Analytical challenges of determining composition and structure in small volumes with applications to semiconductor technology, nanostructures and solid state science

The interplay between structure, composition and properties form the heart of materials science and the understanding of the relationships between them enable the manufacture of both advanced materials and complex devices with defined structure and dimension. As these materials and devices approach atomic-scale dimensions, further opportunity exists to realize new and useful properties not available in macroscopic dimensions. Historically, average values were used to characterize the materials produced, with average structure typically determined using x-ray diffraction techniques and average composition determined from either starting conditions, direct chemical analysis or via energy dispersive/wavelength-dispersive x-ray analysis. Increasingly, however, it has been recognized that advanced processing techniques result in materials that are inhomogenous in either structure, composition or both. Understanding how processing conditions control local nanostructure and composition and how these inhomogeneities result in unique properties is of fundamental and applied importance, and the subject of much current research. Measuring these inhomogeneities at atomic-scale dimensions is an analytical challenge.

A number of new and enhanced analytical techniques have been or are being developed to determine the structure and composition of small volumes. Advancing the metrology of nanoscale materials benefits both basic research (nanoparticles, nanocomposites, ultrafine grain metal alloys containing nanoprecipitates) and cutting edge technology (nanoscale semiconductor structures). Being able to quantify local composition and structure is vital to understand how to control nanoscale properties and critical for advancing applications such as semiconductor device manufacturing. For example, as semiconductor features become smaller, it becomes increasingly more important and more difficult to determine the composition and structure of active regions, to determine structure-function relationships, and to enable efficient and reliable manufacturing. This focus issue contains articles on a variety of techniques including electron microscopy, atom probe tomography, and diffraction, with leaders in the field describing approaches to overcome the challenges in obtaining statistically significant atomic-scale quantification of structure and composition in increasingly small analytical volumes using each technique.

Several papers focus on measuring local composition via a variety of methods. Tsui et al assess and compare four different methods for measuring the average compositions of epitaxial Sc_xGa_1−xN films to determine which was the most reliable and accurate [1]. The compositions of these crystalline films with 0 ≤ x ≤ 0.26 were measured using Rutherford backscattering (RBS), x-ray photoelectron spectroscopy (XPS), c-axis lattice parameter measurements from x-ray diffraction and c/a ratio measurements from electron diffraction patterns. RBS measurements were taken as a standard reference. XPS was found to underestimate the Sc content, whereas c-axis lattice parameter and c/a ratio were not reliable for composition determination due to the unknown degree of strain relaxation in the film. The Sc flux used during growth was found to be proportional to x at a constant Ga flux, enabling a quick way for the Sc content to be estimated. A paper by Aoki et al reports the observation of six different zincblende compound semiconductors in [110] projection using large-collection-angle bright-field imaging with an aberration-corrected scanning transmission electron microscope [2]. The phase contrast is shown to be completely suppressed when the collection semi-angle is set equal to the convergence semi-angle, resulting in no reversals in image contrast with changes in defocus or thickness. The optimum focus for imaging closely separated pairs of atomic columns ('dumbbells') is unique and easily recognized, and the positions of atomic columns occupied by heavier atoms always have darker intensity than those occupied by lighter atoms, enabling the crystal polarity of compound semiconductors to be determined unambiguously. The large-collection-angle bright-field imaging mode provides advantages in studying other more complicated heterostructures at the atomic scale. Kwon et al show that the local composition and the lattice parameter follow Vegard's law in (LaNiO₃)₄/(LaMnO₃)₂ binary oxide superlattices [3]. They obtained the elemental distribution and lattice strain maps from the measured atomic column positions over a large field of view. The maps allow direct observation of compositional defects of the B-sites, which is not possible by Z-contrast alone. This is a promising approach for an atomic scale correlative study of lattice strain and composition, and a method for the calibration of atomic resolution EDS maps. Mitchson et al use variations in HAADF STEM intensities from maps obtained using a low dose to limit beam-induced damage in an amorphous oxide film to determine local composition inhomogeneities [4]. The significantly larger spread of intensities in the substrate's normal direction relative to parallel to the substrate were shown to reflect inhomogeneity perpendicular to the substrate. For an amorphous multi-coat solution-cast oxide sample, the analysis statistically reveals significant variations in the HAADF STEM intensity profile perpendicular to the substrate corresponding to each coat. The inhomogeneous density profiles resulted from the spin-casting of the individual layers, reflecting the chemistry occurring during condensation, which has implications for understanding the properties of multilayer amorphous oxide films from solution. The signal intensity variation parallel to the substrate was used to estimate the signal noise level assuming the films are chemically homogeneous in this direction, providing an estimate of the error of the composition determination. In another amorphous system, Zhu et al demonstrated that precise control of the electron dose rate during state-of-the-art large solid angle energy dispersive x-ray spectroscopy mapping in an aberration-corrected scanning transmission electron microscope is able to determine the Pt and Pd distribution in a peptide-mediated nanosystem [5]. This low-dose-rate recording scheme adds another dimension of flexibility to the design of elemental mapping experiments, and holds significant potential for extending its application to a wide variety of beam sensitive hybrid nanostructures.

Several articles demonstrate how a variety of electron microscopy techniques can be used to determine the local structure. Dankwort et al used electron diffraction and high resolution transmission electron microscopy (HRTEM) in conjunction with image simulation to characterize layered structural motifs in the ternary compound Ni_3−xSn_1−yTe₂ [6]. The layered structural motifs in this compound are related to an average NiAs/Ni₂In-type. Dependent on the stoichiometry, commensurate and incommensurate satellite reflections with respect to the parent NiAs structure were observed in Fourier transform and electron diffraction pattern as a result of occupational modulation of Te and Sn atoms. For the commensurate case a tripling of the c-lattice parameter occurs as a result of Sn–Te–Te stacking. HRTEM micrographs indicate additional ordering phenomena along the c* direction depending on Ni/vacancy ordering, which was rationalized by an alternating filling of van der Waals gaps with Ni. HRTEM investigations prove that morphological defects observed in bright field images are based on domains shifted relative to each other (antiphase boundaries). Mitchson et al used high angle annular dark field scanning transmission electron microscopy (HAADF STEM) image analysis to determine the atomic coordinates of complex semi-crystalline materials containing planes of two different structures that are rotationally disordered with respect to one another [4]. The positions of HAADF STEM intensity peaks in the stacking direction were averaged from multiple images to obtain the atomic coordinates. This approach provides a valuable initial model for a Rietveld refinement of the global c-axis structure of the heterostructures from diffraction data collected over a large area. Soo et al studied the effect of growth temperature on the orientation and crystallographic morphology of Au-catalyzed epitaxial GaAs semiconductor nanopillars grown by metal–organic chemical vapor deposition using electron microscopy [7]. Growth temperature played a significant role on the evolution of side-facets of zinc-blende structured GaAs nanopillars. At a growth temperature of 550 °C, six (112) side-facets are formed; whereas at a higher growth temperature of 600 °C, six (110) side-facets are observed. It is believed that the formation of (112) side-facets is a kinetically dominated process while the formation of (110) side-facets is a thermodynamical process. Diffusion-induced nanopillar foundations present the same (112) edge side-facets regardless of the growth temperature.

Transmission electron microscopy (TEM) has also become indispensable for measuring the critical dimension (CD) of structures as well as local properties in combination with techniques such as scanning probe microscopy. Lee et al discuss a new method for the measurement of transistor gate length using energy-filtered transmission electron microscopy [8]. The electrons transmitted through a plan-view TEM sample of a silicon based device provide diverse information about various overlapped silicon-based materials - silicon, silicon dioxide, and silicon nitride. This information is convoluted in an exceedingly complex manner, however, making it difficult to clearly identify boundaries between materials. Energy-filtered TEM provides a precise and effective measurement condition by determining the maximum value of the integrated area ratio of the electron energy loss spectrum at the boundary to be measured. EF-TEM imaging shows a sharp transition at the boundary when the energy-filter's passband center was set at the optimum condition for the CD measurement of silicon-based materials involving silicon nitride. Electron energy loss spectroscopy (EELS) and EF-TEM images were used to verify this method, which makes it possible to measure the transistor gate length in a dynamic random access memory manufactured using 35 nm process technology. This method can be adapted to measure the CD of other non-silicon-based materials using the EELS area ratio of the boundary materials. In another article, Yang et al used a fixed probe within a scanning electron microscope to measure an average metal-semiconductor diode barrier height of 0.69 ± 0.03 eV (ideality factor 1.48 ± 0.02) for epitaxial Fe contacts fabricated onto the top half of free-standing, Te-doped GaAs nanowires via electrodeposition [9]. Electrical isolation from the substrate via a polymeric layer enabled the measurement of electrical transport through individual wires.

A rapidly developing technique, atom probe tomography, was the topic of several articles addressing the accuracy of analyses and the optimization of sample preparation and analysis conditions. Atom probe tomography combines a field ion microscope with a mass spectrometer having a single particle detection, with the 3D architecture of the sample reconstructed from the time and location of the arrival of individual atoms. The routine use of atom probe tomography as a nano-analysis microscope in the semiconductor industry requires systematic evaluation of the metrological parameters of this instrument (spatial accuracy, spatial precision, composition accuracy and/or composition precision). Vurpillot, et al evaluate the spatial accuracy of atom probe tomography via the analysis of planar structures such as high-k metal gate stacks [10]. They show that the in-depth accuracy of reconstructed APT images is perturbed when analyzing a structure composed of an oxide layer of high electrical permittivity (higher-k dielectric constant) that separates the metal gate and the semiconductor channel of a field emitter transistor. The main sources of the image distortions are the large differences in the evaporation field between these layers, caused by the large differences in material properties between the different layers. They present an analytic model to interpret the inaccuracy in the depth reconstruction of these devices. Douglas et al illustrate the challenges of surface analysis using atom probe tomography through the characterization of near-surface implantation profiles of low concentration phosphorus into single crystal silicon [11]. This system requires particular attention to specimen preparation using a focused ion beam and care during the deposition of various capping layers because phosphorus has significant mass spectra overlaps with silicon species and the near surface location. Implantation profiles of 14 kV phosphorus ions with a predicted peak concentration as low as 0.2 at.% were successfully analyzed using pulsed laser assisted evaporation atom probe tomography. The most important factor in obtaining accurate implantation profiles was to ensure that the phosphorus mass peaks were as free of background noise as possible. The major overlap in the mass spectrum were thermal tails from the Si²⁺ ions obscuring the P²⁺ ions. The initial capping layer selection of nickel was successful in allowing the analysis of the majority of the phosphorus profile but nickel and phosphorus mass spectra overlaps prevented optimum quantification of phosphorus at the surface.

Interest in these and other analytical approaches to determining local composition and structure will continue to grow, as basic research into materials such as 2D layers, heterostructures, nanoprecipitates in advanced metal alloys, and novel architectures in nanoparticles continue to yield emergent properties not found in bulk composites. Continued advancements in analytical capabilities are required to enable cutting edge technology, particularly in the semiconductor industry, as feature sizes decrease to improve device performance and features. The editors hope that this special issue provides a glimpse into the complexity of extending analytical techniques to higher precision and smaller dimensions, and that the articles in this special issue illuminate both the progress in technique and the challenges that remain.

Acknowledgments