of nanoparticles with a special emphasis on informed simplifications and the combination of modelling and experimental methods to achieve this goal. It is our hope that this will provide a structure for future research in the field. We focus on the water compartment, although much of the agenda is general and may prove relevant for exposure assessment in other environmental compartments as well.
The most obvious challenge in ENP exposure assessment originates from the many different properties and types of technical nanoparticle materials on the market, which in turn can be used in a large range of distinct applications. ENPs of the same core material can be present in a large array of sizes and be manufactured in different shapes such as spheres, rods and sheets. Furthermore, depending on the intended application and the manufacturer's preferences, the same core material can be combined with a variety of possible coating materials, greatly influencing its surface properties. In addition, ENPs might be used in their pristine state for certain applications, but will be bound into a matrix in other cases. This results in a potentially infinite number of combinations, constantly increasing with technological development. In addition to the complexity of the ENPs themselves, the complexity of the environment provides a constant challenge for risk assessment, not only of ENPs but also of “ordinary” chemicals.10
Nevertheless, it is possible to simplify the picture in several ways. The key point here is not to become overwhelmed by the details, but to identify main components of the “big picture” and to define what the position of a certain question in the “big picture” is.10 The “big picture” in this context is deriving exposure levels of ENPs through exposure assessment, and the main components of exposure assessment are emissions and environmental fate. The challenges of assessing exposure to ENPs can be fruitfully approached by conducting simplifications that are scientifically informed without losing track of the “big picture”. Below, we provide a number of examples of scientifically informed simplifications with a clear relation to and relevance for the larger picture of assessing environmental exposure to ENPs (Fig. 1).
Fig. 1 Examples of strategies for informed simplifications in ENP exposure assessment. NOM: natural organic matter; SPM: suspended particulate matter. |
An example of a simplifying approach can be to assign different ENPs to distinct groups and types in order to manage the high number of different ENP emissions from the technosphere. The issue of infinite numbers of particle sizes can be addressed by describing the ENPs as a size distribution consisting of finite size classes. The number of size classes will depend on the desired level of resolution; the simplest case of one size class (corresponding to the average or median particle size) can be used when only one average particle size needs to be represented. Using a larger number of size classes approaches a continuous size distribution. Regarding the number of possible coatings, rather than assessing the exposure of every possible ENP-coating combination, coatings can be divided into groups according to their properties and modes of action. Coating groups may for example represent coating characteristics based on steric, electrostatic and electrosteric repulsion properties.11,12
The large amount of ENP product types and applications also leads to a huge number of possibilities with respect to emission scenarios. The way in which ENPs are enclosed in a product matrix will influence their emission rates during use and disposal, and several main types of ENP emissions could represent this. Examples of such types could be ENPs that are chemically stable and ENPs that dissolve rather quickly and thus cease to be ENPs (such as ENPs made from silver and zinc oxide). Arvidsson et al.5,6 made the distinction between two types of ENP emissions from the use phase of products: (1) dissipative emissions of ENPs (such as titanium dioxide ENPs in sunscreen) and (2) ENPs emitted from stocks in the technosphere (such as titanium dioxide ENPs emitted from paint). A third case may be ENPs that are not emitted at all in the use phase but only during waste treatment. Furthermore, an established division of emissions of chemicals is that of point sources and diffuse emissions, which may be relevant for ENPs as well.
Once the ENPs are released from the technosphere and enter the natural environment, one has to consider both their short-term fate directly after release as well as their long-term behaviour, distribution and persistence in the environment. Directly after their release from the industrial product or application, the ENPs are present in a thermodynamically unstable state and may undergo rapid physical and chemical changes after entering a less controlled environment, such as a wastewater treatment plant. In a next step, when the ENPs enter the natural environment, another multitude of processes (such as surface transformations, agglomeration, deposition) acting on the ENPs will influence their fate and transport behaviour.13,14 ENPs will encounter a wide range of natural components, such as suspended particulate matter (SPM), dissolved natural organic matter (NOM), biota, surfactants, and metals, which can potentially interact with the ENPs' surface and affect the ENPs' properties. Instead of trying to describe each possible interaction in detail, it is important to make simplifications in this area as well (Fig. 1). Interaction types can be grouped, such as agglomeration (homo- and heteroagglomeration with SPM or biota), surface interactions (sorption of NOM, surfactants, ions) or transport processes (deposition, flow). Then, dominant processes need to be identified so that research efforts can focus on those first.15
The complexity of processes acting on ENPs is further increased by the fact that the processes do not only depend on the characteristics of the ENPs, but are also strongly influenced by the properties of the surrounding medium. The parameters that have the strongest influence on the ENP behaviour should be identified, for example pH, the concentration of certain ions and concentrations and types of NOM and SPM. Then, typical representative water types can be selected and experiments can be performed in these water types (Fig. 1).
One concrete example of an informed simplification is the speciation of silver ENPs. It is often reasonable to assume that silver species are converted into silver sulphide and that this silver sulphide is the environmentally relevant form of silver ENPs.16 Another example is to omit homoagglomeration from ENP fate modelling. Two different exposure assessment methods have assumed that, because the natural colloids are much more numerous than ENPs in natural waters, homoagglomeration is negligible compared to heteroagglomeration in exposure models.15,17 Recent experimental work seems to support this simplification for many common water types.18
Fig. 2 Contributions of models and experiments to a better understanding of the steps in exposure assessment. STP: sewage treatment plant. |
As a first step the amounts of the ENPs entering the environment need to be assessed. For this purpose, material or substance flow models19 can provide emission estimates by using product inventories and completing them by estimated or measured emission rates from different products.5–7,20,21 Considering the current lack of regulation and declaration of ENP concentration values in products, experimental data on ENP amounts in different relevant consumer products or applications are essential in order to refine mass- or particle-flow models. In addition, experimental validation of emission models could be performed by measuring ENP concentrations at the interface between the technosphere and the environment, for example at the outflow of a wastewater treatment plant. These experimental efforts would inform the development of mass and particle flow models and enable informed simplifications.
Once the ENPs enter the natural environment, we need to understand the transport and transformation processes they undergo to adequately assess their fate. Here, an iterative combination of modelling and empirical work is particularly important. The distinct processes acting on ENPs in different environmental conditions are best studied in controlled laboratory settings, where the influence of individual environmental factors on specific processes can be identified. Then, in order to understand the actual role of the individual processes in the overall fate of the ENPs in environmental systems, the experimental results need to be incorporated into a model. In environmental fate models, the influence of the individual processes on the fate and transport of the ENPs in the environment can be described in detail.15 Thereby, the most relevant processes can be identified and experimental research efforts can be focused on these processes. To ensure commensurability, it is important that modellers and experimental scientists coordinate their research efforts and choose the same metrics and indicators in the models and as output from their experiments. The example mentioned above, where homoagglomeration was shown to be negligible in many natural waters and thus possible to exclude from first approximation exposure models, constitutes a good example of a successful first iteration between modelling and experimental work.18 At a later stage, once analytical techniques are able to measure ENPs in complex natural matrices, analytical campaigns can be used to measure ENP concentrations in the environment and help validate and improve the models, increasing their application domain to more water types.
Another important aspect of combining models and experimental results is the derivation of risk indicators such as the ratio of the predicted environmental concentration (PEC) and the predicted no-effect concentration (PNEC). Environmental fate models can estimate PEC values, whereas the PNECs are assessed in laboratory toxicity tests. As long as environmental ENP concentration ranges cannot be measured experimentally in the environment, modelled concentration ranges for ENPs in different environmental compartments are important information for ecotoxicologists to perform their testing under realistic conditions. Many ENP toxicity studies in the past were performed at unrealistically high ENP concentrations, in the mg L−1 or even g L−1 ranges. First estimates of environmental ENP levels indicate that ENPs are more likely to be present in the environment in ng L−1 to μg L−1 ranges, inciting ecotoxicologists to perform tests at lower ENP concentrations.
In other cases, an endpoint is exposed to a pollutant that originates primarily from one specific product or even one specific phase of a product life cycle. In those cases, the emission model can be limited to the relevant emission sources, thus refocusing the scope of the model. This can also be done if one only wants to know the effects from a certain contribution to the total emissions.
Depending on the system studied and the purpose of the model, the different types of environmental fate models span a wide range. It is possible to model small-scale processes, for example to understand the influence of the physico-chemical properties of a pollutant on a specific environmental process such as sorption of a chemical to soil constituents. On the other hand, mass-flow models have been used to estimate the exposure pathways and quantities of chemicals on a much larger scale, using approximate descriptions of the global environment and the dominant processes.25,26 Between these two extremes, a large variety of environmental fate models can be found, which couple small-scale chemical processes to large-scale environmental processes. These models can predict pollutant's concentrations in local or global settings with different degrees of resolution, depending on the required outcome.25,27–30
The different types of environmental fate models can be used for various purposes. Models aimed at increasing the scientific understanding of the behaviour of a pollutant in the environment can be parameterized with a high degree of detail and an incorporation of complex process descriptions to represent the real world settings as accurately as possible.31 On the other hand, environmental fate models can also serve as an important source of arguments in decision-making on regulatory questions.32 For this purpose, it is important to keep the process descriptions simpler in order to obtain a tool that can be understood and handled by non-scientists as well,33 without losing a significant amount of information or predictive capability.
Several environmental exposure models for ENPs in aquatic environments have already been developed. These models span a range of different levels of complexity and degrees of detail, and they serve different purposes. One way to categorize them is to divide them into (1) bottom-up mechanistic modelling and (2) top-down modelling using partitioning factors. The first category includes models presented in the studies by Arvidsson et al.,4 Praetorius et al.15 and Quik et al.17 In these models, knowledge from the natural sciences, for example colloid chemistry and chemical kinetics, is used to describe the fate of ENPs in water. Mechanisms deemed relevant based on experimental studies, such as agglomeration, sedimentation and dissociation, are included in the models. Although the included mechanisms differ somewhat between the models, the model equations are all such that additional mechanisms can be added. A general way to describe these models mathematically could be
A clear strength of bottom-up mechanistic modelling is that it is possible to incorporate more detailed models of fate processes once they are available. In this respect, these models are truly science-based. However, a challenge with these models is to model processes that have not yet been deciphered by experiments. Another challenge is that some parameters vital to the model can be difficult to obtain relevant values for. One such parameter is the attachment efficiency, describing the probability of a collision between two particles resulting in attachment, normally symbolized with the Greek letter α. αhomo for the homoagglomeration of ENPs can be measured in a straightforward manner (a number of studies on homoagglomeration are summarized in a review by Petosa et al.34). On the other hand, determining values for αhetero describing the heteroagglomeration of ENPs with natural suspended particulate matter (SPM) is much more challenging. Studies on ENP heteroagglomeration are only starting to appear,18,35,36 and reliable methods for determining αhetero have not been established yet. As a result, modelling studies had to use a range of αhetero values representing different scenarios rather than values specific for the water type investigated.4,15 The mechanism of heteroagglomeration thus constitute a specific case that requires creative collaboration between modellers and experimentalists in order to develop models based on informed simplifications.
Top-down modelling of ENPs in the aquatic environment has been conducted by several authors.6,16,37–42 These studies focus more on the emissions part of the exposure assessment and instead of departing from detailed mechanisms of the fate of ENPs, they view the water compartment as a black box. Some of the ENPs entering the water will stay there, whereas some of the ENPs will be transported to other compartments (air, soil, sediment) or become degraded. Exactly by which fate mechanisms this takes place is, however, not described in detail in these studies. A general way to describe these models mathematically could be
The advantage of the top-down approach to exposure modelling of ENPs is that in-depth mechanistic insights into the fate of ENPs are, in principle, not necessary. Partitioning factors can be derived based on rather simple experiments. If an ENP is removed from the water by several different, complex processes simultaneously, it may be very difficult to model this removal mechanistically. However, the lack of mechanistic understanding is a problem also for the top-down approach, particularly in a relatively new field such as ENP exposure assessment, where significant processes might not yet have been understood. If the conditions in the aquatic environment are different from those in the experiments where the partitioning factors were derived (e.g., different pH and concentration of ions), it is unclear to which extent the experimental results are relevant for the conditions of the study. It is thus important to be clear about under which environmental conditions the partitioning factor is valid, which is not always the case in the cited studies. The top-down modelling may thus require a large amount of experimental measurements of partitioning factors under different conditions.
As there are pros and cons with both bottom-up and top-down modelling, an important issue for both modellers and experimentalists is to identify the approach that is most useful in specific cases. It may very well be understood that they are both useful, but in different ways. For example, top-down approaches might be the only feasible method to date for assessing the emissions of ENPs from the technosphere to the environment, because extremely little information on specific ENP processes is available here.7 To derive PEC values on a more local scale and determine most relevant ENP forms in specific environmental compartments, bottom-up approaches including specific processes are necessary. In many cases, a combination of top-down emission models followed by bottom-up fate models or the incorporation of certain detailed fate processes in a larger scale emission model might be the most successful approach for holistic exposure assessment of ENPs.
An interesting difference between the bottom-up and top-down modelling approaches is the use of different exposure indicators. Mass concentration is used in the top-down models and some bottom-up models, whereas particle number concentration is used in some of the bottom-up models. The reason for using particle number instead of mass may be because particle number has been suggested to better reflect the exposure situation and correlate with toxic effects from the ENPs.43 But even more pragmatically, the modelling of some fate mechanisms, such as agglomeration and sedimentation, is typically based on particle number concentration as a unit. Consequently, the choice of exposure indicators is much linked to the modelling approach and the detailed process understanding and thus constitutes another important issue for future collaboration between modellers and experimentalists.
The differences, possibilities and limitations of different modelling types need to be made transparent to the scientific community, so that appropriate input data can be generated for the different models and their predictions can be interpreted in the correct way. Detailed fate models require process-specific input data, such as attachment efficiencies for agglomeration processes, whereas large scale mass- or particle-flow models need more generic partitioning factors, which could for example be obtained by batch-processes simulating a sewage treatment plant (STP).
The most urgent research needs for the exposure assessment of ENPs regard:
• concentrations in and release rates from consumer products and applications
• identification of most relevant ENP types used and focusing research efforts on these products first
• modelling studies that identify dominant environmental processes affecting the fate of ENPs and, at the same time, consider the interplay of various processes in the environment
• laboratory studies focusing on dominant processes under environmentally relevant conditions (e.g., hetero-agglomeration of ENPs with naturally occurring suspended matter)
• consensus between modellers and experimentalists on the necessary output parameters of laboratory studies to be used in models (e.g., attachment efficiencies for agglomeration rather than concentration-dependent aggregation rates)
• validation of models with studies designed by modellers and experimental scientists together (e.g. field or mesocosm studies, fate of ENPs in model STPs)
• development of methods for identifying environmentally relevant forms of different ENPs and then focusing toxicity studies on these ENP forms
• further investigation on the pros, cons and relevance of bottom-up and top-down exposure modelling approaches for ENPs and on mass and particle number concentration as exposure indicators.
With these points in mind, we believe that the field of exposure assessment will be able to move forward significantly and strongly contribute to a well-founded risk assessment of ENPs.
This journal is © The Royal Society of Chemistry 2013 |