Physical characterization of titanium dioxide

International Journal of Cosmetic Science, 2014, 1–12

doi: 10.1111/ics.12113

Physical characterization of titanium dioxide nanoparticles T. A. Egerton* and I. R. Tooley† *School of Chemical Engineering and Advanced Materials, Merz Court, Newcastle University, Newcastle Upon Tyne, NE1 7RU, UK and †Croda Europe Ltd, Sun Care and Biotechnology, Foundry Lane, Ditton, Widnes, Cheshire, WA8 8UB, UK

Received 23 December 2013, Accepted 10 January 2014

Keywords: chemical analysis, claim substantiation in vivo/in vitro, suncare/UV protection

Synopsis OBJECTIVE: The objective of this study was to review six measurement methods (X-ray line broadening, electron microscopy, static light scattering, dynamic light scattering, X-ray sedimentation and surface area determination), which are widely used for the characterization of ultrafine inorganic oxides used in cosmetic formulations. Depending on the processing that they have received and the system in which they are examined, these oxides can exist as primary particles, strongly bound aggregates or weakly bound agglomerates. METHODS: The example of titanium dioxide, TiO2, is used to consider which type of particle is being measured in a particular case, and the factors which influence the ‘size’ that is generated by a particular method. Where appropriate, a correlation is made between results of different measurements. RESULTS: Results for a particular set of four cosmetic grade TiO2’s are presented and examined, in the context of a much broader set of measurements taken from the scientific literature. CONCLUSION: In general, X-ray line broadening, electron microscopy and surface area measurements led to estimates of the size of primary particles. By contrast, both sedimentation and light scattering measurements measured the size of the secondary particles, and the figures which were generated depended on the dispersion conditions used for preparation of the measurement samples. For poorly dispersed or lightly milled samples, the size may be dominated by the presence of weakly bound agglomerates, but even when the sample is well dispersed or heavily milled, the reported size cannot be less than that of the aggregates. sume Re OBJECTIF: L’objectif de ce document est d’examiner six methodes de mesure (elargissement des rayons X, microscopie electronique, diffusion de lumiere statique, diffusion dynamique de la lumiere, sedimentation des rayons X et determination de la surface) qui sont largement utilises pour la caracterisation des oxydes inorganiques ultrafines utilisees dans des formulations cosmetiques. En fonction du traitement qu’ils ont recßu, et le systeme dans lequel ils sont examines, ces oxydes peuvent exister sous forme d’agregats de particules primaires, fortement lies par des liaisons ou des agglomerats faiblement lies.

Correspondence: Ian R. Tooley, Croda Europe Ltd, Sun Care and Biotechnology, Foundry Lane, Ditton, Widnes, Cheshire WA8 8UB, UK. Tel.: 0151 4239287; fax: 0151 4233205; e-mail: ian.tooley@croda. com

METHODES: L’exemple du dioxyde de titane, TiO2, est utilise pour examiner quel type de particules est mesuree dans un cas particulier, et les facteurs qui influent sur la «taille» generee par un procede particulier. Le cas echeant, une correlation est etablie entre les resultats des differentes mesures. RESULTATS: Des resultats pour un ensemble particulier de quatre qualites cosmetiques de TiO2 sont presentes et examines dans le cadre d’un ensemble beaucoup plus large de mesures prises a partir de la litterature scientifique. CONCLUSION: En general, l’elargissement des rayons x -ray, la la mesure de superficie ont conduit a microscopie electronique et a des estimations de la taille des particules primaires. Par contre, la sedimentation ainsi que les mesures de diffusion de lumiere ont mesure la taille des particules secondaires, et les chiffres qui ont ete generes dependent des conditions de dispersion utilisees pour la preparation des echantillons de mesure. Pour les echantillons mal disperses ou broyes grossierement, la taille peut être dominee par la presence d’agglomerats faiblement lies, mais même lorsque l’echantillon est bien disperse ou broye finement, la taille indiquee celle des agregats. ne peut pas être inferieure a

Introduction The requirement to regulate the use of nanoparticles in industry makes it necessary to identify agreed and well-understood methods of characterizing these particles and also to agree and specify the related measurement protocols. This study is a contribution to that goal. Although the term Nanoparticle is widely used in the scientific community, many would struggle to provide a precise definition. A current draft from the International Standards Organisation (ISO) defines ‘nanoscale’ as ~1–100 nm and a nanoparticle as an object having three external dimensions on the nanoscale [1]. (There are analogous definitions of nanorod, nanoplate and nanoobject.) It is important to recognize that crystals or particles within the nanoscale range may form aggregates or agglomerates with external dimensions that are outside the nanoscale range. The ISO proposal describes these as ‘nanostructured aggregates’ or ‘nanostructured agglomerates’. Although the terms aggregate and agglomerate are often confused or used interchangeably, they are in fact quite distinct and are defined as follows:

•

Aggregate: particle comprising strongly bonded or fused particles where the resulting external surface area may be significantly smaller than the sum of calculated surface areas of the individual components.

© 2014 Society of Cosmetic Scientists and the Societe Francßaise de Cosmetologie

1

T. A. Egerton and I. R. Tooley

Physical characterization of nanoparticles

•

Agglomerate: collection of loosely bound particles or aggregates or mixtures of the two where the resulting external surface area is similar to the sum of the surface areas of the individual components.

Much of the current debate around the use of nanotechnology centres on ‘personal care products’, particularly the most commonly used inorganic ‘cosmetic nanomaterials’, TiO2 and ZnO in sunscreens [2]. Fine-particle TiO2 is manufactured as nanoparticles which are typically 5–20 nm in size. However, to quote this as ‘the particle size’ is somewhat misleading, because, as shown in Fig. 1, these nanoparticles may form tightly bound aggregates or more loosely bound agglomerates which obviously have a somewhat larger size. Aggregates formed during drying are sometimes referred to as cemented aggregates, and those formed when the particles are heated are sometimes called sintered aggregates. Thus, Xu et al. [3] report that when primary particles smaller than 10 nm were heated, they clustered together to form larger particles, consisting of many primary particles. The forces required to break apart the aggregates are far greater than those encountered during the production of cosmetic products or application of these products onto skin. Therefore, these aggregates are the smallest particles, which actually occur in a real system. Aggregates may join together to form more loosely bound agglomerates, which often correspond to the flocculates of classical colloid science. For example, the as-supplied dry powder form of inorganic sunscreens may contain agglomerates with particle sizes >1 µ, well outside the nanoscale range. These agglomerates must be broken down to some degree in the final formulations, because Mie theory tells us that such large particles would not be effective as sunscreens and would be very white and opaque on skin [2]. (Seipenbusch et al. [4] reported that specific kinetic energies of 2,700 J kg1 were needed to fracture ~50% of the ~300 nm TiO2 agglomerates of 110 nm aggregates formed by partial sintering of ~6 nm crystals.) As a more general example, it has been demonstrated that milling breaks down agglomerates that are present in dispersions of TiO2 photocatalysts and that this modifies the measured optical properties and hence the photocatalytic activity [5, 6]. The development of techniques to properly characterize materials in final products, such as paints or suncreams, is still far from completion. A plethora of particle sizing techniques is available, but some methods measure the component nanoparticles, whereas others measure the aggregates and/or agglomerates. In addition, quoted particle sizes can vary widely in that any size measurement

depends on how the samples are prepared for measurement. To compare different powders, it is necessary not only to measure them by the same technique, but also to prepare the samples in the same way. This study provides an overview of six commonly used techniques and illustrates their use for characterizing nanotitanium dioxide. To clearly highlight what each technique measures, and to point out anomalies that can occur, measurements of four real nanoparticulate TiO2 grades by five of these techniques are presented and considered in the context of the different structures shown in Fig. 1. Therefore, we briefly summarize the origins of these samples and the details of the measurements. Materials and methods Samples A and B are commercially available materials of titanium dioxide (ex Tayca Corporation) prepared from the same base particles and coated by silica and alumina and alumina and stearic acid, respectively. Sample C was prepared by hydrolysis of TiCl4 and coated by alumina and stearic acid, and sample D was made by a similar process to sample C with an additional processing step to grow the base crystals before applying a coating of alumina and stearic acid. Samples C and D were kindly made available by Croda Europe Ltd.

X-ray diffraction Powder X-ray diffraction patterns were measured with Cu Ka radiation in a Siemens D5000 diffractometer equipped with an energydispersive detector acting as a monochromator. The diffractograms showed no phases other than rutile. The data were analysed by fitting the diffraction pattern between 22 and 48° 2h with a set of peaks corresponding to the reflection positions for rutile. A mean crystallite size of 9 nm as determined for the rutile 110 reflection (at ~27.4° 2h) based on its integral breadth according to the principles of the method of Stokes and Wilson [7].

Electron microscopy Measurements were taken with a Philips 410 transmission electron microscope (TEM) operating at an accelerating voltage of 100 kV. Particle shape was determined by a slightly modified version of the method described in ISO TS 11937-1 (Nanoscale titanium dioxide powdered form – Part 1: characteristics and measurement; Annex C; determination of average primary particle size by TEM). The

Figure 1 Schematic representation of the aggregation and agglomeration of titanium dioxide nanoparticles.

2

© 2014 Society of Cosmetic Scientists and the Societe Francßaise de Cosmetologie International Journal of Cosmetic Science, 1–12



average area-equivalent diameter, median length and median width were calculated from transmission electron micrograph (TEM) measurements based on 600 particles. Dynamic light scattering The DLS particle size measurements were carried out using a Malvern Zetasizer Nano ZS. Samples were diluted to 0.1 wt.% using C12-C15 alkyl benzoate. The measurement duration was set to automatic, and five repeat measurements were taken at 25°C. The TiO2 samples were run using the refractive index measured using a Malvern Mastersizer 2000 [8]. Sedimentation An X-ray Disc Centrifuge, (XRDC; Brookhaven Instruments B1-XDC, Holtsville, NY, U.S.A.) was used to measure the particle size distribution of titanium dioxide particles dispersed in C12-C15 alkyl benzoate. Forty grams of titanium dioxide powder dispersed in 60 g of C12-C15 alkyl benzoate containing dispersing agent was diluted by isopropyl myristate (IPM) to produce a 2% weight/ volume solution of the particles in the sedimentation liquid. The liquid was introduced into the XRDC and run at 3000 r.p.m.; full separation of the particles was achieved after ~2 h.

where b is a constant, typically 0.9, k is the X-ray wavelength, and Bs is the broadening caused by instrumental factors. The smaller the crystal, the broader the diffraction as shown for the 101 line of anatase nanocrystals in Fig. 2. The method is limited, in the case of TiO2 crystals, to D < 150–200 nm, because Bm must be significantly larger than Bs. The above discussion makes clear that the method is restricted to crystalline particles. However, if particles of two crystalline forms are present, the size of each may be separately determined provided their diffraction maxima do not overlap. For non-spherical crystals, the value of D determined by line broadening depends on which diffraction maximum is measured. The measured distance D is in a direction perpendicular to the hkl lattice planes of the, 2hhkl, diffraction maximum. This is shown in Fig. 3 for a tetragonal crystal, such as rutile (a = b = 0.2959, c = 0.4594), for which the most important faces are and which are usually elongated in the c direction. For most rutile crystals, the crystallite size relates to the shortest dimension, the diagonal of the plane which is perpendicular to the c axis. The longest dimension, parallel to the c axis, may then be calculated from the peak width of the diffraction (chosen in preference to the diffraction for practical

Surface area Prior to measurement of adsorption, samples were evacuated at ~1 Pa for 12 h at 110°C and then allowed to cool naturally. Adsorptions were measured at 195°C using a McBain balance. Nitrogen (1.3 kPa) was introduced to the vacuum system and allowed to equilibrate for 60 min. The spring extensions due to the uptake of nitrogen were then measured and the nitrogen pressure recorded, before repeating the process for increasing pressures of nitrogen (1.3–80 kPa). Brunauer–Emmett–Teller (BET) surface areas assumed rN2 = 0.162 nm2 [9]. Results Measurement of crystal size by broadening of X-ray diffraction maxima

bk pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi D¼ cos 2h B2m B2s

Figure 2 The intensity as a function of 2h of the X-ray diffraction peaks corresponding to the 101 line of anatase crystallites of (1) 53, (2) 22, (3) 11, (4) 8.5, (5) 6.5, (6) 5.0 and (7) 3.8 nm. [Reprinted with permission from Anpo et al., J. Phys. Chem. 4305–4310 (1987) 91 Copyright (1987) American Chemical Society.]

001

110

110

The identification of the crystalline structure of nanoparticles, for example whether a TiO2 is anatase or rutile, uses powder X-ray diffraction (PXRD) [9]. Diffraction peaks occur at increasing scattering angles (2h), and each corresponds to particular lattice spacing, dhkl. Thus, the strongest diffraction peak of anatase at ~25° corresponds to reflection from the 101 planes, and the strongest line from rutile corresponds to reflection from the 110 planes at 2h = 27.4°. For sharp peaks, the crystals must be sufficiently large that at very small deviations from the 2hhkl maxima, destructive interference from successive lattice planes reduces the scattered Xray intensity to the background level. If the number of planes is insufficient for this condition to be met (i.e. if the crystals are very small), the diffraction maxima are broadened. The full peak width, Bm, at half the maximum peak height (FWHM) of the 2h diffraction maximum is related to the crystal size, D, by the Scherrer equation:

L002

XRD line

110 Planes

002 Planes

Figure 3 Schematic diagram showing the crystal planes of the rutile form of titanium dioxide. The most important faces are .


3



reasons). Length to width ratios of 2:1 are not uncommon: in most cases, it is the shortest dimension which is quoted. It is emphasized that the method measures the crystallite size of the primary particles (or nanoparticles) as shown in Fig. 1. Even if the particles aggregate and are chemically bonded, or fused together, the method normally measures the size of the individual primary particles. As illustrated diagrammatically in Fig. 4, it would only measure the size of the aggregate if the particles were so closely fused that the lattice planes of individual particles were aligned. Thus, a recent TEM study of ZnO [9] revealed both 200-nm-diameter spherical aggregates and rods ~4 9 1.7 lm long. However, the line broadening (FWHM) of the ZnO (101) peak revealed that the crystallite size of the primary particles was only 30 nm.

As the X-ray line broadening method measures the crystallite sizes of primary particles, it is not affected by the presence of amorphous coatings on the particles’ surface. Thus, the line broadening sizes of two powders, A and B, prepared from the same TiO2 were each 15 nm even though they had different surface coatings (A: silica and alumina; B: aluminium stearate). For equally sized crystals, the FWHM measure of peak breadth is a good size measure. When there is a distribution of sizes, the peak profile is the summation of the signals due to the separate crystal sizes. The measured signal may then be fitted to a simulation, which takes into account the overlaps from a group of signals. For particular nanoparticulate rutiles, designated C and D, the FWHM sizes of 9 and 31 nm corresponded to volume-averaged sizes of 8 and 28 nm, respectively, derived from a simulation programme. In principle, simulations can also be used to determine the amorphous fraction of TiO2 detectable by X-ray diffraction by comparing the calculated intensities with the measured peak area of the diffraction maxima [11]. Electron microscopy

(b) No common

(c) Planes oriented similarly and in register with each other.

orientation of planes (a) No common

orientation of planes

Figure 4 A representation of the need for individual crystal structure to be so clearly fused as to cause the lattice planes of individual particles to adopt a common orientation if the aggregate size is to be measured by line broadening.

B

Electron microscopy, usually transmission electron microscopy (TEM), quickly differentiates between particles of widely different size and/or shape. Qualitative interpretation of TEM images of the type shown in Fig. 5 often provides the standard by which other methods are judged [12]. Inspection of Fig. 5 reveals that the particles are not spherical and do not all have the same size. The range of image sizes allows the particle size distribution, p.s.d, to be assessed but because TEM fields of view are usually restricted to a few hundred particles (compared with the many thousands of crystals sampled by X-ray line broadening, sedimentation, laser scattering or surface area measurement), the influence of TEM sampling errors must be kept in mind. Very high magnification (x~100 000) facilitates the

D

Figure 5 Transmission micrographs of two TiO2 samples (B and D with XRD FWHM crystal sizes of 15 and 31 nm, respectively) illustrate the variability in shape and size of the primary particles and the aggregation and/or agglomeration between particles. Micrographs B and D were taken from an earlier study [2]. Micrograph B was published earlier as ‘B’; micrograph D is one of a set of micrographs of the sample designated ‘C’ in the earlier paper [2].

4




identification of a particular image as a single crystal, an aggregate or agglomerate. However, although the, necessarily, few high magnification images, ≤20, may establish that two types of particles fall within a broad size range, for example ~100 and 50 nm, respectively, there is insufficient statistical reliability to differentiate between closely related size distributions. Typically, a 5% standard deviation in the measurement of the mean crystal size requires 400–500 crystals to be sized. If the particles have a wide size range, the size categories may be arranged logarithmically but even then using too many categories may result in a noisy distribution (e.g. 400 crystals distributed between 40 size categories will introduce unacceptable statistical fluctuation.). The sizing may, of course, be carried out by an autoscanning technique linked to one of the many image-analysis software packages [13]. Such software can record the length and width of individual particles, and/or ‘count’ the micrograph images of different image areas, A. It also reduces the operator bias which may influence manual counts, particularly with respect to deciding whether irregular images represent agglomerate or overlapping primary particles. The influence of sample preparation, particularly the amount of shear applied during preparation, affects whether the measured images correspond to the primary particles, aggregates or agglomerates of Fig. 1. For 250 nm TiO2 pigment, the production of electron micrographs on which most crystals are separate and on which any residual aggregates are sufficiently well defined to allow the individual crystals to be distinguished has become routine [14]. Careful sample preparation also minimizes the influence of aggregation or agglomeration on the measured images of nanoparticulate TiO2.

major and minor axes of the TiO2 images and plotting the results in the form of a bar chart, Fig. 6, as was performed by Liang et al. [15]. As an alternative to averaging the major and minor axes, an area-equivalent diameter, {4A/p}1/2 may be calculated from a ‘count’ of the micrograph images of different image areas, A. Figure 7 shows size averages for sample B, the grade whose micrograph is shown in Fig. 5, measured by displaying the image on a touch-sensitive screen and drawing round the outer edges of the particles with a touch-sensitive pen. The software identifies the mean size (in this case, the number average defined below), the mode size (the most frequently occurring size) and the median (the size of the middle particle in the size sequence).

Non-spherical particles

Figure 6 TEM-derived size distributions based on the average of the major and minor axes of 120 TiO2 particles [15]. The dark bars represent the results, from [15], plotted as a number distribution; the light-coloured bars represent the same results calculated and plotted as a weight distribution.

A comparison of number and weight distribution Number frequency %

30 25 20 15 10 5 0 1

2

3

4

5

6

7

8

9

10

11

12

Average diameter per nm

The problem that most practical powders are not spherical and have a spread of sizes is often met by taking an average of the

LENGTH/WIDTH

AREA EQUIVALENT DIAMETER 30

40 Mean Mode SD/mean Median

25 20

22.5 nm 21.2 nm 20.0% 22.1 nm

Mean Mode SD/mean Median

30

3.77 3.36 25.55% 3.68

20

15 10

10 5 0 0

20

40

60 nm

80

100

120

0 0

2

4

WIDTH

6

8

LENGTH

50

20 Mean Mode SD/mean Median

40 30

12.2 nm 13.4 nm 16.7% 12.2 nm

Mean Mode SD/mean Median

15

45.8 nm 40.1 nm 29.1% 44.0 nm

10 20 5

10 0 0

20

40

60 nm

80

100

120

0 0

20

40

60 nm

80

100

120

Figure 7 The distribution of crystal length and width of a cosmetic-grade ultrafine particle TiO2, sample B, together with the length/width ratios and ‘areaequivalent diameters’. Micrographs of 600 particles were counted. Particle shape was determined by a slightly modified version of the method described in ISO TS 11937-1 Annex C.


5



Spread of sizes When the micrographs show that there is a spread of sizes as shown in Fig. 7 for sample B, an average diameter D is estimated. The simplest average, the number average, Dn, is defined by P ni Di

Dn ¼ P

where ni is the number of particles of diameter Di.

ni

Number averages emerge naturally from measurements of the number of particles of different sizes shown on electron micrographs. For the particle distribution shown in Fig. 6, the number average size is 7.9 nm (close to the value of 7.4 derived from X-ray line broadening). However, most applications of nanoparticles involve P a known weight of the sample, and the weight average wi Di

Dw ¼ P

wi

(where wi is the weight of particles of size i) is then of

interest. Thus, the weight of particles in each size range is plotted in Fig. 6 in addition to the number of particles. For the same results, the weight distributions appear to show a greater proportion of large particles, and the weight average size of 8.2 nm is larger than the number average size of 7.9 nm, reported by Liang et al. Weight averages, Dw, are generally bigger than number averages, Dn, because the contribution from big, heavy, particles is larger and this is so for samples A to D. Distributions presented in the Appendix S1 show that for sample B, Dw is ~2 nm larger than Dn. The difference between the number average Dn and the weight average Dw has been exemplified further in the Appendix S1, which accompanies the online version of this study, and which also summarizes the treatment of size distribution [16, 17] (D.F. Tunstall, Personal communication). Sizes of closely similar particles as determined from electron micrographs Sizing from electron micrographs is exemplified for the four nanoparticulate titania samples A to D whose X-ray line broadening sizes were reported in Section A of the results as 15, 15, 9 and 31 nm. As expected from the X-ray sizes, Fig. 5 shows the images of D to be much larger than those of B (and also, although not shown, of A and C). For sample A, with a silica–alumina surface, and B, with an aluminium stearate surface, an automated estimate of the number-averaged median area-equivalent diameter size gave 21 and 22 nm, respectively. It is clear that the sizes of ~22 nm of A and B suggested by TEM are larger than the, 15 nm, suggested by X-ray line broadening. This is a familiar pattern. In the study by Liang et al., referred to in Fig. 6, four samples of EM size 7.9, 10.1, 7.7 and 9.8 had the line broadening sizes of 7.4, 7.8, 6.7 and 8.5, respectively [15]. Similarly, the EM sizes, shown in Table 1, of a series of anatase samples synthesized Table 1 A comparison of XRD and EM sizes (determined with the accuracy of 10%.) for a series of anatase samples hydrothermally synthesized under varying conditions by Ismagilov et al. [16]

EM Size (nm) XRD Size (nm) EM/XRD Size

6

35

38

20

24

8

16

28

26

22

27

28

16

18

10

14

20

24

20 (Rutile 30) 1.1

1.3

1.4

1.3

1.3

0.8

1.1

1.4

1.1

under varying conditions are consistently larger than the corresponding line broadening results [17]. Large differences between EM and XRD and EM sizes may indicate aggregation of primary particles into well-ordered larger structures. Kumar et al. considered anatase agglomeration to cause their EM size to exceed the line broadening size (50 nm c.f. 14 nm) [18]. A recent study [11] of ZnO found crystallite sizes of ~30 nm when determined from the full-width at half-maximum (FWHM) of the ZnO (101) peak, whereas electron micrographs revealed 200 nm spherical nanoparticles together with rods of ~4 lm long and 1.7 lm wide. Contrasting results such as those of Ridley et al. who reported a HR-TEM size of 4 nm compared with a line broadening size of 7 nm for Ishihara ST01 powder may be a consequence of the necessarily small number of particles that are averaged in high-resolution TEM. Alternatively, the discrepancy may be due to differences in the geometry of the measurements or to a contribution to line broadening from microstrains and faulting within the crystals (which for reasons of space has not been discussed.) [11, 19]. Static light scattering Static light scattering, sometimes known as laser diffraction [20], measures the scattering of monochromatic laser light by a suspension of the powder, using a series of concentric detector elements. The particle size measurement is derived from a comparison of the measured angular dependence of the scattered light intensity with the calculated intensity per unit volume of spherical particles at the different detector elements. This calculation of nanoparticle sizes uses light scattering theory, for example Mie theory, and requires the refractive index of the particles to be known. Because small particles scatter over a wide angle, the lower limit of this type of measurement is typically 30–100 nm. The derived average sizes depend on the intensity of scattered light, and as this varies as d6 the sixth power of the particle diameter, d, the raw results, before correction by the instrumental software, are highly sensitive to the presence of large particles. Kato et al. [21] have pointed out that for a log-normal distribution with Dn = 150 nm, the intensity-average DI would be 227 nm if r = 40 nm, but 167 nm if r = 20 nm. This demonstrates that, just as for Dw/Dn, the broader the size distribution DI/ Dn, the larger the difference between the average sizes. In addition, Allen [22] points out that because non-spherical particles are measured over all orientations, the measured size distribution is broadened. Limitations of space restrict the emphasis on measurements of laser diffraction. However, a typical result, for Ishihara ST-01 anatase derived from published work by Ridley et al. [19], is shown in Fig. 8. A 200 mg kg1 stock dispersion of ST-01 in deionized water at pH 2.7 (HCl) was prepared by ultrasonication and added to the instrument reservoir to achieve an optimum obscuration level. The averaged results from 10 runs were plotted as a mean volumetric size distribution (the vertical bars represent one standard deviation). The median diameter of 2329 nm greatly exceeds the primary particle size of 4.6 nm from TEM, of 3.7 nm from ultra-small-angle Xray scattering (USAXS) and small-angle neutron scattering (SANS) and the 5.0 nm derived from BET analysis of adsorption measurements assuming an anatase density of 3.89 g cm3. Dynamic light scattering (DLS) In this method, also referred to as photon correlation spectroscopy or quasi-elastic light scattering, the sample is irradiated by laser




light and the light scattered at 90° is monitored as a function of time. Because of the Brownian motion of colloidal particles, the intensity of the scattered light fluctuates and an intensity autocorrelation function may be fitted to the measured fluctuation. The diffusion coefficient and hence the hydrodynamic radii of the particles may then be determined from the fitted autocorrelation function. In the simplest case, the intensity autocorrelation functions are fitted to a single exponential assuming the particles to be spherical. If the particles are assumed to be monodisperse, knowledge of the refractive index of the particles is not necessary. A major limitation of this technique is the need to use low concentrations ( 100 nm were excluded). Figure 12 is a plot of the XRD size vs. the size derived from the BET areas. The data cluster around the ideal line calculated from equation (3) confirms the essential validity of the relationship S = 6/(dq) but provides further evidence that surface area measurements reflect the size of the primary particles.

40

20

XRD

7 3.7* 8 7

0 0.00

20.00

40.00 60.00 Size from BET

80.00

100.00

Figure 13 A comparison of published XRD sizes with the BET-derived sizes of the (mainly) anatase samples. Points marked ◊ are from Devi et al. [42]. The other points are from the publications listed in the caption to Fig. 12.


9


that each size measurement method gives about them are compared. It will be recalled that A and B were derived from the same base titanium dioxide crystals but varied in their surface treatments (silica and alumina for A; alumina and stearate for B). By contrast, although the preparation conditions of the base crystals, B, C and D, were deliberately varied, all three had similar surface treatments. A summary of the line broadening, surface area, sedimentation and dynamic light scattering results for these particles is given in Table 5. The sizes inferred from electron microscopy, X-ray line broadening and the surface area measurements are of the same order and much less than the sizes estimated from sedimentation and light scattering. The line broadening sizes follow the pattern D > A, B > C, and despite different surface treatments, samples A and B derived from the same primary particles gave identical line broadening sizes of 15 nm. The micrographs in Fig. 5 demonstrate that D is indeed larger than B. This pattern of results strongly suggests that line broadening is a good measure of the primary particle size. For the silica–alumina-treated A, the line broadening size is in fair agreement with the 12 nm inferred from nitrogen adsorption measurements. This implies that any aggregation of the primary particles does not greatly affect the adsorption of nitrogen at low vapour pressures (P/P0 < 0.3). However, for the alumina/stearic acid-coated samples B to D, the adsorption-derived sizes (C, 19 < B, 23 < D, 48) followed the same sequence but led to a higher estimated sizes than the line broadening (C, 9 < B, 15 < D, 31). We have argued elsewhere that the lower surface energy of the stearate component of the ‘aluminium stearate’ type coatings may reduce the measured nitrogen adsorption [9]. This reduced adsorption is interpreted as an increased particle size of samples. Such results highlight a possible need to take into account differences of surface energetics when interpreting the adsorption data for powders with different surfaces. Adsorption-derived sizes are plotted as a function of the crystal size in Fig. 14. The X-ray disc sedimentation (XRDS) and dynamic light scattering (DLS) sizes, both of which measure aggregated or agglomerated clusters of particles, were, as expected, consistently larger than the sizes derived from XRD and BET measurements. In general, the largest size measurements were those determined by DLS. The different results show that the sample preparation prior to these two measurements has not broken down these clusters to their component crystals – that is, the results are fully consistent with the

Table 5 Sizes (nm) of four TiO2’s A, B, C and D as measured by X-ray line broadening, Specific Surface area (SSA), X-ray disc sedimentation (XRDS) and Dynamic Light Scattering (DLS)

Sample

XRD crystal size (110 reflection) FWHM method

D derived from specific surface area by N2 adsorption (m2 g1) (SSA)

Particle size X-ray disc sedimentation (XRDS)

A B C D

15 15 9 31

12 23 19 48

91 53 41 160

10

(110–120) (50–70) (70–80) (30)

10 (84–98) 6 (47–59) 6 (35–47) 19 (141–179)

Particle size dynamic light scattering (DLS)

133 124 126 201

BET, DLS & sedim,entation size per nm


210 180 150 120 90 60 30 0

0

5

10

15

20

25

30

XRD line broadening size per nm Figure 14 Adsorption-derived (BET) , sedimentation and DLS plotted as a function of the size derived from X-ray line broadening.

sizes

particle model implied by Fig. 1. As the dispersion conditions used for the sedimentation and DLS measurements were chosen to be representative of the likely shear introduced during the manufacture of sunscreen products, these results imply that in sunscreen products, it is the aggregates that attenuate UV light even though the component crystals are too small to effectively block solar UV. It is instructive to comment in more detail on some aspects of the sedimentation results. First, the derivation of size from sedimentation rates requires the effective cluster density to be known; for the results reported in this study, the particle density was determined by helium pycnometry. Second, a comparison of the XRDC results (the middle (violet broken) line of Fig. 14) with the BET results (the lowest (full blue) line) shows that the sizes derived from sedimentation measurements are ~4 times as large as those derived from adsorption measurements. This implies that although the extended clusters sediment more quickly than individual crystals, the proportion of the particle surface which is accessible to vapour molecules remains high. A final comment concerns particle size assessment by electron microscopy. The apparent simplicity of electron micrographs is strongly persuasive, and high-quality image-analysis software has allowed the ready quantification of particle size by removing the operator bias and tedium associated with traditional sizing methods. Therefore, TEM size is often usefully used as the reference against which other sizes are compared – an example is given in Table 1. However, when assessing the relative sizes of closely sized particles, the difficulty of distinguishing between images of (i) overlapping primary particles and (ii) aggregated and/or agglomerated particles remains. The difficulty is worsened by the fact that the relative number of primary particles, agglomerates and aggregates on the microscope grid depends critically on the dispersion processes used to prepare the TEM sample. Often these processes are part of the ‘craft skills’ of the microscopist and may be difficult to formalize or quantify as part of the ‘standard operating procedures’. If the grid preparation was developed and optimized for a particular type of particle, difficulties may be encountered in detailed comparison of particles (such as B and D of Fig. 5) prepared by very different methods. Therefore, despite their widespread use, TEM sizes are not a panacea – they must be treated with the same caution that is necessary for the other methods discussed in this review.




Conclusions X-ray line broadening, surface area measurement and transmission electron microscopy give similar measures of the size of TiO2 nano particles (the primary particles of Fig. 1). However, variation in surface energetics of these nanoparticles can influence the precise size inferred from an adsorption measurement. Electron micrographs also give a measure of the primary particle size. However, careful and reproducible control of the dispersion of powders on the EM grid and of counting of the images is necessary for quantitatively reproducible measures of particle size.

It is exceptional for light scattering or sedimentation to measure the primary size of commercial nano-powders. Instead, the sizes of aggregates and agglomerates are measured. If the agglomerates are broken down by extended milling or improved dispersion, the proportion of aggregates is increased, but usually these are much larger than the size of the constituent primary particles. Acknowledgements The authors would like to thank Mr B Hirthe of Sachtleben for his help with the TEM analysis (Fig. 7).

References 1. ISO/TS 27687:2008 Nanotechnologies – Terminology and Definitions for Nano-objects – Nanoparticle, Nanofibre and Nanoplate. International Organization of Standards, Geneva, Switzerland (2008). 2. Egerton, T.A. and Tooley, I.R. UV absorption and scattering properties of inorganicbased sunscreens. Intl. J. Cosmet. Sci. 34, 117–122 (2012). 3. Xu, N., Shi, Z., Fan, Y., Dong, J., Shi, J. and Hu, M.Z.-C. Effects of particle size of TiO2 on photocatalytic degradation of methylene blue in aqueous suspensions. Ind. Eng. Chem. Res. 38, 373–379 (1999). 4. Seipenbusch, M., Froeschke, S., Weber, A.P. and Kasper, G. Investigations on the fracturing of nanoparticle agglomerates – first results. Proc. Inst. Mech. Eng. – J. Process Mech. Eng. 216, 219–225 (2002). 5. Egerton, T.A. and Tooley, I.R. Effect of changes in TiO2 dispersion on its measured photocatalytic activity. J. Phys. Chem. B 108, 5066–5072 (2004). 6. Egerton, T.A., Harrison, R.W., Hill, S.E., Mattinson, J.A. and Purnama, H. Effects of particle dispersion on the measurement of semi-conductor photocatalytic activity. J. Photochem. Photobiol. A: Chem. 216, 268–274 (2010). 7. Warren, B.E. X-ray Diffraction. Addison-Wesley, New York, NY (1969), pp. 254–257 or Klug, H.P. and Alexander L.E. X-ray Diffraction Procedures for Polycrystalline and Amorphous Materials. Wiley, New York, NY (1954). 8. Abbas, Z., Holmberg, J.P., Hellstrom, A.K. et al. Synthesis, characterization and particle size distribution of TiO2 colloidal nanoparticles. Colloid Surface A 384, 254–261 (2011). 9. Egerton, T.A. and Tooley, I.R. Characterization of TiO2 nanoparticles surface modified with aluminum stearate. Langmuir 21, 3172–3178 (2005). 10. Anpo, M., Shima, T., Kodama, S. and Kubokawa, Y. Photocatalytic hydrogenation

11.

12.

13.

14.

15.

16.

17.

18.

19.

of CH3CCH with H2O on small-particle TiO2,: size quantization effects and reaction intermediates. J. Phys. Chem. 91, 4305– 4310 (1987). Kumari, L., Li, W.Z., Vannoy, C.H. et al. Zinc oxide micro- and nanoparticles: synthesis, structure and optical properties. Mater. Res. Bull. 45, 190–196 (2010). Pope, D., Smith, W.L., Eastlake, M.J. and Moss, R.L. Structure and activity of supported metal catalysts. 6. Measurement of dispersion in palladium-charcoal catalysts. J. Catal. 22, 72–84 (1971). Tsai, M.-C., Tsai, T.-L., Shieh, D.-B. et al. Detecting HER2 on cancer cells by TiO2 spheres mile scattering. Anal. Chem. 81, 7590–7596 (2009). Hornby, M.R. The Particle-Size Measurement of Titanium Dioxide with respect to Coatings Applications. XVth International Conference in Organic Coatings Science and Technology 11–14th July, 1989, Athens Greece, p. 159. Liang, C.H., Shimizu, Y., Sasaki, T. and Koshizaki, N. Synthesis, characterization, and phase stability of ultrafine TiO2 nanoparticles by pulsed laser ablation in liquid media. J. Mater. Res. 19, 15551–15557 (2004). Allen, T. Particle Size Measurement, 5th edn, Vol. 1, Chap. 2. Chapman and Hall, London (1997). Ismagilov, Z.R., Tsikoza, L.T., Shikina, N.V., Zarytova, V.F., Zinoviev, V.V. and Zagrebelnyi, S.N. Synthesis and stabilization of nano-sized titanium dioxide. Russ. Chem. Rev. 78, 873–885 (2009). Kumar, S.M., Dashpande, P.A., Krishna, M., Krupashankara, M.S. and Madras, G. Photocatalytic activity of microwave plasma-synthesized TiO2 nanopowder. Plasma Chem. Plasma Process. 30, 461–470 (2010). Ridley, M., Hackley, V.A. and Machesky, M.L. Characterization and Surface-Reactivity of Nanocrystalline Anatase in Aqueous Solutions. Langmuir 22, 10972–10982 (2006).


20. Merkus, H.G. Particle Size Measurements. Springer, Netherlands (2008). 21. Kato, H., Suzuki, M., Fujita, K. et al. Reliable size determination of nanoparticles using dynamic light scattering method for in vitro toxicology assessment. Toxicol. In Vitro 23, 927–934 (2009). 22. Allen, T. Particle Size Measurement, 5th edn, Vol. 1, Section 10.4.1. Chapman & Hall, London, UK (1997). 23. French, R.A., Jacobson, A.R., Kim, B., Isley, S.L., Penn, R.L. and Baveye, P.C. Influence of ionic strength, pH, and cation valence on aggregation kinetics of titanium dioxide nanoparticles. Environ. Sci. Technol. 43, 1354–1359 (2009). 24. Fatisson, J., Domingos, R.F., Wilkinson, K.J. and Tufenkji, N. Deposition of TiO2 nanoparticles onto silica measured using a quartz crystal microbalance with dissipation monitoring. Langmuir 25, 6062–6069 (2009). 25. Patri, A., Umbreit, T., Zheng, J. et al. Energy dispersive X-ray analysis of titanium dioxide nanoparticle distribution after intravenous and subcutaneous injection in mice. J. Appl. Toxicol. 29, 662–672 (2009). 26. Stebbing, S.R., Hughes, R.W. and Reynolds, R.A. Sizing, stoichiometry and optical absorbance variations of colloidal cadmium sulphide nanoparticles. Adv. Colloid Interface 147–48, 272–280 (2009). 27. Lee, B.C., Kim, K.T., Cho, J.G. et al. Oxidative stress in juvenile common carp (Cyprinus carpio) exposed to TiO2 nanoparticles. Mol. Cell. Toxicol. 8, 357–366 (2012). 28. Day, R.E. and Egerton, T.A. Surface studies of TiO2 pigment with especial reference to the role of coatings. Colloid Surface 23, 137–155 (1987). 29. Gregg, S.J. and Sing, K.S.W. Adsorption, Surface Area and Porosity, 2nd edn. Academic Press, London (1982). 30. Urwin, D., Ottewill, R.H. and Everett, D.H. (eds) Surface Area Determination. Butterworths, London (1970) p. 397.

11



31. Ishihara Techno Corp Product Data Sheet. Ishihara Sangyo Kaisha, Osaka, Japan. 32. Almquist, C.B. and Biswas, P. Role of synthesis method and particle size of nanostructured TiO2 on its photoactivity. J. Catal. 212, 145–156 (2002). 33. Nosaka, A.Y., Fujiwara, T., Yagi, H., Akutsu, H. and Nosaka, Y. Photoinduced changes of surface and adsorbed water in TiO2 photocatalytic systems as studied by solid state H-1-NMR spectroscopy. Chem. Lett. 3, 420–421 (2002). 34. Ohno, T., Mitsui, T. and Matsumura, M. TiO2-photocatalyzed oxidation of adamantane in solutions containing oxygen or hydrogen peroxide. J. Photochem. Photobiol. A 160, 3–9 (2003). 35. Sun, B. and Smirniotis, P.G. Interaction of anatase and rutile TiO2 particles in aqueous

36.

37.

38.

39.

photooxidation. Catal. Today 88, 49–59 (2003). Dung, N.T., Van Khoa, N. and Herrmann, J.-M. Photocatalytic degradation of reactive dye RED-3BA in aqueous TiO2 suspension under UV-visible light. Int. J. Photoenergy 7, 11–15 (2005). Zhang, Z., Wang, C.-C., Zakaria, R. and Ying, J.Y. Role of particle size in nanocrystalline TiO2-based photocatalysts. J. Phys. Chem. B 102, 10871–10878 (1998). Carneiro, J.T., Savenije, T.J., Moulijn, J.A. and Mul, G. Toward a physically sound structure-activity relationship of TiO2-based photocatalysts. J. Phys. Chem. C 114, 327–332 (2010). Zhou, M.H., Yu, J.G. and Yu, H.G. Effects of urea on the microstructure and photocatalytic activity of bimodal mesoporous titania

microspheres. J. Mol. Catal. A-Chem. 313, 107–113 (2009). 40. Chan, C.K., Porter, J.F., Li, Y.G., Guo, W. and Chan, C.M. Effects of calcination on the microstructures and photocatalytic properties of nanosized titanium dioxide powders prepared by vapor hydrolysis. J. Am. Ceram. Soc. 82, 566–572 (1999). 41. Ohtani, B., Ogawa, Y. and Nishimoto, S. Photocatalytic activity of amorphousanatase mixture of titanium (IV) oxide particles suspended in aqueous solutions. J. Phys. Chem. B 101, 3746–3752 (1997). 42. Devi, L.G., Murthy, B.N. and Kumar, S.G. Photocatalytic activity of TiO2 doped with Zn2+ and V5+ transition metal ions: influence of crystallite size and dopant electronic configuration on photocatalytic activity. Mater. Sci. Eng. B-Adv. 166, 1–6 (2010).

Supporting Information Additional Supporting Information may be found in the online version of this article:

12

Appendix S1. Particle size analysis.


Graphical Abstract The contents of this page will be used as part of the graphical abstract of html only. It will not be published as part of main.

Schematic representation of the aggregation and agglomeration of titanium dioxide nanoparticles.