Design and characteristics of n-channel insulated-gate field-effect transistors

An n-channel insulated-gate field-effect transistor technology established at IBM Research has served as the basis for further development leading to FET memory. Designs and characteristics of experimental devices of 500 and 1000 A gate insulator thicknesses are presented, with particular attention to the effects of source-drain spacing. Introduction Because the insulated-gate field-effect transistor [ 1,2] shows great promise for applications in high-density memory and logic, there has been an effort in the IBM Research Center over the last decade to develop materials, devices, and circuit applications for this technology. An early choice was made to pursue n-channel devices because their mobility, which is higher by a factor of two to three than that of p-channel devices, offers a corresponding improvement in performance (delay-power product) [3]. The characteristics of experimental devices that were designed and fabricated during this period are described in this paper. The device technology was incorporated into integrated memory prototypes that provided experience for the development of the F E T memory in IBM System/370, Models 158 and 168. Two different experimental device designs evolved, one using a lOOOA gate insulator thickness and the other using a SOOh; thickness; these designs permitted an exploration of the advantages and problems of thinner insulators. The next section of this paper gives an introductory description of the device operation and some of its main characteristics, using the 500h; device for illustration. The subsequent section gives a more thorough representation of the dc characteristics of both types of devices, including the effects of channel ength (source-drain separation). The conclusion one may draw is that the n-channel device offers very desirable properties for circuit design, having conduction characteristics approximating those predicted by the conventional equations with an effective mobility of 510 cm2/volt-sec. The device threshold tends to be reasonably low, because of the negative work function difference between the aluminum gate and 430 CRITCHLOW. DENNARD AND SCHUSTER the p-type substrate, and can be adjusted to a desired value by choice of an appropriate substrate bias voltage. The threshold is particularly low, well-controlled and relatively insensitive to source-substrate voltage for the 500A device. These devices allow the use of low circuit voltages with a correspondingly low power consumption if gate insulators of this thickness can be reliably fabricated. Device description The n-channel IGFET device, Fig. 1 , consists of a pdoped substrate, n+ diffusions to form the source and drain electrodes and a metal gate electrode insulated from the substrate by a thin layer of oxide. A set of grounded source characteristics of such a device with a 500A gate oxide thickness is shown in Fig. 2 (a) . The operation of the device can be studied under three distinct conditions as shown in Figs. 2 (b) , (c) , and (d). The first condition, that of a turned-off device, occurs when the gate voltage measured with respect to the source, V,,, is less than a critical threshold voltage, V,. The charge in the channel region between the source and drain is depleted as shown in Fig. 2 (b) , or may be accumulated (heavily p-type) in the case of a gate voltage that is much less than V,. Because the p-type substrate is tied to a fixed bias potential below that of the grounded source, both source and drain are reversebiased with respect to the substrate and are thus surrounded by depletion layers. In this condition there is no path for conduction between source and rain. When the gate-to-source voltage, V,,, is greater than V,, the electric field causes the channel region between source and drain to be inverted to n-type as shown in IBM J . RES. DEVELOP. Fig. 2(c) , which allows current conduction from the drain to the source (electron flow from source to drain) when a voltage is applied to the drain. The current increases nearly linearly with drain-source voltage, VDs, for low values of VDs as shown in Fig. 2 (a). This applied drain voltage causes an IR drop to be distributed along the channel, with the potential rising from ground to VDs as the channel is traversed from source to drain. The voltage from gate to channel is thus diminished in the vicinity of the drain, and the reduced field causes the electron concentration to be less as expressed pictorially by the channel depth in Fig. 2 (c). For a large drain voltage (VDs > VGS V,) the channel is "pinched off' completely near the drain as shown in Fig. 2 (d) since there is insufficient field strength from the gate to turn on the channel. Current continues to flow through the space-charge region between the drain and the end of the channel. Further increases in drain voltage are dropped across the space charge region, and, Leff Thin-oxide I 1 Drain L tJ diffusion Source diffusion interconnection Aluminum Underpass A diffusions Figure 1 Structure of the IGFET thin-oxide device (a) and of the unwanted (parasitic) thick-oxide device (b). Figure 2 Typical grounded-source characteristics for a 500A device with illustrations and approximate device equations (not including R D and R , ) for three different regions of operation. (W = 4 mil; Leff = 0.18 mil; V,,, = -7V.)