Characterization of Titanium Dioxide Nanoparticles Dispersed in Organic Ligand Solutions by Using a Diffusion-Ordered Spectroscopy-Based Strategy

Characterization of Titanium Dioxide Nanoparticles in FoodProducts Analytical Methods To Define NanoparticlesRuud J B Petersdagger Greet van Bemmeldagger Zahira Herrera-Riveradagger Hans P F G Helsperdagger

Hans J P Marvindagger Stefan Weigeldagger Peter C Trompsect Agnes G Oomen Anton G Rietveld

and Hans Bouwmeesterdagger

dagger RIKILT Wageningen UR Akkermaalsbos 2 6708 WB Wageningen The Netherlandssect TNO Earth Life and Social Sciences Princetonlaan 6 3584 CB Utrecht The Netherlands National Institute for Public Health and the Environment Antonie van Leeuwenhoeklaan 9 3720 BA Bilthoven The Netherlands

ABSTRACT Titanium dioxide (TiO2) is a common food additive used to enhance the white color brightness and sometimesflavor of a variety of food products In this study 7 food grade TiO2 materials (E171) 24 food products and 3 personal careproducts were investigated for their TiO2 content and the number-based size distribution of TiO2 particles present in theseproducts Three principally different methods have been used to determine the number-based size distribution of TiO2 particleselectron microscopy asymmetric flow field-flow fractionation combined with inductively coupled mass spectrometry and single-particle inductively coupled mass spectrometry The results show that all E171 materials have similar size distributions withprimary particle sizes in the range of 60minus300 nm Depending on the analytical method used 10minus15 of the particles in thesematerials had sizes below 100 nm In 24 of the 27 foods and personal care products detectable amounts of titanium were foundranging from 002 to 90 mg TiO2g product The number-based size distributions for TiO2 particles in the food and personalcare products showed that 5minus10 of the particles in these products had sizes below 100 nm comparable to that found in theE171 materials Comparable size distributions were found using the three principally different analytical methods Although theapplied methods are considered state of the art they showed practical size limits for TiO2 particles in the range of 20minus50 nmwhich may introduce a significant bias in the size distribution because particles lt20 nm are excluded This shows the inability ofcurrent state of the art methods to support the European Union recommendation for the definition of nanomaterials

KEYWORDS titanium dioxide food additive nanomaterial AF4 ICP-MS single-particle ICP-MS

INTRODUCTION

Titanium dioxide (TiO2) is a naturally occurring oxide of theelement titanium also referred to as titania This substanceoccurs naturally as three mineral compounds known as anataserutile and brookite There are a number of industrialapplications for this mineral because of its very high refractionproperties In fact TiO2 is one of the whitest materials knownto exist hence the name ldquotitanium whiterdquo For this reason it isoften included in cosmetic preparations and sunblocks to reflectlight away from the skin It is also incorporated in paints and ina number of construction and building materials Scattering oflight by TiO2 is maximized in particles that are 200minus300 nm indiameter and most commercial products that are used aspigments have primary particle sizes within this range1 The sizedistribution of ultrasonically dispersed primary particles andaggregates generally ranges from lt100 to 500 nm2minus4 UltrafineTiO2 particles range in size from 1 to 150 nm with a meanprimary particle size of 10minus50 nm3 These ultrafine TiO2

particles are mainly used for special applications such asphotocatalysts Because the refractive power of TiO2 particles ishigher at the nanoscale many applications of TiO2 such assunscreens cosmetics coatings and especially photocatalyticapplications would benefit from smaller primary particle sizesand therefore it is expected that the percentage of TiO2 that isproduced in or near the nanosize range will increase56

TiO2 is commonly used as food additive and is authorized forits use in the European Union (EU) as E1717 As a pigmentTiO2 is used to enhance the white color of certain foods suchas dairy products and candy It also lends brightness totoothpaste and some medications However it is also used as afood additive and flavor enhancer in a variety of nonwhitefoods including dried vegetables nuts seeds soups andmustard as well as beer and wine8

In recent years concerns have been raised with regard to thetoxicity of nanosized TiO2 following oral exposure Five dayexposure to 1 or 2 mgkg body weight (bw) per day anataseTiO2 resulted in significantly increased Ti levels in the ovaryand spleen and is suggestive of hormonal effects The primaryparticle size of these TiO2 nanoparticles (NPs) was lt25 nmwhich agglomerated into materials of which 13 of the particleswas smaller than 100 nm The size distribution was dominatedby agglomerates with a mean diameter up to 16 μm9

Furthermore a 90 day intragastric exposure to 25 5 and 10mgkg anatase TiO2 NPs with a crystallite size of 55 nm and ahydrodynamic size around 300 nm (in an 05 hydroxypro-pylmethylcellulose solvent) resulted in several kidney effects

Received March 12 2014Revised June 9 2014Accepted June 13 2014

Article

pubsacsorgJAFC

copy XXXX American Chemical Society A dxdoiorg101021jf5011885 | J Agric Food Chem XXXX XXX XXXminusXXX

(renal inflammation tissue necrosis and renal apoptosis)10 Inaddition the International Agency for Research on Cancer(IARC) has now classified nanosized TiO2 as a potentialcarcinogen on the basis of the rate of incidence of respiratorytract cancer in rats after prolonged inhalation of TiO2 dustparticles11 This indicates the need to evaluate the possiblecarcinogenicity of nanosized TiO2 following other exposureroutesA Monte Carlo human exposure analysis indicated that US

adults are exposed to about 1 mg Tikg bw per day8 Theseauthors indicate that in the assessed food and consumerproducts approximately 36 of the TiO2 particles were lt100nm in one dimension Although TiO2 has been authorized as afood additive7 no recent risk assessment is available In generalthe risk assessment of oral exposure to nanoparticles is difficultbecause of a number of uncertainties and a general lack ofdata12 Specifically for TiO2 knowledge about the particle sizesand size distribution of TiO2 particles in E171 let alone foodand consumer products containing E171 is limited TheEuropean Commission adopted a recommendation for thedefinition of nanomaterials Commission Recommendation2011696EU stating that a material is a nanomaterial if 50or more of the particles in a number-based size distributionhave one or more external dimension in the size range of 1minus100 nm In specific cases the number-based size distributionthreshold of 50 may be replaced by a threshold between 1and 5013 Currently the recommendation is under review thediscussion is focused on the percentage of the number-basedsize distribution below the 100 nm to classify a material asnanomaterial Yet no internationally agreed upon method todetermine the number-based size distribution has beenproposed Therefore we used a combination of analyticalmethods as suggested by Linsinger14

In this study we aimed to provide more information on thenumber-based size distribution of food grade TiO2 ingredientsand their occurrence in consumer products because this isrequired to increase the reliability of the human food exposurestudies as part of the risk assessment We used a combination ofmethods for this asymmetric flow field-flow fractionationonline with inductively coupled plasma mass spectrometry(AF4-ICP-MS) and two methods that provide number-basedsize distributions scanning electron microscopy (SEM) andsingle-particle ICP-MS (sp-ICP-MS) Comparable size distri-butions were found indicating that these methods can reliablybe used to enforce food labeling in line with therecommendation of the EU definition of nanomaterials

EXPERIMENTAL PROCEDURESSamples In this study three types of samples were involved first

several types of TiO2 food additive E171 second food products suchas cakes candy and chewing gum and last personal care productssuch as toothpaste In total seven E171 materials all food grades andfour labeled as E171 were received from suppliers in China GermanyItaly The Netherlands and the United Kingdom and consisted of finewhite powders In total 24 food products and 3 personal care productswere investigated All of these products were purchased in 2012 fromregular shops in The Netherlands Because food labeling for titaniumdioxide may be imprecise 20 products that listed ldquoE171rdquo ldquoTiO2rdquo orldquotitanium dioxiderdquo on the package were selected as well as 7 productswithout this labeling but with a typical white color Samples werestored in a clean dry and dark location and analyzed before theexpiration dataDetermination of Total Titanium Content For the determi-

nation of the total titanium content of the TiO2 materials a suspension

of these materials was prepared according to the method of Jensen15

In short ca 15 mg of bovine serum albumin (BSA) is dissolved anddiluted in Milli-Q water to reach a final concentration of 10 mgmLThis solution is filtered over a 02 μm pore size filter and diluted 20-fold with Milli-Q water to a final concentration of 05 mgmL BSA Anaccurately weighed sample of ca 15 mg of finely powdered TiO2material is brought into a 30 mL vial Then 30 μL of 96 ethanol isadded and distributed equally over the TiO2 material followed by 970μL of the 05 mgmL BSA solution The mixture is shaken manuallyanother 5 mL of the 05 mgmL BSA solution is added and the finalsuspension is shaken manually to obtain a homogeneous suspensionWhile the sample vial is cooled in ice water the suspension issonicated with a Misonix XL-2000 sonicator with a CML-4 needleprobe for 16 min at 225 kHz and 4 W power The final suspension isused as such or after further dilution to the desired concentration withthe 05 mgmL BSA solution Suspensions prepared in this wayshowed to be stable for at least 3 days From each suspension asubsample was collected in a perfluoroalkoxy (PFA) digestion vial towhich 8 mL of nitric acid (70 HNO3) and 2 mL of hydrogen fluoride(50 HF) were added For the determination of the total titaniumcontent of the food and personal care products the whole sample wascut into small pieces and ground and a representative subsample wasused for analysis An analytical sample of ca 05 g was collected fromeach sample and brought into a PFA digestion vial to which 6 mL ofnitric acid (70 HNO3) 2 mL of hydrofluoric acid (50 HF) and 2mL of hydrogen peroxide (30 H2O2) were added All samples weredigested in a MARS microwave system for 50 min The temperatureprogram was as follows at 1200 W power from 20 to 150 degC in 15min then to 180 degC in 15 min and finally constant at 180 degC for 20min Following digestion and cooling to room temperature Milli-Qwater was added to a total volume of 50 mL The extracts were shakenmanually diluted to 100 mL and analyzed with a Thermo X series-2ICP-MS equipped with an autosampler and a conical glass concentricnebulizer and operated at an RF power of 1400 W Data acquisitionwas performed in the selected ion monitoring mode at mz ratios of48 and 49 that are characteristic for titanium Quantification was basedon ionic titanium standard diluted in the same acidic matrix Therecovery of the total Ti method was determined by spiking threeproducts of different matrices in triplicate with known amounts of aTiO2 reference material NM-103

Determination of the Size Distribution in TiO2 MaterialsUsing Electron Microscopy To determine the size distribution ofthe titanium oxide nanoparticles in the TiO2 materials suspensions ofthese materials were studied using electron microscopy TiO2suspensions (25 mgmL) were filtered over an Anopore aluminumoxide filter with a pore size of 20 nm The filters were mounted onaluminum specimen holders with double-sided adhesive carbon tapeand were coated with a 5 nm layer of chromium using an EmitechK575X turbo sputter coater The filters were analyzed with a high-resolution field emission gun scanning electron microscopy incombination with X-ray analysis (FEG-SEMEDX) The microscopeis a Tescan MIRA-LMH FEG-SEM at an accelerating voltage of 15 kVworking distance of 10 mm spot size of 5 nm The EDX spectrometeris a Bruker AXS spectrometer with a Quantax 800 workstation and anXFlash 4010 detector The SEM is equipped with the Scandium SISsoftware package (Olympus Soft Imaging Solutions Germany) forautomated particle analysis With this system the filter area isautomatically inspected on a field-by-field basis In each field of viewparticles are recognized using a preselected grayscale video threshold(detection threshold level) to discriminate between particles and filterbackground The analysis is conducted using the backscatteredelectron (BE) mode From each particlecluster of particles theprojected area equivalent diameter (dpa) is measured Magnificationswere chosen so that their measurable size ranges overlap slightly andcover the particle sizes of interest To measure dpa values from 25 to1600 nm three magnifications of 10000times 25000times and 75000times wereselected These three magnifications cover eight size bins 25minus40 40minus65 65minus100 100minus160 160minus250 250minus400 400minus650 650minus1000 and1000minus1600 nm Per size bin a minimum of 10 particles weremeasured and in total approximately 1000 particles were measured

Journal of Agricultural and Food Chemistry Article

dxdoiorg101021jf5011885 | J Agric Food Chem XXXX XXX XXXminusXXXB

per sample To ensure that there is no overlap of particles the totalloading of the filter with particles was always lt10 Per size bin themean diameter and the 95 confidence interval are calculated on thebasis of the Poisson distributionQuantification and Size Determination of TiO2 Particles

Using AF4-ICP-MS For quantification and size determination ofparticulate TiO2 in TiO2 materials a suspension of these materials wasprepared as described before For quantification and size determi-nation of particulate TiO2 in food and personal care products arepresentative subsample was prepared and 05 g of this sample wasbrought into a 50 mL glass beaker together with 20 mL of hydrogenperoxide (30 H2O2) The content of the beaker was heated to about100 degC and kept just below the boiling point of the H2O2 matrixWhen most of the organic matrix is consumed the content isevaporated until a residue of about 1 mL remains The content isallowed to cool to room temperature and is diluted to 10 mL with the05 mgmL BSA solution Prepared suspensions and extracts areanalyzed with asymmetric flow field-flow fractionation (AF4) onlinecoupled with ICP-MS The AF4 consisted of a metal-free WyattEclipse Dualtec separation system equipped with a 153 times 22 mm flowcell containing a Nadir regenerated cellulose separation membranewith a 10 kDa molecular weight cutoff and a spacer of 350 μm TheAF4 system was further coupled to an Agilent 1100 system with anautoinjector and a binary pump for eluent flow delivery and degassingPrior to injection samples were thoroughly mixed to ensurehomogeneous distribution of suspended particles The injectionvolume was 10 μL and a 7 min focusing time was used A constantcross-flow of 01 mLmin was applied and the flow rate to thedetector was kept at 05 mLmin After 57 min the cross-flow wasreduced to 0 mLmin within 5 min and kept at 0 mLmin for 5 min toremove any residual particles from the separation channel The outletflow of the AF4 system passed through a UV detector and was directlycoupled with the ICP-MS system for the selective determination oftitanium The ICP-MS system and settings were as described beforeBecause no well-characterized titanium oxide particles were availablefor calibration purposes particle size calculation was based on FFFtheory using the Boltzmann constant the column temperature theeluent viscosity and the instrumental settings1617 In short the AF4separation channel consists of an impermeable top block and a bottomblock holding a semipermeable ultrafiltration membrane on top of aporous frit The perpendicular field is caused by restricting the channelflow at the outlet This restriction will force part of the carrier liquid toleave the channel through the bottom block thus causing the cross-flow Prior to elution the analyte is concentrated at a position close tothe injection port during a focusing step Particle equilibrium heightsabove the membrane in AF4 depend on the diffusion coefficient of theparticles and the applied cross-flow rate with larger particles being

driven closer to the membrane Retention times are dependent on theequilibrium height at which the particles travel in the parabolic flowprofile of the channel flow and are expressed by

πη ω= timestdkT

VV2r

2cross

channel (1)

where tr is the retention time of the analyte η the viscosity of theeluent d the hydrodynamic diameter of the particle ω the channelthickness k the Boltzmann constant T the absolute temperature Vcrossthe cross-flow rate and Vchannel the channel flow rate With thisequation the particlersquos hydrodynamic diameter can be calculateddirectly from its retention time This relationship was confirmed bycalibration with polystyrene particles Mass calibration of the ICP-MSwas performed with one of the TiO2 materials for which it wasestablished that the TiO2 content of this material was close to 100

Transformation of Mass-Based to Number-Based SizeDistributions The signal measured by ICP-MS in standard modeis a mass-based signal and the size distribution determined with AF4-ICP-MS is thus also a mass-based distribution The conversion of amass- to a particle number-based distribution is carried out by placingthe detected mass in size bins and calculating the number of particlesas

ρ π=

times timesminusNM

r( 10 ) 1043

7 3 9(2)

where N is the number of particles M is the mass detected in a certainparticle size bin in ng ρ is the density of the particle material in gcm3

(423 gcm3 was used for TiO2) and r is the radius of the particle inthat size bin in nm The numbers are conversion factors to get theunits right

Size Determination of TiO2 Particles Using Single-ParticleICP-MS Single-particle ICP-MS is a relatively new approach for thedetection and characterization of nanoparticles and the principle hasbeen described previously18minus21 The ICP-MS instrument and settingsused were as described before only in this case data acquisition wasdone in the time-resolved analysis (TRA) mode The dwell time wasset at 3 ms with a typical acquisition time of 60 s per measurementBecause of the short dwell time it is not possible to switch betweendifferent isotopes (switching itself takes sim50 ms) and therefore onlyone titanium isotope mz 48 is monitored during measurement Notethat polyatomic interferences are possible for example 32S16O and36Ar12C and also isobaric interference by 48Ca However theseinterferences result in a continuous background whereas TiO2particles will result in discontinuous signals (peaks) and can thus bedistinguished from the background Because only a part of the

Figure 1 Single-particle ICPMS time scan of a diluted sample extract Each peak represents one particle The number of peaks is proportional to theparticle concentration in the sample whereas the peak height is proportional to the particlersquos diameter to the third power

Journal of Agricultural and Food Chemistry Article

dxdoiorg101021jf5011885 | J Agric Food Chem XXXX XXX XXXminusXXXC

nebulized droplets reaches the plasma the nebulization efficiencyneeds to be known to determine the actual particle numberconcentration in the extract The nebulization efficiency is determinedby the analyses of a NIST material SRM8013 a suspension of 60 nmgold nanoparticles in a citrate buffer under the same instrumentalconditions as the samples but monitoring mz 197 for gold Thenebulization efficiency is calculated from the observed number ofparticles in the time scan and the particle flux into the ICP-MS systemusing eq 3 Single-particle data are transferred to and processed inMicrosoft Excel for the calculation of particle sizes particle sizedistributions and particle concentrations Acquiring data for 60 s at adwell time of 3 ms results in 20000 data points consisting ofbackground signals and signals with a much higher intensityoriginating from particles (Figure 1) These particle signals areisolated from the background by plotting a signal distribution that isthe frequency with which a signal height occurs as a function of thatsignal height This allows the determination of a cutoff point toseparate particle signals from the background (Figure 2) From thenumber of the particle signals in the time scan and the nebulizationefficiency determined previously with the gold nanoparticles theparticle number concentration in the diluted sample suspension iscalculated as

η= timesC

N

V1000

pp

n (3)

where Cp = particle number concentration (Lminus1) Np = number ofparticles detected in the time scan (minminus1) ηn = nebulizationefficiency and V = sample input flow (mLmin) The same formula isused to calculate the nebulization efficiency after measurement of the50 ngL 60 nm gold nanoparticle suspension In that case the particlenumber concentration Cp is known (50 ngL of a 60 nm gold particleresults in 2 times 107 particlesL) and the nebulization efficiency iscalculated from the observed number of particles Np in the time scanFrom the intensity of the particle signals and the response factorcalculated from a series of ionic titanium calibration standards that areanalyzed under the same conditions the mass of the individualparticles is calculated as

η= timesm

I t VRF 60p

p d

ion

n

(4)

where mp = particle mass (ng) Ip = particle signal intensity in thesample (cps) RFion = ICP-MS response factor from the calibrationcurve of the ionic titanium standards (cpsμgL) td = dwell time (s)and V = sample flow (mLmin) To calculate the particle massconcentration in the diluted sample suspension the masses of all

Figure 2 Signal distribution graph plotting the frequency of the ICP-MS response in data points as a function of the ICP-MS response ICP-MSresponses to the left of the minimum depict background and ions those to the right the minimum particles

Figure 3 Typical example of a particle size distribution as produced by single-particle ICP-MS analysis

Journal of Agricultural and Food Chemistry Article

dxdoiorg101021jf5011885 | J Agric Food Chem XXXX XXX XXXminusXXXD

individual particles are summed and corrected for nebulizationefficiency and sample flow

η=

sumtimes times

Cm

V 1000mp

n (5)

In eq 5 Cm = particle mass concentration (ngL) Finally the particlesize expressed as the particlersquos diameter (and assuming a sphericalparticle shape) is calculated for each particle as

πρ= timesd

m610p

p

p

43

(6)

where dp = particle diameter in the sample (nm) and ρp = particledensity (gmL) The individual particle sizes can be used to produce asize distribution graph (Figure 3) For all of these calculations atemplate in Microsoft Excel was produced and used throughout

RESULTS AND DISCUSSIONWe compared the size distributions of TiO2 particles asdetermined with three different analytical approaches in threetypes of samples First we characterized TiO2 powders thataccording to the manufacturers are E171 food additivessecond we characterized food products (cakes candy andchewing gum) and finally personal care products (toothpaste)were examinedTitanium Content and Particle Size Determination of

TiO2 Materials Total Titanium Content of TiO2 MaterialsThe total TiO2 content of the E171 materials was determinedas total titanium using HNO3HF digestion and ICP-MSanalysis of the extracts followed by recalculation of results totitanium dioxide The results show that the total TiO2 contentis in the range of 90minus99 for five of the seven E171 materialsand is 74 and 83 for the other two materials Taking intoaccount the measurement uncertainty of the total-Ti method(plusmn6) and the fact that these materials may contain somewater five of the seven E171 materials consist of pure TiO2 Itis known that some TiO2 materials are coated with aluminumand silicon oxides or polymers to increase photostability andprevent aggregation2223 Only the packaging of one of the twosamples with a lower TiO2 content declares that the materialmay contain aluminum hydroxide amorphous silica aluminumphosphate and water No further information about thechemical composition of these two TiO2 materials wasavailableSize Distributions of TiO2 Materials Determined with SEM

The size distribution of the particles in the TiO2 materials wasdetermined by SEM For this the materials were suspended inwater and stabilized by BSA15 Representative SEM images ofthe TiO2 materials are shown in Figure 4 The images on theleft side in Figure 4 give a good picture of the largeagglomerates that are encountered in the samples whereasthe images on the right side allow the estimation of the primaryparticle sizes Detected particles were automatically countedusing image analyses software and divided into size bins of 25minus40 40minus65 65minus100 100minus160 160minus250 250minus400 400minus650650minus1000 and 1000minus1600 nm The number percentage of theparticles in each bin is then plotted against the mean particlesize of the size bin to produce a number-based size distribution(Figure 4) In general the size distributions of the seven TiO2materials are very alike and range from 30 to 600 nm with theapex in all size distributions between 200 and 400 nm About10 of the particles have dimensions below 100 nm The actualpercentage of particles with at least one dimension lt100 nm

will be higher because the image analysis software determinesan area equivalent diameter (diameter projected area dpa) forthe particles As a consequence particles with an aspect ratiodifferent from 1 will be ldquoaveragedrdquo into round particles leadingto an underestimation of the number of particles with onedimension lt100 nm From the image analysis it appears that forgt80 of the particles large and small the aspect ratio is in therange of 08minus12 As a consequence it is expected that theunderestimation of the number of particles with one dimensionlt100 nm is limited to 10minus20 That all materials appear to bemore or less identical probably reflects the production processof this material The most common production method forTiO2 particles is the chloride process or the sulfate process

24 Inindustrialized countries chloride processes appear to be favoredover sulfate processes for environmental economic andqualitative reasons25 Therefore there is a reasonable chancethat all investigated TiO2 materials are produced by the sameprocess

Figure 4 Representative SEM images of TiO2 materials The image onthe left shows the agglomerates whereas the image on the right showsthe individual particles The number-based size distributions producedby image analysis of the SEM images are shown below the SEMimages (the error bars indicate the 95 confidence intervals of thenumber percentage in each particular size bin)

Journal of Agricultural and Food Chemistry Article

dxdoiorg101021jf5011885 | J Agric Food Chem XXXX XXX XXXminusXXXE

The smallest size bin used in our measurements was 25minus40nm Particles in that size range could still be detectedidentified and counted by the image analysis software TiO2particles lt20 nm cannot be reliably identified and counted bythe image analysis software Manually the operator can detectTiO2 particles with diameters down to 10minus20 nm in cleansuspensions However manually detecting small particles is nota reasonable option because many fields have to be viewed toproduce a reliable particle number Thus our measurementsshow that the lower size limit of detection for routine SEManalysis is around 20 nm for TiO2 particlesSize Distributions of TiO2 Materials Determined with AF4-

ICP-MS Strictly speaking AF4 is a separation technique and isnot able to determine size or a number-based size distributionAs there are no well-characterized TiO2 particles of differentsizes (ie reference materials for size calibration) we used atheoretical approach based on the well-established theory offlow field-flow fractionation for size calibration as explainedunder Experimental ProceduresThe result of the AF4-ICP-MS analysis is a chromatogram in

which the signal height is a measure for the mass of titaniumdetected at a certain retention time Following mass calibrationthis signal height is expressed as the mass of Ti or TiO2 andwith the time scale recalculated as a particle size scale based onan equation from flow field-flow fractionation theory describedunder Experimental Procedures this results in a mass-basedsize distribution However to apply the EU definition fornanomaterials this mass-based distribution has to be trans-formed into a number-based distribution Therefore this mass-based size distribution is transformed into a number-based sizedistribution using the equation under Experimental Proceduresand assuming that particles have a spherical shape However ifthe mass is converted into a number of particles this numberdepends on particle size If the mass remains constant as shownin Figure 5 the number of particles increases strongly when

going to smaller particle (notice that the y-scale is a log scale)especially at small particle sizes because particle size and massare related by d3 (see eq 6 under Experimental Procedures)This means that mass to number transformations have largeconversion factors for small particles that easily lead toerroneous results In practice the beginning of chromatogramstends show small spikes noise nonparticle peaks or smallelectrical offsets If these phenomena are also transformed frommass- to number-based small particle sizes will completely anderroneously dominate the number-based distribution resultingin a dramatic shift of the size distribution to smaller particlesizes From the chromatograms in this study it was concludedthat the smallest TiO2 particle size that can be determined and

reliably transformed from mass- to number-based distributionsis around 20 nm

Comparing the Measured Size Distributions of TiO2Materials Conversion of the retention time into particle sizeand the mass-based distribution into a number-baseddistribution allows us to derive number-based size distributionsfor the TiO2 materials The AF4 size distribution of thematerials for which the EM size distribution is shown in Figure4 is shown in Figure 6 In general the number-based size

distribution as determined by EM and AF4 shows two types ofdifferences First the apex of the size distributions originatingfrom AF4 generally is between 200 and 300 nm whereas in EMsize distributions it is between 200 and 400 nm Second theEM size distributions contain larger particles (diameter gt 600nm) than the AF4 size distributions In AF4 large particles(diameter gt 400 nm) may actually drop on and roll over themembrane surface at velocities that are higher than normalelution26 As a result these large particles elute at shorterretention times thereby simulating smaller sized particles Inthe number-based size distribution this will become visible as ashift to smaller particle sizes As before the percentage ofparticles with a size lt100 nm is determined In the AF4 analysisthis ranges from 10 to 15 with an average of 12 This iscomparable with the 10 of particles with sizes below 100 nmthat were found in the EM analyses

Titanium Content and Size Distributions in Food andPersonal Care Products In addition to the TiO2 materialsfood and personal care products were also analyzed for totaltitanium content and with AF4-ICP-MS for titanium content aswell as size distribution An attempt was made to determine thesize distribution of the TiO2 particles in the extracts of theseproducts with SEM However matrix constituents that werestill present made it impossible to make a proper image analysisof the particle sizes As an alternative the extracts wereanalyzed using single-particle ICP-MS (sp-ICPMS) because thistechnique also produces a number-based size distribution Intotal 24 food products and 3 personal care products werepurchased in regular shops in The Netherlands Products that

Figure 5 When the mass remains constant the number of particlesrapidly increases especially at smaller particle sizes

Figure 6 Number-based size distribution of two TiO2 materials asdetermined with AF4-ICP-MS and after recalculation from mass-basedto number-based

Journal of Agricultural and Food Chemistry Article

dxdoiorg101021jf5011885 | J Agric Food Chem XXXX XXX XXXminusXXXF

listed ldquoE171rdquo or ldquotitanium dioxiderdquo on the package as well as afew products with a typical ldquowhite colorrdquo were selected for thisstudyTotal Titanium Content of Food and Personal Care

Products All 27 products were digested with a combination ofHNO3 HF and H2O2 and the total Ti content was determinedusing ICP-MS Total Ti recovery was determined by spikingthree products of different matrices in triplicate with knownamounts of a TiO2 reference material NM-103 The total Tirecovery ranged from 92 to 109 with an average of 96 plusmn 6Method blanks showed only low amounts of titanium that ison average lt2 μg of Ti which corresponds to lt0005 mg Tigproduct The quantification limit of the digestion and detectionmethod was set to 001 mg Tig product The total Ti contentof the products ranges from 54 mg Tig product for a chewinggum to 001 mg Tig for a white-topped cookie Four productshad total Ti levels that were below the quantification limit ofthe method The highest levels of total Ti are found in chewinggums and in toothpastes and the lowest in bakery productssuch as cookies and cupcakes (Figure 7) The results arecomparable to the total Ti content in food items reportedrecently by Weir et al8

Titanium Content of Food and Personal Care ProductsDetermined by AF4-ICP-MS All food and consumer productswith a detectable amount of total Ti have been analyzed withAF4-ICP-MS for particle-based Ti content as well as theparticle size distribution For this purpose a separate samplepreparation was set up using H2O2 to oxidize the sample matrixThe use of strong acids was avoided because these are notcompatible with the AF4 analysis H2O2 oxidation worked forall samples with the exception of the chewing gums wheredebris of the guar gum remained Because the TiO2 particles areexpected to be present in the outer water-soluble coating of thechewing gum8 the removal of this debris probably does notinfluence the result The time between sample preparation andmeasurement with AF4-ICP-MS was kept as short as possibleand was no more than 4 h After sonication of the sampleextract the extract was kept in constant movement on a rollerbank until analysis Prior to injection the sample wasthoroughly mixed to ensure homogeneous distribution ofsuspended particles Because ions are removed in AF4separation during the focusing of the analyte on thesemipermeable membrane mass calibration can be done onlywith a particle-based standard In the absence of a reliable well-

characterized TiO2 particle standard one of the TiO2 powderscharacterized as pure 99 TiO2 was used for this purpose Theparticle-based Ti content of the food and personal careproducts as determined with AF4 and expressed as milligramsof Ti particle per gram of product is presented as the red right-side bars in Figure 7 The side-by-side comparison of the totalTi content and the particle-based Ti content as determinedwith AF4-ICP-MS in each product shows that they arecomparable certainly at Ti concentrations gt05 mg Tigproduct At lower concentration the results from the AF4determination are generally lower than those from the totalICP-MS method This can be explained by adsorptionprocesses in the AF4 separation that become visible in thelower concentration range Another reason may be that verylarge particles aggregates or agglomerates (gt1600 nm) may bepresent in the samples and are lost in the AF4 determinationFor products containing gt05 mg Tig the results show thatmost of the total Ti originates from particles that is TiO2particles and that no large amounts of titanium areunaccounted for in the AF4 analysis

Number-Based Size Distributions in Food and PersonalCare Products Next to the particle-based Ti content of thesamples number-based size distributions were determined(Figure 8) From products containing lt01 mg Tig product noreliable number-based size distributions could be produced Ingeneral the calculated number-based size distributions in fooditems and personal care products based on AF4-ICP-MSanalysis are comparable with the number-based size distributionas determined for TiO2 in E171 materials with EM The apex ofthe size distribution was mostly found around 200 nm which isin the size range of the primary particle sizes found with EMThe results indicate that a part of the TiO2 particles havediameters below 100 nm On a number-based basis the averagepercentage of particles with diameters lt100 nm is around 8comparable with the 12 and 10 that were found for thenumber-based size distributions of TiO2 in E171 with AF4-ICPMS and EM analyses

Size Distributions Determined Using Single-ParticleICPMS Because EM analysis was not feasible in products sp-ICP-MS was used to determine a number-based sizedistribution of the TiO2 particles in food and personal careproducts The sample extracts that were used for the AF4analysis were diluted 10000 times to get in the proper workingrange of the sp-ICP-MS method Counting between 300 and

Figure 7 Total Ti content in food and personal care products determined with ICP-MS (blue left-side bars) and the particle-based Ti content in thesame products determined with AF4-ICP-MS analysis (red right-side bars) The numbers after the product description refer to sample codes

Journal of Agricultural and Food Chemistry Article

dxdoiorg101021jf5011885 | J Agric Food Chem XXXX XXX XXXminusXXXG

3000 particle peaks in each time scan gives enough informationto produce a reliable number-based size distribution WhereasTiO2 particles produce discontinuous signals (peaks) in sp-ICP-MS small particle peaks will disappear in the backgroundnoise and can no longer be distinguished Ti is measured at mz48 or 49 which is not that specific or sensitive respectivelyThe presence of biatomic ions such as 32S16O and 36Ar12C andisobaric interference by 48Ca result in a continuous backgroundThis background noise for Ti measured at mz 48 issubstantially higher than that for silver (mz 107) or gold(mz 197) and therefore the minimum particle size that can bedetermined for titanium is also higher Whereas for gold andsilver 20 nm nanoparticles can be determined the particle sizedetection limit for TiO2 particles in the sp-ICP-MS method is50 nm The use of a collision cell did not improve the situationbecause not only the background noise but also the Ti particlepeaks were reduced resulting in no improvement of the signalnoise ratioComparing the Measured Size Distributions of TiO2

Materials Compared to the calculated AF4-ICP-MS number-based size distribution (Figure 8) the measured sp-ICP-MSnumber-based size distributions (Figure 9) generally are shiftedto the left that is toward smaller particle sizes as illustrated bythe results for chewing gum 623 in Figure 9 Nevertheless theAF4-ICP-MS and sp-ICP-MS number-based size distributionsof TiO2 particles in the products are comparable Furthermorethe distributions of the TiO2 particles in the products are alsocomparable with the number-based size distributions of theTiO2 (E171) materials as determined with AF4-ICP-MS andwith EM In all cases the apex of the size distribution is foundaround 200 nm within the size range of the primary particlesizes found with EM The sp-ICP-MS results indicate that 5minus10 of the TiO2 particles have diameters below 100 nm This iscomparable to or somewhat lower than what was found in theEM and AF4-ICP-MS analysis and may in part be explained bythe difference in the smallest particle that can be detected ForEM in the analysis of the TiO2 materials the smallest TiO2particle that could be detected was around 20 nm for AF4-

ICP-MS in the analysis of TiO2 materials and food and personalcare products it was also around 20 nm whereas in the sp-ICP-MS analysis of the food and personal care products the smallestdetectable TiO2 particle had a diameter of 50 nmThese practical size detection limits for TiO2 particles

introduce a certain and perhaps significant bias because the sizerange between 1 and 20 nm is excluded Whereas Figure 7shows that gt95 of the total titanium (mass-based)concentration is explained by TiO2 particles gt20 nm if evenonly 1 of the total titanium (mass-based) concentration existsof TiO2 particles lt20 nm this will change the determinednumber-based particle size distribution significantly As aconsequence it is not clear whether TiO2 in food products isa nanomaterial and the results especially show the inability ofcurrent state of the art methods to support the EUrecommendation for the definition of nanomaterials

Potential for Consumer Exposure Assessment In this study7 TiO2 materials 24 food products and 3 personal careproducts were studied for their titanium dioxide content andthe number-based size distribution of TiO2 particles thereinThree principally different methods have been used todetermine the number-based size distribution of TiO2 particlesin E171 materials and food and personal care productsComparable size distributions were found and from these itwas determined that 5minus12 of the TiO2 particles in thesematerials and products were lt100 nm These data are suitableto be used in an exposure study further refining the exposureassessment by Weir et al for the US population8 Previouslywe assessed the SiO2 content of food products27 andsubsequently assessed its fate during in vitro digestion28

There we concluded that the intestinal epithelium is exposed tonanoscale material The fate of TiO2 particles while in thehuman digestive tract is unknown and should be evaluated toallow a comparison with the existing oral toxicity studies usingdifferent forms of the TiO2 NPs Although the appliedanalytical methods showed practical size limits all of thesemethods and especially sp-ICP-MS can be elegantly employedfor further studies on the fate of food grade TiO2 to bridge the

Figure 8 Number-based size distribution of the TiO2 material in atoothpaste and chewing gum sample as determined with AF4-ICP-MSand after transformation of the mass-based to a number-baseddistribution

Figure 9 Number-based size distribution of the TiO2 material intoothpaste as determined directly with sp-ICPMS

Journal of Agricultural and Food Chemistry Article

dxdoiorg101021jf5011885 | J Agric Food Chem XXXX XXX XXXminusXXXH

gap between our findings and recent oral toxicity studies toguarantee consumer safety

AUTHOR INFORMATIONCorresponding Author(RJBP) E-mail ruudjpeterswurnl Phone +31 317 48 0671FundingThis research was commissioned and financed by TheNetherlands Food and Consumer Product Safety AuthorityNotesThe authors declare no competing financial interest

REFERENCES(1) Johnson R W Thiele E S French R H Light-scatteringefficiency of white pigments an analysis of model coreminusshell pigmentsvs optimized rutile TiO2 TAPPI J 1997 11 233minus239(2) Braun J H Titanium dioxidea review J Coatings Technol1997 69 59minus72(3) Linak E Schlag S Kishi A Chemical Economics HandbookTitanium Dioxide (Marketing Research Report) SRI InternationalMenlo Park CA USA 2002(4) Swiler D R Pigments inorganic In Kirk-Othmer Encyclopedia ofChemical Technology Wiley New York 2005(5) Robichaud C O Uyar A E Darby M R Zucker L GWiesner M R Estimates of upper bounds and trends in nano-TiO2production as a basis for exposure assessment Environ Sci Technol2009 43 4227minus4233(6) Hendren C O Mesnard X Droge J Wiesner M REstimating production data for five engineered nanomaterials as a basisfor exposure assessment Environ Sci Technol 2011 45 2562minus2569(7) European Commission European Parliament and CouncilDirective 9436EC of 30 June 1994 on colours in foodstuffs Off JEur Communities 1994 L237 13minus29(8) Weir A Westerhoff P Fabricius L Hristovski K van GoetzN Titanium dioxide nanoparticles in food and personal care productsEnviron Sci Technol 2012 46 2242minus2250(9) Tassinari R Cubadda F Moracci G Aureli F DrsquoAmato MValeri M De Berardis B Raggi A Mantovani A Passeri D RossiM Maranghi F Oral short-term exposure to titanium dioxidenanoparticles in Sprague-Dawley rat focus on reproductive andendocrine systems and spleen Nanotoxicology 2013 DOI 103109174353902013822114(10) Gui S Sang X Zheng L Ze Y Zhao X Sheng L Sun QCheng Z Cheng J Hu R Wang L Hong F Tang MIntragastric exposure to titanium dioxide nanoparticles inducednephrotoxicity in mice assessed by physiological and gene expressionmodifications Part Fibre Toxicol 2013 10 4minus10(11) IARC Carbon Black Titanium Dioxide and Talc IARCMonograph 93 on the Evaluation of Carcinogenic Risks to HumansIARC Lyon France 2010 downloadable from httpmonographsiarcfrENGMonographsvol93mono93pdf(12) Bouwmeester H Dekkers S Noordam M Y Hagens W IBulder A S De Heer C Ten Voorde S E C G Wijnhoven S WP Marvin H J P Sips A J A M Review of health safety aspects ofnanotechnologies in food production Regul Toxicol Pharmacol 200953 52minus62(13) European Commission Commission Recommendation of 18October 2011 on the definition of nanomaterial 2011696EUEuropean Commission Luxembourg 2011(14) Linsinger T Roebben G Gilliland D Calzolai L Rossi FGibson N Klein C Requirements for the Implementation of theEuropean Commission Definition of the Term ldquoNanomaterialrdquo JRCReference report EUR 25404 EN EUR Scientific and Technicalresearch series 2012 ISSN 1831-9424 DOI 10278763490(15) Jensen K A Kembouche Y Christiansen E Jacobson N RWallin H Guiot C Spalla O Witschger O Final Protocol for

Producing Suitable Manufactured Nanomaterial Exposure MediaNANOGENOTOX deliverable report 3 July 2011 downloadablefrom httpwwwnanogenotoxeufilesPDFweb20nanogenotox(16) Moon M H Park I Kim Y Size characterization ofliposomes by flow field-flow fractionation and photon correlationspectroscopy J Chromatogr A 1998 813 91minus100(17) Park I Paeng K J Yoon Y Song J H Moon M HSeparation and selective detection of lipoprotein particles of patientswith coronary artery disease by frit-inlet asymmetrical flow field-flowfractionation J Chromatogr B 2002 780 415minus422(18) Degueldre C Favalgar P Y Wold S Gold colloid analysis byinductively coupled plasma-mass spectrometry in a single particlemode Anal Chim Acta 2006 555 263minus268(19) Laborda F Lamanna J J Bolea E Castillo J R Selectiveidentification characterization and determination of dissolved silver(I)and silver nanoparticles based on single particle detection byinductively coupled plasma mass spectrometry J Anal At Spectrom2011 26 1362minus1371(20) Pace H E Rogers N J Jarolimek C Coleman V EHiggins C P Ranville J F Determining transport efficiency for thepurpose of counting and sizing nanoparticles via single particleinductively coupled plasma mass spectrometry Anal Chem 2011 839361minus9369(21) Peters R J Herrera Rivera Z van Bemmel G Marvin H JBouwmeester H Development and validation of single particle ICP-MS for sizing and quantitative determination of nano-silver in chickenmeat Anal Bioanal Chem 2014 DOI 101007s00216-013-7571-0(22) Labille J Feng J H Botta C Borscgneck D Sammutt MCabie M Auffan M Rose J Bottero J Y Aging of TiO2nanocomposites used in sunscreen Dispersion and faste of thedegradation products in aqueous environment Environ Pollut 2010158 3482minus3489(23) Carlotti M E Ugazio E Sapino S Fenoglio I Greco GFubini B Role of particle coating in controlling skin damagephotoinduced by titania nanoparticles Free Radical Res 2009 43312minus322(24) Macwan D P Dave P N Chaturvedi S A review on nano-TiO2 sol-gel type synthesis and its applications J Mater Sci 2011 463669minus3686(25) Linak E Inoguchi Y In Chemical Economics HandbookTitanium Dioxide SRI Consulting Menlo Park CA USA 2005(26) Schimpf M E Caldwell K Giddings J C In Field-FlowFractionation Handbook Wiley New York 2000(27) Dekkers S Krystek P Peters R J Lankveld D X BokkersB G Van Hoeven-Arentzen P H Bouwmeester H Oomen A GPresence and risks of nanosilica in food products Nanotoxicology 20115 393minus405(28) Peters R Kramer E Oomen A G Herrera Rivera Z EOegema G Tromp P C Fokkink R Rietveld A Marvin H J PWeigel S Peijnenburg A A C M Bouwmeester H Presence ofnano-sized silica during in vitro digestion of foods containing silica as afood additive ACS Nano 2012 6 2441minus2451

Journal of Agricultural and Food Chemistry Article

dxdoiorg101021jf5011885 | J Agric Food Chem XXXX XXX XXXminusXXXI