Katsuhiro Tsuji

Evaluation of MOSFET C-V curve variation using test structure for charge-based capacitance measurement

Test structure for charge-based capacitance measurement (CBCM) is improved, to achieve higher accuracy of measuring capacitance-voltage (C-V) curves for actual size MOSFETs. Capacitance mismatch between the device under test (DUT) and the reference is avoided by using charge-injection-induced-error-free CBCM (CIEF CBCM) method. To increase the applicable bias voltage range to DUT, both P- and N-channel MOSFETs are parallel-connected in the pseudo-inverter. It is found that the C-V curves, which are measured with this test structure and are corrected by removing the size effect, are very close to the C-V relation measured by the conventional method, and then, the corrected capacitances give more accurate gate capacitances of MOSFETs.

Measurement of MOSFET C-V Curve Variation Using CBCM Method

The test circuit, in which the cells including CBCMs (Charge-Based Capacitance Measurements) are arrayed in matrix shape, is developed to measure MOSFET capacitance variation. By adjusting the bias condition of the test circuit, it is able to obtain C-V curves for many MOSFETs. Additionally, a variation of threshold voltage is extracted from the estimated C-V curve variation. The obtained threshold voltage variations are close to those which are obtained from current-voltage characteristics.

Effective channel length estimation using charge-based capacitance measurement

An effective channel length is estimated from the capacitance-voltage (C-V) curves of actual size MOSFETs which are measured using charge-based capacitance measurement (CBCM). To evaluate the accurate capacitances between the gate and the channel of sample MOSFETs, their parasitic capacitances are removed by using the test MOSFETs having various channel size and special test structure. A good linear relation between the gate-channel capacitance and the design channel length is obtained and then, the effective channel length is estimated from it. It is found that the obtained effective channel length is shorter than that extracted by the conventional channel resistance method.

Threshold voltage variation extracted from MOSFET C-V curves by charge-based capacitance measurement

The threshold voltage variations for the MOSFETs having various channel structures are evaluated from their measured capacitance-voltage (C-V) curves. It is found that they show reasonable dependence on the channel structure and smaller than those evaluated from the current-voltage (I-V) relations. As one of the reasons that the difference of between the variations extracted from C-V curves and those extracted from I-V relations arises, it is considered that the local channel dopant fluctuation increases the current variation. Furthermore, it is found that the evaluated flat-band voltage variations represent the reasonable behavior.

Effect of Channel Dopant Distribution on Effective Channel Length Extraction

The effect of the channel dopant distribution on the effective channel length (LEFF) of metal–oxide–semiconductor field-effect transistors (MOSFETs) extracted by the channel resistance method (CRM) is studied using simple device models and experimental data. The simple device models explain how the channel dopant distribution affects extracted LEFF. The measured data for various channel structures are qualitatively explained using those models. It is found that an accurate LEFF can probably be extracted if plural samples having different dopant densities can be prepared.

Study on Device Matrix Array structure for MOSFET g m variability evaluation

The effect of Device Matrix Array structure on MOSFET gm-variability measurement is studied. One of the two transfer gates, which are connected to an MOSFET source terminal for both Kelvin measurement and addressable access, is removed. This modification enables us to measure and to reduce the effect of the metal wiring resistance on Kelvin measurement, and thus to estimate the intrinsic MOSFET gm-variability.

Electrical estimation of channel dopant uniformity using test MOSFET array

The dopant uniformity in an MOSFET channel is estimated using the test MOSFET array which includes many MOSFETs with different channel length. Takeuchi coefficient as a function of the channel length is calculated from the measured threshold voltage data. The electrical channel length as a function of the gate voltage is extracted using channel resistance method. It is found that those data show the similar degree of the dopant uniformity for MOSFETs having various channel structures.

Comparison of channel length extracted from gate capacitance with that extracted from channel resistance

The two effective channel lengths L_CAP1 and L_CAP2 which are extracted from the gate capacitances with and without the fringe component, are compared with the effective channel length extracted from channel resistance L_CEF. It is found that L_CAP1 is about 30 nm longer than L_CAP2 and almost coincides with L_CEF for the samples made with 65-nm technology. It is considered that this difference is caused by the extension regions next to source and drain regions.

Reconsideration of effective channel length for metal–oxide–semiconductor field-effect transistor

The effective channel length (LEFF) of a metal–oxide–semiconductor field-effect transistor (MOSFET) as a function of gate voltage is extracted by a channel resistance method. It is observed that LEFF is strongly affected by the channel dopant nonuniformity and gate voltage. It is also weakly affected by the source–drain junction and substrate voltage. Considering these properties, we propose a new channel length definition, which is constant and obeys the classical current equation for MOSFETs with uniform channels, in order to provide a common channel length interpretation for both design and technology engineers.

Effective channel length of MOSFET with halo

The dependences of the effective channel length on the gate voltage for MOSFETs with halo are extracted, and their average behaviors are studied. It is found that they provide the important information about not only the channel structure, but also the effects of the gate voltage and the substrate voltage on the channel, which are helpful for both technology and design engineers to get the common interpretation of “channel length”.