The spread of plastics in the environment has led to the formation of microplastics (MPs) and even smaller nanoplastics (NPs) through processes such as aging, degradation, and fragmentation. While significant research has focused on quantification and understanding the potential harmful effects of larger MPs [1], characterizing NPs proves to be challenging, especially within complex matrices [2]. A comprehensive physicochemical characterization is however imperative, especially for evaluating their potential toxicological impacts. In this study, we establish a foundational framework for such measurements by validating a transmission electron microscopy (TEM)-based approach, in the context of the European project PlasticTrace [3]. This validation encompasses all analysis steps, from sample preparation to image analysis, focusing on reference materials composed of polystyrene nanoplastics. By establishing the full measurement uncertainty balance, we aim to determine the accuracy, precision and reliability of NP size and shape characterization by TEM.
The method is validated supporting on the Nanosphere size standards 3060A (60 ± 4 nm), 3200A (202 ± 4 nm) and 3500A (510 ± 7 nm), purchased from Thermo Fisher Scientific. They are part of a series of polystyrene micro/nanospheres with certified mean diameters traceable to the Standard Meter through the National Institute of Standards and Technology (NIST). The sample preparation consists of optimally diluting the colloidal sample suspension and bringing it on an Alcian blue pre-treated TEM grid. For each material, 15 different TEM specimens are prepared and imaged on 5 different days (3 per day). For each specimen, 10 images are recorded systematically and randomly over the grid surface. Each series of 10 images is analyzed using the ParticleSizer software in ImageJ. The intermediate precision of the quantitative TEM measurement is evaluated using a top-down approach [4] combining the uncertainty related to repeatability (within day) and uncertainty related to day-to-day variations (between day) obtained by ANOVA analysis. Adding the uncertainty related to calibration of the microscope and to trueness, allows to estimate the total combined and expanded uncertainty for the mean, mode and percentiles of parameters including the minimum and maximum Feret diameter (Fmin & Fmax), the equivalent circle diameter (ECD), the maximum inscribed circle diameter (MICD) and the aspect ratio (AR).
Material stability was pertained throughout the validation study and homogeneous distribution of particles on the grid was achieved for all materials. Material 3060A is polydisperse with particle sizes ranging from 10 nm to 70 nm. Materials 3200A and 3500A are more uniform in size, however, a small fraction of particles (<2%) have a significantly larger or smaller diameter. STEM-EDX was performed to verify that all particles have the same elemental composition and thus belong to the polystyrene sample. The intermediate precision obtained is similar for all size parameters. It ranges from 1.1-6.7%, 0.7-1.5% and 0.3-0.7% for materials 3060A, 3200A and 3500A, respectively. The highest values correspond to the d10 percentile and the lowest values to the d75 or d90 perecentiles due to left skewedness of the size histograms (see graphic). The intermediate precision is highest for material 3060A due to the higher degree of polydispersity. For the AR, the intermediate precision is below 1% for all materials and measurands. The main source of uncertainty is related to the trueness uncertainty for all materials. To assess the accuracy of our approach, a comparison with the certified size values (see Methods section) is required [5]. Specific information on how the certified diameter was obtained is lacking, however, we assume it corresponds to the mean ECD of the material. The values and expanded uncertainties we obtained for the mean ECD are 51±4 nm, 198±7 nm and 518±13 nm, for the three materials respectively. Based on a comparison with the certified values, we conclude that the two largest materials are accurately measured, however the smallest material presents a significant difference with respect to the reference value. Since the mode of the distribution is closer to the reference value of 60 nm, a difference in analysis algorithm e.g. leading to exclusion of a fraction of the smaller particles, might be at the origin of this discrepancy. We see indeed that for material 3060A, the outcome of the validation study is less robust against variations in image analysis settings. However, to make a final conclusion on the trueness of our approach, more detailed information on the certification of the materials is required.
This study establishes a robust methodology for the validation of TEM-based measurements on reference materials of polystyrene nanoplastics. By evaluating the measurement uncertainties associated with size and shape analysis, considerable precision is achieved, however, ensuring trueness can be challenging and
requires detailed information on the reported certified diameters. These findings underscore the importance of ongoing refinement and harmonization in analysis algorithms to enhance the accuracy of nanoplastic characterization.
