Classification Strategies for Regulatory Nanodefinitions - Safety of Nanomaterials along Their Lifecycle - page 51

Sedimentation (CLS, acc. to ISO13318) on eight pigments and fillers with a stan-
dardized dispersion protocol based by probe sonication in water with a dispersant
(Gilliland and Hempelmann 2013). Focusing on the same iron oxide Pigment Yellow
42 as previously noted, Figure 3.2d presents the TEM, CLS, DLS, and LD results.
With LD, the resulting D50 in number metrics originates from less than 5% of the
originally measured LD distribution in volume metrics, which is fully within the
error margin. There is hence little confidence on LD results for partially agglomer-
ated pigments. With one exception, the results obtained with CLS are rather uni-
form, even though different manufacturers with different optics (X-ray and turbidity)
(Planken and Cölfen 2010; Wohlleben 2012) were used. The BASF_CLS measure-
ment was dominated by agglomerates, and is displayed in Figure 3.2e as dashed
lines. One identifies several systematic sources of error: deviations in true size distri-
bution after dispersion +, mathematical smoothing between raw data and size distri-
bution +, sphere-equivalent conversion from volume to numbers. Error propagation
and amplification of noise are major concerns since a 1% error in the ability of a
method to accurately describe a mass or volume distribution at the nanoscale could
translate into more than 50% error in a number distribution. An example of such an
error propagation is displayed in Figure 3.2e, where a kink in the originally mea-
sured black solid line (mass metrics) introduces a seemingly dominant peak of the
grey solid line (number metrics) at the lowest diameters, which is neither plausible
for material chemistry nor compatible with TEM images in Figure 3.2a.
In summary from the experiments discussed here, VSSA and CLS are the pre-
ferred methods due to the low scatter between laboratories and due to inherently
agglomerate-tolerant principles, but for CLS the dispersion protocol and metrics
conversion must be further enhanced, as discussed in the following section.
Apart from the aforementioned techniques introduced, many alternatives
(Figure 3.1) have been discussed and ranked elsewhere (Allen 1997a; Anderson et al.
2013; Brown et al. 2013; Calzolai, Gilliland, and Rossi 2012; Linsinger et al. 2012).
A realistic roadmap from 2013 to 2017 will plan for only incremental improvements
for techniques that have been established since decades with several commercial sup-
pliers. The dynamic range of Field-Flow-Fractionation (FFF) per run condition will
likely remain limited and multiple spacers, membranes, and run conditions would be
required to access a dynamic range from a few nm to a few µm (Baalousha and Lead
2012; von der Kammer et al. 2011). Also the uncertainties of the optical detection in
CLS will remain limiting at least for the low-cost commercial equipment, whereas
Analytical Ultracentrifugation (AUC) with interference detection does resolve the
issue but is still too expensive as an investment (Cölfen et al. 2009; Schäfer et al. 2012;
Wohlleben 2012). The coupling of innovative, even counting detectors such as single-
particle inductively coupled plama mass spectrometry (sp-ICP-MS) or condensation
particle counter (CPC) to a fractionation channel does not resolve the necessity for
dispersion. Major improvements in the applicability of FFF and CLS are expected for
software tools that perform validity checks (see the following section) and circumvent
the commercial conversion from raw data to smoothed volume distributions. Instead,
these software tools would perform a direct error propagation to number metrics.
Especially the counting techniques are still in their infancy, often with only
one commercial supplier and no standardization but several research projects; this