- Email: [email protected]
Monitoring nanoparticles in the presence of larger particles in liquids using acoustics and electron microscopy
Journal of Colloid and Interface Science 342 (2010) 18–25
Contents lists available at ScienceDirect
Journal of Colloid and Interface Science www.elsevier.com/locate/jcis
Monitoring nanoparticles in the presence of larger particles in liquids using acoustics and electron microscopy A.S. Dukhin a,*, P.J. Goetz a, Xiaohua Fang b, P. Somasundaran b a b
Dispersion Technology Inc., Bedford Hills, NY 10507, USA Columbia University, New York, NY 10027, USA
a r t i c l e
i n f o
Article history: Received 23 April 2009 Accepted 1 July 2009 Available online 4 July 2009 Keywords: Nanotechnology Nanoparticle Polydispersity Acoustic spectroscopy Zinc oxide Particle size distribution Nanoecology Nanotoxicology
a b s t r a c t Monitoring the presence of nanoparticles in dispersions having broad particle size distributions can be a problem for many measurement techniques because large particles or even aggregates of the smaller particles can mask the presence of the sought after nanoparticles. The ability of many existing techniques to detect the nanoparticles when present in broad polydisperse systems is largely unknown, yet it is critical for proper selection of the measuring technique for characterizing a particular nanodispersion. Acoustic spectroscopy is already a known and proven tool for studying nanoparticles in systems with a narrow size distribution. The purpose of this paper is to evaluate the sensitivity of acoustic spectroscopy for determining the nanoparticle content of very polydisperse systems. We used eight different ZnO powders from different manufacturers to prepare 5 wt.% dispersions, each dispersed in water. The stability of each dispersion was optimized by pH adjustment and addition of sodium hexametaphosphate as determined by maximizing the measured f-potential. According to the acoustic measurement, the median size of these different ZnO dispersions varied from 200 nm to 700 nm. Independent TEM photographs in general confirmed the size variation between the samples. Independent DLS measurements failed to provide particle size data correlating with TEM. The acoustic measurements further showed that each dispersion contained a different relative content in the nanoparticle fraction. The precision with which the nanoparticle fraction could be determined was better than 2% of the total solid loading for all samples. In order to verify consistency of this measurement we performed a mixing study by adding dispersion with the largest nanoparticle content to the dispersion with the smallest nanoparticle content, in small increments. This test confirms that the acoustic sensitivity threshold is about 2% of nanoparticles in the broad polydisperse dispersions of dense metal oxide particles. Ó 2009 Published by Elsevier Inc.
1. Introduction Nanoecology and nanotoxicology are rapidly growing scientific disciplines for studying the impact of nanotechnology on human health and the environment [1–6]. A growing number of consumer products, cosmetic in particular, that contain engineered nanoparticles create an immediate need for careful studies. This is reflected in the law recently adopted by the European Parliament, on March 24, 2009, and supported by the European Cosmetics Association. One problem that must be addressed is the selection of an accurate method for monitoring the nanoparticle content of various products, many being dispersions containing mixtures of not only nanosized particles, but also particles of much larger size. While there are known methods for measuring systems containing only nanoparticles, there are no well established techniques for moni-
* Corresponding author. Fax: +1 914 241 4842. E-mail address: [email protected] (A.S. Dukhin). 0021-9797/$ - see front matter Ó 2009 Published by Elsevier Inc. doi:10.1016/j.jcis.2009.07.001
toring nanoparticles mixed with larger ones. Development of such methods is imperative for reliable investigation of nanoparticles in natural and engineered systems, such as cosmetic products. The first logical step in this regard is an investigation with existing commercially available instruments preferably by an independent institution, which has in possession all the instruments that are relevant for this task. In the absence of such institutions for undertaking a project, instrumentation companies can do such a study with their instruments in collaboration with academic institutions. To make the study reliable, it is necessary that it be conducted with the same set of nanoparticles. This approach also has the advantage of the test being done by staff trained for the particular instrument. Toward this purpose, as a first step researchers at the Langmuir Center for Colloids and Interfaces at Columbia University and Dispersion Technology Inc. have undertaken this study to compare different samples with acoustic spectroscopy and TEM, since these techniques have been reported to be well suited for characterizing nanoparticles with rather narrow size distributions [7–15]. The goal of this study is to determine
19
A.S. Dukhin et al. / Journal of Colloid and Interface Science 342 (2010) 18–25
the sensitivity of the acoustic spectroscopy to nanoparticles in broad particle size distributions with large amounts of coarser particles. Acoustic spectroscopy might have significant advantages over light scattering methods in studying samples with broad polydisperse particle size distributions. Weight basis is innate for this technique because attenuation of ultrasound due to viscous dissipation by submicrometer particles is proportional to the particles weight. This means that raw data for calculating particle size depends on only the third power of the particle size. In contrast, scattered light depends on much higher power of the particles size, as large as the 6th. This makes scattering by large particles dominating raw data and overshadowing contributions of the smaller particles in the nanosize range. Acoustic attenuation’s weaker dependence on the particles size makes contribution of the small (nano) and larger particles more even and the method potentially more sensitive to the nanoparticle content even in the very broad size distributions.
2. Materials and methods ZnO powder was selected for this work since this substance, widely used in sun-screen products, has played an important role in recent nanoecology discussions. In order to make samples easily available for all future studies, they were purchased from Fisher Scientific Inc., which offers powders from a variety of sources. The eight different powders produced by various manufacturers are listed below and in Table 1. Zinc oxide, reagent ACS, and zinc oxide, 99.5+% manufactured by Acros Organics. Z50–500 and Z52-500 USP powder packaged by Fisher Scientific. S80249 by Fisher Scientific. Zinc oxide ACS reagent grade by MO Biomedicals, LLC. Zinc oxide Polystormor by Mallinckrodt Chemicals. Zinc oxide Nanopowder by Americal Elements.
For adjusting the pH of ZnO dispersions we use potassium hydroxide 1.000 N by VWR International. For enhancing the electric surface charge of ZnO particles we use 10 wt.% solution of sodium hexametaphosphate by Fluka dissolved in distilled water. For calibrating electroacoustic probe we use silica Ludox TM-50 produced by Grace–Davison and purchased from Sigma–Aldrich. The original dispersion contains 50% wt silica. We dilute it down to 10 wt.% using 0.01 M KCl solution. Silica has f-potential 38 mV in this solution. Calibration of conductivity probe has been done using distilled water and two KCl solutions with concentrations of 0.01 and 0.1 mol/L.
2.1. Experimental technique We use three different methods: one is based on ultrasound, the second one is electron microscopy, the third is dynamic light scattering. The ultrasound-based instrument is manufactured by Dispersion Technology Inc., Model DT-1201. This instrument is designed for characterizing concentrated dispersions and emulsions with the volume fraction above 1% with no dilution. It has five different sensors. Details are available on the web site www. dispersion.com. Particle size characterization is based on principles of acoustic spectroscopy, as described in ISO Standard ISO 20998-1 [7] and in the literature [8]. Acoustic sensor measures attenuation of ultrasound for a set of frequencies from 1 MHz to 100 MHz. This attenuation frequency spectrum is raw data. It is treated then with a model designed to select the best particle size distribution that generates the theoretical attenuation spectra that fits the experimental data with the least error. Experimental verification of this method with a wide multitude of different materials is discussed elsewhere [8–15]. Electron microscopy photos were made with a JEOL transmission electron microscope (TEM). The particles were first dispersed in pH 10 sodium hydroxide solutions. After equilibrating and mixing the suspension overnight, we prepared the TEM samples by loading a droplet of the dispersion onto a 300 mesh gilder copper TEM grid (Electron Microscopy Sciences, Inc.). After the water in the droplet had evaporated, particles precipitated to the grid. This grid were then put into the TEM sample holder and inserted into the TEM instrument. The accelerating voltage applied was 100 kV. The magnifications for all the samples were kept similar for comparison. We used two different dynamic light scattering instruments. Instructions for both of them suggest the use of filtration for removing large particles. Strong scattering by even a few large particles could cause artifacts with this technique. Filtration would not acceptable for the broad particle size distributions of the current study with large quantities of large particles. Nevertheless, we have performed some trial measurements that failed. Results contradicted TEM photos even qualitatively. We do not discuss them here.
2.2. Sample preparation All samples are prepared at 5 wt.% by adding 5 g of powder to 95 g of distilled water and sonicating to make it homogeneous. Values of pH become naturally from 6 to 7. This sample then was poured into the DT-1201 measuring chamber with a magnetic stirrer for preventing sample sedimentation. All the measurements then are performed with this sample without taking it out of the measuring chamber.
Table 1 Median particle size and percentage of the nanoparticles (<100 nm) in various ZnO samples at 5 wt.% stabilized at pH 10 with hexametaphosphate. Powder name, manufacturer
Median size, micrometers
Cumulative % of nanoparticles according to acoustics, <100 nm
Zinc oxide, 99.5+% by Acros Organics (5 loads, 25 measurements) Zinc oxide, reagent ACS by Acros Organics (1 load, 10 measurements) Z50–500 USP powder packaged by Fisher Scientific (1 load, 10 measurements) Z52-500 USP powder packaged by Fisher Scientific (4 loads, 28 measurements) S80249 by Fisher Scientific (1 load, 5 measurements) Zinc oxide ACS reagent grade by MO Biomedicals, LLC (1 load, 5 measurements) Zinc oxide Polystormor by Mallinckrodt Chemicals (3 loads, 46 measurements) Zinc oxide Nanopowder by American Elements (3 loads, 15 measurements)
0.273 ± 0.01 0.430 ± 0.02 0.561 ± 0.017 0.660 ± 0.037 0.398 ± 0.001 0.349 ± 0.017 0.223 ± 0.009 0.631 ± 0.1
11.0 ± 1.4 7.0 ± 0.5 2.7 ± 0.39 1.9 ± 0.4 6.1 ± 0.2 8.2 ± 2.1 19.6 ± 1.8 4.7 ± 2.5
20
A.S. Dukhin et al. / Journal of Colloid and Interface Science 342 (2010) 18–25
The main purpose of sample preparation is to break up flocs and disperse the particles. This is achieved by increasing the surface charge of the particles and f-potential measurements allow optimization of this procedure. The software option for running multiple measurements continuously is used for f-potential measurements. The initial value of f-potential is about +10 mV for all samples. As it is too low for preventing flocculation, as a first step of dispersion pH is adjusted to 10 by adding small amounts of 1 N KOH. The value of f-potential becomes negative, between 10 and 15 mV, which is not enough to keep particles from aggregating. Hence as a second step small increments of 10% hexametaphosphate (HMPH) solution were added and the sample was sonicated after each addition. Simultaneous continuous measurement of f-potential displays a gradual increase in its absolute value. Under HMPH saturation conditions, the f-potential eventually reaches a sufficiently significant negative value. This saturation value differs from one powder to another, but it is more than 30 mV for all of them but not exceeding 40 mV. Monitoring of the increase in f-potential is an important part of the sample preparation procedure since HMPH overdosing can cause the particles to flocculate again. This preparation usually takes about 1 h, although some particles exhibited much longer equilibration time. This might show up during the particle size measurement as drifting of the measured attenuation spectra with time. In such cases the sample was mixed for a prolonged equilibration time.
Particle size distributions were measured for individual samples and for mixtures of two different samples at different ratios.
particle size and average absolute deviation of this parameter for each ZnO sample. The tested powders were observed to have very different sizes in the range from 220 to 660 nm. One powder (Nanopowder by American Elements) was expected to be the smallest one since it was the only powder that had the term ‘‘nano” in its name. This sample was prepared three times; three different loads and multiple sonications were tried to disperse it without success. In this case the size was observed to vary from one preparation to another, which is reflected in a relatively large variation coefficient. In any case, the amount of nanoparticles in this powder dispersed in water was still very small. The broadness of all the particle size distribution is illustrated on Figs. 1 and 2. Interestingly, all distributions are broad with sizes in the nanoscale, submicrometer, and micrometer scales. Table 1 presents also the averaged content of nanoparticles for each ZnO. Precision of this measurement is defined as ratio of the ‘‘absolute average deviations” to the ‘‘average content.” It is clear that this precision varies significantly from sample to sample. The worst number is 2.5% for Nanopowder by American. However, this sample is not stable. It is strongly aggregated, which reduces the reproducibility of PSD from one sample preparation to another. Precision of practically all other samples is better than 2%. We claim this number as characteristic of the method. The powder with the largest size is Z52-500. This powder has the least amount of nanoparticles, about 2%, according to Table 1. The powder with the most amount of nanoparticles is Mallinckrodt. According to Table 1, it contains 20% of nanoparticles. We used these two powders for making artificial mixtures for consistency testing.
3.1. Individual samples
3.2. Comparison of acoustics with electron microscopy data
Multiple measurements of particle size distribution (PSD) were performed with every sample. Some of the samples were prepared several times. The term ‘‘load” is used in Table 1 for indicating different preparations of the same powder. Each load has been measured at least five times. All these data have been statistically treated. Table 1 presents an arithmetic average of the median
Comparison of the particle size distributions produced by acoustics and electron microscopy is complicated by the difference in their innate basis. Acoustic yields PSD on a weight basis, whereas electron microscopy innate basis is number distribution. Differences between weight and number distribution can be tremendous for a broad PSD. Fig. 7 shows as an example the size
3. Results and discussion
Zinc oxide, reagent ACS by Acros Organics Zinc oxide, 99.5+% Acros Organics Z50-500 USP Z52-500 USP S80249 by Fisher Scientific Zinc oxide 99.99% by Alfa Aesar Zinc oxide ACS MO Biomedicals, LLC PolystormorTM by Mallinckrodt Chemicals Nanopowder America Elements
1.2
PSD,weight basis
1.0
0.8
0.6
0.4
0.2
0.0 -2
10
-1
10
0
10
1
10
Diameter [um] Fig. 1. Differential particle size distributions calculated from acoustic attenuation spectra assuming lognormal distribution shape. Median size and standard deviation are arithmetic averages of the multiple measurements.
21
A.S. Dukhin et al. / Journal of Colloid and Interface Science 342 (2010) 18–25
1.0 0.9
Cumulative,weight basis
0.8 0.7 0.6 0.5 Zinc oxide, ACS Acros Organics Zinc oxide, 99.5+% Acros Organics Z50-500 USP Z52-500 USP S80249 by Fisher Scientific Zinc oxide 99.99% by Alfa Aesar Zinc oxide ACS MO Biomedicals, LLC Polystormor, Mallinckrodt Chemicals Nanopowder America Elements
0.4 0.3 0.2 0.1 0.0 -2
-1
10
10
0
1
10
10
Diameter [um] Fig. 2. Cumulative particle size distributions calculated from acoustic attenuation spectra assuming lognormal distribution shape. Median size and standard deviation are arithmetic averages of the multiple measurements.
Mallinckrodt
Probabilities of Particles in the Bin
1.0 Mallingckrodt 0.8 0.6 0.4 0.2 0.0 0
100 200 300 400 500 600 Size (nm)
Z52-500
Probabilities of Particles in the Bin
1.0 Z52-500
0.8 0.6 0.4 0.2 0.0 0
100
200 300 400 Size (nm)
500
Fig. 3. TEM photographs and size histogram of the Malincrodt and Z52-500 ZnO dispersion. Bar is 500 nm.
600
22
A.S. Dukhin et al. / Journal of Colloid and Interface Science 342 (2010) 18–25
distribution for Z52-500 measured using acoustics and recalculated for both weight and number basis. Weight basis indicates the weight of the particles with a certain size. For instance, Z52500 contains only 2% of ZnO weight in the particles with sizes under 100 nm. Number basis presents number of particles with certain size. For Z52-500 the number of nanoparticles with sizes under 100 nm is about 75%. The same cumulative number is more than 35 times different if expressed either on a number or on a weight basis. TEM photos provide information for calculating PSD on a number basis. In order to compare these statistical data with acoustical PSD it should be converted to a weight basis. Unfortunately, this approach is not realistic due to the complexity of generating complete and representative particle size distribution using TEM. There is an International Standard describing this procedure, ISO 13322– 1 [16]. In particular, this standard determines the number of treated particles required to reach 95% of ‘‘correct” PSD. In the case of broad PSD it is several thousands. For comparison, each photo shown in Figs. 3–6, contains only up to 60 particles. Statistical treatment of hundreds photographs for generating a single reliable PSD is not a realistic approach for resolving problems of monitoring nanoparticle content. The situation would become even more hopeless if we take into account the possibility of various artifacts associated with interpretation of complex particles and aggregates shapes on the photo-
graphs. It is possible that TEM sample preparation aggregated initially separate nanoparticles in aggregates. That is why calculation of number of the nanoparticles on each TEM photo includes particles within aggregates. This might lead to overestimating nanoparticle content. This short analysis brings us to the conclusion that statistical histograms presented in Figs. 3–6 and calculated using just a single photo for each sample cannot be directly quantitatively compared with weight based produced by acoustics. We can use these photos only for some qualitative comparison of these two methods. These TEM photos yield some information that correlates with results of acoustic measurement. Here are some conclusions derived from comparison of the TEM photos that agree with data given in Table 1. First, sample Mallincrodt is clearly the smallest. There are two samples by Acros Organic. Sample ACS has much larger sizes than sample 99.5%, which agrees with Table 1. Samples MO Biomedicals and S80249 have similar size distributions, which agrees with Table 1 data. Sample Nanopowder by American is strongly aggregated. It looks like aggregates consist of nanoparticles. We could not find any contradiction between TEM photos and PSD results presented in Table 1. This comparison of the TEM photos and Table 1 results allows us to conclude that TEM confirms acoustics particle sizing, at least qualitatively.
Nanopowder by American
Probabilities of Particles in the Bin
1.0 AME
0.8 0.6 0.4 0.2 0.0 0
100
200 300 400 Size (nm)
500
600
Z50-500
Probabilities of Particles in the Bin
1.0 Z50-500
0.8 0.6 0.4 0.2 0.0
0
100
200 300 400 Size (nm)
500
600
Fig. 4. TEM photographs and size histogram of the American Nanopowder and Z50–500 ZnO dispersion. Bar is 500 nm.
23
A.S. Dukhin et al. / Journal of Colloid and Interface Science 342 (2010) 18–25
Acros Organic ACS
Probabilities of Particles in the Bin
1.0 acrosacs
0.8 0.6 0.4 0.2 0.0 0
100 200 300 400 500 Size (nm)
600
Acros Organic 99.5%
Probabilities of Particles in the Bin
1.0 Acros995
0.8 0.6 0.4 0.2 0.0 0
100
200
300 400 Size (nm)
500
600
Fig. 5. TEM photographs of two different Acros ZnO dispersions. Bar is 500 nm.
3.3. Mixture There is possibility of verifying acoustic particle sizing consistency by mixing two samples together. Also, this mixing test would yield some information on the sensitivity of the method to the nanoparticle content. The most informative would be additions of the small increments of the sample with the largest nanoparticle content to the sample with the least nanoparticle content. There are two ways one can estimate nanoparticle content in such mixtures. First, one can use known nanoparticle content in the individual samples and also known ratio of the individual samples in a particular mixture. For instance, the Mallinckrodt sample contains 20% of the nanoparticles, whereas Z52-500 sample contains only 2% of nanoparticles. Content of the Mallinckrodt sample in each mixture is also known. This is sufficient for calculating the expected nanoparticle content in each mixture. On the other side, measurement of the attenuation spectra for each mixture yields raw data for calculating particle size distribution. This is the same procedure as for individual samples. It provides nanoparticle content in mixtures similarly to individual samples. If everything is consistent, both methods should yield the same nanoparticle content in a particular mixture.
We have performed this consistency verification test using ZnO sample Z52-500 as a base. This sample contains the least amount of nanoparticles according to the acoustics, only 2%. Sample weight of the initial sample was 100 g, which is sufficient for filling the DT-1201 sample chamber. Additions of the Mallinckrodt sample, with the largest nanoparticle content, to Z52-500 were made in the increments of 2 g and then 4 g. The final mixture contained 26 g of Mallincrodt sample and 100 g of Z52-500 sample. We can calculate nanoparticle content in each mixture using the following algorithm. Initial 100 g of Z52-500 samples contains 5 g of ZnO, which in turns has only 0.1 g (2%) of nanoparticles. Let us assume that we have added M grams of the Mallincrodt sample. This addition contains 5 M/100 g on ZnO because the sample was prepared at 5 wt.% of ZnO. Nanoparticle content would be (5 M/100)/5 because only 20% of Malincrodt ZnO is nano. Total weight of ZnO in such mixture would be (5 g + M/20 g). Total weight of nanoparticles, assuming known individual samples PSD, would be (0.1 g + M/100 g). Percentage of nanoparticles in the mixture of 100 g Z52-500 samples and M grams of Mallincrodt samples P equals
P¼
0:1 þ M=100 100: 5 þ M=20
24
A.S. Dukhin et al. / Journal of Colloid and Interface Science 342 (2010) 18–25
MO Biomedical
Probabilities of Particles in the Bin
1.0 Mo-bio
0.8 0.6 0.4 0.2 0.0 0
100
200
300 400 Size (nm)
500
600
S80249
Probabilities of Particles in the Bin
1.0 S80249
0.8 0.6 0.4 0.2 0.0 0
100
200 300 400 Size (nm)
500
600
Fig. 6. TEM photographs of the MO Biomedical and S80249 ZnO dispersion. Bar is 500 nm.
number basis
1.1
Alternatively, we can estimate the percentage content of nanoparticles in each mixture from acoustic attenuation spectra measured for the mixture. These two numbers should be the same. This is the essence of the performed mixing test. Fig. 7 presents graphically results of the test. Values of the X-axis equals the value of percentage P calculated for each mixture. Actual results of the acoustic sizing measurements are represented with circles in Fig. 8. Solid line represents idea case when measured content equals exactly the expected nanoparticle content P. It is seen that deviation of the acoustically measured nanocontent from the expected values is on a scale of 1%. It is significantly less than 2% precision level, which we determined as precision threshold based on study of individual samples. We can conclude that acoustic spectroscopy yields nanopatricle content in these mixtures that agrees well in trend with expected values.
weight basis
1.0 0.9
particle size distribution
0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1
4. Conclusions
0.0 -2
10
-1
10
0
10 diameter [microns]
1
10
Fig. 7. Particle size distribution of the Z52-500 sample on a weight and number basis.
Particle size characterization using acoustic spectroscopy allows monitoring of nanoparticle content in the dispersions with broad PSD with precision of at least 2%. It can be achieved in concentrated dispersions with no dilution. This conclusion is confirmed by multiple measurements of individual samples of eight
A.S. Dukhin et al. / Journal of Colloid and Interface Science 342 (2010) 18–25
25
different ZnO dispersions and their mixtures. Electron microscopy confirms qualitatively results of PSD characterization using acoustic spectroscopy.
7
6
% nano-particles, measured
References 5
4
3
2
1 1
2
3 4 % nano-particles, estimated
5
6
Fig. 8. Percentage of nanoparticles in Z52-500 sample after incremental additions of Mallinckrodt Chemicals sample. X-axis is a percentage calculated from the known amount of the added Mallinckrodt sample, assuming that it contains 20% on nanoparticles, according to the Table 1. Y-axis is a percentage calculated from the attenuation spectra, which is measured for the mixture.
[1] R. Behra, H. Krug, Nat. Nanotechnol. 3 (2008) 253–254. [2] Groups encourage nanomaterial stewardship, Chemical & Engineering News, July 21, 2008, p. 34. [3] G. Miller, Nanomaterials, sunscreens and cosmetics: small ingredients–big risks, Friends of the Earth, 2006. [4] The appropriateness of existing methodologies to assess the potential risks associated with engineered and adventitious products of nanotechnology. European Commission, Health and Consumer protection Directorate-General, SCENIHR, 2006. [5] Nanotechnology, A Report of the US Food and Drug Administration Nanotechnology Task Force, July 25, 2007. [6] P. Somasundaran, S.C. Mehta, L. Rhein, S. Chakraborty, MRS Bull. 32 (2007) 779–786. [7] Characterization of particles by acoustic methods, ISO 20998, ISO/TC24/SC 4, 2005. [8] A.S. Dukhin, P.J. Goetz, Ultrasound for Characterizing Colloids. Particle Sizing, Zeta Potential, Elsevier, Rheology, 2002. [9] N. Bell, J. Cesarano, J.A. Voight, S.J. Jockwood, D.B. Dimos, J. Mater. Res. 19 (5) (2004) 1333–1340. [10] N. Bell, M.A. Rodriguez, J. Nanosci. Nanotechnol. 4 (3) (2004) 283–290. [11] Y.P. Sun, X. Li, J. Cao, W. Zhang, H.P. Wang, Adv. Colloid Interface Sci. 120 (2006) 47–56. [12] S. Gaydardzhiev, P. Ay, J. Dispers. Sci. Technol. 27 (2006) 413–417. [13] P. Wilhelm, D. Stephan, J. Colloid Interface Sci. 293 (2006) 88–92. [14] Y.P. Sun, X.Q. Li, W.X. Zhang, H.P. Wang, Colloids Surf. 308 (2007) 60–66. [15] F. Stenger, S. Mende, J. Schwedes, W. Peukert, Chem. Eng. Sci. 60 (2005) 4557– 4565. [16] Particle size analysis–image analysis methods, Part 1. Static image analysis methods, ISO 13322-1, 2004.
Contents lists available at ScienceDirect
Journal of Colloid and Interface Science www.elsevier.com/locate/jcis
Monitoring nanoparticles in the presence of larger particles in liquids using acoustics and electron microscopy A.S. Dukhin a,*, P.J. Goetz a, Xiaohua Fang b, P. Somasundaran b a b
Dispersion Technology Inc., Bedford Hills, NY 10507, USA Columbia University, New York, NY 10027, USA
a r t i c l e
i n f o
Article history: Received 23 April 2009 Accepted 1 July 2009 Available online 4 July 2009 Keywords: Nanotechnology Nanoparticle Polydispersity Acoustic spectroscopy Zinc oxide Particle size distribution Nanoecology Nanotoxicology
a b s t r a c t Monitoring the presence of nanoparticles in dispersions having broad particle size distributions can be a problem for many measurement techniques because large particles or even aggregates of the smaller particles can mask the presence of the sought after nanoparticles. The ability of many existing techniques to detect the nanoparticles when present in broad polydisperse systems is largely unknown, yet it is critical for proper selection of the measuring technique for characterizing a particular nanodispersion. Acoustic spectroscopy is already a known and proven tool for studying nanoparticles in systems with a narrow size distribution. The purpose of this paper is to evaluate the sensitivity of acoustic spectroscopy for determining the nanoparticle content of very polydisperse systems. We used eight different ZnO powders from different manufacturers to prepare 5 wt.% dispersions, each dispersed in water. The stability of each dispersion was optimized by pH adjustment and addition of sodium hexametaphosphate as determined by maximizing the measured f-potential. According to the acoustic measurement, the median size of these different ZnO dispersions varied from 200 nm to 700 nm. Independent TEM photographs in general confirmed the size variation between the samples. Independent DLS measurements failed to provide particle size data correlating with TEM. The acoustic measurements further showed that each dispersion contained a different relative content in the nanoparticle fraction. The precision with which the nanoparticle fraction could be determined was better than 2% of the total solid loading for all samples. In order to verify consistency of this measurement we performed a mixing study by adding dispersion with the largest nanoparticle content to the dispersion with the smallest nanoparticle content, in small increments. This test confirms that the acoustic sensitivity threshold is about 2% of nanoparticles in the broad polydisperse dispersions of dense metal oxide particles. Ó 2009 Published by Elsevier Inc.
1. Introduction Nanoecology and nanotoxicology are rapidly growing scientific disciplines for studying the impact of nanotechnology on human health and the environment [1–6]. A growing number of consumer products, cosmetic in particular, that contain engineered nanoparticles create an immediate need for careful studies. This is reflected in the law recently adopted by the European Parliament, on March 24, 2009, and supported by the European Cosmetics Association. One problem that must be addressed is the selection of an accurate method for monitoring the nanoparticle content of various products, many being dispersions containing mixtures of not only nanosized particles, but also particles of much larger size. While there are known methods for measuring systems containing only nanoparticles, there are no well established techniques for moni-
* Corresponding author. Fax: +1 914 241 4842. E-mail address: [email protected] (A.S. Dukhin). 0021-9797/$ - see front matter Ó 2009 Published by Elsevier Inc. doi:10.1016/j.jcis.2009.07.001
toring nanoparticles mixed with larger ones. Development of such methods is imperative for reliable investigation of nanoparticles in natural and engineered systems, such as cosmetic products. The first logical step in this regard is an investigation with existing commercially available instruments preferably by an independent institution, which has in possession all the instruments that are relevant for this task. In the absence of such institutions for undertaking a project, instrumentation companies can do such a study with their instruments in collaboration with academic institutions. To make the study reliable, it is necessary that it be conducted with the same set of nanoparticles. This approach also has the advantage of the test being done by staff trained for the particular instrument. Toward this purpose, as a first step researchers at the Langmuir Center for Colloids and Interfaces at Columbia University and Dispersion Technology Inc. have undertaken this study to compare different samples with acoustic spectroscopy and TEM, since these techniques have been reported to be well suited for characterizing nanoparticles with rather narrow size distributions [7–15]. The goal of this study is to determine
19
A.S. Dukhin et al. / Journal of Colloid and Interface Science 342 (2010) 18–25
the sensitivity of the acoustic spectroscopy to nanoparticles in broad particle size distributions with large amounts of coarser particles. Acoustic spectroscopy might have significant advantages over light scattering methods in studying samples with broad polydisperse particle size distributions. Weight basis is innate for this technique because attenuation of ultrasound due to viscous dissipation by submicrometer particles is proportional to the particles weight. This means that raw data for calculating particle size depends on only the third power of the particle size. In contrast, scattered light depends on much higher power of the particles size, as large as the 6th. This makes scattering by large particles dominating raw data and overshadowing contributions of the smaller particles in the nanosize range. Acoustic attenuation’s weaker dependence on the particles size makes contribution of the small (nano) and larger particles more even and the method potentially more sensitive to the nanoparticle content even in the very broad size distributions.
2. Materials and methods ZnO powder was selected for this work since this substance, widely used in sun-screen products, has played an important role in recent nanoecology discussions. In order to make samples easily available for all future studies, they were purchased from Fisher Scientific Inc., which offers powders from a variety of sources. The eight different powders produced by various manufacturers are listed below and in Table 1. Zinc oxide, reagent ACS, and zinc oxide, 99.5+% manufactured by Acros Organics. Z50–500 and Z52-500 USP powder packaged by Fisher Scientific. S80249 by Fisher Scientific. Zinc oxide ACS reagent grade by MO Biomedicals, LLC. Zinc oxide Polystormor by Mallinckrodt Chemicals. Zinc oxide Nanopowder by Americal Elements.
For adjusting the pH of ZnO dispersions we use potassium hydroxide 1.000 N by VWR International. For enhancing the electric surface charge of ZnO particles we use 10 wt.% solution of sodium hexametaphosphate by Fluka dissolved in distilled water. For calibrating electroacoustic probe we use silica Ludox TM-50 produced by Grace–Davison and purchased from Sigma–Aldrich. The original dispersion contains 50% wt silica. We dilute it down to 10 wt.% using 0.01 M KCl solution. Silica has f-potential 38 mV in this solution. Calibration of conductivity probe has been done using distilled water and two KCl solutions with concentrations of 0.01 and 0.1 mol/L.
2.1. Experimental technique We use three different methods: one is based on ultrasound, the second one is electron microscopy, the third is dynamic light scattering. The ultrasound-based instrument is manufactured by Dispersion Technology Inc., Model DT-1201. This instrument is designed for characterizing concentrated dispersions and emulsions with the volume fraction above 1% with no dilution. It has five different sensors. Details are available on the web site www. dispersion.com. Particle size characterization is based on principles of acoustic spectroscopy, as described in ISO Standard ISO 20998-1 [7] and in the literature [8]. Acoustic sensor measures attenuation of ultrasound for a set of frequencies from 1 MHz to 100 MHz. This attenuation frequency spectrum is raw data. It is treated then with a model designed to select the best particle size distribution that generates the theoretical attenuation spectra that fits the experimental data with the least error. Experimental verification of this method with a wide multitude of different materials is discussed elsewhere [8–15]. Electron microscopy photos were made with a JEOL transmission electron microscope (TEM). The particles were first dispersed in pH 10 sodium hydroxide solutions. After equilibrating and mixing the suspension overnight, we prepared the TEM samples by loading a droplet of the dispersion onto a 300 mesh gilder copper TEM grid (Electron Microscopy Sciences, Inc.). After the water in the droplet had evaporated, particles precipitated to the grid. This grid were then put into the TEM sample holder and inserted into the TEM instrument. The accelerating voltage applied was 100 kV. The magnifications for all the samples were kept similar for comparison. We used two different dynamic light scattering instruments. Instructions for both of them suggest the use of filtration for removing large particles. Strong scattering by even a few large particles could cause artifacts with this technique. Filtration would not acceptable for the broad particle size distributions of the current study with large quantities of large particles. Nevertheless, we have performed some trial measurements that failed. Results contradicted TEM photos even qualitatively. We do not discuss them here.
2.2. Sample preparation All samples are prepared at 5 wt.% by adding 5 g of powder to 95 g of distilled water and sonicating to make it homogeneous. Values of pH become naturally from 6 to 7. This sample then was poured into the DT-1201 measuring chamber with a magnetic stirrer for preventing sample sedimentation. All the measurements then are performed with this sample without taking it out of the measuring chamber.
Table 1 Median particle size and percentage of the nanoparticles (<100 nm) in various ZnO samples at 5 wt.% stabilized at pH 10 with hexametaphosphate. Powder name, manufacturer
Median size, micrometers
Cumulative % of nanoparticles according to acoustics, <100 nm
Zinc oxide, 99.5+% by Acros Organics (5 loads, 25 measurements) Zinc oxide, reagent ACS by Acros Organics (1 load, 10 measurements) Z50–500 USP powder packaged by Fisher Scientific (1 load, 10 measurements) Z52-500 USP powder packaged by Fisher Scientific (4 loads, 28 measurements) S80249 by Fisher Scientific (1 load, 5 measurements) Zinc oxide ACS reagent grade by MO Biomedicals, LLC (1 load, 5 measurements) Zinc oxide Polystormor by Mallinckrodt Chemicals (3 loads, 46 measurements) Zinc oxide Nanopowder by American Elements (3 loads, 15 measurements)
0.273 ± 0.01 0.430 ± 0.02 0.561 ± 0.017 0.660 ± 0.037 0.398 ± 0.001 0.349 ± 0.017 0.223 ± 0.009 0.631 ± 0.1
11.0 ± 1.4 7.0 ± 0.5 2.7 ± 0.39 1.9 ± 0.4 6.1 ± 0.2 8.2 ± 2.1 19.6 ± 1.8 4.7 ± 2.5
20
A.S. Dukhin et al. / Journal of Colloid and Interface Science 342 (2010) 18–25
The main purpose of sample preparation is to break up flocs and disperse the particles. This is achieved by increasing the surface charge of the particles and f-potential measurements allow optimization of this procedure. The software option for running multiple measurements continuously is used for f-potential measurements. The initial value of f-potential is about +10 mV for all samples. As it is too low for preventing flocculation, as a first step of dispersion pH is adjusted to 10 by adding small amounts of 1 N KOH. The value of f-potential becomes negative, between 10 and 15 mV, which is not enough to keep particles from aggregating. Hence as a second step small increments of 10% hexametaphosphate (HMPH) solution were added and the sample was sonicated after each addition. Simultaneous continuous measurement of f-potential displays a gradual increase in its absolute value. Under HMPH saturation conditions, the f-potential eventually reaches a sufficiently significant negative value. This saturation value differs from one powder to another, but it is more than 30 mV for all of them but not exceeding 40 mV. Monitoring of the increase in f-potential is an important part of the sample preparation procedure since HMPH overdosing can cause the particles to flocculate again. This preparation usually takes about 1 h, although some particles exhibited much longer equilibration time. This might show up during the particle size measurement as drifting of the measured attenuation spectra with time. In such cases the sample was mixed for a prolonged equilibration time.
Particle size distributions were measured for individual samples and for mixtures of two different samples at different ratios.
particle size and average absolute deviation of this parameter for each ZnO sample. The tested powders were observed to have very different sizes in the range from 220 to 660 nm. One powder (Nanopowder by American Elements) was expected to be the smallest one since it was the only powder that had the term ‘‘nano” in its name. This sample was prepared three times; three different loads and multiple sonications were tried to disperse it without success. In this case the size was observed to vary from one preparation to another, which is reflected in a relatively large variation coefficient. In any case, the amount of nanoparticles in this powder dispersed in water was still very small. The broadness of all the particle size distribution is illustrated on Figs. 1 and 2. Interestingly, all distributions are broad with sizes in the nanoscale, submicrometer, and micrometer scales. Table 1 presents also the averaged content of nanoparticles for each ZnO. Precision of this measurement is defined as ratio of the ‘‘absolute average deviations” to the ‘‘average content.” It is clear that this precision varies significantly from sample to sample. The worst number is 2.5% for Nanopowder by American. However, this sample is not stable. It is strongly aggregated, which reduces the reproducibility of PSD from one sample preparation to another. Precision of practically all other samples is better than 2%. We claim this number as characteristic of the method. The powder with the largest size is Z52-500. This powder has the least amount of nanoparticles, about 2%, according to Table 1. The powder with the most amount of nanoparticles is Mallinckrodt. According to Table 1, it contains 20% of nanoparticles. We used these two powders for making artificial mixtures for consistency testing.
3.1. Individual samples
3.2. Comparison of acoustics with electron microscopy data
Multiple measurements of particle size distribution (PSD) were performed with every sample. Some of the samples were prepared several times. The term ‘‘load” is used in Table 1 for indicating different preparations of the same powder. Each load has been measured at least five times. All these data have been statistically treated. Table 1 presents an arithmetic average of the median
Comparison of the particle size distributions produced by acoustics and electron microscopy is complicated by the difference in their innate basis. Acoustic yields PSD on a weight basis, whereas electron microscopy innate basis is number distribution. Differences between weight and number distribution can be tremendous for a broad PSD. Fig. 7 shows as an example the size
3. Results and discussion
Zinc oxide, reagent ACS by Acros Organics Zinc oxide, 99.5+% Acros Organics Z50-500 USP Z52-500 USP S80249 by Fisher Scientific Zinc oxide 99.99% by Alfa Aesar Zinc oxide ACS MO Biomedicals, LLC PolystormorTM by Mallinckrodt Chemicals Nanopowder America Elements
1.2
PSD,weight basis
1.0
0.8
0.6
0.4
0.2
0.0 -2
10
-1
10
0
10
1
10
Diameter [um] Fig. 1. Differential particle size distributions calculated from acoustic attenuation spectra assuming lognormal distribution shape. Median size and standard deviation are arithmetic averages of the multiple measurements.
21
A.S. Dukhin et al. / Journal of Colloid and Interface Science 342 (2010) 18–25
1.0 0.9
Cumulative,weight basis
0.8 0.7 0.6 0.5 Zinc oxide, ACS Acros Organics Zinc oxide, 99.5+% Acros Organics Z50-500 USP Z52-500 USP S80249 by Fisher Scientific Zinc oxide 99.99% by Alfa Aesar Zinc oxide ACS MO Biomedicals, LLC Polystormor, Mallinckrodt Chemicals Nanopowder America Elements
0.4 0.3 0.2 0.1 0.0 -2
-1
10
10
0
1
10
10
Diameter [um] Fig. 2. Cumulative particle size distributions calculated from acoustic attenuation spectra assuming lognormal distribution shape. Median size and standard deviation are arithmetic averages of the multiple measurements.
Mallinckrodt
Probabilities of Particles in the Bin
1.0 Mallingckrodt 0.8 0.6 0.4 0.2 0.0 0
100 200 300 400 500 600 Size (nm)
Z52-500
Probabilities of Particles in the Bin
1.0 Z52-500
0.8 0.6 0.4 0.2 0.0 0
100
200 300 400 Size (nm)
500
Fig. 3. TEM photographs and size histogram of the Malincrodt and Z52-500 ZnO dispersion. Bar is 500 nm.
600
22
A.S. Dukhin et al. / Journal of Colloid and Interface Science 342 (2010) 18–25
distribution for Z52-500 measured using acoustics and recalculated for both weight and number basis. Weight basis indicates the weight of the particles with a certain size. For instance, Z52500 contains only 2% of ZnO weight in the particles with sizes under 100 nm. Number basis presents number of particles with certain size. For Z52-500 the number of nanoparticles with sizes under 100 nm is about 75%. The same cumulative number is more than 35 times different if expressed either on a number or on a weight basis. TEM photos provide information for calculating PSD on a number basis. In order to compare these statistical data with acoustical PSD it should be converted to a weight basis. Unfortunately, this approach is not realistic due to the complexity of generating complete and representative particle size distribution using TEM. There is an International Standard describing this procedure, ISO 13322– 1 [16]. In particular, this standard determines the number of treated particles required to reach 95% of ‘‘correct” PSD. In the case of broad PSD it is several thousands. For comparison, each photo shown in Figs. 3–6, contains only up to 60 particles. Statistical treatment of hundreds photographs for generating a single reliable PSD is not a realistic approach for resolving problems of monitoring nanoparticle content. The situation would become even more hopeless if we take into account the possibility of various artifacts associated with interpretation of complex particles and aggregates shapes on the photo-
graphs. It is possible that TEM sample preparation aggregated initially separate nanoparticles in aggregates. That is why calculation of number of the nanoparticles on each TEM photo includes particles within aggregates. This might lead to overestimating nanoparticle content. This short analysis brings us to the conclusion that statistical histograms presented in Figs. 3–6 and calculated using just a single photo for each sample cannot be directly quantitatively compared with weight based produced by acoustics. We can use these photos only for some qualitative comparison of these two methods. These TEM photos yield some information that correlates with results of acoustic measurement. Here are some conclusions derived from comparison of the TEM photos that agree with data given in Table 1. First, sample Mallincrodt is clearly the smallest. There are two samples by Acros Organic. Sample ACS has much larger sizes than sample 99.5%, which agrees with Table 1. Samples MO Biomedicals and S80249 have similar size distributions, which agrees with Table 1 data. Sample Nanopowder by American is strongly aggregated. It looks like aggregates consist of nanoparticles. We could not find any contradiction between TEM photos and PSD results presented in Table 1. This comparison of the TEM photos and Table 1 results allows us to conclude that TEM confirms acoustics particle sizing, at least qualitatively.
Nanopowder by American
Probabilities of Particles in the Bin
1.0 AME
0.8 0.6 0.4 0.2 0.0 0
100
200 300 400 Size (nm)
500
600
Z50-500
Probabilities of Particles in the Bin
1.0 Z50-500
0.8 0.6 0.4 0.2 0.0
0
100
200 300 400 Size (nm)
500
600
Fig. 4. TEM photographs and size histogram of the American Nanopowder and Z50–500 ZnO dispersion. Bar is 500 nm.
23
A.S. Dukhin et al. / Journal of Colloid and Interface Science 342 (2010) 18–25
Acros Organic ACS
Probabilities of Particles in the Bin
1.0 acrosacs
0.8 0.6 0.4 0.2 0.0 0
100 200 300 400 500 Size (nm)
600
Acros Organic 99.5%
Probabilities of Particles in the Bin
1.0 Acros995
0.8 0.6 0.4 0.2 0.0 0
100
200
300 400 Size (nm)
500
600
Fig. 5. TEM photographs of two different Acros ZnO dispersions. Bar is 500 nm.
3.3. Mixture There is possibility of verifying acoustic particle sizing consistency by mixing two samples together. Also, this mixing test would yield some information on the sensitivity of the method to the nanoparticle content. The most informative would be additions of the small increments of the sample with the largest nanoparticle content to the sample with the least nanoparticle content. There are two ways one can estimate nanoparticle content in such mixtures. First, one can use known nanoparticle content in the individual samples and also known ratio of the individual samples in a particular mixture. For instance, the Mallinckrodt sample contains 20% of the nanoparticles, whereas Z52-500 sample contains only 2% of nanoparticles. Content of the Mallinckrodt sample in each mixture is also known. This is sufficient for calculating the expected nanoparticle content in each mixture. On the other side, measurement of the attenuation spectra for each mixture yields raw data for calculating particle size distribution. This is the same procedure as for individual samples. It provides nanoparticle content in mixtures similarly to individual samples. If everything is consistent, both methods should yield the same nanoparticle content in a particular mixture.
We have performed this consistency verification test using ZnO sample Z52-500 as a base. This sample contains the least amount of nanoparticles according to the acoustics, only 2%. Sample weight of the initial sample was 100 g, which is sufficient for filling the DT-1201 sample chamber. Additions of the Mallinckrodt sample, with the largest nanoparticle content, to Z52-500 were made in the increments of 2 g and then 4 g. The final mixture contained 26 g of Mallincrodt sample and 100 g of Z52-500 sample. We can calculate nanoparticle content in each mixture using the following algorithm. Initial 100 g of Z52-500 samples contains 5 g of ZnO, which in turns has only 0.1 g (2%) of nanoparticles. Let us assume that we have added M grams of the Mallincrodt sample. This addition contains 5 M/100 g on ZnO because the sample was prepared at 5 wt.% of ZnO. Nanoparticle content would be (5 M/100)/5 because only 20% of Malincrodt ZnO is nano. Total weight of ZnO in such mixture would be (5 g + M/20 g). Total weight of nanoparticles, assuming known individual samples PSD, would be (0.1 g + M/100 g). Percentage of nanoparticles in the mixture of 100 g Z52-500 samples and M grams of Mallincrodt samples P equals
P¼
0:1 þ M=100 100: 5 þ M=20
24
A.S. Dukhin et al. / Journal of Colloid and Interface Science 342 (2010) 18–25
MO Biomedical
Probabilities of Particles in the Bin
1.0 Mo-bio
0.8 0.6 0.4 0.2 0.0 0
100
200
300 400 Size (nm)
500
600
S80249
Probabilities of Particles in the Bin
1.0 S80249
0.8 0.6 0.4 0.2 0.0 0
100
200 300 400 Size (nm)
500
600
Fig. 6. TEM photographs of the MO Biomedical and S80249 ZnO dispersion. Bar is 500 nm.
number basis
1.1
Alternatively, we can estimate the percentage content of nanoparticles in each mixture from acoustic attenuation spectra measured for the mixture. These two numbers should be the same. This is the essence of the performed mixing test. Fig. 7 presents graphically results of the test. Values of the X-axis equals the value of percentage P calculated for each mixture. Actual results of the acoustic sizing measurements are represented with circles in Fig. 8. Solid line represents idea case when measured content equals exactly the expected nanoparticle content P. It is seen that deviation of the acoustically measured nanocontent from the expected values is on a scale of 1%. It is significantly less than 2% precision level, which we determined as precision threshold based on study of individual samples. We can conclude that acoustic spectroscopy yields nanopatricle content in these mixtures that agrees well in trend with expected values.
weight basis
1.0 0.9
particle size distribution
0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1
4. Conclusions
0.0 -2
10
-1
10
0
10 diameter [microns]
1
10
Fig. 7. Particle size distribution of the Z52-500 sample on a weight and number basis.
Particle size characterization using acoustic spectroscopy allows monitoring of nanoparticle content in the dispersions with broad PSD with precision of at least 2%. It can be achieved in concentrated dispersions with no dilution. This conclusion is confirmed by multiple measurements of individual samples of eight
A.S. Dukhin et al. / Journal of Colloid and Interface Science 342 (2010) 18–25
25
different ZnO dispersions and their mixtures. Electron microscopy confirms qualitatively results of PSD characterization using acoustic spectroscopy.
7
6
% nano-particles, measured
References 5
4
3
2
1 1
2
3 4 % nano-particles, estimated
5
6
Fig. 8. Percentage of nanoparticles in Z52-500 sample after incremental additions of Mallinckrodt Chemicals sample. X-axis is a percentage calculated from the known amount of the added Mallinckrodt sample, assuming that it contains 20% on nanoparticles, according to the Table 1. Y-axis is a percentage calculated from the attenuation spectra, which is measured for the mixture.
[1] R. Behra, H. Krug, Nat. Nanotechnol. 3 (2008) 253–254. [2] Groups encourage nanomaterial stewardship, Chemical & Engineering News, July 21, 2008, p. 34. [3] G. Miller, Nanomaterials, sunscreens and cosmetics: small ingredients–big risks, Friends of the Earth, 2006