9 Experimental Setup

To experimentally characterize BTI and the single traps responsible for the degradation of the device performance, sophisticated measurement tools are necessary. In recent investigations BTI has been studied in a large variety of devices, like (i) conventional or MOSFETs, (ii) high-k MOSFETs, (iii) GaN metal-oxide-semiconductor high-electron-mobility transistors (MOSHEMTs) or (iv) more exotic transistors using 2D materials. Apparently, numerous general purpose instruments are available, but most of them are either very costly or do not cover the large list of requirements. Furthermore, since these measurements typically drive the equipment toward their limits, unexpected behavior is often observed, like undocumented settling periods at zero volts for transitions from positive to negative voltages. This undocumented behavior was observed regularly in our previous attempts and is also regularly reported by our partners. Furthermore, for these setups a range of different tools are required which have to communicate seamlessly. Particularly for long-term experiments it was regularly observed that setups using off-the-shell equipment occasionally produced communication errors or crashed for other reasons. These failures were virtually impossible to track down due to the lacking documentation of the equipment internals. In order to maintain complete control over the delicate measurements setup, an elaborate measurement setup to study single defects in nanoscale MOSFETs has been developed during the previous years. In the following, the requirements for BTI characterization and the basic measurement concepts are discussed, followed by a detailed presentation of the developed equipment.

9.1 Requirements for Time-Dependent Defect Spectroscopy Measurements

To collect experimental data for BTI analysis, very fast measurement methods are necessary because recovery starts immediately after stress, is very fast, and continues for very long times. Furthermore, the threshold voltage shift has to be monitored with a high resolution in time and voltage. While the high time resolution is required to properly resolve RTN signals, the high resolution is required to study very small threshold voltage shifts below . Such shifts typically correspond to shifts in the drain-source current of less than . The high resolution is particularly important because only a few defects are beneficially aligned (from an experimental perspective) to the conducting channel to produce large shifts. Most shifts are close to the measurement limit of conventional setups.

Common ways to measure a drift are the (i) OTF method, the (ii) fast method or the (iii) fast method [86], see Section 4.7. The OTF method allows to monitor the threshold voltage shift during device stress without any interruption. Thus defects with a capture time smaller than the measurement delay, that is , are not accessible by OTF methods, because they require to measure the first-data point, see Figure 9.1 (left).

In contrast to OTF methods, the capture time window is limited by the stress time , that means , and the emission time window is limited by the measurement delay and the recovery time leading to , see Figure 9.1 (right). The measurement delay is thereby defined by the measurement equipment and should be held as short as possible. This can be particularly difficult when studying single charge trapping in nanoscale transistors as this requires a current resolution of several tens of pico-amperes. Furthermore, as will be demonstrated later, the measurement delay is related to the noise level of the recovery traces because the larger the sampling frequency is, the larger the noise level of the sampled signal gets. As a consequence, the minimum detectable step height for single trap BTI investigations increases. In contrast to the measurement delay, there is no theoretical limit for the maximum recovery time.

The fast and the fast method are based on the MSM principle where the device is repeatedly stressed and monitoring of the device recovery is started immediately after stress release, see Figure 9.2.

A feedback loop of an OPAMP is used to control the gate voltage at the fast method, see Figure 4.19. In contrast, the fast method relies on the direct measurement of the drain-source current and the gate current . Afterwards, the drain-source current is converted into a threshold voltage shift using the characteristics of the device recorded prior to the stress/recovery measurements. It has to be noted that the selection of the correct characteristics for the extraction of is of utmost importance, since each transistor accumulates a long term degradation through all measurement cycles and the sub-threshold slope may change during BTI stress. If is converted to using an recorded immediately before each MSM sequence, the remaining degradation from the previous recovery cycle of the total is lost. Conversely, if the conversion of the to is based on an initial changes of the subthreshold slope and the threshold voltage will not be transferred to the . In summary, there are several ways to calculate the whereas the choice of the reference depends on the experimental questions.

In both MSM concepts the measurement delay is primarily defined by the switching transient between stress and recovery phases and has to be held as short as possible. Furthermore, a detailed knowledge of the transient behavior is required to perform accurate device simulations. As previously mentioned, using a general purpose instrument we noticed that some additional delay at the output voltage of was introduced when directly switching from a positive output voltage to a negative output voltage and vice versa. Such additional delays have to be avoided during all measurement routines.