The size characterization of particles present in the form of agglomerates in images measured by scanning electron microscopy (SEM) requires a powerful image segmentation tool in order to properly define the boundaries of each particle. In this work, we propose to use an algorithm from the deep statistical learning community, the Mask-RCNN, coupled with transfer learning to overcome the problem of generalization of the commonly used image processing methods such as watershed or active contour. Indeed, the adjustment of the parameters of these algorithms is almost systematically necessary and slows down the automation of the processing chain. The Mask-RCNN is adapted here to the case study and we present results obtained on titanium dioxide samples (non-spherical particles) with a level of performance evaluated by different metrics such as the DICE coefficient, which reaches an average value of 0.95 on the test images.

At this time, there is no instrument capable of measuring a nano-object along the three spatial dimensions with a controlled uncertainty. The combination of several instruments is thus necessary to metrologically characterize the dimensional properties of a nano-object. This paper proposes a new approach of hybrid metrology taking advantage of the complementary nature of atomic force microscopy (AFM) and scanning electron microscopy (SEM) techniques for measuring the main characteristic parameters of nanoparticle (NP) dimensions in 3D. The NP area equivalent, the minimal and the maximal Feret diameters are determined by SEM and the NP height is measured by AFM. In this context, a kind of new NP repositioning system consisting of a lithographed silicon substrate has been specifically developed. This device makes it possible to combine AFM and SEM size measurements performed exactly on the same set of NPs. In order to establish the proof-of-concept of this approach and assess the performance of both instruments, measurements were carried out on several samples of spherical silica NP populations ranging from 5 to 110 nm. The spherical nature of silica NPs imposes naturally the equality between their height and their lateral diameters. However, discrepancies between AFM and SEM measurements have been observed, showing significant deviation from sphericity as a function of the nanoparticle size.

Scanning Electron Microscopy (SEM) is considered as a reference technique for the determination of nanoparticle (NP) dimensional properties. Nevertheless, the image analysis is a critical step of SEM measuring process and the initial segmentation phase consisting in determining the contour of each nano-object to be measured must be correctly carried out in order to identify all pixels belonging to it. Several techniques can be applied to extract NP from SEM images and evaluate their diameter like thresholding or watershed. However, due to the lack of reference nanomaterials, few papers deals with the uncertainty associated with these segmentation methods. This article proposes a novel approach to extract the NP boundaries from SEM images using a remarkable point. The method is based on the observation that, by varying the electron beam size, the secondary electron profiles crosses each other at this point. First, a theoretical study has been performed using Monte Carlo simulation on silica NP to evaluate the robustness of the method compared with more conventional segmentation techniques (Active Contour or binarization at Full Width at Half-Maximum, FWHM). The simulation results show especially a systematic discrepancy between the NP real size and the measurements performed with both conventional methods. Moreover, generated errors are NP size-dependent. By contrast, it has been demonstrated that a very good agreement between measured and simulated diameters has been obtained with this new technique. As an example, this method of the remarkable point has been applied on SEM images of silica particles. The quality of the segmentation has been shown on silica reference nanoparticles by measuring the modal equivalent projected area diameter and comparing with calibration certificate. The results show that the NP contour can be very accurately delimited with using this point. The measurement uncertainty has been also reduced from 4.3 nm (k = 2) with conventional methods to 2.6 nm (k = 2) using the remarkable point.

The scanning electron microscopy (SEM) technique is widely used for the characterizing of nanoparticle (NP) size, but very few papers deal with NP dimensional metrology. This article reports on a methodology with which to evaluate the uncertainty budget associated with the measurement of the mean diameter of a standard silica NP population by SEM. In this context, the effects of potential error sources have been evaluated though a metrological qualification of the instrument. The measuring method, consisting of determining the area equivalent diameter taken at middle height ( D _eq-FWHM ), has been tested on a reference silica NP with an indicative certification value given by SEM/TEM (number-based modal diameter). Because agglomeration phenomena can cause measurements errors, semi-automatic homemade software has been employed to build the diameter distribution histogram, selecting only isolated particles. Finally, an uncertainty budget, including the main experimental components, has been established for the mean diameter measurement of this silica NP population. The main contributors to this uncertainty budget are the resolution linked to the dimension of the electron beam diameter at the focal plane, the calibration uncertainty on reference NPs, and the measurement repeatability.

Scanning Electron Microscopy (SEM) technique is widely used to characterize nanoparticle (NP) size. The landing energy (LE) of the primary electron beam is considered to be a key parameter related to the ability of electrons to penetrate the sample. However, few studies have been carried out so far on the influence of this parameter on the measurement of NP size by SEM. The increasing needs for reference materials consisting of size-controlled NP suspension for microscope calibration induce new issues. This paper focuses on the effect of electron landing energy on the measurement of the equivalent diameter of several NP populations by SEM. To evaluate the influence of LE, particles of different sizes and chemical compositions were analyzed. The results showed the variation of the measured diameter as a function of LE. SEM secondary electron (SE) yield modeling by the Monte Carlo method allowed us to relate this variation to the information volume in the material. Finally, the use of reference particles and transmission electron microscopy (TEM) allowed us to determine an optimal value of LE to be applied, depending on the chemical composition and particle size to limit the bias in the SEM measurement. We showed that this operating point can be simply determined without reference nanomaterials by scanning an LE range.