3.1 Introduction



next up previous contents
Next: 3.1.1 Charge-Pumping Techniques Up: 3 Analytical and Numerical Previous: 3 Analytical and Numerical

3.1 Introduction

 

The charge-pumping effect in MOS devices has been reported by J.S.Brugler and P.G.A.Jespers in 1969 [43] and by A.Goetzberger and E.H.Nicollian in [148] for the first timegif. They have measured a DC component of the bulk current when periodic voltage pulses are applied to the gate in the circuit shown in Figure 3.1. This current was in the opposite direction and much larger than the leakage current of the reversely biased source and drain junctions. It was proportional to the pulse frequency and the gate area. This current is called the charge-pumping current . In the pioneering work [43] the authors have given a correct qualitative physical explanation of the effect.

 

Neglecting the junction leakage current, the DC component of the bulk current is caused by two effects (for -channel MOSFETs):

  1. By varying the gate pulse sufficiently so that the condition at the interface changes from inversion to accumulation, traps at the interface (and some bulk traps) are successively filled by electrons and holes. This produces a positive net bulk hole current and a negative net source/drain electron current. The effect originates due to the longer emission times for traps compared to the duration of the gate-pulse rise and fall times. The charge-pumping current due to this effect contains information on the traps in MOSFETs.
  2. For slow turn-off of MOSFETs (long fall time of the gate pulse) electrons in the channel have sufficient time to arrive at the source and drain junctions. The inversion layer vanishes before the accumulation occurs at the interface. However, for fast turn-off, when the fall time of the gate pulse is shorter than the time necessary to remove the inversion layer, a significant amount of electrons remain at the interface when the hole accumulation takes place. These electrons are either recombined by holes or transferred to the bulk, which both produce a positive net bulk current of holes and electrons, respectively. The charge-pumping current due to this fast-switching effect does not contain any information on the traps in the device.

The first effect has been extensively used in the last ten years to extract the amount and the distribution (in both, energy and position space) of traps in MOS devices. In these measurements, the second effect introduces a parasitic undesired component which can be removed in most cases. The reasons for an increasing popularity of charge pumping are its simplicity, high sensitivity, good accuracy and its direct applicability on small devices. Charge pumping is mostly applied to analyze the localized degradation after hot-carrier stress in MOSFETs [423][374][288][287][196][194][34][9]. Its ability to be performed on small devices (with real design dimensions) is particularly useful in studying the nonuniform degradation of memory MOS devices (EPROMs and EEPROMs) under working conditions [510][509][214][196]. Uniform degradation caused by E-beam [196] and -rays [495] irradiation and by Fowler-Nordheim injection [495][196][76] has been studied by the charge pumping as well. A particular problem is to study traps in SOI devices, where both, front and back interface play an important role in determining the properties of those devices. Detailed studies are presented in [515] for laser-recrystallized silicon films, in [445] for epitaxial films and in [357][116] for the SIMOX technology. A specific application is to evaluate the traps at grain boundaries in polysilicon-film devices [258][257]. In this case the measured quantity (charge-pumping current) contains a direct information on the traps at grain boundaries. Note that in the conventional current-related techniques to study the polysilicon devices, the device current is affected by the barriers at the grain boundaries, while the barriers are influenced by the traps at grain boundaries, as well as by the dopant concentration. Therefore, the measured drain current contains an indirect information on the traps. The extracted trap density is model-dependent in these cases.

The second application of charge pumping is to employ it for device operation. Much less attention has been paid to this type of application in the past (at least in the literature). Some examples of the so-called charge-pumping devices will be given henceforward. In the work [113] (1975) by an INTEL group, the charge-pumping effect is used to refresh the storage gate capacitance in a two-MOS-transistor bistable memory cell. Both effects, 1 and 2 described above, can be involved at device operation. A 4096-bit memory array is realized with these cells. Here, the feature that the charge-pumping effect produces a DC current in the opposite direction to the junction-leakage current (which discharges the node) is used. The charge-pumping current predominates over the leakage currents and charges the corresponding gate capacitance in a bistable configuration. The same feature is used in [53][50] (and references therein) by U.Cilingiroglu, where charge pumping at the node itself is applied to refresh the floating storage node. The difference between the charge-pumping and the leakage current yields two stable conditions at the characteristics of the gated-diode. By adding a gate resistor [53] or by capacitive coupling [50] the storage node can exhibit bistability with a regenerative loop. A different type of charge-pumping devices is proposed by N.Sasaki in [407][405]. During charge pumping the minority carriers can be injected by effect 2 into the bulk of SOI MOSFETs (SOS technology has been considered in [407][405]). The injected charge remains in the floating substrate, because the substrate potential (its absolute value) increases and reversely biases the source and drain junctions. The increase of the bulk potential is independent of the duration of the top and bottom levels, while it depends on the fall time of the gate pulse (Figure 3.2). The presence of the charge in the bulk can be detected as a change in device conductivity. Using the avalanche effect the charge can be removed from the bulk. This concept has been implemented in an one-device memory cell [407] which has been applied to build a memory array [406].





next up previous contents
Next: 3.1.1 Charge-Pumping Techniques Up: 3 Analytical and Numerical Previous: 3 Analytical and Numerical



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