An international team of researchers has investigated the industrially important atomic layer deposition procedure for deposition of hafnium oxide on indium arsenide. Using ambient pressure X-ray photoelectron spectroscopy, the researchers could follow the oxide formation under industry-relevant conditions. The results of the research play a significant role in the future development of low-power and high-speed electronics.
New materials for faster electronics
An essential component of everyday electronics is transistors. One type of transistor often used is the so-called metal–oxide–semiconductor field-effect transistor or MOSFET for short.
“The MOSFET can be described as a switch. The resistance between source and drain, also called terminals, is controlled by the gate via the gate oxide,” says Giulio D’Acunto from Lund University, the first author of the study.
Today, most transistors are made from the semiconducting material silicon. An important advantage of silicon is that its native oxide, silicon dioxide, forms on its surface with an almost ideal semiconductor-oxide interface, which is important for transistors. However, semiconductors made of materials such as InAs, elements from columns III and V of the periodic table, beat silicon on attractive properties for future high-speed and low-power electronics. They also offer a variety of band gaps. The challenge is that the native oxides on III-V semiconductors, rich of defects, affect the device performance.
“High demands for even faster and smaller devices are more and more difficult to fulfil by the widely used silicon-based devices, and improved materials are needed,” continues D’Acunto. “A combination of hafnium oxide and the III-V semiconductor indium arsenide, which has extremely high electron mobility, would be a perfect candidate to overcome the silicon era. We just need to establish a low-defect interface to the hafnium oxide of the gate.”
Optimising the method
Atomic layer deposition, ALD, is the method used by industry to grow highly uniform oxide materials. In the case of hafnium oxide on indium arsenide, the indium arsenide surface also conveniently gets cleaned of its native oxide by the reaction process. Hafnium oxide is a so-called high-κ oxide. This means that it has extraordinary good insulating properties, which are needed for miniaturized electronic devices. However, the surface cleaning needs to be further improved, as even small remains of native oxide are still limiting the device performance.
“The fundamental knowledge of a process that works kind of and is widely used is of utter importance. Nowadays, the understanding of ALD is based on a theory of an ideal chemical reaction, which may be correctly understood, or not,” explains D’Acunto. “Little is known about the actual reaction mechanism, surface species and chemistry during the ALD process. Understanding the reaction in our study can give insight into how to improve further the interface between III-V semiconductors and high-κ hafnium oxide.”
The ambient pressure X-ray photoelectron spectroscopy method allows for studying the bonds between the atoms in the sample also at elevated gas pressures. Using this method, the researchers could directly investigate what reactions were happening at the indium arsenide’s surface as they deposited a hafnium metalorganic precursor, called TDMA-Hf, to form a hafnium oxide layer.
Unexpectedly. the hafnium oxide is not formed immediately, but through a two-step reaction during the first-half cycle of atomic layer deposition. Here, the native oxide on the semiconductor surface acts as a supply of oxygen. The result is a transition from a rather thick, mixed indium-arsenic-oxide to a layer of hafnium oxide with an atomically thin indium-oxygen-hafnium interface. The high time resolution of the experiment did also help with understanding the dynamics behind the reaction.
Towards a better industrial process
“With our study, we further extended our potential to use ambient pressure X-ray photoelectron spectroscopy to monitor the ALD surface chemical reaction in real-time,” says D’Acunto. “On the other hand, and more interestingly, we extended the ALD model by proving that the reaction actually takes place in more steps than those described in the standard ALD model. We demonstrate a strong correlation between the formation of hafnium dioxide and the reduction of the unwanted native oxide.”
The commercial atomic layer deposition process used by industry is not as effective as the highly controlled ones seen in the present study. The study may, therefore give valuable hints on how to make the industrial process more effective.
“We demonstrate a correlation between the formation of hafnium dioxide and the reduction of the indium arsenide native oxide. Our results indicate that a sufficiently long first pulse of metal precursors guarantee a complete reduction of the unwanted indium arsenide native oxide,” D’Acunto concludes. “At the same time, we form, independently from the initial condition, a single monolayer of hafnium dioxide. Implementing a sufficiently long first pulse also in commercial ALD setups has the potential to improve the performance of indium arsenide-based electronics.”
Giulio D’Acunto, Andrea Troian, Esko Kokkonen, Foqia Rehman, Yen-Po Liu, Sofie Yngman, Zhihua Yong, Sarah R. McKibbin, Tamires Gallo, Erik Lind, Joachim Schnadt, and Rainer Timm, Atomic Layer Deposition of Hafnium Oxide on InAs: Insight from Time-Resolved in Situ Studies, ACS Appl. Electron. Mater. 2, 3915 (2020), DOI: 10.1021/acsaelm.0c00775
All illustrations by Giulio D’Acunto, Lund University