Detection, quanti ﬁ cation and derivation of number size distribution of silver nanoparticles in antimicrobial consumer products †

In 2011 the European Commission published its recommendation for a de ﬁ nition for the term nanomaterial which requires the materials to be characterized in terms of the number size distribution of their constituent particles. More recently, the de ﬁ nition has begun to be applied to the labelling of food and cosmetic products where any components present in the form of engineered nanomaterials must now be clearly indicated in the list of ingredients. The implementation of this de ﬁ nition requires that methods be developed and validated to accurately size particles with at least one external dimension in the range of 1 – 100 nm, and to quantify them on a ‘ number-based ’ particle size distribution. An in-house developed method based on Asymmetric Flow Field Flow Fractionation-Inductively Coupled Plasma Mass Spectrometry (AF4-ICP-MS) for the simultaneous detection and quanti ﬁ cation of citrate-stabilised silver nanoparticles (AgNPs) in water, has been applied to real-world liquid antimicrobial consumer products based on colloidal silver. This transfer of the method from ideal model systems to real products was assessed in light of other techniques including Centrifugal Liquid Sedimentation (CLS), Dynamic Light Scattering (DLS) and Transmission Electron Microscopy (TEM). Five out of six analysed products were found to contain AgNPs in the nano-range by means of a number of techniques including AF4-ICP-MS. Comparative analysis shows that CLS has su ﬃ cient size resolution to size AgNPs in the consumer products while DLS was unsuccessful probably due to sample polydispersivity. Despite the silver nanoparticles having unknown surface properties and stabilisation agents which could have in ﬂ uenced the sizing with AF4, a relatively good agreement between TEM and AF4-ICP-MS was observed. The AF4-ICP-MS data could be converted from mass-based to number-based distributions; this transformation, despite the possibility of experimental artefacts being mathematically ampli ﬁ ed, has shown promising results.


Introduction
The industry of nanotechnology-based ingredients, additives and food contact materials is expected to grow in the near future 1 with alimentary supplements and food packaging nanomaterials already being used in several countries. 2hese fast developments and their potential benets have spurred not only scientic and commercial activity but also highlighted a need to undertake some regulatory harmonization of the nano-science eld.In this context, the European Commission (EC) published in October 2011 its recommendation on the denition of nanomaterials.This recommendation denes a "Nanomaterial" as 'a natural, incidental or manufactured material containing particles, in an unbound state or as an aggregate or as an agglomerate and where, for 50% or more of the particles in the number size distribution, one or more external dimensions is in the size range 1 nm-100 nm'. 3 One important consequence of introducing this denition has been the need to create, develop and validate methods for detection and quantication measurement of nano-materials in food and consumer products.Furthermore, it has highlighted the limitations of current particle sizing methods when applied to measuring the number size distribution of particles both in and above the 1-100 nm size range. 4urrently there is no single analytical method or instrument able to address the requirements of the denition but a variety of methods exist which, in combination, offer the possibility of tackle this problem. 5,6t is well recognised that among nanomaterials, silver nanoparticles (AgNPs) are the largest and fastest growing category in use in consumer products.In particular, their powerful antimicrobial properties 7 and relatively low cost have led to silver nanoparticles being widely used in a number of applications including food contact materials, 8 cosmetics, wall paints, textiles, 9 laundry detergents, biocide sprays and medical devices. 10Over 430 consumer products are currently reported to be on the market 11 with this number being expected to increase signicantly in the coming years.Recently an Inventory of Nanotechnology applications in the agricultural, feed and food sector 12 was published showing that silver (along with nanoencapsulates and titanium dioxide) is among the most exploited nano-technological materials and that food additives and food contact materials are the most common of the current applications.
5][16] However there are still a number of critical points to be claried, including possible in vivo transformation, accumulation and interaction with enzymes and biomolecules, before a denite conclusion about whether or not AgNPs are hazardous to humans can be established. 17In this scenario the development of methods for the simultaneous detection/ quantication of AgNP from ionic and soluble species is crucial.
Among the techniques applicable to the detection and characterization of AgNPs in complex matrices, the use of Asymmetric Flow Field Flow Fractionation 18 (AF4) coupled to Inductively Coupled Plasma-Mass Spectrometry (ICP-MS) appears extremely promising and applicable to a number of sample types including cell culture media 19 and chicken meat. 20his approach combines the ability of AF4 to size-fractionate particles in the nano-range with the high sensitivity and element specicity of the ICP-MS, thus providing an ideal tool for the simultaneous detection and quantication of nanoparticles.Furthermore, being a technique for the analysis of liquid samples, AF4-ICP-MS allows the study of NP size distribution directly in their native dispersions, avoiding possible artefacts deriving from sample preparation and drying as it might happen in electron-microscopy analysis. 4Finally, AF4 is able to isolate AgNP from the ionic/soluble species, providing an appealing tool for ionic/particulate physico-chemical speciation.Single-particle ICP-MS and particle tracking analysis (PTA) are two techniques that can potentially provide the particle size distribution of silver nanoparticles.Both techniques have the advantage to count directly the number of particles, but have a lower limit of size detection of around 20 nm for AgNPs. 21In addition, at the moment, they are somewhat less robust than the AF4-ICP-MS combination and for PTA the measurements are highly operator-dependent. 22ne of the disadvantages of AF4-ICP-MS is that method development can be time consuming leading to undesirable additional costs for testing laboratories. 23In order to reduce FFF method development time, in this work we applied a single AF4-ICP-MS method that has been previously optimised for citrate-stabilised AgNPs in the 10-110 nm range 24 to a collection of commercial antimicrobial AgNPs samples.Thus, in contrast to other similar works published on the topic, 25 our objective was not to adapt the AF4-ICP-MS method to individual products but rather to evaluate the results obtained from the straightforward application of a single standardized method to a range of real life samples.In doing this the performance of the AF4-ICP-MS method was compared to other sizing techniques such as Centrifugal Liquid Sedimentation (CLS), Dynamic Light Scattering (DLS) and Transmission Electron Microscopy (TEM).The data obtained was then used to verify whether or not the tested commercial samples contained nanomaterials according to the EC denition.

Materials
Stock solutions of monomodal sodium citrate-stabilised silver nanoparticles with nominal sizes in the range between 20 nm and 110 nm (Ted Pella Inc., Redding, USA; Pelco Citrate NanoXact™, in 10 nm steps) were purchased as dispersions at a nominal concentration of 0.02 mg mL À1 (AE5%); 10 nm citratestabilised AgNPs were purchased from Sigma Aldrich (Sigma-Aldrich Corp., St. Louis, USA).Bottles were opened in a nitrogen glove-box and divided into ready-to use sub aliquots in amber glass vials.These were then stored at 4 C away from light to prevent oxidation and release of ionic silver from AgNPs. 13 As a quality check immediately aer delivery from the suppliers, the sizes of the AgNPs were veried by mean of CLS.
Nitric acid used was 67-69% ultrapure for trace analysis (CARLO ERBA Reagents S.r.l., Italy); silver and rhodium ICP-MS standards, at a concentration of 1000 mg L À1 in 2% nitric acid were purchased from Absolute Standards (Absolute Standards INC, Hamden, USA); the ultrapure water used throughout the experiments was supplied from a Millipore Advantage System (Merck Millipore, © Merck KGaA, Darmstadt, Germany).Sucrose (puriss.)used to create the liquid density gradient in the CLS was purchased from Sigma Aldrich (Sigma-Aldrich Corp., St. Louis, USA); the AF4 eluent was ultrapure water adjusted to pH 9.2 with 0.1 mM NaOH-solution and was freshly prepared every day.

Consumer products
Six different colloidal silver based consumer products, sold as antimicrobial agent for external use were purchased via internet (Table 1).The products were delivered without any information leaet but with the indication "for external use" on the label.All of them were in liquid form and declared to contain nano or colloidal silver and were delivered in amber or dark bottles (plastic or glass) with or without a nebulizer on top.

Bench UV-VIS spectrophotometry
UV-VIS spectra were recorded at room temperature with a Thermo Nicolet Evolution 300 instrument (Thermo Fisher Scientic, Inc.) in the range 200-600 nm using 1 cm path length quartz cuvettes and 1.5 nm bandwidth.Recorded spectra were not smoothed.Raw data were exported and plotted in OriginPro 7.5 (OriginLab Corporation).Aliquots of consumer products were analysed both when opened and aer 1 week to check their stability.

DLS and Z-potential
Particle size distribution (PSD) was determined by Dynamic Light Scattering (DLS) using a Zetasizer Nano-ZS instrument (Malvern Instruments Ltd, UK) with temperature control (24.9 C).Set parameters were the following: material refractive index 0.56, dispersant 1.330.Material absorption was set to 4.27, and viscosity to 0.8872 cP.Each sample was recorded in duplicate with an equilibration step of 120 s.Acquisition time was 80 s.Soware was set to automatic acquisition mode.Hydrodynamic diameters were calculated using the internal soware analysis.Z-potential was measured using the same instrument and recorded in a DTS1060C disposable cell with an equilibration time of 120 s.Measurements were done just aer pH measurement.A Smulochowski model with a F(Ka) of 1.5 was used.All consumer product samples were analysed without any dilution.

Particle size distribution by CLS
Particle size distribution (PSD) of undiluted consumer products were measured by Disc Centrifuge Photosedimentometer model DC24000UHR (CPS Instruments, Europe).The instrument was operated in line-start mode at a disc rotation speed of 22 000 rpm using an aqueous sucrose gradient (8-24% w/w) capped with dodecane to prevent solvent evaporation.Each measurement was preceded by a calibration step done using an aqueous reference standard of 377 nm diameter PVC spheres.Consumer product samples were analysed fresh and without dilution.

Total silver determination by ICP-MS
An Agilent ICP-MS 7700x (Agilent Technologies, Santa Clara, USA) equipped with platinum sampling and skimmer cones, MicroMist quartz nebuliser and a quartz Scott spray chamber was used for the determination of total silver in consumer products.Argon was used as nebulizer gas.The ICP-MS was operated in full quantication mode.Rhodium at a concentration of 50 mg L À1 in 1% nitric acid was added on-line as internal standard (ISTD), via a T-tube mounted before the nebuliser pump.Seven replicate readings per sample were performed, monitored signals included masses 107 and 109 for Ag and 103 for Rh, and integration time was 0.09 for Rh and 0.6 s for Ag isotopes.A total of 6 silver concentration standards (plus blank) were prepared in 2% nitric acid in the range 0.5-100 mg L À1 .Calibration curves were read three times during the run.A total of 12 procedural blanks were analysed during the run.A number of silver spiked solutions were analysed along with the samples as well as Standard Reference Material 1643e trace elements in water (NIST) for quality control.
Total silver in the consumer products was determined by ICP-MS following dilution in 2% nitric acid.A two-step dilution strategy (average total dilution factor 500Â) was used in which pure nitric acid was rstly allowed to react with the consumer products for 3 hours at room temperature before dilution.All samples, quality controls and standards were prepared gravimetrically on a 4 digit scale.All samples and ionic silver spikes were prepared and analysed in triplicate.Aer preparation samples were promptly analysed for total silver concentration by ICP-MS.

On line AF4-UV-VIS-ICPMS
An Asymmetric Flow Field Flow Fractionation AF2000 MT Multiow FFF, (Postnova Analytics, Germany) coupled to a UV-VIS detector (Postnova SPD-20AV) and ICP-MS were used for size fractionation and quantication of AgNPs and consumer products.AF4 eluent and elution conditions were carefully optimised for the separation mix of AgNPs citrate stabilised as reported elsewhere. 24Key details are provided in Table 2. Since the spectral response of silver nanoparticles varies as a function of their diameter it was not possible to set the UV-VIS detector at a detection wavelength which was optimum for all particle sizes.Instead, the UV-VIS detector was set to a wavelength of 420 nm which was judged to be a suitable compromise value able to detect silver nanoparticles across the size range of 10-110 nm.The AF4 channel was tted with a regenerated cellulose membrane with a cut-off of 10 kDa (Postnova Analytics, Germany, part no.Z-MEM-AQU-527) which was substituted approximately every 50 injections.A 350 mm channel-spacer was used (Postnova Analytics, Germany, part no.Z-AF4-SPA-V-355).For the coupling to the ICP-MS, the outlet tube from the on-line UV-VIS spectrometer was connected to the sample feed tube of the ICP-MS.The outow of the AF4 was set at 0.5 mL min À1 and aspirated by the peristaltic pump of the ICP-MS which was set at a ow of 0.6 mL min À1 .The higher suction ow of the ICP-MS pump served to aspirate, via a T-inlet, from 1% nitric acid Between each analysis run, the connection between AF4 and ICP-MS was manually removed, in order to allow an efficient purge of the AF4 and a ushing with 5% pure nitric acid of the ICP-MS, till a low, stable background for silver was reached.The consumer products were diluted in ultrapure water before injection to avoid overloading the ICP-MS detector or exceeding the highest calibration point.Raw data from the ICP-MS detector data was exported as counts per second versus time and the count ratio (107/103) versus time calculated and used for the quantication of silver in both calibrants and products; UV-VIS data at 420 nm were also exported and plotted versus time.OriginPro 7.5 (OriginLab Corporation) was used for further data treatment.

TEM
TEM analysis of the commercial products was done following liquid spotting on copper support grids.A drop of undiluted product (4 mL) was placed onto ultrathin Formvar-coated 200mesh copper grids (Tedpella Inc.) and le to dry in air at 4 C.For each sample, at least 100 particles were measured to obtain the average and the size distribution.NPs were visualized using a transmission electron microscope (TEM) (JEOL 2100, Japan) at an accelerating voltage of 200 kV.Digital images were analysed with the ImageJ soware and a custom macro performing smoothing (3 Â 3 or 5 Â 5 median lter), manual global threshold and automatic particle analysis provided by the ImageJ.The macro used is available at http:// www.code.google.com/p/psa-macro.The circularity lter of 0.5 was used to exclude agglomerates that occurred during drying.

Total silver determination by ICP-MS
The measured values of total silver in the consumer products are shown in the last column of Table 1.The batch ICP-MS analysis of all six consumer products conrmed that they all contain some degree of silver, but with some discrepancies in the total silver content compared to the amount of nano or colloidal silver content declared by producers.For Product A, B, F the content of total silver is relatively close to the declared values.For Product C and E the content of total silver exceeds the declared value while for Product D the content of total silver was found to be far below the declared value.A comparative graph of declared versus measured values can be found in Fig. 1, while details on quality checks, Limit of Quantication (LoQ) and recovery are contained in ESI.†

Sizing strategies results: CLS, DLS, TEM
Preliminary UV-VIS analysis, allowed a quick and easy screening of consumer products, by identication of the localised Surface Plasmon Resonance (LSPR) peak typical of nano-silver. 26The analysis of freshly opened undiluted consumer products is shown in Fig. 2A and B: Products E, C, B and F showed a strong absorbance in the region of 400 nm that is characteristic of monomodal stocks of sodium citrate stabilised AgNPs in the size-range 10 to 50 nm (data not shown).Product A (Fig. 2B) showed a very weak peak suggesting the presence of a relatively low concentration of silver in the nano-range.From the analysis of Product D, no signicant signal of the LSPR band was detected.
Centrifugal Liquid Sedimentation (CLS) revealed the presence of nanoparticles in 5 out of 6 products analysed (Fig. 2C and Table 3).CLS results conrmed the presence and relative quantities of the particle indicted by the UV-VIS analysis, while sample D did not show any detectable particles.Products B, C, F and E and A were all found to contain nanoparticles with 100% of the particle size distribution (in weight) below 100 nm.In particular, Products B, C, F and E showed very broad peaks  centred below 15 nm, while Product A showed a small peak centred at 17.4 nm.Batch DLS analysis was performed on undiluted consumer products (Table 3 and Fig. S.1 in ESI †).In the majority of cases, the polydispersivity index (PdI) of the products was quite large indicating low monodispersivity and/or possible matrix effects.For Product A, duplicate analysis on batch DLS produced poorly reproducible results and in one of the two measurements a bimodal distribution was detected (Fig. S.1.A of ESI †).All the detected Z-averages are greater than diameters generated by CLS probably due to a small number of larger particles or aggregates which, due to the highly non-linear variation of scattering intensity with particle size which, can introduce errors in DLS biasing the calculated mean size towards larger values. 27n contrast with DLS, CLS shows a greater ability to deal with heterogeneous dispersions, due to size fractionation before particle detection.CLS has proven to be considerably more suitable than DLS for the characterisation of multi-modal AgNP suspensions, and in particular to resolve bimodal mixtures of  AgNPs. 28Nonetheless CLS, being a centrifugation based technique, depends on knowing a value of particle density which is normally assumed to correspond to the bulk metal one (10.49g cm À3 ).While such an approximation may be valid for uncoated particles, it may create a systematic error with polymer or protein coated particles, and this effect might be especially signicant on very small particles. 29Since it was not possible to assess the nature of any capping agent, coating or stabiliser in this study, some error related to the assumed particle density might ultimately affect the sizing by CLS.TEM observations of the six products conrmed the presence of quasi-spherical nanoparticles in all samples with the exception of Product D (Fig. 3), which did not show any sign of nano-or macro-particles (Product D image not reported).Image analysis of Products B, C, E and F produced narrow size distributions with mean size values between 11-14 nm (Fig. 3, Table 3).However, Product A showed a more polydisperse distribution of NPs, which might be approximated as two populations: 1 st with a mean size of 15 nm and a 2 nd of 42.6 nm.However, this bimodal trend was not detected by CLS.For sample C, TEM observations showed a tail up to 30 nm, but the mean size is 11.9 nm.Indeed Product C showed a tendency to change colour from yellow to blue/grey when small aliquots were stored (1 week) in the fridge without rigorous exclusion of oxygen.This was conrmed as a shi of the LSPR peak in the UV-VIS absorbance spectrum (ESI †) which is assumed to be an indication of possible aggregation and/or surface oxidation. 26,30

Z-potential results
All the consumer products found to contain silver nanoparticles were further characterised to determine the Z-potential of particles (Fig. 2D).Initial pH was measured and analysis performed immediately.All the consumer products analysed showed a negative value (from À38.6 mV to À54 mV) at their original pH (ranging from 6.6 to 6.9).Z-potential values of this magnitude would normally be indicative of particles which carry sufficient charge as to be electrostatically stabilised and thus resistant to spontaneous aggregation.Average data are reported in Table 3 while further information is tabulated in Table S.2 of ESI.†

Quantication of particulate silver with AF4-UV-VIS-ICP-MS
The previously developed AF4-UV-VIS-ICP-MS method optimised for citrate-stabilised AgNPs in the range between 10 nm and 110 nm, 24 was applied to the consumer products.For mass calibration a method based on pre-channel injection was adopted 22 in which accurately known quantities of three size AgNPs (nominal diameters 20 nm, 60 nm, 100 nm) on three concentration levels (0, 25 and 60 ng of total silver) were injected using a 50 mL loop before applying the standard elution prole.Three different calibration curves were built for each of three particle sizes respectively.This particular approach has both advantages and disadvantageson one hand it has been previously demonstrated to be effective in compensating for incomplete particle channel recovery compared to ionic silver post-channel calibration 24 but is hindered by the lack of AgNPs certied standards for size and concentration.To overcome this problem it was necessary to in-house determine the concentration of silver particles in the stock solutions. 24he ICP-MS on-line detector showed a Limit of Detection (LoD) (average blank + 3SD blank/slope) between 0.170-0.340mg L À1 and a Limit of Quantication (LoQ) (average blank + 10SD blank/slope) in the range 0.566-1.133mg L À1 for different particle sizes in line with previous results reported in literature. 25ractograms obtained with both the UV-VIS detector and ICP-MS are shown in Fig. 4 which shows an example of the AgNPs mass calibration mixture together with data relative to the ve consumer products.The ve consumer products which showed a detectable amount of silver gave clear signals detectable both on UV-VIS and ICP-MS detectors with main peaks being well separated from the void peak.For the purpose of clarity, the UV-VIS detector signal collected at 420 nm (black-le axes) and ICP-MS (coloured lines, right axes) have been aligned by the void peak.Aer a preliminary run for each sample, dilution factors were chosen to avoid to overload the detector and to ensure quantication within the calibration curve.Exact dilution factors in ultrapure water before injection were the following: Product A: none; Product B: 10.97; Product C: 20.1; Product E: 16.7, Product F: 19.7.
Channel recovery for real consumer products was evaluated by injecting samples with and without cross-ow and comparing the different peak areas obtained on the UV-VIS.Recovery was calculated as: %Recovery ¼ 100 Â peak area EL/peak area NC with EL ¼ under elution condition; NC ¼ no cross ow.
Channel recoveries were the following: 70% for Product A, 88% for Product B, 94% for Product C, 80% for Product F, 96% for E, and 85% for 20 nm AgNPs here used as standard.In our previous work the estimation of the size-related recovery was performed systematically on ve replicates demonstrating the absence of signicant difference for citrate-coated particles of 10, 20, 40 and 60 nm. 24In this context, considering the insig-nicant changes in detected particle size in the consumer products such a variation in recovery might be due more to surface chemistry and consequentially to membrane interaction.
To convert count ratio into concentration, the slope of the 20 nm AgNP calibration curve (count ratio versus area) was used.A curve tting was done using OriginPro.7.5 and the 'Gauss mod' function (an exponentially modied Gaussian peak function for   S.4 of ESI †).From these estimates, Product A was found to contain less than 2% of total silver as particulate silver; on the contrary Product C was found to contain over 85% of particulate silver.Fig. 1 (red bars) shows the AgNP content in the consumer products obtained by this approach; a similar trend was observed from CLS (peak height) and UV-VIS analysis with Product A being the one with the lowest particle concentrations and Product C being the highest, and with Product B and F having an intermediate content of particulate silver.For Product E, a qualitative analysis of UV-VIS and CLS data suggests a content of silver nanoparticles comparable to Product C (Fig. 2A and C), while the data from AF4-ICP-MS shows that it contains around 50% less silver NP than Product C (Table S.4 †).The reason for this underestimation is unclear.

Sizing with AF4-UV-VIS-ICPMS and conversion from mass to number
For both the model mixture (composed of 20 nm, 60 nm and 100 nm citrate-stabilised AgNP) and the consumer products, the following strategy was used to convert the ICP-MS detector information (count ratio) into weight based and number based PSD (Fig. 5 and 6): (i) The relationship between particle size and elution time when using the ICP-MS detector was determined as follows: a size vs. elution time calibration was performed daily with a mixture of citrate-stabilised AgNPs of known size; previous ndings during method development showed elution times to be reproducible with a relative standard deviation generally below 1.5% on 14 replicate measurements performed on both intra-day and inter-day basis for different particle sizes. 24To do this a suspension containing 10 nm, 20 nm, 40 nm, 60 nm and 100 nm citrate stabilised AgNPs was run daily as a channel calibration mixture as in Fig. S.2.A of ESI.† The good reproducibility of the elution time-particle size relationship measured using the ICP-MS detector permitted it to be used for sizing silver nanoparticles in consumer products.An example of applied equation is Size ¼ 0.027(elution time) 2.38 Regression and further details are reported in (iii) By applying the correspondence determined in (ii) to the ICP-MS detector signal for unknown samples the number of counts accumulated in each sampling period was converted in a mass of silver in that time interval.
(iv) Since the mean particle size at each data point in the eluting prole is known from step (i), the volume and theoretical weight of the individual particles at each time point can be calculated by assuming the density of bulk silver of 10.49 g cm À3 31 and spherical geometry.
(v) From the total mass of silver in each given time interval and the corresponding mass expected for each particle the total number of particles in a certain time sampling period can be calculated.
As a conrmation of the quality of this sizing approach a mixture of 20 nm, 60 nm and 100 nm citrate-stabilised AgNPs was sized, providing satisfactory results of 20.3 (5.4) nm, 59.4 (10.4) nm and 95.1 (15.4) nm for peak centre (FWHM) when a gauss function was t on data in OriginPro 7.5.In terms of mass, the 20 nm 60 nm and 100 nm mixture was analysed in a 1 : 1 : 1 mass ratio (Fig. 5A); the experimental values were compared to the expected ones, based on the known injected mass (459.35,463.01 and 444.9 mg L À1 for 20, 60 and 100 nm respectively) and the measured TEM diameters (19.6, 57.4 and 99.4 nm respectively).The measured relative mass-PSD of 33.1%, 31.2%,32.3% compares well with the expected values of 33.6%, 33.9% and 32.5% for the 20, 60, and 100 nm AgNP (Fig. 5B).The above described strategy allowed converting the massbased PSD into the number-based PSD.Such analysis gave an experimental relative number PSD of 95.8%, 3.4% and 0.8% for the small, medium and large AgNP, respectively.Such distribution compares well with the predicted number-PSD of 95.6%, 3.7% and 0.7%.
The effect of conversion from mass to number of particles clearly produces a great enhancement of the relative contribution from 20 nm compared to the 100 nm that very well reects the predicted mass and number ratio calculated (respectively grey and blue bars in Fig. 5B).In performing this operation it is important that the lower size cut-off point be correctly selected to avoid errors from inclusion of materials in the void-peak.Indeed if the cut-off is xed too early, the void peak will be included in the calculation and erroneously considered as small particles.This effect has been studied in detail for the model sample AgNPs 20, 60, 100 nm (Fig. S.3 in ESI †): the smaller the cut-off the higher the risk that a part of the void peak be incorrectly considered to be small sized particles with D50 for number PSD ranging from 3 nm up to 19 nm when going from a cut off of 2 up to 10 nm.For most of the samples analysed in this study the cut-off is set at 8 nm.
It should be noted that although the mass to number transformation is relatively simple to achieve, it suffers from a number of limitations, either intrinsically related to the analytical method or related to data conversion.For the latter case, the fact that the generically named as 'silver nanoparticles' present in consumer products could include a number of different surface chemistries; for example, silver-chitosanpoly(ethylene glycol) nanoparticles, 32 and silver/poly(lactic acid) 33 have been reported for antibacterial applications.Such 'silver nanoparticles' might show an unpredictable behaviour during channel elution, such as repulsion or attraction to the membrane potentially leading to variations in retention time and/or recovery.To some extent, the suitability of an unknown particle type could be veried by measuring its Z-potential under standard elution conditions and comparing it with that found for the citrate-stabilize particles.If the measured Zpotential values are substantially higher of lower than for the size-calibrants, extreme care should be taken in using elution time to derive the particle size since it might produce an erroneous size determination based on retention time only.In the extreme case of positively charged particles our method would likely be ineffective due to the strong interactions with the negatively charged membrane at the working pH of 9.2.
In addition, any inaccuracies in the experimentally determined weight-size distribution are disproportionately magni-ed by the mathematical conversion to the desired number-size distribution.In particular, the smaller the particles the greater is the potential error in number.Finally, the conversion from mass to number assumes that the particles have spherical geometry and that their density corresponds to that of bulk silver.
Bearing in mind the possible limitations discussed above the raw data for the ve consumer products was elaborated to obtain mass and number size distributions together with their

Fig. 1
Fig. 1 Declared content of nano/colloidal silver in the consumer products (black bars), total silver determined by batch ICP-MS in consumer products.Particulate silver in the nano-range detected in consumer products determined by AF4-ICP-MS are represented by red bars.

Fig. 2 (
Fig. 2 (A) UV-VIS spectra of consumer products analysed on the day of opening without any further dilution; (B) magnification of UV-VIS spectra of Products A and D showing a small localised surface plasmon resonance band for Product A; (C) CLS analysis of consumer products analysed undiluted, dotted line shows the 100 nm size threshold; (D) Z-potential versus native pH of consumer products containing AgNPs.

Fig. 3
Fig. 3 TEM images and particle size distributions of consumer products; (A): Product A; (B): Product B; (C): Product C; (D): Product E; (E): Product F. Product D did not show any presence of detectable silver nanoparticles on TEM analysis.
use in Chromatography) (details in TableS.3 of ESI †).Peak areas were used for quantication of the particulate silver (details in paragraph S.2 of ESI †) taking into account channel recovery rate and compared to total silver content determined by batch ICP-MS (Fig.1, Table Fig. S.2.B of ESI.† (ii) The correspondence between counts and mass of silver in each ICP-MS sampling time was calculated from the integrated elution peak areas obtained from injections of known masses of 20 nm AgNPs.

Fig. 5 (
Fig. 5 (A): Transformation of AF4-ICP-MS data into relative (continuous lines) and cumulant (dotted lines) distributions based on mass (black) and number (red) for a citrate-stabilised AgNPs mixture of 19.6 nm, 57.9 nm and 99.4 nm respectively (TEM diameter); D m ¼ cumulant distribution based on mass; D n ¼ cumulant distribution based on number; (B): predicted and measured (by AF4-ICP-MS) relative mass and number particle size distributions.

Table 1
Description of purchased and tested consumer products, declared and measured total concentration of silver, ( ) average of 3 replicates AE standard error solution.Finally, Rh as internal standard was added in line with a T-inlet before reaching the ICP-MS detector.ICP-MS was run in chromatographic mode and m/z signal acquisition in Time Resolved Analysis (TRA).Platinum cones were used.

Table 2
AF4-ICP-MS parameters used in this study

Table 3
Summary of silver particle sizes (nm) determined with different measurement techniques and Z-potential (mV) a 17.4 (1.5) 13.3 (7.3) 12.1 (7.2) -12.6 (6.8) 13.6 (7.1) TEM d P1 15.1 AE 4.5 12.4 AE 5.7 11.9 AE 5.1 -11.6 AE 4.4 13.6 AE 5.a Peak maximum provided by CLS soware from particle size distribution in weight; (s calculated from the peak half height width provided by CLS soware dividing by 2.354).b Z average calculated from DLS soware AEZ average Â OPDI.c Distribution peak centre.d Average AE standard deviation.