[paper] Predictive Hot-Carrier Aging Compact Model

Y. Xiang^1,2, S. Tyaginov^1,3,4, M. Vandemaele^1,2, Z. Wu^1,2, J. Franco¹, E. Bury¹, B. Truijen¹, B.Parvais^1,5, D. Linten¹, B. Kaczer¹

A BSIM-Based Predictive Hot-Carrier Aging Compact Model 

4A.4; IRPS March 21- 24 2021 

¹imec, Leuven (B)
²Department of Electrical Engineering (ESAT), KU Leuven, Leuven (B)
³Institute for Microelectronics (IuE), TU Wien, Vienna (A)
⁴Ioffe Physical-Technical Institute of the Russian Academy of Sciences, Saint Petersburg (RU) 
⁵Department of Electronics and Informatics (ETRO/VUB), Brussels (B)

Abstract: The continued challenge of front-end-of-line transistor reliability has long demanded physics-based SPICE compact models, not only for service lifetime estimation, but also for agingaware device pathfinding with technology scaling and innovation. Here, we present a predictive hot-carrier-degradation (HCD) compact model built upon the industry-standard BSIM model, that conveniently embeds the essential HCD physics within common SPICE simulation flows. We leverage and augment the established, scalable electrostatics and transport in BSIM as the input to an analytical HCD interface states generation formalism, the result of which is in turn injected back into BSIM for a selfconsistent estimation of the threshold voltage (VTH) shift and the mobility degradation. Our approach readily exhibits fundamental, non-empirical predictabilities of the stress timeand the sensing bias- dependency of transistor-level degradation, without having to resort to a priori assumptions. This will further accommodate the irregular, arbitrary voltage waveforms in transient circuit operations, thus enabling efficient evaluation of the power-performance degradation at circuit level. The model ultimately aims to lay the groundwork for a reliability-aware design-technology co-optimization in device pathfinding. 

Fig: Schematic of the Pao-Sah DD current integral method used in commercial CMs [a-e] and the extrapolated piecewise Vch(y) by augmenting the BSIM model. In the Pao-Sah DD formalism, the actual Ids is calculated by the difference of the integral Ξ at the source (channel potential Vch=0) and at the “drift-diffusion limit” (at LDD, where channel potential Vch=VDS,eff), with the latter defined by velocity saturation or pinch-off. The Vch(y) is extrapolated by using the implicit assumptions in BSIM-BULK: the quadratic profile under gradual channel approximation (GCA) and the hyperbolic profile under the drain-side field assumption used in substratecurrent body-effect (SCBE). 

References:

[a] C. K. Dabhi. (2017). BSIM4 4.8.1 MOSFET Model: User’s Manual. [Online]. Available: https://bsim.berkeley.edu/models/bsim4/.

[b] H. Agarwal. (2017). BSIM-BULK106.2.0 MOSFET Compact Model: Technical Manual. [Online]. Available: https://bsim.berkeley.edu/models/bsimbulk/. 

[c] S. Khandelwal. (2015). BSIM-CMG 110.0.0 Multi-Gate MOSFET Compact Model: Technical Manual. [Online]. Available: https://bsim.berkeley.edu/models/bsimcmg/. 

[d] P. Kushwaha. (2017). BSIM-IMG 102.9.1 Independent Multi-Gate MOSFET Compact Model: Technical Manual. [Online]. Available: https://bsim.berkeley.edu/models/bsimimg/. 

[e] W. Grabinski et al., (2019) "FOSS EKV2.6 Verilog-A Compact MOSFET Model," ESSDERC 2019 - 49th European Solid-State Device Research Conference (ESSDERC), Cracow, Poland, 2019, pp. 190-193, doi: 10.1109/ESSDERC.2019.8901822

[Online] Available: https://github.com/ekv26/model