This chapter presents a study of thermoelectric energy harvesting with nano-sized thermopiles (nTE) in a planar 65 nm silicon CMOS process. These devices generated power from a 5C temperature difference at a density comparable to commercially available thermoelectric generators, following a metric used in the research literature (Hudak, 2008). By analyzing these devices as a thermoelectric harvesting system, the authors explore the impact of additional performance metrics such as heat source/sink thermal impedance, available heat flow density, and voltage stacking, providing a more comprehensive set of criteria for evaluating the suitability of a thermal harvesting technology. The authors use their thermoelectric system theory to consider the prospects for several thermoelectric energy harvesting applications.
TopIntroduction
A vast quantity of waste heat is rejected by the heat engines used in industry, transportation, and buildings. The idea of converting some of this heat flow into electricity is compelling. Consider the internal combustion engine powering an automobile. Tens of kW of power is extracted from fuel at an efficiency of less than 40%, meaning that at least 60% of the fuel’s chemical energy would be rejected to the environment as heat. If half of this waste heat could be coupled through a thermoelectric heat engine operating at half the Carnot efficiency, capturing half of the total temperature drop (say 100C), then electrical power equal to (1/2) x (60%) x (1/2) x (100C/500C) ~ 3% of the chemical energy of the fuel being consumed, potentially improving fuel economy by 10%.
This line of reasoning sparks interest in thermoelectric energy harvesting; but one quickly learns that the present state-of-the-art thermoelectric devices are significantly less efficient than assumed above, causing interest in large scale energy thermoelectric harvesting to fade.
However, there are some applications for which the cost of fuel is not the primary driver so efficiency is less important than the ability to achieve a certain power budget in a given usage environment. In situations such as wireless sensor nodes in the Internet of Things (IoT), there is a need for harvested power to avoid the cost of battery replacement. Wearable devices powered by body heat are also of interest, particularly if the harvested heat can extend usage lifetime or enable increased functionality by supplementing battery power.
In both of these applications, more power can be provided when thermoelectric efficiency is improved. So it is still an important topic of materials research to find thermoelectric materials and devices that support higher efficiency.
Recent Developments in Thermoelectrics
In the early 1990s, it was proposed (Hicks, 1993) that by creating layered materials, thermoelectric efficiency could be improved. At the same time, new concepts in thermoelectric devices were introduced (Edwards, 1993; Nahum, 1994), some of which could in principle approach the Carnot limit (Edwards, 1995) under special circumstances.
More recently, nanostructuring has been found to improve silicon thermoelectric efficiency (Boukai, 2008; Hochbaum, 2008) by enhancing acoustic phonon scattering from interfaces. Based on this hypothesis and its compatibility with increased density (hence lower device cost), we developed nano-scale thermoelectric devices (nTE) in a 65 nm planar CMOS process in which silicon ridges as small as 80 nm are fabricated. Because these Si ridges are smaller than the room temperature acoustic phonon mean free path in silicon, it is expected that thermoelectric generating performance can be improved relative to bulk Si.
Material improvements are necessary, but unfortunately not sufficient. As we will see below, it also is important to understand how heat flows through a thermoelectric energy harvesting system and what constraints the heat flow places on the efficiency, size, and other aspects of thermoelectric devices and the thermal system built around them. So a primary objective of this article is to develop a theory of a thermoelectric energy harvesting system. Because nTE technology is an extreme example of device miniaturization, it is a good test case for a theory of a thermoelectric energy harvesting system.