Experiments in patterning of thin films of polycrystalline bismuth telluride (Bi2Te3) on silicon substrates have been performed. Bi2Te3 is representative of a family of thermoelectric materials that are well suited for use in extracting electric energy from thermal gradients associated with flows of waste heat at temperatures in the approximate range of 0 to 150°C. Techniques and processes for fabricating macroscopic thermoelectric devices from bulk thermoelectric materials are mature and well known, but the same cannot yet be said concerning the fabrication of microscopic thermoelectric devices. The experiments reported here were performed as part of a continuing effort to develop capabilities for fabrication (including mass production) of microscopic thermoelectric devices, with a view toward eventually enabling the incorporation of them as integral parts of micro-electromechanical systems (MEMS) that could also include heat exchangers, sensors, actuators, and/or flow channels. Thus, the development of microscopic thermoelectric devices could benefit from the established industrial infrastructure for manufacturing MEMS and other silicon-based microsystems.

Delamination of Bi2Te3 from underlying bare SiO2 was the primary adverse effect observed in patterning experiments. Such delamination is of little concern because in most foreseeable applications, Bi2Te3 would be required to remain only in patterned areas on metal pads, to which Bi2Te3 adheres.
The samples used in the experiments started as 4-in. (10.16-cm) <100> silicon wafers. A 1-μm-thick thermal oxide was grown on each wafer so that the thermoelectric devices intended to be fabricated subsequently would be electrically insulated from the silicon. Metal pads intended to serve as bottom-side electrical interconnections between thermoelectric posts were formed on the oxide by use of a standard lithography-and-liftoff process. After cleaning, blanket undoped Bi2Te3 films were grown by use of solid-source molecular beam epitaxy (MBE). Films as thick as 9 μm were formed, and it appears that greater thicknesses could be achieved by increasing time in the MBE reactor. After growth of the Bi2Te3 films, each wafer was coated with a photoresist to protect the Bi2Te3 in the next step, in which the wafer was cut into 15-by-15-mm dies. The photoresist was then removed by use of a commercial photoresist stripper.

The approach followed in the experiments involved a standard procedure of (1) creating the appropriate patterns by photolithography, then (2) implementing the patterns by use, variously, of wet chemical etching or dry plasma etching to remove the Bi2Te3 from the areas outside the patterns. Established recipes for photolithography and for wet etching and dry plasma etching were used, with suitable modifications to adapt them to the materials and geometry of interest.

The primary adverse effect encountered in the experiments was delamination of Bi2Te3 during the development step of the photolithographic process. As shown in the figure, the Bi2Te3 adhered on and around areas containing the metal pads, but in some open areas in which the Bi2Te3 rested on bare SiO2, the Bi2Te3 was removed. It has been hypothesized that this delamination could be a result of lack of chemical bonding between the Bi2Te3 and the SiO2 surface. However, such delamination is of little concern inasmuch as in most foreseeable applications, metal pads (to which Bi2Te3 adheres, as was shown) must be included under Bi2Te3 posts to provide electrical connections.

This work was done by Brian Morgan and Patrick Taylor of the Army Research Laboratory.