Device-scale AFM Thermoreflectance metrology: Analytics, Probe design, machine Learning, User experience, and Standards (DATAPLUS)

Laser Thermal Analysis, Inc. aims to build a hybrid Atomic Force Microscopy-thermoreflectance (AFM-TR) instrument to generate maps of the thermal resistance, thermal boundary interface resistance, and temperature profiles of microprocessors and wide bandgap semiconductor materials and devices with < 100 nm spatial resolution, < 0.5 ºC temperature resolution, < 10 % measurement uncertainty, and capability of characterizing materials without the need for an applied transducer coating. The dramatic enhancement in resolution provided by AFM-TR comes with some additional experimental considerations to overcome. AFM probes used may have a limited operational lifetime from wear with the surface. Laser Thermal aims to reduce these functional and cost related challenges that currently exist by utilizing novel coatings and tip geometries. This includes utilizing artificial intelligence/machine learning (AI/ML) solutions that offer powerful gains in data throughput and insights through intelligent targeting of high-value material regions. Because AFM-based metrologies have long been considered slow, with measurement times ranging from minutes to hours, this presents an additional economic motivation to efficiently collect data, via intelligent and strategic acquisition methods. These AI/ML concepts can be further leveraged to identify when tips have degraded, allowing for users to quickly understand when data may be compromised and even providing paths toward being able to correct flawed data post facto. These efforts combined will bring new levels of throughput and functionality to AFM-TR as a thermal measurement tool to provide new and useful data to researchers and modelers.