Modeling and control of a silicon substrate heater for carbon nanotube growth experiments

Abstract

The precision engineering research group at MIT is working on carbon nanotube growth experiments on silicon substrates and in microfabricated silicon devices, to try to produce improved bulk nanotube growth. For this thesis, a heating control system was designed and implemented for eventual use in CNT growth experiments. The computer program that controls the heater is user-adaptable, so that an experimenter can easily change the desired temperatures at various points of the process. Later, this heating system will become part of a much larger system that also incorporates a controlled flow rate. The goal of the system is to achieve high-bandwidth control of reaction conditions. In the heating control system designed, a computer controls a power supply attached to a wire-wrapped silicon chip, which is used to heat up the system, and the temperature is measured by a thermocouple. The control algorithm uses proportional gain, and the output is a PWM voltage. For accurate control of the system, a goal was set out to achieve an error of within 10%. For gains above 5, the computer can accurately control the temperature to less than 5.5% of the desired values in steady state, and an error of 0.75% was achieved with a gain of 50.

(cont.) Thus the system meets the desired specification of error. Also, while the error drops dramatically with increasing gain, the overshoot increases much more slowly, making a higher gain desirable. Also, the system still has only reached temperatures of 650 degrees Celsius, although temperatures of 1000 degrees Celsius are required for nanotube growth. In order to achieve this, further tests will be performed with thicker wire and more voltage. Also, contact resistances within the chromel decrease with increasing temperatures, which reduce the percentage of power dissipated in the chromel compared to the lead wires. If the system is modified to eliminate this effect, by wrapping the wire differently or by using doped silicon, higher temperatures can be achieved. This will also make the system more predictable, leading to a better model and better control. Finally, to improve overall performance, one can experiment with changes to the switching time, using a PI or PID controller, and active cooling.