Charge Trapping and Variability in CMOS Technologiesat Cryogenic Temperatures

7.5 RTN in MoS\( _2 \) MOSFETs

The fact that RTN does not freeze out at cryogenic temperatures can be seen across various channel materials and device geometries. In the following section, RTN has been measured at cryogenic temperatures on Tech. C with dimensions \( W\times L = \SI {100}{\nano \meter }\times \SI {70}{\nano \meter } \). For this, RTN signals have been measured within a range of gate voltages for a set of temperatures between 10 K and 75 K at \( \vd =\SI {1}{\volt } \). For every measured RTN trace, the signal was analyzed using the Canny Edge Detector discussed in Section 7.3.1 for detecting charge capture and emission events and extracting the distribution of the corresponding capture and emission times and the step heights. This can be seen for a few examples in Fig. 7.13 for \( T=\SI {25}{\volt } \) at \( \vg []=\SI {-1.51}{\volt } \). The left panel shows an RTN signal with two active traps, one having a mean step height of \( \overline {\Delta I} = \SI {25}{\nano \ampere /\micro \meter } \) and the other of \( \overline {\Delta I} = \SI {12}{\nano \ampere /\micro \meter } \). For every measured trace, the extracted step heights can be plotted against their capture and emission times, as can be seen in the central panel. The detected charge capture and emission events can be clustered by their step heights using a K-means algorithm [236]. The two clusters with two different mean step heights are visualized by two different colors, see Fig. 7.13 (central panel). This data can then be conglomerated to a histogram for every defect showing the exponential distribution of the time constants as can be seen in the right panel.

(-tikz- diagram)

Figure 7.13: Steps on a 100 s RTN signal (left) recorded on Tech. C at \( T=\SI {25}{\kelvin } \) with \( \vg []=\SI {-1.51}{\volt } \) and \( \vd =\SI {1}{\volt } \) have been extracted using the Canny Edge Detector. Two active defects with mean step heights of \( \overline {\Delta I} = \SI {25}{\nano \ampere /\micro \meter } \) and \( \overline {\Delta I} = \SI {12}{\nano \ampere /\micro \meter } \) have been found. The step heights are shown over the capture and emission times in the central figure. Conglomerating this to a histogram shows the expected exponential distributions (right). Figures are reprinted from [241, MJC6].

Repeating this analysis for every bias and temperature condition allows to plot the mean capture and emission times over the gate voltage at every measured \( T \). This can be seen in Fig. 7.14 (left) for \( T=\SI {25}{\kelvin },\SI {50}{\kelvin } \) and \( \SI {75}{\kelvin } \). As this data set has been measured up to comparatively high temperatures, the intersection point where \( \taue =\tauc \) holds, and where the trap level is aligned with the Fermi level \( \Et =\Ef \), cannot be determined for the curves at higher temperatures. Thus, the temperature dependence is compared for a constant gate voltage.

(-tikz- diagram)

Figure 7.14: RTN signals were measured for a range of gate voltages and temperatures between \( T=\SI {10}{\kelvin } \) and \( T=\SI {80}{\kelvin } \). For every measurement, the mean capture and emission time was computed and plotted over (math image), as shown for \( T=25,50 \) and \( \SI {75}{\kelvin } \) in the left panel. The capture and emission times can be plotted in an Arrhenius plot for a constant gate voltage, as it is shown in the right panel for \( \vg =\SI {1.3}{\volt } \). The mean capture and emission times clearly behave non-Arrhenius like towards cryogenic temperatures. This can be explained with the full quantum mechanical transition rate and the fact that nuclear tunneling dominates in cryogenic charge transitions. Figures are reprinted from [241, MJC6].

As can be seen in Fig. 7.14 (right), the charge transition times at the point where \( \tauc =\taue \) holds show a clear non-Arrhenius like behavior and become completely temperature independent towards 10 K. This is the same behavior as measured on Tech. B in the previous section and which agrees well with the full quantum mechanical transition rate. At deep cryogenic temperatures, the only accessible vibrational state is the ground state. The overlap of the ground state vibrational wave function of the initial state with the vibrational wave function of the final state is temperature independent and so is the corresponding lineshape function leading to a vanishing temperature dependence of the transition rates and the corresponding capture and emission times.