Next: 3.3 Analysis of Worst-Case HCD Conditions in the Case of n- and p-channel MOSFETs Up: 3. Defect Creation Previous: 3.1 Microscopic Model for Interface State Creation
The reason of the mentioned in the previous Section model fail for devices with Lch = 1.2 and 2.0um is that the concentration of interface states generated by electrons peaks outside the channel, thereby weakly affecting the device performance. Figure 3.5a, showing the average Nit (i.e. integrated over the interface and divided by the interface length), demonstrates that more severe degradation corresponds to a lower average concentration of interface states. Hence, longer devices are less sensitive to electron-induced Nit, suggesting that another mechanism leading to Nit created closer to the channel must be responsible for this discrepancy. One may envisage that the missing contribution to the damage is triggered by secondary generated (by impact ionization) holes which are accelerated by the electric field and thus create interface states shifted towards the source. Therefore, it is necessary to extend the presented version of the model (described in Section 3.1) in order to make it suitable to represent the degradation in transistors with different channel lengths within the same set of parameters.
(b)
|
As it is proposed in Section 3.1 the superposition of single- and multiple-carrier mechanisms of Si-H bond-breakage is considered. Both mechanisms are controlled by the carrier acceleration integral
Note that we use the acronyms SP-/MP- to distinguish between the two bond-breakage processes and 'e/h' to distinguish the carrier types. The secondary holes are generated by impact ionization caused by the injection of hot electrons. These holes are then accelerated by the electric field towards the source, thereby creating interface states shifted with respect to the electron-induced ones. Also, the Nit fraction induced by holes should be much less than their relative contribution to Idlin change, Figure 3.5b.
For the SP-process first-order kinetics to describe the SP-induced contribution to the total Nit is assumedwhere νSP,e/h are prefactors related to electron/hole contributions.
The MP-induced portion of Nit is calculated analogously to (3.9) Following the procedure proposed in Section 3.1, one may obtain the new expressions for phonon excitation and decay rates with corresponding corrections for secondary carrier type
 |
 |
 |
(3.16) |
 |
|
 |
(3.17) |
The bond-breakage process is associated with the hydrogen hopping from bonded to transport state (with the rate λemi with the backward reaction also taking place (Ppass). For the MP-process, as it was already shown in Section 3.1, even in the case of a high-voltage device the MP-component makes a significant contribution to the total damage. However, in this case the carrier flux is still high enough, i.e. IMP,e + IMP,h ≫ ωe. As a result, the prefactor Pu/Pd in (3.9) is unity and NMP is homogeneously distributed along the lateral coordinate x as in Section 3.1.
(b)
(c)
|
The verification of the extended model is performed using a series of three 5V n-MOSFETs (see Figure 3.2) with identical architecture differing only in channel lengths (Lch = 0.5, 1.2, and 2.0um). Devices were fabricated on a standard 0.35um technology and subjected to a hot-carrier stress at the gate voltage Vgs = 2.0 and the drain voltage of Vds = 6.25V at 25oC. The model was calibrated in a manner which represents the Idlin degradation for all devices with a single set of fitting parameters and reveals a good agreement between experimental and theoretical results, shown in Figure 3.6. AIs for electrons and holes plotted vs. x are depicted in Figure 3.7.
(b)
(c)
|
In Figure 3.8, which demonstrates the total Nit profile and the hole-related component of Nit calculated for 10s and 104s, one may explicitly see that the hole contribution is considerably shifted towards the source. The single-electron component generates traps situated outside the channel which explains why the hole-induced traps have stronger relative impact on Idlin. Only the SP-component was considered while analyzing the relative contribution of holes. For both types of carriers the AI is already too high, thereby saturating the MP rate for this process. As a result, it is impossible to distinguish between multiple-electron and multiple-hole contributions. Additionally, NMP is homogeneously distributed over x and speculations about the position-dependent impact of Nit on the device performance are possible only for the SP-component. The contribution of hot holes to the total concentration Nit (Figure 3.5b) is much less than the corresponding fraction of Idlin change. This trend becomes more pronounced for longer devices. Furthermore, such a behavior is also supported by Figure 3.6 where the experimental Idlin degradation is plotted against the theoretical one as well as the portion of Idlin change induced by electrons and holes only. Note that in the case of Lch = 0.5um the degradation may be represented employing only the electron-component, but this is not possible for longer devices.
(b)
(c)
|
To summarize, the extended model is verified by representing HCD in n-type MOS transistors with various channel lengths using a single number of fitting parameters. The model considers not only channel electrons but also secondary holes generated by impact ionization. It is shown that in spite of a less pronounced hole contribution to the total interface states density, the device behavior is more sensitive to the hole-induced trap generation as compared to the electron one. This is related to the fact that hole-induced Nit is situated closer to the channel, thereby affecting the drain current in a stronger fashion.
