Solid-StateElectronics zyxwvutsrqponmlkjihgfedcbaZYXW Vol. 39, No. zyxwvutsrqponmlkjihgf I, 981-989, 1996 pp. zyxwvutsrqponmlkjihg 003%1101(96)oooo2-0 Copyright 0 1996 Elsevier Science Ltd Printed in Great Britain. All rights resewed 0038-I 101/96 515.00 + 0.00 A NOVEL PROGRAMMING METHOD FOR HIGH SPEED, LOW VOLTAGE FLASH E2PROM CELLS J. RANAWEERA, I. KALASTIRSKY, E. GULERSEN, W. T. NG and C. A. T. SALAMA Department of Electrical and Computer Engineering, University of Toronto, Toronto, Ontario, Canada M5S lA4 (Received 29 September 1995; in revised form 14 December 1995) Aimtract-A novel flash E2PROM cell structure programmed by hot electron injection from the source and drain sides of a MOSFET is presented. The hot electrons are generated by Zener/avaIanche breakdown of heavily doped reverse biased p +R + junctions at the source and drain. Two-dimensional device simulations were used to investigate the effect of the p n + + doping concentrations on the programming time, and the required supply voltages. The cell performance was experimentally verified on test devices fabricated using a standard CMOS process flow in which the n -LDD step was replaced with a p+ boron implant with a doping level of - 1.8 x 10’scm-3. A programming time of 6 ps was obtained on test devices with a gate length of 3 pm. With appropriate p+ doping concentrations and submicron gate lengths, it is possible to achieve a 3.3 V flash E2PROM cell with a programming time in the order of 150 ns. The superior programming speed and low operating voltages of the proposed flash E2PROM cell makes it suitable for a variety of portable applications. Copyright 0 1996 Elsevier Science Ltd INTRODUCTION Conventional flash E2PROM cells[l] are programmed by charging the floating gate by injection of hot electrons from the drain side of the cell. These hot electrons gain sufficient energy from the voltage applied to the drain contact of the memory cell to surmount the energy barrier between the silicon substrate/silicon dioxide insulator interface. A high voltage applied to the control gate attracts these hot electrons to the floating polysilicon gate on top of the thin tunnel oxide, where they accumulate. In this programming method, the magnitude of the lateral electric fields required for hot electron generation is dependent on the channel length and the drain voltage. When the channel length is reduced to the sub-half micron region, the drain voltage also has to be reduced to avoid punch-through breakdown. This results in a maximum lateral field in the range of lo5 V cm-‘, thus setting an upper limit on the hot electrons generation efficiency. This in turn affects the programming speed. In addition, this programming method requires high drain-to-source currents (in the range of milliamperes)[2] which limit the number of cells that can be programmed at the same time. In this paper, a novel programming method for flash E’PROM cells[3] is presented. The electric field available for hot electron generation is significantly higher (- 106Vcm-I), and it is independent of the channel length. The proposed cell can achieve fast programming speed to low supply voltages. In addition, the cell is compatible with conventional CMOS/BiCMOS technology and achieves the same density as conventional flash E2PROM cells[4]. NOVEL PROGRAMMING METHOD AND DEVICE STRUCTURE The proposed flash E2PROM ce11[5], shown in Fig. l(a) has additional p + regions adjoining the n + source and drain of the cell. The front portion of the cell which contains the two p+ regions is called the “program section” and the back portion which is similar to a conventional flash memory cell is called the “sense section” which allows current conduction during the read operation. The heavily doped p +n + junctions in the reverse biased condition are capable of generating hot elec- trons required for programming at relatively low voltages. These hot electrons result from Zener and or avalanche breakdown[6]. Since the doping concen- tration of n + source and drain regions of the cell are in the order of 1020 cmm3, the breakdown mechanism and the voltage required for breakdown depend on the doping concentration of the p+ regions. If the doping densities of both the p and n sides are greater than lOi*cme3, the depletion layer at the junction is very thin and an electric field (> lo6 V cm-‘) necess- ary for Zener breakdown is reached at voltages less than 6 V[7]. During cell programming, as shown in Fig. l(b), a large number of electron-hole pairs are generated (due to junction breakdown) from both the source and drain sides underneath the gate in the program section of the cell. The electric field of the Si-SiO, interface during breakdown of the junctions for doping levels of 1 x 10L9cm-3 and 1 x 1020cn-3 for the p + and n + regions, respectively, is illustrated in Fig. 2. When the source and drain are connected to 981