Skip to main content

Nanoengineering of self-forming diffusion barriers for interconnect technologies

Final Report Summary - NANOS.BIT (Nanoengineering of self-forming diffusion barriers for interconnect technologies)

At present, copper is the material of choice in the current processing technology for advanced semiconductor device interconnects. However, the high diffusion rate between copper and silicon or silicon oxides requires the development of physical barriers to prevent interdiffusion across the interfaces. The thickness of the currently used barriers makes them an unviable option as the semiconductor industry moves from the 45 nm node to 32 nm and beyond, and alternative approaches are required. Some groups have proposed the use of the so called self-forming barriers, which have the potential to overcome the shortcomings of the current approaches. Self-forming Cu (Mn) diffusion barrier layers, recently proposed by a number of groups, have the potential to fulfil these requirements. The formation process involves the deposition of this copper alloy directly onto the SiO2 layer. After annealing, the alloying metal segregates towards the Cu / SiO2 interface and chemically reacts with the insulator, forming a thin barrier layer.

The main objectives for the whole duration of the project were the characterisation and understanding of the formation process of self-forming diffusion barriers layers for transistor interconnects, and the optimisation of the creation process for future generations of device technology. For this, the work of the fellow consists on the development and application of a methodology of analysis based on electron beam related techniques in order to study these structures and interfaces at the atomic scale.

This work has been carried out in the framework of a complete and complementary network of international collaborations. Strong links have been established and consolidated with Dublin City University in Ireland for the fabrication, chemical characterisation and kinetical studies using X-ray photoemission spectroscopy (XPS) of the self-forming diffusion barriers; and with Intel Ireland as industrial mentor.

During this time, I have acquired a high level of expertise and total independence in the use of electron beam related techniques, such as conventional (CTEM) and high-resolution transmission electron microscopy (HRTEM), high-resolution scanning-transmission electron microscopy in high-angle annular dark field mode (HRSTEM-HAADF) in the double aberration corrected machine, energy-dispersive X-ray spectroscopy (EDX) and electron energy loss spectroscopy (EELS). The application of these techniques to the structural and chemical characterisation of the self-formed diffusion barriers in the Cu / Mn system, complemented with the XPS analysis and the feedback with the industry, has led to very relevant results. In a first step, thick Cu / Mn layers were deposited on SiO2 substrates to investigate separately the interaction between Cu and Mn, and between the Mn and the substrate. The results indicate the occurrence of interdiffusion between the Mn and Cu layers, where the Mn tends to diffuse towards the surface of the structure, while the Cu atoms diffuse towards the Mn / SiO2 interface surrounding clusters where higher content of metallic Mn is detected. EELS analysis were used to investigate the chemical state of Mn in the 2-3-nm interfacial layer, showing that it is mainly composed of Mn in +2 and +3 oxidation states. This barrier was also shown to be effective at preventing Cu diffusion into the dielectric layer. The next stage involved the reduction of the thickness of the Mn layer to the minimum achievable while still retaining the diffusion barrier properties. The presence of oxygen turned out to play an important role in the barrier layer formation: a purely metallic Mn film with an approximate thickness of 1 nm cannot be fully converted to Mn silicate following vacuum annealing to 500 degrees of Celsius. TEM analysis suggests the maximum MnSiO3 layer thickness obtainable using metallic Mn is 1.7 nm. In contrast, a 1-nm partially oxidised Mn film can be fully converted to Mn silicate following thermal annealing to 400 degrees of Celsius, forming a MnSiO3 layer with a measured thickness of 2.6 nm. TEM analysis also clearly shows that MnSiO3 growth results in a corresponding reduction in the SiO2 layer thickness. It has also been shown that a fully oxidised Mn oxide thin film can be converted to Mn silicate, in the absence of metallic Mn. It is worth to mention that the thicknesses achieved fulfill the goals established by the international technology roadmap for semiconductors (ITRS) for regarding maximum diffusion barrier thicknesses for the next technology node. The next natural step is the investigation of the barrier formation in Cu(Mn) alloys. For the first time, HAADF, EDX and EELS in STEM mode were used to characterise Cu(Mn) alloys deposited on SiO2 with a sub-nanometric spatial resolution. The deposition process results in a uniform layer of alloy with homogeneous composition and does not involve any interdiffusion or reaction at the interface between the alloy and the dielectric. However, after annealing Mn diffuses to the surface to react with residual O and form MnO, as reported in my previous work and to the interface to react with the SiO2 and form a new layer with lower HAADF contrast. The thickness of this new layer is 2.7 nm, meeting the ITRS targets, and has been formed at the expense of reducing the thickness of the SiO2. EELS composition analysis indicate that this mainly constituted of Mn, O and Si, and no trace of Cu is found in the SiO2 what indicates that it is behaving as an effective barrier. The Mn L23 intensity ratio method was used to investigate the oxidation state of Mn in the diffusion barrier showing that it is purely +2, suggesting that it is almost entirely MnSiO3.

The constant feedback with the collaborators and industry allowed me to redirect the course of the research when necessary to meet the needs and interests of the industry. For this reason, a new research line not included in the original plan has started, as is the investigation of the Ru / Mn as potential new materials for barrier layer formation. The first preliminary results show that it is possible to deposit thin (3 nm) continuous and uniform Ru films. After capping with a 1-nm Mn film and annealing, the Mn diffuses through the grain boundaries and chemically reacts with the SiO2 substrate to form an effective MnSiO3 diffusion barrier layer. This is of great importance especially for the next generation of low-k dielectrics, where the formation of the diffusion barrier is handicapped by their porous structure. The Ru layer, being inert, would seal the pores without interfering with any of the physical or chemical properties of the substrate.

All this work has been summarised in different scientific papers: 2 already published, 2 in press and 2 in preparation; and it has been presented in 5 international conferences. Other ways of disseminations such as presentations in internal seminars, open days and training days have been also carried out.