SiC MOSFETs: Scrutinising the gate atom by atom
Atom probe tomography unveils the composition and the distribution of key elements in the vicinity of the interface SiO2 and SiC.
BY LAKSHMI KANTA BERA, SHIV. KUMAR, NAVAB SINGH, UMESH CHAND, ABDUL HANNAN BIN IBRAHIM ABDULLAH YEO, VOO QIN GUI ROTH AND SURASIT. CHUNG FROM A*STAR AND PIERRE-YVES CORRE FROM CAMECA INSTRUMENTS
Over the last decade, much effort has been devoted to the development of the SiC MOSFET. This is paying off, with sales soaring, driven by the uptake of this device in electric vehicles.
The success of the SiC MOSFET comes from its key attributes. One great strength is its wide bandgap, enabling high-temperature operation and a trimming of energy loss during device operation. Additional assets include its high thermal conductivity, aiding efficient heat dissipation and helping maintain device performance and longevity; and a simple integration flow, following in the footsteps of silicon technology, that involves the growth of SiO2, using thermal oxidation.
However, the SiC MOSFET has an Achilles heel. In stark contrast to the interface of silicon and its native oxide, SiO2, that grown on SiC is riddled with a high interface-state density. The numerous imperfections significantly reduce carrier mobility in the MOS channel and threaten device reliability.
Due to these significant concerns, much effort has been devoted to understanding the origins of the high density of interface states. This has been attributed to carbon-related defects, including C-clustered, C-interstitial, and C-vacancies. Within the scientific community, it is widely agreed that these defects play a significant role in influencing the density of interface states at the SiC/SiO2 interface.
Figure 1.The experimental setup for atom probe tomography.
Many studies, such as those investigating native point defects and carbon clusters in 4H-SiC, as well as the structure and energetics of carbon-related defects in SiC, have provided much insight into the impact of these carbon-related defects on material properties. However, despite many years of investigation by several research groups, the distribution of different elements and compounds in bulk SiO2 and at the SiO2/SiC interface is yet to be completely understood. To a large extent, this comes down to a lack of information at the atomic scale. There is a need to know the identity and the precise location of different species.
Two common techniques for materials characterisation that fall short in this regard are X-ray photoelectron spectroscopy and Auger electron spectroscopy. Both these forms of spectroscopy are hampered by a poor lateral resolution in the micrometre-range, as well as a relatively poor detection limit – it is around 0.1 atomic percent – that leads to a low detection sensitivity, and hampers complex quantification of chemical analysis.
Another widely used technique within the compound semiconductor community, secondary ion mass spectrometry, excels in detecting trace elements down to parts-per-billion levels. However, its sub-micrometre lateral resolution limits atomic-scale analysis. In addition, quantification of results is challenging, due to the dependency on ionization energy for secondary ion yield.
A versatile tool that provides detailed investigations at the sub-nanometre level is high-resolution transmission electron microscopy. Advances in one form of this, high-angle annular dark-field scanning transmission electron microscopy, have enabled, in special cases, three-dimensional characterisation at the atomic scale.
Figure 2. Schematic cross section of sample for interest of investigation of several elements and their distributions.
However, the ultimate characterisation technique for studying materials at the atomic level is atom probe tomography. This incredibly powerful technique, which our team from the Agency for Science, Technology and Research, Singapore, is using to scrutinise the interfaces of SiO2 and SiC, can’t be beaten when it comes to discovering the position of individual atoms in materials with nearly atomic resolution. One of its key merits is equal sensitivity for all elements.
Basic principles
Atom probe tomography operates on the principle of field evaporation, with surface atoms extracted from specimens with an electric field. This process involves directing a laser at the sample, alongside simultaneous high-voltage pulsing, conditions that cause surface atoms to evaporate with near 100 percent ionisation. The evaporated atoms are projected onto a position-sensitive detector for analysis.
To maximise the effectiveness of this technique, there is a need to prepare sharp needle-shaped specimens with a radius of 50-100 nm. Such samples ensure precise analysis through field evaporation of surface atoms. The chemical nature of the atoms is identified with time-of-flight mass spectrometry (see Figure 1 for an overview of this experimental setup).
After loading samples into an atom probe tomography analysis chamber, specimens are cooled to around 50 K and maintained in an ultra-high vacuum, typically 10-11 Torr. Applying a DC voltage of typically between 2 kV and 10 kV results in a
high electric field at the surface of the tip. The field at the tip’s apex is proportional to the applied voltage, and inversely proportional to its radius of curvature.
For electrically conductive materials, atom probe tomography involves the application of voltage pulses to the tip to generate an electrical field that induces field ionization. For less conductive samples a different approach is needed, with laser pulses ensuring the thermal evaporation of ions. The application of laser pulsing is especially advantageous in higher resistivity materials, such as semiconductors, where high-voltage pulses alone may be insufficient for promoting field ionisation.
The detection of ions that are liberated from the sample is correlated to the high-voltage and laser pulses. Ions are converted into a mass-to-charge-state ratio, measured in a unit known as the Dalton. This conversion is crucial for chemical identification of each ion, which is identified by the relationship between its time-of-flight and its mass-to-charge-state ratio.