G.3 Fixed Oxide Charge Spatial Distribution



next up previous contents
Next: Appendix H: Direct Tunneling Rate Up: Appendix G: Relations for the Previous: G.2 Filtering and Differentiation

G.3 Fixed Oxide Charge Spatial Distribution

In order to derive the expressions for the extraction of fixed oxide charge we assume that no interface states are generated while stressing: , as well as are the same before and after stress, while build-up of occurs in time. These conditions are only roughly satisfied for stress at low (hole injection only) and probably also for stress at high (electron injection only) in -channel MOSFETs. They are better fulfilled in -channel devices for stress at low and medium (electron injection).

The charge-pumping current before stress is given by

 

with and being the boundaries for the total capture of holes in a virgin MOSFET. In a symmetric device, holds. A step-approximation is assumed for the transition between the total capture and the zero capture areas. For the charge-pumping technique we chose the constant amplitude method, discussed in Section 3.5.2. In this technique is variable, while is constant. and vary along the interface due to doping, but do not change with during the course of experiment. Therefore, emission times and and emission levels and vary along the -coordinate, but remain constant during the course of experiment, as well. We adopt that and are spatially uniform.
After stress the charge-pumping current reads

 

where and are the capture boundaries after the stress. The impact of on the local surface potential is nearly the same before and after the stress (generally negligible in practice). The stress-generated changes the local and . The difference in the measured current becomes . We will derive a general expression, valid for an arbitrarily large , but with the restriction that only one value exists for each . The assumption that the stress does not change the conditions at the source side is introduced. Since fixed charge roughly induces the same local shift in , , and gif the emission times and consequently, the emission levels and remain unchanged after the stress. Remember that they also do not change by varying in this charge-pumping technique. It follows

 

and , as well as and vary moderately within the short interval from to . The emission levels and change only slightly in this interval because of their logarithmic dependence on the emission times. Therefore, G.21 may well be approximated by

 

The local interface charge at can be calculated by

 

In relationship G.23, is the charge-pumping flat-band potential in the virgin device at and is the charge-pumping flat-band potential at the same position , but after the stress. A positive charge lowers the local , as consistent with G.23. When applying the constant-pulse method we scan the interface by while keeping the top level sufficiently high so that the interface is inverted in complete. At the critical coordinate , holds.
The charge extracted from relationship G.23 will be smaller than the charge actually inducing the local potential shift. Being aware of this fact we may call the extracted to be the apparent interface charge. Differences between the apparent and the real charge are addressed in Appendix F.
From G.22 follows that the coordinate actually scanned is given by

 

If the impurity concentration is known along the interface, is known as well. To find the second quantity in G.23, i.e. which is necessary to calculate , one should observe that holds. This relationship is illustrated in Figure G.1. Simple replacements lead to

 

where the spatial shift is given by

 

For a low density of , the shift is small and G.25 reduces to

 

Before applying G.25 on the experimental data, the spatial distribution must be known for the virgin device. This distribution can be calculated by a simple two-dimensional numerical simulation. The critical concentration which depends on the bottom level duration , is the input to the calculation. The differential current is measured. For given , the results from the known inverse relationship. The follows from G.26. Using G.25, the can be calculated. The only unknown factor is . If is uniform in the channel, this factor may be considered as roughly constant along the whole channel, which enables an estimation

 

In G.28, is the current before stress and is the effective charge-pumping channel length.

Note that also for a uniform distribution, the factor is different for traps around the drain/bulk junction than for traps in the middle of the channel. This difference introduces a proportional error in the determination of . The error is, however, small. A better approach than using G.28 is to extract the spatial distribution of in the virgin device by the methods given in the first part of this appendix. This distribution can be directly employed in G.26.

 

This method for the extraction of fixed oxide charge is sensitive to errors in the gradient of the charge-pumping flat-band potential distribution, as is evident from G.27. Therefore, the doping profile must be known with sufficient accuracy in the region of interest (gate/drain overlap region; around the junction). This is a serious limitation to the method. However, an inspection of the literature shows that only few attempts to extract in MOSFETs have been proposed [480][78]. The approach in [78][74] relies on observing the versus characteristics and can only provide a qualitative estimation of .

A study based on the numerical calculation is necessary to evaluate the potential of the present method to extract fixed charge distribution, as is done for interface charge in Section 3.5.2.



next up previous contents
Next: Appendix H: Direct Tunneling Rate Up: Appendix G: Relations for the Previous: G.2 Filtering and Differentiation



Martin Stiftinger
Sat Oct 15 22:05:10 MET 1994