(image) (image) [Previous] [Next]

Degradation of Electrical Parameters of Power Semiconductor Devices – Process Influences and Modeling

4.5 Experimental pitfalls

To reliably use the poly-heater a number of experimental issues have to be considered. A few of those are addressed in the following.

In order to reach high device temperatures the overall resistance of the poly-heater must be matched to the output characteristic of the SMU for maximum power transfer. For the particular high power SMU (HPSMU) of the Agilent Technologies B1500A parameter analyzer used in the present thesis, the resistance of the poly-heater needed to be designed to be around 20 Ω or 80 Ω in order to transfer the maximum power of 20 W as illustrated in Fig. 4.15.


Fig. 4.15: Output range of the HPSMU used in this thesis. Only for load resistances of 20 Ω or 80 Ω the maximum power of 20 W can be reached. A minimum of 5 W can be supplied for most load resistances.

Another problem can occur if a switching matrix is used to connect the DUT and the poly-heater to the parameter analyzer. As shown in Fig. 4.16, an incorrect connection can cause a ground shift for the potential of the DUT which leads to changes of the device behavior which are not due to temperature.


Fig. 4.16: If both the poly-heater and the source and bulk connection of the DUT are connected via a switching matrix to a ground potential a ground shift can occur. In this particular example, the drain-source voltage of the transistor varies between 2.5 V and 2 V depending on the current flowing through the heater.

Such a connection mistake can be exposed by inversion of the supply polarity of the poly-heater. If this changes the device characteristics a ground shift problem is a likely cause. It is therefore most convenient to use a secondary SMU as a separate virtual ground potential for the device.

Another possible issue regarding the use of the poly-heater is the occurrence of thermally induced strain. However, there are three major arguments that thermally induced strain does not impact the result of the measurements for poly-heater use.

  • • In reference [Lea+12], influences of mechanically stress on transistor performance are discussed. The work mentions stress values of up to \( \sigma = \SI {500}{\mega \pascal } \). By using Young’s modulus of Si of about \( k=\SI {150}{\giga \pascal } \) [BL97; HNK10], the strain in the devices of reference [Lea+12] can be estimated as

    (4.10) \begin{equation} \varepsilon = \frac {1}{k}\sigma = \frac {1}{\SI {1.5e11}{\pascal }} \SI {5e8}{\pascal } \approx 3 \times 10^{-3}.                     \end{equation}

    For comparison, the temperature induced strain in our devices for a deliberately overestimated temperature gradient along the device interface of 10 °C can be estimated from the thermal expansion coefficient of Si \( \alpha = \SI {3e-6}{\per \celsius } \) [OT84] to be about

    (4.11) \begin{equation} \varepsilon = \alpha \times \Delta T = \SI {3e-6}{\per \celsius } \times \SI {10}{\celsius } = 3 \times 10^{-5}.                         \end{equation}

    From this follows that the strain investigated in reference [Lea+12] is about two decades larger than what can be expected for our test device. Still, the changes in the saturation drain current in reference [Lea+12] are around 15 % for 500 MPa, i.e. for typical poly-heater devices the drain current may change less than 0.3 %, which is too small for any significant impact.

  • • In order to check for a potential influence of thermally induced strain on the transistor parameters, test structures with or without a several micrometer thick metal plate on top of the device and heater were constructed. As part of the structure, the metal stack is efficiently heated with the poly-heater because the poly resides between the device and the metal. Metals have usually a much larger thermal expansion coefficient than Si or SiO2 . A device with a metal plate on top should experience different virgin characteristics or degradation behavior with or without a metal plate on top if thermally induced strain is occurring. However, up to now no impact of the metal plate on the device virgin characteristics or degradation behavior [SPN12] has been observed, even with the use of the poly-heater.

  • • To further check for a possible occurrence of parasitic effects of the poly-heater, NBTI experiments were performed where the stress temperature was either provided by using the poly-heater or by using the thermal chuck, see Chapter 5 for details. Both approaches give identical results of the transistor characteristics and their degradation behavior [PG13b]. This strongly indicates that the poly-heater only impacts the device temperature.

For the use of the poly-heater for time critical experiments care has to be taken that the thermal capacitances of the materials surrounding the heater prevent instantaneous temperature switches. However, for the poly-heater devices used in this thesis, a switch of the temperature from e.g. −60 °C to 210 °C occurs in less than 1 s, as shown in Fig. 4.17.


Fig. 4.17: Heating dynamics of the poly-heater [Pob+11b]. The device follows a switch of the power supply to the poly-heater with a small delay because of the finite thermal capacitances of the materials which surround the device and the heater.

Typically, the temperature switch is mostly completed after about 1 ms and the difference to the target temperature becomes unresolvable after a maximum of 10 s [Aic+10c; Pob+11b]. For NBTI experiments the poly-heater is usually used to provide a high stress temperature while recovery is measured at a much lower temperature to prevent undesirable recovery. There, it is especially important to terminate the stress by first switching off the poly-heater and, subsequently, reducing the gate bias. This approach is called degradation quenching [ANG08; Aic10].