ENV/JM/MONO(2015)17/ANN13

Transcription

1 Unclassified ENV/JM/MONO(25)7/ANN3 ENV/JM/MONO(25)7/ANN3 Unclassified Organisation de Coopération et de Développement Économiques Organisation for Economic Co-operation and Development 3-Oct5 English - Or. English ENVIRONMENT DIRECTORATE JOINT MEETING OF THE CHEMICALS COMMITTEE AND THE WORKING PARTY ON CHEMICALS, PESTICIDES AND BIOTECHNOLOGY DOSSIER ON TITANIUM DIOXIDE - GENERAL ANNEXES - ANNEX 3 Series on the Safety of Manufactured Nanomaterials No. 54 This document is only available in PDF format. English - Or. English JT Complete document available on OLIS in its original format This document and any map included herein are without prejudice to the status of or sovereignty over any territory, to the delimitation of international frontiers and boundaries and to the name of any territory, city or area.

2 ENV/JM/MONO(25)7/ANN3 2

3 Project Report Testing the OECD selected alternative nano-tio 2 materials for dispersion stability, environmental behaviour and fate In support of the OECD WPNM Sponsorship Program University of Vienna Department of Environmental Geosciences Dr. Frank von der Kammer Prof. Dr. Thilo Hofmann

4 Content. Introduction Alignment of the work plan with the OECD sponsorship program and the lead sponsor Results (summary) Achievements in the framework of the Project The test procedure for dispersion stability and environmental fate General aspects of the test layout and experimental approach Storage of the particles and dispersion routine (including the decision point ENP is dispersible in environmental media) Static stability tests under controlled conditions (3-36 different hydrochemical combinations per material) Nanoparticle analysis Data evaluation and plotting to produce stability, zeta potential and size maps 4 3. Results from the matrix tests for the alternative TiO 2 materials Influence of natural organic matter (NOM) The sodium chloride matrix: influence of a monovalent cation (with monovalent counter ion) The calcium chloride matrix: influence of a divalent cation (with monovalent counter ion) The sodium sulfate matrix: influence of a divalent anion (with monovalent counterion) Assessment of the associated risk for natural water bodies in Europe Conclusions References Annex Page 2

5 . Introduction Growth in the production and application of engineered nanoparticles (ENPs) will inevitably lead to increased emissions into the environment (Handy et al. 28; Gottschalk et al. 2). One important distribution pathway for most ENPs into the environment is thought to be through wastewater treatment plants (Kiser et al. 29; Kaegi et al. 28). Such emissions will then impact on the aquatic environment, except for the fraction that is retained in sludge, which, if later used in agriculture, may impact on agricultural soil systems and groundwater (Brar et al. 2). Detailed understanding of the behavior on ENPs in aquatic environments is crucial for a comprehensive assessment of their final distribution in the environment and the associated risks (Handy et al. 28; Alvarez et al. 29; Fang et al. 29). The behavior of nanoparticles in general (NPs) in the environment is still not well understood (Lead & Wilkinson 26). Numerous studies have been carried out to elucidate the potential toxicity of ENPs (Handy et al. 28; Hartmann et al. 29; Battin et al. 29). Much less attention however has been given to the exposure assessment of ENPs and methods for the measurement of realistic ENPs concentration in the environment are limited or missing at all. Besides, in order to support exposure models and to enable reproducible toxicity tests, one needs a precise understanding of the mechanisms determining the colloidal stability and related aggregation processes of ENPs in the aquatic environment (v.d Kammer et al. 2; Gao et al. 29; Ji et al. 2). Studies on the dispersion stability of ENPs under conditions that mimic real-world aquatic chemistries should be an essential part of their general characterization (v.d. Kammer et al. 2). The results of such studies would enable the transport and fate of ENPs within natural aquatic systems to be predicted and the results of toxicological tests to be accurately evaluated and compared (Ottofuelling et al. 2). Recent studies have made use of various analytical techniques and experiments to describe and understand the behavior of ENPs in synthetic growth media, natural waters, waste water, and near-natural and well-controlled synthetic waters (see introduction in v.d. Kammer et al 2; Gao et al. 29; Keller et al. 2). Results have indicated that the behavior, transport, and fate of nanoparticles in aqueous systems are controlled both by their surface properties and by the chemistry of the aqueous system, but also that the investigated systems are highly complex and not yet understood in detail. Factors such as electrolyte concentration, the valence of the ions countering the surface charge, the of the system, and the presence and type of organic matter, will cause either aggregation or stabilization of ENPs. Studies on bare and uncoated ENPs cannot be transferred directly to functionalized or capped ENPs, even if they are composed of the same core material due to the different physico-chemical properties of the surface/capping agent. Indeed, the colloidal stability of capped ENPs is not only a function of environmental parameters (e.g.,, ionic strength, electrolyte type), but also depend highly on the nature of the capping material of the ENPs. With regard to the different alternative materials of TiO 2 ENPs which are subject to testing in the OECD program, it was therefore predominantly important to directly compare the bare/uncoated TiO 2 NPs with the coated versions. In order to understand the behavior and characteristics of ENPs in natural aquatic systems researchers have conducted experiments in both complex natural and synthetic waters. Each of the chosen test systems has its advantages and disadvantages. Page 3

6 While results from experiments using real natural waters may be highly realistic, they are often unable to provide information on the processes involved due to the complexity of the water chemistry. Results from synthetic waters on the other hand, with their reduced complexity, can help in understanding the basic principles but are not necessarily representative of natural systems. Furthermore it is important if the behavior of the ENPs in the experiments is continuously monitored or just a stability indicator is given. Quantitative descriptors can be deduced from so called dynamic, continuously monitoring, studies that, depending on the experimental approach used, determine doublet formation or particle growth rate constants, or sedimentation rates. Such experiments are typically very time consuming and hence cover only few hydro-chemical conditions (see also fig. for an overview). Only when applying a wide range of typical concentrations of different water constituents and typical may finally help to understand the behavior of ENPs within aquatic systems. Hence the test system needs to be simple so that a huge number of different conditions can be tested (v.d. Kammer et al. 2). The main objective of this project was to assess the suitability of two OECD endpoints for commercially relevant nanoparticles. Subsequently possible adaption of those endpoints to nanoparticles should be investigated if this would improve the set of OECD endpoints for testing of chemicals towards the testing of nanomaterials. In the framework of the here described project the specific OECD endpoints dispersion stability / environmental behavior and surface water testing were addressed. The two endpoints were tested for commercially available, high volume production titanium dioxide nanoparticles. These two procedures were not available for nanoparticles or not adapted to the intrinsic properties of () nanoparticles in general or/and (2) of the tested titanium dioxide nanoparticles in detail. Within the OECD Working Party of Manufactured Nanomaterials (WPNM), SG3 developed a testing scheme for titanium nanoparticles (Dossier Development Plan (DDP) Documents of the OECD) and a lead material was selected. Lead material for TiO 2 is P25 from Evonik and will be investigated in detail for all (or most) OECD endpoints. Additionally alternative materials were selected which have physico-chemical properties distinct from P25, are used for different purposes or represent a non-nano form of TiO 2 used as a negative control. These alternative titanium dioxide nanostructured materials are Hombikat UV, UV-Titan M22, UV-Titan M262 and PC-5, all from Sachtleben Chemicals and as an non-nano reference Tiona AT- from Crystal Global (table ). Page 4

7 Table : overview of materials and their physic-chemical properties material manufacturer production method composition primary particle size (nm) typical size in dispersion BET surface area comment P25 Evonik gas phase hydr. anatase & rutile 25 ~ 3 8 m2/g lead material Hombikat UV anatase (7 5) ~ 3 34 m2/g hydrophilic Titan M22 sulfate process, rutile 2 ~ m2/g coating Sachtleben precipitation hydrophobic Titan M262 rutile 2 ~ m2/g coating PC 5 anatase 5 25 ND 5 2 m2/g Tiona AT ND = not dispersible Cristal Global sulfate process, precipitation anatase 2 22 ND 8 m2/g non nano The testing procedure for dispersion stability / environmental behavior was developed using P25 first to identify pitfalls of the methodology and then test P25 and the alternative materials in a manner that a direct comparison becomes possible. The OECD was responsible to provide test materials which were approved, characterized and handled in a way that all groups working on the program were assured to have identical materials available for their experiments (usually coming from the same batch providing homogeneity in parameters as size and composition). The materials became successively available during autumn 29 and spring 2. The initial development of procedures for stability testing (v.d. Kammer et al. 2) and surface water transport testing (Batti et al. 29) was performed earlier on random production batches of the two materials (P25 and Hombikat UV). After the OECD approved batches of the test materials were available Evonik P25 was investigated in detail and compared to the random batches to prove comparability of the results from the non- OECD batches. In the following part the preparations for the experimental program, alignment with the other groups in the TiO 2 OECD program and the requirements of the OECD program and the results of the tests are described in detail... Alignment of the work plan with the OECD sponsorship program and the lead sponsor To prevent the duplication of work or efforts not supporting the OECD program a telephone conference was arranged for the 29th September 29 by the lead sponsor (UBA Germany). The international groups working on the TiO 2 materials on environmental fate were participating. Discussion was very constructive with support from Nathalie Tufenkji (McGill University, Canada), a well-known expert on nanoparticles transport in soil and groundwater. A common protocol however could not be established due to different focus of the intended work plans. According to the project description the University of Vienna was asked to look into aggregation in detail and develop a new test protocol, while McGill University s focus was thought to be on subsurface / groundwater transport. Page 5

8 There was a critical mismatch identified between the surface water testing planned at University of Vienna and the protocols of the OECD. For the surface water testing the OECD requests a chemical degradation experiment. All experts in the telephone conference judged this as not relevant for a virtually non-soluble, highly inert material as TiO 2. The decision reached in the conference call was to use limited resources to develop alternative testing strategies which incorporate the special physical properties of nanoparticles instead of performing experiments inadequate for the material in question. University of Vienna therefore developed the outlined multi-parameter matrix testing to achieve a predictive tool for aggregation behavior and mobility..2. Results (summary) In summary the alternative materials have been successfully investigated for their dispersion stability and environmental behavior with a new test protocol. Clear differences and similarities between materials could be demonstrated and linked to their core, their surface chemistry (coating) and even to the possible degradation of the surface coating. The developed testing scheme consists of a synthetic, semi-automated routine, adapted data treatment and comparative procedures to relate the results to environmental conditions. The procedure and results for the lead material and alternative materials were presented on the OECD sponsorship meeting for TiO 2 in January 2 as a keynote presentation. While the comparison of the random production batch of the lead material P25 (not part of this project) with the OECD batch showed very similar behavior for both materials, the alternative material Hombikat UV showed drastic differences between the random batch and the OECD material. The reasons for these differences could not be identified yet..3. Achievements in the framework of the Project a) Development of the synthetic multidimensional test routine for dispersion stability and environmental fate. This included the preparatory testing to find suitable concentrations of the test material (25 mg/l), optimizing the programming of the Metrohm auto-titrator, finding optimal spacings for the and concentration of electrolytes, adapted data treatment and visualization. b) Testing for dispersion stability and environmental fate was completed for all five alternative materials and results in preparation for publication. Work is considered fully supportive to the OECD program. c) By using the results from the proof of principle study and the in-depth testing of the lead material P25 the testing scheme was significantly improved ) by adding a natural buffer to stabilize (bicarbonate), 2) by now using 48 individual measurements per map instead of 2 triplicates or 2 points with 5 repetitions each Page 6

9 e) The full set of alternative materials could be tested in a fully comparative manner. The results show clear differences between the non-oecd batch of Hombikat UV tested during method development and the OECD batch tested in the framework of this project. f) The two Titan M materials reacted highly similar in the stability tests although one was coated to behave more hydrophilic, the other was coated to react more hydrophobic. d) By comparison of the results with European surface water analyses it becomes possible to predict the water bodies which will promote the stability of the tested TiO 2 ENPs (risk map). e) Considering the results and the connection to the surface water data we are now for the first time able to plot detailed risk-maps for ENP dispersion and transport in European surface waters. Page 7

10 2. The test procedure for dispersion stability and environmental fate 2.. General aspects of the test layout and experimental approach The general idea of the matrix testing is to quantify the aquatic behavior of a test particle (here TiO 2 ) under a broad set of realistic environmental conditions by using simple automated dispersion stability tests (so called static tests). The considerations regarding dynamic aggregation experiments and static stability measurements are given in figure together with the advantages and disadvantages of real-world testing (natural waters) and the more generic synthetic test matrix. Fig. : contrasting the main characteristics of static and dynamic aggregation experiments which are performed to derive a measure of dispersion stability. Also given are the limitations of using natural water samples to test the dispersion stability in real waters versus the potentials of using a synthetic multiparameter testing. In the proposed test system a stable dispersion of the to-be-tested particles is separated into 3-45 subsamples and each is then subjected to a different hydrochemical condition. This results in a three dimensional matrix of dispersion stability over and ion concentrationby applying different salts as NaCl, CaCl 2 etc. a set of matrices is obtained, the results become multi-dimensional. The principle of the test is based on a typical industry quality control test. A product containing a dispersion of particles or molecules (e.g. a paint) is checked for phase Page 8

11 separation after a given time span. Sometimes temperature is increased to speed up the undesirable reaction (artificial aging). In the described test a stable dispersion is subjected to a change in water chemistry and the phase separation (aggregation and settling of the particles) is measured once after a given time period. After a reaction and settling time of ~ 2 h 5 h the remaining particle concentration and zeta potential is determined in the supernatant (Fig. 2). In the following the test system with all the individual steps to characterize a sample is described in detail and results are given for the lead TiO 2 material and the alternative materials. The test consists of the four consecutive procedures: ) Dispersion of the particles in ultrapure water preparation of a stable dispersion. 2) Subjection of the stable dispersion to different and ionic strength conditions. 3) Transferring the obtained data into double logarithmic plots (ionic strength versus ). 4) Overlaying the obtained plots with available freshwater data of Europe for purposes of risk assessment Storage of the particles and dispersion routine (including the decision point ENP is dispersible in environmental media) Storage A fraction of the test material is stored as dry powder sample (typically 5 g) in an exsiccator on a Petri dish over a saturated NaCl solution at 2 C to keep the relative humidity constant at 75%. This ensures a pre-wetting of the surfaces and keeps the samples in comparable and stable humidity. Dispersion routine Weighing of 5//25mg in L MQ water, adjusting to with either mol/l HCl or NaOH, 3 sec. ultrasonic bath treatment, followed by a wetting time of 24 hours Using ultrasonic bath to disperse the particles (2W output, constant, 6 min.), adjusting to as before. Caution: dilution of the dispersion may cause a (sometimes reversible, short-term) aggregation (Hartmann et al. 29). Decision Point dispersible in environmental media yes / no: Page 9

12 If the dry particle powder cannot be well dispersed by the above procedure then the procedure is repeated in MQ water with a of 4, 6, 8,, with and without containing 25 mg/l NOM (Suwannee River NOM) (results in 8 tests in total). If the nanoparticles still aggregate and settle in short time (2 h) the further testing is abandoned because the particle is in its pristine and as well in the NOM coated form not likely to disperse in water. Not dispersible is here defined as visibly clean, non-turbid supernatant above a turbid zone or a sediment, 2 h after mixing and ultrasonication. It may be necessary to further adapt this decision point for very small particles (< 5nm) which do not produce a visible turbidity even at elevated concentrations. Due to long term effects it may be advisable that TiO 2 particles which fail this primary test, being non-dispersible, are further aged in water under light for some weeks to check for surface modifications through water contact and daylight which may make the particles more hydrophilic. The before described dispersion test should be repeated then with the aged material. Determination of any dissolved compounds should also accompany this aging test. This aging procedure was not performed with the OECD alternative TiO 2 particles in this test; although two materials failed the first part of this routine and qualified as non-dispersible in water Static stability tests under controlled conditions (3-36 different hydrochemical combinations per material) The general idea behind the multi-parameter stability testing of ENPs (matrix testing) is to subject a stable aqueous dispersion of the to-be-tested nanoparticles to many different water-chemical conditions and different and analyze the particles which remain stable dispersed in the supernatant for concentration, zeta-potential and particle size. According to the particle properties and the water chemical conditions, the nanoparticles will undergo differently rapid aggregation and settling processes. Hence after a well-chosen reaction period the particles in the supernatant are indicative of the speed of the aggregation process. The chosen conditions should be realistic and hence represent typically occurring water chemistries in the natural environment. Physico-chemical background An aqueous dispersion of nanoparticles is not in equilibrium. The equilibrium is reached when all particles are aggregated and surface area minimized. To form an aggregate the particles must first collide with each other and the collision must result in the sticking of particles to each other. The rate of collisions is determined by many factors, like particle number concentration, temperature and viscosity of the medium. Additionally the size and size distribution, particle shape and any shear/turbulent flow of the medium influence the number of collisions per unit time (Gregory 25; Lead & Wilkinson 26). Under so called favorable conditions, particles which collide with each other will to % stick to each other. The system is destabilized and aggregates under diffusion limited conditions in which only the transport of particles to each other and the number of particles present determine the velocity of the aggregation. Page

13 C ENP size ENP ENP C ENP size ENP ENP concentration (aqu. species) CaCl 2 NOM [ ] NaCl CaCl 2 NaCl different types of aquatic species monovalent divalent cation Na + Ca 2+ anion Cl - SO 2-4 buffer NOM bicarbonate/carbonate Suwannee River NOM in total 6 test matrices with 36 datapoints: NaCl CaCl 2 Na 2 SO 4 CaSO 4 NOM (NOM + CaCl 2 ) range 4 8 Fig. 2: Principle of the matrix testing, in red circles the identified master variables. On the other side, if particles repel each other and only a fraction of the collisions between particles are ending up in particles sticking to each other, then the aggregation is slowed down and may even virtually vanish. To repel each other the particles need to build up an energy barrier which approaching particles have to overcome to finally stick to each other. This energy barrier is often of electrostatic nature (two negatively charged or two positively charged particles repel each other). Many other forces that hinder the particles to approach each other very closely are however known and still not fully implemented in mechanistic models which would then be able to describe the process of aggregation quantitatively and based on first principles (Grasso et al 22). One of those forces is steric hindrance which is introduced by large molecules which adsorb or bind to the surface of nanoparticles and provide a repelling mechanism which can be seen as a kind of cushion effect. Situations in which only fractions of the collisions between particles end up in attached particles are socalled stabilized and particles aggregate in a reaction limited regime since the kinetic energy of the particles and the height of the energy barrier control the velocity of the aggregation reaction. If particles finally overcome the energy barrier and manage to come very close to each other (fractions of or even less than a nanometer) then short-range attractive Van der Waals forces take over and the particles will stick. The stability of a nanoparticle dispersion, is hence the resistance against a very close approach of the particles by means of an repulsive energy barrier. The height of this barrier is however depending on the surface charge of the particles, any surface adsorbed ions and molecules and the ionic strength and ionic composition of the surrounding water. Especially the valence of the ions countering the surface charge is of Page

14 great importance with the di- and trivalent ions being far more effective (Schulze-Hardy rule). The plays a role for particles with a -variable surface charge (as TiO 2 ) as it will cause protonation (low, positive surface charge) or deprotonation (high, negative surface charge) of the surface hydroxyl groups of the TiO 2 particles. In the absence of potential determining ions the surface charge becomes zero when the surface hydroxyl groups are saturated with protons, which then is also the iso-electric point IEP. Under these conditions and in the absence of any stabilizing surface coatings the electrostatic repulsion between particles vanishes and the particles aggregate rapidly (diffusion limited regime). Fig. 3: Evolution of the matrix testing layout from preliminary proof-of-principle experiments (stage ) to the final test procedure (stage 2). Grids are indicating the number of variations in ionic strength (y-axis) and (x-axis), numbers in the matrix fields give the number of replicate per condition. Numbers in circles give the total number of tests per matrix. Red text indicates drawbacks of the individual layout, underlined text the main reason for changing the layout. Test design Relevant aquatic species which have a critical influence on the energy barrier and aggregation behavior of nanoparticles are Na +, Ca ++, Cl -, SO 4 -- which should be applied in all four combinations to cover the possible -,, 2, 2- valence combinations. However extensive tests with CaSO 4 and P25 have shown that this combination de-stabilizes the dispersions at all tested conditions (Ottofuelling et al. 2). Hence this combination was not tested on the alternative materials in detail. Special attention was given to the question of parameter spacing ( and ionic strength range) and the number of replicates per test set (matrix) (fig. 3). The different investigated layouts had their advantages and disadvantages. In principle a stepwise improvement was achieved which finally culminated in the proposed layout which was also used for all tests of the lead and the alternative materials. Apart from the automation two major improvements Page 2

15 were introduced in stage 2 and in the transition to stage 3: first the addition of sodium bicarbonate as a buffer, which occurs naturally in all surface waters, reduced the variability of and deviations from the titrated value and second the move from lowresolution with many replicates per point to a high resolution with no replicates. Accuracy is then provided by the large number of close-by neighboring points instead of replicates. Titration procedure Stable particle suspensions containing 5 mg/l solid matter are subjected to the selected conditions by an autotitration procedure which first doses the calculated amount of electrolyte, titrates to the set value and adds as much water as needed to reach a 25 mg/l dispersion. The vessels are then taken out of the sample rack. After a reaction/aggregation and settling time of 5 hours the supernatant is analyzed for particle concentration, particle size and zeta potential. II C concentration in supernatant sampling time ~ 5 h B A sampling time ~ - 2 h time of experiment sampling time ~ 24 h stable conditions (e.g. MQ water) intermediate conditions (e.g. low NaCl) destabilized conditions (e.g. CaCl 2 ) A can be distinguished from B & C A & B & C can be distinguished C can be distinguished from A & B Fig. 4: Graphical representation of the sampling time adjustment. Only in a certain sampling time window it is possible to distinguish the different conditions from stable dispersion (C), medium stable (B) to unstable (A). An auto titration system (Metrohm Titrando 836) equipped with four individual dosing units (Metrohm Dosino 87) and 2 ml burettes (Metrohm Dosino 8) was used for the automated manipulation of the samples, together with a robotic sample processor (Robotic sample Processor XL 85) and a fast reaction electrode (Metrohm Aquatrode Plus). Each of the 25 ml polyethylene titration vessels (sample vials) were dosed with 25 ml of the stock TiO 2 suspension. The electrolyte or NOM solutions were then added automatically, followed by Milli-Q water to reach the desired final particle concentration at a total volume of 5 ml. NaOH and HCl were used to adjust the in the test sets without a buffer; bicarbonate was used in the -buffered tests. A electrode calibration was performed before start, after half of the samples, and at the end of each experimental set. The drift was equal or below. units. All vessels were sealed with plastic lids once the parameters Page 3

16 had been adjusted. Aggregation/sedimentation was allowed to occur for 5 h in non-stirred vessels. After 5 h a sample of ml was taken through a Ø mm hole in the lid of the vessel at a depth of exactly 2 cm using a stainless steel needle and a syringe. The reason for choosing a 5 h reaction period can be seen in figure 4. The sampling should occur after the fast aggregation reaction has leveled out to a nearly constant low concentration in the supernatant and before the slowly aggregating dispersion has left its stable plateau. A period was chosen in between both limits in order to resolve intermediate aggregation rates, and also because of practical concerns of sampling and measurements to be performed during the following day. The of the remaining sample ( 5h ) was measured again and this was further used for data visualization. In the first proof of principle study (v.d. Kammer et al. 2, stage 2, see annex ) the experiments were performed deliberately on unbuffered systems due to concerns of buffer induced reactions through elevated ionic strength and possible specific adsorption of buffer components to the particle surface. The possible impact of an added buffer on experimental results is shown by the well-buffered sodium diphosphate set (positive control used in the proof-of-principle study, annex ), where adsorption of the diphosphate leads to strong stabilization under all applied conditions. However the unbuffered testing was abandoned due to the fact that a natural buffering component, bicarbonate, is present in virtually all surface waters and the addition of bicarbonate improves the stability in the tests Nanoparticle analysis The supernatant taken from the reaction vessels 2 cm below the surface after 5h was analyzed for nanoparticle concentration, particle size and electrophoretic mobility. Particle size and electrophoretic mobility were determined using a Zetasizer Nano ZS (Malvern Instruments, UK). Particle size analysis proved unreliable and little helpful in interpreting the results. The z-potential was calculated from the electrophoretic mobility using the Smoluchowski approximation. Each z-potential data point is the average of 2 repeated measurements of the same sample. The concentration of nano-tio 2 was determined by measuring the nephelometric turbidity (Hach 2 N IS Turbidity meter). The NTU values reported by the turbidity meter were calibrated against known concentrations in a TiO 2 dispersion stabilized by Na 4 P 2 O 7. The addition of NOM does not increase the turbidity in the sample as the NOM is far smaller than the average particle size of measured dispersions Data evaluation and plotting to produce stability, zeta potential and size maps A set of data visualized as a contour plot with the on the x-axis, electrolyte concentration at the y-axis and the zeta potential, supernatant concentration or particle size on the z-axis is defined as a matrix. For each type of electrolyte and the NOM three matrices for z-potential, size and concentration are produced. However, the matrices plotting the size of the Depending on the experiment different datasets were recorded supporting each matrix. For the first proof-ofprinciple study a relatively small set of different and ionic strength values was tested in Page 4

17 each matrix (4 datapoints in y-axis, 3 in x axis) but each single datapoint was supported by 5 replicates, resulting in total 6 measurements per matrix. For the in-detail study of the lead material P25 63 datapoints were generated (7 datapoints in y-axis, 9 in x-axis). However with 63 datapoints a single set cannot be produced with one tray filling of the currently used autosampler. With the results of previous experiments (v.d. Kammer et al. 2) and taking the requirements of realistic routine testings into account, the final test system, as applied to the alternative materials, consists of 6 X 8 single datapoints, 6 in ( ) and 8 in ionic strength (in logarithmic spacing). Since is already a logarithmic derivate of the proton concentration, the plots are of a double logarithmic stile for x- and y-axis. Linear plotting of the concentration, z-potential or size on z- axis was applied. The matrices (concentration, particle size and zeta potential) were plotted as interpolated surface contour plots processed by the software Surfer 7. (Golden Software, Inc.) with the inverse distance weighting algorithm. The interpolated surface is a distance weighted average of the observed parameters, i.e. with increasing distance between data points their relative contribution to the interpolation declines with distance from the grid node according to: Ζ n i j n i h Z i ß ij h ß ij with h ij d 2 ij δ 2 where h ij is the effective separation distance between grid node j and the neighboring point i. Z j is the interpolated value for grid node j; Z i are the neighboring points; d ij is the distance between the grid node j and the neighboring point i; β is the weighting power and δ is the smoothing parameter which was set to.2. This provided less erratic distortions of the surface plot without loosing detail. 3. Results from the matrix tests for the alternative TiO 2 materials In the following the matrix plots of the alternative TiO 2 materials are shown together with the lead material P25. From the alternative materials the Sachtleben Chemicals products Hombikat UV, UV Titan M22 (hydrophilic) and UV Titan M262 (hydrophobic) were easily dispersed in water according to the routine described earlier. The product of Sachtleben Chemicals PC5 and the negative control/non-nano TiO 2 product of Crystal Global Tiona AT- were not dispersible according to the routine described earlier Influence of natural organic matter (NOM) Page 5

18 In fig. 5 it is clearly visible what a dominating influence NOM has on the stability of titanium dioxide nanoparticles. Above mg/l DOC (~2 mg/l NOM) the stability and the z-potential are nearly constant over the entire range. Only below mg/l of DOC there is an effect with to be observed with both UV Titan materials. NOM stabilizes all materials if present above. mg/l DOC, the effect increases with concentration up to about mg/l DOC and from this on remains constant. Hombikat UV UV Titan hydrophobic UV Titan hydrophilic Evonik P25 log mg/l DOC (SRNOM) mg/l DOC stable concentration: C/C in % 5 log mg/l NOM (SRNOM) zeta potential im mv Figure 5: Effect of natural organic matter on dispersion stability and z-potential of the TiO 2 nanopaticles. Stability/concentration (top) and z-potential (bottom) -matrix plots for the respective material. Higher colloidal stability and hence less aggregation is reflected by a higher concentration (brighter shade of the contours) in the dispersion after the reaction and settling period (5h). Elevated absolute values of the zeta potential support the colloidal stability through higher electrostatic repulsion forces between the particles (+/+ or -/-). Hence high absolute values are plotted with bright shades, less magnitude with dark shades (instability, tendency to aggregate). An important observation is that the two UV Titan materials behave nearly identical; especially the zeta-potential plots are so similar that we would argue that the original hydrophobic and hydrophilic coating (see table ) does not play a role here. This may be interpreted as an overcoating of the original coating by NOM, but the results of the other experiments will allow also for a different interpretation: the replacement of the organic coating or the loss of the original coating already in the dispersion and dilution process.. Another striking observation is that Hombikat UV is lost to about 5% from the dispersion, independent of and NOM addition. The z-potential plot suggests a effect on the surface charge of the particles, which is not present in this way with the other materials and Page 6

19 also not observed with the non-oecd variant of the material (random production batch) tested earlier (v.d. Kammer et al. 2) The sodium chloride matrix: influence of a monovalent cation (with monovalent counter ion) The following matrices (fig. 6) show the influence of a non-adsorbing, hence indifferently reacting, monovalent cation and anion (Na + and Cl - ). The zeta-potential plots of both UV Titan and the P25 are underpinning the nearly indifferent (not specifically interacting) character of Na +. The points of zero charge (PZCs) are only slightly shifting towards higher values with increasing NaCl concentration. The PZC of Hombikat UV lies outside of the range tested (< 4) what appears quite unusual for an uncoated Titanium dioxide material. Hombikat UV UV Titan hydrophobic UV Titan hydrophilic Evonik P25 log mm NaCl 2 23 mg/l Na stable concentration: C/C in % log mm NaCl zeta potential im mv Figure 6: Effect of NaCl on dispersion stability and z-potential of the TiO 2 nanopaticles. Stability/concentration (top) and z-potential (bottom) -matrix plots for the respective material. Higher colloidal stability and hence less aggregation is reflected by a higher concentration (brighter shade of the contours) in the dispersion after the reaction and settling period (5h). Elevated absolute values of the zeta potential support the colloidal stability through higher electrostatic repulsion forces between the particles (+/+ or -/-). Hence high absolute values are ploted with bright shades, less magnitude with dark shades (instability, tendency to aggregate). Page 7

20 Again, both the UV Titan materials are clearly reacting different to P25 but very similar among each other. This may be due to the fact that both are coated with Aluminum oxide, what makes them react more like an Aluminum oxide (AlOx) with typically higher PZC between 8 and 9 than a TiO 2 (Appelo & Postma 999). Hombikat UV shows negative z- potentials at all NaCl concentrations and values. Again for the Hombikat UV material, the zeta-potential results are clearly different to the former tests of the non-oecd batch (v.d. Kammer et al. 2) and not easy to explain. To document the differences between a random batch of the Hombikat UV- material and the UV used in the OECD testing, figure 7 shows the results for the NaCl matrix with the random batch material used for the proof-of-principle study (v.d. Kammer et al. 2). It has to be noted that the test matrix in figure 7 was derived with less resolution and covering a narrower range. Additionally it did not contain bicarbonate as a buffering agent to stabilize as the tests in figure 6. The adsorption of carbonate ions from the buffer to the surface of the UV material may alter the behavior of the particles by shifting the PZC to higher values (making the surface more negative). However the differences are striking and do only appear for UV, while P25 behaves quite similar in both studies (figure 7). Figure 7: Taken from v.d. Kammer et al. 2, showing the behavior of the random, non-oecd batches of Degussa P25 (top) and Hombikat UV (bottom for comparison with the here tested OECD batches of the same materials. Shown is the effect of NaCl 2 on dispersion stability and z-potential of the TiO 2 nanoparticles. Stability/ remaining concentration with 25 mg/l initial concentration (left) and z-potential (right). Page 8

21 3.4. The calcium chloride matrix: influence of a divalent cation (with monovalent counter ion) The matrices showing the influence of a partially adsorbing and surface modifying divalent cation with an indifferent counter ion (Cl - ) are shown in figure 8. The effect of Ca 2+ on the zeta-potential by adsorption to the surface of the nanoparticles is evident for all materials. Hombikat UV UV Titan hydrophobic UV Titan hydrophilic Evonik P25 log mm CaCl2-4 mg/l Ca stable concentration: C/C in % log mm CaCl zeta potential im mv Figure 8: Effect of CaCl 2 on dispersion stability and z-potential of the TiO 2 nanopaticles. Stability/concentration (top) and z-potential (bottom) -matrix plots for the respective material. Higher colloidal stability and hence less aggregation is reflected by a higher concentration (brighter shade of the contours) in the dispersion after the reaction and settling period (5h). Elevated absolute values of the zeta potential support the colloidal stability through higher electrostatic repulsion forces between the particles (+/+ or -/-). Hence high absolute values are ploted with bright shades, less magnitude with dark shades (instability, tendency to aggregate). The adsorption of a divalent cation to the surface of TiO 2 particles can increase the positive charge or even induce a charge reversal (v.d. Kammer et al. 2; Fürstenau et al. 98). The IEP for low CaCl 2 concentrations is shifted from 7 to 7.7 for the UV Titan materials and 5.4 to 5.8 for P25. For HOM UV the magnitude of the negative zeta-potential is reduced. Page 9

22 The UV Titan materials show extended regions of positive zeta potential above. mmol/l of CaCl2 and this directly translates into stabilization against aggregation. This stabilization is slightly more pronounced for the hydrophilic type but again both materials are reacting quite similar. P25 is influenced to a much lesser extend but also this material experiences a stabilizing effect especially at higher CaCl 2 concentrations. Hombikat UV shows a slightly reduced stability compared to the NaCl matrices and does not profit from the stabilizing effect of CaCl 2 in the same way as the other materials. The reason might lie in the fact that already the extremely hight negative zeta-potentails in the NaCl and NOM matrices did not translate in relevant colloidal stability. With further reduction of the net-negative charge the aterials I destabilized.. Overall the increasing colloidal stability of the UV Titan materials and P25 through the addition of CaCl 2 would not be expected. The divalent calcium ion is potent in screening the charge of particles surface and reducing repulsive forces. Usually the critical coagulation concentrations (CCC) of Ca 2+ for negatively charged colloids like clay are in the -5 mmol/l range. Here however the addition of Calcium induces an elevated positive charge which has only chloride as a counter ion. This monovalent ion is comparably weak in its potency to screen positive surface charge and the CCC would be expected to lie between 5 2 mmol/l The sodium sulfate matrix: influence of a divalent anion (with monovalent counterion) The matrix plots for sodium sulfate concentrations over reveal that Hombikat UV, the two UVTitan materials and Evonik P25 react different on the addition of a interacting divalent anion. As Sodium is the related counter cation, comparisons should be made with the NaCl matrices. With Hombikat UV the region of negative z-potentials below -3 mv is extended towards lower values for elevated concentrations of sulfate (> mm). Apparently this translates to an extended stability of the materials under high sulfate concentrations and low. The overall increase of aggregation, as observed with the other uncoated material P25, is not observed here. However, the overall interaction of this material with sulfate although seems to be minor compared to the other materials, the impact on stability and zeta-potential is visible but not pronounced. Evonik P25 experiences and overall increase of the negative z-potential by the addition of sulfate, resulting in a shift of the PZC to values smaller that the -range covered by the plots. Only at very low sulfate concentrations the PZC is still in range at 4. The effect of the sulfate addition is apparently stronger at higher concentrations. Interestingly this increase in negative z-potential is not at all translated into a higher stability of the material. With an exception at low and low sulfate concentrations the remaining stable concentration of P25 in the presence of sulfate in extremely low. Now the highest stability of this material is observed at the lowest net zeta-potential what is contradicting classical understanding of how electrostatically stabilized systems behave. The two TitanUV materials again react quite similar, despite their different secondary coating material which is employed on the top of the first layer of AlOx. This secondary coating should give them distinct hydrophilic or hydrophobic properties. It would be expected that these differences can be clearly recognized in the test matrices. For both materials the interaction of the surface with sulfate is clearly visible; the hydrophilic material experiences a slight increase of negative zeta-potential at low sulfate concentrations, which then vanishes with increasing concentrations. At low concentrations of sulfate the hydrophobic material shows nearly no effect compared to the NaCl matrices but both materials show a slight shift of the ZPC towards lower for elevated sulfate concentrations. Extended positively Page 2

23 charged regions as clearly expressed in the NaCl plots are completely missing. Even at 4 the z-potential remains close to zero. This is translated into a low stability and remaining concentration in the supernatant (upper plots). Hence we have three distinct behavioral situations. Hombikat UV is stable and nearly not influenced by the addition of sodium sulfate, P25 experiences a charge increase, which is irrelevant for stability, since the material is nearly fully aggregated. The UV Titan materials experience an overall reduction in net charge and behave accordingly. log mm NaSO4 - Hombikat UV ~ mg/l SO 4 -- UV Titan hydrophobic - - UV Titan hydrophilic - Evonik P stable concentration: C/C in % log mm NaSO zeta potential im mv Figure 7: Effect of Na 2 SO 4 on dispersion stability and z-potential of the TiO 2 nanopaticles. Stability/concentration (top) and z-potential (bottom) matrix plots for the respective material. Higher colloidal stability and hence less aggregation is reflected by a higher concentration (brighter shade of the contours) in the dispersion after the reaction and settling period (5h). Elevated absolute values of the zeta potential support the colloidal stability through higher electrostatic repulsion forces between the particles (+/+ or -/-). Hence high absolute values are ploted with bright shades, less magnitude with dark shades (instability, tendency to aggregate). Page 2

24 4. Assessment of the associated risk for natural water bodies in Europe The matrix tests have shown clearly which hydrochemical conditions promote the dispersion of the TiO 2 nanoparticles and which conditions lead to aggregation. The results for the major water constituents (, Na +, Ca 2+, SO 4 2-, NOM) influencing nanoparticle dispersion and aggregation have been established, covering nearly all water chemistries found in natural waters across Europe. How can we now conclude on the potential risk for European freshwater bodies to stabilize the tested particles in the form of a small primary particle/aggregate and suppress aggregation? Through the initiative of the Forum of European Geological Surveys (FOREGS) to collect hydrochemical data for > 8 water bodies reporting ~ 6 descriptors and the compilation in the European Geochemical Atlas there is direct access to the relevant parameters for a broad set of European Water bodies. The data were condensed and sampling points with missing values excluded. Finally 88 sampling points remained and were plotted over the respective stability matrices of the different TiO 2 materials. These plots enable the identification of material properties which will lead to a large number of cases in which the particle will be well dispersed or vice versa, will be highly aggregated. Independent of the fact which of the two forms bear a larger risk, it is clearly visible which form will be preferred in how many European rivers. Figure 8: Overplot of the Na + and values in 88 surface water samples across Europe and the NaCl stability matrix of the four different TiO 2 materials. For Na + / for example completely different pictures are received from the different materials and it is clearly visible how the matrix plots become relevant for the real world application of the materials. Hombikat UV finds stabilizing conditions in nearly all European surface waters if we account for the Na + / situation alone. UV Titan hydrophilic is nearly everywhere completely aggregated, the other two tend to a more aggregated state, but the situation is not fully clear. Page 22

25 Figure 9: Overplot of the Ca ++ and values in 88 surface water samples across Europe and the CaCl 2 stability matrix of the four different TiO 2 materials. Figure : Overplot of the SO 4 -- and values in 88 surface water samples across Europe and the Na 2 SO 4 stability matrix of the four different TiO 2 materials. The same heterogeneous pattern is seen in all the other plots showing the Ca ++ / and the SO 4 -- / combinations. These plots however cannot be considered individually, the pots for each material must be combined and there is not a single material tested which has not at least one plot where nearly all natural samples fall into the highly aggregated category. While Hombikat UV seems to be stable regarding the existing SO 4 -- / and Na + / combinations, most of the Ca ++ / combinations result in high aggregation. Similar situations for other combinations are observed for the other materials. However, NOM is highly stabilizing and present in all waters and the general overplots shown do not allow an assessment of a certain individual sample. An approach to overcome this is an automated assessment of waterchemical conditions on interpolated matrix plots with a qualified weighing of the different parameters and is under development. Page 23