he Role of Crystal Structures in Semiconductors and the Analytical Applications of X-ray Diffraction

As technology and advanced processes continue to develop, electronic products are becoming smaller and smaller, but the size of semiconductor components is approaching the limits of physical materials. Next generation electronic products are required to be small, fast and multi-functional while also having low power consumption. To meet these requirements and circumvent the physical limitations of component size, surpassing Moore’s Law has become the goal of the current semiconductor industry.

As the size of semiconductor components can no longer be reduced via process technology alone, most of the semiconductor industry has moved towards reducing component sizes through the use of three-dimensional stacking, such as 3D packaging technology. By vertically stacking component layers, the number of components per unit area can be increased, thus enabling the reduction of manufacturing costs. Furthermore, components with different functions can be integrated to form components with more diverse capabilities.

In addition to adapting process technologies, major semiconductor companies are also focused on the research and development of new semiconductor materials, such as gallium nitride (GaN), silicon carbide (SiC), indium phosphide (InP) and other compound semiconductor materials. Compound semiconductor materials are of particular interest because they have a variety of advantageous characteristics, including Direct Bandgap, High Breakdown Voltages, and high Electron Mobility.

The semiconductor industry is one of Taiwan’s most important industries, and the factor with the greatest influence on a semiconductor component’s characteristics is its materials. A material’s performance depends on its structure, and a material’s structure is closely related to factors such as its chemical composition, type of bonding and structure arrangement. This article will focus on introducing application cases for X-ray diffraction analysis (XRD) in the quality control and analysis of the semiconductor manufacturing process.

A material’s structure is comprised of an arrangement of atoms. A single crystal is comprised of atoms arranged regularly in a spatial grid. A polycrystalline structure is comprised of single crystals with different regularities arranged in a spatial grid. Amorphous materials do not have a long-range, ordered structure. Differences in arrangement have a great influence on a material’s applications. Take solar cells as an example. Solar cells are a type of photoelectric component for converting energy; the semiconductor materials absorb sunlight to generate electricity. Silicon solar cells come in three types: single-crystal silicon, polycrystalline silicon and amorphous silicon (see Figure 1). The single-crystal silicon has the best conversion efficiency and a longer service life, but its manufacturing costs are high, making it most suitable for use in places such as power plants and traffic lights, etc.. Though polycrystalline silicon is less efficient at power-generation than single-crystal silicon, it has a lower production cost and a simpler manufacturing process. As a result, it is currently the mainstream in the solar cell market. Amorphous silicon has the lowest conversion efficiency, but it can be produced quickly at low costs in addition to needing only a micron-level coating to be effective. At present, it is widely used in thin-film solar panel batteries suitable for windows and portable power banks, etc. In recent years, it has even been applied to camping vans, the exterior walls of buildings, portable solar panels and more. XRD analysis can be used to determine the crystal structure arrangement of a material, as shown in Figure 2.

Figure 1. The energy conversion efficiency and applications of different types of silicon solar cells (Source: Differences Between the Types of Solar Panels from Transcend)

Figure 2. Materials can be differentiated using XRD or TEM diffraction patterns at select areas

In addition to its crystal structure, a material’s grain size and crystallinity also have an influence on its mechanical properties (such as elasticity, plasticity, stiffness, strength and hardness) and its physical properties (such as its electrical, magnetic, optical and thermal properties).

A material’s grain size is directly related to the surface area of its grain boundary. The smaller the grains, the larger the surface area and the greater the impact on the material’s properties. In regards to the mechanical properties of metal, the smaller the grain size, the higher the strength and hardness and the better the plasticity and toughness. In regards to electrical properties, however, larger grains mean fewer grain boundaries, which in turn means less resistance to electron migration and higher mobility. As a result, larger grains are better for electron transfer. XRD analysis is able to determine grain sizes as well as the crystal orientation distribution at the mm~ cm level (see Figure 3), whereas Electron Back Scatter Diffraction (EBSD) can perform similar analysis of much smaller areas (um level) (see Figure 4).

Figure 3. Analysis of the crystal orientations, average grain size of each orientation, overall average grain size and crystallinity of copper blocks

Figure 4. EBSD analysis of crystal orientations and grain size distributions of copper blocks; It can be observed that the {111} is the main orientation

As technology advances, semiconductor components are getting smaller and smaller. Coating thickness for components needs to be reduced accordingly, which limits the grain growth in the materials. As a result, the influence of crystal orientation on a component’s performance in areas such as electron migration speed is increasing. In undoped silicon wafers, for example, the {111} and {112} crystal orientations have better conductivity than the {100} and {110}, whereas in copper, the {100} crystal orientation has better conductivity than the {111}. As such, the crystal orientation of materials also needs to be considered when developing semiconductor components (see Figures 5 and 6).

Figure 5. A comparison of the differences in silicon material properties due to differences in crystal orientation (Source: The Journal of Physical Chemistry C, 122 (24), 13027-13033.)

Figure 6. XRD Analysis of Different Components: (Left) FRD Substrate with (111) Crystal Orientation; (Right) IGBT with (100) Crystal Orientation

When observing thin film crystal structures of above tens of nanometers in size, it is typical to use Grazing Incidence X-ray Diffraction (GIXRD). The idea is to extend the path taken by the X-ray through the material. This increases the film’s diffraction signal. By setting the material’s theoretical density and incident angle, the X-ray’s depth of penetration into the thin film can be controlled during Grazing Incidence X-ray Diffraction analysis (see Figures 7 and 8).

Figure 7. Schematic Diagrams of Traditional X-ray Diffraction: (a) θ-2θ Scanning Mode and (b) Grazing Incidence Scanning Mode (Source: Scientific Instruments Column _ Applications of Grazing Incidence X-ray Diffraction in 2D Material Crystal Analysis)

Figure 8. (Left) Analysis of Crystal Orientation and Grain Size of Molybdenum Thin Films Using the Grazing Incidence Method; (Right) When the incidence angle is 0.5, the depth of the X-ray’s penetration into the molybdenum thin film is about 126nm

In the case of coatings of only a few nanometers in thickness or the few Å of 2D materials, however, the layers are so thin that effective diffraction signals cannot be obtained even using GIXRD analysis. However, the in-plane grazing incidence diffraction (in-plane GID) analysis method can overcome the problem presented by ultra-thin films and provide nondestructive analysis to obtain information such as material crystal orientations, grain sizes and crystal structures. This method relies mainly on the geometric structures of the thin film/material plane to increase the length of the X-ray’s path through the film, thus obtaining effective diffraction signals. Compared to Grazing Incidence Diffraction, this method significantly reduces the X-ray’s penetration depth and further reduces the signals from the substrate (see Figures 9 and 10).

Figure 9. Schematic Diagram of the In-Plane GID Scanning Mode’s Internal Spatial Structure (Source: (Left) Scientific Instruments Column _ Applications of In-Plane Grazing Incidence Diffraction in 2D Material Crystal Analysis); (Right) In-Plane GID Scanning Diagram (Source: The Rigaku Journal, 26(1), 2010.)

Figure 10. Comparison of Grazing Incidence Diffraction and In-Plane GID Analysis of 2nm Pt on Glass; The In-plane GID can effectively reduce background noise

During the semiconductor film deposition process, it is necessary to adjust the process parameters to control the deposition rate and ensure that the film reaches the specified thickness. Thickness analysis can be performed without damaging the film using the X-ray Reflectometry (XRR) form of XRD. Where TEM analysis can obtain accurate, local (nm level) thickness values as well as observe the subtle changes in the surface, XRR can measure film thickness, density and surface interface roughness over larger (mm level) areas (see Figure 11).

Figure 11. (Left) XRR Spectra and Fitting Analysis of SiGe Before and After Etching; (Right) TEM Slices for SiGe Film Before and After Etching; Note that the trends are similar

Mobile communication has become an integral part of our modern lives. The filter in the communication system is essential to determining communication quality. AlN filters have many advantages, such as high frequency operation, high temperature stability and compatibility with CMOS processes. These qualities enable them to realize the integration of filter chips in order to meet the demand for small, light and thin mobile communications devices. This has led it to be the main force in the production of filter chips. More and more is being required of filters in the current era of 5G communications. They must not only meet the requirements of high frequency operation but also have larger bandwidths and lower Q values in regards to signal leakage. In recent years, some researchers have found that the stress generated by doping AlN with Sc atoms can result in Sc_xAl_1-xN  that demonstrates better piezoelectric performance than AlN (Source: Science and Technology News Issue 233: Sc_xAl_1-xN Piezoelectric Resonator Realizes 5G mm Wave Mobile Communications Integration). As a result, how to determine whether a high quality AlScN film has grown on a silicon substrate has become an important subject for communications device manufacturers. The XRD Rocking Curve method can be used to monitor the quality of coatings. When the crystal structures are arranged regularly, swinging along the axis of a specific, fixed diffraction peak  will greatly reduce its diffraction intensity because it will no longer conform to Bragg’s Law. In contrast, when the lattice is randomly arranged, since there are diffraction peak components in each direction, the intensity of the diffraction peak will decrease more slowly as the axis swings, as shown in Figure 12.

Figure 12. Effect of Coating Lattice Stacking Arrangement on the Rocking Curve

Below, Figure 13 shows that both AlScN stacking structures exhibit the preferred (002) orientation. However, according to the results of the Rocking Curve analysis, it is clear that the full width at half maximum of the structure in the left image is narrower, indicating that its stacking structure is more beneficial to the AlScN surface arrangement.

Figure 13. (Top) XRD coating analysis shows that both AlScN have the preferred (002) orientation; (Bottom) Rocking Curve analysis of the AlScN (002) shows that the full width at half maximum (FWHM) of the structure in the left image is narrower, indicating that the stacking structure is more beneficial to the AlScN surface arrangement.

As the Si MOSFET component processes continue to evolve and shrink, increasing the mobility of electrons and holes in channels is essential to improving device performance. Among the methods for increasing mobility, Strained Engineering (Strained-Si) has proven to be one of the most effective for improving the performance of Si nano-components. Among the available options, SiGe is one of the most attractive PMOS materials for strained engineering because SiGe has higher mobility and better reliability in negative bias temperature instability (NBTI) than Si. It’s also a better match for the Si substrate lattice.

Since Si and SiGe have the same crystal structure and very similar lattice spacing, if the traditional diffraction analysis method is used, the diffraction peaks of the two will all but overlap. At such times, single crystal accessories are needed to improve the angular resolution of the XRD and the collimation of the incident light, making the analysis more sensitive to lattice changes. Figure 14 used HRXRD analysis to obtain the SiGe diffraction peaks then used software to fit the relative position of the diffraction peaks in order to obtain the SiGe ratio of each layer. Furthermore, the periodic oscillation of the satellite peaks around the diffraction peaks can be used to determine the thickness of each layer.

Figure 14. HRXRD Analysis of Multilayer Epitaxial SiGe and Si; By fitting the relative position of the diffraction peak using software, the ratio of each SiGe layer can be obtained; The periodic oscillation of the satellite peak around the diffraction peak can be used to determine the thickness of each layer.

In order to maintain the integrity and continuity of the interfacial atomic bonding between the Si and SiGe during the SiGe epitaxial growth process, the lattice spacing of the SiGe epitaxial layer must be deformed to accommodate the lattice spacing of the Si substrate. This raises the question of lattice matching. Lattice mismatches between Si and SiGe layers can be divided into three categories: completely relaxed, partially relaxed and completely deformed. A variety of epitaxial structural characteristics can be obtained by observing the changes in the diffraction peak distributions and the corresponding directions in the reciprocal space.

However, whether it’s for the epitaxial growth of a deformation layer or a relaxation layer, the measurement of relaxation is very important. It is only by using Reciprocal Space Mapping (RSM) analysis technology that the relaxation degree of heteroepitaxial thin films can be accurately identified (Source: Science and Technology News Vol. 29, No. 1, 96.8_ Analysis of Strain in Silicon Germanium Heteroepitaxial Materials Using X-ray Reciprocal Space Mapping). Figure 15 shows a SiGe buffer layer with a concentration gradient grown on a Si substrate. The Ge concentration of the buffer layer increases from bottom to top. This buffer allows the growth of a Si_0.5 Ge_0.5 high strain layer with high crystal quality. The critical thickness is about 50 nm. When the thickness is ~20 nm greater than the critical thickness, the Si_0.5 Ge_0.5 layer on the surface will begin to release strain.

Figure 15. RSM analysis shows that the critical thickness of the surface SiGe strain layer is about 50nm. As the surface SiGe thickness increases, the strain relief effect also increases. (Source: J Mater Sci: Mater Electron 30, 14130–14135 (2019). A novel three-layer graded SiGe strain relaxed buffer for the high crystal quality and strained Si0.5Ge0.5 layer epitaxial grown)

In addition to observing the distribution of diffraction peaks in the inverted space, RSM spectra can also be analyzed using software to obtain the SiGe composition ratio of each layer and the degree of relaxation between each layer and the Si substrate, as shown in Figure 16.

Figure 16. RSM analysis shows that the closer the SiGe layer is to the Si substrate, the smaller the lattice mismatch, and the farther the SiGe is from the Si substrate, the greater the lattice mismatch. This shows that there is SiGe relaxation and strain relief. (Source: Taiwan Semiconductor Research Center)

As 3D packaging develops, the complex stacking of materials and the numerous processes required mean that it is easy for internal stress to be generated between layers due to differences in film thickness or materials’ physical properties, such as thermal expansion coefficients, densities, and lattice spacing, etc.. After undergoing multiple subsequent processes, such as CMP polishing, it is very likely that local stress concentration areas will experience peeling or cracking, which can cause product failure. As a result, the semiconductor industry has become increasingly aware in recent years of the value of film residual stress analysis.

When a film is squeezed or stretched, the lattice spacing of the material will change. That change can be determined by measuring the shift in the diffraction peak angle (Δ2θ) using XRD or GIXRD (Figure 17). The residual strain of the film can then be calculated according to the solid elastic theory (the method or the method). Then the residual stress can be calculated by entering the material’s Poisson's ratio and Young's modulus (Figure 18).

Figure 17. Schematic Diagram of Lattice Spacing Changes in Various Directions When Compressed or Stretched

Figure 18 shows a GIXRD analysis of Cu thin film. As the  angle changed, it was found that the diffraction peak moved gradually to a lower angle. This means that the lattice spacing has changed, which in turn indicates that there is residual stress in the film that has caused the same lattice plane to exhibit different changes in different directions.

Figure 18. Residual Stress Analysis of Polycrystalline Cu Film; The Δ2θ was obtained using Grazing Incidence Diffraction; According to elastic theory, the residual stress of the film was calculated to be 0.567 GPa.

Electron microscopes and X-ray diffraction are two of the tools most commonly used in the analysis of material microstructures. Electron microscopes are able to observe the fine structures within a specific area, but they require special sample preparation and a vacuum environment. XRD analysis, on the other hand, can be conducted in normal atmosphere and does not require special sample preparation, making it more convenient. In addition, XRD is a form of nondestructive analysis which can be used for analysis in the sample’s normal environment. When combined with the average results of large area analysis, it enables the identification of a material’s overall characteristics.

XRD technology can be used to analyze the crystal phase, crystal orientation, crystallinity, and grain size of materials as well as the preferred orientations of coatings and the residual stress in polycrystalline films. When used in conjunction with high-resolution accessories, it can also perform coating/epitaxial quality analysis and heteroepitaxial thin film composition ratio, thickness, lattice matching and relaxation analysis. Furthermore, the use of total reflection technology enables the analysis of multilayer film thickness, surface interface roughness and material density, etc.. In summary, XRD has a wide range of applications and the capability to provide helpful technical solutions for various advanced process and materials research and development needs.