Change in temperature affects the speed, power and reliability of electronic components and systems. Variation of temperature affects the speed performance, because material characteristics depend on temperature. These dependencies may be normal or reversed based on the type of the semiconductor material. Additionally, these dependencies change with technology scaling, and manufacturers counteract by introducing new processing materials, using metal gates and high-K dielectrics.
For example, temperature influences various performance functions in a MOSFET. These include the carrier density, energy band gap, carrier diffusion, mobility, current density, velocity saturation, leakage current, threshold voltage, electro-migration and interconnect resistance.
Temperature dependence of carrier density for a doped material occurs in three distinct regions. The material has just enough latent energy in the ionization region to push a few of the dopant carriers into the conduction band. When the material is in the extrinsic region, which is the desired region of operation, the carrier concentration remains flat over a wide range of temperatures.
This region has all the dopant carriers energized into the conduction band, and there is minimum generation of additional thermal carriers. However, as the temperature increases, the extrinsic region converts into the intrinsic region, with the number of thermally generated carriers exceeding the number of donor carriers. Typically, the intrinsic carrier concentration in a material is generally much smaller than the concentration of dopant carriers at room temperatures. However, intrinsic carrier concentration is highly temperature dependent and once the number of thermally generated carriers exceeds the number of dopant-generated carriers, the potential for thermal variation problems increases substantially.
At low temperatures, lattice vibrations in the material are small and electrons move more slowly. Thus, ion impurity forces dominate the limit to mobility. As temperature decreases, it takes less time for an electron to pass an impurity ion, which means the mobility decreases. The reverse is true when temperature rises; the carrier’s thermal velocity increases, consequently decreasing the impact of interface charges.
With an increase in temperature, the kinetic energy of particles within the material also increases, effectively increasing the diffusion component of the total current. Two parameters, mobility and carrier density affect the total current through the material. While the carrier density remains nearly fixed with temperature over the extrinsic range or the intended range of operation, the mobility term or the drift component of the total current actually decreases with an increase in temperature.
Since the temperature dependencies of diffusion and drift currents are of opposing nature, the net current change depends on the applied electric field and affects the threshold voltage and leakage current of the MOSFET. Manufacturers typically design the MOSFET such that its threshold voltage decreases linearly with increasing temperature. However, the leakage current doubles for every 10°C rise on temperature.
The resulting change in device current based on temperature can have devastating effects leading to timing failures, systems exceeding power or energy budgets and errors in communication between cores. This is more commonly known as reverse temperature dependence, which is the increase of electrical conduction with increase in temperature, first discovered by C. Park of Motorola, in 1995.