Capacitorless DRAM gain cell

[0002] This invention relates to a capacitorless dynamic random access memory (DRAM) gain cell. In particular, it relates to a self-amplifying large-scale DRAM memory cell that has ultra-high integration density, two field effect transistors, and requires no charge storage capacitor.

[0003] The continuing movement of integrated circuit technology toward smaller scales is making system level integration on a chip both possible and desirable. A system level integration generally merges, on a single chip, memory and logic functions. DRAM cells are attractive for such merged system integrations because DRAM cells occupy a small area, and thereby potentially allow a large number of memory cells to be integrated with logic functions. However, even with DRAMs, merging memory and logic involves problems with process complexity and cost. For example, merging conventional DRAMs with logic, using either stacked or trench capacitor cells, is very complex and costly for several reasons. One reason is that logic is frequently performance driven and, when seeking a compromise between cost and performance, the latter may be favored. DRAMs, on the other hand, are frequently cost driven due, in large part, to the sheer number of such devices used in many commercial systems. Another reason is that deep trench technology, as used for DRAM cells, is not preferred for implementing logic. Still another reason is that stacked capacitor technology, which is a conventional DRAM technology, causes problems with the lithography of standard logic processes, due to its non-planar topography.

[0004] One possible solution is to use static random access memories (SRAMs), which can be easily integrated with complementary metal oxide semiconductor (CMOS) logic. SRAM cells, however, are not area efficient. Thus, there is a need for a memory cell that occupies a very small area, yet does not require extra processing for integration with the logic, especially large capacitor.

[0005] An object of the present invention is to provide a high performance, area-efficient memory cell that can be fabricated by conventional CMOS technology, without the need for special process or structural modifications. Moreover, an object of the present invention is to avoid either stacked capacitor or deep trench storage and, instead, to employ standard devices available in high performance CMOS logic technology.

[0006] Another object of the invention is to provide a high performance, area-efficient memory cell that can use conventional CMOS voltage levels, thereby facilitating merged system level integration.

[0007] Pursuant to these and other objectives, one embodiment of the present invention comprises a storage metal oxide semiconductor field effect transistor (MOSFET), which stores information, and an access MOSFET, which controls the charging and discharging of the gate of the storage MOSFET for writing information. The access MOSFET turns on in response to a write control signal connected to its gate. When the access MOSFET is turned on, a write information signal, representing either a logical “0” or logical “1,” passes through that access MOSFET to the gate of the storage MOSFET. The storage MOSFET is thereby charged to a weak inversion condition or to a depletion (or even majority carrier accumulated) condition in accordance with the write information signal.

[0008] To read information, the access MOSFET is turned off and the storage MOSFET is “off-state” because in weak inversion and depletion conditions a conductive channel is not induced at the surface of the storage MOSFET so a current does not flow across the channel. A read control signal connected to the body of the storage MOSFET is applied with a forward bias to the source. The resulting drain current of the storage MOSFET depends upon its gate charge condition, thereby indicating the state of the stored information.

[0009] A significant and novel feature of the present invention is that when the storage MOSFET is in a weak inversion charge state, the forward bias of the body at the source junction causes a large bipolar drain current. A current gain for bipolar action depends on the condition of the surface of the body. In a weak inversion condition, the current gain is larger than it is in the depletion condition.

[0010] A further embodiment of the present invention is using a different type of MOSFET for the access MOSFET and the stored MOSFET. This results in a one word line and a one bit line circuit.

[0011] Yet a further embodiment of the present invention is a method of operating a MOSFET where a constant voltage below but close to the threshold voltage is applied to the gate to place its surface in a depletion or a weak inversion condition and to input a forward signal bias from the body to the source junction, which causes a large bipolar drain current. A current gain for bipolar action depends on the forward signal current.

[0012] FIG. 1 is a diagram illustrating a certain presently preferred embodiment of a dynamic memory cell according to this invention.

[0013] FIG. 2 is a graph illustrating the relationship between Vg and Id as a parameter of Vb

[0014] FIG. 3 is a diagram illustrating another certain presently preferred embodiment of a dynamic memory cell according to this invention.

[0015] It will be understood from this description that the present invention can be implemented in conventional MOSFET technology, and that the described embodiments will operate accordingly if designed and fabricated in accordance with known CMOS and SOI (silicon on insulator) rules and methodologies. These rules and methodologies are well-known in the art and will not be repeated for this description. SOI materials meeting this criterion are well known in the art.

[0016] Referring to the drawings, and more particularly to FIG. 1, there is shown, by schematic, an example of an embodiment of a memory cell according to the present invention. A storage MOSFET transistor 2, has a gate 2g, a source 2s, a drain 2d, and a body (substrate) 2b, and an access MOSFET transistor 1, has a gate 1g, a source 1s, a drain 1d, and a body 1b. As will be understood, diffusion regions 1sand 1dfunction either as a source or a drain, according to the direction of the current therethrough.

[0017] For this embodiment, the storage transistor 2 is a MOSFET of one conductivity type, i.e., either NFET (N-channel Field Effect Transistor) or PFET (P-channel Field Effect Transistor), and the access transistor 1 is of the same conductivity type, i.e., either both NFET or both PFET. For this particular example, the access transistor 1 and the storage transistor 2 are both NFET.

[0018] As described below, the storage transistor 2 stores a logical “0” and a logical “1” by being charged, through the access transistor 1, to a depletion condition or to a weak inversion condition, respectively. The storage transistor 2 holds the charge by its gate 2gbeing held at the gate voltage for of a logical “0” or a logical “1.” The control signals connected to the transistors 1 and 2, which effect the charging, discharging, and reading of the depletion condition, are the read word signal, labeled as RW, the read bit/precharge signal, labeled as RB, the write word signal, labeled as WW, and the write bit signal, labeled as WB.

[0019] For this description, a weak inversion condition represents a logical “1”, and a depletion condition represents a logical “1.” This definition is, of course, an arbitrary design choice. When storage transistor 2 contains a logic “1” its voltage is close to but below its threshold voltage and the surface under gate 2gis in a weak inversion condition. In this condition, current does not pass from source 2sto drain 2d. A read operation begins when a Read Word line (RW) signal is applied to the body-to-source junction (PN junction) which becomes forward biased. This condition will precipitate an inherent bipolar current flow whose magnitude is proportional to the conditions on the surface of the body of the storage transistor 2, multiplied times the current gain (beta value) of the inherent bipolar transistor (i.e., storage transistor 2). When storage transistor 2 contains a logic “0,” its voltage is far from the threshold voltage of storage transistor 2 and the surface under gate 2gis in a depletion condition. In this condition, the bipolar current is small. In this manner, the effective value of the gate charge, in terms of the magnitude of the discharge current sensed at the Read Bitline (RB), is enhanced by this parasitic bipolar effect.

[0020] FIG. 2 shows the relationship between the gate voltage (Vg) and the drain current (Id) as a parameter of body bias (Vb) for a conventional MOSFET. When Vg is applied at A volts, Id is very small and the surface condition under the gate is in a depletion condition, which means that the surface is charged the same type as the body. When Vg is B volts, Id is larger than in the case of A volts. In this condition, the surface is in a weak inversion condition and the charge of the surface is still a majority carrier, which means that it is the same type as the body, but the ratio between the majority and the minority carriers becomes small. And when Vg is Vth (the threshold voltage), Id is much larger and the surface condition become inverted completely, which means that the charge on the surface is the same type as the source and the drain.

[0021] Parameters (1), (2), and (3) show forward bias conditions of the body-to-source junction (PN junction). (1) is the nearly built-in voltage of a PN junction, about 0.65V, (2) is about 0.4 to about 0.5V , and (3) is the bias at 0V. Experimental data can be found in FIG. 11 on pages 414 to 429 of IEEE Transactions On Electron Devices, Vol. 44, No.3, March 1997, “Dynamic Threshold-Voltage MOSFET(DTMOS) for Ultra-LOw Voltage VLSI.” According to this paper, at Vb=0.5V and Vg=0V, A is 2 nA , at Vb=0.5V and Vg=0.5V, B is 5 uA. As another example, In FIG. 5 of IEEE Transactions On Electron Devices, Vol. 45, No.5, May 1998, pages 1000 to 1009, “Approaches to Extra Low Voltage DRAM Operation by SOI-DRAM,” shows that A is 2 nA at Vb=0.5V, Vg=0V, and B is 20 uA at Vb=0V and Vg=0.5V. Thus, the difference in voltage between “0” and “1” is more than 2 digits. From these data, a conventional MOSFET acts as a storage charge detector in this manner. As that sensitiveness is superior to the current MOSFET operation mode, this invention is applicable to other memory-like conventional floating gate non-volatile memory cells, MONOS (Metal Oxide Nitride Oxide Semiconductor), such as local storage type non-volatile memory cells and the pair input transistors of differential amplifiers.

[0022] In the above explanation of FIG. 2, Vb was constant and Vg was variable, but it is also possible for Vg to be constant and Vb to be variable, which is more sensitive than conventional MOSFET operations. This example is shown in FIG. 2, where Vg is the constant B volt and Vb is changed from C to D volts as an input signal. In FIG. 5 of IEEE Transactions On Electron Devices, Vol. 45, No.5, May 1998, pages, 1000 to 1009, at Vg=0.5V and Vb=−0.5V, Id=0.5 uA. At Vg=0.5V and Vb=+0.5V, Id=20uA In addition, the gate oxide thickness doesn't need to follow the scaling law of its gate length shrink.

[0023] While the actual voltages used will vary with the particular devices, preferably storage transistor 2 has a forward bias voltage of about 0.3 to about 0.6 volts, a depletion voltage of about 0 to about 0.3 volts, and a weak inversion voltage of about 0.7 to about 1.0 volts.

[0024] In FIG. 3, the storage transistor 2 is a MOSFET of one conductivity type, i.e., either NFET or PFET, and the access transistor 1 is a different conductivity type, i.e., either PFET or NFET. WW and RW are one line WL (Word Line). WB and RB are tied one line BL (Bit Line). This operates to write data when WL becomes a negative voltage. Access transistor 1 then becomes active and write data from BL is charged to the gate of storage transistor 2. During this time, the body-source of storage transistor 2 is reversed-biased. To read data, WL becomes a positive voltage, then the body-source of the storage transistor 2 is forward-biased. According to this embodiment, the voltage level of the BL depends on the gate charge of the storage transistor. At this time, access transistor 1 is off-state.

[0025] The present invention is also useful when implemented utilizing SOI technology. However, any technology which allows bodies of transistors to be independently biased could be utilized with the present invention.

[0026] While the invention has been particularly shown and described with reference to a preferred embodiment, it will be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the spirit and scope of the invention.