Gate Leakage vs. NBTI in Plasma Nitrided Oxides: Characterization, Physical Principles, and Optimization

Abstract

Since nitrided oxides improve gate leakage at the expense of NBTI, one must optimize nitrogen concentration in oxinitride samples for reliable performance and reduced power dissipation. Here, we analyze wide range of NBTI stress data to develop a predictive model for gate leakage and first selfconsistent model for field acceleration within R-D framework. This model anticipates a novel design diagram for co-optimization of leakage and NBTI for arbitrary nitrogen concentration and effective oxide thickness.1. Introduction Since nitrided oxides (SiON) improve gate leakage (J G ) [1-6] at the expense of NBTI performance (∆V T ) [7,8], one must necessarily optimize N 2 concentration (%N) in gate-oxides for high-performance ICs. Despite its importance, however, a quantitative analysis of leakage/NBTI trade-off (as a function of %N), has never been reported and the question "Is cooptimization of NBTI/leakage possible at any %N?" has never been answered. In this paper, we simultaneously measure gate leakage and delay-free NBTI over broad range of stress-fields, stress-temperatures and %N, model gate leakage current (J G ) and NBTI degradation within a theoretically consistent framework (hole-assisted thermal generation of interface traps) of field-dependent R-D model, and conclude that although there is no optimum %N for NBTI/leakage, the reduction in J G at NBTI-limited %N (~15-25%, depending on failure criterion) can be significant and would reduce power dissipation without affecting NBTI-margin.2. Gate Leakage Comprehensive simulation [9] (which includes the effects of multi-subband electron/hole quantization, poly-depletion, etc.) of the measured J G -V G for both N-and PMOS (Fig. 1) was done to extract the model parameters as a function of %N (Fig. 2). We assume that any variation in the spatial-profile of nitrogen results only in second-order correction to calculated J G . Contrary to popular belief [2-6, 10], the oxide parameters do not scale linearly with %N. All the parameters have approximately a quadratic fit with %N. Here, effective oxide thickness (EOT) is obtained from simulation of CV, and physical thickness (T PHY ) and %N are determined by XPS [11]. These %N-dependent parameters are used to calculate J G (N,EOT) for arbitrary %N and EOT, as shown in Fig. 10b. -1.0 -0.5 0.0

Critical analysis of short-term negative bias temperature instability measurements: Explaining the effect of time-zero delay for on-the-fly measurements

“…3,7 This "H 2 R-D" model also provides a consistent interpretation of temperature and field dependencies of NBTI for long-term stress, as extensively studied for devices with SiO 2 , plasma SiON, and thin thermal SiON dielectrics. 2,5,6,8 Although the classical "H 2 R-D" model provides an excellent interpretation for long-term stress data, our analysis shows that ͑1͒ the predictions of this model is inconsistent with the short term, sub-10 s, NBTI degradation ͓see Fig. 1͑b͔͒ and ͑2͒ in contrast to long-term degradation, the shortterm time exponent of NBTI degradation depends on the measurement techniques.…”

“…Further, we demonstrate that the existing discrepancy in short-term NBTI data by various measurement techniques can be attributed to contamination by time-zero delay in OTFM systems. Our analysis suggests that lifetime prediction 4,6 and projection back to operating condition 6 must be carried out using long-term NBTI stress data ͑having n ϳ 1/6͒, obtained after t stress ϳ 100 s with t 0 ഛ 1 ms, because higher exponents in short-term degradation can result in unnecessarily pessimistic projections.…”

“…[1][2][3][4][5][6][7] At long stress time ͑t stress Ͼ 10-100 s͒, ⌬V T shows a power law behavior when plotted with respect to time ͑⌬V T ϳ At n ͒ with a consistent time exponent ͑n͒ of ϳ1 / 6. This time exponent is independent of the measurement techniques used to monitor ⌬V T ͓e.g., ultrafast measurement ͑UFM͒, 9 "delay-free" on-the-fly I dlin measurement ͑OTFM͒, 2,3,5,6,8 etc.͔. The R-D model attributes the robust ͑constant over several decades in time 4 ͒ long-term n ϳ 1 / 6 exponent to the diffusion of molecular hydrogen ͑H 2 ͒.…”

“…The origin of such degradation has been debated for the last few years, and it is now generally accepted that, for devices with SiO 2 , plasma SiON, and thin thermal SiON gate dielectrics, NBTI degradation results mainly from depassivation of Si-H bonds at the Si/dielectric interface and resultant diffusion of hydrogen species into gate dielectric and poly-Si. [1][2][3][4][5][6][7][8] As such, several groups have used the reaction-diffusion ͑R-D͒ model to interpret NBTI degradation. [1][2][3][4][5][6][7] At long stress time ͑t stress Ͼ 10-100 s͒, ⌬V T shows a power law behavior when plotted with respect to time ͑⌬V T ϳ At n ͒ with a consistent time exponent ͑n͒ of ϳ1 / 6.…”

“…[1][2][3][4][5][6][7][8] As such, several groups have used the reaction-diffusion ͑R-D͒ model to interpret NBTI degradation. [1][2][3][4][5][6][7] At long stress time ͑t stress Ͼ 10-100 s͒, ⌬V T shows a power law behavior when plotted with respect to time ͑⌬V T ϳ At n ͒ with a consistent time exponent ͑n͒ of ϳ1 / 6. This time exponent is independent of the measurement techniques used to monitor ⌬V T ͓e.g., ultrafast measurement ͑UFM͒, 9 "delay-free" on-the-fly I dlin measurement ͑OTFM͒, 2,3,5,6,8 etc.͔.…”