Non-volatile dynamic RAM cell.

0001This invention relates to non-volatile semiconductor memory cells and more particularly to cells which utilize a device having a floating gate and an enhanced conduction insulator.

0002A number of circuits have evolved which take advantage of the ability of field effect transistors to store charge and thus serve as memory cells. Such cells may be either dynamic or static in nature. The dynamic cells may employ only a single field effect transistor and the static cells may be arranged in a flip-flop configuration, as is well known. Each of these types of cells may be referred to as volatile cells since information stored in these cells is lost when the power supply voltage applied to the memory is lost or turned off. In instances where stored volatile information must be retained, an alternate power supply, such as a battery system, must be coupled to the memory for use in the event of failure of the main power supply.

0003Known devices capable of providing variable threshold voltages, such as field effect transistors having metal-nitride-oxide-silicon (MNOS) and field effect transistors having a floating gate are capable of storing information in a non-volatile manner for long periods of time. By incorporating such non-volatile devices into memory cells, there has been provided normally operating volatile cells which do not require a backup or alternate power supply for preserving information when power interruption or failure occurs in the main power supply.

0004The non-volatile memory cells which use non-volatile MNOS transistors or devices are capable of retaining for long periods of time information stored volatilely in a cell but these devices require high voltage pulses for writing and erasing the information they are slow and they require rather complex processes for their fabrication. An example of a non-volatile semiconductor memory cell is taught in U. S. Patent 3 676 717, filed November 2, 1970.

0005Known non-volatile memory cells which use convention ally arranged floating gate devices are also capable of preserving for long periods of time information stored volatilely in a cell but these devices likewise have required high voltage pulses for writing and erasing the information, they have been slow and required high currents, approximately one milliampere per device, to write. An example of a known non-volatile semiconductor memory cell having incorporated therein a floating gate is taught in U. S. Patent 4 207 615, filed November 17, 1978.

0006In commonly assigned European patent application 81 106 437.7 there is disclosed non-volatile static memories which include a volatile circuit coupled to a non-volatile device having a floating gate and first and second control gates capacitively coupled to the floating gate with a charge injector structure including enhanced conduction insulators disposed between the floating gate and one of the two control gates. A detailed discussion of enhanced conduction insulators may be found in an article entitled "High Current Injection Into Sio₂ from Si rich Sio₂ Films and Experimental Applications" by D. J. DiMaria and _D. W. Dong, Journal of Applied Physics 51(5), May 1980, pp. 2722-2735, and a basic memory cell which utilizes the dual electron injector structure is taught in an article entitled "Electrically-Alterable Memory Using A Dual Electron Injector Structure" by D. J. DiMaria, K. M. DeMeyer and D. W. Dong, IEEE Electron Device Letters, Vol. EDL-1, No. 9, September 1980, pp. 179-181.

0007Highly dense dynamic random access memory (RAM) cells having only a single storage capacitor and a single switch or transistor are disclosed in commonly assigned U. S. Patents 3 387 286, filed on July 14, 1967, and 3 811 076, filed on January 2, 1973.

0008One device dynamic volatile memory cells having the capability of storing data non-volatilely are known. For example, commonly assigned U. S. Patent 3 916 390 filed December 31, 1974, discloses the use of a dual insulator made of silicon dioxide and silicon nitride for storing information non-volatilely during power failure. Other examples of dynamic cells capable of storing non-volatilely by using MNOS structures include U. S. Patent 4 055 837, filed October 22., 1975, and 4 175 291, filed October 31, 1977. These dynamic cells having non-volatile capability can operate satisfactorily, however, they generally require both negative and positive voltages to switch between volatile and non-volatile modes, larger cell areas, larger voltages for the volatile operating mode or backup memory.

0009The invention as claimed is intended to remedy these drawbacks. It solves the problem of how to design a non-volatile dynamic semiconductor memory which is dense, more versatile and faster than such known memories and which is fabricated by a simple process.

0010The advantages offered by the invention are mainly that it operates at low voltages in the volatile mode and requires less time and low power during data transfer between volatile and non-volatile modes that using an enhanced conduction or silicon-rich insulator, it operates faster than known non-volatile memories, that it can store non-volatilely the previously stored data and also can store any new data in a volatile mode, thereby effectively yielding a double dense memory system, that all volatile data from all cells are preferably transferred to a non-volatile mode in a parallel operation or in one cycle of operation, and that both erasure of non-volatile memory and restoration of volatile data occurs simultaneously for all cells of the system and wherein data is transferred directly from the non-volatile mode to the volatile mode in its true or non-inverted form.

0011In accordance with the teachings of this invention improved non-volatile semiconductor memories are provided which comprise a one device dynamic volatile memory circuit including a storage capacitor with a plate and a storage node coupled to a non-volatile device including a voltage divider having first and second serially-connected capacitors, with a floating gate being disposed at the common point between the first and second capacitors. The plate of the storage capacitor.is connected to a reference voltage source. A control gate is capacitively coupled to the floating gate through the first capacitor which includes a dual charge or electron injector structure. The floating gate is coupled to the storage node through the second capacitor. The capacitance of the first capacitor has a value substantially less than that of the second capacitor. The storage node is formed as an inversion layer.

0012One way of carrying out the invention is described in detail below with reference to drawings which illustrate only one specific embodiment, in which: <ul id="ul0001" list-style="none"><li>Fig. 1 is a circuit diagram of the non-volatile memory cell of the present invention,</li><li>Fig. 2 is a sectional view of an embodiment of the circuit of the present invention illustrated in Fig. 1 of the drawing,</li><li>Fig. 3 illustrates a memory system of the present invention having an array of cells each of which may be of the type illustrated in Fig. 2 of the drawings, and</li><li>Fig. 4 is an illustrative pulse program which may be used to operate the cell of Fig. 2 and the system illustrated in Fig. 3 of the drawings.</li></ul>

0013Referring to Fig. 1 of the drawings in more detail, there is shown the circuit diagram of the non-volatile dynamic memory cell of the present invention. This cell includes a storage capacitor C having a storage node 10 in the form of an inversion layer and a plate 12 and an input/output line I/O which may be interconnected by a switching device 14, preferably a field effect transistor. A voltage Vp, e.g., +5 volts, is applied to terminal P connected to capacitor plate 12. Voltage divider circuit 16 having serially-connected first and second capacitors C1 and C2, respectively, is connected between the storage node 10 and a terminal C having control voltage V_C, e.g., +5 volts, applied thereto, with capacitor C2 having a substantially larger capacitance value than that of capacitor C1. A floating gate, identified by terminal FG, is disposed between the first and second capacitors C1 and C2. A parasitic capacitor Cp which is generally present is indicated as being located between the storage node 10 and a semiconductor substrate indicated at S. The first capacitor C1 includes a dual charge or electron injector structure of the type described in the hereinabove cited IEEE Electron Device Letters article.

0014In the normal operation of the circuit or cell of Fig. 1 of the drawings, the storage capacitor C_s, the input/output line I/O and the switching device 14 function as a volatile dynamic one device memory cell. When power failure is detected, the storage capacitor plate 12 is ramped to ground while simultaneously the voltage V_c at terminal C is pulsed to an appropriate positive voltage, such as +20 volts. As a result of this pulse, a large voltage is developed across the first capacitor C1. If the voltage on the storage node 10 is at 0 volts representing a binary digit "0", the voltage across the capacitor C1 is of sufficient value to charge the floating gate FG in a positive direction. On the other hand, if the voltage on storage node 10 is at +5 volts, representing a binary digit "1", no charge transfer takes place. In this manner, the data from storage node 10 is stored in a non-volatile mode in the floating gate FG. In the case of a stored "0", a filled charge potential well extending under both the floating gate FG and the plate 12 is created. As terminal C is biased more positively, the potential well under the floating gate FG becomes deeper. Simultaneously, the plate 12 is being biased more negatively, transferring the charge from under the plate 12 into the floating gate potential well thereby capacitively coupling the floating gate towards ground and allowing a voltage to be generated across the capacitor C1 causing charge injection, resulting in the floating gate being charged positive.

0015In the case of a stored "1", charge is not stored in the potential well, therefore, as the terminal C is ramped positively, the floating gate FG is capacitively coupled positively, not allowing a sufficiently positive voltage to develop across Cl to inject charge.

0016To retrieve the data from the floating gate FG, the control terminal C has applied thereto a 0 volt bias, plate 12 has a +5 volt bias and the input/output line I/O is at zero volts with the switch 14 closed. If the floating gate FG is positively charged, then there is sufficient voltage across the second capacitor C2 to transfer electrons into the potential well under the floating gate FG and under the plate 12. As a result, the voltage on the storage node 10 assumes a lower value, e.g., +0.0 volts due to charge storage. If no charge resides on the floating gate FG, no significant charge transfer takes place and hence the voltage on the storage node 10 remains at about +5 volts. The cell is next refreshed as a normal volatile cell, and is now available for normal volatile operation with data now at the same state as that which existed prior to power failure.

0017An embodiment of the circuit or cell of the present invention is illustrated in a sectional view in Fig. 2 of the drawings. A P type silicon substrate 18 has an N+ diffusion region 20 forming the input/output or bit/sense line connected to terminal BL, a first potential well PW1 located under floating gate FG and a second potential well PW2 located under the storage capacitor plate 12. The potential wells PW1 and _PW2 form the storage node 10. The floating gate FG is separated from the storage node 10 by a thin insulating layer 22 preferably made of silicon dioxide forming the second capacitor C2. The capacitor Cl is formed by the floating gate FG and a first capacitor electrode 24, connected to control terminal C, along with a dual electron injector structure 26 which includes first and second silicon-rich silicon dioxide layers 28 and 30, respectively, separated by a silicon dioxide layer 32. The storage inversion capacitor C s is formed by the capacitor plate 12, an end of which overlaps the floating gate FG, the floating gate FG and the silicon dioxide layer 22. The switching device or transistor 14 is formed by a gate electrode 34, having an end overlapping the floating gate FG, disposed between the bit/sense line diffusion region 20 and the storage node 10 _' and separated from the surface of the silicon substrate 18 by the silicon dioxide layer 22. The gate electrode 34 is connected to a word line terminal WL.

0018Fig. 3 illustrates a 2 x 2 array of non-volatile memory cells of the type shown in Figs. 1 and 2 of the drawings wherein similar elements in the two figures are identified by the same reference characters. The array includes a first word line WL1 to which first and second cells Al and A2 are connected and a second word line WL2 to which a third cell Bl and fourth cell B2 are connected. First and third cells Al and B1 are connected to a first bit line BL1 and the second and fourth cells A2 and B2 are connected to a second bit line BL2. The first and second word lines WL1 and WL2 are connected to word line decoder and driver circuit 36 which may employ conventional circuitry and the first second bit lines BL1 and BL2 are connected to bit line decoder, precharge and sense amplifier circuits 38 which may also utilize conventional circuitry. The control terminal or line C and the capacitor plate terminal or line P are connected to non-volatile write and erase circuits 40, which may be of any known type.

0019In order to better understand the operation of the non-volatile memory cell illustrated in Figs. 1 and 2 and also of the system illustrated in Fig. 3 of the drawings, reference may be had to an illustrative pulse program indicated in Fig. 4 of the drawings. During normal operation of a non-volatile memory cell, e.g., cell A1, illustrated in Fig. 3 of the drawings, the cell performs as a conventional one device storage circuit coupled to the first bit line BL1 and the first word line WL1 and the voltages therein may be those indicated in Fig. 4 of the drawing between times tl and t2, with constant voltages V_c and Vp being applied to terminals C and P, respectively, having a magnitude of, say, +5 volts. For a stored binary bit "0", the voltage on storage node 10 is, say, zero volts and then the voltage on the floating gate FG will be at approximately 1.0 volt. For a stored binary bit "1", the floating gate is preferably at +3 volts for illustration purposes. As indicated in Fig. 4 of the drawings, applied voltage VB_L on the bit/sense line BL and the voltage V_FG produced on the floating gate FG are shown in solid lines for stored binary "1" bits and in dashed lines for stored binary "0" bits.

0020When power failure is detected, the voltage V_c on the control terminal C is increased from +5.0 volts to, say, +20.0 volts while the voltage V_P at the plate terminal P is reduced from +5.0 volts to, say, 0.0 volts, as indicated in Fig. 4 between times t2 and t3. Thus, any charge stored in the potential well PW2 under the storage plate 12 is transferred to the potential well PW1 under the floating gate FG. As a result, for the cell storing a binary bit "1", the storage node voltage V₁₀ is increased to about +10.0 volts, the floating gate voltage V_FG to about +10.0 volts, and, therefore, the voltage V_FG- V_c across the first capacitor C1 is 10.0 volts, which is insufficient to conduct charging current from floating gate FG to the control terminal C. In case of a cell storing binary bit "0", the voltage across the first capacitor C1 is +13.0 volts which is by design sufficient to significantly alter the charge on the floating gate FG since this voltage is equal to or greater than the turn-on voltage of the dual electron injector structure 26. Thus, the floating gate FG of a cell storing a binary "0" now is charged to, say, +3.0 volts after a short period of time ranging from nanoseconds to a few milliseconds and then the voltages V and Vp are removed from the terminals C and P, respectively.

0021Between times t3 and t4, all applied voltages are at 0.0 volts, and thus for the cell that has a stored binary "1", all voltages are at 0.0 volts at all the nodes. The cell that has stored a binary "0", on the other hand, has +3.0 volts on the floating gate FG, and will, therefore, induce an image charge of +3 volts on the storage node 10, if other terminals are at ground or zero potential. The value of the voltage V₁₀ on the storage node 10 may differ depending upon the value of the parasitic capacitances in the circuit.

0022When power is again turned on or resumed, the memory can be used in a volatile mode with new data while still storing the non-volatile data in the floating gates, if desired, or the non-volatile data can be restored without erasing the stored data.

0023After the resumption of power, to retrieve the previously stored information from the floating gate FG back into the volatile circuit, as indicated between times t4 and t5, +5.0 volts is applied to the storage capacitor plate 12 or terminal P and the control terminal C is at 0 volts. With a voltage of +5 volts on the word line and the bit line at ground, the potential well PW2 under the storage capacitor plate 12 will be filled if a "0" or +3 volts is stored on the floating gate. However, if the floating gate is neutral, indicating the storage of a "1", charge will not flow into the potential well under the plate 12. Between times t5 and t6, the memory is reset by lowering the word line voltage V_WL to zero volts and increasing the bit line voltage V_BL to +5 volts.

0024To refresh the stored information, the voltage V_c on the terminal C is increased to +5 volts as is the word line voltage V_WL. If a "1" digit has been stored, the bit line voltage will remain at +5 volts, however, if a "0" digit has been stored, the voltage V_BL on the bit line BL will decrease, as indicated Fig. 4 between times t6 and t7.

0025Erasure of data from the floating gate FG takes place between times t7 and t8. This is accomplished by pulsing the plate terminal P from +5.0 volts to +20.0 volts so that an effective voltage of +13.0 volts appears across the first capacitor C1 of the binary bit "0" cell and 0.0 volts appears across the first capacitor C1 of the binary bit "1" cell. Since 0.0 volts is insufficient to make the first capacitor or dual electron injector structure C1 conductive, no change occurs for the cell storing the previous binary bit "1". Since +13.0 volts is sufficient to conduct through the first capacitor Cl, the floating gate FG of the cell storing the previous binary "0" will lose the +3.0 volt charge, thereby lowering the voltage.

0026The voltage on the plate terminal P is now returned to +5.0 volts, and the cell is refreshed in its normal manner to restore voltages to the storage nodes 10 having full values representing "1" and "0" binary digits.

0027It can be seen that the signal levels between a binary "1" and "0" are independent of the amount of charge stored on the floating gate, as long as the floating gate. has sufficient charge to invert the underlying potential well PW1 between times t4 and t5 on the pulse program of Fig. 4. The signal level at the sense amplifier is about 1 volt for 5 volts on the plate 12 and a 0.2 transfer ratio between the storage node capacitance and the total bit line capacitance.

0028It should be understood that there are numerous other pulse programs that could be implemented to achieve similar results. For example, rather than applying a positive pulse to the storage capacitor plate P, the control terminal C may be pulsed negatively to erase.

0029As indicated hereinabove, each of the cells A1, A2, Bl and B2 illustrated in the system illustrated in Fig. 3 of the drawings may be operated in accordance with the illustrative pulse program shown in Fig. 4. As also noted hereinabove, cell Al is operated by selecting word line WL1 and bit/sense line BL1, while cell A2 is operated by selecting word line WL1 and bit/sense line BL2, cell B1 is operated by selecting word line WL2 and bit/sense line BL1 and cell B2 is operated by selecting word line WL2 and bit/sense line BL2. The voltages V_c and Vp are produced in non-volatile write and erase circuits 40, which may be provided on-chip or by the system using this memory, and are applied to control terminal C and storage capacitor plate terminal P which are common to all cells Al, A2, Bl and B2.

0030It is possible to arrange the storage capacitor plate line or terminal P and control line or terminal C in the same manner as the word line with their own decoder and driver circuits so that only those cells along the energized word line have the write pulses applied to them. In this manner, cells along other word lines are unaffected. Data retrieval, when power resumes, may be carried out as described hereinabove.

0031The cells may be fabricated by any known techniques but it is preferred that the floating gate FG be made from a first layer of doped polysilicon and that the storage capacitor plate 12, the first capacitor electrode 24 and the gate electrode 34 of Fig. 2 be made from a second layer of doped polysilicon. Also, an N type semiconductor substrate may be used instead of the P type substrate, with the polarity of the voltages being opposite to those used in the illustrated examples mentioned hereinabove.

0032It should be understood that the voltages applied to write and erase data from the floating gate FG with the use of the dual electron injector structures are not necessarily symmetrical. This phenomenon has been reported in the above cited article in IEEE Electron Device Letters.

0033Although the voltages applied to the terminal C are indicated as having a maximum magnitude of +20 volts, it should be understood that voltages of significantly lower values can be used by trading off the insulator thicknesses in capacitors Cl, C2 and C . Small voltage increases across the dual electron injector structure produce sharp drops in writing and erase times because of the exponential nature of the dual electron injector and this insulator structure current-voltage characteristics.

0034It should be noted that in the interest of clarity, an array having only four cells has been illustrated in Fig. 3 of the drawing, however, in practice hundreds of word lines would be used with each word line having hundreds of cells coupled thereto to provide an array of thousands of cells.

0035An improved non-volatile dynamic semiconductor memory has been described hereinabove which does not produce data in inverted form when it is transferred from the non-volatile mode and which can operate normally at the high speeds known in random access memories, yet which will not lose its data when a power failure occurs. The cells may retain their previously stored data in a non-volatile structure while new data is being handled in a volatile storage circuit. Volatile circuit data, from either a field effect transistor or bipolar device circuit, is retained in a non-volatile device with the use of lower voltages dissipating only a very small amount of power and with faster data transfer times between the volatile circuitry and the non-volatile device or structure. Furthermore, the process for making the memory of the present invention is simpler than that used to make, e.g., MNOS devices and the memory of the present invention uses substantially lower writing power levels than is required in floating gate devices written by hot electrons.