UC Santa Cruz

Performance, efficiency, and reliability of modern high power, high speed microelectronics and nanoscale structures are strongly influenced by device self-heating on the micron and nanometer scale. Local temperature "hotspots" on a chip indicate areas of high power density and possible failure locations. This dissertation presents results of case studies using high resolution thermoreflectance imaging microscopy to characterize self-heating in high power silicon transistor arrays, gallium nitride high electron mobility power transistors (HEMTs), and nanoscale percolation network devices. Thermoreflectance imaging microscopy was performed with submicron spatial resolution, 800 ps time resolution, and 50 mK temperature resolution.

Thermoreflectance imaging of silicon transistor arrays revealed self-heating nonuniformity greater than a factor of two due to electrical debiasing. At low bias (12 mA/mm), temperature distribution matched state of the art electrothermal simulation. At high bias (47 mA/mm), however, measured self-heating diverged significantly from simulation despite inclusion of temperature dependent material properties in the thermal model. Hotspots with temperature change of 70 K were observed. Results demonstrate nonlinear dependence of current distribution on bias for complex arrayed power devices.

Thermoreflectance imaging of GaN HEMTs revealed for the first time temperature gradients as large at 60 K and 80 K between critical gate, channel, and contact features within the first few microseconds of pulsed excitation at 19 W/mm. Fast transient self-heating gradients in the HEMT were confirmed with 5 ns time resolved thermoreflectance images.

Thermoreflectance imaging revealed nonuniform self-heating in nanoscale percolation network devices consisting of disordered 90 nm diameter silver nanowires. Microscopic hot-spots at selected nanowire-nanowire junctions in the network exhibited nonlinear thermal properties, with temperature dependence of current exceeding the prediction of Joule self-heating for bulk materials. Results encourage a fundamental reevaluation of the transport models and characterization results for network based percolating conductors.

The optical, non-contact thermoreflectance imaging microscopy method quickly measures self-heating distribution on the surface of integrated structures with 50 mK temperature resolution. Transient capable implementations with 50 ns and 800 ps temporal resolution were used to image fast pulsed transient temperature rise and fall in high frequency power integrated circuits. The method is capable of submicron spatial resolution, enabling temperature measurement of smaller device features than can be achieved with similar thermal imaging techniques, such as infrared thermal microscopy. Self-heating in silver nanowire network devices was imaged with 300 nm spatial resolution.