Experiment on MOSFET Subthreshold Leakage with Stress-Dependent

This work has been conducted to corroborate the validity of the STI mechanical stress-dependent diffusion model mentioned in section 2.2 using a MOSFET device with an underlying lightly doped well, which exhibits a significant mechanical stress effect on the subthreshold I-V characteristics. The stress-dependent point defect equilibrium concentration and diffusion, which dominates the transient enhanced diffusion (TED), has also been taken into account.

The results of Diebel’s study [2.20] also show a linear dependence of both point defect equilibrium concentration and diffusivity in a log scale on the mechanical strain. Thus, it is reasonable to express the point defect equilibrium concentration and diffusion in an Arrhenius form:



where is the point defect equilibrium concentration under strain. To investigate the transient enhanced diffusion, only strain-dependent interstitial diffusion is needed. ∆EC for the vacancy extracted from calculation results[2.20] is +7.9eV/unit strain. Extracted ∆EC and ∆ES (in Eq. (2.5)) values for the interstitial are –7.0 and +0.99eV/unit strain, respectively. Furthermore, interstitial diffusivity and equilibrium concentration product, , is reduced under the compressive strain conditions. Two-dimensional process/device simulators, TSUPREM4 and MEDICI, were employed. The stress-dependent diffusion models were incorporated into TSUPREM4 through its user-specified equation interface.

*

A series of n-channel MOSFETs were fabricated using state-of-the-art process technology. Test structures had three active area length (Xactive in Fig. 2.1) values:

0.68µm, 1.46µm, and 20.2µm. Xactive is the design parameter to modulate mechanical stress. The minimum Xactive dimension of 0.68µm takes the presence of one contact window area in the source/drain into account. The gate length and width were 0.17µm and 10µm, respectively. The retrograde well implantations are omitted so as to enhance the sensitivity of the subthreshold characteristics to STI mechanical stress, offering the opportunity to verify the validity of the above mentioned diffusion model. The measured subthreshold I-V characteristics at VD = 1.2V are depicted in Fig. 2.13, with the substrate bias as a parameter. Previous work [2.23] revealed that the substrate bias measurement is a suitable verification index of the MOSFET doping profile because of the high sensitivity of carrier diffusion current to the dopant profile. The procedure for obtaining ∆ES values for various impurities began

by calibrating the one-dimensional dopant profiles in blanket control wafers, using processes that covered the range of device wafer process conditions. The results were taken as stress-free dopant profiles and used to calibrate the dopant diffusion parameters without considering stress-dependent diffusion effect. Two-dimensional MOSFET structures were then simulated in conjunction with the mechanical stress model. Calibrated diffusion parameters were employed to simulate a large Xactive

case, where the stress level is negligible. All major front-end process steps from the STI to the source/drain anneal were considered. The corresponding simulation geometries were calibrated using TEM cross-sectional images. Device simulations were performed and the device model parameters were calibrated to fit the I-V of the large Xactive MOSFET. Next, the process simulations based on VIDAESM for all Xactive values were conducted with an initial set of ∆ES values. Then, the device simulations with smaller Xactive values were performed and compared with the I-V data. The above procedure was iterated until a satisfactory reproduction of subthreshold I-V data was achieved in all cases.

The simulated strain distribution results for Xactive = 0.68µm are given in Fig.

2.14. It can be seen that the magnitude of the total strain, or volume change ratio, (εxx+εyy), is negative in the MOSFET core area, meaning that the device experiences compressive stresses during the process. In addition, the (εxx+εyy) of Xactive = 0.68　m was found to be much larger in magnitude than Xactive = 20.2µm. The compressive stress stems mainly from the lower thermal expansion rate of the STI oxide when compared to silicon, as well as the thermal gate oxidation induced

volume expansion at the STI edge. Thus, as Xactive decreases, the STI approaching the MOSFET core region increases the magnitude of the compressive stress. The extracted the ∆ES for phosphorus, arsenic and boron in the last section are –30, –14, and –7 eV/unit strain, respectively. The negative sign of ∆ES denotes diffusion retardation caused by the compressive stress in pure silicon for these impurities, and is in agreement with the literature [2.15],[2.19],[2.22]for boron and phosphorus. Note that, so far, no conclusive argument has been reached regarding the arsenic diffusion behavior under a general non-uniformly compressive stress in pure silicon.

In the MOSFET structure used in this section, source and drain phosphorus diffusion is much more sensitive to the mechanical stress than boron and arsenic because of the absence of the high concentration retrograde P-type doping well. The experimental silicon and simulation results are shown in Fig. 2.13 and Fig. 2.15. In the absence of the VIDAESM, the simulated leakage current (that is, the flat region in Fig. 2.14) was found to be much higher for MOSFETs with smaller Xactive values as illustrated in dashed lines in Fig. 2.13. The ID value for Xactive= 1.46 µm at VB = −1 V and VG = −0.4 V is 2.5×10⁻¹⁰A/µm without using VIDAESM, which is much larger than the result obtained using VIDAESM (6.3×10⁻¹¹A/µm). The corresponding silicon experimental data is 5.6 ×10⁻¹¹A/µm. The simulations that incorporated VIDAESM revealed that the punchthrough between the deeper part of the phosphorus source and drain is responsible for the leakage current. This means that the dopant diffusion becomes less as Xactive is decreased, and is consistent with the results indicated in Fig.

2.15: a decrease in Xactive produces a substantially large reduction in leakage current.

In Fig 2.15, the gate-edge tunneling current of Xactive = 0.68µm MOSFET prevails in the background current, regardless of high negative substrate biases.

To investigate the impact of the mechanical stress on transient enhanced diffusion, the stress-dependent point defect diffusion and equilibrium concentration models described in Eqs. (2.4) and (2.5) are applied to the numerical simulator. The simulation results show slightly higher subthreshold leakage current, which implies that the dopant diffusion is the stronger. The ID value for Xactive= 1.46 µm at VB = −1 V and VG = −0.4 V increase from 6.3×10⁻¹⁰A/µm to 7.1×10⁻¹⁰A/µm. The explanation for this phenomenon is that the interstitial equilibrium concentration C decreases under the compressive stress and therefore the interstitial supersaturation factor, / , increases after impurity ion implantation, which results in the TED enhancement. The effect is not significant because high ramp rate rapid thermal anneals were applied after ion implantations. To further fit the experimental data after taking the stress-dependent TED effect into account, the final value ∆ES of phosphorus is fine tuned from –30 to –33eV/unit strain. Fig 2.15 depicts the corresponding results of experimental data fitting.

* I

CI C_I^*